How Bayesian Approaches Can Revolutionize Rare Disease Clinical Trials

Randomized controlled trials (RCTs) are the gold standard for generating regulatory-grade evidence on treatment efficacy or safety during drug development. However, they are not always ethical or practical possible in the rare disease space. What works for common conditions with large patient populations can present multiple feasibility challenges for rare disease research, which demands both parsimony and adaptability.

Bayesian trial designs offer the opportunity to enable rare disease research by offering greater flexibility, and a framework to integrate external data, to seamlessly quantify risks and probability of success so to facilitate data-driven decision-making across the trial lifecycle. In January 2026, the FDA released draft guidance on the use of Bayesian methods for drug and biologic development.¹ This marked the first time the regulator had published guidance on Bayesian trials use for drugs, outside of brief mentions within other guidelines on topics like rare disease and adaptive trial designs. An accompanying statement said Bayesian methodologies ‘help address two of the biggest problems of drug development: high costs and long timelines.’²

The guidance outlines the difference between the frequentist methods typically used for RCTs and Bayesian settings, and the need to evaluate frequentist operating characteristics under a Bayesian approach. While frequentist methods depend on fixed hypotheses and predetermined sample sizes, Bayesian approaches incorporate prior knowledge and update the probability of treatment effects as new trial data becomes available. This creates efficiencies without compromising statistical integrity.

In this paper, Billy Amzal, head of strategic consulting at Phastar, and a Bayesian design expert and pioneer, explores the techniques used in Bayesian trial designs, examines regulatory updates, and shares real-world examples of the impact of these innovative methods.

The challenge: Underpowered trials

Most RCTs are underpinned by the frequentist framework, which relies on predefined sample sizes to control for Type I error, and conclusions based on statistical significance thresholds. While rare diseases affect around 300 million people globally, each individual condition affects relatively few individuals.³ This makes recruitment of large sample sizes challenging and impracticable in many rare disease or small population trials.⁴

Failure to meet the sample size target can lead to an increased possibility of incurring a Type II error. Prolonged recruitment periods can also lead to increased costs, increased uncertainty about treatment effectiveness and delays in access to new treatments. Slow acquisition of evidence can also reduce investment by funders who prefer to invest in more rapid approaches to evaluation.⁵

These challenges often call for the adoption of alternative approaches and analytical frameworks capable of extracting maximal information from limited data.

One solution: Bayesian trial designs

Bayesian trial designs provide an alternative to traditional frequentist approaches. While frequentist statistics focuses on observing events to disprove a null hypothesis at a specific p-value and power, Bayesian statistics uses observed events as a way of updating prior beliefs. Prior beliefs can be based on historical data, smaller trials and expert opinion from healthcare professionals. By combining information from all these sources, reliance on large sample sizes can be reduced. Use of Bayesian methods has been shown to reduce the required number of trial participants by 30-2,400% compared to frequentist models.⁴

The new draft guidance from the FDA highlights, at length, the development and modeling of priors. It emphasizes the importance of thoroughly reviewing the data and sources involved within developing a prior, especially if multiple sources need to be combined, and of being able to demonstrate where information came from and its appropriateness.¹

Bayesian trials are also very suitable to adaptive designs. By nature, Bayesian methods allow researchers to adapt to new information as it emerges while maintaining statistical rigor and reduced patient exposure to ineffective treatments through early stopping rules. Bayesian adaptive designs are e.g. suitable for dose selection trials enabling early stopping rules and best performing selection while controlling for operating characteristics and the risks of false conclusion. Probability-based outputs are also more intuitive for sponsors, investigators and regulators.

The use of Bayesian methods is particularly useful in rare disease research or in indication with very low incidence of clinical outcome, situations where data is limited. For example, in this trial evaluating a prevention strategy of mother-to-child HIV transmission,⁶ a Bayesian adaptive design was enabling a single arm design and early stopping rules and compared with external control defined by a Bayesian meta-analysis of historical trials. As a matter of fact, transmission rates were in the range of 1 to 3% with practical challenges to recruit pregnant women at risk of transmission. This approach allowed a division by 5 the sample size compared to a standard frequentist design.

Harnessing historical, real-world data and expert knowledge

Prior distributions typically leverage real-world data and historical data. For example, a new treatment for the rare and aggressive cancer refractory precursor B-cell acute lymphoblastic leukemia was granted accelerated approval based upon findings from a single-arm, open-label Phase II study supported by a historical control arm and external control arms using summary-level outcome estimates from previous trials.⁷

However, in many rare-disease trials, historical data are limited or unavailable. This makes the use of expert priors a vital component of Bayesian trial designs, especially in ultra-rare conditions or early development stages.

