Psychiatric comorbidity in patients with pediatric bipolar disorder – A systematic review

A growing body of evidence suggests that pediatric bipolar disorder (PBD) frequently co‐occurs with comorbid psychiatric disorders that may impact functioning.

K E Y W O R D S adolescent, bipolar disorder, child, comorbidity, pediatric 1 | INTRODUCTION Bipolar disorder is a serious and debilitating mental condition characterized by episodic changes of mood polarity. Over the last decades there has been an increasing scientific interest in bipolar disorder affecting children and adolescents, leading to a growing number of studies on the subject. Bipolar spectrum disorders are estimated to affect 0.003%-3.9% of the child and adolescent population. 1,2 Increasing rates of pediatric bipolar disorder (PBD) has led to a scientific discussion about potential over-diagnosing and the validity of the diagnosis including the relation to Severe Disruptive Mood Disorder or Disruptive Mood Dysregulation Disorder, [3][4][5] but also about a previous lack of recognition of the condition in pediatric patients. 6,7 Although the diagnosis of PBD has been much-discussed in the past decades, recent reviews and task force reports support and manifest the validity of the diagnosis. 6,8 The accurate diagnosis of mood disorders in early stages may help to improve treatment and prognosis, as unrecognized bipolar disorder in children and adolescents may have a serious negative impact on important developmental years where the personal foundation of social relationships, education and adult life is established. 9 A growing body of evidence indicates that PBD frequently co-occurs with comorbid psychiatric disorders such as attention deficit hyperactivity disorder (ADHD), anxiety disorders, substance and alcohol abuse disorder (SUD), and autism spectrum disorder (ASD). 10,11 This indicates the possible presence of specific developmental relationships in context to comorbidity in this group of patients. 11 Comorbidity often has therapeutic implications with regards to treatment choices and response and conversely some pharmacological treatments of comorbid conditions, such as atomoxetine used to treat ADHD, may increase the risk of destabilization of mood. 12,13 Further comorbidity may influence functioning. 11,14,15 Therefore, it is of paramount importance to improve our understanding of comorbid conditions in PBD. Existing reviews concerning the co-occurrence of PBD and other psychiatric disorders include studies that examine the prevalence of PBD in different patient groups, e.g. the prevalence of PBD in patients with ADHD, anxiety or ASD. 11,16,17 To our knowledge this is the first systematic review examining the prevalence of all psychiatric comorbidities in children and adolescents with a primary diagnosis of PBD. We only included studies reporting the prevalence of comorbid disorders in patients with a primary diagnosis of PBD to examine the true patterns of comorbidity in this patient group.
The aim of our study was to systematically review the existing literature on the prevalence of psychiatric comorbidity and general functioning in children and adolescents with a primary diagnosis of PBD. We hypothesized an overall high prevalence of comorbid conditions, and that the prevalence was lower in studies that examined the current prevalence in patients in full or partial remission. 18 Furthermore, we hypothesized that patients with PBD showed a poor general functioning which was specifically decreased in patients with comorbid conditions.

Summations
• No previous review has examined the existing literature on the overall prevalence of all psychiatric comorbidities in children and adolescents with a primary diagnosis of PBD. • We found a high prevalence of comorbidity in patients with PBD, specifically ADHD, ODD and anxiety disorders. • Future original studies should assess current prevalence of comorbidities in patients with PBD who are in remission to obtain more reliable estimates of psychiatric comorbidity in this patient group.

Limitations
• We excluded studies without research-based diagnoses which led to the exclusion of register-based studies and studies that used clinical diagnoses. • Our calculations were based on a limited number of studies. • Included studies may be prone to a selection bias as they mainly include patients from highly specialized psychiatric centers.
2 | MATERIAL AND METHODS

| Information sources and search strategy
We conducted a systematic search on November the 16th, 2022, on the PubMed, Embase and PsycInfo databases with the assistance of an information specialist from Copenhagen University Library. The words and phrases in the search string were written both as indexing terms (e.g. Mesh) and free text phrases with the [Text Word] application. No other database filters were used in the search. We used the following search string: ("children" OR "adolescents") AND ("bipolar disorder") AND ("comorbidity") AND ("attention deficit disorder with hyperactivity" OR "conduct disorder" OR "oppositional defiant disorder" OR "pervasive child development disorder" OR "anxiety disorder" OR "generalized anxiety disorder" OR "obsessive-compulsive disorder" OR "panic disorder" OR "mental retardation" OR "eating disorder" OR "substance-related disorders" OR "tic disorders") The complete search string from PubMed is in Appendix.

