New Algorithms Enhance Cancer Detection in GP Consultations

Researchers from Queen Mary University of London and the University of Oxford have developed two advanced predictive algorithms that use patient health information and routine blood tests to significantly improve early detection of undiagnosed cancers, including hard-to-diagnose types like liver and oral cancer. Using anonymized health records from over 7.4 million adults in England, these models are more sensitive than existing tools, such as the NHS’s QCancer scores, and can identify individuals at high risk for 15 different cancer types. The new algorithms incorporate additional risk factors, including four previously unrecognized medical conditions, two more family history associations, and seven new symptoms linked to cancer risk. They also uniquely utilize seven routine blood tests—measuring full blood count and liver function—as biomarkers, further enhancing their diagnostic accuracy. External validation in multiple UK populations demonstrated superior performance in sensitivity, calibration, and clinical utility compared to older models. These advancements could enable general practitioners to identify and treat cancer at much earlier stages, potentially improving clinical outcomes in the UK.