The concomitants of sleep-disordered breathing in elderly males


Material Information

The concomitants of sleep-disordered breathing in elderly males
Physical Description:
v, 172 leaves : ill. ; 29 cm.
Aber, William Robert, 1959-
Publication Date:


Subjects / Keywords:
Sleep Apnea Syndromes   ( mesh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1988.
Includes bibliographical references (leaves 164-171).
Statement of Responsibility:
by William Robert Aber.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
oclc - 20380658
System ID:

Full Text








ABSTRACT ........................................ iv


ONE INTRODUCTION .......................... ...... 1

Definitions/Terminology...................... 3
Characteristics of SAS Patients...............5
Incidence of Sleep Apnea Activity............. 6
Deficits Associated with SAS and Sleep
Apnea Activity........................... 18
Cardiological Deficits................18
Hypersomnolence ....................... 26
Disruption of Nocturnal Sleep.......... 34
Cognitive Deficits..................... 39
Current Thinking on the Etiology of Apnea-
related Cognitive Deficits................53
Some Methodological and Measurement Issues...61
Statement of the Problem .................... 68

TWO METHOD............................... .......... 70

Subjects............ ........................ 70
Apparatus ........................ ....... ..... 71
Measures.............. .... ........ .... .. 72
Independent Variables..................... 72
Dependent Variables ...................... 82
Procedure.................. ........ .......... 86
Statistical Procedures........................87

THREE RESULTS.................... ................. 88

Incidence of Apnea and Desaturation
Activity........ ........ ........ ...... .... 88
Nocturnal Respiratory and Health Related
Variables.............. .... ....... ..... 94
Nocturnal Respiratory and Sleep/Wake
Variables.................................. 98
EEG Data.................... ...... .. 98
Daytime Sleepiness and Napping Data...103
Subjective Nocturnal Sleep Variables..106
Nocturnal Respiratory and Neuropsychological
Variables......... .... .... .......... ... 108

FOUR DISCUSSION............................... 117

Demographics and Incidence of Nocturnal
Respiratory Disorder..................... 117
Nocturnal Respiratory and Health-Related
Nocturnal Respiratory and Sleep/Wake
Variables................................ 122
EEG Data ............................. 122
Daytime Sleepiness Variables......... 124
Nocturnal Respiratory and Neuropsychological
Variables ............................... 128
The Night-to-Night Consistency of Nocturnal
Respiratory Activity...................... 132
Conclusions................................ 135

APPENDIX RAW DATA TABLES............................ 137

REFERENCES .........................................164

BIOGRAPHICAL SKETCH..................................172


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



William Robert Aber

August, 1988
Chairman: Hugh Davis
Major Department: Clinical and Health Psychology

Sleep Apnea Syndrome (SAS) is a nocturnal respiratory

disorder with demonstrated health, sleep/wake, and cognitive

consequences. Patients with SAS experience apneas (pauses

in breathing of 10 seconds or more), hypopneas (declines in

the amplitude of breathing), and oxygen desaturations while

asleep. Recently, it has become apparent that

sleep-disordered breathing occurs in subjects who have no

prominent health, sleep/wake, or cognitive complaints. In

this study, a nonclinical sample selected to display a wide

range of sleep-disordered breathing (males in good health

above the age of 60) received detailed measurement of their

nocturnal respiration, nocturnal oxygen saturation, sleep,

health status, and cognitive/neuropsychological skills in an

attempt to assess the effects of sleep-disordered breathing

in a nonclinical population.

Thirty-nine elderly males with a mean age of 66 years

comprised the sample. During an experimental evening they

received testing and completed questionnaires before

sleeping overnight with their sleep and breathing

continuously recorded. Fourteen subjects were recorded on

two consecutive nights. Seventy-seven percent of the

subjects experienced at least one episode of apnea or

hypopnea and 20% experienced more than 5 events per hour.

Increasing waking blood pressure was linked to increasing

numbers of nocturnal desaturations. Many significant

relationships between EEG sleep parameters and respiratory

indices were found, indicating that arousals subsequent to

desaturations lead to fragmented nocturnal sleep.

Increasing numbers of desaturations and apneas significantly

predicted increased daytime napping behavior. Lower scores

on tests measuring IQ, immediate nonverbal memory,

visuo-spatial reasoning/organization, and cognitive

flexibility were found to be predicted by increasing

desaturation activity. Nocturnal respiratory activity was

found to remain stable from the first to the second night in

the laboratory.

It is concluded that elderly males who experience

multiple nocturnal desaturations are at increased risk of

sleep fragmentation, daytime sleepiness, increased waking

blood pressure, and negative cognitive/neuropsychological

changes. Elderly males with nocturnal desaturation may fall

on a continuum with Sleep Apnea Syndrome patients.


Sleep apneas are respiratory pauses during sleep.

Sleep Apnea Syndrome (SAS) is a clinical entity diagnosed

primarily on the basis of a laboratory finding of multiple

apneas (often numbering in the hundreds across the course of

a night of sleep) and secondarily on the basis of the

presence of one or more of a cluster of daytime deficits,

including excessive daytime sleepiness, cardiopulmonary

complications, alterations of consciousness, and

intellectual deficits. Although the repetitive nocturnal

respiratory disturbances associated with SAS have long been

thought to be the causes of these waking deficits

(Guilleminault et al., 1976), only during the last several

years have systematic attempts been made to describe and

quantify the nature of the relationship between apnea

activity and daytime sequelae. As will be seen below, sound

evidence establishing the etiologic importance of sleep-

disordered breathing in the development of waking sequelae

in SAS patients has begun to accrue, with the more

sophisticated studies in this area implicating hypoxemia as

the relevant proximal cause.

Paralleling this increase in the number of studies

investigating the concomitants of apnea activity in SAS

patients has been an increased interest in establishing the

incidence of nocturnal respiratory disturbance in

nonclinical populations. These studies have revealed the

occurrence of sleep-disordered breathing in often very high

proportions of segments of the general population. Aging

persons in particular have been found to exhibit striking

incidence rates for sleep apnea activity. Such findings

raise the questions of whether the waking deficits found to

be related to apneas and oxygen desaturations in SAS

patients are also found in nonclinical populations

exhibiting sleep-disordered breathing and to what degree. A

few very recent studies have addressed these questions with

respect to aging persons; however, each of these studies has

focused on only one or a subset of the likely concomitants

of apnea activity, and their findings are in need of both

replication and extension to the full range of likely

daytime sequelae of nocturnal respiratory disturbance.

The present study will examine the correlates of

nocturnal respiratory disturbance in a nonclinical

population expected to display a wide range of apneic

events: elderly males. The relationship between respiratory

indices and detailed measures of demographic, health,

sleep/wake, and cognitive variables in this population will

be examined. Further, a replication of previous findings

indicating a high incidence of apnea activity and a high

prevalence of 'clinically significant' levels of apnea

activity in this population will be attempted. Finally,

certain largely previously neglected methodological and

measurement issues relating to the stability of measures of

respiratory disturbance over time and to the relative

utility of the myriad possible measures of respiratory

disturbance in predicting waking deficits will be examined.

A review of the information currently available on various

aspects of Sleep Apnea Syndrome and the phenomenon of

nocturnal respiratory disturbance in general will be

presented as background for the study. The diagnosis and

characteristics of SAS, the incidence of SAS and sleep apnea

activity in young and old populations, and the deficits

associated with SAS and sleep apnea activity will be

reviewed. As will be seen, the daytime concomitants of

varying levels of sleep apnea activity in healthy younger

subjects are not well understood, and very little research

has examined these concomitants in elderly subjects. This

will be followed by an examination of current thinking

concerning the specific etiology of the possible

neuropsychological concomitants of sleep-disordered

breathing. Finally, some methodological issues will be

raised and literature pertinent to these issues will be

reviewed with an eye toward identifying questions that

remain as yet unanswered.

Sleep apneas have been defined as cessations of airflow

at the mouth and nose which last for 10 seconds or longer

(Guilleminault et al., 1976). Sleep hypopneas have

generally been defined as reductions in airflow of 50% or

more (without complete cessation) at the mouth and nose

(Carskadon et al., 1980). Oxygen desaturation is a frequent

but not invariant concomitant of apneas and hypopneas. When

a respiratory event begins, arterial oxygen saturation

usually begins to fall and may continue to fall until normal

breathing resumes. Sleep Apnea Syndrome (SAS) is a clinical

entity described in a recent nosology published by the

Association of Sleep Disorders Centers (ASDC, 1979) as a

potentially lethal condition characterized by multiple

apneas, repetitive episodes of loud snoring, excessive

daytime sleepiness, and possible cardiopulmonary

complications, alterations of consciousness, and

intellectual deficits. The episodes of apnea are thought to

play a causal role in the production of the other deficits

associated with SAS (Guilleminault et al., 1976); because

those episodes lend themselves to objective quantification

they have taken on increasing importance as diagnostic

markers of the syndrome. Guilleminault et al. (1976)

suggested that the occurrence of at least 30 apneic episodes

during 7 hours of sleep was diagnostic of SAS, and they

validated this criterion through a comparison of symptomatic

(presumably snoring and excessively sleepy) and asymptomatic

younger subjects. These workers later proposed the use of

an Apnea Index score (AI; number of apneas divided by number

of hours of sleep) in excess of 5 as a cut-off for the

diagnosis of SAS (Guilleminault et al., 1978). Carskadon et

al. (1980) combined apneas and hypopneas to form an apnea +

hypopnea index (AHI) and suggested that a score of 5 or

greater be considered diagnostic of SAS. Subjects

exhibiting lower levels of sleep-disordered breathing (that

is, AI or AHI scores greater than 0 but less than 5) have

been labeled as experiencing sleep apnea activity (SAA).

These criteria have been widely adopted in both clinical and

research settings.

Characteristics of SAS Patients

Three relatively large samples of SAS patients have

been described in the literature. Guilleminault et al.

(1978) found that of 50 patients diagnosed with SAS during a

6 year period, 48 were male. Thirty-nine of the 50 were

referred with complaints of excessive daytime sleepiness, 7

were referred by their spouses who described loud snoring

and abnormal movements during sleep, and the remainder were

referred with complaints of morning nausea, morning

headaches/confusion, or hypertension. Sixty percent of

these patients were more than 15% overweight. Other

frequently reported complaints included personality changes

(broadly and anecdotally defined), impotence, and

intellectual deterioration (usually attention/concentration

difficulties; no formal testing was conducted). These

patients ranged in age from 28 to 62. Kales et al. (1985a)

reported on 50 consecutive patients (aged between 25 and 68)

diagnosed with SAS at a sleep disorders center. Eighty-six

percent of these patients were male, all complained of

debilitating daytime sleepiness, and 86% exhibited

cardiovascular abnormalities (hypertension or arrhythmias).

Forty-six of the 50 patients were snorers and all but 9 were

clinically obese. Other frequently reported complaints

included increased body movements during sleep, profuse

sweating during sleep, bedwetting, intense morning

headaches, morning fatigue, impaired morning alertness, and

decreased sexual desire. Weitzman et al. (1980) described a

group of 10 SAS patients who were male, consistently

overweight, hypertensive, excessively sleepy during the day,

and between the ages of 38 and 47. From these data it would

appear that the 'typical' SAS patient is an overweight

middle-aged male who often suffers from deleterious health

and intellectual changes.

Incidence of Sleep Apnea Activity

In studying the incidence of sleep apnea activity and

the prevalence of clinically significant levels of sleep-

disordered breathing in younger and older subjects,

researchers have typically recruited subjects either from

clinical sources (usually sleep disorder clinic patients

with very prominent complaints about their sleep) or from

the community. Some studies using community volunteers

attempted to screen out subjects with sleep complaints,

while others made no such attempt. The 13 studies which

drew subjects from community sources will be discussed


Webb (1974) studied a sample of 20 "non-ill" males

without sleep complaints, 11 of whom were between the ages

of 19 and 45 and 9 of whom were aged 46 to 63. Eighty-two

percent of the subjects aged 46 to 63 exhibited at least one

episode of sleep apnea (range: 1-5) while none of the

younger subjects experienced sleep apnea. The author

concluded that the presence of apnea activity is positively

related to age. Guilleminault et al. (1978) recruited a

sample of 20 noncomplaining subjects between the ages of 40

and 60 and found that the males experienced an average of 7

apneas per night (range: 1-25) and that the females

experienced an average of 2 apneas per night (range: 0-5).

Block et al. (1979) recruited a sample of 49 subjects

without sleep or breathing complaints (#males: 30; age

range: 24-62; #females: 19; age range: 20-51). Forty

percent of the males and 15% of the females experienced at

least one episode of sleep apnea. The males with apnea

exhibited a mean of 4.2 episodes while the females with

apnea exhibited a mean of 3. Block et al. (1979) also

reported data pertaining to the incidence of nocturnal

oxygen desaturation (a frequent concomitant of sleep-

disordered breathing) in these noncomplaining subjects.

Fifty-seven percent of the males experienced at least one

desaturation of 4% or greater; none of the females exhibited

desaturation. Both age and weight:height ratio were found

to be significantly positively related to number of

desaturations but only weight:height ratio (and not age) was

found to be significantly positively related to number of

apnea episodes. Block et al. (1980) compared a group of 20

post-menopausal women (age range: 50-74) with a group of 18

pre-menopausal women (age range: 20-51). All subjects were

without sleep and breathing complaints. There was no

significant difference in the number of women in the two age

groups exhibiting sleep apneas (although 40% of the older

and only 11% of the younger subjects exhibited apneas);

however, a significantly higher number of the older women

exhibited desaturations >4% and hypopneas. The older women

with apneas averaged 39 episodes per night while the younger

subjects with apnea averaged 6.

Bixler et al. (1982) recruited a sample of 100 subjects

who were carefully screened for the absence of sleep

complaints, physical problems, and medication usage. The 41

males and 59 females were grouped into 3 age blocks: 18-29,

30-49, and 50-74. None of these non-complaining subjects

achieved an Al score >5; however, 14.6% of the men and 10.2%

of the women experienced between 3 and 29 apnea episodes per

night (males with apnea averaged 10.5 episodes; females

averaged 8.5 episodes). The incidence of sleep apnea

activity increased across the three age groups (3.3% vs.

12.8% vs. 20%) and there was a significant positive

correlation between age and presence of sleep apnea

activity. The same group of researchers later reported

incidence data from an expanded number of older

non-complaining subjects (Bixler et al., 1985). Sixty

subjects aged 50 or older (many of these were subjects

studied in Bixler et al., 1982) were compared with 69

subjects between the ages of 18 and 49. Eight percent of

the older males and none of the older females and younger

subjects achieved an AI score >5. Thirty-two percent of the

older males and 23% of the older females exhibited between 3

and 29 apnea episodes (the figures for younger males and

females were 10% and 7%, respectively). A significantly

higher number of older subjects (27% vs. 9%) exhibited

between 3 and 29 apnea episodes per night and age was

significantly positively correlated with both the presence

of apnea activity and AI. Age was not significantly related

to apnea activity within the exclusively older group. In

both age groups there was a (non-significant) trend toward a

higher AI score and incidence of apnea activity in males.

These researchers also reported oxygen saturation data

indicating that the minimum saturation reached during sleep

did not differ between males and females and between older

and younger subjects. They did not report correlations

between age and desaturation data.

