- Email: [email protected]
Relationship of awakening and delta sleep in depression
Psychiatry
Research.
297
19, 297-304
Elsevier
Relationship
of Awakening
David
Victoria
J. Kupfer,
Received
December
6. IY8.5;
and Delta Sleep in Depression
J. Grochocinski, revised
version
rec,eived
and Ann May
6. McEachran
5, /%‘6;
accepted Ma_\,29, Iys6.
Abstract. It has been demonstrated that deficiencies in slow-wave sleep occur in the sleep profiles of depressed patients. Recent theories of sleep regulation link a deficiency in the so-called “Process S” to these slow-wave sleep alterations. However, the degree of wakening during sleep has been suggested as the explanation for reduced slow-wave sleep. In the present study, the extent of this relationship was examined in both depressed patients and normal subjects. Only a relatively low level of correlation between the degree of wakening and reduction in slow-wave sleep was noted in depressed patients. This finding on 38 unipolar depressed patients is similar to the findings on a smaller sample of depressed individuals. Key Words.
Delta sleep, awakening,
depression,
sleep regulation.
Abnormalities in a two-process model of sleep regulation (a sleep-dependent process termed “Process S,” and a sleep-independent circadian process termed “Process C”)
have been proposed to account for sleep abnormalities in depressive states (Daan et al., 1984). The major tenets of the two-process model of sleep regulation (BorbCly, 1982) as applied to depression are: the level of Process S, as reflected by electroencephalographic (EEG) slow-wave activity, corresponds to the sleep-dependent facet of sleep propensity; the pathognomic changes of sleep in depressives are a consequence of a deficiency in the buildup of Process S. The application of automated delta-wave analyses in normal subjects and younger depressed patients supports the model to some extent (Kupfer et al., 1984). Time spent asleep is positively correlated with total delta waves (normals and depressives) and average delta waves (depressives); delta sleep is lower in depressives than in normals; the average delta wave count is significanty reduced in younger depressives over the total night and in NREM (non-rapideye-movement) period 1. The model also postulates that measures of phasic REM activity are inversely related to Process S, suggesting that Process S exerts an inhibitory influence on phasic REM activity. Support for this hypothesis has also been aided by studies demonstrating comparability of results using spectral analysis and period analysis (with a delta analyzer). In one study, sleep was recorded in nine drug-free depressive patients and in age- and sex-matched normal control subjects. All-night spectral analysis of the sleep EEG
David J. Kupfer. M.D., is Professor and Chairman; Victoria J. Grochocinski, Ph.D.. is Data Base Manager; and Ann B. McEachran, M.S., is Statistician, Department of Psychiatry, University of Pittsburgh School of Medicine. (Reprint requests to Dr. D.J. Kupfer, Western Psychiatric Institute and Clinic, 3811 O’Hara St., Pittsburgh. PA 15213, USA.) 0165-1781/86/$03.50
@ 1986 Elsevier Science Publishers B.V.
298 showed a significantly reduced power density in the 0.25-25.0 Hr band in the depressive group. Power density values integrated over the entire frequency range (0.25-25.0 Hz) exhibited a decreasing trend over the first three NREM sleep cycles for both groups. In each sleep cycle, depressives had lower values than controls with both types of analysis. The results are consistent with the hypothesis that a buildup of a sleepdependent process is deficient in the sleep regulation of depressive patients (BorbCly et al., 1984). While there is little disagreement about the finding of decreased slow-wave sleep in depressed
patients,
there
is considerable
controversy
about
its specificity.
For exam-
ple, it has been strongly suggested that the amount of awakening found in the sleep of depressed patients, particularly in the NREM periods. is responsible for the decreased delta sleep “production.” The Process S model (which does not make an awake time prediction) itself has been challenged on the basis that if decreased slow-wave sleep is strongly associated with awakening, then the model is not specific to depression. To examine this question, we analyzed the relationship between awakening time and average delta-wave sleep in normal controls and in depressed patients. Methods ‘l‘he depressed group comprised 38 inpatients on the Clinical Research Unit (CRU) at Western Psychiatric Institute and Clinic (WPIC). At admission, in addition to a psychiatric interview and a physical examination, we obtained collateral information from the patient’s family and from case records of previous hospitalirations. During a 2-week drug-free period, patients underwent a series of routine laboratory tests, including thyroid function tests, a Ih-channel EEG. and other tests as indicated by history or physical examination. During the drug-free period, the Schedule for Affective Disorders and Schirophrenia (SADS) (Endicott and Spitrer, 1978) was completed by the patient’s psychiatrist. This structured research interview was used to make diagnoses according to the Research Diagnostic Criteria (Spitter et al., 197X). Diagnoses were determined using information obtained from the initial interview, case record, collateral history from relatives, and observation on the CRU. The patient entered the antidepressant drug protocol if illness severity remained sufficiently high at the end of the 2-week, drug-free period, i.e., if the patient had a minimum score of 30 (sum of two raters) on the l7-item Hamilton Rating Scale for Depression (HRSD) (Hamilton. 