[Federal Register Volume 66, Number 14 (Monday, January 22, 2001)]
[Rules and Regulations]
[Pages 6976-7066]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 01-1668]
[[Page 6975]]
-----------------------------------------------------------------------
Part VIII
Environmental Protection Agency
-----------------------------------------------------------------------
40 CFR Parts 9, 141, and 142
National Primary Drinking Water Regulations; Arsenic and Clarifications
to Compliance and New Source Contaminants Monitoring; Final Rule
Federal Register / Vol. 66, No. 14 / Monday, January 22, 2001 / Rules
and Regulations
[[Page 6976]]
-----------------------------------------------------------------------
ENVIRONMENTAL PROTECTION AGENCY
40 CFR Parts 9, 141 and 142
[WH-FRL-6934-9]
RIN 2040-AB75
National Primary Drinking Water Regulations; Arsenic and
Clarifications to Compliance and New Source Contaminants Monitoring
AGENCY: Environmental Protection Agency (EPA).
ACTION: Final rule.
-----------------------------------------------------------------------
SUMMARY: Today EPA is establishing a health-based, non-enforceable
Maximum Contaminant Level Goal (MCLG) for arsenic of zero and an
enforceable Maximum Contaminant Level (MCL) for arsenic of 0.01 mg/L
(10 g/L). This regulation will apply to non-transient non-
community water systems, which are not presently subject to standards
on arsenic in drinking water, and to community water systems.
In addition, EPA is publishing clarifications for monitoring and
demonstration of compliance for new systems or sources of drinking
water. The Agency is also clarifying compliance for State-determined
monitoring after exceedances for inorganic, volatile organic, and
synthetic organic contaminants. Finally, EPA is recognizing the State-
specified time period and sampling frequency for new public water
systems and systems using a new source of water to demonstrate
compliance with drinking water regulations. The requirement for new
systems and new source monitoring will be effective for inorganic,
volatile organic, and synthetic organic contaminants.
DATES: This rule is effective March 23, 2001, except for the amendments
to Secs. 141.23(i)(1), 141.23(i)(2), 141.24(f)(15), 141.24(h)(11),
141.24(h)(20), 142.16(e), 142.16(j), and 142.16(k) which are effective
January 22, 2004.
The compliance date for requirements related to the clarification
for monitoring and compliance under Secs. 141.23(i)(1), 141.23(i)(2),
141.24(f)(15), 141.24(f)(22), 141.24(h)(11), 141.24(h)(20), 142.16(e),
142.16(j), and 142.16(k) is January 22, 2004. The compliance date for
requirements related to the revised arsenic standard under
Secs. 141.23(i)(4), 141.23(k)(3), 141.23(k)(3)(ii), 141.51(b),
141.62(b), 141.62(b)(16), 141.62(c), 141.62(d), and 142.62(b) is
January 23, 2006. For purposes of judicial review, this rule is
promulgated as of January 22, 2001.
ADDRESSES: Copies of the public comments received, EPA responses, and
all other supporting documents are available for review at the U.S. EPA
Water Docket (4101), East Tower B-57, 401 M Street, SW, Washington DC
20460. For an appointment to review the docket, call 202-260-3027
between 9 a.m. and 3:30 p.m. and refer to Docket W-99-16.
FOR FURTHER INFORMATION CONTACT: The Safe Drinking Water Hotline,
phone: (800) 426-4791, or (703) 285-1093, e-mail: [email protected]
for general information about, and copies of, this document and the
proposed rule. For technical inquiries, contact: Jeff Kempic, (202)
260-9567, e-mail: [email protected] for treatment and costs, and
Dr. John B. Bennett, (202) 260-0446, e-mail: [email protected] for
benefits.
SUPPLEMENTARY INFORMATION:
Regulated Entities
A public water system (PWS), as defined in 40 CFR 141.2, provides
water to the public for human consumption through pipes or ``other
constructed conveyances, if such system has at least fifteen service
connections or regularly serves an average of at least twenty-five
individuals daily at least 60 days out of the year.'' A public water
system is either a community water system (CWS) or a non-community
water system (NCWS). A community water system, as defined in
Sec. 141.2, is ``a public water system which serves at least fifteen
service connections used by year-round residents or regularly serves at
least twenty-five year-round residents.'' The definition in Sec. 141.2
for a non-transient non-community water system (NTNCWS) is ``a public
water system that is not a [CWS] and that regularly serves at least 25
of the same persons over 6 months per year.'' EPA has an inventory
totaling over 54,000 community water systems and approximately 20,000
non-transient non-community water systems nationwide. Entities
potentially regulated by this action are community water systems and
non-transient non-community water systems. The following table provides
examples of the regulated entities under this rule.
Table of Regulated Entities
------------------------------------------------------------------------
Category Examples of regulated entities
------------------------------------------------------------------------
Industry.......................... Privately owned/operated community
water supply systems using ground
water, surface water, or mixed
ground water and surface water.
State, Tribal, and Local State, Tribal, or local government-
Government. owned/operated water supply systems
using ground water, surface water,
or mixed ground and surface water.
Federal Government................ Federally owned/operated community
water supply systems using ground
water, surface water, or mixed
ground water and surface water.
------------------------------------------------------------------------
The table is not intended to be exhaustive, but rather provides a
guide for readers regarding entities likely to be regulated by this
action. This table lists the types of entities that EPA is now aware
could potentially be regulated by this action. Other types of entities
not listed in this table could also be regulated. To determine whether
your facility is regulated by this action, you should carefully examine
the applicability criteria in Secs. 141.11 and 141.62 of the rule. If
you have any questions regarding the applicability of this action to a
particular entity, consult the general information contact listed in
the section listing contacts for further information.
Abbreviations used in this rule
--less than
--less than or equal to
>--greater than
--greater than or equal to
--plus or minus
Sec. --section
--, Greek letter, in statistics represents standard
deviation
g--Microgram, one-millionth of a gram (3.5 x
10-\8\ of an ounce)
g/L--micrograms per liter
AA--Activated alumina
AIC--Akaike Information Criterion
ACWA--Association of California Water Agencies
AMWA--Association of Metropolitan Water Agencies
APHA--American Public Health Association
[[Page 6977]]
ARARs--Applicable or relevant and appropriate requirements
As (III)--Trivalent arsenic. Common inorganic form in water is arsenite
As (V)--Pentavalent arsenic. Common inorganic form in water is arsenate
ASDWA-- Association of State Drinking Water Administrators
AsH3--Arsine
ASTM--American Society for Testing and Materials
ATSDR--Agency for Toxic Substances and Disease Registry, U.S.
Department of Health & Human Services
AWWA--American Water Works Association
AWWARF--American Water Works Association Research Foundation
BAT--Best available technology
BV--Bed volume
CCR--Consumer Confidence Report
CERCLA--Comprehensive Environmental Response, Compensation, and
Liability Act administered by EPA for hazardous substances
C/F--Modified coagulation/filtration
CFR--Code of Federal Regulations
CSFII--Continuing Survey of Food Intakes by Individuals
CWA--Clean Water Act administered by EPA for surface waters of the U.S.
CWS--Community water system
CWSS--Community Water System Survey
DMA--Dimethyl arsinic acid, cacodylic acid,
(CH3)2HAsO2
DNA--Deoxyribonucleic acid
DWSRF--Drinking Water State Revolving Fund
EA--Economic analysis
EDR--Electrodialysis reversal
EEAC--Environmental Economics Advisory Committee
e.g.--exempli gratia, Latin for ``for example''
EPA--U.S. Environmental Protection Agency
et al.--et alia, Latin for ``and others''
FACA--Federal Advisory Committee Act
FR--Federal Register
FRFA--Final Regulatory Flexibility Analysis
FSIS--Federalism Summary Impact Statement
GDP--Gross Domestic Product
GFAA--Graphite furnace atomic absorption
GHAA--Gaseous hydride atomic absorption
GI--Gastrointestinal
GW--Ground water
GWR--Ground Water Rule
HRRCA--Health Risk Reduction and Cost Analysis
ICP-AES--Inductively coupled plasma-atomic emission spectroscopy
ICP-MS--Inductively coupled plasma mass spectroscopy
ICR--Information collection request
i.e.--id est, Latin for ``that is''
IOCs--Inorganic contaminants
ISCV--Intra-system coefficient of variation
IX--Ion exchange
L--Liter, also referred to as lower case ``l'' in older citations
LD50--The dose of a chemical taken by mouth or absorbed by
the skin which is expected to cause death in 50% of the test animals
LS--Modified lime softening
LT1/FBR--Long Term 1 Enhanced Surface Water Treatment and Filter
Backwash Recycling Rule
MCL--Maximum contaminant level
MCLG--Maximum contaminant level goal
MDL--Method detection limit
mg--Milligrams, one-thousandth of a gram, 1 milligram=1,000 micrograms
mg/kg--Milligrams arsenic per kilogram body weight or soil weight
mg/L--Milligrams per liter
MHI--Mean household income
MMA--Monomethyl arsenic, arsonic acid,
CH3H2ASO3
NAOS--National Arsenic Occurrence Survey
NAS--National Academy of Sciences
NAWQA--National Ambient Water Quality Assessment, USGS
NCI--National Cancer Institute
NCWS--Non-community water system
NDWAC--National Drinking Water Advisory Council for EPA
NIRS--National Inorganic and Radionuclide Survey done by EPA
NODA--Notice of Data Availability
NOMS--National Organic Monitoring Survey done by EPA
NPDES--National Pollutant Discharge Elimination System for CWA
NPDWR--National primary drinking water regulation
NR--Not reported
NRC--National Research Council, the operating arm of NAS
NTNCWS--Non-transient non-community water system
NTTAA--National Technology Transfer and Advancement Act
NWIS--National Water Information System of USGS
OGWDW--Office of Ground Water and Drinking Water in EPA
OMB--Office of Management and Budget
PE--Performance evaluation, studies to certify laboratories for EPA
drinking water testing
pH--Negative log of hydrogen ion concentration
PNR--Public Notification Rule
POE--Point-of-entry treatment devices
POTWs--Publicly owned treatment works, treat wastewater
POU--Point-of-use treatment devices
ppb--Parts per billion
ppm--Parts per million
PQL--Practical quantitation level
PRA--Paperwork Reduction Act
psi--Pounds per square inch
PT--Performance testing
PUC--Public utilities commission
PWS--Public water systems
QALYs--Quality adjusted life years
RCRA--Resource Conservation and Recovery Act
REF--Relative exposure factors
RFA--Regulatory Flexibility Act
RIA--Regulatory Impact Analysis
RO--Reverse osmosis
RUS--Rural Utilities Service
RWS--Rural Water Survey
SAB--Science Advisory Board
SBAR--Small Business Advocacy Review
SBREFA--Small Business Regulatory Enforcement Fairness Act
SD--Standard deviation
SDWA--Safe Drinking Water Act
SDWIS--Safe Drinking Water Information System
SEER--Surveillance, Epidemiology, and End Results
SM--Standard Method for Examination of Water and Wastewater
SMF--Standardized monitoring framework
SMRs--Standardized mortality ratios
SO4--Sulfate
SOCs--Synthetic organic contaminants
STP-GFAA--Stabilized temperature platform graphite furnace atomic
absorption
SW--Surface water
TBLLs--Technically based local limits
TC--Toxicity Characteristic, RCRA hazardous waste
TCLP--Toxicity Characteristic Leaching Procedure, tests for hazardous
waste
TDS--Total dissolved solids
TMF--Technical, managerial, financial capacity
TOC--Total organic carbon
UMRA--Unfunded Mandates Reform Act
URTH--Unreasonable risk to health
U.S.--United States
USDA--US Department of Agriculture
USGS--US Geological Survey
UV--Ultraviolet
VOCs--Volatile organic contaminants
VSL--Value of statistical life
VSLY--Value of statistical life year
WHO--World Health Organization
WS--Water supply
WTP--Willingness-to-pay
Table of Contents
I. Background and Summary of the Final Rule
A. What Did EPA Propose?
B. Overview of the Notice of Data Availability (NODA)
C. Does This Regulation Apply to My Water System?
[[Page 6978]]
D. What are the Final Drinking Water Regulatory Standards for
Arsenic (Maximum Contaminant Level Goals and Maximum Contaminant
Levels)?
E. Will There be a Health Advisory?
F. What are the Best Available Technologies For Removing Arsenic
From Drinking Water?
1. BAT technologies
2. Preoxidation
3. Factors affecting listing technologies
4. Other technologies evaluated, but not designated as BAT
5. Waste disposal
G. Treatment Trains Considered For Small Systems
1. Can my water system use point-of-use (POU), point-of-entry
(POE), or bottled water to comply with this regulation?
2. What are the affordable treatment technologies for small
systems?
3. Can my water system get a small system variance from an MCL
under today's rule?
H. Can My System Get a General Variance or Exemption from the MCL
Under Today's Rule?
I. What Analytical Methods are Approved for Compliance Monitoring of
Arsenic and What are the Performance Testing Criteria for Laboratory
Certification?
1. Approved analytical methods
2. Performance testing criteria for laboratory certification
J. How Will I Know if My System Meets the Arsenic Standard?
1. Sampling points and grandfathering of monitoring data
2. Compositing of samples
3. Calculation of violations
4. Monitoring and compliance schedule
K. What do I Need To Tell My Customers?
1. Consumer Confidence Reports
a. General requirements
b. Special informational statement
2. Public Notification
L. What Financial Assistance Is Available for Complying With This
Rule?
M. What is the Effective Date and Compliance Date for the Rule?
N. How Were Stakeholders Involved in the Development of This Rule?
II. Statutory Authority
III. Rationales for Regulatory Decisions
A. What is the MCLG?
B. What is the Feasible Level?
1. Analytical measurement feasibility
2. Treatment
C. How Did EPA Revise Its National Occurrence Estimates?
1. Summary of occurrence data and methodology
2. Corrections and additions to the data
3. Changes to the methodology
4. Revised occurrence results
D. How Did EPA Revise Its Risk Analysis?
1. Health risk analysis
a. Toxic forms of arsenic
b. Effects of acute toxicity
c. Non-cancer effects associated with arsenic.
d. Cancers associated with arsenic
e. How does arsenic cause cancer?
f. What is the quantitative relationship between exposure and
cancer effects that may be projected for exposures in the U.S.?
g. Is it appropriate to assume linearity for the dose-response
assessment for arsenic at low doses given that arsenic is not
directly reactive with DNA?
2. Risk factors/bases for upper- and lower-bound analyses
a. Water consumption
b. Relative Exposure Factors
c. Arsenic occurrence
d. Risk distributions
e. Estimated risk reductions
f. Lower-bound analyses
g. Cases avoided
3. Sensitive subpopulations
4. Risk window
E. What are the Costs and Benefits at 3, 5, 10, and 20 g/L?
1. Summary of cost analysis
a. Total national costs
b. Household costs
2. Summary of benefits analysis
a. Primary analysis
b. Sensitivity analysis on benefits valuation
c. SAB recommendations
d. Analytical approach
e. Results
3. Comparison of costs and benefits
a. Total national costs and benefits
b. National net benefits and benefit-cost ratios
c. Incremental costs and benefits
d. Cost-per-case avoided
4. Affordability
F. What MCL Is EPA Promulgating and What Is the Rationale for This
Level?
1. Final MCL and overview of principal considerations
2. Consideration of health risks
3. Comparison of benefits and costs
4. Rationale for the final MCL
a. General considerations
b. Relationship of MCL to the feasible level (3 g/L)
c. Reanalysis of proposed MCL and comparison to final MCL
d. Consideration of higher MCL options
e. Conclusion
IV. Rule Implementation
A. What are the Requirements for Primacy?
B. What are the Special Primacy Requirements?
C. What are the State Recordkeeping Requirements?
D. What are the State Reporting Requirements?
E. When does a State Have to Apply for Primacy?
F. What are Tribes Required To Do Under This Regulation?
V. Responses to Major Comments Received
A. General Comments
1. Sufficiency of information and adequacy of procedural
requirements to support a final rule
2. Suggestions for development of an interim standard
3. Public involvement and opportunity for comment
4. Relation of MCL to the feasible level
5. Relationship of MCL to other regulatory programs
6. Relation of MCL to WHO standard
7. Regulation of non-transient non-community water systems
(NTNCWSs)
8. Extension of effective date for large systems
B. Health Effects of Arsenic
1. Epidemiology data
2. Dose-response relationship
3. Suggestions that EPA await further health effects research
4. Sensitive subpopulations
5. EPA's risk analysis
6. Setting the MCLG and the MCL
C. Occurrence
1. Occurrence data
2. Occurrence methodology
3. Co-occurrence
D. Analytical Methods
1. Analytical interferences
2. Demonstration of PQL (includes acceptance limits)
3. Acidification of samples
E. Monitoring and Reporting Requirements
1. Compliance determinations
2. Monitoring of POU devices
3. Monitoring and reporting for NTNCWSs
4. CCR health language and reporting date
5. Implementation guidance
6. Rounding analytical results
F. Treatment Technologies
1. Demonstration of technology performance
2. Barriers to technology application
3. Small system technology application
4. Waste generation and disposal
a. Anion exchange
b. Activated alumina
c. Reverse osmosis
5. Emerging technologies
G. Costs
1. Disparity of costs
a. What is EPA's response to major comments on the decision tree
for the proposed rule?
b. What is EPA's response to comments on system level costs?
c. What is EPA's response to comments that state the report
``Cost Implications of a Lower Arsenic MCL'' (Frey et al., 2000), be
used as a basis for reflecting more realistic national costs than
EPA's estimates?
2. Affordability
3. Combined cost of new regulations
4. Projected effects of the new standard on other regulatory
programs.
H. Benefits of Arsenic Reduction
1. Timing of benefits accrual (latency)
2. Use of the Value of Statistical Life (VSL)
3. Use of alternative methodologies for benefits estimation
4. Comments on EPA's consideration of nonquantifiable benefits
5. Comments on EPA's assumption of benefits accrual prior to
rule implementation
I. Risk Management Decision
1. Role of uncertainty in decision making
2. Agency's interpretation of benefits justify costs provision
3. Alternative regulatory approaches
4. Standard for total arsenic vs. species-specific standards
J. Health Risk Reduction and Cost Analysis (HRRCA)
1. Notice and comment requirement
2. Conformance with SDWA requirements
[[Page 6979]]
VI. Administrative and Other Requirements
A. Executive Order 12866: Regulatory Planning and Review
B. Regulatory Flexibility Act (RFA), as Amended by the Small
Business Regulatory Enforcement Fairness Act of 1996 (SBREFA), 5
U.S.C. 601 et seq.
C. Unfunded Mandates Reform Act (UMRA) of 1995
a. Authorizing legislation
b. Cost-benefit analysis
c. Financial assistance
d. Estimates of future compliance costs and disproportionate
budgetary effects
e. Macroeconomic effects
f. Summary of EPA's consultation with State, Tribal, and local
governments
g. Nature of State, Tribal, and local government concerns and how
EPA addressed these concerns
h. Regulatory alternatives considered
i. Selection of the regulatory alternative
D. Paperwork Reduction Act (PRA)
E. National Technology Transfer and Advancement Act (NTTAA)
F. Executive Order 12898: Environmental Justice
G. Executive Order 13045: Protection of Children from Environmental
Health Risks and Safety Risks
H. Executive Order 13132: Federalism
I. Executive Orders 13084 and 13175: Consultation and Coordination with
Indian Tribal Governments
J. Plain Language
K. Congressional Review Act
L. Consultations with the Science Advisory Board, National Drinking
Water Advisory Council, and the Secretary of Health and Human Services
M. Likely Effect of Compliance With the Arsenic Rule on the Technical,
Financial, and Managerial Capacity of Public Water Systems
VI. References
List of Tables
Table I.F-1.--Best Available Technologies and Removal Rates
Table I.G-1.--Treatment Technology Trains
Table I.G-2.--Baseline Values for Small Systems Categories
Table I.G-3.--Available Expenditure Margin for Affordable Technology
Determinations
Table I.G-4.--Design and Average Daily Flows Used for Affordable
Technology Determinations
Table I.G-5.--Affordable Compliance Technology Trains for Small
Systems with population 25-500
Table I.G-6.--Affordable Compliance Technology Trains for Small
Systems with populations 501-3,300 and 3,301 to 10,000
Table I.I-1.--Approved Analytical Methods (40 CFR 141.23) for
Arsenic at the MCL of 0.01 mg/L
Table III.C-1.--Summary of Occurrence Databases for the Proposed and
Final Rules
Table III.C-2.--Alaska PWS Inventories: Baseline Handbook and
Corrected
Table III.C-3.--National Occurrence Exceedance Probability Estimates
Table III.C-4.--Parameters of Lognormal Distributions Fitted to
National Occurrence Distributions
Table III.C-5.--Regional Occurrence Exceedance Probability Estimates
Table III.C-6.--Statistical Estimates of Numbers of Systems with
Average Finished Arsenic Concentrations in Various Ranges
Table III.C-7.--Estimated Intra-System Coefficients of Variation
Table III.C-8.--Comparison of National Arsenic Occurrence Estimates
Table III.D-1.--Life-Long Relative Exposure Factors
Table III.D-2(a).--Cancer Risks for U.S. Populations Exposed At or
Above MCL Options, after Treatment1,2 (Without Adjustment
for Arsenic in Food and Cooking Water)
Table III.D-2(b).--Cancer Risks for U.S. Populations Exposed At or
Above MCL Options, after Treatment1,2 (With Adjustment
for Arsenic Exposure in Food and Cooking Water)
Table III.D-2(c).--Cancer Risks for U.S. Populations Exposed At or
Above MCL Options, after Treatment1 (Lower Bound With
Food and Cooking Water Adjustment, Upper Bound Without Food and
Cooking Water Adjustment)
Table III.D-3.--Annual Total (Bladder and Lung) Cancer Cases Avoided
from Reducing Arsenic in CWSs and NTNCWS
Table III.E-1.--Total Annual National System and State Compliance
Costs
Table III.E-2.--Mean Annual Costs per Household
Table III.E-3.--Estimated Benefits from Reducing Arsenic in Drinking
Water
Table III.E-4.--Sensitivity of the Primary VSL Estimate to Changes
in Latency Period Assumptions, Income Growth, and Other Adjustments
Table III.E-5.--Sensitivity of Combined Annual Bladder and Lung
Cancer Mortality Benefits Estimates to Changes in VSL Adjustment
Factor Assumptions
Table III.E-6.--Sensitivity of Combined Annual Bladder and Lung
Cancer Mortality Benefits Estimates to Changes in VSL Adjustment
Factor Assumptions
Table III.E-7.--Estimated Annual Costs and Benefits from Reducing
Arsenic in Drinking Water
Table III.E-8 Summary of National Annual Net Benefits and Benefit-
Cost Ratios, Combined Bladder and Lung Cancer Cases
Table III.E-9 Estimates of the Annual Incremental Risk Reduction,
Costs, and Benefits of Reducing Arsenic in Drinking Water
Table III.E-10. Annual Cost Per Cancer Case Avoided for the Final
Arsenic Rule--Combined Bladder and Lung Cancer Cases
TABLE V.F-4.1 Treatment Trains in Final Versus Proposed Arsenic Rule
Decision Tree
Table V.F-4.2 New or Revised Treatment Trains
Table VI.B-1. Profile of the Universe of Small Water Systems
Regulated Under the Arsenic Rule
I. Background and Summary of the Final Rule
A. What Did EPA Propose?
On June 22, 2000, the Federal Register published EPA's proposed
arsenic regulation for community water systems and non-transient non-
community water systems (65 FR 38888; EPA, 2000i). EPA proposed a
health-based, non-enforceable goal, or Maximum Contaminant Level Goal
(MCLG), of zero micrograms per liter (g/L) and a Maximum
Contaminant Level (MCL) of 5 g/L. The Agency also requested
comment on alternate MCL levels of 3 g/L, 10 g/L, and
20 g/L. (In the proposed rule EPA expressed arsenic
concentration in milligrams per liter (mg/L) or parts per million,
which matches the units of the former and current standard for arsenic.
Except as noted, the Agency will refer to arsenic concentration in
micrograms per liter (g/L) in this preamble.)
EPA based the June 2000 proposal on extensive analysis including a
careful consideration of the following issues: a nonzero MCLG;
occurrence of arsenic in public water systems; our approach for
estimating national occurrence and co-occurrence; acceptance limits
used to establish the practical quantitation level (PQL); rounding of
measured values for compliance purposes; extending compliance by two
years for systems serving under 10,000 people in order to add capital
improvements; dates for reporting changes in the consumer confidence
reports and public notification; appropriateness of the national
affordability criteria; affordable technologies for small systems;
implementation issues for point-of-use (POU) and point-of-entry (POE)
treatments; appropriateness of non-hazardous residual costing; our
overall analysis of costs; adjusting benefits estimates (e.g., for
factors such as latency); our approach for considering uncertainties
that affected risk; use of the authority to set an MCL at a level other
than the feasible MCL; expression of the MCL as total arsenic;
approaches to regulation of NTNCWSs; State program revisions; selenium
levels as an attenuation factor in arsenic toxicity; impacts on small
entities; use of consensus analytical methods; methods to address
environmental justice concerns; and comments on use of plain
[[Page 6980]]
language. We asked commenters to submit data and comments on these
issues, as well as any other issues raised in the proposal.
The proposal reflected several types of technical evaluations,
including analytical methods performance and laboratory capacity; the
likelihood of different size water systems choosing treatment
technologies based on source water characteristics; and the national
occurrence of arsenic in drinking water supplies. Furthermore, the
Agency assessed the quantifiable and nonquantifiable costs and health
risk reduction benefits likely to occur at the treatment levels
considered, and the effects of arsenic on sensitive subpopulations.
The proposed MCL was consistent with the Agency's use of the new
benefit/cost provisions of the Safe Drinking Water Act (SDWA), as
amended in 1996 (see section II. of this preamble for additional
information about this provision). EPA proposed 3 g/L as the
feasible MCL, after considering treatment costs and efficiency under
field conditions as well as considering the appropriate analytical
methods. Because EPA determined that the benefits of regulating arsenic
at the feasible level would not justify the costs, the Agency proposed
an MCL of 5 g/L, while requesting comment on MCL options of 3
g/L (the feasible level), 10 g/L, and 20 g/
L.
We based our estimates of large system compliance costs primarily
on costs for coagulation/filtration and lime softening, although we
consider several other technologies to be appropriate as best available
technology (BAT) technologies. (See Table I.F-1.) For small-system
(systems serving 10,000 people and less) compliance costs, we
considered the costs for ion exchange, activated alumina, reverse
osmosis, and nanofiltration. EPA proposed extending the effective date
to five years after the final rule issuance for small community water
systems and maintaining the effective date at three years after
promulgation for all other community water systems. EPA proposed that
States applying to adopt the revised arsenic MCL may use their most
recently approved monitoring and waiver plans or note in their primacy
application any revisions to those plans. EPA proposed that NTNCWSs
monitor for arsenic and report exceedances of the MCL.
The Agency also clarified the procedure used for determining
compliance after exceedances for inorganic, volatile organic, and
synthetic organic contaminants in Secs. 141.23(i)(2),
141.24(f)(15)(ii), and 141.24(h)(11)(ii), respectively. Finally, EPA
proposed that new systems and systems using a new source of water be
required to demonstrate compliance with the MCLs using State-specified
time frames. The clarified new source and new system compliance
regulations require that States establish initial sampling frequencies
and compliance periods for inorganic, volatile organic, and synthetic
organic contaminants in Secs. 141.23(c)(9), 141.24(f)(22), and
141.24(h)(20), respectively.
B. Overview of the Notice of Data Availability (NODA)
In the proposed rule, EPA quantified the risk reduction and
benefits of avoiding bladder cancer and noted that a peer-reviewed
quantification of lung cancer risk from arsenic exposure would probably
be available in time to consider for the final rule (65 FR 38888 at
38899; EPA, 2000i). Relying upon a discussion in the National Research
Council (NRC) report (NRC, 1999, pg. 8) about the qualitative risks of
lung cancer (65 FR 38888 at 38944; 2000i), EPA provided a ``What-If''
estimate of lung cancer benefits (65 FR 38888 at 38946, 2000i) in the
proposed rule. On October 20, 2000, the Federal Register published
EPA's Notice of Data Availability (NODA) containing a revised risk
analysis for bladder cancer and new risk information concerning lung
cancer (65 FR 63027; EPA, 2000m), and identified a correction to Table
4 on October 27, 2000 (65 FR 64479; EPA, 2000n). The NODA also provided
information concerning the availability of cost curves used to develop
the costs published in the proposal.
EPA used new risk information for lung and bladder cancer from a
peer-reviewed article written by Morales et al. (2000). In the NODA,
EPA explained that the authors used several alternative statistical
models to estimate cancer risk. EPA explained its reasons for selecting
``Model 1'' with no comparison population for further analysis. We used
daily water consumption (EPA, 2000c) reported by gender, region, age,
economic status, race, and separately for pregnant women, lactating
women, and women in childbearing years combined with weight data to
derive exposure factors for the U.S. We used these exposure factors,
our occurrence estimate (EPA 2000g) of populations exposed to arsenic
at different concentrations, and the risk distributions from the
Morales et al. (2000) paper in Monte Carlo simulations to estimate the
upper bound of risks faced by the U.S. population. The NODA compared
the bladder cancer risks derived for the proposal against the bladder
cancer risks derived from the Morales et al. (2000) study. EPA also
derived lung cancer risks using the same approach and the risk model
contained in the Morales et al. (2000) study.
EPA also used the newly calculated risks to estimate a lower bound
risk in the U.S. This calculation took into account the amount of
additional arsenic people in Taiwan were likely to have ingested from
water used in food preparation. EPA showed the effects on risks for the
U.S. population at both the mean and 90th percentile levels for various
arsenic levels in drinking water. Based on the revised risk assessment,
we updated our assessment of the relative risk of lung cancer as
compared to bladder cancer. The NODA indicated that instead of being 2
to 5 times as many fatal lung cancer cases as bladder cancer cases (as
was cited in NRC's Executive Summary, NRC, 1999, pg. 8 as a qualitative
estimate), the combined risk of excess lung and bladder cancer were
thought to be only about twice that of bladder cancer risk. EPA noted
that, while the new risks were higher than the bladder cancer risk in
the proposal, the monetized benefits of lung cancer would fall within
the lung cancer benefits range estimated using the ``What-If'' analysis
(e.g., $19.6 million--$224 million yearly for an MCL of 10 g/
L) in the proposal (65 FR 38888 at 38959; EPA, 2000m).
In the NODA, EPA also explained that the docket for the proposed
rule had the November 1999 version (EPA, 1999o) of ``Technologies and
Costs for the Removal of Arsenic from Drinking Water'' rather than the
April 1999 version of the document that was the primary source for the
treatment technology cost equations used to generate the national cost
estimate. The national cost estimate was presented in the ``Proposed
Arsenic in Drinking Water Rule Regulatory Impact Analysis'' (EPA,
2000h). The NODA therefore announced the availability of the
``Technologies and Costs for the Removal of Arsenic from Drinking
Water,'' dated April 1999 (EPA,1999b). The NODA also noted that
commenters interested in reproducing the waste disposal curves should
consult the ``Small Water System Byproducts Treatment and Disposal Cost
Document'' (EPA, 1993a) and ``Water System Byproducts Treatment and
Disposal Document (EPA, 1993b).'' In addition to placing these
documents in the docket, the NODA also specified that an electronic
copy of the treatment technology and waste disposal equations used in
the development of the RIA could be found in the docket.
[[Page 6981]]
EPA made the April 1999 version of the document, ``Technologies and
Costs for the Removal of Arsenic from Drinking Water'' (EPA,1999b)
available on its arsenic webpage.
The cost methodology and cost estimates were clearly stated and
explained in the proposal for public review and consideration. Through
a technical oversight, we incorrectly attributed the source for the
cost curves to the November version of the document placed in the
docket (EPA, 1999o). As a result, people could not replicate the
precise analysis we did, should a commenter desire to do so. More
specifically, although the inputs, assumptions, and model methodology
were clearly explained, we incorrectly cited the sources of an
intermediate step of deriving specific cost curves from those
assumptions. Based upon the proposal's detailed discussion of inputs,
assumptions and associated methodology, EPA believes the public was
fully able to review, understand, and comment on the Agency's estimate
of potential impacts. EPA discusses the cost curves further in section
III.E.1 of this preamble.
C. Does This Regulation Apply to My Water System?
The final regulation on arsenic in drinking water promulgated today
applies to all CWSs and NTNCWSs. The regulation not only establishes an
MCLG and MCL for arsenic, but also lists feasible technologies and
affordable technologies for small systems that can be used to comply
with the MCL. However, systems are not required to use the listed
technologies in order to meet the MCL.
D. What are the Final Drinking Water Regulatory Standards for Arsenic
(Maximum Contaminant Level Goals and Maximum Contaminant Levels)?
In today's rule, the MCLG is 0 g/L, and the enforceable
MCL is 0.01 mg/L, which is the same as 10 micrograms per liter
(g/L) or 10 parts per billion (ppb). EPA based the MCL on
total arsenic, because drinking water contains almost entirely
inorganic forms, and the analytical methods for total arsenic are
readily available and capable of being performed by certified
laboratories at an affordable cost.
E. Will There be a Health Advisory?
A health advisory for arsenic is not part of today's rulemaking.
EPA will be considering whether or not to issue a health advisory after
evaluating the recommendations of the Science Advisory Board (SAB)
(EPA, 2000q). The purpose of an advisory would be to provide useful
information to water providers between issuance and implementation of
this rule.
F. What are the Best Available Technologies For Removing Arsenic From
Drinking Water?
Section 1412(b)(4)(E) of the Safe Drinking Water Act states that
each National Primary Drinking Water Regulation (NPDWR) which
establishes an MCL shall list the technology, treatment techniques, and
other means that the Administrator finds to be feasible for purposes of
meeting the MCL. Technologies are judged to be a best available
technology (BAT) when the following criteria are satisfactorily met:
(1) The capability of a high removal efficiency;
(2) A history of full-scale operation;
(3) General geographic applicability;
(4) Reasonable cost based on large and metropolitan water systems;
(5) Reasonable service life;
(6) Compatibility with other water treatment processes; and
(7) The ability to bring all of the water in a system into
compliance.
EPA identified BATs in this section using the listed criteria.
Their removal efficiencies and a brief discussion of the major issues
surrounding the usage of each technology are also given in this
section. More details about the treatment technologies and costs can be
found in ``Technologies and Costs for the Removal of Arsenic From
Drinking Water'' (EPA, 2000t).
1. BAT technologies
EPA reviewed several technologies as BAT candidates for arsenic
removal, e.g., ion exchange, activated alumina, reverse osmosis,
nanofiltration, electrodialysis reversal, coagulation assisted
microfiltration, modified coagulation/filtration, modified lime
softening, greensand filtration, conventional iron and manganese
removal, and several emerging technologies. The Agency determined that,
of the technologies capable of removing arsenic from source water, only
the technologies in Table I.F-1 fulfill the requirements of SDWA for
BAT determinations for arsenic. The maximum percent of arsenic removal
that can be reasonably obtained from these technologies is also shown
in the table. These removal efficiencies are for arsenic (V) removal.
Table I.F-1.-- Best Available Technologies and Removal Rates
------------------------------------------------------------------------
Maximum
Treatment Technology Percent
Removal \1\
------------------------------------------------------------------------
Ion Exchange (sulfate 50 mg/L)................. 95
Activated Alumina.......................................... 95
Reverse Osmosis............................................ >95
Modified Coagulation/Filtration............................ 95
Modified Lime Softening (pH > 10.5)........................ 90
Electrodialysis Reversal................................... 85
Oxidation/Filtration (20:1 iron:arsenic)................... 80
------------------------------------------------------------------------
\1\ The percent removal figures are for arsenic (V) removal. Pre-
oxidation may be required.
2. Preoxidation
In water, the most common valence states of arsenic are As (V), or
arsenate, and As (III), or arsenite. As (V) is more prevalent in
aerobic surface waters and As (III) is more likely to occur in
anaerobic ground waters. In the pH range of 4 to 10, As (V) species
(H2AsO4\-\ and
H2AsO42\-\) are negatively charged,
and the predominant As (III) compound (H3AsO3) is
neutral in charge. Removal efficiencies for As (V) are much better than
removal of As (III) by any of the technologies evaluated because the
arsenate species carry a negative charge and arsenite is neutral under
these pH conditions. To increase the removal efficiency when As (III)
is present, pre-oxidation to the As (V) species is necessary.
As (III) may be converted through pre-oxidation to As (V) using one
of several oxidants. Data on oxidants indicate that chlorine, potassium
permanganate, and ozone are effective in oxidizing As (III) to As (V).
Pre-oxidation with chlorine may create undesirable concentrations of
disinfection byproducts and membrane fouling of subsequent treatments
such as reverse osmosis. EPA has completed research on the chemical
oxidants for As (III) conversion, and is presently investigating
ultraviolet light disinfection technology (UV) and solid oxidizing
media. For POU and POE devices, central chlorination may be required
for oxidation of As (III).
3. Factors affecting listing technologies
Ion Exchange (IX) can effectively remove arsenic using anion
exchange resins. It is recommended as a BAT primarily for sites with
low sulfate because sulfate is preferred over arsenic. Sulfate will
compete for binding sites resulting in shorter run lengths. Due to much
shorter run lengths than activated alumina, anion exchange must be
[[Page 6982]]
regenerated because it is not cost effective to dispose of the resin
after one use. Column bed regeneration frequency is a key factor in the
cost of the process and affects the volume of waste produced by the
process. The proposed rule preamble noted that anion exchange may be
practical up to approximately 120 mg/L of sulfate (Clifford, 1994). The
upper-bound sulfate concentration for the final rule is 50 mg/L. The
selection of this upper bound is based on several factors, including
cost and the ability to dispose of the brine stream.
The proposed rule listed three mechanisms to dispose of the brine
stream used for regeneration. The options were: sanitary sewer,
evaporation pond, and chemical precipitation. Many comments on the
proposed rule were based on the assumption that the waste streams
generated would be considered hazardous waste. Waste streams containing
less than 0.5% solids are evaluated against the toxicity characteristic
directly to determine if the waste is hazardous. Arsenic in the
regeneration brine will likely exceed 5 mg/L for most systems with
arsenic above 10 g/L and sulfate below 50 mg/L. Since the
brine stream would likely be considered hazardous, EPA eliminated the
evaporation pond and the chemical precipitation options from the
decision tree as options for disposal of anion exchange wastes. The
Agency retained discharge to a sanitary sewer because domestic sewage
and any mixture of domestic sewage and other wastes that pass through a
sewer system to a publicly owned treatment works (POTW) for treatment
is excluded from consideration as solid waste (40 CFR 261.4). Domestic
sewage means untreated sanitary wastes that pass through a sewage
system. Discharges meeting the previously stated criteria are excluded
from regulation as hazardous waste. However, these assumptions were
reviewed to substantially reduce projections of brine wastes going to
POTWs from those that were used in support of the proposed rule.
Discharge to a sanitary sewer can be limited by technically based
local limits (TBLLs) for arsenic or total dissolved solids. Since anion
exchange is regenerated more frequently than activated alumina, the
total dissolved solids increase can be significant. Many comments
indicated that significant increases in total dissolved solids would be
unacceptable, especially in the Southwest where water resources are
scarce. Salt is used for regeneration of anion exchange resins. The
upper bound of 50 mg/L sulfate for anion exchange is based on projected
increases of total dissolved solids using the quantity of salt needed
for regeneration and the frequency of regeneration (based on sulfate).
