Prevalence and Correlates of Disruptive Mood Dysregulation Disorder Among Adolescents with Bipolar Disorder

Introduction

Chronic irritable mood and/or severe angry outbursts are among the most common and challenging complaints in child and adolescent psychiatry (Carlson et al. 2009). Disruptive mood dysregulation disorder (DMDD) – characterized by both symptoms of chronic irritability and frequent, severe angry outbursts (American Psychiatric Association 2013) – was added to Diagnostic and Statistical Manual of Mental Disorders, 5th ed. (DSM-5) in an effort to distinguish youth with this phenotype from those with other disorders (Axelson et al. 2011). Conceptualizing this phenotype, however, as part of a diagnostic entity distinct from mood disorders or disruptive behavior disorders (DBD) has not been without controversy (Axelson et al. 2011). The concept of DMDD evolved from concern that youth with nonepisodic irritability were being misdiagnosed with bipolar disorder (BD) (Blader and Carlson 2007; Axelson et al. 2012). Specifically, DMDD was derived from severe mood dysregulation (SMD) – a concept similar to DMDD, but that also includes chronic symptoms of hyperarousal (Leibenluft et al. 2003).

Two points of contention related to DMDD are the limited research using the current criteria, and its questionable test–retest reliability in field trials (Regier et al. 2013). The majority of evidence for the inclusion of DMDD in the DSM-5 was based on SMD. Youth with SMD were less likely than youth with the narrow phenotype of BD (referring to youth with discrete episodes of mania or hypomania) to have parental BD (Brotman et al. 2007; Leibenluft 2011), or to progress to BD-I, II, or not otherwise specified (NOS) (Brotman et al. 2007), but were more likely to develop unipolar depression and anxiety over a 20 year follow-up period (Leibenluft 2011). These findings have been interpreted to mean that SMD is likely not a precursor to BD, or a broader phenotype or subtype of BD (Carlson et al. 2009; Leibenluft 2011). Although informative, these findings may not apply to youth meeting the current DSM-5 criteria for DMDD.

To fill the gap, recent research on DMDD has been conducted via secondary analyses by applying modified DSM-5 criteria post-hoc to clinical (Axelson et al. 2012; Margulies et al. 2012) and epidemiological samples (Copeland et al. 2013; Dougherty et al. 2014). From these studies we have learned, tentatively, that the community prevalence of DMDD ranges from 0.8% to 3.3%, depending upon the age group (Copeland et al. 2013), with reports as high as 8.2% in 6-year-olds (Dougherty et al. 2014). Further, we have learned that the clinical prevalence ranges from 26% to 30% (Axelson et al. 2012; Margulies et al. 2012), that its diagnostic stability over time may be low (Axelson et al. 2012; Copeland et al. 2013), that in both community and clinical samples it is highly comorbid with internalizing and externalizing disorders, particularly with ODD (Axelson et al. 2012; Copeland et al. 2013; Dougherty et al. 2014), and that long- term functional outcome is likely poor (Copeland et al. 2014).

The potential for misdiagnosis of BD and DMDD has not necessarily been reduced with the advent of DMDD in the DSM-5. Since manic and hypomanic symptoms are part of the exclusionary criteria for DMDD, there is now the added risk of missing a BD illness by diagnosing DMDD. This begs the question of whether these disorders can co-occur. The degree of overlap between BD and DMDD is an outstanding question of clinical relevance. Three studies offer preliminary findings. The longitudinal assessment of manic symptoms study (LAMS) examined operational DMDD criteria in youth with elevated symptoms of mania (ESM), and reported that youth with DMDD were not more likely to have BD at baseline or to develop BD over 2 years than were youth without DMDD (Axelson et al. 2012), although 44% of youth with BD-I or BD-II met criteria for DMDD. Among youth admitted to the hospital with a Child Mania Rating Scale parent version score >20, 17.4% met strictly applied DMDD criteria (Margulies et al. 2012). The mean age of participants in these studies was 9.4 ± 2 years and 9.8 ± 2.1 years, respectively– the younger age range for risk of BD onset and when prevalence of BD is lower. A study examining modified DMDD criteria in offspring of parents with BD compared with community controls found that BD offspring were more likely to meet DMDD criteria at a magnitude slightly higher than other non-BD psychopathology (Sparks et al. 2014). Further, among offspring who developed BD, approximately one third displayed the chronic irritability and frequent temper outburst phenotype, although they did not meet full DMDD criteria (Sparks et al. 2014). To date, no studies have examined DMDD in adolescents rigorously diagnosed with BD. The objectives of this study, therefore, were to evaluate the prevalence of DMDD, and to compare clinical correlates in a clinical sample of adolescents with BD-I, II, or NOS.

