Environmental Protection
Agency
Issue Paper: Inorganic
Arsenic Cancer Slope Factor
The July 22, 2005, draft final Issue Paper: Inorganic Arsenic Cancer Slope Factor do not represent final
Agency policy or decisions.
-------
DRAFT Material
Issue Paper: Inorganic Arsenic Cancer
Slope Factor
EXECUTIVE SUMMARY V
INTRODUCTION 1
Background 1
ISSUE 1: CHOICE OF CHRONIC ENDPOINTS 3
EPA's Drinking Water Risk Assessment (2001) 3
NRC 2001 Recommendation 3
Workgroup Conclusion 3
ISSUE 2: INCLUSION OF NON-CANCER ENDPOINTS 4
Workgroup Conclusion 4
ISSUE 3: DETERMINATION OF THE DATA SET TO BE USED FOR MODELING. 5
EPA's Drinking Water Risk Assessment (2001) 6
Newly published related papers 6
Workgroup Conclusion 11
ISSUE 4: CHOICE OF ED0i MODEL 12
EPA's Drinking Water Risk Assessment (2001) 12
NRC 2001 Recommendation 13
Workgroup Conclusion 13
ISSUE 5: LINEAR EXTRAPOLATION TO LOW DOSE 14
EPA's Drinking Water Risk Assessment (2001) 14
NRC 2001 Recommendation 15
Workgroup Conclusion 15
-------
DRAFT Material
ISSUE 6: USE OF A COMPARISON POPULATION 16
EPA's Drinking Water Risk Assessment (2001) 16
NRC 2001 Recommendation 16
Workgroup Conclusion 17
ISSUE 7: BACKGROUND CANCER RATE 18
EPA's Drinking Water Risk Assessment (2001) 18
NRC 2001 Recommendation 18
ISSUE 8: ADJUSTMENT FOR DIETARY INTAKE OF ARSENIC (FOOD) 19
EPA's Drinking Water Risk Assessment (2001) 19
NRC 2001 Recommendation 19
Workgroup Discussion 19
Review of New Studies 19
ISSUE 9: ADJUSTMENT FOR WATER INTAKE (FROM DRINKING WATER ONLY)
22
EPA's Drinking Water Risk Assessment (2001) 22
NRC 2001 Recommendation 22
Review of New Studies 23
Workgroup Conclusions 24
ISSUE 10: ADJUSTMENT FOR DIETARY WATER INTAKE (FROM COOKING
WATER) 26
EPA's Drinking Water Risk Assessment (2001) 26
NRC 2001 Recommendation 26
Workgroup Conclusions 26
ISSUE 11: ADJUSTMENT FOR MORTALITY VS. CANCER INCIDENCE 27
EPA's Drinking Water Risk Assessment (2001) 27
NRC 2001 Recommendation 27
Workgroup Conclusion 27
-------
DRAFT Material
ISSUE 12: CHILD SENSITIVITY ISSUE 28
Working Group Conclusion 28
IV
-------
DRAFT Material
Executive Summary
Since publication of the 2001 NRC report, an intra-Agency workgroup has examined the NRC
recommendations and their potential influence on regulation of inorganic arsenic by the various program
offices. The Arsenic Cancer Slope Factor Workgroup consists of members of the Office of Water, the Office of
Research and Development, the Antimicrobial Division and the Health Effects Division of the Office of
Pesticide Programs, the Office of Children's Health Protection, the Office of Solid Waste, and the Office of
Emergency Response and Remediation. Over the course of eight meetings (April 30, 2003; June 10, 2003;
Aug 12, 2003; Sept 17, 2003; March 10, 2004; Oct 25, 2004; Nov 9, 2004; and November 30, 2004), the
workgroup discussed twelve issues:
1. The choice of chronic endpoints
2. The use of non-carcinogenic endpoints
3. Data to use for modeling
4. The model to use to calculate risk estimates
5. The method of extrapolating data to low doses
6. The use of a comparison population
7. The U.S. lung and bladder cancer background incidence rates
8. Dietary intake of arsenic in Taiwan
9. Consumption of drinking water in Taiwan
10. Consumption of cooking water in Taiwan
11. Adjusting cancer mortality data for incidence
12. Child Sensitivity Issue
This paper summarizes EPA's Drinking Water risk assessment (January 2001), the NRC's 2001
recommendations, and the Arsenic Cancer Slope Factor Workgroup's discussions. In addition, a brief
summary of reports published since the NRC 2001 recommendation is presented.
A summary table showing the issues is shown in Table 1. In general, the workgroup agrees with the
recommendations made by the NRC. Issues for which the workgroup could not reach agreement were
regarding the dietary intake of arsenic in Taiwan, and the consumption of drinking water in Taiwan.
The July 22, 2005, draft final Issue Paper: Inorganic Arsenic Cancer Slope Factor do not represent final
Agency policy or decisions.
-------
Table 1. Summary of Issues and Recommendations
Study
Parameter
1. Choice of Endpoint
2. Inclusion of Non-
Cancer Endpoints
3. Choice of Study
4. Model Choice
5. Linear
Extrapolation to Low
Dose
6. Selection of
Comparison Group
7. Background Cancer
Rate for U.S.
Population
EPA / OW
Lung and bladder cancer
Brief review
Southwestern Taiwanese cancer mortality
data from Chen et al. (1985, 1988, 1992)
Poisson regression model with exponential
linear function of dose anda quadratic
function for age (multiplicative). This
model was chosen because it fit the Akaike
information criteria best.
Linear extrapolation from EDoi
No comparison group used, because the
model that fit the Akaike information
criteria best did not include a comparison
group, and because models with
comparison groups led to supralinear dose-
response relationships.
U.S. background cancer rates are the same
as those in Taiwan as found in You et al.
(2001)
NRC Subcommittee
Lung and bladder cancer
Brief review
Southwestern Taiwanese data from Chen et al. (1985, 1988, 1992)
as primary source; Northwestern Taiwanese data (Chiou et al.
2001) and Chilean lung cancer incidence data (Ferreccio et al.
2000) as supplementary sources
Poisson regression model with linear function of dose (additive).
This model was chosen because it led to the best fit when used
with data from Chiou et al (2001).
Linear extrapolation from EDoi
External comparison group recommended, because data for an
external comparison group exists and because Tsai et al. (1999)
demonstrated that the standardized mortality ratios derived from
the southwestern comparison group are similar to those using the
full Taiwanese population as a comparison group.
Compared to Taiwanese background cancer rates, U.S. background
cancer rates are 1.6 times and 2 times greater for males and
females, respectively. This is based on comparing data from You
et al. (2001) and SEER (2001).
Workgroup
Same as NRC recommendation
For carcinogenic analysis, the
agency will take a qualitative
approach in analyzing non-cancer
effects.
Should use the southwestern
Taiwan data set presented in the
1999 NRC report.
Poisson regression model with
linear function of dose (additive).
Same as NRC recommendation
Same as NRC recommendation
Same as NRC recommendation
-------
DRAFT Material
Study
Parameter
EPA / OW
NRC Subcommittee
Workgroup
8. Adjustments for
Dietary Intake of
Arsenic in Southwest
Taiwan.
Assumed that the Taiwanese ingested 50
ug/day of arsenic from food sources.
Noted that the EPA's value of 50 ug/day was not based on any
scientific data. NRC provided no scientific basis for substituting 30
Ug/day as the alternate for the amount of arsenic ingested from
food sources. NRC concluded that the calculation of ED was not
sensitive to this value.
Regulatory background material
provided the basis for 50 ppb from
food, and some current Asian diets
contain much more than 50 ppb
arsenic. However, some workgroup
members disagreed about the
historic dietary intake of arsenic.
All participants agreed that the
ramifications on the risk
assessment of choosing one value
over the other would be very small.
The workgroup concluded that the
model should be run over a range
arsenic consumption rates
considered by NRC (30 ug/day to
50 ug/day) to confirm that the
calculated risk is insensitive to this
value.
9. Adjustment for
Drinking Water Intake
in Southwest Taiwan
and U.S.
U.S. population: Used lifetime intake
values of 1 L/day and 1.2 L/day in Monte
Carlo analysis based on CSFII (EPA 2000)
water intakes. Taiwan population: assumed
drinking water consumption is 3.5 L for
males and 2.0 L for females plus cooking
water.
Does not recommend any particular drinking water consumption
rates. However, they note that consideration should be given to
variations of drinking water consumption in a given village
population and to the quantity of water consumed, as the
calculation of ED is sensitive to both of these parameters.
The workgroup's review of relevant
literature suggests that the mean
drinking water consumption rate is
between 1 to 4.6 L/day.
10. Adjustment for
Cooking Water Intake
in Southwest Taiwan
and U.S.
Assumed the Tawainese consumed 1 L/day
of cooking water. Water intake accounted
for Americans having 0.49 to 0.54 L water
as cooking water.
Agreed that cooking water should be considered in the
assessment, but noted that EPA did not document their rationale
for using 1 L/day.
The use of 1 L/day is supported by
current literature and EPA's 1988
work group files.
11. Adjustments for
Cancer Mortality vs.
Cancer Incidence
Used Taiwanese mortality data for bladder
and lung cancers. Assumed that 80% of
bladder cancer incidences lead to mortality
and 100% of lung cancer incidences lead
to mortality. Calculated risk of Taiwanese
incidence to derive U.S. incidence. EPA
assumed 26% U.S. diagnosed bladder
cancers and 88% U.S. lung cancers fatal.
Accepted approximation that 100% lung cancers and 80% bladder
cancers were fatal in Taiwan. NRC said that assuming U.S.
bladder cancers are 80% fatal is not appropriate (pg. 199). For
U.S., assumed 21% of bladder cancer incidences and 80% of lung
cancer incidences lead to mortality, based on SEER (2001) data.
Believe that NRC misunderstood
EPA assumptions. However,
difference in U.S. incidence is not
important if relative risk, rather
than absolute risk, is to be
calculated.
12. Child Sensitivity
Issue
No Adjustment is applied "because
carcinogenic effects... are evaluated based
on a lifetime of exposure, which takes into
consideration the elevated dose that
occurs in children (EPA economic
analysis)."
NRC noted that "the evidence is not conclusive" for adverse
reproductive effects and arsenic. Further, the studies on neuro-
cognitive functions have potential confounders. However, "[i]f the
peak concentration represents the most appropriate dose metric of
concern, then use of a lifetime cancer risk model would not be
valid (NRC 2001 at 151).
Children are already addressed in
the epidemiological cancer studies.