The sulfate upper bound for the final rule is significantly lower than
the upper bound from the proposed rule. Due to the potential for an
increase in total dissolved solids, anion exchange would be favored in
areas other than the Southwest where the volume of brine is very small
relative to the total volume of wastewater being treated at the POTW.
Systems that need to treat only a few entry points or can blend a
significant portion of the water to meet the MCL may produce a smaller
brine stream to allow the brine to be discharged to a POTW. Water
systems should check with the POTW to ensure that the brine stream will
be accepted before selecting this option.
Activated Alumina (AA) is an effective arsenic removal technology;
however, the capacity of activated alumina to remove arsenic is very pH
sensitive. High removals can be achieved over a broad range of pH, but
shorter run lengths will be observed at higher pH. Activated alumina
can be operated in one of two ways. The activated alumina can either be
disposed of or regenerated after the media is exhausted. Under the
regeneration option, strong acids and bases are used to remove arsenic
from the media so that it can be used again to remove arsenic. Because
arsenic is strongly adsorbed to the media, only about 50-70% of the
adsorbed arsenic is removed. The brine stream produced by the
regeneration process then requires disposal. The proposed rule listed
discharge to a sanitary sewer as the disposal mechanism for the brines.
Many comments on the proposed rule noted that TBLLs for arsenic or
total dissolved solids might restrict discharge of brine streams to the
sanitary sewer. Since activated alumina run lengths (i.e., number of
bed volumes (BV) per run) are much longer than anion exchange, the
arsenic concentrations in the brine stream would likely be much higher.
Regeneration of activated alumina media is not recommended for larger
systems because: (1) Disposal of the brine may be difficult, (2) the
regeneration process is incomplete which reduces subsequent run
lengths, and (3) for most systems it will be cheaper to replace the
media rather than regenerate it. The option of replacing the spent
media with new media is called disposable activated alumina.
The disposable activated alumina option can be operated both at the
optimal pH of 6 and at higher natural water pH values. It is expected
that larger systems would adjust pH to take advantage of the longer run
lengths. EPA developed several disposable activated alumina options for
the final rule. Two options were based on operating the process at the
natural pH of the water (no pH adjustment). These options are intended
primarily for smaller systems, although larger systems may also be able
to operate at the natural pH if it is low enough to get sufficiently
long run lengths. Two options where the pH was adjusted to pH 6 were
also examined. The longer run length is based on using sulfuric acid to
lower the pH. However, sulfate can compete for adsorption sites with
arsenic. It was recommended that hydrochloric acid be used to obtain a
longer run length (Clifford et al., 1998). When pH is adjusted to pH 6,
post-treatment corrosion control will be necessary.
In our analysis, we assumed that spent media could be safely
disposed of in a non-hazardous landfill. The preamble to the proposed
rule described results from testing of activated alumina media used to
remove arsenic in drinking water systems with arsenic above 50
g/L. The results from the Toxicity Characteristic Leaching
Procedure (TCLP) on these samples was typically less that 50
g/L. The current toxicity characteristic (TC) regulatory level
for designating arsenic as a hazardous waste under the Resource
Conservation and Recovery Act (RCRA) is 5 mg/L (5000 g/L) and
is listed in 40 CFR 261.24(a). The TC regulatory level is one hundred
times higher than the results from the activated alumina samples.
Reverse Osmosis (RO) can provide removal efficiencies of greater
than 95% when operating pressure is ideal. Water rejection (on the
order of 20-25%) may be an issue in water-scarce regions and may prompt
systems employing RO to seek greater levels of water recovery. Water
recovery is the volume of drinking water produced by the process
divided by the influent stream (product water/influent stream).
Increased water recovery is often more expensive, since it can involve
recycling of water through treatment units to allow more efficient
separation of solids from water. This can also produce more
concentrated solid wastes. However, the waste stream will generally not
be as concentrated as anion exchange brines, so it should be easier to
dispose of. Based on the cost of the process, it is unlikely that
reverse osmosis would be installed solely for arsenic removal. Blending
a treated portion with an untreated portion and
[[Page 6983]]
still meeting the MCL would make reverse osmosis more cost effective.
If blending is not an option, post-treatment corrosion control would be
necessary. Since a large portion of the water is wasted, water quantity
could be an issue, especially in the Western U.S. It should be noted
that while reverse osmosis is listed as a BAT, it was not used to
develop national costs because other options are more cost effective
and have much smaller waste streams.
Modified Coagulation/Filtration (C/F) is an effective treatment
process for removal of As (V) according to laboratory, pilot-plant, and
full-scale tests. The type of coagulant and dosage used affects the
efficiency of the process. Below a pH of approximately 7, removals with
alum or ferric sulfate/chloride are similar. Above a pH of 7, removals
with alum decrease dramatically (at a pH of 7.8, alum removal
efficiency is about 40%). Other coagulants are also less effective than
ferric sulfate/chloride. Systems may need to lower pH or add more
coagulant to achieve higher removals.
Modified Lime Softening (LS), operated within the optimum pH range
of greater than 10.5 is likely to provide a high percentage of As
removal. Systems operating lime softening at lower pH will need to
increase the pH to achieve higher removals of arsenic.
Coagulation/Filtration and Lime Softening are unlikely to be
installed solely for arsenic removal. Systems considering installation
of one of these technologies should design the process to operate in
the optimal pH range if high removal efficiencies are needed for
compliance.
Electrodialysis Reversal (EDR) can produce effluent water quality
comparable to reverse osmosis. EDR systems are fully automated, require
little operator attention, and do not require chemical addition. EDR
systems, however, are typically more expensive than nanofiltration and
reverse osmosis systems. These systems are often used in treating
brackish water to make it suitable for drinking. This technology has
also been applied in the industry for wastewater recovery and typically
operates at a recovery of 70 to 80%. Since a large portion of the water
is wasted, water quantity could be an issue, especially in the Western
U.S. It should be noted that while electrodialysis reversal is listed
as a BAT, it was not used to develop national costs because other
options are more cost effective and have much smaller waste streams.
Oxidation/Filtration (including greensand filtration) has an
advantage in that there is not as much competition with other ions.
Arsenic is co-precipitated with the iron during iron removal.
Sufficient iron needs to be present to achieve high arsenic removals.
One study recommended a 20:1 iron to arsenic ratio (Subramanian et al.,
1997). Removals of approximately 80% were achieved when iron to arsenic
ratio was 20:1. When the iron to arsenic ratio was lower (7:1),
removals decreased below 50%. The presence of iron in the source water
is critical for arsenic removal. If the source water does not contain
iron, oxidizing and filtering the water will not remove arsenic. When
the arsenic is present as As(III), sufficient contact time needs to be
provided to convert the As(III) to As(V) for removal by the oxidation/
filtration process. An additional pre-oxidation step is not required
for this process as long as there is sufficient contact time. In
developing national cost estimates, EPA assumed that systems would opt
for this type of technology only if more than 300 g/L of iron
was present. The Agency assumed a removal percentage of 50% when
estimating national costs because the 20:1 ratio could not be verified
due to limitations in the co-occurrence database. However, EPA assumed
a removal percentage of 80% as part of a sensitivity analysis. At
proposal EPA indicated that oxidation filtration was not being listed
as BAT because it has a low removal efficiency, which might not be
appropriate for an MCL of 5. However, the Agency also noted that this
technology may be appropriate for systems that do not require high
arsenic removal and had high iron in their source water. Because this
is an inexpensive technology that is particularly effective for high-
iron, low-arsenic waters, EPA is listing oxidation/filtration as a BAT
with a footnote that the iron-to-arsenic ratio must be at least 20:1.
Systems with greater than 300 g/L of iron will also see
benefits in the aesthetic quality of the water as the iron can be
reduced below the secondary standard. EPA's inclusion of oxidation/
filtration as a BAT in today's final rule is based upon further
evaluation of all available information and studies as well as on
public comments.
4. Other technologies evaluated, but not designated as BAT
Coagulation Assisted Microfiltration. The coagulation process
described previously can be linked with microfiltration to remove
arsenic. The microfiltration step essentially takes the place of a
conventional gravity filter. The University of Houston recently
completed pilot studies at Albuquerque, New Mexico on iron coagulation
followed by a direct microfiltration system. The results of this study
indicated that iron coagulation followed by microfiltration is capable
of removing arsenic (V) from water to yield concentrations that are
consistently below 2 g/L. Critical operating parameters are
iron dose, mixing energy, detention time, and pH (Clifford, 1997).
Coagulation and microfiltration as separate processes have both been
installed full scale, but the combined coagulation/microfiltration
process does not have a full-scale operation history. Since a full-
scale operation history is one of the requirements to list a technology
as a BAT, it is not presently being listed as one. It could be
designated as such in the future if the technology meets that
requirement. EPA used this option in developing the national cost
estimate because we believe coagulation/microfiltration is an
appropriate technology that will be used by certain water systems to
comply with this rule, even though it is not currently listed as BAT
for the reasons mentioned.
Granular ferric hydroxide is a technology that may combine very
long run length without the need to adjust pH. The technology has been
demonstrated for arsenic removal full scale in England (Simms et al.,
2000). A pilot-scale study for activated alumina was also conducted on
that water and showed run lengths much longer than observed in pilot-
scale studies in the United States. Due to the lack of published data
showing performance for a range of water qualities, granular ferric
hydroxide was not designated a BAT. In addition, there is little
published information on the cost of the media, so it is difficult to
evaluate cost. Granular ferric hydroxide is being investigated in
several ongoing studies and may be an effective technology for removing
arsenic. Systems may wish to investigate it and other adsorption
technologies such as modified activated alumina and other iron-based
media. Many of these other new adsorptive media are also being
investigated in several ongoing studies.
5. Waste disposal
Waste disposal will be an important issue for both large and small
drinking water plants. Costs for waste disposal have been added to the
costs of the treatment technologies (in addition to any pre-oxidation
and corrosion control costs), and form part of the treatment trains
that are listed in Tables I.G-1, I.G-5, and I.G-6.
The preamble to the proposed rule summarized toxicity
characteristic leaching procedure (TCLP) data on residuals from
different arsenic removal
[[Page 6984]]
technologies. The arsenic concentrations in TCLP extracts from alum
coagulation, activated alumina, lime softening, iron/manganese removal,
and coagulation-microfiltration residuals were below 0.05 mg/L, which
is two orders of magnitude lower than the current TC regulatory level.
The TCLP data for iron coagulation were mixed--the residuals from an
arsenic removal plant were below 0.05 mg/L, but the residuals from
another iron coagulation plant were above 1 mg/L. However, this is
still below the TC regulatory level of 5 mg/L. Based on these data, EPA
does not believe that drinking water treatment plant residuals would be
classified as hazardous waste. The TCLP data also indicate that most
residuals could meet a much lower TC regulatory level. Options where
the brine stream could be hazardous were eliminated from the final
decision tree. For the purposes of the national cost estimate, it was
assumed that solid residuals would be disposed of at nonhazardous
landfills.
G. Treatment Trains Considered For Small Systems
1. Can my water system use point-of-use (POU), point-of-entry (POE), or
bottled water to comply with this regulation?
Section 1412(b)(4)(E)(ii) of SDWA, as amended in 1996, requires EPA
to issue a list of technologies that achieve compliance with MCLs
established under the Act that are affordable and applicable to typical
small drinking water systems. These small public water systems
categories are: (1) population of more than 25 but less than or equal
to 500; (2) population of more than 500, but less than or equal to
3,300; and (3) population of more than 3,300, but less than or equal to
10,000. Owners and operators may choose any technology or technique
that best suits their conditions, as long as the MCL is met.
The technologies examined for BAT determinations were also
evaluated as small system compliance technologies. Several other
alternatives that are solely small system options were also evaluated
as compliance technologies. Central treatment is not the only option
available to small systems. One of the provisions included in the SDWA
Amendments of 1996 allows the use of POU and POE devices as compliance
technologies for small systems. SDWA stipulates that POU/POE treatment
systems:
shall be owned, controlled and maintained by the public water system
or by a person under contract with the public water system to ensure
proper operation and maintenance and compliance with the MCL or
treatment technique and equipped with mechanical warnings to ensure
that customers are automatically notified of operational problems
(Sec. 1412(b)(4)(E)).
Whole-house, or POE treatment, is necessary when exposure to the
contaminant by modes other than consumption is a concern; this is not
the case with arsenic. Single faucet, or POU treatment, is preferred
when treated water is needed only for drinking and cooking purposes.
POU devices are especially applicable for systems that have a large
flow and only a minor part of that flow directed for potable use such
as at many NTNCWSs. POE/POU options include reverse osmosis, activated
alumina, and ion exchange processes. POU systems are easily installed
and can be easily operated and maintained. In addition, these systems
generally offer lower capital costs and may reduce engineering, legal,
and other fees associated with centralized treatment options. However,
there will be higher administrative costs associated with POU and POE
options. For POU options, the trade-off is lower treatment cost since
only 1% of the water is treated, but higher administrative and
monitoring costs occur. Centrally managed POU options, even with the
higher monitoring and administrative costs, are less expensive than
central treatment for populations up to 150 to 250 people depending
upon the technology and number of households.
Using POU/POE devices introduces some new issues. Adopting a POU/
POE treatment system in a small community requires more record-keeping
to monitor individual devices than does central treatment. POU/POE
systems may require special regulations regarding customer
responsibilities as well as water utility responsibilities. The water
system or person under contract to the system is responsible for
maintaining the devices in customers' homes. This responsibility cannot
be delegated to the customer. Use of POU/POE systems does not reduce
the need for a well-maintained water distribution system. Increased
monitoring may be necessary to ensure that the treatment units are
operating properly. Monitoring POU/POE systems is also more complex
because compliance samples need to be taken after each POU or POE unit
rather than at the entry point to the distribution system to be
reflective of treatment.
EPA examined three technologies as POU and POE devices for the
proposed rule. EPA assumed that systems would more likely choose to use
POU activated alumina (AA) or reverse osmosis (RO), and POE AA in the
proposed rule. POU and POE ion exchange (IX) and POE RO were
considered, but not included as compliance technologies in the proposed
rule. Activated alumina and ion exchange units face a breakthrough
issue. If the activated alumina is not replaced on time, there is a
potential for significantly reduced arsenic removal. However, if the
anion exchange resin is not replaced or regenerated on time, the
previously removed arsenic can be driven off the resin by sulfate. Tap
water arsenic concentrations can be higher than the source water. This
is called chromatographic peaking. Due to the potential for
chromatographic peaking and run lengths that would typically be less
than six months, anion exchange was not listed as a compliance
technology in the proposed rule. POE ion exchange also may present
problems with total dissolved solids since the resin would need to be
regenerated. Since all sites within the system would need treatment,
the total dissolved solids increase from a centrally managed POE ion
exchange system would be similar to that from a central treatment ion
exchange system. EPA did not list POE RO units as compliance
technologies because it could create corrosion control problems. In
addition, water recovery would be no higher than central treatment, so
water quantity issues associated with central treatment reverse osmosis
would be applicable to POE RO.
The proposed rule included POE AA as a small system compliance
technology. Arsenic removal by AA is very sensitive to the pH. The
finished water pH will typically be higher than the optimal pH of 6 to
meet the corrosion control requirements of the lead and copper rule. A
finished water pH for many systems would be in the range of pH 7 to pH
8. Using data on activated alumina run length and pH, it was determined
that viable run lengths were likely only when the finished water pH was
at or below pH 7.5 (Kempic, 2000). Even in this pH range, the media may
need to be replaced more frequently than once a year, which would make
the option very expensive especially compared to the POU AA option. The
run length data used for this analysis were from a site with very
little competing ions (Simms and Azizian, 1997). Studies at other sites
with higher levels of competing ions have much lower run lengths
(Clifford et al., 1998). Based on the limited finished water pH range
where POE AA might be effective and the fact that the POU media needs
replacing much less frequently due to lower water demand, POE AA has
not been listed as a compliance technology
[[Page 6985]]
in the final rule. POE devices utilizing media that are less sensitive
to pH adjustment may be listed as compliance technologies in the future
once data on their performance are generated.
The effect of pH was also examined on POU AA. Under the POU AA
option, the volume of water requiring treatment is much smaller. The
unit will be installed at the kitchen tap and only the water being used
for cooking and consumption is being treated for arsenic removal. Since
the ratio of the daily volume of water being treated to the size of the
unit is much smaller, POU units can be operated for longer periods of
time before the media needs to be replaced. The replacement frequency
assumed for the costs is every six months. Viable run lengths for the
POU option were greater than one year up to pH 8 (Kempic, 2000). This
analysis assumed a large daily usage volume of 24 liters per day. The
average consumption per person per day is just over 1 liter. Even if
competing ions reduced the run length significantly, systems with tap
water at or below pH 8 should meet the MCL of 10 g/L using a
six-month replacement frequency for the media. POU AA is a compliance
technology when the tap water pH is at or below pH 8.
POU RO was listed as a compliance technology in the proposed rule
and it is being listed as a compliance technology in the final rule as
well. Several comments indicated that water rejection would be an issue
with POU devices. Since only about 1% of the total water used in the
household is being treated, POU RO is unlikely to create water quantity
problems. If the water rejection rate was 10:1, this would only
increase the total household water demand by about 10 percent. Where
availability of additional water is limited, systems may want to
consider other alternatives to meet the MCL.
In order to be consistent with 1996 SDWA Amendments, EPA issued a
Federal Register notice on June 11, 1998 (EPA, 1998f) that deleted the
prohibition on the use of POU devices as compliance technologies. This
prohibition was in 40 CFR 141.101. This section now states that public
water systems shall not use bottled water to achieve compliance with an
MCL. Bottled water may be used on a temporary basis to avoid
unreasonable risk to health. Therefore, bottled water cannot be used as
a compliance technology for the arsenic rule.
Likely treatment trains are shown in Table I.G-1. These trains
represent a wide variety of solutions, including BATs, that small
systems may consider when complying with the proposed arsenic MCL. Not
all solutions may be viable for a given system. For example, only those
systems with coagulation/filtration in place will be able to modify
their existing treatment system. The treatment trains include BATs,
waste disposal, and when necessary, pre-oxidation and corrosion
control. While systems could install lime softening at pH > 10.5 or
optimized coagulation/filtration solely for arsenic removal, EPA does
not view this as a likely option. Reverse osmosis and electrodialysis
reversal are also not included in this table because other options are
more cost effective for arsenic removal and do not reject a large
volume of water like these two technologies. RO and EDR may be cost-
effective options if removal of other contaminants is needed and water
quantity is not a concern.
Table I.G-1.-- Treatment Technology Trains for Consideration by Small
Systems in Complying With Final Rule Including BATs
------------------------------------------------------------------------
Treatment Technology Trains for
Train # Consideration by Small Systems
------------------------------------------------------------------------
1............................ Add pre-oxidation [if not in-place] and
modify in-place Lime Softening (pH >
10.5) and modify corrosion control.
2............................ Add pre-oxidation [if not in-place] and
modify in-place Coagulation/Filtration
and modify corrosion control.
3............................ Add pre-oxidation [if not in-place] and
add Anion Exchange and add POTW waste
disposal. Sulfate level 20
mg/L.
4............................ Add pre-oxidation [if not in-place] and
add Anion Exchange and add POTW waste
disposal. Sulfate level: 20 mg/L
sulfate 50 mg/L.
5............................ Add pre-oxidation [if not in-place] and
add Coagulation Assisted Microfiltration
with corrosion control and add
mechanical dewatering/non-hazardous
landfill waste disposal.
6............................ Add pre-oxidation [if not in-place] and
add Coagulation Assisted Microfiltration
with corrosion control and add non-
mechanical dewatering/non-hazardous
landfill waste disposal.
7............................ Add Oxidation/Filtration (Greensand)
(20:1 iron: arsenic) and add POTW for
backwash stream.
8............................ Add pre-oxidation [if not in-place] and
add Activated Alumina and add non-
hazardous landfill (for spent media)
waste disposal. pH 7 pH pH
8.
9............................ Add pre-oxidation [if not in-place] and
add Activated Alumina and add non-
hazardous landfill (for spent media)
waste disposal. pH 8 pH pH 8.3.
10........................... Add pre-oxidation [if not in-place] and
add Activated Alumina with pH adjustment
(to pH 6) and corrosion control and add
non-hazardous landfill (for spent media)
waste disposal. Run length = 23,100 BV.
11........................... Add pre-oxidation [if not in-place] and
add Activated Alumina with pH adjustment
(to pH 6) and corrosion control and add
non-hazardous landfill (for spent media)
waste disposal. Run length = 15,400 BV.
12........................... Add pre-oxidation [if not in-place] and
add POU Reverse Osmosis.
13........................... Add pre-oxidation [if not in-place] and
add POU Activated Alumina. (Finished
water pH pH 8.0)
------------------------------------------------------------------------
Pre-oxidation costs are given as a separate component because they
will be incurred only by some systems. In estimating national costs, it
was assumed that only systems without pre-oxidation in place would need
to add the necessary equipment. It is expected that no surface water
systems will need to install pre-oxidation for arsenic removal and that
fewer than 50% of the ground water systems may need to install pre-
oxidation for arsenic removal. Ground water systems without pre-
oxidation should ascertain if pre-oxidation is necessary by determining
if the arsenic is present as As (III) or As (V). Ground water systems
with predominantly As (V) will probably not need pre-oxidation to meet
the MCL.
2. What are the affordable treatment technologies for small systems?
The 13 treatment trains listed in Table I.G-1 were compared against
the national-level affordability criteria to determine the affordable
treatment trains. The Agency's national-level affordability criteria
were published in the August 6, 1998 Federal Register (EPA, 1998h). In
this notice, EPA discussed the procedure for affordable
[[Page 6986]]
treatment technology determinations for the contaminants regulated
before 1996.
The preamble to the proposed arsenic rule described the derivation
of the national-level affordability criteria (65 FR 38888 at 38926;
EPA, 2000i). A very brief summary follows: First an ``affordability
threshold'' (i.e., the total annual household water bill that would be
considered affordable) was calculated. The total annual water bill
includes costs associated with water treatment, water distribution, and
operation of the water system. In developing the threshold of 2.5%
median household income, EPA considered the percentage of median
household income spent by an average household on comparable goods and
services and on cost comparisons with other risk reduction activities
for drinking water such as households purchasing bottled water or a
home treatment device. The complete rationale for EPA's selection of
2.5% as the affordability threshold is described in ``Variance
Technology Findings for Contaminants Regulated Before 1996'' (EPA,
1998l).
The Variance Technology Findings document also describes the
derivation of the baselines for median household income, annual water
bills, and annual household consumption. Data from the Community Water
System Survey (CWSS) were used to derive the annual water bills and
annual water consumption values for each of the three small system size
categories. The Community Water System Survey data on zip codes were
used with the 1990 Census data on median household income to develop
the median household income values for each of the three small-system
size categories. The median household-income values used for the
affordable technology determinations are not based on the national
median income. The value for each size category is a national median
income for communities served by small water systems within that range.
Table I.G-2 presents the baseline values for each of the three small-
system size categories. Annual water bills and median household income
are based on 1995 estimates.
Table I.G-2.--Baseline Values for Small Systems Categories
----------------------------------------------------------------------------------------------------------------
Annual household
System size category (population consumption (1000 Annual water bills ($/ Median household income
served) gallons/yr) yr) ($)
----------------------------------------------------------------------------------------------------------------
25-500............................... 72 $211 $30,785
501-3,300............................ 74 184 27,058
3,300-10,000......................... 77 181 27,641
----------------------------------------------------------------------------------------------------------------
For each size category, the threshold value was determined by
multiplying the median household income by 2.5%. The annual household
water bills were subtracted from this value to obtain the available
expenditure margin. Projected treatment costs will be compared against
the available expenditure margin to determine if there are affordable
compliance technologies for each size category. The available
expenditure margin for the three size categories is presented in Table
I.G-3.
Table I.G-3.--Available Expenditure Margin for Affordable Technology
Determinations
------------------------------------------------------------------------
Available expenditure
System size category (population served) margin ($/household/
year)
-----------------------------------------------------------------------
25-500....................................... 559
501-3,300.................................... 492
3,301-10,000................................. 510
------------------------------------------------------------------------
The size categories specified in SDWA for affordable technology
determinations are different than the size categories typically used by
EPA in the Economic Analysis. A weighted average procedure was used to
derive design and average flows for the 25-500 category using design
and average flows from the 25-100 and 101-500 categories. A similar
approach was used to derive design and average flows from the 501-1000
and 1001-3300 categories for the 501-3300 category. The Variance
Technology Findings document (EPA, 1998l) describes this procedure in
more detail. Table I.G-4 lists the design and average flows for the
three size categories.
Table I.G-4.-- Design and Average Daily Flows Used for Affordable
Technology Determinations
------------------------------------------------------------------------
System size category (population Design flow Average flow
served) (mgd) (mgd)
------------------------------------------------------------------------
25-500............................ 0.058 0.015
501-3,300......................... 0.50 0.17
3,301-10,000...................... 1.8 0.70
------------------------------------------------------------------------
Capital and operating and maintenance costs were derived for each
treatment train using the flows listed previously and the cost
equations in the Technology and Cost Document. Several conservative
assumptions were made to derive the costs. The influent arsenic
concentration was assumed to be 50 g/L, which was the MCL for
arsenic prior to this rule. The treatment target was 8 g/L,
which is 80% of the MCL. Thus, little blending could be performed to
reduce costs. Capital costs were amortized using the 7% interest rate
preferred by OMB for benefit-cost analyses of government programs and
regulations rather than a 3% interest rate.
The annual system treatment cost in dollars per year was converted
into a rate increase using the average daily flow. The annual water
consumption values listed in Table I.G-2 were multiplied by 1.15 to
account for water lost due to leaks. Since the water lost to leaks is
not billed, the water bills for the actual water used were adjusted to
cover this lost water by increasing the household consumption. The rate
increase in dollars per thousand gallons used was multiplied by the
adjusted annual consumption to determine the annual cost increase for
the household for each treatment train. Several comments on
affordability presented household cost increases that were
[[Page 6987]]
derived by dividing the annual system cost by the number of households.
That is an inappropriate method because residential customers would not
only be paying for the water that they use, but also all the water used
by non-residential customers of the system..
Of the 13 treatment trains in Table I.G-1, the ones identified in
Table I.G-5 are deemed to be affordable for systems serving 25-500
people as the annual household cost was below the available expenditure
margin. The two trains using coagulation-assisted microfiltration are
not affordable for this size category. All 13 treatment trains are
deemed to be affordable for systems serving 501-3,300 and 3,301-10,000
people and are presented in Table I.G-6. Centralized compliance
treatment technologies include ion exchange, activated alumina,
modified coagulation/filtration, modified lime softening, and
oxidation/filtration (e.g. greensand filtration) for source waters high
in iron. In addition, POU and POE devices are also compliance
technology options for the smaller systems.
Table I.G-5.-- Affordable Compliance Technology Trains for Small Systems
With Population 25-500
------------------------------------------------------------------------
Train No. Treatment Technology Trains
------------------------------------------------------------------------
1............................ Add pre-oxidation [if not in-place] and
modify in-place Lime Softening (pH >
10.5) and modify corrosion control.
2............................ Add pre-oxidation [if not in-place] and
modify in-place Coagulation/Filtration
and modify corrosion control.
3............................ Add pre-oxidation [if not in-place] and
add Anion Exchange and add POTW waste
disposal. Sulfate level 20
mg/L.
4............................ Add pre-oxidation [if not in-place] and
add Anion Exchange and add POTW waste
disposal. Sulfate level: 20 mg/L
sulfate 50 mg/l.
7............................ Add Oxidation/Filtration (Greensand)
(20:1 iron: arsenic) and add POTW for
backwash stream.
8............................ Add pre-oxidation [if not in-place] and
add Activated Alumina and add non-
hazardous landfill (for spent media)
waste disposal. pH 7 pH pH
8.
9............................ Add pre-oxidation [if not in-place] and
add Activated Alumina and add non-
hazardous landfill (for spent media)
waste disposal. pH 8 pH pH 8.3.
10........................... Add pre-oxidation [if not in-place] and
add Activated Alumina with pH adjustment
(to pH 6) and corrosion control and add
non-hazardous landfill (for spent media)
waste disposal. Run length = 23,100 BV.
11........................... Add pre-oxidation [if not in-place] and
add Activated Alumina with pH adjustment
(to pH 6) and corrosion control and add
non-hazardous landfill (for spent media)
waste disposal. Run length = 15,400 BV.
12........................... Add pre-oxidation [if not in-place] and
add POU Reverse Osmosis.
13........................... Add pre-oxidation [if not in-place] and
add POU Activated Alumina. (Finished
water pH pH 8.0)
------------------------------------------------------------------------
Table I.G-6.-- Affordable Compliance Technology Trains for Small Systems
With Populations 501-3,300 and 3,301 to 10,000
------------------------------------------------------------------------
Train No. Treatment Technology Trains
------------------------------------------------------------------------
1............................ Add pre-oxidation [if not in-place] and
modify in-place Lime Softening (pH >
10.5) and modify corrosion control.
2............................ Add pre-oxidation [if not in-place] and
modify in-place Coagulation/Filtration
and modify corrosion control.
3............................ Add pre-oxidation [if not in-place] and
add Anion Exchange and add POTW waste
disposal. Sulfate level 20
mg/L.
4............................ Add pre-oxidation [if not in-place] and
add Anion Exchange and add POTW waste
disposal. Sulfate level: 20 mg/L
sulfate 50 mg/l.
5............................ Add pre-oxidation [if not in-place] and
add Coagulation Assisted Microfiltration
with corrosion control and add
mechanical dewatering/non-hazardous
landfill waste disposal.
6............................ Add pre-oxidation [if not in-place] and
add Coagulation Assisted Microfiltration
with corrosion control and add non-
mechanical dewatering/non-hazardous
landfill waste disposal.
7............................ Add Oxidation/Filtration (Greensand)
(20:1 iron: arsenic) and add POTW for
backwash stream.
8............................ Add pre-oxidation [if not in-place] and
add Activated Alumina and add non-
hazardous landfill (for spent media)
waste disposal. pH 7 pH pH
8.
9............................ Add pre-oxidation [if not in-place] and
add Activated Alumina and add non-
hazardous landfill (for spent media)
waste disposal. pH 8 pH pH 8.3.
10........................... Add pre-oxidation [if not in-place] and
add Activated Alumina with pH adjustment
(to pH 6) and corrosion control and add
non-hazardous landfill (for spent media)
waste disposal. Run length = 23,100 BV.
11........................... Add pre-oxidation [if not in-place] and
add Activated Alumina with pH adjustment
(to pH 6) and corrosion control and add
non-hazardous landfill (for spent media)
waste disposal. Run length = 15,400 BV.
12........................... Add pre-oxidation [if not in-place] and
add POU Reverse Osmosis.
13........................... Add pre-oxidation [if not in-place] and
add POU Activated Alumina. (Finished
water pH pH 8.0)
------------------------------------------------------------------------
3. Can My Water System Get a Small System Variance From an MCL Under
Today's Rule?
Section 1415(e)(1) of SDWA allows States to grant variances to
small water systems (i.e., systems having 10,000 customers or less) in
lieu of complying with an MCL if EPA determines that there are no
nationally affordable compliance technologies for that system size/
water quality combination. The system must then install an EPA-listed
variance treatment technology (section 1412(b)(15)) that makes progress
toward the MCL, if not necessarily reaching it. EPA has determined that
affordable technologies exist for all three system size categories and
has therefore not identified a variance technology for any system size
or source water quality combination. Small system variances are not
available for the final arsenic MCL.
H. Can My System Get a General Variance or Exemption From the MCL Under
Today's Rule?
General variances may be granted in accordance with section
1415(a)(1)(A) of SDWA and EPA's regulations. General variances are
available to public water systems that have installed or agree to
install the BAT but, due to source water quality, are or will be unable
to comply with the national primary drinking water standard. The
general variance
[[Page 6988]]
provisions of SDWA are narrowly focused on addressing those rare
circumstances where some unusual characteristic of the source water
available to a system will result in less effective performance of the
BAT. Exemptions may be granted in accordance with section 1416(a) of
SDWA and EPA's regulations. Exemptions are designed to provide a system
facing compelling circumstances, such as economic hardship, additional
time to come into compliance.
Under section 1415(a)(1)(A) of the SDWA, a State that has primary
enforcement responsibility (primacy), or EPA as the primacy agency, may
grant variances from MCLs to those public water systems of any size
that cannot comply with the MCLs because of characteristics of the
water sources. The primacy agency may grant general variances to a
system on condition that the system install the best available
technology, treatment techniques, or other means, and provided that
alternative sources of water are not reasonably available to the
system. At the time this type of variance is granted, the State must
prescribe a schedule for compliance with its terms and may require the
system to implement additional control measures. Furthermore, before
EPA or the State may grant a general variance, it must find that the
variance will not result in an unreasonable risk to health (URTH) to
the public served by the public water system.
Under section 1413(a)(4), States that choose to issue general
variances must do so under conditions, and in a manner, that are no
less stringent than section 1415. Of course, a State may adopt
standards that are more stringent than the EPA's standards. EPA
specifies BATs for general variance purposes. EPA may identify as BAT
different treatments under section 1415 for variances other than the
BAT under section 1412 for MCLs. The BAT findings for section 1415 may
vary depending on a number of factors, including the number of persons
served by the public water system, physical conditions related to
engineering feasibility, and the costs of compliance with MCLs. In this
final rule, EPA is not specifying different BAT for variances under
section 1415(a).
Under section 1416(a), EPA or a State may exempt a public water
system from any requirements related to an MCL or treatment technique
of an NPDWR if it finds that: (1) Due to compelling factors (which may
include a variety of ``compelling'' factors, including economic factors
such as qualification of the PWS as serving a disadvantaged community),
the PWS is unable to comply with the requirement or implement measure
to develop an alternative source of water supply; (2) the exemption
will not result in an URTH; (3) the PWS was in operation on the
effective date of the NPWDR, or for a system that was not in operation
by that date, only if no reasonable alternative source of drinking
water is available to the new system; and (4) management or
restructuring changes (or both) cannot reasonably result in compliance
with the Act or improve the quality of drinking water.
If EPA or the State grants an exemption to a public water system,
it must at the same time prescribe a schedule for compliance (including
increments of progress or measures to develop an alternative source of
water supply) and implementation of appropriate control measures that
the State requires the system to meet while the exemption is in effect.
Under section 1416(b)(2)(A), the schedule prescribed shall require
compliance as expeditiously as practicable (to be determined by the
State), but no later than 3 years after the compliance date for the
regulations established pursuant to section 1412(b)(10). For public
water systems serving 3,300 people or less and needing financial
assistance for the necessary improvements, EPA or the State may renew
an exemption for one or more additional two-year periods, but not to
exceed a total of six years, if the system establishes that it is
taking all practicable steps to meet certain requirements specified in
the statute. Thus, the maximum possible duration of a small systems
exemption is nine years beyond the 5-year compliance schedule specified
in today's rule.
A public water system shall not be granted an exemption unless it
can establish that either: (1) The system cannot meet the standard
without capital improvements that cannot be completed prior to the date
established pursuant to section 1412(b)(10); (2) in the case of a
system that needs financial assistance for the necessary
implementation, the system has entered into an agreement to obtain
financial assistance pursuant to section 1452 or any other Federal or
State program; or (3) the system has entered into an enforceable
agreement to become part of a regional public water system.
EPA believes that exemptions will be an important tool to help
States address the number of systems needing financial assistance to
achieve compliance with the arsenic rule (and other rules) with the
available supply of financial assistance. About 2,300 CWSs and about
1,100 NTNCWSs will need to install treatment to achieve compliance with
today's final rule. CWSs and not-for-profit NTNCWSs are eligible for
assistance from the Drinking Water State Revolving Fund (DWSRF).
Between its inception in Federal Fiscal Year 1997 and June 2000, the
DWSRF program has provided assistance to about 1,100 systems. Given the
many competing demands being placed on financial assistance programs,
the ability to extend the period of time available for a system to
receive financial assistance will provide important flexibility for
States and systems. Exemptions provide an opportunity to extend the
period of time during which a system can achieve compliance, thus
providing needy systems with additional time to qualify for financial
assistance. Under today's action, all systems have 5 years to achieve
compliance. Exemptions for an additional 3 years can be made available
to qualified systems. For those qualified systems serving 3,300 persons
or less, up to 3 additional 2-year extensions to the exemption are
possible, for a total exemption duration of 9 years. When added to the
5 years provided for compliance by the rule, this allows up to 14 years
for small systems serving up to 3,300 people to achieve compliance.
EPA will issue guidance in the near future on considerations
involved in granting exemptions under the arsenic rule, including
making findings of no URTH where exemptions are offered.
I. What Analytical Methods are Approved for Compliance Monitoring of
Arsenic and What are the Performance Testing Criteria for Laboratory
Certification?
1. Approved Analytical Methods
Today's rule lists four analytical technologies that are approved
for compliance determinations of arsenic at the MCL of 0.01 mg/L (see
Table I.I-1). As noted in the June 22, 2000 proposed rule (65 FR 38888,
EPA, 2000i), the methods listed in Table I.I-1 are the same analytical
technologies that were approved for arsenic when the MCL was 0.05 mg/L,
with the exception of the methods that use Inductively Coupled Plasma
Atomic Emission Spectroscopy (ICP-AES) measurement technology. EPA is
withdrawing two ICP-AES methods (EPA Method 200.7 and SM 3120B) because
their detection limits (0.008 mg/L and 0.050 mg/L respectively) are too
high to reliably determine compliance with an MCL of 0.01 mg/L. In the
June 2000 proposed rule, EPA noted that the ICP-AES methods were rarely
used to obtain laboratory certification when analyzing
[[Page 6989]]
low level challenge samples for arsenic. Therefore, we believe
withdrawal of the availability of the ICP-AES methods for compliance
determinations of arsenic in drinking water will not affect laboratory
capacity. EPA did not receive any adverse comment on the proposal to
withdraw approval of these two methods, and today's final rule amends
the CFR to effect this withdrawal.
Table I.I-1.--Approved Analytical Methods (40 CFR 141.23) for Arsenic at
the MCL of 0.01 mg/L
------------------------------------------------------------------------
Methodology Reference method
------------------------------------------------------------------------
Inductively Coupled Plasma Mass 200.8 (EPA)
Spectroscopy (ICP-MS).
Stabilized Temperature Platform 200.9 (EPA)
Graphite Furnace Atomic Absorption
(STP-GFAA).
Graphite Furnace Atomic Absorption 3113B (SM) D-2972-93C (ASTM)
(GFAA).
Gaseous Hydride Atomic Absorption 3114B (SM) D-2972-93B (ASTM)
(GHAA).
------------------------------------------------------------------------
2. Performance Testing Criteria for Laboratory Certification
For purposes of drinking water laboratory certification, the Agency
specifies pass/fail (acceptance) limits for a successful analysis of
the required annual challenge sample, i.e., a performance evaluation
(PE) or performance testing (PT) sample. These acceptance limits have
been historically derived using one of two different approaches:
(a) Variable acceptance limits uniquely derived for each PE
study from a regression analysis of the performance of all
laboratories that participate in that PE-study, or
(b) Fixed acceptance limits derived from a regression analysis
of the laboratory PE sample analysis results in several PE studies.
Variable acceptance limits are analogous to ``grading on a curve''
which means that the pass/fail limit can vary from PE study to study
depending on the quality and experience of the laboratories
participating in the study. These limits are specified in the CFR as
plus or minus two sigma (2 ) where sigma is the standard deviation of
the analytical results reported in the PE study. EPA specifies variable
acceptance limits when a method or measurement technology is new enough
that an insufficient number of experienced laboratories have
participated in the PE studies or when only a few PE studies have been
conducted.
