REGULAR ARTICLES

Predicting Mortality in Mixed

Depression and Dementia
Using EEG Sleep Variables
Carolyn C. Hoch, Ph.D.
Charles F. Reynolds III, M.D.
Patricia R. Houck, M.S.H.
Florence Hall, R.N.
Susan R. Berman, B.A.
Daniel J. Buysse, M.D.
Ronald E. Dahi, M.D.
David J. Kupfer, M.D.

The authors report a study of elect roencephalo- E lderly patients with mixed symptoms of depression
graphic (EEG) sleep predictors of two-year mortal- and cognitive impairment represent a fragile and
heterogeneous patient group for whom it is quite difficult
ity in 26 elderly patients with mixed symptoms of
to provide differential diagnosis, acute treatment, and
depression and cognitive impairment. Patients long-term management. To address the heterogeneous
who had died by two-year follow-up were charac- nature of the phenomenology, etiology, and treatment
terized by significantly longer rapid eye move- response shown by these patients, we have focused our
ment (REM) sleep latencies at baseline, less robust clinical and experimental studies on the differentiation of
REM sleep rebound following all-night sleep depri- patients having primary degenerative dementia with
secondary depression from those with depressive pseu-
vation, and baseline apnea-hypopnea indexes
dodementia.
greater than 3. Logistic regression analysis using We have found that symptom profile, treatment re-
the apnea-hypopnea index value and REM latency sponse, and electroencephalographic (EEC) measures of
correctly predicted 77% of survivors and non- sleep all help to differentiate between these two groups
survivors. Survival time following initial measure- of mixed-symptom patients.13 In the course of perform-
ments was significantly correlated with REM ing two-year follow-up of disease course and diagnosis,
we observed a 35% mortality rate (nine of 26 patients
sleep time (rO.78, p<.O2) and duration of first
died) among mixed-symptom patients, representing a
REM sleep period (r=O.75, p’z.O2). The authors standardized mortality ratio of 450%. According to De-
speculate that changes in these predictor variables partment of Health and Human Services figures,4 the
may indicate impairment in the cholinergic con- expected number of deaths, adjusted for age, sex, and
trol of cognitive function, REM sleep, and respira- race, over two years would have been two. We then asked
tory function. if EEC sleep measures, which can be used to predict
(The Journal of Neuropsychiatry and Clinical pseudodementia,3 are also reliable predictors of two-year
Neurosciences 1989; 1:366-371) mortality.
To answer this question, we retrospectively studied the
baseline clinical and EEC sleep measures in a sample of

Received February 28, 1989; revised June 14, 1989; accepted June 21,
1989. From the Western Psychiatric Institute and Clinic, Pittsburgh,
Pennsylvania. Address reprint requests to Dr. Reynolds, Western Psy-
chiatric Institute and Clinic, 3811 O’Hara Street, Pittsburgh, PA 15213.
Copyright © 1989 American Psychiatric Press, Inc.

366 VOLUME I NUMBER

#{149} FALL
4 #{149} 1989
 HOCH eta!.

