One of the vexing problems often encountered when combining the results of independent studies in a meta-analysis is that the data provided in individual studies are incomplete. Some studies may provide only a risk difference and others only an odds ratio. Of course, if the proportions for treatment and control are reported, then the meta-analyst can carry out a variety of analyses, such as fixed or random effect estimates. Excluding studies with incomplete data carries a risk and should be avoided if it is possible to retrieve the original proportions. Inclusion criteria in some meta-analyses may require that studies contain full data. This requirement can be relaxed if the original data can be retrieved. Copyright © 2012 John Wiley & Sons, Ltd.