Prospective naturalistic studies of children and adolescents with bipolar disorder (mainly subtype I) have shown that while rates of recovery from index episodes are high, in the range of 70%—100%, of those who recover, up to 80% will experience one or more syndromal recurrences over a period of 2 to 5 years (1—4). Moreover, prospective studies have shown that, similar to findings reported for adults (1, 3, 4), youths with bipolar disorder experience frequent mood fluctuations of varying intensities throughout 60%—80% of the follow-up time, particularly depressive and mixed symptoms.

Factors associated with worse longitudinal outcome include early age at illness onset, long illness duration, mixed episodes, rapid cycling, psychosis, subsyndromal symptoms, comorbid disorders, low socioeconomic status, exposure to negative life events, lack of psychotherapy treatment, poor adherence to pharmacological treatment, and family psychopathology (1—6).

A prior study from the Course and Outcome of Bipolar Youth (COBY) program that followed 263 youths with bipolar spectrum disorders for an average of 2 years showed that during most of the follow-up time, youths experienced subsyndromal and syndromal mood symptoms and frequent mood fluctuations (4). Twenty percent of the youths with bipolar II disorder converted to bipolar I disorder, and 25% of those with bipolar disorder not otherwise specified converted to bipolar I or II disorder. The aim of the present study was to extend COBY’s prior findings in a larger sample of youths with bipolar spectrum disorders followed for a longer time. For this purpose, the data were evaluated with two complementary sets of analyses. In the first, to compare our results with those of the existing literature, we conducted analyses using survival analytic techniques and the standard definitions of syndromal recovery and recurrence. In the second, since pediatric bipolar disorder is not manifested only by discrete syndromal recurrences but also by numerous subsyndromal episodes and mood changes, we conducted analyses of the week-by-week mood symptoms over the follow-up period. Subsequent reports will investigate how outcomes are affected by other factors, such as subsyndromal recoveries, suicidal behaviors, health services utilization, psychosocial functioning, exposure to negative life events, specific comorbid disorders, and treatment.

Participants

The methods for COBY have been described in detail elsewhere (4, 7). Briefly, the study included youths ages 7 to 17 years 11 months with DSM-IV bipolar I or II disorder or operationally defined bipolar disorder not otherwise specified. Youths with COBY-defined bipolar disorder not otherwise specified were previously shown to convert to bipolar I or II and to have a comparable but less severe clinical picture, a similar family history, similar rates of comorbid disorders, and a similar longitudinal outcome compared with youths with bipolar I (4, 7).

Youths with schizophrenia, mental retardation, autism, and mood disorders secondary to substances, medications, or medical conditions were excluded from the study.

Participants were recruited from outpatient clinics (67.6%), inpatient units (14.3%), advertisements (13.3%), and referrals from other physicians (4.8%). They were enrolled independent of current mood state or treatment status.

The analyses in this study are based on the prospective evaluation of 413 youths, including 244 (59.1%) with bipolar I disorder, 28 (6.8%) with bipolar II disorder, and 141 (34.1%) with bipolar disorder not otherwise specified who had at least one follow-up assessment. Participants had been prospectively interviewed every 37.5 weeks (SD=20.8) for an average of 191.5 weeks (SD=75.7). Youths with bipolar II were followed significantly longer (227.4 weeks [SD=76.6]) than those with the other two bipolar subtypes (bipolar I: 183.2 weeks [SD=71.8]; bipolar disorder not otherwise specified: 198.7 weeks [SD=79.8]; F=5.4, p=0.005).

At intake, participants with bipolar disorder not otherwise specified were the youngest, followed by those with bipolar I and then those with bipolar II (Table 1). More youths with bipolar disorder not otherwise specified were in Tanner stage I of sexual development than those with bipolar II, and more youths with bipolar II were in Tanner stages IV or V than those with bipolar disorder not otherwise specified. There were no between-group differences in Tanner stages II or III. The mean age at onset of mood symptoms was 8.4 years, and the mean age at onset of DSM-IV mood episodes was 9.3 years (see below for the definition of age at onset). The onset of mood symptoms and episodes was significantly later in youths with bipolar II than in youths with the other two bipolar subtypes. As expected, by definition, the polarity of the index episode reflected the bipolar subtype, with mania or hypomania being more common in youths with bipolar I than in those with bipolar II or bipolar disorder not otherwise specified. However, youths with bipolar II had significantly more depressive index episodes than youths with the other two subtypes. Youths with bipolar I had more lifetime psychosis than those with bipolar disorder not otherwise specified (for all above comparisons, p values were <0.05 and Cohen’s d ranged from 0.3 to 0.9). There were no other significant between-group differences.

