External control arms can make clinical trials feasible in rare and serious conditions, but the approach comes with operational, statistical and regulatory challenges. At a Friends of Cancer Research workshop this week, stakeholders discussed the output of a comparative pilot project and approaches to tackle issues such as data missingness.
Background and context: External control arms
- A product seeking FDA approval is expected, by statue, to present “substantial evidence” to demonstrate that it is safe and effective for the intended use. This evidence, according to the regulations, consists of “adequate and well-controlled investigations, including clinical investigations.” Traditional clinical trials generally satisfy the “well-controlled” criterion by randomizing patients to receive the investigational treatment or a comparator, such as a placebo or standard-of-care regimen.
- Clinical trial design and enrollment are challenging for products intended for small populations, such as certain cancers and other rare diseases. Sponsors can face difficulty identifying enough patients to populate both an investigational group and a control group that will satisfy statistical considerations. In other cases, patients may not wish to enroll in a study if there is a chance they would not benefit, as would be the case if they might receive a placebo. It may also be unethical to operate a control arm in some cases, as when there is no currently marketed treatment option for a life-threatening disease.
- The FDA accepts single-arm trials in certain scenarios, though interpretation of results is limited by risk of bias. Still, the response rate endpoint, measured as tumor shrinkage or disappearance, has supported oncology approvals in the single-arm trial context. The agency’s September 2025 draft guidance on innovative clinical trial designs for cell and gene therapies in small populations endorsed a single-arm approach that uses “an internal, baseline control strategy where a participant’s response to therapy on a given measure is compared to their baseline status on that measure.”
- To address limitations, some sponsors use external control arms. In an externally controlled trial, the control group is made up of people not enrolled in the trial whose data have been captured in some other way – through a registry or through participation in another, unrelated study, for example. These patients did not receive the investigational treatment being studied and may therefore be used as the control group for the new study, permitting investigators to conduct a trial where all qualified participants are enrolled into the investigational arm of the study, as appropriate.
- In 2001, the FDA adopted the International Council for Harmonisation’s E10 scientific guideline on the choice of control group and related issues in clinical trials. The guidance has several sections focused on ECAs. One key concept raised by the E10 guidance was the need to minimize bias in the selection of external controls. To quote the guideline at length: “A control group should be chosen for which there is detailed information, including, where pertinent, individual patient data regarding demographics, baseline status, concomitant therapy, and course on study. The control patients should be as similar as possible to the population expected to receive the test drug in the study and should have been treated in a similar setting and in a similar manner, except with respect to the study therapy. Study observations should use timing and methodology similar to those used in the control patients.”
- In 2022, a group of FDA oncology regulators (including then-Oncology Center of Excellence director RICK PAZDUR), penned a review in ESMO Annals of Oncology describing current use and future directions for ECAs in oncology. The authors explained that data from external sources are often employed to inform study design. They wrote, “However, with increasing clinical studies in specific disease subpopulation areas and availability of vast amounts of [electronic health record] data, there is a renewed interest in using data external to a clinical trial for comparative efficacy and safety analyses. Additionally, the 21st Century Cures Act has led the FDA to develop a framework for evaluating the quality and relevance, also referred to as fit for purpose, of real-world evidence, such that it may be incorporated into regulatory decision making.”
- The FDA subsequently published a draft guidance exclusively focused on ECAs in 2023. The guidance centers on using patient-level data from other clinical trials or from real-world data sources, like registries, electronic health records and medical claims. The agency cautions that the “likelihood of credibly demonstrating” the effectiveness of a drug with an ECA is low, and in many situations, sponsors should use other trial designs that are “more suitable.” External controls should be limited to situations where it otherwise would be impossible to generate evidence “capable of distinguishing the effect of the drug from outcomes attributable to the disease’s natural history, prognostic differences in the study populations, knowledge of treatment assignment (lack of blinding), or other factors such as differences in concomitant therapies.”
- The guidance emphasized that the “suitability of an externally controlled trial design” will be on a case-by-case basis and will depend on a few key factors: the heterogeneity of the disease, preliminary evidence regarding the drug product under investigation, the approach to ascertaining the outcome of interest, and whether the goal of the trial is to show superiority or non-inferiority. If a condition is well-defined and the disease will not improve on its own, then FDA may also accept historical information. [Read full AgencyIQ analysis of the document here.] The document, which remains in draft form, received substantial input from stakeholders.
This week, stakeholders met to dive into the application of ECA in oncology
- Titled “Application of External Control Arms in Oncology Drug Development,” the hybrid event from Friends of Cancer Research, or “Friends,” sought to explore ECAs in oncology, “particularly when traditional randomized trials are impractical,”
- The event’s opening keynote from Highlander Health’s AMY ABERNETHY drew on her experiences in industry, academia and as FDA’s principal deputy commissioner from 2019 to 2021. She spoke to the evolution of thinking around ECAs, which she said began as a “rudimentary attempt to replicate the eligibility criteria of a randomized control trial.” Looking ahead, Abernethy said, “I think one of the things that we’ve still gotten ourselves stuck in is real world data – real world evidence versus clinical trials.” She continued, “Where this story goes is modern evidence generation – how does the whole story come together? A combination of formalized, prospective clinical trials, with or without randomization, and real-world data studies such as external control arms, creating the complete evidence package. And when we start to think about that totality construct, including the multiple ways to get there, we’re going to be in a much faster place.”
