Feasibility of establishing a central repository for the individual participant data from research studies

Tags: Poster
Altman D1, Clarke M2, Dwan K3, Riley R4, Williamson P3
1University of Oxford, UK, 2Queen’s University Belfast, UK, 3University of Liverpool, UK, 4Birmingham University, UK

Meta-analysis of individual participant data (IPD) is widely accepted as the most reliable approach for systematic reviews and has been described as the 'gold standard’. Advantages include standardising outcome definition across studies, increased potential to investigate subgroups, reducing bias by analysing on an intention to treat basis, minimising the possibility of within study selective reporting, thorough analyses of time to event outcomes, opportunities to identify unpublished studies through collaboration with the original researchers, and incorporating additional follow-up. IPD provides a rich source of information that allows clinical and methodological developments to extend beyond exploring the main effects that are traditionally of interest in a single trial or systematic review. However, the resource implications required for an IPD approach are often prohibitive for reviewers. It is therefore essential that as much use as possible is made of IPD that have been collected. We propose that a secure central repository be established to store previously collected IPD. Restricted access to the central repository would only be granted following an approval process that would involve the original reviewers and a nominated committee. The central repository would facilitate exploring additional clinical and methodological questions across a range of studies and reviews. To assess the feasibility of developing and managing a central repository, we have undertaken an on-line survey of 70 IPD reviewers registered with the Cochrane IPD Meta-analysis Methods Group. We asked about their willingness to provide anonymised IPD from their review and asked about practical issues that this may raise. Non-responders have been reminded about the survey up to three times. Analyses are ongoing in April 2011 and will be presented, along with future plans at the Colloquium.