FDA Cautions on Use of ‘Bayesian’ Statistics in Complex Trials

The FDA urges trial sponsors who may be thinking of using Bayesian statistics in complex innovative trials to consider alternatives, in a new draft guidance.

The agency is committed to fostering complex innovative trial design (CID) — defined as trials using adaptive, Bayesian, and other novel statistical approaches — to speed up drug development. But it advises CID sponsors to lay out the benefit-risk tradeoffs when using such statistical methods.

“The use of Bayesian methods can increase the chance of erroneous conclusions,” the agency says, adding that discussions between sponsors and the agency are “critical to the FDA’s evaluation of Bayesian proposals.”