Prevalence and correlates of childhood-onset bipolar disorder among adolescents

Aim: Early-onset bipolar disorder (BD) is associated with a more severe illness as well as a number of clinical factors among adults. Early-onset can be categorized as childhood- (age < 13) or adolescent- (age ≥ 13) onset, with the two displaying different clinical profiles. We set out to examine differences in clinical, and familial characteristics among adolescents with childhood- versus adolescent-onset BD. Methods: The study included 195 adolescents with BD, ages 14 – 18 years. Age of onset was determined retrospectively by self-report. Participants completed the semi-structured K-SADS-PL diagnostic interviews along with self-reported dimensional scales. Analyses examined between-group differences for clinical and familial variables. Variables associated with age of onset at p < 0.1 in univariate analyses were evaluated in a logistic regression model. Results: Approximately one-fifth of participants had childhood-onset BD ( n = 35; 17.9%). A number of clinical and familial factors were significantly associated with childhood-onset BD. However, there were no significant differences in depressive and manic symptom severity. In multivariate analyses, the variables most strongly associated with childhood-onset were police contact, and family history of suicidal ideation. Smoking and psychiatric hospitalization were associated with adolescent-onset. Conclusions: In this large clinical sample of adolescents with BD, one-fifth reported childhood-onset BD. Correlates of childhood-onset generally aligned with those observed in the literature. Future research is warranted to better understand the genetic and environmental implications of high familial loading of psychopathology associated with childhood-onset, and to integrate age-related treatment and prevention strategies

Among adults with BD, childhood-onset is associated with irritability; comorbid attention-deficit/hyperactivity disorder (ADHD) and conduct disorder (CD); rapid cycling; a family history of mood disorders, and a family history of substance abuse (Baldessarini et al., 2012;Holtzman et al., 2015;Lázaro et al., 2007). Comparatively, adolescent-onset BD has been associated with greater prevalence of psychosis (Lázaro et al., 2007) and substance use disorder (SUD) (Perlis et al., 2004).
While well investigated within adult populations, fewer studies have investigated the clinical features of childhood-onset compared to adolescent-onset BD in youth. It is noteworthy that, compared to adults with illness onset often decades prior, youth with recent illness onset, could be less prone to recall bias. Some studies have found childhood-onset BD to be more common than adolescent-onset (58%-75%) (Topor et al., 2013;Wilens et al., 1999). When comparing childhood-onset BD to adolescent-onset in youth, childhood-onset BD is associated with increased rates of ADHD or oppositional defiant disorder (ODD), as well as a higher presence of any comorbid disorder, a more insidious presentation of the illness and different manic symptoms (Faraone et al., 1997;Masi et al., 2006;Song et al., 2010;Topor et al., 2013). Childhood-onset BD is also associated with family history of depression, anxiety, ADHD, CD, and suicidal behaviours, compared to adolescent-onset (Rende et al., 2007). The only characteristic thus far found to be more common in adolescent-onset vs. childhood-onset BD is comorbid SUD (Wilens et al., 1999). Other studies have reported no differences between groups (Findling et al., 2001;Geller et al., 2000;Kennedy et al., 2015).
Despite extant studies, there remain a number of gaps in the literature on this subject. Many studies have focused on BD-I subtype, with analysis of clinical characteristics restricted to mania symptoms, while several studies have relied on chart reviews, rather than direct interviews. Lastly, while there have been large studies on this topic in adults, most studies in adolescents have had modest sample size, which limits the comparative evaluation of clinical correlates. One exception to these limitations is the large Course and Outcome of Bipolar Youth (COBY) study (Birmaher et al., 2009a; which found that: children experienced more irritability during depressive intervals compared to adolescents with either childhood-or adolescent-onset; both children and adolescents with childhood-onset BD experienced more mood lability and comorbid ADHD than adolescents with adolescent-onset BD; and adolescentonset BD was associated with higher rates of comorbid CD, panic disorder and SUD as compared to the other groups .
Overall, there remains a paucity of research comparing childhoodonset to adolescent-onset BD in adolescents, with no studies on this topic, to our knowledge, since the aforementioned COBY paper in 2006. We sought to replicate and expand previous studies by comparing the prevalence as well as the clinical and familial characteristics of childhood-vs. adolescent-onset BD in a large Canadian clinical sample. We hypothesized that adolescent-onset would be more common than childhood-onset BD, and that correlates of childhood-onset would align with those observed in the similarly designed COBY study.

