v>EPA
United States
Environmental Protection
Agency
Office of the Science Advisor
STAFF PAPER
100B04001
-------
-------
EPA/100/B-04/001
March 2004
U.S. Environmental Protection Agency
AN EXAMINATION OF EPA RISK
ASSESSMENT PRINCIPLES AND
PRACTICES
Staff Paper Prepared for the U.S. Environmental Protection Agency
by members of the Risk Assessment Task Force
Office of the Science Advisor
U.S. Environmental Protection Agency
Washington, DC 20460
/"TV Racycled/Rwyclable
Printed with vegetable-based ink on
paper that contains a minimum ol
50% post-consumer fiber content
processed chlorine free,
-------
Page ii
Examination of EPA Risk Assessment Principles and Practices
DISCLAIMERS
This document has been reviewed in accordance with United States Environmental
Protection Agency policy and approved for publication and distribution. Mention of trade names
or commercial products does not constitute endorsement or recommendation for use.
This document presents an analysis of EPA's general risk assessment practices, based on
typical historic and current practice. The document does not establish new Agency policy or
guidance or amend any existing Agency policy or guidance. Nor does the document attempt to
present binding prospective requirements, necessarily applicable to future agency actions. The
use of the words "should," "can," "would," and "may" in this document means that something is
suggested or recommended, but not required.
A particular risk assessment practice described in this document may not apply to an
individual situation based upon the circumstances. Interested parties are free to raise questions
and objections about the substance of the practices discussed in this document and the propriety
of the application of those practices to a particular situation. Any individual or site-specific risk
management decision will be based on the applicable statute and regulations, and on facts
specific to the circumstances at issue. Variance from the approaches outlined in this document
does not necessarily have any significance. EPA and other decision makers retain the discretion
to adopt approaches on a case-by-case basis that differ from those described in this document
where appropriate.
Risk assessments discussed in this staff paper reflect a "snapshot" in time and may not be
reflective of any further assessment activity past the time of a particular description. For
example, the Integrated Risk Information System (IRIS) descriptions, particularly of past
assessments, may not be reflective of the current IRIS data base, as assessments are continuously
updated. Further, it is important to note that current IRIS health assessments are conducted using
the 1999 draft cancer guidelines (as of this examination), and not the 2003 draft final cancer
guidelines.
-------
Examination of EPA Risk Assessment Principles and Practices
Page iii
TABLE OF CONTENTS
ACKNOWLEDGMENTS Page ix
AN EXAMINATION OF EPA RISK ASSESSMENT PRINCIPLES
AND PRACTICES Page 1
1. INTRODUCTION TO EPA RISK ASSESSMENT Page 2
1.1 Overview Page 2
1.1.1 What Is Risk Assessment? Page 2
1.1.2 Why Does EPA Conduct Risk Assessments? Page 2
1.1.3 How Does EPA Use Risk Assessments in Decision Making? .... Page 3
1.1.4 What Is Some Historical Perspective Relevant to EPA
Risk Assessment Practices? Page 4
1.2 EPA Process for Evaluation of Risk Assessment Principles and Practices . Page 6
1.2.1 Why Are We Conducting an Evaluation of Our Risk Assessment
Principles and Practices Now? Page 6
1.2.2 What Is the Purpose and Intent of This Staff Paper? Page 6
1.2.3 Through What Process Is EPA Conducting This Evaluation? Page 7
1.2.4 How Does the 2003 OMB Draft Report to Congress Relate to Our
Evaluation? Page 7
1.2.5 What Is the General Nature of the Comments on EPA Risk
Assessment Practices Submitted to OMB? Page 8
1.3 Organization of This Document Page 8
1.3.1 How Are the General Comments Addressed? Page 8
1.3.2 How Are the Specific Comments and Examples Addressed? Page 9
1.4 Other Components Impacting Risk Assessment, But Not Addressed
in This Document Page 9
1.4.1 What Impact Do the OMB and EPA Information Quality
Guidelines Have on EPA Risk Assessment Practices? Page 9
1.4.2 Why Is Peer Review So Important? Page 10
2. EPA RISK ASSESSMENT AND PUBLIC AND ENVIRONMENTAL HEALTH
PROTECTION Page 11
2.1 Public and Environmental Health Protection ("Public Health
Protection") Page 11
2.1.1 What Is EPA's General Approach for Developing
Risk Assessments? Page 11
2.1.2 What Is the "Conservatism" Issue in Terms of Public
Health Protection? Page 11
-------
Page iv
Examination of EPA Risk Assessment Principles and Practices
2.1.3 Is Science Policy Utilized Within the Risk Assessment
Process? Page 12
2.1.4 What Impact Does Statutory Language Have on EPA Risk
Assessment Practices for Public Health Protection? Page 13
2.1.5 Does EPA Take a Reasonable Approach to Public Health
Protection? Page 16
2.1.6 What Happens When Default Assumptions Are Combined? Page 17
2.2 General Risk Assessment Approaches Used by EPA for Public
and Environmental Health Protection ("Public Health Protection") Page 20
2.2.1 How Comprehensive Are EPA's Risk Assessments? Page 20
2.2.2 Whom Is EPA Trying To Protect? Page 22
2.2.3 Are Risk Assessment and Risk Management Separate? Page 22
2.2.4 How Do "Planning and Scoping" Help Environmental
Risk Assessment? Page 23
2.2.5 How Does EPA Use a Screening Risk Assessment? Page 24
2.2.6 What Happens if EPA Identifies a Potential Risk That Needs
To Be Addressed After a Screening Risk Assessment? Page 25
2.2.7 How Are High-End Exposures Reflected in EPA Evaluations? .. Page 26
3. UNCERTAINTY AND VARIABILITY Page 30
3.1 Overview Page 30
3.2 Uncertainty and Variability Page 30
3.2.1 What Is Uncertainty? Page 30
3.2.2 What Is Parameter Uncertainty? Page 31
3.2.3 What Is Model Uncertainty? Page 32
3.2.4 How Does Variability Differ From Uncertainty? Page 32
3.3 Characterizing Uncertainty and Variability Page 33
3.3.1 Why Is Characterizing Uncertainty and Variability Important? .. Page 33
3.3.2 When Should EPA Conduct Uncertainty Analysis? Page 34
3.3.3 What Is an Appropriate Level of Uncertainty Analysis? Page 35
3.3.4 With What Precision Should EPA Results Be Reported? Page 36
3.4 Issues in Characterizing Uncertainty and Variability Page 36
3.4.1 What Is the Issue of Deterministic Versus Probabilistic
Approaches? Page 37
3.4.2 What Is the Importance of Quantitative Characterization
of Uncertainty in Dose-Response? Page 38
3.4.3 How Does EPA Use Probabilistic Analysis? Page 40
3.4.4 How Does Research Help Reduce Uncertainty? Page 41
3.5 Inherent Variability in Biological Response Page 42
3.5.1 Why Consider Populations and Life-Stages? Page 42
-------
Examination of EPA Risk Assessment Principles and Practices
Pagev
3.5.2 Are Certain Populations and Life-Stages Always at
Greater Risk? Page 43
3.5.3 How Are Sensitivities to Toxic Effects Considered? Page 43
3.5.4 Are Unique or Disproportionate Exposures Considered? Page 45
3.5.5 Is Human Variability Considered in Occupational Cohorts? Page 47
3.5.6 What Needs To Be Done To Address the Risk Assessment
of Populations and Life-Stages? Page 48
3.6 What Is EPA Doing To Improve Assessment of Uncertainty
and Variability? Page 49
3.6.1 How Can EPA Move Forward Toward Quantitative
Characterization of Uncertainty in Dose-Response? Page 49
3.6.2 Will EPA Move Toward Integrated Uncertainty Analysis? Page 49
4. CONSIDERING INFORMATION GAPS IN HEALTH ASSESSMENTS: USE OF
DEFAULT AND EXTRAPOLATION ASSUMPTIONS Page 51
4.1 Default Assumptions Page 51
4.1.1 How Does EPA Use Default Assumptions? Page 51
4.1.2 Are Default Assumptions Science Policy? Page 52
4.1.3 Does Any Change Seen in Animals Indicate There Will Be
a Problem for Humans? Page 53
4.1.4 Are Benign (Histologically Non-Malignant) Tumors Presumed
To Have Potential To Progress to Malignant Tumors,
and Are They Counted as if They Were Malignant? Page 54
4.1.5 Should We Expect Target Organ Concordance Between
Effects Observed in Animals and Those Expected in Humans? .. Page 55
4.2 Extrapolations Page 56
4.2.1 Why Does EPA Use Experiments With Animals To Predict
Effects in Humans? Page 56
4.2.2 How Does EPA Adjust for the Differences Between Animals
and Humans? Page 57
4.2.3 Does EPA Use the Most Sensitive Animal Results To Predict
Effects in Humans? Page 57
4.2.4 Is an Additional Body Scaling Factor for Children Considered? .. Page 59
4.2.5 How Are Effects Observed at High Exposures Used To Predict
Responses to Much Lower Exposures? Page 59
4.2.6 Why Is the Maximum Tolerated Dose Used? Page 61
4.2.7 Does the Use of a Maximum Tolerated Dose Affect Cancer
Potency Estimates? Page 61
4.2.8 Are the Cancer Risks Estimated by EPA the Expected
Incidence of Cancer for a Given Exposure? Page 62
-------
Page vi
Examination of EPA Risk Assessment Principles and Practices
4.2.9 How Does EPA Evaluate Incidence Versus Mortality for
Cancer Rates? Page 63
4.2.10 If Data Are Only Available for One Route of Exposure,
Does EPA Evaluate Other Routes of Exposure? Page 64
4.3 Use of Uncertainty Factors Page 67
4.3.1 How Are NOAELs or LOAELs Used for Extrapolations? Page 67
4.3.2 What Uncertainty Factors Does EPA Use To Reduce the
Experimental NOAEL in Health Assessments? Page 68
4.3.3 Does EPA Consider the Effects of Combining Several UFs? .... Page 69
4.3.4 Does EPA Use UFs Derived for Specific Chemicals? Page 71
4.4 Weight of Evidence Page 71
4.4.1 How Does EPA Consider Weight of Evidence? Page 71
4.4.2 Does EPA Give More Weight to Positive Animal Studies
Than to Negative Animal Studies? Page 72
4.4.3 Does EPA Sometimes Weight Animal Data More Than
Human Data? Page 72
4.5 Toxicity Equivalency Factor Approach Page 83
4.5.1 What Are TEFs? Page 83
4.5.2 How Did the TEF Approach Evolve? Page 85
4.5.3 Is the Additivity Assumption for TEF Justified? Page 87
4.5.4 What Is EPA's Experience With TEFs? Page 87
4.6 Model Use Page 88
4.6.1 Why Does EPA Use Environmental Models? Page 88
4.6.2 How Does EPA Approach Environmental Models? Page 89
4.6.3 What Is EPA Doing To Improve the Use of Environmental
Models? Page 89
4.6.4 How Does EPA Approach Fate and Transport Modeling and
Account for Uncertainty? Page 92
4.6.5 Does EPA Consider or Review Existing Data on Environmental
Concentrations? Page 94
4.6.6 Does EPA Consider and Use Case-Specific Data in Models?
(Or: How Does EPA Decide To Modify Default Assumptions
in Later Tiers Using Models?) Page 95
4.6.7 How Will EPA Improve Pesticide Fate and Transport
Modeling in the Future? Page 95
4.6.8 Does EPA Use Worst-Case Assumptions in Modeling That Are
Unreasonable and Do Not Account for Degradation, Partitioning,
and "Sinks" in the Environment? Page 96
-------
Examination of EPA Risk Assessment Principles and Practices
Pagevii
5. SITE- AND CHEMICAL-SPECIFIC ASSESSMENTS Page 99
5.1 Overview Page 99
5,1.1 What Are the Superfund Program's Purpose and Regulatory
Requirements? Page 101
5.1.2 Whom Does the Superfund Program Seek To Protect? The
Reasonable Maximum Exposure Scenario Page 101
5.1.3 Is the RME Overly Conservative? Page 102
5.2 Superfund Risk Assessment Guidance Page 102
5.2.1 What Are Some Types of Risk Assessment Under Superfund? . Page 102
5.2.2 How Does Superfund Use Site-Specific Concentration and
Maximum Concentration Data in Risk Assessment? Page 103
5.2.3 Has the Superfund Guidance Been Updated and Externally
Peer Reviewed Since It Was First Released in 1989? Page 104
5.3 Default Assumptions and Site-Specific Information Page 105
5.3.1 What Is the Role of Default Assumptions and Site-Specific
Information in Superfund Risk Assessments? Page 105
5.3.2 What Is the Drinking Water Consumption Rate? Page 106
5.3.3 What Is the Inhalation Rate? Page 106
5.3.4 What Is the Exposure Duration for Residences? Page 106
5.3.5 What Is the Exposure Duration for Workers? Page 107
5.3.6 What Is the Ingestion Rate for Construction Workers? Page 108
5.3.7 What Are the Incidental Soil Ingestion Rates for Children? Page 108
5.4 Site-Specific Information Page 109
5.4.1 When Does EPA Use Site-Specific Information? Page 109
5.4.2 How Is Bioavailability Addressed? Page 110
5.4.3 How Are Fish Ingestion Rates Evaluated? Page 111
5.5 Other Factors in Superfund Assessments Page 115
5.5.1 What Is the Role of Stakeholders in Superfund Assessments? .. Page 115
5.5.2 How Are Superfund Assessments Factored in Superfund Risk
Management Decisions? Page 116
5.6 Superfund Site-Specific Conclusions Page 119
5.7 Chemical Mixtures Risk Assessment Methods and Practice Page 120
5.7.1 What Are the Issues Regarding Chemical Mixture Site
Assessments? Page 120
5.7.2 What Are Dose Addition and Response Addition? Page 121
5.7.3 How Are Dose Addition and Response Addition Applied
in Practice? Page 123
5.7.4 Are Interaction Data Used in Mixture Risk Assessment? Page 124
5.7.5 Are Data on Whole Mixtures Used in Risk Assessment? Page 124
5.7.6 What Other EPA Applications Exist for Mixtures Risk
Assessment? Page 125
-------
Page viii Examination of EPA Risk Assessment Principles and Practices
6. ECOLOGICAL ASSESSMENT Page 127
6.1 Overview Page 127
6.1.1 What Is the EPA Ecological Risk Assessment Approach? Page 128
6.2 Organism-Level Versus Population-Level Ecological
Risk Assessments Page 129
6.3 Conservatism and Ecological Risk Assessments Page 131
6.3.1 How Is Conservatism Addressed in Ecological Risk
Assessments? Page 131
6.3.2 How Are Toxicity Reference Values Developed in Ecological
Risk Assessments? Page 131
6.3.3 How Do Screening-Level and Definitive (Baseline)
Assessments Differ? Page 132
6.3.4 How Is Site-Specific Information Used To Derive Exposure
Concentrations in Ecological Risk Assessments? Page 132
6.4 Water Quality Criteria Page 133
6.4.1 What Is the Agency Standard Methodology for Deriving
Aquatic Life Water Quality Criteria? Page 133
6.4.2 What Is the GLI Tier II Methodology? Page 134
6.4.3 Was the GLI Tier II Methodology Made Available for Public
Comment or Vetted Through Other Public Processes? Page 134
6.5 Uncertainty and Pesticide Ecological Assessment Page 135
6.5.1 How Does EPA Address Uncertainty Analysis in Pesticide
Ecological Assessments? Page 135
6.5.2 How Does OPP Address Rodenticide Use Patterns? Page 136
6.6 Summary and Conclusions Page 139
7. SUMMARY AND RECOMMENDATIONS Page 141
7.1 Summary Page 141
7.2 Several Current EPA Activities That Will Enhance EPA Risk
Assessment Principles and Practices Page 142
7.3 Recommendations for Further Improvements to Risk Assessment
Principles and Practices Page 143
LIST OF USEFUL ABBREVIATIONS AND ACRONYMS Page 148
GENERAL REFERENCES Page 154
REFERENCES OF EPA RISK ASSESSMENT GUIDELINES Page 180
ADDITIONAL USEFUL WEB SITES Page 182
-------
Examination of EPA Risk Assessment Principles and Practices
Page ix
ACKNOWLEDGMENTS
Many people worked on and contributed to the examination of risk assessment principles
and practices that ultimately resulted in this document. Many EPA employees from the Agency's
Offices and Regions provided input and we would like to specifically acknowledge the efforts
and contributions made by the following individuals:
Daniel Axelrad (OPEI)
Timothy Benner (ORD)
Michael Beringer (Region 7)
George Bollweg (Region 5)
David Bussard (ORD)
Michael Callahan (Region 6)
Catherine Campbell (OPEI)
Laurel Celeste (OGC)
Weihsueh Chiu (ORD)
Patricia Cirone (Region 10)
Jenny Craig (OAR)
Larry Cupitt (ORD)
Julie Damon (ASPH/EPA Fellow)
Kerry Dearfield (OSA) - Chair, Task Force
Vicki Dellarco (OPP)
Lynn Flowers (ORD)
BrendaFoos(OCHP)
Stiven Foster (ORD)
Gary Foureman (ORD)
Jack Fowle (ORD)
Herman Gibb (ORD)
Andrew Gillespie (detail from Dept of
Agriculture)
David Guinnup (OAR)
Karen Hammerstrom (ORD)
Belinda Hawkins (OEI)
Tala Henry (OW)
Oscar Hernandez (OPPT)
Richard Hertzberg (ORD)
Robert Hetes (ORD) - Chapter 2 lead
James Hetrick (OPP)
Ann Johnson (OPEI)
Myra Karstadt (OPPT)
Aparna Koppikar (ORD)
Michael Kravitz (ORD)
Stephen Kroner (OSWER)
Patricia Lafornara (OEI)
Elizabeth Leovey (OPP)
John Lipscomb (ORD)
Mary Manibusan (ORD)
Carl Mazza (OAR)
Robert McGaughy (ORD)
Michael Metzger (OPP)
Jayne Michaud (OSWER) - Chapter 5 lead
Amy Mills (ORD)
Kenneth Mitchell (OAR)
Jacqueline Moya (ORD)
Debbie Newberry (OSWER)
Susan Norton (ORD)
Marian Olsen (Region 2)
Jennifer Orme-Zavaleta (ORD)
Nicole Paquette (OEI)
Pasky Pascual (ORD)
Resha Putzrath (ORD) - Chapter 4 lead
MaryReiley(OW)
Susan Reith (ORD)
Glenn Rice (ORD)
Charles Ris (ORD)
Bruce Rodan (ORD)
John Schaum (ORD)
RitaSchoeny(OW)
Brad Schultz (Region 5) - Chapter 3 lead
Cheryl Scott (ORD)
Suhair Shallal (SAB)
Michael Slimak (ORD) - Chapter 6 lead
Daniel Stralka (Region 9)
-------
Pagex
Examination of EPA Risk Assessment Principles and Practices
Glenn Suter (ORD)
Jeffrey Swartout (ORD)
Christina Swartz (OPP)
Linda Teuschler (ORD)
Nelson Thurman (OPP)
Douglas Urban (OPP)
Winona Victory (Region 9)
Donn Viviani (OPEI)
Katie Warwick (ORD)
-------
Examination of EPA Risk Assessment Principles and Practices Page 1
U.S. Environmental Protection Agency
AN EXAMINATION OF EPA RISK ASSESSMENT
PRINCIPLES AND PRACTICES
-------
Page 2 Examination of EPA Risk Assessment Principles and Practices
1. INTRODUCTION TO EPA RISK ASSESSMENT
1.1 Overview
1.1.1 What Is Risk Assessment?
The most common basic definition of risk assessment used within the U.S.
Environmental Protection Agency (EPA) is paraphrased from the 1983 report Risk Assessment in
the Federal Government: Managing the Process (NRC, 1983), by the National Academy of
Sciences' (NAS's) National Research Council (NRC):
Risk assessment is a process in which information is analyzed to determine if an
environmental hazard might cause harm to exposed persons and ecosystems.
This process is highly interdisciplinary in that it draws from such diverse fields as
biology, toxicology, ecology, engineering, geology, statistics, and the social sciences to create a
rational framework for evaluating environmental hazards. While this definition has been
somewhat enhanced and elaborated upon through subsequent NAS writings, it still basically
describes risk assessment as it is performed within EPA. EPA uses risk assessment as a tool to
integrate exposure and health effects or ecological effects information into a characterization of
the potential for health hazards in humans or other hazards to our environment.
1.1.2 Why Does EPA Conduct Risk Assessments?
The mission of the EPA is to protect human health and to safeguard the natural
environment air, water, and land upon which life depends. EPA fulfills this mission by,
among other things, developing and enforcing regulations that implement environmental laws
enacted by Congress. The implementation of environmental laws may include grants and other
financial assistance to state and tribal governments carrying out environmental programs
approved, authorized, or delegated by EPA.
Determining environmental standards, policies, guidelines, regulations, and actions
requires making decisions. Environmental decision making is often a controversial process
involving the interplay among many forces: science, social and economic factors, political
considerations, technological feasibility, and statutory requirements. There are often conflicting
interests regarding these various forces than can have a bearing on environmental decisions.
Setting an environmental standard that is too lax may threaten public health, while a standard that
is unnecessarily stringent may impose a significant marginal economic cost for small marginal
gain. Environmental decisions are often time-sensitive, for example when public health is
known or suspected to be at risk. The decisions must frequently be made with incomplete or
imperfect information and many times under the additional pressure of heightened public
-------
Examination of EPA Risk Assessment Principles and Practices Page 3
scrutiny and concern. And, once made, the decisions are often challenged in court and subject to
high levels of public and scientific scrutiny.
EPA conducts risk assessment to provide the best possible scientific characterization of
risks based on a rigorous analysis of available information and knowledge that is, a
description of the nature and magnitude of the risk, an interpretation of the adversity of the risk, a
summary of the confidence or reliability of the information available to describe the risk, areas
where information is uncertain or lacking completely, and documentation of all of the evidence
supporting the characterization of the risk. EPA then incorporates this risk characterization with
all of the other relevant information social, economic, political, and regulatory in making
decisions (policies, regulations) about how to manage the risk. Risk assessment, therefore,
informs decision makers about the science implications of the risk in question. Risk assessments
that meet their objectives can help guide risk managers to decisions that mitigate environmental
risks at the lowest possible cost and which will stand up if challenged in the courts.
1.1.3 How Does EPA Use Risk Assessments in Decision Making?
The primary purpose of a risk assessment is to inform the risk manager's decision making
process. The primary purpose of a risk assessment is not to make or recommend any particular
decisions; rather, it gives the risk manager information to consider along with other pertinent
information. EPA uses risk assessment as a key source of scientific information for making
good, sound decisions about managing risks to human health and the environment. Examples of
such decisions include deciding permissible release levels of toxic chemicals, granting permits
for hazardous waste treatment operations, and selecting methods for remediating Superfund sites.
The use of credible science in risk assessment helps make and support risk management
decisions, but it is not the only factor that the risk manager considers. It is generally recognized
by the science community, by the regulatory community, and by the courts that it is
important to consider other factors along with the science when making decisions about risk
management. In some regulations, the consideration of other factors is mandated (e.g., costs).
Some of these other factors include:
a. Economic factors the costs and benefits of risks and risk mitigation
alternatives.
b. Laws and legal decisions the framework that prohibits or requires some
actions.
c. Social factors attributes of individuals or populations that may affect their
susceptibility to risks from a particular stressor.
-------
Page 4 Examination of EPA Risk Assessment Principles and Practices
d. Technological factors the feasibility, impact, and range of risk management
options
e. Political factors interactions among and between different branches and levels
of government and the citizens they represent.
f. Public factors the attitudes and values of individuals and societies with respect
to environmental quality, environmental risk, and risk management.
1.1.4 What Is Some Historical Perspective Relevant to EPA Risk Assessment
Practices?
EPA was involved with risk assessment practices since EPA's early days, although risk
assessment per se was not a formally recognized process then. EPA completed its first risk
assessment document in December 1975: Quantitative Risk Assessment for Community Exposure
to Vinyl Chloride (Kuzmack and McGaughy, 1975). The next significant document appeared in
1976: Interim Procedures and Guidelines for Health Risk and Economic Impact Assessments of
Suspected Carcinogens (Train, 1976). The preamble of this document, signed by the
Administrator, signaled the Agency's intent that "rigorous assessments of health risk and
economic impact will be undertaken as part of the regulatory process." A general framework
described a process to be followed in analyzing cancer risks of pesticides, and the document
recommended that the health data be analyzed independently of the economic impact analysis.
