Synthetic Control Arms: What If a Clinical Trial Didn't Need a Placebo Group?
Imagine developing a drug for a rare disease that affects only 50 children worldwide. A traditional trial would randomize half of them to placebo. Twenty-five children with a fatal condition would get no treatment. That is ethically impossible. But what if the comparison group could be built from historical data instead, from patients with the same disease whose records already exist? No new placebo arm, no untreated children.
That is the promise of synthetic control arms, and the Clarity Clinical Solutions video makes the case that it is one of the most important innovations in modern trial design. A synthetic control arm is a comparator group constructed from existing patient data rather than by enrolling patients into a separate control arm. Everyone in the trial gets the active treatment. The sponsor uses historical data to estimate what would have happened to those patients without it.
The difference from a randomized trial
A traditional randomized controlled trial enrolls patients into two groups: treatment and placebo or standard of care. Random assignment balances known and unknown confounders between the groups. That is why the RCT remains the gold standard.
A synthetic control arm trial is fundamentally different. All patients receive the active treatment. The comparator is built from historical data. There is no concurrent randomization. The upside is real: the sponsor does not need to recruit control patients, which can cut enrollment requirements by up to 50% and save an estimated $10 to $50 million per study. The downside is equally real: the comparison is observational, not randomized, which introduces the risk of bias from unmeasured differences between historical patients and trial patients.
Where the data comes from
There are three main sources. The first is individual patient data from prior clinical trials, sometimes the sponsor's own completed or failed studies. This is the highest quality source because outcomes were assessed under controlled, standardized conditions. The second is electronic health records from hospitals and health systems. These offer large volumes of real-world data but need careful curation to handle missing data, variable documentation, and non-standardized outcome definitions. The third is natural history registries, which track how a disease progresses without treatment. These are essential in rare diseases where the untreated course of the illness is well documented.
Making the comparison credible
Building a valid synthetic control arm is a serious statistical exercise. The core challenge is making historical patients comparable to trial patients. The most common tool is propensity score matching, which calculates the probability that each patient would have been assigned to the treatment group based on baseline characteristics like age, disease severity, and prior treatments. Each treated patient is then matched with historical patients who have a similar propensity score. Bayesian approaches treat the historical data as an informative prior that is updated as trial data accumulates. Prognostic score methods stratify patients by predicted outcome and compare within strata. All of these methods only work when the important confounders were actually measured.
The advantages
Synthetic control arms reduce the number of patients needed by 30 to 50%, according to the video. They make trials faster because only the treatment arm needs to be enrolled, cutting enrollment time by 40 to 60% and potentially shortening development timelines by one to two years. They save tens of millions of dollars per phase 3 study. And they are more ethical in the settings where they are used: all patients receive the active treatment, avoiding the ethical problem of assigning placebo when effective treatment might exist.
The limits
The limitations deserve equal attention. The most fundamental is that synthetic controls cannot adjust for unmeasured confounders, factors that differ between historical and trial patients but were never recorded. Temporal drift is another problem. If standard of care has improved since the historical data was collected, historical patients will look worse than today's patients would, making the treatment look better than it really is. Data quality from non-trial sources like electronic health records is often incomplete or inconsistent. These are real constraints, and the FDA evaluates each submission on a case-by-case basis.
The regulatory picture
The FDA has become increasingly open to synthetic controls, particularly in rare disease and oncology. Its 2023 guidance on real-world data and real-world evidence describes standards for using non-interventional data to support regulatory decision-making (FDA: Considerations for the Use of Real-World Data and Real-World Evidence to Support Regulatory Decision-Making). The agency's real-world evidence program page tracks this work and the frameworks behind it (FDA: Real-World Evidence Program). External control data can serve at different levels of evidentiary support: exploratory, supportive, or as the primary basis for approval. The bar is highest when the synthetic control is the sole comparator. Then the sponsor needs strong justification, well-characterized natural history, and a large treatment effect. The video reports that as of 2024, more than 40 FDA submissions included external control data, over 25 drugs were approved using synthetic control arms as part of their evidence package, and the number of external control submissions has tripled since 2019.
The landmark example
The most celebrated case is blinatumomab, a treatment for a rare and aggressive form of acute lymphoblastic leukemia in children. The trial that supported the approval was single-arm. All patients received the drug. The control was constructed from over 1,200 historical patients who had received standard chemotherapy. The drug achieved a complete remission rate of about 32% compared with roughly 12% in the historical control. The FDA granted accelerated approval in 2014, a watershed moment that showed the agency would accept synthetic controls when the disease has a well-understood poor prognosis and the treatment effect is large (NCI: FDA Approval for Blinatumomab).
In rare diseases generally, synthetic controls have been transformative. When a condition affects only a few hundred patients worldwide, there are not enough patients to randomize, and when the disease is fatal with no approved treatment, giving placebo is unethical. The FDA has approved roughly 20 rare disease drugs using synthetic control arms as the primary evidence of efficacy, including treatments for fatal childhood neurodegenerative diseases where the control came from natural history registries.
What comes next
Three developments could expand the role of synthetic controls. Generative models may produce synthetic patient data that preserves the statistical properties of real trial data while protecting privacy. Hybrid designs that combine a small internal randomized control arm with synthetic data augmentation could offer the credibility of randomization with the efficiency of external data. And the FDA is developing formal qualification pathways for synthetic control databases and methods, which would reduce regulatory uncertainty once a data source or methodology is pre-approved.
The bottom line
Synthetic control arms do not replace the randomized controlled trial. The RCT remains the standard for establishing efficacy. But synthetic controls expand what we can study when randomization is impossible or unethical. For children with fatal genetic diseases, for patients with ultra-rare cancers, for any condition where a placebo arm is not an option, they offer a path forward that barely existed a decade ago. They are not perfect, and their limitations must be understood. But they are bringing treatments to patients who otherwise would have no trials to join.
This article is based on the Clarity Clinical Solutions video "Synthetic Control Arms —What If a Trial Didn't Need a Placebo Group?" Watch it here: Synthetic Control Arms —What If a Trial Didn't Need a Placebo Group?
References
- Clarity Clinical Solutions — "Synthetic Control Arms —What If a Trial Didn't Need a Placebo Group?" (framework for this article). https://www.youtube.com/watch?v=9kC4U8Q9ZPQ
- FDA — Guidance: Considerations for the Use of Real-World Data and Real-World Evidence to Support Regulatory Decision-Making for Drug and Biological Products (December 2023). https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-use-real-world-data-and-real-world-evidence-support-regulatory-decision-making-drug
- FDA — Real-World Evidence Program; framework for evaluating real-world data in regulatory decisions. https://www.fda.gov/science-research/science-and-research-special-topics/real-world-evidence
- National Cancer Institute (NIH) — FDA approval information for blinatumomab. https://www.cancer.gov/about-cancer/treatment/drugs/blinatumomab