Further Evidence of an Association Between a Positive Child Behavior Checklist-Bipolar Profile and a Diagnosis of Pediatric Bipolar Disorder: A Meta-Analysis
oraz | 22 maj 2023
Article Category: Research Article
Data publikacji: 22 maj 2023
Zakres stron: 44 - 59
DOI: https://doi.org/10.2478/sjcapp-2023-0005
Słowa kluczowe
© 2023 Maura DiSalvo et al., published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Although the diagnosis of pediatric bipolar disorder has been controversial, a meta-analysis of population studies reported that pediatric bipolar (BP) spectrum disorders afflict 3.9% of youth worldwide (1). It included seven studies from the United States, and 12 from South America, Central America, or Europe. The prevalence of BP I disorder was 0.6%. Rates of bipolar spectrum disorders were not higher in the United States than in other Western countries.
Pediatric BP disorder is associated with increased risks of distress and disability including academic, social, and family problems, as well as high rates of hospitalization, suicidal ideation, psychosis, and substance use disorders (2-6). Even subthreshold manifestations of pediatric BP disorder have been associated with significant morbidity and disability (7). Therefore, efforts at improving tools to help identify youth suspected of being at risk for pediatric BP disorder in clinical practice could help mitigate its poor outcomes.
Our research group and others have shown that a unique profile of the Child behavior Checklist (CBCL), an empirically derived, easy to use, parent-report instrument with excellent psychometric properties (8, 9), can be useful in aiding in the identification of youth at risk of having bipolar I (BP-I) disorder (10-14). This unique profile consisting of an aggregated elevated scores on the Attention Problems, Aggressive Behavior, and Anxious/Depressed scales (combined score ≥210 or 2 SDs), has been strongly associated with a clinical diagnosis of pediatric BP-I (11-13), and has been variedly referred to as the CBCL-AAA profile, CBCL-Bipolar (BP) profile, and CBCL-Dysregulation profile (DP) (15-17). For the purposes of clarity, we will refer to this profile as the CBCL-BP/DP profile. The items constituting the profile are in the Supplement.
Two decades ago, we conducted a preliminary meta-analysis (18) with the goal of determining the consistency of behavioral problems, as measured by the CBCL scales, associated with pediatric bipolar disorder rather than the diagnosis of the disorder itself. Results showed that the CBCL scales included in the CBCL-BP/DP profile efficiently distinguished pediatric BP disorder from ADHD (18). However, these CBCL scales were examined individually and as continuous measures. Since that time, the CBCL-BP/DP profile was derived and additional papers have been published examining the association between the CBCL-BP/DP profile and a clinical diagnosis of pediatric BP disorder, calling for a re-examination of this important topic.
Further evidence that the CBCL-BP/DP profile can assist clinicians in identifying youth who are at risk of having a BP diagnosis could aid mental health and primary care providers in ensuring these high-risk youth are referred to an appropriate level of care. This issue is particularly relevant considering the critical importance of the differential diagnosis between pediatric unipolar and bipolar disorder, as misdiagnosis of pediatric BP disorder could lead to inappropriate care, including treatment with antidepressant or stimulant medications, which could result in worsening of mood symptoms, suicidality, and hospitalizations (19-22) in those at risk to have BP disorder.
The main aim of the current study was to re-evaluate the strength of the association between the CBCL-BP/DP profile and the clinical diagnosis of pediatric BP disorder. To this end, we performed a meta-analysis of all available studies examining whether the CBCL-BP/DP profile is associated with a pediatric BP diagnosis. We first examined the rate of a positive CBCL-BP profile/DP in youth with and without BP disorder. We then examined the rate of BP disorder in youth with and without a positive CBCL-BP/DP profile. Based on the previous meta-analysis and our own work, we hypothesized that the CBCL-BP/DP profile would be strongly associated with a clinical diagnosis of pediatric BP disorder.
We conducted a literature search of the peer-reviewed literature published through January 28, 2022, to identify studies that examined the association between a clinical diagnosis of pediatric BP disorder with the CBCL-BP/DP profile. We searched the PubMed and PsychInfo electronic databases using the terms (Child Behavior Checklist [Title/Abstract] OR CBCL[Title/Abstract]) AND (Dysregulation [Title/Abstract] OR Emotional Dysregulation [Title/Abstract] OR Bipolar Disorder [Title/Abstract] OR BPD [Title/Abstract]).