Later, in 1980, EPA announced the availability of water quality criteria documents for 64
contaminants (USEPA, 1980). This was the first application of quantitative procedures
developed by EPA to a large number of carcinogens, and the first EPA document describing
quantitative procedures used in risk assessment.
Then in 1983, the NAS published Risk Assessment in the Federal Government: Managing
the Process (NRC, 1983; commonly referred to as the "Red Book"). EPA has integrated the
principles of risk assessment from this groundbreaking report into its practices to this day. The
following year, EPA published Risk Assessment and Management: Framework for Decision
Making (USEPA, 1984), which emphasizes making the risk assessment process transparent,
describing the assessment's strengths and weaknesses more fully, and providing plausible
alternatives within the assessment.
Shortly after the publication of the Red Book, EPA began issuing a series of guidelines
for conducting risk assessments (e.g., in 1986 for cancer, mutagenicity, chemical mixtures,
developmental toxicology, and in 1992 for estimating exposures). Although EPA efforts focused
initially on human health risk assessment, the basic model was adapted to ecological risk
assessment in the 1990s to deal with a broad array of environmental risk assessments in which
human health impacts are not directly at issue. EPA continues to make a substantial investment
-------
Examination of EPA Risk Assessment Principles and Practices Page 5
in advancing the science and application of risk assessment through continual updates of these
guidelines and the development of newer guidelines as needed. Refer to the section in the
References set aside for the listing of these EPA guidelines.
Over time, the NAS expanded on its risk assessment principles in a series of subsequent
reports, including Pesticides in the Diets of Infants and Children (NRC, 1993), Science and
Judgment in Risk Assessment (NRC, 1994; also known as the "Blue Book"), and Understanding
Risk: Informing Decisions in a Democratic Society (NRC, 1996). For example, the NAS places
equal emphasis on fully characterizing the scope, uncertainties, limitations, and strengths of the
assessment and on the social dimensions of interacting with decision makers and other users of
the assessment in an iterative, analytic-deliberative process. The purpose of this process is to
ensure that the assessments meet the intended objectives and are understandable. EPA risk
assessment practices have evolved over time along with this progression of thought, and in many
cases helped drive the evolution of thinking on risk assessment.
In 1995, EPA updated and issued the current Agency-wide Risk Characterization Policy
(USEPA, 1995a). The Policy calls for all risk assessments performed at EPA to include a risk
characterization to ensure that the risk assessment process is transparent; it also emphasizes that
risk assessments be clear, reasonable, and consistent with other risk assessments of similar scope
prepared by programs across the Agency. Effective risk characterization is achieved through
transparency in the risk assessment process and clarity, consistency, and reasonableness of the
risk assessment product TCCR. EPA's Risk Characterization Handbook (USEPA, 2000a)
was developed to implement the Risk Characterization Policy.
The Congressional/Presidential Commission on Risk Assessment and Risk Management
(CRARM) was created by the Clean Air Act Amendments of 1990 and formed in 1994. Its
mandate was to make a full investigation of the policy implications and appropriate uses of risk
assessment and risk management in regulatory programs, under various federal laws, designed to
prevent cancer and other chronic health effects that may result from exposure to hazardous
substances. More specifically, its mandate was to provide guidance on how to deal with residual
emissions from Section 112 hazardous air pollutants (HAPs) after technology-based controls
have been placed on stationary sources of air pollutants. In 1997, the Commission published its
report in two volumes (CRARM, 1997a; CRARM, 1997b). These discussed the importance of
better understanding and quantification of risks, as well as the importance of evaluating strategies
to reduce human and ecological risks.
EPA's risk assessment principles and practices build on our own risk assessment
guidances and policies such as the Risk Characterization Policy; Guidance for Cumulative
Assessment, Part 1: Planning and Scoping (USEPA, 1997a); the Risk Assessment Guidance for
Superfund, or RAGS (USEPA, 1989a, and subsequent updates); EPA's Information Quality
Guidelines (USEPA, 2002a); and^ Summary of General Assessment Factors for Evaluating the
-------
Page 6 Examination of EPA Risk Assessment Principles and Practices
Quality of Scientific and Technical Information (USEPA, 2003a) as well as the NAS, the
CRARM, and others' concepts. It is understood that risk assessment provides important
information about the nature, magnitude, and likelihood of possible environmental risks to
inform decisions principles that evolved out of these many efforts.
1.2 EPA Process for Evaluation of Risk Assessment Principles and Practices
1.2.1 Why Are We Conducting an Evaluation of Our Risk Assessment Principles
and Practices Now?
EPA constantly evaluates its risk assessment principles and practices, mostly via a
gradual refinement of particular practices that may not be overtly visible to the public. There are
times when EPA takes a concentrated, focused approach: for example, when revising a major
risk assessment guideline such as for cancer assessment. EPA conducts a wider, general review
of its risk assessment principles and practices occasionally to help strengthen core values and
increase its ability to make better decisions.
In early 2002, the position of the EPA Science Advisor was established. The Science
Advisor's overarching responsibility is to coordinate and oversee the scientific activities of the
program and regional offices at EPA. Part of this responsibility is to ensure the best use of
science at the Agency and in its decisions. At the Science Advisor's request, EPA staff began
looking at the Agency's risk assessment practices and training with an eye to update them. When
the Office of Management and Budget (OMB) solicited comments on risk assessment practices
across the federal government (see section 1.2.4), we (EPA staff) took this as an opportunity to
concentrate on a wider review to evaluate current risk assessment practices across programs and
regions.
1.2.2 What Is the Purpose and Intent of This Staff Paper?
This staff paper was developed to give the EPA scientific and technical professional staff
an opportunity to present what we (EPA staff) believe are the current EPA risk assessment
principles and practices. The practices are presented in the context of the public comments
submitted to OMB. The paper's purpose is first to open a dialogue among EPA risk assessors
and risk managers about Agency risk assessment practices. Then, as we engage the public, we
will continue the dialogue about how we can move forward together to clarify and, where
appropriate, strengthen our risk assessment practices.
The staff paper is intended for a wide audience of people who are very familiar with risk
assessment principles and practices risk assessors and risk managers within the EPA as well as
those outside EPA with knowledge of risk assessment practices. The discussion contained here
is not meant to be a primer on risk assessment or an introduction for those not very familiar with
-------
Examination of EPA Risk Assessment Principles and Practices Page 7
these practices. Since the comments address complex and sometimes subtle nuances of risk
assessment, the staff paper attempts to deal with those comments and concepts at that level of
discussion.
1.2.3 Through What Process Is EPA Conducting This Evaluation?
Specifically for this evaluation, on June 17, 2003, an Agency-wide memorandum was
issued from three EPA senior managers (Jessica Furey, Associate Administrator for OPEI; Paul
Oilman, EPA Science Advisor and Assistant Administrator for ORD; and Stephen Johnson,
Assistant Administrator for OPPTS) to start this evaluation. The memorandum called for the
establishment of an Agency-wide workgroup, the Risk Assessment Task Force, to review risk
assessment principles and practices at the Agency. The Risk Assessment Task Force focused on
the practices, assumptions, defaults and principles identified in the comments sent to OMB (see
section 1.2.3 below), as well as issues identified from within EPA. Generation of this report on
the initial analyses of the Task Force (EPA staff) is the first step in a multi-step process. In the
future, EPA expects to communicate with stakeholders about the results of this staff evaluation
and give them opportunities for dialogue, in order to understand their concerns.
1.2.4 How Does the 2003 OMB Draft Report to Congress Relate to Our
Evaluation?
On February 3, 2003, OMB published in the Federal Register (FR) a request for
comments on its Draft 2003 Report to Congress on the Costs and Benefits of Federal
Regulations (USOMB, 2003a). In this FR notice, OMB also sought the public's views on a
number of important issues pertaining to the practice of risk assessment. OMB received many
public comments and these were passed onto EPA. Our effort takes advantage of this
information as we review risk assessment principles and practices at the Agency.
OMB issued its final report to Congress on regulatory policy in September 2003
(USOMB, 2003b) based on the comments it received on the February draft report. The final
report presents findings on major federal rulemakings finalized over the previous 10 years,
specific regulatory reforms, guidance on regulatory analysis, homeland security proposals, and
Agency consultations with states and local governments. The report also deals with the concept
of precaution in U.S. approaches to risk assessment and management. OMB concludes that
precaution plays an important role in risk assessment and risk management, but precaution,
coupled with objective scientific analysis, needs to be applied wisely on a case-by-case basis.
-------
PageS
Examination of EPA Risk Assessment Principles and Practices
1.2.5 What Is the General Nature of the Comments on EPA Risk Assessment
Practices Submitted to OMB?
The vast majority of the comments can be generally summarized from the American
Chemistry Council (ACC) submission. ACC's basic recommendations are not radical or
particularly new. They reiterate three points that can be found in the 1991 Executive Office of
the President document Regulatory Program of the United States Government (EOP, 1991):
a) Risk assessments should not continue an unwarranted reliance on "conservative
(worst-case) assumptions" that distort the outcomes of the risk assessment,
"yielding estimates that may overstate likely risks by several orders of
magnitude."
b) Risk assessments should "acknowledge the presence of considerable uncertainty"
and present the extent to which conservative assumptions may overstate likely
risks.
c) EPA risk assessments must not "intermingle important policy judgments within
the scientific assessment of risk." Rather, the "choice of an appropriate margin of
safety should remain the province of responsible risk-management officials, and
should not be preempted through biased risk assessments."
On the other hand, various other comments submitted suggest that EPA risk assessments
do not fully address all risks. Generally, these latter comments relate to issues of cumulative and
aggregate risk. They state that EPA risk assessments concentrate on single chemical/stressor
risks, failing to account for multiple chemical/stressor exposures and other factors as life-stage,
lifestyle, and increased susceptibility of certain exposed populations. This practice would tend to
underestimate risk in real-world scenarios.
1.3 Organization of This Document
1.3.1 How Are the General Comments Addressed?
This document examines EPA risk assessment principles and practices in light of the
general comments outlined in the preceding section. The comments focus on issues of
conservatism in risk assessment (e.g., overstated/understated risks), use of rigid default
assumptions, poor transparency in the risk assessments we produce, and unacknowledged
uncertainty. The following chapters are designed to address these general comments (issues).
Chapter 2 deals with the issue of conservatism. Chapter 3 discusses the nature of uncertainty and
variability and how EPA deals with these in risk assessment. Chapter 4, on defaults and
extrapolations, discusses how assumptions can be used when chemical- and/or site-specific data
-------
Examination of EPA Risk Assessment Principles and Practices Page 9
are not available or are inadequate to use in performing a risk assessment. The next chapter,
chapter 5, approaches the use of site- and chemical-specific data. This chapter emphasizes the
Superfund approaches many of the specific comments we received were specific to
Superfund.
Most of the comments EPA received focused on human health assessment. Many of the
principles and practices, though, apply to ecological health assessments as well. Chapter 6
briefly discusses and evaluates our ecological health assessments.
The last chapter, chapter 7, provides a summary of the Risk Assessment Task Force
conclusions and recommendations, with transparency in our risk assessment practices as a focus.
1.3.2 How Are the Specific Comments and Examples Addressed?
This document is organized to address the overarching issues of conservatism,
uncertainty, and transparency. However, many comments focused on specific risk assessment
principles and practices as well as specific chemical assessments as examples. We will not
discuss every specific comment in a "one to one" discussion, although there are some instances
in the following chapters for which specific examples are provided to illustrate a point. While
we will mention specific chemicals and specific rulemakings in the context of the general issues
discussion, we will not discuss individual chemical- or site-specific decisions, as these are
generally of high importance to many, are complex, and (in many cases) are still in active
interaction with interested stakeholders.
1.4 Other Components Impacting Risk Assessment, But Not Addressed in This
Document
1.4.1 What Impact Do the OMB and EPA Information Quality Guidelines Have
on EPA Risk Assessment Practices?
The OMB guidelines state that information needs to be "objective, realistic, and
scientifically balanced" (USOMB, 2002). EPA embraced the OMB guidance when developing
its Information Quality Guidelines (USEPA, 2002a). These guidelines point to our reliance on
the extensive use of peer review, the practices found in the various risk assessment guidelines
and the Risk Characterization Policy, and the Agency-wide use of EPA's Quality System. These
EPA practices help ensure that information we use is objective, realistic, and scientifically
balanced. As we stress, EPA uses scientific peer review to help ensure the quality of the risk
assessments we generate and to keep the assessments as objective, and as consistent, as possible.
EPA's Peer Review Handbook (USEPA, 2000b) stresses that peer review panels be balanced in
terms of their expertise and biases so that a reasonable and scientifically balanced review results.
Further, conflicts of interest need to be readily identified. The Risk Characterization Policy
-------
Page 10 Examination of EPA Risk Assessment Principles and Practices
encourages presentation of all plausible viewpoints that are realistic and scientifically
supportable. In 1984, under EPA Order 5360.1, an Agency-wide Quality System was
established. In 2000, Agency-wide policies were defined in EPA Order 5360.1 A2 and EPA
Manual 5360 Al, which expanded the policy to accommodate current and evolving needs of the
Agency. EPA's Quality System helps ensure that we stay objective in the use of information in
any assessment we generate (USEPA, 2000c, 2000d). In association with the Information
Quality Guidelines, we also developed general assessment factors related to quality issues
regarding scientific and technical information (USEPA, 2003a). One should take these
assessment factors into consideration as appropriate when evaluating the quality and relevance of
information, regardless of source.
Many parties believe that EPA risk assessment practices do not follow this guidance. We
believe EPA conducts risk assessments consistent with this guidance, as evidenced by some of
the efforts detailed in the paragraph above. We continually strive to publicly present our risk
assessment practices and guidance documents and subject them to peer review. Our Risk
Characterization Policy directs us to consider all scientifically plausible and supportable
viewpoints, but this information is only part of the full range of information risk managers use to
make a decision.
1.4.2 Why Is Peer Review So Important?
The value of scientific peer review in ensuring the quality of our scientific and technical
products is critical to EPA and is widely understood and accepted across the Agency.
Conscientious use of peer review is essential to the credibility of the risk assessments EPA uses
to support its decisions. Consistent Agency-wide application of peer review has been an EPA
priority for many years and continues to be so. Since issuing the peer review policy in 1993,
EPA has continually supported and strengthened the policy and practices of peer review, which
in turn have supported and strengthened our risk assessment principles and practices. For
example, peer review gave us an opportunity to seek expert review on many of the issues
highlighted in this document, such as the reasonableness of assumptions and methodologies used
in Agency risk assessments.
A peer review is a documented critical review of a specific scientific or technical work
product, conducted by qualified individuals (or organizations) who are independent of those who
performed the work (at a minimum, from a different office) but who are collectively equivalent
in technical expertise (i.e., peers). EPA's Peer Review Handbook (2nd Edition; USEPA 2000b) is
one of the most advanced treatments of peer review for intramural research and
scientific/technical analysis of any federal agency. By utilizing the practices embodied in the
Handbook, EPA ensures the credibility and quality of the risk assessments we generate to support
Agency decisions.
-------
Examination of EPA Risk Assessment Principles and Practices Page 11
2. EPA RISK ASSESSMENT AND PUBLIC AND
ENVIRONMENTAL HEALTH PROTECTION
2.1 Public and Environmental Health Protection ("Public Health Protection")
EPA's risk assessments are conducted in support of its mission to protect public health
and the environment. Given the uncertainty, variability, and data gaps encountered when
conducting any risk assessment, a key objective for EPA's risk assessments is that they avoid
both underestimation of risk and gross overestimation of risk.
2.1.1 What Is EPA's General Approach for Developing Risk Assessments?
Over the years, practices have been developed to assess risk based on available data and
information, and EPA has been at the forefront of much of this development. Risk assessment is
a complex process, requiring the integration of data and information across a broad range of
activities and disciplines, including source characterization, fate and transport, modeling,
exposure assessment, and dose-response assessment EPA seeks to use the available information
(data) in an objective, realistic, and scientifically balanced way. In each specific assessment, the
Agency incorporates the relevant data and information to the extent possible (e.g., see chapter 5).
Where relevant chemical- and/or site-specific data are not available, EPA uses specific default
assumptions and extrapolations to fill in the data gaps and allow the risk assessment to proceed
(see chapter 4) so that the Agency can ultimately make the decisions required under its mandates.
This approach is consistent with the NRC's recommendation about EPA's use of defaults (NRC,
1994). In general, EPA's default assumptions are based on peer reviewed studies, empirical
observations, extrapolation from related observations, or scientific theory.
2.1.2 What Is the "Conservatism" Issue in Terms of Public Health Protection?
Because of data gaps, as well as uncertainty and variability in the available data, risk
cannot be known or calculated with absolute certainty. Further, as Hill (1965) noted, a lack of
certainty or perfect evidence "does not confer upon us a freedom to ignore the knowledge we
already have, or to postpone the action that it appears to demand at a given time." Therefore,
consistent with its mission, EPA risk assessments tend towards protecting public and
environmental health by preferring an approach that does not underestimate risk in the face of
uncertainty and variability. In other words, EPA seeks to adequately protect public and
environmental health by ensuring that risk is not likely to be underestimated. However, because
there are many views on what "adequate" protection is, some may consider the risk assessment
that supports a particular protection level to be "too conservative" (i.e., it overestimates risk),
while others may feel it is "not conservative enough" (i.e., it underestimates risk). This issue
regarding the appropriate degree of "conservatism" in EPA's risk assessments has been a concern
-------
Page 12 Examination of EPA Risk Assessment Principles and Practices
from the inception of the formal risk assessment process and has been a major part of the
discussion and comments surrounding risk assessment.
Even with an optimal cost-benefit solution, in a heterogeneous society, some members of
the population will bear a disproportionate fraction of the costs while others will enjoy a
disproportionate fraction of the benefits (Pacala et al, 2003). Thus, inevitably, different
segments of our society will view EPA's approach to public health and environmental protection
with different perspectives.
The NRC in its 1994 report (NRC, 1994) examined the conservatism issue, particularly as
it relates to the default assumptions EPA uses. The Committee did not reach consensus on the
degree to which "conservatism" should play a role in defining defaults. This issue was heavily
debated to the point where divergent views were offered, one advocating the principle of
"plausible conservatism" and the other the maximum use of scientific information in selecting
defaults. In general, EPA uses defaults that guard against underestimating risk while also being
scientifically plausible given existing uncertainty. EPA's use of various default assumptions is
the basis for many of the differences of opinion about its risk assessment practices. This
concern, identified by the NRC in 1994, continues today as indicated by many of the comments
submitted to OMB.
The question of conservatism is heightened by the ambiguous definitions and uses of the
term "conservatism" by the various concerned parties, including those that feel EPA
overestimates risk and those that feel the Agency systematically understates risks. Some of the
various concepts associated with term "conservatism" include prudence versus misestimation,
conservatism as a response to uncertainty or variability, "level of conservatism," and "amount of
conservatism" (NRC, 1994).
2.1.3 Is Science Policy Utilized Within the Risk Assessment Process?
Science policy positions and choices are by necessity utilized during the risk assessment
process. Two major ways in which the risk assessment process uses science policy are described
below in this section. Note that the utilization of science policy in the risk assessment process is
not meant to "bury" or "hide" risk management decisions within the risk assessment. The use of
any policy choice needs to be transparent in a risk assessment. In addition, although science
policy is utilized in the risk assessment process, it is important to recognize that the policy
positions themselves are developed outside the risk assessment. These policy positions are
usually supported by scientific data and/or consensus and ensure that the risk assessment
proceeds in a way that best serves the needs of the decision maker and the Agency.
First, there are some basic, fundamental policy positions that frame the risk assessment
process to ensure that the risk assessments produced are appropriate for a particular decision.
-------
Examination of EPA Risk Assessment Principles and Practices Page 13
For example, since EPA is a health and environmental protective agency, EPA's policy is that
risk assessments should not knowingly underestimate or grossly overestimate risks. This policy
position prompts risk assessments to take a more "protective" stance given the underlying
uncertainty with the risk estimates generated. Another framing policy position is that EPA will
examine and report on the upper end of a range of risks or exposures when we are not very
certain about where the particular risk lies. For example, in a screening-level risk assessment for
hazardous air pollutants (HAPs), the risk assessment starts with a 70-year exposure to assess an
individual most exposed (see section 2.2.7 for more discussion on this specific example).
Further, when several parameters are assessed, upper-end values and/or central tendency values
are generally combined to generate a risk estimate that falls within the higher end of the
population risk range. Currently, the use of the upper part of a range pertains more often to the
exposure component of the risk assessment than the hazard/dose-response portion. Many
comments to EPA suggest that the combining of upper ends leads to unreasonable estimates of
risk. We generally believe otherwise (e.g., see section 2.1.6), and we feel that this practice
should be explained clearly in the risk characterization to ensure that risk managers can make an
appropriate decision consistent with the Agency's goal of public health and environmental
protection. These policy positions not only shape the risk assessment process, but are also a
factor in the decision making process outside the risk assessment.
Second, default assumptions utilized in any given risk assessment entail science policy
positions or choices. These science policy choices are more specific than the framing science
policies, but generally are consistent with the framing policies. For example, a change that is
considered adverse (i.e., associated with toxicity) in an animal study is assumed to indicate a
problem for humans unless data demonstrate otherwise. As discussed more fully in chapter 4
(particularly sections 4.1.1. and 4.1.2), default assumptions are generally supported by scientific
data and/or scientific consensus. Their use in risk assessments is to allow the risk assessment to
proceed when chemical- and/or site-specific data are missing or not useful.
Most importantly, any science policy position or choice used in the risk assessment
process does not direct the risk assessment itself toward a specific risk management decision,
e.g., the use of a specific risk estimate. Rather, the risk assessment informs the decision maker
about the potential risks and uncertainties around the risk estimate(s). These characterized risks
are then considered in light of the other factors before a decision is made (see section 1.1.3);
science policy is one of the factors that help a risk manager make a decision.
2.1,4 What Impact Does Statutory Language Have on EPA Risk Assessment
Practices for Public Health Protection?
The discretionary power afforded to agencies in making regulatory decisions varies
greatly (OSTP, 1995). EPA faces regulatory, licensing, and other decisions covering a wide
range of environmental issues and pollutants. These decisions are made within a number of EPA
-------
Page 14
Examination of EPA Risk Assessment Principles and Practices
program offices, each responding to a unique mixture of statutes, precedents, and stakeholders.
Congress establishes legal requirements that generally describe the level of protectiveness that
EPA regulations must achieve and, infrequently, Congress imposes specific risk assessment
requirements. In addition, court precedents can affect how EPA considers assessments of risk.
Given this apparent variety of goals, the constraints within which those goals must be reached,
and the discretionary powers afforded, it is to be expected that decisions made in one EPA
program office, for one particular environmental issue, may appear at least on the surface to be
based on different risk considerations and/or public health protection goals from those used in
other EPA program offices and for other environmental issues. For example, there may appear to
be a variation from one EPA risk management decision to another in the way in that public health
protection goals have been balanced against other considerations such as technological
feasibility, precedent, and cost. However, overall, EPA generally interprets its statutes to be
protective of public health and the environment.
Apparent inconsistencies in risk assessment practices across EPA can stem from
differences in statutory language. For example, individual statutes identify varying risks to
evaluate and to protect against (e.g., establish a margin of safety; protect sensitive resources;
reduce overall risks) and mandate different levels of protection (e.g., protect public welfare;
prevent unreasonable risk; reduce overall risks; function without adverse effects). Examples
among major EPA program offices illustrate some of the different Congressional mandates
regarding risk assessment and risk management practices:
a) In the case of threshold effects ... an additional ten-fold margin of safety for the
pesticide chemical residue shall be applied for infants and children ... (OPPTS;
FFDCA §408 (b)(2)(C))
b) The Administrator shall, in a document made available to the public in support of
a regulation promulgated under this section, specify, to the extent practicable:
1) Each population addressed by any estimate of public health effects;
2) The expected risk or central estimate of risk for the specific populations;
3) Each appropriate upper-bound or lower-bound estimate of risk ... (OW;
SDWA § 300g-l (b)(3))
c) The Administrator shall... [add] pollutants which present, or may present,
through inhalation or other routes of exposure, a threat of adverse human health
effects...or adverse environmental effects through ambient concentrations,
bioaccumulation, deposition, or otherwise but not including releases subject to
-------
Examination of EPA Risk Assessment Principles and Practices
PagelS
regulation under subsection (r) of this section as a result of emissions to air...
(OAR;CAA§112(b)(2))
d) ... Provide an ample margin of safety to protect public health or to prevent an
adverse environmental effect (OAR; CAA §112(f)).
