The FDA has asked Congress for $100 million to help build an advanced analytics system the agency says will help it, at long last, master real-world evidence.

The funding, part of President Trump’s fiscal year 2019 budget request, would go toward building a database of at least 10 million patients and provide regulators with “near real-time evidence” on “a broad range of medical products, including drugs, biologics and medical devices,” the FDA says.

Drug sponsors and regulators have long been hopeful that real-world evidence — electronic health records, insurance claims, product and disease registries, mobile health apps — might offer shortcuts through the clinical process and bring safe, cheap drugs to market more quickly by allowing researchers to scale up their trials. But industry leaders and policymakers are still feeling their way forward.

It’s easy to see the appeal of folding real-world evidence into clinical trials. Traditional clinical drug or diagnostic trials have necessarily kept a narrow focus — excluding whole swaths of potential participants in the hope that no other possible variables could cloud results. In practice, though, that means researchers and drug sponsors run the risk of finding out only after the fact — sometimes even after a med has been approved — that a drug may not perform as well or as safely in the real world as it did in a controlled clinical setting.

The explosion of relatively inexpensive, powerful technology — everything from Fitbits to artificial intelligence algorithms — offers sponsors a chance to manage multiple variables through data that, in many cases, has already been collected and, also, is continually updated. Considering using an approved drug for a new indication but worried about its side effects? What could be easier than tapping into a health outcomes database of thousands of patients who took the med to assuage or confirm concerns?

“That data helps you identify down to the personal level who will be eligible for a certain trial and who won’t be eligible,” says Joe Selby, executive director of the Patient-Centered Outcomes Research Institute (PCORI), a Washington, DC, nonprofit that funds trials. “We can identify almost exactly the people we would like to invite to a trial before we ever put a stamp on the postcard or send an email. That’s a huge advance.”

For the FDA, the challenge is to find a way to use that data to speed up approvals without risking public health or safety. Some researchers aren’t waiting for the FDA to catch up. They’re already doing their own large-scale, real-world trials. Russell Rothman, vice president for population health research at Vanderbilt University Medical Center, says a trial he’s conducting (on the effect of aspirin doses on people with high risks for heart disease or stroke) is about to enroll its 10,000th patient. The five-year study, called ADAPTABLE and funded by PCORI, plans to enroll 15,000 people — and researchers are trying to make it all as “real-world” as possible. “We have very broad inclusion criteria,” Rothman says. “We take almost all comers out there.”

Patients are randomly assigned either a daily baby (81 mg) or regular strength (325 mg) dose of aspirin. Participants are being enrolled through some 30 healthcare systems around the country linked by PCORnet (PCORI’s analytic system). The data is collected automatically whenever patients visit their doctors for checkups and researchers check in with study subjects every three-to-six months for a brief survey. After that, researchers basically sit back and wait, Rothman says. “We don’t give the doctors treatment algorithms. We really try to leave these patients in the real-world setting as much as possible. All sorts of things that happen in the real world, we let them happen in the trials,” he says.

The FDA’s budget request suggests the agency is taking real-world evidence seriously — and wants to get the show on the road. It’s the second major step the FDA has taken on the issue in the past year: It also published guidelines on the use of real-world evidence in medical device approvals.

The proposed new data analytics bank would build on the agency’s Sentinel and National Evaluation System for Health Technology (dubbed “NEST”) and represents “the next evolution” in the FDA’s role as Big Data manager, Commissioner Scott Gottlieb said in a recent blog. Used properly, the proposed database could become “a national utility,” he added.

But the budget proposal also illustrates how much work is yet to be done. Sentinel and NEST relied on healthcare payer claims. The new system will rely on electronic health records. Whether or not Congress approves the budget request, the FDA still has to standardize language and measurements across millions of different data points, Gottlieb noted. For example, should patients’ temperatures be recorded in Celsius or Fahrenheit? That may seem like a tiny detail but it’s the kind of thing that can throw off an entry and create cascading errors in a database. That’s partly why some experts are on the fence about real-world data, especially in a drug’s pre-approval phase.

“It’s impossible to learn everything there is to know about a new drug in the pre-marketing phase,” says James Bannon, CEO of Vigilare, a consulting company that helps sponsors and sites manage risk. “The reason for that is because good clinical trial design requires you to eliminate compounding variables.” Real-world evidence, Bannon says, has become a “buzz term” that means different things to different people. He is open to the possibility that advanced analytics might help shorten trials but says their most important role comes after a drug has already gone through regulatory approval. “The major issue becomes, ‘Is it safe?’ And the primary mechanism for proving that it’s safe is the adverse event reporting mechanism that all the regulators can use” after a drug comes to market, he says.

Annette Stemhagen, senior vice president at United BioSource Corporation, is more sanguine about the prospects for real-world evidence in clinical trials. “Epidemiologists have been doing real-world evidence for years. That part is not novel. It’s bigger data now, but the strategies, the techniques are the same,” she says.

It’s not just that there’s already a ton of data out there that can help inform a trial, Stemhagen says. It’s that there’s a seemingly limitless capacity to gather even more information before a drug or diagnostic hits the market. “Really, every company ought to be thinking about natural histories,” Stemhagen says. This kind of deep data dive “improves your development program whether or not it saves your drug,” she adds.

For now, even the most aggressive advocates of real-world evidence are only talking about it augmenting or supplementing clinical trials, not replacing them. Rare diseases, novel treatments — they’ll likely continue to need close observation with a narrow focus. But even in these cases, technology can help researchers do a lot more with a lot less, a lot faster, PCORI’s Selby says. “It’s a great way to find all 20 of them instead of six of them.” In fact, he adds, the enormous scale that modern technology offers probably means we’ll see more not fewer clinical trials.

“The truth is many trials get dragged out longer than necessary because they’re so small,” he says. Modern technologies “make it really feasible to think about faster, larger trials. Given the level of uncertainty in medicine, we should be doing an order of magnitude — 10 times, 100 times —more trials than we’re currently doing.”