Identification of pseudo-immune tolerance for chronic hepatitis B patients: Development and validation of a non-invasive prediction model

Background and aimsPatients with chronic hepatitis B (CHB) in the immune tolerant (IT) phase were previously thought to have no or slight inflammation or fibrosis in the liver. In fact, some CHB patients with normal ALT levels still experience liver fibrosis. This study aimed to develop and validate a non-invasive model for identifying pseudo-immune tolerance (pseudo-IT) of CHB by predicting significant liver fibrosis. MethodsThis multi-center study enrolled a total of 445 IT-phase patients who had undergone liver biopsy for the training cohort (n = 289) and validation cohort (n = 156) during different time periods. A risk model (IT-3) for predicting significant liver fibrosis (Ishak score >= 3) was developed using high-risk factors which were identified using multivariate stepwise logistic regression. Next, an online dynamic nomogram was created for the clinical usage. The receiver operating characteristic (ROC) curve, net reclassification improvement and integrated discrimination improvement were used to assess the discrimination of the IT-3 model. Calibration curves were used to evaluate the models' calibration. The clinical practicability of the model was evaluated using decision curve analysis and clinical impact curves. Results8.8% (39 of 445) patients presented with significant liver fibrosis in this study. Aspartate aminotransferase (AST), hepatitis B e-antigen (HBeAg), and platelet (PLT) were included in the prediction model (IT-3). The IT-3 model showed good calibration and discrimination both in the training and validation cohorts (AUC = 0.888 and 0.833, respectively). The continuous NRI and IDI showed that the IT-3 model had better predictive accuracy than GPR, APRI, and FIB-4 (p < 0.001). Decision curve analysis and clinical impact curves were used to demonstrate the clinical usefulness. At a cut-off value of 106 points, the sensitivity and specificity were 91.7 and 70.2%, respectively. ConclusionThe IT-3 model proved an accurate non-invasive method in identifying pseudo-IT of CHB, which can help to formulate more appropriate treatment strategies.