Abstract

Clinical trials are used to evaluate the health benefit of new health technologies, such as pharmaceuticals, but are quite costly and therefore have been the subject of much study. Health technology adoption decisions are often made based on not only health benefit, but the costs of drugs and treatment processes. Is it possible that this mismatch between incentives at different steps of the health innovation pipeline, clinical effectiveness on the one hand and cost-effectiveness on the other, may lead to suboptimal decisions? We introduce and explore a stream of work that seeks to improve the allocation of resources to clinical trials in a way that balances health value for money for treatments that are ultimately approved. The stream uses work from Bayesian sequential optimal learning and from game theory. We first look at basic trade-offs in a simple two-arm fully sequential trial design, to balance the costs of collecting more trial data with the expected opportunity costs averted by making decisions with better information. We then explore how the theory can apply to UK-NIHR funded clinical trials (including retrospective looks at the ProFHER trial, the CACTUS trial, and the HERO trial), and overview extensions that allow the framework to apply to multiarm trials, precision medicine trials, and explore implications for conditional approval schemes (motivated by the UK Cancer Drugs Fund).