Robustness of the Sequential Efficient Design for Identifying a Target Subpopulation

Precision medicine is an innovative approach for tailoring treatments based on individual characteristics or biomarkers. Enrichment is a main strategy in the development of clinical trials for precision medicine as it “is the prospective use of any patient characteristic to select a study population in which detection of a drug effect (if one is, in fact, present) is more likely than it would be in an unselected population.” (FDA, 2019) [ 1]. Although a binary biomarker can be used to stratify the subjects, the situation is more complex if a continuous biomarker is associated with patient responses. Recently Baldi Antognini et al. [ 2] provided the optimal allocations for inference on the threshold of a continuous biomarker and proposed a new Covariate-Adaptive procedure, called the Sequential Efficient Design (SED), to implement these designs sequentially. In this paper we push forward the results by exploring the robustness of the SED under possible deviation from the linearity assumption of the response-biomarker relationship, taking also into account comparisons with the permuted block design.