A prospective study of arsenic and manganese exposures and maternal blood pressure during gestation

Abstract

Background:

Pregnancy is a sensitive time for maternal cardiovascular functioning and exposures to arsenic or manganese may adversely affect blood pressure (BP).

Objectives:

This study examined the associations between arsenic and manganese exposures and maternal BP measured during pregnancy. Effect modification by pre-pregnancy body mass index (BMI) was evaluated.

Methods:

Pregnant women (N = 1522) were recruited for a prospective cohort study in Bangladesh (2008–2011). Exposure to arsenic and manganese was measured in drinking water at <16 weeks gestation and toenails at one-month postpartum. Systolic and diastolic BP were measured monthly. Linear mixed models estimated mean BP and differences in mean BP over gestation for arsenic or manganese exposures and adjusted for covariates.

Results:

Arsenic levels had an increasing dose-response association with maternal BP after 25 weeks gestation. Effect modification was observed for BMI. Participants with lower BMI (<23 kg/m²) exposed to 50 μg/L arsenic had 2.83 mmHg (95% CI:1.74–3.92) greater mean systolic and 1.96 mmHg (95% CI: 1.02–2.91 mmHg) diastolic BP compared to those exposed to ≤ 1 μg/L arsenic at 40 weeks gestation. Participants with higher BMI (≥23 kg/m²) showed a greater mean systolic BP of 5.72 mmHg (95% CI: 3.18–8.27 mmHg) and diastolic BP change of 6.09 mmHg (95% CI: 4.02–8.16 mmHg) at 40 weeks gestation when exposed to 50 μg/L compared to ≤ 1 μg/L arsenic. Participants with lower BMI exposed to drinking water manganese in the 2nd quartile (181–573 μg/L) had 1.04 mmHg higher mean diastolic BP (95% CI: 0.01–2.07 mmHg) at 40 weeks gestation compared to those in the 1st quartile (0.5–180 μg/L).

Conclusion:

Arsenic exposures during pregnancy were consistently associated with increased average maternal systolic and diastolic BP. The effect of manganese on BP was less consistent.

1. Introduction

Populations around the world are chronically exposed to elevated levels of arsenic and manganese from ingesting contaminated drinking water (Ferreccio et al., 2013; Arsenic contamination of groundwater in, 2019; Hasan and Ali, 2010). In particular, Bangladesh has been greatly impacted by high arsenic and manganese levels from drinking water supplied by groundwater tube wells (Arsenic contamination of groundwater in, 2019; Frisbie et al., 2002). Surveys in Bangladesh report that 32% and 61% of potable tube wells have arsenic >50 μg/L (the Bangladesh standard) and manganese >400 μg/L ( World Health Organization’s recommendation), resulting in approximately 30–70 million people being exposed to high levels of these metals (Hasan and Ali, 2010; Ghosh et al., 2020; Johnston, 2011). The realization of widespread arsenic poisoning in Bangladesh has initiated mitigation efforts such as identifying contaminated tube wells, instillation of filtration systems, and full well remediation (Johnston, 2011). However, mitigating manganese exposure has received considerably less attention. This is partly because mitigation efforts have focused on only testing tube wells for arsenic and partly because of limited data demonstrating adverse health effects from manganese ingestion.

Manganese is an essential mineral that is needed for bone growth, digestion, reproduction, immune function, energy metabolism, and the body’s defense against free radicals (Wood, 2009). Yet experimental studies in animal models show manganese effects mitochondrial and cardia myocyte function through calcium regulation, which negatively alters cardiac contraction (Jiang and Zheng, 2005; Dudek and Pytkowski, 1991). Additionally, occupational inhalation epidemiology studies of manganese report it to be a potent neurotoxicant with additional associated effects on the cardiovascular system (ATSDR, 2019). Population-based cohort studies report that high manganese concentrations in drinking water are associated with increased infant mortality rates and central nervous system dysfunction (Ljung and Vahter, 2007; Hafeman et al., 2007). While a cross-sectional study of a representative sample of the U.S. population showed higher urinary manganese levels are associated with lower systolic and diastolic blood pressure (BP) (Wu et al., 2017). Alternatively, general population studies in India and Korea have found increased blood manganese levels to be correlated with hypertensive participants compared to non-hypertensive participants (Taneja and Mandal, 2007) and shown increased hypertension risks (Lee and Kim, 2011). Most of the research on manganese ingestion from contaminated drinking water has focused on neurotoxicity, and there are contradictory data describing its effects on the cardiovascular system. Arsenic, on the other hand, is a well-known toxicant. Chronic arsenic exposure from drinking water increases the risk of type 2 diabetes, cancers of the lung, skin, liver, and bladder, and cardiovascular diseases (CVD) (Smith et al., 1992; Tsai et al., 1999; Navas-Acien et al., 2005) including carotid artery disease, high BP, endothelial dysfunction, and atherosclerosis (Engel et al., 1994; Abhyankar Lalita et al., 2012; Mateen et al., 2017). Arsenic’s mechanism of cardiotoxicity is proposed to be the result of vascular changes caused by damage to endothelial cells and inhibition of angiogenesis (Prozialeck et al., 2008). Epidemiological studies in Bangladesh have shown that elevated arsenic concentrations in tube well water are associated with increased risk of CVD mortality, especially for smokers or people with higher body mass index (BMI) (Chen et al., 2011, 2019). But these studies have typically relied on adult, male populations and there is only one study that has examined the effects of arsenic exposure on cardiovascular outcomes in pregnant women (Farzan et al., 2015) - an important susceptible population. There is also a need to better understand the effects of arsenic and manganese co-exposures on cardiovascular outcomes in pregnant women (Rodrigues et al., 2015; Wright et al., 2006).