26 mixed-symptom patients. Nine deaths occurred in the able to care for himself or herself, whether appointment
sample over a two-year period; the mean±SD time be- of a legal guardian had become necessary, whether the
tween sleep studies and death was 13.2±12.6 months patient had required transfer from home to a nursing care
(median, 7 months). A priori cut-off points for specific facility, whether any new medical conditions had devel-
predictor variables, such as the apnea-hypopnea index or oped, or whether the patient had died.
REM latency, were not imposed, thereby allowing full
exploration of distributions in relation to outcome mea- Data Analysis
sures. Demographic and clinical data for survivors and non-
survivors were compared with group t-tests, the Mann-
Whitney U statistic, and Fisher’s exact test for probability.
METHODS Individual items from the Hamilton Depression Scale
and the Blessed Dementia Rating Scale were examined
Subjects with group t-tests. The proportion of the groups that
The subjects were 19 inpatients and seven outpatients endorsed each item was tested with Fisher’s exact prob-
who had volunteered to participate in our ongoing stud- ability.
ies of sleep, aging, and neuropsychiatric disorders. De- Baseline sleep variables (means from nights 2 and 3)
tails of recruitment and inclusion criteria have been were compared with group t-tests. Apnea-hypopnea
previously reported.3’56 Briefly, patients who had mixed measures were compared between groups with the
symptoms of depression and cognitive impairment were Mann-Whitney U-test. The apnea-hypopnea index distri-
required to meet either the Schedule for Affective Disor- bution was tested with Fisher’s exact test for probability.
ders and Schizophrenia-Lifetime Version (SADS-L)7 cri- To assess the risk of death, four biological variables were
teria or Research Diagnostic Criteria (RDC)8 for major selected on the basis of univaniate contrasts and entered
depressive disorder and Diagnostic and Statistical Manual into a backward-stepping logistic regression model using
of Mental Disorders, Third Edition (DSM-III)9 criteria for the SAS LOGIST procedure.15 Goodness of fit was evalu-
primary degenerative dementia with depressive features, ated with the chi-square likelihood ratio and by the accu-
or both sets of criteria independently. Patients were also racy of the variables’ abilities to predict survivors and
required to have Hamilton Depression Scale (HAM-DY#{176} nonsurvivors. The relative risk and 95% confidence inter-
scores of 13 or higher on the first 17 items of the scale and val were computed and tested with Fisher’s exact proba-
Mini-Mental State Exam (MMSE) scores of 26 or less. bility test (P). Following identification of potential risk
All patients received thorough medical and psychiatric factors, we determined the mortality rate by calculating
evaluations that included routine laboratory studies, the proportion of patients likely to have died if no risk
waking EEGs, and cranial computerized tomographic factors were considered, compared to using one risk fac-
(CT) scans. Patients were kept free of psychoactive med- tor and two risk factors. These proportions were tested
ication for 14 days prior to and during the six-night EEC using the chi-square statistic.
sleep protocol. Informed consent for the study was ob-
tained for each subject, either directly from the patient or,
for patients with severe cognitive impairment, from fam- RESULTS
ily members.
Patients participated in a sleep-study protocol that has Clinical Measures
been reported in detail elsewhere.5 The protocol spans six When the baseline sleep studies were performed, survi-
consecutive nights and includes three nights of baseline vors and nonsurvivors did not differ on measures of age,
sleep studies, one night of total sleep deprivation, and global severity of depression, or cognitive impairment.
two nights of recovery sleep. Methods of polysomnogra- However, nonsurvivors scored higher on the anergia
phy, including monitoring for sleep apnea, have been item of the HAM-D (t=2.23, p<.O5) and on the item mea-
reported in detail elsewhere.6’12 suring diminished emotional responsiveness on the
Following the sleep studies, follow-up ratings were Blessed Dementia Rating Scale (P=.05, Fisher’s exact test).
repeated every six months for two years. Assessment The final diagnosis a patient received did not predict
scales included the MMSE, the HAM-D, the Blessed De- mortality, and mortality was comparable in patients with
mentia Rating Scale (DRS),13 and the Clinical Dementia pseudodementia (43%) and patients with dementia (32%)
Rating scale (CDR).14 Ratings were performed by a re- (P=0.66, Fisher’s exact test).
search nurse and were based on interviews with patients, Twelve patients had concurrent systemic illnesses (hy-
family members, and primary caregivers. Additional in- pertension, cardiac disease, diabetes, thyroid disease)
formation was obtained about whether the patient was that required maintenance medications; the remaining 14

