[Federal Register: February 21, 2008 (Volume 73, Number 35)]
[Notices]
[Page 9627-9654]
From the Federal Register Online via GPO Access [wais.access.gpo.gov]
[DOCID:fr21fe08-134]
[[Page 9627]]
-----------------------------------------------------------------------
Part II
Environmental Protection Agency
-----------------------------------------------------------------------
Drinking Water Contaminant Candidate List 3--Draft; Notice
[[Page 9628]]
-----------------------------------------------------------------------
ENVIRONMENTAL PROTECTION AGENCY
[EPA-HQ-OW-2007-1189 FRL-8529-7]
RIN 2040-AD99
Drinking Water Contaminant Candidate List 3--Draft
AGENCY: Environmental Protection Agency (EPA).
ACTION: Notice.
-----------------------------------------------------------------------
SUMMARY: EPA is publishing for public review and comment a draft list
of contaminants that are currently not subject to any proposed or
promulgated national primary drinking water regulations, that are known
or anticipated to occur in public water systems, and which may require
regulations under the Safe Drinking Water Act (SDWA). This is the third
Contaminant Candidate List (CCL 3) published by the Agency since the
SDWA amendments of 1996.
This draft CCL 3 includes 93 chemicals or chemical groups and 11
microbiological contaminants. The EPA seeks comment on the draft CCL 3,
the approach used to develop the list, and other specific contaminants.
DATES: Comments must be received on or before May 21, 2008.
ADDRESSES: Submit your comments, identified by Docket ID No. EPA-HQ-OW-
2007-1189, by one of the following methods:
http://www.regulations.gov: Follow the on-line
instructions for submitting comments.
Mail: Water Docket, Environmental Protection Agency,
Mailcode: 2822T, 1200 Pennsylvania Ave., NW., Washington, DC 20460.
Hand Delivery: Water Docket, EPA Docket Center (EPA/DC)
EPA West, Room 3334, 1301 Constitution Ave., NW., Washington, DC. Such
deliveries are only accepted during the Docket's normal hours of
operation, and special arrangements should be made for deliveries of
boxed information.
Instructions: Direct your comments to Docket ID No. EPA-HQ-OW-2007-
1189. EPA's policy is that all comments received will be included in
the public docket without change and may be made available online at
http://www.regulations.gov, including any personal information
provided, unless the comment includes information claimed to be
Confidential Business Information (CBI) or other information whose
disclosure is restricted by statute. Do not submit information that you
consider to be CBI or otherwise protected through http://www.regulations.gov or e-mail. The http://www.regulations.gov Web site
is an ``anonymous access'' system, which means EPA will not know your
identity or contact information unless you provide it in the body of
your comment. If you send an e-mail comment directly to EPA without
going through http://www.regulations.gov your e-mail address will be
automatically captured and included as part of the comment that is
placed in the public docket and made available on the Internet. If you
submit an electronic comment, EPA recommends that you include your name
and other contact information in the body of your comment and with any
disk or CD-ROM you submit. If EPA cannot read your comment due to
technical difficulties and cannot contact you for clarification, EPA
may not be able to consider your comment. Electronic files should avoid
the use of special characters, any form of encryption, and be free of
any defects or viruses. For additional instructions on submitting
comments, go to Unit I.B of the SUPPLEMENTARY INFORMATION section of
this document.
Docket: All documents in the docket are listed in the http://www.regulations.gov
index. Although listed in the index, some
information is not publicly available, e.g., CBI or other information
whose disclosure is restricted by statute. Certain other material, such
as copyrighted material, will be publicly available only in hard copy.
Publicly available docket materials are available either electronically
in http://www.regulations.gov or in hard copy at the Water Docket, EPA/
DC, EPA West, Room 3334, 1301 Constitution Ave., NW., Washington, DC.
The Public Reading Room is open from 8:30 a.m. to 4:30 p.m., Monday
through Friday, excluding legal holidays. The telephone number for the
Public Reading Room is (202) 566-1744, and the telephone number for the
EPA Docket Center is (202) 566-2426.
FOR FURTHER INFORMATION CONTACT: For information on chemical
contaminants contact Thomas Carpenter, Office of Ground Water and
Drinking Water, Standards and Risk Management Division, at (202) 564-
4885 or e-mail carpenter.thomas@epa.gov. For information on microbial
contaminants contact Tracy Bone, Office of Ground Water and Drinking
Water, at 202-564-5257 or e-mail bone.tracy@epa.gov. For general
information contact the EPA Safe Drinking Water Hotline at (800) 426-
4791 or e-mail: hotline-sdwa@epa.gov.
Abbreviations and Acronyms
< --less than
< =--less than or equal to
>--greater than
>=--greater than or equal to
[mu]--microgram, one-millionth of a gram
[mu]g/L--micrograms per liter
ATSDR--Agency for Toxic Substances and Disease Registry
AWWA--American Water Works Association
CASRN--Chemical Abstract Services Registry Number
CDC--Centers for Disease Control and Prevention
CCL--Contaminant Candidate List
CCL 1--EPA's First Contaminant Candidate List
CCL 2--EPA's Second Contaminant Candidate List
CCL 3--EPA's Third Contaminant Candidate List
CFR--Code of Federal Regulations
CUS/IUR--Chemical Update System/Inventory Update Rule
DBP--disinfection byproduct
DWEL--drinking water equivalent level
EPA--United States Environmental Protection Agency
ESA--ethanesulfonic acid
FDA--United States Food and Drug Administration
FR--Federal Register
g--gram
HAAs--haloacetic acids
IOCs--inorganic contaminants
IRIS--Integrated Risk Information System
kg--kilogram
L--liter
LD50--lethal dose 50; an estimate of a single dose that is
expected to cause the death of 50 percent of the exposed animals; it is
derived from experimental data.
lbs--pounds
LOAEL--lowest-observed-adverse-effect level
MCL--maximum contaminant level
MCLG--maximum contaminant level goal
MRDD--maximum recommended daily dose
mg/kg--milligrams per kilogram body weight
mg/kg/day--milligrams per kilogram body weight per day
mg/L--milligrams per liter
MMWR--Morbidity and Mortality Weekly Report
NAS--National Academy of Sciences
NCI--National Cancer Institute
NCOD--National Contaminant Occurrence Database
NDWAC--National Drinking Water Advisory Council
NOAEL--no-observed-adverse-effect level
[[Page 9629]]
NRC--National Academy of Sciences' National Research Council
NPDWR--national primary drinking water regulation
NTP--National Toxicology Program
OPP--Office of Pesticide Programs
PFOA--perfluorooctanoic acid
PFOS--perfluorooctane sulfonic acid
PWS--public water system
RfD--reference dose
SAB--Science Advisory Board
SDWA--Safe Drinking Water Act
TCR--Total Coliform Rule
TD50--tumorigenic dose 50; The dose-rate which if
administered chronically for the standard life-span of the species will
have a 50% probability of causing tumors at some point during that
period.
TRI--Toxics Release Inventory
TDS--training data set
UCM--Unregulated Contaminant Monitoring
UCMR 1--First Unregulated Contaminant Monitoring Regulation
UCMR 2--Second Unregulated Contaminant Monitoring Regulation
US--United States of America
USDA--United States Department of Agriculture
USGS--United States Geological Survey
WBDO--waterborne disease outbreak
WHO--World Health Organization
yr--year
SUPPLEMENTARY INFORMATION:
I. General Information
A. Does this Action Impose Any Requirements on My Public Water
System?
B. What Should I Consider as I Prepare My Comments for EPA?
II. Purpose, Background, and Summary of This Action
A. What is the Purpose of This Action?
B. Background on the CCL, Regulatory Determinations, and
Unregulated Contaminant Monitoring
1. Statutory Requirements for CCL and Regulatory Determinations
2. The First Contaminant Candidate List
3. The Regulatory Determinations for CCL 1
4. The Second Contaminant Candidate List
5. The Regulatory Determinations for CCL 2
6. The Unregulated Contaminant Monitoring Rule
7. The Third Contaminant Candidate List
C. Summary of the Approach Used to Identify and Evaluate
Candidates for CCL 3
D. What is on EPA's Draft CCL 3?
III. What Analyses Did EPA Use To Develop the Draft CCL 3?
A. Classification Approach for Chemicals
1. Identifying the Universe
2. Screening from the Universe to a PCCL
3. Using Classification Models to Develop the CCL 3
4. Selection of the Draft CCL 3--Chemicals
B. Classification Approach for Microbial Contaminants
1. Developing the Universe
2. The Universe to PCCL
3. The PCCL to Draft CCL Process
4. Selection of the Draft CCL 3 Microbes from the PCCL
C. Public Input
1. Nominations & Surveillance
2. External Expert Review and Input
3. How are the CCL and UCMR Interrelated for Specific Chemicals
and Groups?
IV. Request for Comment
A. Pharmaceuticals
B. Perfluorooctanoic acid and Perfluorooctane sulfonic acid
C. Helicobacter pylori
V. EPA's Next Steps
VI. References
I. General Information
A. Does This Action Impose Any Requirements on My Public Water System?
The draft Contaminant Candidate List 3 (CCL 3) or the final CCL 3,
when published, will not impose any requirements on anyone. Instead,
this action notifies interested parties of the availability of EPA's
draft CCL 3 and seeks comment on the contaminants listed.
B. What Should I Consider as I Prepare My Comments for EPA?
You may find the following suggestions helpful for preparing your
comments:
Explain your views as clearly as possible.
Describe any assumptions that you used.
Provide any technical information and/or data you used
that support your views.
Provide specific examples to illustrate your concerns.
Offer alternatives.
Make sure to submit your comments by the comment period deadline.
To ensure proper receipt by EPA, identify the appropriate docket
identification number in the subject line on the first page of your
response. It would also be helpful if you provided the name, date, and
Federal Register citation related to your comments.
II. Purpose, Background, and Summary of This Action
This section briefly summarizes the purpose of this action, the
statutory requirements, previous activities related to the Contaminant
Candidate List (CCL), and the approach used to develop the CCL 3.
A. What Is the Purpose of This Action?
The Safe Drinking Water Act (SDWA), as amended in 1996, requires
EPA to publish a list of currently unregulated contaminants that may
pose risks for drinking water (referred to as the Contaminant Candidate
List, or CCL) and to make determinations on whether to regulate at
least five contaminants from the CCL with a national primary drinking
water regulation (NPDWR) (section 1412(b)(1)). The 1996 SDWA requires
the Agency to publish both the CCL and the regulatory determinations
every five years. The purpose of this action is to present EPA's draft
list of contaminants on the CCL 3, a description of the selection
process, and the rationale used to make the list.
This action also includes a request for comment on the Agency's
draft CCL 3, the approach used to develop the list, and other specific
contaminants.
B. Background on the CCL, Regulatory Determinations, and Unregulated
Contaminant Monitoring
1. Statutory Requirements for CCL and Regulatory Determinations
Section 1412(b) (1) of SDWA, as amended in 1996, requires EPA to
publish the Contaminant Candidate List every five years. SDWA specifies
that the list must include contaminants that are not subject to any
proposed or promulgated NPDWRs, are known or anticipated to occur in
public water systems (PWSs), and may require regulation under SDWA.
The 1996 SDWA Amendments also specify three criteria to determine
whether a contaminant may require regulation:
The contaminant may have an adverse effect on the health
of persons;
The contaminant is known to occur or there is a
substantial likelihood that the contaminant will occur in public water
systems with a frequency and at levels of public health concern; and
In the sole judgment of the Administrator, regulation of
such contaminant presents a meaningful opportunity for health risk
reduction for persons served by public water systems.
In developing the draft CCL 3, the Agency considered the best
available data and information for unregulated contaminants. As
required under the Safe Drinking Water Act, EPA evaluated substances
identified in section 101(14) of the Comprehensive Environmental
Response, Compensation, and Liability Act of 1980 and substances
registered as pesticides under the Federal Insecticide, Fungicide, and
Rodenticide Act. In addition to these required data sources, the Agency
also developed the National Contaminant Occurrence Database (NCOD)
established under section 1445(g) of SDWA. Substances from NCOD were
included in the initial set
[[Page 9630]]
of contaminants considered for the draft CCL 3.
