cutout figures with magnifying glass
Credit: Getty Images

Insufficient participant pools represent a key barrier to conducting randomized controlled trials (RCTs) of potential treatments for rare diseases. Even when researchers are able to recruit an adequate number of participants to receive a candidate therapy, there may be a dearth of additional patients with the same disease to comprise a control group.

An accumulating body of research points to a promising solution to this challenge: the use of synthetic control arms based on patient data from previous RCTs or observational, real-world data, for example from patient registries.1 Also referred to as external control arms, patients in these groups are carefully selected to match the characteristics of patients in a randomized treatment group.

In a case study published in 2020, Davi et al explored the validity of this approach for research investigating treatments for relapsed and refractory multiple myeloma (RR/MM). Noting that a synthetic control arm could be used to “augment a single-arm trial or a RCT compromised by control arm early withdrawal or noncompliance,” they examined the treatment effect of using a synthetic control arm based on patient-level data from previous RR/MM trials compared to the treatment effect of an actual RCT.2

Their results demonstrated a similar treatment effect between both approaches; the hazard ratio [HR] in the original RCT was 0.743 (95% CI, 0.60-0.92), while the HR based on the use of a synthetic control arm matched to the randomized treatment group was 0.758 (95% CI, 0.63-0.91).

In a study published in June 2022 in Nature Communications, Popat et al described the challenges involved in conducting well-powered RCTs for advanced non-small cell lung cancer (aNSCLC), which is “being increasingly divided into rare oncogene-driven subsets.” To address this issue, they constructed a synthetic control arm matched to participants from the single-arm ARROW trial (NCT03037385) that demonstrated the efficacy of the drug pralsetinib for the treatment of RET fusion-positive aNSCLC.3

Using real-world data cohorts that received pembrolizumab monotherapy and pembrolizumab with chemotherapy, they compared the effectiveness of these approaches to the outcomes observed with pralsetinib monotherapy in the ARROW trial. Their findings showed a relative survival benefit with pralsetinib as first-line therapy compared to each of the therapies used in the real-world data cohorts.

Emerging research has also explored the use of synthetic control arms in studies of other diseases including diffuse large B-cell lymphoma, colon cancer, and rheumatoid arthritis, all with favorable results.4,5,6

Below, Rahul Banerjee, MD, FACP, discusses the benefits and drawbacks of using synthetic control arms. Dr. Banerjee is assistant professor in the division of medical oncology at the University of Washington (UW) in Seattle and in the clinical research division at the Fred Hutchinson Cancer Research Center at UW Medicine. He is the first author of a 2021 review that examined the use of this approach in studies of therapies for hematologic malignancies.4

Rahul Banerjee, MD, FACP (Photo courtesy of Rahul Banerjee)

Q: What does the use of a synthetic control arm involve, and how does a trial using this approach differ from a traditional trial design?

A: In a traditional RCT, patients from the population of interest are randomly assigned after study enrollment to receive either the experimental drug or the control intervention. Because of this randomization, RCTs are not generally affected by confounding variables or selection bias.

In contrast, with single-arm trials in which every patient receives the experimental drug, a “synthetic control arm” entails the use of observational or precollected data from a similar cohort of patients to simulate how the control arm of a randomized trial might have behaved. Synthetic control arms can either be created prospectively—ie, collecting data in real time from similar patients who are not enrolled in the trial itself but are involved with a separate observational database—or assembled using previously collected retrospective data.

Q: What initially prompted the development of this approach, and what are examples of diseases that have benefited thus far?

A: Many ethical and logistical considerations go into the feasibility of an RCT: rarity of the disease, ethics of assigning patients to the control arm if the experimental drug is likely to work, and of course, the practical point that a 2-arm trial generally requires twice as many enrolled patients as a single-arm trial, with accompanying financial and accrual considerations.

Synthetic control arms seek to bypass this by allowing all enrolled patients to receive the experimental drug. For example, in 2017 the [US Food and Drug Administration] granted approval to cerliponase alfa to treat a rare pediatric disease called Batten disease based on a single-arm study of 22 patients who received the drug. These patients were then compared against a “natural history cohort” of 42 patients, which effectively constituted a synthetic control arm.7

Q: What are the potential benefits and limitations of this approach for rare disease research?

A: As highlighted above, Batten disease is a rare genetic disorder of children with no other known treatments apart from cerliponase alfa. The use of a synthetic control arm here allowed for all 22 enrolled patients to receive this drug. This disorder is so rare that, according to the accompanying publication about this study, 2 years of collaboration across 4 countries were required to identify just 22 patients.8

Had the investigators tried to enroll twice as many patients for a randomized study employing a control arm, there is no doubt that this trial would have taken much longer to be completed—and therefore, much longer for this drug to reach patients more broadly. And, of course, if such a randomized study wasn’t fully double-blinded, many patients in the control arm would likely have dropped out after realizing that they were in the control arm.

