All Times EDT
Keywords: Bayesian adaptive design, implementing Bayesian trials, efficacy and toxicity tradeoff, late-onset
To expedite drug development, the FDA has issued the guidance “Adaptive Design Clinical Trials and Drugs and Biologics” and launched the Complex Innovative Trial Designs Pilot Program, which recommends the use of novel Bayesian adaptive designs. While in theory these designs are greatly superior to conventional methods, implementing them in practice is challenging. We will illustrate what we have learned in the course of applying a novel Bayesian adaptive phase I-II design in an actual trial. The aim is to optimize the dose of radiation therapy (RT), given either with placebo (P) or an investigational agent (A) for treating advanced pancreatic cancer. The trial is challenging because (1) patients are randomized between the RT+P and RT+A arms, (2) efficacy and toxicity may be late-onset, and (3) within each arm, RT doses are chosen adaptively in real time to optimize the risk-benefit tradeoff between efficacy and toxicity. To address these challenges, we use the Late-Onset Efficacy-Toxicity (LO-ET) design. Implementing the design in this trial has involved double-blind balance-restricted randomization, real-time assessment of patient outcomes to evaluate the efficacy-toxicity tradeoff for each RT dose in each arm in order to optimize each new patient’s RT dose, and transition from a single center trial to a multicenter trial. Conducting the trial has required ongoing interactions between clinic nurses, physicians, the pharmacy, statisticians, data managers, and the sponsor. By March 2020, 36 patients have been randomized. Details, lessons learned, illustrative documentation, and anecdotes will be presented.