Meta-Analysis

Abstract

Problems with meta-analyses are frequent: regressions are often nonlinear; effects are often multifactorial rather than unifactorial; continuous data frequently have to be transformed into binary data for the purpose of comparability; bad studies may be included; coverage may be limited; data may not be homogeneous; failure to relate data to hypotheses may obscure discrepancies. In spite of these well-recognized flaws, the method of meta-analysis is an invaluable scientific activity: Meta-analyses establish whether scientific findings are consistent and can be generalized across populations and treatment variations, or whether findings vary significantly between particular subsets. Explicit methods used limit bias and improve reliability and accuracy of conclusions, and increase the power and precision of estimates of treatment effects and risk exposures. In the past decade, despite reservations on the part of regulatory bodies, the method of meta-analysis has increasingly been employed in drug development programs for the purpose of exploration of changes in treatment effect over time, integrated summaries of safety and efficacy of new treatments, integrating existing information, providing data for rational decision making, and even prospective planning in drug development.