External Control Arms in Single-Arm Trials: Leveraging Real-World Data for Regulatory Submissions

In therapeutic areas defined by small patient populations — rare genetic disorders, orphan oncology indications, pediatric conditions — a traditional randomized controlled trial (RCT) is frequently impossible. Enrolling enough patients into both a treatment and a placebo arm may take years, and withholding a promising therapy from half of an already scarce patient population raises serious ethical concerns. In these settings, sponsors increasingly turn to External Control Arms (ECA), a design in which every enrolled patient receives the investigational treatment, and the comparator is drawn entirely from external data sources.

Regulatory acceptance of ECA-based submissions has expanded substantially over the past several years, driven by the FDA's Real-World Evidence (RWE) Program and comparable EMA guidance on the use of registries and historical data. However, an External Control Arm is not a shortcut around rigorous causal inference. It requires the same discipline as any observational comparison, arguably more, because a single-arm trial affords no internal check on selection bias. This article outlines the methodological framework required to build a defensible ECA that will withstand regulatory and peer-review scrutiny.

1. Why Single-Arm Trials Need External Controls

A single-arm trial without any comparator can only report response rates or survival estimates in absolute terms. Reviewers and clinicians need to know whether that outcome is better than what would have occurred under standard-of-care. Historically, sponsors relied on "objective performance criteria" drawn loosely from the literature, but these comparisons often fail scientific and regulatory scrutiny because they lack transparency about the source population, era effects, and covariate distributions.

An ECA formalizes this comparison by identifying a specific, well-characterized external cohort — often from a disease registry, insurance claims database, or the control arm of a prior completed trial — and applying causal inference methods to make that cohort comparable to the treated population. The goal is to approximate, as closely as observational data allows, what a concurrent randomized control arm would have shown.

2. Selecting the Right External Data Source

The credibility of an ECA begins with the data source. Regulators evaluate three primary types of external control data, each with distinct strengths and limitations.

• Historical Clinical Trial Data: Patient-level data from the control arm of a previously completed RCT in the same indication. This offers the highest data quality and standardized outcome ascertainment but may suffer from era effects if standard-of-care has since evolved.
• Disease Registries: Prospectively maintained, condition-specific databases that often include rich longitudinal clinical detail. Registries are well suited to rare diseases where dedicated infrastructure already exists for natural history documentation.
• Electronic Health Records / Claims Data: Broad real-world data with large sample sizes but frequently inconsistent outcome definitions, missing laboratory values, and coding-based rather than clinically confirmed diagnoses.

3. Evidence Summary Table

4. Balancing Cohorts: Propensity Scores and Beyond

Once an external cohort is identified, the central statistical task is constructing comparability with the treated group. This typically involves propensity score matching, propensity score weighting, or overlap weighting, all of which condition on the probability of receiving the investigational treatment given observed baseline covariates.

Because the external cohort was not collected under the same protocol as the trial, careful attention must be given to eligibility criteria alignment — applying the trial's inclusion and exclusion criteria to the external data before any balancing is performed. Failure to align eligibility first is one of the most common and most damaging errors in ECA construction, since it can introduce immortal time bias or systematically favor healthier external patients.

5. The Threat That Cannot Be Fully Eliminated: Unmeasured Confounding

Unlike a randomized trial, an ECA can only balance covariates that were actually measured and recorded consistently in both datasets. Unmeasured or poorly captured prognostic factors — functional status, disease severity nuances, or supportive care quality — can bias the treatment effect in either direction, and no statistical technique can fully correct for a variable that was never observed.

Because of this irreducible limitation, regulatory reviewers expect a structured quantitative bias analysis, often using an E-value or tipping-point sensitivity analysis, to characterize how strong an unmeasured confounder would need to be to explain away the observed treatment effect. A submission that omits this analysis is at substantially higher risk of regulatory rejection, regardless of how well the observed covariates were balanced.

6. Actionable Steps: Building a Defensible External Control Arm

7. Where ECA Has Delivered Regulatory Approvals

Several oncology and rare disease therapies have achieved marketing authorization on the strength of single-arm trials supported by well-constructed external control arms, particularly in settings with high unmet need and clear historical natural-history data. These precedents demonstrate that regulators are willing to accept ECA-based evidence when the methodological rigor matches what would be expected of a comparable observational study — not a diluted substitute for randomization, but a distinct and defensible causal inference framework in its own right.

Conclusion

External Control Arms extend the reach of clinical evidence into populations where randomization is not a realistic option, but they demand the same methodological discipline as any causal inference study — and an explicit acknowledgment of what cannot be corrected. Sponsors who invest early in eligibility alignment, transparent balancing procedures, and quantitative bias analysis substantially improve their odds of a successful regulatory review. As real-world data infrastructure continues to mature through 2026, External Control Arms are positioned to become an increasingly standard component of the evidence package for rare disease and oncology therapeutics.