Arsenic Exposure from Drinking Water, Dietary Intakes of B Vitamins and Folate, and Risk of High Blood Pressure in Bangladesh: A Population-based, Cross-sectional Study

Abstract

The authors performed a cross-sectional analysis to evaluate the association between arsenic exposure from drinking water and blood pressure using baseline data of 10,910 participants in the Health Effects of Arsenic Longitudinal Study in Bangladesh (October 2000–May 2002). A time-weighted well arsenic concentration (TWA) based on current and past use of drinking wells was derived. Odds ratios for high pulse pressure (≥55 mmHg) by increasing TWA quintiles (≤8, 8.1–40.8, 40.9–91.0, 91.1–176.0, and 176.1–864.0 μg/liter) were 1.00 (referent), 1.39 (95% confidence interval (CI): 1.14, 1.71), 1.21 (95% CI: 0.99, 1.49), 1.19 (95% CI: 0.97, 1.45), and 1.19 (95% CI: 0.97, 1.46). Among participants with a lower than average dietary intake level of B vitamins and folate, the odds ratios for high pulse pressure by increasing TWA quintiles were 1.00 (referent), 1.84 (95% CI: 1.07, 3.16), 1.89 (95% CI: 1.11, 3.20), 1.83 (95% CI: 1.09, 3.07), and 1.89 (95% CI: 1.12, 3.20). The odds ratios for systolic hypertension suggest a similar but weaker association. No apparent associations were observed between TWA and general or diastolic hypertension. These findings indicate that the effect of low-level arsenic exposure on blood pressure is nonlinear and may be more pronounced in persons with lower intake of nutrients related to arsenic metabolism and cardiovascular health. Future research is needed to evaluate the effect of low-level arsenic exposure on specific cardiovascular outcomes.

Epidemiologic studies have associated high-level arsenic exposure from drinking water (average level, >300 μg/liter) with elevated risks of vascular diseases, including peripheral vascular diseases (1), blackfoot disease (2), hypertension (3, 4), ischemic heart disease (5, 6), cerebrovascular disease (7), and carotid atheroslerosis (8). However, little is known about the associations between low-level arsenic exposure from drinking water (<100 μg/liter or <50 μg/liter) and vascular diseases, partly because of lack of detailed and reliable individual-level data on exposure and/or outcome (9).

The elevated groundwater arsenic concentration in Bangladesh has raised major public health concerns. More than 50 million people have been chronically exposed to drinking water with arsenic concentrations exceeding the World Health Organization standard (10 μg/liter) (10). The Health Effects of Arsenic Longitudinal Study was established to evaluate the health effects of arsenic exposure in a rural Bangladeshi population. Specifically, the large variation in arsenic exposure among study participants provides a unique opportunity to evaluate the health effects of arsenic exposure in dose ranges that are relevant to millions of people in many parts of the world, including the United States (11, 12).

Hypertension, or high blood pressure, is the chronic state of elevated pressure in the arteries. Prospective large epidemiologic studies have unequivocally demonstrated a strong, direct relation between high blood pressure and mortality from cardiovascular diseases (13). Blood pressure has two components: Systolic blood pressure (SBP) is measured while the heart contracts and pumps blood into the arteries, and diastolic blood pressure (DBP) is measured while the heart fills with blood. Previous studies of arsenic exposure and blood pressure are limited in that they were conducted in populations with very high arsenic exposure and that they only assessed the risk of general hypertension, defined as having a SBP of ≥140 mmHg and/or a DBP of ≥90 mmHg (14, 15). However, besides general hypertension, several dimensions of blood pressure have been associated with an increased risk of vascular disease. Clinic-based measurements that predict vascular disease include SBP, DBP, mean arterial pressure (defined as ⅓ × (SBP + 2 × DBP)), and pulse pressure (the difference between SBP and DBP) (16–18).

In addition, previous studies did not address the potential of interindividual variability in susceptibility to the adverse effect of arsenic exposure on blood pressure. The methylation of arsenic results in the production of homocysteine (19, 20). Higher levels of plasma homocysteine have been associated with high blood pressure (21, 22). Because the metabolism of homocysteine requires sufficient levels of vitamins B₂, B₁₂, and B₆ and folic acid (23), arsenic exposure may affect blood pressure through its effect on the formation of homocysteine, especially in a background of inadequate intake of folate and B vitamins. In this study, we evaluated the association between arsenic exposure and blood pressure in a population with a wide range of low-level arsenic exposure from drinking water. We also examined the potential modification by dietary intakes of folate and B vitamins.

MATERIALS AND METHODS

Study population

The parent study (Health Effects of Arsenic Longitudinal Study) is an ongoing, prospective cohort study in Araihazar, Bangladesh. Details of the study methodologies have been presented elsewhere (24, 25). Briefly, prior to recruitment of subjects, water samples and geographic positional system data were collected for a set of 5,966 contiguous wells in a well-defined geographic area of 25 km² in Araihazar. Well owners were interviewed to create a roster list of 65,876 regular well users and residents in the area. Eligibility criteria for the cohort study included the following: being married (in order to increase stability of residence), being aged 18 years or more, and having had resided in the study area for 5 or more years (24). Between October 22, 2000, and May 19, 2002, 14,828 potential participants meeting the eligibility criteria were identified from the roster list. Nineteen percent of those eligible (n = 2,778) were not at home during study visits. Of the 12,050 who were available and approached, 11,746 (97.5 percent response rate) participated (24). Verbal consent was obtained from all the participants. The study procedures were approved by the Columbia University Institutional Review Board and the Ethical Committee of the Bangladesh Medical Research Council.

Information on patterns and history of well use, demographics, lifestyle characteristics, and dietary intakes was collected in an extensive baseline interview. Trained physicians completed a comprehensive physical examination that included blood pressure measurements. Both interviewers and physicians were blind to the well arsenic concentration until all the interviews and physical examinations were completed.

Measurements of nutritional intakes

Dietary intakes were measured at baseline by use of a semiquantitative food frequency questionnaire that was designed for the study population. Details of the food frequency questionnaire and a validation study of the food frequency questionnaire have been described elsewhere (26). In the validation study of the food frequency questionnaire, correlations between average daily consumptions measured by the food frequency questionnaire and those measured by two 7-day food diaries indicated that the validity of the food frequency questionnaire in measuring long-term intakes of the nutrients of interest in the present study was moderately good. Correlations for vitamins B₂, B₆, and B₁₂ and folate were 0.37, 0.39, 0.57, and 0.30, respectively (26). We used both the US Department of Agriculture Nutrient Database for Standard Reference (abbreviated version) (27) and the Indian food nutrient database (28) to convert food item intakes to nutrient intake values. All intakes were adjusted for total energy intake using the residual method (29).

Blood pressure measurements

Blood pressure was measured by trained clinicians using an automatic sphygmomanometer (HEM 712-C; Omron Healthcare GmbH, Hamburg, Germany). This model has been validated to have 85 percent of readings falling within 5–10 mmHg of the mercury standard (30). We used an automated sphygmomanometer because it is portable, easy to use, and less prone to observer bias. These strengths are especially important because our fieldwork was conducted in a rural area. Measurements were taken with participants in a seated position after 5 minutes of rest, with the cuff around the upper left arm, in accordance with recommended guidelines.

To reduce false positive measurements of high blood pressure due to the transient stress response to medical professionals (“white coat hypertension”), two additional measurements were taken after 2–3 minutes of rest for respondents found to have a SBP of ≥140 mmHg and/or a DBP of ≥90 mmHg at the first measurement, and the measurement with the lowest blood pressure was recorded.

A reliability study of blood pressure measurements was conducted in 61 subjects with three measurements taken regardless of blood pressure levels at the first measurement. Intraclass correlation coefficients were computed for SBP and DBP separately to evaluate the correlations among the three measurements. The reliability of blood pressure measurement was good, with all intraclass correlation coefficients between 0.92 and 0.94. The absolute differences among measurements were 0.9–2.0 and 0.7–2.8 mmHg for SBP and DBP, respectively.

Information on medicine use was extracted from the questionnaires. Study participants were asked to show all medicines they were taking regularly, and the interviewers recorded generic names. A total of 110 participants were taking antihypertension medicines at the time of baseline interview.

Arsenic exposure measurements

Water samples from the 5,966 wells were collected in 50-ml acid-washed tubes after pumping the well for 5 minutes. The arsenic concentration was measured with graphite furnace atomic-absorption spectrometry with a Hitachi Z-8200 system (Hitachi, Tokyo, Japan) (31). Samples that fell below the detection limit (5 μg/liter) were subsequently analyzed by inductively coupled plasma mass spectrometry, with a detection limit of 0.1 μg/liter. Consistent with similar analysis conducted in other parts of the world (32), analysis for time-series samples from 20 tube wells in the study area showed that well arsenic concentrations were stable over 3 years (33). In addition to information on current well use, participants were also asked about the exact location and duration of use of the previous well. Often the participant led the interviewer to the previous well. Questions were asked of husbands and wives separately, and if the history in a married couple was not the same, participants were asked to give a reason. We derived a time-weighted well arsenic concentration (TWA) as a function of drinking durations and well arsenic concentrations (TWA (μg/liter) = ∑ C_iT_i / ∑ T_i, where C_i and T_i denote the well arsenic concentration and drinking duration for the ith well) (34). The average time for which the arsenic concentration was known was 10.0 years for men and 8.3 years for women, accounting for 25 percent of participants' lifetime, on average, for both sexes.

Statistical methods

We first conducted descriptive analysis to assess differences in blood pressure by levels of demographics and arsenic exposure measures. To account for the correlation of arsenic exposure among participants who consumed water from the same well, we performed linear and logistic regression modeling of clustered data using generalized estimating equations (35). Robust variance estimates were used to calculate standard errors and confidence intervals.

Previous studies suggested a dose-response relation between hypertension and high levels of arsenic exposure. We first evaluated the associations of arsenic exposure with SBP, DBP, mean arterial pressure (⅓ × (SBP + 2 × DBP)), and pulse pressure (SBP − DBP) in linear regression models. Since results did not suggest any linearity of the associations, adjusted odds ratios for general hypertension (SBP, ≥140 mmHg, and/or DBP, ≥90 mmHg (14, 15)), systolic hypertension (SBP, ≥140 mmHg), diastolic hypertension (DBP, ≥90 mmHg), and high pulse pressure (SBP − DBP, ≥55 mmHg) in relation to quintiles of TWA were estimated by use of logistic regression models. High pulse pressure was defined as pulse pressure of ≥55 mmHg, the upper 10th percentile of the population distribution. A pulse pressure of ≥55 mmHg has also been studied as a predictor of cardiovascular diseases (36). Our previous analysis suggested that age, sex, body mass index, tobacco smoking status, water consumption, and markers of socioeconomic status such as educational level may modify arsenic toxicity (34, 37). Theses factors, except for water consumption, were also strong predictors of blood pressure. The final models included all the potential confounders in theory.

Potential effect modification by dietary intakes of vitamins B₂, B₆, and B₁₂ and folate was examined by stratified analysis. Total energy intake was additionally entered in the model as suggested by Willett and Stampfer (29). In addition, a score of 0 or 1 was assigned to participants with less than or greater than or equal to the median intake level of each nutrient, respectively. A composite measure was then created by summing the four individual scores. To formally test whether the association between arsenic exposure and blood pressure differs by levels of nutritional factors, we included cross-product terms representing products of nutrient intake levels and arsenic exposure expressed as a continuous variable in multivariate logistic models.

Among the overall 11,746 participants in the parent cohort study, a total of 11,458 participants with blood pressure measurements were included in descriptive analysis. Participants with available blood pressure measurements did not differ appreciably from those without a blood pressure measurement with respect to demographic and lifestyle attributes (data not shown). Multivariate analysis excluded participants who were taking antihypertension medicines at the time of interview (n = 110) and included 10,910 participants with information on TWA and all the covariate variables. Those with missing intake values on any of the food items in the food frequency questionnaire (n = 296) were further excluded from the stratified analysis by nutritional factors. All analyses were also performed in a subpopulation of 7,960 with a known well arsenic level for ≥5 years (mean, 11.2 years). Sensitivity analysis was also conducted with ≥3 and ≥7 years to define the subpopulation. Results were similar and therefore were not shown. All analyses were conducted using SAS, version 8.0, statistical software (SAS Institute, Inc., Cary, North Carolina).

RESULTS

All of the blood pressure parameters were positively related to age. SBP, DBP, and the mean arterial pressure were positively associated with body mass index and educational attainment, while pulse pressure varied little by these factors (table 1). Compared with never smokers, current smokers and past smokers had lower blood pressure and higher blood pressure, respectively. This finding is in line with findings reported in the literature that cross-sectional studies often observe a transient effect of smoking on blood pressure (38).

TWA was positively related to pulse pressure and SBP, and the associations were stronger in the subpopulation with the exposure level known for an average of 11.2 years. However, we did not observe any linearity of the association. In linear regression models, the increases in pulse pressure by increasing TWA quintiles were 0.00 (referent), 1.93 (95 percent confidence interval (CI): 1.18, 2.68), 1.15 (95 percent CI: 0.39, 1.90), 0.60 (95 percent CI: −0.14, 1.34), and 1.24 (95 percent CI: 0.52, 1.95) mmHg in the overall population and 0.00 (referent), 2.31 (95 percent CI: 1.44, 3.18), 1.57 (95 percent CI: 0.68, 2.46), 1.03 (95 percent CI: 0.16, 1.89), and 1.52 (95 percent CI: 0.69, 2.34) mmHg in the subpopulation. The increases in SBP by increasing TWA quintiles were 0.00 (referent), 1.81 (95 percent CI: 0.70, 2.92), 1.25 (95 percent CI: 0.13, 2.37), 0.73 (95 percent CI: −0.36, 1.81), and 1.81 (95 percent CI: −0.26, 1.87) mmHg in the overall population and 0.00 (referent), 2.03 (95 percent CI: 0.73, 3.33), 1.97 (95 percent CI: 0.64, 3.31), 0.91 (95 percent CI: −0.37, 2.19), and 1.19 (95 percent CI: −0.06, 2.43) mmHg in the subpopulation. The associations of TWA with DBP and mean arterial pressure were not apparent (data not shown).

In logistic regression models, the presence of systolic hypertension and high pulse pressure was positively related to TWA (table 2). However, the positive associations did not follow a strict dose-response relation. The higher four quintiles of arsenic exposure were associated with 19–39 percent and 24–50 percent increases in risk of high pulse pressure among the overall population and the subpopulation, respectively. There were no clear associations between TWA and either diastolic or general hypertension.

The positive associations of TWA with the presence of high pulse pressure and systolic hypertension were somewhat more apparent in participants with low levels of vitamins B₂, B₆, and B₁₂ or folate intake compared with those with higher levels (table 3). The pattern of the odds ratios was consistent in the subpopulation. However, interaction by levels of individual nutrient intake was not statistically significant at p < 0.05. No apparent associations were observed between diastolic hypertension and TWA among participants with any given intake level of single nutrients (data not shown).

Average daily energy-adjusted intakes of vitamins B₂, B₆, and B₁₂ and folate by the composite measure of the four nutrients are presented in table 4. The higher the score on the composite measure, the higher the intake values were for each of the nutrients.

Among participants with all four nutrient intakes lower than the average level (0 score on the composite variable), the odds ratios for high pulse pressure comparing higher quintiles with the bottom quintile of TWA ranged from 1.83 to 1.89, with all of the 95 percent confidence intervals excluding unity (table 5). The associations of higher levels of TWA with systolic hypertension also appear to be stronger, albeit weaker compared with those with high pulse pressure, in participants with lower intake levels of folate and the B vitamins. These results are consistent in the subpopulation. Odds ratios for general hypertension in participants with lower than average intake of B vitamins and folate were greater than those in participants with higher intake levels. However, the associations were not significant; the odds ratios for general hypertension by increasing TWA quintiles were 1.00 (referent), 1.05 (95 percent CI: 0.66, 1.68), 1.24 (95 percent CI: 0.80, 1.94), 1.38 (95 percent CI: 0.88, 2.14), and 1.28 (95 percent CI: 0.76, 1.83), respectively, with lower intake levels of folate and the B vitamins.

DISCUSSION

We found that arsenic exposure was positively associated with systolic hypertension and high pulse pressure, and the associations were more pronounced among participants with lower intake levels of folate and the B vitamins. No apparent association was observed between arsenic exposure and general hypertension.

High systolic blood pressure and a widening pulse pressure have been indicated as a consequence of arteriosclerosis and arterial stiffness (39, 40). It has been found that persons with a high pulse pressure had greater left ventricular mass and greater carotid intimal-medial thickness compared with those with lower pulse pressure, despite a similar mean diastolic blood pressure (41, 42). A high pulse pressure and a high systolic blood pressure have been indicated as independent predictors of cardiovascular disease-related mortality (43–45).

The exact mechanisms by which arsenic induces high systolic blood pressure and pulse pressure are not clear. Arsenic may promote inflammatory activity. Low concentrations of arsenite (As³⁺) have been associated with increased superoxide accumulation in porcine aortic endothelial cells (46, 47). In addition, arsenic may impair formation of endothelial nitric oxide (48, 49). Levels of endothelial nitric oxide play an important role in maintaining vascular tone (50).

Three other cross-sectional studies have investigated the association between arsenic exposure and general hypertension. Chen et al. (3) found a dose-response relation between arsenic exposure and hypertension in southwestern Taiwan, where the median well arsenic concentration in villages ranged from 700 to 930 μg/liter (51). Rahman et al. (4) observed a positive association of hypertension with arsenic exposure in Bangladesh; the range of exposure categories was, however, very wide (0, <500, 500–1,000, >1,000 μg/liter). Recently, a cross-sectional study in Wisconsin found that respondents with well-water arsenic concentrations greater than 10 μg/liter were more likely to report having had high blood pressure than were respondents whose well water had arsenic concentrations less than 2 μg/liter (52). However, analysis was based on self-reported health data.

A systematic review of the literature on the association between arsenic exposure and cardiovascular disease concluded that the evidence from southwestern Taiwan is consistent with a role of high arsenic exposure in atherosclerosis and that the cardiovascular effects of chronic low-level arsenic exposure are unknown (9). Importantly, several studies have assessed the associations between risks of cardiovascular diseases and low-level arsenic exposure. An ecologic study found elevated standardized motorality ratios for diseases of the arteries, arterioles, and capillaries in US counties with a water arsenic concentration greater than 20 μg/liter (53). The mortality of hypertensive heart disease was elevated in members of the Mormons in Millard County, Utah, with <200 μg/liter of arsenic in drinking water compared with that of the general population in Utah (54). No dose-response relation was observed in the two above-mentioned studies.

A dose-response relation may be absent in the range of arsenic exposure in the study population (0–864 μg/liter; mean and median were 101 and 62 μg/liter, respectively). We note that mechanistic studies have demonstrated that the vascular effect of arsenic is nonlinear (55). Alternatively, nondifferential measurement errors in blood pressure and arsenic exposure may have obscured a possible underlying dose-response association. Casual blood pressure readings are not an optimal representation of the entire 24-hour blood pressure pattern. However, we did not detect any systematic measurement errors in blood pressure measurements, and the reliability of blood pressure measurements between consecutive measurements was good. In addition, the relations of blood pressure and conventional risk factors were consistent with those of the literature, further suggesting the validity of the blood pressure measurement in this study.

Many studies used ecologic measures of arsenic exposure or unreliable outcome measures (7, 51–54). On the other hand, arsenic exposure was assessed individually, and blood pressure was measured with a standardized protocol in the present study with a large sample size. The study was population based, and therefore the selection of participants was not dependent on either arsenic exposure or blood pressure status. The validity of self-reported well-use history was good, since the correlation between the arsenic concentration in the current well and urinary arsenic was 0.70 (24). Additional statistical adjustments for factors related to blood pressure, such as contraceptive use in women and occupation, did not change the odds ratios appreciably. However, the possibility of uncontrolled confounding cannot be entirely excluded. We were not able to ascertain a complete lifetime history of well use with which to construct the exposure index. However, the odds ratios for systolic hypertension and high pulse pressure in relation to arsenic exposure were similar, if not greater, in the subpopulation with 11.2 years of known well arsenic on average, suggesting a long-term effect of arsenic exposure.

We found stronger associations of arsenic exposure with systolic hypertension and high pulse pressure among participants with a low intake level of folate and the B vitamins (table 5). The literature has suggested a role of these nutrients and homocysteine in the etiology of vascular diseases (56–58). Methylation of arsenic, a hypothesized detoxification pathway, requires the conversion of S-adenosylmethionine to S-adenosylhomocysteine, which results in homocysteine (19, 20). Homocysteine can be remethylated to methionine by a folate-dependent reaction that is catalyzed by methionine synthase, a vitamin B₁₂-dependent enzyme. Alternately, homocysteine may be metabolized to cysteine in reactions catalyzed by two vitamin B₆-dependent enzymes (23). Our finding is consistent with the hypothesis that individuals with insufficient intakes of nutrients related to both endothelial function and arsenic metabolism are more susceptible to the vascular effect of arsenic exposure.

Dietary intakes were measured by food frequency questionnaire, and therefore measurement errors are expected. Nondifferential measurement errors of nutrient intake (effect modifiers) may distort their modifying effect of the exposure-disease association (59). The absence of both alcohol drinking (due to religious beliefs) and a nutritional fortification program in the population strengthened the validity of dietary intakes measured by food frequency questionnaire. Although there were differences in the observed modifying effects of single nutrients, the heterogeneity was not apparent and may reflect differences in validity of the food frequency questionnaire in measuring different nutrients. We created a composite variable to capture the combined effects of all four nutrients. This method assumes equal contributions of the nutrients. Additional studies are needed to examine the role of each nutrient. In this regard, a recent study from our group showed that a large proportion of people in the study area had low levels of serum folate and vitamin B₁₂ (60).

In conclusion, we found that arsenic exposure from drinking water, even at lower levels (<50 μg/liter and <100 μg/liter), was positively associated with high pulse pressure, and that the association was more apparent among those with lower intake levels of vitamins B₂, B₆, and B₁₂ and folate. Future studies are needed to investigate the effect of low-level arsenic exposure from drinking water on specific atherosclerosis outcomes and cardiovascular diseases.

Abbreviations

Abbreviations
 
• CI

  confidence interval

 
• DBP

  diastolic blood pressure

 
• SBP

  systolic blood pressure

 
• TWA

  time-weighted well arsenic concentration

The authors acknowledge support from grants P42ES10349, P30ES09089, R01CA107431, R01CA102484, and ES000260.

The authors would like to thank colleagues Dr. Lydia Zablotska and Dr. John Barron for their helpful comments.

Conflict of interest: none declared.

References