Joint-Associations of Metal Exposure Biomarkers and Systolic Blood Pressure in the United States: Findings from the 1999-2004 National Health and Nutrition Examination Survey (NHANES)

Abstract

Characterizing the joint toxicity of multiple exposures can be challenging due to the large number of potential exposures and limited prior knowledge of how those combined effects may manifest. One approach to modeling exposure combinations is recursive partitioning, but these algorithms traditionally have not been appropriate for data collected via a complex survey design. We used recursive partitioning for modeling survey data (RPMS) which appropriately accounts for the clustering, strata, and unequal probability of selection inherent in complex national survey data to estimate the joint associations between 11 metals biomarkers with systolic blood pressure (SBP) in the National Health and Examination Survey (1999-2004), adjusting for age, sex, race, and smoking status. We found that those with urinary lead (Pb) > 0.99 μg/g creatinine had a mean SBP = 123.17 mmHg, while those with Pb ≤ 0.99 μg/g creatinine had a mean SBP = 117.79 mmHg, conditional on having higher barium (Ba > 1.19 μg/g creatinine) and antimony (Sb > 0.11 μg/g creatinine). Interestingly, Pb was not significantly associated with SBP in traditional survey linear regression using the above threshold (β1 = 0.70, 95% CI = -0.93, 2.32) or log-transformed Pb (β1 = -0.19, 95% CI = -1,26, 0.88). We also observed that the joint-associations differed by race and Hispanic origin, although Pb was always involved. The highest predicted SBP was conditional on blood Cd > 0.25 μg/L and blood Pb > 1.75 μg/dL for black Americans (SBP = 128.66 mmHg), on urinary Ba > 1.44 μg/g creatinine and urinary Pb > 1.25 μg/g creatinine among Mexican Americans (SBP = 125.99 mmHg) and on urinary Ba > 1.51 μg/g creatinine, urinary Pb > 0.67 μg/g creatinine, and urinary Sb > 0.16 μg/g creatinine for non-Hispanic white Americans (SBP = 123.60 mmHg). This work demonstrates the utility of survey-consistent recursive partitioning to identify joint-associations in large survey datasets with rich environmental exposure data.