In rare diseases where patient populations number in the hundreds or low thousands, a traditional randomized controlled trial with a concurrent placebo arm may be infeasible or unethical. Sponsors have increasingly turned to externally controlled trials, comparing patients receiving the investigational drug to an external cohort drawn from real-world data sources — natural history studies, patient registries, electronic health records, or prior clinical trials.
FDA has encouraged this approach in principle. The agency's draft guidance "Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products" and the 21st Century Cures Act framework for real-world evidence both signal regulatory openness. But in practice, FDA rejected at least three programs in 2025 that relied on external controls, citing bias, design flaws, lack of pre-specification, and unmeasured confounding. The gap between FDA's stated openness and its actual review behavior is where rare disease sponsors must focus. As a Friends of Cancer Research workshop panelist noted in 2026, "sponsors frequently conflate requests for regulatory flexibility with demands for regulatory certainty. While the FDA may exercise flexibility based on clinical context and unmet need, the statutory requirement for demonstrating substantial evidence of effectiveness remains unchanged."
This article is for biotech executives, regulatory affairs leads, clinical development teams, and medical affairs directors building evidence strategies for rare disease programs that may require external control arms.
What FDA's externally controlled trial guidance requires
The draft guidance framework
FDA's draft guidance on externally controlled trials, published in 2023 and still in draft as of May 2026, establishes recommendations for sponsors considering external control designs. The guidance defines an externally controlled trial as one in which "outcomes in participants receiving the test treatment according to a protocol are compared to outcomes in a group of people external to the trial who had not received the same treatment." External control arms can be derived from:
- Patient-level data from prior clinical trials (historical control).
- Real-world data from registries, electronic health records, or medical claims (concurrent control from another setting).
- Natural history studies that document disease progression in untreated patients.
The guidance focuses on patient-level data — not summary-level estimates — and emphasizes that FDA must have access to the raw data from the external control source for independent verification.
When FDA says external controls are most appropriate
FDA's 2023 guidance "Rare Diseases: Considerations for the Development of Drugs and Biological Products" states that external control designs "should be reserved for specific circumstances, such as clinical investigations where the drug effect can be demonstrated in diseases with well-understood and -characterized natural history, high and predictable mortality or progressive and predictable morbidity, and clinical investigations in which the drug effect is large and self-evident." This is a high bar. If the disease trajectory is variable, if the treatment effect is modest, or if the natural history is not well characterized, FDA is signaling that an external control arm is unlikely to be sufficient as the primary evidence basis.
Key requirements from the guidance
Pre-specification: The analysis plan, including the primary endpoint, the external control source, the eligibility criteria for the external cohort, and the statistical method for comparing arms, must be specified before the sponsor unblinds or accesses outcome data.
Comparability: The external control cohort must be comparable to the treatment arm on key prognostic factors — disease severity, age, prior therapies, genetic subtype, and other variables that affect outcome. FDA expects sponsors to demonstrate comparability through propensity score matching, stratification, or other adjustment methods.
Data quality: The external control data must be of sufficient quality to support a regulatory decision. This means complete follow-up, standardized outcome assessments, verified source data, and audit trails. Claims data and unstructured EHR data generally do not meet this standard without significant curation.
Sensitivity analyses: FDA expects sensitivity analyses that test the robustness of the treatment effect estimate against different assumptions about missing data, confounding, and cohort selection.
Early FDA engagement: The guidance recommends pre-IND or Type C meetings to discuss the external control design before the trial is underway, not after data collection is complete.
Where sponsors have failed: the 2025 CRLs
Capricor — deramiocel for Duchenne muscular dystrophy cardiomyopathy
Capricor's BLA for deramiocel was supported by data from the Phase 2 HOPE-2 trial, which included an open-label extension and comparisons to natural history data from FDA-funded datasets. FDA issued a complete response letter in July 2025. The CRL identified the natural history external control comparison as insufficient to support approval and noted that HOPE-2 did not meet its primary efficacy endpoint. Capricor provided topline data from its ongoing Phase 3 HOPE-3 trial to FDA in late 2025; FDA accepted the Class 2 resubmission with a PDUFA target action date of August 22, 2026.
Biohaven — troriluzole for spinocerebellar ataxia
Biohaven submitted an NDA for troriluzole in spinocerebellar ataxia supported by both a one-year placebo-controlled study and a three-year real-world evidence study. The RWE study met its primary and secondary endpoints. FDA issued a CRL in November 2025 citing "issues that can be inherent to real-world evidence and external control studies including potential bias, design flaws, lack of pre-specification and unmeasured confounding factors." The specificity of the CRL language — identifying pre-specification and confounding by name — signals that FDA is not willing to accept post-hoc RWE analyses even when the endpoint results are positive.
