Underlying mechanisms of disruptive mood dysregulation disorder in children: A systematic review by means of research domain criteria

Abstract Background A systematic overview of underlying mechanisms in the new disruptive mood dysregulation disorder (DMDD) diagnosis is needed. The Research Domain Criteria (RDoC) represent a system of six domains of human functioning, which aims to structure the understanding of the nature of mental illnesses. By means of the RDoC framework, the objective of this systematic review is to synthesize available data on children and youths <18 years suffering from DMDD as reported in peer reviewed papers. Methods A literature search guided by PRISMA was conducted using Medline, PsychInfo, and Embase, while the RDoC domains were employed to systematize research findings. Risk of bias in the included studies was examined. Results We identified 319 studies. After study selection, we included 29 studies. Twenty‐one of these had findings relating to >1 RDoC domain. The risk of bias assessment shows limitations in the research foundation of current knowledge on mechanisms of DMDD. Discussion Reviewing self‐report, behavior and neurocircuit findings by means of RDoC domains, we suggest that DMDD youths have a negative interpretation bias in social processes and valence systems. In occurrence of a negative stimuli interpretation, aberrant cognitive processing may arise. However, current knowledge of DMDD is influenced by lack of sample diversity and open science practices. Conclusion We found the six RDoC domains useful in structuring current evidence of the underlying mechanisms of DMDD. Important opportunities for future studies in this field of research are suggested. In clinical practice, this comprehensive summary on DMDD mechanisms can be used in psychoeducation and treatment plans.


INTRODUCTION
Disruptive mood dysregulation disorder (DMDD) can be regarded as laying above a certain threshold on an irritability continuum which needs treatment (Vidal-Ribas et al., 2016). This is also known as clinical irritability. In 2013, DMDD was introduced as a new diagnosis in the DSM-5 within the depressive disorder section (APA, 2013).
DMDD originates from the research syndrome "severe mood dysregulation" (SMD; Leibenluft et al., 2003). DMDD and SMD are characterized by severe, recurrent temper outburst (≥3 per week) and by persistently irritable mood (most of the day in ≥12 months) between the outbursts. Most children with SMD meet DMDD criteria (Deveney et al., 2015;Freeman et al., 2016;Stoddard et al., 2015). In line with previous research (Vidal-Ribas et al., 2016), it seems reasonable to pool these two diagnoses in the present systematic review.
Children with DMDD have severe functional impairment (Copeland et al., 2013;Uran & Kilic, 2020) and adverse outcomes when compared to their treatment-seeking peers without DMDD and children with no psychiatric disorder (Copeland et al., 2014).
Attention deficit hyperactivity disorder (ADHD) is a frequent cooccurring diagnosis in DMDD Rich et al., 2007;Stoddard et al., 2016). However, irritable mood is not a criteria or characteristic in ADHD (APA, 2013). Thus, clinical irritability (i.e., DMDD) needs individual understanding to tailor appropriate treatment.
Keeping abreast of scientific evidence is necessary to adapt and develop effective treatments by combining high quality evidence with clinical expertise (Guyatt et al., 2000). Systematic reviews can provide trustworthy overviews of current evidence (Cipriani & Barbui, 2006). Of note, a narrative review of DMDD in a Research Domain Criteria (RDoC) perspective has been published (Meyers et al., 2017). Contrary to narrative reviews, however, systematic reviews employ explicit methodological strategies to identify relevant studies. This decreases the possibility of erroneous interpretations of study findings (Cipriani & Barbui, 2006). Thus, for professionals trying to help children suffering from the relatively new DMDD diagnosis, there is a specific need for a systematic review of underlying mechanisms of the condition it represents.
The RDoC provides a framework to examine underlying mechanisms of mental disorders (NIMH, 2021). Contrary to diagnostic manuals with focus on symptoms, the RDoC framework integrates several levels of information to capture function in biological and cognitive systems that may create psychopathology. In fact, RDoC Units of Analysis include genetic, neurocircuit, behavioral and selfreport assessments (NIMH, 2021). For instance, the unit behavior includes measurements by cognitive tasks, and neurocircuit analysis by neuroimaging techniques.
To organize findings of different scientific disciplines, RDoC is

