MHRA responds to report on equity in medical devices

This review was first commissioned because oximeters used in the UK were less accurate for those with darker skin tones, leading to a risk of low blood oxygen not being picked up as quickly for these patients. The paper addressed disparities in access to medical devices and set out 18 recommendations to ensure that everyone benefits from technological advancements in healthcare.

Independent review executive summary

The report focused on three types of medical devices that were considered to be particularly susceptible to unfair biases, these being:

• Optical medical devices (including pulse oximeters)
• Medical devices that are assisted by artificial intelligence (AI)
• Polygenic risk scores (PRS) based on genomic screening

The report's high-level findings were that a systemic change was needed due to the risk of biases being inserted at any stage of the medical device life cycle. As such, the recommended changes could affect all involved in the rollout of medical devices.

Recommendations and responses

Some of the key takeaways from the 18 recommendations are summarised as follows:

• Pulse oximeters and optical devices: the report recommends regulators, developers, manufacturers and healthcare professionals take immediate mitigation measures for pulse oximeter bias. The standards for approval of new pulse oximeters should be strengthened to show accuracy on darker skin tones. It also encourages cooperation for better design that produces devices that are not biased. These recommendations are widened in scope to all medical devices.

The Government accepted the issue and risk and has started work on mitigating inaccuracies. It proposed seeking changes to the ISO 80601-2-61:2017 standard to better harmonise standards internationally. It also announced that as part of its reform of UK medical device regulation, it is working to implement changes to strengthen the regulatory framework, core aspects of which it expects to be in place by 2025. The MHRA will also investigate the risk of biases in other medical devices.

• Preventing bias in AI-assisted medical devices: The review calls for improved data collection on patient demographics related to medical device use. Goals of equity, fairness and transparency should be present throughout a device's lifecycle. The report recommends that the NHS influences the deployment of equitable AI-enabled devices through its purchasing power. It also recommends that public funding for research should have equity as a critical driver for investment decisions.

The Government will conduct a gap analysis of the Digital, Artificial Intelligence and Robotics Technologies in Education (DART-Ed) programme to identify further work needed. There are no current plans to make the reporting on the diversity of data used to be a legal requirement, but this will be promoted within guidance. The government agreed that there is scope for equity to be considered within NHS procurement processes but that they would take a proportionate approach that would not excessively limit the number of devices available in the NHS. As a result, careful consideration is needed. Regarding research funding, the government responded that it had created the INCLUDE framework (a guide by the NIHR to help researchers work with under-served groups and to improve their access to clinical research), to factor in equity to such decisions.

• Looking forward—polygenic risk scores: The report recommends widening PRS studies to include lifestyle and other factors that are interlinked with PRS. It recommends undertaking a research programme to understand public perspectives of PRS and how risks are interpreted. Healthcare professionals should have guidance developed to better understand the limitations of PRS testing in patient care and population health.

The Government confirmed that it will continue to promote diversity in genomic research, such as in the case of Our Future Health and Genomics England's Diverse Data Initiative. In collaboration with NHS Race and Health Observatory, Genomics England will develop a common framework to identify and address potential sources of bias in care. PRS are not currently used in the NHS. However, if they are, appropriate guidance will be made available to reflect that PRS may not be as accurate for populations not from European ancestries.

Conclusion

The report makes it clear that due to the risk of bias being introduced at any stage of a medical device's life cycle, all those involved in this life cycle need to be aware of these risks and how to address them. Concerning medical devices generally, the government's confirmation that core aspects of the regulatory framework are to be updated as part of the regulatory changes being made is likely to have the most widespread effect.

Regarding AI as a medical device, for now, additional reporting requirements on diversity data will not be put on a legal footing. It will be interesting to see what further information comes out in the MHRA's Software and AI as a Medical Device Change Programme. According to the roadmap, we can expect MHRA guidance on AI as a medical device later this year.

With such changes on the horizon, businesses working with medical devices should consider building in measures into the development of the device to mitigate the risk of changes, such as adding specific provisions to require that equity in use form part of the development process. Such measures may increase devices' prospects for procurement and leave business better placed to accommodate the medical device reforms in the pipeline.

If you have any medical device related queries that you would like to discuss, please contact Jonathan Bywater in our Pharmaceuticals and Life Sciences team on 07918 496 316, or complete the form below.