EPA prefers the fixed acceptance limits approach because it is the
better indicator of laboratory performance averaged over time and
several different concentrations of the target analyte. Fixed limits
also provide the same pass/fail benchmark in each PE study. As
discussed in the proposed rule, EPA has a large base of PE-study data
from which to derive a practical quantitation limit (PQL) and a fixed
PE-study acceptance limit for arsenic. Thus, as proposed in the June
2000 rule, today's final rule amends Sec. 141.23(k)(3)(ii) to specify
an acceptance limit of 30% in PE (now known as PT) samples
spiked with arsenic at the PQL of 0.003 mg/L or greater. For a brief
discussion of the derivation of the PQL for arsenic, see section
III.B.1, What is the feasible level?
J. How Will I Know if My System Meets the Arsenic Standard?
This section summarizes changes to the arsenic monitoring and
compliance determination requirements. The Agency is also changing the
methods used by a system to determine if it is in violation of an MCL
for all of the regulated inorganic contaminants (IOCs), synthetic
organic contaminants (SOCs), and volatile organic contaminants (VOCs).
See section I.J.3. for more information regarding violation
determinations.
1. Sampling Points and Grandfathering of Monitoring Data
In today's rule, the Agency is moving the requirements associated
with arsenic into Sec. 141.23(c) making it consistent with the
requirements for IOCs regulated under the standardized monitoring
framework. All CWS and NTNCWSs must monitor for arsenic at each entry
point to the distribution system. In some cases, Sec. 142.11(1) allows
States to establish regulations that ``vary from comparable regulations
set forth in part 141 of this chapter, and demonstrate that any
different State regulation is at least as stringent as the comparable
regulation contained in part 141.'' Using this authority, States may
allow systems to collect samples at an alternative location (e.g., the
first point of drinking water consumption in the distribution system)
if the State justifies in its primacy program that the alternative
location is equally or more protective. States could implement the
change in sampling location once the primacy package is approved.
The MCL compliance elements of the rule become effective in 2006.
Some ground water systems will collect samples to comply with the
sampling requirements for all regulated IOCs (including arsenic) in
2005 in accordance with the State monitoring plan. This sampling event
will satisfy the monitoring requirements for the 2005-2007 compliance
period, but the revised arsenic MCL will not become effective until
2006. Ground water systems may use grandfathered data collected after
January 1, 2005 to satisfy the sampling requirements for the 2005-2007
compliance period. The grandfathered data must report results from
analytical methods approved for use by this final rule (e.g., the
method detection limit must be substantially less than the revised MCL
of 10 g/L). Data collected using unacceptably high detection
levels (e.g. using ICP-AES technology) will not be eligible for
grandfathering. If the grandfathered data are used to comply with the
2005-2007 compliance period and the analytical result is greater than
10 g/L, that system will be in violation of the revised MCL on
the effective date of the rule. If systems do not use grandfathered
data, then surface water systems must collect a sample by December 31,
2006 and ground water systems must collect a sample by December 31,
2007 to demonstrate compliance with the revised MCL.
2. Compositing of Samples
Compositing of samples is allowed under the standardized monitoring
framework. The States that allow compositing of samples use the
methodology in the Phase II/V regulations as specified in
Sec. 141.23(a)(4). In today's rule, CWSs and NTNCWSs will still be
allowed to composite samples; however, if arsenic is detected above
one-fifth of the revised MCL (2 g/L), then a follow-up sample
must be taken within 14 days at each sampling point included in the
composite as described in Sec. 141.23(a)(4). Compliance determinations
must be based on the follow up sample result. Water systems may
composite samples (temporally and spatially) until a
[[Page 6990]]
contaminant (arsenic or any other contaminant regulated in the Phase
II/V regulations) is detected. Once a contaminant has been detected in
a composited sample at concentrations greater than one-fifth of the
MCL, the system(s) must discontinue the practice of compositing samples
for all future monitoring.
3. Calculation of Violations
In today's rule, the Agency is clarifying the compliance
determination section for the IOCs (including arsenic), the SOCs, and
the VOCs in Secs. 141.23(i), 141.24(f)(15), and 141.24(h)(11),
respectively.
Systems will determine compliance based on the analytical result(s)
obtained at each sampling point. If any sampling point is in violation
of an MCL, the system is in violation. For systems monitoring more than
once per year, compliance with the MCL is determined by a running
annual average at each sampling point. Systems monitoring annually or
less frequently whose sample result exceeds the MCL for any inorganic
contaminant in Sec. 141.23(c), or whose sample results exceeds the
trigger level for any organic contaminant listed in Sec. 141.24(f) or
Sec. 141.24(h), must revert to quarterly sampling for that contaminant
the next quarter. Systems are only required to conduct quarterly
monitoring at the entry point to the distribution system at which the
sample was collected and for the specific contaminant that triggered
the system into the increased monitoring frequency. Systems triggered
into increased monitoring will not be considered in violation of the
MCL until they have completed one year of quarterly sampling. If any
sample result will cause the running annual average to exceed the MCL
at any sampling point (i.e., the analytical result is greater than four
times the MCL), the system is out of compliance with the MCL
immediately. Systems may not monitor more frequently than specified by
the State to determine compliance unless they have applied to and
obtained approval from the State. If a system does not collect all
required samples when compliance is based on a running annual average
of quarterly samples, compliance will be based on the running annual
average of the samples collected. If a sample result is less than the
method detection limit, zero will be used to calculate the annual
average. States have the discretion to delete results of obvious
sampling or analytic errors.
States still have the flexibility to require confirmation samples
for positive or negative results. States may require more than one
confirmation sample to determine the average exposure over a 3-month
period. Confirmation samples must be averaged with the original
analytical result to calculate an average over the 3-month period. The
3-month average must be used as one of the quarterly concentrations for
determining the running annual average. The running annual average must
be used for compliance determinations.
The rule requires that monitoring be conducted at all entry points
to the distribution system. However, the State has discretion to
require monitoring and determine compliance based on a case-by-case
analysis of individual drinking water systems. The Agency cannot
address all of the possible outcomes that may occur at a particular
water system; therefore, EPA encourages drinking water systems to
inform State regulators of their individual circumstances. Some systems
have implemented elaborate plans including targeted, increased
monitoring that is more representative of the average annual
contaminant concentration to which individuals are being exposed (some
States use a time-weighted or flow-weighted averaging approach to
determine compliance).
Some States require that systems collect samples from wells that
only operate for one month out of the year regardless of whether they
are operating during scheduled sampling times. The State may determine
compliance based on several factors including, but not limited to, the
quantity of water supplied by a source, the duration of service of the
source, and contaminant concentration.
4. Monitoring and Compliance Schedule
Systems must begin complying with the clarified monitoring and
compliance determination provisions of today's rule effective January
22, 2004 for inorganic, volatile organic, and synthetic organic
contaminants. These requirements clarify that for Secs. 141.23(i)(2),
141.24(f)(15)(ii), and 141.24(h)(11)(ii) compliance will be determined
based on the running annual average of the initial MCL exceedance and
any subsequent State-required confirmation samples. In addition, the
clarifications address calculation of compliance when a system fails to
collect the required number of samples. Compliance (determined by the
average concentration) will be based on the total number of samples
collected. Some systems have purposely not collected the required
number of quarterly samples and only incurred monitoring and reporting
violations for the uncollected samples. Any systems that avoid required
sampling will calculate MCL violations by dividing the summed samples
by the actual number of samples taken. This clarification did not
change Secs. 141.23(i)(1) and 141.24(h)(11)(i) which allow systems to
use zero for all non-detects when calculating MCL violations. In
addition, if any one sample would cause the annual average to be
exceeded, the system is out of compliance immediately.
Also in today's rule, the Agency is moving the arsenic monitoring
and compliance requirements from Secs. 141.23(l) to (q) to the
standardized monitoring framework in Sec. 141.23 for other IOCs. States
may grant systems nine-year monitoring waivers using the conditions in
Sec. 141.23(c) for arsenic. The criteria for developing a State waiver
program were published in the Phase II/V rules, and as noted in section
IV.B. of this rule, the Agency is not modifying the waiver criteria in
today's rulemaking. However, the revised arsenic rule is not effective
until January 23, 2006 (see section I.M. for a more detailed discussion
regarding the effective date of the rule.). States and utilities
supported moving arsenic into the standardized monitoring framework.
To use compliance data after the effective date of the 10
g/L MCL, systems must use an approved method with a method
detection limit substantially less than the revised arsenic MCL of 10
g/L. This means that after December 31, 2006 and December 31,
2007 all surface water systems and groundwater systems, respectively,
may not use analytical methods using the ICP-AES technology, because
the detection limits for these methods are 8 g/L or higher.
This restriction means that two ICP-AES methods that were approved when
the MCL was 50 g/L may not be used for compliance
determinations at the revised MCL of 10 g/L. The two methods
are EPA Method 200.7 and SM 3120B. Prior to 2005, systems may have
compliance samples analyzed with these less sensitive methods. However,
EPA advises systems to have compliance samples analyzed and reported at
the laboratory minimum detection limit.
If sampling demonstrates that arsenic exceeds the MCL, a CWS will
be triggered into quarterly monitoring for that sampling point ``in the
next quarter after the violation occurred.'' The State may allow the
system to return to the routine monitoring frequency when the State
determines that the system is reliably and consistently below the MCL.
However, the State cannot make a determination that the system is
reliably and consistently below the MCL until a
[[Page 6991]]
minimum of two consecutive ground water, or four consecutive surface
water samples, have been collected (Sec. 141.23(c)(8)).
The Agency is not promulgating a reduced monitoring approach
similar to the revised radionuclides final rule published on December
7, 2000 (65 FR 76708; EPA, 2000p). As noted above, all systems have to
collect IOC samples once a year or once every three years, depending on
the source water, unless they have a waiver. The Agency believes that
very few States issue waivers for IOCs because the analysis is
relatively inexpensive and most IOCs are naturally occurring elements
that may be found in concentrations above the method detection limit.
Therefore, the majority of systems must collect routine samples for the
regulated IOCs; and most of the methods used for analysis of these
contaminants will measure arsenic as well as antimony, beryllium,
cadmium, chromium, copper, and nickel.
K. What do I Need to tell My Customers?
1. Consumer Confidence Reports
a. General requirements. In 1998, EPA promulgated the Consumer
Confidence Report Rule (CCR) (codified at 40 CFR part 141, subpart O),
a final rule requiring community water systems to issue annual water
quality reports to their customers (63 FR 44512; EPA, 1998i). The
reports are due each year by July 1, and provide a snapshot of water
quality over the preceding calendar year. The reports include
information on levels of detected contaminants and if the system has
violated an MCL or a treatment technique, must also include information
on the potential health effects of contaminants from appendix A to
subpart O. When they have such violations, systems must also include in
their report an explanation of the violation and remedial measures
taken to address it. The arsenic health effects language is currently
required when arsenic levels exceed 25 g/L, one-half the
existing MCL of 50 g/L, required under Sec. 141.154(b).
EPA is today retaining the health effects language for arsenic
issued with the final CCR Rule and updating appendix A to subpart O to
include the MCL and MCLG as revised in this rule, together with special
arsenic-specific reporting requirements.
In addition to the standard reporting of arsenic detects and
arsenic MCL violations, EPA is today finalizing a requirement (proposed
at Sec. 141.154(b); finalized at Sec. 141.154(f)) that CWSs that detect
arsenic between the revised and existing MCL (i.e., above 10
g/L and up to and including 50 g/L) prior to the
effective date for compliance with the revised MCL, include the CCR
Rule health effects language in their reports. This action is required
even though, technically, the systems are not in violation of the
regulations. This requirement will be effective for the five years
after promulgation, when systems are not yet required to comply with
the revised MCL. Then, beginning January 23, 2006, systems out of
compliance must report violations of the revised arsenic MCL under
Sec. 141.153(d)(6) to the public.
Based on stakeholder and commenter input, the Agency decided in the
final CCR Rule that it would use authority granted in SDWA section
1414(c)(4)(B)(vi) to require inclusion of health effects language for
arsenic exceedances before the compliance date. That section allows the
Administrator to require inclusion of health effects language for ``not
more than three regulated contaminants'' other than those found to
violate an MCL. The Agency used this authority for total
trihalomethanes in the Stage 1 Disinfectants and Disinfection
Byproducts Rule (63 FR 69390). The Agency is now using this same
authority for arsenic, because it believes that it is important to
provide customers with the most current understanding of the risk
presented by this contaminant as soon as possible after establishing a
new standard. This provision provides systems the flexibility to put
this health effects information into context and to explain to
customers that the system is complying with existing standards.
EPA modified the language it proposed on June 22, 2000 to reflect
the MCL promulgated today and to clarify what language a system must
include in its report. Systems subject to Sec. 141.154(f) must begin
including the arsenic health effects language in the report due by July
1, 2002.
b. Special informational statement. In addition, in the CCR Rule,
the Agency decided to require that CCRs include additional information
about certain contaminants, one of which was arsenic. As explained in
the preamble to the CCR Rule (63 FR 44512 at 44514; EPA, 1998i),
because of commenters' concerns about the adequacy of the current MCL,
EPA decided that systems that detect arsenic between 25 g/L
and the current MCL must include some information regarding the arsenic
standard (Sec. 141.154(b)). This informational statement is different
from the health effects language required for an MCL violation. EPA
noted in the CCR rule and in the arsenic proposal that the
informational statement requirement would be deleted upon promulgation
of a revised MCL.
In view of the fact that EPA is today finalizing an MCL somewhat
higher than the technologically feasible MCL, and that some commenters
expressed concern about the risk that a higher-than-feasible MCL might
present to certain consumers, EPA is today retaining and revising an
existing Sec. 141.154(b) requirement that systems which find arsenic
below the MCL must provide additional information to their customers.
EPA believes that consumers should be aware of the uncertainties
surrounding the risks presented even by very low levels of arsenic.
While EPA addressed many of the sources of uncertainty in its risk
analysis of arsenic in support of the final rule, several sources of
uncertainty remain. Chief among these is the mode of action (i.e., the
shape of the dose-response curve). EPA continues to research the
effects of arsenic (according to an arsenic research plan required by
the 1996 SDWA Amendments and submitted to Congress) and should have a
better understanding of these effects as the relevant research is
completed. EPA believes that this uncertainty adequately justifies
retaining the existing requirement to provide consumers with
information about low levels of arsenic.
The existing Sec. 141.154(b) requirement is today updated in two
ways. First, the arsenic level that triggers the additional information
is reset from 25 g/L (half the existing MCL) to 5 g/L
(half the revised MCL). In the preamble to the CCR Rule, we explained
that ``[many] commenters agreed that half the MCL would be an
appropriate threshold for requiring additional risk-related
information.'' EPA continues to believe that half the MCL is an
appropriate trigger for special information about certain contaminants.
Beginning with the report due by July 1, 2002, CWSs that find arsenic
above 5 g/L and up to and including 10 g/L must
include Sec. 141.154(b) special health information about arsenic in
their consumer confidence reports.
Second, the suggested text of the special information is updated.
Rather than stating that ``EPA is reviewing the drinking water standard
for arsenic . . .,'' the statement announces clearly that the
consumer's water meets EPA's new standard while also noting the cost-
benefit trade-off involved in setting that standard. The suggested text
further notes that there are uncertainties (described in section III.F
of this notice) surrounding the risks of low levels of arsenic. Systems
retain the flexibility, as defined in the existing requirement, to
[[Page 6992]]
adjust this language in consultation with the Primacy Agency.
2. Public Notification
On May 4, 2000, EPA issued the final Public Notification Rule (PNR)
to revise the minimum requirements that public water systems must meet
for public notification of violations of EPA's drinking water standards
(65 FR 25982; EPA, 2000e). Water systems must begin to comply with the
revised PNR regulations on October 31, 2000 (if they are in
jurisdictions where the program is directly implemented by EPA) or on
the date a primacy State adopts the new requirements (not to exceed May
6, 2002). EPA's drinking water regulation on arsenic affects public
notification requirements and amends the PNR as part of its rulemaking.
Today's final rule will require CWSs and NTNCWSs to provide a Tier
2 public notice for arsenic MCL violations and to provide a Tier 3
public notice for violations of the monitoring and testing procedure
requirements. The new arsenic MCL will become effective January 23,
2006. CWSs and NTNCWSs must provide public notification to consumers
for any violations after the effective date of the revised arsenic MCL.
The PNR requires owners and operators of public water systems to give
notice to persons they serve for all violations when they are operating
under a variance or exemption (or violate conditions of the variance or
exemption).
L. What Financial Assistance is Available for Complying With This Rule?
There are two major sources of Federal financial assistance
available for water systems: the Drinking Water State Revolving Fund
(DWSRF) and the Water and Waste Disposal Loan and Grant Program of the
Rural Utilities Service (RUS) of the U. S. Department of Agriculture.
The 1996 SDWA Amendments authorized (i.e., approved spending) $9.6
billion for the DWSRF program. To date, Congress has appropriated
(i.e., provided) $4.2 billion, which includes $825 million for the
program in Fiscal Year 2001. By the end of September 2000, States had
been awarded $3.2 billion in capitalization grants and, from that, had
provided more than $2.8 billion in assistance to eligible drinking
water systems. The Federal capitalization grant, together with State
matching funds, is currently making available about $1 billion per
year. States have considerable discretion in designing their DWSRF
program, and have the option of offering special assistance to systems
that the State considers to be disadvantaged. Special assistance may
include principal forgiveness, a negative interest rate, an interest
rate lower than that charged to non-disadvantaged systems, and extended
repayment periods of up to 30 years. Federal law allows DWSRF
assistance to be provided to water systems of both public ownership and
private ownership, although some States are unable or choose not to
provide assistance to privately owned systems.
EPA recognizes that public water systems and States face a
significant challenge in implementing new requirements that are needed
to ensure the continued provision of safe drinking water. While the
DWSRF program is proving to be a significant source of funding, it
cannot be viewed as the only source of funding. It will take a
concerted effort on the part of Federal, State and local governments,
private business, and utilities to address the significant
infrastructure needs identified by public water systems. In order to
ensure that the DWSRF program is used to focus attention on the highest
priority needs, all States must give priority to those drinking water
infrastructure improvement projects that will have the greatest public
health benefit or ensure compliance with SDWA. State DWSRF programs are
currently making loans available to the highest ranked projects on
their lists and are also using a portion of the grants to support other
important drinking water program activities.
The RUS program is focused on providing a safe, reliable water
supply and wastewater treatment to residents of rural America. The
program offers a combination of low interest loans and grants to
systems serving rural areas and cities and towns of up to 10,000
persons and which are publicly owned (including Native American
systems) or operated as not-for-profit corporations. In recent years
the RUS program has typically offered assistance totaling about $1.3
billion per year, about 60% of which is directed to drinking water
projects. Thus, about $780 million per year is available for rural
drinking water systems from this program. Together with the
approximately $1 billion per year being made available through the
DWSRF, this results in a total of about $1.78 billion per year of
Federal financial assistance available for drinking water.
Other Federal financial assistance programs exist that may help
systems with SDWA compliance related expenditures. However, these other
programs are not generally as large or focused on drinking water as are
the DWSRF and RUS programs. EPA's Environmental Financial Advisory
Board has developed a ``Guidebook of Financial Tools'' (EPA, 1999c),
which offers a comprehensive summary of public and private programs and
mechanisms for paying for drinking water and other environmental
systems. The handbook is available through EPA's web site at: http://www.epa.gov/efinpage/guidbk98/index.htm.
The Federal financial assistance programs described previously
clearly face numerous, competing demands on their resources. EPA's 1995
Drinking Water Infrastructure Needs Survey (EPA, 1997a) identified a
total 20-year need for all systems of $138.4 billion. The single
largest category of need (accounting for over half of the total need)
is installation and rehabilitation of transmission and distribution
systems. Treatment needs constitute the second largest category of
need, accounting for over \1/4\ of total needs. Storage and source
rehabilitation and development constitute the remaining major
categories of needs. Thus, systems seeking financial assistance for
installation of arsenic treatment are competing for resources with
systems seeking assistance for compliance with other rules and with
systems seeking resources for basic infrastructure repair and
replacement. In seeking to meet these numerous and competing needs, the
Agency recognizes the importance of priority setting for financial
assistance programs. Systems having the financial capability to secure
funding through the capital markets should do so, leaving the Federal
financial assistance programs to assist the truly needy systems. Since
the demand for assistance will likely outstrip the supply of
assistance, States may wish to consider exemptions, which will provide
additional time for systems to secure financial assistance.
M. What is the Effective Date and Compliance Date for the Rule?
In the proposed rule, EPA made a finding that all small systems
(i.e., systems serving 10,000 people or less) would be granted a 2-year
capital improvement extension which extends the MCL effective date for
purposes of compliance with the new MCL to January 23, 2006. EPA
proposed the 2-year capital improvement extension for small systems
because of the time required for systems to plan, finance, design and
construct new treatment systems.
Large systems were not provided this additional time because of the
greater resources these systems have to perform
[[Page 6993]]
capital improvements in a timely manner. However, upon consideration of
information submitted by commenters, EPA has determined that large
systems will also require an additional 2 years to complete the capital
improvements necessary to comply with the arsenic MCL. While large
systems (i.e., systems serving more than 10,000 people) do have greater
resources to implement capital improvements, (e.g., engineering and
construction management staff to manage the projects), these systems
generally also have more entry points to the distribution system that
will require treatment.
A number of treatment technologies are listed as BAT for the
proposed rule: ion exchange, activated alumina, reverse osmosis,
modified coagulation/filtration, modified lime softening and
electrodialysis reversal. There are also several emerging technologies
for arsenic removal, such as nanofiltration and granular ferric
hydroxide. To ensure cost effective compliance with the arsenic MCL,
systems will need to evaluate their treatment technology options as a
first step. This planning step may include pilot studies with potential
treatment systems, or it may be limited to an evaluation of the raw
water characteristics. Systems choosing to conduct pilot testing may
take a year or more to contract with vendors and to perform pilot
testing.
Once the planning step is completed systems must design and
construct the treatment systems. Design and permitting of the treatment
systems can take an additional year, and construction of the treatment
system can take another year. Because systems will also need time to:
obtain funding, obtain local government approval of the project, or
acquire the land necessary to construct these technologies, it is
likely that most large systems will need additional time beyond the
three-year effective date for compliance with the new MCL that EPA
proposed.
Based upon these considerations, EPA determined, in accordance with
section 1412(b)(10) of SDWA, that the compliance date for the new
arsenic MCL, regardless of system size, will be 5 years from the date
of promulgation of the standard. See section I.H. for more information
regarding variance and exemptions.
N. How Were Stakeholders Involved in the Development of This Rule?
EPA met extensively with a broad range of groups during the
development of the arsenic proposal, both at EPA-sponsored meetings and
at other organizations' meetings. The Federal Register published
notices about EPA's arsenic meetings, and we made conference call lines
available for those who chose not to attend in person. In addition, EPA
notified people about regulatory actions via the three Federal Register
notices (proposal, notice of data availability, and correction notice),
by mail and e-mail. Over 600 people asked to be on the mailing list
during the regulatory development period.
EPA held arsenic stakeholders meetings September 11-12, 1997 in
Washington, DC; February 25, 1998 in San Antonio, Texas; May 5, 1998 in
Monterey, California; June 2-3, 1999 in Washington, DC; and August 9,
2000 in Reno, Nevada. For each of these meetings we invited
representatives of States, tribal groups, associations, utilities and
environmental groups. The docket for the proposed rule (W-99-16)
contains the meeting discussion papers, agendas, participants lists,
presentation materials, and executive meeting summaries. All the
meeting materials, except the presentations and attendance list, are
also available on EPA's arsenic in drinking water web page,
www.epa.gov/safewater/arsenic.html.
EPA also presented sessions on drinking water regulations
(including arsenic) at the National Indian Health Board Annual
Conference in Anchorage, Alaska in September 1998. The Inter-tribal
Council of Arizona hosted a consultation for EPA with Tribes February
24-25, 1999 in Las Vegas, NV at which an overview of the proposed
arsenic regulation was presented. EPA also conducted a series of
workshops at the Annual Conference of the National Tribal Environmental
Council May 18-20, 1999 in Eureka, California. The Council distributed
materials and gathered comments on EPA's drinking water regulations
from all recognized Tribal governments.
In addition to the general stakeholder meetings, EPA also had
targeted meetings with States' representatives. In May 1999, State
regulatory representatives from California, Nevada, Michigan, Illinois,
Texas, Indiana, New Mexico, and Louisiana joined EPA in a discussion on
the development of the cost of compliance decision tree. In August
1999, State regulatory representatives from Illinois, Indiana, New
Mexico, and Texas joined EPA workgroup members in a discussion of the
NRC study use, review of the occurrence work, treatment technology
update, and regulatory changes. The interaction from these meetings
with State colleagues improved the regulatory language and the
preamble.
In May 2000, EPA presented a summary of the rule to the National
Governors' Association. In May 2000, EPA held a dialogue in Washington,
DC with State officials and the associations that represent elected
officials. Presentations on arsenic and other drinking water rules
under development were given to representatives of the National
Association of Towns and Townships, National Governors' Association,
National Association of Counties, National League of Cities,
Association of State Drinking Water Administrators, Environmental
Council of the States, Florida Department of Environmental Protection,
Drinking Water Section, Association of State and Territorial Health
Officials, and the International City/County Management. The purpose of
the dialogue was to consult on the expected compliance and
implementation costs of these rules for State, county, and local
governments and gain a better understanding of the views of
representatives of State, county, and local governments and their
elected officials. The meeting materials are in the docket for the
proposed rule.
In addition to the various special meetings and discussions
mentioned previously, EPA representatives delivered arsenic regulatory
development presentations at a variety of meetings held by other
organizations. These included the American Water Works Association
(AWWA) Inorganic Contaminants Meetings in February, 1998 in San
Antonio, TX and in February, 2000 in Albuquerque, NM; meetings of the
Association of State Drinking Water Administrators (ASDWA) in February
and October 1998, March and October 1999, and in October 2000; meetings
of the Association of Metropolitan Water Agencies (AMWA) in January and
March 1998; and a meeting of the Association of California Water
Agencies in March 1998. EPA also gave several technical presentations
and regulatory updates at the AWWA annual meetings as well as at the
AWWA Water Quality and Technology Conferences in 1998, 1999, and 2000.
EPA participated in the Society of Toxicology arsenic workshop in
Philadelphia, PA in March 2000. Finally, EPA co-sponsored and
participated in the four International Conferences on Arsenic Exposure
and Health Effects in July 1993, June 1995, July 1998, and June 2000.
After the proposal was published in the Federal Register, EPA
notified all persons on its electronic mailing list for the arsenic
rule of its availability and sent information. The Regulatory Impact
Analysis went on the arsenic web page a week after the proposal
publication. Similarly, EPA also notified the individuals and
organizations on this
[[Page 6994]]
mailing list about the NODA and the correction notice.
II. Statutory Authority
Section 1401 of SDWA requires a ``primary drinking water
regulation'' to specify a MCL if it is economically and technically
feasible to measure the contaminant and to include testing procedures
to insure compliance with the MCL and proper operation and maintenance.
An NPDWR that establishes an MCL also lists the technologies that are
feasible to meet the MCL, but systems are not required to use the
listed technologies (section 1412(b)(3)(E)(i)). As a result of the 1996
amendments to SDWA, when issuing a NPDWR, EPA must also list affordable
technologies that achieve compliance with the MCL or treatment
technique for three categories of small systems: those serving 10,000
to 3301 persons, 3300 to 501 persons, and 500 to 25 persons. EPA can
list modular (packaged) and POE and POU treatment units for the three
small system sizes, as long as the units are maintained by the public
water system or its contractors. Home units must contain mechanical
warnings to notify customers of problems (section 1412(b)(4)(E)(ii)).
In section 1412(b)(12)(A) of SDWA, as amended August 6, 1996,
Congress directed EPA to propose a national primary drinking water
regulation for arsenic by January 1, 2000 and issue the final
regulation by January 1, 2001. At the same time, Congress directed EPA
to develop a research plan by February 2, 1997 to reduce the
uncertainty in assessing health risks from low levels of arsenic and
conduct the research in consultation with the NAS, other Federal
agencies, and interested public and private entities. The amendments
allowed EPA to enter into cooperative agreements for research. On
October 27, 2000, Public Law 106-377, the bill which included Fiscal
Year 2001 appropriations for EPA, amended the statutory deadline to
direct EPA to promulgate a final arsenic standard by no later than June
22, 2001.
Section 1412(a)(3) requires EPA to propose an MCLG simultaneously
with the NPDWR. The MCLG is defined in section 1412(b)(4)(A) as ``the
level at which no known or anticipated adverse effects on the health of
persons occur and which allows an adequate margin of safety.'' Section
1412(b)(4)(B) specifies that each NPDWR will specify an MCL as close to
the MCLG as is feasible, with two exceptions added in the 1996
amendments. First, the Administrator may establish an MCL at a level
other than the feasible level if the treatment to meet the feasible MCL
would increase the risk from other contaminants or the technology would
interfere with the treatment of other contaminants (section
1412(b)(5)). Second, if benefits at the feasible level would not
justify the costs, EPA may propose and promulgate an MCL ``that
maximizes health risk reduction benefits at a cost that is justified by
the benefits'' (section 1412(b)(6)).
When proposing an MCL, EPA must publish, and seek public comment
on, the health risk reduction and cost analyses (HRRCA) of each
alternative maximum contaminant level considered (section
1412(b)(3)(C)(i)). This includes the quantifiable and nonquantifiable
benefits from reductions in health risk, including those from removing
co-occurring contaminants (not counting benefits resulting from
compliance with other proposed or final regulations), costs of
compliance (not counting costs resulting from other regulations), any
increased health risks (including those from co-occurring contaminants)
that may result from compliance, incremental costs and benefits of each
alternative MCL considered, and the effects on sensitive subpopulations
(e.g., infants, children, pregnant women, elderly, seriously ill, or
other groups at greater risk). EPA must analyze the quality and extent
of the information, the uncertainties in the analysis, and the degree
and nature of the risk. As required by the statute, EPA issued a HRRCA
for arsenic (EPA, 2000i) as section XIII of the June 22, 2000 arsenic
proposal (65 FR 38888 at 38957).
The 1996 amendments also require EPA to base its action on the best
available, peer-reviewed science and supporting studies and to present
health effects information to the public in an understandable fashion.
To meet this obligation, EPA must specify, among other things,
peer-reviewed studies known to the Administrator that support, are
directly relevant to, or fail to support any estimate of public
health effects and the methodology used to reconcile inconsistencies
in the scientific data (section1412(b)(3)(B)(v)).
Section 1413(a)(1) allows EPA to grant States primary enforcement
responsibility (primacy) for NPDWRs when EPA has determined that the
State has adopted regulations that are no less stringent than EPA's.
States must adopt comparable regulations within two years of EPA's
promulgation of the final rule, unless a two-year extension is granted.
State primacy also requires, among other things, adequate enforcement
(including monitoring and inspections) and reporting. EPA must approve
or deny State applications within 90 days of submission (section
1413(b)(2)). In some cases, a State submitting revisions to adopt an
NPDWR has primacy enforcement authority for the new regulation while
EPA action on the revision is pending (section 1413(c)). Section
1451(a) allows EPA to grant primacy enforcement responsibility to
Federally recognized Indian Tribes, providing grant and contract
assistance, using the procedures applied to States.
III. Rationales for Regulatory Decisions
A. What Is the MCLG?
The proposed rule suggested that an MCLG of zero be established for
arsenic in view of the fact that we are currently unable to specify a
safe threshold level due to uncertainty about the mode of action for
arsenic. Today's rule establishes a final MCLG for arsenic of zero.
After full consideration of public comments, EPA continues to believe
that the most scientifically valid approach, given the lack of critical
data, is to use the linear approach to assessing the mode of action.
This approach results in an MCLG of zero. In the proposal and the NODA,
EPA noted that the available data point to several potential
carcinogenic modes of action for arsenic (EPA also requested additional
data on the mode of action). However, which mode(s) of action is
operative is unknown. For this reason, while the Agency recognizes that
the dose-response relationship may be sublinear, the data do not
provide any basis upon which EPA could reasonably construct this
relationship. Thus, EPA has no basis upon which to depart from its
assumption of linearity. The NRC report noted that available data that
could help determine the shape of the dose-response curve are
inconclusive and do not meet EPA's stated criteria for departure from
the default assumption of linearity (NRC, 1999). See section III.D.1
for a thorough discussion of the dose-response assessment.
Because the postulated mode of action for arsenic cannot
specifically be described and the key events are unknown, the Agency
lacks sufficient available, peer-reviewed information to estimate
quantitatively a non-linear mode of action. The Agency has thus decided
not to depart from the assumption of linearity in selecting an MCLG of
zero.
B. What Is the Feasible Level?
1. Analytical Measurement Feasibility
In the development of a drinking water regulation, EPA derives a
practical quantitation limit (PQL) to estimate or evaluate the minimum,
[[Page 6995]]
reliable quantitation level (concentration) that most laboratories can
be expected to meet during day-to-day operations. The PQL accounts for
the limits of current measurement technologies and the laboratories
that use the methods written around these analytical technologies. The
PQL was defined in a November 13, 1985 rule (50 FR 46906, EPA, 1985b)
as ``the lowest concentration of an analyte that can be reliably
measured within specified limits of precision and accuracy during
routine laboratory operating conditions.'' A PQL is determined either
through use of interlaboratory studies or, in absence of sufficient
studies, through the use of a multiplier of 5 to 10 times the method
detection limit (MDL). Interlaboratory data are obtained from water
supply (WS) studies that are conducted by EPA to certify drinking water
laboratories. The WS studies require a candidate laboratory to measure
the concentration of the target analyte within specified limits (e.g.,
30%) of the amount spiked into a PE (now called PT)
challenge sample. Using graphical or linear regression analysis of the
WS data, the Agency sets a PQL at a concentration where at least 75% of
experienced laboratories (generally EPA and State laboratories) could
perform within this acceptable limit for accuracy, e.g.,
30%.
As discussed in the June 22, 2000 proposed rule for arsenic, the
Agency determined that the PQL (i.e., the feasible level of
measurement) for arsenic in drinking water is 0.003 mg/L with an
acceptance limit of 30%. The derivation of the PQL for
arsenic is consistent with the process used to determine PQLs for other
metal contaminants regulated under SDWA and takes into consideration
the recommendations from EPA's SAB (EPA, 1995). Using acceptance limits
of 30% and linear regression analysis of six recent WS
studies, EPA derived a PQL of 0.00258 mg/L for arsenic, which was
rounded to 0.003 mg/L at the 30%. While the PQL represents
a relatively stringent target for laboratory performance, based on the
WS data used to derive the PQL for arsenic, the Agency believes most
laboratories (using appropriate quality assurance and quality control
procedures) can achieve this level on a routine basis.
2. Treatment Feasibility
EPA has determined that 3 g/L is technologically feasible
for large systems based on peer-reviewed treatment information. EPA has
listed seven BATs for arsenic in the final rule. They are: ion exchange
when sulfate 50 mg/L, activated alumina, reverse osmosis,
modified coagulation/filtration, modified lime softening at pH >10.5,
electrodialysis reversal, and oxidation/filtration when the iron to
arsenic ratio is at least 20:1. Bench, pilot and full-scale data were
examined to determine the capabilities of the treatment processes. The
treatment performance data are summarized in ``Technologies and Costs
for the Removal of Arsenic from Drinking Water'' (EPA, 2000t).
C. How Did EPA Revise its National Occurrence Estimates?
1. Summary of Occurrence Data and Methodology
Our data and methodology for estimating arsenic occurrence are
substantially the same as in the proposed rule (65 FR 38888 at 38903;
EPA, 2000i). The data and methodology are described in detail in (EPA,
2000r). Following is a summary of our method. All of the elements of
this summary are the same as in the proposed rule, except where noted.
Our occurrence database consists of arsenic compliance monitoring
samples of finished drinking water, submitted voluntarily by drinking
water agencies in 25 States. The 25 States are distributed throughout
the U.S., with at least one located in each of the seven geographic
regions that we used in our analysis (65 FR 38888 at 38906; EPA, 2000i;
EPA, 2000r). In some States we used data only from a subset of years in
which detection limits were lowest. For each PWS in our database, we
estimated the mean arsenic concentration over time in finished water,
by first ``filling in'' non-detected concentrations, using one of two
statistical methods (EPA, 2000r), then averaging the detected and
filled-in observations from that system. Next, we collected the system
mean estimates into State distributions, then merged the State
distributions into regional and then national distributions. In
combining the regional distributions into a national distribution, we
weighted each region by the total number of systems in the region, not
just the number of systems in the States in our database. This
procedure has the same effect as assigning the regional distributions
to the 25 States for which we have no observations in our database.
In addition to the distributions of system means, we estimated
nationwide intra-system coefficients of variation (ISCV). For a given
water system, the ISCV quantifies the variation of mean arsenic levels
at the system's entry points to the distribution system (i.e., sampling
points of individual wells and treatment points) around the overall
system mean. We estimated a separate ISCV for each ground water (gw)
CWS, surface water (sw) CWS, and, unlike in the proposed rule, ground
water NTNCWS. Each of these ISCVs is assumed to be constant throughout
the U.S.
2. Corrections and Additions to the Data
Some public commenters asked whether our data might have errors in
the classification of water samples as treated or untreated. If that
were the case, then including untreated samples in our database could
cause us to overestimate occurrence in finished water. In order to
determine whether and to what extent these problems exist, we solicited
additional data sets from drinking water agencies in six States
(Alabama, California, Illinois, New Mexico, North Carolina, and Texas)
from whom we already had data in our draft data set. All six States
responded to our request by submitting additional data, including
additional identifiers of untreated observations, as well as some new
observations not contained in our draft data base. In California, once
the newly identified untreated observations were removed from the data
set, the number of surface water observations decreased from 2,488 in
the draft data set to 1,280 in the final data set. For ground water, on
the other hand, the number of samples in California increased from
5,622 to 9,494. The increase resulted in part from the additional data,
and in part because we changed our methodology, as we describe below,
to include samples from both treated and untreated ground water in our
ground water estimates. Changes in the other five States were of
smaller size.
We also updated our data set from Utah. The latest data from Utah
include more observations and covers the years 1980 to 1999. The total
number of observations from Utah in our data set increased from 2,447
to 4,684.
Table III.C-1 compares the number of observations, systems, and
States in our database, by system type and source water type, in the
proposed and final rules. Note that our complete database is larger
than shown in Table III.C-1, but in some States we excluded data from
some years in which analytical detection limits were highest. Table
III.C-1 counts only the data from the years that we used to estimate
occurrence.
[[Page 6996]]
Table III.C-1.--Summary of Occurrence Databases for the Proposed and Final Rules
--------------------------------------------------------------------------------------------------------------------------------------------------------
Proposed rule Final rule
---------------------------------------------------------------------------
System type Source water # of # of # of # of # of # of
observations systems States observations systems States
--------------------------------------------------------------------------------------------------------------------------------------------------------
CWS......................................... GW............................ 44,502 15,640 25 53,307 15,931 25
CWS......................................... SW............................ 15,892 2,360 25 16,212 2,228 25
NTNCWS...................................... GW............................ * 6,420 * 4,662 * 18 7,045 4,382 17
NTNCWS...................................... SW............................ * 420 * 150 * 14 * 409 * 118 * 15
All......................................... All........................... 67,234 22,812 25 76,973 22,659 25
--------------------------------------------------------------------------------------------------------------------------------------------------------
* Data not used in estimating occurrence.