Methods

Results

Eight participants (6.9%) were excluded, as data were missing from the ODD supplement and DMDDP could not be determined, yielding a final analyzable sample of 108 participants. Forty-four participants (40.7%) scored above threshold on the KSADS ODD screening question. Twenty-seven participants (25.0%) met the final DMDDP criteria based on DSM-5 DMDD criteria A–G. Figure 1 illustrates the attrition of participants as the DMDDP criteria were applied. Results from the univariate analysis are summarized in Table 1. Prior to correction for multiple comparisons, the DMDDP group was younger (mean age 15.81 ± 1.5 versus 16.46 ± 1.4; p = 0.04), with a higher proportion of Caucasians (92.6% vs. 77.8%) and of a lower socioeconomic status (SES) (44.58 ± 11.9 vs. 49.91 ± 12.8), although the latter two variables only approached significance. After adjusting for multiple comparisons, there were no significant differences on these demographic variables.

OPEN IN VIEWER

FIG. 1.

Attrition of adolescents with bipolar disorder with application of DSM-5 criteria for disruptive mood dysregulation disorder. DSM-5, Diagnostic and Statistical Manual of Mental Disorders 5^th ed; DMDD, disruptive mood dysregulation disorder; KSADS, Schedule for Affective Disorders and Schizophrenia for School-Aged Children, Present and Lifetime version; ODD, oppositional defiant disorder; BD, bipolar disorder; NOS, not otherwise specified.

Compared with participants without DMDDP, participants with DMDDP were not more likely to be diagnosed with a specific bipolar subtype or to have a different age of onset of BD. Furthermore, although almost 50% of both groups reported having a first-degree relative with BD, participants with DMDDP were not more likely to have a family history of BD. Lifetime history of major depressive episode (MDE) was high in both groups (85.2% of DMDDP and 81.5% of non-DMDDP); and there were similar rates of lifetime suicide attempts (30.8% of DMDDP and 24.1% of non-DMDDP). Participants with DMDDP had higher rates of lifetime active suicidal ideation (70.4% vs. 58.2% of non-DMDDP) and lifetime self-injurious behavior (63.0% vs. 46.8% of non-DMDDP); however, the differences were not statistically significant.

The rate of ODD in the total sample of BD adolescents was 36.2% (42/116, data not shown). Among those with DMDDP, the proportion with ODD was 100.0% (27/27). Among those with ODD in our analyzable sample (38/108), 71.0% had DMDDP. Of the 11 with ODD that did not meet DMDDP criteria, 7 screened negative on the first KSADS ODD screening question (reflective of the DMDD phenotype), but scored positive on the second and/or third ODD screening questions and went on to meet ODD criteria; 3 did not have impairment in 2 out of 3 contexts (for ODD diagnosis only an impairment in one context is required); and 1 did not have sufficient information. When utilizing the FDR approach (Strimmer 2008), with every additional test performed, the false positive rate increases; therefore, because of the 100% overlap of DMDDP and ODD in our sample, and in effort to keep the false positive rate low, ODD was removed from the univariate analyses.

Before correcting for multiple comparisons, those with DMDDP had lower rates of lifetime medication use (55.6% of DMDDP vs. 80.2%, p < 0.01); however, this difference only approached significance after correction (p = 0.05). Similarly, those with DMDDP had lower rates of psychiatric hospitalization approaching significance (36.0% vs. 58.0% in non-DMDDP, p = 0.05) that was not significant after correction (p = 0.16).

The parents of participants with DMDDP were significantly more likely to report conflict in the home than were parents of participants without DMDDP (CBQ = 13.48 ± 5.17 vs. 6.04 ± 6.02, p < 0.0001). There was no difference in adolescent-reported conflict. Participants with DMDD were also more likely to have comorbid attention-deficit and/or hyperactivity disorder (ADHD) (63.0% vs. 28.4%; p = 0.007), an assault history (48.1% vs. 15.4%, p = 0.007), and substance use disorders (SUD) (55.6% vs. 28.4%); however, SUD only approached significance after correction (p = 0.05). Finally, the DMDDP group was significantly and markedly lower functioning than the non-DMDDP group (CGAS mean = 50.74 ± 8.37 vs. 58.91 ± 10.0, p < 0.0001).

