Regulation of Sleep and Growth Hormone in Adolescent Depression RONALD E. DAHL, M.D., NEAL D. RYAN, M.D., DOUGLAS E. WILLIAMSON, B.A., PAUL J. AMBROSINI, M.D., HARRIS RABINOVICH, M.D., HANA NOVACENKO, M.S., BEVERLY NELSON, R.N., AND JOAQUIM PUIG-ANTICH, M.D.
Abstract. This article reviews findings of sleep, growth hormone (GH), and cortisol measures from a number of separate controlled studies of prepubertal and adolescent depression carried out by Puig-Antich and colleagues since 1978. New data are presented comparing 24-hour GH measures in adolescents with major depressive disorder (MDO) (N = 44; mean age = 14.8 ± 2.0) to normal control adolescents (N = 37; mean age = 15.3 ± 1.5). There were no significant overall group differences in summary GH measures between MOD and normal controls. Splitting the MOD group on the basis of suicidality (definite plan or attempt) (N = 20), revealed a significant blunting of sleep GH compared to the nonsuicidal group (N = 24). These results are discussed in the context of the other sleep and neuroendocrine findings in this population, with evidence for dysregulation around slccp onsct. The influences of development on sleep and GH regulation are also considered. J. Am. Acad. Child Adolesc. Psychiatry, 1992,31,4:615-621. Key Words: growth hormone, depression, adolescence, sleep. Among his many accomplishments, Joaquim (Kim) PuigAntich contributed pioneering data and thought to the issues of sleep and neuroendocrine regulation in early onset depression. His approach to this area emphasized three themes: the collection of empirical data with methodological rigor, maintaining a developmental perspective in the analysis and interpretation of those data, and the consideration of the data in a broad framework and large theoretical context. This article will try to be consistent with those themes. We will present previously unpublished data and analysis of growth hormone (GH) secretion in depressed adolescents, review related data on sleep and neuroendocrine regulation in early onset major depressive disorder (MDD), and finally, attempt to place these findings in a larger theoretical (and developmental) framework.
Background GH, Sleep, and Cortisol in Prepubertal MDD
In a series of reports published in 1984, Puig-Antich et a!., showed that prepubertal children with MDD secreted less GH in response to insulin-induced hypoglycemia than nondepressed pathological controls (Puig-Antich et a!., 1984a), a finding that persisted in medication-free recovery (Puig-Antich et a!., 1984b). These MDD subjects also secreted increased sleep-stimulated GH compared with both
Accepted December 10, 1991. From the Western Psychiatric Institute and Clinic, Department of Psychiatry, University of Pittsburgh, School of Medicine, Pittsburgh, PI\., and Eastern Pennsylvania Psychiatric Institute, Philadelphia, PA, and Department of Child Psychiatry, New York State Psychiatric Institute, Department of Psychiatry Columbia University College of Physicians and Surgeons, New York, NY. This article is dedicated to Joaquim (Kim) Puig-Antich. His blend of intellect, passion, humor, and caring created a force with a profound influence on friends, colleagues, and the field ofchild psychiatry. This force has a life of its own and continues to grow. Reprint requests to Dr. Dahl, Western Psychiatric Institute and Clinic, 3811 O'Hara Street, Room E733, Pittsburgh, PA 15213. 0890-8567/92/l304-0615$03.00/0©1992 by the American Academy of Child and Adolescent Psychiatry. J. Am. Acad. Child Adolesc. Psychiatry, 31:4, July 1992
pathological and normal controls (Puig-Antich et a!., 1984c), which again continued after recovery (Puig-Antich et a!., 1984d). EEG sleep measures in this population had shown no significant group differences in major sleep variables during the episode of depression (Puig-Antich et a!., 1982), but on recovery the depressed subjects had significantly reduced rapid eye movement (REM) latency (compared with themselves during the episode, and compared with the normals) (Puig-Antich et a!., 1983). Analysis of serum cortisol measures in these subjects revealed no significant differences compared with normal controls or pathological controls (Puig-Antich et a!., 1989) and no significant differences in dexamethasone suppression between groups (Birmaher et a!., 1992). (Despite an early publication on baseline cortisol measures showing two of four MDD subjects with cortisol hypersecretion [Puig-Antich et a!., 1978], the final conclusion from the large sample studies was that cortisol hypersecretion in prepubertal MDD was an infrequent event.) GH, Sleep, and Cortisol in Adolescent MDD
In another series of studies, Puig-Antich and colleagues examined sleep and neuroendocrine measures in adolescent depression. GH response to an intramuscular injection of desmethylimipramine (DMI) showed that depressed adolescents secreted significantly less GH than normal control adolescents, and that the largest group differences were contributed by the subgroup of suicidal MDD subjects (Ryan et a!., 1988). EEG sleep studies in the adolescents found no significant differences in REM latency or slow wave sleep, but they did report evidence of decreased sleep efficiency in the MDD group compared with normal controls (Goetz et a!., 1987). A second study of EEG sleep in depressed adolescents also failed to find significant group differences in REM and slow wave sleep measures, but in subgroup analysis (based on the findings of Ryan et a!., with DMI GH in the suicidal group), REM differences were evident in the suicidal inpatient subgroup (Dahl et a!., 1990). Cortisol measures (baseline 24-hour levels) in these adolescents revealed no significant group differences between MDD and normal controls in the predominately outpatient study (Dahl
615
DAHL ET AL.
