As the temporal, financial, and ethical cost of randomized clinical trials (RCTs) continues to rise, researchers and regulators in drug discovery and development face increasing pressure to make better use of existing data sources. This pressure is especially high in rare disease, where traditionally designed RCTs are often infeasible due to the inability to recruit enough patients or the unwillingness of patients or trial leaders to randomly assign anyone to placebo. Bayesian statistical methods have recently been recommended in such settings for their ability to combine disparate data sources, increasing overall study power. The use of these methods has received a boost in the United States thanks to a new willingness by regulators at the Food and Drug Administration to consider complex innovative trial designs. These designs allow trialists to change the nature of the trial (eg, stop early for success or futility, drop an underperforming trial arm, incorporate data on historical controls, etc) while it is still running. In this article, we review a broad collection of Bayesian techniques useful in rare disease research, indicating the benefits and risks associated with each. We begin with relatively innocuous methods for combining information from RCTs and proceed on through increasingly innovative approaches that borrow strength from increasingly heterogeneous and less carefully curated data sources. We also offer 2 examples from the very recent literature illustrating how clinical pharmacology principles can make important contributions to such designs, confirming the interdisciplinary nature of this work.
Keywords: Bayesian methods; clinical trial; complex innovative design (CID); drug development; historical data; natural history data; rare disease.
© 2022, The American College of Clinical Pharmacology.