The aim of the Health Technology Assessment (HTA) Programme is to ensure that high quality research information on the clinical effectiveness, cost-effectiveness and broader impact of healthcare treatments and tests are produced in the most efficient way for those who plan, provide or receive care from NHS and social care services. The commissioned workstream invites applications in response to calls for research on specific questions which have been identified and prioritised for their importance to the NHS, patients and social care.
What is the clinical effectiveness and cost effectiveness of the integration of Artificial Intelligence (AI) image interpretation into the breast screening pathway?
- Patient group: All women eligible for breast screening. It is the ambition that all women eligible for breast screening would automatically be enrolled within the study. If this were feasible then information about the use of AI in mammogram interpretation will need to be included in screening invite letters, at the time of the mammogram and at other time points if appropriate. Women can choose to not be part of the study at any time in the pathway.
- Intervention: The use of AI in image interpretation within the breast screening pathway. Several different AI systems should be used across centres. The study should explore the use of AI at multiple stages within image interpretation within the screening pathway. Applicants will be expected to outline how AI will be integrated into the screening services IT system. The choice of AI system(s) will need to be fully justified and evidence presented to show that they are ready for use in a large scale evaluation.
- Comparator: Standard breast screening pathway: each mammogram is read by 2 human readers with disagreements resolved using established protocols. Or another appropriate and justifiable comparator.
- Important Outcomes:
- Accuracy of AI systems (including, but not limited to, sensitivity, specificity, positive and negative predictive values)
- characteristics of cancers detected (such as stage), interval cancers
- define the parameters in which AI performs and understand any inconsistencies in sub-populations, such as density of breast tissue, ethnicity
- incidental findings
- additional tests and procedures, as well as costs
- a model of cost-effectiveness of the use of AI for image interpretation at different stages in the pathway
- safe movement of women through the screening pathway
- acceptability to women screened and workforce, impact on reader behaviour
- workforce utilisation
- impact on inequalities
- NHS efficiency
- Setting: All breast screening units across the UK.
- Study design: An appropriately designed platform study. The study will need to outline how it will fit within the current screening IT system. Applicants will need to consider the design of the study to ensure that it allows for the inclusion of AI for image interpretation at several appropriate points within the pathway. Clear stop/go criteria will need to be defined, if appropriate, within the study.
Breast cancer is a leading cause of death among women worldwide. Approximately 2.4 million women were diagnosed with breast cancer in 2015, and 523,000 women died. Many countries have established screening programmes for breast cancer, as treatment is more likely to be successful when detected early and screening has an appropriate balance of harm and benefit. Screening programmes differ slightly by country but in the UK each mammogram is read by 2 expert film readers (radiologists, radiography advanced practitioners and others). If there are differences in opinion or abnormalities are found there is a process of arbitration to decide next steps. There is considerable interest in the use of AI in image interpretation to improve screening outcomes. Of particular interest is investigating which part of the image interpretation pathway AI may add most benefit, for example the use of AI to improve care quality through reducing unnecessary investigations and supporting staff workload. Patient and staff acceptability is key to the deliverability of AI within the breast screening programme and should also be fully explored within the study.
At present, the evidence base for the use of AI in breast screening is largely based on retrospective studies and the data from these studies generally shows that AI performs less well than the combination of human readers. However, there is a large randomised prospective trial currently ongoing in Sweden. This study will randomise 100,000 to either AI-integrated screening mammography interpretations or standard - double reading by radiologists without AI. The result of this study needs to be considered within the design of any potential future UK prospective study. However, a prospective UK based trial is required to ensure that AI works within the UK screening pathway and show applicability to the UK population directly.
The HTA Programme is interested in funding a prospective evaluation of the use of AI in image interpretation within the breast cancer screening pathway. The study will need to ensure that several AI systems from different manufacturers are utilised for image interpretation at different stages of the pathway.
All applicants are requested to contact NHS England to discuss their proposed application to ensure that the proposed study will be viable within the national IT systems. Applicants are encouraged to contact NHS England early on in the development of their application via email: firstname.lastname@example.org.
Applications including NHS Scotland may also wish to contact: email@example.com.
Making an application
If you wish to submit a Stage 1 application for this call, the online application form can be found on the Funding opportunities page. To select this call, use the filters on the right of the screen or search using the call name and/or number.
Your application must be submitted on-line no later than 1pm on the 15 November 2023. Applications will be considered by the HTA Funding Committee at its meeting in January 2024.
Please note, shortlisted Stage 1 applicants will be given 8 weeks to submit a Stage 2 application. The Stage 2 application will be considered at the Funding Committee in May 2024.
Applications received electronically after 1pm hours on the due date will not be considered.
For commissioned topics, the Programme strongly discourages the practice of the same Co-Applicant joining more than 1 competing team. There may be unusual circumstances where the same person could be included on more than on application, for example a lead from a named charity or a unique national expert in a condition.
For such exceptions, each application needs to state the case as to why the same person is included - the shared Co-Applicant should not divulge application details between teams and both teams should acknowledge in their application that they are aware that 1 of their Co-Applicants is part of a competing application and that study details have not been shared.
Should you have any queries, please email: firstname.lastname@example.org.