et aI., 1989). In the second adolescent study (with a substantial inpatient sample), significant group differences in cortisol measures occurred near sleep onset, with the MDD subjects showing elevated cortisol at this time when cortisol is usually physiologically suppressed (Dahl et aI., 1991). Again, however, subgroup analysis revealed that virtually all of these group differences in cortisol were accounted for by the suicidal inpatient subgroup. In summary, the strongest adolescent findings with DMI GH, EEG sleep, and cortisol, were occurring in a subgroup of MDD subjects with suicidality and/or inpatient status. Because suicidality and inpatient status (as well as severity of depression) were often confounded (overlapped) in these samples, it was impossible to disentangle what specifically was accounting for these findings (see Dahl et aI., 1990, for discussion). Measures of depression severity alone (K-SADS or Hamilton Depression Rating Scale scores), however, did not correlate significantly with the biological measures. These results, taken together, provided the background for the three primary questions with respect to the 24-hour GH measures in this adolescent study: I. Do significant MDD and normal group differences exist in the levels and patterns of 24-hour GH secretion? We had evidence that DMI stimulated GH was blunted in these MDD adolescents (Ryan et aI., 1988), so we wanted to examine unstimulated baseline GH secretion over 24 hours. Within the 24-hour GH pattern, we were specifically interested in GH measures around sleep onset. Sleep onset (or the deep, slow wave sleep that usually occurs within I to 3 hours of sleep onset) is a potent stimulus for GH secretion in humans (Mendelson et aI., 1979; Takahashi et aI., 1968). In contrast, sleep onset strongly suppresses cortisol secretion (Weitzman et aI., 1983). 2. How do the results of sleep-stimulated GH secretion in MDD adolescents compare with prepubertal and adult studies? Puig-Antich et aI. (1984b) had found increased sleep GH in prepubertal MDD, whereas Jarrett et aI. (1990) reported significant blunting of sleepstimulated GH in adults with recurrent depression. Thus, the developmental influences on sleep and GH regulation appeared to be a key issue of interest relevant to this adolescent study. 3. What are the results of sleep-stimulated GH in the suicidal subgroup of adolescents? Given the subgroup findings described previously in adolescent MDD studies of sleep and neuroendocrine regulation (Dahl et aI., 1990, 1991; Ryan et aI., 1988), we planned to carefully examine clinical subgroups, including suicidality, severity, endogenous subtype, and comorbidity. Subjects and Methods of Procedure
Adolescents with screening suggestive of depression were entered into a 2-week diagnostic protocol that included psychiatric, psychosocial, and pediatric assessments. Two independent psychiatric assessments were done, using the Schedule for Affective Disorders and Schizophrenia for 616
School Age Children-Present Episode (K-SADS-P) (Chambers et aI., 1985). The first rater also administered a K-SADS-E (epidemiologic version) (Orvaschel et aI., 1982) to assess the nature of any previous episodes of psychiatric disorder. The second K-SADS-P evaluation was performed 10 to 14 days after the first one and was concerned only with the preceding week. Adolescents were accepted into the study if they were between 12 and 18 years old, had attained at least Tanner stage III (Marshall and Tanner, 1969, 1970) of pubertal development, were determined to be medically healthy by the pediatrician, and fit unmodified Research Diagnostic Criteria (RDC) (Endicott and Spitzer, 1978) for MDD in both psychiatric evaluations. Criteria for exclusion were: (1) mania or bipolar I disorder; (2) medications (a 2-week drug washout period was utilized); (3) significant medical illness; (4) marked obesity (weight or height greater than 95th percentile on the National Center for Health Statistics curve) or severe growth failure (weight or height under the 3rd percentile); (5) IQ below 70; (6) meeting DSM-lll (American Psychiatric Association, 1980) criteria for anorexia nervosa, autism, schizophrenia, or moderate or severe conduct disorder, or RDC for schizoaffective disorder; and (7) pregnancy. In addition to these exclusion criteria, normal control adolescents had no current presence or past history of DSMIII Axis I psychiatric disorder as determined by a single K-SADS-E assessment, which included a semistructured interview with the parent and another with the adolescent. Informed consent was obtained from the parent or legal guardian and from all adolescents. The normal subjects were paid for their participation in the study. After acceptance into the study, the adolescents were admitted to the Sleep-Neuroendocrine Laboratory for 3 consecutive nights. All procedures were explained in advance to the subjects in great detail, and they had visited the laboratory during the diagnostic protocol. Electrode placement for the standard polysomnographic recording was carried out 1 hour before the subject's stated bedtime. Lights-out time varied from subject to subject, depending on their usual bedtime. Further details of recruitment, clinical assessment, inclusion or exclusion criterion, and sleep recording, can be found in a previously published report (Dahl et aI., 1989). GH Determinations
An indwelling venous catheter was inserted the morning after the second night of sleep and remained in place for the next 24 hours. Blood samples were drawn every 20 minutes through a 3-way stopcock system. The samples were centrifuged at 4°C; plasma was separated, then frozen at -20°C until assayed. Plasma growth hormone was analyzed by a double antibody radioimmunoassay. Human GH standard (hGH-RP-l) and the primary antiserum were donated by the National Pituitary Agency. The labeled hGH_l l25 was purchased from New England Nuclear and repurified on the day of assay on a G-IOO Sephadex column. Anti-rabbit globulin serum was used to separate the bound and free fractions. All samples were assayed in duplicate. The intraand interassay coefficients of variation were, respectively, J. Am. Acad. Child Adolesc. Psychiatry, 31:4, July 1992
SLEEP AND GH IN ADOLESCENT DEPRESSION
T ABLE I. Demographic and Clinical Variables Major Depressive Disorder (N = 44)
(N = 37)
21/23 27/6/11 14.8 ± 2.0 44.5 ± 17.9 6/38 20/24 20124
23/14 19/1 0/8 15.3 ± 1.5 38.4 ± 20.3 NA NA NA
1.69 X2 = 2.28 tJ9 = -1.29 U = 791.5 NA NA NA
:":0.19 :":0.32 :":0.20 :":0.24 NA NA NA
N = 20
NA NA NA NA NA NA
NA NA NA NA NA NA
NA NA NA NA NA NA
Sex: Male/female Race: White/black/other Age (X ± SD) Socioeconomic status (X ± SD) Inpatients/outpatients Suicidallnonsuicidal Endogenous/nonendogous Comorbidity: Anxiety Conduct disorder Obsessive compulsive Psychotic symptoms Mean K-SADS 9-item depression (X ± SD) Extracted Hamilton Rating Scale (X ± SD)
