Subscribe
4 min read

FDA is digital health’s gatekeeper amid AI boom

FDA is digital health’s gatekeeper amid AI boom unknown FDA is digital health’s gatekeeper amid AI boom  Modern Healthcare

When it comes to commercializing artificial intelligence solutions, many digital health companies face a long road that runs directly through Silver Spring, Maryland. Silver Spring is the headquarters of the Food and Drug Administration, the agency providing clearances, designations and approvals for an increasing number of AI-enabled medical device and software products.  Read more: Healthcare AI regulation is coming into focusThe healthcare industry is awash in AI-enabled software applications that aim to diagnose and treat clinical conditions. But before many companies can put their AI products on the market, they may need to work through a lengthy, and potentially costly, FDA process. Facing an overload of AI solutions, the agency acts as a gatekeeper, ensuring end users are using safe and effective technology that does what it says it can do.“The FDA process is the absolute minimum to get these devices in clinicians’ hands,” said Connor Landgraf, co-founder and CEO of Eko, a company that’s developed a digital AI-enabled stethoscope. Between 1995 and last October, the FDA approved, designated or cleared 692 AI-enabled devices, according to the most recent data available. The list includes software as medical devices, which is how the FDA evaluates algorithms used for clinical purposes through a standalone software application.

EMAIL ADDRESS Nearly seven out of 10 from the list of 692 occurred after 2020, which coincides with rising interest in clinical AI from providers, investors and developers. It doesn't include AI that's not regulated by the FDA, such as products focused on overall general wellness or that summarize notes in the electronic health record.“There’s just been an explosion of innovation in this space,” said Kyle Faget, healthcare and life sciences lawyer with law firm Foley & Lardner. “What's hard for young innovators is that they have to make a determination at the outset, whether or not the product they're developing is subject to FDA oversight.” The process can be long, but companies say it’s worth it. In May 2023, Eko received FDA 510(k) clearance for one of its digital stethoscopes. The clearance, which the FDA gives when a device is substantially equivalent to a device that is already on the market, allowed Eko to almost immediately start marketing and selling the stethoscopes to clinicians. Eko also received clearance for its AI heart disease detection software in July 2022.  Landgraf said the clearance process meant Eko had to perform rigorous clinical research, beginning in 2014, to prove the stethoscope's safety and effectiveness. Eko, which has raised more than $125 million, spent tens of millions of dollars in research and development during the process, he said.  “With any FDA-approved technology, as an entrepreneur, you realize that it will add probably at least five years to your company’s lifespan. It’s going to be a long journey and healthcare is already a long journey,” Landgraf said. “We went in eyes wide open knowing that if it was going to be a long journey, it would be worth it and we'd build something really durable at the end.” What can complicate matters for digital health companies is AI's fast evolution. The FDA formed a digital health advisory committee in October and former Oracle executive Troy Tazbaz was recruited to oversee the agency's Digital Health Center of Excellence. “FDA has regulated AI for decades,” said Deeona Gaskin, partner at law firm Sidley Austin and former associate chief counsel for enforcement at FDA. “What’s new are these generative AI large language models, where validation is more challenging. I do think FDA has frameworks but has acknowledged that those frameworks were developed decades ago and that maybe they could evolve.” What’s also hard for the agency is AI models can change over time based on data inputs, Faget said. This means evaluating products for safety and guidance can be a moving target that requires companies to submit a predetermined change control plan. The FDA declined to make someone available for an interview on how it regulates AI.In a speech to the Coalition for Health AI on March 5, FDA Commissioner Dr. Robert Califf said the issue of determining whether AI applications are safe and effective is something that's "too large to be contained within FDA."While most digital health companies receive 510(k) clearances from FDA, there are other designations. Startup Cognoa received FDA authorization for its diagnostic aid for autism spectrum disorder in June 2021. The agency assessed the software’s safety and effectiveness during a yearlong study the company conducted with 425 patients. The device uses AI and parent-inputted videos of a child to evaluate a patient’s risk for autism. Getting that authorization was a big reason why Pittsburgh-based insurer Highmark signed on as Cognoa’s first commercial payer partner in January, said Cognoa CEO Dr. Sharief Taraman. Cognoa's study faced more rigor than others going through the FDA process because the solution was new to the marketplace, he said.Still, getting FDA designations and payer coverage doesn't mean companies can overcome the cost and resources of running clinical trials and lobbying additional insurers to cover their products. Digital therapeutics company Pear Therapeutics received FDA clearances for three of its AI-enabled diagnostics and coverage from multiple insurers including Highmark. But even with that going for it, the company went bankrupt in April 2023.   Cleerly, a company that uses AI to diagnose heart disease, received an FDA breakthrough designation for its non-invasive imaging-based investigational software device earlier this month. The voluntary program is designed to expedite how fast a product can to go to market. It requires companies to prove their technology is safe, effective and that no equivalent product exists in the marketplace, said Dr. Udo Hoffmann, Cleerly's chief scientific officer.But even getting this designation is just an initial step. Cleerly is conducting another study, until the end of 2025, that's expected to cost millions of dollars and will cover 7,500 patients. As more clinical AI comes into the marketplace, Hoffmann said it imperative that the FDA work with other stakeholders, such as academic medical centers, to help providers understand which of these devices and algorithms are most effective and safest to use. “If you have 15 products that can detect lung cancer, it’s almost like having 15 TVs to choose from,” Hoffmann said. “Who determines which three or four TVs are worth the price and the quality? That’s the situation we’re going to face in medicine.” 

Published by: