Policy Analysis pubs.acs.org/est
Estimating Potential Increased Bladder Cancer Risk Due to Increased Bromide Concentrations in Sources of Disinfected Drinking Waters Stig Regli,*,† Jimmy Chen,† Michael Messner,† Michael S. Elovitz,‡ Frank J. Letkiewicz,§ Rex A. Pegram,∥ T.J. Pepping,†,⊥ Susan D. Richardson,# and J. Michael Wright∇ †
Office of Ground Water and Drinking Water, U.S. Environmental Protection Agency, Washington, D.C. 20460, United States Office of Research and Development, National Risk Management Laboratory, U.S. Environmental Protection Agency, Cincinnati, Ohio 45268, United States § The Cadmus Group, Bethesda, Maryland 20814, United States ∥ Office of Research and Development, National Health and Environmental Effects Research Laboratory, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina 27711, United States ⊥ Oak Ridge Institute for Science and Education Internship/Research Participation Program, U.S. Department of Energy, Oak Ridge, Tennessee 37830, United States # Department of Chemistry and Biochemistry, University of South Carolina, Columbia, South Carolina 29208, United States ∇ Office of Research and Development, National Center for Environmental Assessment, U.S. Environmental Protection Agency, Cincinnati, Ohio 45268, United States ‡
S Supporting Information *
ABSTRACT: Public water systems are increasingly facing higher bromide levels in their source waters from anthropogenic contamination through coal-fired power plants, conventional oil and gas extraction, textile mills, and hydraulic fracturing. Climate change is likely to exacerbate this in coming years. We estimate bladder cancer risk from potential increased bromide levels in source waters of disinfecting public drinking water systems in the United States. Bladder cancer is the health end point used by the United States Environmental Protection Agency (EPA) in its benefits analysis for regulating disinfection byproducts in drinking water. We use estimated increases in the mass of the four regulated trihalomethanes (THM4) concentrations (due to increased bromide incorporation) as the surrogate disinfection byproduct (DBP) occurrence metric for informing potential bladder cancer risk. We estimate potential increased excess lifetime bladder cancer risk as a function of increased source water bromide levels. Results based on data from 201 drinking water treatment plants indicate that a bromide increase of 50 μg/L could result in a potential increase of between 10−3 and 10−4 excess lifetime bladder cancer risk in populations served by roughly 90% of these plants.
■
INTRODUCTION
also result in increased risk of adverse health effects, such as bladder cancer, due to long-term drinking water usage.
In recent years, drinking water utilities in various regions of the U.S. have expressed concerns over increased bromide levels in their source waters.1−4 In addition to natural sources of bromide from salt water intrusion or geologic formations, there are new anthropogenic bromide releases from coal-fired power plants, conventional oil and gas extraction, textile mills, and hydraulic fracturing activities. Bromide reacts with aqueous chlorine in the form of hypochlorous acid (HOCl) to form hypobromous acid (HOBr) which reacts with organic material to form brominated disinfection byproducts (DBPs) such as brominated trihalomethanes (THMs) and haloacetic acids (HAAs). Increased bromide concentration in the source water may induce changes in concentrations and speciation of DBPs in the finished water. These changes in DBP formation may © 2015 American Chemical Society
■
BACKGROUND Regulatory Context. DBPs are an unintended consequence from disinfection used to minimize risk from pathogens in drinking water. DBPs are complex mixtures in drinking water supplies, most obviously in the formation of chlorination DBPs, and are influenced by source water quality and physicochemical treatment processes. The United States EPA Received: Revised: Accepted: Published: 13094
July 21, 2015 October 16, 2015 October 21, 2015 October 21, 2015 DOI: 10.1021/acs.est.5b03547 Environ. Sci. Technol. 2015, 49, 13094−13102
Policy Analysis
Environmental Science & Technology
