Two arms of a clinical trial, with a synthetic and active arm.

Beyond Placebos: The Promise and Limitations of Synthetic Control Arms

Who doesn’t lie to their doctor about their alcohol intake? OK, maybe not everyone…but it’s a fairly common occurrence and it’s important to this story. It’s a story about human variability and the promise of synthetic control arms.

Synthetic control arms are an incredibly promising initiative within drug development. Instead of randomizing patients to receive either a placebo or the standard of care, synthetic control arms use real-world data (think electronic health records and insurance claims) to build a fictional model for how patients would have progressed without a specific intervention. That data can then be compared to participants in the trial, each of whom received the investigational therapy.

This approach can address many major clinical trial challenges. That’s why the FDA is increasingly on board. For participants, there’s no concern they won’t receive the active drug. That’s huge! As many as 50% of patients in a randomized controlled trial (RCT) get either a placebo or the standard of care that the new drug is trying to beat. It’s a major caveat to clinical trial participation.

For sponsors, there’s a related benefit. Synthetic control arms (sometimes called external control arms) can fast-track trial recruitment. Far fewer participants are needed because there’s no control arm to enroll. And because everyone gets the active drug, it’s an easier sell for participation. The clinical trials can run cheaper and move faster, helping to streamline the biggest bottleneck in drug development.

Control data, but better

The benefits extend far beyond recruitment. Synthetic control arms can help with data quality and downstream analysis.

Sponsors always do their best to match participants in the active and control arms to minimize unexpected results. This includes “controlling” for major factors, such as age, gender, and the nuances of each diagnosis to ensure both groups are balanced. However, there are too many variables to align, and there is no such thing as a perfect match. There’s always an element of luck (or risk) with real human participants.

This can make a drug look better than it is, which is bad news, even for the sponsor. Better-than-expected data usually triggers greater investment and subsequent clinical trials, which will eventually show the drug isn’t that effective.

Or the opposite can happen: Unexpected improvements in a control arm can obscure the comparative benefits of the investigational drug. The hope is that enough participants are enrolled to average out any unexpected outcomes, but that’s not always the case.

Controlling for human behavior

From a data standpoint, synthetic control arms can remove variables that may occur in the placebo arm of a clinical trial. Unfortunately, this can also skew the data in a different way. If we rely on observational data, from electronic records or insurance claims, the control arm of the trial is not subject to the same hoops, tests, and protocol as participants in the actual study.

And sometimes participation in the study itself can influence outcomes. This is known as the Hawthorne effect: Humans change their behavior when they know they’re part of a study or are under scrutiny in some way.

One highly relatable example of this is trials for fatty liver disease. To catch you up: There are two types of fatty liver disease. One is linked to alcohol use and the other––non-alcoholic fatty liver disease or NAFLD––is associated with other lifestyle and metabolic factors. NAFLD is now the most common liver disease in the world and represents a huge market for drug development.

NAFLD trials have struggled in the past with notable improvements in the placebo arms of the trial. Somehow, folks who are not receiving any real intervention “show significant histologic, radiologic, and biochemical responses.” The reason: A significant percentage of participants misreport their alcohol. This led to an incorrect diagnosis of non-alcoholic fatty liver disease and enrollment in the NAFLD clinical trial. When it came time to participate, the testing and additional scrutiny seemed to cause a significant number of these participants to moderate their drinking, improving their liver enzymes and muddying the data.

It’s a problem with the control arm, but it also impacts the active arm of the trial. Everyone’s on their best behavior because they are part of a clinical study. Pulling data from electronic health records won’t provide a decent comparison, as those folks aren’t moderating their behavior based on the study conditions (aka the Hawthorne effect).

Removing some of the human variables

Is the Hawthorne effect a deal-breaker? Far from it. The positives of synthetic control arms far outweigh their limitations.

If you’re not convinced, here’s another way of looking at it: Every now and then, there’s a major phase 3 readout. A drug improved overall survival compared to a standard control arm of human participants. It’s a wonderful advance for medicine, but a measurable improvement in overall survival meant that some participants were randomized not to receive the active drug. And that luck––pure chance, by design––impacted their life expectancy and their final months. The participants who received the active drug benefitted from more time on this Earth. That’s why synthetic control arms are such an important initiative: They give all participants a fighting chance against their disease.