<%BANNER%>

The Effects of prenatal cocaine exposure on attention and reading: a longitudinal study

University of Florida Institutional Repository

PAGE 1

THE EFFECTS OF PRENATAL COCA INE EXPOSURE ON ATTENTION AND READING: A LONGITUDINAL STUDY By TAMARA DUCKWORTH WARNER A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2003

PAGE 2

Copyright 2003 by Tamara Duckworth Warner

PAGE 3

This dissertation is dedicated to the memory of my father, James V. Duckworth, Jr., who taught me to always “dream big" and to my mother, Mary A. Duckworth, who always encouraged to tackle any task that I thought I was "big enough" to handle.

PAGE 4

iv ACKNOWLEDGMENTS I would like to acknowledge the assi stance and patience of my committee members: Eileen B. Fennell, Ph.D, Chair; Duane E. Dede, Ph.D., Co-chair; Fonda Davis Eyler, Ph.D., Kenneth Heilman, M.D., Christia na Leonard, Ph.D.; and Michael Marsiske, Ph.D. I would like to thank Dr. Eyler and her co-principal investig ator, Marylou Behnke, M.D. for allowing me access to their data and for their generous mentorship. Thanks are also due to the entire Project Care staff, pa rticularly Project Direct or Kathie Wobie, Ann Welch, Eric Corpus, and Weir H ou for countless hours of help. My sincere gratitude also goes to the “family” of the Florida Education Fund’s McKnight Doctoral Fellowship Program, who supported my gr aduate studies financially, emotionally, and spiritually. Finally, I want to express my deepest appreciation to my husband, Kenneth D. Warner, for his steadfast love, support, and encouragement without which I would not have been to endure th e sometimes agonizing process of completing a dissertation. Thanks and praise go to God, who constantly su stains me and has "made a way out of no way" many more times than I know.

PAGE 5

v TABLE OF CONTENTS page ACKNOWLEDGMENTS..................................................................................................iv LIST OF TABLES............................................................................................................vii LIST OF FIGURES..........................................................................................................viii ABSTRACT....................................................................................................................... ix CHAPTER 1 INTRODUCTION......................................................................................................1 Overview of Attention...............................................................................................1 Development of Visual Attenti on in Infants and Children........................................6 Measurement of Visual Attenti on in Infants and Children......................................12 Prenatal Cocaine Exposure and Attention...............................................................14 Relationship between Attention and Reading.........................................................17 Study Purpose and Hypotheses................................................................................22 2 METHODS...............................................................................................................23 Participants..............................................................................................................23 Measures..................................................................................................................25 Demographic Variables..................................................................................26 Measures of Cognitive Development.............................................................28 Attention Measures........................................................................................28 Verbal Ability................................................................................................33 Reading Ability..............................................................................................35 Caregiving Environm ent Measures................................................................36 Procedure.................................................................................................................39 Hypotheses...............................................................................................................45 Data Inspection and Analyses..................................................................................46 Data Screening...............................................................................................46 Missing Data..................................................................................................46 Accounting for the Participants......................................................................49 Statistical Analyses........................................................................................50 3 RESULTS.................................................................................................................61

PAGE 6

vi 4 DISCUSSION..........................................................................................................73 Study Summary.......................................................................................................73 Main Findings................................................................................................73 Ancillary Findings..........................................................................................74 Study Findings in the Context of the Literature......................................................76 Study Strengths........................................................................................................83 Study Limitations....................................................................................................86 Future Directions.....................................................................................................87 REFERENCES..................................................................................................................90 BIOGRAPHICAL SKETCH...........................................................................................101

PAGE 7

vii LIST OF TABLES Table page 2-1 Continuous Variables that Differed Si gnificantly Between Mothers in the Two Original Study Groups..........................................................................................54 2-2 Non-continuous Variables that Differe d Significantly Between Mothers in the Two Original Study Groups..................................................................................54 2-3 Variables that Differed Significantly Between Neonates in the Two Original Study Groups.........................................................................................................55 2-4 Summary of Variables for Current Study............................................................55 2-5 Significant Differences Between Partic ipants With and Without TMT Part A Scores....................................................................................................................57 2-6 Demographic Variables Compar ing Groups in Current Study.............................58 3-1 Intercorrelations Between All Atte ntion Measures for Combined Sample..........67 3-2 Group Means and Standard Deviatio ns for Early Childhood Attention and Reading Variables.................................................................................................68 3-3 Group Means and Standard Deviatio ns for All Other Study Variables................69 3-4 Fit Indices for Nested Sequence of Measurement and Structural Models............70

PAGE 8

viii LIST OF FIGURES Figure page 1-1 Theoretical model of the effects of prenatal cocaine exposure on child behavior.................................................................................................................17 2-1 Proposed structural equati on model with factors and...........................................59 2-2 Proposed structural equation model with factor names........................................60 3-1 Final structural model............................................................................................71 3-2 Final structural model showing the variances and residuals of the observed variables with their respective factors...................................................................72

PAGE 9

ix Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE EFFECTS OF PRENATAL COCA INE EXPOSURE ON ATTENTION AND READING: A LONGITUDINAL STUDY By Tamara Duckworth Warner August 2003 Chair: Eileen B. Fennell Cochair: Duane E. Dede Major Department: Clinical and Health Psychology Animal studies and knowledge about the pharmacology of cocaine strongly suggest that maternal cocaine use during pregnancy has negative consequences for fetal development. However, significant neurobeha vioral differences between infants and children with prenatal cocaine exposure (PCE) and well-matched comparison groups have not been found consistently. Attention is one area in which group differences have been reported, but the functi onal implications of these findings for reading or other school-related abilities remain uncertain. A group of prospectively enrolled child ren who were prenatally exposed to cocaine, alcohol, tobacco, and marijuana ( n = 120) were compared to a matched group of children who were exposed to alcohol, tobacco, and marijuana but not cocaine ( n = 120). Both groups were predominantly poor and Af rican American and did not differ by sex.

PAGE 10

x Significant group differences were higher le vels of prenatal drug exposure, higher prenatal risk scores, shorter gestational ages, and smaller head circumferences at birth for the children with PCE. Results indicated that neonatal attenti on as measured by three scales of the Brazelton Neonatal Behavioral Assessment Scal e was not significantly associated with attentional measures administered at ages 5 and 7, including a continuous performance test, short term auditory atte ntion (Digit Span), or three ta sks involving visual scanning and visuomotor coordination (Letter Cancellation, Trail Ma king Test Part A, and the Coding subtest of the Wechsler Intelligence Scale for Children-III). Analyses also revealed that children with PCE did not perform significantly wo rse than non-exposed children on attention measures at ages 5 or 7 or on the Wechsler Individual Achievement Test reading subtests at age 7. Structural equation modeli ng, however, demonstrated that PCE had an indirect effect on reading ability at age 7 that was mediated by head circumference at birth. The effect of PCE on birth head circumference wa s similar to that of prenatal exposure to alcohol or marijuana. Birth head circumfere nce affected age 5 verb al ability which, in turn, was related to age 7 verbal ability and vi sual attention which, in turn, affected age 7 reading ability. The final model accounted fo r 68% of the variance in age 7 reading ability for the combined sample.

PAGE 11

1 CHAPTER 1 INTRODUCTION Overview of Attention This manuscript begins with a brief review of definitions and models of attention, the neuroanatomical correlates of attention, attentional development over the course of infancy and early childhood, and measurement issu es related to attent ion. In attempting to define attention, many writers begin with William James’ now-famous observation that Everyone knows what attention is. It is the taking possession by the mind, in clear and vivid form, of one out of what seem to be several simultaneously possible objects or trains of thought Focalization, concentration, and consciousness are of its essence. (James, 1890, pp. 381-382) Enns (1990) has observed that many definitions of attention share a component of selectivity—that a primary function of attention is to select information for further processing. Similarly, Cohen ( 1993) argues that attention se rves as a gatekeeper in facilitating “the selection of salient information and the al location of cognitive processing appropriate to that information” much like the aperture and lens system of a camera (Cohen, 1993, p. 3). It is generally agreed that attention is not a unitary process, but a multifactorial process, involving a dive rse set of behavioral phenomena Based on this general understanding of attention, several researchers have attempted to construct models of the various components of attention. An empirically validated model of attention that integrat es research from cognitive psychology and neuropsychology has been articulated by Mi rsky, Anthony, Duncan, Ahearn and Kellam (1991). Mirsky et al. (1991) id entified three “elements” of attention—focus, sustain, and

PAGE 12

2 shift. The focus element involves selecting information for enhanced processing, which has been equated with atten tion itself. The focus element also includes an “execute” component, in which motor programs associat ed with focusing attention are activated. The sustain component is synonymous with vigila nce or the ability to maintain focus and alertness over time. The third element, shift, represents the ability to change attentive focus in a flexible and adaptive manner. Neuroanatomically, the three elements of attention in the Mirsky et al. model can be localized to various areas of the brai n based on lesion studies of brain-impaired patients, animal lesion studies, and neuroima ging data (Mirsky et al., 1991). The focusexecute element is associated with the inferior parietal and superior temporal cortex and striatum of the basal ganglia. The parietal ar ea is the primary cerebral locus of the focusexecute aspect of a ttention based on studies of adult neglect patients (Heilman, Watson, & Valenstein, 1993) as well as th e connectivity of the parietal lobe with sensory, motor, limbic, thalamic and brain stem regions of the brain. It is worth noting, however, that unlike adults, children rarely show the full a dult neglect syndrome even with parietal lesions (Ferro, Martins, & Tavora, 1984). The proposed role of superior temporal sulcus in focused attention in Mirsky’s model is supported by studies of the architecture and connectivity of this area as well as studies of epilepsy patients who have undergone anterior temporal lobectomy although there ar e some conflicting data on this latter point (Mirsky et al., 1991). Inclusion of striatum in the anatomy underlying focused attention is based on its modulatory role in motor systems and the role of the caudate in delayed alternation and delayed respons e tasks (Mirsky et al., 1991).

PAGE 13

3 The sustain element of attention is su bserved by rostral midbrain structures, including the mesopontine reticular formation and midline and reticu lar thalamic nuclei. Regulation of arousal and main tenance of consciousness are the primary functions of the brain stem structures. Single unit record ing studies in monkeys trained on a go/no-go visual attention task showed increased firi ng of Type II cells in the midline thalamus, superior colliculus, tectum, pons, and mesen cephalic brain stem, suggesting a role in maintaining attention. In addition, studies invo lving stimulation of reticular nucleus of the thalamus have demonstrated that this thalamic nucleus modifies the influence of reticular formation effects on vi sual signals, specifically one ’s “readiness to respond” to visual stimuli in discrimination paradigms. Neuroanatomically, the shift component of attention in Mirsky’s model is the responsibility of the prefrontal cortex and also perhaps the medial frontal cortex and anterior cingulate gyrus (Mirs ky et al., 1991). The role of the prefrontal cortex role in shifting attention is suppor ted by studies of poor performance on the Wisconsin Card Sorting Test (WCST) by epilepsy patien ts who have undergone resection of the dorsolateral prefrontal cort ex. Additional evidence comes from individuals with schizophrenia who show impaired performan ce on the WCST and reduced activation of the prefrontal areas in imaging studies. Prelim inary inclusion of the medial frontal cortex and anterior cingulate gyrus in attentional shift is based on th e firing patterns of Type II cells in these regions in monkeys during go/ no-go visual discrimination tasks that are thought to measure both shifti ng attention as well as sust ained attention, as discussed earlier (Mirksy et al., 1991).

PAGE 14

4 Additional evidence supporting Mirksy’s model of the neural substrates underlying different aspects of attention come s from neuroimaging st udies of individuals with Attention Deficit/Hyperactivity Disord er (ADHD) and schizophr enia, two clinical disorders with significant at tentional dysfunction. ADHD is a behavioral disorder characterized by hyperactivity, im pulsivity, and inattentiveness. In a recent review of structural and functional ne uroimaging studies of childre n and adolescents with ADHD, Hale, Hariri, and McCracken ( 2000) concluded that three sets of findings are emerging with some consistency: 1) reduced prefront al and caudate volumes, 2) hypoperfusion and hypometabolism in prefrontal and striatal regions, and 3) lower levels of activation in the anterior cingulate during tasks involving s timulus selection and/or response inhibition. That the frontal areas involved in shifting at tention and the striatal areas involved in sustained attention have been implicated in ADHD provides some support for Mirsky’s model of attention. Functional neuroimaging studies on the atte ntional deficits found in those with schizophrenia are another source of support fo r the neuroanatomical outline for Mirsky’s model of attention. In Taylor’s (1996) re view of 24 functional neuroimaging studies using adults, almost half of the studies (11) found that patients with schizophrenia failed to show task-related increases in blood flow in the prefront al area, using the WCST and similar paradigms. Studies of childre n and adolescents with childhood-onset schizophrenia also show reduction of fr ontal metabolism (Hendren, DeBacker, & Pandina, 2000). Thus, the neural mechanisms th at may underlie the atte ntional deficits in schizophrenia for adults appear to be same for children and adolescents. Again, both sets of findings support Mirsky’s loca lization of shifting attention to the prefrontal area.

PAGE 15

5 One serious limitation of Mirsky’s model is its underspecificati on of the role of neurotransmitter systems in attention. In c ontrast, Posner and Petersen’s (1990) model of attention places strong emphasis on norepinephr ine (NE) arising from the locus ceruleus in the rostral brain stem and extending to th e posterior attention system, particularly in the right hemisphere. Norepinephrine is thought to play a crucial role in maintaining an alert state based on animal studies and the e ffects that drugs that manipulate NE levels have on the ability to shift attention (Posner & Petersen, 1990). Posner and Petersen’s model is consistent with current theories th at suggest that the co re deficits in ADHD are in executive functions that re gulate arousal, attention, a nd inhibition and that these deficits may be related etiologically to a bnormalities in dopaminergic and noradrenergic pathways from the brainstem that serve to regulate cortico-striato-thalamo-cortical networks (Hale, Hariri, & McCracken, 2000). It should be noted, however, that pathways involving dopamine and norepinephrine are f ound throughout the brain, and this may be the reason that Mirsky and his colleagues have not included neurotransmitter components in their model of the neuroanato mical substrates of attention. Despite this major limitation, the mode l of attention offered by Mirsky and colleagues has been confirmed empirically us ing principal component s analysis in an epidemiological sample of el ementary school children ( N = 435) and in a sample of normal and neuropsychiatrical ly impaired adults ( N = 203). Importantly, the focusexecute, sustain, and shift fact ors (in addition to an encode factor) were found in both samples, suggesting that attention operates according to similar processes in both children and adults. In both samples, the focus-ex ecute factor was identified by significant loadings from the Digit Symbol Substitution subtest of the Wechsler Adult Intelligence

PAGE 16

6 Scale-Revised (WAIS-R, W echsler, 1981) or Coding subt est from the Wechsler Intelligence Scale for Children-Revised (W ISC-R, Wechsler, 1980) and the Talland Letter Cancellation Test (Talla nd, 1965). The sustain factor was identified by significant loadings from a continuous performance test (CPT), specifically mean number of hits and mean reaction time. Finally, the shift factor wa s identified by the Wisconsin Card Sorting Test (Grant & Berg, 1993) percentage corr ect and number of categories achieved. The adult sample was also administered the Trail Making Test (Reitan & Wolfson, 1985), which had significant primary loadings on the fo cus-execute factor as well as significant secondary loadings on the sustai n factor (Mirsky et al., 1991). Overall, Mirsky’s three factor model of attention, which has been confirmed empirically and has at least partially-specified neuroanatomical substrates for each factor, provides a solid foundation for studying at tention in both children and adults. Development of Visual Attention in Infants and Children This review of the development of a ttention from the neonatal period through early to middle childhood will focus on studies of visual attention. Prior to reviewing the two major conclusions that can be drawn from the developmental literature on attention, a brief review of the measurement of attention in infants is warrante d. Infant attention is generally studied using paradigms that rely on the visual modality and is measured clinically using motor orienti ng responses such as head turn ing and attempts at visual fixation to a stimulus. In the laboratory setting, psychophysiological measures, such as changes in heart rate and evoked brain potenti als, have proven to be useful tools in measuring attention in newborns and young infants who have poor motoric control (Graham, 1992). Visual fixation on a stimulus, ha bituation to the stimulus in the form of

PAGE 17

7 decreasing fixation times after continuous or re peated presentations, and reorientation to a novel stimulus are al so common measures. Generally, two major conclusions can be made from the literature on the development of visual attention in children: children’s attentional capacities improve with age (Cooley & Morris, 1990) a nd some continuity exists between attentional abilities measured at infancy and those measured at later ages (Colombo, 1993; Enns, 1990). Based on their review of the experimental literature, Cooley a nd Morris (1990) suggest that the improvement in child ren’s attentional abilities ove r time can be explained using two different theoretical frameworks. The firs t is an information processing framework that states that the limited attentional capacity of younger ch ildren increases as internal processing mechanisms develop with age. In the case of selective at tention, the filtering mechanism of attention becomes more effici ent and larger proporti ons of attentional resources can be allocated and allocated more fl exibly to relevant rather than irrelevant aspects of a task. Similarly, in the case of sustained atte ntion, a limited capacity model would predict that more effort is required of younger subjects than older subjects in order to perform well (Cooley & Morris, 1990). The second theoretical framework used to explain children’s improved attentional abilities over time is an ecol ogical or perceptual learning framework. According to this framework, children learn with age to become more specific, systematic, economical, and task-directed in their perception and explora tion. In doing so, childre n become better able to differentiate rele vant from irrelevant information needed for a task. Thus, the perceptual learning framework tends to empha size improvements over time in the quality of children’s selective attention abilities (Cooley & Morris, 1990).

PAGE 18

8 Generally, the literature suggests that chil dren begin to develop control over their visual attentional resources between 7 and 13 years of age, but the development of efficient use of strategic methods for allo cating attention conti nues through adulthood. Kaye and Ruskin (1990) conducted a series of three studies to i nvestigate the relative roles of increased information capacity pr ocessing versus strategic allocation of attentional resources in children’s imp roved performance on various attentional measures. Using a paradigm requiring shifting attention to a peripheral visual cue, the researchers found that age differences am ong 3rd graders, 6th graders and college students were only found for general alertness, a non-strategic capacity-related factor. In a divided attention task using two stimulus probability conditions, it was found that both younger and older children demonstrated adultlike strategic proces sing, but there were age differences in their efficient use of these strategies. The third study used a classification task and children at three diffe rent age levels ranging from 5 to 12 years. No qualitative differences in strategy use were found; however, quantitative differences in search rates were found, which in turn a ffected classificatio n. Based on these three studies, the authors concluded th at while adult-like strategies emerge early in childhood, their optimal use in complex tasks is depende nt on the increased information processing efficiency that comes with age (Kaye & Ruskin, 1990). The question of continuity in mental abilities generally, and attention in particular, has been studied in longitudinal samples. The two major paradigms used in these studies are habituation and response to novelty. Habituation is measured as either the rate of decline in looking behavior over repeated presentations of a stimulus or the fixation duration. Response to novelty is meas ured either by looking behavior to a new

PAGE 19

9 stimulus simultaneously paired with an alr eady-familiar stimulus ( novelty preference) or as recovery of fixation to a new stimulus after habituation (recove ry). Bornstein (1990) reviewed eight longitudinal studies that examin ed the predictive vali dity of habituation paradigms administered duri ng the first six months of life for later cognitive performance. The median predictive correlati on of habituation paradigms using different sensory modalities was .49, with a range from .28 to .63. Three-quart ers of these studies focused on predicting IQ, two used language meas ures and one used the Bayley Scales of Infant Development (Bayley, 1969). Notably, thes e studies of infant attention used both normal and at-risk infant samples with the age at the second asse ssment ranging from 2 years to 8.5 years. A similar review by Colo mbo (1993) of 13 studies that specifically used fixation duration to pr edict a variety of outcome measures found correlations ranging from .29 to .77. Colombo (1993) has also reviewed longitudi nal studies of predic tive validity of response to novelty measured in the first year of life with later c ognitive measures. Based on the four studies that used recovery after habituation as their measure of response to novelty, few conclusions can be drawn, in pa rt because recovery measures have poor reliability and stability (Colombo, 1993). In co ntrast, the predictive validity of novelty preference measures with late r intellectual and c ognitive function is more robust. The overall median correlation, based on 14 studies of novelty preference, is .47, with a range from .25 to .66; when only standardized IQ sc ores are used as the outcome measure, the median correlation increases slightly to .49 (Colombo, 1993). The majority of these studies used three IQ outcome measures : the Stanford-Binet (Thorndike, Hagen, & Sattler, 1986), the age-appropr iate Weschler scale, or the Peabody Picture Vocabulary

PAGE 20

10 Test-Revised (Dunn, Dunn, Robertson, & Eisenb erg, 1981). Four studies used the Bayley Scales of Infant Development (Bayley, 1969), four used measures of language development and two examined memory. The ag e at first assessment ranged from 3 to 9 months, while the age at second assessment ranged from 1 to 7 years of age (Colombo, 1993). Overall, the reviews by Bornstein (1990) and Colombo (1993) suggest a moderate degree of predictive valid ity between experimental m easures of attention and standardized cognitive assessmen ts in early to middle childhood. A handful of longitudinal studies have begun to examine the development of particular aspects of atten tion over time and how these components relate to other cognitive functions. In pre-term infants it ha s been found that fixation duration measured at 40 weeks conceptual age was predictive of performance on a focused attention task but not a sustained attention task administer ed at age 12 (Sigman, Cohen, Beckwith, Asarnow, & Parmelee, 1991). Notably, the fo cused attention measure was a signal detection task that also required speede d information processing. The correlations between the infant attentional measures and later attentional measur es were -.36 and -.32, for the less difficult and more difficult version of the task, respectively. To explain their findings, the authors argue that both infant attention meas ures and childhood cognitive assessment may tap the efficiency of informati on processing. If this is the case, then an individual's relative capacity to process information efficientl y would be expected to be stable from infancy to childhood. Furthermore, infant information processing abilities would not be expected to be associated with other measures of childhood attentional ability such as sustained a ttention (Sigman et al., 1991).

PAGE 21

11 In another longitudinal cohort, Rose a nd Feldman (1995) found that an indirect measure of sustained attention (exposure ti me to meet criterion during a visual recognition memory task) acquired at 7 months of age in a sample of pre-term and fullterm infants was significantly correlated with perceptual speed at 11 years of age. When IQ was partialed out, the correlations between the infant sustained attention measure and the perceptual speed measures were .33 for a visual matching task, .34 for a visual search task and .38 for the two measures combined (Rose & Feldman, 1995). The significant correlations between sust ained attention and perceptual speed in the Rose and Feldman (1995) study seem to support increased capacity and increased efficiency theories posited by information pr ocessing accounts. The authors also suggest, however, a second possible explanation for th e significant correlati on between sustained attention and perceptual speed: that the speed with which the child is able to respond is a function of his or her ability to filter out stimuli irrelevant to the task (Rose & Feldman, 1995). In this way, the sustained attention me asure is also a meas ure of appropriately focused attention. Unfortunately, the rese archers did not include any attentional assessments in their longitudinal battery so th e relationship between the infant attentional measure cannot be compared to la ter childhood attentional measures. One major limitation of the experiment al literature on the development of children’s attention, as w ith the models of atten tion reviewed earlier, is its neglect of the development of underlying brain systems. Wh ile neuropsychological studies of childhood attention, which tend to focus on deficits found in clinical po pulations, stress the importance of the developing neural substrat es, they rarely elucidate specific brainbehavior relationships. More important perhaps is the fact that few studies in the

PAGE 22

12 developmental neuropsychology literature ha ve attempted to relate children’s development of various components of attentio n—e.g., focus, shift, and sustain—to brain maturation (Cooley & Morris, 1990). Measurement of Visual Attention in Infants and Children As briefly reviewed in the previous section, a number of different experimental paradigms exist for measuring visual attent ion in infants. The number of clinical measures available for assessing attention in th e first few years of life is not as plentiful. As a result, few developmental studies are av ailable that explore the growth and change of attention using clinical measures. Ar guably, the most commonly used assessment instrument for the study of infant atten tion is the Brazelton Neonatal Behavioral Assessment Scale (BNBAS). First develope d in 1973, the BNBAS is designed to assess the newborn's adaptive responses to his or her new extrauterine environment and has been used for over 25 years in hundreds of re search and clinical setting across the world (Nugent & Brazelton, 2000). Theoretically, de velopment of the BNBAS is based on the assumption that early human developmen t proceeds from a state of relative undifferentiation to one of increasing differentiation, ar ticulation and hierarchic integration. More specificall y, there are hypothesized four primary developmental tasks of the newborn, arranged hierarchically and re lated to increasing sel f-regulation. The first developmental task is organization of au tonomic or physiological behavior, which involves homeostatic adjustment of the centr al nervous system including respiration, the startle response, tremors and temperature regulation. The second development task is control and regulation of motor behavior including inhibiti on of random motor responses, development of better muscle tone and reduction of exce ssive motor activity. Third, the neonate must develop state regula tion, the ability to modulate states of

PAGE 23

13 consciousness including sleep, a nd deal with stress using st rategies such as hand-tomouth movements, communicating with the ca regiver through crying and being consoled with the aid of the caregiver. The fourth and final task for the infant is regulation of affective interactive or social behavior which involves maintaining prolonged alert periods, attending to visual and auditory stim uli within one's range and seeking out and engaging in social interaction with the caregiver (Nugent & Brazelton, 2000). The development of the first three self-re gulatory capacities can be seen as laying the foundation for the fourth, which is the rudimentary development of attention to external stimuli. Following regulation of internal stimuli, the neonate can turn to maintaining alertness or arousal, which is a prerequisite for directing, focusing and sustaining attention. With repeated assessm ents, the BNBAS describe this emerging process during the first couple of months of the newborn' s life in the extrauterine environment. A more detailed description of the scale and its prope rties is found in the Methods chapter. After the neonatal period, sta ndardized measurements of infant attention are more difficult to come by. This is due, in part, to a variety of meth odological challenges associated with developmental research, not the least of which is the problem of measurement equivalence (Hartmann & Geor ge, 1999). Measurement equivalence refers to the question of whether an assessment in strument measures the same construct at various developmental ages. For example, an "intelligence" test may be more a measure of verbal comprehension than reasoning at age 3 as compared to age 10. In addition, children's rapid changes in phys ical and behavioral matura tion and experience can make transient performance levels difficult to captu re (Hartmann & George, 1999). A variety of

PAGE 24

14 design and data analytic techniques can be used to help guard against and test for measurement equivalence, but it is a thorny is sue ever-present in developmental research and very difficult to overcome in infancy and childhood investigations. Prenatal Cocaine Exposure and Attention With the alarming rise in the use of cocaine by pregnant women since the mid1980s, children who have been prenatally e xposed to cocaine have emerged as a new clinical population in which to study the deve lopment of attention. That prenatal cocaine exposure has a teratogenic effect on developm ent has generally been established. Cocaine has the potential to exert deleterious effect s in variety of direct and indirect ways. Cocaine easily crosses both the placenta a nd the blood-brain barrier (Mayes, 1994). Most of what is known about the pharmacologic a nd other effects of cocaine on prenatal development has been learned through resear ch on a variety of animal populations, including rats, mice, rabbits, and monkeys (Kosofsky & Wilkins, 1998; Lidow, 1998; Spear, 1995). There is a general consensus that animal models (despite differences in route of administration and other methodol ogical problems) are useful analogues of prenatal cocaine exposure in humans (Need lman, Frank, Augustyn & Zuckerman, 1995). Three well-studied pharmacological effects of cocaine will be reviewed here. First, cocaine inhibits the reupta ke of dopamine, serotonin and norepinephrine, thereby potentiating their actions (Akbari, Kram er, Whitaker-Azmitia, Spear & Azmitia, 1992; Dow-Edwards, 1995; Factor, Hart & Jonaka it, 1993; Friedman & Wang, 1998; Leslie, Robertson, Jung, Libermann & Bennett, 1994; Mayes, 1994; Mactutus, Herman, & Booze, 1994; Ronnekliev, Fang, Choi & Chai 1998; Vorhees, 1995). Second, cocaine causes vasconstriction and can reduce blood supply and oxygenation to a developing fetus, resulting in chronic or intermitte nt hypoxia (Dow-Edwards, 1995; Woods, 1996).

PAGE 25

15 Third, cocaine acts as a potent local anesth etic, blocking sodium channels, thereby attenuating action poten tials in excitable ce lls (Dow-Edwards, 1995). In addition to its pharmacological effect s, maternal use of cocaine can cause hypertension, placental aburption, spontaneous abortion, poor pregnancy weight gain and undernutrition secondary to appetite loss (Chur ch, Crossland, Holmes, Overbeck & Tilak, 1998). Brain abnormalities reported in animal models of prenatal cocaine exposure include a reduction in number of cortical cells, inappropriate positioning of cortical neurons, altered glial morphology, reduction in length of neurites, and apoptotic neural cell loss (Lidow, 1995, 1998; Nassogne, Evrard & Courtnoy, 1998). These changes are attributed, in part, to alternations in func tion of monoaminergic neurotransmitters that affect synaptogenesis, neural growth, and cell proliferation. However, unlike early media reports of severe developmental consequences associated with cocaine use during pregnancy, research in humans ha s demonstrated that the sequelae of prenatal cocaine exposure ar e subtle but meaningful (Harvey & Kosofsky, 1998; Lester, LaGasse, & Seifer, 1998; Ne uspiel, 1994; Vorhees, 1995). In terms of pregnancy outcome, a few consistent findi ngs have emerged in studies with nondrug using control groups: higher risk for spont aneous abortion, shorter gestational age, smaller head circumference, shorter birth lengt h and lower birth weight (Lester, Freier & LaGasse, 1995; Lutiger, Graham, Einarson, & Ko ren, 1991). In terms of neurobehavioral outcomes, no overall syndrome has been found in infants prenatally exposed to cocaine, in part due to methodologi cal problems (Frank, Augustyn & Zuckerman, 1998; Lester, LaGasse & Bigsby, 1998). A review of ten studi es using the BNBAS, one of the most consistent measures used, suggests that pr oblems with state regulation occurs most

PAGE 26

16 frequently (Frank, Augustyn & Zuckerman, 1998). State regulation, according to the theory underlying the development of the BNBAS, is the third developmental task of the first two months of life, which must be su ccessfully negotiated befo re attention can be directed to stimuli in the external envi ronment (Brazelton, 1984, 1994; Brazelton, Nugent & Lester, 1987; Nugent & Brazelton, 2000). In terms of prenatal cocaine exposure's specific effect on attention, a number of animal models are suggestive. In a mous e model using Pavlovian conditioning and a blocking paradigm in which redundant informa tion must be ignored, learning deficits and behavioral alterations sugges tive of problems in selective attention were found (Kosofsky & Wilkins, 1998). Similarly, a primate model using rhesus monkeys also demonstrated problems in the acquisition of operant behavi ors in cocaine-exposed offspring (Morris, Gillam, Allen & Paule, 1996). Impairments in attention and discriminative learning have been demonstrated in rabbits using a foot shock paradigm (Gabriel & Taylor, 1998). In humans, a handful of studies have begun to emerge indicating that children prenatally exposed to cocaine begin to s how problems with attention and that these problems become more evident after age 4 (Beckwith, Crawford, Moore & Howard, 1995; Chasnoff, Anson, Hatcher, Stenson, Iaukea & Randolph, 1998; Leech, Richardson, Goldschmidt & Day, 1999; Mayes, Grillon, Gr anger & Schottenfeld, 1998). A model for understanding the effects of prenatal cocaine exposure on child behavior, including attention has been presented by Lester, Freier and LaGasse (1995). As shown in Figure 1, cocaine is thought to affect neuroregulatory mechanisms, which in turn result in

PAGE 27

17 disorders of behavioral regulation that manife st as the "Four A's of Infancy": attention, arousal, affect and action. Relationship between Attention and Reading Few theoretical models explicitly link the development of attentional abilities and the development of reading skills in childre n. Models of reading development generally focus on the processing of component skil ls in three areas: phonology, orthography, and semantics (Adams, 1994). A variety of inves tigations, including genetic studies, have examined the association between reading di sability and Attention Deficit/Hyperactivity Disorder (ADHD); (Brock & Knapp, 1996; Felton & Wood, 1989; Fergusson & Horwood, 1992; Gilger, Pennington & DeFries, 1992; Javorsky, 1996; Light, Pennington, Cocaine/Other Drugs Neuroregulatory Mechanisms Disorders of Behavioral Regulation Four A's of Infancy Attention Arousal Affect Action Neurodevelopmental Assessment Battery Figure 1-1. Theoretical model of the effects of prenatal coca ine exposure on child behavior taken from Lester, Frei er and LaGasse (1995).

PAGE 28

18 Gilger & DeFries, 1995; Nahri & Ahonen, 1995; Robins, 1992; Ve lting & Whitehurst, 1997; Willcutt & Pennington, 2000). The overw helming majority of these studies, however, have failed to distinguish s ubtypes with inattentive symptoms and hyperactivity. As Hinshaw (1992) pointed out in his review, while the link has been established between inattention and hype ractivity, on the one hand, and reading underachievement, on the other, causal models have rarely been tested using sufficient methodological controls. Only a small number of investigations have examined the relationship between attenti on and reading in children without diagnosed ADHD or reading disability (e.g., Velting and Whitehurst, 1997; Wood and Felton, 1994). Fortunately, two longitudinal investigations using large samples of normal children have begun to elucidate the relationship between at tention and reading (R abiner, Coie, & the Conduct Problems Prevention Research Group, 2000; Rowe & Rowe, 1992). Both of these longitudinal studies merit detailed review. A longitudinal investigation by Rabiner and colleagues (2000) found that early attention problems predicted reading achieveme nt even after contro lling for prior reading achievement, IQ, and other behavioral probl ems. A heterogeneous group of children ( N = 211) from four sites was fo llowed from kindergarten thr ough fifth grade. Attentional measures were collected from the children’s t eachers using the inattentive items from the Child Attention Problems Scale (Edelbrock, 1990) in kindergarten and the inattention scale of the ADHD Rating Scale (DuPaul, 1991) in grades 1 and 2. Reading achievement was measured in kindergarten and grade 1 us ing the Letter-Word Identification subtest from the Woodcock-Johnson Psychoeducatio nal Battery-Revised (WJ-R, Woodcock & Johnson, 1991); reading achievement in grade 5 was measured using both the Letter-

PAGE 29

19 Word Identification and Passage Comprehensio n subtests of the WJ-R. It was found that kindergarten reading was significantly correlated ( r = -.29) with kinde rgarten inattention after controlling for IQ and parental i nvolvement. First grade reading was also significantly correlated with first grade inattention ( r = -.29) after c ontrolling for IQ, parental involvement, and kindergarten re ading and inattention scores. Inattention accounted for 6% of the total variance in ki ndergarten and first grade reading scores, comparable to the amount of variance explained by IQ. Fifth-grade reading was significantly correlated ( r = -.10) with second grade inattention after controlling for all the previous variable s—IQ, parental involvement, kinde rgarten reading and inattention and first grade reading and inattention. A path analysis using multiple regression procedures found that the model with all of the variables ex plained 66% of the variance in the children’s fifth-gr ade reading achievement. More importantly, the researchers found that children who were highly inattentive first graders (standardardized scores >1.0) we re at greater risk for reading difficulties. Between kindergarten and first grade, the mean standardized reading achievement scores of the highly inattentive child ren declined significantly (-.52 to -.86), making them three times more likely than their peers to show a one standard deviation discrepancy criteria between IQ and reading ability. By fifth grade, the mean standardized reading score for the highly inattentive firs t graders remained substantially below the mean at .71. Another longitudinal study by Rowe and Rowe (1992) used structural equation modeling in a sample of 5,092 normal students ages 5 to 14 years to investigate the relationship between inattentiveness in th e classroom and reading achievement as mediated by family socioeconomic backgr ound factors, reading activity at home and

PAGE 30

20 attitudes towards reading. A st ratified sample of students was drawn from 256 classes in 64 public and 28 private elementary and s econdary schools from four regions (two metropolitan and two rural) in Victoria, Austra lia, reflecting 91% of the target sample. Study measures were collected at five tim e points: year levels 1, 3, 5, 7, and 9. The family socioeconomic indicators used were number of years of mo ther's education and father's education and mothers' and father's occupational classificati on as measured on an 8-point scale. Students' reading activity at home was obtained from self-report responses to four questions measured on a four-point Likert-type scale while students' attitude toward reading was determined using three question measured on a 5-point Likert-type scale. For students ages 5 to 6 years, inte rviews were conducted with classroom teachers to assess reading activity and attitudes. Classroom inattentiveness at all ages was obtained from teacher responses to four item s measured on a 5-point Likert-type scale. Assessment of reading achievement consiste d of two measures: an age-appropriate reading comprehension test and teacher rati ngs on a criterion-re ferenced profile of student reading behaviors (Rowe & Rowe, 1992 ). Data were analyzed using four age categories: 5 to 6 years, 7 to 8 years, 9 to 11 years and 12 to 14 years. Results indicated that across all age groups the measure of inattentiveness accounted for the largest proportion of vari ance, ranging from 13.4% to 22.9%. Attitudes towards reading and reading activity at hom e each explained between approximately 5% and 15%. The proportion of variance in reading achievement accounted by socioeconomic variables was very small, ranging from 0.3 to 3.2%. Analyses using a recursive structural equation model, in which all effects are unidirect ional, indicated that inattentiveness has strong negative influences on students' reading achievement as well as

PAGE 31

21 on the mediating variables of attitude towa rds reading and reading at home. Reading activity at home was found to have a significant influence on students' attitudes towards reading and acted as a strong mediator betw een inattentiveness a nd reading achievement that increased as students pr ogressed through school. Socio economic status was found to have little influence on any of the other f our factors in the model (Rowe & Rowe, 1992). A second set of analyses was conducted us ing a non-recursive structural equation model to examine interdependent effect s between inattentiveness and reading achievement. Goodness-of-fit indices were gr eater than .97 for all age groups, indicating that the model fit the data well. Reciprocal effects between inattentiveness and reading achievement were found to be significant and negative, and this relationship grew stronger over time. In sum, the two principal fi ndings were that: 1) inattentive behaviors in the classroom have a significant negative influence on students' reading achievement and 2) reading achievement, mediated by the direct influence of attitudes toward reading and reading activity at home, has a stronger effect on reducing inat tentive behaviors. Related to the second finding, the authors ar gue that low reading achievement leads to high inattentiveness (Rowe & Rowe, 1992). The results of these two longitudinal studies converge on the notion that significant relationships exist between children's attentional abilities and their reading ability. Both studies also sugge st that early interv ention for inattentive children may help to reduce the chances of late r reading problems. For child ren prenatally exposed to cocaine who may be at greater risk for attentional problems, early detection and intervention could have a significant impact on their academic and overall developmental

PAGE 32

22 outcomes and highlights the need for more research on this potentially vulnerable population. Study Purpose and Hypotheses The purpose of the current study is to inve stigate the developmental trajectory of attention in a sample of children prenatally e xposed to cocaine to determine: 1) whether: an indicator of attentional problems in infa ncy predicts poor atte ntion skill in early childhood and 2) whether attentional measures at birth and early childhood are related to reading ability after controlling for the in fluence of general verbal ability and the caregiving environment. It is hypothesized that after controlling for prenatal obste tric risk and exposure to alcohol, tobacco, and marijuana, the significa nt differences between children prenatally exposed to cocaine and matched controls on an early indicator of attention problems will persist at ages 5 and 7. More specifically, children prenata lly exposed to cocaine will show worse performance on measures of a ttention than their matched controls. In addition, the poor performance of the exposed children will have direct and indirect negative effects on reading achievement at age 7 after controlling fo r verbal ability and the caregiving environment.

