Citation
Effect of recombinant erythropoietin mediated changes in anemia on verbal memory in adult dialysis patients

Material Information

Title:
Effect of recombinant erythropoietin mediated changes in anemia on verbal memory in adult dialysis patients
Creator:
Klein, Wayne L
Publication Date:
Language:
English
Physical Description:
viii, 177 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Anemia ( jstor )
Control groups ( jstor )
Diseases ( jstor )
Epics ( jstor )
Fatigue ( jstor )
Kidney dialysis ( jstor )
Memory ( jstor )
Renal dialysis ( jstor )
Repurchase agreement ( jstor )
Statistical significance ( jstor )
Anemia -- therapy ( mesh )
Department of Clinical and Health Psychology thesis Ph.D ( mesh )
Dissertations, Academic -- College of Health Related Professions -- Department of Clinical and Health Psychology -- UF ( mesh )
Erythropoietin ( mesh )
Memory Disorders ( mesh )
Renal Dialysis ( mesh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1991.
Bibliography:
Includes bibliographical references (leaves 162-175).
Additional Physical Form:
Also available online.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Wayne L. Klein.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Wayne L. Klein. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
26253073 ( OCLC )
026859696 ( ALEPH )

Downloads

This item has the following downloads:


Full Text













EFFECT OF RECOMBINANT ERYTHROPOIETIN MEDIATED CHANGES IN
ANEMIA ON VERBAL MEMORY IN ADULT DIALYSIS PATIENTS

















BY

WAYNE L. KLEIN
















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY


UNIVERSITY OF FLORIDA

1991















ACKNOWLEDGEMENTS

An elderly subject stood out for interest in his performance. A man in his early twenties appeared remarkably uninterested in his standing. A third subject participated in the hope of helping others, despite increasing debilitation. All three died prior to completion of the study. Their names, and the names of the 32 other subjects are not listed out of respect for their privacy. However, without their willingness to endure testing despite the discomforts, and in some cases the agonies, of their illness, this study would not have been possible.

The following, undergraduates at the time, assisted with test construction, test validation, data collection and/or data entry: Bill Dorkowsky, Mike Reiter, Sheri Scott, Katherine Cobb, Edward Suarez and Wylene Bhanji. Without their help my wife, Liz, and child, Sarah, would have suffered even worse neglect. Don Mars, M.D. graciously persuaded his patients to participate as controls. The American Psychological Association provided a Dissertation Research Award which partially funded this project.

My committee, beyond help regarding this project,

consists of those who were most important to my graduate ii






iii


experience. Eileen Fennell, my chair, enabled me to meld my interests with the do-able and the acceptable. Without her the going would have been much less pleasant. It was on Russ Bauer's clinic day that I decided to specialize in clinical neuropsychology. Hugh Davis is forever and agreeably etched in my subconscious. Bob Fennell provided research opportunities with import to me well beyond the findings. Tom Fast became a close friend who taught many things.
















TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS . .............. . . . . ii

ABSTRACT . . ....... . . . . ... . . . . vii

CHAPTERS

1 INTRODUCTION: FOCUS AND RATIONALE . . . . . . . 1

2 LITERATURE REVIEW . ............. . 5

Introduction . .......... ..... . . . 5
Overview of Verbal Memory . . . . . . . . . . . 5
Memory Stages and Levels of Processing . . . 6 Automatic and Effortful Memory . . . . . . . 12
Neuropsychological Dysfunction in Adult Renal
Patients. . . . . . .. .......... . 13
General Intelligence: Verbal and Nonverbal 13
Attention, Level of Activation and Executive
Function. .. . . . . . . . . . 14
Verbal Memory .. . . . . . . . . . . . 16
Nonverbal Memory .. . . . . . . . . . . 20
Conclusions Regarding Cognitive Deficits in
Renal Disease . .. ... . . . . . . .. 21
Overview of Anemia . ....... . . . . 24
Prevalence and Etiology .......... . . 24
Function of Hemoglobin and Adaptation to
Anemia . . . . . . . . . . . . . . . . 26
Erythropoietin Therapy and Iron Deficiency 27
Neuropsychology of Anemia ... ....... . . 27
Effects of Anemia . . . . . . . . . . . .. . 29
Possible Mechanisms. . . . . . . . . . . . 32
Effect of Erythropoietin on Neuropsychological
Function . . . . ..... . ..... . . . . 35
Effect of Erythropoietin on Quality of Life . 36 Previous Neuropsychological Studies . ... . . 37 Results of Pilot Study ........ . . 38 Negative Effects . . .. .......... . . 44
Summary and Conclusions ............ . . 46

3 METHODS . ................. . 50

Subjects . . . . . . . . ... . . . . . . . 50
Measures . ............ . . . . . 52

iv






V


California Verbal Learning Test . .... . . 52 Levels of Processing Task . . . . . . . . . . 52 Frequency Estimation Task . . . . . . . . . . 53 Controlled Word Association Test . . . 54
Cognitive-Affective and Physical Behavior
Questionnaire. ... . . . . . . . . 54
Hemoglobin . ..... . . . . . . . . .. 55
Ferritin . . . . . . . . . . . . . . . . . . 55
Schedule of Testing . ...... . . . . . 56
Statistical Methodology . ....... . . 58 Hypotheses . . ......... . . . 58
Main Hypothesis . . ............ . 58
Secondary Hypotheses ...... . . . . . 59

4 RESULTS. ..... .. . ...... . . . . . . 61

Overview of Analyses . . . . . . . . . . . . . 61
Statistical Assumptions . .......... . 64
Descriptive Statistics . .... ..... .. 69
Age .......... ........ . .. 69
Education ............. . .. . . 69
Sex . . . . . . . . . . . . . . . . . .. . 69
Treatment Modality ........... . . . . 69
Attrition . .. . .. . . . ........ . 70
Time Intervals Between Assessments . . . . . 71 Hemoglobin . ............. . . . . 72
Ferritin . . . . . . . . . . . . . . . . . . 75
Baseline Neuropsychological Functioning ... 79
Test of Main Hypothesis . ........... 82
Post Hoc Analyses .... . . . . . . . 83
Learning Curve . . . . . . . . . . . . . . 83
Exploratory Analyses . .. . . . . . . . . . 85
Sum of CVLT Trials Four and Five . ... . 88 Short Term Memory . . . . . . . . . . . . . 90
Verbal Fluency . ........... . . 90
Secondary Hypotheses . ...... .... . . 91
Levels of Processing ............ 91
Estimation of Frequency of Occurrence . . .. 91 Quality of Life . ..... . . . . . . . . 91
Test Reliability . ..... . ......... 93
CVLT Alternate Form Reliability . . . . . . . 93 Levels of Processing Reliability . . . ... 98 Frequency of Occurrence Reliability . . ... 98
Physical Behavior Self-Report Test-Retest
Reliability . . . . . . . . . . 98
Cognitive-Affective Behavior Self-Report
Test-Retest Reliability . . . . . . . . . 100
Post Hoc Analysis of Blood Urea Nitrogen . . 100 Summary .. . . ................ . 101

5 DISCUSSION . . . . . ...... . . . . . . 104

Overview . . . . . . ......... . . 104
Conclusion . . . . . . . . . . . . . . . . . . 121






vi


APPENDICES

A DEVELOPMENT AND VALIDATION OF FORM III OF
THE CALIFORNIA VERBAL LEARNING TEST . . . . 124

B LEVELS OF PROCESSING VERSIONS II AND III . . . 131 C FREQUENCY ESTIMATION TASK . . . . . . . . . .139

D COGNITIVE-AFFECTIVE AND PHYSICAL BEHAVIOR
QUESTIONNAIRE . . . . . . . . . . . . . . . 141

E COMPARISON OF PARAMETRIC AND NONPARAMETRIC
TESTS OF BASELINE MEASURES BETWEEN GROUPS . 144

F NORMALITY AND HOMOGENEITY OF VARIANCE . . . 146 G GROUP MEANS AND STANDARD DEVIATIONS . . . . . 149 H CHANGE SCORES ...... . . . . . . . . . . 155

I RAW DATA BY SUBJECT . ........ .. . . . 158

BIBLIOGRAPHY .......... . . . . . . . . . . . 162

BIOGRAPHICAL SKETCH . ............... . .175















Abstract of Dissertation Presented to the Graduate School of
the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


EFFECT OF RECOMBINANT ERYTHROPOIETIN MEDIATED CHANGES IN
ANEMIA ON VERBAL MEMORY IN ADULT DIALYSIS PATIENTS By

Wayne L. Klein

August 1991

Chairperson: Eileen B. Fennell Major Department: Clinical and Health Psychology

Cognitive dysfunction has been documented in many

chronic anemic conditions; however, the putative effects of anemia have never been disassociated from the cause of the anemia. Neuropsychological dysfunction of poorly understood origin and severe chronic anemia are nearly universal consequences of chronic renal failure. A small pilot study demonstrated a positive relationship between a recombinant erythropoietin (rEPO) mediated mean rise of 5.46 g/dl of hemoglobin (Hg) and the Sum of trials 1-5 on the California Verbal Learning Test (Sum CVLT). Sum CVLT performance was hypothesized to be enhanced by increased effortful processing secondary to reduced fatigue.

A rEPO treatment group (n=17) and a control group (n=18) underwent three assessments with three alternate forms of the CVLT, one of which was developed for this


vii






viii


study. Also administered were a survey of cognitiveaffective and physical behavior, a verbal fluency task and three alternate forms of two measures likely to be sensitive to hypoxia but not fatigue, the levels of processing and frequency estimation tasks. All subjects were on maintenance dialysis. Mean treatment group Hg at the first assessment was 8.46 g/dl (sd=1.42) and 9.44 g/dl (sd=1.92) in the controls. The rise of 1.34 g/dl Hg between groups in the 54 day mean interval between the first and second assessments was statistically significant (F=7.42, p=.0102), but of doubtful physiological significance. In the 90 days between the second and third assessments Hg was unchanged. The relationship between rise in Hg and Sum CVLT failed to reach statistical significance (F=3.25, p=.0805). Post hoc reclassification of subjects based on direction of Hg change unexpectedly revealed an apparent inverse relationship between Hg and Sum CVLT. Subjects, primarily controls, exhibiting a drop in Hg between the first and second assessments performed below expectations at the first assessment (F=7.82, p=.0027). Whether this was spurious or due to the operation of an unknown factor remains speculative. Hemoglobin related changes in self-reported cognitive-affective and physical behavior were not observed. Current understanding of the relationship between anemia and neuropsychological dysfunction appears insufficient to inform decisions regarding rEPO treatment.















CHAPTER 1

INTRODUCTION: FOCUS AND RATIONALE

The fourth leading health problem in the developed world, renal disease has been estimated to affect approximately 8 million people in the U.S. (Williams, 1985). Uremia, the life threatening condition resulting from renal failure, has been known since ancient times. A variety of symptoms have been associated with uremia, the majority of which have been reported to be secondary to nervous system dysfunction (Ginn et al., 1975). Memory dysfunction has been among the neuropsychological deficits commonly noted. Despite considerable effort, attempts to identify the factors responsible for these deficits have achieved little success (Powell et al., 1986; Schreiner, 1975; Trompeter, Polinsky, Andreoli, & Fennell, 1986).

Chronic anemia has been almost universally observed in chronic renal failure. The primary cause has been generally accepted to be insufficient production of erythropoietin (EPO), a erythropoiesis stimulating factor produced by the kidneys. Recent studies have documented the dramatic effects of recombinant erythropoietin (rEPO) on many physical symptoms until recently attributed to uremia, but now demonstrated to have been at least partially caused by anemia (Nissenson, 1989). One rationale for the present

1






2


study was to test the possibility that a portion of the neuropsychological dysfunction that has been attributed to uremia is actually secondary to anemia.

A further rationale for the present study was to

address the broader question of whether cognitive deficits are directly associated with chronic anemia. Severe acute anemia produces serious symptoms of cerebral dysfunction, but symptoms subside over time as adaptation occurs. Although neuropsychological deficits have been well documented in conditions of chronic anemia, the question of whether the neuropsychological symptoms are caused by the anemia has never been adequately addressed. The neuropsychological study of anemia has been hampered by difficulties in parceling out the effects of rise in hemoglobin (Hg) from the effects of the agent causing the rise. Chronic anemia is the final common result of numerous causative factors including deficiencies of iron, folic acid and cobalamin as well as conditions such as sickle cell disease, all of which have been demonstrated to directly affect the brain. In contrast, rEPO has been generally considered to have no effect on the nervous system. This putative disassociation between neurological and hematological effects suggested that rEPO could potentially be a valuable tool in the study of the neuropsychological effects of chronic anemia.

Pilot work sampled a spectrum of cognitive functioning before and during rEPO treatment. That work, although






3


suffering from small sample size, suggested that rise in Hg was associated with improved long term memory (LTM) as measured by performance on a multitrial verbal list learning task (Klein et al., 1989). The present study sought to confirm and extend these very preliminary findings. An additional purpose of the present study was to generate exploratory data addressing the nature of the mechanisms underlying the putative improvement in LTM. In the pilot study, despite sampling a variety of cognitive functions, the only task that showed significant improvement in performance was the task most sensitive to success in actively organizing material into a meaningful structure. Thus, it was hypothesized that the apparent improvement was mediated by increased cognitive processing, possibly secondary to reduced fatigue.

In the present study, paradigms from human memory

research were used in an attempt to tease out the level at which putative change occurred. Improved automatic learning and increased ability to benefit from semantic processing, two tasks relatively unaffected by functional factors, would have implicated physiological mechanisms affecting the neural substrate of memory. Lack of improvement on these tasks, in conjunction with improved effortful verbal LTM performance, would have lent support to the hypothesis that a functional mechanism such as increased cognitive processing, possibly secondary to reduced fatigue, resulted in the apparent improvement in LTM in the pilot study.






4


Chronic anemia has been reported to have far reaching effects on the organism. In loose association with the severity of the anemia, symptoms may include fatigue, hypoxia, hypocapnea, insomnia, enhanced cardiac output and increased sympathetic activation, reduced appetite, altered blood chemistry, modified lifestyle including the inability to hold gainful employment and, probably, decreased selfesteem. Therefore, determination of the relative contributions of possible underlying mechanisms was considered impossible in a single study of this scope. However, it was hoped that this study would extend our understanding of the neuropsychology of chronic anemia and the clinical significance of chronic anemia with regard to the etiology of the cognitive deficits observed in renal disease.















CHAPTER 2

LITERATURE REVIEW

Introduction

Investigation of the neuropsychology of uremic anemia is necessarily multidisciplinary. The logic underlying the rationale and design of this study is based on information derived from five somewhat distinct areas of investigation. The areas are 1) human memory, 2) clinical studies of neuropsychological function in adults suffering from renal disease, especially with regard to verbal memory, 3) the physiology of anemia, 4) the neuropsychology of anemia and 5) the effects of rEPO mediated improvement in anemia in adult renal patients on neuropsychological function. Each of these will be briefly reviewed in this chapter.

Overview of Verbal Memory

This section briefly overviews current thinking with regard to memory as it relates to the clinical assessment and interpretation of memory function. The application of memory paradigms to the study of the neuropsychology of anemia is discussed. Mounting neurochemical, functional and anatomical evidence suggests that memory may be classified into procedural memory (i.e., motor memory or skill learning) and declarative memory for facts (Nissen, Knopman, & Schacter, 1987). Memory may, to some extent, be divided

5






6


into automatic and the more commonly studied effortful, deliberate or consciously encoded memory (Hasher & Zacks, 1984; Newman, Weingartner, Smallberg, & Calne, 1984; Sanders, Gonzalez, Murphy, Liddle, & Vitina, 1987). Automatic declarative memory.may be further differentiated based on the modality of encoding and retrieval, with the most studied modalities being verbal and visual. Verbal memory may also be partitioned into semantic and episodic memory, the former being dissociated from, and the later bound within, the context of learning (Newman, Weingartner, Smallberg, & Calne, 1984).

Memory Stages and Levels of Processing

Most paradigms appear to demonstrate the division of effortful declarative verbal memory into relatively clearcut temporal categories or stages. The cognitive, neuropsychological and neuroscience traditions have posited long and short term memory stages. The definition of memory stages is dependent on the level of analysis (Squire, 1987). Cognitive science has classified auditory memory of a duration exceeding about 15 seconds as long term memory (LTM) while neuroscience considers this squarely within the domain of short term memory (STM). These differences reflect the criteria by which stages have been determined within the two traditions.

Within the cognitive tradition a succession of models (e.g., capacity, multistore and levels of processing) has stimulated research. In attempting to map memory function






7


onto biological structures and processes, the neuroscience perspective has focused both on the nature of the memory trace and the physical location of memory in the brain. This brief overview will focus primarily on a behavioral (i.e., cognitive and neuropsychological) level of analysis as it relates to neuropsychological assessment.

As conceptualized in the multistore model, memory is a complex series of distinct stages through which information is successively processed (Atkinson & Shiffrin, 1968, 1971). These stages differ along several dimensions including capacity, duration and the nature of the attendant psychological processes. The three generally recognized stages are ultrashort or modality-specific sensory memory (e.g., echoic and iconic memory), immediate or STM and LTM.

Sperling (1960) demonstrated that iconic (i.e., visual sensory) memory decays after about one second. Echoic memory may take several seconds to decay (Darwin, Turvey, & Crowder, 1972). Sensory memory capacity appears to be around nine items (Sperling, 1960). The duration of STM when rehearsal (i.e., replenishment) is prevented has been shown to be about 15 seconds (Peterson & Peterson, 1959). Thus, the repetition of a list of orally presented digits or words utilizes both echoic memory and STM. However, in the case of oral digit span, as traditionally administered, the limiting factor is likely to be attentional.

Oral digit span is the most commonly used measure of

immediate memory span; however, it is frequently insensitive






8



to STM recall deficits (Lezak, 1983). In a series of factor analyses, Wechsler (1987) found digit span to consistently load primarily on attention-concentration rather than on a memory factor. Unpublished data communicated by T. White (June 12, 1990) revealed that for 182 patients seen in a neuropsychology clinic, there was an insignificant correlation (r= .123, R=.098) between digits forward and trial one of the CVLT, a measure of STM recall.

The capacity of STM has been shown to be approximately

7 �2 chunks of information (Miller, 1956). Unfortunately the notion of chunks has not been clearly defined. Recall significantly exceeding the accepted capacity of 7 �2 has been attributed either to the participation of LTM or efficient organization of bits of information into larger chunks. There are no known limits on the duration or capacity of LTM.

An inherent feature of multistore models has been the necessity of elucidating the processes through which information is progressively transferred through the memory storage areas. The mechanism underlying transfer of information from sensory memory to STM has been attributed to attentional processes associated with pattern (i.e., meaning) recognition (Moray, 1959). In contrast, Shiffrin (1975) has proposed that perceptual stimuli are automatically encoded and passed from sensory to STM. From a multistore perspective, transfer from STM to LTM has been viewed as a function of active attention or rehearsal






9


(Rundus, 1971). This model explains the observation that, on a supraspan task, a higher percentage of the first (primacy) and last (recency) words tend to be recalled. The primacy effect is believed to be due to encoding in LTM. The recency effect is hypothesized to be a manifestation of STM. Thus, immediate recall on a supra-span task may be examined for primacy and recency effects to provide indications of the integrity of STM and LTM.

Besides dissimilarities in capacity and duration, there appear to be differences in the cognitive processes requisite to the maintenance of STM and LTM. In general, most studies and clinical observations have suggested that phonological or maintenance rehearsal (i.e., mere parroting) usually serves only to maintain information in STM, whereas more complex, elaborative or semantic processing facilitates transfer to LTM (Craik & Lockhart, 1972). Transfer to LTM has been thought to usually require more active attending and processing.

Demonstration of the differential effects of type or

level of processing led to the proposal that the observation of apparently distinct memory stores could be explained in terms of levels of processing within one memory store (Craik & Lockhart, 1972). The apparently limited storage capacity of STM has been reinterpreted by the levels of processing model as limited processing capacity. The apparent distinction between STM and LTM has been explained by the levels of processing model as partially a function of depth






10


of processing, with semantic processing occurring at greater depth than phonemic processing. However, research in other areas has forced revisions in the original levels of processing models (Cermak, 1982).

To account for studies demonstrating elaborateness of processing to be more salient than type of processing (e.g., semantic vs. phonological), the levels of processing model has been modified to emphasize extent rather than depth of encoding (Craik & Tulving, 1975). Confounding all models, other work has shown that shallowly processed information may at times interfere with and even outlast more deeply processed material (Cermak, 1982). In addition, phonemic processing may result in recall superior to semantic processing when retrieval is phonemic, that is, when processing and retrieval are within the same domain (Morris, Bransford, & Franks, 1977).

Data in STM is easily lost through interference and

decay. In contrast, reflecting the enduring nature of LTM, failures of LTM have usually been considered to be produced by retrieval deficits. The continued presence of the memory may be assessed with recognition tasks. Recall tasks require self-generation of cues. In contrast, in recognition tasks cuing is externally provided. Thus, recognition in the absence of spontaneous recall may be explained by weak or decayed memory traces, failure to generate adequate self-cuing or memory search strategies or






11



insufficient processing during acquisition resulting in inadequate encoding.

Based on the cognitive literature, it may be concluded that LTM is powerfully affected by the manner and extent to which the learner processes the information. If retrieval is based on semantic cuing, then semantic processing will be the most efficient route for encoding. The more elaborate the processing, the higher the probability of effective retrieval. Conditions reducing cognitive processing may be expected to impair LTM. Cognitive processing is likely to be negatively affected by factors including underarousal, overarousal, fatigue, attentional deficits, distracting stimuli, lack of interest and depression (Cohen, Weingartner, Smallberg, Pickar, & Murphy, 1982).

Under some circumstances memory deficits may be

categorized as either primarily functional or organic. Functional factors, those affecting cognitive processing, may either reduce effort expended or increase effort required. Functional memory deficits may, therefore, be considered a product of insufficient or misdirected effort.

Several methods exist by which, under certain

circumstances, various types of memory dysfunction may sometimes be differentiated. By controlling the duration and nature of processing (i.e., semantic, phonological or orthographic), the levels of processing paradigm provides a means of determining whether LTM deficits are characterized by a failure to benefit from semantic processing. Failure






12


to benefit from semantic processing has been attributed to encoding deficits secondary to organic dysfunction of memory structures (e.g., Korsakoff's). Automatic and Effortful Memory

Automatic memory is demonstrated by recall in the absence of effortful encoding. Automatic memory is unconscious and little affected by degree of effort (Jonides & Naveh-Benjamin, 1987). Deficits in automatic memory processes are suggestive of dysfunction at the level of organic memory structures. Reductions in capacity such as illness and, presumably, fatigue, have limited effect on automatic memory (Hasher & Zacks, 1984). The frequency of occurrence task provides a means of assessing automatic memory function. Normal performance on a frequency of occurrence task in conjunction with reduced performance on an effortful task requiring effort for optimal performance would suggest disruption at the level of effortful processing. This pattern of deficits could be the result of frontal lobe dysfunction, poor motivation or failure to employ an effective strategy. One caveat, the use of automatic memory measures to assess the integrity of the neural substrates of effortful memory is predicated on the assumption, not accepted by all, that effortful and automatic memory are largely dependent on identical neuronal systems (Roediger, 1990).





13


Neuropsychological Dysfunction in Adult Renal Patients

This section provides a cursory overview of the general pattern of deficits seen in renal disease and discusses verbal memory within the context of broader findings. General Intelligence: Verbal and Nonverbal

In the absence of confounds, general verbal

intellectual ability has been reported to be relatively spared in adult renal disease (Blatt & Tsushima, 1966; Comty, Leonard, & Shapiro, 1974; English et al., 1978; Fishman & Schneider, 1972; Freeman, Sherrard, Carlsyn, & Paige, 1980; Ryan, Souheaver, & DeWolfe, 1980; Schupak, Sullivan, & Lee, 1967; Trieschmann & Sand, 1971). In a 1982 review, Osberg, Meares, McKee and Burnett noted that in every reported study, verbal IQ scores surpassed performance scores by 5 to 14 points. The verbal-nonverbal discrepancy is partly an artifact of the time constraints on many nonverbal but not verbal intelligence subtests; however, visuospatial and visuomotor deficits have also been widely observed on untimed tasks (Mings, 1987).

The relative sparing of language function is consistent with the general pattern of deficits often seen in diffuse brain dysfunction. Speculations regarding the cause of this verbal-nonverbal disassociation have included: the nondominant hemisphere is more vulnerable to assault, overlearned verbal abilities are more redundantly represented and crystallized intellectual functions tax cognitive capacity less than more novel visuospatial tasks.





14


Interestingly, McDaniel's (1971) data on visual discrimination performance suggested impairment not primarily as a result of difficulties with visual-motor integration, but secondary to interference with cognitive processing. The reported visuospatial motor dysfunctions may be partially mediated by attentional deficits (Fennell, Fennell, Mings, & Morris, 1986; McDaniel, 1971). Attention. Level of Activation and Executive Function

Several authors have suggested the presence of global attention deficits in uremia (Marshall, 1979; Stewart & Stewart, 1979). Data reported by Trieschmann and Sand (1971) on 83 subjects not treated with dialysis demonstrated reduced performance on two measures loading on simple attention and concentration, the Digit Span and Arithmetic subtests of the Wechsler Adult Intelligence Scale (WAIS). Grouping subjects based on severity of illness revealed Digit Span to be the lowest subscale score in the more severely ill group.

In a dialized population with normal premorbid verbal IQ based on vocabulary knowledge, English, Savage, Britton, Ward and Kerr (1978) found reduced performance on WAIS subtests most heavily loading on attention (i.e., Arithmetic, Digit Span, Digit Symbol). In a pediatric population, Digit Span was reported to be significantly reduced regardless of treatment modality. Indeed, in a pediatric population, excluding the renal transplant group, Digit Span discriminated between renal and control groups





15


better than any other measure in the study (Fennell, et al.,1987).

In contrast, Souheaver, Ryan and DeWolfe (1982) found spared auditory attention skills, but on a different task, the Seashore Rhythm test. In the Heilman, Moyer, Melendez, Schwartz and Miller (1975) study, simple attentionconcentration as measured by Digit Span was within normal limits (mean subscale score of 10.04 �2.18) and consistent with verbal intelligence as measured by the Vocabulary subtest (mean 10.00). However, attention was significantly lower in renal patients (n=24) than in controls (n=12). Overall, it may be concluded that attentional deficits are common in uremia.

Uremia appears to depress activation and arousal

(Trompeter, Polinsky, Andreoli, & Fennell, 1986). This underarousal may interfere with attention and increase distractibility (Heilman, Moyer, Melendez, Schwartz, & Miller, 1975). However, reduced level of activation is unlikely to be the only mechanism involved. Assessments at varying time intervals after hemodialysis sessions revealed no change in Digit Span performance, despite changes in speed and reaction time (Lewis, O'Neill, Dustman & Beck, 1980).

Renal disease also appears to negatively affect more complex attentional and executive functioning such as the ability to rapidly generate and/or shift sets. Performance on Trails B, a task requiring maintenance of and alternation





16


between two sequences, speeded visual search and simple visual motor coordination, has been reported to be reduced, at times severely, in renal disease (Heilman, Moyer, Melendez, Schwartz, & Miller, 1975; Ratner, Adams, Levin, & Rourke, 1983; Teschan et al., 1974). Ginn (1975) reported a correlation of .82 between Trails and severity of renal failure as measured by serum creatinine for subjects not on dialysis. Similar results were reported by Teschan and colleagues (1974). Reduced performance on tasks demanding sustained vigilance, such as continuous performance tasks, has been consistently reported (Ginn, 1975; Osberg, Meares, McKee, & Burnett, 1982). Ginn and colleagues (1978) demonstrated worsening of performance on a vigilance task in

8 out of 10 subjects following a reduction in hemodialysis.

Contributions to executive system dysfunction may

include deficits secondary to reduced level of activation, diminished attentional capacity, susceptibility of the frontal-subcortical neural substrate to metabolic abnormalities and dysfunction stemming from a generalized difficulty with tasks demanding fluid as opposed to crystalized intelligence.

Verbal Memory

Reduced vigilance, underarousal, increased

distractibility, increased response latencies and reduced cognitive processing may contribute to global memory deficits. Work in a pediatric renal population has revealed a correlation between learning on a multitrial supra-span





17


verbal task requiring the learning of a list (e.g., Buschke), susceptibility to distraction (i.e., Auditory Consonant Trigrams) and reduced sustained vigilance (Fennell, Fennell, Mings, & Morris, 1986). This led to the suggestion that reduced level of arousal may mediate STM impairment in renal disease (Fennell et al., 1990a). Because they place greater demands on sustained attention, multitrial memory tasks may be more sensitive to uremia.

Overall, initiation of hemodialysis has been reported to improve memory function (Osberg, Meares, McKee, & Burnett, 1982). In adults, the consequences of uremia are largely reversible and appear to be only minimally cumulative in the absence of confounds such as aluminum toxicity, dietary noncompliance, uncontrolled hypertension or diabetes. Comparing dialysis patients and controls, Hart, Pederson, Czerwinski and Adams (1983) found relatively little difference in memory as measured by the Wechsler Memory Scale (WMS). In addition, no significant correlations were found between years of dialysis treatment and memory performance.

Gilli and DeBastiani (1983) reported (n=54) a mild

relationship between duration of hemodialysis and decrements in verbal intellectual ability (WAIS) and memory (WMS). The WMS Memory Quotient, a composite measure of primarily short term verbal and nonverbal memory, was initially above expectations based on WAIS performance. For the 21 hemodialysis subjects who were retested, a minimum of 12





18



months later the WMS Memory Quotient had dropped to a level only slightly above expectations for IQ. The authors interpreted these results as suggesting a decline in memory function in association with time on hemodialysis, perhaps related to elevated parathyroid hormone levels. However, fewer than half of the subjects were retested, and they exhibited initial Memory Quotient scores above the average for the entire subject pool and, as previously noted, above expectations for IQ. The apparent decline in memory may have been an example of regression towards the mean.

Ginn (1975) reported a temporary improvement in verbal LTM on a recognition task the day after hemodialysis, regardless of the level of performance prior to dialysis. A correlation of .687 (p<.001) was reported between latency to response on a single trial word-recognition task and severity of uremia (Ginn et al., 1975). However, increased response latency does not necessarily indicate retrieval difficulties, especially in a population suspected of generalized slowing.

Heilman, Moyer, Melendez, Schwartz and Miller (1975)

found significant verbal STM deficits on the Logical Memory subtest of the WMS. In contrast, loss between immediate and delay was comparable to controls. Ginn and colleagues (1975) reported a significant negative correlation (R<.001) between response latency on a verbal recognition memory task and uremia as measured by serum creatinine. In nondialized uremic subjects (n=23) participating in a longitudinal






19


study, Hagberg (1974) reported verbal STM as measured by paired associate learning to be significantly lower than that predicted by general verbal ability in 23 nondialized uremic subjects. Six months after initiation of hemodialysis, paired associate learning performance had significantly improved in the 21 subjects retested and approached expected levels relative to intelligence. This study used alternate test forms but failed to use a control group to control for possible effects of repeat testing.

Assessments of six subjects at varying time intervals following hemodialysis revealed no changes in STM as measured by a paired associate learning task, despite changes in performance on timed tasks (Lewis, O'Neill, Dustman, & Beck, 1980). In contrast, on a recognition task, STM was shown to improve 24 hours after and deteriorate two days following dialysis (Teschan et al., 1974). Other work by this group has suggested improvement in verbal STM following transplantation, but this has not reached statistical significance (Teschan, Ginn, Bourne, & Ward, 1976).

Based largely on work with a pediatric population, it has been suggested that STM and sensory/motor function are more severely impacted than are more complex cognitive abilities or LTM (Crittenden, Holliday, Piel, & Potter 1985). However, this report failed to provide sufficient data to support this suggestion. It appears likely that the measure of STM was Digit Span from the WAIS-R, which is more






20


of an attentional measure. Other work generally supports the notion of a differential susceptibility for attentional and visuospatial constructional functioning. Also in the child literature, STM and LTM deficits have been reported on multitrial supra-span memory tasks, with impairment in normal developmental improvement in memory (Fennell, et al., 1990a).

Nonverbal Memory

Visual memory as measured by the Benton Visual

Retention Test was impaired in nondialized renal subjects (n=12), but normalized following 12 months of maintenance hemodialysis, although the change was not significant (Hagberg, 1974). In a related study, 20 adults, chronically dialized for a mean of 39.7 months, were followed throughout the hemodialysis cycle. Performance on the Benton Visual Retention Test was mildly impaired on three separate administrations, despite unimpaired ability to copy the drawings (Ratner, Adams, Levin, & Rourke, 1983).

Facial recognition memory as measured by Milner Faces was reported to be unimpaired in nondialized renal patients (Heilman, Moyer, Melendez, Schwartz, & Miller, 1975). Memory For Designs was shown by Hagberg (1974) to be within the normal range for 23 nondialized uremic subjects. Illustrating the possible value of longitudinal work, as well as the need for alternate forms, readministration of the same test form to 21 of the original subjects, following






21


6 months of hemodialysis, revealed a very slight nonsignificant improvement.

In hemodialysis subjects (n=29), impairment in

visuospatial memory as measured by the Block Design Learning Test, in a study lacking a control group, failed to reach statistical significance. However, duration of dialysis did correlate with poor performance (English, Savage, Britton, Ward, & Kerr, 1978). However, it is unclear whether the putative deficit was related to nonverbal memory or to visuospatial constructional ability. In addition, factors such as hypertension, diabetes and age may have contributed to the apparent negative relationship between duration of dialysis and nonverbal memory performance.

Ziestat, Logue and McCarty (1980) reported a

significant correlation between years on hemodialysis and both short term (r=-.39) and long term (r=-.38) visual memory as measured by the WMS. However, again, years on dialysis may have been confounded by factors such as etiology of renal failure. In contrast, verbal memory showed no relationship with years on hemodialysis in this study. The apparent decline in visual memory may have been secondary to the well researched decline in visuospatial processing (Mings, 1987).

Conclusions Regardina Conitive Deficits in Renal Disease

Renal disease differentially affects cognitive

functioning in a manner grossly consistent with that seen in many systemic conditions. The degree of cognitive





22


impairment correlates with severity of uremia, is at times dramatically improved by initiation of dialysis and may be nearly reversed by successful transplantation (Fennell, Rasbury, Fennell, & Morris 1984; Ratner, Adams, Levin, & Rourke, 1983). Adults on maintenance dialysis frequently exhibit deficits in attention, vigilance, reaction time, ability to rapidly generate and/or shift sets, visuospatial ability, visual memory and verbal memory. Weaker evidence suggests variability correlating with the phase of the dialysis cycle, a possible advantage for peritoneal dialysis over hemodialysis and decreased performance in association with time on hemodialysis (Fennell, Fennell, Mings, & Morris, 1986; Osberg, Meares, McKee, & Burnett, 1982). Higher Hg levels in peritoneal dialysis compared to hemodialysis and the possibility of increasing anemia in conjunction with time on hemodialysis leave open the possibility that level of anemia may play a role in the latter two observations. Restoration of near normal cognitive function following transplantation in well dialyzed subjects may also, to an extent yet to be determined, be mediated by the amelioration of anemia.

There are a number of problems with this literature. For example, small sample size, absence of alternate test forms, lack of relevant control groups, and inadequate statistical procedures (Osberg, Meares, McKee, & Burnett, 1982). Additional contributions to discrepancies in the literature include high variance in subject characteristics





23


including differences in medical management, age, age of onset and etiology of renal disease. Many of the conditions such as diabetes, hypertension and cardiovascular disease contributing to or associated with renal failure have neuropsychological consequences in their own right. Discrepant findings in hemodialysis patients may to some extent be a function of the time of assessment relative to dialysis, the adequacy of dialysis and the degree of electrolyte disequilibrium following dialysis. Most of the early studies sampled neuropsychological function at one point in time. There have been very few longitudinal studies.

Progress in the treatment of renal disease has been

charted in the literature on the neuropsychology of uremia. Prior to the availability of dialysis, the neuropsychological consequences of renal failure often included stupor and coma (Arieff, Guisado, & Massry, 1975; Tyler, 1968). Prior to the discovery and acceptance of the neurotoxicity of elevated blood aluminum levels in renal patients, dialysis encephalopathy affected a significant percentage of patients (Sprague et al., 1988). Improvements in the methodology of dialysis as well as other aspects of medical management continue to improve the physiological conditions under which uremic nervous systems function. Concurrently, these advances have also resulted in longer durations of exposure to uremia as well as an aging renal population. Thus, discrepancies in the literature may to





24


some extent be attributed to the differential effects of the passage of time on technology and on individuals.

Overview of Anemia

This section will provide a brief overview of the

significance and etiology of anemia, adaptive mechanisms and the treatment of uremic anemia. Prevalence and Etiology

The world-wide prevalence of anemia has been estimated to be 40% in children, 35% in adult females and 20% in adult males (Lozoff, 1989). The defining feature of anemia is a reduction in blood Hg levels. Anemia is the final common pathway for a large number of diseases, conditions resulting in blood loss and deficiency states, all of which result in decreased Hg in the blood (Bunn, 1980a). The most common form of anemia, iron deficiency anemia has been reported to have a prevalence in the U.S. of 20% in adult women of childbearing years, 50% in pregnant women and 3% in adult males (Lee, Wintrobe, & Bunn, 1980). Probably the second most common type, the anemia of chronic disorders is the mild to moderate anemia frequently associated with chronic inflammatory syndromes and infectious and neoplastic diseases (Wintrobe et al., 1981). Anemia is a very common condition.

Anemia almost invariably accompanies chronic renal failure (Desforges, 1975; Erslev, 1975). Uremic anemia often significantly reduces quality of life, despite adequate dialysis. Prior to the development of rEPO,





25



approximately 25% of dialysis patients suffered from anemia severe enough to require intermittent or regular red-cell transfusions (Eschbach, Egrie, Downing, Browne, & Adamson, 1987).

Uremic anemia results from several interactive

processes. Shortened red blood cell survival is a frequent manifestation of uremia (Jacob, Eaton, & Yawata, 1975). Although controversial, some evidence suggests that uremic toxins inhibit heme synthesis and erythroid progenitor cell formation (Erslev, 1975; Jacob, Eaton, & Yawata, 1975). Blood loss due to platelet dysfunction and as a result of hemodialysis are implicated (Desforges, 1975). The increased level of potentially toxic "middle molecules" in hemodialysis compared to peritoneal dialysis may also contribute to anemia in hemodialysis. Iron deficiency, folate deficiency, B12 deficiency, aluminum-induced microcytosis and hypersplenism have been cited (Paganini, 1989).

One study has demonstrated a significant increase in hematocrit and Hg levels following 12 months of endurance exercise training, which suggests that sedentary life style, a frequent consequence of end-stage renal disease, may contribute to the anemia (Goldberg et al., 1986). However, inadequate erythropoietin (EPO) production relative to the degree of anemia is the major cause of uremic anemia. Thus, uremic anemia is primarily due to an endocrine deficiency state, correctable by rEPO replacement therapy (Chandra,





26


McVicar, & Clemons, 1988). Intravenously administered rEPO has been demonstrated to restore Hg levels to normal in many renal patients (Eschbach, Egrie, Downing, Browne, & Adamson, 1987; Winearls et al., 1986). Function of Hemoqlobin and Adaptation to Anemia

Under normal atmospheric and physiologic conditions,

more than 98% of the total oxygen carried by arterial blood is bound to Hg, with the remainder dissolved in the aqueous portion of blood (P02). To cross cell membranes, oxygen must first disassociate from Hg and dissolve directly in the blood. The function of Hg may be considered to be the maintenance of normal P02 by providing a large reservoir of available oxygen. Thus, in anemia the reservoir of oxygen is reduced.

Anemic organisms may maintain normal tissue oxygenation through a variety of adaptive mechanisms. Blood flow is deviated to more vital areas. The affinity of Hg for oxygen decreases. When Hg falls below 7.5 g/dl, an increase in both heart rate and stroke volume results in a significant rise in resting cardiac output (Bunn, 1980b) mediated by increased sympathetic activation. Although PCO2 is a more potent controller of ventilation than is P02, low P02 resulting from severe anemia may stimulate peripheral chemoreceptors on the carotid bodies resulting in increased ventilation. However, the resulting reduction in PCO2 and developing alkalosis are likely to attenuate this response (Vander, Sherman, & Luciano, 1980).






27


Hypoxia stimulates production of EPO primarily by the kidneys. As the major regulator of erythrocyte production, EPO stimulates erythropoiesis, the process in which erythrocytes develop from stem cells in the bone marrow, manufacture Hg from amino acids and body iron stores and then enter the blood stream as mature erythrocytes. Defective kidneys usually fail to produce a sufficient increase in EPO in response to hypoxia (Erslev, 1987). ErvthroDoietin Theravpy and Iron Deficiency

In the absence of compliance with a regimen of iron supplementation, the rapid erythropoiesis produced by treatment with rEPO will seriously reduce iron levels (Van Wyck, 1989). Subtle iron deficiency may be detected by low serum ferritin levels. With a further decrease in iron stores, low serum iron levels, increased iron binding capacity, and elevated free erythrocyte protoporphyrin are apparent. Iron deficiency anemia is a more severe manifestation of iron deficiency and characterized by reduced Hg levels (Ritchey, 1987). In the case of iron deficient patients treated with rEPO, iron deficiency may block further reduction of anemia and, as discussed in the next section, may have neuropsychological consequences.

Neuropsychology of Anemia

Factors believed to contribute to neuropsychological

dysfunction of renal disease include the buildup of putative neurotoxins, alterations in membrane permeability, and electrolyte disequilibrium (Powell et al., 1986; Trompeter,





28



Polinsky, Andreoli, & Fennell, 1986). Attempts to isolate specific uremic neurotoxins have met with limited success (Schreiner, 1975; Trompeter, Polinsky, Andreoli, & Fennell, 1986).

Prior to the availability of rEPO, Hg values were

occasionally included in neuropsychological studies in renal disease; however, perhaps due to the small range of values, anemia was not found to be contributory (Fennell et al., 1987; Gilli & DeBastiani, 1983). In continuous ambulatory peritoneal dialysis (CAPD) Hg levels tend to run higher, aspects of cognitive performance tend to be closer to normal and "middle molecule" clearance is up to ten times greater. Nissenson (1989) suggests that higher hemoglobin rather than lower "middle molecule" levels may underlie the neuropsychological advantage of CAPD over hemodialysis.

Numerous studies have documented the presence of

neuropsychological dysfunction in other forms of anemia. However, not one of these studies has controlled for the cause of the anemia, which in every case has been reported to produce neuropsychological deficits. As all of the known effects of rEPO are mediated by erythropoiesis, improvement in function following rEPO replacement therapy may be attributed solely to the rise in Hg and hematocrit.

Neuropsychological deficits are well documented in iron deficiency anemia; however, while there is support for the direct effect of iron deficiency on the brain, the role of anemia controlling for iron deficiency has not been examined






29



(Deinard, List, Lindgren, Hunt, & Chang, 1986; Groner, Holtzman, Charney, & Mellits, 1986). Until recently, with the exception of blood transfusions, uremic anemia has only receded following successful renal transplantation. Although transplantation often enhances many neuropsychological functions, there is a huge confound in the simultaneous improvement in both uremia and anemia. Therefore, the role of anemia in neuropsychological dysfunction has not been adequately tested in any population.

Effects of Anemia

Anemic patients may complain of a variety of symptoms including fatigue, dizziness, dyspnea, poor concentration, irritability, weakness, faintness, headache, impotence and tinnitus (Bunn, 1980b; Rapaport, 1987). Some of the complaints reported in conjunction with anemia may be due to the cause of the anemia rather than the effects of low Hg levels (Davies-Jones, Preston, & Timperley, 1980). The relationship between severity of symptoms and Hg concentration is confounded by the effectiveness of a variety of adaptive mechanisms. Symptomatology is dependent on factors including the degree and rate of reduction in oxygen carrying capacity of the blood, degree and rate of change in total blood volume, and the compensatory capacity of the cardiovascular and pulmonary systems (Wintrobe et al., 1981). Following adaptation symptoms may subside. If anemia develops slowly, effective adaptation may avert the





30


appearance of symptoms. Even in cases of severe chronic anemia (6 to 8 g per dl), the patient may be asymptomatic or complain only of fatigue (Wintrobe et al., 1981).

Fatigue appears to be the most common symptom,

regardless of severity of the anemia. Comparing symptom frequency in anemia characterized by Hg levels above or below 8 g/dl, Dawson, Ogston and Fullerton (1969) reported that in anemic subjects fatigue was present in 84.6% and 90.5%, respectively. In severe acute anemia, respiratory and circulatory symptoms suggestive of possible hypoxia (i.e., shortness of breath, "dizziness," faintness and tachycardia) are common (Wintrobe et al., 1981).

Cognitive deficits have been documented in several of the anemias (Davies-Jones, Preston, & Timperley, 1980; Tucker, Sandstead, Penland, Dawson, & Milne, 1984). However, previous studies have been confounded by the failure to control for factors causing the anemia and suspected of producing deleterious effects through mechanisms unrelated to anemia. Nevertheless, existent studies provide some relevant information. As it is the most common, the best studied and the most relevant to the current project (i.e., due to its possible appearance during rEPO treatment), iron deficiency anemia will be emphasized.

Human studies examining the effects of iron

supplementation in deficient subjects have found increased mental development scores in 15 month old infants, improved attention, but not memory in 3-6 year olds, improved problem





31


solving capacity and improved STM and attention in adults (Deinard, List, Lindgren, Hunt, & Chang, 1986; Groner, Holtzman, Charney, & Mellits, 1986; Honig & Oski, 1984; Oski, Honig, Helu, & Howanitz, 1978; Pollitt, Leibel, & Greenfield, 1983; Pollitt, Siemantri, Yunis, & Scrimshaw, 1985; Rybo, Bengtsson, Hallberg, & Oden, 1985; Walter, Kovalskys, & Steel, 1983).

Some authors attribute the neurological complications of iron deficiency to anemia (Davis-Jones, Preston, & Timperley, 1980). However, to date not one study has examined the effects of anemia controlling for the effects of iron deficiency. Anatomical and clinical evidence suggests that iron deficiency may affect dopaminergic systems (Birkmayer & Birkmayer, 1986, 1987; Drayer, et al., 1986; Hallgren & Sourander, 1958). The mechanism may be a reduction in the number of dopamine D2 receptors (BenShachar, Ashkenazi, & Youdim, 1986).

Iron responsive neuropsychological deficits have been demonstrated in nonanemic iron deficient pregnant women and infants (Groner, Holtzman, Charney, & Mellits, 1986; Oski, Honig, Helu, & Howanitz, 1983). Evans (1985) noted that improved cognitive function occurs rapidly following iron supplementation, prior to correction of the anemia; therefore, the neuropsychological deficits associated with iron deficiency do not appear to be mediated by anemia. Reviewing the pediatric literature, Ritchey (1987) concluded that iron deficiency, in the absence of anemia, adversely






32



affects infant behavior and impairs both infant and toddler performance on tests of cognitive function, with these effects magnified by the progression to iron deficiency anemia. However, subjects exhibiting anemia due to iron deficiency anemia are likely to be more iron deficient than nonanemic subjects.

Some studies have found no differences between anemic and nonanemic subjects and attributed the positive findings predominant in the literature to uncontrolled variables such as socioeconomic status and education (Johnson & McGowan, 1983). However, this does not explain the results of iron treatment studies. In addition, the results of animal studies support the predominant human findings (Massaro & Widmayer, 1981; Yehuda, Youdim, & Mostofsky, 1986; Youdim & Ben-Shachar, 1987).

Possible Mechanisms

Mechanisms that might produce neuropsychological dysfunction in anemia include alterations in blood chemistry, cerebral hypoxia and general fatigue. As previously noted, hyperventilation induced by low 02 is possible, but is attenuated by the resulting hypocapnia (i.e., low C02). The effects of severe hyperventilation include increased blood pH, alterations in neuronal excitability, changes in EEG and cerebral vasoconstriction (Fried, 1987). Carbon dioxide concentration is the primary mechanism regulating cerebral blood flow with hypocapnia inducing vasoconstriction and hypercapnia inducing





33


vasodilation (Heistadt, Marcus, & Abboud, 1987). In the anemic state vasoconstriction in response to hypocapnia is likely to result in hypoxia.

Compared to other tissues, the brain both requires a disproportionate amount of oxygen and is more sensitive to hypoxia (Davis-Jones, Preston, & Timperley, 1980). Neuropsychological deficits attributable to hypoxia have been observed in chronic lung disease, normal patients experiencing iatrogenic hypoxia and patients suffering from sleep disordered breathing. The hippocampus is especially vulnerable to hypoxia due to the nature of its blood supply. Hippocampal dysfunction is associated with deficits in LTM encoding (Muramoto, Kuru, Sugishita, & Toyokura, 1979). The disassociation between LTM and STM seen in patients with lesions in the hippocampal area is well known. However, chronic hypoxia may have more widespread effects. Measures of verbal and nonverbal memory, intelligence and verbal fluency have been reported to be affected by hypoxia in association with chronic sleep disordered breathing (Berry, Webb, Block, Bauer, & Switzer, 1986).

The relevance of hyperventilation and cerebral hypoxia to well managed hemodialysis patients is unknown. Hemodialysis increases the degree of alkalosis. Hyperventilation secondary to severe anemia might, theoretically, increase the degree of alkalosis. Although peripheral hypoxia may be common in uremic anemia, the incidence of cerebral hypoxia is unknown. The possibility






34


of an interaction between factors affecting cognitive function should not be discounted. Young adults one to three years post mild head injury (i.e., concussion) exhibit impairment on immediate memory and vigilance tasks relative to controls when all subjects are tested under mildly hypoxic conditions (Ewing, McCarthy, Gronwall, & Wrightson, 1980). Renal patients may be more susceptible to the effects of anemia than are otherwise healthy anemic patients.

In contrast to the unknown incidence of hypoxia and hyperventilation, fatigue, the first and most chronic symptom of anemia, is almost invariably present in end-stage renal disease (Dawson, Ogston, & Fullerton, 1969). Fatigue may be expected to consistently occur prior to the development of cerebral hypoxia, if cerebral hypoxia occurs at all. At a level of anemia severe enough to produce cerebral hypoxia, it appears likely that the level of fatigue will also be greatly increased. Therefore, fatigue may be a more reliable mediator of the putative neuropsychological effects of anemia.

The effects of fatigue on neuropsychological

performance are well known in clinical practice (Lezak, 1983). However, the effects do appear to be variable. Subject factors such as age and physical condition and fatigue factors such as cause and duration may be significant. No changes in neuropsychological test performance were demonstrated in 42 surgical residents when





35


assessed following 2.0 �1.5 hours of sleep vs. 6.5 �1.0 hours. In the acute sleep-deprived state, residents were less vigorous and more fatigued, depressed, tense, confused and angry (p<0.05) than they were in a rested state. These changes in mood had no effect on measures including immediate and delayed recall of logical stories, Trails, immediate memory, or digit symbol (Bartle et al., 1988). However, it is conceivable that more difficult tasks would have shown an effect.

The biological function of the experience of fatigue is hypothesized to be reduction in unnecessary activity. It is hypothesized that the effects of increasing fatigue are likely to appear along a continuum. With increasing fatigue the organism's proclivity to define activity as necessary is likely to decrease. In a testing situation, prior to refusal to perform a task, the subject's definition of the task may change to make the task less taxing. Therefore, tasks not requiring, but enhanced by, additional effortful cognitive activity may be expected to be most sensitive to the effects of fatigue. Tasks overtly demanding attention, vigilance and/or extensive cognitive processing are hypothesized to be affected only as the severity of the fatigue increases.

Effect of Erythropoietin on NeuropDschological Function

The effects of rEPO replacement therapy on quality of life and neuropsychological functioning will be briefly






36


reviewed. The results of the pilot study and possible negative side effects of rEPO will then be discussed. Effect of Ervthropoietin on Quality of Life

Treatment with rEPO has been demonstrated to increase Hg concentration and reported to improve patient well-being (Eschbach, Egrie, Downing, Browne, & Adamson, 1987). In the first clinical trial, out of ten patients, nine reported an improved sense of well-being and eight reported increased exercise tolerance (Winearls et al., 1986).

A study of 37 rEPO treated chronic hemodialysis patients found 16% of the patients returned to work, appetite improved in 81%, 78% reported subjective improvement in exercise tolerance, 70% participated in more social activities, sleep improved in 68% and 84% reported an increased sense of well being. Patients between 40 and 49 years of age showed the most improvement. A hematocrit value of >27% appeared to be the critical level for improved sexual function in men (Delano, 1989).

Evans, Rader and Manninen (1990) found significant increases in self report of energy and strength, greater appetite, improved breathing and reduced tension and anxiety. There were very small but significant increases in patient report of quality of life, affect and life satisfaction.

In an investigation of the effect of rEPO on working

capacity, eight sedentary chronic hemodialysis patients rode a bicycle ergometer before and after elevation of the group






37



mean Hg from 5.9 g/dl to 10.9 g/dl with rEPO treatment. The initial mean anaerobic threshold of 70 watts represented a very limited exercise capacity with inability to perform routine household work over a sustained period. Following rEPO treatment, the mean anaerobic threshold of 106 watts indicated that work capacity had increased enough to allow for full physical rehabilitation with regard to everyday life (Mayer, Thum, Cada, Stummvoll, & Graf, 1988). Previous Neuropsvchological Studies

Several small pilot studies have assessed the impact of rEPO on cognitive function. Wolcott, Schweitzer and Marsh (1988) reported that in a group of nine chronic hemodialysis patients administration of tests that loaded on visual, conceptual and visuomotor tracking and auditory verbal learning demonstrated trends toward improvement. Wolcott, Schweitzer and Nissenson (1989), in a study of the effects of EPO treatment in a group of 17 chronic hemodialysis patients, reported improvements in symbol-digits modality and Trails; however, it is unclear whether these improvements reached clinical or statistical significance.

Decreased P3 latency purportedly represents increased speed and efficiency of information processing. Increased amplitude of event-related potentials suggests improvement in cognitive function. Nissenon, Marsh and Brown (1988) reported that following a mean rEPO mediated improvement in hematocrit from 22.7% to 36.6%, 13 chronic hemodialysis patients exhibited a decrease in P3 latency of response to






38



auditory stimuli, however, these results did not reach statistical significance. A follow-up study (n=17) found no change in P3 latency which suggested that there was no change in speed of processing. The amplitude of the P3 wave increased significantly (p<.03). Reported changes were 19%, 63% and 160% in the parietal, vertex and frontal areas respectively, and are considered by the authors to represent increased attention span, memory and efficiency of cognitive processing (Nissenson, Marsh, Brown, Schweitzer, & Wolcott, 1989).

Results of Pilot Study

In a small pilot study, seven anemic hemodialysis

patients were assessed with a battery of neuropsychological tests and a mood questionnaire (Profile of Mood States) pre and a mean of 98.6 days post (range 77 to 125) initiation of 100 units/kg of rEPO (Klein et al. 1989). The study was prospectively randomized and double-blind; however subjects in most cases were able to accurately report their treatment status, apparently based on subjective experience and behavioral self-observation. Although eleven patients initially began the study, complete data was only obtained on seven due to a variety of factors including medical problems, transplantation and patient refusal to be tested. Results were available for two placebo controls and five patients receiving rEPO. The mean increase in Hg was 0.55 mg/dl in controls and 5.2 mg/dl in those on active treatment, a mean treatment group change from 6.10 �0.97






39



g/dl, severely anemic, to 11.56 +2.41, mildly anemic. At sea level anemia is suspected in men when Hg is <14.0 g/dl and <12.0 g/dl in women (Rapaport, 1987).



Table 1-1. Pilot Study Hemoglobin and Sum CVLT

Hg Hg Hg Sum CVLT Sum CVLT Obs Group Pre Post Change Change SS Change

1 Con 7.0 6.7 -0.3 -6 -0.93 2 Con 8.1 9.5 1.4 -5 -0.55 3 rEPO 6.0 8.4 2.4 1 0.11 4 rEPO 5.1 11.2 6.1 3 0.47 5 rEPO 7.1 13.2 6.1 8 1.08 6 rEPO 5.7 12.0 6.3 0 0.00 7 rEPO 6.6 13.0 6.4 16 1.90 Note: Obs = Observation; SS = Standard Score; Con = control

group; rEPO = treatment group; Change = change score



The Profile of Mood States, a self-report measure, failed to reveal any improvement in fatigue, tension, depression, vigor, or confusion. However, individual patients reported dramatic improvements in circumscribed domains. Individual subjects reported increased capacity for physical activity, a reduction in post dialysis fatigue and decreased problems with impotency. The dramatic reduction in physical symptoms in association with absence of subjective effects is consistent with reports of reduced subjective symptoms when the anemia is chronic, even if it





40


is severe. Despite this, the absence of change on the fatigue and vigor subscales is somewhat suspect and may represent the inability of this subject population to accurately report their subjective state in the manner required by the POMS. Wolcott, Schweitzer and Marsh (1988) also found an absence of significant changes on the PONS following a rEPO mediated increase in mean hematocrit from 23.1% to 36.0% in 13 patients; however, mean POMS vigor and fatigue scales changed in the expected directions and approached statistical significance.

As predicted, there was no improvement on the

Comprehension subtest of the Wechsler Adult Intelligence Scale - Revised. This measure of overlearned semantic stores is usually stable in the absence of severe dysfunction. There was also no improvement in verbal behavioral fluency as measured by the Controlled Oral Word Association Test (FAS). Neither was there any increase in motor speed as measured by finger tapping; however, two patients exhibited quite noticeable differences in capacity for sustained finger activity. During the first assessment finger cramping, pain and fatigue required rests between trials. Rests were not required following initiation of rEPO treatment.

On Trails A and B, a test stressing motor speed, visual search, mental control and decision speed, no improvement was noted. On Trails A, one treated subject moved from borderline to within normal limits. One treated subject,






41


moved from below normal to within normal limits; however, this subject exhibited limited response to rEPO. The other three treated subjects actually were slower post-test. The greatest improvement was seen in one of the two controls, who moved from clearly below normal to within normal limits. Likewise, on Trails B, the greatest improvement was in the treated subject who had only limited response to rEPO and the second greatest improvement was in a control. Three of the four treated, responsive subjects actually performed more poorly on the second testing. These findings contrast with the report of Wolcott, Schweitzer and Nissenson (1989) that Trails improved in treated patients.

Correlation between improved performance on the Sum of five trials on the California Verbal Learning Test (Sum CVLT) supra-span word list learning task and improvement in Hg was r=0.796 (R=.0324). The correlation between improvement in Sum CVLT and rise in hematocrit was r=0.741 (R=.0568). Neither the first two trials of the five trial sequence nor the once presented interference list showed improvement. Improvement was observed primarily in the last two trials. Single and double presentations of supra-span lists with immediate recall are highly dependent on STM (T. White, personal communication, June 12, 1990). They may be considered only weak measures of LTM. Performance on the fourth and fifth presentation is likely to be highly dependent on the degree to which the subject has organized and elaborated the list of target words. Thus, the improved






42


performance appears to have been due to enhanced LTM. This is supported by the lack of improvement on the digit span subtest of the Wechsler Intelligence Scale - Revised. However, the Logical Memory subtest of the WMS showed no change.

Logical Memory is a single trial task and, because, as the name implies, it is already logically organized, it might be expected to less vigorously discriminate between sparse and elaborate processing. Improved performance was observed on the CVLT only after the third presentation. Multitrial supra-span tasks require more active cognitive processing than does the single presentation of a paragraph. Given the hypothesized increase in general vitality, the results may be explained by a greater proclivity, posttreatment, for effortful cognitive activity such as the kind required to organize and elaboratively rehearse a shopping list.

No improvement in immediate or delayed visual memory

was observed based on the Visual Reproduction subtest of the WMS. This test involves only one presentation of the stimuli; therefore, may not be less sensitive than Sum CVLT to changes in degree of cognitive processing. However, the lack of improvement is also consistent with the suggestion that visual-motor integration skills are permanently affected by renal disease, at least if age of onset was during childhood (Mings, 1987).





43


Improved performance on Sum CVLT, in the absence of other changes, is hypothesized to be due to reduction in fatigue. It is hypothesized that the effects of fatigue are likely to appear along a continuum mediated by attributions regarding the importance of effortful activities. With increasing fatigue the saliency of cues requisite to the production of a given level of arousal is likely to change. Nonessential, covert cognitive activity may diminish prior to cognitive activity overtly and undeniably essential to performance of the stated task. Therefore, do to the nature of the tasks, neuropsychological measures directly stressing overt attention and vigilance may be less affected. However, tasks benefiting from but not requiring covert effortful cognitive processing may be most vulnerable to the effects of fatigue.

The dramatic reduction in anemia produced by treatment with EPO has been shown to produce clinically significant improvements in domains such as aerobic capacity, anaerobic threshold, sexual potency, employment status, level of social activity, perceived ability to engage in physical exercise and improved appetite. In contrast significant changes in self-reported mood have been less consistently documented. Preliminary neuropsychological test data suggests improvement on verbal supra-span multitrial list learning tasks. In addition, preliminary reports suggest possible improvement in other cognitive domains as measured






44


by increased amplitude of event-related potentials and paper and pencil tests.

An alternate interpretation of the results of studies examining the cognitive and affective effects of rEPO treatment is that the placebo effect was a major factor in the findings. Examination of Sum CVLT change scores in the study by Klein and colleagues (1989) revealed a significant portion of the variance to have been contributed by the two control subjects, both of whom declined in performance. This may be explained by negative placebo effect in the subjects initially expecting to be on a drug reported to produce major changes in life functioning. Likewise, improvement in the treatment group was consistent with a positive placebo effect. The failure of Nissenson, Marsh, Brown, Schweitzer and Wolcott (1989) to replicate the earlier study of Nissenon, Marsh and Brown (1988), a study which found increased speed of processing based on P3 latency, is also, arguably, consistent the transitory placebo effect often observed with new treatments. Negative Effects

Treatment with rEPO is not without possible negative neuropsychological consequences. Increased hematocrit appears to raise peripheral resistance resulting in increased hypertension in 10% to 22.2% of patients (Hori, Onoyama, Iseki, Fujimi, & Fujishima, 1990). Hypertension is a major problem in the management of renal disease. In the first British clinical trials, Winearls and colleagues





45


(1986) reported that one of their ten patients developed hypertensive encephalopathy. In addition, increased blood clotting may potentially reduce circulation (Winearls et al., 1986). Other side effects reported include anxiety, lethargy, headache, body aches and, in 32% of subjects in one study, iron deficiency (Delano, 1989). In addition, concern has been raised that increasing hematocrit to normal levels may compromise any existing renal function and reduce the efficiency of dialysis (Koene & Frenken, 1990).

Curiously, nervous tissue is capable of producing EPO. After renal carcinoma, cerebellar hemangioblastomas have been reported to be the most frequent cause of elevated EPO in association with neoplasms (Hennessy, Stern, & Herrick, 1967; Race, Finney, Mallams, & Balla, 1964; Waldmann, Levin, & Baldwin, 1961). The proclivity of neural tissue to produce EPO leads to speculation that EPO may normally be produced by and have a function in the brain; however, there is no evidence of EPO mediated neural effects in cancer patients suffering from EPO secreting tumors. However, physiological EPO levels are abnormally low for the degree of anemia in renal failure, but often within the range found in nonanemic populations. Therefore, EPO levels may be raised dramatically during rEPO replacement therapy. However, the suggestion that EPO might affect nervous tissue is purely speculative.

Of more interest is the observation that rEPO treatment may reduce sympathetic arousal. Cardiac output is increased





46


in severe chronic anemia. This adaptive reaction is mediated by sympathetic activation. Recombinant EPO, by reducing anemia, reduces demands for high cardiac stroke volume and heart rate; therefore, may reduce sympathetic activation (Hori, Onoyama, Iseki, Fujimi, & Fujishima, 1990). This could be of significance given suggestions of neuropsychological underactivation in uremia.

Summary and Conclusions

Memory dysfunction has been reported in chronic renal disease with reasonable consistency. Chronic renal disease is characterized by a host of symptoms, many of which stem from nervous system dysfunction. Over the past 25 years numerous studies have examined cognitive dysfunction in uremia. These studies have occurred within the context of medicine's increasing ability to simulate, or in the case of transplantation, restore normal renal function. However, today, the physiological mechanisms and cognitive processes underlying neuropsychological dysfunction in renal disease are still largely unknown.

Recombinant EPO, a potent new treatment for the anemia of renal failure, has produced dramatic improvements in work capacity, sexual potency and quality of life in impressive percentages of treated patients. Because many of the symptoms once attributed to uremia now appear to be dramatically reduced through the treatment of uremic anemia, it seemed reasonable to hope that neuropsychological dysfunction might be similarly affected. A small pilot






47


study found improvement only in effortful verbal LTM. Furthermore, in retrospect, it was noted that the sometimes reported advantage of peritoneal over hemodialysis and decline in performance with time on hemodialysis may be at least partially mediated by degree of anemia. Also, there are suggestions that uremic memory dysfunction is at least partially mediated by reduced vigilance, underarousal, increased distractibility and cognitive slowing. These observations are consistent with the hypothesis that effortful verbal memory may be vulnerable to anemia, secondary to mechanism such as fatigue or hypoxia.

The goal of the present study was to attempt to

replicate the improvement in LTM on a a multitrial supraspan learning task which had been previously shown to correlate with reduction of uremic anemia. Previous work suggested that verbal memory tasks making fewer demands on effortful cognitive processing showed no improvement. If improvement were demonstrated on effortful memory tasks as well as on automatic memory tasks and on the Levels of Processing task, then the most plausible explanation would be that uremic anemia deleteriously affects memory structures, possibly mediated by hypoxia, although other explanations might be posited. Improvement only in effortful memory would suggest that the change was due to increased cognitive processing, possibly secondary to reduced fatigue or increased activation.





48


Failure to replicate would provide information useful in several areas. The extent to which homeostatic mechanisms adapt to chronic anemia is unknown. Data delineating the parameters within which anemia affects cognitive function could potentially contribute to the determination of the optimal Hg at which to maintain patients in end-stage renal disease. This is important given the monetary costs and medical risks associated with utilizing rEPO to treat uremic anemia.

Failure to demonstrate Hg responsive neuropsychological dysfunction in rEPO treated uremic anemia would lend support to the suggestion that the deficits documented in other forms of chronic anemia are mediated by factors other than anemia. In the case of iron deficiency anemia, if the reported deficits are entirely due to the direct effects of iron deficiency on the brain, then, since iron deficiency appears prior to anemia, this would suggest the need for more aggressive screening for iron deficiency, even in the absence of anemia. Beyond scientific value, this would suggest that millions of adults and children may suffer from easily and inexpensively preventable neuropsychological dysfunction.

Increased understanding of the neuropsychological

consequences of uremic anemia may 1) increase understanding of the extent of the organism's ability to adapt to chronic anemia, 2) contribute to the quarter century quest for the uremic neurotoxin(s), 3) show the extent to which anemia






49


itself is likely to contribute to the cognitive deficits observed in other forms of anemia, 4) assist in determining the optimal level of Hg in treated renal patients and 5) in the event of positive findings, provide preliminary data regarding the mechanisms through which anemia undermines neuropsychological functioning.















CHAPTER 3

METHODS

The present study was designed to broadly address the question of whether change in anemia would be reflected by change in memory.

Subjects

All subjects were adults in end-stage renal failure and reasonably stabilized on some form of maintenance dialysis. Subjects were drawn from the Shands Teaching Hospital (STH) Adult Hemodialysis Clinic, the STH Dialysis Home Training Clinic and the Gainesville Veterans Administration Medical Center Hemodialysis Clinic. Age ranged between 23 and 73 years. Subjects with uncontrolled hypertension, mental retardation or known neurological disease were not accepted into the study.

The treatment group consisted of 16 subjects on

maintenance hemodialysis and one subject on peritoneal dialysis. It was planned that subjects would only be accepted into the study into their eighth week of rEPO treatment, based on the assumption that STH adult hemodialysis patient hematocrits would be rising for approximately five months. Therefore, a subject initially tested at 8 weeks was expected to just be reaching asymptote at the third assessment. However, subjects began 50





51


receiving rEPO earlier than anticipated and control subjects became difficult to locate. Therefore, subjects were accepted into the study beyond the eighth week, reasoning that they could serve as controls in the event of minimal additional change in Hg. Mean lag between initiation of rEPO and the first assessment was 5.9 weeks (sd=4.57).

The 18 control subjects did not receive rEPO.

Thirteen subjects, 72.2%, were on some form of peritoneal dialysis (i.e., Peritoneal Dialysis, Continuous Ambulatory Peritoneal Dialysis or Continuous Cycler Peritoneal Dialysis) and five subjects, 27.8%, were on maintenance hemodialysis. Given the small population from which to draw controls, the ideal of matching controls to treated subjects on age, education and ideally, race and sex, was not considered feasible. Despite this, as will be described later, the two groups were nearly identically in domains other than treatment modality. However, mean Hg at the first or baseline assessment was 8.46 g/dl (sd=1.42) in the treatment group and 9.44 g/dl (sd=1.92) in the control group.

All subjects were paid $15 at the completion of the third assessment. Subjects were tested in their home, their dialysis clinic, or in a STH Clinical Psychology testing room.





52


Measures

California Verbal Learning Test

The California Verbal Learning Test (CVLT) is a multitrial supra-span word list learning task designed to be an ecologically valid measure of multiple aspects of learning. The CVLT quantifies parameters including STM, LTM, retention over short and long delays, degree of vulnerability to proactive and retroactive interference, encoding strategies, effects of category cuing and the frequency of perseverations and intrusions (Delis, Kramer, Kaplan, & Ober, 1987).

Three versions were used. Form I is in general

clinical use. Form II is an alternate form developed for research purposes (Delis, McKee, Massman, Kramer, & Gettman, 1990). Form III, the Florida version of the CVLT, was developed for this study. A description of the construction and validation is provided in Appendix A. In keeping with the results of the pilot study the primary measure was the Sum of trials one through five of the CVLT (Sum CVLT).

Levels of Processing Task

A 36 word levels of processing task in use in the

Shands Teaching Hospital Psychology Clinic was used and two alternate forms were developed for this study. Alternate forms two and three were designed with target words of similar frequency of occurrence to that of Form I. Please see Appendix B.






53



The levels of processing paradigm controls the nature of effortful processing through manipulation of the task. Subjects were visually presented a written word and asked to respond to a question that required either orthographic, phonemic or semantic analysis of the visual stimuli. Following an interval of approximately 20 minutes, accuracy in identification of the target words was measured with an oral multiple choice task. Subjects able to benefit from semantic processing may be expected to exhibit a higher rate of correct recognition for semantically processed stimuli. Subjects exhibiting memory deficits due to reduced cognitive processing, because extent of processing is controlled by the task, would be expected to perform like normal subjects on this task. Normal subjects show a recognition advantage for the semantically processed words (Craik & Tulving, 1975). Subjects suffering from neurologically based memory disorders affecting encoding in LTM fail to benefit from semantic processing. Thus, improvement in ability to benefit from semantic processing would suggest physiological improvement. Frequencv Estimation Task

Estimation of frequency of occurence provides a clinical measure of automatic learning believed to be relatively invulnerable to factors such as fatigue, but presumably vulnerable to physiological factors (e.g., hypoxia) directly interfering with the functioning of memory structures (Hasher & Zacks, 1984).





54



After being instructed to try to recall as many words as possible, subjects were read a list of words with some words appearing more than once. Twenty minutes later subjects were asked to estimate how many times each word was read. Three versions were produced and are reproduced in Appendix C. The three versions were administered in consecutive order to all subjects. Prior to estimation of frequency, subjects were tested on free recall. Controlled Word Association Test

Controlled Word Association Test, a verbal fluency task, provides a measure of retrieval unrelated to encoding. It is also dependent on level of arousal, motivation, verbal fluency and ability to initiate behavior. Three commonly used sets of letter triplets are available (Lezak, 1982).

Conitive-Affective and Physical Behavior Questionnaire

A self-report questionnaire designed to assess

cognitive-affective and behavioral changes likely to be caused by medical illness, was developed utilizing the types of questions typically asked of medical patients to evaluate these domains. The questionnaire generates a Cognitive-Affective Scale and a Physical (i.e., physical activity) Scale. The questionnaire was written to avoid obvious response biases. The questionnaire and scoring instructions are displayed in Appendix D.

In the absence of positive results this measure was intended to determine whether there were positive changes






55


in behavior in the treated subjects. The same version was administered at all three assessments. H2moalobin
In most instances, Hg levels were drawn on rEPO

treated subjects on a weekly basis. When blood work was not done on the day of neuropsychological testing, pre and post target date Hg levels were used to calculate status on the day of the assessment. As Hg status tends to change linearly in the absence of major blood loss, estimation is likely to have been reasonably accurate. In the few instances where Hg levels were lacking, estimates were made based on hematocrit (Ritchey, 1987). It should be noted that the experimenter was blind to Hg levels until the completion of all other data collection. Ferritin

When available, ferritin levels were obtained for each subject to rule out iron deficiency as a possible confound (Ritchey, 1987). Ferritin, the most sensitive measure of iron levels, has been used in studies examining the neuropsychological effects of iron deficiency. However, in the absence of ferritin levels, iron saturation and/or transferrin were used if available. When the dates of neuropsychological testing and blood work failed to coincide, estimates were made based on iron measures obtained before and after the time of testing. As iron status tends to change slowly and linearly in the absence of major blood loss, estimation is reasonably accurate.





56


Iron status was used to classify subjects into low and replete categories utilizing lab report guidelines. The normal range for ferritin is frequently cited as around 10 ng/ml to 107 ng/ml. Ferritin levels of under 10 ng/ml were considered indicative of iron deficiency. Ferritin values under 30 ng/ml were classified as low. Normal and elevated iron values were grouped as normal for the purposes of these analyses.

Treatment with rEPO results in a heavy demand on iron stores and induces iron deficiency in the absence of iron supplementation (Van Wyck, 1989). For this reason, as part of routine medical procedures, serum ferritin levels were drawn on most rEPO treated subjects on a monthly basis; however, the experimenter was blind to iron status until the completion of neuropsychological data collection.

Schedule of Testing

Subjects were tested on three occasions. Assessment Two was intended to be administered approximately 30 days following the first assessment. Assessment Three was scheduled for 60 days following the second assessment. Patient illness, noncompliance and difficulty working around transportation schedules significantly altered the assessment timetable in some instances. Because time of testing relative to hemodialysis has been reported to affect results, this was kept consistent for each subject. For most subjects, assessment occurred immediately following hemodialysis.





57


Tests were administered in the same order at all three assessments and all subjects were administered test versions in the same order. Forms for the estimation of frequency of occurrence and levels of processing paradigms were administered in ascending order. For the CVLT, Form II was administered at the first assessment and Form I at the third assessment. This deviation from ideal experimental design occurred because Form III had not been fully validated prior to initiation of the study. Therefore, to minimize losses in the event of poor interform reliability, Form III was administered at the second assessment and forms were not counterbalanced.

Controlled Word Association letter triplets were administered in the following order: CFL, FAS and PRW. Because the letter triplets CFL and PRW can be directly compared, the were placed in what were expected to be the two most important assessments, one and three. FAS is the most widely used letter triplet, but shares a letter with CFL. To minimize possible practice effects for individual letters in the comparison of CFL and PRW, CFL was administered prior to FAS. As change scores rather than absolute performance was compared, error variance stemming from possible differences in difficulty between the three letter triplets was minimized.

Tasks were administered in the following order:

Informed consent, collection of demographic data, CVLT, levels of processing encoding phase, frequency of






58


occurrence encoding phase, Controlled Word Association Test, Cognitive-Affective and Physical Behavior Questionnaire, levels of processing recognition phase, CVLT delayed tasks, frequency of occurrence estimation phase.

Statistical Methodolov

Statistical analyses were performed using the

Statistical Analysis System (SAS) package of statistical procedures available for the IBM personal computer. The SAS General Linear Model procedure was used to test the main hypothesis. Analyses involving variables violating assumptions of normality and homogeneity of variance were performed utilizing nonparametric procedures. The assigned alpha for the main hypothesis was .05, with secondary hypotheses tested more stringently based on the requirements of multiple testing and the extent to which the assumptions underlying parametric procedures were violated.

Hypotheses

There was one main hypothesis and four secondary

hypotheses. Two of the secondary hypotheses were intended, in the event of failure to reject the main hypothesis, to provide preliminary data regarding possible mechanisms mediating the relationship between Hg and cognitive functioning.

Main Hypothesis

The main hypothesis was that adult renal dialysis

patients receiving rEPO replacement therapy would exhibit






59



improved verbal LTM as measured by Sum CVLT, compared to dialysis subjects not receiving rEPO. This improvement was hypothesized to be mediated by rise in Hg. Recombinant EPO itself was not expected to have any effect on Sum CVLT. Recombinant EPO treated subjects typically concurrently receive iron supplementation; therefore, iron status was not expected to change in the vast majority of subjects. Nevertheless, it was necessary to control for iron status because it has been a confound in the majority of studies focusing on the neuropsychology of anemia as deficiency is associated with neuropsychological deficits. A multivariate analysis was expected to demonstrate a main effect for Hg while controlling for rEPO treatment status and iron status.

Secondary Hypotheses

The secondary hypotheses are described below.

The first secondary hypothesis was that self-reported physical activity, as measured by the Cognitive-Affective and Physical Behavior Questionnaire, would improve in response to reduced anemia.

Despite self-reported changes in life functioning, the pilot study failed to demonstrate any alteration in affective state using a self-report adjective checklist, possibly because subjects were unable to adequately perform the task, which required the ability to abstract and introspect. Therefore, it was hypothesized that a concrete, behaviorally oriented self-report device might






60


produce more accurate results. Thus, the second secondary hypothesis was that self-reported behavioral indicators of cognitive-affective state, as measured by the CognitiveAffective and Physical Behavior Questionnaire, would improve in response to reduced anemia.

In the event of positive findings the testing of

several other hypotheses was planned. The intent of these tests was to attempt to characterize the mechanisms mediating improvement in memory. These were preplanned exploratory analyses designed to focus on the question of whether putative improvement in memory following reduction in anemia is mediated by functional or more basic physiological processes. Increased benefit from semantic processing would be suggestive of enhancement of functioning at the level of the physiological substrate of the memory system. However, it was reasoned that improvement in Sum CVLT, if observed, was more likely to be mediated by increased cognitive processing secondary to reduced fatigue; therefore, the third secondary hypothesis was that rise in Hg would not be associated with increased ability to benefit from semantic processing as measured by the semantic score on a levels of processing task.

Reasoning along similar lines, the fourth secondary

hypothesis was that rise in Hg would not be accompanied by improved automatic learning as measured by the ability to estimate frequency of occurrence.















CHAPTER 4

RESULTS

Overview of Analyses

Prior to analysis, the data was inspected for outliers and tested for the assumption of normality using qualitative and quantitative tests available through the SAS Univariate procedure. Hemoglobin and Sum CVLT, the primary independent and dependent variables, respectively, exhibited good approximations of normal distributions and homogeneity of variance. In contrast, the distributions of most other variables deviated from normality to some degree. For variables failing the Shapiro-Wilks W test of normality, the ratio of the standard deviation to the mean was computed. Variables with ratios greater than .25 were further examined quantitatively and graphically (Schlotzhauer & Littell, 1987). Procedures such as SAS General Linear Model are relatively robust with regard to violation of normality; however, great caution is required in interpreting analyses in which variables also violate the assumption of homogeneity of variance.

The assumption of homogeneity of variance between groups was tested with the Folded F test. With the exception of ferritin, all variables necessary for testing of the secondary hypotheses exhibited reasonably good

61






62



between group equality of variance at baseline. To test the impact of these violations of the assumptions of parametric procedures, parametric and nonparametric tests of between group baseline values were calculated for all subjects and the results compared. As displayed in Appendix E, even using an alpha of .05, only 2 tests out of 55 variables, both highly unstable and insignificant variables, produced discrepant results. It was concluded that the violations of normality were unlikely to be of sufficient severity to impact on the validity of SAS General Linear Model parametric procedures, except for ferritin. Since ferritin was used as a classificatory variable inequality of variance had no appreciable effect on statistical procedures.

Demographic and descriptive statistics were calculated using the SAS Means and Frequency procedures. To rule out extraneous between group differences baseline variables were compared. It was planned that the main hypothesis would be tested with a repeated measures ANOVA using the General Linear Model procedure.

A second analysis was planned to test the hypothesis that the expected treatment effect was associated with Hg change rather than rEPO itself and that the treatment effect was not an artifact of variability in iron status. Although a repeated measures analysis with the classificatory terms group membership and iron status and the covariant Hg had been planned, the unexpected finding of no change in Hg status between the second a third assessments led to the






63


dropping of the third assessment from the preplanned analyses. Therefore, the first test of the main hypothesis was a GLM univariate procedure with treatment status as the independent variable and Sum CVLT change score as the dependent variable.

The second planned test was a GLM analysis of covariance, the added independent factors being the classificatory variable iron status and the continuous variable blood Hg level. In actuality, change in iron status affected only a small number of subjects; therefore, controlling for iron status became of minimal importance.

Kruskal-Wallis nonparametric procedures were used to

test the hypotheses that self-reported physical vitality and self-reported cognitive behavior improved in the treatment group. If a main effect for Hg on Sum CVLT had been found then further nonparametric analyses would have been performed in an attempt to provide preliminary data to guide future inquiry into the mechanisms mediating the effect. Specifically, between group differences in ability to benefit from semantic processing and changes in encoding of incidental information were planned. Negative findings eliminated this rationale for these tests. Nonetheless, they were included in post hoc analyses.

Post hoc exploratory analyses focused on attempts to disconfirm the negative findings and to explore other aspects of the data. The observation that treatment status was a poorer predictor of Hg change than had been






64


anticipated, combined with reasoning along the lines that response to anemia may be quite idiosyncratic, led to the decision to reclassify subjects based on Hg change scores. Unfortunately, the small range of Hg change in most subjects necessitated a choice between sacrificing either statistical power or clinical relevance. Retaining statistical power, subjects were first classified based on small distinctions with unlikely and undemonstrated physiological significance. Following this, compromising between sample size and physiological significance, only subjects meeting the criteria of Hg change scores of �1 g/dl were retained. CVLT scores in subjects with clinically significant changes in Hg status were then examined.

The post hoc hypothesis that reduction in anemia might alter the pattern of learning was tested. The slopes of learning curves for trials one through five of the CVLT were calculated with SAS Regression Analysis procedures. A GLM ANOVA then compared changes in slopes between groups. Additional procedures were exploratory. Change scores were calculated for all variables and the performance of specific variables analyzed. Also, relationships between test performance and questionnaire data were explored.

Statistical Assumptions
The basic assumptions underlying most analysis of

variance procedures include 1) normality and 2) homogeneity of within-group variance. Analysis of variance procedures are quite robust with regard to violation of the assumption






65


of normality as well as departures from homogeneity of variance. However, the concurrent violation of both assumptions calls at the very least for conservatism in determination of significance.

As can be seen in Appendix F, Table F-1, based on the Shapiro-Wilks W statistic computed by SAS Univariate, the primary independent variable, Hg, and the major dependent variable, Sum CVLT, were both normally distributed. Hg values were slightly skewed to the right. Sum CVLT was very minimally skewed to the left. Kurtosis in the distribution of Hg values approached zero. For Sum CVLT, the kurtosis value showed the tails to be slightly lighter than expected in a normal distribution.

Iron levels tend to be quite elevated in dialysis patients dependent on repeated blood transfusions. In patients running relatively normal iron levels, rEPO treatment may deplete iron stores. Therefore, it is not surprising that serum ferritin violated the assumption of normality. As it was used only to classify subjects in terms of iron status, this violation was irrelevant to the analyses.

Other measures required for the testing of secondary hypotheses for the most part violated assumptions of normality to some degree. For variables failing the Shapiro-Wilks W test of normality, the ratio of the standard deviation to the mean was computed, as displayed in Appendix F, Table F-i. Variables with ratios greater than .25 were






66


further examined quantitatively (i.e., kurtosis and skew) and graphically (i.e., box plots)

The estimation of frequency of occurrence data violated the test of normality. The ratio of the standard deviation to the mean was 0.262, just over the established limit. The distribution was somewhat skewed to the left; however, Kurtosis was minimal.

The Semantic score generated by the levels of processing task failed the Shapiro-Wilks W test of normality; however, the ratio of the standard deviation to the mean was within acceptable limits. The Orthographic variable from the levels of processing task also failed the test of normality and the standard deviation was large relative to the mean.

The two Cognitive-Affective and Physical Behavior

Questionnaire scales failed the Shapiro-Wilks W test and the standard deviation to mean ratio was .280, slightly above the .25 recommended cut-off. The scores were slightly skewed to the left and the tails were heavier than expected in a normal distribution. The variables age, years of education and years on dialysis all failed the Shapiro-Wilks test. However, the ratio of standard deviation to mean was within accepted limits for the education variable. Slope and intercept of the learning curve for CVLT trials 1 - 5, variables used for post hoc analyses, both passed the Shapiro-Wilks test. Slope had somewhat heavy tails.






67


The assumption of equality of variance between groups was tested with the Folded F test, F'. The results, as displayed in Appendix F, Table F-2, revealed the following CVLT variables to fail the test of homogeneity of variance at the .10 level: Sum of clusters, Sum of intrusions, and Sum of perseverations for trials one through five, Short Delay perseverations and intrusions, and all three CVLT measures of false positives not on the interference list. Other measures failing the test of homogeneity of variance were: levels of processing recall of orthographically processed stimuli, years of education and ferritin.

Many major variables violated the assumption of normality to some degree, but exhibited acceptable approximations of homogeneity of variance; therefore, they were considered candidates for parametric analyses. These variables included: CVLT Short Delay, Long Delay, Short Delay Cued Recall, Long Delay Cued Recall, Recognition Hits, Levels of processing Semantic recognition, estimation of frequency of occurrence and associated free recall and both scales from the Cognitive-Affective and Physical Behavior Questionnaire. Please see Appendix F, Table F-2. The Orthographic variable generated by the levels of processing task was normally distributed but failed the test of homogeneity of variance. Ferritin, Sum CVLT Clusters and years of education were neither normally distributed nor had equality of variance; therefore, parametric procedures were considered unsuitable for their analysis. With the






68


exception of ferritin, all variables necessary for preplanned tests exhibited both reasonably good between group equality of variance and were normally distributed. As the effects of ferritin were judged to be nonlinearly related to ferritin concentration, it was decided to use ferritin as a classificatory variable.

To test the impact of the above violations of the assumptions of parametric procedures, parametric and nonparametric tests of between group baseline values were calculated for all subjects and the results compared. As displayed in Appendix E, even using an alpha of .05, only 2 tests out of 55 variables produced discrepant results. They were CVLT variables that may be expected to be relatively unstable: Short Delay Perseverations and Long Delay Intrusions. This test suggested that the violations of normality were unlikely to be of sufficient severity to impact on the validity of SAS General Linear Model parametric procedures.

The statistical characteristics of the distributions and variance of all variables may be viewed in Appendix F. Tables F-1 and F-2.

There was one debatable outlier. Select analyses were re-run following deletion of this subject. Examination of the raw data revealed that at the first assessment the subject had recalled 8 words on the first trial of the CVLT, but only 3 on the fourth trial and 5 on the fifth trial. Other data for this subject was within expected limits.






69



Descriptive Statistics



The treatment group had a mean age of 43.24 years

(sd=13.45) and consisted of 17 subjects. The mean age of the 18 subjects in the control group was 48.22 years (sd=14.19). A nonparametric Kruskal-Wallis test procedure failed to demonstrate a statistically significant between group difference (X2(l, n=35)=1.13, R<.2948). Education

Mean years of education in the treatment group was 11.41 years (sd=3.02) and 11.82 years (sd=2.15) in the control group. A Kruskal-Wallis test procedure failed to reveal a significant statistical differences between groups (X2(1, n=35)=0.23, g<.6325).

Sex

Overall, the subject population was equally divided by sex with 17 males and 18 females. The treatment group consisted of 7 males and 10 females. The control group consisted of 10 males and 8 females. Thus, the control group was 55.6% male and the treatment group was 41.2% male. A Chi-Square procedure failed to demonstrate a statistically significant difference (X2(l, n=35)= 0.724, R< 0.395). Treatment Modality

In the control group 72.2% were on some form of

peritoneal dialysis (i.e., Peritoneal Dialysis, Continuous Ambulatory Peritoneal Dialysis, or Continuous Cycler Peritoneal Dialysis). Five subjects or 27.8% of the control






70


group were on maintenance hemodialysis. In contrast, 94.1% of the treatment group was on maintenance hemodialysis and one subject representing 5.9% of the treatment group was maintained on peritoneal dialysis. As confirmed by a ChiSquare procedure these differences were highly significant (X2(1, n=35)=16.032, g<.0001).

Determination of time on dialysis was made difficult by intervening periods of renal transplantation as well as inconsistencies between medical records and patient reports. Also, this figure fails to take into account possible differences in the effects of hemodialysis and peritoneal dialysis over time. Mean years on dialysis for the treatment group was 6.64 years (sd=7.69) compared to 2.85 (sd=3.76) years for the control group. These differences approached but failed to reach statistical significance (X2(1, n=35)=2.8425, R<.0918) based on the Kruskal-Wallis Test procedure.

Attrition

All 17 subjects in the treatment group completed the study; however, one 25 year old male died six weeks after completion of the third assessment. Two subjects in the control group, both on hemodialysis, died between the second and third assessments. In addition, one control subject maintained on peritoneal dialysis declined to complete the third assessment after becoming hospitalized in a neighboring city. Eliminating the three subjects not completing the study resulted in a mean age in the control






71



group of 46.49 years (sd=13.61) and a mean of 12.10 (sd=1.63) years of education. The results of a Wilcoxon 2Sample Test (S=267.00, Z=0.7177, R>.4729) failed to demonstrate a significant age difference between the treatment (n=17) and control (n=15) groups. Likewise, the Wilcoxon 2-Sample Test procedure failed to demonstrate a significant difference in years of education between the treatment (n=17) and control (n=15) groups (S= 259.00, Z=0.440, R>.6596).

Cognitive function has been reported to fall prior to nonaccidental death (White & Cunningham, 1988). Consistent with this phenomenon of terminal decline, all three subjects were in the predicted direction, although in two cases the drop was extremely small. For the three deceased subjects, mean Sum CVLT in the last assessment prior to death was 25.56 compared to 28.70 in the previous assessment. The overall effect of terminal drop was to slightly lower control group assessment two and three scores relative to the treatment group.

Time Intervals Between Assessments

The mean time span between assessments one and two was 51.12 days (sd=32.95) for the treatment group. Control group mean time interval between the first and second assessments was 56.50 days (sd=35.98). The 5.38 days greater time interval for the control group failed to reach statistical significance based on a Wilcoxon Rank Sum Test (S= 277.500, Z=-.92647, R>=.3542).






72



Mean days between assessments two and three was 67.80

(sd=34.56) in the control group and 90.41 days (sd=24.03) in the treatment group. Based on the results of a Wilcoxon Rank Sum Test the 22.61 days greater inter-test interval for the treatment group was clearly statistically significant (S=178.50, Z= -2.588, p>=.0096).

The number of days between assessments one and three

was 141.53 days (sd=34.53) in the treatment group (n=17) and 117.67 days (sd=28.24) in the control group (n=15). The results of a Wilcoxon Rank Sum test suggest that the 23.86 day difference between these two groups was statistically significant (S=181.00, Z= -2.495, p>.0126). Mean time lag between initiation of rEPO and the first assessment was 41.4 days (sd=32.1)

Hemoglobin

Mean Hg at the first or baseline assessment was 8.46 g/dl (sd=1.42) in the treatment group and 9.44 g/dl (sd=1.92) in the control group. The nature of these differences suggested a possible trend toward statistical significance as demonstrated by a GLM Analysis of Variance procedure (F (1, 34)=2.94, p=.096). which is consistent with reports of lower Hg levels in hemodialysis than in peritoneal dialysis. Baseline mean hematocrit levels were 28.19% (sd=6.12) in the control group and 25.58% (sd=3.77) in the treatment group. As demonstrated by a GLM ANOVA procedure these differences failed to reach statistical significance (F(1, 34)=2.29, p=.140).






73



As expected, a one-way GLM procedure demonstrated that between the first and second assessments the mean Hg rise of

1.13 g/dl in the treatment group (n=17) was significantly greater (F=7.42, p=.0102) than the -0.21 g/dl Hg change in the control group (n=18). At the time of the second assessment mean Hg was higher in the treatment group (M=9.59, sd=1.84) than in the control group (M=9.23, sd=1.81) but the difference was not statistically significant (F(l, 34)= 0.34, R=.564).

GLM analysis of variance of Hg change scores between the second and third assessments failed to reveal even a trend towards a between group difference (F(1, 31)= 0.16. 2=.6964). Lack of change in the second phase of the study contributed to the failure to find no more than a trend towards a significant rise in Hg status between the first and third assessments (F=3.19, R=.0841). At the third assessment mean Hg was 9.54 g/dl (sd=l1.48) in the treatment group (n=17) and 9.62 g/dl (sd=1.92) in the control group

(n=15).

In this sample the 21 hemodialysis patients exhibited a mean baseline Hg of 8.80 g/dl (SD=1.80) and the 14 peritoneal dialysis patients had a mean Hg level of 9.21 g/dl (sd=1.69). The differences were not statistically significant (F (1, 34)= 0.44, p= .51). Within hemodialysis patients, mean baseline Hg in the controls (n=5) was 9.83 g/dl (sd=2.53). In the treatment group mean Hg at the first assessment was 8.48 g/dl (sd=1.48). Based on a T-Test






74



procedure these differences failed to reach statistical significance (T=1.5099, g=.1475).

Comparison of the maintenance hemodialysis pretreatment Hg of 7.06 g/dl (sd= 1.42) in the rEPO group (n=16) to the baseline Hg of 9.83 g/dl (sd=2.53) in the control group (n=5) revealed a trend toward lower values in the treatment group. However, a T-Test procedure failed to demonstrate a statistically significant difference between groups (T=2.3359, df=4.8, R=.0692).

Prior to initiation of rEPO therapy, mean Hg in the treatment group was 7.39 g/dl (sd=1.94). At the first assessment treatment group Hg had risen to 8.46 g/dl (sd=1.42). Based on a T-Test these differences were not statistically significant (T=-1.8361, df=32.0, R=.0756).

A GLM analysis of variance procedure revealed a main effect for sex (F (1, 34)= 15.24, p=.0004). Mean male (n=17) Hg was 9.96 g/dl (sd=l1.61) and mean female (n=18) Hg was 8.03 g/dl (sd=1.31), which is consistent with the 2 g/dl lower Hg level reported in normal females compared to normal males in the general healthy adult population. There was no treatment by sex interaction.

Excluding the 3 subjects who did not complete the

study, rise in Hg was 0.04 g/dl in the control group (n=15). As anticipated for the control group, mean change in Hg between assessments one and three was not statistically significant (F(1, 32)= 0.07, p=.786). In contrast the rise in Hg was 1.08 g/dl (sd=1.23) between assessments one and






75



three in the rEPO treated group which was statistically significant (F(l, 33)= 4.74, p=.037).

In the time lag between initiation of rEPO and the first assessment mean Hg levels rose 1.07 g/dl (sd-1.90) from an initial level of 7.39 g/dl (sd=1.94). The increase of 1.08 g/dl in the treatment group (n=17) between assessments one and three was 50% of the total increase of 2.15 g/dl (sd=1.23) in Hg over the course of rEPO therapy. However, this total increase of 2.15 g/dl was only 39% of the increase obtained in the pilot study. Please see Tables 4-1 and 4-2 for of Hg levels and change scores, respectively.

For descriptive purposes subjects were classified using the following system: very severe anemia < 6 g/dl Hg, severe anemic 6 - 8 g/dl, moderate anemia 8 - 10 g/dl, mild anemia 10 - 12 g/dl, > 12 g/dl normal. As illustrated in Table 43, between the first and second assessments seven treated subjects showed an improvement in category and one declined. In the controls four rose and three fell. Between the second and third assessments change appeared to be random. Ferritin

Because of the tendency of rEPO treatment to draw down iron stores ferritin levels were assayed more often in rEPO treated subjects; therefore, baseline ferritin levels were available for 94% (n=16) of the treatment group but only 33% (n=6) of the controls. At baseline mean treatment group (n=16) ferritin level was 545.93 ng/ml (sd=734.50) and mean






76



control group (n=6) ferritin level was 523.62 ng/ml (sd=626.13). A Kruskal-Wallis Test failed to reveal a statistically significance difference between these two groups (X2(1, n= 35)= 0.3478, 2<.5553).



Table 4-1. Mean Hemoglobin by Group and Assessment Ass Grp N Min Max Mean Sd

pre trt rEPO 17 4.80 12.70 7.39 1.94 1 rEPO 17 5.60 11.30 8.46 1.42 2 rEPO 17 6.60 14.20 9.59 1.84 3 rEPO 17 6.90 12.25 9.54 1.48 1 Con 18 5.27 13.90 9.44 1.92 2 Con 18 6.30 12.60 9.23 1.81 3 Con 15 6.00 12.30 9.62 1.92 Note: Ass = Assessment; Grp = Group; N = Number of subjects

Min = Minimum; Max = Maximum; Sd = Standard deviation

Con = Control; rEPO = rEPO treated group



Table 4-2. Mean Hq Change by Group and Test Intervals

Test
Interval Grp N Min Max Mean Sd

First Con 18 -2.95 1.70 -0.21 1.34 Second Con 15 -2.10 1.90 0.15 1.14 Pre-Ass rEPO 17 -4.60 4.70 1.07 1.90 First rEPO 17 -0.70 5.90 1.13 1.57 Second rEPO 17 -5.50 1.80 -0.05 1.59 Total rEPO 17 -2.40 5.30 2.15 1.86






77



Table 4-3. Anemia Status Across Assessments Id Assess 1 Assess 2 Assess 3 1 Severe Moderate + Moderate 2 Moderate Moderate Moderate 3 Moderate Moderate Moderate

4 Severe Moderate + Severe

5 Very Severe Severe + Severe

6 Moderate Normal + Moderate 7 Moderate Severe - Moderate + 8 Severe Moderate + Severe

9 Moderate Moderate Moderate 10 Moderate Moderate Moderate 11 Severe Moderate + Moderate 12 Severe Severe Severe 13 Moderate Moderate Moderate 14 Moderate Moderate Moderate 15 Moderate Moderate Normal + 16 Moderate Moderate Moderate 17 Severe Moderate + Moderate 101 Severe Moderate + Moderate 102 Moderate Moderate Moderate 103 Very Severe Severe + Severe 104 Moderate Moderate 105 Normal Normal Moderate 106 Moderate Severe 107 Normal Moderate + Normal + 108 Moderate Moderate Moderate







78



Table 4-3 continued

Id Assess 1 Assess 2 Assess 3

109 Moderate Moderate Moderate 110 Moderate Moderate Moderate 111 Moderate Moderate Moderate 112 Moderate Moderate Moderate 113 Moderate Moderate Moderate 114 Moderate Moderate Moderate 115 Moderate Severe - Very Severe 116 Moderate Moderate Moderate 117 Severe Moderate + 118 Moderate Severe - Severe Note: Valence sign = direction of categorical change;

Id < 100 = treatment group; Id > 100 = controls



As displayed in Table 4-4, subjects were classified based on ferritin level (under 10 Deficient, 10 - 30 Low, over 30 Replete). Classifying subjects into Replete, Low and Deficient categories revealed little variability in iron status across assessments.



Table 4-4. Number of Subjects in each Category of Blood Iron Level

Iron Assess 1 Assess 2 Assess 3 Status Con rEPO Con rEPO Con rEPO

Replete 14 11 14 9 12 11 Low 1 4 1 6 1 3 Deficient 0 1 0 0 0 1






79



At the time of the first assessment, the Control group included one subject with Low iron status and 14 Replete subjects. At Assessment Two this was unchanged. At Assessment Three there was still one subject with Low iron status. In the rEPO group at Assessment One, there was one Deficient subject, four Low iron subjects and 11 Replete subjects. At Assessment Two, there were 6 Low iron and nine Replete subjects. By the third assessment, the treatment group was composed of one Deficient, three Low and 11 Replete subjects. For the subjects on which data was available there was minimal change in iron status; therefore, the need to control for iron deficiency was greatly reduced.

Replete subjects were further divided into Replete and High (over 1,000). As illustrated in Table 4-5 below, the percentage of subjects exhibiting low iron status was consistently greater in the rEPO treatment group as was the percentage of subjects with high iron status. Also apparent from the Table 4-4, iron status was relatively stable across assessments, especially if the top two and bottom two groups were collapsed into Low and Replete categories. Baseline NeuropsycholoQical Functioning

Table 4-6 displays selected raw and standard scores for CVLT II at the baseline assessment. There are no statistically significant differences for any of the scores listed in Table 4-6 or in the more comprehensive list of scores in Appendix G. However, despite the lack of






80



statistically significant differences, examination of raw scores suggests a tendency for somewhat better performance in the treatment group at the first assessment.



Table 4-5. Percentage of Subjects in each Category of Iron Status

Assess 1 Assess 2 Assess 3
Iron Con rEPO Con rEPO Con rEPO

Hi 6.7 23.5 0 20.0 0 18.8 Normal 86.7 41.2 93.7 40.0 92.3 50.0 Lo 6.7 29.4 6.7 40.0 7.7 25.0 Defic 0 5.9 0 0 0 6.3 Note: Hi = Elevated Ferritin; Lo = Low; Defic - Deficient



The mean baseline age corrected control group Sum CVLT score placed the group in the 4th percentile, the borderline range. The mean baseline Sum CVLT score in the treatment group placed it in the 9th percentile, the Low Average range. However, the differences, as previously stated, did not reach statistical significance.