Data do not support children's
susceptibilities at this time
-------
Introduction
Background
In 1975, the U.S. Environmental Protection Agency (EPA) adopted a drinking water regulation for arsenic
based on a U.S. Public Health Service standard set in 1942. The drinking water standard of 50 micrograms
per liter (ug/L), which is equivalent to 50 parts per billion (ppb), remains in effect until 2006. EPA conducted
risk assessments for arsenic-induced skin cancer in 1980, 1988, and 1992. Currently, the Agency's Integrated
Risk Information System (IRIS) carcinogenic risk from oral exposure to arsenic is based on southwestern
Taiwanese skin cancer studies published in 1977 and 1968. IRIS estimates the one-in-a-million additional
skin cancer risk to be 0.02 ppb arsenic in drinking water.
In 1996, EPA charged the National Academy of Sciences (NAS) to review the Agency's characterization of
potential health risks from ingestion of arsenic; the available data on carcinogenic and non-carcinogenic
effects of arsenic in drinking water; the data on metabolism, kinetics, and mode(s) of action of arsenic; and
research priorities. In response, National Research Council (NRC) issued a report in 1999. NRC used data
from Wu et al. 1989 and Chen et al. 1992 to address several risk assessment issues.
EPA applied many of the recommendations from the 1999 NRC report in the risk characterization used to
support the January 2001 revised arsenic drinking water regulation. The Agency based its new 10 ppb
arsenic standard on the risk of bladder and lung cancers from the Taiwanese data used by NRC and estimated
1 - 6 x 10"4 risk to the 90th percentile of the U.S. population.
In April 2001, EPA charged the NRC to review the risk analysis used to support the revised arsenic drinking
water regulation in light of studies published since the 1999 NRC report. NRC released its update report in
September 2001.
Since publication of the 2001 NRC report, an intra-Agency workgroup has examined the NRC
recommendations and their potential influence on regulation of inorganic arsenic by the various program
offices. The Arsenic Cancer Slope Factor Workgroup consists of members of the Office of Water, the Office of
Research and Development, the Antimicrobial Division and the Health Effects Division of the Office of
Pesticide Programs, the Office of Children's Health Protection, the Office of Solid Waste, and the Office of
Emergency Response and Remediation. Over the course of eight meetings (April 30, 2003 to November 30,
2004), the workgroup discussed twelve issues:
1. The choice of chronic endpoints
2. The use of non-carcinogenic endpoints
3. Data to use for modeling
4. The model to use to calculate risk estimates
5. The method of extrapolating data to low doses
6. The use of a comparison population
7. The U.S. lung and bladder cancer background incidence rates
8. Dietary intake of arsenic in Taiwan
9. Consumption of drinking water in Taiwan
10. Consumption of cooking water in Taiwan
11. Adjusting cancer mortality data for incidence
-------
DRAFT Material
12. Child Sensitivity Issue
This paper summarizes EPA's Drinking Water risk assessment (January 2001), the NRC's 2001
recommendations, and the Arsenic Cancer Slope Factor Workgroup's discussions. In addition, a brief
summary of some studies published since the NRC 2001 recommendation is presented.
-------
DRAFT Material
Issue 1: Choice of Chronic Endpoints
There are a number of toxic effects associated with the long-term
Issue: ingestion of inorganic arsenic. Which of these effects should be
used in the risk assessment?
Workgroup Conclusion: Lung and bladder cancers were chosen as the chronic endpoints.
EPA's Drinking Water Risk Assessment (2001)
EPA based its quantitative risk assessment on lung and bladder cancers, consistent with the 1999 NRC report.
The 1999 NRC report concluded that there was sufficient evidence to conclude that ingestion of arsenic in
drinking water causes skin, bladder, and lung cancer. The internal cancers (bladder and lung) were
considered to be the main cancers of concern, based on the NRC 1999 report, which presents the best
science as of its completion, and more recently published studies.
NRC 2001 Recommendation
NRC 2001 noted that new studies link arsenic to hypertension, diabetes, adverse reproductive effects,
respiratory effects, skin lesions, and cognitive effects. However, the epidemiological studies of bladder and
lung cancer provide "a sound and adequate basis for quantitative assessment of cancer risk (NRC 2001, p.
68)." The NRC noted that additional studies would be needed to assess the association of arsenic exposure
with other health effects.
Workgroup Conclusion
The workgroup participants concurred in using bladder and lung cancer for risk assessments.
-------
DRAFT Material
Issue 2: Inclusion of Non-Cancer Endpoints
There are a number of toxic effects associated with the long-term
Issue: ingestion of inorganic arsenic. Which of these effects should be
used in the risk assessment?
, . _ . . Recommends a qualitative approach in considering non-cancer
Workgroup Conclusion: , . .
endpomts.
In 1999, NRC recommended additional study of arsenic-associated skin, cardiovascular, cerebrovascular
effects as well as diabetes and adverse reproductive outcomes. However, in August 2001, EPA's Science
Advisory Board recommended that EPA quantify the benefits of hypertension and diabetes mellitus. In its EPA
2001 risk assessment, EPA listed cardiovascular, pulmonary, immunological, neurological, and endocrine
effects as potential yet nonquantifiable health benefits. In 2001, NRC affirmed that "cancer represents the
most sensitive health endpoint.(NRC 2001, p. 151)", further noting that hypertension, cardiovascular disease,
cerebrovascular disease and diabetes need additional study, as well as reproductive and developmental
effects, and respiratory function.
Workgroup Conclusion
The workgroup agreed that these non-cancer effects will be assessed quantitatively in the non-carcinogenicity
assessment (e.g., IRIS and OPP RED). There was no objection among the workgroup participants that for
carcinogenic analysis, the agency will take a qualitative approach in analyzing non-cancer effects.
-------
DRAFT Material
Issue 3: Determination of the Data Set to be Used for Modeling
Which data set should be used in the quantitative risk assessment
' for long-term exposure to arsenic in drinking water?
, . _ . . The southwestern Taiwan data set presented in the 1999 NRC
Workgroup Conclusion: . . . . .
report should be used.
An important decision in a quantitative risk assessment is the choice of critical study to be used in the dose-
response assessment. NRC reviewed a number of data sets as potential candidates for use in dose-response
modeling for carcinogenic effects, including data from northeastern Taiwan, Chile,and southwestern Taiwan.
Short descriptions of the studies from these regions are provided below. More detailed descriptions of these
studies are provided in Chapter 2 of the NRC 2001 report. In addition, facts such as gender, differences in
diet, cultural background, smoking, occupational exposures, and access to medical care were considered.
Northeastern Taiwan
Chiou et al. (2001) conducted a prospective cohort study of 8,102 persons in the Lanyang basin of
northeastern Taiwan. The primary drinking water supply of all cohort members were wells that had been
contaminated with arsenic from the late 1940s through the mid-1990s. Each subject was monitored for
cancer incidence for 3 to 6 years (the study period was from 1991 to 1996). The study report included
analyses of total urinary cancer rates (including kidney, bladder, and urethral cancer), and of transitional cell
carcinoma (TCC).
Multivariate relative risks were calculated by comparing the results to those of a nonexposed reference group
with arsenic exposures of 10 ug/L or less. The study concluded that the increase in arsenic-induced TCC was
more prominent for those individuals who were exposed to the contaminated drinking water for more than 40
years. The adjusted relative risk of TCC in the highest exposure category (>100.0 ug/L) was 15.1. However,
due to the short duration of follow-up, it only recorded a small number of cancer cases, this value has a
large confidence interval (95% CI = 1.7 - 138.5). NRC (2001 at p. 49) noted that "risk estimates based on
these data might be too imprecise for use in a quantitative risk assessment. However, the data can serve as
supplementary information, along with data from other selected studies."
Chile
Ferreccio et al. (2000) conducted a case-control study of incident lung cancer and bladder cancer in a region
of Chile with a history of increased concentrations of arsenic in drinking water. For each participant, two
controls were selected that matched the participant's sex and closely matched the participant's age. In
addition, the authors attempted to select representative of arsenic exposure in the overall . NRC (2001) pg.
51 said the highest exposure had too many controls, which would tend to underestimate risk, while the next
category (100-300 ppb) having less controls, would increase risk estimates.
Odds ratios were used to estimate the relative risk of exposure to various concentrations of arsenic in
drinking water relative to a referent concentration of 0-10 ug/L Results from the analysis show an increase
in the odds ratio with arsenic concentration, reaching an odds ratio of 8.9 for the highest exposure group
(200-400 ug/L). However, the odds ratio is underestimated for the exposure 200-400 and overestimated for
the next lower category because of the distribution of controls. There was evidence suggestive of a
synergistic interaction between smoking and exposure to arsenic.
-------
DRAFT Material
Southwestern Taiwan
(1) Tsai et al. (1999) evaluated mortality due to several cancer and non-cancer causes in a region of
southwestern Taiwan where blackfoot-disease was endemic. The study was conducted between 1971
through 1994. The area had high concentrations of arsenic in the drinking water in the past. SMRs were
calculated twice, using two referent groups. The first referent was the mortality experience of the whole
of Taiwan. The second referent was the mortality experience of two counties in southwestern Taiwan
where blackfoot disease was endemic. The SMRs for two groups were used to determine the effects of
inorganic arsenic consumption. In addition, factors such as gender, differences in diet, cultural
background, smoking, occupational exposures, and access to medical care were considered.
EPA's Drinking Water Risk Assessment (2001)
EPA used the southwestern Taiwan data provided by NRC 1999 in its quantitative risk assessment for
exposure to arsenic in drinking water.
NRC 2001 Recommendation
The NRC noted that the southwestern Taiwanese ecological studies are the strongest sources of dose-
response information for cancer endpoints. However, they suggested consideration of the northeastern
Taiwanese urinary cancer prospective cohort study (Chiou et al. 2001) and the Chilean lung cancer case-
control study (Ferreccio et al. 2000) for use in a quantitative risk assessment for the following reasons:
Chiou et al. 2001:
1. Large number of subjects;
2. Exposure levels close to concentrations of current regulatory concern (more than half the study
subjects were exposed to concentrations less than 50 ug/L);
3. Data are available at the individual level on potential confounding factors, such as smoking and
socioeconomic status; and
4. Urinary-tract cancer is the endpoint of interest
Ferreccio et al. 2000:
1. Individual information available on residential history, socioeconomic status, occupational history,
and smoking; and
2. Detailed, individual-specific exposure assessments.
Newly published related papers
Since publication of the 2001 NRC report, several published papers may present alternative ways in analyzing
the Taiwan.
