Abstract
BACKGROUND: Early identification of significant liver pathology is crucial for timely antiviral intervention in individuals with chronic hepatitis B (CHB) infection. Current non-invasive methods show limited accuracy in detecting occult liver damage, particularly in those with normal ALT. This study evaluated serum N-glycan profiles for diagnosing significant liver pathology in treatment-naïve CHB patients across clinical phases. METHODS: This cross-sectional study analyzed 626 treatment-naïve CHB patients confirmed by liver biopsy, classified according to 2025 EASL guidelines. Serum N-glycan profiles were determined using DNA sequencer-assisted fluorophore-assisted carbohydrate electrophoresis. Significant liver pathology was defined as inflammation grade ≥ G2 and/or fibrosis stage ≥ S2 (per Scheuer scoring system). Multivariate logistic regression models were developed and compared with traditional non-invasive markers. RESULTS: Among 626 CHB patients, 66.0% had significant inflammation and 58.9% had significant fibrosis. Patients with significant pathology showed characteristic alterations, with elevated P1, P3, P6, P7, P11 peaks and decreased P0, P5, P8, P10 peaks (all p < 0.0001). Compared to respective infection phases, hepatitis phases showed P1 increases of 19.6% and 36% in HBeAg(+) and HBeAg(-) patients, with P11 increases of 82.4% and 73.4%, while P0 decreased by 20.3% and 27.6%, and P10 by 21.6% and 20.3%. Relative to mild pathology (G and S < 2), P1 increased by 27% in significant pathology (G and/or S ≥ 2), reaching 58.7%/48.7% in G4/S4 stages (vs. G0/S0). In ALT-normal HBeAg(+) infection phase, P1 increased by 80.2%/65.8% in G4/S4 stages (vs. G0/S0), with P2 also increasing by 54.1%/45.2%. Multivariate analysis identified P11 as strongest risk factor (OR = 3.84, 95%CI: 1.74-8.45, p = 0.0008), followed by P1 (OR = 2.04, 95%CI: 1.57-2.64, p < 0.0001) and P7 (OR = 1.75, 95%CI: 1.31-2.34, p = 0.0002), while P2 (OR = 0.07, 95%CI: 0.02-0.26, p < 0.0001) and P0 (OR = 0.30, 95%CI: 0.12-0.79, p = 0.0140) served as protective factors. The glycomics combined model (AUC = 0.876 (0.844-0.908)) achieved superior performance and outperformed the clinical model (AUC = 0.818 (0.779-0.857)), LSM (AUC = 0.817 (0.775-0.858)), APRI (AUC = 0.830 (0.792-0.867)), and FIB-4 (AUC = 0.672 (0.621-0.723)) (all p < 0.001), with 78.7% sensitivity and 83.2% specificity. The optimized model reached AUC = 0.917 (0.891-0.942) with accuracy 84.2%, with 78.7% sensitivity and 94.6% specificity. Both glycomics-based models maintained diagnostic capability in ALT-normal patients particularly in HBeAg(+) infection. CONCLUSIONS: Serum N-glycomics demonstrates promising potential for non-invasive identification of significant liver pathology in treatment-naïve CHB patients, providing an alternative approach for early treatment decisions, especially in ALT-normal patients with occult liver damage.