SDWA also directs the Agency to consider the health effects and
occurrence information for unregulated contaminants to identify those
contaminants that present the greatest public health concern related to
exposure from drinking water. In selecting contaminants for the draft
CCL 3, adverse health effects that may pose a greater risk to subgroups
which represent a meaningful portion of the population were considered.
Adverse health effects associated with infants, children, pregnant
women, the elderly, and individuals with a history of serious illness
were evaluated for both chemicals and microbes. The specific analyses
and evaluations used by the Agency are discussed and cited in the
relevant sections of this notice.
2. The First Contaminant Candidate List
Following the 1996 SDWA Amendments, EPA sought input from the
National Drinking Water Advisory Council (NDWAC) on the process that
should be used to identify contaminants for inclusion on the first CCL
(CCL 1). For chemical contaminants, the Agency developed screening and
evaluation criteria based on the recommendations provided by NDWAC. For
microbiological contaminants, NDWAC recommended that the Agency seek
external expertise to identify and select potential waterborne
pathogens. As a result, an external group of microbiologists and public
health experts developed the criteria for screening, conducted an
evaluation of microbial agents, and selected the initial list of
microbiological contaminants for the CCL 1.
The draft CCL 1 was published on October 6, 1997 (62 FR 52193
(USEPA, 1997)). After consideration of all comments, EPA published the
final CCL 1, which included 50 chemical and 10 microbiological
contaminants, on March 2, 1998 (63 FR 10273 (USEPA, 1998 b)). A more
detailed discussion of how EPA developed CCL 1 can be found in the 1997
and the 1998 Federal Register notices (62 FR 52193 (USEPA, 1997) and 63
FR 10273 (USEPA, 1998 b)).
3. The Regulatory Determinations for CCL 1
EPA published its preliminary regulatory determinations for a
subset of contaminants listed on CCL 1 on June 3, 2002 (67 FR 38222
(USEPA, 2002 b)). The Agency published its final regulatory
determinations on July 18, 2003 (68 FR 42898 (USEPA, 2003 a)). EPA
identified 9 contaminants from the 60 contaminants listed on CCL 1 that
had sufficient data and information available to make regulatory
determinations. The 9 contaminants were Acanthamoeba, aldrin, dieldrin,
hexachlorobutadiene, manganese, metribuzin, naphthalene, sodium, and
sulfate. The Agency determined that a national primary drinking water
regulation was not necessary for any of these 9 contaminants. The
Agency issued guidance on Acanthamoeba and health advisories for
magnesium, sodium, and sulfate.
4. The Second Contaminant Candidate List
The Agency published its draft second CCL (CCL 2) Federal Register
notice on April 2, 2004 (69 FR 17406 (USEPA, 2004)) and the final CCL 2
Federal Register notice on February 24, 2005 (70 FR 9071 (USEPA, 2005
b)). The CCL 2 carried forward the 51 remaining chemical and microbial
contaminants that were listed on CCL 1.
5. The Regulatory Determinations for CCL 2
EPA published its preliminary regulatory determinations for a
subset of contaminants listed on CCL 2 on May 1, 2007 (72 FR 24015
(USEPA, 2007 d)). EPA identified 11 contaminants from the 51
contaminants listed on CCL 2 that had sufficient data and information
available to make preliminary regulatory determinations. The 11
contaminants are boron, the dacthal mono- and di-acid degradates, 1,1-
dichloro-2,2-bis (p-chlorophenyl) ethylene (DDE), 1,3-dichloropropene,
2,4-dinitrotoluene, 2,6-dinitrotoluene, s-ethyl propylthiocarbamate
(EPTC), fonofos, terbacil, and 1,1,2,2-tetrachloroethane. The Agency
has made a preliminary determination that a national primary drinking
water regulation is not necessary for any of these 11 contaminants. The
Agency is scheduled to publish its final regulatory determinations in
2008. In the May 1, 2007 FR notice, the Agency indicated that
additional information was needed to make the regulatory determinations
for perchlorate and methyl tertiary butyl ether (MTBE) and provided a
summary of the current health effects, occurrence, and exposure
information.
6. The Unregulated Contaminant Monitoring Rule
SDWA provides EPA with the authority to require all large and a
subset of small systems to monitor for unregulated contaminants. EPA
may require monitoring for up to 30 contaminants under the Unregulated
Contaminant Monitoring Rule (UCMR). Since the 1996 SDWA amendments, the
Agency has issued two UCMRs (UCMR 1 and UCMR 2). UCMR 1 was promulgated
on September 17, 1999 (64 FR 50556 (USEPA, 1999)) and UCMR 2 on January
4, 2007 (72 FR 367 (USEPA, 2007 a)), followed by two revisions
published later in January 2007 (72 FR 3916 (USEPA, 2007 b) and 72 FR
4328 (USEPA, 2007 c)). Monitoring under UCMR 2 will take place during
the 2008-2010 time period.
UCMR 2 requires monitoring for several pesticides and pesticide
degradates, five polybrominated diphenyl ether (PBDE) flame retardants,
a group of nitrosamines and two munitions (TNT and RDX). All of the
chemicals on UCMR 2 were included among the contaminants evaluated for
CCL 3. Data collected under the UCMR are an important source of
occurrence information for the CCL process.
7. The Third Contaminant Candidate List
In 1998, the Agency sought advice from the National Academy of
Sciences' National Research Council (NRC) on how to improve the CCL
process. The NRC published its recommendations on the CCL process in
2001 (NRC, 2001). The NRC proposed a broader, more reproducible process
to identify the CCL than the process used by EPA in the first CCL. The
NRC recommended that EPA develop and use a multi-step process for
creating CCL 3 and future CCLs, whereby a broadly defined ``universe''
of potential drinking water contaminants is identified, assessed, and
reduced to a preliminary CCL (PCCL) using simple screening criteria.
All of the contaminants on the PCCL would then be assessed in more
detail using a classification tool to evaluate the likelihood that
specific contaminants could occur in drinking water at levels and at
frequencies that pose a public health concern.
In 2002, the Agency sought input from the National Drinking Water
Advisory Council (NDWAC) on how to implement the NRC's recommendations
to improve the CCL process. NDWAC agreed that EPA should proceed with
the NRC's recommendations and provided some additional considerations,
including the overarching principles the Agency should follow. The
NDWAC workgroup met 10 times between September 2002 and May 2004. The
NDWAC issued its recommendations in ``The National Drinking Water
Advisory Council Report on the CCL Classification Process to the U.S.
Environmental Protection Agency'' (NDWAC, 2004).
[[Page 9631]]
NDWAC recommended two guiding principles for construction of the
CCL universe, which are:
The universe should include those contaminants that have
demonstrated or have potential occurrence in drinking water, and
The universe should include those contaminants that have
demonstrated or have potential adverse health effects.
These inclusionary principles apply to the selection of
contaminants for initial CCL consideration.
The NDWAC also recommended that the universe of contaminants should
be screened based on widely available data elements that indicate
important health effects and occurrence information. This screening
step should be as simple as possible and capable of identifying
contaminants of the greatest significance for further consideration.
Consideration of a classification approach was also recommended to
increase the transparency and reproducibility of the CCL decision
process. NDWAC recommended that EPA pursue classification models that
build on the screening criteria to further characterize the adverse
health effects and occurrence of chemical contaminants. NDWAC noted
that the classification models are tools to help prioritize
contaminants for the CCL. The model results, available information used
by the model, and expert reviews should be used to determine which
contaminants are listed for the next CCL. The process to develop the
models should be viewed as iterative, and EPA should involve experts
and allow opportunities for meaningful public comment on the evaluation
of contaminants.
NDWAC recommended several overarching principles that EPA should
use to develop the CCL. In addition to the need for transparency and
public participation, these overarching recommendations include:
Integrate expert judgment throughout the CCL process.
Expert judgment is inherent throughout the development of the CCL
process and in implementing that process once it is developed. Critical
reviews, involving various types of expert consultation and
collaboration, will be useful at key points in the new, evolving CCL
process.
Conduct an active surveillance and nomination/evaluation
processes to ensure timely identification of information relevant to
new and emerging agents.
Apply an adaptive management approach (i.e., an approach
that can be refined in future iterations as more knowledge is acquired)
to implement the CCL process. The development of any model should be an
adaptive process, and should be reviewed by experts with consideration
given to updating the process with each successive CCL cycle.
NDWAC also recognized that there were significant differences in
the methods and information used to characterize chemical and
microbiological contaminants. Chemical contaminants tend to be
characterized by toxicological and occurrence data that can be modeled
or estimated if measurement is not possible. These discrete
characteristics are often captured in data sources. For microbes, the
adverse health effects from exposure are characterized by clinical or
epidemiological data and there are few methods to estimate or model
their occurrence. Limited sources of tabular data for microbes may
require evaluation of primary literature, technical reports,
monographs, and reference books to identify a universe of microbes for
consideration. NDWAC recommended the Agency use human pathogens as the
starting point for identifying microorganisms considered for inclusion
in the CCL and apply a two-step evaluation of those pathogens.
C. Summary of the Approach Used To Identify and Evaluate Candidates for
CCL 3
The Agency revised the CCL process used in previous efforts based
on the knowledge and experience it has gained from evaluating
unregulated contaminants and the recommendations and advice from NRC
and NDWAC. Based on these recommendations the Agency developed and
implemented a classification approach that identifies priority drinking
water contaminants in a transparent and reproducible manner that is
amenable to an adaptive management approach.
The Agency's approach to classifying contaminants is based on
available data to characterize the occurrence and adverse health risks
a contaminant may pose to consumers of public water systems. EPA
developed and implemented the following multi-step CCL process to
identify contaminants for inclusion on the Draft CCL 3.
Identify a broad universe of potential drinking water
contaminants (called the CCL 3 Universe). EPA evaluated 284 data
sources that may identify potential chemical and microbial contaminants
and selected a set of approximately 7,500 chemical and microbial
contaminants from these data sources for initial consideration.
Apply screening criteria to the CCL 3 Universe to identify
those contaminants that should be further evaluated. Contaminants not
passing the screening criteria remained in the universe. The screening
criteria EPA developed are based on a contaminant's potential to occur
in public water systems and the potential for public health concern.
Applying these criteria narrows the universe of contaminants to a
Preliminary-CCL (or PCCL).
Identify contaminants from the PCCL to include on the CCL
based on a more detailed evaluation of occurrence and health effects.
For chemicals, EPA used structured classification models as tools to
evaluate and identify drinking water priority contaminants. Decisions
to include chemicals were made using the model results and the best
available data to identify contaminants that may occur in PWSs and may
cause adverse health effects. EPA used a decision tree approach for
microbial contaminants to identify those contaminants that have the
potential to occur in PWSs and transmit waterborne disease. These two
approaches resulted in a draft list of chemicals and microbes for
inclusion on the Draft CCL 3.
Incorporate public input and expert review in the CCL
process. EPA sought public input by asking for nominations of
contaminants to consider for the CCL (71 FR 60704 (USEPA, 2006 b)) and
incorporated these nominations in the three key steps already
discussed. EPA also convened several expert panels for both chemicals
and microbes to review, and provide input and comment, on the CCL 3
process and on a review of a preliminary draft CCL 3.
Exhibit 1 illustrates the CCL multi-step approach that resulted
from the Agency's efforts, input, and collaboration with NRC and NDWAC.
This generalized process is applied to both chemical and microbial
contaminants, though the specific execution of particular steps differs
in detail.
[[Page 9632]]
[GRAPHIC] [TIFF OMITTED] TN21FE08.000
EPA provides a more detailed discussion of the analyses and
decisions it made to develop the Draft CCL 3 in the EPA Water Docket.
EPA prepared several support documents that are available for review at
http://www.regulations.gov. These documents include:
Three comprehensive support documents for the chemicals
entitled, ``Contaminant Candidate List 3 Chemicals: Identifying the
Universe'' (USEPA, 2008 a), ``Contaminant Candidate List 3 Chemicals:
Screening to a PCCL'' (USEPA, 2008 b), and ``Contaminant Candidate List
3 Chemicals: Classification of the PCCL to the CCL'' (USEPA, 2008 c).
These documents describe in detail how the classification process was
developed and used to select the chemicals for the Draft CCL.
Three comprehensive support documents for the microbes
entitled, ``Contaminant Candidate List 3 Microbes: Identifying the
Universe'' (USEPA, 2008 d), ``Contaminant Candidate List 3 Microbes:
Screening to the PCCL'' (USEPA, 2008 e), and ``Contaminant Candidate
List 3 Microbes: PCCL to CCL Process'' (USEPA, 2008 f). These documents
describe the microbial listing process in detail.
