Adaptive Designs for Multi-Cohort Early Phase Oncology Trials

• Activity Number: 30
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Christine Gause*+ and Keaven Anderson
• Companies: Merck and Merck
• Keywords: adaptive designs ; truncated seqeuential probability ratio test ; Bayesian hierarchical models ; Simon two-stage design

Abstract

In early phase oncology drug development, multiple cohorts are often considered simultaneously with the goal to select cohorts for further study as quickly as possible. The Simon two-stage design is frequently used for this purpose, whereby a decision can be made to stop a cohort early for futility after a small number of subjects are enrolled and evaluated for response rate. Flexible monitoring of accumulating data using Bayesian methodology based on the posterior probability that the true response exceeds a particular threshold has also been considered. In addition, designs based on Bayesian hierarchical models have been proposed that permit borrowing of information across cohorts for early decision-making. Alternately, a truncated sequential probability ratio test can be used for continuous monitoring of a binary endpoint where efficacy and futility boundaries are derived using group sequential methods. We investigate these adaptive designs in the context of a multi-cohort oncology trial with a response rate primary endpoint. A simulation study is conducted to compare the approaches with respect to probability of early stopping, power, and sample size.