Interim Monitoring of Longitudinal Outcomes in Clinical Trials

• Activity Number: 631
• Type: Topic Contributed
• Date/Time: Thursday, August 8, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Michael McDermott*+ and Xueya Cai
• Companies: Univ of Rochester Medical Center and Univ of Rochester Medical Center
• Keywords: Interim analyses ; Conditional power ; Simulation

Abstract

Many clinical trials in neurological disease collect data on longitudinal outcomes over a relatively long follow-up period. Often the primary outcome variable is measured at the final time point or visit, with additional data collected at earlier time points. When the duration of follow-up is much longer than the accrual time, there can be substantial accumulation of longitudinal data prior to the primary outcome variable being measured on any subjects at the final time point and it may be of interest to evaluate efficacy or futility on the basis of these interim data. We propose a novel strategy for interim monitoring of the primary outcome variable in this setting. The main idea is to simulate the unobserved data from each subject based on the conditional distribution of the unobserved data given the observed data. Quantities useful for monitoring such as conditional power and predicted intervals can be estimated using this simulation strategy. We evaluate the proposed methods in terms of Type I error control, efficiency, and sensitivity to misspecification of necessary quantities such as the covariance matrix for the longitudinal measurements and the dropout process.