Chronic exposure to inorganic arsenic through contaminated drinking water is a global public health problem, causing a large number of adverse health effects including skin lesions, hypertension, and various forms of cancer.1,2 Epidemiologic studies show associations between arsenic in drinking water and cancers in skin, lung, kidney, bladder, and liver.3–8 Increased risks for cardiovascular and peripheral vascular diseases, as well as several other serious health consequences, have also been associated with ingestion of arsenic through drinking water.9–11 Due to this multitude of morbidity and mortality consequences, overall mortality is likely to be increased in arsenic-affected areas.12 In Northern Chile, 10% of all deaths for men and 5% of all deaths for women, were attributed to arsenic in drinking water in an area where the water supply during a certain time period contained more than 800 μg/L.13 For the purpose of risk assessment, there is a need for better characterization of dose-response relationships, particularly at low concentrations, as well as information on susceptibility factors such as age, sex, and malnutrition.5,14–16
This large cohort study aims to determine whether the chronic arsenic exposure in rural Bangladesh, which has been present for several decades, is associated with increased mortality. Bangladesh, with more than 150 million people, has suffered from water-born diseases, related to the use of polluted surface water.17 In an effort to improve the safety of drinking water, shallow tubewells were installed all over Bangladesh starting in the 1970s. It has recently been estimated that these wells have exposed about half of the population in Bangladesh to arsenic concentrations above 10 μg/L (the WHO guideline value18), with wide variations in exposure.19 Although a large part of the Bangladeshi population has been exposed for long-enough times to cause arsenic-related deaths,20 no study has yet shown such mortality effects.
The study was carried out in Matlab, a rural area located 53 km South-East of Dhaka, Bangladesh. The International Centre for Diarrheal Disease Research, Bangladesh (ICDDR,B) has maintained a Health and Demographic Surveillance System21 in Matlab since 1966—recording demographic and selected health information through home visits, covering 142 villages with a total population of about 220,000. Data bases on birth, migration, marital status, pregnancy outcome, and death are updated based on information collected by community health research workers, who visit all homes on a monthly basis. Each person's place of residence is continuously updated, allowing them to be linked to the household's water sources over time. Socioeconomic surveys were conducted in 1974, 1982, and 1996, providing detailed socioeconomic information, including individual educational level and household asset scores.
Identification and Definition of the Cohort
The cohort comprised all inhabitants in the defined study area aged 15 years or more on 1 January 1991. We followed them until 31 December 2000. This was a closed cohort; due to lack of exposure information, we did not consider any in-migrating persons, nor did we include those who became 15 years of age during this period.
Cases were defined as persons within the cohort who had died of nonaccidental causes during the period 1 January 1991 to 31 December 2000. In developing countries, causes of death are often based on information obtained from relatives or acquaintances of the diseased through systematic retrospective questioning, known as verbal autopsy22,23; this was the method used to record causes of death in this area.21 In a validation study using physicians' reports in Matlab, verbal autopsies were found to be highly specific (>95%), with 85% sensitivity for cancer deaths and 50%–85% sensitivity for cardiovascular deaths.24 Similar results have been observed in India and China.25 There is no reason to believe that misclassification of cause of death would be associated with arsenic exposure levels.
Accidental deaths (injury, burns, drowning, snake bite, homicide, and suicide) were excluded, assuming no association with arsenic exposure. We also excluded anyone missing information on arsenic exposure or socioeconomic status. Follow-up time in person-years was calculated from 1 January 1991 to date of death, out-migration, or end of study period (Fig. 1).
A cross-sectional survey of the history of water sources, the presence of arsenic-related skin lesions, and arsenic concentration in water from all tubewells in the area, was performed from January 2002 to August 2003. Details of study methodology and results are given elsewhere.26 In brief, the survey included all persons above 4 years of age who at the time of the study were living in Matlab at least 1 day a week. A total of 166,934 persons were interviewed regarding lifetime water consumption. A team of field workers visited the area and collected water samples from all functioning tubewells (n = 13,286). Total arsenic concentration was measured by hydride generation atomic absorption spectrophotometry (Shimudzu AA6800, Shimudzu Corporation, Kyoto, Japan) at the ICDDR,B laboratory.27
Details regarding the calculations of individual chronic arsenic exposure of the current population have been presented elsewhere.26 Each person's exposure was calculated for each year since 1970 or from the year of birth if born later. Arsenic concentrations in well water are reasonably stable over time,19,28,29 and so we used current arsenic concentration in tubewells water to construct lifetime exposure. Before 1970, there were very few tubewells in Bangladesh, and most people used surface water.2 In the mid-1970s only 30% of Matlab population used drinking water from tubewells30; this increased to 55% in the 1980s and 95% in the mid-1990s.21 Individual exposure data for deaths were not available because deaths had occurred before the 2002–2003 survey. Similarly, some people had migrated out before the survey. Therefore the study used average household exposure of arsenic from drinking water as a proxy for individual exposure. Average household exposure was calculated for each calendar year from 1970, based on information obtained from the current population present in that specific household for each year (n = 2,997,301 household-years). If data were missing for a specific household, the corresponding information was derived on the bari (compound) level (n = 144,926 compound-years). Exposure data for all persons included in the cohort follow-up were based on the average arsenic levels in the person's drinking water from 1970 (or birth, if born later) up to the end of 1990. The detailed procedure for calculation of exposure is described in the eAppendix (http://links.lww.com/EDE/A343).
To validate the use of average household exposure data as a proxy for individual exposure, 1000 persons were randomly selected from the survey database and their current exposure levels compared with the household exposure data. Mean water arsenic concentration was 108 μg/L both on the individual and household level, and individual and household data were highly correlated (r = 0.935). Similarly, when comparing individual exposure with average bari exposure, mean values were 108 and 109 μg/L, respectively, with a correlation coefficient of 0.792.
Mortality risks in Matlab vary with socioeconomic conditions,31 including education.32 Further, arsenic exposure has been shown to differ among socioeconomic groups, with varying patterns over time.26 In the 1970s and 1980s, higher socioeconomic groups had more opportunity to purchase tubewells. Similarly, in the late 1990s the higher socioeconomic groups were among the first to shift to arsenic-free drinking water. By using exposure information up to the early 1990s but not later, possible confounding by socioeconomic group has only one direction. Asset scores were used based on a household level socioeconomic census in 1996 that contained individual-level and household-level information.33 Each household asset with all available information was assigned a weight or factor. Scores were generated through principal component analysis, with scores categorized into 5 groups ranging from the lowest (poorest) to highest (richest).
We assessed the associations of age, sex, asset score, and education, with average household arsenic exposure. Sex, asset scores, and education were analyzed in relation to exposure and outcome to assess possible associations (P ≤ 0.10), using analysis of variance, χ2, or Spearman correlation coefficients as appropriate for the data. The mortality risks in relation to arsenic exposure were estimated by Cox proportional hazards models, adjusting for potential confounders, and adjusted hazard ratios (HRs) were calculated along with 95% confidence intervals (95% CIs). Potential confounding factors that changed the effect estimates by 5% or more were included in the model. We divided average lifetime exposure into 5 groups, with lowest average arsenic exposure level used as the reference. To evaluate any sex differentials in the association between arsenic exposure and mortality, we also performed stratified analysis by sex. All analyses were done separately for cancer, cardiovascular diseases, infections, and total nonaccidental adult mortality. SPSS 16.0 (SPSS Inc, Chicago, IL) software was used for data analysis. The study was approved by the ethical review committee of ICDDR,B, Dhaka, Bangladesh.
A total of 121,123 persons 15 years of age or more on 1 January 1991 were eligible for the study. After excluding those with missing exposure or socioeconomic information, 115,903 persons were included in the analysis. Of these, 9015 died and 22,488 were lost to follow-up within the study period (Fig. 1).
There were differences in the distributions of age, sex, and education levels between included and excluded persons (Table 1). The male/female ratio was higher for those included, mainly because more women than men were missing SES data. Less educated persons were also more likely to be excluded (Table 1).
Sex, household asset scores, and education differed between deaths and survivors, and these covariates were also associated with arsenic exposure. Among the deaths, there was an excess of men, persons with lower education, and persons with lower asset scores (Table 1). Average arsenic exposure increased with education and socioeconomic status, and declined with age.
Further, women had slightly higher average water arsenic concentrations than men (eTable 1, http://links.lww.com/EDE/A343). The mean level of exposure (μg/L) was 133 (SD = 118; median = 106) for those who died and 131 (SD = 116; median = 105) for survivors. Mean length of exposure was also about the same (12.9 [SD = 7.7] for deaths and 12.6 years [7.5] for survivors).
Infections gave rise to almost one-quarter of all deaths, whereas cardiovascular diseases constituted 13% and cancers 7% (eTable 2, http://links.lww.com/EDE/A343). There was an elevated risk of nonaccidental death with increasing arsenic exposure, even at arsenic concentrations in drinking water as low as 10–49 μg/L (Table 2). The risk of death due to cancer, cardiovascular diseases, and infections all increased with increasing exposure (Table 2, Fig. 2).
We did not observed any sex differences in mortality for cardiovascular diseases or infections. Men had slightly higher risk for arsenic-related death due to cancers, while women had slightly higher risk of death due to nonaccidental causes (eTable 3, http://links.lww.com/EDE/A343). Similar risk estimates were observed when we stratified the analyses by age group and socioeconomic status (data not shown).
This is the first cohort study from Bangladesh to show that chronic exposure to inorganic arsenic via drinking water from tubewells is associated with increased risk of death. We found higher mortality from cancer, cardiovascular diseases, and infections. Increased risk for nonaccidental death was noted even at the lowest dose (10–49 μg/L, which is below the Bangladesh water standard). Mortality from cancer, cardiovascular diseases, and infections all increased with higher arsenic exposure. The results suggest that a few decades of arsenic exposure via drinking water has produced substantial increases in adult mortality.
One of the strengths of our study is the large sample size and the strict cohort approach. Without a death registry the completeness of case ascertainment may be uncertain.34 In this study, mortality cases were prospectively identified within the carefully maintained surveillance system in Matlab, Bangladesh, where all deaths are recorded, and trained personnel classify cause of death based on systematic standardized interviews.21,23 Another strength of this study is that it is stemmed from clear a priori hypotheses that increased risk would be found for these causes of death, based on previous studies in Chile, Taiwan, Argentina, and the United States.1,35 The socioeconomic and demographic characteristics in Matlab are similar to those in other parts of Bangladesh, which implies that inferences from these results may be drawn to other parts of Bangladesh, where elevated arsenic concentrations in well water are a common phenomenon.19
One potential source of bias in the analyses is the use of household exposure data (and to some extent communal, or bari, exposure) as a proxy for individual exposure data. Because the arsenic contamination of well water can vary locally, this could potentially introduce a nondifferential bias. However, a comparison individual and household exposure data in a subsample of the population supports our strategy. Another uncertainty is whether the measurements of arsenic in tubewell water in the 2002–2003 survey provide valid exposure data for the period 1970–1990. Information on temporal variation in water arsenic concentration is limited. In one highly contaminated area, there was a fairly stable arsenic concentration during a 3-year period.28 Similarly, the British Geological Survey followed randomly selected wells all over Bangladesh for 1.5 years without observing changes.19 We followed about 60 tubewells 3 times per year over 3 years without finding major differences over time.29 Thus, we infer that current information on arsenic concentrations in tubewells may be a suitable proxy for previous years. Wealth was measured in 1996, which is the middle of our study period (1991–2000), giving us confidence that it accurately represents household wealth.
Mortality risks in this population vary with socioeconomic conditions31 and education.32 Further, arsenic exposure has been shown to differ among socioeconomic groups, and with a varying pattern over time.26 Those with higher socioeconomic status took the lead in shifting to tubewell water sources in the 1970s and 1980s, and have taken the lead in turning to low-arsenic or arsenic-free water during recent years. Thus, socioeconomic characteristics could potentially confound the associations between arsenic exposure and mortality. Further, a number of studies have indicated that people with poor nutrition (which likely is related to socioeconomic status) are particularly susceptible to arsenic-related health effects.36
Excess mortality burden associated with arsenic exposure was substantial in our study area, with higher risk even at low levels of arsenic exposure (10–49 μg/L) in drinking water. Nonaccidental adult mortality showed a strong association with arsenic exposure. In agreement with the arsenic-related increased mortality risk found in Taiwan,12 we observed 16%–36% increased mortality risk associated with arsenic in the adult population. The risk in Bangladesh may increase even more, because tubewells were still being installed in the late 1990s and some health consequences may not be obvious until more time has passed. In addition, recent studies from Chile suggest that exposure to arsenic very early in life greatly increases the mortality risk for both malignant and nonmalignant lung disease later in adulthood.5
Mortality due to infections showed a strong association with exposure and a clear dose-response relationship. Previous studies have shown elevated risk of chronic respiratory diseases in individuals who drinks arsenic-contaminated water.5,37 We have previously shown an arsenic-related increased risk of infant mortality, most likely due to deaths caused by infections.29 An interaction of arsenic with malnutrition is possible. Malnutrition may decrease the defense against pro-oxidants (such as arsenic) and affect arsenic metabolism.37–39 Arsenic exposure has also been found to damage immune function in children.40 A number of mechanistic studies support an effect of arsenic on immune function.41
Mean exposure to arsenic in our study was 13 years, with a range up to 30 years or more. The reported latency period for cancer and cardiovascular diseases varies from 20 to 40 years.1 Thus, our data may underestimate the eventual burden of mortality from arsenic. In rural Bangladesh, smoking is fairly common among men, while hardly any women smoke.42 Various forms of cancer, cardiovascular disease, and some of the infectious disease mortality are known to be associated with cigarette smoking, which might have contributed to the outcome.4,6 However, data on smoking were not available within this study, which thwarts an adequate analysis of sex differentials in mortality risks.