Reynolds et al. (1985), in a study designed to compare

the incidence of sleep apnea activity in older subjects

falling into one of three groups (demented, depressed,

healthy control), reported that of a group of 23

non-demented, non-depressed subjects without sleep

complaints between the ages of 58 and 82, 4.3% attained an

AI score >5 (11% of the males and 0% of the females). The

average AI score in this sample was 1.3 and the average

Apnea + Hypopnea Index score was 1.8. These workers did not

report data pertaining to the incidence of sub-clinical

levels of apnea activity. In a study similar to Reynolds et

al. (1985), Hoch et al. (1986) recruited 56 healthy subjects

between the ages of 58 and 82 who reported no sleep/wake

complaints and who demonstrated on psychological testing

that they were non-demented and non-depressed. Of these

healthy controls, 5.4% attained an AI score >5. Separate

percentages for males and females were not reported. These

workers also did not report incidence data for subclinical

levels of apnea activity.

Carskadon and Dement (1981) recruited 18 male and 22

female elderly subjects from nonclinical sources. Subjects

who "complained spontaneously" of sleep problems were

excluded, but this criterion was not rigidly applied. The

males ranged in age from 63 to 86 and the females ranged

from 62 to 85 years of age. Eighty-eight percent of the

males and 72% of the females experienced at least one

episode of apnea or hypopnea; 39% of the males and 18% of

the females achieved AI scores >5 (44% and 32% achieved

Apnea + Hypopnea Index scores >5). Males experiencing apnea

exhibited a mean of 51 episodes and females with apnea

averaged 34 episodes. A trend toward an increase with age

(within this elderly group) in number of sleep apnea events

was apparent for the females only. Lavie (1983) studied 78

male industrial workers (age range is not reported, but it

may be assumed from their active employment status that most

subjects were middle-aged or younger). Findings indicated

that 20.5% of these subjects achieved an AI score >5 and 14%

achieved an AI score >10. The average age of the workers

with AI scores >10 was significantly higher than the average

age of the workers without apnea. Ancoli-Israel et al.

(1986) report apnea incidence data derived from 425 randomly

selected subjects aged 65 to 101. The at-home recording

device used in this study provided only very rough measures

of respiratory activity and sleep. Of the 195 males (mean

age: 73.5) and 230 females (mean age: 72.8), 24% achieved AI

scores >5. The mean Apnea Index for the entire sample was

3.97 and for those with AI scores >5 the mean was 13.2. The

males demonstrated a significantly higher prevalence of AI

15 (28% vs. 19.5%) and a significantly higher mean AI (7.5

vs. 3.4); there was no significant difference in the mean

age of those above and below AI=5 within this exclusively

older sample.

McGinty et al. (1982) studied 26 males between the ages

of 55 and 75, selected without reference to sleep or

respiratory complaints. These researchers concentrated

exclusively on oxygen desaturation data; thus, they do not

report on the incidence of sleep apnea activity or on the

prevalence of clinically significant levels of apnea

activity. Sixty-two percent of these elderly males

experienced more than 8 episodes of desaturation >4% per

hour and 46% experienced more than 20 such episodes per

hour. Those subjects exhibiting more than 8 desaturations

per hour experienced an average of 28 such episodes per

hour; the remaining subjects exhibited an average of 4

desaturations per hour. These two groups of older males did

not differ significantly in age. Finally, Smallwood et al.

(1983), in a study designed to compare the prevalence of a

clinically significant level of apnea + hypopnea activity in

elderly controls (age range: 50-80) and elderly Alzheimer's

patients, found that of 24 male and 10 female controls, 46%

and 0%, respectively, attained A+H Index scores greater than

or equal to 5. Apnea + Hypopnea Index was significantly

greater in the elderly male controls than in a group of

young male controls. The elderly males exhibited a mean A+H

Index score of 7.2; the elderly females averaged 0.71.

The above studies report widely divergent incidence

rates for sleep apnea activity and prevalence rates for

clinically significant levels of apnea activity. For

example, within samples of elderly males, separate studies

reported that 32%, 82%, and 88% of the subjects experienced

at least one episode of apnea. Similarly, studies reported

that 8%, 11%, 39%, and 28% of elderly males attained Apnea

Index scores >5. These discrepancies seem attributable to

the differing subject selection strategies used by these

studies. In the case of both sets of percentages the two

latter (higher) figures were derived from samples not

screened for sleep complaints. The other (much lower)

figures were derived from samples composed of subjects who,

after careful screening, were found to have no complaints

about their sleep. It seems apparent that excluding

subjects who are dissatisfied with their sleep greatly

reduces the observed incidence of sleep apnea activity.

Each study which included both male and female older

subjects found the males to exhibit higher incidence rates

for apnea activity (32% vs. 23%; 88% vs. 72%) and a higher

prevalence of AI>5 (8% vs. 0%; 11% vs. 0%; 39% vs 18%; 28%

vs. 19.5%). Two studies reporting on the prevalence of

Apnea + Hypopnea Index scores >5 in older samples found

similar sex differences (46% vs. 0%; 44% vs. 32%). Those

studies reporting incidence rates (ranging between 0% and

11%) and prevalence rates (0%) in younger subjects suggest

that amount of apnea activity is strongly positively related

to age. This conclusion is supported by the findings of the

three studies (Webb, 1974, Bixler et al., 1982, and Bixler

et al., 1985) which found a substantial positive

relationship between age and amount of apnea activity and

the two studies (Block et al., 1979 and Block et al., 1980)

which found a significant relationship between age and both

incidence and amount of nocturnal oxygen desaturation.

These latter studies failed to demonstrate a significant

relationship between age and apnea activity, although

age-related trends were apparent. The findings of McGinty

et al. (1982) indicate that the very high numbers of

respiratory events observed in older males are accompanied

by equally high numbers of desaturations >4%.

The second sampling strategy (i.e., the recruitment of

subjects with prominent sleep complaints) has been utilized

by four studies reporting apnea incidence data. Kales et

al. (1982) wished to examine the possible role of sleep

apnea in the etiology of insomnia. They studied 200

consecutive sleep disorders clinic patients with a primary

complaint of chronic difficulties initiating or maintaining

sleep. These 82 males and 118 females ranged in age from 18

to 78 with a mean age of 42.3. None of these insomniac

subjects exhibited what the authors considered to be a

clinically significant number of apneas per night (30), but

10.5% experienced between 3 and 29 episodes (males: 13.4%;

females: 8.5%). Subjects with sleep apnea activity averaged

9.1 episodes and were significantly older than those

without. Ancoli-Israel et al. (1981) recruited 24 elderly

subjects whose answers to a questionnaire raised suspicion

of the presence of sleep apnea or nocturnal myoclonus. The

11 males and 13 females ranged in age from 58 to 79.

Fifty-four percent of the males and 23% of the females

achieved Al scores greater than 5. Twenty-two of the 24

subjects (91%) exhibited at least one episode of sleep

apnea; within this subsample, the males averaged 53 apneas

and the females averaged 32. Reynolds et al. (1980) studied

a group of 27 patients aged 55 or older (18 males, 9

females) referred to a sleep disorders clinic with

complaints of either excessive daytime sleepiness or chronic

insomnia (most presented with the latter complaint).

Although no mention is made of the exact criterion applied,

the authors report that 18.5% of these older subjects

exhibited a level of apnea activity consistent with a

diagnosis of SAS. All of the patients diagnosable with SAS

were male. Coleman et al. (1981) report apnea incidence

data derived from 83 elderly consecutive referrals to a

sleep disorders clinic. These subjects, aged 60 to 85,

manifested a wide range of sleep/wake complaints.

Seventy-two percent of these complaining elderly patients

experienced at least 30 apnea episodes per night and the

average number of events for all subjects was 150.

In summary, two studies concentrating specifically on

elderly subjects with sleep complaints found a remarkably

high incidence and level of sleep apnea activity, while two

others found lower rates. These latter studies (Kales et

al., 1982 and Reynolds et al., 1980) found that only 0% and

18.5% of older males exhibited very high numbers of apneas.

In contrast, Ancoli-Israel et al. (1981) and Coleman et al.

(1981) found 55% and 72% of their older males to exhibit

high numbers of events. Because the measurement and scoring

methodologies used by these 4 studies were essentially

identical, the wide discrepancy must be attributed to

systematic differences in the samples studied. The most

obvious difference in sample composition lies in the types

of sleep complaints for which patients were referred to the

various sleep disorder clinics. The Kales et al. and

Reynolds et al. samples were made up almost exclusively

of patients with complaints of insomnia; the Ancoli-Israel

et al. and Coleman et al. samples presented with a more

heterogeneous set of complaints (and in particular included

substantial proportions of subjects with complaints

suggestive of sleep apnea). It may be concluded from these

studies that 1) a complaint of insomnia is not predictive of

an increased level of sleep apnea, and 2) heterogeneous

samples of older sleep disorders clinic patients display a

remarkably high prevalence rate for clinically significant

levels of sleep apnea activity. The Kales et al. (1982)

findings, derived from insomniacs of all ages and indicating

a significantly increased incidence of apnea activity in

older patients, suggests (in the absence of direct young vs.

old comparisons in the other three studies) that age is

importantly and positively related to the incidence of sleep

apnea activity in complaining individuals.

The following broad conclusions may be drawn from the

17 above-reviewed studies: 1) sizable proportions of elderly

subjects without prominent sleep or respiratory complaints

experience at least some sleep apnea each night; 2) amount

of apnea activity is age-related regardless of subject

selection strategy; 3) perhaps as many as 39% of healthy

elderly males selected without regard to sleep or

respiratory complaints experience what has come to be

regarded as a clinically significant amount of sleep apnea

activity; and 4) the high numbers of respiratory events

experienced by elderly males are accompanied by high numbers

of oxygen desaturations. These findings are compelling,

given the widespread belief that the waking deficits of SAS

patients are directly attributable to the high numbers of

nocturnal respiratory events they experience, and they

suggest these questions: What, if any, are the

ramifications of the near-ubiquity of sleep apnea and

desaturation activity in elderly subjects, and do

nonclinical elderly subjects with high numbers of

desaturations suffer from SAS-like daytime deficits? The

present study attempted to answer these questions in two

ways. First, by examining correlations between indices of

sleep-disordered breathing and measures of the deficits

commonly seen in SAS patients. Second, by identifying

nonclinical elderly subjects with high numbers of

desaturations and comparing their health, sleep/wake, and

cognitive status to that of subjects exhibiting a lower

level of desaturation activity. The resulting patterns of

between-group differences allowed for judgments to be made

concerning the effects of relatively high numbers of


The cardiopulmonary complications, hypersomnolence, and

assorted other negative clinical features (e.g.,

intellectual deterioration and disrupted nocturnal sleep)

commonly reported in patients diagnosed with Sleep Apnea

Syndrome are thought to be related to the repetitive sleep

apneas, hypopneas, and desaturations those patients

experience. It would seem important to examine the research

findings on which that conclusion is based and to examine

research findings with some bearing on the question of

whether high levels of apnea in nonclinical samples are also

associated with daytime sequelae. If, as the above-reviewed

studies indicate, aging is associated with an increase in

apnea activity, and if higher levels of apnea are associated

with waking sequelae in nonclinical samples (as suggested

by, among others, Ancoli-Israel et al., 1981, and Carskadon

and Dement, 1981), then many elderly individuals would

appear to be at particularly high risk for those deficits.

As will be seen, there is much evidence that strongly

suggests a connection between sleep-disordered breathing and

various deficits in SAS patients; however, there is a

striking paucity of research examining the possibility of a

relationship between apnea and waking deficits in

nonclinical subjects, and almost no research that has looked

specifically at nonclinical older subjects.

Deficits Associated with Sleep Apnea Syndrome
and Sleep Apnea Activity

Cardiological Deficits

Many studies have investigated the relationship between

sleep-disordered breathing and cardiological functioning.

The great majority of these studies have looked only at the

cardiac functioning of subjects with very high levels of

apnea activity (usually sleep disorder clinic patients

diagnosed with SAS). The three studies reviewed at the end

of this section (Berry et al., 1986b; Ancoli-Israel et al.,

1981; Lavie, 1983) compared the incidence of hypertension in

nonclinical groups stratified by level of apnea activity and

are especially pertinent to the justification for, and the

findings of, the present study. The following studies

report two kinds of data pertaining to cardiac functioning:

1) data concerning the possible long-term effects of

sleep-disordered breathing (waking systemic hypertension and

various other waking abnormalities such as arrhythmias,

ventricular and anterior hypertrophy, cardiomegaly, etc.)

and 2) data concerning the possible acute effects of apnea

on cardiac functioning during sleep (bradycardia,

tachycardia, rises in systemic and pulmonary arterial

pressure, PVC, etc.).

Schroeder et al. (1978) studied 22 male SAS patients

aged between 30 and 63 (mean: 47). Six (27%) of these

subjects had waking systemic hypertension and several others

exhibited waking cardiac abnormalities (left ventricular

hypertrophy, left anterior hemiblock, ST-T wave

abnormalities, borderline cardiomegaly). While asleep,

twenty of the 22 patients manifested significant rises in

systemic arterial pressure, and 21 manifested moderate rises

in pulmonary arterial pressure. These rises coincided with

episodes of sleep apnea. Treatments aimed at either

mitigating the effects or eliminating the occurrence of

apneas tracheostomyy or oxygen administered during sleep)

virtually eliminated the transient abnormalities, thus

strongly implicating the apneas in their etiology. Tilkian

et al. (1978) studied 25 male SAS patients aged between 30

and 63 (mean: 44). Forty-eight percent of these patients

exhibited waking systemic hypertension, while 24 of the 25

demonstrated marked sinus arrhythmias, typically

bradycardia, during apneas. Other apnea-contingent

abnormalities included prolonged sinus pauses, ventricular

tachycardia, and atrial tachycardia. Tracheostomy

completely abolished marked sinus arrhythmias and most of

the other transient abnormalities in the 17 patients who

received this treatment, again strongly implicating the

apneas in their etiology. Zwillich et al. (1982) report

cardiopulmonary data derived from 6 male SAS patients aged

23 to 68 (mean: 41.8). None of these subjects reported a

history of systemic hypertension (surprisingly, these

workers failed to measure waking blood pressure); normal

waking sinus rhythm was present in all subjects, but

bradycardia was observed during 95% of all apneas and became

more marked as apnea length increased and arterial oxygen

saturation decreased. Bradycardia was eliminated in all 4

of the patients administered oxygen-enriched air while

asleep. The authors concluded that bradycardia occurs

during almost all apneas and that the degree of bradycardia

depends upon the degree of desaturation resulting from those


Guilleminault et al. (1978) reported that of a group of

48 male and 2 female SAS patients (mean age: 45; range

28-62), 52% were hypertensive. Kales et al. (1985a) found

that 72% of a sample of 50 SAS patients (43 males, 7

females; mean age: 46) were hypertensive. Coverdale et al.

