Objective: Cognitive-behavioural therapy (CBT) delivered face-to-face and via the internet reduces bulimia nervosa (BN) symptoms. However, our empirical understanding of factors affecting patient outcomes is limited. Method: Using data from a randomised, controlled trial comparing internet-based (CBT4BN, n = 78) with face-to-face (CBTF2F, n = 71) group CBT (97% female, M = 28 years), we examined general treatment (across conditions) and modality-specific predictors of end-treatment and 1-year outcomes (abstinence, binge-eating frequency, purging frequency). Results: Improved eating disorder-related quality of life (EDQOL) during treatment and follow-up predicted abstinence at end-treatment and 1-year assessments. Improved EDQOL, disordered eating cognitions, and anxiety symptoms predicted less frequent binge eating and purging. Previous CBT and being employed predicted more frequent binge eating and purging at both assessments. Higher self-transcendence and self-directedness predicted less frequent binge eating. More severe binge eating and purging at baseline and end-treatment predicted more frequent binge eating and purging at subsequent assessments. Improved EDQOL was more strongly associated with positive outcome in CBT4BN; improved depressive symptoms and health-related QOL predicted positive outcome in CBT4BN but not CBTF2F. Discussion: Symptom improvement and certain character traits predicted positive outcome, whereas more severe presentation and prior CBT experience predicted poorer outcome. Consideration of intreatment symptom improvement may facilitate care recommendations, particularly for internet-based modalities.