A novel noninvasive algorithm for the assessment of liver fibrosis in patients with chronic hepatitis B virus infection.
M-Y ZhuX ZouQ LiD-M YuZ-T YangD HuangJie ChenQ-M GongD-H ZhangY ZhangL ChenP-Z ChenX-X ZhangPublished in: Journal of viral hepatitis (2017)
Several noninvasive blood biomarkers have been established for the assessment of liver fibrosis in patients with chronic hepatitis B virus (HBV) infection, but their clinical performance remains inconclusive. Here, we compared the diagnostic performance of these biomarkers and developed a novel algorithm for assessing liver fibrosis. Six hundred and sixteen chronically HBV-infected and treatment-naïve patients who underwent liver biopsy were enrolled and randomly divided into training (N=410) and internal validation cohorts (N=206). One hundred and fifty-nine patients from another centre were recruited as an external validation cohort. Receiver operating characteristic (ROC) curves were used to analyse the performance of the gamma-glutamyltransferase-to-platelet ratio (GPR), red cell volume distribution width-to-platelet ratio (RPR), FIB-4 index, aspartate aminotransferase-to-platelet ratio index (APRI) and HBV DNA level against liver histology, and a novel algorithm was developed using the recursive partitioning and regression tree (RPART) method. In the training cohort, the area under the ROC curve of FIB-4 was significantly higher than that of APRI (P=.038) but was comparable to those of GPR, RPR and HBV DNA; however, the performance of the biomarkers was similar among the validation cohort. The established RPR-HBV DNA algorithm performed better in the training cohort than any individual blood biomarker, and the corresponding sensitivity, specificity, positive predictive value and negative predictive value were 63%, 90%, 72% and 80%, respectively. In the internal and external validation cohorts, the performance of the algorithm in assessing liver fibrosis was also superior to that of other biomarkers. These results suggest that the established RPR-HBV DNA algorithm might improve the diagnostic accuracy of liver fibrosis in treatment-naïve patients with chronic HBV infection, although additional studies are warranted to confirm these findings.
Keyphrases
- liver fibrosis
- hepatitis b virus
- liver failure
- machine learning
- deep learning
- end stage renal disease
- circulating tumor
- ejection fraction
- cell free
- chronic kidney disease
- single molecule
- neural network
- prognostic factors
- virtual reality
- nucleic acid
- single cell
- patient reported outcomes
- circulating tumor cells
- ultrasound guided
- patient reported