The Agency also prepared summaries of stakeholder
involvement and reviews conducted on the CCL process and draft list.
These documents are also available in the EPA Water Docket and at
http://www.regulations.gov.
National Drinking Water Advisory Council Report on the CCL
Classification Process to the U.S. Environmental Protection Agency, May
19, 2004.
A nominations and surveillance report, entitled ``Summary
of the Nominations for the Third Contaminant Candidate List'' (USEPA,
2008 g), which describes the nominations process and the contaminants
that were nominated as part of EPA's process.
Two documents summarizing the expert review of the
chemical and microbial processes, entitled ``Chemical Expert Input and
Review for the Third Contaminant Candidate List'' (USEPA, 2008 h) and
``Microbial Expert Input and Review for the Third Contaminant Candidate
List'' (USEPA, 2008 i).
D. What Is on EPA's Draft CCL 3?
Exhibit 2.--Draft Contaminant Candidate List 3: Microbial Contaminants
------------------------------------------------------------------------
Pathogens
-------------------------------------------------------------------------
Caliciviruses
Campylobacter jejuni
Entamoeba histolytica
Escherichia coli (0157)
Helicobacter pylori
Hepatitis A virus
Legionella pneumophila
Naegleria fowleri
Salmonella enterica
Shigella sonnei
Vibrio cholerae
------------------------------------------------------------------------
Chemical Contaminants
------------------------------------------------------------------------
Common name--registry name CASRN
------------------------------------------------------------------------
alpha-Hexachlorocyclohexane............................. 319-84-6
1,1,1,2-Tetrachloroethane............................... 630-20-6
1,1-Dichloroethane...................................... 75-34-3
1,2,3-Trichloropropane.................................. 96-18-4
1,3-Butadiene........................................... 106-99-0
1,3-Dinitrobenzene...................................... 99-65-0
1,4-Dioxane............................................. 123-91-1
1-Butanol............................................... 71-36-3
2-Methoxyethanol........................................ 109-86-4
2-Propen-1-ol........................................... 107-18-6
3-Hydroxycarbofuran..................................... 16655-82-6
4,4'-Methylenedianiline................................. 101-77-9
Acephate................................................ 30560-19-1
Acetaldehyde............................................ 75-07-0
Acetamide............................................... 60-35-5
Acetochlor.............................................. 34256-82-1
Acetochlor ethanesulfonic acid (ESA).................... 187022-11-3
Acetochlor oxanilic acid (OA)........................... 184992-44-4
Acrolein................................................ 107-02-8
Alachlor ethanesulfonic acid (ESA)...................... 142363-53-9
Alachlor oxanilic acid (OA)............................. 171262-17-2
Aniline................................................. 62-53-3
Bensulide............................................... 741-58-2
Benzyl chloride......................................... 100-44-7
Butylated hydroxyanisole................................ 25013-16-5
Captan.................................................. 133-06-2
Chloromethane (Methyl chloride)......................... 74-87-3
Clethodim............................................... 110429-62-4
Cobalt.................................................. 7440-48-4
Cumene hydroperoxide.................................... 80-15-9
Cyanotoxins (3).........................................
Dicrotophos............................................. 141-66-2
Dimethipin.............................................. 55290-64-7
Dimethoate.............................................. 60-51-5
Disulfoton.............................................. 298-04-4
Diuron.................................................. 330-54-1
Ethion.................................................. 563-12-2
Ethoprop................................................ 13194-48-4
Ethylene glycol......................................... 107-21-1
Ethylene oxide.......................................... 75-21-8
Ethylene thiourea....................................... 96-45-7
Fenamiphos.............................................. 22224-92-6
Formaldehyde............................................ 50-00-0
Germanium............................................... 7440-56-4
HCFC-22................................................. 75-45-6
Hexane.................................................. 110-54-3
Hydrazine............................................... 302-01-2
Methamidophos........................................... 10265-92-6
Methanol................................................ 67-56-1
Methyl bromide (Bromomethane)........................... 74-83-9
Methyl tert-butyl ether................................. 1634-04-4
[[Page 9633]]
Metolachlor............................................. 51218-45-2
Metolachlor ethanesulfonic acid (ESA)................... 171118-09-5
Metolachlor oxanilic acid (OA).......................... 152019-73-3
Molinate................................................ 2212-67-1
Molybdenum.............................................. 7439-98-7
Nitrobenzene............................................ 98-95-3
Nitrofen................................................ 1836-75-5
Nitroglycerin........................................... 55-63-0
N-Methyl-2-pyrrolidone.................................. 872-50-4
N-nitrosodiethylamine (NDEA)............................ 55-18-5
N-nitrosodimethylamine (NDMA)........................... 62-75-9
N-nitroso-di-n-propylamine (NDPA)....................... 621-64-7
N-Nitrosodiphenylamine.................................. 86-30-6
N-nitrosopyrrolidine (NPYR)............................. 930-55-2
n-Propylbenzene......................................... 103-65-1
o-Toluidine............................................. 95-53-4
Oxirane, methyl-........................................ 75-56-9
Oxydemeton-methyl....................................... 301-12-2
Oxyfluorfen............................................. 42874-03-3
Perchlorate............................................. 14797-73-0
Permethrin.............................................. 52645-53-1
PFOA (perfluorooctanoic acid)........................... 335-67-1
Profenofos.............................................. 41198-08-7
Quinoline............................................... 91-22-5
RDX (Hexahydro-1,3,5-trinitro-1,3,5-triazine)........... 121-82-4
sec-Butylbenzene........................................ 135-98-8
Strontium............................................... 7440-24-6
Tebuconazole............................................ 107534-96-3
Tebufenozide............................................ 112410-23-8
Tellurium............................................... 13494-80-9
Terbufos................................................ 13071-79-9
Terbufos sulfone........................................ 56070-16-7
Thiodicarb.............................................. 59669-26-0
Thiophanate-methyl...................................... 23564-05-8
Toluene diisocyanate.................................... 26471-62-5
Tribufos................................................ 78-48-8
Triethylamine........................................... 121-44-8
Triphenyltin hydroxide (TPTH)........................... 76-87-9
Urethane................................................ 51-79-6
Vanadium................................................ 7440-62-2
Vinclozolin............................................. 50471-44-8
Ziram................................................... 137-30-4
------------------------------------------------------------------------
III. What Analyses Did EPA Use To Develop the Draft CCL 3?
A. Classification Approach for Chemicals
1. Identifying the Universe
In the first step in the approach, EPA compiled potential data
sources, including sources identified at a stakeholder workshop
sponsored by the American Water Works Association (AWWA), to develop a
broad universe of potential drinking water contaminants, as shown in
Exhibit 1. This compilation identified the 284 data sources that were
assessed for the CCL Universe.
EPA developed a decision tree for data source selection that was
based on four assessment factors, which were applied to all of the
potential data sources:
Relevance. Ensures that the data source provided
information on demonstrated or potential health effects, occurrence, or
potential occurrence using surrogate information (e.g., environmental
release, environmental fate, and transport properties);
Completeness. Ensures that the data source had minimum
record requirements--contact name, description of the data elements,
and how the data were obtained;
Redundancy. Ensures that the data source does not contain
information identical to other more comprehensive data sources; and
Retrievability. Ensures that the data in the source are
formatted for automated retrieval. Each source was accessed on-line (or
as provided by the source) and reviewed.
Basic information about the source, its purpose, and the data
elements it contained, was compiled and documented. Every source was
evaluated using all assessment factors sequentially. Those sources that
met all four factors became the prime sources that formed the
``Universe of Data Sources.'' Sources that passed the first three
factors, but were not retrievable, were designated as supplemental data
sources, to be consulted as necessary (e.g., to fill in data gaps) in
the development of the CCL. Some of the sources that were not easily
retrievable were identified as ``unique'' or ``exceptional'' because of
the importance of their data (i.e., the Hazardous Substance Database).
EPA included chemicals from these sources in the Universe.
After application of the four assessment factors, 39 sources
(Exhibit 3) met all four factors or were considered as exceptional.
These sources were the primary sources used to develop the CCL Chemical
Universe. The details of the how EPA compiled the list of data sources
is discussed in the document entitled, ``CCL 3 Chemicals: Identifying
the Universe'' (USEPA, 2008 a).
Exhibit 3.--Sources That Comprise the Chemical Universe of Data Sources
for the CCL Process
------------------------------------------------------------------------
Name of data source
-------------------------------------------------------------------------
1. ATSDR CERCLA Priority List.
2. ATSDR Minimal Risk Levels (MRLs).
3. Chemical Toxicity Database--Ministry of Health and Welfare, Japan.
4. Chemical Update System/Inventory Update Rule (CUS/IUR)--EPA.
5. Cumulative Estimated Daily Intake/Acceptable Daily Intake (CEDI/ADI)
Database--FDA.
6. Database of Sources of Environmental Releases of Dioxin-Like
Compounds in the United States--EPA.
7. Distributed Structure Searchable Toxicity Public Database Network
(DSSTox)--EPA.
8. Everything Added to Food in the United States (EAFUS) Database--FDA.
9. Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA) List--
EPA.
10. Generally Regarded As Safe (GRAS) Substance List--FDA.
11. Guidelines for Canadian Drinking Water Quality (CADW): Summary of
Guidelines--Health Canada.
12. Hazardous Substances Data Bank (HSDB)--NLM.
13. Health Advisories (HA) Summary Tables--EPA.
14. High Production Volume (HPV) Chemical List--EPA.
15. Indirect Additives Database--FDA.
16. Integrated Risk Information System (IRIS)--EPA.
17. International Agency for Research on Cancer (IARC) Monographs.
18. International Toxicity Estimates for Risk (ITER) Database--TERA.
19. Joint Meeting On Pesticide Residues (JMPR)--2001 Inventory of
Pesticide Evaluations--WHO, FAO.
20. National Drinking Water Contaminant Occurrence Database (NCOD)--
Round 1&2--EPA.
21. National Drinking Water Contaminant Occurrence Database (NCOD)--
Unregulated Contaminant Monitoring Rule (UCMR)--EPA.
22. National Inorganics and Radionuclides Survey (NIRS)--EPA.
23. National Pesticide Use Database--NCFAP.
[[Page 9634]]
24. National Reconnaissance of Emerging Contaminants (NREC)--USGS Toxic
Substances Hydrology Program.
25. National Toxicology Program (NTP) Studies.
26. National Water Quality Assessment (NAWQA)--USGS.
27. OSHA 1988 Permissible Exposure Limits (PELs)--NIOSH.
28. Pesticide Data Program--USDA.
29. Pesticides Pilot Monitoring Program--USGS/EPA.
30. Risk Assessment Information System (RAIS)--Department of Energy--
Chemical Factors.
31. Risk Assessment Information System (RAIS)--Department of Energy--
Health Effects Data.
32. State of California Chemicals Known to the State to Cause Cancer or
Reproductive Toxicity.
33. Substances Registry System (SRS)--EPA.
34. Syracuse Research Corporation (SRC)--BIODEG.
35. The Toxics Release Inventory (TRI)--EPA.
36. Toxic Substances Control Act (TSCA) List--EPA.
37. Toxicity Criteria Database--California Office of Environmental
Health Hazard Assessment (OEHHA).
38. University of Maryland--Partial List of Acute Toxins/Partial List of
Teratogens.
39. WHO Guidelines for Drinking Water Quality: Summary Tables.
------------------------------------------------------------------------
There were approximately 26,000 unique substances identified from
the 39 data sources. Because of the large number of unique substances
identified, EPA developed an initial universe selection process. In the
first phase of the data evaluation process, EPA identified the
chemicals that were present in both health effects and occurrence data
sources. The Agency queried the data sources and found that
approximately 7,300 chemicals, or about one-third of the chemicals,
were present in both health effects and occurrence data sources.
Occurrence was defined broadly to include production data and
environmental occurrence data. EPA placed these chemicals in the
chemical universe to be further evaluated for screening to the PCCL.
