Early-Onset Mood Disorder

David Brent, M.D., Neal Ryan, M.D., Ronald Dahl, M.D., and Boris Birmaher, M.D.

During the past decade, tremendous advances have been made in our knowledge of the natural history, risk factors, and psychobiology of early-onset affective illness. In this chapter, the epidemiology, natural history, and adult sequelae of early-onset affective illness are reviewed, followed by a discussion of family-genetic and high-risk studies of juvenile affective illness. The status of investigation into biological markers in prepubertal and adolescent depression is discussed, focusing on studies of secretory patterns of cortisol and growth hormone, provocative neuroendocrine challenge studies, and electroencephalographic (EEG) sleep studies. Emphasis is given to clinical and developmental variables that might explain differences between studies or apparent discontinuities between these studies and those of adults with depression. The psychopharmacological treatment studies for prepubertal and adolescent depression are reviewed, and the possible reasons for negative findings in nearly all of them are discussed. Finally, areas of further investigations that may be particularly fruitful for the field are suggested.

Affective illness is less common in prepubertal children than in adolescents. In prepuberty, the point prevalence of major depression has been estimated to range from 1.8 to 2.5%, "minor" forms of depression, including dysthymia, as 2.5%, and bipolar illness as 0.2–0.4% (36). For adolescents, the point prevalence of depression has been estimated at 2.9–4.7%, dysthymic disorder, 1.6–8.0%, and bipolar disorder, 1%. Depressive illness is equally common in males and females prior to puberty, but it more common in females than males post-pubertally (36). On the other hand, bipolar disorder is equally prevalent in both genders.

Early-onset affective illness has become a much more significant public health problem in recent years. There is evidence of a secular trend for depression and suicide. Successive cohorts born after 1950 have an increased risk for depression and suicide relative to earlier birth cohorts, based on retrospectively and prospectively gathered data (50,64,116), with evidence of a secular trend for depression affecting even prepubertal and adolescent children (114).

In clinically referred samples, major depression in childhood and adolescence is a chronic and recurrent condition. According to the results of one naturalistic study of depressed children ages 8 to 13, untreated major depression lasted an average of 7.2 months, and dysthymic disorder, an average of 45.9 months (66). Subsequent follow-up of these children revealed that a 40% cumulative probability of a recurrence of depression could be expected within two years, reaching 72% within five years (67). Earlier age of onset and underlying dysthymic disorder both increased the risk of depressive recurrence (67). Comorbidity with anxiety disorder or conduct disorder did not influence the course of depressive illness (70,71), although other clinical studies suggest that comorbid anxiety disorder and substance abuse may predict longer episode length and greater functional impairment (80,115). Certain features of adolescent depression are predictive of the development of bipolar illness, including: (a) a symptom cluster consisting of rapid onset, psychomotor retardation, and mood-congruent psychotic features, (b) pharmacologically induced hypomania (121), and (c) family history of bipolar disorder and three generations of family members with affective illness. This last circumstance has also been shown to predict bipolarity in prepubertal depressives (41). Comorbid conduct disorder is a common feature of early-onset bipolar disorder, often antedating the onset of bipolarity (71,122).

Both prepubertal and adolescent depressive disorders are associated with social impairment with peers, in school, and with parents (97,98). In a follow-up of remitted prepubertal depressives, some areas of functional impairment persist, particularly with respect to mother-child discord (99). There is evidence that, over time, early-onset depressives are likely to develop additional comorbid conditions, namely conduct disorder, anxiety disorder, and substance abuse (58,68,70). However, anxiety disorders are more likely to antedate early-onset major depression than to develop secondary to depression (68).

Other studies have also supported the view that early-onset depression persists into adulthood. A nine-year follow-up of adolescents with prominent depressive symptomatology predicted the persistence of depression and functional impairment in school, work, and interpersonal relations with spouses and parents (58). Harrington et al. (52) found that child and adolescent-onset depressives were four times more likely than non-affectively ill psychiatric controls to have had an affective disorder on follow-up. Of those with early-onset affective illness who had a second episode before age 17, 95% had a subsequent episode in adulthood. Earlier age of onset is associated with a very high risk of recurrence (67). Pure depression (e.g., no comorbid conduct disorder) tends to persist into adulthood, while those depressives with comorbid conduct disorder do not appear to be at increased risk for depression upon follow-up, but have a course more similar to those with "pure" conduct disorders (35,53). Rao et al. (104) reported a high rate of recurrence in a seven-year, controlled follow-up of adolescent depressives and found that recurrence was associated with significant psychosocial sequelae.

Suicidal behavior and completed suicide are among the most significant sequelae of depression. Clinically significant suicidality (i.e., suicidal ideation with a plan or actual suicide attempt) is quite common among referred samples of affectively disordered youth and is associated with comorbid substance abuse, other non-affective disorders, longer duration of disorder, family discord, and hopelessness (69,89,111). Furthermore, depression has been shown to be the single most significant clinical risk factor for completed suicide in adolescence in several case-control studies using a psychological autopsy approach (10,11,116,117). Some, but not all, studies also suggest that bipolar disorder, particularly in a mixed state, conveys significant risk for completed suicide (10,11). Longitudinal studies of children with early-onset affective illness have also demonstrated a high risk for both attempted and completed suicide associated with affective illness upon follow-up (52,69,105), although the rate of re-attempt was heightened if the affective illness was associated with non-affective comorbidity (69).

