OBJECTIVE: To investigate predictors of relapse two years after a brief cognitive-behavior group therapy in patients with panic disorder who had failed to respond to pharmacologic treatment. METHOD: A total of 56 patients with panic disorder were followed who had met remission criteria at 1 year evaluation after 12 sessions of cognitive-behavior group therapy. Demographic and clinical features and life stressors were investigated as predictors of relapse. RESULTS: At the 2 year assessment, 39 (70%) patients maintained remission status and use of medication was reduced significantly, such that 36 (64%) patients were not undergoing any psychiatric treatment. Among all independent variables investigated, only "conflict" as a stressful life event, RR = 3.20 (CI95% 1.60; 7.20 - p = 0.001), and the severity or residual anxiety symptoms, RR = 3.60 for each scale point (CI95% 1.02; 1.08 - p < 0.001), emerged as nonredundant predictors. CONCLUSION: In spite of the high treatment gains across two years of follow-up, clinicians should pay attention to stress management and to the role of residual symptoms during this period. Results were discussed in the context of treatment cost-efficacy and potential strategies to prolong treatment gains from cognitive-behavior group therapy.