BACKGROUND: Duloxetine has joined venlafaxine on the antidepressant market as a second serotonin-norepinephrine reuptake inhibitor. No previous studies have directly compared these drugs. OBJECTIVE: To compare indirectly the efficacy and safety of extended-release (XR) venlafaxine and duloxetine, the 2 currently available serotonin-norepinephrine reuptake inhibitors (SNRIs) in treating major depressive disorder. METHODS: Outcomes from published, randomized, placebo-controlled trials reporting on moderately to severely depressed patients (Hamilton Rating Scale for Depression [HAM-D] > or =15 or Montgomery-Asberg Depression Rating Scale [MADRS] > or =18). A systematic literature search of Cochrane, EMBASE, and MEDLINE (1996-January 2005) was performed. Two independent reviewers judged the trials for acceptance, and last observation carried forward data were extracted. Differences in remission (8-week HAM-D score < or =7 or MADRS < or =10), response (50% decrease on either scale), and dropout rates from lack of efficacy and adverse events were meta-analyzed using a random effects model. Each rate was contrasted with placebo. Sensitivity analyses were performed to examine the robustness of the results. RESULTS: Data were obtained from 8 trials evaluating 1754 patients for efficacy and 1791 patients for discontinuation/safety. Venlafaxine-XR rates were 17.8% (95% CI 9.0 to 26.5) and 24.4% (95% CI 15.0 to 37.7) greater than those with placebo for remission and response compared with 14.2% (95% CI 8.9 to 26.5) and 18.6% (95% CI 13.0 to 24.2) for duloxetine. Although numerically higher for venlafaxine-XR, no statistically significant differences were found between the drugs; however, both demonstrated overall remission and response rates significantly higher than the rates achieved with placebo (p < 0.001). Reported adverse events were comparable between drugs. CONCLUSIONS: Venlafaxine-XR tends to have a favorable trend in remission and response rates compared with duloxetine. However, dropout rates and adverse events did not differ. A direct comparison is warranted to confirm this tendency.