Pregnancy is a physiologically dynamic and transformative time for the mother’s cardiovascular system resulting in BP fluctuations during gestation (Clapp and Capeless, 1997). Elevated BP during pregnancy increases risks of placental abruption and blood flow restrictions, labor complications, viability of the fetus, and increases the risk of low birth weight (Irwinda et al., 2016; Barbosa et al., 2015; Allen et al., 2004). It may also increase maternal CVD risks later in life (Shen et al., 2017; Magnussen et al., 2009). In the New Hampshire Birth Cohort Study, Farzan et al. demonstrated that total urinary arsenic concentrations measured during pregnancy were associated with increased systolic BP and pulse pressure over the course of gestation (Farzan et al., 2015). There is also evidence that pregnant women tend to accumulate more manganese during pregnancy compared to non-pregnant women (Tholin et al., 1993). A cross-sectional study conducted in Iran demonstrated evidence that pregnant women with preeclampsia had significantly higher average manganese (3.17 μg/L) compared to pregnant women without preeclampsia (2.32 μg/L) (Vigeh et al., 2006). Additionally, Vigeh et al. has demonstrated increased levels of blood manganese in early pregnancy were associated with gestational hypertension (Vigeh et al., 2013, 2016,Vigeh et al., 2016). There is a paucity of extensive research examining the association between manganese and longitudinal changes of BP during pregnancy. It is important to understand the impact of arsenic and manganese on maternal BP during this critical life stage because reducing exposure to these metals during pregnancy may provide an opportunity to prevent detrimental cardiovascular effects.

Given that Bangladesh is experiencing an epidemiologic transition where infectious disease and maternal mortality are yielding to chronic diseases (Ahsan Karar et al., 1986; Mascie-Taylor, 2012), it is important to examine the effects of commonly encountered environmental pollutants like arsenic and manganese on non-communicable diseases like CVD. This study used an existing prospective birth cohort recruited in Bangladesh that included arsenic and manganese measurements in drinking water and toenails and repeated BP measurements during pregnancy to examine the associations between chronic low-level arsenic and manganese exposures and maternal BP. The availability of toenail measurements as non-invasive biomarkers allow for the assessment of 3–12 months of internal exposures to arsenic and manganese, capturing exposures over pregnancy for our cohort (Kile et al., 2007). It was hypothesized that women with higher metal exposures have higher diastolic and systolic BP levels during pregnancy. Additionally, pre-pregnancy BMI has been associated with higher BP levels for women throughout gestation (Savitri et al., 2016; Mrema et al., 2018; Thompson et al., 2009; Miller et al., 2007). We therefore also hypothesized that prepregnancy BMI could act as an effect modifier due to BMI being a known risk factor for high BP.

2. Methods

2.1. Study population

From 2008 to 2011, pregnant mothers were recruited by Dhaka Community Hospital Trust (DCHT) in Sirajdikhan and Pabna Sadar Upazilas, Bangladesh. These Upazilas were selected for recruitment because surveys indicated a range of arsenic concentrations in groundwater and these exposures were more moderate than in other areas of the country, and DCHT operated health care clinics in these areas (Smith et al., 2000). DCHT is a membership-based health care system that offers primary care, prenatal and midwife services, and arsenic exposure awareness and prevention education for people whose water wells have above 50 μg/L of arsenic (Bangladesh standard). DCHT operates a hospital in Dhaka that takes referrals for more advanced health care needs. Additionally, DCHT has worked extensively with communities in both Sirajdikhan and Pabna to provide safe drinking water options and provide water, adequate sanitation, and hygiene training (Joya et al., 2006). DCHT provided trained medical staff and field technicians who performed the data collection including administering questionnaires, measuring biometric data, midwifery, and gathering water and biomarker samples.

Participants were eligible for this birth cohort study if they were: ≥ 18 years of age with an ultrasound confirmed singleton pregnancy <16 weeks gestation, used drinking water from a ground water tube well, planned to live in the same location for the length of pregnancy, used prenatal care provided by DCHT, and agreed that the birth of their child would be at DCHT or by a DCHT-trained midwife (Kile et al., 2014). Women were incentivized to participate by being provided free prenatal care from DCHT and monthly prenatal vitamins. Trained medical staff conducted monthly prenatal home visits and at 4 weeks postpartum for all live births to collect biological and biometrics data. Over the course of this study, participants provided biological samples of toenails and drinking water. Mothers provided demographic information at < 16 weeks gestation, during monthly follow-up visits, and at time of birth. This study was approved by the Human Research Committees/Institutional Review Boards at Harvard TH Chan School of Public Health, Oregon State University (OSU), and Dhaka Community Hospital Trust. All participants gave informed consent before any data collection took place for the studies.

Participants were included in the analysis if they had at least two measurements of BP and measurements of arsenic and manganese. Of the 1613 women who enrolled in the original cohort, 80 women were excluded because they had less than 2 measurements of BP and 2 women were excluded for lacking exposure data. The study also excluded 1 woman with previous history of diabetes and 8 women with twins. Fig. 1 describes our final study sample of n = 1522.

Fig. 1.

Study population for analysis.

The study used all available data. Missing data was likely not completely at random because exposures to arsenic and potentially manganese are risk factors for adverse birth outcomes (Hafeman et al., 2007; Ahmed et al., 2019). Therefore, this study evaluated the characteristics of people who had the missing data compared to the full cohort and conducted sensitivity analyses with and without missing data (Table 1).

Table 1.

Characteristics of pregnancy cohort at enrollment with at least two measurements of systolic and diastolic blood pressure (N = 1522) and among a subset of women within the cohort who had live births (N = 1146).

