A continual reassessment method without undue risk of toxicity

We revisit our earlier extension of the CRM and introduce a new version of the dose escalation procedure. Its performance is assessed by simulation, with respect to a template of five operating measures BEARS (Benchmark, Efficiency, Accuracy, Reliability, Safety) that was used in Zhang, Wang and Yang. Our goal is to correctly identify the maximum tolerated dose with a high probability while protecting trial subjects from undue risk of toxicity. Simulation results show that this may be achieved for some considered scenarios, selective models, and considered competing designs. We discuss the pros and cons of the new version of the dose escalation procedure in the face of treatment versus experimentation dilemma.