| Eligibility criteria
Before our literature search, we established the following eligibility criteria.
Inclusion criteria: Research studies with the aim to estimate the prevalence of psychiatric comorbidity in children and adolescents with a primary diagnosis of bipolar disorder. 3. Research on the following co-morbidities (defined as  psychiatric diagnoses according to DSM-V, DSM-IV,  DSM-IV-TR, DSM-III, DSM-III-R, ICD-9, ICD-10

| Screening and selection process
The screening and selection process was performed by one reviewer (AMF) in the first two steps and by two reviewers in the third step (AMF, KM) in chronological order: 1. Duplicate publications were identified and removed using the duplication function in EndNote and manually. 2. Title and abstract were screened on the remaining reports to select studies that were sought for fullaccess retrieval. 3. Full-text reports were assessed by two reviewers based on the inclusion and exclusion criteria. Records that could not be retrieved online were retrieved through libraries. The full-text assessment led to the selection of the articles included in this review.
Any doubts about eligibility were discussed with a senior investigator (LVK).

| Data extraction
Two reviewers (AMF, KM) made the data extraction independently of each other. We extracted the following data from the included studies on to a preformatted electronic sheet: • Authors and publication year • Number of subjects with bipolar disorder

| Risk of bias of individual studies
We evaluated the risk of bias of the individual studies as we wanted to give each study and their results appropriate weight. We used the STROBE checklist, a 22-item list developed as a guideline for reporting observational studies, as an assessment tool. 19 Each item on the STROBE checklist yielded one point (with a total maximum of 22 points) and half points were given if the item was partially fulfilled. The included articles were independently assessed based on the STROBE checklist by two reviewers (AMF, KM). Afterwards the individual ratings were compared, and any controversy was discussed with a third author (LVK). The STROBE scores were divided into percentages with ≤33% points reflecting low study quality, 33%-66% reflecting moderate study quality and ≥66% reflecting high study quality in accordance with prior procedures. 20

| Functional assessment
If the included studies provided an assessment of the general functioning using Children's Global Assessment Scale (CGAS) 21 or Global Assessment of Functioning (GAF) 22 the scores were categorized as "poor outcome" (score ≤ 50), "moderate outcome" (score 51-70) or "good outcome" (score > 70) in accordance with prior procedures. 23 Both scales use points from 0-100 where increasing scores reflect an improvement in the general functioning. We applied the outcome categories to compare the two assessment scales across multiple studies.

| RESULTS
We identified 4603 records through our database search. After removing 1828 duplicates the title and abstract of the remaining 2775 records were screened. We excluded 2568 records based on their title and/or abstract and the remaining 222 records were sought for retrieval. Two of the reports could not be retrieved online or through libraries. The remaining 220 records were read in their full length and assessed for eligibility. Two hundred of the full-text articles were excluded based on the in-and exclusion criteria. Our search resulted in the inclusion of 20 studies in the systematic review. The studies included a total of 2722 patients with PBD, with a mean age of 12.2 years and 63.3% males. The screening and selection process is depicted in the PRISMA flow diagram in Figure 1.

| Overview of the included studies
A summary of the included studies is shown in Table 1 and contains data according to the variables that were agreed on beforehand. Three of the included studies [24][25][26] have the same study cohort as other included studies, 27,28 but reported results on a few other comorbid diagnoses, which we handpicked to avoid losing the additional results.
Some of the predefined variables are not depicted in Table 1 since few studies reported them. IQ was reported in four studies 25,27,29,30 with a weighted mean IQ of 99.5. Number of hospitalizations was reported in one study 14 with a mean of 1.87 hospitalizations per subject. Number of affective episodes was reported in two studies 15,25 with a mean of 22.7 and 2.28 episodes, respectively.

| Prevalence of comorbid disorder in patients with pediatric bipolar disorder
An overview of the prevalence of comorbid disorders in patients with PBD is presented in Figure 2 and Table 2. The comorbid disorders are presented from most to least frequent. We used the reported prevalences in our calculation of the overall prevalence regardless of prevalence type (lifetime or point). If the study reported both lifetime and point prevalence, we used point prevalence in our calculations, which was the case in Ratheesh et al. 15 If prevalence results were reported in a compound category of multiple diagnoses the results were not used in our calculations, which was the case for the joint category "ADHD/DBDs" in Demeter et al. 31 and "CD/ODD" in Soutullo et al. 32 We also calculated the current (point) prevalence of comorbid disorders in patients while in remission based on the two included studies that reported point prevalence and only assessed patients in remission 15,33 (Figure 2).