Three authors (MD, CHV, JB) screened the articles for relevance and eligibility. Disagreement between the three reviewers was minimal. All three authors (MD, CHV, JB) involved in each step of reviewing the abstracts, full text, and data extraction agreed regarding the final data set. The authors first screened, reviewed, and assessed the reference lists of the retrieved papers. As shown in the PRISMA chart (Figure 1), our initial search identified 417 potentially relevant articles for inclusion. After removing 168 duplicate papers, 249 articles were screened based on the title and abstract and 127 articles were removed because they had titles and abstracts irrelevant to the topic of this review. The remaining 122 articles that had relevant titles and/or abstracts were sought for retrieval and the full reports were assessed for eligibility based on the following criteria: 1) the study assessed the magnitude of the association between the CBCL-BP/DP profile as a categorical measure and diagnosed BP disorder; 2) the study sample was comprised of children and adolescents (ages 6-17); and 3) the study reported the data necessary to calculate odds ratios for meta-analysis (see Data Extraction below for details). Excluded were articles not published in the English language, review articles, editorials, and commentaries. Based on these inclusion and exclusion criteria, 107 articles were excluded, leaving us with 15 relevant articles for meta-analysis. Specific reasons why articles were excluded are detailed in the results.
FIGURE 1.
PRISMA Flow chart of study inclusion
The following variables were extracted from the available studies: sample size, setting (outpatient or inpatient), mean age, and percent males. Detailed information regarding the diagnosis of BP disorder was also extracted and included: structured interview used, informant for the structured interview, individual who administered the structure interview, and what version of DSM was used. For articles that compared youths with a clinical diagnosis of BP disorder with a comparison group, we extracted the number of youths with and without a positive CBCL-BP/DP profile in each group. For articles that compared youths with and without a positive CBCL-BP profile, we extracted the number of youths with and without pediatric BP diagnoses in each group.
We computed two meta-analyses of odds ratios. The first was for studies examining the rates of a positive CBCL-BP/DP profile in youth with BP disorder versus two comparison groups: 1) youth with ADHD, anxiety/depression, or disruptive behavior disorders (DBDs) (termed “youth with other psychiatric disorders” henceforth), and 2) youth who were identified to not have BP disorder and were referred to as “controls.” These comparator groups were not defined a priori and were a result of what had been included in the studies identified for inclusion in the meta-analysis. We combined the multiple psychiatric disorders into one group for analysis given that they are highly comorbid. If a study had multiple comparison groups, we included each in our analysis under the appropriate comparator category. Because of the repeated use of the BP disorder group in studies with multiple comparison groups, measures are not statistically independent of one another and standard statistical procedures will produce inaccurate p-values. To address this intra-sample clustering and to compute accurate p-values for estimates of effect sizes, variance estimates were adjusted using Huber’s (1967) formula as implemented in Stata (23). This formula is a “theoretical bootstrap” that produces robust statistical tests. The method works by entering the cluster scores (i.e., sum of scores within families) into the formula for the estimate of variance. The resulting p-values are valid even when observations are not statistically independent. The second meta-analysis was for studies examining the rates of BP disorder in youth with and without a positive CBCL-BP/DP profile. Our meta-analyses used the random effects model of DerSimonian and Laird(24), which computes a pooled effect size weighted by sample size. We used the
As shown in the PRISMA chart (Figure 1), after screening out duplicate articles (n=168) and articles that did not reference the CBCL-BP/DP profile in the title or abstract (n=127), the literature search identified 122 articles that were relevant. Those articles were retrieved and carefully assessed for eligibility. Of the 122, 18 met all our inclusion and exclusion criteria. Excluded were studies that either 1) were not focused on the relationship between CBCL-BP/DP profile and diagnosed BP disorder (n=73); 2) examined the wrong population (i.e., preschoolers; youth of parents with a clinical diagnosis of BP disorder but no clinical diagnosis themselves, termed youth at risk; or adults) (n=20); 3) examined the CBCL-AAA profile as a continuous measure (n=5); 4) were not original or published research (n=4); or 5) did not give number of youth in each group (n=2). Furthermore, two of the 18 articles that met our inclusion and exclusion criteria were excluded because they used duplicate datasets to other included articles. For the analysis of a positive CBCL-BP/DP profile among youth with and without a BP disorder, we excluded the article that had the lowest number of comparator groups. For the analysis of BP disorder among youth with and without a positive CBCL-BP/DP profile, since the number of comparator groups were the same for each study, we excluded the older study. Additionally, another study that met inclusion and exclusion criteria was exclude because there were zero bipolar disorder cases at follow-up in both the positive and negative CBCL-BP/DP profile groups (n=1). It is standard practice to exclude studies with no events in both arms in meta-analyses of odds ratios, as they provide no indication of either the direction or magnitude of the effect size (27). Thus, 15 articles were included in the final meta-analyses. Given that all studies included in the original review were published prior to the derivation of the CBCL-BP/DP profile, none were included in this meta-analysis and were excluded during the first phase of screening.