JOURNAL OF NEUROPSYCHIATRY 367

PREDICTING MORTALITY WITH EEC SLEEP VARIABLES

patients did not require any medications other than their tency and AHI and accurately predicted survival status
medications for depression or dementia. Seven survivors in 77% of patients (see table 3). When the relative mortal-
(41%) and five nonsurvivors (56%) had systemic ity risk was calculated with the method developed by
illnesses; this difference was not significant (P=.68, Katz et al.,16 we found that patients with a REM latency
Fisher’s exact test). Baseline demographic and clinical greater than 40 minutes and an AHI greater than or equal
data are summarized in table 1.
Information on causes of death was obtained from next TABLE 1. Demographic and dinical measures at baseline for 26
of kin. Four patients had died from pneumonia; three patients with mixed symptoms of depression and
dementiaa
from cardiac disease; one from an infectious disease; and
Survivors Nonsurvivors
one from medication toxicity. Autopsy results were
Measure (n=17) (n=9) t p
available for only three patients and verified a diagnosis
of Alzheimer’s disease. Demographic
Number hospitalized 14 5
Baseline electrocardiogram (ECG) data did not differ
Mean±SD age 71.2±5.7 74.3±9.0 1.08 NSC
between survivors and nonsurvivors, with 76% and 56%, Sex(M/F) 1/16 3/6
respectively, having ECGs within normal limits (P=0.38, Mean±SD years education 11.2±2.7 10.2±4.2 -0.70 NS
Clinical
Fisher’s exact test). Similarly, the two groups did not
Duration of illness (months)
differ on rates of EEC abnormalities: 56% of survivors mean±SD 24.8±18.739.9±35.1 1.44 MS
and 38% of nonsurvivors had normal EECs (P=.43, median 24 36 I.I9’ NS
Fisher’s exact test). However, the proportion of the range 5to72 ltoI2O
Blessed Dementia Rating
groups with abnormal CT scan results was significantly Scale (mean±SD score) 8.7±3.4 9.2±3.2 0.35 NS
different: 7% of survivors had abnormal CT scans, while Hamilton Depression
50% of nonsurvivors had moderate atrophy (P=.03, Scale (mean±SD score) 15.5±6.0 18.3±2.4 1.68 NS
Mini-Mental State
Fisher’s exact test). Exam (mean±SD score) 20.7±6.2 17.1±5.7 -1.45 NS
Modified Ischemic
EEC Sleep Measures Score of Hachinski
(mean±SD score) 1.4±0.8 1.1±0.8 -0.75 NS
Survivors and nonsurvivors did not differ on measures Clinical dementia
of baseline sleep continuity or NREM (non-REM) sleep rating (mean±SD score) 1.5±0.7 1.4±0.6 -0.51 NS
architecture. However, mean REM sleep latency was
apatients are grouped according to whether they survived through-
lower in survivors (16.8 minutes; range, 3.7 to 39.4) than
out a two-year follow-up period.
in nonsurvivors (44.1 minutes; range, 19.2 to 79.2) (t=2.78, bFih, exact test
CNS=not significant
p<.O2). The mean first REM sleep period was longer in
dZ approximation of the Mann-Whitney U-test
survivors (17.9±6.2) than in nonsurvivors (11.1±6.4)
(t=[-2.631, p<.O2). Furthermore, survivors tended to have
a greater percentage of REM sleep (21.6±8.7) than non- TABLE 2. Baseline EEG sleep variables of 26 patients with mixed
survivors (15.4±6.6) (t=1 .85, p<.O8). symptoms of depression and dementiaa

Survivors and nonsurvivors also differed on Apnea- Survivors Nonsurvivors

b
Hypopnea Index (AHI) measures. Six nonsurvivors had Variable (n=17) (n=9)

an AHI of 3 or greater, compared with three survivors REM latencyc (minutes)

(P=.02, Fisher’s exact test). The mean±SD AHI was average 16.8 44.1 .02
range 3.7 to 39.4 19.2 to 79.2
4.0±8.5 (range, 0 to 30.8; median, 0.92) for nonsurvivors
Percent REM sleep 21.6±8.7 15.4±6.6 .08
and 4.6±5.1 (range, 0 to 16.3; median, 4.2) for survivors; (mean±SD)
these values were not significantly different. Data on Duration of first REM period 17.9±6.2 11.1±6.4 .02
sleep variables and AHI are summarized in table 2. (mean±SD minutes)