EPA then examined the rest of the approximately 18,600 chemicals left
in the initial universe more closely to determine whether they were
found only in health effects data sources or only in occurrence data
sources. EPA found that approximately 5,100 chemicals were in health
effects data sources only. Many of these chemicals were biochemical
compounds (e.g., amino acids, sugars, steroids); mixtures and natural
products (e.g., coal tar, petroleum related substances, rocks, stone,
wool); and other entries that were identified as unique ``substances''
in the data sources but were not chemicals (e.g., turbidity, boot and
shoe manufacture, surgical implants). EPA evaluated these to identify
which ones are chemicals of greatest toxicological concern. Many of the
chemicals fell into the category of greatest toxicological concern due
to their classification as carcinogens. This is described in the report
entitled, ``CCL 3 Chemicals: Screening to a PCCL'' (USEPA, 2008 b).
Through this process, a total of 122 chemicals with only toxicity data
were added to the 7,300 chemicals already in the CCL Chemical Universe.
The chemicals found only in occurrence sources were also
categorized. The approximately 13,500 chemicals with only occurrence
data were a diverse group, comprised of many different types of
chemicals. Data sources that provide the amount of an individual
chemical that is manufactured and produced account for 70 percent (or
9,344) of the total. The remaining 30 percent of chemicals are from
various other data sources (i.e., finished water, ambient water,
environmental release, environmental fate and transport properties, and
food additives). EPA grouped these chemicals by the type of occurrence
data for further evaluation. These included the following groupings:
Chemicals with Finished or Ambient Water Data
Chemicals with Release Data
Chemicals with High Production Volumes
EPA added 42 chemicals with finished or ambient water data to the
Universe despite the lack of health effects information in the data
sources because of their demonstrated occurrence in ambient or potable
water. In addition, disinfection byproducts and water treatment
additives were added to the Chemical Universe. While there may not have
been measured occurrence data for these chemicals in the universe of
data sources, they are considered to have ``default'' occurrence data
because they are formed in, or intentionally added to, drinking water
supplies.
EPA also added 36 chemicals with an environmental release data
source (e.g., those on the Toxics Release Inventory or with pesticide
application data) to the Chemical Universe even though they lacked
health effects data.
The largest group of chemicals found only in occurrence data
sources had only production information. These contaminants include:
organometallics, elements, salts of the inorganic elements, salts of
organic acids, natural product organics (including oils, fatty acids,
sugars, intermediary metabolites), and mixtures (e.g., petroleum
related compounds, hydrocarbons, and others). Over half of the
production chemicals are compounds and/or complexes of elemental
constituents; for example, there were about 750 sodium or potassium
salt compounds alone. In these cases, health effects data are not
available for the exact compound, but are generally available for other
related compounds or the key ion or elemental constituent (e.g.,
sodium). Nearly all elements found in inorganic or organic salts are
represented in the Universe by other compounds with both health effects
and occurrence data. EPA found only 10 elements (excluding carbon,
hydrogen, and oxygen, and the inert gasses krypton, neon, and xenon)
that did not otherwise have representative compounds with health
effects data in the Universe. EPA added these compounds (i.e.,
europium, gadolinium, gold, lanthanum, praseodymium, platinum,
polonium, samarium, terbium, and yttrium) to the Universe. After
evaluation of the characteristics of the chemicals with production data
and the amounts produced on a yearly basis, and because the primary
constituents (i.e., elements) of the chemicals were already in the
Chemical Universe, EPA decided to move only those produced at greater
than 1 billion pounds per year to the CCL Chemical Universe when they
lacked health effects information.
[[Page 9635]]
EPA added a total of 269 chemicals with only occurrence data to the
CCL 3 Chemical Universe. The rest of the substances included in the
original data sources were not included in the Universe.
The initial selection process brought into the CCL Chemical
Universe all substances from the data sources that met the defined
selection criteria, described above. Upon further review, EPA found the
Chemical Universe also contained regulated as well as unregulated
compounds, mixtures, and some substances that were not really
chemicals. To further refine the initial list, EPA removed chemicals
with a national primary drinking water regulation. These contaminants
are already regulated; thus, their inclusion in the CCL process is
unnecessary and does not meet the statutory requirement for selection
of the CCL. EPA removed 1,006 chemicals, which is more than the number
of primary drinking water standards. This is because regulated
contaminants can be found in many forms and because many contaminants
are regulated as part of a class or group(s). For example, EPA removed
approximately 780 radionuclides from the initial list, because they are
regulated as alpha and beta emitters. Also removed were various salts
of regulated elements, and entries for individual trihalomethanes,
haloacetic acids, polychlorinated biphenyls and polyaromatic
hydrocarbons that are regulated as a group. The Agency has determined
that it is inappropriate to include aldicarbs (aldicarb, aldicarb
sulfoxide, and aldicarb sulfone) and nickel on the CCL. These
contaminants are subject to regulation under SDWA section 1412(b)(2)
and thus are not part of the contaminant selection process specified
under SDWA section 1412(b)(1). In response to an administrative
petition from the manufacturer Rhone-Poulenc, the Agency issued an
administrative stay of the effective date of the maximum contaminant
levels (MCLs) for aldicarbs, and they never became effective. NPDWRs
for nickel were promulgated on July 17, 1992 (57 FR 31776 (USEPA,
1992)), but the MCL was later vacated and remanded by the D.C. Court of
Appeals in response to a joint motion by EPA and industry parties
challenging the nickel MCL and MCLG. Because these contaminants are
subject to separate regulatory consideration, EPA has not included them
in the CCL process.
EPA also removed substances that are considered a mixture of
chemicals. EPA defines a mixture in this case as a combination of two
or more chemicals/items that are not defined as a unique substance.
Examples of substances in this category include ``chlorinated
compounds, aliphatic alcohols with more than 14 carbon atoms (c>14),
coal-tar-containing shampoo, petroleum-related substances, resin acids,
and rosin acids.'' Undefined mixtures, such as ``diesel engine
exhaust'' were also included in this group.
EPA also removed ``non-chemically defined'' entries from further
consideration for the initial list. Examples include: ``solar
radiation, wood dust, surgical implants, and welding fumes.'' Some of
these substances are present in the data sources because they have been
evaluated for their potential to cause cancer.
The final step removed biological agents from the initial list.
Contaminants in this category are biological organisms that are being
evaluated as part of the CCL 3 Microbiological Universe. Entries for
biological entities were uploaded from the universe of data sources
from various health effects data sources and pesticide data sources.
Many biological entities were also removed as non-chemically defined.
During this phase of the data evaluation, 1,717 chemicals or
substances were removed from the initial Chemical Universe, leaving
approximately 6,000 chemicals that were designated as the CCL 3
Universe. A list of the CCL Chemical Universe is provided in the
docket. EPA further evaluated these 6,000 chemicals in the next key
step of the process.
2. Screening from the Universe to a PCCL
The next step in the CCL selection approach involved narrowing the
Universe of chemicals to a PCCL, as shown in Exhibit 1. EPA considered
and built upon NDWAC recommendations that the screening process be
based on a contaminant's potential to occur in public water systems and
the potential for public health concern, to select those contaminants
that should move to the PCCL for further evaluation. The screening
approach:
Identifies chemicals that have relatively high toxicity
with high potential to occur in PWSs;
Identifies chemicals that have relatively high toxicity
with minimal actual or potential occurrence in drinking water;
Identifies chemicals that have high potential to occur in
PWSs with relatively moderate toxicity; and
Considers and uses as many of the available types of
health effects and occurrence data identified in the data source
evaluations as practical.
EPA compared the chemicals' health effects relative to their
occurrence and developed analyses that specifically incorporate many
types of available data into the screening criteria. The health effects
information included quantitative, descriptive, or categorical
information. Within each of these broad types of health effects
information, there are multiple types of reported health related values
from multiple sources. The health effects analyses conducted by EPA
identified approaches to compare each of these data types and
identified similarities among chemicals that could be used to define
toxicity categories. The occurrence information also included many
types of available data representative of a chemical's potential to
occur in water. Occurrence data ranged from quantified detection in
PWSs, to environmental release, to production data.
The basic framework EPA used in screening is shown in Exhibit 4.
EPA categorized the CCL Chemical Universe contaminants by their
toxicity along the vertical axis and by their occurrence on the
horizontal axis. This allows for separation of chemicals into those
that move to the PCCL based on their toxicity and occurrence properties
(e.g., upper right in Exhibit 4) and those that are not further
evaluated and remain in the CCL Chemical Universe (e.g., lower left in
Exhibit 4).
EPA used a set of test chemicals to develop the screening criteria.
This set of chemicals included regulated and unregulated chemicals that
provided comprehensive information on health effects and occurrence in
finished and/or ambient water as well as environmental release and
production volume. EPA then used these criteria to select chemicals for
the PCCL for further consideration. The following sections summarize
how EPA developed the screening criteria by evaluating the available
data for chemicals in the Universe, using the framework (Exhibit 4) and
the test chemicals. A more detailed discussion is provided in the
support document entitled, ``CCL 3 Chemicals: Screening to a PCCL''
(USEPA, 2008 b).
[[Page 9636]]
[GRAPHIC] [TIFF OMITTED] TN21FE08.001
a. Health Effects Data Elements
EPA evaluated the toxicity information and health effects data
compiled from the data sources in the Universe and these data varied
greatly. Some of these data are quantitative (e.g., RfD, LOAEL, NOAEL,
LD50) and some are descriptive (e.g., cancer classifications
or predictions). EPA designed the screening process to accommodate both
types of health effects data.
The quantitative toxicity elements and values available in the
Universe included the following:
RfDs and equivalent (RfD-eq): RfDs, Minimum Risk Levels
(MRLs) from ATSDR, Tolerable Daily Intakes (TDIs) from the World Health
Organization (WHO), and Public Health Goals (PHGs) from California EPA.
A reference dose is an estimate (with uncertainty spanning perhaps an
order of magnitude) of a daily oral exposure to the human population
(including sensitive subgroups) that is likely to be without an
appreciable risk of deleterious effects during a lifetime. There are
slight differences among Agencies in the methodologies used for some of
the RfD equivalents.
NOAELs--No Observed Adverse Effect Levels. The NOAEL is
the highest dose evaluated in a study or group of studies that does not
have a biologically significant adverse effect on the species evaluated
as compared to controls.
LOAELS--Lowest Observed Adverse Effect Levels. The LOAEL
is the lowest dose evaluated in a study or group of studies that has a
biologically significant adverse effect on the species evaluated as
compared to the controls.
TD50s--Tumorigenic dose 50. The dose-rate which
if administered chronically for the standard life-span of the species
will have a 50 percent probability of causing tumors at some point
during that period.
MRDD--Maximum Recommended Daily Dose. Recommendations for
the maximum adult daily therapeutic doses for pharmaceuticals.
LD50s--Lethal dose 50; an estimate of a single
dose that is expected to cause the death of 50 percent of the exposed
animals; it is derived from experimental data.
EPA used descriptive cancer data to group data elements into
toxicity categories that provide gradation based upon the strength of
the data. Sources for the descriptive cancer data included:
U.S. EPA Cancer Groupings.
IARC Cancer Groupings.
NTP weight-of-evidence findings from cancer bioassays.
National Cancer Institute (NCI) weight-of-evidence
findings from cancer bioassays.
EPA Water Disinfection By-Products with Carcinogenicity
Estimates (DBP-CAN) groupings based on carcinogenic potential derived
from Quantitative Structure Activity Relationship (QSAR) projections.
EPA divided the chemicals in the Universe into five toxicity
categories for screening based upon the distribution of the toxicity
value for each type of quantitative data element and/or the qualitative
information on cancer weight-of evidence. The five toxicity categories
are designated 1 through 5, with Toxicity Category 1 containing
chemicals in the most toxic grouping and Toxicity Category 5 the least
toxic grouping.
Based upon the distribution of the chemicals for each quantitative
data element, EPA selected ranges of toxicity values for each toxicity
category that differed based upon the type of data element. For
example, the range of toxicity values that place a LOAEL in Toxicity
Category 1 differs from the values used for a LD50. Exhibit
5 displays the ranges for each data element and their respective
Toxicity Categories.
Additional information which describes how EPA performed the
analyses to select the toxicity categories is described in the document
entitled, ``CCL 3 Chemicals: Screening to a PCCL'' (USEPA, 2008 b).
