Objective Our aim is to study the relationship between dose of baclofen and effectiveness in alcohol dependence. Methods Two hundred two patients with alcohol dependence, who received baclofen treatment for drinking reduction, were followed up for 1 year. For each patient-month of treatment, the maximum daily dose of baclofen (DDB) and average weekly alcohol consumption (AWAC) were calculated. We defined a favorable drinking outcome as an AWAC under 200 g/w for at least 2 consecutive months. We divided the DDB of each patient-month into 3 categories (low dose: <90 mg/d, medium dose: 90–150 mg/d, and high dose: >150 mg/d) and investigated the relationship between reaching a favorable outcome and the concurrent DDB category in a time-varying Cox regression analysis. Hazard ratios (HRs) were adjusted based on age, sex, and initial AWAC. Results One hundred forty subjects were followed during at least 1 month. Of these patients, 58 (41%) had a favorable drinking outcome. In comparison to low dose, medium dose was associated with a decreased rate of favorable drinking outcome (HR = 0.42; 95% CI [0.20, 0.88]), whereas no difference was found with high dose (HR = 1.31; 95% CI [0.65, 2.64]). Conclusion The relationship between dose of baclofen and favorable drinking outcome was U-shaped, that is, was increased at low and high doses compared to medium doses.