UniQure — AMT-130 for Huntington's disease
UniQure's gene therapy AMT-130 was reviewed under a regulatory pathway where FDA had agreed in 2024 that a Phase 1/2 study using a natural history external control could support a BLA. In November 2025, FDA "no longer agrees" with that position and strongly recommended a prospective, randomized, double-blind, sham surgery-controlled Phase 3 trial. The reversal is particularly notable because it involved FDA changing its own prior agreement with the sponsor — suggesting that the agency's internal review of external control acceptability is evolving and that early agreements may not be binding.
The pattern
All three CRLs share a common thread: the external control evidence was evaluated as insufficient relative to the evidentiary standard FDA applied. In none of these cases did FDA reject the concept of external controls outright. Instead, FDA found the implementation lacking — in pre-specification, in confounding control, in data quality, or in the strength of the natural history baseline.
What makes a rare disease RWE package credible
1. A prospectively designed, well-characterized natural history study
FDA's 2019 draft guidance "Rare Diseases: Natural History Studies for Drug Development" emphasizes that natural history studies should be designed to serve as the basis for external comparisons. This means:
- Prospective longitudinal follow-up with standardized assessments.
- Pre-defined endpoints that match those planned for the treatment trial.
- Sufficient duration to capture disease progression at the same time points as the trial.
- Genetic and biomarker characterization of the cohort that matches the trial population.
Retrospective chart reviews and ad hoc analyses of existing datasets rarely meet this standard. The natural history study should be initiated before or concurrently with the treatment trial, not assembled after trial data are available.
2. Patient-level data with complete audit trail
FDA requires access to patient-level data from the external control source. Summary-level statistics or published literature estimates are not sufficient. The external control dataset must include:
- Individual patient identifiers (de-identified but linkable for audit purposes).
- Source documentation for each data point (lab reports, imaging, clinician assessments).
- Complete follow-up data for the relevant time window.
- Clear documentation of any data imputation or transformation.
3. Rigorous comparator cohort selection
The external control cohort must be selected using transparent, pre-specified criteria that match the trial's inclusion and exclusion criteria as closely as possible. Key considerations:
| Factor | What to document |
|---|---|
| Disease definition | Diagnostic criteria, genetic confirmation, disease stage |
| Prior therapies | Allowed and excluded prior treatments, washout periods |
| Age and demographics | Age at onset, age at baseline, sex, ethnicity |
| Disease severity | Baseline functional scores, biomarker levels, organ involvement |
| Follow-up timing | Interval between baseline and outcome assessments |
| Geographic and temporal factors | Country, care setting, time period of data collection |
4. Statistical methods that address confounding
FDA expects sponsors to use statistical methods that explicitly address the non-randomized nature of the comparison. The 2026 Bayesian methodology guidance provides new regulatory direction on incorporating prior data and external controls using Bayesian frameworks. Accepted approaches include:
- Propensity score matching or weighting to balance prognostic factors between arms.
- Bayesian hierarchical models that borrow strength from external data while accounting for between-study heterogeneity.
- Sensitivity analyses with tipping-point analyses to determine how large an unmeasured confounder would need to be to nullify the treatment effect.
- Multiple imputation for missing data, with pre-specified assumptions about missingness mechanisms.
5. Concordance with confirmatory evidence
FDA's 2026 policy shift to a "single-trial standard with confirmatory evidence" default means that a single externally controlled trial can support approval if it is backed by additional confirmatory evidence. This evidence can include mechanistic data, biomarker data, animal model data, or data from related indications. The plausible mechanism framework for individualized ultra-rare disease therapies, announced in February 2026, creates an additional pathway where mechanistic data can supplement trial evidence when traditional trial designs are not feasible.
For rare disease programs using external controls, the confirmatory evidence package should include:
- Mechanism of action data (target engagement, pharmacodynamic biomarkers).
- Dose-response data from early-phase studies.
- Nonclinical efficacy data from relevant animal models.
- Natural history data showing the disease trajectory in untreated patients.
The operational checklist
Pre-IND meeting (recommended 12–18 months before BLA/NDA submission)
| Item | Purpose |
|---|---|
| Present external control source and justification | Get FDA agreement on the data source |
| Present natural history study design | Confirm endpoints match trial endpoints |
| Present statistical analysis plan | Get FDA buy-in on confounding adjustment methods |
| Discuss confirmatory evidence strategy | Align on what supplemental data is needed |
| Request written minutes | Document FDA's feedback for the BLA/NDA module |
During the trial
- Enroll the external control cohort prospectively using the same outcome assessment schedule as the treatment arm.
- Maintain an independent data monitoring committee for the external control data.
- Do not access outcome data from the external control until the pre-specified analysis window.
- Document any deviations from the pre-specified protocol and their rationale.
BLA/NDA submission
- Include the complete external control dataset with patient-level data in the submission.
- Include the statistical analysis plan as a standalone document with version control.
- Include sensitivity analyses and tipping-point analyses.
- Include a section on limitations of the external control design and how they were mitigated.