| Subjects
Participants were 195 adolescents with BD-I, -II, or Not Otherwise Specified (NOS; akin to Other Specified Bipolar and Related Disorder), aged 14-18, who were recruited from a subspecialty outpatient clinic at a tertiary academic health sciences centre in Toronto, Canada. The study was approved by the institutional research ethics board, and written informed consent was provided by participants and at least one parent/guardian before study commencement. To evaluate the difference in clinical features of childhood-onset and adolescentonset BD, the sample was divided into two subgroups according to the age of BD onset. Age of BD onset was defined as the age at which the individual first met criteria for BD-I, BD-II, or BD-NOS.
Childhood-onset (n = 35) was defined as onset age of <13 years, while adolescent-onset (n = 160) was defined as onset age of 13 years or older. These parameters replicated the criteria used in a number of studies in the field (Birmaher et al., 2009a;Leverich et al., 2007;Post et al., 2010;Post et al., 2017).

| Procedure
All assessments were carried out by research staff who had a Bache-  (Chambers et al., 1985) and K-SADS Mania Rating Scale (MRS) (Axelson et al., 2003). Criteria for BD-NOS were defined according to criteria used in the COBY study (Kennedy et al., 2015). Diagnoses were determined using all available information, including adolescent and parent interviews and any available medical records. Clinical judgement was applied when conflicting information arose. Diagnoses were confirmed by a consensus meeting with a licensed child and adolescent psychiatrist following completion of the K-SADS-PL interview (B.I.G. K-SADS-DRS and K-SADS-MRS were also used to score most severe lifetime individual symptom severity. This measured the severity of each core symptom during the most severe period of experiencing depressive or manic symptoms. The four factor Hollingshead Scale (Hollingshead, 1975) was used to determine socioeconomic status (SES). The Children's Global Assessment Scale (CGAS) was used to score the adolescent's global functioning over the current period, most severe past, and highest level in the past year (Shaffer et al., 1983). Information on comorbid diagnoses and clinical characteristics (e.g., psychosis, psychotropic, and psychosocial treatment history) was obtained from the K-SADS-PL. Anxiety disorders were combined into one variable, "Any Anxiety Disorder(s)," which includes generalized anxiety disorder, social phobia, separation anxiety disorder, agoraphobia, and panic disorder. SUD included alcohol or drug abuse or dependence. Lifetime cigarette smoking, was also ascertained via the K-SADS-PL, and computed as a "yes" or "no" variable. A Safety Form, which was prepared for the particular study, was augmented from the KSADS-PL and used to record any lifetime police contact, lifetime involvement with child protective services, lifetime physical or sexual abuse, as well as any suicidality or non-suicidal self-injury that was not captured during the K-SADS-DRS interview (e.g., occurring outside the context of a depressive interval, such as while intoxicated or situationally dysregulated). The Children's Affective Lability Scale (CALS) was used to assess affect regulation using an adolescent self-report and parent/guardian-report (Gerson et al., 1996). The Life Problems Inventory self-report assessed dimensional traits of identity confusion, impulsivity, emotion dysregulation, and interpersonal problems (Rathus & Miller, 1995). This measure has been considered a proxy measure of borderline personality traits (Courtney-Seidler et al., 2013;Fonseka et al., 2015). Psychiatric history of first-and second-degree relatives was obtained from the adolescent and parent(s) through the Family History Screen (Weissman et al., 2000).

| Data analysis
Clinical and demographic variables were compared across the childhood-and adolescent-onset BD groups using t-tests or chisquared tests. As BD subtype was an omnibus variable, it was dummy coded according to the three subtypes prior to analysis. False discovery rate (FDR) was used to correct for multiple comparisons (Benjamini & Hochberg, 1995). Variables associated with age of onset at p < 0.1 after FDR correction were entered together in a backwards elimination logistic regression analyses. Age, race and sex were included as covariates. Square root transformation was carried out for any skewed variables prior to inclusion in multivariate analysis. All statistical analyses were conducted with Statistical Package for the Social Sciences (SPSS), version 26.0.

