This is an archived version of the Handbook. For the current version, please go to training.cochrane.org/handbook/current or search for this chapter here.

IPD cannot be analysed directly in RevMan. The data need to be first analysed outside of this software, and summary statistics for each study may be entered into RevMan if a two-stage approach is used. For dichotomous and continuous outcomes, data may be entered in the usual way. For time-to-event outcomes, the observed-minus-expected number of events and variance may be entered using the ‘O – E and Variance’ option. Alternatively the generic inverse-variance option may be used to analyse effect estimates such as hazard ratios, rate ratios or adjusted estimates.

Although many standard statistical packages can perform the necessary analyses of IPD from the individual studies, it can be unwieldy and time-consuming to have to analyse each outcome in each study one at a time, and commercially available software is not currently available that supports the direct analysis, pooling and plotting of IPD in a meta-analysis. A non-commercial analysis package, ‘SCHARP’, which analyses each study, pools results and outputs tabulated results and forest plots for dichotomous, continuous and time-to-event IPD, is available free of charge to not-for-profit organizations. This SAS-based package has been developed by the Meta-analysis Group of the UK Medical Research Council Clinical Trials Unit. It is available from the authors, who can be contacted through the IPD Meta-analysis Methods Group (see Box 18.6.a).