- Email: [email protected]
Early return to REM sleep after nocturnal awakening in depression
BIOL PSYCHIATRY 1992;31:171-176
171
Early Return to REM Sleep after Nocturnal Awakening in Depression Michael E. Thase, Charles F. Reynolds III, Ellen Frank, J. Richard Jennings, Gregory L. Garamoni, Amy L. Yeager, and David J. Kupfer
Sixteen male outpatients with major depression and 20 age-equated healthy controls were awakened from rapid eye movement (REM) sleep between 1:30 and 3:30 AM, and the rapidity of return to REM sleep was determined. The time it took to return to REM sleep was reduced in depressives compared with controls: 6•.6 (17.9 SD) min versus 80.6 (24.9 SD) rain, respectively (p = 0.01). The time elapsed until the return to REM sleep was significantly correlated with baseline REM latency in controls (but not depressives). In contrast, return to REM time was significantly correlated with depression severity scores in depressives (but not controls). There was no evidence to support the hypothesis that the more rapid return to REId sleep in depression was caused by a slow wave sleep deficit. The mechanism underlying the rapid return of REM sleep in depression thus may be related to a severity-linked disturbance, such as a proposed increase in REM "pressure."
Introduction Reduced latency to the onset of ~,hefirst rapid eye movement sleep period (REM latency) is a well-described feature of primary or endogenous depressions: approximately 50% of depressed outpatients and up to 80% of inpatients manifest a consistent reduction of REM latency (Thase and Kupfer 1987). However, despite the prevalence of this psychobiologic abnormality, it remains unclear whether reduced REM latency in depression is due to an active, state-dependent process (such as increased REM '~pressure") (Vogel 1983) or reflects a more indirect consequence of diminished slow-wave sleep (Kupfer and Ehlers 1989). As severe depressions are often associated with both increased REM indexes and diminished slow wave sleep (Thase and Kupfer 1987), alternate experimental methods are necessary to unravel these competing hypotheses. One possible method has been described by Schulz and Tetzloff (1982): depressed patients are awakened at 2:30 AM, and the latency (or elapsed time) until the return of REM sleep is determined. This method is novel in that it may obviate potential confounds associated with difficulty initiating
From the Western Psychiatric Institute and Clinic, Department of Psychiatry, University of Pittsburgh School of Medicine, Pittsburgh, PA. Address reprint requests to Dr. M. E. Thase, Western Psychiatric Institute and Clinic, Department of Psychiatry, University of Pittsburgh School of Medicine, 3811 O'Hara St., Pittsburgh, PA 15213. Received February 28, 1991; revised July 24, 1991. © 1992 Society of Biological Psychiatry
0006-3223/92/$05.00
172
BIOL PSYCHIATRY 1992:31:171-176
M.E. Thase et al
sleep and control for possible relationships between sleep onset and clock time. Moreover, the 2:30 AM awakening occurs at a period in the normal sleep cycle in which a majority of slow wave sleep (SWS) already has occurred, thus partially accommodating for potential reductions in SWS observed during the first third of the night in some depressed patients. Using this paradigm, Schulz and Tetzloff (1982) found a reduced latency to the return of REM sleep during both illness and remission in depressed patients with sleep onset REM periods (SOREMPs). The early return of REM sleep observed during remission (i.e., following resolution of the depressive state) is suggestive of a trait-like characteristic, and therefore may reflect a more "passive" process linked to diminished SWS rather than a severity-linked or state-dependent phenomenon. We have had the opportunity to replicate and extend Schulz and Tetzloff's (1982) experimental awakening paradigm in our ongoing duties of REM sleep and nocturnal penile tumescence (NPT) in depression (see, e.g., Thase et al 1988). In this protocol, patients were periodically awakened during the third night of electroencephalographic (EEG) sleep recording for determinations of penile buckling force. Because a number of these awakenings coincided with the 2:30 AM time employed by Schulz and Tetzloff, we compared the return times to REM sleep in these patients against those of matched normal controls studied usi,ig the same methodology. Further, as our sample consisted of less severely depressed outpatients, with relatively normal SWS and without SOREMPs (Thase et al 1988), we were able to extend this method to a sample of patients with both milder depressive symptomatology and less severe sleep disturbances.
