Step layered combination of noninvasive fibrosis models improves diagnostic accuracy of advanced fibrosis in nonalcoholic fatty liver disease

Abstract

Background and Aims: Liver fibrosis is stage-dependently associated with non-alcoholic fatty liver disease (NAFLD) progression. The increased awareness of non-invasive diagnosis has led to the establishment of many fibrosis diagnosis models with various accuracies. We aimed to evaluate the diagnostic performance of nine clinical non-invasive fibrosis models in NAFLD and provide an optimal diagnostic method for advanced fibrosis by step layered combination of non-invasive models. Methods: 453 consecutive patients with biopsy-proven NAFLD were enrolled from three centers and were divided into study cohort and validation cohort randomly. Aspartate aminotransferase-to-platelet ratio index (APRI), BARD, FiB-4, FibroMeter NAFLD, Forns’ Index, Hui model, non-invasive Koeln-Essen- index (NIKEI), S Index and NAFLD fibrosis score (NFS) were calculated. The high area under the receiver operating characteristic curve (AUROC) models were stepwise combined for further diagnosing NAFLD advanced fibrosis. Results: All models had good performance with high negative predictive value (NPV) and specificity for diagnosing fibrosis, while positive predictive value (PPV) and sensitivity were low. APRI, BARD, FibroMeter NAFLD and NIKEI had higher AUROCs and their step layered combination for diagnosing advanced fibrosis showed high specificity, sensitivity, NPV and PPV up to 89.13%, 72.50%, 74.36%, and 88.17%, which also performed well in the validation cohort. Conclusions: APRI, BARD, FibroMeter NAFLD and NIKEI had better diagnostic accuracy, and could be preferred for diagnosing NAFLD fibrosis. The step layered combination of these models performed much better than each single scoring system for diagnosing advanced fibrosis, provides valuable reference for clinical practice and might be a potential substitution of liver biopsy.