United States Science Advisory EPA-SAB-RSAC-01-005
Environmental Board (1400A) May 2001
Protection Agency Washington, DC www.epa.gov/sab
vvEPA ARSENIC RULE
BENEFITS ANALYSIS:
AN SAB REVIEW
PANEL REVIEW DRAFT
AUGUST 9, 2001
A REVIEW BY THE ARSENIC RULE
BENEFITS REVIEW PANEL (ARBRP) OF
THE US EPA SCIENCE ADVISORY
BOARD (SAB)
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
arsenic.wpd
August _, 2001
EPA-SAB-01-
Honorable Christine Todd Whitman
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, NW
Washington, DC 20460
Subject: Arsenic Rule Benefits Analysis; A Science Advisory Board Review
Dear Governor Whitman:
On July 19 and 20, 2001 the Arsenic Rule Benefits Review Panel of the US EPA Science
Advisory Board (SAB) met to review the EPA report EPA 815-R-00-026..
As part of the review process, the ARBRP responded to five charge questions:
Charge Question 1: How should latency be addressed in the benefits estimates when
existing literature does not provide specific quantitative estimates of latency periods
associated with exposure to arsenic in drinking water?
Charge Question 2: How should health endpoints (other than bladder and lung cancer)
be addressed in the analysis, when [existing] literature does not provide specific
quantification, to ensure appropriate consideration by decision makers and the public?
Charge Question 3: Shouldreduction/elimination of exposure be evaluated as a separate
benefits category, in addition to or in conjunction with mortality and morbidity
reduction?
Charge Question 4: How should total benefits and costs and incremental benefits and
costs be addressed in analyzing regulatory alternatives to ensure appropriate
consideration by decision makers and the public?
Charge Question 5: How should uncertainties be addressed in the analysis to ensure
appropriate consideration by decision makers and the public?
Detailed answers to these questions are found in the body of the report. The major
findings and recommendations are:
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
Charge Question 1
A central component in analyzing the benefits of reduced exposure to a carcinogen is to
predict the annual reduction in cancer cases following reduction in exposure. If a population
previously exposed to 50 ppb of arsenic in drinking water is exposed, beginning in 2006, to only
10 ppb, cancer risks in the population will eventually decline to a steady state level associated
with a lifetime of exposure to 10 ppb. How fast this reduction in risk occurs depends on the
cessation lag following reduction in exposure. In the report, we suggest ways in which the
length of this lag could be estimated. When estimates of this lag are unavailable, several possible
assumptions could be made, and the implications of the assumptions for the time pattern of
reduction in cancer cases calculated. For example, the upper bound to benefits (which is the
central case presented in the arsenic benefits analysis) is to assume that the steady state is
reached immediately (i.e., that there is no cessation lag). However, other possible assumptions
could be made and should be included in the primary benefits analysis.
Charge Question 2
The scientific literature on health effects due to arsenic exposure includes studies of a
number of endpoints other than cancer, as well as studies of several cancer sites for which the
risks/benefits have not been quantified (EPA 2000). The quality of these studies varies, as does
the strength of evidence they provide. Nevertheless, this body of evidence is relevant for the
determination of an MCL and needs to be addressed more fully. In some cases, the non-
quantified effects can and should be quantified. Specifically, it appears to us that it should be
possible to quantify mortality from ischemic heart disease, diabetes mellitus and skin cancer.
Serious consideration should also be given to prostate cancer, nephritis and nephrosis,
hypertension, hypertensive heart disease and non-malignant respiratory disease. The literature
that would permit quantification of cases avoided for these endpoints is discussed in Section
2.2.2 of the report.
In cases where a dose-response function has not been estimated, it should be possible to
describe, in tables such as those presented in Appendix 2.2 of this report, the key features of
studies in the literature. Studies must first be selected according to well-defined criteria. The
information that should be provided for each study (grouped by health endpoint of interest)
includes:
Nature of the study design
How exposure was measured
Range of exposures observed
What type of statistical analysis was conducted and what confounding factors were
controlled for in the analysis
In some cases the literature may be so extensive that a summary of results is required in
the text of the report. This summary should focus on health endpoints that have meaning to
humans (e.g., ischemic heart disease rather than electrocardiogram abnormalities), and
should provide some discussion of the mechanism by which the toxin would be expected to
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
exert an effect. The summary should also indicate the level at which effects were observed
in the studies reported and should comment on the likelihood of observing these effects at the
levels relevant to the regulatory decision.
Charge Question 3
Regarding Charge Question 3, we believe that reductions in exposure should not be
considered a separate category of benefits in a benefit cost analysis. The damage function
approach to valuing benefits currently used by the Agency separates the measurement of the
relationship between exposure and response (e.g., risk of fatal or non-fatal cancer) from the
valuation of reductions in risk of death or illness. Epidemiologists estimate dose-response
functions and economists measure the value people place on reductions in risk of death or
illness. Reductions in exposure are already valued under the damage function approach when
people value the reductions in the risk of death or illness associated with them. To add a
separate value for reductions in exposure per se would be double counting.
To abandon the damage function approach and ask people to value reductions in
exposure directly would force lay people to act as epidemiologists, and there is evidence that this
would be difficult. Studies have shown that lay people do not view risk of death or illness as
related to the size of the dose of a toxic substance received; however, the essence of modern risk
assessment is to relate death and illness to size of dose received.
Charge Question 4
We applaud the Agency for presenting the costs and benefits associated with various
possible maximum contaminant levels rather than presenting only the costs and benefits
associated with a single standard that the Agency proposes to implement. We believe, however,
that in the primary analysis (i.e., in the Executive Summary) benefits and costs should be broken
down by system size. Because of the large economies of scale associated with drinking water
treatment, benefit cost ratios are likely to vary substantially by system size, and this information
should be made clear to policy makers and the public.
We also believe that benefits (and incremental benefits associated with different
maximum contaminant levels) should be presented in physical as well as in monetary terms, and
that the age distribution of cases avoided should be presented whenever possible.
Charge Question 5
Benefit-cost analyses of drinking water regulations are likely to entail uncertainties in the
(a) measurement of exposure, (b) measurement of dose-response, (c) valuation of health
outcomes and (d) measurement of costs. The sources of these uncertainties include
measurement error (uncertainty about the average level of arsenic in tap water or of the amount
of tap water consumed) as well as uncertainty about which model to use in describing the
relationship between exposure and response at low doses. In general, there are two approaches
to handling these sources of uncertainty—sensitivity analysis and Monte Carlo simulation. In a
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
sensitivity analysis various assumptions are made about the correct model (e.g., dose response
function) or parameter (e.g., discount rate) to use in the analysis and results are presented for
each set of assumptions. In a Monte Carlo analysis a distribution is assumed for a key parameter
or set of parameters (e.g., the Value of a Statistical Life) and several hundred draws are made
from this distribution. Benefits are calculated for each value of the parameter drawn. This
yields a probability distribution of benefits, whose parameters (e.g., the 10th and 90th percentiles)
can be reported.
We believe that, in the case of model uncertainty, it is appropriate to rely on sensitivity
analysis; however, the assumptions underlying each sensitivity analysis should be clearly spelled
out when presenting results. It is particularly inappropriate to present only the highest and
lowest numbers associated with a set of sensitivity analyses, which may give the reader the false
impression that these constitute the upper and lower bounds of a uniform distribution. For
parameters for which it is possible to specify a distribution, Monte Carlo analysis is desirable.
For example, in the case of the Value of a Statistical Life.)
General Comments on the Benefit-Cost Analysis for Arsenic
The document Arsenic in Drinking Water Rule: Economic Analysis makes a serious
attempt at analyzing the benefits and costs of alternate MCLs for arsenic in drinking water.
Many aspects of the analysis deserve commendation. These include calculating benefits and
costs for different possible MCLs, presenting some breakdown of benefits and costs by system
size, and presenting cost-effectiveness information (cost per cancer case avoided) that would
enable the drinking water standard for arsenic to be compared to other public health programs.
We do, however, have certain criticisms of the computation of the benefits, the
computation of the costs and with the presentation of the results, especially as they appear in the
Executive Summary.
Computation of Benefits
1. In calculating cancer cases avoided, the primary (central case) analysis assumes no
cessation lag between reduction in exposure to arsenic and reduction in cancer risk.
This assumption yields an upper bound io the number of cancer cases avoided by any
MCL. It should be noted that this assumption produces an upper bound to benefits.
Furthermore, alternate assumptions regarding the length of the cessation lag should
be included in the primary analysis and reported in the Executive Summary.
2. Estimates of cancer cases avoided should be broken down by age. The underlying
dose-response function (Morales et al. 2000) predicts reductions in risk by age group;
hence cancer cases avoided can be broken down by age group. It is important for
policy makers and the public to know how many beneficiaries of a regulation are 7
years old and how many are 70.
3. We believe that it is possible to quantify more health endpoints than lung and bladder
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
cancers. Specifically, it appears to us that it should be possible to quantify mortality
from ischemic heart disease, diabetes mellitus and skin cancer. Serious consideration
should also be given to prostate cancer, nephritis and nephrosis, hypertension,
hypertensive heart disease and non-malignant respiratory disease. This
recommendation is, however, subject to approval by the NAS Arsenic Committee.
4. The benefit analysis should present detailed information on non-quantified health
effects in the manner suggested in this report (see Section 2.2 and Appendix 2.2),
rather than simply listing possible health effects.
5. Estimates of avoided non-fatal cancers should be computed in the same fashion as
estimates of avoided fatal cancers. The length of the cessation lag should also be
treated in a parallel fashion. It would be preferable to value avoided non-fatal
cancers using an estimate provided by Magat et al. (1996) of the value of a non-fatal
lymphoma ($3.6 million) rather than using the value of avoiding a case of chronic
bronchitis ($610,000) which is currently used in the analysis.
Computation of Costs
1. When possible, costs should be computed using data for the systems affected
by the proposed arsenic standard(s) rather than national cost data.
2. The costs of complying with the proposed MCLs may be overstated to the extent that
(a) removal of arsenic may also remove other toxic substances; (b) possibilities for
combining ground and surface water to meet the MCL have been overlooked.
3. The capital costs of drinking water treatment should be amortized using the interest
rate that each water system must pay to borrow money, not at the rate of 7% (or 3%)
used in the current analysis.
Presentation of Results
1. The Executive Summary should clearly state the size of the population affected by
each MCL considered in the analysis, as well as the number of systems affected.
2. The Executive Summary should present benefits in physical as well as monetary
terms, including the age distribution of avoided cancers (and other health endpoints,
if possible).
3. The primary case analysis should include sensitivity to the length of the cessation lag
and this should be reported in the Executive Summary.
4. Benefits and costs should be broken down and compared by system size.
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
We appreciate the opportunity to review and provide advice on this important report.
The Science Advisory Board would be pleased to expand on any of the findings described in our
report, and we look forward to your response.
Sincerely,
Dr. William H. Glaze, Chair Dr. Maureen Cropper, Chair
EPA Science Advisory Board Arsenic Rule Benefits Review Panel
EPA Science Advisory Board
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
NOTICE
This report has been written as part of the activities of the Science Advisory Board, a public
advisory group providing extramural scientific information and advice to the Administrator and
other officials of the Environmental Protection Agency. The Board is structured to provide
balanced, expert assessment of scientific matters related to problems facing the Agency. This
report has not been reviewed for approval by the Agency and, hence, the contents of this report
do not necessarily represent the views and policies of the Environmental Protection Agency, nor
of other agencies in the Executive Branch of the Federal government, nor does mention of trade
names or commercial products constitute a recommendation for use.
Distribution and Availability: This Science Advisory Board report is provided to the EPA
Administrator, senior Agency management, appropriate program staff, interested members of the
public, and is posted on the SAB website (www.epa.gov/sab). Information on its availability is
also provided in the SAB's monthly newsletter (Happenings at the Science Advisory Board).
Additional copies and further information are available from the SAB Staff [US EPA Science
Advisory Board (1400A), 1200 Pennsylvania Avenue, NW, Washington, DC 20460-0001; 202-
564-4546].
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
U.S. Environmental Protection Agency
Science Advisory Board
Executive Committee
Arsenic Rule Benefits Review Panel*
CHAIR
Dr. Maureen L. Cropper, Lead Economist, The World Bank, Washington, DC and Professor of
Economics, University of Maryland
Member: Advisory Council on Clean Air Compliance Analysis
OTHER SAB MEMBERS
Dr. Richard Bull, Consulting Toxicologist, MoBull Consulting, Kennewick, WA
Member: Research Strategies Advisory Committee
Drinking Water Committee
Dr. W. Michael Hanemann, Professor, University of California, Berkeley, CA
Member: Environmental Economics Advisory Committee
Dr. V. Kerry Smith, University Distinguished Professor, Department of Agricultural and
Resource Economics, North Carolina State University, Raleigh, NC
Member: Advisory Council on Clean Air Compliance Analysis
CONSULTANTS
Dr. A. Myrick Freeman, Professor, Department of Economics, Bowdoin College, Brunswick,
ME
Dr. Dale Hattis, Research Associate Professor, Center for Technology, Environment, and
Development (CENTED), Clark University, Worcester, MA
Dr. Irva Hertz-Picciotto, Professor, Department of Epidemiology, University of North Carolina,
Chapel Hill, NC.
SCIENCE ADVISORY BOARD STAFF
Ms. Rhonda Fortson, Management Assistant, 1200 Pennsylvania Avenue, NW, Washington, DC,
Phone: (202) 564-4563, Fax: (202) 501-0582, (fortson.rhonda@epa.gov)
Ms. Wanda Fields, Management Assistant, 1200 Pennsylvania Avenue, NW, Washington, DC,
Phone: 202-564-4539, Fax: 202-501-0582, (fields.wanda@,epa.gov)
Mr. Thomas Miller, Designated Federal Officer, 1200 Pennsylvania Avenue, NW, Washington,
DC, Phone: 202-564-4558, Fax: 202-501-0582, (miller.tom@,epa.gov)
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
TABLE OF CONTENTS
1. INTRODUCTION
1.1 Background
1.2 Charge to the Panel
2. RESPONSES TO CHARGE QUESTIONS
2.1 The impact of timing of exposure on avoided cancers (Charge Question 1)
2.1.1 Introduction
2.1.2 Calculation of reduced cancer fatalities associated with reduced exposure to
a carcinogen
2.1.2.1 The timing of the exposure-response relationship
2.1.2.2 Calculating the time path of cancer cases avoided
2.1.3 Quantifying the relationship between exposure and mortality risk
2.2 Characterization of non-quantified health endpoints (Charge Question 2)
2.2.1 Overview
2.2.2 Quantifiability of particular health endpoints
2.2.2.1 Cardiovascular disease endpoints
2.2.2.2 Diseases of the endocrine system
2.2.2.3 Other cancer sites
2.2.2.4 Non-malignant respiratory diseases
2.2.2.5 Reproductive effects
2.2.2.6 Neurologic and neurodevelopmental endpoints
2.2.3 Valuation of non-quantified health endpoints
2.3 Exposure reduction as a benefit category (Charge Question 3)
2.4 Comparison of benefits and costs (Charge Question 4)
2.4.1 Comparison of benefits and costs by system size
2.5 Incorporation of uncertainty into benefits measures (Charge Question 5)
3. GENERAL COMMENTS ON THE ECONOMIC ANALYSIS
3.1 Comments on exposure assessment
3.1.1 Overstatement of reductions in exposure
3.1.2 Characterization of U.S. population exposure in the analysis
3.2 Comments on the computation of benefits
3.2.1 Treatment of latency
3.2.2 Treatment of age
3.2.3 Valuing avoided cancer mortality
3.2.4 Valuing avoided cancer morbidity
111
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
3.3 Comments on the computation of costs
3.3.1 Factors that may cause costs to be overstated
3.3.2 Amortization of costs
References R-1
Appendix 1. Background Documents
Appendix 1.1 Safe Drinking Water Act Provisions
Appendix 1.2 NOW AC Benefits Workgroup Recommendations, October 1998
Appendix 2. Supplementary Information
Appendix 2.1. Appendix to Charge Question 1
Appendix 2.2. Appendix to Charge Question 2
IV
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
1. INTRODUCTION
1.1 Background
According to information provided by EPA (letter from Diane Regas, June 9,
2001), studies have linked long-term exposure to arsenic in drinking water to cancer of
the bladder, lungs, skin, kidney, nasal passages, liver, and prostate. Non-cancer effects of
ingesting arsenic include cardiovascular, pulmonary, immunological, neurological, and
endocrine (e.g., diabetes). The current standard of 50 ppb was set by EPA in 1975, based
on a Public Health Service standard originally established in 1942. A March 1999 report
by the National Academy of Sciences concluded that the current standard does not
achieve EPA's goal of protecting public health and should be lowered as soon as possible.
The Safe Drinking Water Act (SDWA) requires EPA to revise the existing 50
parts per billion (ppb) arsenic standard. In response to this mandate, the Agency
published a standard of 10 ppb to protect consumers against the effects of long-term,
chronic exposure to arsenic in drinking water on January 22, 2001. The rule is
significant in that it is the second drinking water regulation for which EPA has used the
discretionary authority under §1412(b)(6) of the SDWA to set the Maximum
Contaminant Level (MCL) higher than the technically feasible level, which is 3 ppb for
arsenic - based on a determination that the costs would not justify the benefits at this
level. The January 22, 2001 arsenic rule is based on the conclusion that a 10 ppb MCL
maximizes health risk reduction at a cost justified by the benefits.
Key stakeholder concerns on the benefits component of the economic analysis
include the following issues: (1) the timing of health benefits accrual (latency); (2) the
use of the Value of Statistical Life as a measure of health benefits; (3) the use of
alternative methodologies for benefits estimation; (4) how the Agency considered non-
quantifiable benefits in its regulatory decision-making process; (5) the analysis of
incremental costs and benefits; and (6) the Agency's assumption that health risk
reduction benefits will begin to accrue at the same time costs begin to accrue.
The January 22, 2001 rule will apply to all 54,000 community water systems and
requires compliance by 2006. A community water system is a system that serves 15
locations or 25 residents year-round, and includes most cities and towns, apartments, and
mobile home parks with their own water supplies. EPA estimates that roughly five
percent, or 3000, of community water systems, serving 11 million people, will have to
take corrective action to lower the current levels of arsenic in their drinking water. The
new standard will also apply to 20,000 "non-community" water systems that serve at
least 25 of the same people more than six months of the year, such as schools, churches,
nursing homes, and factories. EPA estimates that five percent, or 1,100, of these water
systems, serving approximately 2 million people, will need to take measures to comply
with the January 22, 2001 rule. Of all of the affected systems, 97 percent are small
systems that serve fewer than 10,000 people each.
1.2 Charge to the Panel
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
The Science Advisory Board (SAB) was asked to conduct a review of the benefits
analysis prepared by EPA in support of the arsenic drinking water standard which is
contained in its regulatory support document Arsenic in Drinking Water Rule Economic
Analysis (USEPA, 2000). The Agency asked that the Panel evaluate whether the
components, methodology, criteria and estimates reflected in EPA's analysis are
reasonable and appropriate in light of 1) the Science Advisory Board's (SAB) benefits
transfer report (SAB, 2000, Report on EPA's White Paper, Valuing the Benefits of Fatal
Cancer Risk Reduction), 2) EPA Guidelines for Preparing Economic Analyses (USEPA,
2000a), 3) relevant requirements of SDWA, 4) the Report of the Benefits Working Group
of the National Drinking Water Advisory Council (NDWAC unpublished, October 1998),
and 5) recent literature. Specifically, the Agency asked that the Panel consider the
following issues:
Charge Question 1: How should latency be addressed in the benefits estimates
when existing literature does not provide specific quantitative estimates of latency
periods associated with exposure to arsenic in drinking water?
Charge Question 2: How should health endpoints (other than bladder and lung
cancer) be addressed in the analysis, when [existing] literature does not provide
specific quantification, to ensure appropriate consideration by decision makers
and the public?
Charge Question 3: Should reduction/elimination of exposure be evaluated as a
separate benefits category, in addition to or in conjunction with mortality and
morbidity reduction?
Charge Question 4: How should total benefits and costs and incremental benefits
and costs be addressed in analyzing regulatory alternatives to ensure appropriate
consideration by decision makers and the public?
Charge Question 5: How should uncertainties be addressed in the analysis to
ensure appropriate consideration by decision makers and the public?
Responses to these questions, and to other issues the Committee wishes to
address, are provided to the Agency below.
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
2. RESPONSE TO THE CHARGE QUESTIONS
2.1 The Impact of the Timing of Exposure on Avoided Cancers
Charge Question 1: How should latency be addressed in the benefits estimates
when existing literature does not provide specific quantitative estimates of latency
periods associated with exposure to arsenic in drinking water?
1.1.1. Introduction
A central component in analyzing the benefits of reduced exposure to a
carcinogen is to predict the annual reduction in cancer cases following reduction in
exposure. If a population previously exposed to 50 ppb of arsenic in drinking water is
exposed, beginning in 2006, to only 10 ppb, cancer risks in the population will eventually
decline to a steady state level associated with a lifetime of exposure to 10 ppb. How fast
this reduction in risk occurs depends on the cessation lag following reduction in
exposure.1
In order to explain what should be done when the length of this cessation lag in
unknown, we must describe how the timing of the relationship between exposure and
response (death due to cancer) should be treated in a benefits analysis. As in the case of
arsenic, we analyze a policy that would reduce exposure from a current level of d° (e.g.,
50 ppb) to d0 (e.g., 10 ppb). We assume that this policy would continue into the
indefinite future.
For a benefits analysis we would like to:
(A) Calculate the expected number of cancer fatalities avoided each year, as a result of
the policy, beginning with the year in which the policy is enacted and continuing into
the future.
If benefits are to be monetized:
(B) The expected number of cancer fatalities avoided each year should be multiplied by
the value of a statistical life in that year. This will give the dollar value of benefits
each year, beginning with the year in which the policy in enacted. The dollar value of
benefits in each year should be discounted to the year in which the policy is enacted
and summed. The present discounted value of benefits, so calculated, should be
compared with the present discounted value of costs, calculated over the same period.
The timing of the relationship between exposure and cancer mortality is implicit in
the calculations in (A). As described more fully below, if the lag between reduction in
1 We believe that this is more appropriately termed a "cessation lag," rather than
"latency." This distinction is clarified below.
-------
_ PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
exposure and reduction in risk of death is long, there will be fewer cancer fatalities
avoided in years immediately following the policy than if the lag were shorter.
Uncertainties in the timing of the exposure-response relationship will be reflected in
uncertainties in the number of cancer fatalities reduced each year after the policy is
enacted. These uncertainties should be treated as described in the answer to Charge
Question 5.
2.1.2 Calculation of Reduced Cancer Fatalities Associated with Reduced
Exposure to a Carcinogen
The approach taken here is to relate the age-adjusted risk of death due to cancer to
history of exposure to the carcinogen. This relationship, together with information on the
age distribution of the population affected by the policy, can be used to calculate the
expected number of cancer fatalities avoided by the policy.
The epidemiology underlying the arsenic benefits analysis (Morales et al. 2000)
assumes that the conditional probability of dying from cancer at age t, h(t) is related to
cumulative exposure to a carcinogen as of age t, jq, by a proportional hazard model:
(1) h(t,x) =
where ho(t) = baseline risk of dying from cancer at age t (assuming survival to age t).2
2.1.2.1 The Timing of the Exposure-Response Relationship
The key question is how cumulative exposure (xj depends on the dose of arsenic
received at ages 0 through t. Let dj = dose received at age i. A general form that this
relationship could take is:3
(2) xt = ft(d0,d1,...,dt)
The exact form of this function reflects the answers to the following four
questions (Tollerud et al. 1999):
(a) How long does it take after an exposure until an increase in risk is observed?
(b) How long does the effect of an exposure last after exposure has terminated?
(c) How does the effect of exposure vary by the age at which it was received?
(d) Does the exposure act at an early or late stage in the carcinogenic process?
2 A proportional hazard model (Pope et al. 1995) is also used to measure the association
between particulate matter and all-cause mortality in The Benefits and Costs of the Clean
Air Act, 1970-1990 (USEPA 1997) and The Benefits and Costs of the Clean Air Act,
1990-2010 (USEPA 1999). The issue of the length of the cessation lag after a reduction
in exposure also arises in these studies.
3 The function £( ) could also be conditioned on other factors such as smoking.
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
The relevant questions for the implementation of changes in the drinking water
standard for arsenic are questions (b)-(d). In contrast, most of the epidemiologic literature
addressing the issue of latency has focussed on question (a), which is the usual definition of
latency. The committee wishes to underscore that data addressing question (a) do not
necessarily provide information answering questions (b)-(d). Unfortunately, much less
work has been done to evaluate questions (b)-(d) in the epidemiologic literature in general,
and in the research on arsenic carcinogenicity in particular.
The NAS report Veterans and Agent Orange: Update 1998 (Tollerud et al. 1999)
addresses the second question, regarding how long effects last after cessation of
exposure. With respect to arsenic in drinking water, the charge of our committee is an
expansion of this question: when does the excess risk (compared to a lifetime of exposure
to d' (e.g., 10 ppb)) begin to attenuate and how long does it take until all of the excess is
eliminated? Since the term latency has a traditional usage that is not the charge given to
this committee, we have coined the phrase "cessation-lag" to clarify and emphasize the
difference.
An important point is that the time to benefits from reducing arsenic in drinking
water may not equal the estimated time since first exposure to an adverse effect. A good
example is cigarette smoking: the latency between initiation of exposure and an increase
in lung cancer risk is approximately 20 years. However, after cessation of exposure, risk
for lung cancer begins to decline rather quickly. A benefits analysis of smoking
cessation programs based on the observed latency would greatly underestimate the actual
benefits. We return to the issue of how to estimate the length of the cessation lag below.
2.1.2.2 Calculating the Time Path of Cancer Cases Avoided
If the relationships in (1) and (2) are known, it is, in principle, a simple matter to
compute the expected number of cancer fatalities avoided at age t (and, by analogy, for
all other ages) in each year following the policy. In the first year of the policy it is only
the most recent dose of the carcinogen (dj for persons who are age t in the year the
policy is enacted) that is affected by the policy. The expected reduction in risk of death
due to cancer at age t in the first year of the policy is:
(3) hoCtMKdoA,..,^0)) - g(ft(d0,d1,...,d/))]
where the superscripts ° and ' refer to doses with and without the policy, respectively.
