Using multiple drug exposure levels to optimize power in pharmacogenetic trials

Large-scale pharmacogenetic trials testing tens to hundreds of thousands of single-nucleotide polymorphisms (SNPs) will become possible in the near future given rapidly decreasing costs of genotyping. Devising optimal designs for these trials will be a significant challenge. The author demonstrates how the level of drug exposure may strongly affect the power to detect true associations between genetic polymorphisms and drug response. An analytic model of drug response is described that is used to simulate pharmacogenetic trials. Analytical and numerical sensitivity analyses are performed on the model to demonstrate possible exposure-sensitivity behaviors. This model shows that the power to detect an association can be a nonlinear, nonmonotonic function of drug exposure. This model is further investigated using two clinical trial designs. The conclusion is that trial designs that use more than one drug exposure or dose will have an increased likelihood of discovering statistically significant pharmacogenetic associations.