Logistic regression models to predict DMDDP were first conducted using the three covariates (age, sex, race) and each independent variable. Independent variables included those determined a priori based on the literature (family history BD, intact family, and abuse history) as well as variables that were significant or approaching significance in the univariate analysis (medication use, assault history, lifetime ADHD, lifetime SUD, CGAS score, CBQ parent score). The variables derived from the literature (family history BD, intact family, abuse history) were not statistically significant and therefore, were not included in the final model.

The final logistic regression model was determined using a backwards-stepwise selection procedure using the following variables: Age, sex, race, assault history, CBQ parent score, CGAS score, lifetime ADHD, lifetime SUD, and lifetime medication use as independent variables.

The final logistic regression model was statistically significant (χ² [3] = 34. 659, p < 0.0001), explained 45% (Nagelkerke R ²) of the variance in DMDDP, and correctly classified 84% of cases. Retained variables included: CBQ parent score (odds ratio [OR] = 1.17, confidence interval [CI] = 1.07–1.28; p = 0.001), CGAS score (OR = 0.89, CI = 0.82–0.97); p = 0.006), and lifetime ADHD (OR = 3.3, CI = 0.98–10.94; p = 0.05).

Discussion

In this clinical sample of adolescents with BD, one quarter also met modified DMDD criteria (or DMDDP). This proportion is lower than the proportion of youth with comorbid BD and DMDD reported in the LAMS study (44%) (Axelson et al. 2012); however, we had an older sample and used slightly more conservative DMDD criteria. When only the symptom criteria were applied, the percentage of BD youth with the DMDD phenotype increased to 38%. This is in keeping with Copeland et al. (2013) who reported a decrease in community prevalence as the persistence and duration criteria of DMDD were applied. The overlap of BD and DMDDP of 25% is expected, given that DMDD has more stringent criteria than ODD, and that comorbidity of ODD in the analyzable sample was 36%.

The overlap of BD and DMDDP was not explained by BD subtype. Youth with BD-NOS were no more likely than youth with BD-I to meet DMDDP criteria. Although this finding requires replication, it suggests that DMDD is not a substitute for “misdiagnosed” BD-NOS or BD-II, and that chronic irritability and severe angry outbursts often coexist with discrete episodes of mania or hypomania. As others have emphasized (Axelson et al. 2012; Sparks et al. 2014), this finding reiterates the importance of screening for mania symptoms even when there is a clear DMDD phenotype. It also raises the concern that whereas the potential for misdiagnosing BD may be reduced with DMDD as an alternative diagnosis, the potential for missing a diagnosis of BD may be increased.

Family history of BD was not associated with DMDDP in this sample of youth with BD. Prior related findings have been inconsistent. As mentioned, youth with SMD have lower rates of parental BD than do youth with BD (Brotman et al. 2007; Leibenluft 2011). In the LAMS study, family history of BD was not associated with DMDD at baseline (Axelson et al. 2012). Contrasting our findings, the Pittsburgh Bipolar Offspring Study (BIOS) found that offspring of parents with BD were significantly more likely to have DMDD than offspring of community controls (Sparks et al. 2014). In addition to the primary difference between a high-risk BD offspring study and a study of adolescents affected by BD, a number of methodological differences between the current study and BIOS preclude direct comparison. These differences include older age of the current sample, examination of first- or second-degree family history of BD (vs. parental history), and potential cross-national differences between Canada and the United States. Additional studies are warranted to confirm the familiality of DMDD and BD in relation to one another. In the interim, our study supports the perspective that BD adolescents with DMDDP have a similar rate of familial BD. Taken together with our finding that DMDDP comorbidity was not explained by the BD-NOS subtype, the data supports the conclusion that DMDDP comorbidity is evident among adolescents with equally “classic” BD (i.e., BD-I, familial BD).