We also updated our baseline inventory of the public water systems
in the U.S. and the populations they serve, by type of system, type of
source water, and State. We use this inventory to estimate the numbers
of systems and people affected by different MCL options, by multiplying
the number of people or systems in a given category by the estimated
fraction of systems in that category with mean arsenic greater than the
levels of interest. In the proposed rule, the occurrence and regulatory
impact analyses used different sets of baseline estimates: occurrence
took baseline estimates from EPA's 4th quarter 1997 Safe Drinking Water
Information System (SDWIS) database, while the proposal's regulatory
impact analysis (RIA) used 4th quarter 1998 SDWIS. The result, as some
public commenters pointed out, was that the proposed rule contained two
inconsistent sets of estimates of the numbers of people and systems
affected by different MCL options (65 FR 38888; EPA, 2000i, Table V-3;
EPA, 2000h, Exhibit 4-11). The two estimates of total numbers of
systems affected at various MCLs differed by up to 27%. We corrected
this inconsistency by adopting, with one modification, the baseline
inventory in EPA's Drinking Water Baseline Handbook (EPA, 2000b)
throughout this preamble and all supporting documents for the final
rule. The inventory in the Baseline Handbook is taken from EPA's 4th
quarter 1998 SDWIS database, or the same that was used in the proposed
RIA. The only modification we made to the inventory was in Alaska where
the Baseline Handbook lists zero NTNCWS and zero population served by
NTNCWS. Following public comment from the Alaska Department of
Environmental Conservation, we corrected the inventory of NTNCWS in
Alaska. The Baseline Handbook and corrected Alaska inventories are
shown in Table III.C-2.
Table III.C-2.--Alaska PWS Inventories: Baseline Handbook and Corrected
----------------------------------------------------------------------------------------------------------------
Baseline handbook Corrected
---------------------------------------------------------------
System type Source water Population Population
No. of systems served No. of systems served
----------------------------------------------------------------------------------------------------------------
CWS........................... GW.............. 508 227,874 344 175,367
CWS........................... SW.............. 160 317,155 121 260,792
NTNCWS........................ GW.............. 0 0 161 51,909
NTNCWS........................ SW.............. 0 0 35 56,013
All........................... All............. 668 545,029 661 544,081
----------------------------------------------------------------------------------------------------------------
The revised estimates of numbers of systems affected at different
arsenic concentrations are shown in Table III.C-6. Since the proposed
and final Economic Analysis use the same set of baseline estimates
(except for the small correction in Alaska), changes in Table III.C-6
compared to the proposed RIA (EPA, 2000h, Exhibit 4-11) are due to
changes in the occurrence estimates in Table III.C-3, which follows.
Changes in Table III.C-6 compared to the proposed occurrence analysis
(65 FR 38888; EPA, 2000i, Table V-3) are due to changes in occurrence
estimates and also correction of the baseline.
3. Changes to the Methodology
In September 1999, EPA sponsored a peer review of our occurrence
data and methodology by three independent experts in geochemistry and
statistics. In response to that review and public comments, we have
made minor revisions to our methodology for estimating occurrence in
two ways since the proposed rule.
First, we now estimate the occurrence distribution for ground water
NTNCWSs separately from CWSs. In the proposed rule, we used the CWSs
distribution as a surrogate for NTNCWSs, for both ground and surface
water systems. We now estimate occurrence in ground water NTNCWSs
separately, using the same method as for CWSs, as described previously.
For ground water NTNCWSs we have data from 17 States, compared to 25
States for CWSs, so there are on average fewer States with data in each
region. Moreover we have no data about NTNCWSs from any States in the
Southeast region (Alabama, Florida, Georgia, Mississippi, and
Tennessee). We therefore used the occurrence distribution for ground
water CWSs as a surrogate for ground water NTNCWSs in the Southeast.
The revised occurrence estimates for ground water NTNCWSs are shown in
Table III.C-3.
We still do not estimate a separate occurrence distribution for
surface water NTNCWSs. For surface water NTNCWSs, we did not believe
that the 118 systems for which data were provided for NTNCWSs formed as
strong a basis for estimating occurrence as the much larger CWS surface
water data base, especially in the concentration range of interest. In
addition, there is less reason to believe that surface water NTNCWSs
will differ from surface water CWSs. We thus believe the surface water
CWS estimates provide the soundest basis for estimating impacts given
the types of data available.
Second, we have improved our method for estimating intra-system
[[Page 6997]]
variability. In the proposed rule, we estimated the ISCV by measuring
the total amount of variability of arsenic concentrations around the
system mean within each system. The problem with that approach is that
it fails to distinguish between-source variability (variability of
sampling-point means around the system mean) from within-source
variability (variability of observations at each sampling point around
the sampling-point mean). Within-source variability includes variations
in concentrations through time at a source, and analytical variability
caused by imprecision of the analytical methods used to measure arsenic
in water samples. The ISCV is intended to describe only between-source
variability within a system. Following the recommendations of the peer
review, we corrected our model of intra-system variation to include
separate terms for between-source and within-source variability. As a
result, our estimates of the ISCVs decreased, since we separate out the
within-source variability. The revised ISCV estimates are shown in
Table III.C-7.
A third change to our methodology is that, for ground water
systems, we now include observations on both treated and untreated
ground water in our analysis. With the exception of iron removal
technologies, most treatment in ground water systems has little effect
on arsenic, so one might expect arsenic concentrations to be similar in
treated and untreated samples. This turns out to be the case in our
data: estimates that included untreated samples were either slightly
higher or lower than estimates with only treated samples. We therefore
decided to include both treated and untreated samples in our ground
water occurrence estimates. For surface water estimates, we still use
only samples from treated water.
4. Revised Occurrence Results
Table III.C-3 shows our revised estimates of the national
distribution of arsenic occurrence, by system type and source water
type. The distributions are stated in terms of ``exceedance
probabilities,'' that is, the fraction of systems with mean arsenic
equal to or greater than the given concentration, in finished water.
The ``weighted point estimate'' is the combination of State
distributions into a national distribution, as described previously. We
consider the weighted point estimate to be our best estimate. The
``lognormal fit'' is the result of fitting a lognormal distribution to
the weighted point estimates. The lognormal fit is an approximation to
the weighted point estimate, which we use in our cost and benefit
analyses (sections III.E and III.F). The lognormal approximation
simplifies the simulation studies that we use to derive costs and
benefits, by allowing each distribution to be summarized in terms of
only two parameters. Table III.C-4 lists the parameters of the fitted
lognormal distributions.
Table III.C-3.--National Occurrence Exceedance Probability Estimates
----------------------------------------------------------------------------------------------------------------
Percent of systems with mean finished arsenic exceeding concentrations (g/L) of:
-------------------------------------------------------------------------------
3 5 10 20 50
----------------------------------------------------------------------------------------------------------------
Ground Water CWS
----------------------------------------------------------------------------------------------------------------
Weighted point estimate......... 19.9 12.1 5.3 2.0 0.43
95% confidence interval \1\..... [19.3,21.9] [11.7,13.0] [5.2,5.9] [1.9,2.3] [0.38,0.52]
Lognormal fit................... 19.7 12.0 5.3 2.0 0.43
----------------------------------------------------------------------------------------------------------------
Surface Water CWS
----------------------------------------------------------------------------------------------------------------
Weighted point estimate......... 5.6 3.0 0.80 0.32 0.10
95% confidence interval \1\..... [4.8,20.6] [1.8,9.7] [0.52,1.6] [0.13,0.82] [0.02,0.59]
Lognormal fit................... 5.6 3.0 1.1 0.37 0.067
----------------------------------------------------------------------------------------------------------------
Ground Water NTNCWS
----------------------------------------------------------------------------------------------------------------
Weighted point estimate......... 24.2 15.6 5.3 2.1 0.47
95% confidence interval \1\.....
Lognormal fit................... 23.4 14.2 6.1 2.2 0.42
----------------------------------------------------------------------------------------------------------------
\1\ Brackets indicate confidence intervals which were computed for the proposed rule and have not been updated.
No confidence intervals were computed for NTNCWS.
Table III.C-4.--Parameters of Lognormal Distributions Fitted to National Occurrence Distributions
----------------------------------------------------------------------------------------------------------------
System type Source water Log-mean \1\ Log-SD \2\
----------------------------------------------------------------------------------------------------------------
CWS........................................... GW.............................. -0.25 1.58
CWS........................................... SW.............................. -1.68 1.74
NTNCWS........................................ GW.............................. 0.03 1.47
----------------------------------------------------------------------------------------------------------------
\1\ Log-mean = mean of natural logarithm of arsenic concentrations (g/L).
\2\ Log-SD = standard deviation of natural logarithm of arsenic concentrations (g/L).
Table III.C-3 lists separate distribution estimates for ground and
surface water CWS and for ground water NTNCWSs. As we said previously,
we believe surface water CWSs provide a more sound basis for
estimation.
For CWSs, the estimates in Table III.C-3 have changed only slightly
since the proposed rule. For ground water CWSs, the largest change is
an increase at 10 g/L from 5.3% exceedance to 5.4%. For
surface water CWSs, the largest change is a decrease at 3 g/L
from 6.0% in the proposed rule to 5.6% in Table III.C-3. This decrease
is as expected, since, as we explained previously, our revised database
excludes some observations on untreated water that were included in the
draft database. Our surface water
[[Page 6998]]
occurrence estimates did increase slightly at 5 g/L, however,
as Table III.C-8 shows.
For ground water NTNCWSs, our estimated exceedance probabilities
increased from 19.9% to 24.2% at 3 g/L, and from 12.1% to
15.6% at 5 g/L. The estimates at higher concentrations changed
by at most 0.1% point. The estimates changed because we now estimate a
separate distribution for ground water NTNCWSs, as we described
previously.
The confidence intervals listed in Table III.C-3 were computed for
the proposed rule, using a computationally intensive resampling
procedure, as described in (EPA, 2000r). Since our data set and point
estimates have changed only minimally for the final rule, we did not
recompute the confidence intervals.
Table III.C-5 shows occurrence distributions in seven geographic
regions presented in the proposal and developed by Frey and Edwards
(1997). (The States and names of these geographic regions in Table
III.C-5 are based directly on the authors' designations.) As in the
proposed rule, we find concentrations to be generally highest in the
West, and generally lowest in the Southeast and Mid-Atlantic. In
regions where analytical reporting limits in our database were mostly
higher than 3 g/L or 5 g/L, we did not attempt to
estimate occurrence at the lowest concentrations. These cases are
indicated by dashes in Table III.C-5. In some regions, we were able to
estimate occurrence in fewer States at the lowest concentrations, and
this sometimes led to inconsistencies in our estimates. For example,
for New England surface water CWSs, we estimated occurrence at 3
g/L using only Maine, and at 5 g/L using Maine, New
Hampshire, and New Jersey. The introduction of more States at higher
concentrations led to inconsistent estimates of 6.2% and 11.7% of New
England surface water CWSs with arsenic exceeding 3 g/L and 5
g/L, respectively. We did not try to resolve these
inconsistencies at the regional level, but note that the national
occurrence distributions, listed in Table III.C-3, are consistent.
Table III.C-5.--Regional Occurrence Exceedance Probability Estimates
----------------------------------------------------------------------------------------------------------------
Percent of systems with mean finished arsenic exceeding
concentrations (g/L) of:
---------------------------------------------------------------
3 5 10 20
----------------------------------------------------------------------------------------------------------------
Ground Water CWS
----------------------------------------------------------------------------------------------------------------
Mid-Atlantic.................................... (\2\) *0.4 0.7 0.0
Midwest......................................... 21.2 13.8 6.2 2.4
New England..................................... 21.7 20.8 7.0 2.9
North Central................................... 21.3 13.1 6.0 2.4
South Central................................... 18.6 9.7 3.6 1.1
Southeast....................................... 0.9 0.4 0.1 0.0
West............................................ 31.5 25.2 12.5 5.0
----------------------------------------------------------------------------------------------------------------
Surface Water CWS
----------------------------------------------------------------------------------------------------------------
Mid-Atlantic.................................... (\2\) 0.1 0.0 0.0
Midwest......................................... 3.0 1.6 0.7 0.3
New England..................................... \1\ 6.2 11.7 1.0 0.4
North Central................................... 9.1 3.2 0.6 0.1
South Central................................... 3.8 0.9 0.2 0.1
Southeast....................................... 0.2 0.1 0.0 0.0
West............................................ 12.7 8.2 3.4 1.4
----------------------------------------------------------------------------------------------------------------
Ground Water NTNCWS
----------------------------------------------------------------------------------------------------------------
Mid-Atlantic.................................... (\2\) (\2\) 1.4 0.5
Midwest......................................... 26.2 17.1 8.2 3.3
New England..................................... (\2\) (\2\) 2.1 0.6
North Central................................... 29.8 22.8 15.0 9.3
South Central................................... 24.0 14.4 5.9 1.9
Southeast....................................... 0.9 0.4 0.1 0.0
West............................................ 34.3 21.9 10.5 4.2
----------------------------------------------------------------------------------------------------------------
\1\ Estimate is inconsistent with estimate at the next higher concentration. See text for explanation.
\2\ Means not enough data to form an estimate. See text for explanation.
Table III.C-6 shows our estimates of the numbers of systems with
mean finished arsenic concentrations in various ranges, by system type
and size. As in the proposed rule, we find no evidence of any
consistent difference in mean arsenic among systems of different sizes.
We conclude that the occurrence distributions shown in Table III.C-3
apply to all categories of system size. In Table III.C-6, therefore,
the estimated numbers of systems are computed by multiplying the
baseline inventory of all systems of the given size and type, by the
corresponding probability of falling within the given range, computed
from Table III.C-3 and shown in the ``% of systems'' rows. The
estimates for surface water NTNCWSs were computed by applying the
occurrence distribution for surface water CWSs to the baseline
inventory of surface water NTNCWSs.
[[Page 6999]]
Table III.C-6.--Statistical Estimates of Numbers of Systems With Average Finished Arsenic Concentrations in
Various Ranges
----------------------------------------------------------------------------------------------------------------
Number of systems with mean arsenic concentration (g/
L) in the range of:
System size (population served) ---------------------------------------------------------------
>3 to 5 >5 to 10 >10 to 20 >20
----------------------------------------------------------------------------------------------------------------
Ground Water CWS
----------------------------------------------------------------------------------------------------------------
25 to 500....................................... 2,272 1,980 961 584
501 to 3,300.................................... 811 706 343 208
3,301 to 10,000................................. 192 167 81 49
10,001 to 50,000................................ 95 83 40 24
>50,000......................................... 15 13 6 4
All............................................. 3,384 2,949 1,432 870
% of systems.................................... 7.8% 6.8% 3.3% 2.0%
----------------------------------------------------------------------------------------------------------------
Surface Water CWS
----------------------------------------------------------------------------------------------------------------
25 to 500....................................... 76 68 14 10
501 to 3,300.................................... 92 81 17 12
3,301 to 10,000................................. 47 41 9 6
10,001 to 50,000................................ 41 36 8 5
>50,000......................................... 15 13 3 2
All............................................. 270 239 51 34
% of systems.................................... 2.5% 2.2% 0.5% 0.3%
----------------------------------------------------------------------------------------------------------------
Ground Water NTNCWS
----------------------------------------------------------------------------------------------------------------
25 to 500....................................... 1,440 1,713 545 348
501 to 3,300.................................... 230 274 87 56
3,301 to 10,000................................. 5 6 2 1
10,001 to 50,000................................ 1 1 0 0
>50,000......................................... 0 0 0 0
All............................................. 1,677 1,995 635 405
% of systems.................................... 8.6% 10.3% 3.3% 2.1%
----------------------------------------------------------------------------------------------------------------
Surface Water NTNCWS
----------------------------------------------------------------------------------------------------------------
25 to 500....................................... 14 13 3 2
501 to 3,300.................................... 5 4 1 1
3,301 to 10,000................................. 1 1 0 0
10,001 to 50,000................................ 0 0 0 0
>50,000......................................... 0 0 0 0
All............................................. 20 17 4 2
% of systems.................................... 2.5% 2.2% 0.5% 0.3%
----------------------------------------------------------------------------------------------------------------
Numbers do not add up to totals in some cases due to rounding.
Our proposed and final estimates of intra-system coefficients of
variation are shown in Table III.C-7. The revised estimates are lower,
since, as we described previously, we now better separate out within-
source (time and analytical) variability from the variability of source
means within a system. The ISCV estimate for ground water NTNCWSs also
has changed because we now estimate it separately from that of ground
water CWSs.
Table III.C-7.--Estimated Intra-System Coefficients of Variation (ISCV)
----------------------------------------------------------------------------------------------------------------
Proposed rule Final rule
--------------------------------------------------------
System type Source water 95% confidence
ISCV (percent) ISCV (percent) interval
----------------------------------------------------------------------------------------------------------------
CWS.............................. GW.................. 62.9 37.1 [33.1,40.8]
CWS.............................. SW.................. 68.4 52.6 [31.4,69.6]
NTNCWS........................... GW.................. 62.9 25.2 [9.6,34.7]
----------------------------------------------------------------------------------------------------------------
Table III.C-8 compares our proposed and final national occurrence
estimates to estimates from three other studies: the National Arsenic
Occurrence Survey (NAOS) (Frey and Edwards, 1997), National Inorganics
and Radionuclides Survey (NIRS) (Wade Miller Associates, 1992), and
U.S. Geological Survey (USGS) (USGS, 2000). All of the studies in Table
III.C-8 evaluated drinking water except for USGS, which evaluated
ambient ground water, some of which came from non-drinking water
sources. Wade Miller used surface water estimates from the 1978
Community Water System Survey, which we consider now to be out of date,
so those estimates are not shown. Note that Frey and Edwards (1997)
found significantly different occurrence distributions for small and
large systems, so the NAOS
[[Page 7000]]
estimates are reported separately for small and large systems. The NAOS
included samples from all 50 States, but it was a much smaller study
(468 samples, compared to about 77,000 in our database), and it
analyzed unfinished water samples. Frey and Edwards (1997) applied
estimated efficiencies for the treatments known to be in place at the
sampling locations, to predict the concentrations in finished water.
Table III.C-8.--Comparison of National Arsenic Occurrence Estimates
--------------------------------------------------------------------------------------------------------------------------------------------------------
% of systems with mean arsenic
exceeding concentrations (g/
Study Type of water System types Population served L) of:
---------------------------------------
2 3 5 10 20
--------------------------------------------------------------------------------------------------------------------------------------------------------
Ground Water Systems
--------------------------------------------------------------------------------------------------------------------------------------------------------
EPA-proposed......................... raw + finished......... CWS.................... all.................... 27.2 19.9 12.1 5.4 2.1
EPA-final............................ raw + finished......... CWS.................... all.................... 27.3 19.9 12.1 5.3 2.0
NAOS-small........................... finished \1\........... PWS.................... 10,000..... 23.5 NR 12.7 5.1 NR
NAOS-large........................... finished \1\........... PWS.................... > 10,000............... 28.8 NR 15.4 6.7 NR
NIRS................................. finished............... CWS.................... all.................... 17.4 11.9 6.9 2.9 1.1
USGS................................. raw.................... PWS.................... all.................... 25.0 NR 13.6 7.6 3.1
Surface Water Systems
--------------------------------------------------------------------------------------------------------------------------------------------------------
EPA-proposed......................... finished............... CWS.................... all.................... 9.9 6.0 2.9 0.8 0.3
EPA-final............................ finished............... CWS.................... all.................... 9.8 5.6 3.0 0.8 0.3
NAOS-small........................... finished \1\........... PWS.................... 10,000..... 6.2 NR 1.8 0.0 NR
NAOS-large........................... finished \1\........... PWS.................... > 10,000............... 7.5 NR 1.3 0.6 NR
--------------------------------------------------------------------------------------------------------------------------------------------------------
NR = not reported.
\1\ Predicted from raw water, using estimated efficiency of treatment in place.
Table III.C-8 shows that our proposed and final occurrence
estimates are only slightly different, with the possible exception of
surface water occurrence estimates at 3 g/L, where our
estimate decreased from 6.0% to 5.6% exceedance for the final rule. The
difference is explained by the identification and exclusion of samples
of untreated water from our database for the final rule, as we
described previously. For ground water, our estimates fall within the
range reported in the other three studies. For surface water, our
estimates are somewhat higher than those of the NAOS.
D. How Did EPA Revise its Risk Analysis?
1. Health Risk Analysis
a. Toxic forms of arsenic. Humans are exposed to many forms of
arsenic that have different toxicities. For example, the metallic form
of arsenic (0 valence) is not absorbed from the stomach and intestines
and does not exert adverse effects. On the other hand, a volatile
compound such as arsine (AsH3) is toxic, but is not present
in water or food. Moreover, the primary organic forms (arsenobetaine
and arsenocholine) found in fish and shellfish seem to have little or
no toxicity (Sabbioni et al., 1991). Arsenobetaine quickly passes out
of the body in urine without being metabolized to other compounds
(Vahter, 1994). Little is known about the various arsenic species in
vegetables, grains, and oils (NRC, 1999). Arsenite (+3) and arsenate
(+5) are the most prevalent toxic forms of inorganic arsenic found in
drinking water. In general, the inorganic forms of arsenic have been
considered to be more toxic than the organic forms. In toxicity tests,
the inorganic forms were reported to be more toxic than the organic
forms (NAS, 1977) and the trivalent form was more toxic than the
pentavalent one (Szinicz and Forth, 1988).
In animals and humans, inorganic pentavalent arsenic is converted
to trivalent arsenic that is methylated (i.e., chemically bonded to a
methyl group, which is a carbon atom linked to three hydrogen atoms) to
monomethyl arsenic (MMA) and dimethyl arsinic acid (DMA), which are
organic arsenicals. The primary route of excretion for these four forms
of arsenic is in the urine. The organic arsenicals MMA and DMA were
once thought to be much less toxic than inorganic arsenicals. Many
studies reported organic arsenicals to be less reactive in tissues, to
kill less cells, and to be more easily excreted in urine (NRC, 1999).
However, recent work has shown that the assumption that organic forms
that arise during the metabolism of inorganic arsenic are less toxic
than inorganic forms may not be correct (Aposhian et al., 2000; Petrick
et al., 2000). One reason for this was that earlier toxicity tests were
conducted using pentavalent MMA and DMA because it was believed that
trivalent MMA(III) and DMA(III) were too transient to be found in
urine. Recently, MMA(III) was isolated in human urine (Aposhian et al.,
2000). Tests have demonstrated that MMA(III) is more toxic to
hepatocytes (i.e., liver cells) that inorganic trivalent arsenic
(Petrick et al., 2000; Styblo et al., 2000). These reports indicate
that the metabolism of inorganic arsenic is not necessarily a
detoxification process. As yet, it is not known which form of arsenic
participates in the key events within cells that disrupt cell growth
control and initiate or influence tumor formation. The SAB noted that
``[i]t is not possible to consider contributions of different forms of
arsenic to the overall response based on the data that are available
today'' (EPA, 2000q).
b. Effects of acute toxicity. Inorganic arsenic can exert toxic
effects after acute (short-term) or chronic (long-term) exposure. From
human acute poisoning incidents, the LD50 of arsenic has
been estimated to range from 1 to 4 mg arsenic per kilogram (kg) of
body weight (Vallee et al., 1960, Winship, 1984). This dose would
correspond to a lethal dose range of 70 to 280 mg for 50% of adults
weighing 70 kg. At nonlethal, but high acute doses, inorganic arsenic
can cause gastroenterological effects, shock, neuritis (continuous
pain) and vascular effects in humans (Buchanan, 1962). Such incidents
usually occur after accidental exposures. However, sometimes high dose
acute exposures may be self-administered. For example, inorganic
arsenic is a component of some herbal medicines and adverse effects
have been reported after use. In one report of 74 cases (Tay and Seah,
1975), the primary signs were skin lesions (92%), neurological (i.e.,
nerve) involvement (51%), and
[[Page 7001]]
gastroenterological, hematological (i.e., blood) and renal (i.e.,
kidney) effects (19 to 23%). Although acute or short-term exposures to
high doses of inorganic arsenic can cause adverse effects, such
exposures do not occur from U.S. public water supplies in compliance
with the current MCL of 50 g/L. EPA's drinking water
regulation addresses the long-term, chronic effects of exposure to low
concentrations of inorganic arsenic in drinking water.
c. Non-cancer effects associated with arsenic. A large number of
adverse noncarcinogenic effects has been reported in humans after
exposure to drinking water highly contaminated with inorganic arsenic.
The earliest and most prominent changes are in the skin, e.g.,
hyperpigmentation and keratoses (calus-like growths). Other effects
that have been reported include alterations in gastrointestinal,
cardiovascular, hematological (e.g., anemia), pulmonary, neurological,
immunological and reproductive/developmental function (ATSDR, 1998).
The most common symptoms of inorganic arsenic exposure appear on
the skin and occur after 5-15 years of exposure equivalent to 700
g/day for a 70 kg adult, or within 6 months to 3 years at
exposures equivalent to 2,800 g/day for a 70 kg adult (NRC,
1999, pg. 131). They include alterations in pigmentation and the
development of keratoses that are localized primarily on the palms of
the hands, the soles of the feet, and the torso. The presence of
hyperpigmentation and keratoses on parts of the body not exposed to the
sun is characteristic of arsenic exposure (Yeh, 1973; Tseng, 1977). The
same alterations have been reported in patients treated with Fowler's
solution (1% potassium arsenite; Cuzick et al., 1982), used for asthma,
psoriasis, rheumatic fever, leukemia, fever, pain, and as a tonic (WHO,
1981; NRC, 1999).
Chronic exposure to inorganic arsenic is often associated with
alterations in gastrointestinal(GI) function. For example, noncirrhotic
hypertension is a relatively specific, but not commonly found
manifestation in inorganic arsenic-exposed individuals and may not
become a clinical observation until the patient demonstrates GI
bleeding (Morris et al., 1974; Nevens et al., 1990). Physical
examination may reveal spleen and liver enlargement, and
histopathological examination of tissue specimens may demonstrate
periportal fibrosis (Morris et al., 1974; Nevens et al., 1990; Guha
Mazumder et al., 1997). There have been a few reports of cirrhosis
after inorganic arsenic exposure, but the authors of these studies did
not determine the subjects' alcohol consumption (NRC, 1999).
Development of peripheral vascular disease (hardening of the
arteries to the arms and legs, that can cause pain, numbness, tingling,
infection, gangrene, and clots) after inorganic arsenic exposure has
also been reported. In Taiwan, blackfoot disease (BFD), a severe
peripheral vascular insufficiency which may result in gangrene of the
feet and other extremities) has been the most severe manifestation of
this effect. Tseng (1977) reported over 1,000 cases of BFD in the
arsenic study areas of Taiwan. Less severe cases of peripheral vascular
disease have been described in Chile (Zaldivar et al., 1974) and Mexico
(Cebrian, 1987). In a Utah study, increased standardized mortality
ratios (SMRs) for hypertensive heart disease were noted in both males
and females after exposure to inorganic arsenic-contaminated drinking
water (Lewis et al., 1999). These reports link exposure to inorganic
arsenic effects on the cardiovascular system. Although deaths due to
hypertensive heart disease were roughly twice as high as expected in
both sexes, increases in death did not relate to increases in dose,
calculated as the years of exposure times the median arsenic
concentration. The Utah data indicate that heart disease should be
considered in the evaluation of potential benefits of U.S. regulation.
Vascular effects have also been reported as an effect of arsenic
exposure in another study in the U.S. (Engel et al., 1994), in Taiwan
(Wu et al., 1989) and in Chile (Borgono et al., 1977). The overall
evidence indicating an association of various vascular diseases with
arsenic exposure supports consideration of this endpoint in evaluation
of potential noncancer health benefits of arsenic exposure reduction.
Studies in Taiwan (Lai et al., 1994) and Bangladesh (Rahman et al.,
1998) found an increased risk of diabetes among people consuming
arsenic-contaminated water. Two Swedish studies found an increased risk
of mortality from diabetes among those occupationally exposed to
arsenic (Rahman and Axelson, 1995; Rahman et al., 1998).
Although peripheral neuropathy (numbness, muscle weakness, tremors;
ATSDR, 1998) may be present after exposure to short-term, high doses of
inorganic arsenic (Buchanan, 1962; Tay and Seah, 1975), there are no
studies that definitely document this effect after exposure to levels
of less than 50 g/L of inorganic arsenic in drinking water.
Hindmarsh et al. (1977) and Southwick et al. (1983) have reported
limited evidence of peripheral neuropathy in Canada and the U.S.,
respectively, but it was not reported in studies from Taiwan, Argentina
or Chile (Hotta, 1989, as cited by NRC 1999).
There have been a few, scattered reports in the literature that
inorganic arsenic can affect reproduction and development in humans
(Borzysonyi et al., 1992; Desi et al., 1992; Tabacova et al., 1994;
Hopenhayn-Rich et al., 2000). After reviewing the available literature
on arsenic and reproductive effects, the NRC (1999) wrote that
``nothing conclusive can be stated from these studies.'' Regarding the
Hopenhayn-Rich study, the majority of the SAB panel (EPA, 2000q)
concluded that while:
it is generally reasonable to consider that children are generally
at greater risk for a toxic response to any agent in water because
of their greater drinking water consumption (on a unit-body weight
basis), [the SAB does not] believe that this study demonstrates such
a heightened sensitivity or susceptibility to arsenic.
The EPA agrees with this conclusion.
d. Cancers associated with arsenic. Inorganic arsenic is a multi-
site human carcinogen by the drinking water route. Asian, Mexican and
South American populations with exposures to arsenic in drinking water
generally at or above hundreds of micrograms per liter are reported to
have increased risks of skin, bladder, and lung cancer. The current
evidence also suggests that the risks of liver and kidney cancer may be
increased following exposures to inorganic forms of arsenic. The weight
of evidence for ingested arsenic as a causal factor of carcinogenicity
is much greater now than a decade ago, and the types of cancer
occurring as a result of ingesting inorganic arsenic have even greater
health implications for U.S. and other populations than the occurrence
of skin cancer alone. (Until the late 1980s skin cancer had been the
cancer classically associated with arsenic in drinking water.)
Epidemiologic studies (human studies) provide direct data on arsenic
risks from drinking water at exposure levels much closer to those of
regulatory concern than environmental risk assessments based on animal
toxicity studies.
Skin Cancer. Early reports linking inorganic arsenic contamination
of drinking water to skin cancer came from Argentina (Neubauer, 1947,
reviewing studies published as early as 1925) and Poland (Tseng et al.,
1968). However, the first studies that observed dose-dependent effects
of arsenic associated with skin cancer came from Taiwan (Tseng et al.,
1968; Tseng, 1977). These studies focused EPA's attention on the health
effects of ingested arsenic.
[[Page 7002]]
Physicians administered physical examinations to the study group of
over 40,000 residents from 37 villages, as well as to a reference group
of 7500 residents reported to be exposed to a median level of 0 to
0.017 mg/L arsenic (reference group). The study population was divided
into three groups based on exposure to inorganic arsenic (0 to 0.29,
0.30 to 0.59 and 0.60 mg of inorganic arsenic per liter (mg/
L) measured at the village level. A dose- and age-related increase of
arsenic-induced skin cancer among the villagers was noted. No skin
cancers were observed in the low arsenic reference areas. In both the
EPA 1988 report on skin cancer and the 1999 NRC report, it was noted
that grouping individuals into broad exposure groups (rather than
grouping into village exposures) limited the usefulness of these
studies for quantitative dose-response estimation. However, these Tseng
reports and other corroborating studies such as those by Albores et al.
(1979) and Cebrian et al. (1983) on drinking water exposure and
exposures to inorganic arsenic in medicines (Cuzick et al., 1982) and
in pesticides (Roth, 1956) led the EPA, using skin cancer as the
endpoint, to classify inorganic arsenic as a human carcinogen (Group A)
by the oral route (EPA, 1984).
Internal cancers. Exposure to inorganic arsenic in drinking water
has also been associated with the development of internal cancers. Chen
et al. (1985) used SMRs to evaluate the association between ingested
arsenic and cancer risk in Taiwan. (SMRs, ratios of observed to
expected deaths from specific causes, are standardized to adjust for
differences in the age distributions of the exposed and reference
populations). The authors found statistically significant increased
risks of mortality for bladder, kidney, lung, liver and colon cancers.
A subsequent mortality study in the same area of Taiwan found
significant dose-response relationships for deaths from bladder,
kidney, skin, and lung cancers in both sexes and from liver and
prostrate cancer for males. They also found increases in peripheral and
cardiovascular diseases but not in cerebrovascular accidents (Wu et
al., 1989). There are several corroborating reports of the increased
risk of cancers of internal organs from ingested arsenic including two
from South American countries. In Argentina, significantly increased
risks of death from bladder, lung and kidney cancer were reported
(Hopenhayn-Rich et al., 1996; 1998). In a population of approximately
400,000 in northern Chile, Smith et al. (1998) found significantly
increased risks of bladder and lung cancer mortality.
There have only been a few studies of inorganic arsenic exposure
via drinking water in the U.S., and most have not considered cancer as
an endpoint. The best U.S. study currently available is that of Lewis
et al. (1999) who conducted a mortality study of a population in Utah
whose drinking water contained relatively low concentrations of
arsenic. EPA scientists conducted an epidemiological study of 4,058
Mormons exposed to arsenic in drinking water in seven communities in
Millard County, Utah (Lewis et al., 1999). The 151 samples from their
public and private drinking water sources had arsenic concentrations
ranging from 4 to 620 g/L with seven median (mid-point in
range) community exposure concentrations of 14 to 166
g/L. Observed causes of death in the study group (numbering
2,203) were compared to those expected from the same causes based upon
death rates for the general white male and female population of Utah.
While the study population males had a significantly higher risk of
prostate cancer mortality, females had no significant excess risk of
cancer mortality at any site. Millard County subjects had higher
mortality from kidney cancer, but this was not statistically
significant. Both males and females in the study group had less risk of
bladder, digestive system and lung cancer mortality than the general
Utah population. The Mormon females had lower death rates from breast
and female genital cancers than the State rate. These decreased death
rates were not statistically significant.
Tsai et al. (1999) estimated SMRs for 23 cancer and non-cancer
causes of death in women and 27 causes of death in men in an area of
Taiwan with elevated arsenic exposures. The SMRs in this study are an
expression of the ratio between deaths that were observed in an area
with elevated arsenic levels and those that were expected to occur,
compared to both the mortality of populations in nearby areas without
elevated arsenic levels and to the national population. Drinking water
(250-1,140 g/L) and soil (5.3-11.2
mg/kg) in the Tsai et al. (1999) population study had high arsenic
content. However, the study gives an indication of the types of health
effects that may be associated with arsenic exposure via drinking
water. The study reports a high mortality rate (SMR > 3) for both sexes
from bladder, kidney, skin, lung, and nasal cavity cancers and for
vascular disease. Females also had high mortalities for laryngeal
cancer.
The SMRs calculated by Tsai et al. (1999) used the single cause of
death noted on the death certificates. Many chronic diseases, including
some cancers, are not generally fatal. Consequently, the impact
indicated by the SMR in this study may underestimate the total impact
of these diseases. The causes of death reported in this study are
consistent with what is known about the adverse effects of arsenic.
Tsai et al. (1999) identified ``bronchitis, liver cirrhosis,
nephropathy, intestinal cancer, rectal cancer, laryngeal cancer, and
cerebrovascular disease'' as possibly ``related to chronic arsenic
exposure via drinking water,'' which had not been reported before. In
addition, people in the study area were observed to have nasal cavity
and larynx cancers not caused by occupational exposure to inhaled
arsenic.
A small cohort study in Japan of persons exposed to arsenic in
drinking water provides evidence of the association of cancer and
arsenic among persons exposed for 5 years to 1000
g/L or more and followed for 33 years after cessation of
exposure. The strongest association was for lung and bladder cancer,
similar to results in studies in Taiwan and South America (Tsuda et
al., 1995).
Kurttio et al. (1999) conducted a case-cohort design study of 61
bladder and 49 kidney cancer cases and 275 controls to evaluate the
risk of these diseases with respect to arsenic drinking water
concentrations. In this study the median exposure was 0.1 g/L,
the maximum reported was 64 g/L, and 1% of the exposure was
greater than 10 g/L. The authors reported that very low
concentrations of arsenic in drinking water were significantly
associated with bladder cancer when exposure occurred two to nine years
prior to diagnosis. Arsenic exposure occurring greater than 10 years
prior to diagnosis was not associated with bladder cancer risk. This
raises a question about the significance of the finding about exposures
two to nine years since one would expect earlier exposure to have had
an effect given the Tsuda et al. (1995) study summarized previously.
The two internal cancers consistently seen and best characterized
in epidemiologic studies are those of lung and bladder. EPA considers
the studies summarized before as confirmation of its long-standing view
that arsenic is a known human carcinogen. This rule relies on
assessment of lung and bladder cancers for its quantitative risk
estimates in support of the MCL. EPA recognizes that other internal
cancers as well as skin cancer are important.
[[Page 7003]]
Nonetheless, some issues with other cancer endpoints led to their being
considered qualitatively rather than quantitatively. EPA has considered
skin and liver cancer qualitatively for the following reasons: (1) The
skin cancer endpoint is difficult to analyze because, in the U.S., it
is considered curable; and (2) the liver cancer endpoint is likely to
have been influenced in Taiwan by the prevalence there of viral
hepatitis which is a factor in liver cancer.
How does arsenic cause cancer? EPA sponsored an ``Expert Panel on
Arsenic Carcinogenicity: Review and Workshop'' in May 1997 (EPA,
1997e). The panel evaluated existing data to comment on arsenic's
carcinogenic mode of action and the effect on dose-response
extrapolations. The panel noted that arsenic compounds have not formed
deoxyribonucleic acid (DNA) adducts (i.e., bound to DNA) nor caused
point mutations. Thus, indications are that the mode of action does not
involve direct reaction with DNA. Trivalent inorganic forms inhibit
enzymes, but arsenite and arsenate do not affect DNA replication. The
panel discussed several modes of action, concluding that arsenic
indirectly affects DNA, inducing chromosomal changes. The panel thought
that arsenic-induced chromosomal abnormalities could possibly come from
errors in DNA repair and replication that affect gene expression; that
arsenic may increase DNA hypermethylation and oxidative stress; that
arsenic may affect cell proliferation (cell death appears to be
nonlinear); and that arsenic may act as a co-carcinogen. Arsenite
causes cell transformation but not mutation of cells in culture. It
also induces gene amplification (multiple copies of DNA sequences) in a
way that suggests interference with DNA repair or cell control instead
of direct DNA damage.
In terms of implications for the risk assessment, the panel noted
that risk per unit dose estimates from human studies can be biased
either way (i.e., reduced animal fats in the diet would underestimate
risk). For the Taiwanese study, the ``* * * biases associated with the
use of average doses and with the attribution of all increased risk to
arsenic would both lead to an overestimation of risk (EPA, 1997e, page
31).'' While health effects are most likely observed in people getting
high doses, the effects are assigned to the average dose of the
exposure group. Thus, risk per unit dose estimated from the average
doses would lead to an overestimation of risk (EPA, 1997e, page 31). On
the other hand, basing risk estimates on one or two tumor sites may
underestimate risk as compared to summing risks for all related health
endpoints.