There are few previous epidemiologic studies designed to evaluate sex differences in arsenic-related health risks.43 In the present study, we did not observe any major sex differences in mortality risks, although men had slightly higher risk for arsenic-related cancers (eTable 3, http://links.lww.com/EDE/A343). A previous study in Chile found that arsenic-exposed men had higher cancer risk than women.13 The fact that men smoke much more than women in both these countries, may contribute to the higher risk among men. Obviously, more studies are needed to evaluate sex differences in arsenic-related carcinogenicity, as well as the underlying mechanisms and modifying factors. We have previously shown that men in this population were more susceptible than women to arsenic-induced skin lesions (pigmentation changes and hyperkeratosis).27 This may be related to the less efficient metabolism of arsenic in men compared with women14—a known risk factor for various forms of arsenic toxicity, including skin and bladder cancer.43 Whether this is relevant to the mortality risks seen in the present study remains to be elucidated.
In conclusion, the results show an excess nonaccidental adult mortality in rural populations in Bangladesh even at fairly low (10–49 μg/L) levels of arsenic in drinking water. There was similar excess adult mortality due to cancer, cardiovascular, and infections. About half of the Bangladesh population is exposed to arsenic concentrations exceeding 10 μg/L or above,17,19 suggesting a need for expanded efforts to reduce arsenic exposures in drinking water.
We thank the Matlab population who provided us with information and all staff who were involved with water sample collection, data collection, and processing. We also express our thanks to Samar Kumar Hore for his valuable contribution to field activities. ICDDR,B acknowledges with gratitude the commitment of Sida, WHO, and USAID to the Centre's research efforts.
1. IARC. Some drinking-water disinfectants and contaminants, including arsenic. IARC Monogr Eval Carcinog Risks Hum.
2. WHO/IPCS. Environmental Health Criteria 224, Arsenic and Arsenic Compounds
. 2nd ed. Geneva: World Health Organization; 2001.
3. Tseng WP. Effects and dose–response relationships of skin cancer and blackfoot disease with arsenic. Environ Health Perspect
4. Ferreccio C, Gonzalez C, Milosavjlevic V, Marshall G, Sancha AM, Smith AH. Lung cancer and arsenic concentrations in drinking water in Chile. Epidemiology
5. Smith AH, Marshall G, Yuan Y, et al. Increased mortality from lung cancer and bronchiectasis in young adults after exposure to arsenic in utero and in early childhood. Environ Health Perspect
6. Hopenhayn-Rich C, Biggs ML, Smith AH. Lung and kidney cancer mortality associated with arsenic in drinking water in Cordoba, Argentina. Int J Epidemiol
7. Hopenhayn-Rich C, Biggs ML, Fuchs A, et al. Bladder cancer mortality associated with arsenic in drinking water in Argentina. Epidemiology
8. Chiu HF, Ho SC, Wang LY, Wu TN, Yang CY. Does arsenic exposure increase the risk for liver cancer? J Toxicol Environ Health A
9. Navas-Acien A, Sharrett AR, Silbergeld EK, et al. Arsenic exposure and cardiovascular disease: a systematic review of the epidemiologic evidence. Am J Epidemiol
10. Meliker JR, Wahl RL, Cameron LL, Nriagu JO. Arsenic in drinking water and cerebrovascular disease, diabetes mellitus, and kidney disease in Michigan: a standardized mortality ratio analysis. Environ Health
11. Chen CJ, Wang SL, Chiou JM, et al. Arsenic and diabetes and hypertension in human populations: a review. Toxicol Appl Pharmacol
12. Tsai SM, Wang TN, Ko YC. Mortality for certain diseases in areas with high levels of arsenic in drinking water. Arch Environ Health
13. Smith AH, Smith MM. Arsenic drinking water regulations in developing countries with extensive exposure. Toxicology
14. Lindberg AL, Rahman M, Persson LA, Vahter M. The risk of arsenic induced skin lesions in Bangladeshi men and women is affected by arsenic metabolism and the age at first exposure. Toxicol Appl Pharmacol
15. Smith AH, Arroyo AP, Mazumder DN, et al. Arsenic-induced skin lesions among Atacameno people in Northern Chile despite good nutrition and centuries of exposure. Environ Health Perspect
16. Marshall G, Ferreccio C, Yuan Y, et al. Fifty-year study of lung and bladder cancer mortality in Chile related to arsenic in drinking water. J Natl Cancer Inst
17. Smith AH, Lingas EO, Rahman M. Contamination of drinking-water by arsenic in Bangladesh: a public health emergency. Bull World Health Organ
18. WHO. Arsenic in Drinking-Water. Background Document for Preparation of WHO Guidelines for Drinking-Water Quality
. Geneva: World Health Organization; 2003.