Table 1 describes the 15 studies that were included in the meta-analysis along with the three additional studies that met inclusion criteria but were excluded due to using duplicate samples or having zero bipolar disorder cases. Twelve of the 18 studies were cross-sectional and six were longitudinal. Twelve of the studies were conducted in the United States and six examined international samples. Five studies examined rates of a positive CBCL-BP/DP profile in youth with and without a clinical diagnosis of BP disorder, and all found a higher rate of a positive CBCL-BP/DP profile in youth with a diagnosis of pediatric BP disorder versus comparison groups. Thirteen studies examined the rates of BP disorder in youth with and without a positive CBCL-BP/DP profile, but the rates of BP diagnoses in many studies were low, such that an association between a positive CBCL-BP/DP profile and a clinical diagnosis of BP disorder was only reported by six studies. However, it is also important to note that all 13 studies found that a positive CBCL-BP/DP profile was associated with significant psychopathology and suicidality.
Summary of studies that (A) compare rates of a positive CBCL-Bipolar (BP) profìle/Dysregulation profile (DP) in subjects with and without BP disorder, and (B) compare rates of BP disorder in subjects with and without CBCL-BP profile/DP
| A. Studies that compare rates of a positive CBCL-BP profile in subjects with and without BP disorder | |||||
|---|---|---|---|---|---|
| Diler 2009 (29) | BP: N=157(BP-1 N=79; BP-II N=3; BP-NOS N=79) (+CBCL-BP N=92) |
Structured Interview: K-SADS | +CBCL-BP profile defined as sum of scores on Anxiety/Depression, Attention Problems, and Aggressive Behaviors subscales of the CBCL with a cut off score of 210 or higher (2 SD above normal) | Yes | 60% of children with BP disorder diagnosis had a positive CBCL-BP profile |
| Doerfler 2011 (30) | BP: N=27 (N=13 with mania; N=14 with hypomania); (+CBCL-BP N=10) |
Structured Interview: K-SADS-E | +CBCL-BP profile defined as elevated scores (T score >70) on the Anxiety/Depression, Attention Problems, and Aggressive Behaviors subscales of the CBCL | Yes | Children with a BP disorder diagnosis were significantly more likely to have a positive CBCL-BP profile than children with ADHD |
| Biederman 2013 (17) | BP-I: N=140 (+CBCL-BP N=80) |
Structured Interview: K-SADS-E | +CBCL-BP profile defined as sum of scores on Anxiety/Depression, Attention Problems, and Aggressive Behaviors subscales of the CBCL with a cut off score of 210 or higher (2 SD above normal). “Intermediate” CBCL-BP profile defined as a score of ≥180 and <210 | No, duplicate sample (12) | 62% of children with a BP disorder diagnosis had a positive CBCL-BP profile and 80% of profile-negative BP subjects had an intermediate (3180 <210) score |
| Uchida 2014 (31) | BP-I: N=140 (+CBCL-BP N=80) |
Structured Interview: K-SADS-E | +CBCL-BP profile defined as sum of scores on Anxiety/Depression, Attention Problems, and Aggressive Behaviors subscales of the CBCL with a cut off score of 210 or higher (2 SD above normal) | Yes | Youth with BP disorder diagnosis were significantly more likely than both control subjects and those with ADHD to have a positive CBCL-BP profile |
| Kweon 2016 (32) | BP: N=18 (BP-I N=2; BP-II N=11; BP-NOS N=5) (+CBCL-BP N=3) |
Structured Interview: K-SADS-PL | +CBCL-BP profile defined as sum of scores on Anxiety/Depression, Attention Problems, and Aggressive Behaviors subscales of the CBCL with a cut off score of 210 or higher (2 SD above normal) | Yes | 17% of youth with a BP disorder diagnosis had a positive CBCL-BP profile, compared to 5% of subjects with depression (NS) (limited power due to small sample size) |
