The FDA’s centers for pharmaceuticals and biologics have released a draft guidance specific to Bayesian methods for support of primary inference in clinical trials. The draft guidance was posted approximately five weeks before the agency declared that it would move more aggressively toward approving pharmaceutical and biotech premarket applications on the basis of a single trial, a clear sign of a policy shift at the agency.
The FDA’s Center for Devices and Radiological Health (CDRH) released a guidance on Bayesian clinical trial design in 2010, four years after release of the draft version. CDRH stated that advances in computer hardware had enabled the use of Bayesian mathematics, which the guidance described as computationally intense. Bayesian clinical trial constructs are widely known for their utility in factoring prior information into the design of a clinical trial. CDRH stated, however, that a Bayesian approach has utility even in the absence of prior knowledge as a tool by which a trial could be modified without compromising the validity of the data captured prior to the change in design.
The January 2026 draft guidance by the FDA’s drug and biologics centers applies to biologics license applications (BLAs), new drug applications (NDAs) and investigational new drug applications (INDs). The guidance fulfills a commitment made between the agency and industry under the sixth Prescription Drug User Fee Act (PDUFA) on the use of Bayesian methodologies.
The draft guidance describes scenarios in which Bayesian methods have previously been used, including the well-known use of information gleaned from previous trials of the drug in the same population. However, the agency also highlighted the use of Bayesian trial designs for the evaluation of drugs for pediatric populations based on prior information from studies of the drug in adult populations.
Other examples of the use of prior knowledge include dose-escalation studies and studies of a therapy that has been approved for diseases that are similar to the disease or condition identified in the proposed indication for use statement. The draft specifically calls out oncology drugs where the discussion of dose-escalation studies is concerned.
The 2026 draft complements a 2019 guidance for the use of Bayesian and other adaptive clinical trial methodologies and thus closes an important loop in the centers’ approach to non-traditional clinical trials. The scope of the 2019 guidance is also limited to BLAs, INDs and NDAs, but includes adaptive trial designs that do not rely on Bayesian analytical methods.
As is the case with the 2026 draft, the scope of the 2019 guidance is not explicitly limited to pivotal trials and includes descriptions of several types of study adaptations. Among these are adaptations designed to improve the probability of detecting a true drug effect than might be possible with a non-adaptive study design. However, the guidance points to several hazards associated with adaptive trial designs, such as the potential need for novel analytical methods to suppress the odds of erroneous conclusions. The agency also acknowledged that a change in the trial design could compromise the interpretability of the results.
The 2026 draft emphasizes the need for consultation with the agency before enrollment of the study. Among the considerations the FDA is interested in discussing with sponsors are the reliability of the success criteria, given the differences in how success criteria operate in a frequentist versus a Bayesian model. The docket is open for comment through March 13, 2026.