- Email: [email protected]
Arsenic in public water supplies and cardiovascular mortality in Spain
ARTICLE IN PRESS Environmental Research 110 (2010) 448–454
Contents lists available at ScienceDirect
Environmental Research journal homepage: www.elsevier.com/locate/envres
Arsenic in public water supplies and cardiovascular mortality in Spain$, $$ Ma Jose´ Medrano a,, Raquel Boix a, Roberto Pastor-Barriuso a, Margarita Palau b, Javier Damia´n a, Rebeca Ramis a,c, Jose´ Luis del Barrio d, Ana Navas-Acien e,f a
Centro Nacional de Epidemiologı´a, Instituto de Salud Carlos III, Sinesio Delgado 6, 28029 Madrid, Spain ´n General de Sanidad Ambiental y Salud Laboral, Direccio ´n General de Salud Pu ´ blica y Sanidad Exterior, Ministerio de Sanidad y Polı´tica Social, Madrid, Spain Subdireccio c ´ blica (CIBERESP), Madrid, Spain CIBER en Epidemiologı´a y Salud Pu d ´ blica, Universidad Rey Juan Carlos, Madrid, Spain Departamento de Salud Pu e Department of Environmental Health Sciences, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA f Department of Epidemiology, Welch Center for Prevention, Epidemiology and Clinical Research, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA b
a r t i c l e in f o
a b s t r a c t
Article history: Received 2 June 2009 Received in revised form 22 September 2009 Accepted 1 October 2009 Available online 31 October 2009
Background: High-chronic arsenic exposure in drinking water is associated with increased cardiovascular disease risk. At low-chronic levels, as those present in Spain, evidence is scarce. In this ecological study, we evaluated the association of municipal drinking water arsenic concentrations during the period 1998 2002 with cardiovascular mortality in the population of Spain. Methods: Arsenic concentrations in drinking water were available for 1721 municipalities, covering 24.8 million people. Standardized mortality ratios (SMRs) for cardiovascular (361,750 deaths), coronary (113,000 deaths), and cerebrovascular (103,590 deaths) disease were analyzed for the period 1999 2003. Two-level hierarchical Poisson models were used to evaluate the association of municipal drinking water arsenic concentrations with mortality adjusting for social determinants, cardiovascular risk factors, diet, and water characteristics at municipal or provincial level in 651 municipalities (200,376 cardiovascular deaths) with complete covariate information. Results: Mean municipal drinking water arsenic concentrations ranged from o 1 to 118 mg/L. Compared to the overall Spanish population, sex- and age-adjusted mortality rates for cardiovascular (SMR 1.10), coronary (SMR 1.18), and cerebrovascular (SMR 1.04) disease were increased in municipalities with arsenic concentrations in drinking water 4 10 mg/L. Compared to municipalities with arsenic concentrations o 1 mg/L, fully adjusted cardiovascular mortality rates were increased by 2.2% ( 0.9% to 5.5%) and 2.6% ( 2.0% to 7.5%) in municipalities with arsenic concentrations between 1 10 and 4 10 mg/L, respectively (P-value for trend 0.032). The corresponding figures were 5.2% (0.8% to 9.8%) and 1.5% ( 4.5% to 7.9%) for coronary heart disease mortality, and 0.3% ( 4.1% to 4.9%) and 1.7% ( 4.9% to 8.8%) for cerebrovascular disease mortality. Conclusions: In this ecological study, elevated low-to-moderate arsenic concentrations in drinking water were associated with increased cardiovascular mortality at the municipal level. Prospective cohort studies with individual measures of arsenic exposure, standardized cardiovascular outcomes, and adequate adjustment for confounders are needed to confirm these ecological findings. Our study, however, reinforces the need to implement arsenic remediation treatments in water supply systems above the World Health Organization safety standard of 10 mg/L. & 2009 Elsevier Inc. All rights reserved.
Keywords: Arsenic Drinking water Cardiovascular diseases Mortality
1. Introduction
$ Funding Sources: Research Grant no. PI060656 from the Spanish Fondo de Investigacio´n Sanitaria. Ana Navas-Acien was supported by Grant 1R01HL090863 from the US National Heart Lung and Blood Institute. $$ The study was approved by the Research Committee of the Carlos III Institute of Health. The study complies with the confidentiality rules established by the Spanish National Institute for Statistics (data transfer protocol 24/11/06). Corresponding author. Fax: + 34 91 387 7815. E-mail address: [email protected] (M.J. Medrano).
0013-9351/$ - see front matter & 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.envres.2009.10.002
The vascular effects of high-chronic arsenic exposure in drinking water ( 4100 mg/L) are supported by experimental (States et al., 2009; Simeonova and Luster, 2004; Soucy et al., 2005) and epidemiological evidence from Taiwan (Tseng, 2008; Wang et al., 2007) and Chile (Yuan et al., 2007). Long-term arsenic exposure in Taiwan has been long recognized as the cause of a severe form of peripheral arterial disease called black foot disease (Tseng et al., 2005; Chen et al., 1998) and it has been associated with coronary and cerebrovascular diseases and with subclinical markers of
ARTICLE IN PRESS M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
atherosclerosis (Wu et al., 1989; Chen et al.,1996; Chiou et al., 1997; Wang et al., 2002). In Region II Chile, acute myocardial infarction mortality increased following a period of high exposure to arsenic in drinking water and decreased after arsenic remediation had been implemented (Yuan et al., 2007). In Bangladesh, Mexico and Taiwan, high arsenic exposure in drinking water has been related with cardiovascular risk factors, such as hypertension (Chen et al., 1995, 2007; Rahman et al., 1999) and diabetes (Coronado-Gonza´lez et al., 2007; Rahman et al., 1998; Lai et al., 1994). At moderate or low chronic inorganic arsenic exposure in drinking water ( o100 mg/L), levels that are relevant for most regions in North America, Europe, and other parts of the world, few epidemiologic studies have addressed the association of inorganic arsenic with cardiovascular endpoints (Navas-Acien et al., 2005). Mortality for diseases of the arteries, arterioles, and capillaries were increased in counties with arsenic levels in drinking water 420 mg/L vs. 510 mg/L but not with coronary heart disease and cerebrovascular mortality in a 30-county US study (Engel and Smith, 1994). In Michigan, mortality was increased for all diseases of the circulatory system and cerebrovascular diseases but not for ischemic heart disease in 6 counties with drinking water arsenic 410 mg/L compared to the overall State (Meliker et al., 2007). In a cross-sectional study conducted among residents with private wells in Wisconsin, self-reported circulatory problems including coronary bypass and myocardial infarction were increased comparing arsenic in drinking water 10 410 vs. o2 mg/L (Zierold et al., 2004). In Spain, arsenic concentrations in public water supplies are monitored through a water quality surveillance system that compiles municipal water information since 1992. While arsenic concentrations in drinking water were then generally below 50 mg/L (Ministerio de Sanidad y Consumo, 2000), a relatively important number of communities had concentrations above 10 mg/L, the safety standard for drinking water arsenic that was regulated in 2002. In this study, our goal was to evaluate the association of municipal tap drinking water arsenic concentrations during 1998 2002 with cardiovascular mortality in the population of Spain.
2. Methods 2.1. Study design and population For this ecological study, municipal-level arsenic concentrations in drinking water and mortality data were available for 1721 municipalities located in 49 out of 52 Spanish provinces, covering 24.8 million people (60.7% of the total Spanish population in 2001) (Fig. 1). Non-respondent municipalities (municipalities with no arsenic data during the study period) were located in large low-populated geographical areas, where water control procedures for arsenic and other metals were not fully established in 1998 2002. Compared to respondent municipalities, non-respondent municipalities were less populated (mean 1527 vs. 14,400 inhabitants), older (24.2% vs. 19.1% inhabitants Z 65 years of age), more rural (99.4% vs. 95.5%), and less well-off (5.0 vs. 6.0 in 1 10 ordinal income scale). The study was approved by the Research Committee of the Carlos III Institute of Health and complies with the Spanish National Institute for Statistics confidentiality rules.
2.2. Arsenic in public water supply systems Tap drinking water arsenic concentrations at the municipal level during 1998 2002 were obtained from the National Information System of Consume Water Control, held by the Spanish Ministry of Health (Ministerio de Sanidad y Consumo, 2000). This administrative database was established to track drinking water quality by gathering annual data from local public health authorities responsible for water quality control, including tap water arsenic information from every public water supply system. For this study, a total of 20,074 arsenic determinations were available. The mean (SD) number of arsenic determinations per municipality was 11.7 (41.1), ranging from 1 to 802.
449
> 50 µg/L 10−50 µg/L 1−10 − µg/L < 1 µg/L
Fig. 1. Geographical distribution of arsenic concentrations in municipal drinking water in Spain, 1998–2002. Blank areas represent non-respondent municipalities. Black lines represent provincial boundaries. Arsenic concentrations in municipal water were measured at regional laboratories participating in the System of Consume Water Control, all of them holding accreditation and/or certification following international rules for quality control (UNE-EN ISO/IEC 17025, UNE-EN ISO 9001). Laboratory procedures for sample management, analytical methods, and quality control measures (accuracy, precision, and detection limits) were standardized by Spanish and European Union laws (Royal Decrees 1138/90, 140/03, and 1423/82; EU Directive 98/83/CE). Following these protocols, arsenic concentrations were measured using atomic absorption spectrophotometry in the participant laboratories. The highest detection limit permitted by regulation was 1 mg/L, although the actual detection limit for each laboratory was not available. A total of 1066 (61.9%) municipalities reported all arsenic determinations as undetected and were replaced by a level equal to 1 mg/L divided by the square root of 2. For 270 (15.7%) municipalities with samples analyzed in laboratories with detection limits lower than 1 mg/L, the actual detected values below 1 mg/L were included in the analyses.
2.3. Cardiovascular mortality data Cardiovascular mortality was analyzed for the period 1999 2003 to allow for 1-year delay with respect to arsenic determinations. The 1-year delay was selected based on short-term changes in myocardial infarction rates with increasing and decreasing arsenic levels in drinking water in Region II, Chile (Yuan et al., 2007). Adding a 3-year delay or no delay resulted in similar findings (results not shown). The observed number of deaths from cardiovascular diseases (codes I00 I99 of the tenth revision of the International Classification of Diseases), coronary heart diseases (codes I20 I25), and cerebrovascular diseases (codes I60 I69) at the municipal level for men and women was obtained from the National Institute for Statistics. A total of 361,750 cardiovascular disease deaths were included in the analysis. The number of coronary and cerebrovascular disease deaths was 113,000 and 103,590, respectively. The sex- and age-adjusted cardiovascular mortality rate in respondent municipalities was similar to the overall national rate (standardized mortality ratio [SMR] of 0.99).
2.4. Other relevant variables Municipal-level water characteristics (hardness and magnesium content, pH, and temperature) were obtained from the National Information System of Consume Water Control. Municipal-level socioeconomic indicators (per capita income measured with an ordinal scale from 1 to 10, with 10 reflecting higher ˜ ol de Cre´dito, 1993). income) were obtained from bank statistics (Banco Espan Information for other relevant sociodemographic, dietary, and cardiovascular risk factors was considered at the provincial level because they were not available at the municipal level. Health care access (number of hospital beds per 1000 inhabitants) was obtained from 2001 official statistics (Instituto Nacional de Estadı´stica, 2004); dietary consumption of fish, wine, olive oil, bottled water, and total energy-adjusted folate was obtained from the 1991 National Nutrition Study (Varela et al., 1995); and sex- and age-specific prevalences of smoking, hypertension, high serum cholesterol, diabetes, overweight/obesity, and low physical activity were obtained from the 1995 National Health Survey (Ministerio de Sanidad y Consumo, 1995). Complete information for covariates was only
ARTICLE IN PRESS 450
M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
available in 651 (37.8%) municipalities corresponding to 200,376 (55.4%) cardiovascular disease deaths and 14.4 million (58.1%) population size.
