Public Water Arsenic and Birth Outcomes in the Environmental Influences on Child Health Outcomes Cohort

Abstract

Importance: Inorganic arsenic is associated with adverse birth outcomes, but evidence is limited for public water concentrations (modifiable by federal regulatory action) in US populations.

Objective: To evaluate the association between prenatal public water arsenic exposure below the federal regulatory standard of 10 μg/L and birth outcomes in the US.

Design, setting, and participants: This cohort study analyzed observational pregnancy cohort data from the Environmental Influences on Child Health Outcomes (ECHO) Cohort for birthing parent-infant dyads from 35 pregnancy cohort sites. Infants were born between 2005 and 2020. The data were analyzed between 2024 and 2025.

Exposure: Individual, time-weighted, mean prenatal public water arsenic exposures were estimated by joining Zip Code Tabulation Area-level public water arsenic concentrations with monthly residential history data during pregnancy.

Main outcome and measure: Adjusted risk ratios (RRs) of preterm birth, low birth weight, and small for gestational age were evaluated. Adjusted RRs, mean differences in birth weight-for-gestational age z score and birth weight, and the geometric mean ratio of gestational age at birth were calculated via cubic splines, per 1 μg/L higher prenatal water arsenic, and across policy-relevant categories of exposure.

Results: The cohort comprised 13 998 birthing parents (mean [SD] age, 30.8 [5.6] years) of whom 4.5% were of American Indian, Alaska Native, Native Hawaiian, or Pacific Islander; 7.2% Asian; 12.4% Black; 56.1% White; 4.2% multiple races; and 8.5% another race and 28.1% were of Hispanic/Latino and 70.4% non-Hispanic/Latino ethnicity. Prenatal public water arsenic ranged from less than 0.35 to 37.28 μg/L. In spline models, prenatal public water arsenic was associated with a higher risk of low birth weight, lower birth weight, and lower birth weight-for-gestational age z score, although effect estimates lacked precision. The RR of low birth weight per 1 μg/L higher prenatal water arsenic was higher among Black (1.02; 95% CI, 1.01-1.03), Hispanic/Latino (1.07; 95% CI 1.02-l.12), and White (1.04; 95% CI, 102-1.06) birthing parents, and the RR for preterm birth was higher among Hispanic/Latino birthing parents (1.05; 95% CI, 1.01-1.09). The mean difference of birth weight and birth weight-for-gestational age z score per 1 μg/L higher prenatal water arsenic was more pronounced among White birthing parents (-10 g [95% CI, -17 to -3 g]; -0.02 SDs [95% CI -0.03 to -0.01 SDs]). No evidence that prenatal public water arsenic mediated the association between birthing parent race and ethnicity and adverse birth outcomes was observed.

Conclusions and relevance: In this cohort study of birthing parent-infant dyads across the US, arsenic measured in public water systems was associated with birth outcomes at levels below the current US Environmental Protection Agency's maximum contaminant level. The findings suggest that further reducing the maximum contaminant level for arsenic may decrease the number of infants with low birth weight in the US.

Figure 1.. Zip Code Tabulation Area–Level, Population-Weighted, Mean Public Drinking Water Arsenic Concentrations (2017-2019) Across the US Overlaid With the Number of Participants From Each State (N = 13 998)

Individual-level, time-weighted prenatal exposure estimates were derived using birthing parent residential address, gestational timing, and Zip Code Tabulated Area–level contaminant concentrations averaged across the 3-year periods (available from 2006 to 2019). Public water arsenic concentration categories correspond to actual and potential regulatory thresholds (5.00 μg/L is the maximum contaminant level for Denmark and the states of New Jersey and New Hampshire; 1.00 μg/L is the maximum contaminant level for the Netherlands) and the modal laboratory limit of detection divided by the square root of 2 (0.35 μg/L). Public water arsenic concentrations were not available in Michigan for the 2017-2019 period; therefore, estimates from 2009 to 2011 (most recently available period) are presented.

Figure 2.. Restricted Cubic Spline Models of the Association Between Prenatal Public Drinking Water Arsenic Exposure and Adverse Birth Outcomes in the Environmental Influences on Child Health Outcomes Cohort (N = 13 998)

Models are generalized estimating equations with participants clustered within cohorts (n = 35) using an exchangeable correlation structure. Model 1 was adjusted for birthing parent age (categorical) and education (categorical). Model 2 was additionally adjusted for parity, prepregnancy birthing parent body mass index, prenatal tobacco use, season of conception, and Zip Code Tabulated Area population density. The reference is set to 0.35 μg/L (shown in red, corresponding to undetectable public water arsenic concentrations) with knots at the 67th and 83rd percentiles of the distribution above 0.35 μg/L. Participant prenatal water arsenic concentrations less than 0.35 μg/L represent arsenic concentrations measured below the limit of detection, which were imputed as the limit of detection divided by the square root of 2. Although laboratories reported different limits of detection for different compliance monitoring records, the modal value of the limit of detection divided by the square root of 2 was 0.35 μg/L. The coefficients for the second spline term were −58 (95% CI, −87 to −28) for birth weight; 0% for gestational age; −0.10 (95% CI, −0.16 to −0.03) for birth weight–for–gestational age z score; 1.39 (95% CI, 1.10-1.76) for low birth weight; 1.12 (95% CI, 0.92-1.36) for preterm birth; and 1.22 (95% CI, 1.07-1.40) for small for gestational age.

Figure 3.. Associations Between Prenatal Public Water Arsenic Exposure and Birth Outcomes in the Environmental Influences on Child Health Outcomes Cohort in Stratified Subgroup Analyses (N = 13 998)

Associations represent effect estimates per a 1.00 μg/L higher prenatal public water arsenic exposure in linear models (model 2 adjustments). Square size indicates the number of participants in each subgroup. Effect estimates were not reported for subgroups with less than 50 events or when models did not converge because of small sample/case size. Wald test P values reflect statistical interactions across all strata, including those with fewer than 50 cases and when models did not converge.