1 INTRODUCTION

Arsenic is a naturally occurring metal, widely distributed throughout the environment in air, water, and food. Exposure to arsenic, particularly inorganic arsenic, is associated with numerous adverse health outcomes (Abdul, Jayasinghe, Chandana, Jayasumana, & De Silva, 2015). Animal models show that arsenic crosses the placenta, accumulating in fetal tissues (Devesa et al., 2006), with birth defects observed in offspring following maternal inorganic arsenic exposure (Abdul et al., 2015). Most epidemiologic studies that examined maternal arsenic exposure and birth defects focused on a single defect/defect group (Brender et al., 2006; Jin et al., 2013, 2016; Mazumdar et al., 2015; Rudnai et al., 2014; Suhl et al., 2018; Wang et al., 2019; White et al., 2019; Zierler, Theodore, Cohen, & Rothman, 1988), or assessed exposure from water (Brender et al., 2006; Marie et al., 2018; Mazumdar et al., 2015; Rudnai et al., 2014; Sanders et al., 2014; Zierler et al., 1988). To our knowledge, only one previous study examined exposure in diet (Suhl et al., 2018).

In the United States, diet is the primary source of arsenic exposure (ATSDR, 2007). The limited information on maternal dietary arsenic exposure and birth defects prompted us to use data from the National Birth Defects Prevention Study (NBDPS) to examine maternal dietary exposure to total (all species) and inorganic arsenic and several noncardiac birth defects (Suhl et al., 2022). We observed positive associations between total arsenic and selected musculoskeletal defects, and inorganic arsenic and selected gastrointestinal, genitourinary, and musculoskeletal defects. Herein, we use NBDPS data to examine maternal dietary exposure to arsenic and congenital heart defects (CHDs), the most prevalent group of major birth defects.

2 METHODS

NBDPS was a multisite, population-based case-control study of major birth defects. The NBDPS protocol, approved by the institutional review board at each site, was published previously (Reefhuis et al., 2015). Briefly, case children diagnosed with one or more eligible defects among pregnancies (live births, stillbirths, and terminations) with estimated dates of delivery during 1997–2011 were identified by birth defect surveillance programs. Clinical geneticists reviewed medical record data abstracted for case children to exclude those with chromosomal or monogenic syndromes. Control children were a random sample of live births without a defect diagnosis delivered in the same time frame and geographic areas as case children.

Clinical geneticists classified case children with a CHD by phenotype level, defect complexity, and extracardiac defects (Botto et al., 2007). Phenotype level ranged from detailed (e.g., tetralogy of Fallot [TOF] with pulmonary atresia) to main (e.g., TOF) to broad (e.g., conotruncal defects). Defect complexity was defined as simple (no other cardiac involvement), association (common, uncomplicated combinations of other heart defects), or complex (those not classified as simple or association). Case children without extracardiac defects were classified as isolated; those with extracardiac defects were classified as multiple.

Mothers of case or control children completed a single telephone interview about medical, environmental, and lifestyle information for the 3 months before conception (B3) through the end of pregnancy. The interview included a 58-item food frequency questionnaire (FFQ), adapted from the Willet FFQ (Willett, Reynolds, Cottrell-Hoehner, Sampson, & Browne, 1987), to collect diet during the year before pregnancy; mothers also reported cereal consumption during B3 through the end of pregnancy.

Mothers of 11,829 control and 32,017 case children participated in NBDPS, including 12,584 with a CHD. To improve homogeneity of our analytic sample, we restricted analyses to mothers of 7,343 case children classified as isolated with a main, simple CHD not classified as “other” or “not otherwise specified.” After excluding mothers who did not complete the FFQ (control = 672, case = 390), reported consumption of <500 or > 5,000 cal/d (control = 271, case = 202), or had missing body mass index (control = 440, case = 268), mothers of 10,446 control children and 6,483 case children were available for analysis.

Our methodology to estimate dietary exposure to total and inorganic arsenic was published previously (Suhl et al., 2022). For an FFQ item, we linked mean total arsenic concentration estimates reported in the U.S. Food and Drug Administration Total Diet Study (TDS; U.S. FDA, 2007, 2017) for corresponding TDS items (55 of 58 FFQ items; Table S1) to estimate an overall mean total arsenic concentration for the FFQ item. Arsenic concentrations below the limit of detection (LOD) were entered as 0 μg/g, consistent with estimated concentrations provided by TDS. For dietary inorganic arsenic concentrations, we applied a similar approach for 24 FFQ items (Table S1) using estimates reported from two communities in Texas (Schoof et al., 1999), except estimates below the LOD were entered as one-half the LOD, consistent with the Texas study.