PAGE 33

23 CHAPTER 2 METHODS Participants Participants in the current study were children enrolled in a prospective, longitudinal National Institute of Drug A buse-funded study (DA 05854) to examine the effects of prenatal cocaine exposure on de velopmental outcomes. The original study, entitled Project C.A.R.E. (Cocaine Abuse in the Rural Environmen t), is housed at the University of Florida’s Shands Teaching Hosp ital, and the study’s pr incipal investigators are Marylou Behnke, M.D. and Fonda Davis Eyler, Ph.D. In the original study, 154 cocaine-using pregnant wo men and 154 non-using matched controls were enrolled prospectively soon after they fi rst contacted the hea lth care system, eith er at a prenatal obstetric clinic or at the hospital. Deta iled information regard ing recruitment and enrollment of participants is provided in the Procedures section. Since enrollment in the longitudinal study, there has been a relatively low attrition rate (approximately 10%). All the children who participated in follow-up assessments at ages 5 and 7 and who have complete data for all the study measures were included in the present study. A variety of data is ava ilable on the original study sample based on previous studies (Behnke, Eyler, Conlon, Wobie, Woods, & Cumming, 1998; Behnke, Eyler, Woods, Wobie, & Conlon, 1997; Eyler, Behnke, Conlon, Woods, & Wobie, 1998a, 1998b; Eyler, Behnke, Garvan, Woods, Wobi e, & Conlon, 2001; Woods, Behnke, Eyler, Conlon & Wobie, 1995). Of the 154 cocaine us ers, 70% admitted to using crack, 16% used powder cocaine and 14% denied any co caine use but had a positive urine screen.

PAGE 34

24 Only 6% of the cocaine-using group of wome n received some form of drug treatment during pregnancy. The entire study sample was predominantly African American ( n = 125), lowest Hollingshead socioeconomic status (SES) category (Hollingshead, 1995, n = 118) and had more than one child. As show n in Tables 2-1, 2-2, and 2-3, significant differences on a number of variables were found between the cocaine -using mothers and their babies when compared to their matche d controls. First, the cocaine-using mothers were significantly older than the co mparison group (27.4 years vs. 22.8 years, p = .0001). However, there were no differences between the groups based on the number of women in the over 40 years age range, which has been associated with increased perinatal risk (Eyler et al., 1998a). Second, the cocain e-using mothers entered prenatal care significantly later than the non-using mothers. However, multiple regression analyses controlling for alcohol, tobacco, and mar ijuana use showed that only tobacco use significantly predicted when the mothers ente red prenatal care (Eyler et al., 1998a). Third, the Hobel Total and Perinatal Risk Scores were significantly higher for the cocaine-using mothers, and the difference in the Hobel Total Risk Score was due to higher Prenatal Risk Scores. There were no significant differences between groups on the Labor and Delivery and Neonatal Risk Scales (E yler et al., 1998a). The fourth difference between the groups was in the proportion of mothers who used other substances during their pregnancies. Significantly more cocaine -using mothers also used tobacco, alcohol, and marijuana compared to their matched co ntrols. Lastly, the num ber of infants born before 37 weeks gestation was significantly higher among the cocaine-using mothers, but there was no significant differe nce in mean gestational ag e between the two groups of infants as calculated using the method of Dubowitz, Dubowitz, and Goldberg (1970) ( M

PAGE 35

25 = 38.3 weeks, SD = 2.7 for cocaine-exposed, M = 38.7 weeks, SD = 2.9 for non-exposed; p = .24) (Eyler et al., 1998a). For the neonates, significant differences on all four growth measures—birth weight, length, head circumference, and ch est circumference—were found between those with prenatal cocaine exposure (PCE) and those without PCE. However, the Ponderal Index (Kliegman & King, 1983), calculated us ing the standard formula birth weight (grams) divided by length (cm)3, did not differ between group s although the infants with PCE had significantly lower birth weights a nd significantly shorter birth lengths as compared to the infants without PCE (Eyler et al., 1998a). In multiple regression analyses using cocaine, marijuana, alcohol, and tobacc o, no single drug or combination of drugs was a significant predictor of the Ponderal Index. In contrast, with head and chest circumference, there was an interaction between cocaine and tobacco such that the infants of mothers who used both were significantly smaller than the infant s of mothers who did not use tobacco, who only used tobacco, or who used cocaine but did not use tobacco (Eyler et al., 1998a). Measures Four sets of variables were used in the current study: demographic variables, a birth outcome measure, measures of c ognitive development at age 5 and 7, and caregiving environment measures. The demogr aphic variables, also called exogenous variables, were of interest for the curren t study because of their potential relationship with child cognitive development. The si x demographic variables were: amount of prenatal drug exposure (cocaine, alcohol, tobacco, and marijuana), Hobel prenatal obstetrical risk score, and gender. Child ethni city was excluded as a variable since the overwhelming majority of the sample (mor e than 75%) is African American, the two

PAGE 36

26 study groups were matched for socioeconomic status, and it is necessary to limit the number of variables for the analyses planne d. In terms of birth outcome measures, head circumference at birth has been shown to be predictive of later outcomes in several studies (Chasnoff, Griffit h, Freier, & Murray, 1992; Eyle r, Behnke, Garvan, Wobie, & Hou, 2002). Data from 21 child cognitive meas ures collected at birth, age 5, and age 7 were also used in the current study: 12 atte ntion measures, 8 verbal ability measures, and 2 reading ability measures. Finally, 7 measures of the caregiving environment gathered at age 5 and age 7 were included to examin e the relative contribution of the home environment to child cognitive development. Table 2-4 provides a summary of the all of the variables used in the study, th e construct they were designed to assess, and the ages at which they were assessed. Demographic Variables Prenatal drug exposure. Maternal use of cocaine alcohol, tobacco, and marijuana was obtained using a drug history interview procedure adapted from that of Day, Wagener, and Taylor (1985). Detailed in formation about the drug history interview procedure is provided in the Procedures section. Prenatal cocaine exposure was operationalized as a ratio of the number of weeks of reported cocaine use divided by the total number of weeks of each gestation plus 3 months prior to gestation (the period covered by the substance use interview). Pren atal alcohol exposure was quantified using the average number of ounces of abso lute alcohol consumed per day throughout pregnancy. Similarly, the averag e number of cigarettes smoked per day and the average number of marijuana joints smoked per da y throughout the pregnancy were used to measure prenatal tobacco and prenatal marijuana exposure, respectively.

PAGE 37

27 Head circumference. Orbitofrontal head circumference for each child measured in centimeters using a plastic coated tape. Hobel Obstetric Risk Scale (Hobel, Hyvarinen, Okad a, & Oh, 1973). The Hobel provides a quantitative assessment of 125 pren atal, intrapartum and neonatal factors that are associated with perinatal morbidity and death. For the current study only the Hobel Prenatal Risk Score was used as it was found to differ between the cocaine-using mothers and the comparison mothers in the original sample from which the participants were drawn (Eyler et al., 1998a). Scor es are assigned clinically to 50 historical and developing prenatal items; 40 early, interim, and late intrapartum factors; and 35 neonatal factors. Weights of 1, 5, or 10 are assigned to each of the factors based on their assumed relationship to perinatal morbid ity and death. The initial validity of the Hobel Obstetrical Risk scale was demonstrated in a sample of 738 mixed highand low-risk pregnancies using theoretically assigned weights to each of the variab les (Hobel, Hyvarinen, Okada, & Oh, 1973). Validity was further established in a larger sample of 1,417 mixed highand low-risk pregnancies (t hat included the 738 cases from the previous sample) by comparing the clinically assigned weights to those derived from a logistic regression model (Hobel, Youkeles, & Forsythe, 1979). In this latter study, the clinically assigned scores had a true positive clas sification rate of 82.5% and a true negative classification rate of 49.5%, which compared favorably wi th the logistic model’s predictions. In a review of obstetric risk-sco ring systems, Wall (1988) reporte d that the Hobel system has a sensitivity of .504 and .669, specificity of .685 and .701, and positive predictive validity of .228 and .293 for the antepartum period a nd the intrapartum periods, respectively.

PAGE 38

28 Measures of Cognitive Development Attention Measures Brazelton Neonatal Behavioral Assessment Scale (BNBAS; Brazelton, 1984). Three BNBAS cluster scores, Habituation, Or ientation, and Regulati on of State, were used as measures of infant attention after birth. The BNBAS is a widely used instrument for assessing the neurobehavioral responses of the newborn to his or her new extrauterine environment. It is designed as an interactive assessment for use with newborns from 36 to 44 weeks gestational age. The BNBAS consists of 28 behavioral items scored on a ninepoint scale and 21 reflex items scored on a four-point scale. The behavioral items examine behaviors such as response to vi sual, auditory and tactile stimulation, orientation, alertness, activity, and irritability. Central to th e behavioral assessment is the newborn’s state of consciousness—deep sleep, li ght sleep, drowsy quiet alert, active alert or crying—which serves as the foundation fo r evaluating his or her sensory and motor responses. The scoring summary divides th e items into four general domains of functioning; however, a seven-cluster scor ing method has been developed based on conceptual and empirical methods. The se ven BNBAS clusters are: Habituation, Orientation, Motor, Range of State, Regul ation of State, Autonomic Stability, and Reflexes. Psychometrically, test-rest reliabi lity is difficult to establish for the BNBAS due to rapid changes in the organization of the neonate’s behavior during the first few days and weeks. The BNBAS requires extensive training to obtain interrater reliability of .92 as recommended by the manual. The validity of the BNBAS has been established by more than 25 years of clinical and research use (Brazelton, 1984).

PAGE 39

29 Integrated Visual and Auditory Continuous Performance Test (IVA CPT; Sandford & Turner, 1994; Sandford, 1995). The two IVA CPT composite scores for the visual modality, the Visual Attention Quotient (VAQ) a nd Visual Response Control Quotient (VRCQ), were used as measures of attention at ages 5 and 7. The visual attention composite scores we re chosen over the auditory attention composite scores since the other attention measures in th e study—Letter Cancellati on, Trail Making Test, and the WISC-III Coding subtest—all rely primarily on the visual modality. Test protocols were reviewed to determine va lid profiles based on the IVA CPT manual criteria. The VAQ is a composite score compri sed of three raw scores: Focus (response variability), Vigilance (omission errors) a nd Speed (mean reaction time for all correct trials). The VRCQ is a composite score al so derived from three raw scores: Prudence (impulsivity), Consistency (response variab ility), and Stamina (mean reaction time for correct responses to the fi rst 200 and last 200 trials). The IVA CPT is a 13-minute test of atten tion for children and adults designed to provide data for differentiating between the su btypes of Attention Deficit/Hyperactivity Disorder specified in the Diagnostic a nd Statistical Manual4th Edition (DSM-IV; American Psychiatric Association, 1994). The IVA CPT measures responses to 500 intermixed visual and audito ry stimuli spaced 1.5 seconds apart. The task involves responding by clicking a computer mouse when the stimulus is a visual or auditory 1 and inhibiting responses when the stimulus is a vi sual or auditory 2. The stimuli are presented in pseudo-random order in five sets of 100 tr ials with each set cons isting of two 50-trial blocks. The blocks are counter balanced between visual and auditory stimuli and between frequent presentation of target stimuli (des igned to elicit impulsiv ity) and infrequent

PAGE 40

30 presentation of target stimuli (designed to elicit inattentiveness). Overall, the IVA CPT yields six composite quotient scores on two factors (Response Control and Attention) and 22 raw scores which comprise a Fine Moto r Regulation (Hyperactiv ity) scale, three Attribute scales, and six Va lidity scales (Sandford, 1995). Limited demographic information about th e normative sample for the IVA CPT is available from the manual. The sa mple consisted of volunteers ( N = 487, males = 210, females = 277) ranging in age from five to 90 years. Individuals in the normative sample were not known to have past neurological disorders or current psychological, learning, attentional problems or to demonstrate hype ractivity. In addition, the normative sample was screened for medications other than bi rth control and nasal sprays and was not currently active in psychotherapy or c ounseling. Significant gender differences were found on two scales: males had faster reacti on times but females made fewer commission (impulsive) errors. In addition, significant ag e effects for mean reaction time for correct responses followed a U-shaped curve. The test appears to be more demanding for younger children, as a rapid improvement (reduction in reaction time) was seen for children between the ages of 5 and 7. Reacti on time continues to improve between 8 and 12 years of age then plateaus between the mi d-teen to young adult years. Reaction time was fairly stable through middle age and then slowed down slightly after age 45. Normative information in the computerized database that accompanies the test is reportedly divided into "appropriate age and sex groups (Sanford, 1995). The limited information available about th e psychometric properties of the IVA CPT suggests that it has adequate reliabi lity and validity. The IVA CPT's test-retest reliability was studied using 70 normal volunteers (43 females, 27 males) between 5 and

PAGE 41

31 70 years of age (mean age = 21.8 years). Correl ation for a oneto four-week interval between test administrations for the Visual Attention Quotients wa s very strong at .75. For the visual attention Validity and Attribute scores, the correlations ranged from .34 to .80. Validity studies on the IVA CPT were condu cted in a small sample of 26 children, ages 7 to 12, diagnosed by a physician or psychologist as having ADHD and a comparison group of 31 children with no know n neurological, learning, emotional or ADHD related problems. Results indicated th at IVA CPT shows excellent sensitivity, specificity, positive predictive power, and negative predictive power: 92%, 90%, 89% and 93%, respectively. Concurre nt validity was establishe d by comparing the IVA CPT to two other continuous performance tests and two rating scales. The IVA CPT showed 90% to 100% agreement with th ese other measures and had th e lowest false positive rate at 7.7% (Sandford, 1995). Letter Cancellation (Diller, Ben-Yishay, Gerstman, Goodin, Gordon, & Weinberg, 1974). The Letter Cancellation task tim e to completion is used as a measure of attention at ages 5 and 7. Mi rsky et al. (1991) found, in both adults and children, that the Letter Cancellation test loaded on the “focus-e xecute” factor of attention in their threefactor model of attention. In addition to attention, the Lett er Cancellation test is thought to assess visual scanning, motor speed, and activation and inhibiti on of repetitive motor responses (Lezak, 1995). The task consists of crossing out a target character that is randomly interspersed approximately in an ar ray of at least five different characters. Three scores can be derived based on speed (time to completion), number of omission errors, and number of commission errors. The Letter Cancellation test has been found to be sensitive to a variety of problems in brai n-damaged subjects, including spatial neglect

PAGE 42

32 in right hemisphere stroke patients and temporal processing difficulties of left hemisphere stroke patients (Lezak, 1995). Trail Making Test (TMT; Reitan & Wolfson, 1985). The TMT Trails A time to completion is used as measure of attenti on at age 7. The TMT is one subtest in the Halstead-Reitan Neuropsychologi cal Test Battery and consists of two parts labeled Part A and Part B. The TMT is thought to measur e a variety of functi ons including attention, visual scanning, sequencing, mental flexibili ty, and motor speed and agility (Lezak, 1995; Spreen & Strauss, 1999). In Part A, the subject is instructed to draw lines to conn ect circles containing numbers scattered randomly on a page in numerical order. In Part B, the participant must draw lines alternately between circles contai ning numbers and circles containing letters in numerical and alphabetical order. Only Part A will be used since a meta-analysis of four studies of children ages 9 to 14 found that Part B may be less reliable in younger children (Leckliter, Forster, & Klonoff, 1992). Scoring fo r the test is based on time to completion. In adult studies using a variety of patient populations (except those with schizophrenia), the reliability coefficients generally range from .64 to .94 (Spreen & Strauss, 1998). Mirsky et al (1991) found using principal component analysis on test scores from a mixed sample of adults that both Parts A and B of the TMT loaded most highly on a “perceptual-motor speed” factor (.70 and .63, resp ectively), corresponding to their Focus-Execute component of attention; however, ther e were also significantly secondary loadings on a “vigilance” factor ( .43 and .45 for Parts A and B, respectively), corresponding to their Sustai n component of attention.

PAGE 43

33 Wechsler Intelligence Scale for Children-Third Edition (WISC-III, Wechsler, 1991). The WISC-III Coding and Digit Span subtests were used as a measure of attention at age 7. The WISC-III is a well-validated test of general cognitive functioning for children ages 6 years to 16 years, 11 months The test was standardized on a national sample of 2,220 children stra tified by age, sex, race/ethni city, geographic region and parent education according to the 1988 U.S. Census. The WISC-III has demonstrated good psychometric properties. The validity of the WISC-III is based, in part, on the numerous criterion-related studies conducted on its predecessor, the WISC-R. Factor analytic studies as well as correlational stud ies with three other Wechsler Scales (the WISC-R, WPPSI-R and WAIS-R), other ability tests, neur opsychological tests, and school grades support the validity of the WI SC-III. In addition, the data on the WISC-III has been collected using samples of excep tional children (gifted, mentally retarded, learning disabled, and speech/language delays ) and clinical groups (Attention Deficit/ Hyperactivity Disorder, severe conduct disorder, and epilepsy) (Wechsler, 1991). The reliability coefficient for the Coding subtest at age 7 is .70 with an average reliability of .79 for the range of age 6 to 15 years. Validity for the Coding subtest is based on its factor analytic st udies showing that correlates wi th a Processing Speed factor rather than Verbal or Performance factors. A dditionally, in a factor analytic study of child attention measures, Mirsky et al. (1991) found that the number correct on the Coding subtest loaded on their “focus-execute” fa ctor of attention al ong with number of omissions and time to completion on a Digit Cancellation task. Verbal Ability Wechsler Intelligence Scale for Children-Third Edition (WISC-III; Wechsler, 1991). The WISC-III Comprehension, Information, Similarities, and Vocabulary subtests

PAGE 44

34 were used as measures of verbal ability at age 7. Reliability coefficients for the four subtests ranged from .72 to .79 for the 7-year-olds in the normative sample, and .67 for the composite Verbal Comprehension factor ( N = 200). The validity of the Information, Comprehension, Vocabulary, and Similarities subt ests as measures of verbal ability is based on the moderate intercorrelations betw een the subtests (ranging from .46 to .64 for 7-year-olds in the normative sample), as well as factor analytic studi es showing that the four subtests load together on the same factor (Verbal Comprehension). Again, the overall reliability and validity of the WI SC-III is based on extensive research on its predecessor, the WISC-R, correlational stud ies with other tests of ability and neuropsychological tests, and studies using vari ous special populations. Wechsler Preschool and Primary Scale of Intelligence—Revised (WPPSI-R; Wechsler, 1989). The WPPSI-R Comprehe nsion, Information, Similarities, and Vocabulary subtests were used as measures of verbal ability at age 5. The WPPSI-R is a well-validated test of general cognitive func tioning designed for children ages 3 years to 7 years, 3 months. The test was standard ized on a national sa mple of 1,700 children stratified by age, sex, race/ethnicity, ge ographic region and parent education and occupation based on survey data gathered by U.S. Census Bureau in 1986. The WPPSI-R has demonstrated good psychometric properties. For the age group of interest (5 years), split-half reliability coefficients range from .59 to .86 for the individual subtests. Stability coefficients during a test-rest interval of 3 to 7 weeks ( N = 175) for the individual subtests ranged from .53 to .81. The overall validity of the WPPSI-R is based, in part, on studies of its predecessor, the WPPSI. In addi tion, the WPPSI-R has been evaluated using factor analytic studies and correlational studies with the WISC-R, Stanford-Binet

PAGE 45

35 Intelligence Scale-Fourth Edition (Thorndike Hagen, & Sattler, 1986), the McCarthy Scales of Children’s Abilities (McCarthy, 1972) and the Kaufman Assessment Battery for Children (Kaufman & Kaufman, 1983). Validity studies of the WPPSI-R have also been conducted using samples of gifted, menta lly deficient, learning disabled and speech/language impaired children (Wechsler, 1989). Reading Ability Wechsler Individual Achievement Test (WIAT; Wechsler, 1992). The WIAT Basic Reading and Reading Comprehension subt ests were used as measures of reading ability at age 7. The Basic Reading subt est has 55 items covering picture naming, vocabulary, and single word reading. The Read ing Comprehension subtest consists of 38 items that require the examinee to read a short passage and answer questions presented orally by the examiner. The WIAT is a well-validated test of academic achievement for children ages 5 years to 19 years, 11 months The test was standardized on a national sample of 4,252 children stra tified by age, sex, race/ethni city, geographic region and parent education according to the 1988 U.S. Census ( N = 331 for the age group of interest, 7 years). A subgroup of 1,284 childre n from the WIAT standardization sample was also administered one of the Wechsler intelligence scales. For the age group of interest (7 years), 100 children were admi nistered both the WIAT and the WISC-III. A weighting procedure was used to assure that the scores for the subgroup were comparable to those of the WISC-III standardization sample. The WIAT consists of eight subtests; however, only the psychometric properties of the Basic Reading and Reading Comprehension subtests will be reviewed here. Overall, both subtests have demonstrated good psychometric properties. For 7-year-olds, the split-half reliability coefficients for the Basic Reading and Reading Comprehension

PAGE 46

36 subtests are .95 and .93, respectively. Stability coefficients for the Basic Reading and Reading Comprehension subtests are .91 and .89 for grade 1 ( N = 76) and .94 and .90 for grade 2 ( N = 74), respectively. Reliab ility and stability coefficients are even higher for the overall Reading Composite, which is co mprised of the two subtests. The age 7 reliability coefficient for the Reading Compos ite is .97, and grade 1 and grade 2 stability coefficients are .95 and .96. The content validity of the WIAT is ba sed on reviews by curriculum experts and empirical item analysis studies. The WIAT’s construct and criteri on-related validity was determined by correlational studies using the WIAT subtests and other individuallyadministered achievement tests. Across a vari ety of ages, the Basic Reading and Reading Comprehension subtests were found to correla te from .79 to .87 (median = .82) with the analogous subtests in five other achievement test batteries including the Wide Range Achievement Test-Revised (Jastak & Wilkinson, 1984) and the Woodcock-Johnson Psycho-Educational Battery-Revised Te sts of Achievement (Woodcock & Johnson, 1991). Both subtests also have significant correlations (>.40) with school grades in a sample of children ages 6 to 19 years ( N = 867). In addition, studies of special groups of children (including those identified as gifte d, or having mental retardation, emotional disturbance, learning di sabilities, Attention Deficit Hype ractivity Disorder, and hearing impairment) support the validity of the WIAT (Wechsler, 1992). Caregiving Environment Measures Home Observation for Measu rement of the Environment (HOME; Caldwell & Bradley, 1984). Seven subscales of the HOME were used as measures of the child’s caregiving environment at ages 5 and 7. HOM E scores were determined by observation by trained interviewers during interviews with the child caregivers in their homes. The

PAGE 47

37 four subscales of the Early Childhood vers ion of the HOME (EC HOME) most closely related to literacy development—Learning Stimulation, Language Stimulation, Learning Stimulation, and Variety in Experience—were used to assess the caregiving environment at age 5. The three subscales of the Middle Childhood version of the HOME (MC HOME) most related to literacy devel opment—Growth Fostering Materials and Experiences, Provision for Active Stimulation, and Family Participation—were used to assess the caregiving environment at age 7. The HOME is a screening measure that assesses factors related to th e nurturance and stimulation in a child’s home environment that are believed to be important for cogni tive development. The HOME was designed as an alternative to sociodemogr aphic factors for identifying ch ildren at “high risk” for intellectual/academic problems. Scores are ba sed on both the observer’s visual inspection of the home and self-report of the child's primary caregiv er obtained through a semistructured interview during a 45to 90-mi nute home visit. The HOME has been found to be significantly correlated with longitudinal cognitive test performance and academic achievement in children ages 3 to 10 year s (Bradley, Caldwell, & Rock, 1988; Bradley, 1994; Bradley & Whiteside-Mansell, 1998). Four forms of the HOME are available: an Infant-Toddler versi on (ages birth to 3 years), an Early Childhood version (ages 3 to 6 years), a Middle Childhood version (ages 6 to 10 years), and an Early Adolescent vers ion (ages 10 to 14 years). Since the Early Childhood (EC) and Middle Childhood (MC) vers ions were used in the present study, only they are reviewed here. The EC HOM E contains 55 items clustered in eight subscales: 1) Learning Material s, 2) Language Stimulation, 3) Physical Environment, 4) Parental Responsivity, 5) Learning Stimula tion, 6) Modeling of Social Maturity, 7)

PAGE 48

38 Variety in Experience, and 8) Acceptance of Child. The MC HOME contains 59 items clustered into eight subscales: 1) Parental Responsivity, 2) Physical Environment, 3) Learning Materials, 4) Active Stimulation, 5) Encouraging Maturity, 6) Emotional Climate, 7) Parental Involvement, and 8) Family Participation. The various versions of the HOME and their subscales have undergone name changes, reorganization, or bot h and over time. However, no significant changes in the number of items or item content have b een made for the EC HOME and MC HOME. When the data from the EC HOME were coll ected in the current study, it was called the “Preschool” version. Since the st ructure of the items has not changed since the scale was renamed, the current names for the EC HOME subscales will be used throughout this study. Of the four EC HOME subscales us ed in the current study, only two were relabeled when the parent scale was rename d. The Language Stimula tion and Variety in Experience subscales remained the same while the Learning Materials subscale was previously labeled “Learning Stimulation” and the current Learning Stimulation subscale was called “Academic Stimulation.” The MC HOME, called the “Elementary” version when the data for this study were collected, underwent reorganization of it s items into eight subscales rather than seven subscales. The three subscales used in the current study were not affected by this structural change. Thus, the current names for the MC HOME subscales will be used throughout the study. Of the three MC HOME su bscales used in the current study, one was relabeled when the parent scale was re named. The Learning Materials subscale was previously called “Growth Foster ing Materials and Experiences.”

PAGE 49

39 All four versions of the HOME have been found to have good psychometric properties. Test-retest reliability, as measur ed by coefficient alpha, is above .90 for the total scores and is generally higher for the longer than the sh orter subscales. Interobserver agreement is reported as 90% or higher fo r all versions. Concurrent and predictive validity studies have shown that the HOME is significantly correlated with IQ, as high as r = .58. Low to moderate correlations ( .30 to .60) between EC HOME scores and children's contemporaneous and later inte llectual and academic performance have generally been found (Bradley, 1994). Similar re lationships have been reported for MC HOME scores and children's school perfor mance and classroom behavior (Bradley, Caldwell, Rock, Hamrick, & Harris, 1988). Th ese relationships have been found in African American as well as European American samples (Bradley & Caldwell, 1981; Bradley, Rock, Caldwell, Harris & Hamrick, 1987). While HOME scores have low to modest correlations with a wide variety of demographic variables including race, family structure, neighborhood, and maternal age, two studies have shown that no single demographic factor accounts for much of th e variance in HOME scores and that all the demographic factors together only account for about 50% of the variance (Bradley & Caldwell, 1981; Bradley, Mundfrom et al., 1994). Procedure Detailed information regarding participant recruitment and assessment of birth outcome measures is provided in Eyler et al. (1998a, 1998b) and is summarized here. Recruitment of participants took place be tween July 1991 and July 1993 with the last child born in February, 1994. Institutional Review Board approval was obtained for study procedures and incentives. Informed consent was carefully obtained for all participants, including those who were illiter ate. The consent process incl uded an explanation of child,

PAGE 50

40 maternal, and family measures, drugs tests and interviews, the Federal Certificate of Confidentiality, and the distinction between the researchers and clinical providers in assurance of confidentiality. All participan ts were recruited fro m women designated to deliver at Shands Teaching Hospital, a tertiary care center. Exclusion criteria included rare but majo r maternal illnesses diagnosed before pregnancy that are known to affect pregna ncy or developmental outcome, such as diabetes, sickle cell disease, and mental retardation, as we ll as women who abused legal drugs or used any illicit drugs other than cocaine and marijuana. In addition, only mothers who spoke English and were equal to or greater than 18 years of age were consented for enrollment in the study. A priori participant matching criteria were developed during the original longitudinal study to minimize the effect of possible confounding variables on pregnancy or child outcomes. Four matching criteria fo r the control group we re chosen, three of which were based on characteri stics that significantly diffe red between prenatal cocaine users and the general obstet ric population and which have been shown to relate to pregnancy or developmental outcome. These th ree matching criteria we re: 1) the level of Hollingshead Index of SES, 2) racial/ethnic group membership (African American versus other racial/ethnic categories), and 3) number of previous births (multiparous or primiparous). The fourth matc hing criteria, locati on of prenatal care, was chosen to equate groups on risk factors or complica tions that developed during, but not before, pregnancy. This variable included the local public health unit, out lying clinics (which sent only high-risk women to delive r at Shands) or no prenatal care.

PAGE 51

41 The researchers approached 2,526 potential participants of whom 85% gave informed consent to participate in the st udy. Most cocaine-using participants were enrolled prenatally (75%) from the two clos est county public health prenatal clinics or from the hospital’s high-risk referral prenat al clinic by approach ing all non-excluded potential participants. The addi tional participants were recr uited when they arrived to deliver at the hospital. This latter gro up consisted of women who had received no prenatal care or those whom the researchers had been unable to interview in the prenatal clinic. Of the 2,526 potential pa rticipants, 179 were approach ed for consent at delivery and 89% gave informed consent. The 372 refu sals included 13 women who were willing to give consent for the study but were unabl e or unwilling to provide a urine specimen at enrollment required for continued participat ion. After the first interview, 22 of the women who consented (11 cocaine users a nd 11 nonusers who would have been potential matches) were eliminated from study. Most ( n = 16) were found to have used excluded illicit drugs, while three re ported using confounding prescrip tion medications and another three were no longer able to deliver at Shands Teaching Hospital. As each cocaine-using woman who cons ented and met excl usion criteria was identified, one or two participants from th e pool who consented, who denied prenatal cocaine use, and whose urine specimens s howed no evidence of cocaine use were selected for each match category. The oldest matched control from the appropriate category was then used as the final match. Th e two final groups of participants consisted of 154 cocaine-using women and 154 non-users. The drug use interview, adapted from th at of Day, Wagener, and Taylor (1985), was administered by a one of a number of well-trained, non-judgment al interviewers who

PAGE 52

42 attempted to establish rappor t after informed consent was given by the maternal participant. Interviewers carefully read a nd explained all portions of a detailed drug history due to the low literacy level of the mothers enrolled in the study. Interviews were conducted at the end of each trimester, wh enever possible, and details about drug use during the previous three months was probe d in order to induce less guilt for the participants. Women with very late or no prenatal care were interviewed after birth about drug use throughout their pregnancy. Enrollment ra tes at the end of the first trimester, the end of the second trimester, and at delivery were: 41% (41 cocaine; 84 control), 34% (61 cocaine, 44 control) and 25% (69 cocaine, 9 control), respectively. To trigger memory around real-time events, calendars were used to help women recall their drug use history within the context of her pregnancy. In addi tion to cocaine, participants were queried about their use of drugs from several categor ies, including marijuana, alcohol, tobacco, and other illicit drugs (the latter for exclusi on purposes) using street or slang drug names. The amount (or cost) and timing of each woman’s usual use of each drug were recorded. Increases and decreases in usage patterns were also noted in order to calculate a more accurate average use per trimester. Urine specimens were obtained for drug screening on two occasions that could not be anticipated by the participants. The first specimen was collected on the day of enrollment in the study. Women who consented but refused to provide a urine specimen on the same day were dropped from the st udy. The second specimen was obtained from the mother on the day of the baby’s birth if an infant specimen was unavailable. A full toxicology screen of the urine was co nducted using fluorescence polarization

PAGE 53

43 immunoassay. Positive drug screens were then confirmed with gas chromatography/mass spectroscopy. Study measures were administered to the children and their mothers or another primary caregivers at several different a ssessments. Throughout the course of the longitudinal study, all of the measures have be en administered by trained, certified, or licensed professionals blinded to the study group membership of the mother and child. In the rare cases when a tester was unmaske d, other backup testers were used for the assessments. The first set of child assessments occu rred shortly after birth. Infants were evaluated within the first day or as soon as th ey were well in the Shands Hospital Clinical Research Center, which provided controlled conditions of light, sound, and temperature. In a few cases in which the infants were uns table, they were evaluated in the nursery. Orbitofrontal head circumference measurem ents were obtained by one of a team of neonatal nurse practitioners blinded to th e drug history of the mother. The Hobel Obstetric Risk Scores were determined pos tdelivery by medical pe rsonnel trained by one of the larger study's principal investigat ors, Marylou Behnke, M.D. The BNBAS was administered midway between feedings as close to 40 weeks postconceptual age as possible. In the current sample the majority of infants (69%) were evaluated within 24 hours after birth. Another 20% were admi nistered the BNBAS within 48 hours after birth. The BNBAS was administer ed by certified, reliable eval uators blinded to the drug history of the mothers. The second set of assessments took place when the children were approximately 5 years old. The Early Childhood Home Observat ion for Measurement of the Environment

PAGE 54

44 (EC HOME) was completed by one of two traine d female interviewers during interviews conducted with the primary caregivers in the family's home. The age 5 cognitive test battery, including the IVA, Letter Cancella tion, and WPPSI-R verbal subtests, was administered by one of two licensed school ps ychologists in private practice who were blinded to the child's group membership. Ad ministration of the c ognitive battery took place on the Project Care bus while it was pa rked either on the grounds of the child's school or outside the child's ho me. There was no significant difference in the mean age at which the two study groups were admini stered the age 5 test battery [ t (240) = .08, p = .93] The combined average of the groups' ages were 5.34 years at the time of the first cognitive test battery. The third sets of assessmen ts occurred when the children were approximately 7 years old. The Middle Childhood Home Ob servation for Measurement of the Environment (MC HOME) was completed during interviews conducted with the primary caregivers in the family's home by the sa me interviewers who conducted the age 5 interviews. The age 7 speech and language assessment, which included the Wechsler Individual Achievement Test (WIAT) read ing subtests, was admi nistered following a physical examination by one of two licensed nurse practitioners blinded to the child's group membership. The age 7 speech and language assessment was completed by on the Project Care bus while it was parked on the gr ounds of the child's school or outside the child's home. The age 7 cognitive test batter y also took place on the Project Care bus and was typically done within a day or two of the physical exam and speech and language assessment. The same two blinded, licensed sc hool psychologists who administered the age 5 cognitive test battery also gave the age 7 cognitive test battery. The children were

PAGE 55

45 provided breaks, including snacks, routinely and as needed during administration of the test batteries. There were no significant di fferences in the mean age at which the two study groups were administered the age 7 speech and language or cognitive test batteries [ t s(240) = 1.04, .40, p s = .30, .69, respectively]. The combined averages of the groups' ages were 7.31 and 7.29 years during the spee ch and language and c ognitive batteries, respectively. After assessment protocols were reviewed for scoring accuracy, the scores for each participant were hand-entered into a Mi crosoft Access database. To minimize input errors, the data were entered a second time and checked against the original input. Discrepancies between the two sets of entrie s were reconciled by checking the protocols. Age-corrected scaled scores for the WPPSI-R and WISC-III verbal ability subtests and age-corrected sta ndard scores for the WIAT read ing subtests were used. For age 5 Letter Cancellation, age 7 Letter Cancel lation and TMT Trail A, time in seconds to complete the task was used. Raw scores were used for the EC and MC Home subscales. For structural equation modeling (SEM), it is not necessary for all of the variables to be in the same metric as the solution can be sta ndardized by setting fact or variances to one. Hypotheses Based on a review of the relevant lite rature, three a priori hypotheses were developed to examine the relationship be tween prenatal cocaine exposure and the development of attention and reading skills in children: 1. Performance on the BNBAS Habituati on, Orientation, and State Regulation Supplementary Scales will be significantly correlated with pe rformance on age 5 and 7 attention measures for both groups of children in the study. 2. Children with prenatal cocaine exposure (PCE) will perform significantly worse than matched controls on: a) attentiona l measures at age 5 (IVA CPT and Letter Cancellation), b) attentional measures at age 7 (IVA CPT, Letter Cancellation,

