Drug trials are designed to assess efficacy, so the number of subjects on which FDA approval is based is often not sufficient to assess risks. Meta-analysis thus offers a means to assess safety, as seen in recent controversial studies of cardiovascular risks of rofecoxib (Vioxx) and rosiglitazone (Avandia) and suicidality associated with antidepressants in adolescents reflect this situation. Such meta-analyses cannot be definitive because the number of adverse events is small, so the results depend on the meta-analytic method chosen. In addition, adverse effects are secondary endpoints, and are not reported consistently, leading to heterogeneity and gaps in the data available for analysis. Moreover, clinical judgments about classifying adverse events and attributing them to the drug being evaluated often create an additional source of uncertainty in safety meta-analyses.