Background and aims

Magnetic resonance elastography (MRE) is an accurate, continuous biomarker of liver fibrosis; however, the optimal combination with clinical factors to predict the risk of incident hepatic decompensation is unknown. Therefore, we aimed to develop and validate an MRE-based prediction model for hepatic decompensation for patients with NAFLD.

Approach and results

This international multicenter cohort study included participants with NAFLD undergoing MRE from 6 hospitals. A total of 1254 participants were randomly assigned as training (n = 627) and validation (n = 627) cohorts. The primary end point was hepatic decompensation, defined as the first occurrence of variceal hemorrhage, ascites, or HE. Covariates associated with hepatic decompensation on Cox-regression were combined with MRE to construct a risk prediction model in the training cohort and then tested in the validation cohort. The median (IQR) age and MRE values were 61 (18) years and 3.5 (2.5) kPa in the training cohort and 60 (20) years and 3.4 (2.5) kPa in the validation cohort, respectively. The MRE-based multivariable model that included age, MRE, albumin, aspartate aminotransferase, and platelets had excellent discrimination for the 3- and 5-year risk of hepatic decompensation (c-statistic 0.912 and 0.891, respectively) in the training cohort. The diagnostic accuracy remained consistent in the validation cohort with a c-statistic of 0.871 and 0.876 for hepatic decompensation at 3 and 5 years, respectively, and was superior to Fibrosis-4 in both cohorts ( p < 0.05).

Conclusions

An MRE-based prediction model allows for accurate prediction of hepatic decompensation and assists in the risk stratification of patients with NAFLD.