Welcome
Pharmaceutical & Life Sciences News

Faster Drug Approvals Possible as AI Speeds FDA Reviews

April 29, 2019, 9:26 AM

Artificial intelligence should speed up the decade-long process to bring new drugs to the market as the nation’s top regulators explore methods for automating portions of its reviews.

While the FDA is the shortest leg in the drug review process, regulators are considering the same AI tools that drug and technology companies are using to improve molecule screenings and clinical trials, Sean Khozin, associate director of the Food and Drug Administration’s Oncology Center of Excellence, said.

“Absolutely, drugs can get onto the market much faster” with artificial intelligence, Khozin said. “We have to quantify the risk and we have to quantify the benefit, and that quantification could be much faster as we automate it.”

The FDA is already laying the groundwork for using AI and machine learning—artificial intelligence that uses algorithms and statistical models to perform tasks and predict outcomes without being explicitly programmed—MIT researcher Pratik Shah said. The agency already has a real-world evidence framework and a plan to regulate software as a medical device.

“I think the FDA has done a really good job of actually being proactive and putting guidelines out,” Shah said.

The drug development process costs about $2.58 billion and takes a decade to hit the market, according to the Tufts Center for the Study of Drug Development. That prompted Congress, drug companies, health agencies, and others to find ways to improve this process.

Speeding Data Reviews

The FDA is already working on pilot projects to test AI in its review processes. Khozin also founded an FDA incubator to drive innovations in digital health and big data to advance public health.

Part of the FDA’s review process is to look for signs a medical product application is safe and effective enough for the agency to grant an approval.

“Operationally, we could start to automate a lot of the signal detection work,” he said. “Right now it’s manual. You have to work with spreadsheets and your favorite data analysis tool, and then you manually go through and comb the data.”

There’s also potential to develop machine learning algorithms based on clinical trial data, Leila Pirhaji, chief executive officer of AI-driven drug discovery platform ReviveMed, said. At the same time, about half of the clinical trial results go unreported, often because the results showed a drug is inferior to what’s already on the market, according to he World Health Organization.

“I believe the accumulated data from clinical trials with both positive and negative outcomes would be invaluable for training AI algorithms,” Pirhaji said.

To contact the reporter on this story: Jeannie Baumann in Washington at jbaumann@bloomberglaw.com

To contact the editors responsible for this story: Fawn Johnson at fjohnson@bloomberglaw.com; Andrew Childers at achilders@bloomberglaw.com