| Obsessive-compulsive disorder
Nine of the included studies reported the prevalence of OCD in patients with PBD. 15

| Post-traumatic stress disorder
Five of the included studies reported the prevalence of PTSD in patients with PBD. 15,24,36,38,40 Comorbid PTSD occurred in 0%-9.5% (weighted mean prevalence = 5.5%) of patients with PBD. Current comorbid PTSD occurred in 2% of patients in remission. 15

| Prevalence of comorbid tics or Tourette's
Six of the included studies reported the prevalence of tics or Tourette's in patients with PBD. 26,27,35,[38][39][40] Two of the six studies reported the prevalence of tics alone 26,36 and one study reported the prevalence of Tourette's alone. 39 Comorbid tics or Tourette's occurred in 3.4%-32.1% (weighted mean prevalence = 20.9%) of patients with PBD. No studies reported the current prevalence of tics or Tourette's in patients in remission.

| Prevalence of comorbid mental retardation
One of the included studies reported the prevalence of comorbid mental retardation to be 10.3% in T A B L E 2 Prevalence of comorbid disorders in patients with pediatric bipolar disorder, presented from most to least frequent.  patients with PBD. 29 Mental retardation (defined as an IQ < 70) was an exclusion criterion in 12 of the included studies. 14,15,26,28,31,[33][34][35][36][38][39][40] No studies reported the current prevalence of mental retardation in patients in remission.

| Prevalence of comorbid autism spectrum disorder
Two of the included studies reported the prevalence of ASD in patients with PBD. The studies showed that comorbid ASD occurred in 2.1%-30.3% (weighted mean prevalence = 9.5%) of patients with PBD. 25,34 In Axelson et al. 34 only patients with Asperger's or PDD-NOS were included. Pervasive developmental disorder or ASD was an exclusion criterion in nine of the included studies. 14,15,28,31,33,34,36,38,40 No studies reported the current prevalence of ASD in patients in remission.

| Prevalence of comorbid eating disorders
Three of the included studies reported the prevalence of eating disorders in patients with PBD. The studies showed that comorbid eating disorders occurred in 0%-13.2% (weighted mean prevalence = 5.4%) of patients with PBD. 36

| General functioning of patients with PBD and comorbid disorders
Eleven of the included studies reported a general functioning score on patients with PBD. Seven studies reported a CGAS score 14,15,26,28,34,35,38 and four studies reported a GAF score. 24,25,27,36 Table 3 shows the functional outcome of patients with PBD regardless of comorbidity. The overall general functioning of the children and adolescents with PBD was poor (weighted mean functional score = 45.33). The functional outcome in patients in remission was moderate (weighted mean functional score = 68.3), based the two included studies that assessed patients in remission. 14,15 Table 4 shows the functional outcome of PBD patients with and without a specific comorbidity. In Ratheesh et al. 15 subjects with PBD and comorbid anxiety had a poorer functional outcome ("moderate outcome") compared to PBD subjects without anxiety ("good outcome"). In Dickstein et al. 14 the functional score was lower in subjects with PBD and comorbid anxiety (59 ± 10.6) compared to PBD subjects without anxiety (67 ± 7.9), but both groups were in the "moderate" outcome category. In the remaining studies, involving subjects with PBD and panic disorder, OCD, ASD or ADHD, there was no difference in functional outcome compared to PBD subjects without these specific comorbidities, as all subjects were in the "poor" outcome category.

