Sleep, gender, and depression: An analysis of gender effects on the electroencephalographic sleep of 302 depressed outpatients

BIOL PSYCHIATRY 1990;28:673-684

673

Sleep, Gender, and Depression: An Analysis of Gender Effects on the Electroencephalographic Sleep of 302 Depressed Outpatients Charles F. Reynolds III, David J. Kupfer, Michael E. Thase, Ellen Frank, David B. Jarrett, Patricia A. Coble, Carolyn C. Hoch, Daniel J. Buysse, Anne D. Simons, and Patricia R. Houck

Gender-related differences in electroenceplu~lographic (EEG) sleep were examined in 151 pairs of men and women with major depression, all outpatients, matched for age and severity of depression. Across five decades (age 21-69), depressed men had less slow-wave sleep than did depressed women. Gender differences were small with respect to visually scored measures of slow-wave sleep time and percent, but moderate for gender differences in automated measures of slow-wave density. The time constant of the polygraph preamplifier significantly affected both visually scored and automatically scored slow-wave sleep. Other measures, such as gEM sleep latency, first REM period duration, sleep efficiency, and early morning awakening, showed robust age effects, but no main effects for gender or gender-by-age interactions. Gender effects on slow-wave sleep and delta-wave counts in depression parallel gender effects seen in heclthy aging. The possibility of occult alcohol use by depressed male outpatients cannot be definitely excluded as a partial explanation of the current findings. However, covarying for past alcohol abuse did not negate the statistical significance of the observed gender effects on slowwave sleep and delta-wave density. The possibility of gender differences in slow-wave regulatory mechanisms is suggested, but similarity in temporal distribution of delta-wave density between the first and second non-rapid-eye-movement (NREM) periods does not support gender differences in slow-wave sleep regulation.

Introduction The sleep physiological correlates of major depression, such as shortened rapid-eyemovement (REM) sleep latency and prolonged first REM sleep period, are subject to many sources of variance (for review, see Reynolds and Kupfer 1987). These include age, inpatient/outpatient status, severity of depression, diagnostic subtype (such as endogenous or nonendogenous), and episode duration (Kupfer et al. 1988). Sleep changes

From the Sleep and Biological Rhythms Research Laboratory, Western Psychiatric Institute and Clinic, University of Pittsburgh School of Medicine, 3811 O'Hara Street, Pittsburgh, PA, 15213. Address reprint requests to Dr. C.F. Reynolds Ill, Western Psychiatric Institute and Clinic, University of Pittsburgh School of Medicine, 3811 O'Hara Street, Pittsburgh, PA 15213. Received January 17, 1990; revised April 4, 1990. © 1990 Society of Biological Psychiatry

0006-3223/90/$03.50

674

BIOL PSYCHIATRY 1990;28:673-684

C.F. Reynolds IR et al.

in depressed inpatients become more pronounced with advancing age (Gillin et al. 1981; Kupfer etal. 1982), resulting in a progressive decrease in REM latency, sleep efficiency, and slow-wave sleep. The sleep stigmata of depressgon occur more predictably, and are of greater magnitude, in endogenous than in nonendogenous forms of illness, and in inpatients than in outpatients (Reynolds and Kupfer 1987). Finally~ REM sleep abnormalities may be greater during the first month ot an episode than later in an episode (Kupfer etal. 1988). Despite the well-established increased prevalence of unipolar depression in women, no systematic investigations of sex-related variance in the sleep physiological correlates of depression have been reported. If gender is a significant source of variance in the sleep abnormalities of depression, this would have important methodological and theoretical implications (e.g., possible sex differences in sleep regulatory mechanisms). Gender does have an important impact on sleep in healthy aging (Miles and Dement 1980). For example, both among tl,_e healthy elderly and those with chronic illnesses, slow-wave sleep and overall sleep efficiency show greater diminution in men than in women (Miles and Dement 1980; Reynolds et al. 1985), leading to the suggestion of a differential aging process in men and women (Webb 1982). Studies of sleep-disordered breathing and nocturnal myoclonus have also indicated that these phenomena are more prevalent in elderly men than in elderly women (Ancoli-lsrael et al. 1987). It is therefore important to determine whether the sex-related dLfferences that characterize sleep in healthy aging also characterize sleep in depression (or are removed by depression), and, if so, how large such effects are and whether there are any age-by-gender interactions.

