Objective - To assess the comparative efficacy and safety of pioglitazone and rosiglitazone. Research design and methods - Multiple electronic databases were searched for randomized, controlled trials (RCTs) of efficacy or effectiveness and for studies of any design which reported adverse events. Pooled estimates were calculated using a random effects model. Results - Eighty-seven RCTs fulfilled our inclusion criteria for efficacy or effectiveness and 42 studies examined safety or tolerability. Two head-to-head RCTs of type 2 diabetes demonstrated significant improvements in A1c in both groups at follow-up with no significant difference between groups; a third study found no significant change in A1c in either group. The pooled estimate of effect on A1c for pioglitazone compared to placebo was -0.99% (95% confidence interval [CI], -1.18, -0.81) and for rosiglitazone was -0.92% (95% CI, - 1.2, -0.64). Indirect comparison revealed no significant difference in A1c (between-drug difference -0.07% [95% CI, -0.41, 0.27]). Rosiglitazone increased total cholesterol compared to pioglitazone (net between-drug effect 13.91 mg/dl [95% CI, 1.20 to 26.62]). Both drugs increased weight by 2 to 3 kg and rates of adverse events were similar for the two drugs. Data were insufficient to assess comparative effects on health outcomes such as cardiovascular events. Conclusions - Based largely on indirect evidence, the two thiazolidinediones appear to have similar effects on glycemic control and similar side-effect profiles. Rosiglitazone may increase total cholesterol compared to pioglitazone. Studies are needed which provide direct comparisons between the two drugs, particularly for long-term health outcomes.