(1980) report that of a sample of 8 SAS patients (7 males, 1

female; ages not reported), 50% had waking systemic

hypertension. None of these patients demonstrated severe

bradycardia or tachycardia during apneas. Orr et al. (1979)

studied two small groups of SAS patients, one consisting of

4 men with complaints of excessive daytime sleepiness, and

the other consisting of 4 men with equally high numbers of

apnea events but without complaints of sleepiness. All of

the somnolent patients were found to suffer from waking

right-sided heart failure as well as frequent arrhythmias

during apneas; none of the nonsomnolent patients

demonstrated these deficits. These two groups differed on

one important dimension: the somnolent patients experienced

substantially greater oxygen desaturations during apneas

than the nonsomnolent patients. In a study designed to

directly investigate the relationship between

apnea-contingent oxygen desaturation and transient

apnea-contingent arrhythmias, Shepard et al. (1985) studied

31 male SAS patients with an average age of 55 (range

30-76). Forty-eight percent of these subjects were found to

be hypertensive. Apnea-related premature ventricular

complexes (PVC) were observed during the sleep of 74% of

these subjects. A significant correlation between PVC

frequency and oxygen desaturation was found only for those

subjects who displayed desaturations to 60% or lower. The

authors concluded: "Patients with obstructive sleep apnea

are at relatively low risk of developing ventricular

arrhythmias provided oxygen saturation remains greater than

60%, while those with saturation below 60% are at increased

risk and should be managed accordingly" (p.335). Fujita et

al. (1981) reported that of a sample of 12 male SAS patients

(ages not reported), all suffered cardiac arrhythmias while


Berry et al. (1986b) compared three groups of

middle-aged male subjects (all were nonclinical volunteers;

sample mean age: 50; range: 30-75) formed on the basis of

Apnea + Hypopnea Index scores. The groups did not differ

significantly in either age or weight. The group with AHI

scores >5 did not differ significantly on diastolic or

systolic blood pressure readings or on reports of a history

of systemic hypertension from a group of subjects with AHI

scores between 0 and 5 and a group with AHI=0. For the

sample as a whole, a significant positive correlation was

found between systolic blood pressure and number of

desaturations >4%. Ancoli-Israel et al. (1981) compared 9

elderly subjects who attained AI scores >5 with 15 elderly

subjects with AI scores less than 5. These groups did not

differ significantly in age. All subjects were recruited

from nonclinical sources. The high sleep apnea group did

not report significantly greater frequencies of hypertension

or heart failure than the low sleep apnea group. It should

be noted that these workers did not directly measure their

subjects' blood pressure. Lavie (1983) compared 11

middle-aged males with AI scores >10 with 67 middle-aged

males who attained AI scores less than 10. The high AI

group was found to be significantly older than the low AI

group; the groups did not differ in weight. All subjects

were recruited from nonclinical sources. A significantly

higher proportion of the subjects in the high AI group were

found to have hypertension (36% vs. 7%).

The above studies indicate that SAS patients very

frequently suffer from both long term waking and short term

sleeping cardiac abnormalities. Incidence rates for

diagnosable hypertension ranged from 27% to 72%, with the

majority of studies indicating that roughly 1/2 of all SAS

patients are hypertensive. It is quite difficult, however,

to interpret these findings because none of the studies

looking specifically at SAS patients recruited age- and

weight-matched non-SAS patient control groups. Both age

(see previous section of this chapter) and weight (Bixler et

al., 1982; Kales et al., 1982; Block et al., 1980; Block et

al., 1979) have been found to be positively related to

amount of apnea activity, and both of these variables are

known to predict hypertension. Those studies which simply

report hypertension rates in SAS patients cannot determine

the contribution of sleep-disordered breathing to the

development of hypertension in isolation from other known

causal factors. It seems very likely that the increased

incidence of hypertension in SAS patients is at least partly

attributable to their tendency to be older and heavier than

the general population, and it is possible that age and

weight account completely for the high incidence of

hypertension in SAS patients. Clearly, any comparison of

blood pressure measures in groups differing on amount of

apnea or desaturation activity must control for the effects

of age and weight.

Every study (with the exception of one) reporting data

pertaining to acute apnea-contingent cardiac complications

found very high frequencies of arrhythmias, acute rises in

blood pressure, and PVC. Treatments that abolished apneas

also abolished these acute complications, strongly

implicating the apneas as causal factors. Evidence

indicating that the oxygen desaturation engendered by

sleep-disordered breathing is the key etiologic factor in

the manifestation of sleeping arrhythmia has begun to

accrue. Orr (1986) reviewed these oxygen desaturation

findings and concluded that although hypoxemia is not the

sole predictor of ventricular dysfunction, strong enough

evidence of an important relationship now exists, and

individuals with oxygen saturation levels in the range of

60% or below "... should be singled out for immediate

aggressive treatment intervention" (p.2).

Only three studies have compared incidence rates for

hypertension from nonclinical subjects stratified by level

of apnea activity. One study (Ancoli-Israel, 1981) found no

significant increase in hypertension in high AI elderly

subjects but this is a relatively uninformative finding as

they relied solely on patient reports of a history of

hypertension and did not directly measure blood pressure.

They thus could not conduct correlational analyses relating

amount of apnea or desaturation to waking blood pressure.

Lavie (1983) did find an increased incidence of hypertension

in nonclinical middle-aged subjects with substantial amounts

of apnea compared to subjects with varying lesser degrees of

apnea activity, but this study failed to control for the

significantly higher age of the high apnea group (the groups

did not differ in weight). Finally, Berry et al. (1986b)

found no increased incidence of hypertension in middle-aged

males displaying high levels of apnea compared to age- and

weight-matched subjects with lesser amounts of apnea, but

they did find a significant positive correlation between

number of desaturations and waking systolic pressure

--suggesting that desaturations take a toll on waking

cardiac functioning even in subjects who do not meet

'clinical' criteria for the diagnosis of SAS. This finding

is limited, however, by the fact that no effort was made to

partial out from this correlation the reported significant

positive relationships between both age and weight and

measures of desaturation. The independent contribution of

desaturation to higher blood pressure thus remains unknown

from their data. In general, then, it is simply not clear

from the available evidence whether a high level of

sleep-disordered breathing is predictive of an increased

incidence of hypertension in nonclinical subjects. The

present study attempted to investigate the relationship

between nocturnal respiratory activity and blood pressure

while controlling, where appropriate, for other variables

known to predict blood pressure by recruiting elderly males

with varying levels of desaturation activity and then

determining both the incidence of hypertension in

age-matched groups stratified by level of desaturation and

the partial correlation between various measures of sleep-

disordered breathing and waking blood pressure, controlling

for age and weight.


A complaint of excessive daytime sleepiness (EDS) is

one of the formal requirements for a diagnosis of Sleep

Apnea Syndrome (ASDC, 1979) and from this one might assume

that such complaints are invariably present in patients

labeled 'SAS'. In fact, EDS is described by every SAS

patient studied in reports by Tilkian et al. (1978),

Schroeder et al. (1978), Weitzman et al. (1980), Sullivan

and Issa (1980), Coverdale et al. (1980), Garay et al.

(1981), Fujita et al. (1981), Zwillich et al. (1982), Kales

et al. (1985a), and Guilleminault et al. (1986). However, a

focus on a high level of apnea activity (usually an AI score

55) as diagnostic of SAS has led various investigators to

apply the label 'asymptomatic SAS' to those subjects with Al
scores >5 who do not complain of EDS. Orr et al. (1979),

Lavie (1983), Smirne et al. (1980), Guilleminault et al.

(1978), and Sink et al. (1986) have identified and described

such 'asymptomatic' subjects. Dement et al. (1978) describe

the justification typically offered for this labeling


It is not uncommon to see a patient fall
asleep literally seconds after he has stoutly
maintained that he is 'wide awake'. It is
absolutely clear that some patients deny the
existence of severe sleepiness either because
it is psychologically repugnant or because they
have lost their frame of reference for true
alertness. Some patients are simply and honestly
unaware of their countless microsleeps and lapses
of attention (p.26).

This justification confounds two issues, one

methodological and.the other theoretical. It is at least

implied in the above statement that if a perfect measure of

sleepiness existed (that is, one not susceptible to

distortion by the subject) all persons with high levels of

apnea activity would be shown to be "symptomatic." Although

it is helpful and useful that Dement et al. call attention

to the need for a refinement of measures of sleepiness, in

doing so they regard as settled a question that is in fact

far from settled and which cannot be decided on the basis of

belief but rather on the basis of experimental inquiry.

That question concerns whether it is possible to experience

consistently high numbers of apneas without experiencing

excessive daytime sleepiness. In addition to leading to the

premature closure of an important issue, the use of the term

'asymptomatic' promotes the illogical concept of a "disease"

without symptoms. The promotion of such a concept would

seem to be of benefit only to sleep disorder clinics seeking

to attract patients.

The unreliabilty of global self-reports of non-EDS can

be circumvented through a more objective quantification of

sleepiness, and several studies have in fact attempted to go

beyond a reliance on patient self reports to directly

measure daytime sleepiness. Dement et al. (1978) used the

Multiple Sleep Latency Test (MSLT) to compare the daytime

sleepiness of 16 SAS patients and 6 matched controls without

sleep apnea. The MSLT is a procedure in which subjects are

placed in a quiet, darkened room during the day and

instructed to attempt to fall asleep. Their ability to do

so is measured through continuous monitoring of EEG. Thus,

a patient with EDS will presumably show a very short latency

to sleep in comparison to normals. The MSLT, to be valid,

should be administered repeatedly at regular intervals

throughout the day. Dement et al. found that their SAS

patients fell asleep consistently and significantly more

quickly than the matched controls. The average sleep

latency of the SAS patients was well within the range of

'pathological' sleepiness (less than 5 minutes). Roth et

al. (1980), Hartse et al. (1980), Zorick et al. (1982), and

Knight et al. (1987) have performed replications of the

Dement et al. study; all found significantly greater

sleepiness in SAS patients compared to non-SAS subjects and

very high proportions of SAS subjects meeting the criterion

for pathological sleepiness. Berry et al. (1986b) also

attempted to replicate these findings, using nonclinical

middle-aged male subjects stratified by level of apnea

activity, but found no difference in MSLT results when

subjects with AHI scores >5 were compared with subjects

without apnea activity. This is a somewhat flawed finding,

however, as the MSLT procedure was performed only twice for

each subject (in the late afternoon and then early

evening). Importantly though, Berry et al. did find

evidence of a relationship between apnea activity, in the

sample as a whole, and daytime sleepiness, in the form of

significant positive correlations between both Apnea +

Hypopnea Index score and Number of Desaturations >4% and

questionnaire findings concerning number of daytime naps

typically taken. Similarly, Timms et al. (1985) correlated

various measures of nocturnal oxygen desaturation with MSLT

results derived from 42 SAS patients. They found several

significant correlations indicating that degree of

desaturation is predictive of daytime sleepiness. Roth et

al. (1980) performed the same analyses with a much smaller

sample of SAS patients (n=10) and found no significant

relationship between desaturation indices and MSLT results;

they did, however, find that a higher Apnea Index predicted

greater daytime sleepiness. Two other studies provide

information relevant to the question of whether apnea

activity and/or oxygen desaturation is related to daytime

sleepiness. Orr et al. (1979) compared 4 SAS patients with

complaints of EDS and 4 SAS patients without EDS and found

that the two groups differed, on average, in the degree of

desaturation they experienced during apneas: the EDS group

exhibited much more severe desaturations. This finding

supports the idea that apnea-contingent hypoxemia is an

important cause of daytime sleepiness. Sink et al. (1986)

replicated this study with much larger groups of subjects

with and without complaints of daytime sleepiness (matched

for a high level of apnea) and found essentially the same

group differences in degree of apnea-contingent


The above studies indicate rather clearly that 1) SAS

patients exhibit a very high incidence of excessive daytime

sleepiness, and 2) this sleepiness seems to be related to

the multiple apnea-contingent oxygen desaturations those

patients experience. These studies do not, however, speak

to the nature of the causal relationship between sleep-

disordered breathing and daytime sleepiness. One hypothesis

concerning this causal connection asserts that it is the

high degree of sleep disruption experienced by SAS patients

which leads to their daytime sleepiness. This hypothesis is

derived from research indicating that periodically disturbed

sleep is less restorative than nondisturbed sleep, even when

total sleep time is relatively unaffected (Bonnet, 1985;

Bonnet, 1986). As will be seen in the next section,

research indicates that episodes of apnea and hypopnea are

almost invariably followed by a brief arousal (usually less

than 30 seconds in duration) during the resumption of

breathing. It has been suggested that these arousals which

disrupt the continuity of sleep predict daytime sleepiness

in SAS patients. Three studies have tested this

hypothesis. Roth et al. (1980) found that the variable

arousalss subsequent to respiration" was the single best

predictor (r= -.52) of daytime sleepiness (as measured by

the Multiple Sleep Latency Test) in a group of 10 SAS

patients, and Stepanski et al. (1984) found a significant

correlation (r= -.33) between number of brief arousals and

MSLT results in 15 SAS patients. Bonnet et al. (1986)

divided a group of 12 SAS patients into 3 subgroups on the

basis of AI scores (AI between 5 and 20, AI between 20 and

45, and AI greater than 45) and then compared these groups

on daytime sleepiness, EEG sleep variables, and 3 measures

of 'sleep continuity'. The sleep continuity variables

consisted of the median length of time between each apnea

event and the median of the longest period and of the 5

longest periods of undisturbed sleep within every 100 apnea

events. These variables thus reflected the extent to which

each patient obtained uninterrupted sleep. Results

indicated that for all of the patients apneas tended to

occur in clusters at the rate of 1 per minute, but that

those patients with lesser amounts of apnea accrued a median

of 80 and 24 minutes of undisturbed sleep between clusters;

the patients with severe apnea accrued a median of only 9

minutes of undisturbed sleep between clusters. These

differences between the groups were significant. It is

interesting to note that while the groups also differed in

sleepiness, none of the EEG sleep variables (Time in Bed,

Sleep Latency, Sleep Efficiency, stage percentages) differed

significantly between groups, suggesting that discontinuity

of sleep and not sleep structure predicts increased daytime

sleepiness. This suggestion was confirmed by the finding

that the median of the longest periods between apnea events

correlated significantly (r= -.68) with daytime sleepiness;

Apnea Index and EEG variables did not correlate

significantly with sleepiness. The authors concluded "...

it may be more important to track periods of consolidated

sleep rather than simply numbers of apneic events in

predicting the extent of daytime sleepiness" (p.105). These

findings support the notion that it is the disruption of

sleep attendant upon periodic oxygen insufficiency which

determines the level of daytime sleepiness in SAS patients.

In summary, although a high proportion of SAS patients

complain of excessive daytime sleepiness, some do not.

Dement et al. (1978) suggest that SAS patients who do not

report EDS cannot be relied upon for valid self-report

because their internal norms for sleepiness have been

distorted during years of almost constant daytime

drowsiness. This (in part) is a measurement issue which can

be addressed through the use of multiple strategies for

measuring daytime sleepiness, and the present study utilized

several approaches to the measurement of this variable.

There is some evidence indicating that the key etiologic

factor in the development of apnea-related EDS is the degree

of sleep disruption caused by the arousals which follow

oxygen desaturation. The present study attempted to

replicate these findings (but in an only indirect fashion as

no independent measure of arousal was obtained) through the

computation of correlations between various measures of

nocturnal respiratory disturbance and various measures of

daytime sleep tendency, and through the comparison of groups

differing in amount of nocturnal desaturation on measures of

sleepiness. Finally, and most importantly, there is a

striking paucity of findings relevant to the question of

whether high levels of apnea activity are predictive of

daytime sleepiness in nonclinical subjects. Only one study

(Knight et al., 1987) has recruited non-clinic elderly

subjects (who are at very high risk for multiple apneas)

with a wide range of apnea and desaturation activity for the

purpose of examining correlations between nocturnal

respiratory indices and measures of daytime sleepiness.

This study found a significant group difference in daytime

sleepiness between subjects falling above and below AI=5.