As displayed in Appendix G, Table G-1, on the

Controlled Word Association test, using the letters CFP, the control group (n=17) produced a mean of 11.53 words (sd=3.84). The treatment group (n=17) produced a mean of 10.53 (sd=4.45) words. The difference was not statistically significant.

The levels of processing task revealed that both groups benefited from semantic processing compared to either






81


orthographic or phonemic processing. There was a very slight nonsignificant advantage for phonemic over orthographic processing for both groups. At baseline the treatment group appeared to demonstrate greater benefit from semantic processing than did the control group, although this did not reach statistical significance. Performance on the estimation of frequency of occurrence task was nearly identical between the two groups.


Table 4-6. CVLT Raw and Standard Scores at Baseline

Control Group rEPO Treated Group
n=18 n=17 p
Mean Sd Mean Sd value
Trial 1 5.72 1.74 5.88 2.47 .9464 standard -1.33 0.91 -1.35 1.37 .7700 Sum 1-5 43.22 10.70 45.82 13.03 .5222 T score 33.06 10.31 33.82 16.00 .8662 Short Delay 9.06 2.71 9.65 3.26 .9867 standard -1.11 0.83 -1.06 1.43 .9864 SD Cued 10.94 2.69 10.94 2.90 .7393 standard -0.83 1.04 -0.88 1.36 .9052 Long Delay 9.94 3.44 9.47 3.76 .6660 standard -0.89 1.02 -1.41 1.66 .4344 LD Cued 11.06 3.51 10.82 3.34 .5587 standard -0.78 1.40 -1.24 1.56 .3863 Note: Sd = Standard deviation; SD = Short Delay; LD = Long

Delay; standard = Standard Score






82



Test of Main Hypothesis

The main hypothesis stated that rEPO mediated reduction in chronic anemia in adult dialysis patients is associated with improved verbal LTM as measured by the CVLT Sum of trials one through five.

Mean Sum CVLT change scores between the first and

second assessments were -0.82 (sd=5.38) in the treatment group (n=17) and 4.06 (sd=9.91) in the control group (n=18). These results were in the opposite of the hypothesized direction. A GLM analysis of covariance procedure failed to demonstrate a main effect for either treatment status (F=1.29, p=.2638) or Hg (F=1.32, R=.2590).

With both treatment status and Hg change in the model, few conclusions could be drawn. Therefore, separate univariate GLM analyses were performed for treatment status and for Hg change. However, these analyses failed to show a statistically significant main effect for either Hg ((F=3.25, R=.0805) or for treatment status (F=3.22,

g=.0818).

Examination of mean Sum CVLT scores between groups and across assessments revealed a curious pattern. Although none of the differences reached statistical significance, examination of Table 4-7 reveals that the control group performed poorly on the first assessment relative to the other two assessments. In contrast, assuming equivalence of test forms, treatment group performance was static. That is, change apparently occurred unrelated to rEPO.






83


Table 4-7. Mean Sum CVLT Across Assessments and Between Groups

Grp Variable N Mean Sd Min Max

Con CVLT II 18 43.22 10.70 23 60 Con CVLT III 18 47.28 10.68 30 68 Con CVLT I 15 47.07 10.40 32 68 Trt CVLT II 17 45.82 13.03 21 68 Trt CVLT III 17 45.00 13.36 16 64 Trt CVLT I 17 45.12 11.61 23 69



Representing iron status as a two-level classificatory variable (i.e., Low vs. Replete), a GLM analysis of variance failed to demonstrate a main effect for iron status on Sum CVLT change scores between the first and second assessments (F=0.01, R=.9384). Similar results were obtained introducing Hg into the model (Hg: F=0.26, p=.6126; iron: F=0.01, R=.9337).

Despite the lack of change in Hg status between the

second and third assessments, a GLM Repeated Measures ANOVA was performed. As expected, the hypothesis of a treatment effect was not supported by Wilks' Lambda (F=0.2134, R=.8092) or any other measure. Neither were there between subject effects, within subject effects or treatment by assessment interactions.

Post Hoc Analyses

Learnin Curve

Hypothesizing that change in learning curve might be more sensitive to the effects of anemia than Sum CVLT, SAS






84


Regression procedures were used to generate slopes of the learning curves for trials one through five for each administration of the CVLT.

A GLM procedure analyzing change scores between the first and second assessments failed to demonstrate a relationship between Hg change and change in slope (F=0.08, p=.7766). An identical analysis examining change scores between the second and third assessments produced similar results (F=0.00, p=.9583). A GLM univariate analysis of the effect of treatment status on slope change scores between the first and second assessments failed to reach statistical significance (F= 3.67, p=.0651). Controlling for the increased probability of a Type I error as a result of multiple analyses, the finding can not be considered suggestive of a trend. A Wilcoxon Rank Sum Test produced similar results (Z= 1.83, 2=.0669). Further analysis revealed the change to be primarily in the treatment group, but in the opposite of the expected direction. In the treatment group, mean slope declined -0.84 (sd=1.59), and in the control group slope improved 0.15 (sd=1.27) between the first and second assessments. A GLM univariate analysis of the effect of treatment group status on slope change scores between the second and third assessments failed to suggest a relationship (F=1.30, R=.2628).

Despite extensive analyses, absolutely no evidence was found that even mildly suggested improvement in learning curve in association with improvement in Hg status.






85



Exploratory Analyses

Observation of variability in Hg change scores in both groups led to reclassification based on direction of Hg change, rather than treatment status. A total of 23 subjects exhibited a rise in Hg between the first and second assessments (M=1.22 g/dl, sd=1.26). Twelve subjects exhibited a fall between the first and second assessments (M=-1.05, sd=.96). The difference between groups with rEPO treatment status as the classificatory variable produced a mean difference in Hg change between groups of 1.03 g/dl. In contrast, the alternate system of classification doubled the Hg change score yielding 2.27 g/dl. Difference in Hg change scores between the positive and negative change groups was highly significant (F=29.75, p=.0001).

Control subjects suffering a fall in Hg (n=9) exhibited a mean fall of -1.27 g/dl (sd=l.02). The three rEPO treated subjects exhibiting a fall in Hg had a mean drop of -.38 g/dl (sd=.28). Among rEPO treated subjects with rising Hg (n=14) the mean rise was 1.46 g/dl (sd=l1.54). Subjects from the control group exhibiting a rise in Hg (n=9) showed a mean rise of 0.85 g/dl (sd=0.51).

For exploratory purposes, the main hypothesis was

retested with a GLM procedure utilizing sign of Hg change as the classificatory variable. There was a significant difference in Sum CVLT change scores between the positive (n=23, M=-1.22, sd=7.18) and negative (n=12, M=7.25, sd=7.63) Hg change scores (F=10.52, p=.0027). The analysis






86


was rerun excluding one possible outlier; however, this had little effect on the results (F=7.82, p=.0087). The results were consistent with the original finding (i.e., using treatment status as the classificatory variable) in that they were in the opposite of the predicted direction. In a GLM multiple analysis of variance treatment status was nonsignificant (F=0.59, R=.4482) while sign of Hg change was significant (F=9.45, p=.0043). Also consistent with the original finding, as illustrated in Table 4-8 below, mean Sum CVLT scores revealed poorer performance at the time of the first assessment in the group whose Hg levels would rise.



Table 4-8. Mean Sum CVLT by Sign of Hg Change Between Assessments 1 & 2

Hg Fell n=12 Hg Rose n=23 Asses Mean Sd Mean Sd

1 36.58 9.48 48.61 10.86 2 43.83 11.80 47.39 12.07



In an attempt to reduce error variance contributed by subjects exhibiting very small insignificant changes in Hg, subjects with a Hg change of under 1 g/dl were excluded from the analysis. Mean Sum CVLT change in the negative Hg change group was 9.33 (sd=10.63) and -2.23 (sd=4.30) in the positive Hg change group. A Wilcoxon Rank Sum Test revealed the difference to reach statistical significance at the .05 level (Z=2.373, R=.0176).






87


To test the effect of Hg changes within the range closer to that likely to be considered physiologically relevant, subjects with less than a 2 g/dl Hg change were excluded from the final analysis. The two negative Hg change subjects in the data set had a mean Hg drop change of

2.68 g/dl. The three positive subjects in the data set exhibited a mean Hg rise of 3.77 g/dl. A GLM analysis of this data set (n=5) failed to reveal a main effect for Hg (F= 0.03, R=.8803); however, the small sample size greatly reduced the power of the test and, again, the change was in the opposite of the predicted direction. Mean CVLT improvement was 2 words in the falling Hg group and the rise was 1.33 words in the rising Hg group.

Hypothesizing that the observed effect of Hg change on Sum CVLT might be mediated by affective change, the Cognitive-Affective self-report variable was included in a GLM analysis of variance model. A multivariate analysis demonstrated main effects for both direction of Hg change (F=13.27, p=.0010) and Cognitive-Affective self-report (F=8.29, p=.0072). Further analysis revealed a main effect for direction of Hg change (F=16.97, R=.0003) and an interaction between direction of Hg change and self-reported Cognitive-Affective behavior (F=6.53, R=.0043). When Cognitive-Affective behavior was included in this model it failed to reach significance (F=1.88, R=.1807). Further analyses suggested no relationship between CognitiveAffective behavior and sign of Hg change (r=.11056,






88



p=.5336). Correlation between Sum CVLT and CognitiveAffective self-report was positive (r=.37042, R=.0310). Examination of mean scores by groups revealed that Sum CVLT rose in association with rise in Cognitive-Affective behavior; however, there was less evidence of a relationship when Hg fell.

In a model analyzing change in Sum CVLT, covarying

change in self-reported Physical Activity and direction of change in Hg failed to demonstrate a main effect for Physical Activity (F=0.77, g=.3866). There was no interaction between sign of Hg change and Physical Activity (F=0.75, g=.4827). Neither was there an effect on Sum CVLT for self-reported level of Physical Activity at the time of the first assessment. Analyses failed to demonstrate a significant relationship between self-reported Physical Activity and Sum CVLT

Sum of CVLT Trials Four and Five

Using sign of Hg change scores as the classificatory

variable, a Wilcoxon 2-Sample Rank Sum Test was used to test the post hoc hypothesis of a change in the sum of CVLT trials four and five between the first and second assessment. The difference between groups was significant even controlling for multiple comparisons (Z=2.6366, 2=.0084). Once again the difference was the reverse of the predicted direction. Mean increase in number of words recalled for the sum of trials four and five was 3.75 (sd=5.24) in the group with falling Hg (n=12) and -0.87






89


(sd=4.00) in the group with rising Hg (n=23). Please see Table 4-9.


Table 4-9. Mean Chance in CVLT Trials 4 plus 5 Change Scores Between Assessments I & II Grp N Min Max Mean Sd

Pos 23 -8.00 6.00 -0.87 4.00 Neg 12 -4.00 18.00 3.75 5.24


An identical analysis between the second and third

assessments revealed a mean change of -0.78 words (sd=3.93) in the negative change group (n=9) and a mean change of

-0.15 words (sd=3.96) in the positive Hg change group (n=20). These differences failed to reach statistical significance (Z=-.2845, R=.7760). Analysis of change scores between the first and third assessments revealed a mean change of -3.71 words (sd=7.06) in the negative change group (n=7) and a mean change of -1.25 words (sd=4.23) in the positive Hg change group (n=24). These differences failed to reach statistical significance (Z=1.5613, R=.1184).

Multivariate analysis (n=35) with a General Linear

Model failed to find a main effect for rEPO treatment (i.e., experimental vs. control) on Sum CVLT (F=0.84, R=.3668) in a model including a sign of Hg change factor (F=7.45, p=.0102).

To test the possibility that undiscovered outliers

unduly influenced the results, Sum CVLT change scores were






90


classified based on their sign. A Wilcoxon Rank Sum procedure analyzed differences in sign of Sum CVLT change scores between the first and second assessments. With the sign of Hg change scores serving as the grouping variable, again there was a significant effect for Hg change (Z=2.9180, R=.0035). Analysis using a Chi-Square procedure produced similar results (X2=8.887, df=l, p=.003).

Short Term Memory

Exploratory analysis revealed a significant difference at .01 in CVLT Trial One, a measure loading heavily on STM. However, the change was in the opposite of the expected direction between treatment and control groups between assessments one and two and is likely a type two error. Furthermore, equivalent change, also significant at .01, in the opposite direction, was observed between the second and third assessments. Please see the results displayed in Appendix H.

Verbal Fluency

An analysis of variance procedure suggested that verbal fluency as measured by Controlled Word Association test performance was somewhat low in the control group at the first assessment relative to the overall pattern of performance in both groups across three assessments (F=4.99, p=.0328. Utilizing sign of Hg change as a classificatory variable, Controlled Word Association test performance was not significantly different between assessments (F=0.01, R=.9761).






91


Secondary HYDotheses
Levels of Processing

Treatment status was insignificant in explaining

variability in performance on any portion of the Levels of Processing Task, including the Semantic recall measure (F=0.43, R=.5164). Mean change scores are listed in Appendix H. Re-analysis using direction of Hg change as a classificatory variable produced similar results (F=2.12, R=.1564). This analysis would have been considered a secondary hypothesis if there had been a positive association between Hg and Sum CVLT. Estimation of Freauency of Occurrence

As displayed in Appendix H, there were no significant differences in Frequency Estimation performance between groups or assessments. Parenthetically, LTM, as measured by free recall of the single presentation target word list, improved slightly more in the treatment group than in the control group; however, this failed to reach significance controlling for multiple analyses. Quality of Life

Quality of life is operationally defined here in terms of the items on the Cognitive-Affective and Physical Behavior Questionnaire and is divided into cognitiveaffective and overt physical behavior domains. It was hypothesized that self-reported physical vitality, as measured by the self-report questionnaire, would improve in response to treatment with rEPO.






92


Because Physical Activity was not normally distributed, a Wilcoxon Rank Sum Test procedure was used. Analysis failed to support the hypothesis of a main effect for treatment status on Physical Activity between the first and second assessments (Z=1.1283, R=.2592). Similar results were obtained using Hg change as the independent variable.

It was hypothesized that self-reported behavioral

indicators of Cognitive-Affective state would improve in the treatment group. A Wilcoxon Rank Sum Test procedure failed to demonstrate a change in Cognitive-Affective behavior (Z=1.7654, R=.0775) between the first and second assessments. Because the results could arguably be interpreted as failing to reject the possibility of a trend, the direction of the changes was examined. As displayed in Table 4-10 below change between the first and second assessments was in the opposite of the predicted direction. Direction of change between the second and third assessments was in the predicted direction. Analysis of change scores between the first and second assessments using a GLM procedure failed to demonstrate a main effect for Hg (F=0.40, R=0.5336).

Using sign of Hg change scores as the classificatory variable, a Wicoxon Rank Sum test failed to reveal statistically significant differences in self-reported Physical Activity (Z= -1.0235, p=.3061). Neither were there statistically significant differences in self-reported behavior in the Cognitive-Affective domain (Z= 0.2028, p=.8393).




Full Text
ACKNOWLEDGEMENTS
An elderly subject stood out for interest in his
performance. A man in his early twenties appeared
remarkably uninterested in his standing. A third subject
participated in the hope of helping others, despite
increasing debilitation. All three died prior to
completion of the study. Their names, and the names of the
32 other subjects are not listed out of respect for their
privacy. However, without their willingness to endure
testing despite the discomforts, and in some cases the
agonies, of their illness, this study would not have been
possible.
The following, undergraduates at the time, assisted
with test construction, test validation, data collection
and/or data entry: Bill Dorkowsky, Mike Reiter, Sheri
Scott, Katherine Cobb, Edward Suarez and Wylene Bhanji.
Without their help my wife, Liz, and child, Sarah, would
have suffered even worse neglect. Don Mars, M.D.
graciously persuaded his patients to participate as
controls. The American Psychological Association provided
a Dissertation Research Award which partially funded this
project.
My committee, beyond help regarding this project,
consists of those who were most important to my graduate


APPENDIX C
FREQUENCY ESTIMATION TASK
Table C-l. Encoding Phase
"Now I am going to read a list of [animals, grain & flour
products or means of transportation]. Please remember as
many
of the words as
possible because a
little
later on I
will
ask you to list
as many of them as
you can
II

Assessment 1
ANIMALS
Assessment 2
GRAIN & FLOUR
PRODUCTS
Assessment 3
MEANS OF
TRANSPORTATION
1
kangaroo
1
wheat
1
car
2
dog
2
oatmeal
2
subway
3
snake
3
barley
3
bicycle
4
elephant
4
cake
4
escalator
5
kangaroo
5
barley
5
subway
6
buffalo
6
corn
6
airplane
7
goldfish
7
waffle
7
canoe
8
kangaroo
8
barley
8
subway
9
tiger
9
pancake
9
stagecoach
10
dog
10
oatmeal
10
car
11
moose
11
rice
11
skateboard
12
kangaroo
12
barley
12
subway
13
gopher
13
bagel
13
helicopter
14
kangaroo
14
barley
14
subway
15
dog
15
oatmeal
15
car
139


161
Table 1-4. Raw Verbal Fluency and Self-Report Questionnaire
Data bv Assessment
Controlled Word
Association
CFL
FAS
FAS %
PRW
Assess
Id
1
2
2
3
1
38
38
55-59
29
2
38
48
85-89
33
3
29
23
15-19
26
4
18
8
<4
15
5
25
31
35-39
25
6
33
21
10-14
32
7
17
14
<4
18
8
25
28
25-29
26
9
23
46
85-89
34
10
41
42
80-84
42
11
47
41
75-79
48
12
16
13
5-9
12
13
29
16
>4
29
14
25
28
30-34
23
15
32
30
35-39
27
16
52
51
95-
49
17
21
17
20-24
16
101
23
26
10-14
26
102
43
47
85-89
42
103
28
40
80-84
36
104
22
23
50-54

105
23
40
75-79
40
106



107
43
54
>95
46
108
42
45
75-79
48
109
51

47
110
28
25
25-29
30
111
33
42
80-84
30
112
27
32
65-69
28
113
33
31
50-54
31
114
42
55
95-
51
115
48
49
90-94
44
116
29
26
20-24
35
117
50
60
>95

118
37
37
65-69
29
Cognitive-Affective and
Physical Behavior
Questionnaire
Physical Cognitive
1
2
3
1
2
3
13
17
13
9
8
8
15
13
12
3
1
4
5
4
6
7
5
6
3
5
14
6
6
8
4
7
7
4
5
4
6
7
5
1
3
1
14
14
14
7
8
8
14
14
12
9
9
8
12
12
11
8
8
8
12
14
14
7
8
8
15
14
10
8
8
8
3
3
1
5
6
4
7
4
3
6
3
3
3
3
4
4
4
2
10
8
9
5
4
5
3

1
4

4
12
10
10
6
6
5
12
11
11
5
6
5
5
5
9
7
5
4
12
8
11
6
7
5
8
8

8
7

0
0
2
1
2
1
6
3

8
6

14
14
11
5
7
6
5
7
5
4
6
4
10
13
9
5
7
3
14
13
11
8
7
7
14
17
17
8
9
8
11
9
9
6
8
6
10
6
3
8
8
6
13
13
15
8
8
8
2
2
2
4
5
4
14
11
14
6
14
6
8
6

4
6

14
12
14
8
8
8
Note: Assess = Assessment number


64
anticipated, combined with reasoning along the lines that
response to anemia may be quite idiosyncratic, led to the
decision to reclassify subjects based on Hg change scores.
Unfortunately, the small range of Hg change in most subjects
necessitated a choice between sacrificing either statistical
power or clinical relevance. Retaining statistical power,
subjects were first classified based on small distinctions
with unlikely and undemonstrated physiological significance.
Following this, compromising between sample size and
physiological significance, only subjects meeting the
criteria of Hg change scores of +1 g/dl were retained. CVLT
scores in subjects with clinically significant changes in Hg
status were then examined.
The post hoc hypothesis that reduction in anemia might
alter the pattern of learning was tested. The slopes of
learning curves for trials one through five of the CVLT were
calculated with SAS Regression Analysis procedures. A GLM
ANOVA then compared changes in slopes between groups.
Additional procedures were exploratory. Change scores were
calculated for all variables and the performance of specific
variables analyzed. Also, relationships between test
performance and questionnaire data were explored.
Statistical Assumptions
The basic assumptions underlying most analysis of
variance procedures include 1) normality and 2) homogeneity
of within-group variance. Analysis of variance procedures
are quite robust with regard to violation of the assumption


68
exception of ferritin, all variables necessary for
preplanned tests exhibited both reasonably good between
group equality of variance and were normally distributed.
As the effects of ferritin were judged to be nonlinearly
related to ferritin concentration, it was decided to use
ferritin as a classificatory variable.
To test the impact of the above violations of the
assumptions of parametric procedures, parametric and
nonparametric tests of between group baseline values were
calculated for all subjects and the results compared. As
displayed in Appendix E, even using an alpha of .05, only 2
tests out of 55 variables produced discrepant results. They
were CVLT variables that may be expected to be relatively
unstable: Short Delay Perseverations and Long Delay
Intrusions. This test suggested that the violations of
normality were unlikely to be of sufficient severity to
impact on the validity of SAS General Linear Model
parametric procedures.
The statistical characteristics of the distributions
and variance of all variables may be viewed in Appendix F.
Tables F-l and F-2.
There was one debatable outlier. Select analyses were
re-run following deletion of this subject. Examination of
the raw data revealed that at the first assessment the
subject had recalled 8 words on the first trial of the CVLT,
but only 3 on the fourth trial and 5 on the fifth trial.
Other data for this subject was within expected limits.


151
Table G-2. Means
and Standard
Deviations
at
Second
Assessment bv Group
Control Group
EPO
Treated
Group
Variable
N
Mean
Sd
N
Mean
Sd
Hemoglobin
18
9.23
1.81
17
9.59
1.84
Hematocrit
18
27.09
5.31
17
28.81
5.59
Ferritin
5
337.48
285.93
15
536.47
727.30
Assessment Interval 18
56.50
35.98
17
51.12
32.95
California Verbal
Learning Test
Trial 1
18
6.67
1.91
17
4.88
2.03
Standard Score
18
-0.78
0.88
17
-1.76
0.97
Trial 2
18
9.06
2.46
17
8.18
2.27
Trial 3
18
10.11
2.59
17
9.76
3.05
Trial 4
18
10.72
2.49
17
10.88
3.71
Trial 5
18
10.89
2.89
17
11.29
3.44
Standard Score
18
-1.50
1.34
17
-1.47
1.70
Sum CVLT
18
47.28
10.68
17
45.00
13.36
Standard Score
18
38.56
10.37
17
32.82
14.80
Clusters
18
12.78
9.72
17
13.53
9.84
Perseverations
18
3.61
2.77
17
4.00
3.79
Intrusions
18
1.50
2.20
17
2.06
2.84
List B
17
5.53
2.10
17
5.65
2.32
Standard Score
17
-0.88
1.27
17
-1.12
1.17
Short Delay
17
9.12
3.10
17
9.59
3.99
Standard Score
17
-1.12
1.17
17
-1.00
1.46
Clusters
17
3.88
3.08
17
3.88
2.80
Perseveration
17
0.59
0.80
17
0.41
0.80
Intrusions
17
0.47
0.51
17
0.53
1.01
Short Delay Cued
17
10.53
3.32
17
10.59
2.69
Standard Score
17
-0.94
1.09
17
-1.00
1.22
Perseverations
17
0.00
0.00
17
0.00
0.00
Intrusions
17
1.29
1.49
17
1.47
1.46
Long Delay
17
9.53
3.30
17
9.41
4.17
Standard Score
17
-1.24
1.09
17
-1.47
1.59
Clusters
17
4.53
3.48
17
4.24
2.80
Perseverations
17
0.35
0.79
17
0.12
0.33
Intrusions
17
0.82
1.29
17
1.00
1.17
Long Delay Cued
17
10.41
3.20
17
10.65
2.71
Standard Score
17
-1.24
1.30
17
-1.18
1.47
Perseverations
17
0.12
0.33
17
0.06
0.24
Intrusions
17
1.53
1.23
17
1.35
1.46
Recognition Hits
17
14.47
1.77
17
14.18
1.19
False Positives
from
Interference List
Shared Semantic
17
0.88
1.11
17
0.53
0.94


168
Heistadt, D.D., Marcus, M.L., & Abboud, F.M. (1987).
Experimental attempts to unmask the effects of neural
stimuli on cerebral blood flow. In M.J. Purves, (Ed.),
Cerebral Vascular Smooth Muscle and Its Control (pp. 142-
165). Amsterdam, Netherlands: Elsevier.
Hennessy, T.G., Stern, W.E., & Herrick, S.E. (1967).
Cerebellar Hemangioblastoma: Erythropoietic Activity by
Radio-iron Assay. Journal of Nuclear Medicine. 8, 601-
606.
Honig, A.S., & Oski, F.A. (1984). Solemnity: a clinical risk
index for iron deficient infants. Early Child Development
and Care. 16, 69-83.
Hori, K., Onoyama, K., Iseki, K., Fujimi, S., & Fujishima,
M. (1990). Hemodynamic and volume changes by recombinant
human erythropoietin (rHuEPO) in the treatment of anemic
hemodialysis. Clinical Nephrology. 33. 293-298.
Hunt, K.P., & Hodge, M.H. (1971). Category-item frequency
and category-name meaningfulness (m'): Taxonomic norms
for 84 categories. Psvchonomic Monograph Supplements. 4./
97-121.
Jackson, M., Warrington, E.K., Roe, C.J., & Baker, L.R.I.
(1987). Cognitive function in hemodialysis patients.
Clinical Nephrology. 27. 26-30.
Jacob, H.S., Eaton, J.W., & Yawata, Y. (1975). Shortened red
blood cell survival in uremic patients: beneficial and
deleterious effects of dialysis. Kidney International. 7,
S-139-14 3.
Johnson, D.L., & McGowan, R.J. (1983). Anemia and infant
behavior. Nutrition and Behavior. 1, 185-192.
Jonides, J., & Naveh-Benjamin, M. (1987). Estimating
Frequency of Occurrence. Journal of Experimental
Psychology: Learning. Memory, and Cognition. 13. 230-240.
Kaplan-DeNour, A., Shaltiel, J., & Czaczbes, J.W. (1968).
Emotional reactions of patients on chronic hemodialysis.
Psychosomatic Medicine. 300. 521-533.
Klein, W.L., Frederickson, E.D., Fennell, E.B., Peterson,
J.C., Guzman, S.H., & Tisher, C.C. (1989). Effects of
erythropoietin on neuropsychological function and
reported mood in the anemia of end stage renal disease.
American Federation for Clinical Research (Abstract).


30
appearance of symptoms. Even in cases of severe chronic
anemia (6 to 8 g per dl), the patient may be asymptomatic or
complain only of fatigue (Wintrobe et al., 1981).
Fatigue appears to be the most common symptom,
regardless of severity of the anemia. Comparing symptom
frequency in anemia characterized by Hg levels above or
below 8 g/dl, Dawson, Ogston and Fullerton (1969) reported
that in anemic subjects fatigue was present in 84.6% and
90.5%, respectively. In severe acute anemia, respiratory
and circulatory symptoms suggestive of possible hypoxia
(i.e., shortness of breath, "dizziness," faintness and
tachycardia) are common (Wintrobe et al., 1981).
Cognitive deficits have been documented in several of
the anemias (Davies-Jones, Preston, & Timperley, 1980;
Tucker, Sandstead, Penland, Dawson, & Milne, 1984).
However, previous studies have been confounded by the
failure to control for factors causing the anemia and
suspected of producing deleterious effects through
mechanisms unrelated to anemia. Nevertheless, existent
studies provide some relevant information. As it is the
most common, the best studied and the most relevant to the
current project (i.e., due to its possible appearance during
rEPO treatment), iron deficiency anemia will be emphasized.
Human studies examining the effects of iron
supplementation in deficient subjects have found increased
mental development scores in 15 month old infants, improved
attention, but not memory in 3-6 year olds, improved problem


67
The assumption of equality of variance between groups
was tested with the Folded F test, F'. The results, as
displayed in Appendix F, Table F-2, revealed the following
CVLT variables to fail the test of homogeneity of variance
at the .10 level: Sum of clusters, Sum of intrusions, and
Sum of perseverations for trials one through five, Short
Delay perseverations and intrusions, and all three CVLT
measures of false positives not on the interference list.
Other measures failing the test of homogeneity of variance
were: levels of processing recall of orthographically
processed stimuli, years of education and ferritin.
Many major variables violated the assumption of
normality to some degree, but exhibited acceptable
approximations of homogeneity of variance; therefore, they
were considered candidates for parametric analyses. These
variables included: CVLT Short Delay, Long Delay, Short
Delay Cued Recall, Long Delay Cued Recall, Recognition Hits,
Levels of processing Semantic recognition, estimation of
frequency of occurrence and associated free recall and both
scales from the Cognitive-Affective and Physical Behavior
Questionnaire. Please see Appendix F, Table F-2. The
Orthographic variable generated by the levels of processing
task was normally distributed but failed the test of
homogeneity of variance. Ferritin, Sum CVLT Clusters and
years of education were neither normally distributed nor had
equality of variance; therefore, parametric procedures were
considered unsuitable for their analysis. With the


121
become rare. Therefore, the study was conducted despite
imperfect conditions.
The paucity of changes between assessments provided
the opportunity to examine the alternate form reliability
of the test forms developed for this study. These tests
were the third form of the CVLT as well as the second and
third forms of the levels of processing and estimation of
frequency of occurrence tasks.x The CVLT III performed very
well and appeared to be essentially identical to the two
earlier forms. In contrast, the levels of processing and
estimation of frequency of occurrence alternate forms were
disappointing. The statistically significant correlation
between Cognitive-Affective self-report change scores and
Sum CVLT change scores, especially in the absence of a
significant relationship between Physical Activity and Sum
CVLT, supported the validity of behaviorally oriented
questionnaire data in the study of this population. This
is significant in that previous work with the Profile of
Mood States has been less than supportive of the validity
of self-report data in this population.
Conclusion
Many of the methodological deficiencies noted in the
literature were taken into account in the design of this
study. Thus, this study made use of a longitudinal design,
alternate test forms, a dialysis control group, sufficient
sample size, controlled for iron status, maintained a
consistent time of assessment with respect to dialysis and


122
excluded patients with uncontrolled hypertension.
Nevertheless, flaws severely limit the extent to which
useful conclusions may be drawn.
In the current study, Hg levels were both higher at
baseline and rose much less than in the pilot study. There
was significant intersubject variability in degree of
anemia and in etiology of renal disease. Despite
utilization of dialysis controls, there were big between
group differences in dialysis treatment modality. Although
the experimenter was blind to Hg level, rEPO treatment
status was usually known by subject and experimenter.
The present study produced absolutely no evidence that
rEPO treatment is associated with improvement in
neuropsychological functioning. Given the positive
correlation between rEPO mediated rise in Hg and
improvement in LTM in the Pilot study, the absence of
improvement in response to smaller increases in Hg might be
interpreted as indicating that most patients would benefit
from higher dosages of rEPO. However, the present results
provided data useful in a re-examination of the findings of
the Pilot study. The results of this re-examination
suggested the possibility that the initial findings may
have been spurious. More work is needed with Hg changes of
greater magnitude and lower initial values to speak to the
question of the neuropsychology of anemia.
Despite the absence of Hg change at levels of
demonstrated physiological significance, a possible trend


21
6 months of hemodialysis, revealed a very slight
nonsignificant improvement.
In hemodialysis subjects (n=29), impairment in
visuospatial memory as measured by the Block Design Learning
Test, in a study lacking a control group, failed to reach
statistical significance. However, duration of dialysis did
correlate with poor performance (English, Savage, Britton,
Ward, & Kerr, 1978). However, it is unclear whether the
putative deficit was related to nonverbal memory or to
visuospatial constructional ability. In addition, factors
such as hypertension, diabetes and age may have contributed
to the apparent negative relationship between duration of
dialysis and nonverbal memory performance.
Ziestat, Logue and McCarty (1980) reported a
significant correlation between years on hemodialysis and
both short term (r=-.39) and long term (r=-.38) visual
memory as measured by the WMS. However, again, years on
dialysis may have been confounded by factors such as
etiology of renal failure. In contrast, verbal memory
showed no relationship with years on hemodialysis in this
study. The apparent decline in visual memory may have been
secondary to the well researched decline in visuospatial
processing (Mings, 1987).
Conclusions Regarding Cognitive Deficits in Renal Disease
Renal disease differentially affects cognitive
functioning in a manner grossly consistent with that seen in
many systemic conditions. The degree of cognitive


150
Table G-l continued
Control Group EPO Treated Group
Variable
N
Mean
Sd
N
Mean
Sd
Long Delay
18
9.94
3.44
17
9.47
3.76
Standard Score
18
-0.89
1.02
17
-1.41
1.66
Clusters
18
4.72
3.03
17
4.12
3.16
Perseverations
18
0.33
0.49
17
0.24
0.56
Intrusions
18
0.22
0.73
17
0.76
1.25*
Long Delay Cued
18
11.06
3.51
17
10.82
3.34
Standard Score
18
-0.78
1.40
17
-1.24
1.56
Perseverations
18
0.00
0.00
17
0.00
0.00
Intrusions
18
0.61
1.20
17
1.53
2.43
Recognition Hits
18
14.61
1.61
17
15.00
0.94
False Positives
From
Interference List
Shared Category
18
0.94
0.87
17
0.65
0.79
Unrelated
18
0.39
0.70
17
0.35
0.79
Intrusions not on Interference
List
Shared Semantic
18
0.11
0.32
17
0.41
0.80
Shared Phonemic
18
0.33
0.59
17
0.82
1.47
No Relation
18
0.78
2.16
17
0.29
0.99
Controlled Word Association
C
17
11.53
3.84
17
10.53
4.45
F
17
11.59
4.03
17
10.00
3.77
L
17
12.29
3.70
17
9.41
4.11*
LOP Semantic
16
9.62
1.67
16
10.25
1.98
LOP Orthographic
16
7.69
2.15
16
7.37
2.22
LOP Phonemic
16
7.81
1.64
16
7.44
2.80
Frequency Estimation Task
Free Recall
18
3.61
1.42
17
3.71
1.31
Frequency Estimate
18
11.44
2.20
17
11.24
1.89
Cognitive-Affective and
Physical
Behavior
Questionnaire
Cognitive-Affect
18
6.06
2.01
17
5.82
2.19
Physical
18
9.56
4.42
17
8.88
4.77
Appetite
18
0.67
0.49
17
0.88
0.33
Post Hemodialysis
5
0.80
0.45
15
0.73
0.46
Note: = Significant difference at
LOP = Levels of Processing
05 level


45
(1986) reported that one of their ten patients developed
hypertensive encephalopathy. In addition, increased blood
clotting may potentially reduce circulation (Winearls et
al., 1986). Other side effects reported include anxiety,
lethargy, headache, body aches and, in 32% of subjects in
one study, iron deficiency (Delano, 1989). In addition,
concern has been raised that increasing hematocrit to normal
levels may compromise any existing renal function and reduce
the efficiency of dialysis (Koene & Frenken, 1990).
Curiously, nervous tissue is capable of producing EPO.
After renal carcinoma, cerebellar hemangioblastomas have
been reported to be the most frequent cause of elevated EPO
in association with neoplasms (Hennessy, Stern, & Herrick,
1967; Race, Finney, Mallams, & Balia, 1964; Waldmann, Levin,
& Baldwin, 1961). The proclivity of neural tissue to
produce EPO leads to speculation that EPO may normally be
produced by and have a function in the brain; however, there
is no evidence of EPO mediated neural effects in cancer
patients suffering from EPO secreting tumors. However,
physiological EPO levels are abnormally low for the degree
of anemia in renal failure, but often within the range found
in nonanemic populations. Therefore, EPO levels may be
raised dramatically during rEPO replacement therapy.
However, the suggestion that EPO might affect nervous tissue
is purely speculative.
Of more interest is the observation that rEPO treatment
may reduce sympathetic arousal. Cardiac output is increased


vi
APPENDICES
A DEVELOPMENT AND VALIDATION OF FORM III OF
THE CALIFORNIA VERBAL LEARNING TEST 124
B LEVELS OF PROCESSING VERSIONS II AND III . 131
C FREQUENCY ESTIMATION TASK 139
D COGNITIVE-AFFECTIVE AND PHYSICAL BEHAVIOR
QUESTIONNAIRE 141
E COMPARISON OF PARAMETRIC AND NONPARAMETRIC
TESTS OF BASELINE MEASURES BETWEEN GROUPS . 144
F NORMALITY AND HOMOGENEITY OF VARIANCE . .146
G GROUP MEANS AND STANDARD DEVIATIONS 149
H CHANGE SCORES 155
I RAW DATA BY SUBJECT 158
BIBLIOGRAPHY 162
BIOGRAPHICAL SKETCH 175


101
Sum CVLT observed between the first and second assessments
might be due to an increase in severity of uremia as
measured by blood urea nitrogen (Bun). In the hemodialysis
population, blood values were necessary for the day of the
first and second assessments. However, Bun levels were
unavailable for the dates in question. In the peritoneal
dialysis group, despite a high proportion of subjects having
been tested at home, BUN values were available on or near
the dates of the first and second assessments for 8
subjects. Although the power of the test was unacceptably
low due to the small number of subjects, a univariate
analysis of variance was performed. This procedure failed
to demonstrate a relationship between Sum CVLT and BUN
(F=0.77, p=.4139). However, there was no demonstrable
relationship between Hg and Sum CVLT in this subsample of
subjects (F=0.01, p=.9356).
Summary
Despite extensive post hoc analyses, no evidence of
improvement in memory, self-reported cognitive-affective or
physical functioning was demonstrated in the rEPO treatment
group relative to controls. What appeared to be a trend
towards a decline in Sum CVLT performance in rEPO treated
subjects was found to represent poorer performance on the
assessment prior to a drop in Hg. This phenomenon was
observed primarily in a subset of the controls, but was not
evident until subjects were reclassified based on direction
of Hg change between the first and second assessments.


69
Descriptive Statistics
Age
The treatment group had a mean age of 43.24 years
(sd=13.45) and consisted of 17 subjects. The mean age of
the 18 subjects in the control group was 48.22 years
(sd=14.19). A nonparametric Kruskal-Wallis test procedure
failed to demonstrate a statistically significant between
group difference (X2(l, n=35)=1.13, p<.2948).
Education
Mean years of education in the treatment group was
11.41 years (sd=3.02) and 11.82 years (sd=2.15) in the
control group. A Kruskal-Wallis test procedure failed to
reveal a significant statistical differences between groups
(X2(1, n=35)=0.23, pc.6325).
Sex
Overall, the subject population was equally divided by
sex with 17 males and 18 females. The treatment group
consisted of 7 males and 10 females. The control group
consisted of 10 males and 8 females. Thus, the control
group was 55.6% male and the treatment group was 41.2% male.
A Chi-Square procedure failed to demonstrate a statistically
significant difference (X2(l, n=35)= 0.724, p< 0.395).
Treatment Modality
In the control group 72.2% were on some form of
peritoneal dialysis (i.e., Peritoneal Dialysis, Continuous
Ambulatory Peritoneal Dialysis, or Continuous Cycler
Peritoneal Dialysis). Five subjects or 27.8% of the control


156
Control Group Treatment Group
Variable
Tm
N
Mean
Sd
N
Mean
Sd
Short Delay
Cluster 1
17
0.47
2.67
17
0.06
1.34
Cluster
2
15
-0.27
2.96
17
-0.12
1.17
Psv
1
17
0.35
0.79
17
0.41
2.50
Psv
2
15
-0.40
0.99
17
0.12
2.34
Intrus
1
17
0.29
0.69
17
-0.35
1.22
Intrus
2
15
0.33
0.72
17
0.35
1.17
SD Cued
1
17
-0.35
2.32
17
0.29
1.26
SD Cued
2
15
-0.87
3.14
17
-0.06
0.56
SS
1
17
-0.06
1.14
17
-0.35
2.06
SS
2
15
-0.47
1.73
17
0.18
1.74
Psv
1
17
0.00
0.00
17
-0.12
1.27
Psv
2
15
0.07
0.26
17
0.00
0.87
Intrus
1
17
0.82
1.67
17
-0.29
0.77
Intrus
2
15
-0.33
1.68
17
0.00
0.00
Long Del
1
17
-0.47
2.15
17
0.82
1.67
Long Del
2
15
-0.60
2.50
17
-0.18
1.07
SS
1
17
-0.29
0.92
17
-0.06
2.90
SS
2
15
-0.40
1.35
17
0.24
2.36
Cluster
1
17
-0.18
1.88
17
-0.06
1.34
Cluster
2
15
-0.80
3.08
17
0.24
1.09
Psv
1
17
0.00
0.79
17
0.12
3.06
Psv
2
15
-0.13
0.99
17
0.12
2.67
Intrus
1
17
0.59
1.46
17
-0.12
0.70
Intrus
2
15
0.13
1.51
17
0.12
0.70
LD Cued
1
17
-0.53
2.96
17
0.24
1.48
LD Cued
2
15
-0.47
3.54
17
-0.12
1.32
Psv
1
17
0.12
0.33
17
-0.18
2.38
Psv
2
15
-0.13
0.35
17
0.18
1.29
Intrus
1
17
0.94
1.30
17
0.06
0.24
Intrus
2
15
-0.13
1.73
17
-0.06
0.24
Recog
1
17
-0.12
1.93
17
-0.18
1.91
Recog
2
15
-0.60
1.99
17
0.06
1.14
False +
1
17
0.88
3.10
17
0.59
1.50
False +
2
15
-0.47
2.50
17
0.12
3.04
Verbal Fluency
1
16
5.06
6.20
17
-0.82
1.33 a
2
14
-2.36
5.60
17
-0.59
2.92
Levels of
Semantic
Processing
1 15
0.60
2.03
17
-0.82
8.65
Semantic
2
14
-0.79
2.42
17
-0.65
7.65
Orthograph
1
15
-1.60
2.75
17
0.13
2.00
Orthograph
2
14
0.79
3.56
17
-0.19
1.87
Phonemic
1
15
0.87
2.72
16
-0.50
2.61
Phonemic
2
14
-1.07
2.09
16
0.25
2.46


15
better than any other measure in the study (Fennell, et
al.,1987).
In contrast, Souheaver, Ryan and DeWolfe (1982) found
spared auditory attention skills, but on a different task,
the Seashore Rhythm test. In the Heilman, Moyer, Melendez,
Schwartz and Miller (1975) study, simple attention-
concentration as measured by Digit Span was within normal
limits (mean subscale score of 10.04 +2.18) and consistent
with verbal intelligence as measured by the Vocabulary
subtest (mean 10.00). However, attention was significantly
lower in renal patients (n=24) than in controls (n=12).
Overall, it may be concluded that attentional deficits are
common in uremia.
Uremia appears to depress activation and arousal
(Trompeter, Polinsky, Andreoli, & Fennell, 1986). This
underarousal may interfere with attention and increase
distractibility (Heilman, Moyer, Melendez, Schwartz, &
Miller, 1975). However, reduced level of activation is
unlikely to be the only mechanism involved. Assessments at
varying time intervals after hemodialysis sessions revealed
no change in Digit Span performance, despite changes in
speed and reaction time (Lewis, O'Neill, Dustman & Beck,
1980).
Renal disease also appears to negatively affect more
complex attentional and executive functioning such as the
ability to rapidly generate and/or shift sets. Performance
on Trails B, a task requiring maintenance of and alternation


166
Fennell, R.S., Fennell, E.B., Carter, R.L., Mings, E.L.,
Klausner, A.B., & Hurst, J.R. (1989). Correlations
between performance on neuropsychological tests of
children with chronic renal disease. Unpublished.
Fennell, R.S., Fennell, E.B., Carter, R.L., Mings, E.L.,
Klausner, A.B., & Hurst, J.R. (1990a). Association
between renal function and cognition in childhood chronic
renal failure. Pediatric Nephrology. 4., 16-20.
Fennell, R.S., Fennell, E.B., Carter, R.L., Mings, E.L.,
Klausner, A.B., & Hurst, J.R. (1990b). A longitudinal
study of the cognitive function of children with renal
failure. Pediatric Nephrology. 4, 11-15.
Fennell, R.S., Fennell, E.B., Mings, E. L., Klausner, A. B.,
Hurst, J.R., & Carter, R.L. (1987). Cognitive performance
in children with renal failure as affected by renal
function and therapeutic modalities. In K. Murakami, T.
Kitagawa, K., Yabuta, & T. Sakai (Eds.), Recent Advances
in Pediatric Nephrology (pp. 615-618). New York, New
York: Excerptica Medica.
Fennell, E.B., Fennell, R.S., Mings, E., & Morris, M.K.
(1986). The effects of various modes of therapy for end
stage renal disease on cognitive performance in a
pediatric population: A preliminary report. International
Journal of Pediatric Nephrology. 7, 107-112.
Fennell, R.S., & Rasbury, W.C. (1980). Cognitive functioning
of identical twins discordant for Prune Belly Syndrome
and end-stage renal disease. International Journal of
Pediatric Nephrology. 1, 234-239.
Fennell, R.S., Rasbury, W.C., Fennell, E.B., & Morris, M.K.
(1984). Effects of kidney transplantation on cognitive
performance in a pediatric population. Pediatrics. 74.,
273-278.
Fishman, D.B., & Schneider, C.J. (1972). Predicting
emotional adjustment in home dialysis patients and their
relatives. Journal of Chronic Diseases. 25, 99-109.
Freeman, C.W., Sherrard, D.J., Carlsyn, D.A., & Paige, A.B.
(1980). Psychological assessment of renal dialysis
patients using standard psychometric techniques. Journal
of Counseling and Clinical Psychology. 48, 537-539.
Fried, R. (1987). The Hvperventillation Syndrome: Research
and Clinical Treatment. Baltimore, Maryland: Johns
Hopkins University Press.
Gilli, P., & DeBastiani, P. (1983). Cognitive function and
dialysis treatment. Clinical Neurology. 16, 188-192.


APPENDIX I
RAW DATA BY SUBJECT
Table 1-1. Raw Demographic Data
Hemoglobin by Assessment
Id
Acre
Ed
Sex
Trt
Group
1
2
3
1
24.2
14
M
Hemo
rEPO
7.90
9.00
10.80
2
31.2
13
F
Hemo
rEPO
9.50
10.60
10.10
3
53.7
12
M
Hemo
rEPO
9.00
9.30
9.30
4
29.6
8
F
Hemo
rEPO
7.54
8.10
8.00
5
37.1
14
F
Hemo
rEPO
5.60
6.60
6.90
6
45.2
10
F
Hemo
rEPO
8.30
14.20
8.70
7
43.0
12
F
Hemo
rEPO
8.10
7.90
8.60
8
51.0
12
F
Hemo
rEPO
7.55
8.85
7.25
9
49.3
12
F
Hemo
rEPO
8.50
9.10
9.40
10
25.8
12
M
Hemo
rEPO
9.60
9.75
9.65
11
26.5
12
F
Hemo
rEPO
7.10
10.50
11.10
12
56.3
7
M
Hemo
rEPO
7.20
6.95
7.90
13
46.2
13
M
Hemo
rEPO
11.30
10.60
10.84
14
57.1
12
M
Hemo
rEPO
10.50
10.70
10.81
15
48.0
12
M
Hemo
rEPO
10.20
12.00
12.25
16
37.0
16
F
Perit
rEPO
8.10
10.10
10.30
17
73.8
3
F
Hemo
rEPO
7.80
8.80
10.30
101
57.1
12
F
Hemo
Con
6.90
8.30
10.20
102
48.0
12
M
Hemo
Con
9.40
10.00
11.70
103
34.8
8
F
Perit
Con
5.27
6.48
6.90
104
71.5
8
M
Hemo
Con
9.70
9.61

105
54.7
12
M
Hemo
Con
13.90
12.60
10.50
106
60.5
15
M
Hemo
Con
9.25
6.30

107
43.5
13
M
Perit
Con
12.30
11.20
12.30
108
23.9
12
M
Perit
Con
8.40
9.40
9.90
109
43.5
14
M
Perit
Con
10.90
11.10
11.60
110
57.3
10
F
Perit
Con
9.50
10.00
10.20
111
48.1
14
M
Perit
Con
9.60
9.40
9.70
112
70.0
10
M
Perit
Con
10.71
10.60
8.96
113
65.8
12
F
Perit
Con
9.80
8.10
9.10
114
32.9
12
F
Perit
Con
9.90
11.60
11.60
115
33.9
14
F
Perit
Con
8.90
7.30
6.00
116
29.4
12
F
Perit
Con
8.80
9.10
9.10
117
38.5
8
F
Perit
Con
7.30
8.07

118
54.5
14
M
Perit
Con
9.40
7.00
6.60
Note:
Ed =
Years
of
Education; Trt
= Treatment; Con

Control; Hemo = Hemodialysis;
Perit
= Peritoneal Dialysis
158


I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
filter Cb <~=X/<\/yviLL'
Eileen B. Fennell, Chairperson
Professor of Clinical and Health Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
¡mjMJLI a
/Russell M. Baue^f
Associate Professor of Clinical and Health
Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
A.
Hugh C.
Professo
of Clinical and Health Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
S3 ^4^
Thomas B Fast
Professor of Dentistry
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Robrt^S. Fennell, III
Professor Pharmacology and Therapeutics


169
Koene, R.A.P., & Frenken, L.A.M. (1990). Does treatment of
predialysis patients with recombinant human
erythropoietin compromise renal function? In R.M.
Schaefer, W.H. Horl, & A. Heidland (Eds.), Erythropoietin
in the 90s. New York, New York: Karger.
Lee, G.R., Wintrobe, M.M., & Bunn, H.F. (1980). Iron-
deficiency anemia and the sideroblastic anemias. In K.J.
Isselbacher, R.D. Adams, E. Braunwald, R.G. Petersdorf, &
J.D. Wilson (Eds.), Harrison's Principles of Internal
Medicine (9th ed.). (pp. 1514-1518). New York, New York:
McGraw Hill.
Lewis, E.G., O'Neill, W.M., Dustman, R.E., & Beck, E.C.
(1980). Temporal effects of hemodialysis on measures of
neural efficiency. Kidney International. 17, 357-363.
Lezak, M. D. (1983). Neuropsychological Assessment. (2nd
ed.). New York, New York: Oxford U. Press.
Lozoff, B. (1989). Nutrition and Behavior. American
Psychologist. 44. 231-236.
Lundin, A.P., Delano, B.G., & Quinn-Cefaro, R. (1990).
Perspectives on the improvement of quality of life with
epoetin alfa therapy. Pharmacotherapy. 10, 228-268.
Marshall, J.R. (1979). Neuropsychiatric aspects of renal
failure. Journal of Clinical Psychiatry. 40. 81-85.
Massaro T.F., & Widmayer, P. (1981). The effect of iron
deficiency on cognitive performance in the rat. American
Journal of Clinical Nutrition. 34, 864-870.
Mayer, G., Thum, J., Cada, E.M., Stummvoll, H.K., & Graf, H.
(1988). Working capacity is increased following
recombinant human erythropoietin treatment. Kidney
International. 34. 525-528.
McCarty, S.M., Logue, P.E., Power, D.G., Ziesat, H.A., &
Rosenstiel, A.K. (1980). Alternate form reliability and
age-related scores for Russell's Revised Wechsler Memory
Scale. Journal of Consulting and Clinical Psychology. 48,
296-298.
McDaniel, J.W. (1971). Metabolic and CNS correlates of
cognitive dysfunction with renal failure.
Psychophysiology. 8., 704-713.
Miller, G.A. (1956). The magical number seven, plus or minus
two: Some limits on our capacity to process information.
Psychological Review. 63. 81-97.


BIOGRAPHICAL SKETCH
Wayne L. Klein decided to become a psychologist in
adolescence after giving up on religion and philosophy. As
a college sophomore, academic psychology paled in comparison
to human potential psychology; therefore, in 1971, he
dropped out of SUNY at Albany and spent three months
hitchhiking to Berkeley, California where he spent a seminal
year emersed in Esalen groups, training at the Gestalt
Institute of San Francisco and working in the Berkeley Free
Clinic. After returning to SUNY at Albany to complete
liberal arts requirements, he transferred to SUNY Empire
State College, a university without walls program, and
obtained a B.S. in counselling psychology. Over the next
nine years he worked with a variety of populations in
diverse settings as a counselor and instructor. During that
time he also obtained a B.S. in biology by taking tests
(Regents College), took graduate courses in biopsychology
and earned an M.S. in educational psychology (SUNY at
Albany).
Spurred by interest in cognition-behavior-brain
relationships, he entered the University of Florida Clinical
and Health Psychology doctoral program. After flirting with
psychophysiology, he focused on clinical neuropsychology and
pursued interrelated research interests in the
176


APPENDIX A
DEVELOPMENT AND VALIDATION OF FORM III OF THE CALIFORNIA
VERBAL LEARNING TEST
Introduction
In the pilot study, repeat memory testing with Form I
of the CVLT revealed that some adult hemodialysis subjects
recalled the word lists after several months. Therefore,
alternate forms were considered desirable in the present
study. Form II of the CVLT was recently developed by
Delis, McKee, Massman, Kramer and Gettman (1990). The
construction and validation of a third parallel form will
be described below.
Test Construction
A third form of the CVLT was created for use in this
study. Following the procedures used in the development of
Form I of the CVLT, three criteria were used in selecting
Form III target words. Two of these criteria pertain to
frequency of use of the words and the third category
attempts to produce category members of equivalent
typicality (Delis, Kramer, Kaplan, & Ober, 1987).
The CVLT consists of two lists of 16 words that are
balanced for frequency of appearance in English reading
(Thorndike & Lorge, 1944; Carroll, Davies, & Richman,
1971). Both lists are composed of words belonging to four
124


CHAPTER 4
RESULTS
Overview of Analyses
Prior to analysis, the data was inspected for outliers
and tested for the assumption of normality using qualitative
and quantitative tests available through the SAS Univariate
procedure. Hemoglobin and Sum CVLT, the primary independent
and dependent variables, respectively, exhibited good
approximations of normal distributions and homogeneity of
variance. In contrast, the distributions of most other
variables deviated from normality to some degree. For
variables failing the Shapiro-Wilks W test of normality, the
ratio of the standard deviation to the mean was computed.
Variables with ratios greater than .25 were further examined
quantitatively and graphically (Schlotzhauer & Littell,
1987). Procedures such as SAS General Linear Model are
relatively robust with regard to violation of normality;
however, great caution is required in interpreting analyses
in which variables also violate the assumption of
homogeneity of variance.
The assumption of homogeneity of variance between
groups was tested with the Folded F test. With the
exception of ferritin, all variables necessary for testing
of the secondary hypotheses exhibited reasonably good
61


128
Results
The variable of greatest interest was Sum CVLT. Sum
CVLT is likely to be the most stable measure produced by
the CVLT. As displayed in Table A-3, Forms I and III
produced identical means of 65.1 and a correlation
coefficient of r = .69. Trial 1, a measure of STM,
produced
means of 8.7
and
9.7
for
Forms I
and III
TABLE
A-3
. CVLT
Inter
-Form Reliability
CVLT
III
CVLT I
CVLT I &
Variable
r
Mean
sd
Mean
sd
CVLT Hr
SUM
.69b
65.1
6.2
65.1
4.9
. 84c
T1
.70c
9.7
2.2
8.7
1.7
.54c
T2
.28
12.5
1.9
12.6
1.3
T3
.63b
13.3
1.8
13.8
1.4
T4
.31
14.3
1.2
15.0
1.1
T5
.10
15.1
1.0
15.2
1.2
.75c
List
B
. 56a
8.9
1.9
9.0
1.9
.31a
SD
.63b
14.0
1.6
14.7
1.0
. 82c
Cued
.51a
14.3
1.4
15.2
0.8
.76c
LongDel
.32
14.1
1.3
15.1
1.0
. 80c
LCued
.14
14.9
0.8
15.3
0.7
.79c
Note:
a p<.05
b p<.
01
c
p<.
001
SUM
= Sum
CVLT,
List B =
Interference list;
SD
= short
delay
free recall
.; CUED
= short
delay cued
recall, LONGDEL = long delay free recall;
LCUED = long delay cued recall


APPENDIX F
NORMALITY AND HOMOGENEITY OF VARIANCE
Table F-l. Indices of Normality and Characteristics of the
Distributions
Shapiro-Wilk Test
Variable
W:Normal
Prob Sd/M
Skew Kurtosis
California Verbal
Trial One
Learning
0.956
Test
.010
0.343
.084
-.240
Trial Two
0.959
.016
0.293
.454
.213
Trial Three
0.969
. 120
-
-.189
-.261
Trial Four
0.955
.008
0.282
-.496
.051
Trial Five
0.953
.004
0.272
-.354
-.310
Sum CVLT
0.976
.330

-.160
-.337
Sum Clusters
0.850
.001
0.746
1.631
3.003
Slope
0.994
.997
-
-.141
1.192
Intercept
0.987
.861

-.006
.083
Interference List
0.946
.001
0.372
. 162
-.459
Short Delay
29.520
.0001
0.340
-.392
-.081
Short Delay Cued
0.953
.005
0.260
-.547
. 029
Long Delay
0.951
.003
0.368
-.540
-.078
Long Delay Cued
0.950
.003
0.282
-.594
.180
CVLT Recognition Measures
Correct Positive: 0.741
.001
0.131
-3.206
17.983
False Positive On
No Relation
List B:
0.646
.001
1.681
1.688
2.112
Semantic Shared
0.808
.001
1.027
.668
-0.271
False Positive Not
No Relation
on List
0.501
B:
.001
2.308
4.377
26.366
Semantic Shared
0.561
.001
1.967
1.860
2.411
Phonemic Shared
0.611
.001
1.799
2.324
5.538
Verbal Fluency
Assessment 1
0.975
.287
-
0.362
-.122
Assessment 2
0.965
.062
0.420
0.166
-.402
Assessment 3
0.961
.026
0.427
0.382
-.533
LOP Semantic
0.909
.001
0.170
-.745
-.020
LOP Phonological
0.969
.133
-
-.393
.724
LOP Orthographic
0.962
.045
0.333
.192
-.464
146


41
moved from below normal to within normal limits; however,
this subject exhibited limited response to rEPO. The other
three treated subjects actually were slower post-test. The
greatest improvement was seen in one of the two controls,
who moved from clearly below normal to within normal limits.
Likewise, on Trails B, the greatest improvement was in the
treated subject who had only limited response to rEPO and
the second greatest improvement was in a control. Three of
the four treated, responsive subjects actually performed
more poorly on the second testing. These findings contrast
with the report of Wolcott, Schweitzer and Nissenson (1989)
that Trails improved in treated patients.
Correlation between improved performance on the Sum of
five trials on the California Verbal Learning Test (Sum
CVLT) supra-span word list learning task and improvement in
Hg was r=0.796 (p=.0324). The correlation between
improvement in Sum CVLT and rise in hematocrit was r=0.741
(p=.0568). Neither the first two trials of the five trial
sequence nor the once presented interference list showed
improvement. Improvement was observed primarily in the last
two trials. Single and double presentations of supra-span
lists with immediate recall are highly dependent on STM (T.
White, personal communication, June 12, 1990). They may be
considered only weak measures of LTM. Performance on the
fourth and fifth presentation is likely to be highly
dependent on the degree to which the subject has organized
and elaborated the list of target words. Thus, the improved


167
Ginn, H.E., (1975). Neurobehavioral dysfunction in uremia.
Kidney International. 7(S), 217-221.
Ginn, H.E., Teschan, P.E., Bourne, J.R., Hamel, B., Ward,
J.W., Vaugh, W.K., & Nunnally, J.C. (1978).
Neurobehavioral and clinical responses to hemodialysis.
Trans. Am Soc Artif Intern Organs, 24. 376-378.
Ginn, H.E., Teschan, P.E., Walker, P.J., Bourne, J.R.,
Fristoe, M., Ward, J.W., McLain, L.W., Johnston, H.B., &
Hamel, B. (1975). Neurotoxicity in Uremia. Kidney
International. , S357-S360.
Goldberg, A.P., Geltman, E.M., Gavin, J.R, 3rd., Carney,
R.M., Hagberg, J.M., Delmez, J.A., Naumovich, A.,
Oldfield, M.H., & Harter, H.R. (1986). Exercise training
reduces coronary risk and effectively rehabilitates
hemodialysis patients. Nephron. 42. 311-316.
Groner J.A., Holtzman, N.A., Charney, E., & Mellits, E.D.
(1986). A randomized trial of oral iron on tests of
short-term memory and attention span in young pregnant
women. Journal of Adolescent Health Care. 7, 44-48.
Gutman, R.A., Stead, W.W., & Robinson, R.R. (1981). Physical
activity and employment status of patients on maintenance
hemodialysis. New England Journal of Medicine. 304. 309-
313.
Hagberg, B. (1974). A prospective study of patients in
chronic hemodialysis-III: Predictive value of
intelligence, cognitive deficit, and ego defense
structures in rehabilitation. Journal of Psychosocial
Research. 18. 151-160.
Hallgren, B., & Sourander, P. (1958). The effect of age on
the nonheme iron in the human brain. Journal of
Neurochemistrv. 2, 41-51.
Hart, R.P., Pederson, J.A., Czerwinski, A.W., & Adams, R.L.
(1983). Chronic renal failure, dialysis and
neuropsychological function. Journal of Clinical
Neuropsychology. 5, 301-312.
Hasher. L., & Zacks, R.T. (1984). Automatic Processing of
Fundamental Information. American Psychologist. 39. 1372-
1388.
Heilman, K.M., Moyer, R.S., Melendez, F., Schwartz, H.D., &
Miller, B.D. (1975). A memory deficit in uremic
encephalopathy. Journal of Neurological Science. 26, 245-
249.


APPENDIX B
LEVELS OF PROCESSING VERSIONS II AND III
Table B-l. Levels of Processing Version II Encoding
Questions
Visual Stimuli
Ouestion to be processed
1.
ALTAR
Is this printed in upper case
letters?
2.
ear
Is this part of the body?
3.
sand
Does this rhyme with match?
4.
swine
Is this a type of furniture?
5.
smell
Does this rhyme with sell?
6.
cold
Is this printed in upper case
letters?
7.
sky
Is this printed in upper case
letters?
8.
country
Is this a type of music?
9.
priest
Does this rhyme with fall?
10.
joy
Does this rhyme with boy?
11.
wolf
Is this a type of flower?
12.
BURN
Is this printed in upper case
letters?
13.
silver
Is this a type of metal?
14.
seat
Is this a type of animal?
15.
trot
Does this rhyme with pot?
16.
shoot
Is this printed in upper case
letters?
17.
rope
Does this rhyme with hour?
18.
PEN
Is this printed in upper case
letters?
19.
CHEW
Is this printed in upper case
letters?
131


84
Regression procedures were used to generate slopes of the
learning curves for trials one through five for each
administration of the CVLT.
A GLM procedure analyzing change scores between the
first and second assessments failed to demonstrate a
relationship between Hg change and change in slope (F=0.08,
p=.7766). An identical analysis examining change scores
between the second and third assessments produced similar
results (F=0.00, p=.9583). A GLM univariate analysis of the
effect of treatment status on slope change scores between
the first and second assessments failed to reach statistical
significance (F= 3.67, p=.0651). Controlling for the
increased probability of a Type I error as a result of
multiple analyses, the finding can not be considered
suggestive of a trend. A Wilcoxon Rank Sum Test produced
similar results (Z= 1.83, p=.0669). Further analysis
revealed the change to be primarily in the treatment group,
but in the opposite of the expected direction. In the
treatment group, mean slope declined -0.84 (sd=1.59), and in
the control group slope improved 0.15 (sd=1.27) between the
first and second assessments. A GLM univariate analysis of
the effect of treatment group status on slope change scores
between the second and third assessments failed to suggest a
relationship (F=1.30, p=.2628).
Despite extensive analyses, absolutely no evidence was
found that even mildly suggested improvement in learning
curve in association with improvement in Hg status.


42
performance appears to have been due to enhanced LTM. This
is supported by the lack of improvement on the digit span
subtest of the Wechsler Intelligence Scale Revised.
However, the Logical Memory subtest of the WMS showed no
change.
Logical Memory is a single trial task and, because, as
the name implies, it is already logically organized, it
might be expected to less vigorously discriminate between
sparse and elaborate processing. Improved performance was
observed on the CVLT only after the third presentation.
Multitrial supra-span tasks require more active cognitive
processing than does the single presentation of a paragraph.
Given the hypothesized increase in general vitality, the
results may be explained by a greater proclivity, post
treatment, for effortful cognitive activity such as the kind
required to organize and elaboratively rehearse a shopping
list.
No improvement in immediate or delayed visual memory
was observed based on the Visual Reproduction subtest of the
WMS. This test involves only one presentation of the
stimuli; therefore, may not be less sensitive than Sum CVLT
to changes in degree of cognitive processing. However, the
lack of improvement is also consistent with the suggestion
that visual-motor integration skills are permanently
affected by renal disease, at least if age of onset was
during childhood (Mings, 1987).


75
three in the rEPO treated group which was statistically
significant (F(l, 33)= 4.74, p=.037).
In the time lag between initiation of rEPO and the
first assessment mean Hg levels rose 1.07 g/dl (sd=1.90)
from an initial level of 7.39 g/dl (sd=1.94). The increase
of 1.08 g/dl in the treatment group (n=17) between
assessments one and three was 50% of the total increase of
2.15 g/dl (sd=1.23) in Hg over the course of rEPO therapy.
However, this total increase of 2.15 g/dl was only 39% of
the increase obtained in the pilot study. Please see Tables
4-1 and 4-2 for of Hg levels and change scores,
respectively.
For descriptive purposes subjects were classified using
the following system: very severe anemia < 6 g/dl Hg, severe
anemic 6-8 g/dl, moderate anemia 8-10 g/dl, mild anemia
10 12 g/dl, > 12 g/dl normal. As illustrated in Table 4-
3, between the first and second assessments seven treated
subjects showed an improvement in category and one declined.
In the controls four rose and three fell. Between the
second and third assessments change appeared to be random.
Ferritin
Because of the tendency of rEPO treatment to draw down
iron stores ferritin levels were assayed more often in rEPO
treated subjects; therefore, baseline ferritin levels were
available for 94% (n=16) of the treatment group but only 33%
(n=6) of the controls. At baseline mean treatment group
(n=16) ferritin level was 545.93 ng/ml (sd=734.50) and mean


152
Table G-2 continued
Control Group EPO Treated Group
Variable N
Mean
Sd
N
Mean
Sd
Intrusions not on
Interference
List
Shared Semantic
17
0.35
0.61
17
0.41
0.87
Shared Phonemic
17
0.59
0.80
17
0.76
1.35
No Relation
17
0.29
0.69
17
0.41
0.71
Verbal Fluency
F
16
14.12
5.15
17
11.41
4.78
A
16
10.12
3.81
17
7.29
4.92
S
16
15.25
5.08
17
10.41
5.23
LOP Semantic
15
10.27
1.75
16
10.37
1.89
LOP Orthographic
15
6.20
2.11
16
6.88
2.55
LOP Phonemic
15
8.73
1.94
16
9.06
2.77
Frequency Estimation Task
Free Recall
17
4.24
1.48
17
3.06
1.39
Frequency Estimate
17
10.76
3.09
17
10.18
3.34
Cognitive-Affective
and
Physical
Behavior
Questionnaire
Cognitive-Affective
18
7.00
2.35
16
5.75
2.32
Physical
18
8.78
4.60
16
9.31
4.73
Appetite
18
0.56
0.51
16
0.94
0.25
Post Hemodialysis
6
0.50
0.55
15
0.60
0.51
Note: LOP = Levels of Processing


107
exercise and improved appetite, all previously attributed
to uremia. Despite predictions, compared to these
findings, pilot work with small numbers of rEPO treated
renal patients has been disappointing in that there have
been few improvements in neuropsychological functioning, no
consistently reported improvements and failures to
replicate. In our pilot work (Klein et al., 1989), the
only statistically significant change was in LTM as
measured by multitrial supra-span list learning. Despite
the danger of a Type I error posed by multiple analyses,
credence was placed in this finding because in the past
multitrial list learning had proven particularly sensitive
to treatment changes in renal disease. Therefore, the
present study sought to replicate and extend this finding.
Data analysis focused on change between the first and
second assessments because preliminary analyses revealed no
change in Hg between the second and third assessments.
Univariate analysis of variance of Sum CVLT change scores
failed to confirm either the hypothesized main effect for
Hg (F=3.25, p=0.0805) or a main effect for treatment status
(F=3.22, p=.0818). Combining treatment status and Hg level
in the same model eliminated the putative trend. After
discussing possible reasons for the discrepancy between the
pilot and the present study, post hoc analyses utilizing Hg
change as a classificatory variable will be discussed.
Examination of the literature on the impact of Hg
levels on cognitive function suggested that the Hg change


133
Table
B-2. Multiple Choice
Retrieval
Ouestions
Version II
Multiple Choice
Word List
Processing
Level
1.
seat
coin
maple
nail
SEM-
2.
clip
wood
chew
fern
0RTH+
3.
globe
sand
clamp
scar
PHON-
4.
rug
haze
gift
hire
PH0N+
5.
grass
foot
mast
sheet
ORTH-
6.
tear
country
page
horse
SEM+
7.
joy
cream
oak
fear
PH0N+
8.
path
coal
tie
tent
SEM-
9.
dandelion
bat
ghost
lap
SEM+
10.
moss
altar
bed
cart
0RTH+
11.
trout
egg
doll
swim
PHON-
12.
cable
paste
shoot
cup
ORTH-
13.
band
cord
safe
tune
ORTH+
14.
fur
ear
twig
mate
SEM+
15.
roll
chin
pork
lie
PH0N+
16.
trumpet
key
steak
purple
SEM-
17.
plot
sky
weed
cage
ORTH-
18.
crumb
limb
rope
ash
PHON-
19.
wheel
hide
spade
sheep
SEM+
20.
banana
leaf
pen
nob
ORTH+
21.
tube
roam
spear
art
ORTH-
22.
bullet
smell
cat
oar
PH0N+
23.
bone
pack
rock
wolf
SEM-
24.
prison
corn
book
mire
PHON-
25.
juice
string
silver
acre
SEM+


115
this increased activation on cognitive function is unknown;
however, the curvilinear relationship between performance
and level of activation is well known. Since underarousal
has been suggested as a component of some aspects of
neuropsychological dysfunction in uremia, uremic subjects
may be especially vulnerable to reduced activation
secondary to a reduction in degree of anemia. However,
contrary to this line of reasoning, Hg changes were seldom
in the range and of the magnitude likely to affect degree
of activation.
The Controlled Word Association Test is affected by
degree of activation, at least in grossly neurologically
impaired populations. Superficially, word generation from
within phonemic categories appeared to follow a pattern
similar to that observed for Sum CVLT in that treatment
group performance was relatively uniform across trials and
control group performance was lower at the first
assessment. However, controlling for multiple tests, this
finding failed to reach statistical significance. More
important, unlike Sum CVLT, this variable no longer
exhibited the above mentioned pattern when sign of Hg
change was used for classificatory purposes. In addition,
the self-report questionnaire, although not designed for
this purpose, failed to suggest alteration in degree of
activation.
Examination of the pattern of mean Sum CVLT change
scores led to the suggestion that the apparent decline was


91
Secondary Hypotheses
Levels of Processing
Treatment status was insignificant in explaining
variability in performance on any portion of the Levels of
Processing Task, including the Semantic recall measure
(F=0.43, p=.5164). Mean change scores are listed in
Appendix H. Re-analysis using direction of Hg change as a
classificatory variable produced similar results (F=2.12,
P=.1564). This analysis would have been considered a
secondary hypothesis if there had been a positive
association between Hg and Sum CVLT.
Estimation of Frequency of Occurrence
As displayed in Appendix H, there were no significant
differences in Frequency Estimation performance between
groups or assessments. Parenthetically, LTM, as measured by
free recall of the single presentation target word list,
improved slightly more in the treatment group than in the
control group; however, this failed to reach significance
controlling for multiple analyses.
Quality of Life
Quality of life is operationally defined here in terms
of the items on the Cognitive-Affective and Physical
Behavior Questionnaire and is divided into cognitive-
affective and overt physical behavior domains. It was
hypothesized that self-reported physical vitality, as
measured by the self-report questionnaire, would improve in
response to treatment with rEPO.


10
of processing, with semantic processing occurring at greater
depth than phonemic processing. However, research in other
areas has forced revisions in the original levels of
processing models (Cermak, 1982).
To account for studies demonstrating elaborateness of
processing to be more salient than type of processing (e.g.,
semantic vs. phonological), the levels of processing model
has been modified to emphasize extent rather than depth of
encoding (Craik & Tulving, 1975). Confounding all models,
other work has shown that shallowly processed information
may at times interfere with and even outlast more deeply
processed material (Cermak, 1982). In addition, phonemic
processing may result in recall superior to semantic
processing when retrieval is phonemic, that is, when
processing and retrieval are within the same domain (Morris,
Bransford, & Franks, 1977).
Data in STM is easily lost through interference and
decay. In contrast, reflecting the enduring nature of LTM,
failures of LTM have usually been considered to be produced
by retrieval deficits. The continued presence of the memory
may be assessed with recognition tasks. Recall tasks
require self-generation of cues. In contrast, in
recognition tasks cuing is externally provided. Thus,
recognition in the absence of spontaneous recall may be
explained by weak or decayed memory traces, failure to
generate adequate self-cuing or memory search strategies or


83
Table 4-7. Mean Sum CVLT Across Assessments and Between
Groups
Grp
Variable
N
Mean
Sd
Min
Max
Con
CVLT II
18
43.22
10.70
23
60
Con
CVLT III
18
47.28
10.68
30
68
Con
CVLT I
15
47.07
10.40
32
68
Trt
CVLT II
17
45.82
13.03
21
68
Trt
CVLT III
17
45.00
13.36
16
64
Trt
CVLT I
17
45.12
11.61
23
69
Representing iron status as a two-level classificatory
variable (i.e., Low vs. Replete), a GLM analysis of variance
failed to demonstrate a main effect for iron status on Sum
CVLT change scores between the first and second assessments
(F=0.01, p=.9384). Similar results were obtained
introducing Hg into the model (Hg: F=0.26, p=.6126; iron:
F=0.01, p=.9337).
Despite the lack of change in Hg status between the
second and third assessments, a GLM Repeated Measures ANOVA
was performed. As expected, the hypothesis of a treatment
effect was not supported by Wilks7 Lambda (F=0.2134,
E=.8092) or any other measure. Neither were there between
subject effects, within subject effects or treatment by
assessment interactions.
Post Hoc Analyses
Learning Curve
Hypothesizing that change in learning curve might be
more sensitive to the effects of anemia than Sum CVLT, SAS


19
study, Hagberg (1974) reported verbal STM as measured by
paired associate learning to be significantly lower than
that predicted by general verbal ability in 23 nondialized
uremic subjects. Six months after initiation of
hemodialysis, paired associate learning performance had
significantly improved in the 21 subjects retested and
approached expected levels relative to intelligence. This
study used alternate test forms but failed to use a control
group to control for possible effects of repeat testing.
Assessments of six subjects at varying time intervals
following hemodialysis revealed no changes in STM as
measured by a paired associate learning task, despite
changes in performance on timed tasks (Lewis, O'Neill,
Dustman, & Beck, 1980). In contrast, on a recognition task,
STM was shown to improve 24 hours after and deteriorate two
days following dialysis (Teschan et al., 1974). Other work
by this group has suggested improvement in verbal STM
following transplantation, but this has not reached
statistical significance (Teschan, Ginn, Bourne, & Ward,
1976).
Based largely on work with a pediatric population, it
has been suggested that STM and sensory/motor function are
more severely impacted than are more complex cognitive
abilities or LTM (Crittenden, Holliday, Piel, & Potter
1985). However, this report failed to provide sufficient
data to support this suggestion. It appears likely that the
measure of STM was Digit Span from the WAIS-R, which is more


70
group were on maintenance hemodialysis. In contrast, 94.1%
of the treatment group was on maintenance hemodialysis and
one subject representing 5.9% of the treatment group was
maintained on peritoneal dialysis. As confirmed by a Chi-
Square procedure these differences were highly significant
(X2(1, n=35)=16.032, pc.OOOl).
Determination of time on dialysis was made difficult by
intervening periods of renal transplantation as well as
inconsistencies between medical records and patient reports.
Also, this figure fails to take into account possible
differences in the effects of hemodialysis and peritoneal
dialysis over time. Mean years on dialysis for the
treatment group was 6.64 years (sd=7.69) compared to 2.85
(sd=3.76) years for the control group. These differences
approached but failed to reach statistical significance
(X2(l, n=35)=2.8425, p<.0918) based on the Kruskal-Wallis
Test procedure.
Attrition
All 17 subjects in the treatment group completed the
study; however, one 25 year old male died six weeks after
completion of the third assessment. Two subjects in the
control group, both on hemodialysis, died between the second
and third assessments. In addition, one control subject
maintained on peritoneal dialysis declined to complete the
third assessment after becoming hospitalized in a
neighboring city. Eliminating the three subjects not
completing the study resulted in a mean age in the control


32
affects infant behavior and impairs both infant and toddler
performance on tests of cognitive function, with these
effects magnified by the progression to iron deficiency
anemia. However, subjects exhibiting anemia due to iron
deficiency anemia are likely to be more iron deficient than
nonanemic subjects.
Some studies have found no differences between anemic
and nonanemic subjects and attributed the positive findings
predominant in the literature to uncontrolled variables such
as socioeconomic status and education (Johnson & McGowan,
1983). However, this does not explain the results of iron
treatment studies. In addition, the results of animal
studies support the predominant human findings (Massaro &
Widmayer, 1981; Yehuda, Youdim, & Mostofsky, 1986; Youdim &
Ben-Shachar, 1987).
Possible Mechanisms
Mechanisms that might produce neuropsychological
dysfunction in anemia include alterations in blood
chemistry, cerebral hypoxia and general fatigue. As
previously noted, hyperventilation induced by low 02 is
possible, but is attenuated by the resulting hypocapnia
(i.e., low C02). The effects of severe hyperventilation
include increased blood pH, alterations in neuronal
excitability, changes in EEG and cerebral vasoconstriction
(Fried, 1987). Carbon dioxide concentration is the primary
mechanism regulating cerebral blood flow with hypocapnia
inducing vasoconstriction and hypercapnia inducing


92
Because Physical Activity was not normally distributed,
a Wilcoxon Rank Sum Test procedure was used. Analysis
failed to support the hypothesis of a main effect for
treatment status on Physical Activity between the first and
second assessments (Z=1.1283, p=.2592). Similar results
were obtained using Hg change as the independent variable.
It was hypothesized that self-reported behavioral
indicators of Cognitive-Affective state would improve in the
treatment group. A Wilcoxon Rank Sum Test procedure failed
to demonstrate a change in Cognitive-Affective behavior (Z=-
1.7654, p=.0775) between the first and second assessments.
Because the results could arguably be interpreted as failing
to reject the possibility of a trend, the direction of the
changes was examined. As displayed in Table 4-10 below
change between the first and second assessments was in the
opposite of the predicted direction. Direction of change
between the second and third assessments was in the
predicted direction. Analysis of change scores between the
first and second assessments using a GLM procedure failed to
demonstrate a main effect for Hg (F=0.40, p=0.5336).
Using sign of Hg change scores as the classificatory
variable, a Wicoxon Rank Sum test failed to reveal
statistically significant differences in self-reported
Physical Activity (Z= -1.0235, p=.3061). Neither were there
statistically significant differences in self-reported
behavior in the Cognitive-Affective domain (Z= 0.2028,
P=.8393) .


55
in behavior in the treated subjects. The same version was
administered at all three assessments.
Hemoglobin
In most instances, Hg levels were drawn on rEPO
treated subjects on a weekly basis. When blood work was
not done on the day of neuropsychological testing, pre and
post target date Hg levels were used to calculate status on
the day of the assessment. As Hg status tends to change
linearly in the absence of major blood loss, estimation is
likely to have been reasonably accurate. In the few
instances where Hg levels were lacking, estimates were made
based on hematocrit (Ritchey, 1987). It should be noted
that the experimenter was blind to Hg levels until the
completion of all other data collection.
Ferritin
When available, ferritin levels were obtained for each
subject to rule out iron deficiency as a possible confound
(Ritchey, 1987). Ferritin, the most sensitive measure of
iron levels, has been used in studies examining the
neuropsychological effects of iron deficiency. However, in
the absence of ferritin levels, iron saturation and/or
transferrin were used if available. When the dates of
neuropsychological testing and blood work failed to
coincide, estimates were made based on iron measures
obtained before and after the time of testing. As iron
status tends to change slowly and linearly in the absence
of major blood loss, estimation is reasonably accurate.


8
to STM recall deficits (Lezak, 1983). In a series of factor
analyses, Wechsler (1987) found digit span to consistently
load primarily on attention-concentration rather than on a
memory factor. Unpublished data communicated by T. White
(June 12, 1990) revealed that for 182 patients seen in a
neuropsychology clinic, there was an insignificant
correlation (r= .123, p=.098) between digits forward and
trial one of the CVLT, a measure of STM recall.
The capacity of STM has been shown to be approximately
7 +2 chunks of information (Miller, 1956). Unfortunately
the notion of chunks has not been clearly defined. Recall
significantly exceeding the accepted capacity of 7 +2 has
been attributed either to the participation of LTM or
efficient organization of bits of information into larger
chunks. There are no known limits on the duration or
capacity of LTM.
An inherent feature of multistore models has been the
necessity of elucidating the processes through which
information is progressively transferred through the memory
storage areas. The mechanism underlying transfer of
information from sensory memory to STM has been attributed
to attentional processes associated with pattern (i.e.,
meaning) recognition (Moray, 1959). In contrast, Shiffrin
(1975) has proposed that perceptual stimuli are
automatically encoded and passed from sensory to STM. From
a multistore perspective, transfer from STM to LTM has been
viewed as a function of active attention or rehearsal


73
As expected, a one-way GLM procedure demonstrated that
between the first and second assessments the mean Hg rise of
1.13 g/dl in the treatment group (n=17) was significantly
greater (F=7.42, p=.0102) than the -0.21 g/dl Hg change in
the control group (n=18). At the time of the second
assessment mean Hg was higher in the treatment group
(M=9.59, sd=1.84) than in the control group (M=9.23,
sd=1.81) but the difference was not statistically
significant (F(l, 34)= 0.34, £>=.564).
GLM analysis of variance of Hg change scores between
the second and third assessments failed to reveal even a
trend towards a between group difference (F(l, 31)= 0.16.
P=.6964). Lack of change in the second phase of the study
contributed to the failure to find no more than a trend
towards a significant rise in Hg status between the first
and third assessments (F=3.19, p=.0841). At the third
assessment mean Hg was 9.54 g/dl (sd=1.48) in the treatment
group (n=17) and 9.62 g/dl (sd=1.92) in the control group
(n=15).
In this sample the 21 hemodialysis patients exhibited a
mean baseline Hg of 8.80 g/dl (SD=1.80) and the 14
peritoneal dialysis patients had a mean Hg level of 9.21
g/dl (sd=1.69). The differences were not statistically
significant (F (1, 34)= 0.44, p= .51). Within hemodialysis
patients, mean baseline Hg in the controls (n=5) was 9.83
g/dl (sd=2.53). In the treatment group mean Hg at the first
assessment was 8.48 g/dl (sd=1.48). Based on a T-Test


105
relationship between chronic anemia and cognition would be
potentially relevant in diverse areas of research, public
policy and clinical practice. Therefore, a second goal of
this study was to attempt to speak to the question of
whether chronic anemia affects neuropsychological
functioning.
The primary contribution to uremic anemia is the
failure of diseased kidneys to perform their normal role of
producing the erythropoiesis regulating hormone EPO.
Recently approved for clinical use, rEPO is capable of
effectively eliminating uremic anemia in the majority of
patients. Indeed, in most patients Hg may be maintained at
any desired level. Limitations on the target level of Hg
are a function of economics and severity of deleterious
side effects. Recombinant EPO is quite expensive.
Increasing hematocrit in association with rEPO treatment
has been associated with side effects including
hypertension. At present, there is insufficient data to
engage in cost-benefit and risk-benefit analysis.
Therefore, it was hoped that this study would generate
information regarding the relationship between level of Hg
and neuropsychological functioning to assist in risk-
benefit analysis.
Prior to the development of rEPO, since anemia was
nearly universal in chronic renal disease, any attempt at
clarification of the relative contributions of anemia and
uremia had been a most difficult task. Although several


112
significant main effect for sign of Hg change on Sum CVLT
(F=10.52, p=.0027). However, contrary to expectations, the
data indicated a negative relationship between rise in Hg
and Sum CVLT. Between the first and second assessments,
mean Sum CVLT change score fell 0.82 (Sd=5.38) in the
rising Hg group and improved 4.06 (Sd=9.91) in the falling
group. Physiological, psychometric and psychological
explanations for the apparent finding will be discussed.
The creation of a classificatory variable utilizing
direction of Hg change increased the magnitude and
significance of group differences in Hg change and
provided, arguably, an indirect means of further
characterizing the effects of rEPO on memory function.
This was accomplished both by including treatment status in
a multivariate model and by somewhat more adequately
testing the effects of the only known consequence of rEPO
therapy, increased erythropoiesis.
Recombinant EPO has never been specifically tested
clinically for deleterious neurobehavioral effects. The
drug may eventually be administered to several hundred
thousand people per year. Therefore, potential danger
signals, however improbable, deserve serious attention.
Several mechanisms may be hypothesized by which rEPO might
deleteriously affect cognitive functioning. They are
induction of iron deficiency, unknown hypothetical effects
of rEPO on the brain, cerebrovascular effects including
hypertension and/or increased blood clotting and reduced


102
Reclassification based on direction of Hg change
revealed that Sum CVLT performance improved when Hg fell
between the first and second assessments. Post hoc analysis
revealed a main effect for direction of Hg change (F=16.97,
£=.0003) with an interaction between direction of Hg change
and self-reported Cognitive-Affective behavior (F=6.53,
p=.0043). Decline in cognitive-affective functioning was
associated with decline in list learning performance.
The changes in Hg were of dubious physiological
significance; indeed, the main hypothesis was not adequately
tested due to the small change in Hg. Hypothesizing that
increased uremia may have produced the poor performance,
post hoc data collection was initiated. Despite
insufficient sample size, analyses were performed to explore
the possibility of a trend; however, there was no evidence
of a relationship between Bun and Hg ((F=0.77, p=.4139).
The lack of what is considered to be physiologically
significant change in Hg led to the utilization of this data
set to assess test-retest and/or alternate form reliability
for several instruments. Consistent with the validation
study, the third form of the CVLT developed for this project
was found to be of essentially the same level of difficulty
as Forms I and II. Levels of correlation between CVLT III
and the other two forms were quite acceptable. In contrast,
the two alternate forms of the levels of processing task and
the two alternate forms of the frequency estimation task
exhibited relatively disappointing levels of correlation


58
occurrence encoding phase, Controlled Word Association
Test, Cognitive-Affective and Physical Behavior
Questionnaire, levels of processing recognition phase, CVLT
delayed tasks, frequency of occurrence estimation phase.
Statistical Methodology
Statistical analyses were performed using the
Statistical Analysis System (SAS) package of statistical
procedures available for the IBM personal computer. The
SAS General Linear Model procedure was used to test the
main hypothesis. Analyses involving variables violating
assumptions of normality and homogeneity of variance were
performed utilizing nonparametric procedures. The assigned
alpha for the main hypothesis was .05, with secondary
hypotheses tested more stringently based on the
requirements of multiple testing and the extent to which
the assumptions underlying parametric procedures were
violated.
Hypotheses
There was one main hypothesis and four secondary
hypotheses. Two of the secondary hypotheses were intended,
in the event of failure to reject the main hypothesis, to
provide preliminary data regarding possible mechanisms
mediating the relationship between Hg and cognitive
functioning.
Main Hypothesis
The main hypothesis was that adult renal dialysis
patients receiving rEPO replacement therapy would exhibit


CHAPTER 2
LITERATURE REVIEW
Introduction
Investigation of the neuropsychology of uremic anemia
is necessarily multidisciplinary. The logic underlying the
rationale and design of this study is based on information
derived from five somewhat distinct areas of investigation.
The areas are 1) human memory, 2) clinical studies of
neuropsychological function in adults suffering from renal
disease, especially with regard to verbal memory, 3) the
physiology of anemia, 4) the neuropsychology of anemia and
5) the effects of rEPO mediated improvement in anemia in
adult renal patients on neuropsychological function. Each
of these will be briefly reviewed in this chapter.
Overview of Verbal Memory
This section briefly overviews current thinking with
regard to memory as it relates to the clinical assessment
and interpretation of memory function. The application of
memory paradigms to the study of the neuropsychology of
anemia is discussed. Mounting neurochemical, functional and
anatomical evidence suggests that memory may be classified
into procedural memory (i.e., motor memory or skill
learning) and declarative memory for facts (Nissen, Knopman,
& Schacter, 1987). Memory may, to some extent, be divided
5


4
Chronic anemia has been reported to have far reaching
effects on the organism. In loose association with the
severity of the anemia, symptoms may include fatigue,
hypoxia, hypocapnea, insomnia, enhanced cardiac output and
increased sympathetic activation, reduced appetite, altered
blood chemistry, modified lifestyle including the inability
to hold gainful employment and, probably, decreased self
esteem. Therefore, determination of the relative
contributions of possible underlying mechanisms was
considered impossible in a single study of this scope.
However, it was hoped that this study would extend our
understanding of the neuropsychology of chronic anemia and
the clinical significance of chronic anemia with regard to
the etiology of the cognitive deficits observed in renal
disease.


APPENDIX H
CHANGE SCORES
Control Group Treatment Group
Variable
Tm
N
Mean
Sd
N
Mean
Sd
Hemoglobin
1
18
-0.21
1.34
17
1.13
1.57 b
Hemoglobin
2
15
0.05
2.01
17
1.08
1.23
Hematocrit
1
18
-1.11
4.32
17
3.24
4.54 b
Hematocrit
2
15
-0.43
6.21
17
1.98
3.14
California
Verbal Learning Test
Trial 1
1
18
0.94
2.24
17
-1.00
1.58 b
Trial 1
2
15
-0.67
1.99
17
1.12
1.76 b
Trial 2
1
18
1.17
2.41
0
-0.24
1.56 a
Trial 2
2
15
-0.13
2.53
0
0.76
2.31
Trial 3
1
18
0.94
3.06
17
0.12
2.15
Trial 3
2
15
0.13
2.10
17
0.06
2.28
Trial 4
1
18
0.83
3.01
17
0.18
2.46
Trial 4
2
15
0.07
1.39
17
-1.06
3.25
Trial 5
1
18
0.33
2.89
17
0.06
2.44
Trial 5
2
15
0.13
2.53
17
-0.18
1.78
SS
1
18
-0.11
1.78
17
0.12
1.54
SS
2
15
0.27
1.79
17
-0.06
1.09
Sum CVLT
1
18
4.06
9.91
17
-0.82
5.38
Sum CVLT
2
15
-0.27
6.97
17
0.12
7.85
SS
1
18
5.50
13.61
17
-0.82
5.38
SS
2
15
-0.67
10.88
17
0.12
7.85
Cluster
1
18
1.50
10.25
17
-1.00
7.27
Cluster
2
15
0.67
8.06
17
0.35
10.42
Psv
1
18
0.56
3.76
17
0.29
5.01
Psv
2
15
-0.73
3.13
17
0.24
6.94
Intrus
1
18
0.67
3.20
17
0.35
4.40
Intrus
2
15
-0.33
2.94
17
1.00
4.76
List B
1
17
-0.41
2.69
17
1.18
3.19
List B
2
15
0.67
2.97
17
0.24
2.82
SS
1
17
-0.06
1.60
17
-0.12
1.62
SS
2
15
0.20
1.74
17
-0.76
2.25
Short Del
1
17
0.06
2.75
17
-0.24
0.97
Short Del
2
15
-0.20
2.86
17
-0.24
1.35
SS
1
17
0.06
1.39
17
-0.06
2.66
SS
2
15
0.00
1.56
17
-0.12
2.26
155


37
mean Hg from 5.9 g/dl to 10.9 g/dl with rEPO treatment. The
initial mean anaerobic threshold of 70 watts represented a
very limited exercise capacity with inability to perform
routine household work over a sustained period. Following
rEPO treatment, the mean anaerobic threshold of 106 watts
indicated that work capacity had increased enough to allow
for full physical rehabilitation with regard to everyday
life (Mayer, Thum, Cada, Stummvoll, & Graf, 1988).
Previous Neuropsychological Studies
Several small pilot studies have assessed the impact of
rEPO on cognitive function. Wolcott, Schweitzer and Marsh
(1988) reported that in a group of nine chronic hemodialysis
patients administration of tests that loaded on visual,
conceptual and visuomotor tracking and auditory verbal
learning demonstrated trends toward improvement. Wolcott,
Schweitzer and Nissenson (1989), in a study of the effects
of EPO treatment in a group of 17 chronic hemodialysis
patients, reported improvements in symbol-digits modality
and Trails; however, it is unclear whether these
improvements reached clinical or statistical significance.
Decreased P3 latency purportedly represents increased
speed and efficiency of information processing. Increased
amplitude of event-related potentials suggests improvement
in cognitive function. Nissenon, Marsh and Brown (1988)
reported that following a mean rEPO mediated improvement in
hematocrit from 22.7% to 36.6%, 13 chronic hemodialysis
patients exhibited a decrease in P3 latency of response to


CHAPTER 3
METHODS
The present study was designed to broadly address the
question of whether change in anemia would be reflected by
change in memory.
Subjects
All subjects were adults in end-stage renal failure
and reasonably stabilized on some form of maintenance
dialysis. Subjects were drawn from the Shands Teaching
Hospital (STH) Adult Hemodialysis Clinic, the STH Dialysis
Home Training Clinic and the Gainesville Veterans
Administration Medical Center Hemodialysis Clinic. Age
ranged between 23 and 73 years. Subjects with uncontrolled
hypertension, mental retardation or known neurological
disease were not accepted into the study.
The treatment group consisted of 16 subjects on
maintenance hemodialysis and one subject on peritoneal
dialysis. It was planned that subjects would only be
accepted into the study into their eighth week of rEPO
treatment, based on the assumption that STH adult
hemodialysis patient hematocrits would be rising for
approximately five months. Therefore, a subject initially
tested at 8 weeks was expected to just be reaching
asymptote at the third assessment. However, subjects began
50


This dissertation was submitted to the Graduate Faculty
of the College of Health Related Professions and to the
Graduate School and was accepted as partial fulfillment of
the requirements for the degree of Doctor of Philosophy.
August 1991
(flsUr^Q ft.
Dean, College of Health Related Professions
Dean, Graduate School


23
including differences in medical management, age, age of
onset and etiology of renal disease. Many of the conditions
such as diabetes, hypertension and cardiovascular disease
contributing to or associated with renal failure have
neuropsychological consequences in their own right.
Discrepant findings in hemodialysis patients may to some
extent be a function of the time of assessment relative to
dialysis, the adequacy of dialysis and the degree of
electrolyte disequilibrium following dialysis. Most of the
early studies sampled neuropsychological function at one
point in time. There have been very few longitudinal
studies.
Progress in the treatment of renal disease has been
charted in the literature on the neuropsychology of uremia.
Prior to the availability of dialysis, the
neuropsychological consequences of renal failure often
included stupor and coma (Arieff, Guisado, & Massry, 1975;
Tyler, 1968). Prior to the discovery and acceptance of the
neurotoxicity of elevated blood aluminum levels in renal
patients, dialysis encephalopathy affected a significant
percentage of patients (Sprague et al., 1988). Improvements
in the methodology of dialysis as well as other aspects of
medical management continue to improve the physiological
conditions under which uremic nervous systems function.
Concurrently, these advances have also resulted in longer
durations of exposure to uremia as well as an aging renal
population. Thus, discrepancies in the literature may to


89
(sd=4.00) in the group with rising Hg (n=23). Please see
Table 4-9.
Table 4-9. Mean Change in CVLT Trials 4 plus 5 Change
Scores Between Assessments I & II
Gro
N
Min
Max
Mean
Sd
Pos
23
o
o

CO
6.00
00

o
4.00
Neg
12
o
o

T
18.00
3.75
5.24
An identical analysis between the second and third
assessments revealed a mean change of -0.78 words (sd=3.93)
in the negative change group (n=9) and a mean change of
-0.15 words (sd=3.96) in the positive Hg change group
(n=20). These differences failed to reach statistical
significance (Z=-.2845, p=.7760). Analysis of change scores
between the first and third assessments revealed a mean
change of -3.71 words (sd=7.06) in the negative change group
(n=7) and a mean change of -1.25 words (sd=4.23) in the
positive Hg change group (n=24). These differences failed
to reach statistical significance (Z=1.5613, p=.1184).
Multivariate analysis (n=35) with a General Linear
Model failed to find a main effect for rEPO treatment (i.e.,
experimental vs. control) on Sum CVLT (F=0.84, p=.3668) in a
model including a sign of Hg change factor (F=7.45,
p=.0102).
To test the possibility that undiscovered outliers
unduly influenced the results, Sum CVLT change scores were


CHAPTER 5
DISCUSSION
Overview
A broad spectrum of cognitive abilities have been
reported to be deleteriously affected by chronic renal
disease. Despite considerable effort, the specific
etiologic mechanisms underlying the neuropsychological
deficits associated with end-stage renal disease have not
been satisfactorily explained. Anemia, often severe,
almost invariably accompanies chronic renal failure. One
purpose of this project was to test the hypothesis that
chronic anemia may contribute to the neuropsychological
deficits associated with end-stage renal disease. Chronic
anemia is one of the most common medical conditions in the
world, affecting diverse populations, particularly
including infants, women, the elderly and the malnourished
as well as specific disease populations. Despite the high
prevalence and longstanding recognition of the often
dramatic effects of acute anemia on cognition and brain
function, the neuropsychology of chronic anemia has been
largely ignored. Indeed, presumably because of the
difficulty in disentangling anemia from its etiology, the
question of whether chronic anemia has neuropsychological
consequences has remained unanswered. Elucidation of the
104


145
Kruskal-Wallis Test
General Linear
(Chi-Square Approximation) Model
Nonparametric Parametric
Df=1
Prob >
Variable <
Shi-Sauare
Chi-Scruare
F Value
Pr >F
Long Delay
0.1863
.6660
0.15
.6995
Standard Score
0.6110
.4344
1.27
.2671
Clusters
0.7180
.3968
0.33
.5670
Perseverations
0.8303
.3622
0.31
.5838
Intrusions
4.8724
.0273*
2.48
.1245
Long Delay Cued
0.3420
.5587
0.04
.8425
Standard Score
0.7505
.3863
0.84
.3669
Perseverations
-
-
-
-
Intrusions
1.5373
.2150
2.05
.1613
CVLT Recognition
Variables
Recognition Hits
0.0862
.7691
0.75
.3931
False Positives
From Interference List
Shared Category
1.0593
.3034
1.12
.2981
Unrelated
0.0912
.7627
0.02
.8869
Intrusions not
on Interference List
Shared Semantic
1.2254
.2683
2.19
.1482
Shared Phonemic
0.8235
.3641
1.71
.1996
No Relation
0.6879
.4069
0.71
.4045
Controlled Word Association
CFP
0.7183
.3967
0.49
.4880
FAS
1.2263
.2681
1.41
.2445
LSW
5.1345
.0235*
4.62
.0394*
Levels of Processing
Semantic
1.7209
.1896
0.93
.3424
Orthographic
0.1749
.6757
0.16
.6887
Phonemic
0.0819
.7747
0.21
.6477
Estimation of Frequency of '
Occurrence
Free Recall
0.0188
.8907
0.04
.8390
Estimation
0.2524
.6153
0.09
.7655
Cognitive-Affective and Physical Behavior
Questionnaire
Cognitive-Affect
0.1360
.7123
0.11
.7459
Physical
0.1104
.7396
0.19
.6676
Note: Df = degrees of freedom
* = discrepant results between parametric and
nonparametric tests


108
scores obtained in this study were not within the range
that would be expected to be physiologically significant.
Self-report data regarding Physical Activity had been
obtained so that, in the event of negative findings, a
determination could be made as to whether Hg levels had
been adequately manipulated to produce improvement in life
functioning. Consistent with the small change in Hg, there
were no between group differences in change scores for
self-reported Physical Activity or for Cognitive-Affective
behavior.
In comparison to the mean Hg change of 1.13 (Sd=1.90)
g/dl in the present study, mean treatment group Hg rose
5.46 (Sd=1.72) g/dl in the pilot study. Indeed, despite a
statistically significant change in Hg, (F=7.42, p=.0102),
the independent variable does not appear to have been
sufficiently manipulated; therefore, the main hypothesis
was not adequately tested. In addition, the absence of
change in Sum CVLT in the present study, in conjunction
with a smaller rise in Hg, suggests that Hg change may not
have been sufficient to provide clinical benefit.
Thee are other possible explanations for the
discrepancy between the pilot and the present study. Not
only was change greater, but Hg levels began lower in the
pilot study (M=6.10, Sd=.97 g/dl) than in the present study
(M=8.46, Sd=1.42 g/dl). The physiological effects of
varying degrees of severity of chronic anemia may be
nonlinear. This would suggest that one unit of Hg change


96
Table 4-14. Correlations & Sum CVLT Means Between CVLT
Forms I. II and III
Variable Pearson Correlations Form Mean Sd
I-II
I-III
II-III
Trial
r
.48240
.59965
.47084
I
5.800
2.098
1
P
.0033
.0003
.0065
II
5.800
2.139
N
35
32
32
III
5.938
1.795
Trial
O
r
.56327
.60335
.61043
I
8.143
2.171
c*
P
.0004
.0003
.0002
II
8.629
2.377
N
35
32
32
III
8.969
2.957
Trial
O
r
.59901
. 66308
.65356
I
9.400
3.098
J
P
.0001
.0001
.0001
II
9.943
2.786
N
35
32
32
III
10.063
2.213
Trial
r
.42348
.47538
.64775
I
10.286
2.906
4
P
.0112
.0060
.0001
II
10.800
3.095
N
35
32
32
III
10.344
2.925
Trial
r
.64340
.52880
.75095
I
10.886
3.132
5
P
.0001
.0019
.0001
II
11.086
3.128
N
35
32
32
III
11.031
2.753
Sum
r
.75543
.71063
.80342
I
44.486
11.788
P
.0001
.0001
.0001
II
46.171
11.930
N
35
32
32
III
46.031
10.926
Sum
r
.64837
.80121
.73281
I
12.229
9.540
Clus
P
.0001
.0001
.0001
II
13.143
9.638
N
35
32
32
III
13.781
10.222
Sum
r
.13204
.27777
.07528
I
3.343
2.828
Psv
P
.4496
. 1237
.6822
II
3.800
3.261
N
35
32
32
III
4.219
2.685


Table B-4 continued
138
Processing
Multiole Choice
Word List
Level
26.
silk
shovel
fog
brass
SEM-
27.
dog
lamp
wave
ink
PHON-
28.
ship
fad
gear
tan
ORTH-
29.
truck
cop
prey
camp
0RTH+
30.
soap
pen
spike
chart
PH0N+
31.
float
robot
film
ant
PHON-
32.
paint
hinge
mute
lake
0RTH+
33.
hamster
ditch
surf
ache
SEM-
34.
horse
seem
ramp
dog
ORTH-
35.
floor
hen
stamp
net
PH0N+
36.
spark
hotel
grip
site
SEM+
Note:
+ = Correct
response
is positive;
- = Correct
response
is negative;
ORTHO =
Orthographic; PHON = Phonetic; SEM = Semantic


160
Table 1-3. California Verbal Learning Test Short Delay,
Long Delay and Recognition Raw Scores
Short Delay Long Delay Recognition
Assess
1
2
3
1
2
3
1
2
3
Id
1
15
10
11
16
10
11
16
14
14
2
9
10
8
8
10
10
14
14
16
3
9
10
7
10
11
8
14
15
14
4
8
9
7
7
6
8
15
12
13
5
12
12
11
12
11
13
16
13
14
6
6
8
7
5
10
6
15
14
15
7
8
13
13
11
12
13
16
16
15
8
8
7
10
10
11
9
15
14
14
9
14
13
12
14
13
10
16
14
14
10
13
14
13
12
15
13
15
15
16
11
11
14
14
11
13
15
15
15
16
12
4
3
6
2
2
2
14
15
14
13
10
13
9
9
12
11
16
16
14
14
9
7
8
5
5
8
13
13
2
15
7
5
10
10
6
10
14
14
12
16
15
14
14
14
13
15
16
15
15
17
6
1
1
5
0
2
15
12
13
101
8
7
9
9
10
10
16
16
16
102
9
7
8
9
8
8
14
14
11
103
10
9
6
11
10
6
16
15
12
104
3
6

2
6

11
11

105
6
10
8
8
6
8
16
16
14
106
9


9


15


107
9
8
11
10
8
8
16
16
16
108
11
8
12
11
9
11
16
13
15
109
10
9
10
14
12
13
13
15
14
110
12
10
10
12
9
9
16
16
14
111
10
6
10
12
9
11
15
14
15
112
3
5
2
3
4
3
12
14
11
113
10
13
11
12
13
11
15
15
15
114
13
14
13
15
14
14
14
14
14
115
12
14
12
14
15
12
16
16
16
116
9
14
8
11
15
8
12
16
14
117
10
10

10
8

15
15

118
9
5
6
7
6
7
15
10
14
Note: Assess = Assessment number


154
Table G-3 continued
Control
Group
EPO
Treated
Variable
N
Mean
Sd
N
Mean
Sd
Intrusions not on
Interference
List
Shared Semantic
15
0.33
0.62
17
0.53
0.87
Shared Phonemic
15
0.33
0.82
17
0.59
0.94
Unrelated
15
0.67
0.82
17
0.65
1.06
Verbal Fluency
P
15
15.07
3.71
17
11.00
4.23
R
15
11.60
3.74
17
9.29
3.64
W
15
10.87
2.77
17
8.18
3.97
LOP Semantic
14
9.36
1.65
16
10.19
1.22
LOP Orthographic
14
6.79
2.52
16
7.12
2.55
LOP Phonemic
14
7.57
1.87
16
9.06
2.32
Estimation of Frequency
of Occurrence
Free Recall
15
4.67
1.23
17
3.41
1.87
Frequency Estimate
15
10.33
2.35
17
9.47
3.28
Cognitive-Affective
and
Behavioral Questionnaire
Cognitive-Affective
15
5.40
1.99
17
5.53
2.40
Physical
15
9.53
4.72
17
8.59
4.54
Note: LOP = Levels of Processing


100
Table 4-17. Test Retest Reliability of Physical Activity
Self-Report
Pearson
Correlations between Forms
I & II
I & III
II & III
Form
Mean
Sd
r .89863
.76990
.82068
I
9.230
4.538
p .0001
.0001
.0001
II
9.029
4.596
N 34
32
31
III
9.031
4.575
Note: r = correlation; p = p value; N = sample size
Table 4-18. Test Retest Reliability of Cognitive-Affective
Self-Report
Pearson
Correlations between Forms
I & II
I & III
II & III
Form
Mean
Sd
r
.64224
.86002
.68524
I
5.943
2.071
P
.0001
.0001
.0001
II
6.412
2.388
N
34
32
31
III
5.469
2.185
Note: r = correlation; p = p value; N = sample size
Cognitive-Affective Behavior Self-Report Test-Retest
Reliability
Table 4-18 illustrates significant test-retest
correlations for a series of true or false questions
focusing on cognitive-affective state and behavior.
However, comparison with the physical behavior data revealed
correlations tended to be higher and means less variable
when focusing on physical symptoms and behavior.
Post Hoc Analysis of Blood Urea Nitrogen
After completion of data collection and analysis it was
hypothesized that the inverse relationship between Hg and


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EA8W62V04_30PDZ8 INGEST_TIME 2015-04-06T18:43:54Z PACKAGE AA00029752_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


82
Test of Main Hypothesis
The main hypothesis stated that rEPO mediated reduction
in chronic anemia in adult dialysis patients is associated
with improved verbal LTM as measured by the CVLT Sum of
trials one through five.
Mean Sum CVLT change scores between the first and
second assessments were -0.82 (sd=5.38) in the treatment
group (n=17) and 4.06 (sd=9.91) in the control group (n=18).
These results were in the opposite of the hypothesized
direction. A GLM analysis of covariance procedure failed to
demonstrate a main effect for either treatment status
(F=l. 29, p=.2638) or Hg (F=1.32, p=.2590).
With both treatment status and Hg change in the model,
few conclusions could be drawn. Therefore, separate
univariate GLM analyses were performed for treatment status
and for Hg change. However, these analyses failed to show
a statistically significant main effect for either Hg
((F=3.25, p=.0805) or for treatment status (F=3.22,
p=.0818).
Examination of mean Sum CVLT scores between groups and
across assessments revealed a curious pattern. Although
none of the differences reached statistical significance,
examination of Table 4-7 reveals that the control group
performed poorly on the first assessment relative to the
other two assessments. In contrast, assuming equivalence of
test forms, treatment group performance was static. That
is, change apparently occurred unrelated to rEPO.


