Statistical and Clinical Issues Regarding Choice of the Primary Analysis in Longitudinal Clinical Trials

• Activity Number: 157
• Type: Invited
• Date/Time: Monday, August 4, 2003 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Craig H. Mallinckrodt*+ and Geert Molenberghs and Raymond J. Carroll
• Companies: Eli Lilly & Company and Limburgs University Centrum and Texas A&M University
• Keywords: missing data ; longitudinal data ; maximum likelihood ; mixed-effects models

Abstract

The last observation carried forward (LOCF) approach has for decades been a common method of handling missing data in clinical trials. Considerable advances in statistical methodology and in our ability to implement those methods have been made in recent years. Therefore, we examined from a statistical and clinical perspective the theoretical, empirical, and practical attributes of various methods for handling missing data as compared with the desired attributes of a primary analysis for clinical trials conducted in a regulatory environment. We concluded that likelihood-based, mixed-effects model approaches implemented under the missing at random framework are consistent with the need for a simple, easy to implement primary analyses. These methods are robust to the biases from missing data and provide better control of Type I and Type II errors than LOCF. Conservative behavior of LOCF is not guaranteed and anticonservative behavior is likely in some common scenarios. Therefore, we advocate a shift away from use of LOCF as the primary analysis to likelihood-based, mixed-effects model approaches developed under the missing at random framework.