Five guiding principles for machine-learning-enabled medical devices from MHRA, FDA and Health Canada

MHRA, along with international partners the FDA and Health Canada, has published five guiding principles for the development of predetermined change control plans (PCCPs), with the aim of removing “the regulatory burden for developers of machine-learning-enabled medical devices”.

The guidance highlights that PCCPs allow manufacturers to demonstrate the changes and updates they would seek to make, along with how they plan to ensure that safety and effectiveness are maintained, without need for regulatory intervention.

The five principles include that a PCCP must be “focused and bounded”, describing specific changes that a manufacturer plans to implement. They must be risk-based, driven by an approach adhering to the principles of risk management; and they must be evidence-based, demonstrating that “benefits outweigh the risks throughout the product lifecycle”. In addition, a PCCP should be transparent, providing clear information and plans for ongoing transparency for all stakeholders; and it should incorporate a “total product lifecycle perspective”, improving the quality and integrity of a PCCP by maintaining awareness of the perspective of all stakeholders.

The MHRA notes that at present in the UK, when a manufacturer makes any significant updates or changes to their medical device, they are required to notify their conformity assessment body, with the possibility of their device being reassessed to ensure that the changes do not lead to a negative impact on performance or safety. The nature of medical devices using AI and machine learning means that frequent updates are possible, the guidance adds, which can lead to a “potentially lengthy reassessment every time a change is made”.

The five principles build on the 10 guiding principles for Good Machine Learning Practice, and outline the areas where the MHRA, FDA and Health Canada are “aligned in their expectations of an acceptable PCCP to reduce or remove the need for reassessment”. Each regulator will have their own specific national guidance for manufacturers to follow, with the MHRA set to publish this guidance in 2024.

Dr Paul Campbell, head of software and AI at the MHRA, said: “AI and machine-learning-enabled medical devices are becoming more prevalent, and regulators must adapt their processes to support innovations for patients while continuing to ensure their safety. By collaborating with the FDA and Health Canada on these guiding principles, we can clearly outline where we align on our expectations for a successful change control plan and help reduce the regulatory burden for manufacturers. Collaboration between regulators on these guiding principles demonstrates how working with international partners can help the development of agile regulatory processes that support innovative manufacturers and patients globally.”

Late last year, the MHRA updated its “Software and AI as a Medical Device Change Programme”, which aims to ensure regulatory requirements for software and AI are clear and that patients are protected.

In May of this year, the FDA released a discussion paper designed to generate discussion about artificial intelligence and machine learning in drug development and manufacturing, entitled “Using Artificial Intelligence & Machine Learning in the Development of Drug & Biological Products”.