Forward Longitudinal Data Analysis with Application in Clinical Trials

• Activity Number: 253
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Haobo Ren*+ and Qian Ren and Richard Wu and Yuhwen Soo
• Companies: Regeneron Pharmaceuticals, Inc. and University of Minnesota and Regeneron Pharmaceuticals, Inc. and Regeneron Pharmaceuticals, Inc.
• Keywords: Longitudinal ; MMRM ; Forward LDA

Abstract

Most clinical trials are essentially longitudinal as measurements are taken at multiple sequential time points. When the statistical inferences for the endpoint at some intermediate time point (milestone) at the end of the study, we have choices to use the entire dataset or the data collected prior to the milestone. We call the latter approach Forward Longitudinal Data Analysis and this approach is more clinically meaningful. We focused on the linear mixed model for repeated measures. The distribution for the outcomes belonging to the same individual is specified as the multivariate normal distribution with general covariance matrix. Incomplete longitudinal clinical trial data can then be modeled without any extra handling. For the mean structure, we assume that each group has its own time trend which is characterized by the means at each visit. Under these assumptions, we illustrate that the inference for the treatment effect is the same when using the partial data set compared to using the whole data set. We also discuss the case when monotonic missing pattern is presented. Various additional issues are discussed as well. We illustrate our argument with simulated and real data.