	Cohort at Enrollment (N = 1522)	Cohort with Live Births (N = 1146)
Systolic BP at Baseline
Low (Below 90 mmHg)	42 (2.8%)	33 (2.9%)
Normal (90–120 mmHg)	1316 (86.5%)	996 (86.9%)
Elevated (Over 120 mmHg)	164 (10.8%)	117 (10.2%)
Diastolic BP at Baseline
Low (Below 60 mmHg)	33 (2.2%)	28 (2.4%)
Normal (60–80 mmHg)	1266 (83.2%)	958 (83.6%)
Elevated (Above 80 mmHg)	223 (14.7%)	160 (14.0%)
Age (years)
Mean (SD)	22.9 (4.19)	23.0 (4.21)
BMI (kg/m2)
Mean (SD)	20.5 (3.19)	20.5 (3.23)
Education
No school attendance	226 (14.8%)	165 (14.4%)
Primary School	509 (33.4%)	367 (32.0%)
Secondary School and Higher Education	787 (51.7%)	614 (53.6%)
Monthly Household Income (Taka)
>3000	247 (16.2%)	193 (16.8%)
3001–4000	382 (25.1%)	306 (26.7%)
4001–5000	467 (30.7%)	338 (29.5%)
5001+	401 (26.3%)	302 (26.4%)
Missing	25 (1.6%)	7 (0.6%)
Cook Fuel Type
Clean Fuel	1038 (68.2%)	791 (69.0%)
Less Clean Fuel	476 (31.3%)	349 (30.5%)
Refused	5 (0.3%)	3 (0.3%)
Missing	3 (0.2%)	3 (0.3%)
Gestational Age at Enrollment (Weeks)
Mean (SD)	11.4 (3.09)	11.4 (3.03)
Median [Min, Max]	11.0 [4.00, 16.0]	12.0 [4.00, 16.0]
Missing	4 (0.3%)	1 (0.1%)
Prenatal Vitamin Use
Not at all	2 (0.1%)	1 (0.1%)
Yes, only 3–4 times a week	3 (0.2%)	2 (0.2%)
Yes, everyday	1416 (93.0%)	1128 (98.4%)
Missing	101 (6.6%)	15 (1.3%)
Clinic Location
Sirajdikhan	815 (53.5%)	581 (50.7%)
Pabna	707 (46.5%)	565 (49.3%)
Drinking Water Arsenic at enrollment (μg/L)
Mean (SD)	43.4 (102)	45.8 (105)
Median [Min, Max]	2.00 [0.500, 1400]	2.30 [0.500, 1400]
Drinking Water Manganese at enrollment (μg/L)
Mean (SD)	735 (731)	723 (705)
Median [Min, Max]	596 [0.500, 5300]	591 [0.500, 4720]
Toenail Arsenic at <4 weeks postpartum (μg-As/g)
Mean (SD)	2.06 (3.13)	2.07 (3.14)
Median [Min, Max]	0.958 [0.02, 40.6]	0.962 [0.02, 40.6]
Missing	406 (26.7%)	47 (4.1%)
Toenail Manganese at <4 weeks postpartum (μg-Mn/g)
Mean (SD)	8.32 (8.56)	8.29 (8.53)
Median [Min, Max]	6.00 [0.05, 132]	5.97 [0.05, 132]
Missing	410 (26.9%)	51 (4.5%)

2.2. Blood pressure measurements

BP was taken by trained clinicians using a manual arm cuff around the upper portion of the arm, targeting the brachial artery following protocols. Participants rested approximately 10 min prior to measurement. BP readings were recorded with systolic over diastolic measurements in mmHg. BP readings were taken every month during gestation either during a home or clinic visit from the time of enrollment through the month prior to delivery or miscarriage. BP was taken in a home environment and clinic settings. This resulted in up to eight BP readings per participant depending on the gestational age at which participants enrolled in the study and the length of follow up. Mean and median gestational age at enrollment are described in Table 1. Cut off levels of BP are based on American Heart Association readings of systolic blood pressure: < 90 mm Hg as low BP, 90–120 mm Hg as normal BP, and over 120 mm Hg as elevated BP and diastolic as < 60 mm Hg as low BP, 60–80 mm Hg as normal BP, and over 80 mm Hg as elevated BP (Monitoring Your Blood Pressure at Home [Internet], 2021).

2.3. Exposure assessment drinking water arsenic and manganese

Drinking water samples were collected from the groundwater tube wells that were used by participants at enrollment as their primary source of drinking water. Water samples were preserved in the field to a pH < 2 using ultrapure nitric acid. Samples were kept at room temperature and shipped to Harvard TH Chan School of Public Health for analysis. Arsenic and manganese concentrations were measured following EPA methods 200.8 using inductively-coupled plasma mass spectrometry. Quality control measures included field and laboratory blanks, standard reference material, and calibration standards for arsenic and manganese. The limit of detection (LOD) for both drinking water arsenic and manganese levels was 1.0 μg/L. There were 94 samples below the LOD for arsenic and 2 samples below the LOD for manganese. Samples below the limit of detection were given half the LOD of 0.5 μg/L (Lubin et al., 2004).

2.4. Internal exposure assessment toenail arsenic and manganese

Toenail measurements are non-invasive biomarkers that assess approximately 3–12 months of internal exposures to arsenic and manganese (Kile et al., 2007). Nail clippings were collected from all ten toes of the participant one month postpartum. External contamination was removed by sonication in 1% Triton X-100 solution (Sigma-Aldrich, Inc., St. Louis, MO) for 30 min and then rinsed repeatedly with Milli-Q water (Millipore Corporation, Billerica, MA). Samples were microwave acid digested and analyzed for total arsenic and manganese using inductive coupled plasma mass spectrometry (PerkinElmer Model DRC-11 6100, Norwalk, CT). Details of the sample analysis have been published previously (Rodrigues et al., 2015). Quality control procedures included method blanks, certified human hair reference material (CRM Hair; Shanghai Institute of Nuclear Research, Academia Sinica, China), and calibration curves. The concentrations of arsenic and manganese were normalized for nail mass yielding a concentration of μg-As per gram and μg-Mg per gram of toenail. The LOD for toenail arsenic was 0.04 μg/g with 2 samples below the LOD, and the LOD for manganese was 0.1 μg/g with 1 sample below the LOD. Samples below the LOD were given the value of half the LOD of 0.02 μg/g for toenail arsenic and 0.05 μg/g for toenail manganese.

2.5. Covariates

Covariates were chosen a priori from previous literature based on their associations with arsenic, manganese, and cardiovascular outcomes (Farzan et al., 2015; Rahman et al., 2009; Gilbert-Diamond et al., 2011). These covariates are: maternal age (continuous), maternal education as self-reported in categories of highest level of education achieved (no school attendance; primary education; secondary education or higher education), cook fuel type within home (wood or natural gas; crop residue or dung), BMI derived from weight at time of enrollment (kg) and height (m) (continuous), monthly family income reported by the financial provider of the mother (Taka: > 3,000, 3001–4,000, 4001–5,000, and +5001), and clinic location (Pabna or Sirajdikhan).