- The workshop provided a status update on the ongoing ECA pilot project, launched in October 2025. The project’s goal is to “collaboratively define a robust process for ECA construction using patient-level RWD and prior clinical trial data to evaluate its applicability for replicating the population and efficacy results observed in target trial data.” The project relies on a coalition of multiple data partners that independently constructed ECAs using a shared statistical analysis plan.
- The pilot centers around one disease area, metastatic pancreatic ductal adenocarcinoma, or mPDAC, with the aim to compare the constructed ECAs to the well-characterized control arm of the completed Phase 3 RESOLVE trial. In that study, patients were randomized to receive the investigational drug or placebo in combination with the standard of care first-line chemotherapy regimen of gemcitabine and nab-paclitaxel. Friends Senior Science Policy Analyst BERNAT NAVARRO explained that the pilot project working group translated the RESOLVE inclusion and exclusion criteria into “definitions that could be applied across external data sources, distinguishing between both real-world data and historical clinical trial data sets.”
- The working group participants explained why they chose mPDAC to explore ECAs. From a practical standpoint, the mPDAC setting offered data availability from the RESOLVE trial, which had relatively straightforward inclusion/exclusion criteria, with no specific biomarkers used for selection. In addition, multiple external partners had relevant, patient-level data. The Pancreatic Cancer Action Network’s CASSADIE MORAVEK described unmet need for the relatively rare cancer, which has an annual incidence of about 67,000 and a five-year survival rate of approximately 13%. Ontada’s JANET ESPIRITO explained how this need translated into feasibility of the use case. “Because [OS] was relatively short, we were able to reasonably assess overall survival versus requiring data sets that required much longer longitudinal follow-up for endpoints that take longer to assess.”
- In advance of the workshop, Friends released a brief overview of the project’s preliminary findings. During the workshop, Navarro provided a deeper look at comparative results from the six data partners that could apply the primary analysis approach. The presentation offered side-by-side comparisons between the RESOLVE control arm and the constructed ECAs across clinical eligibility criteria, lab eligibility criteria, and demographic characteristics that allow for comparison of baseline balance and attrition. Navarro explained, “We use a common [statistical analysis plan] to ensure methodological consistency, but there was flexibility in data set specific choices. This reflects the realities of working with heterogeneous external data sources, and it’s important to document those choices in order to contextualize these results.”
- Each of the six constructed ECAs was used to generate an OS estimation and Kaplan-Meier curve. According to Navarro, “a preliminary takeaway is that alignment with the target trial is closely tied to data fitness, including availability of key prognostic variables, cohort size and ability to achieve peripheral balance.” Panelist RUTHIE DAVI, representing Medidata, came to a similar conclusion, saying “for me, the biggest finding is not a surprising one, and that is that you need complete – or nearly complete – baseline information about patients in order to fully evaluate the eligibility criteria and balance the external control with the randomized control.”
- Handling missing data remains a primary challenge in constructing ECAs, ConcertAI’s JENNIFER RIDER said, “We know that if we just blindly exclude patients from the ECA because they are missing information on eligibility criteria, that not only will lead to unviable sample size – you can start with a very, very large real world data source, but by the time you exclude patients for that missingness, you end up with a very modest number of patients – but there are also a lot of concerns about selection bias with that approach.” Following a comment on data extraction, AstraZeneca’s BRAD KARALIUS said, “I think here’s a real opportunity for AI, where we have natural language processing algorithms that can abstract from unstructured data. Of course, that work needs to be validated, but I think that’s a that’s a great use case for these contemporary tools.”
- In terms of next steps, Navarro said that the working group could conduct additional sensitivity analyses “to better distinguish the effects of data, source differences, missingness, sample characteristics and analytic methods.” Karalius echoed this call, noting the importance of these analyses given the heterogeneity in data sources. He said the group should “dig into the data of each data partner.” He challenged attendees to think critically about approaches to resolve heterogeneity and missingness problems, asking “What if we did matching instead of weighting? Would that change anything? What if it was missing this covariate which is present? Would that change anything?”
Broader implementation and regulatory considerations came under review in two subsequent panels
- Across the board, participants identified data missingness as a key hurdle for ECAs. Representing J&J Innovative Medicine, ASHITA BATAVIA observed “a really interesting connection between the missingness of the data – and the less missing data you have, perhaps you have a little bit more flex with how creative on the methodology side you need to be.”
