BackgroundMetabolic dysfunction-associated steatohepatitis (MASH) is associated with more than a 10-fold increase in liver-related mortality. However, biomarkers predicting both MASH and mortality are missing. We developed a metabolome-derived prediction score for MASH and examined whether it predicts mortality in Chinese and European cohorts. MethodsThe MASH prediction score was developed using a multi-step machine learning strategy, based on 44 clinical parameters and 250 plasma metabolites measured by proton nuclear magnetic resonance (1H-NMR) in 311 Chinese adults undergoing a liver biopsy. External validation was conducted in a Finnish liver biopsy cohort (n=305). We investigated association of the score with all-cause and cause-specific mortality in the population-based Shanghai Changfeng Study (n=5,893) and the UK Biobank (n=111,673). ResultsA total of 24 clinical parameters and 194 1H-NMR metabolites were significantly associated with MASH in the Chinese liver biopsy cohort. The final MASH score included body mass index, aspartate transaminase, tyrosine, and the phospholipids-to-total lipids ratio in very-low density lipoprotein. The score identified patients with MASH with AUROCs of 0.87 (95% CI, 0.83-0.91) and 0.81 (95% CI, 0.75-0.87) in the Chinese and Finnish cohorts, with high negative predictive values. Participants with a high or intermediate risk of MASH based on the score had a markedly higher risk of MASLD-related mortality than those with a low risk in Chinese (HR, 23.19; 95%CI, 4.80-111.97) and European individuals (HR, 27.80; 95%CI, 15.08-51.26) after 7.4 and 12.6 years of follow-up. The MASH prediction score was superior to the FIB-4 index and the NAFLD Fibrosis Score in predicting MASLD-related mortality. ConclusionThe metabolome-derived MASH prediction score accurately predicts risk of MASH and MASLD-related mortality in both Chinese and European individuals. Impact and implicationsMetabolic dysfunction-associated steatohepatitis (MASH) is associated with more than a 10-fold increase in liver-related death. However, biomarkers predicting not only MASH, but also death due to liver disease, are missing. We established a MASH prediction score based on 44 clinical parameters and 250 plasma metabolites using a machine learning strategy. This metabolome-derived MASH prediction score could accurately identify patients with MASH among both Chinese and Finnish individuals, and it was superior to the FIB-4 index and the NAFLD Fibrosis Score in predicting MASLD-related death in the general population. Thus, the new MASH prediction score is a useful tool for identifying individuals with a markedly increased risk of serious liver-related outcomes among at-risk and general populations.
Read full abstract