Zhaohui Su, VP of Biostatistics at Ontada, shared a post on LinkedIn:
“The FDA’s 2026 draft guidance on Bayesian methodology represents a significant advancement in the integration of real-world evidence (RWE) within clinical trials. This guidance supports the use of Bayesian models for primary inference, early stopping, subgroup evaluation, and the incorporation of external or non-concurrent controls, provided that data sources and operating characteristics are transparently justified.
The importance of this guidance lies in its ability to combine randomized controlled trial (RCT) data with high-quality real-world data (RWD) in a scientifically coherent manner. Rather than initiating trials from the ground up, we can leverage informative priors derived from registries, electronic health records (EHR), claims, natural-history studies, and prior clinical experiences.
Additionally, Bayesian augmented-control designs facilitate the reduction of randomized control arm sizes by borrowing from external data while automatically down-weighting non-exchangeable information. This capability is particularly imporrtant in the context of rare diseases and small populations where conducting large trials is not feasible.
In summary, Bayesian models yield decision-focused outputs, including:
- Probability the treatment works
- Probability a future patient benefits
- Probability a future study replicates
These outputs address the questions that regulators seek to answer. For more details, refer to the FDA‘s guidance here. Welcome your comments. Thank you.”
More posts featuring Zhaohui Su.