Exhibit 5.--Potency Measures for Universe Data Elements Partitioned Based on Toxicity
[mg/kg/day or mg/kg]
----------------------------------------------------------------------------------------------------------------
RfD NOAEL LOAEL MRDD LD50
----------------------------------------------------------------------------------------------------------------
Toxicity Category 1...................... < 0.0001 < 0.01 < 0.01 < 0.01 < 1
Toxicity Category 2...................... 0.0001-< 0.001 0.01-< 1 0.01-< 1 0.01-< 1 1-< 50
Toxicity Category 3...................... 0.001-< 0.05 1-< 10 1-< 10 1-< 10 50-< 500
Toxicity Category 4...................... 0.05-< 0.1 10-< 1000 10-< 1000 10-< 1000 500-5000
Toxicity Category 5...................... >0.1 >1000 >1000 >1000 >5000
----------------------------------------------------------------------------------------------------------------
EPA partitioned the cancer-related data elements in the Universe
into the Toxicity Categories as shown in Exhibit 6. The cancer data
placed chemicals in only the three highest Toxicity Categories. EPA did
not use quantitative measures of dose-response for carcinogenicity in
the screening criteria because more chemicals have categorical data and
can be analyzed using this descriptive data than by cancer slope
factors. In addition, EPA
[[Page 9637]]
did not use descriptors indicating lack of carcinogenic potential or
insufficient data to determine carcinogenic potential in categorizing
chemicals because those descriptors apply only to the cancer endpoint
and do not consider noncancer effects associated with exposure to the
chemical.
Exhibit 6.--Partitioning of Cancer Data Based on TD50 Values and Weight-of-Evidence Descriptors
--------------------------------------------------------------------------------------------------------------------------------------------------------
TD50 EPA IARC/HC NTP NCI DSS-Tox
--------------------------------------------------------------------------------------------------------------------------------------------------------
Toxicity Category 1**........... < 0.1 Group A; Human Group 1............ CE 2 species/2 P 2 species/2 H.
Carcinogen. sexes; or 2 sexes; or 2
species; or 2 species; or 2
sexes. sexes.
Toxicity Category 2............. 0.1-100 Groups B1 and B2; Group 2A........... Combinations of CE, Combinations of P, HM.
likely carcinogens. SE, EE, and NE. E and N.
Toxicity Category 3............. >100 Group C; Suggestive Group 2B........... Combinations of SE, Combinations of E M and LM.
evidence of EE, and NE. and N.
carcinogenicity.
--------------------------------------------------------------------------------------------------------------------------------------------------------
** Cancer data placed chemicals in only the three highest Toxicity Categories.
CE = clear evidence, SE = some evidence, EE = equivocal evidence, NE = no evidence.
P = positive, N = Negative, E = equivocal.
H = high probability, HM = high to medium probability, M = medium probability, LM = medium to low probability.
EPA chose a conservative approach in the screening process to
categorize each chemical's toxicity and evaluated all the available
health effects dose-response and categorical data elements for a given
chemical. Chemicals were assigned to the highest toxicity category
indicated after an evaluation of all the available data. Accordingly,
if a chemical had just one data element that places it in Toxicity
Category 1, it was categorized as such even if some of the other data
elements for that same chemical may place it in a lower toxicity
category. For example, if a chemical is classified as a 2A carcinogen
by IARC, it was placed in Toxicity Category 2 using the descriptive
cancer data even if a quantified LOAEL from a different study places it
in Toxicity Category 3.
b. Occurrence Data Elements
EPA evaluated the occurrence data elements for each chemical and
placed them on the horizontal axis of the screening table. In assessing
the data, EPA found that the data elements that represent a chemical's
potential to occur in drinking water vary greatly. EPA's goal was to
determine which data elements best represented the potential to occur
in drinking water. EPA considered and evaluated data elements in the
following categories:
Finished Water--measures of concentration and frequency of
detections.
Ambient Water--measures of concentration and frequency of
detections.
Total Releases in the Environment--pounds per year and
number of States.
Pesticide Application Rates--pounds per year and number of
States.
Production volume--pounds per year.
In addition to evaluating quantitative data elements listed above,
EPA also considered chemicals with descriptive data based upon their
likelihood of occurring in drinking water. Examples of descriptive
occurrence data elements include characterization as a disinfection
byproduct or a drinking water treatment chemical.
EPA used the following hierarchal approach to select the occurrence
data element used to screen a chemical: Finished Water or Ambient Water
> Environmental Release Data > Production Data.
The highest data elements in the hierarchy are the finished and
ambient water data; the lowest, the production data. Environmental
release data from the Toxics Release Inventory (TRI) and pesticide
application amounts occupy the middle position in the hierarchy.
EPA also decided that when multiple data values exist for the
chemicals within a given component of the hierarchy, the most
conservative data value is used. For example, in the case of a chemical
that has finished water data and ambient water data, EPA selected the
highest reported concentration as the occurrence value used in
screening.
EPA obtained the finished water data elements from the National
Contaminant Occurrence Database (NCOD), the Unregulated Contaminant
Monitoring (UCM) Rounds 1 and 2, the National Inorganic Radionuclides
Survey (NIRS), the Unregulated Contaminant Monitoring Regulation (UCMR)
monitoring, the Information Collection Rule database for disinfection
byproducts, the U.S. Department of Agriculture (USDA) Pesticide Data
Program (PDP), and the U.S. Geological Survey (USGS) Pesticides Pilot
Monitoring Program (PPMP). These sources included data elements such as
percent samples with detections, percent drinking water systems with
detections, mean and/or median detected concentrations, and highest
observed concentrations.
EPA obtained ambient water values from the USGS National Water
Quality Assessment Program (NAWQA), the USGS Toxics Substances
Hydrology program's National Reconnaissance of Emerging Contaminants
(NREC) and related studies, and the PPMP. These sources included data
elements such as percent samples with detections, percent sites with
detections, mean and/or median detected concentrations, and highest
observed concentrations.
The environmental release data are those reported for 2004 from the
TRI and the National Pesticide Use Database, developed by the National
Center for Food and Agricultural Policy (NCFAP). The available
environmental release data elements include: total releases to the
environment (lbs/yr), number of States with releases, pesticide total
mass active ingredient applied nationally (lbs/yr), and number of
States with pesticide application. EPA chose to use the pounds released
per year into the environment for screening because the mass applied to
the environment was more directly related to a potential concentration
in water than the number of States where a chemical is released or
applied.
EPA used the Toxic Substances Control Act (TSCA) chemical
production volume ranges reported under the Chemical Update System/
Inventory Update Rule (CUS/IUR) to assess production volume. EPA
selected the most recent year of data available for each particular
chemical. CUS/IUR reports chemical production volume ranges rather than
as exact values of
[[Page 9638]]
release, and provides production data for all chemicals produced in
volumes exceeding 10,000 lbs/yr. The production data are reported in 5
categories that range from less than 10,000 lbs/yr to greater than 1
billion lbs/yr. Therefore, EPA chose to use those ranges as the
occurrence subdivisions for the production data.
The occurrence data were grouped by powers of 10 and arrayed from
low to high across the horizontal axis of the screening table (Exhibit
4). The document entitled ``CCL 3 Chemicals: Screening to a PCCL''
(USEPA, 2008b) describes the analyses in greater detail.
In some cases, disinfection byproducts and water treatment
chemicals lacked quantitative data elements in the Universe. However,
both groups have a strong potential to be present in drinking water.
EPA moved chemicals in these two categories forward to the PCCL for
further evaluation even when limited health effects and/or occurrence
information were available.
c. Selection of the PCCL
The last step in the screening process used the intersections
between health effects and occurrence data elements in the screening
table (Exhibit 4) to establish the PCCL selection line. As noted above,
the health data elements were grouped by the 5 toxicity categories with
the element showing the highest potency determining placement in the
screening table. EPA selected the highest available data element in the
occurrence hierarchy to determine placement of a chemical on the
horizontal axis in the screening table. Because the chemicals were
evaluated using a hierarchical approach for their occurrence elements,
EPA developed separate criteria for each of the occurrence elements,
and used the placement of a group of test chemicals that had all or
nearly all of the occurrence data elements, to establish the position
of the PCCL selection line. The test chemicals were selected from
regulated and past CCL chemicals. Each had data to illustrate whether
it was or was not of concern as a drinking water contaminant.
As a secondary analysis, EPA evaluated existing Drinking Water
Equivalent Levels (DWELs) to confirm whether they would make the PCCL.
The DWELS were derived from the lower RfD potency for each of the RfD
Toxicity Categories. The DWEL (mg/L) is calculated from the RfD in mg/
kg/day by multiplying the RfD by an adult body weight of 70 kg and
dividing by a drinking water intake of 2 L/day (rounded to one
significant figure).When comparing the position of the set of DWELs to
the PCCL selection line, all four toxicity categories would be put on
the PCCL. This analysis supports the position of the PCCL selection
line for chemicals with finished or ambient water concentration data.
EPA also used the test chemicals to determine the PCCL selection
line for the other occurrence data elements--total releases to the
environment (i.e., TRI, pesticide application data) and production
data. For example, the test chemicals were placed in Exhibit 4 based on
their release data to guide the placement of the line that separated
the ``pass to the PCCL'' chemicals from the ``do not pass to the PCCL''
chemicals. In general, the PCCL selection line was positioned so that
regulated and most prior CCL chemicals would be selected for the PCCL.
EPA also analyzed the test chemicals with respect to occurrence,
releases, and production data. The test data fit well for the former
two categories. For the latter, the fit was not as good so EPA chose to
set the PCCL selection line at the point where all chemicals produced
at greater than 100 million pounds per year pass to the PCCL even if
they fall in the lowest toxicity category.
The criteria for moving a chemical with finished or ambient water,
environmental release, and production data to the PCCL are displayed in
Exhibit 7.
Exhibit 7.--Criteria for a Chemical To Pass Screening to the PCCL
----------------------------------------------------------------------------------------------------------------
Occurrence (by data type)
--------------------------------------------------------------------------
Health effects Finished/ambient water Release amount (per Production volume (per
concentrations year) year)
----------------------------------------------------------------------------------------------------------------
Toxicity Category 1.................. All Concentrations..... All Amounts............ All Amounts.
Toxicity Category 2.................. [gteqt]1 [mu]g/l....... [gteqt]10,000 lbs/yr... [gteqt]500,000 lbs/yr.
Toxicity Category 3.................. [gteqt]10 [mu]g/l...... [gteqt]100,000 lbs/yr.. [gteqt]10 M lbs/yr.
Toxicity Category 4.................. [gteqt]100 [mu]g/l..... [gteqt]1 M lbs/yr...... [gteqt]50 M lbs/yr.
Toxicity Category 5.................. [gteqt]1000 [mu]g/l.... [gteqt]10 M lbs/yr..... [gteqt]100 M lbs/yr.
----------------------------------------------------------------------------------------------------------------
EPA added DBPs and drinking water additives that lacked
quantitative occurrence data but fell in the Toxicity Category 1 or
Toxicity Category 2 groupings to the PCCL because of their high
probability for being present in disinfected and treated drinking
water.
The screening process provides a data-driven, objective, and
transparent process for selecting the PCCL from the Universe. All
Toxicity Category 1 chemicals (i.e., most toxic) were captured
regardless of their occurrence category. The occurrence threshold
required for the PCCL selection became less inclusive as the
contaminant toxicity decreased. The screening of the CCL 3 Universe
resulted in the selection of 532 chemical contaminants for the PCCL
from the approximately 6,000 chemicals that were screened. The
categorical summary of chemicals that passed the screening is
illustrated in Exhibit 8. A complete chemical PCCL list can be found in
Appendix B of the document entitled, ``CCL 3 Chemicals: Screening to a
PCCL'' (USEPA, 2008b). The 532 PCCL chemicals were further scrutinized
as part of the next key step in the process. Some of the contaminants
on the PCCL had limited data available for the scoring protocols and
could not be run through the models. The 32 contaminants that had
limited data identified in the appendixes to the ``Classification of
the PCCL to the CCL'' support document (EPA 2008c) and will remain on
the PCCL until new data are identified for further evaluation.
