Bayesian Adaptive Designs for Efficient Targeted Agent Development
View Presentation View Presentation
*J. Jack Lee, University of Texas MD Anderson Cancer Center 


Advances in biomedicine have fueled the development of targeted agents in cancer therapy. Targeted agents, however, do not work for everyone and may not work at all. Hence, the development of target agents requires the evaluation of treatment efficacy as well as prognostic and predictive markers. In addition, upon the identification of each patient’s marker profile, it is desirable to treat patients with best available treatments in the clinical trial accordingly. We have developed Bayesian adaptive designs to (1) test the treatment efficacy, (2) identify prognostic and predictive markers, and (3) provide better treatments for patients enrolled in the trial. Via the outcome-adaptive randomization (AR), Bayesian AR designs can treat more patients with more effective treatments based on the available data at the time. With frequent interim monitoring, ineffective treatments can be stopped early for futility, new treatments can be added, and effective treatments can graduate for further evaluation. Through simulations, design parameters can be chosen to yield the desirable operating characteristics and to control the types I and II errors. Compared with the traditional designs, the proposed designs can be more efficient, more ethical, and also more flexible in the study conduct. Additional infrastructure must be set up to allow timely and frequent monitoring of interim results. Bayesian adaptive randomization designs are distinctively suitable for the development of multiple targeted agents in phase II settings. Examples and lessons learned from the recently completed BATTLE trial in non-small cell lung cancer will be given.