(l)Lamm SH, Byrd DM, Kruse MB, Feinleib M, Lai S. Bladder Cancer and Arsenic Exposure: Differences
in the Two Populations Enrolled in a Study in Southwest Taiwan. Biomedical and Environmental
Sciences 2003; 16:355-368
(2) Lamm SH, Engel A, Kruse MB, Feinleib M, Byrd DM, Lai S, Wilson R. Arsenic in Drinking Water and
Bladder Cancer Mortality in the U.S.: An analysis based on 133 U.S. counties and thirty years of
observation. Journal of Occupational and Environmental Medicine 2004; 46(3):298-306
-------
DRAFT Material
(3)Steinmaus C, Yuan Y, Bates M, Smith A. Case-Control Study of Bladder Cancer and Drinking Water
Arsenic in the Western United States. American Journal of Epidemiology 2003; 158:1193-1201
(4) Tollestrup K, Frost FJ, Harter LC, McMillan GP. Mortality among children residing near the American
Smelting and Refining Company (ASARCO) copper smelter in Ruston, Washington. Arch Environ
Health. 2003 Nov;58(ll):683-91.
(5)Chen CL, Hsu LI, Chiou HY, Hsueh YM, Chen SY, Wu MM, Chen CJ. Ingested arsenic, cigarette
smoking, and lung cancer risk follow-up study in arseniasis-endemic areas in Taiwan. JAMA. 2004
Dec 22;292(24):2984-90.
The Agency reviewed these five papers. The summary and review are listed below:
Lamm SH, Byrd DM, Kruse MB, Feinleib M, Lai S. Bladder Cancer and Arsenic Exposure: Differences
in the Two Populations Enrolled in a Study in Southwest Taiwan. Biomedical and Environmental
Sciences 2003; 16:355-368
Summary:
This study was a reanalysis of the underlying data used by the National Research Council (1999, 2001), and
the U.S. Environmental Protection Agency (2001) to assess carcinogenic risk from arsenic ingestion. The
original data were from studies by Wu et al. (1989) and Chen et al. (1992) with a reanalysis and risk
assessment from Morales et al. (2000). Based upon median arsenic concentrations of wells within the
villages of Southwest Taiwan, Lamm et al. (2003) categorized villages by water source as either artesian
wells, shallow wells, or mixed, as suggested in earlier papers by Chen et al. (1985). Lamm et al. (2003)
included this categorization in their re-analyses examining drinking water arsenic and bladder cancer
mortality. Although a significant dose response relationship was confirmed for the artesian wells, no such
trend was observed for the other two well types, which had arsenic concentrations generally less than 325
ppb. The authors concluded that previous risk assessments may have been flawed by failing to account for
well type in the analysis. Two possibilities were suggested to explain the results: 1.) the trend was only
present at high exposures; 2.) the presence of a co-carcinogen in the artesian wells that was not present in
the shallow wells.
EPA Reviewers' Comments:
Lamm et al. (2003) reclassified wells based solely on the median arsenic concentration of wells within a
village. Specifically, the authors artificially classified village well types into three categories (shallow, <
0.325 ppm; mixed, wells above and below 0.325 ppm; and artesian, > 0.325 ppm) according to the
arsenic concentration in the well water. The validity of Lamm's reclassification is impossible to assess
with the information provided. Assuming that all the authors had was the information in Chen et al.
(1985), the validity of the classification is suspect.
Age is an important risk factor related both to duration of exposure and cancer risk, but age was not
considered in the Lamm et al. (2003) re-analysis. Although the authors acknowledge that their analysis
does not account for age or smoking as an influence because the data are not available to the public, it
is evident from past literature that age and smoking can dramatically affect the onset of adverse health
effects due to arsenic exposure (Tseng et al., 2000, 2003) and is commonly adjusted for in most
epidemiological studies.
In this manuscript, Lamm et. al. (2003) used the fact that the Taiwanese studies (Wu et al., 1989;
Tseng et al., 2000) of arsenic and bladder cancer were ecologic to suggest that factors other than
arsenic may be responsible for bladder cancer. Little evidence exists for the co-carcinogen explanation.
Previous studies have reported little difference in the chemical make-up of the wells that could explain
the results. The presence of another bladder carcinogen in well water seems unlikely.
-------
DRAFT Material
Lamm SH, Engel A, Kruse MB, Feinleib M, Byrd DM, Lai S, Wilson R. Arsenic in Drinking Water and
Bladder Cancer Mortality in the U.S.: An analysis based on 133 U.S. counties and thirty years of
observation. Journal of Occupational and Environmental Medicine 2004; 46(3):298-306
Summary:
This ecologic study examined white male bladder cancer mortality in the U.S. in relation to drinking water
arsenic concentrations. This study was designed to be analogous to the Wu et al. (1989) Southwest Taiwan
Study. Arsenic exposures were based on U.S. Geological Survey county-specific data. U.S. counties were
evaluated for inclusion in the study if they exclusively used groundwater as source of drinking water and had
measured arsenic levels >3 ppb (n=268 counties, arsenic range 3-60 ppb). These exposures were
considerably lower than those in the previous studies conducted in Taiwan. White male bladder cancer
mortality data from 1950-1979 were extracted from the NCI/EPA (1983). Of the 268 counties identified from
the U.S.GS, 133 counties had at least one male bladder cancer death. More than 4500 white male bladder
cancer deaths were observed across those 133 counties for the 30-year period. County level standardized
mortality ratios (SMR) were calculated. No increase in SMR was observed as the median arsenic level of the
county increased.
EPA Reviewers' Comments:
Serious limitations in this ecological study exist with not enough detail or acknowledgement in the
discussion by the authors. Exposure assignment was likely a major source of misclassification, even at
the county level. A major assumption of this study was groundwater arsenic concentrations for a
county were stable over time. The U.S.GS exposure information was collected between 1973-1998 with
the majority of groundwater arsenic measures obtained after 1980. Groundwater arsenic concentrations
may not be stable over time, and no evidence is provided to support this assumption. Also, little
variance was observed across counties in median groundwater arsenic concentrations. Only two
counties had median arsenic concentrations above 50 mg/L, five counties had exposures above 20
mg/L; and only 15 counties over 10 mg/L. The statistical power to observe SMRs below 1.7 is severely
limited in the counties with arsenic greater than 50 mg/L. The power to detect an SMR of 1.5 in these
counties is 46%.
Individual heterogeneity with respect to consumption of arsenic contaminated fluids was not taken into
account which could be due to migration, use of alternative water sources, and simple variation in
intake. Although the Steinmaus et. al. (2003) analysis demonstrates how important these variables are
in evaluating the effect of low level arsenic exposure, individual information on age, duration of
exposure, latency, occupation, and smoking cannot be taken into account in an ecologic analysis.
Although the sample size may have been large enough to overcome the problems of studying a rare
disease, the averaging effect of ecologic studies such as this make it impossible to evaluate subgroups
in the population that may be more sensitive to low-level arsenic. Another averaging effect comes from
the use of person-years. Not incorporating age information implies that 100 person-years from five 20-
year old individuals would be the same as two 50-year old individuals. This averaging effect could have
significant impact on the results of this study.
Counties with zero cases were excluded from the analyses without any justification given. Presumably,
this was because a SMR could not be calculated for these counties. However, this exclusion may have
resulted in biases, especially if these counties were also areas with low arsenic exposure.
The method of calculating the expected numbers of cases was questionable. Expected deaths were
calculated by multiplying the ratio of the decade specific death rate for each county to the decade
specific death rate for the state by the observed number of deaths. With this unconventional method, it
was not clear as to whether Lamm et al. (2004) were referring to bladder cancer mortality rates or to
overall death rates. The use of the SMR was unnecessary and confusing. If the authors were indeed
referring to county level bladder cancer mortality rates in the ratio, why not compare the bladder rates
directly using a Poisson model?
-------
DRAFT Material
Steinmaus C, Yuan Y, Bates M, Smith A. Case-Control Study of Bladder Cancer and Drinking Water
Arsenic in the Western United States. American Journal of Epidemiology 2003; 158:1193-1201
Summary:
An incident case-control study of bladder cancer in relation to drinking water arsenic was conducted in seven
counties in the western United States. All incident cases of bladder cancer, which occurred between the
years 1994 and 2000 were identified and enrolled (n=181). Controls were frequency matched to cases by
age and gender and were selected either through random digit dialing or through the Health Care Financing
Administration rolls if the case was 365 years old. Lifetime exposure to drinking water arsenic was estimated
using 7000 measurements from community and domestic wells obtained from the Nevada and California state
health departments. Subjects were asked to estimate their drinking water intake as well as use of water
filters and bottled water over the last 60 years. If drinking water estimates were not obtained or could not
be estimated, drinking water estimates for that well were assigned a value of zero. No statistically elevated
risks were identified for arsenic intakes greater than 80 ppb, although there was some evidence of elevated
risks among smokers with a 40-year exposure lag at intakes greater than 80 ppb (OR=3.67, 95%CI 1.43-
9.42).
EPA Reviewers' Comments:
The reviewers have identified two issues that may have influenced the validity of the exposure
assessment in this study. First, the authors have chosen to use a default of 0 ng/day arsenic exposure
for several types of circumstances in the exposure assessment. These circumstances were as follows:
1) arsenic exposures for residences outside the study area were assigned a value of zero; 2) the use of
bottled water and water treated with a filter known to remove arsenic was assigned an arsenic level of
zero; 3) for residences where arsenic measurements could not be located or estimated, arsenic levels
were assigned a value of zero. For these default situations it is an appropriate assumption to treat
them in the same manner, but these default assumptions could skew the result towards the null
hypothesis if the cases or controls were classified as false-negatives.
Second, misclassification of arsenic exposure as well as recall bias may be present within this study.
Individual exposures were obtained by interview. Both cases and controls were asked to recall living
locations over their lifetime and recall the type of water as well as amount of water consumed so that
the authors could estimate a quantitative dose of arsenic over time. However, more cases than controls
(19% versus 6%) were interviewed using next of kin. The likelihood that a person can recall fluid
intake over lifetime (let alone have that recalled by the next of kin), considering the mean age for both
cases and controls was approximately 70 years, may be improbable.
The appropriateness of controls was questionable. Controls were significantly more likely to be in a
higher income bracket than cases. The authors state that since income was not a risk factor for bladder
cancer than this could not be a confounder. While true, the possibility that controls were a biased
selection of the study population still remains. If income were strongly associated with higher or lower
arsenic exposures, the results would be biased relative to the general population.
The most significant problem with this study may be the interpretation and conclusion. For a small
study with limited statistical power, a great deal of attention was made to the statistical significance of
results. Little attention was provided to other potential trends in the data, despite the lack of
significance. In Table 3, at a 40-year lag, elevated odds ratios compared to the referent group (under
10 ppb) were detected in all classifications of exposure, with the exception of 10-80 ppb in the highest
1-year and 5-year average. While attention to statistical significance is important, the elevated odds
ratios may also be relevant, particularly given the low power of the study. For example, the power to
detect an odds ratio of 1.78 for the >80 ppb category in the highest one year average for a 40-year lag
was only 46%.
