Robert A. Beckman - Comment

Chen, Clark and I have developed methods which may be apply to the Guidance. Mention of new methods in Guidances increases the likelihood of appropriate application, facilitating therapeutics development. I attach a detailed comment letter and two references. 1.Decision analysis-guided Phase II-III predictive biomarker transition(1). Integration into Guidance: Lines 519-536 discuss using trends in an earlier study to decide on enrichment of the current study when the selection is empirically based, and lines 964-971 discuss the same problem for a molecular biomarker. This method provides quantitative guidance for such decisions, and could be referenced in either or both locations. 2.Adaptive, predictive performance-based hypothesis prioritization in Phase 3 (1, 2). This is a data-driven method for optimizing the alpha split between full population and predictive subset hypotheses when the truth of the latter is still uncertain. It may be done for Phase 3 power optimization using mature Phase 2 data according to prespecified algorithms by an independent committee prior to Phase 3 unblinding, without affecting alpha, or altering patient selection or management. We propose that prospective, adaptive alpha splitting based on external data be considered as a possible statistically valid approach in the current Guidance Document. Integration into guidance: Lines 1064-1067 discuss alpha splitting in studies with both biomarker positive and negative patients, and lines 1127-1138 discuss Type I error control, including alpha splitting. Data-driven, optimized alpha-splitting could be referenced in either or both of those locations. 1.Beckman RA., Clark J, Chen C. Integrating Predictive Biomarkers and Classifiers into Oncology Clinical Development Programmes. Nature Rev. Drug Disc., 10: 735-49(2011). 2.Chen C, Beckman RA. Hypothesis Testing in a Confirmatory Phase III Trial with a Possible Subset Effect, Statistics in Biopharm. Res. 1:431-40(2009).