After statistical correction, there were no differences between BD youth with and without DMDDP on the demographic variables of age, sex, and race. Interestingly, the high ratio of males to females seen in ODD (Loeber et al. 2000; Burke et al. 2002) has not yet been demonstrated in DMDD (Axelson et al. 2012; Copeland et al. 2013; Dougherty et al. 2014; Sparks et al. 2014). Similarly, although there is considerable evidence to support that a disadvantaged background such as a nonintact family and/or lower SES are risk factors for ODD (Loeber et al. 2000; Burke et al. 2002), these factors were not associated with DMDDP in our sample, and findings from the extant literature are inconsistent (Axelson et al. 2012; Copeland et al. 2013). It is of note that in our sample, there was a high proportion of youth from nonintact families (∼50%), potentially limiting signal detection.

Although the structure and constituents of the family unit were not necessarily associated with DMDDP risk, high conflict at home as reported by parents was one of the independently significant correlates of DMDDP in the logistic regression model. This is consistent with reports from epidemiological studies (Copeland et al. 2013; Dougherty et al. 2014); and Dougherty et al. (2014) reported that parental hostility when a child was 3 years of age was predictive of DMDD at 6 years of age. Because of the paucity of longitudinal data, however, it is unclear if high conflict is a cause or a consequence of DMDD. Regardless, this is an important finding, as it points to the degree of stress in the home environment when an adolescent has DMDD. Why conflict was significant by parental report and not by adolescent report is unclear. We can speculate that youth have less insight into the severity of the conflict at home and/or that they may have a tendency to minimize the problem. Although we are at the preliminary stages of understanding effective management of DMDD, family-targeted interventions are likely a critical component.

Not surprisingly – although not significant in the final model – youth with BD and DMDDP were more likely to have a history of assaulting others, perhaps reflective of poor impulse control. Another telling finding replicated in the literature was the greater functional impairment of BD youth with, as compared to without, DMDDP (Axelson et al. 2012; Copeland et al. 2013; Dougherty et al. 2014; Sparks et al. 2014). Contrary to what one would expect with this level of impairment, however, particularly given the high rates of suicidal ideation, self-injurious behavior and suicide attempts in both groups, youth with BD and DMDDP as compared to youth with BD without DMDDP had lower, not higher, rates of lifetime medication use and psychiatric hospitalization (55.6% vs. 80.2% and 36% vs. 58%, respectively). Interpretation of these findings must be done with caution, however, as these findings were not retained after statistical correction. The extant literature is again inconsistent: Copeland et al. (2013) reported generally increased mental health service use by youth with DMDD, whereas Axelson et al. (2012) reported no difference, also in spite of greater impairment across DMDD groups in both studies. Future studies are needed to determine whether this discrepancy is the result of stigma, noncompliance, methodological limitations, or other reasons.

DSM-5 stipulates that if both ODD and DMDD are present, the diagnosis of DMDD prevails. The clinical utility of DMDD over ODD, however, has been widely debated. In our sample of BD adolescents, 100% of youth with BD and DMDDP also met criteria for ODD. This high rate of overlap was predicted by the Childhood and Adolescent Disorders Work Group (Axelson et al. 2011), and has been reported in other studies. In BD offspring (Sparks et al. 2014) there was a 92% overlap of DMDD and ODD if exclusion criteria for CD were omitted; in the LAMS study (Axelson et al. 2012), there was an overlap of 96% of DMDD and either ODD or CD; and in epidemiological studies, DMDD had the highest co-occurrence with ODD, with ORs ranging from 25 to 103 (Copeland et al. 2013; Dougherty et al. 2014).

What is notable is that the degree of overlap diminishes when examining the rate of DMDD in those with ODD (Axelson et al. 2012; Copeland et al. 2013). For example, 29% of youth with BD and ODD in our sample did not meet criteria for DMDDP. As outlined, the majority of BD youth with ODD did not meet DMDDP criteria because they screened negative on the KSADS-PL ODD question related to temper control, but screened positive on the screening questions related to arguing with adults or defying rules. In other words, youth can meet ODD criteria without displaying temper dysregulation. Our study did not have sufficient power to evaluate what may differentiate these youth. The LAMS study (Axelson et al. 2012) reported no baseline or long-term differences on a multitude of variables, and one small epidemiological study reported that those with ODD alone were more globally impaired (Dougherty et al. 2014). DMDD has been proposed as a mood disorder with chronic irritability as the defining symptom. ODD is characterized as a behavioral disorder strongly associated with and predictive of mood disorders (Stringaris et al. 2009); however, mood symptoms in and of themselves are not among the criteria. Whether this distinction is clinically relevant, particularly in terms of management, remains to be seen.

