Oncology drug developers often decide to initiate Phase III trials at risk after significant preliminary anti-tumor activities are observed in small Phase I/II trials. The preliminary data can hardly provide the much-needed information for selecting a biomarker cutpoint or prioritizing a biomarker hypothesis in Phase III. To address this issue, Magnusson and Turnbull (2013) proposed a group sequential enrichment design that de-selects non-performing biomarker subpopulations at an interim analysis and pools the remaining ones in final analysis. The same endpoint was used for interim and final analyses therein. In this presentation, we propose a more general approach in that different endpoints may be used (e.g., PFS for interim analysis and OS for final analysis) and sample size for remaining ones is subject to increase after the interim analysis. An interesting multiplicity issue will be discussed. The use of a sensitive intermediate endpoint for population de-selection increases the study power after multiplicity adjustment, which is further improved with sample size adjustment. Our proposed design paves the way for expedited development of personalized medicines in confirmatory trials with limited prior data.