Future Directions in Medical Device Research: AI/ML and Cybersecurity
The future of medical device research is focused on two areas, AI/ML and Cybersecurity. This brief post focused on AI/ML. A second will focus on Cybersecurity.
The folks at Greenlight Guru host an informative podcast on Global Medical Devices. They produced a recent episode on AI/ML (Artificial Intelligence/Machine Learning) with Mike Drues, President of Vascular Sciences, in which they discussed recent developments in AI/ML and their implications for medical device production, function and review. In April 2023 the FDA released a Draft Guidance for Industry document titled “Marketing Submission recommendations for a Predetermined Change Control Plan for Artificial Intelligence/Machine Learning (AI/ML)-Enabled Device Software Functions” highlighting the things to consider and account for when seeking to market medical devices informed by AI/ML.
AI refers to a technology’s ability to change and evolve on its own. While to some degree, there is little new other than the use of software in these changes, AI/ML may nonetheless still affect any of the core questions we ask when determining the need for FDA clearance and pre-market approval. When seeking clearance based on substantial equivalence, any difference in technology cannot introduce new questions of safety and efficacy, or the overall risk. If we can pose a convincing argument to the FDA that the introduction of AI does not introduce new questions in the areas of safety, efficacy and risk, then a case for substantial equivalence may be supported. If this is not the case, then a De Novo request may be appropriate. In some cases, a De Novo request may make more sense and be easier in the long-run. Existing parameters determined by a previously approved device may introduce unwanted constraints to later modifications and related devices. Future iterations may be hindered by the initial substantial equivalence framing. Further, if AI/ML is modeled after a previously-cleared device, but modifications to the device introduce changes to the user interface, users, uses, use environments, training and/or labeling, you’ll need human factors research for pre-FDA market approval.
These are questions you’d want to discuss with your research team or consultants. It may be worth setting up a pre-submission meeting with the FDA early on, as this will help you avoid preventable problems that can derail your timeline. Moreover, it allows you the space to tell regulators about your device and to control the initial framing. A huge advantage to the speed and success of your later clearance requests and pre-market approvals.