BACKGROUND: Reviews have compared the efficacy and tolerability of newer second-generation antidepressants with those of placebo or older treatments, but comparative evidence for use of second-generation antidepressants to treat major depressive disorder has not been evaluated. PURPOSE: To systematically evaluate comparative data on the efficacy, effectiveness, and tolerability of commonly prescribed second-generation antidepressants (selective serotonin reuptake inhibitors, bupropion, duloxetine, mirtazapine, and venlafaxine) in the treatment of major depressive disorder. DATA SOURCES: MEDLINE, EMBASE, and PsychLit; the Cochrane Library; and the International Pharmaceutical Abstracts were searched from January 1980 through February 2005 for reviews; randomized, controlled trials; meta-analyses; and observational studies. STUDY SELECTION: The authors reviewed 46 head-to-head randomized, controlled trials comparing one second-generation antidepressant with another. Twenty-four observational studies and placebo-controlled trials were also included for assessment of safety and tolerability. DATA EXTRACTION: Two researchers independently reviewed titles and abstracts. Trained reviewers abstracted data from each study and assigned an initial quality rating. A second reviewer verified the data extraction and quality rating. DATA SYNTHESIS: According to fair to good evidence, the second-generation antidepressants that were compared had only minimal differences in efficacy, and 88% of comparative efficacy studies reported no statistically significant difference in any outcome measure at the end of the study. One effectiveness trial rated good and 2 effectiveness trials rated fair reported no statistically significant differences in primary outcome measures for compared drugs. Meta-analyses showed a modest but statistically significant additional treatment effect for sertraline and venlafaxine compared with fluoxetine. About 96% of comparative trials were sponsored by or had at least 1 author affiliated with a pharmaceutical company; the remaining trials did not report funding sources. Adverse event profiles differed among drugs; however, the degree and quality of adverse event assessment varied and only 13% of trials used a standardized scale to assess adverse events. LIMITATIONS: Quantitative analyses could not be done for many drug comparisons because the quantity and quality of the evidence were inadequate. Most published evidence was from trials sponsored by pharmaceutical companies, and publication bias may have occurred. CONCLUSIONS: Overall, second-generation antidepressants probably do not differ substantially for treatment of major depressive disorder. Choosing the agent that is most appropriate for a given patient is difficult.