7
onto biological structures and processes, the neuroscience
perspective has focused both on the nature of the memory
trace and the physical location of memory in the brain.
This brief overview will focus primarily on a behavioral
(i.e., cognitive and neuropsychological) level of analysis
as it relates to neuropsychological assessment.
As conceptualized in the multistore model, memory is a
complex series of distinct stages through which information
is successively processed (Atkinson & Shiffrin, 1968, 1971).
These stages differ along several dimensions including
capacity, duration and the nature of the attendant
psychological processes. The three generally recognized
stages are ultrashort or modality-specific sensory memory
(e.g., echoic and iconic memory), immediate or STM and LTM.
Sperling (1960) demonstrated that iconic (i.e., visual
sensory) memory decays after about one second. Echoic
memory may take several seconds to decay (Darwin, Turvey, &
Crowder, 1972). Sensory memory capacity appears to be
around nine items (Sperling, 1960). The duration of STM
when rehearsal (i.e., replenishment) is prevented has been
shown to be about 15 seconds (Peterson & Peterson, 1959).
Thus, the repetition of a list of orally presented digits or
words utilizes both echoic memory and STM. However, in the
case of oral digit span, as traditionally administered, the
limiting factor is likely to be attentional.
Oral digit span is the most commonly used measure of
immediate memory span; however, it is frequently insensitive


165
Delis, D.C., McKee, R., Massman, P., Kramer, J.H., &
Gettman, D. (1990). Alternate Form of the California
Verbal Learning Test: Development and Reliability,
unpublished manuscript.
Desforges, J.F. (1975). The role of hemodialysis in the
hematologic disorders of uremia. Kidney International. 7,
S-123-128.
Drayer, B.P., Burger, P., Darwin, R., Reiderer, S.,
Herfkens, R., & Johnson, G.A. (1986). Magnetic resonance
imaging of brain iron. Americal Journal of
Neuroradioloqy. 7, 373-380.
English, A., Savage, R.D., Britton, P.G., Ward, M.K., &
Kerr, D.N.S. (1978). Intellectual impairment in chronic
renal failure. British Medical Journal. 1, 888-890.
Erslev, A.J. (1975). The effect of uremic toxins on the
production and metabolism of erythropoietin. Kidney
Internation. 7, S129-S133.
Erslev, A.J. (1987). Erythropoietin coming of age. New
England Journal of Medicine. 316. 101-103.
Eschbach, J.W. (1989). The anemia of chronic renal failure:
pathophysiology and the effects of the effrects of
recombinant erythropoietin. Kidney International. 35.
134-148.
Eschbach, J.W., & Adamson, J.W. (1985). Anemia of end-stage
renal disease (ESRD). Kidney International. 28, 1-5.
Eschbach, J.W., Egrie, J.C., Downing, M.R., Browne, J.K., &
Adamson, J.W. (1987). Correction of the anemia of end-
stage renal disease with recombinant human
erythropoietin. New England Journal of Medicine. 316. 73-
78.
Evans, D.I. (1985). Cerebral function in iron deficiency: A
review. Child-Care, Health and Development 11. 105-112.
Evans, R.W., Rader, B., & Manninen (1990). The quality of
life of hemodialysis recipients treated with recombinant
human erythropoietin. Journal of the American Medical
Association, 263, 825-830.
Ewing, R., McCarthy, D., Gronwall, D., & Wrightson, P.
(1980). Persisting effects of minor head injury
observable during hypoxic stress. Journal of Clinical
Neuropsychology. 2, 147-155.


APPENDIX D
COGNITIVE-AFFECTIVE AND PHYSICAL BEHAVIOR QUESTIONNAIRE
Table D-l. Questionnaire
Circle only the sentences true over the past two days.
1. I don't mind climbing stairs.
2. Walking quickly tires me out.
3. I sit during much of the day.
4. I take one or more naps during the day.
5. I feel useless, like I am a burden on others.
6. I easily get out of breath.
7. I have difficulty concentrating.
8. My sleep is sounder than it used to be.
9. I am irritable and impatient with myself.
10. I do work around the house only for short periods.
11. I stay away from home only for brief periods of time.
12. My sex drive is stronger than it was one year ago.
13. I am not doing any daily work around the house.
14. I feel terrible when I get off of hemodialysis.
15. I feel tired most of the time.
16. I do not finish things I start.
17. I react slowly to things that are said or done.
18. I work a job outside the home.
19. I spend much of the day lying down in order to rest.
20. I enjoy eating.
141


106
studies controlled for Hg, possible confounds such as level
of physical exercise, degree of uremia, extent of uremic
anorexia and years on hemodialysis, as well as additional
sources of error such as the inability to use subjects as
their own controls, may have hidden the effect. The
development of rEPO, because it presumably alleviates
uremic anemia without directly affecting the brain,
suggested a method of disassociating anemia from the usual
confounds and presented the opportunity to use subjects as
their own controls. Thus, rEPO treatment was seen as a
powerful new method for studying the neuropsychology of
chronic anemia.
Several lines of evidence hinted at a possible
relationship between anemia and cognitive dysfunction.
Neuropsychological function has in some studies been
reported to be somewhat better in peritoneal dialysis than
in hemodialysis. Although this has been attributed to
greater middle molecule clearance in peritoneal dialysis,
the offending middle molecule has never been found, and Hg
levels tend to be higher in peritoneal dialysis. Also,
years on hemodialysis has been reported to be associated
with decreasing cognitive function and increasing anemia.
The dramatic reduction in anemia produced by treatment
with rEPO has been shown to produce clinically significant
improvements in domains such as aerobic capacity, anaerobic
threshold, sexual potency, employment status, level of
social activity, perceived ability to engage in physical


98
Table 4-14 continued
Note: Under Correlations, the first value listed is the
correlation (r) under that the p value, and then, N,
the number of subjects; Psv = Perseveration; Intrus =
Intrusions; Recog = Recognition
Levels of Processing Reliability
Forms II and III of the Levels of Processing Task were
developed for use in this study. As illustrated in Table 4-
15, correlations between recall of orthographically
processed words on Levels of Processing Forms I, II and III
were all positive, but low and nonsignificant. Correlations
between phonemically processed stimuli followed an identical
pattern between Forms I and II and I and III. However, the
correlation between Forms II and III reached statistical
significance. The opposite was true of the semantically
processed words. Correlations between Forms I and II and
Forms I and III reached statistical significance.
Frequency of Occurrence Reliability
Table 4-16 illustrates a finding identical to that
of the Semantic Levels of Processing results, in that
correlations reached statistical significance between Forms
I and II and between Forms I and III, but not between II and
III.
Physical Behavior Self-Report Test-Retest Reliability
The data displayed in Table 4-17 demonstrates quite
acceptable test-retest reliability for the Physical Behavior
scale. Correlation coefficients ranged from r=.820 to


175
Wolcott, D.L., Schweitzer, S., & Nissenson, A.R. (1989).
Recombinant erythropoietin (r-EPO) improves cognitive
function (CF) and quality of life (QL) of chronic
hemodialysis (CHD) patients. Kidney International. 35.
266. Abstract.
Yehuda, S., Youdim, M., & Mostofsky, D.I. (1986). Brain iron
deficiency causes reduced learning capacity in rats.
Pharmacology. Biochemistry and Behavior. 25, 141-144.
Youdim, M.B.H., & Ben-Schachar, D. (1987). Minimal brain
damage induced by early iron deficiency: modified
dopaminergic neurotransmission. Israel Journal of Medical
Sciences, 2_3, 19-23.
Ziestat, H.A. Jr., Logue, P.E., & McCarty, S.M. (1980).
Psychological measurement of memory deficits in dialysis
patients. Perceptual Motor Skills. 50, 311-318.


TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS ii
ABSTRACT vii
CHAPTERS
1 INTRODUCTION: FOCUS AND RATIONALE 1
2 LITERATURE REVIEW 5
Introduction 5
Overview of Verbal Memory 5
Memory Stages and Levels of Processing ... 6
Automatic and Effortful Memory 12
Neuropsychological Dysfunction in Adult Renal
Patients 13
General Intelligence: Verbal and Nonverbal 13
Attention, Level of Activation and Executive
Function 14
Verbal Memory 16
Nonverbal Memory 20
Conclusions Regarding Cognitive Deficits in
Renal Disease 21
Overview of Anemia 24
Prevalence and Etiology 24
Function of Hemoglobin and Adaptation to
Anemia 26
Erythropoietin Therapy and Iron Deficiency 27
Neuropsychology of Anemia 27
Effects of Anemia 29
Possible Mechanisms 32
Effect of Erythropoietin on Neuropsychological
Function 35
Effect of Erythropoietin on Quality of Life 36
Previous Neuropsychological Studies 37
Results of Pilot Study 38
Negative Effects 44
Summary and Conclusions 46
3METHODS 50
Subjects 50
Measures 52
iv


25
approximately 25% of dialysis patients suffered from anemia
severe enough to require intermittent or regular red-cell
transfusions (Eschbach, Egrie, Downing, Browne, & Adamson,
1987).
Uremic anemia results from several interactive
processes. Shortened red blood cell survival is a frequent
manifestation of uremia (Jacob, Eaton, & Yawata, 1975).
Although controversial, some evidence suggests that uremic
toxins inhibit heme synthesis and erythroid progenitor cell
formation (Erslev, 1975; Jacob, Eaton, & Yawata, 1975).
Blood loss due to platelet dysfunction and as a result of
hemodialysis are implicated (Desforges, 1975). The
increased level of potentially toxic "middle molecules" in
hemodialysis compared to peritoneal dialysis may also
contribute to anemia in hemodialysis. Iron deficiency,
folate deficiency, B12 deficiency, aluminum-induced
microcytosis and hypersplenism have been cited (Paganini,
1989) .
One study has demonstrated a significant increase in
hematocrit and Hg levels following 12 months of endurance
exercise training, which suggests that sedentary life style,
a frequent consequence of end-stage renal disease, may
contribute to the anemia (Goldberg et al., 1986). However,
inadequate erythropoietin (EPO) production relative to the
degree of anemia is the major cause of uremic anemia. Thus,
uremic anemia is primarily due to an endocrine deficiency
state, correctable by rEPO replacement therapy (Chandra,


14
Interestingly, McDaniel's (1971) data on visual
discrimination performance suggested impairment not
primarily as a result of difficulties with visual-motor
integration, but secondary to interference with cognitive
processing. The reported visuospatial motor dysfunctions
may be partially mediated by attentional deficits (Fennell,
Fennell, Mings, & Morris, 1986; McDaniel, 1971).
Attention. Level of Activation and Executive Function
Several authors have suggested the presence of global
attention deficits in uremia (Marshall, 1979; Stewart &
Stewart, 1979). Data reported by Trieschmann and Sand
(1971) on 83 subjects not treated with dialysis demonstrated
reduced performance on two measures loading on simple
attention and concentration, the Digit Span and Arithmetic
subtests of the Wechsler Adult Intelligence Scale (WAIS).
Grouping subjects based on severity of illness revealed
Digit Span to be the lowest subscale score in the more
severely ill group.
In a dialized population with normal premorbid verbal
IQ based on vocabulary knowledge, English, Savage, Britton,
Ward and Kerr (1978) found reduced performance on WAIS
subtests most heavily loading on attention (i.e.,
Arithmetic, Digit Span, Digit Symbol). In a pediatric
population, Digit Span was reported to be significantly
reduced regardless of treatment modality. Indeed, in a
pediatric population, excluding the renal transplant group,
Digit Span discriminated between renal and control groups


iii
experience. Eileen Fennell, my chair, enabled me to meld
my interests with the do-able and the acceptable. Without
her the going would have been much less pleasant. It was
on Russ Bauer's clinic day that I decided to specialize in
clinical neuropsychology. Hugh Davis is forever and
agreeably etched in my subconscious. Bob Fennell provided
research opportunities with import to me well beyond the
findings. Tom Fast became a close friend who taught many
things.


88
2=.5336). Correlation between Sum CVLT and Cognitive-
Affective self-report was positive (r=.37042, p=.0310).
Examination of mean scores by groups revealed that Sum CVLT
rose in association with rise in Cognitive-Affective
behavior; however, there was less evidence of a relationship
when Hg fell.
In a model analyzing change in Sum CVLT, covarying
change in self-reported Physical Activity and direction of
change in Hg failed to demonstrate a main effect for
Physical Activity (F=0.77, p=.3866). There was no
interaction between sign of Hg change and Physical Activity
(F=0.75, p=.4827). Neither was there an effect on Sum CVLT
for self-reported level of Physical Activity at the time of
the first assessment. Analyses failed to demonstrate a
significant relationship between self-reported Physical
Activity and Sum CVLT
Sum of CVLT Trials Four and Five
Using sign of Hg change scores as the classificatory
variable, a Wilcoxon 2-Sample Rank Sum Test was used to test
the post hoc hypothesis of a change in the sum of CVLT
trials four and five between the first and second
assessment. The difference between groups was significant
even controlling for multiple comparisons (Z=2.6366,
P=.0084). Once again the difference was the reverse of the
predicted direction. Mean increase in number of words
recalled for the sum of trials four and five was 3.75
(sd=5.24) in the group with falling Hg (n=12) and -0.87


Abstract of Dissertation Presented to the Graduate School of
the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
EFFECT OF RECOMBINANT ERYTHROPOIETIN MEDIATED CHANGES IN
ANEMIA ON VERBAL MEMORY IN ADULT DIALYSIS PATIENTS
By
Wayne L. Klein
August 1991
Chairperson: Eileen B. Fennell
Major Department: Clinical and Health Psychology
Cognitive dysfunction has been documented in many
chronic anemic conditions; however, the putative effects of
anemia have never been disassociated from the cause of the
anemia. Neuropsychological dysfunction of poorly understood
origin and severe chronic anemia are nearly universal
consequences of chronic renal failure. A small pilot study
demonstrated a positive relationship between a recombinant
erythropoietin (rEPO) mediated mean rise of 5.46 g/dl of
hemoglobin (Hg) and the Sum of trials 1-5 on the California
Verbal Learning Test (Sum CVLT). Sum CVLT performance was
hypothesized to be enhanced by increased effortful
processing secondary to reduced fatigue.
A rEPO treatment group (n=17) and a control group
(n=18) underwent three assessments with three alternate
forms of the CVLT, one of which was developed for this
vii


74
procedure these differences failed to reach statistical
significance (T=1.5099, £=.1475).
Comparison of the maintenance hemodialysis pretreatment
Hg of 7.06 g/dl (sd= 1.42) in the rEPO group (n=16) to the
baseline Hg of 9.83 g/dl (sd=2.53) in the control group
(n=5) revealed a trend toward lower values in the treatment
group. However, a T-Test procedure failed to demonstrate a
statistically significant difference between groups
(T=2.3359, df=4.8, p=.0692).
Prior to initiation of rEPO therapy, mean Hg in the
treatment group was 7.39 g/dl (sd=1.94). At the first
assessment treatment group Hg had risen to 8.46 g/dl
(sd=1.42). Based on a T-Test these differences were not
statistically significant (T=-1.8361, df=32.0, p=.0756).
A GLM analysis of variance procedure revealed a main
effect for sex (F (1, 34)= 15.24, p=.0004). Mean male
(n=17) Hg was 9.96 g/dl (sd=1.61) and mean female (n=18) Hg
was 8.03 g/dl (sd=1.31), which is consistent with the 2 g/dl
lower Hg level reported in normal females compared to normal
males in the general healthy adult population. There was no
treatment by sex interaction.
Excluding the 3 subjects who did not complete the
study, rise in Hg was 0.04 g/dl in the control group (n=15).
As anticipated for the control group, mean change in Hg
between assessments one and three was not statistically
significant (F(l, 32)= 0.07, p=.786). In contrast the rise
in Hg was 1.08 g/dl (sd=1.23) between assessments one and


174
Uyeda, K.M., & Mandler, G. (1980). Prototypicality norms for
28 semantic categories. Behavior Research Methods &
Instrumentation. 12. 587-595.
Van Wyck, D.B. (1989). Iron deficiency in patients with
dialysis-associated anemia during erythropoietin
replacement therapy: Strategies for assessment and
management. Seminars in Nephrology. 9, 21-24.
Vander, A.J., Sherman, J.H., & Luciano, D.S. (1980). Human
Physiology. New York, New York: McGraw-Hill.
Waldmann, T.A., Levin, E.H., & Baldwin, M. (1961). The
association of polycythemia with a cerebellar
hemangioblastoma. American Journal of Medicine. 31. 318-
324.
Walter, T., Kovalskys, J., & Steel, A. (1983). Effect of
mild iron deficiency on infant mental development scores.
Journal of Pediatrics. 102. 519-522.
Wechsler, D. (1987). Wechsler Memory Scale Revised Manual.
New York, New York: The Psychological Corporation.
White, N., & Cunningham, W.R. (1988). Is Terminal Drop
Pervasive or Specific? Journal of Gerontology;
Psychological Sciences. 43. 141-144.
Williams, S.R. (1985). Nutrition and Diet Therapy (5th ed.).
St. Louis, Missouri: Times Mirror/Mosby.
Winearls, C.G., Oliver, D.O., Pippard, M.J., Reid, C.,
Downing, M.R., & Cotes, P.M. (1986). Effect of human
erythropoietin derived from recombinant DNA on the anemia
of patients maintained by chronic hemodialysis. Lancet.
2, 1175-1177.
Wintrobe, M.W., Lee, G.R., Boggs, D.R., Bithell, T.C.,
Foerster, J., Athens, J.W., & Lukens, J.N. (1981).
Clinical Hematology (4th ed.). Philadelphia,
Pennsylvania: Lea & Febiger.
Wolcott, D.L., Marsh, J.T., La-Rue, A., Carr, C., &
Nissenson, A.R. (1989). Recombinant erythropoietin
treatment may improve quality of life and cognitive
function in chronic hemodialysis patients. American
Journal of Kidney Disease. 14, 478-485.
Wolcott, D.L., Schweitzer, S., & Marsh, J.T. (1988).
Recombinant erythropoietin improves cognitive function
and quality of life of chronic hemodialysis patients.
Kidney International. 33. 242. Abstract.


Ill
In summary, insufficient magnitude of Hg change, too
high an initial level of Hg in the present study or a
combination of both conditions may have resulted in the
discrepancy between the pilot and the present study.
However, closer examination of the pilot data in the
context of the present study suggested that the results of
the pilot could have been spurious. The present data set
provided insufficient information to determine which
hypothesis might account for the discrepancy between the
present and the pilot studies. Therefore, subjects were
reclassified to maximized between group differences in Hg
levels.
Observation of change in Hg levels in both groups
suggested the possibility of reclassification of subjects
based on Hg change scores to maximize between group
differences. The symptoms of acute anemia are quite
variable at a given Hg level and, if chronic anemia
followed a similar pattern, then, arguably, collapsing Hg
scores into a two-level classificatory system might
potentially reduce error variance. Unfortunately, only a
small number of subjects exhibited Hg change scores of
sufficient magnitude to justify classification based on
differences of demonstrated physiological significance.
Therefore, the first analysis simply reclassified based on
sign of Hg change.
Reclassification based on sign of Hg change doubled
the mean Hg change score and resulted in a statistically


49
itself is likely to contribute to the cognitive deficits
observed in other forms of anemia, 4) assist in determining
the optimal level of Hg in treated renal patients and 5) in
the event of positive findings, provide preliminary data
regarding the mechanisms through which anemia undermines
neuropsychological functioning.


90
classified based on their sign. A Wilcoxon Rank Sum
procedure analyzed differences in sign of Sum CVLT change
scores between the first and second assessments. With the
sign of Hg change scores serving as the grouping variable,
again there was a significant effect for Hg change
(Z=2.9180, p=.0035). Analysis using a Chi-Square procedure
produced similar results (X2=8.887, df=l, p=.003).
Short Term Memory
Exploratory analysis revealed a significant difference
at .01 in CVLT Trial One, a measure loading heavily on STM.
However, the change was in the opposite of the expected
direction between treatment and control groups between
assessments one and two and is likely a type two error.
Furthermore, equivalent change, also significant at .01, in
the opposite direction, was observed between the second and
third assessments. Please see the results displayed in
Appendix H.
Verbal Fluency
An analysis of variance procedure suggested that verbal
fluency as measured by Controlled Word Association test
performance was somewhat low in the control group at the
first assessment relative to the overall pattern of
performance in both groups across three assessments (F=4.99,
E=.0328. Utilizing sign of Hg change as a classificatory
variable, Controlled Word Association test performance was
not significantly different between assessments (F=0.01,
P=.9761).


48
Failure to replicate would provide information useful
in several areas. The extent to which homeostatic
mechanisms adapt to chronic anemia is unknown. Data
delineating the parameters within which anemia affects
cognitive function could potentially contribute to the
determination of the optimal Hg at which to maintain
patients in end-stage renal disease. This is important
given the monetary costs and medical risks associated with
utilizing rEPO to treat uremic anemia.
Failure to demonstrate Hg responsive neuropsychological
dysfunction in rEPO treated uremic anemia would lend support
to the suggestion that the deficits documented in other
forms of chronic anemia are mediated by factors other than
anemia. In the case of iron deficiency anemia, if the
reported deficits are entirely due to the direct effects of
iron deficiency on the brain, then, since iron deficiency
appears prior to anemia, this would suggest the need for
more aggressive screening for iron deficiency, even in the
absence of anemia. Beyond scientific value, this would
suggest that millions of adults and children may suffer from
easily and inexpensively preventable neuropsychological
dysfunction.
Increased understanding of the neuropsychological
consequences of uremic anemia may 1) increase understanding
of the extent of the organism's ability to adapt to chronic
anemia, 2) contribute to the quarter century quest for the
uremic neurotoxin(s), 3) show the extent to which anemia


159
Table 1-2. California Verbal Learning Test Trials One.
Five. Sum and Sum of Clustering Raw Data by Assessment
Number
Raw Score Raw Score T Score
Raw Score
Trial 1 Trial 5 Sum CVLT Sum Cluster
Assess
1
2
3
1
2
3
1
2
3
1
2
3
Id
1
8
6
7
15
12
13
49
41
35
23
21
14
2
3
5
5
10
11
11
14
12
11
4
5
8
3
9
8
8
12
10
10
51
44
38
4
10
3
4
5
4
4
10
11
10
6
12
5
11
11
7
5
6
5
5
15
13
10
40
33
32
15
8
21
6
4
6
7
7
9
8
17
25
22
7
14
8
7
5
4
8
10
13
15
22
26
46
12
14
15
8
5
4
6
9
12
13
26
26
39
9
6
13
9
7
7
5
14
15
12
50
51
29
22
20
15
10
10
6
8
14
16
14
49
59
52
30
33
35
11
9
7
7
13
14
14
40
43
33
14
17
13
12
3
0
3
4
7
7
14
19
26
1
8
2
13
4
3
8
12
13
11
34
42
40
4
12
5
14
5
5
6
13
10
10
49
37
37
12
11
9
15
4
4
3
9
7
11
30
19
34
6
1
11
16
10
7
9
16
16
16
59
53
61
46
37
49
17
3
2
3
8
3
4
25
16
24
5
2
6
101
5
4
6
10
9
10
33
35
35
10
7
11
102
5
4
5
10
8
7
27
19
24
14
5
10
103
8
6
6
13
11
8
30
19
20
17
12
12
104
3
6

8
8

31
40

4
3

105
4
5
5
9
9
9
34
42
34
14
12
10
106
5
9

9
12

37
49

5
15

107
6
7
6
10
12
9
34
40
37
9
11
6
108
7
7
9
13
11
14
43
41
54
15
11
15
109
6
7
7
13
13
14
49
44
51
11
16
33
110
8
5
5
14
11
12
55
36
36
25
9
16
111
8
10
6
10
9
11
39
43
43
7
11
10
112
3
5
4
5
6
9
26
34
36
0
4
2
113
8
6
5
5
13
14
22
50
54
4
7
12
114
8
8
9
16
14
14
43
45
56
29
22
23
115
5
9
7
15
16
14
27
56
35
14
44
28
116
5
10
5
14
9
11
38
11
12
23
10
117
5
7

1
14

26
43

3
15

118
4
5
3
9
6
10
28
20
26
10
3
9


86
was rerun excluding one possible outlier; however, this had
little effect on the results (F=7.82, p=.0087). The results
were consistent with the original finding (i.e., using
treatment status as the classificatory variable) in that
they were in the opposite of the predicted direction. In a
GLM multiple analysis of variance treatment status was
nonsignificant (F=0.59, p=.4482) while sign of Hg change was
significant (F=9.45, p=.0043). Also consistent with the
original finding, as illustrated in Table 4-8 below, mean
Sum CVLT scores revealed poorer performance at the time of
the first assessment in the group whose Hg levels would
rise.
Table 4-8. Mean Sum CVLT by Sian of Ha Change Between
Assessments 1 & 2
Hg Fell n=12 Hg Rose n=23
Asses Mean Sd Mean Sd
1 36.58 9.48 48.61 10.86
2 43.83 11.80 47.39 12.07
In an attempt to reduce error variance contributed by
subjects exhibiting very small insignificant changes in Hg,
subjects with a Hg change of under 1 g/dl were excluded from
the analysis. Mean Sum CVLT change in the negative Hg
change group was 9.33 (sd=10.63) and -2.23 (sd=4.30) in the
positive Hg change group. A Wilcoxon Rank Sum Test
revealed the difference to reach statistical significance at
the .05 level (Z=2.373, p=.0176).


109
may be of greater physiological significance in more
severely anemic individuals. Conceivably, change in Hg
equivalent to that observed in the present study, but
occurring at a lower baseline level, might have resulted in
significant cognitive effects. The present study may have
failed to demonstrate cognitive change because baseline Hg
levels were above the threshold for cognitive dysfunction.
Indeed, based on the literature review, it is quite
conceivable that the nervous system is able to adapt to all
but severe anemia.
Despite plausible explanations for a failure to
replicate, re-examination of the pilot data in the light of
present findings suggests the possibility that the results
of the pilot study may have been spurious.
In the pilot study, four of the five treated subjects
had Hg increases of over 6 g/dl. In contrast, the greatest
Hg rise in the present study was 5.90 g/dl. However,
significance was reached in the pilot study only with the
contribution of the putative relationship between negative
LTM change scores in the control group and negative Hg
change in the same two subjects. The greatest fall in Hg
was 0.93 g/dl. The current study had 19 subjects with
changes in Hg of 1.0 g/dl or greater between assessments
one and two. Re-examination of the Pilot study data
revealed that mean change in the control group was -5.5
words recalled over five trials and 5.6 words in the
treatment group. Thus, nearly 50% of the change in the


9
(Rundus, 1971). This model explains the observation that,
on a supraspan task, a higher percentage of the first
(primacy) and last (recency) words tend to be recalled. The
primacy effect is believed to be due to encoding in LTM.
The recency effect is hypothesized to be a manifestation of
STM. Thus, immediate recall on a supra-span task may be
examined for primacy and recency effects to provide
indications of the integrity of STM and LTM.
Besides dissimilarities in capacity and duration, there
appear to be differences in the cognitive processes
requisite to the maintenance of STM and LTM. In general,
most studies and clinical observations have suggested that
phonological or maintenance rehearsal (i.e., mere parroting)
usually serves only to maintain information in STM, whereas
more complex, elaborative or semantic processing facilitates
transfer to LTM (Craik & Lockhart, 1972). Transfer to LTM
has been thought to usually require more active attending
and processing.
Demonstration of the differential effects of type or
level of processing led to the proposal that the observation
of apparently distinct memory stores could be explained in
terms of levels of processing within one memory store (Craik
& Lockhart, 1972). The apparently limited storage capacity
of STM has been reinterpreted by the levels of processing
model as limited processing capacity. The apparent
distinction between STM and LTM has been explained by the
levels of processing model as partially a function of depth


APPENDIX E
COMPARISON OF PARAMETRIC AND NONPARAMETRIC TESTS OF BASELINE
MEASURES BETWEEN GROUPS
Kruskal-Wallis Test
General Linear
(Chi-Square Approximation)
Nonparametric
Model
Parametric
Df=l Prob >
Variable Chi-Square Chi-Square F Value Pr >F
Hemoglobin
3.0616
.0802
2.94
.0959
Hematocrit
3.1778
.0746
2.29
. 1401
Ferritin
0.3478
.5553
0.00
.9483
Age
1.1868
.2760
1.13
.2948
Education
0.06695
.7958
0.23
.6325
Dialysis years
2.8425
. 0918
3.51
.0699
California Verbal
Learning
Test
Trial 1
0.0045
.9464
0.05
.8253
Standard Score
0.0855
.7700
0.00
.9602
Trial 2
0.1244
.7243
0.50
.4847
Trial 3
0.2496
.6173
0.21
.6535
Trial 4
1.3349
.2479
0.68
.4140
Trial 5
0.4435
.5054
0.40
.5291
Standard Score
0.0045
.9464
0.11
.7400
Sum CVLT
0.3157
.5742
0.42
.5222
Standard Score
0.0221
.8817
0.03
.8662
Clusters
0.0175
.8947
0.36
.5519
Perseverations
0.0805
.7765
0.38
.5442
Intrusions
0.1496
.6989
0.01
.9409
List B
0.0914
.7623
0.09
.7687
Standard Score
0.0012
.9724
0.10
.7587
Short Delay
0.0002
.9867
0.34
.5624
Standard Score
0.0002
.9864
0.02
.8951
Clusters
0.2849
.5935
0.00
.9760
Perseverations
2.7324
.0983
4.67
.0380*
Intrusions
0.2658
.6061
0.14
.7086
Short Delay Cued
0.1107
.7393
0.00
.9973
Standard Score
0.0141
.9052
0.01
.9053
Perseverations
3.3690
.0664
2.62
.1151
Intrusions
0.4843
.4865
0.04
.8343
144


34
of an interaction between factors affecting cognitive
function should not be discounted. Young adults one to
three years post mild head injury (i.e., concussion) exhibit
impairment on immediate memory and vigilance tasks relative
to controls when all subjects are tested under mildly
hypoxic conditions (Ewing, McCarthy, Gronwall, & Wrightson,
1980). Renal patients may be more susceptible to the
effects of anemia than are otherwise healthy anemic
patients.
In contrast to the unknown incidence of hypoxia and
hyperventilation, fatigue, the first and most chronic
symptom of anemia, is almost invariably present in end-stage
renal disease (Dawson, Ogston, & Fullerton, 1969). Fatigue
may be expected to consistently occur prior to the
development of cerebral hypoxia, if cerebral hypoxia occurs
at all. At a level of anemia severe enough to produce
cerebral hypoxia, it appears likely that the level of
fatigue will also be greatly increased. Therefore, fatigue
may be a more reliable mediator of the putative
neuropsychological effects of anemia.
The effects of fatigue on neuropsychological
performance are well known in clinical practice (Lezak,
1983). However, the effects do appear to be variable.
Subject factors such as age and physical condition and
fatigue factors such as cause and duration may be
significant. No changes in neuropsychological test
performance were demonstrated in 42 surgical residents when


120
Unfortunately, several interrelated factors weakened
the design of the present study. Initial plans called for
integration of this project into a larger study examining
the clinical efficacy and safety of rEPO prior to U.S. Food
and Drug Administration approval. However, approval of
rEPO for clinical use was unexpectedly expedited. This
resulted in reduced control over the independent variables.
Hemoglobin change scores were much smaller than in the
pilot study. Indeed, follow-up revealed that over one year
after initiation of rEPO, the majority of subjects were
still moderately anemic. Because the experimenter was
blind to Hg levels until completion of neuropsychological
testing, the existence of only minimal Hg change scores was
undetected until completion of the study.
In the Pilot study, data collection was coordinated
with the timetable of a multicenter clinical trial.
Baseline assessment occurred prior to any EPO mediated rise
in Hg. Approval for clinical use resulted in the majority
of hemodialysis subjects being placed on rEPO, many prior
to baseline assessment. A reduced pool of rEPO-free
hemodialysis subjects necessitated the inclusion of
peritoneal dialysis subjects as controls. Although
undesirable, the impact of a nonequivalent control group
was minimized by the design. The controls merely served to
control for the effects of multiple testing with different
test forms. The approval of rEPO raised the possibility
that suitable rEPO-free long-term dialysis subjects would


114
mechanisms, cognitive deficits might be expected to appear
prior to rise in Hg. As previously mentioned, analysis
using a model containing either treatment status or Hg
change suggested a possible trend; however, a model
containing both terms failed to shed light on the "direct
effect of EPO on the brain" hypothesis. Following
reclassification of subjects based on direction of Hg
change, a multivariate analysis demonstrated treatment
status to be unambiguously noncontributory (F=0.59,
P=.4482), while direction of Hg change was significant
(F=9.45, p=.0043). Thus, the post hoc hypothesis that EPO
directly and deleteriously affects the brain through a
channel unrelated to Hg finds no support in this data.
Cognitive dysfunction might hypothetically be mediated
by cerebral vascular mechanisms. Recombinant EPO mediated
increases in Hg frequently produce undesirable effects
including headache and hypertension (Delano, 1989; Hori,
Onoyama, Iseki, Fujimi, & Fujishima, 1990). Increased
platelet aggregation may further increase risk of
cerebrovascular accidents. Although these possibilities
could not be tested within the confines of the present data
set, it may be noted that Hg rise was modest compared to
clinical trials where the effect of rEPO on clotting time
and blood pressure was studied.
In severe anemia, the need for greater cardiac output
results in increased sympathetic activation (Hori, Onoyama,
Iseki, Fujimi, & Fujishima, 1990). The effect, if any, of


63
dropping of the third assessment from the preplanned
analyses. Therefore, the first test of the main hypothesis
was a GLM univariate procedure with treatment status as the
independent variable and Sum CVLT change score as the
dependent variable.
The second planned test was a GLM analysis of
covariance, the added independent factors being the
classificatory variable iron status and the continuous
variable blood Hg level. In actuality, change in iron
status affected only a small number of subjects; therefore,
controlling for iron status became of minimal importance.
Kruskal-Wallis nonparametric procedures were used to
test the hypotheses that self-reported physical vitality and
self-reported cognitive behavior improved in the treatment
group. If a main effect for Hg on Sum CVLT had been found
then further nonparametric analyses would have been
performed in an attempt to provide preliminary data to guide
future inquiry into the mechanisms mediating the effect.
Specifically, between group differences in ability to
benefit from semantic processing and changes in encoding of
incidental information were planned. Negative findings
eliminated this rationale for these tests. Nonetheless,
they were included in post hoc analyses.
Post hoc exploratory analyses focused on attempts to
disconfirm the negative findings and to explore other
aspects of the data. The observation that treatment status
was a poorer predictor of Hg change than had been


173
Souheaver, G.T., Ryan, J.J., & DeWolfe, A.S. (1982).
Neuropsychological patterns in uremia. Journal of
Clinical Psychology. 38. 490-496.
Sperling, G. (1960). The information available in brief
visual presentations. Psychological Monographs. 74 (whole
no. 498).
Sprague, S.M., Corwin, H.L., Tanner, C.M., Wilson, R.S.,
Green, B.J., & Goetz, C.G. (1988). Relationship of
aluminum to neurocognitive dysfunction in chronic
dialysis patients. Archives of Internal Medicine. 148.
2169-2172.
Squire, L.R. (1987). Memory and Brain. New York, New York:
Oxford University Press.
Stewart, R.S., & Stewart, R.M. (1979). Neuropsychiatric
aspects of chronic renal disease. Psychosomatic. 20. 524-
527.
Teschan, P.E., Ginn, H.E., Bourne, J.R., & Ward, J.W.
(1976). Neurobehavioral responses to "middle molecule"
dialysis and transplantation. Transactions of American
Society of Artificial Internal Organs. 22. 190-194.
Teschan, P.E., Ginn, H.E., Walker, P.J., Bourne, J.R.,
Fristoe, M., & Ward, J.W. (1974). Quantified function in
uremic patients on maintenance dialysis. Transactions of
American Society of Artificial Internal Organs. 20, 388-
394.
Thorndike, E.L., & Lorge, I. (1944). The teacher/s word book
of 30.000 words. New York, New York: Teachers College.
Trieschmann, R.B., & Sand, P.L. (1971). WAIS and MMPI
correlates of increasing renal failure in adult medical
patients. Psychology Report. 29. 1251-1262.
Trompeter, R.S., Polinsky, M.S., Andreoli, S.A., & Fennell,
R.S. (1986). Neurological complications of kidney
failure. American Journal of Kidney Diseases. 7, 318-323.
Tucker, D.M., Sandstead, H.H., Penland, J.G., Dawson, S.L.,
& Milne, D.B. (1984). Iron status and brain functions:
serum ferritin levels associated with asymmetries of
cortical electrophysiology and cognitive performance.
American Journal of Clinical Nutrition. 39. 105-113.
Tyler, H.R. (1968). Neurologic disorders in renal failure.
American Journal of Medicine. 44., 734-748.


52
Measures
California Verbal Learning Test
The California Verbal Learning Test (CVLT) is a multi-
trial supra-span word list learning task designed to be an
ecologically valid measure of multiple aspects of learning.
The CVLT quantifies parameters including STM, LTM,
retention over short and long delays, degree of
vulnerability to proactive and retroactive interference,
encoding strategies, effects of category cuing and the
frequency of perseverations and intrusions (Delis, Kramer,
Kaplan, & Ober, 1987).
Three versions were used. Form I is in general
clinical use. Form II is an alternate form developed for
research purposes (Delis, McKee, Massman, Kramer, &
Gettman, 1990). Form III, the Florida version of the CVLT,
was developed for this study. A description of the
construction and validation is provided in Appendix A. In
keeping with the results of the pilot study the primary
measure was the Sum of trials one through five of the CVLT
(Sum CVLT).
Levels of Processing Task
A 36 word levels of processing task in use in the
Shands Teaching Hospital Psychology Clinic was used and two
alternate forms were developed for this study. Alternate
forms two and three were designed with target words of
similar frequency of occurrence to that of Form I. Please
see Appendix B.


142
Table D-l continued
21. My future looks bright.
22. I feel tired all the time.
23. Everything is an effort.
24. I enjoy visiting friends.
25. I do heavy work around the house.
26. I do alot of cooking.
27. Simple chores quickly tire me out.
28. My muscles quickly get tired.
Table D-2. Instructions for Scoring: Corrections
If 13 is endorsed then consider 10 endorsed.
If 19 is endorsed then consider 3 endorsed.
If 22 is endorsed then consider 15 endorsed.
Table D-3. Cognitive-Affective Scale Scoring Instructions
Starting with 9 points, subtract one point for each of the
following if endorsed:
5. I feel useless, like I am a burden on others.
7. I have difficulty concentrating.
9. I am irritable and impatient with myself.
16. I do not finish things I start.
17. I react slowly to things that are said or done.
Subtract one point for each that is not endorsed:
8. My sleep is sounder than it used to be.
20. I enjoy eating.
21. My future looks bright.
24. I enjoy visiting friends.


130
Randall, & Georgemiller, 1986; McCarty, Logue, Power,
Ziesat & Rosenstiel, 1980).
Sum of Trials 1-5 and Trial 1 both worked well and
they are measures of special interest in the study for
which Form III was designed. In contrast, many of the
other correlations are rather low, especially Trial 5 (r =
.10) and Long Delay Cued Recall (r = .14). Also, in every
instance except for Trial 1, correlation coefficients
between Forms I and II are higher, sometimes markedly
higher. However, it is likely that the low correlations
for some measures are part of a general suppression of all
correlation coefficients due to the homogeneity and very
high functioning of the subject pool. Indeed, recognition
measures could not be calculated because nearly every
subject had a perfect score. Likewise, a mean of 15.1 for
the fifth trial on Form III clearly illustrates the problem
of a ceiling effect. Higher alternate form correlation
coefficients might be expected with a more representative
cross-section of the population.
Construction of this version of CVLT Form III is
deficient in that some target words have prototypicality
ratings (Battig and Montague, 1969) of 4 or lower. This
theoretically increases the possibility that correct
responses may occur by chance in subjects prone to
semantically related intrusions. However, the present data
suggests CVLT Form III to be an excellent alternate form.


UNIVERSITY OF FLORIDA
1262 08554 8112


126
TABLE A-2. Form III Word Frequency and Typicality Values
Primary List
a b c
Interference
a b
List
C
Dresser
7
45.5
8
Rocker
2
44.1
36
Rug
40
51.5
12
Bureau
44
45.3
11
Couch
28
47.1
7
Stool
16
49.2
10
Bookcase
3
40.5
14
Cabinet
26
48.6
15
Orchids
3
43.8
6
Carnation
1
36.6
3
Lilies
33
45.3
8
Pansies
7
42.5
9
Gardenias
1
29.6
11
Daisies
28
44.1
4
Daffodils
3
40.7
12
Geraniums
4
39.6
16
Sapphire
6
39.4
4
Liver
10
48.7
-
Pearl
47
47.7
5
Bacon
12
50.7
-
Garnet
2
37.2
11
Ham
17
49.8
-
Jade
4
42.8
7
Steak
14
46.2
-
Socks
12
50.3
4
Silk
75
55.1
3
Slippers
20
47.5
5
Flannel
9
44.2
15
Sandals
5
44.8
3
Burlap
1
41.1
12
Boots
37
47.5
2
Denim
1
41.1
13
Mean
15.
7
43.8
8.1
16.7
45.4
12.
sd
15.
9
5.5
4.3
19.5
4.8
8.
Note: word frequency (Thorndike & Lorge, 1944)
b = word frequency (Carroll, Davies & Richman, 1971)
c cat rank = typicality ranking (Battig & Montague,
1969); M = mean, sd = standard deviation


35
assessed following 2.0 +1.5 hours of sleep vs. 6.5 +1.0
hours. In the acute sleep-deprived state, residents were
less vigorous and more fatigued, depressed, tense, confused
and angry (p<0.05) than they were in a rested state. These
changes in mood had no effect on measures including
immediate and delayed recall of logical stories, Trails,
immediate memory, or digit symbol (Bartle et al., 1988).
However, it is conceivable that more difficult tasks would
have shown an effect.
The biological function of the experience of fatigue is
hypothesized to be reduction in unnecessary activity. It is
hypothesized that the effects of increasing fatigue are
likely to appear along a continuum. With increasing fatigue
the organism's proclivity to define activity as necessary is
likely to decrease. In a testing situation, prior to
refusal to perform a task, the subject's definition of the
task may change to make the task less taxing. Therefore,
tasks not requiring, but enhanced by, additional effortful
cognitive activity may be expected to be most sensitive to
the effects of fatigue. Tasks overtly demanding attention,
vigilance and/or extensive cognitive processing are
hypothesized to be affected only as the severity of the
fatigue increases.
Effect of Erythropoietin on Neuropsychological Function
The effects of rEPO replacement therapy on quality of
life and neuropsychological functioning will be briefly


CHAPTER 1
INTRODUCTION: FOCUS AND RATIONALE
The fourth leading health problem in the developed
world, renal disease has been estimated to affect
approximately 8 million people in the U.S. (Williams, 1985).
Uremia, the life threatening condition resulting from renal
failure, has been known since ancient times. A variety of
symptoms have been associated with uremia, the majority of
which have been reported to be secondary to nervous system
dysfunction (Ginn et al., 1975). Memory dysfunction has
been among the neuropsychological deficits commonly noted.
Despite considerable effort, attempts to identify the
factors responsible for these deficits have achieved little
success (Powell et al., 1986; Schreiner, 1975; Trompeter,
Polinsky, Andreoli, & Fennell, 1986).
Chronic anemia has been almost universally observed in
chronic renal failure. The primary cause has been generally
accepted to be insufficient production of erythropoietin
(EPO), a erythropoiesis stimulating factor produced by the
kidneys. Recent studies have documented the dramatic
effects of recombinant erythropoietin (rEPO) on many
physical symptoms until recently attributed to uremia, but
now demonstrated to have been at least partially caused by
anemia (Nissenson, 1989). One rationale for the present
1


93
Table 4-10. Self-reported Cognitive-Affective State Change
Scores bv Group & Assessment Interval
Test Control Group Treatment Group
Interv N Mean Sd N Mean Sd
One 18 0.94 2.21 16 -0.19 1.33
Two 15 -1.73 2.02 16 -0.13 1.41
Note: Interv = Change between assessments one and two or
between two and three.
Test Reliability
Two alternate forms of the Levels of Processing Task
and Frequency of Occurrence Task were developed specifically
for this study. A self-report questionnaire was written to
sample physical and cognitive-affective domains thought
likely to be affected by change in severity of anemia. None
of these new measures underwent a prior validation study.
Also, one alternate form of the CVLT was developed and
validated for use in this study, but validated on a younger
and better educated population. Because the only apparent
changes in test performance were in the controls, controls
which might have been used in a test validation study, the
lack of clear-cut findings produced the opportunity to
examine the reliability of these instruments in a medical
population.
CVLT Alternate Form Reliability
As displayed in Table 4-11, for the overall group the
correlation between Forms I and III and between Forms III
and II was quite good relative to correlations between Forms


129
respectively. The correlation between forms for Trial 1
was r = .70. The mean interference trial scores of 9.0 and
8.9 for Forms I and III, respectively, were nearly
identical and produced a correlation of r = .56. It is
desirable that the interference trial be of equivalent
difficulty to the primary list as measured by Trial 1 of
the primary list. The CVLT Form III primary list appeared
slightly but not significantly easier. The inter-form
correlation for Trial 5 was r= .50; however, means were
15.2 and 15.1. Please see Table A-3 for a tabular listing
of the results.
In Table A-3, the correlations between Forms I and II
were reported by Delis, McKee, Massman, Kramer and Gettman
(1990).
Discussion
A comparison of mean scores between the two forms
suggests that the two versions are quite similar. Indeed,
this is illustrated by summing the 10 primary measures.
Mean word recall for Form I was 134.6 and 131.1 for Form
III. As this is out of a total of 160 words the 3.5 word
recall advantage for Form I is rather insignificant. The
correlations for Sum of Trials 1 through 5 (.69) and Trial
1 (.70) compare favorably with the alternate form
reliability coefficients reported for the Rey Auditory
Verbal Learning Test, which ranged from .60 to .77 and with
Russell's revised Wechsler Memory Scale, which had
correlations ranging from .60 to .74 (Ryan, Geisser,


72
Mean days between assessments two and three was 67.80
(sd=34.56) in the control group and 90.41 days (sd=24.03) in
the treatment group. Based on the results of a Wilcoxon
Rank Sum Test the 22.61 days greater inter-test interval for
the treatment group was clearly statistically significant
(S=178.50, Z= -2.588, £>=.0096).
The number of days between assessments one and three
was 141.53 days (sd=34.53) in the treatment group (n=17) and
117.67 days (sd=28.24) in the control group (n=15). The
results of a Wilcoxon Rank Sum test suggest that the 23.86
day difference between these two groups was statistically
significant (S=181.00, Z= -2.495, p>.0126). Mean time lag
between initiation of rEPO and the first assessment was 41.4
days (sd=32.1)
Hemoglobin
Mean Hg at the first or baseline assessment was 8.46
g/dl (sd=l.42) in the treatment group and 9.44 g/dl
(sd=1.92) in the control group. The nature of these
differences suggested a possible trend toward statistical
significance as demonstrated by a GLM Analysis of Variance
procedure (F (1, 34)=2.94, p=.096). which is consistent
with reports of lower Hg levels in hemodialysis than in
peritoneal dialysis. Baseline mean hematocrit levels were
28.19% (sd=6.12) in the control group and 25.58% (sd=3.77)
in the treatment group. As demonstrated by a GLM ANOVA
procedure these differences failed to reach statistical
significance (F(l, 34)=2.29, £=.140).


47
study found improvement only in effortful verbal LTM.
Furthermore, in retrospect, it was noted that the sometimes
reported advantage of peritoneal over hemodialysis and
decline in performance with time on hemodialysis may be at
least partially mediated by degree of anemia. Also, there
are suggestions that uremic memory dysfunction is at least
partially mediated by reduced vigilance, underarousal,
increased distractibility and cognitive slowing. These
observations are consistent with the hypothesis that
effortful verbal memory may be vulnerable to anemia,
secondary to mechanism such as fatigue or hypoxia.
The goal of the present study was to attempt to
replicate the improvement in LTM on a a multitrial supra-
span learning task which had been previously shown to
correlate with reduction of uremic anemia. Previous work
suggested that verbal memory tasks making fewer demands on
effortful cognitive processing showed no improvement. If
improvement were demonstrated on effortful memory tasks as
well as on automatic memory tasks and on the Levels of
Processing task, then the most plausible explanation would
be that uremic anemia deleteriously affects memory
structures, possibly mediated by hypoxia, although other
explanations might be posited. Improvement only in
effortful memory would suggest that the change was due to
increased cognitive processing, possibly secondary to
reduced fatigue or increased activation.


137
Table
B-4. Multiple Choice
Retrieval
Ouestions
Version III
Multiple Choice
Word List
Processing
Level
1.
hill
coin
grass
nail
SEM-
2.
clip
wood
TOWEL
fern
0RTH+
3.
globe
garden
clamp
scar
PHON-
4.
rug
haze
gift
bath
PH0N+
5.
pencil
foot
mast
sheet
ORTH-
6.
tear
violin
page
horse
SEM+
7.
palace
cream
oak
fear
PH0N+
8.
path
coal
arm
tent
SEM-
9.
strawberry
bat
ghost
lap
SEM+
10.
moss
college
bed
cart
ORTH+
11.
trout
egg
doll
plate
PHON-
12.
string
paste
star
cup
ORTH-
13.
band
cord
shore
tune
ORTH+
14.
fur
finger
twig
mate
SEM+
15.
roll
chin
pork
bed
PHON+
16.
peel
key
steak
purple
SEM-
17.
plot
animal
weed
cage
ORTH-
18.
crumb
limb
mile
ash
PHON-
19.
wheel
lie
spade
white
SEM+
20.
peddle
leaf
knife
nob
ORTH+
21.
deck
roam
spear
art
ORTH-
22.
bullet
clock
cat
oar
PH0N+
23.
bone
pack
rock
canoe
SEM-
24.
dime
grass
book
mire
PHON-
25.
juice
rope
salt
acre
SEM+


135
Table B-3. Levels of Processing Version III Encoding
Questions
Visual Stimuli
Ouestion to be orocessed
1.
COLLEGE
Is this printed in upper case
letters?
2.
finger
Is this part of the body?
3.
garden
Does this rhyme with match?
4.
hamster
Is this a type of furniture?
5.
clock
Does this rhyme with sock?
6.
ship
Is this printed in upper case
letters?
7.
animal
Is this printed in upper case
letters?
8.
violin
Is this a type of musical instrument?
9.
robot
Does this rhyme with fall?
10.
palace
Does this rhyme with chalice?
11.
canoe
Is this a type of flower?
12.
LAKE
Is this printed in upper case
letters?
13.
salt
Is this a seasoning?
14.
hill
Is this a type of animal?
15.
spike
Does this rhyme with bike?
16.
star
Is this printed in upper case
letters?
17.
mile
Does this rhyme with hour?
18.
KNIFE
Is this printed in upper case
letters?
19.
TOWEL
Is this printed in upper case
letters?
20.
dime
Does this rhyme with board?
21.
shovel
Is this a type of tool?
22.
bed
Does this rhyme with head?
23.
strawberry
Is this a type of fruit?
24.
deck
Is this printed in upper case
letters?
25.
pencil
Is this printed in upper case
letters?


62
between group equality of variance at baseline. To test the
impact of these violations of the assumptions of parametric
procedures, parametric and nonparametric tests of between
group baseline values were calculated for all subjects and
the results compared. As displayed in Appendix E, even
using an alpha of .05, only 2 tests out of 55 variables,
both highly unstable and insignificant variables, produced
discrepant results. It was concluded that the violations of
normality were unlikely to be of sufficient severity to
impact on the validity of SAS General Linear Model
parametric procedures, except for ferritin. Since ferritin
was used as a classificatory variable inequality of variance
had no appreciable effect on statistical procedures.
Demographic and descriptive statistics were calculated
using the SAS Means and Frequency procedures. To rule out
extraneous between group differences baseline variables were
compared. It was planned that the main hypothesis would be
tested with a repeated measures ANOVA using the General
Linear Model procedure.
A second analysis was planned to test the hypothesis
that the expected treatment effect was associated with Hg
change rather than rEPO itself and that the treatment effect
was not an artifact of variability in iron status. Although
a repeated measures analysis with the classificatory terms
group membership and iron status and the covariant Hg had
been planned, the unexpected finding of no change in Hg
status between the second a third assessments led to the


17
verbal task requiring the learning of a list (e.g.,
Buschke), susceptibility to distraction (i.e., Auditory
Consonant Trigrams) and reduced sustained vigilance
(Fennell, Fennell, Mings, & Morris, 1986). This led to the
suggestion that reduced level of arousal may mediate STM
impairment in renal disease (Fennell et al., 1990a).
Because they place greater demands on sustained attention,
multitrial memory tasks may be more sensitive to uremia.
Overall, initiation of hemodialysis has been reported
to improve memory function (Osberg, Meares, McKee, &
Burnett, 1982). In adults, the consequences of uremia are
largely reversible and appear to be only minimally
cumulative in the absence of confounds such as aluminum
toxicity, dietary noncompliance, uncontrolled hypertension
or diabetes. Comparing dialysis patients and controls,
Hart, Pederson, Czerwinski and Adams (1983) found relatively
little difference in memory as measured by the Wechsler
Memory Scale (WMS). In addition, no significant
correlations were found between years of dialysis treatment
and memory performance.
Gilli and DeBastiani (1983) reported (n=54) a mild
relationship between duration of hemodialysis and decrements
in verbal intellectual ability (WAIS) and memory (WMS). The
WMS Memory Quotient, a composite measure of primarily short
term verbal and nonverbal memory, was initially above
expectations based on WAIS performance. For the 21
hemodialysis subjects who were retested, a minimum of 12


26
McVicar, & Clemons, 1988). Intravenously administered rEPO
has been demonstrated to restore Hg levels to normal in many
renal patients (Eschbach, Egrie, Downing, Browne, & Adamson,
1987; Winearls et al., 1986).
Function of Hemoglobin and Adaptation to Anemia
Under normal atmospheric and physiologic conditions,
more than 98% of the total oxygen carried by arterial blood
is bound to Hg, with the remainder dissolved in the aqueous
portion of blood (P02). To cross cell membranes, oxygen
must first disassociate from Hg and dissolve directly in the
blood. The function of Hg may be considered to be the
maintenance of normal P02 by providing a large reservoir of
available oxygen. Thus, in anemia the reservoir of oxygen
is reduced.
Anemic organisms may maintain normal tissue oxygenation
through a variety of adaptive mechanisms. Blood flow is
deviated to more vital areas. The affinity of Hg for oxygen
decreases. When Hg falls below 7.5 g/dl, an increase in
both heart rate and stroke volume results in a significant
rise in resting cardiac output (Bunn, 1980b) mediated by
increased sympathetic activation. Although PC02 is a more
potent controller of ventilation than is P02, low P02
resulting from severe anemia may stimulate peripheral
chemoreceptors on the carotid bodies resulting in increased
ventilation. However, the resulting reduction in PC02 and
developing alkalosis are likely to attenuate this response
(Vander, Sherman, & Luciano, 1980).


13
Neuropsychological Dysfunction in Adult Renal Patients
This section provides a cursory overview of the general
pattern of deficits seen in renal disease and discusses
verbal memory within the context of broader findings.
General Intelligence: Verbal and Nonverbal
In the absence of confounds, general verbal
intellectual ability has been reported to be relatively
spared in adult renal disease (Blatt & Tsushima, 1966;
Comty, Leonard, & Shapiro, 1974; English et al., 1978;
Fishman & Schneider, 1972; Freeman, Sherrard, Carlsyn, &
Paige, 1980; Ryan, Souheaver, & DeWolfe, 1980; Schupak,
Sullivan, & Lee, 1967; Trieschmann & Sand, 1971). In a 1982
review, Osberg, Meares, McKee and Burnett noted that in
every reported study, verbal IQ scores surpassed performance
scores by 5 to 14 points. The verbal-nonverbal discrepancy
is partly an artifact of the time constraints on many
nonverbal but not verbal intelligence subtests; however,
visuospatial and visuomotor deficits have also been widely
observed on untimed tasks (Mings, 1987).
The relative sparing of language function is consistent
with the general pattern of deficits often seen in diffuse
brain dysfunction. Speculations regarding the cause of this
verbal-nonverbal disassociation have included: the
nondominant hemisphere is more vulnerable to assault,
overlearned verbal abilities are more redundantly
represented and crystallized intellectual functions tax
cognitive capacity less than more novel visuospatial tasks.


134
Table B-2 continued
Processing
Multiple Choice
Word List
Level
26.
silk
hatchet
fog
brass
SEM-
27.
dog
lamp
wave
knot
PHON-
28.
cold
fad
gear
tan
ORTH-
29.
surprise
cop
prey
camp
0RTH+
30.
soap
pond
trot
chart
PH0N+
31.
float
priest
film
ant
PHON-
32.
paint
hinge
mute
burn
0RTH+
33.
swine
ditch
surf
ache
SEM-
34.
horse
seem
ramp
rain
ORTH-
35.
floor
hen
stamp
pad
PH0N+
36.
spark
pear
grip
site
SEM+
Note:
+ = Correct
response
is positive;
- = Correct
response
is negative;
ORTHO =
Orthographic; PHON = Phonetic; SEM = Semantic


103
with the original forms. However, the Physical Activity
scale on the self-report questionnaire developed for this
study exhibited excellent test-retest reliability. Test-
retest reliability was somewhat lower for the Cognitive-
Affective scale.


18
months later the WMS Memory Quotient had dropped to a level
only slightly above expectations for IQ. The authors
interpreted these results as suggesting a decline in memory
function in association with time on hemodialysis, perhaps
related to elevated parathyroid hormone levels. However,
fewer than half of the subjects were retested, and they
exhibited initial Memory Quotient scores above the average
for the entire subject pool and, as previously noted, above
expectations for IQ. The apparent decline in memory may
have been an example of regression towards the mean.
Ginn (1975) reported a temporary improvement in verbal
LTM on a recognition task the day after hemodialysis,
regardless of the level of performance prior to dialysis. A
correlation of .687 (pc.001) was reported between latency to
response on a single trial word-recognition task and
severity of uremia (Ginn et al., 1975). However, increased
response latency does not necessarily indicate retrieval
difficulties, especially in a population suspected of
generalized slowing.
Heilman, Moyer, Melendez, Schwartz and Miller (1975)
found significant verbal STM deficits on the Logical Memory
subtest of the WMS. In contrast, loss between immediate and
delay was comparable to controls. Ginn and colleagues (1975)
reported a significant negative correlation (pc.OOl) between
response latency on a verbal recognition memory task and
uremia as measured by serum creatinine. In nondialized
uremic subjects (n=23) participating in a longitudinal


46
in severe chronic anemia. This adaptive reaction is
mediated by sympathetic activation. Recombinant EPO, by
reducing anemia, reduces demands for high cardiac stroke
volume and heart rate; therefore, may reduce sympathetic
activation (Hori, Onoyama, Iseki, Fujimi, & Fujishima,
1990). This could be of significance given suggestions of
neuropsychological underactivation in uremia.
Summary and Conclusions
Memory dysfunction has been reported in chronic renal
disease with reasonable consistency. Chronic renal disease
is characterized by a host of symptoms, many of which stem
from nervous system dysfunction. Over the past 25 years
numerous studies have examined cognitive dysfunction in
uremia. These studies have occurred within the context of
medicine's increasing ability to simulate, or in the case of
transplantation, restore normal renal function. However,
today, the physiological mechanisms and cognitive processes
underlying neuropsychological dysfunction in renal disease
are still largely unknown.
Recombinant EPO, a potent new treatment for the anemia
of renal failure, has produced dramatic improvements in work
capacity, sexual potency and quality of life in impressive
percentages of treated patients. Because many of the
symptoms once attributed to uremia now appear to be
dramatically reduced through the treatment of uremic anemia,
it seemed reasonable to hope that neuropsychological
dysfunction might be similarly affected. A small pilot


157
Control Group Treatment Group
Variable
Tm
N
Mean
Sd
N
Mean
Sd
Estimation
of
Frequency of Occurrence
Recall
1
17
0.65
1.90
16
1.62
3.07 a
Recall
2
15
0.47
1.73
16
0.00
2.25
Estimation
1
17
-0.59
2.35
16
-0.65
1.37
Estimation
2
15
-0.67
2.66
16
0.35
1.32
Cognitive-Affective
and Behavior Questionnaire
Physical
Physical
1
18
-0.78
2.10
16
0.06
1.95
2
15
0.13
2.50
16
-0.25
2.98
Cognitive
1
18
0.94
2.21
17
-1.06
3.19
Cognitive
2
15
-1.73
2.02
17
-0.71
4.96 a
Appetite
1
18
-0.11
0.32
17
0.06
0.25
Appetite
2
15
-0.13
0.52
17
0.00
0.00
Note: a = p < .05; b p < .01
Tm= Assessment Interval; Psv = Perseverations;
Del = Delay; Intrus = Intrusions; SS = Standard
Score; Recog = Recognition; Orthograph = Orthographic;
False + = False positive


116
not due to a decrease in treatment group performance, but
rather, to improved performance between the first and
second assessments in subjects exhibiting a fall in Hg.
The majority of these subjects were in the control group.
Mean Sum CVLT in the treatment group was 45.8, 45.0 and
45.1 for assessments one through three respectively. With
a standard deviation varying between 11.6 and 13.4, these
scores were remarkably similar. In contrast, in the
control group, mean Sum CVLT was 43.2 at the first testing
but 47.3 and 47.1 at assessments two and three,
respectively. With a standard deviation varying between
10.4 and 10.7, mean performance at the first assessment was
nearly .4 of a standard deviation below expectations.
Hypothesizing that increased Hg may have reduced the
efficiency of dialysis resulting in increased uremia;
therefore, produced poorer cognitive performance secondary
to uremia, post hoc collection of BUN and creatinine levels
was attempted. However, few subjects were consistently
assessed on the same days as laboratory tests were
performed. Therefore, the small size created the strong
possibility, if not probability, of a Type II error.
Consistent with this, analysis failed to suggest a
relationship between BUN and Sum CVLT.
It might be argued that the control group, rather than
performing more poorly at the first assessment, did better
at the second assessment. That is, the control group
exhibited a practice effect while the treatment group,


54
After being instructed to try to recall as many words
as possible, subjects were read a list of words with some
words appearing more than once. Twenty minutes later
subjects were asked to estimate how many times each word
was read. Three versions were produced and are reproduced
in Appendix C. The three versions were administered in
consecutive order to all subjects. Prior to estimation of
frequency, subjects were tested on free recall.
Controlled Word Association Test
Controlled Word Association Test, a verbal fluency
task, provides a measure of retrieval unrelated to
encoding. It is also dependent on level of arousal,
motivation, verbal fluency and ability to initiate
behavior. Three commonly used sets of letter triplets are
available (Lezak, 1982).
Cognitive-Affective and Physical Behavior Questionnaire
A self-report questionnaire designed to assess
cognitive-affective and behavioral changes likely to be
caused by medical illness, was developed utilizing the
types of questions typically asked of medical patients to
evaluate these domains. The questionnaire generates a
Cognitive-Affective Scale and a Physical (i.e., physical
activity) Scale. The questionnaire was written to avoid
obvious response biases. The questionnaire and scoring
instructions are displayed in Appendix D.
In the absence of positive results this measure was
intended to determine whether there were positive changes


85
Exploratory Analyses
Observation of variability in Hg change scores in both
groups led to reclassification based on direction of Hg
change, rather than treatment status. A total of 23
subjects exhibited a rise in Hg between the first and second
assessments (M=1.22 g/dl, sd=1.26). Twelve subjects
exhibited a fall between the first and second assessments
(M=-1.05, sd=.96). The difference between groups with rEPO
treatment status as the classificatory variable produced a
mean difference in Hg change between groups of 1.03 g/dl.
In contrast, the alternate system of classification doubled
the Hg change score yielding 2.27 g/dl. Difference in Hg
change scores between the positive and negative change
groups was highly significant (F=29.75, p=.0001).
Control subjects suffering a fall in Hg (n=9) exhibited
a mean fall of -1.27 g/dl (sd=1.02). The three rEPO treated
subjects exhibiting a fall in Hg had a mean drop of -.38
g/dl (sd=.28). Among rEPO treated subjects with rising Hg
(n=14) the mean rise was 1.46 g/dl (sd=1.54). Subjects from
the control group exhibiting a rise in Hg (n=9) showed a
mean rise of 0.85 g/dl (sd=0.51).
For exploratory purposes, the main hypothesis was
retested with a GLM procedure utilizing sign of Hg change as
the classificatory variable. There was a significant
difference in Sum CVLT change scores between the positive
(n=23, M=-l.22, sd=7.18) and negative (n=12, M=7.25,
sd=7.63) Hg change scores (F=10.52, p=.0027). The analysis


95
Table 4-12. Correlation Between Treatment Group Sum CVLT I.
II and III
Pearson Correlations
CVLT II
CVLT III
CVLT I
CVLT II
r
1.00000
0.91732
0.77220
P
0.0000
0.0001
0.0003
N
17
17
17
CVLT III
r
0.91732
1.00000
0.81127
P
0.0001
0.0000
0.0001
N
17
17
17
CVLT I
r
0.77220
0.81127
1.00000
P
0.0003
0.0001
0.0000
N
17
17
17
Note: r = correlation; p = probability value
Table 4-13. Correlation Between Sum CVLT I. II and III for
Controls
Pearson Correlations
CVLT II
CVLT III
CVLT I
CVLT II r
1.00000
0.57038
0.64116
P
0.0000
0.0134
0.0100
N
18
18
15
CVLT III r
0.57038
1.00000
0.78866
P
0.0134
0.0000
0.0005
N
18
18
15
CVLT I r
0.64116
0.78866
1.00000
P
0.0100
0.0005
0.0000
N
15
15
15
Note: r = correlation; p = probability; N = subject number


24
some extent be attributed to the differential effects of the
passage of time on technology and on individuals.
Overview of Anemia
This section will provide a brief overview of the
significance and etiology of anemia, adaptive mechanisms and
the treatment of uremic anemia.
Prevalence and Etiology
The world-wide prevalence of anemia has been estimated
to be 40% in children, 35% in adult females and 20% in adult
males (Lozoff, 1989). The defining feature of anemia is a
reduction in blood Hg levels. Anemia is the final common
pathway for a large number of diseases, conditions resulting
in blood loss and deficiency states, all of which result in
decreased Hg in the blood (Bunn, 1980a). The most common
form of anemia, iron deficiency anemia has been reported to
have a prevalence in the U.S. of 20% in adult women of
childbearing years, 50% in pregnant women and 3% in adult
males (Lee, Wintrobe, & Bunn, 1980). Probably the second
most common type, the anemia of chronic disorders is the
mild to moderate anemia frequently associated with chronic
inflammatory syndromes and infectious and neoplastic
diseases (Wintrobe et al., 1981). Anemia is a very common
condition.
Anemia almost invariably accompanies chronic renal
failure (Desforges, 1975; Erslev, 1975). Uremic anemia
often significantly reduces quality of life, despite
adequate dialysis. Prior to the development of rEPO,


79
At the time of the first assessment, the Control group
included one subject with Low iron status and 14 Replete
subjects. At Assessment Two this was unchanged. At
Assessment Three there was still one subject with Low iron
status. In the rEPO group at Assessment One, there was one
Deficient subject, four Low iron subjects and 11 Replete
subjects. At Assessment Two, there were 6 Low iron and nine
Replete subjects. By the third assessment, the treatment
group was composed of one Deficient, three Low and 11
Replete subjects. For the subjects on which data was
available there was minimal change in iron status;
therefore, the need to control for iron deficiency was
greatly reduced.
Replete subjects were further divided into Replete and
High (over 1,000). As illustrated in Table 4-5 below, the
percentage of subjects exhibiting low iron status was
consistently greater in the rEPO treatment group as was the
percentage of subjects with high iron status. Also apparent
from the Table 4-4, iron status was relatively stable across
assessments, especially if the top two and bottom two groups
were collapsed into Low and Replete categories.
Baseline Neuropsychological Functioning
Table 4-6 displays selected raw and standard scores for
CVLT II at the baseline assessment. There are no
statistically significant differences for any of the scores
listed in Table 4-6 or in the more comprehensive list of
scores in Appendix G. However, despite the lack of


113
sympathetic arousal secondary to decreased need for high
cardiac output. In addition, at high levels of Hg, the
efficiency of dialysis may be compromised resulting in
increased uremia.
Iron deficiency is a frequent consequence of rEPO
therapy in the absence of adequate iron supplementation or
assimilation. Iron deficiency has been shown to have
neurobehavioral consequences unrelated to heme synthesis.
Quite conceivably, stepped-up erythropoiesis could bind
enough nonheme iron into Hg to lower iron levels
sufficiently to deleteriously affect brain function.
Hypothetically, this could result in reduced cognitive
performance, and several weeks later, a rise in Hg.
However, in treatment subjects, vigorous iron
supplementation minimized changes in iron status;
therefore, very few subjects exhibited changes in iron
stores of potential clinical significance. Consistent with
this observation, a GLM univariate analysis unequivocally
failed to demonstrate a main effect for iron status
(F=0.01, p=0.9384).
The notion that rEPO may affect brain function through
mechanisms unrelated to erythropoiesis is contrary to
general assumptions. However, cancerous neural tissue is
capable of producing EPO (Hennessy, Stern, & Herrick,
1967) This capability suggests the possibility that EPO
may serve a function in the nervous system. If rEPO
negatively affected brain function through nonheme


53
The levels of processing paradigm controls the nature
of effortful processing through manipulation of the task.
Subjects were visually presented a written word and asked
to respond to a guestion that required either orthographic,
phonemic or semantic analysis of the visual stimuli.
Following an interval of approximately 20 minutes, accuracy
in identification of the target words was measured with an
oral multiple choice task. Subjects able to benefit from
semantic processing may be expected to exhibit a higher
rate of correct recognition for semantically processed
stimuli. Subjects exhibiting memory deficits due to
reduced cognitive processing, because extent of processing
is controlled by the task, would be expected to perform
like normal subjects on this task. Normal subjects show a
recognition advantage for the semantically processed words
(Craik & Tulving, 1975). Subjects suffering from
neurologically based memory disorders affecting encoding in
LTM fail to benefit from semantic processing. Thus,
improvement in ability to benefit from semantic processing
would suggest physiological improvement.
Frequency Estimation Task
Estimation of frequency of occurence provides a
clinical measure of automatic learning believed to be
relatively invulnerable to factors such as fatigue, but
presumably vulnerable to physiological factors (e.g.,
hypoxia) directly interfering with the functioning of
memory structures (Hasher & Zacks, 1984).


51
receiving rEPO earlier than anticipated and control
subjects became difficult to locate. Therefore, subjects
were accepted into the study beyond the eighth week,
reasoning that they could serve as controls in the event of
minimal additional change in Hg. Mean lag between
initiation of rEPO and the first assessment was 5.9 weeks
(sd=4.57).
The 18 control subjects did not receive rEPO.
Thirteen subjects, 72.2%, were on some form of peritoneal
dialysis (i.e., Peritoneal Dialysis, Continuous Ambulatory
Peritoneal Dialysis or Continuous Cycler Peritoneal
Dialysis) and five subjects, 27.8%, were on maintenance
hemodialysis. Given the small population from which to
draw controls, the ideal of matching controls to treated
subjects on age, education and ideally, race and sex, was
not considered feasible. Despite this, as will be
described later, the two groups were nearly identically in
domains other than treatment modality. However, mean Hg at
the first or baseline assessment was 8.46 g/dl (sd=1.42) in
the treatment group and 9.44 g/dl (sd=1.92) in the control
group.
All subjects were paid $15 at the completion of the
third assessment. Subjects were tested in their home,
their dialysis clinic, or in a STH Clinical Psychology
testing room.


127
Lorge (1944) and Carroll, Davies and Richman (1971) as well
as the rank order of each word as an exemplar of its
category (Battig and Montague, 1969). Table A-2 lists Form
III words and their frequency values.
Methods
Eighteen normal, native English speaking volunteers
were administered Forms I and III of the CVLT. Seventeen
subjects were between 18 and 23 years of age. One subject
was 48 years old. Mean age was 21.9 with a standard
deviation of 6.7 years. There were 11 female and 7 male
subjects. Mean years of college education was 14.5
(SD=1.2). Sixteen of the 18 subjects had at least some
college education.
The majority of testing was performed by an
undergraduate senior psychology major in his last semester
of study, who had received extensive training prior to data
collection. All subjects were administered Forms I and III
of the CVLT. Every other subject began with Form I. Mean
interval between testing sessions was 5 days. Minimum time
between sessions was 48 hours.
All assessments were performed individually. During
the 20 minute delay between immediate and delayed recall
subjects either conversed with the experimenter or
performed activities in which they had been engaged prior
to the assessment (e.g., studying). Administration of
other measures during the delay was rejected to reduce the
difficulties associated with obtaining volunteer subjects.


16
between two sequences, speeded visual search and simple
visual motor coordination, has been reported to be reduced,
at times severely, in renal disease (Heilman, Moyer,
Melendez, Schwartz, & Miller, 1975; Ratner, Adams, Levin, &
Rourke, 1983; Teschan et al., 1974). Ginn (1975) reported a
correlation of .82 between Trails and severity of renal
failure as measured by serum creatinine for subjects not on
dialysis. Similar results were reported by Teschan and
colleagues (1974). Reduced performance on tasks demanding
sustained vigilance, such as continuous performance tasks,
has been consistently reported (Ginn, 1975; Osberg, Meares,
McKee, & Burnett, 1982). Ginn and colleagues (1978)
demonstrated worsening of performance on a vigilance task in
8 out of 10 subjects following a reduction in hemodialysis.
Contributions to executive system dysfunction may
include deficits secondary to reduced level of activation,
diminished attentional capacity, susceptibility of the
frontal-subcortical neural substrate to metabolic
abnormalities and dysfunction stemming from a generalized
difficulty with tasks demanding fluid as opposed to
crystalized intelligence.
Verbal Memory
Reduced vigilance, underarousal, increased
distractibility, increased response latencies and reduced
cognitive processing may contribute to global memory
deficits. Work in a pediatric renal population has revealed
a correlation between learning on a multitrial supra-span


143
Table D-4. Physical Scale Scoring Instructions
Start with 18 points and subtract one point for each
endorsed.
2. Walking quickly tires me out.
3. I sit during much of the day.
4. I take one or more naps during the day.
6. I easily get out of breath.
10. I do work around the house only for short periods.
11. I stay away from home only for brief periods of time.
13. I am not doing any daily work around the house.
15. I feel tired most of the time.
19. I spend much of the day lying down in order to rest.
22. I feel tired all the time.
23. Everything is an effort.
27. Simple chores quickly tire me out.
28. My muscles quickly get tired.
Subtract one point for any of the following not endorsed:
1. I don't mind climbing stairs.
12. My sex drive is stronger than it was one year ago.
18. I work a job outside the home.
25. I do heavy work around the house.
26. I do alot of cooking.


170
Mings, E. L. (1987) Visual Motor Integration Deficits in
Pediatric Renal Disease. Doctoral dissertation,
University of Florida.
Miskin, M. (1978). Memory in monkeys severely impaired by
combined but not by separate removal of amygdala and
hippocampus. Nature. 273. 297-298.
Moray, N. (1959). Attention in dichotic listening: Affective
cues and the influence of instructions. Quarterly Journal
of Experimental Psychology. 11, 59-60.
Morris, C.D., Bransford, J.D., & Franks, J.J. (1977). Levels
of processing vs. transfer appropriate processing.
Journal of Verbal Learning and Verbal Behavior. 16 519-
533.
Muramoto, 0., Kuru, Y., Sugishita, M., & Toyokura, Y.
(1979). Pure memory loss with hippocampal lesions. A
pneumoencephalographic study. Archives of Neurology. 36.
54-56.
Newman, R.P., Weingartner, H., Smallberg, S.A., & Caine,
D.B. (1984). Effortful and automatic memory: Effects of
dopamine. Neurology. 34. 805-807.
Nissen, M.J. Knopman, D.S., & Schacter, D.L. (1987).
Neurochemical dissociation of memory systems. Neurology.
789-794.
Nissenson, A.R. (1989). Recombinant human erythropoietin:
Impact on brain and cognitive function, exercise
tolerance, sexual potency, and quality of life. Seminars
in Nephrology. 9, 25-31.
Nissenson, A.R., Levin, M.L., Klawans, H.L., & Nausieda,
P.L. (1977). Neurological sequelae of end-stage renal
disease. Journal of Chronic Disease. 30, 705-733.
Nissenson, A.R., Marsh, J.T., & Brown, W.S. (1988).
Recombinant erythropoietin improves brain function in
chronic hemodialysis patients. Kidney International. 33.
232. Abstract.
Nissenson, A.R., Marsh, J.T., Brown, W.S., Schweitzer, S., &
Wolcott (1989). Brain function improves in chronic
hemodialysis patients after recombinant erythropoietin.
Kidney International. 35. 257. Abstract.
Osberg, J. W., Meares, G.J., McKee, D.C., & Burnett, G.B.
(1982). Intellectual functioning in renal failure and
chronic dialysis. Journal of Chronic Diseases. 35, 445-
457.


29
(Deinard, List, Lindgren, Hunt, & Chang, 1986; Groner,
Holtzman, Charney, & Mellits, 1986). Until recently, with
the exception of blood transfusions, uremic anemia has only
receded following successful renal transplantation.
Although transplantation often enhances many
neuropsychological functions, there is a huge confound in
the simultaneous improvement in both uremia and anemia.
Therefore, the role of anemia in neuropsychological
dysfunction has not been adequately tested in any
population.
Effects of Anemia
Anemic patients may complain of a variety of symptoms
including fatigue, dizziness, dyspnea, poor concentration,
irritability, weakness, faintness, headache, impotence and
tinnitus (Bunn, 1980b; Rapaport, 1987). Some of the
complaints reported in conjunction with anemia may be due to
the cause of the anemia rather than the effects of low Hg
levels (Davies-Jones, Preston, & Timperley, 1980). The
relationship between severity of symptoms and Hg
concentration is confounded by the effectiveness of a
variety of adaptive mechanisms. Symptomatology is dependent
on factors including the degree and rate of reduction in
oxygen carrying capacity of the blood, degree and rate of
change in total blood volume, and the compensatory capacity
of the cardiovascular and pulmonary systems (Wintrobe et
al., 1981). Following adaptation symptoms may subside. If
anemia develops slowly, effective adaptation may avert the


39
g/dl, severely anemic, to 11.56 +2.41, mildly anemic. At
sea level anemia is suspected in men when Hg is <14.0 g/dl
and <12.0 g/dl in women (Rapaport, 1987).
Table
1-1.
Pilot Studv Hemoglobin and
Sum CVLT
Sum CVLT
SS Change
Obs
Grouo
Hg
Pre
Hg
Post
Hg
Chancre
Sum CVLT
Change
1
Con
7.0
6.7
-0.3
-6
-0.93
2
Con
8.1
9.5
1.4
-5
-0.55
3
rEPO
6.0
8.4
2.4
1
0.11
4
rEPO
5.1
11.2
6.1
3
0.47
5
rEPO
7.1
13.2
6.1
8
1.08
6
rEPO
5.7
12.0
6.3
0
0.00
7
rEPO
6.6
13.0
6.4
16
1.90
Note:
Obs =
Observation; SS =
Standard Score; Con
= control
group
; rEPO
= treatment
group;
Change = change score
The Profile of Mood States, a self-report measure,
failed to reveal any improvement in fatigue, tension,
depression, vigor, or confusion. However, individual
patients reported dramatic improvements in circumscribed
domains. Individual subjects reported increased capacity
for physical activity, a reduction in post dialysis fatigue
and decreased problems with impotency. The dramatic
reduction in physical symptoms in association with absence
of subjective effects is consistent with reports of reduced
subjective symptoms when the anemia is chronic, even if it


164
Comty, C.M., Leonard, A., & Shapiro, F.C. (1974).
Psychosocial problems in dialized diabetic patients.
Kidney International. 1, Supplement 144-151.
Craik, F.I.M., & Lockhart, R.S. (1972). Levels of
processing: a framework for memory research. Journal of
Verbal Learning and Verbal Behavior. 11. 671-684.
Craik, F.I.M., & Tulving, E. (1975). Depth of processing and
the retention of words in episodic memory. Journal of
Experimental Psychology: General. 104. 268-294.
Crittenden, M.R., Holliday, M.A., Piel, C.F., & Potter, D.E.
(1985). Intellectual development of children with renal
insufficiency and end stage disease. International
Journal of Pediatric Nephrology. 6, 275-280.
Darwin, C.J., Turvey, M.T., & Crowder, R.G. (1972). An
auditory analogue of the Sperling partial report
procedure: Evidence for brief auditory storage, Cognitive
Psychology. 3, 255-267.
Davies-Jones, G.A.B., Preston, F.E., & Timperley, W.R.
(1980). Neurological Complications in Clinical
Hematology. Boston, Massachusetts: Blackwell Scientific
Publications.
Dawson, A.A., Ogston, D., & Fullerton, H.W. (1969).
Evaluation of diagnostic significance of certain symptoms
and physical signs in anemic patients. British Medical
Journal, 3, 436-439.
Deinard A.S., List, A., Lindgren, B., Hunt, J.V., & Chang,
P. (1986). Cognitive deficits in iron-deficient and iron-
deficient anemic children. Journal of Pediatrics. 108.
681-689.
Delano, B.G. (1989). Improvements in quality of life
following treatment with r-HuEPO in anemic hemodialysis
patients. American Journal of Kidney Diseases. 14, 14-18.
Delis, D.C., Freeland, J., Kramer, J.H., & Kaplan, E.
(1988). Integrating clinical assessment with cognitive
neuroscience: Construct validation of the California
Verbal Learning Test. Journal of Consulting and Clinical
Psychology. 56. 123-130.
Delis, D.C., Kramer, J.H., Kaplan, E., & Ober, B.A. (1987).
California Verbal Learning Test Adult Version Research
Edition Manual. New York, New York: Psychological
Corporation.


153
Table G-3. Means and Standard Deviations at Third
Assessment by Group
Control Group
EPO Treated
Variable
N
Mean
Sd
N
Mean
Sd
Hemoglobin
15
9.62
1.92
17
9.54
1.48
Hematocrit
15
28.12
5.55
17
27.55
4.06
Ferritin
3
484.30
266.67
14
560.34
814.39
Assessment Interval 15
67.80
34.56
17
90.41
24.03
California Verbal
Learning Test
Trial 1
15
5.87
1.64
17
6.00
1.97
Standard Score
15
-1.13
0.74
17
-1.24
1.03
Trial 2
15
9.00
2.80
17
8.94
3.17
Trial 3
15
10.33
2.23
17
9.82
2.24
Trial 4
15
10.93
2.66
17
9.82
3.13
Trial 5
15
10.93
2.52
17
11.12
3.02
Standard Score
15
-1.47
1.85
17
-1.53
1.59
Sum CVLT
15
47.07
10.40
17
45.12
11.61
Standard Score
15
36.80
13.21
17
33.18
13.65
Clusters
15
13.80
8.27
17
13.76
11.95
Perseverations
15
3.33
2.13
17
5.00
2.94
Intrusions
15
1.47
2.10
17
2.29
3.80
List B
15
6.40
2.16
17
4.88
2.50
Standard Score
15
-0.53
1.13
17
-1.35
1.37
Short Delay
15
9.07
2.87
17
9.47
3.39
Standard Score
15
-1.13
1.19
17
-1.12
1.32
Clusters
15
3.87
2.17
17
4.00
2.76
Perseverations
15
0.20
0.56
17
0.76
1.03
Intrusions
15
0.80
0.86
17
0.47
1.01
Short Delay Cued
15
9.87
2.83
17
10.76
2.39
Standard Score
15
-1.40
1.55
17
-1.00
1.12
Perseverations
15
0.07
0.26
17
0.00
0.00
Intrusions
15
0.73
0.88
17
1.29
1.65
Long Delay
15
9.27
2.87
17
9.65
3.82
Standard Score
15
-1.53
1.41
17
-1.24
1.48
Clusters
15
4.00
2.59
17
4.35
3.16
Perseverations
15
0.27
0.59
17
0.24
0.56
Intrusions
15
0.93
1.03
17
0.88
1.05
Long Delay Cued
15
10.13
2.95
17
10.82
2.53
Standard Score
15
-1.47
1.60
17
-1.24
1.35
Perseverations
15
0.00
0.00
17
0.00
0.00
Intrusions
15
1.20
1.15
17
1.41
1.54
Recognition Hits
15
14.07
1.62
17
13.59
3.18
False Positives
from
Interference List
Shared Category
15
0.87
0.99
17
0.82
0.88


56
Iron status was used to classify subjects into low and
replete categories utilizing lab report guidelines. The
normal range for ferritin is frequently cited as around 10
ng/ml to 107 ng/ml. Ferritin levels of under 10 ng/ml were
considered indicative of iron deficiency. Ferritin values
under 30 ng/ml were classified as low. Normal and elevated
iron values were grouped as normal for the purposes of
these analyses.
Treatment with rEPO results in a heavy demand on iron
stores and induces iron deficiency in the absence of iron
supplementation (Van Wyck, 1989). For this reason, as part
of routine medical procedures, serum ferritin levels were
drawn on most rEPO treated subjects on a monthly basis;
however, the experimenter was blind to iron status until
the completion of neuropsychological data collection.
Schedule of Testing
Subjects were tested on three occasions. Assessment
Two was intended to be administered approximately 30 days
following the first assessment. Assessment Three was
scheduled for 60 days following the second assessment.
Patient illness, noncompliance and difficulty working
around transportation schedules significantly altered the
assessment timetable in some instances. Because time of
testing relative to hemodialysis has been reported to
affect results, this was kept consistent for each subject.
For most subjects, assessment occurred immediately
following hemodialysis.


APPENDIX G
GROUP MEANS AND STANDARD DEVIATIONS
Table G-l. Means and Standard Deviations at Assessment One
bv Group
Control Group rEPO Treated Group
Variable
N
Mean
Sd
N
Mean
Sd
Age
18
48.22
14.19
17
43.24
13.45
Education in years
18
11.83
2.09
17
11.41
3.02
Days since EPO begun
-
-
16
41.37
32.08
Years on Dialysis
18
2.84
3.76
17
6.64
7.69
Ferritin
6
523.62
626.13
16
545.93
734.50
Hemoglobin
18
9.44
1.92
17
8.46
1.42
Hematocrit
18
28.19
6.12
17
25.58
3.77
California Verbal
Learning Test
Trial 1
18
5.72
1.74
17
5.88
2.47
Standard Score
18
-1.33
0.91
17
-1.35
1.37
Trial 2
18
7.89
2.00
17
8.41
2.37
Trial 3
18
9.17
2.73
17
9.65
3.52
Trial 4
18
9.89
2.91
17
10.71
2.93
Trial 5
18
10.56
3.07
17
11.24
3.25
Standard Score
18
-1.39
1.33
17
-1.59
2.12
Sum CVLT
18
43.22
10.70
17
45.82
13.03
Standard Score
18
33.06
10.31
17
33.82
16.00
Clusters
18
11.28
7.43
17
13.24
11.51
Perseverations
18
3.06
2.60
17
3.65
3.10
Intrusions
18
0.83
2.04
17
0.88
1.83
List B
18
5.94
1.66
17
5.76
1.92
Standard Score
18
-0.78
0.94
17
-0.88
1.05
Short Delay
18
9.06
2.71
17
9.65
3.26
Standard Score
18
-1.11
0.83
17
-1.06
1.43
Clusters
18
3.44
2.01
17
3.47
3.02
Perseverations
18
0.22
0.43
17
0.76
0.97*
Intrusions
18
0.17
0.51
17
0.24
0.56
Short Delay Cued
18
10.94
2.69
17
10.94
2.90
Standard Score
18
-0.83
1.04
17
-0.88
1.36
Perseverations
18
0.00
0.00
17
0.29
0.77
Intrusions
18
0.56
1.34
17
0.65
1.22
149


80
statistically significant differences, examination of raw
scores suggests a tendency for somewhat better performance
in the treatment group at the first assessment.
Table 4-5. Percentage of Subjects in each Category of Iron
Status
Iron
Assess
Con
1
rEPO
Assess
Con
2
rEPO
Assess
Con
3
rEPO
Hi
6.7
23.5
0
20.0
0
18.8
Normal
86.7
41.2
93.7
40.0
92.3
50.0
Lo
6.7
29.4
6.7
40.0
7.7
25.0
Defic
0
5.9
0
0
0
6.3
Note: Hi = Elevated Ferritin; Lo = Low; Defic = Deficient
The mean baseline age corrected control group Sum CVLT
score placed the group in the 4th percentile, the borderline
range. The mean baseline Sum CVLT score in the treatment
group placed it in the 9th percentile, the Low Average
range. However, the differences, as previously stated, did
not reach statistical significance.
As displayed in Appendix G, Table G-l, on the
Controlled Word Association test, using the letters CFP, the
control group (n=17) produced a mean of 11.53 words
(sd=3.84). The treatment group (n=17) produced a mean of
10.53 (sd=4.45) words. The difference was not statistically
significant.
The levels of processing task revealed that both groups
benefited from semantic processing compared to either


Id
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
101
102
103
104
105
106
107
108
77
4-3. Anemia Status Across Assessments
Assess 1 Assess 2 Assess 3
Severe
Moderate
Moderate
Severe
Very Severe
Moderate
Moderate
Severe
Moderate
Moderate
Severe
Severe
Moderate
Moderate
Moderate
Moderate
Severe
Severe
Moderate
Very Severe
Moderate
Normal
Moderate
Normal
Moderate
Moderate
+
Moderate
Moderate
Moderate
+
Severe
+
Normal
+
Severe
-
Moderate
+
Moderate
Moderate
Moderate
+
Severe
Moderate
Moderate
Moderate
Moderate
Moderate
+
Moderate
+
Moderate
Severe
+
Moderate
Normal
Severe
-
Moderate
+
Moderate
3
Moderate
Moderate
Moderate
Severe
Severe
Moderate
Moderate +
Severe
Moderate
Moderate
Moderate
Severe
Moderate
Moderate
Normal +
Moderate
Moderate
Moderate
Moderate
Severe

Moderate

Normal +
Moderate


EFFECT OF RECOMBINANT ERYTHROPOIETIN MEDIATED CHANGES IN
ANEMIA ON VERBAL MEMORY IN ADULT DIALYSIS PATIENTS
BY
WAYNE L. KLEIN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE
UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE
REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1991

ACKNOWLEDGEMENTS
An elderly subject stood out for interest in his
performance. A man in his early twenties appeared
remarkably uninterested in his standing. A third subject
participated in the hope of helping others, despite
increasing debilitation. All three died prior to
completion of the study. Their names, and the names of the
32 other subjects are not listed out of respect for their
privacy. However, without their willingness to endure
testing despite the discomforts, and in some cases the
agonies, of their illness, this study would not have been
possible.
The following, undergraduates at the time, assisted
with test construction, test validation, data collection
and/or data entry: Bill Dorkowsky, Mike Reiter, Sheri
Scott, Katherine Cobb, Edward Suarez and Wylene Bhanji.
Without their help my wife, Liz, and child, Sarah, would
have suffered even worse neglect. Don Mars, M.D.
graciously persuaded his patients to participate as
controls. The American Psychological Association provided
a Dissertation Research Award which partially funded this
project.
My committee, beyond help regarding this project,
consists of those who were most important to my graduate

iii
experience. Eileen Fennell, my chair, enabled me to meld
my interests with the do-able and the acceptable. Without
her the going would have been much less pleasant. It was
on Russ Bauer's clinic day that I decided to specialize in
clinical neuropsychology. Hugh Davis is forever and
agreeably etched in my subconscious. Bob Fennell provided
research opportunities with import to me well beyond the
findings. Tom Fast became a close friend who taught many
things.

TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS ii
ABSTRACT vii
CHAPTERS
1 INTRODUCTION: FOCUS AND RATIONALE 1
2 LITERATURE REVIEW 5
Introduction 5
Overview of Verbal Memory 5
Memory Stages and Levels of Processing ... 6
Automatic and Effortful Memory 12
Neuropsychological Dysfunction in Adult Renal
Patients 13
General Intelligence: Verbal and Nonverbal 13
Attention, Level of Activation and Executive
Function 14
Verbal Memory 16
Nonverbal Memory 20
Conclusions Regarding Cognitive Deficits in
Renal Disease 21
Overview of Anemia 24
Prevalence and Etiology 24
Function of Hemoglobin and Adaptation to
Anemia 26
Erythropoietin Therapy and Iron Deficiency 27
Neuropsychology of Anemia 27
Effects of Anemia 29
Possible Mechanisms 32
Effect of Erythropoietin on Neuropsychological
Function 35
Effect of Erythropoietin on Quality of Life 36
Previous Neuropsychological Studies 37
Results of Pilot Study 38
Negative Effects 44
Summary and Conclusions 46
3METHODS 50
Subjects 50
Measures 52
iv

V
California Verbal Learning Test 52
Levels of Processing Task 52
Frequency Estimation Task 53
Controlled Word Association Test 54
Cognitive-Affective and Physical Behavior
Questionnaire 54
Hemoglobin 55
Ferritin 55
Schedule of Testing 56
Statistical Methodology 58
Hypotheses 58
Main Hypothesis 58
Secondary Hypotheses 59
4 RESULTS 61
Overview of Analyses 61
Statistical Assumptions 64
Descriptive Statistics 69
Age 69
Education 69
Sex 69
Treatment Modality 69
Attrition 70
Time Intervals Between Assessments 71
Hemoglobin 72
Ferritin 75
Baseline Neuropsychological Functioning ... 79
Test of Main Hypothesis 82
Post Hoc Analyses 83
Learning Curve 83
Exploratory Analyses 85
Sum of CVLT Trials Four and Five 88
Short Term Memory 90
Verbal Fluency 90
Secondary Hypotheses 91
Levels of Processing 91
Estimation of Frequency of Occurrence .... 91
Quality of Life 91
Test Reliability 93
CVLT Alternate Form Reliability 93
Levels of Processing Reliability 98
Frequency of Occurrence Reliability 98
Physical Behavior Self-Report Test-Retest
Reliability 98
Cognitive-Affective Behavior Self-Report
Test-Retest Reliability 100
Post Hoc Analysis of Blood Urea Nitrogen . 100
Summary 101
5 DISCUSSION 104
Overview 104
Conclusion 121

vi
APPENDICES
A DEVELOPMENT AND VALIDATION OF FORM III OF
THE CALIFORNIA VERBAL LEARNING TEST 124
B LEVELS OF PROCESSING VERSIONS II AND III . 131
C FREQUENCY ESTIMATION TASK 139
D COGNITIVE-AFFECTIVE AND PHYSICAL BEHAVIOR
QUESTIONNAIRE 141
E COMPARISON OF PARAMETRIC AND NONPARAMETRIC
TESTS OF BASELINE MEASURES BETWEEN GROUPS . 144
F NORMALITY AND HOMOGENEITY OF VARIANCE . .146
G GROUP MEANS AND STANDARD DEVIATIONS 149
H CHANGE SCORES 155
I RAW DATA BY SUBJECT 158
BIBLIOGRAPHY 162
BIOGRAPHICAL SKETCH 175

Abstract of Dissertation Presented to the Graduate School of
the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
EFFECT OF RECOMBINANT ERYTHROPOIETIN MEDIATED CHANGES IN
ANEMIA ON VERBAL MEMORY IN ADULT DIALYSIS PATIENTS
By
Wayne L. Klein
August 1991
Chairperson: Eileen B. Fennell
Major Department: Clinical and Health Psychology
Cognitive dysfunction has been documented in many
chronic anemic conditions; however, the putative effects of
anemia have never been disassociated from the cause of the
anemia. Neuropsychological dysfunction of poorly understood
origin and severe chronic anemia are nearly universal
consequences of chronic renal failure. A small pilot study
demonstrated a positive relationship between a recombinant
erythropoietin (rEPO) mediated mean rise of 5.46 g/dl of
hemoglobin (Hg) and the Sum of trials 1-5 on the California
Verbal Learning Test (Sum CVLT). Sum CVLT performance was
hypothesized to be enhanced by increased effortful
processing secondary to reduced fatigue.
A rEPO treatment group (n=17) and a control group
(n=18) underwent three assessments with three alternate
forms of the CVLT, one of which was developed for this
vii

viii
study. Also administered were a survey of cognitive-
affective and physical behavior, a verbal fluency task and
three alternate forms of two measures likely to be sensitive
to hypoxia but not fatigue, the levels of processing and
frequency estimation tasks. All subjects were on
maintenance dialysis. Mean treatment group Hg at the first
assessment was 8.46 g/dl (sd=1.42) and 9.44 g/dl (sd=1.92)
in the controls. The rise of 1.34 g/dl Hg between groups in
the 54 day mean interval between the first and second
assessments was statistically significant (F=7.42, p=.0102),
but of doubtful physiological significance. In the 90 days
between the second and third assessments Hg was unchanged.
The relationship between rise in Hg and Sum CVLT failed to
reach statistical significance (F=3.25, p=.0805). Post hoc
reclassification of subjects based on direction of Hg change
unexpectedly revealed an apparent inverse relationship
between Hg and Sum CVLT. Subjects, primarily controls,
exhibiting a drop in Hg between the first and second
assessments performed below expectations at the first
assessment (F=7.82, p=.0027). Whether this was spurious or
due to the operation of an unknown factor remains
speculative. Hemoglobin related changes in self-reported
cognitive-affective and physical behavior were not observed.
Current understanding of the relationship between anemia and
neuropsychological dysfunction appears insufficient to
inform decisions regarding rEPO treatment.

