Generalized Functional Linear Models: Efficient Modeling for High-dimensional Correlated Mixture Exposures.

Area of Science:

• Environmental epidemiology
• Toxicology
• Biostatistics

Background:

• Human health is impacted by complex environmental chemical mixtures.
• Analyzing these mixtures poses challenges like high dimensionality and correlated exposures.

Purpose of the Study:

• To introduce and evaluate a novel statistical approach, the generalized functional linear model (GFLM), for analyzing health effects of exposure mixtures.
• To demonstrate the GFLM's ability to handle correlated exposures and provide interpretable results.

Main Methods:

• The generalized functional linear model (GFLM) was developed to treat mixture effects as smooth functions.
• GFLM reorders exposures based on mechanisms and captures internal correlations for estimation.
• The model's robustness and efficiency were assessed through extensive simulations.

Main Results:

• Applied to NHANES data, GFLM identified significant nutrient mixture effects on BMI, with fiber and fat showing the strongest negative and positive impacts.
• In analyzing per- and polyfluoroalkyl substances (PFAS) and gout risk, GFLM revealed no significant association, highlighting its robustness to multicollinearity.

Conclusions:

• The GFLM framework is a powerful tool for mixture exposure analysis in environmental epidemiology.
• It offers improved handling of correlated exposures and interpretable results, advancing understanding of complex environmental health impacts.