Several family-genetic studies have demonstrated continuity between child and adult affective disorders. Puig-Antich et al. (96) conducted a blinded family history study that compared the morbid risk of disorders in the relatives of prepubertal major depressives, non-affectively ill "neurotic" psychiatric controls, and normal controls. The rates of depression were higher in the relatives of the depressive probands, particularly in male relatives. Looking within the major depressive group, familial loading for depression was greater in endogenous than in non-endogenous depressives, and the rate of mania in first-degree relatives was higher in psychotic than non-psychotic depressives. Major depression, when comorbid with conduct disorder, was associated with a lower familial morbid risk of depression than when depression occurred without conduct disorder (0.14 versus 0.39); the same was true when comparing the morbid risk of depression in the relatives of suicidal versus non-suicidal prepubertal depressives (0.20 versus 0.46). A companion family history study of adolescent depression also showed familial aggregation of depression (127).

Strober et al. (122) reported on a family study of bipolar I disorder in children and adolescents, with schizophrenic probands as controls. Higher familial rates of bipolar disorder were found in the relatives of both prepubertal and adolescent-onset bipolars, compared to schizophrenics. The morbid risk of bipolar disorder was 3.5-fold higher in the relatives of prepubertal-onset bipolars than in those of adolescent-onset bipolars. Prepubertal onset was also associated with greater non-affective comorbidity and resistance to the therapeutic effect of lithium.

Results from several high-risk studies have revealed that the offspring of affectively ill parents are at higher risk for affective illnesses than psychiatric or normal controls (51,88). Both an increased risk and an earlier onset of disorder were noted. Orvaschel et al. (87) recently demonstrated that the greatest risk for the development of affective illness is in offspring who come from families with affectively ill parents and grandparents. On follow-up, at least half of the offspring of affectively ill parents will experience a depression by late adolescence (51,125). Parent-child discord, low family cohesion, and affectionless control were associated with development of conduct disorder and substance abuse in offspring of parents with a history of depression, but these variables were not associated with the onset of depression or anxiety. However, in the offspring of parents without history of depression, these factors were associated with depression, as well as with conduct disorder and substance abuse (34). Greater functional impairment and more prolonged depressive episodes in high-risk children and adolescents were associated with the degree of exposure to parental affective illness, early age of onset of affective illness in the parent, chronicity, severity, recurrence of the illness in the parent, psychopathology in the non-affectively ill parent, and parental divorce (62,124).

Several high risk studies of offspring of bipolar disordered parents also indicated familial transmission of bipolar disorder. Akiskal et al. (1), in an uncontrolled, high-risk study of the offspring and siblings of bipolar patients, found that fully 50% showed signs of bipolarity, often manifested by cyclothymic disorder. One controlled, high-risk study found an association between parental bipolarity and affective and cyclothymic disorders (63), whereas another found a more non-specific increase in a broad range of non-affective disorders (49). The severity and chronicity of parental bipolar disorder, as well as psychopathology in the non-bipolar parent, were associated with psychopathology and functional impairment in offspring (49).

These family-genetic and longitudinal high-risk studies support a significant role for the familial transmission of affective illness, probably through both environmental and genetic mechanisms. These studies also support the view that depression is heterogenous in etiology. Disorders that are both most recurrent and familial appear to be psychotic depression, endogenous depression, and bipolar disorder, whereas the least familial and recurrent disorders are depression comorbid with conduct disorder. It may be that the secular trend in affective illness is accounted for primarily by the latter type of early-onset depressives, who are genetic "phenocopies", presenting with comorbid conduct disorder, substance abuse, and suicidality.

Other life stressors may also be associated with risk for depression, including loss of a parent (126), loss of a sibling or friend to suicide (12), or physical or sexual abuse (59,60). With respect to loss, depression is much more likely to occur if a personal or family history of depression is present (12,126). Parent-child discord has also been associated with both prepubertal and adolescent depression, but it is hard to discern to what extent discord is a cause of depression or a by-product of the irritability of depression in the child and its frequent co-occurrence in an affectively ill parent (97,98,99).

Chronic medical conditions have been associated with an increased risk of early-onset depression and depressive symptomatology. These include juvenile-onset diabetes, inflammatory bowel disease, and epilepsy (9,13). The etiologies of depression in these diverse groups are likely to be heterogenous and to relate to the general stress of chronic illness, neurobiological interplay between systemic, metabolic diseases and psychopathology, and the effects of medications such as steroids and phenobarbital on mood (9).