The retention rate was 86%, with 93% of participants completing at least one follow-up interview. Except for lower rates of anxiety disorders in youths who dropped out of the study (54.5% compared with 38.7%; p=0.02), there were no other demographic or clinical differences between those who continued in the study and those who withdrew.

Procedures

Each participating university’s institutional review board approved the study, and consent was obtained from the participating youths and their parents. At intake, youths and parents were directly interviewed for the presence of current and lifetime psychiatric disorders in the youths. The instruments used were the Schedule for Affective Disorders and Schizophrenia for School-Age Children—Present and Lifetime Version (K-SADS-PL) (8), the Kiddie Mania Rating Scale (K-MRS) (9), and the depression section of the K-SADS-PL (10).

Longitudinal changes in psychiatric symptoms since the previous evaluation were assessed using the Longitudinal Interval Follow-Up Evaluation (11) and tracked on a week-by-week basis using this instrument’s Psychiatric Status Rating Scales (12). These scales use numeric values that have been operationally linked to the DSM-IV criteria; DSM-IV criteria information is gathered in the interview and then translated into ratings for each week of the follow-up period. The ratings indicate the severity level of an episode as well as whether the patient has recovered or had a recurrence. For mood disorders, scores on the Psychiatric Status Rating Scales range from 1 for no symptoms to 2—4 for varying levels of subthreshold symptoms and impairment to 5—6 for meeting full criteria with different degrees of severity or impairment. The consensus scores obtained after interviewing parents and their children were used for the analyses.

Family history of mood disorders was ascertained using the Family History Screen (13). The Petersen Pubertal Development Scale (14) and the equivalent Tanner stages were used to evaluate and categorize pubertal stages. Socioeconomic status was ascertained using the four-factor Hollingshead scale (15).

All assessments were conducted by research staff trained to reliably administer the interviews; interview results were presented to child psychiatrists or psychologists, who confirmed the diagnoses and the Psychiatric Status Rating Scales scores. The overall K-SADS-PL kappa coefficients for psychiatric disorders were ≥0.8. The intraclass correlation coefficients for the K-MRS and the K-SADS-P depression section were ≥0.95. The intraclass correlation coefficients for syndromal and subsyndromal mood disorders ascertained through the Psychiatric Status Rating Scales (using methods described elsewhere [12]) were ≥0.75; the intraclass coefficient correlations and the Kendall’s coefficients of concordance were between 0.74 and 0.79 for a major depressive episode and between 0.60 and 0.67 for mania/hypomania.

During the follow-up, the K-MRS and the depression section of the K-SADS-P were used as an additional assessment of mood symptom severity for the week when symptoms were most severe during the month prior to interview. A comparison of the maximum scores for depression and mania on the Psychiatric Status Rating Scales for the 4 weeks prior to each follow-up assessment and the maximum scores on the K-MRS and the depression section of the K-SADS-P for the same period showed Spearman correlations of 0.82 (p<0.0001) and 0.77 (p<0.0001), respectively.

Definitions of Clinical Course

Age at onset of bipolar disorder was defined as the age at onset of a DSM-IV mood episode or an episode fulfilling the COBY’s modified DSM-IV criteria for bipolar disorder not otherwise specified. The minimum age at onset was arbitrarily set at age 4. The duration of bipolar disorder was calculated from the age at onset. The index episode was defined as the current or most recent DSM-IV mood episode.

The percentage of follow-up weeks spent asymptomatic or symptomatic in the different mood symptom categories during the entire follow-up period was computed for each participant, based on Psychiatric Status Rating Scales scores. Full recovery was defined as 8 consecutive weeks with a score ≤2 (minimal or no mood symptoms) (4, 16). Time to recovery from the index episode was measured from the onset of the index episode. Participants were considered to have a recurrence (new episode) if they had a Psychiatric Status Rating Scales score ≥5 with a duration of 1 week for mania/hypomania or 2 weeks for depression. Similar to previous studies (16), "change in mood polarity" was defined as a switch between depression (rating ≥3) and mania/hypomania (rating ≥3) or vice versa with or without any intervening weeks with no symptoms or when ratings for 1 week included both mania/hypomania and depression scores ≥3. This definition of change in mood polarity is not the equivalent of DSM rapid cycling. For rapid cycling as well as for mixed episodes, COBY used the DSM-IV definitions.