9.6% and 12.4% at 1.5 ng/mL, 3.2% and 3.7% at the level of 7.5 ng/mL, and 2.6% and 3.9% at the level of 33.3 ng/mL. Data Analysis All variables were examined using the W statistic (Shapiro and Wilkes, 1965) to determine if they fit a Gaussian distribution. If the variable was not Gaussian, a logarithmic transformation was used to normalize it before applying parametric statistics. The demographic variability from the two samples were compared using pooled and separate t tests, X2 , and Fisher's exact test as appropriate. Summary measures of the 24-hour GH secretion included: area under the curve (AUC), using the trapezoidal rule for the entire 24-hour period; maximum or peak secretion during the 24hour sampling; AUC for the first half of sleep (4 hours postsleep onset); AUC for the sleep period (8 hours postsleep onset); and the percentage of growth hormone (for the 24-hour period) that was secreted during sleep. All between group comparisons were made using pooled or separate t tests, as appropriate, with a = 0.05. Relationships between the dependent growth hormone variables and other continuous variables such as age were tested using Pearson correlations (after log transformations to achieve Gaussian distribution, when necessary). One-way analysis of variance with covariance methods were used to control for possible confounders such as age, sex, and to test interactions. Results Sample
From the original sample of 88 adolescents participating in the protocol, six subjects (two MDD and four normal controls) were eliminated because of missing data (difficulties with blood draws from the intravenous catheter). One additional subject (MDD) was dropped because of erratic GH assay values (>3 SD from the mean). This left a final sample of 44 MDD adolescents and 37 normal controls. There were no significant group differences in age, race, J. Am. Acad. Child Adolesc. Psychiatry, 31 :4, July 1992
N=7 N=3 N=7 3.5 ± 0.6 23.2 ± 6.7
Normal Statistic
x2 =
p
sex, or socioeconomic status (Hollingshead Scale) as shown in Table 1. Within the MDD group, 20 met criteria for endogenous subtype, six were inpatients, seven had psychotic features, and 20 were suicidal (12 had attempted suicide and eight had significant suicidal ideation with a definite suicidal plan but had not made a clear attempt). Comorbid anxiety (generalized anxiety, separation anxiety, or phobia) was present in 20 subjects, conduct disorder in seven, and obsessive compulsive disorder in three. The mean severity score using the nine-item K-SADS scale was 3.5 ± 0.6, with the extracted Hamilton Depression Rating Scale (Williamson et a!., in press) mean score was 23.2 ± 6.7. GH Results
There were no significant group differences in any of the summary GH variables as shown in Table 2, including peak GH, 24-hour integrated levels (AUC), mean GH over the first half of the night, mean GH over the sleep period, or the percentage of total GH secreted during the sleep period. As shown in Figure 1, the overall secretory patterns between the groups were quite similar. We also examined the effects of age and sex on the GH measures. Age was negatively correlated with GH secretion during sleep: r = -0.35, p < 0.01; and total 24-hour GH secretion; r = -0.24, P < 0.03. Sex was a significant covariate for total 24-hour GH secretion (F 1•77 = 5.2, P S 0.03), and percentage of GH secreted during sleep (F 1•77 = 8.4, P S 0.01), with the females secreting more total GH, but a lower ratio of sleep GH. With the addition of age and sex as covariates, MDDs were still not significantly different from normal controls on these
measures. Analysis of age by diagnosis interaction revealed one significant effect. Within the MDD group, the percentage of GH secreted during sleep showed a negative correlation with age (r = -0.45, p < 0.01), compared to the normal group (r = +0.16, P = NS); the interaction was found to be significant (F 1•77 = 6.59, p S 0.03). There were no significant interactions between diagnoses and sex or age and sex. The MDD group was dichotomized into suicidal and nonsuicidal subgroups, using a score of 4 or greater on the 617
DAHL ET AL. TABLE
2. Growth Hormone Summary Variables of Major Depressive Disorder vs. Normal Comparisons
Major Depressive Disorder (N = 44) 24-hour area under the curve 24-hour peak Area under the curve first 4 hours of sleep Area under the curve 8 hours of sleep Area under the curve 8 hours of sleep (as a percentage of 24-hour secretion)
250.2 ± 116.5 21.3 ± 9.9 92.6 ± 42.7 131.3 ± 65.5 51.9± 13.9
K-SADS items for suicidality (this indicates a definite suicidal plan or an actual suicidal attempt and was used as the cutoff in previous analyses by Ryan et al., 1988 and Dahl et al., 1990). This comparison revealed a significant blunting in sleep-stimulated GH secretion in the suicidal MDD group as illustrated in Figure 2. Summary GH variables (shown in Table 3) also indicated that the suicidal subgroup secreted significantly less GH in the first half of the night compared with the nonsuicidal group (76.4 ± 36.9 vs. 106.1 ± 43.2; t42 = -2.43, P ~ 0.02). Including age and sex as covariates did not change the significance of this finding (F 1,42 = 7.1, p ~ 0.02). The suicidal group also secreted less GH over the total sleep period (107.1 ± 55.0 vs. 151.5 ± 67.7; t42 = -2.37, p ~ 0.03) and a trend for less total 24-hour GH than the nonsuicidal group (t42 = 1.88, P ~ 0.07). Splitting the MDD group on endogenous subtype, comorbidity of anxiety, conduct disorder, psychotic features, or inpatient status did not reveal significant subgroup differences for GH secretion. An analysis of severity of depression (using K-SADS scores) showed no significant correlation between severity of depression and GH summary values. Discussion
14,---------------------,
,..,
MDDs(n=44)
I'
Normals (n=37)
3 FIG.
618
5
7
256.0 21.6 96.4 146.7
± ± ± ±
179.9 10.5 76.8 109.3
54.4± 15.1
Statistic
p
=
0.21 -0.06 0.89 -0.13
s; 0.83 S; 0.95 S; 0.38 S; 0.90
=
-0.78
s; 0.43
t79
=
t79 t79 t79
=
t79
=
subgroup. Other subgroup splits did not reveal significant GH differences. (However, the inpatient sample in this study was very small, N = 6.) The subgroup of suicidal MDD adolescents had also revealed blunted GH secretion after 1M DMI stimulation (Ryan et al., 1988). As discussed by Ryan, these results may indicate either a noradrenergic or serotonergic alteration of GH secretion in the suicidal MDD adolescents. Increased somatostatin (endogenous inhibitor of GH secretion) is also a possible mechanism for findings of blunted GH response across multiple tests. These results must also be considered in the context of other related GH, particularly in relation to developmental influences. As stated in the introduction, Puig-Antich et al., had shown that prepubertal MDD subjects had increased sleep-stimulated GH compared to controls. Kutcher et al., (1988) also reported increased nocturnal GH associated with MDD in a small sample of adolescents and controls. Jarrett et al. (1990), however, showed that adults with recurrent MDD had significant blunting of sleep-stimulated GH that persisted upon drug-free recovery. These findings suggest that age and development may be playing a central role in the effects of depression on sleep-stimulated GH. Also, as
There were no significant group differences in sleep-stimulated GH or total 24-hour GH secretion between the MDD and normal groups. Splitting the MDD group on suicidality revealed significant blunting of sleep GH in the suicidal
I I
Normal (N = 37)