Similarly, individual statutory requirements regarding the appropriate level of protection
can have a significant impact on the focus (the purpose and scope) of a risk assessment, which
can lead to the appearance of inconsistency in risk assessment practices. Such requirements vary
across Agency programs; for example:
a) ... To assure chemical substances and mixtures do not present an unreasonable
risk of injury to health or the environment (OPPTS; TSCA §2(b)(3)).
b) ... Function without unreasonable and adverse effects on human health and the
environment (OPPTS; FIFRA §3).
c) ... Necessary to protect human health and the environment (OSWER; RCRA
§3005 as amended).
d) ... Provide the basis for the development of protective exposure levels (OSWER;
NCP §300.430(d)).
e) ... Adequate to protect public health and the environment from any reasonably
anticipated adverse effects (OW; CWA §405(d)(2)(D)).
Even the statutory language used for different statutes administered within one major office,
EPA's Office of Air and Radiation (OAR), shows differences:
a) ... Protect public health with an adequate margin of safety (OAR; CAA § 109).
b) ... Provide an ample margin of safety to protect public health or to prevent an
adverse environmental effect (OAR; CAA §112(f)).
c) ... Protect the public welfare from any known or anticipated adverse effects
(OAR; CAA §109).
d) ... [Not] cause or contribute to an unreasonable risk to public health, welfare, or
safety (OAR; CAA §202(a)(4)).
-------
Page 16
Examination of EPA Risk Assessment Principles and Practices
e) ... Protect sensitive and critically sensitive aquatic and terrestrial resources (OAR;
CAAA §404 (Appendix B)).
f) ... Reduce overall risks to human health and the environment (OAR; Title VI of
CAA).
g) ... Actions to mitigate environmental and health risks (OAR; SARA Title IV).
For example, within OAR's National Ambient Air Quality Standards Program, the
primary standards are health-based and do not consider costs or technological feasibility. The
secondary standards consider the impacts of pollutants on human economic well-being such as
visibility, agricultural productivity, and ecological impacts. Under the Clean Water Act, EPA's
Office of Water (OW) publishes ambient water quality criteria based on protecting human health;
these risk assessments do not consider the cost or technological feasibility of meeting these
criteria. Under the Safe Drinking Water Act, OW conducts risk assessments to determine non-
enforceable Maximum contaminant level goals (MCLGs). OW then sets enforceable Maximum
contaminant levels (MCLs) as close as technically feasible to the MCLGs after taking costs into
consideration.
In its report on chemical risk assessment, the General Accounting Office, or GAO,
(USGAO, 2001) recognized how contextual differences in statutes may affect the focus of risk
assessments. By taking an overall public health protective stance (as a response to uncertainty
and variability), EPA's approach to risk assessment takes into account the variety of language
found in the various statutes and allows some overall consistency in the Agency's risk
assessment practices, while allowing specific EPA offices to follow their particular mandates.
2,1.5 Does EPA Take a Reasonable Approach to Public Health Protection?
Since uncertainty and variability are present in risk assessments, EPA usually
incorporates a "high-end" hazard and/or exposure level in order to ensure an adequate margin of
safety for most of the potentially exposed, susceptible population, or ecosystem. EPA's high-end
levels are around 90% and above a reasonable approach that is consistent with the NRC
discussion (NRC, 1994). This policy choice is consistent with EPA's legislative mandates (e.g.,
adequate margin of safety, see section 2.1.4). Even with a high-end value, there will be exposed
people or environments at greater risk and at lower risk. In addition to the high-end values, EPA
programs typically estimate central tendency values for risk managers to evaluate. This provides
a reasonable sense of the range of risk that usually lies on the actual distribution.
When EPA has usable data which show that protection at a higher level (e.g., 99%) or
lower level (e.g., 90% or lower) is appropriate, then managers may use the information to make
the risk management decision. For example, exposure data used for risk assessment are
-------
Examination of EPA Risk Assessment Principles and Practices Page 17
frequently based on studies, such as controlled field trials for pesticides, that reflect high-end
exposures. Usually, the relationship between these levels and the actual levels to which people
are exposed is not known. This leads to uncertainty in estimates about the percentile of the
population exposed in a given assessment. Therefore, the risk management decision regarding
the percentile at which to regulate needs to consider how close the exposure data are likely to
approximate actual exposures in the population being considered. Typically, percentiles chosen
for pesticides regulation can range from possibly the 70th or 80th percentile in situations (e.g.,
occupational) where it is known that use of the available data will clearly overestimate exposure
to the 99.9th percentile when the data are likely to closely approximate the population exposures
(e.g., specific dietary exposures).
Whatever hazard and exposure values are used to estimate risk, it is important to be
transparent in characterizing the range of possible risks. For example, if we identify certain
populations that will be at greater risk than the high end we propose, then it is important to
highlight these populations so decision makers can make appropriate decisions regarding their
possible risk. The risk assessment should not make that decision; it should characterize that risk
and who or how much is at risk. Consequently, managers may decide on a greater level of action
in a certain locale, if conditions warrant, than may be appropriate elsewhere. On the other hand,
risk assessments should also characterize the range of risk, including the lower levels of risk.
EPA's policy is to characterize the range of risk and to also highlight exceptionally susceptible
populations (USEPA, 1995a, 2000a).
While the need for a full characterization of risk is stated in EPA's policies and guidance,
the actual characterization in risk assessments may not be so explicit. Therefore, we may need to
more consistently characterize the range of risk and highlight susceptible populations in our risk
assessments. This greater transparency will help make the reasonableness of the estimated risk
for the exposed population or ecosystem more understandable. Related to this, the NRC pointed
out in 1994 that there is little empirical evidence to suggest that EPA's potency, exposure, or risk
estimates are markedly higher than estimates embodying a reasonable degree of prudence (i.e.,
the conventional benchmarks of the 95th or 99th percentiles that statisticians use) (NRC, 1994).
2.1.6 What Happens When Default Assumptions Are Combined?
Some comments assert that EPA has so overemphasized conservatism that most risk
estimates are alarmingly false, meaningless, and unscientific. It should be noted that the use of
default assumptions does not render the process or results non-scientific. The basis of the
argument that EPA's risk estimates are alarmingly false suggests that combining several values
(e.g., use of values at a default level of 95%) results in excessive overestimates of risk. For
example, it is implied that combining two 95th percentile defaults results in an estimate above the
99th percentile, that combining three 95th percentile defaults results in an estimate above the 99.9th
-------
Page 18 Examination of EPA Risk Assessment Principles and Practices
percentile, and so forth. Suggestions have also been made that using the mean instead of the 95th
percentile in these types of calculations would result in a less conservative risk estimate.
However, the final distribution in these types of calculations is influenced differently by
the different inputs, and just multiplying the numbers as implied above will not lead to the
answers above. In estimating exposure or risk using percentiles, how much "influence" using a
particular 95th percentile value has in the ultimate exposure or risk calculation depends on more
than just its position in the percentile rankings. It depends also on the variability of the data
within the distribution for the input factors, the shape of the input distributions, and even the
number of data points. If all the input variables show the same variability, shape, etc., then the
above reasoning about compounding values is true. This is rarely the case in actual situations,
however.
For illustrative purposes, consider a case where the calculated exposure is a product of
several input factors. If most of these factors have little variability (e.g., the average lifetime of
70 years, or the volume of air breathed per day varies within plus or minus 25% for almost all the
population subgroups) and the concentration of a chemical that the population is exposed to
varies by three orders of magnitude, then the resulting percentile position of the calculated
exposure within the resulting exposure distribution will be much more influenced by which value
is selected from the input distribution of concentrations than it will be for which value is selected
for the other factors. Selecting the mean value for the concentration input value and 95th
percentile values for the others will result in a calculated exposure that is much closer to the
mean of the resulting distribution than the 95th percentile (or higher), because the resulting
distribution is heavily influenced by the concentration input. Conversely, selecting the 95th
percentile from the concentration input distribution and the means of the others will result in a
calculated exposure that is close to the 95th percentile of the resulting exposure distributions.
Consequently, in the cases where all the input distributions are not the same in variability, where
the final estimate falls on the combined distribution depends on which input variable is selected
as 95th percentile.
In fact, the example above is similar to the RME (reasonable maximum exposure)
calculated in Superfund assessments. Typically, the concentrations in environmental media are
highly variable and the other parameters are less variable. Selecting the mean of the
concentration distribution, while setting one or more of the other parameters at the 95th percentile
value, usually results in an estimate that is reasonably conservative (with the acknowledgment
that each case, due to its unique data set, will be different).
An additional factor within the RME calculation is that the 95% upper confidence limit
(UCL) on the mean of the concentrations is used instead of the mean itself. Using the mean
concentration for an area with contaminated soil is in effect setting a scenario where it is equally
probable that the exposed individual will be at any given location in the area. When the site has
-------
Examination of EPA Risk Assessment Principles and Practices Page 19
"attractive nuisances" with higher concentrations, the risk assessor factors this into the
assessment. The coverage of the data set within the area affects the confidence in the mean
calculated from the data. For example, confidence in calculating the true mean for a one-acre site
with three data points will be much less than for a one-acre site with 300 data points. The 95%
UCL allows the assessor to have some confidence that the mean is not being underestimated for a
relatively sparse data set. On the other hand, for a robust data set, the 95% UCL will be quite
close to the mean itself, so it does not introduce appreciable additional conservatism into the
estimate of exposure. Further, the upper confidence limit on the mean is not the same as the 95th
percentile of observed concentrations.
For many other exposure calculations that EPA performs, the Agency's 1992 Guidelines
for Exposure Assessment (USEPA, 1992a) provide specific guidance on how to construct
estimates using distributions of contributing factors, taking into consideration that the resulting
exposure estimate will be more sensitive to some factors than others. These guidelines suggest
that when exposure data or probabilistic simulations are not available, an exposure estimate that
lies between the 90th percentile and the maximum exposure in the exposed population be
constructed "by using maximum or near-maximum values for one or more of the most sensitive
variables, leaving others at their mean values" (USEPA, 1992a).
Nor is just using the mean instead of the 95th percentile necessarily less "conservative."
A percentile distribution, such as the collection of data associated with soil concentrations on a
specific site, is an ordered series of data values. The values for the data maybe such that the 95th
percentile value may be higher or lower than the mean. While the mean of a data set is most
often below the 95th percentile, consider a data set of 100 soil samples in which 97 samples are at
the background level, 1 part per million (ppm), for a given pollutant, while the remaining
samples are at 50 ppm, 100 ppm, and 1,700 ppm. The 95th percentile of this data set is 1 ppm,
which is actually numerically indistinguishable from the minimum value; the mean, 1.947 ppm,
lies between the 96th and 97th percentile. Consequently, use of the 95th percentile value in a
calculation instead of the mean does not necessarily make the calculation more or less
conservative. While the mean does indeed fall below the 95th percentile for the vast majority of
cases EPA sees, the fact remains that mathematical manipulation of percentile data and selection
of values from percentile distributions for purposes of estimating resulting exposures in a given
scenario needs to be considered on an individual basis to determine the degree of "conservatism."
Risk estimates often involve adding risks from various chemicals. This can be done by
adding doses for chemicals with the same mode of action (the approach used by EPA's Office of
Pesticide Programs, or OPP), or it can be approximated by adding, for example, the resulting
single-chemical cancer risks for various chemicals (the approach EPA uses in permitting and in
its Superfund program). Adding individual chemical risks to estimate a combined, cumulative
risk presents several complexities (USEPA, 2003b). For example, chemical interactions, which
can lead to synergism or antagonism, are poorly understood. Also, there is a statistical
-------
Page 20 Examination of EPA Risk Assessment Principles and Practices
complexity because the risks themselves are expressed as upper bound estimates of risk and not
most likely estimates. Cogliano (1997) showed that adding upper bound estimates may make the
resulting risk sums more conservative, but that the actual resulting risk values are not misleading
and are probably within a factor of 2 or 3 of the estimates that would result from calculating the
95% UCL of the sum of the most likely estimates. Cogliano states;
(A)s the number of carcinogens increases, the sum of upper bounds becomes increasingly improbable as an
estimate of overall risk. At the same time, however, the analysis shows that the sum of upper bounds is not
a misleading estimate of overall risk. Obtaining similar results for different case studies suggests that these
conclusions apply to more typical mixes of carcinogens.
Central estimates of the overall risk can differ from the sum of upper bounds by a factor of 2 - 5, as the ratio
between overall upper and lower bounds decreases ...
In conclusion, this analysis shows that sums of plausible upper bound risk estimates do provide useful
information about the overall risk from several carcinogens. The overall risk depends on the independence,
additivity, and number of risk estimates, as well as shapes of the underlying risk distributions.
In response to both uncertainty and variability, EPA develops risk estimates using default
assumptions based on empirical evidence or based on scientifically sound extrapolations.
Further, EPA risk assessments are in fact a combination of both high-end and central tendency
estimates. Consequently, the resulting risk estimates are expected to be on the high end of the
range of risks but within the range of plausible outcomes. The combination of default
assumptions is therefore reasonable, especially for independent factors, and does not result in
exaggerated estimates. On balance, while the resulting estimates are likely to be reasonable,
without a detailed uncertainty analysis it is not possible to determine where on the range of
plausible outcomes the estimates actually reside.
2.2 General Risk Assessment Approaches Used by EPA for Public and Environmental
Health Protection ("Public Health Protection")
2.2.1 How Comprehensive Are EPA's Risk Assessments?
EPA cannot perform a time- and resource-intensive risk assessment for every situation
and EPA decision. Consequently, for each risk assessment, EPA selects an approach that is
consistent with the nature and scope of the decision being made. The appropriate approach
depends on the needs of the decision maker and/or the role that risk information plays in the
decision, balancing uncertainty and resources. Even using the best models and data, uncertainty
is still inherent in the process. Given that uncertainty is inherent, there is a continuing tension
between improving our understanding in order to make a decision and the reality of limited
resources to perform the analysis and the desire for timely decision making. Figure 2-1
illustrates this risk assessment continuum and the balance of resources and uncertainty as the
assessment becomes more complex.
-------
Examination of EPA Risk Assessment Principles and Practices
Page 21
Figure 2-1. The Risk Assessment Continuum
The following graphic illustrates that risk assessment can be performed with low levels of
data and relatively little effort to develop conservative estimates of risk. Depending on the
outcome and the needs of the risk manager, higher levels of analysis may be performed. Note
that as one moves up the risk assessment continuum, the data needs and costs also rise.
However, the quality of the result should also rise as well. (The following graphic is intended to
be illustrative of the concept of tiered approaches. The actual modifications to the risk
assessment that occur as the assessment is refined may vary from the sequence described here
and are dependent on study-specific circumstances. In addition, a tiered approach is not
prescriptive: one may begin with a screening-level assessment or begin with a higher level of
analysis, as needed.)
Complete study-specific data, no assumptions; higher cost, lower uncertainty
Add quantitative uncertainty/variability analysis
More refined exposure assessment (e.g.,
probabilistic risk assessment)
More refined dispersion modeling and exposure
modeling
Simple dispersion model, no exposure modeling
No data, all assumptions; lower cost, high uncertainly
-------
Page 22 Examination of EPA Risk Assessment Principles and Practices
2.2.2 Whom Is EPA Trying To Protect?
EPA typically cannot protect every individual but rather attempts to protect individuals
who represent high-end exposures (typically around the 90th percentile and above) or those who
have some underlying biological sensitivity; in doing so, EPA protects the rest of the population
as well. In general, EPA tries to protect sensitive individuals based on normal distribution of
sensitivities. EPA considers the most sensitive individuals where there are data, but does not
necessarily attempt to protect "hypersensitive" individuals. The degree to which sensitive
individuals are protected, or explicitly defined, may vary between programs based on factors
such as the need to balance risk reductions and costs as directed and constrained by statutory
authority. Programs may approach the problem semi-quantitatively (e.g., selecting individual
parameter values at specified percentiles of a distribution) or qualitatively (e.g., making
conservative assumptions to ensure protection for most individuals), though no overall degree of
protection can be explicitly stated.
2.2.3 Are Risk Assessment and Risk Management Separate?
At EPA, risk assessment (evaluation of the science) and risk management (decision
making, setting of policy) are not necessarily separate. We believe it is appropriate to involve
decision makers from the beginning, as they typically initiate requests for risk assessments or
analyses. Consequently, separating them entirely from the risk assessment process is neither
logical nor desirable. Also, risk assessments typically are coordinated with the evaluations of
economics, feasibility of remedies, and community concerns, for example, so that their results
can be factored into decisions. EPA's Risk Characterization Handbook (USEPA, 2000a)
describes in detail the roles of the risk assessor and risk manager in the risk assessment process.
Further, the NRC report on understanding risk supports the concept that risk assessment is
conducted for the purpose of supporting risk management, and risk management considerations
shape what is addressed in a risk characterization (NRC, 1996).
Briefly, risk assessors are best qualified to understand the quality and nature of the data
and to use that data to determine what the risk is, who/what is affected, the level of comfort with
the conclusions, the uncertainty and variability inherent in the assessment, and the strengths and
weaknesses of the assessment. In other words, they are the best qualified to make the scientific
judgments necessary in risk assessments, including selecting models and data and assigning
defaults where data gaps exist. Risk managers are responsible for valuing the risk and determine
the amount of protection (as well as conservatism) to be applied in a decision. As a decision
maker, the risk manager integrates the risk assessment with other considerations in order to make
and justify regulatory decisions.
However, the comments point out that many of the default assumptions and policy
choices inherent in the risk assessment may frequently not be apparent to the risk managers. It is
-------
Examination of EPA Risk Assessment Principles and Practices Page 23
the role of the risk assessors to transparently characterize such details (e.g., default assumptions,
data selected, policy choices) so as to make clear the range of plausible risk. The risk managers,
in their role, should inquire about the use of defaults and the choices made, if not characterized
clearly, in order to be fully informed before using the risk assessment in making a decision.
2.2.4 How Do "Planning and Scoping" Help Environmental Risk Assessment?
In environmental risk assessments, planning and scoping are performed prior to the main
analytic work. This initial work defines the questions and issues to be addressed, the analysis
needed to address these questions/issues, and the knowledge and information needs of the
analysis.
Several general planning and scoping steps are relevant for many environmental risk
assessments (see below), but specific circumstances are also important. EPA's recently
completed Framework for Cumulative Risk Assessment (USEPA, 2003b) provides some general
planning and scoping steps:
a) Defining the Purpose of the Assessment
b) Defining the Scope of Analysis and Products Needed
c) Agreeing on Participants, Roles and Responsibilities
d) Agreeing on the Depth of the Assessment and the Analytical Approach
e) Agreement on the Resources Available and Schedule
f) Problem Formulation
g) Developing the Conceptual Model
h) Constructing the Analysis Plan
Problem formulation is a systematic planning step, linked to the regulatory and policy
context of the assessment, that identifies the major factors to be considered in a particular
environmental assessment. The problem formulation process results in a conceptual model that
identifies the sources, stressors, exposed populations, and the relationships among them.
The conceptual model and the associated narrative show the basic rationale for the
decisions made concerning the course of action for the risk assessment. Specifically, the
conceptual model and associated narrative provide: (1) the scientific rationale for selecting the
-------
Page 24 Examination of EPA Risk Assessment Principles and Practices
stressors, sources, receptors, exposed populations, exposure or environmental pathways, and
endpoints/effects; (2) the scientific, technical, economic, or sociologic basis for the conceptual
model; and (3) the scientific implications of additional data gathering.
The final stage in the planning and scoping process is the development of the analysis
plan (see discussion in USEPA, 1998a). This plan describes how hypotheses about the
relationships among the sources, stressors, exposure conditions, populations, and adverse
effects/endpoints presented in the conceptual model and narrative will be considered during the
risk analysis phase of the assessment. The plan includes a rationale for which relationships are to
be addressed and which methods and models will be used, and discusses data gaps and
uncertainties. The plan may also compare the level of confidence needed for the management
decision with the confidence levels expected from alternative analyses in order to determine data
needs and evaluate which analytical approach is best.
As stated in the EPA Science Policy Council's 2002 Lessons Learned on Planning and
Scoping for Environmental Risk Assessments (USEPA, 2002b), formal planning and dialogue can
improve the final risk assessment product by making it more specific to the needs of decision
makers and other stakeholders. Many questions and issues may be candidates for consideration
and analysis in a risk assessment. Planning and scoping help define the boundaries of the
analytic work (i.e. "what's in and what's out" of the assessment), reducing ambiguities about
what can and cannot be done with assessment results, and helping to reduce analytic work to
manageable segments.
As important as how planning and scoping is done is that it is done. In general, planning
and scoping are performed, to some degree, for all assessments. However, they may not be
inclusive or explicit on all factors. For example, simplified analytical approaches have been
developed and implemented over time. For each of these methods or analytical approaches,
choices of what's in and what's out have already been made. These choices made may have been
explicit when these approaches were developed, but may have become accepted over time and
therefore no longer acknowledged. This may be especially true for the toxicity reference values.
Some of the criticisms related to these components may stem from the lack of explicit
characterizations of what they represent.
2.2.5 How Does EPA Use a Screening Risk Assessment?
Where data are sparse and uncertainty great, EPA carries out a screening risk assessment
that tends to use default assumptions to avoid underestimating risk. These screening assessments
typically provide high-end and bounding estimates. Pathways of trivial importance are then
eliminated, and the remaining estimates are refined. This approach either demonstrates with
minimal effort that no risk is large enough to consider reducing or, if that is not the case, it
eliminates further work on refining estimates for pathways or chemicals that are clearly not
-------
Examination of EPA Risk Assessment Principles and Practices Page 25
important. This is consistent with the NRC's recommendation that EPA should use bounding
estimates for screening assessments to determine whether further analysis is necessary (NRC,
1994). If risks are not of concern even with these high-end or bounding screening estimates, then
one can be fairly confident that the risks are not of concern. However, if these screening tests
show that the risks are potentially of concern, then a more refined risk assessment may be
warranted that uses more detailed data, models, etc., though at greater expense. An example is
provided in section 2.2.7. These high-end screening assessments usually contain many default
assumptions since data are generally not available. However, when usable data are available,
they are considered instead of the defaults.
Many comments that focused on the screening assessments of risks may have
misunderstood the purpose of these assessments. For example, the degree to which default
assumptions are used in exposure assessments depends on the purpose of the assessment. By
their very nature, screening assessments are broad in scope and based on relatively sparse data.
In these cases, EPA attempts to clearly identify that the assessments are for screening purposes
and to explain the meaning of the results and the utility of the assessment in the context of
whatever decision is at hand.
2.2.6 What Happens if EPA Identifies a Potential Risk That Needs To Be
Addressed After a Screening Risk Assessment?
When a screening assessment identifies the potential for a non-trivial risk, EPA decides if
pursuing that risk is appropriate based on its current priorities and available resources. If the
Agency decides to pursue the risk, more detailed, refined risk assessments are then performed,
though the degree of refinement (i.e., where the risk assessment falls along the continuum shown
in figure 2-1) depends on the type of decision, the available resources, and the needs of the
decision maker. For those pathways or chemicals that were shown to be non-trivial by bounding
estimates, we work to refine our estimates of exposure and dose. At this point we estimate
exposures, doses, and responses that fall on the distribution of actual exposures pertinent to the
population under study. In performing this continued analysis, we use a combination of data,
ranges of data, distributions of data, and assumptions about each of the factors needed to estimate
risk. Generally, we perform both central tendency and high-end estimates (and, increasingly, we
develop fully probabilistic risk distributions). Each of these estimates is surrounded by
uncertainty (perhaps unquantifiable); the degree of uncertainty depends on the quality and
comprehensiveness of the available data.
For the more refined assessment, EPA would like to use appropriate and available data to
generate a more data-based assessment. This poses great difficulty when the data are not
available and/or adequate. If data are simply not available, then we usually employ basic default
assumptions. More frequently, data are available but deemed inadequate by EPA. This creates a
potentially confrontational situation if the generator of the data claims the data are adequate.
-------
Page 26 Examination of EPA Risk Assessment Principles and Practices
Peer review has been extremely helpful for determining the adequacy of the data, and EPA
makes extensive use of this process. When peer review judges the data to be inadequate, we
generally fall back on using defaults to help assess the risk. This situation probably is the basis
for many of the comments that EPA does not use current data and/or chemical- or site-specific
data. In fact, we do try to use the most relevant information as validated by peer review.