Subjects and Methods In order to answer this question, we performed a retrospective examination of EEG sleep among outpatients with major depression as defined by Research Diagnostic Criteria (RDC) (Spitzer et al. 1978). The sample was pooled from five different NIMH-funded studies conducted in our laboratory during the past 7 years (Maintenance Therapies of Recurrent Depression; Maintenance Therapy of Late-Life Depression; Sleep, Aging, and Mental Illness: Nocturnal Penile Tumescence in Depression; and Psychobiology of Reco" e~y ~om L~pression). Because ~I of the protocols involve common methodology with respect to subject recruitment, and ev~uation, ~ d because o ~ Clinical Research Center (MH30915) has implemented procedures to ensure diagnostic reliability across protocols, a pool of 420 usable patients (269 female, 151 male) was available, from which a study sample of 151 male to female pairs of patients, matched on age ( - 5 years, all within the same decade) and Hamilton Depression Severity ( + 5 points), was constituted. Sample characteristics are shown in greater detail in Table 1. All patients met SADS/RDC (Spitzer et al. 1978) for current major depression, unipolar and nonpsychotic, with a Hamilton severity rating (Hamilton 1967) of at least 15 after a 2-week psychotropic drug-free washout. Other nonpsychotropic prescription use was allowed (e.g., antihypertensives, thyroid supplementation, or nonsteroidal anti-inflammatory compounds), to ensure that patients' medical conditions were stable and under optimal management. Some patients also met SADS-L criteria for a past history (>~2 years before entry into protocol) of alcohol abuse: 27.1% of men versus 12.5% of women (X2 = 15.2 p < 0.001). Age groups were defined by decade (20-29, 30-39, 40-49, 50-59, 60-69), with 11-32 pairs per decade. We selected 20 representative dependent variables from the domains of sleep conti-

Sleep, Gender, and Depression

BIOL PSYCHIATRY 1990;28:673-684

675

Table 1. Sample Description Variable Age (Mean _ SD) (Range) Hamilton Depression Rating ( M e ~ 4- SD) (Range) Recurrent depression" Single episode Episode number (median)" p < 0.001 (Range) Endogenous No Probable Yes

Men (n = 151)

Women (n = 151)

40.0 4- 11.6 (21--69) 21.1 4- 4.2

40.0 4- 11.4 (21--68) 21.5 .4- 3.8

(15-34) 83% 17% 3.0 (1--48)

(15-34) 94% 6% 5.0 (1-54)

17% 18% 65%

12% 28% 60%

p value NS NS

X2 = 8.31 p < 0.005 Z -- - 5.00

NS NS NS

~l'he difference in proportion of men and women having single-episode versus rectwrent depression reflects the fact that patients wf'-redrawn from different protocols having different entry criteria with respect io a history of recurrent depression.

nuity, sleep architecture, and REM sleep organization, all of which in previous studies had been shown to be affected in depression. These included 14 variables derived from visual scoring of sleep records and 6 from automated REM sleep and slow-wave sleep analysis. Definition of dependent variables, method of polysomnography and automated slow-wave sleep and REM sleep analysis, and scoring reliability have been published previously (Ganguli et al. 1987; Kupfer et al. 1988). Sleep studies were performed using Grass model 78D polygraphs, equipped with a 0.5-A low-frequency switch with adjustable settings of 0.3 and 1.0 Hz. Two different settings had b~cn ~ e d in the five protocols: 149 patients (104 male, 45 female) were studied with a switch setting of 0.3 (fall time constant: 250 msec), and 153 patients (106 female, 47 male) were studied using a switch setting of 1.0 (fall time constant 100 msec). The 153 patients studied at a low-filter setting of 1.9 were all participants in the Maintenance Therapies of Recurrent Depression study and were required to be minimally in their third lifetime episode of depression. Patients from the remaining four protocols were studied at a switch setting of 0.3 and had less highly recurrent illness. Because switch setting affects the amount of slow-wave sleep that is scored, with more slow-wave sleep in patients recorded at 0.3 than 1.0, we performed separate subgroup analyses on visual and automated slow-wave sleep variables. In reporting results, we will sometimes use the term "time constant," for convenience, when referring to the different 0.5-A lowfrequency switch settings. The fall time constants corresponding to the 0.5-A low-frequency values of 1.0 and 0.3 are 0.1 and 0.25 msec, respectively. The longer fall time constant of the 0.3-Hz setting permits more accurate reproduction of slow-wave activity between 0.5 and 1 Hz than does the shorter fall time constant of the 1.0-Hz setting. Analyses were performed on data from the second night in the laboratory, in order to ensure comparability with published data sets. We chose not to use first-night data in order to avoid first-night effects. Third-night data could not be used because of different experimental procedures across protocols. The existence of a first-night effect in this sample was confirmed for some variables, but not for others, using paired t-tests. Thus, with respect to sleep efficiency, a first-night effect was present in both wen (paired t =

676

BIOL PSYCHIATRY 1990;28:673-684

C.F. Reynolds [] et al.

