Random Effects Linear Models in Cross-Over Trials

• Activity Number: 470
• Type: Topic Contributed
• Date/Time: Wednesday, August 7, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Michael Kenward*+
• Companies: University of London
• Keywords: Cross-over trials ; Personalized medicine ; Random effects ; Missing data ; Linear mixed model ; Dropouts

Abstract

Cross-over trials involve the collection of repeated measurements from individual subjects, and as such, represent an obvious setting in which random effects models have a potentially important role. Indeed, many proposed analyses for such data are based on the linear mixed model. However several issues do arise when using such models in this context, and not all are well known. In this talk an overview is given that touches on the most important of these. The talk will cover the role of such models in the weighted combination of between- and within-individual information, and the implications of their use when data are missing, for example with dropouts, and when baseline measurements from each treatment period are incorporated in the analysis. By their nature, cross-over trials are often small compared with conventional parallel group trials, and so particular attention will be given small sample issues both in terms of estimation and subsequent inference.