Of all sleep-disordered breathing events in non- Apnea-Hypopnea Index

number of patients with 3 6 02d
survivors, 77% originated in the central nervous system,
a score 3
and 10% were caused by obstructions in the airway. In number of patients with 14 3
survivors, 50% of sleep-disordered events were central, a score <3

and 9% were obstructive. Because oxygen desaturation apatients were grouped according to whether they survived
measures were available for only six subjects, these data throughout a two-year follow-up period.
bGroup t-test
were not analyzed.
CStatistics were performed on the square root transformation of
When REM latency, percent of REM sleep, duration of
REM latency. The retransformed means are presented.
the first REM sleep period, and AHI were entered into a dFiher exact test

logistic regression analysis, the model retained REM la-

368 VOLUME I #{149}

NUMBER FALL
4 #{149} 1989
 HOCH et a!.

to 3 had a 3.7-fold increase in mortality risk (95% confi- depression, or degree of cognitive impairment. However,
dence interval, L,=1.19, L2=11.45; p<.O5). Goodness of fit nonsurvivors did score higher on measures of anergia
was checked by plotting the predicted proportion of non- and diminished emotional responsiveness. Systemic
survivors versus the true number of nonsurvivors. The illnesses requiring medication occurred at the same rate
-2 log likelihood ratio test value was 17.35 (p<.OOI), in survivors and nonsurvivors, but nonsurvivors had
which demonstrates an adequate joint association of vari-
ables in the model. The raw data on which the logistic FIGURE 1. Classification of survivors and nonsurvivors indicates
regression is based are shown in figure 1. Rates of mor- that having a REM latency longer than 40 minutes and
an Apnea-Hypopnea Index score greater than or equal
tality associated with zero, one, or two risk factors were to 3 was associated with a 3.7-fold increase in relative
0%, 55%, and 100%, respectively (2=6.82, p<.O5). risk of death by two-year follow-up (p<.O5). The mor-
We also examined whether sleep variables predicted tality rate was 100% with both risk factors, 55% with
one risk factor, and 0% with neither risk factor
the period of survival following baseline measurement.
(x’=6.82, p<.05).
The analysis indicated that time to mortality was signifi-
150

cantly correlated with baseline REM sleep time (rho=0.78,

p<.O2), percent REM sleep (rho=0.73, p.cz.O3), and dura-
tion of first REM sleep period (rho=0.75, p<.O2).
Survivors and nonsurvivors did not differ on measures
of sleep continuity and NREM sleep architecture on the
A o
two recovery nights after sleep deprivation. However,
mean±SD REM sleep time was higher on the second
recovery night in survivors (81.7±33.2 minutes) than in 3 0

50 0
nonsurvivors (55.2±28.9 minutes) (t=2.02, p<.O5) (figure A1

2). This reflects mean increases in REM sleep time of 41% Ab

in survivors and 21 % in nonsurvivors. When entered into Legend

A 00d
the logistic regression model, measures of REM sleep
I A AMI,.
rebound did not improve prediction of mortality. 0 5 30 ‘5 20 25 30 35
Apn.o-I4ypopn.a Index

DISCUSSION FIGURE 2. Scatterplot of REM sleep time in minutes on the sec-

ond recovery night after total sleep deprivation. Survi-
vors averaged 81.7±33.2 minutes total REM sleep;
Patients with mixed symptoms of depression and demen- nonsurvivors averaged 55.2±28.9 minutes (p<.O5).
tia who died during the follow-up period did not differ
significantly from survivors at baseline in age, severity of

leo
TABLES. Logistic regression ana1ysis and computation of risk of
mortality at two-year follow-up in 26 patients with
mixed symptoms of depression and dementia 160

Standard Standardized
Risk Factor1’ Coefficient Error Coefficient 140

Square root of 0.84 0.37 1.92

-
REM latency (SQRL) 320

Apnea-Hypopnea 3.95 1.71 2.10 0

Index (AHI)c 100

‘
Value of risk (px)
0. 0
S
.! Co
px=1+e (-6.77#{247}SQRL+AHI)