Method Our methods for recruitment and assessment of depressed patients and normal controls for participation in the NIT protocol have been reported previously, as have the main findings of this project (Thase et al 1988). To summarize briefly, depressed men between the ages of 20 and 60 presenting for treatment at the Western Psychiatric Institute and Clinic who met both clinical (American Psychiatric Association 1980) and research criteria [Schedule for Affecti ve Disorders and Schizophrenia (SADS) (Endicott and Spitzer 1978 ); Research Diagnostic Criteria (RDC) (Spitzer et al 1978)] for a diagnosis of major depressive disorder were eligible to enroll in the study. Controls were recruited from media advertisements and hospital staff. Potential control subjects were interviewed to exclude individuals with diagnosable psychopathology using the lifetime version of the SADS (Spitzer et al 1975). All subjects were further screened with a comprehensive physical examination and appropriate laboratory studies to rule out medical conditions that might either cause depression or erectile dysfunction, or require treatment with medications that would invalidate EEG sleep/NIT studies. A battery of clinical assessment measures that included the Hamilton Rating Scale for Depression (HRSD) (Hamilton 1960) and a selfreport c,f depressive symptoms derived from the Brief Symptom Inventory (BSI) (Derogatis 1975) also was completed. All subjects provided written informed consent for research participation and underwent a supervised 14-day drug-free ahd alcohol-free interval. Following the washout period, subjects underwent a 3-night EEG sleep/NPT protoc~ J. During the third night of protocol (but not during nights I and 2), subjects were awaken~:d during each episode of NPT for visual estimates of penile rigidity and measuremert of buckling force (see Thase et al 1988). The~e procedures generally took less than 3 ~ain to complete, and during this time the techn~,logist spoke with the subjects to
Post-Arousal REM Sleep in Depression
BIOL PSYCHIATRY 1992;31:171-176
173
confirm their awakening. Thereafter, subjects were permitted to return to sleep immediately. The sleep EEG/NPT polygraphic records of 24 consecutively studied depressed men and 24 healthy controls were initially matched on the basis of age ( _+ 5 years); selection of matched pairs for possible inclusion in this analysis was determined by the availability of the original sleep record for manual re-scoring. The availability of records was directly related to the recency of the recordings; we do not store polygraphic records for longer than 12-18 months. Among these 48 potential cases, a total of 16 depressed and 20 control subjects met the following criteria: (1) they were awakened within 60 min (_+) of the 2:30 AM time utilized by Schulz and Tetzloff (1982), and (2) they were awakened from an REM period. This latter modification served to equate the patients with respect to both clock time and position within the REM-NREM cycle. Demographic and clinical characteristics of the same of 36 subjects are summarized in Table 1, as are selected baseline (night 2) EEG sleep characteristics. We used night 2 for baseline EEG sleep recordings, allowing night 1 for accommodation to the laboratory. The key dependent measure was min of sleep elapsed from the end of the experimental awakening until the onset of REM sleep (->3 min in duration). The prediction that more rapid return to REM sleep would occur in depression was tested by comparing the elapsed time values for the two groups with two-tailed t-tests. Pearson correlation coefficients between baseline REM latency (taken from night 2) and elapsed time until return to REM sleep (night 3) were computed, whereas nonparametric (Spearman rho) correlations were t~sed to assess the relationship between severity of depression (HRSD and BSI-D) versus baseline REM latency (from night 2) and time until return to REM sleep. (The nonparametric correlations were used for analyses of symptomatic measures because of differences in the distribution of scores on these measures in the depressed and normal control samples.) The proposed relationship of a more rapid return to REM sleep in patients with SWS deficits was examined by computing the number of minutes (and percent) of hand-scored SWS measured between sleep onset and the experimental awakening, and between the experimental awakening and return to REM sleep. We also performed parallel analyses in 33 of the subjects (15 depressed and 18 controls), for whom we were able to use more refined automated scoring methods for total delta wave counts and delta wave counts per min (i.e., delta waves in the 0.5-3.0 Hz range; Kupfer et al 1986): The hypothesis that early return to REM sleep in depression would be associated with a severity-linked process would be supported by an association with clinical measures of depression in the absence of, or independent of, a slow-wave sleep deficiency.