METHOD
The present review was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) for systematic reviews (Moher et al., 2009;Shamseer et al., 2015). The PICO method was used to define eligibility criteria, with the applied formulation: Population (children <12 years), Indicator (SMD, DMDD, or oppositional defiant disorder with chronic irritability and anger (ODD-IA; see Supporting Information A), i.e., the equivalent to DMDD in ICD-11), Comparison group (children without SMD, DMDD, or ODD-IA), and Outcome (genes, molecules, cells, circuits, physiology, behaviors, reports, and paradigms units of analysis by means of RDoC domains). Due to no involvement of patients or members of the public, an approval from the regional ethical committee was not applied for.
The search terms were developed and optimized in Embase (see Supporting Information B) and then translated, tested, and reiterated in MEDLINE and PsycInfo according to Bramer et al. (2017). This process was guided by a librarian. The final search terms used in Embase were: ((Disruptive OR severe) ADJ mood dysregulation) OR ((persistent* OR chronic) ADJ3 irritab*) AND (temper ADJ3 (outburst* OR tantrum*)) AND (child* OR youth* OR pediatric*).tw.
AND ((cogniti* OR social OR arousal OR threat OR reward OR motor).tw. OR exp cognition/OR exp Social behavior/), (see Supporting Information C for the translation into MEDLINE and PsycInfo).

Selection criteria
The age criteria were adjusted to <18 years as an initial inspection indicated that only a few numbers of published reports had age criteria as <12 years. Therefore, in this review "youths" refer to children up to 18 years. Exclusion criteria included samples with any individual's age >18 years, full-text not available, non-English language, publication type ≠ peer-reviewed journal article, studies not using DSM or ICD diagnostic criteria, results not related to SMD, DMDD or ODD-IA or RDoC domains, and if IQ of participants <70.

Inclusion and exclusion criteria are detailed in Supporting
Information D.

Identification, selection, and data extraction
Abstracts and titles of the retrieved references were screened independently by the first author. Relevant articles based on selection criteria were screened in full-text versions, including supplementary online content (see Supporting Information), and the data extraction was done in collaboration with last author (see Supporting Information E, for articles excluded). Identification of the associated RDoC domain(s) with each study was performed based on information about the tasks or methods including outcome measures. For example, studies using a face-emotion processing task were identified as associated with Social Processes and Valence Systems due to involving both the ability to recognize face emotions and that the identification of an angry face can elicit responses to aversive situations.

Risk of bias
A practical guide for addressing risk of bias in observational, etiology studies (Savitz et al., 2019) was conducted on the extracted papers.
Contrary to more objective tools for assessing intervention studies, which were also considered (see Supporting Information F), this nonalgorithmic, tailored approach emphasizes identifying a few of the most likely influential sources of bias which can provide meaningful insights to consider in the design of new studies. Based on the authors familiarity with the literature, the sample characteristics (e.g., sample sizes and affiliations) of the included studies were hypothesized as having major influence on the overall knowledge on underlying mechanisms of DMDD.

RESULTS
In the following section we present the results, structured by the six RDoC domains. The initial search yielded 318 articles. After duplicates were removed, 209 articles were left for consideration. Manual editing and selection were performed and resulted in 29 included studies. These were assessed for risk of bias. PRISMA flow chart of identification, screening, eligibility assessment and inclusion are presented in Figure 1.
Citations with identified study characteristics are presented in Table 1. For transparency and verifiability purposes, the indicator (DMDD or SMD) used in each study is specified. No study of ODD-IA was found. The aggregated comorbidity rates of the included studies are 80% for ADHD, 65% for ODD/conduct disorders, 46% for anxiety disorders, and 20% for depressive disorders.
Twenty-one of the included studies were identified as related to more than one RDoC domain. Number of articles per RDoC domain is presented in Figure 2.

Risk of bias
As presented in Table 1, the study samples are characterized by a predominantly U.S. NIMH affiliation and insufficient information to identify similar sample sets (see also Supporting Information G). The sample size median of the included studies is 91 participants and the median of participants in each comparison group is 32. Some studies have not corrected for multiple comparisons in their analyses (Guyer et al., 2007;Thomas et al., 2013Thomas et al., , 2014 whereas other have not provided adequate information to conclude whether such corrections are done (Adleman et al., 2011;Brotman et al., 2010;Deveney, Brotman, et al., 2012a;Freeman et al., 2016;Kim et al., 2013;Pagliaccio et al., 2017;Rich et al., 2011). The aggregated IQ score of the samples is 109, and 111 for the control groups and 106 for the SMD and DMDD groups. There is a lack of knowledge regarding socioeconomic factors as only five studies provide such characteristics (Kircanski et al., 2018;Perlman et al., 2015;Stoddard et al., 2017;Tseng et al., 2019;Uran & Kilic, 2015). Consequences for current knowledge on DMDD mechanisms are detailed in the discussion.

