Meta-Analysis with Binary Outcomes

Abstract

This chapter describes how to perform meta-analysis with binary data using R. We introduce the usual effect measures for binary outcomes and discuss issues raised by sparse binary data. We describe how to perform meta-analysis using the inverse variance method [17] and the DerSimonian–Laird method [12]. Furthermore, we introduce the Mantel–Haenszel method [24] and the Peto method [36] which are specific to binary outcomes. Several examples use base R commands. We also describe the metabin function from R package meta [31, 32] which provides a unified syntax for all methods in this chapter.