Antibody illustration in orange and teal on a light ivory background

Externally Controlled Trials: When FDA Accepts External Controls

Table of Contents

What an externally controlled trial is, when FDA accepts an external or natural history control arm, and what the 2023 draft guidance asks sponsors to do.

Last updated: November 22, 2026

An externally controlled trial compares patients on a test drug with a group outside the trial, such as a natural history cohort or registry. FDA treats these designs as a special case: they work best when the disease course is predictable, the effect is large and the endpoint is objective.

Scope: the regulatory basis, FDA’s February 2023 draft guidance, three approvals that relied on external comparators, and a checklist for sponsors weighing the design.

What is an externally controlled trial?

An externally controlled trial gives the test drug to patients under a protocol and compares their outcomes with a group that was not part of the trial. The comparison group can come from an earlier period or from the same period in a different setting. Nobody is randomized to the control.

FDA’s draft guidance, Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products, defines the design as one where outcomes in participants receiving the test treatment according to a protocol are compared with outcomes in an external group. That group can be a historical control, drawn from people treated or untreated at an earlier time, or a concurrent control from the same time period in another setting.

The guidance focuses on patient level data from other clinical trials or from real world data sources such as registries. Natural history studies, which follow untreated patients to document how a disease progresses, are a common source in rare disease.

Is the FDA externally controlled trials guidance final?

No. As of November 22, 2026, the guidance is still a draft. FDA’s landing page lists a February 2023 issue date, carries the “Not for implementation” label used on drafts, and shows content current as of January 31, 2023.

The guidance was issued jointly by CDER, CBER and the Oncology Center of Excellence under docket FDA-2022-D-2983. FDA accepts comments on it at any time. Being a draft does not make it irrelevant: it is the clearest written statement of how reviewers think about these designs, and sponsors should expect review teams to apply its logic.

What does the law say about historical controls?

FDA’s regulation on adequate and well controlled studies, 21 CFR 314.126, lists historical control as one of five acceptable control types, but says “historical control designs are usually reserved for special circumstances.”

The regulation requires a design that permits a valid comparison with a control to provide a quantitative assessment of drug effect. It names four concurrent control types (placebo, dose comparison, no treatment and active treatment) plus the historical control. In our paraphrase of paragraph (b)(2)(v), a historical control compares results on the test drug with experience elsewhere, either the documented natural history of the disease or active treatment in comparable patients. The regulation notes that historical groups cannot be assessed on pertinent variables as well as concurrent groups, and points to diseases with high and predictable mortality and to drugs whose effect is self evident as the typical uses.

The international standard says the same thing more bluntly. ICH E10, Choice of Control Group, calls the inability to control bias the major and well recognized limitation of externally controlled trials, and limits their use to cases where “the effect of treatment is dramatic and the usual course of the disease highly predictable.”

When does FDA accept an external control arm?

FDA is most receptive when three conditions line up: a disease with a well defined natural history, an expected effect that is large, and an endpoint that is objective and reliably recorded outside a trial. When those conditions are missing, FDA says sponsors should choose a different design.

The draft guidance is direct on the downside. In many cases, it says, the likelihood of credibly demonstrating effectiveness with an external control is low, and “sponsors should choose a more suitable design.” Death, measured survival and some hard functional milestones travel well into external datasets. Symptom scales, investigator judged outcomes and anything sensitive to how closely patients are watched do not.

Factor Favors an external control Argues against it
Disease course Well documented, predictable progression or mortality Variable course, spontaneous improvement
Expected effect Large, well beyond natural variation Modest or incremental
Endpoint Objective: death, survival, defined motor milestones Subjective scales, patient reported symptoms
Comparator data Patient level data with source access Summary statistics, publications only
Standard of care Stable over the comparison period Changed between control era and trial era
Population size Ultra rare, randomization infeasible or unethical Enough patients for a randomized trial

What biases does FDA worry about?

The core problems are confounding the sponsor cannot measure and the absence of blinding. The draft guidance says these cannot be eliminated, so the job is to design around them and show FDA how.

The guidance flags specific traps:

  • Cherry picking the comparator. Choosing a control arm from a completed trial whose results are already known is a particular concern.
  • Index date mismatch. If the start of follow up is defined differently in the treated and external arms, FDA says the effect estimate may be biased.
  • Intercurrent events. Events like treatment switching or rescue therapy may be captured differently in routine records than in a trial.
  • Missing data. Sponsors should assess missingness in the external source before the trial, as a feasibility check.

What does FDA expect sponsors to submit?

FDA expects a protocol and statistical analysis plan locked before the trial starts, design decisions made without looking at the external data, and patient level data for both arms in the marketing application, with access to source records.

The draft guidance says sponsors should finalize a study protocol before initiating the externally controlled trial and submit the statistical analysis plan with it. Design and analysis choices should be blinded to any observed external control data. Marketing applications must include the relevant patient level data for both arms, and FDA should be able to see the source documents behind the external arm. Sponsors should also document every data source they considered, not just the one they used, and consult the review division early.

Real world data sit inside the same logic. FDA’s December 2018 Framework for its Real World Evidence Program says external controls, such as historical controls, are a possible type of control arm in an adequate and well controlled study. FDA’s August 2023 guidance on using real world data and real world evidence for regulatory decisions says real world data can serve as a comparator arm in an externally controlled trial, and treats such a trial as an interventional study generally subject to the investigational drug rules in 21 CFR part 312.

Which FDA approvals used external controls?

The cleanest examples are ultra rare diseases with lethal or steadily progressive courses. Three approvals show the pattern: Brineura in 2017, Zokinvy in 2020 and Lembaz, approved November 17, 2026.

