A0292
Title: Computing constrained optimal designs with applications to dose-finding
Authors: Lenka Filova - Comenius University in Bratislava (Slovakia) [presenting]
Radoslav Harman - Comenius University in Bratislava (Slovakia)
Pal Somogyi - Comenius University in Bratislava (Slovakia)
Abstract: The utility of mathematical programming methods in computing optimal designs for clinical trials is demonstrated. These methods offer a flexible approach that is particularly beneficial for dose finding, as they readily incorporate constraints that are naturally inherent to the problem. An important application of this approach is in computing optimal designs for multivariate observations. This is crucial in various scenarios, such as in cases where efficacy and toxicity are jointly investigated or when monitoring multiple indicators. As an illustration, we focus on computing locally optimal and adaptive designs in a multinomial logistic model, particularly relevant to clinical trials. We achieve this using algorithms from the R library OptimalDesign, which enables users to compute both approximate and exact designs based on several frequently used optimality criteria.