9 11 13 15 17 Hours Post Sleep Onset
19
1. Twenty-four hour GH aligned by sleep onset.
21
23
presented in the introduction, pharmacological tests of GH secretion (including insulin, clonidine, and DMI) have revealed blunted GH associated with MDD across prepubertal, adolescent, and adult studies (Annseau et al., 1984; Eriksson et al., 1988; Jarrett et al., 1990; Jensen and Garfinkel, 1990; Lesch et al., 1988; Matussek et al., 1980; PuigAntich 1984a, b, c, d; Ryan, 1988). Because blunted GH response to pharmacological stimulation appears stable across development, the variance in sleep GH findings across puberty raises questions with respect to how development may be influencing the regulation of sleep and GH in MDD. A few possible mechanisms must be considered. There is evidence that sleep, pharmacological probes, and exercise act through separate neurotransmitters or peptide mechanisms in stimulating growth hormone secretion at the pituitary (Mendelson, 1982). For example, adrenergic modulation of GH secretion appears to be much less significant during sleep (Lucke and Glick, 1971). It is possible that the age effects on sleep and GH findings represent changes in one specific component of this network of regulation. Another possible explanation, however, is that age-related changes in sleep GH findings are reflecting developmental changes in the regulation of sleep itself. This paradigm of age-related changes in sleep as a central J. Am. Acad. Child Adolesc. Psychiatry, 31 :4, July 1992
SLEEP AND GH IN ADOLESCENT DEPRESSION
Non-suicidal (n=24) Suicidal (n=20) Normals (n=37)
3
FIG.
5
7
9 11 13 15 17 19 21 23 Hours Post Sleep Onset 2. Twenty-four hour GH (suicidal split) aligned by sleep onset.
component to these findings is consistent with results from the other sleep and neuroendocrine studies described in the introduction. Specifically, EEG sleep changes were not evident in the prepubertal group studied during the episode of depression (Puig-Antich et aI., 1982). In the adolescent sample, EEG sleep differences occurred in the suicidal inpatient subgroup clustered around sleep onset (increased latency to sleep and decreased latency from sleep onset to the first REM period) (Dahl et aI., 1990). The baseline night cortisol measures followed the same pattern-there were no significant differences in prepuberty, but group differences appeared in the subgroup of suicidal adolescent MDD subjects in the interval of time around sleep onset (at a time when cortisol is normally suppressed by sleep onset, the suicidal MDD subjects had increased cortisol) (Dahl et aI., in press). These results from baseline GH measures also fit this pattern, with blunted GH findings emerging around sleep onset in the subgroup of suicidal adolescents with MDD. Taken together, these results suggest dysregulation around sleep onset. All of the findings (GH, cortisol, and EEG sleep) are consistent with a sleep onset mechanism which is weakened (or impaired) in the subgroups of affected adolescents. That is, because sleep onset stimulates growth hormone and suppresses cortisol, impaired sleep onset is consistent with the blunted growth hormone and increased cortisol. The increased latency to sleep onset and TABLE
reduced REM latency seen in these studies also fit with this model. The deep, slow wave sleep in the first 1 to 2 hours after sleep onset delays the onset of the first REM period, thus dysregulation at this time may allow an earlier REM period, resulting in reduced REM latency. This paradigm of dysregulation in the first 1 to 2 hours of sleep also fits well with psychobiological studies in adult MDD, as discussed previously by Kupfer et al. (1986). This paradigm is also consistent with our knowledge of the age and developmental influences on sleep regulation. Across the human life span, increasing age is strongly associated with a steady decrease in the duration and intensity of slow wave sleep, decreased efficiency and continuity of sleep, decreased threshold of arousal, and decreased REM latency (Dement et aI., 1982; Williams et aI., 1974; Zeppelin, 1983). Thus, as many previous investigators have noted, EEG sleep changes associated with MDD mimic, to some degree, changes in sleep with normal aging (Reynolds and Kupfer, 1987). A recent meta-analysis by Knowles and MacLean (1990), reviewing studies of EEG sleep and MDD, indicated that normal age-related changes in sleep appeared to be accelerated in the MDD subjects. One possibility is that at the early end of the age spectrum, sleep is "protected." Children are very deep sleepers, have high sleep efficiency, are difficult to arouse from sleep, and have long sleep durations (Busby and Pivik, 1985; Carskadon et aI., 1987; Williams et aI., 1974). There are substantial changes in the regulation of sleep across adolescence, with a 40% decrease in slow wave sleep, 30% to 40% decrease in REM latency, and significantly increased daytime sleepiness (Carskadon et aI., 1983; Carskadon and Dement, 1987; Dahl et aI., 1990; Williams et aI., 1974). The emergence of disturbances in GH, cortisol, and EEG sleep in a subgroup of MDD adolescents may be caused by a physiological interaction between depression and developmental changes in the regulation of sleep resulting in dysregulation around sleep onset. These physiological disturbances near sleep onset may also provide a potential focus contributing to treatment and/ or prevention of adolescent MDD. Recent evidence from the National Institute of Mental Health epidemiological catchment area study of sleep disturbance in psychiatric disorders in adults (Ford and Kamerow, 1989), highlighted evidence that sleep disturbances may be critical in the pathogenesis of depression. That report also raised the possibility that early recognition and treatment of sleep disturbances
3. Growth Hormone Summary Variables of Major Depressive Disorder (MDD) Suicidal vs. MDD Nonsuicidal Comparisons
MDD Suicidal (N = 20) 24-hour area under the curve 24-hour peak Area under the curve first 4 hours of sleep Area under the curve 8 hours of sleep Area under the curve 8 hours of sleep (as a percentage of 24-hour secretion) J. Am. Acad. Child Adolesc. Psychiatry, 31 :4, July 1992
217.4 19.3 76.4 107.1
± ± ± ±
110.8 10.5 36.9 55.0
50.1 ± 14.8
Nonsuidical (N = 24) 277.5 23.0 106.1 151.5
± ± ± ±
116.2 9.2 43.2 67.7
53.3 ± 13.3
Statistic
p
=
-1.88 -1.25 -2.43 -2.37
:":0.22 :":0.02 :":0.03
142 =
-0.74
Abstract. This article reviews findings of sleep, growth hormone (GH), and cortisol measures from a number of separate controlled studies of prepubertal and adolescent depression carried out by Puig-Antich and colleagues since 1978. New data are presented comparing 24-hour GH measures in adolescents with major depressive disorder (MDO) (N = 44; mean age = 14.8 ± 2.0) to normal control adolescents (N = 37; mean age = 15.3 ± 1.5). There were no significant overall group differences in summary GH measures between MOD and normal controls. Splitting the MOD group on the basis of suicidality (definite plan or attempt) (N = 20), revealed a significant blunting of sleep GH compared to the nonsuicidal group (N = 24). These results are discussed in the context of the other sleep and neuroendocrine findings in this population, with evidence for dysregulation around slccp onsct. The influences of development on sleep and GH regulation are also considered. J. Am. Acad. Child Adolesc. Psychiatry, 1992,31,4:615-621. Key Words: growth hormone, depression, adolescence, sleep. Among his many accomplishments, Joaquim (Kim) PuigAntich contributed pioneering data and thought to the issues of sleep and neuroendocrine regulation in early onset depression. His approach to this area emphasized three themes: the collection of empirical data with methodological rigor, maintaining a developmental perspective in the analysis and interpretation of those data, and the consideration of the data in a broad framework and large theoretical context. This article will try to be consistent with those themes. We will present previously unpublished data and analysis of growth hormone (GH) secretion in depressed adolescents, review related data on sleep and neuroendocrine regulation in early onset major depressive disorder (MDD), and finally, attempt to place these findings in a larger theoretical (and developmental) framework.