CHAPTER 1
INTRODUCTION: FOCUS AND RATIONALE
The fourth leading health problem in the developed
world, renal disease has been estimated to affect
approximately 8 million people in the U.S. (Williams, 1985).
Uremia, the life threatening condition resulting from renal
failure, has been known since ancient times. A variety of
symptoms have been associated with uremia, the majority of
which have been reported to be secondary to nervous system
dysfunction (Ginn et al., 1975). Memory dysfunction has
been among the neuropsychological deficits commonly noted.
Despite considerable effort, attempts to identify the
factors responsible for these deficits have achieved little
success (Powell et al., 1986; Schreiner, 1975; Trompeter,
Polinsky, Andreoli, & Fennell, 1986).
Chronic anemia has been almost universally observed in
chronic renal failure. The primary cause has been generally
accepted to be insufficient production of erythropoietin
(EPO), a erythropoiesis stimulating factor produced by the
kidneys. Recent studies have documented the dramatic
effects of recombinant erythropoietin (rEPO) on many
physical symptoms until recently attributed to uremia, but
now demonstrated to have been at least partially caused by
anemia (Nissenson, 1989). One rationale for the present
1

2
study was to test the possibility that a portion of the
neuropsychological dysfunction that has been attributed to
uremia is actually secondary to anemia.
A further rationale for the present study was to
address the broader question of whether cognitive deficits
are directly associated with chronic anemia. Severe acute
anemia produces serious symptoms of cerebral dysfunction,
but symptoms subside over time as adaptation occurs.
Although neuropsychological deficits have been well
documented in conditions of chronic anemia, the question of
whether the neuropsychological symptoms are caused by the
anemia has never been adequately addressed. The
neuropsychological study of anemia has been hampered by
difficulties in parceling out the effects of rise in
hemoglobin (Hg) from the effects of the agent causing the
rise. Chronic anemia is the final common result of numerous
causative factors including deficiencies of iron, folic acid
and cobalamin as well as conditions such as sickle cell
disease, all of which have been demonstrated to directly
affect the brain. In contrast, rEPO has been generally
considered to have no effect on the nervous system. This
putative disassociation between neurological and
hematological effects suggested that rEPO could potentially
be a valuable tool in the study of the neuropsychological
effects of chronic anemia.
Pilot work sampled a spectrum of cognitive functioning
before and during rEPO treatment. That work, although

3
suffering from small sample size, suggested that rise in Hg
was associated with improved long term memory (LTM) as
measured by performance on a multitrial verbal list learning
task (Klein et al., 1989). The present study sought to
confirm and extend these very preliminary findings.
An additional purpose of the present study was to generate
exploratory data addressing the nature of the mechanisms
underlying the putative improvement in LTM. In the pilot
study, despite sampling a variety of cognitive functions,
the only task that showed significant improvement in
performance was the task most sensitive to success in
actively organizing material into a meaningful structure.
Thus, it was hypothesized that the apparent improvement was
mediated by increased cognitive processing, possibly
secondary to reduced fatigue.
In the present study, paradigms from human memory
research were used in an attempt to tease out the level at
which putative change occurred. Improved automatic learning
and increased ability to benefit from semantic processing,
two tasks relatively unaffected by functional factors, would
have implicated physiological mechanisms affecting the
neural substrate of memory. Lack of improvement on these
tasks, in conjunction with improved effortful verbal LTM
performance, would have lent support to the hypothesis that
a functional mechanism such as increased cognitive
processing, possibly secondary to reduced fatigue, resulted
in the apparent improvement in LTM in the pilot study.

4
Chronic anemia has been reported to have far reaching
effects on the organism. In loose association with the
severity of the anemia, symptoms may include fatigue,
hypoxia, hypocapnea, insomnia, enhanced cardiac output and
increased sympathetic activation, reduced appetite, altered
blood chemistry, modified lifestyle including the inability
to hold gainful employment and, probably, decreased self
esteem. Therefore, determination of the relative
contributions of possible underlying mechanisms was
considered impossible in a single study of this scope.
However, it was hoped that this study would extend our
understanding of the neuropsychology of chronic anemia and
the clinical significance of chronic anemia with regard to
the etiology of the cognitive deficits observed in renal
disease.

CHAPTER 2
LITERATURE REVIEW
Introduction
Investigation of the neuropsychology of uremic anemia
is necessarily multidisciplinary. The logic underlying the
rationale and design of this study is based on information
derived from five somewhat distinct areas of investigation.
The areas are 1) human memory, 2) clinical studies of
neuropsychological function in adults suffering from renal
disease, especially with regard to verbal memory, 3) the
physiology of anemia, 4) the neuropsychology of anemia and
5) the effects of rEPO mediated improvement in anemia in
adult renal patients on neuropsychological function. Each
of these will be briefly reviewed in this chapter.
Overview of Verbal Memory
This section briefly overviews current thinking with
regard to memory as it relates to the clinical assessment
and interpretation of memory function. The application of
memory paradigms to the study of the neuropsychology of
anemia is discussed. Mounting neurochemical, functional and
anatomical evidence suggests that memory may be classified
into procedural memory (i.e., motor memory or skill
learning) and declarative memory for facts (Nissen, Knopman,
& Schacter, 1987). Memory may, to some extent, be divided
5

6
into automatic and the more commonly studied effortful,
deliberate or consciously encoded memory (Hasher & Zacks,
1984; Newman, Weingartner, Smallberg, & Caine, 1984;
Sanders, Gonzalez, Murphy, Liddle, & Vitina, 1987).
Automatic declarative memory may be further differentiated
based on the modality of encoding and retrieval, with the
most studied modalities being verbal and visual. Verbal
memory may also be partitioned into semantic and episodic
memory, the former being dissociated from, and the later
bound within, the context of learning (Newman, Weingartner,
Smallberg, & Caine, 1984).
Memory Stages and Levels of Processing
Most paradigms appear to demonstrate the division of
effortful declarative verbal memory into relatively clear-
cut temporal categories or stages. The cognitive,
neuropsychological and neuroscience traditions have posited
long and short term memory stages. The definition of memory
stages is dependent on the level of analysis (Squire, 1987).
Cognitive science has classified auditory memory of a
duration exceeding about 15 seconds as long term memory
(LTM) while neuroscience considers this squarely within the
domain of short term memory (STM). These differences
reflect the criteria by which stages have been determined
within the two traditions.
Within the cognitive tradition a succession of models
(e.g., capacity, multistore and levels of processing) has
stimulated research. In attempting to map memory function

7
onto biological structures and processes, the neuroscience
perspective has focused both on the nature of the memory
trace and the physical location of memory in the brain.
This brief overview will focus primarily on a behavioral
(i.e., cognitive and neuropsychological) level of analysis
as it relates to neuropsychological assessment.
As conceptualized in the multistore model, memory is a
complex series of distinct stages through which information
is successively processed (Atkinson & Shiffrin, 1968, 1971).
These stages differ along several dimensions including
capacity, duration and the nature of the attendant
psychological processes. The three generally recognized
stages are ultrashort or modality-specific sensory memory
(e.g., echoic and iconic memory), immediate or STM and LTM.
Sperling (1960) demonstrated that iconic (i.e., visual
sensory) memory decays after about one second. Echoic
memory may take several seconds to decay (Darwin, Turvey, &
Crowder, 1972). Sensory memory capacity appears to be
around nine items (Sperling, 1960). The duration of STM
when rehearsal (i.e., replenishment) is prevented has been
shown to be about 15 seconds (Peterson & Peterson, 1959).
Thus, the repetition of a list of orally presented digits or
words utilizes both echoic memory and STM. However, in the
case of oral digit span, as traditionally administered, the
limiting factor is likely to be attentional.
Oral digit span is the most commonly used measure of
immediate memory span; however, it is frequently insensitive

8
to STM recall deficits (Lezak, 1983). In a series of factor
analyses, Wechsler (1987) found digit span to consistently
load primarily on attention-concentration rather than on a
memory factor. Unpublished data communicated by T. White
(June 12, 1990) revealed that for 182 patients seen in a
neuropsychology clinic, there was an insignificant
correlation (r= .123, p=.098) between digits forward and
trial one of the CVLT, a measure of STM recall.
The capacity of STM has been shown to be approximately
7 +2 chunks of information (Miller, 1956). Unfortunately
the notion of chunks has not been clearly defined. Recall
significantly exceeding the accepted capacity of 7 +2 has
been attributed either to the participation of LTM or
efficient organization of bits of information into larger
chunks. There are no known limits on the duration or
capacity of LTM.
An inherent feature of multistore models has been the
necessity of elucidating the processes through which
information is progressively transferred through the memory
storage areas. The mechanism underlying transfer of
information from sensory memory to STM has been attributed
to attentional processes associated with pattern (i.e.,
meaning) recognition (Moray, 1959). In contrast, Shiffrin
(1975) has proposed that perceptual stimuli are
automatically encoded and passed from sensory to STM. From
a multistore perspective, transfer from STM to LTM has been
viewed as a function of active attention or rehearsal

9
(Rundus, 1971). This model explains the observation that,
on a supraspan task, a higher percentage of the first
(primacy) and last (recency) words tend to be recalled. The
primacy effect is believed to be due to encoding in LTM.
The recency effect is hypothesized to be a manifestation of
STM. Thus, immediate recall on a supra-span task may be
examined for primacy and recency effects to provide
indications of the integrity of STM and LTM.
Besides dissimilarities in capacity and duration, there
appear to be differences in the cognitive processes
requisite to the maintenance of STM and LTM. In general,
most studies and clinical observations have suggested that
phonological or maintenance rehearsal (i.e., mere parroting)
usually serves only to maintain information in STM, whereas
more complex, elaborative or semantic processing facilitates
transfer to LTM (Craik & Lockhart, 1972). Transfer to LTM
has been thought to usually require more active attending
and processing.
Demonstration of the differential effects of type or
level of processing led to the proposal that the observation
of apparently distinct memory stores could be explained in
terms of levels of processing within one memory store (Craik
& Lockhart, 1972). The apparently limited storage capacity
of STM has been reinterpreted by the levels of processing
model as limited processing capacity. The apparent
distinction between STM and LTM has been explained by the
levels of processing model as partially a function of depth

10
of processing, with semantic processing occurring at greater
depth than phonemic processing. However, research in other
areas has forced revisions in the original levels of
processing models (Cermak, 1982).
To account for studies demonstrating elaborateness of
processing to be more salient than type of processing (e.g.,
semantic vs. phonological), the levels of processing model
has been modified to emphasize extent rather than depth of
encoding (Craik & Tulving, 1975). Confounding all models,
other work has shown that shallowly processed information
may at times interfere with and even outlast more deeply
processed material (Cermak, 1982). In addition, phonemic
processing may result in recall superior to semantic
processing when retrieval is phonemic, that is, when
processing and retrieval are within the same domain (Morris,
Bransford, & Franks, 1977).
Data in STM is easily lost through interference and
decay. In contrast, reflecting the enduring nature of LTM,
failures of LTM have usually been considered to be produced
by retrieval deficits. The continued presence of the memory
may be assessed with recognition tasks. Recall tasks
require self-generation of cues. In contrast, in
recognition tasks cuing is externally provided. Thus,
recognition in the absence of spontaneous recall may be
explained by weak or decayed memory traces, failure to
generate adequate self-cuing or memory search strategies or

11
insufficient processing during acquisition resulting in
inadequate encoding.
Based on the cognitive literature, it may be concluded
that LTM is powerfully affected by the manner and extent to
which the learner processes the information. If retrieval
is based on semantic cuing, then semantic processing will be
the most efficient route for encoding. The more elaborate
the processing, the higher the probability of effective
retrieval. Conditions reducing cognitive processing may be
expected to impair LTM. Cognitive processing is likely to
be negatively affected by factors including underarousal,
overarousal, fatigue, attentional deficits, distracting
stimuli, lack of interest and depression (Cohen,
Weingartner, Smallberg, Pickar, & Murphy, 1982).
Under some circumstances memory deficits may be
categorized as either primarily functional or organic.
Functional factors, those affecting cognitive processing,
may either reduce effort expended or increase effort
required. Functional memory deficits may, therefore, be
considered a product of insufficient or misdirected effort.
Several methods exist by which, under certain
circumstances, various types of memory dysfunction may
sometimes be differentiated. By controlling the duration
and nature of processing (i.e., semantic, phonological or
orthographic), the levels of processing paradigm provides a
means of determining whether LTM deficits are characterized
by a failure to benefit from semantic processing. Failure

12
to benefit from semantic processing has been attributed to
encoding deficits secondary to organic dysfunction of memory
structures (e.g., Korsakoff's).
Automatic and Effortful Memory
Automatic memory is demonstrated by recall in the
absence of effortful encoding. Automatic memory is
unconscious and little affected by degree of effort (Jonides
& Naveh-Benjamin, 1987). Deficits in automatic memory
processes are suggestive of dysfunction at the level of
organic memory structures. Reductions in capacity such as
illness and, presumably, fatigue, have limited effect on
automatic memory (Hasher & Zacks, 1984). The frequency of
occurrence task provides a means of assessing automatic
memory function. Normal performance on a frequency of
occurrence task in conjunction with reduced performance on
an effortful task requiring effort for optimal performance
would suggest disruption at the level of effortful
processing. This pattern of deficits could be the result of
frontal lobe dysfunction, poor motivation or failure to
employ an effective strategy. One caveat, the use of
automatic memory measures to assess the integrity of the
neural substrates of effortful memory is predicated on the
assumption, not accepted by all, that effortful and
automatic memory are largely dependent on identical neuronal
systems (Roediger, 1990).

13
Neuropsychological Dysfunction in Adult Renal Patients
This section provides a cursory overview of the general
pattern of deficits seen in renal disease and discusses
verbal memory within the context of broader findings.
General Intelligence: Verbal and Nonverbal
In the absence of confounds, general verbal
intellectual ability has been reported to be relatively
spared in adult renal disease (Blatt & Tsushima, 1966;
Comty, Leonard, & Shapiro, 1974; English et al., 1978;
Fishman & Schneider, 1972; Freeman, Sherrard, Carlsyn, &
Paige, 1980; Ryan, Souheaver, & DeWolfe, 1980; Schupak,
Sullivan, & Lee, 1967; Trieschmann & Sand, 1971). In a 1982
review, Osberg, Meares, McKee and Burnett noted that in
every reported study, verbal IQ scores surpassed performance
scores by 5 to 14 points. The verbal-nonverbal discrepancy
is partly an artifact of the time constraints on many
nonverbal but not verbal intelligence subtests; however,
visuospatial and visuomotor deficits have also been widely
observed on untimed tasks (Mings, 1987).
The relative sparing of language function is consistent
with the general pattern of deficits often seen in diffuse
brain dysfunction. Speculations regarding the cause of this
verbal-nonverbal disassociation have included: the
nondominant hemisphere is more vulnerable to assault,
overlearned verbal abilities are more redundantly
represented and crystallized intellectual functions tax
cognitive capacity less than more novel visuospatial tasks.

14
Interestingly, McDaniel's (1971) data on visual
discrimination performance suggested impairment not
primarily as a result of difficulties with visual-motor
integration, but secondary to interference with cognitive
processing. The reported visuospatial motor dysfunctions
may be partially mediated by attentional deficits (Fennell,
Fennell, Mings, & Morris, 1986; McDaniel, 1971).
Attention. Level of Activation and Executive Function
Several authors have suggested the presence of global
attention deficits in uremia (Marshall, 1979; Stewart &
Stewart, 1979). Data reported by Trieschmann and Sand
(1971) on 83 subjects not treated with dialysis demonstrated
reduced performance on two measures loading on simple
attention and concentration, the Digit Span and Arithmetic
subtests of the Wechsler Adult Intelligence Scale (WAIS).
Grouping subjects based on severity of illness revealed
Digit Span to be the lowest subscale score in the more
severely ill group.
In a dialized population with normal premorbid verbal
IQ based on vocabulary knowledge, English, Savage, Britton,
Ward and Kerr (1978) found reduced performance on WAIS
subtests most heavily loading on attention (i.e.,
Arithmetic, Digit Span, Digit Symbol). In a pediatric
population, Digit Span was reported to be significantly
reduced regardless of treatment modality. Indeed, in a
pediatric population, excluding the renal transplant group,
Digit Span discriminated between renal and control groups

15
better than any other measure in the study (Fennell, et
al.,1987).
In contrast, Souheaver, Ryan and DeWolfe (1982) found
spared auditory attention skills, but on a different task,
the Seashore Rhythm test. In the Heilman, Moyer, Melendez,
Schwartz and Miller (1975) study, simple attention-
concentration as measured by Digit Span was within normal
limits (mean subscale score of 10.04 +2.18) and consistent
with verbal intelligence as measured by the Vocabulary
subtest (mean 10.00). However, attention was significantly
lower in renal patients (n=24) than in controls (n=12).
Overall, it may be concluded that attentional deficits are
common in uremia.
Uremia appears to depress activation and arousal
(Trompeter, Polinsky, Andreoli, & Fennell, 1986). This
underarousal may interfere with attention and increase
distractibility (Heilman, Moyer, Melendez, Schwartz, &
Miller, 1975). However, reduced level of activation is
unlikely to be the only mechanism involved. Assessments at
varying time intervals after hemodialysis sessions revealed
no change in Digit Span performance, despite changes in
speed and reaction time (Lewis, O'Neill, Dustman & Beck,
1980).
Renal disease also appears to negatively affect more
complex attentional and executive functioning such as the
ability to rapidly generate and/or shift sets. Performance
on Trails B, a task requiring maintenance of and alternation

16
between two sequences, speeded visual search and simple
visual motor coordination, has been reported to be reduced,
at times severely, in renal disease (Heilman, Moyer,
Melendez, Schwartz, & Miller, 1975; Ratner, Adams, Levin, &
Rourke, 1983; Teschan et al., 1974). Ginn (1975) reported a
correlation of .82 between Trails and severity of renal
failure as measured by serum creatinine for subjects not on
dialysis. Similar results were reported by Teschan and
colleagues (1974). Reduced performance on tasks demanding
sustained vigilance, such as continuous performance tasks,
has been consistently reported (Ginn, 1975; Osberg, Meares,
McKee, & Burnett, 1982). Ginn and colleagues (1978)
demonstrated worsening of performance on a vigilance task in
8 out of 10 subjects following a reduction in hemodialysis.
Contributions to executive system dysfunction may
include deficits secondary to reduced level of activation,
diminished attentional capacity, susceptibility of the
frontal-subcortical neural substrate to metabolic
abnormalities and dysfunction stemming from a generalized
difficulty with tasks demanding fluid as opposed to
crystalized intelligence.
Verbal Memory
Reduced vigilance, underarousal, increased
distractibility, increased response latencies and reduced
cognitive processing may contribute to global memory
deficits. Work in a pediatric renal population has revealed
a correlation between learning on a multitrial supra-span

17
verbal task requiring the learning of a list (e.g.,
Buschke), susceptibility to distraction (i.e., Auditory
Consonant Trigrams) and reduced sustained vigilance
(Fennell, Fennell, Mings, & Morris, 1986). This led to the
suggestion that reduced level of arousal may mediate STM
impairment in renal disease (Fennell et al., 1990a).
Because they place greater demands on sustained attention,
multitrial memory tasks may be more sensitive to uremia.
Overall, initiation of hemodialysis has been reported
to improve memory function (Osberg, Meares, McKee, &
Burnett, 1982). In adults, the consequences of uremia are
largely reversible and appear to be only minimally
cumulative in the absence of confounds such as aluminum
toxicity, dietary noncompliance, uncontrolled hypertension
or diabetes. Comparing dialysis patients and controls,
Hart, Pederson, Czerwinski and Adams (1983) found relatively
little difference in memory as measured by the Wechsler
Memory Scale (WMS). In addition, no significant
correlations were found between years of dialysis treatment
and memory performance.
Gilli and DeBastiani (1983) reported (n=54) a mild
relationship between duration of hemodialysis and decrements
in verbal intellectual ability (WAIS) and memory (WMS). The
WMS Memory Quotient, a composite measure of primarily short
term verbal and nonverbal memory, was initially above
expectations based on WAIS performance. For the 21
hemodialysis subjects who were retested, a minimum of 12

18
months later the WMS Memory Quotient had dropped to a level
only slightly above expectations for IQ. The authors
interpreted these results as suggesting a decline in memory
function in association with time on hemodialysis, perhaps
related to elevated parathyroid hormone levels. However,
fewer than half of the subjects were retested, and they
exhibited initial Memory Quotient scores above the average
for the entire subject pool and, as previously noted, above
expectations for IQ. The apparent decline in memory may
have been an example of regression towards the mean.
Ginn (1975) reported a temporary improvement in verbal
LTM on a recognition task the day after hemodialysis,
regardless of the level of performance prior to dialysis. A
correlation of .687 (pc.001) was reported between latency to
response on a single trial word-recognition task and
severity of uremia (Ginn et al., 1975). However, increased
response latency does not necessarily indicate retrieval
difficulties, especially in a population suspected of
generalized slowing.
Heilman, Moyer, Melendez, Schwartz and Miller (1975)
found significant verbal STM deficits on the Logical Memory
subtest of the WMS. In contrast, loss between immediate and
delay was comparable to controls. Ginn and colleagues (1975)
reported a significant negative correlation (pc.OOl) between
response latency on a verbal recognition memory task and
uremia as measured by serum creatinine. In nondialized
uremic subjects (n=23) participating in a longitudinal

19
study, Hagberg (1974) reported verbal STM as measured by
paired associate learning to be significantly lower than
that predicted by general verbal ability in 23 nondialized
uremic subjects. Six months after initiation of
hemodialysis, paired associate learning performance had
significantly improved in the 21 subjects retested and
approached expected levels relative to intelligence. This
study used alternate test forms but failed to use a control
group to control for possible effects of repeat testing.
Assessments of six subjects at varying time intervals
following hemodialysis revealed no changes in STM as
measured by a paired associate learning task, despite
changes in performance on timed tasks (Lewis, O'Neill,
Dustman, & Beck, 1980). In contrast, on a recognition task,
STM was shown to improve 24 hours after and deteriorate two
days following dialysis (Teschan et al., 1974). Other work
by this group has suggested improvement in verbal STM
following transplantation, but this has not reached
statistical significance (Teschan, Ginn, Bourne, & Ward,
1976).
Based largely on work with a pediatric population, it
has been suggested that STM and sensory/motor function are
more severely impacted than are more complex cognitive
abilities or LTM (Crittenden, Holliday, Piel, & Potter
1985). However, this report failed to provide sufficient
data to support this suggestion. It appears likely that the
measure of STM was Digit Span from the WAIS-R, which is more

20
of an attentional measure. Other work generally supports
the notion of a differential susceptibility for attentional
and visuospatial constructional functioning. Also in the
child literature, STM and LTM deficits have been reported on
multitrial supra-span memory tasks, with impairment in
normal developmental improvement in memory (Fennell, et al.,
1990a).
Nonverbal Memory
Visual memory as measured by the Benton Visual
Retention Test was impaired in nondialized renal subjects
(n=12), but normalized following 12 months of maintenance
hemodialysis, although the change was not significant
(Hagberg, 1974). In a related study, 20 adults, chronically
dialized for a mean of 39.7 months, were followed throughout
the hemodialysis cycle. Performance on the Benton Visual
Retention Test was mildly impaired on three separate
administrations, despite unimpaired ability to copy the
drawings (Ratner, Adams, Levin, & Rourke, 1983).
Facial recognition memory as measured by Milner Faces
was reported to be unimpaired in nondialized renal patients
(Heilman, Moyer, Melendez, Schwartz, & Miller, 1975).
Memory For Designs was shown by Hagberg (1974) to be within
the normal range for 23 nondialized uremic subjects.
Illustrating the possible value of longitudinal work, as
well as the need for alternate forms, readministration of
the same test form to 21 of the original subjects, following

21
6 months of hemodialysis, revealed a very slight
nonsignificant improvement.
In hemodialysis subjects (n=29), impairment in
visuospatial memory as measured by the Block Design Learning
Test, in a study lacking a control group, failed to reach
statistical significance. However, duration of dialysis did
correlate with poor performance (English, Savage, Britton,
Ward, & Kerr, 1978). However, it is unclear whether the
putative deficit was related to nonverbal memory or to
visuospatial constructional ability. In addition, factors
such as hypertension, diabetes and age may have contributed
to the apparent negative relationship between duration of
dialysis and nonverbal memory performance.
Ziestat, Logue and McCarty (1980) reported a
significant correlation between years on hemodialysis and
both short term (r=-.39) and long term (r=-.38) visual
memory as measured by the WMS. However, again, years on
dialysis may have been confounded by factors such as
etiology of renal failure. In contrast, verbal memory
showed no relationship with years on hemodialysis in this
study. The apparent decline in visual memory may have been
secondary to the well researched decline in visuospatial
processing (Mings, 1987).
Conclusions Regarding Cognitive Deficits in Renal Disease
Renal disease differentially affects cognitive
functioning in a manner grossly consistent with that seen in
many systemic conditions. The degree of cognitive

22
impairment correlates with severity of uremia, is at times
dramatically improved by initiation of dialysis and may be
nearly reversed by successful transplantation (Fennell,
Rasbury, Fennell, & Morris 1984; Ratner, Adams, Levin, &
Rourke, 1983). Adults on maintenance dialysis frequently
exhibit deficits in attention, vigilance, reaction time,
ability to rapidly generate and/or shift sets, visuospatial
ability, visual memory and verbal memory. Weaker evidence
suggests variability correlating with the phase of the
dialysis cycle, a possible advantage for peritoneal dialysis
over hemodialysis and decreased performance in association
with time on hemodialysis (Fennell, Fennell, Mings, &
Morris, 1986; Osberg, Meares, McKee, & Burnett, 1982).
Higher Hg levels in peritoneal dialysis compared to
hemodialysis and the possibility of increasing anemia in
conjunction with time on hemodialysis leave open the
possibility that level of anemia may play a role in the
latter two observations. Restoration of near normal
cognitive function following transplantation in well
dialyzed subjects may also, to an extent yet to be
determined, be mediated by the amelioration of anemia.
There are a number of problems with this literature.
For example, small sample size, absence of alternate test
forms, lack of relevant control groups, and inadequate
statistical procedures (Osberg, Meares, McKee, & Burnett,
1982). Additional contributions to discrepancies in the
literature include high variance in subject characteristics

23
including differences in medical management, age, age of
onset and etiology of renal disease. Many of the conditions
such as diabetes, hypertension and cardiovascular disease
contributing to or associated with renal failure have
neuropsychological consequences in their own right.
Discrepant findings in hemodialysis patients may to some
extent be a function of the time of assessment relative to
dialysis, the adequacy of dialysis and the degree of
electrolyte disequilibrium following dialysis. Most of the
early studies sampled neuropsychological function at one
point in time. There have been very few longitudinal
studies.
Progress in the treatment of renal disease has been
charted in the literature on the neuropsychology of uremia.
Prior to the availability of dialysis, the
neuropsychological consequences of renal failure often
included stupor and coma (Arieff, Guisado, & Massry, 1975;
Tyler, 1968). Prior to the discovery and acceptance of the
neurotoxicity of elevated blood aluminum levels in renal
patients, dialysis encephalopathy affected a significant
percentage of patients (Sprague et al., 1988). Improvements
in the methodology of dialysis as well as other aspects of
medical management continue to improve the physiological
conditions under which uremic nervous systems function.
Concurrently, these advances have also resulted in longer
durations of exposure to uremia as well as an aging renal
population. Thus, discrepancies in the literature may to

24
some extent be attributed to the differential effects of the
passage of time on technology and on individuals.
Overview of Anemia
This section will provide a brief overview of the
significance and etiology of anemia, adaptive mechanisms and
the treatment of uremic anemia.
Prevalence and Etiology
The world-wide prevalence of anemia has been estimated
to be 40% in children, 35% in adult females and 20% in adult
males (Lozoff, 1989). The defining feature of anemia is a
reduction in blood Hg levels. Anemia is the final common
pathway for a large number of diseases, conditions resulting
in blood loss and deficiency states, all of which result in
decreased Hg in the blood (Bunn, 1980a). The most common
form of anemia, iron deficiency anemia has been reported to
have a prevalence in the U.S. of 20% in adult women of
childbearing years, 50% in pregnant women and 3% in adult
males (Lee, Wintrobe, & Bunn, 1980). Probably the second
most common type, the anemia of chronic disorders is the
mild to moderate anemia frequently associated with chronic
inflammatory syndromes and infectious and neoplastic
diseases (Wintrobe et al., 1981). Anemia is a very common
condition.
Anemia almost invariably accompanies chronic renal
failure (Desforges, 1975; Erslev, 1975). Uremic anemia
often significantly reduces quality of life, despite
adequate dialysis. Prior to the development of rEPO,

25
approximately 25% of dialysis patients suffered from anemia
severe enough to require intermittent or regular red-cell
transfusions (Eschbach, Egrie, Downing, Browne, & Adamson,
1987).
Uremic anemia results from several interactive
processes. Shortened red blood cell survival is a frequent
manifestation of uremia (Jacob, Eaton, & Yawata, 1975).
Although controversial, some evidence suggests that uremic
toxins inhibit heme synthesis and erythroid progenitor cell
formation (Erslev, 1975; Jacob, Eaton, & Yawata, 1975).
Blood loss due to platelet dysfunction and as a result of
hemodialysis are implicated (Desforges, 1975). The
increased level of potentially toxic "middle molecules" in
hemodialysis compared to peritoneal dialysis may also
contribute to anemia in hemodialysis. Iron deficiency,
folate deficiency, B12 deficiency, aluminum-induced
microcytosis and hypersplenism have been cited (Paganini,
1989) .
One study has demonstrated a significant increase in
hematocrit and Hg levels following 12 months of endurance
exercise training, which suggests that sedentary life style,
a frequent consequence of end-stage renal disease, may
contribute to the anemia (Goldberg et al., 1986). However,
inadequate erythropoietin (EPO) production relative to the
degree of anemia is the major cause of uremic anemia. Thus,
uremic anemia is primarily due to an endocrine deficiency
state, correctable by rEPO replacement therapy (Chandra,

26
McVicar, & Clemons, 1988). Intravenously administered rEPO
has been demonstrated to restore Hg levels to normal in many
renal patients (Eschbach, Egrie, Downing, Browne, & Adamson,
1987; Winearls et al., 1986).
Function of Hemoglobin and Adaptation to Anemia
Under normal atmospheric and physiologic conditions,
more than 98% of the total oxygen carried by arterial blood
is bound to Hg, with the remainder dissolved in the aqueous
portion of blood (P02). To cross cell membranes, oxygen
must first disassociate from Hg and dissolve directly in the
blood. The function of Hg may be considered to be the
maintenance of normal P02 by providing a large reservoir of
available oxygen. Thus, in anemia the reservoir of oxygen
is reduced.
Anemic organisms may maintain normal tissue oxygenation
through a variety of adaptive mechanisms. Blood flow is
deviated to more vital areas. The affinity of Hg for oxygen
decreases. When Hg falls below 7.5 g/dl, an increase in
both heart rate and stroke volume results in a significant
rise in resting cardiac output (Bunn, 1980b) mediated by
increased sympathetic activation. Although PC02 is a more
potent controller of ventilation than is P02, low P02
resulting from severe anemia may stimulate peripheral
chemoreceptors on the carotid bodies resulting in increased
ventilation. However, the resulting reduction in PC02 and
developing alkalosis are likely to attenuate this response
(Vander, Sherman, & Luciano, 1980).

27
Hypoxia stimulates production of EPO primarily by the
kidneys. As the major regulator of erythrocyte production,
EPO stimulates erythropoiesis, the process in which
erythrocytes develop from stem cells in the bone marrow,
manufacture Hg from amino acids and body iron stores and
then enter the blood stream as mature erythrocytes.
Defective kidneys usually fail to produce a sufficient
increase in EPO in response to hypoxia (Erslev, 1987).
Erythropoietin Therapy and Iron Deficiency
In the absence of compliance with a regimen of iron
supplementation, the rapid erythropoiesis produced by
treatment with rEPO will seriously reduce iron levels (Van
Wyck, 1989). Subtle iron deficiency may be detected by low
serum ferritin levels. With a further decrease in iron
stores, low serum iron levels, increased iron binding
capacity, and elevated free erythrocyte protoporphyrin are
apparent. Iron deficiency anemia is a more severe
manifestation of iron deficiency and characterized by
reduced Hg levels (Ritchey, 1987). In the case of iron
deficient patients treated with rEPO, iron deficiency may
block further reduction of anemia and, as discussed in the
next section, may have neuropsychological consequences.
Neuropsychology of Anemia
Factors believed to contribute to neuropsychological
dysfunction of renal disease include the buildup of putative
neurotoxins, alterations in membrane permeability, and
electrolyte disequilibrium (Powell et al., 1986; Trompeter,

28
Polinsky, Andreoli, & Fennell, 1986). Attempts to isolate
specific uremic neurotoxins have met with limited success
(Schreiner, 1975; Trompeter, Polinsky, Andreoli, & Fennell,
1986).
Prior to the availability of rEPO, Hg values were
occasionally included in neuropsychological studies in renal
disease; however, perhaps due to the small range of values,
anemia was not found to be contributory (Fennell et al.,
1987; Gilli & DeBastiani, 1983). In continuous ambulatory
peritoneal dialysis (CAPD) Hg levels tend to run higher,
aspects of cognitive performance tend to be closer to normal
and "middle molecule" clearance is up to ten times greater.
Nissenson (1989) suggests that higher hemoglobin rather than
lower "middle molecule" levels may underlie the neuro
psychological advantage of CAPD over hemodialysis.
Numerous studies have documented the presence of
neuropsychological dysfunction in other forms of anemia.
However, not one of these studies has controlled for the
cause of the anemia, which in every case has been reported
to produce neuropsychological deficits. As all of the known
effects of rEPO are mediated by erythropoiesis, improvement
in function following rEPO replacement therapy may be
attributed solely to the rise in Hg and hematocrit.
Neuropsychological deficits are well documented in iron
deficiency anemia; however, while there is support for the
direct effect of iron deficiency on the brain, the role of
anemia controlling for iron deficiency has not been examined

29
(Deinard, List, Lindgren, Hunt, & Chang, 1986; Groner,
Holtzman, Charney, & Mellits, 1986). Until recently, with
the exception of blood transfusions, uremic anemia has only
receded following successful renal transplantation.
Although transplantation often enhances many
neuropsychological functions, there is a huge confound in
the simultaneous improvement in both uremia and anemia.
Therefore, the role of anemia in neuropsychological
dysfunction has not been adequately tested in any
population.
Effects of Anemia
Anemic patients may complain of a variety of symptoms
including fatigue, dizziness, dyspnea, poor concentration,
irritability, weakness, faintness, headache, impotence and
tinnitus (Bunn, 1980b; Rapaport, 1987). Some of the
complaints reported in conjunction with anemia may be due to
the cause of the anemia rather than the effects of low Hg
levels (Davies-Jones, Preston, & Timperley, 1980). The
relationship between severity of symptoms and Hg
concentration is confounded by the effectiveness of a
variety of adaptive mechanisms. Symptomatology is dependent
on factors including the degree and rate of reduction in
oxygen carrying capacity of the blood, degree and rate of
change in total blood volume, and the compensatory capacity
of the cardiovascular and pulmonary systems (Wintrobe et
al., 1981). Following adaptation symptoms may subside. If
anemia develops slowly, effective adaptation may avert the

30
appearance of symptoms. Even in cases of severe chronic
anemia (6 to 8 g per dl), the patient may be asymptomatic or
complain only of fatigue (Wintrobe et al., 1981).
Fatigue appears to be the most common symptom,
regardless of severity of the anemia. Comparing symptom
frequency in anemia characterized by Hg levels above or
below 8 g/dl, Dawson, Ogston and Fullerton (1969) reported
that in anemic subjects fatigue was present in 84.6% and
90.5%, respectively. In severe acute anemia, respiratory
and circulatory symptoms suggestive of possible hypoxia
(i.e., shortness of breath, "dizziness," faintness and
tachycardia) are common (Wintrobe et al., 1981).
Cognitive deficits have been documented in several of
the anemias (Davies-Jones, Preston, & Timperley, 1980;
Tucker, Sandstead, Penland, Dawson, & Milne, 1984).
However, previous studies have been confounded by the
failure to control for factors causing the anemia and
suspected of producing deleterious effects through
mechanisms unrelated to anemia. Nevertheless, existent
studies provide some relevant information. As it is the
most common, the best studied and the most relevant to the
current project (i.e., due to its possible appearance during
rEPO treatment), iron deficiency anemia will be emphasized.
Human studies examining the effects of iron
supplementation in deficient subjects have found increased
mental development scores in 15 month old infants, improved
attention, but not memory in 3-6 year olds, improved problem

31
solving capacity and improved STM and attention in adults
(Deinard, List, Lindgren, Hunt, & Chang, 1986; Groner,
Holtzman, Charney, & Mellits, 1986; Honig & Oski, 1984;
Oski, Honig, Helu, & Howanitz, 1978; Pollitt, Leibel, &
Greenfield, 1983; Pollitt, Siemantri, Yunis, & Scrimshaw,
1985; Rybo, Bengtsson, Hallberg, & Oden, 1985; Walter,
Kovalskys, & Steel, 1983).
Some authors attribute the neurological complications
of iron deficiency to anemia (Davis-Jones, Preston, &
Timperley, 1980). However, to date not one study has
examined the effects of anemia controlling for the effects
of iron deficiency. Anatomical and clinical evidence
suggests that iron deficiency may affect dopaminergic
systems (Birkmayer & Birkmayer, 1986, 1987; Drayer, et al.,
1986; Hallgren & Sourander, 1958). The mechanism may be a
reduction in the number of dopamine D2 receptors (Ben-
Shachar, Ashkenazi, & Youdim, 1986).
Iron responsive neuropsychological deficits have been
demonstrated in nonanemic iron deficient pregnant women and
infants (Groner, Holtzman, Charney, & Mellits, 1986; Oski,
Honig, Helu, & Howanitz, 1983). Evans (1985) noted that
improved cognitive function occurs rapidly following iron
supplementation, prior to correction of the anemia;
therefore, the neuropsychological deficits associated with
iron deficiency do not appear to be mediated by anemia.
Reviewing the pediatric literature, Ritchey (1987) concluded
that iron deficiency, in the absence of anemia, adversely

32
affects infant behavior and impairs both infant and toddler
performance on tests of cognitive function, with these
effects magnified by the progression to iron deficiency
anemia. However, subjects exhibiting anemia due to iron
deficiency anemia are likely to be more iron deficient than
nonanemic subjects.
Some studies have found no differences between anemic
and nonanemic subjects and attributed the positive findings
predominant in the literature to uncontrolled variables such
as socioeconomic status and education (Johnson & McGowan,
1983). However, this does not explain the results of iron
treatment studies. In addition, the results of animal
studies support the predominant human findings (Massaro &
Widmayer, 1981; Yehuda, Youdim, & Mostofsky, 1986; Youdim &
Ben-Shachar, 1987).
Possible Mechanisms
Mechanisms that might produce neuropsychological
dysfunction in anemia include alterations in blood
chemistry, cerebral hypoxia and general fatigue. As
previously noted, hyperventilation induced by low 02 is
possible, but is attenuated by the resulting hypocapnia
(i.e., low C02). The effects of severe hyperventilation
include increased blood pH, alterations in neuronal
excitability, changes in EEG and cerebral vasoconstriction
(Fried, 1987). Carbon dioxide concentration is the primary
mechanism regulating cerebral blood flow with hypocapnia
inducing vasoconstriction and hypercapnia inducing

33
vasodilation (Heistadt, Marcus, & Abboud, 1987). In the
anemic state vasoconstriction in response to hypocapnia is
likely to result in hypoxia.
Compared to other tissues, the brain both requires a
disproportionate amount of oxygen and is more sensitive to
hypoxia (Davis-Jones, Preston, & Timperley, 1980).
Neuropsychological deficits attributable to hypoxia have
been observed in chronic lung disease, normal patients
experiencing iatrogenic hypoxia and patients suffering from
sleep disordered breathing. The hippocampus is especially
vulnerable to hypoxia due to the nature of its blood supply.
Hippocampal dysfunction is associated with deficits in LTM
encoding (Muramoto, Kuru, Sugishita, & Toyokura, 1979). The
disassociation between LTM and STM seen in patients with
lesions in the hippocampal area is well known. However,
chronic hypoxia may have more widespread effects. Measures
of verbal and nonverbal memory, intelligence and verbal
fluency have been reported to be affected by hypoxia in
association with chronic sleep disordered breathing (Berry,
Webb, Block, Bauer, & Switzer, 1986).
The relevance of hyperventilation and cerebral hypoxia
to well managed hemodialysis patients is unknown.
Hemodialysis increases the degree of alkalosis.
Hyperventilation secondary to severe anemia might,
theoretically, increase the degree of alkalosis. Although
peripheral hypoxia may be common in uremic anemia, the
incidence of cerebral hypoxia is unknown. The possibility

34
of an interaction between factors affecting cognitive
function should not be discounted. Young adults one to
three years post mild head injury (i.e., concussion) exhibit
impairment on immediate memory and vigilance tasks relative
to controls when all subjects are tested under mildly
hypoxic conditions (Ewing, McCarthy, Gronwall, & Wrightson,
1980). Renal patients may be more susceptible to the
effects of anemia than are otherwise healthy anemic
patients.
In contrast to the unknown incidence of hypoxia and
hyperventilation, fatigue, the first and most chronic
symptom of anemia, is almost invariably present in end-stage
renal disease (Dawson, Ogston, & Fullerton, 1969). Fatigue
may be expected to consistently occur prior to the
development of cerebral hypoxia, if cerebral hypoxia occurs
at all. At a level of anemia severe enough to produce
cerebral hypoxia, it appears likely that the level of
fatigue will also be greatly increased. Therefore, fatigue
may be a more reliable mediator of the putative
neuropsychological effects of anemia.
The effects of fatigue on neuropsychological
performance are well known in clinical practice (Lezak,
1983). However, the effects do appear to be variable.
Subject factors such as age and physical condition and
fatigue factors such as cause and duration may be
significant. No changes in neuropsychological test
performance were demonstrated in 42 surgical residents when

35
assessed following 2.0 +1.5 hours of sleep vs. 6.5 +1.0
hours. In the acute sleep-deprived state, residents were
less vigorous and more fatigued, depressed, tense, confused
and angry (p<0.05) than they were in a rested state. These
changes in mood had no effect on measures including
immediate and delayed recall of logical stories, Trails,
immediate memory, or digit symbol (Bartle et al., 1988).
However, it is conceivable that more difficult tasks would
have shown an effect.
The biological function of the experience of fatigue is
hypothesized to be reduction in unnecessary activity. It is
hypothesized that the effects of increasing fatigue are
likely to appear along a continuum. With increasing fatigue
the organism's proclivity to define activity as necessary is
likely to decrease. In a testing situation, prior to
refusal to perform a task, the subject's definition of the
task may change to make the task less taxing. Therefore,
tasks not requiring, but enhanced by, additional effortful
cognitive activity may be expected to be most sensitive to
the effects of fatigue. Tasks overtly demanding attention,
vigilance and/or extensive cognitive processing are
hypothesized to be affected only as the severity of the
fatigue increases.
Effect of Erythropoietin on Neuropsychological Function
The effects of rEPO replacement therapy on quality of
life and neuropsychological functioning will be briefly

36
reviewed. The results of the pilot study and possible
negative side effects of rEPO will then be discussed.
Effect of Erythropoietin on Quality of Life
Treatment with rEPO has been demonstrated to increase
Hg concentration and reported to improve patient well-being
(Eschbach, Egrie, Downing, Browne, & Adamson, 1987). In the
first clinical trial, out of ten patients, nine reported an
improved sense of well-being and eight reported increased
exercise tolerance (Winearls et al., 1986).
A study of 37 rEPO treated chronic hemodialysis
patients found 16% of the patients returned to work,
appetite improved in 81%, 78% reported subjective
improvement in exercise tolerance, 70% participated in more
social activities, sleep improved in 68% and 84% reported an
increased sense of well being. Patients between 40 and 49
years of age showed the most improvement. A hematocrit
value of >27% appeared to be the critical level for improved
sexual function in men (Delano, 1989).
Evans, Rader and Manninen (1990) found significant
increases in self report of energy and strength, greater
appetite, improved breathing and reduced tension and
anxiety. There were very small but significant increases in
patient report of quality of life, affect and life
satisfaction.
In an investigation of the effect of rEPO on working
capacity, eight sedentary chronic hemodialysis patients rode
a bicycle ergometer before and after elevation of the group

37
mean Hg from 5.9 g/dl to 10.9 g/dl with rEPO treatment. The
initial mean anaerobic threshold of 70 watts represented a
very limited exercise capacity with inability to perform
routine household work over a sustained period. Following
rEPO treatment, the mean anaerobic threshold of 106 watts
indicated that work capacity had increased enough to allow
for full physical rehabilitation with regard to everyday
life (Mayer, Thum, Cada, Stummvoll, & Graf, 1988).
Previous Neuropsychological Studies
Several small pilot studies have assessed the impact of
rEPO on cognitive function. Wolcott, Schweitzer and Marsh
(1988) reported that in a group of nine chronic hemodialysis
patients administration of tests that loaded on visual,
conceptual and visuomotor tracking and auditory verbal
learning demonstrated trends toward improvement. Wolcott,
Schweitzer and Nissenson (1989), in a study of the effects
of EPO treatment in a group of 17 chronic hemodialysis
patients, reported improvements in symbol-digits modality
and Trails; however, it is unclear whether these
improvements reached clinical or statistical significance.
Decreased P3 latency purportedly represents increased
speed and efficiency of information processing. Increased
amplitude of event-related potentials suggests improvement
in cognitive function. Nissenon, Marsh and Brown (1988)
reported that following a mean rEPO mediated improvement in
hematocrit from 22.7% to 36.6%, 13 chronic hemodialysis
patients exhibited a decrease in P3 latency of response to

38
auditory stimuli, however, these results did not reach
statistical significance. A follow-up study (n=17) found no
change in P3 latency which suggested that there was no
change in speed of processing. The amplitude of the P3 wave
increased significantly (p<.03). Reported changes were 19%,
63% and 160% in the parietal, vertex and frontal areas
respectively, and are considered by the authors to represent
increased attention span, memory and efficiency of cognitive
processing (Nissenson, Marsh, Brown, Schweitzer, & Wolcott,
1989).
Results of Pilot Study
In a small pilot study, seven anemic hemodialysis
patients were assessed with a battery of neuropsychological
tests and a mood questionnaire (Profile of Mood States) pre
and a mean of 98.6 days post (range 77 to 125) initiation of
100 units/kg of rEPO (Klein et al. 1989). The study was
prospectively randomized and double-blind; however subjects
in most cases were able to accurately report their treatment
status, apparently based on subjective experience and
behavioral self-observation. Although eleven patients
initially began the study, complete data was only obtained
on seven due to a variety of factors including medical
problems, transplantation and patient refusal to be tested.
Results were available for two placebo controls and five
patients receiving rEPO. The mean increase in Hg was 0.55
mg/dl in controls and 5.2 mg/dl in those on active
treatment, a mean treatment group change from 6.10 +0.97

39
g/dl, severely anemic, to 11.56 +2.41, mildly anemic. At
sea level anemia is suspected in men when Hg is <14.0 g/dl
and <12.0 g/dl in women (Rapaport, 1987).
Table
1-1.
Pilot Studv Hemoglobin and
Sum CVLT
Sum CVLT
SS Change
Obs
Grouo
Hg
Pre
Hg
Post
Hg
Chancre
Sum CVLT
Change
1
Con
7.0
6.7
-0.3
-6
-0.93
2
Con
8.1
9.5
1.4
-5
-0.55
3
rEPO
6.0
8.4
2.4
1
0.11
4
rEPO
5.1
11.2
6.1
3
0.47
5
rEPO
7.1
13.2
6.1
8
1.08
6
rEPO
5.7
12.0
6.3
0
0.00
7
rEPO
6.6
13.0
6.4
16
1.90
Note:
Obs =
Observation; SS =
Standard Score; Con
= control
group
; rEPO
= treatment
group;
Change = change score
The Profile of Mood States, a self-report measure,
failed to reveal any improvement in fatigue, tension,
depression, vigor, or confusion. However, individual
patients reported dramatic improvements in circumscribed
domains. Individual subjects reported increased capacity
for physical activity, a reduction in post dialysis fatigue
and decreased problems with impotency. The dramatic
reduction in physical symptoms in association with absence
of subjective effects is consistent with reports of reduced
subjective symptoms when the anemia is chronic, even if it

40
is severe. Despite this, the absence of change on the
fatigue and vigor subscales is somewhat suspect and may
represent the inability of this subject population to
accurately report their subjective state in the manner
reguired by the POMS. Wolcott, Schweitzer and Marsh (1988)
also found an absence of significant changes on the POMS
following a rEPO mediated increase in mean hematocrit from
23.1% to 36.0% in 13 patients; however, mean POMS vigor and
fatigue scales changed in the expected directions and
approached statistical significance.
As predicted, there was no improvement on the
Comprehension subtest of the Wechsler Adult Intelligence
Scale Revised. This measure of overlearned semantic
stores is usually stable in the absence of severe
dysfunction. There was also no improvement in verbal
behavioral fluency as measured by the Controlled Oral Word
Association Test (FAS). Neither was there any increase in
motor speed as measured by finger tapping; however, two
patients exhibited guite noticeable differences in capacity
for sustained finger activity. During the first assessment
finger cramping, pain and fatigue reguired rests between
trials. Rests were not reguired following initiation of
rEPO treatment.
On Trails A and B, a test stressing motor speed, visual
search, mental control and decision speed, no improvement
was noted. On Trails A, one treated subject moved from
borderline to within normal limits. One treated subject,

41
moved from below normal to within normal limits; however,
this subject exhibited limited response to rEPO. The other
three treated subjects actually were slower post-test. The
greatest improvement was seen in one of the two controls,
who moved from clearly below normal to within normal limits.
Likewise, on Trails B, the greatest improvement was in the
treated subject who had only limited response to rEPO and
the second greatest improvement was in a control. Three of
the four treated, responsive subjects actually performed
more poorly on the second testing. These findings contrast
with the report of Wolcott, Schweitzer and Nissenson (1989)
that Trails improved in treated patients.
Correlation between improved performance on the Sum of
five trials on the California Verbal Learning Test (Sum
CVLT) supra-span word list learning task and improvement in
Hg was r=0.796 (p=.0324). The correlation between
improvement in Sum CVLT and rise in hematocrit was r=0.741
(p=.0568). Neither the first two trials of the five trial
sequence nor the once presented interference list showed
improvement. Improvement was observed primarily in the last
two trials. Single and double presentations of supra-span
lists with immediate recall are highly dependent on STM (T.
White, personal communication, June 12, 1990). They may be
considered only weak measures of LTM. Performance on the
fourth and fifth presentation is likely to be highly
dependent on the degree to which the subject has organized
and elaborated the list of target words. Thus, the improved

42
performance appears to have been due to enhanced LTM. This
is supported by the lack of improvement on the digit span
subtest of the Wechsler Intelligence Scale Revised.
However, the Logical Memory subtest of the WMS showed no
change.
Logical Memory is a single trial task and, because, as
the name implies, it is already logically organized, it
might be expected to less vigorously discriminate between
sparse and elaborate processing. Improved performance was
observed on the CVLT only after the third presentation.
Multitrial supra-span tasks require more active cognitive
processing than does the single presentation of a paragraph.
Given the hypothesized increase in general vitality, the
results may be explained by a greater proclivity, post
treatment, for effortful cognitive activity such as the kind
required to organize and elaboratively rehearse a shopping
list.
No improvement in immediate or delayed visual memory
was observed based on the Visual Reproduction subtest of the
WMS. This test involves only one presentation of the
stimuli; therefore, may not be less sensitive than Sum CVLT
to changes in degree of cognitive processing. However, the
lack of improvement is also consistent with the suggestion
that visual-motor integration skills are permanently
affected by renal disease, at least if age of onset was
during childhood (Mings, 1987).

43
Improved performance on Sum CVLT, in the absence of
other changes, is hypothesized to be due to reduction in
fatigue. It is hypothesized that the effects of fatigue are
likely to appear along a continuum mediated by attributions
regarding the importance of effortful activities. With
increasing fatigue the saliency of cues requisite to the
production of a given level of arousal is likely to change.
Nonessential, covert cognitive activity may diminish prior
to cognitive activity overtly and undeniably essential to
performance of the stated task. Therefore, do to the nature
of the tasks, neuropsychological measures directly stressing
overt attention and vigilance may be less affected.
However, tasks benefiting from but not requiring covert
effortful cognitive processing may be most vulnerable to the
effects of fatigue.
The dramatic reduction in anemia produced by treatment
with EPO has been shown to produce clinically significant
improvements in domains such as aerobic capacity, anaerobic
threshold, sexual potency, employment status, level of
social activity, perceived ability to engage in physical
exercise and improved appetite. In contrast significant
changes in self-reported mood have been less consistently
documented. Preliminary neuropsychological test data
suggests improvement on verbal supra-span multitrial list
learning tasks. In addition, preliminary reports suggest
possible improvement in other cognitive domains as measured

44
by increased amplitude of event-related potentials and paper
and pencil tests.
An alternate interpretation of the results of studies
examining the cognitive and affective effects of rEPO
treatment is that the placebo effect was a major factor in
the findings. Examination of Sum CVLT change scores in the
study by Klein and colleagues (1989) revealed a significant
portion of the variance to have been contributed by the two
control subjects, both of whom declined in performance.
This may be explained by negative placebo effect in the
subjects initially expecting to be on a drug reported to
produce major changes in life functioning. Likewise,
improvement in the treatment group was consistent with a
positive placebo effect. The failure of Nissenson, Marsh,
Brown, Schweitzer and Wolcott (1989) to replicate the
earlier study of Nissenon, Marsh and Brown (1988), a study
which found increased speed of processing based on P3
latency, is also, arguably, consistent the transitory
placebo effect often observed with new treatments.
Negative Effects
Treatment with rEPO is not without possible negative
neuropsychological consequences. Increased hematocrit
appears to raise peripheral resistance resulting in
increased hypertension in 10% to 22.2% of patients (Hori,
Onoyama, Iseki, Fujimi, & Fujishima, 1990). Hypertension is
a major problem in the management of renal disease. In the
first British clinical trials, Winearls and colleagues

45
(1986) reported that one of their ten patients developed
hypertensive encephalopathy. In addition, increased blood
clotting may potentially reduce circulation (Winearls et
al., 1986). Other side effects reported include anxiety,
lethargy, headache, body aches and, in 32% of subjects in
one study, iron deficiency (Delano, 1989). In addition,
concern has been raised that increasing hematocrit to normal
levels may compromise any existing renal function and reduce
the efficiency of dialysis (Koene & Frenken, 1990).
Curiously, nervous tissue is capable of producing EPO.
After renal carcinoma, cerebellar hemangioblastomas have
been reported to be the most frequent cause of elevated EPO
in association with neoplasms (Hennessy, Stern, & Herrick,
1967; Race, Finney, Mallams, & Balia, 1964; Waldmann, Levin,
& Baldwin, 1961). The proclivity of neural tissue to
produce EPO leads to speculation that EPO may normally be
produced by and have a function in the brain; however, there
is no evidence of EPO mediated neural effects in cancer
patients suffering from EPO secreting tumors. However,
physiological EPO levels are abnormally low for the degree
of anemia in renal failure, but often within the range found
in nonanemic populations. Therefore, EPO levels may be
raised dramatically during rEPO replacement therapy.
However, the suggestion that EPO might affect nervous tissue
is purely speculative.
Of more interest is the observation that rEPO treatment
may reduce sympathetic arousal. Cardiac output is increased

46
in severe chronic anemia. This adaptive reaction is
mediated by sympathetic activation. Recombinant EPO, by
reducing anemia, reduces demands for high cardiac stroke
volume and heart rate; therefore, may reduce sympathetic
activation (Hori, Onoyama, Iseki, Fujimi, & Fujishima,
1990). This could be of significance given suggestions of
neuropsychological underactivation in uremia.
Summary and Conclusions
Memory dysfunction has been reported in chronic renal
disease with reasonable consistency. Chronic renal disease
is characterized by a host of symptoms, many of which stem
from nervous system dysfunction. Over the past 25 years
numerous studies have examined cognitive dysfunction in
uremia. These studies have occurred within the context of
medicine's increasing ability to simulate, or in the case of
transplantation, restore normal renal function. However,
today, the physiological mechanisms and cognitive processes
underlying neuropsychological dysfunction in renal disease
are still largely unknown.
Recombinant EPO, a potent new treatment for the anemia
of renal failure, has produced dramatic improvements in work
capacity, sexual potency and quality of life in impressive
percentages of treated patients. Because many of the
symptoms once attributed to uremia now appear to be
dramatically reduced through the treatment of uremic anemia,
it seemed reasonable to hope that neuropsychological
dysfunction might be similarly affected. A small pilot

47
study found improvement only in effortful verbal LTM.
Furthermore, in retrospect, it was noted that the sometimes
reported advantage of peritoneal over hemodialysis and
decline in performance with time on hemodialysis may be at
least partially mediated by degree of anemia. Also, there
are suggestions that uremic memory dysfunction is at least
partially mediated by reduced vigilance, underarousal,
increased distractibility and cognitive slowing. These
observations are consistent with the hypothesis that
effortful verbal memory may be vulnerable to anemia,
secondary to mechanism such as fatigue or hypoxia.
The goal of the present study was to attempt to
replicate the improvement in LTM on a a multitrial supra-
span learning task which had been previously shown to
correlate with reduction of uremic anemia. Previous work
suggested that verbal memory tasks making fewer demands on
effortful cognitive processing showed no improvement. If
improvement were demonstrated on effortful memory tasks as
well as on automatic memory tasks and on the Levels of
Processing task, then the most plausible explanation would
be that uremic anemia deleteriously affects memory
structures, possibly mediated by hypoxia, although other
explanations might be posited. Improvement only in
effortful memory would suggest that the change was due to
increased cognitive processing, possibly secondary to
reduced fatigue or increased activation.

48
Failure to replicate would provide information useful
in several areas. The extent to which homeostatic
mechanisms adapt to chronic anemia is unknown. Data
delineating the parameters within which anemia affects
cognitive function could potentially contribute to the
determination of the optimal Hg at which to maintain
patients in end-stage renal disease. This is important
given the monetary costs and medical risks associated with
utilizing rEPO to treat uremic anemia.
Failure to demonstrate Hg responsive neuropsychological
dysfunction in rEPO treated uremic anemia would lend support
to the suggestion that the deficits documented in other
forms of chronic anemia are mediated by factors other than
anemia. In the case of iron deficiency anemia, if the
reported deficits are entirely due to the direct effects of
iron deficiency on the brain, then, since iron deficiency
appears prior to anemia, this would suggest the need for
more aggressive screening for iron deficiency, even in the
absence of anemia. Beyond scientific value, this would
suggest that millions of adults and children may suffer from
easily and inexpensively preventable neuropsychological
dysfunction.
Increased understanding of the neuropsychological
consequences of uremic anemia may 1) increase understanding
of the extent of the organism's ability to adapt to chronic
anemia, 2) contribute to the quarter century quest for the
uremic neurotoxin(s), 3) show the extent to which anemia

49
itself is likely to contribute to the cognitive deficits
observed in other forms of anemia, 4) assist in determining
the optimal level of Hg in treated renal patients and 5) in
the event of positive findings, provide preliminary data
regarding the mechanisms through which anemia undermines
neuropsychological functioning.

CHAPTER 3
METHODS
The present study was designed to broadly address the
question of whether change in anemia would be reflected by
change in memory.
Subjects
All subjects were adults in end-stage renal failure
and reasonably stabilized on some form of maintenance
dialysis. Subjects were drawn from the Shands Teaching
Hospital (STH) Adult Hemodialysis Clinic, the STH Dialysis
Home Training Clinic and the Gainesville Veterans
Administration Medical Center Hemodialysis Clinic. Age
ranged between 23 and 73 years. Subjects with uncontrolled
hypertension, mental retardation or known neurological
disease were not accepted into the study.
The treatment group consisted of 16 subjects on
maintenance hemodialysis and one subject on peritoneal
dialysis. It was planned that subjects would only be
accepted into the study into their eighth week of rEPO
treatment, based on the assumption that STH adult
hemodialysis patient hematocrits would be rising for
approximately five months. Therefore, a subject initially
tested at 8 weeks was expected to just be reaching
asymptote at the third assessment. However, subjects began
50

51
receiving rEPO earlier than anticipated and control
subjects became difficult to locate. Therefore, subjects
were accepted into the study beyond the eighth week,
reasoning that they could serve as controls in the event of
minimal additional change in Hg. Mean lag between
initiation of rEPO and the first assessment was 5.9 weeks
(sd=4.57).
The 18 control subjects did not receive rEPO.
Thirteen subjects, 72.2%, were on some form of peritoneal
dialysis (i.e., Peritoneal Dialysis, Continuous Ambulatory
Peritoneal Dialysis or Continuous Cycler Peritoneal
Dialysis) and five subjects, 27.8%, were on maintenance
hemodialysis. Given the small population from which to
draw controls, the ideal of matching controls to treated
subjects on age, education and ideally, race and sex, was
not considered feasible. Despite this, as will be
described later, the two groups were nearly identically in
domains other than treatment modality. However, mean Hg at
the first or baseline assessment was 8.46 g/dl (sd=1.42) in
the treatment group and 9.44 g/dl (sd=1.92) in the control
group.
All subjects were paid $15 at the completion of the
third assessment. Subjects were tested in their home,
their dialysis clinic, or in a STH Clinical Psychology
testing room.

52
Measures
California Verbal Learning Test
The California Verbal Learning Test (CVLT) is a multi-
trial supra-span word list learning task designed to be an
ecologically valid measure of multiple aspects of learning.
The CVLT quantifies parameters including STM, LTM,
retention over short and long delays, degree of
vulnerability to proactive and retroactive interference,
encoding strategies, effects of category cuing and the
frequency of perseverations and intrusions (Delis, Kramer,
Kaplan, & Ober, 1987).
Three versions were used. Form I is in general
clinical use. Form II is an alternate form developed for
research purposes (Delis, McKee, Massman, Kramer, &
Gettman, 1990). Form III, the Florida version of the CVLT,
was developed for this study. A description of the
construction and validation is provided in Appendix A. In
keeping with the results of the pilot study the primary
measure was the Sum of trials one through five of the CVLT
(Sum CVLT).
Levels of Processing Task
A 36 word levels of processing task in use in the
Shands Teaching Hospital Psychology Clinic was used and two
alternate forms were developed for this study. Alternate
forms two and three were designed with target words of
similar frequency of occurrence to that of Form I. Please
see Appendix B.

53
The levels of processing paradigm controls the nature
of effortful processing through manipulation of the task.
Subjects were visually presented a written word and asked
to respond to a guestion that required either orthographic,
phonemic or semantic analysis of the visual stimuli.
Following an interval of approximately 20 minutes, accuracy
in identification of the target words was measured with an
oral multiple choice task. Subjects able to benefit from
semantic processing may be expected to exhibit a higher
rate of correct recognition for semantically processed
stimuli. Subjects exhibiting memory deficits due to
reduced cognitive processing, because extent of processing
is controlled by the task, would be expected to perform
like normal subjects on this task. Normal subjects show a
recognition advantage for the semantically processed words
(Craik & Tulving, 1975). Subjects suffering from
neurologically based memory disorders affecting encoding in
LTM fail to benefit from semantic processing. Thus,
improvement in ability to benefit from semantic processing
would suggest physiological improvement.
Frequency Estimation Task
Estimation of frequency of occurence provides a
clinical measure of automatic learning believed to be
relatively invulnerable to factors such as fatigue, but
presumably vulnerable to physiological factors (e.g.,
hypoxia) directly interfering with the functioning of
memory structures (Hasher & Zacks, 1984).

54
After being instructed to try to recall as many words
as possible, subjects were read a list of words with some
words appearing more than once. Twenty minutes later
subjects were asked to estimate how many times each word
was read. Three versions were produced and are reproduced
in Appendix C. The three versions were administered in
consecutive order to all subjects. Prior to estimation of
frequency, subjects were tested on free recall.
Controlled Word Association Test
Controlled Word Association Test, a verbal fluency
task, provides a measure of retrieval unrelated to
encoding. It is also dependent on level of arousal,
motivation, verbal fluency and ability to initiate
behavior. Three commonly used sets of letter triplets are
available (Lezak, 1982).
Cognitive-Affective and Physical Behavior Questionnaire
A self-report questionnaire designed to assess
cognitive-affective and behavioral changes likely to be
caused by medical illness, was developed utilizing the
types of questions typically asked of medical patients to
evaluate these domains. The questionnaire generates a
Cognitive-Affective Scale and a Physical (i.e., physical
activity) Scale. The questionnaire was written to avoid
obvious response biases. The questionnaire and scoring
instructions are displayed in Appendix D.
In the absence of positive results this measure was
intended to determine whether there were positive changes

55
in behavior in the treated subjects. The same version was
administered at all three assessments.
Hemoglobin
In most instances, Hg levels were drawn on rEPO
treated subjects on a weekly basis. When blood work was
not done on the day of neuropsychological testing, pre and
post target date Hg levels were used to calculate status on
the day of the assessment. As Hg status tends to change
linearly in the absence of major blood loss, estimation is
likely to have been reasonably accurate. In the few
instances where Hg levels were lacking, estimates were made
based on hematocrit (Ritchey, 1987). It should be noted
that the experimenter was blind to Hg levels until the
completion of all other data collection.
Ferritin
When available, ferritin levels were obtained for each
subject to rule out iron deficiency as a possible confound
(Ritchey, 1987). Ferritin, the most sensitive measure of
iron levels, has been used in studies examining the
neuropsychological effects of iron deficiency. However, in
the absence of ferritin levels, iron saturation and/or
transferrin were used if available. When the dates of
neuropsychological testing and blood work failed to
coincide, estimates were made based on iron measures
obtained before and after the time of testing. As iron
status tends to change slowly and linearly in the absence
of major blood loss, estimation is reasonably accurate.

56
Iron status was used to classify subjects into low and
replete categories utilizing lab report guidelines. The
normal range for ferritin is frequently cited as around 10
ng/ml to 107 ng/ml. Ferritin levels of under 10 ng/ml were
considered indicative of iron deficiency. Ferritin values
under 30 ng/ml were classified as low. Normal and elevated
iron values were grouped as normal for the purposes of
these analyses.
Treatment with rEPO results in a heavy demand on iron
stores and induces iron deficiency in the absence of iron
supplementation (Van Wyck, 1989). For this reason, as part
of routine medical procedures, serum ferritin levels were
drawn on most rEPO treated subjects on a monthly basis;
however, the experimenter was blind to iron status until
the completion of neuropsychological data collection.
Schedule of Testing
Subjects were tested on three occasions. Assessment
Two was intended to be administered approximately 30 days
following the first assessment. Assessment Three was
scheduled for 60 days following the second assessment.
Patient illness, noncompliance and difficulty working
around transportation schedules significantly altered the
assessment timetable in some instances. Because time of
testing relative to hemodialysis has been reported to
affect results, this was kept consistent for each subject.
For most subjects, assessment occurred immediately
following hemodialysis.

57
Tests were administered in the same order at all three
assessments and all subjects were administered test
versions in the same order. Forms for the estimation of
frequency of occurrence and levels of processing paradigms
were administered in ascending order. For the CVLT, Form
II was administered at the first assessment and Form I at
the third assessment. This deviation from ideal
experimental design occurred because Form III had not been
fully validated prior to initiation of the study.
Therefore, to minimize losses in the event of poor inter
form reliability, Form III was administered at the second
assessment and forms were not counterbalanced.
Controlled Word Association letter triplets were
administered in the following order: CFL, FAS and PRW.
Because the letter triplets CFL and PRW can be directly
compared, the were placed in what were expected to be the
two most important assessments, one and three. FAS is the
most widely used letter triplet, but shares a letter with
CFL. To minimize possible practice effects for individual
letters in the comparison of CFL and PRW, CFL was
administered prior to FAS. As change scores rather than
absolute performance was compared, error variance stemming
from possible differences in difficulty between the three
letter triplets was minimized.
Tasks were administered in the following order:
Informed consent, collection of demographic data, CVLT,
levels of processing encoding phase, frequency of

58
occurrence encoding phase, Controlled Word Association
Test, Cognitive-Affective and Physical Behavior
Questionnaire, levels of processing recognition phase, CVLT
delayed tasks, frequency of occurrence estimation phase.
Statistical Methodology
Statistical analyses were performed using the
Statistical Analysis System (SAS) package of statistical
procedures available for the IBM personal computer. The
SAS General Linear Model procedure was used to test the
main hypothesis. Analyses involving variables violating
assumptions of normality and homogeneity of variance were
performed utilizing nonparametric procedures. The assigned
alpha for the main hypothesis was .05, with secondary
hypotheses tested more stringently based on the
requirements of multiple testing and the extent to which
the assumptions underlying parametric procedures were
violated.
Hypotheses
There was one main hypothesis and four secondary
hypotheses. Two of the secondary hypotheses were intended,
in the event of failure to reject the main hypothesis, to
provide preliminary data regarding possible mechanisms
mediating the relationship between Hg and cognitive
functioning.
Main Hypothesis
The main hypothesis was that adult renal dialysis
patients receiving rEPO replacement therapy would exhibit

59
improved verbal LTM as measured by Sum CVLT, compared to
dialysis subjects not receiving rEPO. This improvement was
hypothesized to be mediated by rise in Hg. Recombinant EPO
itself was not expected to have any effect on Sum CVLT.
Recombinant EPO treated subjects typically concurrently
receive iron supplementation; therefore, iron status was
not expected to change in the vast majority of subjects.
Nevertheless, it was necessary to control for iron status
because it has been a confound in the majority of studies
focusing on the neuropsychology of anemia as deficiency is
associated with neuropsychological deficits. A
multivariate analysis was expected to demonstrate a main
effect for Hg while controlling for rEPO treatment status
and iron status.
Secondary Hypotheses
The secondary hypotheses are described below.
The first secondary hypothesis was that self-reported
physical activity, as measured by the Cognitive-Affective
and Physical Behavior Questionnaire, would improve in
response to reduced anemia.
Despite self-reported changes in life functioning, the
pilot study failed to demonstrate any alteration in
affective state using a self-report adjective checklist,
possibly because subjects were unable to adeguately perform
the task, which required the ability to abstract and
introspect. Therefore, it was hypothesized that a
concrete, behaviorally oriented self-report device might

60
produce more accurate results. Thus, the second secondary
hypothesis was that self-reported behavioral indicators of
cognitive-affective state, as measured by the Cognitive-
Affective and Physical Behavior Questionnaire, would
improve in response to reduced anemia.
In the event of positive findings the testing of
several other hypotheses was planned. The intent of these
tests was to attempt to characterize the mechanisms
mediating improvement in memory. These were preplanned
exploratory analyses designed to focus on the question of
whether putative improvement in memory following reduction
in anemia is mediated by functional or more basic
physiological processes. Increased benefit from semantic
processing would be suggestive of enhancement of
functioning at the level of the physiological substrate of
the memory system. However, it was reasoned that
improvement in Sum CVLT, if observed, was more likely to be
mediated by increased cognitive processing secondary to
reduced fatigue; therefore, the third secondary hypothesis
was that rise in Hg would not be associated with increased
ability to benefit from semantic processing as measured by
the semantic score on a levels of processing task.
Reasoning along similar lines, the fourth secondary
hypothesis was that rise in Hg would not be accompanied by
improved automatic learning as measured by the ability to
estimate frequency of occurrence.

CHAPTER 4
RESULTS
Overview of Analyses
Prior to analysis, the data was inspected for outliers
and tested for the assumption of normality using qualitative
and quantitative tests available through the SAS Univariate
procedure. Hemoglobin and Sum CVLT, the primary independent
and dependent variables, respectively, exhibited good
approximations of normal distributions and homogeneity of
variance. In contrast, the distributions of most other
variables deviated from normality to some degree. For
variables failing the Shapiro-Wilks W test of normality, the
ratio of the standard deviation to the mean was computed.
Variables with ratios greater than .25 were further examined
quantitatively and graphically (Schlotzhauer & Littell,
1987). Procedures such as SAS General Linear Model are
relatively robust with regard to violation of normality;
however, great caution is required in interpreting analyses
in which variables also violate the assumption of
homogeneity of variance.
The assumption of homogeneity of variance between
groups was tested with the Folded F test. With the
exception of ferritin, all variables necessary for testing
of the secondary hypotheses exhibited reasonably good
61

62
between group equality of variance at baseline. To test the
impact of these violations of the assumptions of parametric
procedures, parametric and nonparametric tests of between
group baseline values were calculated for all subjects and
the results compared. As displayed in Appendix E, even
using an alpha of .05, only 2 tests out of 55 variables,
both highly unstable and insignificant variables, produced
discrepant results. It was concluded that the violations of
normality were unlikely to be of sufficient severity to
impact on the validity of SAS General Linear Model
parametric procedures, except for ferritin. Since ferritin
was used as a classificatory variable inequality of variance
had no appreciable effect on statistical procedures.
Demographic and descriptive statistics were calculated
using the SAS Means and Frequency procedures. To rule out
extraneous between group differences baseline variables were
compared. It was planned that the main hypothesis would be
tested with a repeated measures ANOVA using the General
Linear Model procedure.
A second analysis was planned to test the hypothesis
that the expected treatment effect was associated with Hg
change rather than rEPO itself and that the treatment effect
was not an artifact of variability in iron status. Although
a repeated measures analysis with the classificatory terms
group membership and iron status and the covariant Hg had
been planned, the unexpected finding of no change in Hg
status between the second a third assessments led to the

63
dropping of the third assessment from the preplanned
analyses. Therefore, the first test of the main hypothesis
was a GLM univariate procedure with treatment status as the
independent variable and Sum CVLT change score as the
dependent variable.
The second planned test was a GLM analysis of
covariance, the added independent factors being the
classificatory variable iron status and the continuous
variable blood Hg level. In actuality, change in iron
status affected only a small number of subjects; therefore,
controlling for iron status became of minimal importance.
Kruskal-Wallis nonparametric procedures were used to
test the hypotheses that self-reported physical vitality and
self-reported cognitive behavior improved in the treatment
group. If a main effect for Hg on Sum CVLT had been found
then further nonparametric analyses would have been
performed in an attempt to provide preliminary data to guide
future inquiry into the mechanisms mediating the effect.
Specifically, between group differences in ability to
benefit from semantic processing and changes in encoding of
incidental information were planned. Negative findings
eliminated this rationale for these tests. Nonetheless,
they were included in post hoc analyses.
Post hoc exploratory analyses focused on attempts to
disconfirm the negative findings and to explore other
aspects of the data. The observation that treatment status
was a poorer predictor of Hg change than had been

64
anticipated, combined with reasoning along the lines that
response to anemia may be quite idiosyncratic, led to the
decision to reclassify subjects based on Hg change scores.
Unfortunately, the small range of Hg change in most subjects
necessitated a choice between sacrificing either statistical
power or clinical relevance. Retaining statistical power,
subjects were first classified based on small distinctions
with unlikely and undemonstrated physiological significance.
Following this, compromising between sample size and
physiological significance, only subjects meeting the
criteria of Hg change scores of +1 g/dl were retained. CVLT
scores in subjects with clinically significant changes in Hg
status were then examined.
The post hoc hypothesis that reduction in anemia might
alter the pattern of learning was tested. The slopes of
learning curves for trials one through five of the CVLT were
calculated with SAS Regression Analysis procedures. A GLM
ANOVA then compared changes in slopes between groups.
Additional procedures were exploratory. Change scores were
calculated for all variables and the performance of specific
variables analyzed. Also, relationships between test
performance and questionnaire data were explored.
Statistical Assumptions
The basic assumptions underlying most analysis of
variance procedures include 1) normality and 2) homogeneity
of within-group variance. Analysis of variance procedures
are quite robust with regard to violation of the assumption

65
of normality as well as departures from homogeneity of
variance. However, the concurrent violation of both
assumptions calls at the very least for conservatism in
determination of significance.
As can be seen in Appendix F, Table F-l, based on the
Shapiro-Wilks W statistic computed by SAS Univariate, the
primary independent variable, Hg, and the major dependent
variable, Sum CVLT, were both normally distributed. Hg
values were slightly skewed to the right. Sum CVLT was very
minimally skewed to the left. Kurtosis in the distribution
of Hg values approached zero. For Sum CVLT, the kurtosis
value showed the tails to be slightly lighter than expected
in a normal distribution.
Iron levels tend to be quite elevated in dialysis
patients dependent on repeated blood transfusions. In
patients running relatively normal iron levels, rEPO
treatment may deplete iron stores. Therefore, it is not
surprising that serum ferritin violated the assumption of
normality. As it was used only to classify subjects in
terms of iron status, this violation was irrelevant to the
analyses.
Other measures required for the testing of secondary
hypotheses for the most part violated assumptions of
normality to some degree. For variables failing the
Shapiro-Wilks W test of normality, the ratio of the standard
deviation to the mean was computed, as displayed in Appendix
F, Table F-l. Variables with ratios greater than .25 were

66
further examined quantitatively (i.e., kurtosis and skew)
and graphically (i.e., box plots)
The estimation of frequency of occurrence data violated
the test of normality. The ratio of the standard deviation
to the mean was 0.262, just over the established limit. The
distribution was somewhat skewed to the left; however,
Kurtosis was minimal.
The Semantic score generated by the levels of
processing task failed the Shapiro-Wilks W test of
normality; however, the ratio of the standard deviation to
the mean was within acceptable limits. The Orthographic
variable from the levels of processing task also failed the
test of normality and the standard deviation was large
relative to the mean.
The two Cognitive-Affective and Physical Behavior
Questionnaire scales failed the Shapiro-Wilks W test and the
standard deviation to mean ratio was .280, slightly above
the .25 recommended cut-off. The scores were slightly
skewed to the left and the tails were heavier than expected
in a normal distribution. The variables age, years of
education and years on dialysis all failed the Shapiro-Wilks
test. However, the ratio of standard deviation to mean was
within accepted limits for the education variable. Slope
and intercept of the learning curve for CVLT trials 1-5,
variables used for post hoc analyses, both passed the
Shapiro-Wilks test. Slope had somewhat heavy tails.

67
The assumption of equality of variance between groups
was tested with the Folded F test, F'. The results, as
displayed in Appendix F, Table F-2, revealed the following
CVLT variables to fail the test of homogeneity of variance
at the .10 level: Sum of clusters, Sum of intrusions, and
Sum of perseverations for trials one through five, Short
Delay perseverations and intrusions, and all three CVLT
measures of false positives not on the interference list.
Other measures failing the test of homogeneity of variance
were: levels of processing recall of orthographically
processed stimuli, years of education and ferritin.
Many major variables violated the assumption of
normality to some degree, but exhibited acceptable
approximations of homogeneity of variance; therefore, they
were considered candidates for parametric analyses. These
variables included: CVLT Short Delay, Long Delay, Short
Delay Cued Recall, Long Delay Cued Recall, Recognition Hits,
Levels of processing Semantic recognition, estimation of
frequency of occurrence and associated free recall and both
scales from the Cognitive-Affective and Physical Behavior
Questionnaire. Please see Appendix F, Table F-2. The
Orthographic variable generated by the levels of processing
task was normally distributed but failed the test of
homogeneity of variance. Ferritin, Sum CVLT Clusters and
years of education were neither normally distributed nor had
equality of variance; therefore, parametric procedures were
considered unsuitable for their analysis. With the

68
exception of ferritin, all variables necessary for
preplanned tests exhibited both reasonably good between
group equality of variance and were normally distributed.
As the effects of ferritin were judged to be nonlinearly
related to ferritin concentration, it was decided to use
ferritin as a classificatory variable.
To test the impact of the above violations of the
assumptions of parametric procedures, parametric and
nonparametric tests of between group baseline values were
calculated for all subjects and the results compared. As
displayed in Appendix E, even using an alpha of .05, only 2
tests out of 55 variables produced discrepant results. They
were CVLT variables that may be expected to be relatively
unstable: Short Delay Perseverations and Long Delay
Intrusions. This test suggested that the violations of
normality were unlikely to be of sufficient severity to
impact on the validity of SAS General Linear Model
parametric procedures.
The statistical characteristics of the distributions
and variance of all variables may be viewed in Appendix F.
Tables F-l and F-2.
There was one debatable outlier. Select analyses were
re-run following deletion of this subject. Examination of
the raw data revealed that at the first assessment the
subject had recalled 8 words on the first trial of the CVLT,
but only 3 on the fourth trial and 5 on the fifth trial.
Other data for this subject was within expected limits.

69
Descriptive Statistics
Age
The treatment group had a mean age of 43.24 years
(sd=13.45) and consisted of 17 subjects. The mean age of
the 18 subjects in the control group was 48.22 years
(sd=14.19). A nonparametric Kruskal-Wallis test procedure
failed to demonstrate a statistically significant between
group difference (X2(l, n=35)=1.13, p<.2948).
Education
Mean years of education in the treatment group was
11.41 years (sd=3.02) and 11.82 years (sd=2.15) in the
control group. A Kruskal-Wallis test procedure failed to
reveal a significant statistical differences between groups
(X2(1, n=35)=0.23, pc.6325).
Sex
Overall, the subject population was equally divided by
sex with 17 males and 18 females. The treatment group
consisted of 7 males and 10 females. The control group
consisted of 10 males and 8 females. Thus, the control
group was 55.6% male and the treatment group was 41.2% male.
A Chi-Square procedure failed to demonstrate a statistically
significant difference (X2(l, n=35)= 0.724, p< 0.395).
Treatment Modality
In the control group 72.2% were on some form of
peritoneal dialysis (i.e., Peritoneal Dialysis, Continuous
Ambulatory Peritoneal Dialysis, or Continuous Cycler
Peritoneal Dialysis). Five subjects or 27.8% of the control

70
group were on maintenance hemodialysis. In contrast, 94.1%
of the treatment group was on maintenance hemodialysis and
one subject representing 5.9% of the treatment group was
maintained on peritoneal dialysis. As confirmed by a Chi-
Square procedure these differences were highly significant
(X2(1, n=35)=16.032, pc.OOOl).
Determination of time on dialysis was made difficult by
intervening periods of renal transplantation as well as
inconsistencies between medical records and patient reports.
Also, this figure fails to take into account possible
differences in the effects of hemodialysis and peritoneal
dialysis over time. Mean years on dialysis for the
treatment group was 6.64 years (sd=7.69) compared to 2.85
(sd=3.76) years for the control group. These differences
approached but failed to reach statistical significance
(X2(l, n=35)=2.8425, p<.0918) based on the Kruskal-Wallis
Test procedure.
Attrition
All 17 subjects in the treatment group completed the
study; however, one 25 year old male died six weeks after
completion of the third assessment. Two subjects in the
control group, both on hemodialysis, died between the second
and third assessments. In addition, one control subject
maintained on peritoneal dialysis declined to complete the
third assessment after becoming hospitalized in a
neighboring city. Eliminating the three subjects not
completing the study resulted in a mean age in the control

71
group of 46.49 years (sd=13.61) and a mean of 12.10
(sd=1.63) years of education. The results of a Wilcoxon 2-
Sample Test (S=267.00, Z=0.7177, p>.4729) failed to
demonstrate a significant age difference between the
treatment (n=17) and control (n=15) groups. Likewise, the
Wilcoxon 2-Sample Test procedure failed to demonstrate a
significant difference in years of education between the
treatment (n=17) and control (n=15) groups (S= 259.00,
Z=0.440, p>.6596).
Cognitive function has been reported to fall prior to
nonaccidental death (White & Cunningham, 1988). Consistent
with this phenomenon of terminal decline, all three subjects
were in the predicted direction, although in two cases the
drop was extremely small. For the three deceased subjects,
mean Sum CVLT in the last assessment prior to death was
25.56 compared to 28.70 in the previous assessment. The
overall effect of terminal drop was to slightly lower
control group assessment two and three scores relative to
the treatment group.
Time Intervals Between Assessments
The mean time span between assessments one and two was
51.12 days (sd=32.95) for the treatment group. Control
group mean time interval between the first and second
assessments was 56.50 days (sd=35.98). The 5.38 days
greater time interval for the control group failed to reach
statistical significance based on a Wilcoxon Rank Sum Test
(S= 277.500, Z=-.92647, p>=.3542).

72
Mean days between assessments two and three was 67.80
(sd=34.56) in the control group and 90.41 days (sd=24.03) in
the treatment group. Based on the results of a Wilcoxon
Rank Sum Test the 22.61 days greater inter-test interval for
the treatment group was clearly statistically significant
(S=178.50, Z= -2.588, £>=.0096).
The number of days between assessments one and three
was 141.53 days (sd=34.53) in the treatment group (n=17) and
117.67 days (sd=28.24) in the control group (n=15). The
results of a Wilcoxon Rank Sum test suggest that the 23.86
day difference between these two groups was statistically
significant (S=181.00, Z= -2.495, p>.0126). Mean time lag
between initiation of rEPO and the first assessment was 41.4
days (sd=32.1)
Hemoglobin
Mean Hg at the first or baseline assessment was 8.46
g/dl (sd=l.42) in the treatment group and 9.44 g/dl
(sd=1.92) in the control group. The nature of these
differences suggested a possible trend toward statistical
significance as demonstrated by a GLM Analysis of Variance
procedure (F (1, 34)=2.94, p=.096). which is consistent
with reports of lower Hg levels in hemodialysis than in
peritoneal dialysis. Baseline mean hematocrit levels were
28.19% (sd=6.12) in the control group and 25.58% (sd=3.77)
in the treatment group. As demonstrated by a GLM ANOVA
procedure these differences failed to reach statistical
significance (F(l, 34)=2.29, £=.140).

73
As expected, a one-way GLM procedure demonstrated that
between the first and second assessments the mean Hg rise of
1.13 g/dl in the treatment group (n=17) was significantly
greater (F=7.42, p=.0102) than the -0.21 g/dl Hg change in
the control group (n=18). At the time of the second
assessment mean Hg was higher in the treatment group
(M=9.59, sd=1.84) than in the control group (M=9.23,
sd=1.81) but the difference was not statistically
significant (F(l, 34)= 0.34, £>=.564).
GLM analysis of variance of Hg change scores between
the second and third assessments failed to reveal even a
trend towards a between group difference (F(l, 31)= 0.16.
P=.6964). Lack of change in the second phase of the study
contributed to the failure to find no more than a trend
towards a significant rise in Hg status between the first
and third assessments (F=3.19, p=.0841). At the third
assessment mean Hg was 9.54 g/dl (sd=1.48) in the treatment
group (n=17) and 9.62 g/dl (sd=1.92) in the control group
(n=15).
In this sample the 21 hemodialysis patients exhibited a
mean baseline Hg of 8.80 g/dl (SD=1.80) and the 14
peritoneal dialysis patients had a mean Hg level of 9.21
g/dl (sd=1.69). The differences were not statistically
significant (F (1, 34)= 0.44, p= .51). Within hemodialysis
patients, mean baseline Hg in the controls (n=5) was 9.83
g/dl (sd=2.53). In the treatment group mean Hg at the first
assessment was 8.48 g/dl (sd=1.48). Based on a T-Test

74
procedure these differences failed to reach statistical
significance (T=1.5099, £=.1475).
Comparison of the maintenance hemodialysis pretreatment
Hg of 7.06 g/dl (sd= 1.42) in the rEPO group (n=16) to the
baseline Hg of 9.83 g/dl (sd=2.53) in the control group
(n=5) revealed a trend toward lower values in the treatment
group. However, a T-Test procedure failed to demonstrate a
statistically significant difference between groups
(T=2.3359, df=4.8, p=.0692).
Prior to initiation of rEPO therapy, mean Hg in the
treatment group was 7.39 g/dl (sd=1.94). At the first
assessment treatment group Hg had risen to 8.46 g/dl
(sd=1.42). Based on a T-Test these differences were not
statistically significant (T=-1.8361, df=32.0, p=.0756).
A GLM analysis of variance procedure revealed a main
effect for sex (F (1, 34)= 15.24, p=.0004). Mean male
(n=17) Hg was 9.96 g/dl (sd=1.61) and mean female (n=18) Hg
was 8.03 g/dl (sd=1.31), which is consistent with the 2 g/dl
lower Hg level reported in normal females compared to normal
males in the general healthy adult population. There was no
treatment by sex interaction.
Excluding the 3 subjects who did not complete the
study, rise in Hg was 0.04 g/dl in the control group (n=15).
As anticipated for the control group, mean change in Hg
between assessments one and three was not statistically
significant (F(l, 32)= 0.07, p=.786). In contrast the rise
in Hg was 1.08 g/dl (sd=1.23) between assessments one and

75
three in the rEPO treated group which was statistically
significant (F(l, 33)= 4.74, p=.037).
In the time lag between initiation of rEPO and the
first assessment mean Hg levels rose 1.07 g/dl (sd=1.90)
from an initial level of 7.39 g/dl (sd=1.94). The increase
of 1.08 g/dl in the treatment group (n=17) between
assessments one and three was 50% of the total increase of
2.15 g/dl (sd=1.23) in Hg over the course of rEPO therapy.
However, this total increase of 2.15 g/dl was only 39% of
the increase obtained in the pilot study. Please see Tables
4-1 and 4-2 for of Hg levels and change scores,
respectively.
For descriptive purposes subjects were classified using
the following system: very severe anemia < 6 g/dl Hg, severe
anemic 6-8 g/dl, moderate anemia 8-10 g/dl, mild anemia
10 12 g/dl, > 12 g/dl normal. As illustrated in Table 4-
3, between the first and second assessments seven treated
subjects showed an improvement in category and one declined.
In the controls four rose and three fell. Between the
second and third assessments change appeared to be random.
Ferritin
Because of the tendency of rEPO treatment to draw down
iron stores ferritin levels were assayed more often in rEPO
treated subjects; therefore, baseline ferritin levels were
available for 94% (n=16) of the treatment group but only 33%
(n=6) of the controls. At baseline mean treatment group
(n=16) ferritin level was 545.93 ng/ml (sd=734.50) and mean

76
control group (n=6) ferritin level was 523.62 ng/ml
(sd=626.13). A Kruskal-Wallis Test failed to reveal a
statistically significance difference between these two
groups
Table 4
(X2(1, n= 35)= 0.3478, p<.5553).
-1. Mean Hemocrlobin bv Grouo and
Assessment
Ass
Gro
N
Min
Max
Mean
Sd
pre trt
rEPO
17
4.80
12.70
7.39
1.94
1
rEPO
17
5.60
11.30
8.46
1.42
2
rEPO
17
6.60
14.20
9.59
1.84
3
rEPO
17
6.90
12.25
9.54
1.48
1
Con
18
5.27
13.90
9.44
1.92
2
Con
18
6.30
12.60
9.23
1.81
3
Con
15
6.00
12.30
9.62
1.92
Note: Ass = Assessment;
Grp =
Group; N =
Number of
subje<
Min = Minimum; Max = Maximum; Sd = Standard deviation
Con = Control; rEPO = rEPO treated group
Table 4-2. Mean Ho Change by Group and Test Intervals
Test
Interval
Gro
N
Min
Max
Mean
Sd
First
Con
18
-2.95
1.70
-0.21
1.34
Second
Con
15
-2.10
1.90
0.15
1.14
Pre-Ass
rEPO
17
-4.60
4.70
1.07
1.90
First
rEPO
17
-0.70
5.90
1.13
1.57
Second
rEPO
17
-5.50
1.80
-0.05
1.59
Total
rEPO
17
-2.40
5.30
2.15
1.86

Id
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
101
102
103
104
105
106
107
108
77
4-3. Anemia Status Across Assessments
Assess 1 Assess 2 Assess 3
Severe
Moderate
Moderate
Severe
Very Severe
Moderate
Moderate
Severe
Moderate
Moderate
Severe
Severe
Moderate
Moderate
Moderate
Moderate
Severe
Severe
Moderate
Very Severe
Moderate
Normal
Moderate
Normal
Moderate
Moderate
+
Moderate
Moderate
Moderate
+
Severe
+
Normal
+
Severe
-
Moderate
+
Moderate
Moderate
Moderate
+
Severe
Moderate
Moderate
Moderate
Moderate
Moderate
+
Moderate
+
Moderate
Severe
+
Moderate
Normal
Severe
-
Moderate
+
Moderate
3
Moderate
Moderate
Moderate
Severe
Severe
Moderate
Moderate +
Severe
Moderate
Moderate
Moderate
Severe
Moderate
Moderate
Normal +
Moderate
Moderate
Moderate
Moderate
Severe

Moderate

Normal +
Moderate

78
Table 4-3 continued
Id
Assess 1
Assess 2
Assess 3
109
Moderate
Moderate
Moderate
110
Moderate
Moderate
Moderate
111
Moderate
Moderate
Moderate
112
Moderate
Moderate
Moderate
113
Moderate
Moderate
Moderate
114
Moderate
Moderate
Moderate
115
Moderate
Severe
Very Severe
116
Moderate
Moderate
Moderate
117
Severe
Moderate
+
118
Moderate
Severe
Severe
Note:
Valence sign = direction of
Id < 100 = treatment group;
categorical change;
Id > 100 = controls
As displayed in Table 4-4, subjects were classified
based on ferritin level (under 10 Deficient, 10 30 Low,
over 30 Replete). Classifying subjects into Replete, Low
and Deficient categories revealed little variability in iron
status across assessments.
Table 4-4. Number of Subjects in each Category of Blood
Iron Level
Iron Assess 1 Assess 2 Assess 3
Status Con rEPO Con rEPO Con rEPO
Replete 14 11 14 9 12 11
Low 14 16 13
Deficient 01 00 01

79
At the time of the first assessment, the Control group
included one subject with Low iron status and 14 Replete
subjects. At Assessment Two this was unchanged. At
Assessment Three there was still one subject with Low iron
status. In the rEPO group at Assessment One, there was one
Deficient subject, four Low iron subjects and 11 Replete
subjects. At Assessment Two, there were 6 Low iron and nine
Replete subjects. By the third assessment, the treatment
group was composed of one Deficient, three Low and 11
Replete subjects. For the subjects on which data was
available there was minimal change in iron status;
therefore, the need to control for iron deficiency was
greatly reduced.
Replete subjects were further divided into Replete and
High (over 1,000). As illustrated in Table 4-5 below, the
percentage of subjects exhibiting low iron status was
consistently greater in the rEPO treatment group as was the
percentage of subjects with high iron status. Also apparent
from the Table 4-4, iron status was relatively stable across
assessments, especially if the top two and bottom two groups
were collapsed into Low and Replete categories.
Baseline Neuropsychological Functioning
Table 4-6 displays selected raw and standard scores for
CVLT II at the baseline assessment. There are no
statistically significant differences for any of the scores
listed in Table 4-6 or in the more comprehensive list of
scores in Appendix G. However, despite the lack of

80
statistically significant differences, examination of raw
scores suggests a tendency for somewhat better performance
in the treatment group at the first assessment.
Table 4-5. Percentage of Subjects in each Category of Iron
Status
Iron
Assess
Con
1
rEPO
Assess
Con
2
rEPO
Assess
Con
3
rEPO
Hi
6.7
23.5
0
20.0
0
18.8
Normal
86.7
41.2
93.7
40.0
92.3
50.0
Lo
6.7
29.4
6.7
40.0
7.7
25.0
Defic
0
5.9
0
0
0
6.3
Note: Hi = Elevated Ferritin; Lo = Low; Defic = Deficient
The mean baseline age corrected control group Sum CVLT
score placed the group in the 4th percentile, the borderline
range. The mean baseline Sum CVLT score in the treatment
group placed it in the 9th percentile, the Low Average
range. However, the differences, as previously stated, did
not reach statistical significance.
As displayed in Appendix G, Table G-l, on the
Controlled Word Association test, using the letters CFP, the
control group (n=17) produced a mean of 11.53 words
(sd=3.84). The treatment group (n=17) produced a mean of
10.53 (sd=4.45) words. The difference was not statistically
significant.
The levels of processing task revealed that both groups
benefited from semantic processing compared to either

81
orthographic or phonemic processing. There was a very
slight nonsignificant advantage for phonemic over
orthographic processing for both groups. At baseline the
treatment group appeared to demonstrate greater benefit from
semantic processing than did the control group, although
this did not reach statistical significance. Performance on
the estimation of frequency of occurrence task was nearly
identical between the two groups.
Table 4-6.
CVLT Raw and
Standard
Scores at
Baseline
Control Group
n=18
Mean Sd
rEPO Treated Group
n=17
Mean Sd
P
value
Trial 1
5.72
1.74
5.88
2.47
.9464
standard
-1.33
0.91
-1.35
1.37
.7700
Sum 1-5
43.22
10.70
45.82
13.03
.5222
T score
33.06
10.31
33.82
16.00
.8662
Short Delay
9.06
2.71
9.65
3.26
.9867
standard
-1.11
0.83
-1.06
1.43
.9864
SD Cued
10.94
2.69
10.94
2.90
.7393
standard
-0.83
1.04
-0.88
1.36
.9052
Long Delay
9.94
3.44
9.47
3.76
. 6660
standard
-0.89
1.02
-1.41
1.66
.4344
LD Cued
11.06
3.51
10.82
3.34
.5587
standard
-0.78
1.40
-1.24
1.56
.3863
Note: Sd = Standard deviation; SD = Short Delay; LD = Long
Delay; standard = Standard Score

82
Test of Main Hypothesis
The main hypothesis stated that rEPO mediated reduction
in chronic anemia in adult dialysis patients is associated
with improved verbal LTM as measured by the CVLT Sum of
trials one through five.
Mean Sum CVLT change scores between the first and
second assessments were -0.82 (sd=5.38) in the treatment
group (n=17) and 4.06 (sd=9.91) in the control group (n=18).
These results were in the opposite of the hypothesized
direction. A GLM analysis of covariance procedure failed to
demonstrate a main effect for either treatment status
(F=l. 29, p=.2638) or Hg (F=1.32, p=.2590).
With both treatment status and Hg change in the model,
few conclusions could be drawn. Therefore, separate
univariate GLM analyses were performed for treatment status
and for Hg change. However, these analyses failed to show
a statistically significant main effect for either Hg
((F=3.25, p=.0805) or for treatment status (F=3.22,
p=.0818).
Examination of mean Sum CVLT scores between groups and
across assessments revealed a curious pattern. Although
none of the differences reached statistical significance,
examination of Table 4-7 reveals that the control group
performed poorly on the first assessment relative to the
other two assessments. In contrast, assuming equivalence of
test forms, treatment group performance was static. That
is, change apparently occurred unrelated to rEPO.

