Background and goals: Liver fibrosis influences treatment and surveillance strategies in chronic hepatitis B (CHB). This multicenter study aimed to examine the accuracy of serum fibrosis models in CHB patients including those with low alanine aminotransferase (ALT) levels and serially in those undergoing treatment.

Method: We examined noninvasive fibrosis models [Hepascore, Fibrotest, APRI, hepatitis e antigen (HBeAg)-positive and -negative models] in 179 CHB patients who underwent liver biopsy and fibrosis assessment by METAVIR and image morphometry. Serial Hepascore measurements were assessed in 40 subjects for up to 8.7 years.

Results: Hepascore was more accurate than Fibrotest [area under the curve (AUC) 0.83 vs. 0.72, P = 0.05] and HBeAg-positive model (AUC 0.83 vs. 72, P = 0.03) for significant fibrosis but was not significantly different to APRI or HBeAg-negative scores. Fibrosis area assessed by morphometry was correlated with Hepascore (r = 0.603, P < 0.001), Fibrotest (r = 0.392, P = 0.03), and HBeAg-positive (r = 0.492, P = 0.001) scores only. Among 73 patients with an ALT <60 IU/L, noninvasive models were useful to predict fibrosis (PPV 80–90%) or exclude significant fibrosis (NPV 79–100%). Hepascore increased significantly among patients monitored without treatment and reduced among patients undergoing therapy (0.05/year ± 0.03 vs. −0.04/year ± 0.02, P = 0.007).

Conclusions: Serum fibrosis models are predictive of fibrosis in CHB and assist in identifying subjects with low–normal ALT levels for treatment.