Associations between urine glyphosate levels and metabolic health risks: insights from a large cross-sectional population-based study.

Abstract

Background: The prevalence of metabolic syndrome (MetS) in American adults increased from 37.6% in the 2011-12 period to 41.8% in 2017-2018. Environmental exposure, particularly to common compounds such as glyphosate, has drawn increasing attention as a potential risk factor.

Methods: We employed three cycles of data (2013-2018) from the National Health and Nutrition Examination Survey (NHANES) in a cross-sectional study to examine potential associations between urine glyphosate measurements and MetS incidence. We first created a MetS score using exploratory factor analysis (EFA) of the International Diabetes Federation (IDF) criteria for MetS, with data drawn from the 2013-2018 NHANES cycles, and validated this score independently on an additional associated metric, the albumin-to-creatinine (ACR) ratio. The score was validated via a machine learning approach in predicting the ACR score via binary classification and then used in multivariable regression to test the association between quartile-categorized glyphosate exposure and the MetS score.

Results: In adjusted multivariable regressions, regressions between quartile-categorized glyphosate exposure and MetS score showed a significant inverted U-shaped or saturating dose‒response profile, often with the largest effect for exposures in quartile 3. Exploration of potential effect modification by sex, race, and age category revealed significant differences by race and age, with older people (aged > 65 years) and non-Hispanic African American participants showing larger effect sizes for all exposure quartiles.

Conclusions: We found that urinary glyphosate concentration is significantly associated with a statistical score designed to predict MetS status and that dose-response coefficient is nonlinear, with advanced age and non-Hispanic African American, Mexican American and other Hispanic participants exhibiting greater effect sizes.