Lower Serum Albumin Level: A Prospective Risk Predictor of Hepatocellular Carcinoma in Patients With Chronic Hepatitis B Virus Infection

Hepatocellular carcinoma (HCC) is one of the most common malignant tumours in China, at high annual incidence and mortality. Chronic hepatitis B virus infection (CHB) is considered as a leading cause to bring about HCC in China. Serum albumin (ALB) level has been adopted to verify its risk with HCC development as a combination variable with other factors. However, the predictive value of a single ALB level on HBV-related HCC risk remained unclear. The aim of this study was to evaluate the prediction ability of serum ALB concentration on the risk of HBV-related HCC development. A prospectively enrolled clinical cohort compromising 2932 cases of CHB patients with at least 1-year exclusion window was selected to explore the predictive role of serum ALB level on incident HCC risk. Baseline clinical data including host characters and laboratory test were collected at the initial period of hospitalisation. The hazard ratio of ALB level associated with HCC development was assessed by Cox proportional hazards regression model using univariate and multivariate analyses. We evaluated the discrimination accuracy of ALB level in predicting HCC development by receiver operating characteristic (ROC) curves. Dose-dependent and time-dependent effects of ALB level on HCC risk prediction were demonstrated, respectively, using a restricted cubic spline and a Fine and Grey competing risk model. Referred to patients with higher ALB level, those with lower ALB level exhibited significantly increased risk of HCC development after adjustment for host variables (dichotomised analyses: hazard ratio = 3.12, 95% confidence interval 1.63-5.97, p = 8.23 x 10(-4), p(log-rank) = 5.97 x 10(-4); tertile analyses: hazard ratio = 2.07, 95% confidence interval 1.63-2.64, p = 3.77 x 10(-9), p(log-rank) < 2.00 x 10(-16); quartile analyses: hazard ratio = 2.10, 95% confidence interval 1.56-2.84, p = 9.87 x 10(-7), p(log-rank) < 2.00 x 10(-16)). There was a statistically increasing trend on HCC risk which was found following by the decrease of ALB level (p(trend) < 0.0001). Similar findings were present by the Kaplan-Meier analysis, cumulative incidences of HCC development were significantly higher in patients with lower ALB levels, with the p value obtained from log-rank test were all < 0.0001. The result of dose-dependent effect showed hazard ratio (HR) value of HCC risk was gradually decreasing as the increasing of ALB level, with non-linear correlation being statistically significant (Wald chi(2) = 20.59, p = 0.000). HR value in lower ALB level remained persistently prominent by fluctuating around 2.73 in the whole follow-up time by adjusting for host variables. Sub-cohort analysis by ROC revealed that the discrimination ability of the ALB model was performed better than Child-Pugh (C-P) model in both cohort of patients with 1-year (area under curve [AUC] 0.762 vs. 0.720) and 2-year exclusion window (AUC 0.768 vs. 0.728). The AUC added by ALB level was demonstrated significantly from host model to full model. Lower ALB level was significantly associated with an increased risk of HBV-related HCC and could provide extra useful clinical utility to other host features, which might be a promising non-invasive indicator for surveillance on HCC development.