Combination therapies are becoming increasingly common and are standard practice in anesthesia and oncology. This talk focuses on studies most relevant to anesthesia where it is of interest to describe the pharmacodynamic effect of two drugs given simultaneously. This effect is generally described via a response surface where parameters exist nonlinearly in the model and typically vary between individuals. The current approach to design these studies is to implement a criss-cross design. However, given the limited information available prior to the study, such designs have been shown to perform quite poorly.
We present statistical methodology for the construction of optimized designs for drug interaction studies. The goal is to choose two sets of drug concentrations for each individual in order to yield accurate parameter estimates, despite the lack of strong prior knowledge. Methods used include D- and product design optimality with space filling techniques. Currently, no software is available for this problem so adapted code from the software package POPT was used to optimize designs. The results from simulation-estimations in NONMEM are presented to (empirically) show how the optimized design outperforms the criss-cross design.