A machine learning algorithm for stratification of risk of cardiovascular disease in metabolic dysfunction-associated steatotic liver disease

Abstract

Background

Steatotic liver disease (SLD) is associated with adverse cardiac events. Metabolic dysfunction-associated steatotic liver disease (MASLD) has emerged as a condition characterized by the abnormal accumulation of hepatic lipids that is closely linked to five metabolic disorders: overweight or obesity, impaired glucose regulation, hypertension, hypertriglyceridemia, and low high-density lipoprotein-cholesterol. This retrospective study aimed to stratify the risk of cardiac events in patients with MASLD.

Methods

Patients diagnosed with MASLD through ultrasonography were evaluated. We implemented a machine learning-based approach using a survival classification and regression tree (CART) model to stratify patients based on age, and the number of risk scores was investigated as a predictor of adverse outcomes in the derivation cohort. The primary outcomes were major adverse cardiac events (MACE) including cardiac death, nonfatal myocardial infarction, and revascularization due to coronary artery disease.

Results

Among 2,962 patients (median age, 62 years; men, 53.5 %), the distribution of risk factors was as follows: one (10.8 %), two (28.5 %), three (33.0 %), four (19.9 %), and five (7.8 %). Over a median follow-up period of 6.8 years, 170 (5.7 %) patients experienced MACE. In the derivation cohort of 2,073 patients, the CART model identified age ≥60 years old and risk factors ≥4 as significant predictors of MACE. These findings were corroborated in a validation cohort of 889 patients. Patients meeting both criteria exhibited the highest risk of MACE (log-rank test, p < 0.001).

Conclusions

Patients aged ≥60 years old with risk factors ≥4 indicates at high risk of MACE in patients with MASLD. This risk stratification system provides a practical tool for identifying high-risk individuals in the MASLD population.