- Patient advocate JANE PERLMUTTER pointed out an area in oncology constituting more of a data desert: real-world evidence generally does not include patient reported outcomes. She said that “increasingly, as we have more and more drugs available to patients, knowing about the tolerability of drugs is really important to patients.” In the following panel, AMY COMSTOCK RICK, head of the FDA’s Rare Disease Innovation Hub, described this issue as a “tension,” explaining that rare disease stakeholders have increasing interest and pressure to conduct studies using endpoints “based on patient preference, as opposed to traditional endpoints.” She noted that “some of those measures may not have been measured in your historical studies and your registries.”
- Data access and use agreements are a driver of operational friction. The FDA’s MARIE BRADLEY, Senior Advisor for RWE in CDER’s Office of Medical Policy, emphasized that the agency requires patient-level data for ECAs that leverage real-world data. “While that might seem simple, it’s very far from it,” she said, “Often the sponsors who conduct the studies don’t own the data. Even the third-party vendor needs to get permission from the data owner and there are concerns with data use agreements and what we can and can’t assign.” Later during the panel, IQVIA’s JACLYN BOSCO described the contracting phase for third-party data as “one of the biggest, lengthy parts of doing an ECA.” Bradley encouraged early engagement to ensure the agency will be able to reanalyze the data during review. “We’re working around the solutions. Hopefully you’ll hear more about that soon,” she said.
- The FDA’s Rick also emphasized the importance of sponsor engagement, commenting on the difficulty in providing desired “certainty” around regulatory flexibilities. She explained that the agency can “give more guidance on natural history studies and even what maybe should be collected in control arms for other trials, so that the data can be reused.” Rick also underscored a point she often makes in public-facing meetings. “I think in this era of innovation and evolution, sponsors need to be particularly careful about assuming that they know what FDA’s answer is going to be. FDA is evolving, just like any other good scientific organization does, and just because it was an answer you or your colleague got five years ago,” she said, “doesn’t necessarily mean it’ll be the answer now. And so please don’t assume that you know the answer to a question you haven’t just asked.”
- Looking ahead, Bradley said the FDA is prioritizing hybrid externally controlled trials. According to Bradley, the agency is currently funding “several complementary demonstration research projects” on this methodology, which addresses challenges by having a “randomization anchor” that can increase internal validity and enable adaptive borrowing approaches. While she emphasized the agency’s interest, she pointed out that the results of the demonstration projects will inform decision-making around the need for additional guidance.
- From a more philosophical perspective, the discussion addressed how contemporaneous ECAs could improve clinical trial design. J&J’s Batavia pointed out the potential utility of this approach in multiple myeloma. “Some of those trials, when you enroll a randomized controlled trial, the standard of care at the point of enrollment is rarely the standard of care at the end of the trial, because we’ve made such tremendous progress in extending lives in populations, and there’s so much rich innovation in that space,” she said, continuing, “So, when you read out a trial having that real-world control arm to help contextualize, what is this – in the future time – contemporaneous standard of care for comparison, can also be a really interesting use case and application.”
Analysis
- At a high level, the meeting provided a forum for general discussion of a topic that, in practice, is highly context-specific. The ECA Pilot Program represents an effort to pull common methodological threads to increase understanding and facilitate implementation of ECAs. While the pilot’s preliminary results revealed several areas that warrant additional sensitivity analyses, it’s worth noting that the output comes just six months after the project’s formal launch.
- ECA acceptability for regulatory decision-making has been the source of recent controversy, though this situation was not discussed during the workshop. In March, uniQure announced that FDA feedback will likely substantially delay its Huntington’s disease treatment’s chance at a regulatory review. The agency asked uniQure to run a randomized, controlled study of its gene therapy for Huntington’s disease, arguing that recent Phase 1/2 data comparing the experimental treatment to an external control insufficiently demonstrated effectiveness. UniQure claimed this demand is a reversal of earlier agreement between it and the FDA on what supporting evidence would be necessary for an approval [Read AgencyIQ analysis here]. Controversy over the agency’s handling of that application appears to have played some role in VINAY PRASAD’s coming departure as director of the FDA’s biologics center, which oversees cell and gene therapies such as uniQure’s. [Read AgencyIQ’s analysis of Prasad’s exit here.]
- The FDA’s publication of Complete Response Letters offers stakeholders additional insight into cases where the agency found an ECA to be insufficient. AgencyIQ’s original analysis of the letters released through the openFDA repository has found the agency redacted varying levels of information regarding the key approvability issues associated with an application, with clinical and statistical deficiencies relatively lightly redacted. Through its transparency in revealing assessment of these issues, the FDA has disclosed its thinking on when it questions validity or finds other concerns with an external control arm. Examples include this May 2025 letter to Stealth BioTherapeutics and this August 2025 letter to Saol Therapeutics. On the other hand, the agency expressed openness to an externally controlled trial as a means to address identified deficiencies in this December 2024 letter to Lexicon Pharmaceuticals.
Featuring previous research by Alexander Gaffney and Sebastian Godoy.
To contact the author of this item, please email Amanda Conti ( aconti@agencyiq.com).
To contact the editor of this item, please email Kari Oakes ( koakes@agencyiq.com).