Statistical Analyses

The syndromal recoveries and recurrences manifested in bipolar disorder were evaluated using survival analyses and Cox proportional hazards regressions (17). The numerous ongoing mood changes and periods of subsyndromal symptoms of bipolar disorder were evaluated using within-group and between-group analyses of the week-by-week syndromal and subsyndromal symptoms, stratifying by bipolar subtype. Differences within and between groups were analyzed using standard parametric and nonparametric univariate tests.

About 14% of the sample had their most recent mood episode offset and recovered before intake. Since the analyses with and without these participants yielded similar results, all participants were included in the analyses.

The data were censored after participants with bipolar II and bipolar disorder not otherwise specified converted to other bipolar subtypes.

For all of these analyses, the effects of demographic variables, age, bipolar subtype, psychosis, age at bipolar onset, duration of bipolar disorder, presence of any comorbid disorder, and family history of mood disorders, as well as interactions between these variables, were evaluated. Variables were examined univariately, and those that were significantly associated with the outcome of interest were analyzed using multiple linear regressions. Analyzing the data using generalized linear models yielded similar results. To assess the effects of age on outcome, the sample was divided into three groups: children <12 years old, adolescents ≥12 years old with onset of their episodes prior to age 12, and adolescents with onset of their episodes at or after age 12. Analyses including age or pubertal status yielded similar results. Thus, only the results including age are presented. Also, since bipolar subtype and polarity of the index bipolar episode were highly associated (φ=0.75, p≤0.0001) (also see Table 1), only the bipolar subtype was included in the analyses.

All p values are based on two-tailed tests with a set at 0.05.

Survival Analyses

Recovery from the index episode

Overall, 81.4% of the participants had full recovery, a median of 123.7 weeks after the onset of the index episode (Table 2). Youths with bipolar I showed significantly higher rates of recovery compared to those with bipolar disorder not otherwise specified, and those with bipolar I and II had shorter times to recovery compared to those with bipolar disorder not otherwise specified (Figure 1). In addition to the standard 8-week duration criterion for recovery, in analyses using 2, 4, and 6 weeks with minimal or no mood symptoms, rates of recovery were 88%, 84%, and 81%, respectively.

A lower likelihood of full recovery was associated with a diagnosis of bipolar disorder not otherwise specified (versus bipolar I and II, hazard ratio=0.62, 95% CI=0.49—0.97); children with childhood onset (versus adolescents with adolescent onset, hazard ratio=0.50, 95% CI=0.37—0.67; versus adolescents with childhood onset, hazard ratio=0.70, 95% CI=0.50—0.99); non-Caucasian race (hazard ratio=0.60, 95% CI=0.42—0.84); longer duration of illness (hazard ratio=0.83, 95% CI=0.78—0.88); and a family history of mania/hypomania in first- or second-degree relatives (hazard ratio=0.79, 95% CI=0.63—0.99). Conversely, the interaction of living with both biological parents and having higher socioeconomic status was associated with a greater likelihood of recovery (hazard ratio=1.23, 95% CI=1.01—1.51). No other significant predictors or interactions were found.

Recurrence after recovery from index episode

Of the youths who recovered, 62.5% had a syndromal recurrence a median of 71 weeks after recovering from their index episode (Table 2). Higher rates of, and shorter times to, recurrence occurred among youths with bipolar I and II as compared to those with bipolar disorder not otherwise specified (Figure 2).