Results Baseline REM latency differed significantly between groups: depressed ~ = 56.6 (14.4) min versus control ~ = 92.9 (44.8) rain, p < 0.0001 (see Table 1). The samples did not differ with respect to either the mean time of the experimental awakening on night 3 [control it = 02:32 AM (0:32), depressed ~ = 02:28 AM (0:26); t = 0.47, df = 34, p = NS] or min of time elapsed from sleep onset until the awakening [control ~ = 183.9 min (35.9 min), depressed ~ - 169.3 min (56.7 min); t = 1.03, p = NS]. Among the depressed patients, l was awakened from REM period 1, 12 from REM period 2, and 3 from REM period 3. Among the controls, the proportions were as follows: 2 (REM 1
174
M.E. Thase et al
BIOL PSYCHIATRY 1992;31:!71-176
Table 1. Demographic, Clinical, and EEG Sleep Characteristics of Depressed and Healthy Control Subjects* Controls
Depressives
(n = 20)
(n = 16)
Demographics Age
Mean -+ SD Range Race (white/other) Education (yr) Marital status (married/partner/no partner) Number of episodes Duration of current episode (weeks) Median -- SD Range RDC subtypes
Clinical Hamilton (17-item) Brief Symptoms Inventory Depression Bseline EEG sleep (from night 2) Sleep efficiency (%) REM time (rain) REM (%) REM latency (rain) SWS (min)a SWS (%)"
Return to REM time (from night 3)
37.5 (11.6) 22-59 18/2 16.4 (2.8) 10/4/6
38.0 (11.4) 21-58 15/1 14.8 (2.9) 5/2/9
NS b NS c NS b NS c
2.3 (!.6) 31.2 (37.8) 4-150 12 unipolar, 4 bipolar 11 4 nonendogenous 8 endogenous 0.5 (1.2) 0.2 (0.3)
90.3 79.6 20.0 92.9 60. I 15.5 80.6
(6,0) (26.9) (6.2) (44.8) (45.5) (I !.6) (24.9)
18.9 (4.3) 1.4 (0.6)
86.5 81.9 21.4 56.6 36.9 9.4 61.6
(9.3) (19.5) (3.4) (14.4) (29.8) (7.3) (17.9)
0.001 d
0.001 e
NS b NS b NS b
o.0olf NS b NS b 0.01 ~
*Standarddeviations are in pmentheses. "Sleep stages 3 and 4. ~Oroupt-test, 'X' for contingency tables, dGroup t-test, t tGroup t-test, t fGroup t-test, t ~Groupt-test, t
= = = =
- 16.6, df - 17. I, p < 0.0001 (adjusted for unequal variances), -6.5, df = 14.2, p < 0.0001 (adjusted for unequal variances). 3.4, df = 23.7, p < 0.0001 (adjusted for unequal variances). 2.7, df = 34, p = 0.01.
period), 14 (REM 2 period), and 4 (REM 3 period). These proportions did not differ between the two groups (x = = 0.19, p = 0.91). The depressed and control subjects also did not differ with respect to the number of min of REM sleep elapsed in the episode they were awakened from (depressed ~ = 11.3 (2.9 SD) min versus controls ~ = 13.1 (6.3 SD) min, t = 1.1, df = 34, p = 0.31). However, despite these similarities in the timing of the awakening and its position within the REM-NREM cycle, latency to return to REM sleep was reduced in the depressed patients ~ = 61.6 (17.9 SD) mini compared with normal controls [~ = 80.6 (24.9 $D) rain, p = 0.OI] (see Table 1). Baseline REM latency was significantly correlated with the latency of return to REM sleep in controls (r = 0.54, p < 0.05), but not in the depressives (r = 0.18, p = NS). A comparison of the strength of these two correlations using Fisher's r to z score transformation method indicated that the two groups' values significantly differed from each other (z = 2.10, p = 0.04). Spearman rho correlations between the time to return to
Post-Arousal REM Sleep in Depression
BIOL PSYCHIATRY 1992;31:i71-176
175
REM sleep and measures of severity of depression revealed a significant association in the depressed sample (BSI-D: r = - 0 . 5 6 , p < 0.02; HRSD: r = - 0 . 6 1 , p < 0.01), but not in normal controls (BSI-D: r = 0.00, p = NS; HRSD: r = - 0 . 3 2 , p = NS). By contrast, baseline REM latency was not significantly correlated with these symptomatic measures in either depressed patients (HRSD: r = 0.03, p = NS; BSI-D: r = - 0 . 1 0 , p = NS) or normal controls (HRSD: r = - 0 . 2 2 , p = NS; BSI-D: r = 0.02, p = NS). Percentage of sleep time occupied by SWS stages 3 and 4 did not significantly differ between controls and depressives. This was true with respect to both SWS values from baseline (night 2) and analyses of the percentage of total SWS recorded from sleep onset until the experimental awakening on night 3 [control ~ = 67.4% (36.0%), depressed = 78.6% (32.3%); t = - 0 . 9 7 , df = 34, p = NS], as well as the SWS recorded between the 2:30 AM awakening and the return to REM sleep [controls ~ = 12.3% (17.7%), depressed ~ = 7.8% (15.4%); t = 0.81, df = 34, p = NS]. Similarly, neither average delta wave counts per min (cpm) nor total delta cpm differed significantly between groups. Thus, between sleep onset and the experimental awakening, depressive~ averaged 27.3 (14.1) cpm, whereas controls averaged 27.6 (17.1) cpm (t = 0.06, df = 31, p = NS). Similarly, during the interval between experimental awakening and return to REM sleep, depressives averaged 16.7 (13.8) cpm and controls averaged 17.1 (14.7) cpm ( t = 0.08, df = 31, p = N S ) . Discussion These findings replicate and extend the earlier work of Schulz and Tetzloff (1982). We have found the experimental awakening protocol to be applicable to an outpatient sample of depressed patients. This paradigm may offer additional advantages in future research on the relationship of REM sleep and depression by eliminating potential confounds associated with severe sleep onset difficulties and "missed" fivJt REM sleep periods. Moreover, it provides