Former Director at CDRH Says Pre-Determined Change Plans Not Needed for AI/ML

While most comments on the FDA’s draft guidance on providing predetermined change plans for devices that include use of AI or machine learning (ML) offer improvements of one sort or another, one commenter doesn’t believe such change plans are even needed.

Victor Krauthamer, adjunct professor in medical device evaluation and rehabilitation medicine engineering at George Washington University and former director of CDRH’s Division of Biomedical Physics, commented that “it would be dangerous to permit predetermined change plans for such software changes in medical devices,” since there are “major questions about bias, trustworthiness, understandability and reliability of ML/AI.”

“The predetermined change control plan may give companies less incentive to get their software correct with initial approval or clearance,” Krauthamer states.

Related Topics