FDA, Duke Workshop Examines Trial Design for Rare Diseases

September 29, 2026

Speakers discussed randomization, external data, adaptive methods, and endpoints for clinical investigations with small patient populations

Regulators, statisticians, researchers, industry representatives, and patient advocates met September 29 at the National Press Club in Washington to discuss how clinical investigations can produce reliable evidence when rare-disease populations are too small for conventional trials.

Tagline Strategies attended the workshop “RISE to the Challenge: Statistical Considerations for Rare Disease Clinical Investigations,” co-convened by Duke-Margolis Institute for Health Policy and the FDA Rare Disease Innovation Hub, to gain insight and help clients navigate communications challenges surrounding rare-disease trials.

Rare-disease studies often involve small, clinically diverse patient populations. Researchers may have limited information about a disease’s natural history, little agreement on appropriate endpoints, and difficulty observing disease progression within a practical study period. These conditions can reduce statistical power and leave investigators with inconclusive results.

The workshop did not establish a new evidentiary standard or produce formal recommendations. Speakers instead considered how trial sponsors could adapt study designs while controlling bias and producing results that patients, clinicians, and regulators can interpret.

Dr. Rebecca Rothwell Chiu, a supervisory mathematical statistician with the FDA’s Center for Drug Evaluation and Research (CDER), opened the statistical discussion. She welcomed how often speakers discussed endpoints that evaluate “things that happen to patients.” Her point was that trials should assess outcomes meaningful to patients, such as changes in how they feel and function, while still producing reliable evidence about whether a treatment works.

Randomization prompted some of the day’s most direct exchanges. Dr. Thomas Fleming, a biostatistician at the University of Washington, cautioned that a trial designed to be easier to conduct—such as one in which everyone receives the experimental treatment and success is measured by a short-term biomarker change—may leave patients unable to make an informed choice. He argued that patients need reliable evidence of a treatment’s clinical benefits and risks, and that randomization should be used when possible because it reduces the chance that differences between patients will be mistaken for a treatment effect. He also described circumstances in which nonrandomized evidence may be persuasive, such as when spontaneous improvement is unlikely, or the observed treatment effect is unusually large.

Dr. A.J. Allen, a former researcher at Lilly, described the competing perspectives surrounding trial participation. A rare disease community may want definitive evidence, while an individual patient may focus on the chance of receiving experimental treatment. That tension can shape enrollment, retention, and acceptance of different allocation ratios. Allen encouraged sponsors to discuss those choices with patient communities early.

Speakers also reviewed Bayesian borrowing, adaptive and smooth designs, within-patient controls, unequal randomization, and statistical tests that combine evidence from several outcomes.

Dr. Wonyul Lee, a senior statistical reviewer in CDER’s Office of Biostatistics, discussed global tests for diseases with multiple clinical manifestations. His simulations showed potential power gains when a treatment has similar effects across distinct endpoints. Those gains can shrink when endpoints are highly correlated or when the analysis includes outcomes on which the treatment is unlikely to have an effect. Dr. Lee also cautioned that a positive global test does not show that any single endpoint reached statistical significance.

Natural history research can help resolve some of these design questions before a pivotal trial begins. Dr. Bradley Williams, the director of research at the Jain Foundation, called a natural history study a “dress rehearsal for a clinical trial.” Such research can identify reliable outcomes, inform assessment schedules, and help determine the treatment difference a later study should be designed to detect.

Dr. Antara Majumdar, the director of oncology statistics at GSK, described using external control data to supplement the internal control arm of a randomized study. Dr. Majumdar said similar eligibility criteria do not, by themselves, make historical and current data comparable. Investigators must also examine endpoint definitions, data-collection methods, index dates, geography, changes in standards of care, and when the data were collected.

Several speakers warned against treating statistical efficiency as a substitute for evidence quality. Dr. Scott Berry, a statistical scientist, put the point plainly: “Efficiency is not a shortcut.” The director of the FDA’s Center for Devices and Radiological Health Office of Clinical Evidence and Analysis, Dr. Daniel Caños, gave a similar warning about the assumptions behind Bayesian methods: “There’s no free lunch in statistics.”

Early-feasibility studies can generate repeated measurements on technical performance, reliability, usability, assessment timing, and participant burden. Speakers noted that many observations from one participant can produce a more precise estimate for that person. They do not replace evidence from independent participants.

The discussion underlined that there is no single design formula for rare-disease studies. The appropriate method depends on the disease, the expected treatment effect, the available data, and the clinical question. Speakers urged sponsors to plan early, build strong natural-history datasets, involve patients in study design, state assumptions clearly, and consult the appropriate FDA review division before protocols and operational plans are fixed.

Tagline’s View

For Tagline Strategies, the workshop offered a clearer way to help clients communicate the choices behind rare-disease trials. We can help sponsors explain why a study uses a particular endpoint, control group, or external dataset; what those choices mean for patients; and what the resulting evidence can and cannot show. Bringing patients’ perspectives into those conversations early, and communicating uncertainty plainly as studies progress, can help clients build understanding and trust without overstating the strength of their findings.

Back to All News

Stay in the Know with
Our Latest Insights

Sign up for Tagline’s Newsletter, a digest of regulatory, policy, and communications developments shaping the industries we serve. We promise not to spam you! Our aim is to publish a quarterly newsletter with special features once in a while.

Sign Up