a novel method to assess REM sleep pressure somewhat independently from disturbances of slow wave sleep. It i~ intriguing that the return to REM sleep was significantly related to baseline REM latency in normal controls but not in the depressed patients. Though we had not anticipated this finding, it appears to be consistent with Kupfer and Ehlers' (1989) two-factor model of REM latency disturbance in depression. This model postulates the existence of at least two mechanisms underlying REM latency disturbances in depression. In type I disorders, which are hypothesized to predominate in less severely depressed patients, ItEM latency is proposed to reflect a stable, state-independent abnormality. As such, REM latency values in the 50-70 min range often persist despite remission of the depressive syndrome, suggesting either a "scar" or a trait-like phenomenon (Thase an0 Simons in press). By contrast, Kupfer and Ehlers (1989) propose a second (type 2) disturbance in which a process associated with marked depressive severity causes a state-dependent reduction of REM latency, with values often less than 40 min. The early return to REM sleep observed in the current ambulatory sample may thus be "driven" by the severity-linked (type 2) disturbance which is unrelated to the baseline REM latency (type 1) of our ambulatory sample. This would account for the significant correlations between selfreported and observer-rated measures of severity and return to REM time in the depressed patients (but not the controls), as well as the absence of a relationship between baseline REM latency and return to REM time in depressed patients relative to the unaffected controls. Of course, an alternate, more parsimonious (and admittedly skeptical) expla-
176
BIOL PSYCHIATRY 1992;31:!71-176
M.E. Thase et al
nation of these findings is that selectively restricted ranges of variability observed on the key measures (i.e., REM latency for depressives and symptom scores for controls) may have artificially produced these findings. Further study of Kupfer and Ehlers' (1989) twofactor model obviously is necessary. On the other hand, it is fairly clear that the significantly reduced latency of return to REM sleep observed in our study was not due to a deficit of slow wave sleep in this outpatient sample of depressed men. Indeed, the depressed and control groups did not differ with respect to either hand-scored or automated measures of SWS. The absence of an association between diminished slow wave sleep and an early return to REM sleep may provide further indirect support of an unidentified (as of yet) state-dependent abnormality related to the dysregulation of REM sleep in depression. We believe that this observation, coupled with the observed relationship between illness severity and time to return of REM sleep, support the alternate hypothesis that the REM sleep pressure may predominate during acute depressive illness. Restudy of our sample following remission from depression will be necessary to definitively ascertain whether such findings are related to a state-dependent process (i.e., severity-linked) or reflect a more enduring traitlike abnormality. This research was supported in part by Grants MH-30915 (Mental Health Clinic Reseat,:, Center), MH-40023 and MH-00295 (CFR) and MH-41884 (MET) from the National Institute of Mental Health. 'We wish to thank Ms. Lisa Stupar for her assistance in preparing this manuscript.
References American Psychiatric Association (1980): Diagnostic and Statistical Manual of Me~:~alDisorders, 3rd ed. Washington DC: American Psychiatric Press. Derogatis LR (1975): The brief symptom inventory. Baltimore, MD: Clinical Psychometrics. Endicott J, Spitzer RL (1978): A diagnostic interview: The scheduled tbr affective disorders and schizophrenia. Arch Gen Psychiatry 35:837-848. Hamilton M (1960): A rating scale for depression. J Neurol Neurosurg Psychiatry 12:56-62. Kupfer DJ, Ehlers CL (1989): Two roads to REM latency. Arch Gen Psychiatry 46:945-948. Kupfer DJ, Reynolds CF, Grochocinski VJ (1986): Comparison of automated REM and slow wave sleep analysis in young and middle-aged depressed subjects. Biol Psychiatry 21:189-200. Schulz H, Tetzioff W (1982): Distribution of REM latencies after sleep interruption in depressive patients and control subjects. Biol Psychiatry 17:1367-1376. Spitzer RL, Endicott J, Robins E (1975): Schedule for Affective Disorders and SchizophreniaLifetime Version. New York: New York State Psychiatric Institute, Biometrics Research. Spitzer RL, Endicott J, Robins E (1978): Research diagnostic criteria: Rationale and reliability. Arch Gen Psychiatry 35:773-782. Thase ME, Kupfer DJ (1987): Current status of EEG sleep in the assessment and treatment of depression. In Burrows GD, Werry JS (eds), Advances in Human Psychopharmacology, Vol 4. Greenwich, CT: JAI Press, pp 93-148. Thase ME, Simons AD (199~): The applied use of psychotherapy to study the psychobiology of depression. Psychother Practice Res (in press). Thase ME, Reynolds CF Ill, Jennings JR, et al (1988): Nocturnal penile tumescence is diminished in depression. Bi, I Psychiatry 24:33-46. Vogel GW (1983): Evidence for REM sleep deprivation as the mechanism of action of antidepressant drugs. Prog Neuropsychopharmacol Biol Psychiatry 7:343-349.