[[Page 9639]]
Exhibit 8.--Summary of Total Chemicals That Passed Screening for PCCL by Screening Categories
----------------------------------------------------------------------------------------------------------------
Finished or
Toxicity categories ambient water Pesticide Total Production Totals
concentration app releases volume
----------------------------------------------------------------------------------------------------------------
Toxicity Category 1...................... 29 4 56 38 127
Toxicity Category 2...................... 33 26 32 61 152
Toxicity Category 3...................... 36 31 21 66 154
Toxicity Category 4...................... 5 4 10 63 82
Toxicity Category 5...................... 0 0 0 17 17
----------------------------------------------------------------------------------------------------------------
3. Using Classification Models To Develop the CCL 3
The 532 PCCL chemicals were further scrutinized as part of this key
step in the process by using classification models as tools to aid in
the selection of the draft CCL 3. As experience is gained, the EPA
expects to modify and improve the development of the classification
process for future CCLs.
From the inception of the development of the CCL classification
process, EPA intended to use classification models as a decision
support tool. EPA envisioned that, after testing and evaluation, models
would be used to process complex data in a consistent, objective, and
reproducible manner and provide a prioritized listing of candidate
contaminants for the last stage of the CCL process--an expert review
and evaluation. Model application also would help EPA focus resources
for the expert review and evaluation of the highest priority potential
contaminants.
An overview of the classification model approach used to further
evaluate chemicals on the PCCL is described in the following sections.
A detailed discussion of the process is provided in a document
entitled, ``Contaminant Candidate List 3 Chemicals: Classification of
the PCCL to the CCL'' (USEPA, 2008c). The development of this
classification process involves the following steps:
Development of the Attribute Scoring Protocols.
Development of the Training Data Set.
Application of the Classification Models.
Evaluation of Classification Model Output and Selection of
the CCL.
To use models to evaluate and classify the PCCL contaminants for
listing on the CCL, EPA needed to develop methods to interrelate the
important measures (i.e., attributes) that represent a contaminant's
health effects and potential for occurrence in drinking water. Four
attributes were selected: Potency, severity, prevalence, and magnitude.
Protocols were developed for scoring each attribute.
EPA also tested and evaluated the results of several classification
models to determine which ones might provide the best decision support
tools. To make this evaluation, EPA developed a chemical data set and
used the data set to ``train'' the classification models. The selected
models were utilized to process the data for the PCCL chemicals and
provide a prioritized listing of candidate contaminants for the expert
review and evaluation.
a. Development of the Attribute Scoring Protocols
EPA used attributes to characterize different chemicals on the
basis of similar qualities or traits. These qualities or traits
represent the likelihood of occurrence or potential for adverse health
effects of each contaminant. Throughout the process of evaluating the
attributes EPA recognized that a wide range of data elements would have
to be used for each attribute to characterize chemicals on the PCCL. To
evaluate PCCL chemicals with differing types of occurrence and health
effects data as potential CCL contaminants, one must be able to
establish consistent relationships among the different types of data
that represent measures of the attributes. If the same data were
available for all contaminants, the comparison and prioritization of
candidates would be less complex. To consistently apply the best
available data for PCCL chemicals, EPA normalized the different types
of data into scales and scoring protocols that accept a variety of
input data, apply a consistent framework, and compare different types
of data. The following sections describe how EPA developed the scales
and scoring protocols for the health effects and occurrence attributes.
i. Health Effects Attributes
Potency and severity are the attributes used to describe health
effects. EPA defines potency as the lowest dose of a chemical that
causes an adverse health effect and severity is based on the adverse
health effect associated with the dose used to define the measure of
potency. In other words, potency was scored on the dose that produced
the adverse effect and severity was scored based on the health-related
significance of the adverse effect (e.g., from dermatitis to organ
effects to cancer). These two attributes are interrelated, in that the
severity is linked to the measure of potency.
The following toxicological parameters were used to evaluate
potency:
Reference Dose (RfD) or equivalent.
Cancer potency (concentration in water for 10-4
cancer risk).
No-Observed-Adverse-Effect Level (NOAEL).
Lowest-Observed-Adverse-Effect Level (LOAEL).
Rat oral median Lethal Dose (LD50).
EPA developed a ``learning set'' of about two hundred chemicals to
calibrate the potency scoring protocols. Once the data for the learning
set of chemicals was collected, EPA arrayed and graphically displayed
the data to analyze their range and distribution. EPA selected a
distribution based on logarithms (base 10) of the toxicity parameters
rounded to the nearest integer because it provided a spread of the
chemical toxicity parameters across the range and the curve was roughly
log-normal.
EPA used a log-based distribution to establish a potency scoring
equation for each toxicity parameter. This was accomplished by
assigning the most frequent (modal) value in each distribution a score
of 5 on a 10 point scale. When the toxicity parameter was one log more
toxic than the modal value, a score of 6 was assigned. Similarly, when
the parameter was one log less toxic than the modal value a score of 4
was given, and so on. EPA developed an equation for each toxicity
parameter that equated the modal value to a score of 5 and calculated
the potency score. Because the modal rounded log differed for the
different measures of toxicity, it was necessary to use a different
equation for each to normalize the mode to a score of 5. The
[[Page 9640]]
resultant equations are summarized in Exhibit 9.
Exhibit 9.--Scoring Equations for Potency
------------------------------------------------------------------------
-------------------------------------------------------------------------
RfD Score = 10 - (Log10 of RfD + 7).
NOAEL Score = 10 - (Log10 of NOAEL + 4).
LOAEL Score = 10 - (Log10 of LOAEL + 4).
LD50 Score = 10 - (Log10 of LD50 + 2).
10-\4\ cancer risk Score = 10 - (Log10 of the 10-4 cancer risk + 6).
------------------------------------------------------------------------
For distributions that spanned more than 5 orders of magnitude
above or below the mode, scores for the tails of the distribution were
truncated at 1 and 10. Conversely, for distributions that did not span
5 full orders of magnitude above and below the mode, not all scores
between 1 and 10 were used. For example, the distribution of the 10-4
values for cancer risk was skewed, with values up to 5 orders of
magnitude above the modal value (more potent carcinogens) but only 2
orders of magnitude below the mode (less potent carcinogens). This
meant that the lowest potency score for this toxicity parameter was a
``3.''
EPA tested the scoring process by using a subset of contaminants
with values from multiple data elements considered in the process. In
the testing of the potency scoring process, EPA scored all of the
chemicals in the learning set for each toxicity parameter to examine
the consistency across scores for the non-cancer measures of potency.
EPA evaluated the agreement of non-cancer scores across the RfD, NOAEL,
LOAEL and LD50 inputs and found the scores for any given compound to be
generally consistent across parameters. Because of the general
consistency among scores, EPA determined that a hierarchy of RfD>
NOAEL> LOAEL> LD50 would be used in the scoring of potency. This
hierarchy gives preference to the potency value with the richest
supporting data set (the RfD--or equivalent values) and gives the
lowest ranking to the LD50 because it is a measure of acute rather than
chronic toxicity. If data are available for both the cancer and
noncancer endpoints, the higher of the cancer or noncancer potency is
selected and the critical effect of the higher measure of potency is
used to score the severity.
Severity refers to the relative impact of an adverse health affect.
Just as toxicity increases with dose, the severity of the observed
effect also increases. A low dose effect could be a simple increase in
liver weight while the same chemical at a higher dose could cause
cirrhosis of the liver. For consistency, the measure of severity that
was used for scoring the PCCL chemicals was the effect or effects seen
at the LOAEL. Restricting severity scores to the effects at the LOAEL
ties them to the data used to derive the potency score.
The severity measures used to score the PCCL chemicals differ from
those used for potency, prevalence, and magnitude because they are
descriptive rather than quantitative. Accordingly, they are less
amenable to automation and often require more scientific judgment in
their application. To guide scoring for severity, EPA developed the
nine-point scale displayed in Exhibit 10, and a compendium of nearly
250 descriptions of critical effects grouped by their severity scores
(e.g., ``Chronic irritation without histopathology changes'' equals a
score of 3).
Exhibit 10.--Final Nine-Point Scoring Protocol for Severity
------------------------------------------------------------------------
Score Critical effect Interpretation
------------------------------------------------------------------------
1................. No adverse effect........
2................. Cosmetic effects......... Considers those effects
that alter the
appearance of the body
without affecting
structure or functions.
3................. Reversible effects; Transient, adaptive
differences in organ effects.
weights, body weights or
changes in biochemical
parameters with minimal
clinical significance.
4................. Cellular/physiological Considers cellular/
changes that could lead physiological changes in
to disorders (risk the body that are used
factors or precursor as indicators of disease
effects). susceptibility.
5................. Significant functional Considers those disorders
changes that are in which the removal of
reversible or permanent chemical exposure will
changes of minimal restore health back to
toxicological prior condition.
significance.
6................. Significant, Considers those disorders
irreversible, non-lethal that persist for over a
conditions or disorders. long period of time but
do not lead to death.
7................. Developmental or Considers those chemicals
reproductive effects. that cause developmental
effects or that impact
the ability of a
population to reproduce.
8................. Tumors or disorders Considers chemical
likely leading to death. exposures that result in
a fatal disorder and all
types of tumors.
9................. Death....................
------------------------------------------------------------------------
Severity scores 1 through 6 represent a progression in the severity
of the observed effect. Severity score 7 is used for all studies where
the effect observed is a reproductive and/or developmental effect
allowing the Agency to track the chemicals that pose developmental or
reproductive concerns consistent with the 1996 SDWA. A severity score
of 8 was used to track all cases where cancer is the basis for the
potency score.
ii. Occurrence Attributes
EPA used prevalence and magnitude to describe the potential to
occur in drinking water. Prevalence measures how widespread the
occurrence of the contaminant is in the environment or how widely the
contaminant may be distributed. The prevalence measure indicates the
percent of public water systems or monitoring sites across the nation
with detections, number of States with releases, or the total pounds
produced nationally. Magnitude relates to the quantity of a contaminant
that may be found in the environment. The magnitude measures include
the median concentration of detections in water or the total pounds of
the chemical released into the environment. In most cases the same data
element (e.g., detections in drinking water or amount released into the
environment) could be used to determine the prevalence, based on the
spatial distribution and magnitude based on the amounts. However, where
production data were used to determine prevalence, there was no
corresponding direct measure of magnitude, so persistence and mobility
data were used as surrogate indicators of potential magnitude.
Production/persistence and mobility data are assigned the lowest
level in the hierarchy of data available for prevalence and magnitude.
Persistence-mobility is determined by chemical
[[Page 9641]]
properties that measure or estimate environmental fate characteristics
of a contaminant and affect their likelihood to occur and persist in
the water environment. Data sources that could provide occurrence data
ranged from direct measure of concentrations in water to annual
measures of environmental release or production. EPA compiled a second
subset or learning set of 207 chemicals, with available data for all of
the occurrence attribute data elements that measured prevalence and
each of the data elements that measured magnitude, to calibrate
protocols for prevalence and magnitude.
The data available for the prevalence attribute consisted of
measurements of a contaminant's occurrence across the United States.
The prevalence measures have finite ranges such as zero to 100 percent
of samples/sites or 1 to 50 States depending on the reporting
requirements of the available data source. Accordingly, the scaling of
scores for prevalence focused on establishing appropriate groupings of
the number of sites or States impacted across the 1 to 10 scoring
scale.
The relationship between production or even environmental release
data and the actual occurrence in drinking water is complex. Where
actual water measurements are available, they are the preferred data
element to score prevalence because they are the most direct measures
of occurrence in drinking water. EPA selected the following hierarchy
for scoring prevalence:
Percent of PWSs with detections (national scale data).
Percent of ambient water sites or samples with detections
(national scale data).
Number of States reporting application of the contaminant
as a pesticide.
Number of States reporting releases (total) of the
chemical.
Production volume in lbs/yr.
The production data provide the pounds produced annually of a chemical
product in the United States. To some extent, this production rate
represents the commercial importance of the chemical, so EPA
interpreted the high production tonnage as a likely indication of wide
use of a commodity chemical and used this information to score
prevalence. For example, a chemical produced at a billion lbs/yr is
more likely to be used and released more widely than a compound
produced at only 10,000 lbs/yr.
Magnitude represents the quantity of a contaminant that may be in
the environment. The data sources that provided the first four levels
of the prevalence hierarchy provided direct measurements of water and
environmental release that could be used to score magnitude. However,
the production categories did not supply an appropriate measure for
magnitude. EPA used the persistence and mobility for chemicals with
only production data as the basis of the magnitude attribute.