and bladder cancer risk in the Cantor et al. 2010 study may be more pronounced than what might be found for the general US population and for which we later model (based on a metaanalysis which does not include the Cantor et al.16 data). Consistent with the findings for the BrTHMs, brominated DBPs, in general, are more genotoxic, cytotoxic, and carcinogenic than chlorinated DBPs based on toxicological data.18,19 A comprehensive genotoxicity database containing more than 80 DBPs was created that examined DNA damage in Chinese hamster ovary cells.18,19 This database allowed genotoxicity comparisons of several different DBP classes, along with the effect of bromide vs chloride substituents and monohalo vs dihalo vs trihalo substituents. Brominated species were found generally to be more genotoxic than chlorinated species. For example, bromoacetic acid is more than an order of magnitude more genotoxic than chloroacetic acid. This same pattern is also observed in Salmonella, as well as in new human cell data.29,30 However, there is not a consistent relationship between toxicity and the degree of bromine substitution in dihalo and trihalocompounds.21,31 The significantly higher genotoxicity of brominated versus chlorinated DBP species is also apparent when the bromo- and chloro-DBPs of different classes (beyond only THMs and HAAs) are summed together.18,19 Bromide DBP Chemistry and Exposure Implications. Additional occurrence of bromide in source waters increases the formation of brominated DBPs with disinfection using chlorine,32−40 chloramines,41−46 ozone,39−41,47 and chlorine dioxide.41,48 The resultant speciation shift from primarily chlorinated compounds to more heavily brominated compounds occurs across every compound class, including the THMs, HAAs, haloacetonitriles, halonitromethanes, haloaldehydes, haloketones, haloamides, and halofuranones, as well as new DBP classes recently discovered, including halobenzoquinones49,50 and other new brominated aromatic DBPs.51 As such, source waters having higher levels of bromide cause a shift from species like chloroform, dichloroacetic acid, trichloronitromethane, and MX (3-chloro-4-(dichloromethyl)-5-hydroxy2(5H)-furanone) to bromoform, dibromoacetic acid, tribromonitromethane, and brominated MX analogues resulting in higher mass of DBPs. Increased bromine substitution on a carbon atom generally renders these DBPs less stable due to steric hindrance and electronic effects. It is also important to note that increased bromide levels not only increases the overall mass of DBPs due to the higher mass of bromine (80 Da) relative to chlorine (35.5 Da), but research has also shown that the molar mass of the DBPs formed increases especially for THMs, HAAs, and haloacetonitriles.52−54 This can be explained by the increased reactivity, as reaction rates of HOBr with natural organic matter are at least 10 times faster than with HOCl.52,53,55 In addition, HOBr is more efficient at substitution reactions than chlorine, such that aqueous bromine is substituted more into organic structures, whereas chlorine tends to cleave carbon bonds and produce relatively less halogen substitution.53 DBP formation and degradation is complex and exposure assessment remains a key challenge in epidemiological studies. The impact of certain water-use activities and specific exposure routes are important considerations when evaluating the relationship between DBPs and bladder cancer. The physicochemical properties of DBPs can determine exposure route and impact absorption, metabolism, excretion, and distribution in the body (i.e., bioavailability). HAAs, for example, are neither volatile nor very permeable to skin; therefore, ingestion is
has regulated DBPs in public drinking water systems since 19795−7 mainly due to concern for their carcinogenicity in humans. Epidemiological studies indicate an association between exposure to chlorinated drinking water and bladder cancer.8−13 Increased incidence of bladder cancer has also been associated with exposure to higher THM4 levels in drinking water.10,11,14−17 In its economic analysis to support the Stage 2 DBP regulation,7 EPA used estimated bladder cancer cases avoided as the health effects metric for quantifying the benefits under the new regulation. Since they are the predominant DBPs, THMs and HAAs are considered surrogate measures for the co-occurrence of other DBPs in systems using chlorination. Why Brominated DBPs are a Concern. Given the multitude of DBPs present in drinking water, a key remaining challenge is to determine which mixtures of DBPs are of most toxicological concern. Among DBPs formed in chlorinated drinking water, bromine-containing species may pose greater health risks than those containing only chlorine.17−19 DNA damage and induction of gene mutations are important basic mechanisms (genotoxicity) by which chemicals can cause cancer. In many cases, chemical carcinogens are transformed by metabolizing enzymes to “activated” metabolites