Interval-Censored Time-to-Event Data in Clinical Trials: Analyzing and Designing

• Activity Number: 72
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Xing Sun*+ and Cong Chen
• Companies: Merck & Co., Inc. and Merck & Co., Inc.
• Keywords: clinical trial ; survival analysis ; interval-censored ; group sequential design

Abstract

Interval-censored time-to-event data occur frequently in randomized clinical trials. These data often pose challenges for both trial design and statistical analysis. Standard methods, generally accepted and practiced by pharmaceutical industry, are often simple imputation based methods. In this study, we demostrate by simulation that under typical clinical trial settings, these simple methods are less optimal than other more sophisticated nonparametric likelihood based methods proposed two decades ago. Also, unlike for right-censored survival data, expected sample size/power depends heavily on baseline survival distribution and interval frequency for interval-censored data. A discussion on how to design a trial with interval censored data to accommodate this property, especially for group sequential design, will also be included.