Initial Guesses of Toxicity Probabilities

This chapter discusses the calibration of the initial guesses of toxicity probabilities associated with the test doses, or the skeleton. Section 8.2 presents an approach by which the initial guesses are specified via a single design parameter, namely, the halfwidth δ of the indifference intervals of the CRM model. Section 8.3 describes how δ can be chosen for a CRM application, and gives some numerical recommendations for the logistic model. Section 8.4 applies the techniques to redesign the bortezomib trial as a case study.