171
Early Return to REM Sleep after Nocturnal Awakening in Depression Michael E. Thase, Charles F. Reynolds III, Ellen Frank, J. Richard Jennings, Gregory L. Garamoni, Amy L. Yeager, and David J. Kupfer
Sixteen male outpatients with major depression and 20 age-equated healthy controls were awakened from rapid eye movement (REM) sleep between 1:30 and 3:30 AM, and the rapidity of return to REM sleep was determined. The time it took to return to REM sleep was reduced in depressives compared with controls: 6•.6 (17.9 SD) min versus 80.6 (24.9 SD) rain, respectively (p = 0.01). The time elapsed until the return to REM sleep was significantly correlated with baseline REM latency in controls (but not depressives). In contrast, return to REM time was significantly correlated with depression severity scores in depressives (but not controls). There was no evidence to support the hypothesis that the more rapid return to REId sleep in depression was caused by a slow wave sleep deficit. The mechanism underlying the rapid return of REM sleep in depression thus may be related to a severity-linked disturbance, such as a proposed increase in REM "pressure."
Introduction Reduced latency to the onset of ~,hefirst rapid eye movement sleep period (REM latency) is a well-described feature of primary or endogenous depressions: approximately 50% of depressed outpatients and up to 80% of inpatients manifest a consistent reduction of REM latency (Thase and Kupfer 1987). However, despite the prevalence of this psychobiologic abnormality, it remains unclear whether reduced REM latency in depression is due to an active, state-dependent process (such as increased REM '~pressure") (Vogel 1983) or reflects a more indirect consequence of diminished slow-wave sleep (Kupfer and Ehlers 1989). As severe depressions are often associated with both increased REM indexes and diminished slow wave sleep (Thase and Kupfer 1987), alternate experimental methods are necessary to unravel these competing hypotheses. One possible method has been described by Schulz and Tetzloff (1982): depressed patients are awakened at 2:30 AM, and the latency (or elapsed time) until the return of REM sleep is determined. This method is novel in that it may obviate potential confounds associated with difficulty initiating
From the Western Psychiatric Institute and Clinic, Department of Psychiatry, University of Pittsburgh School of Medicine, Pittsburgh, PA. Address reprint requests to Dr. M. E. Thase, Western Psychiatric Institute and Clinic, Department of Psychiatry, University of Pittsburgh School of Medicine, 3811 O'Hara St., Pittsburgh, PA 15213. Received February 28, 1991; revised July 24, 1991. © 1992 Society of Biological Psychiatry
0006-3223/92/$05.00
172
BIOL PSYCHIATRY 1992:31:171-176
M.E. Thase et al
sleep and control for possible relationships between sleep onset and clock time. Moreover, the 2:30 AM awakening occurs at a period in the normal sleep cycle in which a majority of slow wave sleep (SWS) already has occurred, thus partially accommodating for potential reductions in SWS observed during the first third of the night in some depressed patients. Using this paradigm, Schulz and Tetzloff (1982) found a reduced latency to the return of REM sleep during both illness and remission in depressed patients with sleep onset REM periods (SOREMPs). The early return of REM sleep observed during remission (i.e., following resolution of the depressive state) is suggestive of a trait-like characteristic, and therefore may reflect a more "passive" process linked to diminished SWS rather than a severity-linked or state-dependent phenomenon. We have had the opportunity to replicate and extend Schulz and Tetzloff's (1982) experimental awakening paradigm in our ongoing duties of REM sleep and nocturnal penile tumescence (NPT) in depression (see, e.g., Thase et al 1988). In this protocol, patients were periodically awakened during the third night of electroencephalographic (EEG) sleep recording for determinations of penile buckling force. Because a number of these awakenings coincided with the 2:30 AM time employed by Schulz and Tetzloff, we compared the return times to REM sleep in these patients against those of matched normal controls studied usi,ig the same methodology. Further, as our sample consisted of less severely depressed outpatients, with relatively normal SWS and without SOREMPs (Thase et al 1988), we were able to extend this method to a sample of patients with both milder depressive symptomatology and less severe sleep disturbances.