However, no significant correlations between their limited

measures of oxygen desaturation and MSLT results were

found. Berry et al. (1986b) attempted to investigate this

question (using middle-aged males); they found no

relationship between nocturnal respiratory indices and MSLT

results, but did find that at-home napping tendency is

significantly related to increasing oxygen desaturation.

The somewhat contradictory findings of these two studies are

in need of further investigation. Through the recruitment

of nonclinical elderly males, the present study represents

an attempt to clarify these findings and to test the Block

et al. (1979) conjecture that older subjects may show a more

reliable relationship between level of apnea activity and

daytime sleepiness than do younger subjects.

Disruption of Nocturnal Sleep

Although the disruption of nocturnal sleep is not a

formally recognized 'symptom' of Sleep Apnea Syndrome, many

authors have drawn attention to the short-term arousals

caused by apneas and hypopneas and to the apparent

distortion of sleep stage distribution attendant upon

multiple respiratory events. Further, several authors have

suggested that the excessive daytime sleepiness and commonly

noted intellectual changes reported by SAS patients are at

least partly attributable to the fragmented sleep those

patients experience (Guilleminault et al., 1976; Carskadon

et al., 1980; Bonnet, 1985). Their thinking on this matter

of etiology, as well as the thinking of authors who

emphasize hypoxemia as an etiological factor, will be

discussed in a later section. To inform that discussion,

studies reporting data concerning the stages in which

respiratory events tend to cluster, the differences between

the sleep of SAS patients and non-SAS patients, and the

effects on sleep of treatments which substantially reduce

apnea activity in SAS patients, will be reviewed in this


Block et al. (1979) studied the sleep of 49 non-SAS

subjects without sleep and breathing complaints and found

that apnea events and desaturations occurred more frequently

in Stages 1 and REM and less frequently in slow wave sleep

than would be expected from the amount of time their

subjects spent in those stages. Block et al. (1980) found

that in a sample of 38 women aged 20-74 without sleep and

breathing complaints, apneas and desaturations occurred

disproportionately more frequently in stages 2 and REM.

Bixler et al. (1982) studied the sleep of 100 non-SAS

patients aged 18-74 and found apnea events to cluster

disproportionately in Stages 1 and REM and to occur less

frequently than would be expected in slow wave sleep. The

same group of researchers (Bixler et al., 1985) reported

data from an expanded number of subjects and found similar

results. Krieger et al. (1983) report data derived from 40

non-SAS patients aged 20-76 and, as did the above studies,

found apneas to cluster disproportionately in Stages 1, 2,

and REM.

Guilleminault et al. (1976) were one of the first sets

of workers to draw attention to the fragmented sleep of SAS

patients. They studied the sleep of 50 SAS patients aged 22

to 70 and found that most episodes of apnea were associated

with brief (usually less than 1 minute in duration)

arousals. These arousals were noted through the appearance

of EEG alpha waves and an increase in chin muscle tonus upon

the resumption of breathing at the end of a respiratory

event. They reported that their SAS subjects experienced

significantly reduced amounts of slow wave sleep (Stages 3

and 4) and significantly increased amounts of Stage 2 sleep,

in comparison to age norms. Roth et al. (1980) compared the

sleep of 10 SAS patients and 10 age-matched controls and

found that the SAS patients experienced significantly less

REM sleep and significantly more Stage 1 sleep. Further,

the SAS patients manifested more shifts between sleep stages

and a shorter latency to sleep upon retiring. These workers

found short-term arousals to almost invariably accompany the

resumption of breathing after an episode of apnea.

Similarly, McGinty et al. (1982) found short term arousals

to be invariably associated with episodes of apnea and

hypopnea in healthy elderly males. Hesla et al. (1985)

compared the sleep of 146 SAS patients and 20 controls of

similar mean age and found the SAS patients to have

significantly reduced slow wave and REM sleep percentages.

These workers also found that an increasing Apnea + Hypopnea

Index score significantly predicted less time spent in REM

and slow wave sleep, more shifts between stages, and more

time spent awake. Bixler et al. (1985) compared the sleep

of subjects with and without sleep apnea activity and found

no differences in amounts of wakefulness or Stage 1 sleep

(they did not, however, perform comparisons for amounts of

other sleep stages). Krieger et al. (1983), in a study of

the sleep of 40 non-SAS patients of all ages, found number

of apneas to be significantly positively correlated with

number of shifts between sleep stages. Finally, Berry et

al. (1986b) compared the sleep of nonclinical subjects

stratified by level of apnea activity and found no

significant differences in sleep stage amounts. Further,

they found no significant correlations between indices of

nocturnal respiratory disturbance and sleep stage

variables. Berry et al. (and most of the other studies

mentioned above) did not measure short arousals; it

therefore remains unknown from their data whether high

numbers of respiratory events in nonclinical subjects are

predictive of high numbers of post-event arousals. However,

the findings of Guilleminault et al. (1976) and Roth et al.

(1980), derived from SAS patients, and the findings of

McGinty et al. (1982), derived from healthy elderly males,

suggest that arousals are an almost invariant result of

apneas and hypopneas.

With regard to the effects on sleep of treatments aimed

at reducing apnea activity in SAS patients, Weitzman et al.

(1980) found that in 10 patients whose sleep was recorded

before and after receiving tracheostomy, the amount of time

spent in slow wave sleep was significantly increased and

Stage 1 percent, number of awakenings, and number of shifts

between stages was significantly decreased. In a similar

study, Roth et al. (1980) found number of stage shifts and

percentage of time spent in Stage 1 to be decreased after

tracheostomy. Holzer et al. (1985) treated 5 SAS patients

with continuous positive airway pressure (CPAP) and found

slow wave and REM sleep to be significantly increased and

Stage 1 sleep to be significantly decreased after

treatment. Issa and Sullivan (1986) also used CPAP to treat

12 SAS patients and found similar significant increases in

slow wave and REM sleep and significant decreases in stage 1

and 2 sleep after treatment. They found CPAP to have no

effect on wakefulness.

In summary, there is a striking consensus across

studies concerning the temporal placement of apneas within

sleep: apneas tend to cluster in light sleep (Stages 1 and

2) and in REM sleep, and they tend not to occur during slow

wave sleep. From the sparse information available

concerning the differences in the sleep of SAS and non-SAS

subjects, it may be tentatively concluded that SAS patients

experience significantly reduced amounts of either REM or

slow wave sleep and correspondingly increased amounts of

light sleep (Berry et al., 1986b found no such differences

in a nonclinical sample). Because no differences in amount

of wakefulness have been found, it appears that the effect

of multiple apneas is simply a redistribution of sleep stage

percentages: away from REM or slow wave sleep and toward

light sleep. This redistribution is reflected in an

increased number of shifts between sleep stages. Multiple,

but very brief arousals (so brief that they are not

traditionally scorable as 'wakefulness') almost invariably

result from multiple apneas. Significantly reducing the

number of nocturnal respiratory events through treatment

leads directly to an increase in time spent in REM and slow

wave sleep and to decreases in both time spent in light

sleep and number of shifts between stages. These findings

argue strongly for the view that sleep-disordered breathing

causes the above-outlined distortions of nocturnal sleep.

The possible relevance of these distortions and the brief

arousals which appear to cause them as etiologic factors in

the development of the intellectual deficits of SAS patients

will be discussed shortly; it should suffice at this point

to note that very little experimental work has directly

tested the relationship between the disrupted sleep

associated with sleep apnea activity and cognitive deficits;

the present study attempted to do so.

Cognitive Deficits

Until very recently, evidence for the assertion that

SAS patients suffer from intellectual deficits was

exclusively anecdotal in nature (cf., Guilleminault, et al.

1976). In 1985, studies attempting to test this assertion

began to appear, as did studies concerned with the

possibility that sleep apnea activity is predictive of

cognitive functioning in nonclinical subjects. The findings

of the 8 currently-available studies providing information

relevant to these two issues will be reviewed in this


Kales et al. (1985b) administered the WAIS-R, Wechsler

Memory Scale (WMS), and Bender Test of Visual-Motor

Integration to 50 SAS patients (mean age: 46; range 25-68).

No control group of non-SAS patients was used. They found

that 1) 39% of these patients were impaired on the WAIS-R

(defined as PIQ at least 15 points lower than VIQ); 2) 50%

fell in the impaired range on the WMS (defined as a Memory

Quotient less than 86); and 3) 24% showed moderate to severe

impairment on the Bender (defined as a Z score below 72).

These workers reported no correlations relating measures of

nocturnal respiratory disturbance to cognitive functioning.

Findley et al. (1986) recruited 26 SAS patients with a mean

age of 52. They split their subjects into 2 age- and

education-matched groups: those whose median oxygen

saturation at night was below 90% were labeled Hypoxemics,

and those whose median saturation was greater than 90% were

labeled Non-hypoxemics. They then compared these two groups

on 8 tests of neuropsychological functioning (including a

short form of the WAIS-R, Trail-making part B, a measure of

vigilance/eye-hand coordination, a test of rapid, complex

problem-solving, and tests of immediate and delayed memory

for verbal and non-verbal information). The hypoxemic group

performed significantly more poorly on 4 of the 8 tests,

displaying deficits in immediate and delayed verbal memory,

attention/concentration (Trails B), and vigilance/eye-hand

coordination. Also found was a significant negative

correlation between median sleeping oxygen saturation and a

rating of overall cognitive impairment made by a

neuropsychologist blind to the group membership of the

subjects. The correlation between this rating and number of

desaturations >4% was not significant. These correlations

did not control for age and education. Surprisingly, no

correlations between individual test scores and measures of

desaturation were computed. Norman et al. (1986) recruited

33 male SAS patients aged 19-68 and administered a 'battery

of neuropsychological tests designed to assess attention,

memory, the ability to shift cognitive sets, and

visual-motor tracking skills' (the names of these tests are

not listed). They correlated scores on these tests with

Apnea + Hypopnea Index scores and mean oxygen saturation and

found AHI to significantly predict poorer verbal memory and

a poorer ability to shift cognitive set; they also found

lower mean oxygen saturation to predict poorer visual-motor

tracking skills. No effort was made in this study to

partial out the effect of age and education on cognitive


In an attempt to investigate the contribution of

excessive daytime sleepiness to the intellectual deficits of

SAS patients, Greenberg et al. (1987) studied 14 SAS

patients, 10 sleep-disordered patients without sleep apnea

activity but matched with the SAS patients on level of

daytime sleepiness (these were patients diagnosed with

narcolepsy or nocturnal myoclonus), and 14 controls without

sleep apnea or excessive daytime sleepiness. All groups

were age- and education-matched and all subjects were aged

less than 50. An important flaw in this study relates to

the measures of daytime sleepiness used for the purpose of

matching SAS and non-SAS patients. These authors relied on

nocturnal sleep latency (as opposed to daytime Multiple

Sleep Latency Tests) and on a 'clinical rating of observed

sleepiness obtained from the sleep lab director' to gauge

the sleepiness of each of the subjects. These are very

problematic measures of daytime sleepiness nocturnal sleep

latency is not, strictly speaking, a measure of daytime

sleep tendency, and a non-blind experimenter-offered rating

of the sleepiness of the subjects is obviously open to

systematic bias. It is interesting to note that MSLT data

were available for 7 of the 10 sleepy controls. The mean

MSLT latency for these subjects was 6.9 minutes; this is a

value outside the range of pathological sleepiness,

suggesting that these sleepy controls suffered from less

than extreme daytime somnolence. Although no MSLT data were

obtained from the SAS patients, an estimate of their

probable level of sleepiness may be derived from their mean

Apnea Index of 48. Roth et al. (1980) found that for a

group of 10 SAS patients with a mean Apnea Index of 50 the

average MSLT latency was 1.9 minutes, well within the range

of pathological sleepiness. It can thus be concluded with

some confidence that the SAS patients studied by Greenberg

et al. were much sleepier that the 'sleepy' controls. This

less than adequate matching of daytime sleepiness in the SAS

and non-SAS patients obviously hinders the ability of this

study to eliminate daytime sleepiness as an explanation for

any relative performance deficits observed in the SAS

patients. Each subject was administered a battery of

neuropsychological tests consisting of 4 subtests of the

WAIS-R, the Bender, Trail making B, the Purdue Pegboard

(right, left, and both hands summed), a letter cancellation

test, the controlled oral word association test, and the

Wechsler Memory Scale Logical Memory and Figural Memory

subtests (both immediate and delayed recall). One-way ANOVA

comparisons found the SAS patients to perform significantly

more poorly than the 'sleepy' controls on 5 of the 15

measures (Digit Span, Bender, Purdue Pegboard right and left

hand, and letter cancellation). The SAS patients performed

significantly more poorly than the non-sleepy controls on 3

of the measures (WMS delayed recall of Figures, Purdue

Pegboard right and left hand). Interestingly, the sleepy

controls and the non-sleepy controls were not found to

differ significantly on any of the cognitive measures. This

suggests that these sleepy controls were not significantly

more sleepy than the healthy controls and/or that

traditional neuropsychological measures are not especially

sensitive to the effects of moderate daytime sleepiness.

Both conclusions seem justifiable. These workers also

computed zero-order correlations between 4 measures of

nocturnal respiratory disturbance (these 4 measures were

derived from apnea data only; hypopneas were not scored and

desaturations not associated with apneas were ignored) and

the 15 cognitive measures, using data from the 14 SAS

patients only. Four of the 60 correlations were significant

(a number barely exceeding that expected by chance alone);

poorer performance on Block Design and Purdue Pegboard (both

hands summed) was significantly predicted by increasing Time

Not Breathing and by a lower Lowest Desaturation achieved.

It seems possible that the number of significant

correlations was limited both by the small number of

subjects from which the correlations were computed and by

the failure of these workers to quantify desaturations not

associated with apneas. Given the above outlined problems

with the sleepiness matching procedure, the following

conclusion offered by the authors does not seem fully

warranted: "This study's main finding is that hypoxemic

sleep apnea patients experience neuropsychological

dysfunction beyond what can be attributed to the effects of

either excessive daytime sleepiness or aging" (p.260). It

can in fact only be safely concluded from these findings

that pathologically sleepy SAS patients perform more poorly

than moderately sleepy non-SAS patients. No judgment can be

made concerning the relative contribution of hypoxemia to

this difference in performance.

Yesavage et al. (1985) recruited 41 exclusively elderly

males (mean age: 69). All subjects were demonstrated

through testing to be non-demented. Roughly half of the

subjects were recruited through advertisements (the text of

which is not described) and the remainder were sleep

disorder clinic patients referred by their physicians for

unstated reasons. Because the authors report that 73% of

these subjects achieved Apnea + Hypopnea Index scores

greater than 5, this sample must be considered

unrepresentative of elderly males in general, although

potentially interesting in the sense that a wide range of

apnea activity was probably captured. A battery of

neuropsychological tests consisting of the Raven Colored

Matrices, the Peabody Picture Vocabulary Test (PPVT), the

WAIS-R Digit Symbol subtest, the Benton Visual Retention

Test, the Face-hand test, the Bushke list learning test, a

test of reading comprehension, the Stroop color naming test,

and Trail making parts A and B was administered to each

subject and these scores were correlated with Apnea +

Hypopnea Index (these workers did not measure oxygen

saturation). Five of these 11 zero-order correlations were

significant (no effort was made to control for the possible

effects of age and education on these correlations):

increasing AHI predicted poorer performance on the PPVT, the

Raven, the Stroop, Trail making B, and Digit Symbol. Berry

et al. (1986a) recruited 46 middle-aged male snorers from

nonclinical sources (mean age: 50; range: 30-75) and

administered a battery of cognitive tests consisting of the

WAIS-R, Wisconsin Card Sort, Hooper Visual Organization

Test, Wechsler Memory Scale, Rey-Osterreith Complex Figure,

Verbal Fluency, and Finger Tapping. These subjects

displayed a wide range of sleep apnea activity, with 13% of

the subjects displaying a high level of apnea/hyopopnea

activity (AHI>5). Partial correlations (controlling for age

and education) between 3 measures of nocturnal respiratory

disturbance and the 12 scores derived from the battery were

computed. Ten of these 36 correlations were significant;

all indicated a negative relationship between apnea activity

and cognitive skill. Number of desaturations >4% proved to

be the most consistent predictor of neuropsychological

functioning, correlating significantly (and negatively) with

WAIS-R VIQ and PIQ, Verbal Fluency, Wechsler Memory Scale

Memory Quotient, Wechsler Memory Scale delayed recall of

Logical Stories, and delayed recall of the Rey-Osterreith

figure. Apnea + Hypopnea Index also correlated negatively

with WAIS-R PIQ; Apnea Index was significantly negatively

correlated with WMS Memory Quotient, delayed recall of WMS

Logical Stories, and delayed recall of WMS Figures.

