Exploring a Subpopulation of MASLD Associated With New Onset of CKD Using Supervised Clustering Techniques

Area of Science:

• Hepatology and Nephrology
• Metabolic Syndrome Research
• Machine Learning in Medicine

Background:

• Metabolic dysfunction-associated steatotic liver disease (MASLD) is linked to cardiovascular-kidney-metabolic syndrome, including chronic kidney disease (CKD).
• The heterogeneity of MASLD and its impact on CKD development are not fully understood.
• Accurate risk stratification for CKD in MASLD patients is crucial.

Purpose of the Study:

• To investigate the relationship between MASLD subgroups and new-onset CKD.
• To utilize machine learning for identifying distinct MASLD phenotypes.
• To improve risk stratification for CKD development in MASLD.

Main Methods:

• Employed supervised clustering using SHapley Additive exPlanations (SHAP) and uniform manifold approximation and projection (UMAP) on steatotic liver disease data.
• Analyzed health examination data from 12,168 Japanese subjects over a 10-year follow-up.
• Utilized Gaussian mixture models to detect MASLD subclusters and Kaplan-Meier analysis for CKD incidence.

Main Results:

• Identified 10 subclusters and 3 distinct subgroups of MASLD using machine learning.
• Demonstrated significant differences in cumulative CKD incidence among the MASLD subgroups.
• An obese MASLD subgroup characterized by atherogenic lipid profiles, high fatty liver index, and elevated uric acid exhibited the highest risk for CKD development.

Conclusions:

• Supervised clustering of MASLD reveals phenotypically distinct subpopulations, enhancing CKD risk stratification.
• An obese MASLD subgroup with atherogenic dyslipidemia and hyperuricemia is associated with significantly increased CKD risk.
• These findings aid in personalized risk assessment and management of MASLD patients to prevent CKD.