Issues and Approaches for the Analysis of Longitudinal Data in Confirmatory Clinical Trials
View Presentation View Presentation
*Craig H Mallinckrodt, Eli Lilly 

Keywords: longitudinal data, mixed-effects models, clinical trials

Treatment effects are often evaluated by comparing change in outcomes over time. However, valid analyses of longitudinal data can be problematic. In choosing the primary hypothesis for confirmatory clinical trials, debate exists as to whether effectiveness or efficacy is most relevant. The pros and cons of each hypothesis will be discussed; leading to the conclusion that efficacy is most relevant. Given this hypothesis, the characteristics of the data can be used to quide development an appropriate analysis. These data characteristics and regulatory considerations will be reviewed. Theory and empirical evidence support the utility of a likelihood-based mixed-effects model repeated measures (MMRM) approach. Key aspects of the MMRM analysis will be contrasted with other formulations of mixed effects models.