When exposure or dose estimates have sufficiently narrow uncertainty relative to the
needs of the decision maker, we can develop the final risk assessment. Otherwise, the data or
assumptions used usually have to be even further refined, if resources allow, in an attempt to
further reduce uncertainty and bring the estimated exposure or dose closer to the actual values in
the population. Refining the estimates usually requires that new data be considered. These data
may come from other studies in the literature, information previously developed for a related
purpose and adapted, or new survey, laboratory, or field data. The decision about which
particular parts of the information base to refine should be based both on which data will most
significantly reduce the uncertainty of the overall exposure or dose estimate of interest, and on
which data are in fact obtainable either technologically or within resource constraints. After
refinement of the estimate, we again determine whether the estimates provided will be sufficient
to answer the questions posed to an acceptable degree, given the uncertainties that may be
associated with those estimates. Refinements proceed iteratively until the assessment provides
an adequate answer for the decision maker within the resources available.
2.2.7 How Are High-End Exposures Reflected in EPA Evaluations?
Although populations experience a range (or distribution) of exposures, a number of
environmental statutes require that EPA consider those exposures at the high end of the
distribution when making certain decisions. Utilizing high end exposures as one component of
the risk decision making process helps to ensure equitable protection across an exposed
population (i.e., protecting a high end person in a population helps ensure protection of most of
the population).
One example of this is EPA's approach for evaluating a population's high end exposure
to hazardous air pollutants (HAPs). This has sometimes been referred to as evaluation of "the
porch potato" (i.e., the assumption that someone lives outdoors at the point of maximum
concentration at or beyond the fence line of a facility for 24 hours a day for a lifetime). This
section discusses the statutory and analytical frameworks that provide the basis for EPA's risk
assessments for HAPs pursuant to certain provisions of the Clean Air Act (CAA).
-------
Examination of EPA Risk Assessment Principles and Practices
Page 27
What Is the Statutory Basis for EPA's Selection of Exposure Scenarios When Assessing
Hazardous Air Pollutant Risks'?
The Clean Air Act identifies the risk to the individual most exposed (IME) as the risk of
interest when making certain decisions about the regulation of HAPs. Specifically, the 1990
CAA Amendments direct EPA to consider risk to the IME when determining whether a source
category may be deleted from the list of sources of HAPs (Section 112(c)(9)(B)(i)), and when
determining whether residual risk standards are necessary (Section 112(f)(2)(A)). Residual risk
standards are the mechanism that Congress provided EPA for addressing public health and
environmental risks that may remain after the regulated community has implemented technology-
based standards for the control of HAPs.
In addition, Section 112(f)(2)(B) of the 1990 CAA Amendments incorporates by
reference the use of a two-step risk assessment framework for setting residual risk standards
under Section 112(f)(2)(A). (Note: EPA has not yet proposed any residual risk standards, but is
working on a number of residual risk determinations and expects to issue the first proposal in
2004.) The two-step framework was articulated in EPA's 1989 Benzene NESHAP (54 Federal
Register 38044) and consists of:
1) A first step, in which EPA ensures that risks are "acceptable." As explained in the
Benzene NESHAP, in this step EPA generally limits the maximum individual risk,
or "MIR," to no higher than approximately 1 in 10 thousand. The benzene
NESHAP defines the MIR as the estimated risk that a person living near a plant
would have if he or she were continuously exposed to the maximum pollutant
concentrations for a lifetime (70 years).
2) A second step, in which EPA establishes an "ample margin of safety." In this
step, EPA strives to protect the greatest number of persons possible to an
estimated individual excess lifetime cancer risk level no higher than
approximately 1 in 1 million.
In judging whether risks are acceptable and whether an ample margin of safety is
provided, the benzene NESHAP states that EPA will consider not only the magnitude of
individual risk, but "the distribution of risks in the exposed population, incidence, the science
policy assumptions and uncertainties associated with the risk measures, and the weight of
evidence that a pollutant is harmful to health." Therefore, decisions under Section 112(f)(2)(A)
typically will include consideration of both population risk and individual risk, as well as other
factors. (Note that the ample margin of safety analysis in the second step of the residual risk
framework also includes consideration of additional factors relating to the appropriate level of
control, "including costs and economic impacts of controls, technological feasibility,
uncertainties, and any other relevant factors.")
-------
Page 28 Examination of EPA Risk Assessment Principles and Practices
What Analytical Framework Does EPA Use to Estimate the "1MB " and "MIR ? "
Risk assessment for air toxics involves using models to estimate HAP concentrations in
air (and other media, as necessary), and then combining these HAP concentrations with other
exposure assumptions and measures of pollutant toxicity to estimate risk (usually
deterministically). As a result, OAR's risk estimates are ultimately a function of the values
selected for numerous parameters, including: HAP-specific emission factors, meteorological
parameters (e.g., wind speed, wind direction, precipitation, temperature), stack parameters (e.g.,
height, diameter, temperature, exit velocity), distance to receptor, and duration of exposure (e.g.,
residential occupancy period).
The exposure scenario that EPA uses to estimate the MIR is provided in the Benzene
NESHAP, The Benzene NESHAP specifies that the MIR be based on exposure to the maximum
pollutant concentrations for 70 years. The Benzene NESHAP indicates that it is appropriate to
account for habitability when identifying the location of maximum pollutant concentrations.
Therefore, EPA may characterize the maximum off-site annual average concentration in
habitable areas (e.g., excluding such areas as lakes) to estimate the MIR.
OAR typically uses a tiered analytical approach when estimating risk to the ME. As
discussed above, because there are many parameters that influence an individual's exposure, it is
not possible to identify the actual individual in the population who is most exposed. OAR
recognizes, however, that it is very unlikely that there exists an individual who should be
characterized using the worst-case values of all exposure and toxicity parameters. Nevertheless,
a tiered analysis often begins with a "worst-case" or bounding analysis that generally sets
parameters at values that maximize the estimate of risk (e.g., exposure is assumed to continue for
a lifetime). If risks estimated using such an analysis are not of concern, then there is no need to
refine the analysis further, and EPA may proceed with the appropriate action (i.e., EPA may
delete the source pursuant to 112(c)(9)(B)(i) or make a determination not to propose residual risk
standards under 112(f)(2)(A)). If risks estimated in the initial analysis are of potential concern,
analysts may make successive refinements in modeling methodologies and input data to derive
successively less conservative, more realistic, estimates of the risk to the IME.
In refining the risk estimate for the IME, the exposure assumptions for which OAR has
the best alternative information generally are modified first. For example, where facility-specific
information is available, OAR will use the actual locations of residences (e.g., placing the
receptor in the geographic center of a populated census block near a facility) to estimate the
concentrations of HAPs in air. OAR also may conduct additional data gathering and analysis to
further refine the characteristics of the actual emissions and emission points. The goal of
successive tiers of the analysis is to start with a bounding or worst-case exposure estimate and
then move away from the worst-case parameters until the combination of parameter values
represents, in the judgment of the assessor, the exposure experienced by the individual in the
-------
Examination of EPA Risk Assessment Principles and Practices
Page 29
population who is most exposed. However, where the risk estimates are below the level of
concern, there is no need to conduct additional analysis to refine the estimates further, even if the
estimates might still overestimate the risk to the individual in the population who is most
exposed.
As described above, the risk estimate is a function of the values of a number of
parameters. In refining the bounding or worst-case risk analysis, OAR has not modified the
assumption of 70-year, 24-hour per day, outdoor exposure. Although OAR recognizes that the
majority of people do not reside outdoors and in one location for their entire lives, and that there
is a general trend of increased population mobility, OAR believes that the data available for
refining assumptions for exposure duration and frequency are less certain than the data available
for refining other parameter values. For example:
a) Residential occupancy periods may be influenced by factors (e.g., economic,
geographic) that may cause local population mobility patterns to differ from
national estimates of population mobility.
b) If a source of concern occurs in the majority of communities in the country, then it
is possible that an individual may be exposed to the source for a longer period of
time than one might predict using national estimates of population mobility. That
is, even though an individual moves, the individual's new residence may be
located near a similar source of concern.
c) If a single source impacts a large geographic area, then it is possible that an
individual may move or travel from one point of exposure to the source to another
point of exposure to the same source.
d) Exposure to HAPs may not diminish when individuals are indoors. Empirical
data for many pollutants show that long-term average indoor concentrations of
outdoor air pollutants are roughly equivalent to long-term average outdoor
concentrations of those pollutants (e.g., Sexton et al., 2004).
Moreover, OAR is cognizant that NRC acknowledged some of these issues in its 1994 report,
which recommended that: "EPA should use the mean of current life expectancy as the
assumption for the duration of individual residence time in a high-exposure area, or a distribution
of residence times which accounts for the likelihood that changing residences might not result in
significantly lower exposure. Similarly, EPA should use a conservative estimate for the number
of hours a day an individual is exposed, or develop a distribution of the number of hours per day
an individual spends in different exposure situations."
-------
Page 30 Examination of EPA Risk Assessment Principles and Practices
3. UNCERTAINTY AND VARIABILITY
3.1 Overview
Uncertainty and variability exist in all risk assessments. Even at its best, risk assessment
does not estimate risk with absolute certainty. Thus, it is important that the risk assessment
process handle uncertainties in a predictable way that is scientifically defensible, consistent with
the Agency's statutory mission, and responsive to the needs of decision makers (NRC, 1994).
Instead of explicitly quantifying how much confidence there is in a risk estimate, EPA attempts
to increase the confidence that risk is not underestimated by using several options to deal with
uncertainty and variability when data are missing. For example, in exposure assessment, the
practice at EPA is to collect new data, narrow the scope of the assessment, use default
assumptions, use models to estimate missing values, use surrogate data (e.g., data on a parameter
that come from a different region of the country than the region being assessed), and/or use
professional judgment. The use of individual assumptions can range from qualitative (e.g.,
assuming one is tied to the residence location and does not move through time or space) to more
quantitative (e.g., using the 95th percentile of a sample distribution for an ingestion rate). This
approach can also fit the practice of hazard assessment when data are missing. Confidence in
ensuring that risk is not underestimated has often been qualitatively ensured through the use of
default assumptions.
EPA has been increasingly concerned about characterizing uncertainty in its risk
estimates. EPA's 1986 set of Risk Assessment Guidelines explicitly stated the importance of
characterizing uncertainty. EPA's Exposure Assessment Guidelines developed this theme further
for the exposure assessment part of risk assessment. EPA's Risk Characterization Policy
provided even more direction for describing uncertainty in risk estimates. For probabilistic
analysis specifically, EPA made significant efforts in recent years to use probabilistic techniques
to characterize uncertainty; these include the March 1997 Guiding Principles for Monte Carlo
Analysis (USEPA, 1997b), the May 1997 Policy Statement (USEPA, 1997c), and the December
2001 Superfund document Risk Assessment Guidance for Superfund: Volume III Part A,
Process for Conducting Probabilistic Risk Assessment (USEPA, 200 la)
3,2 Uncertainty and Variability
3.2.1 What Is Uncertainty?
Uncertainty can be defined as a lack of precise knowledge as to what the truth is, whether
qualitative or quantitative.
-------
Examination of EPA Risk Assessment Principles and Practices
Page 31
Numerous schemes for classifying uncertainty have been proposed. The preferred
approach of the NRC (1994) focused on two broad categories: parameter uncertainty and model
uncertainty. These are defined below in sections 3.2.2 and 3.2.3.
3.2.2 What Is Parameter Uncertainty?
Risk assessments depict reality interpreted through mathematical representations that
describe major processes and relationships. Process or mechanistic models use equations to
describe the processes that an environmental agent undergoes in the environment in traveling
from the source to the target organism. Mechanistic models have also been developed to
represent the toxicokinetic and toxicodynamic processes that take place inside the organism,
leading to the toxic endpoint. The specific parameters of the equations found in these models are
factors that influence the release, transport, and transformation of the environmental agent, the
exposure of the target organism to the agent, transport and metabolism of the agent in the body,
and interactions on the cellular and molecule levels. Empirical models are also used to define
relationships between two values, such as the dose and the response. Uncertainty in parameter
estimates stem from a variety of sources, including:
a)
Measurement errors:
b)
c)
d)
e)
use of
1) Random errors in analytical devices (e.g., imprecision of continuous
monitors that measure stack emissions).
2) Systemic bias (e.g., estimating inhalation from indoor ambient air without
considering the effect of volatilization of contaminants from hot water
during showers).
Use of surrogate data for a parameter instead of direct analysis of it (e.g.,
standard emission factors for industrialized processes).
Misclassification (e.g., incorrect assignment of exposures of subjects in historical
epidemiologic studies due to faulty or ambiguous information).
Random sampling error (e.g., estimation of risk to laboratory animals or exposed
workers in a small sample).
Non-representativeness (e.g., developing emission factors for dry cleaners based
on a sample of "dirty" plants).
-------
Page 32
Examination of EPA Risk Assessment Principles and Practices
3.2.3 What Is Model Uncertainty?
Model uncertainties arise because of gaps in the scientific theory that is required to make
predictions on the basis of causal inferences. Common types of model uncertainties in various
risk assessment-related activities include;
a) Relationship errors (e.g., incorrectly inferring the basis of correlations between
chemical structure and biological activity).
b) Oversimplified representations of reality (e.g., representing a three-dimensional
aquifer with a two-dimensional mathematical model).
c) Incompleteness, i.e., exclusion of one or more relevant variables (e.g., relating
asbestos to lung cancer without considering the effect of smoking on both those
exposed to asbestos and those unexposed).
d) Use of surrogate variables for ones that cannot be measured (e.g., using wind
speed at the nearest airport as a proxy for wind speed at the facility site).
e) Failure to account for correlations that cause seemingly unrelated events to occur
more frequently than expected by chance (e.g., two separate components of a
nuclear plant are both missing a particular washer because the same newly hired
assembler put them together).
f) Extent of (dis)aggregation used in the model (e.g., whether to break up fat
compartment into subcutaneous and abdominal fat in a physiologically based
pharmacokinetic, or PBPK, model).
Model uncertainty is often difficult to quantify. Further, it is inherent in risk assessment
that seeks to capture the complex processes impacting release, environmental fate and transport,
exposure, and exposure-response. EPA's models are often incomplete and knowledge of specific
processes limited. As a result, EPA relies on specific default assumptions as a response to
uncertainty. Again, these represent scientifically plausible choices that are intended to more
likely guard against underestimating risks.
3.2.4 How Does Variability Differ From Uncertainty?
Variability is considered with uncertainty since, like uncertainty, it does not allow for a
specific single correct value. However, they differ significantly in terms of their impact and role
in risk assessments. Uncertainty the lack of knowledge can be reduced through additional
-------
Examination of EPA Risk Assessment Principles and Practices Page 33
investigation. Variability is inherent heterogeneity across space, in time, or among individuals; it
cannot be reduced with additional investigation, only better understood or characterized.
"Human variability" refers to person-to-person differences in biological susceptibility or
in exposure. Although both human variability and uncertainty can be characterized as ranges or
distributions, they are fundamentally different concepts. Again, uncertainty can be reduced by
further research that supports a model or improves a parameter estimate, but human variation is a
reality that can be better characterized, but not reduced, by further research. (Note that fields
other than risk assessment use "variation" or "variability" to mean dispersion about a central
value, including measurement errors and other random errors that risk assessors address as
uncertainty.)
EPA addresses variability by assessing the risk to the sensitive portions of the population
or ecosystem (see section 3.6 for fuller discussion). Accordingly, EPA typically makes explicit
choices to characterize the risks at the upper end of the expected distribution. Typically, EPA
focuses on the critical parameter or assumption in any particular aspect of an assessment and
based on the particular needs of the decision maker, including the mandates and constraints in
statutory language, and selects the appropriate choice such as the maximum or a specified
percentile. In general, EPA's approach to variability has focused on exposure (e.g.,
characterizing the risks to the individual most exposed within the Air Toxics or Hazardous Air
Pollutant program, section 2.2.7), but the approach may also be reflected in the toxicity part of
the assessment (e.g., UF for human variability).
3.3 Characterizing Uncertainty and Variability
One of the major comments on EPA risk assessment practices is that they do not
characterize uncertainty and variability transparently enough. This is an issue EPA is attempting
to address. The comments also encourage EPA to begin using probabilistic analysis to describe
quantitatively the uncertainty in risk estimates where appropriate.
3.3.1 Why Is Characterizing Uncertainty and Variability Important?
The very heart of risk assessment is the responsibility to use whatever information is at
hand or can be generated to produce an estimate, a range, a probability distribution whatever
best expresses the present state of knowledge about the effects of some hazard in some specific
setting. To ignore the uncertainty in any process is almost sure to leave critical parts of the
process incompletely examined and hence to increase the probability of generating a risk estimate
that is incorrect, incomplete, or misleading (NRC, 1994).
The NRC (1994) further noted that risk assessments that do not pay sufficient attention to
uncertainty are vulnerable to four common, potentially serious pitfalls:
-------
Page 34 Examination of EPA Risk Assessment Principles and Practices
a) They do not allow for optimal weighting of the probabilities and consequences of
error.
b) They do not permit a reliable comparison of alternative decisions.
c) They fail to communicate the range of control options that would be comparable
with different assessments of the true state of nature.
d) They preclude the opportunity for identifying research initiatives.
Further, uncertainty analysis will play a more prominent and formal role in regulatory
decision making. For example, OMB's recent revisions to its regulatory analysis guidelines
(USOMB, 2003c) state that formal quantitative uncertainty analysis be performed for economic
assessment in support of overall regulatory analysis. Whenever possible, appropriate statistical
techniques should be used to determine the probability distribution of relevant outcomes. For
major rules involving annual economic effects of $1 billion or more, a formal quantitative
analysis of uncertainty is required. The OMB guidelines outline analytical approaches of varying
levels of complexity, which could be used for uncertainty analysis such as qualitative disclosure,
numerical sensitivity analysis, and formal probabilistic analysis (required for rules with impacts
greater than $1 billion).
3.3.2 When Should EPA Conduct Uncertainty Analysis?
As the NRC notes, EPA's analysis of uncertainty has tended to be piecemeal and highly
focused on an assessment's sensitivity to the accuracy of a few specified assumptions rather than
a full exploration of the process from data collection to final risk assessments (NRC, 1994).
Further, EPA tends to conduct uncertainty analysis a posteriori, focusing on specific assumptions
or variables. The level of treatment again depends on the decision and the value such an analysis
would provide to that decision. EPA believes this is an appropriate approach to balance
resources with value added. Some have suggested that EPA should integrate analysis and
uncertainty into the basic development of the primary study (see USEPA, 2000e). Until such
methods and supporting data are developed, though, it is not feasible to do a full and integrated
assessment for every analysis. Further, in most instances EPA is not the data developer; much
EPA analysis is based upon third-party literature.
Ideally, an uncertainty analysis would be built into the study design; hopefully, as
methods and supporting data continue to be developed, we will move toward such an approach in
the future.
-------
Examination of EPA Risk Assessment Principles and Practices Page 35
3.3.3 What Is an Appropriate Level of Uncertainty Analysis?
Over the years, improved computer capabilities have created more opportunities to
characterize uncertainty. As a result, advocates promote such characterization in all cases. We
need to be judicious in which methods we apply, such as Monte Carlo analysis. Uncertainty
analysis is not a panacea, and fiill formal assessments can still be time- and resource-intensive.
Further, the time and resources needed to collect an adequate database for such analyses can be a
problem. While uncertainty analysis arguably provides significant information to aid in decision
making, its relative value is case-specific and depends on the characteristics of the assessment
and the decision being made. In some cases, a full probabilistic assessment may add little value
relative to simpler forms. This may occur where more detailed uncertainty analysis (or analysis
focused on non-critical uncertainties) does not provide information which has any impact on the
overall decision.
Accordingly, EPA's practice is to use a "tiered approach" to conducting uncertainty
analysis; that is, EPA starts as simply as possible (e.g., with qualitative description) and
sequentially employs more sophisticated analyses (e.g., sensitivity analysis to full probabilistic),
but only as warranted by the value added to the analysis and the decision process. Questions
regarding the appropriate way to characterize uncertainty include:
a) Will the quantitative analysis improve the risk assessment?
b) What are the major sources of uncertainty?
c) Are there time and resources for a complex analysis?
d) Does this project warrant this level of effort?
e) Will a quantitative estimate of uncertainty improve the decision? How will the
uncertainty analysis affect the regulatory decision?
f) How available are the skills and experience needed to perform the analysis?
g) Have the weaknesses and strengths of the methods involved been evaluated?
h) How will the uncertainty analysis be communicated to the public and decision
makers?
-------
Page 36 Examination of EPA Risk Assessment Principles and Practices
3.3.4 With What Precision Should EPA Results Be Reported?
Uncertainty is inherent in almost all the data EPA reviews. However, at the end of many
risk assessments, values are provided that appear to have several significant digits. This implies
a certain level of precision that probably is not justified. EPA is currently addressing this
practice of characterizing the values.
With the increased use of computational methods and readily available software,
estimates can be reported with apparently unlimited precision. Nonetheless, the precision to
which a value is reported should be commensurate with the level of confidence in that number.
EPA has been criticized in the past for reporting risk estimates with unnecessary and
unsupportable degrees of precision. Often, this is the result of automated reporting within
commercial software packages, or of incremental stages of analysis that are not reviewed and
edited in accordance with their incremental nature.
EPA agrees that results should be reported to their proper level of precision and the
value(s) well characterized. EPA guidance on presentation of risk information comes from
various documents developed since 1989. The Risk Assessment Guidance for Superfund:
Volume I Part A (USEP A, 1989a) provides guidance on the number of significant figures to
use in presenting cancer risks in the risk characterization. RAGS Part A's page 8-7 (Exhibit 8-2,
footnote b) states that "All cancer risks should be expressed as one significant figure only." In
addition, theRisk Assessment Guidance for Superfund: Volume 1 Part D (USEP A, 2001b)
provides standardized tables for presenting risk information. The Record of Decision guidance
provides standardized tables for presenting risk information (USEPA, 1999a); the same
document provides guidance on presenting non-cancer hazard indices.
In a final risk description, all cancer risks and non-cancer hazard indices should be
presented with one significant figure only. More figures, however, should be carried along the
way to minimize rounding errors and to make it possible for others to verify calculations. The
number of significant figures presented during a risk assessment changes to reflect the needs of
the assessment. For example, risk information presented in RAGS Part D tables (tables 6
through 9 in RAGS) typically present cancer risk information with two significant figures. This
summary information allows risk assessors reviewing the document to check the accuracy of the
cancer risk calculations and provide appropriate comments.
3,4 Issues in Characterizing Uncertainty and Variability
EPA has been examining its practice of uncertainty analysis over the years. Generally
speaking, EPA has applied a qualitative approach to characterizing uncertainty (and variability
for that matter). Quantitative characterization has been used more unevenly. Also, EPA
-------
Examination of EPA Risk Assessment Principles and Practices
Page 37
typically uses deterministic approaches to characterize risk although, increasingly often, EPA
applies probabilistic techniques for characterization of risk, usually within exposure assessments.
3.4.1 What Is the Issue of Deterministic Versus Probabilistic Approaches?
Risk assessments may consider both non-cancer and cancer endpoints when data are
available. A non-cancer assessment includes an oral reference dose (RfD) and an inhalation
reference concentration (RfC). A cancer assessment includes a cancer weight-of-evidence
narrative statement and, if data permit, a quantitative cancer risk estimate. Many of the risk
estimates from these assessments are found in the Agency's Integrated Risk Information System
(IRIS) as well as in EPA offices that also develop health assessments.,
As an example for the practice of characterizing uncertainty, we examine the traditional
approach to dose-response assessment analysis of non-cancer endpoints (RfD or RfC). The
approach is to identify the no-observed-adverse-effect level (NOAEL) and lowest-observed-
adverse-effect level (LOAEL) from an appropriate study. The NOAEL (or LOAEL, if a NOAEL
is not available) is adjusted downward by uncertainty factors (UFs) intended to account for
uncertainties in the available data, producing an exposure that is likely to pose no increased risk
of adverse effects for chronic exposure. EPA currently also uses the benchmark dose (BMD)
approach to deriving a RfD/RfC when appropriate data are available.
While there may be significant statistical analysis and consideration of (both qualitative
and quantitative) uncertainty, an estimated reference value or exposure is largely developed via a
deterministic process and yields a deterministic estimate. Part of this approach is the use of UFs
that are applied to adjust for uncertainties in extrapolating from the type of study serving as the
basis for the RfD/RfC to the situation of interest for the risk assessment. UFs are used to account
for each of the extrapolations used in the assessment, such as use of animal data (interspecies
extrapolation) and protection of susceptible individuals (intraspecies variability).
A deterministic value for "a dose or exposure not expected to cause adverse effects" may
convey an inappropriate degree of precision, and some think that a single number is not
appropriate. A single number is frequently presented as "the risk assessment"; it is clear this is
not an adequate way to characterize any potential risk. Several considerations need to be
characterized to increase confidence in the assessment (i.e., to reduce uncertainty);
a) The degree of confidence needed in stating that susceptible/sensitive individuals
are affected or not above the resulting exposures.
b) The condition of the database from which the original BMD, NOAEL, or LOAEL
was developed (i.e,} the quality of the individual study on which the value is
based).
