BACKGROUND AND AIM: Liver biopsy remains the gold standard to evaluate liver histology. However, it has several limitations. This study aims to construct a noninvasive model to predict liver histology for commencing antiviral therapy in HBeAg-positive chronic hepatitis B (CHB) with aminotransferase (ALT) ≤ 2 upper limit of normal (ULN). METHODS:Two hundred and ninety-eight patients with HBeAg-positive CHB, ALT ≤2ULN and HBV-DNA≥20 000 IU/ml were enrolled and randomly divided into a training group and a validation group. A noninvasive model was constructed in the training group to predict significant liver histological change [necroinflammatory activity grade (G) ≥ 2 or fibrosis stage (S) ≥ 2] and then validated in the validation group. RESULTS:Aspartate aminotransferase, HBsAg, platelet, and albumin were identified as independent predictors. A model was constructed by them. It had an area under the receiver operating characteristic curve of 0.875 in the training group, 0.858 in the validation group and 0.868 in the entire cohort. Using a cut-off point of -0.96, it showed 93% sensitivity, 90% negative predictive value (NPV) in the training group and 95% sensitivity, 94% NPV in the validation group. Using a cut-off point of 0.96, it showed 95% specificity, 91% positive predictive value (PPV) in the training group and 89% specificity, 80% PPV in the validation group. CONCLUSIONS: This study constructed a noninvasive model to predict liver histology in HBeAg-positive CHB with ALT ≤ 2ULN, which might reduce the clinical need for liver biopsy.
RCT Entities:
BACKGROUND AND AIM: Liver biopsy remains the gold standard to evaluate liver histology. However, it has several limitations. This study aims to construct a noninvasive model to predict liver histology for commencing antiviral therapy in HBeAg-positive chronic hepatitis B (CHB) with aminotransferase (ALT) ≤ 2 upper limit of normal (ULN). METHODS: Two hundred and ninety-eight patients with HBeAg-positive CHB, ALT ≤ 2ULN and HBV-DNA ≥20 000 IU/ml were enrolled and randomly divided into a training group and a validation group. A noninvasive model was constructed in the training group to predict significant liver histological change [necroinflammatory activity grade (G) ≥ 2 or fibrosis stage (S) ≥ 2] and then validated in the validation group. RESULTS: Aspartate aminotransferase, HBsAg, platelet, and albumin were identified as independent predictors. A model was constructed by them. It had an area under the receiver operating characteristic curve of 0.875 in the training group, 0.858 in the validation group and 0.868 in the entire cohort. Using a cut-off point of -0.96, it showed 93% sensitivity, 90% negative predictive value (NPV) in the training group and 95% sensitivity, 94% NPV in the validation group. Using a cut-off point of 0.96, it showed 95% specificity, 91% positive predictive value (PPV) in the training group and 89% specificity, 80% PPV in the validation group. CONCLUSIONS: This study constructed a noninvasive model to predict liver histology in HBeAg-positive CHB with ALT ≤ 2ULN, which might reduce the clinical need for liver biopsy.