Main Organiser

Julius Centre University of Malaya

Co-organiser

Department of Social and Preventive Medicine, Faculty of Medicine, University of Malay

Supported by

University of Malaya

COMBINING DATA IN MULTI-COUNTRY STUDIES: META-ANALYSIS VERSUS MEGA- ANALYSIS OF INDIVIDUAL PATIENT DATA

Author

M T Bazelier

Institution

Utrecht University

Abstract

Objectives: To evaluate the difference between a traditional meta-analysis and an individual patient data mega-analysis in multi-country studies.

Methods: In the UK General Practice Research Database (GPRD), there was an almost 3-fold increased risk of hip fracture for incident patients with multiple sclerosis (MS) versus population- based controls (adjusted hazard ratio (adj HR) 2.8 [95% CI 1.8-4.3]). However, in incident MS patients from Denmark, hip fracture risk was not significantly increased: adj HR 1.8 [0.9-3.5]. We investigated whether an inverse-variance fixed effect meta-analysis of the two studies provided different results than an individual patient data mega-analysis with comparable characteristics.

Results: A meta-analysis of the two original studies yielded an adj HR of 2.5 [1.7-3.5]. When we made the two study populations comparable in terms of calendar time, definition of the index date, proportion of incident / prevalent patients and duration of follow-up, we found an adj HR of hip fracture with MS of 2.4 [95% CI 1.4-4.2] in GPRD and adj HR of 3.4 [2.7-4.3] in Denmark. An individual patient data mega-analysis of these study populations resulted in adj HR 2.9 [2.4-3.6].

Conclusion: Individual choices in study design may explain substantial amounts of heterogeneity when data from multi-country studies are combined.