Heavy metal mixtures and kidney function: multi-scale evidence from a nationally representative cohort

Abstract

Background and objective

Chronic kidney disease (CKD) is a major global health burden, and exposure to heavy metals is increasingly recognized as a potentially modifiable risk factor.

Methods

Using data from a nationally representative U.S. survey (NHANES 2011–2016), we evaluated associations between blood concentrations of seven metals (Pb, Cd, Cu, Hg, Zn, Se, Mn) and kidney outcomes—estimated glomerular filtration rate (eGFR), urine albumin-to-creatinine ratio (UACR), and CKD status. We fitted single-metal regression models and mixed-exposure models, including directional weighted quantile sum (WQS) regression and Bayesian kernel machine regression (BKMR). To generate mechanistic hypotheses, we performed functional enrichment and molecular docking analyses.

Results

In single-metal models, Pb was associated with lower eGFR, and Cu was associated with higher UACR. In mixture analyses, WQS showed a positive association with UACR and (in the positive-direction model) higher eGFR, whereas the WQS association with CKD was suggestive but not statistically significant. BKMR indicated non-linear and context-dependent (co-exposure-dependent) patterns for selected metals across outcomes, and bivariate surfaces suggested potential interaction structure. Exploratory bioinformatic analyses highlighted extracellular matrix-related processes and cell–matrix signaling pathways, and docking suggested plausible metal–protein contacts for selected targets.

Conclusion

In this cross-sectional analysis, associations between metal exposures and kidney outcomes varied by outcome metric and statistical approach, with more consistent signals for Pb–eGFR and Cu–UACR. Mixture models suggested joint-exposure patterns and potential non-linearity/interaction structure, but causal inference is limited by the cross-sectional design and single-time blood measurements; reverse causation and residual confounding remain possible. The downstream bioinformatic and docking results are hypothesis-generating and intended to prioritize pathways for future mechanistic research.

Graphical Abstract