3.90, p < 0.001) and women (paired t = 3.03, p < 0.003). With respect to REM latency (square-root transformation), a first-night effect was present in women (paired t = - 4.17, p < 0.001) but not in men (paired t = - 1.61, p < 0.11). In neither variable was there a night-by-sex interaction. With respect to slow-wave sleep measures (both minutes and percent), neither men nor women showed a first-night effect. A two-way analysis of variance examined the main effects of gender and age, as well as any interactions. As a check on type-I error inflation, we performed three multivariate analyses of variance (MANOVAs) on three conceptually and empirically linked groups of variables: sleep continuity, REM sleep, and slow-wave sleep. Because of the different proportions of men and women having a past history of alcohol abuse (27.1% versus 12.5%, respectively), we used a history of alcoholism as a covariate ("dummy" variable) in further ANOVAs of variables showing significant sex effects. Also, because of differences in the proportion of men and women having single-episode versus recurrent depression, we performed an additional ANOVA on variables showing significant sex effects, limited to patients with recurrent depression only. However, recent reports by Giles et al. (1989) and by Buysse et al. (1988) have suggested little, if any effect of single-vs-recurrent episode on the EEG sleep correlates of depression. Finally, for variables showing sex effects, we examined critical effect sizes, following procedures of Kraemer and Thieman (1988), in order to characterize any differences as small, medium, or large. Critical effect size is calculated as the quantitative difference between the two group means, divided by the measure of variability (pooled standard deviation) and adjusted for unbalanced sample sizes. Results

Main Effects due to Gender The multivariate analysis of variance (MANOVA) showed a significant effect of gender on slow-wave sleep measures for all age groups (n = 267, Hotelling's t = 2.88, p < 0.02) (Figures 1, 2, Tables 2-7). The effect of low filter setting (time constant) was also confirmed (Hotellhag's t = 6.25, p < 0.001). No significant effect of gender was detected in the MANOVAs using measures of sleep architecture or REM sleep.

Visual Measures of Slow-Wave Sleep. In the univariate two-way ANOVAs, minutes of slow-wave sleep showed an effect of gender (p < 0.02), with depressed men consistently having fewer minutes of slow-wave sleep across all five decades (Table 3, Figure 1). A nearly significant sex effect was also demonstrated for slow-wave sleep percent (p < 0.06). These results were derived from the subsample of 149 subjects recorded at a low filter setting of 0.3. Similar results were obtained in the analysis of 153 patients recorded at a low filter setting of 1.0 (Table 4), in which trends due to sex were found for both minutes of slow-wave sleep (p < 0.07) and percent of slow-wave sleep (p < 0.08). As shown in Table 7, the critical effect sizes for these visually scored measures of slow-wave sleep ranged from 0.14 to 0.18 and would thus be characterized as "small" (range 0.1--0.24) by Cohen (1969). Automated Measures of Slow-Wave Sleep. At a low filter setting of 1.0, but not 0.3, depressed women were found to have higher total delta-wave counts and greater deltawave counts oer minute than depressed men. An examination of total delta-wave counts and counts per minute in the first non-REM (NREM) sleep period (low filter setting =

Sleep, Gender, and Depression

BIOL PSYCHIATRY

677

1990;28:673--684

TIMECONSTANT= 1.0

TIMECONSTANT= 0.3 100-

'°°t

'°1

90-

"~

80.

80"

70-

70-

\

a. IP

(~ ~p

.

~ \

6O-

6o50-

5040-

o ~

\

\

•

3o-

30.

20-

20'

10-

10-

Legend X MALES

o

~..~'~..0'

~.o..~.~'.~o..~'a,o..J,~'

Age Group

4- FEMALES

o Age Group,

Figure 1. Visual slow-wave sleep in ndnutes by gender across decade at both 0,3 and 1.0 time constant. A gender and age effect is detected at both time constants.