Although marginally significant in multivariable analyses, ADHD was associated with a threefold increased risk of DMDDP in this sample of BD adolescents, independent of other factors, and the prevalence of ADHD in this comorbid BD and DMDDP sample was 63%. Because 100% of youth with BD and DMDDP also met criteria for ODD, this means that a large proportion of this sample met criteria for BD, DMDDP, ODD, and ADHD. This high rate of ADHD is consistent with the general DMDD literature (Axelson et al. 2012; Sparks et al. 2014), and is expected given that both BD and ODD are highly comorbid with ADHD (Kuhne et al. 1997; Youngstrom et al. 2010; Mitchell et al. 2014). The fundamental question then becomes: What does the diagnosis of DMDD add? Prior to inclusion of DMDD in DSM-5, the option of adding a severe explosive angry outbursts specifier to relevant diagnoses was considered. With the high rates of comorbidity and the competing importance of highlighting the DMDD phenotype, the specifier option may warrant reconsideration.

The rationale for removing the hyperarousal criteria from SMD in the transition to DMDD was to reduce the high rates of comorbid ADHD and SMD, and to isolate a subgroup of chronically irritable youth who may not have hyperarousal symptoms (Axelson et al. 2011). Therefore, in this and other studies, it is curious that hyperarousal symptoms, as part of the ADHD diagnosis, are nevertheless a predominant component of the DMDD phenotype.

There are several limitations to this study. First, as the DSM-5 criteria were not established at the time of study inception, DMDD criteria were applied post-hoc using non-DMDD specific scales. Furthermore, our results may not be comparable to some prior findings – such as those reported by Copeland et al. (2013) and Margulies et al. (2012) – based on different proxy measures for DMDD. Our methods, however, were similar to those of Sparks et al. (2014) and Axelson et al. (2012), and the final DMDDP criteria were the same as DSM-5 criteria, with the exception of age of onset and BD comorbidity. Second, because the DSM-5 criteria for DMDD exclude history of mania or hypomania, the DMDDP in our sample may have been influenced by the presence of episodic mood symptoms of mania and/or depression, and may not be reflective of DMDD without this influence. An objective of this study, however, was to measure the degree of overlap of BD and DMDD, as their delimitation has been controversial. Third, although the overall sample was large (n = 116), the DMDD group was relatively small (n = 27), increasing the chance of type II error caused by insufficient power. With statistical correction using the FDR approach, however, the risk of type I error with multiple testing was minimized. Fourth, the analysis was cross-sectional and we can only comment on associations with DMDD and not longitudinal predictors. Finally, our sample was derived from referrals to a tertiary care center, and may not be generalizable to other clinical or community samples of BD youth.

Conclusions

This study is the first to examine the prevalence and correlates of DMDD in a sample of BD youth. The prevalence of DMDDP was ∼25%, and DMDDP was independently associated with comorbid ADHD, high conflict at home, and greater functional impairment. Despite the positioning of DMDD as phenotypically and biologically distinct from BD, these phenotypes commonly overlap in clinical settings. This overlap is not explained by BD-NOS or nonfamilial BD. Further, the strong association of ADHD with DMDDP in this sample draws into question whether arousal symptoms should have been retained as originally elaborated in the SMD phenotype. To date, no treatments have been specifically recommended for youth with DMDD, as research is limited. Our findings tentatively suggest that comorbid DMDD warrants increased resources and targeted interventions to mitigate the associated functional impairment and family conflict. Larger, longitudinal studies using DMDD-specific measures focusing on predictors of DMDD, DMDD as a precursor to BD, family psychiatric history of BD and other disorders, as well as treatment interventions for DMDD, are required.

Clinical Significance

This study tentatively suggests that comorbid DMDD is common among adolescents with BD, and that these phenotypes are not necessarily distinct: Chronic irritability and severe angry outbursts may often coexist, with discrete episodes of mania or hypomania. As such, it is important to screen for mania symptoms even where there is a clear DMDDP. Clinicians should be aware that although it remains possible that the potential for misdiagnosing BD may be reduced with DMDD as an alternative diagnosis, the potential for missing a diagnosis of BD might be increased unless clinicians recognize that these two phenotypes are not mutually exclusive.