| B. Studies that compare rates of BP disorder in subjects with and without CBCL-BP profile | |||||
| Volk 2007 (33) | +CBCL-BP: N=33(BP N=0)
|
Missouri Assessment for Genetics Interview for Children | +CBCL-BP profile defined as T-scores ≥70 on the Anxious/Depressed, Aggressive Behavior, and Attention Problems subscales | Yes | A positive CBCL-BP profile correlated with diagnoses of ADHD, ODD and CD as well as high rates of suicidality |
| Holtmann 2008 (34) | +CBCL-BP: N=62(BP N=0); |
ICD-10 diagnoses based on the diagnostic guidelines of the German society for child and adolescent psychiatry | +CBCL-BP profile defined as T-scores ≥70 on the Anxious/Depressed, Aggressive Behavior, and Attention Problems subscales | Yes | A positive CBCL-BP profile correlated with disruptive behavior disorders |
| McGough 2008 (35) | +CBCL-BP: N=45 (BP N=3) |
Structured Interview: K-SADS-PL | +CBCL-BP profile defined as T-scores ≥70 on the Anxious/ Depressed, Aggressive Behavior, and Attention Problems subseales |
Yes | A positive CBCL-BP profile was associated with increased generalized anxiety disorder, oppositional defiant disorder, conduct disorder, and parental substance abuse |
| Meyer 2009 (36) | +CBCL-BP: N=16 (BP diagnosis at Time 4 N=2; Time 5 N=5). |
Structured Interview | +CBCL-BP profile defined as T-score ≥60 on the Attention Problems, Aggressive Behavior, and Anxious/Depressed subscales at least once during childhood and/or adolescence (1 SD) | Yes | 31% of youth with a positive CBCL-BP profile developed BP disorder overtime |
| Biederman 2009 (12) | +CBCL-BP: N=28 (BP N=10) |
Structured Interview: Age <18: K-SADS-E Age ≥18: Structured Clinical Interview for DSM-III-R | +CBCL-BP profile defined as sum of scores on Anxiety/Depression, Attention Problems, and Aggressive Behaviors subscales of the CBCL with a cut off score of 210 or higher (2 SD above normal) | Yes | Over a mean follow up period of 7.4 years, a positive CBCL-BP profile predicted subsequent diagnoses of bipolar disorder, major depressive disorder and conduct disorder, as well as impaired psychosocial functioning and higher risk for psychiatric hospitalization |
| Althoff 2010 (15) | +CBCL-BP: N=57 (BP N=O) Comparison subjects: N= 1516 (BP N=6) |
Structured interview: Composite International Diagnostic Interview (ClDI) | +CBCL-EP profile defined by latent class analysis (LCA) as the class with the highest elevations on the Attention Problems, Aggressive Behavior, and Anxious/Depressed subscales. No specific cutoff was reported in the paper | Yes | The presence of a positive CBCL-BP profile at Wave 1 in childhood was associated with increased rates of adult anxiety disorders, mood disorders, disruptive behavior disorders, and drug abuse 14 years later |
| Biederman 2012 (13) | ADHD and +CBCL-BP: N=45 |
Structured Interview: Age <18: K-SADS-E Age ≥18: Structured Clinical Interview for DSM-III-R and DSM-IV Axis 1 Disorders | +CBCL-EP profile defined as sum of scores on Anxiety/Depression, Attention Problems, and Aggressive Behaviors subscales of the CBCL with a cut off score of 210 or higher (2 SD above normal). CBCL-DESR defined as a score of ≥180 and <210 | No, duplicate sample (12) | Children with ADHD and a positive CBCL-BP profile had more impairing ADHD symptoms, higher rates of psychiatric hospitalizations, and higher rates of CD, ODD, and bipolar disorder compared to ADHD and control subjects |
