JMIR Publications

ABSTRACT

Background:

Metabolic syndrome comprises metabolic abnormalities that significantly increases the risk of cardiovascular disease (CVD) and other chronic conditions. Metabolic syndrome is affecting a growing proportion of the global population, particularly in aging and urban communities. Traditional methods for predicting metabolic syndrome often rely on invasive tests, which limits their application in large-scale health management. There is a growing need for more accessible and noninvasive prediction models.

Objective:

This study aimed to develop a predictive model for metabolic syndrome using noninvasive body composition data. Additionally, we assessed the utility of the model in predicting long-term CVD risk, enhancing its practical application in both clinical and nonclinical environments for early intervention and prevention strategies.

Methods:

We developed a predictive model for metabolic syndrome using non-invasive data from the Korea National Health and Nutrition Examination Survey (KNHANES) and Korean Genome and Epidemiology Study (KoGES), evaluating five algorithms. Dual-energy X-ray absorptiometry (DEXA) data from the 2008 to 2011 KNHANES cohort were used for training, whereas bioelectrical impedance analysis (BIA) data from the KNHANES 2022 (internal validation) and KoGES (external validation) were employed. Cox proportional hazards regression was used to assess the model’s ability to predict the long-term CVD risk.

Results:

The predictive model demonstrated strong performance across the internal and external validation cohorts. The performance of the model, based on the area under the receiver operating characteristic curve (AUROC) of the five best-performing models, ranged from 0.8338–0.8447 for metabolic syndrome prediction in the internal validation, 0.8066–0.8138 in external validation 1, and 0.8039–0.8123 in external validation 2. Additionally, the Cox proportional hazards regression analysis confirmed that patients with metabolic syndrome had a 1.51-fold higher risk of developing CVD, highlighting the potential for long-term cardiovascular risk prediction.

Conclusions:

This study successfully developed a predictive model for metabolic syndrome using noninvasive body composition data, demonstrating robust performance across internal and external validation cohorts. The use of both DEXA and BIA data enhanced the generalizability of the model, making it applicable to various clinical settings. Although this model is promising for widespread use, future research should focus on incorporating more diverse populations and additional clinical features to improve its predictive utility beyond CVD.