There is much research underway about the mode of action for
arsenic. In order to understand the shape of the dose-response
relationship in the range of exposure typical of the U.S., that is
significantly below the range of observation of epidemiologic studies,
one needs to identify which one or more of the possible modes of action
is operative. If this can be elucidated, it will become possible to
study and quantify the key events within cells that influence cell
growth control and how they may quantitatively relate to eventual tumor
incidence. Until then the shape of the dose-response relationship and
whether there is any threshold cannot be known.
f. What is the quantitative relationship between exposure and
cancer effects that may be projected for exposures in the U.S.? The
Agency chose to make its quantitative estimates of risk based on the
Chen et al. (1988; 1992) and Wu et al. (1989) Taiwan studies. This
choice was endorsed by the NRC and EPA's SAB (EPA, 2000q; NRC, 1999).
The database from Taiwan has the following advantages: mortality data
were drawn from a cancer registry; arsenic well water concentrations
were measured for each of the 42 villages; there was a large,
relatively stable study population that had life-time exposures to
arsenic; there are limited measured data for the food intake of arsenic
in this population; age- and dose-dependent responses with respect to
arsenic in the drinking water were demonstrated; the collection of
pathology data was unusually thorough; and the populations were quite
homogeneous in terms of lifestyle.
EPA recognizes that there are problems with the Taiwan study that
introduce uncertainties to the risk analysis such as: the use of median
exposure data at the village level; the low income and relatively poor
diet of the Taiwanese study population (high levels of carbohydrates,
low levels of protein, selenium and other essential nutrients); and
high exposure to arsenic via food and cooking water. These are
discussed more thoroughly in the following paragraphs. The available
studies from Taiwan are ecological studies and have exposure
uncertainties that are recognized. Ecological studies are problematic
as bases for quantitative risk assessment. Errors in assigning persons
to exposures are difficult to avoid. Moreover, all confounding factors
that may have contributed to risk may not be adequately accounted for.
These uncertainties have to be remembered since they lead to
uncertainty in the quantitative dose-response relationship estimated in
the observed range of data and in any extrapolation to estimate the
potential risk at exposures significantly below the observed range.
There is not a way to take all confounding factors into account
quantitatively. (see section III.F.)
Notwithstanding these concerns, the Taiwan epidemiological studies
provide the basis for assessing potential risk from lower
concentrations of inorganic arsenic in drinking water, without having
to adjust for cross-species toxicity interpretation. Ordinarily, the
characteristics of human carcinogens can be explored and experimentally
defined in test animals. Dose-response can be measured, and animal
studies may identify internal transport, metabolism, elimination, and
subcellular events that explain the carcinogenic process. Arsenic
presents unique problems for quantitative risk assessment because there
is no test animal species in which to study its carcinogenicity. While
such studies have been undertaken, it appears that test animals do not
respond to inorganic arsenic exposure in a way that makes them useful
as a model for human cancer assessment. Their metabolism of inorganic
arsenic is also quantitatively different than humans.
There are issues with the extrapolation of the dose-response from
the observed range of exposure in Taiwan to estimate Taiwan cancer risk
below the observed data range and application of the same risk estimate
to U.S. populations. The following issues have been addressed:
The Taiwan population ingested more arsenic in food and
via cooking with contaminated water than is typical for the U.S.
population. This is because the staples of the Taiwan diet were rice
and sweet potatoes. Rice and sweet potatoes are high in arsenic and
both staples absorb water upon cooking. EPA did a sensitivity analysis
of the effect of exposure to arsenic through water used in preparing
food in Taiwan. EPA also analyzed the effect of exposure to arsenic
through food.
The Taiwan data on exposure were uncertain because the
association of individuals with contaminated wells was made by grouping
persons in a village and assuming they had a lifetime of exposure to
the median of the concentration of arsenic measured in the wells
serving that village. Wells within each village had varying arsenic
levels so that people using certain wells had much higher exposures
than others in the same village. Not all wells serving all villages
were measured. However, all villagers were assigned a single median
[[Page 7004]]
concentration for exposure. In addition, moves made from village to
village were not accounted for. When villages with only one arsenic
measurement were removed from the data set (on the theory that the
exposure data were too uncertain), or when village means instead of
medians were used for the exposure estimates, there was no
statistically significant change in the estimated point of departure,
using Model 1 of Morales et al. (2000).
The Taiwan population was a rural population that was not
well nourished, having deficits of selenium, possibly methionine or
choline (methyl donors), zinc and other essential nutrients. This
malnourishment is not typical of the U.S. population, although some
U.S. populations may have one or another of the same deficits. The
Taiwanese population may also have some genetic differences from the
general U.S. population. These issues cannot be quantitatively
accounted for. However, deficits in selenium in the diet, in
particular, are a known risk factor for cancer and indicate possible
overestimation of risk when the Taiwan data are applied. EPA has
qualitatively taken this into account. (See section III.F.)
The Utah study (Lewis et al., 1999) did not find any
excess bladder or lung cancer risk after exposure to arsenic at
concentrations of 14 to 166 g/L. An important feature of the
study is that it estimated excess risk by comparing cancer rates among
the study population, in Millard County, Utah to background rates in
all of Utah. But the cancer rates observed among the study population,
even those who consumed the highest levels of arsenic, were lower, in
many cases significantly lower, than in all of Utah. This is evidence
that there are important differences between the study and comparison
populations besides their consumption of arsenic. One such difference
is that Millard County is mostly rural, while Utah as a whole contains
some large urban populations. Another difference is that the subjects
of the Utah study were all members of the Church of Jesus Christ of
Latter Day Saints, who for religious reasons have relatively low rates
of tobacco and alcohol use. For these reasons, the Agency believes that
the comparison of the study population to all of Utah is not
appropriate for estimating excess risks. An alternative method of
analysis is to compare cancer rates only among people within the study
population who had high and low exposures. The Agency performed such an
analysis on the Utah data, using the statistical technique of Cox
proportional hazard regression (US EPA, 2000x; Cox and Oakes, 1984).
The results showed no detectable increased risk of lung or bladder
cancers due to arsenic, even among subjects exposed to more than 100
g/L on average. On the other hand, the excess risk could also
not be distinguished statistically from the levels predicted by model 1
of Morales et al. (2000). What these results show is that the Utah
study is not powerful enough to estimate excess risks with enough
precision to be useful for the Agency's arsenic risk analysis.
Furthermore, the SAB noted that ``(a)lthough the data provided in
published results of the Lewis, et al., 1999 study imply that there was
no excess bladder or lung cancer in this population, the data are not
in a form that allows dose-response to be assessed dependably'' (EPA,
2000q). The indications of Lewis et al. study have been taken into
account in the judgments of the impact of scientific uncertainties on
the final MCL.
g. Is it appropriate to assume linearity for the dose-response
assessment for arsenic at low doses given that arsenic is not directly
reactive with DNA? Independent scientific panels (EPA, 2000q; NRC,
1999; EPA, 1997e; EPA, 1988) who have considered the Taiwan study have
raised the caution that using the Taiwan study to estimate U.S. risk at
lower levels may result in an overly conservative estimation of U.S.
risk. The independent panels have each said that below the observed
range of the high level of contamination in Taiwan the shape of the
dose-response relationship may prove to be sublinear when there is
adequate data to characterize the mode of action. If so, an assumption
that the effects seen per dose increment remain the same from high to
low levels of dose may overstate the U.S. risk. In evaluating the
benefits of alternative MCLs, EPA weighed both the qualitative and
quantitative uncertainties about risk magnitude (see section III.F.)
The use of a linear procedure to extrapolate from a higher,
observed data range to a lower range beyond observation is a science
policy approach that has been in use by Federal agencies for four
decades. Its basis is both science and policy. The policy objectives
are to avoid underestimating risk in order to protect public health and
be consistent and clear across risk assessments. The science components
include its applicability to generally available data sets (animal
tests and human studies) and its basis in the fact that cancer is a
consequence of genetic changes coupled with the assumption that direct
reaction with DNA is a basic mode of action for chemicals causing
important genetic changes (Cogliano et al., eds., 1999).
The linear approach is intended to identify a level of risk that is
an upper limit on what the risk might be. There are two biological
situations in which the linear approach can be a particularly uncertain
estimate of risk. One is when the metabolism and toxicokinetics of the
agent being assessed cause a nonlinear relationship between the dose of
the active form and the dose of the applied form of the agent. If this
is not quantitatively dealt with in the dose part of the dose-response
estimation, the linear extrapolation will have added uncertainties. In
the case of arsenic, it is known that metabolism and toxicokinetics are
complex, but the active form(s) is not known. The resulting
complexities of estimating dose cannot, therefore, be accounted for in
dose-response modeling.
The other situation is when the mode of action of the agent is
indirect; that is, when there is not a one-to-one reaction between the
active form of the agent and DNA, but, instead, the active form affects
other cell components or processes that, in turn, causes genetic
change. In such cases, the rates of these secondary processes are
limiting, not the dose of the active form. With few exceptions, the
rates of these secondary processes are thought not to be a linear
function of applied dose. In the case of arsenic, it is known that
arsenic does cause genetic changes in short-term tests, but these are
indirect genetic changes (not one-to-one reactions between arsenic and
DNA).
If there are both complex toxicokinetics and secondary effects, the
upper-limit risk estimate from the linear approach provides may be
overly conservative. However, there simply are not sufficient data to
quantify the effect of these two features of arsenic on risk. While
some commenters assert that the Agency can simply use models that have
sublinear structures to address the issue of secondary nature of
effects, the Agency does not agree. There are no data on the effects of
arsenic that may be precursors to cancer. Without such biological data,
the exercise of blindly applying models has no anchor, in EPA's
judgment. Such modeled extrapolations could take numerous shapes and
there is no way to decide how shallow or steep the curve would be or
where on the dose gradient the zero risk level might be, given the
hundreds of possibilities. There are also certain modes of action that
do not involve DNA reactivity, but are thought to be linear in dose
response, such as effects on growth-control signals within cells. Since
we do not know what the mode of action of arsenic is, we cannot in fact
rule out linearity. Therefore, in
[[Page 7005]]
accordance with the 1986 cancer guidelines, and subsequent guidance
discussed later, the Agency cannot reasonably use anything other than a
linear mode of action to estimate the upper bound of risk associated
with arsenic exposure. Nevertheless, the uncertainties about both of
these facets (the toxicokinetics and secondary effects) of risk
estimation have been taken into account qualitatively in the Agency's
final decision as a perspective on the linear dose-response estimation
(see section III.F.).
The Agency considered mode-of-action information as a basis for
departing from the assumption of linearity and in the process,
developed a framework for judging the adequacy of mode of action data
(EPA, 1996a). This framework has been reviewed and supported by the SAB
(EPA, 1997f; EPA, 1999g). The framework was applied to the assessment
of chloroform (EPA, 2000d).
In order to decide whether a particular mode of action is operative
for an agent, the database on mode of action must be rich and able to
both describe the sequence of key events in the putative mode of action
and demonstrate it experimentally. The elements of the framework
analysis include:
Summary description of postulated mode of action (the
postulated sequence of cellular/physiological events leading to cancer
must be described.)
Identification of key events (the specific events that are
key to carcinogenesis must described in order to be experimentally
examined.)
Strength, consistency, specificity of association (the
experimental observation of the key events and their relationship to
tumor development must be described.)
Dose-response relationship (the dose-response relationship
between the key events and tumor incidence must be described and
evaluated.)
Temporal relationship (the key events must be shown to
precede tumor development.)
Biological plausibility and coherence (the postulated mode
of action and the data must be in accord with general, accepted
scientific evidence about the causes of cancer.)
Other modes of action (alternative modes of action that
are suggested must be examined and their contribution, if any,
described.)
Conclusion (an overall conclusion is made as to whether
the postulated mode of action is accurate given the results of
evaluation of the evidence under the previous elements.)
Human relevance, including subpopulations (if the evidence
of mode of action of carcinogenicity is from animal studies, its human
relevance is examined.)
In the case of chloroform, there was sufficient information to
describe key events and undertake mode of action analysis. In the case
of arsenic, the postulated mode of action cannot be specifically
described, the key events are unknown, and no analysis of the remaining
elements of the mode of action framework can be made. Several possible
influences of arsenic on the carcinogenic process have been postulated,
but there are insufficient experimental data either to show that any
one of the possible modes is the influence actually at work or to test
the dimensions of its influence as the framework requires.
For chloroform there are extensive data on metabolism that identify
the likely active metabolite. The key events--cell toxicity followed by
sustained cell proliferation and eventually tumor effects--have been
extensively studied in many experiments. The key events have been
empirically demonstrated to precede and consistently be associated with
tumor effects. In sum, a very large number of studies have satisfied
the requirements of the framework analysis. By contrast, the arsenic
database fails to even be able to satisfy the first element of the
framework; the key events are unknown. While there are a number of
possible modes of action implied by existing data, none of them has
been sufficiently studied to be analyzed under the Agency's framework.
For this reason the comparison of the ``best available, peer reviewed
data'' for arsenic and chloroform shows quite different results. There
are not sufficient data on arsenic to describe a mode of action as
there were for chloroform. This was also the conclusion of the SAB
review of arsenic (EPA, 2000q).
Overall, the NRC and SAB reports agreed that the best available
science provides no alternative to use of a linear dose-response
process for arsenic because a specific mode (or modes) of action has
not been identified. Unlike chloroform, the Agency lacks sufficient
available, peer-reviewed information on arsenic to estimate
quantitatively a non-linear mode of action. The Agency thus has decided
not to depart from the assumption of linearity in selecting an MCLG of
zero.
2. Risk factors/bases for upper- and lower-bound analyses
EPA calculated upper- and lower-bound risk estimates for the U.S.
population exposed to arsenic concentrations. The approach for this
analysis included five components. First, we developed relative
exposure factor distributions, which incorporate data from the recent
EPA water consumption study with age, sex, and weight data. Second, the
Agency calculated the arsenic occurrence distributions for the
population exposed to arsenic levels above 3 g/L. Third, we
chose risk distributions for bladder and lung cancer for the analysis
from Morales et al. (2000). Fourth, EPA developed estimates of the
projected bladder and lung cancer risks faced by exposed populations
using Monte-Carlo simulations, bringing together the relative exposure
factor, occurrence, and risk distributions. These simulations resulted
in upper bound estimates of the risks faced by U.S. populations exposed
to arsenic concentrations at or above 3 g/L in their drinking
water. Finally, EPA made adjustments to the lower-bound risk estimates
to reflect exposure to arsenic in cooking water and in food in Taiwan.
A more detailed description of the risk methodology is provided in
Appendix B of the Economic Analysis (EPA, 2000o).
a. Water consumption. EPA recently updated its estimates of per
capita daily average water consumption (EPA, 2000c). The estimates used
data from the combined 1994, 1995, and 1996 Continuing Survey of Food
Intakes by Individuals (CSFII), conducted by the U.S. Department of
Agriculture (USDA). The CSFII is a complex, multi-stage area
probability sample of the entire U.S. and is conducted to survey the
food and beverage intake of the U.S. Per capita water consumption
estimates are reported by source. Sources include community tap water,
bottled water, and water from other sources, including water from
household wells and rain cisterns, and household and public springs.
For each source, the mean and percentiles of the distribution of
average daily per capita consumption are reported. The estimates are
based on an average of 2 days of reported consumption by survey
respondents. The estimated mean daily average per capita consumption of
``community tap water'' by individuals in the U.S. population is 1
liter/person/day. For ``total water'', which includes bottled water,
the estimated mean daily average per capita consumption is 1.2 liters
per/person/day. These estimates of water consumption are based on a
sample of 15,303 individuals in the 50 States and the District of
Columbia. The sample was selected to represent the entire
[[Page 7006]]
population of the U.S. based on 1990 census data.
The estimated 90th percentile of the empirical distribution of
daily average per capita consumption of community tap water for the
U.S. population is 2.1 liters/person/day; the corresponding number for
the 90th percentile of daily average per capita consumption of total
water is 2.3 liters/person/day. In other words, current consumption
data indicate that 90% of the U.S. population consumes approximately 2
liters/person/day, or less.
Water consumption estimates for selected subpopulations in the U.S.
are described in the CSFII, including per capita water consumption by
source for gender, region, age categories, economic status, race, and
residential status and separately for pregnant women, lactating women,
and women in childbearing years. The water consumption estimates by age
and sex were used in the computation of the relative exposure factors
discussed later.
b. Relative Exposure Factors. Lifetime male and female relative
exposure factors (REFs) for each of the broad age categories used in
the water consumption study were calculated, where the life-long REFs
indicate the sensitivity of exposure to an individual relative to the
sensitivity of exposure of an ``average'' person weighing 70 kilograms
and consuming 2 liters of water per day, a ``high end'' water
consumption estimate according to the EPA water consumption study
referred to previously (EPA, 2000c). In these calculations, EPA
combined the water consumption data with data on population weight from
the 1994 Statistical Abstract of the U.S. Distributions for both
community tap water and total water consumption were used because the
community tap water estimates may underestimate actual tap water
consumption. The weight data included a mean and a distribution of
weight for male and females on a year-to-year basis. The means and
standard deviations of the life-long REFs derived from this analysis
are shown in Table III.D-1.
Table III.D-1.--Life-Long Relative Exposure Factors
------------------------------------------------------------------------
Community water Total water
consumption data consumption data
------------------------------------------------------------------------
Male........................ Mean = 0.60......... Mean = 0.73
s.d. = 0.61......... s.d. = 0.62
Female...................... Mean = 0.64......... Mean = 0.79
s.d. = 0.6.......... s.d. = 0.61
------------------------------------------------------------------------
c. Arsenic occurrence. EPA recently updated its estimates of
arsenic occurrence, and calculated separate occurrence distributions
for arsenic found in ground water and surface water systems. These
occurrence distributions were calculated for systems with arsenic
concentrations of 3 g/L or above. Arsenic occurrence estimates
are described in more detail in section III.C.
d. Risk distributions. In its 1999 report, ``Arsenic in Drinking
Water,'' the NRC analyzed bladder cancer risks using data from Taiwan.
In addition, NRC examined evidence from human epidemiological studies
in Chile and Argentina, and concluded that risks of bladder and lung
cancer had comparable risks to those ``in Taiwan at comparable levels
of exposure'' (NRC, 1999). The NRC also examined the implications of
applying different statistical analyses to the newly available
Taiwanese data for the purpose of characterizing bladder cancer risk.
While the NRC's work did not constitute a formal risk analysis, they
did examine many statistical issues (e.g., measurement errors, age-
specific probabilities, body weight, water consumption rate, comparison
populations, mortality rates, choice of model) and provided a starting
point for additional EPA analyses. The report noted that ``poor
nutrition, low selenium concentrations in Taiwan, genetic and cultural
characteristics, and arsenic intake from food'' were not accounted for
in their analysis (NRC, 1999, pg. 295). In the June 22, 2000 proposed
rule, EPA calculated bladder cancer risks and benefits using the
bladder cancer risk analysis from the NRC report (NRC, 1999). We also
estimated lung cancer benefits in a ``What If'' analysis based on the
statement in the 1999 NRC report that ``some studies have shown that
excess lung cancer deaths attributed to arsenic are 2-5 fold greater
than the excess bladder cancer deaths'' (NRC, 1999).
In July, 2000, a peer reviewed article by Morales et al. (2000) was
published, which presented additional analyses of bladder cancer risks
as well as estimates of lung and liver cancer risks for the same
Taiwanese population analyzed in the NRC report. EPA summarized and
analyzed the new information from the Morales et al. (2000) article in
a NODA published on October 20, 2000 (65 FR 63027; EPA, 2000m).
Although the data used were the same as used by the NRC to analyze
bladder cancer risk in their 1999 publication, Morales et al. (2000)
considered more dose-response models and evaluated how well they fit
the Taiwanese data for both bladder cancer risk and lung cancer risk.
Ten risk models were presented in Morales et al. (2000) used with and
without one of two comparison populations. After consultation with the
primary authors (Morales and Ryan), EPA chose Model 1 with no
comparison population for further analysis.
EPA believes that the models in Morales et al. (2000) without a
comparison population are more reliable than those with a comparison
population. Models with no comparison population estimate the arsenic
dose-response curve only from the study population. Models with a
comparison population include mortality data from a similar population
(in this case either all of Taiwan or part of southwestern Taiwan) with
low arsenic exposure. Most of the models with comparison populations
resulted in dose-response curves that were supralinear (higher than a
linear dose response) at low doses. The curves were ``forced down''
near zero dose because the comparison population consists of a large
number of people with low risk and low exposure. EPA believes, based on
discussions with the authors of Morales et al. (2000), that models with
a comparison population are less reliable, for two reasons. First,
there is no basis in data on arsenic's carcinogenic mode of action to
support a supralinear curve as being biologically plausible. To the
contrary, the conclusion of the NRC panel (NRC, 1999) was that the mode
of action data led one to expect dose responses that would be either
linear or less than linear at low dose. However, the NRC indicated that
available data are inconclusive and `` * * * do not meet EPA's 1996
stated criteria for departure from the default assumption of
linearity.'' (NRC, 1999)
[[Page 7007]]
Second, models that include comparison populations assume that the
study and comparison populations are the same in all important respects
except for arsenic exposure. Yet Morales et al. (2000) agree that
``[t]here is reason to believe that the urban Taiwanese population is
not a comparable population for the poor rural population used in this
study.'' Moreover, because of the large amount of data in the
comparison populations, the model results are sensitive to assumptions
about this group. Evidence that supports these arguments are that the
risks in the comparison groups are substantially lower than in
similarly exposed members of the study group and the shape of the
estimated dose-response changes sharply as a result. For these reasons,
EPA believes that the models without comparison populations are more
reliable than those with them. Of the models that did not include a
comparison population, EPA believes that Model 1 best fits the data,
based on the Akaike Information Criterion (AIC), a standard criterion
of model fit, applied to Poisson models. In Model 1, 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.
Morales et al. (2000) reported that two other models without
comparison populations also fit the Taiwan data well: Model 2, another
Poisson model with a nonparametric instead of quadratic age effect, and
a multi-stage Weibull (MSW) model. Under Model 2, the points of
departure for male and female bladder and lung cancer are from 1% to
11% lower than under Model 1, but within the 95% confidence bounds from
Model 1. Model 2 therefore implies essentially the same bladder and
lung cancer risks as Model 1. Under the MSW model, compared to Model 1,
points of departure are 45% to 60% higher for bladder cancer and for
female lung cancer, and 38% lower for male lung cancer. EPA did not
consider the MSW model for further analysis, because this model is more
sensitive to the omission of individual villages (Morales et al., 2000)
and to the grouping of responses by village (NRC, 1999), as occurs in
the Taiwanese data. However, if the MSW model were correct, it would
imply a 14% lower combined risk of lung and bladder cancers than Model
1, among males and females combined.
Considering all of these results, the Agency decided that the more
exhaustive statistical analysis of the data provided by Morales et al.
(2000), as analyzed by EPA, would be the basis for the new risk
calculations for the final rule (with further consideration of
additional risk analyses) and other pertinent information. The Agency
views the results of the alternative models described above as an
additional uncertainty which was considered in the decision concerning
the selection of the final MCL (see section III.F. of today's
preamble).
e. Estimated risk reductions. Estimated risk reductions for bladder
and lung cancer at various MCL levels were developed using Monte-Carlo
simulations. Monte-Carlo analysis is a technique for analyzing problems
where there are a large number of combinations of input values which
makes it impossible to calculate every possible result. A random number
generator is used to select input values from pre-defined
distributions. For each set of random numbers, a single scenario's
result is calculated. As the simulation runs, the model is recalculated
for each new scenario that continues until a stopping criteria is
reached. These simulations combined the distributions of relative
exposure factors (REFs), occurrence at or above 3 g/L, and
risks of bladder and lung cancer taken from the Morales et al. (2000)
article. The simulations resulted in upper-bound estimates of the
actual risks faced by populations exposed to arsenic concentrations at
or above 3 g/L in their drinking water.
f. Lower-bound analyses. Two adjustments were made to the risk
distributions resulting from the simulations described previously,
reflecting uncertainty about the actual arsenic exposure in the Taiwan
study area. First, the Agency made an adjustment to the lower bound
risk estimates to take into consideration the effect of exposure to
arsenic through water used in preparing food in Taiwan. The Taiwanese
staple foods were dried sweet potatoes and rice (Wu et al., 1989). Both
the 1988 EPA ``Special Report on Ingested Inorganic Arsenic'' report
(EPA,1988) and the 1999 NRC report assumed that an average Taiwanese
male weighed 55 kg and drank 3.5 liters of water daily, and that an
average Taiwanese female weighed 50 kg and drank 2 liters of water
daily. Using these assumptions, along with an assumption that Taiwanese
men and women ate one cup of dry rice and two pounds of sweet potatoes
a day, the Agency re-estimated risks for bladder and lung cancer, using
one additional liter water consumption for food preparation (i.e., the
water absorbed by hydration during cooking). This adjustment was
discussed and used in the October 20, 2000 NODA (65 FR 63027; EPA,
2000m).
Second, an adjustment was made to the lower-bound risk estimates to
take into consideration the relatively high arsenic concentration in
the food consumed in Taiwan as compared to the U.S. The food consumed
daily in Taiwan contains about 50 g of arsenic, versus about
10 g in the U.S. (NRC, 1999, pp. 50-51). Thus the total
consumption of inorganic arsenic (from food preparation and drinking
water) is considered, per kilogram of body weight, in the process of
these adjustments. To carry them out, the relative contribution of
arsenic in the drinking water that was consumed as drinking water, on a
g arsenic per kilogram body weight per day (g/kg/day)
basis, was compared to the total amount of arsenic consumed in drinking
water, drinking water used for cooking, and in food, on a g/
kg/day basis.
Other factors contributing to lower bound uncertainty include the
possibility of a sub-linear dose-response curve below the point of
departure. The NRC noted ``Of the several modes of action that are
considered most plausible, a sub-linear dose response curve in the low-
dose range is predicted, although linearity cannot be ruled out.''
(NRC, 1999). The recent Utah study (Lewis et al., 1999), described in
section V.G.1(b), provides some evidence that the shape of the dose-
response curve may well be sub-linear at low doses. Because sufficient
mode of action data were not available, an adjustment was not made to
the risk estimates to reflect the possibility of a sub-linear dose-
response curve. Additional factors contributing to uncertainty include
the use of village well data rather than individual exposure data,
deficiencies in the Taiwanese diet relative to the U.S. diet (selenium,
choline, etc.), and the baseline health status in the Taiwanese study
area relative to U.S. populations. The Agency did not make adjustments
to the risk estimates to reflect these uncertainties because applicable
peer-reviewed, quantitative studies on which to base such adjustments
were not available.
Estimated risk levels for bladder and lung cancer combined at
various MCL levels are shown in Tables III.D-2(a-c). The risk estimates
without adjustments for exposure uncertainty through cooking water and
food are shown Table III.D-2 (a). These estimates incorporate
occurrence data, water consumption data, and male and female risk
estimates. Lower bounds show estimates using community water
consumption data; upper bounds show estimates using total water
consumption data. Table III.D-2 (b) shows estimated risk
[[Page 7008]]
levels for bladder and lung cancer combined at various MCL levels with
adjustments for exposure uncertainty through cooking water and food.
These estimates incorporate occurrence data, water consumption data,
and male risk estimates, with lower bounds reflecting community water
consumption data and upper bounds reflecting total water consumption
data. There are no adjustments for other factors which contribute to
uncertainty, such as the use of village well data as opposed to
individual exposure data. Tablet III.D-2 (c) is a combination of Table
III.D-2(a) and Table III.D-2 (b), with the lower bounds taken from
Table III.D-3 (b), and the upper bounds taken from Table III.D-2 (a).
Thus Table III.D-2(c) reflects the range of estimates before and after
the exposure uncertainty adjustments for cooking water and for food,
along with the incorporation of water consumption data, occurrence
data, and cancer risk estimates. These estimates were used to estimate
the range of potential cases avoided at the various MCL levels.
The lower-bound risk estimates in Tables III.D-2(a-c) reflect the
following:
--The community (tap) water consumption from the EPA water consumption
study (EPA, 2000c)
--The occurrence distributions of arsenic in U.S. ground and surface
water systems
--Male risk estimates from Morales et al. (2000)
--Arsenic exposure from cooking water in Taiwan
--Arsenic exposure from food in Taiwan
The upper-bound risk estimates in Tables III.D-2(a-c) reflect the
following:
--The total water consumption estimates from the EPA water consumption
study (EPA, 2000c)
--The occurrence distributions of arsenic in U.S. ground and surface
water systems
--Male and female risk estimates from Morales et al. (2000)
Table III.D-2(a).--Cancer Risks for U.S. Populations Exposed At or Above MCL Options, After Treatment 1,2
[without adjustment for arsenic in food and cooking water]
----------------------------------------------------------------------------------------------------------------
Mean exposed 90th percentile exposed
MCL (g/L) population risk population risk
----------------------------------------------------------------------------------------------------------------
3............................................................. 0.93 -1.25 x 10-4 1.95 -2.42 x 10-4
5............................................................. 1.63 -2.02 x 10-4 3.47 -3.9 x 10-4
10............................................................ 2.41 -2.99 x 10-4 5.23 -6.09 x 10-4
20............................................................ 3.07 -3.85 x 10-4 6.58 -8.37 x 10-4
----------------------------------------------------------------------------------------------------------------
\1\ Actual risks could be lower, given the various uncertainties discussed, or higher, as these estimates assume
that the probability of illness from arsenic exposure in the U.S. is equal to the probability of death from
arsenic exposure among the arsenic study group.
\2\ The estimated risks are male and female risks combined.
Table III.D-2(b).--Cancer Risks for U.S. Populations Exposed At or Above MCL Options, After Treatment 1, 2
[without adjustment for arsenic in food and cooking water]
----------------------------------------------------------------------------------------------------------------
Mean exposed 90th percentile exposed
MCL (g/L) population risk population risk
----------------------------------------------------------------------------------------------------------------
3............................................................. 0.11 -0.13 x 10-4 0.22 -0.26 x 10-4
5............................................................. 0.27 -0.32 x 10-4 0.55 -0.62 x 10-4
10............................................................ 0.63 -0.76 x 10-4 1.32 -1.54 x 10-4
20............................................................ 1.1 -1.35 x 10-4 2.47 -2.89 x 10-4
----------------------------------------------------------------------------------------------------------------
\1\ Actual risks could be lower, given the various uncertainties discussed, or higher, as these estimates assume
that the probability of illness from arsenic exposure in the U.S. is equal to the probability of death from
arsenic exposure among the arsenic study group.
\2\ The estimated risks are for males.
Table III.D-2(c).--Cancer Risks for U.S. Populations Exposed At or Above MCL Options, After Treatment 1,2
[lower bound with food and cooking water adjustment, upper bound withough food and cooking water adjustment]
----------------------------------------------------------------------------------------------------------------
Mean exposed 90th percentile exposed
MCL (g/L) population risk population risk
----------------------------------------------------------------------------------------------------------------
3............................................................. 0.11 -1.25 x 10-4 0.22 -2.42 x 10-4
5............................................................. 0.27 -2.02 x 10-4 0.55 -3.9 x 10-4
10............................................................ 0.63 -2.99 x 10-4 1.32 -6.09 x 10-4
20............................................................ 1.1 -3.85 x 10-4 2.47 -8.37 x 10-4
----------------------------------------------------------------------------------------------------------------
\1\ Actual risks could be lower, given the various uncertainties discussed, or higher, as these estimates assume
that the probability of illness from arsenic exposure in the U.S. is equal to the probability of death from
arsenic exposure among the arsenic study group.
g. Cases avoided. The lower and upper bound risk estimates from
Table III.D-2(c) were applied to the exposed population to generate
cases avoided for CWSs serving less than a million customers. Because
the actual arsenic occurrence was known for the very large systems
(those serving over a million customers), their system-specific arsenic
occurrence distributions could be directly computed. The system-
specific arsenic distributions allowed direct calculation of avoided
cancer cases. The process, described in detail in the Economic Analysis
(EPA, 2000o), utilizes the same risk estimates from Morales et al.
(2000) that were used in deriving the number of cases avoided in
smaller CWSs. Cases avoided for NTNCWSs were also computed separately,
utilizing factors developed to
[[Page 7009]]
account for the intermittent nature of the exposure. These factors are
described in the Economic Analysis.
An upper-bound adjustment was made to the number of bladder cancer
cases avoided to reflect a possible lower mortality rate in Taiwan than
was assumed in the risk assessment process described earlier. We also
made this adjustment in the June 22, 2000 proposal. In the Taiwan study
area, information on arsenic-related bladder and lung cancer deaths was
reported. In order to use these data to determine the probability of
contracting bladder and lung cancer as a result of exposure to arsenic,
a probability of mortality, given the onset of arsenic-induced bladder
and lung cancer among the Taiwanese study population, must be assumed.
The study area in Taiwan is a section where arsenic concentrations in
the water are very high by comparison to those in the U.S., and an area
of low incomes and poor diets, where the availability and quality of
medical care is not of high quality, by U.S. standards. In its estimate
of bladder cancer risk, the Agency assumed that within the Taiwanese
study area, the probability of contracting bladder cancer was
relatively close to the probability of dying from bladder cancer (i.e.,
that the bladder cancer incidence rate was equal to the bladder cancer
mortality rate).
We do not have data on the rates of survival for bladder cancer in
the Taiwanese villages in the study at the time of data collection. We
do know that the relative survival rates for bladder cancer in
developing countries overall ranged from 23.5% to 66.1% in 1982-1992
(WHO, 1998). We also have some information on annual bladder cancer
mortality and incidence for the general population of Taiwan in 1996.
The age-adjusted annual incidence rates of bladder cancer for males and
females, respectively, were 7.36 and 3.09 per 100,000, with
corresponding annual mortality rates of 3.21 and 1.44 per 100,000
(correspondence from Chen to Herman Gibb, January 3, 2000). Assuming
that the proportion of males and females in the population is equal,
these numbers imply that the mortality rate for bladder cancer in the
general population of Taiwan, at present, is 45%. Since survival rates
have most likely improved over the years since the original Taiwanese
study, this number represents a lower bound on the survival rate for
the original area under study (i.e., one would not expect a higher rate
of survival in that area at that time). This has implications for the
bladder cancer risk estimates from the Taiwan data. If there were any
persons with bladder cancer who recovered and died from some other
cause, then our estimate underestimated risk; that is, there were more
cancer cases than cancer deaths. Based on the previous discussion, we
think bladder cancer incidence could be no more than two-fold bladder
cancer mortality; and that an 80% mortality rate would be plausible.
Thus, we have adjusted the upper bound of cases avoided, which is used
in the benefits analysis, to reflect a possible mortality rate for
bladder cancer of 80 percent. Because lung cancer mortality rates are
quite high, about 88% in the U.S. (EPA, 1998n), the assumption was made
that all lung cancers in the Taiwan study area resulted in fatalities.
The total number of bladder and lung cases avoided at each MCL is
shown in Table III.D-3. These cases avoided include CWSs and NTNCWSs
cases. The number of bladder and lung cancer cases avoided ranges from
57.2 to 138.3 at an MCL of 3 g/L, 51.1 to 100.2 at an MCL of 5
g/L, 37.4 to 55.7 at an MCL of 10 g/L, and 19.0 to
19.8 at an MCL of 20 g/L. The cases avoided were divided into
premature fatality and morbidity (i.e., illness) cases based on U.S.
mortality rates. In the U.S. approximately one out of four individuals
who is diagnosed with bladder cancer actually dies from bladder cancer.
The mortality rate for the U.S. is taken from a cost of illness study
recently completed by EPA (EPA, 1999j). For those diagnosed with
bladder cancer at the average age of diagnosis (70 years), the
probability for dying of that disease during each year post-diagnosis
was summed over a
20-year period to obtain the value of 26 percent. Mortality rates for
U.S. bladder cancer patients have decreased overall by 24% from 1973 to
1996. For lung cancer, mortality rates are much higher. The comparable
mortality rate for lung cancer in the U.S. is 88% (EPA, 1998n).
Table III.D-3.--Annual Total (Bladder and Lung) Cancer Cases Avoided
From Reducing Arsenic in CWSs and NTNCWS
------------------------------------------------------------------------
Total
Reduced Reduced cancer
Arsenic level (g/L) mortality morbidity cases
cases\1\ cases\1\ avoided
------------------------------------------------------------------------
3................................ 32.6-74.1 24.6-64.2 57.2-138.3
5................................ 29.1-53.7 22.0-46.5 51.1-100.2
10............................... 21.3-29.8 16.1-25.9 37.4-55.7
20............................... 10.2-11.3 8.5-8.8 19.0-19.8
------------------------------------------------------------------------
\1\ Based on U.S. mortality rates given in the text.
3. Sensitive Subpopulations
The 1996 SDWA amendments include specific provisions in section
1412(b)(3)(C)(i)(V) that require EPA to assess the effects of a
contaminant not just on the general population but on groups within the
general population such as infants, children, pregnant women, the
elderly, individuals with a history of serious illness, or other
subpopulations 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. The NRC subcommittee noted that
there is a marked variation in susceptibility to arsenic-induced toxic
effects that may be influenced by factors such as genetic polymorphisms
(especially in metabolism), life stage at which exposures occur, sex,
nutritional status, and concurrent exposures to other agents or
environmental factors. The NRC report concluded that there is
insufficient scientific information to permit separate cancer risk
estimates for potential subpopulations such as pregnant women,
lactating women, and children and that factors that influence
sensitivity to or expression of arsenic-associated cancer and noncancer
effects need to be better characterized. EPA agrees with the NRC that
there is not enough information to make risk conclusions on any
specific subpopulations.
4. Risk Window
EPA has historically considered 10-4 to 10-6
as a target risk range protective of public health in its drinking
water program. However, the risk-range
[[Page 7010]]
represents a policy goal for EPA, and is not a statutory factor in
setting an MCL. Note that the procedure EPA uses to estimate such risks
provides an upper-bound estimate. In the case of arsenic, EPA performed
a benefit-cost analysis as required by the statute. This analysis is
discussed in more detail in section III.F.
E. What Are the Costs and Benefits at 3, 5, 10, and 20 g/L?
In accordance with section 1412 (b)(3)(C) of SDWA, EPA must analyze
the costs and benefits of a proposed NPDWR. To comply with this
provision, EPA included the complete analysis in the proposed rule.
Also, in accordance with Executive Order 12866, Regulatory Planning and
Review, EPA must estimate the costs and benefits of the arsenic rule in
an Economic Analysis in conjunction with publishing the final rule. EPA
has prepared an Economic Analysis to comply with the requirements of
this Order. This section provides a summary of the information from the
Arsenic Economic Analysis (EPA, 2000o).
1. Summary of Cost Analysis
National cost estimates of compliance with the arsenic rule were
derived from estimates of utility treatment costs, monitoring and
reporting costs, and start-up costs for both CWS and NTNCWSs. Utility
treatment costs were derived using occurrence data, treatment train
unit costs, and decision trees. The occurrence data provide a measure
of the number of systems that would need to install treatment in each
size category. The treatment train unit cost estimates provide a
measure of how much a technology will cost to install. Decision trees
vary by system size and are used as a prediction of the treatment
technology trains facilities would likely install to comply with
options considered for the revised arsenic standard. Detailed
descriptions of the methodologies used in determining the costs of this
rule are found in the ``Technologies and Cost for Removal of Arsenic in
Drinking Water'' document (EPA, 2000t) and also the ``Arsenic Economic
Analysis'' (EPA, 2000o), both of which are in the docket for this final
rulemaking.
a. Total national costs. Under the MCL of 10 g/L, the
Agency estimates that total national costs to CWSs are $172.3 million
(1999 dollars) annually at a 3% discount rate. This total national cost
includes annual treatment costs ($169.6 million), annual monitoring and
administrative costs ($1.8 million), and annual State costs ($0.9
million). Assuming a 7% discount rate, total national costs to CWSs are
estimated at $196.6 million annually.