| Wilens 2013 (37) | +CBCL-BP: N=43 (BP N=35) |
Structured Interview: Age <18: K-SADS-E |
+CBCL-EP profile defined as sum of scores on Anxíety/Depression, Attention Problems, and Aggressive Behaviors subscales of the CBCL with a cut off score of 210 or higher (2 SD above normal). CBCL-DESR defined as a score of ≥180 and <210 | Yes | Subjects with a positive CBCL-BP profile had higher risk for psychiatric disorders including BP disorder, CD, multiple anxiety disorders, and ADHD |
| Mbekou 2014 (38) | +CBCL-BP: N=150 (BP N=10) |
DSM-IV-TR diagnostic information was obtained from patients’ clinical files and included for analyses. | +CBCL-BP profile defined as sum of scores on Anxiety/Depression, Attention Problems, and Aggressive Behaviors subscales of the CBCL with a cut off score of 210 or higher (2 SD above normal) | Yes | There were no significant differences in the number of youths diagnosed with BP disorder in children with a positive CBCL-BP profile versus those with a negative profile |
| Peyre 2015 (39) | +CBCL-BP: N=42(BP N=1); |
Structured Interview: K-SADS-PL | +CBCL-BP profile defined as T-scores >70 on the Attention Problems, Aggressive Behavior, and Anxious/Depressed subscales of the CBCL | Yes | A positive CBCL-BP profile was associated with ODD, anxiety disorders, MDD, emotionality, and “self-directness” |
| Caporino 2016 (28) | +CBCL-BP: N=17 (BP N=0) |
Structured interview: Composite International Diagnostic Interview (ClDI) | Latent profile analysis (LPA) was used to stratify patients into dysregulated (+CBCL-BP) versus non-dysregulated based on CBCL Attention Problems, Aggressive Behavior, and Anxious/Depressed subscale scores. |
No, neither group developed BP disorder by follow-up | At the 7- to 19- year follow-up, the dysregulated group had significantly higher rates of PTSD, agoraphobia, panic attacks, and QCD within the last year, as well as rates of lifetime PTSD |
| Dolitzsch 2016 (40) | +CBCL-BP: N=32 (BP N=2); |
Structured Interview: K-SADS-PL | +CBCL-BP profile defined as having a T-score ≥67 on the Anxious/Depressed, Attention Problems, and Aggressive Behavior subscales of the CBCL | Yes | A positive CBCL-BP profile carried a high (>2) relative risk for manic episodes, bipolar disorder, cyclothymia, hyperkinetic conduct disorder, mixed disorders of conduct and emotion, and suicidal ideation |
| Joshi 2018 (41) | ASD with +CBCL-BP: N=44 (BP N=20) |
Structured Interview: K-SADS-E | +CBCL-BP profile defined as sum of scores on Anxiety/Depression, Attention Problems, and Aggressive Behaviors subscales of the CBCL with a cut off score of 210 or higher (2 SD above normal) | Yes | ASD youth with a positive CBCL-BP profile had a greater severity of autism, significantly higher rates of disruptive behavior disorders, MDD, and BP disorder, and had more psychosocial dysfunction |
Two of the eighteen articles were excluded from the meta-analysis because they contained a data set that was duplicated by another paper, and another article was excluded because the rate of BP disorder at follow-up was zero in those with and without a positive CBCL-BP/DP (28). Thus, fifteen of the eighteen studies were included in the meta-analysis; four studies were included in the first analysis and eleven studies were included in the second analysis.
Five studies (17, 29-32) compared the rates of a positive CBCL-BP/DP profile in youth with and without a clinical diagnosis of BP disorder, but one study was excluded due to use of a duplicate data set(17). Two studies included in the meta-analysis had multiple comparison groups (29, 31).