2.5. Statistical analysis For each municipality, all available arsenic determinations in drinking water during the period 1998 2002 were averaged as an approximation to long-term arsenic exposure. Municipalities were classified according to their mean arsenic concentrations as o1, 1 10, and 410 mg/L. Mean levels of sociodemographic and water characteristics, as well as of dietary and cardiovascular risk factors, were compared across categories of municipal arsenic concentration using linear regression models. For each municipality, the expected number of cardiovascular, coronary, and cerebrovascular disease deaths among men and women for the period 1999 2003 was calculated by applying national sex- and age-specific mortality rates in 2001 to the estimated number of person-years by sex and age in each municipality over the period 1999 2003. Assuming a Poisson distribution for the observed number of deaths (Greenland and Rothman, 2008), SMRs and their 95% confidence intervals (CIs) were computed comparing observed to expected number of deaths within each category of municipal arsenic concentration. For risk assessment, the association of municipal arsenic concentration in drinking water with cardiovascular, coronary, and cerebrovascular disease mortality was evaluated using two-level hierarchical Poisson models with random intercept (Gelman and Hill, 2007). More specifically, at the first municipal level, the observed number of deaths Oij in municipality i of province j was assumed to follow the log-linear Poisson model
3. Results Mean municipal arsenic concentrations in drinking water ranged from o1 to 118 mg/L, with 7 villages (joint population 10,613 inhabitants) registering mean arsenic concentrations 450 mg/L (Fig. 1), the Spanish safety standard for arsenic in drinking water during the study period. Differences in sociodemographic, dietary, and cardiovascular risk factors across water arsenic categories were generally small but statistically significant (Table 1). Compared to municipalities with arsenic concentrations o1 mg/L, municipalities with concentrations 410 mg/L tended to be younger, had higher levels of water hardness and pH, and were located in provinces with less hospital beds per 1000 inhabitants, higher sex- and age-adjusted prevalence of cardiovascular risk factors, and lower intake of fish, wine, olive oil, and total energyadjusted folate. Compared to the overall Spanish population, sex- and ageadjusted mortality rates for cardiovascular disease (SMR 1.10, 95% CI 1.08–1.12), coronary heart disease (SMR 1.18, 95% CI 1.15–1.22), and cerebrovascular disease (SMR 1.04, 95% CI 1.01–1.08) were increased in the 89 municipalities with mean arsenic concentrations in drinking water 410 mg/L, with no substantial differences by sex (Table 2). No increased cardiovascular, coronary, or
Oij PoissonfEij expðaj þ bxij þ d0 zij Þg; where Eij is the expected number of deaths under national sex- and age-specific mortality rates, aj is the province-specific intercept, and b and d are the coefficients associated to the log-transformed municipal arsenic concentration xij and other municipal-level adjustment factors zij, whose effects were constrained to be the same for all provinces. At the second provincial level, the intercept aj was allowed to vary randomly across provinces according to a normal distribution with mean depending on provincial-level covariates wj and with common betweenprovince variance t2,
aj Nðg0 þ c0 wj ; t2 Þ; where g0 is the overall intercept and c are the coefficients associated with provincial-level adjustment factors wj. In addition to the fixed effects for municipal- and provincial-level explanatory variables, the model includes a random intercept term that accounts for unexplained provincial-level variations in the expected number of deaths. The model was fitted using penalized quasilikelihood methods (Breslow and Clayton, 1993), and the resulting regression coefficients were interpreted in terms of percentage change in mortality by applying the transformation 100{exp(b) 1}. To explore the association between arsenic concentration in drinking water and cardiovascular mortality endpoints without assuming a log-linear doseresponse relation, we also estimated percentage changes in mortality and their 95% CIs comparing municipalities in the highest arsenic categories (1 10 and 410 mg/L) to those in the lowest one (o 1 mg/L), as well as using restricted quadratic splines for log-transformed municipal arsenic concentrations with knots at 0.5, 2, and 10 mg/L, which corresponded to the 21st, 67th and 88th percentiles of the arsenic distribution in the 651 municipalities with complete covariate information (Greenland, 1995). Selecting knots at different percentiles resulted in similar dose-response relations (results not shown). The above models were fitted with increasing levels of adjustment. First, we adjusted for sex, age, per capita municipal income, and health care access at provincial level (hospital beds per 1000 inhabitants). Second, we further adjusted for sex- and age-adjusted prevalence of cardiovascular risk factors at provincial level (smoking, hypertension, high serum cholesterol, diabetes, overweight/obesity, and low physical activity). Third, we further adjusted for dietary factors at provincial level (fish, wine, olive oil, bottled water, and total energy-adjusted folate) and water characteristics at municipal level (hardness and magnesium content, pH, and temperature). Because cardiovascular mortality rates in Spain are much higher in men than women, we also performed separate analyses by sex. Sensitivity analyses were conducted to assess the potential impact of data quality in a subset of 356 municipalities (covering 16.4 million people and 232,074 cardiovascular disease deaths) with at least 8 arsenic determinations (mean of 46.7 determinations per municipality), yielding similar results to those obtained in the whole sample (not shown). All analyses were performed in R statistical software (R Foundation for Statistical Computing, Vienna, Austria), using the glmmPQL function in MASS package to fit random-intercept hierarchical Poisson models. The spatial distribution of municipal arsenic concentrations was depicted using MapInfo (MapInfo Corporation, New York).
Table 1 Characteristics of study municipalities by arsenic concentration in drinking water. Characteristic
Total
Arsenic category (lg/L) o1
1 10 410
Pvaluea
No. of municipalities
1721
1336
296
89
Sociodemographic characteristics Total population (thousands) Mean population (thousands) Rural (%) Z 65 years (%) Per capita incomeb Hospital beds per 1000 inhabitants
24,793 14.4 95.5 19.1 6.0 9.1
18,978 14.2 95.9 19.9 6.0 9.2
4803 16.2 93.6 16.2 5.9 8.8
1011 11.4 96.6 16.5 5.7 8.4
0.89 0.19 o 0.001 0.24 o 0.001
Cardiovascular risk factorsc Smoking (%) Hypertension (%) High serum cholesterol (%) Diabetes (%) Obesity (%) Low physical activity (%)
27.5 17.1 12.1 6.3 11.5 61.2
27.1 16.7 11.8 6.2 11.2 60.9
28.4 18.1 13.1 6.7 12.1 60.8
29.7 19.9 13.4 7.2 13.3 66.6
o 0.001 o 0.001 o 0.001 o 0.001 o 0.001 o 0.001
Dietary factorsd Fish (g/person-day) Wine (g/person-day) Olive oil (g/person-day) Bottled water (g/person-day) Total energy-adjusted folatee
73.7 63.9 34.5 20.8 3.5
73.8 63.6 35.3 20.8 3.6
75.6 70.1 33.1 20.0 3.4
65.6 48.1 27.4 24.3 2.7
o 0.001 o 0.001 o 0.001 0.22 o 0.001
Water characteristicsf Total no of arsenic determinations Mean no of arsenic determinations Arsenic (mg/L) Hardness (mg/L) Magnesium (mg/L) pH Temperature (1C)
20,074 11.7 2.4 75.1 23.9 7.2 16.8
14,796 11.1 0.7 58.6 25.9 7.2 16.6
3890 13.1 3.9 137.3 17.4 6.8 16.7
1388 15.6 23.3 115.0 17.3 7.6 17.9
0.48 o 0.001 0.56 o 0.001 0.10
a P-value for heterogeneity of means across categories of municipal arsenic concentration. b Ordinal variable from 1 (lowest income) to 10 (highest income). c Sex- and age-adjusted prevalence of cardiovascular risk factors according to the national population distribution in 2001. d Food content of family diet. e Ordinal variable from 1 (lowest quintile of total energy-adjusted folate) to 5 (highest quintile). f Measurements of hardness, magnesium, pH, and temperature were available in 1608, 1576, 973, and 773 municipalities, respectively.
ARTICLE IN PRESS M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
451
Table 2 Standardized mortality ratio (95% confidence interval) for cardiovascular, coronary, and cerebrovascular disease by municipal arsenic concentration in drinking water.a Arsenic category (lg/L)
Total
Men
Women
No. of deaths
SMR (95% CI)b
No. of deaths
SMR (95% CI)D
No. of deaths
SMR (95% CI)D
285,049 62,739 13,962 361,750
0.98 (0.976–0.983) 0.99 (0.985–1.000) 1.10 (1.078–1.115) 0.99 (0.983–0.989)
128,504 28,214 6567 163,285
1.00 (0.993–1.004) 0.98 (0.964–0.987) 1.10 (1.071–1.124) 1.00 (0.993–1.003)
156,545 34,525 7395 198,465
0.97 (0.960–0.970) 1.01 (0.996–1.018) 1.10 (1.071–1.121) 0.98 (0.972–0.981)
Coronary heart disease mortality o1 88,566 1 10 19,709 410 4725 Total 113,000
0.99 (0.980–0.993) 1.00 (0.984–1.011) 1.18 (1.148–1.216) 0.99 (0.989–1.001)
50,121 11,045 2734 63,900
0.99 (0.985–1.003) 0.97 (0.950–0.986) 1.16 (1.119–1.206) 1.00 (0.988–1.003)
38,445 8664 1991 49,100
0.98 1.04 1.21 0.99
(0.966–0.986) (1.017–1.061) (1.157–1.264) (0.985–1.003)
Cerebrovascular disease mortality o1 81,368 1 10 18,327 410 3895 Total 103,590
0.95 0.99 1.04 0.96
33,321 7464 1623 42,408
0.97 0.98 1.02 0.98
48,047 10,863 2272 61,182
0.93 1.00 1.06 0.95
(0.923–0.940) (0.978–1.015) (1.017–1.104) (0.939–0.954)
Cardiovascular mortality o1 1 10 410 Total
a b
(0.941–0.954) (0.974–1.003) (1.012–1.077) (0.952–0.964)
(0.962–0.983) (0.954–0.999) (0.975–1.074) (0.966–0.984)
Analyses based on all 1721 municipalities with available arsenic concentrations in drinking water. Standardized mortality ratio (95% confidence interval) comparing observed to expected number of deaths under national sex- and age-specific mortality rates.
Table 3 Percentage change (95% confidence interval) in cardiovascular, coronary, and cerebrovascular disease mortality by municipal arsenic concentration in drinking water.a 2-Fold increase in arsenic concentration
P-value for trendb
Arsenic category (lg/L) o1
1 10
410
No. of municipalities
651
383
193
75
Cardiovascular mortality No. of male/female deaths Model 1—total Model 2—total Model 3—total Model 3—men Model 3—women
90,195/110,181 0.7 ( 0.1 to 1.6) 0.7 ( 0.1 to 1.5) 0.8 (0.1 to 1.6) 0.7 ( 0.1 to 1.5) 1.0 (0.0 to 1.9)
65,723/80,844 0 (reference) 0 (reference) 0 (reference) 0 (reference) 0 (reference)
18,844/23,113 2.0 ( 1.2 to 5.4) 1.4 ( 1.9 to 4.7) 2.2 ( 0.9 to 5.5) 0.4 ( 2.8 to 3.8) 4.0 (0.2 to 8.0)
5628/6224 1.8 ( 2.9 to 6.8) 1.7 ( 3.0 to 6.7) 2.6 ( 2.0 to 7.5) 2.4 ( 2.5 to 7.5) 3.2 ( 2.5 to 9.1)
0.080 0.090 0.032 0.095 0.048
Coronary heart disease mortality No. of male/female deaths Model 3—total Model 3—men Model 3—women
35,862/27,595 0.9 ( 0.1 to 2.0) 0.6 ( 0.6 to 1.8) 1.3 ( 0.2 to 2.7)
26,117/20,065 0 (reference) 0 (reference) 0 (reference)
7394/5819 5.2 (0.8 to 9.8) 0.4 ( 4.4 to 5.4) 11.3 (5.1 to 17.8)
2351/1711 1.5 ( 4.5 to 7.9) 1.5 ( 5.3 to 8.8) 3.3 ( 4.8 to 12.2)
0.091 0.31 0.084
Cerebrovascular disease mortality No. of male/female deaths Model 3—total Model 3—men Model 3—women
22,041/31,985 0.7 ( 0.4 to 1.8) 0.7 ( 0.7 to 2.1) 0.7 ( 0.6 to 2.0)
15,831/23,122 0 (reference) 0 (reference) 0 (reference)
4850/7012 0.3 ( 4.1 to 4.9) 0.1 ( 5.5 to 5.7) 0.8 ( 4.5 to 6.3)
1360/1851 1.7 ( 4.9 to 8.8) 1.5 ( 6.6 to 10.3) 1.4 ( 6.3 to 9.8)
0.20 0.34 0.31
Model 1 adjusted for sex, age, per capita municipal income, and hospital beds per population at provincial level. Model 2 further adjusted for prevalence of cardiovascular risk factors at provincial level (smoking, hypertension, high serum cholesterol, diabetes, overweight/obesity, and low physical activity). Model 3 further adjusted for dietary factors at provincial level (fish, wine, olive oil, bottled water, and total energy-adjusted folate) and water characteristics at municipal level (hardness, magnesium, pH, and temperature). a b
Analyses restricted to 651 municipalities (14.4 million people) with complete information on adjustment factors. P-value for trend based on log-transformed arsenic concentrations.