Background GH, Sleep, and Cortisol in Prepubertal MDD
In a series of reports published in 1984, Puig-Antich et a!., showed that prepubertal children with MDD secreted less GH in response to insulin-induced hypoglycemia than nondepressed pathological controls (Puig-Antich et a!., 1984a), a finding that persisted in medication-free recovery (Puig-Antich et a!., 1984b). These MDD subjects also secreted increased sleep-stimulated GH compared with both
Accepted December 10, 1991. From the Western Psychiatric Institute and Clinic, Department of Psychiatry, University of Pittsburgh, School of Medicine, Pittsburgh, PI\., and Eastern Pennsylvania Psychiatric Institute, Philadelphia, PA, and Department of Child Psychiatry, New York State Psychiatric Institute, Department of Psychiatry Columbia University College of Physicians and Surgeons, New York, NY. This article is dedicated to Joaquim (Kim) Puig-Antich. His blend of intellect, passion, humor, and caring created a force with a profound influence on friends, colleagues, and the field ofchild psychiatry. This force has a life of its own and continues to grow. Reprint requests to Dr. Dahl, Western Psychiatric Institute and Clinic, 3811 O'Hara Street, Room E733, Pittsburgh, PA 15213. 0890-8567/92/l304-0615$03.00/0©1992 by the American Academy of Child and Adolescent Psychiatry. J. Am. Acad. Child Adolesc. Psychiatry, 31:4, July 1992
pathological and normal controls (Puig-Antich et a!., 1984c), which again continued after recovery (Puig-Antich et a!., 1984d). EEG sleep measures in this population had shown no significant group differences in major sleep variables during the episode of depression (Puig-Antich et a!., 1982), but on recovery the depressed subjects had significantly reduced rapid eye movement (REM) latency (compared with themselves during the episode, and compared with the normals) (Puig-Antich et a!., 1983). Analysis of serum cortisol measures in these subjects revealed no significant differences compared with normal controls or pathological controls (Puig-Antich et a!., 1989) and no significant differences in dexamethasone suppression between groups (Birmaher et a!., 1992). (Despite an early publication on baseline cortisol measures showing two of four MDD subjects with cortisol hypersecretion [Puig-Antich et a!., 1978], the final conclusion from the large sample studies was that cortisol hypersecretion in prepubertal MDD was an infrequent event.) GH, Sleep, and Cortisol in Adolescent MDD
In another series of studies, Puig-Antich and colleagues examined sleep and neuroendocrine measures in adolescent depression. GH response to an intramuscular injection of desmethylimipramine (DMI) showed that depressed adolescents secreted significantly less GH than normal control adolescents, and that the largest group differences were contributed by the subgroup of suicidal MDD subjects (Ryan et a!., 1988). EEG sleep studies in the adolescents found no significant differences in REM latency or slow wave sleep, but they did report evidence of decreased sleep efficiency in the MDD group compared with normal controls (Goetz et a!., 1987). A second study of EEG sleep in depressed adolescents also failed to find significant group differences in REM and slow wave sleep measures, but in subgroup analysis (based on the findings of Ryan et a!., with DMI GH in the suicidal group), REM differences were evident in the suicidal inpatient subgroup (Dahl et a!., 1990). Cortisol measures (baseline 24-hour levels) in these adolescents revealed no significant group differences between MDD and normal controls in the predominately outpatient study (Dahl
615
DAHL ET AL.
