Abstract

Background & Aims Risk scores for hepatocellular carcinoma (HCC) developed in Asians offer poor-moderate predictability in Caucasian patients with chronic hepatitis B (CHB). This nine center cohort study aimed to develop and validate an accurate HCC risk score in Caucasian CHB patients treated with the current oral antivirals, entecavir/tenofovir. Methods We included 1815 adult Caucasians with CHB and no HCC at baseline who received entecavir/tenofovir for ⩾12 months. Using data from eight centers (derivation dataset, n = 1325), a HCC risk score was developed based on multivariable Cox models and points system for simplification. Harrell’s c-index was used as discrimination, bootstrap for internal validation and the data from the 9th and largest center (validation dataset, n = 490) for external validation. Results The 5-year cumulative HCC incidence rates were 5.7% and 8.4% in the derivation and validation dataset, respectively. In the derivation dataset, age, gender, platelets and cirrhosis were independently associated with HCC. The PAGE-B score was developed based on age, gender and platelets (c-index = 0.82, 0.81 after bootstrap validation). The addition of cirrhosis did not substantially improve the discrimination (c-index = 0.84). The predictability of PAGE-B score was similar (c-index = 0.82) in the validation dataset. Patients with PAGE-B ⩽9, 10–17, ⩾18 had 5-year cumulative HCC incidence rates of 0%, 3%, 17% in the derivation and 0%, 4%, 16% in the validation dataset. Conclusion PAGE-B, which is based only on baseline patients’ age, gender and platelets, represents a simple and reliable score for prediction of the 5-year HCC risk in Caucasian CHB patients under entecavir/tenofovir. Risk scores for hepatocellular carcinoma (HCC) developed in Asians offer poor-moderate predictability in Caucasian patients with chronic hepatitis B (CHB). This nine center cohort study aimed to develop and validate an accurate HCC risk score in Caucasian CHB patients treated with the current oral antivirals, entecavir/tenofovir. We included 1815 adult Caucasians with CHB and no HCC at baseline who received entecavir/tenofovir for ⩾12 months. Using data from eight centers (derivation dataset, n = 1325), a HCC risk score was developed based on multivariable Cox models and points system for simplification. Harrell’s c-index was used as discrimination, bootstrap for internal validation and the data from the 9th and largest center (validation dataset, n = 490) for external validation. The 5-year cumulative HCC incidence rates were 5.7% and 8.4% in the derivation and validation dataset, respectively. In the derivation dataset, age, gender, platelets and cirrhosis were independently associated with HCC. The PAGE-B score was developed based on age, gender and platelets (c-index = 0.82, 0.81 after bootstrap validation). The addition of cirrhosis did not substantially improve the discrimination (c-index = 0.84). The predictability of PAGE-B score was similar (c-index = 0.82) in the validation dataset. Patients with PAGE-B ⩽9, 10–17, ⩾18 had 5-year cumulative HCC incidence rates of 0%, 3%, 17% in the derivation and 0%, 4%, 16% in the validation dataset. PAGE-B, which is based only on baseline patients’ age, gender and platelets, represents a simple and reliable score for prediction of the 5-year HCC risk in Caucasian CHB patients under entecavir/tenofovir.