Euro Convergence: Market access pathways for AI now visible

Regulatory NewsRegulatory News | 12 May 2021 |  By 

The pathways companies will use to bring artificial intelligence or machine learning (AI/ML) medical devices and diagnostics to market in the future are now becoming clear, Loganathan Kumarasamy, head of US compliance services at Zifo Technologies, said at RAPS Euro Convergence 2021.
Devices and diagnostics that make use of AI and ML need regulation specific to the field, Kumarasamy said, pointing out that software in medical devices has long been used safely under existing frameworks. Additional regulation for AI/ML is needed because the technologies are adaptive and have advanced capabilities, which are not covered by existing frameworks, creating new risks.
“If we need to monitor or update a medical device that is already in the market, a proper change control process would have to happen,” Kumarasamy said. “But these machine learning techniques have capabilities to modify those, which brings in a lot of patient safety risk.”
Kumarasamy set out how the US and European Union are trying to manage those risks. He described the risk-based pre-certification program the US Food and Drug Administration (FDA) has set up for medical software. The program, which Kumarasamy called a “huge change” for the medical device industry, will enable organizations that show FDA they have a proactive quality culture and organizational excellence to benefit from streamlined regulatory reviews.
The scheme is still at the pilot stage but Kumarasamy sees it coming to shape how FDA regulates AI and ML in medical devices in the future. Over the next 3 to 5 years, Kumarasamy envisages the pre-certification process becoming the first regulatory step for developers of AI/ML medical devices. (RELATED: FDA’s AI/ML action plan includes ‘tailored’ regulatory framework for SaMD, Regulatory Focus 18 January 2021)
“Once you complete the development following the [good machine learning practices] and computer software assurance, if it is a low-risk device...the medical device manufacturer can release it into the market without FDA review, because you are pre-certified. And if it is a high-risk medical device, then you will be eligible for accelerated review,” Kumarasamy said.
Kumarasamy also laid out how he expects the policies of the European Medicines Agency (EMA) and other EU bodies to reshape the regulatory process for AI/ML medical devices on the other side of the Atlantic in the years to come.
The European Commission recently published a proposal for the regulation of AI in the EU, across all industries. The policy will sit alongside the incoming regulations on medical devices and in vitro diagnostics (IVDs), which lack provisions specific to AI and ML. In the document, the Commission proposed treating medical devices and IVDs that use AI or ML as high-risk products that are subject to additional regulation.
EU bodies are still working out the details but Kumarasamy has discerned the broad outline of how the region will regulate AI/ML medical devices. “MDR would act as a base, and then the AI general requirements would be an add on,” Kumarasamy said. “Then, we can assume that [the International Medical Device Regulators Forum] would give us some guidance or policy on how the quality management principles need to be followed for AI-based medical devices.”
Another panelist, Pat Baird, senior regulatory specialist at Philips, shared details of how the World Health Organization (WHO) is approaching AI and ML. Baird said WHO has identified AI and ML as a way to improve healthcare in developing countries, leading it to set up an organization that features a clinical expert panel, engineers developing a benchmarking platform and other groups working on different aspects of what it will take to deploy safe, effective AI and ML systems in healthcare.
RAPS Euro Convergence 2021


© 2023 Regulatory Affairs Professionals Society.

Discover more of what matters to you

No taxonomy