Phase II clinical trials in oncology are used for initial evaluation of the therapeutic efficacy of a new treatment regimen. Simon's two-stage or Fleming designs are commonly used for such trials. Treatment effect dilution may be observed because of heterogeneity of the population leading to negative conclusion on the whole population and termination of drug development while patients with particular characteristics may have benefit. Several authors have proposed alternative strategies based on Simon's design, including stratification on the patients' characteristics. In this paper, the authors develop a new stratified phase II design, allowing early stop of the trial for efficacy as it is an extension of Fleming's design. An example based on real data is given and the results from exact binomial calculations are developed to explore several assumptions on response rates, prevalence of each population and errors. Copyright © 2011 John Wiley & Sons, Ltd.