et aI., 1989). In the second adolescent study (with a substantial inpatient sample), significant group differences in cortisol measures occurred near sleep onset, with the MDD subjects showing elevated cortisol at this time when cortisol is usually physiologically suppressed (Dahl et aI., 1991). Again, however, subgroup analysis revealed that virtually all of these group differences in cortisol were accounted for by the suicidal inpatient subgroup. In summary, the strongest adolescent findings with DMI GH, EEG sleep, and cortisol, were occurring in a subgroup of MDD subjects with suicidality and/or inpatient status. Because suicidality and inpatient status (as well as severity of depression) were often confounded (overlapped) in these samples, it was impossible to disentangle what specifically was accounting for these findings (see Dahl et aI., 1990, for discussion). Measures of depression severity alone (K-SADS or Hamilton Depression Rating Scale scores), however, did not correlate significantly with the biological measures. These results, taken together, provided the background for the three primary questions with respect to the 24-hour GH measures in this adolescent study: I. Do significant MDD and normal group differences exist in the levels and patterns of 24-hour GH secretion? We had evidence that DMI stimulated GH was blunted in these MDD adolescents (Ryan et aI., 1988), so we wanted to examine unstimulated baseline GH secretion over 24 hours. Within the 24-hour GH pattern, we were specifically interested in GH measures around sleep onset. Sleep onset (or the deep, slow wave sleep that usually occurs within I to 3 hours of sleep onset) is a potent stimulus for GH secretion in humans (Mendelson et aI., 1979; Takahashi et aI., 1968). In contrast, sleep onset strongly suppresses cortisol secretion (Weitzman et aI., 1983). 2. How do the results of sleep-stimulated GH secretion in MDD adolescents compare with prepubertal and adult studies? Puig-Antich et aI. (1984b) had found increased sleep GH in prepubertal MDD, whereas Jarrett et aI. (1990) reported significant blunting of sleepstimulated GH in adults with recurrent depression. Thus, the developmental influences on sleep and GH regulation appeared to be a key issue of interest relevant to this adolescent study. 3. What are the results of sleep-stimulated GH in the suicidal subgroup of adolescents? Given the subgroup findings described previously in adolescent MDD studies of sleep and neuroendocrine regulation (Dahl et aI., 1990, 1991; Ryan et aI., 1988), we planned to carefully examine clinical subgroups, including suicidality, severity, endogenous subtype, and comorbidity. Subjects and Methods of Procedure
Adolescents with screening suggestive of depression were entered into a 2-week diagnostic protocol that included psychiatric, psychosocial, and pediatric assessments. Two independent psychiatric assessments were done, using the Schedule for Affective Disorders and Schizophrenia for 616
School Age Children-Present Episode (K-SADS-P) (Chambers et aI., 1985). The first rater also administered a K-SADS-E (epidemiologic version) (Orvaschel et aI., 1982) to assess the nature of any previous episodes of psychiatric disorder. The second K-SADS-P evaluation was performed 10 to 14 days after the first one and was concerned only with the preceding week. Adolescents were accepted into the study if they were between 12 and 18 years old, had attained at least Tanner stage III (Marshall and Tanner, 1969, 1970) of pubertal development, were determined to be medically healthy by the pediatrician, and fit unmodified Research Diagnostic Criteria (RDC) (Endicott and Spitzer, 1978) for MDD in both psychiatric evaluations. Criteria for exclusion were: (1) mania or bipolar I disorder; (2) medications (a 2-week drug washout period was utilized); (3) significant medical illness; (4) marked obesity (weight or height greater than 95th percentile on the National Center for Health Statistics curve) or severe growth failure (weight or height under the 3rd percentile); (5) IQ below 70; (6) meeting DSM-lll (American Psychiatric Association, 1980) criteria for anorexia nervosa, autism, schizophrenia, or moderate or severe conduct disorder, or RDC for schizoaffective disorder; and (7) pregnancy. In addition to these exclusion criteria, normal control adolescents had no current presence or past history of DSMIII Axis I psychiatric disorder as determined by a single K-SADS-E assessment, which included a semistructured interview with the parent and another with the adolescent. Informed consent was obtained from the parent or legal guardian and from all adolescents. The normal subjects were paid for their participation in the study. After acceptance into the study, the adolescents were admitted to the Sleep-Neuroendocrine Laboratory for 3 consecutive nights. All procedures were explained in advance to the subjects in great detail, and they had visited the laboratory during the diagnostic protocol. Electrode placement for the standard polysomnographic recording was carried out 1 hour before the subject's stated bedtime. Lights-out time varied from subject to subject, depending on their usual bedtime. Further details of recruitment, clinical assessment, inclusion or exclusion criterion, and sleep recording, can be found in a previously published report (Dahl et aI., 1989). GH Determinations
An indwelling venous catheter was inserted the morning after the second night of sleep and remained in place for the next 24 hours. Blood samples were drawn every 20 minutes through a 3-way stopcock system. The samples were centrifuged at 4°C; plasma was separated, then frozen at -20°C until assayed. Plasma growth hormone was analyzed by a double antibody radioimmunoassay. Human GH standard (hGH-RP-l) and the primary antiserum were donated by the National Pituitary Agency. The labeled hGH_l l25 was purchased from New England Nuclear and repurified on the day of assay on a G-IOO Sephadex column. Anti-rabbit globulin serum was used to separate the bound and free fractions. All samples were assayed in duplicate. The intraand interassay coefficients of variation were, respectively, J. Am. Acad. Child Adolesc. Psychiatry, 31:4, July 1992
SLEEP AND GH IN ADOLESCENT DEPRESSION
T ABLE I. Demographic and Clinical Variables Major Depressive Disorder (N = 44)
(N = 37)
21/23 27/6/11 14.8 ± 2.0 44.5 ± 17.9 6/38 20/24 20124
23/14 19/1 0/8 15.3 ± 1.5 38.4 ± 20.3 NA NA NA
1.69 X2 = 2.28 tJ9 = -1.29 U = 791.5 NA NA NA
:":0.19 :":0.32 :":0.20 :":0.24 NA NA NA
N = 20
NA NA NA NA NA NA
NA NA NA NA NA NA
NA NA NA NA NA NA
Sex: Male/female Race: White/black/other Age (X ± SD) Socioeconomic status (X ± SD) Inpatients/outpatients Suicidallnonsuicidal Endogenous/nonendogous Comorbidity: Anxiety Conduct disorder Obsessive compulsive Psychotic symptoms Mean K-SADS 9-item depression (X ± SD) Extracted Hamilton Rating Scale (X ± SD)
