ABSTRACT
This chapter discusses Seamless Population Selection design and Unblinded Sample Size Reestimation for Oncology confirmatory trials. It describes the key statistical considerations and relevant methodologies. Many recent cancer drugs target one or more specific biological pathways. A patient is more likely to benefit from the drug if his tumors are mediated through aberration of the specific pathway(s). The main statistical considerations for a seamless population selection design are good decision criteria for population selection after the learning stage and controlling the overall type-I error rates. Combination test along with a closed testing procedure restricts an initial set of hypotheses after each interim analysis. An alternative approach is the conditional error function in conjunction with the closed test principle. The conditional invariance principle allows for flexible interim decisions while controlling type-I error rates. In general, bias-corrected methods are not routinely used in clinical trials and the uncorrected point estimates and confidence intervals are usually reported.
