Exposure to Inorganic Arsenic in Drinking Water and CHD
In this prospective study, we found that lifetime exposure to low levels of inorganic arsenic in drinking water (10–100 μg/L) was associated with increased risk for CHD. We estimated that for every 15-μg/L increase in arsenic concentration in residential drinking water, the risk for CHD increased by 38%; and across increasing levels of exposure, risk increased in a dose-dependent fashion (trend p = 0.0007) after adjusting for sex, family history of CHD, and serum LDL levels.
The wide spectrum of longitudinal clinical, behavioral, and demographic data in SLVDS, plus a low rate of out-migration, along with variability in inorganic arsenic exposure in the San Luis Valley, renders this region and cohort particularly suitable for this research. Inorganic arsenic in groundwater in the San Luis Valley is natural, resulting primarily from weathering and erosion of rock formations (Neely 2002), and spatial variation is due to long-term patterns of rainfall and physio-chemical conditions (Abernathy et al. 2003; Hinwood et al. 2003).
We used a thorough residential and employment history, coupled with a comprehensive spatial prediction model of groundwater concentrations of inorganic arsenic, to characterize a life-course time-weighted average arsenic exposure at the individual level. The selection of residential arsenic concentration as the exposure metric was based on a correlation analysis with speciated arsenic concentrations in historically collected urine samples from this same cohort.
One plausible mechanism for arsenic cardiotoxicity is through the creation of oxygen radicals including lipid peroxidase, which can initiate endothelial cell proliferation, function, and apoptosis, a precursor to atherosclerosis (Chen Y et al. 2009; Hirano et al. 2003; Navas-Acien et al. 2005; Pi et al. 2002; Ratnaike 2003; Santra 2000; Waalkes et al. 2000). Studies in high-arsenic areas of Asia have found increased levels of circulating reactive oxygen species such as hydrogen peroxide, hydroxyl radicals, and superoxide radicals (Yamanaka et al. 1990) and higher blood levels of lipid peroxidase (Pi et al. 2002) versus low-exposure comparison groups. Other arsenic toxicity mechanisms that have been suggested include vascular smooth muscle cell proliferation and dysfunction (Bae et al. 2008), inhibited endothelial nitric oxide synthase activity (Kao et al. 2003; Lee et al. 2003), smooth muscle cell migration (Simeonova and Luster 2004), and enhanced platelet aggregation (Lee et al. 2002).
Past research has documented an association between hypertension and inorganic arsenic exposure (Chen CJ et al. 1995, 2007; Rahman et al. 1999); a systematic review of 11 studies (Abhyankar et al. 2012) also corroborated an association between arsenic exposure in drinking water and hypertension, even at low concentrations of arsenic. These findings suggest that arsenic may be related to CHD through a pathway that includes hypertension, a known risk factor for CHD; consequently, hypertension was not included as an independent risk factor in our models. We assessed hypertension as a potential confounder and found no change in the association between CHD and arsenic exposure. Also, although the study was adequately powered to investigate risk from arsenic exposure, there may be concern that well-known risk factors for CHD, including BMI and smoking, were not associated with CHD in this small cohort; however, other known CHD risk factors (family history of CHD, serum LDL levels, and sex) were significantly associated.
Recent research has suggested that intake of folate and selenium can influence arsenic metabolism and the association between cardiovascular disease (George et al. 2013); however, in this study, folate and selenium intake levels did not significantly contribute to the hazards model nor significantly change the association. This difference in finding could be attributable to variations in estimation of micronutrient intake in this study compared with others, and therefore should be a consideration to improve measurements for future research.
There exists the possibility of misclassification bias due to the exposure estimation in the use of exposure predication models and residential history reconstruction. Although our groundwater modeled predictions were correlated with arsenic concentrations measured in urine samples (ρ = 0.63), misclassification cannot be ruled out. We also found in a limited cohort that cases had a statistically higher level of toxic urine arsenic species, suggesting that in a different metric of exposure an association between inorganic arsenic exposure and CHD also exists. Specific to the residential reconstruction, the primary method for data collection (interview) varied in response by case status (30% noncases, 45% cases) although not by exposure status, which could induce misclassification bias. We believe that misclassification bias would be small given the low migration of this population [5% migrated to the SLV as children, and only 10 participants (war veterans) lived outside of the SLV for > 6 months (< 3 years) through 1998] and the validation of clerk records and SLVDS database to complete residential history.
The exposure assessment does not include exposure resulting from ingesting contaminated food, inhalation of dust or soil, or use of tobacco products. In a comprehensive review of literature and analysis of arsenic, the Agency for Toxic Substances and Disease Registry (2007) noted that in areas of the United States where arsenic levels in drinking water are > 10 μg/L, ingestion of drinking water is the dominant source of inorganic arsenic exposure relative to the U.S. diet and inhalation through air, and therefore confirms our use of drinking water as the main source of exposure (Tao and Bolger 1999).