What changed in 2026
The plausible mechanism framework
On February 23, 2026, FDA issued a draft guidance on the "Plausible Mechanism Framework" for individualized therapies targeting specific genetic conditions with known biological cause. The guidance applies primarily to gene editing and RNA-based therapies (e.g., antisense oligonucleotides) for ultra-rare diseases. Under this framework, sponsors can build approval cases from mechanistic data when traditional trials are not feasible, provided the mechanism is well-understood and the disease cause is known.
The framework explicitly acknowledges the role of natural history data and external controls as components of the evidence package, but it does not replace the need for rigorous methodology. Applied Clinical Trials noted in its 2026 analysis of FDA actions that "sponsors should reassess late-stage designs against the single-trial standard and identify where Bayesian methods could improve efficiency or make better use of existing data."
The single-trial standard
FDA Commissioner Makary and CBER Director Prasad announced in a February 2026 NEJM article that the agency's default position is now "one adequate and well-controlled study, combined with confirmatory evidence" for new drug approvals. This is particularly relevant for rare disease programs using external controls: a single externally controlled trial with strong confirmatory evidence may now be sufficient, whereas FDA previously expected at least two independent studies.
Bayesian methodology guidance
FDA's 2026 guidance on Bayesian methods in drug development provides clearer regulatory direction on incorporating prior data, real-world evidence, and external controls into adaptive trial designs. For rare disease sponsors, this means that Bayesian frameworks can formally weight external control data according to its relevance and quality, rather than treating it as equivalent to a concurrent randomized control.
What to monitor next
- Finalization of the externally controlled trials guidance: The guidance is still in draft. When finalized, it may tighten or loosen the requirements for pre-specification, data quality, and sensitivity analyses.
- FDA review outcomes for HOPE-3 (Capricor): The August 2026 PDUFA date for deramiocel will be a bellwether for whether a more rigorously designed external control package can survive FDA review.
- UniQure Phase 3 design for AMT-130: If FDA requires a full sham-controlled trial for Huntington's gene therapy, it will signal limited tolerance for external controls in CNS indications even with strong natural history data.
- Evolving ECA use in oncology: Friends of Cancer Research continues to pilot external control methodologies in oncology, and FDA's oncology review divisions may develop different standards than rare disease review divisions.
Sources
- FDA Draft Guidance, "Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products" (2023): Regulatory framework for external control arm design. Available at https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-design-and-conduct-externally-controlled-trials-drug-and-biological-products.
- FDA Draft Guidance, "Rare Diseases: Natural History Studies for Drug Development" (March 2019): Natural history study design recommendations. Available at https://www.fda.gov/regulatory-information/search-fda-guidance-documents/rare-diseases-natural-history-studies-drug-development.
- FDA Draft Guidance, "Considerations for the use of the Plausible Mechanism Framework to Develop Individualized Therapies that Target Specific Genetic Conditions with Known Biological Cause" (February 2026): Ultra-rare disease evidence framework. Available at https://www.hhs.gov/press-room/fda-launches-framework-accelerating-development-individualized-therapies-ultra-rare-diseases.html.
- FDA, "Rare Diseases: Considerations for the Development of Drugs and Biological Products": Overall rare disease drug development guidance. Available at https://www.fda.gov/drugs/guidances-drugs/guidance-documents-rare-disease-drug-development.
- FDA Real-World Evidence program page (updated April 2026): RWE guidance documents and policy framework. Available at https://www.fda.gov/science-research/science-and-research-special-topics/real-world-evidence.
- 21st Century Cures Act, Section 3022: Real-world evidence statutory authority.
- BioSpace, "FDA Reversals in Rare Disease Space Highlight Confusion Around External Controls" (2025): Capricor, Biohaven, and UniQure CRL analysis.
- Applied Clinical Trials, "Four FDA Actions Reshaping Drug Development in 2026": Single-trial standard, plausible mechanism framework, and Bayesian guidance analysis.
- Arnold & Porter Advisory, "FDA Advances a Plausible Mechanism Framework for Rare Disease Drug Development" (February 2026): Legal analysis of FDA policy shift.
- Drug Discovery and Development, "Rethinking Evidence Generation for Ultra-Rare Diseases" (April 2026): Operational implications for rare disease evidence strategies.
- Evidinno, "Advancing Evidence Generation with External Control Arms" (November 2025): Technical framework for ECA design and implementation.
- Friends of Cancer Research, External Control Arms Workshop (2026): Oncology ECA pilot project and regulatory discussion.
- Khachatryan A, Read SH, Madison T. "External control arms for rare diseases: building a body of supporting evidence." J Pharmacokinet Pharmacodyn. 2023;50(6):501-506.
- Jahanshahi M, Gregg K, Davis G, et al. "The use of external controls in FDA regulatory decision making." Ther Innov Regul Sci. 2021;55(5):1019-1035.