Total national costs to NTNCWSs are estimated at $8.1 million
annually at a 3% discount rate. This includes annual treatment costs
($7.0 million), annual monitoring and administrative costs ($0.9
million), and annual State costs ($0.1 million). Total national costs
to NTNCWSs, assuming a 7% discount rate, are estimated at $9.1 million
annually.
Table III.E-1 shows the total national cost breakdown for the
arsenic MCL and also for three other arsenic levels considered in the
proposed rule. Expected system costs include treatment costs,
monitoring costs, and administrative costs of compliance. State costs
include monitoring and administrative costs of implementation. As
expected, aggregate arsenic compliance costs increase with decreasing
arsenic MCL levels as more systems are affected.
Table III.E-1.--Total Annual National System and State Compliance Costs
[$ millions, 1999]
----------------------------------------------------------------------------------------------------------------
CWS NTNCWS Total
Discount rate -----------------------------------------------------------------------------
3 percent 7 percent 3 percent 7 percent 3 percent 7 percent
----------------------------------------------------------------------------------------------------------------
MCL = 3 g/L
----------------------------------------------------------------------------------------------------------------
System Costs...................... $668.1 $759.5 $28.2 $31.0 $696.3 $790.4
Treatment......................... 665.9 756.5 27.2 29.6 693.1 786.0
Monitoring/Administrative......... 2.2 3.0 1.0 1.4 3.2 4.4
State Costs....................... 1.4 1.6 0.1 0.2 1.5 1.7
-----------------------------------------------------------------------------
Total \1\..................... 669.4 761.0 28.3 31.1 697.8 792.1
----------------------------------------------------------------------------------------------------------------
MCL = 5 g/L
----------------------------------------------------------------------------------------------------------------
System Costs...................... 396.4 451.1 17.3 18.9 413.5 470.2
Treatment......................... 394.4 448.3 16.3 17.6 410.6 466.1
Monitoring/Administrative......... 2.0 2.8 1.0 1.3 2.9 4.1
State Costs....................... 1.1 1.3 0.1 0.2 1.2 1.4
----------------------------------------------------------------------------------------------------------------
Total \1\..................... 397.5 452.5 17.3 19.1 414.8 471.7
----------------------------------------------------------------------------------------------------------------
Final MCL = 10 g/L
----------------------------------------------------------------------------------------------------------------
System Costs...................... 171.4 195.5 7.9 8.9 179.4 204.4
Treatment......................... 169.6 193.0 7.0 7.6 176.7 200.6
Monitoring/Administrative......... 1.8 2.5 0.9 1.3 2.7 3.8
State Costs....................... 0.9 1.0 0.1 0.2 1.0 1.2
----------------------------------------------------------------------------------------------------------------
Total \1\..................... 172.3 196.6 8.1 9.1 180.4 205.6
----------------------------------------------------------------------------------------------------------------
MCL = 20 g/L
----------------------------------------------------------------------------------------------------------------
System Costs...................... 62.4 71.4 3.5 4.1 65.9 75.5
Treatment......................... 60.7 69.0 2.6 2.8 63.3 71.8
Monitoring/Administrative......... 1.7 2.4 0.9 1.3 2.6 3.7
[[Page 7011]]
State Costs....................... 0.7 0.8 0.1 0.2 0.9 1.0
-----------------------------------------------------------------------------
Total \1\..................... 63.2 72.3 3.6 4.2 66.8 76.5
----------------------------------------------------------------------------------------------------------------
\1\ Total may not match detail due to rounding.
b. Household costs. Table III.E-2 shows mean annual costs per
household for those households that are served by systems that may need
to treat under today's rule. As discussed in Table III.C-6 of today's
preamble and Table 8-2 of the Economic Analysis, of the approximately
74,000 systems that are covered by today's rule, EPA estimates that
only about 3,433 of these systems will require treatment. Table III.E-2
refers only to the households served by systems expected to need
treatment. The average household cost increase resulting from today's
rule is $31.85. However, due to economies of scale, costs per household
are higher in the smaller size categories, and lower in the larger size
categories. For today's rule (10 g/L), costs are expected to
be $326.82 per household for systems serving 100 people, and $162.50
per household for systems serving 101-500 people. Costs per households
in systems larger than those are substantially lower: From $70.72 to
$0.86 per household. As shown in Table III.E-2, the costs per household
do not vary dramatically across MCL options although Table III.E-1
shows that total national costs are significantly different. This
divergence is attributable to the total number of households affected
by each MCL level and not the cost of treatment. For example,
approximately eleven million households would be affected by an MCL of
3 g/L compared to approximately three million affected by the
today's final rule MCL of 10 g/L. In addition, the household
costs change relatively little among MCL options because while each
progressively lower MCL option brings in a larger number of systems
subject to the rule, the majority of those systems generally need only
minimal removal of arsenic. This fact offsets, to an extent, the
increased costs as a result of more systems covered at lower MCL
options. A more detailed discussion of household costs can be found in
Chapter 6 of the ``Arsenic Economic Analysis'' document (EPA, 2000o).
Table III.E-2.--Mean Annual Costs Per Household
[in 1999 dollars] \1\
----------------------------------------------------------------------------------------------------------------
3 g/L m>g/L m>g/L m>g/L
----------------------------------------------------------------------------------------------------------------
100......................................................... $317.00 $318.26 $326.82 $351.15
101-500..................................................... 166.91 164.02 162.50 166.72
501-1,000................................................... 74.81 73.11 70.72 68.24
1,001-3.300................................................. 63.76 61.94 58.24 54.36
3.301-10,000................................................ 42.84 40.18 37.71 34.63
10,001-50,000............................................... 38.40 36.07 32.37 29.05
50,001-100,000.............................................. 31.63 29.45 24.81 22.63
100,001-1,000,000........................................... 25.29 23.34 20.52 19.26
>1,000,000.................................................. 7.41 2.79 0.86 0.15
All categories.............................................. 41.34 36.95 31.85 23.95
----------------------------------------------------------------------------------------------------------------
\1\ Only households served by those systems expected to install treatment.
2. Summary of Benefits Analysis
Arsenic ingestion has been linked to a multitude of health effects,
both cancerous and non-cancerous. These health effects include cancer
of the bladder, lungs, skin, kidney, nasal passages, liver, and
prostate. Arsenic ingestion has also been attributed to cardiovascular,
pulmonary, immunological, and neurological, endocrine effects. A
complete list of the arsenic-related health effects reported in humans
is discussed in section III. D of this preamble. Current research on
arsenic exposure has only been able to provide enough information to
conduct a quantitative assessment of bladder and lung cancers. The
other health effects and possible non-health benefits remain
unquantified in this analysis but are discussed qualitatively. It is
important to note that if the Agency were able to quantify additional
arsenic-related health effects and non-health effects, the quantified
benefits estimates may be significantly higher than the estimates
presented in this analysis. In addition, the SDWA amendments of 1996
require that EPA fully consider both quantifiable and non-quantifiable
benefits that result from drinking water regulations and has done this
for today's arsenic rule.
a. Primary analysis. Quantifiable benefits. Although arsenic in
drinking water has been associated with numerous health effects (see
section III.D), the quantified benefits that result from today's rule
are associated only with reductions in arsenic-related bladder and lung
cancers. A complete discussion of risk assessment methodology and
assumptions can be found in Chapter 5 of the ``Arsenic Economic
Analysis'' document (EPA, 2000o).
The quantified benefits for today's rule for both CWSs and NTNCWSs
range from $140 million to $198 million and consider both lower- and
upper-bound risk levels. Specifically, the benefits to the CWSs are
approximately $138.2 million to $193.2 million and $1.4 million to $4.5
million for NTNCWSs. Table III.E-3 shows the complete range of
quantified benefits for the other MCL levels considered by the Agency.
Section III.D.2. of this preamble explains the derivation of the upper-
and lower-bound estimates
[[Page 7012]]
In order to monetize the benefit from the bladder and lung cancers
cases avoided, the Agency used two different values. First, a value of
statistical life (VSL) estimate was applied to those cancer cases that
result in a mortality. EPA assumed a 26% mortality rate for bladder
cancer and an 88% mortality rate for lung cancer (EPA, 1999j; EPA,
1998n). The current VSL value used by the Agency is $6.1 million, in
1999 dollars. This value of $6.1 million does not reflect any
adjustments to account for national real income growth that occurred
subsequent to the completion of the wage-risk studies on which EPA's
VSL estimate is derived. Were the Agency to adjust the VSL to account
for this growth in real income, the VSL would be approximately $6.77
million (assuming a 1.0 income elasticity).
Second, a willingness-to-pay value (WTP) is used to monetize the
cancer cases that do not result in a mortality. The WTP value for
avoiding a non-fatal cancer is not currently available; therefore, the
Agency used a WTP estimate to reduce a case of chronic bronchitis as a
proxy. The use of this value may understate the true benefit if the WTP
to avoid a nonfatal cancer is greater than the WTP to avoid a case of
chronic bronchitis. The mean value of this WTP estimate is $607,000 (in
1999 dollars). A complete discussion of the VSL and WTP values and how
they are calculated can be found in Chapter 5 for the ``Arsenic
Economic Analysis'' document (EPA, 2000o).
--Non-quantifiable benefits. There are a number of important non-
quantified benefits that EPA considered in its analysis. Chief among
these are certain health impacts known to be caused by arsenic, though,
while they may be substantial, the extent to which these impacts occur
at levels below 50 g/L is unknown. These additional health
effects include cancers, other than bladder and lung cancers, as well
as non-cancer health effects. In addition, EPA has identified non-
health benefits that may result from today's rule, which are discussed
next.
EPA was not able to quantify many of the health effects potentially
associated with arsenic due to data limitations. These health effects
include other cancers such as skin and prostate cancer and non-cancer
endpoints such as cardiovascular, pulmonary, and neurological impacts.
These health effects and the relevant studies linking these health
effects to arsenic in drinking water are discussed in section III.D. of
today's rule. For example, a number of epidemiologic studies conducted
in several countries (e.g., Taiwan, Japan, England, Hungary, Mexico,
Chile, and Argentina) report an association between arsenic in drinking
water and skin cancer in exposed populations. Studies conducted in the
U.S. have not demonstrated an association between inorganic arsenic in
drinking water and skin cancer. However, these studies may not have
included enough people in their design to detect these types of
effects.
Other potential benefits not quantified or monetized in today's
rule include reduced uncertainty about becoming ill from consumption of
arsenic in drinking water and the ability for some treatment
technologies to eliminate multiple contaminants. The reduced
uncertainty concept depends on several factors including consumer's
degree of risk aversion, their perceptions about the drinking water
quality (degree to which they will be affected by the regulatory
action), and the expected probability and severity of human heath
effects associated with arsenic contamination of drinking water.
Another non-quantified benefit is the effect on those systems that
install treatment technologies that can address multiple contaminants.
For example, membrane systems, such as reverse osmosis, can be used for
arsenic removal but can also remove many other contaminants that EPA is
in the process of regulating or considering regulating. Therefore, by
installing a reverse osmosis system, a system may not have to make any
additional changes to comply with these future regulations.
Table III.E-3.--Estimated Benefits From Reducing Arsenic in Drinking
Water
[$ millions 1999]
------------------------------------------------------------------------
Total quantified Potential non-quantified
Arsenic level (g/ health benefits health benefits includes
L) \1\ reductions in:
------------------------------------------------------------------------
3......................... $213.8-$490.9 Skin Cancer.
5......................... $191.1-$355.6 Kidney Cancer.
10........................ $139.6-$197.7 Cancer of the
Nasal Passages.
20........................ $66.2-$75.3 Liver Cancer.
Prostate Cancer.
Cardiovascular
Effects.
Pulmonary
Effects.
Immunological
Effects.
Neurological
Effects.
Endocrine
Effects.
------------------------------------------------------------------------
\1\ Benefits from reduction in bladder and lung cancer. The range
represents both a lower and upper bound risk as discussed in section
III. D. of this preamble.
b. Sensitivity analysis on benefits valuation. For the final
rulemaking analysis, some commenters have argued that the Agency should
consider an assumed time lag or latency period in its benefits
calculations. The term ``latency'' can be used in different ways,
depending on the context. For example, health scientists tend to define
latency as the period beginning with the initial exposure to the
carcinogen and ending when the cancer is initially manifested (or
diagnosed), while others consider latency as the period between
manifestation of the cancer and death. Latency, in this case, refers to
the difference between the time of initial exposure to environmental
carcinogens and the actual mortality. Use of such an approach might
reduce significantly the present value of health risk reduction
benefits estimates.
In the proposed arsenic rule, the Agency included qualitative
language on the latency issue, including descriptions of other
adjustments which may influence the estimate of economic benefits
associated with avoided cancer fatalities. The Agency also agreed to
ask the SAB to conduct a review of the benefits' transfer issues and
possible adjustment factors associated with economic valuation of
mortality risks. A summary of the SAB's recommendations is shown in the
following section.
[[Page 7013]]
c. SAB recommendations. EPA brought this issue before the
Environmental Economics Advisory Committee (EEAC) of EPA's SAB in a
meeting held on February 25, 2000 in Washington, DC. The SAB submitted
a final report on its findings and recommendations to EPA on July 27,
2000. The Panel's report made a number of recommendations on the
adjustment factors and benefit-cost analysis in general. A copy of the
final SAB report (EPA, 2000j) is in the record for this rulemaking.
The SAB Panel noted that benefit-cost analysis, as described in the
Agency's Guidelines for Preparing Economic Analysis (EPA, 2000k), is
not the only analytical tool nor is efficiency the only appropriate
criterion for social decision making. The SAB Panel also stated that it
is important to carry out such analyses in an unbiased manner with as
much precision as possible. In its report, the SAB recommended that the
Agency continue to use a wage-risk-based VSL as its primary estimate;
any appropriate adjustments that are made for timing (e.g., latency)
and income growth should be part of the Agency's main analysis while
any other proposed adjustments should be accounted for in sensitivity
analyses to show how results would change if the VSL were adjusted for
some of the major differences in the characteristics of the risk and of
the affected populations. The SAB recommended including only
adjustments for latency and income growth in the main analysis because
it did not believe any of the other proposed adjustments were
adequately supported in the literature at the present time.
Specifically, the SAB report recommended that (1) Health benefits
brought about by current policy initiatives (i.e., after a latency
period) should be discounted to present value using the same rate that
is used to discount other future benefits and costs in the primary
analysis; and any other proposed adjustments should be accounted for in
a sensitivity analysis including adjustments to the VSL for a ``cancer
premium,'' voluntariness and controllability, altruism, risk aversion,
and ages of the affected population. No adjustment should be made to
the VSL to reflect health status of persons whose cancer risks are
reduced. (2) Estimates of VSLs accruing in future years should be
adjusted in the primary analysis to reflect anticipated income growth,
using a range of income elasticities.
After considering the SAB's recommendations, EPA has developed a
sensitivity analysis of the latency structure and associated benefits
for the arsenic rule, as described in the next section and in the
Economic Analysis for the final rule. This analysis consists of health
risk reduction benefits that reflect adjustments for discounting,
incorporation of a range of latency period assumptions, adjustments for
growth in income, and incorporation of other factors such as
voluntariness and controllability. Although the SAB recommended
accounting for latency in a primary benefits analysis, the Agency
believes that, in the absence of any sound scientific evidence on the
duration of particular latency periods for arsenic related cancers,
discounted benefits estimates for arsenic are more appropriately
accounted for in a sensitivity analysis. Sensitivity analyses are
generally reserved for examining the effects of accounting for highly
uncertain factors, such as the estimation of latency periods, on health
risk reduction benefits estimates.
Defining a latency period is highly uncertain because the length of
the latency period is often poorly understood by health scientists. In
some cases, information on the progression of a cancer is based on
animal studies, and extrapolation to humans is complex and uncertain.
Even when human studies are available, the dose considered may differ
significantly from the dose generally associated with drinking water
contaminants (e.g., involve a high level of exposure over a short time
period, rather than a long term, low level of exposure). The magnitude
of the dose, may in turn, affect the resulting latency period.
Information on latency may be unavailable in many cases or, if
available, may be highly uncertain and vary significantly across
individuals. The Agency recognizes, however, that despite significant
uncertainty in the latency period associated with arsenic exposure
through drinking water, it is unlikely that all cancer reduction
benefits would be realized immediately upon exposure reduction. To the
extent that there are delays due to latency in the realization of these
benefits, monetized cancer reduction benefits would be discounted;
although, as discussed above, this may be offset by other adjustments.
d. Analytical approach. For the latency sensitivity analysis, the
health benefits have been broken into separate treatments of morbidity
and mortality. The mortality component of the total benefits is
examined in this analysis because a cancer latency period (i.e., the
time period between initial exposure to environmental carcinogens and
the actual fatality) impacts arsenic-related fatalities to a greater
extent than arsenic-related morbidity. For purposes of this analysis,
the Agency examined the impacts of various latency period assumptions,
adjustments for income growth, and incorporation of other adjustments
such as a voluntariness and controllability, on bladder and lung cancer
fatalities associated with arsenic in drinking water (EPA, 2000k).
Because the latency period for arsenic related bladder and lung
cancers is unknown, EPA has assumed a range of latency periods from 5
to 20 years. While both lung and bladder cancer have relatively long,
average latencies, the lower end of the latency period is substantially
less. As can be seen by inspection of the Surveillance, Epidemiology,
and End Results (SEER) data of the National Cancer Institute,
significant incidence of both cancers occurs in individuals in the 15-
19 year old age groups (NCI, 2000). This strongly indicates a short
latency period for whatever the cause of the cancer may have been.
Moreover, the mode of action for arsenic is suspected to be one
that operates at a late stage of the cancer process that may advance
the expression of cancers initiated by other causes (sometimes referred
to as ``promoting out'' the cancerous effect). Therapeutic treatment
with the drug cyclophosphamide, which causes cell toxicity, has been
seen to induce bladder cancer in as little as 7 months to 15 years in
affected patients. This was of course a high dose treatment, but the
example serves to illustrate the ability of an agent to advance the
development of cancer.
For these reasons, we believe latency periods of 5, 10, and 20
years serve as reasonable approximations, in the absence of definitive
data on arsenic-induced cancers, of the latency periods for the
sensitivity analysis.
Table III.E-4 shows the sensitivity of the primary analysis VSL
estimate ($6.1 million, 1999 dollars) to changes in latency period
assumptions and also with the incorporation of an adjustment to reflect
changes in WTP based on real income growth and other adjustment
factors. As is shown in Table III.E-4, the adjusted VSL is greater than
the primary VSL ($6.77 million versus $6.1 million) at an income
elasticity of 1.0, with adjustments for income growth only. Assuming a
3% discount rate, the lowest adjusted VSL value ($3.44 million) is
yielded over a 20-year latency period that includes discounting and
income growth only (income elasticity = 0.22). Assuming a 7% discount
rate, the highest adjusted VSL is also $6.77 million (adjusted for
income growth only (income elasticity = 1.0)). The lowest adjusted VSL
is $1.61 million (discounted over 20 years).
[[Page 7014]]
Table III.E-4.-- Sensitivity of the Primary VSL Estimate to Changes in Latency Period Assumptions, Income
Growth, and Other Adjustments
[$ millions, 1999]
----------------------------------------------------------------------------------------------------------------
Latency period (Years)
Adjustment factor ----------------------------------------------------------------
5 10 20
----------------------------------------------------------------------------------------------------------------
3% Discount Rate
----------------------------------------------------------------------------------------------------------------
Primary Analysis (No VSL Adjustment)........... 6.1 6.1 6.1
Adjusted for Income Growth: \1\
elasticity = 0.22.......................... 6.22 6.22 6.22
elasticity = 1.0........................... 6.77 6.77 6.77
Adjusted for Income Growth \1\ and Discounting:
elasticity = 0.22.......................... 5.37 4.63 3.44
elasticity = 1.0........................... 5.84 5.04 3.75
Adjusted for Income Growth,\1\ Discounting, and
7% Increase for Voluntariness and
Controllability;
elasticity = 0.22.......................... 5.74 4.95 3.69
elasticity = 1.0........................... 6.25 5.39 4.01
Break-Even for Other Characteristics (as a
percentage of the primary VSL estimate);
elasticity = 0.22.......................... 6 percent 19 percent 40 percent
elasticity = 1.0........................... -2 percent 12 percent 34 percent
----------------------------------------------------------------------------------------------------------------
7% Discount Rate
----------------------------------------------------------------------------------------------------------------
Primary Analysis (No VSL Adjustment)........... 6.1 6.1 6.1
Adjusted for Income Growth:\1\
elasticity = 0.22.......................... 6.22 6.22 6.22
elasticity = 1.0........................... 6.77 6.77 6.77
Adjusted for Income Growth \1\ and Discounting:
elasticity = 0.22.......................... 4.44 3.16 1.61
elasticity = 1.0........................... 4.83 3.44 1.75
Adjusted for Income Growth, \1\ Discounting,
and 7% Increase for Voluntariness and
Controllability:
elasticity = 0.22.......................... 4.75 3.38 1.72
elasticity = 1.0........................... 5.17 3.68 1.87
----------------------------------------------------------------------------------------------------------------
Break-Even for Other Characteristics (as a
percentage of the primary VSL estimate):
elasticity = 0.22.......................... 22 percent 45 percent 72 percent
elasticity = 1.0........................... 15 percent 40 percent 69 percent
----------------------------------------------------------------------------------------------------------------
\1\ This adjustment reflects the change in WTP based on real income growth from 1990 to 1999.
The first row of both the 3% and 7% discount rate panels in Table
III.E-4 shows the VSL used in the primary analysis. Because this value
has not been adjusted for discounting over an assumed and unknown
latency period, this value does not deviate from the original $6.1
million used in the primary benefits analysis. The second and third
rows of both the 3 and 7 percent panels show the adjustments to the
primary VSL to account for changes in WTP for fatal risk reductions
associated with real income growth from 1990 to 1999. As real income
grows, the WTP to avoid fatal risks is also expected to increase at a
rate corresponding to the income elasticity of demand, as discussed
below. This income growth, from the years 1990 to 1999, accounts for
the differences in incomes of the VSL study population versus the
population affected by the arsenic rule. This does not include any
income adjustments over a latency period because of methodological
issues that have not yet been resolved. However, pending the resolution
of these issues, EPA may include an adjustment for income growth over a
latency period in future analyses, as recommended by the SAB.
The fourth and fifth rows of both the 3% and 7% panels illustrates
the impacts of adjusting the primary VSL for discounting and WTP
changes based on real income growth over a range of assumed latency
periods. As is shown in Table III.E-4, this value decreases from $5.84
million assuming a five-year latency period to $3.75 million assuming a
20-year latency period (at a 3% discount rate and income elasticity of
1.0). At a 7% discount rate, this value decreases from $4.83 million to
$1.75 million.
The sixth and seventh rows of the 3% and 7% panels illustrate the
effects of incorporating a 7% increase for voluntariness and
controllability. The 7% adjustment is based on a study by Cropper and
Subramanian (1999) that indicates individuals may place a slightly
higher Willingness to Pay (WTP) on risks where exposure is neither
voluntary nor controllable by the individual.
In adjusting for WTP changes based on real income growth, EPA used
a range of income elasticities from the economics literature. Income
elasticity is the % change in demand for a good (in this case, WTP for
fatal risk reductions) for every 1% change in income. For example, an
income elasticity of 1.0 implies that a 10 percent higher income level
results in a 10% higher WTP for fatal risk reductions. In a recent
study (EPA, 2000l), EPA reviewed the literature related to the income
elasticity of demand for the prevention of fatal health impacts. Based
on data from cross-sectional studies of wage premiums, a range of
elasticity estimates for serious health impacts was developed, ranging
from a lower-end estimate of 0.22 to an upper-end estimate of 1.0.
There are several other characteristics that differ between the VSL
estimates used in the primary analysis and an ideal estimate specific
to the case of cancer risks from arsenic. These might
[[Page 7015]]
include a cancer premium, differences in risk aversion, altruism, age
of the individual affected, and a morbidity component of the VSL
mortality estimate. Very little empirical information is available on
the impact that these characteristics have on VSL estimates so they are
not accounted for directly in this sensitivity analysis. A more
complete discussion of the other characteristics identified by
economists as having a potential impact on willingness to pay to reduce
mortality risks can be found in chapter seven of the Agency's
``Guidelines for Preparing Economic Analyses'' (EPA 2000k), which is
available in the docket for this final rulemaking.
However, it is possible to use a different type of analysis to
address the question: what would the impact on VSL of these additional
characteristics need to be to produce the $6.1 million VSL used in the
primary benefits analysis? (See primary benefits analysis in section
III.E.2.a of today's rule.) The last two rows of the 3% and 7% panels
of Table III.E-4 attempt to answer this question in percentage terms.
For example, at a 3% discount rate over a 10-year latency period,
income elasticity of 1.0, and a 7% adjustment for controllability and
voluntariness, a factor of 12% (as shown in the bottom row of the 3%
panel of Table III.E-4) indicates that if accounting for these
characteristics would increase VSL by more than 12% then the primary
analysis will tend to understate the value of risk reductions. If
accounting for these characteristics would not increase VSL by at least
12%, then the primary analysis may overstate benefits (a negative %
indicates that the primary analysis understates benefits unless the
combined impact of these additional characteristics actually reduces
VSL estimates).
Some researchers believe that the value of some of these
characteristics will substantially add to the unadjusted VSL (one study
suggests that a cancer premium alone may be worth an additional 100% of
primary VSL value (Revesz, 1999)). Some researchers also believe that
some of these characteristics have a negative effect on VSL, suggesting
that some of these factors offset one another. Until we know more about
these various factors we cannot explicitly make adjustments to existing
VSL estimates. The SAB noted in its report that these characteristics
require more empirical research prior to incorporation into the
Agency's primary benefits analysis, but could be explored as part of a
sensitivity analysis.
e. Results. Table III.E-5 illustrates the impacts of changes in VSL
adjustment factor assumptions on the estimated benefits for the range
of fatal bladder and lung cancer cases avoided in the final arsenic
rule, assuming a 3% discount rate. The results of this analysis at a 7%
discount rate are given in Table III.E-6. These results were calculated
by applying the adjusted VSL from Table III.E-4 to the lower- and
upper-bound estimates of fatal bladder and lung cancer cases avoided as
shown in Table III.E-3 in section III.D.2 of today's rule. For purposes
of this sensitivity analysis, EPA presented combined bladder and lung
cancer cases avoided in Tables III.E-5 and III.E-6. Health risk
reduction benefits attributable to reduced arsenic levels in both CWSs
and NTNCWSs are presented in these tables as well.
It is important to note that the monetized benefits estimates shown
in this section reflect quantifiable benefits only. As shown in section
III.E.2.a, there may be a number of nonquantifiable benefits associated
with regulating arsenic in drinking water. Were EPA able to quantify
some of the currently nonquantifiable health effects and other benefits
associated with arsenic regulation, monetized benefits estimates would
be higher than what is shown in the table. A more complete discussion
of how risks from arsenic in drinking water and the corresponding
health benefits were calculated is provided in the ``Arsenic Economic
Analysis'' (EPA, 2000o), which is available in the docket for this
final rulemaking.
Table III.E-5.--Sensitivity of Combined Annual Bladder and Lung Cancer Mortality Benefits Estimates to Changes
in VSL Adjustment Factor Assumptions
[$ millions, 1999, 3% discount rate] \1\
----------------------------------------------------------------------------------------------------------------
Arsenic Level (g/L) 3 5 10 20
----------------------------------------------------------------------------------------------------------------
5-Year Latency Period Assumption
----------------------------------------------------------------------------------------------------------------
Primary Analysis (No VSL Adjustment)........................ 199-452 176-328 130-182 62-69
Adjusted for Income Growth \2\
E = 0.22................................................ 203-461 181-334 133-186 63-70
E = 1.0................................................. 221-502 197-364 144-202 69-77
Adjusted for Income Growth\2\ and Discounting:
E = 0.22................................................ 175-398 156-288 114-160 55-61
E = 1.0................................................. 190-433 170-314 124-174 60-66
Adjusted for Income Growth,\2\ Discounting, and 7% Increase
for Voluntariness and Controllability:
E = 0.22................................................ 187-425 167-308 122-171 59-65
E = 1.0................................................. 204-463 182-336 133-186 64-71
----------------------------------------------------------------------------------------------------------------
10-Year Latency Period Assumption
----------------------------------------------------------------------------------------------------------------
Primary Analysis (No VSL Adjustment)........................ 199-452 176-328 130-182 62-69
Adjusted for Income Growth: \2\
E = 0.22................................................ 203-461 181-334 133-186 63-70
E = 1.0................................................. 221-502 197-364 144-202 69-77
Adjusted for Income Growth,\2\ and Discounting:
E = 0.22................................................ 151-343 135-249 99-138 47-52
E = 1.0................................................. 164-373 147-271 107-150 51-57
Adjusted for Income Growth,\2\ Discounting, and 7% Increase
for Voluntariness and Controllability:
E = 0.22................................................ 161-367 144-266 105-148 50-56
E = 1.0................................................. 176-399 157-289 115-161 55-61
[[Page 7016]]
20-Year Latency Period Assumption
----------------------------------------------------------------------------------------------------------------
Primary Analysis (No VSL Adjustment)........................ 199-452 176-328 130-182 62-69
Adjusted for Income Growth: \2\
E = 0.22................................................ 203-461 181-334 133-186 63-70
E = 1.0................................................. 221-502 197-364 144-202 69-77
Adjusted for Income Growth \2\ and Discounting:
E = 0.22................................................ 112-255 100-185 73-103 35-39
E = 1.0................................................. 122-278 109-201 80-112 38-42
Adjusted for Income Growth,\2\ Discounting, and 7% Increase
for Voluntariness and Controllability:
E = 0.22................................................ 120-273 107-198 79-110 38-42
E = 1.0................................................. 131-297 117-215 85-119 41-45
----------------------------------------------------------------------------------------------------------------
\1\ The lower- and upper-bound benefits estimates correspond to the lower- and upper-bound risk estimates and
cancer cases avoided as shown in section III.D.2 of this preamble.
\2\ This adjustment reflects the change in WTP based on real income growth from 1990 to 1999. E = income
elasticity.
Table III.E-6.--Sensitivity of Combined Annual Bladder and Lung Cancer Mortality Benefits Estimates to Changes
in VSL Adjustment Factor Assumptions
[$ millions, 1999, 7% discount rate] \1\
----------------------------------------------------------------------------------------------------------------
Arsenic Level (g/L) 3 5 10 20
----------------------------------------------------------------------------------------------------------------
5-Year Latency Period Assumption
----------------------------------------------------------------------------------------------------------------
Primary Analysis (No VSL Adjustment)........................ 199-452 178-328 130-182 62-69
Adjusted for Income Growth: \2\
E = 0.22................................................ 203-461 181-334 133-186 63-70
E = 1.0................................................. 221-502 197-364 144-202 69-77
Adjusted for Income Growth,\2\ and Discounting:
E = 0.22................................................ 145-329 129-238 95-132 45-50
E = 1.0................................................. 157-358 141-259 103-144 50-55
Adjusted for Income Growth,\2\ Discounting, and 7% Increase
for Voluntariness and Controllability:
E = 0.22................................................ 155-352 138-255 102-142 49-54
E = 1.0................................................. 168-383 150-278 110-154 53-58
----------------------------------------------------------------------------------------------------------------
10-Year Latency Period Assumption
----------------------------------------------------------------------------------------------------------------
Primary Analysis (No VSL Adjustment)........................ 199-452 178-328 130-182 62-69
Adjusted for Income Growth: \2\
E = 0.22................................................ 203-461 181-334 133-186 63-70
E = 1.0................................................. 221-502 197-364 144-202 69-77
Adjusted for Income Growth \2\ and Discounting:
E = 0.22................................................ 103-234 92-170 67-94 32-36
E = 1.0................................................. 112-255 100-185 73-103 35-39
Adjusted for Income Growth,\2\ Discounting, and 7% Increase
for Voluntariness and Controllability:
E = 0.22................................................ 110-251 98-182 72-101 35-38
E = 1.0................................................. 120-273 107-198 78-110 38-42
----------------------------------------------------------------------------------------------------------------
20-Year Latency Period Assumption
----------------------------------------------------------------------------------------------------------------
Primary Analysis (No VSL Adjustment)........................ 199-452 178-328 130-182 62-69
Adjusted for Income Growth: \2\
E = 0.22................................................ 203-461 181-334 133-186 63-70
E = 1.0................................................. 221-502 197-364 144-202 69-77
Adjusted for Income Growth \2\ and Discounting:
E = 0.22................................................ 53-119 47-86 34-48 16-18
E = 1.0................................................. 57-130 51-94 37-52 18-20
Adjusted for Income Growth,\2\ Discounting, and 7% Increase
for Voluntariness and Controllability:
E = 0.22................................................ 56-127 50-92 37-51 18-20
E = 1.0................................................. 61-139 54-100 40-56 19-21
----------------------------------------------------------------------------------------------------------------
\1\ The lower- and upper-bound benefits estimates correspond to the lower- and upper-bound risk estimates and
cancer cases avoided as shown in section III.D.2 of this preamble.
\2\ This adjustment reflects the change in WTP based on real income growth from 1990 to 1999. E = income
elasticity.
[[Page 7017]]
As shown in Tables III.E-5 and III.E-6, the highest range of
adjusted benefits estimates at the 10 g/L MCL ($144-$202
million) are yielded when benefits are adjusted for changes in WTP
based on real income growth only with an income elasticity of 1.0. The
lowest adjusted benefits estimates at the 10 g/L MCL ($73-$103
million at 3%, $34-$48 million at 7%) are yielded under the assumption
of a 20-year latency period that includes adjustments for discounting
and WTP changes based on real income growth (income elasticity = 0.22).
These results indicate the high degree of sensitivity of benefits
estimates to different assumptions of a latency period, discount rate,
and income elasticity and also the inclusion of adjustments for income
growth and voluntariness and controllability.
3. Comparison of Costs and Benefits
This section presents a comparison of quantifiable total national
costs and benefits for each of the arsenic regulatory options
considered. Three separate analyses are considered, including a direct
comparison of aggregate national costs and benefits, a summary of
benefit-cost ratios and net benefits, and the results of a cost-
effectiveness analysis of each regulatory option.
a. Total national costs and benefits. Table III.E-7 shows the
annual costs and benefits associated with the 10 g/L MCL and
also with three other arsenic levels considered in the proposed rule.
Both costs and benefits increase as arsenic levels decrease. Costs
increase over decreasing arsenic levels because of the increasing
number of systems that must treat to lower arsenic levels. Benefits
estimates increase as arsenic levels decrease due to the greater number
of both fatal and non-fatal cancer cases avoided at lower arsenic
levels. Additionally, other potential non-quantifiable health benefits
are summarized in Table III.E-7.
Table III.E-7 Estimated Annual Costs and Benefits From Reducing Arsenic in Drinking Water
[1999, $ millions]
----------------------------------------------------------------------------------------------------------------
Total national Total bladder Total lung Total combined
Arsenic level costs to CWSs cancer health cancer health cancer health Potential nonquantifiable
(g/L) and NTNCSs \1\ benefits \2\ benefits \2\ benefits \2\ health benefits
----------------------------------------------------------------------------------------------------------------
3.................. 697.8-792.1 58.2-156.4 155.6-334.5 213.8-490.9 Skin Cancer; Kidney Cancer;
Cancer of the Nasal
Passages; Liver Cancer;
Prostate Cancer;
Cardiovascular Effects;
Pulmonary Effects;
Immunological Effects;
Neurological Effects;
Endocrine Effects.
5.................. 414.8-471.7 52.0-113.3 139.1-242.3 191.1-355.6 ...........................
10................. 180.4-205.6 38.0-63.0 101.6-134.7 139.6-197.7 ...........................
20................. 66.8-76.5 20.1-21.5 46.1-53.8 66.2-75.3 ...........................
----------------------------------------------------------------------------------------------------------------
\1\ Costs include treatment, monitoring, O&M, and administrative costs to CWSs and NTNCWSs and State costs for
administration of water programs. The lower number shows costs annualized at a consumption rate of interest of
3%, EPA's preferred approach. The higher number shows costs annualized at 7%, which represents the standard
discount rate preferred by OMB for benefit-cost analyses of government programs and regulations.
\2\ The lower- and upper-bound bladder, lung, and combined cancer benefits estimates correspond to the lower-
and upper-bound risk estimates and cancer cases avoided as shown in section III.D.2 of this preamble; these
estimates include both mortality and morbidity.
b. National net benefits and benefit-cost ratios. Table III.E-8
describes the quantifiable net benefits and the benefit-cost ratios
under various regulatory levels for both CWSs and NTNCWSs at 3% and 7%
discount rates. The net benefits and benefit-cost ratios do not include
any of the potential nonquantifiable health benefits that are listed in
the previous table. As shown in Table III.E-8, under both the lower-and
upper-bound estimates of avoided lung and bladder cancer cases, the net
benefits decrease as the arsenic rule MCL options become increasingly
more stringent. Similarly, the benefit-cost ratios decrease with each
more stringent MCL option. Costs outweigh the quantified benefits for
the lower-bound benefits estimates under all four MCL options. Benefit-
cost ratios are equal to or greater than 1.0 for the upper-bound
benefits estimates (at both 3% and 7% discount rates) for arsenic
levels of 10 g/L and 20 g/L.
Table III.E--8. Summary of National Annual Net Benefits and Benefit-Cost Ratios, Combined Bladder and Lung
Cancer Cases
[1999, $ millions]\1\ \2\ \3\
----------------------------------------------------------------------------------------------------------------
Arsenic level (g/L)
-------------------------------------------------------
3 5 10 20
----------------------------------------------------------------------------------------------------------------
3% Discount Rate
----------------------------------------------------------------------------------------------------------------
Lower Bound.................... Net Benefits........... (484.0) (223.7) (40.8) (0.6)
B/C Ratio.............. 0.3 0.5 0.8 1.0
Upper Bound.................... Net Benefits........... (206.8) (59.2) 17.3 8.5
B/C Ratio.............. 0.7 0.9 1.1 1.1
----------------------------------------------------------------------------------------------------------------
7% Discount Rate
----------------------------------------------------------------------------------------------------------------
Lower Bound.................... Net Benefits........... (578.3) (280.6) (66.0) (10.3)
B/C Ratio.............. 0.3 0.4 0.7 0.9
Upper Bound.................... Net Benefits........... (301.1) (116.1) (7.9) (1.2)
[[Page 7018]]
B/C Ratio.............. 0.6 0.8 1.0 1.0
----------------------------------------------------------------------------------------------------------------
\1\ Costs include treatment, monitoring, O&M, and administrative costs to CWSs and NTNCWSs and State costs for
administration of water programs. The lower number shows costs annualized at a consumption rate of interest of
3%, EPA's preferred approach. The higher number shows costs annualized at 7%, which represents the standard
discount rate preferred by OMB for benefit-cost analyses of government programs and regulations.
\2\ The lower- and upper-bound bladder, lung, and combined cancer benefits estimates correspond to the lower-
and upper-bound risk estimates and cancer cases avoided as shown in section III.D.2 of this preamble;
unquantified benefits are not included.
