Best practices when conducting and reporting a meta-analysis

Abstract

Meta-analysis, a statistical technique that combines the results of different studies to obtain a pooled or summary estimate of data, has become an integral part of evidence-based decision-making processes. This tool is being increasingly used in medical and nonmedical disciplines as it is able to quantitatively synthesize a large and often complex or sometimes conflicting body of literature. However, when performing a meta-analysis, one should be aware of the factors that may impact on the results and which statistical method and model should be used for each scenario. Usually, the results of a pairwise meta-analysis are depicted in a forest plot and reported as the summary effect size (i.e., point estimate) combined with its confidence interval (i.e., lower and upper limits), a P value, and a between-study heterogeneity value. Additionally, there is increasing value in routinely reporting further metrics (e.g., prediction interval, number needed to treat, fragility index) which allow more informative inferences about the data. This chapter briefly reviews the methods to perform and interpret a meta-analysis, including concepts of effect-size measures, heterogeneity, and subgroup and sensitivity analyses. The benefits and drawbacks of a meta-analysis and the international recommendations and standards for conducting and reporting are also discussed. (c) 2022 Elsevier Inc. All rights reserved.