83
Table 4-7. Mean Sum CVLT Across Assessments and Between
Groups
Grp
Variable
N
Mean
Sd
Min
Max
Con
CVLT II
18
43.22
10.70
23
60
Con
CVLT III
18
47.28
10.68
30
68
Con
CVLT I
15
47.07
10.40
32
68
Trt
CVLT II
17
45.82
13.03
21
68
Trt
CVLT III
17
45.00
13.36
16
64
Trt
CVLT I
17
45.12
11.61
23
69
Representing iron status as a two-level classificatory
variable (i.e., Low vs. Replete), a GLM analysis of variance
failed to demonstrate a main effect for iron status on Sum
CVLT change scores between the first and second assessments
(F=0.01, p=.9384). Similar results were obtained
introducing Hg into the model (Hg: F=0.26, p=.6126; iron:
F=0.01, p=.9337).
Despite the lack of change in Hg status between the
second and third assessments, a GLM Repeated Measures ANOVA
was performed. As expected, the hypothesis of a treatment
effect was not supported by Wilks7 Lambda (F=0.2134,
E=.8092) or any other measure. Neither were there between
subject effects, within subject effects or treatment by
assessment interactions.
Post Hoc Analyses
Learning Curve
Hypothesizing that change in learning curve might be
more sensitive to the effects of anemia than Sum CVLT, SAS

84
Regression procedures were used to generate slopes of the
learning curves for trials one through five for each
administration of the CVLT.
A GLM procedure analyzing change scores between the
first and second assessments failed to demonstrate a
relationship between Hg change and change in slope (F=0.08,
p=.7766). An identical analysis examining change scores
between the second and third assessments produced similar
results (F=0.00, p=.9583). A GLM univariate analysis of the
effect of treatment status on slope change scores between
the first and second assessments failed to reach statistical
significance (F= 3.67, p=.0651). Controlling for the
increased probability of a Type I error as a result of
multiple analyses, the finding can not be considered
suggestive of a trend. A Wilcoxon Rank Sum Test produced
similar results (Z= 1.83, p=.0669). Further analysis
revealed the change to be primarily in the treatment group,
but in the opposite of the expected direction. In the
treatment group, mean slope declined -0.84 (sd=1.59), and in
the control group slope improved 0.15 (sd=1.27) between the
first and second assessments. A GLM univariate analysis of
the effect of treatment group status on slope change scores
between the second and third assessments failed to suggest a
relationship (F=1.30, p=.2628).
Despite extensive analyses, absolutely no evidence was
found that even mildly suggested improvement in learning
curve in association with improvement in Hg status.

85
Exploratory Analyses
Observation of variability in Hg change scores in both
groups led to reclassification based on direction of Hg
change, rather than treatment status. A total of 23
subjects exhibited a rise in Hg between the first and second
assessments (M=1.22 g/dl, sd=1.26). Twelve subjects
exhibited a fall between the first and second assessments
(M=-1.05, sd=.96). The difference between groups with rEPO
treatment status as the classificatory variable produced a
mean difference in Hg change between groups of 1.03 g/dl.
In contrast, the alternate system of classification doubled
the Hg change score yielding 2.27 g/dl. Difference in Hg
change scores between the positive and negative change
groups was highly significant (F=29.75, p=.0001).
Control subjects suffering a fall in Hg (n=9) exhibited
a mean fall of -1.27 g/dl (sd=1.02). The three rEPO treated
subjects exhibiting a fall in Hg had a mean drop of -.38
g/dl (sd=.28). Among rEPO treated subjects with rising Hg
(n=14) the mean rise was 1.46 g/dl (sd=1.54). Subjects from
the control group exhibiting a rise in Hg (n=9) showed a
mean rise of 0.85 g/dl (sd=0.51).
For exploratory purposes, the main hypothesis was
retested with a GLM procedure utilizing sign of Hg change as
the classificatory variable. There was a significant
difference in Sum CVLT change scores between the positive
(n=23, M=-l.22, sd=7.18) and negative (n=12, M=7.25,
sd=7.63) Hg change scores (F=10.52, p=.0027). The analysis

86
was rerun excluding one possible outlier; however, this had
little effect on the results (F=7.82, p=.0087). The results
were consistent with the original finding (i.e., using
treatment status as the classificatory variable) in that
they were in the opposite of the predicted direction. In a
GLM multiple analysis of variance treatment status was
nonsignificant (F=0.59, p=.4482) while sign of Hg change was
significant (F=9.45, p=.0043). Also consistent with the
original finding, as illustrated in Table 4-8 below, mean
Sum CVLT scores revealed poorer performance at the time of
the first assessment in the group whose Hg levels would
rise.
Table 4-8. Mean Sum CVLT by Sian of Ha Change Between
Assessments 1 & 2
Hg Fell n=12 Hg Rose n=23
Asses Mean Sd Mean Sd
1 36.58 9.48 48.61 10.86
2 43.83 11.80 47.39 12.07
In an attempt to reduce error variance contributed by
subjects exhibiting very small insignificant changes in Hg,
subjects with a Hg change of under 1 g/dl were excluded from
the analysis. Mean Sum CVLT change in the negative Hg
change group was 9.33 (sd=10.63) and -2.23 (sd=4.30) in the
positive Hg change group. A Wilcoxon Rank Sum Test
revealed the difference to reach statistical significance at
the .05 level (Z=2.373, p=.0176).

87
To test the effect of Hg changes within the range
closer to that likely to be considered physiologically
relevant, subjects with less than a 2 g/dl Hg change were
excluded from the final analysis. The two negative Hg
change subjects in the data set had a mean Hg drop change of
2.68 g/dl. The three positive subjects in the data set
exhibited a mean Hg rise of 3.77 g/dl. A GLM analysis of
this data set (n=5) failed to reveal a main effect for Hg
(F= 0.03, p=.8803); however, the small sample size greatly
reduced the power of the test and, again, the change was in
the opposite of the predicted direction. Mean CVLT
improvement was 2 words in the falling Hg group and the rise
was 1.33 words in the rising Hg group.
Hypothesizing that the observed effect of Hg change on
Sum CVLT might be mediated by affective change, the
Cognitive-Affective self-report variable was included in a
GLM analysis of variance model. A multivariate analysis
demonstrated main effects for both direction of Hg change
(F=13.27, p=.0010) and Cognitive-Affective self-report
(F=8.29, p=.0072). Further analysis revealed a main effect
for direction of Hg change (F=16.97, p=.0003) and an
interaction between direction of Hg change and self-reported
Cognitive-Affective behavior (F=6.53, p=.0043). When
Cognitive-Affective behavior was included in this model it
failed to reach significance (F=1.88, p=.1807). Further
analyses suggested no relationship between Cognitive-
Affective behavior and sign of Hg change (r=.11056,

88
2=.5336). Correlation between Sum CVLT and Cognitive-
Affective self-report was positive (r=.37042, p=.0310).
Examination of mean scores by groups revealed that Sum CVLT
rose in association with rise in Cognitive-Affective
behavior; however, there was less evidence of a relationship
when Hg fell.
In a model analyzing change in Sum CVLT, covarying
change in self-reported Physical Activity and direction of
change in Hg failed to demonstrate a main effect for
Physical Activity (F=0.77, p=.3866). There was no
interaction between sign of Hg change and Physical Activity
(F=0.75, p=.4827). Neither was there an effect on Sum CVLT
for self-reported level of Physical Activity at the time of
the first assessment. Analyses failed to demonstrate a
significant relationship between self-reported Physical
Activity and Sum CVLT
Sum of CVLT Trials Four and Five
Using sign of Hg change scores as the classificatory
variable, a Wilcoxon 2-Sample Rank Sum Test was used to test
the post hoc hypothesis of a change in the sum of CVLT
trials four and five between the first and second
assessment. The difference between groups was significant
even controlling for multiple comparisons (Z=2.6366,
P=.0084). Once again the difference was the reverse of the
predicted direction. Mean increase in number of words
recalled for the sum of trials four and five was 3.75
(sd=5.24) in the group with falling Hg (n=12) and -0.87

89
(sd=4.00) in the group with rising Hg (n=23). Please see
Table 4-9.
Table 4-9. Mean Change in CVLT Trials 4 plus 5 Change
Scores Between Assessments I & II
Gro
N
Min
Max
Mean
Sd
Pos
23
o
o

CO
6.00
00

o
4.00
Neg
12
o
o

T
18.00
3.75
5.24
An identical analysis between the second and third
assessments revealed a mean change of -0.78 words (sd=3.93)
in the negative change group (n=9) and a mean change of
-0.15 words (sd=3.96) in the positive Hg change group
(n=20). These differences failed to reach statistical
significance (Z=-.2845, p=.7760). Analysis of change scores
between the first and third assessments revealed a mean
change of -3.71 words (sd=7.06) in the negative change group
(n=7) and a mean change of -1.25 words (sd=4.23) in the
positive Hg change group (n=24). These differences failed
to reach statistical significance (Z=1.5613, p=.1184).
Multivariate analysis (n=35) with a General Linear
Model failed to find a main effect for rEPO treatment (i.e.,
experimental vs. control) on Sum CVLT (F=0.84, p=.3668) in a
model including a sign of Hg change factor (F=7.45,
p=.0102).
To test the possibility that undiscovered outliers
unduly influenced the results, Sum CVLT change scores were

90
classified based on their sign. A Wilcoxon Rank Sum
procedure analyzed differences in sign of Sum CVLT change
scores between the first and second assessments. With the
sign of Hg change scores serving as the grouping variable,
again there was a significant effect for Hg change
(Z=2.9180, p=.0035). Analysis using a Chi-Square procedure
produced similar results (X2=8.887, df=l, p=.003).
Short Term Memory
Exploratory analysis revealed a significant difference
at .01 in CVLT Trial One, a measure loading heavily on STM.
However, the change was in the opposite of the expected
direction between treatment and control groups between
assessments one and two and is likely a type two error.
Furthermore, equivalent change, also significant at .01, in
the opposite direction, was observed between the second and
third assessments. Please see the results displayed in
Appendix H.
Verbal Fluency
An analysis of variance procedure suggested that verbal
fluency as measured by Controlled Word Association test
performance was somewhat low in the control group at the
first assessment relative to the overall pattern of
performance in both groups across three assessments (F=4.99,
E=.0328. Utilizing sign of Hg change as a classificatory
variable, Controlled Word Association test performance was
not significantly different between assessments (F=0.01,
P=.9761).

91
Secondary Hypotheses
Levels of Processing
Treatment status was insignificant in explaining
variability in performance on any portion of the Levels of
Processing Task, including the Semantic recall measure
(F=0.43, p=.5164). Mean change scores are listed in
Appendix H. Re-analysis using direction of Hg change as a
classificatory variable produced similar results (F=2.12,
P=.1564). This analysis would have been considered a
secondary hypothesis if there had been a positive
association between Hg and Sum CVLT.
Estimation of Frequency of Occurrence
As displayed in Appendix H, there were no significant
differences in Frequency Estimation performance between
groups or assessments. Parenthetically, LTM, as measured by
free recall of the single presentation target word list,
improved slightly more in the treatment group than in the
control group; however, this failed to reach significance
controlling for multiple analyses.
Quality of Life
Quality of life is operationally defined here in terms
of the items on the Cognitive-Affective and Physical
Behavior Questionnaire and is divided into cognitive-
affective and overt physical behavior domains. It was
hypothesized that self-reported physical vitality, as
measured by the self-report questionnaire, would improve in
response to treatment with rEPO.

92
Because Physical Activity was not normally distributed,
a Wilcoxon Rank Sum Test procedure was used. Analysis
failed to support the hypothesis of a main effect for
treatment status on Physical Activity between the first and
second assessments (Z=1.1283, p=.2592). Similar results
were obtained using Hg change as the independent variable.
It was hypothesized that self-reported behavioral
indicators of Cognitive-Affective state would improve in the
treatment group. A Wilcoxon Rank Sum Test procedure failed
to demonstrate a change in Cognitive-Affective behavior (Z=-
1.7654, p=.0775) between the first and second assessments.
Because the results could arguably be interpreted as failing
to reject the possibility of a trend, the direction of the
changes was examined. As displayed in Table 4-10 below
change between the first and second assessments was in the
opposite of the predicted direction. Direction of change
between the second and third assessments was in the
predicted direction. Analysis of change scores between the
first and second assessments using a GLM procedure failed to
demonstrate a main effect for Hg (F=0.40, p=0.5336).
Using sign of Hg change scores as the classificatory
variable, a Wicoxon Rank Sum test failed to reveal
statistically significant differences in self-reported
Physical Activity (Z= -1.0235, p=.3061). Neither were there
statistically significant differences in self-reported
behavior in the Cognitive-Affective domain (Z= 0.2028,
P=.8393) .

93
Table 4-10. Self-reported Cognitive-Affective State Change
Scores bv Group & Assessment Interval
Test Control Group Treatment Group
Interv N Mean Sd N Mean Sd
One 18 0.94 2.21 16 -0.19 1.33
Two 15 -1.73 2.02 16 -0.13 1.41
Note: Interv = Change between assessments one and two or
between two and three.
Test Reliability
Two alternate forms of the Levels of Processing Task
and Frequency of Occurrence Task were developed specifically
for this study. A self-report questionnaire was written to
sample physical and cognitive-affective domains thought
likely to be affected by change in severity of anemia. None
of these new measures underwent a prior validation study.
Also, one alternate form of the CVLT was developed and
validated for use in this study, but validated on a younger
and better educated population. Because the only apparent
changes in test performance were in the controls, controls
which might have been used in a test validation study, the
lack of clear-cut findings produced the opportunity to
examine the reliability of these instruments in a medical
population.
CVLT Alternate Form Reliability
As displayed in Table 4-11, for the overall group the
correlation between Forms I and III and between Forms III
and II was quite good relative to correlations between Forms

94
I and II. Closer examination revealed a similar pattern for
the treatment group. This pattern is displayed in Table 4-
12. However, as shown in Table 4-13, the control group
exhibited poorer correlations between Form II and Form III
and between Form II and Form I than did the treatment group.
Table 4-11. Correlation of Sum CVLT Between Forms I, II and
III for All Subjects
CVLT II
Pearson Correlations
CVLT III
CVLT I
CVLT II
r
1.00000
0.75543
0.71063
P
0.0000
0.0001
0.0001
N
35
35
32
CVLT III
r
0.75543
1.00000
0.80342
P
0.0001
0.0000
0.0001
N
35
35
32
CVLT I
r
0.71063
0.80342
1.00000
P
0.0001
0.0001
0.0000
N
32
32
32
Note: r = correlation; p = probability value;
N = subject number
Table 4-14 demonstrates that Form III, the version
created for this study, correlates as well with both Forms I
and with II as Form I correlates with Form II. In addition,
means scores were remarkably similar between the three forms
of the CVLT.

95
Table 4-12. Correlation Between Treatment Group Sum CVLT I.
II and III
Pearson Correlations
CVLT II
CVLT III
CVLT I
CVLT II
r
1.00000
0.91732
0.77220
P
0.0000
0.0001
0.0003
N
17
17
17
CVLT III
r
0.91732
1.00000
0.81127
P
0.0001
0.0000
0.0001
N
17
17
17
CVLT I
r
0.77220
0.81127
1.00000
P
0.0003
0.0001
0.0000
N
17
17
17
Note: r = correlation; p = probability value
Table 4-13. Correlation Between Sum CVLT I. II and III for
Controls
Pearson Correlations
CVLT II
CVLT III
CVLT I
CVLT II r
1.00000
0.57038
0.64116
P
0.0000
0.0134
0.0100
N
18
18
15
CVLT III r
0.57038
1.00000
0.78866
P
0.0134
0.0000
0.0005
N
18
18
15
CVLT I r
0.64116
0.78866
1.00000
P
0.0100
0.0005
0.0000
N
15
15
15
Note: r = correlation; p = probability; N = subject number

96
Table 4-14. Correlations & Sum CVLT Means Between CVLT
Forms I. II and III
Variable Pearson Correlations Form Mean Sd
I-II
I-III
II-III
Trial
r
.48240
.59965
.47084
I
5.800
2.098
1
P
.0033
.0003
.0065
II
5.800
2.139
N
35
32
32
III
5.938
1.795
Trial
O
r
.56327
.60335
.61043
I
8.143
2.171
c*
P
.0004
.0003
.0002
II
8.629
2.377
N
35
32
32
III
8.969
2.957
Trial
O
r
.59901
. 66308
.65356
I
9.400
3.098
J
P
.0001
.0001
.0001
II
9.943
2.786
N
35
32
32
III
10.063
2.213
Trial
r
.42348
.47538
.64775
I
10.286
2.906
4
P
.0112
.0060
.0001
II
10.800
3.095
N
35
32
32
III
10.344
2.925
Trial
r
.64340
.52880
.75095
I
10.886
3.132
5
P
.0001
.0019
.0001
II
11.086
3.128
N
35
32
32
III
11.031
2.753
Sum
r
.75543
.71063
.80342
I
44.486
11.788
P
.0001
.0001
.0001
II
46.171
11.930
N
35
32
32
III
46.031
10.926
Sum
r
.64837
.80121
.73281
I
12.229
9.540
Clus
P
.0001
.0001
.0001
II
13.143
9.638
N
35
32
32
III
13.781
10.222
Sum
r
.13204
.27777
.07528
I
3.343
2.828
Psv
P
.4496
. 1237
.6822
II
3.800
3.261
N
35
32
32
III
4.219
2.685

97
Table 4-14 continued
Pearson
Variable
Correlations
Form
Mean
Sd
I-II
I-III
II-III
Sum
r
-0.001
.02699
.50981
I
0.857
1.912
Intrus
P
0.996
.8834
.0029
II
1.771
2.510
N
35
32
32
III
1.906
3.104
List B
r
.40311
.54579
.33823
I
5.857
1.768
P
.0181
.0012
.0583
II
5.588
2.176
N
34
32
32
III
5.594
2.434
Short
r
.67782
.73638
.72612
I
9.343
2.960
Delay
P
.0001
.0001
.0001
II
9.353
3.524
N
34
32
32
III
9.281
3.113
Short
Delay
r
.72150
.67170
.58327
I
10.943
2.754
Cued
P
.0001
.0001
.0005
II
10.559
2.977
N
34
32
32
III
10.344
2.598
Long
r
.76124
.79658
.77339
I
9.714
3.553
Delay
P
.0001
.0001
.0001
II
9.471
3.703
N
34
32
32
III
9.469
3.360
Long
r
.65964
.70126
.56638
I
10.943
3.378
Delay
Cued
P
.0001
.0001
. 0007
II
10.529
2.926
N
34
32
32
III
10.500
2.712
Recog
True
r
.30641
.44776
.31827
I
14.800
1.324
Positive
P
.0780
. 0102
.0759
II
14.324
1.492
N
34
32
32
III
13.813
2.546
Recog
r
.77802
.73247
.67942
I
2.543
3.576
False
Positive
P
.0001
.0001
.0001
II
3.236
3.593
N
34
32
32
III
3.000
3.312

98
Table 4-14 continued
Note: Under Correlations, the first value listed is the
correlation (r) under that the p value, and then, N,
the number of subjects; Psv = Perseveration; Intrus =
Intrusions; Recog = Recognition
Levels of Processing Reliability
Forms II and III of the Levels of Processing Task were
developed for use in this study. As illustrated in Table 4-
15, correlations between recall of orthographically
processed words on Levels of Processing Forms I, II and III
were all positive, but low and nonsignificant. Correlations
between phonemically processed stimuli followed an identical
pattern between Forms I and II and I and III. However, the
correlation between Forms II and III reached statistical
significance. The opposite was true of the semantically
processed words. Correlations between Forms I and II and
Forms I and III reached statistical significance.
Frequency of Occurrence Reliability
Table 4-16 illustrates a finding identical to that
of the Semantic Levels of Processing results, in that
correlations reached statistical significance between Forms
I and II and between Forms I and III, but not between II and
III.
Physical Behavior Self-Report Test-Retest Reliability
The data displayed in Table 4-17 demonstrates quite
acceptable test-retest reliability for the Physical Behavior
scale. Correlation coefficients ranged from r=.820 to

99
r=.898. Examination of the means and standard deviations
shows essentially unchanged scores between administrations.
Table 4-15. Correlations and Means Between Levels of
Processinq
Test Forms I
. II & III
Form
Mean
Sd
I & II
Correlations
I & III
II & III
Orthoqraohic
r .28992
.12925
.23972
I
7.531
2.155
p .1136
.4961
.2020
II
6.548
2.336
N 31
30
30
III
6.967
2.498
Phonetic
r .23765
.26180
.54589
I
7.625
2.268
p .1980
.1623
.0018
II
8.903
2.371
N 31
30
30
III
8.367
2.220
Semantic
r .40394
.56396
.16411
I
9.938
1.831
p .0242
.0012
.3862
II
10.323
1.796
N 31
30
30
III
9.800
1.472
Note: r = correlation;
p = probability; N =
subject number
Table 4-16. Correlations & Means Between Estimation of
Frequency of Occurrence Test Forms I, II & III
Pearson
Correlations between
Forms
Form
Mean
Sd
I & II
I & III
II & III
r
.51046
.43870
.10965
I
11.343
2.029
P
.0020
.0120
.5502
II
10.471
3.184
N
34
32
32
III
9.875
2.871

100
Table 4-17. Test Retest Reliability of Physical Activity
Self-Report
Pearson
Correlations between Forms
I & II
I & III
II & III
Form
Mean
Sd
r .89863
.76990
.82068
I
9.230
4.538
p .0001
.0001
.0001
II
9.029
4.596
N 34
32
31
III
9.031
4.575
Note: r = correlation; p = p value; N = sample size
Table 4-18. Test Retest Reliability of Cognitive-Affective
Self-Report
Pearson
Correlations between Forms
I & II
I & III
II & III
Form
Mean
Sd
r
.64224
.86002
.68524
I
5.943
2.071
P
.0001
.0001
.0001
II
6.412
2.388
N
34
32
31
III
5.469
2.185
Note: r = correlation; p = p value; N = sample size
Cognitive-Affective Behavior Self-Report Test-Retest
Reliability
Table 4-18 illustrates significant test-retest
correlations for a series of true or false questions
focusing on cognitive-affective state and behavior.
However, comparison with the physical behavior data revealed
correlations tended to be higher and means less variable
when focusing on physical symptoms and behavior.
Post Hoc Analysis of Blood Urea Nitrogen
After completion of data collection and analysis it was
hypothesized that the inverse relationship between Hg and

101
Sum CVLT observed between the first and second assessments
might be due to an increase in severity of uremia as
measured by blood urea nitrogen (Bun). In the hemodialysis
population, blood values were necessary for the day of the
first and second assessments. However, Bun levels were
unavailable for the dates in question. In the peritoneal
dialysis group, despite a high proportion of subjects having
been tested at home, BUN values were available on or near
the dates of the first and second assessments for 8
subjects. Although the power of the test was unacceptably
low due to the small number of subjects, a univariate
analysis of variance was performed. This procedure failed
to demonstrate a relationship between Sum CVLT and BUN
(F=0.77, p=.4139). However, there was no demonstrable
relationship between Hg and Sum CVLT in this subsample of
subjects (F=0.01, p=.9356).
Summary
Despite extensive post hoc analyses, no evidence of
improvement in memory, self-reported cognitive-affective or
physical functioning was demonstrated in the rEPO treatment
group relative to controls. What appeared to be a trend
towards a decline in Sum CVLT performance in rEPO treated
subjects was found to represent poorer performance on the
assessment prior to a drop in Hg. This phenomenon was
observed primarily in a subset of the controls, but was not
evident until subjects were reclassified based on direction
of Hg change between the first and second assessments.

102
Reclassification based on direction of Hg change
revealed that Sum CVLT performance improved when Hg fell
between the first and second assessments. Post hoc analysis
revealed a main effect for direction of Hg change (F=16.97,
£=.0003) with an interaction between direction of Hg change
and self-reported Cognitive-Affective behavior (F=6.53,
p=.0043). Decline in cognitive-affective functioning was
associated with decline in list learning performance.
The changes in Hg were of dubious physiological
significance; indeed, the main hypothesis was not adequately
tested due to the small change in Hg. Hypothesizing that
increased uremia may have produced the poor performance,
post hoc data collection was initiated. Despite
insufficient sample size, analyses were performed to explore
the possibility of a trend; however, there was no evidence
of a relationship between Bun and Hg ((F=0.77, p=.4139).
The lack of what is considered to be physiologically
significant change in Hg led to the utilization of this data
set to assess test-retest and/or alternate form reliability
for several instruments. Consistent with the validation
study, the third form of the CVLT developed for this project
was found to be of essentially the same level of difficulty
as Forms I and II. Levels of correlation between CVLT III
and the other two forms were quite acceptable. In contrast,
the two alternate forms of the levels of processing task and
the two alternate forms of the frequency estimation task
exhibited relatively disappointing levels of correlation

103
with the original forms. However, the Physical Activity
scale on the self-report questionnaire developed for this
study exhibited excellent test-retest reliability. Test-
retest reliability was somewhat lower for the Cognitive-
Affective scale.

CHAPTER 5
DISCUSSION
Overview
A broad spectrum of cognitive abilities have been
reported to be deleteriously affected by chronic renal
disease. Despite considerable effort, the specific
etiologic mechanisms underlying the neuropsychological
deficits associated with end-stage renal disease have not
been satisfactorily explained. Anemia, often severe,
almost invariably accompanies chronic renal failure. One
purpose of this project was to test the hypothesis that
chronic anemia may contribute to the neuropsychological
deficits associated with end-stage renal disease. Chronic
anemia is one of the most common medical conditions in the
world, affecting diverse populations, particularly
including infants, women, the elderly and the malnourished
as well as specific disease populations. Despite the high
prevalence and longstanding recognition of the often
dramatic effects of acute anemia on cognition and brain
function, the neuropsychology of chronic anemia has been
largely ignored. Indeed, presumably because of the
difficulty in disentangling anemia from its etiology, the
question of whether chronic anemia has neuropsychological
consequences has remained unanswered. Elucidation of the
104

105
relationship between chronic anemia and cognition would be
potentially relevant in diverse areas of research, public
policy and clinical practice. Therefore, a second goal of
this study was to attempt to speak to the question of
whether chronic anemia affects neuropsychological
functioning.
The primary contribution to uremic anemia is the
failure of diseased kidneys to perform their normal role of
producing the erythropoiesis regulating hormone EPO.
Recently approved for clinical use, rEPO is capable of
effectively eliminating uremic anemia in the majority of
patients. Indeed, in most patients Hg may be maintained at
any desired level. Limitations on the target level of Hg
are a function of economics and severity of deleterious
side effects. Recombinant EPO is quite expensive.
Increasing hematocrit in association with rEPO treatment
has been associated with side effects including
hypertension. At present, there is insufficient data to
engage in cost-benefit and risk-benefit analysis.
Therefore, it was hoped that this study would generate
information regarding the relationship between level of Hg
and neuropsychological functioning to assist in risk-
benefit analysis.
Prior to the development of rEPO, since anemia was
nearly universal in chronic renal disease, any attempt at
clarification of the relative contributions of anemia and
uremia had been a most difficult task. Although several

106
studies controlled for Hg, possible confounds such as level
of physical exercise, degree of uremia, extent of uremic
anorexia and years on hemodialysis, as well as additional
sources of error such as the inability to use subjects as
their own controls, may have hidden the effect. The
development of rEPO, because it presumably alleviates
uremic anemia without directly affecting the brain,
suggested a method of disassociating anemia from the usual
confounds and presented the opportunity to use subjects as
their own controls. Thus, rEPO treatment was seen as a
powerful new method for studying the neuropsychology of
chronic anemia.
Several lines of evidence hinted at a possible
relationship between anemia and cognitive dysfunction.
Neuropsychological function has in some studies been
reported to be somewhat better in peritoneal dialysis than
in hemodialysis. Although this has been attributed to
greater middle molecule clearance in peritoneal dialysis,
the offending middle molecule has never been found, and Hg
levels tend to be higher in peritoneal dialysis. Also,
years on hemodialysis has been reported to be associated
with decreasing cognitive function and increasing anemia.
The dramatic reduction in anemia produced by treatment
with rEPO has been shown to produce clinically significant
improvements in domains such as aerobic capacity, anaerobic
threshold, sexual potency, employment status, level of
social activity, perceived ability to engage in physical

107
exercise and improved appetite, all previously attributed
to uremia. Despite predictions, compared to these
findings, pilot work with small numbers of rEPO treated
renal patients has been disappointing in that there have
been few improvements in neuropsychological functioning, no
consistently reported improvements and failures to
replicate. In our pilot work (Klein et al., 1989), the
only statistically significant change was in LTM as
measured by multitrial supra-span list learning. Despite
the danger of a Type I error posed by multiple analyses,
credence was placed in this finding because in the past
multitrial list learning had proven particularly sensitive
to treatment changes in renal disease. Therefore, the
present study sought to replicate and extend this finding.
Data analysis focused on change between the first and
second assessments because preliminary analyses revealed no
change in Hg between the second and third assessments.
Univariate analysis of variance of Sum CVLT change scores
failed to confirm either the hypothesized main effect for
Hg (F=3.25, p=0.0805) or a main effect for treatment status
(F=3.22, p=.0818). Combining treatment status and Hg level
in the same model eliminated the putative trend. After
discussing possible reasons for the discrepancy between the
pilot and the present study, post hoc analyses utilizing Hg
change as a classificatory variable will be discussed.
Examination of the literature on the impact of Hg
levels on cognitive function suggested that the Hg change

108
scores obtained in this study were not within the range
that would be expected to be physiologically significant.
Self-report data regarding Physical Activity had been
obtained so that, in the event of negative findings, a
determination could be made as to whether Hg levels had
been adequately manipulated to produce improvement in life
functioning. Consistent with the small change in Hg, there
were no between group differences in change scores for
self-reported Physical Activity or for Cognitive-Affective
behavior.
In comparison to the mean Hg change of 1.13 (Sd=1.90)
g/dl in the present study, mean treatment group Hg rose
5.46 (Sd=1.72) g/dl in the pilot study. Indeed, despite a
statistically significant change in Hg, (F=7.42, p=.0102),
the independent variable does not appear to have been
sufficiently manipulated; therefore, the main hypothesis
was not adequately tested. In addition, the absence of
change in Sum CVLT in the present study, in conjunction
with a smaller rise in Hg, suggests that Hg change may not
have been sufficient to provide clinical benefit.
Thee are other possible explanations for the
discrepancy between the pilot and the present study. Not
only was change greater, but Hg levels began lower in the
pilot study (M=6.10, Sd=.97 g/dl) than in the present study
(M=8.46, Sd=1.42 g/dl). The physiological effects of
varying degrees of severity of chronic anemia may be
nonlinear. This would suggest that one unit of Hg change

109
may be of greater physiological significance in more
severely anemic individuals. Conceivably, change in Hg
equivalent to that observed in the present study, but
occurring at a lower baseline level, might have resulted in
significant cognitive effects. The present study may have
failed to demonstrate cognitive change because baseline Hg
levels were above the threshold for cognitive dysfunction.
Indeed, based on the literature review, it is quite
conceivable that the nervous system is able to adapt to all
but severe anemia.
Despite plausible explanations for a failure to
replicate, re-examination of the pilot data in the light of
present findings suggests the possibility that the results
of the pilot study may have been spurious.
In the pilot study, four of the five treated subjects
had Hg increases of over 6 g/dl. In contrast, the greatest
Hg rise in the present study was 5.90 g/dl. However,
significance was reached in the pilot study only with the
contribution of the putative relationship between negative
LTM change scores in the control group and negative Hg
change in the same two subjects. The greatest fall in Hg
was 0.93 g/dl. The current study had 19 subjects with
changes in Hg of 1.0 g/dl or greater between assessments
one and two. Re-examination of the Pilot study data
revealed that mean change in the control group was -5.5
words recalled over five trials and 5.6 words in the
treatment group. Thus, nearly 50% of the change in the

110
dependent variable occurred in the control group. Although
change in the control group might ordinarily be attributed
to error variance, clouding the issue, both control
subjects exhibited small drops in Hg over the course of the
study, .93 g/dl and .55 g/dl, respectively. Therefore, the
drop in Sum CVLT in the control group was consistent with
the hypothesized relationship between Hg and LTM.
Interestingly, Hg changes of this magnitude were
nonsignificant in the present study. Furthermore, a rise
of 6.3 g/dl, an order of magnitude greater Hg change,
resulted in no LTM improvement in one of the five
experimental subjects.
The mean Sum CVLT change in the pilot study was 2.43
(Sd=7.63). Mean CVLT change in the present study was 1.69
(Sd=8.30) between the first and second assessments and
-0.06 (Sd=7.34) between the second and third assessments.
Variability was high and of similar magnitude in both
studies. In the pilot study, if the high correlation
between Hg and Sum CVLT change scores is to be accepted,
then a substantial portion of the variance in the Pilot
study was due to changes in Hg. Curiously, if one is to
accept Hg change as a primary source of variance, despite
smaller changes in Hg the present study, both studies
exhibited similar levels of variance in the dependent
measure. Excellent CVLT alternate form reliability makes
it unlikely that the utilization of alternate forms in the
present study introduced significant additional variance.

Ill
In summary, insufficient magnitude of Hg change, too
high an initial level of Hg in the present study or a
combination of both conditions may have resulted in the
discrepancy between the pilot and the present study.
However, closer examination of the pilot data in the
context of the present study suggested that the results of
the pilot could have been spurious. The present data set
provided insufficient information to determine which
hypothesis might account for the discrepancy between the
present and the pilot studies. Therefore, subjects were
reclassified to maximized between group differences in Hg
levels.
Observation of change in Hg levels in both groups
suggested the possibility of reclassification of subjects
based on Hg change scores to maximize between group
differences. The symptoms of acute anemia are quite
variable at a given Hg level and, if chronic anemia
followed a similar pattern, then, arguably, collapsing Hg
scores into a two-level classificatory system might
potentially reduce error variance. Unfortunately, only a
small number of subjects exhibited Hg change scores of
sufficient magnitude to justify classification based on
differences of demonstrated physiological significance.
Therefore, the first analysis simply reclassified based on
sign of Hg change.
Reclassification based on sign of Hg change doubled
the mean Hg change score and resulted in a statistically

112
significant main effect for sign of Hg change on Sum CVLT
(F=10.52, p=.0027). However, contrary to expectations, the
data indicated a negative relationship between rise in Hg
and Sum CVLT. Between the first and second assessments,
mean Sum CVLT change score fell 0.82 (Sd=5.38) in the
rising Hg group and improved 4.06 (Sd=9.91) in the falling
group. Physiological, psychometric and psychological
explanations for the apparent finding will be discussed.
The creation of a classificatory variable utilizing
direction of Hg change increased the magnitude and
significance of group differences in Hg change and
provided, arguably, an indirect means of further
characterizing the effects of rEPO on memory function.
This was accomplished both by including treatment status in
a multivariate model and by somewhat more adequately
testing the effects of the only known consequence of rEPO
therapy, increased erythropoiesis.
Recombinant EPO has never been specifically tested
clinically for deleterious neurobehavioral effects. The
drug may eventually be administered to several hundred
thousand people per year. Therefore, potential danger
signals, however improbable, deserve serious attention.
Several mechanisms may be hypothesized by which rEPO might
deleteriously affect cognitive functioning. They are
induction of iron deficiency, unknown hypothetical effects
of rEPO on the brain, cerebrovascular effects including
hypertension and/or increased blood clotting and reduced

113
sympathetic arousal secondary to decreased need for high
cardiac output. In addition, at high levels of Hg, the
efficiency of dialysis may be compromised resulting in
increased uremia.
Iron deficiency is a frequent consequence of rEPO
therapy in the absence of adequate iron supplementation or
assimilation. Iron deficiency has been shown to have
neurobehavioral consequences unrelated to heme synthesis.
Quite conceivably, stepped-up erythropoiesis could bind
enough nonheme iron into Hg to lower iron levels
sufficiently to deleteriously affect brain function.
Hypothetically, this could result in reduced cognitive
performance, and several weeks later, a rise in Hg.
However, in treatment subjects, vigorous iron
supplementation minimized changes in iron status;
therefore, very few subjects exhibited changes in iron
stores of potential clinical significance. Consistent with
this observation, a GLM univariate analysis unequivocally
failed to demonstrate a main effect for iron status
(F=0.01, p=0.9384).
The notion that rEPO may affect brain function through
mechanisms unrelated to erythropoiesis is contrary to
general assumptions. However, cancerous neural tissue is
capable of producing EPO (Hennessy, Stern, & Herrick,
1967) This capability suggests the possibility that EPO
may serve a function in the nervous system. If rEPO
negatively affected brain function through nonheme

114
mechanisms, cognitive deficits might be expected to appear
prior to rise in Hg. As previously mentioned, analysis
using a model containing either treatment status or Hg
change suggested a possible trend; however, a model
containing both terms failed to shed light on the "direct
effect of EPO on the brain" hypothesis. Following
reclassification of subjects based on direction of Hg
change, a multivariate analysis demonstrated treatment
status to be unambiguously noncontributory (F=0.59,
P=.4482), while direction of Hg change was significant
(F=9.45, p=.0043). Thus, the post hoc hypothesis that EPO
directly and deleteriously affects the brain through a
channel unrelated to Hg finds no support in this data.
Cognitive dysfunction might hypothetically be mediated
by cerebral vascular mechanisms. Recombinant EPO mediated
increases in Hg frequently produce undesirable effects
including headache and hypertension (Delano, 1989; Hori,
Onoyama, Iseki, Fujimi, & Fujishima, 1990). Increased
platelet aggregation may further increase risk of
cerebrovascular accidents. Although these possibilities
could not be tested within the confines of the present data
set, it may be noted that Hg rise was modest compared to
clinical trials where the effect of rEPO on clotting time
and blood pressure was studied.
In severe anemia, the need for greater cardiac output
results in increased sympathetic activation (Hori, Onoyama,
Iseki, Fujimi, & Fujishima, 1990). The effect, if any, of

115
this increased activation on cognitive function is unknown;
however, the curvilinear relationship between performance
and level of activation is well known. Since underarousal
has been suggested as a component of some aspects of
neuropsychological dysfunction in uremia, uremic subjects
may be especially vulnerable to reduced activation
secondary to a reduction in degree of anemia. However,
contrary to this line of reasoning, Hg changes were seldom
in the range and of the magnitude likely to affect degree
of activation.
The Controlled Word Association Test is affected by
degree of activation, at least in grossly neurologically
impaired populations. Superficially, word generation from
within phonemic categories appeared to follow a pattern
similar to that observed for Sum CVLT in that treatment
group performance was relatively uniform across trials and
control group performance was lower at the first
assessment. However, controlling for multiple tests, this
finding failed to reach statistical significance. More
important, unlike Sum CVLT, this variable no longer
exhibited the above mentioned pattern when sign of Hg
change was used for classificatory purposes. In addition,
the self-report questionnaire, although not designed for
this purpose, failed to suggest alteration in degree of
activation.
Examination of the pattern of mean Sum CVLT change
scores led to the suggestion that the apparent decline was

116
not due to a decrease in treatment group performance, but
rather, to improved performance between the first and
second assessments in subjects exhibiting a fall in Hg.
The majority of these subjects were in the control group.
Mean Sum CVLT in the treatment group was 45.8, 45.0 and
45.1 for assessments one through three respectively. With
a standard deviation varying between 11.6 and 13.4, these
scores were remarkably similar. In contrast, in the
control group, mean Sum CVLT was 43.2 at the first testing
but 47.3 and 47.1 at assessments two and three,
respectively. With a standard deviation varying between
10.4 and 10.7, mean performance at the first assessment was
nearly .4 of a standard deviation below expectations.
Hypothesizing that increased Hg may have reduced the
efficiency of dialysis resulting in increased uremia;
therefore, produced poorer cognitive performance secondary
to uremia, post hoc collection of BUN and creatinine levels
was attempted. However, few subjects were consistently
assessed on the same days as laboratory tests were
performed. Therefore, the small size created the strong
possibility, if not probability, of a Type II error.
Consistent with this, analysis failed to suggest a
relationship between BUN and Sum CVLT.
It might be argued that the control group, rather than
performing more poorly at the first assessment, did better
at the second assessment. That is, the control group
exhibited a practice effect while the treatment group,

117
starting at a higher level, failed to benefit from the
experience. However, a simple practice effect would not be
expected as test forms contained different words and
categories. Learning how to learn list learning from
minimal exposure to a list learning task and generalization
to subsequent list learning weeks later would be quite
extraordinary. Also, neither group showed an improvement
between the second and third assessments. It further
strains credulity to consider that maximal learning
occurred following a single exposure to a list learning
task. In addition, the practice effect hypothesis requires
the dubious assumption that prior to exposure to the list
learning task, the control group was poorer at list
learning than was the treatment group, but benefited more
from practice. Based on subject composition the control
group would be expected to perform slightly better; indeed,
they did on the second and third assessments. The
hypothesis of differential response to practice must be
rejected.
Form III of the CVLT, the version developed for this
study, was administered second to all subjects. If Sum
CVLT III were easier than Form I and Form II, then this
might explain the appearance of improved performance in the
control group at the second assessment and suggest
decreased performance in the treatment group. One of the
benefits of the present design is that it directly controls
for possible intertest form differences through the

118
creation of difference scores. However, absence of
randomization eliminated the ability to statistically test
for interactions between test instrument and group or order
effects.
Although CVLT III might theoretically be somewhat
easier based on the word frequency rates of the target
words, as described in Appendix A, administration of Forms
I and III to 18 normal, native English speaking volunteers,
primarily university undergraduates, revealed identical Sum
CVLT mean scores for both groups. Moreover, scores were
remarkably similar between Forms I, II and III in the
treatment group. In addition, scores between Forms II and
III were essentially identical in the controls. In the
treatment group, the correlation was an impressive .92
between Sum CVLT Forms II and III and .81 between Forms I
and III. In contrast, the correlation between Forms I and
II, tests arguably considered equivalent, was .77 for Sum
CVLT. For comparison, in the validation of Form II, the
correlation between Sum CVLT on Forms I and II was .84.
Taken together, these results suggest that Sum CVLT III is
equivalent to the two previous forms. Differences in test
instruments appear unable to account for the observed
pattern of scores.
Psychological explanations may be posited; however,
none elegantly fit the data. Nevertheless, it should be
noted that subject expectations were poorly controlled.
Most hemodialysis patients spent three half-days per week

119
in a milieu in which information and misinformation was
routinely shared. Most peritoneal dialysis patients were
encouraged to participate by their physician. Hemodialysis
patients were approached directly by the experimenter.
This difference may have inadvertently affected patient
performance at the first assessment.
Another source of error that may have contributed to
the pattern of results is that the majority of subjects
knew if they were receiving rEPO. Indeed, at higher
dosages of rEPO subjects experience such dramatic effects
that their status is obvious to all concerned. However, in
this study both experimenter and subject were blind to Hg
level.
In a multivariate model There was a main effect for
direction of Hg change (F=16.97, p=.0003) and an
interaction between direction of Hg change and self-
reported Cognitive-Affective behavior (F=6.53, p=.0043).
Improvement in self-reported Cognitive-Affective behavior
was associated with improved performance on Sum CVLT. As
the nature of the main effect is unexplained no attempt
will be made to explain the interaction.
The levels of processing and estimation of frequency
paradigms were intended to probe the mechanisms underlying
any observed improvement in memory function. Analyses
failed to suggest that there were changes in either
automatic memory or ability to benefit from semantic
processing.

120
Unfortunately, several interrelated factors weakened
the design of the present study. Initial plans called for
integration of this project into a larger study examining
the clinical efficacy and safety of rEPO prior to U.S. Food
and Drug Administration approval. However, approval of
rEPO for clinical use was unexpectedly expedited. This
resulted in reduced control over the independent variables.
Hemoglobin change scores were much smaller than in the
pilot study. Indeed, follow-up revealed that over one year
after initiation of rEPO, the majority of subjects were
still moderately anemic. Because the experimenter was
blind to Hg levels until completion of neuropsychological
testing, the existence of only minimal Hg change scores was
undetected until completion of the study.
In the Pilot study, data collection was coordinated
with the timetable of a multicenter clinical trial.
Baseline assessment occurred prior to any EPO mediated rise
in Hg. Approval for clinical use resulted in the majority
of hemodialysis subjects being placed on rEPO, many prior
to baseline assessment. A reduced pool of rEPO-free
hemodialysis subjects necessitated the inclusion of
peritoneal dialysis subjects as controls. Although
undesirable, the impact of a nonequivalent control group
was minimized by the design. The controls merely served to
control for the effects of multiple testing with different
test forms. The approval of rEPO raised the possibility
that suitable rEPO-free long-term dialysis subjects would

121
become rare. Therefore, the study was conducted despite
imperfect conditions.
The paucity of changes between assessments provided
the opportunity to examine the alternate form reliability
of the test forms developed for this study. These tests
were the third form of the CVLT as well as the second and
third forms of the levels of processing and estimation of
frequency of occurrence tasks.x The CVLT III performed very
well and appeared to be essentially identical to the two
earlier forms. In contrast, the levels of processing and
estimation of frequency of occurrence alternate forms were
disappointing. The statistically significant correlation
between Cognitive-Affective self-report change scores and
Sum CVLT change scores, especially in the absence of a
significant relationship between Physical Activity and Sum
CVLT, supported the validity of behaviorally oriented
questionnaire data in the study of this population. This
is significant in that previous work with the Profile of
Mood States has been less than supportive of the validity
of self-report data in this population.
Conclusion
Many of the methodological deficiencies noted in the
literature were taken into account in the design of this
study. Thus, this study made use of a longitudinal design,
alternate test forms, a dialysis control group, sufficient
sample size, controlled for iron status, maintained a
consistent time of assessment with respect to dialysis and

122
excluded patients with uncontrolled hypertension.
Nevertheless, flaws severely limit the extent to which
useful conclusions may be drawn.
In the current study, Hg levels were both higher at
baseline and rose much less than in the pilot study. There
was significant intersubject variability in degree of
anemia and in etiology of renal disease. Despite
utilization of dialysis controls, there were big between
group differences in dialysis treatment modality. Although
the experimenter was blind to Hg level, rEPO treatment
status was usually known by subject and experimenter.
The present study produced absolutely no evidence that
rEPO treatment is associated with improvement in
neuropsychological functioning. Given the positive
correlation between rEPO mediated rise in Hg and
improvement in LTM in the Pilot study, the absence of
improvement in response to smaller increases in Hg might be
interpreted as indicating that most patients would benefit
from higher dosages of rEPO. However, the present results
provided data useful in a re-examination of the findings of
the Pilot study. The results of this re-examination
suggested the possibility that the initial findings may
have been spurious. More work is needed with Hg changes of
greater magnitude and lower initial values to speak to the
question of the neuropsychology of anemia.
Despite the absence of Hg change at levels of
demonstrated physiological significance, a possible trend

123
toward an inverse relationship between Hg change and Sum
CVLT was noted, and confirmed statistically by post hoc
analyses. Curiously, Sum CVLT performance was temporarily
below expectations in subjects who would later exhibit a
drop in Hg. Stated differently, between the first and
second assessments, Sum CVLT rose when Hg fell. After
exploring a variety of hypotheses, it was concluded that
the cause is unknown at this time.
Future studies of the impact of rEPO on nervous system
function in the anemic patient need address a number of
problems discussed in these pages. Many of the more
serious flaws in the present study were a function of the
release of rEPO for clinical use. Given that under current
medical protocols it may be that the severely anemic
chronic renal patient may become rare, future studies
focusing on the neuropsychology of chronic anemia may have
to work with other populations. Nevertheless, despite
increasing difficulties in working with this population,
optimal care of the renal patient demands well controlled
studies of the effect of rEPO mediated changes in anemia on
neuropsychological function.

APPENDIX A
DEVELOPMENT AND VALIDATION OF FORM III OF THE CALIFORNIA
VERBAL LEARNING TEST
Introduction
In the pilot study, repeat memory testing with Form I
of the CVLT revealed that some adult hemodialysis subjects
recalled the word lists after several months. Therefore,
alternate forms were considered desirable in the present
study. Form II of the CVLT was recently developed by
Delis, McKee, Massman, Kramer and Gettman (1990). The
construction and validation of a third parallel form will
be described below.
Test Construction
A third form of the CVLT was created for use in this
study. Following the procedures used in the development of
Form I of the CVLT, three criteria were used in selecting
Form III target words. Two of these criteria pertain to
frequency of use of the words and the third category
attempts to produce category members of equivalent
typicality (Delis, Kramer, Kaplan, & Ober, 1987).
The CVLT consists of two lists of 16 words that are
balanced for frequency of appearance in English reading
(Thorndike & Lorge, 1944; Carroll, Davies, & Richman,
1971). Both lists are composed of words belonging to four
124

125
categories with two of the categories shared by both lists.
Words are balanced with regard to the rank order of each
word as an exemplar of the category as measured by Battig
and Montague (1969), Hunt and Hodge (1971) and Uyeda and
Mandler (1980). The primary and interference lists of each
form contain two categories in common and two unrelated
categories.
TABLE A-l. Mean Word Frequency and Typicality Across Forms
FORM I
a b
word word
freer freer
c
cat
rank
FORM II
a b
word word
freer freo
C
cat
rank
FORM III
a be
word word cat
freg freg rank
Primarv list
M 10.81 41.66
13.31
12.2
43.6
-
15.7
43.8
8.1
Sd 9.91 8.65
6.73
8.2
10.5
-
15.9
5.5
4.3
Interference list
M 14.13 43.56
14.94
7.7
42.8
-
16.7
45.4
12.3
Sd 13.9 9.83
9.28
7.8
9.5
-
19.5
4.8
8.8
Note: a word freq = word frequency
(Thorndike &
Lorge,
1944); b word freq = word frequency (Carroll, Davies,
& Richman, 1971) ; c cat rank = typicality ranking
(Battig & Montague, 1969); M = mean; Sd = standard
deviation
Table A-l compares the word frequency characteristics
of Forms I, II and III with regard to frequency of
appearance in English reading as measured by Thorndike and

126
TABLE A-2. Form III Word Frequency and Typicality Values
Primary List
a b c
Interference
a b
List
C
Dresser
7
45.5
8
Rocker
2
44.1
36
Rug
40
51.5
12
Bureau
44
45.3
11
Couch
28
47.1
7
Stool
16
49.2
10
Bookcase
3
40.5
14
Cabinet
26
48.6
15
Orchids
3
43.8
6
Carnation
1
36.6
3
Lilies
33
45.3
8
Pansies
7
42.5
9
Gardenias
1
29.6
11
Daisies
28
44.1
4
Daffodils
3
40.7
12
Geraniums
4
39.6
16
Sapphire
6
39.4
4
Liver
10
48.7
-
Pearl
47
47.7
5
Bacon
12
50.7
-
Garnet
2
37.2
11
Ham
17
49.8
-
Jade
4
42.8
7
Steak
14
46.2
-
Socks
12
50.3
4
Silk
75
55.1
3
Slippers
20
47.5
5
Flannel
9
44.2
15
Sandals
5
44.8
3
Burlap
1
41.1
12
Boots
37
47.5
2
Denim
1
41.1
13
Mean
15.
7
43.8
8.1
16.7
45.4
12.
sd
15.
9
5.5
4.3
19.5
4.8
8.
Note: word frequency (Thorndike & Lorge, 1944)
b = word frequency (Carroll, Davies & Richman, 1971)
c cat rank = typicality ranking (Battig & Montague,
1969); M = mean, sd = standard deviation

127
Lorge (1944) and Carroll, Davies and Richman (1971) as well
as the rank order of each word as an exemplar of its
category (Battig and Montague, 1969). Table A-2 lists Form
III words and their frequency values.
Methods
Eighteen normal, native English speaking volunteers
were administered Forms I and III of the CVLT. Seventeen
subjects were between 18 and 23 years of age. One subject
was 48 years old. Mean age was 21.9 with a standard
deviation of 6.7 years. There were 11 female and 7 male
subjects. Mean years of college education was 14.5
(SD=1.2). Sixteen of the 18 subjects had at least some
college education.
The majority of testing was performed by an
undergraduate senior psychology major in his last semester
of study, who had received extensive training prior to data
collection. All subjects were administered Forms I and III
of the CVLT. Every other subject began with Form I. Mean
interval between testing sessions was 5 days. Minimum time
between sessions was 48 hours.
All assessments were performed individually. During
the 20 minute delay between immediate and delayed recall
subjects either conversed with the experimenter or
performed activities in which they had been engaged prior
to the assessment (e.g., studying). Administration of
other measures during the delay was rejected to reduce the
difficulties associated with obtaining volunteer subjects.

128
Results
The variable of greatest interest was Sum CVLT. Sum
CVLT is likely to be the most stable measure produced by
the CVLT. As displayed in Table A-3, Forms I and III
produced identical means of 65.1 and a correlation
coefficient of r = .69. Trial 1, a measure of STM,
produced
means of 8.7
and
9.7
for
Forms I
and III
TABLE
A-3
. CVLT
Inter
-Form Reliability
CVLT
III
CVLT I
CVLT I &
Variable
r
Mean
sd
Mean
sd
CVLT Hr
SUM
.69b
65.1
6.2
65.1
4.9
. 84c
T1
.70c
9.7
2.2
8.7
1.7
.54c
T2
.28
12.5
1.9
12.6
1.3
T3
.63b
13.3
1.8
13.8
1.4
T4
.31
14.3
1.2
15.0
1.1
T5
.10
15.1
1.0
15.2
1.2
.75c
List
B
. 56a
8.9
1.9
9.0
1.9
.31a
SD
.63b
14.0
1.6
14.7
1.0
. 82c
Cued
.51a
14.3
1.4
15.2
0.8
.76c
LongDel
.32
14.1
1.3
15.1
1.0
. 80c
LCued
.14
14.9
0.8
15.3
0.7
.79c
Note:
a p<.05
b p<.
01
c
p<.
001
SUM
= Sum
CVLT,
List B =
Interference list;
SD
= short
delay
free recall
.; CUED
= short
delay cued
recall, LONGDEL = long delay free recall;
LCUED = long delay cued recall

129
respectively. The correlation between forms for Trial 1
was r = .70. The mean interference trial scores of 9.0 and
8.9 for Forms I and III, respectively, were nearly
identical and produced a correlation of r = .56. It is
desirable that the interference trial be of equivalent
difficulty to the primary list as measured by Trial 1 of
the primary list. The CVLT Form III primary list appeared
slightly but not significantly easier. The inter-form
correlation for Trial 5 was r= .50; however, means were
15.2 and 15.1. Please see Table A-3 for a tabular listing
of the results.
In Table A-3, the correlations between Forms I and II
were reported by Delis, McKee, Massman, Kramer and Gettman
(1990).
Discussion
A comparison of mean scores between the two forms
suggests that the two versions are quite similar. Indeed,
this is illustrated by summing the 10 primary measures.
Mean word recall for Form I was 134.6 and 131.1 for Form
III. As this is out of a total of 160 words the 3.5 word
recall advantage for Form I is rather insignificant. The
correlations for Sum of Trials 1 through 5 (.69) and Trial
1 (.70) compare favorably with the alternate form
reliability coefficients reported for the Rey Auditory
Verbal Learning Test, which ranged from .60 to .77 and with
Russell's revised Wechsler Memory Scale, which had
correlations ranging from .60 to .74 (Ryan, Geisser,

130
Randall, & Georgemiller, 1986; McCarty, Logue, Power,
Ziesat & Rosenstiel, 1980).
Sum of Trials 1-5 and Trial 1 both worked well and
they are measures of special interest in the study for
which Form III was designed. In contrast, many of the
other correlations are rather low, especially Trial 5 (r =
.10) and Long Delay Cued Recall (r = .14). Also, in every
instance except for Trial 1, correlation coefficients
between Forms I and II are higher, sometimes markedly
higher. However, it is likely that the low correlations
for some measures are part of a general suppression of all
correlation coefficients due to the homogeneity and very
high functioning of the subject pool. Indeed, recognition
measures could not be calculated because nearly every
subject had a perfect score. Likewise, a mean of 15.1 for
the fifth trial on Form III clearly illustrates the problem
of a ceiling effect. Higher alternate form correlation
coefficients might be expected with a more representative
cross-section of the population.
Construction of this version of CVLT Form III is
deficient in that some target words have prototypicality
ratings (Battig and Montague, 1969) of 4 or lower. This
theoretically increases the possibility that correct
responses may occur by chance in subjects prone to
semantically related intrusions. However, the present data
suggests CVLT Form III to be an excellent alternate form.

APPENDIX B
LEVELS OF PROCESSING VERSIONS II AND III
Table B-l. Levels of Processing Version II Encoding
Questions
Visual Stimuli
Ouestion to be processed
1.
ALTAR
Is this printed in upper case
letters?
2.
ear
Is this part of the body?
3.
sand
Does this rhyme with match?
4.
swine
Is this a type of furniture?
5.
smell
Does this rhyme with sell?
6.
cold
Is this printed in upper case
letters?
7.
sky
Is this printed in upper case
letters?
8.
country
Is this a type of music?
9.
priest
Does this rhyme with fall?
10.
joy
Does this rhyme with boy?
11.
wolf
Is this a type of flower?
12.
BURN
Is this printed in upper case
letters?
13.
silver
Is this a type of metal?
14.
seat
Is this a type of animal?
15.
trot
Does this rhyme with pot?
16.
shoot
Is this printed in upper case
letters?
17.
rope
Does this rhyme with hour?
18.
PEN
Is this printed in upper case
letters?
19.
CHEW
Is this printed in upper case
letters?
131

132
Table B-l continued
Visual Stimuli
Ouestion to be processed
20.
prison
Does this rhyme with board?
21.
hatchet
Is this a type of tool?
22.
lie
Does this rhyme with cry?
23.
dandelion
Is this a type of flower?
24.
tube
Is this printed in upper case
letters?
25.
grass
Is this printed in upper case
letters?
26.
pear
Is this a type of fruit?
27.
swim
Does this rhyme with rug?
28.
pad
Does this rhyme with fad?
29.
trumpet
Is this a type of sport?
30.
SURPRISE
Is this printed in upper case
letters?
31.
rain
Is this printed in upper case
letters?
32.
tie
Is this a type of furniture?
33.
knot
Does this rhyme with cab?
34.
SAFE
Is this printed in upper case
letters?
35.
sheep
Is this a type of animal?
36.
hire
Does this rhyme with wire?

133
Table
B-2. Multiple Choice
Retrieval
Ouestions
Version II
Multiple Choice
Word List
Processing
Level
1.
seat
coin
maple
nail
SEM-
2.
clip
wood
chew
fern
0RTH+
3.
globe
sand
clamp
scar
PHON-
4.
rug
haze
gift
hire
PH0N+
5.
grass
foot
mast
sheet
ORTH-
6.
tear
country
page
horse
SEM+
7.
joy
cream
oak
fear
PH0N+
8.
path
coal
tie
tent
SEM-
9.
dandelion
bat
ghost
lap
SEM+
10.
moss
altar
bed
cart
0RTH+
11.
trout
egg
doll
swim
PHON-
12.
cable
paste
shoot
cup
ORTH-
13.
band
cord
safe
tune
ORTH+
14.
fur
ear
twig
mate
SEM+
15.
roll
chin
pork
lie
PH0N+
16.
trumpet
key
steak
purple
SEM-
17.
plot
sky
weed
cage
ORTH-
18.
crumb
limb
rope
ash
PHON-
19.
wheel
hide
spade
sheep
SEM+
20.
banana
leaf
pen
nob
ORTH+
21.
tube
roam
spear
art
ORTH-
22.
bullet
smell
cat
oar
PH0N+
23.
bone
pack
rock
wolf
SEM-
24.
prison
corn
book
mire
PHON-
25.
juice
string
silver
acre
SEM+

134
Table B-2 continued
Processing
Multiple Choice
Word List
Level
26.
silk
hatchet
fog
brass
SEM-
27.
dog
lamp
wave
knot
PHON-
28.
cold
fad
gear
tan
ORTH-
29.
surprise
cop
prey
camp
0RTH+
30.
soap
pond
trot
chart
PH0N+
31.
float
priest
film
ant
PHON-
32.
paint
hinge
mute
burn
0RTH+
33.
swine
ditch
surf
ache
SEM-
34.
horse
seem
ramp
rain
ORTH-
35.
floor
hen
stamp
pad
PH0N+
36.
spark
pear
grip
site
SEM+
Note:
+ = Correct
response
is positive;
- = Correct
response
is negative;
ORTHO =
Orthographic; PHON = Phonetic; SEM = Semantic

135
Table B-3. Levels of Processing Version III Encoding
Questions
Visual Stimuli
Ouestion to be orocessed
1.
COLLEGE
Is this printed in upper case
letters?
2.
finger
Is this part of the body?
3.
garden
Does this rhyme with match?
4.
hamster
Is this a type of furniture?
5.
clock
Does this rhyme with sock?
6.
ship
Is this printed in upper case
letters?
7.
animal
Is this printed in upper case
letters?
8.
violin
Is this a type of musical instrument?
9.
robot
Does this rhyme with fall?
10.
palace
Does this rhyme with chalice?
11.
canoe
Is this a type of flower?
12.
LAKE
Is this printed in upper case
letters?
13.
salt
Is this a seasoning?
14.
hill
Is this a type of animal?
15.
spike
Does this rhyme with bike?
16.
star
Is this printed in upper case
letters?
17.
mile
Does this rhyme with hour?
18.
KNIFE
Is this printed in upper case
letters?
19.
TOWEL
Is this printed in upper case
letters?
20.
dime
Does this rhyme with board?
21.
shovel
Is this a type of tool?
22.
bed
Does this rhyme with head?
23.
strawberry
Is this a type of fruit?
24.
deck
Is this printed in upper case
letters?
25.
pencil
Is this printed in upper case
letters?

136
Table B-3 continued
Visual Stimuli
Ouestion to be processed
26.
hotel
Is this a type of building?
27.
plate
Does this rhyme with rug?
28.
net
Does this rhyme with pet?
29.
peel
Is this a type of sport?
30.
TRUCK
Is this printed in upper case
letters?
31.
dog
Is this printed in upper case
letters?
32.
arm
Is this a type of furniture?
33.
ink
Does this rhyme with cab?
34.
SHORE
Is this printed in upper case
letters?
35.
white
Is this a type of color?
36.
bath
Does this rhyme with math?

137
Table
B-4. Multiple Choice
Retrieval
Ouestions
Version III
Multiple Choice
Word List
Processing
Level
1.
hill
coin
grass
nail
SEM-
2.
clip
wood
TOWEL
fern
0RTH+
3.
globe
garden
clamp
scar
PHON-
4.
rug
haze
gift
bath
PH0N+
5.
pencil
foot
mast
sheet
ORTH-
6.
tear
violin
page
horse
SEM+
7.
palace
cream
oak
fear
PH0N+
8.
path
coal
arm
tent
SEM-
9.
strawberry
bat
ghost
lap
SEM+
10.
moss
college
bed
cart
ORTH+
11.
trout
egg
doll
plate
PHON-
12.
string
paste
star
cup
ORTH-
13.
band
cord
shore
tune
ORTH+
14.
fur
finger
twig
mate
SEM+
15.
roll
chin
pork
bed
PHON+
16.
peel
key
steak
purple
SEM-
17.
plot
animal
weed
cage
ORTH-
18.
crumb
limb
mile
ash
PHON-
19.
wheel
lie
spade
white
SEM+
20.
peddle
leaf
knife
nob
ORTH+
21.
deck
roam
spear
art
ORTH-
22.
bullet
clock
cat
oar
PH0N+
23.
bone
pack
rock
canoe
SEM-
24.
dime
grass
book
mire
PHON-
25.
juice
rope
salt
acre
SEM+

Table B-4 continued
138
Processing
Multiole Choice
Word List
Level
26.
silk
shovel
fog
brass
SEM-
27.
dog
lamp
wave
ink
PHON-
28.
ship
fad
gear
tan
ORTH-
29.
truck
cop
prey
camp
0RTH+
30.
soap
pen
spike
chart
PH0N+
31.
float
robot
film
ant
PHON-
32.
paint
hinge
mute
lake
0RTH+
33.
hamster
ditch
surf
ache
SEM-
34.
horse
seem
ramp
dog
ORTH-
35.
floor
hen
stamp
net
PH0N+
36.
spark
hotel
grip
site
SEM+
Note:
+ = Correct
response
is positive;
- = Correct
response
is negative;
ORTHO =
Orthographic; PHON = Phonetic; SEM = Semantic

APPENDIX C
FREQUENCY ESTIMATION TASK
Table C-l. Encoding Phase
"Now I am going to read a list of [animals, grain & flour
products or means of transportation]. Please remember as
many
of the words as
possible because a
little
later on I
will
ask you to list
as many of them as
you can
II

Assessment 1
ANIMALS
Assessment 2
GRAIN & FLOUR
PRODUCTS
Assessment 3
MEANS OF
TRANSPORTATION
1
kangaroo
1
wheat
1
car
2
dog
2
oatmeal
2
subway
3
snake
3
barley
3
bicycle
4
elephant
4
cake
4
escalator
5
kangaroo
5
barley
5
subway
6
buffalo
6
corn
6
airplane
7
goldfish
7
waffle
7
canoe
8
kangaroo
8
barley
8
subway
9
tiger
9
pancake
9
stagecoach
10
dog
10
oatmeal
10
car
11
moose
11
rice
11
skateboard
12
kangaroo
12
barley
12
subway
13
gopher
13
bagel
13
helicopter
14
kangaroo
14
barley
14
subway
15
dog
15
oatmeal
15
car
139

140
Table C-2. Free Recall Instructions
"Remember the list of [animals, grains & flour products or
means of transportation] that I read earlier. Please tell
me all the words from that list that you can remember, in
any order."
Table C-3. Frequency Estimation
"When I read you that list some of the words were said more
than once. I will now read you words from that list. When
I say a word tell how many times you think it appeared the
first time I read it.
Guess when
not
sure."
Assessment 1
ANIMALS
Assessment 2
GRAIN & FLOUR
PRODUCTS
Assessment 3
MEANS OF
TRANSPORTATION
1
kangaroo
cake
escalator
2
snake
oatmeal
canoe
3
dog
wheat
bicycle
4
elephant
rice
stagecoach
5
buffalo
corn
airplane
6
goldfish
waffle
car
7
tiger
pancake
subway
8
moose
barley
skateboard
9
gopher
bagel
helicopter

APPENDIX D
COGNITIVE-AFFECTIVE AND PHYSICAL BEHAVIOR QUESTIONNAIRE
Table D-l. Questionnaire
Circle only the sentences true over the past two days.
1. I don't mind climbing stairs.
2. Walking quickly tires me out.
3. I sit during much of the day.
4. I take one or more naps during the day.
5. I feel useless, like I am a burden on others.
6. I easily get out of breath.
7. I have difficulty concentrating.
8. My sleep is sounder than it used to be.
9. I am irritable and impatient with myself.
10. I do work around the house only for short periods.
11. I stay away from home only for brief periods of time.
12. My sex drive is stronger than it was one year ago.
13. I am not doing any daily work around the house.
14. I feel terrible when I get off of hemodialysis.
15. I feel tired most of the time.
16. I do not finish things I start.
17. I react slowly to things that are said or done.
18. I work a job outside the home.
19. I spend much of the day lying down in order to rest.
20. I enjoy eating.
141

142
Table D-l continued
21. My future looks bright.
22. I feel tired all the time.
23. Everything is an effort.
24. I enjoy visiting friends.
25. I do heavy work around the house.
26. I do alot of cooking.
27. Simple chores quickly tire me out.
28. My muscles quickly get tired.
Table D-2. Instructions for Scoring: Corrections
If 13 is endorsed then consider 10 endorsed.
If 19 is endorsed then consider 3 endorsed.
If 22 is endorsed then consider 15 endorsed.
Table D-3. Cognitive-Affective Scale Scoring Instructions
Starting with 9 points, subtract one point for each of the
following if endorsed:
5. I feel useless, like I am a burden on others.
7. I have difficulty concentrating.
9. I am irritable and impatient with myself.
16. I do not finish things I start.
17. I react slowly to things that are said or done.
Subtract one point for each that is not endorsed:
8. My sleep is sounder than it used to be.
20. I enjoy eating.
21. My future looks bright.
24. I enjoy visiting friends.

143
Table D-4. Physical Scale Scoring Instructions
Start with 18 points and subtract one point for each
endorsed.
2. Walking quickly tires me out.
3. I sit during much of the day.
4. I take one or more naps during the day.
6. I easily get out of breath.
10. I do work around the house only for short periods.
11. I stay away from home only for brief periods of time.
13. I am not doing any daily work around the house.
15. I feel tired most of the time.
19. I spend much of the day lying down in order to rest.
22. I feel tired all the time.
23. Everything is an effort.
27. Simple chores quickly tire me out.
28. My muscles quickly get tired.
Subtract one point for any of the following not endorsed:
1. I don't mind climbing stairs.
12. My sex drive is stronger than it was one year ago.
18. I work a job outside the home.
25. I do heavy work around the house.
26. I do alot of cooking.

APPENDIX E
COMPARISON OF PARAMETRIC AND NONPARAMETRIC TESTS OF BASELINE
MEASURES BETWEEN GROUPS
Kruskal-Wallis Test
General Linear
(Chi-Square Approximation)
Nonparametric
Model
Parametric
Df=l Prob >
Variable Chi-Square Chi-Square F Value Pr >F
Hemoglobin
3.0616
.0802
2.94
.0959
Hematocrit
3.1778
.0746
2.29
. 1401
Ferritin
0.3478
.5553
0.00
.9483
Age
1.1868
.2760
1.13
.2948
Education
0.06695
.7958
0.23
.6325
Dialysis years
2.8425
. 0918
3.51
.0699
California Verbal
Learning
Test
Trial 1
0.0045
.9464
0.05
.8253
Standard Score
0.0855
.7700
0.00
.9602
Trial 2
0.1244
.7243
0.50
.4847
Trial 3
0.2496
.6173
0.21
.6535
Trial 4
1.3349
.2479
0.68
.4140
Trial 5
0.4435
.5054
0.40
.5291
Standard Score
0.0045
.9464
0.11
.7400
Sum CVLT
0.3157
.5742
0.42
.5222
Standard Score
0.0221
.8817
0.03
.8662
Clusters
0.0175
.8947
0.36
.5519
Perseverations
0.0805
.7765
0.38
.5442
Intrusions
0.1496
.6989
0.01
.9409
List B
0.0914
.7623
0.09
.7687
Standard Score
0.0012
.9724
0.10
.7587
Short Delay
0.0002
.9867
0.34
.5624
Standard Score
0.0002
.9864
0.02
.8951
Clusters
0.2849
.5935
0.00
.9760
Perseverations
2.7324
.0983
4.67
.0380*
Intrusions
0.2658
.6061
0.14
.7086
Short Delay Cued
0.1107
.7393
0.00
.9973
Standard Score
0.0141
.9052
0.01
.9053
Perseverations
3.3690
.0664
2.62
.1151
Intrusions
0.4843
.4865
0.04
.8343
144

145
Kruskal-Wallis Test
General Linear
(Chi-Square Approximation) Model
Nonparametric Parametric
Df=1
Prob >
Variable <
Shi-Sauare
Chi-Scruare
F Value
Pr >F
Long Delay
0.1863
.6660
0.15
.6995
Standard Score
0.6110
.4344
1.27
.2671
Clusters
0.7180
.3968
0.33
.5670
Perseverations
0.8303
.3622
0.31
.5838
Intrusions
4.8724
.0273*
2.48
.1245
Long Delay Cued
0.3420
.5587
0.04
.8425
Standard Score
0.7505
.3863
0.84
.3669
Perseverations
-
-
-
-
Intrusions
1.5373
.2150
2.05
.1613
CVLT Recognition
Variables
Recognition Hits
0.0862
.7691
0.75
.3931
False Positives
From Interference List
Shared Category
1.0593
.3034
1.12
.2981
Unrelated
0.0912
.7627
0.02
.8869
Intrusions not
on Interference List
Shared Semantic
1.2254
.2683
2.19
.1482
Shared Phonemic
0.8235
.3641
1.71
.1996
No Relation
0.6879
.4069
0.71
.4045
Controlled Word Association
CFP
0.7183
.3967
0.49
.4880
FAS
1.2263
.2681
1.41
.2445
LSW
5.1345
.0235*
4.62
.0394*
Levels of Processing
Semantic
1.7209
.1896
0.93
.3424
Orthographic
0.1749
.6757
0.16
.6887
Phonemic
0.0819
.7747
0.21
.6477
Estimation of Frequency of '
Occurrence
Free Recall
0.0188
.8907
0.04
.8390
Estimation
0.2524
.6153
0.09
.7655
Cognitive-Affective and Physical Behavior
Questionnaire
Cognitive-Affect
0.1360
.7123
0.11
.7459
Physical
0.1104
.7396
0.19
.6676
Note: Df = degrees of freedom
* = discrepant results between parametric and
nonparametric tests

APPENDIX F
NORMALITY AND HOMOGENEITY OF VARIANCE
Table F-l. Indices of Normality and Characteristics of the
Distributions
Shapiro-Wilk Test
Variable
W:Normal
Prob Sd/M
Skew Kurtosis
California Verbal
Trial One
Learning
0.956
Test
.010
0.343
.084
-.240
Trial Two
0.959
.016
0.293
.454
.213
Trial Three
0.969
. 120
-
-.189
-.261
Trial Four
0.955
.008
0.282
-.496
.051
Trial Five
0.953
.004
0.272
-.354
-.310
Sum CVLT
0.976
.330

-.160
-.337
Sum Clusters
0.850
.001
0.746
1.631
3.003
Slope
0.994
.997
-
-.141
1.192
Intercept
0.987
.861

-.006
.083
Interference List
0.946
.001
0.372
. 162
-.459
Short Delay
29.520
.0001
0.340
-.392
-.081
Short Delay Cued
0.953
.005
0.260
-.547
. 029
Long Delay
0.951
.003
0.368
-.540
-.078
Long Delay Cued
0.950
.003
0.282
-.594
.180
CVLT Recognition Measures
Correct Positive: 0.741
.001
0.131
-3.206
17.983
False Positive On
No Relation
List B:
0.646
.001
1.681
1.688
2.112
Semantic Shared
0.808
.001
1.027
.668
-0.271
False Positive Not
No Relation
on List
0.501
B:
.001
2.308
4.377
26.366
Semantic Shared
0.561
.001
1.967
1.860
2.411
Phonemic Shared
0.611
.001
1.799
2.324
5.538
Verbal Fluency
Assessment 1
0.975
.287
-
0.362
-.122
Assessment 2
0.965
.062
0.420
0.166
-.402
Assessment 3
0.961
.026
0.427
0.382
-.533
LOP Semantic
0.909
.001
0.170
-.745
-.020
LOP Phonological
0.969
.133
-
-.393
.724
LOP Orthographic
0.962
.045
0.333
.192
-.464
146

147
Table F-l continued
Shapiro-Wilk Test
Variable W:Normal Prob Cognitive-Affective and Physical Behavior Questionnaire
Cognitive-Affective
0.932
.001
0.375
-.086
.844
Physical
0.922
.001
0.280
-.393
.724
Frequency of Occurrence
0.936
.001
0.262
-.731
.237
Hemoglobin
0.984
.763

.146
-.038
Hematocrit
0.979
.467
-
.379
.849
Ferritin
0.715
.001
1.297
1.660
1.625
Age
0.941
.001
0.300
.199
-.774
Education
0.865
.001
0.221
-1.245
2.169
Note: Sd/M = ratio of standard deviation to mean; The
above data, except for Verbal Fluency, is collapsed
across assessments; LOP = Levels of Processing
Table F-2. SAS Folded F Test of Homogeneity of Variance
Between Groups
F'
Degrees
P
Variable
Value
of Freedom
value
California Verbal
Learning Test
Trial One
1.48
50
50
0.1656
Trial Two
1.14
50
50
0.6346
Trial Three
1.33
50
50
0.3228
Trial Four
1.46
50
50
0.1823
Trial Five
1.29
50
50
0.3779
Sum CVLT
1.39
50
50
0.2520
Sum Clusters
1.68
50
50
0.0711 *
Sum Intrusions
1.97
50
50
0.0178 *
Sum Perseverations
1.73
50
50
0.0553 *
Interference List
1.31
50
49
0.3453
Short Delay
1.51
50
49
0.1481
Clusters
1.35
50
49
0.2960
Perseverations
2.23
50
49
0.0058 *
Intrusions
1.68
50
49
0.0725 *
Short Delay Cued
1.25
49
50
0.4341
Long Delay
1.47
50
49
0.1829
Clusters
1.03
49
50
0.9233
Intrusions
1.15
50
49
0.6202
Long Delay Cued
1.28
49
50
0.3814

148
Table F-2 continued
F'
Degrees
P
Variable
Value
of Freedom
value
Recognition Memory
Correct Positive:
1.58
50
49
0.1127
False Positive On List
. B:
No Relation
1.26
49
50
0.4118
Semantic Shared
1.16
49
50
0.6079
False Positive Not on
List B
Semantic Shared
2.49
50
49
0.0017 *
Phonemic Shared
2.93
50
49
0.0002 *
No Relation
2.35
49
50
0.0032 *
Verbal Fluency
CFP
1.02
47
50
0.9395
FAS
1.21
50
47
0.5181
LSW
1.08
50
47
0.7843
Levels of Processing
Semantic
1.00
47
44
0.9916
Phonological
1.10
47
44
0.7607
Orthographic
2.14
47
44
0.0124 *
Frequency of Occurrence
Free Recall
1.16
50
49
0.6092
Estimation
1.32
50
49
0.3361
Cognitive-Affective and Physical
Behavior Questionnaire
Cognitive-Affective
1.06
49
50
0.8251
Physical
1.04
49
50
0.8790
Appetite
2.72
50
49
0.0006 *
Hemoglobin
1.26
50
50
0.4119
Hematocrit
1.44
50
50
0.1983
Ferritin
2.82
44
13
0.0460 *
Age
1.11
17
16
0.8333
Education
2.10
16
17
0.1400
Note: = Unequal variance

APPENDIX G
GROUP MEANS AND STANDARD DEVIATIONS
Table G-l. Means and Standard Deviations at Assessment One
bv Group
Control Group rEPO Treated Group
Variable
N
Mean
Sd
N
Mean
Sd
Age
18
48.22
14.19
17
43.24
13.45
Education in years
18
11.83
2.09
17
11.41
3.02
Days since EPO begun
-
-
16
41.37
32.08
Years on Dialysis
18
2.84
3.76
17
6.64
7.69
Ferritin
6
523.62
626.13
16
545.93
734.50
Hemoglobin
18
9.44
1.92
17
8.46
1.42
Hematocrit
18
28.19
6.12
17
25.58
3.77
California Verbal
Learning Test
Trial 1
18
5.72
1.74
17
5.88
2.47
Standard Score
18
-1.33
0.91
17
-1.35
1.37
Trial 2
18
7.89
2.00
17
8.41
2.37
Trial 3
18
9.17
2.73
17
9.65
3.52
Trial 4
18
9.89
2.91
17
10.71
2.93
Trial 5
18
10.56
3.07
17
11.24
3.25
Standard Score
18
-1.39
1.33
17
-1.59
2.12
Sum CVLT
18
43.22
10.70
17
45.82
13.03
Standard Score
18
33.06
10.31
17
33.82
16.00
Clusters
18
11.28
7.43
17
13.24
11.51
Perseverations
18
3.06
2.60
17
3.65
3.10
Intrusions
18
0.83
2.04
17
0.88
1.83
List B
18
5.94
1.66
17
5.76
1.92
Standard Score
18
-0.78
0.94
17
-0.88
1.05
Short Delay
18
9.06
2.71
17
9.65
3.26
Standard Score
18
-1.11
0.83
17
-1.06
1.43
Clusters
18
3.44
2.01
17
3.47
3.02
Perseverations
18
0.22
0.43
17
0.76
0.97*
Intrusions
18
0.17
0.51
17
0.24
0.56
Short Delay Cued
18
10.94
2.69
17
10.94
2.90
Standard Score
18
-0.83
1.04
17
-0.88
1.36
Perseverations
18
0.00
0.00
17
0.29
0.77
Intrusions
18
0.56
1.34
17
0.65
1.22
149

150
Table G-l continued
Control Group EPO Treated Group
Variable
N
Mean
Sd
N
Mean
Sd
Long Delay
18
9.94
3.44
17
9.47
3.76
Standard Score
18
-0.89
1.02
17
-1.41
1.66
Clusters
18
4.72
3.03
17
4.12
3.16
Perseverations
18
0.33
0.49
17
0.24
0.56
Intrusions
18
0.22
0.73
17
0.76
1.25*
Long Delay Cued
18
11.06
3.51
17
10.82
3.34
Standard Score
18
-0.78
1.40
17
-1.24
1.56
Perseverations
18
0.00
0.00
17
0.00
0.00
Intrusions
18
0.61
1.20
17
1.53
2.43
Recognition Hits
18
14.61
1.61
17
15.00
0.94
False Positives
From
Interference List
Shared Category
18
0.94
0.87
17
0.65
0.79
Unrelated
18
0.39
0.70
17
0.35
0.79
Intrusions not on Interference
List
Shared Semantic
18
0.11
0.32
17
0.41
0.80
Shared Phonemic
18
0.33
0.59
17
0.82
1.47
No Relation
18
0.78
2.16
17
0.29
0.99
Controlled Word Association
C
17
11.53
3.84
17
10.53
4.45
F
17
11.59
4.03
17
10.00
3.77
L
17
12.29
3.70
17
9.41
4.11*
LOP Semantic
16
9.62
1.67
16
10.25
1.98
LOP Orthographic
16
7.69
2.15
16
7.37
2.22
LOP Phonemic
16
7.81
1.64
16
7.44
2.80
Frequency Estimation Task
Free Recall
18
3.61
1.42
17
3.71
1.31
Frequency Estimate
18
11.44
2.20
17
11.24
1.89
Cognitive-Affective and
Physical
Behavior
Questionnaire
Cognitive-Affect
18
6.06
2.01
17
5.82
2.19
Physical
18
9.56
4.42
17
8.88
4.77
Appetite
18
0.67
0.49
17
0.88
0.33
Post Hemodialysis
5
0.80
0.45
15
0.73
0.46
Note: = Significant difference at
LOP = Levels of Processing
05 level

151
Table G-2. Means
and Standard
Deviations
at
Second
Assessment bv Group
Control Group
EPO
Treated
Group
Variable
N
Mean
Sd
N
Mean
Sd
Hemoglobin
18
9.23
1.81
17
9.59
1.84
Hematocrit
18
27.09
5.31
17
28.81
5.59
Ferritin
5
337.48
285.93
15
536.47
727.30
Assessment Interval 18
56.50
35.98
17
51.12
32.95
California Verbal
Learning Test
Trial 1
18
6.67
1.91
17
4.88
2.03
Standard Score
18
-0.78
0.88
17
-1.76
0.97
Trial 2
18
9.06
2.46
17
8.18
2.27
Trial 3
18
10.11
2.59
17
9.76
3.05
Trial 4
18
10.72
2.49
17
10.88
3.71
Trial 5
18
10.89
2.89
17
11.29
3.44
Standard Score
18
-1.50
1.34
17
-1.47
1.70
Sum CVLT
18
47.28
10.68
17
45.00
13.36
Standard Score
18
38.56
10.37
17
32.82
14.80
Clusters
18
12.78
9.72
17
13.53
9.84
Perseverations
18
3.61
2.77
17
4.00
3.79
Intrusions
18
1.50
2.20
17
2.06
2.84
List B
17
5.53
2.10
17
5.65
2.32
Standard Score
17
-0.88
1.27
17
-1.12
1.17
Short Delay
17
9.12
3.10
17
9.59
3.99
Standard Score
17
-1.12
1.17
17
-1.00
1.46
Clusters
17
3.88
3.08
17
3.88
2.80
Perseveration
17
0.59
0.80
17
0.41
0.80
Intrusions
17
0.47
0.51
17
0.53
1.01
Short Delay Cued
17
10.53
3.32
17
10.59
2.69
Standard Score
17
-0.94
1.09
17
-1.00
1.22
Perseverations
17
0.00
0.00
17
0.00
0.00
Intrusions
17
1.29
1.49
17
1.47
1.46
Long Delay
17
9.53
3.30
17
9.41
4.17
Standard Score
17
-1.24
1.09
17
-1.47
1.59
Clusters
17
4.53
3.48
17
4.24
2.80
Perseverations
17
0.35
0.79
17
0.12
0.33
Intrusions
17
0.82
1.29
17
1.00
1.17
Long Delay Cued
17
10.41
3.20
17
10.65
2.71
Standard Score
17
-1.24
1.30
17
-1.18
1.47
Perseverations
17
0.12
0.33
17
0.06
0.24
Intrusions
17
1.53
1.23
17
1.35
1.46
Recognition Hits
17
14.47
1.77
17
14.18
1.19
False Positives
from
Interference List
Shared Semantic
17
0.88
1.11
17
0.53
0.94

152
Table G-2 continued
Control Group EPO Treated Group
Variable N
Mean
Sd
N
Mean
Sd
Intrusions not on
Interference
List
Shared Semantic
17
0.35
0.61
17
0.41
0.87
Shared Phonemic
17
0.59
0.80
17
0.76
1.35
No Relation
17
0.29
0.69
17
0.41
0.71
Verbal Fluency
F
16
14.12
5.15
17
11.41
4.78
A
16
10.12
3.81
17
7.29
4.92
S
16
15.25
5.08
17
10.41
5.23
LOP Semantic
15
10.27
1.75
16
10.37
1.89
LOP Orthographic
15
6.20
2.11
16
6.88
2.55
LOP Phonemic
15
8.73
1.94
16
9.06
2.77
Frequency Estimation Task
Free Recall
17
4.24
1.48
17
3.06
1.39
Frequency Estimate
17
10.76
3.09
17
10.18
3.34
Cognitive-Affective
and
Physical
Behavior
Questionnaire
Cognitive-Affective
18
7.00
2.35
16
5.75
2.32
Physical
18
8.78
4.60
16
9.31
4.73
Appetite
18
0.56
0.51
16
0.94
0.25
Post Hemodialysis
6
0.50
0.55
15
0.60
0.51
Note: LOP = Levels of Processing

153
Table G-3. Means and Standard Deviations at Third
Assessment by Group
Control Group
EPO Treated
Variable
N
Mean
Sd
N
Mean
Sd
Hemoglobin
15
9.62
1.92
17
9.54
1.48
Hematocrit
15
28.12
5.55
17
27.55
4.06
Ferritin
3
484.30
266.67
14
560.34
814.39
Assessment Interval 15
67.80
34.56
17
90.41
24.03
California Verbal
Learning Test
Trial 1
15
5.87
1.64
17
6.00
1.97
Standard Score
15
-1.13
0.74
17
-1.24
1.03
Trial 2
15
9.00
2.80
17
8.94
3.17
Trial 3
15
10.33
2.23
17
9.82
2.24
Trial 4
15
10.93
2.66
17
9.82
3.13
Trial 5
15
10.93
2.52
17
11.12
3.02
Standard Score
15
-1.47
1.85
17
-1.53
1.59
Sum CVLT
15
47.07
10.40
17
45.12
11.61
Standard Score
15
36.80
13.21
17
33.18
13.65
Clusters
15
13.80
8.27
17
13.76
11.95
Perseverations
15
3.33
2.13
17
5.00
2.94
Intrusions
15
1.47
2.10
17
2.29
3.80
List B
15
6.40
2.16
17
4.88
2.50
Standard Score
15
-0.53
1.13
17
-1.35
1.37
Short Delay
15
9.07
2.87
17
9.47
3.39
Standard Score
15
-1.13
1.19
17
-1.12
1.32
Clusters
15
3.87
2.17
17
4.00
2.76
Perseverations
15
0.20
0.56
17
0.76
1.03
Intrusions
15
0.80
0.86
17
0.47
1.01
Short Delay Cued
15
9.87
2.83
17
10.76
2.39
Standard Score
15
-1.40
1.55
17
-1.00
1.12
Perseverations
15
0.07
0.26
17
0.00
0.00
Intrusions
15
0.73
0.88
17
1.29
1.65
Long Delay
15
9.27
2.87
17
9.65
3.82
Standard Score
15
-1.53
1.41
17
-1.24
1.48
Clusters
15
4.00
2.59
17
4.35
3.16
Perseverations
15
0.27
0.59
17
0.24
0.56
Intrusions
15
0.93
1.03
17
0.88
1.05
Long Delay Cued
15
10.13
2.95
17
10.82
2.53
Standard Score
15
-1.47
1.60
17
-1.24
1.35
Perseverations
15
0.00
0.00
17
0.00
0.00
Intrusions
15
1.20
1.15
17
1.41
1.54
Recognition Hits
15
14.07
1.62
17
13.59
3.18
False Positives
from
Interference List
Shared Category
15
0.87
0.99
17
0.82
0.88

154
Table G-3 continued
Control
Group
EPO
Treated
Variable
N
Mean
Sd
N
Mean
Sd
Intrusions not on
Interference
List
Shared Semantic
15
0.33
0.62
17
0.53
0.87
Shared Phonemic
15
0.33
0.82
17
0.59
0.94
Unrelated
15
0.67
0.82
17
0.65
1.06
Verbal Fluency
P
15
15.07
3.71
17
11.00
4.23
R
15
11.60
3.74
17
9.29
3.64
W
15
10.87
2.77
17
8.18
3.97
LOP Semantic
14
9.36
1.65
16
10.19
1.22
LOP Orthographic
14
6.79
2.52
16
7.12
2.55
LOP Phonemic
14
7.57
1.87
16
9.06
2.32
Estimation of Frequency
of Occurrence
Free Recall
15
4.67
1.23
17
3.41
1.87
Frequency Estimate
15
10.33
2.35
17
9.47
3.28
Cognitive-Affective
and
Behavioral Questionnaire
Cognitive-Affective
15
5.40
1.99
17
5.53
2.40
Physical
15
9.53
4.72
17
8.59
4.54
Note: LOP = Levels of Processing

APPENDIX H
CHANGE SCORES
Control Group Treatment Group
Variable
Tm
N
Mean
Sd
N
Mean
Sd
Hemoglobin
1
18
-0.21
1.34
17
1.13
1.57 b
Hemoglobin
2
15
0.05
2.01
17
1.08
1.23
Hematocrit
1
18
-1.11
4.32
17
3.24
4.54 b
Hematocrit
2
15
-0.43
6.21
17
1.98
3.14
California
Verbal Learning Test
Trial 1
1
18
0.94
2.24
17
-1.00
1.58 b
Trial 1
2
15
-0.67
1.99
17
1.12
1.76 b
Trial 2
1
18
1.17
2.41
0
-0.24
1.56 a
Trial 2
2
15
-0.13
2.53
0
0.76
2.31
Trial 3
1
18
0.94
3.06
17
0.12
2.15
Trial 3
2
15
0.13
2.10
17
0.06
2.28
Trial 4
1
18
0.83
3.01
17
0.18
2.46
Trial 4
2
15
0.07
1.39
17
-1.06
3.25
Trial 5
1
18
0.33
2.89
17
0.06
2.44
Trial 5
2
15
0.13
2.53
17
-0.18
1.78
SS
1
18
-0.11
1.78
17
0.12
1.54
SS
2
15
0.27
1.79
17
-0.06
1.09
Sum CVLT
1
18
4.06
9.91
17
-0.82
5.38
Sum CVLT
2
15
-0.27
6.97
17
0.12
7.85
SS
1
18
5.50
13.61
17
-0.82
5.38
SS
2
15
-0.67
10.88
17
0.12
7.85
Cluster
1
18
1.50
10.25
17
-1.00
7.27
Cluster
2
15
0.67
8.06
17
0.35
10.42
Psv
1
18
0.56
3.76
17
0.29
5.01
Psv
2
15
-0.73
3.13
17
0.24
6.94
Intrus
1
18
0.67
3.20
17
0.35
4.40
Intrus
2
15
-0.33
2.94
17
1.00
4.76
List B
1
17
-0.41
2.69
17
1.18
3.19
List B
2
15
0.67
2.97
17
0.24
2.82
SS
1
17
-0.06
1.60
17
-0.12
1.62
SS
2
15
0.20
1.74
17
-0.76
2.25
Short Del
1
17
0.06
2.75
17
-0.24
0.97
Short Del
2
15
-0.20
2.86
17
-0.24
1.35
SS
1
17
0.06
1.39
17
-0.06
2.66
SS
2
15
0.00
1.56
17
-0.12
2.26
155

156
Control Group Treatment Group
Variable
Tm
N
Mean
Sd
N
Mean
Sd
Short Delay
Cluster 1
17
0.47
2.67
17
0.06
1.34
Cluster
2
15
-0.27
2.96
17
-0.12
1.17
Psv
1
17
0.35
0.79
17
0.41
2.50
Psv
2
15
-0.40
0.99
17
0.12
2.34
Intrus
1
17
0.29
0.69
17
-0.35
1.22
Intrus
2
15
0.33
0.72
17
0.35
1.17
SD Cued
1
17
-0.35
2.32
17
0.29
1.26
SD Cued
2
15
-0.87
3.14
17
-0.06
0.56
SS
1
17
-0.06
1.14
17
-0.35
2.06
SS
2
15
-0.47
1.73
17
0.18
1.74
Psv
1
17
0.00
0.00
17
-0.12
1.27
Psv
2
15
0.07
0.26
17
0.00
0.87
Intrus
1
17
0.82
1.67
17
-0.29
0.77
Intrus
2
15
-0.33
1.68
17
0.00
0.00
Long Del
1
17
-0.47
2.15
17
0.82
1.67
Long Del
2
15
-0.60
2.50
17
-0.18
1.07
SS
1
17
-0.29
0.92
17
-0.06
2.90
SS
2
15
-0.40
1.35
17
0.24
2.36
Cluster
1
17
-0.18
1.88
17
-0.06
1.34
Cluster
2
15
-0.80
3.08
17
0.24
1.09
Psv
1
17
0.00
0.79
17
0.12
3.06
Psv
2
15
-0.13
0.99
17
0.12
2.67
Intrus
1
17
0.59
1.46
17
-0.12
0.70
Intrus
2
15
0.13
1.51
17
0.12
0.70
LD Cued
1
17
-0.53
2.96
17
0.24
1.48
LD Cued
2
15
-0.47
3.54
17
-0.12
1.32
Psv
1
17
0.12
0.33
17
-0.18
2.38
Psv
2
15
-0.13
0.35
17
0.18
1.29
Intrus
1
17
0.94
1.30
17
0.06
0.24
Intrus
2
15
-0.13
1.73
17
-0.06
0.24
Recog
1
17
-0.12
1.93
17
-0.18
1.91
Recog
2
15
-0.60
1.99
17
0.06
1.14
False +
1
17
0.88
3.10
17
0.59
1.50
False +
2
15
-0.47
2.50
17
0.12
3.04
Verbal Fluency
1
16
5.06
6.20
17
-0.82
1.33 a
2
14
-2.36
5.60
17
-0.59
2.92
Levels of
Semantic
Processing
1 15
0.60
2.03
17
-0.82
8.65
Semantic
2
14
-0.79
2.42
17
-0.65
7.65
Orthograph
1
15
-1.60
2.75
17
0.13
2.00
Orthograph
2
14
0.79
3.56
17
-0.19
1.87
Phonemic
1
15
0.87
2.72
16
-0.50
2.61
Phonemic
2
14
-1.07
2.09
16
0.25
2.46

157
Control Group Treatment Group
Variable
Tm
N
Mean
Sd
N
Mean
Sd
Estimation
of
Frequency of Occurrence
Recall
1
17
0.65
1.90
16
1.62
3.07 a
Recall
2
15
0.47
1.73
16
0.00
2.25
Estimation
1
17
-0.59
2.35
16
-0.65
1.37
Estimation
2
15
-0.67
2.66
16
0.35
1.32
Cognitive-Affective
and Behavior Questionnaire
Physical
Physical
1
18
-0.78
2.10
16
0.06
1.95
2
15
0.13
2.50
16
-0.25
2.98
Cognitive
1
18
0.94
2.21
17
-1.06
3.19
Cognitive
2
15
-1.73
2.02
17
-0.71
4.96 a
Appetite
1
18
-0.11
0.32
17
0.06
0.25
Appetite
2
15
-0.13
0.52
17
0.00
0.00
Note: a = p < .05; b p < .01
Tm= Assessment Interval; Psv = Perseverations;
Del = Delay; Intrus = Intrusions; SS = Standard
Score; Recog = Recognition; Orthograph = Orthographic;
False + = False positive

APPENDIX I
RAW DATA BY SUBJECT
Table 1-1. Raw Demographic Data
Hemoglobin by Assessment
Id
Acre
Ed
Sex
Trt
Group
1
2
3
1
24.2
14
M
Hemo
rEPO
7.90
9.00
10.80
2
31.2
13
F
Hemo
rEPO
9.50
10.60
10.10
3
53.7
12
M
Hemo
rEPO
9.00
9.30
9.30
4
29.6
8
F
Hemo
rEPO
7.54
8.10
8.00
5
37.1
14
F
Hemo
rEPO
5.60
6.60
6.90
6
45.2
10
F
Hemo
rEPO
8.30
14.20
8.70
7
43.0
12
F
Hemo
rEPO
8.10
7.90
8.60
8
51.0
12
F
Hemo
rEPO
7.55
8.85
7.25
9
49.3
12
F
Hemo
rEPO
8.50
9.10
9.40
10
25.8
12
M
Hemo
rEPO
9.60
9.75
9.65
11
26.5
12
F
Hemo
rEPO
7.10
10.50
11.10
12
56.3
7
M
Hemo
rEPO
7.20
6.95
7.90
13
46.2
13
M
Hemo
rEPO
11.30
10.60
10.84
14
57.1
12
M
Hemo
rEPO
10.50
10.70
10.81
15
48.0
12
M
Hemo
rEPO
10.20
12.00
12.25
16
37.0
16
F
Perit
rEPO
8.10
10.10
10.30
17
73.8
3
F
Hemo
rEPO
7.80
8.80
10.30
101
57.1
12
F
Hemo
Con
6.90
8.30
10.20
102
48.0
12
M
Hemo
Con
9.40
10.00
11.70
103
34.8
8
F
Perit
Con
5.27
6.48
6.90
104
71.5
8
M
Hemo
Con
9.70
9.61

105
54.7
12
M
Hemo
Con
13.90
12.60
10.50
106
60.5
15
M
Hemo
Con
9.25
6.30

107
43.5
13
M
Perit
Con
12.30
11.20
12.30
108
23.9
12
M
Perit
Con
8.40
9.40
9.90
109
43.5
14
M
Perit
Con
10.90
11.10
11.60
110
57.3
10
F
Perit
Con
9.50
10.00
10.20
111
48.1
14
M
Perit
Con
9.60
9.40
9.70
112
70.0
10
M
Perit
Con
10.71
10.60
8.96
113
65.8
12
F
Perit
Con
9.80
8.10
9.10
114
32.9
12
F
Perit
Con
9.90
11.60
11.60
115
33.9
14
F
Perit
Con
8.90
7.30
6.00
116
29.4
12
F
Perit
Con
8.80
9.10
9.10
117
38.5
8
F
Perit
Con
7.30
8.07

118
54.5
14
M
Perit
Con
9.40
7.00
6.60
Note:
Ed =
Years
of
Education; Trt
= Treatment; Con

Control; Hemo = Hemodialysis;
Perit
= Peritoneal Dialysis
158

159
Table 1-2. California Verbal Learning Test Trials One.
Five. Sum and Sum of Clustering Raw Data by Assessment
Number
Raw Score Raw Score T Score
Raw Score
Trial 1 Trial 5 Sum CVLT Sum Cluster
Assess
1
2
3
1
2
3
1
2
3
1
2
3
Id
1
8
6
7
15
12
13
49
41
35
23
21
14
2
3
5
5
10
11
11
14
12
11
4
5
8
3
9
8
8
12
10
10
51
44
38
4
10
3
4
5
4
4
10
11
10
6
12
5
11
11
7
5
6
5
5
15
13
10
40
33
32
15
8
21
6
4
6
7
7
9
8
17
25
22
7
14
8
7
5
4
8
10
13
15
22
26
46
12
14
15
8
5
4
6
9
12
13
26
26
39
9
6
13
9
7
7
5
14
15
12
50
51
29
22
20
15
10
10
6
8
14
16
14
49
59
52
30
33
35
11
9
7
7
13
14
14
40
43
33
14
17
13
12
3
0
3
4
7
7
14
19
26
1
8
2
13
4
3
8
12
13
11
34
42
40
4
12
5
14
5
5
6
13
10
10
49
37
37
12
11
9
15
4
4
3
9
7
11
30
19
34
6
1
11
16
10
7
9
16
16
16
59
53
61
46
37
49
17
3
2
3
8
3
4
25
16
24
5
2
6
101
5
4
6
10
9
10
33
35
35
10
7
11
102
5
4
5
10
8
7
27
19
24
14
5
10
103
8
6
6
13
11
8
30
19
20
17
12
12
104
3
6

8
8

31
40

4
3

105
4
5
5
9
9
9
34
42
34
14
12
10
106
5
9

9
12

37
49

5
15

107
6
7
6
10
12
9
34
40
37
9
11
6
108
7
7
9
13
11
14
43
41
54
15
11
15
109
6
7
7
13
13
14
49
44
51
11
16
33
110
8
5
5
14
11
12
55
36
36
25
9
16
111
8
10
6
10
9
11
39
43
43
7
11
10
112
3
5
4
5
6
9
26
34
36
0
4
2
113
8
6
5
5
13
14
22
50
54
4
7
12
114
8
8
9
16
14
14
43
45
56
29
22
23
115
5
9
7
15
16
14
27
56
35
14
44
28
116
5
10
5
14
9
11
38
11
12
23
10
117
5
7

1
14

26
43

3
15

118
4
5
3
9
6
10
28
20
26
10
3
9

160
Table 1-3. California Verbal Learning Test Short Delay,
Long Delay and Recognition Raw Scores
Short Delay Long Delay Recognition
Assess
1
2
3
1
2
3
1
2
3
Id
1
15
10
11
16
10
11
16
14
14
2
9
10
8
8
10
10
14
14
16
3
9
10
7
10
11
8
14
15
14
4
8
9
7
7
6
8
15
12
13
5
12
12
11
12
11
13
16
13
14
6
6
8
7
5
10
6
15
14
15
7
8
13
13
11
12
13
16
16
15
8
8
7
10
10
11
9
15
14
14
9
14
13
12
14
13
10
16
14
14
10
13
14
13
12
15
13
15
15
16
11
11
14
14
11
13
15
15
15
16
12
4
3
6
2
2
2
14
15
14
13
10
13
9
9
12
11
16
16
14
14
9
7
8
5
5
8
13
13
2
15
7
5
10
10
6
10
14
14
12
16
15
14
14
14
13
15
16
15
15
17
6
1
1
5
0
2
15
12
13
101
8
7
9
9
10
10
16
16
16
102
9
7
8
9
8
8
14
14
11
103
10
9
6
11
10
6
16
15
12
104
3
6

2
6

11
11

105
6
10
8
8
6
8
16
16
14
106
9


9


15


107
9
8
11
10
8
8
16
16
16
108
11
8
12
11
9
11
16
13
15
109
10
9
10
14
12
13
13
15
14
110
12
10
10
12
9
9
16
16
14
111
10
6
10
12
9
11
15
14
15
112
3
5
2
3
4
3
12
14
11
113
10
13
11
12
13
11
15
15
15
114
13
14
13
15
14
14
14
14
14
115
12
14
12
14
15
12
16
16
16
116
9
14
8
11
15
8
12
16
14
117
10
10

10
8

15
15

118
9
5
6
7
6
7
15
10
14
Note: Assess = Assessment number

161
Table 1-4. Raw Verbal Fluency and Self-Report Questionnaire
Data bv Assessment
Controlled Word
Association
CFL
FAS
FAS %
PRW
Assess
Id
1
2
2
3
1
38
38
55-59
29
2
38
48
85-89
33
3
29
23
15-19
26
4
18
8
<4
15
5
25
31
35-39
25
6
33
21
10-14
32
7
17
14
<4
18
8
25
28
25-29
26
9
23
46
85-89
34
10
41
42
80-84
42
11
47
41
75-79
48
12
16
13
5-9
12
13
29
16
>4
29
14
25
28
30-34
23
15
32
30
35-39
27
16
52
51
95-
49
17
21
17
20-24
16
101
23
26
10-14
26
102
43
47
85-89
42
103
28
40
80-84
36
104
22
23
50-54

105
23
40
75-79
40
106



107
43
54
>95
46
108
42
45
75-79
48
109
51

47
110
28
25
25-29
30
111
33
42
80-84
30
112
27
32
65-69
28
113
33
31
50-54
31
114
42
55
95-
51
115
48
49
90-94
44
116
29
26
20-24
35
117
50
60
>95

118
37
37
65-69
29
Cognitive-Affective and
Physical Behavior
Questionnaire
Physical Cognitive
1
2
3
1
2
3
13
17
13
9
8
8
15
13
12
3
1
4
5
4
6
7
5
6
3
5
14
6
6
8
4
7
7
4
5
4
6
7
5
1
3
1
14
14
14
7
8
8
14
14
12
9
9
8
12
12
11
8
8
8
12
14
14
7
8
8
15
14
10
8
8
8
3
3
1
5
6
4
7
4
3
6
3
3
3
3
4
4
4
2
10
8
9
5
4
5
3

1
4

4
12
10
10
6
6
5
12
11
11
5
6
5
5
5
9
7
5
4
12
8
11
6
7
5
8
8

8
7

0
0
2
1
2
1
6
3

8
6

14
14
11
5
7
6
5
7
5
4
6
4
10
13
9
5
7
3
14
13
11
8
7
7
14
17
17
8
9
8
11
9
9
6
8
6
10
6
3
8
8
6
13
13
15
8
8
8
2
2
2
4
5
4
14
11
14
6
14
6
8
6

4
6

14
12
14
8
8
8
Note: Assess = Assessment number

BIBLIOGRAPHY
Abram, H.S., Moore, G.L., & Westervelt, F.B. (1971).
Suicidal behavior in chronic dialysis patients. American
Journal of Psychiatry. 127. 1199-1204.
Afifi, A.K., & Bergman, R.A. (1980). Basic Neuroscience.
Baltimore, Maryland: Urban & Schwarzenberg.
Alfred, H.J., Graff, L.G., Cohen, A.J., & Shuster, A. (1981).
Treatment of Renal Failure. In B.D. Rose (Ed.),
Pathophysiology of Renal Disease (pp.475-573). New York,
New York: McGraw-Hill.
Arieff, A.I., Guisado, R., & Massry, S.G. (1975). Uremic
encephalopathy: Studies on biochemical alterations in the
brain. Kidney International. 7, Supplement 194-200.
Atkinson, R.C., & Shiffrin, R.M. (1968). Human memory: A
proposed system and its control processes. In K.W. Spence &
J.T. Spence (Eds.), The Psychology of Learning and
Motivation: Advances in Research and Theory. 2, New York,
New York: Academic Press.
Atkinson, R.C., & Shiffrin, R.M. (1971). The control of short
term memory. Scientific American. 225. 82-90.
Bartle, E.J, Sun, J.H, Thompson, L., Light, A.I., McCool C., &
Heaton S. (1988). The effects of acute sleep deprivation
during residency training. Surgery. 104. 311-316.
Battig, W.F., & Montague, W.E. (1969). Category norms for
verbal items in 56 categories: A replication and extension
of the Connecticut category norms. Journal of Experimental
Psychology Monograph. 80. 1-46.
Ben-Shachar, D., Ashkenazi, R., & Youdim, M.B. (1986). Long
term consequence of early iron-deficiency on dopaminergic
neurotransmission in rats. International Journal of
Developmental Neuroscience. 4, 81-88.
Berry, D.T., Webb, W.B., Block, A.J., Bauer, R.M., & Switzer,
D.A. (1986). Nocturnal hypoxia and neuropsychological
variables. Journal of Clinical & Experimental
Neuropsychology. 8, 229-238. (Medline Abstract).
162

163
Birkmayer, W., & Birkmayer, J.D. (1986). Iron, a new aid in
the treatment of Parkinson patients. Journal of Neural
Transmission, 67, 287-292.
Birkmayer, W., & Birkmayer, J.D. (1987). Dopamine action and
disorders of neurotransmitter balance. Gerontology 33. 168-
71.
Blatt, B., & Tsushima, W.T. (1966). A psychological survey of
uremic patients being considered for the chronic
hemodialysis program. Intellectual and emotional problems
in uremic patients. Nephrology. 3., 206-208.
Brenner, B.M., & Lazarus, J.M. (1980). Chronic Renal Failure
In K. Isselbacher, R. Adams, E. Braunwald, R. Petersdorf, &
J. Wilson (Eds.), Harrison/s Principles of Internal
Medicine (9th ed.). (pp. 1299-1307). New York, New York:
Mcgraw-Hill.
Bunn, H.F. (1980a). Anemia associated with chronic systemic
disorders. In K. Isselbacher, R. Adams, E. Braunwald, R.
Petersdorf, & J. Wilson (Eds.), Harrisons Principles of
Internal Medicine (9th ed.). (pp. 1530-1532). New York,
New York: Mcgraw-Hill.
Bunn, H. F. (1980b). Pallor and Anemia. In K. Isselbacher,
R. Adams, E. Braunwald, R. Petersdorf, & J. Wilson
(Eds.), Harrisons Principles of Internal Medicine (9th
ed.). (pp. 262-272). New York, New York: Mcgraw-Hill.
Canadian Erythropoietin Study Group (1989). The effect of
recombinant erythropoietin (EPO) upon quality of life
(QL) and functional capacity (FC) of anemic patients on
chronic hemodialysis. Kidney Internatinal. 35, 195.
Abstract.
Carroll, J.B., Davies, P., & Richman, B. (1971). The
American Heritage Word Freguencv Book. New York, New
York: American Heritage Publishing Company.
Cermak, L.S. (1982). The Long and Short of It in Amnesia. In
L.S. Cermak (Ed.), Human Memory and Amnesia (pp. 43-59).
Hillsdale, New Jersey: Lawrence Erlbaum Associates.
Chandra, M., McVicar, M., & Clemons, G. K. (1988).
Pathogenesis of the anemia of chronic renal failure: the
role of erythropoietin. Adv. Pediatr [200] 5, 361-389.
Cohen, R.M., Weingartner, H., Smallberg, S.A., Pickar, D., &
Murphy, D.L. (1982). Archives of General Psychiatry. 39.
593-597.

