Mechanisms of Cognitive Behavioral Therapy for Bulimia Nervosa: Correlated Factors of Treatment Effect and Predictors of Treatment Response Based on Resting-State Functional Connectivity

Background Bulimia nervosa (BN) has substantial mortality risk, and cognitive behavioral therapy (CBT) is the first-line treatment; however, remission rates remain modest, and the neurobiological mechanisms of CBT response are unclear. Resting-state functional connectivity (rsFC) may clarify CBT-related brain changes and pretreatment predictors of response, yet longitudinal whole-brain rsFC studies in BN are lacking. Methods Thirty Japanese female with BN received multi-site structured CBT (CBT-E: n = 24; videoconference CBT: n = 4; guided internet-based CBT: n = 2). Resting-state fMRI and symptom measures, including the Eating Disorder Examination Questionnaire (EDE-Q), were obtained pre- and post-treatment. Whole-brain region of interest (ROI)-to-ROI rsFC (142×142) was computed in CONN and harmonized across sites using neuroHarmonize/ComBat. Multiple regression analyses controlling for age and psychotropic medication examined (1) rsFC changes associated with symptom improvement and (2) pretreatment rsFC predictors of response. Significance was assessed using seed-level FDR correction (q < 0.05), applied across all connections from each seed ROI to all target ROIs. A sensitivity analysis was conducted on the CBT-E subgroup (n = 24) using both q < 0.05 and q < 0.1 thresholds. Results Across both the primary analysis including all CBT modalities and the CBT-E–restricted analysis, two findings were consistent (q < 0.05): (i) greater EDE-Q improvement was associated with decreased rsFC between the right orbitofrontal cortex (OFC) and right posterior supramarginal gyrus (SMG), and (ii) stronger pretreatment rsFC between the left postcentral gyrus and the temporo-occipital inferior temporal gyrus predicted greater symptom improvement. At the network level, both analyses linked symptom improvement to increased cerebellum–temporal visual connectivity, and better response to stronger pretreatment connectivity between parietal (SMG/superior parietal lobule) and frontal association regions. Although some edge-level results differed between analyses at q < 0.05, sensitivity analyses using a more lenient threshold (q < 0.1) yielded patterns that largely encompassed the primary findings. Conclusions This is the first whole-brain longitudinal analysis to demonstrate that CBT-related BN improvement may involve rsFC modulation regarding reward-control and perceptual-affective food-cue processing. Pretreatment connectivity related to body perception and cognitive control may also be relevant to responsiveness. These findings provide rsFC markers of CBT-related change and preliminary prognostic features in BN. Trial registration: UMIN000039841. Registered 18 March 2020.