| Socio-demographic and clinical characteristics
Socio-demographic characteristics of all participants can be seen in Table 1. Of the 195 participants enrolled, 17.9% had childhood-onset (n = 35) and 82.1% had adolescent-onset (n = 160). Participants with childhood-onset were significantly younger at the time of the study than those with adolescent-onset. SES, sex, race, and living with both natural parents were not significantly different between the two groups.
In terms of clinical characteristics, as per Table 2, participants with childhood-onset were significantly more likely to have BD-NOS, and less likely to have BD-I. Participants with childhood-onset had significantly higher lifetime prevalence of ADHD, ODD, police contact, and stimulant treatment, and higher parent-reported affective lability (CALS). Childhood-onset was also significantly associated with a lower lifetime prevalence of smoking. Lower lifetime prevalence of psychosis, second-generation antipsychotics use, and psychiatric hospitalization were no longer significant after correction for multiple comparisons. In terms of family psychiatric history, per Table 3, childhood-onset participants had significantly more family history of depression, suicidal ideation, ADHD, SUD, alcohol use disorder (AUD), and CD.

| Most severe lifetime individual symptom severity
There were no significant between group differences in severity ratings during most severe lifetime depression or manic episodes. There was also no difference for individual symptom severity during these episodes.  Prior studies have found a higher prevalence of adolescent-onset compared to childhood-onset among adults with BD (Holtzman et al., 2015;Leverich et al., 2007;Perlis et al., 2004;Post et al., 2010).

