Keywords: dual-agent, binary efficacy and toxicity outcomes, Bayesian CRM, posterior distribution, dose-finding, phase 1
In Phase I oncology trials with targeted and/or immunotherapy dual agents, finding best dose combination/s (BDC) is/are generally required than finding maximum tolerated dose combination (MTDC). Identifying an acceptable level of toxicity in conjunction with early clinical anti-tumor activity may be helpful for such purpose. The dual-agent Bayesian continual reassessment method (2D Bayesian CRM) for binary toxicity outcomes is well established and a similar approach may be useful for consideration of preliminary binary anti-tumor activity outcomes. In assumed dual-agent dose-toxicity model, escalation with overdose principle (EWOC) may be utilized to approximate the potential uncertainty in the corresponding posterior distribution of the toxicity rate. In assumed dual-agent dose-efficacy model, an indifference interval may be utilized to approximate the potential variability in mean posterior anti-tumor activity rate. Given small sample sizes and limited available data in early dose escalation steps in clinical trials, incorporating correlation between toxicity and anti-tumor activity outcomes in the same model may not be realistic. Nonetheless, if two models are run independently, devising some measure to combine the corresponding posterior mean quantities may be achievable. Simulations under variety of scenarios suggest the proposed algorithm has reasonable operating characteristics compared to 2D Bayesian CRM with only binary toxicity outcomes.