|
Abstract:
|
Phase I cancer clinical trials are designed in two parts: Part I of the study is designed to identify a maximum tolerated Dose (MTD) or a range to be used in Part II (and beyond), which is designed to collect a set of efficacy data (in addition to safety data on the MTD). There are several challenges encountered in the design and conduct of dose escalation studies that are carried out using fixed or adaptive methods. The simplest and most widely used approach for determining the MTD classifies safety events into two categories, DLT and non-DLTs, with the aim of finding the dose with probability of DLT closest to targeted probability, usually 25 to 30 percent. In particular, model-based methods define the MTD as a model parameter to be estimated. Bayesian model-based approaches include the continual reassessment method (CRM), a framework similar to CRM with a set of decision rules called toxicity probability interval (TPI) approach and a dose escalation with overdose control (EWOC) method. In this paper, we will explore combining the existing approaches and compare their properties. Simulation and examples will be used as supportive details for the comparison.
|