1960). The HRSD was used twice weekly throughout the investigation to evaluate severity of depression. The overall sleep-scoring techniques have been previously described (Kupfer et al.. 1982); all records were scored independently and without knowledge of the clinical diagnosis and clinical course for specific patients. All-night sleep recordings were made using a Grass (model 78R) polygraph and were simultaneously recorded on l-inch magnetic tapes (Honeywell recorder, model 101). REM and Delta Sleep Analyses. Automated measures were derived from software pr-ograms developed at the University of Pittsburgh and descrihcd previously (Mcl’artland ct al.. 1978). Briefly. these programs arc implemented on a PDI’ I I /44 minicomputer using the LPS-I I laboratory peripheral system. Compressed analog tapes of each subject’s sleep record are played directly into an analog-to-digital converter; the program yields both a minute-by-minute printout and a disk file. The disk file is later matched on a DEC-IO system computer to the visually scored minute-b!,-minute Irecord so that nonsleep minutes. as well as movement artifact, can be accounted for and removed. Our data analyzer program has two main sections: Section I. written in Fortran. provides the operator interaction for program startup and all disk I; 0 functions: Section 2, written in PDP-I I assembly language. provides the real-time data-collection functions (Kupfer ct al.. 1981). Automated
299
The compressed analog tape recordings of the patients’ EECi waveforms are played directly into the analog-to-digital converter of the LPS-I I at a speed I6 times that of the recording speed, allowing an &hour period to be processed in only 30 minutes. The EEG channel is monitored, and data are collected at a real-time rate of 200 samples per second. Each sample is stored in a buffer until 200 samples have been collected. The average of these stored samples is computed and then subtracted from the oldest sample in the buffer. (The effect is to filter the signal digitally with a high-pass characteristic and a roll-off frequency of 0.1 HI, thereby eliminating problems associated with baseline drift.) The standard delineation of delta waves indicative of stage 3 and stage 4 sleep is 0.5-2.0 Hr at 75-200 mV. The adjusted samples (from the buffer) go directly to a baseline crossing detector and then to an amplitude (peak) detector. With each baseline crossing, the period of the preceding half wave is computed from the prior crossing, and the peak is noted and checked. The program normally runs between two calibration marks (which the program detects at the start and end of a record) and provides both a minute-by-minute printout and a disk file. While night counts can then be examined, or the average delta counts per minute analyred. correlations with visually scored sleep can be made for each minute of the night during NREM or REM periods or on any other subset ofthe EEG sleep recording. For automated analyses, each REM period contains a minimum of 5 minutes, although the REM onset time is derived in the usual fashion in which the first REM period needs to be only 3 minutes in length. For this study, we used baseline night 4 for depressed subjects and night 2 for normal controls. Subject Characteristics: Patients. Of the 38 (29 female, 9 male) nondelusional patients with a diagnosis of major depressive syndrome, 3 were bipolar. In addition, IX patients showed psychomotor retardation. I I had psychomotor agitation, and 6 were both agitated and retarded. Patients had an age range of 24-61 years. with a mean age of 41.6 (SD = 10.6) years. The mean initial H RSD score was 40.6 (SD q 9.4) (sum of two raters). The median length of the current episode was 30.0 weeks and the median number of previous episodes was 2.0. The mean REM latency for these subjects was 39.6 (SD = 28.5). Six of the 38 patients experienced a first NREM period shorter than 5 minutes. To ensure reliability. the data for these patients are not included in Table I. Characteristics: Normal Control Subjects. Fifty-tw*o (32 female, 20 male) normal control subjects. with an average age of 34.6 (SD q I I .5) years (age range IX-63 years) formed the comparison group. The mean REM latency for the group was 69.8 (SD = 21.8). Selection criteria demanded that all individuals have a normal sleep-wake cycle; for premenopausal women, the 3 consecutive night studies were in the folhcular phase of their menstrual cycle. Exclusion criteria were as follows: normals with obesity were ruled out, as were dieters who showed significant weight loss: normals could not be on birth control pills or lactating, currently or within the last 6 months. Subjects were also excluded if they had any current medical illness requiring pharmacological intervention, with the exception of well-controlled thyroid or hypertensive disease. Any abnormal physical examination features or laboratory data excluded individuals. Subjects were excluded if they, or their first-degree relatives, had a history of seizure disorder or major psychiatric illness. All individuals were required to be free of drugs at least 2 weeks before the biological studies began and had to have an established regular sleep-wake cycle with a “good-morning” time by 8 a.m. During the drug-free period, all drugs and alcohol were eliminated, with the exception of Tylenol or small doses of aspirin. Sleep investigations were then carried out on the 13th Floor Sleep Laboratory for 3 consecutive nights with the analyses of night 2 being used for this particular investigation. Subject
Statistical analysis involved parametric correlation and analysis of covariance (ANCOVA) techniques. Correlation analysis was used to examine the relationship between minutes of awakening and average delta waves within each of the two groups (depressed and controls). as well as across the two groups as a whole. ANCOVA was used to compare within each NREM period the amount of average delta wavjes in the depressed and control subjects, Statistics.
