Effectiveness of Antiviral Treatment in Cirrhotic Patients with Low-level Hepatitis B Virus DNA Levels

Multicenter, Open-label, Single arm Trial with Matched Historical controls. Male and female adults with compensated liver cirrhosis due to chronic hepatitis B virus infection who have low-level viremia. To assess the efficacy of Tenofovir Alafenamide (TAF) in reducing liver-related events (hepatocellular carcinoma, liver-related events and death, decompensated liver cirrhosis) in cirrhotic chronic hepatitis B patients with low-level viremia compared with matched historical controls.

1. This clinical trial is a multicenter, open label, single arm study in cirrhotic chronic hepatitis B patients with low-level viremia. 2. Approximately 200 subjects meeting eligibility criteria will be enrolled and randomized (1:2) to Treatment Arm (A) or , Matched Historical Controls Arm (B), as below: * Treatment Arm (A): 200 subjects, TAF 25mg once daily with food for 3 years * Matched Historical Controls Arm (B): 400 subjects, patients who did not receive antiviral treatment during their follow-up period, and matched with the treatment group in a 1:1 ratio according to their baseline characteristics 3. Treatment Arm is scheduled to be followed up to 3 years. The analysis population of this study consists of the Full Analysis Set (FAS), which can evaluate efficacy. The treatment group includes patients who received any dose of study medication (TAF), while the control group consists of patients who did not receive antiviral treatment and were matched by propensity score (PS) from those under observation. Unless otherwise specified, the FAS will be used for primary and secondary efficacy endpoints. Statistical analyses for the primary efficacy endpoint will be performed at a two-sided significance level of 5%, aligned with the sample size calculation. PS will be estimated using logistic regression, incorporating variables such as age, gender, HBeAg positivity, HBV DNA level, ALT, platelet count, albumin, total bilirubin, creatinine, prothrombin time, diabetes, hypertension, and family history of hepatocellular carcinoma. The groups will be matched 1:2 using nearest neighbor matching. Baseline characteristics between the matched groups will be compared using standardized mean difference (SMD), with an absolute SMD \<0.1 indicating good balance. The 3-year cumulative incidence of the primary endpoint will be estimated using the Kaplan-Meier method. For comparisons between PS-matched groups, a Cox proportional hazards model with robust variance estimation will be used to account for matching, reporting HR and 95% confidence intervals. For hepatocellular carcinoma incidence, mortality, liver transplantation, decompensated liver function (Child-Pugh score ≥8), and cirrhosis-related complications, 1-, 2-, and 3-year incidence rates will be estimated with Kaplan-Meier. Cox models with robust variance will be used for group comparisons, reporting HR and 95% confidence intervals. HBsAg loss and viral response (HBV DNA \<15 IU/mL) are binary outcomes, summarized by frequency and proportion in each group. Comparisons will be made using a Poisson regression model with Generalized Estimating Equations (GEE), reporting relative risk (RR) and 95% confidence intervals. Fibroscan changes, treatment response, duration, and health-related quality of life by disease status (hepatitis, cirrhosis, hepatocellular carcinoma) will be analyzed using GEE, accounting for the matched-pair design. In the PS-matched retrospective control group, the 1-, 2-, and 3-year cumulative probabilities of initiating antiviral therapy due to serum HBV DNA ≥2,000 IU/mL will be estimated via Kaplan-Meier. Two official analyses will be conducted: an interim analysis (using data collected by October 2024) and a final analysis. No correction for type 1 error inflation from multiple testing will be applied in the interim analysis. For sensitivity analysis, multivariable analyses will be performed using the unmatched dataset for primary and secondary endpoints. Depending on outcome type, Cox models and Poisson regression will be employed, using the covariates from the PS estimation for adjustment.