Associations between prenatal arsenic exposure with adverse pregnancy outcome and child mortality

Abstract

1. Introduction

Arsenic is ubiquitous in the environment, with human exposure occurring through dietary intake, inhalation of contaminated air, and ingestion of contaminated soil/dust (Joseph et al. 2015). However, the consumption of arsenic-contaminated drinking water is the major exposure route that affects more than 200 million people worldwide, including approximately 77 million in Bangladesh and 17 million in the United States (US) (BBS/UNICEF 2014; IARC 2004). With respect to its frequency, toxicity, and potential for human exposure, arsenic holds the highest ranking since 1997 on the US Agency for Toxic Substances and Disease Registry (ATSDR) substance priority list.

Chronic exposure to arsenic has been associated with a number of health outcomes, including increased risk of cancers (skin, lung, liver, bladder, and kidney), cardiovascular disease (coronary heart disease, acute myocardial infarction, and hypertension), respiratory disease, and diabetes mellitus (Abdul et al. 2015; Maull et al. 2012; Moon et al. 2012; Sanchez et al. 2016). While there is extensive literature on the health impacts of arsenic exposure in adult populations and a growing literature on impaired neurodevelopment function in children (Tsuji et al. 2015), there is relatively little epidemiologic research evaluating the effects of in utero arsenic exposure on pregnancy outcomes and early life. Arsenic readily crosses the placental barrier and thus may influence fetal development. Strong correlations have been observed between concentrations of arsenic in placenta and cord blood arsenic levels in an Argentine population (Concha et al. 1998), maternal blood and cord blood arsenic levels in a Bangladeshi population (Hall et al. 2007), and placental, maternal, and infant arsenic levels in a US population (Punshon et al. 2015).

A recent meta-analysis evaluating associations of arsenic exposure with adverse pregnancy outcomes and infant mortality reported significantly elevated risks of spontaneous abortion, stillbirth, neonatal mortality, and infant death based on published research from Asia and the US (Quansah et al. 2015). However, the majority of the existing studies utilized ecologic exposure assessments of arsenic concentrations in drinking water. The best evidence to date comes from a prospective cohort study of women in rural Bangladesh with individual-level measures of urinary total arsenic concentrations during pregnancy. The study observed a significant dose-response relationship with infant mortality as well as elevated risks of spontaneous abortion and stillbirth, although these were associated with wide confidence intervals and no clear dose-response association (Rahman et al. 2010).

Thus, prospective, individual-level evidence supporting adverse effects of arsenic exposure with pregnancy outcomes and infant mortality is still limited. In this study, we sought to evaluate the association of individual-level prenatal arsenic exposure based on maternal creatinine-adjusted urinary total arsenic concentration with risk of adverse pregnancy outcome (stillbirth, spontaneous abortion, and therapeutic/elective abortion) as well as child mortality in a pregnancy study among 498 women nested in a larger population-based cohort in rural Bangladesh.

2. Methods

3. Results

Maternal characteristics for all observed pregnancies are shown in Table 1. The mean maternal age of the study sample was 32.2 years. The median creatinine-adjusted urinary arsenic concentration was 555 μg/g creatinine, ranging from 17 to 3,712 μg/g creatinine. Adverse pregnancy outcomes were reported in 109 (18.3%) of the reported pregnancies in the study, including 23 (3.9%) stillbirths, 60 (10.0%) spontaneous abortions, and 26 (4.4%) therapeutic/elective abortions. While almost all females in the study were non-smokers (n=594, 99.2%), 200 (39.1%) women reported regular exposure to environmental tobacco smoke inside their homes during pregnancy. Increased maternal age (OR=1.12; 95% CI=1.07, 1.17) and prior stillbirth pregnancy (OR=3.64; 95% CI=1.66, 7.98) were associated with having an adverse pregnancy outcome; whereas, parity (OR=0.68; 95% CI=0.50, 0.94) was inversely associated with having an adverse pregnancy outcome.

Associations between prenatal arsenic exposure and adverse pregnancy outcomes are summarized in Table 2. Higher prenatal urinary total arsenic concentration (>555 μg/g creatinine) was associated with increased risk (adjusted OR=1.59; 95% CI=1.02, 2.46) of adverse pregnancy outcome, defined as stillbirth, spontaneous abortion, or therapeutic/elective abortion. When considering creatinine-adjusted arsenic concentration as a continuous exposure measure, a 50 μg/g increase in arsenic exposure was associated with a 2% (95% CI=1.01, 1.04) increased risk of adverse pregnancy outcome. Additionally, we conducted subgroup analyses to evaluate the associations between prenatal arsenic exposure and each individual adverse pregnancy outcome. A significantly increased risk of stillbirth was associated with high prenatal arsenic exposure (adjusted OR=2.50; 95% CI=1.04, 6.01). A 50 μg/g increase in creatinine-adjusted arsenic concentration was associated with a 2% (95% CI=1.00, 1.05) increased risk of stillbirth when modelling arsenic as a continuous exposure measure.

