A systematic review of analytical methods used to study subgroups in (individual patient data) meta-analyses

Objectives: To determine whether individual patient data meta-analyses (IPDMA) are used to perform subgroup analyses and to study whether the analytical methods regarding subgroup analyses differ between IPDMA and conventional meta-analyses (CMA).

Study Design and Setting: IPDMA were identified with a comprehensive literature search, subsequently, CMA on similar research questions were traced. Methods for studying subgroups were compared for IPDMA and CMA that were matched with respect to domain, type of treatment, and outcome measure.

Results: Of all 171 identified IPDMA and 102 CMA, 80% and 45% presented subgroup analyses, respectively. For 35 IPDMA and 37 "matched" CMA, subgroup analytic methods could be compared. The number of performed subgroup analyses did not differ between IPDMA and CMA. Both IPDMA and CMA often do not report adequate information on methods of analyses. Interaction tests were often not performed in IPDMA (69%) and individual patient data was often not directly modelled (74%).

Conclusion: Many IPDMA performed subgroup analyses, but overall treatment effects were more emphasized than subgroup effects. To study subgroups, a wide variety of analytical methods was used in both IPDMA and CMA. In general, the use and reporting of appropriate methods for subgroup analyses should be promoted. Recommendations for improvement of methods of analyses are provided. (C) 2007 Elsevier Inc. All rights reserved.