Synthetic Control Arm in Clinical Studies & AI

15 February 2024
Lev Korolkov , Product Manager
Synthetic Control Arm in Clinical Studies & AI

While placebo-controlled randomized control trials (RCTs) continue to be considered the gold standard, single-arm trials have gained acceptance in certain contexts for evaluating new treatment interventions. More recently, there has been increased interest in innovative methods to address the challenges associated with traditional trial designs. Thus, the method of using external or synthetic control arms is especially relevant in situations where recruiting a sufficient number of participants is difficult, for example, in certain oncology indications, pediatric conditions and rare therapeutic indications. However, their inclusion in pivotal trials requires careful evaluation of factors such as data availability and comparability. 

The use of synthetic control arms can be particularly beneficial in Phase 2 trials for internal decision-making and may be especially relevant in rare diseases with small patient populations. It’s also useful in cases where determining an appropriate active control is difficult due to the lack of established treatments, leading to trials with insufficient statistical power or single-arm designs that make comparisons challenging. There's significant interest in receiving innovative treatments rather than being assigned to a placebo group among cancer patients. In randomized controlled trials, patients are typically assigned randomly to either an experimental or a control intervention arm, commonly comprising either a placebo or standard-of-care (SOC). What has been previously said is the use of placebos can complicate the process of recruitment of placebo-controlled trials due to patient reluctance. Many clinical trials for rare diseases involve small patient populations, resulting in insufficient statistical power, or are conducted as single-arm trials, making comparison with other therapeutic options challenging. Given the ethical concerns surrounding denying patients necessary therapy, researchers are exploring alternatives such as external controls or synthetic control arms. Synthetic control arms offer a solution to these difficulties by integrating external historical and real-world data into trials, providing a less risky and cost-effective approach for sponsors.


The use of Synthetic Control Arms in clinical trials offers various advantages. Using historical or real-world data or mixing data of enrolled patients with external data to replace or combine control arms can reduce the patient burden and free up resources, which increases the probability of trial success. To maintain scientific accuracy, appropriate statistical methodology is required to address the difference between RCT and SCA data, because non-randomised treatment comparisons can potentially be biased. Each clinical trial which uses synthetic control arms can have its own specific issues regarding the quality of data. The statistical methodology used to handle such issues depends on the complexity of matching it with external data. For more complex situations, multivariate regression and propensity score, Bayesian methods and ML algorithms can be implemented.

Senior Biostatistician at Therapyte 


While synthetic control arms are a relatively new concept, these designs have already been used in regulatory decision-making in both Europe and the United States. However, they cannot replace control arms in all trials, and their use requires careful consideration of multiple factors such as patient population matching, similarity of medical imaging protocols and standard care. The validity of using synthetic control groups can be heavily influenced by the application of AI techniques to extract and validate data from various sources. Machine learning and natural language processing contribute to extract relevant information from structured and unstructured data and increase the quality and validity of compared groups. 

Utilizing RWD to construct control arms can be highly beneficial, but it is crucial to find a competent data provider for this purpose. Collaborating with a reliable RWD partner ensures the utilization of accurate, extensive, and representative evidence, ultimately leading to more informed healthcare decisions and improved patient outcomes. Therapyte’s dedicated team of data scientists and analysts are always willing to advise potential clients on the implementation of ECAs and other RWD/RWE services.