Four studies contributing five comparison groups compared the rates of a positive CBCL-BP/DP profile in youth with BP disorder and those with other psychiatric disorders (ADHD, anxiety/depression, or DBD) (30, 31). The pooled OR was greater than one and statistically significant, indicating significantly increased odds of having a positive CBCL-BP/DP profile in youth with a clinical diagnosis of BP disorder compared to those with other psychiatric disorders (pooled OR=4.34, 95% CI=2.82, 8.27; p<0.001) (Figure 2). Results remained significant after correcting for intra-sample clustering (p=0.02). Overall, heterogeneity was high and significant (I2=72.1%, p=0.006), suggesting significant variability between study effects. There was no evidence of publication bias as determined by Egger’s test (p=0.81).
FIGURE 2.
Forest plot for the meta-analysis of positive CBCL-BP profiles in youths with bipolar disorder vs. youth with ADHD, controls, youths with disruptive behavior disorders, and youths with depression/anxiety. Odds ratios >1 indicate increased odds for the CBCL-BP profile in youth with bipolar disorder.
Two studies compared the rates of a positive CBCL-BP/DP profile in youth with BP disorder and in control groups (29, 31). These were clinical healthy control groups who were free of any lifetime episode of any major psychiatric disorder in the Diler et al. (29) study and free of ADHD and mood disorders in the Uchida et al.(31) study. The pooled OR was greater than one and statistically significant, indicating significantly increased odds of having a positive CBCL-BP/DP profile in youth with a clinical diagnosis of BP disorder compared to control groups (pooled OR=34.77, 95% CI=2.87, 420.95; p=0.005) (Figure 2). Overall, heterogeneity was high and significant (I2=82.7%, p=0.02), suggesting significant variability between study effects. There were too few studies to assess publication bias. It should be noted that the pooled OR calculated for youth with BP disorder versus control groups is less precise due to the small number of studies contributing to the analysis. Because the large pooled OR has a wide confidence interval, its magnitude needs to be interpreted with caution.
Thirteen studies (12, 13, 15, 28, 33-41) compared the rates of BP disorder in youth with and without a positive CBCL-BP/DP profile, however one study was excluded due to use of a duplicate data set (13) and another study was excluded because the rate of BP disorder at follow-up was zero in those with and without a positive CBCL-BP/DP profile (28). Thus, eleven studies had extractable data and were included in the meta-analysis. The pooled OR was greater than one and statistically significant, indicating significantly increased odds of having a clinical diagnosis of BP disorder in youth with a positive CBCL-BP/DP profile compared to those without it (pooled OR=4.25, 95% CI=2.12, 8.52; p<0.001) (Figure 3). Heterogeneity was high and significant (I2=60.9%, p=0.004), suggesting significant variability between study effects. There was no evidence of publication bias as determined by Egger’s test (p=0.21).
FIGURE 3.
Forest plot for the meta-analysis of bipolar disorder diagnoses in youths with and without a positive CBCL-BP profile. Odds ratios >1 indicate increased odds for a bipolar disorder diagnosis in youth with the CBCL-BP profile compared to youth without it.
The results of our literature search and meta-analyses showed that youth with a diagnosis of BP disorder were at significantly increased odds of having a positive CBCL-BP/DP profile when compared to those with ADHD, DBD, anxiety/depression, and controls. Furthermore, youth with a positive CBCL-BP/DP profile were at significantly increased odds of having a clinical diagnosis of BP disorder when compared to those without it. These findings are consistent with and expand upon a previous meta-analysis, performed almost 20 years ago, which also found that the CBCL-BP/DP profile could distinguish between youth with BP disorder and those with ADHD(18). These findings are also consistent with previous data showing that a positive CBCL-BP/DP profile is associated with a clinical diagnosis of pediatric BP disorder both cross-sectionally and longitudinally(11-13, 42).
While other groups have found that a positive CBCL-BP/DP profile has limited ability to differentiate BP disorder from other psychiatric diagnoses, they have noted that a positive CBCL-BP/DP profile was associated with high levels of psychopathology, multiple psychiatric diagnoses, suicidality, and poor psychosocial functioning (16, 29, 30, 32-36, 43, 44). For example, Holtmann et al. (43) found that youth with a positive CBCL-BP/DP profile at age 8-11 had increased rates of ADHD, mood and substance use disorders, suicidal ideation and attempts, and poorer functioning at age 19, but noted that none of the patients in the sample were diagnosed with BP disorder. Such differences among studies are understandable given that some controversy remains about how to implement the DSM criteria for BP disorder in youth, especially for young children (45).