300 controlling for age and minutes 01 awakenmg. I he square root ot awake and of average delta was used in all analyses to compensate for heteroscedasticity and skewness. Table I contains the results of the ANCOVA, with age and square root of minutes of awake time as the covariates. The last column contains the probabilities associated with (I) the adjusted means (after adjustments for age and awake time, are the means different?); (2) the slopes (is there a difference in the correlations between awake time and delta in the controls and
Table 1. Results covariates
of ANCOVA,
with age and minutes
Normal NREM 1 n Average delta -&i 1 SD Range: Delta Adjusted means: Delta Average minutes of awake INREM 1, Range. Minutes of awake Correlation between sqrt of awake & sqrt of average delta Correlation between age & sqrt of average delta
52 21.2 10.7-35.4 1 .o-54.3 20.2
NREM 2 ” Average delta -&+lSD Range: Delta Adjusted means: Delta Average minutes of awake INREM 21 Range: Minutes of awake Correlation between sqrt of awake & sqrt of average delta Correlation between age & sqrt of average delta NREM 3 n Average delta -&t~l SD Range: Delta Adjusted means: Delta Average mmutes of awake INREM 31 Range: Minutes of awake Correlation between sqrt of awake & sqrt of average delta Correlation between age & sqrt of average delta Note.
ANCOVA
analysis
1. Correlation
slgnlficant
at p
0.01.
2
slgnlflcant
at p
0.05
Correlatlon
32 10.2 2.9-21 .a 0.4-41.9 11.4
0.4 o-29
0.4 O-27
-0.27
-0.03
-0.491
-0.22
52 15.5 7.3-26.8 2.5-49.3 14.7
32 12.6 5.0-23.9 1 .l-45.4 13.8
1 .o o-29
1.6 O-30
-0.421
-0.392
-0.332
-0.33
52 a.4 3.0-16.4 1.5-50.7 a.2
32 a.1 3.8-13.9 0.7-21.6 a.4
1.3
o-a
1.3 o-37
-0.29
0.04
-0.13
of covar~ance,
Depressive
NREM
of awake time as
Probability Equality of: Adjusted means 0.0008 Slopes 0.3353 Zero slopes. 0.0008 [age & awake)
Equality of: Adjusted means 0.6662 Slopes 0.8839 Zero slopes: iO.0001 (age & awake)
Equality of: Adjusted means 0.9143 Slopes 0.2079 Zero slopes: 0.2110 ‘age & awake’
-0.18 non-rapld-eye-movement
sleep.
sqrt
square
root
301 slope (given that the correlations in the two groups do not correlation between awake time and delta, or age and delta’!). ANCOVA was also used to examine whether there were significant linear or curvilinear trends across the N REM periods for average delta in either the normal or the depressed subjects. depressives’!);
and (3) the test for Let-0
differ, is there in general a significant
Results Our particular interest was focused in the first NREM period, where the ma.jor differences between the slow-wave sleep of normals and depressed patients are traditionally found. No significant correlation between awake time in the first NREM period and average delta (Y = -0.03, NS, depressives, n = 32; Y= -0.27, NS, normals, n = 52) was found (Table I; Figs. I and 2). Even though there was no significant correlation for awake and delta in the first NREM period, there was still a significant difference in the amount of average delta in the depressives and normals (p < 0.0001) (Table I). Further, there was a representative range of awake time for each group (O-27 minutes, depressives; O-29 minutes, normals (Table I). In the second NREM period, there was a significant correlation between awake time and delta, but no difference in the means of the average delta between normals and depressives. The BMDP program (Dixon et al., 1983) P2V was used to test for linear and nonlinear trends within the normals and the depressives across the three NREM periods. The normals showed a significant linear trend (JJ < 0.000 l), but no significant quadratic term (JI = 0.0562). This result correlates with the decreasing means across the NREM periods (21.2, 15.5, and 8.4) (Table 1). The depressives showed a different Fig. 1. Average delta for depressed
subjects in NREM 1
I-
l-
0 L-
0
0
l-
Q
0
2-
i!
0
0 0 0
0.5
1
1.5
7.
1.5
5
Minutes of Awako NREM
1
non-rapld-eye-movement
sleep
period
#l
3.5
4
(sqrt)
4.5
5
5.5
6
302 Fig. 2. Average delta for normal controls in NREM 1 IX 7-
X
:Iii:;.; 4
z-
x
1
X
X
X
i 0’
I
0
I
0.5
I
1
1.5
2
1.5
3
3.5
4
4.5
5
5.5
s
Minutes of Awako (sq t-t) NREM
1
non-rapld-eye-movement
sleep
period
Ul
trend. with the linear term being nonsignificant (/I = 0.2392) and the quadratic term being highly significant (J) = 0.0045). This corresponds to the obser\,ed means ( 10.2. 12.6, and X.1) (Table I). Again, awake time and age were used as covariates. Awake time for the night as a whole showed a wide range in both depressed ( l-164 minutes) and normal subjects (O-84 minutes) (Fig. 3). ANCOVA, with age and aivakc time as covariates. examined differences between the two groups. Awake time was limited to 90 minutes to ensure comparable ranges between the two gr-oups. No difference was found between dcpressikcs and controls in average delta for the whole night (11 0.1 160) after age and awake time were co\aried (I’able 2). q
Discussion Neither normal controls nor depressed patients provided any evidence that awake time explains differences in slow-wave sleep between normals and depressives, a conclusion reached earlier in a small sample of depressed patients and normal controls (Horbkly et al.. 1984). Further. differences in the first NREM period in slow-wave sleep bet\veen normals and depressed subjects persisted even when allowance was made for time spent awake. We used covariance techniques to ascertain the effect of awake time on delta sleep patterning throughout thenight in normalanddepressed subjects. In bothgroups. wakefulness seemed to havevery littleimpact on the first NREM period but asignificant effect on the second NREM period. This is certainly meaningful in thedepressed group since the hypothesis of reduced slow-wave sleep or Process S in depression has been predicated on this finding not simply being secondag to increased wakefulness.