The search for psychobiological markers is exceedingly important, both for research in early-onset affective illness and for a better understanding of affective illness during the life span (109). First, such approaches have helped validate the diagnosis of major depression in prepubertal patients. Second, such markers may shed light on the pathophysiology of depression by indicating which neurotransmitter systems may be most intimately involved with depression. Third, markers may be useful in predicting the course, symptomatology (e.g., mania, psychosis, suicidal behavior), risk of recurrence, and treatment response in patients with early-onset depression. Fourth, the identification of biological markers may make it possible to study more homogeneous populations for family-genetic and treatment studies. Fifth, the study of these markers in children at high risk for depression may identify trait markers, thereby enabling researchers and clinicians to identify who is vulnerable to depression before it occurs. This may lead to greater understanding of the etiology of depression and to the development of preventive strategies. Furthermore, the identification of trait markers may lead to more informative family-genetic studies which include children who have not yet passed through the age of risk. If such children are unaffected, but have a given marker, then they may be counted as informative family members, thereby increasing the power of genetic-linkage studies. Sixth, the study of psychobiological markers in prepubertal depression enables one to examine the biological correlates of affective illness unadulterated by substance abuse, personality disorder, psychopharmacologic exposure, or chronic medical illness. Furthermore, longitudinal studies of biological probes in early-onset depression can shed light on the interactions among factors such as normal development, environment, and intrinsic vulnerability to depression. In specific, the longitudinal study of prepubertal depressives (and those at high risk for depression) through pubescence is critical to an understanding of the relationship between puberty and increased risk of depression. In studying the developmental sequence of the appearance of psychobiological markers, it may be that early-onset markers are more intrinsic to the pathophysiology of depression, whereas later appearing markers may be more related to dysregulation that occurs with chronic stress (92,109). Among the studies to be reviewed are neuroendocrine and sleep studies of prepubertal children and adolescents with major depression.

Cortisol hypersecretion is rare in prepubertal children with major depression, occurring in only about 10% of cases (91). This was true for prepubertal depressives when comparing 24-hour cortisol secretory patterns during an episode and in recovery, and when comparing these patients to psychiatric and normal controls. A subgroup of melancholic children may exhibit hypersecretion of cortisol (5). In contrast, nocturnal cortisol levels were actually lower in depressed children (31).

In a study comparing adolescent outpatients with depression and normal controls, no group differences were found with respect to the 24-hour pattern of cortisol secretion (20,74). In a second sample of adolescent inpatients and outpatients, Dahl et al. (25) examined the 24-hour cortisol secretory pattern by symptomatic subgroup of adolescent depressives. In this study, inpatient and suicidal subgroups of patients did not have the usual nadir of cortisol secretion near sleep onset, compared to other adolescents with depression and to normal controls. It is uncertain whether this subgroup was under greater stress, which led to both suicidality and inpatient status, whether inpatient status allowed for greater circadian entrainment, thus revealing this abnormality, or whether these findings represent an intrinsic biological difference between suicidal and non-suicidal adolescents with major depression. The results suggest that cortisol hypersecretion associated with depression is developmentally mediated, is rarely seen in childhood, emerges inconsistently with adolescence, and is not observed commonly until adulthood. However, additional analyses of these data, which incorporated 5- to 9-year clinical follow-up, indicated that elevated cortisol near sleep onset was associated with future depressive episodes (103). Depression in maltreated children was associated with the loss of the expected diurnal pattern of cortisol secretion (54,60). While these studies did not examine 24-hour cortisol secretion, they do point to the possibility that severe stress in a subgroup of children may lead to both depression and cortisol hypersecretion, and that in fact, the nature of the stress response may predict which children under severe stress will develop affective illness.

The early claims for the dexamethasone suppression test (DST) as a sensitive and specific test for endogenous major depression have been qualified considerably in more recent studies of adults (3). Overall, as reviewed by Dahl et al. (18), Casat et al. (14), and Ryan and Dahl (109), the sensitivity of the DST across studies was low in both children (58%) and adolescents (44%). Studies on inpatients yielded much greater sensitivity than did studies on outpatients (61% vs. 29%). Specificity was somewhat higher in adolescents than in children, when compared to normal controls (84% vs. 74%) and to psychiatric controls (solely inpatients, 85% vs. 60%).

Other clinical and methodological variables were associated with DST non-suppression, most notably concurrent suicidal behavior (90,107), subsequent suicidal lethality (107), and endogenous subtype (106). The association of suicidality with DST non-suppression among inpatients was consistent with a report by Dahl et al. (25) of a loss of the usual decline in cortisol secretion at sleep onset in suicidal adolescent inpatients with major depression. The studies associating psychotic and endogenous subtypes have not been replicated in DST studies that employed an indwelling catheter to obtain blood samples (8,18). This may be related to the fact that these studies were conducted primarily on outpatients in a relatively low-stress environment. As in studies of 24-hour cortisol secretion, it is likely that age, pubertal status, and degree of stress interact to yield abnormalities in cortisol secretion in depression, and that such abnormalities are less common in children and among outpatients than among adolescents and inpatients.

Corticotropin-releasing hormone (CRH) infusion results in release of adrenocorticotrophic hormone (ACTH) from the pituitary, followed by cortisol secretion from the adrenal. In adults with depression, the ACTH response was blunted after CRH challenge, despite normal to high cortisol levels (46). This was not the case in studies of prepubertal children with major depression. Birmaher and colleagues (5) found no differences between prepubertal major depressives and normal controls for baseline cortisol or ACTH, or the response of cortisol or ACTH to CRH. However, subgroup analyses revealed that ten inpatient depressed children and four melancholic depressed subjects had significantly lower ACTH levels, but normal cortisol, after CRH, consistent with findings in depressed adults. Another related subanalysis in this sample found that prior sexual abuse was associated with a reduced response of ACTH to CRH (61), a finding also reported by DeBellis et al. (30) in a different sample. In both of the above-noted samples, the abused children were living in currently stable environments. In contrast, abused depressed children living in conditions of chronic adversity demonstrated increased ACTH secretion following a CRH challenge.