PAGE 56

46 TMT Part A, WISC-III Coding, and WISC -III Digit Span), and c) reading measures at age 7 (WIAT Basic Reading and Reading Comprehension subtests). 3. It is hypothesized, regardless of signi ficant group differences on measures of attention and reading, that performance on attentional measures at birth, age 5, and age 7 will be significant predictors of reading at age 7 after controlling for verbal ability and the caregiving envir onment. It is hypothesized that PCE will have both a direct effect on attention and an indirect e ffect on attention that is mediated by birth head circumference. It is also hypothesized that PCE will have an indirect effect on r eading at age 7 that is mediated through attention. Data Inspection and Analyses Data Screening The first step in the data analysis plan involved screening the data for violations of normality and for missing values. Data screening was conducted using SPSS and PRELIS 2.52 (Jreskog & Srbom, 2001b). W ith the exception of the four drug variables, data with significant skewness, kur tosis, or both were transformed into normal scores. The drug variables were not transformed as it was expe cted that these data would not have a normal distribution. Missing Data A large number of missing values we re found for the Brazelton Neonatal Behavioral Assessment Scale (BNBAS) Habitu ation, Orientation, and Regulation of State scores. Specifically 83 participants ( n = 39 for PCE group, n = 44 for non-exposed group) were missing the Habituation score, 45 partic ipants were missing the Orientation score ( n = 30 for PCE group, n = 15 for non-exposed group), eight were missing both the Habituation and Orientation scores ( n = 5 for the PCE group, n = 3 for the non-exposed group), and 20 participants were missing all three scores ( n = 15 for PCE group, n = 5 for non-exposed group). As reported in Eyler et al (1998b), a significantl y larger proportion of infants with PCE than non-exposed infants fa iled to come to a quiet, alert state so that

PAGE 57

47 the Orientation items could be administer ed (25% vs. 12%). Independent sample t -tests using Bonferroni correction for multiple co mparisons were conducted between the groups of children with missing BNBAS scores and th e rest of the sample Results indicated no significant differences between the groups mi ssing the Habituation score, Orientation score, or both scores and the rest of the sa mple on any of the demographic variables, head circumference, or gestational age. For the group missing all three BNBAS scores, one significant difference was found from the re st of the sample. There was a larger proportion of African Americans in the group mi ssing all three BNBAS sc ores than in the rest of the sample [ t (26) = -2.41, p = .019]. It was decided to use all available BNBAS data to evaluate the first hypothesis but to exclude the BNBAS from the path analysis needed to evaluate the third hypothesis. A number of missing values were also found for the Intermediate Visual and Auditory Continuous Performance Test (IVA CP T) Visual Attention Quotient (VAQ) and Visual Response Control Quotient (VRCQ) sc ores at both ages 5 and 7. Specifically, 17 children were missing the IVA CPT at age 5 ( n = 12 for PCE group, n = 5 for nonexposed group), 17 children were missing the IVA CPT at age 7 ( n = 8 for PCE group, n = 9 for non-exposed group), and one child (non-exposed) was missing both sets of IVA CPT scores. Independent sample t -tests using Bonferroni correction for multiple comparisons were conducted between the groups of children with missing IVA CPT scores and the rest of the sample. Results i ndicated no significant di fferences between the groups at age 5 or at age 7 on any of the dem ographic variables, head circumference, or measures of attention, reading, verbal abi lity, or the caregiving environment. It was

PAGE 58

48 decided to use all available IVA CPT data to evaluate the second hypothesis but to exclude these data from the path analysis needed to evaluate the third hypothesis. Not including the BNBAS or IVA CPT scor es, 29 participants were found to be missing only one of the remaining 30 data point s. The distribution of missing scores was as follows: Hobel prenatal risk score ( n = 6), age 5 Letter Cancellation ( n = 11), age 5 HOME Academic Stimulation subscale ( n = 1), and age 7 Trail Making Test (TMT) Part A ( n = 11). With the exception of the TMT, th e missing data points appeared to be random, so the mean score based on the child's group membership (PCE vs. nonexposed) was substituted for the missing data. For the TMT, the missing scores were gene rally the result of the child not being able to count to 15 or not being ab le to complete the practice item ( n = 5 for PCE group, n = 6 for non-exposed group). Since it is some what unusual for a 7-year-old not to be able to count to 15, independent-samples t -tests between the group of children missing the TMT and the rest of the sample were conducted. Results revealed significant differences on 14 variables as shown in Tabl e 2-4. A significant difference was found for ethnicity; all of the children who could not do the TMT Part A were African American. These children also scored significantly wors e than the rest of the sample on the Hobel prenatal risk score; EC HOM E Language Stimulation subscale; age 5 and age 7 Letter Cancellation; WPPSI-R Comprehension, Info rmation, and Similarities; WISC-III Digit Span, Comprehension, Information, Similar ities, and Vocabulary; and WIAT Basic Reading and Reading Comprehension. Since the differences between the group missing the TMT and the rest of the sample were si gnificant, a decision was made to replace the

PAGE 59

49 missing TMT Part A scores with a arbitrar y score of 300 seconds, the time limit for discontinuance of the test. Accounting for the Participants After imputing values for participants who were missing only one data point, a total sample size of 240 participants ( n = 120 for both groups) remained for further analyses. Thus, a total of 68 of the 308 participan ts originally enrolled in the prospective, longitudinal study were excluded from the cu rrent study. Of these 68 participants, 22 had more than one missing data point, 12 died pr ior to age 7, eight dr opped out of the study, 10 were lost to follow up, six had moved out of the area, eight refused to participate in one or both of the age 5 and age 7 assessmen ts, one child was deaf, and one child was profoundly retarded and could not complete the assessments. To determine whether the differences found in the original study sample were also present in the smaller sample used in the current study, independent samples t -tests with Bonferroni correction for multiple comparisons were performed for the demographic variables and head circumference. Table 26 shows the results of the analyses, which revealed that the sample for the current study was very similar to the original sample. As in the original sample, the group with PC E had significant greater mean amounts of prenatal cocaine exposure [ t (238) = 17.43, p = .000], prenatal alcohol exposure [ t (238) = 5.69, p = .000], and prenatal tobacco exposure [ t (238) = 7.30, p = .000] than the nonexposed group. In addition, the group with PCE had significantly higher mean Hobel prenatal risk score than the non-expose d group, as in the original sample [ t (238) = 5.11, p = .000] and significantly smaller mean head circumference as compared to the nonexposed group [ t (238) = -3.26, p = .001]. Finally, statistica lly similar proportions of

PAGE 60

50 females and African Americans we re found in both study groups [ t s (238) = 1.42 and 1.20, p s = .156 and .232, respectively]. There were, however, two differences between the current sample and the original study sample. Unlike the original sample, ther e was no significant difference between the groups in mean amount of prenatal marijuana exposure [ t (238) = 2.06, p = .041]. Another difference between the current sample and the original sample was the significantly shorter mean gestational age for the group w ith PCE compared to the non-exposed group [ t (238) = -3.27, p = .001]. The average gestational age of the infants with PCE was 38.50 weeks compared to 39.29 weeks for the nonexposed infants. Wh ile statistically significant, the difference between the groups was less than one week. Statistical Analyses SPSS and LISREL 8.52 (Jreskog & Srbom, 2001a) were used to conduct all statistical analyses. The criterion for signifi cance tests for all a priori hypotheses was set at = .05 with Bonferroni correction for mu ltiple comparisons. To test the first hypothesis, a correlational analysis was c onducted to determine whether the BNBAS is significantly related to measur es of attention at age 5 a nd age 7. To test the second hypothesis of group differences on measures of attention and reading, a cross-sectional analysis using the independent samples t -test was conducted. Finally, to test for longitudinal associations between measures of attention administered at birth, age 5 and age 7 and reading ability at age 7 and the hypothesis that head circumference mediates the effect of PCE on attenti on and reading, structural equa tion modeling (SEM) was used in a combined sample of the children with PCE and without PCE.

PAGE 61

51 SEM is a powerful statistical technique that involves multiple regression analyses of factors. Factors, called la tent constructs, were derive d from the reliable shared variance of one or more observed variables in dicators and thus are free of measurement error. SEM allows for the examination of complex relationships between multiple continuous and discrete independent variab les and multiple continuous and discrete dependent variables. After spec ification of a model, which is a type of confirmatory factor analysis, SEM can be used to test a model, test specific hypotheses about a model (including mediational hypotheses) modify an existing model, or test a set of related models (Ullman, 2001). To ensure sufficient identif ication to conduct structural equation modeling (SEM), the number of unknown parameters must be less than or equal to the number of known pieces of information supplied to the program. In general, then, the number of indicator variables should be limited. In addition, to an alyze longitudinal data, a stable number of participants are required across all time points. For the 18 participants who were randomly missing one of the variables in th e study, the missing data was imputed using the mean score based on group membership (PCE vs. non-exposed). For the 11 participants who were missing the TMT, an arbitrary score of 300 seconds, the time limit for discontinuance of the test, was substituted for the missing value. Participants missing more than one data point were excluded from the analyses. Figures 2-1 and 2-2 provide diagrams of the proposed structural model of the longitudinal relationship between PCE, attention, and reading. Eight of the factors in the model are indicated by a single variable: prenat al cocaine exposure (COCAINE), prenatal alcohol exposure (ALCOHOL), prenatal tobacco exposure (TOBACCO), prenatal

PAGE 62

52 marijuana exposure (MARIJUA), Hobel pren atal risk (HOBEL), sex (SEX), head circumference at birth (HEADC), Letter Can cellation at age 5 (LCAN5), and Digit Span at age 7 (DSPAN7). Attention at age 5 is singly indicated by Letter Cancellation (LCAN5). Attention at age 7 was divided in to two separate factors. Digit Span was allowed to be a singly indicated factor (D SPAN7) because it does not have a visuomotor component while the other three attentional measures—Letter Cancellation, TMT Part A, and WISC-III Coding subtest—which have a visu omotor component were combined into a Visual Attention factor (VATTN7). The ag e 5 caregiving environm ent factor (HOME5) was indicated by four subscales from th e EC HOME while the age 7 caregiving environment factor (HOME7) was indicated by three subscales from the MC HOME. The age 5 and age 7 Verbal Ability factors (VERBAL5 and VERBAL7) were indicated by the four age-appropriate Wechsler subtests—C omprehension, Information, Similarities, and Vocabulary. The structure of the model was based on the predictive relationships that are expected to exist between the various factors. The six endoge nous variables were used to predict birth head circumference, which in turn is used to predict the three age 5 factors (LCAN5, VERBAL5, and HOME5). As expected with longitudinal data, each of the three age 5 factors (attention, verbal ability, and caregiving environment) was used to predict their respective age 7 f actors. As the only measure of attention obtained at age 5, Letter Cancellation factor was used to predict the two age 7 attention factors, DSPAN7 and VATTN7. The other two age 5 factors, VERBAL5 and HOME5, are used to predict their respective age 7 factors, VERBAL7 a nd HOME7. Finally, the four age 7 factors

PAGE 63

53 (DSPAN7, VATTN7, VERBAL7, and HOME7) were used to predict the final outcome, age 7 reading ability (READ7).

PAGE 64

54 Table 2-1 Continuous Variables that Differe d Significantly Between Mothers in the Two Original Study Groups Groupa Variable Cocaine Users Matched Controls p -value M SD M SD Age 27.6 4.8 23.8 5.5 .0001*** Week entered prenatal care 14.8 7.6 12.1 7.2 .003** Total Hobel score 94.2 72.1 78.5 48.2 .0276* Prenatal Hobel score 54.5 20.1 43.0 19.3 .00001*** an = 154 for both groups. p < .05, ** p < .01, *** p < .001, two-tailed using the independent samples t -test. Table 2-2 Non-continuous Variables that Differed Significantly Between Mothers in the Two Original Study Groups Groupa Variable Cocaine Users Matched Controls p -value Tobacco users 123 37 .0001*** Alcohol users 118 47 .0001*** Marijuana users 68 11 .0001*** Births < 37 weeks gestation 28 14 .03* an = 154 for both groups. p < .05, ** p < .01, *** p < .001, two-tailed using the independent samples t -test.

PAGE 65

55 Table 2-3 Variables that Di ffered Significantly Between Neon ates in the Two Original Study Groups Groupa Variable Cocaine-Exposed Matched Controls p -value M SD M SD Birth weight (g) 2985 668 3179 700 .03* Length (cm) 48.7 3.2 49.7 3.3 .007** Chest circumference (cm) 31.6 33.6 .01* Head circumference (cm) 33.6 34.8 .007** an = 154 in both groups. p < .05, ** p < .01, two-tailed using the independent samples t -test. Table 2-4 Summary of Variables for Current Study Variable Name Variable Label Construct Assessed Age Demographic Variables WUCFULL Average ratio of weeks of cocaine use Cocaine exposure prenatal + 3 months prior POAALC Average ounces of absolute alcohol consumed per day Alcohol exposure prenatal POATOB Average number of cigarettes smoked per day Tobacco exposure prenatal POAMAR Average number of marijuana joints smoked per day Marijuana exposure prenatal HEADC Head circumference Head circumference birth HOBPRE Hobel Prenatal Risk Prenatal risk birth SEX Sex Sex birth Child Cognitive Variables BRHABIT BNBAS Habituation Attention <1 week BRORIENT BNBAS Orientation Attention <1 week BRREGSTA BNBAS Regulation of State Attention <1 week IVAVAQ5 IVA CPT Vi sual Attention Quotient Attention 5 years IVAVRCQ5 IVA CPT Visual Response Control Quotient Attention 5 years LCANT5 Letter Cancellation time Attention 5 years IVAVAQ7 IVA CPT Vi sual Attention Quotient Attention 7 years IVAVRCQ7 IVA CPT Visual Response Control Quotient Attention 7 years LCANT7 Letter Cancellation time Attention 7 years TRAILA TMT Trail A time Attention 7 years W3COD WISC-III Coding Attention 7 years

PAGE 66

56 Table 2-4. Continued Variable Name Variable Label Construct Assessed Age W3DSS WISC-III Digit Span Attention 7 years WPCOM WPPSI-R Comprehensi on Verbal ability 5 years WPINF WPPSI-R Information Verbal ability 5 years WPSIM WPPSI-R Similarities Verbal ability 5 years WPVOC WPPSI-R Vocabulary Verbal ability 5 years W3COM WISC-III Comprehension Verbal ability 7 years W3INF WISC-III Information Verbal ability 7 years W3SIM WISC-III Similarities Verbal ability 7 years W3VOC WISC-III Vocabulary Verbal ability 7 years WIATBR WIAT Broad Reading Reading ability 7 years WIATRC WIAT Reading Comprehension Reading ability 7 years Caregiving Environment Variables H5LEAR EC HOME Learning Material s Home environment 5 years H5LANG EC HOME Language Stimulation Home environment 5 years H5ACAD EC HOME Language Stimula tion Home environment 5 years H5VARI EC HOME Variety in Expe rience Home environment 5 years H7GROW MC HOME Learning Material s Home environment 7 years H7ACTI MC HOME Active Stimulati on Home environment 7 years H7FAMI MC HOME Family Participation Home environment 7 years Note BNBAS = Brazelton Neonatal Behavior Assessment Scale, IVA CPT = Intermediate Visual and Auditory Continuous Performance Test, TMT = Trail Making Test, WISC-III = Wechsler Intelligence Scale for Children-Third Edition WPPSI-R = Wechsler Preschool and Primary Scale of Intelligence-Revised, WIAT = Wechsler Individual Achievement Test, EC HOME = Early Childhood Home Observation for Measurement of the Environment, MC HOME = Middle Childhood Home Observation for Measurement of the Environment.

PAGE 67

57 Table 2-5 Significant Differences Between Pa rticipants With and Without TMT Part A Scores Groupa Variable With TMT Without TMT p -value M SD M SD Ethnicityb .82 .38 1.00 .00 .000*** Hobel prenatal risk score 49.83 20.50 41.36 11.64 .042* EC HOME Language Stimulation 6.36 .96 5.73 1.49 .038* Letter Cancellation ag e 5 (secs.) 102.53 43.30 131.09 38.78 .033* WPPSI-R Comprehension 7.92 2.48 5.27 1.19 .001** WPPSI-R Information 7.22 2.48 4.18 1.25 .000** WPPSI-R Similarities 7.85 2.24 5.91 1.97 .005** Letter Cancellation ag e 7 (secs.) 58.70 20.94 80.27 28.89 .001** WISC-III Digit Span 8.94 2.59 4.18 1.47 .000* WISC-III Comprehension 8.20 3.27 4.90 2.47 .001** WISC-III Information 8.17 2.54 4.81 .98 .000*** WISC-III Similarities 8.27 4.49 3.00 2.45 .000*** WISC-III Vocabulary 8.44 2.62 5.00 2.41 .000*** WIAT Basic Reading 96.52 12.35 82.82 1.47 .000*** WIAT Reading Comprehension 92.24 11.67 0.64 2.91 .000*** Note. EC HOME = Early Childhood Home Observation for Measurement of the Environment, TMT = Trail Making Test, WPPSI-R = Wechsler Preschool and Primary Scale of IntelligenceRevised, WISC-III = Wechsler Intelligence Scale fo r Children-Third Edition, WIAT = Wechsler Individual Achievement Test. an = 229 for group with TMT scores, n = 11 for group without TMT scores. bEthnicity was coded so that African American = 1 and others = 0. Thus, the number in the table represents the proportion of the group that was African American. p < .05, ** p < .01, *** p < .001, two-tailed using the independent samples t -test and Bonferroni correction for familywise error rate.

PAGE 68

58 Table 2-6 Demographic Variables Comparing Groups in Current Study Groupa Variable PCE Non-exposed p -value M SD M SD Prenatal cocaine exposure .458 .288 .000.000 .000*** Prenatal alcohol exposure .232 .414 .002.005 .000*** Prenatal tobacco exposure 8.96 9.23 1.865.33 .000*** Prenatal marijuana exposure .009 .369 .002.127 .041* Hobel prenatal risk score 55.79 19.66 43.0918.84 .000** Gestational age (weeks) 38.50 2.10 39.291.61 .001* Head circumference (cm) 33.49 1.93 34.241.66 .001** Sexb .52 .50 .43.50 .156 Ethnicityc .80 .40 .86.35 .232 Note. PCE = prenatal cocaine exposure. an = 120 for both groups. bSex was coded so that female = 1, male = 0. Thus, the numbers in this row indicate the proportion of the group that was female. cEthnicity was coded so that African American = 1, others = 0. Thus, the numbers in this row indicate the proportion of the group that was African American.

PAGE 69

59 59 Figure 2-1. Proposed structural equation model with factors and indicators for relationships between prenatal cocaine exposur e, birth head circumference, attention, and reading while controlling for other prenatal drug exposure, Hobel prenatal risk, sex, verbal ability, and caregi ving environment. Primary relationships of interest are indicated by bold lines. Ellipses indicate latent variables (factors) while boxes indicat e measured variables. BNBAS = Brazelton Neonatal Behavioral Assessment Scale. EC HOME = Early Childhood Home Observation Measurement of the Environment. MC HOME = Middle Childhood Home Observation Measur ement of the Environment. IVA CPT = Intermediate Visual and Auditory Continuous Performance Test. TMT = Trail Making Test. WIAT = Wechsler Individual Achievement Test. WISC-III = Wechsle r Intelligence Scale for Children-Third Edi tion. WPPSI-R = Wechsler Primary and Preschool Scale of Intelligence Revised. Verbal ability at age 7 WPPSI-R subtests WISC-III subtests Caregiving environment at a g e 5 EC HOME subscales Caregiving environment at a g e 7 Reading at age 7 Letter Cancellation at age 5 Letter Cancellation TMT Part A WISC-III Coding Visual attention at age 7 Prenatal cocaine ex p osure Prenatal marijuana exposure Hobel prenatal risk Verbal ability at age 5 Head circumference at birth WISC-III Digit Span at a g e 7 Prenatal tobacco exposure Prenatal alcohol ex p osure MC HOMEsubscales WIAT subtests Sex

PAGE 70

60 Figure 2-2. Proposed structural equation model with factor na mes for relationships between prenatal cocaine exposure, birth h ead circumference, attention, and reading while controlling for other prenatal drug exposure, Hobel prenatal risk, sex, verbal ability, and caregiving enviro nment. Primary relationships of interest are indicated by bold lines. COCAINE = prenatal cocaine ex posure, ALCOHOL = prenatal alcohol exposure TOBACCO = prenatal tobacco exposure, MARIJUANA = prenatal mariju ana exposure, HOBEL = Hobel prenatal risk score, SEX = child se x, HEADC = head circumference at birth, LCAN5 = Letter Cancellation at age 5, VERBAL5 = verbal ability at age 5, HOME5 = caregivin g environment at age 5, DSPAN7 = Digit Span at age 7, VATTN7 = visual attention at age 7, VERBAL7 = verbal ability at age 7, HOME 7 = caregiving environment at age 7, and READ7 = reading ability at age 7. VERBAL7 HOME5 HOME7 READ7 LCAN5 VATTN7 COCAINE SEX MARIJUANA HOBEL VERBAL5 HEADC DSPAN7 TOBACCO ALCOHOL

PAGE 71

61 CHAPTER 3 RESULTS Hypothesis #1 To test the first hypothesis—that th e BNBAS Habituati on, Orientation, and Regulation of State scores obtained during the first week of life would have significant relationships with measures of childhood attention at ages 5 and 7—a correlational analysis was performed using all available data for both gr oups. There was only one si gnificant correlation between one of the BNBAS measures and one of the early childhood at tention measures. The BNBAS Orientation score was sign ificantly correlated with the age 5 IVA CPT visual attention quotient; while significant, this association was small ( r = .149, p < .05, N = 191). No significant relationships were found between the three BNBAS scores and Le tter Cancellation at age 5, or any of the age 7 attention measures (Letter Ca ncellation, IVA CPT scores, TMT Part A, WISCIII Coding, or WISC-III Digit Span). Intercorrela tions between all of the proposed attention measures in the study for the combined sample of children with PCE and non-exposed children are shown in Table 3-1. A post-hoc analysis was conducted to ex amine the relationship of the BNBAS Orientation score with the age 5 IVA CPT visual attention quotie nt separately by group (PCE vs. non-exposed). When analyzed separately by group, the relationship between the BNBAS Orientation score and the age 5 IVA CPT visual attention qu otient became non-significant. However, it appears that the si gnificant relationship found in the combined sample was driven largely by the PCE group ( n = 80, r = .217, p = .053) rather than the non-exposed group ( n = 111, r = .095, p = .32).

PAGE 72

62 Hypothesis #2 To evaluate the second hypothesis that children with PCE would perform significantly worse than non-exposed children on measures of atte ntion at ages 5 and 7 and on measures of reading at age 7, independent samples t -tests with Bonferroni correction for multiple comparisons were performed. As shown in Ta ble 3-2, no significant differences were found between the groups on any of the attention or reading measures. For age 5 Letter Cancellation, the PCE group took an average of 108 seconds to complete the task, compared to 100 seconds for the non-exposed group [ t (238) = 1.44, p = .15]. For each of the age 7 attention measures, the PCE and non-exposed groups obtained remarkably similar mean scores. For Digit Span and Coding, the scaled scores for the PCE group we re 8.71 and 10.14, respectively, while the scores were 8.73 and 10.56 for the non-exposed group, respectively [ t s(238) = -.06, and -.93, p s = .95 and .35, respectively]. For the ot her two time-based measures, Le tter Cancellation and TMT Part A, the PCE group completed the tasks in 60 and 52 seconds on average while the non-exposed group required 59 and 54 seconds on average [ t s(238) = -.36 and -.21, p s = .71 and .83, respectively]. Finally, on the two reading measur es, both groups performed in the average range, with the PCE group having slightly higher mean s scores than the non-exposed group. Scaled scores were 96.20 and 95.59 for the Basic Read ing subtest and 92.02 and 91.39 for the Reading Comprehension subtest for the PCE and non-exposed groups, respectively ( t s(238) = .39 and .42, p s = .70 and .68). Post hoc comparisons between the two groups were made for the remainder of the study variables using independent samples t -tests with Bonferroni corrections for multiple comparisons. As shown in Table 3-3, only two sta tistically significant group differences were found. One was for the EC HOME Learning Stimulation subscale [ t (238) 2.62, p = .01]; the other was for the MC HOME Growth Fostering Materials subscale [ t (238) = 3.12, p = .00].

PAGE 73

63 Ironically, the PCE group scored slightly higher than the non-exposed group on both measures. The PCE group obtained average scores of 7.41 for the EC HOME Learning Stimulation subscale and 5.11 for MC HOME Growth Fostering Materials subscale compared to respective means of 6.58 and 4.67 for the non-exposed group. This disparity is likely due to the fact that a significantly larger proportion of the children with PCE were living in placements away from their biological mothers at age 5 compared to non-exposed childr en (64% vs. 9%). Since there was only two significant between-groups diffe rences, combining the PCE and non-exposed groups for all further analyses was deemed appropriate to maximize statistical power. Hypothesis #3 To evaluate the third set of hypothe ses that a) performance on measures of attention will be significant pr edictors of performance in read ing, b) PCE will have a direct effect on attention and an indir ect effect on attention mediated by head circumference, and c) PCE will have an indirect effect on reading me diated by attention, stru ctural equation modeling (SEM) was performed on the combined sample of children with and without PCE. As noted earlier, the Brazelton Neonatal Be havioral Assessment Scale (BNB AS) and Intermediate Visual and Auditory Continuous Performance Test (IVA CPT) were dropped from th ese analyses due to large numbers of missing data. The first step in SEM is the development of the measurement model that, by definition, allows all the factors in the model to be correl ated. Several indices were used to assess the goodness-of-fit of both the measurement and struct ural models. One index used was the ratio between the chi-square statistic and the degrees of freedom for the model. Generally, a model is considered to fit well if chi-square is less than twice the degrees of freedom ( 2 < 2 df ). Other indices used to assess fit in the current study are Jreskog a nd Srbom's (1989) goodness-of-fit (GFI), Bentler's (1990) normed comparativ e fit (CFI) and Bentle r and Bonett's (1980)

PAGE 74

64 nonnormed fit (NNFI) indices. For each of these indices, better f it is associated with higher values, and .90 is generally considered a mi nimum acceptable level (Bentler & Bonett, 1980). Finally, the root mean square error of approxi mation (RMSEA) takes into account the error of approximation in the population, while the root mean square residual (RMR) provides a measure of the average size of the residual difference between the actual covariances among the observed indicators and the covariances predicted by a pa rticular model. For both the RMSEA and RMR, a value of 0.05 is considered an indicator of good fit (Browne and Cudeck, 1993; Bryant and Yarold, 1995). The initial measurement model was checked for signs of underiden tification including negative variances, correlations greater than 1.0, a nd factor loadings or correlations that seemed to have the wrong sign or were much smaller or much larger than expected. No signs of model underidentification were detected. Table 3-4 shows fit indices for the iterative process used to determine the final measurement model. The basic measurement model (M1), in which all variables were allowed to correlate, fit the data very well: 2 (281, N = 240) = 376.72, GFI = .90, CFI = .98, NNFI = .98, RMSEA = .04, and RMR = .05. No modifications of the measurement model were needed since all of the fit indi ces met their respective criteria for good fit. The next step in SEM is construction of a struct ural model that fits the data as well as the final measurement model with fewer estimated parameters. The hypothesized structural model (S1) also fit the data moderately well [ 2 (345, N = 240) = 536.37, GFI = .87, CFI = .97, NNFI = .96, RMSEA = .05, and RMR = .09] but significantly worse than the final measurement model. An iterative process was undertaken to improve the model's fit first by dropping insignificant paths one at a time, then inspecting the modifica tion indices to determine additional parameters to freely estimate. Table 3-4 displays the fit indices of the intermediate models between the

PAGE 75

65 initial hypothesized structural model and the final st ructural model. In the first four model steps, insignificant paths were dropped between the following factors: 1) Hobel and Head Circumference, 2) Head Circum ference and age 5 Letter Cance llation, 3) Head Circumference and age 5 caregiving environment, and 4) age 7 Digit Span and Reading. As expected, these changes did not significantly improve the fit of the model but did increase the degrees of freedom for subsequent tests of model fit. In the fifth model step (S6), a path was a dded between age 5 Verbal Ability and age 7 Digit Span, revealing a moderate re lationship between the two factors ( = .52). Adding this path significantly improved the model fit as comp ared to the previous structural model: 2 difference (1, N = 240) = 41.53, p = .00. Further modifications were still needed, however, because the structural model was still significant different from the measurement model: 2 difference (67, N = 240) = 120.53, p = .00. In the sixth model step (S7), th e insignificant path between age 5 Letter Cancellation and age 7 Digit Span was dropped with no significant change in the model's fit. In the seventh model step (S8), estimating th e path between age 5 Verbal Ability and age 7 Visual Attention reve aled another moderately strong relationship ( = .47). Again, estimating this additional path significantly improved the model fit: 2 difference (1, N = 240) = 25.09, p = .00. However, this model was still significan tly different from the measurement model: 2 difference (67, N = 240) = 96.55, p < .05. In the final model step, a path was added between Sex to age 7 Visual Attention. The magnit ude of this relationship was small ( = .29), indicating that girls performed better than boys on the visual at tention measures (females were coded as 1, males were coded as 0). Adding this final path also significantly improved the model fit over the previous structural model: 2 difference (1, N = 240) = 15.74, p = .00). No further attempts were

PAGE 76

66 made to improve the fit of the structural model si nce it was no longer differed significantly from the measurement model: 2 difference (66, N = 240) = 80.81, p > .05). The final structural model fit the data well [ 2 (347, N = 240) = 457.53, GFI = .89, CFI = .98, NNFI = .98, RMSEA = .04, and RMR = .06] and was a significant improvement over the initial hypothesized structural model with improvement s in all five fit indices. It should be noted, however, that two of the fit indices did not m eet the criteria for good fit: GFI was < .90 and the RMR was > .05. Nevertheless, the final structural model was able to account for 44% of the variance in the age 7 Visual Attention factor and 68% of the variance in the age 7 Reading factor. As predicted, the Visual Attention factor was the strong est predictor of age 7 Reading even after controlling for verbal ab ility and the caregiving environment ( s = .44, .41, and .14). In summary, the first hypothesis that the BN BAS scores collected during the first week would be significantly correlate d with early childhood measures of attention a nd reading could not be evaluated due to a large amount of missing data. The sec ond hypothesis, that children with PCE would perform significantly worse than non-exposed children on measures of attention at ages 5 and 7 and reading at age 7, was not supported. There was mixed support for the third set of hypotheses regarding relationships between attention and reading, PCE and attention, and PCE and reading. Visual attention at age 7 was f ound to be the strongest predictor of reading at age 7; however, PCE had no direct relationship with attention at age 5 or 7. PCE was found to have an indirect effect on readi ng at age 7 mediated by head circum ference at birth, verbal ability and visual attention.

PAGE 77

67Table 3-1 Intercorrelations Between All A ttention Measures for Combined Sample 1 2 3 4 5 6 7 8 9 10 11 12 1. BNBAS Habituation -.076 .090 .051 .051 .090 -.028 -.021 .070 -.023 .010 .058 2. BNBAS Orientation -.273** .149* .081 -.051 .033 -.019 -.014 -.069 .008 -.095 3. BNBAS State Regulation -.070 .018 -.087 .040 -.028 .003 -.118 .086 -.014 4. IVA CPT VAQ age 5 -.630** -.122 .129 .071 .024 -.215** .089 .216** 5. IVA CPT VRCQ age 5 --.158* .075 .000 -.056 -.178** .020 .219** 6. Letter Cancel age 5 --.127 -.152* .204** .298** -.277** .211** 7. IVA CPT VAQ age 7 -.374** -.031 -.161* .131* .247** 8. IVA CPT VRCQ age 7 --.009 -.100 .150* .138* 9. Letter Cancel age 7 -.148* -.284** -.105 10. TMT Part A --.205** -.303** 11. WISC-III Coding -.185** 12. WISC-III Digit Span -Note n s vary by cell due to missing data. BNBAS = Brazelton Neonatal Behavi oral assessment Scale, IVA CPT = Intermediate Visual and A uditory Continuous Performance Test, VAQ = Visual Attention Quotient, VRCQ = Visual Response Control Quotient, TMT= Trail Making Test, and WISC-II I = Wechsler Intelligence Scale for Children-Third Edition. p < .05, ** p <.01, two-tailed.

PAGE 78

68 Table 3-2 Group Means and Standard Devi ations for Early Childhood Attention and Reading Variables Groupa Variable PCE Non-exposed p -value M SD M SD IVA CPT VAQ age 5 b 44.96 30.31 49.12 27.01 .261 IVA CPT VRCQ age 5 b 62.15 37.46 66.72 34.29 .323 Letter Cancellation – age 5 (secs.) 108.31 43.28 99.41 43.15 .112 IVA CPT VAQ age 7c 56.31 18.32 58.04 20.87 .488 IVA CPT VRCQ age 7 c 73.80 25.69 77.97 23.17 .183 Letter Cancellation – age 7 (secs.) 60.22 21.75 59.15 21.86 .705 TMT Part A (secs.) 54.92 51.16 57.2 63.08 .629 WISC-III Coding 10.14 3.62 10.56 3.32 .347 WISC-III Digit Span 8.71 2.79 8.72 2.68 .978 WIAT Broad Reading 96.34 12.04 95.43 12.79 .569 WIAT Reading Comprehension 92.22 11.48 91.20 11.87 .499 Note PCE = prenatal cocaine exposure, IVA CPT = Intermediate Visual and Auditory Continuous Performance Test, VAQ = Visual Attention Quotient, VRCQ = Visual Response Control Quotient, TMT = Trail Making Test, WISC-III = Wechsler Intelligence Scale for Children-Third Edition, WIAT = Wechsler Individual Achievement Test. an = 120 for both groups for all measures except IVA CPT scores. bn = 116 for PCE group, n = 126 for nonexposed group, cn = 123 for both groups.

PAGE 79

69 Table 3-3 Group Means and Standard De viations for All Other Study Variables Groupa Variable PCE Non-exposed p -value M SD M SD Caregiving Environment– age 5 EC HOME Academic Stimulation 4.49 .81 4.41 .93 .456 EC HOME Language Stimulation 6.48 .89 6.19 1.07 .022 EC HOME Learning Stimulation 7.40 2.36 6.59 2.51 .010* EC HOME Variety in Experience 6.39 1.54 6.17 1.31 .237 Caregiving Environment– age 7 MC HOME Active Stimulation 3.80 1.89 3.67 1.77 .581 MC HOME Family Participation 6.38 2.08 6.48 2.10 .719 MC HOME Growth Fostering Materials and Experiences 5.12 1.65 4.45 1.53 .001** Verbal Ability – age 5 WPPSI-R Comprehension 7.87 2.43 7.71 2.57 .574 WPPSI-R Information 7.27 2.45 6.90 2.59 .259 WPPSI-R Similarities 8.92 2.52 8.46 2.42 .155 WPPSI-R Vocabulary 7.82 2.31 7.71 2.23 .698 Verbal Ability – age 7 WISC-III Comprehension 7.86 3.36 8.24 3.25 .383 WISC-III Information 8.08 2.63 7.96 2.55 .723 WISC-III Similarities 8.05 4.32 8.00 4.80 .944 WISC-III Vocabulary 8.15 2.80 8.41 2.61 .462 Note PCE = prenatal cocaine exposure, TMT = Tra il Making Test, WIAT = Wechsler Individual Achievement Test, EC HOME = Early Childhood Home Observation for Measurement of the Environment, WPPSI-R = Wechsler Preschool and Primary Scal e of Intelligence-Revise d, MC HOME = Middle Childhood Home Observation for Measurement of th e Environment, WISC-III = Wechsler Intelligence Scale for Children-Third Edition. an = 120 for both groups. p < .05, ** p < .01, *** p < .001, one-tailed using the independent samples t -test and Bonferroni correction for familywise error rate.

PAGE 80

70Table 3-4 Fit Indices for Nested Sequen ce of Measurement and Structural Models COMPARISON TO M1 measurement model Comparison to previous model Model Step 2 df p GFI CFI NNFI RMSEA RMR 2 df p 2 df p MEASUREMENT MODEL M1. Basic measurement model 376.72 281 .00 .90 .98 .98 .04 .05 Structural Models S1. Hypothesized structural model 536.37 345 .00 .87 .97 .96 .05 .09 159.65 64 .00 S2. Drop path from HOBEL to HEADC 538.30 346 .00 .87 .97 .96 .05 .09 161.58 65 .00 1.931 NS S3. Drop path from HEADC to LCAN5 538.32 347 .00 .87 .97 .96 .05 .09 161.60 66 .00 .021 NS S4. Drop path from HEADC to HOME5 538.77 348 .00 .87 .97 .96 .05 .09 162.05 67 .00 .451 NS S5. Drop path from DSPAN7 to READ7 538.78 349 .00 .87 .97 .96 .05 .09 162.06 68 .00 .011 NS S6. Add path from VERBAL5 to DSPAN7 497.25 348 .00 .88 .98 .97 .04 .07 120.53 67 .00 41.531 .00 S7. Drop path from LCAN5 to DSPAN7 498.36 349 .00 .88 .98 .97 .04 .08 121.64 68 .00 1.111 NS S8. Add path from VERBAL5 to VATTN7 473.27 348 .00 .88 .98 .98 .04 .06 96.55 67 NS 25.091 .00 S9. Add path from SEX to VATTN7 457.53 347 .00 .89 .98 .98 .04 .06 80.81 66 NS 15.741 .00 Note. N = 240. GFI = goodness of fit index, CFI = comparative fit index, NNFI = non-normed fit index, RMSEA = root mean square error of approximation, RMR = root mean square residual, M = measurement model, S = structural model, VATTN7 = age 7 visual attention, HOME5 = age 5 ca regiving environment, HOBEL = Hobel prenatal risk, HOME7 = age 7 caregiving environment, LCan5 = age 5 Letter Cancellation, DSPAN7 = age 7 Digit Span READ7 = age 7 reading, VERBAL5 = age 5 verbal ability, VERBAL7 = age 7 verbal ability.