-------
DRAFT Material
Tollestrup K, Frost FJ, Harter LC, McMillan GP. Mortality among children residing near the
American Smelting and Refining Company (ASARCO) copper smelter in Ruston, Washington. Arch
Environ Health. 2003 Nov;58(ll):683-91.
Summary:
Tollestrup et al. (2003) conducted a retrospective cohort study to determine whether childhood exposure to
ambient arsenic was associated with increased mortality rates. The cohort was comprised of children who
had lived within 2 miles of a copper smelter and arsenic refinery (American Smelting and Refining Company)
in Ruston, Washington, for at least 2 years from 1907-1932. The subjects were identified from school census
records, and included 1,827 boys and 1,305 girls with an age limit of 14 years. Exposure intensity was
calculated as the total number of days spent at a residence within 1 mile of the smelter stack, and grouped
by the number of years spent at the residence: 0 < 1.0 year, 1.0-3.9 years, 4.0-4.9 years, and > 10.0 years.
A total of 3,336 potential subjects were identified, and 196 were excluded because they had worked at the
smelter. Crude mortality rates were based on person-years of follow-up, and calculated for 10 general causes
of death. The highest crude mortality rate for boys was for ischemic heart disease in all exposure intensity
groups, but no evidence of a dose-response relationship was found. The 2nd highest mortality rate for boys
was for malignant neoplasms, with a range of 12.5/10,000 person-years to 21.9/10,000 person-years. A
dose-response was observed only for the mortality rate for "external causes," such as motor vehicle
accidents. Cox proportional hazard ratios adjusted for year of birth found only one exposure group (> 10.0
years) for which the mortality rations were significantly higher than 1.00. These included all causes of death
(1.52, 95% CI 1.23-1.86), ischemic heart disease (1.77, 95% CI 1.21-2.58), and external causes (1.93, 95%
CI 1.03-3.62). Although girls also had the highest crude mortality rates for malignant neoplasms and
ischemic heart disease, no dose-response relationships were observed. This study did not find consistent
patterns of adverse health effects from childhood exposure to ambient arsenic at levels much lower than
occupational settings.
EPA Reviewers' Comments:
A deficiency of the study was the truncation of the study period to 1932 which could result in
exposure misclassification.
Other limitations include ambiguous exposure data (exposures to arsenic were not chronic and were
unknown since air and soil levels were not quantified), poor follow-up (34.7% of boys and 46.5% of
girls were not found after their last date of exposure), the use of crude mortality rates, and lack of
information on smoking within the cohort and on family members who worked at the smelter and
could have brought arsenic into the household.
Chen CL, Hsu LI, Chiou HY, Hsueh YM, Chen SY, Wu MM, Chen CJ. Ingested arsenic, cigarette
smoking, and lung cancer risk follow-up study in arseniasis-endemic areas in Taiwan. JAMA. 2004
Dec 22;292(24):2984-90.
Summary:
Chen et al. (2004) investigated the relationship between ingested arsenic and lung cancer and the effect of
smoking on the relationship. A total of 2,503 residents in southwestern and 8,088 residents in northeastern
Taiwan were followed for an average period of 8 years. These were areas where residents had been drinking
well water contaminated with high concentrations of arsenic until the establishment of public water systems.
Questionnaires were administered to all participants in the study eliciting information on residential and
occupational history, history of drinking well water, cigarette smoking and alcohol consumption. Water
measurements taken in the 1960s of shared artesian wells in the southwestern area were used in conjunction
with information derived from the questionnaire to derive an average arsenic concentration for each
participant, which was used as an exposure metric in subsequent analysis. Average arsenic concentrations
for participants in the northeastern region, who derived their drinking water from shallow wells, were
determined by direct measurement of individual wells. The incidence of lung cancer was ascertained from
10
-------
DRAFT Material
national registry data for the period the period January 1985-December 2000. During the follow up period of
83,783 person-years, 139 lung cancer cases were diagnosed.
After adjusting for cigarette smoking and other risk factors such as age, alcohol consumption, and years of
schooling, a significant (p <0.001) increasing trend in lung cancer was shown to result from increasing
average levels of arsenic in well water. With levels <10 ug/L as the referent, relative risks (with 95%
confidence intervals) for those consuming drinking water with arsenic concentrations of 10-99, 100-299, 300-
699, and >700 ug/L, were respectively, 1.09 (0.63-1.91), 2.28 (1.22-4.27), 3.03 (1.62-5.69), 3.29 (1.60-
6.78). It was further shown that 32% to 55% of lung cancer cases were attributable to both arsenic
exposure and cigarette smoking. The synergism was shown to be additive; multiplicative interaction was not
statistically demonstrated.
Workgroup Conclusion
The workgroup agreed that the southwestern Taiwan study should be used as the primary data source. The
large confidence intervals (CI) on the Chiou et al. 2001 data and Ferreccio et al. 2001 data (CI values overlap
considerably) are not precise enough for quantified risk assessments
11
-------
DRAFT Material
Issue 4: Choice of ED0i Model
What model should be used to calculate the 1% effective dose
' (EDpi) and other risk estimates?
, . _ . . Poisson regression model with linear function of dose (additive)
Workgroup Conclusion: . . . . .
should be used.
The 1% effective dose (ED0i) is the dose at which 1% of the population is affected. In trying to determine a
dose-response curve, previous modeling attempts have started by estimating ED0i for exposure to inorganic
arsenic.
EPA's Drinking Water Risk Assessment (2001)
Morales et al. (2000) calculated cancer risk estimates using 10 risk models to estimate the risk of cancer
mortality. Morales et al. (2000) used large number of models to help determine the sensitivity of the models.
Poisson models were used in Morales et al. (2000) based on a recommendation from NRC (1999), which
suggested that Poisson models were more stable than other modeling methods that had been used previously
(mainly multi-stage Weibull models). After taking public comment on the models presented by Morales et al.
(2000), EPA selected one of the ten models to use for estimating bladder and lung cancer ED0i values. The
Morales et al. (2000) models were all based on the following general equation:
M = f(t)*g(d)
where:
M = Cancer hazard function
f(t) = a function describing the baseline risk in the absence of exposure to the chemical in question.
Although the baseline function can incorporate a number of factors (e.g., age, gender, smoking/non-smoking,
etc.), the Taiwanese data used by Morales et al. only contained gender and age (Q information.
g(d) = a function describing the effect of exposure to the chemical. The term d\s the chemical dose.
Models that use this general equation frequently fall into one of two categories: multiplicative models or
additive models. In a multiplicative model, the function g(d) takes the following general form:
where ct, c2, and c?are constants. If ct = 0, the multiplicative model is exponential linear; otherwise, it is
exponential quadratic. In contrast, an additive model takes the form:
g(d) = c^d2 + c2d + c3
where clt c2, and c?are constants. If d = 0, then additive model is linear; otherwise, it is quadratic.
In determining which of the ten models to use, EPA did not consider any model using an external comparison
group because such models tended to produce supralinear dose-response curves, and superlinear curves were
not considered to be sufficiently conservative (see Issue 5). The final model chosen was a quadratic function
of age and a linear dose with no dose transformation (multiplicative model). This model was chosen because
it did not use an external comparison population, and because the model best fit the Akaike information
criterion.
12
-------
DRAFT Material
NRC 2001 Recommendation
NRC notes that although the general Poisson approach is appropriate for calculating ED0i values, the
applications had some limitations. Table 5-3 in NRC (2001) illustrated the differences in ED0i values
calculated for the U.S. population for three studies using the Beir IV approach to calculate lifetime risk,
assuming that the Chilean, Taiwanese, and U.S. populations drank the same amount of water, with the first
two populations weighing 50 kg and assuming that the risks of cancer incidence were the same as the risks of
cancer mortality. (Table 5-5 shows the difference between the using different drinking water rates in Taiwan
and the U.S.) For the Chilean data (Ferreccio et al. 2000), NRC calculated ED0i values using linear regression
of relative risks. NRC used two multiplicative models (either with linear or log dose) and one additive model
with linear dose for both the northeastern Taiwan data (Chiou et al. 2001) and the southwestern Taiwan data
(Chen et al. 1985, 1992). "The BEIR IV formula allows a useful approach to computing an ED0i for the
United States based on relative risks obtained from a different population (NRC 2001 at p. 207)." In
contrast, Morales et al. (2000) calculated ED0i values directly from the dose-response function, which was
estimated using Taiwanese baseline risks.
In Tables 5-7 and 5-8 NRC calculated risk estimates for U.S. populations exposed to arsenic through drinking
water using U.S. background incidence rates (vs. Table 5-9 using the background incidence rate of Taiwan).
Tables 5-7 calculated U.S. lung cancer risk based on the Chilean and Taiwanese data. Table5-8 assumed that
the Taiwanese drank 1 L/day or 2 L/day. These estimates were determined using a Poisson model with
linear dose (additive model) and using the BEIR IV approach to apply relative risks from Taiwan to U.S.
baseline population risks. NRC notes that they calculated risks "using assumptions considered to be
reasonable by the subcommittee; it is possible to get higher and lower estimates using other assumptions."
"...[A] wide range of different models can be used to fit the arsenic carcinogenicity data currently available,
and no clear biological basis exists for distinguishing among them (NRC 2001 at p. 151)." The top three
models in Table 5-4 based on high PMP values are the multiplicative with quadratic dose model (used by OW
in the 2001 risk assessment), the additive log dose, and the additive linear dose. Based on low AIC values,
the top three models are the additive linear dose, the multiplicative with quadratic dose model, and the
additive linear dose. NRC concluded that the information known regarding mode of action does not "justify
the choice of any specific dose-response model..." and among the "reasonable model choices, the estimates
do not vary by more than an order of magnitude (NRC 2001, at pp. 207 and 208)." Finally, the recommended
model "is a biologically plausible model."
Workgroup Conclusion
In general, it is more difficult to use external control information in a relative risk model when individual (not
grouped) data are used such as the case of proportional hazard model where partial likelihood is used. When
external control rates are in the form of vital statistics, it is difficult to use them in that model. Note that the
reason is statistical, not mathematical; while one can transform "mathematically" from relative to additive
model by log transform, it is difficult to incorporate external control when individual data must be used to
construct a model. In the additive model, one can always use the external control data to estimate
background parameters in a model, but not always easy to do so when some relative risk model is used. After
discussion, the workgroup agreed that the additive Poisson regression model with a linear dose, as
recommended by the NRC, is the preferred model.
13
-------
DRAFT Material
Issue 5: Linear Extrapolation to Low Dose
Issue:
The arsenic toxicity data needs to be extrapolated in order to
estimate risk for the potential concentrations of concern. Should
the extrapolation at low concentrations be linear or nonlinear?
