Medical device software and AI – key new guidance

We now have a new key document providing guidance on the EU Medical Devices Regulation (MDR), which comes into full force in 2020. While it’s a highly technical document, developers and researchers targeting medical device software and AI algorithms with medical purposes need to understand its contents.

The Medical Devices Regulation (EU Regulation 2017/745) is a major update to the way that medical devices are regulated in Europe. Nearly 20 years after the approval of its predecessor (the Medical Devices Directive, MDD), the MDR is a substantial step forward in regulation, tightening many loopholes and introducing new processes that aim to improve safety and transparency.

The original MDD had basic provisions for how software could form part of a medical device, either as a component, an accessory, or a device on its own. Over the years, it was tightened up and improved through further EU directives, consensus statements, informational documents from competent bodies like the UK’s MHRA, and in particular the Guidance MEDDEVs. Together, these provided pretty good coverage for most applications, and procedures to provide clarification in borderline or new cases. The MDR completely replaces these with a new set of processes.

But for software medical devices, we’ve been missing the guidance that notified bodies need to interpret the MDR – the equivalent of MEDDEV 2.1/6. We knew that regulators are generally converging on something similar to the 2014 IMDRF SaMD framework, but that has no legal force and arguably fudged some critical distinctions. The new document fills this gap.

The new MDCG 2019-11 Guidance, issued in October 2019, provides flowcharts and case studies to help illustrate the rules encoded in MDR Annex VIII and the corresponding rules of the In-Vitro Devices Regulation (IVDR). If you’re looking for light-touch regulation or think that your new app or algorithm has an easy ride to Class I self-certification, it’s bad news. We’re now working through its implications for our own development of AI for cancer diagnosis, but welcome the clarity that it provides and highlight that it is essentially continuing the MDR’s drive towards improved patient safety.

The key implication is that many software applications (including AI, ML and statistical tools like risk calculators) will fall into Class IIa or Class IIb, and for high-risk devices or poorly drawn claims, potentially the most highly regulated Class III.

Ethics and regulation for medical devices

Details of applications vary greatly, so you need to get specialist advice, as early as possible, to understand the impact of the new guidance and to plan your route to CE marking. Contact us if we might be able to help you.

Our CEO, Dr Antony Rix, will be discussing this further in his talk at the Cambridge Medical Device Developers’ Meetup on 26 November on Ethics & Regulation for medical devices: making software and AI good for clinical use.

Details and registration

Further reading on medical device software and AI