Analysis of variance results from unpublished data derived

from the same study (Berry, 1985/1986) indicate that when

these subjects were stratified into age- and

education-equivalent groups based on level of apnea/hypopnea

activity, the group with a high level of apnea activity

performed significantly worse than the group with no

apnea/hypopnea activity on 4 measures (WAIS-R PIQ, WMS

delayed recall of Logical Stories, delayed recall of Rey

Figure, and Verbal Fluency) and significantly worse than the

group with a subclinical level of apnea activity on 1

measure (delayed recall of Rey Figure).

In an important, very recent study similar to the

present one in many respects, Knight et al. (1987) recruited

27 elderly subjects of both sexes from community sources.

All subjects were aged 65 or older (mean age: 76) and were

selected without reference to sleep complaints. Potential

subjects with known dementia or heart disease were excluded

from participation. The subjects were divided into 2

age-equivalent groups based on level of apnea activity

(Apnea Index less than 5, n=17; Apnea Index >5, n=10).

Hypopneas were not scored and desaturations not associated

with apneas were ignored. A battery of neuropsychological

tests similar to those used by previous studies was

administered to each subject and the intellectual

functioning of the two groups was compared using T-tests.

No group differences were found on any of the 14 measures of

cognitive skill. Further, when the High AI group was split

along the median of apnea activity (i.e., 5 subjects with AI

between 5 and 18, and 5 subjects with AI greater than 18),

no significant differences between these small subgroups on

the neuropsychological measures were found. The authors did

derive two (rather limited) measures of oxygen desaturation:

mean desaturation during apneas and lowest saturation

attained during apneas; and correlated these, AI, and mean

apnea duration with the scores derived from the

neuropsychological battery. None of these correlations was

significant. The authors concluded that an AI score greater

than 5 is not predictive of cognitive deficits in an

elderly, nonclinical population, and suggested that a higher

cutting score for a judgment of 'clinical significance' is

required for this population. The authors do not directly

discuss or attempt to explain the failure of their

correlational analyses to find any of the relationships

between nocturnal respiratory dysfunction and cognitive

functioning reported by the other studies of nonclinical

subjects reviewed above. Two comments must be made

concerning the findings of this study. First, their

conclusion is warranted although not particularly

enlightening. They have demonstrated that by ignoring the

wider range of possible respiratory events (i.e.,

hypopneas), with which sizable desaturations have repeatedly

been shown to be associated and which occur with great

frequency in older subjects, groups formed with reference to

apneas only do not differ in cognitive skill. Apnea Index

alone is not a complete measure of the hypoxemic status of

these subjects and it therefore cannot be inferred from

these findings that multiple nocturnal desaturations are not

predictive of cognitive dysfunction. These findings argue

persuasively for a grouping strategy that relies on

information pertaining to number of desaturations. The

application of such a strategy in this study would probably

have shifted several subjects from the low group to the high

group, and might or might not have resulted in group

differences in neuropsychological functioning with either

outcome a more valid test of the effects of desaturation on

cognitive functioning would have been performed. A second,

related comment concerns the limited nature of the

desaturation indices derived by these authors and the

resultant limited utility of the correlations computed with

those indices. As mentioned above, any desaturation not

directly associated with an apnea, no matter how large, was

ignored. From the finding of no significant correlations it

can therefore only be concluded that apnea-related

desaturations, taken in isolation, are not predictive of

neuropsychological functioning. Further, the authors made

no attempt to quantify the cumulative severity of

apnea-related desaturations; they simply noted the average

apnea-related drop in saturation for each subject and the

lowest saturation achieved by each during an apnea. Thus, a

subject with hundreds of exclusively moderately severe

desaturations would have scored 'better' on their measures

than a subject with only one very large desaturation. Given

what is known about the deleterious effects of chronic

low-level hypoxemia (to be reviewed in the following

section), this is an illogical way to rate the severity of a

subject's nocturnal hypoxemia. Clearly, what is needed is a

measure reflecting the absolute number of desaturations a

subject experiences across the course of a night. It thus

remains unknown whether the inclusion of information about

the repetitive desaturations associated with hypopneas, and

about the cumulative severity of desaturations, would have

improved the predictive strength of the relationship between

respiratory disturbance and cognitive skill. The present

study addressed all of these criticisms through a much more

thorough assessment of apneas, hypopneas, and desaturations

of varying degrees and durations.

Finally, Klonoff et al. (1987), intending to test the

hypothesis that the commonly-reported cognitive deficits of

SAS patients are due not to disease-specific etiological

factors but instead to situational anxiety and depression

secondary to being labeled as suffering from a 'disease',

recruited 11 SAS patients scheduled for imminent surgery to

treat their nocturnal respiratory disorder and 11

age-matched patients diagnosed with significant heart

disease who were scheduled for imminent coronary-bypass

surgery. They were administered a wide-ranging battery of

neuropsychological tests, similar to those used by the above

studies, both before surgery and 3 months post-surgery.

T-test comparisons revealed no group differences in

cognitive skill before or after surgery. No correlations

between indices of nocturnal respiratory disturbance and the

measures of intellectual functioning were computed. The

authors concluded: "... there were no cognitive changes that

were specific to SAS. ... The findings of this study

indicate that many of the previously reported cognitive

sequelae associated with SAS are situational reactions and

coping mechanisms to disease per se and resolve with

successful treatment..." (p. 212). These conclusions are

not warranted by the data. First, without a control group

of SAS patients not scheduled for surgery it is not possible

to evaluate whether (as the authors assume) these cognitive

data are generalizable to all SAS patients, or whether the

pre-surgery test data were adversely affected by anxiety,

making the results non-representative of SAS patients in

general and applicable only to SAS patients about to undergo

major surgery. Second, without a control group of

age-matched healthy controls it is not possible to establish

whether both SAS patients and patients with serious heart

disease suffer from real neuropsychological deficits

(perhaps resulting from the common etiological pathway of

long-term oxygen insufficiency). A comparison of the two

groups' scores post-surgery with data from healthy controls

would have controlled for emotional factors and provided an

answer to this second unresolved issue.

The results of the above studies are difficult to

summarize. As with any relatively young literature,

investigators have not yet arrived at a consensus on issues

concerning which variables are of most interest, which

subjects are of most interest, and which questions should be

asked of their data. A few trends are, however, apparent.

When the neuropsychological functioning of subjects with

high levels of nocturnal respiratory disturbance has been

compared to that of subjects with no sleep-disordered

breathing or to that of subjects with high levels of

sleep-disordered breathing not accompanied by untoward

desaturation, significant differences in functioning have

been found. The two studies not finding such differences

were importantly flawed by either a failure to measure the

full range of nocturnal respiratory events or by a failure

to use appropriate control groups. Of the three different

studies performing appropriate comparisons, at least two

have found subjects with high levels of apnea/desaturation

activity to perform significantly more poorly on measures of

intelligence, short-term memory, attention/concentration,

visual-motor integration, and motor efficiency. One study

found verbal fluency to be depressed in high AHI subjects.

Further, several studies found significant negative

correlations between either AHI or meaningful measures of

oxygen desaturation and measures of verbal and non-verbal

intelligence, verbal and non-verbal short-term memory,

attention/concentration, motor efficiency, and visuo-spatial

reasoning/organization. No study has as of yet provided the

information necessary to settle the question of whether

otherwise healthy elderly subjects who experience high

levels of desaturation suffer from cognitive deficits

relative to their non-apneic peers, no study has directly

tested the notion that desaturation predicts cognitive

performance in this age group, and no study has as of yet

provided data that would allow for an estimate to be made of

the relative contributions of desaturation and sleepiness to

cognitive deficits. The present study was designed

primarily to investigate the first 2 of these 3 issues; an

effort to investigate the third issue would be necessary

only when and if the results of preliminary studies such as

the present one suggested a connection between desaturation

and neuropsychological abilities.

Current Thinking on the Etiology of
Apnea-related Cognitive Deficits

One of the issues the present study attempted to

address concerns the etiology of apnea-related cognitive

deficits. Two explanations have been offered for the

apparent covariation of nocturnal respiratory disturbance

and neuropsychological performance. Brief outlines of those

explanations and of findings relevant to them will presented

in this section.

Carskadon et al. (1980), among others, suggest that the

cognitive deficits associated with severe sleep apnea may be

the result of disturbed nocturnal sleep. Recall that

research comparing the sleep of control subjects and SAS

patients indicates that SAS patients typically experience

significantly reduced amounts of either REM or slow wave

sleep and significantly increased amounts of light sleep,

brief arousals, and shifts between sleep stages. Carskadon

et al. conjecture that these disturbances of sleep

architecture and continuity affect intellectual performance

through the medium of increased daytime sleepiness. This

notion has recently been refined by Bonnet (1985; 1986), who

has proposed a "sleep continuity" theory of the restorative

function of sleep. This theory asserts that it is neither

total sleep time nor amounts of specific stages of sleep

but rather the experiencing of periods of sleep undisturbed

by arousal which determines the restorative value of sleep.

Bonnet has suggested that these periods of undisturbed sleep

must be at least 10 minutes in duration, and has shown that

regular arousals at a rate greater than once per 10 minutes

of sleep lead to daytime sleepiness and to cognitive

performance decrements very similar to those produced by

total sleep deprivation. Because sleep apneics experience

sleep fragmentation (caused by periodic pauses in

respiration) similar to that which he modeled experimentally

in healthy volunteers, Bonnet has suggested that the

sleepiness- and performance-related problems seen in

patients with sleep apnea may be secondary to sleep

disturbance rather than to reduced blood oxygen levels

(Bonnet, 1987).

One way to test this idea would entail comparing the

neuropsychological performance of two age-matched groups of

subjects with equally pathological levels of daytime

sleepiness but differing on the presence of high levels of

apnea activity. Although such a study has been attempted

(Greenberg et al., 1987; see preceding section), a detailed

examination of the method used to match the daytime

sleepiness of the high and low apnea groups suggested that

sleepiness was only partially controlled. Of the 5 measures

on which the two groups differed (Digit Span, Letter

Cancellation, Purdue Pegboard Right and Left hand, Bender),

the first 4 contain 2 of the several task requirements that

are sensitive to the lapses of attention attendant upon

sleepiness: short term memory and speed of performance

(Johnson, 1982). Because the SAS group was probably

substantially more sleepy than the control group, the

observed group differences on cognitive measures (all in the

direction of poorer performance in the SAS subjects) could

conceivably have been the result of daytime sleepiness.

The possible relevance of daytime sleepiness to

performance deficits in SAS patients has been demonstrated

by a study conducted by Bonnet (1986). In this repeated

measures study healthy, non-sleep disordered volunteers were

subjected to awakenings at regular intervals of 1 minute for

two consecutive nights and then one week later to awakenings

at intervals of 10 minutes for two consecutive nights. The

two conditions modeled the sleep-fragmenting effects of very

high and above average numbers of apnea-related arousals

(subjects in the 1 minute condition were aroused an average

of 134 times across the night, which can be translated into

an Apnea Index score of 20; subjects in the 10 minute

condition were aroused 34 times equivalent to an Apnea

Index of 5). A sleep latency test and performance tests

were administered to the subjects on a baseline morning as

well as on the second morning of each arousal condition.

The performance tests were not traditional

neuropsychological measures but were instead measures known

to be particularly sensitive to the effects of sleep

fragmentation and sleep deprivation: a reaction time test, a

prolonged test of auditory vigilance, and a work-paced,

prolonged test of addition. The two experimental conditions

each produced significant increases in daytime sleepiness

compared to baseline (the 1 minute condition produced a

level of MSLT sleepiness within the pathological range 2.9

minutes, while the 10 minute condition did not 5.9

minutes). Only the every-minute arousal condition produced

significant performance decrements (on Vigilance and

Additions); the 10 minute condition showed intermediate but

not statistically significant effects on performance. These

findings seem to indicate that the sleep fragmentation which

would accompany an Apnea Index of 20 or greater results in

significant daytime performance decrements, while the

sleepiness produced by an Apnea Index of 5 or less is not

sufficient, in and of itself, to produce performance

decrements even on measures especially sensitive to


In summary, daytime sleepiness resulting from the very

fragmented sleep of severe sleep apneics must be considered

a likely contributor to the cognitive performance deficits

they exhibit. While it is not clear whether the sleepiness

resulting from a more moderate level of sleep fragmentation

(such as that produced in patients with Apnea Indexes

between 5 and 20) has a significant effect on cognitive test

performance, it has been shown that fragmentation consistent

with an AI of 5 or less does not lead to significant

performance decrements. These findings point to the need to

control for the effects of daytime sleepiness in any study

attempting to validate additional explanations (such as

chronic hypoxemia) for observed performance differences

between subjects with differing amounts of sleep-disordered

breathing. The present study attempted to control for the

effects of daytime sleepiness through partial correlations

between indices of nocturnal respiratory activity and

measures of neuropsychological skill, controlling for amount

of napping behavior.

The second explanation for the apparent covariation of

nocturnal respiratory disturbance and neuropsychological

functioning emphasizes the etiologic importance of the

multiple episodes of oxygen desaturation which accompany

sleep-disordered breathing. It has been clearly established

that subjects with high levels of apnea/hypopnea activity

experience higher numbers of desaturations >4% and >10% than

subjects with both lesser amounts of apnea/hypopnea activity

and subjects without apnea/hypopnea activity (cf., Berry et

al., 1986b), and it has been shown that the maximum

desaturation achieved during apnea/hypopnea events by

subjects with high levels of apnea/hypopnea activity is at

least twice as great, on average, as that achieved by

subjects with lower levels of apnea activity (cf., Block et

al., 1979; Dolly and Block, 1982; Sink et al., 1986; McGinty

et al., 1982). Evidence establishing a connection between

the frequent, severe desaturations experienced by both SAS

patients (Findley et al., 1986; Norman et al., 1986;

Greenberg et al., 1987) and middle-aged snorers (Berry et

al., 1986a), and neuropsychological performance has begun to

accrue. These findings are not surprising, given what is

known about the cognitive effects of chronic hypoxemia in

other populations. The neuropsychological functioning of

one such population, consisting of persons who suffer from

chronic obstructive pulmonary disease (COPD), has been

relatively well studied.