\3\ Numbers in parentheses indicate negative numbers.
c. Incremental costs and benefits. Incremental costs and benefits
are those that are incurred or realized in reducing arsenic exposures
from one level to the next more stringent level (e.g., from 20
g/L to 10 g/L). Estimates of incremental costs are
useful in developing estimates of the cost-effectiveness of
successively more stringent requirements.
Table III.E-9 shows the incremental total national risk reduction,
arsenic mitigation costs, and monetized health benefits for the various
arsenic levels valued using discount rates of three and seven percent.
Table III.E-9--Estimates of the Annual Incremental Risk Reduction, Costs, and Benefits of Reducing Arsenic in
Drinking Water
[$ millions, 1999]
----------------------------------------------------------------------------------------------------------------
Arsenic level (g/L)
Benefit-cost element ---------------------------------------------------
20 10 5 3
----------------------------------------------------------------------------------------------------------------
Incremental Risk Reduction:
Fatal Cancers Avoided per Year \1\...................... 10.2-11.3 11.1-18.5 7.8-23.9 3.5-20.4
Incremental Risk Reduction:
Non-Fatal Cancers Avoided per Year \1\.................. 8.5-8.8 7.6-17.1 5.9-20.6 2.6-17.7
Annual Incremental Monetized Benefits \2\................... $66.2-$75.3 $73.4-$122. $51.5-$157. $22.7-$135.
4 9 4
Annual Incremental Costs (3%) \3\........................... $66.8 $113.6 $234.4 $283.0
Annual Incremental Costs (7%) \3\........................... $76.5 $129.1 $266.0 $320.5
----------------------------------------------------------------------------------------------------------------
\1\ Total fatal and non-fatal cancer cases avoided are discussed in section III.D.2 of this preamble.
\2\ The lower- and upper-bound combined cancer benefits estimates correspond to the lower- and upper-bound risk
estimates and cancer cases avoided as shown in section III.D.2 of this preamble.
\3\ Costs include treatment, monitoring, O&M, and administrative costs to CWSs and NTNCWSs and State costs for
administration of water programs.
d. Cost-per-case avoided. Cost-per-case avoided is a commonly used
measure of the economic efficiency with which regulatory options are
meeting the intended regulatory objectives. Table III.E-10 shows the
results of an analysis in which the average national cost of achieving
each unit of reduction in cases of bladder and lung cancer avoided, was
calculated. The average annual cost per case avoided was computed at
each MCL option for both 3% and 7% discount rates.
As shown in Table III.E-10, the cost per bladder and lung cancer
case avoided ranges from $4.8 million down to $3.2 million at the 10
g/L MCL, assuming a 3% discount rate. At a 7% discount rate,
the cost per bladder and lung cancer case avoided ranges from $5.5
million down to $3.7 million at the 10 g/L MCL. As expected,
the cost per bladder and lung cancer case avoided decreases with
increasing arsenic levels. This is due to lower compliance costs at
higher levels for the standard.
Table III.E-10.--Annual Cost Per Cancer Case Avoided for the Final
Arsenic Rule--Combined Bladder and Lung Cancer Cases
$ millions, 1999]
------------------------------------------------------------------------
Lower-bound Upper-bound
Arsenic level (g/L) estimate 1 estimate 1
------------------------------------------------------------------------
3 % Discount Rate
------------------------------------------------------------------------
3....................................... 12.2 5.0
5....................................... 8.1 4.1
10...................................... 4.8 3.2
20...................................... 3.5 3.4
------------------------------------------------------------------------
7 % Discount Rate
------------------------------------------------------------------------
3....................................... 13.8 5.7
5....................................... 9.2 4.7
10...................................... 5.5 3.7
20...................................... 4.0 3.9
------------------------------------------------------------------------
1 The lower- and upper-bound cost per cancer case avoided corresponds to
the range of combined cancer benefits estimates as shown in Table
III.E-3.
4. Affordability
As noted previously, section 1412(b)(4)(E)(ii) of SDWA, as amended,
requires EPA, when promulgating a national primary drinking water
regulation which establishes a maximum contaminant level (MCL), to
[[Page 7019]]
list technology (considering source water quality) that achieves
compliance with the MCL and is affordable for systems in three specific
population size categories: 25-500, 501-3300, and 3301-10,000. If, for
any given size category/source water quality combination, an affordable
compliance technology cannot be identified, section 1412(b)(15)(A)
requires the Agency to list a variance technology. Variance
technologies may not achieve full compliance with the MCL but they must
achieve the maximum contaminant reduction that is affordable
considering the size of the system and the quality of the source water.
In order for the technology to be listed, EPA must determine that this
level of contaminant reduction is protective of public health.
A determination of national level affordability is concerned with
identifying, for each of the given size categories, some central
tendency or typical circumstance relating to their financial abilities.
The metric EPA selected for this purpose is the median household income
(MHI) for communities of the specified sizes. The household is thus the
focus of the national-level affordability analysis. EPA considers
treatment technology costs affordable to the typical household if they
represent a percentage of MHI that appears reasonable when compared to
other household expenditures. This approach is based on the assumption
that the affordability to the median household served by the CWS can
serve as an adequate proxy for the affordability of technologies to the
system itself. The national-level affordability criteria have two major
components: current annual water bills (baseline) and the affordability
threshold (total % of MHI directed to drinking water). Current annual
water bills were derived directly from the 1995 Community Water System
Survey. Based on 1995 conditions, 0.75-0.78% of MHI is being directed
to water bills for systems serving fewer than 10,000 persons.
The fundamental, core question in establishing national-level
affordability criteria is: what is the threshold beyond which drinking
water would no longer be affordable for the typical household in each
system size category? Based upon careful analysis EPA believes this
threshold to be 2.5% of MHI. In establishing this threshold, the Agency
considered baseline household expenditures (as documented in the 1995
Consumer Expenditure Survey, Bureau of Labor Statistics) for piped
water relative to expenditure benchmarks for other household goods,
including those perceived as substitutes for piped water treated to
higher standards, such as bottled water and point-of-use and point-of-
entry devices. Based on these considerations, EPA concluded that
current household water expenditures are low enough, relative to other
expenditures, to support the cost of additional risk reductions. The
detailed rationale for the selection of 2.5% MHI as the affordability
threshold is provided in the guidance document entitled ``Variance
Technology Findings for Contaminants Regulated Before 1996.'' The
difference between the affordability threshold and current water bills
is the available expenditure margin. This represents the dollar amount
by which the water bill of the typical (median) household could
increase before exceeding the affordability threshold of 2.5% of MHI.
By definition, the MHI is the income value exactly in the middle of
the income distribution. The median is a measure of central tendency;
its purpose is to help characterize the nature of a distribution of
values. In the case of income, which tends not to be evenly
distributed, the median is a much better indicator of central tendency
than the mean, or arithmetic average, that could be significantly
skewed by a few large values. The Agency recognizes that there will be
half the households in each size category with incomes above the
median, and half the households with incomes below the median. The
objective of a national-level affordability analysis is to look across
all the households in a given size category of systems and determine
what is affordable to the typical, or ``middle of the road'' household.
The Agency recognizes that baseline costs change over time as water
systems comply with new regulations and otherwise update and improve
their systems. To take account of this upward movement in the baseline,
the Agency plans to adjust the baseline it employs in its calculation
in two ways. First, actual changes in the baseline will be measured
approximately every 5 years by the Community Water System Survey. These
changes will reflect not only the increased costs resulting from EPA
drinking water rules, but also any changes resulting from other factors
that could affect capital or operating and maintenance costs. Second,
to the extent practical and appropriate during the period between
Community Water System Surveys, the baseline will be adjusted to
reflect the cost of rules promulgated during that period.
MHI also changes from year to year, generally increasing in
constant dollar terms. For example, since 1995 MHI has increased (in
1999$) by 9.6%. Thus, to determine the available expenditure margin
(the difference between the affordability threshold and the baseline)
for each successive rule, adjustments would need to be made in both the
baseline and the MHI.
Given the narrow and specific purpose for which the national-level
affordability criteria are used, the Agency is not adjusting either the
baseline or the MHI for its analysis for the final arsenic rule. As
noted previously, MHI has increased by 9.6%. The rules, which have been
promulgated since the baseline was developed, are the Interim Enhanced
Surface Water Treatment Rule, the Stage 1 Disinfectants and
Disinfection ByProducts Rule, the revised Radionuclides Rule, the
Consumer Confidence Report Rule and the revised Public Notification
Rule. The Interim Enhanced Surface Water Treatment Rule applies only to
systems serving greater than 10,000 persons, so it has essentially no
impact on the baseline costs for smaller systems. The Stage 1
Disinfectants and Disinfection ByProducts Rule does apply to small
systems, and it has an impact on only 12% of the nearly 68,200 ground
water systems serving 10,000 persons; and on 70% of the nearly 5200
surface water systems serving 10,000 persons. The revised Radionuclides
Rule has limited impact since it, for the most part, reaffirmed long-
standing MCLs. The Consumer Confidence Rule and revised Public
Notification Rule result in no capital expenditures and only very
modest administrative costs.
The Agency believes that, for purposes of assessing national-level
affordability of the arsenic rule, the unadjusted baseline and
unadjusted MHI are appropriate. Making adjustments to these two factors
would not materially alter the outcome of the analysis.
The distinction between national-level affordability criteria and
affordability assessments for individual systems cannot be over-
emphasized. The national-level affordability criteria serve only to
guide EPA on the listing of an affordable compliance technology versus
a variance technology for a given system size/source water combination
for a given contaminant. In the case of arsenic, EPA has determined
that nationally affordable technologies exist for all system size
categories and has therefore not identified a variance technology for
any system size/source water combination. This means that EPA believes
that the typical household in each system size category can afford the
costs associated with the listed compliance technologies. EPA
[[Page 7020]]
recognizes that individual water systems may serve a preponderance of
households with incomes well below the median or may face unusually
high treatment costs due to some unusual local circumstance.
SDWA provides a number of tools that States can use to address
affordability concerns for these individual water systems. Two of these
tools are financial assistance under the Drinking Water State Revolving
Fund (DWSRF) and extended compliance time-frames under an exemption.
SDWA allows States to provide special assistance to water systems that
the State determines to be disadvantaged, using State-developed
affordability criteria. This special assistance may include forgiveness
of principal, a negative interest rate, an interest rate lower than
that charged to non-disadvantaged systems, and extended repayment
periods of up to 30 years. To date, about half of the States have
implemented disadvantaged community programs as part of their DWSRF.
Almost one quarter of all loans made under the DWSRF have been made to
systems classified as disadvantaged by the States.
In addition to special financial assistance through the DWSRF, as
discussed previously, systems facing affordability concerns may also be
eligible for extended time to achieve compliance under the terms of a
State-issued exemption or may receive assistance under the Rural
Utilities Service (RUS) program of the United States Department of
Agriculture (see section I.L). Together with the approximately $1
billion per year being made available through the DWSRF, this results
in a total of about $1.78 billion per year of Federal financial
assistance available for drinking water.
Decisions that a drinking water system makes about how to allocate
its costs to users and how to design rates can also have a significant
effect on affordability for low-income households. A traditional
declining block rate structure would be regressive and might result in
the households with the least income subsidizing excessive water use by
more affluent households. Numerous alternative rate designs are
possible that are more progressive. Of particular interest in
addressing affordability concerns is lifeline rates. Lifeline rates are
a rate structure applicable to qualified residential customers that
includes a specified block of water use priced below the standard
charge for the customer class. Such rates are primarily designed to aid
the poor in obtaining some minimum level of service at an affordable
price.
The basic organizational or institutional structure of the drinking
water system is another very important factor that influences the
affordability of water service. The key issue here is the extent to
which a given organizational or institutional structure is capable of
achieving economic and operational efficiency. An especially important
element of this efficiency relates to the degree to which a system
seeks to work together with other systems. Systems that effectively
work together, perhaps by combining management, will realize lower
overall costs compared to the same systems working independently.
F. What MCL Is EPA Promulgating and What Is the Rationale for This
Level?
1. Final MCL and Overview of Principal Considerations
EPA is today promulgating a final arsenic MCL of 10 g/L.
EPA's selection of this MCL is based on the SDWA statutory requirements
for establishing an MCL and reflects the Agency's detailed evaluation
and careful consideration of thousands of pages of comments. As part of
this process, we have evaluated new data and analysis on occurrence,
unit treatment costs, small system impacts, treatment technology
availability, waste disposal options, and uncertainties regarding
exposure and health effects data. Based on this new information, the
Agency has revisited technical analyses, calculations, and judgments
underlying the proposed MCL of 5 g/L. As discussed in section
III.E. in this preamble, the Agency has conducted a thorough
revaluation of costs and has carefully considered substantial new
analysis on this subject submitted by commenters. In addition, EPA has
completed a detailed reassessment of the risks of arsenic in drinking
water, and has made significant adjustments to provide a more
quantitative evaluation of major sources of uncertainty discussed at
proposal and emphasized by commenters from a number of different
perspectives.
Today's rule, with a final MCL of 10 g/L, reflects the
application of several provisions under SDWA, the first of which
generally requires that EPA set the MCL for each contaminant as close
as feasible to the MCLG, based on available technology and taking costs
to large systems into account. The 1996 SDWA amendments also require
that the Administrator determine whether or not the quantifiable and
nonquantifiable benefits of an MCL justify the quantifiable and
nonquantifiable costs. This determination is to be based on the Health
Risk Reduction and Cost Analysis (HRRCA) required under section
1412(b)(3)(C). The HRRCA must include consideration of seven analyses:
(1) The quantifiable and nonquantifiable benefits from treatment
to the new MCL;
(2) The quantifiable and non quantifiable benefits resulting from
reductions of co-occurring contaminants;
(3) The quantifiable and nonquantifiable costs resulting directly
from the MCL;
(4) The incremental costs and benefits at the new MCL and
alternatives considered;
(5) The health risks posed by the contaminant, including risks to
vulnerable populations;
(6) Any increased risk resulting from compliance, including risks
associated with co-occurring contaminants; and
(7) Any other relevant factor, including the uncertainties in the
analyses and the degree and nature of risk.
Finally, the 1996 SDWA amendments provide new discretionary
authority for the Administrator to set an MCL less stringent than the
feasible level if the benefits of an MCL set at the feasible level
would not justify the costs (section 1412(b)(6)) based on the HRRCA
analysis. Today's rule establishing an MCL of 10 g/L for
arsenic is the second time EPA has invoked this new authority. (The
first such time was in the final rule for uranium, which was published
on December 7, 2000; EPA, 2000p.)
In addition to the feasible MCL of 3 g/L, the Agency
evaluated MCL options of 5 g/L, 10 g/L, and 20
g/L and the various comments offered concerning these levels
in response to the proposed rule. EPA has determined that a final MCL
of 10 g/L more appropriately meets the relevant statutory
criteria referred to above, particularly after considering the
following: Available information relating to the various health effects
associated with arsenic; new analysis regarding the projected risk to
the population of adverse health effects that would remain after
implementation; the revised costs and benefits of the various options;
the incremental costs and benefits; and the uncertainties in the
benefit-cost and risk analyses. A summary of the results of the
Agency's reanalysis of these various factors follows.
2. Consideration of Health Risks
The fifth and seventh HRRCA analyses focus on the health risks to
be addressed by a new MCL. Estimates of risk levels to the population
remaining
[[Page 7021]]
after the regulation is in place provide a perspective on the level of
public health protection and associated benefits. SDWA clearly places a
particular focus on public health protection afforded by MCLs. For
instance, where EPA decides to use its discretionary authority after a
determination that the benefits of an MCL would not justify the costs,
section 1412(b)(6) requires EPA to set the MCL at a level that
``maximizes health risk reduction benefits at a cost that is justified
by the benefits.'' (EPA does not believe the sixth HRRCA analysis,
consideration of increased risk likely to result from compliance is a
significant factor in connection with selection of a final MCL; rather,
we believe that many of the appropriate technologies for reducing
arsenic will reduce many other co-occurring inorganic contaminants as
well thereby decreasing, rather than increasing risk.)
The Agency based its evaluation of the risk posed by arsenic at the
MCL options of 3 g/L, 5 g/L, 10 g/L and 20
g/L on a number of considerations, including the bladder
cancer risk analysis developed by the National Research Council (NRC)
of the National Academy of Sciences (NRC, 1999); the NRC's qualitative
assessment of other possible adverse health effects; the lung cancer
risk analysis developed by Morales et al. (2000); and findings of other
relevant national and international studies. This information included,
but was not limited to, findings from epidemiological studies in South
America cited in the NRC report (NRC, 1999) and a study of a population
exposed to high levels of arsenic in Millard County, Utah conducted by
Lewis, et al. (1999).
Among the factors EPA considered in choosing the final MCL was
Congress' intent that EPA ``reduce * * * [scientific] uncertainty'' in
promulgating the arsenic regulation reflected in section 1412(b)(12)
arsenic research plan provisions and the legislative history on the
arsenic provision (S. Rep. 104-169, 104th Cong., 1st Sess. at 39-40).
The uncertainties in the analyses of costs, benefits and risks are also
a factor required to be considered in the HRRCA. All assessments of
risk are characterized by an amount of uncertainty. Some of this
uncertainty can be reduced by collecting more data or data of a
different sort. For other types of uncertainty, improved data or
assessment methods can allow one to define the degree to which an
estimate is likely to be above or below the ``true'' risk. For the
arsenic risk assessment, there are several definable sources of
uncertainty that were taken into account. These include, but are not
limited to, the following:
Uncertainty about the exact exposure of individuals in the
study population to arsenic in drinking water, water used in cooking,
and food;
Uncertainties associated with applying data from a
population in rural Taiwan to the heterogenous population of the U.S.
(including differences in health status and diet between the Taiwanese
and the U.S. population); and
Uncertainties concerning precisely how a chemical causes
cancer in humans (the mode of action) that affects assessments of the
extent and severity of health effects at low doses.
Section III.D. of the preamble to today's final rule provides a
detailed explanation of how these uncertainties associated with the
risk analysis were taken into account in developing a revised estimate
of the risk of arsenic in drinking water. Based on comments and
available information, the Agency has focused, in particular, on the
first uncertainty bullet, and made two adjustments to its risk analysis
to reduce uncertainty and more accurately apply data from the Taiwan
study to the U.S. population. EPA has revised its quantified estimate
of the risks of arsenic in drinking water to adjust for exposure to
arsenic in both cooking water and food in the Taiwanese study and has
also developed a risk range for the combined effects of bladder and
lung cancer to reflect the scope of uncertainty underlying these
estimates. Thus, one of the previously listed uncertainties has
specifically been taken into account quantitatively, while others
continue to be considered in a qualitative sense.
In EPA's judgment, use of a risk range more clearly supports a
qualitative consideration and recognition of the uncertainties that are
inherent in any risk analysis that substantially relies upon
epidemiological information. EPA believes that the health risk analysis
presented in section III.D. of today's rule comprises a plausible range
of likely risk associated with various concentrations of arsenic in
drinking water. As just suggested, we do not believe it is appropriate
to select a central or ``best estimate'' of the risk, due to the
uncertainties associated with the underlying health effects studies and
the various plausible assumptions used in considering these
uncertainties for our risk analysis. This revised analysis of risks was
used in recalculating the benefits attributable to reducing arsenic in
drinking water from its present levels. EPA also recognizes that the
latter two bulleted sources of uncertainty may operate to reduce the
risk estimates if it were possible to account for them quantitatively.
3. Comparison of Benefits and Costs
Under HRRCA analyses one and two, the Agency must consider both
quantifiable and nonquantifiable health risk reduction benefits.
Benefits considered in our analysis include those about which
quantitative information is known and can be monetized as well as those
which are more qualitative in nature (such as some of the non-cancer
health effects potentially associated with arsenic) and which cannot
currently be monetized. Important assumptions inherent in EPA's revised
analysis of the benefits estimates include the value of a statistical
life and willingness to pay to avoid illness. These assumptions and
various adjustment factors considered for our benefits analysis are
explained in detail in section III.E. of this preamble.
EPA considered the relationship of the monetized benefits to the
monetized costs for each the regulatory levels it considered. While
strict equality of monetized benefits and costs is not a requirement
under section 1412(b)(6)(A), this relationship is an important
consideration in the regulatory development process. The monetized
costs and monetized benefits of this final rule, and the methodologies
used to calculate them, are discussed in detail in section III. E. of
this preamble and in the arsenic Economic Analysis.
EPA believes, however, that reliance on only an arithmetic analysis
of whether monetized benefits outweigh monetized costs is inconsistent
with the statute's instruction to consider both quantifiable and
nonquantifiable costs and benefits. The Agency therefore examined and
considered qualitative and non-monetized benefits in establishing the
final MCL, as well as other factors discussed previously. These
benefits are associated with avoiding certain adverse health impacts
known to be caused by arsenic at higher concentrations, which may also
be associated with low level concentrations, and include skin and
prostate cancer as well as cardiovascular, pulmonary, neurological and
other non-cancer effects. (These health effects are discussed in
Section III.D. of this preamble.)
Other potential benefits not monetized for today's final rule
include customer peace of mind from knowing drinking water has been
treated for arsenic and reduced treatment costs for contaminants that
may be co-treated with arsenic. (For example, increased use of
coagulation and micro filtration
[[Page 7022]]
by surface water systems will offer benefits with respect to removal of
microbial contaminants and disinfection byproducts.)
HRRCA analyses three and four require EPA to consider the costs of
compliance with the rule and the incremental costs and benefits. EPA
has also revised the cost of compliance estimates associated with the
various possible regulatory levels considered for today's final
rulemaking. The central estimate of costs has risen modestly since the
proposed rule based on our further analysis of the information and data
provided by commenters. However, in response to comments, we have also
performed a sensitivity analysis that addresses a number of variables
in our analysis and which indicates that the costs of compliance could
exceed our central estimate by as much as 22%.
In comparing monetized costs and benefits, we conducted several
types of analyses, including:
Comparison of total national costs and benefits (Table
III.E-7);
Analysis of incremental costs and benefits (comparing one
regulatory option to another) (Table III.E-9);
Estimates of net benefits (Table III.E-8); and
Examination of benefit-cost ratios (Table III.E-8).
Detailed descriptions of our analyses appear in section III.E. of this
preamble and in the Economic Analysis supporting today's rule. Our
consideration of these analyses in support of the rationale for the
final MCL is discussed below.
4. Rationale for the Final MCL
The rationale for the final MCL promulgated with today's rule is
based on the HRRCA analyses outlined previously and the statutory
criteria for setting an alternative (higher than feasible) MCL under
section 1412(b)(6). These analyses include:
A revised risk analysis of arsenic in drinking water;
A revised analysis of total costs;
A revised analysis of total benefits;
A comparison of costs and benefits using various metrics
at various MCL options (including incremental costs and benefits); and
Other pertinent factors (including uncertainties and the
degree and nature of risk).
In the proposed rule, EPA indicated a preference for a standard at
5 g/L, but solicited comment on MCL options of 3 g/L,
10 g/L, and 20 g/L, depending upon how uncertainties
were addressed in the risk analysis as well in the calculation of costs
and benefits. However, EPA also noted that, between the time of
proposal and promulgation of the final rule, it would work to resolve
as much of this uncertainty as possible. As described earlier, the
principal revised analyses conducted since the rule was proposed and
considered in our selection of the final MCL include: A revised
analysis of the uncertainties of the health effects that has generated
a revised risk range for the various MCL options considered; a revised
range of benefits associated with our current estimates of the risks;
and a revised analysis of costs, including uncertainty and sensitivity
analyses. These revised analyses allow an updated comparison of the
costs and benefits for the various regulatory options considered.
a. General considerations. As explained in section III.E. of
today's preamble, both our benefits and cost estimates involve ranges,
rather than point estimates, due to a variety of factors. Thus, our
consideration of costs and benefits involved an examination and
comparison of these ranges. As can be seen from Table III.E-7, both
total costs and benefits increase as one examines progressively lower
(i.e., more stringent) regulatory options compared to higher options.
However, the benefits and costs do not increase proportionately across
the range of regulatory options as shown by a comparison of net
benefits (defined as costs minus benefits). Progressively more
stringent regulatory options become considerably more expensive, from a
cost standpoint, than the corresponding increases in benefits, as
reflected in decreasing net benefits. (see Table III.E-8.)
b. Relationship of MCL to the feasible level (3 g/L). The
MCL must be set as close as feasible to the MCLG, unless EPA invokes
its discretionary authority under section 1412(b)(6) of SDWA to set an
alternative MCL, which must then be set at a level that maximizes
health risk reduction benefits at a cost that is justified by the
benefits. As explained earlier in this preamble, the MCLG is zero and
the feasible level is 3 g/L. The Agency believes that there
are several important considerations in examining the feasible level.
In comparing the benefits and the costs at this level (see Table III.E-
7), we note that it has the highest projected total national costs
(relative to the other MCL options considered). In addition, while the
benefits are highest at this level relative to the other MCL options,
both the net benefits and the benefit/cost disparity at the feasible
level are the least favorable of the regulatory options considered. For
these reasons, we believe benefits of the feasible level do not justify
the costs. Almost all commenters agreed with this conclusion in the
proposal.
c. Reanalysis of proposed MCL and comparison to final MCL. Based on
substantial public comment, EPA has reexamined the proposed MCL of 5
g/L. In comparing this level to 10 g/L, we note that
both the net benefits and the benefit-cost relationships are less
favorable for 5 g/L as compared to 10 g/L. Total
national costs at 5 g/L are also approximately twice the costs
of an MCL of 10 g/L. At 10 g/L, EPA notes that the
lung and bladder cancer risks to the exposed population after the
rule's implementation are within the Agency's target risk range for
drinking water contaminants of 1 x 10 -\6\ to 1 x 10
-\4\ or below. EPA recognizes that there is uncertainty in
this quantification of cancer risk (as well as other health endpoints)
and this risk estimate includes a number of assumptions, as discussed
previously. EPA did not directly rely on the risk range in selecting
the final MCL, since it is not part of the section 1412(b)(6) criteria;
however, it is an important consideration, because it has a direct
bearing on our estimates of the benefits of the rule.
d. Consideration of higher MCL options. EPA does not believe an MCL
less stringent 10 g/L is warranted from the standpoint of
benefit-cost comparison. While total national costs associated with 20
g/L are the lowest of the regulatory options considered,
benefits are also the lowest of these options. Both regulatory options
of 10 g/L and 20 g/L have relatively favorable
benefit-cost relationships relative to lower regulatory options but are
not significantly different from one another based on this comparison
metric. However, the incremental, upper-bound benefits at 10
g/L are more than twice those of 20 g/L; and 10
g/L is clearly the more protective level. Thus, we do not
believe that an MCL of 20 g/L would ``maximize health risk
reduction benefits'' as required for an MCL established pursuant to
section 1412(b)(6).
e. Conclusion. Strict parity of monetized costs and monetized
benefits is not required to find that the benefits of a particular MCL
option are justified under the statutory provisions of section
1412(b)(6) of SDWA. However, EPA believes that, based on comparisons of
cost and benefits (using the various benefit-cost comparison tools
discussed), the monetized benefits of a regulatory level of 10
g/L best justify the costs. In addition, as discussed in
section III.D. and elsewhere in today's preamble, our further
qualitative consideration of the various sources of
[[Page 7023]]
uncertainty in our understanding of arsenic since the proposal (e.g.,
such as that surrounding the mode of action), has led us to conclude
that our estimate of risk (for the risks we have quantified) is most
likely an upper bound of risks and that the higher MCL of 10
g/L is appropriate. Finally, as discussed in section III.E. of
this preamble EPA believes that there are a number of not yet
quantified adverse health effects and potentially substantial non-
monetized benefits at 10 g/L that increase the overall
benefits at this level.
In summary, based on our reanalysis of costs, benefits, and health
risk reduction, and factoring in the uncertainties in these analyses
and the degree and nature of risk, EPA believes the final MCL of 10
g/L represents the level that best maximizes health risk
reduction benefits at a cost that is justified by the benefits and that
the other regulatory options considered in the proposed rule do not
satisfy the statutory requirements of section 1412(b)(6) of SDWA. We
are therefore exercising our discretionary authority under the statute
to establish an MCL at a level higher than the feasible level and
setting that level at 10 g/L.
IV. Rule Implementation
A. What Are the Requirements for Primacy?
States must revise their programs to adopt any part of today's rule
that is more stringent than the approved State program. Primacy
revisions must be completed in accordance with 40 CFR 142.12, and
142.16. States must submit their revised primacy application to the
Administrator for approval. A State's request for final approval must
be submitted to the Administrator no later than 2 years after
promulgation of a new standard unless the State requests and is granted
an additional 2-year extension.
For revisions of State programs, Sec. 142.12 requires States to
submit, among other things, ``[a]ny additional materials that are
listed in Sec. 142.16 of this part for a specific EPA regulation, as
appropriate.'' Today's rule does not require States to submit
information in Sec. 142.16(e) for primacy revisions associated with the
revised arsenic MCL. The final rule notes that Sec. 142.16(e) primacy
revision information will only be required for new contaminants, not
revisions of existing regulated contaminants.
B. What Are the Special Primacy Requirements?
Today's rule adds special primacy requirements in Sec. 142.16(j)
and Sec. 142.16(k) to the State special primacy requirement section.
Section 142.16(j) clarifies that for an existing regulated contaminant
such as arsenic, States may indicate in the primacy application that
they will use the existing monitoring plans and waiver criteria
approved for primacy under the National Primary Drinking Water
Standards (NPDWRs) for organic and inorganic contaminants (the Phase
II/V rules). Alternatively, the State may inform the Agency in its
application of any changes to the monitoring plans and waiver
procedures.
Section 142.16(k) requires States to establish initial monitoring
requirements for new systems and new sources. Many States already have
developed monitoring programs for new systems and for systems that are
using new sources of water. To meet the requirements of Sec. 142.16(k),
States that have existing requirements may simply explain to EPA in
their primacy revision package their monitoring schedule and how the
State can ensure that all new systems and new sources will comply with
the existing MCLs and monitoring requirements. Some States may wish to
explain that monitoring for new systems is established on a case-by-
case basis. States should explain the factors that are considered as
case-by-case determinations are made.
When a State develops or modifies an initial monitoring program for
new systems and new sources, it should ensure that the program reflects
the contaminant(s) of concern for that State, known contaminant use,
historical data, and vulnerability. Because of varying contaminant uses
and sources, some contaminants occur at higher levels in some regions
of the country than in other regions. Additionally, the concentrations
of some contaminants are known to show clear seasonal peaks, while
others remain constant throughout the year. For example, some States
may be concerned with atrazine and require multiple samples during a
specified vulnerable period (e.g., May 1-July 31), while another State
may only require one sample for the entire year. Alternatively, another
State may be concerned about trichloroethylene and require four
quarterly samples.
C. What Are the State Recordkeeping Requirements?
The standard record keeping requirements for States under SDWA
apply to the arsenic rule (Sec. 142.14). Today's rule does not modify
or require additional recordkeeping requirements. States with primacy
must keep all records of current monitoring requirements and the most
recent monitoring frequency decision pertaining to each contaminant,
including the monitoring results and other data supporting the
decision, and the State's findings based on the supporting data and any
additional bases for such decision. These records must be kept in
perpetuity or until a more recent monitoring frequency decision has
been issued.
D. What are the State Reporting Requirements?
Currently, States with primary enforcement responsibility must
report to EPA information under Sec. 142.15 regarding violations,
variances and exemptions, and enforcement actions and general
operations of State public water supply programs. Today's rule does not
modify or require additional reporting requirements. The State
reporting requirements that will apply to the arsenic standard are the
same as all other regulated inorganic contaminants.
E. When Does a State Have To Apply for Primacy?
To maintain primacy for the Public Water Supply Supervision (PWSS)
program and to be eligible for interim primacy enforcement authority
for future regulations, States must adopt today's final rule. A State
must submit a request for approval of program revisions that adopt the
revised MCL and implement regulations within two years of promulgation,
unless EPA approves an extension per Sec. 142.12(b). Interim primacy
enforcement authority allows States to implement and enforce drinking
water regulations once State regulations are effective and the State
has submitted a complete and final primacy revision application. To
obtain interim primacy, a State must have primacy with respect to each
existing NPDWR. Under interim primacy enforcement authority, States are
effectively considered to have primacy during the period that EPA is
reviewing their primacy revision application.
F. What Are Tribes Required To Do Under This Regulation?
Currently, the Navajo Nation is the only Tribe with primacy for all
the National Primary Drinking Water Regulations, and it will be subject
to the same requirements as a State. There are no other Federally
recognized Indian tribes with primacy to enforce any of the drinking
water regulations. EPA's Regions have responsibility for implementing
the rules for all Tribes except the Navajo Nation under section
1451(a)(1) of SDWA. To obtain primacy authority for the revised arsenic
MCL, Tribes must submit a primacy
[[Page 7024]]
application to regulate inorganic contaminants (i.e., the Phase II/V
rule).
V. Responses to Major Comments Received
A. General Comments
1. Sufficiency of Information and Adequacy of Procedural Requirements
To Support a Final Rule
A number of commenters challenged EPA's basis for promulgating a
final rule, arguing that (1) there was insufficient technical
information provided with the proposed rule, (2) various expert
technical evaluations were not adequately considered, or (3) procedural
requirements (e.g., Unfunded Mandates Reform Act (UMRA), Small Business
Regulatory and Enforcement Flexibility Act (SBREFA)) have not been
fully satisfied. EPA respectfully disagrees, and we believe that the
record of our actions is sufficient to support a final rulemaking.
Other portions of the preamble to today's rule explain the technical
evaluations performed in support of the proposed rule and the revised
analyses conducted, based on comments and information submitted in
response to the proposal. EPA recognizes that various questions about
different aspects of this rulemaking have been the subject of an array
of analyses and reports by various investigators. This area of
investigation has also been dynamic, and there will undoubtedly be
additional analyses after promulgation of the final rule that the
Agency will need to consider in light of the requirement to
periodically review (and revise as appropriate) all final drinking
water regulations as provided by section 1412(b)(9) of SDWA. However,
we believe that we have fully and appropriately considered all
available and relevant information for the final rulemaking and do not
need to repropose as several commenters suggest. We also believe that
we have fully satisfied the procedural requirements of the pertinent
statutory and Executive Order requirements. Section VI. of the preamble
to today's final rule discusses these procedural requirements in more
detail.
2. Suggestions for Development of an Interim Standard
Several commenters advocated an interim standard in view of the
uncertainties associated with the health effects data, the costs of
compliance with the final rule, and concerns over the interpretation of
the ``anti-backsliding'' provision of SDWA related to review and
revision of existing standards (section 1412(b)(9)). While EPA
appreciates these concerns, we do not believe that they provide a
sufficient basis for concluding that an interim standard be set. We
agree with the recommendation of the National Academy of Sciences that
there is sufficient information available now to develop a new lower
drinking water standard for arsenic. We further believe that available
information is sufficient to support a final, rather than an interim,
standard. Finally, there is simply no authority in SDWA to establish an
interim standard that does not comply with sections 1412(b)(4) and
1412(b)(6). However, we are committed to reviewing and revising, if
appropriate, the final standard every six years (or sooner, if
pertinent new information becomes available). In so doing, we must
ensure that the revised standard provides for ``equal or greater
protection to the health of persons'' as compared to the standard it
replaces.
3. Public Involvement and Opportunity for Comment
Some commenters questioned whether the extent of public involvement
in the development of today's rule was sufficient. Some commenters also
suggested that the Agency use a negotiated rulemaking process for the
final rule pursuant to the Federal Advisory Committee Act (FACA). EPA
believes that public involvement throughout the development of this
rule, has been extensive and far-reaching. As discussed in section I.N.
earlier in this preamble, during the period 1996-2000, EPA conducted a
number of Agency workgroup meetings on arsenic and advertised six
stakeholder meetings (held in five locations) in the Federal Register.
Five States also provided written comments on implementation issues
during the workgroup process. Representatives of eight Federal
agencies, 19 State offices, 16 associations representing the breadth of
the public water system community, 13 corporations, 14 consulting
engineering companies, two environmental organizations, three members
of the press, 37 public utilities and cities, four universities, and
one Indian tribe attended the stakeholder meetings on arsenic. EPA
presented an overview of the arsenic rulemaking to over 900 Tribal
representatives in 1998 and provided more detailed information in 1999
to 25 Tribal council members and water utility operators from 12 Indian
tribes. In addition, EPA provided updates on our rulemaking activities
at national and regional meetings of various groups and trade
associations. We also participated in the American Water Works
Association's (AWWA) technical workgroup meetings. As part of the Small
Business Regulatory and Enforcement Flexibility Act (SBREFA) process,
EPA also received valuable input from discussions with small entity
representatives during SBREFA consultations for the arsenic rule. EPA
obtained recommendations from the National Drinking Water Advisory
Council (NDWAC) on the rule as a whole as well as on our approach
benefits analysis and small systems affordability. We also posted
discussion papers produced for our stakeholder interactions on the EPA
Office of Ground Water and Drinking Water (OGWDW) Internet site and
sent them directly to participants at stakeholder meetings and others
who expressed interest. EPA also received over 1,100 comments on the
June 22, 2000 proposed rule. EPA took these comments into consideration
in developing today's final rule.
EPA agrees that the FACA-negotiated rulemaking process has been an
effective one in the past for other complex rulemakings. However, EPA
does not believe that a negotiated rulemaking at this point is
consistent with the deadlines set by Congress for this rulemaking. We
would point out, however, that the Agency has taken a number of active
steps to ensure broad-based stakeholder involvement, as described
previously, and has solicited expert points of view outside the Agency.
Some of these actions included a charge to the National Academy of
Sciences (NAS) to fully explore the most current health effects issues.
A charge was also given to EPA's Science Advisory Board (SAB) to review
key aspects of the proposed rule and EPA's underlying rationale. EPA
believes that this combination of actions ensured that full and
complete stakeholder involvement occurred, and that further
negotiations would be unnecessary.
4. Relation of MCL to the Feasible Level
Several commenters questioned the feasible level of 3 g/L
contained in the proposed rule. Commenters believed that EPA has not
accurately assessed the capabilities of laboratories to achieve the
practical quantitation level (PQL) or of treatment technologies to
reliably and consistently treat down to the feasible level. EPA
disagrees and still believes that 3 g/L is feasible from the
standpoints of both analytical methods and treatment technologies. EPA
discusses these issues in more detail in section III.B. of the preamble
to today's final rule. Many of the comments on the proposed rule were
concerned by the close proximity of the proposed
[[Page 7025]]
standard (5 g/L) to the proposed feasible level (3 g/
L). However, comments regarding whether or not the proposed standard of
5 g/L is feasible are not particularly germane to the setting
of the final standard, which is well above any level identified by most
commenters as being feasible.
5. Relationship of MCL to Other Regulatory Programs
Many commenters expressed concerns about the possible impact of a
new revised drinking water standard for arsenic on other regulatory
standards for arsenic. In particular, several commenters recommended
that EPA consider the prospective costs of future Comprehensive
Environmental Response, Compensation and Liability Act (CERCLA) site
clean-up actions, RCRA hazardous waste management costs, or national
permit discharge elimination system (NPDES) permits to the extent that
a new arsenic in drinking water standard leads to more stringent
regulatory actions under those respective statutes. EPA disagrees and
notes that SDWA specifically excludes from consideration under the
HRRCA such prospective, ancillary costs in developing a drinking water
standard (see section 1412(b)(3)(C) of SDWA).
6. Relation of MCL to WHO Standard
Several commenters on the proposed rule expressed a concern that
the drinking water standard in the U.S. should be no more stringent
than the standard developed for the World Health Organization (WHO).
