Optimisation of Clinical Trials using Bayesian Decision Theory

A decision maker confronted with the task of designing a clinical trial has to consider a multitude of aspects. Large trials lead to more evidence,which in turn makes it less likely that harmful decisions are taken when deciding on future treatments for patients. On the other hand, large trials typically require substantial financial resources and may take a long time to finish. This trade-off between the cost of pharmaceutical R&D and the value of the data generated motivates a study of how clinical trials can be optimised in practice. The size of the trial is but one aspect of its design. In a setting where there is some prior evidence that a treatment works better for a subpopulation defined by a biomarker, it is natural to ask whether the resources available are best spent by restricting trial recruitment via biomarker screening. The optimal trial design also depends on the type of the decision maker. For a commercial sponsor, it is vital that invested resources may eventually be recouped via incomes from sales. On theother hand, a publicly funded trial might instead be optimised purely from a public health perspective. By viewing the trial as a stage in a sequential problem, post-trial decisions regarding pricing for a new treatment may also affect optimality. Bayesian decision theory is a flexible framework that may be applied when searching for an optimal course of action in an uncertain environ-ment. In particular, it allows for different beliefs prior to the trial and different preferences for the trial outcomes. This thesis presents two pa-pers in which Bayesian decision theory is used to find the optimal trial design under two different models for the contemporary regulatory envi-ronment. In addition to providing a methodology for trial design in the specific situations considered, the analysis also leads to insights regarding the impact of typical regulatory rules on the behaviour of trial sponsors