A Group Sequential Procedure for Continuous Longitudinal Data in Multi-Armed Trials

For ethical, administrative and economical considerations, group sequential methods are essentially used in clinical trials to compare treatments in a series of interim analyses. A group sequential method based on linear mixed effects models is developed to test the rates of change among g(ges 3) treatment arms for continuous longitudinal data, which is a generalization of the test proposed by Lee and DeMets (1991). The asymptotic joint distribution of the sequential proposed statistics is derived, and the behavior of the joint distribution of the proposed statistics is discussed by utilizing the critical values provided by Jennison and Turnbull (1991). The testing procedure is also illustrated by an example.