AI update – the regulatory approach to AI and medical products in the UK

Overview

The government has set out in its White Paper a "pro-innovation approach to AI regulation" and a cross-industry light touch regulatory framework based on five principles for the safe, transparent, fair, accountable and contestable use of AI (Principles) – see our 2023 insight on the White Paper for details. Rather than proposing a stand-alone piece of AI legislation that would apply across all sectors, the government has tasked individual regulators with implementing the Principles and regulating the use of AI in their specific fields.

At the end of April the MHRA published its response to the White Paper, in which it has considered the impact of AI on the regulation of medical products from three angles: (1) as a regulator of AI products, (2), as an organisation that makes evidence-based decisions that affect public and patient safety and (3) as a public service organisation delivering time-critical decisions. The MHRA has confirmed in this response that it considers AI to be within the remit of the existing regulations, which are already closely aligned with the Principles, but that it will also factor-in AI and its unique challenges to new regulations and guidance (specifically in its programme of regulatory reform for medical devices).

MHRA as regulator of AI products

It is already well established that software can be considered a "medical device" which is therefore regulated by the MHRA under existing legislation (the UK Medical Devices Regulations 2002) and the MHRA has confirmed in this response that, where AI is used for a medical purpose, it is very likely to also fall within the definition of a "medical device". The MHRA is already undertaking a programme of regulatory reform for medical devices generally (we expect to see the new UK regulations put in place next year) and it has confirmed that it this includes ensuring there is proportionate regulation of AI as a medical device (AIaMD) that takes into account the novel risks and benefits of AIaMD without stifling the potential to transform healthcare. The key challenges that the MHRA identifies with respect to regulating the use of these AIaMD products in line with the Principles are:

• Transparency and explainability – it is a specific challenge to ensure that the labelling and information provided with the device adequately sets out how the device works and its intended purpose, and the MHRA has issued some specific guidance to assist manufacturers with this (see here).
• Fairness – the use of AI must ensure that harm is not caused through inherent bias against certain groups in the population e.g. women, ethnic minorities and disadvantage socio-economic groups.
• Accountability and governance – there are already obligations for manufacturers in the existing regulations but these will be strengthened and clarified in the new medical devices regulations. It will also be important to ensure that AI models are designed to meet their intended use and that full traceability and accountability of manufacturers is supported.

The MHRA is also collaborating with international partners, such as the US FDA and Health Canada, to develop best practice and alignment for AIaMD products. 

The MHRA is launching a pilot regulatory sandbox (AI-Airlock) for AIaMD in spring 2024 to enable it to start to identify and address the unique challenges posed by AIaMD from a regulatory perspective and will work with key partners including the UK Approved Bodies, the NHS and other regulators – we expect to hear more about this soon.

The MHRA as an organisation that makes evidence-based decisions that affect public and patient safety

The MHRA is engaging with the pharmaceutical industry and other stakeholders on how AI is used to generate and process evidence for medical products, such as in clinical trials, in silico data, and real world data. The MHRA is also using AI to enhance its own ability to understand the relationship between exposure to medical products and clinical outcomes, and to monitor and report safety incidents. 

The regulator has also clarified that the use of AI by the pharmaceutical industry will not impact on its approach to regulation as the fundamental questions remain the same to determine wither a product is safe or not.

MRHA as a public service organisation delivering time-critical decisions

Finally, the MHRA is looking to its own internal operations and exploring the opportunities and challenges of using AI to improve the efficiency and effectiveness of its regulatory services, across all its functions from regulatory science, to market access, to post-market surveillance and enforcement. It is also developing a data strategy that prioritises applying advanced analytics and AI within the regulator safely and responsibly. It is also specifically looking at how it can use AI to improve the quality of applications for medicines licences by using a machine learning tool to assist with the initial screening of applications for completeness, errors and consistency. The aim is for this to reduce human involvement at the early stages and free up resource for the assessments themselves and therefore speed up the process of applying for medicines licences and carrying out assessments generally.

The MHRA is also looking into developing a medicines website checking tool which will allow members of the public to report websites suspected of selling fake or illegal medicines or devices to the MHRA to protect consumers from fraudulent medical products, and in parallel is looking into prototyping similar products that use AI.
 

Key takeaways

• The MHRA is intending to incorporate its regulation reform and guidance for AIaMD products in the new medical devices regulations – we are in somewhat of a holding pattern for now.
• The MHRA considers its role in ensuring safety, quality and effectiveness of medicines and medical products is closely aligned to the Principles and that it will continue to consider the Principles as a key priority driving regulatory reform, especially in a world where AI is increasingly used in all corners of the industry.
• It will be important to pursue early engagement and dialogue with the MHRA on AIaMD products, especially if they involve novel or complex features, such as adaptive or learning algorithms, predetermined change control plans, or in silico data. Resourcing and expertise with the MHRA may be an issue.
• The potential impact of AI on evidence generation and submission of applications for medical products could be huge and it is worth companies in this industry considering how they can use AI to enhance the quality, robustness, and efficiency of their data and analysis whilst always maintaining the core principles of transparency and explainability.
• One of the biggest challenges to AI in this sector is the ethical/social implications of using AI for medical purposes, and how to address the potential risks of bias, discrimination, or harm to individuals or groups, especially those who are vulnerable or underrepresented.