303
Fig. 3. Distribution control subjects
of awake time in 38 depressed
and 52 normal
Logend I72 DURESSIVES 0
Minutor of Awake
Table 2. Average delta count for the whole night Minutes of awake:
o-5
6-30
31-90
Depressives n Mean iSD
11 11.2 6.3-19.9
14 9.1 4.6-18.0
IO 6.7 3.2-14.0
18 13.6 6.8-27.3
25 11.7 6.7-20.3
9 10.3 6.3-16.8
Normals n Mean LSD
Nore. The square root transformation was used in the analysis. and the means and standard deviations - 8 + 1 SD1 have been retransformed in th1.s table.
NORMALS
304
Correcting forawake time reduces the likelihood that changes in Process S in depression are secondary to wakefulness or arousal. A strong linear trend in the normals and a strong quadratic trend in thedepressivcsacrossthenightwhenageandawakctimewerc corrected for also suggests there may be inherent differences in Process S between the two groups Further investigations on this issue are needed to demonstrate conclusively the “minimal” role of wakefulness on the reduced slow-wave sleep in depressed patients. One strategy would be to examine a group of anergic depressed patients who tend to demonstrate no sleep maintenance difficulty, but who have in the past been reported to show reductions in slow-wave sleep comparable to depressed individuals with major sleep continuity disruption. To some extent, this issue has been addressed in the present sample since almost one-half ofthedepressedgroup had<10mihutes ofwakefulness. A second strategy would be to compare another group of psychiatric patients with reported deficiencies in slow-wave sleep. Patients with schirophrenia could be studied to address the “specificity” issue and also the contribution that wakefulness makes to deficiencies in slow-wave sleep. Intensive examinations of both of these strategies arc currently underway in our laboratory. Acknowledgments. This work was suppor-ted. in part, by N I M H grants 3OYI5 and 24652, and also by a grant from the John D. and Catherine I‘. MacArthur Research Network on the Psychobiology of Depression.
References Borbcly,
A.A. A two-process
model of sleep regulation. Hurr?urr Nrurohiolog.\~, 1, 195 (19X2). I.. Leopfe. M., Kupfer, D.J.. Ulrich. R.F.. Grochocinski. V.J.. Doman, J.. and Matthews, G. All-night spectral analysis of the sleep EEG in untreated depressives and normal controls. P.\~~chiutt:1. Rr.\eurc~h, 12, 27 (1984). L>aan. S.. Beersma. D.G.. and Borbely, ,4.A. I iming of human sleep: Recovery process gatcd by a circadian pacemaker. Anwricun Journul of’ P/7_v.\io/og~., 246, R 16 I ( 1984). Dixon. W.J.. Brown. M.H., Engelman. L.. Franc, J.W.. Hill, M.A.. Jennrick. R.I., and Toporek. J.D.. eds. BML>P .Sfuri.v/ic,u/ Sq/ib\.urc. University of California Press. Berkeley (10X3). Endicott. J.. and Spitrcr. R.l.. A diagnostic interview: Schedule for AfTective Disorders and Schizophrenia. Archirw o/. C;rnrwl P.s~l7iatr.1~. 35, X37 ( 197X). Hamilton, M. A rating scale for depression. Jolcrrrul O/ ,Yruro/oX~~, :Veuro.\ur~erl’ ur~tl Ps,d7iar,lr~, 23, 56 ( 1960). Kupfcr. D.J.. Shaw, D.H., Ulrich. R.J.., Coblc. P.A., and Spiker-. L1.G. Application ol automated delta analysis. .S/PP/I Re.wurc~h, 10, I73 (19X1). Kupfer, D.J., Shaw, D.H.. Ulrich. R.F., C‘oble, P.A., and Spiker. D.G. Application of automated REM analysis in depression. Archives o/. Grnrrul P.ql.chiutr~‘, 39, 569 (1982). Kupfer. D.J., Ulrich, K.F.. Cable. P.A., Jarrett. D.B.. Grochocinski. V.J.. Doman, J.. Matthews. G., and Horbely. A.A. Application ot’automated REM and slow wave sleep analysis: II. I esting the assumptions of the two-process model of sleep regulation in normal and dcprcsscd subjects. P.swhiurt:l, RmwrcI~. 13, 335 ( 1984). Mcl’artland. R.J.. Kupfer. D.J.. Cable. P.A., Spiker-. D.ti.. and Matthews. G. REM sleep in primary depression: A computerized analysis. ~/C,c,/roc,rtc.L/tu/(~~ru/)h.l, uncl C‘/ini(.u/ ,Yclrropll~~.\iolo,!g, 44, 5 I3 ( 197X). Spitter. K.L.. Endicott. J.. and Robins. E. Rcscarch Diagnostic C‘ritcria: Kationalc and reliability. if,-clrir,c,v of’ GetwruI P.\j’.c,l7iutr,\., 35, 773 ( 197X).
