Two-stage individual participant data meta-analysis and generalized forest plots

Abstract

In this article, I describe a command, ipdmetan, that facilitates two-stage individual participant data meta-analysis of any measure of effect and its standard error by fitting a specified model to data from each study. The command can estimate random effects and heterogeneity statistics and include additional covariates and interactions. If individual participant data are available for certain studies and aggregate data for others, ipdmetan allows them to be combined in one analysis. This command can produce detailed and flexible forest plots, including ones outside the context of formal meta-analysis.