Predictive model for fibrosis and cirrhosis in chronic hepatitis B using liver stiffness measurement

Abstract

AIM: To correlate liver stiffness measurements with demographic and laboratory parameters, and assess the predictive value of these parameters as non-invasive markers of fibrosis and cirrhosis in chronic hepatitis B (CHB). METHODS: Transient elastography was performed in 265 CHB patients using Fibroscan (Echosens, France). Liver stiffness scores of >8.1 and >10.3 kPa were used as cut-off values for significant fibrosis and cirrhosis respectively. To derive a new index using commonly measured laboratory markers, the study sample was randomly split into a training set and a validation set. RESULTS: Of the 265 patients, 170(64%) were male. Ninety-six patients (36%) were HBeAg-positive, and 90 patients (34%) had severe fibrosis or cirrhosis. The median liver stiffness score was 6.3 kPa (range, 3.0 to 56.3). Liver stiffness correlated positively with ALT, bilirubin and AFP, and negatively with albumin levels. Using 13 parameters (age, sex, platelet, AST, ALT, GGT, AFP, albumin, globulin, bilirubin, ALP, HBV DNA, and HBeAg), the sequence of variables in order of their associations with liver stiffness (co-effi cient path) was determined using L1 regularized regression. The AUROC were calculated for each number of variables used for the prediction of fibrosis and cirrhosis, and the number of variables used was determined as the one that additional variables would not give a relatively higher accuracy. Four best predictive parameters (AST, Platelet, GGT and AFP) were used to derive a model for fibrosis (the APGA index). Additional variables did not improve accuracy significantly. The model using log(index)=1.44 + 0.1490(GGT) + 0.3308 log(AST) – 0.5846 log(platelets) + 0.1148 log (AFP+1) to predict significant fibrosis had an AUROC of 0.85 in both training and validation groups. An optimal cutoff value of 6.9 had sensitivity of 82%, specificity of 69%, with negative predictive value of 91%. The AUROC for predicting cirrhosis were 0.89 and 0.85 in training and validation groups respectively. Using an optimal cutoff of 8.9 was associated with sensitivity of 64% and specificity of 89% with a negative predictive value of 92%. The AUROC for the APGA index was higher compared to that of the AST-toplatelet ratio index, AST/ALT ratio, and age-platelet index in the training group (0.85, 0.81, 0.50, and 0.77 respectively) and the validation group (0.85, 0.80. 0.38 and 0.68 respectively) CONCLUSION: Measurement of liver stiffness with transient elastography correlates well with known factors associated with the severity of disease in Asian CHB patients. The APGA index is a reliable non-invasive method for predicting significant fibrosis and cirrhosis in CHB.