-------
Page 38
Examination of EPA Risk Assessment Principles and Practices
c) The numerical values for the UFs employed in calculating the RfC/RfD.
d) The procedure by which the BMD, NOAEL, or LOAEL from the experimental
system may have been adjusted to account for known differences in the human
population of interest (e.g., dosimetric adjustments).
e) The shape of the dose-response curve.
A probabilistic approach could be advantageous in that it would eliminate the need to
define a single value and might be less likely to imply undue precision. In such a framework, a
probability distribution would be used to express the belief that any particular value represents
the dose or exposure concentration that would pose no appreciable risk of adverse effects. In
contrast with the imprecision of a deterministic approach, several researchers still attempt to
interpret what is meant by the RfD or RfC and what is the degree of protection afforded (e.g.,
Swartout et al., 1998; Baird et al., 1996; Gaylor and Kodell, 2000).
3.4.2 What Is the Importance of Quantitative Characterization of Uncertainty in
Dose-Response?
EPA has stressed the importance of attention to uncertainty in risk assessment. We quote
extensively here from the successive versions of EPA's Cancer Guidelines, since they articulate
the issue well in terms of cancer risk assessment. Many of the points apply to this issue for risk
assessment in general. The 1986 Cancer Guidelines state:
It should be emphasized in every quantitative risk estimation that the results are uncertain. Uncertainties
due to experimental and epidemiologic variability as well as uncertainty in the exposure assessment can be
important. There are major uncertainties in extrapolating both from animals to humans and from high to
low doses. There are important species differences in uptake, metabolism, and organ distribution of
carcinogens, as well as species and strain differences in target-site susceptibility. Human populations are
variable with respect to genetic constitution, diet, occupational and home environment, activity patterns,
and other cultural factors. Risk estimates should be presented together with the associated hazard
assessment (section III.C.3.) to ensure that there is an appreciation of the weight of evidence for
carcinogenicity that underlies the quantitative risk estimates.
At that time EPA recognized the need for further development of methodologies to
address uncertainties;
It is also recognized that there is a need for new methodology that has not been addressed in this document
in a number of areas, e.g., the characterization of uncertainty. As this knowledge and assessment
methodology are developed, these Guidelines will be revised whenever appropriate.
-------
Examination of EPA Risk Assessment Principles and Practices Page 39
The 1999 draft revised Cancer Guidelines contain substantial additional discussion about
uncertainty. In the key area of dose-response assessment, they state;
The characterization presents the results of analyses of dose data, of response data, and of dose-response.
When alternative approaches are plausible and persuasive in selecting dose data, response data, or
extrapolation procedures, the characterization follows the alternative paths of analysis and presents the
results.
This discussion goes on to address specific areas where uncertainties, including
quantitative uncertainties, are to be considered:"Uncertainty analyses, qualitative or quantitative
if possible, are highlighted in the characterization." The guidelines continue with a list of
specific factors to include in the characterization.
The 2003 draft final Cancer Guidelines contain a further expansion in the discussion of
uncertainty issues:
Some aspects of model uncertainty that should be addressed in an assessment include the use of animal
models as a surrogate for humans, the influence of cross-species differences in metabolism and physiology,
the use of effects observed at high doses as an indicator of the potential for effects at lower doses, the effect
of using linear or nonlinear extrapolation to estimate risks, the use of small samples and subgroups to make
inferences about entire human populations or subpopulations with differential susceptibilities, and the use of
experimental exposure regimens to make inferences about different human exposure scenarios (NRC,
2002).
Toxicokinetic and toxicodynamic models are generally premised on site concordance across species ... The
assessment should discuss the relevant data that bear on this form of model uncertainty. ...
Probabilistic risk assessment, informed by expert judgment, has been used in exposure assessment to
estimate human variation and uncertainty in lifetime average daily dose. Probabilistic methods can be used
in this exposure assessment application because the pertinent variables (for example, concentration, intake
rate, exposure duration, and body weight) have been identified, their distributions can be observed, and the
formula for combining the variables to estimate the lifetime average daily dose is well defined (see USEPA,
1992a). Similarly, probabilistic methods can be applied in dose-response assessment when there is an
understanding of the important parameters and their relationships, such as identification of the key
determinants of human variation (for example, metabolic polymorphisms, hormone levels, and cell
replication rates), observation of the distributions of these variables, and valid models for combining these
variables. With appropriate data and expert judgment, formal approaches to probabilistic risk assessment
can be applied to provide insight into the overall extent and dominant sources of human variation and
uncertainty. In doing this, it is important to note that analyses that omit or underestimate some principal
sources of variation or uncertainty could provide a misleadingly narrow description of the true extent of
variation and uncertainty and give decision-makers a false sense of confidence in estimates of risk.
Specification of joint probability distributions is appropriate when variables are not independent of each
other. In each case, the assessment should carefully consider the questions of uncertainty and human
variation and discuss the extent to which there are data to address them.
Probabilistic risk assessment has been used in dose-response assessment to determine and distinguish the
degree of uncertainty and variability in toxicokinetic and toxicodynamic modeling. Although this field is
-------
Page 42
Examination of EPA Risk Assessment Principles and Practices
general public or the population at risk, which may not have the resources to promote such
research.
3.5 Inherent Variability in Biological Response
3.5.1 Why Consider Populations and Life-Stages?
Consideration of the variability among humans is a critical aspect of risk assessment. It is
the goal of EPA risk assessments to identify all potentially affected populations, including human
populations (e.g., gender, nutritional status, genetic predisposition) and life-stages (e.g.,
childhood, pregnancy, old age) that may be more susceptible to toxic effects or are highly or
disproportionately exposed (e.g., children, ethnic groups; see USEPA, 1992a, 2000a, 2002c).
While susceptible ecosystems and ecological entities (e.g., endangered species) are also
considered in ecological assessments, the following discussion focuses on human populations
and life-stages.
The term "susceptible" is also used to describe sensitive populations/life-stages, as these
two terms are often used interchangeably and no convention for their use is widely accepted.
Although "susceptible" has been used to describe susceptibility to toxic effect(s) and
disproportionate or unique exposures, it is more transparent to identify the two issues separately
in risk assessment.
The NRC recognized that many types of variability enter into risk assessment, including
the nature and intensity of exposure and susceptibility to toxic insult as affected by age, lifestyle,
genetic background, ethnicity, and other factors (NRC, 1994). The NRC outlined several
recommendations for considering such variability in risk assessment. The NRC also made
recommendations specific to risk assessments for children and infants as a potentially susceptible
group (NRC, 1993).
The consideration of populations and life-stages has also been described through the
concept of vulnerability, or the propensity for the system to suffer harm (USEPA, 2003b).
Factors included in vulnerability are:
a)
b)
Susceptibility or sensitivity to adverse effect: increased likelihood of sustaining an
adverse effect as a relationship to a factor describing the population (e.g., genetic
polymorphisms, prior immune reactions, disease state, or prior damage) and/or
life-stage.
Differential exposure: differences in current exposure, historical body burden, and
background exposure.
-------
Examination of EPA Risk Assessment Principles and Practices Page 43
c) Differential preparedness to withstand a stressor: linked to coping systems and
resources of an individual, population, or community with respect to prospective
mitigation efforts (e.g., immunization).
d) Differential ability to recover: linked to coping systems (e.g., differential survival
rates).
Legislation has called on EPA to consider potentially susceptible populations and life-
stages. The SDWA Amendments mandate that EPA consider risks to groups within the general
population that are identified as being at greater risk of adverse health effects, including children,
the elderly, and people with serious illness (SDWAA, 1996). Similarly, the Food Quality
Protection Act (FQPA) contains special provisions for the consideration of pesticide risks to
children (FQPA, 1996). In addition to legislative mandates, EPA has further guidance in
considering health and safety risks to children (EO 13045,1997; USEPA, 1995b).
3.5.2 Are Certain Populations and Life-Stages Always at Greater Risk?
When conducting risk assessments, EPA examines populations and life-stages that may
be especially sensitive to the stressor(s) being assessed. This does not imply that these examined
populations and life-stages are always at greatest risk. For each stressor and exposure scenario,
different data are available such that there is not a single or exact method for examining potential
susceptibility and associated risk. Risk assessment often uses an iterative approach. Populations
and life-stages may be assessed using defaults and assumptions in screening-level assessments,
while more detailed analyses of these groups will be performed for more refined assessments.
While the question "Are certain populations and life-stages at greater risk?" needs to be asked in
all iterations of risk assessment, the answer may not remain the same throughout.
3.5.3 How Are Sensitivities to Toxic Effects Considered?
When data are available to describe the toxicological differences for a susceptible
population or life-stage, then those data are summarized and analyzed, and the decisions based
on this information are presented. It is preferable to have population- and chemical-specific data
to describe a susceptibility to toxic effects. For example, the IRIS RfC for beryllium is based on
the human subpopulation that is susceptible to chronic beryllium disease (USEPA, 1998b; also
see section 3.5.5 below). Similarly, if data are available to indicate that susceptible populations
or life-stages are not at risk, those data are also used.
When one is analyzing and describing the available toxicity and effects data for
susceptible populations or life-stages, pragmatic considerations include:
-------
Page 44
Examination of EPA Risk Assessment Principles and Practices
a) Are epidemiologic, toxicologic, mechanistic, or other studies available that
investigate reproductive effects, developmental effects or windows of
susceptibility, multi-generation effects, gender or race differences, genetic
predisposition, modulation of existing disease, old age, or other possible
susceptibilities? Were qualitative (i.e., different effect) or quantitative (i.e., same
effect at lower dose or more severe effect at same dose) differences in
susceptibility observed? Both?
b) Have the available studies examined the likely and plausible endpoints?
c) Has a distribution been used in assessing toxicity (e.g., benchmark dose lower
confidence level)? Is it feasible that an entirely different distribution exists for
susceptible populations or life-stages, such that selecting a central value or even
tail of the existing distribution is not necessarily representative of susceptibility?
d) How have the findings of available studies been incorporated in the dose-response
assessment? Is it clear how susceptible populations and life-stages have been
considered in the quantitative assessment of toxicity?
e) What data are absent/needed? Can information from existing studies be
synthesized to identify specific data gaps?
It has been recognized that limited data are currently available for the a priori
identification of susceptible populations and life-stages for many chemicals and risk assessments
(USEPA, 2002c). In these situations, it is important that risk assessments clearly identify and
summarize data needs and uncertainties, in addition to the available data. Typically, when data
are limited, default practices are used:
a) Non-cancer effects: An intraspecies UF is used to account for variations in
susceptibility within the human population (USEPA, 2002c). This UF typically
has a value of 10-fold, but can be increased or reduced when sufficient data are
available. One can apply the same UF to carcinogens using a non-linear dose-
response model. For example, the IRIS chloroform oral carcinogenesis
assessment considers the non-cancer assessment to be protective against cancer
risk, and the same intraspecies UF is applied in the chloroform oral cancer and
non-cancer assessments (USEPA, 2001c).
A database UF may also be applied for deficiencies in the available data or when
existing data suggest that additional data may yield a lower reference value
(USEPA, 2002c). This UF is most often used when developmental or two-
generation reproduction studies are not available, but it may be applied in other
-------
Examination of EPA Risk Assessment Principles and Practices
Page 45
situations to account for the lack of data for potentially susceptible populations or
life-stages.
b) Cancer effects: An evaluation should be made as to whether low-dose linear
extrapolation is sufficient to protect susceptible populations (USEPA, 2003c).
For example, available data indicate that early life exposure to mutagenic
carcinogens may lead to greater incidence of cancer than the same lifetime
average daily dose received later in life. This information has been reviewed by
EPA and published in draft (USEPA, 2003d).
3.5.4 Are Unique or Disproportionate Exposures Considered?
Unique or disproportionate exposures are considered in parallel to the toxicity
considerations described in the preceding section. Available data describing potential unique or
disproportionate exposure are summarized and decisions based on this information are presented.
Again, it is preferable to have population- and chemical-specific data to address specific
exposure scenarios. For example, the Integrated Exposure Uptake Biokinetic Model (IEUBK) is
a very specialized tool that allows detailed analysis of lead exposures for children ages 6 months
to 7 years (USEPA, 200Id).
While some risk assessments consider exposure to the population as a whole, many only
consider exposed populations, not unexposed members of the general population. For example,
a hazardous waste assessment only considers the people who live near or on the site, or near the
incinerator or point source involved.
For exposure assessment, pragmatic considerations for unique and disproportionate
exposures include:
a) Are there unique or disproportionate exposures based on age, race, ethnicity,
gender, lifestyle, cultural practices, economic status, or other considerations?
b) How well do existing studies or distributions of exposure consider identified
populations or life-stages?
c) Have the exposures within these groups been considered in detail? For example:
pica soil ingestion by children, consumption of breast milk by infants, traditional
or subsistence diets, mouthing of objects by toddlers, different/increased food and
water consumption rates for children, restrictions on mobility (e.g., children
cannot leave a location or residence by choice as adults can; elderly populations
often have reduced mobility).
-------
Page 46
Examination of EPA Risk Assessment Principles and Practices
d) Have unique or disproportionate childhood exposures been integrated for lifetime
exposure? Children are not children for a chronic duration, but all adults spent a
portion of their lifetime in childhood; this needs to be taken into account as a
source of unique and disproportionate exposure as a part of a continuous lifetime.
When exposure data for certain populations and life-stages are limited, exposure
assumptions and default values are used to assess plausible current and/or future exposure
scenarios. The exposure factors published in EPA's Exposure Factors Handbook are widely
used in Agency exposure assessments (USEPA, 1997d, 2002d). In some scenarios, additional
default data may be used to describe situations the handbook does not cover. For example,
exposure parameters used to estimate childhood exposures due to mouthing exposures have been
used in pesticide exposure assessments (USEPA, 1997d).
It is essential that risk assessments, and exposure assessment in particular, explicitly
address environmental justice concerns. The goal of environmental justice is to ensure that all
people, regardless of race, national origin, or income, are protected from disproportionate
impacts of environmental hazards (EO 12898, 1994). For example, concerns regarding risks
associated with subsistence and the practice of traditional lifeways have been voiced by Native
American people across the United States and by the National Environmental Justice Advisory
Council (NEJAC, 2002). Executive Order 13175, "Consultation and Coordination With Indian
Tribal Governments," requires EPA to consult with tribes regarding protection of their land and
health. Tribes have specifically requested that EPA enter into formal consultation with them on
environmental issues, including exposures unique to tribes. These exposures are a result of
cultural or behavioral preferences of certain native populations. EPA typically bases its decisions
on the general U.S. population; statistics derived from the general population's exposure patterns
do not necessarily represent exposure patterns for all peoples. For example, the exposure
distribution for four Columbia River Treaty Tribes (CRITFC, 1994; also, see table 3-1, below)
shows that these peoples' average fish consumption rate is much higher than the rate for the
average U.S. person. This disproportionate exposure leaves Native Americans in these
communities with exposures that are underestimated or ignored in risk assessments.
-------
Examination of EPA Risk Assessment Principles and Practices
Page 47
Table 3-1: Fish Consumption Patterns for General U.S
Exposure Parameter
Tissue chemical concentration
Ingestion rate offish tissue (g/day)
Adults
Children under 15
Children under 6
. and Native American Populations
General Public
AFC
Average
HFC
Average
Columbia River Treaty Tribes
AFC
Average
HFC
Average
7.5*
2.83*
142.4b
77.95b
63.2°
24.8°
389"
162"
AFC average fish consumption; HFC high fish consumption
" Mean U.S. per capita consumption rate of uncooked freshwater and estuarine fish (USEPA, 2000f).
b 99'h percentile U.S. per capita consumption rate of uncooked freshwater and estuarine fish (USEPA, 2000f).
c Mean consumption rate for fish consumers in the Umatilla, Nez Perce, Yakama, and Warm Springs Tribes of the Columbia
River Basin (CRITFC, 1994).
d 99'h percentile consumption rate for fish consumers in the Umatilla, Nez Perce, Yakama, and Warm Springs Tribes of the
Columbia River Basin (CRITFC, 1994).
Situations for unique exposures also need to be considered. For example, the Asian
Pacific American population typically consumes foods that are not included in the U.S. diet
survey (Sechena et al., 2003). Thus, their exposure to contaminants in seaweed or moon snails
are not included in routine market basket surveys or the resulting exposure estimates.
Sometimes disproportionate exposures occur because of racial or socioeconomic
conditions. For example, exposure to environmental and household agents can trigger asthma
attacks (AAP, 2003); the Centers for Disease Control and Prevention reported that asthma
mortality rates among African Americans are 2.5-fold higher than among Caucasians (CDC,
2003). Similarly, children of low income families constitute 83% of the children ages 1 to 5 with
elevated blood lead levels that may result in health effects.
3.5.5 Is Human Variability Considered in Occupational Cohorts?
Occupational epidemiologic cohorts are an important source of human toxicity
information for use in risk assessment. Occupational studies are conducted with select
populations. Because of their presence in the workforce, these persons are believed to be healthy
(this is the "healthy worker effect"), predominately male, and generally adult. These factors
obviously limit a population's diversity with respect to susceptibility. As a result, the general
approach is to use an intraspecies UF greater than 1 for non-cancer endpoints when the RfD or
RfC is based on occupational data. This practice is consistent with the recommendation that the
reduction of the intraspecies UF of 10-fold be considered only if available data are sufficiently
representative of the exposure/dose-response data for the most sensitive populations and life-
stages (USEPA, 2002c).
-------
Page 48 Examination of EPA Risk Assessment Principles and Practices
The 2003 IRIS database was examined for all instances in which human occupational
data had been used to set an RfD and/or RfC. In the six instances in which occupational data had
been used to set the RfD, an intraspecies factor of 10 was used three times, an intraspecies factor
of 3 was used twice, and an intraspecies factor of 1 was used once. In the latter case (that of
beryllium and beryllium compounds), the health effect involved is beryllium sensitization and
progress to CBD. The intraspecies factor of 1 was chosen because only a small percentage of the
population is sensitive to CBD and sensitization: the affected population already represented a
sensitive group of people.
Occupational epidemiologic cohorts are also used in cancer assessments. For example, in
the health assessment of 1,3-butadiene, the cancer slope factor (CSF) was derived using an
occupationally exposed male cohort. In animal studies, mammary gland cancer was the only
common tumor in mice and rats (USEPA, 2002e), so a factor of 2 was used to protect the female
population. This is an example of observations from the occupational cohort which did not
include women and children not necessarily representing the risks to the larger, more diverse
human population.
3.5.6 What Needs To Be Done To Address the Risk Assessment of Populations and
Life-Stages?
Risk characterization is the opportunity to bring together all population and life-stage
considerations, including the toxicity and exposure concerns discussed in the preceding sections.
A risk characterization considers all information, integrated through a logical and transparent
approach, and provides a thorough discussion of the findings and related uncertainties (USEPA,
2000a). A risk characterization needs to provide a clear answer to the question, "Are there
susceptible populations and life-stages?"
Overarching guidance is needed to improve the ease and consistency of risk assessment
for susceptible populations and life-stages. Such guidance needs to provide more detail,
methods, and advice for considering the issues discussed in this section. As a part of such
efforts, EPA needs to continue developing a strategy and methodologies to address the
cumulative risks of unique or disproportionate exposures.
Continued research and the availability of new types of information (e.g., genomics,
improved early life animal toxicity testing, cultural practices awareness, demographics) will
provide an improved ability to identify and describe susceptible populations and life-stages. It is
anticipated that these advances will improve the quality of risk assessments and guidance
documents both for risk assessment in general and for susceptible groups in particular. Until
such data are available, the current practice of considering available data and following Agency
practice where data are limited is a reasonable approach. The Agency continues to strive to be
-------
Examination of EPA Risk Assessment Principles and Practices Page 49
clear and transparent in explaining the assessment of populations, uncertainties, and life-stages in
risk assessment.
3.6 What Is EPA Doing To Improve Assessment of Uncertainty and Variability?
3.6.1 How Can EPA Move Forward Toward Quantitative Characterization of
Uncertainty in Dose-Response?
In successive versions of its Cancer Guidelines, EPA expressed increasing emphasis on a
full examination of uncertainties, with the recognition that both qualitative and quantitative
approaches to uncertainty assessment are important and can (applied appropriately) help clarify
the nature of assessment findings. The use of sophisticated uncertainty tools also involves
substantial issues of science and mathematics, as well as specialized issues such as the
appropriate presentation and characterization of probabilistic estimates in the decision making
context where appropriate. We note that active research is underway in this field, and that EPA
guidelines are intended to be flexible to allow use of advances in this field as they develop. As
also discussed in various places in the guidelines, substantial quantitative uncertainty analyses
are also appropriate (and at a more advanced stage of development) for a number of important
components of cancer risk assessments, including exposure assessment, characterization of
statistical uncertainty in fits of both descriptive and biologically based dose-response models, and
dosimetry estimates using pharmacokinetic data and modeling. It is not, however, EPA's intent
to suggest that full probabilistic models of cancer risks are generally feasible at this time, or that
the role of a qualitative presentation of uncertainties should be diminished.
EPA health scientists currently conduct research that both generates necessary
quantitative data and integrates data-driven hypotheses on mode of action into biologically based
models. These models simulate key biological processes and provide quantitative predictions
that improve risk assessment, including cross-route extrapolation and animal-to-human
extrapolation. One can use these models to develop better estimates of variability in the human
population, to identify susceptible subpopulations, to help determine the applicability of
proposed modes of action, and to derive likelihoods of adverse effects in humans.
3.6.2 Will EPA Move Toward Integrated Uncertainty Analysis?
EPA recognizes the importance of uncertainty analysis and the value of having an
integrated framework for evaluating uncertainty within an overall analysis, rather than as an a
posteriori exercise that generates deterministic estimates. EPA is working toward integrated
probabilistic frameworks within risk models now being developed. Once these are implemented,
EPA will be able to conduct probabilistic analyses as part of any original analysis, though it is
recognized that probabilistic frameworks will be pointless without adequate data to insert.
-------
Page 50 Examination of EPA Risk Assessment Principles and Practices
Models that include integrated probabilistic frameworks include APEX, TRIM, and
CALENDEX.
In addition, EPA is developing a flexible modeling framework, the Multimedia Integrated
Modeling System (MEVIS), which provides an infrastructure for modeling that supports
composing, configuring, applying, and evaluating models. It can support connecting models in a
controlled manner, provide graphical user interfaces to help users configure models, and manage
the execution of complex systems of models. MIMS has been designed to support a wide range
of models. One of the benefits of using MIMS to manage model executions is that it can initiate
repetitive executions, including simulation of multiple scenarios and sensitivity, uncertainty, and
optimization studies.
MIMS's sensitivity and uncertainty capabilities are designed to work with a broad range
of models or modeling systems. The initial sensitivity capability will support local evaluation.
Uncertainty capabilities will initially include simple Monte Carlo and Latin Hypercube
Sampling. More sophisticated capabilities, such as analysis of variance for sensitivity studies,
are planned for the future. In addition to handling parameter sampling and model execution,
MIMS will provide data analysis tools to help the user interpret the results of the computations.
-------
Examination of EPA Risk Assessment Principles and Practices Page 51
4. CONSIDERING INFORMATION GAPS IN HEALTH
ASSESSMENTS: USE OF DEFAULT AND EXTRAPOLATION
ASSUMPTIONS
4.1 Default Assumptions
Given the nature of uncertainty and data gaps, it is accepted practice to use default
assumptions (sometimes simply called defaults) to address these uncertainties. A default
assumption is "the option chosen on the basis of risk assessment policy that appears to be the
best choice in the absence of data to the contrary" (NRC, 1983). TheNRC, in its review of EPA
risk assessment practices titled Science and Judgment in Risk Assessment (NRC, 1994),
supported EPA's use of defaults as a reasonable way to deal with uncertainty. That report stated
that EPA should have principles for choosing default options and for judging when and how to
depart from them. It identified a number of criteria that NRC believed ought to be taken into
account in formulating such principles, including protecting the public health, ensuring scientific
validity, minimizing serious errors in estimating risks, maximizing incentives for research,
creating an orderly and predictable process, and fostering openness and trustworthiness.
The first two sections (4.1.1 and 4.1.2) discuss default assumptions in general. The rest
of section 4,1 discusses specific examples of defaults that have been highlighted in the comments
we received.