Workgroup Conclusion:
Linear extrapolation should be used.
Dose-response models can be classified in terms of the way they represent risks at low doses. In a linear
model, the response is directly proportional to the dose administered, while in nonlinear models, the response
and dose are not directly proportional. Two types of nonlinear models are supralinear, which is concave
downward, and sublinear, which is concave upward. These dose-relationship models are illustrated in Figure
1.
Figure 1. Linear and NonLinear Dose-Response Models
Supralinear
Linear
Superlinear
Dose
EPA's Drinking Water Risk Assessment (2001)
NRC (1999) concluded that the existing studies did not identify specific modes of action. When studies show
no DNA reactivity, without sufficient evidence to support nonlinear modes of action, EPA's current interim
cancer guidelines (EPA 2001) uses linear extrapolation (EPA 1999 at 1-16).
EPA's cancer guidelines noted that animal studies generally don't detect less than a 10% tumor incidence
(1999 at 1-13), which is equivalent to 100 tumors in a population of 1,000. The cancer guidelines state that
data are used to identify the lower limit of the observed dose estimated to cause an adverse health effect for
10% of the population (LED10). "The linear approach is to draw a straight line between a point of departure
from observed data, generally, as a default, the LED10, and the origin...(EPA 1999 at 1-16)." Linear
extrapolation produces the upper bound on risk at low doses assuming no need to add a factor to account for
human variability (EPA 1999 at 1-17).
14
-------
DRAFT Material
EPA accepted the 1999 NRC conclusions about modes of action for arsenic. Therefore, EPA estimated the
risks of cancer from exposure to arsenic in drinking water using a linear extrapolation from the southwestern
Taiwanese epidemiological studies down to the origin.
NRC 2001 Recommendation
The NRC subcommittee concluded that the available mode of action studies do not indicate the shape of the
dose-response curve, which will be a composite of the curves for specific biochemical endpoints (pg. 119).
Statistical goodness-of-fit criteria applied to both the southwestern Taiwanese data and the northeastern
Taiwan data support supralinear models with a log-transform of dose (NRC, 2001at p. 192). The Chilean data
set (Ferreccio et al. 2000) yielded some evidence of supralinear dose-response relationship. On the other
hand, dose misclassification of the southwestern Taiwan data may cause an apparent supralinear relationship.
NRC also recommended a linear dosed because of the inter-individual variation of response in humans
In determining which dose-response model to use, EPA had removed from consideration models that
incorporated a comparison population because they tended to produce supralinear dose-response curves.
The NRC criticized this practice, noting that an apparent supralinear curve may be caused by dose
misclassification. Exposure to high concentrations of arsenic in food or movement of people among the
different villages might cause such effects to appear. NRC noted that the presence of a supralinear curve in
modeling scenarios was not enough justification to remove a model from consideration. NRC did not criticize
the practice of using linear extrapolation below ED0i.
Workgroup Conclusion
The workgroup participants agreed that a linear extrapolation to low dose should be used.
15
-------
DRAFT Material
Issue 6: Use of a Comparison Population
, Shou d mterna or externa comparison groups be used in dose-
Issue: K u K
response modeling?
Workgroup Conclusion: An external comparison population should be used.
In the 1999 NRC report, it was suggested that two different comparison groups be considered for models of
the southwestern Taiwan data set: (1) using only the 42 villages and using the variation in cancer rates from
village-to-village to determine the nature of the estimated dose-response relationship (internal comparison
group); and (2) using data from an external comparison population, such as nationwide data (external
comparison group). NRC noted that an external comparison group is classically used in the analysis of cohort
data and provides more accurate estimates of the baseline cancer rates than an internal comparison group.
However, if a comparison group differs from the study population in important ways, the results will be
biased.
Morales et al. (2000) derived standardized mortality ratios (SMRs) from the ecological Taiwanese data set
with comparison populations of southwestern and all of Taiwan. The authors noted that SMRs correspond to
maximum likelihood estimates (MLEs) of risk ratios derived from Poisson models. The authors obtained the
baseline hazard three ways: modeling without a comparison population, assuming the comparison population
has no exposure, or by using empirical estimates of baseline hazard based on the comparison population. "In
general, models with no transformation on dose and an exponential linear dose effect fit well ...[with] no
comparison population.... [Using] southwestern ... or the entire Taiwnese population, models with the square
root and log transformation fit well (Morales et al. 2000 at p. 658)." Although log transformation without a
comparison population had a good model fit, that model was instable at low dose. In addition, multiplicative
models gave a better fit than additive models. The authors note that ED0i values for male bladder cancer
ranged from 21 to 633 g/L depending on the model choice and on the use of a comparison population.
EPA's Drinking Water Risk Assessment (2001)
EPA based its risk assessment on a model that did not use any comparison population. This model was
chosen partly because the no-comparison group models were more stable (i.e., less sensitive to model
choice) and the comparison-group models yielded ED0i values much lower than seemed reasonable.
Moreover, the differences between the poor rural study population and the more prosperous urban
comparison population might cause bias. Finally, models incorporating a comparison population tended to
produce a supralinear dose-response relationship, even though the mechanistic data suggested a sublinear
model.
NRC 2001 Recommendation
The NRC recommended using the whole Southwest Taiwan region (of which only a small minority is exposed
to increased arsenic concentrations in drinking water) as an external comparison group for the following
reasons:
1. Although it has been argued that the southwestern Taiwan population differs significantly from that
of the whole of Taiwan (other than in the amount of arsenic in the drinking water), the standard
mortality rates from the arsenic-endemic area are similar to those in the rest of Taiwan.
2. Models incorporating a comparison group tended to produce supralinear dose-response curves.
Although the mechanistic data suggest sublinear dose, there are other factors to consider that might
cause the curve to be supralinear (see Issue 5).
16
-------
DRAFT Material
The shape of the dose reponse curve in an individual will be a composite of the curves for the biological
endpoints. NRC notes that even though the modes of action suggest sublinear dose-response curves,
especially extremely susceptible populations may produce a supralinear dose-response curve composite.
NRC (2001, p. 192) noted that supralinear curves can result from understimating the arsenic exposure from
food, from movement of people between villages, and underestimation of the people exposed to high arsenic
wells in villages assigned low exposure. Twenty of the 42 villages only had one well tested, and the rest of
the villages had measurements for 2 to 47 wells (NRC 1999). Using no comparison population would
understimate the slope of the dose-response curve, and using an external comparison population would
decrease the measurement error because the large control group anchors the model fit (NRC 2001 at p. 192).
Workgroup Conclusion
There was no objection among the workgroup participants in using the Southwest Taiwan region as the
external comparison group. However, OW questions whether NRC (1999 and 2001) fully addressed village
exposure uncertainty in their analyses because the measured wells may not have represented all the wells
used in the village during the period of exposure. Wu et al. (1989) collected water samples from wells used
1964-1966, and calculated person-years for 1973-1986, the years analyzed for death certificates. Village 4-7
had one well and 20,856 person-years, while village 6-C had one well and 24,694 person-years.
The Yang and Blackwell (1961) paper examined by the workgroup noted that: "the village of Fong-chia, which
has a population of about 1500 people, has seven wells. As many as twenty or more families use one well
(pg. 114)." "The average life of a deep well is two to three years; however, some of the wells in villages
further from the sea coast occasionally last longer (pg. 114-115)."
17
-------
DRAFT Material
Issue 7: Background Cancer Rate
What background lung and bladder cancer incidence rates should
' be used for the United States population?
, . _ . . U.S. background cancer rates should be used in calculating lifetime
Workgroup Conclusion: . .
risk.
The hazard (or relative risk) of cancer due to arsenic exposure has been estimated based on Taiwanese data,
including the Taiwanese background rate. The results are then used to calculate lifetime risk for the U.S.
population, using the BEIR IV formula and the U.S. background cancer rates.
EPA's Drinking Water Risk Assessment (2001)
EPA used the ED0iS derived by NRC (1999) from Poisson regression models. Furthermore, EPA derived cancer
incidence from the Taiwanese mortality data by assuming 80% mortality for bladder cancer and 100%
mortality for lung cancer. Then the Agency applied the increased lifetime cancer risks (absolute risks) in
Taiwan in Monte Carlo simulations of the U.S. population, to estimate the average individual lifetime bladder
and lung cancer risks in the U.S. caused by arsenic exposure. The average risk value, multiplied by the
number of people exposed to arsenic at 10 ppb in the U.S., provided the number of U.S. cases expected.
Running the Monte Carlo simulation at existing arsenic levels provides the total cases, which when subtracted
from the number of cases at 10 ppb, estimates the number of cases avoided by lowering the MCL to 10 ppb.
The Agency assumed that 26% of the bladder cancers are fatal and 88% of the lung cancers are fatal when
costing out the health benefits for the drinking water regulation.
NRC 2001 Recommendation
NRC noted that the Poisson regression approach cannot readily incorporate the baseline risk of the U.S.
population in deriving the relative risk. NRC estimated U.S. excess lifetime lung and bladder cancer incidence
using U.S. background cancer incidence data from SEER 2001 in Tables 5-7 and 5-8, respectively, and
adjusted by the difference in background cancer incidence between the U.S. and Taiwan using the data from
You et al. 2001 in Table 5-9. Using U.S. background rates increases the bladder risk about 3 times for males
and 2 times for females; 2.3 times for males and 3 times for females for lung cancer.
Workgroup Conclusion
After discussion, the workgroup agreed that relative risk should be used in calculating risk estimates for the
U.S. population. When applying this relative risk data to the U.S. population, it is necessary to use U.S.
background data instead of Taiwanese background data. Therefore, use of U.S. background cancer rates is
recommended.
18
-------
DRAFT Material
Issue 8: Adjustment for Dietary Intake of Arsenic (Food)
How should the dietary intake of arsenic in southwestern Taiwan
be taken into account?
, . _ . . The model should be run multiple times over a range of possible
Workgroup Conclusion: , .
food consumption rates.
Because there is a high general background level of inorganic arsenic in food, it has been suggested that
effective exposures to arsenic in Taiwan are higher than represented simply by the amount of well water
drunk. For this paper, the issue of intake of arsenic from food (e.g., dry rice, sweet potatoes) has been
distinguished from the issue of intake of arsenic from drinking water (Issue 9) and intake of arsenic from
water used in cooking, such as water used to boil rice and potatoes (Issue 10).
EPA's Drinking Water Risk Assessment (2001)
To account for background levels of arsenic in food, EPA assumed that the inorganic arsenic consumption due
to food in Taiwan was 50 ug/day, compared to 10 ug/day in the United States. NRC (1999) cited results of
Schoof et al. (1998) as estimating Taiwanese daily intake from yams as 31 ug/day and rice as 19 ug/day.
NRC (1999) also noted that the Li et al. (1979) study found 95% of the rice crop to contain arsenic primarily
100 to 700 ug/kg, with some up to 1.43 mg/kg. The soil had probably been treated with arsenical pesticides.
NRC 2001 Recommendation
The NRC found little evidence to support EPA's assumption that food contributed 50 ug/day of inorganic
arsenic to the Taiwanese diet. NRC addressed the issue by determining how sensitive the calculation of ED0i
was to the consumption rate. NRC found that changing the consumption rate from 50 ug/day to 30 ug/day
did not change the calculated ED0i significantly (about 1% difference). This lack of sensitivity was not
unexpected, since the southwestern Taiwanese population, which was used as a comparison group, had a
similar dietary intake as the exposed population.
Workgroup Discussion
Review of New Studies
Since the publication of NRC 2001, the workgroup reviewed a number of studies, delving into the literature
used to derive EPA's 50 ug/day value and examining new Asian literature.
Taiwanese Studies
(Irgo/ic 1988) summarized Yang and Blackwell's 1961 study of 41 families in Southwest Taiwan.
(Yang and Blackwell 1961) interviewed 41 families in the affected region of Southwest Taiwan. Based on
these interviews, the study presents consumption rates of rice and sweet potatoes. Although the
information is based on anecdotal data, and no statistics are presented, the study was considered worth
reviewing because the data are directly from the population of concern.
(EPA 1989) An internal workgroup paper which slightly modified the food consumption patterns in the
Irgolic 1988 and Yang and Blackwell 1969 papers to account for the higher sweet potato consumption of
the majority of the study population. Using Yang and Blackwell's data for cooked rice and sweet potatoes,
EPA's food adjustment decreases the caloric average consumption that Yang and Blackwell developed in
the field, so EPA's arsenic intake may underestimate actual consumption patterns at the time.
(Li et al, 1979) In 1975 the authors collected rice samples from 86 townships in Taiwan, when Taiwan
routinely used arsenical pesticides on rice. The mean was 0.3 to 0.53 ppm or mg/kg .
19
-------
DRAFT Material
(Schoof et al,, 1988) The authors analyzed the dry weight concentrations of arsenic in rice and yams in
South West Taiwan and applied EPA's 1989 daily consumption rate of 225 g/day of rice and 500 g/day
yams (Irgolic 1988) to estimate daily dietary intake of arsenic.
Studies in India and Bangladesh
Duxbury et al. (2003), Bae et al. (2003), Alam et al. (2003), and Watanabe et al. (2004) reported arsenic
content in foods grown in West Bengal, India and Bangladesh. Chowdhury et al. (2001) estimate that
adults in West Bengal, India obtain about 285 ug/day of inorganic arsenic from their primary staples of
rice (750 g/day) and vegetables (500 g/day). Roychowdhury et al. (2002) report concentrations of arsenic
in potatoes and raw and cooked rice from 6 villages in West Bengal where arsenic concentrations in
groundwater are above 50 mg/L.
Food consumption rates, arsenic concentrations in food, and arsenic ingestion rates from these studies are
presented in Tables 2, 3, and 4, respectively. The arsenic ingestion rates calculated from Chowdhury et al.
(2001) and Watanabe (2004) are 285 and 214 ug/day, respectively. These rates are much higher than 30-50
l^g/day consumption rates used by NRC. However, the studies from which these rates were derived are from
locations where both dietary habits and arsenic concentrations in food differ from those in Taiwan.
Table 2. Summary of Food Consumption Studies
Study
Yang and Blackwell, 1961
U.S. EPA 1989
Chowdhury et al. 2001
Bae et al., 2002
Watanabe et al., 2004
Population
Southwes Taiwan,
male adults
Taiwan
West Bengal, India
Bangladesh
NW Bangladesh
Method
Informal interview
survey, n=41
-------
DRAFT Material
Table 3. Summary of Studies of Arsenic Concentrations in Food
Study
Li et al. 1979
Schoof et al. 1998
Duxbury et al., 2003
Alam et al., 2003
Bae et al., 2002
Roy Chowdhury et al., 2002
Population
Taiwan
South west
Taiwan
Bangladesh
Bangladesh
Bangladesh
West Bengal, India
Method
Samples from 85
townships
8 rice samples & 19 yam
samples
150 rice samples tested
from winter and rainy
season
-
-
--
Food Type
Rice
Rice
Yams
Dry Rice
Potatoes (not reported as
yams or as white potatoes)
Dry Rice
Dry Rice
Potato skins
Concentration
mean 0.3 to 0.53
mg/kg
0.15 mg/kg
0.11 mg/kg
0.1-0.42 mg/kg
<0.1 mg/kg (dry
weight)
0.173 mg/kg
0.04-0.7 mg/kg
0.06-0.7 mg/kg
Table 4. Summary of Studies of Arsenic Consumption Per Day
Study
Schoof et al. 1998
Chowdhury et al., 2001
Watanabe et al., 2004
Population
Taiwan
West Bengal, India
NW Bangladesh
Method
-
-
~
Food Type
Rice and Yams, Adults
Rice and Vegetables, Adults
Rice and Vegetables, Children
Rice, Bread, Potato, Fish, Male
Rice, Bread, Potato, Fish, Female
Concentration
50 ng/day
285.0 jig/day
153.2 ng/day
214 jig/day
120 ng/day
Workgroup Conclusion
While there was disagreement among the workgroup participants as to the dietary intake of arsenic, all
participants agreed that the ramifications on the risk assessment of choosing one value over the other would
be very small. The workgroup concluded that the model should be run over a range arsenic consumption
rates considered by NRC (30 ug/day to 50 ug/day) to confirm that the calculated risk is insensitive to this
value.
21
-------
DRAFT Material
Issue 9: Adjustment for Water Intake (From Drinking Water Only)
How much water do people drink in southwestern Taiwan and in
SSUe! the United States?
, . _ . . The model should be run multiple times over a range of possible
Workgroup Conclusion: .... .
drinking water consumption rates
To estimate cancer risks associated with a given arsenic concentration in drinking water, a value must be
determined to account for the quantity of drinking water consumed. The drinking water consumption rates
used in the model are important, as assumptions about the total arsenic exposure in the study population can
have a large impact on risk assessments. This paper addressed the issue of intake of arsenic from drinking
water (Issue 9) and intake of arsenic from water used in cooking, such as water used to boil rice and
potatoes (Issue 10).
EPA's Drinking Water Risk Assessment (2001)
Based on data from the 1994-1996 Continuing Survey of Food Intakes by Individuals (CSFII) (EPA 2000), EPA
estimated that the mean daily average per capita consumption of tap water by individuals in the United
States was 1 L/person/day for "community tap water" and 1.2 L/person/day for "total water" (which includes
bottled water). The values represented a lifetime average tap water intake, which included the amount of
tap water added during food preparation (discussed further in issue 10). U.S. consumption of tap water does
not include water added by manufacturers during processing (e.g., beer, soft drinks, ready-to-eat canned
soup) because manufacturers often process the water (e.g., reverse osmosis) for product consistency, which
would remove contaminants such as arsenic. However, instead of using a point estimate (e.g., a 70-kg adult
drinking 2 L/day, and about 90% of the population drinks less than 2 L/day) for its risk assessment, EPA
conducted a Monte Carlo analysis generate lifetime risks of water intake.
For the Taiwanese population, EPA assumed that the consumption was 3.5 and 2.0 L/day for men and
women, respectively. The Agency also used these consumption rates in the 1988 risk assessment. A 1989
EPA workgroup report noted that 3 to 4 adults estimated their daily water intake as up to 3.75 L/day during
EPA's 1988 visit to the affected Taiwan area.
NRC 2001 Recommendation
NRC was concerned about two issues:
1. The Taiwanese study used by EPA measured exposure at the village level, rather than at the
individual level. Individuals in the same village may vary dramatically in terms of the quantity of
water consumed.
2. Although, on average, the Taiwanese population should have a higher drinking water consumption
than the U.S. population, the idea that the two populations differ dramatically has been questioned
(Mushak and Crocetti, 1995).
NRC (2001at 140) noted that "...no appropriate data on the distribution of water consumption in the
Taiwanese study populations are available at this time.... However, it seems likely that the water
consumption pattern of the people in southwestern Taiwan... has changed with time as the socioeconomic
situation has improved.... In the absence of reliable data on water consumption in the Taiwanese study
populations, the sensitivity of the risk estimates to those assumptions should be assessed to quantify some of
the uncertainty in the risk assessment."
22
-------
DRAFT Material
To determine the impact of individual variability in the Taiwanese drinking water rate, NRC modeled
Taiwanese risk based on the U.S. population water intake profile of the CSFII data, which has a mean of 21
milliliters per kilogram body weight per day (mL/kg/day) and a standard deviation of 15 mL/kg/day.
Second, for the sensitivity analysis in Table 5-5, NRC selected three water consumption rates for a 50 kg
Taiwanese adult: a consumption rate equal to the U.S. mean rate (i.e., 21 mL/kg/day x 50 kg= 1 L/day), a
rate 2.2 times that of the U.S. mean, as used in EPA 1988 (21 mL/kg/day x 50 kg x 2.2 = 2.3 L/day, or 46
mL/kg/day), and a rate 3 times the U.S. mean (to implicitly account for additional arsenic exposure through
food and cooking water [see issue 10], 3.2 L/day, or 64 mL/kg/day. Taking individual variability of Taiwanese
into account and increasing the consumption rate increased the central tendency estimates for ED0i and
widened the 95% confidence interval for the calculated ED0i values. Therefore, NRC concluded that the
drinking water consumption rate, as well as variability of that rate in both U.S. and Taiwanese populations,
are important factors to consider.
In calculating risk estimates for U.S. populations exposed to arsenic through drinking water, NRC used a
drinking water consumption rate of 1 L/day for the U.S. population and two possible consumption rates for
the Taiwanese population: 1 L/day (identical to the U.S. population) and 2.2 L/day. NRC noted that they
calculated risk estimates to compare results from Taiwan, Chile, and the U.S., noting that the risk values
presented "should not be considered bounds on the possible risk estimates, because other assumptions could
be made that would result in higher or lower values (NRC 2001 at 203)." The report also states that NRC
calculated risks "using assumptions considered to be reasonable by the subcommittee. (NRC 2001 at p. 203)."
Review of New Studies
Since the publication of NRC 2001, the workgroup reviewed a number of other studies not cited by NRC.
These studies are summarized in Table 5.