Method Our methods for recruitment and assessment of depressed patients and normal controls for participation in the NIT protocol have been reported previously, as have the main findings of this project (Thase et al 1988). To summarize briefly, depressed men between the ages of 20 and 60 presenting for treatment at the Western Psychiatric Institute and Clinic who met both clinical (American Psychiatric Association 1980) and research criteria [Schedule for Affecti ve Disorders and Schizophrenia (SADS) (Endicott and Spitzer 1978 ); Research Diagnostic Criteria (RDC) (Spitzer et al 1978)] for a diagnosis of major depressive disorder were eligible to enroll in the study. Controls were recruited from media advertisements and hospital staff. Potential control subjects were interviewed to exclude individuals with diagnosable psychopathology using the lifetime version of the SADS (Spitzer et al 1975). All subjects were further screened with a comprehensive physical examination and appropriate laboratory studies to rule out medical conditions that might either cause depression or erectile dysfunction, or require treatment with medications that would invalidate EEG sleep/NIT studies. A battery of clinical assessment measures that included the Hamilton Rating Scale for Depression (HRSD) (Hamilton 1960) and a selfreport c,f depressive symptoms derived from the Brief Symptom Inventory (BSI) (Derogatis 1975) also was completed. All subjects provided written informed consent for research participation and underwent a supervised 14-day drug-free ahd alcohol-free interval. Following the washout period, subjects underwent a 3-night EEG sleep/NPT protoc~ J. During the third night of protocol (but not during nights I and 2), subjects were awaken~:d during each episode of NPT for visual estimates of penile rigidity and measuremert of buckling force (see Thase et al 1988). The~e procedures generally took less than 3 ~ain to complete, and during this time the techn~,logist spoke with the subjects to
Post-Arousal REM Sleep in Depression
BIOL PSYCHIATRY 1992;31:171-176
173
confirm their awakening. Thereafter, subjects were permitted to return to sleep immediately. The sleep EEG/NPT polygraphic records of 24 consecutively studied depressed men and 24 healthy controls were initially matched on the basis of age ( _+ 5 years); selection of matched pairs for possible inclusion in this analysis was determined by the availability of the original sleep record for manual re-scoring. The availability of records was directly related to the recency of the recordings; we do not store polygraphic records for longer than 12-18 months. Among these 48 potential cases, a total of 16 depressed and 20 control subjects met the following criteria: (1) they were awakened within 60 min (_+) of the 2:30 AM time utilized by Schulz and Tetzloff (1982), and (2) they were awakened from an REM period. This latter modification served to equate the patients with respect to both clock time and position within the REM-NREM cycle. Demographic and clinical characteristics of the same of 36 subjects are summarized in Table 1, as are selected baseline (night 2) EEG sleep characteristics. We used night 2 for baseline EEG sleep recordings, allowing night 1 for accommodation to the laboratory. The key dependent measure was min of sleep elapsed from the end of the experimental awakening until the onset of REM sleep (->3 min in duration). The prediction that more rapid return to REM sleep would occur in depression was tested by comparing the elapsed time values for the two groups with two-tailed t-tests. Pearson correlation coefficients between baseline REM latency (taken from night 2) and elapsed time until return to REM sleep (night 3) were computed, whereas nonparametric (Spearman rho) correlations were t~sed to assess the relationship between severity of depression (HRSD and BSI-D) versus baseline REM latency (from night 2) and time until return to REM sleep. (The nonparametric correlations were used for analyses of symptomatic measures because of differences in the distribution of scores on these measures in the depressed and normal control samples.) The proposed relationship of a more rapid return to REM sleep in patients with SWS deficits was examined by computing the number of minutes (and percent) of hand-scored SWS measured between sleep onset and the experimental awakening, and between the experimental awakening and return to REM sleep. We also performed parallel analyses in 33 of the subjects (15 depressed and 18 controls), for whom we were able to use more refined automated scoring methods for total delta wave counts and delta wave counts per min (i.e., delta waves in the 0.5-3.0 Hz range; Kupfer et al 1986): The hypothesis that early return to REM sleep in depression would be associated with a severity-linked process would be supported by an association with clinical measures of depression in the absence of, or independent of, a slow-wave sleep deficiency.
