Validation of a data-driven clustering model for MASLD: Evidence from three large-scale Asian cohorts

Background & Aims: Metabolic dysfunction-associated steatotic liver disease (MASLD) is a heterogeneous condition that presents varying risks for liver-related and cardiovascular complications. Clustering methods have identified distinct MASLD subtypes, yet their applicability to Asian populations remains unclear. This study aims to validate a MASLD clustering model using clinical variables from three Asian cohorts: Wenzhou Real-World (WRW), Hong Kong Clinical Data Analysis and Reporting System (CDARS), and SingHealth Diabetes Registry. Methods: Clustering analysis was conducted based on age, BMI, hemoglobin A(1c), alanine aminotransferase, LDL-cholesterol, and triglycerides. Outcomes included major adverse cardiovascular events (MACE), liver-related events (LRE), and new-onset type 2 diabetes (T2DM). They were analyzed using Cox regression risk models and Kaplan-Meier analyses to assess risk and incident events across MASLD clusters. Results: Across the three cohorts, distinct risk patterns emerged for MACE, LRE, and T2DM among various MASLD clusters. For MACE, the cardiometabolic cluster exhibited the highest risk in all cohorts: WRW (hazard ratio [HR] 1.315, p <0.001), Hong Kong CDARS (HR 1.559, p <0.001), and SingHealth Diabetes Registry (HR 1.262, p <0.001). For LRE, the liver-specific cluster showed the highest risk in the WRW (HR 1.578, p = 0.002) and SingHealth Diabetes Registry cohorts (HR 2.403, p <0.001). In contrast, in the Hong Kong CDARS cohort, both the cardiometabolic (HR 1.818, p <0.001) and liver-specific clusters (HR 1.557, p <0.001) exhibited similarly increased risks. For T2DM, the cardiometabolic cluster showed the highest risk in the WRW (HR 3.418, p <0.001) and Hong Kong CDARS cohorts (HR 2.761, p <0.001). Conclusions: The proposed MASLD clustering model is applicable to Asian populations, facilitating personalized treatment and optimizing outcomes.