Background: Meta-analysis using individual patient data (IPD) is now considered the gold standard for systematic reviews. Up until recently most IPD meta-analyses have been conducted in areas where mortality or another time-to-event outcome has been the only or primary outcome collected and few IPD meta-analyses have been conducted in areas such as surgery. The Hernia Trialists Collaboration, an EU Concerted Action, is currently collecting individual patient data for hernia repair trials. The Collaboration previously conducted an overview based on published data which found that formal statistical meta-analysis was usually not possible using this published information alone.

Methods: Reports of relevant trials up to April 2000 were identified using electronic database searching and through the Collaboration. International project meetings were arranged and the collaborative group agreed in principle to contribute individual patient data from their trials to allow a central re-analysis of source data for inclusion in a meta-analysis.

Results: 53 eligible trials were identified: 45 were reported as full papers, seven as abstracts only and one was unpublished. The trials included a total of 11,298 randomised patients. 37 trials compared laparoscopic with open groin hernia repair and 16 trials compared open mesh with non-mesh repair. 26 trial datasets have now been collected with a total of 5110 randomised patients. A further six datasets have been promised and the remainder are under negotiation. 15 datasets were received on floppy disk, nine by e-mail, one as paper format and one was centrally located. Although the majority of datasets were re-coded into the standard format for analysis by the Secretariat three trials were re-coded locally by Collaborators.

Conclusion: The main barriers identified in obtaining trial data included misinterpretation or unfamiliarity with the concepts of IPD meta-analysis, difficulties reassuring trialists who have not fully published their findings about the Collaboration's publication policy, and difficulties in tracing trialists and locating the data. Data may be received in many different computer formats and there may be many subsequent queries on each dataset. Clarification of data requires careful management and perseverance. There is also frequent need to translate foreign language publications and text data. In conclusion, the collection of IPD is possible for complex minimum datasets but requires sufficient resources as well as support from a multi-disciplinary team. Despite maximum effort it may not be possible to collect IPD from all trials. However, the benefits of IPD, including the ability to conduct formal meta-analysis where this may not have been previously possible, may outweigh these difficulties.