PAGE 81

71 R2 = .13 R 2 = .07 .93 .29 -.22 .85 .26 -.11 .14 Figure 3-1. Final structural model of the relationship between prenatal cocaine exposure (COCAI NE), other prenatal drug expos ure (ALCOHOL, TOBACCO, and MARIJUANA), Hobel prenatal risk score (HOBEL), sex (SEX), birt h head circumference (HEADC), ag e 5 Letter Cancellation (LCAN 5), age 5 verbal ability (VERBAL5), age 5 caregiving environment (HOME5), age 7 Digit Span (DSPAN7), age 7 visual attention (VATTN7), a ge 7 verbal ability (VERBAL7), age 7 caregiving environment (HOME7), and age 7 reading ability (READ7). The bold lines highlight the effec t of prenatal cocaine exposure (PCE) on age 7 reading, which is mediated throug h birth head circumference and verbal ability. Age 7 visual a ttention was the largest predictor of age 7 reading followed closely by age 7 verbal ability. Interestingly, sex was also a significant predict or of visual attention at age 7, with girls outperforming boys. .14 HOME7 HOME5 R2 = .00 R2 = .70 .41 VERBAL5 VERBAL7 R2 = .86 R2 = .68 R2 = 1.00 HOBEL R2 = 1.00 SEX R2 = 1.00 R2 = 1.00 MARIJUANA R2 = 1.00 TOBACCO HEADC READ7 .52 .47 -.13 -.13 R2 = .00 44 .27 DSPAN7 LCAN5 VATTN7 R2 = .27 R2 = 1.00 COCAINE ALCOHOL R2 = .44

PAGE 82

72 .42 .65 1.00 1.00 poamar .00 .00 1.00 poaalc 1.00 wucfull .00 ALCOHOL MARIJUANA 1.00 sex SEX .00 .33 VATTN7 .58 .83 traila w3cod .65 .70 .82 .82 VERBAL5 .33 .32 wpinf wpcom .57 wpvoc wpsim .51 .63 .66 .59 .39 h5lang h5acad h5vari h5lear .61 .58 .64 .78 HOME5 .61 .70 .49 .63 HOME7 .50 .76 h7grow h7acti h7fami Figure 3-2. Final structural model showing the variances and residuals of the observed variables with their respective factors. 1.00 DSPAN7 dspan7 .00 1.00 LCAN5 lcant5 .00 1.00 hobpre .00 HOBEL TOBACCO COCAINE .78 .76 .80 .65 VERBAL7 .58 .35 w3inf w3com .43 w3voc w3sim .40 poaalc .00 .24 .06 wiatbr wiatrc .87 .97 READ7 1.00 HEADC .00 headc .89 lcant7

PAGE 83

73 CHAPTER 4 DISCUSSION Study Summary In the current study, a group of children ( n = 120) who were prenatally exposed to cocaine, alcohol, tobacco, and marijuana were compared to a matched group of children ( n = 120) who were exposed to alcohol, t obacco, and marijuana but not cocaine. Statistical analyses showed that the group of children with prenatal cocaine exposure (PCE) had significantly higher levels of expos ure to alcohol, tobacco, and marijuana and significantly higher prenat al obstetrical risk scores than their matched co ntrols. The PCE group also had significantly shorter mean gestat ional age and mean head circumference at birth than the matched control group. The composition of the two groups did not differ significantly by ethnicity or se x. Both groups were predominantly African American and were almost equally split between boys and girls. Main Findings Three hypotheses were developed to exam ine the proposed relationships between prenatal cocaine exposure, attention, and reading ability. The first hypothesis, that measures of neonatal attenti on would be significantly rela ted to measures of early childhood attention was not supported. With one exception, neonatal attention as measured by three scales of the Braze lton Neonatal Behavioral Assessment Scale (BNBAS) was not significantl y associated with a variet y of attentional measures administered at ages 5 and 7, including a continuous performance test, short term auditory attention (Dig it Span), or three at tentional tasks involving visual scanning and

PAGE 84

74 visuomotor coordination (Letter Cancel lation, TMT Part A, and WISC-III Coding subtest). The second hypothesis, that children with PCE would perform significantly worse than non-exposed children on measures of early childhood attention and reading ability was also not supported. No significant group differences were found on visual attention indices of a continuous performance test, shor t term auditory atten tion (Digit Span), or three visuomotor attentional tasks (Lette r Cancellation, TMT Part A, and Coding). Support was found, however, for the third hypothesis that PCE would have an indirect effect on reading ability that wa s mediated by head circumference and its subsequent effects of verbal ability and visual attention. Re sults of structural equation modeling on the combined sample of cocai ne-exposed and nonexposed groups showed that PCE had a small negative effect on h ead circumference, w ith higher amounts of exposure associated with smaller head sizes at birth ( = -.13). Head circumference, in turn, predicted age 5 verbal ability ( = .26), which was highly predictive of age 7 verbal ability ( = .93). Age 5 verbal ability was also mo derately related to two different age 7 attention factors, one comprised of Digit Span and the other by Letter Cancellation, TMT and Coding ( s = .52 and .47). Finally, the age 7 visu al attention factor comprised of Letter Cancellation, TMT, and C oding was found to be the larg est predictor age 7 reading ( = .44) followed closely by age 7 verbal ability ( = .41). Overall, the structural equation model accounted for 68% of the variance in age 7 reading ability. Ancillary Findings Between-group post hoc analyses revealed that the two study groups did not differ on verbal ability measures or 6 of 7 measur es of the caregiving environment. The only

PAGE 85

75 significant difference was found on the Growth Fostering Materials and Experiences subscale of the Middle Childhood Home Ob servation for Measurement of the Environment. Contrary to expectations, the difference favored the PCE group, probably because a much larger proportion of these children were in placements away from their biological mothers. Several other findings from the structural equation modeling analysis also deserve comment. First, the small negative ef fect of PCE on head circumference ( = -.13) was very similar to that of prenatal exposure to alcohol and marijuana ( s = -.13 and -.11, respectively). Unexpectedly, prenatal tobacco exposure had a paradoxical effect on head circumference with higher levels of expos ure associated with larger head sizes ( = .14). The similarity of the coefficients for the f our drug exposure variables and the paradoxical positive coefficient for the tobacco exposure variable may be due to multicollinearity among these variables. An alternative appro ach would have been to combine the alcohol, marijuana, and tobacco exposure variables into a single "other drug exposure" variable. A decision was made not to follow this approa ch so that the potential effects of PCE could be compared directly with the effects of exposure to the other drugs measured in the study. More than any of the drug variables, however, sex was the single strongest predictor of head circumfere nce with girls having smalle r heads at birth than boys ( = .22). Sex also independently predicted the ag e 7 visual attention factor with girls outperforming boys ( = .29). None of the drug variables nor head circumference were direct predictors of the age 5 or age 7 attention factors.

PAGE 86

76 Study Findings in the Context of the Literature Hypothesis #1. That the three BNBAS scores us ed in the current study were not predictive of early childhood a ttention measures is not w holly surprising. First, the majority of studies that have assessed the predictive validity of infant attention for later cognitive abilities have no t utilized the BNBAS. Mo st studies have employed experimental habituation paradigms that are not widely used in the clinical assessment of infants. Thus, while useful as a broad-based clinical tool for assessing the adaptation of infants to their extrauterine environment, th e BNBAS and its specific cluster scores may not have the sensitivity or specificity needed for research applications. Second, the BNBAS has typically been used to predict genera l cognitive abilities, such as IQ or language development, rather th an attention in partic ular. While there are a handful of studies that have begun to trace the developmental trajectory of attention from infancy to childhood, none have used the BNBAS. For the studies reviewed in Chapter 1, two experimental paradigms, fixation dur ation and exposure time needed to meet criterion during a visual recogni tion memory task, were the meas ures used to predict later attentional abilities, respectively (Sigman et al., 1991, Rose & Feldman, 1995). Thus, the attempt in the current stu dy to predict childhood attenti onal performance using the BNBAS was a venture into rela tively unchartered territory. Third, studies comparing infant and earl y childhood performance on attentional or other cognitive tasks generally do not use ne onates. In the review by Colombo (1993), the age at the first assessment ranged from 3 to 9 months. The reason may be that scores based on assessments of newborns using m easures such as the BNBAS are likely to reflect the state of the child at the time of the assessment rath er than a more enduring trait that is not expected to ch ange significantly over time. Using the BNBAS in newborns,

PAGE 87

77 then, is somewhat akin to using a state m easure to predict a childhood trait (attention). Such an undertaking is less lik ely to produce significant resu lts, especially considering that attention is not a stable attribute in childhood but c ontinues to develop with age (Cooley & Morris, 1990). Finally, the hypothesis that BN BAS scores would be predic tive of later attentional measures was based, in part, on group di fferences on the BNBAS reported in the literature between children with PCE and non-exposed children. A meta-analysis of group differences on the BNBAS between child ren with PCE and non-exposed children has revealed that the larges t reliable differences were for motor performance and abnormal reflexes and not for the Orientation, Habituation, or Regulation of State scores (Held, Riggs, & Dorman, 1999). It was also found that while the Orientation and Autonomic Regulation clusters produced small, significant effects at birth and at 3 to 4 weeks of age, the effects were small, decrea sed over time, and were likely due to a large sample size (Held, Riggs, & Dorman, 1999). T hus, attentional diffe rences found between infants with PCE and comparison groups in so me studies did not hold up across samples. In sum, considering the lack of reliable between-group differences on the BNBAS for cocaine-exposed and non-exposed child ren, combined with the three BNBAS measurement issues just outlined, it would ha ve been quite remarkable if the three BNBAS scores used in the present study were significantly correla ted with attention measures at age 5 and 7. Hypothesis #2. The lack of significant group di fferences on attention and reading measures between children with PCE and nonexposed children in the current study is consistent with the majority of the literatu re in this area. In a recent review, Frank,

PAGE 88

78 Augustyn, Knight, Pell, and Zuckerman (2001) concluded that after controlling for confounding factors, PCE has no consistent ne gative association with physical growth, developmental test scores, or language skil ls. However, it should be noted that the majority of studies included in the revi ew examined outcomes for children age 3 or younger. Only two other groups of researchers have published data on children with PCE over age 3 using assessment instruments simila r to those in the current study. Hurt et al. (1997) found no group differences at age 4 betw een a Philadelphia inner-city sample of children with PCE and non-exposed childre n on the Wechsler Preschool and Primary Scale of Intelligence-Revised (WPPSI-R) mean Verbal, Performance, or Full Scale IQ scores. Richardson, Conroy, and Day ( 1996), based in Pittsburgh, found no group differences at age 6 between children with PCE and non-exposed children for any scores on the Stanford-Binet Intelligence ScaleFourth Edition (Thorndike, Hagen, & Sattler, 1986) or on the Reading, Spelling, or Ar ithmetic subtests of the Wide Range Achievement Test-Revised (Jastak & Wilkinson, 1984). As in the current study, the Pittsburgh research team found that both gr oups of children genera lly scored in the average range on IQ and achievement measures (Richardson et al., 1996). Three published studies of a ttentional abilities of sc hool-age children with PCE and non-exposed children were found in the extant literature. The first two, from the same group of researchers (Richardson et al ., 1996; Leech et al., 1999). In the first study, the authors compared exposed and nonexposed groups the using a continuous performance task (CPT) involving shapes and colors that is not widely available for commercial use. In their sample, the 6-year-old s with PCE made significantly more errors of omission across all three bl ocks of trials than did nonexposed children; however, no

PAGE 89

79 group differences were found for errors of co mmission (Richardson et al., 1996). It was later reported that among those excluded from the analyses were 13 children who did not complete the test to due "impulse control and attention problems" and an unspecified number of children refused to do the task (Leech et al., 1999). The proportion of these children who had PCE and thei r characteristics are not repo rted. In a follow-up study, the authors found that PCE during first trimester predicted more errors of omission; however, marijuana use during second trimester and tobacco use during the second and third trimesters were also predictive of more er rors of omission (Leech et al., 1999). For all drug exposure variables, the regression coeffi cients were small ranging from .09 to .10, raising the question about whethe r these effects are meaningful in terms of the everyday functioning (e.g., school performanc e) of children with PCE. In the third study, Bandstra, Morrow, Anthony, Accornero, and Fried (2001) conducted a longitudinal investig ation of attention in school-a ge children with PCE using two different CPT tasks. They used the Test of Variables of Atte ntion (TOVA; Greenberg et al., 1996) at age 5 and th e Conners’ Continuous Performa nce Test (CCPT; Conners, 1995) at age 7. Using omission errors as the criterion variable, th ey found that age 5 scores predicted age 7 scores. While the noncocaine-exposed cont rol group performed better on both measures than th e children with PCE, between -groups statistical analyses were not conducted to determine whether th e differences were significant. A between groups analysis, combining the two CPT measur es with an age 3 "time on task" measure of attention, revealed that the estimated group difference was signi ficant. Hierarchical models that included birth outcome measur es such as head circumference did not significantly attenuate the between-groups difference. Us ing SEM, the authors also found

PAGE 90

80 that the amount of PCE signifi cantly predicted omission error scores at age 7 even after controlling for other prenatal drug expos ure, child’s age, and child’s sex ( = .26) (Bandstra et al., 2001). The current study found no differences between children with PCE and nonexposed children for two indices of visual attention on the Intermediate Visual and Auditory Continuous Performance Test (IVA CPT). Between-group differences on errors of omission and errors of commission were not examined but such data are available for the sample used in the current study and could be the focus of a future study. Hypothesis #3 -Relationship be tween PCE and head circumference. The finding that the relationship between PCE and cognitive variables is mediated, in part, by head circumference is consiste nt with the published literature. Currently, only four other studies have examined the relationships between PCE, head circumference, and developmental outcomes. Two of these studi es focused on 24-month outcomes, and two studies examined 36-month outcomes. In the first study, Chasnoff et al. (1992) found significant correlations between head circumfere nce at various ages up to 24 months and Mental and Psychomotor indices of the Bayl ey Scales of Infant Development (Bayley, 1969). While prenatal exposure to cocaine alcohol, tobacco, and marijuana all contributed to head circumference measurem ents during the first 2 years, only cocaine was a significant predictor as a single variable in their Chicago-ba sed sample (Chasnoff et al., 1992). Another group of researchers found that he ad circumference at birth is correlated with BSID scores at 6.5, 12, and 24 months of age in a mixed sample of cocaine-exposed and non-exposed children ( N = 415) (Singer et al., 2002) Significant correlation

PAGE 91

81 coefficients between head circumference a nd BSID Index scores ranged from .12 to .22 (Singer et al., 2002). Using step wise regression to develop a model predicting scores at 24 months, it was found that the negative eff ects of cocaine on c ognitive outcome were mediated through smaller head circumference at birth (Singer et al., 2002). No other measures—gestational age, birth weight, birth length, Apgar scores, or the Hobel Neonatal Risk score—mediated the effects of cocaine on BSID scores obtained at 24 months (Singer et al., 2002). In the third study, Azuma and Chasnoff (1993) used path analysis to determine whether head circumference was a significant predictor of age 3 developmental outcome assessed using the Stanford-Binet Intelligen ce Scale-Fourth Editi on (SBIS; Thorndike, Hagen, & Sattler, 1986). While head circumfere nce at age 3 was not found to be a direct predictor of the SBIS composite IQ score at age 3, it did have an indirect effect on composite IQ that was mediated by poor pers everance as measured by a combination of five-point behavior ra ting scales completed by the bli nded examiners who administered the SBIS (Azuma & Chasnoff, 1993). The regres sion coefficients were -.30 between head circumference and perseverance and -.60 between perseverance and composite IQ (Azuma & Chasnoff, 1993). The fourth study to examine head circ umference as a predictor variable was conducted by principal investig ators of the grant from whic h the sample for current study was drawn. Using structural equation mode ling, it was found that age 3 developmental outcome measured by a factor comprised of BSID scores and four subtests of the Vineland Adaptive Behavior Scales (VABS; Sparrow, Balla, & Cicchetti, 1984) was predicted by birth head circumference ( = .14), which itself was predicted by prenatal

PAGE 92

82 cocaine exposure ( = -.18) after controlling for pren atal alcohol and tobacco exposure (Eyler, Behnke, Garvan, Wobie, & Hou, 2002). While interpretation is made somewhat more difficult by the study's use of an outcome variable that combined scores from both a child behavioral measure (BSID) and a caregiv er report measure (VAB S), it is clear that head circumference was a mediator between PCE and overall developmental outcome. The results of the current study and the f our studies just reviewed suggest that head circumference may serve as a proxy fo r the effects of PCE on in utero brain development. While three different groups of researchers have found significant relationships between head circumference and BSID scores, these stud ies do not elucidate the mechanism underlying the relationship. In the Azuma and Chasnoff (1993) study, however, perseverance was the factor linking head circumfe rence to Stanford-Binet composite IQ. In the current study, head circ umference predicted verbal ability, verbal ability predicted visual attention, and visual attention was found to be the most significant predictor of reading performa nce. If sustained effort (perseverance) is considered somewhat analogous to sustained attention, th en the results of the Azuma and Chasnoff (1993) study and the current study arguably converge to s uggest that PCE's effect on head circumference may be related to probl ems with self-regulation that can affect cognitive performance. These findings warrant further studies using head circumference, particularly at birth, as a pred ictor variable in future stud ies of developmental outcomes of children with PCE. Hypothesis #3 Relationship Between Attention and Reading. While there are no current published studies on the relationship between attention and reading in children with PCE, the finding that attention is strongly related to reading ability is consistent with

PAGE 93

83 other longitudinal studies using samples of elementary school children. Using path analysis, Rabiner, Coie, and The Conduct Problems Prevention Research Group (2000) found that teacher ratings of inattentiveness in second grade had a regression coefficient of -.10 with fifth-grade reading achievement in a combined sample of both "at-risk" and "normal" children. Using structural equati on modeling and a much larger sample of children drawn from the normal school popul ation, Rowe and Rowe (1992) found that teacher ratings of inattentiveness had a str ong negative influence across four age groups with regression coefficients ranging from -.21 to -.39. The much larger regression coefficients in the current study may be the re sult of using direct be havioral measures of attention rather th an teacher ratings. Overall, the results of the current study suggest that th e effects of PCE, if any, on school-age outcomes such as attention and read ing are subtle and diffi cult to detect using traditional pencil and paper assessment met hods. The effects of PCE that can be identified are of similar magnitude as those as sociated with prenatal exposure to alcohol and tobacco, which are used by pregnant wome n more often and in larger quantities. Distinguishing the effects of prenatal dr ug exposure on later childhood development is complicated by the fact that children born to mothers who abused substances during pregnancy are subject to other multiple risk factors for poor outcome, including poverty. Studies on children with PCE, including the curren t one, seem to point to the resiliency of humans to develop normally despite multiple risk factors rather than the strength of the teratogenic model to pr edict negative outcomes. Study Strengths The strengths of the current study result from the nature of the sample, control for potentially confounding variable s, and the use of structural equation modeling. The

PAGE 94

84 sample is distinguished from that of othe r prospective, longitudinal studies in that examiners were blinded to the children's cocai ne exposure status and the attrition rate was very low (approximately 10%). In the Frank et al. (2001) review, th e authors excluded 20 studies because they failed to guard agai nst examiner bias by masking the cocaine exposure status of the children. The distortio ns that can result from examiner bias, particularly in behavior al research, are well-kno wn (Kazdin, 1980). Moreover, researchers have documented specific negative examiner bias with children labeled as prenatally exposed to cocaine (Thurman, Brobeil, & Ducette; 1994; Woods, Eyler, Conlon, Behnke, & Wobie, 1998). Obtaining data from blinded examiners has not been a universal feature of studies of children with PCE and makes the results of the current study more trustworthy. Careful analysis of the characteristics of the sample and how it compares to the originally enrolled sample is another impor tant feature of the current study. For the studies included in the Frank et al. (2001) review, retention rates ranged from 39% to 94%, characteristics of thos e lost to follow-up were not reported for 6 of the 17 independent cohorts, and four studies failed to report attrition at all (Frank et al., 2001). A wide variety of factors can affect which pa rticipants remain in longitudinal studies, including the severity of th e problems being studied. Thus in order to ascertain the potential longer-term effects of any variable, it is critical to unders tand the nature of the sample and how it changes over time. The samp le in the current study represented 78% of the original sample, including the children w ho died, and replicated the original group differences in prenatal drug exposure, Hobel pr enatal risk, and birth head circumference. The current study groups were also similar to the original study groups in that they did

PAGE 95

85 not differ in terms of ethnicity or sex. The one statistically significant difference between the current study groups that was not found in the original st udy groups—gestational age—was less than one week and did not re present a significant clinical difference. A third strength of the current study is control for, and comparison of, confounding variables. Failure to control for confounding variables, pa rticularly exposure to other drugs of abuse, is one of the most prevalent methodological fl aws in the literature on children with PCE. The original st udy groups were matched for demographic variables including ethnicity, and the demographic similari ties between the groups were maintained in the current study sample, thereby controlling for these variables. Using separate factors for prenatal exposure to cocaine, alcohol, tobacco, and marijuana in the structural equation model allowed direct comp arison of the relative effect of each drug. These direct comparisons are important because they help to contextu alize the effects of PCE in terms of alcohol and tobacco—drugs that are more familiar to health professionals and the public and are mo re commonly used by pregnant women. Considering that women who use cocaine during their pregnancies are, in some states, currently subject to punitive consequences (while women who use alcohol and tobacco are not), the ability to make comparisons between various drugs of abuse could have important policy implications. The use of structural equation modeli ng is not yet common in research on children with PCE and represents a significant contribution to the l iterature. Structural equation modeling is a powerful statistical t echnique that enables researchers to pose and test hypotheses that cannot easily be evaluated using more traditional statistical methods. In the current study, no statis tically significant differences were found between children

PAGE 96

86 with PCE and non-exposed children in a cr oss-sectional analysis using traditional statistical techniques. However, the rela tionship between atten tion and reading and between PCE and attention could still be expl ored using structural equation modeling. Study Limitations Measurement issues. One of the primary limitations of the current study relates to the methods used to measure attention. Attention is a complex, multi-dimensional process that involves a number of distinct areas of the brain, depending on the nature of the task. In the current study attention was measured using one auditory task and three pencil and paper tasks: Digit Span, Letter Cancellation, TMT Part A, and Coding. Since the Digit Span factor did not turn out to be a significant predic tor of reading, this discussion will focus on the latter three tasks. Letter Cancellation, TMT Part A, and Coding were combined into a factor labeled "visual attention;" however, each of these tasks require much more than "attention" to be completed skillfully and efficiently. All th ree tasks involve visual scanning, have a significant fine motor component, and require fa miliarity with letters or numbers or both. Moreover, all three tasks are timed, and sc ores are based primarily on speed with a secondary emphasis on accuracy. Arguably, then the factor comprised of these tasks could have been called a "processing speed" meas ure rather than an "attention" measure. Additionally, it should be noted that there was some vari ability in how the children performed on each of the tests. While both gr oups of children scored in the average range on Coding, mean scores for TMT Part A for bot h groups of children were more than 1.5 standard deviations below published norms. This highlights the fact that these tests may tap different aspects of the multidim ensional construct called attention.

PAGE 97

87 Generalizability issues The participants in the current study were drawn from rural north central Flor ida. By contrast, the majority of the prospective longitudinal studies on the effects of PCE funded by the National Institut e on Drug Abuse are based in large urban centers, namely Atlanta, Baltimor e, Chicago, Cleveland, Detroit, Miami, and Pittsburgh. This significant geographical differe nce could affect a variety of factors that impact child developmental outcome, ranging from the amount and quality of drugs used by the mothers to the amount and quality of social supports available to mothers and intervention programs available to the child ren. Although the overwhelming majority of the participant families in the NIDA-funded studies are poor, poverty is likely to be experienced quite differently in a rural area th an in an urban area. The potential impact that a rural versus an urban setting may ha ve on developmental outcomes of children with PCE remains an empirical question that warrants further research. Future Directions Several of the suggestions for helpin g to elucidate the relationships found between PCE, attention, and r eading in the current study rela te to measurement issues. First, no significant relationship was found between the Digit Span factor and the Reading factor, and this may be due to th e fact that scaled scores that combine performance on Digits Forward and Digits Backward were used. Digits Forward is traditionally considered an attentional task, while Digits Backward, which requires manipulating the information presented, is c onsidered a working memory task and more difficult than Digits Forward. Two children c ould obtain identical sc aled scores on the combined Digit Span task, with considerably different performances on Digits Forward vs. Digits Backward. Thus, using raw scores for Digits Forward and Digits Backward

PAGE 98

88 may have revealed significant between-group differences or resulted in significant predictive relationships between Digits Forward or Digits Backward and reading ability. Second, as discussed earlier, the Visual A ttention factor was comprised of three tasks that involved both visual scanning and visuomotor processing speed. To assess the relative importance of these tw o sets of component skills to reading ability, a task that relies primarily on motor speed, such as th e Finger Tapping Test from the HalsteadReitan Neuropsychological Test Battery (Reita n & Wolfson, 1985) or mean reaction time on the IVA CPT (Sandford, 1995), could have been used as a covariate in the analyses. Although no consistent negative effects of PC E have been found for motor development (Frank et al., 2001; Singer et al, 2002), analys is of scores on motor and visual-motor tasks could help to clarify the variability in performance across the different measures of visual attention and how they may affect reading skills. Third, future studies should examine singl e-word reading skills separately from reading comprehension. In the current study, the Reading factor wa s comprised of two subtests, Basic Reading and Reading Compre hension, that assess different aspects of reading ability. The Basic Reading subtest is largely a measure of single-word reading and relies on sight word recognition for familiar words and phonological decoding skills for unfamiliar words. The other subtest, R eading Comprehension, involves reading short passages and answering questions posed by th e examiner. It may be that the visual scanning aspects of the Visual Attenti on factor, comprised of Coding, Letter Cancellation, and TMT Part A, may be more related to single word reading while the working memory aspects of the Digit Span factor and the general information knowledge and vocabulary skills captured in the Verbal Ability factor would be more associated with

PAGE 99

89 reading comprehension, which requires higher orde r synthesis of material as it is read. In short, more detailed analysis of which as pects of attention map onto which aspects of reading could be the focus of future studies. Fourth, it is likely that there are subg roups among the children with PCE who differ in ways that may affect outcome. For example, there was some variability in the amount and timing of cocaine exposure am ong the children with PCE. By combining children with lowand high-exposure or chil dren with predominantly first-trimester versus third-trimester expos ure in the analyses, signifi cant subgroups differences may have been diluted or even eliminated. A nother example of poten tially important subgroup differences has to do with exposure to multiple teratogens. At least one animal study has reported that maternal use of alcohol and co caine during pregnancy has more detrimental effects on offspring outcome than use of e ither drug alone (Randall, Cook, Thomas, & White, 1999). Since human maternal cocaine us e during pregnancy most often occurs in conjunction with other drugs of abuse, studi es of the coteratology of cocaine with alcohol, tobacco, and marijuana should beco me the research wave of the future.

PAGE 100

90 REFERENCES Adams, M. A. (1994). Beginning to read: Thinki ng and learning about print Cambridge, MA: MIT Press. Akbari, H. M., Kramer, H. K., Whitaker-Azmiti a, A. P., Spear, L. P., & Azmitia, E. C. (1992). Prenatal cocaine e xposure disrupts the development of the serotonergic system. Brain Research, 572 57-63. American Psychiatric Association. (1994 ). Diagnostic and Statistical Manual of Mental Disorders (4th ed.). Washington, DC: Author. Azuma, S. D., & Chasnoff, I. J. (1993). Outcom e of children prenatally exposed to cocaine and other drugs: A path anal ysis of three-year data. Pediatrics, 92 (3), 396-402. Bandstra, E. S., Morrow, C. E., Anthony, J. C., Accornero, V. H., & Fried, P. A. (2001). Longitudinal investigation of task persistence and sustained attention in children with prenatal cocaine exposure. Neurotoxicology and Teratology, 23 545-559. Bayley, N. (1969). Bayley Scal es of Infant Development. San Antonio, TX: Psychological Corp. Beckwith, L., Crawford, S., Moore, J. A., & Howard, J. (1995). Attentional and social functioning of preschool-age children exposed to PCP and cocaine in utero. In M. Lewis & M. Bendersky (Eds.), Mothers, babies, and cocaine: The role of toxins in development (pp. 287-303). Hillsdale, NJ: Erlbaum. Behnke, M., Eyler, F. D., Conlon, M., Wobie, K., Woods, N. S., & Cumming, W. (1998). Incidence and description of structural br ain abnormalities in newborns exposed to cocaine. Journal of Pediatrics, 132 291-294. Behnke, M., Eyler, F. D., Garvan, C. W., W obie, K., & Hou, W. (2002). Cocaine exposure and developmental outcome from birth to 6 months Neurotoxicology and Teratology, 24 283-295. Behnke, M., Eyler, F. D., Woods, N. S., Wobi e, K., & Conlon, M. (1997). Rural pregnant cocaine users: An in-depth sociodemographic comparison. Journal of Drug Issues, 27 501-524. Bentler, P. M. (1990). Comparative fit indices in structural models. Psychological Bulletin, 107 238-246.

PAGE 101

91 Bentler, P. M., & Bonett, D. G. (1980). Signifi cance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin, 99 588-606. Bornstein, M. H. (1990). Attention in infancy and the prediction of cognitive capacities in childhood. In J. T. Enns (Ed.), The development of attention: Research and theory (pp. 3-20). New York: Elsevier. Bradley, R. H. (1994). The HOME Inventory: Review and refl ections. In H. Reese (Ed.), Advances in child development and behavior (pp. 241-288). San Diego: Academic Press. Bradley, R. H., & Caldwell, B. M. (1981). The HOME Inventory: A validation of the preschool scale for Black children. Child Development, 52 708-710. Bradley, R. H., Caldwell, B. M., & Rock, S. L. (1988). Home environment and school performance: A ten-year follow-up and examin ation of three models of environmental action. Child Development, 59 852-867. Bradley, R. H., Caldwell, B. M., Rock, S. L ., Hamrick, H. M., & Harris, P. (1988). Home Observation for Measurement of the Environment: Development of a HOME inventory for use with families having children 6 to 10 years old. Contemporary Educational Psychology, 13 58-71. Bradley, R. H., Mundfrom, D. J ., Whiteside, L., Caldwell, B. M ., Casey, P. H., Kirby, R. S., & Hansen, S. (1994). The demography of parenting: A re-examination of the association between HOME scores and income. Nursing Research, 43 260-266. Bradley, R. H., Rock, S. L., Caldwell, B. M., Harris, P. T., & Hamrick, H. M. (1987). Home environment and school performance am ong Black elementary school children. Journal of Negro Education, 56 499-509. Bradley, R. H., & Whiteside-Mansell, L. (1998). Home environment and children's development: Age and demographic differen ces. In M. Lewis & C. Feiring (Eds.), Families, risk, and competence (pp. 133-157). Mahwah, NJ: Erlbaum. Brazelton, T. B. (1984). Neonatal behavioral assessment scale Philadelphia: Lippincott. Brazelton, T. B. (1994). Neonatal behavioral assessment scale (2nd ed.). London :Spastics International Medi cal Publications. Brazelton, T. B., Nugent, J. K., & Lester, B. M. (1987). Neonatal Behavioral Assessment Scale. In J. D. Osofsky (Ed.), Handbook of infant development (pp. 780-817). New York: Wiley. Brock, S. E., & Knapp, P. K. (1996). Reading comprehension abilities of children with Attention-Deficit/Hyperactivity Disorder. Journal of Attention Disorders, 1 173-185.

PAGE 102

92 Browne, M. W., & Cudeck, R. (1993). Bootstra pping goodness-of-fit measures in structural equation models. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 111-135). Newbury Park, CA: Sage. Bryant, F. B., & Yarnold, P. R. (1995). Principal-components and exploratory and confirmatory factor analysis. In L. G. Grimm & P. R. Yarnold (Eds.), Reading and understanding multivariate statistics (pp. 99-136). Washington, DC: American Psychological Association. Caldwell, B. M., & Bradley, R. H. (1984). Home observation for measurement of the environment Little Rock, AR: University of Arkansas Press. Chasnoff, I. J., Anson, A., Hatcher, R., St enson, H., Iaukea, K., & Randolph, L. A. (1998). Prenatal exposure to cocaine and other drugs: Outcome at four to six years. In J. A. Harvey & B. E. Kosofsky (Eds.), Cocaine: Effects on the developing brain (pp. 314328). New York: New York Academy of Sciences. Chasnoff, I. J., Griffith, D. R., Freier, C ., & Murray, J. (1992). Cocaine/polydrug use in pregnancy: Two year follow-up. Pediatrics, 89 284-289. Church, M. W., Crossland, W. J., Holmes, P. A., Overbeck, G. W., & Tilak, J. P. (1998). Effects of prenatal cocaine on hearing, visi on, growth and behavior. In J. A. Harvey & B. E. Kosofsky (Eds.), Cocaine: Effects on the Developing Brain (pp. 12-28). New York: New York Academy of Sciences. Cohen, R. A. (1993). The neuropsychology of attention New York: Plenum Press. Colombo, J. (1993). Infant cognition: Predicting later intellectual functioning Newbury Park, CA: Sage. Conners, C. K. (1995). Conners’ Continuous Performance Test (CPT), 2nd ed. Toronto, Canada: Multi-Health Systems. Cooley, E. L., & Morris, R. D. (1990). Attent ion in children: A ne uropsychologically based model for assessment. Developmental Neuropsychology, 6 239-274. Day, N. L., Wagener, D. K., & Taylor, P. M. (1985). Measurement of substance use during pregnancy: Methodologic issues In T. M. Pinkert (Ed.) Current research on the consequences of maternal drug abuse. NIDA research monograph 59 36-47. Diller, L., Ben-Yishay,Y., Gerstman, L. J., Goodin, R., Gordon, W., & Weinberg, J. (1974). Studies of scanning behavior in hemiplegia. Rehabilitation monograph No. 50: Studies in cognition and re habilitation in hemiplegia New York: New York University Medical Center Institute of Rehabilitation Medicine. Dow-Edwards, D. L. (1995). Developmental toxi city of cocaine: Mech anisms of action. In M. Lewis & M. Bendersky (Eds.), Mothers, babies, and cocaine: The role of toxins in development (pp. 5-17). Hillsdale, NJ: Erlbaum.

PAGE 103

93 Dubowitz, L. M. S., Dubowitz, V. & Goldberg. C. (1970). Clinical assess ment of gestational age in the newborn infant. Journal of Pediatrics, 77 1-10. Dunn, L. M., Dunn, L. M., Robertson, G. J., & Eisenberg, J. L. (1981). Peabody Picture Vocabulary Test-Revised. Circle Pines, MN: AGS/American Guidance Service. DuPaul, G. (1991). Parent and teacher rati ngs of ADHD symptoms: Psychometric properties in a community-based sample. Journal of Child and Adolescent Psychopharmacology, 20 245-253. Edelbrock, C. (1990). The Child Attention Pr oblems Scale. In R. A. Barkley (Ed.), Attention Deficit Hyperactivity Disorder: A handbook for diagnosis and treatment (pp. 302305). New York: Guilford. Enns, J. T. (Ed.) (1990). The development of attention: Research and theory New York: Elsevier. Eyler, F. D., Behnke, M., Conlon, M., Woods, N. S., & Wobie, K. ( 1998a). Birth outcome from a prospective, matched study of prenat al crack/cocaine use: I. Interactive and dose effects on health and growth. Pediatrics, 101 229-237. Eyler, F. D., Behnke, M., Conlon, M., Woods, N. S., & Wobie, K. (1998b). Birth outcome from a prospective, matched study of prenat al crack/cocaine use: II. Interactive and dose effects on neurobeha vioral assessment. Pediatrics, 101 237-241. Eyler, F. D., Behnke, M., Garvan, C. W., Wobie, K., & Hou, W. (2002, November). Analysis of the direct and indirect effects of prenatal co caine exposure on 36-month developmental outcome Poster session presented at th e 22nd annual conference of the National Academy of Neuropsychology, Miami, FL. Eyler, F. D., Behnke, M., Garvan, C. W., W oods, N. S., Wobie, K. & Conlon, M. (2001). Newborn evaluations of toxic ity and withdrawal related to prenatal cocaine exposure. Neurotoxicology and Teratology, 23 399-411. Factor, E. M., Hart, R. P., & Jonakait, G. M. (1993). Neurochemical development of the raphe after continuous pren atal cocaine exposure. Brain Research Bulletin, 31 49-56. Felton, R. H., & Wood, F. B. (1989). Cognitive de ficits in reading di sability and Attention Deficit Disorder. Journal of Learning Disabilities, 22 3-13. Fergusson, D. M., & Horwood, J. L. (1992). A ttention deficit and r eading achievement. Journal of Child Psychology and Psychiatry, 33 375-395. Ferro, J. M., Martins, I. P., & Ta vora, L. (1984). Neglect in children. Annals of Neurology, 15 281-284.

PAGE 104

94 Frank, D. A., Augustyn, M., Knight, W. G., Pell, T. & Zuckerman, B. (2001). Growth, development, and behavior in early chil dhood following prenatal cocaine exposure: A systematic review. Journal of the American Medical Association, 285 1613-1625. Frank, D. A., Augustyn, M., & Zuckerman, B. S. (1998). Neonatal neurobehavioral and neuroanatomic correlates of prenatal cocain e exposure. In J. A. Harvey & B. E. Kosofsky (Eds.), Cocaine: Effects on the developing brain (pp. 40-50). New York: New York Academy of Sciences. Friedman, E., & Wang, H.-Y. (1998). Prenatal cocai ne exposure alters signal transduction in the brain D1 dopamine receptor system. In J. A. Harvey & B. E. Kosofsky (Eds.), Cocaine: Effects on the developing brain (pp. 238-247). New York: New York Academy of Sciences. Gabriel, M., & Taylor, C. (1998) Prenatal exposure to cocaine impairs neuronal coding of attention and discriminative learning. In J. A. Harvey & B. E. Kosofsky (Eds.), Cocaine: Effects on the developing brain (pp. 194-212). New York: New York Academy of Sciences. Gilger, J. W., Pennington, B. F., & DeFries, J. C. (1992). A twin st udy of the etiology of comorbidity: Attention Deficit-Hyperactivity Disorder and dyslexia. Journal of the American Academy of Child and Adolescent Psychiatry, 31 343-348. Graham, F. K. (1992). Attention: The heartbeat, the bl ink, and the brain. In B. A. Campbell, H. Hayne, & R. Richardson (Eds.), Attention and information processing in infants and adults: Perspectives from human and animal research Hillsdale, NJ: Erlbaum. Grant, D. A., & Berg, E. A. (1993). Wisconsin Card Sorting Test San Antonio, TX: Psychological Corp. Greenberg, L., Leark, R., Dupuy, T., Corman, C., Kindschi, C., & Cenedela, M. (1996 ). Test of Variables of Attention (TOVA) Los Alamitos, CA: Universal Attention Disorders. Griffith, D. R., Azuma, S. D., & Chasnoff, I. J. (1994). Three-year outcome of children exposed prenatally to drugs. Journal of the American Academy of Child and Adolescent Psychiatry, 33 20-27. Hale, T. S., Hariri, A. R., & McCracken, J. T. (2000). Attention-Deficit/Hyperactivity Disorder: Perspectives from neuroimaging. Mental Retardation and Developmental Disabilities Research Reviews, 6 214-219. Hartmann, D. P., & George, T. P. (1999). Design, measurement, and analysis in developmental research. In M. H. Bornstein & M. E. Lamb (Eds.), Developmental Psychology: An advanced textbook (pp. 125-198). Mahwah, NJ: Erlbaum. Harvey, J. A., & Kosofsky, B. E. (Eds.) (1998). Cocaine: Effects on the developing brain New York: New York Academy of Sciences.