4.1.1 How Does EPA Use Default Assumptions?
EPA's current practice is to examine all relevant and available data first when performing
a risk assessment. When the chemical- and/or site-specific data are unavailable (i.e., when there
are data gaps) or insufficient to estimate parameters or resolve paradigms, EPA uses a default
assumption in order to continue with the risk assessment. Under this practice EPA invokes
defaults only after the data are determined to be not usable at that point in the assessment this
is a different approach from choosing defaults first and then using data to depart from them. The
default assumptions are not chemical- or site-specific, but are relevant to the data gap in the risk
assessment. They are based on peer reviewed studies and extrapolation to address specific data
gaps. These defaults are based on published studies, empirical observations, extrapolation from
related observations, and/or scientific theory.
EPA's use of defaults is appropriate. These choices are well within the range of plausible
outcomes and often at specific percentiles (for variability) within that range of observation. The
aim of the risk assessment is to ensure that EPA does not underestimate risk to a chemical or
stressor; the default assumptions used in the risk assessment are used to help pursue this goal.
-------
Page 52 Examination of EPA Risk Assessment Principles and Practices
Conversely, the intent of using default assumptions is not to produce a risk assessment that
appreciably overestimates the risk being assessed,
A default is usually based on some data and/or some consensus agreement for the generic
situation pertinent to the data gap being assessed. Defaults usually have undergone peer review,
in which interested parties are involved or have the opportunity to be involved in those reviews.
For example, many of the exposure defaults we use are found in EPA's Exposure Factors
Handbook, a resource that was peer reviewed and continues to be examined and re-evaluated.
When chemical- or site-specific information becomes available and is adequate to use, our risk
assessments attempt to use those data rather than the default(s) assuming those data are
relevant for the specific risk being assessed, and the peer review supports their use.
There is always room for improvement in highlighting where these defaults are used in
actual assessments, and one should not assume that people who read an assessment will
automatically know that a default position is used (NRC, 1994). Where and when we use default
assumptions to address a data gap may not be clear to many. We need to point to the places
where the default was developed and explain why it is a reasonable assumption.
4.1.2 Are Default Assumptions Science Policy?
One of the major comments EPA received suggests that default assumptions are actually
conservative policy decisions embedded in risk assessments. While default assumptions have
data and/or scientific consensus supporting them, use of defaults to address data gaps in a risk
assessment is a science policy decision. In keeping with the EPA's goal of protecting public
health and the environment, the default assumptions are used ensure that we do not
underestimate risk to chemicals and stressors. (Again, though, they are not intended to overtly
overestimate a risk.) Defaults are ideally developed in a transparent manner, and we believe this
is the practice of the Agency. Therefore, while defaults are indeed used in risk assessments, they
are not buried (hidden) policy decisions. Recently, the GAO reflected that we use assumptions
(defaults) as an unavoidable part of risk assessment because science cannot always provide clear
answers to questions at various stages of an assessment (USGAO, 2001). Particular assumptions
are chosen through an evaluation of available scientific information or policy decisions related to
our regulatory mission and mandates. Due to the complexity of risk assessments, transparency in
choice and use of assumptions is critical in risk assessments and communication efforts.
If a default assumption is a policy position that some deem too conservative or not
conservative enough, this issue can be addressed during the peer review. Generally speaking,
EPA's current default assumptions are positions based on data or scientific consensus and
supported through peer review. Some older risk assessments may contain defaults that have not
yet been reexamined in light of new data. Thus, in the context of current data, they appear
inadequate (i.e., to some, they can be too conservative or not conservative enough). This is not
-------
Examination of EPA Risk Assessment Principles and Practices Page 53
an inherent problem with the original risk assessments, but a process problem for EPA: EPA
needs to determine the need to reexamine older assessments based on current priorities, needs,
and resources.
4.1.3 Does Any Change Seen in Animals Indicate There Will Be a Problem for
Humans?
It is generally accepted that there can be numerous changes to the recipient organism (the
animal under study) following exposure to a chemical, some of which maybe beneficial,
adaptive, early manifestations on a continuum to toxicity, overtly toxic, or several of these things
in combination. Unless there are data to indicate otherwise, a change that is considered adverse
(i.e., associated with toxicity) is assumed to indicate a problem for humans.
It is recognized that a diversity of opinion exists regarding what is "adverse" versus
"adaptive," both within EPA and in the general scientific community. At present, there is no
Agency-wide guidance from which all health assessors can draw when making a judgment about
adversity. Therefore, various experts may have differing opinions on what constitutes an adverse
effect for some changes. Moreover, as the purpose of a risk assessment is to identify risk (harm,
adverse effect, etc.), effects that appear to be adaptive, non-adverse, or beneficial may not be
mentioned.
As a further look at this issue, an "adaptive" example is used. The human body is capable
of adapting to certain toxic insults. When adaptive responses become adverse and irreversible is
not yet defined. In cases where data are not available to determine when the capacities of repair
mechanisms are exceeded and adaptive responses become toxic, health assessments are based on
any adverse response that is deemed biologically significant. As a general principle, our practice
is not to base risk assessments on adaptive, non-adverse, or beneficial events.
The IRIS database includes some examples of effects reported at the designated NOAEL
level, but their lexicological significance was questioned and not relied upon in setting
RfDs/RfCs; for example, mild nasal lesions and hyaline droplet degeneration in the nasal cavity.
These effects were not considered of sufficient toxicological significance to warrant designating
the levels that produced them as LOAELs. In these instances, the LOAELs were set at the next
highest dose level, based on more severe nasal responses.
There are also differences of opinion regarding the severity of effects and their qualitative
and quantitative relationship to more overt toxicity. To explore past Agency practice, we
performed a series of IRIS database searches on a selected example: "liver weight" as the sole
critical effect, expanded slightly to incorporate examples in which liver weight appears to be
listed as the predominant critical effect, along with other effects. The results indicate that liver
weight change has been listed as a sole critical effect in the IRIS database in seven profiles (and
-------
Page 54 Examination of EPA Risk Assessment Principles and Practices
as a predominant effect in three more; examples dating from 1987 to 1994). More in-depth
examination of these IRIS records generally indicates that at least one additional toxicity
consideration was incorporated into the Agency decision at that time. This additional toxicity
information comes in three forms: (1) listing of co-critical adverse events; (2) supporting
information on adverse effects from other studies, at lower, similar, or slightly higher doses, that
have not been reported in the IRIS critical events column; and/or (3) a clear progression of
toxicity at slightly higher doses. All three sources of additional toxicity information may be
discussed in a particular IRIS file, with the predominant information being supportive toxicity
data at similar doses in other studies. The extent to which additional toxicity information is
available and/or reported varies among the chemical assessments. This examination does not
rule out the possibility of using liver weight (or other changes) as a critical effect if it is deemed
to contribute to adversity in an assessment.
4.1.4 Are Benign (Histologically Non-Malignant) Tumors Presumed To Have
Potential To Progress to Malignant Tumors, and Are They Counted as if
They Were Malignant?
When evaluating the results of long-term bioassays in rodents for evidence of potential
carcinogenicity, we do not always presume that all benign tumor types have the potential to
progress to malignant tumors. Although the total incidence of benign tumors is considered in
evaluations of carcinogenic potential, the weight placed on evidence of an increase in a benign
tumor response is subject to consideration of the potential to progress to a malignant tumor.
For example, a uterine polyp, a benign tumor, is not considered to have the potential to
progress to a malignant tumor; in the absence of evidence of an increase in malignant tumors in
the same tissue (or in other tissues), evidence of carcinogenic potential would be considered to be
weak. In such cases, the chemical would likely be assigned the descriptor "Suggestive Evidence
of Carcinogenicity, but Not Sufficient To Assess Human Carcinogenic Potential" or "Not Likely
To Be Carcinogenic to Humans" and a quantitative risk assessment would not be recommended
(USEPA, 1999b).
On the other hand, hepatocellular adenomas are presumed to have the potential to
progress to hepatocellular carcinomas. This is because there is considerable evidence showing
that hepatocellular adenomas may progress to carcinomas in both rodents and humans. In
addition, there is no clear histological demarcation between hepatocellular adenomas and
carcinomas. Thus, when the results of rodent bioassays give evidence of the induction of both
adenomas and carcinomas, the combined incidence of the two tumor types is considered in a
weight-of-evidence evaluation of potential carcinogenicity. This approach is consistent with that
of the National Toxicology Program (NTP), which recommends combining hepatocellular
adenomas and hepatocellular carcinomas for an overall analysis of carcinogenicity (in addition to
analyzing them separately).
-------
Examination of EPA Risk Assessment Principles and Practices Page 55
4.1.5 Should We Expect Target Organ Concordance Between Effects Observed in
Animals and Those Expected in Humans?
Thus far, there is evidence that growth control mechanisms at the level of the cell are
homologous among mammals, but there is no evidence that these mechanisms are necessarily
target organ concordant. For example, Haseman and Lockhart (1993) examined a database of
379 long-term carcinogenicity studies in rats and mice to evaluate sex and species correlations in
site-specific carcinogenic responses. Within a species, most target sites showed a strong
correlation between males and females. If a chemical produced a site-specific carcinogenic effect
in female rats or mice, there was a 65% probability that it would be carcinogenic at the same site
in males. On the other hand, the interspecies correlation was lower. Approximately 36% of the
site-specific carcinogenic effects observed in one species were also observed in the other species.
EPA's cancer risk assessment guidelines recommend that site concordance of tumor
effects between animals and humans be considered in each case. Site concordance is not
assumed a priori. On the other hand, in establishing a mode of action that involves certain
processes with consequences for particular tissue sites (e.g., disruption of thyroid function,
hormonal effects leading to mammary rumors), one would expect site concordance when
evaluating the weight of evidence. We address the issue of target organ concordance on a case-
by-case basis, considering information on the mode of action. For example, based on a
significant body of data, EPA determined that male rat kidney rumors are not relevant to the
assessment of human cancer risk where tumors arise as a result of accumulation of a protein
(alpha-2u-globulin) unique to the male rat (USEPA, 199la). Similarly, the reproductive and
developmental endpoints need to be judged considering what is known about the mode of action
for an individual chemical as well as how the effects in animals may or may not be predictive of
effects in humans.
Similarly, the expectation of concordance of fetal effects would depend on knowledge
about mode of action. Most fetal effects are exquisitely dependent on time as well as dose. As
the developmental process of laboratory animals and humans differs, especially for rodents,
slight differences in the pattern of exposure or the toxicokinetics in the mother may greatly affect
the observed outcome. In this regard, it is worth noting that the standard metric of daily dosing
may have different implications for a laboratory animal than a human. Moreover, certain
reproductive endpoints that are useful for evaluating fetal effects in rodents have no human
equivalent, e.g., number and percent of live offspring per litter. With regard to fetal effects, for
example, we consider both individual effects and total affected implants. Changes in dose may
not only change the specific malformation, but also cause fetolethality rather than malformation.
Thus, concordance of endpoint may not be the best predictor of developmental effects in people.
-------
Page 56 Examination of EPA Risk Assessment Principles and Practices
4.2 Extrapolations
Some federal regulatory agencies can estimate risks based on historical data about those
risks. Thus, we have information on risks such as house fires, traffic accidents, and occupational
injuries. At EPA, we are often faced with the problem of estimating the risk from chemicals for
which we have limited or no historical data, e.g., new chemicals. Even when such a chemical
has been in use, most if not all of the available information about its toxicity comes from studies
with laboratory animals that use concentrations that are designed to produce effects. Similarly,
available human data usually involve either exposures at high concentrations for short durations
(e.g., accidents) or occupational exposures. Our concerns often involve situations where people
might be exposed to much lower concentrations of the chemical for various periods of time. We,
therefore, need to use procedures to extrapolate from responses (or lack thereof) at exposures
where data are available to exposures of interest. Frequently, we also need to extrapolate from
responses of laboratory animals to those that might reasonably be expected in people.
The rest of section 4.2 discusses how EPA extrapolates from animals to humans.
4.2.1 Why Does EPA Use Experiments With Animals To Predict Effects in
Humans?
As a general rule, we assume that toxic responses observed in laboratory animals are
indicative of toxic responses that are likely to occur in people. To predict responses in people
more accurately, we would prefer information from animals that are as similar to people as
possible. The use of other mammals, such as dogs, rats, and mice, as models for responses in
humans is based on the assumption that there are important similarities among mammals in the
way they respond to chemicals. A qualitative similarity has been established in the response of
laboratory animals and humans to carcinogenic substances. For example, one analysis of
chemicals that have been tested by the NTP for their ability to produce cancer (Huff et al, 1991)
indicates that rats and mice predict responses for each better than 80% of the time for liver
cancers. Most known human carcinogens have been shown to be positive for tumorigenicity in
well-conducted animal studies. Data are not available to determine whether the opposite is true.
Nevertheless, important differences can exist. Metabolic patterns of both activation and
deactivation along with differences in pharmacokinetics, for example, can give rise to significant
differences in sensitivity between species and within species. Unless we know the rates of the
activation-deactivation processes, for example, it is impossible to predict the differences in
response among species. Presently, there is a paucity of comparative data on metabolism for
specific chemicals and other interspecies differences that can affect toxicity. Important
differences require experiments on the toxicokinetics or toxicodynamics within each species.
We cannot be certain which results are most like those for humans in the absence of these data.
-------
Examination of EPA Risk Assessment Principles and Practices Page 57
The increasing use of mode of action information will make it clearer whether a similarity or
difference is apparent between test animals and humans.
4.2.2 How Does EPA Adjust for the Differences Between Animals and Humans?
If we have sufficient information about how the environmental agent is absorbed,
distributed, metabolized, and excreted in the laboratory animal and the human and about the
physiologies of the two species, we can develop a model (such as a PBPK model) that will allow
us to compare how differences in these factors and processes affect the biologically effective
dose to a target tissue. For some chemicals, this process has resulted in a very specific
adjustment from the animal to the human. When information is limited, we can only make
general adjustments.
Our practice is to interpret the findings of long-term rodent bioassays in conjunction with
results of subchronic studies along with toxicokinetic studies and other pertinent information, if
available. When data are appropriate, we use metabolic and toxicokinetic data to:
a) Identify and compare the relative activities of metabolic pathways in animals and
humans and at different ages.
b) Describe anticipated distribution within the body and possibly identify target
organs.
c) Identify changes in toxicokinetics and metabolic pathways with changes in dose.
d) Determine the bioavailability via different routes of exposure by analyzing uptake
processes under various exposure conditions.
This issue is addressed on a case-by-case basis where data are available. When species-
comparative data are not available, we assume concordance between the animals tested and
humans.
4.2.3 Does EPA Use the Most Sensitive Animal Results To Predict Effects in
Humans?
Use of the most sensitive species, strain, sex, age, etc., was originally justified on the
need to be health protective in the absence of data. From the 1970s through the 1990s, the most
sensitive responding species (given several data sets to choose from) were frequently selected:
very little information was available as to what was scientifically the most representative choice
for human risk prediction. Combined with UFs and other upper-bound estimates, basing cancer
and non-cancer risks on the most sensitive animal data gives reasonable assurance that the
-------
Page 58
Examination of EPA Risk Assessment Principles and Practices
potential for harm will not be underestimated, most likely even when some toxicity endpoints
have not been evaluated. Our early cancer risk assessment guidelines (USEPA, 1986a) comment
on the use of the most sensitive animal response in the absence of appropriate human studies.
The most recent draft guidelines (USEPA, 2003c) suggest a wide range of considerations based
on knowledge and uncertainty for selecting the appropriate data set(s), assuming several sets are
available.
Other federal agencies use a similar approach. According to Rhomberg (1997), the
Occupational Safety and Health Administration (OSHA) also generally uses the data set
demonstrating the highest estimated sensitivity, although it tends to "present several alternatives
together or do several analyses and present the median result." The Consumer Product Safety
Commission (CPSC) also uses similar methods, but differs in its approach to combining tumors
from different sites.
The availability of more biological knowledge or insight potentially enables the risk
assessor to make more scientifically informed choices among the available experiments with
animals for use in dose-response analysis. The recent draft cancer guidelines (USEPA, 2003c)
ask for a full display of all data options as well as the rationale for why one data set is selected
and why others are rejected. Data from humans are preferred, although they are not always
amenable to quantitative risk assessment. Alternatively, data from species that respond most like
humans should be used if information to support this selection exists. If this is not known, all of
the available data sets are considered and compared. An informed, scientific judgment is made
as to what data best represent the observed data and important biological features such as mode
of action. When faced with critical uncertainties or data gaps, however, we will favor a selection
that ensures against underestimating risk as a policy choice.
We are continuing our trend toward using relevant data before using a default assumption
to address data gaps. Use of toxicokinetic and toxicodynamic data for interspecies extrapolation
should better address this source of uncertainty. Using the available data should accommodate
some of the differences among species. For example:
a) When data on mode of action indicate that most the sensitive species/strain/sex is
not an appropriate model for humans, e.g., alpha-2u-globulin, those data are not
used for estimating human risks.
b) In some cases, use of the data that estimate the highest potential for harm can
result in the use of less-well-designed studies, a concern that should be attenuated
by increased use of the benchmark dose procedure.
c) Additionally, we have begun to investigate the utility of meta-analysis as one way
to evaluate an entire database quantitatively.
-------
Examination of EPA Risk Assessment Principles and Practices Page 59
4.2.4 Is an Additional Body Scaling Factor for Children Considered?
The draft final cancer guidelines (USEPA, 2003c) suggest that different adjustments may
need to be made for children inhaling carcinogenic substances than for adults, but the guidelines
do not specify how this should be done. For example, their language has been incorrectly
interpreted as suggesting the use of a scaling factor of body weight to the 0.75 power before
adjusting for a child's breathing rate. These draft guidelines recognize that correlations exist
"between breathing rate, respiratory tract dimensions, and body weight."
The methods that could be used for interspecies extrapolation with appropriate
adjustments for children are the subject of draft guidance (USEPA, 2003d), ongoing discussions,
and further assessment by the Agency. The supplemental guidance for early-life exposures, for
example, illustrates the importance of information regarding the mode of action of the carcinogen
on evaluating the potential risks to children. We understand the need to address this issue further
and will do so, in part, based on the comments received on the draft guidance.
4.2.5 How Are Effects Observed at High Exposures Used To Predict Responses to
Much Lower Exposures?
The role of animal tests is to provide maximum detectability of effects within the narrow
constraints of test sensitivity. This is typically achieved through the use of "high" exposures or
doses to increase the probability that a possible adverse effect will be observed. This practice is
usually used when sample sizes are too small to be sensitive to effects that are not overtly toxic
or seen in response to potent stressors. Generally, the methods used to extrapolate from high to
lower exposures differ for cancer and non-cancer effects. Both are discussed below. It should be
noted that as more mode of action information becomes available, the distinction between cancer
and non-cancer endpoints will dissipate and the dose-response evaluation is more likely to
distinguish between linear and nonlinear extrapolations from high exposures to lower exposures.
Cancer
Carcinogenic effects are typically evaluated using a long-term rodent carcinogenicity
bioassay. Current practice for these tests suggests using at least 50 animals per sex per dose
group in each of three treatment groups and in a concurrent control group, usually for 18 to 24
months, depending on the rodent species tested. One group is exposed to a high dose, often the
maximum tolerated dose, or MTD (discussed further below). Two groups are given lower doses
of the test compound. The high dose in long-term studies is generally selected to provide the
maximum ability to detect treatment-related carcinogenic effects while not compromising the
outcome of the study through excessive toxicity or inducing inappropriate toxicokinetics (e.g.,
overwhelming absorption or detoxification mechanisms). The purpose of the two or more lower
-------
Page 60 Examination of EPA Risk Assessment Principles and Practices
doses is to provide some information on the shape of the dose-response curve, including possible
subtle, precursor, and/or early events that can provide a more meaningful dose-response
relationship relative to anticipated human exposures. Similar protocols have been and continue
to be used by many laboratories worldwide.
Because of biological variability and the relative low sensitivity of most bioassay designs,
positive evidence at the high dose is considered a rebuttable presumption of carcinogenicity.
These findings are not altered by negative data at lower exposures unless there are supporting
studies that provide a biological basis to conclude that the mode of action responsible for the
high-dose effects is not relevant at lower doses and therefore for lower levels of exposure
experienced by humans.
There is often a need to estimate risks for exposures significantly below exposures for
which we have data. Two methods are used for this process. For mutagenic chemicals and those
for which insufficient data exist to determine a mode of action, a point of departure (POD) at the
low end of the exposures for which we have data is selected and a straight line is drawn from (the
lower confidence limit on) that dose to zero. A nonlinear approach is selected when there are
sufficient data to ascertain the mode of action and conclude that it is not linear at low doses and
the agent does not demonstrate mutagenic or other activity consistent with linearity at low doses.
The use of low-dose linear assumptions for mutagenic carcinogens has been the peer
reviewed, common practice for decades for many federal regulatory agencies including EPA.
Modifications for non-mutagenic carcinogens are currently under review.
Non-Cancer
For non-cancer effects, we routinely use the absence of an adverse effect (a NOAEL) in a
relevant animal study as the basis for estimating a negligible risk exposure level for humans.
Adjustments are made to the NOAEL to account for limitations in knowledge, i.e., uncertainty,
or limitations inherent in the test system, e.g., the use of relatively small numbers of genetically
relatively homogeneous animals. Similarly, mode of action, physiological differences, or other
factors have been used to improve the accuracy of estimation of the RfD or RfC for non-cancer
endpoints.
The Agency has also been using and is continuing to refine the BMD methodology
(USEPA, 2000g) for estimating non-cancer risks. In this procedure, the POD is derived from the
dose-response curve, and further adjustments analogous to those for the NOAEL may be
necessary to account for similar limitations of the data.
The EPA RfD/RfC review (USEPA, 2002c) delineates several options that have been
used as the Agency developed more accurate methods for evaluating these parameters from
-------
Examination of EPA Risk Assessment Principles and Practices Page 61
responses in animals. It also provides a series of examples, as well as lists of recommendations
that address the predictivity of results from animal studies.
4.2.6 Why Is the Maximum Tolerated Dose Used?
Because tests for cancer are long and expensive, they are designed to increase the
likelihood of detecting a positive response with a limited number of animals. The highest dose
that does not produce other overt toxicity (i.e., the MTD) and fractions of that dose are selected
for the dosing regimen. The MTD is usually determined from a shorter, e.g., 90-day, experiment
with the same animal species and chemical.
The use of such high doses in animal cancer tests has been the subject of considerable
debate. Limitations inherent in using the MTD approach and suggestions for improvement have
been controversial since its use became standard. Testing at the MTD has been criticized
because it may cause damage which may lead to cellular proliferation, increased mitosis, and
eventually carcinogenicity. It has been argued that biochemical and physiological distortions
occurring at high doses may lead to toxicity-induced carcinogenic effects that might not be
expected at lower doses (Ames and Gold, 1990). Haseman (1985) has shown that more than
two-thirds of the carcinogenic effects detected in feeding studies conducted under the NTP would
have been missed if the highest dose had been restricted to doses below the MTD, if the
chemicals were, in fact, animal carcinogens. Additionally, Bickis and Krewski (1989) found the
upper confidence limit on the linear term (qt*) from the linearized, multi-stage model based on
263 data sets to be highly correlated with the maximum dose tested.
The draft final cancer guidelines (USEPA, 2003c) cautions the assessor in the use of
results from the MTD and recommends that this issue be addressed on a case-by-case basis in the
context of other study results and other lines of evidence. The guidelines state that the results of
such studies would not be considered suitable for dose-response extrapolation if it is determined
that the mode of action underlying the tumorigenic response at high doses is not operative at
lower doses. For example, EPA developed a science policy position for agents that cause thyroid
follicular cell tumors as a result of high-dose effects that are not present at low doses (USEPA,
1998c). In addition, the recent chloroform risk assessment was predicated on the recognition of a
high-dose effect leading to cell proliferation and then carcinogenicity. The MCL was therefore
developed based on an assumption of nonlinearity for carcinogenicity.
4.2.7 Does the Use of a Maximum Tolerated Dose Affect Cancer Potency
Estimates?
The draft final cancer guidelines (USEPA, 2003c) include an extensive discussion of the
role of dose selection in the conduct and interpretation of data from cancer tests carried out in
animals. Section 2,2.2.1 deals with "Long-Term Carcinogenicity Studies," and section 2.2.2.1.1,
-------
Page 62 Examination of EPA Risk Assessment Principles and Practices
"Dosing Issues," is in essence a discussion of the role of the MTD in cancer testing in animals
and interpretation of data from such tests.
Section 2.2.2.1.1 makes the following critical points, among others, to justify the use of
MTDs: "...failure to reach a sufficient dose reduces the sensitivity of the studies." Because
animal tests are conducted with a relatively small number of animals, dosing at the MTD is
necessary to provide the best chance that cancer will be seen in test animals if a chemical is
carcinogenic. Further, the document recognizes that "...overt toxicity or inappropriate
toxicokinetics due to excessively high doses may result in tumor effects that are secondary to the
toxicity rather than directly attributable to the agent."