In Table 3, associations between prenatal arsenic exposure and child mortality are summarized. Among the 489 livebirths, we ascertained vital status through November 2015 on 483 (98.8%) children, with 21 deaths observed. Although the confidence intervals included the null, an elevated risk of child mortality was observed in relation to higher prenatal arsenic exposure (adjusted HR=1.92; 95% CI=0.78, 4.68). When considering creatinine-adjusted arsenic concentration as a continuous exposure measure, a 50 μg/g increase in creatinine-adjusted arsenic concentration was associated with a 4% (95% CI=1.01, 1.07) increased risk of child mortality. In subset analyses evaluating the association with infant mortality (death before 1 year of age), 12 infant deaths were observed. An elevated risk of infant mortality was also observed in relation to higher prenatal arsenic exposure (adjusted HR=2.30; 95% CI=0.63, 8.41). A 50 μg/g increase in creatinine-adjusted arsenic concentration was associated with a 7% (95% CI=0.99, 1.15) increased risk of infant mortality when modelling arsenic as a continuous exposure measure.

Sensitivity analyses were conducted to evaluate associations between prenatal arsenic exposure and adverse pregnancy outcomes as well as child mortality among pregnancies with prenatal arsenic exposure derived from within one year of the reported pregnancy (restricted to 260 (43.5%) pregnancies and 203 (41.5%) livebirths from the primary analyses). Findings from these sensitivity analyses were not appreciably different from the overall analyses presented (data not shown). We also conducted the overall analyses unadjusted for urinary creatinine as well as separately adjusting for urinary creatinine as an independent covariate in the model (Supplemental Table 1), with no appreciable difference in the results from those presented.

4. Discussion

In the present study, we observed an association between high prenatal arsenic exposure (>555 μg/g creatinine) in relation to adverse pregnancy outcomes, particularly stillbirths. There was also modest evidence of elevated risk of child mortality in relation to high prenatal arsenic exposure.

An increased risk of adverse pregnancy outcomes in relation to prenatal arsenic exposure has been observed previously and evaluated in a recent meta-analysis (Quansah et al. 2015). The strongest association in the present analyses was observed for stillbirths in relation to prenatal arsenic exposure. In previous studies, no significant association with stillbirths was observed in relation to well water arsenic concentration (Myers et al. 2010; Rudnai et al. 2006) and prenatal urinary total arsenic concentration (Rahman et al. 2010). However, increased associations were observed with well water arsenic concentration in other studies (Ahmad et al. 2001; Cherry et al. 2008; Hopenhayn-Rich et al. 2000; Milton et al. 2005; von Ehrenstein et al. 2006) as well as with airborne arsenic emissions (Ihrig et al. 1998). For spontaneous abortion, a significant increased risk was observed in relation to higher arsenic levels in drinking water in a large ecological study (Rudnai et al. 2006). However, in a population-based prospective cohort study, no significant association between prenatal urinary total arsenic concentration and spontaneous abortion was observed, although there was an increased association with infant mortality (Rahman et al. 2010). A recent case-control study conducted in Romania also similarly showed no association between drinking water arsenic and spontaneous abortion (Bloom et al. 2014).

Increased risks have been previously observed with child mortality (Hall et al. 2007; Hopenhayn-Rich et al. 2000; Myers et al. 2010; Rahman et al. 2010) in relation to arsenic levels in well water, although some of these associations were associated with wide confidence intervals. While we did not have data available on cause of death in this study sample, it is possible that in utero arsenic exposure is associated with an increased risk of acute lower respiratory infections, the leading cause of childhood morbidity and mortality globally (Liu et al. 2015). Prior studies have shown that prenatal arsenic exposure is associated with both the frequency and severity of respiratory tract infections among infants (Farzan et al. 2016; Rahman et al. 2011).

