ABSTRACT
This chapter introduces ordinal and continuous toxicity extensions for phase I clinical trials. It includes variations of algorithm-based designs, such as the 3 + 3 design, and model-based designs including the continual reassessment method (CRM) and escalation with overdose control (EWOC). The chapter presents ordinal extensions of these designs with an example of how to design a trial and re-estimate dose levels as patients accrue based on the proportional odds CRM. It discusses nonpara-metric methods that relax model assumptions in trial designs as well as designs incorporating continuous toxicities scores as an alternative to ordinal toxicity grading in these early-phase trials. The chapter introduces two commonly used models to incorporate ordinal safety outcomes, namely the proportional odds model and the probit model, and their applications in early-phase trials. It also conceptually introduces maximum likelihood estimation (MLE) and Bayesian estimation. The chapter provides a phase I dose-finding example using the proportional odds model CRM with MLE estimation.
