Machine Learning-Driven Metabolic Syndrome Prediction: An International Cohort Validation Study

Area of Science:

• Computational biology
• Epidemiology
• Biostatistics

Background:

• Metabolic syndrome (MetS) poses a significant global health challenge.
• Accurate prediction of MetS risk is crucial for timely intervention.
• Existing prediction models may require further refinement for diverse populations.

Purpose of the Study:

• To develop and validate a machine learning (ML)-based model for predicting metabolic syndrome (MetS) risk.
• To identify key predictors of MetS using LASSO regression.
• To assess the generalizability of the developed model across different ethnic and geographical cohorts.

Main Methods:

• Utilized data from the China Health and Retirement Longitudinal Study (CHARLS) for model development (n=6155).
• Employed LASSO regression for feature selection to identify significant MetS predictors.
• Trained and evaluated nine ML algorithms, including multilayer perceptron (MLP) and xgboost.
• Validated model performance on independent datasets from Korea (KNHANES), UK Biobank, and US (NHANES).

Main Results:

• The MLP-based model demonstrated strong predictive performance in the CHARLS (AUC=0.8908) and NHANES (AUC=0.9055) cohorts.
• Logistic regression showed high accuracy in the KNHANES cohort (AUC=0.9101).
• Xgboost achieved notable performance in the UK Biobank cohort (AUC=0.8556).
• The MLP model consistently showed robust performance across multiple validation datasets.

Conclusions:

• The developed MLP-based model shows significant potential for clinical application in MetS risk detection.
• The model's performance across diverse cohorts suggests its generalizability.
• This ML approach offers a promising tool for early identification and management of metabolic syndrome.