Abstract
Sarcopenia, characterized by progressive loss of muscle mass and function, is a growing public health concern. The ZJU index, a novel metabolic marker, integrates lipid metabolism and glucose regulation parameters. While its association with metabolic disorders has been established, its relationship with sarcopenia remains underexplored, especially in middle-aged adults. This cross-sectional study analyzed data from 4,012 U.S. adults aged 20-59 years in the 2011-2018 NHANES dataset. The association between ZJU and sarcopenia was assessed using multivariable logistic regression, restricted cubic splines (RCS) for smooth curve fitting, and subgroup analyses. To improve risk stratification and identify key predictors, machine learning techniques-including Random Forest, SHAP, and the Boruta algorithm-were applied. Each standard deviation increase in ZJU was associated with an 13% higher likelihood of sarcopenia [OR = 1.13, 95% CI 1.08-1.17]. Individuals in the highest ZJU quartile faced a 12.6-fold greater likelihood than those in the lowest quartile [OR = 13.6, 95% CI 3.08-60.2]. Subgroup analysis showed notable interactions with gender and diabetes (p < 0.05). Machine learning models consistently ranked ZJU, education level, and race as the most influential predictors of sarcopenia, emphasizing the interplay between metabolic health and socioeconomic factors. Higher ZJU scores are linked to increased sarcopenia risk in adults aged 20-59 years, supporting its role as an early metabolic biomarker. Machine learning identified ZJU, education, and race as key predictors, underscoring the impact of socioeconomic factors.