Meta-analysis using individual participant data from randomised trials: opportunities and limitations created by access to raw data
Meta-analysis based on individual participant data (IPD), often described as the ‘gold standard’ for effectiveness evidence synthesis, is increasingly being deployed despite being more resource intensive than collating study-level results. Its professed virtues include the ability to incorporate unreported data and to standardise variables and their definitions across trials. In reality, the unreported data, although present in shared datasets, might still not be usable in the analysis. The characteristics of trial participants and their outcomes may be too diversely captured for harmonisation and too time and resource consuming to standardise. Embarking on an IPD meta-analysis can lead to unanticipated challenges which ought to be handled with pragmatism. The aim of this article is to discuss the opportunities created by access to IPD and the practical limitations placed on such meta-analyses, using an international IPD meta-analysis of trials on the effect of lifestyle interventions in pregnancy as an example. Despite the increasing uptake of IPD meta-analysis, they encounter old problems shared by other research methods. When embarking on IPD meta-analysis, it is essential to evaluate the trade-offs between the ambitions, and what is achievable due to constraints imposed by the condition of collected IPD. Furthermore, incorporation of aggregate data from trials where IPD was not available should be a mandatory sensitivity analysis that makes the evidence synthesis up-to-date.