BorMly, A.A., Tobler.
Research.
297
19, 297-304
Elsevier
Relationship
of Awakening
David
Victoria
J. Kupfer,
Received
December
6. IY8.5;
and Delta Sleep in Depression
J. Grochocinski, revised
version
rec,eived
and Ann May
6. McEachran
5, /%‘6;
accepted Ma_\,29, Iys6.
Abstract. It has been demonstrated that deficiencies in slow-wave sleep occur in the sleep profiles of depressed patients. Recent theories of sleep regulation link a deficiency in the so-called “Process S” to these slow-wave sleep alterations. However, the degree of wakening during sleep has been suggested as the explanation for reduced slow-wave sleep. In the present study, the extent of this relationship was examined in both depressed patients and normal subjects. Only a relatively low level of correlation between the degree of wakening and reduction in slow-wave sleep was noted in depressed patients. This finding on 38 unipolar depressed patients is similar to the findings on a smaller sample of depressed individuals. Key Words.
Delta sleep, awakening,
depression,
sleep regulation.
Abnormalities in a two-process model of sleep regulation (a sleep-dependent process termed “Process S,” and a sleep-independent circadian process termed “Process C”)
have been proposed to account for sleep abnormalities in depressive states (Daan et al., 1984). The major tenets of the two-process model of sleep regulation (BorbCly, 1982) as applied to depression are: the level of Process S, as reflected by electroencephalographic (EEG) slow-wave activity, corresponds to the sleep-dependent facet of sleep propensity; the pathognomic changes of sleep in depressives are a consequence of a deficiency in the buildup of Process S. The application of automated delta-wave analyses in normal subjects and younger depressed patients supports the model to some extent (Kupfer et al., 1984). Time spent asleep is positively correlated with total delta waves (normals and depressives) and average delta waves (depressives); delta sleep is lower in depressives than in normals; the average delta wave count is significanty reduced in younger depressives over the total night and in NREM (non-rapideye-movement) period 1. The model also postulates that measures of phasic REM activity are inversely related to Process S, suggesting that Process S exerts an inhibitory influence on phasic REM activity. Support for this hypothesis has also been aided by studies demonstrating comparability of results using spectral analysis and period analysis (with a delta analyzer). In one study, sleep was recorded in nine drug-free depressive patients and in age- and sex-matched normal control subjects. All-night spectral analysis of the sleep EEG
David J. Kupfer. M.D., is Professor and Chairman; Victoria J. Grochocinski, Ph.D.. is Data Base Manager; and Ann B. McEachran, M.S., is Statistician, Department of Psychiatry, University of Pittsburgh School of Medicine. (Reprint requests to Dr. D.J. Kupfer, Western Psychiatric Institute and Clinic, 3811 O’Hara St., Pittsburgh. PA 15213, USA.) 0165-1781/86/$03.50
@ 1986 Elsevier Science Publishers B.V.
298 showed a significantly reduced power density in the 0.25-25.0 Hr band in the depressive group. Power density values integrated over the entire frequency range (0.25-25.0 Hz) exhibited a decreasing trend over the first three NREM sleep cycles for both groups. In each sleep cycle, depressives had lower values than controls with both types of analysis. The results are consistent with the hypothesis that a buildup of a sleepdependent process is deficient in the sleep regulation of depressive patients (BorbCly et al., 1984). While there is little disagreement about the finding of decreased slow-wave sleep in depressed
patients,
there
is considerable
controversy
about
its specificity.
For exam-
ple, it has been strongly suggested that the amount of awakening found in the sleep of depressed patients, particularly in the NREM periods. is responsible for the decreased delta sleep “production.” The Process S model (which does not make an awake time prediction) itself has been challenged on the basis that if decreased slow-wave sleep is strongly associated with awakening, then the model is not specific to depression. To examine this question, we analyzed the relationship between awakening time and average delta-wave sleep in normal controls and in depressed patients. Methods ‘l‘he depressed group comprised 38 inpatients on the Clinical Research Unit (CRU) at Western Psychiatric Institute and Clinic (WPIC). At admission, in addition to a psychiatric interview and a physical examination, we obtained collateral information from the patient’s family and from case records of previous hospitalirations. During a 2-week drug-free period, patients underwent a series of routine laboratory tests, including thyroid function tests, a Ih-channel EEG. and other tests as indicated by history or physical examination. During the drug-free period, the Schedule for Affective Disorders and Schirophrenia (SADS) (Endicott and Spitrer, 1978) was completed by the patient’s psychiatrist. This structured research interview was used to make diagnoses according to the Research Diagnostic Criteria (Spitter et al., 197X). Diagnoses were determined using information obtained from the initial interview, case record, collateral history from relatives, and observation on the CRU. The patient entered the antidepressant drug protocol if illness severity remained sufficiently high at the end of the 2-week, drug-free period, i.e., if the patient had a minimum score of 30 (sum of two raters) on the l7-item Hamilton Rating Scale for Depression (HRSD) (Hamilton. 