cerebrovascular mortality was observed for the category of arsenic 1 10 mg/L, except for coronary heart disease mortality among women (SMR 1.04, 95% CI 1.02–1.06). In multivariable models adjusted for social determinants, cardiovascular risk factors, diet, and water characteristics, cardiovascular mortality increased 0.8% (95% CI 0.1–1.6%) for each doubling in arsenic concentrations (Table 3, model 3). Compared to municipalities with arsenic concentrations o1 mg/L, cardiovascular disease mortality was increased by 2.2% (95% CI 0.9% to 5.5%) in municipalities with arsenic concentrations between 1 and 10 mg/L and by 2.6% (95% CI 2.0% to 7.5%) in municipalities with arsenic concentrations 410 mg/L, with
consistent findings by sex, except a stronger association for the arsenic 1 10 mg/L category among women. Increased cardiovascular mortality with increasing arsenic concentrations in municipal drinking water was also observed using restricted quadratic splines (Figs. 2 and 3). For coronary heart disease mortality, although the number of events was smaller, fully adjusted relative risks were consistent with increased risk, particularly among women. For cerebrovascular disease mortality, the association was weaker and not statistically significant. We further evaluated whether increased mortality in municipalities with arsenic concentrations in drinking water between 1
ARTICLE IN PRESS M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
30 5 20 0 10 -5
0 0.2 0.5 1 2 5 10 20 Arsenic concentration in drinking water (µg/L)
Fig. 2. Percentage change in cardiovascular mortality by municipal arsenic concentration in drinking water. Curves represent the adjusted percentage change in cardiovascular mortality (solid curve) and its 95% confidence interval (dashed curves) based on restricted quadratic splines for log-transformed arsenic concentrations with knots at 0.5, 2, and 10 mg/L. The reference value (percentage change= 0) was set at 0.5 mg/L. Percentage changes were adjusted for sex, age, per capita municipal income, hospital beds per population at provincial level, prevalence of cardiovascular risk factors at provincial level (smoking, hypertension, high serum cholesterol, diabetes, overweight/obesity, and low physical activity), dietary factors at provincial level (fish, wine, olive oil, bottled water, and total energy-adjusted folate), and water characteristics at municipal level (hardness, magnesium, pH, and temperature). Bars represent the histogram of municipal arsenic distribution.
and 10 mg/L could be influenced by 16 municipalities that were geographically located in the arsenic contaminated area of the aquifer Los Arenales (Sahu´n et al., 2004; Garcı´a-Villanova et al., 2005), where inhabitants may drink from private uncontrolled wells with higher arsenic concentrations. Excluding these municipalities from multivariable adjusted analyses, the percentage increase in coronary heart disease mortality among women was 11.9% (95% CI 5.7–18.7%) comparing municipalities with arsenic concentrations between 1 and 10 mg/L to those o1 mg/L, similar to the result obtained in the whole sample.
4. Discussion In this ecologic study covering nearly 60% of the total Spanish population, cardiovascular mortality rates in municipalities with arsenic concentrations in drinking water 410 mg/L were 3.2% increased in women and 2.4% increased in men compared to municipalities with concentrations o1 mg/L, after adjustment for differences in social determinants, cardiovascular risk factors, diet, and water characteristics, measured at the municipal or provincial level. For municipalities with arsenic concentrations between 1 and 10 mg/L, the corresponding figures were 4.0% in women and 0.4% in men. For coronary heart disease mortality, the multivariable adjusted association was stronger for women compared to men. For cerebrovascular disease mortality, multivariable associations were weaker and not statistically significant. Our study extends previous findings in populations with high arsenic concentrations in drinking water (Tseng, 2008; NavasAcien et al., 2005; Wang et al., 2007; Yuan et al., 2007) to populations with low-to-moderate concentrations, supporting a role for low-chronic arsenic exposure in cardiovascular and coronary heart disease development. The association was modest and only borderline statistically significant after multivariable adjustment. However, more than 1 million people in the study lived in municipalities with arsenic concentrations 4 10 mg/L and almost 5 millions lived in municipalities with arsenic concentrations between 1 and 10 mg/L. Moreover, inherent to the ecological
Men 40
10
30 5 20 0 10 -5
0 0.2
0.5
1
2
5
10
20
Women 40
10
30
Percentage of municipalities
40
Percentage change in cardiovascular mortality
Percentage change in cardiovascular mortality
10
Percentage of municipalities
452
5 20 0 10 -5
0 0.2
0.5
1
2
5
10
20
Arsenic concentration in drinking water (µg/L) Fig. 3. Percentage change in sex-specific cardiovascular mortality by municipal arsenic concentration in drinking water. Curves represent the adjusted percentage changes in cardiovascular mortality (solid curves) and their 95% confidence intervals (dashed curves) based on restricted quadratic splines for log-transformed arsenic concentrations with knots at 0.5, 2, and 10 mg/L. The reference value (percentage change=0) was set at 0.5 mg/L. Percentage changes were adjusted for age, per capita municipal income, hospital beds per population at provincial level, sex-specific prevalence of cardiovascular risk factors at provincial level (smoking, hypertension, high serum cholesterol, diabetes, overweight/obesity, and low physical activity), dietary factors at provincial level (fish, wine, olive oil, bottled water, and total energy-adjusted folate), and water characteristics at municipal level (hardness, magnesium, pH, and temperature). Bars represent the histogram of municipal arsenic distribution.
design, the associations could have been underestimated because of substantial measurement error in both arsenic exposure and cardiovascular mortality measures. Arsenic is an established carcinogen and a high priority contaminant for screening in drinking water sources (World Health Organization, 2003). Low-to-moderate exposure to inorganic arsenic in drinking water affects many populations around the world. In Spain, arsenic concentrations in drinking water are relatively low, being o1 mg/L for the great majority of the population. In our study, 4.1% of the population was 410 mg/L, similar to those reported in Europe and North America (Smith et al., 2002). The World Health Organization (WHO), the European Union, and many countries around the world have established 10 mg/L as the safety standard for arsenic concentrations in drinking water, mostly based on risk benefit analysis for the carcinogenic effects of arsenic and practical/economic difficulties to remove arsenic in drinking water o10 mg/L. However, because many uncertainties remain at low exposure levels, in particular for non-cancer endpoints, the WHO designated 10 mg/L as a provisional standard. In the US, the National Research Council
ARTICLE IN PRESS M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
concluded that more research was needed to evaluate the association of chronic arsenic exposure with cardiovascular disease (National Research Council, 1999, 2001). Ecological and cross-sectional studies evaluating low-chronic arsenic exposure in drinking water and cardiovascular endpoints have been contradictory. Our findings are consistent with increased mortality for diseases of the arteries, arterioles, and capillaries in US counties with arsenic levels in drinking water 420 mg/L (Engel and Smith, 1994), increased mortality for diseases of the circulatory system in six Michigan counties with drinking water arsenic 410 mg/L (Meliker et al., 2007), and increased prevalence of self-reported circulatory problems with arsenic in drinking water 410 mg/L in Wisconsin (Zierold et al., 2004). In the study by Zierold et al. (2004), participants exposed to arsenic 2 10 mg/L in drinking water were also more likely to report a history of heart disease (odds ratio 1.52, 95% CI 1.00– 2.35). Prospective cohort studies with individual measures of arsenic exposure, standardized cardiovascular outcomes and adequate adjustment for confounders are needed to confirm the ecological findings at low-chronic exposure levels (Navas-Acien et al., 2005). Experimental and mechanistic studies have shown that arsenic can induce atherosclerosis, up-regulate inflammatory signals and enhance oxidative stress (Bunderson et al., 2004; Simeonova and Luster, 2004; States et al., 2009). The relevance of those findings at low-chronic levels has been uncertain because of typically high arsenic concentrations used in these experiments. Recent studies using lower and lower arsenic concentrations are identifying important vascular effects, including neovascularization, angiogenesis, and vessel remodeling at levels o10 mg/L in drinking water (Soucy et al., 2005; Straub et al., 2007, 2008). In our study, the strength of the association for the arsenic category 410 mg/L with cardiovascular and coronary mortality were comparable by gender. For arsenic 1 10 mg/L, however, the association with cardiovascular and coronary mortality was stronger for women compared to men. Reasons for gender differences at lower arsenic exposure concentrations could be related to differential confounding by gender, competing mortality by other conditions related to arsenic toxicity that could be more common among men (lung cancer, bladder cancer, chronic respiratory disease, for instance, because of a strong interaction between arsenic and smoking that is more common among men (Ministerio de Sanidad y Consumo, 1995)), metabolic reasons, or other reasons. Because of its design, this study evaluates the public health impact of arsenic in drinking water with cardiovascular mortality at the municipal level, but is limited to infer the relation of arsenic with cardiovascular mortality at the individual level. Important non-differential exposure misclassification could have substantially underestimated the strength of the associations. Other limitations must be considered. Firstly, selection bias induced by selective participation in the National Information System of Consume Water Control cannot be ruled out, although is unlikely given similar cardiovascular mortality, diet, and prevalence of risk factors in municipalities with and without arsenic exposure information. Secondly, the quality of arsenic assays and number of measures could have potentially affected the study, although consistent results were found in sensitivity analysis restricted to municipalities with at least 8 arsenic determinations. Thirdly, other sources of inorganic arsenic exposure, such as food and past-use of arsenic containing pesticides were not included. Finally, although we used hierarchical Poisson models with province-specific random intercepts to adjust the association between municipal arsenic concentration in drinking water and cardiovascular mortality for between-province differences in any relevant confounding factor, residual confounding induced by
453
within-province variations in social determinants, cardiovascular risk factors, and diet cannot be ruled out. Compared to other ecological studies, this study adjusted for potential confounders at the municipal or provincial level. Lack of adjustment for differences in socioeconomic and cardiovascular factors across communities could maybe explain some of the inconsistencies in previous ecological analyses (Engel and Smith, 1994; Meliker et al., 2007). Other strengths include the large study area covering almost 25 million individuals across Spain, multiple water arsenic determinations with consistent concentrations over time for most municipalities, and the use of procedures for water sampling and chemical analyses that were standardized by Spanish regulations. Statistical strengths include the use of multilevel (municipal and provincial) methods, the inclusion of a random intercept accounting for unexplained provincial-level variations in mortality, dose-response analyses using policy relevant cut-offs and flexible splines, and stratification by gender. In conclusion, this ecological study suggests that elevated arsenic concentrations in drinking water were associated with increased cardiovascular mortality at the municipal level in a country characterized by low-to-moderate arsenic exposure levels. Among women, the association with cardiovascular and coronary disease mortality was observed at arsenic concentrations below current standards (1 10 mg/L). Chronic exposure to low/moderate inorganic arsenic from drinking water and diet is widespread. Given increasing evidence of environmental exposures as cardiovascular disease risk factors (Bhatnagar, 2006), the investigation of the cardiovascular impact of arsenic in population-based prospective cohort studies is a public health priority. The study findings, together with evidence of the carcinogenic (IARC, 2004), developmental (Wasserman et al., 2004), and metabolic (Navas-Acien et al., 2008, 2009) effects of arsenic, reinforce the need to implement arsenic remediation treatments in water supply systems that are above the current WHO safety standard of 10 mg/L (Aragone´s-Sanz et al., 2001; Blanco Herna´ndez et al., 1998; Rodrı´guez et al., 2006).