Participants who recovered had a mean of 1.1 syndromal recurrences (SD=1.2) during the 4-year follow-up; 33% (110/336) had one recurrence, 20% (66/336) had two recurrences, and 10% (34/336) had three or more recurrences during the follow-up period. Across all bipolar subtypes, most syndromal recurrences after the index episode were major depressive episodes (59.5%), followed by hypomanic (20.9%), manic (14.8%), and mixed (4.8%) episodes. Interestingly, for youths whose index episode was a major depression, a mixed episode, hypomania, or not otherwise specified, 64% (109/171) of the first recurrences were major depressions, followed by mania/hypomania (30%; 52/171) and then mixed episodes (6%; 10/171). For youths whose index episode was mania, the majority of first recurrences were mania/hypomania (59%; 23/39), followed by depression (41%; 16/39). A subanalysis including only participants with bipolar I in an age band similar to that of a prior bipolar I longitudinal study (3) yielded similar results.

An increased likelihood of recurrence was associated with bipolar I and bipolar II (versus bipolar disorder not otherwise specified, hazard ratio=1.37, 95% CI=1.01—1.88); lower socioeconomic status (hazard ratio=0.87, 95% CI=0.77—0.97); and a family history of mania/hypomania in first- or second-degree relatives (hazard ratio=1.38, 95% CI=1.04—1.84). No other significant predictors or interactions were found.

Week-by-Week Mood Analyses

Analyses for the entire sample

Overall, participants spent approximately 40% of the time during the follow-up period asymptomatic and 60% symptomatic (41.8% subsyndromal and 16.6% with syndromal symptomatology) (Table 3). Participants spent significantly more time in syndromal depression or mixed/cycling than in syndromal mania/hypomania. There were no significant differences in time spent in each subsyndromal polarity.

Within-group analyses

Analyses within each bipolar subtype showed that youths with bipolar I and bipolar disorder not otherwise specified spent significantly more follow-up time with subsyndromal than with syndromal symptoms. For youths with bipolar II, there was no difference in the proportion of follow-up time spent with syndromal and subsyndromal symptoms. While experiencing syndromal symptoms, all three bipolar subtypes spent more time with depression and mixed/cycling symptoms than with mania/hypomania. While experiencing subsyndromal symptoms, youths with bipolar I spent more time with subsyndromal mania and mixed symptoms than with subsyndromal depression, and youths with bipolar disorder not otherwise specified spent more time with subsyndromal mixed symptoms than with the other two subsyndromal polarities. For youths with bipolar II, there were no significant differences in the amount of follow-up time spent in each subsyndromal polarity. For all comparisons, p values were ≤0.05 and values for Cohen’s d_s ranged from 0.45 to 1.3.

Between-group analyses

Comparisons between bipolar subtypes showed that youths with bipolar I spent more follow-up time asymptomatic than did those with bipolar disorder not otherwise specified. Youths with bipolar I and II spent similar amounts of time with syndromal symptoms but more time than those with bipolar disorder not otherwise specified. In contrast, youths with bipolar disorder not otherwise specified spent significantly more time with subsyndromal symptoms than did those with the other two bipolar subtypes. Within syndromal periods, youths with bipolar I and II spent more time with hypomania than did those with bipolar disorder not otherwise specified, youths with bipolar II spent more time in major depression episodes than did those with the other two bipolar subtypes, and youths with bipolar I spent more time with mixed/cycling symptoms than did those with bipolar disorder not otherwise specified. By definition, youths with bipolar II and bipolar disorder not otherwise specified did not spend any weeks with syndromal mania or mixed symptoms. Within subsyndromal periods, participants with bipolar I spent more weeks with subsyndromal manic symptoms compared to those with bipolar II. For all comparisons, p values were ≤0.05 and values for Cohen’s d_s ranged from 0.24 to 0.63.

Multiple linear regression models

The following variables were associated with more follow-up weeks with any mood symptoms: low socioeconomic status (t=4.28, p<0.001), children with childhood onset (versus adolescents with adolescent onset, t=5.07, p<0.001), adolescents with childhood onset (versus adolescents with adolescent onset, t=4.05, p<0.001), and presence of any comorbid disorder (t=2.62, p=0.009). No other significant predictors or interactions were found. Except for bipolar I and II predicting more follow-up time with syndromal symptomatology, and bipolar disorder not otherwise specified and being non-Caucasian predicting more time with subsyndromal symptomatology, separate linear regressions for syndromal and subsyndromal symptomatology yielded similar results.

