This study aimed to describe attachment styles and autonomic response patterns (i.e., vagal tone) to frustration in children diagnosed with DMDD in comparison to typically developing controls without a history of mental health problems. The hypothesis was that children with DMDD would differ in attachment scores and physiological state regulation compared to typically developing matched controls when faced with a frustrating task. The final objective was to examine the relationship between RSA and DMDD symptom severity, as reported by primary caregivers. The hypothesis was that reduced RSA would be associated with more severe symptoms.

Participants were assessed in a single session at their respective schools in Johannesburg, South Africa, between 08:00 and 10:00.

This quasi-experimental study compared children diagnosed with DMDD to healthy age, gender and grade-matched controls. English-proficient participants, between the ages of 7 and 13, were recruited via purposive sampling (Table 1). Attention-deficit/hyperactivity disorder was the most prevalent comorbid disorder in DMDD participants (n = 14). Other comorbid disorders included conduct disorder (n = 1), anxiety (n = 1) and unspecified mood disorder (n = 1). Exclusion criteria for this group were compromised neurology (e.g. epilepsy) or serious illness (e.g. meningitis) in the last 3 months. Control group exclusions included meeting any of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria for DMDD or having immediate family members with a history of mental health conditions (e.g. depressive or anxiety disorders) to minimise potential genetic and environmental counfounds.³⁸ The final sample comprised 30 children and adolescents, who were matched on age, gender and ethnicity (15 per group; 14 male; MAge_Control = 10.13, SDAge_Control = 1.85; MAge_DMDD = 10.13, SDAge_DMDD = 1.73). Sample size was determined using G*Power, aiming to detect medium to large effect sizes between the DMDD and control groups. Practical constraints limited the study to 30 participants, which was sufficient to detect medium effect sizes (dz ≥ 0.53).

The study’s main outcome was to establish whether attachment style and the parasympathetic component of the autonomic nervous system differentiate the response profiles between children diagnosed with DMDD and controls when confronted with a frustrating task. A baseline electrocardiogram (ECG) was recorded before the commencement of the experiment and again after each condition of the experimental task (Figure 1). Participants then completed a self-report measure of attachment style, and parents completed a questionnaire on their perceptions of their child’s social and behavioural difficulties.

The ECG was recorded at 1024 Hz using a wireless single-channel waveform recorder. The inter-beat interval (IBI) series was extracted from the ECG,³⁹ examined visually undetected R-waves and artefacts and edited in CardioEdit.⁴⁰ Editing was required for < 5% of each data file. CardioBatch⁴¹ was used to calculate RSA, which determines the contribution of the vagus nerve to overall heart period (HP). The detrended residual series underwent band-pass filtering to isolate HP pattern variance at the respiratory frequency for children (0.12–1.0 Hz).

A rigged version of the Affective Posner Task^42,43 test elicited frustration in participants by manipulating feedback about response time. The task involved three games with various trials, each consisting of visual cues, targets and feedback. Game 1 served as a baseline with accurate feedback but no reward or punishment. Game 2 provided accurate feedback and rewarded or punished performance. Game 3 induced frustration by giving negative feedback regardless of performance. Frustration was induced only in the final game to increase arousal gradually and prevent carry-over.

The Attachment Style Classification Questionnaire (ASCQ)^44,45 assessed participants’ attachment styles. It comprises 15 items, divided into three factors that correspond to Ainsworth’s²⁵ attachment patterns. These categories include secure attachment (e.g. ‘I usually believe that others who are close to me will not leave me’), anxious or ambivalent attachment (e.g. ‘I’m sometimes afraid that no one really loves me’) and avoidant attachment (e.g. ‘I find it uncomfortable and get annoyed when someone tries to get too close to me’). Responses were scored on a 5-point Likert scale (1 = ‘All wrong’, 5 = ‘Very right’). Avoidant and anxious attachment scores were calculated by averaging their respective items.

The Social Behaviour and Communication: Parent Questionnaire, adapted from the Listening Project Parent Questionnaire,⁴⁶ is an informal structured assessment tool. It evaluates parental observations of their child’s developmental and behavioural challenges across five areas: gestural and facial expression; language and social interaction; auditory processing; emotional regulation and reciprocity and behaviour. It encompasses areas such as hearing sensitivity, speech, spontaneity, emotional control and relatedness (Table 2). Parents reported any difficulties and changes in their child’s behaviour since starting medication, providing examples where possible. Scores ranged from -1 to 1, indicating improvement (-1), no or vague improvement (0) or deterioration (1) in different domains. The total count of problems was determined for each domain.

Baseline demographic and clinical variables not used for matching were compared between groups using the paired t-test for years of exposure to English and McNemar’s test for paired categorical data for the remaining variables.

The effect of group on each of the attachment scores was determined using a repeated measures mixed model with attachment score as the dependent variable and group as independent variable. Group was treated as a repeated measure to capture the paired (matched case-control) nature of the data.