Acknowledgments The authors are grateful to the Spanish Ministry of Health and Social Policy for providing the data on water composition and to the Spanish National Institute for Statistics and Mrs. Margarita Garcı´a Ferruelo for providing mortality files. References Aragone´s-Sanz, N., Palacios Dı´ez, M., Avello de Miguel, A., Go´mez Rodrı´guez, P., Martı´nez Corte´s, M., Rodrı´guez Bernabeu, M.J., 2001. Nivel de arse´nico en abastecimientos de agua de consumo de origen subterra´neo en la Comunidad de Madrid. Rev. Esp. Salud Pu´blica 75, 421–432. ˜ ol de Cre´dito, 1993. In: Anuario del Mercado Espan ˜ ol. Banco Espan ˜ ol Banco Espan de Cre´dito, Madrid. Bhatnagar, A., 2006. Environmental cardiology. Studying mechanistic links between pollution and heart disease. Circ. Res. 99, 692–705. Blanco Herna´ndez, A.L., Alonso Gutie´rrez, D., Jime´nez de Blas, O., Santiago Guervo´s, M., de Miguel Manzano, B., 1998. Estudio de los niveles de plomo, cadmio, zinc y arse´nico, en aguas de la provincia de Salamanca. Rev. Esp. Salud Pu´blica 72, 53–65. Breslow, N.E., Clayton, D.G., 1993. Approximate inference in generalized linear mixed models. J. Am. Stat. Assoc. 88, 9–25. Bunderson, M., Brooks, D.M., Walker, D.L., Rosenfeld, M.E., Coffin, J.D., Beall, H.D., 2004. Arsenic exposure exacerbates atherosclerotic plaque formation and increases nitrotyrosine and leukotriene biosynthesis. Toxicol. Appl. Pharmacol. 201, 32–39. Chen, C.-J., Hsueh, Y.-M., Lai, M.-S., Shyu, M.-P., Chen, S.-Y., Wu, M.-M., Kuo, T.-L., Tai, T.-Y., 1995. Increased prevalence of hypertension and long-term arsenic exposure. Hypertension 25, 53–60. Chen, C.-J., Chiou, H.-Y., Chiang, M.-H., Lin, L.-J., Tai, T.-Y., 1996. Dose-response relationship between ischaemic heart disease mortality and long-term arsenic exposure. Arterioscler. Thromb. Vasc. Biol. 16, 504–510.
ARTICLE IN PRESS 454
M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
Chen, C.-J., Wu, M.-M., Lee, S.-S., Wang, J.-D., Cheng, S.-H., Wu, H.-Y., 1998. Atherogenicity and carcinogenicity of high-arsenic artesian well water. Multiple risk-factors and related malignant neoplasms of blackfoot disease. Arteriosclerosis 8, 452–460. Chen, Y., Factor-Litvak, P., Howe, G.R., Graziano, J.H., Brandt-Rauf, P., Parvez, F., van Geen, A., Ahsan, H., 2007. 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. Am. J. Epidemiol. 165, 541–552. Chiou, H.-Y., Huang, W.-I., Hsu, C.-L., Chang, S.-F., Hsu, Y.-H., Chen, C.-J., 1997. Doseresponse relationship between prevalence of cerebrovascular disease and ingested inorganic arsenic. Stroke 28, 1717–1723. Coronado-Gonza´lez, J.A., del Razo, L.M., Garcı´a-Vargas, G., Sanmiguel-Salazar, F., ˜ a, J., 2007. Inorganic arsenic exposure and type 2 diabetes Escobedo-de la Pen mellitus in Mexico. Environ. Res. 104, 383–389. Engel, R., Smith, A., 1994. Arsenic in drinking water and mortality from vascular disease: an ecologic analysis in 30 counties in the United States. Arch. Environ. Health 49, 418–427. ˜ o Fonseca, I., Gonza´lez Parama´s, Garcı´a-Villanova, R.J., Rodrı´guez Torrero, P., Torun A.M., Go´mez Ba´rez, J.A., Cordo´n Marcos, C., 2005. Arsenic in ground drinking water. The regional monitoring plan of Castilla y Leo´n (Spain). Eur. J. Water Qual. 36, 107–117. Gelman, A., Hill, J., 2007. In: Data Analysis using Regression and Multilevel/ Hierarchical Models. Cambridge University Press, Cambridge. Greenland, S., 1995. Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology 6, 356–365. Greenland, S., Rothman, K.J., 2008. Introduction to categorical statistics. In: Rothman, K.J., Greenland, S., Lash, T.L. (Eds.), Modern Epidemiology third ed Wolters Kluwer/Lippincott Williams & Wilkins, Philadelphia. IARC Working Group on the Evaluation of Carcinogenic Risks to Humans, 2004. In: Some Drinking-Water Disinfectants and Contaminants, Including Arsenic. IARC Press, Lyon, France. Instituto Nacional de Estadı´stica, 2004. In: Estadı´stica de Establecimientos Sanitarios con Re´gimen de Internado 2001. Instituto Nacional de Estadı´stica, Madrid. Lai, M.-S., Hsueh, Y.-M., Chen, C.-J., Shyu, M.-P., Chen, S.-Y., Kuo, T.-L., Wu, M.-M., Tai, T.-Y., 1994. Ingested inorganic arsenic and prevalence of diabetes mellitus. Am. J. Epidemiol. 139, 484–492. Meliker, J.R., Wahl, R.L., Cameron, L.L., Nriagu, J.O., 2007. Arsenic in drinking water and cerebrovascular disease, diabetes mellitus, and kidney disease in Michigan: a standardized mortality ratio analysis. Environ. Health 6, 4. ˜ a. Ministerio de Sanidad y Consumo, 1995. In: Encuesta Nacional de Salud de Espan Ministerio de Sanidad y Consumo, Madrid. Ministerio de Sanidad y Consumo, 2000. In: Calidad del agua de Consumo en ˜ a. 1er Informe Nacional. An ˜ o 1993/1995. Ministerio de Sanidad y Espan Consumo, Madrid, pp. 47–48. National Research Council, 1999. Arsenic in Drinking Water. Subcommittee on Arsenic in Drinking Water. National Academy Press, Washington, DC. National Research Council, 2001. Arsenic in Drinking Water 2001 Update. Subcommittee on Arsenic in Drinking Water. National Academy Press, Washington, DC. Navas-Acien, A., Sharrett, A.R., Silbergeld, E.K., Schwartz, B.S., Nachman, K.E., Burke, T.A., 2005. Arsenic exposure and cardiovascular disease: a systematic review of the epidemiologic evidence. Am. J. Epidemiol. 162, 1037–1049. Navas-Acien, A., Silbergeld, E.K., Pastor-Barriuso, R., Guallar, E., 2008. Arsenic exposure and prevalence of type 2 diabetes in US adults. JAMA 300, 814–822. Navas-Acien, A., Silbergeld, E.K., Pastor-Barriuso, R., Guallar, E., 2009. Arsenic exposure and prevalence of type 2 diabetes: updated findings from the National Health Nutrition and Examination Survey, 2003–2006. Epidemiology101097/EDE.0b013e3181afef88.
Rahman, M., Tondel, M., Ahmad, S.A., Axelson, O., 1998. Diabetes mellitus associated with arsenic exposure in Bangladesh. Am. J. Epidemiol. 148, 198–203. Rahman, M., Tondel, M., Ahmad, S.A., Chowdhury, I.A., Faruquee, M.H., Axelson, O., 1999. Hypertension and arsenic exposure in Bangladesh. Hypertension 33, 74–78. Rodrı´guez, V., Morote, P., Arbesu´, E., Alvarez, P., Esteban, A.I., Suarez, P., 2006. Enfermedades relacionadas con la ingesta cro´nica de arse´nico a trave´s de aguas de consumo. Gac. Sanit. 20 (Espec Congr), 48. Sahu´n, B., Go´mez, J.J., Lillo, T., del Olmo, P., 2004. Arse´nico en aguas subterra´neas e interaccio´n agua-roca: un ejemplo en la cuenca terciaria del Duero (Castilla y ˜ a). Rev. Soc. Geol. Espan ˜ a 17, 137–155. Leo´n, Espan Simeonova, P.P., Luster, M.I., 2004. Arsenic and atherosclerosis. Toxicol. Appl. Pharmacol. 198, 444–449. Smith, A.H., Lopipero, P.A., Bates, M.N., Steinmaus, C.M., 2002. Arsenic epidemiology and drinking water standards. Science 296, 2145–2146. Soucy, N.V., Mayka, D., Klei, L.R., Nemec, A.A., Bauer, J.A., Barchowsky, A., 2005. Neovascularization and angiogenic gene expression following chronic arsenic exposure in mice. Cardiovasc. Toxicol. 5, 9–41. States, J.C., Srivastava, S., Chen, Y., Barchowsky, A., 2009. Arsenic and cardiovascular disease. Toxicol. Sci. 107, 312–323. Straub, A.C., Stolz, D.B., Vin, H., Ross, M.A., Soucy, N.V., Klei, L.R., Barchowsky, A., 2007. Low level arsenic promotes progressive inflammatory angiogenesis and liver blood vessel remodeling in mice. Toxicol. Appl. Pharmacol. 222, 327–336. Straub, A.C., Clark, K.A., Ross, M.A., Chandra, A.G., Li, S., Gao, X., Pagano, P.J., Stolz, D.B., Barchowsky, A., 2008. Arsenic-stimulated liver sinusoidal capillarization in mice requires NADPH oxidase-generated superoxide. J. Clin. Invest. 118, 3980–3989. Tseng, C.H., Huang, Y.K., Huang, Y.L., Chung, C.J., Yang, M.H., Chen, C.J., Hsueh, Y.M., 2005. Arsenic exposure, urinary arsenic speciation, and peripheral vascular disease in blackfoot disease-hyperendemic villages in Taiwan. Toxicol. Appl. Pharmacol. 206, 299–308. Tseng, C.H., 2008. Cardiovascular disease in arsenic-exposed subjects living in the arseniasis-hyperendemic areas in Taiwan. Atherosclerosis 199, 12–18. Varela, G., Moreiras, O., Carbajal, A., Campo, M., 1995. In: Encuesta de Presupuestos Familiares. Estudio Nacional de Nutricio´n y Alimentacio´n. Instituto Nacional de Estadı´stica, Madrid. Wang, C.H., Jeng, J.S., Yip, P.K., Chen, C.L., Hsu, L.I., Hsueh, Y.M., Chiou, H.Y., Wu, M.M., Chen, C.J., 2002. Biological gradient between long-term arsenic exposure and carotid atherosclerosis. Circulation 105, 1804–1809. Wang, C.H., Hsiao, C.K., Chen, C.L., Hsu, L.I., Chiou, H.Y., Chen, S.Y., Hsueh, Y.M., Wu, M.M., Chen, C.J., 2007. A review of the epidemiologic literature on the role of environmental arsenic exposure and cardiovascular diseases. Toxicol. Appl. Pharmacol. 222, 315–326. Wasserman, G.A., Liu, X., Parvez, F., Ahsan, H., Factor-Litvak, P., van Geen, A., Slavkovich, V., LoIacono, N.J., Cheng, Z., Hussain, I., Momotaj, H., Graziano, J.H., 2004. Water arsenic exposure and children’s intellectual function in Araihazar, Bangladesh. Environ. Health Perspect. 112, 1329–1333. World Health Organization, 2003. In: Guidelines for Drinking-Water Quality third ed. World Health Organization, Geneva. Wu, M.-M., Kuo, T.-L., Hwang, Y.-H., Chen, C.-J., 1989. Dose-response relation between arsenic concentration in well water and mortality from cancers and vascular disease. Am. J. Epidemiol. 130, 1123–1132. Yuan, Y., Marshall, G., Ferreccio, C., Steinmaus, C., Selvin, S., Liaw, J., Bates, M.N., Smith, A.H., 2007. Acute myocardial infarction mortality in comparision with lung and bladder cancer mortality in arsenic-exposed region II of Chile from 1950 to 2000. Am. J. Epidemiol. 166, 1381–1391. Zierold, K.M., Knobeloch, L., Anderson, H., 2004. Prevalence of chronic diseases in adults exposed to arsenic-contaminated drinking water. Am. J. Public Health 94, 1936–1937.