8
= [1+e- (-6.77+84+3.95)] 0
aThe model accurately predicted survival for 77% of patients: there o
were 14 predicted survivors, 17 actual survivors, three predicted non- 40 0

survivors, and nine actual nonsurvivors (95% CI: L,=I.19, L2=11.45;

p<.O5 [method of Katz et al.’6]). 0
bPatients with REM latency 40 minutes and an AHI3 had a 3.7- 20

0
fold increased risk of mortality at two-year follow-up (p<.OS).
AHI value: 0 for AHI3; I for AHI>3
Noneurvvors Survivors

JOURNAL OF NEUROPSYCHIATRY 369

PREDICTING MORTALITY WITH EEC SLEEP VARIABLES

significantly higher rates of at least moderate brain atro- dent sample. Finally, one cannot generalize results of the
phy on CT scan. Although the two groups did not differ logistic regression analysis to other patient populations,
on measures of EEC sleep continuity or NREM sleep such as nondepressed but demented patients or de-
architecture, nonsurvivors were characterized at base- pressed but nondemented patients.
line by significantly longer REM sleep latencies, shorter Bearing these caveats in mind, the current data do
first REM sleep periods, and a less robust REM sleep appear consistent with recent studies suggesting in-
rebound following sleep deprivation. A greater propor- creased mortality rates among elderly community resi-
tion of nonsurvivors also had AHI values greater than 3. dents20 and nursing home residents2’ who have
The Apnea-Hypopnea Index and REM latency correctly sleep-disordered breathing and among mid-life patients
predicted 77% of survivors versus nonsurvivors at two- with obstructive sleep apnea syndrome. Other studies
year follow-up, and the time to mortality was signifi- have also reported a modest relationship between sever-
cantly and positively related to baseline REM sleep time, ity of Alzheimer’s dementia and extent of sleep-disor-
percent REM sleep, and the duration of the first REM dered breathing’2; a possible relationship between
sleep period. overnight mental-status deterioration and severity of
Although diminished REM sleep duration and sleep- sleep apnea24; and decreases in REM sleep duration as
disordered breathing emerged as predictors of mortality, dementia progresses.25
alterations in these variables may be viewed as indica- Our data raise the possibility that, apart from the added
tions either of central nervous system impairment (brain mental-status morbidity imposed by sleep-disordered
failure), of systemic medical illness, or of both. However, breathing, EEC sleep alterations may also predict mortal-
because equal proportions of survivors and non- ity. The specific changes that we observed reflect both the
survivors had systemic illnesses, the presence of systemic capacity to generate REM sleep and the capacity to main-
medical illness may not adequately explain the relation- tain respiratory control during sleep. Impairment in both
ships observed between REM sleep variables and sleep- capacities may result, we speculate, from degenerative
disordered breathing or between REM sleep variables changes in brainstem cholinergic neuronal mechanisms
and mortality at two years. regulating REM sleep and respiration. This hypothesis
An alternative interpretation is that diminished basal could be tested by determining if cholinergic muscarinic
REM sleep, a reduced REM sleep rebound capacity, and agonists potentiate REM sleep and diminish sleep apnea
higher rates of sleep-disordered breathing represent spe- in patients with mixed symptoms of depression and
cific central nervous system changes, particularly in dementia.
brainstem areas thought to be anatomic substrates for the The finding of a higher mortality rate among elderly
control of REM sleep and respiration. Cholinergic mech- subjects with an AHI greater than 3 suggests that lower
anisms have been implicated in the control of both.’7’9 levels of sleep-disordered breathing may represent a
The higher incidence of brain atrophy seen on CT scans higher risk for patients with mixed cognitive impairment
among nonsurvivors is also consistent with the view that and depression than was previously thought. This find-
altered REM sleep and sleep-disordered breathing reflect ing awaits cross-validation in an independent sample.
brain failure rather than systemic illness. Finally, the finding that nonsurvivors were character-
Finally, the survivors had EEC sleep findings that re- ized by diminished response capacities, both during
sembled those seen in elderly patients who are depressed wakefulness (they had higher scores on measures of
(a short REM sleep latency, long first REM sleep period), anergia and blunted emotional response) and during
while sleep EEC findings in nonsurvivors resembled sleep (their REM sleep rebound was diminished after
those seen in patients with dementia (diminished REM total sleep deprivation) may be noteworthy.
sleep at baseline and during recovery from sleep depri- These results invite some preliminary comparisons
vation, and increased sleep apnea).2”2 with studies of prolonged sleep deprivation in rats, re-
These data must be viewed as preliminary for several cently reported by Rechtschaffen et al.26 and Everson et
reasons. The small sample size means that caution should al.27 Total sleep deprivation and selective sleep-stage de-
be used when interpreting the results, and a cross-vali- privations (paradoxical sleep deprivation) led to death in
dation of the logistic regression model is necessary. The 11 to 54 days. In one experiment, eight rats who appeared
study was an exploratory, post-hoc examination of exist- near death were allowed to “recover” by sleeping; five
ing data. Univariate contrasts were initially computed, rats recovered, and three died. Interestingly, the three
and variables that differed significantly between survi- rats that died had less sleep during recovery and fewer
vors and nonsurvivors were then entered into a logistic long bouts of sleeping than the rats that recovered. Even
regression model. Hence, confidence in the predictor more intriguing is the observation that the five surviving
model is limited until it is cross-validated in an indepen- rats had very large rebounds in paradoxical (rapid-eye

