To evaluate the predictive value of different risk models for hepatocellular carcinoma (HCC) in patients with chronic hepatitis B (CHB)-related cirrhosis receiving antiviral therapy with entecavir (ETV), tenofovir disoproxil fumarate (TDF), and tenofovir alafenamide fumarate (TAF), and to clarify the effects of pharmacological differences among various nucleos(t)ide analogs (NAs) on HCC risk, so as to provide a reference for clinical diagnosis and treatment.

A total of 252 treatment-naive patients with CHB-related cirrhosis without a history of HCC who received NA antiviral therapy from May 2015 to May 2020 were retrospectively enrolled. They were divided into the ETV group (n=108), TDF group (n=96), and TAF group (n=48) according to the NA used. Cox proportional hazards model was used to construct an HCC risk prediction model. Patients were classified into low-risk group (0-3 points), intermediate-risk group (4-7 points), and high-risk group (8-13 points) based on the Asia-Pacific Association for the Study of the Liver (APASL) REALB score. All patients continued to receive corresponding NA therapy and were followed up until the study endpoint. The predictive performance of each risk score was compared using time-dependent area under the receiver operating characteristic curve (AUROC), and the differences in HCC incidence among different NA groups were analyzed. The main outcome measures included HCC incidence, international normalized ratio (INR), alpha-fetoprotein (AFP), diabetes mellitus, drinking history, predictive performance of the REALB score, and prognosis-related effects of different NAs.

A total of 252 patients were screened, and finally 108 patients were assigned to the ETV group, 96 to the TDF group, and 48 to the TAF group. The median follow-up duration was 56.96 months, and hepatocellular carcinoma (HCC) was diagnosed in 19.00% (48 cases/252 cases) of patients. The 1year incidence rates of HCC in the ETV, TDF, and TAF groups were 4.63% (5 cases/108 cases), 5.21% (5 cases/96 cases), and 4.17% (2 cases/48 cases), respectively; the 3 year rates were 12.96% (14 cases/108 cases), 13.54% (13 cases/96 cases), and 12.50% (6 cases/48 cases), respectively; and the 5 year rates were 17.59% (19 cases/108 cases), 18.75% (18 cases/96 cases), and 16.67% (8 cases/48 cases), respectively. No statistically significant differences were observed among the three groups (all P>0.05). Multivariate stepwise regression analysis revealed that INR (HR=2.77, 95% CI: 1.46-5.25, P<0.01), AFP (HR=1.00, 95% CI: 1.00-1.00, P<0.05), diabetes mellitus (HR=3.06, 95% CI: 1.54-6.08, P<0.01), and history of alcohol consumption (HR=2.25, 95% CI: 1.04-4.86, P<0.05) were independent risk factors for HCC development. The predictive performance of the REALB score at 3 years (AUROC=0.74) and 5 years (AUROC=0.70) remained higher than that of other models except the RWSHCC model, while its 1year AUROC was similar to those of other models. Statistically significant differences in the 1, 3, and 5 year HCC incidence rates were found between the intermediaterisk group and the highrisk group (all P<0.001).

Different NAs (ETV, TDF, TAF) have comparable efficacy in controlling HCC risk in patients with CHB-related cirrhosis, and differences in their pharmacological properties do not significantly affect HCC risk. The REALB score shows stable and excellent predictive ability for HCC in patients receiving NA antiviral therapy and can be used as a preferred tool for clinical prognosis evaluation.