that react with DNA to produce mutations. A hypothesis has been developed linking bladder cancer risk to population variation in the ability to metabolize brominated trihalomethanes (BrTHMs) to genotoxic metabolites. BrTHMs, but not chloroform, were found to be mutagenic in Salmonella via metabolic activation by the enzyme, glutathione S-transferase theta-1 (GSTT1).20,21 Ross and Pegram22,23 subsequently demonstrated that GSTT1 is active in the urinary tract and that GSTT1-mediated metabolism of bromodichloromethane (BDCM) produces reactive intermediates that covalently bind DNA and deoxyguanosine, which is consistent with the possibility it is a mutagenic carcinogen. GSTT1 is polymorphically expressed in humans, which has enabled epidemiologists to compare bladder cancer risks in people that have (i.e., GSTT1(+)) and do not have (i.e., GSTT1-null) this enzyme. Cantor et al.16 found that people with the GSTT1(+) genotype were at significantly greater risk (odds ratio (OR) = 2.2; 95% confidence interval (CI): 1.1−4.3) for developing bladder cancer when exposed to the upper THM4 exposure quartile (>49 μg/L) compared to GSTT1-null participants who had no increased risk at the same exposure level. Cantor et al.16 also examined the role of genetic variation in another enzyme, GST zeta 1 (GSTZ1), which is involved in HAA metabolism and elimination.24,25 Study participants in the highest THM4 exposure quartile (>49 μg/L) had a 1.8 (95%CI: 0.9−3.5) increased risk of bladder cancer compared to those in the lowest quartile (35 μg/day of THM4 versus those with no reported consumption of chlorinated water (odds ratios (OR) = 1.35; 95%CI: 0.92 1.99). These effect estimates were smaller than the metric based on duration of shower or bath (OR = 1.83; 95%CI: 1.17 2.87) for the highest showering/bathing exposure quartile (duration × THM4 water concentration) as compared to the lowest quartile. This suggests that for this study population, which was essentially the same as that of Cantor et al.,16 the bladder cancer risk associated with exposure to DBPs may be more attributed to the dermal/inhalation route than ingestion.
■
ANALYSIS Overview. Given the association between bladder cancer and THM4 noted in epidemiological studies, and the impact of bromide on BrTHMs and other DBPs, we postulate that a population served by a utility producing drinking water with increased THM4 due to increased bromide levels in its source water may be subject to increased bladder cancer risk. Increased bromide levels in source waters from anthropogenic activities are a concern, since bromide is difficult to remove by treatment.1 We estimate potential increased bladder cancer risk for U.S. populations using disinfected public drinking water systems with increased bromide in their source waters by I. estimating potential increase in bladder cancer risk as a function of increased THM4 levels in finished water, II. estimating increased THM4 levels in finished waters attributed to increased bromide levels in source waters, and III. consolidating “I” and “II” to estimate potential increase in bladder cancer risk as a function of increased bromide levels in source water on a national level. In sum, we are estimating the potential increase in bladder cancer risk that might result from an increase in THM4 concentration (and associated DBPs) due specifically to increased bromide in the source water. Data Analysis IBladder Cancer Risk Estimates. In the Economic Analysis (EA) for the Stage 2 Disinfectants and Disinfection Byproduct Rule (DBPR), the EPA estimated the
OR = 1 + 0.00581 × C
where C is the average (lifetime) concentration of THM4 (μg/L). 13096
DOI: 10.1021/acs.est.5b03547 Environ. Sci. Technol. 2015, 49, 13094−13102
Policy Analysis
Environmental Science & Technology
- s = (ln(upper 95% confidence limit) − ln(lower 95% confidence limit))/(z0.975 − z0.025) - z = quantile of the standard normal distribution and z0.975 − z0.025 = 3.92 • The variance of the slope (ln(OR)/THM4) estimated from a single record of Table 1 is then s2/THM42 • The unadjusted weight for each row of the table is the inverse of the variance of the slope (weight = THM42/s2) and these are multiplied by a constant (the sum of unadjusted weights) so that the sum of adjusted weights is 1. • Finally, the estimated overall slope is the sum, across rows, of adjusted weight times ln(OR)/THM4. The result is 0.00427 and the resulting function is shown as a dashed line in Figure 1. The weighted slope (0.00427) is smaller than EPA’s unweighted estimate (0.00581),60 because it gives less weight to the three lowest nonzero exposure levels (THM4 = 10, 20, and 30) and also to the highest (THM4 = 130). The following equation relates ORs to average THM4 exposure:
Table 1. Odds Ratio Estimates and Weights Used for Bladder Cancer Associated with THM4 Exposure average THM4 OR (μg/L) (95% CI)a 0 10 20 30 40 50 60 70 80 90 100 110 120 130
1.00 1.13 (0.96, 1.33) 1.16 (0.98, 1.38) 1.17 (1.00, 1.37) 1.19 (1.02, 1.39) 1.22 (1.04, 1.43) 1.26 (1.08, 1.47) 1.32 (1.12, 1.55) 1.38 (1.14, 1.68) 1.46 (1.13, 1.89) 1.55 (1.11, 2.17) 1.66 (1.07, 2.55) 1.77 (1.03, 3.06) 1.90 (0.98, 3.66)
ln(OR) (95% CI) 0 0.122 (−0.041, 2.85) 0.148 (−0.020, 0.322) 0.157 (0.000, 0.315) 0.174 (0.020, 0.329) 0.199 (0.039, 0.358) 0.231 (0.077, 0.385) 0.278 (0.113, 0.438) 0.322 (0.131, 0.519) 0.378 (0.122, 0.637) 0.438 (0.104, 0.775) 0.507 (0.068, 0.936) 0.571 (0.030, 1.118) 0.641 (−0.20, 1.297)
SEb
weightc
adjusted weightd
0 0.083
0 14.5
0.0035
0.087
52.5
0.0125
0.080
139.5
0.0333
0.079
256.7
0.0613
0.081
378.8
0.0905
0.079
582.0
0.1390
0.083
713.2
0.1704
0.099
654.0
0.1562
0.131
470.4
0.1124
0.171
341.9
0.0817
0.222
246.5
0.0589
0.278
186.6
0.0446
0.336
149.6
0.0357
OR(THM4) = e THM4 × 0.00427
Based on this function and assuming average pre-Stage 1 exposure at 38.05 μg/L which EPA estimated in its EA for the Stage 2 DBPR,60 baseline risk (r0 = lifetime bladder cancer risk with zero THM4 exposure) can be derived as follows: OR38.05μg/L = e38.05 × 0.00427
a Odds Ratios and confidence intervals were provided by Kogevinas and Villanueva.63 bStandard Error (SE) = Interval width divided by 3.92. cWeight = ((TTHM4/SE)2)/1000. dAdjusted weight is the weight divided by the sum of the weights.
pre‐Stage 1 odds/baseline odds = 1.176
0.02459/(r0/(1 − r0)) = 1.176
Solving for ro:
r0 = 0.02047 where baseline odds are associated with exposure to 0 μg/L. Baseline odds is therefore 0.02047/(1−0.02047) = 0.02090. From the function relating ORs to THM4, the additional risk posed by a 1 μg/L increase in THM4 (r1) will depend on the initial conditions (THM4 exposure level and lifetime bladder cancer risk). The additional risk will be smallest for the case with zero THM4 exposure (r0 = 0.02047): r1/(1 − r1) = 0.02090 × e1 × 0.00427 = 0.02099 r1 = 0.02099/(1 + 0.02099)
r1 = 0.02056 Figure 1. Bladder cancer risk from six epidemiological studies as a function of THM exposure from residential concentration data (derived from personal communication with Kogevinas and Villanueva63).
r1 − r0 = 0.02056 − 0.02047 = 0.00009
If lifetime risk is greater than r0, then the risk associated with a 1 μg/L increase is slightly increased. Consider the case where the average THM4 level changes from the pre-Stage 1 mean of 38.05 μg/L to 39.05 μg/L and pre-Stage 1 lifetime risk is r38.05 = 0.024. From above, the pre-Stage 1 odds is 0.0246 and the additional risk is calculated as shown below:
The values in Table 1 were weighted equally, although the confidence limits show that uncertainty varies across the range of THM4 exposure. To more properly account for this, we derived a weighted mean slope as follows: • The response variable was changed to ln(OR). As shown in Figure 1, this normalized the errors so that the upper and lower confidence limits are symmetric about ln(OR). • The weight used for each record (row) of Table 1 is based on its uncertainty, expressed as standard error. The standard error (s) is derived from the confidence interval width:
r39.05/(1 − r39.05) = 0.0246 × e1 ×\ 0.00427 = 0.0247
r39.05 = 0.0241
The additional risk due to the 1 μg/L increase is 0.0241− 0.024 = 0.0001 or 10 −4 . The simple relationship of approximately 10−4 added lifetime risk per 1 μg/L increase in THM4 holds over the THM4 range expected for systems in 13097
DOI: 10.1021/acs.est.5b03547 Environ. Sci. Technol. 2015, 49, 13094−13102
Policy Analysis
Environmental Science & Technology compliance with Stage 2 (see Table S1). We believe this is the smallest increase that one would reasonably consider, given that the magnitude of uncertainty is difficult to estimate. Data Analysis IITHM4 Increases from Increased Bromide Levels. We used the Surface Water Analytical Tool (SWAT), version 1.1 (see SI for further information) that the EPA used in its Stage 2 EA60,64 to assess the impacts of an increase of bromide levels in source waters on THM4 levels in public drinking water distribution systems. This version of SWAT was used by the EPA to estimate treatment and associated water quality changes for different regulatory options. SWAT uses source water and treatment data collected under the Information