To keep the process straightforward, EPA used one scale for all
water concentration data. EPA distributed scores across the range of
values so that organic contaminants could receive high scores as well
as the inorganic contaminants (IOCs). Comparisons and adjustments were
made until there was a reasonable distribution of the scores for
organic and inorganic contaminants by using a semi-logarithmic scale.
EPA selected the single scale approach and this is discussed in more
detail in the report entitled ``CCL 3 Chemicals: Classification of the
PCCL to the CCL'' (USEPA, 2008 c).
When developing the calibration scales for the release data, the
ranges of data were similarly arrayed using a scale based on half-log
units with a distribution of scores that reflected the distribution of
the data in the learning set.
EPA based the persistence and mobility scores on chemical and
physical properties combined with environmental fate parameters.
Persistence and mobility act as measures of potential magnitude because
both fate (i.e., persistence) and transport (i.e., mobility) affect the
amount of a contaminant to be found in water. The length of time a
chemical remains in the environment before it is degraded (persistence)
affects its concentration in water. Similarly, the mobility of a
chemical, or its ability to be transported to and in water, affects its
potential to reach and dissolve in the source waters, and thus, the
ultimate concentration of the chemical in the water.
EPA considered a number of data elements to measure the mobility of
a chemical in the environment. The physical/chemical parameters that
were chosen for the CCL process are:
Organic Carbon Partition Coefficient (Koc)
Octanol/Water Partition Coefficient (Kow)
Soil/Water Distribution Coefficient (Kd)
Henry's Law Coefficient (KH)
Solubility
The first 4 measures of mobility represent the equilibrium ratio for
the partitioning of the contaminant from one medium to another:
Koc (soil/sediment organic carbon: water), Kow
(octanol: water), Kd (soil/sediment: water) and Henry's Law
Coefficient (air: water). Koc, Kow and Kd
are sometimes expressed as logs of the original measurements. The
measures of persistence reflect the time the chemical will remain
unchanged in the environment. Persistence is reflected in the following
measures of environmental fate:
Half-Life
Measured Degradation Rate
Modeled Degradation Rate
Each of the mobility and persistence data elements listed above are
presented in hierarchical order, with the most desirable at the top
(i.e., the first data to be used if available).
As was the case with prevalence, EPA used a hierarchy in scoring
magnitude. The hierarchy uses finished water occurrence data if
available, and if not, the highest available element in the hierarchy
of finished water data > ambient water data > environmental release
data > persistence and mobility data. The data elements used in scoring
magnitude follow:
Median value of detections from finished water systems
(PWSs) (national scale data)
Median value of detections from ambient water sites or
samples (national scale data)
Amount of pesticide applied (annual, in pounds)
Amount of total releases (annual, in pounds)
Persistence and mobility data
EPA developed attribute scoring protocols through a step-wise
process of data selection, data analysis, calibration of scales, and
evaluation of the functionality of the scores in PCCL to CCL decision-
making. This is discussed in more detail in the report entitled
``Contaminant Candidate List 3 Chemicals: Classification of the PCCL to
the CCL'' (USEPA, 2008 c). EPA used the attribute protocols to
normalize the data for the PCCL chemicals and develop a set of scores
for the four attributes that are the input into the models. By
normalizing the data elements, EPA developed a process that can use
different kinds of data and information (e.g., quantitative and
descriptive) to develop input to the models and provide a relative
score for potential contaminants using the attribute scores.
b. Training Data Set for the Classification Models
The training data set (TDS) for chemicals is the set of data used
to train
[[Page 9642]]
(or teach) the classification models to mimic EPA expert list-not list
decisions for PCCL chemicals. EPA compiled this data set in addition to
the two learning sets to represent the types of chemicals likely to
move forward to the PCCL. This data set also represents the range of
possible attribute scores and listing decisions needed to train and
calibrate the classification models. The TDS used to train the models
for CCL 3 was comprised of 202 discrete sets of attribute scores for
chemicals and consensus list-not list decisions made by a team of EPA
subject matter experts.
Classification models use statistical approaches for pattern
recognition and derive mathematical relationships among input variables
(e.g., measurements or descriptive data) and output from a TDS. EPA
used classification models to develop a relationship between the
contaminant attribute scores (input variables) and the classification
of these contaminants into list-not list categories (output). EPA
subject matter experts familiar with the technical aspects of the
attribute data and the selection of drinking water contaminants for
listing and regulation made the list-not list decisions for the TDS.
EPA then applied the models to the PCCL to predict likely list-not list
decisions.
EPA considered the following key factors in developing the training
data set:
Selection of contaminants representing a range of outcomes
and decisions likely to be encountered in developing a CCL;
A variety of input data ensuring adequate coverage of
attribute scores and combinations of scores;
Chemicals that, when present in drinking water, would
present a meaningful opportunity for public health improvement if
regulated; and
Contaminants that would likely be selected for the PCCL.
The TDS used for training the classification models consisted of
202 combinations of attribute scores and the decisions made by EPA
experts. The TDS included some of the contaminants from the learning
sets used in developing the scoring protocols for toxicity and
occurrence. It also included additional contaminants to meet the key
factor requirements described above. The set of known chemicals chosen
for the TDS was supplemented with a set of attribute scores and
decisions that were selected to balance the range of scored attributes
the classification model would need to evaluate as described further
below.
Initially, EPA selected ``data rich'' contaminants from among
regulated contaminants and previous CCLs because they had a range of
readily available occurrence and health effects information. EPA
drinking water subject matter experts and stakeholders reviewed the
initial list of contaminants and identified additional candidates for
the TDS. This initial selection process identified 51 chemical
contaminants. Subsequently, EPA randomly chose 50 contaminants from
chemicals in the CCL 3 Universe with high health effects potency values
and accompanying occurrence data because they represented contaminants
likely to make it to the PCCL. The addition of these 50 contaminants
resulted in 101 contaminants with data to score attributes.
The performance of the classification models using the initial TDS
gave an indication of gaps in the possible attribute space that the set
of 101 TDS contaminants did not adequately cover. This led EPA to add
the sets of possible attribute scores to the TDS based on Latin
hypercube sampling (NIST, 2006; http://www.itl.nist.gov/div898/handbook/glossary.htm#LHC
). Using this approach, EPA added 101 specific
combinations of attribute scores to fill in gaps in the space defined
by total possible attribute scores and improve the performance of the
models. This set of 202 scores and decisions ensured good coverage of
both ``list'' and ``not list'' outcomes and became the TDS. Models
trained with the TDS with 202 decisions had greater agreement with EPA
subject matter experts than those trained with the TDS of 101
contaminants.
List-not list decisions were a key component of the TDS. EPA
subject matter experts made list-not list decisions as individuals and
as a group, based on attribute scores and based on data that had not
been converted to attribute scores (actual or raw data). The
development of the list-not list decisions was an iterative process
that incorporated revisions to the attribute scoring protocols as
experience was gained by the EPA experts. EPA resolved differences
between the decisions based on the scored attributes and the raw data
by revising the scoring protocols based on the EPA experts' experience
to improve the correlation of decisions based on scores to those based
on raw data.
EPA subject matter experts reviewed and evaluated the health
effects and occurrence data for each contaminant. Each individual
reviewer made decisions about how to classify the contaminant and then
met as a group to discuss their decisions. Early in the process the
reviewers recognized that clear list or not-list decisions could easily
be made for some contaminants, but not for other contaminants. For the
chemicals where the decision whether to list contaminants was not
clear, two categories were added to the analyses. The categories of
List? (L?) or Not List? (NL?) allowed the group to identify chemicals
that were close to the boundary for a List-Not List decision. That is
L? signifies that the decision is leaning towards listing but with some
uncertainty, and NL? signifies that the decision is leaning towards not
listing but with some uncertainty. These additional two categories were
incorporated into the evaluation and model training process.
The EPA subject matter experts also reached a consensus decision
for each contaminant. This consensus decision was used to train the
models. This is discussed in more detail in the report entitled
``Contaminant Candidate List 3 Chemicals: Classification of the PCCL to
the CCL'' (USEPA, 2008c).
c. Evaluation of Classification Models
EPA identified several different models for possible use in
selecting contaminants from the PCCL for the CCL: Artificial neural
networks, classification decision trees, linear models, and
multivariant adaptive regression splines. EPA evaluated the
classification models in a two-step process. The first step was the
evaluation and selection of models from within each of the model
classes that best predicted the consensus decisions of the subject
matter experts. The second step was the evaluation of the performance
of the best models selected from each class (USEPA, 2008c).
EPA evaluated models based on the 4 attributes that the model was
able to consider, the types of relationships or mathematical functions
that the model utilized, and the model's ability to predict
classifications of the TDS. The iterative training process minimized
the model's predictive error, thereby reducing incorrect model
predictions. EPA also evaluated the impact of the attributes used by
the models and the effects of missing data on the performance of the
models during the various stages of development.
EPA evaluated the performance of five models. Three models,
Artificial Neural Network (ANN), Quick, Unbiased and Efficient
Statistical Tree (QUEST), and Linear Regression demonstrated consistent
performance when trained and evaluated with the TDS. The classification
models were assessed and compared with respect to:
[[Page 9643]]
The number of correct and incorrect classifications for
the 202 TDS contaminants.
The number of ``large'' misclassifications (off by more
than one category).
The weighted sum of TDS classification errors.
Ability to identify intermediate classifications.
Consistent behavior (e.g., no decreasing classification as
attribute scores increase).
This is discussed in more detail in the report entitled ``Contaminant
Candidate List 3 Chemicals: Classification of the PCCL to the CCL''
(USEPA, 2008c).
d. Application and Use of Model Results
From the inception of the development of the CCL classification
process, EPA intended to use classification models as decision support
tools. It was envisioned that the models would be used to process
complex data in a consistent, objective, and reproducible manner and
provide a prioritized listing of contaminants, allowing EPA to focus
resources on the expert review and evaluation of the highest priority
potential contaminants. The ANN, Linear, and QUEST models are three
different classes of models, with three different mathematical
approaches, yet they all provided similar results and logical
determinations. EPA explored simple ways to combine the results of all
three models, to capture both agreement among models and unique
results. Both a straightforward, additive approach, and a collective,
rank-order approach were utilized to provide a prioritized listing of
contaminants to be considered further and evaluated for possible
inclusion on the draft CCL 3.
e. Model Outcome and Expert Evaluation
In the last step of the process, the chemicals on the PCCL were
scored for their attributes and evaluated by the three models. Some of
the contaminants on the PCCL had limited data available for the scoring
protocols and could not be run through the models. The 32 contaminants
that had limited data are identified in the appendixes to the
``Classification of the PCCL to the CCL'' support document (EPA 2008c)
and will remain on the PCCL until new data are identified for further
evaluation. As part of the evaluation of model output, EPA formulated
several post-model refinements that were added to the CCL selection
process. Exhibit 11 illustrates the results of the model output for the
PCCL contaminants. The PCCL consisted of chemicals with variable health
effects data, ranging from reference doses (RfD) to Lethal Dose 50s
(LD50), and occurrence data ranging from measured water
concentration data from Public Water Systems (PWS) to production volume
data.
Exhibit 11.--Model Results for the PCCL Chemicals
----------------------------------------------------------------------------------------------------------------
Total Finished
3-Models decision % of PCCL or ambient Release Production
PCCL water
----------------------------------------------------------------------------------------------------------------
L............................................. 9 44 3 24 17
L-L?.......................................... 12 58 9 29 20
L?............................................ 33 163 26 64 73
NL?-L?........................................ 6 30 6 11 13
NL?........................................... 28 139 29 28 82
NL?-NL........................................ 4 20 7 9 4
NL............................................ 9 46 21 7 18
N (all)....................................... 100 500 101 172 227
----------------------------------------------------------------------------------------------------------------
Four of the seven decision categories, L, L?, NL?, NL, in the first
column of Exhibit 11 signify that all of the models were in unanimous
agreement with the listing decision. The other categories (e.g., NL?-
L?) represent varied agreement where one or two of the models chose one
category and the other model(s) resulted in a different category. Note
that none of the models placed a contaminant in a category more than
one category higher or lower than the other models. That is, no
contaminants were categorized as ``L'' by one model and as ``NL?'' by
one of the other models, or visa versa. The models categorized
approximately one-half of the chemicals on the PCCL as L? or above.
