Symptom Outcome in Early-Onset Bipolar Disorder: Could Be Better, Could Be Worse

The study of bipolar disorder in children has had its ups and downs. The diagnosis went from almost nonexistent to being so common that the DSM-5 committee felt compelled to sharpen the criteria and add a new diagnosis, disruptive mood dysregulation disorder, to decrease what was thought to be overdiagnosis. In this issue, the latest report from the Course and Outcome of Bipolar Youth (COBY) study, by Birmaher et al. (1), also reverses the thinking that childhood-onset bipolar disorder has an invariably gloomy course. Briefly, children ages 7–17 with initially scrupulously diagnosed bipolar I or II disorder or bipolar disorder not otherwise specified were interviewed every 6 months or so about their weekly psychiatric status for the previous 6 months (i.e., prospectively collected retrospective data) for an average follow-up period of 8 years. Their moods and other clinical information were recorded and, for the purposes of this particular study, the presence or absence and the severity of mood symptoms on the mania and depression subscales of the Schedule for Affective Disorders and Schizophrenia for School-Age Children were ascertained.

Using latent class growth analysis, a statistical procedure that helps identify homogeneous populations within a larger heterogeneous group, the authors identified four groups of study participants based on their 8-year clinical course. At the extremes were those who were “predominantly euthymic” during most of the follow-up period (24%) and those where were “predominantly ill” (22.3%). The two in-between groups were “moderately euthymic” (34.6%) or were “ill with improving course” (19.1%). In fact, the “ill with improving course” group took about 3 years to change trajectories. Latent class growth analysis has been used to illuminate trajectories in a number of large samples (such as studies of children with aggression [2]), and in general they follow the lines of good, poor, and changing trajectory outcomes, with the proportion of these varying from sample to sample.

Besides telling us that at least for some young people, the outlook for early-onset bipolar disorder is not as grim as we’ve been led to believe, the predictors of good versus poor outcomes were not too surprising. Keeping with the “cup half full” theme, the “predominantly euthymic” group, compared with the more chronic group, was older at first onset of symptoms and first episode and was less likely to have experienced behaviors associated with depression, such as self-injurious and suicidal behavior. Their condition overall was less complicated, with lower rates of comorbid attention deficit hyperactivity disorder (ADHD) and anxiety and thus less concomitant use of stimulants and antidepressants. Rates of all mental disorders in their parents were lower (bipolar disorder, ADHD, and other disorders), socioeconomic status was higher, intact families were more common, and rates of prior physical and sexual abuse lower. Part of the reason this group was “predominantly euthymic” is that half the group had no additional mood episodes over the follow-up period (49.9%, compared with 8.5% for the “predominantly ill” group). We can’t tell if the index episode for the study was their first episode, but they appeared to have had fewer episodes prior to entering the study.

Using latent class growth analysis to understand the possible range of outcomes is clearly a step forward. There are a number of other questions we would like answered: What differences in functional outcomes accompany these symptom groups? How many children have only one or two episodes and remit entirely (as opposed to remaining subsyndromal)? What accounts for improvement at around 3 years after an episode in the “improving” group, and does treatment matter? Were index manic episodes in the “predominantly euthymic” group more in keeping with the shorter manic episode durations seen in adults? In the Collaborative Depression Study (CDS), on which the COBY study was patterned, median time to recovery from intake manic/mixed episode was much shorter (3) than in the COBY study (4) (7–16 weeks compared with 78 weeks) and rates of prior other psychiatric disorders much lower (e.g., for ADHD, 28% compared with at least 58% [5]). Were the parents’ bipolar disorders responsive to lithium in the “predominantly euthymic sample” and more complex and comorbid in the “predominantly ill” group (6)? Finally, both my clinical and research experience (7) has been that diagnoses change even with standardized research diagnostic procedures. How many children in the COBY study, now adolescents and young adults, would be independently diagnosed with something else?

The question of how these findings compare with what we know of adult bipolar disorder samples is a more difficult one to answer than it might appear. Although there are many 1- to 2-year follow-up studies of adults with bipolar disorder, there are relatively few that are really long-term (8). These adult studies have acquired samples at hospitalization and have higher rates of psychosis. What is needed is a latent class growth analysis of the CDS to determine whether percent time ill is lower in adults because more adults are predominantly euthymic than youths or because rates of chronic outcomes are lower (5).

Outcome is also measured differently from study to study. When our group examined outcomes with a gross metric of 1) recovered/remitted/good outcome, 2) incomplete remission/fair outcome, and 3) chronic or poor outcome (i.e., chronically mood disordered, psychotic, or dead by suicide), we found that a little less than half the study participants do quite well in that they experience few if any recurrences, or function well between episodes. About one-quarter remain chronically ill, usually depressed and functionally disabled, and the remainder are somewhere in-between (8). Overall, Birmaher and colleagues’ findings thus seem somewhat similar to what we observed in a study of young adult bipolar patients. The question is whether predictors of good and poor outcome are similar in the two samples.

How do the older participants in the COBY study compare with adolescents with mania in other studies? Again, sample differences make it difficult to draw conclusions. Those few studies with at least a 4-year follow-up were again often derived from inpatient samples with high rates of psychosis (28%−63%), relatively short manic episodes (8–16 weeks), and low rates of ADHD comorbidity (4%−15%). Overall, outcomes sound similar—about half had a good outcome, and 20% were chronically ill (8).

Do the chronically ill COBY children map onto young samples reported by Geller et al. (9) and Wozniak et al. (10)? In those samples, rates of comorbidity with ADHD were also high (72% in the COBY study, compared with 92% [9] and 85% [10]). There are important methodological differences in sample assessment (8), so it really isn’t yet possible to draw conclusions.

There is a need to reconcile the differences between studies in order to provide an informed picture of the future of the illness for families concerned with the early onset of a severe mental disorder. We need to know what to say when parents (or young patients) ask whether they will be “sick for the rest of their lives.” Perhaps as with cancer, we can give them a comparable “5-year survival rate,” hedging a bit depending on how complicated their clinical and psychosocial picture is. What we can’t yet tell them is how to change the trajectory of the illness into the “predominantly euthymic” one—and that, let’s face it, is what they really want to know.