Obtaining IPD often enables inclusion of studies that could not be included in a standard systematic review because they are either unpublished or do not report sufficient information to allow them to be included in the analyses. This may help avoid many types of publication bias (Stewart 2002). However, one must ensure that by restricting analyses to those studies that can supply IPD, bias is not introduced through selective availability of study data.
The success and validity of the IPD approach requires that data from all or nearly all studies will be available. If unavailability is related to the study results, for example if investigators are keen to supply data from studies with promising results but reluctant to provide data from those that were less encouraging, then ignoring the unavailable studies could bias the results of the IPD review. If a large proportion of the data have been obtained, perhaps 90% or more of individuals randomized, we can be relatively confident of the results. However, with less information we need to be suitably circumspect in drawing conclusions. Sensitivity analysis combining the results of any unavailable studies (as extracted from publications or obtained in tabular form) and comparing these with the main IPD results are a useful aid to interpreting the data. Reports of IPD reviews that were unable to obtain IPD from all studies should state reasons why IPD were not available, and the likelihood of ensuing bias.
As for other types of Cochrane review, IPD meta-analyses should clearly state what studies were not included and the reasons why. If only a limited number of studies are able to provide IPD for analysis, then the value of the approach is questionable. Experiences in cancer have been good and in most cases perseverance has led to data being available from a high proportion of eligible trials. This can make it especially important to explore the ability and willingness of the primary investigators to supply IPD at an early stage in the project.