Background: Anxiety frequently coexists with depression, and benzodiazepines are often prescribed together with antidepressants. However, benzodiazepines themselves have little or no antidepressive effects and we lack firm evidence for or against this combination therapy. We therefore conducted a meta-analysis of relevant randomized controlled trials to date. Methods: All randomized controlled trials that compared antidepressant-benzodiazepine treatment with antidepressant alone for adult patients with major depression were sought by electronic searches of Medline and several other databases (January 1972 to December 1998), combined with hand searching, reference searching and SciSearch. Two reviewers independently assessed the eligibility and quality of the studies. Relative risks were estimated with random effects model. Results: Aggregating nine studies with a total of 679 patients, the combination therapy group was 37% (95%CI: 19-51%) less likely to drop out than the antidepressant alone group. The intention-to-treat analysis showed that the former were 63% (18-127%) to 38% (15-66%) more likely to show response (defined as 50% or greater reduction in the depression scale from baseline) up to 4 weeks. Limitations: None of the included RCTs followed the patients beyond 8 weeks. Conclusions: The potential benefits of adding a benzodiazepine to an antidepressant must be balanced judiciously against possible harm, including development of dependence and accident proneness, on the one hand, and against continued suffering following no response and drop-out, on the other. (C) 2001 Elsevier Science B.V. All rights reserved.