Effects of environmental phenols on eGFR: machine learning modeling methods applied to cross-sectional studies

Area of Science:

• Environmental Health
• Toxicology
• Nephrology

Background:

• Limited research exists on the relationship between environmental phenol exposure and estimated glomerular filtration rate (eGFR).
• Environmental phenols are widespread contaminants with potential health implications.
• Understanding these associations is crucial for public health assessments.

Purpose of the Study:

• To develop a robust and explainable machine learning (ML) model associating environmental phenol exposure with eGFR.
• To identify specific environmental phenols that impact kidney function.

Main Methods:

• Utilized data from the National Health and Nutrition Examination Survey (NHANES, 2013-2016).
• Developed and fine-tuned five ML models for eGFR regression based on phenol exposure.
• Employed Shapley Additive Explanations (SHAP) for model interpretation.

Main Results:

• A Random Forest (RF) regressor demonstrated high performance (R-squared: 0.998).
• Urinary triclosan (TCS) and bisphenol S (BPS) concentrations were positively correlated with eGFR.
• SHAP analysis identified BPS, bisphenol F (BPF), and bisphenol A (BPA) as significant contributors to the model.

Conclusions:

• The RF model effectively identified correlations between phenol exposure and eGFR in the NHANES cohort.
• Findings suggest potential inverse associations between BPA, BPF, and BPS exposure and eGFR.