1.0) also revealed significantly more delta activity in women than in men. However, the temporal distribution of delta-wave activity, as measured by delta ratio (the ratio of deltawave counts per minute in the first NREM period to the second NREM period), showed no significant sex effects at either time constant. The temporal profile of slow-wave sleep density is shown in Figure 2. Regardless of time constant, a significant effect of NREM period (p < 0.0001) is present m both groups, reflecting the progressive decrease in slow-wave sleep density across the night. In patients recorded at a switch setting of 1.0, a significant gender effect was present for both the 0.5-2 Hz (p < 0.05) and 0.5-3 Hz (p < 0.02) band widths. Also, a NREM-period-bygetider interaction was present in the 0.5-2.0-Hz band width (p < 0.05), with a trend in the 0.5-3.0-Hz band widths (p < 0.09). In patients recorded at a switch setting of 0.3, trendworthy sex effects were noted in both band widths (p < 0.09), but no NREMperiod-by-gender interactions were found. At neither time constant did we find a significant three-way interaction (NREM period by gender by age) for delta counts per minute (0.53 Hz). As noted in Table 7, critical effect sizes for gender differences in slow-wave density were moderate (~>0.25 but <0.4) among patients recorded at a time constant of 1.0, but not significant i~ those recorded at 0.3.

Effects of Past Alcohol Abuse. As shown in Tables 5 and 6, significant effects of a history of alcohol abuse were noted in measures of first NREM period delta-wave counts (total, counts per minute), in patients studied at a low filter setting 1.0. In general, however, the same pattern of sex-related differences in slow-wave sleep and delta-wave counts was observed both before and after the effects of a history of alcohol abuse were

678

C.F. Reynolds HI et al.

BIOL PSYCHIATRY 1990;28:673-684

m o a ~ qc~oz)

N~ \

6v,n~

Leg~O

~ 2O-

-

o _~-J_~_~_

~

o

~

"o

Nocr-4~l~ Pciod

Non-q~[~ F~riod

wu~s O~to)

maazs t~=to)

I o-

o

5a.-.~.fe~g.

o

i z0-a6~=~

o~m_f~g

[] _~-._~_f~.~

o ~..L-W.

o L~--s._~.

m-

Non-R~ Period

Figure 2. Distribution of delta counts per minute by sex and by decade across NREM periods at both 0.2 and 1.0 time constants. A significant gender effect and a period-by-gender interaction is present at t ~ e constant 1.0, but not at 0.3.

covaried. Furthermore, the proportion of men with a positive SADS-L history of alcohol abuse did not differ between the subgroup recorded at a low-filter setting of 0.3 (29.3%) and that recorded at a setting of 1.0 (24.8%). The chi-square statistic was 0.52 (NS).

Effects of Recurrent Depression. In the secondary analysis restricted to patients with recurrent depression, a significant gender effect for minutes of slow-wave sleep was seen in the subgroup recorded at a low-filter setting of 0.3 (n = 90, p < 0.05) and a trend was observed in the subgroup recorded at a setting of 1.0 (n = 152, p < 0.09). Similarly, when we examined slow-wave sleep and automated delta-wave measures in patients (47 male, 106 female) studied in our Depression Prevention Program (MH29618), a program with very restrictive criteria that selects patients with highly recurrent illness, robust sex differences (F > M) were found, as indicated in Table 4 and Figure 2. No significant gender effects were found for any other EEG sleep variables that are

Sleep, Gender, and Depression

679

BIOL P S Y ~ T R Y

1990;28:673-684

~ d .R~M Sleep in

Table 2. Effects o f G e n d e r a n d A g e o n Selected M e a s u r e s o f Sleep C o n t i n u i ~ D e p r e s s e d Outpatients Male Measure

Mean ± SD

Female (n - 151) Mean ± SD

Hamilton 17-1tem Visual derived measures Total recording period (min) Sleep latency (min) Sleep efficiency (%) Maintenance (%) Early morning awake (min) REM latency (min) REM time (%) REM density REM activity (units) First REM period REM time (rain) REM activity REM density Automated REM measures Total REM counts REM counts per minnte

21.1 -- 4.1

21.5 _-+ 3.7

(n -

428.1 20.9 87.0 91.5 18.5 65.4 22.4 1.4 119.3

151)

__ 44.8 _+ 21.3 __ 10.5 ± 9.8 ± 24.5 ± 32.6 ± 6.2 ± 0.4 ± 54.7

430.0 23.3 87.7 92.8 16.6 64.4 23.2 1.3 119.1

__ 46.7 -4- 26.6 __ 10.9 ± 9.5 ± 23.7 ± 30.3 ± 6.0 ± 0.4 ± 59.4

Effects

Age by Gender -~ --0.06 m -~ -~

Age

Gender

m

m

0.0001 0.0001 0.0001 0.005

m

n

0.05 0.08

18.1 ± 11.1 23.5 __ 22.0 1.2 ± 0.5

19.7 ± 11.8 23.6 ± 21.8 1.1 ± 0.5

~ ----.

