A New Adaptive Simon-Based Design Focusing on Subpopulation Heterogeneity

Abstract

IntroductionIn phase II clinical trials in oncology, determining inclusion criteria of a potentially sensitive population is a critical issue in drug development. Firstly, investigators must define the patient-related criteria: performance status, an independent risk factor for increased toxicity and reduced efficacy of treatment [1] and the age limit permitted, since it has been shown that some drugs induce severe side effects more frequently in elderly than in younger subjects [2]. The second consideration is the tumor-related criteria. It is common practice to include tumors with an identical primary site, such as breast or lung, and with identical histology. These criteria are chosen after examination of the pre-clinical data on drug efficacy [3]. However, this approach has been shown to have a low positive predictive value and a low negative predictive value of drug efficacy in humans [4].Despite the extensive patient-related and tumor-related data available, population selection for phase II trials remains challenging and these trials frequently fail to demonstrate efficacy. This can be due to the heterogeneity of two subpopulations included in the same phase II trial. The drug may show efficacy in only one subpopulation: for instance, overexpression of c-erbB2 (but not lack of expression) is predictive of a response to trastuzumab [5], the cytotoxic agent vinflunine has been shown to be active in patients with good performance status but not in those with poor performance status [6] and KRAS mutation is predictive of a response to cetuximab in colon cancer [7] but not in lung cancer [8].Studying efficacy in heterogeneous subpopulations may therefore prove extremely useful to limit the risk of failure. AbstractBackground: Phase III trials can fail, leading to termination of drug development. This can result from heterogeneous subpopulations such as a drug-sensitive and a drug-insensitive subpopulation of patients or biological subtypes.