Social anxiety disorder (SAD) is common, first-line treatments are often only partially effective, and reliable predictors of treatment response are lacking. Here, we assessed resting state functional connectivity (rsFC) at pre-treatment and during early treatment as a potential predictor of response to a novel attention bias modification procedure, gaze-contingent music reward therapy (GC-MRT).

Thirty-two adults with SAD were treated with GC-MRT. rsFC was assessed with multi-voxel pattern analysis of fMRI at pre-treatment and after 2–3 weeks. For comparison, 20 healthy control (HC) participants without treatment were assessed twice for rsFC over the same time period. All SAD participants underwent clinical evaluation at pre-treatment, early-treatment (week 2–3), and post-treatment.

SAD and depressive symptoms improved significantly from pre-treatment to post-treatment. After 2–3 weeks of treatment, decreased connectivity between the executive control network (ECN) and salience network (SN), and increased connectivity within the ECN predicted improvement in SAD and depressive symptoms at week 8. Increased connectivity between the ECN and default mode network (DMN) predicted greater improvement in SAD but not depressive symptoms at week 8. Connectivity within the DMN decreased significantly after 2–3 weeks of treatment in the SAD group, while no changes were found in HC over the same time interval.

We identified early changes in rsFC during a course of GC-MRT for SAD that predicted symptom change. Connectivity changes within the ECN, ECN-DMN, and ECN-SN may be related to mechanisms underlying the clinical effects of GC-MRT and warrant further study in controlled trials.