Design considerations for phase I/II dose finding clinical trials in Immuno-oncology and cell therapy.

Immuno-oncology (IO) and cell therapy, the frontier of cancer treatment, is a rapidly developing area that brings new opportunities to patients. In IO and cell therapy clinical trial development, it is critical to identify the right dose level in early phase of trials thus improving the probability of success in confirmatory trials to test the superiority over other therapies. Given the complex mechanism interacting with immune system for IO drugs especially cell therapy, the traditional oncology dose finding trial designs may not serve the purpose. Specifically, it is questionable to believe the monotone relationship between dose level and safety/efficacy, which will likely result in inappropriate dose selection using designs with monotone assumption. Additionally, considering the immune system pathway, designs ignoring the heterogeneity of the patient populations may provide misleading dose decisions, which could be either unsafe or lead to selection mistakes for targeted population. Therefore, in our paper, we review and present the limitations of the traditional dose finding designs. Then we discuss improved dose finding designs that consider both safety and efficacy outcomes simultaneously. Furthermore, we propose novel dose finding designs for multiple populations: BNP-mTPI and fBNP-mTPI, which extend the modified toxicity probability interval designs by utilizing Bayesian non-parametric priors. The proposed designs consider patient population differences meanwhile flexibly borrowing information across populations with similar profiles to improve the efficiency of dose search and accuracy of estimation of optimal dose level. Simulations are provided to demonstrate the model performance. Finally, we conclude the recommendations for IO and cell therapy dose finding designs in the discussion and offer insights for future research direction.