| Summary
Our aim was to systematically review the existing literature on the prevalence of psychiatric comorbidity and general functioning in children and adolescents with a primary diagnosis of PBD. The search resulted in the inclusion of 20 studies with a total study population of 2722 PBD patients, with a mean age of 12.2 years and 63.3% males.
Our hypothesis of an overall high prevalence of comorbid disorders in patients with PBD was confirmed. The most common comorbidities were ADHD (60%) and ODD (47%). Furthermore, we found a high prevalence of anxiety disorders such as separation anxiety (29%), agoraphobia (27%), generalized anxiety disorder (24%), social phobia (24%), panic disorder (19%), simple phobia (18.7%) and PTSD (5.5%). The prevalence of OCD, conduct disorder, and Tourette's/tics disorder was 21%. The least common comorbid disorders were SUD (13.2%), mental retardation (10%), ASD (9.5%) and eating disorders (5.4%). The prevalence of comorbid disorders was lower in studies that assessed the current prevalence in patients in full or partial remission, 15,33 with the exception of SUD which was higher in patients in remission.
Our hypothesis that patients with PBD showed a poor general functioning was confirmed. However, our hypothesis that general functioning was specifically decreased in patients with comorbid conditions compared with those without was confirmed only in the two studies assessing patients in remission, and not in the four studies where this was unclear (Table 4). Furthermore, our findings reflect that general functioning is specifically decreased during an affective phase. Nevertheless, and more intriguing, our results suggest that comorbid anxiety adds to decreased general functioning of PBD patients even when in remission. Although based on very few studies, this could have clinical implications as it indicates the need for screening and treatment of comorbid disorders in PBD patients, also when in remission, as comorbid conditions can have a negative impact on the patient's wellbeing and overall functioning.
Few previous reviews have reviewed the existing literature on the overall prevalence of psychiatric comorbidity in PBD 11,16,17 and none included a review of all psychiatric comorbidities in children and adolescents with a primary diagnosis of PBD. Our results replicate previous findings of an overall high prevalence of comorbid conditions and a high heterogeneity in prevalence data in patients with PBD. To understand the high prevalence and heterogeneity in prevalence data of comorbid disorders in patients with PBD various factors should be taken into consideration.
First, most of the studies included patients in a manic or depressed state. It has previously been shown that the prevalence of anxiety is lower in adult bipolar patients in remission compared to adult patients during an affective episode. 18 The prevalence of comorbid disorders in patients during an affective episode may not reflect the pattern of comorbidity in the remitted state and the prevalence of many comorbid diseases may be lower in remission, 15,33 which our findings support.
Second, many of the included studies report the lifetime prevalence of comorbid disorders, meaning the prevalence of the disorder from birth until inclusion in the study. When reporting the lifetime prevalence, we cannot know if the comorbid disease is present at the same time as PBD was present in the study. The comorbid condition could have been present before the onset of PBD and may not be relevant anymore at the time of study inclusion. Furthermore, the comorbid disorder could also reflect a diagnosis that later was recognized as bipolar disorder. In both cases, such data may not reflect a "true" comorbid disorder being present during the same time period. An early and false diagnosis could be due to subsyndromal or overlapping symptoms, for example symptoms of aggression or irritability that are often seen in patients with PBD in a manic, depressed or mixed state, but also in patients with ODD. 42 In other words, we cannot know the timewise relationship between the disorders when assessing lifetime prevalence. The timewise relationship is of particularly interest for clinicians, as they would like to know when certain comorbid disorders are present and relevant in PBD.
Third, the included studies may be prone to a selection bias as they mainly include patients from highly specialized psychiatric centers and the study population represents the most complex and ill patients possibly leading to a false high prevalence of comorbid disorders. The prevalence of comorbid disorders may not be as high in the general population of young patients with bipolar disorder with different illness severity. Further, most studies derived from the US, which may also reduce the generalizability of our findings.
It has previously been suggested that the high prevalence of ADHD and ODD in patients with PBD could be due to overlapping symptoms such as hyperactivity, distractibility, and irritability. 3,43 Our study population was mainly males, and the gender distribution may influence the prevalence of comorbid disorders because some disorders such as ADHD and ODD are more common in male subjects. 44 The overall high prevalence of comorbid disorders could also reflect shared etiological components or genetic dispositions to bipolar disorder and comorbid conditions. The surprisingly high prevalence of e.g. OCD and Tourette's/tics disorder could be explained by the socalled Berkson's paradox. It is the paradox, that although two conditions are independent in the community, a misleading positive association between these two conditions may be observed in patients who seek care. This false association occurs because a patient with several diseases is more likely to seek treatment in hospital settings than a patient with only a single disease. 45 Tics affect up to 20% of children at some point 46 and the co-occurrence with PBD may be accidental and due to Berkson's paradox rather than reflecting a shared origin between the two conditions. Another explanation is that early onset of bipolar disorder in childhood or adolescence might resemble a specific phenotype with a more severe course, and therefore more comorbid disorders, than patients with a later onset.
Almost all the included studies were categorized as high or moderate quality based on the STROBE checklist. However, the STROBE score does not necessarily reflect the quality of a report's content, but rather structural elements as determined by the checklist, which we furthermore weighted equally. Because the STROBE checklist was developed as a reporting tool, and not a risk of bias tool, scores should be interpreted in the context of additional information in the tables and main text, for example sample size, medication status and affective phase.
Our review of the current literature implies that research about the prevalence of ASD, mental retardation and eating disorders in patients with a primary diagnosis of PBD is relatively sparse. A small number of the studies in our review included patients with these comorbidities (Table 2). This was mainly because patients with autism and mental retardation were excluded from most studies. There are studies that examine the prevalence of PBD in patients with primary ASD, 47,48 but the prevalence of ASD in patients with primary PBD has not been examined in depth.