164
Comty, C.M., Leonard, A., & Shapiro, F.C. (1974).
Psychosocial problems in dialized diabetic patients.
Kidney International. 1, Supplement 144-151.
Craik, F.I.M., & Lockhart, R.S. (1972). Levels of
processing: a framework for memory research. Journal of
Verbal Learning and Verbal Behavior. 11. 671-684.
Craik, F.I.M., & Tulving, E. (1975). Depth of processing and
the retention of words in episodic memory. Journal of
Experimental Psychology: General. 104. 268-294.
Crittenden, M.R., Holliday, M.A., Piel, C.F., & Potter, D.E.
(1985). Intellectual development of children with renal
insufficiency and end stage disease. International
Journal of Pediatric Nephrology. 6, 275-280.
Darwin, C.J., Turvey, M.T., & Crowder, R.G. (1972). An
auditory analogue of the Sperling partial report
procedure: Evidence for brief auditory storage, Cognitive
Psychology. 3, 255-267.
Davies-Jones, G.A.B., Preston, F.E., & Timperley, W.R.
(1980). Neurological Complications in Clinical
Hematology. Boston, Massachusetts: Blackwell Scientific
Publications.
Dawson, A.A., Ogston, D., & Fullerton, H.W. (1969).
Evaluation of diagnostic significance of certain symptoms
and physical signs in anemic patients. British Medical
Journal, 3, 436-439.
Deinard A.S., List, A., Lindgren, B., Hunt, J.V., & Chang,
P. (1986). Cognitive deficits in iron-deficient and iron-
deficient anemic children. Journal of Pediatrics. 108.
681-689.
Delano, B.G. (1989). Improvements in quality of life
following treatment with r-HuEPO in anemic hemodialysis
patients. American Journal of Kidney Diseases. 14, 14-18.
Delis, D.C., Freeland, J., Kramer, J.H., & Kaplan, E.
(1988). Integrating clinical assessment with cognitive
neuroscience: Construct validation of the California
Verbal Learning Test. Journal of Consulting and Clinical
Psychology. 56. 123-130.
Delis, D.C., Kramer, J.H., Kaplan, E., & Ober, B.A. (1987).
California Verbal Learning Test Adult Version Research
Edition Manual. New York, New York: Psychological
Corporation.

165
Delis, D.C., McKee, R., Massman, P., Kramer, J.H., &
Gettman, D. (1990). Alternate Form of the California
Verbal Learning Test: Development and Reliability,
unpublished manuscript.
Desforges, J.F. (1975). The role of hemodialysis in the
hematologic disorders of uremia. Kidney International. 7,
S-123-128.
Drayer, B.P., Burger, P., Darwin, R., Reiderer, S.,
Herfkens, R., & Johnson, G.A. (1986). Magnetic resonance
imaging of brain iron. Americal Journal of
Neuroradioloqy. 7, 373-380.
English, A., Savage, R.D., Britton, P.G., Ward, M.K., &
Kerr, D.N.S. (1978). Intellectual impairment in chronic
renal failure. British Medical Journal. 1, 888-890.
Erslev, A.J. (1975). The effect of uremic toxins on the
production and metabolism of erythropoietin. Kidney
Internation. 7, S129-S133.
Erslev, A.J. (1987). Erythropoietin coming of age. New
England Journal of Medicine. 316. 101-103.
Eschbach, J.W. (1989). The anemia of chronic renal failure:
pathophysiology and the effects of the effrects of
recombinant erythropoietin. Kidney International. 35.
134-148.
Eschbach, J.W., & Adamson, J.W. (1985). Anemia of end-stage
renal disease (ESRD). Kidney International. 28, 1-5.
Eschbach, J.W., Egrie, J.C., Downing, M.R., Browne, J.K., &
Adamson, J.W. (1987). Correction of the anemia of end-
stage renal disease with recombinant human
erythropoietin. New England Journal of Medicine. 316. 73-
78.
Evans, D.I. (1985). Cerebral function in iron deficiency: A
review. Child-Care, Health and Development 11. 105-112.
Evans, R.W., Rader, B., & Manninen (1990). The quality of
life of hemodialysis recipients treated with recombinant
human erythropoietin. Journal of the American Medical
Association, 263, 825-830.
Ewing, R., McCarthy, D., Gronwall, D., & Wrightson, P.
(1980). Persisting effects of minor head injury
observable during hypoxic stress. Journal of Clinical
Neuropsychology. 2, 147-155.

166
Fennell, R.S., Fennell, E.B., Carter, R.L., Mings, E.L.,
Klausner, A.B., & Hurst, J.R. (1989). Correlations
between performance on neuropsychological tests of
children with chronic renal disease. Unpublished.
Fennell, R.S., Fennell, E.B., Carter, R.L., Mings, E.L.,
Klausner, A.B., & Hurst, J.R. (1990a). Association
between renal function and cognition in childhood chronic
renal failure. Pediatric Nephrology. 4., 16-20.
Fennell, R.S., Fennell, E.B., Carter, R.L., Mings, E.L.,
Klausner, A.B., & Hurst, J.R. (1990b). A longitudinal
study of the cognitive function of children with renal
failure. Pediatric Nephrology. 4, 11-15.
Fennell, R.S., Fennell, E.B., Mings, E. L., Klausner, A. B.,
Hurst, J.R., & Carter, R.L. (1987). Cognitive performance
in children with renal failure as affected by renal
function and therapeutic modalities. In K. Murakami, T.
Kitagawa, K., Yabuta, & T. Sakai (Eds.), Recent Advances
in Pediatric Nephrology (pp. 615-618). New York, New
York: Excerptica Medica.
Fennell, E.B., Fennell, R.S., Mings, E., & Morris, M.K.
(1986). The effects of various modes of therapy for end
stage renal disease on cognitive performance in a
pediatric population: A preliminary report. International
Journal of Pediatric Nephrology. 7, 107-112.
Fennell, R.S., & Rasbury, W.C. (1980). Cognitive functioning
of identical twins discordant for Prune Belly Syndrome
and end-stage renal disease. International Journal of
Pediatric Nephrology. 1, 234-239.
Fennell, R.S., Rasbury, W.C., Fennell, E.B., & Morris, M.K.
(1984). Effects of kidney transplantation on cognitive
performance in a pediatric population. Pediatrics. 74.,
273-278.
Fishman, D.B., & Schneider, C.J. (1972). Predicting
emotional adjustment in home dialysis patients and their
relatives. Journal of Chronic Diseases. 25, 99-109.
Freeman, C.W., Sherrard, D.J., Carlsyn, D.A., & Paige, A.B.
(1980). Psychological assessment of renal dialysis
patients using standard psychometric techniques. Journal
of Counseling and Clinical Psychology. 48, 537-539.
Fried, R. (1987). The Hvperventillation Syndrome: Research
and Clinical Treatment. Baltimore, Maryland: Johns
Hopkins University Press.
Gilli, P., & DeBastiani, P. (1983). Cognitive function and
dialysis treatment. Clinical Neurology. 16, 188-192.

167
Ginn, H.E., (1975). Neurobehavioral dysfunction in uremia.
Kidney International. 7(S), 217-221.
Ginn, H.E., Teschan, P.E., Bourne, J.R., Hamel, B., Ward,
J.W., Vaugh, W.K., & Nunnally, J.C. (1978).
Neurobehavioral and clinical responses to hemodialysis.
Trans. Am Soc Artif Intern Organs, 24. 376-378.
Ginn, H.E., Teschan, P.E., Walker, P.J., Bourne, J.R.,
Fristoe, M., Ward, J.W., McLain, L.W., Johnston, H.B., &
Hamel, B. (1975). Neurotoxicity in Uremia. Kidney
International. , S357-S360.
Goldberg, A.P., Geltman, E.M., Gavin, J.R, 3rd., Carney,
R.M., Hagberg, J.M., Delmez, J.A., Naumovich, A.,
Oldfield, M.H., & Harter, H.R. (1986). Exercise training
reduces coronary risk and effectively rehabilitates
hemodialysis patients. Nephron. 42. 311-316.
Groner J.A., Holtzman, N.A., Charney, E., & Mellits, E.D.
(1986). A randomized trial of oral iron on tests of
short-term memory and attention span in young pregnant
women. Journal of Adolescent Health Care. 7, 44-48.
Gutman, R.A., Stead, W.W., & Robinson, R.R. (1981). Physical
activity and employment status of patients on maintenance
hemodialysis. New England Journal of Medicine. 304. 309-
313.
Hagberg, B. (1974). A prospective study of patients in
chronic hemodialysis-III: Predictive value of
intelligence, cognitive deficit, and ego defense
structures in rehabilitation. Journal of Psychosocial
Research. 18. 151-160.
Hallgren, B., & Sourander, P. (1958). The effect of age on
the nonheme iron in the human brain. Journal of
Neurochemistrv. 2, 41-51.
Hart, R.P., Pederson, J.A., Czerwinski, A.W., & Adams, R.L.
(1983). Chronic renal failure, dialysis and
neuropsychological function. Journal of Clinical
Neuropsychology. 5, 301-312.
Hasher. L., & Zacks, R.T. (1984). Automatic Processing of
Fundamental Information. American Psychologist. 39. 1372-
1388.
Heilman, K.M., Moyer, R.S., Melendez, F., Schwartz, H.D., &
Miller, B.D. (1975). A memory deficit in uremic
encephalopathy. Journal of Neurological Science. 26, 245-
249.

168
Heistadt, D.D., Marcus, M.L., & Abboud, F.M. (1987).
Experimental attempts to unmask the effects of neural
stimuli on cerebral blood flow. In M.J. Purves, (Ed.),
Cerebral Vascular Smooth Muscle and Its Control (pp. 142-
165). Amsterdam, Netherlands: Elsevier.
Hennessy, T.G., Stern, W.E., & Herrick, S.E. (1967).
Cerebellar Hemangioblastoma: Erythropoietic Activity by
Radio-iron Assay. Journal of Nuclear Medicine. 8, 601-
606.
Honig, A.S., & Oski, F.A. (1984). Solemnity: a clinical risk
index for iron deficient infants. Early Child Development
and Care. 16, 69-83.
Hori, K., Onoyama, K., Iseki, K., Fujimi, S., & Fujishima,
M. (1990). Hemodynamic and volume changes by recombinant
human erythropoietin (rHuEPO) in the treatment of anemic
hemodialysis. Clinical Nephrology. 33. 293-298.
Hunt, K.P., & Hodge, M.H. (1971). Category-item frequency
and category-name meaningfulness (m'): Taxonomic norms
for 84 categories. Psvchonomic Monograph Supplements. 4./
97-121.
Jackson, M., Warrington, E.K., Roe, C.J., & Baker, L.R.I.
(1987). Cognitive function in hemodialysis patients.
Clinical Nephrology. 27. 26-30.
Jacob, H.S., Eaton, J.W., & Yawata, Y. (1975). Shortened red
blood cell survival in uremic patients: beneficial and
deleterious effects of dialysis. Kidney International. 7,
S-139-14 3.
Johnson, D.L., & McGowan, R.J. (1983). Anemia and infant
behavior. Nutrition and Behavior. 1, 185-192.
Jonides, J., & Naveh-Benjamin, M. (1987). Estimating
Frequency of Occurrence. Journal of Experimental
Psychology: Learning. Memory, and Cognition. 13. 230-240.
Kaplan-DeNour, A., Shaltiel, J., & Czaczbes, J.W. (1968).
Emotional reactions of patients on chronic hemodialysis.
Psychosomatic Medicine. 300. 521-533.
Klein, W.L., Frederickson, E.D., Fennell, E.B., Peterson,
J.C., Guzman, S.H., & Tisher, C.C. (1989). Effects of
erythropoietin on neuropsychological function and
reported mood in the anemia of end stage renal disease.
American Federation for Clinical Research (Abstract).

169
Koene, R.A.P., & Frenken, L.A.M. (1990). Does treatment of
predialysis patients with recombinant human
erythropoietin compromise renal function? In R.M.
Schaefer, W.H. Horl, & A. Heidland (Eds.), Erythropoietin
in the 90s. New York, New York: Karger.
Lee, G.R., Wintrobe, M.M., & Bunn, H.F. (1980). Iron-
deficiency anemia and the sideroblastic anemias. In K.J.
Isselbacher, R.D. Adams, E. Braunwald, R.G. Petersdorf, &
J.D. Wilson (Eds.), Harrison's Principles of Internal
Medicine (9th ed.). (pp. 1514-1518). New York, New York:
McGraw Hill.
Lewis, E.G., O'Neill, W.M., Dustman, R.E., & Beck, E.C.
(1980). Temporal effects of hemodialysis on measures of
neural efficiency. Kidney International. 17, 357-363.
Lezak, M. D. (1983). Neuropsychological Assessment. (2nd
ed.). New York, New York: Oxford U. Press.
Lozoff, B. (1989). Nutrition and Behavior. American
Psychologist. 44. 231-236.
Lundin, A.P., Delano, B.G., & Quinn-Cefaro, R. (1990).
Perspectives on the improvement of quality of life with
epoetin alfa therapy. Pharmacotherapy. 10, 228-268.
Marshall, J.R. (1979). Neuropsychiatric aspects of renal
failure. Journal of Clinical Psychiatry. 40. 81-85.
Massaro T.F., & Widmayer, P. (1981). The effect of iron
deficiency on cognitive performance in the rat. American
Journal of Clinical Nutrition. 34, 864-870.
Mayer, G., Thum, J., Cada, E.M., Stummvoll, H.K., & Graf, H.
(1988). Working capacity is increased following
recombinant human erythropoietin treatment. Kidney
International. 34. 525-528.
McCarty, S.M., Logue, P.E., Power, D.G., Ziesat, H.A., &
Rosenstiel, A.K. (1980). Alternate form reliability and
age-related scores for Russell's Revised Wechsler Memory
Scale. Journal of Consulting and Clinical Psychology. 48,
296-298.
McDaniel, J.W. (1971). Metabolic and CNS correlates of
cognitive dysfunction with renal failure.
Psychophysiology. 8., 704-713.
Miller, G.A. (1956). The magical number seven, plus or minus
two: Some limits on our capacity to process information.
Psychological Review. 63. 81-97.

170
Mings, E. L. (1987) Visual Motor Integration Deficits in
Pediatric Renal Disease. Doctoral dissertation,
University of Florida.
Miskin, M. (1978). Memory in monkeys severely impaired by
combined but not by separate removal of amygdala and
hippocampus. Nature. 273. 297-298.
Moray, N. (1959). Attention in dichotic listening: Affective
cues and the influence of instructions. Quarterly Journal
of Experimental Psychology. 11, 59-60.
Morris, C.D., Bransford, J.D., & Franks, J.J. (1977). Levels
of processing vs. transfer appropriate processing.
Journal of Verbal Learning and Verbal Behavior. 16 519-
533.
Muramoto, 0., Kuru, Y., Sugishita, M., & Toyokura, Y.
(1979). Pure memory loss with hippocampal lesions. A
pneumoencephalographic study. Archives of Neurology. 36.
54-56.
Newman, R.P., Weingartner, H., Smallberg, S.A., & Caine,
D.B. (1984). Effortful and automatic memory: Effects of
dopamine. Neurology. 34. 805-807.
Nissen, M.J. Knopman, D.S., & Schacter, D.L. (1987).
Neurochemical dissociation of memory systems. Neurology.
789-794.
Nissenson, A.R. (1989). Recombinant human erythropoietin:
Impact on brain and cognitive function, exercise
tolerance, sexual potency, and quality of life. Seminars
in Nephrology. 9, 25-31.
Nissenson, A.R., Levin, M.L., Klawans, H.L., & Nausieda,
P.L. (1977). Neurological sequelae of end-stage renal
disease. Journal of Chronic Disease. 30, 705-733.
Nissenson, A.R., Marsh, J.T., & Brown, W.S. (1988).
Recombinant erythropoietin improves brain function in
chronic hemodialysis patients. Kidney International. 33.
232. Abstract.
Nissenson, A.R., Marsh, J.T., Brown, W.S., Schweitzer, S., &
Wolcott (1989). Brain function improves in chronic
hemodialysis patients after recombinant erythropoietin.
Kidney International. 35. 257. Abstract.
Osberg, J. W., Meares, G.J., McKee, D.C., & Burnett, G.B.
(1982). Intellectual functioning in renal failure and
chronic dialysis. Journal of Chronic Diseases. 35, 445-
457.

171
Oski, F.A., & Honig, A. (1983). The effects of therapy on
the developmental scores of iron-deficient infants.
Journal of Pediatrics. 92. 21-25.
Oski F.A., Honig A.S., Helu B.M., & Howanitz P. (1978).
Effect of iron therapy on behavior performance in non-
anemic iron-deficient infants. Pediatrics. 71. 877-880.
Paganini, E.P. (1989). Overview of anemia associated with
chronic renal disease: Primary and scondary mechanisms.
Seminars in Nephrology. 9, Supplement 3-8.
Peterson, L.R., & Peterson, M.J. (1959). Short-term
retention of individual verbal items. Journal of
Experimental Psychology. 58, 193-198.
Pollitt, E., Leibel, R.L., & Greenfield, D.B. (1983). Iron
deficiency and cognitive test performance in preschool
children. Nutrition and Behavior. 1, 137-146.
Pollitt, E., Siemantri, A.G., Yunis, F., & Scrimshaw, N.S.
(1985). Cognitive effects of iron-deficiency anemia
(Letter to the editor). Lancet. 1, 158.
Powell. D., Bergstrom, J., Rastislav, D., Gulyassy, P.,
Lockwood, D., & Phillips, L. (1986). Toxins and
inhibitors in chronic renal failure. American Journal of
Kidney Diseases. 7(4) 292-299.
Race, G.J., Finney, J.W., Mallams, J.T., & Balia, G.A.
(1964). Hematopoietic stimulating effect of a Cerebellar
Hemangioblastoma. Journal of the American Medical
Association. 187. 150-151.
Rapaport, S.I. (1987). Introduction to Hematology (2nd ed.).
Philadelphia, Pennsylvania: J.B. Lippincott.
Rasbury, W.C., Fennell, R.S., Fennell, E.B., & Morris, M.K.
(1986). Cognitive functioning in children with end stage
renal disease pre- and post-dialysis session.
International Journal of Pediatric Nephrology. 7, 45-50.
Rasbury, W.C, Fennell, R.S, & Morris, M.K (1983). Cognitive
functioning of children with end-stage renal disease pre
and post successful transplant. Journal of Pediatrics.
102. 589-592.
Ratner, D.P., Adams, K.M., Levin, N.W., & Rourke, B.P.
(1983). Effects of hemodialysis on the cognitive and
sensory-motor functioning of the adult chronic
hemodialysis patient. Journal of Behavioral Medicine. 6,
291-311.

172
Ritchey, A.K. (1987). Iron deficiency in children.
Postgraduate Medicine. 82. 59-69.
Roediger, H.L. (1990). Implicit Memory: Retention without
remembering. American Psychologist. 45. 1043-1056.
Rundus, D. (1971). Analysis of rehearsal processes in free-
recall. Journal of Experimental Psychology. 89. 63-77.
Rundus, D. (1977). Maintenance rehearsal and single-level
processing. Journal of Verbal Learning and Verbal
Behavior. 16. 665-681.
Ryan, J.J., Geisser, M.E., Randall, D.M., & Georgemiller,
R.J. (1986). Alternate form reliability and equivalency
of the Rey Auditory Verbal Learning Test. Journal of
Clinical and Experimental Neuropsychology. 8, 611-616.
Ryan, J.J., Souheaver, G.T., & DeWolfe, A.S. (1980).
Intellectual deficit in chronic renal failure: A
comparison with neurological and medical-psychiatric
patients. Journal of Nervous and Mental Diseases. 168.
763-767.
Ryan, J.J., Souheaver, G.T., & DeWolfe, A.S. (1981).
Halstead-Reitan test results in chronic hemodialysis.
Journal of Nervous and Mental Diseases. 169. 311-314.
Rybo, E., Bengtsson, C., Hallberg, L., & Oden, A. (1985).
Effect of iron supplementation on women with iron
deficiency. Scandinavian Journal of Haematology. 34.
(Supplement 43), 103-114.
Sanders, R.E., Gonzalez, E.G., Murphy, M.D., Liddle, C.L., &
Vitina, J.R. (1987). Frequency of Occurrence and the
criteria for automatic processing. Journal of
Experimental Psychology: Learning. Memory, and Cognition.
13, 241-250.
Schlotzhauer, S.D., & Littell, R.C. (1987). SAS System for
Elementary Statistical Analysis. Cary, North Carolina:
SAS Institute.
Schreiner, G.E. (1975). The search for the uremic toxin(s).
Kidney International. 7, S270-S271.
Schupak, E., Sullivan. J.F., & Lee, D.Y. (1967). Chronic
hemodialysis in "unselected" patients. Annals of Internal
Medicine. 67, 708-717.
Shiffrin, R. (1975). Short-term store: The basis for a
memory system. In F. Restle, R.M. Shiffrin, N.J.
Castellan, H.R. Lindman, & D.B. Pisoni (Eds.), Cognitive
Theory. 1, Hillsdale, New Jersey: Erlbaum.

173
Souheaver, G.T., Ryan, J.J., & DeWolfe, A.S. (1982).
Neuropsychological patterns in uremia. Journal of
Clinical Psychology. 38. 490-496.
Sperling, G. (1960). The information available in brief
visual presentations. Psychological Monographs. 74 (whole
no. 498).
Sprague, S.M., Corwin, H.L., Tanner, C.M., Wilson, R.S.,
Green, B.J., & Goetz, C.G. (1988). Relationship of
aluminum to neurocognitive dysfunction in chronic
dialysis patients. Archives of Internal Medicine. 148.
2169-2172.
Squire, L.R. (1987). Memory and Brain. New York, New York:
Oxford University Press.
Stewart, R.S., & Stewart, R.M. (1979). Neuropsychiatric
aspects of chronic renal disease. Psychosomatic. 20. 524-
527.
Teschan, P.E., Ginn, H.E., Bourne, J.R., & Ward, J.W.
(1976). Neurobehavioral responses to "middle molecule"
dialysis and transplantation. Transactions of American
Society of Artificial Internal Organs. 22. 190-194.
Teschan, P.E., Ginn, H.E., Walker, P.J., Bourne, J.R.,
Fristoe, M., & Ward, J.W. (1974). Quantified function in
uremic patients on maintenance dialysis. Transactions of
American Society of Artificial Internal Organs. 20, 388-
394.
Thorndike, E.L., & Lorge, I. (1944). The teacher/s word book
of 30.000 words. New York, New York: Teachers College.
Trieschmann, R.B., & Sand, P.L. (1971). WAIS and MMPI
correlates of increasing renal failure in adult medical
patients. Psychology Report. 29. 1251-1262.
Trompeter, R.S., Polinsky, M.S., Andreoli, S.A., & Fennell,
R.S. (1986). Neurological complications of kidney
failure. American Journal of Kidney Diseases. 7, 318-323.
Tucker, D.M., Sandstead, H.H., Penland, J.G., Dawson, S.L.,
& Milne, D.B. (1984). Iron status and brain functions:
serum ferritin levels associated with asymmetries of
cortical electrophysiology and cognitive performance.
American Journal of Clinical Nutrition. 39. 105-113.
Tyler, H.R. (1968). Neurologic disorders in renal failure.
American Journal of Medicine. 44., 734-748.

174
Uyeda, K.M., & Mandler, G. (1980). Prototypicality norms for
28 semantic categories. Behavior Research Methods &
Instrumentation. 12. 587-595.
Van Wyck, D.B. (1989). Iron deficiency in patients with
dialysis-associated anemia during erythropoietin
replacement therapy: Strategies for assessment and
management. Seminars in Nephrology. 9, 21-24.
Vander, A.J., Sherman, J.H., & Luciano, D.S. (1980). Human
Physiology. New York, New York: McGraw-Hill.
Waldmann, T.A., Levin, E.H., & Baldwin, M. (1961). The
association of polycythemia with a cerebellar
hemangioblastoma. American Journal of Medicine. 31. 318-
324.
Walter, T., Kovalskys, J., & Steel, A. (1983). Effect of
mild iron deficiency on infant mental development scores.
Journal of Pediatrics. 102. 519-522.
Wechsler, D. (1987). Wechsler Memory Scale Revised Manual.
New York, New York: The Psychological Corporation.
White, N., & Cunningham, W.R. (1988). Is Terminal Drop
Pervasive or Specific? Journal of Gerontology;
Psychological Sciences. 43. 141-144.
Williams, S.R. (1985). Nutrition and Diet Therapy (5th ed.).
St. Louis, Missouri: Times Mirror/Mosby.
Winearls, C.G., Oliver, D.O., Pippard, M.J., Reid, C.,
Downing, M.R., & Cotes, P.M. (1986). Effect of human
erythropoietin derived from recombinant DNA on the anemia
of patients maintained by chronic hemodialysis. Lancet.
2, 1175-1177.
Wintrobe, M.W., Lee, G.R., Boggs, D.R., Bithell, T.C.,
Foerster, J., Athens, J.W., & Lukens, J.N. (1981).
Clinical Hematology (4th ed.). Philadelphia,
Pennsylvania: Lea & Febiger.
Wolcott, D.L., Marsh, J.T., La-Rue, A., Carr, C., &
Nissenson, A.R. (1989). Recombinant erythropoietin
treatment may improve quality of life and cognitive
function in chronic hemodialysis patients. American
Journal of Kidney Disease. 14, 478-485.
Wolcott, D.L., Schweitzer, S., & Marsh, J.T. (1988).
Recombinant erythropoietin improves cognitive function
and quality of life of chronic hemodialysis patients.
Kidney International. 33. 242. Abstract.

175
Wolcott, D.L., Schweitzer, S., & Nissenson, A.R. (1989).
Recombinant erythropoietin (r-EPO) improves cognitive
function (CF) and quality of life (QL) of chronic
hemodialysis (CHD) patients. Kidney International. 35.
266. Abstract.
Yehuda, S., Youdim, M., & Mostofsky, D.I. (1986). Brain iron
deficiency causes reduced learning capacity in rats.
Pharmacology. Biochemistry and Behavior. 25, 141-144.
Youdim, M.B.H., & Ben-Schachar, D. (1987). Minimal brain
damage induced by early iron deficiency: modified
dopaminergic neurotransmission. Israel Journal of Medical
Sciences, 2_3, 19-23.
Ziestat, H.A. Jr., Logue, P.E., & McCarty, S.M. (1980).
Psychological measurement of memory deficits in dialysis
patients. Perceptual Motor Skills. 50, 311-318.

BIOGRAPHICAL SKETCH
Wayne L. Klein decided to become a psychologist in
adolescence after giving up on religion and philosophy. As
a college sophomore, academic psychology paled in comparison
to human potential psychology; therefore, in 1971, he
dropped out of SUNY at Albany and spent three months
hitchhiking to Berkeley, California where he spent a seminal
year emersed in Esalen groups, training at the Gestalt
Institute of San Francisco and working in the Berkeley Free
Clinic. After returning to SUNY at Albany to complete
liberal arts requirements, he transferred to SUNY Empire
State College, a university without walls program, and
obtained a B.S. in counselling psychology. Over the next
nine years he worked with a variety of populations in
diverse settings as a counselor and instructor. During that
time he also obtained a B.S. in biology by taking tests
(Regents College), took graduate courses in biopsychology
and earned an M.S. in educational psychology (SUNY at
Albany).
Spurred by interest in cognition-behavior-brain
relationships, he entered the University of Florida Clinical
and Health Psychology doctoral program. After flirting with
psychophysiology, he focused on clinical neuropsychology and
pursued interrelated research interests in the
176

177
neuropsychology of mineral deficiency and anemia,
chemosensation and food preferences, pica and factors
underlying anorexia. He also tested the claims that
subliminal audio tapes affect mood and engaged in test
construction.
His training in neuropsychology continues at the Tufts
University School of Medicine Boston V.A. Medical Center
Predoctoral Internship Consortium. Following internship he
will complete an N.I.H. Postdoctoral Fellowship at the
Boston University Aphasia Research Center. He has been
married since 1980 and has one daughter, who was born in
Gainesville.

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
filter Cb <~=X/<\/yviLL'
Eileen B. Fennell, Chairperson
Professor of Clinical and Health Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
¡mjMJLI a
/Russell M. Baue^f
Associate Professor of Clinical and Health
Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
A.
Hugh C.
Professo
of Clinical and Health Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
S3 ^4^
Thomas B Fast
Professor of Dentistry
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Robrt^S. Fennell, III
Professor Pharmacology and Therapeutics

This dissertation was submitted to the Graduate Faculty
of the College of Health Related Professions and to the
Graduate School and was accepted as partial fulfillment of
the requirements for the degree of Doctor of Philosophy.
August 1991
(flsUr^Q ft.
Dean, College of Health Related Professions
Dean, Graduate School

UNIVERSITY OF FLORIDA
1262 08554 8112



2
study was to test the possibility that a portion of the
neuropsychological dysfunction that has been attributed to
uremia is actually secondary to anemia.
A further rationale for the present study was to
address the broader question of whether cognitive deficits
are directly associated with chronic anemia. Severe acute
anemia produces serious symptoms of cerebral dysfunction,
but symptoms subside over time as adaptation occurs.
Although neuropsychological deficits have been well
documented in conditions of chronic anemia, the question of
whether the neuropsychological symptoms are caused by the
anemia has never been adequately addressed. The
neuropsychological study of anemia has been hampered by
difficulties in parceling out the effects of rise in
hemoglobin (Hg) from the effects of the agent causing the
rise. Chronic anemia is the final common result of numerous
causative factors including deficiencies of iron, folic acid
and cobalamin as well as conditions such as sickle cell
disease, all of which have been demonstrated to directly
affect the brain. In contrast, rEPO has been generally
considered to have no effect on the nervous system. This
putative disassociation between neurological and
hematological effects suggested that rEPO could potentially
be a valuable tool in the study of the neuropsychological
effects of chronic anemia.
Pilot work sampled a spectrum of cognitive functioning
before and during rEPO treatment. That work, although


V
California Verbal Learning Test 52
Levels of Processing Task 52
Frequency Estimation Task 53
Controlled Word Association Test 54
Cognitive-Affective and Physical Behavior
Questionnaire 54
Hemoglobin 55
Ferritin 55
Schedule of Testing 56
Statistical Methodology 58
Hypotheses 58
Main Hypothesis 58
Secondary Hypotheses 59
4 RESULTS 61
Overview of Analyses 61
Statistical Assumptions 64
Descriptive Statistics 69
Age 69
Education 69
Sex 69
Treatment Modality 69
Attrition 70
Time Intervals Between Assessments 71
Hemoglobin 72
Ferritin 75
Baseline Neuropsychological Functioning ... 79
Test of Main Hypothesis 82
Post Hoc Analyses 83
Learning Curve 83
Exploratory Analyses 85
Sum of CVLT Trials Four and Five 88
Short Term Memory 90
Verbal Fluency 90
Secondary Hypotheses 91
Levels of Processing 91
Estimation of Frequency of Occurrence .... 91
Quality of Life 91
Test Reliability 93
CVLT Alternate Form Reliability 93
Levels of Processing Reliability 98
Frequency of Occurrence Reliability 98
Physical Behavior Self-Report Test-Retest
Reliability 98
Cognitive-Affective Behavior Self-Report
Test-Retest Reliability 100
Post Hoc Analysis of Blood Urea Nitrogen . 100
Summary 101
5 DISCUSSION 104
Overview 104
Conclusion 121


11
insufficient processing during acquisition resulting in
inadequate encoding.
Based on the cognitive literature, it may be concluded
that LTM is powerfully affected by the manner and extent to
which the learner processes the information. If retrieval
is based on semantic cuing, then semantic processing will be
the most efficient route for encoding. The more elaborate
the processing, the higher the probability of effective
retrieval. Conditions reducing cognitive processing may be
expected to impair LTM. Cognitive processing is likely to
be negatively affected by factors including underarousal,
overarousal, fatigue, attentional deficits, distracting
stimuli, lack of interest and depression (Cohen,
Weingartner, Smallberg, Pickar, & Murphy, 1982).
Under some circumstances memory deficits may be
categorized as either primarily functional or organic.
Functional factors, those affecting cognitive processing,
may either reduce effort expended or increase effort
required. Functional memory deficits may, therefore, be
considered a product of insufficient or misdirected effort.
Several methods exist by which, under certain
circumstances, various types of memory dysfunction may
sometimes be differentiated. By controlling the duration
and nature of processing (i.e., semantic, phonological or
orthographic), the levels of processing paradigm provides a
means of determining whether LTM deficits are characterized
by a failure to benefit from semantic processing. Failure


3
suffering from small sample size, suggested that rise in Hg
was associated with improved long term memory (LTM) as
measured by performance on a multitrial verbal list learning
task (Klein et al., 1989). The present study sought to
confirm and extend these very preliminary findings.
An additional purpose of the present study was to generate
exploratory data addressing the nature of the mechanisms
underlying the putative improvement in LTM. In the pilot
study, despite sampling a variety of cognitive functions,
the only task that showed significant improvement in
performance was the task most sensitive to success in
actively organizing material into a meaningful structure.
Thus, it was hypothesized that the apparent improvement was
mediated by increased cognitive processing, possibly
secondary to reduced fatigue.
In the present study, paradigms from human memory
research were used in an attempt to tease out the level at
which putative change occurred. Improved automatic learning
and increased ability to benefit from semantic processing,
two tasks relatively unaffected by functional factors, would
have implicated physiological mechanisms affecting the
neural substrate of memory. Lack of improvement on these
tasks, in conjunction with improved effortful verbal LTM
performance, would have lent support to the hypothesis that
a functional mechanism such as increased cognitive
processing, possibly secondary to reduced fatigue, resulted
in the apparent improvement in LTM in the pilot study.


87
To test the effect of Hg changes within the range
closer to that likely to be considered physiologically
relevant, subjects with less than a 2 g/dl Hg change were
excluded from the final analysis. The two negative Hg
change subjects in the data set had a mean Hg drop change of
2.68 g/dl. The three positive subjects in the data set
exhibited a mean Hg rise of 3.77 g/dl. A GLM analysis of
this data set (n=5) failed to reveal a main effect for Hg
(F= 0.03, p=.8803); however, the small sample size greatly
reduced the power of the test and, again, the change was in
the opposite of the predicted direction. Mean CVLT
improvement was 2 words in the falling Hg group and the rise
was 1.33 words in the rising Hg group.
Hypothesizing that the observed effect of Hg change on
Sum CVLT might be mediated by affective change, the
Cognitive-Affective self-report variable was included in a
GLM analysis of variance model. A multivariate analysis
demonstrated main effects for both direction of Hg change
(F=13.27, p=.0010) and Cognitive-Affective self-report
(F=8.29, p=.0072). Further analysis revealed a main effect
for direction of Hg change (F=16.97, p=.0003) and an
interaction between direction of Hg change and self-reported
Cognitive-Affective behavior (F=6.53, p=.0043). When
Cognitive-Affective behavior was included in this model it
failed to reach significance (F=1.88, p=.1807). Further
analyses suggested no relationship between Cognitive-
Affective behavior and sign of Hg change (r=.11056,



PAGE 1

EFFECT OF RECOMBINANT ERYTHROPOIETIN MEDIATED CHANGES IN ANEMIA ON VERBAL MEMORY IN ADULT DIALYSIS PATIENTS BY WAYNE L. KLEIN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1991

PAGE 2

ACKNOWLEDGEMENTS An elderly subject stood out for interest in his performance. A man in his early twenties appeared remarkably uninterested in his standing. A third subject participated in the hope of helping others, despite increasing debilitation. All three died prior to completion of the study. Their names, and the names of the 32 other subjects are not listed out of respect for their privacy. However, without their willingness to endure testing despite the discomforts, and in some cases the agonies, of their illness, this study would not have been possible. The following, undergraduates at the time, assisted with test construction, test validation, data collection and/or data entry: Bill Dorkowsky, Mike Reiter, Sheri Scott, Katherine Cobb, Edward Suarez and Wylene Bhanji. Without their help my wife, Liz, and child, Sarah, would have suffered even worse neglect. Don Mars, M.D. graciously persuaded his patients to participate as controls. The American Psychological Association provided a Dissertation Research Award which partially funded this project. My committee, beyond help regarding this project, consists of those who were most important to my graduate ii \

PAGE 3

• • • 111 experience. Eileen Fennell, my chair, enabled me to meld my interests with the do-able and the acceptable. Without her the going would have been much less pleasant. It was on Russ Bauer's clinic day that I decided to specialize in clinical neuropsychology. Hugh Davis is forever and agreeably etched in my subconscious. Bob Fennell provided research opportunities with import to me well beyond the findings. Tom Fast became a close friend who taught many things . H

PAGE 4

TABLE OF CONTENTS Page ACKNOWLEDGEMENTS ii ABSTRACT vii CHAPTERS 1 INTRODUCTION: FOCUS AND RATIONALE 1 2 LITERATURE REVIEW 5 Introduction 5 Overview of Verbal Memory 5 Memory Stages and Levels of Processing ... 6 Automatic and Effortful Memory 12 Neuropsychological Dysfunction in Adult Renal Patients 13 General Intelligence: Verbal and Nonverbal . 13 Attention, Level of Activation and Executive Function 14 Verbal Memory 16 Nonverbal Memory 20 Conclusions Regarding Cognitive Deficits in Renal Disease 21 Overview of Anemia 24 Prevalence and Etiology 24 Function of Hemoglobin and Adaptation to Anemia 26 Erythropoietin Therapy and Iron Deficiency . 27 Neuropsychology of Anemia 27 Effects of Anemia 29 Possible Mechanisms 32 Effect of Erythropoietin on Neuropsychological Function 35 Effect of Erythropoietin on Quality of Life . 36 Previous Neuropsychological Studies 37 Results of Pilot Study 38 Negative Effects 44 Summary and Conclusions 4 6 3 METHODS 50 Subjects 50 Measures 52 iv

PAGE 5

California Verbal Learning Test 52 : Levels of Processing Task 52 Frequency Estimation Task 53 Controlled Word Association Test 54 Cognitive-Affective and Physical Behavior Questionnaire 54 '' Hemoglobin 55 Ferritin 55 Schedule of Testing 56 ^ Statistical Methodology 58 "s Hypotheses 58 * Main Hypothesis 58 ^ Secondary Hypotheses 59 "' 4 RESULTS 61 Overview of Analyses . 61 Statistical Assumptions 64 i Descriptive Statistics 69 Age 69 ^ Education 69 *| Sex 69 i Treatment Modality 69 J Attrition 70 ^ Time Intervals Between Assessments 71 ] Hemoglobin 72 Ferritin 75 Baseline Neuropsychological Functioning ... 79 Test of Main Hypothesis 82 Post Hoc Analyses 83 Learning Curve 83 Exploratory Analyses 85 Sum of CVLT Trials Four and Five 88 ' Short Term Memory 90 Verbal Fluency 90 Secondary Hypotheses 91 Levels of Processing 91 Estimation of Frequency of Occurrence .... 91 Quality of Life 91 i Test Reliability 93 | CVLT Alternate Form Reliability 93 i Levels of Processing Reliability 98 ] Frequency of Occurrence Reliability 98 i Physical Behavior Self-Report Test-Retest 1 Reliability 98 ] Cognitive-Affective Behavior Self -Report .) Test-Retest Reliability 100 Post Hoc Analysis of Blood Urea Nitrogen . . . 100 Summary lOi ! 5 DISCUSSION 104 Overview 104 Conclusion 121

PAGE 6

VI APPENDICES A DEVELOPMENT AND VALIDATION OF FORM III OF THE CALIFORNIA VERBAL LEARNING TEST 124 B LEVELS OF PROCESSING VERSIONS II AND III ... 131 C FREQUENCY ESTIMATION TASK 139 D COGNITIVE-AFFECTIVE AND PHYSICAL BEHAVIOR QUESTIONNAIRE 141 E COMPARISON OF PARAMETRIC AND NONPARAMETRIC TESTS OF BASELINE MEASURES BETWEEN GROUPS . . 144 F NORMALITY AND HOMOGENEITY OF VARIANCE . . . .14 6 6 GROUP MEANS AND STANDARD DEVIATIONS 149 H CHANGE SCORES 155 I RAW DATA BY SUBJECT 158 BIBLIOGRAPHY 162 BIOGRAPHICAL SKETCH .... 175 J i i

PAGE 7

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy I EFFECT OF RECOMBINANT ERYTHROPOIETIN MEDIATED CHANGES IN ANEMIA ON VERBAL MEMORY IN ADULT DIALYSIS PATIENTS \ \ Wayne L. Klein I August 1991 ., Chairperson: Eileen B. Fennell \ Major Department: Clinical and Health Psychology "\ Cognitive dysfunction has been documented in many l . . \ chronic anemic conditions; however, the putative effects of j 1 anemia have never been disassociated from the cause of the i anemia. Neuropsychological dysfunction of poorly understood •' origin and severe chronic anemia are nearly universal consequences of chronic renal failure. A small pilot study demonstrated a positive relationship between a recombinant erythropoietin (rEPO) mediated mean rise of 5.46 g/dl of hemoglobin (Hg) and the Sum of trials 1-5 on the California Verbal Learning Test (Sura CVLT) . Sum CVLT performance was hypothesized to be enhanced by increased effortful processing secondary to reduced fatigue. A rEPO treatment group (n=17) and a control group (n=18) underwent three assessments with three alternate forms of the CVLT, one of which was developed for this Vll

PAGE 8

Vr ;: . :i?t Vlll study. Also administered were a survey of cognitiveaffective and physical behavior, a verbal fluency task and three alternate forms of two measures likely to be sensitive to hypoxia but not fatigue, the levels of processing and frequency estimation tasks. All subjects were on maintenance dialysis. Mean treatment group Hg at the first assessment was 8.46 g/dl (sd=1.42) and 9.44 g/dl (sd=1.92) in the controls. The rise of 1.34 g/dl Hg between groups in the 54 day mean interval between the first and second assessments was statistically significant (F=7.42, e=.0102), but of doubtful physiological significance. In the 90 days between the second and third assessments Hg was unchanged. The relationship between rise in Hg and Sum CVLT failed to reach statistical significance (F=3.25, e=.0805). Post hoc reclassification of subjects based on direction of Hg change unexpectedly revealed an apparent inverse relationship between Hg and Sum CVLT. Subjects, primarily controls, exhibiting a drop in Hg between the first and second assessments performed below expectations at the first assessment (F=7.82, p=.0027). Whether this was spurious or due to the operation of an unknown factor remains speculative. Hemoglobin related changes in self-reported cognitive-affective and physical behavior were not observed. Current understanding of the relationship between anemia and neuropsychological dysfunction appears insufficient to inform decisions regarding rEPO treatment.

PAGE 9

CHAPTER 1 INTRODUCTION: FOCUS AND RATIONALE The fourth leading health problem in the developed world, renal disease has been estimated to affect approximately 8 million people in the U.S. (Williams, 1985). Uremia, the life threatening condition resulting from renal failure, has been known since ancient times. A variety of symptoms have been associated with uremia, the majority of which have been reported to be secondary to nervous system dysfunction (Ginn et al., 1975). Memory dysfunction has been among the neuropsychological deficits commonly noted. Despite considerable effort, attempts to identify the factors responsible for these deficits have achieved little success (Powell et al., 1986; Schreiner, 1975; Trompeter, Polinsky, Andreoli, & Fennell, 1986) . Chronic anemia has been almost universally observed in chronic renal failure. The primary cause has been generally accepted to be insufficient production of erythropoietin (EPO) , a erythropoiesis stimulating factor produced by the kidneys. Recent studies have documented the dramatic effects of recombinant erythropoietin (rEPO) on many physical symptoms until recently attributed to uremia, but now demonstrated to have been at least partially caused by anemia (Nissenson, 1989) . One rationale for the present 1

PAGE 10

study was to test the possibility that a portion of the neuropsychological dysfunction that has been attributed to uremia is actually secondary to anemia. A further rationale for the present study was to address the broader question of whether cognitive deficits are directly associated with chronic anemia. Severe acute anemia produces serious symptoms of cerebral dysfunction, but symptoms subside over time as adaptation occurs. Although neuropsychological deficits have been well documented in conditions of chronic anemia, the question of whether the neuropsychological symptoms are caused by the anemia has never been adequately addressed. The neuropsychological study of anemia has been hampered by difficulties in parceling out the effects of rise in hemoglobin (Hg) from the effects of the agent causing the rise. Chronic anemia is the final common result of numerous causative factors including deficiencies of iron, folic acid and cobalamin as well as conditions such as sickle cell disease, all of which have been demonstrated to directly affect the brain. In contrast, rEPO has been generally considered to have no effect on the nervous system. This putative disassociation between neurological and hematological effects suggested that rEPO could potentially be a valuable tool in the study of the neuropsychological effects of chronic anemia. Pilot work sampled a spectrum of cognitive functioning before and during rEPO treatment. That work, although

PAGE 11

suffering from small sample size, suggested that rise in Hg was associated with improved long term memory (LTM) as measured by performance on a multitrial verbal list learning task (Klein et al., 1989). The present study sought to confirm and extend these very preliminary findings. An additional purpose of the present study was to generate exploratory data addressing the nature of the mechanisms underlying the putative improvement in LTM. In the pilot study, despite sampling a variety of cognitive functions, the only task that showed significant improvement in performance was the task most sensitive to success in actively organizing material into a meaningful structure. Thus, it was hypothesized that the apparent improvement was mediated by increased cognitive processing, possibly secondary to reduced fatigue. In the present study, paradigms from human memory research were used in an attempt to tease out the level at which putative change occurred. Improved automatic learning and increased ability to benefit from semantic processing, two tasks relatively unaffected by functional factors, would have implicated physiological mechanisms affecting the neural substrate of memory. Lack of improvement on these tasks, in conjunction with improved effortful verbal LTM performance, would have lent support to the hypothesis that a functional mechanism such as increased cognitive processing, possibly secondary to reduced fatigue, resulted in the apparent improvement in LTM in the pilot study.

PAGE 12

Chronic anemia has been reported to have far reaching effects on the organism. In loose association with the severity of the anemia, symptoms may include fatigue, hypoxia, hypocapnea, insomnia, enhanced cardiac output and increased sympathetic activation, reduced appetite, altered blood chemistry, modified lifestyle including the inability to hold gainful employment and, probably, decreased selfesteem. Therefore, determination of the relative contributions of possible underlying mechanisms was considered impossible in a single study of this scope. However, it was hoped that this study would extend our understanding of the neuropsychology of chronic anemia and the clinical significance of chronic anemia with regard to the etiology of the cognitive deficits observed in renal disease.

PAGE 13

CHAPTER 2 LITERATURE REVIEW Introduction Investigation of the neuropsychology of uremic anemia is necessarily multidisciplinary. The logic underlying the rationale and design of this study is based on information derived from five somewhat distinct areas of investigation. The areas are 1) human memory, 2) clinical studies of neuropsychological function in adults suffering from renal disease, especially with regard to verbal memory, 3) the physiology of anemia, 4) the neuropsychology of anemia and 5) the effects of rEPO mediated improvement in anemia in adult renal patients on neuropsychological function. Each of these will be briefly reviewed in this chapter. Overview of Verbal Memory This section briefly overviews current thinking with regard to memory as it relates to the clinical assessment and interpretation of memory function. The application of memory paradigms to the study of the neuropsychology of anemia is discussed. Mounting neurochemical, functional and anatomical evidence suggests that memory may be classified into procedural memory (i.e., motor memory or skill learning) and declarative memory for facts (Nissen, Knopman, & Schacter, 1987) . Memory may, to some extent, be divided 5

PAGE 14

into automatic and the more commonly studied effortful, deliberate or consciously encoded memory (Hasher & Zacks, 1984; Newman, Weingartner, Smallberg, & Calne, 1984; Sanders, Gonzalez, Murphy, Liddle, & Vitina, 1987). Automatic declarative memory may be further differentiated based on the modality of encoding and retrieval, with the most studied modalities being verbal and visual. Verbal memory may also be partitioned into semantic and episodic memory, the former being dissociated from, and the later bound within, the context of learning (Newman, Weingartner, Smallberg, & Calne, 1984). Memory Stages and Levels of Processing Most paradigms appear to demonstrate the division of effortful declarative verbal memory into relatively clearcut temporal categories or stages. The cognitive, neuropsychological and neuroscience traditions have posited long and short term memory stages. The definition of memory stages is dependent on the level of analysis (Squire, 1987) . Cognitive science has classified auditory memory of a duration exceeding about 15 seconds as long term memory (LTM) while neuroscience considers this squarely within the domain of short term memory (STM) . These differences reflect the criteria by which stages have been determined within the two traditions. Within the cognitive tradition a succession of models (e.g., capacity, multistore and levels of processing) has stimulated research. In attempting to map memory function

PAGE 15

onto biological structures and processes, the neuroscience perspective has focused both on the nature of the memory trace and the physical location of memory in the brain. This brief overview will focus primarily on a behavioral (i.e., cognitive and neuropsychological) level of analysis as it relates to neuropsychological assessment. As conceptualized in the multistore model, memory is a complex series of distinct stages through which information is successively processed (Atkinson & Shiffrin, 1968, 1971). These stages differ along several dimensions including capacity, duration and the nature of the attendant psychological processes. The three generally recognized stages are ultrashort or modality-specific sensory memory (e.g., echoic and iconic memory), immediate or STM and LTM. Sperling (1960) demonstrated that iconic (i.e., visual sensory) memory decays after about one second. Echoic memory may take several seconds to decay (Darwin, Turvey, & Crowder, 1972) . Sensory memory capacity appears to be around nine items (Sperling, 1960) . The duration of STM when rehearsal (i.e., replenishment) is prevented has been shown to be about 15 seconds (Peterson & Peterson, 1959) . Thus, the repetition of a list of orally presented digits or words utilizes both echoic memory and STM. However, in the case of oral digit span, as traditionally administered, the limiting factor is likely to be attentional. Oral digit span is the most commonly used measure of immediate memory span; however, it is frequently insensitive

PAGE 16

8 to STM recall deficits (Lezak, 1983) . In a series of factor analyses, Wechsler (1987) found digit span to consistently load primarily on attention-concentration rather than on a memory factor. Unpublished data communicated by T. White (June 12, 1990) revealed that for 182 patients seen in a neuropsychology clinic, there was an insignificant correlation (r= .123, e='098) between digits forward and trial one of the CVLT, a measure of STM recall. The capacity of STM has been shown to be approximately 7 +2 chunks of information (Miller, 1956) . Unfortunately the notion of chunks has not been clearly defined. Recall significantly exceeding the accepted capacity of 7 +2 has been attributed either to the participation of LTM or efficient organization of bits of information into larger chunks. There are no known limits on the duration or capacity of LTM. An inherent feature of multistore models has been the necessity of elucidating the processes through which information is progressively transferred through the memory storage areas. The mechanism underlying transfer of information from sensory memory to STM has been attributed to attentional processes associated with pattern (i.e., meaning) recognition (Moray, 1959) . In contrast, Shiffrin (1975) has proposed that perceptual stimuli are automatically encoded and passed from sensory to STM. From a multistore perspective, transfer from STM to LTM has been viewed as a function of active attention or rehearsal

PAGE 17

(Rundus, 1971) . This model explains the observation that, on a supraspan task, a higher percentage of the first (primacy) and last (recency) words tend to be recalled. The primacy effect is believed to be due to encoding in LTM. The recency effect is hypothesized to be a manifestation of STM. Thus, immediate recall on a supra-span task may be examined for primacy and recency effects to provide indications of the integrity of STM and LTM. Besides dissimilarities in capacity and duration, there appear to be differences in the cognitive processes requisite to the maintenance of STM and LTM. In general, most studies and clinical observations have suggested that phonological or maintenance rehearsal (i.e., mere parroting) usually serves only to maintain information in STM, whereas more complex, elaborative or semantic processing facilitates transfer to LTM (Craik & Lockhart, 1972) . Transfer to LTM has been thought to usually require more active attending and processing. Demonstration of the differential effects of type or level of processing led to the proposal that the observation of apparently distinct memory stores could be explained in terms of levels of processing within one memory store (Craik & Lockhart, 1972) . The apparently limited storage capacity of STM has been reinterpreted by the levels of processing model as limited processing capacity. The apparent distinction between STM and LTM has been explained by the levels of processing model as partially a function of depth

PAGE 18

10 of processing, with semantic processing occurring at greater depth than phonemic processing. However, research in other areas has forced revisions in the original levels of processing models (Cermak, 1982) . To account for studies demonstrating elaborateness of processing to be more salient than type of processing (e.g., semantic vs. phonological) , the levels of processing model has been modified to emphasize extent rather than depth of encoding (Craik & Tulving, 1975) . Confounding all models, other work has shown that shallowly processed information may at times interfere with and even outlast more deeply processed material (Cermak, 1982) . In addition, phonemic processing may result in recall superior to semantic processing when retrieval is phonemic, that is, when processing and retrieval are within the same domain (Morris, Bransford, & Franks, 1977) . Data in STM is easily lost through interference and decay. In contrast, reflecting the enduring nature of LTM, failures of LTM have usually been considered to be produced by retrieval deficits. The continued presence of the memory may be assessed with recognition tasks. Recall tasks require self-generation of cues. In contrast, in recognition tasks cuing is externally provided. Thus, recognition in the absence of spontaneous recall may be explained by weak or decayed memory traces, failure to generate adequate self-cuing or memory search strategies or

PAGE 19

^ 11 insufficient processing during acquisition resulting in inadequate encoding. Based on the cognitive literature, it may be concluded that LTM is powerfully affected by the manner and extent to which the learner processes the information. If retrieval is based on semantic cuing, then semantic processing will be the most efficient route for encoding. The more elaborate the processing, the higher the probability of effective retrieval. Conditions reducing cognitive processing may be expected to impair LTM. Cognitive processing is likely to be negatively affected by factors including underarousal, overarousal, fatigue, attentional deficits, distracting stimuli, lack of interest and depression (Cohen, Weingartner, Smallberg, Pickar, & Murphy, 1982) . Under some circumstances memory deficits may be categorized as either primarily functional or organic. Functional factors, those affecting cognitive processing, may either reduce effort expended or increase effort required. Functional memory deficits may, therefore, be considered a product of insufficient or misdirected effort. Several methods exist by which, under certain circumstances, various types of memory dysfunction may sometimes be differentiated. By controlling the duration and nature of processing (i.e., semantic, phonological or orthographic) , the levels of processing paradigm provides a means of determining whether LTM deficits are characterized by a failure to benefit from semantic processing. Failure

PAGE 20

fr. 12 to benefit from semantic processing has been attributed to encoding deficits secondary to organic dysfunction of memory structures (e.g., Korsakoff's). Automatic and Effortful Memory Automatic memory is demonstrated by recall in the absence of effortful encoding. Automatic memory is unconscious and little affected by degree of effort (Jonides & Naveh-Benjamin, 1987) . Deficits in automatic memory processes are suggestive of dysfunction at the level of organic memory structures. Reductions in capacity such as illness and, presumably, fatigue, have limited effect on automatic memory (Hasher & Zacks, 1984) . The frequency of occurrence task provides a means of assessing automatic memory function. Normal performance on a frequency of occurrence task in conjunction with reduced performance on an effortful task requiring effort for optimal performance would suggest disruption at the level of effortful processing. This pattern of deficits could be the result of frontal lobe dysfunction, poor motivation or failure to employ an effective strategy. One caveat, the use of automatic memory measures to assess the integrity of the neural substrates of effortful memory is predicated on the assumption, not accepted by all, that effortful and automatic memory are largely dependent on identical neuronal systems (Roediger, 1990) .

PAGE 21

13 Neuropsychological Dysfunction in Adult Renal Patients This section provides a cursory overview of the general pattern of deficits seen in renal disease and discusses verbal memory within the context of broader findings. General Intelligence: Verbal and Nonverbal In the absence of confounds, general verbal intellectual ability has been reported to be relatively spared in adult renal disease (Blatt & Tsushima, 1966; Comty, Leonard, & Shapiro, 1974; English et al., 1978; Fishman & Schneider, 1972; Freeman, Sherrard, Carlsyn, & Paige, 1980; Ryan, Souheaver, & DeWolfe, 1980; Schupak, Sullivan, & Lee, 1967; Trieschmann & Sand, 1971). In a 1982 review, Osberg, Meares, McKee and Burnett noted that in every reported study, verbal IQ scores surpassed performance scores by 5 to 14 points. The verbal -nonverbal discrepancy is partly an artifact of the time constraints on many nonverbal but not verbal intelligence subtests; however, visuospatial and visuomotor deficits have also been widely observed on untimed tasks (Mings, 1987) . The relative sparing of language function is consistent with the general pattern of deficits often seen in diffuse brain dysfunction. Speculations regarding the cause of this verbal-nonverbal disassociation have included: the nondominant hemisphere is more vulnerable to assault, overlearned verbal abilities are more redundantly represented and crystallized intellectual functions tax cognitive capacity less than more novel visuospatial tasks.

PAGE 22

14 Interestingly, McDaniel's (1971) data on visual discrimination performance suggested impairment not primarily as a result of difficulties with visual-motor integration, but secondary to interference with cognitive processing. The reported visuospatial motor dysfunctions may be partially mediated by attentional deficits (Fennell, Fennell, Mings, & Morris, 1986; McDaniel, 1971). Attention. Level of Activation and Executive Function Several authors have suggested the presence of global attention deficits in uremia (Marshall, 1979; Stewart & Stewart, 1979) . Data reported by Trieschmann and Sand (1971) on 83 subjects not treated with dialysis demonstrated reduced performance on two measures loading on simple attention and concentration, the Digit Span and Arithmetic subtests of the Wechsler Adult Intelligence Scale (WAIS) . Grouping subjects based on severity of illness revealed Digit Span to be the lowest subscale score in the more severely ill group. In a dialized population with normal premorbid verbal IQ based on vocabulary knowledge, English, Savage, Britton, Ward and Kerr (1978) found reduced performance on WAIS subtests most heavily loading on attention (i.e.. Arithmetic, Digit Span, Digit Symbol) . In a pediatric population. Digit Span was reported to be significantly reduced regardless of treatment modality. Indeed, in a pediatric population, excluding the renal transplant group, Digit Span discriminated between renal and control groups J

PAGE 23

15 better than any other measure in the study ( Fennel 1, et al. ,1987) , In contrast, Souheaver, Ryan and DeWolfe (1982) found spared auditory attention skills, but on a different task, the Seashore Rhythm test. In the Heilman, Moyer, Melendez, Schwartz and Miller (1975) study, simple attentionconcentration as measured by Digit Span was within normal limits (mean subscale score of 10.04 +2.18) and consistent with verbal intelligence as measured by the Vocabulary subtest (mean 10.00). However, attention was significantly lower in renal patients (n=24) than in controls (n=12). Overall, it may be concluded that attentional deficits are common in uremia. Uremia appears to depress activation and arousal (Trompeter, Polinsky, Andreoli, & Fennell, 1986). This underarousal may interfere with attention and increase distract ibility (Heilman, Moyer, Melendez, Schwartz, & Miller, 1975) . However, reduced level of activation is unlikely to be the only mechanism involved. Assessments at varying time intervals after hemodialysis sessions revealed no change in Digit Span performance, despite changes in speed and reaction time (Lewis, O'Neill, Dustman & Beck, 1980) . Renal disease also appears to negatively affect more complex attentional and executive functioning such as the ability to rapidly generate and/or shift sets. Performance on Trails B, a task requiring maintenance of and alternation

PAGE 24

16 between two sequences, speeded visual search and simple visual motor coordination, has been reported to be reduced, at times severely, in renal disease (Heilman, Moyer, Melendez, Schwartz, & Miller, 1975; Ratner, Adams, Levin, & Rourke, 1983; Teschan et al., 1974). Ginn (1975) reported a correlation of .82 between Trails and severity of renal failure as measured by serum creatinine for subjects not on dialysis. Similar results were reported by Teschan and colleagues (1974) . Reduced performance on tasks demanding sustained vigilance, such as continuous performance tasks, has been consistently reported (Ginn, 1975; Osberg, Meares, McKee, & Burnett, 1982) . Ginn and colleagues (1978) demonstrated worsening of performance on a vigilance task in 8 out of 10 subjects following a reduction in hemodialysis. Contributions to executive system dysfunction may include deficits secondary to reduced level of activation, diminished attentional capacity, susceptibility of the frontal-subcortical neural substrate to metabolic abnormalities and dysfunction stemming from a generalized difficulty with tasks demanding fluid as opposed to crystalized intelligence. Verbal Memory Reduced vigilance, underarousal , increased distractibility, increased response latencies and reduced cognitive processing may contribute to global memory deficits. Work in a pediatric renal population has revealed a correlation between learning on a multitrial supra-span

PAGE 25

17 verbal task requiring the learning of a list (e.g., Buschke), susceptibility to distraction (i.e.. Auditory Consonant Trigrams) and reduced sustained vigilance (Fennell, Fennell, Mings, & Morris, 1986). This led to the suggestion that reduced level of arousal may mediate STM impairment in renal disease (Fennell et al., 1990a). Because they place greater demands on sustained attention, multitrial memory tasks may be more sensitive to uremia. Overall, initiation of hemodialysis has been reported to improve memory function (Osberg, Meares, McKee, & Burnett, 1982) . In adults, the consequences of uremia are largely reversible and appear to be only minimally cumulative in the absence of confounds such as aluminum toxicity, dietary noncompliance, uncontrolled hypertension or diabetes. Comparing dialysis patients and controls. Hart, Pederson, Czerwinski and Adams (1983) found relatively little difference in memory as measured by the Wechsler Memory Scale (WMS) . In addition, no significant correlations were found between years of dialysis treatment and memory performance. Gilli and DeBastiani (1983) reported (n=54) a mild relationship between duration of hemodialysis and decrements in verbal intellectual ability (WAIS) and memory (WMS) . The WMS Memory Quotient, a composite measure of primarily short term verbal and nonverbal memory, was initially above expectations based on WAIS performance. For the 21 hemodialysis subjects who were retested, a minimum of 12

PAGE 26

18 months later the WMS Memory Quotient had dropped to a level only slightly above expectations for IQ. The authors interpreted these results as suggesting a decline in memory function in association with time on hemodialysis, perhaps related to elevated parathyroid hormone levels. However, fewer than half of the subjects were retested, and they exhibited initial Memory Quotient scores above the average for the entire subject pool and, as previously noted, above expectations for IQ. The apparent decline in memory may have been an example of regression towards the mean. Ginn (1975) reported a temporary improvement in verbal LTM on a recognition task the day after hemodialysis, regardless of the level of performance prior to dialysis. A correlation of .687 (e<.001) was reported between latency to response on a single trial word-recognition task and severity of uremia (Ginn et al., 1975). However, increased response latency does not necessarily indicate retrieval difficulties, especially in a population suspected of generalized slowing. Heilman, Moyer, Melendez, Schwartz and Miller (1975) found significant verbal STM deficits on the Logical Memory subtest of the WMS. In contrast, loss between immediate and delay was comparable to controls. Ginn and colleagues (1975) reported a significant negative correlation (e<.001) between response latency on a verbal recognition memory task and uremia as measured by serum creatinine. In nondialized uremic subjects (n=23) participating in a longitudinal

PAGE 27

19 study, Hagberg (1974) reported verbal STM as measured by paired associate learning to be significantly lower than that predicted by general verbal ability in 23 nondialized uremic subjects. Six months after initiation of hemodialysis, paired associate learning performance had significantly improved in the 21 subjects retested and approached expected levels relative to intelligence. This study used alternate test forms but failed to use a control group to control for possible effects of repeat testing. Assessments of six subjects at varying time intervals following hemodialysis revealed no changes in STM as measured by a paired associate learning task, despite changes in performance on timed tasks (Lewis, O'Neill, Dustman, & Beck, 1980) . In contrast, on a recognition task, STM was shown to improve 24 hours after and deteriorate two days following dialysis (Teschan et al., 1974). Other work by this group has suggested improvement in verbal STM following transplantation, but this has not reached statistical significance (Teschan, Ginn, Bourne, & Ward, 1976) . Based largely on work with a pediatric population, it has been suggested that STM and sensory/motor function are more severely impacted than are more complex cognitive abilities or LTM (Crittenden, Holliday, Piel, & Potter 1985) . However, this report failed to provide sufficient data to support this suggestion. It appears likely that the measure of STM was Digit Span from the WAIS-R, which is more

PAGE 28

KB20 of an attentional measure. Other work generally supports the notion of a differential susceptibility for attentional and visuospatial constructional functioning. Also in the child literature, STM and LTM deficits have been reported on multitrial supra-span memory tasks, with impairment in normal developmental improvement in memory (Fennell, et al., 1990a) . Nonverbal Memory Visual memory as measured by the Benton Visual Retention Test was impaired in nondialized renal subjects (n=12), but normalized following 12 months of maintenance hemodialysis, although the change was not significant (Hagberg, 1974) . In a related study, 20 adults, chronically dialized for a mean of 39.7 months, were followed throughout the hemodialysis cycle. Performance on the Benton Visual Retention Test was mildly impaired on three separate administrations, despite unimpaired ability to copy the drawings (Ratner, Adams, Levin, & Rourke, 1983) . Facial recognition memory as measured by Milner Faces was reported to be unimpaired in nondialized renal patients (Heilman, Moyer, Melendez, Schwartz, & Miller, 1975). Memory For Designs was shown by Hagberg (1974) to be within the normal range for 23 nondialized uremic subjects. Illustrating the possible value of longitudinal work, as well as the need for alternate forms, readministration of the same test form to 21 of the original subjects, following

PAGE 29

21 6 months of hemodialysis, revealed a very slight nonsignificant improvement. In hemodialysis subjects (n=29) , impairment in visuospatial memory as measured by the Block Design Learning Test, in a study lacking a control group, failed to reach statistical significance. However, duration of dialysis did correlate with poor performance (English, Savage, Britton, Ward, & Kerr, 1978) . However, it is unclear whether the putative deficit was related to nonverbal memory or to visuospatial constructional ability. In addition, factors such as hypertension, diabetes and age may have contributed to the apparent negative relationship between duration of dialysis and nonverbal memory performance. Ziestat, Logue and McCarty (1980) reported a significant correlation between years on hemodialysis and both short term (r=-.39) and long term (r=-.38) visual memory as measured by the WMS. However, again, years on dialysis may have been confounded by factors such as etiology of renal failure. In contrast, verbal memory showed no relationship with years on hemodialysis in this study. The apparent decline in visual memory may have been secondary to the well researched decline in visuospatial processing (Mings, 1987) . Conclusions Regarding Cognitive Deficits in Renal Disease Renal disease differentially affects cognitive functioning in a manner grossly consistent with that seen in many systemic conditions. The degree of cognitive

PAGE 30

22 impairment correlates with severity of uremia, is at times dramatically improved by initiation of dialysis and may be nearly reversed by successful transplantation (Fennell, Rasbury, Fennell, & Morris 1984; Ratner, Adams, Levin, & Rourke, 1983) . Adults on maintenance dialysis frequently exhibit deficits in attention, vigilance, reaction time, ability to rapidly generate and/or shift sets, visuospatial ability, visual memory and verbal memory. Weaker evidence suggests variability correlating with the phase of the dialysis cycle, a possible advantage for peritoneal dialysis over hemodialysis and decreased performance in association with time on hemodialysis (Fennell, Fennell, Mings, & Morris, 1986; Osberg, Meares, McKee, & Burnett, 1982) . Higher Hg levels in peritoneal dialysis compared to hemodialysis and the possibility of increasing anemia in conjunction with time on hemodialysis leave open the possibility that level of anemia may play a role in the latter two observations. Restoration of near normal cognitive function following transplantation in well dialyzed subjects may also, to an extent yet to be determined, be mediated by the amelioration of anemia. There are a number of problems with this literature. For example, small sample size, absence of alternate test forms, lack of relevant control groups, and inadequate statistical procedures (Osberg, Meares, McKee, & Burnett, 1982) . Additional contributions to discrepancies in the literature include high variance in subject characteristics

PAGE 31

23 including differences in medical management, age, age of onset and etiology of renal disease. Many of the conditions such as diabetes, hypertension and cardiovascular disease contributing to or associated with renal failure have neuropsychological consequences in their own right. Discrepant findings in hemodialysis patients may to some extent be a function of the time of assessment relative to dialysis, the adequacy of dialysis and the degree of electrolyte disequilibrium following dialysis. Most of the early studies sampled neuropsychological function at one point in time. There have been very few longitudinal studies. Progress in the treatment of renal disease has been charted in the literature on the neuropsychology of uremia. Prior to the availability of dialysis, the neuropsychological consequences of renal failure often included stupor and coma (Arieff, Guisado, & Massry, 1975; Tyler, 1968) . Prior to the discovery and acceptance of the neurotoxicity of elevated blood aluminum levels in renal patients, dialysis encephalopathy affected a significant percentage of patients (Sprague et al., 1988). Improvements in the methodology of dialysis as well as other aspects of medical management continue to improve the physiological conditions under which uremic nervous systems function. Concurrently, these advances have also resulted in longer durations of exposure to uremia as well as an aging renal population. Thus, discrepancies in the literature may to

PAGE 32

!» 24 some extent be attributed to the differential effects of the passage of time on technology and on individuals. Overview of Anemia This section will provide a brief overview of the significance and etiology of anemia, adaptive mechanisms and the treatment of uremic anemia. Prevalence and Etiology The world-wide prevalence of anemia has been estimated to be 40% in children, 35% in adult females and 20% in adult males (Lozoff , 1989) . The defining feature of anemia is a reduction in blood Hg levels. Anemia is the final common pathway for a large number of diseases, conditions resulting in blood loss and deficiency states, all of which result in decreased Hg in the blood (Bunn, 1980a) . The most common form of anemia, iron deficiency anemia has been reported to have a prevalence in the U.S. of 20% in adult women of childbearing years, 50% in pregnant women and 3% in adult males (Lee, Wintrobe, & Bunn, 1980) . Probably the second most common type, the anemia of chronic disorders is the mild to moderate anemia frequently associated with chronic inflammatory syndromes and infectious and neoplastic diseases (Wintrobe et al., 1981). Anemia is a very common condition. Anemia almost invariably accompanies chronic renal failure (Desforges, 1975; Erslev, 1975). Uremic anemia often significantly reduces quality of life, despite adequate dialysis. Prior to the development of rEPO,

PAGE 33

25 approximately 25% of dialysis patients suffered from anemia severe enough to require intermittent or regular red-cell transfusions (Eschbach, Egrie, Downing, Browne, & Adamson, 1987) . Uremic anemia results from several interactive processes. Shortened red blood cell survival is a frequent manifestation of uremia (Jacob, Eaton, & Yawata, 1975) . Although controversial, some evidence suggests that uremic toxins inhibit heme synthesis and erythroid progenitor cell formation (Erslev, 1975; Jacob, Eaton, & Yawata, 1975). Blood loss due to platelet dysfunction and as a result of hemodialysis are implicated (Desforges, 1975) . The increased level of potentially toxic "middle molecules" in hemodialysis compared to peritoneal dialysis may also contribute to anemia in hemodialysis. Iron deficiency, folate deficiency, B12 deficiency, aluminum-induced microcytosis and hypersplenism have been cited (Paganini, 1989) . One study has demonstrated a significant increase in hematocrit and Hg levels following 12 months of endurance exercise training, which suggests that sedentary life style, a frequent consequence of end-stage renal disease, may contribute to the anemia (Goldberg et al., 1986). However, inadequate erythropoietin (EPO) production relative to the degree of anemia is the major cause of uremic anemia. Thus, uremic anemia is primarily due to an endocrine deficiency state, correctable by rEPO replacement therapy (Chandra,

PAGE 34

26 McVicar, & demons, 1988) . Intravenously administered rEPO has been demonstrated to restore Hg levels to normal in many renal patients (Eschbach, Egrie, Downing, Browne, & Adamson, 1987; Winearls et al., 1986). Function of Hemoglobin and Adaptation to Anemia Under normal atmospheric and physiologic conditions, more than 98% of the total oxygen carried by arterial blood is bound to Hg, with the remainder dissolved in the aqueous portion of blood (P02) . To cross cell membranes, oxygen must first disassociate from Hg and dissolve directly in the blood. The function of Hg may be considered to be the maintenance of normal P02 by providing a large reservoir of available oxygen. Thus, in anemia the reservoir of oxygen is reduced. Anemic organisms may maintain normal tissue oxygenation through a variety of adaptive mechanisms. Blood flow is deviated to more vital areas. The affinity of Hg for oxygen decreases. When Hg falls below 7.5 g/dl, an increase in both heart rate and stroke volume results in a significant rise in resting cardiac output (Bunn, 1980b) mediated by increased sympathetic activation. Although PC02 is a more potent controller of ventilation than is P02, low P02 resulting from severe anemia may stimulate peripheral chemoreceptors on the carotid bodies resulting in increased ventilation. However, the resulting reduction in PC02 and developing alkalosis are likely to attenuate this response (Vander, Sherman, & Luciano, 1980) .

PAGE 35

27 Hypoxia stimulates production of EPO primarily by the kidneys. As the major regulator of erythrocyte production, EPO stimulates erythropoiesis, the process in which erythrocytes develop from stem cells in the bone marrow, manufacture Hg from amino acids and body iron stores and then enter the blood stream as mature erythrocytes. Defective kidneys usually fail to produce a sufficient increase in EPO in response to hypoxia (Erslev, 1987) . Erythropoietin Therapy and Iron Deficiency In the absence of compliance with a regimen of iron supplementation, the rapid erythropoiesis produced by treatment with rEPO will seriously reduce iron levels (Van Wyck, 1989) . Subtle iron deficiency may be detected by low serum ferritin levels. With a further decrease in iron stores, low serum iron levels, increased iron binding capacity, and elevated free erythrocyte protoporphyrin are apparent. Iron deficiency anemia is a more severe manifestation of iron deficiency and characterized by reduced Hg levels (Ritchey, 1987) . In the case of iron deficient patients treated with rEPO, iron deficiency may block further reduction of anemia and, as discussed in the next section, may have neuropsychological consequences. Neuropsychology of Anemia Factors believed to contribute to neuropsychological dysfunction of renal disease include the buildup of putative neurotoxins, alterations in membrane permeability, and electrolyte disequilibrium (Powell et al., 1986; Trompeter,

PAGE 36

K 28 Polinsky, Andreoli, & Fennell, 1986). Attempts to isolate specific uremic neurotoxins have met with limited success (Schreiner, 1975; Trompeter, Polinsky, Andreoli, & Fennell, 1986) . Prior to the availability of rEPO, Hg values were occasionally included in neuropsychological studies in renal disease; however, perhaps due to the small range of values, anemia was not found to be contributory (Fennell et al., 1987; Gilli & DeBastiani, 1983). In continuous ambulatory peritoneal dialysis (CAPD) Hg levels tend to run higher, aspects of cognitive performance tend to be closer to normal and "middle molecule" clearance is up to ten times greater. Nissenson (1989) suggests that higher hemoglobin rather than lower "middle molecule" levels may underlie the neuropsychological advantage of CAPD over hemodialysis. Numerous studies have documented the presence of neuropsychological dysfunction in other forms of anemia. However, not one of these studies has controlled for the cause of the anemia, which in every case has been reported to produce neuropsychological deficits. As all of the known effects of rEPO are mediated by erythropoiesis, improvement in function following rEPO replacement therapy may be attributed solely to the rise in Hg and hematocrit. Neuropsychological deficits are well documented in iron deficiency anemia; however, while there is support for the direct effect of iron deficiency on the brain, the role of anemia controlling for iron deficiency has not been examined

PAGE 37

29 (Deinard, List, Lindgren, Hunt, & Chang, 1986; Groner, Holtzman, Charney, & Mellits, 1986) . Until recently, with the exception of blood transfusions, uremic anemia has only receded following successful renal transplantation. Although transplantation often enhances many neuropsychological functions, there is a huge confound in the simultaneous improvement in both uremia and anemia. Therefore, the role of anemia in neuropsychological dysfunction has not been adequately tested in any population. Effects of Anemia Anemic patients may complain of a variety of symptoms including fatigue, dizziness, dyspnea, poor concentration, irritability, weakness, faintness, headache, impotence and tinnitus (Bunn, 1980b; Rapaport, 1987) . Some of the complaints reported in conjunction with anemia may be due to the cause of the anemia rather than the effects of low Hg levels (DaviesJones, Preston, & Timperley, 1980) . The relationship between severity of symptoms and Hg concentration is confounded by the effectiveness of a variety of adaptive mechanisms. Symptomatology is dependent on factors including the degree and rate of reduction in oxygen carrying capacity of the blood, degree and rate of change in total blood volume, and the compensatory capacity of the cardiovascular and pulmonary systems (Wintrobe et al., 1981). Following adaptation symptoms may subside. If anemia develops slowly, effective adaptation may avert the

PAGE 38

30 appearance of symptoms. Even in cases of severe chronic anemia (6 to 8 g per dl) , the patient may be asymptomatic or complain only of fatigue (Wintrobe et al., 1981). Fatigue appears to be the most common symptom, regardless of severity of the anemia. Comparing symptom frequency in anemia characterized by Hg levels above or below 8 g/dl, Dawson, Ogston and Fullerton (1969) reported that in anemic subjects fatigue was present in 84.6% and 90.5%, respectively. In severe acute anemia, respiratory and circulatory symptoms suggestive of possible hypoxia (i.e., shortness of breath, "dizziness," faintness and tachycardia) are common (Wintrobe et al., 1981). Cognitive deficits have been documented in several of the anemias (DaviesJones, Preston, & Timperley, 1980; Tucker, Sandstead, Penland, Dawson, & Milne, 1984) . However, previous studies have been confounded by the failure to control for factors causing the anemia and suspected of producing deleterious effects through mechanisms unrelated to anemia. Nevertheless, existent studies provide some relevant information. As it is the most common, the best studied and the most relevant to the current project (i.e., due to its possible appearance during rEPO treatment) , iron deficiency anemia will be emphasized. Human studies examining the effects of iron supplementation in deficient subjects have found increased mental development scores in 15 month old infants, improved attention, but not memory in 3-6 year olds, improved problem

PAGE 39

31 solving capacity and improved STM and attention in adults (Deinard, List, Lindgren, Hunt, & Chang, 1986; Groner, Holtzman, Charney, & Mellits, 1986; Honig & Oski, 1984; Oski, Honig, Helu, & Howanitz, 1978; Pollitt, Leibel, & Greenfield, 1983; Pollitt, Siemantri, Yunis, & Scrimshaw, 1985; Rybo, Bengtsson, Hallberg, & Oden, 1985; Walter, Kovalskys, & Steel, 1983). Some authors attribute the neurological complications of iron deficiency to anemia (DavisJones, Preston, & Timperley, 1980) . However, to date not one study has examined the effects of anemia controlling for the effects of iron deficiency. Anatomical and clinical evidence suggests that iron deficiency may affect dopaminergic systems (Birkmayer & Birkmayer, 1986, 1987; Drayer, et al., 1986; Hallgren & Sourander, 1958) . The mechanism may be a reduction in the number of dopamine D2 receptors (BenShachar, Ashkenazi, & Youdim, 1986) . Iron responsive neuropsychological deficits have been demonstrated in nonanemic iron deficient pregnant women and infants (Groner, Holtzman, Charney, & Mellits, 1986; Oski, Honig, Helu, & Howanitz, 1983). Evans (1985) noted that improved cognitive function occurs rapidly following iron supplementation, prior to correction of the anemia; therefore, the neuropsychological deficits associated with iron deficiency do not appear to be mediated by anemia. Reviewing the pediatric literature, Ritchey (1987) concluded that iron deficiency, in the absence of anemia, adversely

PAGE 40

32 affects infant behavior and impairs both infant and toddler performance on tests of cognitive function, with these effects magnified by the progression to iron deficiency anemia. However, subjects exhibiting anemia due to iron deficiency anemia are likely to be more iron deficient than nonanemic subjects. Some studies have found no differences between anemic and nonanemic subjects and attributed the positive findings predominant in the literature to uncontrolled variables such as socioeconomic status and education (Johnson & McGowan, 1983) . However, this does not explain the results of iron treatment studies. In addition, the results of animal studies support the predominant human findings (Massaro & Widmayer, 1981; Yehuda, Youdim, & Mostofsky, 1986; Youdim & Ben-Shachar, 1987) . Possible Mechanisms Mechanisms that might produce neuropsychological dysfunction in anemia include alterations in blood chemistry, cerebral hypoxia and general fatigue. As previously noted, hyperventilation induced by low 02 is possible, but is attenuated by the resulting hypocapnia (i.e., low C02) . The effects of severe hyperventilation include increased blood pH, alterations in neuronal excitability, changes in EEG and cerebral vasoconstriction (Fried, 1987) . Carbon dioxide concentration is the primary mechanism regulating cerebral blood flow with hypocapnia inducing vasoconstriction and hypercapnia inducing

PAGE 41

33 vasodilation (Heistadt, Marcus, & Abboud, 1987) . In the anemic state vasoconstriction in response to hypocapnia is likely to result in hypoxia. Compared to other tissues, the brain both requires a disproportionate amount of oxygen and is more sensitive to hypoxia (Davis-Jones, Preston, & Timperley, 1980) . Neuropsychological deficits attributable to hypoxia have been observed in chronic lung disease, normal patients experiencing iatrogenic hypoxia and patients suffering from sleep disordered breathing. The hippocampus is especially vulnerable to hypoxia due to the nature of its blood supply. Hippocampal dysfunction is associated with deficits in LTM encoding (Muramoto, Kuru, Sugishita, & Toyokura, 1979) . The disassociation between LTM and STM seen in patients with lesions in the hippocampal area is well known. However, chronic hypoxia may have more widespread effects. Measures of verbal and nonverbal memory, intelligence and verbal fluency have been reported to be affected by hypoxia in association with chronic sleep disordered breathing (Berry, Webb, Block, Bauer, & Switzer, 1986) . The relevance of hyperventilation and cerebral hypoxia to well managed hemodialysis patients is unknown. Hemodialysis increases the degree of alkalosis. Hyperventilation secondary to severe anemia might, theoretically, increase the degree of alkalosis. Although peripheral hypoxia may be common in uremic anemia, the incidence of cerebral hypoxia is unknown. The possibility

PAGE 42

34 of an interaction between factors affecting cognitive function should not be discounted. Young adults one to three years post mild head injury (i.e., concussion) exhibit impairment on immediate memory and vigilance tasks relative to controls when all subjects are tested under mildly hypoxic conditions (Ewing, McCarthy, Gronwall, & Wrightson, 1980) . Renal patients may be more susceptible to the effects of anemia than are otherwise healthy anemic patients . In contrast to the unknown incidence of hypoxia and hyperventilation, fatigue, the first and most chronic symptom of anemia, is almost invariably present in end-stage renal disease (Dawson, Ogston, & Fullerton, 1969) . Fatigue may be expected to consistently occur prior to the development of cerebral hypoxia, if cerebral hypoxia occurs at all. At a level of anemia severe enough to produce cerebral hypoxia, it appears likely that the level of fatigue will also be greatly increased. Therefore, fatigue may be a more reliable mediator of the putative neuropsychological effects of anemia. The effects of fatigue on neuropsychological performance are well known in clinical practice (Lezak, 1983) . However, the effects do appear to be variable. Subject factors such as age and physical condition and fatigue factors such as cause and duration may be significant. No changes in neuropsychological test performance were demonstrated in 42 surgical residents when

PAGE 43

35 assessed following 2.0 ±1.5 hours of sleep vs. 6.5+1.0 hours. In the acute sleep-deprived state, residents were less vigorous and more fatigued, depressed, tense, confused and angry (p<0.05) than they were in a rested state. These changes in mood had no effect on measures including immediate and delayed recall of logical stories. Trails, immediate memory, or digit symbol (Bartle et al., 1988). However, it is conceivable that more difficult tasks would have shown an effect. The biological function of the experience of fatigue is hypothesized to be reduction in unnecessary activity. It is hypothesized that the effects of increasing fatigue are likely to appear along a continuum. With increasing fatigue the organism's proclivity to define activity as necessary is likely to decrease. In a testing situation, prior to refusal to perform a task, the subject's definition of the task may change to make the task less taxing. Therefore, tasks not requiring, but enhanced by, additional effortful cognitive activity may be expected to be most sensitive to the effects of fatigue. Tasks overtly demanding attention, vigilance and/or extensive cognitive processing are hypothesized to be affected only as the severity of the fatigue increases. Effect of Ervthropoietin on Neuropsychological Function The effects of rEPO replacement therapy on quality of life and neuropsychological functioning will be briefly

PAGE 44

36 reviewed. The results of the pilot study and possible negative side effects of rEPO will then be discussed. Effect of Erythropoietin on Quality of Life Treatment with rEPO has been demonstrated to increase Hg concentration and reported to improve patient well-being (Eschbach, Egrie, Downing, Browne, & Adamson, 1987) . In the first clinical trial, out of ten patients, nine reported an improved sense of well-being and eight reported increased exercise tolerance (Winearls et al., 1986). A study of 37 rEPO treated chronic hemodialysis patients found 16% of the patients returned to work, appetite improved in 81%, 78% reported subjective improvement in exercise tolerance, 70% participated in more social activities, sleep improved in 68% and 84% reported an increased sense of well being. Patients between 4 and 49 years of age showed the most improvement. A hematocrit value of >27% appeared to be the critical level for improved sexual function in men (Delano, 1989) . Evans, Rader and Manninen (1990) found significant increases in self report of energy and strength, greater appetite, improved breathing and reduced tension and anxiety. There were very small but significant increases in patient report of quality of life, affect and life satisfaction. In an investigation of the effect of rEPO on working capacity, eight sedentary chronic hemodialysis patients rode a bicycle ergometer before and after elevation of the group

PAGE 45

BT 37 mean Hg from 5.9 g/dl to 10.9 g/dl with rEPO treatment. The initial mean anaerobic threshold of 70 watts represented a very limited exercise capacity with inability to perform routine household work over a sustained period. Following rEPO treatment, the mean anaerobic threshold of 106 watts indicated that work capacity had increased enough to allow for full physical rehabilitation with regard to everyday life (Mayer, Thum, Cada, Stummvoll, & Graf, 1988). Previous Neuropsychological Studies Several small pilot studies have assessed the impact of rEPO on cognitive function. Wolcott, Schweitzer and Marsh (1988) reported that in a group of nine chronic hemodialysis patients administration of tests that loaded on visual, conceptual and visuomotor tracking and auditory verbal learning demonstrated trends toward improvement. Wolcott, Schweitzer and Nissenson (1989) , in a study of the effects of EPO treatment in a group of 17 chronic hemodialysis patients, reported improvements in symbol-digits modality and Trails; however, it is unclear whether these improvements reached clinical or statistical significance. Decreased P3 latency purportedly represents increased speed and efficiency of information processing. Increased amplitude of event-related potentials suggests improvement in cognitive function. Nissenon, Marsh and Brown (1988) reported that following a mean rEPO mediated improvement in hematocrit from 22.7% to 36.6%, 13 chronic hemodialysis patients exhibited a decrease in P3 latency of response to

PAGE 46

38 auditory stimuli, however, these results did not reach statistical significance. A follow-up study (n=17) found no change in P3 latency which suggested that there was no change in speed of processing. The amplitude of the P3 wave increased significantly (p<.03). Reported changes were 19%, 63% and 160% in the parietal, vertex and frontal areas respectively, and are considered by the authors to represent increased attention span, memory and efficiency of cognitive processing (Nissenson, Marsh, Brown, Schweitzer, & Wolcott, 1989) . Results of Pilot Study In a small pilot study, seven anemic hemodialysis patients were assessed with a battery of neuropsychological tests and a mood questionnaire (Profile of Mood States) pre and a mean of 98.6 days post (range 77 to 125) initiation of 100 units/kg of rEPO (Klein et al. 1989). The study was prospectively randomized and double-blind; however subjects in most cases were able to accurately report their treatment status, apparently based on subjective experience and behavioral self -observation. Although eleven patients initially began the study, complete data was only obtained on seven due to a variety of factors including medical problems, transplantation and patient refusal to be tested. Results were available for two placebo controls and five patients receiving rEPO. The mean increase in Hg was 0.55 mg/dl in controls and 5.2 mg/dl in those on active treatment, a mean treatment group change from 6.10 +0.97

PAGE 47

39 g/dl, severely anemic, to 11.56 +2.41, mildly anemic. At sea level anemia is suspected in men when Hg is <14.0 g/dl and <12.0 g/dl in women (Rapaport, 1987). Obs GrouD Hg Pre Hg Post Hg Chanae Sum CVLT Chanae Sum CVLT SS Chanae -0.93 1 Con 7.0 6.7 -0.3 -6 2 Con 8.1 9.5 1.4 -5 -0.55 3 rEPO 6.0 8.4 2.4 1 0.11 4 rEPO 5.1 11.2 6.1 3 0.47 5 rEPO 7.1 13.2 6.1 8 1.08 6 rEPO 5.7 12.0 6.3 0.00 7 rEPO 6.6 13.0 6.4 16 1.90 Note: Obs = Observation; SS = Standard Score; Con = control group; rEPO = treatment group; Change = change score The Profile of Mood States, a self-report measure, failed to reveal any improvement in fatigue, tension, depression, vigor, or confusion. However, individual patients reported dramatic improvements in circumscribed domains. Individual subjects reported increased capacity for physical activity, a reduction in post dialysis fatigue and decreased problems with impotency. The dramatic reduction in physical symptoms in association with absence of subjective effects is consistent with reports of reduced subjective symptoms when the anemia is chronic, even if it

PAGE 48

40 is severe. Despite this, the absence of change on the fatigue and vigor subscales is somewhat suspect and may represent the inability of this subject population to accurately report their subjective state in the manner required by the POMS. Wolcott, Schweitzer and Marsh (1988) also found an absence of significant changes on the POMS following a rEPO mediated increase in mean hematocrit from 23.1% to 36.0% in 13 patients; however, mean POMS vigor and fatigue scales changed in the expected directions and approached statistical significance. As predicted, there was no improvement on the Comprehension subtest of the Wechsler Adult Intelligence Scale Revised. This measure of overlearned semantic stores is usually stable in the absence of severe dysfunction. There was also no improvement in verbal behavioral fluency as measured by the Controlled Oral Word Association Test (FAS) . Neither was there any increase in motor speed as measured by finger tapping; however, two patients exhibited quite noticeable differences in capacity for sustained finger activity. During the first assessment finger cramping, pain and fatigue required rests between trials. Rests were not required following initiation of rEPO treatment. On Trails A and B, a test stressing motor speed, visual search, mental control and decision speed, no improvement was noted. On Trails A, one treated subject moved from borderline to within normal limits. One treated subject.

PAGE 49

Pt* 41 moved from below normal to within normal limits; however, this subject exhibited limited response to rEPO. The other three treated subjects actually were slower post-test. The greatest improvement was seen in one of the two controls, who moved from clearly below normal to within normal limits. Likewise, on Trails B, the greatest improvement was in the treated subject who had only limited response to rEPO and the second greatest improvement was in a control. Three of the four treated, responsive subjects actually performed more poorly on the second testing. These findings contrast with the report of Wolcott, Schweitzer and Nissenson (1989) that Trails improved in treated patients. Correlation between improved performance on the Sum of five trials on the California Verbal Learning Test (Sum CVLT) supra-span word list learning task and improvement in Hg was r=0.796 (e=.0324). The correlation between improvement in Sum CVLT and rise in hematocrit was r=0.741 (E=.0568). Neither the first two trials of the five trial sequence nor the once presented interference list showed improvement. Improvement was observed primarily in the last two trials. Single and double presentations of supra-span lists with immediate recall are highly dependent on STM (T. White, personal communication, June 12, 1990). They may be considered only weak measures of LTM. Performance on the fourth and fifth presentation is likely to be highly dependent on the degree to which the subject has organized and elaborated the list of target words. Thus, the improved

PAGE 50

or. 42 performance appears to have been due to enhanced LTM. This is supported by the lack of improvement on the digit span subtest of the Wechsler Intelligence Scale Revised. However, the Logical Memory subtest of the WMS showed no change . Logical Memory is a single trial task and, because, as the name implies, it is already logically organized, it might be expected to less vigorously discriminate between sparse and elaborate processing. Improved performance was observed on the CVLT only after the third presentation. Multitrial supra-span tasks require more active cognitive processing than does the single presentation of a paragraph. Given the hypothesized increase in general vitality, the results may be explained by a greater proclivity, posttreatment, for effortful cognitive activity such as the kind required to organize and elaboratively rehearse a shopping list. No improvement in immediate or delayed visual memory was observed based on the Visual Reproduction subtest of the WMS. This test involves only one presentation of the stimuli; therefore, may not be less sensitive than Sum CVLT to changes in degree of cognitive processing. However, the lack of improvement is also consistent with the suggestion that visual-motor integration skills are permanently affected by renal disease, at least if age of onset was during childhood (Mings, 1987) .

