Modeling and treating internalizing psychopathology in a clinical trial: a latent variable structural equation modeling approach.

Abstract

Clinical trials are typically designed to test the effect of a specific treatment on a single diagnostic entity. However, because common internalizing disorders are highly correlated ('co-morbid'), we sought to establish a practical and parsimonious method to characterize and quantify changes in a broad spectrum of internalizing psychopathology targeted for treatment in a clinical trial contrasting two transdiagnostic psychosocial interventions. Alcohol dependence treatment patients who had any of several common internalizing disorders were randomized to a six-session cognitive-behavioral therapy (CBT) experimental treatment condition or a progressive muscle relaxation training (PMRT) comparison treatment condition. Internalizing psychopathology was characterized at baseline and 4 months following treatment in terms of the latent structure of six distinct internalizing symptom domain surveys. Exploratory structural equation modeling (ESEM) identified a two-factor solution at both baseline and the 4-month follow-up: Distress (measures of depression, trait anxiety and worry) and Fear (measures of panic anxiety, social anxiety and agoraphobia). Although confirmatory factor analysis (CFA) demonstrated measurement invariance between the time-points, structural models showed that the latent means of Fear and Distress decreased substantially from baseline to follow-up for both groups, with a small but statistically significant advantage for the CBT group in terms of Distress (but not Fear) reduction. The approach demonstrated in this study provides a practical solution to modeling co-morbidity in a clinical trial and is consistent with converging evidence pointing to the dimensional structure of internalizing psychopathology.