Patients with chronic symptoms

In addition to analyzing the percentage of time participants spent symptomatic (syndromal plus subsyndromal symptoms), we calculated chronicity, as measured by the percentage of participants who had syndromal and/or subsyndromal mood symptoms ≥75% of the follow-up time. Approximately 38% of the participants, particularly those with bipolar I (bipolar I > bipolar disorder not otherwise specified, χ²=5.66, p=0.02), experienced chronic mood symptoms, with only 3% of participants meeting full syndromal criteria for mood episodes. Most of these chronic symptoms were mixed (46%), followed by depression (33.8%) and mania/hypomania (21%).

Psychosis

Clinically relevant psychotic symptoms were defined as having a score of 3 (definitely present) for delusions and/or hallucinations on the Psychiatric Status Rating Scales. Overall, 16% of the participants experienced psychotic symptoms during the follow-up period. For these youths, psychotic symptoms were manifested during 17.1% of the follow-up time. Participants with bipolar I spent significantly more time with psychotic symptoms than did those with bipolar disorder not otherwise specified.

Change in mood polarity

Shifts in mood polarity occurred a mean of 36.2 times (SD=46.9) during the entire follow-up period, or 12.1 times per year (SD=15.0) (Table 3). A change in mood polarity once per year or less was observed in 31.2% of the sample, five or more times per year in 51.1%, 10 times or more per year in 38.7%, and more than 20 times per year in 23.7%. Except for youths with bipolar disorder not otherwise specified being more likely to have at least 10 changes in mood polarity per year compared to those with either bipolar I or II (p values ≤0.03, values for Cohen’s d_s ranging from 0.23 to 0.45), there were no other between-group differences. In a multiple linear regression, lower socioeconomic status (t=3.90, p=0.001), children with childhood onset (versus adolescents with adolescent onset, t=4.02, p=0.001), adolescents with childhood onset (versus adolescents with adolescent-onset, t=3.73, p=0.001), and presence of any comorbid disorder (t=2.04, p=0.04) were significant predictors of having a greater number of changes in mood polarity per year. There were no other significant predictors or interactions.

Conversion From Bipolar II to Bipolar I and From Bipolar Disorder Not Otherwise Specified to Bipolar I or II

Of the 169 youths with bipolar II and bipolar disorder not otherwise specified, 61 (36.1%) converted to a different bipolar subtype over the follow-up period. Of these, 25% (7/28) with bipolar II converted to bipolar I, 19.9% (28/141) with bipolar disorder not otherwise specified converted to bipolar I, and 18.4% (26/141) with bipolar disorder not otherwise specified converted to bipolar II (38.3% of youths with bipolar disorder not otherwise specified converted overall).

Corroborating prior COBY findings (4), this study showed that bipolar spectrum disorders in youths are episodic disorders characterized most often by subsyndromal episodes and less frequently by syndromal episodes, with mainly depressive and mixed symptoms and rapid mood changes.

Survival analyses using the standard definitions of syndromal recovery and recurrence (based on DSM-IV and the literature) indicated that approximately 80% of youths with bipolar spectrum disorders achieved full recovery about 2.5 years after onset of their index episode. However, 1.5 years after full recovery, approximately 60% of the participants had at least one syndromal recurrence. Compared to youths with bipolar disorder not otherwise specified, those with bipolar I and II were more likely to recover but also to have a less durable recovery.

During the entire follow-up period, one-third of the participants had at least one syndromal recurrence and 30% experienced more than two syndromal recurrences. Most of these syndromal recurrences were major depressions, followed by hypomanic, manic, and mixed episodes. In general, the polarity of the index episode predicted the polarity of subsequent episodes.

In addition to the analyses focusing only on recovery and recurrence of syndromal symptoms, week-by-week analyses provided a more in-depth clinical picture of the course of bipolar disorder, showing that distinct periods of full syndromal mood episodes exist in youths with bipolar spectrum disorders. However, these episodes are embedded in more prevailing and longer periods of subsyndromal mood symptomatology. Youths with bipolar spectrum disorders were symptomatic during 60% of the follow-up period, during which they spent about 2.5 times more time with subsyndromal than with syndromal symptomatology. Mixed/cycling and depressive symptoms accounted for the greatest proportion of time ill. In contrast, purely manic symptomatology, especially at the full syndromal level, was less common. Rapid mood changes were ubiquitous, and psychotic symptoms were relatively common, particularly in youths with bipolar I. Chronic symptoms, defined as having any type of symptoms during 75% or more of the follow-up period, were present in 38% of the participants. Almost all of these chronic symptoms were subsyndromal and of the depressive type.