9.6% and 12.4% at 1.5 ng/mL, 3.2% and 3.7% at the level of 7.5 ng/mL, and 2.6% and 3.9% at the level of 33.3 ng/mL. Data Analysis All variables were examined using the W statistic (Shapiro and Wilkes, 1965) to determine if they fit a Gaussian distribution. If the variable was not Gaussian, a logarithmic transformation was used to normalize it before applying parametric statistics. The demographic variability from the two samples were compared using pooled and separate t tests, X2 , and Fisher's exact test as appropriate. Summary measures of the 24-hour GH secretion included: area under the curve (AUC), using the trapezoidal rule for the entire 24-hour period; maximum or peak secretion during the 24hour sampling; AUC for the first half of sleep (4 hours postsleep onset); AUC for the sleep period (8 hours postsleep onset); and the percentage of growth hormone (for the 24-hour period) that was secreted during sleep. All between group comparisons were made using pooled or separate t tests, as appropriate, with a = 0.05. Relationships between the dependent growth hormone variables and other continuous variables such as age were tested using Pearson correlations (after log transformations to achieve Gaussian distribution, when necessary). One-way analysis of variance with covariance methods were used to control for possible confounders such as age, sex, and to test interactions. Results Sample
From the original sample of 88 adolescents participating in the protocol, six subjects (two MDD and four normal controls) were eliminated because of missing data (difficulties with blood draws from the intravenous catheter). One additional subject (MDD) was dropped because of erratic GH assay values (>3 SD from the mean). This left a final sample of 44 MDD adolescents and 37 normal controls. There were no significant group differences in age, race, J. Am. Acad. Child Adolesc. Psychiatry, 31 :4, July 1992
N=7 N=3 N=7 3.5 ± 0.6 23.2 ± 6.7
Normal Statistic
x2 =
p
sex, or socioeconomic status (Hollingshead Scale) as shown in Table 1. Within the MDD group, 20 met criteria for endogenous subtype, six were inpatients, seven had psychotic features, and 20 were suicidal (12 had attempted suicide and eight had significant suicidal ideation with a definite suicidal plan but had not made a clear attempt). Comorbid anxiety (generalized anxiety, separation anxiety, or phobia) was present in 20 subjects, conduct disorder in seven, and obsessive compulsive disorder in three. The mean severity score using the nine-item K-SADS scale was 3.5 ± 0.6, with the extracted Hamilton Depression Rating Scale (Williamson et a!., in press) mean score was 23.2 ± 6.7. GH Results
There were no significant group differences in any of the summary GH variables as shown in Table 2, including peak GH, 24-hour integrated levels (AUC), mean GH over the first half of the night, mean GH over the sleep period, or the percentage of total GH secreted during the sleep period. As shown in Figure 1, the overall secretory patterns between the groups were quite similar. We also examined the effects of age and sex on the GH measures. Age was negatively correlated with GH secretion during sleep: r = -0.35, p < 0.01; and total 24-hour GH secretion; r = -0.24, P < 0.03. Sex was a significant covariate for total 24-hour GH secretion (F 1•77 = 5.2, P S 0.03), and percentage of GH secreted during sleep (F 1•77 = 8.4, P S 0.01), with the females secreting more total GH, but a lower ratio of sleep GH. With the addition of age and sex as covariates, MDDs were still not significantly different from normal controls on these
measures. Analysis of age by diagnosis interaction revealed one significant effect. Within the MDD group, the percentage of GH secreted during sleep showed a negative correlation with age (r = -0.45, p < 0.01), compared to the normal group (r = +0.16, P = NS); the interaction was found to be significant (F 1•77 = 6.59, p S 0.03). There were no significant interactions between diagnoses and sex or age and sex. The MDD group was dichotomized into suicidal and nonsuicidal subgroups, using a score of 4 or greater on the 617
DAHL ET AL. TABLE
2. Growth Hormone Summary Variables of Major Depressive Disorder vs. Normal Comparisons
Major Depressive Disorder (N = 44) 24-hour area under the curve 24-hour peak Area under the curve first 4 hours of sleep Area under the curve 8 hours of sleep Area under the curve 8 hours of sleep (as a percentage of 24-hour secretion)
250.2 ± 116.5 21.3 ± 9.9 92.6 ± 42.7 131.3 ± 65.5 51.9± 13.9
K-SADS items for suicidality (this indicates a definite suicidal plan or an actual suicidal attempt and was used as the cutoff in previous analyses by Ryan et al., 1988 and Dahl et al., 1990). This comparison revealed a significant blunting in sleep-stimulated GH secretion in the suicidal MDD group as illustrated in Figure 2. Summary GH variables (shown in Table 3) also indicated that the suicidal subgroup secreted significantly less GH in the first half of the night compared with the nonsuicidal group (76.4 ± 36.9 vs. 106.1 ± 43.2; t42 = -2.43, P ~ 0.02). Including age and sex as covariates did not change the significance of this finding (F 1,42 = 7.1, p ~ 0.02). The suicidal group also secreted less GH over the total sleep period (107.1 ± 55.0 vs. 151.5 ± 67.7; t42 = -2.37, p ~ 0.03) and a trend for less total 24-hour GH than the nonsuicidal group (t42 = 1.88, P ~ 0.07). Splitting the MDD group on endogenous subtype, comorbidity of anxiety, conduct disorder, psychotic features, or inpatient status did not reveal significant subgroup differences for GH secretion. An analysis of severity of depression (using K-SADS scores) showed no significant correlation between severity of depression and GH summary values. Discussion
14,---------------------,
,..,
MDDs(n=44)
I'
Normals (n=37)
3 FIG.
618
5
7
256.0 21.6 96.4 146.7
± ± ± ±
179.9 10.5 76.8 109.3
54.4± 15.1
Statistic
p
=
0.21 -0.06 0.89 -0.13
s; 0.83 S; 0.95 S; 0.38 S; 0.90
=
-0.78
s; 0.43
t79
=
t79 t79 t79
=
t79
=
subgroup. Other subgroup splits did not reveal significant GH differences. (However, the inpatient sample in this study was very small, N = 6.) The subgroup of suicidal MDD adolescents had also revealed blunted GH secretion after 1M DMI stimulation (Ryan et al., 1988). As discussed by Ryan, these results may indicate either a noradrenergic or serotonergic alteration of GH secretion in the suicidal MDD adolescents. Increased somatostatin (endogenous inhibitor of GH secretion) is also a possible mechanism for findings of blunted GH response across multiple tests. These results must also be considered in the context of other related GH, particularly in relation to developmental influences. As stated in the introduction, Puig-Antich et al., had shown that prepubertal MDD subjects had increased sleep-stimulated GH compared to controls. Kutcher et al., (1988) also reported increased nocturnal GH associated with MDD in a small sample of adolescents and controls. Jarrett et al. (1990), however, showed that adults with recurrent MDD had significant blunting of sleep-stimulated GH that persisted upon drug-free recovery. These findings suggest that age and development may be playing a central role in the effects of depression on sleep-stimulated GH. Also, as
There were no significant group differences in sleep-stimulated GH or total 24-hour GH secretion between the MDD and normal groups. Splitting the MDD group on suicidality revealed significant blunting of sleep GH in the suicidal
I I
Normal (N = 37)