In many of the previous studies, prenatal arsenic exposure was defined based on an ecologic exposure measure from only a single source (Ahmad et al. 2001; Aschengrau et al. 1989; Cherry et al. 2008; Rahman et al. 2007); whereas, in the present study, creatinine-adjusted urinary total arsenic concentrations were used. This biomarker of arsenic exposure reflects arsenic exposure from multiple sources, including drinking water, food consumption, soil, and dust (Hughes 2006).There are certain limitations in the present study that we consider. First, the proportion of spontaneous abortion events reported in the population is lower than expected. Since pregnancy was self-reported by female study participants, it is likely that participants were unaware of some early pregnancy losses. The underreporting and misclassification of spontaneous abortion outcomes in this study could have led to an underestimation of the association between prenatal arsenic exposure and early pregnancy loss. Second, previous studies have shown that arsenic metabolism efficiency among pregnant women varies across trimesters, with evidence to suggest that metabolism efficiency increases in the first trimester (Gardner et al. 2011; Hopenhayn et al. 2003). However, it was not possible to evaluate the effect of trimester-specific arsenic exposure in this study. Women enrolled in the BEST cohort had a spot urine sample collected every two years for the measurement of urinary total arsenic concentration, from which prenatal arsenic exposure was assigned for the present study. Therefore, the potential misclassification of arsenic exposure might obscure trimester-specific associations. However, since the correlation of urinary total arsenic concentration between biennial samples among women in this study is high (r=0.62-0.68), we infer that the classification of prenatal arsenic exposure in this study is reasonable despite not being systematically measured during pregnancy. Given this, arsenic exposure across trimesters is also likely to be highly correlated; therefore, we would not be able to evaluate trimester-specific effects in this study population. Third, spot urine samples are more influenced by factors such as hydration status and physical activity than 24-hour urine. In spite of the potential higher intra-individual variation, it has been reported that spot urine samples may be adequate biomarkers to derive reasonably accurate measurements of environmental exposure (Rivera-Nunez et al. 2010; Woods et al. 1998). Urinary creatinine concentrations were also used to adjust urinary arsenic for variation in hydration status. Although creatinine adjustment could introduce bias to the association estimates since urinary creatinine is affected by age, sex, and body size (Nermell et al. 2008), no appreciable differences in results were observed based on the unadjusted urinary total arsenic concentration (μg/L) analyses. Moreover, urinary arsenic concentration is primarily reflective of recent arsenic exposure and not cumulative tissue burden. However, urinary total arsenic concentration may be a relatively good marker of exposure in this study population since exposure is chronic and primarily through drinking water. It will be important for future research to investigate the relationship between other biomarkers of arsenic exposure (e.g., toenail or blood arsenic concentrations) and adverse pregnancy outcome. Fourth, it is possible that other variables, such as other environmental exposures or genetic factors that were unmeasured in this study population, may be unaccounted for in our statistical analyses. Additionally, data after baseline was not available on time-varying covariates (e.g., exposure to passive tobacco smoke) for the childhood mortality analyses, which may also have resulted in residual confounding in those analyses. Fifth, logistic regression analyses were conducted in the present study to estimate odds ratios, which may be an overestimate of the relative risks since the prevalence of adverse birth outcomes is common in this population (18.3%). In a comparison of the crude odds ratio estimates to crude relative risk estimates from a log-binomial model, no appreciable overestimation was observed; therefore, we deem the odds ratios to be a good approximation of the underlying relative risks in this analysis. Finally, the study sample included only women with manifest arsenic skin lesions. These women may be more susceptible to arsenic due to genetics or other risk factors, which may have implications for the generalizability of these results to less susceptible populations.

Despite the limitations noted, there is biological support for our findings stemming from several animal studies. Specifically, mice treated by sodium arsenite showed that meiotic aberration caused by arsenite may contribute to decreased preimplantation development and further disrupt embryo development (Navarro et al. 2004). Furthermore, a study in pregnant mice indicated that arsenic exposure is associated with mammalian spontaneous abortion by aberrant placental vasculogenesis and placental insufficiency (He et al. 2007). Particular strengths of the study include individual-level measurement of urinary total arsenic concentration and the prospective study design.

5. Conclusions

The epidemiologic literature evaluating prenatal arsenic exposure and pregnancy outcomes is still equivocal. The present study provides evidence supporting an adverse association between prenatal arsenic exposure and stillbirth in a population with moderate to high arsenic exposure, as well as a possible adverse association with child mortality. These findings encourage further research to evaluate the health impacts of prenatal and early life arsenic exposure in children as well as arsenic remediation strategies for women of child-bearing age.

Supplementary Material

Acknowledgments

Footnotes

References