Medidata Blog
Cross-industry Historical Clinical Trial Data: The Secret to Regulatory Success with a Synthetic Control Arm®
Ruthie Davi, our VP of data science, and Jacob Aptekar, our senior director and trial design solutions lead, share their perspectives on the use of a Synthetic Control Arm, pioneered by Medidata Acorn AI, for more efficient trials in a regulatory setting.
What is a Synthetic Control Arm?
The randomized, controlled trial is well known as the gold standard for how studies for new therapies are conducted. Yet prospectively randomizing patients to standard of care may not be possible for every study due to many factors. For example, it may be unacceptable to assign seriously ill patients to a control arm where standard of care therapy is not sufficiently effective. In these types of situations, a randomized clinical trial may fail to provide an appropriate scientific answer for investigators and place an undue burden on patients participating in the clinical trial. For these reasons, investigators are sometimes limited to less than optimal single-arm trial designs.
A Synthetic Control Arm (SCA) offers a design option that can reduce—or even eliminate—the need for patients in a control arm. This involves leveraging patient-level data from historical clinical trials in the same indication. Historical patients who meet eligibility criteria and were assigned to receive an appropriate standard of care are carefully selected. These patients are intentionally chosen to have a status at baseline that matches those in the current-day experimental arm in order to create an apples-to-apples fair comparison.
How does Medidata Acorn AI build SCAs?
First, we assess the eligibility of the patients for the SCA using screening and baseline measurements from historical trials to determine which patients meet the eligibility criteria of the current trial. Essentially, we identify patients who would qualify for the investigational trial if they existed today. Then, in order to make a direct and fair comparison, statistical methods, such as propensity score matching or weighting, are used to select the right patients from the eligible historical group to align with the baseline characteristics in the investigational arm of the new trial. We then proceed much like a randomized controlled trial. We make comparisons between the investigationally treated patients and the SCA, in terms of efficacy and safety, and attribute differences to the investigational product.
Can an SCA ever truly replace a traditional control arm?
Medidata Acorn AI has demonstrated that in at least some cases, an SCA can provide the necessary information to replace a randomized control arm without altering the scientific understanding of the effect of the medical product. Working in collaboration with the Friends of Cancer Research, Medidata Acorn AI conducted case studies to evaluate whether an SCA could match the results of a randomized control. Working with a small set of historical clinical trials in a single indication, the researchers first chose one trial as the target and completely set it aside. Then control data from all other available trials were used to create an SCA. Finally, the researchers were unblinded to patient outcomes and the overall survival of the randomized control from the target trial was compared to the SCA. In both case studies (one in non-small-cell lung cancer and one in relapsed refractory multiple myeloma) the overall survival of the randomized control was closely mimicked by the SCA.
By reusing patient-level data from historical clinical trials, it is possible to create a well-balanced SCA. And in these two case studies, the SCA provided highly similar estimates of overall survival as the randomized control arm. This carries important implications for use in indications where randomization is problematic.
How are SCAs viewed by regulators? Can you get FDA agreement to use it?
SCAs have already benefited sponsors in the real world. For example, Medidata Acorn AI worked with Celsion while the oncology drug developer was conducting an early-phase, single-arm trial for an ovarian cancer candidate. Although patients treated with the investigational product had promising results, without a control arm, it was difficult to put them in context. Medidata Acorn AI built an SCA for Celsion to help estimate the magnitude of the treatment effect, which was used to make a go-decision for the development of the investigational product and then to design their Phase 2 trial. Because of the more precise estimate of the treatment effect afforded by using an SCA, the company was able to save about 20 patients in that upcoming trial.
“We are extremely impressed with the high quality of the matched data from the Medidata SCA. Using an SCA for a portion of the study will reduce costs and should improve the rate of enrollment as patients will be more likely to receive GEN-1 rather than placebo.”
— Michael H. Tardugno, Chairman, President, and Chief Executive Officer, Celsion
In another groundbreaking case, the US Food and Drug Administration (FDA) agreed to consider the use of SCA in Medicenna Therapeutics’ phase 3 registrational trial in recurrent glioblastoma (rGBM). This was a precedent-setting regulatory decision to consider a hybrid external control (combining synthetic control arm patients with randomized patients) in a phase 3 trial is an indication that previously used traditional randomized controls.
“We are extremely impressed with the [Medidata Acorn AI] team for providing a scientifically rigorous rationale for the design of an innovative registration trial incorporating an external control arm for the treatment of recurrent glioblastoma (rGBM) with MDNA55. Their expertise and collaborative effort with thought leaders was instrumental in demonstrating to the FDA the validity of a well designed external control in a registration trial.”
— Fahar Merchant, Ph.D., President and CEO, Medicenna Therapeutics