PAGE 105

95 Heilman, K. M., Watson, R. T., & Valenstein, E. (1993). Neglect and related disorders. In K. M. Heilman & E. Valenstein (Eds.), Clinical neuropsychology (pp. 279-336). New York: Oxford University Press. Held, J. R., Riggs, M. L., & Dorman, C. (1999). The effect of prenatal cocaine exposure on neurobehavioral outcome: A meta-analysis. Neurotoxicology and Teratology 21 619625. Hendren, R. L., DeBacker, I., & Pandina, G. J. (2000). Review of neuroimaging studies of child and adolescent psychiatric di sorders from the past 10 years. Journal of the American Academy of Child and Adolescent Psychiatry, 39 815-828. Hinshaw, S. P. (1992). Externalizing behavior problems and academic underachievement in childhood and adolescence: Causal rela tionships and underlying mechanisms. Journal of Consulting and Clinical Psychology, 59 289-294. Hobel, C. J., Hyvarinen, M. A., Okada, D. M ., & Oh, W. (1973). Pren atal and intrapartum high risk screening: I. Predic tion of the high-risk neonate. American Journal of Obstetrics and Gynecology, 117 1-9. Hobel, C. J., Youkeles, L., & Forsythe, A. (1979). Prenatal and intrapartum high risk screening: II. Risk factors assessed. American Journal of Obst etrics and Gynecology, 135 1051-1056. Hollingshead, A. B. (1995). Four factor index of social status Unpublished manuscript. Hurt, H., Malmud, E., Betancourt, L., Braitma n, L. E., Brodsky, N. L., & Giannetta, J. (1997). Children with in uter o cocaine exposure do not differ from control subjects on intelligence testing. Archives of Pediatrics and Adolescent Medicine, 151 1237-1241. James, W. (1890). The principles of psychology (Vol. 1). New York: Holt. Jastak, S., & Wilkinson, G. S. (1984). Wide Range Achievement Test-Revised. Wilmington, DE: Jastak Associates/Wide Range. Javorsky, J. (1996). An examination of youth w ith Attention-Deficit/Hyperactivity Disorder and language learning disabi lities; A clinical study. Journal of Learning Disabilities, 29 247-258. Jreskog, K. G., & Srbom, D. G. (1989). LISREL 7 user's reference guide Chicago: Scientific Software. Jreskog, K. G., & Srbom, D. G. (2 001a) [Computer software]. LISREL 8.52. Lincolnwood, IL: Scientific Software. Jreskog, K. G., & Srbom, D. G. (2 001b) [Computer software]. PRELIS 2.52. Lincolnwood, IL: Scientific Software.

PAGE 106

96 Kaufman, A. S., & Kaufman, N. L. (1983). Kaufman Assessment Battery for Children Circle Pines, MN: AGS/American Guidance Service. Kaye, D. B., & Ruskin, E. M. (1990). The develo pment of attentional control mechanisms. In J. T. Enns (Ed.), The development of attention: Research and theory (pp. 227-244). New York: Elsevier. Kazdin, A. S. (1980). Research design in clinical psychology New York: Harper & Row. Kliegman, R. M., & King, K. C. (1983). Intraute rine growth retardation: Determinants of aberrant fetal growth. In A. A. Fanaroff & R. J. Martin (Eds.), Berhman's NeonatalPerinatal Medicine St. Louis, MO: C. V. Mosby. Kosofsky, B. E., & Wilkins, A. S. (1998). A mouse model of tran splacental cocaine exposure: Clinical implications for exposed in fants and children. In J. A. Harvey & B. E. Kosofsky (Eds.), Cocaine: Effects on the developing brain (pp. 248-261). New York: New York Academy of Sciences. Leckliter, I. N., Forster, A. A., & Klonoff, H ., Knights, R. M. (1992). A review of reference group data from normal children for the Ha lstead-Reitan Batter y for older children. The Clinical Neuropsychologist, 6 201-229. Leech, S. L., Richardson, G. A., Goldschmidt, L., & Day, N. L. (1999). Prenatal substance exposure: Effects on attention a nd impulsivity of 6-year-olds. Neurotoxicology and Teratology, 21 109-118. Leslie, C. A., Robertson, M. W., Jung, A. B., Liebermann, J., & Bennett, J. P., Jr. (1994). Effects of prenatal cocaine exposure upon postnatal development of neostriatal dopaminergic function. Synapse, 17 210-215. Lester, B. M., Freier, K., & LaGasse, L. (1995). Prenatal cocaine exposure and child outcome: What do we really know? In M. Lewis & M. Bendersky (Eds.), Mothers, babies, and cocaine: The role of toxins in development (pp. 19-39). Hillsdale, NJ: Erlbaum. Lester, B. M., LaGasse, L. L., & Bigsby, R. (1998). Prenatal cocaine exposure and child development: What do we know and what do we do? Seminars in Speech and Language, 19 123-146. Lester, B. M., LaGasse, L. L., & Seifer, R. (1998). Cocaine exposure and children: The meaning of subtle effects. Science, 282 633-634. Lezak, M. D. (1995). Neuropsychological assessment (3rd ed.). New York: Oxford University Press. Lidow, M. S. (1995). Prenatal co caine exposure adversely affect s development of the primate cerebral cortex. Synapse, 21 332-341.

PAGE 107

97 Lidow, M. S. (1998). Nonhuman prim ate model of the effect of prenatal cocaine exposure on cerebral cortical development. In J. A. Harvey & B. E. Kosofsky (Eds.), Cocaine: Effects on the development brain (pp. 182-193). New York: New York Academy of Sciences. Light, J. G., Pennington, B. F., Gilger, J. W., & DeFries, J. C. (1995). Reading disability and hyperactivity disorder: Evidence for a common genetic etiology. Developmental Neuropsychology, 11 323-335. Lutiger, B., Graham, K., Einarson, T. R., & Koren, G. (1991). Relationship between gestational cocaine use and pre gnancy outcome: A meta-analysis. Teratology, 44 405-414. Mactutus, C. F. (1999). Prenatal intravenous cocaine adversely affects attentional processing in preweanling rats. Neurotoxicology and Teratology, 21 539-550. Mactutus, C. F., Booze, R. M., & Dowell, R. T. (2000). The influence of route of administration on the acute cardiovascul ar effects of cocaine in conscious unrestrained pr egnant rats. Neurotoxicology and Teratology, 22 357-368. Mactutus, C. F., Herman, A. S., & Booze, R. M. (1994). Chronic intravenous model for studies of drug (ab)use in the pregnant a nd/or group-housed rat: An initial study with cocaine. Neurotoxicology and Teratology, 16 183-191. Mayes, L. C. (1994). Neurobiology of prenatal cocaine exposure e ffect on developing monoamine systems. Infant Mental Health Journal, 15, 121-133. Mayes, L.C., Grillon, C., Granger, R., & Schot tenfeld, R. (1998). Regulation of arousal and attention in preschool children exposed to cocaine prenatally. In J. A. Harvey & B. E. Kosofsky (Eds.), Cocaine: Effects on the developing brain (pp. 126-143). New York: New York Academy of Sciences. McCarthy, D. (1972). McCarthy Scales of Children's Abilities San Antonio, TX: Psychological Corp. Mirsky, A. F., Anthony, B. J., Duncan, C. C ., Ahearn, M. B., & Kellam, S. G. (1991). Analysis of the elements of atte ntion: A neuropsychological approach. Neuropsychology Review, 2 109-145. Morris, P., Gillam, M. P., Allen, R. R., & Paule, M. G. (1996). The effect of chronic cocaine exposure during pregnancy on the acquisition of operant behaviors by rhesus monkey offspring. Neurotoxicology and Teratology, 18 155-166. Narhi, V., & Ahonen, T. (1995). Reading disa bility with or without Attention Deficit Hyperactivity Disorder: Do attentio nal problems make a difference? Developmental Neuropsychology, 11 337-349.

PAGE 108

98 Nassogne, M.-C., Evrard, P., & Courtnoy, P. J. ( 1998). Selective direct toxicity of cocaine on fetal mouse neurons: Teratogenic implications of neurite and apoptotic neuronal loss. In J. A. Harvey & B. E. Kosofsky (Eds.), Cocaine: Effects on the developing brain (pp. 51-68). New York: New York Academy of Sciences. Needlman, R., Frank, D. A., Augustyn, M., & Zuckerman, B. S. (1995). Neurophysiological effects of prenatal cocaine exposur e: Comparison of human and animal investigations. In M. Lewis & M. Bendersky (Eds.), Mothers, babies, and cocaine: The role of toxins in development (pp. 229-250). Hillsdale, NJ: Erlbaum. Neuspiel, D. R. (1994). Behavior in cocaine-exposed infants a nd children: Association versus causality. Drug and Alcohol Dependence, 36 101-107. Nugent, J. K., & Brazelton, T. B. (2000). Preven tive infant mental health: Uses of the Brazelton Scale. In J. D. Osofsky & H. E. Fitzgerald (Eds.), WAIMH handbook of infant mental health volume two: Ea rly intervention, eval uation, and assessment (pp. 157-202). New York: Wiley. Posner, M. I., & Petersen, S. E. (1990). The attention system of the human brain. Annual Review of Neuroscience, 13 25-42. Rabiner, D., Coie, J. D., & The Conduct Probl ems Prevention Research Group. (2000). Early attention problems and children's reading achievement: A longitudinal investigation. Journal of the American Academy of Child and Adolescent Psychiatry, 39 859-867. Randall, C. L., Cook, J. L., Thomas, S. E., & White, N. M. (1999). Alcohol plus cocaine prenatally is more deleteri ous than either drug alone. Neurotoxicology and Teratology, 21 673-678. Reitan, R. M., & Wolfson, D. (1985). The Halstead-Reitan Neurops ychological Test Battery: Theory and interpretation Tucson: Neuropsychology Press. Richardson, G. A., Conroy, M. L., & Day, N. L. (1996). Prenatal cocai ne exposure: Effects on the development of school-age children. Neurotoxicology and Teratology, 18 627634. Robins, P. M. (1992). A compar ison of behavioral and atten tional functioning in children diagnosed as hyperactive or learning-disabled. Journal of Abnormal Child Psychology, 20 65-92. Ronnekliev, O. K., Fang, Y., Choi, W. S., & Ch ai, L. (1998). Changes in the midbrain-rostral forebrain dopamine circuitry in the cocai ne-exposed primate fetal brain. In J. A. Harvey & B. E. Kosofsky (Eds.), Cocaine: Effects on the developing brain (pp. 165181). New York: New York Academy of Sciences. Rose, S. A., & Feldman, J. F. (1995). Prediction of IQ and specific cognitive abilities at 11 years from infancy measures. Developmental Psychology, 31 685-696.

PAGE 109

99 Rowe, K. J., & Rowe, K. S. (1992). The re lationship between inat tentiveness in the classroom and reading achievemen t (Part B): An explanatory study. Journal of the American Academy of Child and Adolescent Psychiatry, 31 357-368. Sandford, J. A. (1995). IVA: Computerized visual and auditory continuous performance task administration manual Richmond, VA: BrainTrain. Sandford, J. A., & Turner, A. (1994). Integrated Visual and Auditory (IVA) Continuous Performance Test Richmond, VA: BrainTrain. Sigman, M., Cohen, S. E., Beckwith, L., Asarnow R., & Parmelee, A. H. (1991). Continuity in cognitive abilities from infancy to 12 years of age. Cognitive Development, 6 47-57. Singer, L. T., Arendt, R., Minnes, S., Farkas, K., Salvator, A., Kirchner, H. L., & Kliegman, R. (2002). Cognitive and motor outcomes of cocaine-exposed infants. JAMA: Journal of the American Medical Association, 287 1952-1960. Spear, L. P. (1995). Alterations in cognitive f unctioning following prenatal cocaine exposure: Studies in an animal model. In M. Lewis & M. Bendersky (Eds.), Mothers, babies, and cocaine: The role of toxins in development (pp. 207-227). Hillsdale, NJ: Erlbaum. Spreen, O., & Strauss, E. (1998). A compendium of neuropsychol ogical tests: Administration, norms, and commentary (2nd ed.). New York: Oxford University Press. Talland, G. A. (1965). Deranged memory. New York: Academic Press. Taylor, S. F. (1996). Cerebral blood flow activat ion and functional lesion s in schizophrenia. Schizophrenia Research, 19 129-140. Thorndike, R. L., Hagen, E. P ., & Sattler, J. M. (1986). Stanford-Binet Intelligence ScaleFourth Edition Chicago: Riverside. Thurman, S. K., Brobeil, R. A., & Ducette, J. P. (1994). Prenatally expos ed to cocaine: Does the label matter? Journal of Early Intervention, 18 119-130. Ullman, J. B. (2001). Structural equation mode ling. In B. G. Tabachnick & L. S. Fidell (Eds.), Using multivariate statistics (4th ed.) (pp. 653-771). Boston: Allyn & Bacon. Velting, O. N., & Whitehurst, G. J. (1997) Inattention-hyperac tivity and reading achievement in children from low in come families: A longitudinal model. Journal of Abnormal Child Psychology, 25 321-331. Vorhees, C. V. (1995). A review of developm ental exposure models for CNS stimulants: Cocaine. In M. Lewis & M. Bendersky (Eds.), Mothers, babies, and cocaine: The role of toxins in development (pp. 71-94). Hillsdale, NJ: Erlbaum. Wall, E. M. (1998). Assessing obstetric risk: A review of obstetric risk-scoring systems. Journal of Family Practice, 27 (2), 153-163.

PAGE 110

100 Wechsler, D. (1974). Wechsler Intelligence Sc ale for Children-Revised New York, NY: Psychological Corp. Wechsler, D. (1981). Wechsler Adult Intelligence Scale-Revised San Antonio, TX: Psychological Corp. Weschler, D. (1989). Wechsler Preschool and Primary Scal e of Intelligence-Revised manual. San Antonio, TX: Psychological Corp. Weschler, D. A. (1991). Weschler Intelligence Scale fo r Children-Third Edition manual San Antonio, TX: Psychological Corp. Weschler, D. A. (1992). Manual for the Weschler Individual Achievement Test San Antonio, TX: Psychological Corp. Willcutt, E. G., & Pennington, B. F. (2000). Comorbidity of reading disability and AttentionDeficit/Hyperactivity Disorder: Differences by gender and subtype. Journal of Learning Disabilities, 33 179-191. Wood, F. B., & Felton, R. H. (1994). Separate linguistic and attenti onal factors in the development of reading. Topics in Language Disorders, 14 42-57. Woodcock, R. W., & Johnson, M. B. (1989). Woodcock-Johnson Psycho-Educational Battery-Revised Allen, TX: DLM Teaching Resources. Woods, J. R., Jr. (1996). Adverse conseque nces of prenatal illicit drug exposure. Current Opinion in Obstetrics and Gynecology, 8 403-411. Woods, N. S., Behnke, M., Eyler, F. D., C onlon, M., & Wobie, K. (1995). Cocaine use among pregnant women: Socioeconomic, obste trical, and psychologi cal issues. In M. Lewis & M. Bendersky (Eds.), Mothers, babies, and cocaine: The role of toxins in development (pp. 305-332). Hillsdale, NJ: Erlbaum. Woods, N. S., Eyler, F. D., Conlon, M., Behnke M., & Wobie, K. (1998). Pygmalion in the cradle: Observer bias agai nst cocaine-exposed infants. Journal of Developmental and Behavioral Pediatrics, 19 283-285.

PAGE 111

101 BIOGRAPHICAL SKETCH Tamara Duckworth Warner was born and ra ised in Charleston, West Virginia. She earned her bachelor of arts degree cum laude with a concentration in Afro-American Studies from Harvard and Radcliffe Colleges in 1992. She earned a master of arts degree from the Program in the American Culture at the University of Michigan in 1996. A master of science degree in clinical ps ychology was earned by Ms. Warner from the University of Florida in 1999. Her research interests include the development of neuropsychological normative data for Af rican American children and adults.


Permanent Link: http://ufdc.ufl.edu/UFE0001153/00001

Material Information

Title: The Effects of prenatal cocaine exposure on attention and reading: a longitudinal study
Physical Description: Mixed Material
Creator: Warner, Tamara Duckworth ( Author, Primary )
Publication Date: 2003
Copyright Date: 2003

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0001153:00001

Permanent Link: http://ufdc.ufl.edu/UFE0001153/00001

Material Information

Title: The Effects of prenatal cocaine exposure on attention and reading: a longitudinal study
Physical Description: Mixed Material
Creator: Warner, Tamara Duckworth ( Author, Primary )
Publication Date: 2003
Copyright Date: 2003

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0001153:00001


This item has the following downloads:


Full Text











THE EFFECTS OF PRENATAL COCAINE EXPOSURE ON ATTENTION AND
READING: A LONGITUDINAL STUDY















By

TAMARA DUCKWORTH WARNER


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


2003




























Copyright 2003

by

Tamara Duckworth Warner




























This dissertation is dedicated to the memory of my father, James V. Duckworth, Jr., who
taught me to always "dream big" and to my mother, Mary A. Duckworth, who always
encouraged to tackle any task that I thought I was "big enough" to handle.















ACKNOWLEDGMENTS

I would like to acknowledge the assistance and patience of my committee

members: Eileen B. Fennell, Ph.D, Chair; Duane E. Dede, Ph.D., Co-chair; Fonda Davis

Eyler, Ph.D., Kenneth Heilman, M.D., Christiana Leonard, Ph.D.; and Michael Marsiske,

Ph.D. I would like to thank Dr. Eyler and her co-principal investigator, Marylou Behnke,

M.D. for allowing me access to their data and for their generous mentorship. Thanks are

also due to the entire Project Care staff, particularly Project Director Kathie Wobie, Ann

Welch, Eric Corpus, and Weir Hou for countless hours of help.

My sincere gratitude also goes to the "family" of the Florida Education Fund's

McKnight Doctoral Fellowship Program, who supported my graduate studies financially,

emotionally, and spiritually. Finally, I want to express my deepest appreciation to my

husband, Kenneth D. Warner, for his steadfast love, support, and encouragement without

which I would not have been to endure the sometimes agonizing process of completing a

dissertation. Thanks and praise go to God, who constantly sustains me and has "made a

way out of no way" many more times than I know.
















TABLE OF CONTENTS
page

A CK N O W LED G M EN T S .............. ............................... ............................................... iv

LIST OF TABLES .................... .......... ......... ........ ..... ...... ............ vii

LIST OF FIGURES .............. ............................... .......... ............. viii

A B S T R A C T ................................. ........... ............................. ............... ix

CHAPTER

1 IN TR OD U CTION ...................................................... ..... .. .... .......

O verview of A attention ................................................................................ .... 1
Development of Visual Attention in Infants and Children...................................... 6
Measurement of Visual Attention in Infants and Children............... ................. 12
Prenatal Cocaine Exposure and Attention..... .......... ...................................... 14
Relationship between Attention and Reading ............................................... 17
Study Purpose and H ypotheses...................................................... .... .. .............. 22

2 M E TH O D S............................. .............. ...... 23

P a rtic ip a n ts ..................................................................................................... 2 3
M measures ................................................................ ..... ...... ....... 25
Demographic Variables..................................... 26
Measures of Cognitive Development............................. .................... 28
A attention M measures ........................................ 28
Verbal Ability .. ...................................................... ...... .............. 33
Reading Ability ................................... ......................... ........ 35
Caregiving Environment Measures............................. .......... 36
P procedure ............................................................................................................. 39
H ypotheses.......................................................... 45
D ata Inspection and Analyses............................................... .................... 46
D ata S creen in g ....................................................................... 4 6
M issin g D ata ....................................................................... 4 6
Accounting for the Participants............................ ..................... 49
Statistical A analyses ........................ ............................... ........................... 50

3 R E S U L T S ................................................................ ....................... .............. 6 1









4 DISCUSSION ......... ........................ ......... ........ ....... 73

Study Summary ......................................... 73
M ain Findings .. .... .................................................. ...... .............. 73
Ancillary Findings ........................................ .... ................................ 74
Study Findings in the Context of the Literature .................................................. 76
Study Strengths ..................................................................... ........ 83
Study L im stations ......... ........................ ............ ........ ........... 86
Future D directions .................. ...................... ........... ... ............. 87

REFEREN CES ....................................................... ......... .......... ....... 90

BIOGRAPHICAL SKETCH .............................................................. .............. 101
















LIST OF TABLES


Table pge

2-1 Continuous Variables that Differed Significantly Between Mothers in the Two
O rigin al Stu dy G rou p s ........................................ ............................................ 54

2-2 Non-continuous Variables that Differed Significantly Between Mothers in the
Tw o O original Study G roups ........................................................ ......... ..... 54

2-3 Variables that Differed Significantly Between Neonates in the Two Original
S tu dy G rou p s.................. .................................. ................ 5 5

2-4 Summary of Variables for Current Study ................................................... 55

2-5 Significant Differences Between Participants With and Without TMT Part A
Scores ............................................................. ..... ...... ........ 57

2-6 Demographic Variables Comparing Groups in Current Study .......................... 58

3-1 Intercorrelations Between All Attention Measures for Combined Sample.......... 67

3-2 Group Means and Standard Deviations for Early Childhood Attention and
R leading V ariables ....... ........................ .................................. ................ 68

3-3 Group Means and Standard Deviations for All Other Study Variables .............. 69

3-4 Fit Indices for Nested Sequence of Measurement and Structural Models............ 70
















LIST OF FIGURES


Figure page

1-1 Theoretical model of the effects of prenatal cocaine exposure on child
b eh av ior .............................. .......... ....... ....... ................ .. ........ ..... 17

2-1 Proposed structural equation model with factors and ....................................... 59

2-2 Proposed structural equation model with factor names ...................................... 60

3-1 Final structural m odel.................... ...... .... ........ ............ ................ ............. .. 71

3-2 Final structural model showing the variances and residuals of the observed
variables w ith their respective factors............................................................... 72















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

THE EFFECTS OF PRENATAL COCAINE EXPOSURE ON ATTENTION AND
READING: A LONGITUDINAL STUDY

By

Tamara Duckworth Warner

August 2003


Chair: Eileen B. Fennell
Cochair: Duane E. Dede
Major Department: Clinical and Health Psychology

Animal studies and knowledge about the pharmacology of cocaine strongly

suggest that maternal cocaine use during pregnancy has negative consequences for fetal

development. However, significant neurobehavioral differences between infants and

children with prenatal cocaine exposure (PCE) and well-matched comparison groups

have not been found consistently. Attention is one area in which group differences have

been reported, but the functional implications of these findings for reading or other

school-related abilities remain uncertain.

A group of prospectively enrolled children who were prenatally exposed to

cocaine, alcohol, tobacco, and marijuana (n = 120) were compared to a matched group of

children who were exposed to alcohol, tobacco, and marijuana but not cocaine (n = 120).

Both groups were predominantly poor and African American and did not differ by sex.









Significant group differences were higher levels of prenatal drug exposure, higher

prenatal risk scores, shorter gestational ages, and smaller head circumferences at birth for

the children with PCE.

Results indicated that neonatal attention as measured by three scales of the

Brazelton Neonatal Behavioral Assessment Scale was not significantly associated with

attentional measures administered at ages 5 and 7, including a continuous performance

test, short term auditory attention (Digit Span), or three tasks involving visual scanning

and visuomotor coordination (Letter Cancellation, Trail Making Test Part A, and the

Coding subtest of the Wechsler Intelligence Scale for Children-III). Analyses also

revealed that children with PCE did not perform significantly worse than non-exposed

children on attention measures at ages 5 or 7 or on the Wechsler Individual Achievement

Test reading subtests at age 7.

Structural equation modeling, however, demonstrated that PCE had an indirect

effect on reading ability at age 7 that was mediated by head circumference at birth. The

effect of PCE on birth head circumference was similar to that of prenatal exposure to

alcohol or marijuana. Birth head circumference affected age 5 verbal ability which, in

turn, was related to age 7 verbal ability and visual attention which, in turn, affected age 7

reading ability. The final model accounted for 68% of the variance in age 7 reading

ability for the combined sample.














CHAPTER 1
INTRODUCTION

Overview of Attention

This manuscript begins with a brief review of definitions and models of attention,

the neuroanatomical correlates of attention, attentional development over the course of

infancy and early childhood, and measurement issues related to attention. In attempting to

define attention, many writers begin with William James' now-famous observation that

Everyone knows what attention is. It is the taking possession by the mind, in clear
and vivid form, of one out of what seem to be several simultaneously possible
objects or trains of thought. Focalization, concentration, and consciousness are of
its essence. (James, 1890, pp. 381-382)

Enns (1990) has observed that many definitions of attention share a component of

selectivity-that a primary function of attention is to select information for further

processing. Similarly, Cohen (1993) argues that attention serves as a gatekeeper in

facilitating "the selection of salient information and the allocation of cognitive processing

appropriate to that information" much like the aperture and lens system of a camera

(Cohen, 1993, p. 3). It is generally agreed that attention is not a unitary process, but a

multifactorial process, involving a diverse set of behavioral phenomena

Based on this general understanding of attention, several researchers have

attempted to construct models of the various components of attention. An empirically

validated model of attention that integrates research from cognitive psychology and

neuropsychology has been articulated by Mirsky, Anthony, Duncan, Ahearn and Kellam

(1991). Mirsky et al. (1991) identified three "elements" of attention-focus, sustain, and









shift. The focus element involves selecting information for enhanced processing, which

has been equated with attention itself. The focus element also includes an "execute"

component, in which motor programs associated with focusing attention are activated.

The sustain component is synonymous with vigilance or the ability to maintain focus and

alertness over time. The third element, shift, represents the ability to change attentive

focus in a flexible and adaptive manner.

Neuroanatomically, the three elements of attention in the Mirsky et al. model can

be localized to various areas of the brain based on lesion studies of brain-impaired

patients, animal lesion studies, and neuroimaging data (Mirsky et al., 1991). The focus-

execute element is associated with the inferior parietal and superior temporal cortex and

striatum of the basal ganglia. The parietal area is the primary cerebral locus of the focus-

execute aspect of attention based on studies of adult neglect patients (Heilman, Watson,

& Valenstein, 1993) as well as the connectivity of the parietal lobe with sensory, motor,

limbic, thalamic and brain stem regions of the brain. It is worth noting, however, that

unlike adults, children rarely show the full adult neglect syndrome even with parietal

lesions (Ferro, Martins, & Tavora, 1984). The proposed role of superior temporal sulcus

in focused attention in Mirsky's model is supported by studies of the architecture and

connectivity of this area as well as studies of epilepsy patients who have undergone

anterior temporal lobectomy although there are some conflicting data on this latter point

(Mirsky et al., 1991). Inclusion of striatum in the anatomy underlying focused attention is

based on its modulatory role in motor systems and the role of the caudate in delayed

alternation and delayed response tasks (Mirsky et al., 1991).









The sustain element of attention is subserved by rostral midbrain structures,

including the mesopontine reticular formation and midline and reticular thalamic nuclei.

Regulation of arousal and maintenance of consciousness are the primary functions of the

brain stem structures. Single unit recording studies in monkeys trained on a go/no-go

visual attention task showed increased firing of Type II cells in the midline thalamus,

superior colliculus, tectum, pons, and mesencephalic brain stem, suggesting a role in

maintaining attention. In addition, studies involving stimulation of reticular nucleus of

the thalamus have demonstrated that this thalamic nucleus modifies the influence of

reticular formation effects on visual signals, specifically one's "readiness to respond" to

visual stimuli in discrimination paradigms.

Neuroanatomically, the shift component of attention in Mirsky's model is the

responsibility of the prefrontal cortex and also perhaps the medial frontal cortex and

anterior cingulate gyms (Mirsky et al., 1991). The role of the prefrontal cortex role in

shifting attention is supported by studies of poor performance on the Wisconsin Card

Sorting Test (WCST) by epilepsy patients who have undergone resection of the

dorsolateral prefrontal cortex. Additional evidence comes from individuals with

schizophrenia who show impaired performance on the WCST and reduced activation of

the prefrontal areas in imaging studies. Preliminary inclusion of the medial frontal cortex

and anterior cingulate gyms in attentional shift is based on the firing patterns of Type II

cells in these regions in monkeys during go/no-go visual discrimination tasks that are

thought to measure both shifting attention as well as sustained attention, as discussed

earlier (Mirksy et al., 1991).









Additional evidence supporting Mirksy's model of the neural substrates

underlying different aspects of attention comes from neuroimaging studies of individuals

with Attention Deficit/Hyperactivity Disorder (ADHD) and schizophrenia, two clinical

disorders with significant attentional dysfunction. ADHD is a behavioral disorder

characterized by hyperactivity, impulsivity, and inattentiveness. In a recent review of

structural and functional neuroimaging studies of children and adolescents with ADHD,

Hale, Hariri, and McCracken (2000) concluded that three sets of findings are emerging

with some consistency: 1) reduced prefrontal and caudate volumes, 2) hypoperfusion and

hypometabolism in prefrontal and striatal regions, and 3) lower levels of activation in the

anterior cingulate during tasks involving stimulus selection and/or response inhibition.

That the frontal areas involved in shifting attention and the striatal areas involved in

sustained attention have been implicated in ADHD provides some support for Mirsky's

model of attention.

Functional neuroimaging studies on the attentional deficits found in those with

schizophrenia are another source of support for the neuroanatomical outline for Mirsky's

model of attention. In Taylor's (1996) review of 24 functional neuroimaging studies

using adults, almost half of the studies (11) found that patients with schizophrenia failed

to show task-related increases in blood flow in the prefrontal area, using the WCST and

similar paradigms. Studies of children and adolescents with childhood-onset

schizophrenia also show reduction of frontal metabolism (Hendren, DeBacker, &

Pandina, 2000). Thus, the neural mechanisms that may underlie the attentional deficits in

schizophrenia for adults appear to be same for children and adolescents. Again, both sets

of findings support Mirsky's localization of shifting attention to the prefrontal area.









One serious limitation of Mirsky's model is its underspecification of the role of

neurotransmitter systems in attention. In contrast, Posner and Petersen's (1990) model of

attention places strong emphasis on norepinephrine (NE) arising from the locus ceruleus

in the rostral brain stem and extending to the posterior attention system, particularly in

the right hemisphere. Norepinephrine is thought to play a crucial role in maintaining an

alert state based on animal studies and the effects that drugs that manipulate NE levels

have on the ability to shift attention (Posner & Petersen, 1990). Posner and Petersen's

model is consistent with current theories that suggest that the core deficits in ADHD are

in executive functions that regulate arousal, attention, and inhibition and that these

deficits may be related etiologically to abnormalities in dopaminergic and noradrenergic

pathways from the brainstem that serve to regulate cortico-striato-thalamo-cortical

networks (Hale, Hariri, & McCracken, 2000). It should be noted, however, that pathways

involving dopamine and norepinephrine are found throughout the brain, and this may be

the reason that Mirsky and his colleagues have not included neurotransmitter components

in their model of the neuroanatomical substrates of attention.

Despite this major limitation, the model of attention offered by Mirsky and

colleagues has been confirmed empirically using principal components analysis in an

epidemiological sample of elementary school children (N= 435) and in a sample of

normal and neuropsychiatrically impaired adults (N= 203). Importantly, the focus-

execute, sustain, and shift factors (in addition to an encode factor) were found in both

samples, suggesting that attention operates according to similar processes in both children

and adults. In both samples, the focus-execute factor was identified by significant

loadings from the Digit Symbol Substitution subtest of the Wechsler Adult Intelligence









Scale-Revised (WAIS-R, Wechsler, 1981) or Coding subtest from the Wechsler

Intelligence Scale for Children-Revised (WISC-R, Wechsler, 1980) and the Talland

Letter Cancellation Test (Talland, 1965). The sustain factor was identified by significant

loadings from a continuous performance test (CPT), specifically mean number of hits and

mean reaction time. Finally, the shift factor was identified by the Wisconsin Card Sorting

Test (Grant & Berg, 1993) percentage correct and number of categories achieved. The

adult sample was also administered the Trail Making Test (Reitan & Wolfson, 1985),

which had significant primary loadings on the focus-execute factor as well as significant

secondary loadings on the sustain factor (Mirsky et al., 1991).

Overall, Mirsky's three factor model of attention, which has been confirmed

empirically and has at least partially-specified neuroanatomical substrates for each factor,

provides a solid foundation for studying attention in both children and adults.

Development of Visual Attention in Infants and Children

This review of the development of attention from the neonatal period through

early to middle childhood will focus on studies of visual attention. Prior to reviewing the

two major conclusions that can be drawn from the developmental literature on attention, a

brief review of the measurement of attention in infants is warranted. Infant attention is

generally studied using paradigms that rely on the visual modality and is measured

clinically using motor orienting responses such as head turning and attempts at visual

fixation to a stimulus. In the laboratory setting, psychophysiological measures, such as

changes in heart rate and evoked brain potentials, have proven to be useful tools in

measuring attention in newborns and young infants who have poor motoric control

(Graham, 1992). Visual fixation on a stimulus, habituation to the stimulus in the form of









decreasing fixation times after continuous or repeated presentations, and reorientation to

a novel stimulus are also common measures.

Generally, two major conclusions can be made from the literature on the

development of visual attention in children: children's attentional capacities improve with

age (Cooley & Morris, 1990) and some continuity exists between attentional abilities

measured at infancy and those measured at later ages (Colombo, 1993; Enns, 1990).

Based on their review of the experimental literature, Cooley and Morris (1990) suggest

that the improvement in children's attentional abilities over time can be explained using

two different theoretical frameworks. The first is an information processing framework

that states that the limited attentional capacity of younger children increases as internal

processing mechanisms develop with age. In the case of selective attention, the filtering

mechanism of attention becomes more efficient and larger proportions of attentional

resources can be allocated and allocated more flexibly to relevant rather than irrelevant

aspects of a task. Similarly, in the case of sustained attention, a limited capacity model

would predict that more effort is required of younger subjects than older subjects in order

to perform well (Cooley & Morris, 1990).

The second theoretical framework used to explain children's improved attentional

abilities over time is an ecological or perceptual learning framework. According to this

framework, children learn with age to become more specific, systematic, economical, and

task-directed in their perception and exploration. In doing so, children become better able

to differentiate relevant from irrelevant information needed for a task. Thus, the

perceptual learning framework tends to emphasize improvements over time in the quality

of children's selective attention abilities (Cooley & Morris, 1990).









Generally, the literature suggests that children begin to develop control over their

visual attentional resources between 7 and 13 years of age, but the development of

efficient use of strategic methods for allocating attention continues through adulthood.

Kaye and Ruskin (1990) conducted a series of three studies to investigate the relative

roles of increased information capacity processing versus strategic allocation of

attentional resources in children's improved performance on various attentional

measures. Using a paradigm requiring shifting attention to a peripheral visual cue, the

researchers found that age differences among 3rd graders, 6th graders and college

students were only found for general alertness, a non-strategic capacity-related factor. In

a divided attention task using two stimulus probability conditions, it was found that both

younger and older children demonstrated adult-like strategic processing, but there were

age differences in their efficient use of these strategies. The third study used a

classification task and children at three different age levels ranging from 5 to 12 years.

No qualitative differences in strategy use were found; however, quantitative differences

in search rates were found, which in turn affected classification. Based on these three

studies, the authors concluded that while adult-like strategies emerge early in childhood,

their optimal use in complex tasks is dependent on the increased information processing

efficiency that comes with age (Kaye & Ruskin, 1990).

The question of continuity in mental abilities generally, and attention in

particular, has been studied in longitudinal samples. The two major paradigms used in

these studies are habituation and response to novelty. Habituation is measured as either

the rate of decline in looking behavior over repeated presentations of a stimulus or the

fixation duration. Response to novelty is measured either by looking behavior to a new









stimulus simultaneously paired with an already-familiar stimulus (novelty preference) or

as recovery of fixation to a new stimulus after habituation (recovery). Bornstein (1990)

reviewed eight longitudinal studies that examined the predictive validity of habituation

paradigms administered during the first six months of life for later cognitive

performance. The median predictive correlation of habituation paradigms using different

sensory modalities was .49, with a range from .28 to .63. Three-quarters of these studies

focused on predicting IQ, two used language measures and one used the Bayley Scales of

Infant Development (Bayley, 1969). Notably, these studies of infant attention used both

normal and at-risk infant samples with the age at the second assessment ranging from 2

years to 8.5 years. A similar review by Colombo (1993) of 13 studies that specifically

used fixation duration to predict a variety of outcome measures found correlations

ranging from .29 to .77.

Colombo (1993) has also reviewed longitudinal studies of predictive validity of

response to novelty measured in the first year of life with later cognitive measures. Based

on the four studies that used recovery after habituation as their measure of response to

novelty, few conclusions can be drawn, in part because recovery measures have poor

reliability and stability (Colombo, 1993). In contrast, the predictive validity of novelty

preference measures with later intellectual and cognitive function is more robust. The

overall median correlation, based on 14 studies of novelty preference, is .47, with a range

from .25 to .66; when only standardized IQ scores are used as the outcome measure, the

median correlation increases slightly to .49 (Colombo, 1993). The majority of these

studies used three IQ outcome measures: the Stanford-Binet (Thomdike, Hagen, &

Sattler, 1986), the age-appropriate Weschler scale, or the Peabody Picture Vocabulary









Test-Revised (Dunn, Dunn, Robertson, & Eisenberg, 1981). Four studies used the Bayley

Scales of Infant Development (Bayley, 1969), four used measures of language

development and two examined memory. The age at first assessment ranged from 3 to 9

months, while the age at second assessment ranged from 1 to 7 years of age (Colombo,

1993). Overall, the reviews by Bomstein (1990) and Colombo (1993) suggest a moderate

degree of predictive validity between experimental measures of attention and

standardized cognitive assessments in early to middle childhood.