Taken together, the findings from 24-hour cortisol secretion studies and the DST and the CRH challenge studies suggest that abnormalities of cortisol secretion are relatively rare in children and begin to emerge, albeit inconsistently, in adolescence. Children may be developmentally protected against manifesting these abnormalities associated with adult depression, and such abnormalities may only consistently emerge after exposure to severe stress (e.g., physical or sexual abuse), with past and ongoing stress possibly exerting different effects.

Several critical questions emerge from contemplation of these results: 1) At what point in adolescence or adulthood do psychobiological abnormalities of cortisol homeostasis consistently emerge? 2) Why are children with depression "protected" from more commonly manifesting abnormalities in cortisol homeostasis? 3) What are the pathways from stress to perturbations in cortisol dynamics to the development of depression? Longitudinal studies of depressed and control samples with careful measures of both exogenous stress and the above-noted psychobiological probes will be critical to addressing these questions. On the basis of recent work, it is critical that we take into account the presence of multiple feedback loops involving the hypothalamus, pituitary, and adrenals (13).

Consistent evidence exists for blunted growth hormone (GH) response to provocative pharmacologic challenges in prepubertal depressed patients. Puig-Antich et al. (100) showed a diminished growth hormone response to insulin-induced hypoglycemia in prepubertal children with endogenous major depression, compared with a non-depressed, neurotic psychiatric control group. This reduction in GH response to insulin persisted after a medication-free period of recovery (101). Furthermore, this abnormality has also been found in a subset of prepubertal children who, by virtue of familial loading, were at high risk for depression but were as yet unaffected (Ryan, unpublished data). Taken together, these studies support the view that this finding represents a trait marker for major depression. Jensen & Garfinkel (57) found blunted growth hormone response to oral clonidine and L-DOPA in prepubertal depressed vs. normal subjects. A similar response was observed by Meyer et al. (85) to arginine, oral clonidine, and to insulin-induced hypoglycemia, but not to L-dopa or 5-hydroxytryptophan in a comparison of 7–13-year-old depressed boys and age-matched, normal controls.

Ryan and colleagues (110) replicated and extended the findings of Puig-Antich et al. (100,101) in a comparison of 38 prepubertal patients with major depression and 19 normal age- and Tanner stage-matched controls. As in the previous studies, GH release was blunted after insulin-induced hypoglycemia in the depressed subjects, but the effect was unrelated to endogenicity. A trend was found for reduced GH release with administration of 1.3 mg/kg of intravenous clonidine, although this result did not attain statistical significance. Interestingly, GH release was also significantly diminished after challenge with growth hormone releasing hormone (GHRH). This unexpected finding suggested that the abnormality in GH regulation was not solely at the level of the a2 adrenergic system in the hypothalamus. The origin of this abnormality is currently under investigation by Ryan and colleagues, who are examining the role of somatostatin and insulin-like growth factor (somatomedin C) in this process.

Less has been published on provocative GH challenges in adolescents with major depression. What has been published is not as conclusive as the above-noted findings in prepubertal children. Ryan et al. (113) found blunted GH response to desmethylimipramine (DMI) in depressed adolescents, compared with normal controls, but the difference between groups was entirely accounted for a reduction in GH response in depressed, suicidal adolescents. Jensen & Garfinkel (57) found no difference in the GH response to oral clonidine or L-dopa in a small sample of adolescents (8 depressed and 5 normals). The variability of the results obtained in adolescents, relative to prepubertal children, may be attributable to the mediating effects of sex steroids on GH response to provocative challenges. For example, Matussek et al. (82) noted that the GH response to clonidine was most consistently diminished in postmenopausal females. In addition, some studies have indicated that estrogen potentiated the effects of exercise and arginine on the release of GH (84), whereas others have indicated that androgens, but not estrogens, potentiated the effect of GHRH on GH release (120).

These findings are consistent with those noted for adults. Adults with major depression exhibit a reduced GH response to clonidine (15,118) – a finding that persists after recovery (118). More recently, studies have indicated that depressed adults show a blunted GH response to GHRH (79,86), which was related to a similar GH response after clonidine (79) and DMI (86) administration and to higher levels of circulating somatomedin (79).

Puig-Antich et al. (94,95) found that prepubertal patients with major depression hypersecreted GH during sleep compared to non-depressed, neurotic controls, both during an episode and after a sustained, medication-free recovery. Meyer et al. (85), in a comparison of 24-hour patterns of GH secretion in depressed and normal boys, found decreased GH secretion association with depression, both during the daytime and nocturnal phases of GH secretion. More recently, DeBellis et al. (31) found no difference between prepubertal depressed children and normal controls in nocturnal growth hormone secretion.

Studies of GH secretion in adolescent major depression are also inconsistent. Kutcher et al. (75), in a comparison of nine depressed adolescents and nine normal controls, found increased nocturnal secretion of GH, mostly in the first half of the night (midnight and 1:00 a.m.). This result was replicated in a second study comparing 12 adolescent depressives and 12 normal adolescents (74). In both studies, nocturnal GH hypersecretion was unrelated to suicidality or endogenicity. In contrast, Dahl and colleagues (26) found decreased nocturnal growth hormone in depressed, suicidal adolescents, compared to normal controls, although non-suicidal depressed adolescents were no different than normal controls. Earlier timing of GH secretion relative to sleep onset was associated with depression in prepubertal children, adolescents, and adults, and predicted recurrence of depression in adults (19,38,128).