The week-by-week analyses also shed light on the similarities and differences in the longitudinal patterns of symptom phenomenology of youths with bipolar I and II and bipolar disorder not otherwise specified. Bipolar I was manifested by more time with subsyndromal than with syndromal symptoms. Most of the syndromal time was characterized by mixed/cycling or depressive symptoms, and most of the subsyndromal time was with subsyndromal manic or mixed symptoms. Youths with bipolar II spent equal amounts of time in syndromal and subsyndromal states. The syndromal episodes were most commonly depression or mixed states, but there were no significant differences in the proportion of time spent with any type of subsyndromal symptoms. Finally, bipolar disorder not otherwise specified was mainly manifested by periods of subsyndromal mixed symptoms, closely followed by periods of subsyndromal manic or depressive symptoms.

Between-group comparisons provided preliminary validation for the subtyping of bipolar disorder in youths. In general, in comparison with other subtypes, each bipolar subtype continued to show some category-specific symptomatology. For example, during follow-up, youths whose initial diagnosis was bipolar I showed more syndromal, mixed/rapid cycling, and manic/hypomanic symptoms than did those with bipolar disorder not otherwise specified; youths with bipolar II spent more time in hypomania than did those with bipolar disorder not otherwise specified and more time in depression than did those with bipolar I and bipolar disorder not otherwise specified; and youths with bipolar disorder not otherwise specified spent significantly more time with subsyndromal symptoms than did those with bipolar I and II. However, there was some symptom overlap among the different bipolar subtypes, especially bipolar I and II. Moreover, 25% of youths with bipolar II converted to bipolar I, and 38% of those with bipolar disorder not otherwise specified converted to bipolar I and II.

Although there were some differences in the demographic and clinical factors associated with the outcome variables measured (recovery, time symptomatic, and changes in polarity), in general, early onset of bipolar disorder, presence of comorbid disorders, family history of mood disorders (particularly mania/hypomania), low socioeconomic status, and non-Caucasian race were associated with worse outcome. Long duration of illness was also associated with a lower likelihood of recovery, with each year of illness decreasing the likelihood of recovery by 20%.

Before continuing the discussion of COBY’s findings, it is important to note the limitations of this study. First, despite efforts to obtain precise information, the data collected through the Longitudinal Interval Follow-Up Evaluation is subject to retrospective recall bias. Although it appears that this instrument has adequate psychometric properties (1, 4, 12), further studies using the methods described by Warshaw et al. (12) and including blind interviewers are warranted. Second, although COBY used the standard definitions of course for recovery and recurrence (1, 3, 18), the rates and duration of the mood episodes may change according to the duration criteria and symptom threshold severity chosen. Third, the results pertaining to youths with bipolar II should be considered tentative given the relatively small size of this group. However, across all bipolar subtypes, after depression most syndromal recurrences were hypomanias. Finally, as most participants were Caucasian and were recruited primarily from outpatient and, to a lesser extent, inpatient settings, the generalizability of the observations to other populations remains uncertain. Nevertheless, nonreferred adolescents with bipolar disorder have been shown to have a similar course and high morbidity (5).

Despite methodological differences, all existing studies of the course of bipolar disorder in youths, regardless of country and source of ascertainment, show that the likelihood of recovery from the index episode is high (1—4). However, as with adult populations, despite the high recovery rate, the rates of recurrence, persistence of subsyndromal clinical morbidity, and rapid and frequent changes in mood polarity are also high, and most syndromal and subsyndromal recurrences are depressions (16, 19—23). It does appear that the polarity of the index episode predicts the polarity of the subsequent episodes (21, 24—28), which suggests the possibility of using specific psychosocial and pharmacological treatments based on the polarity of the index episode.

The results of this and other emerging pediatric studies suggest strong general similarities in the longitudinal course of bipolar disorder in youths and adults, which is mainly manifested by subsyndromal symptomatology and rapid mood changes (1—4, 29). However, there is evidence that very early onset confers greater liability for a more chronic and fluctuating course, mixed/cycling episodes, high rates of comorbid disorders, and increased rates of mood disorders in families (1, 3, 30—32). Converging with these accounts are reports indicating that adults whose onset of bipolar disorder dates to childhood have a more severe and chronic course, lower quality of life, and more episodes, changes in mood polarity, suicidality, and comorbidity (33—35).

