Learning, then confirming

There has been broad agreement for years among scientists at FDA and drug developers about the feasibility of using adaptive trial designs to increase the probability of success while reducing the time and cost of clinical development.

Yet despite the clear and urgent need to break free of the frequentist trial paradigm, use of alternative approaches such as Bayesian statistics to learn from and modify a trial while it is underway have been confined to the pages of academic journals and conference presentations. Instead, a combination of regulatory uncertainty and inertia has prevented the adoption of adaptive techniques. ...