Background & aims: Current guidelines recommend biannual surveillance for hepatocellular carcinoma (HCC) in all patients with cirrhosis, regardless of etiology. However, HCC incidence is not well established for many causes of cirrhosis. We aimed to assess the disease-specific incidence of HCC in a large cohort of patients with cirrhosis and to develop a scoring system to predict HCC risk.

Methods: A derivation cohort of patients with cirrhosis diagnosed by biopsy or non-invasive measures was identified through retrospective chart review. The disease-specific incidence of HCC was calculated according to etiology of cirrhosis. Factors associated with HCC were identified through multivariable Cox regression and used to develop a scoring system to predict HCC risk. The scoring system was evaluated in an external cohort for validation.

Results: Of 2,079 patients with cirrhosis and ≥6months follow-up, 226 (10.8%) developed HCC. The 10-year cumulative incidence of HCC varied by etiologic category from 22% in patients with viral hepatitis, to 16% in those with steatohepatitis and 5% in those with autoimmune liver disease (p<0.001). By multivariable Cox regression, age, sex, etiology and platelets were associated with HCC. Points were assigned in proportion to each hazard ratio to create the Toronto HCC Risk Index (THRI). The 10-year cumulative HCC incidence was 3%, 10% and 32% in the low-risk (<120points), medium-risk (120-240) and high-risk (>240) groups respectively, values that remained consistent after internal validation. External validation was performed on a cohort of patients with primary biliary cirrhosis, hepatitis B viral and hepatitis C viral cirrhosis (n=1,144), with similar predictive ability (Harrell's c statistic 0.77) in the validation and derivation cohorts.

Conclusion: HCC incidence varies markedly by etiology of cirrhosis. The THRI, using readily available clinical and laboratory parameters, has good predictive ability for HCC in patients with cirrhosis, and has been validated in an external cohort. This risk score may help to guide recommendations regarding HCC surveillance among patients with cirrhosis.

Lay summary: HCC incidence varies markedly depending on the underlying cause of cirrhosis. Herein, using readily available clinical and laboratory parameters we describe a risk score, THRI, which has a good predictive ability for HCC in patients with cirrhosis, and has been validated in an external cohort. This risk score may help to guide recommendations regarding HCC surveillance among patients with cirrhosis.