Total and inorganic arsenic concentration estimates were multiplied by the corresponding grams consumed of each FFQ item/day to determine arsenic consumed/FFQ item/day; summing all consumption estimates produced overall total and inorganic dietary arsenic estimates for each mother (μg/d). Using reported prepregnancy body weight, μg/kg-body weight (bw)/d total and inorganic arsenic estimates were estimated for all analyses. Exposure to arsenic through drinking water was not included in our exposure assessment because of the low proportion (8%) of NBDPS mothers reporting well water use (Suhl et al., 2022).

We evaluated several child and maternal covariates. Child covariates were sex, plurality, and first-degree family history of CHDs. Maternal covariates were age and educational attainment at delivery, race/ethnicity, gravidity, NBDPS site, prepregnancy dietary folate equivalents and tertiles of prepregnancy total energy intake based on distributions among control mothers, along with folic acid supplementation, alcohol consumption, and tobacco smoke exposure during the critical exposure period for heart development (first 3 months of pregnancy), as well as the month before pregnancy to ensure identification of exposures during that month that may have carried over into pregnancy.

Our primary analyses included estimating frequencies and proportions for each covariate and examining categorical tertiles of total and inorganic dietary arsenic exposure (low, medium, and high) based on cutoffs of the distributions among control mothers. Unconditional multivariable logistic regression was applied to estimate odds ratios (ORs) and 95% confidence intervals (CIs), with the low tertile of exposure as the referent category. Multivariable models were constructed by including each covariate individually in a model for each arsenic exposure–CHD pair; covariates that altered the crude estimate for a pair by >10% were included in the final model. Analyses were restricted to CHDs for which there were at least five case mothers included in each tertile of exposure for total or inorganic arsenic. When sparse data prevented model convergence, Firth's logistic regression was used to estimate ORs and 95% profile likelihood CIs (Firth, 1993).

Secondary analyses included restricting to mothers without prepregnancy diabetes and those with singleton pregnancies due to known associations with CHDs, and to mothers with deficient folate intake (no folic acid supplementation and dietary folate <600 μg/d) due to the role of folate in arsenic metabolism (Na et al., 2020). Additional secondary analyses examined additive interaction effects between maternal dietary arsenic (total, inorganic) and tobacco smoke exposure (none, any), which contains high levels of arsenic, during the critical exposure period. Additive interaction was evaluated by examining relative excess risk due to interaction (RERI) and bootstrap 95% CIs for the middle and high tertiles of exposures. Finally, we examined alternate, high-dose thresholds of total and inorganic arsenic exposure (90th percentile among control mothers) and alternate exposure estimates for total arsenic using the maximum concentrations reported in TDS items and including arsenic in breakfast cereal consumption during the 3 months before pregnancy; no inorganic arsenic estimates were available for breakfast cereals. All analyses were conducted using SAS v. 9.4 (SAS Institute Inc., 2013).

3 RESULTS

Frequencies and proportions of child and maternal covariates are presented in Table 1; missing data ranged from 0% to 1.3% for covariates. Tertile cutoffs (μg/kg-bw/d) were 0.07 and 0.22 for total and 0.05 and 0.08 for inorganic arsenic. Compared with the low tertile of total arsenic exposure, positive associations (OR ≥ 1.2) were observed in both middle and high tertiles for perimembranous ventricular septal defect (VSD) and in the high tertile only for double outlet right ventricle-transposition of the great arteries (DORV-TGA), partial anomalous pulmonary venous return (PAPVR), and tricuspid atresia (Table 2). For inorganic arsenic, positive associations were observed in both tertiles for triscuspid atresia and in the high tertile only for DORV-TGA, conoventricular VSD, PAPVR, and pulmonary atresia. Most remaining estimates for total or inorganic arsenic were near or below unity.

