Bergrun T. Magnusdottir: Optimal designs for finding the single best dose of a drug

Finding a suitable dose is among the most difficult tasks during clinical development of a new drug. In early phases dose finding studies usually focus on finding a safe dose. Safety variables are thus of main interest. In later phases the focus is shifted towards efficacy. Typically a primary efficacy variable is defined and modeled. Various dose response models have been suggested. For continuous responses the most successful one is the Sigmoid Emax model. Here both efficacy and safety are considered simultaneously and the Emax model is extended to a model with a bivariate response, one response being a primary efficacy variable and one being a primary safety variable. The focus is on locally c-optimal designs for this nonlinear, bivariate response model. More specifically the locally c-optimal designs minimize the asymptotic variance for the estimate of the dose that maximizes patient’s utility. The utility is a function of the expected efficacy and the expected safety and referred to as the Clinical Utility Index (CUI).