A weighted kernel machine regression approach to environmental pollutants and infertility

Area of Science:

• Environmental Epidemiology
• Reproductive Health
• Statistical Modeling

Background:

• Couple-based epidemiological studies collect extensive environmental pollutant data for both partners.
• Analyzing the combined effects of numerous chemical mixtures on infertility presents statistical challenges due to complex exposure relationships.

Purpose of the Study:

• To develop a novel statistical method, weighted kernel machine regression (wKRM), to effectively analyze partner-specific environmental exposures in human infertility studies.
• To account for the unique couple-based data structure and model the joint effects of chemical mixtures on infertility.

Main Methods:

• Developed weighted kernel machine regression (wKRM), a nonparametric approach combining female and male partner exposure data using linear weighting.
• Kernel machine regression was adapted to incorporate partner-specific exposure information, addressing limitations of standard methods.

Main Results:

• The wKRM method effectively reduces the number of exposures analyzed and provides an importance index for female versus male partner exposures.
• Simulation studies confirmed wKRM's efficacy in estimating joint exposure effects and modeling infertility outcomes.
• Application to a prospective study indicated that male partner exposure to polychlorinated biphenyls may be more strongly associated with infertility than female partner exposure.

Conclusions:

• Weighted kernel machine regression (wKRM) offers an improved approach for analyzing couple-based environmental exposure data in infertility research.
• The method highlights the differential impact of partner-specific exposures, suggesting male exposure to certain pollutants like PCBs may significantly influence infertility risk.