In the only study reported to date of serotonergic functioning in early-onset depression, Ryan et al. (108) compared 37 depressed children with 23 psychiatrically normal children using an intravenous challenge of L-5 hydroxytryptophan (L-5HTP), preceded by oral preloading with carbidopa to prevent peripheral degradation of L-5HTP. In response to the L-5HTP challenge, the depressed children showed significantly a greater prolactin response and a smaller rise in cortisol, compared with controls. Cortisol response was unaffected by gender, but the difference in prolactin secretion between depressives and normals was seen only in females. There was no association between suicidality and either prolactin or cortisol response to L-5HTP, nor were these two markers correlated with aggression, endogenicity, or melancholia. Children at high risk for depression because of familial loading had a profile similar to depressed children using the identical challenge paradigm, suggesting that this response may antedate the onset of depression (7).

These findings are consistent with studies in adults, insofar as several other studies have shown differences between adult depressives and controls with respect to neuroendocrine responses to serotonergic probes. Similar to the findings of Ryan et al. (108), Maes et al. (81) found enhanced prolactin response to oral L-5HTP in melancholic women. In contrast to the blunted cortisol response reported by Ryan et al. (108), enhanced cortisol response to L-5HTP was associated with depression in another study (81). While Ryan et al. (108) found gender differences for prolactin response, but not cortisol, Maes et al. (81) reported that female, but not male depressives differed from controls on both prolactin and cortisol response to L-5HTP.

The variability in responses reported in different studies may be a function of age, gender, the specificity of the serotonergic probes, and the balance of different subpopulations of serotonergic receptors in a given subject. The latter is likely to be the case, because there is evidence that both prolactin and cortisol responses to serotonergic agonists are governed by both pre- and post-synaptic effects.

Sleep EEG changes are one of the most frequently found psychobiological markers associated with depression in adults (65). These include reduced rapid eye movement (REM) latency, decreased delta (stage 3 and 4) sleep, sleep continuity disturbances, increased REM density, and more even distribution of REM density across the entire night, as compared to the pattern of increased REM pressure during the second half of the night in normal individuals. While these sleep characteristics are typical in adult depressives, they are much less consistently found in juvenile populations.

Puig-Antich et al. (92) compared 54 prepubertal depressives, 25 psychiatric controls with "emotional" disorders, and 11 normal controls and found no group differences with respect to any of the sleep parameters noted above. In a follow-up study (93), 28 fully recovered prepubertal depressed children were studied in a medication-free state. The recovered depressed subjects showed shorter REM latency upon follow-up compared to within-subject, in-episode data and to the two control groups upon follow-up. However, those followed up, compared to those who were not, had more normal sleep to begin with, so that the interpretation of these results is not straightforward. To our knowledge, a replication of this study has not been published. In contrast, Emslie et al. (33) compared 35 hospitalized depressed children and 20 age-matched controls and found evidence of decreased REM latency, increased REM time, and increased sleep latency in the depressed group. Dahl et al. (21) compared 36 prepubertal depressed children with 18 age- and Tanner staged-matched normal controls. Overall, there were no group differences, but one subgroup of 8 depressed subjects had reduced REM latency, decreased stage 4 sleep, increased REM time, and symptomatically were more severely depressed and more likely to be melancholic. Recent findings by our research group with a larger sample of depressed children and controls found increased sleep latency, and decreased delta sleep in the depressed sample, although the effect size was small (22). In the same study, there were no REM differences despite the large sample size. However, when depressed and normal control children were challenged with a cholinergic agonist (arecoline), the depressed children had significantly shorter REM latency (24), perhaps "unmasking" REM latency differences usually not seen in depressed young subjects.

The picture in adolescent depression appears to be intermediate between that of prepuberty and adulthood. Lahmeyer et al. (78), comparing 13 older adolescents with depression and 13 age matched controls, found evidence of decreased REM latency in the depressed group. Moreover, REM latency and age were negatively correlated, suggesting that this marker may not emerge until later in adolescence. Goetz et al. (45) studied 49 non-bipolar, mostly outpatient adolescents with major depression and 40 normal adolescent controls. The adolescent depressive group had longer sleep latency, but there were no differences between depressives and controls with respect to REM latency, REM density, or delta sleep. Age was correlated with REM latency and delta sleep, particularly in the depressives, compared with the controls. These results were similar to those of Lahmeyer et al. (78). A comparison of 27 adolescents with major depression and 30 normal controls revealed no age by diagnosis interactions for any of the relevant sleep parameters (23). Also, no group differences were seen when the depressed group was compared with the control group. However, the suicidal depressed adolescents showed increased sleep latency and decreased REM latency, compared to both the non-suicidal depressives and to normal controls. Substantial overlap occurred between suicidality and inpatient status, such that it was uncertain whether the findings could be attributed to suicidality or to inpatient status (23). Kutcher et al. (76) compared 23 endogenously depressed older adolescents and 23 normal controls. The depressed group had decreased REM latency and increased sleep latency. Increased REM eye movement density was predictive of depression in a 5- to 9-year clinical follow-up study of adolescent depressives (104).

