Development and validation of a new predictive model for the immune tolerance stage of chronic HBV infection based on the liver histopathological changes.

OBJECTIVE

To identify clinical and viral indicators for the development of a new model to accurately differentiate the stages of chronic hepatitis B virus (HBV) infection based on histopathological changes in the liver.

METHODS

Clinical and liver pathology data from chronic hepatitis B (CHB) patients who underwent liver biopsy were retrospectively collected. The patients were allocated into test and validation groups. The area under the receiver operating characteristic (ROC) curve (AUC) was calculated to idneitfy the optimal diagnostic value for differentiating the stages of chronic HBV infection.

RESULTS

A total of 118 patients and 73 patients who met the diagnostic and inclusion criteria were selected as the test group and validation group, respectively. Multivariate analysis revealed that HBeAg was independently correlated with the IT and IC stages. The cutoff value of HBeAg used to quantitatively differentiate between IT and IC was 1335 S/CO. The AUC values were 0.921 (95% confidence interval (CI): 0.836-0.971) and 0.846 (95% CI: 0.726-0.967) in the test and validation groups, respectively. A new prediction model of the IT stage was established by using three indicators, namely, HBeAg, HBsAg and HBV DNA. The AUC values were 0.923 (95% CI: 0.864-0.982, p < 0.001) and 0.89 (95% CI: 0.787-0.994, p < 0.01) in the test and validation groups, respectively, when this prediction model was used. For the new model, CMA guidelines (2019 version), EASL guidelines (2017 version) and AASLD guidelines (2018 version), the error rates in the test group were 4.65%, 11.62%, 23.26%, and 46.51%, respectively, while the errors rates in the validation group were 20.0%, 25.0%, 40.0%, and 45.0%, respectively.

CONCLUSIONS

High levels of HBeAg, rather than HBeAg positivity, may serve as a predictor of the IT stage. A predictive model for the immune tolerance stage was established by combining three indicators. Compared with the recommended standards from multiple current guidelines, the new prediction model has a significantly lower error rate.