FibroBox: a novel noninvasive tool for predicting significant liver fibrosis and cirrhosis in HBV infected patients

Abstract

Abstract Background China is a highly endemic area of chronic hepatitis B (CHB). The accuracy of existed noninvasive biomarkers including TE, APRI and FIB-4 for staging fibrosis is not high enough in Chinese cohort. Methods Using liver biopsy as a gold standard, a novel noninvasive indicator was developed using laboratory tests, ultrasound measurements and liver stiffness measurements with machine learning techniques to predict significant fibrosis and cirrhosis in CHB patients in north and east part of China. We retrospectively evaluated the diagnostic performance of the novel indicator named FibroBox, Fibroscan, aspartate transaminase-to-platelet ratio index (APRI), and fibrosis-4 index (FIB-4) in CHB patients from Jilin and Huai’an (training sets) and also in Anhui and Beijing cohorts (validation sets). Results Of 1289 eligible HBV patients who had liver histological data, 63.2% had significant fibrosis and 22.5% had cirrhosis. In LASSO logistic regression and filter methods, fibroscan results, platelet count, alanine transaminase (ALT), prothrombin time (PT), type III procollagen aminoterminal peptide (PIIINP), type IV collagen, laminin, hyaluronic acid (HA) and diameter of spleen vein were finally selected as input variables in FibroBox. Consequently, FibroBox was developed of which the area under the receiver operating characteristic curve (AUROC) was significantly higher than that of TE, APRI and FIB-4 to predicting significant fibrosis and cirrhosis. In the Anhui and Beijing cohort, the AUROC of FibroBox was 0.88 (95% CI, 0.72–0.82) and 0.87 (95% CI, 0.83–0.91) for significant fibrosis and 0.87 (95% CI, 0.82–0.92) and 0.90 (95% CI, 0.85–0.94) for cirrhosis. In the validation cohorts, FibroBox accurately diagnosed 81% of significant fibrosis and 84% of cirrhosis. Conclusions FibroBox has a better performance in predicting liver fibrosis in Chinese cohorts with CHB, which may serve as a feasible alternative to liver biopsy.

Keywords

• Liver fibrosis
• HBV
• Noninvasive diagnosis
• Machine learning