Comparing Methods of Estimating Treatment Effects on a Continuous Outcome in Multicenter Randomized Controlled Trials: A Simulation Study

• Activity Number: 44
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Rong Chu*+ and Lehana Thabane and Jinhui Ma and Eleanor Pullenayegum and Anne Holbrook
• Companies: McMaster University and McMaster University and McMaster University and McMaster University and McMaster University
• Keywords: multicentre trials ; clustering ; intraclass correlation ; stratification ; fixed effects ; random effects

Abstract

Multicentre clinical trials routinely use randomization and analysis stratified by centre to control for differences between centres and to improve precision. This approach overlooks the potential for clustering within centres. Assuming no treatment by centre interaction, we compared six methods (ignoring centres, including centres as fixed effects, including centres as random effects, GEE, and fixed- or random-effects centre-level analysis) to analyze continuous outcomes in multicentre trials using simulation technique over a wide spectrum of intra-class correlation (ICC) and centre size values. While all methods yield unbiased treatment effects estimates, ignoring centres leads to loss of power. The random effects model is more efficient and has nominal coverage. Fixed effect model is less precise when the number of centres is large and allocation is subject to chance imbalance.