Alternative methods for dealing with within trial clustering have been proposed for conducting individual participant data (IPD) ‘one-stage’ meta-analysis. Does it matter which one is used? Will it affect the estimates and their precision? 

Tom Fanshawe and Rafael Perera

Meta-analysis of studies using IPD is increasingly common and is likely to grow in popularity as a result of increased data sharing with repositories for collaborative data sharing become more established. IPD meta-analysis allows systematic reviews to pool data, beyond what is possible based only on the summary statistics, for example, intervention effects, reported in published papers and also provides the opportunity of fitting appropriate statistical methods to obtain summary estimates.

Two-stage and one-stage meta-analyses are the commonly used statistical approaches in IPD. In the ‘two-stage’ meta-analysis, the …