A handful of longitudinal studies have begun to examine the development of

particular aspects of attention over time and how these components relate to other

cognitive functions. In pre-term infants it has been found that fixation duration measured

at 40 weeks conceptual age was predictive of performance on a focused attention task but

not a sustained attention task administered at age 12 (Sigman, Cohen, Beckwith,

Asarnow, & Parmelee, 1991). Notably, the focused attention measure was a signal

detection task that also required speeded information processing. The correlations

between the infant attentional measures and later attentional measures were -.36 and -.32,

for the less difficult and more difficult version of the task, respectively. To explain their

findings, the authors argue that both infant attention measures and childhood cognitive

assessment may tap the efficiency of information processing. If this is the case, then an

individual's relative capacity to process information efficiently would be expected to be

stable from infancy to childhood. Furthermore, infant information processing abilities

would not be expected to be associated with other measures of childhood attentional

ability such as sustained attention (Sigman et al., 1991).









In another longitudinal cohort, Rose and Feldman (1995) found that an indirect

measure of sustained attention (exposure time to meet criterion during a visual

recognition memory task) acquired at 7 months of age in a sample of pre-term and full-

term infants was significantly correlated with perceptual speed at 11 years of age. When

IQ was partialed out, the correlations between the infant sustained attention measure and

the perceptual speed measures were .33 for a visual matching task, .34 for a visual search

task and .38 for the two measures combined (Rose & Feldman, 1995).

The significant correlations between sustained attention and perceptual speed in

the Rose and Feldman (1995) study seem to support increased capacity and increased

efficiency theories posited by information processing accounts. The authors also suggest,

however, a second possible explanation for the significant correlation between sustained

attention and perceptual speed: that the speed with which the child is able to respond is a

function of his or her ability to filter out stimuli irrelevant to the task (Rose & Feldman,

1995). In this way, the sustained attention measure is also a measure of appropriately

focused attention. Unfortunately, the researchers did not include any attentional

assessments in their longitudinal battery so the relationship between the infant attentional

measure cannot be compared to later childhood attentional measures.

One major limitation of the experimental literature on the development of

children's attention, as with the models of attention reviewed earlier, is its neglect of the

development of underlying brain systems. While neuropsychological studies of childhood

attention, which tend to focus on deficits found in clinical populations, stress the

importance of the developing neural substrates, they rarely elucidate specific brain-

behavior relationships. More important perhaps is the fact that few studies in the









developmental neuropsychology literature have attempted to relate children's

development of various components of attention-e.g., focus, shift, and sustain-to brain

maturation (Cooley & Morris, 1990).

Measurement of Visual Attention in Infants and Children

As briefly reviewed in the previous section, a number of different experimental

paradigms exist for measuring visual attention in infants. The number of clinical

measures available for assessing attention in the first few years of life is not as plentiful.

As a result, few developmental studies are available that explore the growth and change

of attention using clinical measures. Arguably, the most commonly used assessment

instrument for the study of infant attention is the Brazelton Neonatal Behavioral

Assessment Scale (BNBAS). First developed in 1973, the BNBAS is designed to assess

the newborn's adaptive responses to his or her new extrauterine environment and has

been used for over 25 years in hundreds of research and clinical setting across the world

(Nugent & Brazelton, 2000). Theoretically, development of the BNBAS is based on the

assumption that early human development proceeds from a state of relative

undifferentiation to one of increasing differentiation, articulation and hierarchic

integration. More specifically, there are hypothesized four primary developmental tasks

of the newborn, arranged hierarchically and related to increasing self-regulation. The first

developmental task is organization of autonomic or physiological behavior, which

involves homeostatic adjustment of the central nervous system including respiration, the

startle response, tremors and temperature regulation. The second development task is

control and regulation of motor behavior, including inhibition of random motor

responses, development of better muscle tone and reduction of excessive motor activity.

Third, the neonate must develop state regulation, the ability to modulate states of









consciousness including sleep, and deal with stress using strategies such as hand-to-

mouth movements, communicating with the caregiver through crying and being consoled

with the aid of the caregiver. The fourth and final task for the infant is regulation of

affective interactive or social behavior, which involves maintaining prolonged alert

periods, attending to visual and auditory stimuli within one's range and seeking out and

engaging in social interaction with the caregiver (Nugent & Brazelton, 2000).

The development of the first three self-regulatory capacities can be seen as laying

the foundation for the fourth, which is the rudimentary development of attention to

external stimuli. Following regulation of internal stimuli, the neonate can turn to

maintaining alertness or arousal, which is a prerequisite for directing, focusing and

sustaining attention. With repeated assessments, the BNBAS describe this emerging

process during the first couple of months of the newborn's life in the extrauterine

environment. A more detailed description of the scale and its properties is found in the

Methods chapter.

After the neonatal period, standardized measurements of infant attention are more

difficult to come by. This is due, in part, to a variety of methodological challenges

associated with developmental research, not the least of which is the problem of

measurement equivalence (Hartmann & George, 1999). Measurement equivalence refers

to the question of whether an assessment instrument measures the same construct at

various developmental ages. For example, an "intelligence" test may be more a measure

of verbal comprehension than reasoning at age 3 as compared to age 10. In addition,

children's rapid changes in physical and behavioral maturation and experience can make

transient performance levels difficult to capture (Hartmann & George, 1999). A variety of









design and data analytic techniques can be used to help guard against and test for

measurement equivalence, but it is a thorny issue ever-present in developmental research

and very difficult to overcome in infancy and childhood investigations.

Prenatal Cocaine Exposure and Attention

With the alarming rise in the use of cocaine by pregnant women since the mid-

1980s, children who have been prenatally exposed to cocaine have emerged as a new

clinical population in which to study the development of attention. That prenatal cocaine

exposure has a teratogenic effect on development has generally been established. Cocaine

has the potential to exert deleterious effects in variety of direct and indirect ways.

Cocaine easily crosses both the placenta and the blood-brain barrier (Mayes, 1994). Most

of what is known about the pharmacologic and other effects of cocaine on prenatal

development has been learned through research on a variety of animal populations,

including rats, mice, rabbits, and monkeys (Kosofsky & Wilkins, 1998; Lidow, 1998;

Spear, 1995). There is a general consensus that animal models (despite differences in

route of administration and other methodological problems) are useful analogues of

prenatal cocaine exposure in humans (Needlman, Frank, Augustyn & Zuckerman, 1995).

Three well-studied pharmacological effects of cocaine will be reviewed here. First,

cocaine inhibits the reuptake of dopamine, serotonin and norepinephrine, thereby

potentiating their actions (Akbari, Kramer, Whitaker-Azmitia, Spear & Azmitia, 1992;

Dow-Edwards, 1995; Factor, Hart & Jonakait, 1993; Friedman & Wang, 1998; Leslie,

Robertson, Jung, Libermann & Bennett, 1994; Mayes, 1994; Mactutus, Herman, &

Booze, 1994; Ronnekliev, Fang, Choi & Chai, 1998; Vorhees, 1995). Second, cocaine

causes vasconstriction and can reduce blood supply and oxygenation to a developing

fetus, resulting in chronic or intermittent hypoxia (Dow-Edwards, 1995; Woods, 1996).









Third, cocaine acts as a potent local anesthetic, blocking sodium channels, thereby

attenuating action potentials in excitable cells (Dow-Edwards, 1995).

In addition to its pharmacological effects, maternal use of cocaine can cause

hypertension, placental aburption, spontaneous abortion, poor pregnancy weight gain and

undernutrition secondary to appetite loss (Church, Crossland, Holmes, Overbeck & Tilak,

1998). Brain abnormalities reported in animal models of prenatal cocaine exposure

include a reduction in number of cortical cells, inappropriate positioning of cortical

neurons, altered glial morphology, reduction in length of neurites, and apoptotic neural

cell loss (Lidow, 1995, 1998; Nassogne, Evrard, & Courtnoy, 1998). These changes are

attributed, in part, to alternations in function of monoaminergic neurotransmitters that

affect synaptogenesis, neural growth, and cell proliferation.

However, unlike early media reports of severe developmental consequences

associated with cocaine use during pregnancy, research in humans has demonstrated that

the sequelae of prenatal cocaine exposure are subtle but meaningful (Harvey & Kosofsky,

1998; Lester, LaGasse, & Seifer, 1998; Neuspiel, 1994; Vorhees, 1995). In terms of

pregnancy outcome, a few consistent findings have emerged in studies with nondrug

using control groups: higher risk for spontaneous abortion, shorter gestational age,

smaller head circumference, shorter birth length and lower birth weight (Lester, Freier &

LaGasse, 1995; Lutiger, Graham, Einarson, & Koren, 1991). In terms of neurobehavioral

outcomes, no overall syndrome has been found in infants prenatally exposed to cocaine,

in part due to methodological problems (Frank, Augustyn & Zuckerman, 1998; Lester,

LaGasse & Bigsby, 1998). A review often studies using the BNBAS, one of the most

consistent measures used, suggests that problems with state regulation occurs most









frequently (Frank, Augustyn & Zuckerman, 1998). State regulation, according to the

theory underlying the development of the BNBAS, is the third developmental task of the

first two months of life, which must be successfully negotiated before attention can be

directed to stimuli in the external environment (Brazelton, 1984, 1994; Brazelton, Nugent

& Lester, 1987; Nugent & Brazelton, 2000).

In terms of prenatal cocaine exposure's specific effect on attention, a number of

animal models are suggestive. In a mouse model using Pavlovian conditioning and a

blocking paradigm in which redundant information must be ignored, learning deficits and

behavioral alterations suggestive of problems in selective attention were found (Kosofsky

& Wilkins, 1998). Similarly, a primate model using rhesus monkeys also demonstrated

problems in the acquisition of operant behaviors in cocaine-exposed offspring (Morris,

Gillam, Allen & Paule, 1996). Impairments in attention and discriminative learning have

been demonstrated in rabbits using a foot shock paradigm (Gabriel & Taylor, 1998).

In humans, a handful of studies have begun to emerge indicating that children

prenatally exposed to cocaine begin to show problems with attention and that these

problems become more evident after age 4 (Beckwith, Crawford, Moore & Howard,

1995; Chasnoff, Anson, Hatcher, Stenson, Iaukea & Randolph, 1998; Leech, Richardson,

Goldschmidt & Day, 1999; Mayes, Grillon, Granger & Schottenfeld, 1998). A model for

understanding the effects of prenatal cocaine exposure on child behavior, including

attention has been presented by Lester, Freier and LaGasse (1995). As shown in Figure 1,

cocaine is thought to affect neuroregulatory mechanisms, which in turn result in









disorders of behavioral regulation that manifest as the "Four A's of Infancy": attention,

arousal, affect and action.



Cocaine/Other Drugs



Neuroregulatory Mechanisms



Disorders of Behavioral Regulation



Four A's of Infancy



Attention Arousal Affect Action



Neurodevelopmental
Assessment Battery



Figure 1-1. Theoretical model of the effects of prenatal cocaine exposure on child behavior
taken from Lester, Freier and LaGasse (1995).

Relationship between Attention and Reading

Few theoretical models explicitly link the development of attentional abilities and

the development of reading skills in children. Models of reading development generally

focus on the processing of component skills in three areas: phonology, orthography, and

semantics (Adams, 1994). A variety of investigations, including genetic studies, have

examined the association between reading disability and Attention Deficit/Hyperactivity

Disorder (ADHD); (Brock & Knapp, 1996; Felton & Wood, 1989; Fergusson &

Horwood, 1992; Gilger, Pennington & DeFries, 1992; Javorsky, 1996; Light, Pennington,









Gilger & DeFries, 1995; Nahri & Ahonen, 1995; Robins, 1992; Velting & Whitehurst,

1997; Willcutt & Pennington, 2000). The overwhelming majority of these studies,

however, have failed to distinguish subtypes with inattentive symptoms and

hyperactivity. As Hinshaw (1992) pointed out in his review, while the link has been

established between inattention and hyperactivity, on the one hand, and reading

underachievement, on the other, causal models have rarely been tested using sufficient

methodological controls. Only a small number of investigations have examined the

relationship between attention and reading in children without diagnosed ADHD or

reading disability (e.g., Velting and Whitehurst, 1997; Wood and Felton, 1994).

Fortunately, two longitudinal investigations using large samples of normal children have

begun to elucidate the relationship between attention and reading (Rabiner, Coie, & the

Conduct Problems Prevention Research Group, 2000; Rowe & Rowe, 1992). Both of

these longitudinal studies merit detailed review.

A longitudinal investigation by Rabiner and colleagues (2000) found that early

attention problems predicted reading achievement even after controlling for prior reading

achievement, IQ, and other behavioral problems. A heterogeneous group of children (N=

211) from four sites was followed from kindergarten through fifth grade. Attentional

measures were collected from the children's teachers using the inattentive items from the

Child Attention Problems Scale (Edelbrock, 1990) in kindergarten and the inattention

scale of the ADHD Rating Scale (DuPaul, 1991) in grades 1 and 2. Reading achievement

was measured in kindergarten and grade 1 using the Letter-Word Identification subtest

from the Woodcock-Johnson Psychoeducational Battery-Revised (WJ-R, Woodcock &

Johnson, 1991); reading achievement in grade 5 was measured using both the Letter-









Word Identification and Passage Comprehension subtests of the WJ-R. It was found that

kindergarten reading was significantly correlated (r = -.29) with kindergarten inattention

after controlling for IQ and parental involvement. First grade reading was also

significantly correlated with first grade inattention (r = -.29) after controlling for IQ,

parental involvement, and kindergarten reading and inattention scores. Inattention

accounted for 6% of the total variance in kindergarten and first grade reading scores,

comparable to the amount of variance explained by IQ. Fifth-grade reading was

significantly correlated (r = -.10) with second grade inattention after controlling for all

the previous variables-IQ, parental involvement, kindergarten reading and inattention

and first grade reading and inattention. A path analysis using multiple regression

procedures found that the model with all of the variables explained 66% of the variance

in the children's fifth-grade reading achievement.

More importantly, the researchers found that children who were highly inattentive

first graders (standardardized scores >1.0) were at greater risk for reading difficulties.

Between kindergarten and first grade, the mean standardized reading achievement scores

of the highly inattentive children declined significantly (-.52 to -.86), making them three

times more likely than their peers to show a one standard deviation discrepancy criteria

between IQ and reading ability. By fifth grade, the mean standardized reading score for

the highly inattentive first graders remained substantially below the mean at .71.

Another longitudinal study by Rowe and Rowe (1992) used structural equation

modeling in a sample of 5,092 normal students ages 5 to 14 years to investigate the

relationship between inattentiveness in the classroom and reading achievement as

mediated by family socioeconomic background factors, reading activity at home and









attitudes towards reading. A stratified sample of students was drawn from 256 classes in

64 public and 28 private elementary and secondary schools from four regions (two

metropolitan and two rural) in Victoria, Australia, reflecting 91% of the target sample.

Study measures were collected at five time points: year levels 1, 3, 5, 7, and 9. The

family socioeconomic indicators used were number of years of mother's education and

father's education and mothers' and father's occupational classification as measured on an

8-point scale. Students' reading activity at home was obtained from self-report responses

to four questions measured on a four-point Likert-type scale while students' attitude

toward reading was determined using three question measured on a 5-point Likert-type

scale. For students ages 5 to 6 years, interviews were conducted with classroom teachers

to assess reading activity and attitudes. Classroom inattentiveness at all ages was

obtained from teacher responses to four items measured on a 5-point Likert-type scale.

Assessment of reading achievement consisted of two measures: an age-appropriate

reading comprehension test and teacher ratings on a criterion-referenced profile of

student reading behaviors (Rowe & Rowe, 1992). Data were analyzed using four age

categories: 5 to 6 years, 7 to 8 years, 9 to 11 years and 12 to 14 years.

Results indicated that across all age groups the measure of inattentiveness

accounted for the largest proportion of variance, ranging from 13.4% to 22.9%. Attitudes

towards reading and reading activity at home each explained between approximately 5%

and 15%. The proportion of variance in reading achievement accounted by

socioeconomic variables was very small, ranging from 0.3 to 3.2%. Analyses using a

recursive structural equation model, in which all effects are unidirectional, indicated that

inattentiveness has strong negative influences on students' reading achievement as well as









on the mediating variables of attitude towards reading and reading at home. Reading

activity at home was found to have a significant influence on students' attitudes towards

reading and acted as a strong mediator between inattentiveness and reading achievement

that increased as students progressed through school. Socioeconomic status was found to

have little influence on any of the other four factors in the model (Rowe & Rowe, 1992).

A second set of analyses was conducted using a non-recursive structural equation

model to examine interdependent effects between inattentiveness and reading

achievement. Goodness-of-fit indices were greater than .97 for all age groups, indicating

that the model fit the data well. Reciprocal effects between inattentiveness and reading

achievement were found to be significant and negative, and this relationship grew

stronger over time. In sum, the two principal findings were that: 1) inattentive behaviors

in the classroom have a significant negative influence on students' reading achievement

and 2) reading achievement, mediated by the direct influence of attitudes toward reading

and reading activity at home, has a stronger effect on reducing inattentive behaviors.

Related to the second finding, the authors argue that low reading achievement leads to

high inattentiveness (Rowe & Rowe, 1992).

The results of these two longitudinal studies converge on the notion that

significant relationships exist between children's attentional abilities and their reading

ability. Both studies also suggest that early intervention for inattentive children may help

to reduce the chances of later reading problems. For children prenatally exposed to

cocaine who may be at greater risk for attentional problems, early detection and

intervention could have a significant impact on their academic and overall developmental









outcomes and highlights the need for more research on this potentially vulnerable

population.

Study Purpose and Hypotheses

The purpose of the current study is to investigate the developmental trajectory of

attention in a sample of children prenatally exposed to cocaine to determine: 1) whether:

an indicator of attentional problems in infancy predicts poor attention skill in early

childhood and 2) whether attentional measures at birth and early childhood are related to

reading ability after controlling for the influence of general verbal ability and the

caregiving environment.

It is hypothesized that after controlling for prenatal obstetric risk and exposure to

alcohol, tobacco, and marijuana, the significant differences between children prenatally

exposed to cocaine and matched controls on an early indicator of attention problems will

persist at ages 5 and 7. More specifically, children prenatally exposed to cocaine will

show worse performance on measures of attention than their matched controls. In

addition, the poor performance of the exposed children will have direct and indirect

negative effects on reading achievement at age 7 after controlling for verbal ability and

the caregiving environment.














CHAPTER 2
METHODS

Participants

Participants in the current study were children enrolled in a prospective,

longitudinal National Institute of Drug Abuse-funded study (DA 05854) to examine the

effects of prenatal cocaine exposure on developmental outcomes. The original study,

entitled Project C.A.R.E. (Cocaine Abuse in the Rural Environment), is housed at the

University of Florida's Shands Teaching Hospital, and the study's principal investigators

are Marylou Behnke, M.D. and Fonda Davis Eyler, Ph.D. In the original study, 154

cocaine-using pregnant women and 154 non-using matched controls were enrolled

prospectively soon after they first contacted the health care system, either at a prenatal

obstetric clinic or at the hospital. Detailed information regarding recruitment and

enrollment of participants is provided in the Procedures section. Since enrollment in the

longitudinal study, there has been a relatively low attrition rate (approximately 10%). All

the children who participated in follow-up assessments at ages 5 and 7 and who have

complete data for all the study measures were included in the present study.

A variety of data is available on the original study sample based on previous

studies (Behnke, Eyler, Conlon, Wobie, Woods, & Cumming, 1998; Behnke, Eyler,

Woods, Wobie, & Conlon, 1997; Eyler, Behnke, Conlon, Woods, & Wobie, 1998a,

1998b; Eyler, Behnke, Garvan, Woods, Wobie, & Conlon, 2001; Woods, Behnke, Eyler,

Conlon & Wobie, 1995). Of the 154 cocaine users, 70% admitted to using crack, 16%

used powder cocaine and 14% denied any cocaine use but had a positive urine screen.









Only 6% of the cocaine-using group of women received some form of drug treatment

during pregnancy. The entire study sample was predominantly African American (n =

125), lowest Hollingshead socioeconomic status (SES) category (Hollingshead, 1995, n =

118) and had more than one child. As shown in Tables 2-1, 2-2, and 2-3, significant

differences on a number of variables were found between the cocaine-using mothers and

their babies when compared to their matched controls. First, the cocaine-using mothers

were significantly older than the comparison group (27.4 years vs. 22.8 years, p = .0001).

However, there were no differences between the groups based on the number of women

in the over 40 years age range, which has been associated with increased perinatal risk

(Eyler et al., 1998a). Second, the cocaine-using mothers entered prenatal care

significantly later than the non-using mothers. However, multiple regression analyses

controlling for alcohol, tobacco, and marijuana use showed that only tobacco use

significantly predicted when the mothers entered prenatal care (Eyler et al., 1998a).

Third, the Hobel Total and Perinatal Risk Scores were significantly higher for the

cocaine-using mothers, and the difference in the Hobel Total Risk Score was due to

higher Prenatal Risk Scores. There were no significant differences between groups on the

Labor and Delivery and Neonatal Risk Scales (Eyler et al., 1998a). The fourth difference

between the groups was in the proportion of mothers who used other substances during

their pregnancies. Significantly more cocaine-using mothers also used tobacco, alcohol,

and marijuana compared to their matched controls. Lastly, the number of infants born

before 37 weeks gestation was significantly higher among the cocaine-using mothers, but

there was no significant difference in mean gestational age between the two groups of

infants as calculated using the method of Dubowitz, Dubowitz, and Goldberg (1970) (M









= 38.3 weeks, SD = 2.7 for cocaine-exposed, M= 38.7 weeks, SD = 2.9 for non-exposed;

p = .24) (Eyler et al., 1998a).

For the neonates, significant differences on all four growth measures-birth

weight, length, head circumference, and chest circumference-were found between those

with prenatal cocaine exposure (PCE) and those without PCE. However, the Ponderal

Index (Kliegman & King, 1983), calculated using the standard formula birth weight

(grams) divided by length (cm)3, did not differ between groups although the infants with

PCE had significantly lower birth weights and significantly shorter birth lengths as

compared to the infants without PCE (Eyler et al., 1998a). In multiple regression analyses

using cocaine, marijuana, alcohol, and tobacco, no single drug or combination of drugs

was a significant predictor of the Ponderal Index. In contrast, with head and chest

circumference, there was an interaction between cocaine and tobacco such that the infants

of mothers who used both were significantly smaller than the infants of mothers who did

not use tobacco, who only used tobacco, or who used cocaine but did not use tobacco

(Eyler et al., 1998a).

Measures

Four sets of variables were used in the current study: demographic variables, a

birth outcome measure, measures of cognitive development at age 5 and 7, and

caregiving environment measures. The demographic variables, also called exogenous

variables, were of interest for the current study because of their potential relationship

with child cognitive development. The six demographic variables were: amount of

prenatal drug exposure (cocaine, alcohol, tobacco, and marijuana), Hobel prenatal

obstetrical risk score, and gender. Child ethnicity was excluded as a variable since the

overwhelming majority of the sample (more than 75%) is African American, the two









study groups were matched for socioeconomic status, and it is necessary to limit the

number of variables for the analyses planned. In terms of birth outcome measures, head

circumference at birth has been shown to be predictive of later outcomes in several

studies (Chasnoff, Griffith, Freier, & Murray, 1992; Eyler, Behnke, Garvan, Wobie, &

Hou, 2002). Data from 21 child cognitive measures collected at birth, age 5, and age 7

were also used in the current study: 12 attention measures, 8 verbal ability measures, and

2 reading ability measures. Finally, 7 measures of the caregiving environment gathered at

age 5 and age 7 were included to examine the relative contribution of the home

environment to child cognitive development. Table 2-4 provides a summary of the all of

the variables used in the study, the construct they were designed to assess, and the ages at

which they were assessed.

Demographic Variables

Prenatal drug exposure. Maternal use of cocaine, alcohol, tobacco, and

marijuana was obtained using a drug history interview procedure adapted from that of

Day, Wagener, and Taylor (1985). Detailed information about the drug history interview

procedure is provided in the Procedures section. Prenatal cocaine exposure was

operationalized as a ratio of the number of weeks of reported cocaine use divided by the

total number of weeks of each gestation plus 3 months prior to gestation (the period

covered by the substance use interview). Prenatal alcohol exposure was quantified using

the average number of ounces of absolute alcohol consumed per day throughout

pregnancy. Similarly, the average number of cigarettes smoked per day and the average

number of marijuana joints smoked per day throughout the pregnancy were used to

measure prenatal tobacco and prenatal marijuana exposure, respectively.









Head circumference. Orbitofrontal head circumference for each child measured

in centimeters using a plastic coated tape.

Hobel Obstetric Risk Scale (Hobel, Hyvarinen, Okada, & Oh, 1973). The Hobel

provides a quantitative assessment of 125 prenatal, intrapartum and neonatal factors that

are associated with perinatal morbidity and death. For the current study only the Hobel

Prenatal Risk Score was used as it was found to differ between the cocaine-using mothers

and the comparison mothers in the original sample from which the participants were

drawn (Eyler et al., 1998a). Scores are assigned clinically to 50 historical and developing

prenatal items; 40 early, interim, and late intrapartum factors; and 35 neonatal factors.

Weights of 1, 5, or 10 are assigned to each of the factors based on their assumed

relationship to perinatal morbidity and death. The initial validity of the Hobel Obstetrical

Risk scale was demonstrated in a sample of 738 mixed high- and low-risk pregnancies

using theoretically assigned weights to each of the variables (Hobel, Hyvarinen, Okada,

& Oh, 1973). Validity was further established in a larger sample of 1,417 mixed high-

and low-risk pregnancies (that included the 738 cases from the previous sample) by

comparing the clinically assigned weights to those derived from a logistic regression

model (Hobel, Youkeles, & Forsythe, 1979). In this latter study, the clinically assigned

scores had a true positive classification rate of 82.5% and a true negative classification

rate of 49.5%, which compared favorably with the logistic model's predictions. In a

review of obstetric risk-scoring systems, Wall (1988) reported that the Hobel system has

a sensitivity of .504 and .669, specificity of .685 and .701, and positive predictive validity

of .228 and .293 for the antepartum period and the intrapartum periods, respectively.









Measures of Cognitive Development

Attention Measures

Brazelton Neonatal Behavioral Assessment Scale (BNBAS; Brazelton, 1984).

Three BNBAS cluster scores, Habituation, Orientation, and Regulation of State, were

used as measures of infant attention after birth. The BNBAS is a widely used instrument

for assessing the neurobehavioral responses of the newborn to his or her new extrauterine

environment. It is designed as an interactive assessment for use with newborns from 36 to

44 weeks gestational age. The BNBAS consists of 28 behavioral items scored on a nine-

point scale and 21 reflex items scored on a four-point scale. The behavioral items

examine behaviors such as response to visual, auditory and tactile stimulation,

orientation, alertness, activity, and irritability. Central to the behavioral assessment is the

newborn's state of consciousness-deep sleep, light sleep, drowsy quiet alert, active alert

or crying-which serves as the foundation for evaluating his or her sensory and motor

responses. The scoring summary divides the items into four general domains of

functioning; however, a seven-cluster scoring method has been developed based on

conceptual and empirical methods. The seven BNBAS clusters are: Habituation,

Orientation, Motor, Range of State, Regulation of State, Autonomic Stability, and

Reflexes. Psychometrically, test-rest reliability is difficult to establish for the BNBAS

due to rapid changes in the organization of the neonate's behavior during the first few

days and weeks. The BNBAS requires extensive training to obtain interrater reliability of

.92 as recommended by the manual. The validity of the BNBAS has been established by

more than 25 years of clinical and research use (Brazelton, 1984).









Integrated Visual and Auditory Continuous Performance Test (IVA CPT;

Sandford & Turner, 1994; Sandford, 1995). The two IVA CPT composite scores for the

visual modality, the Visual Attention Quotient (VAQ) and Visual Response Control

Quotient (VRCQ), were used as measures of attention at ages 5 and 7. The visual

attention composite scores were chosen over the auditory attention composite scores

since the other attention measures in the study-Letter Cancellation, Trail Making Test,

and the WISC-III Coding subtest-all rely primarily on the visual modality. Test

protocols were reviewed to determine valid profiles based on the IVA CPT manual

criteria. The VAQ is a composite score comprised of three raw scores: Focus (response

variability), Vigilance (omission errors) and Speed (mean reaction time for all correct

trials). The VRCQ is a composite score also derived from three raw scores: Prudence

impulsivityy), Consistency (response variability), and Stamina (mean reaction time for

correct responses to the first 200 and last 200 trials).

The IVA CPT is a 13-minute test of attention for children and adults designed to

provide data for differentiating between the subtypes of Attention Deficit/Hyperactivity

Disorder specified in the Diagnostic and Statistical Manual-4th Edition (DSM-IV;

American Psychiatric Association, 1994). The IVA CPT measures responses to 500

intermixed visual and auditory stimuli spaced 1.5 seconds apart. The task involves

responding by clicking a computer mouse when the stimulus is a visual or auditory 1 and

inhibiting responses when the stimulus is a visual or auditory 2. The stimuli are presented

in pseudo-random order in five sets of 100 trials with each set consisting of two 50-trial

blocks. The blocks are counterbalanced between visual and auditory stimuli and between

frequent presentation of target stimuli (designed to elicit impulsivity) and infrequent









presentation of target stimuli (designed to elicit inattentiveness). Overall, the IVA CPT

yields six composite quotient scores on two factors (Response Control and Attention) and

22 raw scores which comprise a Fine Motor Regulation (Hyperactivity) scale, three

Attribute scales, and six Validity scales (Sandford, 1995).

Limited demographic information about the normative sample for the IVA CPT is

available from the manual. The sample consisted of volunteers (N= 487, males = 210,

females = 277) ranging in age from five to 90 years. Individuals in the normative sample

were not known to have past neurological disorders or current psychological, learning,

attentional problems or to demonstrate hyperactivity. In addition, the normative sample

was screened for medications other than birth control and nasal sprays and was not

currently active in psychotherapy or counseling. Significant gender differences were

found on two scales: males had faster reaction times but females made fewer commission

(impulsive) errors. In addition, significant age effects for mean reaction time for correct

responses followed a U-shaped curve. The test appears to be more demanding for

younger children, as a rapid improvement (reduction in reaction time) was seen for

children between the ages of 5 and 7. Reaction time continues to improve between 8 and

12 years of age then plateaus between the mid-teen to young adult years. Reaction time

was fairly stable through middle age and then slowed down slightly after age 45.

Normative information in the computerized database that accompanies the test is

reportedly divided into "appropriate" age and sex groups (Sanford, 1995).

The limited information available about the psychometric properties of the IVA

CPT suggests that it has adequate reliability and validity. The IVA CPT's test-retest

reliability was studied using 70 normal volunteers (43 females, 27 males) between 5 and









70 years of age (mean age = 21.8 years). Correlation for a one- to four-week interval

between test administrations for the Visual Attention Quotients was very strong at .75.

For the visual attention Validity and Attribute scores, the correlations ranged from .34 to

.80. Validity studies on the IVA CPT were conducted in a small sample of 26 children,

ages 7 to 12, diagnosed by a physician or psychologist as having ADHD and a

comparison group of 31 children with no known neurological, learning, emotional or

ADHD related problems. Results indicated that IVA CPT shows excellent sensitivity,

specificity, positive predictive power, and negative predictive power: 92%, 90%, 89%

and 93%, respectively. Concurrent validity was established by comparing the IVA CPT

to two other continuous performance tests and two rating scales. The IVA CPT showed

90% to 100% agreement with these other measures and had the lowest false positive rate

at 7.7% (Sandford, 1995).

Letter Cancellation (Diller, Ben-Yishay, Gerstman, Goodin, Gordon, &

Weinberg, 1974). The Letter Cancellation task time to completion is used as a measure of

attention at ages 5 and 7. Mirsky et al. (1991) found, in both adults and children, that the

Letter Cancellation test loaded on the "focus-execute" factor of attention in their three-

factor model of attention. In addition to attention, the Letter Cancellation test is thought

to assess visual scanning, motor speed, and activation and inhibition of repetitive motor

responses (Lezak, 1995). The task consists of crossing out a target character that is

randomly interspersed approximately in an array of at least five different characters.

Three scores can be derived based on speed (time to completion), number of omission

errors, and number of commission errors. The Letter Cancellation test has been found to

be sensitive to a variety of problems in brain-damaged subjects, including spatial neglect









in right hemisphere stroke patients and temporal processing difficulties of left hemisphere

stroke patients (Lezak, 1995).

Trail Making Test (TMT; Reitan & Wolfson, 1985). The TMT Trails A time to

completion is used as measure of attention at age 7. The TMT is one subtest in the

Halstead-Reitan Neuropsychological Test Battery and consists of two parts labeled Part

A and Part B. The TMT is thought to measure a variety of functions including attention,

visual scanning, sequencing, mental flexibility, and motor speed and agility (Lezak,

1995; Spreen & Strauss, 1999).

In Part A, the subject is instructed to draw lines to connect circles containing

numbers scattered randomly on a page in numerical order. In Part B, the participant must

draw lines alternately between circles containing numbers and circles containing letters in

numerical and alphabetical order. Only Part A will be used since a meta-analysis of four

studies of children ages 9 to 14 found that Part B may be less reliable in younger children

(Leckliter, Forster, & Klonoff, 1992). Scoring for the test is based on time to completion.

In adult studies using a variety of patient populations (except those with

schizophrenia), the reliability coefficients generally range from .64 to .94 (Spreen &

Strauss, 1998). Mirsky et al. (1991) found using principal component analysis on test

scores from a mixed sample of adults that both Parts A and B of the TMT loaded most

highly on a "perceptual-motor speed" factor (.70 and .63, respectively), corresponding to

their Focus-Execute component of attention; however, there were also significantly

secondary loadings on a "vigilance" factor (.43 and .45 for Parts A and B, respectively),

corresponding to their Sustain component of attention.









Wechsler Intelligence Scale for Children-Third Edition (WISC-III, Wechsler,

1991). The WISC-III Coding and Digit Span subtests were used as a measure of attention

at age 7. The WISC-III is a well-validated test of general cognitive functioning for

children ages 6 years to 16 years, 11 months. The test was standardized on a national

sample of 2,220 children stratified by age, sex, race/ethnicity, geographic region and

parent education according to the 1988 U.S. Census. The WISC-III has demonstrated

good psychometric properties. The validity of the WISC-III is based, in part, on the

numerous criterion-related studies conducted on its predecessor, the WISC-R. Factor

analytic studies as well as correlational studies with three other Wechsler Scales (the

WISC-R, WPPSI-R and WAIS-R), other ability tests, neuropsychological tests, and

school grades support the validity of the WISC-III. In addition, the data on the WISC-III

has been collected using samples of exceptional children (gifted, mentally retarded,

learning disabled, and speech/language delays) and clinical groups (Attention Deficit/

Hyperactivity Disorder, severe conduct disorder, and epilepsy) (Wechsler, 1991).

The reliability coefficient for the Coding subtest at age 7 is .70 with an average

reliability of .79 for the range of age 6 to 15 years. Validity for the Coding subtest is

based on its factor analytic studies showing that correlates with a Processing Speed factor

rather than Verbal or Performance factors. Additionally, in a factor analytic study of child

attention measures, Mirsky et al. (1991) found that the number correct on the Coding

subtest loaded on their "focus-execute" factor of attention along with number of

omissions and time to completion on a Digit Cancellation task.

Verbal Ability

Wechsler Intelligence Scale for Children-Third Edition (WISC-III; Wechsler,

1991). The WISC-III Comprehension, Information, Similarities, and Vocabulary subtests









were used as measures of verbal ability at age 7. Reliability coefficients for the four

subtests ranged from .72 to .79 for the 7-year-olds in the normative sample, and .67 for

the composite Verbal Comprehension factor (N = 200). The validity of the Information,

Comprehension, Vocabulary, and Similarities subtests as measures of verbal ability is

based on the moderate intercorrelations between the subtests (ranging from .46 to .64 for

7-year-olds in the normative sample), as well as factor analytic studies showing that the

four subtests load together on the same factor (Verbal Comprehension). Again, the

overall reliability and validity of the WISC-III is based on extensive research on its

predecessor, the WISC-R, correlational studies with other tests of ability and

neuropsychological tests, and studies using various special populations.

Wechsler Preschool and Primary Scale of Intelligence-Revised (WPPSI-R;

Wechsler, 1989). The WPPSI-R Comprehension, Information, Similarities, and

Vocabulary subtests were used as measures of verbal ability at age 5. The WPPSI-R is a

well-validated test of general cognitive functioning designed for children ages 3 years to

7 years, 3 months. The test was standardized on a national sample of 1,700 children

stratified by age, sex, race/ethnicity, geographic region and parent education and

occupation based on survey data gathered by U.S. Census Bureau in 1986. The WPPSI-R

has demonstrated good psychometric properties. For the age group of interest (5 years),

split-half reliability coefficients range from .59 to .86 for the individual subtests. Stability

coefficients during a test-rest interval of 3 to 7 weeks (N= 175) for the individual

subtests ranged from .53 to .81. The overall validity of the WPPSI-R is based, in part, on

studies of its predecessor, the WPPSI. In addition, the WPPSI-R has been evaluated using

factor analytic studies and correlational studies with the WISC-R, Stanford-Binet









Intelligence Scale-Fourth Edition (Thorndike, Hagen, & Sattler, 1986), the McCarthy

Scales of Children's Abilities (McCarthy, 1972) and the Kaufman Assessment Battery for

Children (Kaufman & Kaufman, 1983). Validity studies of the WPPSI-R have also been

conducted using samples of gifted, mentally deficient, learning disabled and

speech/language impaired children (Wechsler, 1989).