| Limitations and direction for future research
Our findings should be seen in the light of certain limitations. Overall, it should be noted that the diagnosis of PBD and its the relation to Severe Disruptive Mood Disorder or Disruptive Mood Dysregulation Disorder is still debated [3][4][5] and the results of the present paper on comorbidity should be regarded in relation to this uncertainty. The high comorbidity rates in PBD are likely not replicated in adult bipolar disorder. Further, studies have noted cross-cultural differences in diagnosis such that US psychiatrists conceptualize explosive temper, chronic behavioral and social and cognitive dysregulation/deficits as a form of bipolar disorder, while UK and EU psychiatrists preferential diagnose pervasive developmental disorder (not PBD). 49,50 On the other hand, one may argue that all included studies used validated diagnostic tools, such as the K-SADS, to categorize patient groups. Furthermore, we have previously shown that the diagnosis of PBD is stable over time. 9 We included studies which used different versions of the DSM and while there are considerable areas of overlap in the different versions, there are also areas of substantial differences. It is not possible to clarify how this may have affected the prevalence of psychiatric comorbidity in PBD in the included studies. We made no eligibility requirements for the prevalence type in the included studies, which led to the inclusion of articles that used point and lifetime prevalence. It would have been interesting to only include articles with point prevalence to calculate an overall current prevalence of comorbid disorders in patients with PBD, as this yields the most useful information about the true prevalence of comorbid disorders. Such an eligibility criterion was not realistic due to the small sample of studies that used point prevalence (N = 7). We excluded studies without research-based diagnoses which led to the exclusion of register-based studies and studies that used clinical diagnoses, however we did this to ensure high diagnostic validity in relation to the diagnostically challenging group of patients with PBD and comorbidities. We included studies examining patients under the age of 18, but this is a relatively broad age span which includes both pre-and postpubescent children and adolescents. Furthermore, the data we collected did not lend itself to sub-analysis with regards to pubertal stage or chronological age. It is possible that phenotypical differences exist between these groups. We only included studies examining patients with a primary diagnosis of bipolar disorder because we wanted to estimate the prevalence of comorbid disorders in patients with PBD. This may have led to the exclusion of articles that examine the bidirectional relationship between bipolar disorder and various comorbidities. Furthermore, our calculations of the overall prevalence of comorbid disorders were based on a limited number of studies, because not every study reported results on all comorbid disorders ( Table 2). Especially our results on the current prevalence while in remission and general functioning were based on few studies.
Future original studies on comorbidity in PBD should aim at making a clear distinction between lifetime and point prevalence to better understand the true prevalence of psychiatric comorbidity in this patient group. They should also aim at including patients in remission because an active affective episode may influence the prevalence of comorbid diseases such as anxiety. 51 It is also important to include patients with ASD and mental retardation in future research, as this subgroup of patients is poorly understood and thus are at risk of being overlooked in clinical settings.
The prevalence of comorbidity in PBD patients can have clinical and therapeutic implications. The treatment of comorbid disorders with psychotropic medication such as SSRI or atomoxetine may induce mania and destabilization of mood. 52,53 Furthermore, an unrecognized or untreated comorbidity in patients with PBD may result in more severe mood symptoms, less response to treatment and a decreased general functioning. 11 To conclude, our review highlights the clinical and scientific importance of comorbidity in PBD. Overall, we found a high prevalence of comorbidity in patients with PBD, specifically ADHD, ODD and anxiety disorders. However, assessing comorbidity in children and adolescents who are not in remission may lead to an overestimation. We also found that patients with PBD generally have a poor general functioning regardless of comorbidities and that it is specifically decreased during an affective phase. Furthermore, our results suggest that comorbid anxiety adds to decreased general functioning of PBD patients even when in remission. Comorbidity may have therapeutic implications with regards to treatment choices and response and conversely the treatment of comorbid conditions may increase the risk of destabilization of mood. Furthermore, an unrecognized or untreated comorbidity in patients with PBD may result in more severe mood symptoms, less response to treatment and a decreased general functioning. Future studies should make a clear distinction between lifetime and point prevalence and assess the prevalence of comorbidity in patients in remission to better understand the true prevalence of psychiatric comorbidity in this patient group. Greater knowledge of the patterns and prevalence of comorbidity in patients with PBD will help clinicians recognize and treat otherwise overlooked comorbid disorders.

CONFLICT OF INTEREST STATEMENT
Lars Vedel Kessing has within the preceding 3 years been a consultant for Lundbeck and Teva. Other authors have no conflicts of interest to declare.

DATA AVAILABILITY STATEMENT
Data sharing is not applicable to this article as no new data were created or analyzed in this study.