Expert prior elicitation can also enhance rare disease research by enabling teams to formally incorporate expert opinion into Bayesian trial designs. Prior elicitation follows a structured interview process in which experts are asked questions to ascertain their beliefs on, for example, a treatment effect. This allows researchers to transform clinical expertise into formal prior probability distributions that incorporate uncertainty.

Preparation is key to ensure good quality priors are generated from an elicitation. Key considerations include the creation of a clear endpoint, comparability between the elicited prior and the trial, and careful selection of experts. All experts should also have an evidence dossier to ensure they all have the same, unbiased information.

During elicitation, the use of graphical representations helps experts to explore their own understanding. Group discussions allow experts to argue for their own beliefs and hear arguments from others. The eventual aim is to elicit a single aggregate prior which represents the collective belief of all the experts present based on discussion of their individual priors. Techniques like the SHELF (Sheffield Elicitation Framework) protocol can help to create visualizations and ensure reproducibility when producing priors.⁸

Regulatory considerations for Bayesian trials

The benefits of Bayesian trials are increasingly being recognized by regulators. The new draft guidance issued by the FDA in January is specifically designed to facilitate the use of Bayesian methodologies.⁹ It aims to provide ‘clarity on modern statistical methods,’ helping sponsors bring more cures and meaningful treatments to patients faster and more affordably. For example, the guidance recognizes Bayesian borrowing as a powerful approach for extrapolation, provided a thorough evaluation takes place.

An accompanying statement issued by the FDA further highlights the benefits of Bayesian methods, including making better use of available data, more efficient clinical trials and delivering safe and effective treatments to patients sooner. It says Bayesian methods may be particularly valuable in rare indications.²

Similarly, the EMA has noted the evolution of Bayesian methodologies, and their potential uses in early clinical development and when there are rare indications.¹⁰ Comments on the draft ICH M15 guideline on general principles for model informed drug development have also highlighted the need to specifically include Bayesian methods in the final guidance.⁹

Current adoption of Bayesian methods

Despite its advantages, particularly in rare disease research, and increasing regulatory backing, Bayesian statistics remain underutilized. A review of rare disease and small population trials found just 6% used Bayesian methods. The rest of the trials used frequentist approaches and the majority of these failed to achieve their recruitment targets.⁴

Analysis of the use of Bayesian methods over the past 20 years also found uptake is still scarce and mostly applied to the analysis of treatment efficacy in single-arm trials with binary endpoints.¹¹ However, the same study highlighted the opportunity to make use of the advantages of Bayesian methods, particularly in settings with small populations and severe conditions with high unmet needs, like many rare diseases.

Conclusion

Bayesian trial designs offer a powerful, flexible alternative to traditional approaches, supporting interim analyses, early stopping for efficacy or futility, and adaptive modifications—all without compromising statistical integrity.

When designed, justified and implemented correctly, Bayesian methods can reduce risk and speed up access to potentially life-changing treatments for patients, while also offering sponsors a more transparent, responsive, and informative pathway to generating high-quality clinical evidence.

With rare disease a growing area of clinical research and clearer expectations set by the FDA, the industry can no longer afford to wait when it comes to adopting Bayesian methods. We need to leverage adaptive, data-driven designs if we are to harness new information, reduce patient exposure to ineffective treatments and optimize rare disease research.

Billy Amzal, Head of Strategic Consulting at Phastar

References

1. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/use-bayesian-methodology-clinical-trials-drug-and-biological-products
2. https://www.fda.gov/news-events/press-announcements/fda-issues-guidance-modernizing-statistical-methods-clinical-trials
3. https://www.thelancet.com/journals/langlo/article/PIIS2214-109X(24)00056-1/fulltext
4. https://pubmed.ncbi.nlm.nih.gov/34910979/
5. https://pmc.ncbi.nlm.nih.gov/articles/PMC1559709/
6. https://pmc.ncbi.nlm.nih.gov/articles/PMC9741956/ 
7. https://pmc.ncbi.nlm.nih.gov/articles/PMC10673956/
8. https://link.springer.com/chapter/10.1007/978-3-319-65052-4_4
9. https://www.ema.europa.eu/en/ich-m15-guideline-general-principles-model-informed-drug-development-step-2b-scientific-guideline#current-version-71848
10. https://health.ec.europa.eu/system/files/2022-06/medicinal_qa_complex_clinical-trials_en.pdf
11. https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2025.1548997/full