370 VOLUME 1 #{149}

NUMBER 4 FALL
#{149} 1989
 HOCH et a!.

movement) sleep during recovery. function. Future sleep deprivation experiments in elderly
As noted, the survivors in our study had approximately humans should include concurrent measurement of core
twice as much REM sleep during recovery from sleep body temperature to further explore these hypotheses.
deprivation as the nonsurvivors. Thus, the studies by
Rechtschaffen et al.26 and Everson et al.27 and our own
study suggest that the ability to generate recovery sleep- This study was supported in part by National Institute of
particularly REM sleep-following sleep deprivation is Mental Health grants MH37869 and MH00295 to Dr. Reyn-
associated with improved survival. The underlying olds, grant MH00792 to Dr. Hoch, and grant MH3091 5 to Dr.
mechanism for such an association is unclear. Vitiello et Kupfer. The authors thank Mathias Berger, M.D., and Allan
al. and Rechtschaffen et al.26 speculate that sleep may be Rechtschaffen, Ph.D.,for their review of the data presented here.
necessary for effective thermoregulation, while our stud- This study was presented at the Conference on Sleep and
ies suggest that impaired REM sleep generation indicates Health Risk, held March 11, 1989, at Philipps University in
the presence of cerebral or brainstem cholinergic dys- Marburg, West Germany.