In the second year of the policy, for persons of age t, both d^ and dt are affected by the
policy, and the formula in (3) would be adjusted accordingly. Eventually, a steady-state
will be reached in which persons of age t face the same mortality risk from cancer as
people who have been exposed to the lower level of the carcinogen (d') throughout their
lifetime.
In each year, the number of fatalities avoided by the policy among persons of age
t would be the expression similar to (3) multiplied by the number of persons of age t.
Similar computations would be performed for persons of all ages. In this manner, it
-------
_ PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
should be possible to compute the expected number of fatalities avoided, by age (or age-
group), in each year following the enactment of the policy. Because the age distribution
of avoided cancer fatalities is calculated, it should be reported in a benefits analysis even
if information on the age distribution of avoided fatalities is not used in valuing avoided
mortality.
2.1.3 Quantifying the Relationship Between Exposure and Mortality Risk
Most epidemiologic studies ignore the time pattern of exposure in estimating the
proportional hazard model in equation (1). For example, Morales et al. (2000)
effectively assume that
(4) x^Sdi.
i=0
Estimating the time pattern of exposure and effect in the context of equations (1)
and (2) is not trivial. In order to properly study effects of protracted exposures, detailed
exposure histories for each study subject, including the dates and ages when the
individual was exposed and the level of exposure at all points in time, are needed.
Appropriate statistical methods have been developed for an investigation of the effect of
exposure accrued as a function of time since that exposure (Thomas, 1983; Breslow and
Day, 1987; Thomas, 1988). In general, the ability to investigate the issues of timing of
exposure in a given data set will depend on the quality of the exposure measure, the
quality of the timing of exposure information, the number of people developing the
disease of interest, and variation of exposure over time within the study group. These
aspects of study quality are, of course, important in evaluating any epidemiologic
investigation. But there are special problems that arise in the evaluation of time-related
factors (Enterline and Henderson, 1973; Peto, 1985; Thomas, 1987).
Appendix 2.1 to this report further discusses possible methods for estimating the
time pattern of exposure and response. If, however, such estimation is impossible (as the
charge question assumes), what can be done?
One extreme assumption that would yield an upper bound to the benefits of the
program is to assume that the program immediately attains the steady-state result, i.e.,
that the reduction in the age-t mortality rate is given by:
(5) hoCtMfldMV.A0)) - g(ft(d0',d/,...,dt'))].
This assumption is implicit the Agency's primary analysis.
In the absence of data that would make it possible to estimate the cessation lag
and account for it as described above, it would still be desirable to examine the influence
of this lag by performing sensitivity analyses similar to those carried out for the PM -
mortality relationship in the Agency's analysis of The Benefits and Costs of the Clean
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
Air Act: 1990-2010 (USEPA 1999). The Agency assumed alternative time patterns for
the reduction in mortality risk following the reduction in PM concentrations, for
example, assuming an equal percentage reduction in risk each year for a 15 year period,
and then calculated the present value of the stream of deaths avoided. We recommend
that similar sensitivity analyses be conducted here.
2.2. Characterization of Non-Quantified Health Endpoints
Charge Question 2: How should health endpoints (other than bladder and lung
cancer) be addressed in the analysis, when [existing] literature does not provide
specific quantification, to ensure appropriate consideration by decision makers
and the public?
2.2.1 Overview
The scientific literature on health effects due to arsenic exposure includes studies
of a number of endpoints other than cancer, as well as studies of several cancer sites for
which the risks/benefits have not been quantified (USEPA 2000). The quality of these
studies varies, as does the strength of evidence they provide. Nevertheless, this body of
evidence is relevant for the determination of an MCL and needs to be addressed more
fully. In some cases, the non-quantified effects can and should be quantified. In other
words, the lack of quantification by EPA, to date, of these effects should not be construed
to mean that they are not quantifiable.
Of the 49 non-quantified non-carcinogenic health effects listed in the Benefits
Analysis (USEPA 2000), some would not be relevant at low exposure levels, e.g., at or
below the current standard. These would include gangrene in adults or children, hepatic
enlargement, Raynaud's syndrome and others. The main categories for which there may
be concern at lower exposure levels are: several cardiovascular and cerebrovascular
diseases, endocrine effects (diabetes mellitus), reproductive health outcomes, and non-
malignant respiratory diseases. Some data have emerged for neurologic or
neurodevelopmental outcomes, but this evidence is currently somewhat sparse.
Studies addressing the major categories of concern at lower exposure levels are
listed in the tables in Appendix 2.2 (which are not comprehensive, but rather,
representative). These studies demonstrate a broad array of related endpoints and
indicate the range and weight of evidence, qualitatively, as well as the consistency with
which these effects are related to arsenic exposure. Such consistency, particularly when
at least some of the studies are of high quality and have adjusted for individual-level
confounders, strengthens the evidence for causality.
Given (a) the consistency of results, including supportive in vivo animal
experiments; (b) epidemiologic studies with individual level data on exposure, outcomes,
and confounders; and (c) evidence suggesting plausibility of effects at low exposures: the
Panel finds that for several of these health endpoints, the benefits can and should be
quantified. These include, at a minimum, mortality from ischemic heart disease, diabetes
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
mellitus, and skin cancer. Serious consideration should also be given to prostate cancer,
nephritis and nephrosis, hypertension, hypertensive heart disease, and non-malignant
respiratory disease. The discussion below briefly assesses the broad groupings of
outcomes, highlighting those for which quantification appears to be eminently feasible.4
We also discuss in Appendix 2.2 a "public health based" approach that suggests potential
order-of-magnitude effects for deaths due to other cancers and to cardiovascular disease.
The type of information that should be provided in a benefit-cost analysis about
endpoints that have not been quantified is listed in the tables in Appendix 2.2. For each
health endpoint (e.g., cardiovascular morbidity), studies that pass certain scientific
criteria should be listed.5 The information that should be provided for each study
includes:
• Nature of the study design
• How exposure was measured
• Range of exposures observed
• What type of statistical analysis was conducted and what confounding factors
were controlled for in the analysis
• Measure of association (e.g., odds ratio) and level of statistical significance of the
association
In some cases the literature may be so extensive that a summary of results is required
in the text of the report. This summary should focus on health endpoints that have
meaning to humans (e.g., ischemic heart disease rather than electrocardiogram
abnormalities), and should provide some discussion of the mechanism by which the toxin
would be expected to exert an effect. The summary should also indicate the level at
which effects were observed in the studies reported and should comment on the
likelihood of observing these effects at the levels relevant to the regulatory decision.
2.2.2 Quantifiability of Particular Health Endpoints
2.2.2.1. Cardiovascular disease endpoints (see tables I, II, and III in Appendix
2.2).
Both human and animal studies provide evidence that arsenic affects the
4 Notably, these outcomes are not all independent. For instance, arsenic is associated
with increased prevalence of hypertension, and with increased incidence of ischemic
heart disease. Within the studies assessing the latter, hypertension was a strong risk
factor. Thus, hypertension may be one step along one or more pathways by which
arsenic increases risk for ischemic heart disease. Nonetheless, hypertension can itself be
a cause of death, though this occurs much more rarely than death due to ischemic heart
disease.
5 For an example of such criteria see Table 5-2 in The Benefits and Costs of the Clean Air
Act, 1990-2010 (USEPA 1999) which lists the criteria used to select studies that examine
the health effects of the criteria air pollutants.
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
cardiovascular system, possibly via several mechanisms. The human studies have
included both occupational cohorts for which exposure is primarily by inhalation, and
communities for which exposure is primarily via drinking water. Both morbidity
(Lagerkvist et al.1986, Chen et al.1988, Chen et al.1995, Tseng et al.1996, Chiou et
al.1997, Rahman et al.1999, Hsueh et al.1998, Tsai et al.1999), and mortality (Axelson et
al.1978, Wu et al.1989, Engel et al.1994, Chen et al.1996, Tsai et al.1999, Lewis et
al. 1999, Hertz-Picciotto et al.2000) have been addressed in these investigations. Several
tables in Appendix 2.2 illustrate the range of types of studies and exposure levels at
which these effects have been observed.
The Taiwanese study by Chen et al., 1996 on mortality from ischemic heart
disease is particularly interesting, in that a wide range of individual-level confounding
factors were adjusted in the analysis, including age, sex, smoking, body mass index,
serum cholesterol level, serum triglyceride level, blackfoot disease, hypertension and
diabetes. Their adjustment for the latter two chronic diseases that may themselves
contribute to ischemic heart disease risk could have attenuated the effects, although the
relative risks are reduced only modestly by the inclusion of the confounders other than
blackfoot disease. Nevertheless, there is a strong dose-response relationship, rising from
2-fold to 5-fold increased risks according to the cumulative exposure level.
Another study from Taiwan, by Tsai et al. (1999), relied on vital statistics, and
hence did not collect the individual-level confounding data included by Chen and
colleagues. However, these authors present analyses for a broader list of causes of
mortality, including diabetes, hypertension, pulmonary heart disease, cerebrovascular
disease, liver cirrhosis, and a host of other non-cancer and cancer endpoints. The
findings on lung and bladder cancer confirm those of numerous other investigators;
results for ischemic heart disease are similarly consistent with those of Chen et al.(1996)
and others. Additionally, the study presents information on some health outcomes some
outcomes not previously observed in arsenic-exposed populations.
Whereas most of the studies on cardiovascular endpoints have been conducted in
communities with long and heavy exposures, a few were conducted in a population with
more relevant levels. For instance, Lewis et al.(1999) examined records from the
Mormon Church from towns in Utah with concentrations in drinking water of 18-164
ppb. These authors found mortality due to hypertensive heart disease to be elevated in
both males and females. Although individual-level confounder data were not available,
the church's prohibitions on consumption of alcohol and caffeine would tend to minimize
this problem; the extremely low rates of respiratory cancer and non-malignant respiratory
disease attest to the low level of smoking in this population, which may also explain the
low incidence of ischemic heart disease. In another study relevant for low level
exposures, Gomez-Caminero (2001) examined several biomarkers of subclinical
cardiovascular damage comparing a population exposed at 45 • g/L in drinking water to
one with negligible exposures (<2 • g/L). Among pregnant women residing in the
exposed community, the levels of von Willebrand factor were significantly reduced as
compared with those in unexposed pregnant women. The important point is that these
data suggest damage to the endothelium of the arterial walls at levels just under the
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
current standard of 50 • g/L. The vascular endothelium serves as a barrier between blood
plasma and the arterial smooth muscle and regulates the flow of lipoproteins between
these compartments. Arsenic may damage the endothelium directly or restrict its repair or
regenerative capacity, by inhibiting endothelial cell hyperplasia. Reduced von
Willebrand factor could play a role in this process.
It is also notable that, in the past, clinical cardiovascular effects normally only
seen in adults were observed in children at very high exposure levels. The possibility
that subclinical damage to the cardiovascular system occurs in early life, setting the stage
for severe and potentially fatal events at older ages, should be considered.
The Panel concludes that cardiovascular effects of arsenic are plausible at current
levels in drinking water. Despite uncertainty in the shape of the dose-response curve, a
benchmark dose approach would be a reasonable method for estimating benefits from
reduction of the MCL. To place the epidemiologic findings with regard to ischemic heart
disease in context, over 500,000 deaths occurred in the U.S. in 1999 due to this cause, or
22% of all deaths. Undoubtedly the overwhelming majority of these are not due to
arsenic. However, the same can be said for lung and bladder cancer in the general
population. Given the large background incidence of ischemic heart disease, the
committee believes these effects/benefits should be quantified. A similar argument
would apply to the morbidity and mortality from hypertension.
Peripheral vascular disease is a well-established effect of high exposures to
arsenic, to the extent that the presence of one severe form of this condition, blackfoot
disease, has been used as an indicator of exposure. There is probably little direct
relevance of the extreme manifestations of this condition for lower exposures. The
likelihood of less severe conditions at low exposures is uncertain.
2.2.2.2 Diseases of the Endocrine System (see table IV, Appendix 2.2).
Most of the epidemiologic literature demonstrating increased risk of diabetes in
association with arsenic exposure has been published in the last five years (Tsai et
all999, Lai et all994, Tseng et al.2000, Rahman et all996). Studies include
occupational and drinking water sources for exposure, and both mortality and morbidity
studies have found significant excesses. Generally speaking, because diabetes is not a
common cause-of-death, mortality studies would be expected to observe only the tip of
the iceberg in terms of increased incidence. However, even when not fatal, diabetes
engenders large medical costs and has a serious, lifelong impact on the quality of life.
Besides clinical disease, glucosuria and elevated glycosylated hemoglobin have
both been found in association with arsenic exposure (Jensen and Hansen, 1998, Rahman
et all999, Gomez-Caminero 2001). These are biologically significant markers of
impaired glucose metabolism. Glycosylated hemoglobin represents an indicator of long-
term glycemic control. The Chilean population examined by Gomez-Caminero (2001),
for which exposures were -45 • g/L, was found to have significantly elevated
glycosylated hemoglobin, both when this biomarker was treated as a continuous measure
10
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
(% of hemoglobin glycosylated), and when it was dichotomized (>6.5% vs. <6.5%).
Since these women were pregnant, the age range was fairly young and therefore the
majority were born after levels were reduced to about 110 • g/L, which occurred around
1970 (Hopenhayn-Rich et al., 2000). As the risk of diabetes increases with age, the
findings may indicate that the effects of arsenic on glycemic status could begin early,
laying the basis for clinical disease that manifests primarily beyond the reproductive
years (i.e., Type II diabetes).