A recent study by our research group evaluated sleep in depressed and normal adolescents in a protocol where all subjects followed a rigidly imposed sleep/wake schedule before and during the study. The depressed adolescents showed significantly longer sleep latencies and significantly reduced REM latencies, compared with controls (19). These results agree with a published meta-analysis (4) which showed that, in studies with clinical populations less than 20 years of age, increased sleep latency was the only sleep variable that differentiated depressed from normal subjects.

Therefore, consistent sleep EEG findings in depression do not emerge until late adolescence. When they appear earlier, the subjects are inpatients, suicidal or severely depressed, and possibly under considerable stress with past and ongoing stress possibly exerting different effects. As with cortisol HPA abnormalities, it may take an unusual amount of exogenous stress for children and adolescents to manifest sleep-EEG abnormalities. Moreover, as noted in the analyses of several of these studies (45,76,78), REM and sleep continuity changed as a function of age, with some evidence that the trend was accelerated in the depressive group. This is also consistent with results from the meta-analysis of Knowles & MacLean (65), who noted that the differences in the sleep findings between depressed and normal subjects seemed to increase as a function of age.

A growing body of literature suggests that asymmetry of activation in frontal brain regions is associated with negative affective states (27,29,37). A wide range of studies have supported the concept that relative right frontal activation is associated with negative or withdrawal-related affective states (and/or a trait with a strong tendency toward affective states). The most common experimental approach to these areas involves comparisons of EEG power in the alpha band (which is the inverse of activity). These studies indicated that greater right (relative to left) activation was associated with greater negative mood (often fear) in non-human primates, infants, toddlers, and adults (27,28,29,37). Further, studies of infants of depressed mothers found greater relative right frontal activation, compared with results from infants of control mothers (29). Recently, Graae et al. (48) reported similar right frontal EEG asymmetries in female adolescent suicide attempters, compared with normal control adolescents (48).

While controlled studies have generally failed to find differences of effect between tricyclic antidepressants and placebo in depressed children and adolescents, a relatively large study with fluoxetine (a specific serotonin reuptake inhibitor [SSRI]) found this drug to be significantly superior to placebo (32). No controlled trials have yet been done of treatment of bipolar disorder, although prepubertal-onset bipolar disease was associated with lithium resistance (122).

The largest study of prepubertal depression examined 38 depressed children assigned to either imipramine or placebo treatment (102). No difference in treatment response was found between imipramine (56% response) and placebo (68%). In a related study examining the relationship of plasma levels of imipramine and its metabolites to clinical response, Puig-Antich et al. (102) noted that non-response was associated with lower plasma level of imipramine and desipramine, severe depression, and the presence of hallucinations.

An inpatient treatment study of imipramine vs. placebo in prepubertal depressed children found that children with "pure" depression, or depression complicated by anxiety, responded much better to imipramine than to placebo (57% vs. 20%, respectively), whereas depression comorbid with conduct disorder responded better to placebo (33% vs. 67% for imipramine) [55]. A study of severely and chronically ill prepubertal depressed children also showed no difference in response rate between nortriptyline (NT) and placebo (39). As defined by a score on the Child Depression Rating Scale (CDRS) of < 20, 31% of those treated with NT vs. 17% of those given placebo responded. Using the criteria of Puig-Antich et al. (102), the response rates were 46% vs. 58% for medication and placebo, respectively.

Two methodological issues were addressed in subsequent follow-up studies by Geller et al. (39,42). First, a follow-up of subjects who had responded to placebo and hence "failed" the placebo washout phase demonstrated that the subsequent risks of depression and mania were the same in the placebo responders and non-responders (41). Moreover, these two groups were similar with respect to other clinical parameters, namely, severity, chronicity, age of onset, and comorbid psychopathology, and differed only insofar as placebo washout responders were more likely to be female. These results indicated that the placebo-washout phase may not be necessary for clinical trials in this age group.

Second, Geller et al. (43) noted an association between the development of mania and tricyclic use. If the exposure to nortriptyline was greater than 25 weeks, the risk for mania was increased two-fold. Bipolar I disorder developed only in subjects exposed to tricyclic antidepressants. However, the rate of bipolar I, II, mania or hypomania was not different between those who did and did not receive tricyclic antidepressants. However, an extension of this follow-up found no association between tricyclic exposure and the onset of mania (41). A pedigree heavily loaded with affective illness in three generations conveyed a greatly increased risk for the development of mania (40,41), as was noted a decade earlier by Strober and Carlson (121). This may be a critical treatment and design issue in samples of clinically referred, early-onset depressives, a population that is often enriched with those having a family history of bipolarity.

There have been even fewer published studies of adolescents. Ryan et al. (112), in an open-label study of adolescents, found only a 44% response rate to imipramine, with non-response associated with comorbid separation anxiety, endogenicity, and female gender. In one small, controlled study of depressed adolescent inpatients, 80% responded to amitriptyline vs. 60% of subjects who received placebo (72). Three recent studies also noted no difference between antidepressant and placebo (73,77, R. Klein, unpublished data). Geller et al. (40) studied chronically and severely ill adolescents with major depression and found only an 8% response rate for nortriptyline vs. 21% for those treated with placebo. A comparison of adolescent depressives treated with fluoxetine or placebo also found no difference between the two treatment groups, with approximately two-thirds responding in each group (119).