When analyzed by data type, the majority of chemicals in the List
category used LD50 data for health effects. This was a
concern and became an important issue for consideration. The role
LD50 played in the health effects scoring was discussed
extensively during the post-model evaluation process.
As part of the last stage in the CCL classification process, the
model output was reviewed by a group of internal EPA experts
representing several offices. This step involved a detailed review of
the data used for the models and the available supplemental data for
the chemicals. The EPA experts also deliberated on the method of using
the model data to produce a draft proposal for CCL 3. The function of
this review was to critically compare the results from the model to the
data for the chemicals for a cross section of the modeled contaminants.
Based upon issues identified by the evaluators, several post model
refinements were added to the CCL process. Three major issues and
refinements are described below.
The relationship between potency and concentration was important
when deciding whether to list a chemical. However this ratio could only
be developed when water concentration data were available. Accordingly,
calculation of the ratio between the health-based value and the 90th
percentile concentration in finished or ambient water was added as a
post-model process. The potency/concentration ratio serves as a
benchmark that suggests a greater concern for a contaminant if the
ratio is low and a lesser concern when it is high.
The addition of modeled occurrence data for pesticides and
estimated concentration in surface and ground water was obtained from
the EPA Office of Pesticide Programs (OPP). The modeled estimates of
concentration in water for pesticides are part of the EPA's pesticide
registration and re-registration evaluations. Once the availability of
the OPP data for some of the pesticides was confirmed, the data were
extracted from OPP documents and used to generate a potency/
concentration ratio similar to that used with the water concentration
data.
Data certainty was factored into the decision process by
characterizing health effect and occurrence data
[[Page 9644]]
elements and their relative certainty based upon the type of data that
was used to score the attribute for the model classification. This
characterization tagged data elements with high certainty and low
certainty. The combined certainty measure for a single contaminant
(i.e., health effects and occurrence tags) was used to place
contaminants in bins of high, medium and low certainty.
The high certainty bin consisted of chemicals with direct
occurrence measured in water and well-studied data for health effects.
Such contaminants are expected to be good candidates for regulatory
determination because they provide information that can be considered
in that process and have minimal research needs. Examples of the data
used to characterize chemicals in the high certainty bin include
chemicals with RfDs, LOAELs, and NOAELs, and water concentration data.
The medium bin consists of chemicals that will need further occurrence
and/or health effects research. For example, chemicals with well
studied health effects that only have environmental release data are
included in the medium bin. Chemicals that are released to the
environment and need further health effects research are also included
in the medium bin. The low certainty bin consists of chemicals that
have limited data, yet these data suggest that further evaluation
should be pursued. These chemicals may need extensive health effects
and occurrence research that may require significant resources before
regulatory determinations can be made. Examples include chemicals with
only LD50 and/or production volume data. The CCL should
consist both of chemicals that provide sufficient data to support
regulatory determinations as well as chemicals that are of concern and
need to be targeted for additional drinking water research.
Contaminants from each bin were scrutinized separately in selecting
which ones should be listed on the CCL 3.
4. Selection of the Draft CCL 3--Chemicals
The chemicals for the draft CCL 3 were selected from within the
three certainty bins with the emphasis placed on the source of the
occurrence data (e.g., measured concentrations, release, and
production). Four groups of chemicals were placed on the CCL based on
their modeled scores, the potency-concentration ratios, where
available, and the estimate of data certainty. They included:
36 chemicals in the high certainty bin with finished or
ambient water data and a potency/90th percentile concentration ratio
< =10.
24 pesticide chemicals in the medium certainty bin with
modeled surface and/or ground water data that yielded a potency/
concentration ratio < =10.
27 chemicals in the medium certainty bin with release data
that gave modeled L or L-L? rankings.
8 chemicals in the low certainty bin that were added to
the CCL as recommended by the public in response to EPA's Federal
Register notice (71 FR 60704, USEPA, 2006b). The notice requested that
the public submit chemical and microbial contaminant nominations that
should be considered for CCL 3. This process is discussed in section
III.C.1.
The potency and concentration were compared to develop a ratio that
was used to select contaminants for the draft CCL 3 from the high
certainty bin. A ratio between the health-based value and the 90th
percentile was taken for chemicals with measurements in finished and
ambient water. Contaminants for this bin were selected for the draft
CCL 3 when the ratio was < =10, representing occurrence in water at a
level of concern related to its health effects data.
The pesticides in the medium bin, where modeled data was obtained
from OPP, were selected for the draft CCL 3 based on their potency/
concentration ratios. Similar to the chemicals in the high certainty
bin, pesticides were selected for the draft CCL 3 when the potency/
concentration ratio was < 10, representing potential occurrence in water
at a level of concern related to its health effects data. The other
chemicals in the medium bin were selected for the draft CCL 3 based on
a review of their data and their prioritization from the classification
models.
Chemicals in the low certainty bin were selected for the draft CCL
3 based on a review of their supplemental data and the data submitted
through the nominations process. Some of the chemicals identified
through the nominations process were already on the draft CCL 3 based
on the data EPA collected for the universe. The supplemental data
provided with the nominations were used to screen the nominated
chemicals and score the attributes for those that passed the screen.
The scored attributes were then processed through the models and the
post-model evaluations. Those that were listed demonstrated adverse
health effects and a potential to occur in PWSs. Chemicals not selected
for the draft CCL 3 will remain on the PCCL until additional occurrence
or health effects data become available to support their reevaluation.
B. Classification Approach for Microbial Contaminants
As discussed in CCL 2 (USEPA, 2005b), the Agency evaluated the
NDWAC, NRC and other recommendations, and used the information to
develop a pragmatic approach for classifying the microorganisms on the
draft CCL 3. The CCL 3 approach for microbes, like the approach used
for chemicals, uses the attributes of occurrence and health effects to
select the microbial contaminants. EPA's objective is to target
microorganisms with the highest potential for human exposure and the
most serious adverse health effects. Parallel to the chemical selection
process, the Agency considers a broad universe of microbial
contaminants and systematically narrows that universe down to develop
the draft CCL 3 in a transparent and scientifically sound CCL process.
The first step of the CCL 3 approach for microbes identifies a universe
of potential drinking water contaminants. The second step screens that
universe of microbiological contaminants to a Preliminary Contaminant
Candidate List (PCCL). Lastly, EPA selects the draft CCL 3 microbial
list by ranking the PCCL contaminants based on occurrence in drinking
water (including waterborne disease outbreaks) and human health
effects.
1. Developing the Universe
EPA defined the microbial Universe for the draft CCL 3 as all known
human pathogens. The Universe process began with the list of 1,415
recognized human pathogens compiled by Taylor et al. (2001). The Agency
added organisms to the Universe and updated nomenclature in Taylor et
al. (2001) to account for emerging pathogens and new taxonomy research.
As EPA reviewed Taylor et al. (2001), additional pathogens were
also identified. EPA surveyed fungi in drinking water and identified
six fungi reported to occur in drinking water distribution systems that
did not appear on the Taylor list. The added fungi are shown in Exhibit
12. EPA also added reovirus to the Universe based on additional health
effects information (Tyler, et al., 2004).
In October 2006, EPA published a notice (71 FR 60704 (USEPA,
2006b)) requesting chemical and microbial contaminant nominations as
part of the process to identify emerging contaminants that should be
considered for the CCL. As a result of the
[[Page 9645]]
nominations process, 24 microbial contaminants were nominated by the
public. Twenty-two of the microbes were previously identified by Taylor
et al. (2001) and are already in the Universe. The two additional
pathogens nominated were Methylobacterium (with two species) and
Mimivirus. These two bacterial species, two viral groups and six fungal
species were added to the Microbial Universe which brings the Microbial
Universe list to 1,425 pathogens. The full Universe list is available
in the document, ``Contaminant Candidate List 3 Microbes: Identifying
the Universe'' (USEPA, 2008d).
Exhibit 12.--Fungi Added to the Microbial Universe
------------------------------------------------------------------------
Pathogen
-------------------------------------------------------------------------
Arthrographis kelrae
Chryosporium zontatum
Geotrichum candidum
Sporotrichum pruinosum
Stachybotrys chartarum
Stemphylium macrosporoideum
------------------------------------------------------------------------
2. The Universe to PCCL
EPA developed screening criteria to reduce the Universe of all
human pathogens to just those pathogens that could be transmitted
through drinking water. For example, pathogens transmitted solely by
animals, such as the virus that causes rabies, were screened out of the
Universe and are not included on the PCCL. Screening is based on a
pathogen's epidemiology, geographical distribution, and biological
properties in their host and in the environment. EPA moved pathogens
forward to the PCCL if there was any evidence linking a pathogen to a
drinking water-related disease. The screening criteria restrict the
microbial PCCL to human pathogens that may cause drinking water-related
diseases resulting from ingestion of, inhalation of, or dermal contact
with drinking water. EPA used 12 screening criteria (Exhibit 13) to
reduce the pathogens in the microbial CCL universe to the PCCL.
Exhibit 13.--CCL Screening Criteria for Pathogens
------------------------------------------------------------------------
-------------------------------------------------------------------------
1. All anaerobes.
2. Obligate intracellular fastidious pathogens.
3. Transmitted by contact with blood or body fluids.
4. Transmitted by vectors.
5. Indigenous to the gastrointestinal tract, skin and mucous membranes.
6. Transmitted solely by respiratory secretions.
7. Life cycle incompatible with drinking water transmission.
8. Drinking water-related transmission is not implicated.
9. Natural habitat is in the environment without epidemiological
evidence of drinking water-related disease.
10. Not endemic to North America.
11. Represented by a pathogen for the entire genus or species (that are
closely related).
12. Current taxonomy changed from taxonomy used in Universe.
------------------------------------------------------------------------
Pathogens meeting any single criterion of the 12 criteria were
removed from further consideration and not moved forward to the PCCL.
Based upon this screening exercise, 1,396 of the 1,425 pathogens were
excluded and 29 pathogens moved on to the PCCL. The results of the
screening process are summarized in Exhibit 14. The screening criteria
and results of the screening process are discussed in greater detail in
the supporting document titled ``Contaminant Candidate List 3 Microbes:
Screening to the PCCL'' (USEPA, 2008 e).
Exhibit 14.--Application of Twelve Screening Criteria to Pathogens in the Microbial CCL Universe
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Screening Criteria Pathogens
Pathogen class Total ------------------------------------------------------------------------------------------------ screened On PCCL
1 2 3 4 5 6 7 8 9 10 11 12 out
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Bacteria.................................................... 540 125 14 10 37 117 7 0 29 154 2 28 5 528 12
Viruses..................................................... 219 0 0 26 104 0 19 1 18 0 36 8 0 212 7
Protozoa.................................................... 66 0 0 1 29 3 0 4 7 7 0 6 0 57 7*
Helminths................................................... 287 0 0 0 25 0 0 106 0 0 156 0 0 287 0
Fungi....................................................... 313 0 0 0 0 12 1 0 0 297 0 0 0 310 3
Total................................................... 1,425 125 14 37 195 132 27 111 54 458 194 42 5 1,394 29*
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
* Two additional protozoa,Cryptosporidium and Giardia were not considered for CCL 3 and they are discussed in more detail later.
3. The PCCL to Draft CCL Process
Pathogens on the PCCL were scored for placement on the draft CCL.
EPA devised a scoring system to assign a numerical value to each
pathogen on the PCCL.
Each of the pathogens on the PCCL was scored using three scoring
protocols, one protocol each for waterborne disease outbreaks (WBDO),
occurrence in drinking water, and health effects. The higher of the
WBDO score or the occurrence score is added to the normalized health
effects score to produce a composite pathogen score. Pathogens
receiving high scores were considered for placement on the CCL.
EPA normalized the health effects score so that occurrence and
health effects have equal value in determining the ranking of the CCL.
The equal weighting of occurrence and health effects information
closely mirrors the risk estimate methods used by EPA during drinking
water regulation development. This scoring system prioritizes and
restricts the number of pathogens on the CCL to only those that have
been strongly associated with drinking water-related disease. Pathogens
that scored low will remain on the PCCL until additional occurrence
data, epidemiological surveillance data, or health effects data become
available to support their reevaluation. It is important to note that
pathogens for which there are no data documenting a waterborne disease
outbreak in drinking water earn a low score under the protocols. EPA
believes that pathogens that have caused a WBDO and have health effects
data