0.05 0.005 0.005

603.8 ± 363.5 7.2 ± 3,6

654.6 ± 406.7 7.3 ± 3.9

0.08 --

m

n

m

n

Table 3. Effects o f G e n d e r a n d A g e on S l o w - W a v e S l e e p M e a s u r e s ( T i m e C o n s t a n t = 0.3) Effects

Measure Hamilton 17-item Slow-wave sleep (min) Slow-wave sleep (%) Total counts (0.5-2.0 Hz) Total counts (0.5-3.0 Hz) Total cpm (0.5-2.0 Hz) Total cpm (0.5-3.0 Hz) First NREM period counts (0.5-2.0 Hz) First NREM period counts (0.2-3.0 Hz) First NREM period counts (0.5-2.0 Hz) First NREM period counts (0.5-3.0 Hz) Delta ratio (0.5-2.0 Hz) Delta ratio (0.5-3.0 Hz)

Male (n = 104) Mean __ SD 20.9 ~4.1 ± 12.0.45032.9 6966.3 !7.4 ± 24.2 1816.5 -

3.6 33.7 9.2 3280.2 4715.3 10.6 15.4 1575.9

Female (n = 45) Mean - SD 21.9 59.2 15.0 5410.6 7655.8 !8.5 26.2 1611.4

+ 4.2 _ 41.8 ± 9.6 - 3172.3 - 4916.5 -4- 9.8 .4- 15.6 + 1215.4

Age by Gender 0

.

0

~

0.02 0.06

Age m

0.01 0.05 0.001

0.002 0.01 0.02 0.05

2557.0 - 2206.0

2305.5 ± 1760.1

0.06

26.8 + 17.7

26.1 ± 16.6

0.01

37.9 - 24.9

37.8 - 25.0

0.01

1.2 ± 0.5 1.3 - 0.5

1.3 - 0.7 1.3 ± 0.6

Gender m

n

i

n

m

w

m

D

680

C.F. Reynolds met

toOL PSYCHIATRY 1990;28:673-684

al.

Table 4. Effects o f G e n d e r a n d A g e o n S l o w - W a v e Sleep M e a s u r e s ( T i m e C o n s t a n t = 1.0) Effects Male

Female (n = 106) Mean ± SD

(n = 4 7 )

Measure

Mean _+ SD

Hamilton 17-item Slow-wave sleep (min) Slow-wave sleep (%) Total counts (0.5-2.0 Hz) Total counts (0.5-3.0 I-Iz) Total cpm (0.5-2.0 Hz) Total cpm (0.5-3.0 Hz) First NREM period counts (0.5-2.0 Hz) First NREM period counts (0.5-3.0 Hz) First NREM period counts (0.5-2.0 Hz) First 1VREM period counts (0.5-3.0 Hz) Delta ratio (0.5-2.0 Hz) Delta ratio (0.5-,3.0 Hz)

21.7 19.4 5.3 2006.4 4112.7 7.0 14.3 720.7

- 5.0 - 27.0 _ 7.8 _ 1557.7 ± 3308.7 ± 5.3 _ 11.4 __ 775.7

21.3 31.8 8.5 2948.1 6280.6 I0.I 21.5 1221.5

.4.4.4.4± ± ± ±

Gender

3.5 31.5 8.4 1721.8 3745.9 5.5 12.2 924.7

Age

~ 0.07 0.08 0.02 0.01 0.02 0.01 0.01

Age by Gender

0.02 0.02 0.02 0.02

m

0.001

1470.1 ± 1480.3

2552.2 ± 1807.4

0.002

0.02

10.6 - 9.4

17.9 ± 11.9

0.002

0.001

0.08

22.2 ± 18.4

37.0 ± 22.2

0.001

0.~i}5

0.05

1.5 ± 1.5 1.5 ± 1.0

1.8 ± 2.2 1.7 ± 1.3

-~

m

w

characteristically abnormal in depression, such as REM latency, duration of first REM period, sleep efficiency, and early morning awakening (Table 2).

:fain Effects Due to Age Consistent with previous studies of inpatients (e.g., Gillin et al. 1981), many visual and automated sleep measures from the current sample of outpatients showed robust age

Table 5. Effects o f G e n d e r a n d A g e o n S l o w - W a v e S l e e p M e a s u r e s C o v a r i e d for A l c o h o l History ( T i m e C o n s t a n t = 0.3)

Measure

Alcohol

Gender

Age

Hamilton 17-item Slow-wave sleep (min) Slow-wave sleep (%) Total counts (0.5-2.0 l-lz) Total counts (0.5-3.0 Hz) Total cpm (0.5-2.0 Hz) Total cpm (0.5-3.0 Hz) First period (0.5-2.0 Hz) First period (0.5-3.0 Hz) First period cpm (0.5-2.0 Hz) First period c p m (0.5-3.0 Hz) Delta ratio (0.5-2.0 Hz) Delta ratio (0.5-3.0 Hz)