That being said, synthetic control arms must be approached with much more caution in diseases that aren’t as rare or for diseases where new treatments are emerging over time. In our analysis of synthetic control arms in lymphoma and myeloma, for example, we found that several studies employed a historical synthetic control with data from patients treated as many as 14 years previously.4 Many advances in cancer treatment have emerged during that timeframe, and even meticulous matching for baseline demographics and disease characteristics cannot correct the fact that patients diagnosed more recently will generally live longer.

Another limitation for nongenetic diseases is some heterogeneity in how the specific patient population is identified. For example, we identified 1 study where patients in a single-arm study of relapsed/refractory multiple myeloma were allowed to enroll if they had intolerance to previous therapies rather than, strictly speaking, only disease that had become refractory to previous therapies. The synthetic control arm had a stricter definition of refractoriness and thus potentially might have enrolled patients with slightly more aggressive disease overall. These conclusions are speculative, of course, but both types of differences might have inflated the margins by which a new treatment outperformed a synthetic control arm.

Q: What is needed to further develop and refine this research design?

A: There is no doubt that synthetic control arms are here to stay, particularly as the field of medicine—and in particular oncology—moves increasingly toward single-arm trials and rich individual-patient-level datasets. Several best practices can be used to improve the accuracy and generalizability of synthetic controls. Indeed, a quality checklist for studies using synthetic control arms has been proposed.9

In our analysis, we did identify an example of a synthetic control arm that was created specifically to be paired with a single-arm study of a new therapy in multiple myeloma.10 Rather than reusing the same historical data set across multiple studies, which raises the risk of false-positive signals as well as the issues highlighted above, the synthetic control arm used here was guaranteed to match the interventional study’s patient population in terms of eligibility criteria and treatment timeframes.

This approach, of course, still doesn’t replicate the rigor of an RCT—nor does it allow for patient-reported outcomes and other important endpoints that only a unified 2-arm trial can collect—but it does enhance the reliability of comparisons made with the synthetic control arm.


  1. Mishra-Kalyani PS, Amiri Kordestani L, Rivera DR, et al. External control arms in oncology: current use and future directions. Ann Oncol. 2022;33(4):376-383. doi:10.1016/j.annonc.2021.12.015
  2. Davi R, Yin X, Stewart M. Exploring the validity of a synthetic control arm (SCA) for augmentation or replacement of a randomized control in difficult-to-study indications: a case study in relapsed or refractory multiple myeloma (R/R MM). J Clin Oncol. 2020;38(15_suppl):e20521-e20521
  3. Popat S, Liu SV, Scheuer N, et al. Addressing challenges with real-world synthetic control arms to demonstrate the comparative effectiveness of Pralsetinib in non-small cell lung cancer. Nat Commun. 2022;13(1):3500. doi:10.1038/s41467-022-30908-1
  4. Banerjee R, Midha S, Kelkar AH, Goodman A, Prasad V, Mohyuddin GR. Synthetic control arms in studies of multiple myeloma and diffuse large B-cell lymphoma. Br J Haematol. 2022;196(5):1274-1277. doi:10.1111/bjh.17945
  5. Azizi Z, Zheng C, Mosquera L, Pilote L, El Emam K; GOING-FWD Collaborators. Can synthetic data be a proxy for real clinical trial data? a validation study. BMJ Open. 2021;11(4):e043497. doi:10.1136/bmjopen-2020-043497
  6. Wang Z, Yu Z, Chen S, Zhang L. Investigating synthetic controls with randomized clinical trial data in rheumatoid arthritis studies. J Clin Trials. 2021;11:466. doi:10.35248/2167-0870.21.11.466
  7. US Food and Drug Administration. FDA approves first treatment for a form of Batten disease. Published April 17, 2017. Accessed September 11, 2022.
  8. Schulz A, Ajayi T, Specchio N, et al; CLN2 Study Group. Study of intraventricular cerliponase alfa for CLN2 disease. N Engl J Med. 2018;378(20):1898-1907. doi:10.1056/NEJMoa1712649
  9. Thorlund K, Dron L, Park JJH, Mills EJ. Synthetic and external controls in clinical trials – a primer for researchers. Clin Epidemiol. 2020;12:457-467. doi:10.2147/CLEP.S242097
  10. Jagannath S, Lin Y, Goldschmidt H, et al. KarMMa-RW: comparison of idecabtagene vicleucel with real-world outcomes in relapsed and refractory multiple myeloma. Blood Cancer J. 2021;11(6):116. doi:10.1038/s41408-021-00507-2