Results Baseline REM latency differed significantly between groups: depressed ~ = 56.6 (14.4) min versus control ~ = 92.9 (44.8) rain, p < 0.0001 (see Table 1). The samples did not differ with respect to either the mean time of the experimental awakening on night 3 [control it = 02:32 AM (0:32), depressed ~ = 02:28 AM (0:26); t = 0.47, df = 34, p = NS] or min of time elapsed from sleep onset until the awakening [control ~ = 183.9 min (35.9 min), depressed ~ - 169.3 min (56.7 min); t = 1.03, p = NS]. Among the depressed patients, l was awakened from REM period 1, 12 from REM period 2, and 3 from REM period 3. Among the controls, the proportions were as follows: 2 (REM 1
174
M.E. Thase et al
BIOL PSYCHIATRY 1992;31:!71-176
Table 1. Demographic, Clinical, and EEG Sleep Characteristics of Depressed and Healthy Control Subjects* Controls
Depressives
(n = 20)
(n = 16)
Demographics Age
Mean -+ SD Range Race (white/other) Education (yr) Marital status (married/partner/no partner) Number of episodes Duration of current episode (weeks) Median -- SD Range RDC subtypes
Clinical Hamilton (17-item) Brief Symptoms Inventory Depression Bseline EEG sleep (from night 2) Sleep efficiency (%) REM time (rain) REM (%) REM latency (rain) SWS (min)a SWS (%)"
Return to REM time (from night 3)
37.5 (11.6) 22-59 18/2 16.4 (2.8) 10/4/6
38.0 (11.4) 21-58 15/1 14.8 (2.9) 5/2/9
NS b NS c NS b NS c
2.3 (!.6) 31.2 (37.8) 4-150 12 unipolar, 4 bipolar 11 4 nonendogenous 8 endogenous 0.5 (1.2) 0.2 (0.3)
90.3 79.6 20.0 92.9 60. I 15.5 80.6
(6,0) (26.9) (6.2) (44.8) (45.5) (I !.6) (24.9)
18.9 (4.3) 1.4 (0.6)
86.5 81.9 21.4 56.6 36.9 9.4 61.6
(9.3) (19.5) (3.4) (14.4) (29.8) (7.3) (17.9)
0.001 d
0.001 e
NS b NS b NS b
o.0olf NS b NS b 0.01 ~
*Standarddeviations are in pmentheses. "Sleep stages 3 and 4. ~Oroupt-test, 'X' for contingency tables, dGroup t-test, t tGroup t-test, t fGroup t-test, t ~Groupt-test, t
= = = =
- 16.6, df - 17. I, p < 0.0001 (adjusted for unequal variances), -6.5, df = 14.2, p < 0.0001 (adjusted for unequal variances). 3.4, df = 23.7, p < 0.0001 (adjusted for unequal variances). 2.7, df = 34, p = 0.01.
period), 14 (REM 2 period), and 4 (REM 3 period). These proportions did not differ between the two groups (x = = 0.19, p = 0.91). The depressed and control subjects also did not differ with respect to the number of min of REM sleep elapsed in the episode they were awakened from (depressed ~ = 11.3 (2.9 SD) min versus controls ~ = 13.1 (6.3 SD) min, t = 1.1, df = 34, p = 0.31). However, despite these similarities in the timing of the awakening and its position within the REM-NREM cycle, latency to return to REM sleep was reduced in the depressed patients ~ = 61.6 (17.9 SD) mini compared with normal controls [~ = 80.6 (24.9 $D) rain, p = 0.OI] (see Table 1). Baseline REM latency was significantly correlated with the latency of return to REM sleep in controls (r = 0.54, p < 0.05), but not in the depressives (r = 0.18, p = NS). A comparison of the strength of these two correlations using Fisher's r to z score transformation method indicated that the two groups' values significantly differed from each other (z = 2.10, p = 0.04). Spearman rho correlations between the time to return to
Post-Arousal REM Sleep in Depression
BIOL PSYCHIATRY 1992;31:i71-176
175
REM sleep and measures of severity of depression revealed a significant association in the depressed sample (BSI-D: r = - 0 . 5 6 , p < 0.02; HRSD: r = - 0 . 6 1 , p < 0.01), but not in normal controls (BSI-D: r = 0.00, p = NS; HRSD: r = - 0 . 3 2 , p = NS). By contrast, baseline REM latency was not significantly correlated with these symptomatic measures in either depressed patients (HRSD: r = 0.03, p = NS; BSI-D: r = - 0 . 1 0 , p = NS) or normal controls (HRSD: r = - 0 . 2 2 , p = NS; BSI-D: r = 0.02, p = NS). Percentage of sleep time occupied by SWS stages 3 and 4 did not significantly differ between controls and depressives. This was true with respect to both SWS values from baseline (night 2) and analyses of the percentage of total SWS recorded from sleep onset until the experimental awakening on night 3 [control ~ = 67.4% (36.0%), depressed = 78.6% (32.3%); t = - 0 . 9 7 , df = 34, p = NS], as well as the SWS recorded between the 2:30 AM awakening and the return to REM sleep [controls ~ = 12.3% (17.7%), depressed ~ = 7.8% (15.4%); t = 0.81, df = 34, p = NS]. Similarly, neither average delta wave counts per min (cpm) nor total delta cpm differed significantly between groups. Thus, between sleep onset and the experimental awakening, depressive~ averaged 27.3 (14.1) cpm, whereas controls averaged 27.6 (17.1) cpm (t = 0.06, df = 31, p = NS). Similarly, during the interval between experimental awakening and return to REM sleep, depressives averaged 16.7 (13.8) cpm and controls averaged 17.1 (14.7) cpm ( t = 0.08, df = 31, p = N S ) . Discussion These findings replicate and extend the earlier work of Schulz and Tetzloff (1982). We have found the experimental awakening protocol to be applicable to an outpatient sample of depressed patients. This paradigm may offer additional advantages in future research on the relationship of REM sleep and depression by eliminating potential confounds associated with severe sleep onset difficulties and "missed" fivJt REM sleep periods. Moreover, it