Collection Rule (ICR)65 from July 1997 through December 1998 from 273 large plants using surface water sources representing a total population of over 100 million people contained in the AUX8 ICR data set. An overarching consideration for our analysis was to estimate the potential impact of increasing source water bromide on a subset of these plants that were assumed to be in compliance with the Stage 2 DBPR. Post-Stage 2 DBPR national level data sets are not available. Therefore, we attempted to utilize the AUX8 ICR data set in such a way as to include plants that, based on their ICR data and SWAT modeling, would likely have met the Stage 2 DBPR MCL of 80/60 μg/L (THM4/HAA5) at maximum residence time sample sites in the distribution system. In other words, if a plant was predicted (SWAT model predictions of THM4 and HAA5, based on ICR data inputs) to be within the annual average 80/60 compliance metric, then that plant would not likely make treatment changes (e.g., switch to chloramines) and would be operating today as it did during the ICR. Based on these criteria, 201 plants, comprising 2089 plant months, were identified and included in the analysis (see SI). For the purpose of our modeling, we assumed that any ICR data entry for bromide concentration that was below the minimum reporting level (MRL) was equivalent to the reporting limit (i.e., 20 μg/L). This is consistent with the EPA’s approach when conducting SWAT modeling to support the Stage 2 EA. The modeling with an assumption of half of the reporting limit is included in the SI to illustrate a sensitivity analysis of different values assigned to those below MRL. We also used the THM4 values predicted by SWAT at the average residence time in the distribution system (vs the maximum residence time, which would likely overestimate mean exposure for the entire residential population served by the plant) to estimate increases in THM4 concentration in treated drinking water (i.e., concentrations that would occur in households) as a function of increased bromide concentration in the source water. This assumption should better represent most of the exposed population for a given system. To predict the impact of increased bromide on THM4 formation, for each of the plant months, we increased the source water bromide concentration by different increments (see Table 2) while keeping the other THM4 formation equation input parameters unchanged from their original ICR values. Changes in THM4 concentrations from increases in influent bromide (denoted as “ΔBr−”) are denoted as “ΔTHM4.” Table 2 shows the summary statistics of ΔTHM4 as a function of ΔBr− predicted by the SWAT, indicating the predicted ΔTHM4 concentrations consistently increased, on average, by roughly an order of magnitude less than the increase in bromide. Such predicted results are consistent with a theoretical analysis that a simple replacement of chloride atom by bromide atom leads to a higher mass-based
Table 2. THM4 increases (ΔTHM4) as a Function of Increased Bromide Levels in Source Water (ΔBr−) ΔBr− (μg/L) ΔTHM4 (μg/L) (plant months)
statistics
10
30
50
75
100
mean minimum lower 95% percentile median upper 95% percentile maximum
1.3 0.0 0.1
3.2 0.0 0.3
4.6 0.0 0.5
6.0 0.0 0.6
7.1 0.0 0.8
1.1 3.4
2.6 8.3
3.7 11.6
4.9 14.8
5.8 17.5
10.1
23.7
33.2
42.1
49.3
concentration of DBPs due to a higher molecular weight of bromide. However, there is a diminishing rate of increased ΔTHM4 as the ΔBr− increases, indicating that organic precursors could be a limiting factor for formation of THMs with higher bromide concentration. For instance, the mean ΔTHM4 is 1.3 μg/L at the ΔBr−of 10 μg/L (i.e., 0.13 μg/L ΔTHM4 per μg/L ΔBr−), compared to the mean ΔTHM4 7.1 μg/L at the ΔBr− of 100 μg/L (i.e., 0.07 μg/L ΔTHM4 per μg/L ΔBr−). Table 2 also indicates that ΔTHM4 at a given ΔBr− varies across the 2089 plant-months. This variability is due not just to the variability in baseline bromide, but due to variations of other factors affecting THM4 formation, including source water quality (total organic carbon (TOC), ultraviolet absorbance at wavelength of 254 nanonmeters (UV254), temperature), operational conditions (disinfectant type, disinfectant dose, pH adjustments, changes in TOC and UV254), and residence time in the distribution system, characteristics essentially inherent in the SWAT model. These variations can occur from plant to plant, or within a specific plant due to temporal fluctuations. Given the wide range of initial baseline bromide or THM4 concentrations among the 201 plants examined, ranging from
