FDA Support Drives Shift Toward Bayesian Trial Design

In a recent video interview with Applied Clinical Trials, David Morton, PhD, director of biostatistics, Certara, discussed how the FDA’s increased support for Bayesian methods is expected to drive a shift from traditional frequentist approaches toward more flexible, probability-based trial designs. He emphasized the growing role of external data borrowing, adaptive decision-making, and simulation-driven planning in improving efficiency, particularly in rare disease and small population studies. Morton also highlighted the need for stronger upfront design, including justification of prior data and robust analytical frameworks, as well as greater cross-functional alignment and early regulatory engagement to support transparency, reduce uncertainty, and accelerate more data-driven development models.

Editor's note: This transcript is a lightly edited rendering of the original audio/video content. It may contain errors, informal language, or omissions as spoken in the original recording.

ACT: How could Bayesian approaches change how trials are designed, monitored, or adapted during execution?

Morton: A lot of teams using Bayesian models can integrate prior knowledge from previous studies into sample size and decision criteria. This leads to more efficient designs, especially for small populations, and can inform design elements like dose selection for subsequent trials.

Bayesian calculations also govern the timing and adaptation rules for interim analyses. It can support stop, go, or no-go decisions, and it’s really more of a continuous learning approach versus a fixed interim analysis.

One of the key aspects is that it’s all probability-based decision making, based on posterior probability distributions of success. So trials become more responsive and more information-driven. It also allows for adaptive designs where you can use different types of priors to control early stopping for efficacy or futility.

We’ve run simulations to see when stopping rules might occur, and it gives you probabilities of outcomes rather than relying on traditional confidence intervals, which are based on repeated trials. So it really shifts from confidence to probability.