Contents lists available at ScienceDirect
Environmental Research journal homepage: www.elsevier.com/locate/envres
Arsenic in public water supplies and cardiovascular mortality in Spain$, $$ Ma Jose´ Medrano a,, Raquel Boix a, Roberto Pastor-Barriuso a, Margarita Palau b, Javier Damia´n a, Rebeca Ramis a,c, Jose´ Luis del Barrio d, Ana Navas-Acien e,f a
Centro Nacional de Epidemiologı´a, Instituto de Salud Carlos III, Sinesio Delgado 6, 28029 Madrid, Spain ´n General de Sanidad Ambiental y Salud Laboral, Direccio ´n General de Salud Pu ´ blica y Sanidad Exterior, Ministerio de Sanidad y Polı´tica Social, Madrid, Spain Subdireccio c ´ blica (CIBERESP), Madrid, Spain CIBER en Epidemiologı´a y Salud Pu d ´ blica, Universidad Rey Juan Carlos, Madrid, Spain Departamento de Salud Pu e Department of Environmental Health Sciences, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA f Department of Epidemiology, Welch Center for Prevention, Epidemiology and Clinical Research, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA b
a r t i c l e in f o
a b s t r a c t
Article history: Received 2 June 2009 Received in revised form 22 September 2009 Accepted 1 October 2009 Available online 31 October 2009
Background: High-chronic arsenic exposure in drinking water is associated with increased cardiovascular disease risk. At low-chronic levels, as those present in Spain, evidence is scarce. In this ecological study, we evaluated the association of municipal drinking water arsenic concentrations during the period 1998 2002 with cardiovascular mortality in the population of Spain. Methods: Arsenic concentrations in drinking water were available for 1721 municipalities, covering 24.8 million people. Standardized mortality ratios (SMRs) for cardiovascular (361,750 deaths), coronary (113,000 deaths), and cerebrovascular (103,590 deaths) disease were analyzed for the period 1999 2003. Two-level hierarchical Poisson models were used to evaluate the association of municipal drinking water arsenic concentrations with mortality adjusting for social determinants, cardiovascular risk factors, diet, and water characteristics at municipal or provincial level in 651 municipalities (200,376 cardiovascular deaths) with complete covariate information. Results: Mean municipal drinking water arsenic concentrations ranged from o 1 to 118 mg/L. Compared to the overall Spanish population, sex- and age-adjusted mortality rates for cardiovascular (SMR 1.10), coronary (SMR 1.18), and cerebrovascular (SMR 1.04) disease were increased in municipalities with arsenic concentrations in drinking water 4 10 mg/L. Compared to municipalities with arsenic concentrations o 1 mg/L, fully adjusted cardiovascular mortality rates were increased by 2.2% ( 0.9% to 5.5%) and 2.6% ( 2.0% to 7.5%) in municipalities with arsenic concentrations between 1 10 and 4 10 mg/L, respectively (P-value for trend 0.032). The corresponding figures were 5.2% (0.8% to 9.8%) and 1.5% ( 4.5% to 7.9%) for coronary heart disease mortality, and 0.3% ( 4.1% to 4.9%) and 1.7% ( 4.9% to 8.8%) for cerebrovascular disease mortality. Conclusions: In this ecological study, elevated low-to-moderate arsenic concentrations in drinking water were associated with increased cardiovascular mortality at the municipal level. Prospective cohort studies with individual measures of arsenic exposure, standardized cardiovascular outcomes, and adequate adjustment for confounders are needed to confirm these ecological findings. Our study, however, reinforces the need to implement arsenic remediation treatments in water supply systems above the World Health Organization safety standard of 10 mg/L. & 2009 Elsevier Inc. All rights reserved.
Keywords: Arsenic Drinking water Cardiovascular diseases Mortality
1. Introduction
$ Funding Sources: Research Grant no. PI060656 from the Spanish Fondo de Investigacio´n Sanitaria. Ana Navas-Acien was supported by Grant 1R01HL090863 from the US National Heart Lung and Blood Institute. $$ The study was approved by the Research Committee of the Carlos III Institute of Health. The study complies with the confidentiality rules established by the Spanish National Institute for Statistics (data transfer protocol 24/11/06). Corresponding author. Fax: + 34 91 387 7815. E-mail address: [email protected] (M.J. Medrano).
0013-9351/$ - see front matter & 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.envres.2009.10.002
The vascular effects of high-chronic arsenic exposure in drinking water ( 4100 mg/L) are supported by experimental (States et al., 2009; Simeonova and Luster, 2004; Soucy et al., 2005) and epidemiological evidence from Taiwan (Tseng, 2008; Wang et al., 2007) and Chile (Yuan et al., 2007). Long-term arsenic exposure in Taiwan has been long recognized as the cause of a severe form of peripheral arterial disease called black foot disease (Tseng et al., 2005; Chen et al., 1998) and it has been associated with coronary and cerebrovascular diseases and with subclinical markers of
ARTICLE IN PRESS M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
atherosclerosis (Wu et al., 1989; Chen et al.,1996; Chiou et al., 1997; Wang et al., 2002). In Region II Chile, acute myocardial infarction mortality increased following a period of high exposure to arsenic in drinking water and decreased after arsenic remediation had been implemented (Yuan et al., 2007). In Bangladesh, Mexico and Taiwan, high arsenic exposure in drinking water has been related with cardiovascular risk factors, such as hypertension (Chen et al., 1995, 2007; Rahman et al., 1999) and diabetes (Coronado-Gonza´lez et al., 2007; Rahman et al., 1998; Lai et al., 1994). At moderate or low chronic inorganic arsenic exposure in drinking water ( o100 mg/L), levels that are relevant for most regions in North America, Europe, and other parts of the world, few epidemiologic studies have addressed the association of inorganic arsenic with cardiovascular endpoints (Navas-Acien et al., 2005). Mortality for diseases of the arteries, arterioles, and capillaries were increased in counties with arsenic levels in drinking water 420 mg/L vs. 510 mg/L but not with coronary heart disease and cerebrovascular mortality in a 30-county US study (Engel and Smith, 1994). In Michigan, mortality was increased for all diseases of the circulatory system and cerebrovascular diseases but not for ischemic heart disease in 6 counties with drinking water arsenic 410 mg/L compared to the overall State (Meliker et al., 2007). In a cross-sectional study conducted among residents with private wells in Wisconsin, self-reported circulatory problems including coronary bypass and myocardial infarction were increased comparing arsenic in drinking water 10 410 vs. o2 mg/L (Zierold et al., 2004). In Spain, arsenic concentrations in public water supplies are monitored through a water quality surveillance system that compiles municipal water information since 1992. While arsenic concentrations in drinking water were then generally below 50 mg/L (Ministerio de Sanidad y Consumo, 2000), a relatively important number of communities had concentrations above 10 mg/L, the safety standard for drinking water arsenic that was regulated in 2002. In this study, our goal was to evaluate the association of municipal tap drinking water arsenic concentrations during 1998 2002 with cardiovascular mortality in the population of Spain.
2. Methods 2.1. Study design and population For this ecological study, municipal-level arsenic concentrations in drinking water and mortality data were available for 1721 municipalities located in 49 out of 52 Spanish provinces, covering 24.8 million people (60.7% of the total Spanish population in 2001) (Fig. 1). Non-respondent municipalities (municipalities with no arsenic data during the study period) were located in large low-populated geographical areas, where water control procedures for arsenic and other metals were not fully established in 1998 2002. Compared to respondent municipalities, non-respondent municipalities were less populated (mean 1527 vs. 14,400 inhabitants), older (24.2% vs. 19.1% inhabitants Z 65 years of age), more rural (99.4% vs. 95.5%), and less well-off (5.0 vs. 6.0 in 1 10 ordinal income scale). The study was approved by the Research Committee of the Carlos III Institute of Health and complies with the Spanish National Institute for Statistics confidentiality rules.
2.2. Arsenic in public water supply systems Tap drinking water arsenic concentrations at the municipal level during 1998 2002 were obtained from the National Information System of Consume Water Control, held by the Spanish Ministry of Health (Ministerio de Sanidad y Consumo, 2000). This administrative database was established to track drinking water quality by gathering annual data from local public health authorities responsible for water quality control, including tap water arsenic information from every public water supply system. For this study, a total of 20,074 arsenic determinations were available. The mean (SD) number of arsenic determinations per municipality was 11.7 (41.1), ranging from 1 to 802.
449
> 50 µg/L 10−50 µg/L 1−10 − µg/L < 1 µg/L
Fig. 1. Geographical distribution of arsenic concentrations in municipal drinking water in Spain, 1998–2002. Blank areas represent non-respondent municipalities. Black lines represent provincial boundaries. Arsenic concentrations in municipal water were measured at regional laboratories participating in the System of Consume Water Control, all of them holding accreditation and/or certification following international rules for quality control (UNE-EN ISO/IEC 17025, UNE-EN ISO 9001). Laboratory procedures for sample management, analytical methods, and quality control measures (accuracy, precision, and detection limits) were standardized by Spanish and European Union laws (Royal Decrees 1138/90, 140/03, and 1423/82; EU Directive 98/83/CE). Following these protocols, arsenic concentrations were measured using atomic absorption spectrophotometry in the participant laboratories. The highest detection limit permitted by regulation was 1 mg/L, although the actual detection limit for each laboratory was not available. A total of 1066 (61.9%) municipalities reported all arsenic determinations as undetected and were replaced by a level equal to 1 mg/L divided by the square root of 2. For 270 (15.7%) municipalities with samples analyzed in laboratories with detection limits lower than 1 mg/L, the actual detected values below 1 mg/L were included in the analyses.
2.3. Cardiovascular mortality data Cardiovascular mortality was analyzed for the period 1999 2003 to allow for 1-year delay with respect to arsenic determinations. The 1-year delay was selected based on short-term changes in myocardial infarction rates with increasing and decreasing arsenic levels in drinking water in Region II, Chile (Yuan et al., 2007). Adding a 3-year delay or no delay resulted in similar findings (results not shown). The observed number of deaths from cardiovascular diseases (codes I00 I99 of the tenth revision of the International Classification of Diseases), coronary heart diseases (codes I20 I25), and cerebrovascular diseases (codes I60 I69) at the municipal level for men and women was obtained from the National Institute for Statistics. A total of 361,750 cardiovascular disease deaths were included in the analysis. The number of coronary and cerebrovascular disease deaths was 113,000 and 103,590, respectively. The sex- and age-adjusted cardiovascular mortality rate in respondent municipalities was similar to the overall national rate (standardized mortality ratio [SMR] of 0.99).
2.4. Other relevant variables Municipal-level water characteristics (hardness and magnesium content, pH, and temperature) were obtained from the National Information System of Consume Water Control. Municipal-level socioeconomic indicators (per capita income measured with an ordinal scale from 1 to 10, with 10 reflecting higher ˜ ol de Cre´dito, 1993). income) were obtained from bank statistics (Banco Espan Information for other relevant sociodemographic, dietary, and cardiovascular risk factors was considered at the provincial level because they were not available at the municipal level. Health care access (number of hospital beds per 1000 inhabitants) was obtained from 2001 official statistics (Instituto Nacional de Estadı´stica, 2004); dietary consumption of fish, wine, olive oil, bottled water, and total energy-adjusted folate was obtained from the 1991 National Nutrition Study (Varela et al., 1995); and sex- and age-specific prevalences of smoking, hypertension, high serum cholesterol, diabetes, overweight/obesity, and low physical activity were obtained from the 1995 National Health Survey (Ministerio de Sanidad y Consumo, 1995). Complete information for covariates was only
ARTICLE IN PRESS 450
M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
available in 651 (37.8%) municipalities corresponding to 200,376 (55.4%) cardiovascular disease deaths and 14.4 million (58.1%) population size.