provides a novel method to assess REM sleep pressure somewhat independently from disturbances of slow wave sleep. It i~ intriguing that the return to REM sleep was significantly related to baseline REM latency in normal controls but not in the depressed patients. Though we had not anticipated this finding, it appears to be consistent with Kupfer and Ehlers' (1989) two-factor model of REM latency disturbance in depression. This model postulates the existence of at least two mechanisms underlying REM latency disturbances in depression. In type I disorders, which are hypothesized to predominate in less severely depressed patients, ItEM latency is proposed to reflect a stable, state-independent abnormality. As such, REM latency values in the 50-70 min range often persist despite remission of the depressive syndrome, suggesting either a "scar" or a trait-like phenomenon (Thase an0 Simons in press). By contrast, Kupfer and Ehlers (1989) propose a second (type 2) disturbance in which a process associated with marked depressive severity causes a state-dependent reduction of REM latency, with values often less than 40 min. The early return to REM sleep observed in the current ambulatory sample may thus be "driven" by the severity-linked (type 2) disturbance which is unrelated to the baseline REM latency (type 1) of our ambulatory sample. This would account for the significant correlations between selfreported and observer-rated measures of severity and return to REM time in the depressed patients (but not the controls), as well as the absence of a relationship between baseline REM latency and return to REM time in depressed patients relative to the unaffected controls. Of course, an alternate, more parsimonious (and admittedly skeptical) expla-
176
BIOL PSYCHIATRY 1992;31:!71-176
M.E. Thase et al
nation of these findings is that selectively restricted ranges of variability observed on the key measures (i.e., REM latency for depressives and symptom scores for controls) may have artificially produced these findings. Further study of Kupfer and Ehlers' (1989) twofactor model obviously is necessary. On the other hand, it is fairly clear that the significantly reduced latency of return to REM sleep observed in our study was not due to a deficit of slow wave sleep in this outpatient sample of depressed men. Indeed, the depressed and control groups did not differ with respect to either hand-scored or automated measures of SWS. The absence of an association between diminished slow wave sleep and an early return to REM sleep may provide further indirect support of an unidentified (as of yet) state-dependent abnormality related to the dysregulation of REM sleep in depression. We believe that this observation, coupled with the observed relationship between illness severity and time to return of REM sleep, support the alternate hypothesis that the REM sleep pressure may predominate during acute depressive illness. Restudy of our sample following remission from depression will be necessary to definitively ascertain whether such findings are related to a state-dependent process (i.e., severity-linked) or reflect a more enduring traitlike abnormality. This research was supported in part by Grants MH-30915 (Mental Health Clinic Reseat,:, Center), MH-40023 and MH-00295 (CFR) and MH-41884 (MET) from the National Institute of Mental Health. 'We wish to thank Ms. Lisa Stupar for her assistance in preparing this manuscript.
References American Psychiatric Association (1980): Diagnostic and Statistical Manual of Me~:~alDisorders, 3rd ed. Washington DC: American Psychiatric Press. Derogatis LR (1975): The brief symptom inventory. Baltimore, MD: Clinical Psychometrics. Endicott J, Spitzer RL (1978): A diagnostic interview: The scheduled tbr affective disorders and schizophrenia. Arch Gen Psychiatry 35:837-848. Hamilton M (1960): A rating scale for depression. J Neurol Neurosurg Psychiatry 12:56-62. Kupfer DJ, Ehlers CL (1989): Two roads to REM latency. Arch Gen Psychiatry 46:945-948. Kupfer DJ, Reynolds CF, Grochocinski VJ (1986): Comparison of automated REM and slow wave sleep analysis in young and middle-aged depressed subjects. Biol Psychiatry 21:189-200. Schulz H, Tetzioff W (1982): Distribution of REM latencies after sleep interruption in depressive patients and control subjects. Biol Psychiatry 17:1367-1376. Spitzer RL, Endicott J, Robins E (1975): Schedule for Affective Disorders and SchizophreniaLifetime Version. New York: New York State Psychiatric Institute, Biometrics Research. Spitzer RL, Endicott J, Robins E (1978): Research diagnostic criteria: Rationale and reliability. Arch Gen Psychiatry 35:773-782. Thase ME, Kupfer DJ (1987): Current status of EEG sleep in the assessment and treatment of depression. In Burrows GD, Werry JS (eds), Advances in Human Psychopharmacology, Vol 4. Greenwich, CT: JAI Press, pp 93-148. Thase ME, Simons AD (199~): The applied use of psychotherapy to study the psychobiology of depression. Psychother Practice Res (in press). Thase ME, Reynolds CF Ill, Jennings JR, et al (1988): Nocturnal penile tumescence is diminished in depression. Bi, I Psychiatry 24:33-46. Vogel GW (1983): Evidence for REM sleep deprivation as the mechanism of action of antidepressant drugs. Prog Neuropsychopharmacol Biol Psychiatry 7:343-349.