1960). The HRSD was used twice weekly throughout the investigation to evaluate severity of depression. The overall sleep-scoring techniques have been previously described (Kupfer et al.. 1982); all records were scored independently and without knowledge of the clinical diagnosis and clinical course for specific patients. All-night sleep recordings were made using a Grass (model 78R) polygraph and were simultaneously recorded on l-inch magnetic tapes (Honeywell recorder, model 101). REM and Delta Sleep Analyses. Automated measures were derived from software pr-ograms developed at the University of Pittsburgh and descrihcd previously (Mcl’artland ct al.. 1978). Briefly. these programs arc implemented on a PDI’ I I /44 minicomputer using the LPS-I I laboratory peripheral system. Compressed analog tapes of each subject’s sleep record are played directly into an analog-to-digital converter; the program yields both a minute-by-minute printout and a disk file. The disk file is later matched on a DEC-IO system computer to the visually scored minute-b!,-minute Irecord so that nonsleep minutes. as well as movement artifact, can be accounted for and removed. Our data analyzer program has two main sections: Section I. written in Fortran. provides the operator interaction for program startup and all disk I; 0 functions: Section 2, written in PDP-I I assembly language. provides the real-time data-collection functions (Kupfer ct al.. 1981). Automated
299
The compressed analog tape recordings of the patients’ EECi waveforms are played directly into the analog-to-digital converter of the LPS-I I at a speed I6 times that of the recording speed, allowing an &hour period to be processed in only 30 minutes. The EEG channel is monitored, and data are collected at a real-time rate of 200 samples per second. Each sample is stored in a buffer until 200 samples have been collected. The average of these stored samples is computed and then subtracted from the oldest sample in the buffer. (The effect is to filter the signal digitally with a high-pass characteristic and a roll-off frequency of 0.1 HI, thereby eliminating problems associated with baseline drift.) The standard delineation of delta waves indicative of stage 3 and stage 4 sleep is 0.5-2.0 Hr at 75-200 mV. The adjusted samples (from the buffer) go directly to a baseline crossing detector and then to an amplitude (peak) detector. With each baseline crossing, the period of the preceding half wave is computed from the prior crossing, and the peak is noted and checked. The program normally runs between two calibration marks (which the program detects at the start and end of a record) and provides both a minute-by-minute printout and a disk file. While night counts can then be examined, or the average delta counts per minute analyred. correlations with visually scored sleep can be made for each minute of the night during NREM or REM periods or on any other subset ofthe EEG sleep recording. For automated analyses, each REM period contains a minimum of 5 minutes, although the REM onset time is derived in the usual fashion in which the first REM period needs to be only 3 minutes in length. For this study, we used baseline night 4 for depressed subjects and night 2 for normal controls. Subject Characteristics: Patients. Of the 38 (29 female, 9 male) nondelusional patients with a diagnosis of major depressive syndrome, 3 were bipolar. In addition, IX patients showed psychomotor retardation. I I had psychomotor agitation, and 6 were both agitated and retarded. Patients had an age range of 24-61 years. with a mean age of 41.6 (SD = 10.6) years. The mean initial H RSD score was 40.6 (SD q 9.4) (sum of two raters). The median length of the current episode was 30.0 weeks and the median number of previous episodes was 2.0. The mean REM latency for these subjects was 39.6 (SD = 28.5). Six of the 38 patients experienced a first NREM period shorter than 5 minutes. To ensure reliability. the data for these patients are not included in Table I. Characteristics: Normal Control Subjects. Fifty-tw*o (32 female, 20 male) normal control subjects. with an average age of 34.6 (SD q I I .5) years (age range IX-63 years) formed the comparison group. The mean REM latency for the group was 69.8 (SD = 21.8). Selection criteria demanded that all individuals have a normal sleep-wake cycle; for premenopausal women, the 3 consecutive night studies were in the folhcular phase of their menstrual cycle. Exclusion criteria were as follows: normals with obesity were ruled out, as were dieters who showed significant weight loss: normals could not be on birth control pills or lactating, currently or within the last 6 months. Subjects were also excluded if they had any current medical illness requiring pharmacological intervention, with the exception of well-controlled thyroid or hypertensive disease. Any abnormal physical examination features or laboratory data excluded individuals. Subjects were excluded if they, or their first-degree relatives, had a history of seizure disorder or major psychiatric illness. All individuals were required to be free of drugs at least 2 weeks before the biological studies began and had to have an established regular sleep-wake cycle with a “good-morning” time by 8 a.m. During the drug-free period, all drugs and alcohol were eliminated, with the exception of Tylenol or small doses of aspirin. Sleep investigations were then carried out on the 13th Floor Sleep Laboratory for 3 consecutive nights with the analyses of night 2 being used for this particular investigation. Subject
Statistical analysis involved parametric correlation and analysis of covariance (ANCOVA) techniques. Correlation analysis was used to examine the relationship between minutes of awakening and average delta waves within each of the two groups (depressed and controls). as well as across the two groups as a whole. ANCOVA was used to compare within each NREM period the amount of average delta wavjes in the depressed and control subjects, Statistics.