Ref erences
I. Reynolds CF. Perel JM, Kupfer DJ, et al: Open-trial response to 15. Harrell FE: The LOGIST procedure, in SUGI Supplemental Library
antidepressant treatment in elderly patients with mixed depression Users Guide. Edited by Hastings RP. Cary, NC, SAS Institute, 1986,
and cognitive impairment. Psychiatry Res 1987; 21:111-122 pp 269-293
2. Reynolds CF. Hoch CC, Kupfer DJ, et al: Bedside differentiation of 16. Katz D, Baptista J, Azen SP, et al: Obtaining confidence intervals for
depressive pseudodementia from dementia. Am J Psychiatry 1988; the risk ratio in cohort studies. Biometrics 1987; 34:469-474
145:1099-1103 17. Dev NB, Loefschke HH: A cholinergic mechanism involved in the
3. Buysse DJ, Reynolds CF. Kupfer DJ, et al: Electroencephalographic respiratory chemosensitivity of the medulla oblongata in the cat.
sleep in depressive pseudodementia. Arch Gen Psychiatry 1988; Pfluger’s Arch 1979; 379:29-36
45:568-575 18. Willenberg IM, Dermietzel R, Leibstein AG, et al: Mapping of
4. U.S. Department of Health and Human Services: Health United cholinoceptive (nicotinoceptive) neurons in the lower brain stem,
States, 1987 (publication no. DHHS 88-1232). Hyattsville, MD, Na- with special reference to the ventral surface of the medulla. J Auton
tional Center for Health Statistics, 1988, pp 56-57 Nerv Syst 1985; 14:287-298
5. Reynolds CF, Kupfer DJ, Hoch CC, etal: Sleep deprivation as a probe 19. Hobson JA, McCarley RW, Wyzinski PW: Sleep cycle oscillation:
in the elderly. Arch Gen Psychiatry 1987; 44:982-990 reciprocal discharge by two brainstem neuronal groups. Science
6. Reynolds CF, Kupfer DJ, Houck PR, et al: Reliable discrimination of 1975; 189:55-58
elderly depressed and demented patients by EEC sleep data. Arch 20. Bliwise DL, Bliwise NC, Partinen M, et al: Sleep apnea and mortality
Gen Psychiatry 1988; 45:258-264 in an aged cohort. Am J Public Health 1988; 78:544-547
7. Endicott J, Spitzer RL: A diagnostic interview: the Schedule for 21. Ancoli-Israel 5, Klauher MR. Kripke DF, et al: Increased risk of
Affective Disorders and Schizophrenia. Arch Gen Psychiatry 1978; mortality with sleep apnea in nursing home patients: preliminary
35:837-844 data. Sleep Res 1988; 17:140
8. Spitzer RL, Endicott J,
Robins E: Research Diagnostic Criteria: ratio- 22. He J, Kryger MH, Zorick FJ, et al: Mortality and apnea index in
nale and reliability. Arch Gen Psychiatry 1978; 35:773-782 obstructive sleep apnea. Chest 1989; 94:9-14
9. American Psychiatric Association: Diagnostic and Statistical Manual 23. Hoch CC, Reynolds CF, Nebes RD. et al: Clinical significance of
of Mental Disorders, Third Ed. Washington, DC, American Psychi- sleep-disordered breathing in Alzheimer’s disease: preliminary
atric Association, 1980 data. Journal of the American Geriatrics Society 1989; 37:138-144
10. Hamilton M: Development of a rating scale for primary depressive 24. Moldofsky H, Goldstein R, McNicholas WI, et al: Disordered breath-
illness. British Journal of Social and Clinical Psychology 1967; 6:278- ing during sleep and overnight intellectual deterioration in patients
296 with pathological aging, in Sleep/Wake Disorders: Natural History,
II. Folstein MF, Folstein SE, McHugh PR: Mini-Mental State: a practical Epidemiology, and Long-Term Evolution. Edited by Guilleminault
method for grading the cognitive state of patients for the clinician. J C, Lugaresi E. New York, Raven, 1983, pp 143-150
Psychiatr Res 1975; 12:189-198 25. Vitiello MV, Bokan JA, Kukull WA, et al: Rapid eye movement sleep
12. Hoch CC, Reynolds CF. Kupfer DJ, et al: Sleep-disordered breathing measures of Alzheimer’s-type dementia patients and optimally
in normal and pathologic aging. J Clin Psychiatry 1986; 47:499-503 healthy aged individuals. Biol Psychiatry 1984; 19:721-734
13. Blessed C, Tomlinson BE, Roth M: The association between quanti- 26. Rechtschaffen A, Bergmann BM, Everson CA, et al: Sleep depriva-
tative measures of dementia and of senile change in the cerebral gray tion in the rat: x. integration and discussion of the findings. Sleep
matter of elderly subjects. Br J Psychiatry 1968; 144:797-811 1989; 12:68-87
14. Hughes CP, Berg L, Cohen LA, et al: A new clinical scale for the 27. Everson CA, Gilliland MA, Kushida CA, et al: Sleep deprivation in
staging of dementia. Br J Psychiatry 1982; 140:566-572 the rat: IX. recovery. Sleep 1989; 12:60-67

JOURNAL OF NEUROPSYCHIATRY 371