20. Ahsan H, Chen Y, Parvez F, et al. Arsenic exposure from drinking water and risk of premalignant skin lesions in Bangladesh: baseline results from the Health Effects of Arsenic Longitudinal Study. Am J Epidemiol
21. Razzaque A, Streatfield PK. Matlab Demographic Surveillance System, Bangladesh: Population and Health in Developing Countries
. Vol. 1. Ottawa, ON: IDRC & INDEPTH; 2002;387–395.
22. Chandramohan D, Maude GH, Rodrigues LC, Hayes RJ. Verbal autopsies for adult deaths: issues in their development and validation. Int J Epidemiol
23. Baqui AH, Black RE, Arifeen SE, Hill K, Mitra SN, al Sabir A. Causes of childhood deaths in Bangladesh: results of a nationwide verbal autopsy study. Bull World Health Organ
24. Alam N, Chowdhury HR, Bhuiyan MA, Streatfield PK. Special supplement on verbal autopsy and cause of death. In: Health and Demographic Surveillance System – Matlab
. Vol. 37. Dhaka, Bangladesh: ICDDR, B, Centre for Health and Population Research; 2004:61:–77.
25. Gajalakshmi V, Peto R. Commentary: verbal autopsy procedure for adult deaths. Int J Epidemiol
26. Rahman M, Vahter M, Wahed MA, et al. Prevalence of arsenic exposure and skin lesions. A population based survey in Matlab, Bangladesh. J Epidemiol Community Health
27. Rahman M, Vahter M, Sohel N, et al. Arsenic exposure and age and sex specific risk for skin lesions: A population-based case-referent study in Bangladesh. Environ Health Perspect
28. Cheng Z, van Geen A, Seddique AA, Ahmed KM. Limited temporal variability of arsenic concentrations in 20 wells monitored for 3 years in Araihazar, Bangladesh. Environ Sci Technol
29. Rahman A, Vahter M, Ekstrom EC, et al. Association of arsenic exposure during pregnancy with fetal loss and infant death: A cohort study in Bangladesh. Am J Epidemiol
30. Ruzicka LT, Chowdhury AKMA. Demographic Surveillance System – Matlab
. Vol. 2. Dacca, Bangladesh: Cholera Research Laboratory; 1978.
31. Mostafa G, van Ginneken JK. Trends in and determinants of mortality in the elderly population of Matlab, Bangladesh. Soc Sci Med
32. Hurt LS, Ronsmans C, Saha S. Effects of education and other socioeconomic factors on middle age mortality in rural Bangladesh. J Epidemiol Community Health
34. Morgenstern H. Uses of ecologic analysis in epidemiologic research. Am J Public Health
35. Yoshida T, Yamauchi H, Fan Sun G. Chronic health effects in people exposed to arsenic via the drinking water: dose-response relationships in review. Toxicol Appl Pharmacol
36. Milton AH, Hasan Z, Shahidullah SM, et al. Association between nutritional status and arsenicosis due to chronic arsenic exposure in Bangladesh. Int J Environ Health Res
37. Mazumder DN, Steinmaus C, Bhattacharya P, et al. Bronchiectasis in persons with skin lesions resulting from arsenic in drinking water. Epidemiology
38. Hall M, Gamble M, Slavkovich V, et al. Determinants of arsenic metabolism: blood arsenic metabolites, plasma folate, cobalamin, and homocysteine concentrations in maternal-newborn pairs. Environ Health Perspect
39. Vahter M, Marafante E. Effects of low dietary intake of methionine, choline or proteins on the biotransformation of arsenite in the rabbit. Toxicol Lett
40. Soto-Pena GA, Luna AL, Acosta-Saavedra L, et al. Assessment of lymphocyte subpopulations and cytokine secretion in children exposed to arsenic. Faseb J
41. Lemarie A, Morzadec C, Bourdonnay E, Fardel O, Vernhet L. Human macrophages constitute targets for immunotoxic inorganic arsenic. J Immunol
42. Cohen N. Smoking, health, and survival: prospects in Bangladesh. Lancet
© 2009 Lippincott Williams & Wilkins, Inc.
43. Vahter M, Akesson A, Liden C, Ceccatelli S, Berglund M. Gender differences in the disposition and toxicity of metals. Environ Res