300 controlling for age and minutes 01 awakenmg. I he square root ot awake and of average delta was used in all analyses to compensate for heteroscedasticity and skewness. Table I contains the results of the ANCOVA, with age and square root of minutes of awake time as the covariates. The last column contains the probabilities associated with (I) the adjusted means (after adjustments for age and awake time, are the means different?); (2) the slopes (is there a difference in the correlations between awake time and delta in the controls and
Table 1. Results covariates
of ANCOVA,
with age and minutes
Normal NREM 1 n Average delta -&i 1 SD Range: Delta Adjusted means: Delta Average minutes of awake INREM 1, Range. Minutes of awake Correlation between sqrt of awake & sqrt of average delta Correlation between age & sqrt of average delta
52 21.2 10.7-35.4 1 .o-54.3 20.2
NREM 2 ” Average delta -&+lSD Range: Delta Adjusted means: Delta Average minutes of awake INREM 21 Range: Minutes of awake Correlation between sqrt of awake & sqrt of average delta Correlation between age & sqrt of average delta NREM 3 n Average delta -&t~l SD Range: Delta Adjusted means: Delta Average mmutes of awake INREM 31 Range: Minutes of awake Correlation between sqrt of awake & sqrt of average delta Correlation between age & sqrt of average delta Note.
ANCOVA
analysis
1. Correlation
slgnlficant
at p
0.01.
2
slgnlflcant
at p
0.05
Correlatlon
32 10.2 2.9-21 .a 0.4-41.9 11.4
0.4 o-29
0.4 O-27
-0.27
-0.03
-0.491
-0.22
52 15.5 7.3-26.8 2.5-49.3 14.7
32 12.6 5.0-23.9 1 .l-45.4 13.8
1 .o o-29
1.6 O-30
-0.421
-0.392
-0.332
-0.33
52 a.4 3.0-16.4 1.5-50.7 a.2
32 a.1 3.8-13.9 0.7-21.6 a.4
1.3
o-a
1.3 o-37
-0.29
0.04
-0.13
of covar~ance,
Depressive
NREM
of awake time as
Probability Equality of: Adjusted means 0.0008 Slopes 0.3353 Zero slopes. 0.0008 [age & awake)
Equality of: Adjusted means 0.6662 Slopes 0.8839 Zero slopes: iO.0001 (age & awake)
Equality of: Adjusted means 0.9143 Slopes 0.2079 Zero slopes: 0.2110 ‘age & awake’
-0.18 non-rapld-eye-movement
sleep.
sqrt
square
root
301 slope (given that the correlations in the two groups do not correlation between awake time and delta, or age and delta’!). ANCOVA was also used to examine whether there were significant linear or curvilinear trends across the N REM periods for average delta in either the normal or the depressed subjects. depressives’!);
and (3) the test for Let-0
differ, is there in general a significant
Results Our particular interest was focused in the first NREM period, where the ma.jor differences between the slow-wave sleep of normals and depressed patients are traditionally found. No significant correlation between awake time in the first NREM period and average delta (Y = -0.03, NS, depressives, n = 32; Y= -0.27, NS, normals, n = 52) was found (Table I; Figs. I and 2). Even though there was no significant correlation for awake and delta in the first NREM period, there was still a significant difference in the amount of average delta in the depressives and normals (p < 0.0001) (Table I). Further, there was a representative range of awake time for each group (O-27 minutes, depressives; O-29 minutes, normals (Table I). In the second NREM period, there was a significant correlation between awake time and delta, but no difference in the means of the average delta between normals and depressives. The BMDP program (Dixon et al., 1983) P2V was used to test for linear and nonlinear trends within the normals and the depressives across the three NREM periods. The normals showed a significant linear trend (JJ < 0.000 l), but no significant quadratic term (JI = 0.0562). This result correlates with the decreasing means across the NREM periods (21.2, 15.5, and 8.4) (Table 1). The depressives showed a different Fig. 1. Average delta for depressed
subjects in NREM 1
I-
l-
0 L-
0
0
l-
Q
0
2-
i!
0
0 0 0
0.5
1
1.5
7.
1.5
5
Minutes of Awako NREM
1
non-rapld-eye-movement
sleep
period
#l
3.5
4
(sqrt)
4.5
5
5.5
6
302 Fig. 2. Average delta for normal controls in NREM 1 IX 7-
X
:Iii:;.; 4
z-
x
1
X
X
X
i 0’
I
0
I
0.5
I
1
1.5
2
1.5
3
3.5
4
4.5
5
5.5
s
Minutes of Awako (sq t-t) NREM
1
non-rapld-eye-movement
sleep
period
Ul
trend. with the linear term being nonsignificant (/I = 0.2392) and the quadratic term being highly significant (J) = 0.0045). This corresponds to the obser\,ed means ( 10.2. 12.6, and X.1) (Table I). Again, awake time and age were used as covariates. Awake time for the night as a whole showed a wide range in both depressed ( l-164 minutes) and normal subjects (O-84 minutes) (Fig. 3). ANCOVA, with age and aivakc time as covariates. examined differences between the two groups. Awake time was limited to 90 minutes to ensure comparable ranges between the two gr-oups. No difference was found between dcpressikcs and controls in average delta for the whole night (11 0.1 160) after age and awake time were co\aried (I’able 2). q
Discussion Neither normal controls nor depressed patients provided any evidence that awake time explains differences in slow-wave sleep between normals and depressives, a conclusion reached earlier in a small sample of depressed patients and normal controls (Horbkly et al.. 1984). Further. differences in the first NREM period in slow-wave sleep bet\veen normals and depressed subjects persisted even when allowance was made for time spent awake. We used covariance techniques to ascertain the effect of awake time on delta sleep patterning throughout thenight in normalanddepressed subjects. In bothgroups. wakefulness seemed to havevery littleimpact on the first NREM period but asignificant effect on the second NREM period. This is certainly meaningful in thedepressed group since the hypothesis of reduced slow-wave sleep or Process S in depression has been predicated on this finding not simply being secondag to increased wakefulness.
