A Bayesian Repeated Measures Design with Temporal Trend for Delayed Toxicity Outcomes

• Activity Number: 701
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Rui Qin* and Jun Yin and Sumithra J. Mandrekar and Daniel Sargent
• Companies: Mayo Clinic and Mayo Clinic and Mayo Clinic and Mayo Clinic
• Keywords: clinical trial ; phase I ; longitudinal model ; Bayesian design

Abstract

Dose-limiting toxicity (DLT), typically assessed during the first treatment cycle, is the primary endpoint for most Phase I clinical trials in oncology. This is based on the assumption that DLT in the first cycle generally represents the overall toxicity risk. Delayed toxicity outcomes (frequent with targeted therapy or radiation) pose many statistical design challenges. Delaying new patient enrollment to account for delayed toxicity outcomes to be observed is problematic. We propose an adaptive dose-finding design using a repeated measures model for longitudinal data based on a quasi-continuous toxicity score. This model accommodates both the dose-toxicity curve and the temporal trend for cumulative toxicity from subsequent cycles, and at the same time, captures the overall severity of multiple toxicities. The design is implemented in the Bayesian framework, including the statistical models, computation, loss functions and decision making. Simulation results from various clinically relevant scenarios will be presented to understand the operating characteristics of the proposed design.