Bias in Oncology Time-to-Event Data Analysis

• Activity Number: 583
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Din Chen*+ and Yuhlong Lio and Yibin Wang
• Companies: Georgia Southern University and The University of South Dakota and Novartis
• Keywords: clinical trial ; time-to-event ; interval-censored ; MLE

Abstract

In all oncology clinical trials, the time-to-event data is generated from cancer scan with a specific interval (such as about 8 weeks) resulting interval-censored data. However the common practice to analyze the data from this type of trial is to approximate the interval-censored data using the midpoint or left of the interval-censored observation to get an exact failure time data rather than interval-censored data so that the well-known statistical methods developed for right-censored data could be used for the suggested analyses. Bias then could be introduced. In this talk we will conduct a simulation study to investigate the bias in parameter estimation and inference. The appropriate statistical models for interval-censored data are reviewed and suggested.