Statistical Properties of a Modified Accelerated Design for Phase I Cancer Clinical Trials

Abstract

In recent years, the accelerated design has been proposed as an improvement to the standard algorithm-based design (Simon et al., 1997; Storer, 1989). There may be situations under the proposed accelerated designs by Storer (1989) and Simon et al. (1997) in which the MTD may be declared with undesirable features. We propose a modified accelerated design in this article. Lin and Shih (2001) studied several key statistical properties for the standard algorithm-based A + B design based on the exact mathematical derivation. In this article, we derive similar key statistical properties for the accelerated design. Under certain design settings, the modified accelerated design requires fewer expected number of patients, and hence less time (sometimes significantly less time) to complete the study; however, it may incur a higher overall expected toxicity rate.