Grant et al., (1982) administered a battery of tests

consisting of the WAIS, the Halstead-Reitan Battery,

Trailmaking, the Aphasia Screening Test, and the Logical

Story and Drawing subtests of the Wechler Memory Scale (30

minute delayed recall) to very large age-, sex-, and

education-matched groups of COPD patients and healthy

controls. The COPD patients demonstrated significantly

poorer performance on 11 of the 14 measures (the three for

which there was no significant difference were: Reitan

Rhythm, Aphasia Screening, and delayed recall of Logical

Stories). The authors noted that the COPD patients

performed especially poorly on tasks demanding flexible

thought, ability to think abstractly, visual-motor

integration, and motor speed, strength, and coordination.

The authors concluded from significant negative correlations

between measures of both oxygen saturation and of oxygen

transport capability and two ratings of global

neuropsychological impairment that "... cerebral disorder in

hypoxemic COPD is caused, at least in part, by insufficient

oxygenation of brain tissue" (p. 1474).

Prigatano et al. (1983), noting that the COPD patients

studied by Grant et al. (1982) typically experienced

moderate to severe levels of chronic hypoxemia, wished to

determine if COPD patients with only mild levels of chronic

hypoxemia suffer from similar neuropsychological deficits

relative to healthy controls. They recruited 100 mildly

hypoxemic COPD patients and compared their

neuropsychological test performance to that of 25 age- and

education-equivalent healthy controls. All subjects were

administered the Halstead-Reitan Battery, the WAIS, the

Wechsler Memory Scale (with both immediate and delayed

recall of Stories, Figures, and Paired Associates), the

Aphasia Screening Test, Trailmaking, and the Lafayette

Repeatable Neuropsychological Test Battery. These mildly

hypoxemic COPD patients performed significantly more poorly

than the controls on a remarkable 19 of 25 individual

measures. Most notably, the COPD patients showed

significant deficits on PIQ, all subtests of the Wechsler

Memory Scale, Trailmaking B, the Aphasia Screening Test, and

the Category, Tactual Performance, and Memory subtests of

the Halstead-Reitan. The authors concluded that mildly

hypoxemic COPD patients suffer from impairments in memory,

abstract reasoning, and speed of performance. Finally, Krop

et al. (1973) compared the neuropsychological performance of

two groups of COPD patients: one with severe hypoxemia and

the other with only mild hypoxemia. A test battery

consisting of the WAIS, the Wechsler Memory Scale, the

Bender (both with and without Background Interference), the

Facial Recognition Test, and Finger Tapping was administered

to 10 severely hypoxemic and 12 mildly hypoxemic COPD

patients. The severely hypoxemic patients performed

significantly more poorly on both of the Bender-Gestalt

scores and both right- and left-hand Finger Tapping. Only

left-hand finger tapping remained significantly different

between groups after the severely hypoxemic patients were

treated for one month with oxygen-enriched air. The results

of this study indicate that many of the cognitive deficits

demonstrated by severely hypoxemic COPD patients are also

present (perhaps to a more moderate degree) in mildly

hypoxemic patients, and that deficits in motor speed and

visual-motor integration may only result when the degree of

chronic hypoxemia experienced is severe.

In summary, some questions concerning the relative

contributions of disturbed nocturnal sleep and hypoxemia to

the cognitive sequelae of nocturnal respiratory disturbance

remain as yet unanswered; the present study attempted to

address those questions. Findings concerning the

neuropsychological deficits of another chronically hypoxemic

population, COPD patients, indicate that many areas of

intellectual functioning can be affected, although memory,

abstract thinking, perceptual-motor integration, attention,

and speed of performance seem particularly vulnerable. The

present study measured each of these and several other areas

of intellectual functioning in a population expected to

demonstrate a wide range of cumulative severity of chronic

nocturnal hypoxemia.

Some Methodological and Measurement Issues

The design of the present study allowed for an

examination of two methodological questions: 1) Is there a

'first night effect' for indices of nocturnal respiratory

disturbance that is similar in magnitude to that which has

been demonstrated for EEG indices of nocturnal sleep?, and

2) Which of the many possible ways of measuring and

summarizing apnea/hypopnea activity and oxygen desaturation

activity are most useful in predicting daytime sequelae?

In 1966, Agnew, Webb, and Williams published the

results of a study in which they examined the sleep of 43

subjects who slept in a sleep laboratory for 4 consecutive

nights. It was their intention to establish whether the

common sleep-research practice of discarding the first night

of laboratory sleep as unrepresentative is justified. In

brief, they found that this practice is justified; the first

night of sleep, in comparison to the second night, was

characterized by significantly more awake time,

significantly more time spent in Stage 1, and significantly

less time spent in Stage REM. Further, they found that the

latencies to onset of Stages 4 and REM were significantly

delayed, and that the number of shifts between stages was

significantly increased. The sleep obtained by these

subjects on nights 2-4 was very similar, with no significant

differences found among the 3 nights in wakefulness, amounts

of time spent in the various stages, and latencies to onset

of the various stages. The number of shifts between stages

declined significantly on each successive night. The

authors concluded that the untoward effects of sleeping in a

laboratory for the first time adapt out substantially by the

second consecutive night. Several other studies have

attempted to replicate these findings (Webb and Campbell,

1979; Hartmann, 1968; Kupfer et al., 1974; Schmidt and

Kaelbling, 1971). With the exception of Kupfer et al.

(1974), who studied the sleep of psychiatric inpatients,

each found significant differences between nights 1 and 2

and each found that these differences had disappeared by the

second or third night.

Given the above findings indicating that the measures

of sleep obtained during the first night in a laboratory are

unrepresentative of typical sleep, it makes sense to wonder

if indices of sleep-disordered breathing obtained from only

one night of laboratory sleep are similarly

unrepresentative. This is not an idle question, as studies

designed specifically to investigate one or another aspect

of the phenomenon of sleep-disordered breathing rarely

(perhaps only 10-15% of the time) perform recordings of

their subjects for more than one night. Only four studies

to date have attempted to quantify the night-to-night

variability of apnea/hypopnea activity. Block et al. (1981)

measured the nocturnal respiratory disturbance of 10

post-menopausal women at two different times, each separated

by an average of 37 days. These subjects constituted the

placebo-treated control group in a study designed to assess

whether the administration of progesterone decreases

apnea/hypopnea or desaturation activity. The authors

derived the following respiratory-related measures: number

of apneas, number of hypopneas, number of desaturations >4%,

mean low saturation, mean maximum change in saturation

during events, and maximum duration of apneas, hypopneas,

and desaturations >4%. The authors report that there were

no significant differences between the two nights for any of

these variables. Although these data do establish that

measures of apnea activity and desaturation are stable, the

stability they establish pertains only to stability between

two 'first nights' of measurement. The length of time

between measurement nights was so long (roughly 1 month)

that the question of whether a habituation to laboratory

conditions occurs for indices of nocturnal respiration could

not be reasonably answered.

Bliwise et al. (1983) measured apneas and hypopneas in

66 healthy older subjects (mean age: 67; range: 44-88) on

two consecutive nights. Oxygen saturation was not

measured. Apnea + iypopnea Index differed significantly on

the two nights, showing an average increase of 1.9 from the

first to the second night. The correlation between AHI

scores on the two nights was .74 (p< .005). These authors

also examined the nightly variation in number of subjects

meeting the criterion of AHI >5. While 11 subjects met this

criterion on the first night, twice as many met the

criterion on the second night. In only one subject was AHI

>5 on night 1 accompanied by an AHI score <5 on night 2. A

subsample of 19 subjects were recorded on a third

consecutive night; no difference in AHI was found for these

subjects between nights 2 and 3. The authors concluded that

a single night of polysomnography within healthy older

subjects is likely to underestimate the absolute level of

respiratory disturbance seen in a second or third night of

recording, although night 1 apnea/hypopnea data are clearly

significantly predictive of data obtained on subsequent

nights. Based on these findings, the authors caution

against the occasional practice of inconsistently using

different single night findings (that is, using night 1 data

from some subjects and night 2 data from others) when

attempting to relate individual levels of respiratory

disturbance to daytime and other sequelae.

Wittig et al. (1984) reviewed the records of 22

sleep-disorder clinic patients who had been studied on at

least two nights separated by no more than 90 days. The

authors do not specify the proportion of subjects studied on

consecutive nights and they did not perform separate

reliability analyses for subjects studied at varying

intervals of time. These subjects were divided into two

age-equivalent groups based on amount of apnea activity: 11

subjects who experienced more than 100 apnea episodes on

night 1 and 11 subjects who experienced less than 100.

Neither hypopneas nor desaturations were measured.

Correlations between night 1 and night 2 were performed

within each group for the following variables: Apnea Index,

mean duration of apneas, and duration of longest apnea. No

T-tests comparing the two nights were calculated. All 3

night-to-night correlations were significant within the High

Apnea group (correlations ranged from .60 to .88), while 2

of 3 were significant within the 'Low' Apnea group (Apnea

Index did not correlate significantly). The results of this

study are more easily interpretable for the High Apnea

group: subjects with very high numbers of apneas (these

subjects demonstrated an average Apnea Index of 54 + 21)

show very high night-to-night reliability for amount of

apnea activity regardless of the interval of time that

elapses between measurement occasions. The finding that AI

is not significantly reliable between nights in the 'low'

apnea group is almost impossible to interpret. The authors

make no attempt to investigate this finding more fully by

separately analyzing data obtained from consecutive nights

and data obtained at longer intervals; without such an

analysis it is impossible to evaluate the degree of

habituation to laboratory conditions which may or may not

occur in subjects with less than 100 apneas. Further, the

authors make no attempt to draw distinctions between the

members of this group and then separately analyze

reliability data within subgroups; clearly, many of these

low apnea subjects had many apneas and many had very few

(the average Apnea Index within this group was 6.5, but S.D.

was a very high 4.6). Without making such distinctions it

is not possible to determine to whom these findings of

night-to-night unreliability can be generalized. The

failure of these authors to perform T-test comparisons

between nights 1 and 2 is puzzling.

Finally, Kramer and Silva (1986) studied 20

sleep-disorder clinic patients (with complaints of either

excessive daytime sleepiness or loud snoring; ages are not

given) for two consecutive nights. They quantified both

number of apneas and maximum oxygen desaturation and

compared nights 1 and 2 data using T-tests. No correlations

were calculated. No significant difference between the two

nights was found for either variable. However, 45% of these

patients demonstrated an increase in AI from night 1 to

night 2 of greater than 50%; 20% of the subjects

demonstrated an AI <5 on the first night and an AI >5 on the

second night.

In summary, the findings of only 2 of the 4 above

studies are relevant to the question of whether a

habituation of measures of nocturnal respiratory disturbance

occurs across consecutive nights of measurement. The

findings of Kramer and Silva (1986) indicate that the

night-to-night reliability of AI and desaturation in

sleep-disorders clinic patients is relatively high, although

membership in a group defined by AI >5 varies from night 1

to night 2 for 20% of such patients. Bliwise et al. (1983)

found that in healthy older subjects AHI increased

significantly from night 1 to night 2, that 17% of subjects

not diagnosable as SAS on night 1 became so diagnosable on

night 2, and that night 1 AHI was significantly predictive

of night 2 AHI. These are important findings and they are

in need of replication. Bliwise et al. did not measure

oxygen saturation and the night-to-night reliability of this

important aspect of sleep-disordered breathing therefore

remains unknown for healthy older subjects. The present

study attempted to assess the reliability of a wide range of

nocturnal respiratory indices in healthy older males.

Finally, the second measurement issue this study

attempted to address concerns the question of which of the

many possible ways of measuring nocturnal respiratory

disturbance are useful in predicting waking deficits. A

review of the many studies which have attempted to

investigate one or another aspect of sleep-disordered

breathing reveals that there is no consensus concerning

which variables should be quantified and which should be

ignored. Investigators often ignore hypopneas and

desaturations without offering justification. Similarly,

investigators frequently rely solely on counts of apneas,

hypopneas, and desaturations while ignoring the temporal

aspects of these (that is, average duration and total amount

of time spent in events). Investigators who do measure

desaturation often inexplicably fail to offer counts of

desaturations of differing degrees of severity and to

compute the absolute amount of time spent in desaturations

of varying degrees and the average length of desaturation

episodes of varying degrees. Finally, investigators rarely

if ever calculate either the average saturation change which

accompanies apneas and hypopneas or the average duration of

the saturation changes accompanying apneas and hypopneas.

Clearly, the only way to determine if these and other

approaches to the quantification of sleep-disordered

breathing are useful or enlightening is to select a

homogeneous sample of subjects and examine the

intercorrelations among these independent variables and the

correlations between those quantifications and other

variables of interest. The results of such a study might

help future researchers interested in the same population to

better allocate the expenditure of limited measurement

resources. The present study measured apneas, hypopneas,

and desaturations, and derived and analyzed many of the

possible ways of summarizing those events.

Statement of the Problem

Sleep Apnea Syndrome is a clinical entity with

demonstrated serious consequences. The sleep apneas and

related arterial oxygen desaturations frequently observed in

SAS patients are thought to play a causal role in the

development of associated deficits. Research indicating

that sleep apneas and desaturations occur with some

frequency in nonclinical samples suggests the hypothesis

that specific nonclinical populations at high risk for

nocturnal events may suffer from deficits similar to those

documented in clinical samples. Convincing evidence

supporting this hypothesis in male middle-aged snorers now

exists. A review of extant research findings indicates that

elderly males comprise a population that is perhaps more at

risk for the occurrence of nocturnal respiratory events than

any other non-medical population. At present, almost no

information is available concerning the concomitants of

those events in elderly subjects.

The present study attempted to gather such

information. Older males without acute illness were

recruited and a comprehensive analysis of their respiratory,

oxygen saturation, neuropsychological, sleep/wake, and

health status was carried out. Potential relationships

between respiratory disturbances and variables known to be

disrupted in SAS patients were thus examined in this

nonclinical sample.


Older males were recruited through newspaper

advertisements, posted notices, and phone calls to aging

subjects studied previously in the laboratory of W.B. Webb.

Participants were required to be male, at least 60 years of

age, literate, and in general good health (i.e., self-

described as healthy and not under the active care of a

physician for acute illness). Potential subjects with a

history of head trauma, neurologic or pulmonary disease,

alcoholism, or sleep disorder (specifically, narcolepsy or

sleep-related nocturnal myoclonus), and those taking

medications known to affect sleep (i.e., major and minor

tranquilizers, anxiolytics, and anti-depressants), were

excluded. Potential subjects who met these criteria were

offered an opportunity to participate in a 'sleep study' in

which they would complete testing during an evening, sleep

overnight in a laboratory with several physiological

parameters recorded, and complete a 2 week sleep diary at

home, for a payment of $50. Volunteers were scheduled on a

first-come, first-serve basis. During recruitment, 19

subjects (chosen on the basis of practical considerations

such as laboratory and technician availability and with an

eye toward obtaining a wide sub-sample age range) were asked

if they would be willing to return to the laboratory for a

second night of physiological recording. Fourteen of these

subjects agreed and were recorded on two consecutive

nights. A total of 40 subjects were studied; one subject

who was unable to sleep in the laboratory was dropped.

Table 2-1 lists the demographic characteristics of the

subjects. The final sample of 39 males had a mean age of

65.8 years and was drawn from a wide cross section of the

population, ranging from the retired to the unemployed to

university professors. All subjects read and signed

informed consent forms and were debriefed at the conclusion

of their participation.

Table 2-1. Demographic variables from 39 older males.


