Meet the Company Trying to Democratize Clinical Trials With AI

A decade ago, Pablo Graiver was working as a VP at Kayak, the online airfare aggregator, when he sat down to dinner with an old friend—a heart surgeon from his home country of Argentina. The talk turned to how tech was doing more to save folks a few bucks on a flight to Rome than to save people’s lives. The biggest problem in healthcare? “Clinical trials,” she said. “They’re a disaster.”

Right now, the US has exactly 19,816 clinical trials open and ready to recruit patients—trials of promising new therapeutics to fight everything from HIV to cancer to Alzheimer’s. About 18,000 of them will get stuck on the tarmac because they won’t get enough people enrolled. And a third of those will never get off the ground at all, for the same reason.

So where are all the patients? Well, the vast majority of them either don’t know the trials exist, or don’t know they can participate. Since 2000, the government has kept details of every clinical drug trial in a national registry, but it’s a nightmare for the average human to navigate. So most pharma companies use recruitment firms to painstakingly comb through patient medical records and find people who might be a good fit—geographically, genetically, and generationally. Each patient hunt is basically a one-off. Like, say if every time you wanted to fly somewhere you had to search on the websites of United, Delta, American, Frontier, Alaska, and Southwest one at a time. And then do the same thing for hotels. (Man, the early aughts were bleak, weren’t they?)

Graiver’s new company, Antidote, does for clinical trials what Kayak and Orbitz and Priceline did for travel. It gives that painful patient matching problem an e-commerce solution. “Fundamentally, it’s just a question of structuring information,” says Graiver. “Which is something the tech world is great at. I was shocked no one had done it already.”

The information that most needed help was something called inclusion/exclusion criteria. It’s what makes a patient eligible to enroll (or not) in a trial: things like age, sex, prior treatment regimes, and current health status. When drugmakers submit new trial details to ClinicalTrials.gov, most of it gets entered as structured data, the kind of thing you enter in a drop-down menu. But eligibility criteria gets entered in a free text field, where you can write whatever you want. That lack of structure means a machine can’t read it—unless it’s been properly trained.

That’s what Antidote does. Graiver’s company started by amassing thousands of clinical studies from ClinicalTrials.gov and the World Health Organization, and they hired clinical experts to manually standardize all that free-wheeling trial jargon into structured language a search engine could understand. Then they trained it to categorize and identify studies using that language.

If you search for adult onset diabetes, it will know to pull up trials for Type 2 diabetes, and diabetes mellitus 2, and T2DM—since they’re all ways to describe the same disease. Called TrialReach at the time, the company proceeded slowly, focusing first only on diabetes and Alzheimer’s studies.

Then in 2015, Graiver’s platform got a big boost from big pharma. For two years prior, Novartis, Pfizer, and Eli Lilly had worked together to organize their trial data to be machine-readable. But as they looked to expand the consortium, the three pharma giants realized a need for a more neutral host organization. So they gave the tech to Graiver. Today, three years and a new name later, Antidote has annotated more than 14,000 trials—about 50 percent of what’s listed on ClinicalTrials.gov—spanning 726 conditions.

The result of all this data structuring is that Antidote can take a number (say, 50) and return studies that say something like this: “Ages Eligible for Study: Child, Adult, Senior” but not studies like this: “Ages Eligible for Study: 75 years and older.” And the interface is pretty slick. You type in your condition, where you live, then choose your age and sex. For a 50-year-old woman living in St. Louis, Missouri with lung cancer, 617 trials pop up. On the next screen, Antidote asks how far you’d be willing to travel; within 20 miles the trial options narrow to 69. If you know what kind of mutation is causing your lung cancer, Antidote can winnow down the number even further. At this point, you could print out a list of the trials, take them to your oncologist, and discuss your options.

Or, you can click on any trials you’re interested in, register your email with Antidote, and they’ll send you contact information for the trial organizers, along with next steps. They’ll also keep you updated on any new trials for which you might be a match.

The service is totally free for patients, who can find it on their own or through a widget on websites for patient organizations. Through 231 of those partnerships, including with the American Kidney Fund, Muscular Dystrophy Association, and Lung Cancer Alliance, Antidote says it reaches more than 15 million people per month. On the website of JDRF—the leading Type 1 Diabetes research fund in the world—27,863 people have searched for a trial using the Antidote widget since it launched in 2016. That’s more than in the previous 10 years combined using JDRF’s existing search tool.

More on Clinical Trials

• Megan Molteni

  At ClinicalTrials.Gov, Untested Stem Cell Clinics Advertise for Free!

• Adam Rogers

  Scientists Hate the NIH’s New Rules for Experimenting on Humans

• Megan Molteni

  The Most Promising Cancer Treatments In a Century Have Arrived—But Not For Everyone

“It makes it less of a wild goose chase for patients,” says Esther Schorr, COO of PatientPower, an online cancer news site and Antidote partner. Surveys of their 30,000 member community have shown an uptick in trial enrollment since the widget went up about a year ago. “There’s just so much information for the common man or woman to get through. Technology can really make a patient’s journey easier.”

It’s also making things easier (and cheaper) for drugmakers. Antidote makes money chiefly by selling limited access to this user database to the world’s biggest pharma companies and clinical research institutions, helping them to fill their own trials.¹ When you enter your email address, you’re consenting not just to having your personal information shared with the sponsor of a particular trial, but to having your deidentified data shared with third parties.

Antidote maintains that it still keeps up some kind of a firewall; pharma companies can’t just contact you out of the blue—they have to place a request through Antidote, that you can accept or deny. But the broad consent language in the company’s privacy policy gives Antidote a lot of latitude with how it can use your name, age, sex, location, and any other details you provide about your medical condition.

It’s a tradeoff between privacy and care that many patients are confronting these days. Like the seniors filling their homes and wardrobes with IoT-enabled sensors to keep track of their movement and heart rates. Or the record number of Americans letting companies mine their DNA, so they can know if they’re at higher risk for genetic diseases like Alzheimer’s or cancer. For Antidote’s users, the promise of a cure—however distant—is well worth the risk.

_1 Correction appended 01/30/18 5:40pm EST This story was changed to clarify how Antidote earns revenues by providing clinical trial sponsors access to eligible patients.

Related Video

Technology

The Robot Will See You Now – AI and Health Care

Artificial intelligence is now detecting cancer and robots are doing nursing tasks. But are there risks to handing over elements of our health to machines, no matter how sophisticated?