2.6. Statistical analysis

General descriptive statistics were calculated including means, medians, frequencies, and Pearson correlations. Spaghetti plots of gestational age by systolic or diastolic BP (mm Hg) were examined for fit of data with semi-parametric smoothing via generalized additive modeling (GAM) for drinking water arsenic and manganese quartiles of exposure (Supplemental Figs. S1-S4). Linear mixed models were fit to examine for associations between drinking water and internal (toenail) measurements of arsenic/manganese levels and longitudinal measures of diastolic and systolic BP over gestation. Differences in mean BP were estimated to compare pregnant women who were exposed to different levels of arsenic or manganese. Arsenic and manganese concentrations were evaluated as continuous explanatory variables and as quartiles. Model selection was based on Akaike’s Information Criterion (AIC). When modeled as continuous explanatory variables, drinking water arsenic and manganese, and toenail arsenic and manganese were evaluated on the natural log scale. Additionally, gestational age was modeled as a quadratic function in models relating drinking water arsenic to diastolic BP; otherwise, gestational age was modeled as a linear function. Two-way interactions were evaluated between continuous arsenic or manganese exposures and all covariates (including arsenic and manganese interactions) using likelihood ratio tests (LRT) of α = 0.05. There were significant interactions between arsenic, manganese and BMI. Subsequently, final models were stratified by lower baseline BMI (less than 23 kg/m²) and higher baseline BMI (23 kg/m² and higher). The results of models examining the mean difference of BP in the entire cohort are presented in Supplemental Tables S1 and S2. The stratified models examining the difference in blood pressure are in Tables 2 and 3.

Table 2.

Covariate-adjusted difference in mean systolic and diastolic blood pressures over trimesters of pregnancy comparing women exposed to different concentrations of drinking water arsenic and different quartiles of internal (toenail) arsenic concentrations stratified by body mass index (BMI).

Body Mass Index	Arsenic Exposure Comparisona	Trimesterb	Systolic Blood Pressurec		Diastolic Blood pressurec
			Estimate (95% CI)	P-valued	Estimate (95% CI)	P-valued
Drinking Water Measurements
BMI <23 kg/m2	10 to ≤ 1 μg/L	Trimester 1	−0.92 (−1.45–−0.38)	<0.001	−0.67 (−1.05–−0.30)	<0.001
		Trimester 2	0.19 (−0.23–0.60)	0.4	−0.29 (−0.59–−0.0003)	0.05
		Trimester 3	1.67 (1.02–2.31)	<0.001	1.16 (0.60–1.71)	<0.001
	50 to ≤ 1 μg/L	Trimester 1	−1.56 (−2.47–−0.65)	<0.001	−1.14 (−1.72–−0.50)	<0.001
		Trimester 2	0.32 (−0.38–1.03)	0.4	−0.50 (−0.99–−0.0006)	0.05
		Trimester 3	2.83 (1.74–3.92)	<0.001	1.96 (1.02–2.91)	<0.001
	250 to ≤ 1 μg/L	Trimester 1	−2.20 (−3.48–−0.92)	<0.001	−1.61 (−2.51–−0.71)	<0.001
		Trimester 2	0.45 (−0.54–1.45)	0.4	−0.70 (−1.40–−0.0008)	0.05
		Trimester 3	3.99 (2.45–5.53)	<0.001	2.77 (1.44–4.10)	<0.001
	500 to ≤ 1 μg/L	Trimester 1	−2.48 (−3.92–−1.03)	<0.001	−1.81 (−2.83–−0.80)	<0.001
		Trimester 2	0.51 (−0.61–1.63)	0.4	−0.79 (−1.58–−0.0009)	0.05
		Trimester 3	4.50 (2.76–6.23)	<0.001	3.12 (1.62–4.62)	<0.001
BMI ≥23 kg/m2	10 to ≤ 1 μg/L	Trimester 1	−2.32 (−3.68–−0.96)	< 0.001	−1.56 (−2.42–−0.70)	<0.001
		Trimester 2	0.12 (−1.01–1.24)	0.9	−0.22 (−0.85–0.41)	0.5
		Trimester 3	3.37 (1.87–4.87)	< 0.001	3.58 (2.37–4.80)	<0.001
	50 to ≤ 1 μg/L	Trimester 1	−3.94 (−6.26–−1.63)	< 0.001	−2.65 (−4.10–−1.19)	<0.001
		Trimester 2	0.20 (−1.71–2.11)	0.9	−0.37 (−1.44–0.70)	0.5
		Trimester 3	5.72 (3.18–8.27)	< 0.001	6.09 (4.02–8.16)	<0.001
	250 to ≤ 1 μg/L	Trimester 1	−5.56 (−8.83–−2.30)	< 0.001	−3.74 (−5.79–−1.68)	<0.001
		Trimester 2	0.28 (−2.42–2.98)	0.9	−0.52 (−2.04–0.99)	0.5
		Trimester 3	8.08 (4.49–11.67)	< 0.001	8.60 (5.68–11.51)	<0.001
	500 to ≤ 1 μg/L	Trimester 1	−6.26 (−9.94–−2.59)	< 0.001	−4.21 (−6.52–−1.89)	<0.001
		Trimester 2	0.32 (−2.72–3.36)	0.9	−0.59 (−2.29–1.12)	0.5
		Trimester 3	9.09 (5.06–13.13)	< 0.001	9.67 (6.39–12.96)	<0.001
Toenail Measurements
BMI <23 kg/m2	Q2 to Q1	Trimester 1	−2.34 (−3.79–−0.89)	0.002	−1.68 (−2.67–−0.69)	<0.001
		Trimester 2	−1.40 (−2.43–−0.37)	0.008	−0.91 (−1.55–−0.27)	0.006
		Trimester 3	−0.15 (−1.94–1.65)	0.9	0.12 (−1.05–1.29)	0.8
	Q3 to Q1	Trimester 1	−2.18 (−3.66–−0.71)	0.004	−1.27 (−2.27–−0.28)	0.01
		Trimester 2	−1.19 (−2.26–−0.12)	0.03	−0.78 (−1.45–−0.12)	0.02
		Trimester 3	0.13 (−1.69–1.95)	0.9	−0.13 (−1.31–1.05)	0.8
	Q4 to Q1	Trimester 1	−2.79 (−4.29–−1.29)	<0.001	−1.28 (−2.29–−0.26)	0.01
		Trimester 2	−1.02 (−2.12–0.08)	0.07	−0.54 (−1.23–0.14)	0.1
		Trimester 3	1.33 (−0.50–3.17)	0.2	0.43 (−0.76–1.63)	0.5
BMI ≥23 kg/m2	Q2 to Q1	Trimester 1	−0.51 (−4.03–3.02)	0.8	−0.60 (−2.75–1.56)	0.6
		Trimester 2	0.76 (−1.83–3.36)	0.6	0.71 (−0.62–2.05)	0.3
		Trimester 3	2.46 (−1.31–6.22)	0.2	2.46 (−0.06–4.98)	0.06
	Q3 to Q1	Trimester 1	−1.25 (−4.80–2.29)	0.5	−1.04 (−3.18–1.10)	0.3
		Trimester 2	−0.16 (−2.79–2.48)	0.9	0.02 (−1.33–1.38)	0.9
		Trimester 3	1.30 (−2.49–5.10)	0.5	1.44 (−1.08–3.97)	0.3
	Q4 to Q1	Trimester 1	−4.94 (−8.63–−1.26)	0.009	−2.22 (−4.44–−0.001)	0.05
		Trimester 2	−1.55 (−4.35–1.25)	0.3	0.01 (−1.43–1.45)	0.9
		Trimester 3	2.98 (−0.96–6.91)	0.1	2.98 (0.38–5.58)	0.02

^aQuartiles of toenail arsenic concentration: Q1 = 0.03–0.50 μg/g; Q2 = 0.51–0.98 μg/g; Q3 = 0.99–2.28 μg/g; Q4 = 2.29–40.55 μg/g.

^bTrimester was estmiated for the week at the end of each trimester where trimester 1:12 weeks, trimester 2: 24 weeks, trimester 3: 40 weeks.

^cAdjusted for age, gestational week, income, education, household cook fuel type, location of clinicgestational week² for drinking water diastolic models, and natural log manganese.

^dP-values denotes significant of difference between arsenic exposure comparison from linear mixed models.

Table 3.

The covariate-adjusted difference in mean systolic and diastolic blood pressures over trimesters of pregnancy comparing women exposed to different quartiles of drinking water and internal (toenail) manganese concentrations stratified by body mass index (BMI).

Body Mass Index	Manganese Quartile Comparisona	Trimesterb	Systolic Blood Pressurec		Diastolic Blood pressurec
			Estimate (95% CI)	P-valued	Estimate (95% CI)	P-valued
Drinking Water Measurements
BMI <23 kg/m2	Q2 to Q1	Trimester 1	−1.15 (−2.42–−0.12)	0.08	−0.70 (−1.56–0.16)	0.1
		Trimester 2	−0.07 (−0.99–0.85)	0.9	0.05 (−0.53–0.63)	0.9
		Trimester 3	1.38 (−0.17–2.93)	0.08	1.04 (0.01–2.07)	0.05
	Q3 to Q1	Trimester 1	−0.66 (−1.89–0.57)	0.3	−0.53 (−1.37–0.30)	0.2
		Trimester 2	−1.21 (−2.09–−0.33)	0.007	−0.25 (−0.81–0.30)	0.4
		Trimester 3	−1.94 (−3.47–−0.41)	0.01	0.12 (−0.90–1.13)	0.8
	Q4 to Q1	Trimester 1	−0.05 (−1.28–1.19)	0.9	−0.28 (−1.11–0.56)	0.5
		Trimester 2	−0.62 (−1.50–0.26)	0.2	−0.26 (−0.81–0.29)	0.4
		Trimester 3	−1.38 (−2.91–0.15)	0.08	−0.24 (−1.25–0.76)	0.6
BMI ≥23 kg/m2	Q2 to Q1	Trimester 1	0.60 (−2.54–3.75)	0.7	−0.40 (−2.34–1.55)	0.7
		Trimester 2	0.51 (−1.93–2.95)	0.7	0.39 (−0.86–1.65)	0.5
		Trimester 3	0.39 (−3.20–3.99)	0.8	1.45 (−0.88–3.77)	0.2
	Q3 to Q1	Trimester 1	2.67 (−0.29–5.64)	0.08	0.37 (−1.49–2.23)	0.7
		Trimester 2	1.11 (−1.11–3.34)	0.3	0.40 (−0.74–1.54)	0.5
		Trimester 3	−0.97 (−4.36–2.42)	0.6	0.43 (−1.78–2.65)	0.7
	Q4 to Q1	Trimester 1	−0.49 (−3.54–2.56)	0.8	−0.79 (−2.68–1.11)	0.4
		Trimester 2	0.11 (−2.23–2.44)	0.9	−0.07 (−1.27–1.12)	0.9
		Trimester 3	0.90 (−2.57–4.37)	0.4	0.88 (−1.37–3.13)	0.4
Toenail Measurements
BMI <23 kg/m2	Q2 to Q1	Trimester 1	−0.24 (−1.67–1.20)	0.7	0.16 (−0.81–1.13)	0.7
		Trimester 2	0.43 (−0.58–1.45)	0.4	0.36 (−0.27–0.98)	0.3
		Trimester 3	1.33 (−0.46–3.12)	0.2	0.61 (−0.55–1.78)	0.3
	Q3 to Q1	Trimester 1	−0.29 (−1.75–1.17)	0.7	−0.1 (−1.09–0.89)	0.8
		Trimester 2	1.17 (0.13–2.22)	0.03	0.56 (−0.08–1.21)	0.09
		Trimester 3	3.13 (1.32–4.94)	<0.001	1.44 (0.27–2.62)	0.02
	Q4 to Q1	Trimester 1	0.04 (−1.43–1.50)	0.9	−1.10 (−2.09–0.11)	0.03
		Trimester 2	0.33 (−0.73–1.39)	0.5	−0.42 (−1.07–0.24)	0.2
		Trimester 3	0.72 (−1.09–2.53)	0.4	0.50 (−0.68–1.68)	0.4
BMI ≥23 kg/m2	Q2 to Q1	Trimester 1	0.88 (−2.68–4.44)	0.6	−0.11 (−2.24–2.02)	0.9
		Trimester 2	1.64 (−0.98–4.27)	0.2	0.30 (−1.04–1.64)	0.7
		Trimester 3	2.66 (−1.14–6.46)	0.2	0.85 (−1.73–3.43)	0.5
	Q3 to Q1	Trimester 1	−0.08 (−3.65–3.50)	0.9	−0.20 (−2.33–1.93)	0.9
		Trimester 2	1.81 (−0.83–4.46)	0.2	0.48 (−0.87–1.83)	0.5
		Trimester 3	4.33 (0.55–8.11)	0.03	1.38 (−1.17–3.93)	0.3
	Q4 to Q1	Trimester 1	−3.35 (−6.92–0.22)	0.07	−1.57 (−3.70–0.57)	0.1
		Trimester 2	−0.23 (−2.87–2.41)	0.9	−0.15 (−1.50–1.20)	0.8
		Trimester 3	3.94 (0.14–7.74)	0.04	1.73 (−0.84–4.31)	0.2

