Effect Measures in Meta-Analysis

Abstract

When conducting a meta-analysis of randomised controlled trials outcomes, appropriate choice of the effect measure is important. This article demonstrates on various types of effect measures in meta-analysis, for example, binary, continuous and ordinal outcomes. A general fixed effects model and a random effects model are employed for combining these outcomes in meta-analysis. Six trials totaling 1876 patients from a meta-analysis of randomised controlled trials evaluating the efficacy and drawbacks of limited (D1) versus extended lymphadenectomy (D2) for proven gastric adenocarcinoma are analysed for binary and continuous outcomes. An individual patient data consisting of five randomised trials of anti-cholinesterase drug tacrine in patients with Alzheimer's disease is also discussed for ordinal outcomes.