In contrast, a recent double-blind, randomized trial (32) compared 96 non-psychotic depressed children and adolescents treated with fluoxetine or placebo for eight weeks. A greater proportion of patients on fluoxetine, compared with placebo (56% vs. 33%, respectively; p=0.024) showed improvement on the Clinical Global Impressions scale. Significant differences were noted on the Children's Depression Rating Scale (CDRS) after five weeks of treatment. However, complete symptomatic remission occurred for only 31% and 23% of the fluoxetine- and placebo-treated subjects, respectively. This lack of difference in response to tricyclic antidepressants and placebo in early-onset depression is a critical area for further research for several reasons. The above-noted placebo-controlled study of fluoxetine (32) is the sole study demonstrating the efficacy of any clinical intervention, either pharmacologic or psychosocial, in clinically referred, affectively ill children or adolescents. Therefore, replication of this study is crucial. Antidepressants continue to be used widely, and, while most clinicians agree that on the basis of their clinical experience that treatment with antidepressants is warranted, no accepted guidelines exist for optimal dosage, length of treatment, or indications. Finally, the apparent difference in the antidepressant/placebo response in children and adolescents compared with adults raises questions as to the possibility of developmentally mediated biological differences in depression across the life span. As reviewed in this chapter, strong evidence supports a continuity between child-onset and adult depression, based on family genetics, high risk, longitudinal, and neuroendocrine challenge studies. Therefore, it is particularly puzzling why depressed children and adolescents show different responses to tricyclic antidepressants than do depressed adults.

First, it should be noted that, in general, children and adolescents do respond well to antidepressants—it is just that the placebo response rate is also high. The exception to this is the study of Geller et al. (40), where 8% of adolescents with depression responded to antidepressants, while 21% responded to placebo. This particular population was a very severely and chronically ill sample, enriched for family history of bipolar disorder—a population that, in adults, has not done well on tricyclic antidepressants (123). Puig-Antich et al. (102) also noted that those prepubertal, depressed patients with either severe depression and/or hallucinations did not respond well to imipramine. It is possible that this subpopulation might respond better to lithium or a monoamine oxidase inhibitor.

A second possibility is developmental, and two possibilities have been advanced. Primate studies have revealed that the noradrenergic neurotransmitter system continues to mature and develop throughout puberty (47). It is possible that antidepressants do not exert the same impact on developing systems as on adult systems. However, the neuroendocrine studies noted above, namely the insulin tolerance and clonidine challenge studies of GH, are consistent with a noradrenergic deficit of neurotransmission, at least in endogenously depressed prepubertal children. Therefore, one could posit that noradrenergic agents should be effective in prepubertal endogenous depressives. Fluoxetine, the sole agent with a large study demonstrating efficacy, is an almost purely serotonergic agent, while all of the tricyclic antidepressants have a strong noradrenergic component. A related issue is that perhaps the high level of circulating sex steroids in adolescents may explain their failure to respond to antidepressants. However, this would not explain the high placebo response that has been reported in some studies of adolescents (72,119).

A third possibility, as yet untested, is that younger populations of affectively ill patients are enriched for atypical depression, a group that has not responded as well to tricyclic antidepressants as to monoamine oxidase inhibitors (123). With the availability of reversible monoamine oxidase inhibitors, studies using these much safer agents should be considered.

A fourth and related issue is that endogenomorphic features have been found to predict tricyclic antidepressant response in adults (123), yet endogenous features are relatively rare in clinical samples of early-onset affectively ill patients who participate in protocols. However, endogenicity did not predict response to imipramine in the Puig-Antich et al. study of prepubertal depressives (102). In addition, endogenicity was associated with non-response in the open-label study of depressed adolescents by Ryan et al. (112).

Fifth, it is possible that the psychosocial environment of depressed children and adolescents is sufficiently different from depressed adults enrolled in clinical trials that the outcomes may differ by age. Specifically, family conflict, history of abuse, and parental depression are noted as potential risk factors for early-onset depression. It is possible that unless these potent factors are addressed in treatment, the sociotoxic impact of parent-child discord and parental depression may override any individually focused intervention, whether psychosocial or psychopharmacologic. This issue would not, however, explain the high placebo response rate seen in the reported studies.

