Keywords: Utility, efficacy-toxicity tradeoff, Phase I/II, optimal biological dose, immunotherapy, targeted agent
The advent of novel targeted agent and immunotherapy has challenged the traditional paradigm for early phase clinical trial that focuses on the maximum tolerated dose (MTD). For these novel agents, efficacy does not necessarily monotonically increase with the dose. Thus, to achieve optimal therapeutic effect, targeted and immunotherapy agents are not necessarily administrated at the MTD. We propose a utility-based Bayesian seamless phase I/II design (USPD) to identify the optimal biological dose (OBD), defined as the dose with the highest desirability in terms of the efficacy-toxicity tradeoff. The USPD is consisted of two stages. In the first stage, we use the Bayesian optimal interval design to explore and identify the set of safe doses; and in the second stage, we adaptively randomize patients to these doses based on their desirability, quantified by a utility function that accounts for the efficacy-toxicity trade-off. To accommodate the common case that efficacy may take a relatively long time to be evaluated, we propose a novel approach that uses observed biological activity to predict the unobserved clinical efficacy outcome to facilitate the real-time decision making. Simulation shows that the USPD is capable of safeguarding patients from toxic doses and has a higher probability to identify the OBD. We develop a user-friendly Shiny App to facilitate the application of the proposed design.