Baseline low-level viremia in HBV-related hepatocellular carcinoma: prognostic significance, risk stratification, and clinical implications.

Introduction: The prognostic relevance of baseline low-level viremia (LLV) after hepatitis B virus-related hepatocellular carcinoma (HBV-HCC) diagnosis remains uncertain. We evaluated the association between baseline LLV and overall survival (OS), explored related biological features, and developed survival prediction models for patients with baseline LLV. Methods: This retrospective study included 921 patients with HBV-HCC: 611 with baseline LLV and 310 with HBV DNA <20 IU/mL. Propensity score matching (PSM), inverse probability of treatment weighting (IPTW), and Cox regression were used to assess the association between baseline HBV DNA status and OS. Peripheral T cell subsets and selected markers of hepatic functional reserve, hepatocellular injury, and systemic inflammation were explored. Patients with baseline LLV were randomly divided into training and reserved internal validation sets. Model 1 used the full candidate variable set, whereas Model 2 used non-imaging variables. Predictors were selected using LASSO Cox regression and Boruta, and five survival algorithms were compared within the training set. Results: Baseline LLV was associated with poorer OS. After PSM, HBV DNA <20 IU/mL was associated with lower adjusted mortality than LLV (hazard ratio, 0.688; 95% confidence interval, 0.543-0.872; P = 0.002), with similar findings after IPTW. Patients with LLV had a lower peripheral CD4/CD8 ratio, and exploratory analyses identified significant indirect associations through albumin, aspartate aminotransferase, and systemic immune-inflammation response index. Model 1 retained tumor size, tumor multiplicity, portal vein tumor thrombus, cholinesterase, alkaline phosphatase, and alpha-fetoprotein; Model 2 retained gamma-glutamyl transferase, cholinesterase, alkaline phosphatase, and alpha-fetoprotein. Random survival forest was selected based on overall training performance. In the reserved validation set, the 1-, 3-, and 5-year time-dependent AUCs were 0.824, 0.777, and 0.765 for Model 1 and 0.769, 0.742, and 0.723 for Model 2, respectively. Both models showed good calibration and positive net benefit. Discussion: Baseline LLV remained prognostically relevant after HCC diagnosis and was accompanied by exploratory differences in systemic inflammation and the relative balance of peripheral T cell subsets. These findings support closer virological monitoring and individualized risk assessment.