Environment-Wide Association Study of Chemical Biomarkers and Health Outcomes in NHANES 2017–2018: Discovery, Validation, and Dose–Response Analysis

Abstract Background Environment-wide association studies (ExWAS) offer a systematic approach to identifying chemical biomarker–health outcome associations, yet few have applied rigorous multi-stage validation. Methods We screened 92 chemical biomarkers against 48 health outcomes in NHANES 2017–2018 (2,796 tests across four screening rounds; not all chemicals were crossed with all outcomes). Associations passing an initial FDR screen were subjected to cross-cycle validation in NHANES 2015–2016—the primary inferential safeguard given the adaptive screening design—followed by dose–response analysis and multiple sensitivity specifications. Survey-weighted regression models adjusted for age, sex, race/ethnicity, poverty–income ratio, BMI, and smoking. Results Of 26 associations passing FDR correction, 21 were testable in cross-cycle validation; of these, 15 (71%) replicated with concordant direction and p < 0.05 in a temporally independent NHANES 2015–2016 sample. Of these 15, 14 remained robust after analyte-specific sensitivity checks; urinary creatinine adjustment identified one association (iodine–BMI) as a dilution artifact. Two novel findings emerged: dimethylarsonic acid with uric acid (β = 0.20 mg/dL per log-unit DMA, 95% CI: 0.15–0.26) and urinary perchlorate with BUN (β = 1.21 mg/dL per log-unit perchlorate, 95% CI: 0.97–1.45); a third high-novelty association (methylmercury–waist circumference) is likely explained by fish consumption patterns. Conclusions Multi-stage ExWAS with cross-cycle validation identified 14 robust chemical–health associations. Two novel findings—DMA–uric acid and perchlorate–BUN—survived all sensitivity checks and warrant prospective investigation.