What explains the failure to identify replicable moderators of symptom change in social anxiety disorder?

Trials of cognitive behaviour therapy (CBT) for social anxiety disorder (SAD) have struggled to identify replicable moderators of treatment outcome. This could be due to a genuine lack of effects, or a spurious finding caused by methodological factors such as inadequate testing of theory-driven moderators, use of small homogenous samples, failure to model non-linear relationships, and over-reliance on significance testing. We probed explanations for the field's failure to detect moderators by testing whether 15 theory-driven and atheoretical variables moderated treatment outcome in a large heterogeneous sample treated with group CBT for SAD. Moderation was not assessed by only using p-values for linear models, but also by considering effect sizes, plots, and non-linear relationships. Despite using a comprehensive approach to assess moderation, only two variables - the baseline severity of SAD symptoms and fear of negative evaluation (FNE) - were found to moderate social anxiety symptom trajectories. FNE had a non-linear relationship with symptom change that would have been missed using common research methods. Our findings suggest both a genuine lack of effects and limitations of research methods have contributed to the field's inability to identify moderators. We provide suggestions that may increase the likelihood of future researchers detecting genuine effects.