Social processes
Fourteen studies are identified as describing Social Processes. DMDD youths rate themselves as having significantly more social problems than youths with other psychiatric disorders (Freeman et al., 2016).
Also, SMD youths report being more afraid of neutral faces than controls . On behavior level, SMD youths have UNDERLYING MECHANISMS OF DISRUPTIVE MOOD DYSREGULATION DISORDER IN CHILDREN a bias toward angry faces, but not toward or away from happy faces (Hommer et al., 2014), and they have difficulties in identifying emotional prosody (Deveney, Brotman, et al., 2012a). Three studies find that SMD youths have face-emotion labelling difficulties (Guyer et al., 2007;Kim et al., 2013;Rich et al., 2008), whereas others do not (Thomas et al., 2012(Thomas et al., , 2014Tseng et al., 2016). One study finds observable, but not statistically, differences in implicit face-emotion accuracy with SMD having the lowest accuracy compared to healthy controls (HC) and youths with BD (Thomas et al., 2013).
On a neurocircuit level of analysis, some studies find abnormalities in left or right amygdala activation during face-emotion processing Thomas et al., 2012Thomas et al., , 2013 whereas others do not (Kircanski et al., 2018;Stoddard et al., 2017;Thomas et al., 2014;Tseng et al., 2016). Noteworthy, differences in face-emotion processing tasks makes it difficult to compare these results directly. For example, one study examines attention orienting to angry versus neutral faces (Kircanski et al., 2018) whereas another investigates passive viewing and rating of emotional and nonemotional aspects of happy, angry, fearful, and neutral faces . Similar ambiguity yields for other neurocircuit findings. For example, some studies do not detect association between neural activity using fMRI and threat orienting (Kircanski et al., 2018) or responses to face-emotions (Stoddard et al., 2017) in DMDD, whereas others find that SMD youths show hyperactivation in superior temporal gyrus compared to HC when viewing angry faces (Thomas et al., 2014;Tseng et al., 2016).
In sum, social processing difficulties are present in DMDD youths by self-report. By task performance (e.g., behavior) some studies find face-emotion labelling deficits in these youths whereas others do not. There is, however, indication of a bias toward threatening or angry faces. On neurocircuit level of analysis, Examine aware versus non-aware faceemotion labelling deficits.
Behavior + fMRI + Selfreport: Affective priming task: fMRI during "Aware" and "non-aware" priming of shapes by emotional faces (angry, fearful, happy, neutral, blank oval). Subjects rated how much they liked the shape.
Behavior + Self-report: No significant differences between diagnosis on effect of emotions on ratings.
BD responded more quickly than SMD and HV.
Social processes (reception of facial communication) Observable, but not statistically, differences in accuracy with lowest accuracy in SMD youths (see Table 3 in paper).
Social processes (reception of facial communication)  face-emotion processing are inconsistent, possibly due to different processing tasks.

Valence systems
As relevant to Negative Valence System and Positive Valence System, 19 and 15 studies respectively are identified. Responses in negative or positive valence systems can arise from the valuation of angry or happy faces, indicating an overlap between valence systems and social processes. We refer to the social processes section for results associated with face emotion-processing.
Our results show that SMD youths do not have abnormalities on behavior level by task performance in reward or punishment processing (Rau et al., 2008;Rich et al., 2005). In response to frustration, SMD youths report high levels of arousal (Rich et al., , 2011 whereas another study finds no differences in mood ratings in response to reward or frustration (Perlman et al., 2015). However, one study does not report arousal ratings using the same task paradigm in DMDD youths as Rich and colleagues (Tseng et al., 2019). There is some indication of SMD youths having slower reaction time to frustration, but the results are equivocal (Rich et al., , 2011. In response to frustration one study finds increased neuronal activity in anterior cingulate cortex (ACC) and medio-frontal gyrus (MFG) in SMD youths using magnetoencephalography (Rich et al., 2011), whereas another study finds no neuronal activity or connectivity abnormalities in DMDD using fMRI (Tseng et al., 2019). Furthermore, a third study finds that SMD have decreased activation during frustration in ACC and MFG and increased activation during reward (Perlman et al., 2015). The same study found no effects in the amygdala.
One study finds that SMD youths have reduced attention interference from (nonsocial) emotional distracters .
However, during frustration SMD youths demonstrates a disability in shifting spatial attention with associated left amygdala hypoactivation and decreased striatal response, but no prefrontal regions abnormalities . Furthermore, a recent study finds no abnormalities in attention orienting to, or distraction by, threat in DMDD youths on behavioral or neural level (Kircanski et al., 2018).
Taken together, in DMDD there is indication of abnormal responses to frustration by self-report, behavior paradigms and neurocircuit activity. However, the results are ambiguous in terms of different results on reports of arousal in response to frustration. No abnormalities in reward or punishment processing are demonstrated on behavior level.