Meyer et al. (36) found that 31% of high risk offspring of parents with BP disorder who had a positive CBCL-BP/DP profile developed BP disorder over the course of a 20 year follow up, but also concluded that several other conditions (ADHD, cluster B personality disorder, anxiety) were equally or more likely to be predicted by this profile. Given that pediatric BP disorder has a high overlap with psychiatric disorders such as ADHD, anxiety, and mood and substance use disorders, and has high rates of suicidality, it seems reasonable to hypothesize that a profile that would predict pediatric BP disorder would also predict a similar constellation of psychiatric impairments.
Studies that examined pediatric BP disorder among youth with and without a positive CBCL-BP/DP profile were less likely to find an association between a positive CBCL-BP/DP profile and a BP disorder diagnosis than those studies that examined a positive CBCL-BP/DP profile among youth with and without BP disorder. Because the prevalence of pediatric BP disorder in many of the studies examining pediatric BP as an outcome was very low (<2%; (33-35, 38) they may have had low power to detect the association. In addition, prevalence and the positive predictive value (PPV; i.e., the probability that a person with a positive test has the disorder) of an assessment tool have a direct relationship; as prevalence decreases, so does the PPV. Thus, in these studies with a low prevalence of BP disorder, it is not surprising that the relationship between a positive CBCL-BP/DP profile and a BP disorder diagnosis was not robust.
None of the studies we reviewed provided information about disruptive mood dysregulation disorder (DMDD). Given that that disorder had been created, in part, to reduce diagnoses of pediatric BP disorder (46), it is reasonable to hypothesize that the CBCL-BP/DP profile would be positive for many patients diagnosed with DMDD. We should also consider the possibility that the CBCL-BP/DP profile may be a non-specific indicator of severe emotional dysregulation, which is often seen in ADHD and other disorders (47). For example, in a study of 417 newly referred youth, we compared youth positive for the CBCL-BP/DP profile, those with more modest scores on the sum of the constituent subscales (≥ 180 and <210) and those with scores < 180. Patients with the CBCL-BP/DP profile were significantly more impaired on all measures of social and executive functioning compared to the other two groups, but the group with modest elevations also showed impairments compared with those with no elevation (48). Future work will need to further address these issues.
Our findings have significant clinical and public health implications. While the results support the idea that the CBCL-BP/DP profile may be a useful tool for identifying youth who may be at high risk of having a clinical diagnosis of BP disorder, they also support the notion that it could be a useful marker of the presence of significant psychopathology and dysfunction in youth. As the CBCL is easy to administer, reliable, inexpensive, and available in multiple languages, it could easily be used in both pediatric and child psychiatry clinics to aid in identifying youth at high risk for compromised outcomes. In addition, six out of the 18 studies presented in our paper utilized international samples, providing evidence that the CBCL may be a useful tool for international collaboration in the field of pediatric BP disorder.
Although the CBCL-BP/DP profile is not diagnostic of pediatric BP disorder, a diagnosis that requires careful psychiatric evaluation to diagnose, it may be a useful tool for helping clinicians identify youth at high risk of severe psychopathology and could potentially be used as part of a triage system in primary care or community clinics to assists in determining whether referral to a specialist is needed for a particular patient.
