The Food and Drug Administration is sticking its toe in the water of artificial intelligence, providing its first guidance on the emerging development of applications for the technology in healthcare.
The FDA released model 1.0 of its software precertification pilot Monday to provide an initial tool to test these programs. The agency noted that with AI and machine learning technology advancing rapidly, the health IT community must move quickly to ensure their safety in practical applications.
“Software is increasingly used in healthcare to treat and diagnose conditions and diseases, aid clinical decision making and manage patient care,” the FDA wrote in its working model (PDF). "Under this program, software developers would be assessed (by FDA or by an FDA-accredited third party) for the rigor of their practices in software design, testing, clinical assessment, and real-world performance monitoring, along with other appropriate capabilities."
The new program aims to provide companies with an optional "Excellence Appraisal" and thereby give them a "pre-check" approval. This would support the company's ability to gain approval for new applications in the future.
The pilot will start testing the safety of AI applications during the first half of the year, and the FDA plans to collect public comments on the plan by March 8. Of course, by that point, the ongoing government shutdown may start impacting the FDA's workflow.
FDA Commissioner Scott Gottlieb said this week that user fees collected through the end of 2018 can keep the agency going for about one more month. The partial shutdown doesn't allow the FDA to collect 2019 user fees, so if the shutdown continues into February, certain FDA functions will have to cease.
The FDA acknowledged it will have its work cut out managing an "agile" regulatory program—one that leaves room for innovation while maintaining safety.
“Unlike manufacturers of hardware devices who modify their products every few months to years, developers of software modify their products in response to real-world performance and user feedback every few weeks to months," the agency wrote. "It is important for public health to address these distinctive aspects of digital health technology."
The FDA will implement the new model under de novo classification, which is designed to bring new drugs and applications to market exceptionally quickly. On Feb. 7, the agency plans to offer a user session to answer questions from the developer community.