__ -0.09 --~ ~ ~ _ -~ ---

m 0.05 0.07 -m --~ u ~ ~ ~ u

0.08 0.06 -0.002 0.01 0.01 0.02

0.01 0.02

Gender by sex/age interaction

q

u

m

m

Sleep, Gender, and Depression

niOL PSYCHIATRY 1990;28:673--684

681

Table 6. Effects of Gender and Age on Slow-Wave Sleep Measures Covaried for Alcohol History (Time Constant = 1.0)

Measure Hamilton 17-item Slow-wave sleep (min) Slow-wave sleep (%) Total counts (0.5-2.0 Hz) Total counts (0.5-3.0 Hz) Total cpm (0.5-2.0 Hz) Total cpm (0.5-3.0 l-lz) First period (0.5-2.0 Hz) First period (0.5-3.0 HT.) First period cpm (0.5-2.0 Hz) First period cpm (0.5-3.0 Hz) Delta ratio (0.5-2.0 Hz) Delta ratio (0.5-3.0 Hz)

Alcohol

Gender

Age

0.06

0.07 0.09 0.01

0.02 0.02 0.05

0.01

--

0.02

0.05

0.01

0.05 0.09 0.01 0.05

Gender by sex/age interaction

~

0.002 0.001 0.001 0.001

0.01 0.05 0.01 0.09

0.09 0.05 0.05

effects. The MANOVA showed significant effects for age on sleep contLauity variables (Hotelling's t = 5.05, p < 0.001), on REM sleep measures (Hotelling's r = 3.51, p < 0.001), and on slow-wave sleep measures (Hotclling's t = 1.92, p < 0.02). In the univariate two-factor ANOVAs, sleep efficiency and sleep maintenance, as well as REM latency and slow-wave sleep percent, all decreased with age. Similarly, most automated delta measures (at both time constants) declined with age, although delta ratio was stable across the five decades examined. (Data for each variable, i.e., means +_ SD by decade, are available upon request.)

Gender-by-Age Interactions The MANOVA showed no significant gender-by-age interactions in sk,ep continuity, REM, or slow-wave sleep measures. Table 7. Critical Effect Sizes a of Gender on Slow-Wave Sleep Measures Time constant Measure Slow-wave sleep (min) Slow-wave sleep (%) Total counts (0.5-2.0 Hz) Total counts (0.5-3.0 Hz) Total cpm (0.5-2.0 Hz) Total cpm (0.5-3.0 Hz) Delta counts first period (0.5-2.0 Delta counts first period (0.5-3.0 Delta counts first pe~od (0.5-2.0 Delta counts first period (0.5-3.0 Delta ratio (0.5-2.0 Hz) Delta ratio (0.5-3.0 Hz) aKraemer and Thiemann (1988).

Hz) Hz) Hz) Hz)

0.3

1.0

0.18 0.14 --~ ~ ~ -~ -~ ~

0.18 0.17 0.24 0.26 0.25 0.26 0.25 0.27 0.28 0.29

682

BIOL PSYCHIATRY 1990;28:673-684

C.F. Reynolds Ill et al.

Discussion Outpatient men in an episode of unipolar, nonpsychotic depression had less slow-wave sleep (minutes, percent) and lower delta-wave counts (total counts and counts per minute) during the entire night and during the first NREM period than did outpatient women. The visually scored measures showed small critical effect sizes, while automated, measu~s showed moderate effect sizes. The automated findings were more robust in the subgroup of patients with highly recurrent depression studied at a low-filter setti~;g of 1.0 than in the subgroup of patients with less recurrent forms of illness studied at a setting of 0.3. No other gender~elated differences in such key dependent variables as sleep efficiency, early-morning awakening, REM sleep latency, and first REM period duration were detected. Although we cannot exclude the possibility of a type 1 error, the MANOVA did show significant effects of gender on slow-wave sleep variables. Moreover, we think that these findings are valid because they partially parallel gender findings seen in healthy older controls, that is, healthy older women show higher levels of slow-wave sleep than healthy older men (Reynolds et al. 1985). On the other hand, however, significant differences in v/sua/measures of slow-wave sleep between younger men and women were not reported in the extensive normative studies of Williams et al. (1974). Hence, the current findings in younger depressives represent a depa_~!re from normative findings in younger adults. Why should slow-wave sleep be lower in depressed men than in depressed women, across the adult life cycle from age 20 to 70? Are there sex differences in the regulatory mechanisms governing the amount, density, and temporal distribution of slow-wave sleep in depression? Factors that may be related to the age-dependent decrease of slow-wave sleep (and possibly to the sex-related differences as well) also include the increase in respiratory disturbances and abnormal movements during sleep that become more prevalent in late life, and that occur more frequently in men than in women (e.g., AncoliIsrael et al. 1987). Other factors that may be related to the age-dependent decrease of slow-wave sleep with age (and that could possibly be relevant to the sex differences observed hetty,) include increasing levels of monoamine oxidase activity, the altered hormonal milieu in late life, loss of dendrite systems in pyramidal cell layers two, three, and five, increased prevalence of ~aytime napping, altered circadian rhythms, and noncurrent medical problems (for review, see Miles and Dement 1980). We are inclined to doubt that sleep fragmentation per se drives down slow-rate sleep; moreover, slow-rate sleep and delta-wave count differences between men and women were much more robust than the small difference in sleep maintenance. We observed that gender differences in slow-wave sleep and delta-wave counts were already present in 20- and 30-year-old patients. This observation is consistent with a recent report by Dijk et al. (1989) of significantly higher power densities during NREM sleep (0.25-11.0 Hz) in healthy young women than men. These authors also reported higher power densities during REM sleep in young women. The presence of sex differences in sleep as early as the third decade, both in healthy controls and in depressives, argues against the cgncept of a gender-dependent differential aging process affecting slow-wave-sleep regulatory mechanisms. Moreover, the general absence of significant age-by-gender interaction, particularly in slow-wave-sleep measures, is further evidence against a gender-dependent differential aging process. Dijk et al. (1989) suggested that sex differences in power spectral densities might be related "to anatom-

Sleep, Gender, and Depression

BIOL PSYCHIATRY

1990;28:673-684

683

ical differences between the sexes, much as skull thickness." In this study we used a minimum amplitude criterion of 75 pN for the automatic counting of delta waves (75200 pLV) in order to be consistent with the conventional criteria for scoring slow-wave sleep (Rechtschaffen and Kales 1968). Therefore, skull differences, which are likely to induce amplitude differences, could very well explain the observed differences in delta-wave counts. We initially questioned whether the sex difference in slow-wave sleep and deltawave counts was an effect of the higher rate of past alcohol abuse in men versus women. Diminution in slow-wave sleep has been consistently reported in alcohol abusers (Giliin et al. 1990), including in subjects studied after prolonged alcohol ~,bstinence (Snytier and Karacan 1985). However, although an alcohol-abuse history was detectable ~n some delta-wave count measures, sex effects were still present after covarying for this history. Moreover, in the subgroup of patients with highly recurrent depression recruited from our Depression Prevention Program, slow-wave sleep and delta-wave count measures were significantly higher Jn women (n = !06) than in men (n = 47). However, we cannot exclude the possibility of continuing, covert alcohol use by the outpatient depressives as an explanation for the gender differences in slow-wave sleep reported here. The current findings relating to gender differences in slow-wave sleep in depressed patients have methodological and theoretical implications. Methodologically, in studies where measures of slow-wave sleep are used as independent or as dependent variables, effects of sex will need to be controlled initially in sample composition or subsequently in statistical analysis. Likewise, investigators need to pay attention to the effect of time constant on the amount of EEG delta activity and slow-wave sleep recorded. As confirmed in the MANOVA, the use of the 0.3 low-frequency switch setting results in recording more delta activity between 0.5 and 1.0 Hz and more slow-wave sleep than does the use of the 1.0 switch setting. With respect to the analyses depicted in Figure 2, we cannot exclude the possibility that gender differences in slow-wave density are inflated by the 1.0 switch setting used in the Depression Prevention Clinic (DPC) subgroup, as compared to the low-frequency switch setting of 0.3 used in the non-DPC suogroup with less highly recurrent depression. In both subgroups, however, there were small but significant effect sizes for differences in visual measures. Theoretically, it may be important that a higher prevalence of depression in women exists despite the better preservation of slow-wave sleep across five decades of life. For example, this may suggest that the better preservation of slow-wave sleep may not be consistent with the concept of slow-wave sleep as a "protective" factor against depression, as we have previously thought (Kupfer and Reynolds 1989). In a related context, provided by the two-process model of sleep regulation (Borb~ly 1982), we would ask whether the higher level of delta-wave counts in depressed women at the start of the night (NREMperiod-by-gender interaction, Figure 2) suggests a higher level of "process S" in women. This may be the case. If so, it suggests the possibility of sex differences in slow-wavesleep regulatory mechanisms that permit a greater rate of slow-wave generation in depressed women. On the otL~r hand, the similarity of delta ratio in depressed men and women (i.e., ratio of delta-wave counts per minute in the first to the second NREM period) and the generally similar "decay" cur~,es depicted in Figure 2 mean that the two groups have similar temporal distribution of delta activity, even if they differ in total delta-wave counts. Similarity in temporal distribution of delta-wave counts may argue against the hypothesis of different regulatory mechanisms.