Our meta-analyses found substantial heterogeneity among studies. This likely reflects the unresolved controversies about how to best diagnose BP disorder in youth. Our group defines pediatric bipolar disorder according to DSM criteria for BP-I; however, criteria for defining pediatric bipolar disorder varied across studies. Six of the fifteen studies included in the meta-analysis used a diagnostic interview and KSADS-E (12, 31, 37) or K-SADS-PL (29, 32, 35) to assess for bipolar spectrum disorders (BP-I, BP-II, and cylothymia)(35). BP-I (12, 29, 31, 32, 37), BP-II (29, 32, 37), or bipolar not otherwise specified (BP-NOS) (29, 32) according to DSM-III-R (12) and DSM-IV criteria (12, 29, 31, 32, 35, 37). Additionally, three studies used a diagnostic interview and K-SADS-E (30, 41) or K-SADS-PL (39) to diagnose for bipolar disorder according to DSM-IV criteria but did not specify how bipolar disorder was defined (30, 39, 41), two studies assessed for ICD-10 diagnoses of bipolar disorders using the diagnostic guidelines of the German society for child and adolescent psychiatry (34) or the K-SADS-PL (40), one study only used diagnostic interviews to assess for BP-I or BP-II according to DSM-IV criteria (36), one study used a diagnostic interview and the Composite International Diagnostic Interview (CIDI) to assess for BP-I or BP-II according to DSM-IV criteria (15), one study used the Missouri Assessment for Genetics Interview for Children (MAGIC) to assess for BP-I and BP-II according to DSM-IV criteria (33), and one study obtained DSM-IV-TR diagnostic information from patients’ clinical files and did not specify how bipolar disorder was defined (38).
As reviewed by Goldstein et al. (45), controversies about pediatric BP include what is the best measurement tool for diagnosing BP disorder and whether informants are essential for diagnosis. Marangoni et al. (49) note that when conduct and oppositional disorders are comorbid with ADHD, those disorders will increase aggressive behaviors and make the diagnosis of BP disorder less sure. They also point out that some symptoms characteristic of BD can be difficult to differentiate from the severe end of emotional and cognitive states common in youth. These uncertainties likely added to the heterogeneity of results among studies.
Our findings should be reviewed considering some methodological limitations. Because this was a meta-analysis it is limited by the quality of the data available in the published literature and the data they provided. Because the studies did not provide consistent data about age or studied a wide age range, we could not conduct separate analyses for children and adolescents. The number of available papers examining this subject was relatively small and the number of papers that were able to be used in the meta-analysis was even smaller. As a result, the small number of studies included in the analysis of a positive CBCL-BP/DP profile among youth with BP disorder versus controls produced a pooled OR with an extremely wide confidence interval, indicating a lack of precision. Thus, this result should be interpreted with caution. Additionally, we had to exclude one study that had zero patients with BP disorder at follow-up in both the positive and negative CBCL-BP/DP profile groups (28). Studies with zero events provide no indication of either the direction or magnitude of the effect size, thus we do not know how this paper would have impacted the results. Furthermore, there was substantial heterogeneity in the samples, which may make our results not easily generalizable to all samples. One potential source of heterogeneity could be the different definitions used to define a positive CBCL-BP/DP profile. While most studies used a CBCL-AAA T-score ≥210 or T-scores ≥70 on all three scales that make up the CBCL-AAA to define a positive CBCL-BP/DP profile, four studies used different definitions (Table 1). For these four studies, a positive CBCL-BP/DP profile was defined by T-scores ≥60 on all three CBCL-AAA scales (36), T-scores ≥67 on all three CBCL-AAA scales(40), or by latent class analysis or latent profile analysis with no specific cut-off reported (15, 28). The number of studies was too small to perform an analysis stratified by CBCL-BP/DP profile definitions, thus we do not know how the differences in definitions impacted the results. Given the discordance in how the CBCL-BP/DP profile is defined throughout the literature, more work needs to be done to identify the most appropriate cutoff for this profile.
Another potential source of heterogeneity is from combining ADHD, anxiety/depression, and DBDs into one comparator group called “other psychiatric disorders.” The number of studies were too small to perform stratified analyses based on the different comparator disorders, but these disorders are highly comorbid, supporting the decision to combine them into one comparator group. Future research would benefit from more studies examining the rates of a positive CBCL-BP/DP profile in youth with a clinical diagnosis of BP disorder versus youth with other specific psychiatric diagnoses. Additionally, we did not utilize a protocol to perform this meta-analysis and did not perform a systematic review of each paper for biases, which may increase the risk of bias in our findings.
Despite these considerations, our systematic review and meta-analysis showed that the presence of a positive CBCL-BP/DP profile in youth may be a useful tool to help in identifying youth with a likely diagnosis of BP disorder and other significant psychopathology and dysfunction.