^aQuartiles of drinking water manganese at BMI <23 kg/m²: Q1 = 0.5–180 μg/L, Q2 = 181–573 μg/L, Q3 = 574–960 μg/L, Q4 = 961–5300 μg/L and at BMI ≥23 kg/m²: Q1 = 0.5–175 μg/L, Q2 = 176–630 μg/L, Q3 = 631–985 μg/L, Q4 = 986–4720 μg/L. Quartiles of toenail manganese at BMI <23 kg/m²: Q1 = 0.07–3.0 μg/g; Q2 = 3.1–6.0 μg/g; Q3 = 6.1–10.8 μg/g; Q4 = 10.9–132 μg/g and at BMI ≥23 kg/m²: Q1 = 0.07–2.6 μg/g; Q2 = 2.7–5.5 μg/g; Q3 = 5.6–10.9 μg/g; Q4 = 11.0–59 μg/g.

^bTrimester was estmiated for the week at the end of each trimester where trimester 1: 12 weeks, trimester 2: 24 weeks, trimester 3: 40 weeks.

^cAdjusted for age, gestational week, income, education, household cook fuel type, location of clinic and natural log arsenic.

^dP-values denotes significant of difference between manganese quartile comparison from linear mixed models.

The linear mixed model stratified by pre-pregnancy BMI levels was:

$$Y_{ij}=[{\beta }_0+b_{0i}]+t_{ij}[{\beta }_1+b_{1i}]+{\beta }_2{M}_i+{\beta }_3\left(M_i{}^{\star }{t}_{ij}\right)+{\overline{Z}}_i^T{\beta }_Z+{\epsilon }_{ij}$$

where for mother i, at gestational age t_ij Y_ij is the systolic BP or diastolic BP, b_0i and b_1i are the random intercept and slope, M_i is the selected value of drinking water arsenic or quartile of toenail arsenic and manganese exposures, ${\overline{Z_i}}^T$ is the row vector for covariates at baseline, and ε_ij is the residual of the model.

Models were adjusted for all measured covariates, based on literature and previous studies. All models contained random effects for intercepts and the linear component of gestational age that were modeled as bivariate normal with mean vector (0,0) and unknown variances and correlation. Residuals were modeled independent of random effect vectors, and according to a normal distribution with mean 0 and unknown variance. Model assumptions of normality and homoscedasticity were inspected visually through residual plots and empirically through descriptive statistics comparing means and medians. Statistical significance was set to α = 0.05. All analyses were performed in R version 4.0.3.

3. Results

3.1. Baseline blood pressure and exposures in study population

The characteristics of the study participants with at least two measurements of BP showed minimal differences between the enrollment cohort and the cohort with live births at baseline (Table 1). Participants who dropped out of the study or experienced fetal loss did not have postnatal toenail data available, birth outcomes, nor as many monthly BP measurements as participants who had a live birth. Toenail arsenic concentrations (range: 0.03–40.6 μg-As/g) and drinking water arsenic concentrations (range: 0.5–1400 μg/L) were moderately correlated (r = 0.41). Twenty per cent of the participants used water wells that contained arsenic concentrations above 50 μg per liter of water (the Bangladesh cutoff standard). Toenail manganese concentrations were weakly correlated with drinking water manganese concentrations (r = 0.12). Also, drinking water manganese and arsenic concentrations were weakly correlated (r = 0.09), as were toenail arsenic and manganese concentrations (r = 0.09). Drinking water arsenic concentration was slightly higher in the population who had a live birth compared to the larger cohort (45.8 μg/L to 43.3 μg/L), whereas drinking water manganese was slightly higher in the larger cohort compared to those who had a live birth (735 μg/L to 723 μg/L). Toenail arsenic and manganese were relatively similar between the two groups.

3.2. Arsenic exposure and blood pressure

Time-varying associations between arsenic concentrations in drinking water and systolic and diastolic BP were such that pregnant women exposed to higher levels of arsenic had lower mean BPs up to 20–25 weeks of gestation, after which mean BPs were higher for those with increasingly higher arsenic exposure (Fig. 2). Pregnant women in their third trimester (week 40) with prepregnancy BMI of ≥23 kg/m² and exposed to 500 ≥g/L of arsenic in drinking water had higher covariate-adjusted mean systolic BP of 9.09 mmHg (95% CI: 5.06–13.13 mmHg) and higher covariate-adjusted mean diastolic BP of 9.67 mmHg (95% CI: 6.39–12.96 mmHg) compared to those exposed to ≤1 μg/L (Table 2). The corresponding increased mean BPs for those with BMI <23 kg/m² are 4.50 mmHg (95% CI: 2.76–6.23 mmHg) and 3.12 mmHg (95% CI: 1.62–4.62 mmHg) for systolic and diastolic BP, respectively.

Fig. 2.

Predicted mean systolic and diastolic blood pressures over gestational age at five selected concentrations of drinking water arsenic for pregnant women in Bangledesh who have body mass index <23 kg/m² (A1 and A2) or body mass index ≥23 kg/m² (B1 and B2). The five drinking water arsenic concentrations were selected to represent a wide-range of exposures encountered in this population. Estimates are for women located in Sirajdikhan Upazila, secondary school or higher education, 22 years of age, average drinking water manganese intake (5.67 μg/L), and income of 4000–5000+ Taka per month (n = 1169). The models for diastolic blood pressure (A2 and B2) included a quadratic function of gestational age.

Higher BMI was associated with greater differences in mean BP at selected levels of arsenic exposure compared to lower BMI (Fig. 2 and Table 2). Among pregnant women with BMI <23 kg/m² in their third trimester (week 40), those exposed to arsenic in drinking water exceeding the WHO cutoff recommendation (>10 μg/L) had higher covariate-adjusted mean systolic BP of 1.67 mmHg (95% CI: 1.02–2.31 mmHg) compared to those exposed to ≤ 1 μg/L of arsenic. The corresponding increase in mean systolic BP for those with higher BMI was 3.37 mmHg (95% CI: 1.87–4.87 mmHg). Those participants exposed to arsenic at and above the Bangladesh drinking water cutoff (50 μg/L) compared to 1 μg/L of arsenic in drinking water, the covariate-adjusted mean systolic BP was 2.83 mmHg (95% CI: 1.74–3.92 mmHg) higher in the third trimester for pregnant women with lower BMI and it was 5.72 mmHg (95% CI: 3.18–8.27 mmHg) higher for pregnant women with higher BMI. Similar trends occurred for diastolic BP (Table 2).