684

BIOL PSYCHIATRY 1990;28:673-684

C.F. Reynolds HI et al.

Supported in part by NIMH Grants 37869, 00295 (CFR), 40023 (CFR, MET, EF), 43832 (CFR, EF), 29618 (DJK, EF), 41884 (IVl~T, ADS), C~327 (DBJ), 24652, and 30915 (DJK). The authors thank Donna E. Giles, Ph.D., and Timothy H. Monk, Ph.D., for their reviews of the data in this manuscript.

References Ancoli-lsrael S, Kripke DF, Mason W (1987): Characteristics of obstructive and central sleep apnea in the elderly: An interim report. Biol Psychiatry 22:741-750. Borb~ly AA (1982): A two-process model of sleep regulation. Hum Neurobioi 1:195-204. Buysse DJ, Reynolds CF, Houck PR, Stack JA, Kupfer DJ (1988): Age of illness onset and sleep EEG variables in elderly depressives. Biol Psychiatry 24:355-359. Cohen J (1969): Statistical Power Analysis for the Behavioral Sciences. New York: Academic. Dijk DJ, Beersma DGM, Bloem GM (1989): Sex differences in the sleep EEG of young adults: Visual scoring and spectral analysis. Sleep 12:500-507. Ganguli R, Reynolds CF, Kupfer DJ (t 987): Electroencephalographic sleep in young ever medicated schizophrenics: A comparison with delusional and nondelusional depressives and with healthy controls. Arch Gen Psychiatry 44:36-44. Giles DE, Etzel BA, Reynolds CF, Kupfer DJ (1989): Stability of polysomnographic parameters in unipolar depressives: A cross-sectional report. Biol Psychiatry 25:807-810. Gillin JC, Duncan WC, Murphy DL, et al (1981): Age-related changes in sleep in depressed and normal subjects. Psychiatry Res 4:73-78. Gillin JC, Smith T, Irwin M, Kripke DF, Brown S, Schuckit M (1990): Short REM latency in primary alcoholics with secondary depression. Am J Psychiatry 147:106-109. Hamilton M (1967): Development of a rating scale for primary depressive illness. Br J Soc Clin Psychol 6:278-296. Kraemer HC, Thieman S (1988): H.~w Many Subjects: Statistical Power Analysis in Research. Newbury Park, CA. Kupfer DJ, Reynolds CF (1989): Slow wave sleep as a "protective" factor. In Stunkard AJ, Baum A (eds), Perspectives in Behavioral Medicine: Eating, Sleeping, amt Sex. Hillsdale, NJ: Lawrence Erlbaum, pp 131-145. Kupfer DJ, Reyr,olds CF, Ulrich RF, Shaw DH, Coble PA (1982): Sleep, depression, and aging. Neurobiol Aging Exp Clin Res 3:351-360. Kupfer DJ, Frank E, Grochocinski VJ, Gregor M, McEachran AB (1988): Electroencephalographic sleep profile in recurrent depression: A longitudinal investigation. Arch Gen Psychiatry 45:678688. Miles LE, Dement WC (1980): Sleep and aging. Sleep 3:119-220. Rechtschaffen A, Kales A (1968): A Manual of Standardized Terminology, Techniques and Scoring System for Sleep Stages of Human Subjects. Los Angeles, CA: UCLA Brain Information Service, Brain Research Instit-ate. Reynolds CF, Kupfer DJ (1987): Sleep research in affective illness: State-of-the art circa 1987. Sleep 10:199-215. Reynolds CF, Kupfer DJ, Taska LS, Hoch CC, Sewitch DE, Spiker DG (1985): Sleep of healthy seniors: A revisit. Sleep 8:20-29. Snyder S, Karacan I (1985): Sleep patterns of s ~ r chronic alcoholics. Neuropsychobiology 13:97100. Spitzer RL, Endicott J, Robins E (1978): Research Diagnostic Criteria: Rationale and reliability. Arch Gen Psychiatry 36:773-782. Webb WB (1982): Sleep in older persons: Sleep structure of 50-6(, year-old men and women. J Gerontol 37:581-586. Williams RL, Karacan I, Hursch CJ (1974): Electroencephalography of Human Sleep: Clinical Applications. New York: J Wiley.