Background & aimsIdentification of disease severity remains a challenge in the management of non-alcoholic steatohepatitis (NASH). Cytokeratin-18 (CK18), is a recently developed non-invasive biomarker for NASH. We aimed to assess the performance of CK18 in disease severity prediction among Taiwanese NASH patients.MethodsA total of 76 biopsy-proven NASH patients (54 males, age = 41.0 ± 13.5 years) were consecutively recruited. The optimal cutoff values of CK18 for each stage of fibrosis were correlated with their histopathological manifestations.ResultsThere were 23 (30.3%) patients of Metavir fibrosis stage 0 (F0), 32 (42.1%) patients of F1, 14 (18.4%) patients of F2, and 7 (9.2%) patients of F3-4, respectively. The CK18 levels among those patients of F0, F1, F2, F3-4 were 86.7 ± 75.6 U/L, 122.4 ± 123.8 U/L, 160.7 ± 120.4 U/L, and 507.3 ± 343 U/L, respectively (trend for P<0.001). The adjusted optimal cutoff value for F2 prediction was 312.5 U/L, yielding the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the accuracy of 96.4%, 28.6%, 77.9%, 75%, and 77.6%, respectively (P = 0.009). For the prediction of advanced fibrosis (F3-4), the adjusted optimal cutoff value was 374.5 U/L, yielding the sensitivity, specificity, PPV, NPV, and the accuracy of 97.1%, 54.1%, 95.7%, 66.7%, and 77.6%, respectively (P = 0.003). Among those patients without hyperuricemia, the PPV, NPV, and accuracy of CK18 reached 100%, 95.8%, and 96%, respectively (P<0.001).ConclusionsCK18 combined with uric acid measurement is a promising non-invasive biomarker for prediction of disease severity in NASH patients.Trial registrationClinicalTrials.gov NCT01068444.