The risk of liver cancer is increased in patients with cirrhosis and cured hepatitis C virus (HCV) infection. These patients are screened regularly with the aim of detecting any cancers that do develop as early as possible and improving patient outcomes. However, the risk of cancer varies substantially between these individuals. If an individual patient’s risk could be better predicted, clinicians may be able to target their surveillance resources to those who might benefit the most.
Several liver cancer risk prediction models have been developed already but their accuracy has not yet been fully tested. Dr Hamish Innes (Glasgow Caledonian University and a member of the DeLIVER early cancer detection consortium led by University of Oxford’s Professor Ellie Barnes) and colleagues assessed the relative performance of six risk prediction models in two external “validation” cohorts.
Using data from the Scotland HCV clinical database and the STOP-HCV study, the team found that the “aMAP” model performed best in terms of calibration and discriminating between patients who go on to develop liver cancer versus those who do not. They observed that discrimination of the models varied by cohort, age and HCV genotype (for genetic risk scores), and that some risk models underpredicted liver cancer risk. These results highlight the importance of this validation step and further research before models can be used to support clinical decision-making.