Evaluation of Common Statistical Methods in Randomized Delayed-Start Design of Progression Disease Clinical Trials

• Activity Number: 129
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Thomas Kelleher*+ and Bongin Yoo
• Companies: Bristol-Myers Squibb
• Keywords: delayed-start design ; linear mixed effects model ; type I error ; power ; missing data

Abstract

Randomized delayed-start design have been used in clinical trials of Alzheimer's Disease or Parkinson's Disease in order to evaluate whether treatment is modifying underlying disease courses rather than simply relieving short term symptoms. In the design, subjects are randomly assigned to either active or placebo and followed a period of time (Phase I) and then all the subjects in the placebo are switched to active for a period of time (Phase II). A simulation study is conducted to evaluate two most widely used linear mixed effects models, mixed-effects model repeated measures (MMRM) and random-coefficient mixed effets model (RCM) for the analysis of simulated data from the design. Our simulations suggest that MMRM provides well preserved type I error rate for Phase II as well as reasonable power estimates for Phase I regardless of missing patterns and provide less bias.