AI tool improves early cancer diagnosis in primary care

AI cancer detection tool C the Signs has reported a 98% sensitivity for cancer, according to interim findings, showing the potential for AI to improve early diagnosis in primary care.

Speaking a Pulse PCN event in Birmingham on 24 April, CEO and co-founder of the AI tool, Dr Bea Bakshi, said there were many opportunities that digital tools present for early diagnosis of cancer, which can then improve survival rates and reduce the costs of treatment for the NHS.

C the Signs, for example, is a tool that works by evaluating the risk factors of patients, including signs, symptoms, clinical markers, genetics, lifestyle and environmental factors, in order to help diagnose cancer at an early stage.

It is currently used in around 1,400 GP practices, and is integrated with SystmOne, EMIS and Vision, for which it has to go through ‘extensive approvals’, according to Dr Bakshi, including as a class 1 medical device with the MHRA.

Interim findings presented to the American Society of Clinical Oncology in October found that C the Signs had a 98% sensitivity for spotting cancer in the 111,421 patients who were risk-assessed by the tool. Of the 7,360 who were diagnosed with cancer, C the Signs spotted 7,257 of them.

It does this by flagging cancer risk and then referring patients to diagnostics. The system runs prospectively in everyday consultations, using both historic data from the electronic medical record, and new data from the patient.

The platform also predicted cancer origin accurately in 93.2% of patients, with most accurate predictions with brain, central nervous system, breast, lung, gynaecological, head and neck, bowel, skin and prostate cancers.

The C the Signs observational study was conducted between January 2021 and October 2022 in the NHS, with no pre-selection criteria for patients assessed.

‘GPs are generalists,’ said Dr Bakshi, who is also a GP. ‘Cancer is less than 2% of our workload, so the average GP only sees eight new cases of cancer per year and one rare case of cancer in [their] lifetime. So, our own internal algorithm is not able to learn at the same pace as AI to say what are the signs and symptoms of cancer, and then be able to recognise that at appointment.’

Recent analysis from The Nuffield Trust and Health Foundation found that one in five people with cancer had to see their GP at least three times before they received a diagnosis.

This research, published 24 April, found that there were ‘stark inequalities’ in the proportion of people diagnosed with advanced cancer.

The authors, Dr Elizabeth Fisher and Jessica Morris, said: ‘Health systems in each area need to work together and with deprived groups; communities of Mixed, Asian and Black ethnicities; and other vulnerable groups to understand and tackle barriers to cancer diagnosis.

‘In particular, improving screening rates would help reduce inequalities for some cancers. More support and tailored responses for some groups, such as people with learning disabilities and trans people, is needed.

‘Younger people are more likely to be diagnosed at an early stage in their cancer, but also more likely to see a GP multiple times before being diagnosed. Enabling patients to see the same GP can help serious problems to be spotted earlier. The NHS must do better in communicating with patients, both prior to diagnosis and once they know whether they have cancer or not. Patients often feel unable to speak up for themselves – something that becomes more important when the whole system is under pressure.’

Access funding

C the Signs as a tool costs between 18p per patient and 28p per patient, according to the government’s catalogue of cloud suppliers, G Cloud.

But Dr Bakshi added that there are funding opportunities for PCNs who want to run initiatives in early diagnosis or work to improve referral practice.

She told attendees: ‘[Cancer] Alliances should be funding your projects,’ she said. ‘If you want to do anything with cancer, their mandate is to delegate funding directly to place, PCN or federation level.’

In the recent Network DES, it stated that PCNs should be working with cancer alliances.

It said: ‘A PCN should work with cancer alliances to increase early diagnosis rates and improve referral practice. Alliances have been allocated funding to support this work.’

A spokesperson for Cheshire and Merseyside Cancer Alliance added: ‘The Cancer Alliance instigates, co-ordinates and funds a wide range of programmes and projects to improve cancer services, treatment and care and one important area of this is working with primary care and PCNs.

‘We do fund projects involving PCNs, but these are usually linked to wider areas of improvement, including raising awareness of cancer and diagnosing it earlier or around fulfilling requirements set out in the national planning guidance or the PCN-DES.’

There are 21 cancer alliances working across England.

The C the Signs study is pending publication in the Journal of Clinical Oncology.