The document also stresses the importance of establishing that the MTD has not been
exceeded. Criteria for determining this include whether "...an adequate high dose would
generally be one that produces some toxic effects without unduly affecting mortality from effects
other than cancer or producing significant adverse effects on the nutrition and health of the test
animal." Thus, our practice is that the experimental conditions are scrutinized to ensure that the
MTD has been reached but not exceeded. We further believe that effects seen at the MTD may
be appropriate for use in risk assessment when the data have been critically evaluated.
4.2.8 Are the Cancer Risks Estimated by EPA the Expected Incidence of Cancer
for a Given Exposure?
Our risk estimates are designed to ensure that risks are not underestimated, which means
that a risk estimate is the upper bound on the estimated risk. In past guidelines (USEPA, 1986a),
we have explicitly stated that the true cancer potency "could be as low as zero." Since the CSF is
generally multiplied by the exposure estimate to generate an estimated risk, a CSF of zero would
result in a risk estimate of zero.
If other procedures, e.g., odds ratios from epidemiologic studies, result in risk estimates
that should be interpreted differently, these should be clearly delineated. In particular, if the
method either requires or results in maximum likelihood estimates (MLEs) rather than bounding
estimates being used, this should be clearly stated.
According to Rhomberg (1997), OSHA uses the same methodology as EPA, but it
presents and focuses on the MLE dose-response curve. CPSC also uses the same model as EPA,
but modifies its constraints so that, with the limited data usually available, the MLE of the dose-
response function is linear at low doses. Rhomberg asserts that this procedure also results in an
upper-bound estimate, although no proof is given in the cited document.
-------
Examination of EPA Risk Assessment Principles and Practices Page 63
4.2.9 How Does EPA Evaluate Incidence Versus Mortality for Cancer Rates?
Incidence of a disease is a measure of disease occurrence. Incidence rate measures the
frequency of a given disease in a population. A concept closely related to incidence rate is that of
the risk or probability of developing a disease over a period of time. The incidence rate for a
disease is also referred to as the hazard rate for developing that disease (Kalbfleisch and Prentice,
1980). Mortality rate, on the other hand, measures the frequency of death rate in a population
from a given disease. Both incidence and mortality rates are measures of risk and have been used
to calculate risk in different risk assessments. The appropriate measure depends upon the
particular risk of interest and the type of data available. It is important that this be generally
made clear in risk assessments.
Incidence rates give a more accurate indication of risk of a disease in a population.
However, obtaining information about incidence rates in a population is usually difficult. One
place to look for incidence rates of site-specific cancers in a population is the Surveillance,
Epidemiology, and End Results (SEER) data. Obtaining mortality data and computing rates is
comparatively easy, because every death is recorded. Thus, over the years, the most commonly
used data for risk assessments have been the mortality rates. An adjustment between mortality
and incidence can be carried out using information from databases such as SEER.
In diseases for which survival is poor and mortality is 95% to 100% within a short period
of time (e.g., lung cancer, pancreatic cancer), the mortality rates are good surrogates for incidence
rates. A risk assessment based on these mortality rates provides a good estimation of true risk in
a given population. In diseases for which survival is higher and mortality is lower (e.g., non-
melanoma skin cancer, urinary bladder cancer), the mortality rates are poor surrogates for
incidence rates. A risk assessment based on these mortality rates therefore provides an
underestimation of risk in a given population.
EPA's draft final Guidelines for Carcinogen Risk Assessment (USEPA, 2003c)
recognize this issue. Section 3.2.1 of the guidelines states that "Analysis of epidemiologic
studies depends on the type of study and quality of data, particularly the availability of
quantitative measures of exposure. The objective is a dose-response curve that estimates the
incidence of cancer attributable to exposure to the agent." It further notes that "The analysis is
tailored to the nature of each study, with due consideration of the consequences of study design.
For example, many studies collect information from death certificates, which leads to estimates
of mortality rather than incidence. Because survival rates vary for different cancers, the analysis
can be improved by adjusting mortality figures to reflect the relationship between incidence and
mortality." This was the approach used in EPA's assessment of 1,3-butadiene, found in IRIS
(USEPA, 2002e). A similar approach was used in the NRC's assessment of the cancer risk from
arsenic ingestion (NRC, 2001).
-------
Page 64
Examination of EPA Risk Assessment Principles and Practices
It should be noted that when the CSF is derived from animal data, the extrapolation uses
the tumor incidence data.
4.2.10 If Data Are Only Available for One Route of Exposure, Does EPA Evaluate
Other Routes of Exposure?
Our general assumption is that chemicals that cause internal toxicity by one route of
exposure will do so by any other route if that route also produces an internal dose. This position
is based on the scientific principle that the internal dose to the tissue of interest is the ultimate
determinant of toxicity. We advocate development and use of agent-specific pharmacokinetic
information to inform this position.
In its 1994 report, the NRC (NRC, 1994) states that "... the target-site dose is the ultimate
determinant of risk ..." Thus, the manner or route of systemic exposure is secondary (except as it
affects internal doses, e.g., by first-pass metabolism) to the internal dose. Others in government
also reflect this fundamental principle (e.g., Corish and Fitzpatrick, 2002; ACGIH, 1991).
Some chemicals have demonstrated the same (or very close to the same) target organs in
response to chemical exposure, regardless of route. For example, for cumene, a kidney effect is
observed after oral exposure, implying an internal dose of cumene or its metabolites to the kidney
via the oral route; a kidney effect is also seen after an inhalation exposure to cumene, implying
an internal dose of cumene or its metabolites to the kidney via the inhalation route. This
demonstrates a target organ dosing independent of route of administration and a fundamental role
for establishing an internal dose leading to toxicity. Three other examples from IRIS of chemical
assessments having a common critical target organ via inhalation (RfC) and oral (RfD) routes of
administration are shown below, in table 4-1.
Table 4-1: Route Comparison
Agent
Cumene (CAS # 98-82-8)
1,1-dichloroethylene (CAS # 75-35-4)
Vinyl chloride (CAS # 75-01-4)
Target Organ Affected via the
Oral Route
Kidney
Liver
Liver
Inhalation Route
Kidney
Liver
Liver
Several examples from studies measuring kinetic endpoints of compounds administered
by various routes of exposure found results to be mostly parallel across routes. This indicates
similar kinetic processes regardless of route.
-------
Examination of EPA Risk Assessment Principles and Practices
Page 65
a) After exposing rats to benzo(a)pyrene via either oral, inhalation, or intravenous
routes, Rarnesh et al. (2002) observed data indicating metabolism to all principal
metabolites across all routes.
b) Kaneko et al. (1995) reported metabolism in rats of xylene by inhalation, oral, and
intraperitoneal (ip) routes of exposure, with a significant "first-pass" effect via the
oral route. This latter effect was observed only when the liver enzymes were
altered by compounds known to induce activating enzymes.
c) After administration of alkyl ketones to rats via oral or inhalation route, Duguay
and Plaa (1995) observed dose-related increases in parent and metabolite in all
tissues examined (plasma, liver, and lung).
d) Gospe and Albayati (1994) demonstrated that oral administration of toluene
produces blood toluene concentrations that can simulate blood levels achieved
after inhalation exposures of this agent.
e) Timchalket al. (1991) observed that radiolabeled dichloropropane administered to
rats was readily absorbed, similarly metabolized, and excreted after both oral and
inhalation exposures.
Studies measuring both toxicokinetics and toxicity administered by various routes of
exposure found, in general, parallel results regardless of route.
a) In a study measuring toxic in addition to kinetic endpoints, Gansewendt et al.
(1991) reported that both oral and inhalation methyl bromide administration in
rats resulted in the same levels and species of DNA adducts, with the highest
levels found in the stomach and forestomach for both routes.
b) After administering rats hepatotoxins that were either poorly (carbon
tetrachloride) or highly (chloroform) metabolized, Wang et al. (1997) observed
and reported that toxicodynamics were parallel and toxicity was present for both
compounds via oral, inhalation, or ip routes.
c) McEuen et al. (1995) noted that oral and ip routes of administration of
dinitrobenzene produced significantly higher levels of active metabolite via the
oral route but only subtle differences in target organ (testes) toxicity in rats.
d) Using a PBPK model for cumene to examine measures of dose associated with
renal toxicity across an oral and an inhalation study in rats, Foureman and Clewell
-------
Page 66
Examination of EPA Risk Assessment Principles and Practices
(1999) noted that measures of dose generated from either study were associated
with the toxicity in a dose-related manner, regardless of route.
What Data May Set Aside the Route-to-Route Extrapolation ?
The draft final cancer guidelines (USEPA, 2003c) discuss some evidence that could
modify the application of this extrapolation, stating that an exception would occur when there is
convincing toxicokinetic evidence that absorption for a given chemical does not occur by another
route of interest. The basis for this exception is establishing the absence of an internal dose such
that the extrapolation may not be applied and such a chemical could be characterized as not being
toxic via this particular route of administration.
In the absence of data to the contrary, the route-to-route extrapolation assumption used is
that if a chemical is absorbed via one route, it will also be absorbed by all other routes.
Demonstration of any degree of uptake for each of the routes of interest is sufficient to allow the
qualitative judgment to apply the route-to-route extrapolation. The level of information
necessary to set aside the extrapolation (i.e., "contrary" evidence ) may largely be a function of
the individual stressor being tested. The similarities of processes and structures involved in
absorptive processes across biological systems may be regarded as an unstated principle that
underlies the extrapolation. Absorption obligates passing through a tissue and, fundamentally,
tissues have more similarities than differences. The presumption is that if the stressor can
penetrate tissue via one route, it can penetrate the tissue if it arrives there via another route. Thus
it is anticipated that most agents are absorbed to some degree into biological systems regardless
of route (e.g., see results summarized in Owen, 1990).
Chemicals causing point-of-entry toxicity are often very reactive, and thus are less likely
than other chemicals to produce an internal dose. Chemicals that do not establish an internal
dose by a given route would not be presumed to give internal toxicity by that route. Evidence of
robust point-of-entry toxicity by an agent may thus be considered contrary evidence for the
contention that toxicity by one route of exposure by that agent is presumed to do so by another
route. IRIS lists assessments for a number of highly reactive agents that elicit only clear and
robust portal-of-entry effects in the context of this issue, contrary evidence. Nearly all of
these assessments explicitly consider that these agents would not be anticipated to elicit internal
toxicity and discount the need for information on systemic toxicity. This discounting is done by
decreasing or even eliminating the "database" UF typically evoked at a value of 10 when studies
such as developmental and reproductive toxicity (toxicities typically associated with an internal
dose) are absent. Table 4-2 gives some examples. The practice of omitting the database UF
when portal-of-entry effects occur may not be followed in some older IRIS assessments that have
not yet been updated. For example, a database UF of 3 is applied specifically for missing
developmental data in the assessment of one other of the isocyanates, TDI (CAS # 26471-62-5).
-------
Examination of EPA Risk Assessment Principles and Practices
Page 67
A database UF of 10 is also applied in the RfC for 2-chloroacetophenone (CAS # 532-27-4), a
highly reactive tearing agent.
Table 4-2: Point-of-Entry Toxicity Examples
Agent (CAS#)
Acrolein( 107-02-8)
Ammonia (7664-41-7)
Epoxybutane
(106-88-7)
HC1 (7647-0 1-0)
HDI (822-06-0)
H3PO4 (7664-38-2)
MDI (101-68-8)
Methacrylate (80-62-6)
Value"
UF = 1
and Commentary of "Database" UF (Online IRIS)
, not applied due to lack of systemic distribution (2002)
UF = not greater than 3, no significant distribution likely to occur
(1991)
UF = 1
UF = 1
UF = 3
(1994)
UF=1
UF = 3
UF = 1
, extrarespiratory circulation of EBU minimal (1992)
, expected portal-of-entry effect, UF not applied (1995)
, unlikely that HDI is distributed in significant amounts
, no UF to be applied (1995)
(1998)
, reactive, potential for systemic effects is remote (1998)
a This UF has a maximum value of 10.
4.3 Use of Uncertainty Factors
4.3.1 How Are NOAELs or LOAELs Used for Extrapolations?
When developing a non-cancer reference value (i.e., a RfD or RfC) for a chemical
substance, EPA surveys the scientific literature and selects a critical study and a critical effect to
serve as the point of departure for the assessment. The critical effect is defined as the first
adverse effect, or its known precursor, that occurs in the most sensitive species as the dose rate of
an agent increases (USEPA, 2002c). Such a study, whether an occupational human study, a
deliberately dosed animal study, or some other study, typically involves exposure at a range of
doses. The highest exposure level at which there are no statistically or biologically significant
increases in the frequency or severity of adverse effects between the exposed population and its
appropriate control is as mentioned above called the "no-observed-adverse-effect level"
(NOAEL). When a NOAEL can be identified in a critical study, it becomes the basis of the
reference value derivation. The NOAEL is divided by appropriate UFs (e.g., for intraspecies
-------
Page 68 Examination of EPA Risk Assessment Principles and Practices
variation or study duration) to derive the final reference value. If NOAEL cannot be identified,
then a "lowest-observed-adverse-effect level" (LOAEL) also mentioned above is identified
instead. A LOAEL is the lowest exposure level at which there are biologically significant
increases (with or without statistical significance) in frequency or severity of adverse effects
between the exposed population and its appropriate control group. The NOAEL is generally
presumed to lie between zero and the LOAEL, so an UF (generally 10 but sometimes 3 or 1) is
applied to the LOAEL to derive a nominal NOAEL. Other factors are then applied to derive the
reference value. More recently, some assessments have used a HMD approach instead of the
traditional NOAEL/LOAEL approach (USEPA, 2000g).
For each non-cancer reference value (i.e., RfC or RfD) for a chemical substance in IRIS,
the following table shows whether that value was based on a NOAEL, a LOAEL, or a benchmark
dose lower confidence level. The table clearly shows that the majority of non-cancer reference
values in IRIS are based on NOAELs.
Table 4-3: Effect Levels Used To Derive Non-Cancer Reference Values in IRIS
NOAEL
LOAEL
BMD
Total
RfCs
35
21
13
69
RfDs
294
51
10
355
4.3.2 What Uncertainty Factors Does EPA Use To Reduce the Experimental
NOAEL in Health Assessments?
The Risk Assessment Forum RfD/RfC Technical Panel report (USEPA, 2002c) defines
an uncertainty factor as:
one of several, generally 10-fold, default factors used in operationally deriving the RfD and the RfC from
animal experimental data. The factors are intended to account for (1) variation in sensitivity among the
members of the human population; (2) the uncertainty in extrapolating animal data to humans; (3) the
uncertainty in extrapolating from data obtained in a study with less-than-lifetime exposure to lifetime
exposure; (4) the uncertainty in extrapolating from a LOAEL rather than from a NOAEL; and (5) the
uncertainty associated with extrapolation when the database is incomplete. The exact value of the UFs
chosen depend on the quality of the studies available, the extent of the database, and scientific judgment.
Some investigators evaluated the accuracy and limitations of allocating a value of 10 for
each area of uncertainty (Dourson and Stara, 1983; Woutersen et al, 1984; Kadry et al, 1995;
-------
Examination of EPA Risk Assessment Principles and Practices Page 69
Dourson et al., 1996). For example, the subchronic-to-chronic UF, some conclude, is probably
closer to a two- to threefold difference of uncertainty (95% of the time) and a UF of 10 should be
considered as a loose upper-bound estimate of the overall uncertainty. Others (see Dourson and
Stara, 1983, for survey of arguments representing various points of view) conclude that a 10-fold
interspecies UF could account for all animal to human differences and likewise for interhuman
variability. There are opposing arguments that EPA's application of UFs may not be
conservative enough (Dourson and Stara, 1983).
EPA applies UFs in health assessments based on the available data and the scientific
judgment of EPA risk assessors and internal and external peer reviewers. In cases where
chemical-specific data are lacking, each UF is typically no greater than 10. For example, the
majority of IRIS Toxicological Reviews provide justifications for the individual UFs selected for
a particular chemical substance. The choices of critical health endpoint, principal study, and UFs
undergo rigorous internal and independent, external scientific peer review before being presented
in the IRIS database.
For several years, EPA has used a more qualitative approach to modify the usual 10-fold
default values. For example, in deriving inhalation RfC values, the interspecies variability UF of
10 is used in the absence of data, where the distribution is assumed to be log-normally
distributed. While EPA has not yet established guidance for the use of chemical-specific data for
derived UFs, the reference concentration methodology guidance (USEPA, 1994a) provides
opportunities for using data-derived interspecies UFs by subdividing the factor of 10 to allow for
separate evaluations of toxicokinetics and toxicodynamics. The advantage to such subdivision is
a default UF of 10 for interspecies variability that can now be reduced to 3 when animal data are
dosimetrically adjusted to account for toxicokinetics (USEPA, 2002c).
4.3.3 Does EPA Consider the Effects of Combining Several UFs?
In its review of the RfD/RfC practice in the Agency, the RfD/RfC Technical Panel
reviewed the application of UFs in health assessments (USEPA, 2002c). The Panel recognized
the potential for overlap in the individual UFs and concluded that the application of five UFs of
10 for the RfD/RfC is inappropriate. Therefore, due the uncertainty inherent in values when so
many UFs are applied, the Panel recommended that no reference value (RfD/RfC) for any
particular chemical substance be derived if the composite UF is greater than 3,000. It further
recommended avoiding the derivation of a reference value that involves application of the full
10-fold in four or more areas of extrapolation. The report also recommended to discontinue the
use of the modifying factor as a UF.
-------
Page 70
Examination of EPA Risk Assessment Principles and Practices
The two graphs below show EPA's historical practice for combining UFs to derive a RfD
or RfC. As can be seen, when reference values were originally being generated, some
combinations of UFs reached as high as 10,000. As we gained greater and greater experience
and obtained more usable data, the trend has shown a decrease in the combination of UFs to such
high levels. The more recent RfDs and RfCs are more in line with the recommendation from the
Technical Panel: no derived reference value exceeds 3,000.
Figure 4-1: Use of Uncertainty Factors Over Time (to Set RfDs and RfCs)
RD UFs overtime
964
1969
1994
Year
1999
2D04
-------
Examination of EPA Risk Assessment Principles and Practices
Page 71
RfCUFs established,between 1989..and'.2GQ3
10000
1989
4.3.4 Does EPA Use UFs Derived for Specific Chemicals?
With respect to departing from the default UFs with data-derived or chemical-specific
adjustment factors, the RfD/RfC Technical Panel indicated that guidance on how to use
chemical-specific adjustment factors in risk assessment will be developed. Toward that end,
EPA has recently begun to consider data-derived approaches for chemical-specific adjustment
factors in risk assessments.
4.4 Weight of Evidence
4.4.1 How Does EPA Consider Weight of Evidence?
Risk assessment involves consideration of the weight of evidence provided by all
available scientific data. In other words, "weight of evidence evaluation is a collective
evaluation of all pertinent information so that the full impact of biological plausibility and
coherence is adequately considered" (USEPA, 1999b). Judgment on the weight of evidence
involves consideration of the quality and adequacy of data and consistency of responses induced
by the stressor. The weight-of-evidence judgment requires combined input of relevant
disciplines: toxicology, biology, chemistry, epidemiology, statistics, etc. Initial views of the
database may change significantly when other data are brought into consideration. For example,
-------
Page 72 Examination of EPA Risk Assessment Principles and Practices
the impact of a positive animal carcinogenicity study may be diminished by high-quality negative
studies; or a weak association in human epidemiologic studies maybe bolstered by consideration
of other key data from animal or other assays. Generally, no single study, whether positive or
negative, drives the overall weight-of-evidence judgment. And study findings are not scored by
any mathematical algorithm; rather, they are based on professional scientific judgment.
4.4.2 Does EPA Give More Weight to Positive Animal Studies Than to Negative
Animal Studies?
EPA evaluates all data, including both "positive" and "negative" studies, in assessing the
weight of evidence for toxicity of a chemical. Assessment decisions are, by nature, subject to
professional judgment and require input from experts in a variety of disciplines including
toxicology, pathology, and statistics. Generally, EPA does consider positive animal studies even
in the presence of negative studies. Negative studies do not "outweigh" positive studies, though,
unless they present a robust, compelling outcome relative to the positive studies.
To examine the comment that EPA always gives more weight to positive animal studies
than to negative studies, 108 chemicals classified as Class D (not classifiable as to human
carcinogenicity) in IRIS were searched for evidence as to whether they included any positive
animal studies. Seventeen of those substances were found to have statistically significant
positive animal data, but it was the professional judgment of those performing the assessment
and those who scientifically peer reviewed it that these substances were indeed nonclassifiable as
to human carcinogenicity. Thus, we do not automatically translate positive animal
carcinogenicity data into classification as a human carcinogen. Rather, it is EPA's practice to
base decisions on the best available science.
4.4.3 Does EPA Sometimes Weight Animal Data More Than Human Data?
EPA prefers high-quality human studies over animal studies because they provide the
most relevant kind of information for human hazard identification. In the absence of usable
human data, the default assumption is that positive effects in animal cancer studies indicate
carcinogenic potential in humans. Risk analysis takes all studies into account, whether they show
positive associations, null results, or even protective effects. In weighing positive versus null
studies, those judged to be of high quality are given more weight than those judged
methodologically less sound, and possible reasons for inconsistent results should be sought. Null
results from a single epidemiologic study cannot prove the absence of health effects because of
the limitations that may be associated with the study inadequate statistical power, inadequate
design, imprecise estimates, confounding factors, and others. However, null results from a well-
designed and well-conducted epidemiologic study that contains usable exposure data can help
define upper limits for the estimated dose of concern for human exposure if the overall weight of
evidence indicates the agent is a potential human carcinogen.
-------
Examination of EPA Risk Assessment Principles and Practices Page 73
EPA tries to use human studies whenever feasible and appropriate. As the NRC pointed
out (NRC, 1994), "Epidetniologic studies clearly provide the most relevant kind of information
for hazard identification, simply because they involve observations of human beings, not
laboratory animals." As a result, human data form the basis of a number of highly visible
assessments such as those of arsenic, chromium, nickel, benzene, and vinyl chloride. It is true,
however, that we have relied more heavily on animal data for some assessments even when
human data were available. The case of epichlorohydrin is an example of such an instance.
Despite the fact that two retrospective cohort mortality studies of exposed workers found no
evidence of an increased cancer risk, epichlorohydrin is classified as a "probable human
carcinogen." The apparent discounting of the human data is explained as follows:
The first study is inadequate for valid carcinogenicity assessment because of low exposure, short exposure
duration, short study period, and the young age of the cohort. The second study suffers from some of the
same limitations, as well as that of a small cohort size with at least 10 years exposure (274 individuals).
While pointing out the benefits of epidemiologic evidence, the NRC noted that the
obvious and substantial advantage (of human studies) is offset to various degrees by the
difficulties associated with obtaining and interpreting epidemiologic information (NRC, 1994).
As the Office of Science and Technology Policy indicated (OSTP, 1985), the epidemiologic
method (in the assessment of cancer risk) is often hampered by the long latent period between
exposure to a carcinogenic agent and the development of cancer, by the inability to control for
the confounding influences of unknown risk factors, by problems in assessing specific agents
when the human exposures are to mixtures, by the frequent absence of appropriate groups for
study, and by a variety of difficulties associated with accurate and unbiased historical exposure
assessment or disease ascertainment. In short, the Agency, while recognizing the clear
advantages of epidemiologic studies, also recognizes the limitations of these studies; and it is
these limitations which sometimes constrain the use of human data in an assessment.
For some comparisons of the use of animal versus human studies in risk assessment
decisions within the IRIS database, please see the tables below.
Animal Versus Human Studies Tables
All substances in the IRIS database were searched for the keyword "human," which
provided 541 matches (essentially the whole database). From these matches, the first 60 were
examined to identify compounds for which human data are available, yet animal studies serve as
the basis for assigning RfCs, RfDs, or cancer risk estimates.
Within the first 60, there were 15 with RfC listings. Of those 15, 8 list specific human
data in the inhalation exposure section. Four of those eight use human data as the principal study
guiding determination of the RfC, and four did not. Table 4-4 outlines the explanations provided
for why the available human data were not a principal study for the latter four substances. The
-------
Page 74
Examination of EPA Risk Assessment Principles and Practices
table also presents brief statements of positive and negative results from listed human studies and
interspecies UFs.
