A Value-Based and Highly Sequential Approach to Learning in Clinical Trials

Information Systems and Operations Management (ISOM)

Speaker : Stephen Chick 

From : INSEAD

Room: Bernard Ramanantsoa

"Clinical trials are necessary for evaluating the benefit of new health technologies but are quite costly, and have therefore been the subject of much study. This work explores the logical implication for clinical trial design of two important trends: (a) greater accountability in terms of cost and health benefits for new health technologies, and (b) adaptive clinical trial designs and advances in other sectors for highly sequential optimal learning. We propose and solve a Bayesian decision-theoretic model of a fully sequential experiment in which the end point is observed with delay, and where the trial design is based on health economic criteria rather than abstract statistical criteria. We identify sequential experiments which maximize the expected benefits of technology adoption decisions, minus sampling costs. The solution yields a unified policy defining the optimal `do not experiment'/`fixed sample size experiment'/`sequential experiment' regions and optimal stopping boundaries for sequential sampling, as a function of the prior mean benefit and the size of the delay. The model can value the expected benefits accruing to study units and the fixed costs of switching from control to treatment. We apply the model to the field of medical statistics, using data retrospectively from published trials from pragmatic trials in Europe, as well as for highly-sequential, highly multi-arm trial design for dose optimization. We highlight potential benefits of the approach."