Drug Disease Treated patients External comparator Result
Lembaz (adrabetadex), Mandos LLC, a Beren Therapeutics subsidiary, 2026 Niemann-Pick disease type C with neurological symptoms before age 6 36, across three studies and expanded access Matched external control group 55% reduction in risk of death; median survival 8.5 vs 4.9 years
Zokinvy (lonafarnib), Eiger BioPharmaceuticals, 2020 Hutchinson-Gilford progeria syndrome 62 Untreated patients in a natural history cohort, matched 1:1 on mutation, sex and continent Death HR 0.30 at 3 years; 0.40 at last follow up
Brineura (cerliponase alfa), BioMarin, 2017 CLN2 disease (TPP1 deficiency) 24 in a single arm study Natural history cohort of 42 untreated patients; matched on age, genotype and baseline motor score At week 96, 16 of 17 matched treated patients had no motor decline vs 6 of 17 controls

Lembaz is the newest. FDA approved it to reduce the risk of death in adults and children weighing at least 7 kg whose neurological symptoms began before age 6. FDA said patients treated with Lembaz lived longer than a matched external control group, and the benefit held whether or not patients also took miglustat. The label carries a boxed warning for hearing impairment, including permanent hearing loss. FDA’s announcement does not describe how the control group was assembled, so the review documents will be the place to see how the matching held up.

Zokinvy shows the honest arithmetic of these designs. At three years, 5 of 62 treated patients (8.1%) had died versus 12 of 62 untreated matches (19.4%). The label translates that into an average lifespan gain of about 3 months over the first three years and 2.5 years through the last follow up, which ran up to 11 years.

How are externally controlled trials different from single arm trials?

Every externally controlled trial has a single treated arm, but not every single arm trial needs an external control. In oncology, FDA often grants accelerated approval on response rate from a single arm study because tumors rarely shrink on their own.

That is why the draft guidance uses tumor shrinkage as an example of an outcome where the natural course is well understood. The response rate itself carries the comparison. An external control arm earns its keep when the endpoint is something like survival or a functional milestone, where the reader needs to know what would have happened without the drug. For how FDA weighs response rate, progression free survival and overall survival, see our endpoints guide.

Our framework: a sponsor checklist before choosing an external control

  1. Test the disease. Can you show, with published or registry data, that the untreated course is predictable enough that a large effect would stand out?
  2. Test the endpoint. Is it recorded the same way in routine care as in your protocol? Death and dated milestones pass; rating scales usually do not.
  3. Secure patient level data. Get contractual access to individual records and source documents before you commit. Summary data will not satisfy FDA.
  4. Lock the plan first. Finalize the protocol and statistical analysis plan, including the matching method and index date, before anyone looks at outcomes in the external set.
  5. Document the search. List every candidate data source and why you rejected it, so reviewers cannot suspect a favorable pick.
  6. Plan sensitivity analyses. Show how much unmeasured confounding it would take to erase the effect.
  7. Meet the division early. Bring the design to a pre IND or end of phase meeting, and ask directly whether a randomized trial is feasible in FDA’s view.
  8. Budget for a no. If FDA says the effect is not large enough to read through bias, you need a randomized option ready.

What to watch next

Watch whether FDA finalizes the 2023 draft, how the Lembaz review documents describe the external control construction, and whether more rare disease sponsors file on natural history comparators. Finalization would be the first formal signal on how the agency weighs comments it has collected since 2023.

Frequently asked questions

What is an external control arm?

An external control arm is a comparison group drawn from outside the trial, such as patients in a natural history study, a registry or an earlier clinical trial, used in place of a randomized control.

What is the difference between a historical control and a concurrent external control?

A historical control comes from patients treated or untreated at an earlier time. A concurrent external control comes from the same time period but a different setting, such as a registry running alongside the trial.

Is FDA’s externally controlled trials guidance final?

No. The February 2023 guidance is still a draft as of November 2026, under docket FDA-2022-D-2983, issued by CDER, CBER and the Oncology Center of Excellence.

When does FDA accept a natural history control?

Mostly when the disease course is well documented and predictable, the expected effect is large and the endpoint is objective, such as survival in a lethal rare disease.

Can real world data be used as a control arm?

Yes. FDA’s August 2023 guidance on real world data says it can serve as a comparator arm in an externally controlled trial, provided the data are fit for purpose and the design controls bias.

What is the biggest weakness of external controls?

Bias. Without randomization and blinding, treated and external patients can differ in ways nobody measured. ICH E10 calls the inability to control bias the major limitation.

What did FDA approve Lembaz on?

FDA approved Lembaz (adrabetadex) on November 17, 2026 based on survival in 36 treated patients compared with a matched external control group: a 55% reduction in risk of death and median survival of 8.5 versus 4.9 years.

Does a single arm cancer trial need an external control?

Not usually. FDA often accepts response rate from a single arm trial for accelerated approval because tumors rarely shrink on their own, so the endpoint carries its own comparison.

Do sponsors have to give FDA patient level data for the external arm?

Yes. The draft guidance says marketing applications must include relevant patient level data for both arms, and FDA should be able to access source records for the external data.

What regulation covers historical controls?

21 CFR 314.126(b)(2)(v) lists historical control as an acceptable control type in adequate and well controlled studies, while noting such designs are usually reserved for special circumstances.

Sources

Related coverage

Get the daily briefing: subscribe to BioMed Nexus.

Featured Articles

Join 65,000+ Biotech, MedTech, and Pharma Leaders

Your Daily Edge in Biotech, MedTech, and Pharma

Get trusted, high-signal updates every morning
Breakthroughs, trial data, deals, and the news that matters