PAGE 51

'•ssa 43 Improved performance on Sum CVLT, in the absence of other changes, is hypothesized to be due to reduction in fatigue. It is hypothesized that the effects of fatigue are likely to appear along a continuum mediated by attributions regarding the importance of effortful activities. With increasing fatigue the saliency of cues requisite to the production of a given level of arousal is likely to change. Nonessential, covert cognitive activity may diminish prior to cognitive activity overtly and undeniably essential to performance of the stated task. Therefore, do to the nature of the tasks, neuropsychological measures directly stressing overt attention and vigilance may be less affected. However, tasks benefiting from but not requiring covert effortful cognitive processing may be most vulnerable to the effects of fatigue. The dramatic reduction in anemia produced by treatment with EPO has been shown to produce clinically significant improvements in domains such as aerobic capacity, anaerobic threshold, sexual potency, employment status, level of social activity, perceived ability to engage in physical exercise and improved appetite. In contrast significant changes in self -reported mood have been less consistently documented. Preliminary neuropsychological test data suggests improvement on verbal supra-span multitrial list learning tasks. In addition, preliminary reports suggest possible improvement in other cognitive domains as measured

PAGE 52

44 by increased amplitude of event-related potentials and paper and pencil tests. An alternate interpretation of the results of studies examining the cognitive and affective effects of rEPO treatment is that the placebo effect was a major factor in the findings. Examination of Sum CVLT change scores in the study by Klein and colleagues (1989) revealed a significant portion of the variance to have been contributed by the two control subjects, both of whom declined in performance. This may be explained by negative placebo effect in the subjects initially expecting to be on a drug reported to produce major changes in life functioning. Likewise, improvement in the treatment group was consistent with a positive placebo effect. The failure of Nissenson, Marsh, Brown, Schweitzer and Wolcott (1989) to replicate the earlier study of Nissenon, Marsh and Brown (1988) , a study which found increased speed of processing based on P3 latency, is also, arguably, consistent the transitory placebo effect often observed with new treatments. Negative Effects Treatment with rEPO is not without possible negative neuropsychological conseguences . Increased hematocrit appears to raise peripheral resistance resulting in increased hypertension in 10% to 22.2% of patients (Hori, Onoyama, Iseki, Fujimi, & Fujishima, 1990). Hypertension is a major problem in the management of renal disease. In the first British clinical trials, Winearls and colleagues

PAGE 53

ff 45 (1986) reported that one of their ten patients developed hypertensive encephalopathy. In addition, increased blood clotting may potentially reduce circulation (Winearls et al., 1986). Other side effects reported include anxiety, lethargy, headache, body aches and, in 32% of subjects in one study, iron deficiency (Delano, 1989) . In addition, concern has been raised that increasing hematocrit to normal levels may compromise any existing renal function and reduce the efficiency of dialysis (Koene & Frenken, 1990) . Curiously, nervous tissue is capable of producing EPO. After renal carcinoma, cerebellar hemangioblastomas have been reported to be the most frequent cause of elevated EPO in association with neoplasms (Hennessy, Stern, & Herrick, 1967; Race, Finney, Mallams, & Balla, 1964; Waldmann, Levin, & Baldwin, 1961) . The proclivity of neural tissue to produce EPO leads to speculation that EPO may normally be produced by and have a function in the brain; however, there is no evidence of EPO mediated neural effects in cancer patients suffering from EPO secreting tumors. However, physiological EPO levels are abnormally low for the degree of anemia in renal failure, but often within the range found in nonanemic populations. Therefore, EPO levels may be raised dramatically during rEPO replacement therapy. However, the suggestion that EPO might affect nervous tissue is purely speculative. Of more interest is the observation that rEPO treatment may reduce sympathetic arousal. Cardiac output is increased

PAGE 54

46 in severe chronic anemia. This adaptive reaction is mediated by sympathetic activation. Recombinant EPO, by reducing anemia, reduces demands for high cardiac stroke volume and heart rate; therefore, may reduce sympathetic activation (Hori, Onoyama, Iseki, Fujimi, & Fujishima, 1990) . This could be of significance given suggestions of neuropsychological underactivation in uremia. Summary and Conclusions Memory dysfunction has been reported in chronic renal disease with reasonable consistency. Chronic renal disease is characterized by a host of symptoms, many of which stem from nervous system dysfunction. Over the past 25 years numerous studies have examined cognitive dysfunction in uremia. These studies have occurred within the context of medicine's increasing ability to simulate, or in the case of transplantation, restore normal renal function. However, today, the physiological mechanisms and cognitive processes underlying neuropsychological dysfunction in renal disease are still largely unknown. Recombinant EPO, a potent new treatment for the anemia of renal failure, has produced dramatic improvements in work capacity, sexual potency and quality of life in impressive percentages of treated patients. Because many of the symptoms once attributed to uremia now appear to be dramatically reduced through the treatment of uremic anemia, it seemed reasonable to hope that neuropsychological dysfunction might be similarly affected. A small pilot

PAGE 55

47 study found improvement only in effortful verbal LTM. Furthermore, in retrospect, it was noted that the sometimes reported advantage of peritoneal over hemodialysis and decline in performance with time on hemodialysis may be at least partially mediated by degree of anemia. Also, there are suggestions that uremic memory dysfunction is at least partially mediated by reduced vigilance, underarousal , increased distract ibility and cognitive slowing. These observations are consistent with the hypothesis that effortful verbal memory may be vulnerable to anemia, secondary to mechanism such as fatigue or hypoxia. The goal of the present study was to attempt to replicate the improvement in LTM on a a multitrial supraspan learning task which had been previously shown to correlate with reduction of uremic anemia. Previous work suggested that verbal memory tasks making fewer demands on effortful cognitive processing showed no improvement. If improvement were demonstrated on effortful memory tasks as well as on automatic memory tasks and on the Levels of Processing task, then the most plausible explanation would be that uremic anemia deleteriously affects memory structures, possibly mediated by hypoxia, although other explanations might be posited. Improvement only in effortful memory would suggest that the change was due to increased cognitive processing, possibly secondary to reduced fatigue or increased activation.

PAGE 56

48 Failure to replicate would provide information useful in several areas. The extent to which homeostatic mechanisms adapt to chronic anemia is unknown. Data delineating the parameters within which anemia affects cognitive function could potentially contribute to the determination of the optimal Hg at which to maintain patients in end-stage renal disease. This is important given the monetary costs and medical risks associated with utilizing rEPO to treat uremic anemia. Failure to demonstrate Hg responsive neuropsychological dysfunction in rEPO treated uremic anemia would lend support to the suggestion that the deficits documented in other forms of chronic anemia are mediated by factors other than anemia. In the case of iron deficiency anemia, if the reported deficits are entirely due to the direct effects of iron deficiency on the brain, then, since iron deficiency appears prior to anemia, this would suggest the need for more aggressive screening for iron deficiency, even in the absence of anemia. Beyond scientific value, this would suggest that millions of adults and children may suffer from easily and inexpensively preventable neuropsychological dysfunction. Increased understanding of the neuropsychological consequences of uremic anemia may 1) increase understanding of the extent of the organism's ability to adapt to chronic anemia, 2) contribute to the quarter century quest for the uremic neurotoxin (s) , 3) show the extent to which anemia

PAGE 57

49 itself is likely to contribute to the cognitive deficits observed in other forms of anemia, 4) assist in determining the optimal level of Hg in treated renal patients and 5) in the event of positive findings, provide preliminary data regarding the mechanisms through which anemia undermines neuropsychological functioning.

PAGE 58

CHAPTER 3 METHODS The present study was designed to broadly address the question of whether change in anemia would be reflected by change in memory. Subjects All subjects were adults in end-stage renal failure and reasonably stabilized on some form of maintenance dialysis. Subjects were drawn from the Shands Teaching Hospital (STH) Adult Hemodialysis Clinic, the STH Dialysis Home Training Clinic and the Gainesville Veterans Administration Medical Center Hemodialysis Clinic. Age ranged between 2 3 and 73 years. Subjects with uncontrolled hypertension, mental retardation or known neurological disease were not accepted into the study. The treatment group consisted of 16 subjects on maintenance hemodialysis and one subject on peritoneal dialysis. It was planned that subjects would only be accepted into the study into their eighth week of rEPO treatment, based on the assumption that STH adult hemodialysis patient hematocrits would be rising for approximately five months. Therefore, a subject initially tested at 8 weeks was expected to just be reaching asymptote at the third assessment. However, subjects began 50

PAGE 59

K 51 receiving rEPO earlier than anticipated and control subjects became difficult to locate. Therefore, subjects were accepted into the study beyond the eighth week, reasoning that they could serve as controls in the event of minimal additional change in Hg. Mean lag between initiation of rEPO and the first assessment was 5.9 weeks (sd=4.57) . The 18 control subjects did not receive rEPO. Thirteen subjects, 72.2%, were on some form of peritoneal dialysis (i.e.. Peritoneal Dialysis, Continuous Ambulatory Peritoneal Dialysis or Continuous Cycler Peritoneal Dialysis) and five subjects, 27.8%, were on maintenance hemodialysis. Given the small population from which to draw controls, the ideal of matching controls to treated subjects on age, education and ideally, race and sex, was not considered feasible. Despite this, as will be described later, the two groups were nearly identically in domains other than treatment modality. However, mean Hg at the first or baseline assessment was 8.46 g/dl (sd=1.42) in the treatment group and 9.44 g/dl (sd=1.92) in the control group . All subjects were paid $15 at the completion of the third assessment. Subjects were tested in their home, their dialysis clinic, or in a STH Clinical Psychology testing room.

PAGE 60

'« 52 Measures California Verbal Learning Test The California Verbal Learning Test (CVLT) is a multitrial supra-span word list learning task designed to be an ecologically valid measure of multiple aspects of learning. The CVLT quantifies parameters including STM, LTM, retention over short and long delays, degree of vulnerability to proactive and retroactive interference, encoding strategies, effects of category cuing and the frequency of perseverations and intrusions (Delis, Kramer, Kaplan, & Ober, 1987). Three versions were used. Form I is in general clinical use. Form II is an alternate form developed for research purposes (Delis, McKee, Massman, Kramer, & Gettman, 1990). Form III, the Florida version of the CVLT, was developed for this study. A description of the construction and validation is provided in Appendix A. In keeping with the results of the pilot study the primary measure was the Sum of trials one through five of the CVLT (Sxm CVLT) . Levels of Processing Task A 36 word levels of processing task in use in the Shands Teaching Hospital Psychology Clinic was used and two alternate forms were developed for this study. Alternate forms two and three were designed with target words of similar frequency of occurrence to that of Form I. Please see Appendix B.

PAGE 61

^ 53 The levels of processing paradigm controls the nature of effortful processing through manipulation of the task. Subjects were visually presented a written word and asked to respond to a question that required either orthographic, phonemic or semantic analysis of the visual stimuli. Following an interval of approximately 20 minutes, accuracy in identification of the target words was measured with an oral multiple choice task. Subjects able to benefit from semantic processing may be expected to exhibit a higher rate of correct recognition for semantical ly processed stimuli. Subjects exhibiting memory deficits due to reduced cognitive processing, because extent of processing is controlled by the task, would be expected to perform like normal subjects on this task. Normal subjects show a recognition advantage for the semantically processed words (Craik & Tulving, 1975) . Subjects suffering from neurologically based memory disorders affecting encoding in LTM fail to benefit from semantic processing. Thus, improvement in ability to benefit from semantic processing would suggest physiological improvement. Frequency Estimation Task Estimation of frequency of occurence provides a clinical measure of automatic learning believed to be relatively invulnerable to factors such as fatigue, but presumably vulnerable to physiological factors (e.g., hypoxia) directly interfering with the functioning of memory structures (Hasher & Zacks, 1984) .

PAGE 62

54 After being instructed to try to recall as many words as possible, subjects were read a list of words with some words appearing more than once. Twenty minutes later subjects were asked to estimate how many times each word was read. Three versions were produced and are reproduced in Appendix C. The three versions were administered in consecutive order to all subjects. Prior to estimation of freguency, subjects were tested on free recall. Controlled Word Association Test Controlled Word Association Test, a verbal fluency task, provides a measure of retrieval unrelated to encoding. It is also dependent on level of arousal, motivation, verbal fluency and ability to initiate behavior. Three commonly used sets of letter triplets are available (Lezak, 1982) . Cognitive-Affective and Physical Behavior Questionnaire A self-report guestionnaire designed to assess cognitive-affective and behavioral changes likely to be caused by medical illness, was developed utilizing the types of guestions typically asked of medical patients to evaluate these domains. The guestionnaire generates a Cognitive-Affective Scale and a Physical (i.e., physical activity) Scale. The guestionnaire was written to avoid obvious response biases. The guestionnaire and scoring instructions are displayed in Appendix D. In the absence of positive results this measure was intended to determine whether there were positive changes

PAGE 63

55 in behavior in the treated subjects. The same version was administered at all three assessments. Hemoglobin In most instances, Hg levels were drawn on rEPO treated subjects on a weekly basis. When blood work was not done on the day of neuropsychological testing, pre and post target date Hg levels were used to calculate status on the day of the assessment. As Hg status tends to change linearly in the absence of major blood loss, estimation is likely to have been reasonably accurate. In the few instances where Hg levels were lacking, estimates were made based on hematocrit (Ritchey, 1987) . It should be noted that the experimenter was blind to Hg levels until the completion of all other data collection. Ferritin When available, ferritin levels were obtained for each subject to rule out iron deficiency as a possible confound (Ritchey, 1987). Ferritin, the most sensitive measure of iron levels, has been used in studies examining the neuropsychological effects of iron deficiency. However, in the absence of ferritin levels, iron saturation and/or transferrin were used if available. When the dates of neuropsychological testing and blood work failed to coincide, estimates were made based on iron measures obtained before and after the time of testing. As iron status tends to change slowly and linearly in the absence of major blood loss, estimation is reasonably accurate.

PAGE 64

err 56 Iron status was used to classify subjects into low and replete categories utilizing lab report guidelines. The normal range for ferritin is frequently cited as around 10 ng/ml to 107 ng/ml. Ferritin levels of under 10 ng/ml were considered indicative of iron deficiency. Ferritin values under 30 ng/ml were classified as low. Normal and elevated iron values were grouped as normal for the purposes of these analyses. Treatment with rEPO results in a heavy demand on iron stores and induces iron deficiency in the absence of iron supplementation (Van Wyck, 1989) . For this reason, as part of routine medical procedures, serum ferritin levels were drawn on most rEPO treated subjects on a monthly basis; however, the experimenter was blind to iron status until the completion of neuropsychological data collection. Schedule of Testing Subjects were tested on three occasions. Assessment Two was intended to be administered approximately 30 days following the first assessment. Assessment Three was scheduled for 60 days following the second assessment. Patient illness, noncompliance and difficulty working around transportation schedules significantly altered the assessment timetable in some instances. Because time of testing relative to hemodialysis has been reported to affect results, this was kept consistent for each subject. For most subjects, assessment occurred immediately following hemodialysis.

PAGE 65

57 Tests were administered in the same order at all three assessments and all subjects were administered test versions in the same order. Forms for the estimation of frequency of occurrence and levels of processing paradigms were administered in ascending order. For the CVLT, Form II was administered at the first assessment and Form I at the third assessment. This deviation from ideal experimental design occurred because Form III had not been fully validated prior to initiation of the study. Therefore, to minimize losses in the event of poor interform reliability, Form III was administered at the second assessment and forms were not counterbalanced. Controlled Word Association letter triplets were administered in the following order: CFL, FAS and PRW. Because the letter triplets CFL and PRW can be directly compared, the were placed in what were expected to be the two most important assessments, one and three. FAS is the most widely used letter triplet, but shares a letter with CFL. To minimize possible practice effects for individual letters in the comparison of CFL and PRW, CFL was administered prior to FAS. As change scores rather than absolute performance was compared, error variance stemming from possible differences in difficulty between the three letter triplets was minimized. Tasks were administered in the following order: Informed consent, collection of demographic data, CVLT, levels of processing encoding phase, frequency of

PAGE 66

m 58 occurrence encoding phase, Controlled Word Association Test, Cognitive-Affective and Physical Behavior Questionnaire, levels of processing recognition phase, CVLT delayed tasks, frequency of occurrence estimation phase. Statistical Methodology Statistical analyses were performed using the Statistical Analysis System (SAS) package of statistical procedures available for the IBM personal computer. The SAS General Linear Model procedure was used to test the main hypothesis. Analyses involving variables violating assumptions of normality and homogeneity of variance were performed utilizing nonparametric procedures. The assigned alpha for the main hypothesis was .05, with secondary hypotheses tested more stringently based on the requirements of multiple testing and the extent to which the assumptions underlying parametric procedures were violated. H ypotheses There was one main hypothesis and four secondary hypotheses. Two of the secondary hypotheses were intended, in the event of failure to reject the main hypothesis, to provide preliminary data regarding possible mechanisms mediating the relationship between Hg and cognitive functioning. Main Hypothesis The main hypothesis was that adult renal dialysis patients receiving rEPO replacement therapy would exhibit

PAGE 67

59 improved verbal LTM as measured by Sum CVLT, compared to dialysis subjects not receiving rEPO. This improvement was hypothesized to be mediated by rise in Hg. Recombinant EPO itself was not expected to have any effect on Sum CVLT. Recombinant EPO treated subjects typically concurrently receive iron supplementation; therefore, iron status was not expected to change in the vast majority of subjects. Nevertheless, it was necessary to control for iron status because it has been a confound in the majority of studies focusing on the neuropsychology of anemia as deficiency is associated with neuropsychological deficits. A multivariate analysis was expected to demonstrate a main effect for Hg while controlling for rEPO treatment status and iron status. Secondary Hypotheses The secondary hypotheses are described below. The first secondary hypothesis was that self -reported physical activity, as measured by the Cognitive-Affective and Physical Behavior Questionnaire, would improve in response to reduced anemia. Despite self -reported changes in life functioning, the pilot study failed to demonstrate any alteration in affective state using a self -report adjective checklist, possibly because subjects were unable to adequately perform the task, which required the ability to abstract and introspect. Therefore, it was hypothesized that a concrete, behaviorally oriented self -report device might

PAGE 68

60 produce more accurate results. Thus, the second secondary hypothesis was that self -reported behavioral indicators of cognitive-affective state, as measured by the CognitiveAffective and Physical Behavior Questionnaire, would improve in response to reduced anemia. In the event of positive findings the testing of several other hypotheses was planned. The intent of these tests was to attempt to characterize the mechanisms mediating improvement in memory. These were preplanned exploratory analyses designed to focus on the question of whether putative improvement in memory following reduction in anemia is mediated by functional or more basic physiological processes. Increased benefit from semantic processing would be suggestive of enhancement of functioning at the level of the physiological substrate of the memory system. However, it was reasoned that improvement in Sum CVLT, if observed, was more likely to be mediated by increased cognitive processing secondary to reduced fatigue; therefore, the third secondary hypothesis was that rise in Hg would not be associated with increased ability to benefit from semantic processing as measured by the semantic score on a levels of processing task. Reasoning along similar lines, the fourth secondary hypothesis was that rise in Hg would not be accompanied by improved automatic learning as measured by the ability to estimate frequency of occurrence.

PAGE 69

}l ^* CHAPTER 4 RESULTS Overview of Analyses Prior to analysis, the data was inspected for outliers and tested for the assumption of normality using qualitative and quantitative tests available through the SAS Univariate procedure. Hemoglobin and Sum CVLT, the primary independent and dependent variables, respectively, exhibited good approximations of normal distributions and homogeneity of variance. In contrast, the distributions of most other variables deviated from normality to some degree. For variables failing the Shapiro-Wilks W test of normality, the ratio of the standard deviation to the mean was computed. Variables with ratios greater than .25 were further examined quantitatively and graphically (Schlotzhauer & Littell, 1987) . Procedures such as SAS General Linear Model are relatively robust with regard to violation of normality; however, great caution is required in interpreting analyses in which variables also violate the assumption of homogeneity of variance. The assumption of homogeneity of variance between groups was tested with the Folded F test. With the exception of ferritin, all variables necessary for testing of the secondary hypotheses exhibited reasonably good 61

PAGE 70

62 between group equality of variance at baseline. To test the impact of these violations of the assumptions of parametric procedures, parametric and nonparametric tests of between group baseline values were calculated for all subjects and the results compared. As displayed in Appendix E, even using an alpha of .05, only 2 tests out of 55 variables, both highly unstable and insignificant variables, produced discrepant results. It was concluded that the violations of normality were unlikely to be of sufficient severity to impact on the validity of SAS General Linear Model parametric procedures, except for ferritin. Since ferritin was used as a classif icatory variable inequality of variance had no appreciable effect on statistical procedures. Demographic and descriptive statistics were calculated using the SAS Means and Frequency procedures. To rule out extraneous between group differences baseline variables were compared. It was planned that the main hypothesis would be tested with a repeated measures ANOVA using the General Linear Model procedure. A second analysis was planned to test the hypothesis that the expected treatment effect was associated with Hg change rather than rEPO itself and that the treatment effect was not an artifact of variability in iron status. Although a repeated measures analysis with the classif icatory terms group membership and iron status and the covariant Hg had been planned, the unexpected finding of no change in Hg status between the second a third assessments led to the

PAGE 71

63 dropping of the third assessment from the preplanned analyses. Therefore, the first test of the main hypothesis was a GLM univariate procedure with treatment status as the independent variable and Sum CVLT change score as the dependent variable. The second planned test was a GLM analysis of covariance, the added independent factors being the classificatory variable iron status and the continuous variable blood Hg level. In actuality, change in iron status affected only a small number of subjects; therefore, controlling for iron status became of minimal importance. Kruskal-Wallis nonparametric procedures were used to test the hypotheses that self -reported physical vitality and self-reported cognitive behavior improved in the treatment group. If a main effect for Hg on Sum CVLT had been found then further nonparametric analyses would have been performed in an attempt to provide preliminary data to guide future inquiry into the mechanisms mediating the effect. Specifically, between group differences in ability to benefit from semantic processing and changes in encoding of incidental information were planned. Negative findings eliminated this rationale for these tests. Nonetheless, they were included in post hoc analyses. Post hoc exploratory analyses focused on attempts to disconfirm the negative findings and to explore other aspects of the data. The observation that treatment status was a poorer predictor of Hg change than had been

PAGE 72

64 anticipated, combined with reasoning along the lines that response to anemia may be quite idiosyncratic, led to the decision to reclassify subjects based on Hg change scores. Unfortunately, the small range of Hg change in most subjects necessitated a choice between sacrificing either statistical power or clinical relevance. Retaining statistical power, subjects were first classified based on small distinctions with unlikely and undemonstrated physiological significance. Following this, compromising between sample size and physiological significance, only subjects meeting the criteria of Hg change scores of +1 g/dl were retained. CVLT scores in subjects with clinically significant changes in Hg status were then examined. The post hoc hypothesis that reduction in anemia might alter the pattern of learning was tested. The slopes of learning curves for trials one through five of the CVLT were calculated with SAS Regression Analysis procedures. A GLM ANOVA then compared changes in slopes between groups. Additional procedures were exploratory. Change scores were calculated for all variables and the performance of specific variables analyzed. Also, relationships between test performance and questionnaire data were explored. Statistical Assumptions The basic assumptions underlying most analysis of variance procedures include 1) normality and 2) homogeneity of within-group variance. Analysis of variance procedures are quite robust with regard to violation of the assumption

PAGE 73

65 of normality as well as departures from homogeneity of variance. However, the concurrent violation of both assumptions calls at the very least for conservatism in determination of significance. As can be seen in Appendix F, Table F-1, based on the Shapiro-Wilks W statistic computed by SAS Univariate, the primary independent variable, Hg, and the major dependent variable, Sum CVLT, were both normally distributed. Hg values were slightly skewed to the right. Sum CVLT was very minimally skewed to the left. Kurtosis in the distribution of Hg values approached zero. For Sum CVLT, the kurtosis value showed the tails to be slightly lighter than expected in a normal distribution. Iron levels tend to be quite elevated in dialysis patients dependent on repeated blood transfusions. In patients running relatively normal iron levels, rEPO treatment may deplete iron stores. Therefore, it is not surprising that serum ferritin violated the assumption of normality. As it was used only to classify subjects in terms of iron status, this violation was irrelevant to the analyses. Other measures required for the testing of secondary hypotheses for the most part violated assumptions of normality to some degree. For variables failing the Shapiro-Wilks W test of normality, the ratio of the standard deviation to the mean was computed, as displayed in Appendix F, Table F-1. Variables with ratios greater than .25 were

PAGE 74

66 further examined quantitatively (i.e., kurtosis and skew) and graphically (i.e., box plots) The estimation of frequency of occurrence data violated the test of normality. The ratio of the standard deviation to the mean was 0.2 62, just over the established limit. The distribution was somewhat skewed to the left; however, Kurtosis was minimal. The Semantic score generated by the levels of processing task failed the Shapiro-Wilks W test of normality; however, the ratio of the standard deviation to the mean was within acceptable limits. The Orthographic variable from the levels of processing task also failed the test of normality and the standard deviation was large relative to the mean. The two Cognitive-Affective and Physical Behavior Questionnaire scales failed the Shapiro-Wilks W test and the standard deviation to mean ratio was .280, slightly above the .25 recommended cut-off. The scores were slightly skewed to the left and the tails were heavier than expected in a normal distribution. The variables age, years of education and years on dialysis all failed the Shapiro-Wilks test. However, the ratio of standard deviation to mean was within accepted limits for the education variable. Slope and intercept of the learning curve for CVLT trials 1-5, variables used for post hoc analyses, both passed the Shapiro-Wilks test. Slope had somewhat heavy tails.

PAGE 75

67 The assumption of equality of variance between groups was tested with the Folded F test, F' . The results, as displayed in Appendix F, Table F-2 , revealed the following CVLT variables to fail the test of homogeneity of variance at the .10 level: Sum of clusters, Sum of intrusions, and Sum of perseverations for trials one through five. Short Delay perseverations and intrusions, and all three CVLT measures of false positives not on the interference list. Other measures failing the test of homogeneity of variance were: levels of processing recall of orthographically processed stimuli, years of education and ferritin. Many major variables violated the assumption of normality to some degree, but exhibited acceptable approximations of homogeneity of variance; therefore, they were considered candidates for parametric analyses. These variables included: CVLT Short Delay, Long Delay, Short Delay Cued Recall, Long Delay Cued Recall, Recognition Hits, Levels of processing Semantic recognition, estimation of frequency of occurrence and associated free recall and both scales from the Cognitive-Affective and Physical Behavior Questionnaire. Please see Appendix F, Table F-2. The Orthographic variable generated by the levels of processing task was normally distributed but failed the test of homogeneity of variance. Ferritin, Sum CVLT Clusters and years of education were neither normally distributed nor had equality of variance; therefore, parametric procedures were considered unsuitable for their analysis. With the

PAGE 76

68 exception of ferritin, all variables necessary for preplanned tests exhibited both reasonably good between group equality of variance and were normally distributed. As the effects of ferritin were judged to be nonlinearly related to ferritin concentration, it was decided to use ferritin as a classif icatory variable. To test the impact of the above violations of the assumptions of parametric procedures, parametric and nonparametric tests of between group baseline values were calculated for all subjects and the results compared. As displayed in Appendix E, even using an alpha of .05, only 2 tests out of 55 variables produced discrepant results. They were CVLT variables that may be expected to be relatively unstable: Short Delay Perseverations and Long Delay Intrusions. This test suggested that the violations of normality were unlikely to be of sufficient severity to impact on the validity of SAS General Linear Model parametric procedures. The statistical characteristics of the distributions and variance of all variables may be viewed in Appendix F. Tables F-1 and F-2. There was one debatable outlier. Select analyses were re-run following deletion of this subject. Examination of the raw data revealed that at the first assessment the subject had recalled 8 words on the first trial of the CVLT, but only 3 on the fourth trial and 5 on the fifth trial. Other data for this subject was within expected limits.

PAGE 77

69 Descriptive Statistics Age The treatment group had a mean age of 43.24 years (sd=13.45) and consisted of 17 subjects. The mean age of the 18 subjects in the control group was 48.22 years (sd=14.19). A nonpararaetric Kruskal-Wallis test procedure failed to demonstrate a statistically significant between group difference (X2(l, n=35)=1.13, p<.2948). Education Mean years of education in the treatment group was 11.41 years (sd=3.02) and 11.82 years (sd=2.15) in the control group. A Kruskal-Wallis test procedure failed to reveal a significant statistical differences between groups (X2(l, n=35)=0.23, e<.6325). Sex Overall, the subject population was equally divided by sex with 17 males and 18 females. The treatment group consisted of 7 males and 10 females. The control group consisted of 10 males and 8 females. Thus, the control group was 55.6% male and the treatment group was 41.2% male. A Chi-Square procedure failed to demonstrate a statistically significant difference (X2(l, n=35)= 0.724, 2< 0.395). Treatment Modality In the control group 72.2% were on some form of peritoneal dialysis (i.e.. Peritoneal Dialysis, Continuous Ambulatory Peritoneal Dialysis, or Continuous Cycler Peritoneal Dialysis). Five subjects or 27.8% of the control

PAGE 78

70 group were on maintenance hemodialysis. In contrast, 94.1% of the treatment group was on maintenance hemodialysis and one subject representing 5.9% of the treatment group was maintained on peritoneal dialysis. As confirmed by a ChiSquare procedure these differences were highly significant (X2(l, n=35)=16.032, E<.0001). Determination of time on dialysis was made difficult by intervening periods of renal transplantation as well as inconsistencies between medical records and patient reports. Also, this figure fails to take into account possible differences in the effects of hemodialysis and peritoneal dialysis over time. Mean years on dialysis for the treatment group was 6.64 years (sd=7.69) compared to 2.85 (sd=3.76) years for the control group. These differences approached but failed to reach statistical significance (X2(l, n=35)=2.8425, E<.0918) based on the Kruskal-Wallis Test procedure. Attrition All 17 subjects in the treatment group completed the study; however, one 25 year old male died six weeks after completion of the third assessment. Two subjects in the control group, both on hemodialysis, died between the second and third assessments. In addition, one control subject maintained on peritoneal dialysis declined to complete the third assessment after becoming hospitalized in a neighboring city. Eliminating the three subjects not completing the study resulted in a mean age in the control

PAGE 79

rr71 group of 46.49 years (sd=13.61) and a mean of 12.10 (sd=1.63) years of education. The results of a Wilcoxon 2Sample Test (S=267.00, Z=0.7177, e>.4729) failed to demonstrate a significant age difference between the treatment (n=17) and control (n=15) groups. Likewise, the Wilcoxon 2-Sample Test procedure failed to demonstrate a significant difference in years of education between the treatment (n=17) and control (n=15) groups (S= 259.00, Z=0.440, E>.6596) . Cognitive function has been reported to fall prior to nonaccidental death (White & Cunningham, 1988) . Consistent with this phenomenon of terminal decline, all three subjects were in the predicted direction, although in two cases the drop was extremely small. For the three deceased subjects, mean Sum CVLT in the last assessment prior to death was 25.56 compared to 28.70 in the previous assessment. The overall effect of terminal drop was to slightly lower control group assessment two and three scores relative to the treatment group. Time Intervals Between Assessments The mean time span between assessments one and two was 51.12 days (sd=32.95) for the treatment group. Control group mean time interval between the first and second assessments was 56.50 days (sd=35.98). The 5.38 days greater time interval for the control group failed to reach statistical significance based on a Wilcoxon Rank Sum Test (S= 277.500, Z=-. 92647, E>=.3542).

PAGE 80

^ 72 Mean days between assessments two and three was 67.80 (sd=34.56) in the control group and 90.41 days (sd=24.03) in the treatment group. Based on the results of a Wilcoxon Rank Sum Test the 22.61 days greater inter-test interval for the treatment group was clearly statistically significant (S=178.50, Z= -2.588, E>=.0096). The number of days between assessments one and three was 141.53 days (sd=34.53) in the treatment group (n=17) and 117.67 days (sd=28.24) in the control group (n=15) . The results of a Wilcoxon Rank Sum test suggest that the 23.86 day difference between these two groups was statistically significant (S=181.00, Z= -2.495, e>.0126). Mean time lag between initiation of rEPO and the first assessment was 41.4 days (sd=32.1) Hemoglobin Mean Hg at the first or baseline assessment was 8.46 g/dl (sd=1.42) in the treatment group and 9.44 g/dl (sd=1.92) in the control group. The nature of these differences suggested a possible trend toward statistical significance as demonstrated by a GLM Analysis of Variance procedure (F (1, 34)=2.94, £=.096). which is consistent with reports of lower Hg levels in hemodialysis than in peritoneal dialysis. Baseline mean hematocrit levels were 28.19% (sd=6.12) in the control group and 25.58% (sd=3.77) in the treatment group. As demonstrated by a GLM ANOVA procedure these differences failed to reach statistical significance (F(l, 34)=2.29, £=.140).

PAGE 81

73 As expected, a one-way GLM procedure demonstrated that between the first and second assessments the mean Hg rise of 1.13 g/dl in the treatment group (n=17) was significantly greater (F=7.42, £=.0102) than the -0.21 g/dl Hg change in the control group (n=18) . At the time of the second assessment mean Hg was higher in the treatment group (M=9.59, sd=1.84) than in the control group (M=9.23, sd=1.81) but the difference was not statistically significant (F(l, 34)= 0.34, £=.564). GLM analysis of variance of Hg change scores between the second and third assessments failed to reveal even a trend towards a between group difference (F(l, 31)= 0.16. E=.6964). Lack of change in the second phase of the study contributed to the failure to find no more than a trend towards a significant rise in Hg status between the first and third assessments (F=3.19, e=.0841). At the third assessment mean Hg was 9.54 g/dl (sd=1.48) in the treatment group (n=17) and 9.62 g/dl (sd=1.92) in the control group (n=15) . In this sample the 21 hemodialysis patients exhibited a mean baseline Hg of 8.80 g/dl (SD=1.80) and the 14 peritoneal dialysis patients had a mean Hg level of 9.21 g/dl (sd=1.69). The differences were not statistically significant (F (1, 34)= 0.44, £= .51). Within hemodialysis patients, mean baseline Hg in the controls (n=5) was 9.83 g/dl (sd=2.53). In the treatment group mean Hg at the first assessment was 8.48 g/dl (sd=1.48). Based on a T-Test

PAGE 82

74 procedure these differences failed to reach statistical significance (T=1.5099, £=.1475). Comparison of the maintenance hemodialysis pretreatment Hg of 7.06 g/dl (sd= 1.42) in the rEPO group (n=16) to the baseline Hg of 9.83 g/dl (sd=2.53) in the control group (n=5) revealed a trend toward lower values in the treatment group. However, a T-Test procedure failed to demonstrate a statistically significant difference between groups (T=2.3359, df=4.8, £=.0692). Prior to initiation of rEPO therapy, mean Hg in the treatment group was 7.39 g/dl (sd=1.94). At the first assessment treatment group Hg had risen to 8.46 g/dl (sd=1.42). Based on a T-Test these differences were not statistically significant (T=-1.8361, df=32.0, E=.0756). A GLM analysis of variance procedure revealed a main effect for sex (F (1, 34)= 15.24, £=.0004). Mean male (n=17) Hg was 9.96 g/dl (sd=1.61) and mean female (n=18) Hg was 8.03 g/dl (sd=1.31), which is consistent with the 2 g/dl lower Hg level reported in normal females compared to normal males in the general healthy adult population. There was no treatment by sex interaction. Excluding the 3 subjects who did not complete the study, rise in Hg was 0.04 g/dl in the control group (n=15) . As anticipated for the control group, mean change in Hg between assessments one and three was not statistically significant (F(l, 32)= 0.07, p=.786). In contrast the rise in Hg was 1.08 g/dl (sd=1.23) between assessments one and

PAGE 83

75 three in the rEPO treated group which was statistically significant (F(l, 33)= 4.74, £=.037). In the time lag between initiation of rEPO and the first assessment mean Hg levels rose 1.07 g/dl (sd=1.90) from an initial level of 7.39 g/dl (sd=1.94). The increase of 1.08 g/dl in the treatment group (n=17) between assessments one and three was 50% of the total increase of 2.15 g/dl (sd=1.23) in Hg over the course of rEPO therapy. However, this total increase of 2.15 g/dl was only 39% of the increase obtained in the pilot study. Please see Tables 4-1 and 4-2 for of Hg levels and change scores, respectively. For descriptive purposes subjects were classified using the following system: very severe anemia < 6 g/dl Hg, severe anemic 6-8 g/dl, moderate anemia 8-10 g/dl, mild anemia 10 12 g/dl, > 12 g/dl normal. As illustrated in Table 43, between the first and second assessments seven treated subjects showed an improvement in category and one declined. In the controls four rose and three fell. Between the second and third assessments change appeared to be random. Ferritin Because of the tendency of rEPO treatment to draw down iron stores ferritin levels were assayed more often in rEPO treated subjects; therefore, baseline ferritin levels were available for 94% (n=16) of the treatment group but only 33% (n=6) of the controls. At baseline mean treatment group (n=16) ferritin level was 545.93 ng/ml (sd=734.50) and mean

PAGE 84

76 control group (n=6) ferritin level was 523.62 ng/ml (sd=626.13). A Kruskal-Wallis Test failed to reveal a statistically significance difference between these two groups (X2(l, n= 35)= 0.3478, E<.5553). Ass Grp N Min Max Mean Sd pre trt rEPO 17 4.80 12.70 7.39 1.94 1 rEPO 17 5.60 11.30 8.46 1.42 2 rEPO 17 6.60 14.20 9.59 1.84 3 rEPO 17 6.90 12.25 9.54 1.48 1 Con 18 5.27 13.90 9.44 1.92 2 Con 18 6.30 12.60 9.23 1.81 3 Con 15 6.00 12.30 9.62 1.92 Note: Ass = Assessment; Grp = Group; N = Number of subjects Min = Minimum; Max = Maximum; Sd = Standard deviation Con = Control; rEPO = rEPO treated group Test Interval Grp N Min Max Mean -0.21 Sd First Con 18 -2.95 1.70 1.34 Second Con 15 -2.10 1.90 0.15 1.14 Pre-Ass rEPO 17 -4.60 4.70 1.07 1.90 First rEPO 17 -0.70 5.90 1.13 1.57 Second rEPO 17 -5.50 1.80 -0.05 1.59 Total rEPO 17 -2.40 5.30 2.15 1.86

PAGE 85

77 Table 4-3. Anemia Status Across Assessments Id Assess 1 1 Severe 2 Moderate 3 Moderate 4 Severe 5 Very Severe 6 Moderate 7 Moderate 8 Severe 9 Moderate 10 Moderate 11 Severe 12 Severe 13 Moderate 14 Moderate 15 Moderate 16 Moderate 17 Severe 101 Severe 102 Moderate 103 Very Severe 104 Moderate 105 Normal 106 Moderate 107 Normal 108 Moderate Assess 2 Assess 3 Moderate + Moderate Moderate Moderate + Severe + Normal + Severe Moderate + Moderate Moderate Moderate + Severe Moderate Moderate Moderate Moderate Moderate + Moderate + Moderate Severe + Moderate Normal Severe Moderate + Moderate Moderate Moderate Moderate Severe Severe Moderate Moderate + Severe Moderate Moderate Moderate Severe Moderate Moderate Normal + Moderate Moderate Moderate Moderate Severe • Moderate • Normal + Moderate

PAGE 86

78 Tabl e 4-3 continued Id Assess 1 Moderate Assess 2 Assess 3 Moderate 109 Moderate 110 Moderate Moderate Moderate 111 Moderate Moderate Moderate 112 Moderate Moderate Moderate 113 Moderate Moderate Moderate 114 Moderate Moderate Moderate 115 Moderate Severe Very Severe 116 Moderate Moderate Moderate 117 Severe Moderate + • 118 Moderate Severe Severe Note: : Valence sign = direction of categorical change; Id < 100 = treatment group; Id > 100 = controls As displayed in Table 4-4, subjects were classified based on ferritin level (under 10 Deficient, 10 30 Low, over 30 Replete) . Classifying subjects into Replete, Low and Deficient categories revealed little variability in iron status across assessments. Table Iron 1 4-4. Number Level of Subiects in each Cate icforv of Blood Iron Assess 1 Status Con rEPO Assess 2 Con rKPO Assess 3 Con rFPO Replete 14 11 14 9 12 11 Low 1 4 1 6 1 3 Deficient 1 1

PAGE 87

f ". '^^ 79 At the time of the first assessment, the Control group included one subject with Low iron status and 14 Replete subjects. At Assessment Two this was unchanged. At Assessment Three there was still one subject with Low iron status. In the rEPO group at Assessment One, there was one Deficient subject, four Low iron subjects and 11 Replete subjects. At Assessment Two, there were 6 Low iron and nine Replete subjects. By the third assessment, the treatment group was composed of one Deficient, three Low and 11 Replete subjects. For the subjects on which data was available there was minimal change in iron status; therefore, the need to control for iron deficiency was greatly reduced. Replete subjects were further divided into Replete and High (over 1,000). As illustrated in Table 4-5 below, the percentage of subjects exhibiting low iron status was consistently greater in the rEPO treatment group as was the percentage of subjects with high iron status. Also apparent from the Table 4-4, iron status was relatively stable across assessments, especially if the top two and bottom two groups were collapsed into Low and Replete categories. Baseline Neuropsychological Functioning Table 4-6 displays selected raw and standard scores for CVLT II at the baseline assessment. There are no statistically significant differences for any of the scores listed in Table 4-6 or in the more comprehensive list of scores in Appendix G. However, despite the lack of

PAGE 88

80 statistically significant differences, examination of raw scores suggests a tendency for somewhat better performance in the treatment group at the first assessment. Table 4-5. Percentage of Subjects in each Category of Iron Status Iron Assess Con 1 rEPO Assess Con 2 rEPO Assess Con 3 rEPO Hi 6.7 23.5 20.0 18.8 Normal 86.7 41.2 93.7 40.0 92.3 50.0 Lo 6.7 29.4 6.7 40.0 7.7 25.0 Defic 5.9 6.3 Note: Hi = Elevated Ferritin; Lo = Low; Defic = Deficient The mean baseline age corrected control group Sum CVLT score placed the group in the 4th percentile, the borderline range. The mean baseline Sum CVLT score in the treatment group placed it in the 9th percentile, the Low Average range. However, the differences, as previously stated, did not reach statistical significance. As displayed in Appendix G, Table G-1, on the Controlled Word Association test, using the letters CFP, the control group (n=17) produced a mean of 11.53 words (sd=3.84). The treatment group (n=17) produced a mean of 10.53 (sd=4.45) words. The difference was not statistically significant. The levels of processing task revealed that both groups benefited from semantic processing compared to either

PAGE 89

81 orthographic or phonemic processing. There was a very slight nonsignificant advantage for phonemic over orthographic processing for both groups. At baseline the treatment group appeared to demonstrate greater benefit from semantic processing than did the control group, although this did not reach statistical significance. Performance on the estimation of frequency of occurrence task was nearly identical between the two groups. '* * Control Group n=18 Mean Sd t-* ^-^ »-' J^^ fc3 d \^ rEPO Trea n=17 Mean ted Group Sd 2.47 P value Trial 1 5.72 1.74 5.88 .9464 standard -1.33 0.91 -1.35 1.37 .7700 Sum 1-5 43.22 10.70 45.82 13.03 .5222 T score 33.06 10.31 33.82 16.00 .8662 Short Delay 9.06 2.71 9.65 3.26 .9867 standard -1.11 0.83 -1.06 1.43 .9864 SD Cued 10.94 2.69 10.94 2.90 .7393 standard -0.83 1.04 -0.88 1.36 .9052 Long Delay 9.94 3.44 9.47 3.76 .6660 standard -0.89 1.02 -1.41 1.66 .4344 LD Cued 11.06 3.51 10.82 3.34 .5587 standard -0.78 1.40 -1.24 1.56 .3863 Note: Sd = Standard deviation; SD = Short Delay; LD = ] Jong Delay; standard = Standard Score

PAGE 90

82 Test of Main Hypothesis The main hypothesis stated that rEPO mediated reduction in chronic anemia in adult dialysis patients is associated with improved verbal LTM as measured by the CVLT Sum of trials one through five. Mean Sum CVLT change scores between the first and second assessments were -0.82 (sd=5.38) in the treatment group (n=17) and 4.06 (sd=9.91) in the control group (n=18) . These results were in the opposite of the hypothesized direction. A GLM analysis of covariance procedure failed to demonstrate a main effect for either treatment status (F=1.29, p=.2638) or Hg (F=1.32, £=.2590). With both treatment status and Hg change in the model, few conclusions could be drawn. Therefore, separate univariate GLM analyses were performed for treatment status and for Hg change. However, these analyses failed to show a statistically significant main effect for either Hg ((F=3.25, E=.0805) or for treatment status (F=3.22, E=.0818) . Examination of mean Sum CVLT scores between groups and across assessments revealed a curious pattern. Although none of the differences reached statistical significance, examination of Table 4-7 reveals that the control group performed poorly on the first assessment relative to the other two assessments. In contrast, assuming equivalence of test forms, treatment group performance was static. That is, change apparently occurred unrelated to rEPO.

PAGE 91

Con CVLT II 18 43.22 10.70 23 Con CVLT III 18 47.28 10.68 30 Con CVLT I 15 47.07 10.40 32 Trt CVLT II 17 45.82 13.03 21 Trt CVLT III 17 45.00 13.36 16 Trt CVLT I 17 45.12 11.61 23 83 Table 4-7. Mean Sum CVLT Across Assessments and Between Groups Grp Variable N Mean Sd Min Max 60 68 68 68 64 69 Representing iron status as a two-level classif icatory variable (i.e., Low vs. Replete), a GLM analysis of variance failed to demonstrate a main effect for iron status on Sum CVLT change scores between the first and second assessments (F=0.01, E=.9384). Similar results were obtained introducing Hg into the model (Hg: F=0.26, £=.6126; iron: F=0.01, E='9337) . Despite the lack of change in Hg status between the second and third assessments, a GLM Repeated Measures ANOVA was performed. As expected, the hypothesis of a treatment effect was not supported by Wilks' Lambda (F=0.2134, E=.8092) or any other measure. Neither were there between subject effects, within subject effects or treatment by assessment interactions. Post Hoc Analyses Learning Curve Hypothesizing that change in learning curve might be more sensitive to the effects of anemia than Sum CVLT, SAS

PAGE 92

84 Regression procedures were used to generate slopes of the learning curves for trials one through five for each administration of the CVLT. A GLM procedure analyzing change scores between the first and second assessments failed to demonstrate a relationship between Hg change and change in slope (F=0.08, E=.7766) . An identical analysis examining change scores between the second and third assessments produced similar results (F=0.00, E=.9583). A GLM univariate analysis of the effect of treatment status on slope change scores between the first and second assessments failed to reach statistical significance (F= 3.67, £=.0651). Controlling for the increased probability of a Type I error as a result of multiple analyses, the finding can not be considered suggestive of a trend. A Wilcoxon Rank Sum Test produced similar results (Z= 1.83, £=.0669). Further analysis revealed the change to be primarily in the treatment group, but in the opposite of the expected direction. In the treatment group, mean slope declined -0.84 (sd=1.59), and in the control group slope improved 0.15 (sd=1.27) between the first and second assessments. A GLM univariate analysis of the effect of treatment group status on slope change scores between the second and third assessments failed to suggest a relationship (F=1.30, £=.2628). ' Despite extensive analyses, absolutely no evidence was found that even mildly suggested improvement in learning curve in association with improvement in Hg status.

PAGE 93

r< '* ^.,'i^-' -i 'm 85 Exploratory Analyses Observation of variability in Hg change scores in both groups led to reclassification based on direction of Hg change, rather than treatment status. A total of 23 subjects exhibited a rise in Hg between the first and second assessments (M=1.22 g/dl, sd=1.26). Twelve subjects exhibited a fall between the first and second assessments (M=-1.05, sd=.96). The difference between groups with rEPO treatment status as the classif icatory variable produced a mean difference in Hg change between groups of 1.03 g/dl. In contrast, the alternate system of classification doubled the Hg change score yielding 2.27 g/dl. Difference in Hg change scores between the positive and negative change groups was highly significant (F=29.75, £=.0001). Control subjects suffering a fall in Hg (n=9) exhibited a mean fall of -1.27 g/dl (sd=1.02). The three rEPO treated subjects exhibiting a fall in Hg had a mean drop of -.38 g/dl (sd=.28). Among rEPO treated subjects with rising Hg (n=14) the mean rise was 1.46 g/dl (sd=1.54). Subjects from the control group exhibiting a rise in Hg (n=9) showed a mean rise of 0.85 g/dl (sd=0.51). For exploratory purposes, the main hypothesis was retested with a GLM procedure utilizing sign of Hg change as the classif icatory variable. There was a significant difference in Sum CVLT change scores between the positive (n=23, M=-1.22, sd=7.18) and negative (n=12, M=7.25, sd=7.63) Hg change scores (F=10.52, p=.0027). The analysis

PAGE 94

86 was rerun excluding one possible outlier; however, this had little effect on the results (F=7.82, £=.0087). The results were consistent with the original finding (i.e., using treatment status as the classificatory variable) in that they were in the opposite of the predicted direction. In a GLM multiple analysis of variance treatment status was nonsignificant (F=0.59, p=.4482) while sign of Hg change was significant (F=9.45, e=.0043). Also consistent with the original finding, as illustrated in Table 4-8 below, mean Sum CVLT scores revealed poorer performance at the time of the first assessment in the group whose Hg levels would rise. Table 4-8. Mean Sum CVLT by Sicfn of Hg Change Between Assessments 1 & 2 Asses 1 2 Hg Fell n= =12 Hg Rose n=2 3 Mean Sd Mean Sd 36.58 9.48 48.61 10.86 43.83 11.80 47.39 12.07 In an attempt to reduce error variance contributed by subjects exhibiting very small insignificant changes in Hg, subjects with a Hg change of under 1 g/dl were excluded from the analysis. Mean Sum CVLT change in the negative Hg change group was 9.3 3 (sd=10.63) and -2.23 (sd=4.3 0) in the positive Hg change group. A Wilcoxon Rank Sum Test revealed the difference to reach statistical significance at the .05 level (Z=2.373, £=.0176).

PAGE 95

87 To test the effect of Hg changes within the range closer to that likely to be considered physiologically relevant, subjects with less than a 2 g/dl Hg change were excluded from the final analysis. The two negative Hg change subjects in the data set had a mean Hg drop change of 2.68 g/dl. The three positive subjects in the data set exhibited a mean Hg rise of 3.77 g/dl. A GLM analysis of this data set (n=5) failed to reveal a main effect for Hg (F= 0.03, E=.8803) ; however, the small sample size greatly reduced the power of the test and, again, the change was in the opposite of the predicted direction. Mean CVLT improvement was 2 words in the falling Hg group and the rise was 1.33 words in the rising Hg group. Hypothesizing that the observed effect of Hg change on Sum CVLT might be mediated by affective change, the Cognitive-Affective self-report variable was included in a GLM analysis of variance model. A multivariate analysis demonstrated main effects for both direction of Hg change (F=13.27, E=.0010) and Cognitive-Affective self-report (F=8.29, E=.0072). Further analysis revealed a main effect for direction of Hg change (F=16.97, p=.0003) and an interaction between direction of Hg change and self -reported Cognitive-Affective behavior (F=6.53, £=.0043). When Cognitive-Affective behavior was included in this model it failed to reach significance (F=1.88, £=.1807). Further analyses suggested no relationship between CognitiveAffective behavior and sign of Hg change (r=. 11056,

PAGE 96

88 E=.5336). Correlation between Sum CVLT and CognitiveAffective self-report was positive (r=. 37042, e='0310). Examination of mean scores by groups revealed that Sum CVLT rose in association with rise in Cognitive-Affective behavior; however, there was less evidence of a relationship when Hg fell. In a model analyzing change in Sum CVLT, covarying change in self-reported Physical Activity and direction of change in Hg failed to demonstrate a main effect for Physical Activity (F=0.77, £=.3866). There was no interaction between sign of Hg change and Physical Activity (F=0.75, 2='4827). Neither was there an effect on Sum CVLT for self-reported level of Physical Activity at the time of the first assessment. Analyses failed to demonstrate a significant relationship between self-reported Physical Activity and Sum CVLT • Sum of CVLT Trials Four and Five Using sign of Hg change scores as the classif icatory variable, a Wilcoxon 2 -Sample Rank Sum Test was used to test the post hoc hypothesis of a change in the sum of CVLT trials four and five between the first and second assessment. The difference between groups was significant even controlling for multiple comparisons (Z=2.6366, E=.0084). Once again the difference was the reverse of the predicted direction. Mean increase in number of words recalled for the sum of trials four and five was 3.75 (sd=5.24) in the group with falling Hg (n=12) and -0.87

PAGE 97

89 (sd=4.00) in the group with rising Hg (n=23) . Please see Table 4-9. Table 4-9. Mean Chancre in CVLT Trials 4 plus 5 Change Scores Between Assessments I & II Grp N Min Max Mean Sd Pos 23 -8.00 6.00 -0.87 4.00 Neg 12 -4.00 18.00 3.75 5.24 An identical analysis between the second and third assessments revealed a mean change of -0.78 words (sd=3.93) in the negative change group (n=9) and a mean change of -0.15 words (sd=3.96) in the positive Hg change group (n=20) . These differences failed to reach statistical significance (Z=-.2845, e=.7760) . Analysis of change scores between the first and third assessments revealed a mean change of -3.71 words (sd=7.06) in the negative change group (n=7) and a mean change of -1.25 words (sd=4.23) in the positive Hg change group (n=24) . These differences failed to reach statistical significance (Z=1.5613, e=.1184). Multivariate analysis (n=35) with a General Linear Model failed to find a main effect for rEPO treatment (i.e., experimental vs. control) on Sum CVLT (F=0.84, £=.3668) in a model including a sign of Hg change factor (F=7.45, E=.0102) . To test the possibility that undiscovered outliers unduly influenced the results, Sum CVLT change scores were

PAGE 98

90 classified based on their sign. A Wilcoxon Rank Sum procedure analyzed differences in sign of Sum CVLT change scores between the first and second assessments. With the sign of Hg change scores serving as the grouping variable, again there was a significant effect for Hg change (Z=2.9180, E=.0035). Analysis using a Chi-Square procedure produced similar results (X2=8.887, df=l, e=.003). Short Term Memory Exploratory analysis revealed a significant difference at .01 in CVLT Trial One, a measure loading heavily on STM. However, the change was in the opposite of the expected direction between treatment and control groups between assessments one and two and is likely a type two error. Furthermore, equivalent change, also significant at .01, in the opposite direction, was observed between the second and third assessments. Please see the results displayed in Appendix H. Verbal Fluency An analysis of variance procedure suggested that verbal fluency as measured by Controlled Word Association test performance was somewhat low in the control group at the first assessment relative to the overall pattern of performance in both groups across three assessments (F=4.99, E=.0328. Utilizing sign of Hg change as a classif icatory variable. Controlled Word Association test performance was not significantly different between assessments (F=0.01, E=.9761).

PAGE 99

91 Secondary Hypotheses Levels of Processing Treatment status was insignificant in explaining variability in performance on any portion of the Levels of Processing Task, including the Semantic recall measure (F=0.43, E=.5164). Mean change scores are listed in Appendix H. Re-analysis using direction of Hg change as a classif icatory variable produced similar results (F=2.12, E=.1564). This analysis would have been considered a secondary hypothesis if there had been a positive association between Hg and Sum CVLT. Estimation of Frequency of Occurrence As displayed in Appendix H, there were no significant differences in Frequency Estimation performance between groups or assessments. Parenthetically, LTM, as measured by free recall of the single presentation target word list, improved slightly more in the treatment group than in the control group; however, this failed to reach significance controlling for multiple analyses. Quality of Life Quality of life is operationally defined here in terms of the items on the Cognitive-Affective and Physical Behavior Questionnaire and is divided into cognitiveaffective and overt physical behavior domains. It was hypothesized that self-reported physical vitality, as measured by the self -report questionnaire, would improve in response to treatment with rEPO.

PAGE 100

92 Because Physical Activity was not normally distributed, a Wilcoxon Rank Sum Test procedure was used. Analysis failed to support the hypothesis of a main effect for treatment status on Physical Activity between the first and second assessments (Z=1.1283, e=.2592). Similar results were obtained using Hg change as the independent variable. It was hypothesized that self-reported behavioral indicators of Cognitive-Affective state would improve in the treatment group. A Wilcoxon Rank Sum Test procedure failed to demonstrate a change in Cognitive-Affective behavior (Z=1.7654, p=.0775) between the first and second assessments. Because the results could arguably be interpreted as failing to reject the possibility of a trend, the direction of the changes was examined. As displayed in Table 4-10 below change between the first and second assessments was in the opposite of the predicted direction. Direction of change between the second and third assessments was in the predicted direction. Analysis of change scores between the first and second assessments using a GLM procedure failed to demonstrate a main effect for Hg (F=0.40, £=0.5336). Using sign of Hg change scores as the classificatory variable, a Wicoxon Rank Sum test failed to reveal statistically significant differences in self-reported Physical Activity (Z= -1.0235, e=-3061). Neither were there statistically significant differences in self-reported behavior in the Cognitive-Affective domain (Z= 0.2028, E=.8393).

PAGE 101

93 Table 4-10. Self-reported Cognitive-Affective State Change Scores by Group & Assessment Interval Test Control Group Treatment Group Interv N Mean Sd N Mean Sd One 18 0.94 2.21 16 -0.19 1.33 Two 15 -1.73 2.02 16 -0.13 1.41 Note: Interv = Change between assessments one and two or between two and three. Test Reliability Two alternate forms of the Levels of Processing Task and Frequency of Occurrence Task were developed specifically for this study. A self-report questionnaire was written to sample physical and cognitive-affective domains thought likely to be affected by change in severity of anemia. None of these new measures underwent a prior validation study. Also, one alternate form of the CVLT was developed and validated for use in this study, but validated on a younger and better educated population. Because the only apparent changes in test performance were in the controls, controls which might have been used in a test validation study, the lack of clear-cut findings produced the opportunity to examine the reliability of these instruments in a medical population. CVLT Alternate Form Reliability As displayed in Table 4-11, for the overall group the correlation between Forms I and III and between Forms III and II was quite good relative to correlations between Forms

PAGE 102

94 I and II. Closer examination revealed a similar pattern for the treatment group. This pattern is displayed in Table 412. However, as shown in Table 4-13, the control group exhibited poorer correlations between Form II and Form III and between Form II and Form I than did the treatment group. Table 4-11. Correlation of Sum CVLT Between Forms I. II and III for All Subjects Pearson Correlations CVLT II CVLT III CVLT I CVLT II r 1.00000 0.75543 0.71063 p 0.0000 0.0001 0.0001 N 33 35 32 CVLT III r 0.75543 1.00000 0.80342 P 0.0001 0.0000 0.0001 N 35 .35 32 CVLT I r 0.71063 0.80342 1.00000 P 0.0001 0.0001 0.0000 N 32 32 32 Note: r = correlation; p = probability value; N = subject number ..,* Table 4-14 demonstrates that Form III, the version created for this study, correlates as well with both Forms I and with II as Form I correlates with Form II. In addition, means scores were remarkably similar between the three forms of the CVLT. " "

PAGE 103

95 II and III Pearson Correlations r CVLT II CVLT III CVLT I CVLT II 1.00000 0.91732 0.77220 P 0.0000 0.0001 0.0003 N 17 17 17 CVLT III r 0.91732 1.00000 0.81127 P 0.0001 0.0000 0.0001 N 17 17 17 CVLT I r 0.77220 0.81127 1.00000 P 0.0003 0.0001 0.0000 N 17 17 17 Note: r = correlation; p = probability value Table 4-13. Correlation Between Sum CVLT I. II and III for Controls Pearson Correlations CVLT II CVLT III CVLT I CVLT II r 1.00000 P 0.0000 N 18 CVLT III r 0.57038 P 0.0134 N 18 CVLT I r 0.64116 P 0.0100 N 15 0.57038 0.0134 18 1.00000 0.0000 18 0.78866 0.0005 15 0.64116 0.0100 15 0.78866 0.0005 15 1.00000 0.0000 15 Note: r = correlation; p = probability; N = subject number

PAGE 104

96 Table 4-14. Correlations & Sum CVLT Means Between CVLT Forms I, II and III Variable Pearson Correlat ions Form Mean Sd I-II I-III II-III Trial 1 r .48240 .59965 .47084 I 5.800 2.098 P .0033 .0003 .0065 II 5.800 2.139 N 35 33 32 III 5.938 1.795 Trial 2 r .56327 .60335 .61043 I 8.143 2.171 P .0004 .0003 .0002 II 8.629 2.377 N 35 32 32 III 8.969 2.957 Trial 3 r .59901 .66308 .65356 I 9.400 3.098 P .0001 .0001 .0001 ' II 9.943 2.786 N 35 32 32 III 10.063 2.213 Trial 4 r .42348 .47538 .64775 I 10.286 2.906 P .0112 .0060 .0001 II 10.800 3.095 N 35 33 32 III 10.344 2.925 Trial 5 r .64340 .52880 .75095 I 10.886 3.132 P .0001 .0019 .0001 II 11.086 3.128 N 35 33 32 III 11.031 2.753 Sum r .75543 .71063 .80342 X 44.486 11.788 P .0001 .0001 .0001 ZI 46.171 11.930 N 35 32 32 III 46.031 10.926 Sum Clus r .64837 .80121 .73281 I 12.229 9.540 P .0001 .0001 .0001 II 13.143 9.638 N 35 33 32 III 13.781 10.222 Sum Psv r .13204 .27777 .07528 I 3.343 2.828 P .4496 .1237 .6822 II 3.800 3.261 N 35 33 32 III 4.219 2.685

PAGE 105

97 Table 414 continued > -: . -' '.' :-': Variable Pearson Correlations I-II I-III II-III Form Mean Sd Sum Intrus r P -0.001 0.996 .02699 .8834 .50981 .0029 I II 0.857 1.771 1.912 2.510 N 35 32 32 III 1.906 3.104 List B r .40311 .54579 .33823 I 5.857 1.768 P .0181 .0012 .0583 II 5.588 2.176 N 34 3f 32 III 5.594 2.434 Short Delay r P .67782 .0001 .73638 .0001 .72612 .0001 I II 9.343 9.353 2.960 3.524 N 34 32 ^2 III 9.281 3.113 Short Delay Cued r P .72150 .0001 .67170 .0001 .58327 .0005 I II 10.943 10.559 2.754 2.977 K 34 32 32 III 10.344 2.598 Long Delay r P .76124 .0001 .79658 .0001 .77339 .0001 I II 9.714 9.471 3.553 3.703 N 34 32 32 III 9.469 3.360 Long Delay Cued r P .65964 .0001 .70126 .0001 .56638 .0007 , I tt 10.943 10.529 3.378 2.926 N 34 r^ -^ ^ . H**'' III 10.500 2.712 Recog True Positive r P .30641 .0780 .44776 .0102 .31827 .0759 I II 14.800 14.324 1.324 1.492 N 34 ' 312 ^32^'III 13.813 2.546 Recog False Positive r P .77802 .0001 .73247 .0001 .67942 .0001 I II 2.543 3.236 3.576 3.593 N 34 32 32 III 3.000 3.312

PAGE 106

98 Table 4-14 continued Note: Under Correlations, the first value listed is the correlation (r) under that the p value, and then, N, the number of subjects; Psv = Perseveration; Intrus = Intrusions; Recog = Recognition Levels of Processing Reliability Forms II and III of the Levels of Processing Task were developed for use in this study. As illustrated in Table 415, correlations between recall of orthographically processed words on Levels of Processing Forms I, II and III were all positive, but low and nonsignificant. Correlations between phonemically processed stimuli followed an identical pattern between Forms I and II and I and III. However, the correlation between Forms II and III reached statistical significance. The opposite was true of the semantically processed words. Correlations between Forms I and II and Forms I and III reached statistical significance. Frequency of Occurrence Reliability Table 4-16 illustrates a finding identical to that of the Semantic Levels of Processing results, in that correlations reached statistical significance between Forms I and II and between Forms I and III, but not between II and III. Physical Behavior Self-Report Test-Retest Reliability The data displayed in Table 4-17 demonstrates quite acceptable test-retest reliability for the Physical Behavior scale. Correlation coefficients ranged from r=.820 to

PAGE 107

99 r=.898. Examination of the means and standard deviations shows essentially unchanged scores between administrations, Table 4-15. Correlations and Means Between Levels of Processing Test Forms I. II & III Correlations I & II I & III II & III Form Orthoaraphic r .28992 .12925 p .1136 .4961 N 31 30 Phonetic r .23765 .26180 p .1980 .1623 N 31 30 Semantic r .40394 .56396 p .0242 .0012 N 31 30 .23972 .2020 30 .54589 .0018 30 .16411 .3862 30 I II III I II III I II III Mean 7.531 6.548 6.967 9.938 10.323 9.800 Sd 2.155 2.336 2.498 7.625 2.268 8.903 2.371 8.367 2.220 1.831 1.796 1.472 Note: r = correlation; p = probability; N = subject number Table 4-16. Correlations & Means Between Estimation of Frequency of Occurrence Test Forms I. II & III Pearson Correlations between Forms I & II I & III II & III r .51046 p .0020 N 34 .43870 .0120 32 .10965 .5502 32 Form I II III Mean 11.343 10.471 9.875 Sd 2.029 3.184 2.871

PAGE 108

100 Table 4-17. Test Retest Reliability of Physical Activity Self -Report Pearson Correlations between Forms I & II I & III II & III Form Mean Sd r .89863 .76990 .82068 I 9.230 4.538 p .0001 .0001 .0001 II 9.029 4.596 N 34 32 31 III 9.031 4.575 Note: r = correlation; p = p value; N = sample size Table 4-18. Test Retest Reliability of Cognitive-Affective Self-Report Pearson Correlations between Forms I & II I & III II & III Form Mean Sd r .64224 .86002 .68524 I 5.943 2.071 p .0001 .0001 .0001 II 6.412 2.388 N 34 32 31 III 5.469 2.185 Note: r = correlation; p = p value; N = sample size Cognitive-Affe ctive Behavior Self-Report Test-Retest Reliability Table 4-18 illustrates significant test-retest correlations for a series of true or false questions focusing on cognitive-affective state and behavior. However, comparison with the physical behavior data revealed correlations tended to be higher and means less variable when focusing on physical symptoms and behavior. Post Hoc Ana lysis of Blood Urea Nitrog en After completion of data collection and analysis it was hypothesized that the inverse relationship between Hg and *-'.A > ^ k >?&i i*i

PAGE 109

101 Sum CVLT observed between the first and second assessments might be due to an increase in severity of uremia as measured by blood urea nitrogen (Bun) . In the hemodialysis population, blood values were necessary for the day of the first and second assessments. However, Bun levels were unavailable for the dates in question. In the peritoneal dialysis group, despite a high proportion of subjects having been tested at home, BUN values were available on or near the dates of the first and second assessments for 8 subjects. Although the power of the test was unacceptably low due to the small number of subjects, a univariate analysis of variance was performed. This procedure failed to demonstrate a relationship between Sum CVLT and BUN (F=0.77, p=.4139). However, there was no demonstrable relationship between Hg and Sum CVLT in this subsample of subjects (F=0.01, E=.9356). Summary Despite extensive post hoc analyses, no evidence of improvement in memory, self -reported cognitive-affective or physical functioning was demonstrated in the rEPO treatment group relative to controls. What appeared to be a trend towards a decline in Sum CVLT performance in rEPO treated subjects was found to represent poorer performance on the assessment prior to a drop in Hg. This phenomenon was observed primarily in a subset of the controls, but was not evident until subjects were reclassified based on direction of Hg change between the first and second assessments.

PAGE 110

102 Reclassification based on direction of Hg change revealed that Sum CVLT performance improved when Hg fell between the first and second assessments. Post hoc analysis revealed a main effect for direction of Hg change (F=16.97, E=.0003) with an interaction between direction of Hg change and self-reported Cognitive-Affective behavior (F=6.53, E=.0043). Decline in cognitive-affective functioning was associated with decline in list learning performance. The changes in Hg were of dubious physiological significance; indeed, the main hypothesis was not adequately tested due to the small change in Hg. Hypothesizing that increased uremia may have produced the poor performance, post hoc data collection was initiated. Despite insufficient sample size, analyses were performed to explore the possibility of a trend; however, there was no evidence of a relationship between Bun and Hg ((F=0.77, e=.4139). The lack of what is considered to be physiologically significant change in Hg led to the utilization of this data set to assess test-retest and/or alternate form reliability for several instruments. Consistent with the validation study, the third form of the CVLT developed for this project was found to be of essentially the same level of difficulty as Forms I and II. Levels of correlation between CVLT III and the other two forms were quite acceptable. In contrast, the two alternate forms of the levels of processing task and the two alternate forms of the frequency estimation task exhibited relatively disappointing levels of correlation

PAGE 111

103 with the original forms. However, the Physical Activity scale on the self-report questionnaire developed for this study exhibited excellent test-retest reliability. Testretest reliability was somewhat lower for the CognitiveAffective scale.

PAGE 112

CHAPTER 5 DISCUSSION Overview A broad spectrum of cognitive abilities have been reported to be deleteriously affected by chronic renal disease. Despite considerable effort, the specific etiologic mechanisms underlying the neuropsychological deficits associated with end-stage renal disease have not been satisfactorily explained. Anemia, often severe, almost invariably accompanies chronic renal failure. One purpose of this project was to test the hypothesis that chronic anemia may contribute to the neuropsychological deficits associated with end-stage renal disease. Chronic anemia is one of the most common medical conditions in the world, affecting diverse populations, particularly including infants, women, the elderly and the malnourished as well as specific disease populations. Despite the high prevalence and longstanding recognition of the often dramatic effects of acute anemia on cognition and brain function, the neuropsychology of chronic anemia has been largely ignored. Indeed, presumably because of the difficulty in disentangling anemia from its etiology, the question of whether chronic anemia has neuropsychological consequences has remained unanswered. Elucidation of the 104

PAGE 113

^-* 105 relationship between chronic anemia and cognition would be potentially relevant in diverse areas of research, public policy and clinical practice. Therefore, a second goal of this study was to attempt to speak to the question of whether chronic anemia affects neuropsychological functioning. The primary contribution to uremic anemia is the failure of diseased kidneys to perform their normal role of producing the erythropoiesis regulating hormone EPO. Recently approved for clinical use, rEPO is capable of effectively eliminating uremic anemia in the majority of patients. Indeed, in most patients Hg may be maintained at any desired level. Limitations on the target level of Hg are a function of economics and severity of deleterious side effects. Recombinant EPO is quite expensive. Increasing hematocrit in association with rEPO treatment has been associated with side effects including hypertension. At present, there is insufficient data to engage in cost-benefit and risk-benefit analysis. Therefore, it was hoped that this study would generate information regarding the relationship between level of Hg and neuropsychological functioning to assist in riskbenefit analysis. Prior to the development of rEPO, since anemia was nearly universal in chronic renal disease, any attempt at clarification of the relative contributions of anemia and uremia had been a most difficult task. Although several

PAGE 114

106 studies controlled for Hg, possible confounds such as level of physical exercise, degree of uremia, extent of uremic anorexia and years on hemodialysis, as well as additional sources of error such as the inability to use subjects as their own controls, may have hidden the effect. The development of rEPO, because it presumably alleviates uremic anemia without directly affecting the brain, suggested a method of disassociating anemia from the usual confounds and presented the opportunity to use subjects as their own controls. Thus, rEPO treatment was seen as a powerful new method for studying the neuropsychology of chronic anemia. Several lines of evidence hinted at a possible relationship between anemia and cognitive dysfunction. Neuropsychological function has in some studies been reported to be somewhat better in peritoneal dialysis than in hemodialysis. Although this has been attributed to greater middle molecule clearance in peritoneal dialysis, the offending middle molecule has never been found, and Hg levels tend to be higher in peritoneal dialysis. Also, years on hemodialysis has been reported to be associated with decreasing cognitive function and increasing anemia. The dramatic reduction in anemia produced by treatment with rEPO has been shown to produce clinically significant improvements in domains such as aerobic capacity, anaerobic threshold, sexual potency, employment status, level of social activity, perceived ability to engage in physical

PAGE 115

107 exercise and improved appetite, all previously attributed to uremia. Despite predictions, compared to these findings, pilot work with small numbers of rEPO treated renal patients has been disappointing in that there have been few improvements in neuropsychological functioning, no consistently reported improvements and failures to replicate. In our pilot work (Klein et al., 1989), the only statistically significant change was in LTM as measured by multitrial supra-span list learning. Despite the danger of a Type I error posed by multiple analyses, credence was placed in this finding because in the past multitrial list learning had proven particularly sensitive to treatment changes in renal disease. Therefore, the present study sought to replicate and extend this finding. Data analysis focused on change between the first and second assessments because preliminary analyses revealed no change in Hg between the second and third assessments. Univariate analysis of variance of Sum CVLT change scores failed to confirm either the hypothesized main effect for Hg (F=3.25, E=0.0805) or a main effect for treatment status (F=3.22, E=.0818). Combining treatment status and Hg level in the same model eliminated the putative trend. After discussing possible reasons for the discrepancy between the pilot and the present study, post hoc analyses utilizing Hg change as a classif icatory variable will be discussed. Examination of the literature on the impact of Hg levels on cognitive function suggested that the Hg change

PAGE 116

108 scores obtained in this study were not within the range that would be expected to be physiologically significant. Self-report data regarding Physical Activity had been obtained so that, in the event of negative findings, a determination could be made as to whether Hg levels had been adequately manipulated to produce improvement in life functioning. Consistent with the small change in Hg, there were no between group differences in change scores for self-reported Physical Activity or for Cognitive-Affective behavior. In comparison to the mean Hg change of 1.13 (Sd=1.90) g/dl in the present study, mean treatment group Hg rose 5.46 (Sd=1.72) g/dl in the pilot study. Indeed, despite a statistically significant change in Hg, (F=7.42, £=.0102), the independent variable does not appear to have been sufficiently manipulated; therefore, the main hypothesis was not adequately tested. In addition, the absence of change in Sum CVLT in the present study, in conjunction with a smaller rise in Hg, suggests that Hg change may not have been sufficient to provide clinical benefit. Thee are other possible explanations for the discrepancy between the pilot and the present study. Not only was change greater, but Hg levels began lower in the pilot study (M=6.10, Sd=.97 g/dl) than in the present study (M=8.46, Sd=1.42 g/dl). The physiological effects of varying degrees of severity of chronic anemia may be nonlinear. This would suggest that one unit of Hg change

PAGE 117

109 may be of greater physiological significance in more severely anemic individuals. Conceivably, change in Hg equivalent to that observed in the present study, but occurring at a lower baseline level, might have resulted in significant cognitive effects. The present study may have failed to demonstrate cognitive change because baseline Hg levels were above the threshold for cognitive dysfunction. Indeed, based on the literature review, it is quite conceivable that the nervous system is able to adapt to all but severe anemia. Despite plausible explanations for a failure to replicate, re-examination of the pilot data in the light of present findings suggests the possibility that the results of the pilot study may have been spurious. In the pilot study, four of the five treated subjects had Hg increases of over 6 g/dl. In contrast, the greatest Hg rise in the present study was 5.90 g/dl. However, significance was reached in the pilot study only with the contribution of the putative relationship between negative LTM change scores in the control group and negative Hg change in the same two subjects. The greatest fall in Hg was 0.93 g/dl. The current study had 19 subjects with changes in Hg of 1.0 g/dl or greater between assessments one and two. Re-examination of the Pilot study data revealed that mean change in the control group was -5.5 words recalled over five trials and 5.6 words in the treatment group. Thus, nearly 50% of the change in the

PAGE 118

110 dependent variable occurred in the control group. Although change in the control group might ordinarily be attributed to error variance, clouding the issue, both control subjects exhibited small drops in Hg over the course of the study, .93 g/dl and .55 g/dl, respectively. Therefore, the drop in Sum CVLT in the control group was consistent with the hypothesized relationship between Hg and LTM. Interestingly, Hg changes of this magnitude were nonsignificant in the present study. Furthermore, a rise of 6.3 g/dl, an order of magnitude greater Hg change, resulted in no LTM improvement in one of the five experimental subjects. The mean Sum CVLT change in the pilot study was 2.43 (Sd=7.63). Mean CVLT change in the present study was 1.69 (Sd=8.30) between the first and second assessments and -0.06 (Sd=7.34) between the second and third assessments. Variability was high and of similar magnitude in both studies. In the pilot study, if the high correlation between Hg and Sum CVLT change scores is to be accepted, then a substantial portion of the variance in the Pilot study was due to changes in Hg. Curiously, if one is to accept Hg change as a primary source of variance, despite smaller changes in Hg the present study, both studies exhibited similar levels of variance in the dependent measure. Excellent CVLT alternate form reliability makes it unlikely that the utilization of alternate forms in the present study introduced significant additional variance.

PAGE 119

Ill In summary, insufficient magnitude of Hg change, too high an initial level of Hg in the present study or a combination of both conditions may have resulted in the discrepancy between the pilot and the present study. However, closer examination of the pilot data in the context of the present study suggested that the results of the pilot could have been spurious. The present data set provided insufficient information to determine which hypothesis might account for the discrepancy between the present and the pilot studies. Therefore, subjects were reclassified to maximized between group differences in Hg levels. Observation of change in Hg levels in both groups suggested the possibility of reclassification of subjects based on Hg change scores to maximize between group differences. The symptoms of acute anemia are quite variable at a given Hg level and, if chronic anemia followed a similar pattern, then, arguably, collapsing Hg scores into a two-level classif icatory system might potentially reduce error variance. Unfortunately, only a small number of subjects exhibited Hg change scores of sufficient magnitude to justify classification based on differences of demonstrated physiological significance. Therefore, the first analysis simply reclassified based on sign of Hg change. Reclassification based on sign of Hg change doubled the mean Hg change score and resulted in a statistically

PAGE 120

112 significant main effect for sign of Hg change on Sum CVLT (F=10.52, E='0027). However, contrary to expectations, the data indicated a negative relationship between rise in Hg and Sum CVLT. Between the first and second assessments, mean Sum CVLT change score fell 0.82 (Sd=5.38) in the rising Hg group and improved 4.06 (Sd=9.91) in the falling group. Physiological, psychometric and psychological explanations for the apparent finding will be discussed. The creation of a classif icatory variable utilizing direction of Hg change increased the magnitude and significance of group differences in Hg change and provided, arguably, an indirect means of further characterizing the effects of rEPO on memory function. This was accomplished both by including treatment status in a multivariate model and by somewhat more adequately testing the effects of the only known consequence of rEPO therapy, increased erythropoiesis. Recombinant EPO has never been specifically tested clinically for deleterious neurobehavioral effects. The drug may eventually be administered to several hundred thousand people per year. Therefore, potential danger signals, however improbable, deserve serious attention. Several mechanisms may be hypothesized by which rEPO might deleteriously affect cognitive functioning. They are induction of iron deficiency, unknown hypothetical effects of rEPO on the brain, cerebrovascular effects including hypertension and/or increased blood clotting and reduced

PAGE 121

C!? 113 sympathetic arousal secondary to decreased need for high cardiac output. In addition, at high levels of Hg, the efficiency of dialysis may be compromised resulting in increased uremia. Iron deficiency is a frequent consequence of rEPO therapy in the absence of adequate iron supplementation or assimilation. Iron deficiency has been shown to have neurobehavioral consequences unrelated to heme synthesis. Quite conceivably, stepped-up erythropoiesis could bind enough nonheme iron into Hg to lower iron levels sufficiently to deleteriously affect brain function. Hypothetically, this could result in reduced cognitive performance, and several weeks later, a rise in Hg. However, in treatment subjects, vigorous iron supplementation minimized changes in iron status; therefore, very few subjects exhibited changes in iron stores of potential clinical significance. Consistent with this observation, a GLM univariate analysis unequivocally failed to demonstrate a main effect for iron status (F=0.01, E=0.9384) . The notion that rEPO may affect brain function through mechanisms unrelated to erythropoiesis is contrary to general assumptions. However, cancerous neural tissue is capable of producing EPO (Hennessy, Stern, & Herrick, 1967) . This capability suggests the possibility that EPO may serve a function in the nervous system. If rEPO negatively affected brain function through nonheme

PAGE 122

114 mechanisms, cognitive deficits might be expected to appear prior to rise in Hg. As previously mentioned, analysis using a model containing either treatment status or Hg change suggested a possible trend; however, a model containing both terms failed to shed light on the "direct effect of EPO on the brain" hypothesis. Following reclassification of subjects based on direction of Hg change, a multivariate analysis demonstrated treatment status to be unambiguously noncontributory (F=0.59, E=.4482), while direction of Hg change was significant {F=9.45, E=.0043). Thus, the post hoc hypothesis that EPO directly and deleteriously affects the brain through a channel unrelated to Hg finds no support in this data. Cognitive dysfunction might hypothetically be mediated by cerebral vascular mechanisms. Recombinant EPO mediated increases in Hg frequently produce undesirable effects including headache and hypertension (Delano, 1989; Hori, Onoyama, Iseki, Fujimi, & Fujishima, 1990). Increased platelet aggregation may further increase risk of cerebrovascular accidents. Although these possibilities could not be tested within the confines of the present data set, it may be noted that Hg rise was modest compared to clinical trials where the effect of rEPO on clotting time and blood pressure was studied. In severe anemia, the need for greater cardiac output results in increased sympathetic activation (Hori, Onoyama, Iseki, Fujimi, & Fujishima, 1990). The effect, if any, of

PAGE 123

115 this increased activation on cognitive function is unknown; however, the curvilinear relationship between performance and level of activation is well known. Since underarousal has been suggested as a component of some aspects of neuropsychological dysfunction in uremia, uremic subjects may be especially vulnerable to reduced activation secondary to a reduction in degree of anemia. However, contrary to this line of reasoning, Hg changes were seldom in the range and of the magnitude likely to affect degree of activation. The Controlled Word Association Test is affected by degree of activation, at least in grossly neurologically impaired populations. Superficially, word generation from within phonemic categories appeared to follow a pattern similar to that observed for Sum CVLT in that treatment group performance was relatively uniform across trials and control group performance was lower at the first assessment. However, controlling for multiple tests, this finding failed to reach statistical significance. More important, unlike Sum CVLT, this variable no longer exhibited the above mentioned pattern when sign of Hg change was used for classif icatory purposes. In addition, the self-report questionnaire, although not designed for this purpose, failed to suggest alteration in degree of activation. Examination of the pattern of mean Sum CVLT change scores led to the suggestion that the apparent decline was

PAGE 124

.1} 116 not due to a decrease in treatment group performance, but rather, to improved performance between the first and second assessments in subjects exhibiting a fall in Hg. The majority of these subjects were in the control group. Mean Sum CVLT in the treatment group was 4 5.8, 45.0 and 45.1 for assessments one through three respectively. With a standard deviation varying between 11.6 and 13.4, these scores were remarkably similar. In contrast, in the control group, mean Sum CVLT was 43.2 at the first testing but 47.3 and 47.1 at assessments two and three, respectively. With a standard deviation varying between 10.4 and 10.7, mean performance at the first assessment was nearly .4 of a standard deviation below expectations. Hypothesizing that increased Hg may have reduced the efficiency of dialysis resulting in increased uremia; therefore, produced poorer cognitive performance secondary to uremia, post hoc collection of BUN and creatinine levels was attempted. However, few subjects were consistently assessed on the same days as laboratory tests were performed. Therefore, the small size created the strong possibility, if not probability, of a Type II error. Consistent with this, analysis failed to suggest a relationship between BUN and Sum CVLT. It might be argued that the control group, rather than performing more poorly at the first assessment, did better at the second assessment. That is, the control group exhibited a practice effect while the treatment group.

PAGE 125

117 Starting at a higher level, failed to benefit from the experience. However, a simple practice effect would not be expected as test forms contained different words and categories. Learning how to learn list learning from minimal exposure to a list learning task and generalization to subsequent list learning weeks later would be quite extraordinary. Also, neither group showed an improvement between the second and third assessments. It further strains credulity to consider that maximal learning occurred following a single exposure to a list learning task. In addition, the practice effect hypothesis requires the dubious assumption that prior to exposure to the list learning task, the control group was poorer at list learning than was the treatment group, but benefited more from practice. Based on subject composition the control group would be expected to perform slightly better; indeed, they did on the second and third assessments. The hypothesis of differential response to practice must be rejected. Form III of the CVLT, the version developed for this study, was administered second to all subjects. If Sum CVLT III were easier than Form I and Form II, then this might explain the appearance of improved performance in the control group at the second assessment and suggest decreased performance in the treatment group. One of the benefits of the present design is that it directly controls for possible intertest form differences through the

PAGE 126

118 creation of difference scores. However, absence of randomization eliminated the ability to statistically test for interactions between test instrument and group or order effects. Although CVLT III might theoretically be somewhat easier based on the word frequency rates of the target words, as described in Appendix A, administration of Forms I and III to 18 normal, native English speaking volunteers, primarily university undergraduates, revealed identical Sum CVLT mean scores for both groups. Moreover, scores were remarkably similar between Forms I, II and III in the treatment group. In addition, scores between Forms II and III were essentially identical in the controls. In the treatment group, the correlation was an impressive .92 between Sum CVLT Forms II and III and .81 between Forms I and III. In contrast, the correlation between Forms I and II, tests arguably considered equivalent, was .77 for Sum CVLT. For comparison, in the validation of Form II, the correlation between Sum CVLT on Forms I and II was .84. Taken together, these results suggest that Sum CVLT III is equivalent to the two previous forms. Differences in test instruments appear unable to account for the observed pattern of scores. Psychological explanations may be posited; however, none elegantly fit the data. Nevertheless, it should be noted that subject expectations were poorly controlled. Most hemodialysis patients spent three half-days per week

PAGE 127

119 in a milieu in which information and misinformation was routinely shared. Most peritoneal dialysis patients were encouraged to participate by their physician. Hemodialysis patients were approached directly by the experimenter. This difference may have inadvertently affected patient performance at the first assessment. Another source of error that may have contributed to the pattern of results is that the majority of subjects knew if they were receiving rEPO. Indeed, at higher dosages of rEPO subjects experience such dramatic effects that their status is obvious to all concerned. However, in this study both experimenter and subject were blind to Hg level. In a multivariate model There was a main effect for direction of Hg change (F=16.97, £=.0003) and an interaction between direction of Hg change and selfreported Cognitive-Affective behavior (F=6.53, £=.0043). Improvement in self-reported Cognitive-Affective behavior was associated with improved performance on Sum CVLT. As the nature of the main effect is unexplained no attempt will be made to explain the interaction. The levels of processing and estimation of frequency paradigms were intended to probe the mechanisms underlying any observed improvement in memory function. Analyses failed to suggest that there were changes in either automatic memory or ability to benefit from semantic processing.

PAGE 128

120 Unfortunately, several interrelated factors weakened the design of the present study. Initial plans called for integration of this project into a larger study examining the clinical efficacy and safety of rEPO prior to U.S. Food and Drug Administration approval. However, approval of rEPO for clinical use was unexpectedly expedited. This resulted in reduced control over the independent variables. Hemoglobin change scores were much smaller than in the pilot study. Indeed, follow-up revealed that over one year after initiation of rEPO, the majority of subjects were still moderately anemic. Because the experimenter was blind to Hg levels until completion of neuropsychological testing, the existence of only minimal Hg change scores was undetected until completion of the study. In the Pilot study, data collection was coordinated with the timetable of a multicenter clinical trial. Baseline assessment occurred prior to any EPO mediated rise in Hg. Approval for clinical use resulted in the majority of hemodialysis subjects being placed on rEPO, many prior to baseline assessment. A reduced pool of rEPO-free hemodialysis subjects necessitated the inclusion of peritoneal dialysis subjects as controls. Although undesirable, the impact of a nonequivalent control group was minimized by the design. The controls merely served to control for the effects of multiple testing with different test forms. The approval of rEPO raised the possibility that suitable rEPO-free long-term dialysis subjects would

PAGE 129

121 become rare. Therefore, the study was conducted despite imperfect conditions. The paucity of changes between assessments provided the opportunity to examine the alternate form reliability of the test forms developed for this study. These tests were the third form of the CVLT as well as the second and third forms of the levels of processing and estimation of frequency of occurrence tasks. x The CVLT III performed very well and appeared to be essentially identical to the two earlier forms. In contrast, the levels of processing and estimation of frequency of occurrence alternate forms were disappointing. The statistically significant correlation between Cognitive-Affective self -report change scores and Sum CVLT change scores, especially in the absence of a significant relationship between Physical Activity and Sum CVLT, supported the validity of behaviorally oriented questionnaire data in the study of this population. This is significant in that previous work with the Profile of Mood States has been less than supportive of the validity of self-report data in this population. Conclusion Many of the methodological deficiencies noted in the literature were taken into account in the design of this study. Thus, this study made use of a longitudinal design, alternate test forms, a dialysis control group, sufficient sample size, controlled for iron status, maintained a consistent time of assessment with respect to dialysis and

PAGE 130

122 excluded patients with uncontrolled hypertension. Nevertheless, flaws severely limit the extent to which useful conclusions may be drawn. In the current study, Hg levels were both higher at baseline and rose much less than in the pilot study. There was significant intersubject variability in degree of anemia and in etiology of renal disease. Despite utilization of dialysis controls, there were big between group differences in dialysis treatment modality. Although the experimenter was blind to Hg level, rEPO treatment status was usually known by subject and experimenter. The present study produced absolutely no evidence that rEPO treatment is associated with improvement in neuropsychological functioning. Given the positive correlation between rEPO mediated rise in Hg and improvement in LTM in the Pilot study, the absence of improvement in response to smaller increases in Hg might be interpreted as indicating that most patients would benefit from higher dosages of rEPO. However, the present results provided data useful in a re-examination of the findings of the Pilot study. The results of this re-examination suggested the possibility that the initial findings may have been spurious. More work is needed with Hg changes of greater magnitude and lower initial values to speak to the question of the neuropsychology of anemia. Despite the absence of Hg change at levels of demonstrated physiological significance, a possible trend

PAGE 131

123 toward an inverse relationship between Hg change and Sum CVLT was noted, and confirmed statistically by post hoc analyses. Curiously, Sum CVLT performance was temporarily below expectations in subjects who would later exhibit a drop in Hg. Stated differently, between the first and second assessments. Sum CVLT rose when Hg fell. After exploring a variety of hypotheses, it was concluded that the cause is unknown at this time. Future studies of the impact of rEPO on nervous system function in the anemic patient need address a number of problems discussed in these pages. Many of the more serious flaws in the present study were a function of the release of rEPO for clinical use. Given that under current medical protocols it may be that the severely anemic chronic renal patient may become rare, future studies focusing on the neuropsychology of chronic anemia may have to work with other populations. Nevertheless, despite increasing difficulties in working with this population, optimal care of the renal patient demands well controlled studies of the effect of rEPO mediated changes in anemia on neuropsychological function. ifr ».

PAGE 132

».wsw • T Vn'tvr^W APPENDIX A DEVELOPMENT AND VALIDATION OF FORM III OF THE CALIFORNIA VERBAL LEARNING TEST Introduction In the pilot study, repeat memory testing with Form I of the CVLT revealed that some adult hemodialysis subjects recalled the word lists after several months. Therefore, alternate forms were considered desirable in the present study. Form II of the CVLT was recently developed by Delis, McKee, Massman, Kramer and Gettman (1990) . The construction and validation of a third parallel form will be described below. Test Construction A third form of the CVLT was created for use in this study. Following the procedures used in the development of Form I of the CVLT, three criteria were used in selecting Form III target words. Two of these criteria pertain to frequency of use of the words and the third category attempts to produce category members of equivalent typicality (Delis, Kramer, Kaplan, & Ober, 1987) . The CVLT consists of two lists of 16 words that are balanced for frequency of appearance in English reading (Thorndike & Lorge, 1944; Carroll, Davies, & Richman, 1971) . Both lists are composed of words belonging to four 124 '» '
PAGE 133

125 categories with two of the categories shared by both lists. Words are balanced with regard to the rank order of each word as an exemplar of the category as measured by Battig and Montague (1969), Hunt and Hodge (1971) and Uyeda and Handler (1980) . The primary and interference lists of each form contain two categories in common and two unrelated categories. "" '^ ^ TABLE A-1. Mean Word Frequency and Typicality Across Forms FORM I FORM II a b c word word cat freq frea rank FORM III a be word word cat freq frea rank 12.2 43.6 8.2 10.5 15.7 15.9 43.8 8.1 5.5 4.3 a be word word cat freq freq rank Primary list M 10.81 41.66 13.31 Sd 9.91 8.65 6.73 Interference list M 14.13 43.56 14.94 7.7 42.8 16.7 45.4 12.3 Sd 13.9 9.83 9.28 7.8 9.5 19.5 4.8 8.8 Note: a word freq = word frequency (Thorndike & Lorge, 1944); b word freq = word frequency (Carroll, Davies, & Richman, 1971) ; c cat rank = typicality ranking (Battig & Montague, 1969) ; M = mean; Sd = standard deviation . ' Table A-1 compares the word frequency characteristics of Forms I, II and III with regard to frequency of appearance in English reading as measured by Thorndike and

PAGE 134

126 TABLE A-2 . Form III Word Frequency and Typicality Values Primary List a b c Interference List a b c Dresser Rug Couch Bookcase Orchids Lilies Gardenias Daffodils Sapphire Pearl Garnet Jade Socks Slippers Sandals Boots 7 40 28 3 3 33 1 3 6 47 2 4 12 20 5 37 45.5 51.5 47.1 40.5 43.8 45.3 29.6 40.7 39.4 47.7 8 12 7 14 6 8 11 12 4 5 37.2 11 42.8 7 50.3 4 47.5 5 44.8 3 47.5 2 Rocker 2 Bureau 44 Stool 16 Cabinet 26 Carnation 1 Pansies 7 Daisies 28 Geraniums 4 Liver 10 Bacon 12 Ham 17 Steak 14 Silk 75 Flannel 9 Burlap 1 Denim 1 44.1 45.3 49.2 48.6 36.6 42.5 44.1 39.6 48.7 50.7 49.8 46.2 55.1 44.2 41.1 41.1 36 11 10 15 3 9 4 16 3 15 12 13 Mean 15.7 43.8 8.1 16.7 45.4 12.3 Sd 15.9 5.5 4.3 19.5 4.8 8.8 Note: word frequency (Thorndike & Lorge, 1944) b = word frequency (Carroll, Davies & Richman, 1971) c cat rank = typicality ranking (Battig & Montague, 1969) ; M = mean, sd = standard deviation

PAGE 135

127 Lorge (1944) and Carroll, Davies and Richman (1971) as well as the rank order of each word as an exemplar of its category (Battig and Montague, 1969) . Table A-2 lists Form III words and their frequency values. Methods Eighteen normal, native English speaking volunteers were administered Forms I and III of the CVLT. Seventeen subjects were between 18 and 23 years of age. One subject was 4 8 years old. Mean age was 21.9 with a standard deviation of 6.7 years. There were 11 female and 7 male subjects. Mean years of college education was 14.5 (SD=1.2). Sixteen of the 18 subjects had at least some college education. The majority of testing was performed by an undergraduate senior psychology major in his last semester of study, who had received extensive training prior to data collection. All subjects were administered Forms I and III of the CVLT. Every other subject began with Form I. Mean interval between testing sessions was 5 days. Minimum time between sessions was 48 hours. All assessments were performed individually. During the 20 minute delay between immediate and delayed recall subjects either conversed with the experimenter or performed activities in which they had been engaged prior to the assessment (e.g., studying). Administration of other measures during the delay was rejected to reduce the difficulties associated with obtaining volunteer subjects.

PAGE 136

128 Results The variable of greatest interest was Sum CVLT. Sum CVLT is likely to be the most stable measure produced by the CVLT. As displayed in Table A-3, Forms I and III produced identical means of 65.1 and a correlation coefficient of r = .69. Trial 1, a measure of STM, produced means of 8.7 and 9.7 for Forms I and III TABLE A-3. CVLT Inter-Form Reliability CVLT III CVLT I CVLT I & Variable r Mean sd Mean sd CVLT II r SUM .69b 65.1 6.2 65.1 4.9 .84c Tl .70c 9.7 2.2 8.7 1.7 .54c T2 .28 12.5 1.9 12.6 1.3 T3 .63b 13.3 1.8 13.8 1.4 T4 .31 14.3 1.2 15.0 1.1 T5 .10 15.1 1.0 15.2 1.2 .75c List B .56a 8.9 1.9 9.0 1.9 .31a SD .63b 14.0 U« 14.7 1.0 .82c Cued .51a 14.3 1.4 15.2 0.8 .76c LongDel .32 14.1 1.3 15.1 1.0 .80c LCued .14 14.9 0.8 15.3 0.7 .79c Note : a p<.05 b p<.01 c p<.001 SUM = Sum CVLT, List B = Interference list; SD = short delay free recall; CUED = short delay cued recall, LONGDEL = long delay free recall; LCUED = long delay cued recall

PAGE 137

129 respectively. The correlation between forms for Trial 1 was r = .70. The mean interference trial scores of 9.0 and 8.9 for Forms I and III, respectively, were nearly identical and produced a correlation of r = .56. It is desirable that the interference trial be of equivalent difficulty to the primary list as measured by Trial 1 of the primary list. The CVLT Form III primary list appeared slightly but not significantly easier. The inter-form correlation for Trial 5 was r= .50; however, means were 15.2 and 15.1. Please see Table A-3 for a tabular listing of the results. In Table A-3, the correlations between Forms I and II were reported by Delis, McKee, Massman, Kramer and Gettman (1990) . Discussion A comparison of mean scores between the two forms suggests that the two versions are quite similar. Indeed, this is illustrated by summing the 10 primary measures. Mean word recall for Form I was 134.6 and 131.1 for Form III. As this is out of a total of 160 words the 3.5 word recall advantage for Form I is rather insignificant. The correlations for Sum of Trials 1 through 5 (.69) and Trial 1 (.70) compare favorably with the alternate form reliability coefficients reported for the Rey Auditory Verbal Learning Test, which ranged from .60 to .77 and with Russell's revised Wechsler Memory Scale, which had correlations ranging from .60 to .74 (Ryan, Geisser,

PAGE 138

130 Randall, & Georgemiller, 1986; McCarty, Logue, Power, Ziesat & Rosenstiel, 1980). Sum of Trials 1-5 and Trial 1 both worked well and they are measures of special interest in the study for which Form III was designed. In contrast, many of the other correlations are rather low, especially Trial 5 (r = .10) and Long Delay Cued Recall (r = .14). Also, in every instance except for Trial 1, correlation coefficients between Forms I and II are higher, sometimes markedly higher. However, it is likely that the low correlations for some measures are part of a general suppression of all correlation coefficients due to the homogeneity and very high functioning of the subject pool. Indeed, recognition measures could not be calculated because nearly every subject had a perfect score. Likewise, a mean of 15.1 for the fifth trial on Form III clearly illustrates the problem of a ceiling effect. Higher alternate form correlation coefficients might be expected with a more representative cross-section of the population. Construction of this version of CVLT Form III is deficient in that some target words have prototypical ity ratings (Battig and Montague, 1969) of 4 or lower. This theoretically increases the possibility that correct responses may occur by chance in subjects prone to semantically related intrusions. However, the present data suggests CVLT Form III to be an excellent alternate form.