Reading Ability

Wechsler Individual Achievement Test (WIAT; Wechsler, 1992). The WIAT

Basic Reading and Reading Comprehension subtests were used as measures of reading

ability at age 7. The Basic Reading subtest has 55 items covering picture naming,

vocabulary, and single word reading. The Reading Comprehension subtest consists of 38

items that require the examinee to read a short passage and answer questions presented

orally by the examiner. The WIAT is a well-validated test of academic achievement for

children ages 5 years to 19 years, 11 months. The test was standardized on a national

sample of 4,252 children stratified by age, sex, race/ethnicity, geographic region and

parent education according to the 1988 U.S. Census (N= 331 for the age group of

interest, 7 years). A subgroup of 1,284 children from the WIAT standardization sample

was also administered one of the Wechsler intelligence scales. For the age group of

interest (7 years), 100 children were administered both the WIAT and the WISC-III. A

weighting procedure was used to assure that the scores for the subgroup were comparable

to those of the WISC-III standardization sample.

The WIAT consists of eight subtests; however, only the psychometric properties

of the Basic Reading and Reading Comprehension subtests will be reviewed here.

Overall, both subtests have demonstrated good psychometric properties. For 7-year-olds,

the split-half reliability coefficients for the Basic Reading and Reading Comprehension









subtests are .95 and .93, respectively. Stability coefficients for the Basic Reading and

Reading Comprehension subtests are .91 and .89 for grade 1 (N= 76) and .94 and .90 for

grade 2 (N = 74), respectively. Reliability and stability coefficients are even higher for

the overall Reading Composite, which is comprised of the two subtests. The age 7

reliability coefficient for the Reading Composite is .97, and grade 1 and grade 2 stability

coefficients are .95 and .96.

The content validity of the WIAT is based on reviews by curriculum experts and

empirical item analysis studies. The WIAT's construct and criterion-related validity was

determined by correlational studies using the WIAT subtests and other individually-

administered achievement tests. Across a variety of ages, the Basic Reading and Reading

Comprehension subtests were found to correlate from .79 to .87 (median = .82) with the

analogous subtests in five other achievement test batteries including the Wide Range

Achievement Test-Revised (Jastak & Wilkinson, 1984) and the Woodcock-Johnson

Psycho-Educational Battery-Revised Tests of Achievement (Woodcock & Johnson,

1991). Both subtests also have significant correlations (>.40) with school grades in a

sample of children ages 6 to 19 years (N= 867). In addition, studies of special groups of

children (including those identified as gifted, or having mental retardation, emotional

disturbance, learning disabilities, Attention Deficit Hyperactivity Disorder, and hearing

impairment) support the validity of the WIAT (Wechsler, 1992).

Caregiving Environment Measures

Home Observation for Measurement of the Environment (HOME; Caldwell &

Bradley, 1984). Seven subscales of the HOME were used as measures of the child's

caregiving environment at ages 5 and 7. HOME scores were determined by observation

by trained interviewers during interviews with the child caregivers in their homes. The









four subscales of the Early Childhood version of the HOME (EC HOME) most closely

related to literacy development-Learning Stimulation, Language Stimulation, Learning

Stimulation, and Variety in Experience-were used to assess the caregiving environment

at age 5. The three subscales of the Middle Childhood version of the HOME (MC

HOME) most related to literacy development-Growth Fostering Materials and

Experiences, Provision for Active Stimulation, and Family Participation-were used to

assess the caregiving environment at age 7. The HOME is a screening measure that

assesses factors related to the nurturance and stimulation in a child's home environment

that are believed to be important for cognitive development. The HOME was designed as

an alternative to sociodemographic factors for identifying children at "high risk" for

intellectual/academic problems. Scores are based on both the observer's visual inspection

of the home and self-report of the child's primary caregiver obtained through a semi-

structured interview during a 45- to 90-minute home visit. The HOME has been found to

be significantly correlated with longitudinal cognitive test performance and academic

achievement in children ages 3 to 10 years (Bradley, Caldwell, & Rock, 1988; Bradley,

1994; Bradley & Whiteside-Mansell, 1998).

Four forms of the HOME are available: an Infant-Toddler version (ages birth to 3

years), an Early Childhood version (ages 3 to 6 years), a Middle Childhood version (ages

6 to 10 years), and an Early Adolescent version (ages 10 to 14 years). Since the Early

Childhood (EC) and Middle Childhood (MC) versions were used in the present study,

only they are reviewed here. The EC HOME contains 55 items clustered in eight

subscales: 1) Learning Materials, 2) Language Stimulation, 3) Physical Environment, 4)

Parental Responsivity, 5) Learning Stimulation, 6) Modeling of Social Maturity, 7)









Variety in Experience, and 8) Acceptance of Child. The MC HOME contains 59 items

clustered into eight subscales: 1) Parental Responsivity, 2) Physical Environment, 3)

Learning Materials, 4) Active Stimulation, 5) Encouraging Maturity, 6) Emotional

Climate, 7) Parental Involvement, and 8) Family Participation.

The various versions of the HOME and their subscales have undergone name

changes, reorganization, or both and over time. However, no significant changes in the

number of items or item content have been made for the EC HOME and MC HOME.

When the data from the EC HOME were collected in the current study, it was called the

"Preschool" version. Since the structure of the items has not changed since the scale was

renamed, the current names for the EC HOME subscales will be used throughout this

study. Of the four EC HOME subscales used in the current study, only two were

relabeled when the parent scale was renamed. The Language Stimulation and Variety in

Experience subscales remained the same while the Learning Materials subscale was

previously labeled "Learning Stimulation" and the current Learning Stimulation subscale

was called "Academic Stimulation."

The MC HOME, called the "Elementary" version when the data for this study

were collected, underwent reorganization of its items into eight subscales rather than

seven subscales. The three subscales used in the current study were not affected by this

structural change. Thus, the current names for the MC HOME subscales will be used

throughout the study. Of the three MC HOME subscales used in the current study, one

was relabeled when the parent scale was renamed. The Learning Materials subscale was

previously called "Growth Fostering Materials and Experiences."









All four versions of the HOME have been found to have good psychometric

properties. Test-retest reliability, as measured by coefficient alpha, is above .90 for the

total scores and is generally higher for the longer than the shorter subscales. Interobserver

agreement is reported as 90% or higher for all versions. Concurrent and predictive

validity studies have shown that the HOME is significantly correlated with IQ, as high as

r = .58. Low to moderate correlations (.30 to .60) between EC HOME scores and

children's contemporaneous and later intellectual and academic performance have

generally been found (Bradley, 1994). Similar relationships have been reported for MC

HOME scores and children's school performance and classroom behavior (Bradley,

Caldwell, Rock, Hamrick, & Harris, 1988). These relationships have been found in

African American as well as European American samples (Bradley & Caldwell, 1981;

Bradley, Rock, Caldwell, Harris & Hamrick, 1987). While HOME scores have low to

modest correlations with a wide variety of demographic variables including race, family

structure, neighborhood, and maternal age, two studies have shown that no single

demographic factor accounts for much of the variance in HOME scores and that all the

demographic factors together only account for about 50% of the variance (Bradley &

Caldwell, 1981; Bradley, Mundfrom et al., 1994).

Procedure

Detailed information regarding participant recruitment and assessment of birth

outcome measures is provided in Eyler et al. (1998a, 1998b) and is summarized here.

Recruitment of participants took place between July 1991 and July 1993 with the last

child born in February, 1994. Institutional Review Board approval was obtained for study

procedures and incentives. Informed consent was carefully obtained for all participants,

including those who were illiterate. The consent process included an explanation of child,









maternal, and family measures, drugs tests and interviews, the Federal Certificate of

Confidentiality, and the distinction between the researchers and clinical providers in

assurance of confidentiality. All participants were recruited from women designated to

deliver at Shands Teaching Hospital, a tertiary care center.

Exclusion criteria included rare but major maternal illnesses diagnosed before

pregnancy that are known to affect pregnancy or developmental outcome, such as

diabetes, sickle cell disease, and mental retardation, as well as women who abused legal

drugs or used any illicit drugs other than cocaine and marijuana. In addition, only

mothers who spoke English and were equal to or greater than 18 years of age were

consented for enrollment in the study.

A priori participant matching criteria were developed during the original

longitudinal study to minimize the effect of possible confounding variables on pregnancy

or child outcomes. Four matching criteria for the control group were chosen, three of

which were based on characteristics that significantly differed between prenatal cocaine

users and the general obstetric population and which have been shown to relate to

pregnancy or developmental outcome. These three matching criteria were: 1) the level of

Hollingshead Index of SES, 2) racial/ethnic group membership (African American versus

other racial/ethnic categories), and 3) number of previous births multiparouss or

primiparous). The fourth matching criteria, location of prenatal care, was chosen to

equate groups on risk factors or complications that developed during, but not before,

pregnancy. This variable included the local public health unit, outlying clinics (which

sent only high-risk women to deliver at Shands) or no prenatal care.









The researchers approached 2,526 potential participants of whom 85% gave

informed consent to participate in the study. Most cocaine-using participants were

enrolled prenatally (75%) from the two closest county public health prenatal clinics or

from the hospital's high-risk referral prenatal clinic by approaching all non-excluded

potential participants. The additional participants were recruited when they arrived to

deliver at the hospital. This latter group consisted of women who had received no

prenatal care or those whom the researchers had been unable to interview in the prenatal

clinic. Of the 2,526 potential participants, 179 were approached for consent at delivery

and 89% gave informed consent. The 372 refusals included 13 women who were willing

to give consent for the study but were unable or unwilling to provide a urine specimen at

enrollment required for continued participation. After the first interview, 22 of the

women who consented (11 cocaine users and 11 nonusers who would have been potential

matches) were eliminated from study. Most (n = 16) were found to have used excluded

illicit drugs, while three reported using confounding prescription medications and another

three were no longer able to deliver at Shands Teaching Hospital.

As each cocaine-using woman who consented and met exclusion criteria was

identified, one or two participants from the pool who consented, who denied prenatal

cocaine use, and whose urine specimens showed no evidence of cocaine use were

selected for each match category. The oldest matched control from the appropriate

category was then used as the final match. The two final groups of participants consisted

of 154 cocaine-using women and 154 non-users.

The drug use interview, adapted from that of Day, Wagener, and Taylor (1985),

was administered by a one of a number of well-trained, non-judgmental interviewers who









attempted to establish rapport after informed consent was given by the maternal

participant. Interviewers carefully read and explained all portions of a detailed drug

history due to the low literacy level of the mothers enrolled in the study. Interviews were

conducted at the end of each trimester, whenever possible, and details about drug use

during the previous three months was probed in order to induce less guilt for the

participants. Women with very late or no prenatal care were interviewed after birth about

drug use throughout their pregnancy. Enrollment rates at the end of the first trimester, the

end of the second trimester, and at delivery were: 41% (41 cocaine; 84 control), 34% (61

cocaine, 44 control) and 25% (69 cocaine, 9 control), respectively. To trigger memory

around real-time events, calendars were used to help women recall their drug use history

within the context of her pregnancy. In addition to cocaine, participants were queried

about their use of drugs from several categories, including marijuana, alcohol, tobacco,

and other illicit drugs (the latter for exclusion purposes) using street or slang drug names.

The amount (or cost) and timing of each woman's usual use of each drug were recorded.

Increases and decreases in usage patterns were also noted in order to calculate a more

accurate average use per trimester.

Urine specimens were obtained for drug screening on two occasions that could

not be anticipated by the participants. The first specimen was collected on the day of

enrollment in the study. Women who consented but refused to provide a urine specimen

on the same day were dropped from the study. The second specimen was obtained from

the mother on the day of the baby's birth if an infant specimen was unavailable. A full

toxicology screen of the urine was conducted using fluorescence polarization









immunoassay. Positive drug screens were then confirmed with gas chromatography/mass

spectroscopy.

Study measures were administered to the children and their mothers or another

primary caregivers at several different assessments. Throughout the course of the

longitudinal study, all of the measures have been administered by trained, certified, or

licensed professionals blinded to the study group membership of the mother and child. In

the rare cases when a tester was unmasked, other backup testers were used for the

assessments.

The first set of child assessments occurred shortly after birth. Infants were

evaluated within the first day or as soon as they were well in the Shands Hospital Clinical

Research Center, which provided controlled conditions of light, sound, and temperature.

In a few cases in which the infants were unstable, they were evaluated in the nursery.

Orbitofrontal head circumference measurements were obtained by one of a team of

neonatal nurse practitioners blinded to the drug history of the mother. The Hobel

Obstetric Risk Scores were determined postdelivery by medical personnel trained by one

of the larger study's principal investigators, Marylou Behnke, M.D. The BNBAS was

administered midway between feedings as close to 40 weeks postconceptual age as

possible. In the current sample the majority of infants (69%) were evaluated within 24

hours after birth. Another 20% were administered the BNBAS within 48 hours after

birth. The BNBAS was administered by certified, reliable evaluators blinded to the drug

history of the mothers.

The second set of assessments took place when the children were approximately 5

years old. The Early Childhood Home Observation for Measurement of the Environment









(EC HOME) was completed by one of two trained female interviewers during interviews

conducted with the primary caregivers in the family's home. The age 5 cognitive test

battery, including the IVA, Letter Cancellation, and WPPSI-R verbal subtests, was

administered by one of two licensed school psychologists in private practice who were

blinded to the child's group membership. Administration of the cognitive battery took

place on the Project Care bus while it was parked either on the grounds of the child's

school or outside the child's home. There was no significant difference in the mean age at

which the two study groups were administered the age 5 test battery [t(240) = .08, p =

.93] The combined average of the groups' ages were 5.34 years at the time of the first

cognitive test battery.

The third sets of assessments occurred when the children were approximately 7

years old. The Middle Childhood Home Observation for Measurement of the

Environment (MC HOME) was completed during interviews conducted with the primary

caregivers in the family's home by the same interviewers who conducted the age 5

interviews. The age 7 speech and language assessment, which included the Wechsler

Individual Achievement Test (WIAT) reading subtests, was administered following a

physical examination by one of two licensed nurse practitioners blinded to the child's

group membership. The age 7 speech and language assessment was completed by on the

Project Care bus while it was parked on the grounds of the child's school or outside the

child's home. The age 7 cognitive test battery also took place on the Project Care bus and

was typically done within a day or two of the physical exam and speech and language

assessment. The same two blinded, licensed school psychologists who administered the

age 5 cognitive test battery also gave the age 7 cognitive test battery. The children were









provided breaks, including snacks, routinely and as needed during administration of the

test batteries. There were no significant differences in the mean age at which the two

study groups were administered the age 7 speech and language or cognitive test batteries

[ts(240) = 1.04, .40, ps = .30, .69, respectively]. The combined averages of the groups'

ages were 7.31 and 7.29 years during the speech and language and cognitive batteries,

respectively.

After assessment protocols were reviewed for scoring accuracy, the scores for

each participant were hand-entered into a Microsoft Access database. To minimize input

errors, the data were entered a second time and checked against the original input.

Discrepancies between the two sets of entries were reconciled by checking the protocols.

Age-corrected scaled scores for the WPPSI-R and WISC-III verbal ability

subtests and age-corrected standard scores for the WIAT reading subtests were used. For

age 5 Letter Cancellation, age 7 Letter Cancellation and TMT Trail A, time in seconds to

complete the task was used. Raw scores were used for the EC and MC Home subscales.

For structural equation modeling (SEM), it is not necessary for all of the variables to be

in the same metric as the solution can be standardized by setting factor variances to one.

Hypotheses

Based on a review of the relevant literature, three a priori hypotheses were

developed to examine the relationship between prenatal cocaine exposure and the

development of attention and reading skills in children:

1. Performance on the BNBAS Habituation, Orientation, and State Regulation
Supplementary Scales will be significantly correlated with performance on age 5
and 7 attention measures for both groups of children in the study.

2. Children with prenatal cocaine exposure (PCE) will perform significantly worse
than matched controls on: a) attentional measures at age 5 (IVA CPT and Letter
Cancellation), b) attentional measures at age 7 (IVA CPT, Letter Cancellation,









TMT Part A, WISC-III Coding, and WISC-III Digit Span), and c) reading
measures at age 7 (WIAT Basic Reading and Reading Comprehension subtests).

3. It is hypothesized, regardless of significant group differences on measures of
attention and reading, that performance on attentional measures at birth, age 5,
and age 7 will be significant predictors of reading at age 7 after controlling for
verbal ability and the caregiving environment. It is hypothesized that PCE will
have both a direct effect on attention and an indirect effect on attention that is
mediated by birth head circumference. It is also hypothesized that PCE will have
an indirect effect on reading at age 7 that is mediated through attention.

Data Inspection and Analyses

Data Screening

The first step in the data analysis plan involved screening the data for violations

of normality and for missing values. Data screening was conducted using SPSS and

PRELIS 2.52 (Joreskog & Sorbom, 200 b). With the exception of the four drug

variables, data with significant skewness, kurtosis, or both were transformed into normal

scores. The drug variables were not transformed as it was expected that these data would

not have a normal distribution.

Missing Data

A large number of missing values were found for the Brazelton Neonatal

Behavioral Assessment Scale (BNBAS) Habituation, Orientation, and Regulation of State

scores. Specifically 83 participants (n = 39 for PCE group, n = 44 for non-exposed group)

were missing the Habituation score, 45 participants were missing the Orientation score (n

= 30 for PCE group, n = 15 for non-exposed group), eight were missing both the

Habituation and Orientation scores (n = 5 for the PCE group, n = 3 for the non-exposed

group), and 20 participants were missing all three scores (n = 15 for PCE group, n = 5 for

non-exposed group). As reported in Eyler et al. (1998b), a significantly larger proportion

of infants with PCE than non-exposed infants failed to come to a quiet, alert state so that









the Orientation items could be administered (25% vs. 12%). Independent sample t-tests

using Bonferroni correction for multiple comparisons were conducted between the groups

of children with missing BNBAS scores and the rest of the sample. Results indicated no

significant differences between the groups missing the Habituation score, Orientation

score, or both scores and the rest of the sample on any of the demographic variables, head

circumference, or gestational age. For the group missing all three BNBAS scores, one

significant difference was found from the rest of the sample. There was a larger

proportion of African Americans in the group missing all three BNBAS scores than in the

rest of the sample [t(26) = -2.41, p = .019]. It was decided to use all available BNBAS

data to evaluate the first hypothesis but to exclude the BNBAS from the path analysis

needed to evaluate the third hypothesis.

A number of missing values were also found for the Intermediate Visual and

Auditory Continuous Performance Test (IVA CPT) Visual Attention Quotient (VAQ) and

Visual Response Control Quotient (VRCQ) scores at both ages 5 and 7. Specifically, 17

children were missing the IVA CPT at age 5 (n = 12 for PCE group, n = 5 for non-

exposed group), 17 children were missing the IVA CPT at age 7 (n = 8 for PCE group, n

= 9 for non-exposed group), and one child (non-exposed) was missing both sets of IVA

CPT scores. Independent sample t-tests using Bonferroni correction for multiple

comparisons were conducted between the groups of children with missing IVA CPT

scores and the rest of the sample. Results indicated no significant differences between the

groups at age 5 or at age 7 on any of the demographic variables, head circumference, or

measures of attention, reading, verbal ability, or the caregiving environment. It was









decided to use all available IVA CPT data to evaluate the second hypothesis but to

exclude these data from the path analysis needed to evaluate the third hypothesis.

Not including the BNBAS or IVA CPT scores, 29 participants were found to be

missing only one of the remaining 30 data points. The distribution of missing scores was

as follows: Hobel prenatal risk score (n = 6), age 5 Letter Cancellation (n = 11), age 5

HOME Academic Stimulation subscale (n = 1), and age 7 Trail Making Test (TMT) Part

A (n = 11). With the exception of the TMT, the missing data points appeared to be

random, so the mean score based on the child's group membership (PCE vs. non-

exposed) was substituted for the missing data.

For the TMT, the missing scores were generally the result of the child not being

able to count to 15 or not being able to complete the practice item (n = 5 for PCE group,

n = 6 for non-exposed group). Since it is somewhat unusual for a 7-year-old not to be

able to count to 15, independent-samples t-tests between the group of children missing

the TMT and the rest of the sample were conducted. Results revealed significant

differences on 14 variables as shown in Table 2-4. A significant difference was found for

ethnicity; all of the children who could not do the TMT Part A were African American.

These children also scored significantly worse than the rest of the sample on the Hobel

prenatal risk score; EC HOME Language Stimulation subscale; age 5 and age 7 Letter

Cancellation; WPPSI-R Comprehension, Information, and Similarities; WISC-III Digit

Span, Comprehension, Information, Similarities, and Vocabulary; and WIAT Basic

Reading and Reading Comprehension. Since the differences between the group missing

the TMT and the rest of the sample were significant, a decision was made to replace the









missing TMT Part A scores with a arbitrary score of 300 seconds, the time limit for

discontinuance of the test.

Accounting for the Participants

After imputing values for participants who were missing only one data point, a

total sample size of 240 participants (n = 120 for both groups) remained for further

analyses. Thus, a total of 68 of the 308 participants originally enrolled in the prospective,

longitudinal study were excluded from the current study. Of these 68 participants, 22 had

more than one missing data point, 12 died prior to age 7, eight dropped out of the study,

10 were lost to follow up, six had moved out of the area, eight refused to participate in

one or both of the age 5 and age 7 assessments, one child was deaf, and one child was

profoundly retarded and could not complete the assessments.

To determine whether the differences found in the original study sample were also

present in the smaller sample used in the current study, independent samples t-tests with

Bonferroni correction for multiple comparisons were performed for the demographic

variables and head circumference. Table 2-6 shows the results of the analyses, which

revealed that the sample for the current study was very similar to the original sample. As

in the original sample, the group with PCE had significant greater mean amounts of

prenatal cocaine exposure [t(238) = 17.43,p = .000], prenatal alcohol exposure [t(238) =

5.69, p = .000], and prenatal tobacco exposure [t(238) = 7.30, p = .000] than the non-

exposed group. In addition, the group with PCE had significantly higher mean Hobel

prenatal risk score than the non-exposed group, as in the original sample [t(238) = 5.11, p

= .000] and significantly smaller mean head circumference as compared to the non-

exposed group [t(238) = -3.26, p = .001]. Finally, statistically similar proportions of









females and African Americans were found in both study groups [ts (238) = 1.42 and -

1.20, ps = .156 and .232, respectively].

There were, however, two differences between the current sample and the original

study sample. Unlike the original sample, there was no significant difference between the

groups in mean amount of prenatal marijuana exposure [t(238) = 2.06, p = .041]. Another

difference between the current sample and the original sample was the significantly

shorter mean gestational age for the group with PCE compared to the non-exposed group

[t(238) = -3.27, p = .001]. The average gestational age of the infants with PCE was 38.50

weeks compared to 39.29 weeks for the non-exposed infants. While statistically

significant, the difference between the groups was less than one week.

Statistical Analyses

SPSS and LISREL 8.52 (Joreskog & Sorbom, 2001a) were used to conduct all

statistical analyses. The criterion for significance tests for all a priori hypotheses was set

at c = .05 with Bonferroni correction for multiple comparisons. To test the first

hypothesis, a correlational analysis was conducted to determine whether the BNBAS is

significantly related to measures of attention at age 5 and age 7. To test the second

hypothesis of group differences on measures of attention and reading, a cross-sectional

analysis using the independent samples t-test was conducted. Finally, to test for

longitudinal associations between measures of attention administered at birth, age 5 and

age 7 and reading ability at age 7 and the hypothesis that head circumference mediates

the effect of PCE on attention and reading, structural equation modeling (SEM) was used

in a combined sample of the children with PCE and without PCE.









SEM is a powerful statistical technique that involves multiple regression analyses

of factors. Factors, called latent constructs, were derived from the reliable shared

variance of one or more observed variables indicators and thus are free of measurement

error. SEM allows for the examination of complex relationships between multiple

continuous and discrete independent variables and multiple continuous and discrete

dependent variables. After specification of a model, which is a type of confirmatory

factor analysis, SEM can be used to test a model, test specific hypotheses about a model

(including mediational hypotheses), modify an existing model, or test a set of related

models (Ullman, 2001).

To ensure sufficient identification to conduct structural equation modeling (SEM),

the number of unknown parameters must be less than or equal to the number of known

pieces of information supplied to the program. In general, then, the number of indicator

variables should be limited. In addition, to analyze longitudinal data, a stable number of

participants are required across all time points. For the 18 participants who were

randomly missing one of the variables in the study, the missing data was imputed using

the mean score based on group membership (PCE vs. non-exposed). For the 11

participants who were missing the TMT, an arbitrary score of 300 seconds, the time limit

for discontinuance of the test, was substituted for the missing value. Participants missing

more than one data point were excluded from the analyses.

Figures 2-1 and 2-2 provide diagrams of the proposed structural model of the

longitudinal relationship between PCE, attention, and reading. Eight of the factors in the

model are indicated by a single variable: prenatal cocaine exposure (COCAINE), prenatal

alcohol exposure (ALCOHOL), prenatal tobacco exposure (TOBACCO), prenatal









marijuana exposure (MARIJUA), Hobel prenatal risk (HOBEL), sex (SEX), head

circumference at birth (HEADC), Letter Cancellation at age 5 (LCAN5), and Digit Span

at age 7 (DSPAN7). Attention at age 5 is singly indicated by Letter Cancellation

(LCAN5). Attention at age 7 was divided into two separate factors. Digit Span was

allowed to be a singly indicated factor (DSPAN7) because it does not have a visuomotor

component while the other three attentional measures-Letter Cancellation, TMT Part A,

and WISC-III Coding subtest-which have a visuomotor component were combined into

a Visual Attention factor (VATTN7). The age 5 caregiving environment factor (HOMES)

was indicated by four subscales from the EC HOME while the age 7 caregiving

environment factor (HOME7) was indicated by three subscales from the MC HOME. The

age 5 and age 7 Verbal Ability factors (VERBAL5 and VERBAL7) were indicated by the

four age-appropriate Wechsler subtests-Comprehension, Information, Similarities, and

Vocabulary.

The structure of the model was based on the predictive relationships that are

expected to exist between the various factors. The six endogenous variables were used to

predict birth head circumference, which in turn is used to predict the three age 5 factors

(LCAN5, VERBALS, and HOMES). As expected with longitudinal data, each of the

three age 5 factors (attention, verbal ability, and caregiving environment) was used to

predict their respective age 7 factors. As the only measure of attention obtained at age 5,

Letter Cancellation factor was used to predict the two age 7 attention factors, DSPAN7

and VATTN7. The other two age 5 factors, VERBALS and HOMES, are used to predict

their respective age 7 factors, VERBAL7 and HOME7. Finally, the four age 7 factors






53


(DSPAN7, VATTN7, VERBAL7, and HOME7) were used to predict the final outcome,

age 7 reading ability (READ7).









Table 2-1 Continuous Variables that Differed Significantly Between Mothers in the Two
Original Study Groups
Group


Variable

Age
Week entered prenatal care
Total Hobel score
Prenatal Hobel score


Cocaine Users
M SD
27.6 4.8
14.8 7.6
94.2 72.1
54.5 20.1


Matched Controls
M SD
23.8 5.5
12.1 7.2
78.5 48.2
43.0 19.3


an = 154 for both groups.
* p < .05, ** p < .01, *** p < .001, two-tailed using the independent samples t-test.


Table 2-2 Non-continuous Variables that Differed Significantly Between Mothers in the
Two Original Study Groups
Group
Variable Cocaine Users Matched Controls p-value
Tobacco users 123 37 .0001***

Alcohol users 118 47 .0001***
Marijuana users 68 11 .0001***

Births < 37 weeks gestation 28 14 .03*

an = 154 for both groups.
* p < .05, ** p < .01, *** p < .001, two-tailed using the independent samples t-test.


p-value

.0001***
.003**
.0276*
.00001***









Table 2-3 Variables that Differed Significantly Between Neonates in the Two Original
Study Groups
Group


Variable

Birth weight (g)
Length (cm)
Chest circumference (cm)
Head circumference (cm)


Cocaine-Exposed
M SD
2985 668
48.7 3.2
31.6
33.6


Matched
M
3179
49.7
33.6
34.8


Controls
SD
700
3.3


p-value

.03*
.007**
.01*
.007**


an = 154 in both groups.
* p < .05, ** p < .01, two-tailed using the independent samples t-test.


Table 2-4 Summary of Variables for Current Study


Variable Name

WUCFULL

POAALC

POATOB

POAMAR

HEADC
HOBPRE
SEX

BRHABIT
BRORIENT
BRREGSTA
IVAVAQ5

IVAVRCQ5

LCANT5
IVAVAQ7

IVAVRCQ7

LCANT7
TRAILA
W3COD


Variable Label Coi
Demographic Variables
Average ratio of weeks of cocaine C
use
Average ounces of absolute A
alcohol consumed per day
Average number of cigarettes T
smoked per day
Average number of marijuana M
joints smoked per day
Head circumference He
Hobel Prenatal Risk
Sex


Child Cognitive Variables
BNBAS Habituation
BNBAS Orientation
BNBAS Regulation of State
IVA CPT Visual Attention
Quotient
IVA CPT Visual Response
Control Quotient
Letter Cancellation time
IVA CPT Visual Attention
Quotient
IVA CPT Visual Response
Control Quotient
Letter Cancellation time
TMT Trail A time
WISC-III Coding


instruct Assessed

cocaine exposure

alcohol exposure

tobacco exposure

marijuana exposure

ad circumference
Prenatal risk
Sex


Attention
Attention
Attention
Attention

Attention

Attention
Attention

Attention

Attention
Attention
Attention


Age


prenatal + 3
months prior
prenatal

prenatal

prenatal

birth
birth
birth


<1 week
<1 week
<1 week
5 years

5 years

5 years
7 years

7 years

7 years
7 years
7 years









Table 2-4. Continued


Variable Name
W3DSS
WPCOM
WPINF
WPSIM
WPVOC
W3COM
W3INF
W3SIM
W3VOC
WIATBR
WIATRC


Variable Label
WISC-III Digit Span
WPPSI-R Comprehension
WPPSI-R Information
WPPSI-R Similarities
WPPSI-R Vocabulary
WISC-III Comprehension
WISC-III Information
WISC-III Similarities
WISC-III Vocabulary
WIAT Broad Reading
WIAT Reading Comprehension


Construct Assessed
Attention
Verbal ability
Verbal ability
Verbal ability
Verbal ability
Verbal ability
Verbal ability
Verbal ability
Verbal ability
Reading ability
Reading ability


Caregiving Environment Variables
H5LEAR EC HOME Learning Materials Home environment 5 years
H5LANG EC HOME Language Stimulation Home environment 5 years
H5ACAD EC HOME Language Stimulation Home environment 5 years
H5VARI EC HOME Variety in Experience Home environment 5 years
H7GROW MC HOME Learning Materials Home environment 7 years
H7ACTI MC HOME Active Stimulation Home environment 7 years
H7FAMI MC HOME Family Participation Home environment 7 years


Note. BNBAS


Brazelton Neonatal Behavior Assessment Scale, IVA CPT = Intermediate Visual


and Auditory Continuous Performance Test, TMT = Trail Making Test, WISC-III = Wechsler
Intelligence Scale for Children-Third Edition WPPSI-R = Wechsler Preschool and Primary
Scale of Intelligence-Revised, WIAT = Wechsler Individual Achievement Test, EC HOME =
Early Childhood Home Observation for Measurement of the Environment, MC HOME = Middle
Childhood Home Observation for Measurement of the Environment.


Age
7 years
5 years
5 years
5 years
5 years
7 years
7 years
7 years
7 years
7 years
7 years










Table 2-5 Significant Differences Between Participants With and Without TMT Part A
Scores
Group


Variable


Ethnicityb
Hobel prenatal risk score
EC HOME Language Stimulation
Letter Cancellation age 5 secss.)
WPPSI-R Comprehension
WPPSI-R Information
WPPSI-R Similarities
Letter Cancellation age 7 secss.)
WISC-III Digit Span
WISC-III Comprehension
WISC-III Information
WISC-III Similarities
WISC-III Vocabulary
WIAT Basic Reading
WIAT Reading Comprehension


With TMT
M SD
.82 .38
49.83 20.50
6.36 .96
102.53 43.30
7.92 2.48
7.22 2.48
7.85 2.24
58.70 20.94
8.94 2.59
8.20 3.27
8.17 2.54
8.27 4.49
8.44 2.62
96.52 12.35
92.24 11.67


Without TMT


M
1.00
41.36
5.73
131.09
5.27
4.18
5.91
80.27
4.18
4.90
4.81
3.00
5.00
82.82
0.64


SD
.00
11.64
1.49
38.78
1.19
1.25
1.97
28.89
1.47
2.47
.98
2.45
2.41
1.47
2.91


Note. EC HOME = Early Childhood Home Observation for Measurement of the Environment,
TMT = Trail Making Test, WPPSI-R = Wechsler Preschool and Primary Scale of Intelligence-
Revised, WISC-III = Wechsler Intelligence Scale for Children-Third Edition, WIAT = Wechsler
Individual Achievement Test.
an = 229 for group with TMT scores, n = 11 for group without TMT scores.
bEthnicity was coded so that African American = 1 and others = 0. Thus, the number in the table
represents the proportion of the group that was African American.
* p < .05, **p < .01, ***p < .001, two-tailed using the independent samples t-test and Bonferroni
correction for familywise error rate.


p-value

.000***
.042*
.038*
.033*
.001**
.000**
.005**
.001**
.000*
.001**
.000***
.000***
.000***
.000***
.000***










Table 2-6 Demographic Variables Comparing Groups in Current Study
Group


Variable


Prenatal cocaine exposure
Prenatal alcohol exposure
Prenatal tobacco exposure
Prenatal marijuana exposure
Hobel prenatal risk score
Gestational age (weeks)
Head circumference (cm)
Sexb
Ethnicityc


PCE
M SD
.458 .288
.232 .414
8.96 9.23
.009 .369
55.79 19.66
38.50 2.10
33.49 1.93
.52 .50
.80 .40


Non-exposed


M
.000
.002
1.86
.002
43.09
39.29
34.24
.43
.86


SD
.000
.005
5.33
.127
18.84
1.61
1.66
.50
.35


Note. PCE = prenatal cocaine exposure.
an = 120 for both groups.
bSex was coded so that female = 1, male = 0. Thus, the numbers in this row indicate the
proportion of the group that was female.
CEthnicity was coded so that African American = 1, others = 0. Thus, the numbers in this row
indicate the proportion of the group that was African American.


p-value

.000***
.000***
.000***
.041*
.000**
.001*
.001**
.156
.232













C a
IIa 41


0ca




Ca 0


SoCao C
s "







cn
co




..2






o















S2 C
C a c o


0 II t


l '








2 -< a -




So aum Ls






T3 '0 C O



Wm ^^
."""e~
F-.2 I












.9 a '3r














ct "
cEa


o C o ao



o 0
a II a



0 3 4 -

" -





00








4.o o
tII
0 -3 0 c O











"" *a o
ca&9 I
n 0 8 II
2 e 0







8.0
Sa ac,




03 a
-,r


2 a o 2 ) 1
a~g 0 of 1 l!
60 a a1
itt < E
&S2Hoo
FI ^Pfi














CHAPTER 3
RESULTS

Hypothesis #1. To test the first hypothesis-that the BNBAS Habituation, Orientation,

and Regulation of State scores obtained during the first week of life would have significant

relationships with measures of childhood attention at ages 5 and 7-a correlational analysis was

performed using all available data for both groups. There was only one significant correlation

between one of the BNBAS measures and one of the early childhood attention measures. The

BNBAS Orientation score was significantly correlated with the age 5 IVA CPT visual attention

quotient; while significant, this association was small (r = .149, p < .05, N= 191). No significant

relationships were found between the three BNBAS scores and Letter Cancellation at age 5, or

any of the age 7 attention measures (Letter Cancellation, IVA CPT scores, TMT Part A, WISC-

III Coding, or WISC-III Digit Span). Intercorrelations between all of the proposed attention

measures in the study for the combined sample of children with PCE and non-exposed children

are shown in Table 3-1.

A post-hoc analysis was conducted to examine the relationship of the BNBAS

Orientation score with the age 5 IVA CPT visual attention quotient separately by group (PCE vs.

non-exposed). When analyzed separately by group, the relationship between the BNBAS

Orientation score and the age 5 IVA CPT visual attention quotient became non-significant.

However, it appears that the significant relationship found in the combined sample was driven

largely by the PCE group (n = 80, r = .217, p = .053) rather than the non-exposed group (n =

111, r = .095,p = .32).









Hypothesis #2. To evaluate the second hypothesis that children with PCE would perform

significantly worse than non-exposed children on measures of attention at ages 5 and 7 and on

measures of reading at age 7, independent samples t-tests with Bonferroni correction for multiple

comparisons were performed. As shown in Table 3-2, no significant differences were found

between the groups on any of the attention or reading measures. For age 5 Letter Cancellation,

the PCE group took an average of 108 seconds to complete the task, compared to 100 seconds

for the non-exposed group [t(238) = 1.44, p = .15]. For each of the age 7 attention measures, the

PCE and non-exposed groups obtained remarkably similar mean scores. For Digit Span and

Coding, the scaled scores for the PCE group were 8.71 and 10.14, respectively, while the scores

were 8.73 and 10.56 for the non-exposed group, respectively [ts(238) = -.06, and -.93, ps = .95

and .35, respectively]. For the other two time-based measures, Letter Cancellation and TMT Part

A, the PCE group completed the tasks in 60 and 52 seconds on average while the non-exposed

group required 59 and 54 seconds on average [ts(238) = -.36 and -.21, ps = .71 and .83,

respectively]. Finally, on the two reading measures, both groups performed in the average range,

with the PCE group having slightly higher means scores than the non-exposed group. Scaled

scores were 96.20 and 95.59 for the Basic Reading subtest and 92.02 and 91.39 for the Reading

Comprehension subtest for the PCE and non-exposed groups, respectively (ts(238) = .39 and .42,

ps = .70 and .68).

Post hoc comparisons between the two groups were made for the remainder of the study

variables using independent samples t-tests with Bonferroni corrections for multiple

comparisons. As shown in Table 3-3, only two statistically significant group differences were

found. One was for the EC HOME Learning Stimulation subscale [t(238) 2.62, p = .01]; the

other was for the MC HOME Growth Fostering Materials subscale [t(238) = 3.12, p = .00].