The exposure assessment remains limited by potential misclassification bias. Thirty-three percent of the subjects had a residential history created through records at the county clerk office because they were not available (e.g., deceased) for interview. However, the use of county clerk records was confirmed in participants who were not interviewed, where we found that data collected from county clerk records had strong agreement with self-reported residential history. For the subjects with imputed residence (n = 18 subjects, 126 person-years), we looked at city for the address before and after the period with missing residence and found an 89% agreement, suggesting that many residents may move houses, but not necessarily out of the city or out of the San Luis Valley. Self-reported estimates of lifetime residential, employment, and schooling locations and duration, and number of cups of water consumed per day also are likely limited by inaccuracies leading to misclassification bias which could bias the findings.
Another limitation is that the arsenic exposure estimates were included in the proportional-hazards model under the assumption of no error. Past research has incorporated bootstrap methods to incorporate an error term for the estimate in logistic regression models, but to date has not been done in a proportional-hazards model. A future step would be to develop the statistical methodology for incorporating the error term associated with the predicted arsenic exposure into the proportional-hazards model.
A recent study from a high arsenic area reported an HR for CHD of 1.22 (95% CI: 0.65, 2.32) at arsenic levels of 12.1–62.0 μg/L, similar to levels found in the SLV, while controlling for known CHD risk factors (Chen Y et al. 2011). Using a comprehensive exposure assessment, our study found consistent results at lower levels (1–100 μg/L), with a proportional-hazards ratio of 1.75 for exposure levels from 30 to 45 μg/L relative to < 20 μg/L. Our findings plus those by Moon et al. (2013), who identified a similar association between incident cardiovascular disease and exposure to low to moderate arsenic levels with exposure defined through urine biomarkers, indicate that a dose–response relationship between arsenic and CHD exists at levels of arsenic that are not uncommon in many areas.
Inorganic arsenic exposure in drinking water has been identified as a cardiotoxic element at concentrations seen in drinking water supplies around the world which strengthens the importance of ensuring public water supplies meet the U.S. EPA maximum contaminant level (MCL) of 10 μg/L (U.S. EPA 1998). Currently, many areas of the United States have levels above the U.S. EPA MCL, including western states of Nevada, Colorado, and Arizona; Midwest areas including Michigan; and Northeastern areas of New Hampshire, Maine, and Connecticut (U.S. EPA 2011).
In conclusion, we observed an association between CHD risk and inorganic arsenic exposure in a chronic low-level arsenic area in the southwestern United States. Because arsenic in drinking water remains a common exposure in the United States, the risk of CHD should provide motivation to public health officials to bring drinking water levels into compliance (< 10 μg/L) and to conduct further research to elucidate the role of arsenic in the pathobiology of CHD.
Discussion
In this prospective study, we found that lifetime exposure to low levels of inorganic arsenic in drinking water (10–100 μg/L) was associated with increased risk for CHD. We estimated that for every 15-μg/L increase in arsenic concentration in residential drinking water, the risk for CHD increased by 38%; and across increasing levels of exposure, risk increased in a dose-dependent fashion (trend p = 0.0007) after adjusting for sex, family history of CHD, and serum LDL levels.
The wide spectrum of longitudinal clinical, behavioral, and demographic data in SLVDS, plus a low rate of out-migration, along with variability in inorganic arsenic exposure in the San Luis Valley, renders this region and cohort particularly suitable for this research. Inorganic arsenic in groundwater in the San Luis Valley is natural, resulting primarily from weathering and erosion of rock formations (Neely 2002), and spatial variation is due to long-term patterns of rainfall and physio-chemical conditions (Abernathy et al. 2003; Hinwood et al. 2003).
We used a thorough residential and employment history, coupled with a comprehensive spatial prediction model of groundwater concentrations of inorganic arsenic, to characterize a life-course time-weighted average arsenic exposure at the individual level. The selection of residential arsenic concentration as the exposure metric was based on a correlation analysis with speciated arsenic concentrations in historically collected urine samples from this same cohort.
One plausible mechanism for arsenic cardiotoxicity is through the creation of oxygen radicals including lipid peroxidase, which can initiate endothelial cell proliferation, function, and apoptosis, a precursor to atherosclerosis (Chen Y et al. 2009; Hirano et al. 2003; Navas-Acien et al. 2005; Pi et al. 2002; Ratnaike 2003; Santra 2000; Waalkes et al. 2000). Studies in high-arsenic areas of Asia have found increased levels of circulating reactive oxygen species such as hydrogen peroxide, hydroxyl radicals, and superoxide radicals (Yamanaka et al. 1990) and higher blood levels of lipid peroxidase (Pi et al. 2002) versus low-exposure comparison groups. Other arsenic toxicity mechanisms that have been suggested include vascular smooth muscle cell proliferation and dysfunction (Bae et al. 2008), inhibited endothelial nitric oxide synthase activity (Kao et al. 2003; Lee et al. 2003), smooth muscle cell migration (Simeonova and Luster 2004), and enhanced platelet aggregation (Lee et al. 2002).
