Correcting Bias Caused by Unequal Efficacy Assessment Intervals Between Treatment Groups

• Activity Number: 53 - New Developments in Survival Analysis
• Type: Contributed
• Date/Time: Sunday, August 8, 2021 : 3:30 PM to 5:20 PM
• Sponsor: Biometrics Section
• Author(s): Qian Li* and Daniel Li and Zhihong Yang and Yizhou Fei
• Companies: Bristol Myers Squibb and Bristol-Myers Squibb and Bristol Myers Squibb and Bristol Myers Squibb
• Keywords: Control of Type I error; Correcting Bias; Unequal Efficacy Assessment Intervals

Abstract

In randomized and controlled oncology clinical trials, we may encounter situations where the treatment cycles are not consistent between treatment groups. Taking an example of a two-arm randomized clinical trial, the treatment cycles are 28 and 21 days, in Arms A and B, respectively. Prior to the start of each treatment cycle, patients will need to be assessed for treatment efficacy to determine the subsequent treatment plan. The timing of such assessments will also be used for time-to-event analysis such as progression free survival. The difference in efficacy assessment interval may lead to bias against the treatment arm which has shorter intervals. In the above example when Arm A is a new drug to be tested and Arm B is a control, such difference will lead to type I error rate inflation. To correct bias and type I error inflation, we propose an algorithm to virtually realign the shorter treatment interval to the longer one. Using this algorithm, the bias and inflation of type I error can be effectively eliminated. Simulations are used to illustrate the control of bias and type I error rate and show the maintenance of power.