The Medicines and Healthcare products Regulatory Agency (MHRA) has won over £900,000 in funding for two projects exploring the use of AI in clinical trials.
The executive agency of the Department for Health and Social Care has received the total of £970,688 from the Regulators’ Pioneer Fund run by the Department of Business, Energy and Industrial Strategy (BEIS) to support innovation in regulation.
The projects are aimed at introducing complex AI safely into clinical settings.
The largest share of the money – £750,387 – is going to explore the potential of synthetic datasets to replace control groups of real patients in clinical trials. It is aimed at creating artificial control groups for clinical trials , which would lower their costs and improve how new treatments are tested before applied in the NHS.
Currently, patients could be randomly assigned to control groups and denied access to treatments that could improve their symptoms or extend their lives, and it is often difficult to recruit enough patients in trials for rare diseases.
In response, MHRA has made progress in creating artificial patients with similar health information to real ones. This could boost the number for small trials and better reflect groups who are not well represented.
It is thought the project will take 12-18 months to complete.
High fidelity data
Dr Puja Myles, director of the Clinical Practice Research Datalink at MHRA, said: “This provides researchers with high fidelity data that replicates the complex clinical relationships in real primary care data while protecting patient privacy, as they are wholly synthetic.
“This funding could revolutionise how clinical trials are conducted, giving patients access to potentially life changing treatments, saving money that could be better spent on more advanced research while providing highly reliable results for the clinical researchers.”
The second project, receiving £167,863, is aimed at safely introducing ‘black box’ AI products into clinical settings so that clinicians will be confident that the devices are helping them to make suitable and appropriate decisions.
It will focus on producing a workable methodology for clinicians to understand the factors in the algorithms and the weight assigned to them so they can assess whether or not to overrule it. This will take forward elements of the recently published Software and AI as Medical Device Change Programme.
Professor Johan Ordish, head of software and AI at MHRA, said: “AI as a medical device is crucial in supporting the development of next-gen healthcare. However, the transparency and explainability of AI has a key relationship with the safety as well as usability of these devices, impacting whether clinicians adopt and trust AI as a medical device.
“Patient safety is at the heart of everything we do, and we’re determined to be well prepared to get AI technology to the bedside safely, effectively and with confidence in this assurance. By producing a methodology that supports the regulation of both transparent and complex AI models, it will help us to make this vision a reality.”