303
Fig. 3. Distribution control subjects
of awake time in 38 depressed
and 52 normal
Logend I72 DURESSIVES 0
Minutor of Awake
Table 2. Average delta count for the whole night Minutes of awake:
o-5
6-30
31-90
Depressives n Mean iSD
11 11.2 6.3-19.9
14 9.1 4.6-18.0
IO 6.7 3.2-14.0
18 13.6 6.8-27.3
25 11.7 6.7-20.3
9 10.3 6.3-16.8
Normals n Mean LSD
Nore. The square root transformation was used in the analysis. and the means and standard deviations - 8 + 1 SD1 have been retransformed in th1.s table.
NORMALS
304
Correcting forawake time reduces the likelihood that changes in Process S in depression are secondary to wakefulness or arousal. A strong linear trend in the normals and a strong quadratic trend in thedepressivcsacrossthenightwhenageandawakctimewerc corrected for also suggests there may be inherent differences in Process S between the two groups Further investigations on this issue are needed to demonstrate conclusively the “minimal” role of wakefulness on the reduced slow-wave sleep in depressed patients. One strategy would be to examine a group of anergic depressed patients who tend to demonstrate no sleep maintenance difficulty, but who have in the past been reported to show reductions in slow-wave sleep comparable to depressed individuals with major sleep continuity disruption. To some extent, this issue has been addressed in the present sample since almost one-half ofthedepressedgroup had<10mihutes ofwakefulness. A second strategy would be to compare another group of psychiatric patients with reported deficiencies in slow-wave sleep. Patients with schirophrenia could be studied to address the “specificity” issue and also the contribution that wakefulness makes to deficiencies in slow-wave sleep. Intensive examinations of both of these strategies arc currently underway in our laboratory. Acknowledgments. This work was suppor-ted. in part, by N I M H grants 3OYI5 and 24652, and also by a grant from the John D. and Catherine I‘. MacArthur Research Network on the Psychobiology of Depression.
References Borbcly,
A.A. A two-process
model of sleep regulation. Hurr?urr Nrurohiolog.\~, 1, 195 (19X2). I.. Leopfe. M., Kupfer, D.J.. Ulrich. R.F.. Grochocinski. V.J.. Doman, J.. and Matthews, G. All-night spectral analysis of the sleep EEG in untreated depressives and normal controls. P.\~~chiutt:1. Rr.\eurc~h, 12, 27 (1984). L>aan. S.. Beersma. D.G.. and Borbely, ,4.A. I iming of human sleep: Recovery process gatcd by a circadian pacemaker. Anwricun Journul of’ P/7_v.\io/og~., 246, R 16 I ( 1984). Dixon. W.J.. Brown. M.H., Engelman. L.. Franc, J.W.. Hill, M.A.. Jennrick. R.I., and Toporek. J.D.. eds. BML>P .Sfuri.v/ic,u/ Sq/ib\.urc. University of California Press. Berkeley (10X3). Endicott. J.. and Spitrcr. R.l.. A diagnostic interview: Schedule for AfTective Disorders and Schizophrenia. Archirw o/. C;rnrwl P.s~l7iatr.1~. 35, X37 ( 197X). Hamilton, M. A rating scale for depression. Jolcrrrul O/ ,Yruro/oX~~, :Veuro.\ur~erl’ ur~tl Ps,d7iar,lr~, 23, 56 ( 1960). Kupfcr. D.J.. Shaw, D.H., Ulrich. R.J.., Coblc. P.A., and Spiker-. L1.G. Application ol automated delta analysis. .S/PP/I Re.wurc~h, 10, I73 (19X1). Kupfer, D.J., Shaw, D.H.. Ulrich. R.F., C‘oble, P.A., and Spiker. D.G. Application of automated REM analysis in depression. Archives o/. Grnrrul P.ql.chiutr~‘, 39, 569 (1982). Kupfer. D.J., Ulrich, K.F.. Cable. P.A., Jarrett. D.B.. Grochocinski. V.J.. Doman, J.. Matthews. G., and Horbely. A.A. Application ot’automated REM and slow wave sleep analysis: II. I esting the assumptions of the two-process model of sleep regulation in normal and dcprcsscd subjects. P.swhiurt:l, RmwrcI~. 13, 335 ( 1984). Mcl’artland. R.J.. Kupfer. D.J.. Cable. P.A., Spiker-. D.ti.. and Matthews. G. REM sleep in primary depression: A computerized analysis. ~/C,c,/roc,rtc.L/tu/(~~ru/)h.l, uncl C‘/ini(.u/ ,Yclrropll~~.\iolo,!g, 44, 5 I3 ( 197X). Spitter. K.L.. Endicott. J.. and Robins. E. Rcscarch Diagnostic C‘ritcria: Kationalc and reliability. if,-clrir,c,v of’ GetwruI P.\j’.c,l7iutr,\., 35, 773 ( 197X).
BorMly, A.A., Tobler.