Taiwanese Study (Yang and Blackwell, 1961)- EPA had not previously cited this source, although the 1989
workgroup members had file copies. The study authors conducted interviews in Southwest Taiwan with
families in the affected region. Based on these interviews, the study estimates adult water intake to be 1 to
3 liters a day, depending on temperature and physical activity. "[M]ost of the population are engaged in
relatively heavy labor.... [T]he proportion of poor people in the total population of the area appears to be
higher than the national average and the economic status of the poorest families in the area certainly is
extremely low.... The two principal sources of calories ... are sweet potatoes and rice
Studies in India and Bangladesh - Two recent studies have reported on arsenic exposure in rural villages of
Bangladesh (Watanabe et al., 2004) and West Bengal, India (Chowdhury et al., 2001). The rural settings and
socioeconomic classes share some similarities with the Taiwanese study population. However, these studies
have higher temperatures, different foods and preparation methods, longer growing seasons, and Watanabe
reported short work days. A larger data set was collected in the Chowdhury study; however reporting of
methods and analyses were limited.
International Commission on Radiological Protection (ICRP)- (1981), as cited in EPA 1997. Although the
U.S. populations (adult and children) and study design are not clearly provided, the results are consistent
with other studies. Increasing temperature and activity increased fluid intake in the U.S. (3.7 L/day for a
70 kg adult = 52 mL/kg/day) to levels approaching the Asian studies.
Ershow and Cantor (1989) - as cited in EPA 1997. The Ershow and Cantor (1989) based their estimates
on the U.S.DA 1977-1978 Nationwide Food Consumption Survey. Total water was defined as tapwater
plus water in purchased food and beverages (e.g, soft drinks and beer). A 70 kg American adult has a
total water consumption of 30 mL/kg/day.
2004 U.S. Studies- Two large, well conducted surveys of drinking water rates in the U.S. have been
conducted (NHANES and CSFII). These studies provide useful benchmarks for evaluating Taiwanese
drinking water rates. In addition to water intake, the NHANES survey provides information on all
beverages consumed and attributes 19% water contribution inherent in food sources.
23
-------
DRAFT Material
Data from these studies are summarized in Table 5. The studies present values for the mean drinking water
rate ranging from 2 to 3.5L/day in Asia. This range is consistent with the range used in the NRC report (1 to
3.1 L/day). In addition, unlike the Taiwanese, the Americans obtain fluids other than tap water (e.g., soft
drinks, beer), so that total U.S. water ranges from 2.1 to 2.8 L/day and higher for active adults or adults in
hot environments. Water added to products prior to merchandising (e.g., bottled ice tea) is not included in
the 2001 EPA risk assessment consumption, and the National Health and Nutrition Examination Survey (N
HANES) tracks beverage water in the 2.8 L/day value. Furthermore, NHANES tracks water contained in food
in its 3.2 L/day total water estimate.
Workgroup Conclusions
The workgroup agrees with the NRC conclusion that the selection of the drinking water consumption rates
should consider, as possible, the uncertainty associated with trying to accurately determine the mean
consumption rate of the populations, and of the variability of individuals within the populations. The
workgroup agrees that the model is highly sensitive to the term selected. The workgroup's review of relevant
literature suggests that the mean adult drinking water consumption rate for Asian population is between 1 to
4.6 L/day.
24
-------
DRAFT Material
Table 5. Summary of Water Consumption Studies
Study
Yang and Blackwell, 1961
EPA 1989
Chowdhury et al., 2001(a)
Watanabe et al., 2004
ICRP 1981, as cited in EPA
1997.
Ershow and Cantor (1989),
as cited in EPA 1997.
U.S. DA CSFII as reported in
EPA, 2004
NHANES III as reported in
NRC, 2004
Population
Southwest Taiwan,
adults
Southwest Taiwan,
adults
West Bengal, India
- adults
Bangladesh -
adults working 1-4
hours a day mean,
6 max.
U.S. adults
U.S. adults 20-64
U.S. all ages
U.S. adults 18-30
yr
Method
Informal interview
survey, n=48
informal questioning of
3-4
Formal interview survey,
n = 4,613 (a)
Formal interview survey,
n=38
6 days of hot weather
at the end of the rainy
season.
Normal conditions
Up to 90°F
Moderately active
U.S. DA survey, included
1,731 adults in that age
group
Formal questionnaire
survey, n = 20,000
Formal questionnaire
survey, n= ~3,700
Water Type
Drinking water
Drinking water
Well water drunk
Well water drunk
Includes tap water, coffee,
soft drinks, beer, etc.
Tap water
Total water: includes water
in food & beverages
-------
DRAFT Material
Issue 10: Adjustment for Dietary Water Intake (From Cooking Water)
How should the dietary intake of arsenic in southwestern Taiwan
' via cooking water be taken into account?
Workgroup Conclusion: The use of 1 L/day is justified by literature.
For this paper, the issue of intake of arsenic from drinking water (Issue 9) and intake of arsenic from water
used in cooking, such as water used to boil rice and potatoes (Issue 10).
EPA's Drinking Water Risk Assessment (2001)
EPA assumed that 1 L/day of cooking water was consumed by both men and women in Southwest Taiwan.
The value was based on anecdotal information regarding water requirements to cook rice and potatoes, and
information regarding rice and potato consumption rates (U.S. EPA 1989).
NRC 2001 Recommendation
Although the NRC agreed with EPA's method for accounting for the extra water consumption due to use of
drinking water in cooking food, the NRC noted that the rationale for using 1 L/day was not documented.
Workgroup Conclusions
Three recent studies have reported on arsenic exposure in rural villages of Bangladesh (Watanabe et al.,
2004, and Bae at al., 2003) and West Bengal, India (Chowdhury et al., 2001). The rural settings and
populations have similarities with the Taiwanese study population. However, climatic and cultural differences
exist (e.g, different foods in diet and different rice preparation).
Chowdhury et al. (2001), Bae et al. (2003), and Watanabe et al. (2004) reported that 1 L of water was used
for food preparation (rice, potatoes, drinks) for adults in West Bengal and Bangladesh.
The lifetime per capita U.S. mean of tap water and bottled water of 1.0-1.2 L/day, respectively, includes the
water added to foods and beverages during preparation (indirect water used in food preparation ranges from
0.5 to 0.55 L/day of the U.S. water consumption).
26
-------
DRAFT Material
Issue 11: Adjustment for Mortality vs. Cancer Incidence
Issue: How are cancer mortality rates related to cancer incidence rates?
. . _ , . Since the workgroup decided to determine relative risks instead of
Workgroup Conclusion: , . . . . ... . . .
absolute risks, this issue is irrelevant.
EDoi calculations used by EPA were based on data from southwestern Taiwan and referred to lifetime cancer
mortality from bladder or lung cancer. In order to compare these data to the northwestern Taiwan data, it
was necessary to make assumptions regarding the relationship between cancer incidence and cancer
mortality.
EPA's Drinking Water Risk Assessment (2001)
EPA converted risks calculated from Taiwanese mortality data to cancer incidence by assuming an 80%
mortality for bladder cancer and a 100% mortality for lung cancer. EPA applied the absolute risk increase
seen in Taiwan to the U.S. arsenic exposure to derive increased occurrence of lung and bladder cancer
incidence. In its 2001 risk assessment, EPA applied 26% bladder mortality and 88% lung cancer mortality to
the U.S. cancer estimates.
NRC 2001 Recommendation
NRC used the lifetime baseline (background) cancer risk in the U.S. to derive cancer risk for arsenic
exposure, which increases the risk estimates. Lung cancer incidence is three times higher for females and
two times higher for males in the U.S. than in Taiwan. Likewise, bladder cancer is about three times higher
in males and two times higher in females for American.
"[T]he subcommittee was split on whether using the U.S. background rates was preferable to using the
Taiwanese background rates for estimating arsenic risks in the United States.... The subcommittee agreed,
however, that if there was a multiplicative interaction between a complex array of risk factors, including
smoking,... then using the U.S. background cancer incidence rates would be preferred over the Taiwanese
background rates for estimating arsenic cancer risks in the U.S. population (NRC 2001 at 221)."
NRC believes that 100% mortality for lung cancer is probably an appropriate estimate, but that 80% mortality
for bladder cancer in Taiwan may not be appropriate for the U.S. population. Data from SEER (2001) suggest
that the mortality rate in the U.S. is around 20%.
Workgroup Conclusion
Because the workgroup decided to use relative risks instead of absolute risks, no adjustment is necessary to
account for the difference in background cancer incidences between the U.S. and Taiwan.
27
-------
DRAFT Material
Issue 12: Child Sensitivity Issue
, What food and water consumption rates should be used for
Issue: ....
children?
Children should already be addressed in the epidemiological
, . _ . . studies. However, the time to tumor may have longer time to
Workgroup Conclusion: , . . . . ...... , ,. , . ., . . .
develop into a tumor. Will be discussed in the uncertainty
analysis. No further adjustment is needed.
The Agency has published Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a); this document
mentions the need to address early-life exposures from carcinogens. In addition, the recent EPA
Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens (EPA, 2005b)
identifies age-dependent adjustment factors to be applied to carcinogens with a mutagenic mode of action
when chemical-specific data are unavailable. (U.S. EPA, 2005b).
Much toxicity data are available on arsenic; however, the data needed to account for an accurate
representation of early-life exposure to arsenic appears to be insufficient. For example, the National Resource
Council (NRC, 2001) summarizes the few studies of infant mortality, spontaneous abortions, and still births in
chapter 2. NRC (2001) concluded: "There are no reliable data that indicate heightened susceptibility of
children to arsenic.... [IJnfants and children might be at greater risk for cancer and noncancer effects
because of greater water consumption on a body-weight basis. However, ... the lifetime cancer risk
estimates account for the greater childhood exposures by deriving risk estimates from epidemiology studies
of cancer among populations exposed to cancer from birth...." (NRC 2001 at p. 8) "However, the [adverse
reproductive effects] evidence is not conclusive, because the studies suffer from such limitations as a lack of
information on lifestyle and other exposures that could affect reproductive outcomes (NRC 2001 at 66). NRC
also concluded "that although a large amount of research is available on arsenic's mode of action, the exact
nature of the carcinogenic action is not clear" (NRC, 2001). Finally, NRC concluded that inorganic arsenic and
its metabolites have been shown to induce chromosomal alterations and large deletion type mutations, but
not point mutations.
Although there is some new evidence indicating that exposure to arsenic from drinking water during pregnancy may be
associated with decreased birth weights of newborns (Hopenhayn, 2003) and may increase the cancer incidence of the child
in the later stage of life (Waalkes, 2003), the data needed to account for an accurate representation of early-life exposure
of arsenic appears to be insufficient (NRC, 2001).
