Multiparametric US for Identifying Patients with High-Risk NASH: A Derivation and Validation Study

Background Nonalcoholic fatty liver disease (NAFLD) is common in the general population but identifying patients with high-risk nonalcoholic steatohepatitis (NASH) who are candidates for pharmacologic therapy remains a challenge. Purpose To develop a score to identify patients with high-risk NASH, defined as NASH with an NAFLD activity score (NAS) of 4 or greater and clinically significant fibrosis (stage 2 [F2] or higher). Materials and Methods This was a cross-sectional secondary analysis of data prospectively collected between April 2017 and March 2019 for a group of patients with NAFLD in Japan (Japan NAFLD, the derivation data set) with contemporaneous two-dimensional shear-wave elastography and biopsy-proven NAFLD (age range, 20-89 years). Three US markers (liver stiffness [LS, measured in kilopascals], attenuation coefficient [AC, measured in decibels per centimeter per megahertz], and dispersion slope [DS, measured in meters per second per kilohertz]) were determined, together with aspartate aminotransferase (AST) and alanine aminotransferase (ALT) levels and the AST-to-ALT ratio. The best-fit multivariate logistic regression model for identifying patients with high-risk NASH was determined. Diagnostic performance was assessed by using the area under the receiver operating characteristic curve (AUC). The findings were validated in an independent data set (Korea NAFLD; age range, 20-78 years). Results The Japan NAFLD data set included 111 patients (mean age, 53 years ± 18 [standard deviation]; 57 men), 84 (76%) with NASH. The Korea NAFLD data set included 102 patients (mean age, 48 years ± 18; 43 men), 55 (36%) with NASH. The most predictive model (LAD NASH score) combined LS, AC, and DS. Performance was satisfactory in both the derivation sample (AUC, 0.86; 95% CI: 0.79, 0.93) and the validation sample (AUC, 0.88; 95% CI: 0.80, 0.95). The LAD NASH score showed a positive predictive value of 86.5% and a negative predictive value of 87.5% for high-risk NASH in the derivation sample. Conclusion A score combining three US markers may be useful for noninvasive identification of patients with high-risk nonalcoholic steatohepatitis for inclusion in clinical trials and pharmacologic therapy. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Lockhart in this issue.