Association between visceral fat accumulation and sarcopenia: A cross-sectional study

Abstract

Background: The incidence of sarcopenia is increasing annually, and tools for assessing its risk remain limited. Visceral fat accumulation is closely associated with sarcopenia.

Methods: Data from 5200 participants in NHANES 2011-2018 were analyzed. Six visceral fat accumulation indicators, namely relative fat mass (RFM), lipid accumulation product (LAP), weight-adjusted waist index (WWI), triglyceride glucose-waist-to-height ratio (TyG-WHtR), metabolic score for insulin resistance (METS-IR), and metabolic score for visceral fat (METS-VF), were evaluated and compared for their associations with sarcopenia using multivariable logistic regression, smoothed curve fitting and threshold effect analysis. This study aimed to develop nine machine learning (ML) models incorporating visceral fat indicators to predict the risk of sarcopenia, with Shapley Additive Explanations (SHAP) applied to enhance model interpretability.

Results: Visceral fat accumulation indicators were substantially associated with the risk of sarcopenia. Threshold effect analysis revealed that the saturation points for RFM, LAP, WWI, TyG-WHtR, METS-IR, and METS-VF in sarcopenia were 41.844, 76.747, 11.352, 4.777, 50.525, and 6.806, respectively. The logistic regression model exhibited the highest predictive performance with an area under the receiver operating characteristic curve (AUC-ROC) of 0.878. WWI was identified as the strongest predictor of sarcopenia risk in the SHAP analysis.

Conclusion: All visceral fat accumulation indicators were positively associated with sarcopenia risk and WWI identified as the most important predictor. The ML model achieved high predictive accuracy, highlighting the role of visceral fat accumulation in sarcopenia risk and healthy aging promotion.

Keywords: Machine learning; NHANES; Sarcopenia; Visceral fat accumulation.