The working group also understands that whether children are more sensitive to a carcinogen is not necessary related to
the mechanism of how it causes cancer. For the chemical with same potency, exposure at the early stages of life would
allow the body more time to express and turn into final cancer. A time to tumor analysis would help to demonstrate this
kind of difference.
Working Group Conclusion
Because the cancer slope factor used in this cancer risk assessment is derived from the epidemiology study using the
Southwestern Taiwan data, it is generally believed that the sensitive population exposed to inorganic arsenic through
drinking water during the most sensitive period of time is already included in the exposed population. Therefore, given the
lack of additional data, the working group agreed that an adjustment factor does not appear to be appropriate in the cancer
risk assessment associated with inorganic arsenic exposure in the current approach.
28
-------
DRAFT Material
References
Alam MGM et al. 2003. "Arsenic and heavy metal contamination of vegetables grown in Samta village,
Bangladesh/The Science of Total Environment. 308, pp. 83-96.
Bae, M. C. Watanabe, T. Inaoka, M. Sekiyama, N. Sudo, M..H. Bokul, R. Ohtsuka. 2002. Arsenic in cooked
rice in Bangladesh. The Lancet v. 360, December 7, 2002, 1839-1840.
Chen CJ, CW Chen, MM Wu, TL Kuo. 1992. Cancer potential in liver, lung, bladder, and kidney due to
ingested inorganic arsenic in drinking water. Br. J. Cancer 66(5):888-892.
Chen CJ, MM Wu, SS Lee, JD Wang, SH Cheng, and HY Wu. 1988. Atherogenicity and carcinogenicity of
high-arsenic artesian well water. Multiple risk factors and related malignant neoplasms of blackfoot
disease. Arteriosclerosis 8:452-460
Chen CJ, YC Chuang, TM Lin, and HY Wu. 1985. Malignant neoplasms among residents of a blackfoot
disease-endemic area in Taiwan: high-arsenic artesian well water and cancers. Cancer Res. 45(11 Pt 2):
5895-5899.
Chiou HY, ST Chiou, YH Hsu, YL Chou, CH Tseng, ML Wei, CJ Chen. 2001. Incidence of transitional cell
carcinoma and arsenic in drinking water: a follow-up study of 8,102 residents in an arseniasis-endemic area
in northeastern Taiwan. Am. J. Epidemiol. 153(5)411-418.
Chowdhury et. al. 2001. "Groundwater Arsenic Contamination and Human Suffering in West Bengal - India
and Bangladesh." Environmental Sciences 8(5), 2001, pp. 393-415.
Duxbury JM, AB Mayer, JG Lauren, N. Hassan. 2003. "Food chain aspects of arsenic contamination in
Bangladesh: Effects on quality and productivity of rice." in Journal of Environmental Science and Health, Vol.
A38, No. 1, pp.61-69.
Ershow and Cantor. 1989. Total Water and Tapwater Intake in the United States: Population-Based
Estimates of Quantities and Sources. As cited in U.S. EPA 1997.
Ferreccio C, C Gonzales, V Milosavlijevic, G Marshall, AM Sancha, and AH Smith. 2000. Lung cancer and
arsenic concentrations in drinking water in Chile. Epidemiology ll(6):673-679.
Hopenhayn,C.; Ferreccio, C.; Browning, S.R.; Huang, B. Peralta, C.; Gibb, H.; and Hertz-Picciotto, I.
2003. Arsenic Exposure from Drinking Water and Birth Weight. Epidemiology. 14:593-602.
Institute of Medicine of the National Academies (2004). Dietary reference intakes: water, potassium, sodium,
chloride and sulfate. National Academies Press, Washington, DC.
Lamm SH, Byrd DM, Kruse MB, Feinleib M, Lai S. Bladder Cancer and Arsenic Exposure: Differences in the
Two Populations Enrolled in a Study in Southwest Taiwan. Biomedical and Environmental Sciences 2003;
16:355-368
Lamm SH, Engel A, Kruse MB, Feinleib M, Byrd DM, Lai S, Wilson R. Arsenic in Drinking Water and Bladder
Cancer Mortality in the U.S.: An analysis based on 133 U.S. counties and thirty years of observation. Journal
of Occupational and Environmental Medicine 2004; 46(3):298-306
Lewis DR, JW Southwick, R Ouellet Hellstrom, J Rench, and RL Calderon. 1999. Drinking water arsenic in
Utah: a cohort mortality study. Environmental Health Perspectives 107(5):359-365.
29
-------
DRAFT Material
Morales KH, L Ryan, TL Kuo, MM Wu, and CJ Chen. 2000. Risk of Internal Cancers from Arsenic in Drinking
Water. Env. Health Persp. V108, n7, July 2000.
Mushak P and AF Crocetti, 1995. Risk and revisionism in arsenic cancer risk assessment. Environ. Health
Perspect. 103 (7-8):684-689.
NRC, 1999. Arsenic in Drinking Water. Washington, DC: National Academy Press.
NRC, 2001. Arsenic in Drinking Water: 2001 Update. Washington, DC: National Academy Press.
Roychowdhury T, T Uchino, H Tokunaga, and M. Ando. 2002. Survey of arsenic in food composites from an
arsenic-affected area of West Bengal, India. Food and Chemical Toxicology 40:1611-1621.
Schoof RA, LJ Youst, E Crecelius, K Irgolic, W Goessler, HR Guo, and H Greene. 1998. Dietary arsenic intake
in Taiwanese districts with elevated arsenic in drinking water. Human and Ecological Risk Assessment
SEER (Surveillance, Epidemiology, and End Results). 2001. Surveillance, Epidemiology, and End Results
Program Public-Use Data (1973-1998), National Cancer Institute, DCCPS, Surveillance Research Program,
Cancer Statistics Branch. [Online]. Available: http://www-seer.ims.nci.nih.gov/Publications/ [August 20,
2001]
Steinmaus C, Yuan Y, Bates M, Smith A. Case-Control Study of Bladder Cancer and Drinking Water Arsenic in
the Western United States. American Journal of Epidemiology 2003; 158:1193-1201
Tsai SM, TN Wang, and YC Ko. 1999. Mortality for certain diseases in areas with high Ivels of arsenic in
drinking water. Arch. Enivron. Health 54(3): 186-193.
Tollestrup K, Frost FJ, Harter LC, McMillan GP. Mortality among children residing near the American Smelting
and Refining Company (ASARCO) copper smelter in Ruston, Washington. Arch Environ Health. 2003
Nov;58(ll):683-91.
U.S. EPA. 1980. Water Quality Criteria Documents Availability. November 28, 1980 Federal Register (45 FR
79318).
U.S. EPA. 1988. Special Report on Ingested Inorganic Arsenic: Skin Cancer; Nutritional Essentiality.
EPA/625/3-87/013.
U.S. EPA. 1989. Report on Arsenic Work Group Meetings. From Charles Abernathy, William Marcus, Chao
Chen, Herman Gibb, and Paul White to Peter Cook and Peter Preuss. February 23, 1989.
U.S. EPA. 1992. Water Quality Standards; Establishment of Numeric Criteria for Priority Toxic Pollutants;
States' Compliance; Final Rule. December 22, 1992 Federal Register (57 FR 60848).
U.S. EPA. 1997. Exposure Factors Handbook: Volume I - General Factors. EPA/600/P-95/002Fa. August
1997. Office of Research and Development. National Center for Environmental Assessment.
U.S. EPA. 1999. Draft Guidelines for Carcinogen Risk Assessment. NCEA-F-0644. July 1999. Review draft.
Available at www.epa.gov/ncea/raf/pdfs/cancer_gls.pdf.
U.S. EPA. 2001a. National Primary Drinking Water Regulations; Arsenic and Clarifications to Compliance and
New Source Contaminants Monitoring; Final Rule. January 22, 2001 Federal Register (66 FR 6976).
U.S. EPA. 2001b. Arsenic Rule Benefits Analysis: An SAB Review. A Review by the Arsenic Benefits Review
Panel (ABRP) of the U.S. EPA Science Advisory Board (SAB) EPA-SAB-EC-01-008.
30
-------
DRAFT Material
U.S. EPA. 2003a. Notice of Availability and Opportunity to Provide Comment on the Draft Final Guidelines for
Carcinogen Risk Assessment and the Draft Supplemental Guidance for Assessing Cancer Susecptibility from
Early-Life Exposure to Carcinogens. March 3, 2003 Federal Register (68 FR 10012).
U.S. EPA. 2003b. Draft Final Guidelines for Carcinogen Risk Assessment. EPA/630/P-03/001A. NCEA-F-
0644A. February 2003. Available at www.epa.aov/ncea/raf/cancer2003.htm.
U.S. EPA. 2003c. Supplemental Guidance for Assessing Cancer Susceptibility from Early-Life Exposure to
Carcinogens. EPA/630/R-03/003. External Review Draft. February 2003. Available at
www.epa.qov/ncea/raf/cancer2003.htm.
U.S. EPA. 2004. Estimated Per Capita Water Ingestion in the United StatesAn Update. Based on Data
Collected by the United States Department of Agriculture's (U.S.DA) 1994-1996 and 1998 Continuing Survey
of Food Intakes by Individuals. EPA-822-00-001. Office of Water, Office of Standards and Technology, U.S.
Environmental Protection Agency. October 2004.
U.S. EPA. 2005a. Guidelines for Carcinogen Risk Assessment. EPA/630/P-03/001F March 2005. Available at
http://cfpub.epa.aov/ncea/cfm/recordisplav.cfm?deid= 116283
U.S. EPA. 2005b. Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to
Carcinogens. EPA/630/R-03/003F March 2005. Available at
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=116283
Waalkes, M.P.; JM Ward, J Liu and BA Diwan 2003. Transplacental carcinogenicity of inorganic arsenic in the
drinking water: induction of hepatic, ovarian, pulmonary, and adrenal tumors in mice. Toxicology and Applied
Pharmacology: 186:7-17
Watanabe et al 2004. Water intake in an Asian population living in arsenic-contaminated area. Toxicology
and Applied Pharmacology. 198:272-282
Yang TI and RQ Blackwell (1961). Nutritional and environmental conditions in the endemic blackfoot area.
Formosan Science. Volume 15, number 3, pages 101 - 129.
You SL, TW Huang, and CJ Chen. 2001. Cancer registration in Taiwan. Asian Pacific J. Cancer Prev. 2(IACR
suppl.): 75-78.
Chen CL, Hsu LI, Chiou HY, Hsueh YM, Chen SY, Wu MM, Chen CJ. Ingested arsenic, cigarette smoking, and
lung cancer risk follow-up study in arseniasis-endemic areas in Taiwan. JAMA. 2004 Dec 22;292(24):2984-90.
31
------- |