2.5. Statistical analysis For each municipality, all available arsenic determinations in drinking water during the period 1998 2002 were averaged as an approximation to long-term arsenic exposure. Municipalities were classified according to their mean arsenic concentrations as o1, 1 10, and 410 mg/L. Mean levels of sociodemographic and water characteristics, as well as of dietary and cardiovascular risk factors, were compared across categories of municipal arsenic concentration using linear regression models. For each municipality, the expected number of cardiovascular, coronary, and cerebrovascular disease deaths among men and women for the period 1999 2003 was calculated by applying national sex- and age-specific mortality rates in 2001 to the estimated number of person-years by sex and age in each municipality over the period 1999 2003. Assuming a Poisson distribution for the observed number of deaths (Greenland and Rothman, 2008), SMRs and their 95% confidence intervals (CIs) were computed comparing observed to expected number of deaths within each category of municipal arsenic concentration. For risk assessment, the association of municipal arsenic concentration in drinking water with cardiovascular, coronary, and cerebrovascular disease mortality was evaluated using two-level hierarchical Poisson models with random intercept (Gelman and Hill, 2007). More specifically, at the first municipal level, the observed number of deaths Oij in municipality i of province j was assumed to follow the log-linear Poisson model
3. Results Mean municipal arsenic concentrations in drinking water ranged from o1 to 118 mg/L, with 7 villages (joint population 10,613 inhabitants) registering mean arsenic concentrations 450 mg/L (Fig. 1), the Spanish safety standard for arsenic in drinking water during the study period. Differences in sociodemographic, dietary, and cardiovascular risk factors across water arsenic categories were generally small but statistically significant (Table 1). Compared to municipalities with arsenic concentrations o1 mg/L, municipalities with concentrations 410 mg/L tended to be younger, had higher levels of water hardness and pH, and were located in provinces with less hospital beds per 1000 inhabitants, higher sex- and age-adjusted prevalence of cardiovascular risk factors, and lower intake of fish, wine, olive oil, and total energyadjusted folate. Compared to the overall Spanish population, sex- and ageadjusted mortality rates for cardiovascular disease (SMR 1.10, 95% CI 1.08–1.12), coronary heart disease (SMR 1.18, 95% CI 1.15–1.22), and cerebrovascular disease (SMR 1.04, 95% CI 1.01–1.08) were increased in the 89 municipalities with mean arsenic concentrations in drinking water 410 mg/L, with no substantial differences by sex (Table 2). No increased cardiovascular, coronary, or
Oij PoissonfEij expðaj þ bxij þ d0 zij Þg; where Eij is the expected number of deaths under national sex- and age-specific mortality rates, aj is the province-specific intercept, and b and d are the coefficients associated to the log-transformed municipal arsenic concentration xij and other municipal-level adjustment factors zij, whose effects were constrained to be the same for all provinces. At the second provincial level, the intercept aj was allowed to vary randomly across provinces according to a normal distribution with mean depending on provincial-level covariates wj and with common betweenprovince variance t2,
aj Nðg0 þ c0 wj ; t2 Þ; where g0 is the overall intercept and c are the coefficients associated with provincial-level adjustment factors wj. In addition to the fixed effects for municipal- and provincial-level explanatory variables, the model includes a random intercept term that accounts for unexplained provincial-level variations in the expected number of deaths. The model was fitted using penalized quasilikelihood methods (Breslow and Clayton, 1993), and the resulting regression coefficients were interpreted in terms of percentage change in mortality by applying the transformation 100{exp(b) 1}. To explore the association between arsenic concentration in drinking water and cardiovascular mortality endpoints without assuming a log-linear doseresponse relation, we also estimated percentage changes in mortality and their 95% CIs comparing municipalities in the highest arsenic categories (1 10 and 410 mg/L) to those in the lowest one (o 1 mg/L), as well as using restricted quadratic splines for log-transformed municipal arsenic concentrations with knots at 0.5, 2, and 10 mg/L, which corresponded to the 21st, 67th and 88th percentiles of the arsenic distribution in the 651 municipalities with complete covariate information (Greenland, 1995). Selecting knots at different percentiles resulted in similar dose-response relations (results not shown). The above models were fitted with increasing levels of adjustment. First, we adjusted for sex, age, per capita municipal income, and health care access at provincial level (hospital beds per 1000 inhabitants). Second, we further adjusted for sex- and age-adjusted prevalence of cardiovascular risk factors at provincial level (smoking, hypertension, high serum cholesterol, diabetes, overweight/obesity, and low physical activity). Third, we further adjusted for dietary factors at provincial level (fish, wine, olive oil, bottled water, and total energy-adjusted folate) and water characteristics at municipal level (hardness and magnesium content, pH, and temperature). Because cardiovascular mortality rates in Spain are much higher in men than women, we also performed separate analyses by sex. Sensitivity analyses were conducted to assess the potential impact of data quality in a subset of 356 municipalities (covering 16.4 million people and 232,074 cardiovascular disease deaths) with at least 8 arsenic determinations (mean of 46.7 determinations per municipality), yielding similar results to those obtained in the whole sample (not shown). All analyses were performed in R statistical software (R Foundation for Statistical Computing, Vienna, Austria), using the glmmPQL function in MASS package to fit random-intercept hierarchical Poisson models. The spatial distribution of municipal arsenic concentrations was depicted using MapInfo (MapInfo Corporation, New York).
Table 1 Characteristics of study municipalities by arsenic concentration in drinking water. Characteristic
Total
Arsenic category (lg/L) o1
1 10 410
Pvaluea
No. of municipalities
1721
1336
296
89
Sociodemographic characteristics Total population (thousands) Mean population (thousands) Rural (%) Z 65 years (%) Per capita incomeb Hospital beds per 1000 inhabitants
24,793 14.4 95.5 19.1 6.0 9.1
18,978 14.2 95.9 19.9 6.0 9.2
4803 16.2 93.6 16.2 5.9 8.8
1011 11.4 96.6 16.5 5.7 8.4
0.89 0.19 o 0.001 0.24 o 0.001
Cardiovascular risk factorsc Smoking (%) Hypertension (%) High serum cholesterol (%) Diabetes (%) Obesity (%) Low physical activity (%)
27.5 17.1 12.1 6.3 11.5 61.2
27.1 16.7 11.8 6.2 11.2 60.9
28.4 18.1 13.1 6.7 12.1 60.8
29.7 19.9 13.4 7.2 13.3 66.6
o 0.001 o 0.001 o 0.001 o 0.001 o 0.001 o 0.001
Dietary factorsd Fish (g/person-day) Wine (g/person-day) Olive oil (g/person-day) Bottled water (g/person-day) Total energy-adjusted folatee
73.7 63.9 34.5 20.8 3.5
73.8 63.6 35.3 20.8 3.6
75.6 70.1 33.1 20.0 3.4
65.6 48.1 27.4 24.3 2.7
o 0.001 o 0.001 o 0.001 0.22 o 0.001
Water characteristicsf Total no of arsenic determinations Mean no of arsenic determinations Arsenic (mg/L) Hardness (mg/L) Magnesium (mg/L) pH Temperature (1C)
20,074 11.7 2.4 75.1 23.9 7.2 16.8
14,796 11.1 0.7 58.6 25.9 7.2 16.6
3890 13.1 3.9 137.3 17.4 6.8 16.7
1388 15.6 23.3 115.0 17.3 7.6 17.9
0.48 o 0.001 0.56 o 0.001 0.10
a P-value for heterogeneity of means across categories of municipal arsenic concentration. b Ordinal variable from 1 (lowest income) to 10 (highest income). c Sex- and age-adjusted prevalence of cardiovascular risk factors according to the national population distribution in 2001. d Food content of family diet. e Ordinal variable from 1 (lowest quintile of total energy-adjusted folate) to 5 (highest quintile). f Measurements of hardness, magnesium, pH, and temperature were available in 1608, 1576, 973, and 773 municipalities, respectively.
ARTICLE IN PRESS M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
451
Table 2 Standardized mortality ratio (95% confidence interval) for cardiovascular, coronary, and cerebrovascular disease by municipal arsenic concentration in drinking water.a Arsenic category (lg/L)
Total
Men
Women
No. of deaths
SMR (95% CI)b
No. of deaths
SMR (95% CI)D
No. of deaths
SMR (95% CI)D
285,049 62,739 13,962 361,750
0.98 (0.976–0.983) 0.99 (0.985–1.000) 1.10 (1.078–1.115) 0.99 (0.983–0.989)
128,504 28,214 6567 163,285
1.00 (0.993–1.004) 0.98 (0.964–0.987) 1.10 (1.071–1.124) 1.00 (0.993–1.003)
156,545 34,525 7395 198,465
0.97 (0.960–0.970) 1.01 (0.996–1.018) 1.10 (1.071–1.121) 0.98 (0.972–0.981)
Coronary heart disease mortality o1 88,566 1 10 19,709 410 4725 Total 113,000
0.99 (0.980–0.993) 1.00 (0.984–1.011) 1.18 (1.148–1.216) 0.99 (0.989–1.001)
50,121 11,045 2734 63,900
0.99 (0.985–1.003) 0.97 (0.950–0.986) 1.16 (1.119–1.206) 1.00 (0.988–1.003)
38,445 8664 1991 49,100
0.98 1.04 1.21 0.99
(0.966–0.986) (1.017–1.061) (1.157–1.264) (0.985–1.003)
Cerebrovascular disease mortality o1 81,368 1 10 18,327 410 3895 Total 103,590
0.95 0.99 1.04 0.96
33,321 7464 1623 42,408
0.97 0.98 1.02 0.98
48,047 10,863 2272 61,182
0.93 1.00 1.06 0.95
(0.923–0.940) (0.978–1.015) (1.017–1.104) (0.939–0.954)
Cardiovascular mortality o1 1 10 410 Total
a b
(0.941–0.954) (0.974–1.003) (1.012–1.077) (0.952–0.964)
(0.962–0.983) (0.954–0.999) (0.975–1.074) (0.966–0.984)
Analyses based on all 1721 municipalities with available arsenic concentrations in drinking water. Standardized mortality ratio (95% confidence interval) comparing observed to expected number of deaths under national sex- and age-specific mortality rates.
Table 3 Percentage change (95% confidence interval) in cardiovascular, coronary, and cerebrovascular disease mortality by municipal arsenic concentration in drinking water.a 2-Fold increase in arsenic concentration
P-value for trendb
Arsenic category (lg/L) o1
1 10
410
No. of municipalities
651
383
193
75
Cardiovascular mortality No. of male/female deaths Model 1—total Model 2—total Model 3—total Model 3—men Model 3—women
90,195/110,181 0.7 ( 0.1 to 1.6) 0.7 ( 0.1 to 1.5) 0.8 (0.1 to 1.6) 0.7 ( 0.1 to 1.5) 1.0 (0.0 to 1.9)
65,723/80,844 0 (reference) 0 (reference) 0 (reference) 0 (reference) 0 (reference)
18,844/23,113 2.0 ( 1.2 to 5.4) 1.4 ( 1.9 to 4.7) 2.2 ( 0.9 to 5.5) 0.4 ( 2.8 to 3.8) 4.0 (0.2 to 8.0)
5628/6224 1.8 ( 2.9 to 6.8) 1.7 ( 3.0 to 6.7) 2.6 ( 2.0 to 7.5) 2.4 ( 2.5 to 7.5) 3.2 ( 2.5 to 9.1)
0.080 0.090 0.032 0.095 0.048
Coronary heart disease mortality No. of male/female deaths Model 3—total Model 3—men Model 3—women
35,862/27,595 0.9 ( 0.1 to 2.0) 0.6 ( 0.6 to 1.8) 1.3 ( 0.2 to 2.7)
26,117/20,065 0 (reference) 0 (reference) 0 (reference)
7394/5819 5.2 (0.8 to 9.8) 0.4 ( 4.4 to 5.4) 11.3 (5.1 to 17.8)
2351/1711 1.5 ( 4.5 to 7.9) 1.5 ( 5.3 to 8.8) 3.3 ( 4.8 to 12.2)
0.091 0.31 0.084
Cerebrovascular disease mortality No. of male/female deaths Model 3—total Model 3—men Model 3—women
22,041/31,985 0.7 ( 0.4 to 1.8) 0.7 ( 0.7 to 2.1) 0.7 ( 0.6 to 2.0)
15,831/23,122 0 (reference) 0 (reference) 0 (reference)
4850/7012 0.3 ( 4.1 to 4.9) 0.1 ( 5.5 to 5.7) 0.8 ( 4.5 to 6.3)
1360/1851 1.7 ( 4.9 to 8.8) 1.5 ( 6.6 to 10.3) 1.4 ( 6.3 to 9.8)
0.20 0.34 0.31
Model 1 adjusted for sex, age, per capita municipal income, and hospital beds per population at provincial level. Model 2 further adjusted for prevalence of cardiovascular risk factors at provincial level (smoking, hypertension, high serum cholesterol, diabetes, overweight/obesity, and low physical activity). Model 3 further adjusted for dietary factors at provincial level (fish, wine, olive oil, bottled water, and total energy-adjusted folate) and water characteristics at municipal level (hardness, magnesium, pH, and temperature). a b
Analyses restricted to 651 municipalities (14.4 million people) with complete information on adjustment factors. P-value for trend based on log-transformed arsenic concentrations.