This comment dealt primarily with the proposed level of 5 g/L
and does not apply to the final MCL of 10 g/L, which is
identical to the WHO standard. However, while the thrust of the comment
is now moot, EPA notes that the basis for the final MCL and the WHO
standard are different. EPA's standard is based on consideration of all
of the risk management factors required to be evaluated under SDWA
(e.g., risk, costs, benefits, treatment technology and analytical
method capabilities, small systems affordability, etc.) while the WHO
standard is based solely on health effects, without regard to any
implementation considerations. Further, the health basis for the WHO
standard is primarily an assessment of arsenic-induced skin cancer,
whereas there are a number of health endpoints of concern in EPA's
analysis including lung and bladder cancer. In summary, the two levels
(the WHO standard and EPA's final MCL) happen to be the same but a
possible future change in the WHO standard would not necessarily
require a revision to EPA's MCL, for the reasons just discussed.
7. Regulation of Non-Transient Non-Community Water Systems (NTNCWSs)
Several commenters objected to the approach outlined in the
proposed rule for addressing NTNCWSs (monitoring and reporting only)
and pointed out the need for consistency in coverage of NTNCWSs in
EPA's rules. These commenters noted that the rules originally
promulgated in 1976 (arsenic and radionuclides) have not required
coverage of NTNCWSs, whereas more recently promulgated rules have. In
addition, EPA's proposed radon rule suggested not covering NTNCWSs and
the recently promulgated radionuclides rule did not require coverage of
NTNCWSs, but instead deferred this issue for future resolution. EPA
agrees that the outcomes of its recent decisions with respect to
coverage of NTNCWSs have been different. However, we considered the
merits of each rulemaking on a case-by-case basis using a consistent
set of criteria, namely the cost/benefit analysis required under
section 1412(b)(4).
For the proposed arsenic rule, EPA carefully examined the risks
posed by NTNCWSs and concluded preliminarily that the risks were such
that, without coverage, consumers of water from NTNCWSs were projected
to be within the target risk range. EPA acknowledges, however, that
there is uncertainty associated with its information about exposure
patterns for consumers of water from NTNCWSs and the demographics of
these facilities. Thus, our understanding of the health risks (and
associated possible benefits of removal) to consumers of water from
NTNCWSs is uncertain. In the case of arsenic, EPA believes the
additional uncertainty in the overall risk analysis argues against any
finding at this point that these systems are substantially different in
terms of exposure than community water systems. EPA also believes the
decision to cover these facilities in today's rule is supported by
consideration of the risks to certain subpopulations within the general
population, such as children who consume water at day care facilities
or schools that are served by NTNCWSs.
Concerns were also expressed about whether commenters were provided
with sufficient information about the costs of full coverage. These
commenters noted that EPA could not, without violating the notice and
comment provisions of the Administrative Procedure Act, move to full
coverage of these facilities in the final rule. EPA disagrees with this
comment. The proposal clearly indicated that full coverage of NTNCWSs
was an option on which comment was being requested and the supporting
documents provided complete information about the costs of full
coverage. (EPA, 2000h, see Table 6-9).
8. Extension of Effective Date for Large Systems
Commenters were generally supportive of EPA's proposed national
determination (pursuant to section 1412(b)(10) of SDWA) that water
systems covered by the rule, serving less than 10,000 persons, and
needing to make capital improvements to comply with the new standard
would need more than 3 years from the time of rule promulgation to
accomplish this. Thus, the proposed rule suggested allowing a two-year
extension for compliance with the new standard, beyond the three years
provided after the promulgation date. However, several commenters
suggested that this finding and the additional two years for compliance
should be applicable to all systems, including those serving more than
10,000 persons, since extensive planning, design, and new equipment
will also generally be needed by larger systems in a similar situation
to comply with the new standard. EPA was persuaded by these comments,
and has, as part of the implementation requirements for today's final
rule, elected to apply this two-year extension to all facilities
covered by today's rule.
B. Health Effects of Arsenic
1. Epidemiology Data
Many commenters were critical of the Taiwan epidemiologic study as
a basis for EPA decision making, quantitative dose-response assessment,
extrapolation of the dose-response from the observed range of exposure,
and application of the same risk estimate to the U.S. population. No
commenters challenged the EPA conclusion that this study and the other
epidemiologic studies together show that arsenic is carcinogenic to
humans. Some supported the risk analysis in the proposed rule and the
notice of data availability (NODA) because it is relatively risk
averse; others had criticisms.
The following issues were raised about the use of the Taiwan risk
assessment to represent U.S. risk: Arsenic exposure from food and via
cooking with contaminated water in Taiwan is higher than is typical for
the U.S. population; exposure groupings were made at the village level
and were assigned the median of the
[[Page 7026]]
concentration of arsenic measured in the wells serving that village;
not all wells serving all villages were measured and well
concentrations varied seasonally; the Taiwan population was a rural
population that was not well nourished, having deficits of selenium,
possibly methionine or choline (methyl donors), zinc and other
essential nutrients; and the Taiwan population may have unknown
differences in genetic polymorphisms from the U.S. population. Similar
concerns were raised about the South American studies.
Commenters also cited studies in the U.S. (Lewis et al., 1999, Utah
population) and Europe (Buchet et al., 1999; Kurttio et al., 1999) as
support for the position that the risks from the Taiwan study
overestimated the risks in the U.S.
Many commenters were convinced that the Lewis et al. (1999) study
of a U.S. population is the best study to use in estimating U.S. risk.
Since the Utah study did not observe cancer outcomes that one would
expect if risks were as large as the Taiwan or South American studies
suggest, these commenters believe that risks estimates from studies of
populations outside of the U.S. overestimate U.S. risks.
Scientists generally agree that high doses of arsenic are
associated with various cancer and noncancer health effects in humans.
Epidemiology studies in humans demonstrate that arsenic induces skin
and internal (e.g., bladder and lung) cancers and non-cancer effects
such as skin keratoses and vascular abnormalities when ingested in
drinking water at high doses.
The epidemiologic investigations that have been most thorough in
investigating the exposure and effects on humans of ingesting ground
water contaminated with arsenic are those of populations in Taiwan
(Chen et al., 1985; 1988; 1992; Wu et al., 1989), Argentina (Hopenhayn-
Rich et al., 1996; 1998), Chile (Smith et al., 1998), and the U.S.
(Lewis et al., 1999). All of these and other, smaller studies have been
considered in the Agency's deliberations on this rule.
The studies from Taiwan, Chile, Argentina and the U.S. employed the
proper endpoints, selected correct study groups and grouped the people
into discrete exposure groups. They also used acceptable methods and
accounted for some known confounders. These studies, due to their
relative sizes, varied in their statistical power to detect
differences. The Utah study (Lewis et al., 1999) contained 4,000 people
while the Taiwan study had approximately 40,000 people and the two
South American studies each had over 200,000 people. All of these
epidemiology studies were ecological and retrospective studies. The
Taiwan and South American studies had no individual exposure data. The
Utah study associated persons with wells that had measured
concentrations though exposure was calculated based on both level of
arsenic and length of exposure. The Utah study followed exposed
individuals to discern causes of later disease through carefully kept
church records.
The Agency chose to make its quantitative estimates of risk based
on the Taiwan study. This choice was endorsed by the EPA Science
Advisory Board (SAB, 2000q; NRC, 1999). The database from Taiwan has
the following advantages: Mortality data were drawn from a cancer
registry; arsenic well water concentrations were measured for each of
the 42 villages; there was a large, relatively stable study population
that had life-time exposures to arsenic; there are limited measured
data for the food intake of arsenic in this population; age-and dose-
dependent responses with respect to arsenic in the drinking water were
demonstrated; the collection of pathology data was unusually thorough;
and the populations were quite homogeneous in terms of lifestyle.
Studies in Argentina and Chile also showed lung and bladder risk of
similar magnitude at comparable levels of exposure. EPA recognizes that
there are problems with the Taiwan study that introduce uncertainties
to the risk analysis such as: the ecological study design; the use of
median exposure data at the village level; the low income and
relatively poor diet of the Taiwanese study population (high levels of
carbohydrates, low levels of protein, selenium and other essential
nutrients); and high exposure to arsenic via food and cooking water.
As urged by many commenters, the Agency has considered and made
adjustments in its dose-response assessment to reflect the quantitative
effect of the high Taiwanese exposure to arsenic via food and cooking
water. The Agency made an adjustment to the lower bound risk estimates
to take into consideration the effect of exposure to arsenic through
water used in preparing food in Taiwan. In addition, an adjustment was
made to the lower bound risk estimates to take into consideration the
relatively high arsenic concentration in the food consumed in Taiwan as
compared to the U.S. We also considered several additional factors
qualitatively in our final decision. These included the effect of the
median well exposure data from the Taiwan study and the effects of
nutritional factors such as selenium and methyl donors. However, we did
not feel that there were sufficient data to account for these factors
quantitatively.
The U.S. population cannot be considered to be made up entirely of
well-nourished, genetically uniform persons. People of the Asian and
Pacific Islander group make up about 4% (approximately 11 million) of
the more than 270 million people in the U.S. (U.S. Census Bureau,
2000). In addition, there is a significant portion of the U.S.
population living in poverty with poor nutrition. Thus, the Agency
continues to believe that the Taiwan study is appropriate as a basis
for risk assessment. The fact that the whole of the Taiwanese
population was nutritionally vulnerable is a factor that the Agency has
considered qualitatively as an uncertainty in risk assessment that may
on average lead to overestimation of risk when applied to the U.S.
The Utah study (Lewis et al., 1999) did not find any excess bladder
or lung cancer risk after exposure to arsenic at concentrations of 14
to 166 g/L. An important feature of the study is that it
estimated excess risk by comparing cancer rates among the study
population in Millard County, Utah to background rates in all of Utah.
But the cancer rates observed among the study population, even those
who consumed the highest levels of arsenic, were significantly lower
than in all of Utah. This is evidence that there are important
differences between the study and comparison populations besides their
consumption of arsenic. One such difference is that Millard County is
mostly rural, while Utah as a whole contains some large urban
populations. Another difference is that the subjects of the Utah study
were all members of the Church of Jesus Christ of Latter Day Saints,
who for religious reasons have relatively low rates of tobacco and
alcohol use. For these reasons, the Agency believes that the comparison
of the study population to all of Utah is not appropriate for
estimating excess risks. An alternative method of analysis is to
compare cancer rates only among people within the study population who
had high and low exposures. The Agency performed such an analysis on
the Utah data, using the statistical technique of Cox proportional
hazard regression (U.S. EPA, 2000x; Cox and Oakes, 1984). The results
showed no detectable increased risk of lung or bladder cancers due to
arsenic, even among subjects exposed to more than 100 g/L on
average. On the other hand, the excess risk could also not be
distinguished statistically from the
[[Page 7027]]
levels predicted by model 1 of Morales et al. (2000). These results
show that the Utah study is not powerful enough to estimate excess
risks with enough precision to be useful for the Agency's quantitative
arsenic risk analysis. Furthermore, the SAB noted that ``(a)lthough the
data provided in published results of the Lewis, et al., 1999 study
imply that there was no excess bladder or lung cancer in this
population, the data are not in a form that allows dose-response to be
assessed dependably'' (EPA, 2000q). Other studies in the U.S. (Morton
et al., 1976; Valentine et al., 1992; Wong et al., 1992) and Europe
(Buchet et al., 1999; Kurttio et al., 1999) were also considered in
EPA's evaluation of the risk from arsenic. However, these studies were
not sufficient to develop a dose-response relationship.
2. Dose-Response Relationship
Numerous comments were received about the quantitative estimation
of potential cancer risks to U.S. populations from drinking water
exposure to arsenic. Concerns were raised about the extrapolation of
the dose (exposure)--response relationship observed in a study of
cancer incidence in an arseniasis-endemic area of Taiwan with high
levels of arsenic in water (Chen et al., 1988; Wu et al., 1989; Chen et
al., 1992) to estimate potential response in the U.S. to arsenic in
water at lower levels.
Some commenters asked whether it is appropriate to assume a linear
dose response for arsenic given that arsenic does not appear to be
directly reactive with DNA. Other commenters urged strict adherence to
the linear approach, and recommended choosing an MCL that is below the
1/10,000 level of estimated risk based on that approach.
Some commenters also noted that independent scientific panels (EPA,
2000q; NRC, 1999; EPA, 1997e; EPA, 1988) who have considered the Taiwan
study have raised the caution that using the Taiwan study to estimate
U.S. risk at lower levels may result in an overly conservative
estimation of U.S. risk. The independent panels have each said that
below the observed range of the high level of contamination in Taiwan
the shape of the dose-response relationship is likely to be sublinear.
Thus, an assumption that the effects seen per dose increment remain the
same from high to low levels of dose may overstate the U.S. risk. Some
commenters have urged that the Agency model the dose-response
relationship as a sublinear one, rather than as a linear one as in the
proposal and NODA for the rule. These commenters consider adherence to
the linear model as a failure of the Agency to use the best available,
peer-reviewed science as required by SDWA.
After consideration of the arguments made by the commenters, the
Agency continues to believe that the best approach, given the
uncertainties associated with the available data, is to use the linear
approach to set the MCLG for arsenic. In the proposal and the NODA, EPA
discussed the fact that the available data on arsenic's carcinogenic
mode of action point to several potential modes of action, but which
one is operative is unknown. For this reason, the data do not support
use of an alternative to linearity. The Agency recognizes that the
dose-response relationship may be sublinear. The Agency has considered
both a linear extrapolation and a nonlinear approach in the selection
of an MCL in this final rule. (see section III.D.1.g. and the comment
response document for a thorough discussion of the Agency's position on
the dose-response assessment for arsenic.)
3. Suggestions That EPA Await Further Health Effects Research
Several commenters expressed the opinion that EPA should delay
setting a standard for arsenic until more research studies have been
completed. These commenters focused on research areas such as health
effects (especially at low doses), the mode of action, and the dose-
response curve. Other commenters questioned EPA's support of new
research and tracking of ongoing research.
Since developing the Arsenic Research Plan as required by the 1996
SDWA amendments, EPA and stakeholders have established a substantial
research program. Significant research has been completed, and further
research is underway. EPA is tracking the progress of ongoing research
and will make research results available to the public. EPA is
committed to issuing the arsenic regulation based on best available
science and believes that the research currently available is
sufficient to do so.
EPA believes that the research underway may provide important new
data for future rulemakings on arsenic. However, EPA does not believe
that a determination on the arsenic MCL must be delayed until this
research is complete. Indeed, the U.S. Court of Appeals for the
District of Columbia Circuit found that EPA:
cannot reject the ``best available'' evidence simply because of the
possibility of contradiction in the future by evidence unavailable
at the time of action--a possibility that will always be present''
and that ``[a]ll scientific conclusions are subject to some doubt;
future hypothetical findings always have the potential to resolve
the doubt. What is significant is Congress's requirement that the
action be taken on the basis of the best available evidence at the
time of rulemaking. The word ``available'' would be senseless if
construed to mean ``expected to be available at some future date''
(Chlorine Chemistry Council v. EPA, 206 F.3d 1286, 1290-91 (D.C.
Cir. 2000)).
In the future, as part of the 6-year review process, the Agency will
evaluate new data to determine if the MCLG and/or MCL promulgated in
today's regulation should be revised.
Research pertaining to arsenic in the drinking water is a priority
for the EPA. In addition, EPA supports and encourages other
organizations to sponsor new epidemiology and toxicology studies on
arsenic. The nature of scientific research is that as each study
attempts to address or resolve a particular issue, it also raises more
questions for investigation. EPA recognizes that even when the ongoing
set of studies are complete, more are likely to follow. Uncertainty is
inherent in science; at no point will ``all'' research be finished and
``all'' questions be answered.
4. Sensitive Subpopulations
Some commenters encouraged EPA to set the arsenic standard as low
as possible to protect vulnerable populations. These commenters felt
that EPA should consider human development and reproduction and
variously defined vulnerable populations as persons with immune,
cardiovascular, and nervous system disorders, children, low-income
people, Native Americans, diabetics, and geriatric populations.
The 1996 SDWA amendments include specific provisions in section
1412(b)(3)(C)(i)(V) that require EPA to assess the effects of a
contaminant on the general population and on groups within the general
population such as infants, children, pregnant women, the elderly,
individuals with a history of serious illness, or other subpopulations
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. The NRC subcommittee (NRC, 1999) noted that there
is a marked variation in susceptibility to arsenic-induced toxic
effects which may be influenced by factors such as genetic
polymorphisms (especially in metabolism), life stage at which exposures
occur, sex, nutritional status, and concurrent exposures to other
agents or environmental factors. EPA shares the view of the NRC report
[[Page 7028]]
which concluded that there is insufficient scientific information to
permit separate cancer risk estimates for potential subpopulations such
as pregnant women, lactating women, and children and that factors that
influence sensitivity to or expression of arsenic-associated cancer and
noncancer effects need to be better characterized. The EPA agrees with
NRC that there is not enough information to make risk conclusions
regarding any specific subpopulations. However, EPA believes it is
appropriate to consider effects on infants due to their greater
consumption of water per body weight and is considering whether to
issue a health advisory that will address this issue.
A study of a population in Chile exposed to about 800 g/L
in its drinking water for a period of years showed significant
association with this exposure and fetal and infant mortality that
declined to background when the water was treated to remove arsenic.
This study was cited by a commenter as indicating more general
sensitivity of fetuses and infants. The dose was one that had a range
of significant arsenic toxicity effects on the adult population. It is
logical that fetuses of mothers so exposed would be affected and
infants would have received several times the adult exposure per kg
body weight and, consequently, more toxicity. This study does not
indicate disproportionate effects on fetuses or infants at low doses.
Once the water was treated the effects declined to background
(Hopenhayn-Rich et al., 2000).
5. EPA's risk analysis
Several commenters felt that EPA did not follow the NRC
recommendations that ``the final calculated risk should be supported by
a range of analyses over a fairly broad feasible range of
assumptions'', misinterpreted the NRC report, or relied solely on the
NRC report and thus did not do an appropriate risk assessment for
arsenic. Others viewed the NRC report as lacking peer review or as
being politically motivated.
The SAB (EPA, 2000q, pgs. 2-3) discussed EPA's use of the NRC
report. In the cover letter to the Administrator they stated:
* * * The NRC also noted a number of factors that likely differ
between the Taiwanese study population and the U.S. population and
which might influence the validity of arsenic cancer risk estimates
in the United States. Even though the Agency did its own risk
characterization (i.e., they combined the NRC risk factors with U.S.
exposure information and arsenic occurrence distributions to obtain
a range of risks for use in their benefits analysis), they chose not
to quantitatively take any of these factors into account at this
time.
The Panel agrees with conclusions reached by the NRC in its 1999
report on arsenic, especially their conclusion that ``there is
sufficient evidence from human epidemiological studies * * * that
chronic ingestion of inorganic arsenic arsenic [sic] causes bladder
and lung, as well as skin cancer.'' The NRC also stated that
currently the Taiwanese data are the best available for quantifying
risk * * *. We note, however, that this Panel does not believe that
resolution of all these factors can nor must be accomplished before
EPA promulgates a final arsenic rule in response to the current
regulatory deadlines. However, resolution of the critical factors *
* *. in time for the next evaluation cycle for the arsenic
regulation should be considered as a goal.
In closing the cover letter to the Administrator, the SAB stated:
Specifically, the majority of the Panel members felt that there
is adequate basis for the Agency to consider use of its
discretionary authority under the Safe Drinking Water Act of 1996 to
consider MCLs other than the proposed 5 g/L.
* * * The ultimate risk number derived from the Taiwanese study
has proven very sensitive to the decision about the appropriateness
of the comparison population. This of course, has important
implications for the use of the data to estimate risk in the U.S.
Also a study in Utah suggests that some U.S. populations may be less
susceptible to the development of cancer, than those in Taiwan * *
*. Also, a recently published study suggests that the incremental
increases in lung and bladder cancers observed in the Taiwan study
are of roughly the same magnitude, rather than the NRC's inference
of a potentially two- to five-fold greater rate of lung cancer
relative to bladder cancer.
As noted by the NRC, the mechanisms associated with arsenic-
induced cancer most likely have a sublinear character, which implies
that linear models, such as those used by the Agency, overestimate
risk * * *. Nonetheless, the Panel agrees with the NRC that
available data do not yet meet EPA's new criteria for departing from
linear extrapolation of cancer risk.
The NRC Subcommittee on Arsenic in Drinking Water explored a number
of model approaches using the Taiwan epidemiology data for bladder
cancer. Although there are indications that the dose-response
relationship for arsenic may be nonlinear at low doses, a convincing
biological argument for selecting a nonlinear model is not yet
available. Thus, according to EPA's draft 1996 guidelines and
consistent with the 1986 guidelines, EPA determined that a point of
departure approach was most appropriate to estimate low-dose risks. EPA
agreed with NRC's choice of the Poisson model. In the NODA, based on
the Morales et al. study (2000), EPA conducted a re-analysis of the
bladder and lung cancer data using a Poisson model with no comparison
population to estimate points of departure for each health endpoint. In
addition to the re-analysis of bladder and lung cancer risk, EPA did a
sensitivity analysis of the effect of exposure to arsenic through water
used in preparing food in Taiwan. In response to comments received on
the proposed rule and the NODA, EPA has also analyzed the effect of
exposure to arsenic through food and considered the effect of village
level exposure data. In summary, EPA's final risk calculation is
supported by analyses of the effect of various assumptions and
uncertainties on the risk estimate and reflects the best available
science.
EPA believes that it has done a thorough risk analysis on arsenic.
Arsenic health risks have remained a high priority at EPA for over 20
years, and EPA scientists have closely followed all scientific
developments. EPA established four independent scientific panels to
evaluate arsenic health risks (EPA, 2000q; NRC, 1999; EPA, 1997e; EPA,
1988) and provided a sense of the views of the broader scientific
community. EPA participated in conducting one of the major cancer
mortality studies available on arsenic (Lewis et al., 1999). In the
proposed rule and NODA, EPA used the 1999 NRC report's analysis of the
Taiwan data as well as other published scientific papers to
characterize the potential health hazards of ingested arsenic. The NRC
report represents a thorough examination of the best available, peer
reviewed science through the late 1990s. Other studies that were
important in EPA's analysis were the Utah study (Lewis et al., 1999)
and the Morales et al. (2000) study. In selecting the proposed MCL, EPA
considered the uncertainties of the quantitative dose-response
assessment, particularly the possible nonlinearity of the dose-
response. EPA also considered the unquantifiable risks from arsenic
such as noncancer effects. In response to commenters, EPA expanded its
analysis of the Utah study (U.S. EPA, 2000x) and delved further into
the uncertainties in the Taiwan data. The Agency made an adjustment to
the lower-bound risk estimates to take into consideration the effect of
exposure to arsenic through water used in preparing food in Taiwan. In
addition, an adjustment was made to the lower-bound risk estimates to
take into consideration the relatively high arsenic concentration in
the food consumed in Taiwan as compared to the U.S. EPA also
investigated the effect of the ecological exposure data on its risk
estimates. When villages with only one arsenic measurement were removed
[[Page 7029]]
from the data set (on the theory that the exposure data was too
uncertain), or when village means instead of medians were used for the
exposure estimates, there was no statistically significant change in
the estimated point of departure, using Model 1 of Morales et al.
(2000). In summary, EPA believes that it has completed a thorough risk
analysis on arsenic that used the best available, peer reviewed
science.
The NRC subjected their draft report to a very rigorous external
peer review using its own procedures that are well established and
generally acknowledged as being independent and objective. The SAB also
reviewed the NRC report and EPA's risk analysis, which was, in part,
based on the NRC report. In addition, the public was provided an
opportunity to comment on the EPA risk analysis as a part of the
arsenic proposal and NODA.
EPA disagrees that the NRC report was politically motivated. The
NRC Subcommittee on Arsenic in Drinking Water was composed of 16 highly
respected scientific experts. EPA believes that this panel produced an
impartial analysis of the data available on the toxicity of arsenic.
6. Setting the MCLG and the MCL
Some commenters were confused about the difference between MCLGs
and MCLs, how EPA sets MCLGs and MCLs based on legal, scientific, and
policy principles, and the relationship between the MCLG and costs and
benefits. Other commenters were concerned about a perceived ``anti-
backsliding'' provision for MCLGs and MCLs in SDWA.
In accordance with SDWA, standards set for contaminants consist of
two components, a maximum contaminant level goal (MCLG) and a national
primary drinking water regulation (NPDWR) (section 1412(b)(1)(A)),
which specifies either ``a maximum contaminant level (MCL) for such
contaminant which is generally set as close to the maximum contaminant
level goal as is feasible'' (section 1412(b)(4)(B)) or a treatment
technique if ``it is not economically or technologically feasible to
ascertain the level of the contaminant'' (section 1412(b)(7)(A)).
SDWA defines an MCLG as ``the level at which no known or
anticipated adverse effects on the health of persons occur and which
allows an adequate margin of safety'' (section 1412(b)(4)(A)). MCLGs
for all carcinogens are set at zero unless adequate scientific data
support a higher MCLG. In accordance with the SDWA, the MCLG is based
on the best available peer reviewed science. An MCLG is a goal, not a
regulatory limit that the Agency expects to be attained by water
systems.
The MCLG must be proposed simultaneously with a national primary
drinking water regulation (section 1412(a)(3)), which specifies a
maximum contaminant level (MCL) as close to the MCLG as technically
feasible. The MCL is the enforceable standard. SDWA allows EPA to make
an exception to setting the MCL as close to the MCLG as is feasible
where the ``Administrator determines * * * that the benefits of a
maximum contaminant level * * * would not justify the costs of
complying with the level.'' In this case, EPA may propose and
promulgate an MCL ``that maximizes health risk reduction benefits at a
cost that is justified by the benefits'' (section 1412(b)(6)). This
exception was used to set the MCL for arsenic. EPA found that at the
feasible level of 3 g/L, the benefits of compliance did not
justify the costs. The Agency determined that an MCL of 10 g/L
maximizes the health risk reduction benefits at a cost that is
justified by the benefits (see preamble discussion of the risk
management decision that was made for arsenic in section III.F.)
Some commenters argued that EPA sets the MCL within a risk-range of
10-\4\ to 10-\6\ without proper regard to the
statutory requirements discussed above. This is not the case. As noted
in the proposal, EPA has historically considered this risk range as
protective of public health, and accordingly has sought to ensure that
drinking water standards are within this risk range. However, the risk-
range represents a policy goal for EPA, and is not a statutory factor
in setting an MCL. In the case of arsenic, EPA did the benefit-cost
analysis required by the statute. Having found that the benefits of an
MCL at the feasible level were not justified by the costs, EPA set the
MCL at 10 g/L. This MCL maximizes health risk reduction
benefits at a cost that is justified by the benefits.
EPA is required to review and revise as appropriate, each national
primary drinking water regulation, at least every 6 years. Revisions to
current regulations ``shall maintain, or provide for greater protection
of the health of persons'' (section 1412(b)(9)). When new scientific
data become available, the Agency may reevaluate the MCLG and MCL.
C. Occurrence
The principal concerns raised by the commenters and our responses
are as follows:
1. Occurrence data
Several commenters expressed concern that EPA estimated occurrence
using data from only 25 States and that the national estimate was thus
not as robust as it should have been. Many of these commenters
suggested that EPA should request data from all States/more systems
before issuing the final rule.
It is true that we based our occurrence estimate on data from only
25 States. However, we believe that we have compiled the most
comprehensive and accurate occurrence estimate possible with currently
available data, and that this estimate adequately supports our various
analyses and final decisions.
For our occurrence analysis, we relied on data submitted
voluntarily by State drinking water agencies. In doing so, we collected
the largest available database on arsenic in drinking water, consisting
of almost 77,000 observations from more than 26,600 public water
systems in 25 States. We received but did not use data from six States
(Florida, Idaho, Iowa, Louisiana, Pennsylvania, and South Dakota),
because the data either could not be linked to PWSs; did not indicate
if results were censored; were all zero; did not provide analytical or
reporting limits; or were rounded to the nearest 10 g/L.
In response to our request in the proposed rule for additional
occurrence data, we received additional data from several States.
However, in each case, the submitted data either corresponded closely
to observations already in our data set (California, New Mexico, Utah),
or were of the wrong kind or insufficient quantity to use in our
estimation (Iowa, Maryland, Nebraska, Oklahoma, Vermont, West
Virginia).
Of the States from whom we did not receive usable data, we believe
that many do not have databases of the kind and quality that we would
need for our occurrence analysis. We therefore could not have obtained
such information from other States without requiring, in some
instances, new monitoring to be undertaken and new data to be compiled.
In forming our occurrence estimate, we did not ignore States for
which we have no suitable data. We accounted for these States by
assigning regional occurrence distributions to them. Our resulting
national estimates compare relatively closely with those developed by
the utility industry and by the U.S. Geological Survey (EPA, 2000r).
Some commenters indicated EPA should not use data from the U.S.
Geological Survey's National Ambient Water Quality Assessment (NAWQA)
or EPA's NIRS, SDWIS, or Rural Water
[[Page 7030]]
Survey (RWS) to estimate occurrence. In forming our occurrence
estimates, we used arsenic concentrations drawn only from our 25-State
arsenic compliance monitoring database. We did not use observations
from NAWQA, SDWIS, RWS, NAOS, NIRS, NOMS, Community Water System Survey
(CWSS) or any other surveys or studies. As the preamble of the proposed
rule (65 FR 38888 at 38903) states, we used National Organic Monitoring
Survey (NOMS), RWS, and the 1978 CWSS in previous arsenic occurrence
analyses, but did not use them for the present analysis because of
their age and relatively high detection limits. The only information we
used from SDWIS was the type and size of particular systems, and the
numbers of systems and population served in different categories of
systems. We used NAWQA, NAOS and NIRS only for comparison to our
finished results.
2. Occurrence Methodology
Some commenters stated their belief that EPA had underestimated
national occurrence because they believe that EPA did not have enough
data with which to develop the estimate. Commenters also believed that,
since the national occurrence is underestimated, noncompliance/co-
occurrence are also underestimated.
We do not agree that we have underestimated arsenic occurrence. We
have the largest existing database of arsenic in drinking water, with
almost 77,000 observations from more than 26,600 public water systems
in 25 States. We did not ignore States for which we have no data, but
accounted for them by assigning regional occurrence distributions to
them. Our data and methodology have been approved by an independent
expert peer review panel. Our occurrence estimates are close to those
of the NAOS and USGS.
Some commenters believe EPA's occurrence methodology is
inconsistent with the way compliance is determined and that EPA should
use a running annual average for estimating noncompliance.
We acknowledged in the proposed rule (65 FR 38888 at 38907) that
our method of estimating occurrence is different from the method used
for determining compliance with the MCL. Our method usually gives
higher estimates, because we substitute non-zero values for non-
detects, while under the regulatory definition of compliance, non-
detects are assumed to equal zero. We believe our method is the best
one despite the difference, for two reasons. First, our goal is to
characterize arsenic occurrence as accurately as possible. Given a
sound characterization of system-mean occurrence and of intra-system
and intra-source variability, the numbers of systems and points of
entry expected to fail the regulatory definition of compliance at some
MCL option can be determined. The reverse calculation, on the other
hand, is generally not possible. Second, as analytical methods improve
and detection limits decrease, the difference between the two methods
will decrease.
To the extent that our estimates disagree with those used for
determining compliance, our estimates will be higher and thus will
cause us to slightly overestimate the costs associated with any MCL
option. Our estimates of benefits, on the other hand, should not be
biased one way or the other by our occurrence estimate, since health
risks are mainly determined by mean exposure over time, which we
accurately characterize. The same would not be true if we used the
regulatory definition of non-detects, which underestimates mean
occurrence.
Commenters also pointed out that occurrence estimates in different
parts of the rule and support documents are inconsistent. Although the
analysis is internally consistent, apparent inconsistencies in the
numbers arise from three sources: System versus site considerations,
year of the SDWIS inventory, and use of best point or regressed
estimates. With respect to the first point, because most large ground
water systems have multiple entry points, some systems which have
average concentrations below the MCL will still have impacted entry
points. As a consequence, the number of impacted systems is much larger
than the number of systems with mean concentrations above the MCL. In
the proposal, this difference amounted to several hundred systems.
In connection with the second point, year of the SDWIS inventory,
it is not unusual for there to be a change from year to year in the
inventory of hundreds of water systems. This results from restructuring
and consolidations, among other factors. In the final rule and
supporting documents, we have tried to address this issue by
consistently using a single set of baseline estimates, taken from EPA's
Drinking Water Baseline Handbook (EPA, 2000b). Regardless, this factor
is only responsible for a one or two percent variation in the impact
estimate, and is not of sufficient significance to impact the decision
making process.
The third issue relates to the representation of the mean system
arsenic occurrence. In many tables, mean arsenic concentrations are
presented which reflect our best point estimates. Nevertheless, the
best estimate of national cost impacts derives from use of a best fit
equation which incorporates all of the data. We have used these
regressed fits in the development of the costs and benefits. The two
sets of estimates are described in section III.C.4.
3. Co-Occurrence
Some commenters believe EPA has underestimated the co-occurrence of
arsenic with radon.We agree that, based on the NWIS data, most systems
with arsenic greater than 10 g/L will also have radon greater
than 300 pCi/L. However, only about 8% of all systems exceed both
standards. Moreover, about 85% of such systems (again based on NWIS)
have radon in the range of 300 to 1000 pCi/L, where incidental removal
of radon will be most effective. We expect, for example, that systems
with 300 to 1000 pCi/L of radon will be more likely to treat for
arsenic by coagulation and microfiltration, which removes most radon
incidentally by aeration. Therefore, we believe that the impact of co-
occurrence of radon and arsenic will be small.
Some commenters believed that EPA did not evaluate the effect of
different sulfate levels in its decision tree. We did evaluate several
ranges of concentrations of sulfate and arsenic against each other (see
65 FR 38888 at 38938). The sulfate concentration ranges included 0 to
25, 25 to 120, 120 to 250, 250 to 500, and >500 mg/L. The arsenic
concentration ranges included 0 to 2, 2 to 5, 5 to 10, 10 to 20, and
>20 g/L. For these ranges, there was no apparent change in co-
occurrence of sulfate and arsenic as the concentrations increased.
However, the Agency took the co-occurrence of arsenic and sulfate and
the impact on anion exchange technology into consideration in the
decision tree at sulfate levels of 20, 20 to 90, 90 to 120, and >120
mg/L. The revised decision tree for today's final rule only applies
anion exchange when sulfate levels are less than 50 mg/L.
Some commenters expressed their belief that NWIS is inadequate to
estimate national co-occurrence of arsenic and radon and that NWIS data
should be verified as representative of PWS water use by requesting
data from States. It is true that NWIS includes samples from non-
drinking water supplies. NWIS is, however, the largest and best data
base available for studying co-occurrence with over 40,000 ambient
water samples. To the extent that non-drinking water samples affect our
estimates, they should cause us to
[[Page 7031]]
overestimate occurrence and therefore also co-occurrence. We realize
that NWIS may not reflect conditions in any given State or water
system; we use it only for deriving national estimates.
D. Analytical Methods
1. Analytical Interferences
Commenters expressed concern about the potential for matrix
interferences in the analysis of arsenic at low levels. A potential for
chloride interference when using ICP-MS with samples containing high
levels of chloride was specifically noted by commenters. A commenter
also stated that some investigators had reported arsenic results in
drinking water samples that differed depending on the valence state of
the arsenic in the sample (i.e., As (III) or (V)) when using methods
that used GHAA technology. The Agency agrees that interferences may be
encountered when determining arsenic using the methods proposed in the
June 2000 rule (including the GHAA technique). However, the Agency
disagrees that the interferences are unexpected or impede compliance
with the arsenic MCL of 0.01 mg/L. Four different measurement
technologies are approved for the analysis of arsenic: AA furnace, AA-
Platform, GHAA and ICP-MS with respective MDLs of 0.001 mg/L, 0.0005
mg/L, 0.001 mg/L, and 0.0014 mg/L. These technologies have been used
for compliance determinations of arsenic for many years. The methods
written around each of these technologies identify potential
interferences and contain corrective procedures. In particular, the
ICP-MS method warns of potential interferences from chloride and
provides instructions to eliminate this problem.
2. Demonstration of PQL (Includes Acceptance Limits)
Several commenters agreed with the 30% acceptance limit
and the 0.003 mg/L PQL derived and proposed for arsenic. Other
commenters expressed concerns that the PQL was not correctly derived or
that the acceptance limits were too broad.
A commenter stated that the Agency should set the PQL at 5 to 10
times the method detection limits of 0.001mg/L which would result in a
PQL range of 0.005 to 0.010 mg/L. As previously explained in section
III.B.1 of this preamble, EPA only uses the MDL multiplier approach to
derive a PQL when there is insufficient interlaboratory data to
statistically derive a PQL. For arsenic, the Agency had ample WS data
to derive a PQL using the interlaboratory approach.
Several commenters were concerned that the ``PQL study is not
realistic and does not account for matrix interference in real drinking
water samples.'' In addition, some commenters stated that the ``PQL
should be set at a level that is achievable by laboratories on a
routine basis.'' EPA disagrees that the PQL for arsenic is unrealistic,
or that it has been set at a level that is unachievable on a routine
basis. As explained in section III.B.1 of this preamble, EPA used the
interlaboratory data from six recent WS studies to derive the arsenic
PQL. The WS studies utilize reagent grade water (i.e., blank water free
of interferences) for the PE-samples that are analyzed in the WS study.
Use of reagent water to prepare a test sample conforms with an accepted
and longstanding practice in which a method developer validates an
analytical method in blank water before looking for possible
inaccuracies from matrix effects when the method is applied to a sample
matrix (e.g., a compliance drinking water sample). Reagent water is
used as an initial benchmark for method development and testing,
because it is interference-free and can be readily produced in any
competent laboratory. A lab subsequently identifies and corrects for
matrix effects by comparing its performance on reagent water to the
results on the matrix (contaminated drinking water) or spiked matrix
(clean drinking water spiked with arsenic) sample.
All of the methods approved for SDWA and Clean Water Act (CWA)
compliance monitoring require that laboratories demonstrate acceptable
performance in reagent grade water before drinking water samples are
tested. A study conducted by Eaton (Eaton, 1994) found that the type of
matrix and the analytical method used had no significant effect on the
derivation of their PQL. This study included drinking waters with high
total dissolved solids and total organic carbon, and arsenic
concentrations that ranged from 0.001 to 0.010 mg/L. Thus, EPA
disagrees with the comment that the PQL would be significantly
different if derived in various drinking waters instead of in reagent
water.
The Agency also believes that the derived PQL of 0.003 mg/L is
realistic and is achievable on a routine basis. The derivation of the
PQL for arsenic is consistent with the longstanding process used to
determine PQLs for other metal contaminants regulated under SDWA. In
deriving the PQL for arsenic, the Agency took into consideration the
issue of laboratory capability, laboratory capacity, and the ability of
laboratories to achieve a quantitation level on a routine basis. The
PQL for arsenic was derived from data collected in WS studies in which
PE-samples were prepared with reagent water spiked with low
concentrations, 0.006 mg/L, of arsenic. These studies were conducted
from 1992 to 1995. The number of EPA Regional and State laboratories
that participated in each study ranged from 26 to 45 laboratories.
Using acceptance limits of 30% a linear regression analysis
of this data yielded a PQL of 0.00258 mg/L. The Agency rounded up to
derive the proposed PQL of 0.003 mg/L (3 g/L) with a
30% acceptance limit. Over 75% of the EPA Regional and
State laboratories were able to report arsenic concentrations within
30% of 3