Does prediction error during exposure relate to clinical outcomes in cognitive behavior therapy for social anxiety disorder? A longitudinal and observational study.

Exposure therapy is an effective treatment for anxiety disorders, but its mechanisms remain unclear. We explored the violation of threat-predictions during exposure, which has been hypothesized as necessary for symptom change. Forty-three adults undergoing exposure therapy for social anxiety disorder recorded their experiences during 523 in- and between-session exposures by completing smartphone surveys. Participants rated their threat-predictions, threat-outcomes, anxiety levels, surprise, and learning outcome for each exposure. We used multilevel models to explore the effect of prediction error (threat-prediction from outcome) during exposure on anxiety symptoms at the next session, and on symptom change from baseline to three months follow-up. Exposures successfully generated threat-prediction error and unexpected exposure outcomes were experienced as more surprising and likely to stimulate learning. However, larger threat-prediction error during exposure did not predict lower symptoms at the next session, nor overall symptom improvement. These findings suggest that the magnitude of threat-prediction error is not, in itself, sufficient for successful exposure therapy.