cerebrovascular mortality was observed for the category of arsenic 1 10 mg/L, except for coronary heart disease mortality among women (SMR 1.04, 95% CI 1.02–1.06). In multivariable models adjusted for social determinants, cardiovascular risk factors, diet, and water characteristics, cardiovascular mortality increased 0.8% (95% CI 0.1–1.6%) for each doubling in arsenic concentrations (Table 3, model 3). Compared to municipalities with arsenic concentrations o1 mg/L, cardiovascular disease mortality was increased by 2.2% (95% CI 0.9% to 5.5%) in municipalities with arsenic concentrations between 1 and 10 mg/L and by 2.6% (95% CI 2.0% to 7.5%) in municipalities with arsenic concentrations 410 mg/L, with
consistent findings by sex, except a stronger association for the arsenic 1 10 mg/L category among women. Increased cardiovascular mortality with increasing arsenic concentrations in municipal drinking water was also observed using restricted quadratic splines (Figs. 2 and 3). For coronary heart disease mortality, although the number of events was smaller, fully adjusted relative risks were consistent with increased risk, particularly among women. For cerebrovascular disease mortality, the association was weaker and not statistically significant. We further evaluated whether increased mortality in municipalities with arsenic concentrations in drinking water between 1
ARTICLE IN PRESS M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
30 5 20 0 10 -5
0 0.2 0.5 1 2 5 10 20 Arsenic concentration in drinking water (µg/L)
Fig. 2. Percentage change in cardiovascular mortality by municipal arsenic concentration in drinking water. Curves represent the adjusted percentage change in cardiovascular mortality (solid curve) and its 95% confidence interval (dashed curves) based on restricted quadratic splines for log-transformed arsenic concentrations with knots at 0.5, 2, and 10 mg/L. The reference value (percentage change= 0) was set at 0.5 mg/L. Percentage changes were adjusted for sex, age, per capita municipal income, hospital beds per population at provincial level, prevalence of cardiovascular risk factors at provincial level (smoking, hypertension, high serum cholesterol, diabetes, overweight/obesity, and low physical activity), dietary factors at provincial level (fish, wine, olive oil, bottled water, and total energy-adjusted folate), and water characteristics at municipal level (hardness, magnesium, pH, and temperature). Bars represent the histogram of municipal arsenic distribution.
and 10 mg/L could be influenced by 16 municipalities that were geographically located in the arsenic contaminated area of the aquifer Los Arenales (Sahu´n et al., 2004; Garcı´a-Villanova et al., 2005), where inhabitants may drink from private uncontrolled wells with higher arsenic concentrations. Excluding these municipalities from multivariable adjusted analyses, the percentage increase in coronary heart disease mortality among women was 11.9% (95% CI 5.7–18.7%) comparing municipalities with arsenic concentrations between 1 and 10 mg/L to those o1 mg/L, similar to the result obtained in the whole sample.
4. Discussion In this ecologic study covering nearly 60% of the total Spanish population, cardiovascular mortality rates in municipalities with arsenic concentrations in drinking water 410 mg/L were 3.2% increased in women and 2.4% increased in men compared to municipalities with concentrations o1 mg/L, after adjustment for differences in social determinants, cardiovascular risk factors, diet, and water characteristics, measured at the municipal or provincial level. For municipalities with arsenic concentrations between 1 and 10 mg/L, the corresponding figures were 4.0% in women and 0.4% in men. For coronary heart disease mortality, the multivariable adjusted association was stronger for women compared to men. For cerebrovascular disease mortality, multivariable associations were weaker and not statistically significant. Our study extends previous findings in populations with high arsenic concentrations in drinking water (Tseng, 2008; NavasAcien et al., 2005; Wang et al., 2007; Yuan et al., 2007) to populations with low-to-moderate concentrations, supporting a role for low-chronic arsenic exposure in cardiovascular and coronary heart disease development. The association was modest and only borderline statistically significant after multivariable adjustment. However, more than 1 million people in the study lived in municipalities with arsenic concentrations 4 10 mg/L and almost 5 millions lived in municipalities with arsenic concentrations between 1 and 10 mg/L. Moreover, inherent to the ecological
Men 40
10
30 5 20 0 10 -5
0 0.2
0.5
1
2
5
10
20
Women 40
10
30
Percentage of municipalities
40
Percentage change in cardiovascular mortality
Percentage change in cardiovascular mortality
10
Percentage of municipalities
452
5 20 0 10 -5
0 0.2
0.5
1
2
5
10
20
Arsenic concentration in drinking water (µg/L) Fig. 3. Percentage change in sex-specific cardiovascular mortality by municipal arsenic concentration in drinking water. Curves represent the adjusted percentage changes in cardiovascular mortality (solid curves) and their 95% confidence intervals (dashed curves) based on restricted quadratic splines for log-transformed arsenic concentrations with knots at 0.5, 2, and 10 mg/L. The reference value (percentage change=0) was set at 0.5 mg/L. Percentage changes were adjusted for age, per capita municipal income, hospital beds per population at provincial level, sex-specific prevalence of cardiovascular risk factors at provincial level (smoking, hypertension, high serum cholesterol, diabetes, overweight/obesity, and low physical activity), dietary factors at provincial level (fish, wine, olive oil, bottled water, and total energy-adjusted folate), and water characteristics at municipal level (hardness, magnesium, pH, and temperature). Bars represent the histogram of municipal arsenic distribution.
design, the associations could have been underestimated because of substantial measurement error in both arsenic exposure and cardiovascular mortality measures. Arsenic is an established carcinogen and a high priority contaminant for screening in drinking water sources (World Health Organization, 2003). Low-to-moderate exposure to inorganic arsenic in drinking water affects many populations around the world. In Spain, arsenic concentrations in drinking water are relatively low, being o1 mg/L for the great majority of the population. In our study, 4.1% of the population was 410 mg/L, similar to those reported in Europe and North America (Smith et al., 2002). The World Health Organization (WHO), the European Union, and many countries around the world have established 10 mg/L as the safety standard for arsenic concentrations in drinking water, mostly based on risk benefit analysis for the carcinogenic effects of arsenic and practical/economic difficulties to remove arsenic in drinking water o10 mg/L. However, because many uncertainties remain at low exposure levels, in particular for non-cancer endpoints, the WHO designated 10 mg/L as a provisional standard. In the US, the National Research Council
ARTICLE IN PRESS M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
concluded that more research was needed to evaluate the association of chronic arsenic exposure with cardiovascular disease (National Research Council, 1999, 2001). Ecological and cross-sectional studies evaluating low-chronic arsenic exposure in drinking water and cardiovascular endpoints have been contradictory. Our findings are consistent with increased mortality for diseases of the arteries, arterioles, and capillaries in US counties with arsenic levels in drinking water 420 mg/L (Engel and Smith, 1994), increased mortality for diseases of the circulatory system in six Michigan counties with drinking water arsenic 410 mg/L (Meliker et al., 2007), and increased prevalence of self-reported circulatory problems with arsenic in drinking water 410 mg/L in Wisconsin (Zierold et al., 2004). In the study by Zierold et al. (2004), participants exposed to arsenic 2 10 mg/L in drinking water were also more likely to report a history of heart disease (odds ratio 1.52, 95% CI 1.00– 2.35). Prospective cohort studies with individual measures of arsenic exposure, standardized cardiovascular outcomes and adequate adjustment for confounders are needed to confirm the ecological findings at low-chronic exposure levels (Navas-Acien et al., 2005). Experimental and mechanistic studies have shown that arsenic can induce atherosclerosis, up-regulate inflammatory signals and enhance oxidative stress (Bunderson et al., 2004; Simeonova and Luster, 2004; States et al., 2009). The relevance of those findings at low-chronic levels has been uncertain because of typically high arsenic concentrations used in these experiments. Recent studies using lower and lower arsenic concentrations are identifying important vascular effects, including neovascularization, angiogenesis, and vessel remodeling at levels o10 mg/L in drinking water (Soucy et al., 2005; Straub et al., 2007, 2008). In our study, the strength of the association for the arsenic category 410 mg/L with cardiovascular and coronary mortality were comparable by gender. For arsenic 1 10 mg/L, however, the association with cardiovascular and coronary mortality was stronger for women compared to men. Reasons for gender differences at lower arsenic exposure concentrations could be related to differential confounding by gender, competing mortality by other conditions related to arsenic toxicity that could be more common among men (lung cancer, bladder cancer, chronic respiratory disease, for instance, because of a strong interaction between arsenic and smoking that is more common among men (Ministerio de Sanidad y Consumo, 1995)), metabolic reasons, or other reasons. Because of its design, this study evaluates the public health impact of arsenic in drinking water with cardiovascular mortality at the municipal level, but is limited to infer the relation of arsenic with cardiovascular mortality at the individual level. Important non-differential exposure misclassification could have substantially underestimated the strength of the associations. Other limitations must be considered. Firstly, selection bias induced by selective participation in the National Information System of Consume Water Control cannot be ruled out, although is unlikely given similar cardiovascular mortality, diet, and prevalence of risk factors in municipalities with and without arsenic exposure information. Secondly, the quality of arsenic assays and number of measures could have potentially affected the study, although consistent results were found in sensitivity analysis restricted to municipalities with at least 8 arsenic determinations. Thirdly, other sources of inorganic arsenic exposure, such as food and past-use of arsenic containing pesticides were not included. Finally, although we used hierarchical Poisson models with province-specific random intercepts to adjust the association between municipal arsenic concentration in drinking water and cardiovascular mortality for between-province differences in any relevant confounding factor, residual confounding induced by
453
within-province variations in social determinants, cardiovascular risk factors, and diet cannot be ruled out. Compared to other ecological studies, this study adjusted for potential confounders at the municipal or provincial level. Lack of adjustment for differences in socioeconomic and cardiovascular factors across communities could maybe explain some of the inconsistencies in previous ecological analyses (Engel and Smith, 1994; Meliker et al., 2007). Other strengths include the large study area covering almost 25 million individuals across Spain, multiple water arsenic determinations with consistent concentrations over time for most municipalities, and the use of procedures for water sampling and chemical analyses that were standardized by Spanish regulations. Statistical strengths include the use of multilevel (municipal and provincial) methods, the inclusion of a random intercept accounting for unexplained provincial-level variations in mortality, dose-response analyses using policy relevant cut-offs and flexible splines, and stratification by gender. In conclusion, this ecological study suggests that elevated arsenic concentrations in drinking water were associated with increased cardiovascular mortality at the municipal level in a country characterized by low-to-moderate arsenic exposure levels. Among women, the association with cardiovascular and coronary disease mortality was observed at arsenic concentrations below current standards (1 10 mg/L). Chronic exposure to low/moderate inorganic arsenic from drinking water and diet is widespread. Given increasing evidence of environmental exposures as cardiovascular disease risk factors (Bhatnagar, 2006), the investigation of the cardiovascular impact of arsenic in population-based prospective cohort studies is a public health priority. The study findings, together with evidence of the carcinogenic (IARC, 2004), developmental (Wasserman et al., 2004), and metabolic (Navas-Acien et al., 2008, 2009) effects of arsenic, reinforce the need to implement arsenic remediation treatments in water supply systems that are above the current WHO safety standard of 10 mg/L (Aragone´s-Sanz et al., 2001; Blanco Herna´ndez et al., 1998; Rodrı´guez et al., 2006).