Similar patterns occurred from linear mixed models in which toenail arsenic was a categorical explanatory variable, namely higher quartiles of arsenic concentration were associated with lower mean BPs early in pregnancy but higher mean BPs later in pregnancy, although covariate-adjusted differences in mean BPs later in pregnancy tended not to be significant (Table 2). Associations between arsenic exposure and mean BP over pregnancy were similar between the larger cohort and the smaller cohort that did not include women who experienced fetal loss, withdrew from the study, or were lost to follow up (data not shown).

3.3. Manganese exposure and blood pressure

Linear mixed models with manganese as a categorical predictor variable had improved fit based on AIC over models with a continuous variable for manganese. Pregnant women exposed to manganese in drinking water showed non-linear, time-varying associations between quartiles of exposure and systolic and diastolic BP (Fig. 3). Pregnant women in their third trimester (week 40) with lower prepregnancy BMI <23 kg/m² and quartile 2 (181–573 μg/L) drinking water manganese exposures had higher covariate-adjusted mean diastolic BP of 1.04 mmHg (95% CI: 0.01–2.07 mmHg) compared to those exposed to quartile 1 (0.5–180 μg/L) (Table 3). However, toenail manganese had a different pattern of association by BMI. Namely, participants with prepregnancy BMI <23 kg/m² and quartile 3 (6.1–10.8 μg/g) of manganese toenail measurements had higher covariate-adjusted mean systolic BP of 3.13 mmHg (95% CI: 1.32–4.94) and diastolic BP of 1.44 mmHg (95% CI: 0.27–2.62) compared to quartile 1 (0.07–3.0 μg/g). Those with higher prepregnancy BMI ≥23 kg/m² and quartile 3 (5.6–10.9 μg/g) and quartile 4 (11.0–59 μg/g) manganese toenail measurements had higher covariate-adjusted mean sytolic BP (4.33 mmHg; 95% CI: 0.55–8.11 mmHg and 3.94 mmHg; 95% CI: 0.14–7.74) respectively, during trimester 3 (week 40) of gestation compared to those with quartile 1 (0.07–2.6 μg/g) measurements.

Fig. 3.

Predicted mean systolic and diastolic blood pressures over gestational age at quartiles of drinking water manganese for pregnant women in Bangledesh who have body mass index (BMI) < 23 kg/m² (A1 and A2) or body mass index ≥23 kg/m² (B1 and B2). Estimates are for women located in Sirajdikhan Upazila, secondary school or higher education, 22 years of age, average drinking water arsenic intake of 1.03 μg/L, and income of 4000–5000+ Taka per month. Quartiles of drinking water manganese at BMI <23 kg/m2: Q1 = 0.5–180 μg/L, Q2 = 181–573 μg/L, Q3 = 574–960 μg/L, Q4 = 961–5300 μg/L; Quartiles of drinking water manganese at BMI ≥23 kg/m²: Q1 = 0.5–175 μg/L, Q2 = 176–630 μg/L, Q3 = 631–985 μg/L, Q4 = 986–4720 μg/L.

4. Discussion

Arsenic and manganese exposures were associated with increased average changes in maternal systolic and diastolic BP over gestation based on BMI levels in this prospective birth cohort recruited in Bangladesh. The shape of these exposure-response relationships differed between the metals. Arsenic had a linear association with maternal BP over gestation, but manganese had a non-linear association with maternal BP over gestation. This finding is significant because mothers exposed to high levels of arsenic and manganese could potentially increase their baseline systolic BPs over 9 mmHg during pregnancy. Other environmental pollutants have also been shown to increase maternal blood pressure. For example, epidemiological studies report that exposures to particulate matter air pollution increases for BP in the third trimester by 2.11 mmHg (95% CI 1.34–2.89) (van den Hooven et al., 2011). Increases in blood pressure are precursors to serious pregnancy-related complications like pre-eclampsia which is one of the leading causes of death for pregnant mothers, especially in low-income countries around the world (Firoz et al., 2011). Maternal hypertensive disorders are common complications during pregnancy and one of the main causes of maternal mortality (Sibai et al., 2005). Gestational hypertension affects nearly 10% of mothers, especially those carrying their first child (Sibai et al., 2005; Moreira et al., 2009).

Findings that environmental drinking water arsenic exposure over 10 μg/L was associated with increased BP during pregnancy is consistent with other studies demonstrating arsenic is a risk factor for cardiovascular diseases (Navas-Acien et al., 2005; Engel et al., 1994; Abhyankar Lalita et al., 2012; Simeonova and Luster, 2004). A cross-sectional study in Bangladesh showed low-level (<50 μg/L) exposures to drinking water arsenic were associated with 1.39 higher odds (95%CI: 1.10–1.75) of systolic hypertension compared to those with <8 μg/L of drinking water arsenic (Chen et al., 2007). Within the U.S., researchers found that populations exposed to over 10 μg/L of arsenic in drinking water had a 1.68 higher odds of high BP diagnosis than those exposed to drinking water arsenic between 2 and 10 μg/L, after adjusting for age, sex, BMI, and smoking status (Zierold et al., 2004). Similar to the findings of our present study a longitudinal study by Farzan et al. found that pregnant women exposed to background levels of arsenic in drinking water had a 0.15 mmHg increase (95% CI: 0.02–0.29) in systolic BP per month of gestation, after adjusting for age, BMI, smoking, marital status, gestational diabetes, education, parity, and number of BP measurements (Farzan et al., 2015). This study was able to adjust for many of the same confounders as Farzan et al. (age, BMI, education, and considering gestational diabetes and parity) and was also able to include manganese as a co-exposure. There is experimental evidence which supports arsenic being toxic to the cardiovascular system. Studies in pregnant mice strains that are susceptible to atherosclerosis showed pregnant dams exposed to arsenic in their drinking water gave birth to male pups with early onset of atherogenesis and altered heart tissue and muscle tone (Srivastava et al., 2007, 2009). In a similar study, pregnant mice exposed to arsenic birthed pups with altered liver development and changes in stress and inflammatory responses, contributing to early onset atherosclerosis (States et al., 2012).