Evidence for the diabetogenicity of arsenic is mounting, plausible mechanisms
have been shown, subclinical markers of altered glycemic control have been observed,
and there appears to be relevance at low exposures. Diabetes was directly responsible for
68,000 deaths in the U.S. in 1999, representing 2.9% of deaths, more than five times as
many as occurred due to bladder cancer. Quantification of the benefits of reducing the
arsenic MCL in terms of diabetes mortality, as well as the multidimensional costs
associated with chronic illness, is appropriate. Any effect that arsenic has in increasing
the incidence or advancing the onset of Type II diabetes will contribute to the risks of
many other diseases associated with arsenic exposure (e.g. hypertension, cardiovascular
disease, liver cancer, peripheral vascular disease).
2.2.2.3 Other cancer sites (see table V, Appendix 2.2).
Increased risks for kidney, liver, skin, bone, prostate, laryngeal, nasal and other
sites are observed to occur in arsenic-exposed populations (Lewis et al. 1999, Smith et
al.1992, Tsai et al.1999). A comprehensive accounting of benefits from the reduction in
the arsenic MCL should quantitate at least the strongest of these effects, accounting for
uncertainly. Recent studies on the mechanisms for arsenic carcinogenicity do not suggest
that lung and bladder would be the only sites affected. An excess of prostate cancer was
associated with cumulative arsenic exposures above 1 ppm year in Utah.
2.2.2.4 Non-malignant respiratory diseases (see table VI, Appendix 2.2).
The increased incidence of bronchitis, emphysema, respiratory symptoms, and
chronic airway obstruction are surprising for exposures that do not occur via inhalation.
At high exposures, strong dose-response relationships were found for respiratory
symptoms (Mazumder et al.2000). Plausibility for these effects at low exposures is
uncertain. Shortness of breath was elevated at 50-199 • g/L in West Bengal (Mazumder et
al.2000), and an ecologic study in the U.S. found mortality was increased from chronic
airways obstruction and emphysema at levels as low as 5-10 • g/L, with the highest risk at
>20 • g/L (Engel and Smith 1994). This latter finding suggests the possibility that
communities with somewhat higher arsenic concentrations in drinking water (e.g., >20
• g/L) may also include a higher proportion of smokers. Two concerns are: first, that
smoking could be a confounder, and second, that smoking and arsenic could have
synergistic effects. Since smoking acts synergistically with arsenic in producing lung
cancer (Hertz-Picciotto et al.1992), a similar interaction for non-malignant respiratory
diseases is possible. Although smoking is a voluntary risk, smokers do constitute a
susceptible subgroup.
11
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
2.2.2.5 Reproductive effects (see table VII, Appendix 2.2).
Few reproductive endpoints have been examined in more than one study. Most of
the spontaneous abortion studies were conducted in populations with high exposures;
those that were not did not have individual data on confounders, and hence little
confidence can be placed in the results. The time trend analyses by Hopenhayn-Rich et
al.(2000) suggest that stillbirths and postneonatal mortality are increased at high
exposures but not at levels between 40 and 70 • g/L; the decline in rates in the exposed
region after arsenic levels are reduced may be partially attributable to other
improvements in water quality and standard of living. In contrast, an effect on
birthweight may be seen at lower levels, based on the studies to date. Transfer of arsenic
to the fetus has been shown; interestingly, blood plasma arsenic was essentially all in the
form of DMA, and pregnant women had a higher proportion of their urinary arsenic as
DMA than nonpregnant women (Concha et al. 1998), suggesting more efficient
methylation during pregnancy.
2.2.2.6 Neurologic and Neurodevelopmental Endpoints (see table VIII, Appendix
2.2).
There have been studies indicating associations between environmental exposures
and pathologies, symptoms, and developmental deficit.
2.2. 3 Valuation of Non-Quantified Health Endpoints
The preceding discussion suggests that some health endpoints affected by arsenic
exposure, including skin cancer and ischemic heart disease could be quantified. That is, the
expected reduction in cases could be calculated for each endpoint (possibly by age group) for
each year following the reduction in exposure. If the magnitudes of these effects can be
characterized, valuation should be done in the same way as for bladder and lung cancers. (See
Charge Question 1.)
Two issues, however, arise: (1) Do unit values exit for all of the health endpoints that
can be quantified? (2) Should valuation be done if effects cannot be quantified?
To answer the first question, unit values that measure what individuals would pay to
avoid adverse health effects (Willingness-to-Pay estimates) do not exist for all health endpoints
mentioned in our answer to Charge Question 2. The Benefits and Costs of the Clean Air Act,
1990-2010 (USEPA 1999) contains a recent review of the available data for at least some of the
relevant endpoints. Where only cost of illness estimates are available, they can be used but
should be clearly described as providing lower bounds on true willingness to pay. (The EPA
Cost of Illness Handbook is a recent source for cost of illness data for some relevant endpoints.)
To make economic valuation possible, it is important to express and characterize these
other endpoints in terms of effects on people's activity levels and sense of well-being, as much
12
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
as possible. There is a fairly extensive body of data on the economic values of reducing days
experiencing various symptoms, restricted activity days, hospitalizations, required treatments,
etc. It would be difficult to use this body of data to value many of the health effects listed in
Exhibit 5-1 (p. 5-4 of the arsenic economic analysis) such as hepatic enlargement, anemia,
leukopenia, peripheral neuropathy, since the clinical significance and impact on individuals'
activities of these effects is not clear.
To answer the second question raised above, it is not possible to value health effects that
have not been quantified.
2.3 Exposure Reduction as a Benefit Category
Charge Question 3: Should reduction/elimination of exposure be evaluated as a
separate benefits category, in addition to or in conjunction with mortality and morbidity
reduction?
Regarding Charge Question 3, the Panel believes that reductions in exposure should not
be considered a separate category of benefits in a benefit cost analysis. The damage function
approach to valuing benefits currently used by the Agency separates the measurement of the
relationship between exposure and response (e.g., risk of fatal or non-fatal cancer) from the
valuation of reductions in risk of death or illness. Epidemiologists estimate dose-response
functions and economists measure the value people place on reductions in risk of death or
illness. Reductions in exposure are already valued under the damage function approach when
people value the reductions in the risk of death or illness associated with them. To add a
separate value for reductions in exposure per se would be double counting.
One might argue that if some benefits from reducing arsenic exposure have not been
quantified (or monetized), then an additional value should be added for reductions in exposure
per se. There is, however, no practical way of doing this. Extending the set of health endpoints
in terms of mortality and morbidity effects that can be quantified in some way (as we
recommend in our answer to Charge Question 2) is the appropriate basis for developing a more
complete benefit analysis, not attaching an ad hoc value to reductions in exposure.
It might be argued that EPA should abandon the damage function approach to valuing
health benefits and ask people to value reductions in exposure directly. This, however, seems
unwise. To abandon the damage function approach and ask people to value reductions in
exposure directly would force lay people to act as epidemiologists, and there is evidence that this
is difficult. Malmfors et al. (19XX) have shown that lay people do not view risk of death or
illness as related to the size of the dose of a toxic substance received; any dose, however small,
poses an equivalent risk. This is consistent with other studies that show that people attach a
premium to reducing risks of adverse outcomes to zero (Viscusi et al., 19XY). The essence of
modern risk assessment is, however, to relate health outcomes to the size of the dose received.
13
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
2.4 Comparison of Benefits and Costs
Charge Question 4: How should total benefits and costs and incremental benefits and
costs be addressed in analyzing regulatory alternatives to ensure appropriate
consideration by decision makers and the public?
2.4.1 Comparison of Benefits and Costs by System Size
One noteworthy feature of the arsenic in drinking water problem is that for the most part,
those who would receive the health benefits from reductions in the concentrations of arsenic in
drinking water will also bear the costs of achieving them. These costs will take the form of
higher rates and prices for water supply and/or higher taxes to cover these costs. Thus it is
important to try to determine whether those who receive these benefits would be willing to bear
the costs of reducing arsenic concentrations in their drinking water. This is the question that
benefit-cost analysis attempts to answer, because in principle the benefits of a program are
defined as the sum of the affected individuals' willingness to pay for these improvements. If all
benefits and costs of a regulation are measured accurately, and if benefits are less than costs, this
is a signal that if the people receiving the benefits had to pay these costs, they would consider
themselves to be made worse off. Conversely, if benefits exceed costs, the policy would make
them better off.
For this reason, we recommend that benefits and costs should be calculated on a water
supply system basis. Because of both the variability of costs and benefits across systems and the
non-linearities in how benefits and costs vary with alternative regulatory standards, aggregation
can produce inaccurate results. Therefore, rather than calculating the total benefits across all
affected systems and the total costs across all affected systems, and then using these aggregate
results to calculate total net benefits, marginal benefits, marginal costs and marginal net benefits,
we recommend that total benefits, costs and net benefits and marginal benefits, costs and net
benefits should be calculated for each system that is affected by the standard, and the system-
level results should then be aggregated to the national level.
While there are too many affected systems to perform a separate cost analysis tailored to
the specific circumstances of every system, nevertheless the existing cost analysis appears to be
too generic and too little tailored to the specific circumstances of the particular utilities affected
by arsenic regulation (e.g., water supply systems in the west and southwest that use
groundwater). Rather than using national cost functions, an attempt should be made to employ
cost functions tailored to these affected utilities. Grouping utilities into size classes and
conducting an analysis by size class is acceptable if this is done with specific reference to size
classes that are meaningful for the systems affected by the arsenic regulation and using data
specific to these systems. In the existing analysis, individual cost analyses were only performed
for water utilities that serve more than a million people ("very large systems"); we recommend
lowering the threshold population size for performing individual cost analyses, for example to a
service population of 250,000 or more.
14
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
2.5 Incorporation of Uncertainty into Benefits Measures
Charge Question 5: How should uncertainties be addressed in the analysis to ensure
appropriate consideration by decision makers and the public?
Benefit-cost analyses of drinking water regulations are likely to entail uncertainties in the
(a) measurement of exposure, (b) measurement of dose-response, (c) valuation of health
outcomes and (d) measurement of costs. The sources of these uncertainties include
measurement error (uncertainty about the average level of arsenic in tap water or of the amount
of tap water consumed) as well as uncertainty about which model to use in describing the
relationship between exposure and response at low doses. In general, there are two approaches
to handling these sources of uncertainty—sensitivity analysis and Monte Carlo simulation. In a
sensitivity analysis various assumptions are made about the correct model (e.g., dose response
function) or parameter (e.g., discount rate) to use in the analysis and results are presented for
each set of assumptions. In a Monte Carlo analysis a distribution is assumed for a key parameter
or set of parameters (e.g., the Value of a Statistical Life) and several hundred draws are made
from this distribution. Benefits are calculated for each value of the parameters drawn. This
yields a probability distribution of benefits, whose parameters (e.g., the 10th and 90th percentiles)
can be reported.
We believe that, in the case of model uncertainty, it is appropriate to rely on sensitivity
analysis; however, the assumptions underlying each sensitivity analysis should be clearly spelled
out when presenting results. It is particularly inappropriate to present only the highest and
lowest numbers associated with a set of sensitivity analyses, which may give the reader the false
impression that these constitute the upper and lower bounds of a uniform distribution. For
parameters for which it is possible to specify a distribution, Monte Carlo analysis is desirable
(for example, in the case of the Value of a Statistical Life).
The EPA analysis of the Arsenic in Drinking Water Rule discusses some of the sources
of uncertainty in benefit estimates and handles them by performing sensitivity analyses.
Specifically, it focuses on the impact of alternate assumptions about the parameters of the dose-
response function, which will vary depending on what fraction of arsenic in the Taiwanese
population (the population used to estimate the dose response function) is assumed to come from
drinking water. A "high" and "low" estimate of benefits are generated based on alternate
assumptions about the sources of arsenic exposure in Taiwan.
The other set of sensitivity analyses that are performed pertain to the Value of a
Statistical Life (VSL). This is varied to allow for (a) changes in the VSL as incomes grow, (b)
the involuntary nature of drinking water risks and (c) the length of the latency period. As we
explain in more detail in the next section, latency (or, more correctly, the cessation lag between
reduction in exposure and reduction in risk) is not handled correctly in the arsenic benefits
analysis. We also have criticisms of the treatment of adjustments for income growth and for the
involuntary nature of drinking water risks. In principle., however, there is nothing wrong with
handling these sources of uncertainty through a sensitivity analysis. The choice of discount rate
is also correctly handled via sensitivity analysis.
15
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
The report could, however, improve in its reporting of the results of these sensitivity
analyses in two ways. First, the presentation of the details of the analysis in the Executive
Summary and in the body of the report does not provide a sufficiently clear description of the
specific details of all aspects of the uncertainty analysis. With considerable effort it is possible
to develop a more complete understanding of how the analysis was undertaken by studying the
appendices to the report. Second, when the results of two alternate assumptions are presented,
for example, the "high" and "low" benefit estimates in the Executive Summary, it is important to
state that these are not the endpoints of a uniform distribution.
16
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
3. GENERAL COMMENTS ON THE ECONOMIC ANALYSIS
3.1 Comments on Exposure Assessment
3.1.1 Overstatement of Reductions in Exposure
The benefits analysis is based on an assumption that the mean concentrations of arsenic
will be 80% of the MCL. The argument is that systems will design treatment to meet an 80%
standard so as to assure that realized concentrations will "never" exceed the MCL. This kind of
overdesign is apparently standard practice in the drinking water industry. However, since the
"overdesign" is to assure that the realized concentrations of arsenic do not exceed the MCL, the
expectation that realized concentrations will at least sometimes exceed 80% of the MCL should
be reflected in the exposure analysis. To the extent that the mean concentration of arsenic is
greater than 80% of the MCL, benefits will be overstated. It is necessary to make an estimate of
the mean concentration actually realized at each MCL.
3.1.2 Characterization of U.S. Population Exposure in the Analysis
There are a few opportunities to improve the presentation of arsenic exposures in the
benefits analysis. First, although the report gives national estimates of the proportion of water
systems of various types that exceed various average arsenic levels, and Tables ni.C-5 and C-6
give helpful breakdowns by geographic region and the system size (population served per
system), there does not appear to be an accessible presentation of the national or regional
numbers of people or population aggregate exposures (people exposed X • g/liter X years/years
of system operation) broken down in the same ways. A breakdown of the numbers of people in
these categories is important for understanding the distributional burdens of both current arsenic
exposures/health harm and the prospective compliance costs. A breakdown of the amounts of
population aggregate exposure in these categories is very important for understanding the extent
to which the national aggregate arsenic-in-drinking water problem would be reduced by different
MCLs.
3.2 Comments on the Computation of Benefits
3.2.1 Treatment of 'Latency'
As the answer to Charge Question 1 implies, we do not believe that the lag between
reduction in exposure and reduction in fatal cancers has been treated correctly in the benefits
analysis. The correct approach is to predict the number of fatal cancers avoided each year based
on an assumption about the percent of the steady-state reduction in cancer cases that will be
achieved each year following the policy. For example, in The Benefits and Costs of the Clean
Air Act, 1990-2010, it was assumed that 25% of the steady state benefits from reducing air
pollution would be achieved in the first year of the policy, 50% by the second year, and
(increasing gradually), 100% of the benefits by the end of the 5th year of the policy.
17
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
Once this time path is established, the number of fatal cancers avoided in year t should
be multiplied by the Value of a Statistical Life in year t and the result discounted to the first
year of the policy. The sum of these present discounted values over the horizon of the analysis
yields the present discounted value of benefits of the policy. It is, of course, possible to
annualize this number by calculating the constant annual value of benefits that produces the
same present discounted value of benefits.
In its primary analysis the Agency makes no adjustment for the cessation lag in its
calculation of cancer mortalities avoided. It simply assumes that the cancer mortality risk will
drop immediately to the new steady state level upon implementation of the new standard. Then
in a sensitivity analysis (Section 5.5), it accounts for the cessation lag not with alternative
calculations of cancer mortalities avoided, but by discounting the Value of A Statistical Life
applied to these avoided deaths for three alternative lag periods, 5, 10, and 20 years. In terms of
the calculated monetary benefits, this is equivalent to assuming there is no reduction in cancer
mortalities avoided for the first 5, 10, 20 years after the regulation is implemented, after which
the cancer mortality risk drops immediately to the new steady state level.
3.2.2 Treatment of Age
There is sufficient information in the dose-response function in Morales et al. (2000) to
calculate cancer cases avoided by age group. We believe that this should be done. The dose-
response function used to compute the number of cancer cases avoided in the benefits analysis
(Model 1 of Morales et al.) is a special case of equation (1) in which "the relative risk of
mortality at any time is assumed to increase exponentially with a linear function of dose and a
quadratic function of age (p. B-7)." Instead of using this equation to predict risks by age group,
the information contained in the equation is aggregated to compute a lifetime cancer risk.
3.2.3 Valuing Avoided Cancer Mortality
(1) The Agency should recognize the uncertainty in the estimated VSL used to value fatal
cancers either by sensitivity analysis or incorporating the uncertainty in Monte Carlo analyses.
(2) The committee does not believe that the adjustments to the VSL for income growth
and the voluntaries/controllability of risk are entirely correct.
The arsenic benefits analysis uses elasticities of 0.22 and 1.0 as lower and upper bounds
(p. 5-31) to adjust the Value of a Statistical Life for income growth. It cites EPA (2000c),
claiming it is a review of the literature on elasticities and establishes the 0.22 -1.0 as the best
range. But EPA 2000c is the "Guidelines," and there is nothing in the guidelines about adjusting
for income or suggesting specific elasticities. The citation is incorrect. In addition, The Benefits
and Costs of the Clean Air Act, 1990-2010 (USEPA 1999), which was reviewed by the Council
for Clean Air Act Compliance and Analysis, used 0.08, 0.40, and 1.0 as low, central, and high
estimates.
18
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
Regarding the adjustments for voluntariness/controllability of risk, the SAB Review of
the EPA's White Paper, Valuing the Benefits of Fatal Cancer Risk Reductions recommended that
no such adjustments be made.
3.2.4 Valuing Avoided Cancer Morbidity
With respect to nonfatal bladder cancers, an alternative to using the value for chronic
bronchitis used in the Section 812 analyses is to use the value for nonfatal lymphoma obtained
by Magat, Viscusi, and Huber (1996). This value is $3.6 million (in 1999$). This value was
derived from a mall intercept survey, rather than a random sample of the U.S. population. But
the end point being valued more closely corresponds to nonfatal bladder cancer than does
chronic bronchitis. Estimates of avoided non-fatal cancers should be computed in the same
fashion as estimates of avoided fatal cancers. The length of the cessation lag should also be
treated in a parallel fashion.
3.3 Comments on the Computation of Costs
3.3.1 Factors that May Cause Costs to Be Overstated
Two features of the existing cost analysis may lead it to overstate the costs of arsenic
regulation, at least to some degree: We recommend, that the Agency attempt to take account of
these factors. (1) To the extent that arsenic removal is a joint product of water treatment together
with the removal of other contaminants, the existing cost analysis may overstate the costs (or
understate the benefits) of arsenic regulation. Utilities may already have pre-existing installed
treatment processes for other contaminants that lower the cost of arsenic removal in a manner
not reflected in the current analysis, or utilities may adopt new treatment processes in response
to arsenic regulation that yield other improvements in drinking water quality as a by-product. (2)
In two of three cases, the existing cost analyses for the very large systems affected by the
arsenic regulations note that the costs may be overstated because they do not account for options
that may be available to lower costs associated with ground water entry points: "Depending on
the spatial distribution of the wells, it may be possible to implement centralized treatment, with
reduced compliance costs. It may also be possible to achieve compliance without treatment by
blending ground water with surface water. Finally, depending on the additional capacity
available from surface water and unaffected well, the city could shut down affected wells."
Presumably, the same considerations apply to some of the other systems affected by arsenic
regulation.
3.3.2 Amortization of Costs
In the arsenic benefits analysis capital costs are amortized (expressed as annual
equivalent flows) by using a discount rate of 7%. An alternative calculation based on a 3% rate
is also presented. However, what matters for the impact on utility finances and utility customers
is the actual interest rate at which the affected utilities will finance these investments. We
19
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
recommend that the Agency estimate this when calculating the regulatory costs (Freeman,
Measurement..., 1993, pp. 213-216; Kolb and Schemata, JPAM, 1990).
Exhibit 6-7 of the arsenic economic analysis presents data showing recommended cost of
capital estimates for various types of water utility ranging from 4.17% to 5.94%. Having
reviewed the report from which they derive, we do not believe these estimates are adequate.
First, while the analysis allows for the use of different sources of capital by non-small utilities of
different sizes (those serving 10,001 - 50,000 and those serving over 50,000) it assumes that the
costs of various types of capital - long-term debt, short-term debt, equity capital, municipal
bonds - are the same regardless of size for all systems serving over 10,000. We do not believe
this assumption is likely to be accurate. Second, with investor owned utilities the report states
that an after-tax figure is appropriate for the required analysis. We disagree and instead
recommend (1) using a before-tax figure for the cost of capital for investor owned utilities, and
(2) using a separate account to track the revenue gains to the government sector from taxes from
the water system debt.
By way of illustration, suppose an investor owned water utility and a public owned water
utility both need to borrow $1 million. Suppose the investor owned utility issues bonds with an
interest rate of 8.5%. The publicly owned utility can borrow at a lower interest rate since the
interest paid on its bonds is tax exempt; it can borrow at 5.19%, to use the figure from page 29 of
the report on Public Water System Cost of Capital. The difference of 3.31% (= 8.5 - 5.1) is the
savings due to the tax exemption on publicly owned system debt. The report recommends using
5.19% as the cost of capital for investor owned utility debt as well as publicly owned utility debt,
because it views the 3.31% interest increment as merely a transfer payment. While this is not
incorrect, it is misleading with respect to the policy implications. Because the investor owned
utility pays a higher interest rate for its debt than the publicly owned utility, its customers will
face a larger cost increase than those of the publicly owned utility. We believe this should be
made explicit in the analysis.
Third, for similar reasons we disagree with the way in which the report treats the
financing of capital costs on a pay-as-you-go basis out of current revenues or accumulated
capital reserves. This type of financing accounts for about 20-30% of cost of capital expenditures
for non-small systems, and 20-60% for small systems. The report imputes an opportunity cost of
capital to funds from this source. For example, if a small system needs to fund $1 million of
water supply improvement from cash flow, the report recommends amortizing this as though the
funds were being borrowed with unrated or low rated general obligation bonds at an interest rate
of 5.47%. Suppose the investment were being made over a 5-year period. If the utility had made
no provision for a sinking fund, it would need to raise the $1 million from higher water rates
over the 5-year period. To the extent there is a sinking fund, the impact on water rates will be
less severe. It is clear, however, that using an imputed cost of capital may not give an accurate
assessment of the short-term impact on water rates when financing water system investments
from cash flow.
20
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
REFERENCES
Axelson O, Dahlgren E, Jansson C-D, Rehnlund SO. Arsenic exposure and mortality: a case-
referent study from a Swedish copper smelter. Brit J Indus Med 1978;35:8-15.
Breslow and Day, 1987
Enterline and Henderson, 1973
Chen C-J, Chiou H-Y, Chiang M-H, Lin L-J, Tai T-Y. Dose response relationship between
ischemic heart disease mortality and long-term arsenic exposure. Arterioscler thromb Vase Biol
1996;16:504-510
Chen C, Hsueh, YM, Lai, MS, Shyu, MP, Chen, SY, Wu, MM, Kuo, TL, Tai, TY. Increased
prevalence of hypertension and long-term arsenic exposure. Hypertension 25:53-60(1995).
Chen C-J, Wu M-M, Lee S-S, Wang J-D, Cheng S-H,Wu H-Y. Atherogenicity and
carcinogenicity of high arsenic Artesian well water. Arteriosclerosis 1988;8:452-460.
Chiou H, Wei-I, H, Che-Long, S, Shu-Feng, C, Yi-Hsiang, H, Chi en-Jen ,C. Dose-response
relationship between prevalence of cerebrovascular disease and ingested inorganic arsenic.
Stroke 28:1717-23(1997).
Concha G, Vogler G, Lezcano D, Nermell B, Vahter M. Exposure to inorganic arsenic
metabolites during early human development, lexicological Sci 1998;44:185-190.
Engel RR, Smith AH. Arsenic in drinking water and mortality from vascular disease: an
ecologic analysis in 30 counties in the United States. Arch Environ Health 1994;49:418-427.
Engel RR, Hopenhayn-Rich, Receveur O, Smith AH. Vascular effects of chronic arsenic
exposure: a review. Epidemiologic Rev 1994;16:184-209.
Freeman, AM 1993 (cited in section 3.3.2)
Gomez-Caminero A. Cardiovascular Effects of Arsenic During Pregnancy. Doctoral
Dissertation, University of North Carolina, Chapel Hill, 2001.
Hertz-Picciotto I, Smith AH, Holtzman D, Lipsett M, Alexeeff G (1992) Synergism between
occupational arsenic exposure and smoking in lung cancer induction. Epidemiology 3:23-31.
Hertz-Picciotto I, Arrighi HM, Hu S-W (2000). Does arsenic exposure increase the risk for
circulatory disease? Amer J Epidemiol 151:174-181
R-l
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
Hopenhayn-Rich C, Browning SR, Hertz-Picciotto I, et al. Chronic arsenic exposure and risk of
infant mortality in two areas of Chile. Environ Health Persp 2000; 108:667-673.
Hsueh Y-M, Wu W-L, Huang Y-L, Chiou H-Y, Tseng C-H, Chen C-J. Low serum carotene
level and increased risk of ischemic heart disease related to long-term arsenic exposure.
Atherosclerosis 1998; 141:249-257.
Jensen GE, Hansen ML. Occupational arsenic exposure and glycosylated haemoglobin. Analyst
1998;123:77-80.
Kolb and Schemata (cited in section 3.3.2)
Lagerkvist BEA, Linderholm H, Nordberg GF. Arsenic and Raynaud's phenomenon. Int Arch
Occup Environ Health 1988;60:361-364
Lagerkvist B, Linderholm H, Nordberg GF. Vasospastic tendency and Raynaud's phenomenon
in smleter workers exposed to arsenic. Environ Res 1986;39:465-474.
Lai M-S, Hsueh Y-M, Chen C-J, et al. Ingested inorganic arsenic and prevalence of diabetes
mellitus. Am JEpidemiol 139:484-492(1994).
Lewis, et al, 1999
Magat, Viscusi and Huber 1996
Malmfors 19XX
Morales, et al., 2000
NDWAC. Benefits Workgroup Recommendations, October, 1998.
Peto 1995
Rahman M, Tondel M, Chowduhury JA, Axelson O. Relations between exposure to arsenic,
skin lesions, and glucosuria. Occup Environ Med 1999;56:277-281.
Rahman M, Tondel, M, Ahmad, A, Chowdhury, I, Faruquee, M, Axelson, O. Hypertension and
arsenic exposure in Bangladesh. Hypertension 33:74-78(1999).
Regas, Diane. Request for review of the benefits assessment for the arsenic in drinking water
regulation. Memorandum from the Acting Assistant Administrator for Water, US EPA, June 8,
2001.
SAB. An SAB Report on EPA's White Paper Valuing the Benefits of Fatal Cancer Risk
Reductions, EPA-SAB-EEAC-00-013, July 2000.
R-2
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
Thomas, 1983
Thomas, 1987
Thomas, 1988
Tollerud 1999
Tsai S-M, Wang T-N, Ko Y-C. Mortality for certain diseases in areas with high levels of arsenic
in drinking water. Arch Environ Health 1999; 54:186-193.
Tseng C-H, Chong C-K, Chen C-J, Tai T-Y. Dose-response relationship between peripheral
vascular disease and ingested inorganic arsenic among residents in blackfoot disease endemic
villages in Taiwan. Atherosclerosis 1996; 120:125-33.
Tseng C-H, Tai T-Y, Chong C-K, et al. Long-term arsenic exposure and incidence of non-
insulin-dependent diabetes mellitus: a cohort study in arseniasis-hyperendemic villages in
Taiwan. Environ Health Persp 2000; 108:846-851.
USEPA. Arsenic in Drinking Water Rule Economic Analysis. EPA 815-R-00-026, December
2000.
USEPA. The Benefits and Costs of the Clean Air Act, 1990-2010 (USEPA 1999)
USEPA. Cost of Illness Handbook.
USEPA. Guidelilnes for Preparing Economic Analyses. EPA 240-R-00-003, September 2000a.
USEPA. National Primary Drinking Water Regulations; Arsenic and Clarifications to
Compliance and New Source contaminants Monitoring: Final Rule. EPA-815-Z-01, Federal
Register, Vol. 66(14)6976-7066. January 22, 2001
Viscusi, K., et al.
Wu M-M, Kuo T-L, Hwang Y-H, Chen C-J. Dose-response relation between arsenic
concentration in well water and mortality from cancers and vascular diseases. Amer J Epidemiol
1989;130:1123-1132.
R-3
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
APPENDICES
APPENDIX 1 - BACKGROUND
Appendix 1.1 SAFE DRINKING WATER ACT PROVISIONS
SDWA Requirements for Setting the Standard
The Safe Drinking Water Act requires EPA to establish a Maximum Contaminant Level
Goal (MCLG) and to promulgate a National Primary Drinking Water Regulation (NPDWR) if
the Administrator determines that: i) the contaminant may have an adverse effect on the health
of persons; ii) the contaminant is know to occur or there is substantial likelihood that the
contaminant will occur with a frequency and at levels of public health concern; and in the sole
judgment of the Administrator, regulation of such contaminant presents a meaningful
opportunity for health risk reduction for persons served by public water systems.
The MCLG is to be set at the level at which no known or anticipated adverse effects on
the health of persons occur and which allows an adequate margin of safety. Further, the
regulation for a contaminant with an MCLG shall specify a Maximum Contaminant Level
(MCL) which is as close to the MCLG as is feasible. If it is not economically or technically
feasible to measure the contaminant, a treatment technique can be promulgated in lieu of an
MCL.
SDWA further defines feasible to mean with the use of best technology, treatment
techniques, and other means are available taking cost into consideration. And when the
Administrator proposes a NPDWR she must also publish a determination as to whether the
benefits of the MCL justify, or do not justify, the costs. Among other factors, this determination
is to be based on the analysis an analysis of each of the following: i) quantifiable and
nonquantifiable health risk reduction benefits for which there is a factual basis in the rulemaking
record to conclude that such benefits are likely to occur as the result of treatment to comply with
each level for the contaminant; ii) quantifiable and nonquantifiable costs for which there is a
factual basis in the rulemaking record to conclude that such costs are likely to occur as a result of
compliance with the MCL; iii) the incremental costs and benefits associated with each
alternative MCL; iv) effects of the contaminant on the general population and groups within the
population that are identified as likely to be at greater risk of adverse health effects due to
exposure to contaminants in drinking water than the general population.
And v) other relevant factors, including quality/extent of information, uncertainties in the
analyses above, and factors with respect to the degree and nature of the risk.
The Administrator is explicitly given the authority to establish a MCL at a level other
than the feasible level, if the technology, treatment techniques, and other means used to
determine the feasible level would result in an increase in the health risk from drinking water by
increasing the concentration of other contaminants or interfering with the efficacy of techniques
used to comply with other NPDWRs or if she determines as above that the benefits of a MCL
would not justify the costs of complying with the level. In that case, the Administrator may
A-l
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
promulgate a MCL that maximizes health risk reduction benefits at a cost that is justified by the
benefits.
A-2
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
Appendix 1.2 NDWAC Benefits Workgroup Recommendations, October, 1998
The National Drinking Water Advisory Council (NDWAC) was charged with providing
EPA with recommendations on which benefits should be routinely considered in developing its
regulations. They were to address what categories of benefits should be considered, how to
consider qualitative benefits, and how to compare the results of benefits assessments with cost
analyses. NDWAC adopted the following recommendations from the Working Group:
Recommendation 1: EPA should focus its benefits analysis efforts primarily on
assessing effects on human health, defining these effects as clearly as possible and using
the best available data to value them. It is also recommended that EPA consider 1) health
risk reductions, 2) taste and odor improvements, 3) reduction in water system materials
damage, 4) commercial water treatment cost reductions, 5) benefits due to source water
protection, and 6) benefits derived from the provision of information on drinking water
quality.
Recommendation 2: EPA should devote substantial efforts to better understanding the
health effects of drinking water contaminants, including the types of effects, their
severity and affected sensitive subpopulations. Better information is also needed on
exposures and the effects of different exposure levels, particularly for contaminants with
threshold effects. These efforts should pay particular attention to obtaining improved
information concerning impacts on children and other sensitive populations.
Recommendation 3: EPA should clearly identify and describe the uncertainties in the
benefits and costs analysis, including descriptions of factors that may lead the analysis to
significantly understate or overstate total benefits and costs. Factors that may have
significant but indeterminate effects on the benefits and costs estimates should also be
described.
Recommendation 4: EPA should consider both quantified and non-quantified benefits in
regulatory decision making. The information about quantified and non-quantified
(qualitative) benefits should be presented together in a format, such as a table, to ensure
that decision-makers consider both kinds of information.
Recommendation 5: EPA should consider incremental benefits and costs, total benefits
and costs, the distribution of benefits and costs, and cost-effectiveness in regulatory
decision-making. This information should be presented together in a format, such as a
table, to ensure its consideration by decision-makers.
Recommendation 6: Whenever EPA considers regulation of a drinking water
contaminant, it should evaluate and consider, along with water treatment requirements to
remove a contaminant, source water protection options to prevent such contaminant from
occurring. The full range of benefits of those options should be considered.
A-3
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
APPENDIX 2
Appendix 2.1 Supplemental Information to Charge Question 1
Estimates of latency can be approached by developing classical Armitage-Doll multi-
stage models of the morbidity and mortality from various cancers in the U.S. population and then
exploring mathematically the expected distributions of times to diagnosis and death from various
cancers, making various plausible assumptions about where arsenic might act in the sequence of
genetic changes leading to the different cancers. Recent (1994-98) U.S. morbidity and mortality
data for different cancers are available from the "SEER" program [Ries, L. A. G., Eisner, M. P.,
Kosary, C. L. Hankey, B. F., Miller, B. A., Clegg, L., and Edwards, B. K. (2001) SEER Cancer
Statistics Review, 1973-1998, National Cancer Institute, Bethesda, Md.].
The most straightforward approach to specifying the models is to do a simple set of
weighted regression analyses to these data of the form:
Log(Incidence or Mortality Rate in cases/100,000 population per year) = k*Log(Age - L) + b
In this equation, L is a lag period that represents the typical time between the unobserved birth of
the first cancer cell and either cancer diagnosis or cancer death (for morbidity v. mortality data,
respectively), and k + 1 is the number of "stages" (sequential genetic changes) in the cancer
model. Some fits derived from the data from Taiwan are contained in Attachment A. The "U.S.
incidence data" worksheet contains SEER incidence and mortality data for lung and bladder
cancer for each sex, but the model fitting has not yet been done. The "5-stage male smoker"
worksheet shows an example of a 5-stage lung cancer model created several years ago to
represent the expected time pattern of development of lung cancer in smokers who began
smoking at age 13. [See Hattis, D., and Silver, K. "Use of Mechanistic Data in Occupational
Health Risk Assessment—The Example of Diesel Particulates," in Chemical Risk Assessment
and Occupational Health-Current Applications. Limitations, and Future Prospects. C. Mark
Smith, David C. Christiani, and Karl T. Kelsey, eds., Greenwood Publishing Group, Inc.,
Westport CT, 1994, pp. 167-177 for an example of prior use of this approach]
Such a model makes it straightforward to explore the implications of different
assumptions about which stages are affected by arsenic exposures. Additional data available in
the literature may help judge the relative likelihood of different stage-of-action assumptions. In
addition to the Chen et al. (1991) paper cited above, the following by Tsai et al. (1998) might be
useful in estimating the rates at which risks for various health effects might decrease when
exposure is decreased [Tsai, SM, Wang, TN, and Ko, YC Cancer mortality trends in a blackfoot
disease endemic community of Taiwan following water source replacement. J. Toxicol Environ.
Health 55(6):389-404 1998]. It is important that the estimate the latent benefits from lowering
exposure to individuals that have had prior arsenic exposure be estimated utilizing the same
model utilized to estimate potency. Mode of action has implications for how rapidly and
completely the effects in the exposed population are reversed as it does when exposure increases
to increase the risk of cancer. Thus, it is important to be consistent in the utilization of mode of
action information in the final treatment of risks.
A-4
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
As indicated above, in the ideal circumstance there needs to be some consideration or at
least acknowledgment of the different ages at the time the rule is put into effect. Benefits will
accrue over a lifetime for children conceived after treatment is instituted. However, at that
moment there will be people of different ages who will gain some benefit. Benefits to these
individuals could be significantly larger if arsenic were largely a late stage carcinogen. This
appears to be the basis of the reduction in lifetime risks associated with discontinuation of
smoking even after several years. Arsenic produces a variety of effects at the molecular and
cellular level that can contribute to cancer risk. It is probable that there will be insufficient data
to come to hard conclusions about how different modes of action are contributing to the cancer
incidence at different doses or dose rates. Because the experimental data (i.e. mechanistic data)
that is available today indicate the possibility of several distinctly different modes of action with
different metabolic forms of arsenic at different doses such an exercise will be viewed as being
highly speculative by scientists. Thus, unless more certainty can be brought to the analysis than
was apparent in the Panel's brief review of the literature, it is suggested that such analyses be
confined to the uncertainty analysis as it has the distinct possibility of confusing the more
straightforward derivation of latency information from existing data. It is strongly suggested
that the sophistication of the methodology applied be limited by and consistent with
recommendations of the National Research Council (NRC) panel, which has been charged with
making recommendations on the risk assessment methodology that should be used.
A-5
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
APPENDIX 2.2
Supplement to Charge Question 2
Studies addressing the major categories of concern at lower exposure levels are listed in the tables (which are not comprehensive, but
rather, representative). These studies demonstrate a broad array of related endpoints and indicate the range and weight of evidence,
qualitatively, as well as the consistency with which these effects are related to arsenic exposure. Such consistency, particularly when
at least some of the studies are of high quality and have adjusted for individual-level confounders, strengthens the evidence for
causality.
A-6
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
I. Human morbidity studies of cardiovascular endpoints
Outcome
Cerebrovas
cular
disease/cere
bral
infarction
Ischemic
heart
disease
Electrocard
io-graphic
abnormaliti
es
Hypertensi
on
ii
Systolic
blood
pressure
Vasospastic
tendency
(finger
systolic
pressure,
Authors/year &
location
Chiou et al 1997
Taiwan
Hsueh et al 1998
Taiwan
Ohnishi et al 2000
Japan
Chen et al 1995
Taiwan
Rahman et al 1999
Bangladesh
Jensen & Hansen
1998
Denmark
Lagerkvist et al 1986
Sweden
Design
Retrospective
cohort
Retrospective
cohort
Prospective,
patients with
promyelocytic
leukemia
Retrospective
cohort
Retrospective
cohort
Retrospective
cohort
X-sectional
Exposure assessment
Cumulative exposure
Avg concentr'n in H,0
Duration of exposure via
H20
As Tx for promyelocytic
leukemia
Cumulative exposure
[Avg cone in B,0]*
Cumulative exposure
Avg concentr'n in H,0
Job with arsenic
exposure, urinary As
Urinary As available but
not used-
Estimated exposure at
300 ug/day, or 4 g over
23 years
Dose-response analysis:
Significant; adjusted for
age, sex, cigarettes,
alcohol
Significant, adjusted for
total cholesterol, BMI,
hypertension, serum oc-
and (3-carotene
Prolonged QT intervals in
all 8 patients, serious
arrhythmias in 4
Significant; adjusted for
age, sex, diabetes,
proteinuria, BMI
Significant; adjusted for
age, sex, BMI
No dose-response analysis
conducted
Measure of
association
Odds ratio
Odds ratio
Odds ratio
Prevalence
ratio
Difference in
means
Difference in
prevalence
Range of exposures
<0. 1,0.1 -4.9, >5.0mg/L-
year;
<0. 1,0.1-50, 50.1-2999.9,
>300 ug/L
<13, 13-29, >30 years
drinking artesian well
water
1 5 mg/kg for 20-79 days
0,0.1-6.3,6.4-10.8, 10.9-
14.7 mg/L-years;
0, .01-.70, >.70mg/L
0, <1.0, 1.0-5.0, >5. 0-10.0
mg/L-years;
<0.5, 0.5 to 1.0, >1.0mg/L
Mean of 22.3 nmol/mmol
As in creatinine vs. 12.0
nmol/mmol for referents
10-340 ug/L (mean=70) in
urine among exposed; 5-20
ug/L among referents,
highest quartile had mean
of 180 ug/L
A-7
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
upon
cooling)
A-8
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
I. Human morbidity studies of cardiovascular endpoints (con't)
Outcome
Blackfoot
disease**
Peripheral
vascular
disease***
Raynaud
phenomenon,
numbness &
other symptoms
von Willebrand
factor
Authors/year &
location
Chen et al 1988
Taiwan
Tseng et al 1996
Taiwan
Lagerkvist et al
1988 Sweden
Gomez-
Caminero 2001
Chile
Design
Retrospective
cohort
Retrospective
cohort
Time trend -
start to end of
vacation
Prospective
cohort of
pregnant
women
Exposure assessment
Duration of exposure
viaH.0
Cumulative exposure
Duration well water use
Duration living in Bf
area
Exposed vs. unexposed
town
Dose-response analysis:
Significant in highest
exposure group, adjusted
for age, sex, BMI,
cigarette smoking,
diabetes hypertension,
serum total cholesterol, &
trislycerides
No dose-response analysis
conducted. Significant
difference in numbness &
other signs,
Significant vs. referents
Measure of
association
Odds ratio
Difference in
prevalence
Difference in
means, odds
ratio for
lowest tertile
Range of exposures
0 (referent) 1-29, >30 years
drinking artesian well
water
0 (referent), 0.1-19.9, >20
mg/L-years
0, 1-19, 20-29, >30 years
drinking artesian well
water
Exposed: mean of 61 ug/L
urine
<2 ug/L (referent),
~45 ug/L (exposed)
* The analysis for this exposure metric did not adjust for all factors in the next column
** Blackfoot disease has been used as an indicator of exposure to arsenic &/or susceptibility to the effects of arsenic, due to its close
association with elevated arsenic exposures.
***Diagnosed by Doppler ultrasound, ABI<0.9 on either side of extremity
A-9
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
II. Human mortality studies of cardiovascular & renal endpoints
Outcome
Circulatory
disease
it
Cardiovascular
disease
it
<.<.
Ischemic heart
Disease
(.(.
Hypertensive
heart disease
Authors/year &
location
Tsai et al 1999
Taiwan
Hertz-Picciotto et
al, 2000
US smelter
workers
Wu et al 1989
Taiwan
Axelson et al
1978
Sweden, area
around smelter
Hertz-Picciotto et
al, 2000
US smelter
workers
Chen et al 1996
Taiwan
Tsai et al 1999
Taiwan
Lewis et al 1999
Utah, USA
Design
Retrospective
cohort, 1971-
1994
Retrospective
cohort
Retrospective
cohort 1973-
1986
Case-control
Retrospective
cohort
Two
prospective
cohorts, 1985-
1993, and 1988-
1995
Retrospective
cohort, 1971-
1994
Retrospective
cohort
Exposure assessment
Townships with arsenic
contaminated water
from 1900's to mid-
1970's
Cumulative
occupational exposure
over the worklife
Villages with arsenic
contaminated water
Employment in exposed
jobs
Cumulative
occupational exposure
over the worklife
Avg concentr'n in HjO
Cumulative exposure
Townships with arsenic
contaminated water
from 1900's to mid-
1970's
Cumulative exposure.
Means in towns ranged
Dose-response analysis:
Significant in both sexes,
adjusted for age, calendar
year
Significant dose response
adjusted for age, year of
hire, and the healthy
worker survivor effect
Significant, adjusted for
age, sex
Significant dose response
Significant dose response
adjusted for age, year of
hire, and the healthy
worker survivor effect
Monotonic dose response,
models adjusted for age,
sex, baseline BMI,
cigarette smoking, serum
cholesterol, triglycerides,
diabetes, hypertension,
blackfoot disease*
Significant in both sexes,
adjusted for age, calendar
year
Significant excess in men
and women
Measure of
association
Standardized
mortality
ratio
Rate ratio
Mortality
ratio
Mantel-
Haenszel
rate ratio
Rate ratio
Hazard ratio
from Cox
proportional
hazards
model
Standardized
mortality
ratio
Standardized
mortality
Range of exposures
0.78 ppm,
referents: local county, and
national rates
<750 (referent), 750-1999,
2000-3999, 4000-7999,
8000-19,999, >20,000
ug/m3 -years
0.3,0.3-0.59, >.60 ppm
Not employed at smelter
(referent), employed at
smelter: 'close to' 0.5
mg/m3
<750 (referent), 750-1999,
2000-3999, 4000-7999,
8000-19,999, >20,000
ug/m3 -years
0 (referent), 0.1-9.9, 10.0-
19.9, 20.0+ mg/L years
0.78 ppm,
referents: local county, and
national rates
<1, 1-4.999, >5.0ppm-
years,
A-10
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
I
I
jfrom 18.1-164.4 ppb I
[ratio
I range
II. Human mortality studies of cardiovascular & renal endpoints (con't)
Outcome
Cerebrovascular
disease
it
Peripheral
vascular disease
<.<.
it
Pulmonary heart
disease
**
Nephritis,
nephrosis
it
Authors/year &
location
Wu et al 1989
Taiwan
Tsai et al 1999
Taiwan
Wu et al 1989
Taiwan
Tsai et al 1999
Taiwan
Engel & Smith
1994
USA
Tsai et al 1999
Taiwan
Engel et al 1994
Tsai et al 1999
Taiwan
Lewis et al 1999
Utah, USA
Design
Retrospective
cohort 1973-
1986
Retrospective
cohort, 1971-
1994
Retrospective
cohort 1973-
1986
Retrospective
cohort, 1971-
1994
Ecologic study
at the county
level
Retrospective
cohort, 1971-
1994
Retrospective
cohort, 1971-
1994
Retrospective
cohort
Exposure assessment
Villages with arsenic
contaminated water
Townships with arsenic
contaminated water
from 1900's to mid-
1970's
Concentr'n in H in
villages with arsenic
contaminated water
Townships with arsenic
contaminated water
from 1900's to mid-
1970's
Avg concentr'n in Hfl
Townships with arsenic
contaminated water
from 1900's to mid-
1970's
Townships with arsenic
contaminated water
from 1900's to mid-
1970's
Cumulative exposure.
Means in towns ranged
from 18. 1-164.4 ppb
Dose-response analysis:
Significant, adjusted for
age, sex
Significant in both sexes,
adjusted for age, calendar
year
Significant, adjusted for
age, sex
No dose measure used,
adjusted for age, sex,
calendar year
No clear monotonic dose
response, but elevated risk
at each level >5 ug/L
No dose measure used,
adjusted for age, sex,
calendar year
No dose measure used,
adjusted for age, sex,
calendar year
Significant excess in men
and women
Measure of
association
Mortality
ratio
Standardized
mortality
ratio
Mortality
ratio
Standardized
mortality
ratio
Standardized
mortality
ratio
Standardized
mortality
ratio
Standardized
mortality
ratio
Standardized
mortality
ratio
Range of exposures
<0.3,0.3-0.59, >.60 ppm
0.78 ppm,
referents: local county, and
national rates
<0.3, 0.3-0.59, >.60 ppm
0.78 ppm,
referents: local county, and
national rates
5-10, 10-20, >20 ug/L
0.78 ppm,
referents: local county, and
national rates
0.78 ppm,
referents: local county, and
national rates
<1, 1-4.999, >5.0ppm-
years,
range
A-ll
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
*Adjustment for Blackfoot disease attenuated but did not eliminate the association of arsenic exposure with ISHD
**For further mortality and morbidity studies of cardiovascular endpoints, see Table 6, Engel et al., 1994.
A-12
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
III. Animal morbidity studies of cardiovascular endpoints
Outcome
Authors/year
Design
Exposure assessment
Dose-response analysis
adjusted for:
Measure of
association
Exposure level
Animal Studies
Vasoreactivity
Vasoreactivity
Potentiation of
P-
adrenoreceptor
stimulation
Stroke
volume,
cardiac output
Vasoreactivity
*
Bekemeir &
Hirschelmann
1989
Carmignano et al
1983
ii
Carmignano et al
1985
Experiment
Experiment
Experiment
Not applicable -
controlled dosing
tt
ct
ct
tt
Only one dose group
Only one dose group
Only one dose group
Only one dose group
Only one dose group
15 mg/kg, orally
50 ug/mL drinking water
50 ug/mL drinking water
* after administration of isoprenaline, clonidine, tyramine, etc.
A-13
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
IV. Human mortality and morbidity studies of endocrinologic/metabolic conditions and biomarkers
Outcome
Diabetes mellitus
mortality
Diabetes mellitus
incidence
it
<.<.
Glycosylated
hemoglobin
it
Glucosuria
Hepatic function:
bilirubin
excretion, ALP
activity
Authors/year &
location
Tsai et al 1999
Taiwan
Lai et al 1994
Taiwan
Rahman et al
1996
Sweden
Tseng et al
2000
Taiwan
Jensen & Hansen
1998
Denmark
Gomez-
Caminero 2001
Chile
Rahman et al
1999
Bangladesh
Hernandez-
Zavala et al 1998
Mexico
Design
Retrospective
cohort, 1971-
1994
Retrospective
cohort
Retrospective
cohort
Prospective
cohort, ~2.5
years follow-up
Retrospective
cohort
Prospective
cohort of
pregnant
women
Retrospective
cohort
Retrospective
cohort
Exposure assessment
Townships with arsenic
contaminated water
from 1900's to mid-
1970's
Cumulative exposure
Duration well water
use*
Job in glassworks with
likely exposure
Cumulative exposure
from H,0
Jobs with arsenic
exposure (taxidermists,
construction workers,
wood & electric pylon
impregnators
Exposed vs. unexposed
town
Avg concentr'n in Hfl
Cumulative exposure
Mean water
concentration in each of
three towns
Dose-response analysis:
No dose measure used,
adjusted for age, sex,
calendar year
Significant, adjusted for
age, sex, BMI, physical
activity
Significant in those with
highest exposure, adjusted
for age
Significant, adjusted for
age, sex, BMI
Significant vs. referents
Significant vs. referents
Significant, adjusted for
age and sex, using
cumulative exposure
Significant differences,
adjusted for age, alcohol,
tobacco, pesticides
Measure of
association
Standardized
mortality
ratio
Odds ratio
Odds ratio
Hazard ratio
from Cox
model
Difference in
medians
Difference in
means, odds
ratio for
>6.5%
Prevalence
ratio
Difference in
means
Range of exposures
0.78 ppm,
referents: local county, and
national rates
0 (referent), 0.1-15.0,
>15.1 ppm-yrs;
0 (referent, 1-10, 11-20,
>21 years drinking artesian
well water
No quantitation available
<17 mg/L years (referent),
>17 mg/L years
6-44 nmol/mmol urinary
As in creatinine (referents);
12-295 nmol/mmol
(exposed)
<2 ug/L (referent),
~45 ug/L (exposed)
<0.5,0.5-1.0, >1.0 mg/L;
<1.0, 1.0-5.0, >5. 0-10.0,
>10.0 mg/L-years
Means: 14.0 ug/L
(referent),
116 ug/L and 239 ug/L in
two exposed towns
; for all factors in the next column
A-14
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
V. Human studies of cancers other than lung and bladder
Outcome
Kidney cancer
Liver cancer
Prostate cancer
it
Stomach cancer*
Colon cancer*
Rectum cancer*
Liver cancer*
Nasal cancer*
Laryngeal ca*
Skin cancer*
Bone cancer*
Lymphoma*
Authors/year &
location
Smith et al 1992
Taiwan
"-»
Tsai et al 1999
Taiwan
Lewis et al 1999
Utah, USA
Tsai et al 1999
Taiwan
"-»
"-»
"-»
"-»
"-»
"-»
"-»
"-»
Design
Retrospective
cohort
"-»
Retrospective
cohort 1971-
1994
Retrospective
cohort
Retrospective
cohort 1971-
1994
"-»
"-»
"-»
"-»
"-»
"-»
"-»
"-»
Exposure assessment
Cumulative exposure in
H,0
"-»
Townships with arsenic
contaminated water
from 1900's to mid-
1970's
Cumulative exposure.