Mean SD Range
Age 65.82 4.9 60-83
Education (years) 14.59 2.7 9-20
Weight (lbs.) 178.89 30.3 121-267
Weight:Height Ratio (lbs.:in.) 2.59 0.4 1.7-3.7


During the evening testing all subjects received a

blood pressure check, made from a seated position, with a

standard hospital cuff. Overnight recordings were made in a

quiet, darkened room with the subject sleeping on a standard

hospital bed. Electrodes for electroencephalographic

recording were mounted in three pairs using sites F2/F8,

02/06, and RE/LE from the international 10/20 system. These

electrodes were affixed to the scalp with collodion-soaked

gauze dried with an air hose. Three Modified Bipolar EKG

leads were attached to the subjects' chests. These 4 EEG

and EKG channels were recorded on a Grass Polygraph model

79D at a chart speed of 15 mm/sec. Oral and nasal airflow

was measured by a combined thermister affixed to the upper

lip. Chest and abdominal movement was measured by impedance

pneumography through surface electrodes placed on the lower

chest and just above the navel. Arterial oxygen saturation

was monitored continuously from a non-invasive earclip

attached to the earlobe and analyzed by a Biox 3700

Pulse-Oximeter. These 4 respiratory channels were routed to

a strip chart polygraph (Desk Model Physiograph) and

recorded at a chart speed of .25 cm/sec.

Independent Variables

Three groups of independent variables were extracted

from the above-described physiological recordings.

Respiratory variables (those derived primarily from the

airflow thermister and chest/abdominal movement data)

included number of apneas and hypopneas. Apneas were scored

using the standard criterion of complete cessation of

airflow for at least 10 consecutive seconds. Hypopneas were

scored using the dual criteria of a reduction in airflow of

at least 50% (but less than 100%) and a reduction in blood

oxygen saturation of at least 7%. The latter criterion was

added when an informal inter-rater reliability check

suggested that it was not possible to completely and

objectively operationalize guidelines for making a judgment

of 50% airflow reduction. The variables derived from counts

of numbers of apneas and hypopneas were Apnea Index

(#Apneas/Sleep Period Time), Hypopnea Index

(#Hypopneas/Sleep Period Time) and Apnea plus Hypopnea Index
(#Apneas + #Hypopneas/Sleep Period Time). Oxygen saturation

variables (those derived primarily from the continuous

recording of arterial oxygen saturation) included Mean

Lowest Saturation, Number of Desaturations >4%, >7%, and

10%, and Duration of Desaturations >4%, and >10%. Mean

Lowest Saturation was derived from minute by minute noting

of the lowest saturation attained. Those variables

reflecting number and duration of desaturations of varying

degrees were adjusted for Sleep Period Time. Desaturations

were scored regardless of whether or not they were

associated with a reduction in the amplitude of breathing.

The third group of variables is made up of measures which

take into account both respiratory and oxygen saturation

information. These include Mean Saturation Change in Apneas

and Mean Saturation Change in Hypopneas. The above

variables provide complete coverage of all potentially

important aspects of nocturnal respiratory activity: number

of both types of respiratory events, chronic amount of

variability in oxygen saturation, number of desaturations of

varying degrees and their duration, and separate measures of

the degree of desaturation accompanying apneas and


In an effort to identify any redundancy of measurement

within the independent variables (the indices of nocturnal

respiratory activity), a correlation matrix containing the

10 variables reflecting apnea activity, hypopnea activity,

desaturation associated with apnea and hypopnea activity,

and time spent in, and number of, desaturations of varying

severity was calculated. Table 2-2 lists the

inter-correlations of these independent variables. An

examination of this table suggests that two variables, Apnea

Index and Mean Saturation Change in Apneas/Hypopneas, are

not well predicted by the other variables, manifesting

average inter-correlations of only .54 and .60,

respectively. The remaining 8 variables are highly

inter-correlated. The redundancy of measurement in this

subgroup of variables is apparent from the high average

correlations within the group (ranging from .82 to .94).

Mean Low Saturation stands somewhat apart from this subgroup

in that the average correlation of this variable with the

other 7 (.82) is relatively low. In the correlational

analyses reported in the following chapter, these 3

variables (Mean Low Saturation, Apnea Index, and Mean

Saturation Change in Apneas/Hypopneas) will be retained. In

addition, Number of Desaturations >4% will be retained as

representative of the 7 remaining highly inter-correlated

desaturation measures. All other variables will be ignored.

Table 2-2. Correlations among the 10 measures of nocturnal
respiratory activity.

AI HI AHI LSat MSatAH D4 D7 D10 Sec4 SeclO

HI .33
AHI .65 .93
LSat -.47 -.79 -.82
MSatAH .52 .45 .56 -.55
D4 .50 .97 .97 -.83 .51
D7 .52 .97 .98 -.84 .56 .98
D10 .61 .87 .94 -.81 .58 .90 .96
Sec4 .60 .94 .99 -.85 .59 .98 .98 .92
SeclO .67 .81 .91 -.78 .62 .84 .92 .98 .88

Corr. .54 .72 .81 -.71 .60 .77 .80 .79 .81 .78

Mean Corr.
in redundant
group. .90 .93 .82 .92 .94 .91 .93 .87

Legend: AI=Apnea Index; HI=Hypopnea Index; AHI=Apnea +
Hypopnea Index; LSat=Mean Low Saturation; MSatAH=Mean
Saturation Change in Apneas/Hypopneas; D4=# of
desaturations >4%; D7=# of desaturations >7%; D10=# of
desaturations >10%; Sec4=Seconds in desaturations >4%;
SeclO=Seconds in desaturations >10%.

Many of the findings to be reported in the next chapter

would be more difficult to interpret if the independent

variables of interest (nocturnal respiratory activity) were

shown to be phenomena with little or limited night-to-night

reliability. Such a finding would call into question the as

yet prevalent practice of recruiting subjects for only one

night of sleep in studies of the relationship between

apnea/hypopnea activity and other variables of interest. It

is now common practice, in studies of EEG sleep, to obtain

several consecutive nights of sleep data from subjects and

then discard the first night's data as unrepresentative.

This became common practice after it was shown that there

are significant and predictable differences between the

sleep obtained by subjects on the first night in a sleep

laboratory and the sleep obtained on subsequent nights.

This "first night effect" is characterized by longer sleep

latencies, greater numbers of awakenings, and less time

spent in REM sleep compared to other nights. In essence it

will be the aim of the following analyses to determine if a

similar first night effect exists for nocturnal respiratory

activity. Fourteen of the 39 subjects slept for two

consecutive nights in the laboratory; all physiological

measurements (respiratory and EEG) were repeated on the

second night. Correlations and T-tests between the same

respiratory variables measured on the two nights will be

reported, and these will be compared to the findings of

similar analyses conducted with the EEG variables, based on

the reasoning that if the reliability of respiratory

variables does not substantially exceed that of the EEG

variables, a first night effect will have been found.

Table 2-3 lists the Night 1 means and standard

deviations for demographic, nocturnal respiratory, and sleep

variables from the subsample of subjects who were recorded

twice. Also listed for purposes of comparison are the full

sample means for the same variables. Although the subsample

is clearly representative of the full sample across all

demographic and EEG variables, there is evidence indicating

that the subsample manifested somewhat higher numbers of

apneas, hypopneas, and desaturations. This mild bias is

fortuitous because most of the studies for which

night-to-night reliability would be an issue make active

attempts to recruit subjects who manifest higher than normal

levels of respiratory activity (i.e., by selecting subjects

on the basis of weight, snoring, complaints of daytime

sleepiness, age, etc.).

Table 2-4 lists the Night 1 and Night 2 means from the

14 subjects for all respiratory and sleep variables. An

examination of the EEG data suggests that a typical first

night effect was found: these subjects took longer to fall

asleep, spent more time awake and in light sleep after

falling asleep, and spent less time in REM and slow wave

sleep on the first night relative to the second. Paired

T-tests between Night 1 and Night 2 data were significant

for 6 of the 14 EEG variables: Pure Sleep Time, Sleep

Efficiency, Sleep Latency, Time % Stage 0, Time % Stage 1,

and Time % Stage REM. In contrast, although an examination

of the two night means for the respiratory variables

suggests a moderate tendency for numbers of respiratory

events and desaturations to increase from the first to the

second night, no paired T-test comparison was significant

for the 10 variables, indicating that these variables are

substantially more stable across the first two nights in a

laboratory than are EEG variables.

Table 2-3.

Means and standard deviations of demographic,
first night respiratory, and first night EEG
variables from full sample of 39 older males
and subsample of 14 subjects who spent two
nights in laboratory.



Demographic Variables

Height:Weight Ratio

Respiratory Variables

Apnea Index
Hypopnea Index
Ap. + Hypop. Index
Mean Low Saturation
Mean Sat. Change: Ap/Hyp
Number of Desats. >4%
Number of Desats. >7%
Number of Desats. >10%
Seconds in Desats. >4%
Seconds in Desats. >10%





Sleep Variables

Pure Sleep Time
Sleep Efficiency Index
Sleep Latency
Number Stage 0 Periods
Time % Stage 0
Time % Stage 1
Time % Stage 2
Time % Stage 3
Time % Stage 4
Time % Stage REM
Time % Slow Wave Sleep
Latency 1st REM Period
Number of REM Periods
Mean REM Period Length



Table 2-4. Respiratory and sleep variable means for Night 1
and Night 2 in 14 older males.


Night 1 Night 2
Respiratory Variables

Apnea Index 3.59 4.17
Hypopnea Index 3.01 3.48
Apnea + Hypopnea Index 6.61 7.65
Mean Low Saturation 92.70 92.19
Mean Sat. Change: Ap/Hyp 6.09 6.43
Number of Desats. >4% 11.17 11.68
Number of Desats. >7% 4.81 5.86
Number of Desats. >10% 2.48 4.23
Seconds in Desats. >4% 330.47 354.26
Seconds in Desats. >10% 90.70 154.87

Sleep Variables

Pure Sleep Time 347.93 367.21*
Sleep Efficiency .849 .906 *
Sleep Latency 12.64 6.36*
Number Stage 0 Periods 1.33 1.35
Time % Stage 0 11.47 6.93*
Time % Stage 1 2.72 1.96*
Time % Stage 2 58.05 57.27
Time % Stage 3 2.01 2.41
Time % Stage 4 13.51 16.43
Time % Stage REM 12.23 14.98*
Time % Slow Wave Sleep 17.17 20.29
Latency 1st REM Period 157.00 129.21
Number of REM Periods 2.78 3.28
Mean REM Period Length 18,41 18.09

*=Significant T-test.

Table 2-5 lists the results of Pearson correlations

between Night 1 and Night 2 respiratory and sleep

variables. As might be expected, all of the significant

correlations are positive. There is a striking

preponderance of significant correlations within the group

of respiratory variables: 10 of the 10 variables displayed

significant night-to-night reliability and 7 of these

correlations were strong enough to be significant at the

p<.005 level. In contrast, only 8 of the 14 correlations

between Night 1 and Night 2 EEG variables were significant,

and only one met the p<.005 criterion.

Finally, it can be seen from Table 2-5 that two of the

three respiratory variables found, through intercorrelation,

Table 2-5. Results of Pearson correlations between Night 1
and Night 2 respiratory and sleep variables in
14 older males.

Pearson r
Respiratory Variables

Apnea Index .607
Hypopnea Index .755*
Apnea + Hypopnea Index .655
Mean Low Saturation .625
Mean Sat. Change: Ap/Hyp .886*
Number of Desats. >4% .787*
Number of Desats. >7% .759*
Number of Desats. >10% .862*
Seconds in Desats. >4% .765*
Seconds in Desats. >10% .840*

Sleep Variables

Pure Sleep Time .519
Sleep Efficiency .578
Sleep Latency .685*
Number Stage 0 Periods Not Significant
Time % Stage 0 .572
Time % Stage 1 Not Significant
Time % Stage 2 .637
Time % Stage 3 Not Significant
Time % Stage 4 Not Significant
Time % Stage REM .526
Time % Slow Wave Sleep Not Significant
Latency 1st REM Period Not Significant
Number of REM Periods .639
Mean REM Period Length .501


to be relatively independent of the remaining 7 variables

(Apnea Index, Mean Low Saturation, Mean Saturation Change in

Apneas/Hypopneas; see Table 2-2) display a relatively low

night-to-night reliability. Apnea Index and Mean Low

Saturation are the only variables for which reliability did

not reach .650. This suggests that their relative

independence from other respiratory measures is at least

partly attributable to measurement error. Whether this

error is systematic (derived, perhaps, from these variables

being more sensitive than the others to the effects of

sleeping in a laboratory for the first time), or whether

apnea activity and chronic fluctuation in oxygen saturation

are simply more variable across nights, cannot be determined

from the design of the present study. Measurement conducted

across several consecutive nights of sleep would be required

to settle this issue. One predictable result of this

relatively low reliability/high measurement error will be a

suppression of the correlations between Apnea Index and Mean

Low Saturation and any dependent variables of interest.

That is, it can be predicted from these reliability data

that it will be more difficult to find significant

correlations between these two variables and the

demographic, health, sleep/wake, and neuropsychological

dependent variables than it will be for the other

respiratory variables.

In summary, T-test and correlation analyses indicate

that nocturnal respiratory activity, measured in a variety

of ways, remains stable from the first to the second night

in the laboratory. Informal comparison of the reliabilities

of respiratory and EEG variables suggests that no "first

night effect" of the magnitude observed by this and many

other studies for sleep variables exists for indices of

nocturnal respiratory activity. These findings provide

statistical justification for the single-night design of the

present study.

Dependent Variables

The dependent measures fall into three broad classes:

measures of health status, measures of sleep/wake status,

and measures of neuropsychological status. The measures of

health status include height, weight, systolic and diastolic

blood pressure, and the Cornell Medical Index (CMI). The

CMI (Broadman, et al., 1949) is a wide-ranging, face-valid

symptom checklist providing information about past and

present physical difficulties. It yields an overall score

indicating total number of symptoms endorsed, along with

subscale scores for several symptom categories

(neurological, pulmonary, and cardiological). These four

scores were used in the analyses described below. In

addition, subjects' responses to a specific question

concerning diagnosis of hypertension were noted and used.

Together, the CMI, blood pressure, height, and weight

measures provided a broad health screening as well as

detailed information concerning health deficits possibly

related to apnea/hypopnea activity.

The measures of sleep/wake status include the 2-week

Sleep Log, a Sleep Questionnaire, the mean of several

Stanford Sleepiness Scale ratings, and overnight EEG. The

Sleep Log is a self-monitoring device used extensively in

the laboratory of W.B. Webb. Subjects completed the daily

Sleep Log during the 2 week period immediately following

their visit to the laboratory; return rate was 100%.

Variables derived from the Logs include: mean Total Bed

Time, mean estimated Sleep Latency, mean estimated number

and duration of nocturnal Awakenings, mean number of reports

of Daytime Sleepiness, and mean number and duration of Naps.

The Sleep Questionnaire is a largely multiple choice

measure which asks for general estimates of a number of

sleep-related parameters. Variables derived from the Sleep

Questionnaire include: typical Total Bed Time, typical Sleep

Latency, typical number and duration of nocturnal

Awakenings, typical frequency of feelings of Daytime

Sleepiness, and typical number and duration of Naps. The

Stanford Sleepiness Scale (SSS) is a frequently-used measure

of subjective sleepiness (Hoddes, et al., 1973). It is a 7

point Likert-type scale with each point anchored by

statements reflecting increasing degrees of sleepiness

(e.g., 1-feel active and vital. ... 7-almost in reverie,

sleep onset soon.). The SSS was administered at 1/2 hour

intervals beginning with arrival at the laboratory and

ending at bedtime. A mean sleepiness rating was derived

from these repeated administrations.

