Predicting hepatic decompensation in patients with metabolic dysfunction associated steatotic liver disease-related cirrhosis: the ABID-LSM Model

Abstract

Background & aims: Predicting the risk of hepatic decompensation guides prognostication and therapy; however, it is challenging in patients with cirrhosis due to metabolic dysfunction-associated steatotic liver disease (MASLD). We aimed to improve a previously developed predictive tool of hepatic decompensation in MASLD cirrhosis (ABIDE) by incorporating liver stiffness measurement (LSM). Methods: A multi-centre retrospective cohort of patients with compensated cirrhosis due to MASLD was identified, with decompensation incidence assessed using competing risk regression. The prognostic accuracy of a modified ABIDE model incorporating LSM (ABID-LSM) was assessed using time-dependent AUC (tAUC) and compared with other predictive models. Results: Out of 388 patients, 273 (70.4%) had available LSM. Hepatic decompensation occurred in 54 (20%) patients during follow-up (median 31 months, range: 20-60). The predictive accuracy at 5 years of ABID-LSM (tAUC 0.80) was better than ABIDE (tAUC 0.75, p = 0.03) and LSM (tAUC 0.63, p < 0.001). The ABID-LSM model calibrated well (slope 0.99) with excellent overall performance (Integrated Brier Score 0.15). A cut-off of 8.1 separated those at high and low risk of hepatic decompensation at 5 years (24% vs. 5%, respectively, sHR = 4.8, p < 0.001). The ABID-LSM model had better predictive ability at 5 years than ALBI, FIB-4, NAFLD Decompensation Risk Score and ANTICIPATE models (all p < 0.001) as well as hepatic vein pressure gradient measurement (tAUC 0.78 vs. 0.71, p < 0.001, n = 60). Conclusions: The ABID-LSM model has greater accuracy in predicting hepatic decompensation in patients with cirrhosis due to MASLD than existing predictive models. If externally validated, ABID-LSM may identify those who benefit from pharmacotherapy and close monitoring.