Introduction

Raised blood pressure is a major risk factor for coronary heart disease and the leading risk factor for stroke. It is responsible for an estimated 7.5 million deaths annually, 12.8% of all deaths worldwide,¹ and is the leading risk factor for global disease burden (7.0% of global DALYs [Disability Adjusted Life Years]).² The increasing prevalence of hypertension reflects both ageing populations³ and underlying risk factors, such as unhealthy diet (including high sodium intake), lack of physical activity, obesity, stress, and excessive alcohol intake.^4,5 In particular, there is persuasive and concordant evidence from epidemiological studies, clinical trials, and animal studies that sodium intake (mainly through food) plays an important role in the global burden of raised blood pressure.^6–10

In 2008, maternal health authorities in Dacope, a rural subdistrict of Bangladesh (Figure 1), reported an unusual increase in the incidence of hypertensive events (such as gestational hypertension and preeclampsia) in pregnant women.¹¹ Dacope is a relatively poor subdistrict, and literacy is estimated to be <40%. Most people work in agriculture and cultivate rice. Smoking rates and alcohol intake are extremely low, and physical activity is high—particularly during harvest times.¹² Extreme weather events such as storm surges and tides have led to contamination of fresh drinking water sources with sea water. High sodium levels were found in various drinking water sources (mainly tube wells and ponds) with seasonal fluctuations.¹³ Therefore, the increased hypertension problems were hypothesized to be linked to consumption of saline drinking water rather than to the well known risk factors listed above. Salinization of drinking water is believed to be associated with climate change^14,15; however, changes in river flow from an upstream barrage, faulty management of polders, shrimp farming, and ground water extraction may all contribute further to this process.¹⁶ A recent study in the area reported a dose–response relationship between water sodium concentrations and risk of preeclampsia, a complication of pregnancy characterized by hypertension.¹⁷

• Download figure
• Download PowerPoint

Apart from preeclampsia, there are several other adverse outcomes associated with increased blood pressure in pregnancy; therefore, we investigated—in the present study—blood pressures of 701 nonhypertensive pregnant women residing in Dacope with respect to their sources of drinking water during pregnancy. We hypothesize that drinking water sodium concentrations are positively associated with systolic and diastolic blood pressure in these women. If we were to find evidence for an association between drinking water sodium and blood pressure in this cross-sectional study, this would contribute to the arguments in favor of prioritizing low-saline drinking water alternatives—particularly for pregnant women—to avoid preventable hypertension-related burden of disease in salinity affected coastal areas.

We specifically looked at healthy pregnant women and did not include women with preeclampsia in the analysis because preeclampsia could be both a consequence and a cause of increased blood pressure. This makes it difficult to establish whether high blood pressure leads to preeclampsia in these women or that preeclampsia was caused by other risk factors and induced high blood pressure over the course of the illness.

Methods

The study area is situated in the subdistrict of Dacope, part of the Khulna district in Bangladesh’s southwest coastal region (Figure 1). Dacope is divided into 9 administrative unions, with a total population of ≈158 000 people. The study population comprised 1020 nonhypertensive controls from a previously conducted case–control study of preeclampsia in Dacope. Women recruited in the wet season (n=310) and women with known complications in their pregnancy (n=9) were excluded, leaving 701 women. The dry season was chosen because it is associated with the highest contrasts in drinking water salinization¹³ and minimizes potential confounding by season (mainly related to temperature and [agriculture related] physical activity). In the case–control study, all hypertensive pregnant women in the 20th week of their pregnancy and residing in Dacope were eligible during a period of 18 months (January 2009 to June 2010). For each hypertensive woman, 5 healthy pregnant controls were selected from the same village and also recruited in their 20th gestational week. Blood pressure, weight, height, and mid-upper arm circumference (a proxy for nutritional status¹⁸) were measured by trained data collectors. Blood pressure was measured in the left arm (resting, with palm up) using a manual sphygmomanometer. Each of the 9 data collectors was responsible for one union in the subdistrict. All participants were interviewed using a structured questionnaire to obtain data on age, parity, socioeconomic status, seasonal changes of water source, underlying illnesses, physical activity, smoking and environmental exposures such as insecticides, and use of various types of local drugs and medicines. All participants were asked to collect a 24-hour urine sample: 92% of the participants (n=645) agreed to participate. They were given a container and given instructions to start the collection the next morning, after the first urine excretion. Urinary sodium was measured by Ion Selective Electrode method. All data collection methods are described in more detail elsewhere.¹⁷

Average sodium concentrations per water source were based on 407 available water samples. These samples were collected by the same research group in the dry season of 2010 and 2011. All samples were analyzed at the University of Dhaka, where sodium was measured by the Atomic Absorption method (AAnalyst 800, Perkin Elmer).¹⁷ The analysis showed that tube wells had the highest sodium concentrations, with a mean concentration exceeding 700 mg/L, followed by pond water with an average concentration of ≈400 mg/L. Harvested rainwater was (as expected) lowest in sodium content.

Participants’ water source was categorized as rainwater, pond water, tube well, or a combination of sources. Cooking water sources were categorized in the same way with the addition of river water. Information on type of water source drunk in the 2 weeks before the interview was available for all women and was used as proxy for drinking water sodium (high, medium, and low).

Data on average and maximum daily temperatures and accumulated rainfall (if any) on the days of data collection were abstracted from the Bangladesh Meteorologic Department.

Statistical Analysis

We used multiple linear regression models to investigate association of water source with systolic and diastolic blood pressure. Where possible, missing data for all variables—apart from weather parameters—were imputed using Multivariate Imputation by Chained Equations techniques.¹⁹ Moving average techniques were used to impute missing values in daily maximum temperature because temperature tables showed smooth increases and decreases over time and season. Models were adjusted for potential confounders in 3 phases: first adjustment for age, followed by adjustment for age and well-established risk factors (nutritional status, physical activity, parity), and finally also for local social, behavioral, and environmental factors. The models were not adjusted for gestational age because all women were in their 20th gestational week when recruited and examined. Stepwise methods were used for building the models. Log likelihood, Akaike’s Information Criterion, and Bayesian Information Criterion were used for model selection. In addition, the union of residence of the participants was included as a random effects parameter in the mixed models. The random effects were assumed to be normally distributed on the log odds scale. A likelihood ratio test was used to test whether the mixed model fit the data better than the linear regression models. Random effects per union were calculated using the Best Linear Unbiased Predictor method.²⁰ Sensitivity analysis was performed excluding unions with significantly higher or lower blood pressure values than average.

Results

Characteristics of the study population are shown in Table S1 in the online-only Data Supplement. Approximately a third of the participants drank exclusively tube well water (N=241, 34%); another 33% drank pond water (N=229), and 22% reported drinking a combination of those 2 (N=153). Rain water accounted for 11% of drinking water sources (78 participating women). None of the respondents used river water as drinking water source, but it was used by 25 women (4%) as cooking water source. Women using pond and tube well water as drinking water source had similar baseline characteristics as rain water users, but were more likely to be anemic and were more often users of the local stimulant gull (smokeless tobacco). The daily temperatures on days that tube well and pond users were interviewed were on average somewhat cooler than the days that rainwater users were interviewed (−1.4°C and −4.1°C, respectively). Figure 2 shows drinking water sodium concentration by water source. Median sodium concentration was 31 mg/L for rainwater, 208 mg/L for pond water, and 704 mg/L for tube well water. The median concentration in water ingested by those reported to drink a combination of sources was difficult to establish because the consumption of each source would vary greatly per person and over time. The concentrations per source were reflected in 24-hour urinary sodium (Figure S1).

• Download figure
• Download PowerPoint

Figure 3 shows median systolic and diastolic blood pressure by water source. Median blood pressure values were higher with increasing sodium concentration by water source: 100/65, 105/70, and 110/70 mm Hg for rain, pond, and tube well water users, respectively. Users of different water sources were similar in terms of physical activity (Table S1) and diet, and therefore, differences in these potential confounders are unlikely to explain our findings.

• Download figure
• Download PowerPoint

Mixed models (Table) showed a better fit to the data than linear regression models, though results were similar. Drinking tube well water, pond water, and combined sources were significantly associated with higher systolic pressure (+3.62 mm Hg [confidence interval: 1.20, 6.04] and +4.85 mm Hg [2.55, 7.25] for pond and tube well users, respectively) compared with drinking rain water after adjustment for known and context-specific confounders. Results of the mixed linear regression models for diastolic blood also showed higher values for pond and tube well users as compared with rainwater users (+1.72 mm Hg [−0.36, 3.80] and +2.30 mm Hg [0.33, 4.23] for pond and tube well users, respectively; Table). The uncertainty in the diastolic blood pressure results is possibly (partly) a reflection more variability in the measurement of diastolic compared with systolic blood pressure. As expected, blood pressure was in all models associated with the known risk factors age, nutritional status, physical activity, and parity (Table S1). Furthermore, daily temperature was significantly and inversely associated with blood pressure. Sensitivity analysis excluding participants from Loudobe, Koilashgonj, or Kamarkhola—where blood pressures were significantly above or below average—did not materially alter the results in any of the models.

Although the sodium levels in the examined water samples are exceptionally high, concentrations of other water contaminants, such as arsenic or other chemicals, were found to be relatively low in the area: arsenic was absent in surface water sources such as ponds, and there were relatively low concentrations in shallow surface water sources.²¹

Discussion

Our study adds to previously reported evidence that highly saline drinking water is associated with higher blood pressure levels.^13,22–26 Although some studies failed to show this association (eg,Robertson et al²⁷ and Welty et al²⁸), sodium concentrations evaluated in this study by far exceeded the sodium levels evaluated in those studies. This is of particular interest because blood pressure is directly related to adverse pregnancy outcomes and may increase the chance of low-for-gestational-age births.^29,30 Although in monsoon season women can often rely on rainwater, they will have to switch to a higher saline source if their pregnancy also covers (part of) the dry season. The increases in blood pressure that can be caused by this change in source could significantly increase the risk of adverse pregnancy outcomes especially because such increase would push the blood pressure of the women into the (pre)hypertensive range.

Concerning guidelines for drinking water quality, the WHO states that: “No health-based guideline value has been derived […], as the contribution from drinking-water to daily intake is small.”³¹ Mention is made of a taste threshold of 200 mg sodium/L and a recommendation of 20 mg/L; in Dacope, most pond and tube well users are drinking sodium levels well beyond this identified taste threshold. Based on the findings of the present study and the study by Khan et al,¹⁷ drinking water is an important source of ingested sodium in Dacope, and this may be the case also in other coastal areas with drinking water salinity problems around the world. Drinking water sodium should in these situations be considered an important potential source of sodium intake and a risk factor for hypertension and related disorders. Developing international guidelines for maximum sodium content of drinking water would help to prioritize this potential health hazard and facilitate improved drinking water sources in salinity prone (and often poor) areas, such as Dacope.

The drinking water salinity problems in coastal Bangladesh, and similar coastal areas worldwide, reflect high vulnerability to extreme weather events and long-term climatic changes.¹⁵ The results of the current study may already (partly) be interpreted in the context of extreme weather events: in May 2009, cyclone Aila hit Bangladesh and inundated part of the subdistrict of Dacope.³² Although our analysis was restricted to the dry season only (from November 2009 onward), the cyclone and resultant flooding may have increased the salinity levels in ponds and tube wells, which was possibly still evident during the dry season measurement campaign in the present study.

In its 5th assessment report, the Intergovernmental Panel on Climate Change stressed strong evidence for significant sea level rise and an increase in the occurrence of sea level extremes.³³ Climate change–induced extreme precipitation increases the frequency that river embankments will be overtopped³⁴ and, hence, the risk that ponds and shallow wells within the embankment boundaries will be flooded by saline river water.³⁵ Together with the predicted rise in sea levels, this will lead to contamination of the drinking water sources for the population. It is expected that not only the frequency and intensity of the salinity contamination will increase, but because of the moving tide line, a larger area of the delta will be affected in the future.³⁵ Therefore, drinking water salinity problems in areas like Dacope are expected to increase in the near future.¹⁵ These trends will have a high impact on drinking water quality of vulnerable coastal areas, such as Dacope.

Strengths of the study include its large size, stringent measurement protocols, and detailed attention to potential confounders. Observer bias is unlikely because water sodium concentrations of sources for each participant were unknown at the time blood pressure data were collected. The study was done in a relatively homogenous group (pregnant women) with high levels of physical activity and low rates of obesity; any differences in these variables and food intake patterns (possible confounders of the water–blood pressure association) should have been balanced after adjustment for socioeconomic status. Our study also had some limitations. First, information on water source rather than actual drinking water sodium concentration was used. Although there may be considerable variability in sodium concentrations within the same water source type, tube wells and ponds were always found to be significantly more saline than rainwater,¹¹ and similar differences were observed in urinary sodium excretion within each of the drinking water source groups (Figure S1). Dietary sodium was not measured in this study. However, an estimated dietary sodium intake of 2.23 g sodium per day (including seasoning habits) was obtained from an ongoing study and was relatively homogeneous across socioeconomic class and village (“Drinking water salinity and dietary sodium,” P.F.D. Scheelbeek, unpublished data, 2015).

Because this study is conducted in the same area, it is unlikely that dietary sodium intake varies substantially across water source: sodium intake through water could therefore be considered as important source of sodium intake in this population. Also potassium concentrations in water (and food) were not measured. Stepwise regression methods were used in the analysis phase: although this is used frequently in epidemiological studies, more recently stepwise methods have been challenged by some research groups, arguing that it should not be used for causal analysis.³⁶ As in all epidemiological studies, the study might be subject to some residual confounding that was not adjusted for. However, arsenic concentrations—which play an important role in drinking water quality elsewhere in Bangladesh—are low in the areas investigated in this study. In a recent study, 2 ponds in the same area have been monitored extensively during a period of 18 months. No high or unusual concentrations of chemicals, metal, and minerals (including arsenic), pathogens, or other (unexpected) contaminants other than salt were measured (P.F.D. Scheelbeek, unpublished data, 2015). Finally, this study was restricted to pregnant women only. Because of natural blood pressure changes in the course of the pregnancy,³⁷ the effect of water sodium on blood pressure in the general population may be different. Future work might focus on blood pressures in the general population to obtain a better understanding of the public health impact of drinking water salinity in coastal areas, such as Dacope.

Hypertensive women were excluded from the existing database for the analyses in study, despite the fact that they are potentially an important group to consider when studying the association between drinking water sodium and blood pressure. Excluding these women may have introduced a small bias in the study and—most probably—lead to a slight underestimation of the estimated effect size. The exclusion of these women was done, however, because the vast majority of women in the higher blood pressure ranges were also preeclamptic; including preeclamptic women in the analysis could have introduced reverse causality problems (see introduction). Nonetheless, despite removing this important group, a positive association between drinking water source and blood pressure was found, even in the group of women with blood pressures in the normotensive range only: the effect size of this association in women with higher blood pressure values would be expected to be larger and more significant.

The design of the study allowed only one measurement during the 20th gestational week: different effects of saline water on blood pressure may be found in earlier or later stages of the pregnancy.

Perspectives

Our study found evidence that drinking water with a high sodium concentration from saline contamination is associated with higher blood pressures in normotensive pregnant women in coastal areas of Bangladesh. These changes could be there at the start of the pregnancy—because sodium concentrations of women may differ even before they conceive or could be induced during pregnancy when a woman is forced to change drinking water source from a rainwater-based source to a high-saline alternative, because of depletion of rainwater storage. The likely rise in salinity will put added pressure on safe drinking water sources for millions of people living in vulnerable coastal areas around the world. Recognition of the health risks associated with water salinity by national and international bodies, as well as setting guideline values for water sodium concentrations, should encourage governments to monitor drinking water quality in coastal areas, explore adaptation strategies, and accelerate safe water provision where sodium concentrations exceed guideline values.

Acknowledgments

We thank Dr Adrian Butler and Dr Mohammad Hoque for their input in the manuscript; Dr Marc Chadeau-Hyam and Gianluca Campanella for their valuable comments on the data analysis; and all data collectors for the excellent work in Dacope.

Footnotes