The overnight EEG variables were derived from

minute-by-minute scoring of the electroencephalographic

recordings, using the system of Agnew and Webb (1972) with

modification for the scoring of slow wave sleep in older

persons (Webb and Dreblow, 1982). A 5-minute combining rule

for REM periods was used. Scoring was conducted by a

trained and experienced sleep technician. EEG variables

include: Sleep Period Time, Pure Sleep Time, Sleep

Efficiency Index, Sleep Latency, Number of Stage 0 periods

lasting 3 minutes or longer, time % stages 0, 1, 2, 3, 4,

and REM, time % Slow Wave Sleep, Latency to 1st REM Period,

Number of REM Periods, and Mean REM Period Length.

The measures of neuropsychological status were chosen

in an attempt to sample from each of several loosely bounded

areas of cognitive skill. These include both verbal and

nonverbal intelligence (WAIS-R FSIQ, VIQ, and PIQ),

immediate verbal memory (age-corrected Digit Span scale

score; Wechsler Memory Scale Paired Associates; and a

subscale of the California Verbal Learning Test), delayed

verbal memory (30 min. delayed recall of Wechsler Memory

Scale Logical Stories; a subscale of the California Verbal

Learning Test), remote memory (Memory for Historical

Events), immediate nonverbal memory (Spatial Recognition

Test), delayed nonverbal memory (30 min. delayed recall of

the Wechsler Memory Scale Figures; 30 min. delayed recall of

the Rey-Osterrieth Complex Figure), visuo-perceptual/

organizational skills (Hooper Visual Organization Test;

age-corrected Block Design scale score), language ability

(Controlled Word Association Test), frontal/self-regulatory

skills (Wisconsin Card Sord), and motor functioning (left

and right Finger Tapping). Appropriate scores were derived

from each of these tests and together they provide both a

broad screening of cognitive functioning and a detailed

assessment of skills thought to be particularly sensitive to

hypoxemia (memory, visual-organizational skills, abstract

reasoning, and motor coordination.


The following schedule was maintained on the experimental


6:00 Sign 2 Informed Consent forms
Stanford Sleepiness Scale
Height and Weight measurement
Blood Pressure measurement
Wechsler Memory Scale (these subtests only: immed.
recall of Stories and Drawings; Paired Associates)

6:30 Stanford Sleepiness Scale (interrupt WMS if needed)
WAIS-R (Satz-Mogel Short Form)

6:50 Wechsler Memory Scale (interrupt WAIS-R; 30 min.
delayed recall of stories and drawings)

7:00 Stanford Sleepiness Scale (interrupt WAIS-R)

7:30 Stanford Sleepiness Scale (interrupt WAIS-R if
Rey-Osterreith Figure (Copy and Immediate Recall)
Cornell Medical Index
Sleep Questionnaire
Verbal Fluency

8:00 Stanford Sleepiness Scale
California Verbal Learning Test
Rey-Osterreith Figure (30 min. Delayed Recall)

BREAK (approx. 15 mins.)

8:30 Stanford Sleepiness Scale
California Verbal Learning Test (30 min. del.
Wisconsin Card Sort

9:00 Stanford Sleepiness Scale (interrupt WCS if needed)
Remote Memory Questionnaire

9:30 Stanford Sleepiness Scale (interrupt RMQ if needed)
Spatial Recognition Test
Finger Tapping (10 secs.; first right, then left
hand 3 times)
10:00 Stanford Sleepiness Scale
Begin wiring for EEG, Thermister, Resp., Oximeter

11:00 Lights out, begin recording.

6:00 am to 7:00 am wakeup

A sleep technician remained with the subjects throughout the

night and the subjects were allowed to determine their own

wakeup times.

Statistical Procedures
The primary statistical method utilized in evaluating

the present data was correlational. This was considered to

be appropriate because of the exploratory nature of the

study and because the independent variables of interest do

not lend themselves well to experimental manipulation.

Other approaches were used in an effort to add confidence,

including: 1) partial correlation to separate out variance

attibutable to third factors; 2) stratification of the

sample by level of desaturation activity and the use of

T-tests to evaluate between-group differences; 3) the

elimination of redundant independent variables.

Non-parametric procedures were utilized where appropriate.

The use of multivariate statistics was not feasible due to

the limited subjects:variables ratio. Although it must be

conceded that a correlational approach carries the burden of

possibly producing spurious findings, it is believed that

the various additional procedures used lend confidence to

the results.


Incidence of Apnea and Desaturation Activity

Subjects were selected to fulfill several sampling

requirements, including male sex, age of at least 60 years,

and self report of general good health. Potential subjects

were excluded if they reported a history of head trauma,

neurologic or pulmonary disease, or alcoholism. Subjects

under the active care of a physician for acute illness and

those taking medications known to affect sleep (i.e., major

and minor tranquilizers, anxiolytics, and anti-depressants)

were also excluded from participation. A total of 40

subjects were studied; the data of one subject who was

unable to sleep in the laboratory were not included in the

analyses reported below. This group of older males had a

mean age of 66 years, a mean education level of 14 1/2

years, and a mean weight of approximately 179 pounds.

The respiratory records were divided into 90 minute

intervals beginning with the onset of sleep. From a total

of 1061 respiratory events scored (hypopneas accompanied by

an oxygen desaturation >7%, plus apneas), 250 (24%) occurred

during the first 90 minutes, 207 (19%) occurred during the

second 90 minutes, 245 (23%) occurred during the third 90

minutes, 299 (28%) occurred during the fourth 90 minutes,

and 60 (6%) occurred during the fifth period. The

attenuated number of events beyond the 6th hour of sleep is

attributable to the fact that only a small proportion of

subjects slept longer than 6 1/2 hours.

Table 3-1 provides a percentage breakdown of

respiratory events according to the stages of sleep in which

they occurred, along with the relative distribution of time

spent in each stage by these subjects. Most events occurred

during stage 2 sleep, with the bulk of the remainder being

fairly evenly divided between stages 4 and REM. This

distribution closely mirrors the proportion of time spent in

the various stages and a Goodness of Fit Chi-square analysis

of these proportions was not significant (X2=3.67; p=NS).

Table 3-1. Proportions of apnea/hypopnea events in sleep
stages compared with distribution of sleep time
among stages.


Sleep Stage Events % of Sleep Time
Stage 1 2.83% 3.50%
Stage 2 78.42% 64.34%
Stage 3 1.98% 2.20%
Stage 4 7.07% 15.59%
REM 9.70% 14.38%

Table 3-2 presents a frequency distribution for Apnea +

Hypopnea Index (AHI) scores. Seventy-seven percent of the

subjects experienced at least one episode of apnea or

hypopnea and 20.5% achieved an Apnea + Hypopnea Index (AHI)

score >5. It is apparent from this table, however, that AHI

scores are distributed in a highly skewed manner; roughly

one-quarter of the subjects experienced no apneas or

hypopneas and the great majority experienced very few (less

than 3 or 4) such episodes per hour.

Table 3-2. Distribution of number of subjects with varying
amounts of apnea + hypopnea activity.

AHI # Subs. % Cumul.% AHI # Subs. i Cumul.%

0 9 23 23 9-10 0 0 82
0-1 9 23 46 10-11 1 2.5 84.5
1-2 5 13 59 11-12 0 0 84.5
2-3 4 10 69 12-13 0 0 84.5
3-4 3 8 77 13-14 2 5 89.5
4-5 1 2.5 79.5 14-15 1 2.5 92
5-6 0 0 79.5 15-22 1 2.5 94.5
6-7 0 0 79.5 22-29 0 0 94.5
7-8 1 2.5 82 29-36 1 2.5 97
8-9 0 0 82 36-43 1 2.5 100

In an effort to determine whether increasing age is

related to an increasing incidence of apnea activity,

subjects were divided into groups falling above and below

the mean sample age (65.8 years) and into 2 groups formed on

the basis of AHI scores (AHI less than 5; AHI greater than

or equal to 5). Table 3-3 presents the results of this

partitioning. An examination of this table suggests that

there is little to no relationship between age and incidence

of apnea/hypopnea activity within this sample of exclusively

older males and a Goodness of Fit Chi-square analysis of

these proportions was not significant (X2=.17; p=NS).

Similarly, a T-test comparison of the mean age of subjects

Table 3-3. Incidence of different levels of Apnea +
Hypopnea Index scores in 39 older males grouped
by age.

Age Group N AHI<5 AHI>5

Below Mean Age 22 82% 18%
Above Mean Age 17 76% 23%
Overall 39 80% 20%

Mean AHI(SD) 1.2(1.3) 18.7(10.6)

with AHI scores above and below 5 revealed no significant

difference (T=.125; p=NS).

Table 3-4 presents a frequency distribution for numbers

of subjects who experienced various amounts of desaturations

>4% per hour. This too is a highly skewed distribution,

indicating that most subjects experienced very few (less

than 5 or 6) such desaturations per hour of sleep, while

22.5% of the subjects experienced 9 or more desaturations

>4% per hour. The 9 subjects with 9 or more desaturations

4%/hour averaged 31 episodes per hour; the remaining

subjects averaged 2.2 episodes per hour.

Table 3-4. Distribution of number of subjects with varying
amounts of desaturations >4% per hour.

Desats. # Subs. % Cumul.% Desats. # Subs. I Cumul.%

0 3 8 8 7-8 1 2.5 75
0-1 6 15.5 23.5 8-9 1 2.5 77.5
1-2 10 26 49.5 9-10 2 5 82.5
2-3 4 10 59.5 10-17 1 2.5 85
3-4 2 5 64.5 17-24 2 5 90
4-5 0 0 64.5 24-31 1 2.5 92.5
5-6 3 8 72.5 31-38 1 2.5 95
6-7 0 0 72.5 38-85 2 5 100

A Pearson correlation matrix relating the demographic
variables with the 4 representative nocturnal respiratory

indices revealed that while age and education were unrelated

to apnea/hypopnea or desaturation activity, weight:height

ratio was significantly related to 3 of the 4 indices,

indicating that increasing weight predicts increasing

desaturation activity.

Table 3-5 presents the results of T-test analyses

comparing the above-described 2 groups of subjects (that is,

groups formed according to level of desaturation activity:

less than 9 desaturations >4% per hour, or more than 9 such

desaturations per hour) on demographic and nocturnal

respiratory variables. The cutting score used in the

formation of these groups, although somewhat arbitrary, was

deemed appropriate because it produced proportions in this

sample very similar to those produced by the

traditionally-used criterion for clinical significance of

AHIZ5, and because it produced a High group with an N large

enough for the valid use of small-group T-tests. While

there were no significant between-group differences for age

and education [T(37)=.05, .04; p=NS], the T-test for

weight:height ratio was significant [T(37)=3.5; p=.000],

indicating that the High desaturation group had a

significantly higher mean weight:height ratio than the Low

desaturation group. As would be expected from the method

used to form the groups, all indices reflecting number of

desaturations (number of Desaturations >4% and 10%),

severity of desaturations (mean Saturation Change in Apneas

and Hypopneas), duration of desaturations (total seconds

spent in Desaturations >4% and >10%), and average nadir of

saturation (Mean Low Saturation; recall that this measure

was obtained by noting the lowest saturation in each minute

of sleep and then averaging across the night), were

significantly different between the two groups [T(37)=6.59,

6.58, 5.56, 4.39, 8.69, 7.07, 6.03; all p=.000]. In

addition, the between-group differences for Apnea Index and

Hypopnea Index were also significant [T(37)=5.93, 5.19;

p=.000]. For each of the above respiratory indices the High

desaturation group demonstrated a significantly more

disordered nocturnal respiratory process than the Low group.

Table 3-5. Means and standard deviations for selected demo-
graphic and nocturnal respiratory variables in
39 older males grouped by level of desaturation

Low Desaturation High Desaturation*
Group (n=30) Group (n=9)

Age 65.8(4.9) 65.9(4.8)
Education 14.6(2.9) 14.5(2.1)
Weight:Height Ratio 2.5(0.3) 2.9(0.5)*
Apnea Index 0.9(1.1) 6.2(4.6)*
Hypopnea Index 0.2(0.2) 10.7(11.5)*
Mean Low Saturation 93.8(1.1) 90.3(2.4)*
Mean Sat. Change: Apn. 2.7(3.0) 9.0(2.9)*
Mean Sat. Change: Hyp. 3.6(4.0) 9.6(1.3)*
Number of Desats. >4% 2.2(2.2) 31.0(24.4)*
Number of Desats. >10% 0.1(0.4) 6.9(5.7)*
Seconds in Desats. >4% 63.9(58.8) 794.4(462.2)*
Seconds in Desats. >10% 5.5(15.6) 225.3(173.3)*

*=Significant T-test.

In summary, the present sample of 39 males with an

average age of 66 years, an average education level of 14

1/2 years, and an average weight of 179 lbs. exhibited a

wide range of apnea/hypopnea and desaturation activity. A

total of 1061 respiratory events were scored, with 77% of

the subjects experiencing at least one apnea or hypopnea and

20% exhibiting a level of apnea/hypopnea activity

traditionally considered to be clinically significant.

Similarly, 82% of the subjects experienced at least one

desaturation >4% and 22% experienced an average of more than

9 desaturations per hour of sleep. Respiratory events were

evenly distributed across the course of the night and they

occurred most frequently during Stage 2 sleep.

Non-parametric, T-test, and correlational analyses failed to

reveal a significant relationship between age and

apnea/hypopnea or desaturation activity within this

exclusively older sample, while correlational and T-test

analyses found weight:height ratio to be significantly and

positively related to respiratory activity. When the sample

was stratified by level of desaturation activity, subjects

with a high number of desaturations had significantly more

respiratory events and significantly more severe and

long-lasting desaturations than the remaining subjects.

Nocturnal Respiratory and Health Related Variables

Health variables included seated blood pressure

readings and several scores derived from the Cornell Medical

Index (overall number of symptoms endorsed, as well as

number of symptoms endorsed from the respiratory,

cardiological, and neurological subscales). Twenty-four

Pearson correlations relating these variables to the 4

representative respiratory indices described above resulted

in 5 significant correlations (a number greater than would

be expected by chance alone). Diastolic blood pressure and

number of respiratory and cardiological symptoms endorsed on

the Cornell Medical Index (CMI) were unrelated to any of the

indices. Table 3-6 lists the remaining variables and the

significant correlations found. Systolic blood pressure was

significantly positively related to number of desaturations,

and was negatively related to Mean Low Saturation. The

number of neurological symptoms endorsed on the CMI was

positively related to indices reflecting both number of

apneas and number of desaturations. Because both age and

weight could serve as possible third-variable explanations

for the observed significant relationships between Systolic

Table 3-6. Significant (p<.05) Pearson correlations between
nocturnal respiratory variables and health
variables in 39 older males.

BP Sysa CMI Ovb CMI Neurc
Apnea Index .32
Mean Low Saturation -.61*
Mean Sat. Change: Apneas
and Hypopneas
Number of Desats. >4% .59* .28 .29

aBP Sys=Systolic blood pressure.
CMI Ov=Cornell Medical Index overall score.
cCMI Neur=Cornell Medical Index neurological score.