Model robust designs for survival trials

Abstract

The exponential-based proportional hazards model is often assumed in time- to-event experiments but may only approximately hold. We consider deviations in different neighbourhoods of this model that include other widely used paramet- ric proportional hazards models and we further assume that the data are subject to censoring. Minimax designs are then found explicitly based on criteria corre- sponding to classical c- and D-optimality. We provide analytical characterisations of optimal designs which, unlike optimal designs for related problems in the litera- ture, have finite support and thus avoid the issues of implementing a density-based design in practice. Finally, our designs are compared with the balanced design that is traditionally used in practice, and recommendations for practitioners are given.