There have very few studies in prospective birth cohorts examining the effect of manganese on the cardiovascular system. Vigeh et al. identified blood manganese to be correlated and associated with hypertension in pregnant women located in Iran and Japan (Vigeh et al., 2013, 2016). The authors found the correlations between higher blood manganese levels and hypertensive women using a healthy population and assuming a linear relationship for hypertensive individuals (n = 16) (Vigeh et al., 2013). The associations between second trimester blood manganese levels and higher the odds of pre-hypertension (OR: 3.28, 95% CI:1.21–8.95) reviewed similar covariates to this current study including BMI and maternal age (Vigeh et al., 2016). Our study built upon the Vigeh et al. studies by including regression models that examined the longitudinal association between manganese and BP per gestational month, as well as the potential for this association to be non-linear. Additional epidemiologic studies are needed to review effect modification and non-parametric dose-response relationships of manganese and possible toxicity on blood pressure (Hasan and Ali, 2010).

Experimental studies provide biological plausibility that manganese can exert toxicity on the cardiovascular system. Specifically, high levels of manganese impair mitochondrial function in rat heart tissue with impaired calcium channel flows (Jiang and Zheng, 2005; Naoki et al., 1992). Rats oral gavaged with 0.25 mm/kg daily manganese showed after 14 days increases in exchangeable calcium within the heart and a 35% and 228% increase in ventricle pressure (Dudek and Pytkowski, 1991). These studies support a hypothesis that the mechanism of action for manganese on the cardiovascular system might be through changes in mitochondrial and cardia myocyte function through calcium regulation negatively altering cardiac contraction (Jiang and Zheng, 2005; Dudek and Pytkowski, 1991; Naoki et al., 1992). Given the potential impacts of manganese on cardiovascular tissue in experimental models, it is plausible that analogous effects could occur in humans and excessive manganese exposures could impact the functioning of the cardiovascular system and manifest as changes in blood pressure. There are limitations on interpreting the metabolism of manganese in rodents to humans. Rodents have shorter gestations with litters resulting in further development occurring outside of the womb compared to humans (Carter, 2020). Subsequently, these inter species differences makes translation manganese toxicity between rodent models and humans challenging. From previous epidemiological and experimental models, our study further supports a growing concern on the role of excess manganese and possible toxicity on cardiovascular health (Hasan and Ali, 2010).

Our study found that drinking water and toenail measurements for arsenic were strongly correlated but weakly correlated for manganese. This suggests that there were alternative sources of exposures to manganese beyond drinking water in this population. This likely includes dietary sources such as legumes, rice, and leafy vegetables (ATSDR, 2019). It is also important to note that drinking water is indicative of current exposures whereas toenail measurements indicate cumulative exposure over the previous 3–12 months (Karagas et al., 2000). Only a few women in our study were categorized as hypertensive during pregnancy, had a history of diabetes, or experienced symptoms that would indicate the possibility of developing pre-eclampsia. Thus, our population did not have many risk factors for cardiovascular disease. Our study found that BMI was an effect modifier. Overall, individuals with higher BMI at the time of enrollment (≤16 weeks gestational age) tended to have greater changes in blood pressure over gestation. Previous studies with prospective cohorts have shown that women with higher pre-pregnancy BMI had significantly higher BP throughout gestation (Savitri et al., 2016; Mrema et al., 2018). Our analysis investigated changes in BP in each trimester. This provides insights into how gestational age acts on arsenic and manganese related changes to BP.

Other strengths of this study include that the cohort was healthy with little or no reported use of tobacco products. Generally, women in our study adhered to the prenatal care specifically taking prenatal vitamins which helps possibly reduce the impacts of arsenic toxicity by increasing the efficiency of 1-carbon metabolism (Vahter, 2007). Arsenic, manganese, and blood pressure were measured prospectively with arsenic and manganese measured both in drinking water at the time of enrollment which occurred early in pregnancy and in toenails collected <4 weeks postnatal which would reflect cumulative internal dose over the course of gestation. This dual exposure assessment approach yielded similar findings which suggests that the risk of exposure misclassification is minimized. Drinking water is a proxy for current arsenic and manganese exposures, whereas toenail measurements provide an integrated proxy for exposures of all routes that occurred on average nine months before the sample was collected (Kile et al., 2007; Karagas et al., 2000). Our study has repeated measures of blood pressure that were taken in a home environment by trained professionals which can reduce stress experienced by participants (Monitoring Your Blood Pressure at Home [Internet], 2021; Shahab et al., 2019). There are several limitations in this study. The first BP reading was taken at enrollment for ≤16 weeks gestation which varied across participants. Subsequently, the start time of measurements and duration of observations varied across the population. Also, we did not include a BP reading at the time of delivery which means our models might be missing the effect of arsenic and manganese on BP at birth. Our study offered prenatal vitamins, prenatal care, birthing services, and arsenic exposure education which could potentially allow our population to be different than the general Bangladeshi population not receiving these services and potentially underestimate the effects of arsenic and manganese exposure. We found that manganese did not have a linear response pattern through drinking water exposures, and modeling assumptions make it difficult to examine potential additive or multiplicative interactions of arsenic (linear) and manganese (nonlinear) in the linear-mixed model with random slopes framework. Future studies must consider possible alternative statistical approaches for evaluating interaction and non-linear dose-responses.

5. Conclusions

Pregnancy is a physiologically dynamic life stage where both the mother and fetus are vulnerable to environmental contamination. Exposure to arsenic and manganese during pregnancy increased maternal BP and may exacerbated by maternal pre-existing conditions such as increased body mass index. Both arsenic and manganese can be removed from drinking water by water treatment or by adopting safe drinking water options making them a modifiable risk factor for maternal BP. The consistency of research showing that arsenic increases the risk of CVD suggests that clinical providers help educate the public on how to test drinking water for arsenic or exposures to arsenic and manganese contaminated drinking water is a global issue and may increase the risks of elevated BP during pregnancy. As CVD continue to be leading causes of morbidity and mortalities around the world, it is important to consider environmental contamination that could amplify risks of CVD in pregnant populations.