Acknowledgments The authors are grateful to the Spanish Ministry of Health and Social Policy for providing the data on water composition and to the Spanish National Institute for Statistics and Mrs. Margarita Garcı´a Ferruelo for providing mortality files. References Aragone´s-Sanz, N., Palacios Dı´ez, M., Avello de Miguel, A., Go´mez Rodrı´guez, P., Martı´nez Corte´s, M., Rodrı´guez Bernabeu, M.J., 2001. Nivel de arse´nico en abastecimientos de agua de consumo de origen subterra´neo en la Comunidad de Madrid. Rev. Esp. Salud Pu´blica 75, 421–432. ˜ ol de Cre´dito, 1993. In: Anuario del Mercado Espan ˜ ol. Banco Espan ˜ ol Banco Espan de Cre´dito, Madrid. Bhatnagar, A., 2006. Environmental cardiology. Studying mechanistic links between pollution and heart disease. Circ. Res. 99, 692–705. Blanco Herna´ndez, A.L., Alonso Gutie´rrez, D., Jime´nez de Blas, O., Santiago Guervo´s, M., de Miguel Manzano, B., 1998. Estudio de los niveles de plomo, cadmio, zinc y arse´nico, en aguas de la provincia de Salamanca. Rev. Esp. Salud Pu´blica 72, 53–65. Breslow, N.E., Clayton, D.G., 1993. Approximate inference in generalized linear mixed models. J. Am. Stat. Assoc. 88, 9–25. Bunderson, M., Brooks, D.M., Walker, D.L., Rosenfeld, M.E., Coffin, J.D., Beall, H.D., 2004. Arsenic exposure exacerbates atherosclerotic plaque formation and increases nitrotyrosine and leukotriene biosynthesis. Toxicol. Appl. Pharmacol. 201, 32–39. Chen, C.-J., Hsueh, Y.-M., Lai, M.-S., Shyu, M.-P., Chen, S.-Y., Wu, M.-M., Kuo, T.-L., Tai, T.-Y., 1995. Increased prevalence of hypertension and long-term arsenic exposure. Hypertension 25, 53–60. Chen, C.-J., Chiou, H.-Y., Chiang, M.-H., Lin, L.-J., Tai, T.-Y., 1996. Dose-response relationship between ischaemic heart disease mortality and long-term arsenic exposure. Arterioscler. Thromb. Vasc. Biol. 16, 504–510.
ARTICLE IN PRESS 454
M.J. Medrano et al. / Environmental Research 110 (2010) 448–454
Chen, C.-J., Wu, M.-M., Lee, S.-S., Wang, J.-D., Cheng, S.-H., Wu, H.-Y., 1998. Atherogenicity and carcinogenicity of high-arsenic artesian well water. Multiple risk-factors and related malignant neoplasms of blackfoot disease. Arteriosclerosis 8, 452–460. Chen, Y., Factor-Litvak, P., Howe, G.R., Graziano, J.H., Brandt-Rauf, P., Parvez, F., van Geen, A., Ahsan, H., 2007. 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. Am. J. Epidemiol. 165, 541–552. Chiou, H.-Y., Huang, W.-I., Hsu, C.-L., Chang, S.-F., Hsu, Y.-H., Chen, C.-J., 1997. Doseresponse relationship between prevalence of cerebrovascular disease and ingested inorganic arsenic. Stroke 28, 1717–1723. Coronado-Gonza´lez, J.A., del Razo, L.M., Garcı´a-Vargas, G., Sanmiguel-Salazar, F., ˜ a, J., 2007. Inorganic arsenic exposure and type 2 diabetes Escobedo-de la Pen mellitus in Mexico. Environ. Res. 104, 383–389. Engel, R., Smith, A., 1994. Arsenic in drinking water and mortality from vascular disease: an ecologic analysis in 30 counties in the United States. Arch. Environ. Health 49, 418–427. ˜ o Fonseca, I., Gonza´lez Parama´s, Garcı´a-Villanova, R.J., Rodrı´guez Torrero, P., Torun A.M., Go´mez Ba´rez, J.A., Cordo´n Marcos, C., 2005. Arsenic in ground drinking water. The regional monitoring plan of Castilla y Leo´n (Spain). Eur. J. Water Qual. 36, 107–117. Gelman, A., Hill, J., 2007. In: Data Analysis using Regression and Multilevel/ Hierarchical Models. Cambridge University Press, Cambridge. Greenland, S., 1995. Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology 6, 356–365. Greenland, S., Rothman, K.J., 2008. Introduction to categorical statistics. In: Rothman, K.J., Greenland, S., Lash, T.L. (Eds.), Modern Epidemiology third ed Wolters Kluwer/Lippincott Williams & Wilkins, Philadelphia. IARC Working Group on the Evaluation of Carcinogenic Risks to Humans, 2004. In: Some Drinking-Water Disinfectants and Contaminants, Including Arsenic. IARC Press, Lyon, France. Instituto Nacional de Estadı´stica, 2004. In: Estadı´stica de Establecimientos Sanitarios con Re´gimen de Internado 2001. Instituto Nacional de Estadı´stica, Madrid. Lai, M.-S., Hsueh, Y.-M., Chen, C.-J., Shyu, M.-P., Chen, S.-Y., Kuo, T.-L., Wu, M.-M., Tai, T.-Y., 1994. Ingested inorganic arsenic and prevalence of diabetes mellitus. Am. J. Epidemiol. 139, 484–492. Meliker, J.R., Wahl, R.L., Cameron, L.L., Nriagu, J.O., 2007. Arsenic in drinking water and cerebrovascular disease, diabetes mellitus, and kidney disease in Michigan: a standardized mortality ratio analysis. Environ. Health 6, 4. ˜ a. Ministerio de Sanidad y Consumo, 1995. In: Encuesta Nacional de Salud de Espan Ministerio de Sanidad y Consumo, Madrid. Ministerio de Sanidad y Consumo, 2000. In: Calidad del agua de Consumo en ˜ a. 1er Informe Nacional. An ˜ o 1993/1995. Ministerio de Sanidad y Espan Consumo, Madrid, pp. 47–48. National Research Council, 1999. Arsenic in Drinking Water. Subcommittee on Arsenic in Drinking Water. National Academy Press, Washington, DC. National Research Council, 2001. Arsenic in Drinking Water 2001 Update. Subcommittee on Arsenic in Drinking Water. National Academy Press, Washington, DC. Navas-Acien, A., Sharrett, A.R., Silbergeld, E.K., Schwartz, B.S., Nachman, K.E., Burke, T.A., 2005. Arsenic exposure and cardiovascular disease: a systematic review of the epidemiologic evidence. Am. J. Epidemiol. 162, 1037–1049. Navas-Acien, A., Silbergeld, E.K., Pastor-Barriuso, R., Guallar, E., 2008. Arsenic exposure and prevalence of type 2 diabetes in US adults. JAMA 300, 814–822. Navas-Acien, A., Silbergeld, E.K., Pastor-Barriuso, R., Guallar, E., 2009. Arsenic exposure and prevalence of type 2 diabetes: updated findings from the National Health Nutrition and Examination Survey, 2003–2006. Epidemiology101097/EDE.0b013e3181afef88.
Rahman, M., Tondel, M., Ahmad, S.A., Axelson, O., 1998. Diabetes mellitus associated with arsenic exposure in Bangladesh. Am. J. Epidemiol. 148, 198–203. Rahman, M., Tondel, M., Ahmad, S.A., Chowdhury, I.A., Faruquee, M.H., Axelson, O., 1999. Hypertension and arsenic exposure in Bangladesh. Hypertension 33, 74–78. Rodrı´guez, V., Morote, P., Arbesu´, E., Alvarez, P., Esteban, A.I., Suarez, P., 2006. Enfermedades relacionadas con la ingesta cro´nica de arse´nico a trave´s de aguas de consumo. Gac. Sanit. 20 (Espec Congr), 48. Sahu´n, B., Go´mez, J.J., Lillo, T., del Olmo, P., 2004. Arse´nico en aguas subterra´neas e interaccio´n agua-roca: un ejemplo en la cuenca terciaria del Duero (Castilla y ˜ a). Rev. Soc. Geol. Espan ˜ a 17, 137–155. Leo´n, Espan Simeonova, P.P., Luster, M.I., 2004. Arsenic and atherosclerosis. Toxicol. Appl. Pharmacol. 198, 444–449. Smith, A.H., Lopipero, P.A., Bates, M.N., Steinmaus, C.M., 2002. Arsenic epidemiology and drinking water standards. Science 296, 2145–2146. Soucy, N.V., Mayka, D., Klei, L.R., Nemec, A.A., Bauer, J.A., Barchowsky, A., 2005. Neovascularization and angiogenic gene expression following chronic arsenic exposure in mice. Cardiovasc. Toxicol. 5, 9–41. States, J.C., Srivastava, S., Chen, Y., Barchowsky, A., 2009. Arsenic and cardiovascular disease. Toxicol. Sci. 107, 312–323. Straub, A.C., Stolz, D.B., Vin, H., Ross, M.A., Soucy, N.V., Klei, L.R., Barchowsky, A., 2007. Low level arsenic promotes progressive inflammatory angiogenesis and liver blood vessel remodeling in mice. Toxicol. Appl. Pharmacol. 222, 327–336. Straub, A.C., Clark, K.A., Ross, M.A., Chandra, A.G., Li, S., Gao, X., Pagano, P.J., Stolz, D.B., Barchowsky, A., 2008. Arsenic-stimulated liver sinusoidal capillarization in mice requires NADPH oxidase-generated superoxide. J. Clin. Invest. 118, 3980–3989. Tseng, C.H., Huang, Y.K., Huang, Y.L., Chung, C.J., Yang, M.H., Chen, C.J., Hsueh, Y.M., 2005. Arsenic exposure, urinary arsenic speciation, and peripheral vascular disease in blackfoot disease-hyperendemic villages in Taiwan. Toxicol. Appl. Pharmacol. 206, 299–308. Tseng, C.H., 2008. Cardiovascular disease in arsenic-exposed subjects living in the arseniasis-hyperendemic areas in Taiwan. Atherosclerosis 199, 12–18. Varela, G., Moreiras, O., Carbajal, A., Campo, M., 1995. In: Encuesta de Presupuestos Familiares. Estudio Nacional de Nutricio´n y Alimentacio´n. Instituto Nacional de Estadı´stica, Madrid. Wang, C.H., Jeng, J.S., Yip, P.K., Chen, C.L., Hsu, L.I., Hsueh, Y.M., Chiou, H.Y., Wu, M.M., Chen, C.J., 2002. Biological gradient between long-term arsenic exposure and carotid atherosclerosis. Circulation 105, 1804–1809. Wang, C.H., Hsiao, C.K., Chen, C.L., Hsu, L.I., Chiou, H.Y., Chen, S.Y., Hsueh, Y.M., Wu, M.M., Chen, C.J., 2007. A review of the epidemiologic literature on the role of environmental arsenic exposure and cardiovascular diseases. Toxicol. Appl. Pharmacol. 222, 315–326. Wasserman, G.A., Liu, X., Parvez, F., Ahsan, H., Factor-Litvak, P., van Geen, A., Slavkovich, V., LoIacono, N.J., Cheng, Z., Hussain, I., Momotaj, H., Graziano, J.H., 2004. Water arsenic exposure and children’s intellectual function in Araihazar, Bangladesh. Environ. Health Perspect. 112, 1329–1333. World Health Organization, 2003. In: Guidelines for Drinking-Water Quality third ed. World Health Organization, Geneva. Wu, M.-M., Kuo, T.-L., Hwang, Y.-H., Chen, C.-J., 1989. Dose-response relation between arsenic concentration in well water and mortality from cancers and vascular disease. Am. J. Epidemiol. 130, 1123–1132. Yuan, Y., Marshall, G., Ferreccio, C., Steinmaus, C., Selvin, S., Liaw, J., Bates, M.N., Smith, A.H., 2007. Acute myocardial infarction mortality in comparision with lung and bladder cancer mortality in arsenic-exposed region II of Chile from 1950 to 2000. Am. J. Epidemiol. 166, 1381–1391. Zierold, K.M., Knobeloch, L., Anderson, H., 2004. Prevalence of chronic diseases in adults exposed to arsenic-contaminated drinking water. Am. J. Public Health 94, 1936–1937.