A body composition-based clustering study and its association with metabolic phenotypes among the general population...

Metabolic phenotypes are linked to various metabolic diseases, but current classification methods have limitations. This study aims to directly cluster obese populations based on body composition and machine learning, enhancing understanding of lipid metabolism and disease associations.

A retrospective analysis included participants who underwent InBody examinations at Taikang Xianlin Drum Tower Hospital in 2023. Subjects were categorized into four phenotypes: MHNW, MHO, MUNW, and MUO, based on BMI and metabolic syndrome criteria. Correlations between InBody indexes and clinical data were analyzed. Machine learning cluster analysis identified subgroupings, and associations with metabolic diseases were examined.

InBody indexes correlated strongly with medical history and lab results. Clustering classified males into two groups and females into three, with significant differences in age, weight, height, BMI, and InBody scores (all p < 0.001). The prevalence of hypertension and hyperlipidemia varied notably among male subgroups, while hypertension and diabetes showed significant differences among female subgroups.

The InBody-based clustering analysis showed males could be categorized into 2 subgroups while females could be classified into 3 subgroups, indicating that the population with a specific InBody clustering profile could be at higher risk of metabolic diseases.