Contradictory results among randomized clinical trials addressing similar questions are common and occur when the conclusions of different groups of investigators disagree, or when the results of several trials are statistically inconclusive. Meta-analysis, a term used to describe the process of evaluating and combining the results of conflicting studies, has been proposed as a method for reconciling the contradictory results. In this review of meta-analyses, we distinguish between the pooled and methodologic techniques, described the highly variable strategies used, and propose guidelines for improving the conduct of meta-analyses. In pooled analyses the results of multiple clinical trials are combined and the outcome is compared for patients receiving the principal and comparative therapy. In methodologic analyses the clinical trials are judged according to a set of standards used to assess scientific validity and clinical applicability. Since neither technique alone appeared satisfactory for resolving the conflicting results, we propose an approach to meta-analysis that requires methodologic criteria to identify scientifically valid studies, and pooling criteria to combine data from each of the studies. We believe this new strategy of meta-analysis will have enhanced scientific validity and clinical applicability.