Metal mixtures and kidney function: An application of machine learning to NHANES data.

Background: Exposure to heavy metals may increase risk of kidney disease, but most studies have examined individual metals or two-way interactions. There is increasing recognition of the importance of studying exposure to metal mixtures and health outcomes.

Objectives: We used Bayesian kernel machine regression (BKMR) to examine associations between a mixture of four heavy metals and indicators of kidney function.

Methods: We used NHANES 2015-16 data on 1435 adults aged 40 and over to study cross-sectional associations between blood levels of four heavy metals (Co, Cr, Hg and Pb) and kidney function. Kidney function was assessed by estimated glomerular filtration rate (eGFR) and by albumin to creatinine ratio (ACR), measured continuously and dichotomized into indicators of chronic kidney disease (CKD) and albuminuria, respectively. BKMR tested for non-linearity in the exposure-specific responses to evaluate dose-response relationships between mixtures and outcomes and possible interaction effects among exposures. Interactions among continuous outcomes were identified using the NLinteraction package in R.

Results: A higher metals mixture was significantly associated with all four measures of kidney function in dose-response patterns. Pb had the strongest association with eGFR, albuminuria and ACR, and the second strongest association with CKD. We also observed an interaction between Pb and Co for eGFR and an interaction between Pb and Cd for ACR.

Conclusions: Exposure to a co-occurring heavy metals mixture was associated with indicators of poor kidney function. Within this mixture, Pb, Co and Cd considered singly and jointly made the greatest contributions to the observed effects. Future prospective study is needed to confirm the association between metal mixtures and kidney function.