To determine whether the DMDD and matched control groups differed in their autonomic reactivity during the Affective Posner Task, a repeated measures mixed effects model was employed with the outcome (i.e. RSA or HP) as dependent variable and group, test (i.e. game 1, game 2 and game 3) the group-test interaction and age as independent variables.⁴⁷ Group was treated as a repeated measure. Autonomic reactivity was assessed as the difference between baseline and RSA and HP measured during each game. In all analyses, non-normality of the model residuals was established by inspecting model diagnostics. Post hoc comparisons used the Tukey-Kramer adjustment for multiple comparisons.

The correlation between the number of social interaction problems and RSA for Game 3 of the Posner task was measured by Spearman’s correlation coefficient (as data were not normally distributed).

Ethical clearance to conduct this study was obtained from the University of the Witwatersrand Human Research Ethics Committee (Medical) (No. M150526) and the Gauteng Department of Education (No. D2016/049, D2016/335A and D2017/319AA). Participants and their legal guardians gave written informed consent before participation, with appropriate opportunities for withdrawal without prejudice. With permission from the Ethics Committee, participants were not informed of the frustration-inducing aspect of the study, which involved some misleading information. All participants were debriefed immediately after the experiment. Anonymity and confidentiality of results were assured.

There were no significant group differences in the number of years of exposure to English (F(1, 12.9) = 3.48, p = 0.085), home language (χ²(1) = 0.33, p = 0.56), medical aid (χ²(1) = 1.29, p = 0.26) or parental level of education ((χ12(3) = 4.83, p = 0.18; χ12(1) = 2.67, p = 0.10); Table 1).

Disruptive mood dysregulation disorder participants scored significantly higher on avoidant attachment than controls (DMDD: Least squares mean (LSM) = 3.0, 95% CI 2.5–3.5; Control: LSM = 2.0; 95% CI 1.5–2.5; d = 0.98). There were no significant group differences in anxious attachment scores (DMDD: LSM = 2.9, 95% CI 2.5–3.0; Control: LSM = 2.5; 95% CI 2.1–2.9).

Table 3 summarises the descriptive statistics for each group’s RSA and HP measurements before, during and following administration of the Affective Posner Task.

Controlling for baseline and age,⁴⁸ there was a significant game effect for RSA in both groups, with a reduction in RSA from baseline to game 1 (F[1,26] = 9.20, p = 0.005; d = 0.40) and baseline to game 2 (F[1,26] = 6.07, p = 0.021; d = 0.42). The reductions in RSA from baseline to game 3 did not reach significance (Table 4). No significant group differences in RSA emerged from baseline to game 1, game 2 or game 3. Group-game interaction effects on RSA from baseline-to-game were also non-significant. No significant group, game or group-game effects emerged for RSA for game-to-recovery conditions.

A significant group effect for HP was evident throughout all game conditions, with the control group consistently having a higher HP than the DMDD group. From baseline to game 1, the estimated LS-mean HP was 80 msec higher in controls compared to DMDD participants (F[1,14] = 10.27, p = 0.006; d = 1.24). Similar patterns emerged from baseline to game 2 (LSM = 71 msec higher; F[1,14] = 8.11, p = 0.013; d = 1.04) and from baseline to game 3 (LSM = 74 msec higher; F[1,14] = 9.86, p = 0.007; d = 1.16). The group effect on HP was also apparent during game-to-recovery conditions, with the control group exhibiting higher HP than the DMDD group (game to recovery 1: LSM = 74 msec higher; F[1,14] = 9.48, p = 0.008; d = 1.18; game to recovery 2: LSM = 58 msec higher; F[1,15] = 6.95, p = 0.019; d = 0.89; game to recovery 3: LSM = 67 msec higher; F[1,14] = 8.49, p = 0.012; d = 1.14) There were no significant game or group-game interaction effects on HP from baseline-to-game or game-to-recovery conditions (Table 4).

Because of age-appropriate social skills in the control group, the analysis was limited to the DMDD group. A significant negative correlation was found between vagal tone (RSA) during frustration and social interaction issues (rho(28) = 0.52, p = 0.049; Figure 2). Children with lower vagal tone during the frustration condition had the highest score for interaction problems on the Social Behaviour and Communication: Parent Questionnaire.

The results are limited by a few issues. Although not unusual for studies involving clinical populations,^11,59 findings cannot be generalised because of the small sample size. Girls were underrepresented in the sample, which further limits the external validity of the findings. The study was conducted in Johannesburg, South Africa, and findings cannot be extrapolated to the wider South African context. Anthropometric factors (e.g., body weight, height, and waist-hip ratio) that affect RSA were not included because of the sample’s vulnerability and the settings where the data were collected (schools). Participants were also not medication naïve. Additionally, the lack of randomisation and blinding may have introduced selection and observer biases. Future studies should investigate the relationship between RSA and these variables in a similar sample.