The Food and Drug Administration said Tuesday that it will launch a new research project focused on real-world evidence — data collected by insurance companies, in electronic health records, and in other places in medicine — to learn more about Covid-19, including how diagnostics and medications are being used in the pandemic and how best to design studies to test them.
The project is a collaboration with Aetion, a New York health tech startup that specializes in real-world evidence.
“Right now, I think that the most intriguing part of the Aetion story is the ability to have a very consistent analytic approach that can be applied across multiple datasets,” Amy Abernethy, the FDA’s principal deputy commissioner, said in an interview.
advertisement
The exact questions the collaboration will explore are yet to be determined, Abernethy said, but she said the agency felt it had to “pressure test” the company’s technology.
The partnership emerged from the Covid-19 Evidence Accelerator, an initiative between the Reagan-Udall Foundation and Friends of Cancer Research that allowed teams from government, academia, and industry to present real-world evidence efforts to one another.
advertisement
“The goal is to figure out how we can advance Covid-19 evidence out of that,” said Carolyn Magill, Aetion’s CEO. “This relationship with the FDA materialized to look at safety and effectiveness of treatments, vaccines, and diagnostics. And we’re really excited about it.”
Abernethy said she has been thinking a great deal about how real-world evidence can be used against Covid-19. The disease presents a genuine problem for researchers seeking to conduct observational studies, because patients can vary so dramatically in the severity of their disease, and because doctors really don’t have good ways to predict who will recover and who will get worse.
This means that two groups of patients who appear similar based on their electronic health records might in fact be very different. The best approach is to randomly assign patients ahead of time to get one treatment or another, which creates two nearly identical groups even if researchers don’t understand what is causing the disease to worsen.
The question is whether there’s a way to draw conclusions after the fact, for patients who have already been treated. That’s the problem that real-world evidence researchers — as well as companies such as Aetion and Flatiron Health, where Abernethy previously worked — are trying to solve.
This can be difficult. Abernethy points to a study conducted by researchers in the Department of Veterans Affairs on using the drug hydroxychloroquine to treat Covid-19 that was published in April. It showed the drug was not only ineffective, but potentially harmful.
“It basically was a paper that showed that sick patients were more likely to get prescribed hydroxychloroquine, and the sick patients die,” Abernethy said. “Now is it that it’s hydroxychloroquine that’s causing the death, or is it the fact that sick patients are the ones that more frequently die of Covid-19?”
Since then, two studies in New York City published in the New England Journal of Medicine and the Journal of the American Medical Association both showed no efficacy for hydroxychloroquine, but also not the same signs of significant harm, because they controlled for how sick patients were. Randomized trials for hydroxychloroquine in Covid-19 are still being completed.
There are problems that will and won’t be addressable using real-world evidence, Abernethy said. She doesn’t think that the FDA will be trying to use “synthetic controls” — creating control arms so that drugs can be studied without randomized trials. But she does think there will be some treatments — she names convalescent plasma, the use of blood plasma from recovered Covid-19 patients to treat the sick — where real-world data may emerge must faster than randomized studies.
More than that, she said, she envisions using real-world evidence to prioritize which medicines are tested, and to understand exactly which patients they should be tried in.
“It really helps you zero in on the work that you should be doing, quickly, to resolve whether or not identified treatments bear out in clinical trials,” Abernethy said.
