Development and Validation of a Blood-Biomarker-Based Predictive Model for HBV-Associated Hepatocellular Carcinoma

Abstract

Objective This study aims to explore the optimal predictors of HBV-associated HCC using Lasso, and establish a prediction model. Methods A retrospective analysis was conducted on patients who underwent CBC and CRP testing between January 2016 and March 2024. The study population comprised 5441 cases divided into three cohorts: non-HBV-infected (1333 cases), HBV-infected (1023 cases), and HBV-associated HCC (3085 cases). A value of CRP <10 mg/L was used to exclude cases of acute bacterial infections. Baseline data and blood parameters were compared across the three groups (control group (n = 1049), the HBV-infected group (n = 789), and the HBV-associated HCC group (n = 1367)). HBV-infected group and the HBV-associated HCC group were used as modeling subjects which 70% were classified as training set (n = 1512) and 30% were classified as validation set (n = 644). Lasso regression and logistic regression were employed to identify the most effective predictors of HBV-associated HCC, which were subsequently incorporated into a predictive model by training set. Results Significant variations in age, gender, and blood parameter indices were observed between individuals with acute bacterial infections and non-infections in the study population, and also between three groups. The optimal predictors identified for HBV-associated HCC included gender, age, MONO, EO%, MCHC, MPV, and PCT. Conclusions The study highlights the significant impact of acute bacterial infections on immune status, erythrocyte system, and platelet system. After excluding acute bacterial infections, factors such as gender, age, MONO, EO%, MCHC, MPV, and PCT are effective predictors for clinical prediction of HCC development in HBV-infected patients.