Performance of non-invasive models of fibrosis in predicting mild to moderate fibrosis in patients with non-alcoholic fatty liver disease

• Published in: Liver international 2016-04, Vol.36 (4), p.572-579
• Main Authors: Siddiqui, Mohammad S., Patidar, Kavish R., Boyett, Sherry, Luketic, Velimir A., Puri, Puneet, Sanyal, Arun J.
• Format: Article
• Language: English
• Subjects: Adult; Alanine Transaminase - blood; Area Under Curve; Aspartate Aminotransferases - blood; Biomarkers - blood; Biopsy; Blood Platelets; Body Mass Index; Clinical Enzyme Tests; Decision Support Techniques; Female; FibroMeter non-alcoholic fatty liver disease; fibrosis; fibrosis 4; Humans; Liver Cirrhosis - blood; Liver Cirrhosis - diagnosis; Liver Cirrhosis - etiology; Male; Middle Aged; non-alcoholic fatty liver disease; Non-alcoholic Fatty Liver Disease - blood; Non-alcoholic Fatty Liver Disease - complications; Non-alcoholic Fatty Liver Disease - diagnosis; Platelet Count; Predictive Value of Tests; Prognosis; Reproducibility of Results; Retrospective Studies; Risk Factors; ROC Curve; Severity of Illness Index; Time Factors
• ISSN: 1478-3223; 1478-3231
• DOI: 10.1111/liv.13054

Background & Aims In non‐alcoholic fatty liver disease, presence of fibrosis is predictive of long‐term liver–related complications. Currently, there are no reliable and non‐invasive means of quantifying fibrosis in those with non‐alcoholic fatty liver disease. Therefore, we aimed to evaluate the performance of a panel of non‐invasive models in predicting fibrosis in non‐alcoholic fatty liver disease. Methods The accuracy of FibroMeter non‐alcoholic fatty liver disease, fibrosis 4 and four other non‐invasive models in predicting fibrosis in those with biopsy proven non‐alcoholic fatty liver disease was compared. These models were constructed post hoc in patients who had necessary clinical information collected within 2 months of a liver biopsy. The areas under receiver operating characteristics curves were compared for each model using Delong analysis. Optimum cut‐off for each model and fibrosis stage were calculated using the Youden index. Results The area under receiver operating characteristics curves for F ≥ 1 fibrosis for fibrosis 4 and FibroMeter non‐alcoholic fatty liver disease was 0.821 and 0.801 respectively. For F ≥ 3, the area under receiver operating characteristics curves was 0.866 for fibrosis 4 and 0.862 for FibroMeter non‐alcoholic fatty liver disease. Delong analysis showed the area under receiver operating characteristics curves was statistically different for fibrosis 4 and FibroMeter non‐alcoholic fatty liver disease compared with BARD, BAAT and aspartate aminotransferase:alanine aminotransferase ratio for F ≥ 1 and F ≥ 3. Area under receiver operating characteristics curves were significantly different for fibrosis 4 and FibroMeter non‐alcoholic fatty liver disease for F ≥ 3 compared with non‐alcoholic fatty liver disease fibrosis score. At a fixed sensitivity of 90%, FibroMeter non‐alcoholic fatty liver disease had the highest specificity for F ≥ 1 (52.4%) and F ≥ 3 (63.8%). In contrast, at a fixed specificity of 90%, fibrosis 4 outperformed other models with a sensitivity of 60.2% for F ≥ 1 and 70.6% for F ≥ 3 fibrosis. Conclusion These non‐invasive models of fibrosis can predict varying degrees of fibrosis from routinely collected clinical information in non‐alcoholic fatty liver disease.