How AI can lower the costs of clinical trials

Drug costs are unsustainably high, but using AI in the recruitment phase of clinical trials could play a hand in lowering them.
article cover

Natalia Gdovskaia/Getty Images

· 4 min read

When you think of artificial intelligence (AI), you may think of the machines that take over the world in The Matrix and use a dashing young Keanu Reeves as a battery. We’re not here to weigh in on the likelihood of a future AI robot versus human apocalypse, but one thing we do know is that—when it comes to clinical trials—AI has the potential to do a lot of good.

Clinical trials are extremely expensive—they cost, on average, $700 million to $1 billion, according to Stefan Harrer, chief innovation officer at the Digital Health Cooperative Research Centre, an Australian investment fund and incubator for the digital health industry. Patient recruitment makes up 32%, or $224 million to $320 million, of that cost, according to a brief from the Deloitte Centre for Health Solutions.

Yet, a lot of that money ends up down the (metaphorical) toilet since only 10% of drug trials end with an approval, according to Deloitte.

“Right now, the world can’t afford the medicines and the devices that are coming out,” said Michelle Gallaher, CEO of Opyl, which created an AI platform that’s used for trial recruitment. “We could repurpose literally hundreds of billions of dollars a year to go back into medical research or back into healthcare.”

And tackling the recruitment process would make a big dent in lowering overall clinical trial costs since it makes up such a big chunk of the cost.

Why is the recruitment phase so expensive?

A lot of costs go into the recruitment phase. There are costs for marketing, so participants know there’s a trial to join. Nurses need to screen participants to make sure they fit the trial’s qualifications. And don’t forget the recruitment events.

“For each day that a drug or a device is delayed to market, it costs a pharmaceutical or a med-tech company money. But more importantly, it costs them time on their patent as their patent starts to run down. So they need to get the product to market as successfully and as fast as possible,” said Gallaher.

But recruitment takes time and is difficult for a number of reasons. For one, trials need people who meet extremely specific guidelines.

“Medicines are becoming more and more specialized,” said Lucas Glass, vice president of artificial intelligence at life sciences research and analytics firm IQVIA. “When you can target people with specific gene populations…it becomes much more difficult to find the right patients.”

Navigate the healthcare industry

Healthcare Brew covers pharmaceutical developments, health startups, the latest tech, and how it impacts hospitals and providers to keep administrators and providers informed.

Another is a lack of incentive for people to participate. A reason a lot of healthy people choose to join Phase 1 clinical trials is the financial incentive, since you can make thousands of dollars from participating in a trial. The median compensation for healthy volunteers in phase one clinical trials is $3,070, according to a 2021 study.

But when the job market is high, as it’s been this year, there’s often less of a financial incentive, Gallaher said.

So how does AI solve these problems?

AI can make it way easier to identify patients who meet the qualifications for a trial, mainly by sorting through massive troves of data that no human would be able to go through on their own, said Arun Bhatt, an independent clinical research and drug development consultant.

Clinical trials often look for participants by using electronic health records (EHR) to see who fits the qualifications. If a health system has an EHR database of 100,000 patients, it would take a very long time to sort through all of them manually. AI can quickly pinpoint who’s eligible for a trial, said Bhatt.

Opyl uses AI to identify potential trial participants via social media, Gallaher said. Maybe someone is talking about their diabetes on social media. Opyl’s AI platform can identify that person as a potential participant for a clinical trial testing a new diabetes treatment.

Are clinical trials actually using AI yet?

In practice, not a lot of clinical trials are actually using AI yet, at least not for recruitment. Gallaher estimated that 10%-15% of trials use AI for recruitment.

“Healthcare is one of the last industries to come to digital transformation. And therefore, I think clinical trials is one of the very, very last to come into this digital economy,” said Gallaher.

One reason AI may not be used as much as it could is a lack of trust, Bhatt said.

“There are these questions about how much trust we can put today in this technology,” he said. “One has to put some faith in the technology to see what happens.”

Navigate the healthcare industry

Healthcare Brew covers pharmaceutical developments, health startups, the latest tech, and how it impacts hospitals and providers to keep administrators and providers informed.