Means in towns ranged
from 18. 1-164.4 ppb
Townships with arsenic
contaminated water
from 1900's to mid-
1970's
"-»
"-»
"-»
"-»
"->
"^
"->
"->
Dose-response analysis:
Significant, adjusted for
age, sex
"-»
Adjusted for age, sex,
calendar year
Significant excess
Adjusted for age, sex,
calendar year
"-»
"^
"^
"-»
"->
"^
"->
"->
Measure of
association
Rate ratio
"-»
Standardized
mortality
ratio
Standardized
mortality
ratio
Standardized
mortality
ratio
"-»
"^
"^
"->
"->
"^
"->
"->
Range of exposures
<.<.
0.78 ppm,
referents: local county, and
national rates
<1, 1-4.999, >5.0ppm-
years,
ranse
0.78 ppm,
referents: local county, and
national rates
(.(.
<.<.
<.<.
tt
tt
(.(.
tt
tt
*Excess observed in both genders. Cancers found in excess in only one gender not included.
A-15
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
VI. Human morbidity & mortality studies of non-malignant respiratory endpoints
Outcome
Respiratory
effects: cough,
shortness of
breath
Bronchitis
Chronic airways
obstruction
Emphysema
Authors/year &
location
Mazumder et al
2000
West Bengal,
India
Tsai et al 1999
Taiwan
Engel & Smith
1994
USA
"-»
Design
X-sectional
Retrospective
cohort 1971-
1994
Ecologic study
at county level
"-»
Exposure assessment
Current concentration
measured in well water
Townships with arsenic
contaminated water
from 1900's to mid-
1970's
Avg concentr'n in f^O
"-»
Dose-response analysis:
Significant, adjusted for
age & sex, smokers
excluded
Adjusted for age, sex,
calendar year
Adjusted for age, sex, and
calendar year
"-»
Measure of
association
Prevalence
odds ratio
Standardized
mortality
ratio
Standardized
mortality
ratio
"-»
Range of exposures
<50, 50-199, 200-499,
500-799, >800 ug/L
0.78 ppm,
referents: local county, and
national rates
5-10, 10-20, >20 ug/L
it
A-16
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
VII. Human reproductive studies
Outcome
Spontaneous
abortion
it
<.<.
it
<.<.
Stillbirth
it
it
Preterm birth
Authors/year &
location
Nordstrom et al
1978
Sweden
Nordstrom et al
1979
Sweden
Borzsonyi et al
1992
Hungary
Ahmad et al
2001
Bangladesh
Aschengrau et al
1989
Massachusetts
"-»
Borzsonyi et al
1992
Hungary
Hopenhayn-Rich
et al 2000
Chile
Ahmad et al
2001
Bangladesh
Design
Retrospective
cohort of
pregnancies
Retrospective
cohort of
pregnancies
Retrospective
cohort
Retrospective
cohort of
pregnancies
Case-control
"-»
Retrospective
cohort
Retrospective
vital statistics
Retrospective
cohort of
pregnancies
Exposure assessment
Residential proximity to
a smelter
Employment in smelter
prior to or during
pregnancy
Concentration in HjO
Concentration in HjO
Duration of residence in
high arsenic area
Concentration in HjO
"-»
Concentration in HjO
Concentration in HjO
Comparison of two
communities
Concentration in HjO
Duration of residence in
high arsenic area
Dose-response analysis:
Trend in frequency by
distance of region to
smelter
Highest prevalence among
those living near the
smelter during or after
their employment
Significant difference
comparing high vs. low
arsenic region
Significant difference
comparing high vs. low
arsenic region, and for
those with longer duration
Trend in risk
"-»
Significant difference
comparing high vs. low
arsenic region
Significant difference
during period when
exposures were very high
Significant difference
comparing high vs. low
arsenic region, and for
those with longer duration
Measure of
association
Prevalence
ratio
Prevalence
ratio
Prevalence
rate
difference
Prevalence
rate
difference
Odds ratio
"-»
Prevalence
rate
difference
Mortality
rate
difference
and ratio
Prevalence
rate
difference
Range of exposures
No quantitation
it
Low (not quantitated
referent), 170-330 ug/L
<20 (referent), >100 ug/L
0.8,0.8-1.3, 1.4-1.9 ug/L
it
Low (not quantitated
referent), 170-330 ug/L
<5 (referent), various
levels to >800 ug/L
<20 (referent), >100 ug/L
A-17
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
VII. Human reproductive studies (con't)
Outcome
Birthweight
Low birthweight
Congenital
malformations
Coarctation of
the aorta
Neonatal
mortality
Postneonatal
mortality
Authors/year &
location
Nordstrom et al
1978
Sweden
Hopenhayn et al
2001
Chile
Nordstrom et al
1979
Sweden
Zierler et al
1988
Massachusetts
Hopenhayn-Rich
et al 2000
Chile
"-»
Design
Retrospective
cohort of
pregnancies
Prospective
cohort & review
of vital statistics
Retrospective
cohort of
pregnancies
Case-control
Retrospective
vital statistics
"-»
Exposure assessment
Residential proximity to
smelter or employment
Concentration in HjO
Comparison of two
communities
Employment in the
smelter
Routine monitoring of
water
Concentration in t^O
Comparison of two
communities
"-»
Dose-response analysis:
Lowest birthweight
among those living
nearest the smelter
Significantly increased
risk of low birth weight
Higher prevalence of
congenital malformations
among employed mothers
Above vs. below the limit
of detection, three-fold
increased risk, adjusted
for seven other
contaminants, source of
water, maternal education
Significant difference
during period when
exposures were very high
"-»
Measure of
association
Difference in
birthweight
Odds ratio
for low
birthweight
Prevalence
ratio
Odds ratio
Mortality
rate
difference
and ratio
"-»
Range of exposures
No quantitation
<2 (referent), 40-50 ug/L
<.<.
< limit of detection (0.8
ug/L), >limit of detection
<5 (referent), various
levels to >800 ug/L
it
A-18
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
VIII. Human studies of neurologic and neurodevelopmental endpoints
Outcome
Peripheral
neuropathy
Various
neurobehavioral
parameters*
Verbal IQ
Authors/year &
location
Gerr et al 2000
Georgia, USA
"-»
Calderon et al
2001
Mexico
Design
Cross-sectional
"-»
Cross-sectional
Exposure assessment
Dust & soil arsenic
measurements
"-»
Urinary arsenic
Dose-response analysis:
Significant trend, adjusted
for age, education, sex,
verbal intellectual score,
alcohol
"-»
Significant inverse
correlation
Measure of
association
Odds ratio
Linear
regression
Partial
correlation
coefficient
Range of exposures
House dust: 1-1200 ug/g
Window sill dust: 0.5-192
Attic dust 1.2-2635 ug/g
Soil 2.0-1845 ug/g
it
<50, 50-100, >100 ug As/g
creatinine;
Range: 27.5-186.2 ug/g
creatinine
*Vibrotactile threshold, standing steadiness, tremor intensity
A Public Health Based Approach to Calculating the Magnitude of Unquantified Health Effects
Several of the analyses of the health effects of arsenic in Taiwan use Standardized Mortality Ratios (SMRs) to compare
death rates in villages with high levels of arsenic in drinking water to death rates in unexposed areas. The analysis below
compares the number of excess deaths due to lung and bladder cancers (based on SMRs) with excess deaths due to other
cancers and due to vascular disease. The goal is to compare the magnitude of excess deaths for endpoints for which dose-
response has not been quantified to excess deaths for endpoints for which dose-response functions exist. This suggests the
possible magnitude of effects that might be established if dose-response functions were estimated.
The spreadsheet in Attachment 1 to Appendix 2.2, performs this analysis using data reported in Wu et al. (1989) and Tsai
et al (1999). For the Wu et al. data the basic findings are that (1) cancers other than lung and bladder have similar aggregate
excess deaths as the sum of lung plus bladder cancer excess deaths, and (2) vascular deaths are comparable in number to the
sum of lung plus bladder cancer excess deaths. This suggests that the total mortality effect at the high exposure levels in the
Wu et al. study is about three times the effect of the previously quantified lung and bladder cancers. For the Tsai et al. data,
the basic findings are similar for total excess cancer deaths—about double those from lung plus bladder cancer by themselves.
However, the vascular excess deaths for these data are just over half the excess deaths from lung plus bladder cancers. This
A-19
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite"
apparent difference from the Wu et al. results may be related to the fact that more of the Tsai et al. data are from a somewhat
later period relative to the end of exposure than the earlier Wu et al. data. One possible interpretation of this is that the
vascular deaths may tend to have a shorter average lag time relative to exposures than the cancer deaths.
A-20
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
Attachment 1 to Appendix 2.2
A-21
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
N u mbers ofDeaths (Table 2)
<
Cancers
All sites
Bladder
Kidney
Skin
Lung
Liver
Prostate
Leukemia
Nasopharynx
Esophagus
Stomach
C ol on
Uterine Cervix
:Uni dentifi ed sit e's
^» ""«'"•""'""
Data from
-Tsai, S. M. , Wang, T. N , andK
Q
All Causes
All sites
Oral
Nasopharyngeal
Esophagus
Stomach
Intestine
Colon
Rectum
Liver
Gallbladder
Pancreas
Nasal
Laryngeal
Lung
Bone
Skin
Breast
Cervical
Ovary
Prostate
Bladder
.From Tables 3 and 4
Males Females Males Females
3ppm ; .3 -.5 9 ppm >i= .6 ppm .:< .3 ppm '.3 -.5 9 ppm ;= .6 ppm ;< .3 ppm .3-.59ppm := .6 ppm ,-< .3 ppm i. 3 -.5 9 ppm -= .6 ppm
243 244 150 191 180 144 224 56 405 12 534 61 162 22 277 2 472
23;.; 36J 26; 30 361 30 'i 22.64; 61.02s 92.71; 25.6; 57.02" 11.3
9 11? 6 :• 4; 13 'i 16;; 8 . 4 2 \ 18.9 25.26' 3 . 4 2 ; 19.42 798'
2 8 9 2 10 5 2 03 14 01 32 41 1 73 14 75 8 66
53 i 62 i 32 43' 40 t 38 49.16s 100.67 104.08 36.71,; 60.82'' 12.16
54' 42;-, 27:; 25' 16 \ 10* 4 7 . 7 8 i 67.62 86.73= 21.4;- 24.18-? 1.75
153 0 95 9 9 18
6 4S 1 ? 4" 3* f 4.871 6.52 i 2.69 3.03 i ' 4.55 ? 0.00-
4 5 2 2 4 1 358 816 858 159 581, 489
8 Si 2? 2 \ 2: '* 7.62. 9.37 6.55' 1 . 8 3 i 3.64 O.Oo'
26 10 10 8 11 2 25 66 17 82 56 42 6 71 18 72 5 98
8 6 3 11 5 5 7 94 83 12 51 9 05 8 16 17 21
1 4 1 091546392
49i 50" 29" 59; 36 i 36; 43.91 83.73; 97.49:- 50.24-* 54.67" 113.35
363 230 136 320 226 93 364 1 421 47 572 68 277 5 370 79 386 41
21 ' 29 14" 21 29 8 22 54 57 8 60 4 182 48 00 35 82
127 85,{ 62? 105= 93; 3 / •' 125.87' 153.98 259.51-' 91.14s' 153.07 14474
137 81 44 106 60 30 137 8 145 36 175 72 92 42 98 11 120 68
..,..np.,,-....,..P "
2774 1263 95 2 19 211 2 28 1510 2.01 ' 2029 843 9 24 2 3 2 51
23~' 20' ;; >) 3 •? 0 . 0 0 s 12 7.46
60 " 50 59 90 01 29 31 13
69 41 2 1 67 13 2 12 28 0 04 12 7 59
195s 143.84' 1.36? 1.17: 1.461 51 : 0.07 '. :' 1 1 1 i 79.46 1 . 4 { 1.15" 168
15s 7.15 i; >; ; , 8 . 0.01 f' • 8 > 5.81
91 6105' 30 x 004 83 5847 142" 113 176
46 31 96 14 0 02 33 21 98 15 1 03 2 11
631 ; 345.27 1.83. 1.695 1.98*" 286 * 0.38 i • 2 2 4 ; 119.28* 1.88,, 1.641 2.14
131168 1 000 111218
30 ' 24 57 5 ' 0 01 1 9 19 75
40 13 3 3 2 14 4 09 27 0 04 29 5 82 98 3 33 7 15
30 J 16. 8 1 : 1 . 7 8 " 1.2! 2.55s 13 ';: 0.02 • 13 ; 2.73- . 7 6 •• 2.53) 8.15
699 225 39 31 2 88 3 34 474 0 63 471 114 02 13 3 77 4 52
41'; 16.64*' 2.46= 1.77; 3.34-! 24 < 0.03 34.' 15.li; .25" 1.56;' 3.14
66' 1365' 483 374 615 52 007 68 1196 68 441 721
47 46.48
122; 96.09t 1.27? 1 . 0 5 >' 1.52
15 13 78
48 19 07 2 52 1 86 3 34 " 29 0 04
312; 34.99? 8.92;-: 7.96' 9.96; 277 •? 0.37 ; 295- 20.96? 14.07:- 12.51*" 1578
Males
3 - .5 9 pp
1 80
3
1
1
5
1
3
57
3
2
Deaths
2 54 5
1 185
5
-2
4
32
2
25
11
105
-1
-1
23
10
357
19
56
1
26
1
274
38.38
10.48
A-22
-------
PANEL REVIEW DRAFTa; August 9, 2001 "Do Not Quote or Cite'
A-23
------- |