Experimental Design for Combination Studies

• Activity Number: 305
• Type: Contributed
• Date/Time: Wednesday, August 11, 2004 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Ming T. Tan*+ and Hongbin Fang and Guoliang Tian
• Companies: University of Maryland and Medicine and University of Maryland and Medicine and University of Maryland and Medicine

Abstract

Drug combination is central to cancer therapy. A statistical experimental design is necessary since even the administration of precisely the same dose to virtually identical animals may result in different responses. Such data variation needs to be controlled in the experimental design and accounted for in the analysis to allow proper inference on the efficacy of combinations. However, methods currently available are derived under some restrictive assumptions. For example, Abdelbasit and Plackett (1982) proposed an optimal experimental design assuming that the dose-response relationship follows a specified linear model. Tallarida et al. (1997) derived a fixed-ratio design and used t-test to detect the simple similar action. However, in reality, we usually do not have enough information on what kind of synergistical effect on the dose-response relationship is before experiment. We first propose a novel (nonparametric) models that does not impose such strong assumption on the possible joint action. We then propose a experimental design for the joint action with nonparametric models based on the uniform measure. This design is optimal in the sense that it reduces the variability in modeling synergy while allocates the mixtures reasonably to extract maximum information on the joint action of the compounds. Based on this experimental design, a robust F-test is proposed to detect the simple similar action of two compounds. The method is illustrated with the study of the combination of two anti-cancer agents: temozolomide and irinotecan.