Ironically, the PCE group scored slightly higher than the non-exposed group on both measures.

The PCE group obtained average scores of 7.41 for the EC HOME Learning Stimulation

subscale and 5.11 for MC HOME Growth Fostering Materials subscale compared to respective

means of 6.58 and 4.67 for the non-exposed group. This disparity is likely due to the fact that a

significantly larger proportion of the children with PCE were living in placements away from

their biological mothers at age 5 compared to non-exposed children (64% vs. 9%). Since there

was only two significant between-groups differences, combining the PCE and non-exposed

groups for all further analyses was deemed appropriate to maximize statistical power.

Hypothesis #3. To evaluate the third set of hypotheses that a) performance on measures

of attention will be significant predictors of performance in reading, b) PCE will have a direct

effect on attention and an indirect effect on attention mediated by head circumference, and c)

PCE will have an indirect effect on reading mediated by attention, structural equation modeling

(SEM) was performed on the combined sample of children with and without PCE. As noted

earlier, the Brazelton Neonatal Behavioral Assessment Scale (BNBAS) and Intermediate Visual

and Auditory Continuous Performance Test (IVA CPT) were dropped from these analyses due to

large numbers of missing data.

The first step in SEM is the development of the measurement model that, by definition,

allows all the factors in the model to be correlated. Several indices were used to assess the

goodness-of-fit of both the measurement and structural models. One index used was the ratio

between the chi-square statistic and the degrees of freedom for the model. Generally, a model is

considered to fit well if chi-square is less than twice the degrees of freedom (x2 < 2df). Other

indices used to assess fit in the current study are Joreskog and Sorbom's (1989) goodness-of-fit

(GFI), Bentler's (1990) normed comparative fit (CFI) and Bentler and Bonett's (1980)









nonnormed fit (NNFI) indices. For each of these indices, better fit is associated with higher

values, and .90 is generally considered a minimum acceptable level (Bentler & Bonett, 1980).

Finally, the root mean square error of approximation (RMSEA) takes into account the error of

approximation in the population, while the root mean square residual (RMR) provides a measure

of the average size of the residual difference between the actual covariances among the observed

indicators and the covariances predicted by a particular model. For both the RMSEA and RMR,

a value of 0.05 is considered an indicator of good fit (Browne and Cudeck, 1993; Bryant and

Yarold, 1995).

The initial measurement model was checked for signs of underidentification including

negative variances, correlations greater than 1.0, and factor loadings or correlations that seemed

to have the wrong sign or were much smaller or much larger than expected. No signs of model

underidentification were detected. Table 3-4 shows fit indices for the iterative process used to

determine the final measurement model. The basic measurement model (Ml), in which all

variables were allowed to correlate, fit the data very well: X2 (281, N= 240) = 376.72, GFI = .90,

CFI= .98, NNFI = .98, RMSEA = .04, and RMR = .05. No modifications of the measurement

model were needed since all of the fit indices met their respective criteria for good fit.

The next step in SEM is construction of a structural model that fits the data as well as the

final measurement model with fewer estimated parameters. The hypothesized structural model

(S1) also fit the data moderately well [X2 (345, N= 240) = 536.37, GFI= .87, CFI= .97, NNFI=

.96, RMSEA = .05, and RMR = .09] but significantly worse than the final measurement model.

An iterative process was undertaken to improve the model's fit first by dropping insignificant

paths one at a time, then inspecting the modification indices to determine additional parameters

to freely estimate. Table 3-4 displays the fit indices of the intermediate models between the









initial hypothesized structural model and the final structural model. In the first four model steps,

insignificant paths were dropped between the following factors: 1) Hobel and Head

Circumference, 2) Head Circumference and age 5 Letter Cancellation, 3) Head Circumference

and age 5 caregiving environment, and 4) age 7 Digit Span and Reading. As expected, these

changes did not significantly improve the fit of the model but did increase the degrees of

freedom for subsequent tests of model fit.

In the fifth model step (S6), a path was added between age 5 Verbal Ability and age 7

Digit Span, revealing a moderate relationship between the two factors (3 = .52). Adding this path

significantly improved the model fit as compared to the previous structural model: X2 difference

(1, N= 240) = 41.53, p = .00. Further modifications were still needed, however, because the

structural model was still significant different from the measurement model: X2 difference (67, N

= 240) = 120.53, p = .00. In the sixth model step (S7), the insignificant path between age 5 Letter

Cancellation and age 7 Digit Span was dropped with no significant change in the model's fit.

In the seventh model step (S8), estimating the path between age 5 Verbal Ability and age

7 Visual Attention revealed another moderately strong relationship (P = .47). Again, estimating

this additional path significantly improved the model fit: X2 difference (1, N= 240) = 25.09, p =

.00. However, this model was still significantly different from the measurement model: X2

difference (67, N= 240) = 96.55,p < .05. In the final model step, a path was added between Sex

to age 7 Visual Attention. The magnitude of this relationship was small (P = .29), indicating that

girls performed better than boys on the visual attention measures (females were coded as 1,

males were coded as 0). Adding this final path also significantly improved the model fit over the

previous structural model: X2 difference (1, N= 240) = 15.74, p = .00). No further attempts were









made to improve the fit of the structural model since it was no longer differed significantly from

the measurement model: X2 difference (66, N= 240) = 80.81, p > .05).

The final structural model fit the data well [2 (347, N= 240) = 457.53, GFI= .89, CFI=

.98, NNFI = .98, RMSEA = .04, and RMR = .06] and was a significant improvement over the

initial hypothesized structural model with improvements in all five fit indices. It should be noted,

however, that two of the fit indices did not meet the criteria for good fit: GFI was < .90 and the

RMR was > .05. Nevertheless, the final structural model was able to account for 44% of the

variance in the age 7 Visual Attention factor and 68% of the variance in the age 7 Reading

factor. As predicted, the Visual Attention factor was the strongest predictor of age 7 Reading

even after controlling for verbal ability and the caregiving environment (ps = .44, .41, and .14).

In summary, the first hypothesis that the BNBAS scores collected during the first week

would be significantly correlated with early childhood measures of attention and reading could

not be evaluated due to a large amount of missing data. The second hypothesis, that children with

PCE would perform significantly worse than non-exposed children on measures of attention at

ages 5 and 7 and reading at age 7, was not supported. There was mixed support for the third set

of hypotheses regarding relationships between attention and reading, PCE and attention, and

PCE and reading. Visual attention at age 7 was found to be the strongest predictor of reading at

age 7; however, PCE had no direct relationship with attention at age 5 or 7. PCE was found to

have an indirect effect on reading at age 7 mediated by head circumference at birth, verbal ability

and visual attention.
















































































o o~ o~ ha


c-I l p, p


Cc,
b0


cl H

iut
IQ)


'U

00

00


CCC#

IQ)Q


cK Cl
C3Cj


II
H .





























So
Pi











H


0




H





0.
o
*aV




-v









Table 3-2 Group Means and Standard Deviations for Early Childhood Attention and
Reading Variables
Group
Variable PCE Non-exposed p-value
M SD M SD
IVA CPT VAQ age 5b 44.96 30.31 49.12 27.01 .261
IVA CPT VRCQ age 5b 62.15 37.46 66.72 34.29 .323
Letter Cancellation age 5 secss.) 108.31 43.28 99.41 43.15 .112
IVA CPT VAQ age 7' 56.31 18.32 58.04 20.87 .488
IVA CPT VRCQ age 7 73.80 25.69 77.97 23.17 .183
Letter Cancellation age 7 secss.) 60.22 21.75 59.15 21.86 .705
TMT Part A secss.) 54.92 51.16 57.2 63.08 .629
WISC-III Coding 10.14 3.62 10.56 3.32 .347
WISC-III Digit Span 8.71 2.79 8.72 2.68 .978
WIAT Broad Reading 96.34 12.04 95.43 12.79 .569
WIAT Reading Comprehension 92.22 11.48 91.20 11.87 .499

Note. PCE = prenatal cocaine exposure, IVA CPT = Intermediate Visual and Auditory Continuous
Performance Test, VAQ = Visual Attention Quotient, VRCQ = Visual Response Control Quotient, TMT =
Trail Making Test, WISC-III = Wechsler Intelligence Scale for Children-Third Edition, WIAT = Wechsler
Individual Achievement Test.
an = 120 for both groups for all measures except IVA CPT scores. bn = 116 for PCE group, n = 126 for
nonexposed group, "n = 123 for both groups.









Table 3-3 Group Means and Standard Deviations for All Other Study Variables
Group
Variable PCE Non-exposed p-value
M SD M SD
Caregiving Environment- age 5
EC HOME Academic Stimulation 4.49 .81 4.41 .93 .456
EC HOME Language Stimulation 6.48 .89 6.19 1.07 .022
EC HOME Learning Stimulation 7.40 2.36 6.59 2.51 .010*
EC HOME Variety in Experience 6.39 1.54 6.17 1.31 .237
Caregiving Environment- age 7
MC HOME Active Stimulation 3.80 1.89 3.67 1.77 .581
MC HOME Family Participation 6.38 2.08 6.48 2.10 .719
MC HOME Growth Fostering 5.12 1.65 4.45 1.53 .001**
Materials and Experiences
Verbal Ability age 5
WPPSI-R Comprehension 7.87 2.43 7.71 2.57 .574
WPPSI-R Information 7.27 2.45 6.90 2.59 .259
WPPSI-R Similarities 8.92 2.52 8.46 2.42 .155
WPPSI-R Vocabulary 7.82 2.31 7.71 2.23 .698
Verbal Ability age 7
WISC-III Comprehension 7.86 3.36 8.24 3.25 .383
WISC-III Information 8.08 2.63 7.96 2.55 .723
WISC-III Similarities 8.05 4.32 8.00 4.80 .944
WISC-III Vocabulary 8.15 2.80 8.41 2.61 .462

Note. PCE = prenatal cocaine exposure, TMT = Trail Making Test, WIAT = Wechsler Individual
Achievement Test, EC HOME = Early Childhood Home Observation for Measurement of the Environment,
WPPSI-R = Wechsler Preschool and Primary Scale of Intelligence-Revised, MC HOME = Middle
Childhood Home Observation for Measurement of the Environment, WISC-III = Wechsler Intelligence
Scale for Children-Third Edition.
an = 120 for both groups.
* p < .05, **p < .01, ***p < .001, one-tailed using the independent samples t-test and Bonferroni correction
for familywise error rate.













CA CA CA CA C CA C


a r r tl-


E
C)
-3
0

Ci







0
E


E
C)




C3
-3
0



C)

C)


- 0>
- Ci
im


CA CAl


O tl O N c00 0

\O xO \o \o \o O xO


0> 0> 0> 0> 0> 0> 0> 0> 0>




0> 0> 0> 0> 0> 0> 0> 0> 0>




N^ N^ N^ N^ N^ 0C C C 0>
OC OC OC C C C C00 0 C00 0C
00 00 000 0000000 0


0 0 0
0 0 0


0 0 0 0 0 0
0 0 0 0 0 a


tn xoC N- oC a^ o0 all oo N
~r ~r ~ 0r 0 0r 0 0r
en n n n n e e e e


tim
0




o1
0




00
^



00
0a


0



0
0


00
(N

(N
Ni










a,
N
en







.a,

2s 0
4-


S(N en
VI VI VI


br tim xO N OC 0
CA CA CA CA CA CAl


pc


ti)
o oI
I cIz





















7 0















to
CZ r
i |

















Oi~i
an co
r o

















0










OCZ,
Z ^
B a-
^ o~
'0







kn
c3U
I I
























)r)
?z






















E1


B S Si S S f S~ fl c

aX _W PI ^ a rc^
4i J 4J 4tJ tim 0tim^




a Q
-o a 0 0 0k~a











U F

ca
Oa




o M









0 a



a s ca

ca i1




0 coa
a* ca



















.s Fi o 0 .
ca ca














Ca
u I *





5= a .5 .2















C) C 0

p^^ll









^ -^



























! A I s





00 o o


o --

-0
\-110 I















i~ 1^ |
a




























0-








o n
0 0A
_ 0








I l l I



t














CHAPTER 4
DISCUSSION

Study Summary

In the current study, a group of children (n = 120) who were prenatally exposed to

cocaine, alcohol, tobacco, and marijuana were compared to a matched group of children

(n = 120) who were exposed to alcohol, tobacco, and marijuana but not cocaine.

Statistical analyses showed that the group of children with prenatal cocaine exposure

(PCE) had significantly higher levels of exposure to alcohol, tobacco, and marijuana and

significantly higher prenatal obstetrical risk scores than their matched controls. The PCE

group also had significantly shorter mean gestational age and mean head circumference at

birth than the matched control group. The composition of the two groups did not differ

significantly by ethnicity or sex. Both groups were predominantly African American and

were almost equally split between boys and girls.

Main Findings

Three hypotheses were developed to examine the proposed relationships between

prenatal cocaine exposure, attention, and reading ability. The first hypothesis, that

measures of neonatal attention would be significantly related to measures of early

childhood attention was not supported. With one exception, neonatal attention as

measured by three scales of the Brazelton Neonatal Behavioral Assessment Scale

(BNBAS) was not significantly associated with a variety of attentional measures

administered at ages 5 and 7, including a continuous performance test, short term

auditory attention (Digit Span), or three attentional tasks involving visual scanning and









visuomotor coordination (Letter Cancellation, TMT Part A, and WISC-III Coding

subtest).

The second hypothesis, that children with PCE would perform significantly worse

than non-exposed children on measures of early childhood attention and reading ability

was also not supported. No significant group differences were found on visual attention

indices of a continuous performance test, short term auditory attention (Digit Span), or

three visuomotor attentional tasks (Letter Cancellation, TMT Part A, and Coding).

Support was found, however, for the third hypothesis that PCE would have an

indirect effect on reading ability that was mediated by head circumference and its

subsequent effects of verbal ability and visual attention. Results of structural equation

modeling on the combined sample of cocaine-exposed and nonexposed groups showed

that PCE had a small negative effect on head circumference, with higher amounts of

exposure associated with smaller head sizes at birth (3 = -.13). Head circumference, in

turn, predicted age 5 verbal ability (3 = .26), which was highly predictive of age 7 verbal

ability (3 = .93). Age 5 verbal ability was also moderately related to two different age 7

attention factors, one comprised of Digit Span and the other by Letter Cancellation, TMT

and Coding (3s = .52 and .47). Finally, the age 7 visual attention factor comprised of

Letter Cancellation, TMT, and Coding was found to be the largest predictor age 7 reading

(3 = .44) followed closely by age 7 verbal ability (3 = .41). Overall, the structural

equation model accounted for 68% of the variance in age 7 reading ability.

Ancillary Findings

Between-group post hoc analyses revealed that the two study groups did not differ

on verbal ability measures or 6 of 7 measures of the caregiving environment. The only









significant difference was found on the Growth Fostering Materials and Experiences

subscale of the Middle Childhood Home Observation for Measurement of the

Environment. Contrary to expectations, the difference favored the PCE group, probably

because a much larger proportion of these children were in placements away from their

biological mothers.

Several other findings from the structural equation modeling analysis also deserve

comment. First, the small negative effect of PCE on head circumference (3 = -.13) was

very similar to that of prenatal exposure to alcohol and marijuana (3s = -.13 and -.11,

respectively). Unexpectedly, prenatal tobacco exposure had a paradoxical effect on head

circumference with higher levels of exposure associated with larger head sizes (3 = .14).

The similarity of the coefficients for the four drug exposure variables and the paradoxical

positive coefficient for the tobacco exposure variable may be due to multicollinearity

among these variables. An alternative approach would have been to combine the alcohol,

marijuana, and tobacco exposure variables into a single "other drug exposure" variable.

A decision was made not to follow this approach so that the potential effects of PCE

could be compared directly with the effects of exposure to the other drugs measured in

the study.

More than any of the drug variables, however, sex was the single strongest

predictor of head circumference with girls having smaller heads at birth than boys (3 = -

.22). Sex also independently predicted the age 7 visual attention factor with girls

outperforming boys (3 = .29). None of the drug variables nor head circumference were

direct predictors of the age 5 or age 7 attention factors.









Study Findings in the Context of the Literature

Hypothesis #1. That the three BNBAS scores used in the current study were not

predictive of early childhood attention measures is not wholly surprising. First, the

majority of studies that have assessed the predictive validity of infant attention for later

cognitive abilities have not utilized the BNBAS. Most studies have employed

experimental habituation paradigms that are not widely used in the clinical assessment of

infants. Thus, while useful as a broad-based clinical tool for assessing the adaptation of

infants to their extrauterine environment, the BNBAS and its specific cluster scores may

not have the sensitivity or specificity needed for research applications.

Second, the BNBAS has typically been used to predict general cognitive abilities,

such as IQ or language development, rather than attention in particular. While there are a

handful of studies that have begun to trace the developmental trajectory of attention from

infancy to childhood, none have used the BNBAS. For the studies reviewed in Chapter 1,

two experimental paradigms, fixation duration and exposure time needed to meet

criterion during a visual recognition memory task, were the measures used to predict later

attentional abilities, respectively (Sigman et al., 1991, Rose & Feldman, 1995). Thus, the

attempt in the current study to predict childhood attentional performance using the

BNBAS was a venture into relatively unchartered territory.

Third, studies comparing infant and early childhood performance on attentional or

other cognitive tasks generally do not use neonates. In the review by Colombo (1993), the

age at the first assessment ranged from 3 to 9 months. The reason may be that scores

based on assessments of newborns using measures such as the BNBAS are likely to

reflect the state of the child at the time of the assessment rather than a more enduring trait

that is not expected to change significantly over time. Using the BNBAS in newborns,









then, is somewhat akin to using a state measure to predict a childhood trait (attention).

Such an undertaking is less likely to produce significant results, especially considering

that attention is not a stable attribute in childhood but continues to develop with age

(Cooley & Morris, 1990).

Finally, the hypothesis that BNBAS scores would be predictive of later attentional

measures was based, in part, on group differences on the BNBAS reported in the

literature between children with PCE and non-exposed children. A meta-analysis of

group differences on the BNBAS between children with PCE and non-exposed children

has revealed that the largest reliable differences were for motor performance and

abnormal reflexes and not for the Orientation, Habituation, or Regulation of State scores

(Held, Riggs, & Dorman, 1999). It was also found that while the Orientation and

Autonomic Regulation clusters produced small, significant effects at birth and at 3 to 4

weeks of age, the effects were small, decreased over time, and were likely due to a large

sample size (Held, Riggs, & Dorman, 1999). Thus, attentional differences found between

infants with PCE and comparison groups in some studies did not hold up across samples.

In sum, considering the lack of reliable between-group differences on the BNBAS for

cocaine-exposed and non-exposed children, combined with the three BNBAS

measurement issues just outlined, it would have been quite remarkable if the three

BNBAS scores used in the present study were significantly correlated with attention

measures at age 5 and 7.

Hypothesis #2. The lack of significant group differences on attention and reading

measures between children with PCE and non-exposed children in the current study is

consistent with the majority of the literature in this area. In a recent review, Frank,









Augustyn, Knight, Pell, and Zuckerman (2001) concluded that after controlling for

confounding factors, PCE has no consistent negative association with physical growth,

developmental test scores, or language skills. However, it should be noted that the

majority of studies included in the review examined outcomes for children age 3 or

younger. Only two other groups of researchers have published data on children with PCE

over age 3 using assessment instruments similar to those in the current study. Hurt et al.

(1997) found no group differences at age 4 between a Philadelphia inner-city sample of

children with PCE and non-exposed children on the Wechsler Preschool and Primary

Scale of Intelligence-Revised (WPPSI-R) mean Verbal, Performance, or Full Scale IQ

scores. Richardson, Conroy, and Day (1996), based in Pittsburgh, found no group

differences at age 6 between children with PCE and non-exposed children for any scores

on the Stanford-Binet Intelligence Scale-Fourth Edition (Thomdike, Hagen, & Sattler,

1986) or on the Reading, Spelling, or Arithmetic subtests of the Wide Range

Achievement Test-Revised (Jastak & Wilkinson, 1984). As in the current study, the

Pittsburgh research team found that both groups of children generally scored in the

average range on IQ and achievement measures (Richardson et al., 1996).

Three published studies of attentional abilities of school-age children with PCE

and non-exposed children were found in the extant literature. The first two, from the

same group of researchers (Richardson et al., 1996; Leech et al., 1999). In the first study,

the authors compared exposed and non-exposed groups the using a continuous

performance task (CPT) involving shapes and colors that is not widely available for

commercial use. In their sample, the 6-year-olds with PCE made significantly more errors

of omission across all three blocks of trials than did non-exposed children; however, no









group differences were found for errors of commission (Richardson et al., 1996). It was

later reported that among those excluded from the analyses were 13 children who did not

complete the test to due "impulse control and attention problems" and an unspecified

number of children refused to do the task (Leech et al., 1999). The proportion of these

children who had PCE and their characteristics are not reported. In a follow-up study, the

authors found that PCE during first trimester predicted more errors of omission; however,

marijuana use during second trimester and tobacco use during the second and third

trimesters were also predictive of more errors of omission (Leech et al., 1999). For all

drug exposure variables, the regression coefficients were small ranging from .09 to .10,

raising the question about whether these effects are meaningful in terms of the everyday

functioning (e.g., school performance) of children with PCE.

In the third study, Bandstra, Morrow, Anthony, Accornero, and Fried (2001)

conducted a longitudinal investigation of attention in school-age children with PCE using

two different CPT tasks. They used the Test of Variables of Attention (TOVA; Greenberg

et al., 1996) at age 5 and the Conners' Continuous Performance Test (CCPT; Conners,

1995) at age 7. Using omission errors as the criterion variable, they found that age 5

scores predicted age 7 scores. While the noncocaine-exposed control group performed

better on both measures than the children with PCE, between-groups statistical analyses

were not conducted to determine whether the differences were significant. A between

groups analysis, combining the two CPT measures with an age 3 "time on task" measure

of attention, revealed that the estimated group difference was significant. Hierarchical

models that included birth outcome measures such as head circumference did not

significantly attenuate the between-groups difference. Using SEM, the authors also found









that the amount of PCE significantly predicted omission error scores at age 7 even after

controlling for other prenatal drug exposure, child's age, and child's sex (3 = .26)

(Bandstra et al., 2001).

The current study found no differences between children with PCE and non-

exposed children for two indices of visual attention on the Intermediate Visual and

Auditory Continuous Performance Test (IVA CPT). Between-group differences on errors

of omission and errors of commission were not examined but such data are available for

the sample used in the current study and could be the focus of a future study.

Hypothesis #3 -- Relationship between PCE and head circumference. The

finding that the relationship between PCE and cognitive variables is mediated, in part, by

head circumference is consistent with the published literature. Currently, only four other

studies have examined the relationships between PCE, head circumference, and

developmental outcomes. Two of these studies focused on 24-month outcomes, and two

studies examined 36-month outcomes. In the first study, Chasnoff et al. (1992) found

significant correlations between head circumference at various ages up to 24 months and

Mental and Psychomotor indices of the Bayley Scales of Infant Development (Bayley,

1969). While prenatal exposure to cocaine, alcohol, tobacco, and marijuana all

contributed to head circumference measurements during the first 2 years, only cocaine

was a significant predictor as a single variable in their Chicago-based sample (Chasnoff

et al., 1992).

Another group of researchers found that head circumference at birth is correlated

with BSID scores at 6.5, 12, and 24 months of age in a mixed sample of cocaine-exposed

and non-exposed children (N= 415) (Singer et al., 2002). Significant correlation









coefficients between head circumference and BSID Index scores ranged from .12 to .22

(Singer et al., 2002). Using stepwise regression to develop a model predicting scores at

24 months, it was found that the negative effects of cocaine on cognitive outcome were

mediated through smaller head circumference at birth (Singer et al., 2002). No other

measures-gestational age, birth weight, birth length, Apgar scores, or the Hobel

Neonatal Risk score-mediated the effects of cocaine on BSID scores obtained at 24

months (Singer et al., 2002).

In the third study, Azuma and Chasnoff (1993) used path analysis to determine

whether head circumference was a significant predictor of age 3 developmental outcome

assessed using the Stanford-Binet Intelligence Scale-Fourth Edition (SBIS; Thorndike,

Hagen, & Sattler, 1986). While head circumference at age 3 was not found to be a direct

predictor of the SBIS composite IQ score at age 3, it did have an indirect effect on

composite IQ that was mediated by poor perseverance as measured by a combination of

five-point behavior rating scales completed by the blinded examiners who administered

the SBIS (Azuma & Chasnoff, 1993). The regression coefficients were -.30 between head

circumference and perseverance and -.60 between perseverance and composite IQ

(Azuma & Chasnoff, 1993).

The fourth study to examine head circumference as a predictor variable was

conducted by principal investigators of the grant from which the sample for current study

was drawn. Using structural equation modeling, it was found that age 3 developmental

outcome measured by a factor comprised of BSID scores and four subtests of the

Vineland Adaptive Behavior Scales (VABS; Sparrow, Balla, & Cicchetti, 1984) was

predicted by birth head circumference (3 = .14), which itself was predicted by prenatal









cocaine exposure (3 = -.18) after controlling for prenatal alcohol and tobacco exposure

(Eyler, Behnke, Garvan, Wobie, & Hou, 2002). While interpretation is made somewhat

more difficult by the study's use of an outcome variable that combined scores from both a

child behavioral measure (BSID) and a caregiver report measure (VABS), it is clear that

head circumference was a mediator between PCE and overall developmental outcome.

The results of the current study and the four studies just reviewed suggest that

head circumference may serve as a proxy for the effects of PCE on in utero brain

development. While three different groups of researchers have found significant

relationships between head circumference and BSID scores, these studies do not elucidate

the mechanism underlying the relationship. In the Azuma and Chasnoff (1993) study,

however, perseverance was the factor linking head circumference to Stanford-Binet

composite IQ. In the current study, head circumference predicted verbal ability, verbal

ability predicted visual attention, and visual attention was found to be the most significant

predictor of reading performance. If sustained effort (perseverance) is considered

somewhat analogous to sustained attention, then the results of the Azuma and Chasnoff

(1993) study and the current study arguably converge to suggest that PCE's effect on

head circumference may be related to problems with self-regulation that can affect

cognitive performance. These findings warrant further studies using head circumference,

particularly at birth, as a predictor variable in future studies of developmental outcomes

of children with PCE.

Hypothesis #3 Relationship Between Attention and Reading. While there are

no current published studies on the relationship between attention and reading in children

with PCE, the finding that attention is strongly related to reading ability is consistent with









other longitudinal studies using samples of elementary school children. Using path

analysis, Rabiner, Coie, and The Conduct Problems Prevention Research Group (2000)

found that teacher ratings of inattentiveness in second grade had a regression coefficient

of -.10 with fifth-grade reading achievement in a combined sample of both "at-risk" and

"normal" children. Using structural equation modeling and a much larger sample of

children drawn from the normal school population, Rowe and Rowe (1992) found that

teacher ratings of inattentiveness had a strong negative influence across four age groups

with regression coefficients ranging from -.21 to -.39. The much larger regression

coefficients in the current study may be the result of using direct behavioral measures of

attention rather than teacher ratings.

Overall, the results of the current study suggest that the effects of PCE, if any, on

school-age outcomes such as attention and reading are subtle and difficult to detect using

traditional pencil and paper assessment methods. The effects of PCE that can be

identified are of similar magnitude as those associated with prenatal exposure to alcohol

and tobacco, which are used by pregnant women more often and in larger quantities.

Distinguishing the effects of prenatal drug exposure on later childhood development is

complicated by the fact that children born to mothers who abused substances during

pregnancy are subject to other multiple risk factors for poor outcome, including poverty.

Studies on children with PCE, including the current one, seem to point to the resiliency of

humans to develop normally despite multiple risk factors rather than the strength of the

teratogenic model to predict negative outcomes.

Study Strengths

The strengths of the current study result from the nature of the sample, control for

potentially confounding variables, and the use of structural equation modeling. The









sample is distinguished from that of other prospective, longitudinal studies in that

examiners were blinded to the children's cocaine exposure status and the attrition rate was

very low (approximately 10%). In the Frank et al. (2001) review, the authors excluded 20

studies because they failed to guard against examiner bias by masking the cocaine

exposure status of the children. The distortions that can result from examiner bias,

particularly in behavioral research, are well-known (Kazdin, 1980). Moreover,

researchers have documented specific negative examiner bias with children labeled as

prenatally exposed to cocaine (Thurman, Brobeil, & Ducette; 1994; Woods, Eyler,

Conlon, Behnke, & Wobie, 1998). Obtaining data from blinded examiners has not been a

universal feature of studies of children with PCE and makes the results of the current

study more trustworthy.

Careful analysis of the characteristics of the sample and how it compares to the

originally enrolled sample is another important feature of the current study. For the

studies included in the Frank et al. (2001) review, retention rates ranged from 39% to

94%, characteristics of those lost to follow-up were not reported for 6 of the 17

independent cohorts, and four studies failed to report attrition at all (Frank et al., 2001). A

wide variety of factors can affect which participants remain in longitudinal studies,

including the severity of the problems being studied. Thus, in order to ascertain the

potential longer-term effects of any variable, it is critical to understand the nature of the

sample and how it changes over time. The sample in the current study represented 78% of

the original sample, including the children who died, and replicated the original group

differences in prenatal drug exposure, Hobel prenatal risk, and birth head circumference.

The current study groups were also similar to the original study groups in that they did









not differ in terms of ethnicity or sex. The one statistically significant difference between

the current study groups that was not found in the original study groups-gestational

age-was less than one week and did not represent a significant clinical difference.

A third strength of the current study is control for, and comparison of,

confounding variables. Failure to control for confounding variables, particularly exposure

to other drugs of abuse, is one of the most prevalent methodological flaws in the literature

on children with PCE. The original study groups were matched for demographic

variables including ethnicity, and the demographic similarities between the groups were

maintained in the current study sample, thereby controlling for these variables. Using

separate factors for prenatal exposure to cocaine, alcohol, tobacco, and marijuana in the

structural equation model allowed direct comparison of the relative effect of each drug.

These direct comparisons are important because they help to contextualize the effects of

PCE in terms of alcohol and tobacco-drugs that are more familiar to health

professionals and the public and are more commonly used by pregnant women.

Considering that women who use cocaine during their pregnancies are, in some states,

currently subject to punitive consequences (while women who use alcohol and tobacco

are not), the ability to make comparisons between various drugs of abuse could have

important policy implications.

The use of structural equation modeling is not yet common in research on

children with PCE and represents a significant contribution to the literature. Structural

equation modeling is a powerful statistical technique that enables researchers to pose and

test hypotheses that cannot easily be evaluated using more traditional statistical methods.

In the current study, no statistically significant differences were found between children









with PCE and non-exposed children in a cross-sectional analysis using traditional

statistical techniques. However, the relationship between attention and reading and

between PCE and attention could still be explored using structural equation modeling.

Study Limitations

Measurement issues. One of the primary limitations of the current study relates

to the methods used to measure attention. Attention is a complex, multi-dimensional

process that involves a number of distinct areas of the brain, depending on the nature of

the task. In the current study attention was measured using one auditory task and three

pencil and paper tasks: Digit Span, Letter Cancellation, TMT Part A, and Coding. Since

the Digit Span factor did not turn out to be a significant predictor of reading, this

discussion will focus on the latter three tasks.

Letter Cancellation, TMT Part A, and Coding were combined into a factor labeled

"visual attention;" however, each of these tasks require much more than "attention" to be

completed skillfully and efficiently. All three tasks involve visual scanning, have a

significant fine motor component, and require familiarity with letters or numbers or both.

Moreover, all three tasks are timed, and scores are based primarily on speed with a

secondary emphasis on accuracy. Arguably, then, the factor comprised of these tasks

could have been called a "processing speed" measure rather than an "attention" measure.

Additionally, it should be noted that there was some variability in how the children

performed on each of the tests. While both groups of children scored in the average range

on Coding, mean scores for TMT Part A for both groups of children were more than 1.5

standard deviations below published norms. This highlights the fact that these tests may

tap different aspects of the multidimensional construct called attention.









Generalizability issues. The participants in the current study were drawn from

rural north central Florida. By contrast, the majority of the prospective longitudinal

studies on the effects of PCE funded by the National Institute on Drug Abuse are based in

large urban centers, namely Atlanta, Baltimore, Chicago, Cleveland, Detroit, Miami, and

Pittsburgh. This significant geographical difference could affect a variety of factors that

impact child developmental outcome, ranging from the amount and quality of drugs used

by the mothers to the amount and quality of social supports available to mothers and

intervention programs available to the children. Although the overwhelming majority of

the participant families in the NIDA-funded studies are poor, poverty is likely to be

experienced quite differently in a rural area than in an urban area. The potential impact

that a rural versus an urban setting may have on developmental outcomes of children with

PCE remains an empirical question that warrants further research.

Future Directions

Several of the suggestions for helping to elucidate the relationships found

between PCE, attention, and reading in the current study relate to measurement issues.

First, no significant relationship was found between the Digit Span factor and the

Reading factor, and this may be due to the fact that scaled scores that combine

performance on Digits Forward and Digits Backward were used. Digits Forward is

traditionally considered an attentional task, while Digits Backward, which requires

manipulating the information presented, is considered a working memory task and more

difficult than Digits Forward. Two children could obtain identical scaled scores on the

combined Digit Span task, with considerably different performances on Digits Forward

vs. Digits Backward. Thus, using raw scores for Digits Forward and Digits Backward









may have revealed significant between-group differences or resulted in significant

predictive relationships between Digits Forward or Digits Backward and reading ability.

Second, as discussed earlier, the Visual Attention factor was comprised of three

tasks that involved both visual scanning and visuomotor processing speed. To assess the

relative importance of these two sets of component skills to reading ability, a task that

relies primarily on motor speed, such as the Finger Tapping Test from the Halstead-

Reitan Neuropsychological Test Battery (Reitan & Wolfson, 1985) or mean reaction time

on the IVA CPT (Sandford, 1995), could have been used as a covariate in the analyses.

Although no consistent negative effects of PCE have been found for motor development

(Frank et al., 2001; Singer et al, 2002), analysis of scores on motor and visual-motor

tasks could help to clarify the variability in performance across the different measures of

visual attention and how they may affect reading skills.

Third, future studies should examine single-word reading skills separately from

reading comprehension. In the current study, the Reading factor was comprised of two

subtests, Basic Reading and Reading Comprehension, that assess different aspects of

reading ability. The Basic Reading subtest is largely a measure of single-word reading

and relies on sight word recognition for familiar words and phonological decoding skills

for unfamiliar words. The other subtest, Reading Comprehension, involves reading short

passages and answering questions posed by the examiner. It may be that the visual

scanning aspects of the Visual Attention factor, comprised of Coding, Letter

Cancellation, and TMT Part A, may be more related to single word reading while the

working memory aspects of the Digit Span factor and the general information knowledge

and vocabulary skills captured in the Verbal Ability factor would be more associated with









reading comprehension, which requires higher order synthesis of material as it is read. In

short, more detailed analysis of which aspects of attention map onto which aspects of

reading could be the focus of future studies.

Fourth, it is likely that there are subgroups among the children with PCE who

differ in ways that may affect outcome. For example, there was some variability in the

amount and timing of cocaine exposure among the children with PCE. By combining

children with low- and high-exposure or children with predominantly first-trimester

versus third-trimester exposure in the analyses, significant subgroups differences may

have been diluted or even eliminated. Another example of potentially important subgroup

differences has to do with exposure to multiple teratogens. At least one animal study has

reported that maternal use of alcohol and cocaine during pregnancy has more detrimental

effects on offspring outcome than use of either drug alone (Randall, Cook, Thomas, &

White, 1999). Since human maternal cocaine use during pregnancy most often occurs in

conjunction with other drugs of abuse, studies of the coteratology of cocaine with

alcohol, tobacco, and marijuana should become the research wave of the future.
















REFERENCES


Adams, M. A. (1994). Beginning to read: Thinking and learning about print. Cambridge,
MA: MIT Press.

Akbari, H. M., Kramer, H. K., Whitaker-Azmitia, A. P., Spear, L. P., & Azmitia, E. C.
(1992). Prenatal cocaine exposure disrupts the development of the serotonergic
system. Brain Research, 572, 57-63.

American Psychiatric Association. (1994). Diagnostic and Statistical Manual of Mental
Disorders (4th ed.). Washington, DC: Author.

Azuma, S. D., & Chasnoff, I. J. (1993). Outcome of children prenatally exposed to cocaine
and other drugs: A path analysis of three-year data. Pediatrics, 92(3), 396-402.

Bandstra, E. S., Morrow, C. E., Anthony, J. C., Accornero, V. H., & Fried, P. A. (2001).
Longitudinal investigation of task persistence and sustained attention in children with
prenatal cocaine exposure. Neurotoxicology and Teratology, 23, 545-559.

Bayley, N. (1969). Bayley Scales of Infant Development. San Antonio, TX: Psychological
Corp.

Beckwith, L., Crawford, S., Moore, J. A., & Howard, J. (1995). Attentional and social
functioning of preschool-age children exposed to PCP and cocaine in utero. In M.
Lewis & M. Bendersky (Eds.), Mothers, babies, and cocaine: The role of toxins in
development (pp. 287-303). Hillsdale, NJ: Erlbaum.

Behnke, M., Eyler, F. D., Conlon, M., Wobie, K., Woods, N. S., & Cumming, W. (1998).
Incidence and description of structural brain abnormalities in newborns exposed to
cocaine. Journal ofPediatrics, 132, 291-294.

Behnke, M., Eyler, F. D., Garvan, C. W., Wobie, K., & Hou, W. (2002). Cocaine exposure
and developmental outcome from birth to 6 months. Neurotoxicology and Teratology,
24, 283-295.

Behnke, M., Eyler, F. D., Woods, N. S., Wobie, K., & Conlon, M. (1997). Rural pregnant
cocaine users: An in-depth sociodemographic comparison. Journal of Drug Issues,
27, 501-524.

Bentler, P. M. (1990). Comparative fit indices in structural models. Psychological Bulletin,
107, 238-246.