Objective: To identify predictors of resumed benzodiazepine use after participation in a benzodiazepine discontinuation trial. Method: We performed multiple Cox regression analyses to predict the long-term outcome of a 3-condition, randomized, controlled benzodiazepine discontinuation trial in general practice. Results: Of 180 patients, we completed follow-up for 170 (94%). Of these, 50 (29%) achieved long-term success, defined as no use of benzodiazepines during follow-up. Independent predictors of success were as follows: offering a taper-off program with group therapy (hazard ratio [HR] 2.4; 95% confidence interval [CI], 1.5 to 3.9) or without group therapy (HR 2.9; 95%CI, 1.8 to 4.8); a lower daily benzodiazepine dosage at the start of tapering off (HR 1.5; 95%CI, 1.2 to 1.9); a substantial dosage reduction by patients themselves just before the start of tapering off (HR 2.1; 95%CI, 1.4 to 3.3); less severe benzodiazepine dependence, as measured by the Benzodiazepine Dependence Self-Report Questionnaire Lack of Compliance subscale (HR 2.4; 95%CI, 1.1 to 5.2); and no use of alcohol (HR 1.7; 95%CI, 1.2 to 2.5). Patients who used over 10 mg of diazepam equivalent, who had a score of 3 or more on the Lack of Compliance subscale, or who drank more than 2 units of alcohol daily failed to achieve long-term abstinence. Conclusions: Benzodiazepine dependence severity affects long-term taper outcome independent of treatment modality, benzodiazepine dosage, psychopathology, and personality characteristics. An identifiable subgroup needs referral to specialized care.