Cognitive systems
Thirteen studies are identified as relating to Cognitive Systems. By self-report, one study finds that DMDD youths experience more attentional problems than youths without DMDD in psychiatric clinical assessment (Freeman et al., 2016). Two studies have examined attention without social or emotional interference on behavior level. These studies indicate that DMDD youths might have impairments in selective attention and visual attention, but the results are equivocal (Pagliaccio et al., 2017;Uran & Kilic, 2015). On a neurocircuit level, youths with DMDD or ADHD fail to mobilize parietal and related brain regions during long reaction time in a selective attention task (Pagliaccio et al., 2017). The association between attention and frustration or face-emotion processing appears ambiguous Hommer et al., 2014;Kim et al., 2013;Kircanski et al., 2018;Rich et al., 2007Rich et al., , 2010Tseng et al., 2019), and is described under social processes and valence systems as overlap with these domains systems exists. Additionally, on behavior level (performance-based), SMD youths seem to have impaired cognitive flexibility and normal motor inhibition abilities, but the results are obscure (Adleman et al., 2011;Deveney, Connolly, et al., 2012b;Dickstein et al., 2007Dickstein et al., , 2010Uran & Kilic, 2015).
In brief, current research suggests that DMDD youths have impaired cognitive flexibility in addition to attention difficulties.

Arousal, regulatory, and sensorimotor systems
Few findings (four and two, respectively) associated with Arousal/ Regulatory and Sensorimotor Systems were identified. Some studies find that SMD youths report more arousal than HC during frustration (Rich et al., , 2011 whereas others do not . During a motor inhibition task, no differences between SMD and HC, is found on behavior or neural activation measures (Deveney, Connolly, et al., 2012b). tions of others to include deficits in emotional self-monitoring . Hence, DMDD youths might have more general difficulties in processing of self and others. Comparable scores on social awareness in SMD and autism spectrum disorder (Sturm, et al., 2018) highlights this possibility. However, research on DMDD youths understanding of the self and mental states of others is scarce.

DISCUSSION
Problems with attention and cognitive control (i.e., subconstructs of cognitive systems) is indicated in DMDD youths but the results are equivocal. Social and valence mechanisms, which can provide insights to cognitive processes due to overlap between the domains, point to possible perception bias in reception of communication and frustration inducing stimuli in these youths. A recent study suggests that youths with ADHD Combined Type are more inattentive than DMDD youths, but DMDD youths more emotionally labile than ADHD youths (Uran & Kilic, 2020).
These results indicate a difference in mechanisms related to attention with DMDD having a context specific and ADHD having a general deficit in attention. Notably, this also imply that emotional hyperarousal, that is, hyper-lability, likely linked to the presence or activation of the perception bias, might be a unique mechanism of DMDD. In essence, DMDD might be particularly characterized by emotional hyperarousal, and ADHD by cognitive hyperarousal. Indeed, the same study found that children with ≥2 psychiatric comorbidities in DMDD and ADHD Combined Type had significantly higher scores in indexes on "Oppositional," "Inattention," and "ADHD index." This imply that the inattention symptomology worsens when a general inattention deficit and emotional lability interact making inattention problems higher in both emotional and nonemotional contexts. Thus, it may be feasible to examine attention and constructs such as perception and cognitive control in conjunction with other domains as interaction may create the specific symptomology observed in DMDD.
Previous literature argues that DMDD youths exhibit low frustration tolerance, supporting the role of frustrative nonreward processes from the negative valence domain (Meyers et al., 2017).
As apparent from the present review, there is not conclusive evidence for abnormal responses to frustration in these youths.
By building on previous work Meyers et al., 2017;Stringaris et al., 2018) and accounting for the reasoning in the present systematic review, it is still possible to argue that DMDD youths have a specific negative interpretation bias in both social processes and valence systems (i.e., "hot" cognitive abnormalities), and that "cold" cognitive system abnormalities occur primarily in conjunction with such interpretations.
Seemingly inconsistent results of the association between cognitive processes with or without emotional interference (e.g., the involvement of amygdala and the ACC, face-emotion labelling deficits, and responses to frustration), might depend on the instrument's achievement in eliciting "hot" and "cold" processes.
Consistent with a developmental system perspective, DMDD youths might have an immature socioemotional system relative to their cognitive control system, that is, a significant discrepancy in the maturation and connections of their socioemotional system and cognitive control system (Casey et al., 2008;Steinberg, 2008).
Findings of suicidal attempt as unplanned and impulsive in DMDD (Benarous, et al., 2020) speaks to the potential severe consequences of an immature socioemotional system on cognitive control.
Despite the fact that hyperarousal was a criterion for SMD, we conclude that research on arousal and regulatory processes is scarce.
DSM-5 did not include hyperarousal as a DMDD criteria, but as sensitivity to stimuli and activation of neural systems is intertwined with social, valence and cognitive processes and perceptions, regulatory processes and arousal are clearly relevant. If emotional (hyper)lability linked to a negative interpretation tendency is a mechanism of DMDD, as we propose, continued efforts and innovative methods are needed to examine this unique chain of interactions and neural sensitivity. As initially suggested by the SMD criteria, hyperarousal might be a key element in clinical irritability and anger.
Researchers have raised fundamental concerns regarding the validity of DMDD as a diagnostic group (Malhi & Bell, 2019 Uran & Kilic, 2020) did not get a natural affinity in our systematic review. Childhood maltreatment is associated with altered stress responses and symptoms coinciding with DMDD such as irritability and a negative interpretation bias (Bérubé et al., 2021;Teicher et al., 2003). Even if reports of childhood stressors not directly uncover regulatory and arousal mechanisms in DMDD, such results are important in understanding the potential trajectories from normal to abnormal irritability and anger. In keeping with the RDoC framework, future research may benefit from direct measurement of stress responses, such as cortisol levels, in clinical irritability.
Childhood stressors are linked to socioeconomic factors (Farah, 2017;Johnson et al., 2016). In our risk of bias assessment, we identified a pressing issue to address such factors to

