Synthetic data may revolutionize clinical trials – and save lives

Brandon Ryan and Thomas McLaughlin were two cousins facing one cancer. Their heartbreaking fight against melanoma in 2010 highlighted a tragic, often overlooked element of randomized clinical trials: control groups.

Although they both faced the same type of aggressive melanoma and enrolled into clinical trials for the same experimental medication, only Thomas would end up receiving the drug that would ultimately prolong his life.

Randomized control trials are essential when investigating novel medications. Control arms are especially important, since researchers must have comparators when measuring the effects of a new drug compared to current standards of care. But recent innovations in machine learning have challenged the idea that human control arms are truly necessary.

Synthetic control arms refer to datasets generated from patient health records using artificial intelligence that statistically represent patient parameters. These datasets may provide enough statistical rigor to replace the need for control groups while bypassing serious privacy risks associated with using real patient data. As a result, some clinical trials may have the option of using these synthetic datasets in place of data generated by real patients.

With continued innovation in this intersection between data science and clinical research, patients looking to receive experimental medications would no longer face the risk of being placed in the control group.

Indeed, there are already considerations in place to support patients in the human control arm who would benefit from the experimental therapy. In many clinical trials, when experimental medications are recognized to have a definitively superior benefit early in the study, the trial can be stopped and patients in the control groups are also provided with the new medication. But for patients with aggressive diseases like cancer, not everyone can afford to wait.

For patients with aggressive diseases like cancer, not everyone can afford to wait.

For patients facing such dire circumstances, selection into the human control arm of a study would take their final chance to fight away. In Brandon’s instance, despite enrolling in the same clinical study, Brandon was randomly assigned to receive chemotherapy medications as the standard of care. He would pass away that very same year. Meanwhile, Thomas was randomized to the experimental arm of the study – the synthetic control arm – and would live on for five more years.

Although the application of synthetic data to clinical trials is promising, challenges remain. For one, this technology is novel and requires more time and success cases to fuel adoption. Additionally, the creation of these datasets can be time consuming, and the technical expertise required in this process may not always be readily available to clinical trial teams.

So far, the track record for its use is promising. Both the European Medicine Agency and the US Food and Drug Administration recognize the use of synthetic control arms for drug approvals. In fact, some medications have already been able to move through the pharmaceutical pipeline using synthetic data. One example is blinatumomab, a medication used to treat leukemia, in which synthetic data was used to gain timely regulatory approval.

Behind each clinical trial, there are real human lives at stake. Although the application of synthetic data is significant from a technical perspective, the most critical element of this technology is the impact it can have on patients. For this reason, investment into its adoption is paramount.

With sufficient adoption, advancements in synthetic data could lead us to a future in which every cancer patient enrolled in a clinical trial is guaranteed hope. A future where the rigours of science no longer leaves patients behind.