| DISCUSSION
In contrast, prior studies in adolescents have found that childhoodonset is at least as common as adolescent-onset among adolescents with BD (Rende et al., 2007;Topor et al., 2013;Wilens et al., 1999).
This difference could be in part due to differences in sample size, age, and/or factors related to nationality. Regarding nationality, there is evidence of earlier onset of BD in U.S. samples compared to European samples, which occurs in the context of higher familial loading, more medical comorbidity, and more adversity in U.S. samples (Post et al., 2017). Another factor that likely contributes to differences in prevalence of adolescent-onset compared to childhood-onset is the definition of BD onset. Whereas the current study defined age of BD onset as the age at which the individual first experienced an episode of mania or hypomania, or when study criteria for BD-NOS were met, other studies use different definitions. For example, the COBY study defines BD onset as age at first mood episode (including depression).
Together with the younger age of participants in COBY, this is likely a contributing factor to the higher rate of childhood-onset in COBY.
Analysis of the clinical characteristics associated with childhoodonset largely aligned with prior studies in adolescents with BD. As expected, adolescents with childhood-onset were younger. We found that BD-NOS was significantly more likely than BD-I in childhood onset. This could be due to the nature of subtype progression of the illness. That is, in contrast to adults, there are relatively high rates of diagnostic progression, with 25%-45% of adolescents with BD-NOS later manifesting BD-I or BD-II (Axelson et al., 2011;. Comparatively, COBY found higher rates of BD-I among children (Rende et al., 2007). Although both studies used similar instruments and approach the diagnosis of BD in similar ways, there are a number of factors that could potentially explain this difference.
The current sample is meaningfully older, with an average age in midlate adolescence. In addition, most participants were enrolled within 1-2 years of BD onset. Finally, the recruitment frame, with minimum age 13 years, was constrained by the fact the setting did not include paediatrics or psychiatry services for pre-adolescent children. As such, the current sample is somewhat biased toward adolescent onset. In terms of factors that are thought to explain higher rates of early-onset BD in the United States versus Europe, Canada may be intermediate (Post et al., 2010;Post et al., 2017). Compared to the current sample, the COBY sample also had a lower SES which may explain in part the earlier emergence of BD-I (Birmaher et al., 2009a). Studies in the United States have found that when adversity is higher and coupled with high familial loading, BD onset often occurs earlier (Rende et al., 2007).
The finding that childhood-onset was associated with a comorbid ADHD (and with stimulant treatment) and comorbid ODD aligns with the younger age of these conditions in the general population, and is consistent with both adult (Lázaro et al., 2007), and adolescent BD literature (Faraone et al., 1997;Masi et al., 2006). Those with childhood-onset were more likely to have experienced police contact.
While this has not been investigated in relation to childhood-versus adolescent-onset specifically, an earlier age of BD onset has been associated with police contact within both adult populations (Fazel et al., 2010;Fovet et al., 2015) and described in a prior publication based on the current sample (Barton et al., 2021). Childhood-onset was also negatively associated with lifetime smoking, which aligns with the epidemiology of smoking in adolescents in the general population and in BD specifically (Goldstein et al., 2008;Vermeulen et al., 2019;Wilens et al., 2008).
There was a consistent pattern of higher loading of familial psychiatric history in the childhood-onset group. In univariate analyses, childhood-onset was significantly associated with increased family history of depression, suicidal ideation, ADHD, SUD, AUD, and CD, and the same non-significant trend was observed for mania/hypomania, anxiety, and psychosis. Studies among adults have found Note: Backwards elimination logistic regression analyses were undertaken including variables associated with age of onset (p < .1) in univariate analyses. Age, race and sex were included as covariates.
childhood-onset to be associated with increased family history of mood disorders and SUD compared to adolescent-onset (Baldessarini et al., 2012;Holtzman et al., 2015). In line with our findings, in the COBY cohort, both children and adolescents with childhood-onset had higher rates of familial depression, suicidal behaviour, ADHD, SUD, CD, and anxiety, compared to adolescents with adolescentonset (Rende et al., 2007). In general, across the BD populations, early-onset BD has been associated with higher familial loading compared to later-onset BD (Benazzi, 2004;Birmaher et al., 2009a;Strober et al., 2006). The mechanism underlying earlier BD onset in adolescents with higher loading of familial psychopathology is thought to comprise a combination of genetic and environmental factors (Alloy et al., 2006;Kennedy et al., 2015).
Lastly, in terms of mood symptoms, there were no betweengroup differences in severity of lifetime most severe hypo/manic or depressive episodes, nor were there any between-group differences in individual manic or depressive symptoms within these episodes.
Within adult populations, early onset, whether childhood-or adolescent-, is associated with greater symptom severity for depressive and manic episodes (Leverich et al., 2007;Perlis et al., 2004). Akin to our study, COBY did not find any difference in lifetime most severe manic or depressive intervals ). However, COBY did find between-group differences in specific mood symptom severity. In terms of depression symptoms, adolescents with childhood-onset had more guilt, negative self-image, insomnia, and non-suicidal self-injury as compared to adolescents with adolescent-onset. In terms of manic symptoms, elation, decreased need for sleep, racing thoughts, increased goal-directed activity, and increased energy were all more severe in adolescents with adolescent-onset vs. adolescents with childhood-onset (Birmaher et al., 2009b). Relatedly, a recent long-term follow-up study of the COBY cohort found equally high rates of hypo/manic episodes well into young adulthood among those with childhood-and adolescentonset BD (Hafeman et al., 2020).
The current study has a number of limitations which should be considered. First, while less of a concern than in adult studies, the current study was cross-sectional and retrospective. Second, we do not have data regarding inter-rater reliability of age of BD onset. Third, in contrast to the COBY study, our study did not include a third group of children with BD, and had a small sample size for the childhood-onset group. Fourth, the sample was recruited from a tertiary outpatient setting, limiting generalizability of present findings to primary care or community samples. Lastly, as the childhood-onset group had a significantly longer illness duration than the adolescent-onset group, there was potential for differences in recall between youth with child-onset vs. adolescent-onset BD. This difference in illness duration, despite the relatively large sample, suggests that still larger samples are required to parse the effects of age of onset and stage of illness.
Relatedly, we did not systematically evaluate number of episodes or total amount of time in episodes.
Despite the acknowledged limitations, our study addresses a gap in the literature by investigating a large number of clinical factors associated with childhood-onset BD, in a sample that includes multiple BD subtypes. The pattern of findings overall converge with the prior literature, while also prompting discussion of factors that could explain differences in age of onset, BD subtype, and their interrelationships across different populations. Future longitudinal research is warranted to compare the overall and specific mood symptom manifestations of BD in childhood-vs. adolescent-onset. Gaining a better understanding of how the observed differences manifest, and change with age, can help better understand the trajectory of earlyonset BD. Finally, future research is warranted to unpack the genetic and environmental impact of high familial psychopathology in relation to the emergence, treatment, and perhaps prevention or delay of childhood-onset BD.