To determine the efficacy of continuing medical education (CME), we collected 248 original articles describing studies of CME interventions. These articles were reviewed for applicability and scientific credibility by applying preset methodological criteria. Thirteen percent of articles described randomized trials, but only 7% of all articles and 20% of randomized trials assessed the impact of CME on patient outcomes. Seven articles met all our criteria and were reviewed in detail. These studies provide convincing evidence that CME can improve physician behaviors. However, only three of these methodologically sound studies assessed patient outcomes and only one demonstrated any improvement in outcomes.