Data Analysis

All statistical analyses were conducted using IBM SPSS Statistics 27 (IBM Corp, 2020). Descriptive statistics were calculated for demographic variables, tinnitus presence and severity, sensory triggers, and misophonia presence. Independent samples t-tests were conducted to evaluate potential differences in variable scores between individuals with clinically relevant levels of misophonia and individuals without significant misophonia symptoms as an exploratory and descriptive endeavor. Evaluation of potential sex effects was conducted through multivariate analyses of variance (ANOVA) based on reports of phenotypical differences between males and females with neurodevelopmental disorders sharing symptom characteristics (Ethridge et al., 2017, 2019; May et al., 2019). Age was included as a covariate and retained when significant to control for age-related factors that potentially influence symptom experiences, presentation, and quality of life (Schroder et al., 2013; Palumbo et al., 2018; Jager et al., 2020).

Scored variables from clinical measures (N = 16, see Table 2 for a full list of included measures) were standardized using z-scores for cluster analyses. Variables were selected for clustering based on hypothesized relationships to misophonia or psychiatric risk. Subscales were selected in lieu of total scores to avoid issues with interpreting outcomes associated with combining subscales measuring different symptoms thus preventing a less accurate assessment of sub-phenotypes (e.g., MQ: used the subscale for emotional behaviors and trigger responses over total score). Subgroup formation was determined with the use of Two-Step cluster analysis and silhouette plot evaluation as a data-driven approach to determining the initial input for k-means clustering. The Two-Step cluster analysis outcome was confirmed using silhouette plot evaluation, as the results of silhouette plotting are representations of clustering method outputs. A Two-Step cluster approach identifies sub-groups by running pre-clustering followed by hierarchical clustering methods and provides an estimation for the optimal cluster definition. The Two-Step cluster algorithm outcome suggested two subgroups splitting on the presence or absence of clinically significant levels of misophonia symptoms, however silhouette plots suggested the presence of a third subgroup. Due to this discrepancy, we conducted a two-cluster solution and a three-cluster solution via K-Means Cluster analyses to explore and address potential splits on variable types. K-means clustering provides cluster centroids based on minimizing the sum of squared simple Euclidian distance for the pre-defined cluster number. The k-means algorithm achieved stability after 26-iterations for the three-cluster solution and after 5-iterations for the two-cluster solution. Univariate ANOVAs were run to address group differences by cluster on variables entered into the k-means cluster analysis according to the three-cluster solution. Fisher’s Least Significant Difference (Fisher’s LSD) post-hoc test determined significance between clusters. Current diagnoses were also evaluated by cluster membership according to the three-cluster solution using chi-square analyses.

Results of independent samples t-tests comparing misophonia groups on scored clinical variables.

Mean values for all clinical variables by group and t-scores. T-scores accompanied by non-whole number degrees of freedom are t-tests without assumed variance. p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001.

To investigate the role of anxiety on the relationship between misophonia symptoms severity and emotional behaviors measured via the MQ a bootstrapped simple mediation analysis was performed using PROCESS (Hayes, 2018). Simple self-reported anxiety frequency and intensity were used as individual mediation variables. Anxiety was assessed by asking participants the frequency of which anxiety was experienced (5-point scale from never – all the time) and the intensity of anxiety experienced in a typical day (5-point scale from none – extreme distress). Further bootstrapped mediation analyses were conducted to evaluate anxiety intensity and frequency mediation by cluster. All mediation analysis were bootstrapped 5,000 times.