Another, as yet untested possibility remains, raised by Geller et al. (39): that the failure to find differences in response to antidepressant vs. placebo in more juvenile populations may be related to the recent secular trend in affective illness. It is likely that the secular trend has not occurred due to changes in the genetic make-up of the population, but rather that certain exogenous factors (e.g., social stressors) have contributed to the increased risk of depression in younger populations. Therefore, it is possible that, via the secular trend, the proportion of phenocopies with depressive symptoms has increased, and that most clinical samples are, in fact, enriched with them. One would predict that phenocopies might not show high familial loading for depression, would be more likely to be suicidal and to have comorbid symptoms of disruptive disorders and substance abuse, and might not show the core differences in neuroendocrine challenge studies (e.g., L-5HTP cortisol and prolactin, GH-GHRH). Among prepubertal children, there is some preliminary evidence to support this, insofar as suicidal and conduct-disordered children had lower family loading for affective disorder (96). Also, certain neurobiological findings have been reported by some (100,101), but not others (108,109), to be present only in endogenously depressed prepubertal children. Other biological changes (e.g., in blunted or exaggerated ACTH response to CRH) are found in association with maltreatment or sexual abuse (30,60,61). Additional evidence to support the role of family discord in "phenocopies" comes from the work of Fendrich et al. (34), in which family environmental factors like "affectionless control" were associated with increased risk for depression only in offspring of non-depressed parents. Such familial risk factors were associated with comorbid substance abuse and conduct disturbance in both the high and low risk groups. Hence, the presence of comorbid disruptive disorder may signal the presence of a phenocopy with low familial loading. Longitudinal studies (35,53) also support the thesis that "pure" depression is much more likely to be continuous with adult depression than depression comorbid with conduct disorder. Additionally, Hughes et al. (55) found that the presence of conduct disorder predicted a lower response to antidepressant and a better response to placebo than were observed in the "pure" depressive group, although these findings have not been replicated by others. The role of the exogenous stressors, family loading for affective illness, and the neuroendocrine profile of early-onset depressives may be helpful in delineating the subgroups of depressed patients who will respond more favorably to antidepressants than to placebo. From the extant literature, one might posit that children with strong familial loading for recurrent unipolar depression, who present with prominent endogenous features without extensive non-affective comorbidity, would be the most likely to respond to antidepressants rather than to placebo. Conversely, such studies may point the way to more precise ways of identifying those patient who might benefit from psychosocial interventions.

During the past decade, tremendous strides have been made in our knowledge of early-onset affective disorders. The following are some of the "unsolved mysteries" in this fascinating field that should be addressed in the coming years:

1. The etiology of the secular trend in early-onset depression. As noted above, a better understanding of the secular trend has important implications for treatment and prevention. This could possibly be achieved by stratifying epidemiologic and clinical data sets on the basis of family history, psychobiologic markers, or antidepressant response, to learn if the secular trends are really influencing all subpopulations equally. For example, Giles et al. (44) found that the rate of unipolar depression was the same in siblings and in parents of unipolar depressed probands with reduced REM latency, whereas the rate was higher in siblings than in parents of depressed probands without reduced REM latency.

2. The differential psychobiology of bipolar and unipolar depression. This will perhaps best be achieved by following early-onset depressives studied psychobiologically to ascertain the predictors of the course of illness and of the onset of mania (109).

3. Predictors of onset and course of affective illness. This will probably best be understood by following populations at high risk for depression and studying them from family-genetic, family-environmental, and psychobiological points of view, in order to learn which factors predict the onset of depression and its subsequent course. In this regard, the identification of trait markers that antedate the development of affective illness will be a critical determinant of progress in this area. Such studies may also lead to empirically based programs to prevent the onset of affective illness in those at risk.

4. The genetics of early-onset affective illness. If informative, trait psychobiological markers for depression can be identified, then they can be utilized to render unaffected members of pedigrees who are not yet through the age of risk much more informative, thereby increasing statistical power for linkage studies. For example, pedigrees now include many unaffected members who are not through the age of risk, and therefore do not contribute to estimates of linkage. If a reliable marker for vulnerability to affective illness could be identified, this would enable a greater proportion of family members in a pedigree to become informative. Use of animal, particularly primate, models for depression with subsequent linkage studies may also shed light on the genetic etiology of affective illness.

5. The role of puberty in the increased risk of depression. With the onset of puberty, the risk of depression increases dramatically, particularly in females, yet the role of puberty in the etiology of depression is poorly understood. Basic neurobiological studies of the role of puberty in the development of neurotransmitter systems relevant to risk for affective disorder are critical, as are clinical psychobiological studies that follow normal, depressed, and high-risk children through puberty. Such studies are also likely to shed light on the gender difference in the rate of depression that occurs postpubertally.

6. Predictor of suicidal behavior and suicide within the early-onset affectively ill. Some psychobiological studies have indicated specific psychobiologic correlates of differential suicidality with depressed subjects (23,25,26,107,113), although it has been difficult to disentangle the impact of inpatient status, depressive severity, and suicidality. Further cross-sectional and longitudinal studies of early-onset affectively ill subjects that assess in psychosocial, family-genetic, and psychobiological domains are likely to be the most informative in this regard.

7. Integrating studies of individual variability in stress responsivity with studies of risk for depression. Animal models and follow-up studies of behaviorally inhibited infants suggest that individual differences in stress-responsivity may be related to subsequent risk for behavioral syndrome analogous to depression or anxiety. Such paradigms should also be applied to clinical populations, and particularly to children at risk for the development of depression, beginning with very young children.

8. Better prediction of treatment response. Up to this point, assessment and treatment studies of early-onset depression have not been well integrated. In order to address the issues of heterogeneity, it is vital to integrate family-genetic, family-environmental, and psychobiological assessments with treatment studies in order to learn what characteristics from these different domains predict treatment response to either psychosocial or psychopharmacologic agents.

Preparation of this chapter was made possible in part by funding from NIMH #MH41712, "Psychobiology of Depression in Children and Adolescents" (Program Project) (Dr. Ryan), NIMH #MH46510, "EEG Sleep Changes in Adolescent Depression" (Dr. Dahl), NIMH #MH44711, "Youth Exposed to Suicide" (Dr. Brent), and NIMH #MH30915, "Mental Health Clinical Research Center for the Study of Affective Disorders" (Dr. Kupfer).

published 2000