Model-Assisted Designs for Identifying the Optimal Biological Dose

Abstract

Compared to the model-assisted designs introduced in Chapter 4, model-based designs are complicated statistically and computationally, making them more challenging to implement in practice. This chapter introduces two model-assisted phase I/II designs, BOIN12 and U-BOIN, to find the optimal biological dose (OBD). These two designs simultaneously consider toxicity and efficacy, and use the utility to quantify the risk-benefit tradeoff. Based on the accrued toxicity and efficacy data, the designs adaptively assign patients according to the estimated utility. As model-assisted designs, BOIN12 and U-BOIN have the advantages of being simple to implement and meanwhile yielding competitive performances. Conducting the trial does not require complicated model estimation. The decision of dose transition can be easily made by looking up the pre-generated decision table. Examples and software are provided to illustrate BOIN12 and U-BOIN.