PAGE 139

HQtr '. ,1 r < .y fc* APPENDIX B LEVELS OF PROCESSING VERSIONS II AND III Table B-1. Levels of Processing Version II Encoding Questions Visual Stimuli 1 . ALTAR 2 . ear 3. sand 4. swine 5. smell 6. cold 7 . sky 8 . country 9. priest 10. joy 11. wolf 12 . BURN 13. silver 14. seat 15. trot 16. shoot 17. rope 18. PEN 19. CHEW Question to be processed Is this printed in upper case letters? Is this part of the body? Does this rhyme with match? Is this a type of furniture? Does this rhyme with sell? Is this printed in upper case letters? Is this printed in upper case letters? Is this a type of music? Does this rhyme with fall? Does this rhyme with boy? Is this a type of flower? Is this printed in upper case letters? Is this a type of metal? Is this a type of animal? Does this rhyme with pot? Is this printed in upper case letters? Does this rhyme with hour? Is this printed in upper case letters? Is this printed in upper case letters? 131 % > {

PAGE 140

132 Table B-1 continued Visual Stimuli Question to be processed Does this rhyme with board? Is this a type of tool? Does this rhyme with cry? Is this a type of flower? Is this printed in upper case letters? Is this printed in upper case letters? Is this a type of fruit? Does this rhyme with rug? Does this rhyme with fad? Is this a type of sport? Is this printed in upper case letters? Is this printed in upper case letters? Is this a type of furniture? Does this rhyme with cab? Is this printed in upper case letters? Is this a type of animal? Does this rhyme with wire? 20. prison 21. hatchet 22. lie 23. dandelion 24. tube 25. grass 26. pear 27. swim 28. pad 29. trumpet 30. SURPRISE 31. rain 32. tie 33. knot 34. SAFE 35. sheep 36. hire

PAGE 141

133 Table B-2. Multiple Choice Retrieval Questions Version II Processing Level SEMORTH+ PHONPHON+ ORTHSEM+ PHON+ SEMSEM+ ORTH+ PHONORTHORTH+ SEM+ PHON+ SEMORTHPHONSEM+ ORTH+ ORTHPHON+ SEMPHONSEM+ Multii 3le Choice Word List 1. seat coin maple nail 2. clip wood chew fern 3. globe sand clamp scar 4. rug haze gift hire 5. grass foot mast sheet 6. tear country page horse 7. joy cream oak fear 8. path coal tie tent 9. dandelion bat ghost lap 10. moss altar bed cart 11. trout egg doll swim 12. cable paste shoot cup 13. band cord safe tune 14. fur ear twig mate 15. roll chin pork lie 16. trumpet key steak purple 17. plot sky weed cage 18. crumb limb rope ash 19. wheel hide spade sheep 20. banana leaf pen nob 21. tube roam spear art 22. bullet smell cat oar 23. bone pack rock wolf 24. prison corn book mire 25. juice string silver acre

PAGE 142

134 Table B-2 continued Multicle Choice Word List Processing Level 26. silk hatchet fog brass SEM27. dog lamp wave knot PHON28. cold fad gear tan ORTH29. surprise cop prey camp ORTH+ 30. soap pond trot chart PHON+ 31. float priest film ant PHON32. paint hinge mute burn ORTH+ 33. swine ditch surf ache SEM34. horse seem ramp rain ORTH35. floor hen stamp .':: pad PHON+ 36. spark pear grip site SEM+ Note: + = Correct response is positive; = Correct response is negative; ORTHO = Orthographic; PHON = Phonetic; SEM = Semantic t .i j

PAGE 143

135 Table B-3. Levels of Processing Version III Encoding Question to be processed Is this printed in upper case letters? Is this part of the body? Does this rhyme with match? Is this a type of furniture? Does this rhyme with sock? Is this printed in upper case letters? Is this printed in upper case letters? Is this a type of musical instrument? Does this rhyme with fall? Does this rhyme with chalice? Is this a type of flower? Is this printed in upper case letters? Is this a seasoning? Is this a type of animal? Does this rhyme with bike? Is this printed in upper case letters? Does this rhyme with hour? Is this printed in upper case letters? Is this printed in upper case letters? Does this rhyme with board? Is this a type of tool? Does this rhyme with head? Is this a type of fruit? Is this printed in upper case letters? Is this printed in upper case letters? Questions Visual Stimuli 1. COLTiEGE 2. finger 3. garden 4. hamster 5. clock 6. ship 7. animal 8. violin 9. robot 10. palace 11. canoe 12. LAKE 13. salt 14. hill 15. spike 16. star 17. mile 18. KNIFE 19. TOWEL 20. dime 21. shovel 22. bed 23. strawberry 24. deck 25. pencil

PAGE 144

136 Table B-3 continued Visual Stimuli Question to be processed Is this a type of building? Does this rhyme with rug? Does this rhyme with pet? Is this a type of sport? Is this printed in upper case letters? Is this printed in upper case letters? Is this a type of furniture? Does this rhyme with cab? Is this printed in upper case letters? Is this a type of color? Does this rhyme with math? 26. hotel 27. plate 28. net 29. peel 30. TRUCK 31. dog 32. arm 33. ink 34. SHORE 35. white 36. bath ?„. .-f -».!,

PAGE 145

137 Table B-4. Multiple Choice Retrieval Questions Version III Multiple Choice Word List 1. hill 2. clip 3. globe 4. rug 5. pencil 6. tear 7. palace 8. path 9. strawb 10. laoss 11. trout 12. string 13. band 14. fur 15. roll 16. peel 17. plot 18. crumb 19. wheel 20. peddle 21. deck 22. bullet 23. bone 24. dime 25. juice com wood garden haze foot violin cream coal bat college egg paste cord finger chin key animal limb lie leaf roam clock pack grass rope grass TOWEL ^ clamp gift mast page oak arm ghost bed doll star shore twig pork steak weed mile spade knife spear cat rock book salt nail fern scar bath sheet horse fear tent lap cart plate cup tune mate bed purple cage ash white nob art oar canoe mire acre Processing Level SEMORTH+ PHONPHON+ ORTHSEM+ PHON+ SEMSEM+ ORTH+ PHONORTHORTH+ SEM+ PHON+ SEMORTHPHONSEM+ ORTH+ ORTHPHON+ SEMPHONSEM+

PAGE 146

138 Table B-4 continued Multiple Choice Word List Processing Level 26. silk shovel fog brass SEM27. dog lamp wave ink PHON28. ship fad gear tan ORTH29. truck cop prey camp ORTH+ 30. soap pen spike chart PHON+ 31. float robot film ant PHON32. paint hinge mute lake ORTH+ 33. hamster ditch surf ache SEM34. horse seem ramp dog ORTH35. floor hen stamp net PHON+ 36. spark hotel grip site SEM+ Note: + = Correct response is positive; = Correct response is negative; ORTHO = Orthographic; PHON = Phonetic; SEM = Semantic jf' J w

PAGE 147

APPENDIX C FREQUENCY ESTIMATION TASK Table C-1. Encoding Phase "Now I am going to read a list of [animals, grain & flour products or means of transportation] . Please remember as many of the words as possible because a little later on I will ask you to list as many of them as you can." Assessment 1 ANIMATuS 1 kangaroo 2 dog 3 snake 4 elephant 5 kangaroo 6 buffalo 7 goldfish 8 kangaroo 9 tiger 10 dog 11 moose 12 kangaroo 13 gopher 14 kangaroo 15 dog As sessment 2 GRAIN & FLOUR PRODUCTS 1 wheat 2 oatmeal ' 3 barley 4 cake 5 barley 6 corn 7 waffle 8 barley 9 pancake 10 oatmeal 11 rice 12 barley 13 bagel 14 barley 15 oatmeal 139 As sessment 3 MEANS OF TRANSPORTATION 1 car t:^ subway 3 bicycle 4 escalator 5 subway 6 airplane 7 canoe 8 subway 9 stagecoach 10 car 11 skateboard 12 subway 13 helicopter 14 subway 15 car

PAGE 148

:t 140 Table C-2. Free Recall Instructions "Remember the list of [animals, grains & flour products or means of transportation] that I read earlier. Please tell me all the words from that list that you can remember, in any order." Table C-3. Frequency Estimation "When I read you that list some of the words were said more than once. I will now read you words from that list. When I say a word tell how many times you think it appeared the first time I read it. Guess when not sure." Assessment 1 ANIMALS 1 kangaroo 2 snake 3 dog 4 elephant 5 buffalo 6 goldfish 7 tiger 8 moose 9 gopher Assessment 2 GRAIN & FLOUR PRODUCTS cake oatmeal_ wheat rice corn waffle pancake_ barley bagel Assessment 3 MEANS OF TRANSPORTATION escalator_ canoe bicycle_ stagecoach_ airplane car subway_ skateboard_ helicopter_ J

PAGE 149

APPENDIX D COGNITIVE-AFFECTIVE AND PHYSICAL BEHAVIOR QUESTIONNAIRE Table D-1. Questionnaire Circle only the sentences true over the past two days. 1. I don't mind climbing stairs. 2. Walking quickly tires me out. 3. I sit during much of the day. 4. I take one or more naps during the day. 5. I feel useless, like I am a burden on others. 6. I easily get out of breath. "' 7. I have difficulty concentrating. 8. My sleep is sounder than it used to be. 9. I am irritable and impatient with myself. 10. I do work around the house only for short periods. 11. I stay away from home only for brief periods of time. 12. My sex drive is stronger than it was one year ago. 13. I am not doing any daily work around the house. 14. I feel terrible when I get off of hemodialysis. 15. I feel tired most of the time. ^ 16. I do not finish things I start. "" * w; 17. I react slowly to things that are said or done. 18. I work a job outside the home. ^ , 19. I spend much of the day lying down in order to rest. 20. I enjoy eating. 141

PAGE 150

142 Table D-1 continued 21. My future looks bright. 1 22. I feel tired all the time. g ^ i 23. Everything is an effort. 24. I enjoy visiting friends. ] 25. I do heavy work around the house. 26. I do alot of cooking. ; 27. Simple chores quickly tire me out. ;• 28. My muscles quickly get tired. Table D-2. Instructions for Scoring; Corrections If 13 is endorsed then consider 10 endorsed. If 19 is endorsed then consider 3 endorsed. ' If 22 is endorsed then consider 15 endorsed. i Table D-3 . Cognitive-Affective Scale Scoring Instructions Starting with 9 points, subtract one point for each of the following if endorsed: 5. I feel useless, like I am a burden on others. ; 7. I have difficulty concentrating. ] 9. I am irritable and impatient with myself. 16. I do not finish things I start. 1 17. I react slowly to things that are said or done. Subtract one point for each that is not endorsed: ] 8. My sleep is sounder than it used to be. 5 20. I enjoy eating. 21. My future looks bright. 24. I enjoy visiting friends. i

PAGE 151

,J «; hijf9 ..r". . t ^ ^vr mf-tfi-'-'t* -^ ,f '.f •«"' 143 Table D-4 . Physical Scale Scoring Instructions Start with 18 points and subtract one point for each endorsed. . ' 2. Walking quickly tires me out. 3. I sit during much of the day. 4. I take one or more naps during the day. 6. I easily get out of breath. 10. I do work around the house only for short periods. 11. I stay away from home only for brief periods of time. 13. I am not doing any daily work around the house. 15. I feel tired most of the time. 19. I spend much of the day lying down in order to rest. ( 22. I feel tired all the time, 1 23. Everything is an effort. ^ 27. Simple chores quickly tire me out. j 28. My muscles quickly get tired. Subtract one point for any of the following not endorsed: j 1. I don't mind climbing stairs. < 5 12. My sex drive is stronger than it was one year ago. 1 18. I work a job outside the home. ; 25. I do heavy work around the house. 1 J 26. I do alot of cooking.

PAGE 152

APPENDIX E COMPARISON OF PARAMETRIC AND NONPARAMETRIC TESTS OF BASELINE MEASURES BETWEEN GROUPS Kruskal-Wallis Test (Chi-Square Approximation) Nonparametric Df=l Prob > Variable Chi-Sauare Chi-Sauare Hemoglobin 3.0616 .0802 Hematocrit 3.1778 .0746 Ferritin 0.3478 .5553 Age 1.1868 .2760 Education 0.06695 .7958 Dialysis years 2.8425 .0918 California Verbal Learning Test Trial 1 0.0045 .9464 Standard Score 0.0855 .7700 Trial 2 0.1244 .7243 Trial 3 0.2496 .6173 Trial 4 1.3349 .2479 Trial 5 0.4435 .5054 Standard Score 0.0045 .9464 Sum CVLT 0.3157 .5742 Standard Score 0.0221 .8817 Clusters 0.0175 .8947 Perseverations 0.0805 .7765 Intrusions 0.1496 .6989 List B 0.0914 .7623 Standard Score 0.0012 .9724 Short Delay 0.0002 .9867 Standard Score 0.0002 .9864 Clusters 0.2849 i ., .5935 Perseverations 2.7324 .0983 Intrusions 0.2658 .6061 Short Delay Cued 0.1107 .7393 Standard Score 0.0141 .9052 Perseverations 3.3690 r .0664 . Intrusions 0.4843 ; .4865 M General Linear Model Parametric F Value Pr >F 2.94 .0959 2.29 .1401 0.00 .9483 1.13 .2948 0.23 .6325 3.51 .0699 0.05 .8253 0.00 .9602 0.50 .4847 0.21 .6535 0.68 .4140 0.40 .5291 0.11 .7400 0.42 .5222 0.03 .8662 0.36 .5519 0.38 .5442 0.01 .9409 0.09 .7687 0.10 .7587 0.34 .5624 0.02 .8951 0.00 .9760 4.67 .0380* 0.14 .7086 0.00 .9973 0.01 .9053 2.62 .1151 0.04 .8343 144

PAGE 153

145 Kruskal-Wallis Test (Chi-Square Approximation) Nonparametric Df=l Prob > Variable Chi-Sauare Chi-Sauare Long Delay 0.1863 .6660 Standard Score 0.6110 .4344 Clusters 0.7180 .3968 Perseverations 0.8303 .3622 Intrusions 4.8724 .0273* Long Delay Cued 0.3420 .5587 Standard Score 0.7505 .3863 Perseverations — Intrusions 1.5373 .2150 7691 ,3034 .7627 .3967 .2681 .0235* .1896 .6757 .7747 CVLT Recognition Variables Recognition Hits 0.0862 False Positives From Interference List Shared Category 1.0593 Unrelated 0.0912 Intrusions not on Interference List Shared Semantic 1.2254 .2683 Shared Phonemic 0.8235 .3641 No Relation 0.6879 .4069 Controlled Word Association CFP 0.7183 FAS 1.2263 LSW 5.1345 Levels of Processing Semantic 1.7209 Orthographic 0.1749 Phonemic 0.0819 Estimation of Frequency of Occurrence Free Recall 0.0188 .8907 Estimation 0.2524 , „ .6153 Cognitive-Affective and Physical Behavior Questionnaire Cognitive-Affect 0.1360 .7123 Physical 0.1104 .7396 . Note: Df = degrees of freedom * = discrepant results between parametric and nonparametric tests :« General Linear Model Parametric F Value Pr >F 0.15 .6995 1.27 .2671 0.33 .5670 0.31 .5838 2.48 .1245 0.04 .8425 0.84 .3669 2.05 .1613 0.75 .3931 1.12 .2981 0.02 .8869 2.19 .1482 1.71 .1996 0.71 .4045 0.49 .4880 1.41 .2445 4.62 .0394* 0.93 .3424 0.16 .6887 0.21 .6477 0.04 .8390 0.09 .7655 sstionnaire 0.11 .7459 0.19 .6676 itric and

PAGE 154

APPENDIX F NORMALITY AND HOMOGENEITY OF VARIANCE Table F-1. Indices of Normality and Characteristics of the Distributions ShapiroWilk Test Variable W: Normal Prob
PAGE 155

147 Table F-1 continued Variable Shapiro-Wilk Test WtNormal Prob
PAGE 156

148 Table F-2 continued Variable Recognition Memory Correct Positive: F' Value 1.58 False Positive On List B: No Relation 1.2 6 Semantic Shared 1.16 False Positive Not on List B Semantic Shared 2.49 Phonemic Shared 2.93 . No Relation 2.35 Degrees of Freedom 50 49 49 50 50 49 49 50 50 49 49 50 P value 0.1127 0.4118 0.6079 0.0017 * 0.0002 * 0.0032 * Verbal Fluency CFP 1.02 FAS 1.21 TnSW ^ * • • 1.08 Levels of Processing -. Semantic 1.00 Phonological 1.10 Orthographic 2.14 Frequency of Occurrence Free Recall 1.16 Estimation 1.32 47 50 50 47 47 47 50 50 50 47 47 44 44 44 49 49 0.9395 0.5181 0.7843 0.9916 0.7607 0.0124 * 0.6092 0.3361 Cognitive-Affective and Physical Behavior Questionnaire Cognitive-Affective 1.06 49 50 0.8251 Physical .. 1.04 49 50 0.8790 Appetite 2.72 50 49 0.0006 * Hemoglobin Hematocrit Ferritin Age Education 1.26 50 50 0.4119 1.44 50 50 0.1983 2.82 44 13 0.0460 * 1.11 17 16 0.8333 2.10 16 17 0.1400 Note: * = Unequal variance

PAGE 157

APPENDIX G Table G-1. Means and Standard nit Ljr\s\ij i^u V J. Deviations .rtJ. X at Assessment One • by Group Variable Control Group N Mean Sd rEPO Treated Group N Mean Sd Age 18 Education in years 18 Days since EPO begun Years on Dialysis 18 48.22 11.83 2.84 14.19 2.09 3.76 17 17 16 17 43.24 11.41 41.37 6.64 13.45 3.02 32.08 7.69 Ferritin 6 523.62 626.13 16 545.93 734.50 Hemoglobin 18 9.44 1.92 17 8.46 1.42 Hematocrit 18 28.19 6.12 17 25.58 3.77 California Verbal Learr ling Test Trial 1 18 5.72 1.74 17 5.88 2.47 Standard Score 18 -1.33 0.91 17 -1.35 1.37 Trial 2 18 7.89 2.00 17 8.41 2.37 Trial 3 18 9.17 2.73 17 9.65 3.52 Trial 4 18 9.89 2.91 17 10.71 2.93 Trial 5 18 10.56 3.07 17 11.24 3.25 Standard Score 18 -1.39 1.33 17 -1.59 2.12 Sum CVLT 18 43.22 10.70 17 45.82 13.03 Standard Score 18 33.06 10.31 17 33.82 16.00 Clusters 18 11.28 7.43 17 13.24 11.51 Perseverations 18 3.06 2.60 17 3.65 3.10 Intrusions 18 0.83 2.04 17 0.88 1.83 List B is 5.94 1.66 17 5.76 1.92 Standard Score 18 -0.78 0.94 17 -0.88 1.05 Short Delay 18 9.06 2.71 17 9.65 3.26 Standard Score 18 -1.11 0.83 17 -1.06 1.43 Clusters 18 3.44 2.01 17 3.47 3.02 Perseverations 18 0.22 0.43 17 0.76 0.97* Intrusions 18 0.17 0.51 17 0.24 0.56 Short Delay ' Cued 18 10.94 2.69 17 10.94 2.90 Standard Score 18 -0.83 1.04 17 -0.88 1.36 Perseverations 18 0.00 0.00 17 0.29 0.77 Intrusions 18 0.56 1.34 17 0.65 1.22 149

PAGE 158

150 Table G-1 continued Variable Long Delay Standard Score Clusters Perseverations Intrusions Long Delay Cued Standard Score Perseverations Intrusions Control Group EPO Treated Group 1 N Mean Sd N Mean Sd ^ 18 9.94 3.44 17 9.47 i 3.76 -i 18 -0.89 1.02 17 -1.41 1.66 I 18 4.72 3.03 17 4.12 3.16 1 0.56 1 18 0.33 0.49 17 0.24 18 0.22 0.73 17 0.76 1.25* 1 18 11.06 3.51 17 10.82 3.34 ^ 18 -0.78 1.40 17 -1.24 1.56 I 18 0.00 0.00 17 0.00 0.00 ' 18 0.61 1.20 17 1.53 2.43 Recognition Hits 18 14.61 1.61 False Positives From Interference List Shared Category 18 0.94 0.87 Unrelated 18 0.39 0.70 Intrusions not on Interference List Shared Semantic 18 0.11 0.32 Shared Phonemic 18 0.33 0.59 No Relation 18 0.78 2.16 Controlled Word Association C 17 11.53 3.84 F 17 11.59 4.03 L 17 12.29 3.70 17 15.00 0.94 17 17 17 17 17 0, 0, 0, 0, 0. 65 35 41 82 29 17 10.53 17 10.00 17 9.41 0, 0, 0. 1. 0. 79 79 80 47 99 4.45 3.77 4.11* LOP Semantic 16 9.62 1.67 16 10.25 LOP Orthographic 16 7.69 2.15 16 7.37 LOP Phonemic 16 7.81 1.64 16 7.44 1.98 2.22 2.80 Frequency Estimation Task Free Recall 18 3.61 1.42 17 3.71 1.31 Frequency Estimate 18 11.44 2.20 17 11.24 1.89 Cognitive-Affective and Physical Behavior Questionnaire Cognitive-Affect 18 6.06 2.01 17 5.82 2 Physical is 9.56 4.42 17 8.88 4, Appetite 18 0.67 0.49 17 0.88 0, Post Hemodialysis 5 0.80 0.45 15 0.73 0, Note: * = Significant difference at .05 level LOP = Levels of Processing 19 77 33 46

PAGE 159

151 Table G-2. Means and Standard Deviations at Second Assessment by Group Control Group EPO Treated Group Variable N Mean Sd N Mean Sd Hemoglobin 18 9.23 1.81 17 9.59 1.84 Hematocrit 18 27.09 5.31 17 28.81 5.59 Ferritin 5 337.48 285.93 15 536.47 727.30 Assessment Interval 18 56.50 35.98 17 51.12 32.95 California Verbal Learning Test Trial 1 18 6.67 1.91 17 4.88 2.03 Standard Score 18 -0.78 0.88 17 -1.76 0.97 Trial 2 18 9.06 2.46 17 8.18 2.27 Trial 3 18 10.11 2.59 17 9.76 3.05 Trial 4 18 10.72 2.49 17 10.88 3.71 Trial 5 18 10.89 2.89 17 11.29 3.44 Standard Score 18 -1.50 1.34 17 -1.47 1.70 Sum CVLT 18 47.28 10.68 17 45.00 13.36 Standard Score 18 38.56 10.37 17 32.82 14.80 Clusters 18 12.78 9.72 17 13.53 9.84 Perseverations 18 3.61 2.77 17 4.00 3.79 Intrusions 18 1.50 2.20 17 2.06 2.84 List B 17 5.53 2.10 17 5.65 2.32 Standard Score 17 -0.88 1.27 17 -1.12 1.17 Short Delay 17 9.12 3.10 17 9.59 3.99 Standard Score 17 -1.12 1.17 17 -1.00 1.46 Clusters 17 3.88 3.08 17 3.88 2.80 Perseveration 17 0.59 0.80 17 0.41 0.80 Intrusions 17 0.47 0.51 17 0.53 1.01 Short Delay Cued 17 10.53 3.32 17 10.59 2.69 Standard Score 17 -0.94 1.09 17 -1.00 1.22 Perseverations 17 0.00 0.00 17 0,00 0.00 Intrusions X7 1.29 1.49 17 1.47 1.46 Long Delay 17 9.53 3.30 17 9.41 4.17 Standard Score 17 -1.24 1.09 17 -1.47 1.59 Clusters 17 4.53 3.48 17 4.24 2.80 Perseverations 17 0.35 0.79 17 0.12 0.33 Intrusions 17 0.82 1.29 17 1.00 1.17 Long Delay Cued 17 10.41 3.20 17 10.65 2.71 Standard Score 17 -1.24 1.30 17 -1.18 1.47 Perseverations 17 0.12 0.33 17 0.06 0.24 Intrusions 17 1.53 1.23 17 1.35 1.46 Recognition Hits 17 14.47 1.77 False Positives from Interference List Shared Semantic 17 0.88 1.11 17 17 14.18 0.53 1.19 0.94

PAGE 160

152 Table G-2 continued Note: LOP = Levels of Processing i ' 1 ' ft t '•s . jf '.i^ -i/''"'• •*i'-' u. J fi^ >^ t; • • ^M ^ If'. I >. . V . 1. A. ' 3 Control Group EPO Treated Group Variable 1 ^ Mean Sd N Mean Sd Intrusions not on Interference List Shared Semantic 17 0.35 0.61 17 0.41 0.87 Shared Phonemic 17 0.59 0.80 17 0.76 1.35 No Relation 17 0.29 0.69 17 0.41 0.71 1 Verbal Fluency 1 F 16 14.12 5.15 17 11.41 4.78 i A 16 10.12 3.81 17 7.29 4.92 1 S 16 15.25 5.08 17 10.41 5.23 ; LOP Semantic 15 10.27 1.75 16 10.37 i, 1.89 LOP Orthographic 15 6.20 2.11 16 6.88 2.55 i LOP Phonemic 15 8.73 1.94 16 9.06 2.77 Frequency Estimation Task Free Recall 17 4.24 1.48 17 3.06 1.39 Frequency Estimate 17 10.76 3.09 17 10.18 3.34 1 Cognitive-Affective and Physical Behavior Questionnaire Cognitive-Affective 18 7.00 2.35 16 5.75 2.32 Physical 18 8.78 4.60 16 9.31 4.73 Appetite 18 0.56 0.51 16 0.94 0.25 Post Hemodialysis 6 0.50 0.55 15 0.60 0.51 i

PAGE 161

153 Table G-3. Means and Standard Deviations at Third Assessment by Group Control Group EPO Treated Variable N Mean Sd N Mean Sd Hemoglobin 15 9.62 1.92 17 9.54 1.48 Hematocrit 15 28.12 5.55 17 27.55 4.06 Ferritin 3 484.30 266.67 14 560.34 814.39 Assessment Interval 15 67.80 34.56 17 90.41 24.03 California Verbal Learning Test Trial 1 15 5.87 1.64 17 6.00 1.97 Standard Score 15 -1.13 0.74 17 -1.24 1.03 Trial 2 15 9.00 2.80 17 8.94 3.17 Trial 3 15 10.33 2.23 17 9.82 2.24 Trial 4 15 10.93 2.66 17 9.82 3.13 Trial 5 15 10.93 2.52 17 11.12 3.02 Standard Score 15 -1.47 1.85 17 -1.53 1.59 Sum CVLT 15 47.07 10.40 17 45.12 11.61 Standard Score 15 36.80 13.21 17 33.18 13.65 Clusters 15 13.80 8.27 17 13.76 11.95 Perseverations 15 3.33 2.13 17 5.00 2.94 Intrusions 15 1.47 2.10 17 2.29 3.80 List B 15 6.40 2.16 17 4.88 2,50 Standard Score 15 -0.53 1.13 17 -1.35 1.37 Short Delay 15 9.07 2.87 17 9.47 3.39 Standard Score 15 -1.13 1.19 17 -1.12 1.32 Clusters 15 3.87 2.17 17 4.00 2.76 Perseverations 15 0.20 0.56 17 0.76 1.03 Intrusions 15 0.80 0.86 17 0.47 1.01 Short Delay Cued 15 9.87 2.83 17 10.76 2.39 Standard Score 15 -1.40 1.55 17 -1.00 1.12 Perseverations 15 0.07 0.26 17 0.00 0.00 Intrusions 15 0.73 0.88 17 1.29 1.65 Long Delay 15 9.27 2.87 17 9.65 3.82 Standard Score 15 -1.53 1.41 17 -1.24 1.48 Clusters 15 4.00 2.59 17 4.35 3.16 Perseverations 15 0.27 0.59 17 0.24 0.56 Intrusions 15 0.93 1.03 17 0.88 1.05 Long Delay Cued 15 10.13 2.95 17 10.82 2.53 Standard Score 15 -1.47 1.60 17 -1.24 1.35 Perseverations 15 0.00 0.00 17 0.00 0.00 Intrusions 15 1.20 1.15 17 1.41 1.54 Recognition Hits 15 14.07 1.62 False Positives from Interference List Shared Category 15 0.87 0.99 17 17 13.59 0.82 3.18 0.88

PAGE 162

154 Table G-3 continued Variable Control Group JI Mean Sd Intrusions not on Interference List Shared Semantic 15 0.33 0.62 Shared Phonemic 15 0.33 0.82 Unrelated 15 0.67 0.82 Verbal Fluency P R W EPO Treated N Mean Sd 17 17 17 0.53 0.59 0.65 0.87 0.94 1.06 15 15.07 3.71 17 11.00 4.23 15 11.60 3.74 17 9.29 3.64 15 10.87 2.77 17 8.18 3.97 LOP Semantic LOP Orthographic LOP Phonemic 14 9.36 14 6.79 14 7.57 1.65 2.52 1.87 16 16 16 10.19 7.12 9.06 Estimation of Frequency of Occurrence Free Recall 15 4.67 1.23 17 3.41 Frequency Estimate 15 10.33 2.35 17 9.47 Cognitive-Affective and Behavioral Questionnaire Cognitive-Affective 15 5.40 1.99 17 5.53 Physical 15 9.53 4.72 17 8.59 1.22 2.55 2.32 1.87 3.28 2.40 4.54 Note: LOP = Levels of Processing

PAGE 163

APPENDIX H CHANGE SCORES Control Group Treatment Group Variable Tm N Mean S^ H Mean Sd Hemoglobin 1 18 -0.21 1.34 17 1.13 1.57 b Hemoglobin 2 15 0.05 2.01 17 1.08 1.23 Hematocrit 1 18 -1.11 4.32 17 3.24 4.54 b Hematocrit 2 15 -0.43 6.21 17 1.98 3.14 California Verbal Learning Test Trial 1 1 18 0.94 2.24 17 -1.00 1.58 b Trial 1 2 15 -0.67 1.99 17 1.12 1.76 b Trial 2 1 18 1.17 2.41 -0.24 1.56 a Trial 2 2 15 -0.13 2.53 0.76 2.31 Trial 3 1 18 0.94 3.06 17 0.12 2.15 Trial 3 2 15 0.13 2.10 17 0.06 2.28 Trial 4 1 18 0.83 3.01 17 0.18 2.46 Trial 4 2 15 0.07 1.39 17 -1.06 3.25 Trial 5 1 18 0.33 2.89 17 0.06 2.44 Trial 5 2 15 0.13 2.53 17 -0.18 1.78 SS 1 18 -0.11 1.78 17 0.12 1.54 SS 2 15 0.27 1.79 17 -0.06 1.09 Sum CVLT 1 18 4.06 9.91 17 -0.82 5.38 Sum CVLT 2 15 -0.27 6.97 17 0.12 7.85 SS 1 18 5.50 13.61 17 -0.82 5.38 SS 2 15 -0.67 10.88 17 0.12 7.85 Cluster 1 18 1.50 10.25 17 -1.00 7.27 Cluster 2 15 0.67 8.06 17 0.35 10.42 Psv 1 18 0.56 3.76 17 0.29 5.01 Psv 2 15 -0.73 3.13 17 0.24 6.94 Intrus 1 18 0.67 3.20 17 0.35 4.40 Intrus 2 15 -0.33 2.94 17 1.00 4.76 List B 1 17 -0.41 2.69 17 1.18 3.19 List B 2 15 0.67 2.97 17 0.24 2.82 SS 1 17 -0.06 1.60 17 -0.12 1.62 SS 2 15 0.20 1.74 17 -0.76 2.25 Short Del 1 17 0.06 2.75 17 -0.24 0.97 Short Del 2 15 -0.20 2.86 17 -0.24 1.35 SS 1 17 0.06 1.39 17 -0.06 2.66 SS 2 15 0.00 1.56 17 -0.12 2.26 155

PAGE 164

156 Control Group T reatme int Group Variable Tm N Mean £d U Mean Sd Short Delay Cluster 1 17 0.47 2.67 17 0.06 1.34 Cluster 2 15 -0.27 2.96 17 -0.12 1.17 Psv 1 17 0.35 0.79 17 0.41 2.50 Psv 2 15 -0.40 0.99 17 0.12 2.34 Intrus 1 17 0.29 0.69 17 -0.35 1.22 Intrus 2 15 0.33 0.72 17 0.35 1.17 SD Cued 1 17 -0.35 2.32 17 0.29 1.26 SD Cued 2 15 -0.87 3.14 17 -0.06 0.56 SS 1 17 -0.06 1.14 17 -0.35 2.06 SS 2 15 -0.47 1.73 17 0.18 1.74 Psv 1 17 0.00 0.00 17 -0.12 1.27 Psv 2 15 0.07 0.26 17 0.00 0.87 Intrus 1 17 0.82 1.67 17 -0.29 0.77 Intrus 2 15 -0.33 1.68 17 0.00 0.00 Long Del 1 17 -0.47 2.15 17 0.82 1.67 Long Del 2 15 -0.60 2.50 17 -0.18 1.07 SS 1 17 -0.29 0.92 17 -0.06 2.90 SS 2 15 -0.40 1.35 17 0.24 2.36 Cluster 1 17 -0.18 1.88 17 -0.06 1.34 Cluster 2 15 -0.80 3.08 17 0.24 1.09 Psv 1 17 0.00 0.79 17 0.12 3.06 Psv 2 15 -0.13 0.99 17 0.12 2.67 Intrus 1 17 0.59 1.46 17 -0.12 0.70 Intrus 2 15 0.13 1.51 17 0.12 0.70 LD Cued 1 17 -0.53 2.96 17 0.24 1.48 LD Cued 2 15 -0.47 3.54 17 -0.12 1.32 Psv 1 17 0.12 0.33 17 -0.18 2.38 Psv 2 15 -0.13 0.35 17 0.18 1.29 Intrus 1 17 0.94 1.30 17 0.06 0.24 Intrus 2 15 -0.13 1.73 17 -0.06 0.24 Recog 1 17 -0.12 1.93 17 -0.18 1.91 Recog 2 15 -0.60 1.99 17 0.06 1.14 False + 1 17 0.88 3.10 17 0.59 1.50 False + 2 15 -0.47 2.50 17 0.12 3.04 Verbal Fluency 1 16 5.06 6.20 17 -0.82 1.33 a 2 14 -2.36 5.60 17 -0.59 2.92 Levels of : Processing Semantic 1 15 0.60 2.03 17 -0.82 8.65 Semantic 2 14 -0.79 2.42 17 -0.65 7.65 Orthograph 1 15 -1.60 2.75 17 0.13 2.00 Orthograph 2 14 0.79 3.56 17 -0.19 1.87 Phonemic 1 15 0.87 2.72 16 -0.50 2.61 Phonemic 2 14 -1.07 2.09 16 0.25 2.46

PAGE 165

t^ J " M 157 i> < ;," i,! : '"•'< Control Group Treatment Group Variable Tm N Mean Sd N Mean Sd Estimation of Frequency of Occurrence Recall 1 17 0.65 1.90 16 1.62 3.07 a Recall 2 15 0.47 1.73 16 0.00 2.25 Estimation 1 17 -0.59 2.35 16 -0.65 1.37 Estimation 2 15 -0.67 2.66 16 0.35 1.32 Cognitive-i Effective and Behavior Quest; ionnaire Physical 1 18 -0.78 2.10 16 0.06 1.95 Physical 2 15 0.13 2.50 16 -0.25 2.98 Cognitive 1 18 0.94 2.21 17 -1.06 3.19 Cognitive 2 15 -1.73 2.02 17 -0.71 4.96 a Appetite 1 18 -0.11 0.32 17 0.06 0.25 Appetite 2 15 -0.13 0.52 17 0.00 0.00 Note: a=p<.05;bp<.01 Tm= Assessment Interval; Psv = Perseverations; Del = Delay; Intrus = Intrusions; SS = Standard Score; Recog = Recognition; Orthograph = Orthographic; False + = False positive

PAGE 166

APPENDIX I RAW DATA BY SUBJECT Table I-l. Raw Demographic Data Hemoglob in by Assessmen Id Acfe Ed Sex Trt Group 1 2 3 1 24.2 14 M Hemo rEPO 7.90 9.00 10.80 2 31.2 13 F Hemo rEPO 9.50 10.60 10.10 3 53.7 12 M Hemo rEPO 9.00 9.30 9.30 4 29.6 8 F Hemo rEPO 7.54 8.10 8.00 5 37.1 14 F Hemo rEPO 5.60 6.60 6.90 6 45.2 10 F Hemo rEPO 8.30 14.20 8.70 7 43.0 12 F Hemo rEPO 8.10 7.90 8.60 8 51.0 12 F Hemo rEPO 7.55 8.85 7.25 9 49.3 12 F Hemo rEPO 8.50 9.10 9.40 10 25.8 12 M Hemo rEPO 9.60 9.75 9.65 11 26.5 12 F Hemo rEPO 7.10 10.50 11.10 12 56.3 7 M Hemo rEPO 7.20 6.95 7.90 13 46.2 13 M Hemo rEPO 11.30 10.60 10.84 14 57.1 12 M Hemo rEPO 10.50 10.70 10.81 15 48.0 12 M Hemo rEPO 10.20 12.00 12.25 16 37.0 16 F Perit rEPO 8.10 10.10 10.30 17 73.8 3 F Hemo rEPO 7.80 8.80 10.30 101 57.1 12 F Hemo Con 6.90 8.30 10.20 102 48.0 12 M Hemo Con 9.40 10.00 11.70 103 34.8 8 F Perit Con 5.27 6.48 6.90 104 71.5 8 M Hemo Con 9.70 9.61 • 105 54.7 12 M Hemo Con 13.90 12.60 10.50 106 60.5 15 M Hemo Con 9.25 6.30 • 107 43.5 13 M Perit Con 12.30 11.20 12.30 108 23.9 12 M Perit Con 8.40 9.40 9.90 109 43.5 14 M Perit Con 10.90 11.10 11.60 110 57.3 10 F Perit Con 9.50 10.00 10.20 111 48.1 14 M Perit Con 9.60 9.40 9.70 112 70.0 10 M Perit Con 10.71 10.60 8.96 113 65.8 12 F Perit Con 9.80 8.10 9.10 114 32.9 12 F Perit Con 9.90 11.60 11.60 115 33.9 14 F Perit Con 8.90 7.30 6.00 116 29.4 12 F Perit Con 8.80 9.10 9.10 117 38.5 8 F Perit Con 7.30 8.07 • 118 54.5 14 M Perit Con 9.40 7.00 6.60 Note: Ed = Years of Education; Trt Control; Hemo = Hemodialysis; Perit Treatment; Con = Peritoneal Dialysis 158

PAGE 167

159 Table 1-2. California Verbal Learning Test Trials One. Five, Sum and Sum of Clustering Raw Data by Assessment Number, Raw Score Raw Score T Score Raw Score Trial 1 Trial 5 Sum CVLT Sum Cluster Assess 2^ 2 3 I 2 3 1 2 3 1 2 3 Id 1 8 6 7 15 12 13 49 41 35 23 21 14 2 3 5 5 10 11 11 14 12 11 4 5 8 3 9 8 8 12 10 10 51 44 38 4 10 3 4 5 4 4 10 11 10 6 12 5 11 11 7 5 6 5 5 15 13 10 40 33 32 15 8 21 6 4 6 7 7 9 8 17 25 22 7 14 8 7 5 4 8 10 13 15 22 26 46 12 14 15 8 5 4 6 9 12 13 26 26 39 9 6 13 9 7 7 5 14 15 12 50 51 29 22 20 15 10 10 6 8 14 16 14 49 59 52 30 33 35 11 9 7 7 13 14 14 40 43 33 14 17 13 12 3 3 4 7 7 14 19 26 1 8 2 13 4 3 8 12 13 11 34 42 40 4 12 5 14 5 5 6 13 10 10 49 37 37 12 11 9 15 4 4 3 9 7 11 30 19 34 6 1 11 16 10 7 9 16 16 16 59 53 61 46 37 49 17 3 2 3 8 3 4 25 16 24 5 2 6 101 5 4 6 10 9 10 33 35 35 10 7 11 102 5 4 5 10 8 7 27 19 24 14 5 10 103 8 6 6 13 11 8 30 19 20 17 12 12 104 3 6 • 8 8 • 31 40 • 4 3 • 105 4 5 5 9 9 9 34 42 34 14 12 10 106 5 9 • 9 12 • 37 49 • 5 15 • 107 6 7 6 10 12 9 34 40 37 9 11 6 108 7 7 9 13 11 14 43 41 54 15 11 15 109 6 7 7 13 13 14 49 44 51 11 16 33 110 8 5 5 14 11 12 55 36 36 25 9 16 111 8 10 6 10 9 11 39 43 43 7 11 10 112 3 5 4 5 6 9 26 34 36 4 2 113 8 6 5 5 13 14 22 50 54 4 7 12 114 8 8 9 16 14 14 43 45 56 29 22 23 115 5 9 7 15 16 14 27 56 35 14 44 28 116 5 10 5 14 9 11 38 11 12 23 10 117 5 7 • 1 14 • 26 43 • 3 15 • 118 4 5 3 9 6 10 28 20 26 10 3 9 j

PAGE 168

160 Table 1-3. California Verbal Learning Test Short Delay. Long Delay and Recognition Raw Scores Short Delay Assess 1 2 3 Id 1 15 10 11 2 9 10 8 3 9 10 7 4 8 9 7 5 12 12 11 6 6 8 7 7 8 13 13 8 8 7 10 9 14 13 12 10 13 14 13 11 11 14 14 12 4 3 6 13 10 13 9 14 9 7 8 15 7 5 10 16 15 14 14 17 6 1 1 101 8 7 9 102 9 7 8 103 10 9 6 104 3 6 • 105 6 10 8 106 9 • • 107 9 8 11 108 11 8 12 109 10 9 10 110 12 10 10 111 10 6 10 112 3 5 2 113 10 13 11 114 13 14 13 115 12 14 12 116 9 14 8 117 10 10 • 118 9 5 6 Long Delay Recognit: ion 1 2 3 1 2 3 16 10 11 16 14 14 8 10 10 14 14 16 10 11 8 14 15 14 7 6 8 15 12 13 12 11 13 16 13 14 5 10 6 15 14 15 11 12 13 16 16 15 10 11 9 15 14 14 14 13 10 16 14 14 12 15 13 15 15 16 11 13 15 15 15 16 2 2 2 14 15 14 9 12 11 16 16 14 5 5 8 13 13 2 10 6 10 14 14 12 14 13 15 16 15 15 5 2 15 12 13 9 10 10 16 16 16 9 8 8 14 14 11 11 10 6 16 15 12 2 6 • 11 11 • 8 6 8 16 16 14 9 • • 15 • • 10 8 8 16 16 16 11 9 11 16 13 15 14 12 13 13 15 14 12 9 9 16 16 14 12 9 11 15 14 15 3 4 3 12 14 11 12 13 11 15 15 15 15 14 14 14 14 14 14 15 12 16 16 16 11 15 8 12 16 14 10 8 • 15 15 • 7 6 7 15 10 14 Note: Assess = Assessment number

PAGE 169

,v 161 Table 1-4. Raw Verbal Fluency and Self-Report Questionnaire Data by Assessment Controlled Word Association Cognitive-Affective and Physical Behavior Questionnaire CFL FAS FAS % PRW Phy sica 1 Cognitive Assess 1 2_ 2 1 1 2 3 1 2 3 Id 1 38 38 55-59 29 13 17 13 9 8 8 2 38 48 85-89 33 15 13 12 3 1 4 3 29 23 15-19 26 5 4 6 7 5 6 4 18 8 <4 15 3 5 14 6 6 8 5 25 31 35-39 25 4 7 7 4 5 4 6 33 21 10-14 32 6 7 5 1 3 1 7 17 14 <4 18 14 14 14 7 8 8 8 25 28 25-29 26 14 14 12 9 9 8 9 23 46 85-89 34 12 12 11 8 8 8 10 41 42 80-84 42 12 14 14 7 8 8 11 47 41 75-79 48 15 14 10 8 8 8 12 16 13 5-9 12 3 3 1 5 6 4 13 29 16 >4 29 7 4 3 6 3 3 14 25 28 30-34 23 3 3 4 4 4 2 15 32 30 35-39 27 10 8 9 5 4 5 16 52 51 9549 3 • 1 4 • 4 17 21 17 20-24 16 12 10 10 6 6 5 101 23 26 10-14 26 12 11 11 5 6 5 102 43 47 85-89 42 5 5 9 7 5 4 103 28 40 80-84 36 12 8 11 6 7 5 104 22 23 50-54 8 8 • 8 7 • 105 23 40 75-79 40 2 1 2 1 106 • • • 6 3 • 8 6 • 107 43 54 >95 46 14 14 11 5 7 6 108 42 45 75-79 48 5 7 5 4 6 4 109 51 « 47 10 13 9 5 7 3 110 28 25 25-29 30 14 13 11 8 7 7 111 33 42 80-84 30 14 17 17 8 9 8 112 27 32 65-69 28 11 9 9 6 8 6 113 33 31 50-54 31 10 6 3 8 8 6 114 42 55 9551 13 13 15 8 8 8 115 48 49 90-94 44 2 2 2 4 5 4 116 29 26 20-24 35 14 11 14 6 14 6 117 50 60 >95 8 6 • 4 6 • 118 37 37 65-69 29 14 12 14 8 8 8 Note: Asses s = Assessment numb er ^v.!

PAGE 170

BIBLIOGRAPHY Abram, H.S., Moore, G.L., & Westervelt, F.B. (1971). Suicidal behavior in chronic dialysis patients. American Journal of Psychiatry . 127 . 1199-1204. Afifi, A.K., & Bergman, R.A. (1980). Basic Neuroscience . Baltimore, Maryland: Urban & Schwarzenberg. Alfred, H.J., Graff, L.G., Cohen, A. J. , & Shuster, A. (1981). Treatment of Renal Failure. In B.D. Rose (Ed.), Pathophysiology of Renal Disease (pp. 475-573) . New York, New York: McGraw-Hill. Arieff, A.I., Guisado, R. , & Massry, S.G. (1975). Uremic encephalopathy: Studies on biochemical alterations in the brain. Kidney International . 7, Supplement 194-200. Atkinson, R.C., & Shiffrin, R.M. (1968). Human memory: A proposed system and its control processes. In K.W. Spence & J.T. Spence (Eds.), The Psychology of Learning and Motivation; Advances in Research and Theory ^ 2, New York, New York: Academic Press. Atkinson, R.C, & Shiffrin, R.M. (1971). The control of shortterm memory. Scientific American . 225 , 82-90. Bartle, E.J, Sun, J.H, Thompson, L. , Light, A.I., McCool C. , & Heaton S. (1988). The effects of acute sleep deprivation during residency training. Surgery, 104 . 311-316. Battig, W.F., & Montague, W.E. (1969). Category norms for verbal items in 56 categories: A replication and extension of the Connecticut category norms. Journal of Experimental Psychology Monograph , 80 , 1-46. Ben-Shachar, D. , Ashkenazi, R. , & Youdim, M.B. (1986). Longterm consequence of early iron-deficiency on dopaminergic neurotransmission in rats. International Journal of Developmental Neuroscience . 4, 81-88. Berry, D.T., Webb, W.B., Block, A.J., Bauer, R.M. , & Switzer, D.A. (1986) . Nocturnal hypoxia and neuropsychological variables. Journal of Clinical & Experimental Neuropsy chology . 8, 229-238. (Medline Abstract) . 162

PAGE 171

163 \ 1 Blrkmayer, W. , & Birkmayer, J.D. (1986). Iron, a new aid in ] the treatment of Parkinson patients. Journal of Neural » Transmission, 67, 287-292. i Birkmayer, W. , & Birkmayer, J.D. (1987). Dopamine action and disorders of neurotransmitter balance. Gerontology 33 . 16871. . i Blatt, B. , & Tsushima, W.T. (1966). A psychological survey of ] uremic patients being considered for the chronic 'I hemodialysis program. Intellectual and emotional problems ' in uremic patients. Nephrology . 2/ 206-208. Brenner, B.M. , & Lazarus, J.M. (1980). Chronic Renal Failure In K. Isselbacher, R. Adams, E. Braunwald, R. Petersdorf, & > J. Wilson (Eds.), Harrison^s Principles of Internal Medicine (9th ed.). (pp. 1299-1307). New York, New York: Mcgraw-Hill. Bunn, H.F. (1980a). Anemia associated with chronic systemic disorders. In K. Isselbacher, R. Adams, E. Braunwald, R. 1 Petersdorf, & J. Wilson (Eds.), Harrison^s Principles of Internal Medicine (9th ed.). (pp. 1530-1532). New York, New York: Mcgraw-Hill. | •j Bunn, H. F. (1980b) . Pallor and Anemia. In K. Isselbacher, R. Adams, E. Braunwald, R. Petersdorf, & J. Wilson (Eds.), Harrison^ s Principles of Internal Medicine (9th j ed.). (pp. 262-272). New York, New York: Mcgraw-Hill. Canadian Erythropoietin Study Group (1989). The effect of recombinant erythropoietin (EPO) upon quality of life (QL) and functional capacity (FC) of anemic patients on chronic hemodialysis. Kidney Internatinal , 35 ^ 195. Abstract. Carroll, J.B., Davies, P., & Richman, B. (1971). The American Heritage Word Frequency Book . New York, New i York: American Heritage Publishing Company. i \ Cermak, L.S. (1982). The Long and Short of It in Amnesia. In L.S. Cermak (Ed.), Human Memory and Amnesia (pp. 43-59). Hillsdale, New Jersey: Lawrence Erlbaum Associates. Chandra, M. , McVicar, M. , & demons, G. K. (1988). Pathogenesis of the anemia of chronic renal failure: the role of erythropoietin. Adv. Pediatr [200] 35, 361-389. Cohen, R.M., Weingartner, H. , Smallberg, S.A., Pickar, D. , & 1 Murphy, D.L. (1982). Archives of General Psvchiatry . 39 . ] 593-597. ^

PAGE 172

164 Comty, CM., Leonard, A., Se Shapiro, F.C. (1974). j Psychosocial problems in dialized diabetic patients. I Kidney International . 1, Supplement 144-151. I Craik, F.I.M., & Lockhart, R.S. (1972). Levels of I processing: a framework for memory research. Journal of Verbal Learning and Verbal Behavior . 11 , 671-684. \ Craik, F.I.M., & Tulving, E. (1975). Depth of processing and 1 the retention of words in episodic memory. Journal of I Experimental Psvchology; General , 104 , 268-294. Crittenden, M.R., Holliday, M.A. , Piel, C.F., & Potter, D.E. (1985). Intellectual development of children with renal insufficiency and end stage disease. International ^ Journal of Pediatric Nephrology . 6, 275-280. \ 'i } Darwin, C.J., Turvey, M.T., & Crowder, R.G. (1972). An i auditory analogue of the Sperling partial report t procedure: Evidence for brief auditory storage. Cognitive I Psychology, 3, 255-267. fli Davies-Jones, G.A.B., Preston, F.E., & Timperley, W.R. | (1980) . Neurological Complications in Clinical ' Hematology . Boston, Massachusetts: Blackwell Scientific | Publications. :; i Dawson, A. A., Ogston, D. , & Fullerton, H.W. (1969). ^ Evaluation of diagnostic significance of certain symptoms and physical signs in anemic patients. British Medical Journal, 2, 436-439. Deinard A.S., List, A., Lindgren, B. , Hunt, J.V. , & Chang, P. (1986) . Cognitive deficits in iron-deficient and irondeficient anemic children. Journal of Pediatrics . 108 , 681-689. ' Delano, B.G. (1989) . Improvements in quality of life following treatment with r-HuEPO in anemic hemodialysis patients. American Journal of Kidney Diseases , 14, 14-18. l Delis, D.C., Freeland, J., Kramer, J.H. , & Kaplan, E. ^ (1988) . Integrating clinical assessment with cognitive ] neuroscience: Construct validation of the California !. Verbal Learning Test. Journal of Consulting and Clinical \ Psvchology . 56. 123-130. ] Delis, D.C., Kramer, J.H., Kaplan, E. , & Ober, B.A. (1987). I California Verbal Learning Test Adult Version Research i Edition Manual . New York, New York: Psychological | Corporation. J

PAGE 173

165 Delis, D.C., McKee, R. , Massman, P., Kramer, J.H., & Gettman, D. (1990) . Alternate Form of the California Verbal Learning Test: Development and Reliability, unpublished manuscript. Desforges, J.F. (1975). The role of hemodialysis in the hematologic disorders of uremia. Kidney International . 7, S-123-128. Drayer, B.P., Burger, P., Darwin, R. , Reiderer, S., Herfkens, R. , & Johnson, G.A. (1986). Magnetic resonance imaging of brain iron. Americal Journal of Neuroradiology f 7, 373-380. English, A., Savage, R.D., Britton, P.O., Ward, M.K., & Kerr, D.N.S. (1978). Intellectual impairment in chronic renal failure. British Medical Journal . 1, 888-890. Erslev, A.J. (1975). The effect of uremic toxins on the production and metabolism of erythropoietin. Kidney Internation . 7, S129-S133. Erslev, A.J. (1987) . Erythropoietin coming of age. New England Journal of Medicine . 316 . 101-103. Eschbach, J.W. (1989) . The anemia of chronic renal failure: pathophysiology and the effects of the effrects of recombinant erythropoietin. Kidney International . 35, 134-148. Eschbach, J.W., & Adamson, J.W. (1985). Anemia of end-stage renal disease (ESRD) . Kidney International . 28, 1-5. Eschbach, J.W., Egrie, J.C, Downing, M.R., Browne, J.K., & Adamson, J.W. (1987) . Correction of the anemia of endstage renal disease with recombinant human erythropoietin. New England Journal of Medicine . 316 . 7378. Evans, D.I. (1985). Cerebral function in iron deficiency: A review. Child-Care. Health and Development 11, 105-112. Evans, R.W., Rader, B. , & Manninen (1990). The quality of life of hemodialysis recipients treated with recombinant human erythropoietin. Journal of the American Medical Association. 263, 825-830. Ewing, R. , McCarthy, D. , Gronwall, D. , & Wrightson, P. (1980) . Persisting effects of minor head injury observable during hypoxic stress. Journal of Clinical Neuropsychology , 2, 147-155.

PAGE 174

166 I ' ~ . /••. Fennell, R.S., Fennell, E.B., Carter, R.L., Mings, E.L., Klausner, A.B., & Hurst, J.R. (1989). Correlations between performance on neuropsychological tests of children with chronic renal disease. Unpublished. Fennell, R.S., Fennell, E.B., Carter, R.L., Mings, E.L., Klausner, A.B., & Hurst, J.R. (1990a). Association between renal function and cognition in childhood chronic renal failure. Pediatric Nephrology . 4, 16-20. Fennell, R.S., Fennell, E.B., Carter, R.L., Mings, E.L., Klausner, A.B., & Hurst, J.R. (1990b). A longitudinal study of the cognitive function of children with renal failure. Pediatric Nephrology ,, 4./ 11-15. Fennell, R.S., Fennell, E.B., Mings, E. L. , Klausner, A. B., Hurst, J.R., & Carter, R.L. (1987). Cognitive performance in children with renal failure as affected by renal function and therapeutic modalities. In K. Murakami, T. Kitagawa, K. , Yabuta, & T. Sakai (Eds.), Recent Advances in Pediatric Nephrology (pp. 615-618) . New York, New York: Excerptica Medica. Fennell, E.B., Fennell, R.S., Mings, E. , & Morris, M.K. (1986) . The effects of various modes of therapy for end stage renal disease on cognitive performance in a pediatric population: A preliminary report. International Journal of Pediatric Nephrology . 7, 107-112. Fennell, R.S., & Rasbury, W.C. (1980). Cognitive functioning of identical twins discordant for Prune Belly Syndrome and end-stage renal disease. International Journal of Pediatric Nephrology , 1, 234-239. Fennell, R.S., Rasbury, W.C, Fennell, E.B., & Morris, M.K. (1984) . Effects of kidney transplantation on cognitive performance in a pediatric population. Pediatrics . 74, 273-278. Fishman, D.B., & Schneider, C.J. (1972). Predicting emotional adjustment in home dialysis patients and their relatives. Journal of Chronic Diseases . 25 , 99-109. Freeman, C.W., Sherrard, D.J., Carlsyn, D.A., & Paige, A.B. (1980) . Psychological assessment of renal dialysis patients using standard psychometric techniques. Journal of Counseling and Clinical Psychology . 48 . 537-539. Fried, R. (1987) . The Hvperventillation Syndrome; Research and Clinical Treatment . Baltimore, Maryland: Johns Hopkins University Press. Gilli, P., & DeBastiani, P. (1983). Cognitive function and dialysis treatment. Clinical Neurology , 16 , 188-192.

PAGE 175

167 Ginn, H.E., (1975). Neurobehavioral dysfunction in uremia. Kidney International ^ 2(S) , 217-221. Ginn, H.E., Teschan, P.E., Bourne, J.R., Hamel, B. , Ward, J.W., Vaugh, W.K., & Nunnally, J.C. (1978). Neurobehavioral and clinical responses to hemodialysis. Trans. Am Soc Artif Intern Organs. 24., 376-378. Ginn, H.E., Teschan, P.E., Walker, P.J., Bourne, J.R., Fristoe, M. , Ward, J.W., McLain, L.W. , Johnston, H.B., & Hamel, B. (1975). Neurotoxicity in Uremia. Kidney International . 7, S357-S360. Goldberg, A. P., Geltman, E.M., Gavin, J.R, 3rd., Carney, R.M., Hagberg, J.M., Delmez, J.A. , Naumovich, A., Oldfield, M.H., & Harter, H.R. (1986). Exercise training reduces coronary risk and effectively rehabilitates hemodialysis patients. Nephron . 42 . 311-316. Groner J.A. , Holtzman, N.A., Charney, E., & Mellits, E.D. (1986) . A randomized trial of oral iron on tests of short-term memory and attention span in young pregnant women. Journal of Adolescent Health Care , 7, 44-48. Gutman, R.A., Stead, W.W., & Robinson, R.R. (1981). Physical activity and employment status of patients on maintenance hemodialysis. New England Journal of Medicine . 304 . 309313. Hagberg, B. (1974) . A prospective study of patients in chronic hemodialysis-III : Predictive value of intelligence, cognitive deficit, and ego defense structures in rehabilitation. Journal of Psvchosocial Research . 18, 151-160. Hallgren, B. , & Sourander, P. (1958). The effect of age on the nonheme iron in the human brain. Journal of Neurochemistry , 2, 41-51. Hart, R.P., Pederson, J.A., Czerwinski, A.W., & Adams, R.L. (1983) . Chronic renal failure, dialysis and neuropsychological function. Journal of Clinical Neuropsychology , 5, 301-312. Hasher. L. , & Zacks, R.T. (1984). Automatic Processing of Fundamental Information. American Psychologist . 39, 13721388. Heilman, K.M., Moyer, R.S., Melendez, F., Schwartz, H.D., & Miller, B.D. (1975). A memory deficit in uremic encephalopathy. Journal of Neurological Science . 26, 245249.

PAGE 176

168 Heistadt, D.D., Marcus, M.L., & Abboud, F.M. (1987). Experimental attempts to unmask the effects of neural stimuli on cerebral blood flow. In M.J. Purves, (Ed.), Cerebral Vascular Smooth Muscle and Its Control (pp. 142165) . Amsterdam, Netherlands: Elsevier. Hennessy, T.G., Stern, W.E., & Herrick, S.E. (1967). Cerebellar Hemangioblastoma: Erythropoietic Activity by Radio-iron Assay. Journal of Nuclear Medicine . 8, 601606. Honig, A.S., & Oski, F.A. (1984). Solemnity: a clinical risk index for iron deficient infants. Early Child Development and Care . 16, 69-83. Hori, K., Onoyama, K. , Iseki, K. , Fujimi, S., & Fujishima, M. (1990) . Hemodynamic and volume changes by recombinant human erythropoietin (rHuEPO) in the treatment of anemic hemodialysis. Clinical Nephrology ^ 33 . 293-298. Hunt, K.P., & Hodge, M.H. (1971). Category-item frequency and category-name meaningfulness (m'): Taxonomic norms for 84 categories. Psychonomic Monograph Supplements . 4, 97-121. Jackson, M. , Warrington, E.K., Roe, C.J., & Baker, L.R.I. (1987). Cognitive function in hemodialysis patients. Clinical Nephrology ^ 27, 26-30. Jacob, H.S., Eaton, J.W. , & Yawata, Y. (1975). Shortened red blood cell survival in uremic patients: beneficial and deleterious effects of dialysis. Kidney International . 7, S-139-143. Johnson, D.L., & McGowan, R.J. (1983). Anemia and infant behavior. Nutrition and Behavior . 1, 185-192. Jonides, J., & Naveh-Benjamin, M. (1987). Estimating Frequency of Occurrence. Journal of Experimental Psychology: Learning. Memory, and Cognition . 13 . 230-240. Kaplan-DeNour, A., Shaltiel, J., & Czaczbes, J.W. (1968). Emotional reactions of patients on chronic hemodialysis. Psychosomatic Medicine . 300 , 521-533. Klein, W.L., Frederickson, E.D., Fennell, E.B., Peterson, J.C, Guzman, S.H., & Tisher, C.C. (1989). Effects of erythropoietin on neuropsychological function and reported mood in the anemia of end stage renal disease. American Federat ion for Clinical Research (Abstract) .

PAGE 177

169 Koene, R.A.P., & Frenken, L.A.M. (1990). Does treatment of predialysis patients with recombinant human erythropoietin compromise renal function? In R.M. Schaefer, W.H. Horl, & A. Heidland (Eds.)/ Erythropoietin in the 90s . New York, New York: Karger. Lee, G.R., Wintrobe, M.M. , & Bunn, H.F. (1980). Irondeficiency anemia and the sideroblastic anemias. In K.J. Isselbacher, R.D. Adams, E. Braunwald, R.G. Petersdorf, & J.D. Wilson (Eds.), Harrison^ s Principles of Internal Medicine (9th ed.). (pp. 1514-1518). New York, New York: McGraw Hill. Lewis, E.G., O'Neill, W.M. , Dustman, R.E., & Beck, E.G. (1980) . Temporal effects of hemodialysis on measures of neural efficiency. Kidney International . 17 . 357-363. Lezak, M. D. (1983) . Neuropsychological Assessment . (2nd ed.). New York, New York: Oxford U. Press. Lozoff , B. (1989) . Nutrition and Behavior. American Psychologist . 44 . 231-236. Lundin, A. P., Delano, B.G., & Quinn-Cefaro, R. (1990). Perspectives on the improvement of quality of life with epoetin alfa therapy. Pharmacotherapy . 10 . 228-268. Marshall, J.R. (1979) . Neuropsychiatric aspects of renal failure. Journal of Clinical Psychiatry . 40 . 81-85. Massaro T.F., & Widmayer, P. (1981). The effect of iron deficiency on cognitive performance in the rat. American Journal of Clinical Nutrition ^ 34 . 864-870. Mayer, G., Thum, J., Cada, E.M., Stummvoll, H.K., & Graf, H. (1988) . Working capacity is increased following recombinant human erythropoietin treatment. Kidney International , 34 . 525-528. Mccarty, S.M., Logue, P.E., Power, D.G., Ziesat, H.A., & Rosenstiel, A.K. (1980) . Alternate form reliability and age-related scores for Russell's Revised Wechsler Memory Scale. Journal of Consulting and Clinical Psychology , 48, 296-298. McDaniel, J.W. (1971). Metabolic and CNS correlates of cognitive dysfunction with renal failure. Psvchophysiology . 8, 704-713. Miller, G.A. (1956) . The magical number seven, plus or minus two: Some limits on our capacity to process information. Psychological Review , 63 , 81-97.

PAGE 178

170 Mings, E. L. (1987) . Visual Motor Integraton Deficits in Pediatric Renal Disease . Doctoral dissertation, University of Florida. Miskin, M. (1978) . Memory in monkeys severely impaired by combined but not by separate removal of amygdala and hippocampus. Nature ^ 273 . 297-298. Moray, N. (1959) . Attention in dichotic listening: Affective cues and the influence of instructions. Quarterly Journal of Experimental Psychology . 11 , 59-60. Morris, CD., Bransford, J.D., & Franks, J.J. (1977). Levels of processing vs. transfer appropriate processing. Journal of Verbal Learning and Verbal Behavior . 16 . 519533. Muramoto, O. , Kuru, Y., Sugishita, M. , & Toyokura, Y. (1979) . Pure memory loss with hippocampal lesions. A pneumoencephalographic study. Archives of Neurology , 36 , 54-56. Newman, R.P., Weingartner, H. , Smallberg, S.A., & Calne, D.B. (1984) . Effortful and automatic memory: Effects of dopamine. Neurology , 34 , 805-807. Nissen, M.J. Knopman, D.S., & Schacter, D.L. (1987). Neurochemical dissociation of memory systems. Neurology , 789-794. Nissenson, A.R. (1989) . Recombinant human erythropoietin: Impact on brain and cognitive function, exercise tolerance, sexual potency, and quality of life. Seminars in Nephrology , 9, 25-31. Nissenson, A.R., Levin, M.L., Klawans, H.L., & Nausieda, P.L. (1977) . Neurological sequelae of end-stage renal disease. Journal of Chronic Disease , 30 . 705-733. Nissenson, A.R., Marsh, J.T., & Brown, W.S. (1988). Recombinant erythropoietin improves brain function in chronic hemodialysis patients. Kidney International . 33 , 232. Abstract. Nissenson, A.R., Marsh, J.T., Brown, W.S., Schweitzer, S., & Wolcott (1989) . Brain function improves in chronic hemodialysis patients after recombinant erythropoietin. Kidney International . 35 , 257. Abstract. Osberg, J. W. , Meares, G.J., McKee, D.C., & Burnett, G.B. (1982). Intellectual functioning in renal failure and chronic dialysis. Journal of Chronic Diseases . 35, 445457.

PAGE 179

171 Oski, F.A., & Honig, A. (1983). The effects of therapy on the developmental scores of iron-deficient infants. Journal of Pediatrics . 92, 21-25. Oski F.A., Honig A.S., Helu B.M. , & Howanitz P. (1978). Effect of iron therapy on behavior performance in nonanemic iron-deficient infants. Pediatrics . 71 . 877-880. Paganini, E.P. (1989) . Overview of anemia associated with chronic renal disease: Primary and scondary mechanisms. Seminars in Nephrology ^ 9, Supplement 3-8. Peterson, L.R., & Peterson, M.J. (1959). Short-term retention of individual verbal items. Journal of Experimental Psychology . 58 . 193-198. Pollitt, E., Leibel, R.L., & Greenfield, D.B. (1983). Iron deficiency and cognitive test performance in preschool children. Nutrition and Behavior . 1, 137-146. Pollitt, E., Siemantri, A.G. , Yunis, F. , & Scrimshaw, N.S. (1985) . Cognitive effects of iron-deficiency anemia (Letter to the editor). Lancet , 1, 158. Powell. D. , Bergstrom, J., Rastislav, D. , Gulyassy, P., Lockwood, D., & Phillips, L. (1986). Toxins and inhibitors in chronic renal failure. American Journal of Kidney Diseases , 7(4) 292-299. Race, G.J., Finney, J.W., Mallams, J.T., & Balla, G.A. (1964) . Hematopoietic stimulating effect of a Cerebellar Hemangioblastoma. Journal of the American Medical Association . 187, 150-151. Rapaport, S.I. (1987). Introduction to Hematology (2nd ed.), Philadelphia, Pennsylvania: J.B. Lippincott. Rasbury, W.C, Fennell, R.S., Fennell, E.B., & Morris, M.K. (1986) . Cognitive functioning in children with end stage renal disease preand post-dialysis session. International Journal of Pediatric Nephrology , 7, 45-50. Rasbury, W.C, Fennell, R.S, & Morris, M.K (1983). Cognitive functioning of children with end-stage renal disease pre and post successful transplant. Journal of Pediatrics . 102 . 589-592. Ratner, D.P., Adams, K.M. , Levin, N.W. , & Rourke, B.P. (1983) . Effects of hemodialysis on the cognitive and sensory-motor functioning of the adult chronic hemodialysis patient. Journal of Behavioral Medicine . 6, 291-311.

PAGE 180

172 Ritchey, A.K. (1987) . Iron deficiency in children. Postgraduate Medicine . 82 . 59-69. Roediger, H.L. (1990). Implicit Memory: Retention without remembering. American Psychologist. 45 ,, 1043-1056. Rundus, D. (1971) . Analysis of rehearsal processes in freerecall. Journal of Experimental Psychology , 89 . 63-77. Rundus, D. (1977) . Maintenance rehearsal and single-level processing. Journal of Verbal Learning and Verbal Behavior . 16 . 665-681. Ryan, J.J., Geisser, M.E., Randall, D.M., & Georgemiller, R.J. (1986) . Alternate form reliability and equivalency of the Rey Auditory Verbal Learning Test. Journal of Clinical and Experimental Neuropsychology ^ 8, 611-616. Ryan, J.J., Souheaver, G.T., & DeWolfe, A.S. (1980). Intellectual deficit in chronic renal failure: A comparison with neurological and medical-psychiatric patients. Journal of Nervous and Mental Diseases , 168 , 763-767. Ryan, J.J., Souheaver, G.T., & DeWolfe, A.S. (1981). Halstead-Reitan test results in chronic hemodialysis. Journal of Nervous and Mental Diseases , 169 . 311-314. Rybo, E., Bengtsson, C. , Hallberg, L. , & Oden, A. (1985). Effect of iron supplementation on women with iron deficiency. Scandinavian Journal of Haematology . 34 , (Supplement 43), 103-114. Sanders, R.E., Gonzalez, E.G., Murphy, M.D., Liddle, C.L., & Vitina, J.R. (1987) . Frequency of Occurrence and the criteria for automatic processing. Journal of Experimental Psychology : Learning. Memory, and Cognition . 13, 241-250. i ^ Schlotzhauer, S.D., & Littell, R.C. (1987). SAS System for Elementary Statistical Analysis . Cary, North Carolina: SAS Institute. ' ^ ' .:' ' *.i'%.' . Schreiner, G.E. (1975). The search for the uremic toxin(s) . Kidney International . 7, S270-S271. Schupak, E., Sullivan. J.F., & Lee, D.Y. (1967). Chronic hemodialysis in "unselected" patients. Annals of Internal Medicine , 67, 708-717. Shiffrin, R. (1975) . Short-term store: The basis for a memory system. In F. Restle, R.M. Shiffrin, N.J. Castellan, H.R. Lindman, & D.B. Pisoni (Eds.), Cognitive Theory . 1, Hillsdale, New Jersey: Erlbaum.

PAGE 181

niKf;'; -'• 173 Souheaver, G.T., Ryan, J.J., & DeWolfe, A.S. (1982). Neuropsychological patterns in uremia. Journal of Clinical Psychology . 38 . 490-496. Sperling, G. (1960) . The information available in brief visual presentations. Psychological Monographs ^ 74 (whole no. 498) . Sprague, S.M., Corwin, H.L., Tanner, CM., Wilson, R.S., Green, B.J., & Goetz, C.G. (1988). Relationship of aluminum to neurocognitive dysfunction in chronic dialysis patients. Archives of Internal Medicine . 148 ,, 2169-2172. Squire, L.R. (1987) . Memory and Brain . New York, New York: Oxford University Press. Stewart, R.S., & Stewart, R.M. (1979). Neuropsychiatric aspects of chronic renal disease. Psychosomatic . 20, 524527. Teschan, P.E., Ginn, H.E., Bourne, J.R., & Ward, J.W. (1976) . Neurobehavioral responses to "middle molecule" dialysis and transplantation. Transactions of American Society of Artificial Internal Organs ^ 22 . 190-194. Teschan, P.E., Ginn, H.E., Walker, P.J., Bourne, J.R., Fristoe, M. , & Ward, J.W. (1974). Quantified function in uremic patients on maintenance dialysis. Transactions of American Society of Artificial Internal Organs . 20 . 388394. Thorndike, E.L., & Lorge, I. (1944). The teacher ^s word book of 30.000 words . New York, New York: Teachers College. Trieschmann, R.B., & Sand, P.L. (1971). WAIS and MMPI correlates of increasing renal failure in adult medical patients. Psychology Report , 29, 1251-1262. Trompeter, R.S., Polinsky, M.S., Andreoli, S.A., & Fennell, R.S. (1986) . Neurological complications of kidney failure. American Journal of Kidney Diseases , 7, 318-323. Tucker, D.M. , Sandstead, H.H., Penland, J.G., Dawson, S.L., & Milne, D.B. (1984). Iron status and brain functions: serum ferritin levels associated with asymmetries of cortical electrophysiology and cognitive performance. American Journal of Clinical Nutrition ^ 39 . 105-113. Tyler, H.R. (1968) . Neurologic disorders in renal failure. American Journal of Medicine . 44 , 734-748.

PAGE 182

174 Uyeda, K.M., & Handler, G. (1980). Prototypicality norms for 28 semantic categories. Behavior Research Methods & Instrumentation . 12, 587-595. Van Wyck, D.B. (1989). Iron deficiency in patients with dialysis-associated anemia during erythropoietin replacement therapy: Strategies for assessment and management. Seminars in Nephrolocfy . 9, 21-24. Vander, A.J., Sherman, J.H., & Luciano, D.S. (1980). Human Physiology . New York, New York: McGraw-Hill. Waldmann, T.A., Levin, E.H., & Baldwin, M. (1961). The association of polycythemia with a cerebellar hemangioblastoma. American Journal of Medicine . 31 , 318324. Walter, T., Kovalskys, J., & Steel, A. (1983). Effect of mild iron deficiency on infant mental development scores. Journal of Pediatrics . 102 ^ 519-522. Wechsler, D. (1987) . Wechsler Memory Scale Revised Manual . New York, New York: The Psychological Corporation. White, N. , & Cunningham, W.R. (1988). Is Terminal Drop Pervasive or Specific? Journal of Gerontology; Psychological Sciences . 43., 141-144. Williams, S.R. (1985). Nutrition and Diet Therapv (5th ed.). St. Louis, Missouri: Times Mirror/Mosby. Winearls, C.G., Oliver, D.O., Pippard, M.J., Reid, C, Downing, M.R., & Cotes, P.M. (1986). Effect of human erythropoietin derived from recombinant DNA on the anemia of patients maintained by chronic hemodialysis. Lancet, 2, 1175-1177. Wintrobe, M.W. , Lee, G.R., Boggs, D.R. , Bithell, T.C., Foerster, J., Athens, J.W., & Lukens, J.N. (1981). Clinical Hematology (4th ed.). Philadelphia, Pennsylvania: Lea & Febiger. Wolcott, D.L., Marsh, J.T. , La-Rue, A., Carr, C. , & Nissenson, A.R. (1989) . Recombinant erythropoietin treatment may improve quality of life and cognitive function in chronic hemodialysis patients. American Journal of Kidney Disease . 14 . 478-485. Wolcott, D.L., Schweitzer, S., & Marsh, J.T. (1988). Recombinant erythropoietin improves cognitive function and quality of life of chronic hemodialysis patients. Kidney International ^ 33 , 242. Abstract.

PAGE 183

175 Wolcott, D.L. , Schweitzer, S., & Nissenson, A.R. (1989). Recombinant erythropoietin (r-EPO) improves cognitive function (CF) and quality of life (QL) of chronic hemodialysis (CHD) patients. Kidney International . 35 . 266. Abstract. Yehuda, S., Youdim, M. , & Mostofsky, D.I. (1986). Brain iron deficiency causes reduced learning capacity in rats. Pharmacology. Biochemistry and Behavior ^ 25 . 141-144. Youdim, M.B.H., & Ben-Schachar, D. (1987). Minimal brain damage induced by early iron deficiency: modified dopaminergic neurotransmission. Israel Journal of Medical Sciences . 23 . 19-23. Ziestat, H.A. Jr., Logue, P.E., & McCarty, S.M. (1980). Psychological measurement of memory deficits in dialysis patients. Perceptual Motor Skills . 50 ^ 311-318. •.,.-. r . . . • .", "•• -"» /' ' •» * ' ;

PAGE 184

BIOGRAPHICAL SKETCH Wayne L. Klein decided to become a psychologist in adolescence after giving up on religion and philosophy. As a college sophomore, academic psychology paled in comparison to human potential psychology; therefore, in 1971, he dropped out of SUNY at Albany and spent three months hitchhiking to Berkeley, California where he spent a seminal year emersed in Esalen groups, training at the Gestalt Institute of San Francisco and working in the Berkeley Free Clinic. After returning to SUNY at Albany to complete liberal arts requirements, he transferred to SUNY Empire State College, a university without walls program, and obtained a B.S. in counselling psychology. Over the next nine years he worked with a variety of populations in diverse settings as a counselor and instructor. During that time he also obtained a B.S. in biology by taking tests (Regents College) , took graduate courses in biopsychology and earned an M.S. in educational psychology (SUNY at Albany) . , Spurred by interest in cognition-behavior-brain relationships, he entered the University of Florida Clinical and Health Psychology doctoral program. After flirting with psychophysiology, he focused on clinical neuropsychology and pursued interrelated research interests in the 176

PAGE 185

,.j -..^v i^,V* t-, . ', J ' "... ^ 177 • . '.? , -w J neuropsychology of mineral deficiency and anemia, chemosensation and food preferences, pica and factors underlying anorexia. He also tested the claims that subliminal audio tapes affect mood and engaged in test construction. ^ His training in neuropsychology continues at the Tufts University School of Medicine Boston V.A. Medical Center Predoctoral Internship Consortium. Following internship he will complete an N.I.H. Postdoctoral Fellowship at the Boston University Aphasia Research Center. He has been married since 1980 and has one daughter, who was born in Gainesville. .;V

PAGE 186

w I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. (^}
PAGE 187

This dissertation was submitted to the Graduate Faculty of the College of Health Related Professions and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. August 1991 (iLSU^.O Q. "KlSoSb . . .^JUDean, College of Health Related Professions Dean, Graduate School

PAGE 188

UNIVERSITY OF FLORIDA 3 1262 08554 8112


94
I and II. Closer examination revealed a similar pattern for
the treatment group. This pattern is displayed in Table 4-
12. However, as shown in Table 4-13, the control group
exhibited poorer correlations between Form II and Form III
and between Form II and Form I than did the treatment group.
Table 4-11. Correlation of Sum CVLT Between Forms I, II and
III for All Subjects
CVLT II
Pearson Correlations
CVLT III
CVLT I
CVLT II
r
1.00000
0.75543
0.71063
P
0.0000
0.0001
0.0001
N
35
35
32
CVLT III
r
0.75543
1.00000
0.80342
P
0.0001
0.0000
0.0001
N
35
35
32
CVLT I
r
0.71063
0.80342
1.00000
P
0.0001
0.0001
0.0000
N
32
32
32
Note: r = correlation; p = probability value;
N = subject number
Table 4-14 demonstrates that Form III, the version
created for this study, correlates as well with both Forms I
and with II as Form I correlates with Form II. In addition,
means scores were remarkably similar between the three forms
of the CVLT.


31
solving capacity and improved STM and attention in adults
(Deinard, List, Lindgren, Hunt, & Chang, 1986; Groner,
Holtzman, Charney, & Mellits, 1986; Honig & Oski, 1984;
Oski, Honig, Helu, & Howanitz, 1978; Pollitt, Leibel, &
Greenfield, 1983; Pollitt, Siemantri, Yunis, & Scrimshaw,
1985; Rybo, Bengtsson, Hallberg, & Oden, 1985; Walter,
Kovalskys, & Steel, 1983).
Some authors attribute the neurological complications
of iron deficiency to anemia (Davis-Jones, Preston, &
Timperley, 1980). However, to date not one study has
examined the effects of anemia controlling for the effects
of iron deficiency. Anatomical and clinical evidence
suggests that iron deficiency may affect dopaminergic
systems (Birkmayer & Birkmayer, 1986, 1987; Drayer, et al.,
1986; Hallgren & Sourander, 1958). The mechanism may be a
reduction in the number of dopamine D2 receptors (Ben-
Shachar, Ashkenazi, & Youdim, 1986).
Iron responsive neuropsychological deficits have been
demonstrated in nonanemic iron deficient pregnant women and
infants (Groner, Holtzman, Charney, & Mellits, 1986; Oski,
Honig, Helu, & Howanitz, 1983). Evans (1985) noted that
improved cognitive function occurs rapidly following iron
supplementation, prior to correction of the anemia;
therefore, the neuropsychological deficits associated with
iron deficiency do not appear to be mediated by anemia.
Reviewing the pediatric literature, Ritchey (1987) concluded
that iron deficiency, in the absence of anemia, adversely


117
starting at a higher level, failed to benefit from the
experience. However, a simple practice effect would not be
expected as test forms contained different words and
categories. Learning how to learn list learning from
minimal exposure to a list learning task and generalization
to subsequent list learning weeks later would be quite
extraordinary. Also, neither group showed an improvement
between the second and third assessments. It further
strains credulity to consider that maximal learning
occurred following a single exposure to a list learning
task. In addition, the practice effect hypothesis requires
the dubious assumption that prior to exposure to the list
learning task, the control group was poorer at list
learning than was the treatment group, but benefited more
from practice. Based on subject composition the control
group would be expected to perform slightly better; indeed,
they did on the second and third assessments. The
hypothesis of differential response to practice must be
rejected.
Form III of the CVLT, the version developed for this
study, was administered second to all subjects. If Sum
CVLT III were easier than Form I and Form II, then this
might explain the appearance of improved performance in the
control group at the second assessment and suggest
decreased performance in the treatment group. One of the
benefits of the present design is that it directly controls
for possible intertest form differences through the


66
further examined quantitatively (i.e., kurtosis and skew)
and graphically (i.e., box plots)
The estimation of frequency of occurrence data violated
the test of normality. The ratio of the standard deviation
to the mean was 0.262, just over the established limit. The
distribution was somewhat skewed to the left; however,
Kurtosis was minimal.
The Semantic score generated by the levels of
processing task failed the Shapiro-Wilks W test of
normality; however, the ratio of the standard deviation to
the mean was within acceptable limits. The Orthographic
variable from the levels of processing task also failed the
test of normality and the standard deviation was large
relative to the mean.
The two Cognitive-Affective and Physical Behavior
Questionnaire scales failed the Shapiro-Wilks W test and the
standard deviation to mean ratio was .280, slightly above
the .25 recommended cut-off. The scores were slightly
skewed to the left and the tails were heavier than expected
in a normal distribution. The variables age, years of
education and years on dialysis all failed the Shapiro-Wilks
test. However, the ratio of standard deviation to mean was
within accepted limits for the education variable. Slope
and intercept of the learning curve for CVLT trials 1-5,
variables used for post hoc analyses, both passed the
Shapiro-Wilks test. Slope had somewhat heavy tails.


22
impairment correlates with severity of uremia, is at times
dramatically improved by initiation of dialysis and may be
nearly reversed by successful transplantation (Fennell,
Rasbury, Fennell, & Morris 1984; Ratner, Adams, Levin, &
Rourke, 1983). Adults on maintenance dialysis frequently
exhibit deficits in attention, vigilance, reaction time,
ability to rapidly generate and/or shift sets, visuospatial
ability, visual memory and verbal memory. Weaker evidence
suggests variability correlating with the phase of the
dialysis cycle, a possible advantage for peritoneal dialysis
over hemodialysis and decreased performance in association
with time on hemodialysis (Fennell, Fennell, Mings, &
Morris, 1986; Osberg, Meares, McKee, & Burnett, 1982).
Higher Hg levels in peritoneal dialysis compared to
hemodialysis and the possibility of increasing anemia in
conjunction with time on hemodialysis leave open the
possibility that level of anemia may play a role in the
latter two observations. Restoration of near normal
cognitive function following transplantation in well
dialyzed subjects may also, to an extent yet to be
determined, be mediated by the amelioration of anemia.
There are a number of problems with this literature.
For example, small sample size, absence of alternate test
forms, lack of relevant control groups, and inadequate
statistical procedures (Osberg, Meares, McKee, & Burnett,
1982). Additional contributions to discrepancies in the
literature include high variance in subject characteristics


177
neuropsychology of mineral deficiency and anemia,
chemosensation and food preferences, pica and factors
underlying anorexia. He also tested the claims that
subliminal audio tapes affect mood and engaged in test
construction.
His training in neuropsychology continues at the Tufts
University School of Medicine Boston V.A. Medical Center
Predoctoral Internship Consortium. Following internship he
will complete an N.I.H. Postdoctoral Fellowship at the
Boston University Aphasia Research Center. He has been
married since 1980 and has one daughter, who was born in
Gainesville.


40
is severe. Despite this, the absence of change on the
fatigue and vigor subscales is somewhat suspect and may
represent the inability of this subject population to
accurately report their subjective state in the manner
reguired by the POMS. Wolcott, Schweitzer and Marsh (1988)
also found an absence of significant changes on the POMS
following a rEPO mediated increase in mean hematocrit from
23.1% to 36.0% in 13 patients; however, mean POMS vigor and
fatigue scales changed in the expected directions and
approached statistical significance.
As predicted, there was no improvement on the
Comprehension subtest of the Wechsler Adult Intelligence
Scale Revised. This measure of overlearned semantic
stores is usually stable in the absence of severe
dysfunction. There was also no improvement in verbal
behavioral fluency as measured by the Controlled Oral Word
Association Test (FAS). Neither was there any increase in
motor speed as measured by finger tapping; however, two
patients exhibited guite noticeable differences in capacity
for sustained finger activity. During the first assessment
finger cramping, pain and fatigue reguired rests between
trials. Rests were not reguired following initiation of
rEPO treatment.
On Trails A and B, a test stressing motor speed, visual
search, mental control and decision speed, no improvement
was noted. On Trails A, one treated subject moved from
borderline to within normal limits. One treated subject,


28
Polinsky, Andreoli, & Fennell, 1986). Attempts to isolate
specific uremic neurotoxins have met with limited success
(Schreiner, 1975; Trompeter, Polinsky, Andreoli, & Fennell,
1986).
Prior to the availability of rEPO, Hg values were
occasionally included in neuropsychological studies in renal
disease; however, perhaps due to the small range of values,
anemia was not found to be contributory (Fennell et al.,
1987; Gilli & DeBastiani, 1983). In continuous ambulatory
peritoneal dialysis (CAPD) Hg levels tend to run higher,
aspects of cognitive performance tend to be closer to normal
and "middle molecule" clearance is up to ten times greater.
Nissenson (1989) suggests that higher hemoglobin rather than
lower "middle molecule" levels may underlie the neuro
psychological advantage of CAPD over hemodialysis.
Numerous studies have documented the presence of
neuropsychological dysfunction in other forms of anemia.
However, not one of these studies has controlled for the
cause of the anemia, which in every case has been reported
to produce neuropsychological deficits. As all of the known
effects of rEPO are mediated by erythropoiesis, improvement
in function following rEPO replacement therapy may be
attributed solely to the rise in Hg and hematocrit.
Neuropsychological deficits are well documented in iron
deficiency anemia; however, while there is support for the
direct effect of iron deficiency on the brain, the role of
anemia controlling for iron deficiency has not been examined


59
improved verbal LTM as measured by Sum CVLT, compared to
dialysis subjects not receiving rEPO. This improvement was
hypothesized to be mediated by rise in Hg. Recombinant EPO
itself was not expected to have any effect on Sum CVLT.
Recombinant EPO treated subjects typically concurrently
receive iron supplementation; therefore, iron status was
not expected to change in the vast majority of subjects.
Nevertheless, it was necessary to control for iron status
because it has been a confound in the majority of studies
focusing on the neuropsychology of anemia as deficiency is
associated with neuropsychological deficits. A
multivariate analysis was expected to demonstrate a main
effect for Hg while controlling for rEPO treatment status
and iron status.
Secondary Hypotheses
The secondary hypotheses are described below.
The first secondary hypothesis was that self-reported
physical activity, as measured by the Cognitive-Affective
and Physical Behavior Questionnaire, would improve in
response to reduced anemia.
Despite self-reported changes in life functioning, the
pilot study failed to demonstrate any alteration in
affective state using a self-report adjective checklist,
possibly because subjects were unable to adeguately perform
the task, which required the ability to abstract and
introspect. Therefore, it was hypothesized that a
concrete, behaviorally oriented self-report device might


71
group of 46.49 years (sd=13.61) and a mean of 12.10
(sd=1.63) years of education. The results of a Wilcoxon 2-
Sample Test (S=267.00, Z=0.7177, p>.4729) failed to
demonstrate a significant age difference between the
treatment (n=17) and control (n=15) groups. Likewise, the
Wilcoxon 2-Sample Test procedure failed to demonstrate a
significant difference in years of education between the
treatment (n=17) and control (n=15) groups (S= 259.00,
Z=0.440, p>.6596).
Cognitive function has been reported to fall prior to
nonaccidental death (White & Cunningham, 1988). Consistent
with this phenomenon of terminal decline, all three subjects
were in the predicted direction, although in two cases the
drop was extremely small. For the three deceased subjects,
mean Sum CVLT in the last assessment prior to death was
25.56 compared to 28.70 in the previous assessment. The
overall effect of terminal drop was to slightly lower
control group assessment two and three scores relative to
the treatment group.
Time Intervals Between Assessments
The mean time span between assessments one and two was
51.12 days (sd=32.95) for the treatment group. Control
group mean time interval between the first and second
assessments was 56.50 days (sd=35.98). The 5.38 days
greater time interval for the control group failed to reach
statistical significance based on a Wilcoxon Rank Sum Test
(S= 277.500, Z=-.92647, p>=.3542).


172
Ritchey, A.K. (1987). Iron deficiency in children.
Postgraduate Medicine. 82. 59-69.
Roediger, H.L. (1990). Implicit Memory: Retention without
remembering. American Psychologist. 45. 1043-1056.
Rundus, D. (1971). Analysis of rehearsal processes in free-
recall. Journal of Experimental Psychology. 89. 63-77.
Rundus, D. (1977). Maintenance rehearsal and single-level
processing. Journal of Verbal Learning and Verbal
Behavior. 16. 665-681.
Ryan, J.J., Geisser, M.E., Randall, D.M., & Georgemiller,
R.J. (1986). Alternate form reliability and equivalency
of the Rey Auditory Verbal Learning Test. Journal of
Clinical and Experimental Neuropsychology. 8, 611-616.
Ryan, J.J., Souheaver, G.T., & DeWolfe, A.S. (1980).
Intellectual deficit in chronic renal failure: A
comparison with neurological and medical-psychiatric
patients. Journal of Nervous and Mental Diseases. 168.
763-767.
Ryan, J.J., Souheaver, G.T., & DeWolfe, A.S. (1981).
Halstead-Reitan test results in chronic hemodialysis.
Journal of Nervous and Mental Diseases. 169. 311-314.
Rybo, E., Bengtsson, C., Hallberg, L., & Oden, A. (1985).
Effect of iron supplementation on women with iron
deficiency. Scandinavian Journal of Haematology. 34.
(Supplement 43), 103-114.
Sanders, R.E., Gonzalez, E.G., Murphy, M.D., Liddle, C.L., &
Vitina, J.R. (1987). Frequency of Occurrence and the
criteria for automatic processing. Journal of
Experimental Psychology: Learning. Memory, and Cognition.
13, 241-250.
Schlotzhauer, S.D., & Littell, R.C. (1987). SAS System for
Elementary Statistical Analysis. Cary, North Carolina:
SAS Institute.
Schreiner, G.E. (1975). The search for the uremic toxin(s).
Kidney International. 7, S270-S271.
Schupak, E., Sullivan. J.F., & Lee, D.Y. (1967). Chronic
hemodialysis in "unselected" patients. Annals of Internal
Medicine. 67, 708-717.
Shiffrin, R. (1975). Short-term store: The basis for a
memory system. In F. Restle, R.M. Shiffrin, N.J.
Castellan, H.R. Lindman, & D.B. Pisoni (Eds.), Cognitive
Theory. 1, Hillsdale, New Jersey: Erlbaum.


148
Table F-2 continued
F'
Degrees
P
Variable
Value
of Freedom
value
Recognition Memory
Correct Positive:
1.58
50
49
0.1127
False Positive On List
. B:
No Relation
1.26
49
50
0.4118
Semantic Shared
1.16
49
50
0.6079
False Positive Not on
List B
Semantic Shared
2.49
50
49
0.0017 *
Phonemic Shared
2.93
50
49
0.0002 *
No Relation
2.35
49
50
0.0032 *
Verbal Fluency
CFP
1.02
47
50
0.9395
FAS
1.21
50
47
0.5181
LSW
1.08
50
47
0.7843
Levels of Processing
Semantic
1.00
47
44
0.9916
Phonological
1.10
47
44
0.7607
Orthographic
2.14
47
44
0.0124 *
Frequency of Occurrence
Free Recall
1.16
50
49
0.6092
Estimation
1.32
50
49
0.3361
Cognitive-Affective and Physical
Behavior Questionnaire
Cognitive-Affective
1.06
49
50
0.8251
Physical
1.04
49
50
0.8790
Appetite
2.72
50
49
0.0006 *
Hemoglobin
1.26
50
50
0.4119
Hematocrit
1.44
50
50
0.1983
Ferritin
2.82
44
13
0.0460 *
Age
1.11
17
16
0.8333
Education
2.10
16
17
0.1400
Note: = Unequal variance


76
control group (n=6) ferritin level was 523.62 ng/ml
(sd=626.13). A Kruskal-Wallis Test failed to reveal a
statistically significance difference between these two
groups
Table 4
(X2(1, n= 35)= 0.3478, p<.5553).
-1. Mean Hemocrlobin bv Grouo and
Assessment
Ass
Gro
N
Min
Max
Mean
Sd
pre trt
rEPO
17
4.80
12.70
7.39
1.94
1
rEPO
17
5.60
11.30
8.46
1.42
2
rEPO
17
6.60
14.20
9.59
1.84
3
rEPO
17
6.90
12.25
9.54
1.48
1
Con
18
5.27
13.90
9.44
1.92
2
Con
18
6.30
12.60
9.23
1.81
3
Con
15
6.00
12.30
9.62
1.92
Note: Ass = Assessment;
Grp =
Group; N =
Number of
subje<
Min = Minimum; Max = Maximum; Sd = Standard deviation
Con = Control; rEPO = rEPO treated group
Table 4-2. Mean Ho Change by Group and Test Intervals
Test
Interval
Gro
N
Min
Max
Mean
Sd
First
Con
18
-2.95
1.70
-0.21
1.34
Second
Con
15
-2.10
1.90
0.15
1.14
Pre-Ass
rEPO
17
-4.60
4.70
1.07
1.90
First
rEPO
17
-0.70
5.90
1.13
1.57
Second
rEPO
17
-5.50
1.80
-0.05
1.59
Total
rEPO
17
-2.40
5.30
2.15
1.86


43
Improved performance on Sum CVLT, in the absence of
other changes, is hypothesized to be due to reduction in
fatigue. It is hypothesized that the effects of fatigue are
likely to appear along a continuum mediated by attributions
regarding the importance of effortful activities. With
increasing fatigue the saliency of cues requisite to the
production of a given level of arousal is likely to change.
Nonessential, covert cognitive activity may diminish prior
to cognitive activity overtly and undeniably essential to
performance of the stated task. Therefore, do to the nature
of the tasks, neuropsychological measures directly stressing
overt attention and vigilance may be less affected.
However, tasks benefiting from but not requiring covert
effortful cognitive processing may be most vulnerable to the
effects of fatigue.
The dramatic reduction in anemia produced by treatment
with EPO has been shown to produce clinically significant
improvements in domains such as aerobic capacity, anaerobic
threshold, sexual potency, employment status, level of
social activity, perceived ability to engage in physical
exercise and improved appetite. In contrast significant
changes in self-reported mood have been less consistently
documented. Preliminary neuropsychological test data
suggests improvement on verbal supra-span multitrial list
learning tasks. In addition, preliminary reports suggest
possible improvement in other cognitive domains as measured


118
creation of difference scores. However, absence of
randomization eliminated the ability to statistically test
for interactions between test instrument and group or order
effects.
Although CVLT III might theoretically be somewhat
easier based on the word frequency rates of the target
words, as described in Appendix A, administration of Forms
I and III to 18 normal, native English speaking volunteers,
primarily university undergraduates, revealed identical Sum
CVLT mean scores for both groups. Moreover, scores were
remarkably similar between Forms I, II and III in the
treatment group. In addition, scores between Forms II and
III were essentially identical in the controls. In the
treatment group, the correlation was an impressive .92
between Sum CVLT Forms II and III and .81 between Forms I
and III. In contrast, the correlation between Forms I and
II, tests arguably considered equivalent, was .77 for Sum
CVLT. For comparison, in the validation of Form II, the
correlation between Sum CVLT on Forms I and II was .84.
Taken together, these results suggest that Sum CVLT III is
equivalent to the two previous forms. Differences in test
instruments appear unable to account for the observed
pattern of scores.
Psychological explanations may be posited; however,
none elegantly fit the data. Nevertheless, it should be
noted that subject expectations were poorly controlled.
Most hemodialysis patients spent three half-days per week


123
toward an inverse relationship between Hg change and Sum
CVLT was noted, and confirmed statistically by post hoc
analyses. Curiously, Sum CVLT performance was temporarily
below expectations in subjects who would later exhibit a
drop in Hg. Stated differently, between the first and
second assessments, Sum CVLT rose when Hg fell. After
exploring a variety of hypotheses, it was concluded that
the cause is unknown at this time.
Future studies of the impact of rEPO on nervous system
function in the anemic patient need address a number of
problems discussed in these pages. Many of the more
serious flaws in the present study were a function of the
release of rEPO for clinical use. Given that under current
medical protocols it may be that the severely anemic
chronic renal patient may become rare, future studies
focusing on the neuropsychology of chronic anemia may have
to work with other populations. Nevertheless, despite
increasing difficulties in working with this population,
optimal care of the renal patient demands well controlled
studies of the effect of rEPO mediated changes in anemia on
neuropsychological function.


57
Tests were administered in the same order at all three
assessments and all subjects were administered test
versions in the same order. Forms for the estimation of
frequency of occurrence and levels of processing paradigms
were administered in ascending order. For the CVLT, Form
II was administered at the first assessment and Form I at
the third assessment. This deviation from ideal
experimental design occurred because Form III had not been
fully validated prior to initiation of the study.
Therefore, to minimize losses in the event of poor inter
form reliability, Form III was administered at the second
assessment and forms were not counterbalanced.
Controlled Word Association letter triplets were
administered in the following order: CFL, FAS and PRW.
Because the letter triplets CFL and PRW can be directly
compared, the were placed in what were expected to be the
two most important assessments, one and three. FAS is the
most widely used letter triplet, but shares a letter with
CFL. To minimize possible practice effects for individual
letters in the comparison of CFL and PRW, CFL was
administered prior to FAS. As change scores rather than
absolute performance was compared, error variance stemming
from possible differences in difficulty between the three
letter triplets was minimized.
Tasks were administered in the following order:
Informed consent, collection of demographic data, CVLT,
levels of processing encoding phase, frequency of


125
categories with two of the categories shared by both lists.
Words are balanced with regard to the rank order of each
word as an exemplar of the category as measured by Battig
and Montague (1969), Hunt and Hodge (1971) and Uyeda and
Mandler (1980). The primary and interference lists of each
form contain two categories in common and two unrelated
categories.
TABLE A-l. Mean Word Frequency and Typicality Across Forms
FORM I
a b
word word
freer freer
c
cat
rank
FORM II
a b
word word
freer freo
C
cat
rank
FORM III
a be
word word cat
freg freg rank
Primarv list
M 10.81 41.66
13.31
12.2
43.6
-
15.7
43.8
8.1
Sd 9.91 8.65
6.73
8.2
10.5
-
15.9
5.5
4.3
Interference list
M 14.13 43.56
14.94
7.7
42.8
-
16.7
45.4
12.3
Sd 13.9 9.83
9.28
7.8
9.5
-
19.5
4.8
8.8
Note: a word freq = word frequency
(Thorndike &
Lorge,
1944); b word freq = word frequency (Carroll, Davies,
& Richman, 1971) ; c cat rank = typicality ranking
(Battig & Montague, 1969); M = mean; Sd = standard
deviation
Table A-l compares the word frequency characteristics
of Forms I, II and III with regard to frequency of
appearance in English reading as measured by Thorndike and


163
Birkmayer, W., & Birkmayer, J.D. (1986). Iron, a new aid in
the treatment of Parkinson patients. Journal of Neural
Transmission, 67, 287-292.
Birkmayer, W., & Birkmayer, J.D. (1987). Dopamine action and
disorders of neurotransmitter balance. Gerontology 33. 168-
71.
Blatt, B., & Tsushima, W.T. (1966). A psychological survey of
uremic patients being considered for the chronic
hemodialysis program. Intellectual and emotional problems
in uremic patients. Nephrology. 3., 206-208.
Brenner, B.M., & Lazarus, J.M. (1980). Chronic Renal Failure
In K. Isselbacher, R. Adams, E. Braunwald, R. Petersdorf, &
J. Wilson (Eds.), Harrison/s Principles of Internal
Medicine (9th ed.). (pp. 1299-1307). New York, New York:
Mcgraw-Hill.
Bunn, H.F. (1980a). Anemia associated with chronic systemic
disorders. In K. Isselbacher, R. Adams, E. Braunwald, R.
Petersdorf, & J. Wilson (Eds.), Harrisons Principles of
Internal Medicine (9th ed.). (pp. 1530-1532). New York,
New York: Mcgraw-Hill.
Bunn, H. F. (1980b). Pallor and Anemia. In K. Isselbacher,
R. Adams, E. Braunwald, R. Petersdorf, & J. Wilson
(Eds.), Harrisons Principles of Internal Medicine (9th
ed.). (pp. 262-272). New York, New York: Mcgraw-Hill.
Canadian Erythropoietin Study Group (1989). The effect of
recombinant erythropoietin (EPO) upon quality of life
(QL) and functional capacity (FC) of anemic patients on
chronic hemodialysis. Kidney Internatinal. 35, 195.
Abstract.
Carroll, J.B., Davies, P., & Richman, B. (1971). The
American Heritage Word Freguencv Book. New York, New
York: American Heritage Publishing Company.
Cermak, L.S. (1982). The Long and Short of It in Amnesia. In
L.S. Cermak (Ed.), Human Memory and Amnesia (pp. 43-59).
Hillsdale, New Jersey: Lawrence Erlbaum Associates.
Chandra, M., McVicar, M., & Clemons, G. K. (1988).
Pathogenesis of the anemia of chronic renal failure: the
role of erythropoietin. Adv. Pediatr [200] 5, 361-389.
Cohen, R.M., Weingartner, H., Smallberg, S.A., Pickar, D., &
Murphy, D.L. (1982). Archives of General Psychiatry. 39.
593-597.


78
Table 4-3 continued
Id
Assess 1
Assess 2
Assess 3
109
Moderate
Moderate
Moderate
110
Moderate
Moderate
Moderate
111
Moderate
Moderate
Moderate
112
Moderate
Moderate
Moderate
113
Moderate
Moderate
Moderate
114
Moderate
Moderate
Moderate
115
Moderate
Severe
Very Severe
116
Moderate
Moderate
Moderate
117
Severe
Moderate
+
118
Moderate
Severe
Severe
Note:
Valence sign = direction of
Id < 100 = treatment group;
categorical change;
Id > 100 = controls
As displayed in Table 4-4, subjects were classified
based on ferritin level (under 10 Deficient, 10 30 Low,
over 30 Replete). Classifying subjects into Replete, Low
and Deficient categories revealed little variability in iron
status across assessments.
Table 4-4. Number of Subjects in each Category of Blood
Iron Level
Iron Assess 1 Assess 2 Assess 3
Status Con rEPO Con rEPO Con rEPO
Replete 14 11 14 9 12 11
Low 14 16 13
Deficient 01 00 01


140
Table C-2. Free Recall Instructions
"Remember the list of [animals, grains & flour products or
means of transportation] that I read earlier. Please tell
me all the words from that list that you can remember, in
any order."
Table C-3. Frequency Estimation
"When I read you that list some of the words were said more
than once. I will now read you words from that list. When
I say a word tell how many times you think it appeared the
first time I read it.
Guess when
not
sure."
Assessment 1
ANIMALS
Assessment 2
GRAIN & FLOUR
PRODUCTS
Assessment 3
MEANS OF
TRANSPORTATION
1
kangaroo
cake
escalator
2
snake
oatmeal
canoe
3
dog
wheat
bicycle
4
elephant
rice
stagecoach
5
buffalo
corn
airplane
6
goldfish
waffle
car
7
tiger
pancake
subway
8
moose
barley
skateboard
9
gopher
bagel
helicopter


12
to benefit from semantic processing has been attributed to
encoding deficits secondary to organic dysfunction of memory
structures (e.g., Korsakoff's).
Automatic and Effortful Memory
Automatic memory is demonstrated by recall in the
absence of effortful encoding. Automatic memory is
unconscious and little affected by degree of effort (Jonides
& Naveh-Benjamin, 1987). Deficits in automatic memory
processes are suggestive of dysfunction at the level of
organic memory structures. Reductions in capacity such as
illness and, presumably, fatigue, have limited effect on
automatic memory (Hasher & Zacks, 1984). The frequency of
occurrence task provides a means of assessing automatic
memory function. Normal performance on a frequency of
occurrence task in conjunction with reduced performance on
an effortful task requiring effort for optimal performance
would suggest disruption at the level of effortful
processing. This pattern of deficits could be the result of
frontal lobe dysfunction, poor motivation or failure to
employ an effective strategy. One caveat, the use of
automatic memory measures to assess the integrity of the
neural substrates of effortful memory is predicated on the
assumption, not accepted by all, that effortful and
automatic memory are largely dependent on identical neuronal
systems (Roediger, 1990).


viii
study. Also administered were a survey of cognitive-
affective and physical behavior, a verbal fluency task and