9 11 13 15 17 Hours Post Sleep Onset
19
1. Twenty-four hour GH aligned by sleep onset.
21
23
presented in the introduction, pharmacological tests of GH secretion (including insulin, clonidine, and DMI) have revealed blunted GH associated with MDD across prepubertal, adolescent, and adult studies (Annseau et al., 1984; Eriksson et al., 1988; Jarrett et al., 1990; Jensen and Garfinkel, 1990; Lesch et al., 1988; Matussek et al., 1980; PuigAntich 1984a, b, c, d; Ryan, 1988). Because blunted GH response to pharmacological stimulation appears stable across development, the variance in sleep GH findings across puberty raises questions with respect to how development may be influencing the regulation of sleep and GH in MDD. A few possible mechanisms must be considered. There is evidence that sleep, pharmacological probes, and exercise act through separate neurotransmitters or peptide mechanisms in stimulating growth hormone secretion at the pituitary (Mendelson, 1982). For example, adrenergic modulation of GH secretion appears to be much less significant during sleep (Lucke and Glick, 1971). It is possible that the age effects on sleep and GH findings represent changes in one specific component of this network of regulation. Another possible explanation, however, is that age-related changes in sleep GH findings are reflecting developmental changes in the regulation of sleep itself. This paradigm of age-related changes in sleep as a central J. Am. Acad. Child Adolesc. Psychiatry, 31 :4, July 1992
SLEEP AND GH IN ADOLESCENT DEPRESSION
Non-suicidal (n=24) Suicidal (n=20) Normals (n=37)
3
FIG.
5
7
9 11 13 15 17 19 21 23 Hours Post Sleep Onset 2. Twenty-four hour GH (suicidal split) aligned by sleep onset.
component to these findings is consistent with results from the other sleep and neuroendocrine studies described in the introduction. Specifically, EEG sleep changes were not evident in the prepubertal group studied during the episode of depression (Puig-Antich et aI., 1982). In the adolescent sample, EEG sleep differences occurred in the suicidal inpatient subgroup clustered around sleep onset (increased latency to sleep and decreased latency from sleep onset to the first REM period) (Dahl et aI., 1990). The baseline night cortisol measures followed the same pattern-there were no significant differences in prepuberty, but group differences appeared in the subgroup of suicidal adolescent MDD subjects in the interval of time around sleep onset (at a time when cortisol is normally suppressed by sleep onset, the suicidal MDD subjects had increased cortisol) (Dahl et aI., in press). These results from baseline GH measures also fit this pattern, with blunted GH findings emerging around sleep onset in the subgroup of suicidal adolescents with MDD. Taken together, these results suggest dysregulation around sleep onset. All of the findings (GH, cortisol, and EEG sleep) are consistent with a sleep onset mechanism which is weakened (or impaired) in the subgroups of affected adolescents. That is, because sleep onset stimulates growth hormone and suppresses cortisol, impaired sleep onset is consistent with the blunted growth hormone and increased cortisol. The increased latency to sleep onset and TABLE
reduced REM latency seen in these studies also fit with this model. The deep, slow wave sleep in the first 1 to 2 hours after sleep onset delays the onset of the first REM period, thus dysregulation at this time may allow an earlier REM period, resulting in reduced REM latency. This paradigm of dysregulation in the first 1 to 2 hours of sleep also fits well with psychobiological studies in adult MDD, as discussed previously by Kupfer et al. (1986). This paradigm is also consistent with our knowledge of the age and developmental influences on sleep regulation. Across the human life span, increasing age is strongly associated with a steady decrease in the duration and intensity of slow wave sleep, decreased efficiency and continuity of sleep, decreased threshold of arousal, and decreased REM latency (Dement et aI., 1982; Williams et aI., 1974; Zeppelin, 1983). Thus, as many previous investigators have noted, EEG sleep changes associated with MDD mimic, to some degree, changes in sleep with normal aging (Reynolds and Kupfer, 1987). A recent meta-analysis by Knowles and MacLean (1990), reviewing studies of EEG sleep and MDD, indicated that normal age-related changes in sleep appeared to be accelerated in the MDD subjects. One possibility is that at the early end of the age spectrum, sleep is "protected." Children are very deep sleepers, have high sleep efficiency, are difficult to arouse from sleep, and have long sleep durations (Busby and Pivik, 1985; Carskadon et aI., 1987; Williams et aI., 1974). There are substantial changes in the regulation of sleep across adolescence, with a 40% decrease in slow wave sleep, 30% to 40% decrease in REM latency, and significantly increased daytime sleepiness (Carskadon et aI., 1983; Carskadon and Dement, 1987; Dahl et aI., 1990; Williams et aI., 1974). The emergence of disturbances in GH, cortisol, and EEG sleep in a subgroup of MDD adolescents may be caused by a physiological interaction between depression and developmental changes in the regulation of sleep resulting in dysregulation around sleep onset. These physiological disturbances near sleep onset may also provide a potential focus contributing to treatment and/ or prevention of adolescent MDD. Recent evidence from the National Institute of Mental Health epidemiological catchment area study of sleep disturbance in psychiatric disorders in adults (Ford and Kamerow, 1989), highlighted evidence that sleep disturbances may be critical in the pathogenesis of depression. That report also raised the possibility that early recognition and treatment of sleep disturbances
3. Growth Hormone Summary Variables of Major Depressive Disorder (MDD) Suicidal vs. MDD Nonsuicidal Comparisons
MDD Suicidal (N = 20) 24-hour area under the curve 24-hour peak Area under the curve first 4 hours of sleep Area under the curve 8 hours of sleep Area under the curve 8 hours of sleep (as a percentage of 24-hour secretion) J. Am. Acad. Child Adolesc. Psychiatry, 31 :4, July 1992
217.4 19.3 76.4 107.1
± ± ± ±
110.8 10.5 36.9 55.0
50.1 ± 14.8
Nonsuidical (N = 24) 277.5 23.0 106.1 151.5
± ± ± ±
116.2 9.2 43.2 67.7
53.3 ± 13.3
Statistic
p
=
-1.88 -1.25 -2.43 -2.37
:":0.22 :":0.02 :":0.03
142 =
-0.74