Limitations
By focusing on the six RDoC domains, other interesting results might have been insufficiently emphasized as exemplified with childhood stressors. Also, it is difficult to identify appropriate measures of RDoC constructs (Watson et al., 2017). Whether and how units of analysis relate to RDoC domains and constructs is ambiguous. Nevertheless, by tabulating the results in this review including the associated RDoC domains, investigating our evaluations are readily available.
Even though SMD was used to understand DMDD in the present review, the relatively small difference between DMDD and SMD criteria might be significant and becomes evident when examining youths on a group level. Nevertheless, due to scarcity of DMDD research, it seemed reasonable to include SMD research to gain insight into DMDD mechanisms. Indeed, RDoC advocates dimensional approaches to psychopathology. As such, a categorical comparison by means of RDoC can be questionable. However, an overview of underlying mechanisms specific to DMDD as provided in this systematic review is necessary to clinicians giving psychoeducation and treatment to youths currently being diagnosed with DMDD. We argue that both diagnostic categories and dimensional approaches to psychopathology contributes to our understanding mental disorders and serves different and compatible purposes to researchers and clinicians.
Our systematic review does not include non-English or non-

Conclusion
By building on previous literature and accounting for the results of this systematic review by means of RDoC, we argue that DMDD youths have a negative interpretation bias in social processes and valence systems, that is, primarily abnormalities in emotion (vs. cognitive) related perceptions and underlying processes. This is in line with a developmental system perspective with DMDD youths having an immature socioemotional system relative to their cognitive system, and accounts for RDoC domains as overlapping and highly interconnected. Important areas for future research on DMDD mechanisms are continued examination of "hot" and "cold" processes and interactions, using both diagnostic and dimensional approaches to irritability. Additionally, there is a pressing issue to address the associations with arousal and regulatory systems and childhood stressors. Already, the description of DMDD mechanisms in the present review can be helpful in psychoeducation and to develop and advance effective treatment programs. This includes support for clinical trials targeting the negative interpretation bias to improve irritability and anger tolerance and promoting emotion regulation techniques.
Importantly, all future studies on mechanisms of DMDD are encouraged to increase sample diversity and statistical transparency, following open science practices, which can lead to significant improvements in this new field of research.

ACKNOWLEDGMENT
The authors thank Ellen Bjørnstad for her contribution in quality assurance of the research strategy in Embase Classic + Embase, MEDLINE(R) ALL and APA PsycInfo. Supported by South-Eastern Norway Regional Health Authority and Oslo University Hospital.