Most associations between total or inorganic arsenic exposure and CHDs identified in our primary analyses persisted in secondary analyses (Tables S2–S7). Additional positive associations were observed among mothers with deficient folate intake for dextro-TGA (d-TGA), TOF, and secundum atrial septal defect (ASD2; Table S4). Restricting the sample to non-diabetic mothers (Table S2) or those with singleton pregnancies (Table S3) did not produce additional positive associations, nor did alternate exposure thresholds (Table S5; total = 0.51 μg/kg-bw/d; inorganic = 0.16 μg/kg-bw/d) and estimations using maximum concentrations reported in TDS items and including arsenic in breakfast cereal consumption (Tables S6 and S7). No statistically significant RERIs were observed for high tertiles of arsenic exposure in secondary analyses examining additive interaction effects with tobacco smoke exposure (data not shown). In both primary and secondary analyses, many estimates were imprecise.

4 DISCUSSION

CHDs are among the most common birth defects in the United States (Mai et al., 2019) and suspected to have a multifactorial etiology. To elucidate potential risk factors for these defects, we explored associations between maternal dietary exposure to total and inorganic arsenic and isolated, simple CHDs using data from a large, population-based case-control study. Positive associations were observed in the middle and high tertiles of dietary exposure to total arsenic and perimembranous VSD, and middle and high tertiles of exposure to inorganic arsenic and tricuspid atresia; positive associations were also observed in the high tertile of total or inorganic arsenic exposure and several additional CHDs.

Our findings add to a currently limited literature on maternal arsenic exposure and CHDs in children, including the lack of previous studies of maternal dietary arsenic exposure. Three studies examined maternal drinking water exposure to arsenic and CHDs, with reports of positive associations with VSD and coarctation of the aorta (Zierler et al., 1988), as well as with ASD or any CHD (Marie et al., 2018; Rudnai et al., 2014). Another study reported associations between elevated arsenic in maternal hair samples and simple CHDs, and groups of septal, conotruncal, and left and right ventricular outflow tract defects (Jin et al., 2016). We observed increased, positive associations for total arsenic and d-TGA, TOF, perimembranous VSD, and ASD2 among mothers with deficient folate intake. A recent recovery animal study of CHDs in offspring exposed to arsenic during periconception showed a decrease in occurrence of CHDs following administration of folic acid (Na et al., 2020). The findings support epigenetic changes caused by arsenic toxicity in cardiac tissue.

Primary limitations of our study include examination of dietary arsenic exposure only, which may have underestimated total maternal arsenic exposure; reliance on retrospective self-reports of diet during the year before pregnancy, which may have led to recall bias and missed changes in dietary habits during early pregnancy; and limited availability of inorganic arsenic concentrations in food items, which were derived from foods collected in two communities in Texas and corresponded to only 24 NBDPS FFQ items. Primary strengths of our study include analysis of a large, population-based and racial/ethnically diverse study sample of live births, stillbirths, and elective terminations with CHDs; systematic phenotype classification of case children by clinical geneticists; examination of dietary sources of total and inorganic arsenic, which has not been previously examined for CHDs, and detailed data on folic acid intake, which allowed for exploration of the potential role of folic acid in the relation between dietary arsenic and CHDs.

5 CONCLUSION

We explored associations between maternal prepregnancy dietary exposure to arsenic and isolated, simple CHDs in offspring. Although we observed positive associations for dietary exposure to total or inorganic arsenic and a small number of CHDs, these could be spurious findings and need to be interpreted with caution. Also, our exposure analysis for inorganic arsenic was limited by available food arsenic concentrations. Future studies examining maternal dietary arsenic exposure and CHDs require improved dietary assessment extending into early pregnancy, improved estimates of inorganic arsenic concentrations in foods, and controlling for nondietary sources of arsenic exposure.

FUNDING INFORMATION

This work was funded by the Centers for Disease Control and Prevention cooperative agreements under PA #96043 PA #02081 FOA #DD09-001 FOA #DD13-003 NOFO #DD18-001 to the Centers for Birth Defects Research and Prevention participating in the National Birth Defects Prevention Study (NBDPS) and/or the Birth Defects Study To Evaluate Pregnancy exposureS (BD-STEPS), and grants (U01DD001035 U01DD001223) awarded to the Iowa Center for Birth Defects Research and Prevention.

CONFLICT OF INTEREST

The authors declare no potential conflict of interest.