Past research has documented an association between hypertension and inorganic arsenic exposure (Chen CJ et al. 1995, 2007; Rahman et al. 1999); a systematic review of 11 studies (Abhyankar et al. 2012) also corroborated an association between arsenic exposure in drinking water and hypertension, even at low concentrations of arsenic. These findings suggest that arsenic may be related to CHD through a pathway that includes hypertension, a known risk factor for CHD; consequently, hypertension was not included as an independent risk factor in our models. We assessed hypertension as a potential confounder and found no change in the association between CHD and arsenic exposure. Also, although the study was adequately powered to investigate risk from arsenic exposure, there may be concern that well-known risk factors for CHD, including BMI and smoking, were not associated with CHD in this small cohort; however, other known CHD risk factors (family history of CHD, serum LDL levels, and sex) were significantly associated.
Recent research has suggested that intake of folate and selenium can influence arsenic metabolism and the association between cardiovascular disease (George et al. 2013); however, in this study, folate and selenium intake levels did not significantly contribute to the hazards model nor significantly change the association. This difference in finding could be attributable to variations in estimation of micronutrient intake in this study compared with others, and therefore should be a consideration to improve measurements for future research.
There exists the possibility of misclassification bias due to the exposure estimation in the use of exposure predication models and residential history reconstruction. Although our groundwater modeled predictions were correlated with arsenic concentrations measured in urine samples (ρ = 0.63), misclassification cannot be ruled out. We also found in a limited cohort that cases had a statistically higher level of toxic urine arsenic species, suggesting that in a different metric of exposure an association between inorganic arsenic exposure and CHD also exists. Specific to the residential reconstruction, the primary method for data collection (interview) varied in response by case status (30% noncases, 45% cases) although not by exposure status, which could induce misclassification bias. We believe that misclassification bias would be small given the low migration of this population [5% migrated to the SLV as children, and only 10 participants (war veterans) lived outside of the SLV for > 6 months (< 3 years) through 1998] and the validation of clerk records and SLVDS database to complete residential history.
The exposure assessment does not include exposure resulting from ingesting contaminated food, inhalation of dust or soil, or use of tobacco products. In a comprehensive review of literature and analysis of arsenic, the Agency for Toxic Substances and Disease Registry (2007) noted that in areas of the United States where arsenic levels in drinking water are > 10 μg/L, ingestion of drinking water is the dominant source of inorganic arsenic exposure relative to the U.S. diet and inhalation through air, and therefore confirms our use of drinking water as the main source of exposure (Tao and Bolger 1999).
The exposure assessment remains limited by potential misclassification bias. Thirty-three percent of the subjects had a residential history created through records at the county clerk office because they were not available (e.g., deceased) for interview. However, the use of county clerk records was confirmed in participants who were not interviewed, where we found that data collected from county clerk records had strong agreement with self-reported residential history. For the subjects with imputed residence (n = 18 subjects, 126 person-years), we looked at city for the address before and after the period with missing residence and found an 89% agreement, suggesting that many residents may move houses, but not necessarily out of the city or out of the San Luis Valley. Self-reported estimates of lifetime residential, employment, and schooling locations and duration, and number of cups of water consumed per day also are likely limited by inaccuracies leading to misclassification bias which could bias the findings.
Another limitation is that the arsenic exposure estimates were included in the proportional-hazards model under the assumption of no error. Past research has incorporated bootstrap methods to incorporate an error term for the estimate in logistic regression models, but to date has not been done in a proportional-hazards model. A future step would be to develop the statistical methodology for incorporating the error term associated with the predicted arsenic exposure into the proportional-hazards model.
A recent study from a high arsenic area reported an HR for CHD of 1.22 (95% CI: 0.65, 2.32) at arsenic levels of 12.1–62.0 μg/L, similar to levels found in the SLV, while controlling for known CHD risk factors (Chen Y et al. 2011). Using a comprehensive exposure assessment, our study found consistent results at lower levels (1–100 μg/L), with a proportional-hazards ratio of 1.75 for exposure levels from 30 to 45 μg/L relative to < 20 μg/L. Our findings plus those by Moon et al. (2013), who identified a similar association between incident cardiovascular disease and exposure to low to moderate arsenic levels with exposure defined through urine biomarkers, indicate that a dose–response relationship between arsenic and CHD exists at levels of arsenic that are not uncommon in many areas.
Inorganic arsenic exposure in drinking water has been identified as a cardiotoxic element at concentrations seen in drinking water supplies around the world which strengthens the importance of ensuring public water supplies meet the U.S. EPA maximum contaminant level (MCL) of 10 μg/L (U.S. EPA 1998). Currently, many areas of the United States have levels above the U.S. EPA MCL, including western states of Nevada, Colorado, and Arizona; Midwest areas including Michigan; and Northeastern areas of New Hampshire, Maine, and Connecticut (U.S. EPA 2011).
In conclusion, we observed an association between CHD risk and inorganic arsenic exposure in a chronic low-level arsenic area in the southwestern United States. Because arsenic in drinking water remains a common exposure in the United States, the risk of CHD should provide motivation to public health officials to bring drinking water levels into compliance (< 10 μg/L) and to conduct further research to elucidate the role of arsenic in the pathobiology of CHD.
SHARE