Comparable with other findings in the literature (1—4, 6), we found that childhood-onset bipolar disorder, comorbid disorders, positive family history for mood disorders, and low socioeconomic status were associated with poorer outcome. Moreover, the probability of recovery was inversely related to duration of illness, which further underscores the importance of early detection of illness and rapid implementation of stabilizing treatments. Such efforts may be of even greater urgency for youths with bipolar disorder who have risk factors associated with poorer outcome.

Similar to findings in the literature on major depression (36), it appears that there are some differences in the factors associated with the various indices of clinical outcome (recovery, recurrence, and amount of time symptomatic). Moreover, as the polarity of the index episode was shown to convey different prognostic characteristics, it may be that these observations will be informative to clinical practice. For example, since youths with depression were seen to have more depressive recurrences, they may require more aggressive and specific therapies to reduce the risk of future depressive episodes.

To our knowledge, COBY is the first naturalistic, prospective study of this affective subtype in youths—an important avenue of research given the modal age of onset of bipolar II illness during adolescence (18, 27, 37). Consistent with adult data on high levels of morbidity associated with this subtype, youths with bipolar II had a greater overall risk of recurrence and more depressive morbidity compared to youths with bipolar I, and had rates of nonaffective comorbidity, suicidality, nonsuicidal self-injurious behaviors, functional impairment, and family history of bipolar and other mood disorders comparable to those with bipolar I (6, 7, 32, 38—40). Also, bipolar II in youths appears to be a far less stable phenotype than in adults; in this cohort, 25% of the bipolar II youths converted to bipolar I, a rate higher than that reported in adult studies (41). Since this disorder is mainly characterized by episodes of syndromal and subsyndromal depression across the lifespan, it is well appreciated that periods of hypomania can be misconstrued as normal fluctuations in mood or as erratic behavior, and perhaps more so in adolescents (27, 42), resulting in a high risk of misclassification as recurrent unipolar depression or other, nonaffective disorders.

Bipolar disorder not otherwise specified is characterized by rates of comorbidity, suicidality, functional impairment (except hospitalizations), and family history of mood disorders equivalent to those in youths with bipolar I and II (6, 7, 32, 38—40). These findings, together with the high rates of conversion to bipolar I or II, provide preliminary validation of its nosological affinity with bipolar I and II disorder. It should be emphasized, however, that our classification relied on the presence of an affective phenotype that differed from bipolar I or II due to failure to meet the DSM-IV duration requirements for these subtypes. Thus, our findings are in accord with studies in the adult literature (18, 42, 43) emphasizing the existence of clinically relevant episodes of mania/hypomania that last for less time than required by DSM criteria. These episodes are often overlooked because of the predominance of syndromal depression and subsyndromal manic/mixed symptoms of short duration in its expression. Results from COBY suggest that bipolar disorder not otherwise specified is an episodic illness, albeit one that often comprises subsyndromal episodes that should be considered distinct from the symptomatology of youths with behavior disorders and "severe mood dysregulation" (44).

In summary, although distinct episodes of full syndromal mood symptomatology as well as durable periods of euthymia can be identified in youths with bipolar disorder, the course of bipolar spectrum disorders in children and adolescents is predominantly characterized by subsyndromal and, much less frequently, syndromal episodes. Rapid mood changes are evident during these episodes, which are mainly of depressive and mixed polarity. During follow-up each bipolar subtype showed some distinct clinical characteristics and course, but there was overlap in their symptoms and substantial conversion from bipolar II and bipolar disorder not otherwise specified into other bipolar subtypes. The course of bipolar disorder, the relative infrequency of syndromal DSM manic episodes, the effects of development in symptom manifestation, and the high prevalence of comorbid disorders may account, at least in part, for the difficulties in recognizing and managing this illness in youths. The recurrence, chronicity, and psychosocial morbidity associated with this illness in critical developmental stages call for its prompt recognition and the development of more efficacious treatments, particularly since each year of illness appears to decrease the likelihood of recovery.