Among the first 60, there were 42 with RfD listings. Of these 42, nine list specific human
data in the oral exposure section. Five use the human data as the principal study in guiding
determination of the RfD; four do not. Table 4-5 outlines the explanations provided for these
four studies as to why they were not used to set the RfD.
There were 27 among the first 60 with classifications of carcinogenicity. Of these 27, 10
identify human data in the assessment. Four of these assessments use human data to make the
classification of carcinogenicity; six do not. Table 4-6 provides the explanations as to why, for
the latter six substances, human data were not used to make the carcinogenic classification.
Table 4-4: Explanations of Why Human Results Were Not Used To Estimate the RfC for Four
Substances in the IRIS Database8
Chemical
Human
Data (HD)
Listed
HD =
Principal
Study
Explanation
Bromomethane
Yes
No
Several studies have been conducted on the longer-term
effects of occupational exposure bromomethane. None
of these studies can serve as the basis for the
derivation of an RfC for bromomethane because of
concurrent exposures to other chemicals, Inadequate
quantitation of exposure levels and/or durations, and
other deficits in study design.
[UF] of 3 for interspecies extrapolation because
dosimetric adjustments have been applied.
Positive human occupational study: neurological
effects (Anger et al,, 1986).
Four cases of reported occupational toxicity (Herzstein
and Cullen, 1990).
-------
Examination of EPA Risk Assessment Principles and Practices
Page 75
Dichlorvos
Yes
No
Justification not provided. A UF of 3 is used for
interspecies extrapolation.
Limited reporting of mixed results in study of
residential fumigators and residents (Gold et al.,
1984).
Plasma cholinesterase activity and vitamin A decrease
in five of six pesticide manufacturing plant workers
(Ember et al., 1972).
Case studies of poisonings report clinical and self-
reported symptoms (Low et al., 1980; Reeves et al,,
1981).
Decrease in plasma cholinesterase activity, but no
significant difference in reported clinical symptoms
between test and control subjects in volunteer human
studies (Rider et al., 1967; Slomka and Hine, 1981).
-------
Page 76
Examination of EPA Risk Assessment Principles and Practices
Table 4-4: Explanations of Why Human Results Were Not Used To Estimate the RfC for Four
Substances in the IRIS Database"
Chemical
Human
Data (HD)
Listed
HD =
Principal
Study
Explanation
Acrylonitrile
Yes
No
Justification not provided.
An uncertainty factor of 3 for interspecies variability is
used because the use of the dosimetric adjustments
account for part of this area of uncertainty.
No excess mortality for non-cancer endpoints in human
occupational epi studies (O'Berg, 1980; Ott et al.,
1980; Werner and Carter, 1981; O'Berg et al., 1985;
Chenetal, 1987).
Self-reported and clinical symptoms apparent in
occupational study (Wilson et al., 1948).
No effects in human volunteer study (Jakubowski et al.,
1987).
Propylene gtycol
monomethyl ether
(PGME)
Yes
No
Comment on use of human data; A human exposure study
(Stewart et al., 1970) indicates that the RfC should
prevent irritation. Additional justification not provided.
[An uncertainty] factor of 3 is used for interspecies
extrapolation given the dosimetric adjustment and the
zero-order pharmacokinetic rate of elimination of PGME
that is suggestive of an adaptive metabolic response
(Morgott and Nolan, 1987).
Human volunteer study was positive for self-reported
symptoms but negative for clinical effects (Stewart et
al, 1970),
a These four substances were identified from the first 60 matches in the IRIS data file for which the word "human"
was identified. RfCs were estimated for 15 of these 60 substances. Human data relevant to the RfC were
available for 8 of the 15 substances. The human data were used as the principal study for the RfC for four of
those eight substances. This table describes the reasoning provided as to why the human data were not used in the
other four instances. Interspecies UFs are listed. Results of human studies are included in italics.
-------
Examination of EPA Risk Assessment Principles and Practices
Page 77
Table 4-5: Explanations of Why Human Results Were Not Used To Estimate the RfD for Four
Substances in the IRIS Database3
Chemical
Chromium
(VI)
Human
Data (HD)
Listed
Yes
Principal
Study
No
Explanation
In 1965, a study of 155 subjects exposed to drinking water at
concentrations of approximately 20 mg/L was conducted outside
Jinzhou, China ... Precise exposure concentrations, exposure
durations, and confounding factors were not discussed, and
this study does not provide a NOAEL for the observed effects.
However, the study suggests that gastrointestinal effects may occur
in humans following exposures to hexavalent chromium at levels
of 20 ppm in drinking water (Zhang and Li, 1987).
[UF] two 10-fold decreases in dose to account for both the
expected interhuman and interspecies variability in the toxicity of
the chemical in lieu of specific data.
Positive drinking water study (Zhang and Li, 1987).
Studies of positive allergic response (Bruynzeel et al, 1988;
Polak, 1983; Cronin, 1980; Hunter, 1974).
-------
Page 78
Examination of EPA Risk Assessment Principles and Practices
Table 4-5: Explanations of Why Human Results Were Not Used To Estimate the RfD for Four
Substances in the IRIS Database"
Chemical
Human
Data (HD)
Listed
HD =
Principal
Study
Explanation
Uranium,
soluble salts
Yes
No
Uranium is a classical nephrotoxic. The toxicity of this chemical
to humans has been of interest since the 1800's when uranium was
used as a homeopathic cure for diabetes mellitus (Hodge, 1973).
These early reports demonstrate the susceptibility of humans to the
nephrotoxicity of ingested uranium, but provide inadequate basis
for estimating the threshold dose for toxic effects.
[UF of] 10 for both intraspecies and interspecies variability to the
toxicity of the chemical in lieu of specific data,
Reports of nephrotoxicity with therapeutic use (Hodge, 1973).
Mixed results from experiments on uranium excretion and
toxicity (Hursh and Spoor, 1973; Lussenhop et al, 1958; Hursh
etal, 1969).
Naphthalene
Yes
No
Humans exposed via inhalation, combined inhalation and dermal
exposure, and combined inhalation and oral exposure have
developed hemolytic anemia. Hemolytic anemia is characterized
by findings of lowered hemoglobin, hematocrit, and erythrocyte
values, elevated reticulocyte counts, Heinz bodies, elevated serum
bilirubin, and fragmentation of erythrocytes. In severe cases, the
hemolytic anemia was accompanied by jaundice, high serum levels
of bilirubin, cyanosis, and kernicterus with pronounced
neurological signs. Neither oral nor inhalation exposure levels
were available in human studies reporting anemia (Melzer-
Lange and Walsh-Kelly, 1989; Owa, 1989; Owaetal., 1993).
[UF of] 10 to extrapolate from rats to humans.
Human studies positive for hemolytic anemia (Melzer-Lange
and Walsh-Kelly, 1989; Owa, 1989; Owaetal, 1993).
-------
Examination of EPA Risk Assessment Principles and Practices
Page 79
Table 4-5: Explanations of Why Human Results Were Not Used To Estimate the RfD for Four
Substances in the IRIS Database"
Chemical
1,3-
Dichloro-
propene
Human
Data (HD)
Listed
Yes
HD =
Principal
Study
No
Explanation
There are no chronic human studies suitable for dose-response
assessment. Chronic feeding studies in rats (Stott et al., 1995)
and mice (Redmond et al., 1995) and chronic gavage studies
(NTP, 1985) using both species are available. The feeding studies
are favored over the gavage studies because the route of
administration is more relevant to human exposure.
The toxicokinetics in humans are similar to those observed in
rats. Waechter etal. (1992) showed that the absorption of 1,3-
dichloropropene from inhalation exposure of humans (72%-82%)
was similar to absorption in rats (82%; Stott and Kastl, 1986).
The default uncertainty factor of 10 for interspecies extrapolation
is applied because there are no data on the relative sensitivity of
rats and humans to stomach irritation.
Human studies demonstrate contact dermatitis (Bousema et al.,
1991; Nater and Gooskens, 1976).
* Human poisonings result in neurotoxic symptoms (Hernandez et
al., 1994).
a These four substances were identified from the first 60 matches in the IRIS data file for which the word "human"
was identified. RfDs were estimated for 42 of the 60 substances. Human data relevant to the RfD were available
for 9 of the 42 substances. The human data were used as the principal study for the RfC for five of the nine
substances, This table describes the reasoning provided as to why the human study were not used in the other
four instances. Interspecies uncertainty factors are listed. Results of human studies are included in italics.
-------
Page 80
Examination of EPA Risk Assessment Principles and Practices
Table 4-6: Explanations of Why Human Results Were Not Used for the Carcinogenic
Classification for Six Substances in the IRIS Database"
Chemical
Bromomethane
Human Data
(HP) Listed
Yes
Explanation
D (Not classifiable as to human carcinogenicity)
Inadequate human data. A prospective mortality study was reported for
a population of 3579 white male chemical workers. The authors noted
that it was difficult to draw definitive conclusions as to causality
because of the lack of exposure information and the likelihood that
exposure was to many brominated compounds.
(Wong et al., 1984) study looked at cancer mortality in the above
mentioned occupational cohort. Only increased cancer mortality was
from testicular cancer.
P,P'-
Dichlorodiphenyl-
trichloroethane
(DDT)
Yes
B2 (Probable human carcinogen based on sufficient evidence of
carcinogenicity in animals)
Inadequate human data. The existing epidemiological data are
inadequate.
Autopsy studies relating tissue levels of DDT to cancer incidence
have yielded conflicting results. Three studies reported that tissue
levels of DDT and DDE were higher in cancer victims than in those
dying of other diseases (Casarett et al., 1968; Dacre and Jennings,
1970; Wasserman et al., 1976). In other studies no such relationship
was seen (Maier-Bode, 1960; Robinson et al, 1965; Hoffman et al.,
1967), Studies of occupationally exposed workers and volunteers
have been of insufficient duration to be useful in assessment of the
carcinogenicity of DDT to humans.
-------
Examination of EPA Risk Assessment Principles and Practices
Page 81
Table 4-6: Explanations of Why Human Results Were Not Used for the Carcinogenic
Classification for Six Substances in the IRIS Database"
Chemical
Human Data
(HP) Listed
Explanation
Polychlorinated
biphenyls (PCBs)
Yes
B2 (Probable human carcinogen
carcinogenicity in animals)
based on sufficient evidence of
Inadequate human data. The human studies are being updated; currently
available evidence is inadequate, but suggestive.
Significant cancer excesses in human cohort studies (Bertazzi et al,
1987; Brown, 1987; Sinks et al, 1992).
Inconclusive results of occupational studies due to design limitations
(NIOSH, 1977; Gustavsson et al, 1986; Shalat et al, 1989).
Liver cancer and skin disorders in populations exposed to PCB-
contaminated rice oil in Japan and Taiwan (ATSDR, 1993; Safe,
1994).
Dieldrin
Yes
B2 (Probable human carcinogen based on sufficient evidence of
carcinogenicity in animals)
Inadequate human data. Two studies of workers exposed to aldrin and
to dieldrin reported no increased incidence of cancer. Both studies were
limited in their ability to detect an excess of cancer deaths ... Expo sure
was not quantified, and workers were also exposed to other
organochlorine pesticides (endrin and telodrin). The number of workers
studied was small, the mean age of the cohort (47.7 years) was young,
the number of expected deaths was not calculated, and the duration of
exposure and of latency was relatively short.
No statistically significant excess in cancer deaths among
occupational cohort (Van Raalte, 1977; Ditraglia et al, 1981).
~ same studies as cited in the aldrin assessment
-------
Page 82
Examination of EPA Risk Assessment Principles and Practices
Table 4-6: Explanations of Why Human Results Were Not Used for the Carcinogenic
Classification for Six Substances in the IRIS Database'1
Chemical
Human Data
(HP) Listed
Explanation
Acrylamide
Yes
B2 (Probable human carcinogen based on sufficient evidence of
carcinogenicity in animals)
Inadequate human data. There are two studies on the relationship of
workers exposed to acrylamide and cancer mortality. A basic limitation
of both studies is that the design is insufficient to derive an inference of
relative risk.
Based on inadequate human data and sufficient evidence of
carcinogenicity in animals; significantly increased incidences of benign
and/or malignant tumors at multiple sites in both sexes of rats, and
carcinogenic effects in a series of one-year limited bioassays in mice by
several routes of exposures. The classification is supported by positive
genotoxicity data, adduct formation activity, and structure-activity
relationships to vinyl carbamate and acrylonitrile.
No statistically significant increase in cancer mortality in human
study (Collins, 1984).
Confounded results in human study (Sobel et al., 1986).
Aldrin
Yes
B2 (Probable human carcinogen based on sufficient evidence of
carcinogenicity in animals)
Inadequate human data. Two studies of workers exposed to aldrin and
dieldrin (a metabolite of aldrin) did not find these workers to have an
excess risk of cancer. B oth studies, however, were limited in their
ability to detect an excess of deaths from cancer ... Exposure was not
quantified, and workers were also exposed to other organochlorine
pesticides (endrin and telodrin). A small number of workers was
studied, the mean age of the cohort (47.7 years) was low, the number of
expected deaths was not calculated, and the duration of exposure and of
latency was relatively short.
No statistically significant excess in cancer deaths among
occupational cohort (Van Raalte, 1977; Ditraglia et al., 1981).
~ same studies as cited in the dieldrin assessment
8 These six substances were identified from the first 60 matches in the IRIS data file in which the word "human"
was identified. Of these 60, 27 were classified as to their carcinogenicity. Of the 27, 10 identify human data in
the assessment, and 4 use human data for identifying a carcinogenic classification. This table describes the
reasoning provided as to why the human data were not used in the other six instances. Results of human studies
are included in italics.
-------
Examination of EPA Risk Assessment Principles and Practices Page 83
4.5 Toxicity Equivalency Factor Approach
Several comments stated that the use of Toxicity Equivalency Factors (TEFs) is
inappropriate because TEFs are based on the unproven assumption of additive toxic effects,
compounded by the uncertainties inherent in the generation of TEFs, their application, and the
heterogeneity of effects for which they are being used. However, some comments acknowledge
the value of TEFs in certain circumstances.
4.5.1 What Are TEFs?
In 1986, EPA published the Guidelines for the Health Risk Assessment of Chemical
Mixtures (USEPA, 1986b). These provide the bases for evaluating human risk from exposure to
combinations of two or more chemical substances, regardless of source or of spatial or temporal
proximity. They were extensively peer reviewed, submitted for public comments, and reviewed
again by EPA's Science Advisory Board. The guidelines have been supplemented by a technical
support document (USEPA, 1988a) and the Supplementary Guidance for Conducting Health
Risk Assessment of Chemical Mixtures (USEPA, 2000h). These guidance documents refer the
assessor to several different methods for evaluating mixtures data of varying types;
epidemiologic or toxicological data on the environmental mixture (the whole mixture) to which
exposures occur; test data on a tested mixture judged to be sufficiently similar to the
environmental mixture; data from a group of similar mixtures; and data on the mixture
components. TEFs are used in evaluating mixture toxicity from data on components. Their use
is based on an assumption of additivity among similarly acting components of a mixture.
In the supplementary mixtures guidance (USEPA, 2000h), the Agency provided
procedures for developing the relative potency factor (RPF) method and described TEFs as a
special case of RPFs. The RPF method is component-based and relies on two types of
information: toxicological dose-response data for at least one component of the mixture being
addressed (referred to as the index compound or 1C) and scientific judgment as to the toxicity of
the other individual compounds in the mixture and of the mixture as a whole. If data are limited,
the applicability of a set of RPFs may be restricted to certain effects, a specific route of exposure,
or exposure duration. Application may also be limited to a certain portion of the dose-response
curve. TEFs are a special case of RPFs, because they can be applied to any effect, route, or
exposure duration, and across the entire dose-response curve (USEPA, 2000h). They are thus
used only in relatively data-rich situations. Both RPFs and the special case of TEFs are intended
to serve as interim approaches for addressing any mixture, pending the development of new data
on the mixture's toxicity. The supplementary mixtures guidance (USEPA, 2000h) noted an early
Agency application of an RPF-like approach for polycyclic aromatic hydrocarbons (PAHs)
(USEPA, 1993a; Schoeny et al, 1998). EPA has since applied the RPF method to
organophosphate pesticides in the Guidance on Cumulative Risk Assessment of Pesticide
-------
Page 84
Examination of EPA Risk Assessment Principles and Practices
Chemicals That Have a Common Mechanism ofToxicity (USEPA, 2002f). This was followed
with the publication of biostatistical criteria for developing RPFs (USEPA, 2003e).
The RPF method (including TEFs) is based on dose addition and assumes that the
chemicals in a mixture share a common toxic mode of action that is relevant to the health
endpoints being assessed (see section 5.7.2 for a complete discussion of dose addition).
Operationally, this means that mixture components tested in the same bioassay should have dose-
response curves of similar shape between the toxicity thresholds and the maximum response.
The components are assumed to be true toxicological representations of each other, although
their relative toxic potencies may differ.
When one uses the RPF method, it is necessary to identify the constraints of its
application. For example, a set of RPFs may be restricted to oral exposures and not be usable for
exposures to the same mixture through the inhalation route; this was the recommendation for
RPFs applied to PAHs. TEFs for chlorinated dibenzo-p-dioxins, dibenzo-p-furans, and coplanar
polychlorinated biphenyls (PCBs) have no identified constraints.
To apply the method (see the following formula), one estimates an RPF for each mixture
component; that is, one estimates component toxicity relative to that of the 1C.
Equation 4.5-1
Where:
&m
/(*)
A
RPFt
risk posed by chemical mixture
dose-response function of index chemical
dose of the ith mixture component (/=!,...,«)
toxicity proportionality constant relative to index chemical for the fth mixture
component (*' = !, ..., n)
number of chemical components in the mixture
One estimates the RPF by choosing a response level and calculating the ratio of the dose causing
that response for the component and the 1C. For example, say that the ED10 (the effective dose at
which 10% of a test population exhibits an effect) for component A is 500 micrograms (^g) and
that of the index chemical is 5 ug. If the index chemical RPF is set at 1, the component RPF
would be 5/500 or 0.01. The next step is to calculate the 1C equivalent dose for the mixture. The
equivalent dose of an individual component is the product of the amount of that component in
the mixture and the RPF of the component. In our example, if there is 100 ug of component A in
the mixture, then its 1C equivalent dose is 100 ug x 0.01, or 1 ug. These equivalent doses are
summed across all components to determine the mixture total equivalent dose. Lastly, one
-------
Examination of EPA Risk Assessment Principles and Practices Page 85
estimates the risk posed by the mixture by comparing its summed 1C equivalent dose to the dose-
response function (potency) of the 1C.
TEFs have been developed for three structurally related groups of compounds which have
been demonstrated to exert similar biochemical and toxic endpoints: the polychlorinated
dibenzo-p-dioxins (PCDDs), polychlorinated dibenzofurans (PCDFs), and (co-planar) PCBs. Of
these, 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) is the most extensively researched and has
served as the 1C. Many studies support the TEF approach for these compounds because they
share a common mechanism of action involving binding to a cellular receptor (AhR). TEFs were
derived for 29 dioxin-like agents, which express each congener's toxicity as a fraction of that of
TCDD, the 1C (van den Berg et al., 1998).
4.5.2 How Did the TEF Approach Evolve?
In the 1970s and 1980s, human health risk assessments of complex mixtures of PCDDs
and PCDFs were generally done only by considering 2,3,7,8-TCDD or by assuming that all
congeners were equally potent to 2,3,7,8-TCDD (USEPA, 1987a, 1989b). It appears that PCDD
and PCDF congeners (and co-planar PCBs) act through a common mechanism of action and
induce similar biochemical and toxicological effects. However, the toxicity of individual
congeners was shown to vary in different bioassays, leading to a recognition of the uncertainty
associated with the earlier approach.
Eadon et al. (1986) described the first TEF-like method as a means to estimate potential
human health risks associated with a PCB transformer fire in Binghamton, New York.
Subsequently, EPA concluded that a TEF approach was the best available interim science policy
for dealing with complex emissions of dioxins and forans from waste incineration. In 1987, EPA
adopted an interim procedure, based on TEFs, for estimating the hazard and dose-response of
complex mixtures containing PCDDs and PCDFs in addition to 2,3,7,8-TCDD (USEPA, 1987a).
Following adoption of the TEF methodology in the United States and Canada, the North
Atlantic Treaty Organization Committee on the Challenges of Modern Society (NATO/CCMS)
examined the methodology and concluded that it was the best available interim method for
PCDD/PCDF human health risk assessment (NATO, 1988a, b). NATO/CCMS refined the TEFs
by including more recent data sets and more in vivo data. The NATO/CCMS panel assigned
TEFs to OCDD and OCDF, and removed TEFs for all congeners lacking chlorine in the 2,3,7,8-
positions. Although it is theoretically possible to detect nearly all of the 210 PCDD/PCDF
isomers in the environment, only the seventeen 2,3,7,8-substituted congeners are known to
bioaccumulate. EPA officially adopted the revised TEFs in 1989 (TEFs-NATO89), with the
caveat that the methodology remain interim and continued revisions be made (USEPA, 1989b;
Kutz et al., 1990). The use of the TEF methodology for human health risk assessment and risk
-------
Page 86 Examination of EPA Risk Assessment Principles and Practices
management purposes has since been formally adopted by a number of other countries (e.g.,
Canada, Germany, Italy, the Netherlands, Sweden, and the United Kingdom) (Yrjanheiki, 1992).
The dioxin-like activity of co-planar PCBs has also been established (Safe, 1990, 1994).
Studies show that only PCB congeners substituted in the meta and para positions are approximate
stereoisomers of 2,3,7,8-TCDD and induce dioxin-like biochemical and lexicological effects
(Leece et al, 1985). In 1991, EPA convened a workshop which concluded that a small subset of
the PCBs displayed dioxin-like activity and met the criteria for TEFs (Barnes et al., 1991;
USEPA, 1991b).
In the years since initial adoption of the TEF methodology, additional data have been
developed on the lexicological potency of individual PCDDs, PCDFs, and PCBs relative to
2,3,7,8-TCDD. A joint project conducted by the World Health Organization European Centre for
Environmental Health (WHO-ECEH) and the International Programme on Chemical Safety
(IPCS) resulted in development of a database of all available relevant toxicological data for
dioxin-like compounds available through 1993. Following a review of almost 1,200 peer
reviewed publications, 146 were selected and used to calculate TEFs for PCBs (TEFs-WHO94).
A panel of experts from eight different countries recommended interim TEFs for 13 dioxin-like
PCBs, based on the reported results for 14 different biological and toxicological parameters, from
a total of 60 articles (Ahlborg et al., 1994). EPA reaffirmed application of this methodology in
human health risk assessment in its dioxin reassessment (USEPA, 2000i).
In 1997, a second WHO-ECEH group expanded the TEF methodology to include class-
specific TEFs for mammals, birds, and fish. The resulting report (van den Berg et al., 1998)
included TEFs for 7 PCDD, 10 PCDF, and 12 PCB congeners for mammals, birds, and fish
(TEFs-WHO98).
In summary, the TEF concept and associated TEF factors for PCDDs, PCDFs, and PCBs
were developed through an international consensus process that has
a) Included input from numerous international experts, agencies, and stakeholders.
b) Re-examined the assumptions and limitations of the concept and the TEFs.
c) Revised TEFs as new scientific information and data became available.
This method of adding weighted doses for related compounds is considered by the international
scientific community to be an improvement over other options, such as (1) ignoring risks for
non-TCDD congeners, (2) treating each congener as a separate chemical evaluation (i.e., ignoring
the risk posed by the mixture), or (3) assigning the full toxicity of TCDD to all similar congeners.
-------
Examination of EPA Risk Assessment Principles and Practices Page 87
4.5.3 Is the Additivity Assumption for TEF Justified?
A substantial effort has been made to test the assumptions of additivity and the ability of
the TEF approach to predict the effects of mixtures of dioxin-like chemicals. Current
experimental evidence shows that for PCDDs, PCDFs, co-planar dioxin-like PCBs, and strictly
AhR-mediated events, the concept of TEF additivity adequately estimates the dioxin-like toxicity
of either synthetic mixtures or environmental extracts, despite the variations in relative
contributions of each congener (see chapter 9 of the Dioxin Reassessment, USEPA, 2000a
particularly van Birgelen et al. [1994a, b] and van der Plas et al. [1999, 2000]; also Hamm et al.,
2003).
Studies in fish and wildlife species of mixtures of PCDDs, PCDFs, and PCBs (including
binary mixtures, synthetic complex mixtures, and environmental mixtures) support the additivity
assumption (Zabel et al., 1995; Walker et al., 1996; Tillitt and Wright, 1997). Further, numerous
studies that have examined the effects of environmental mixtures in marine mammals and avian
species show a correlation between toxic effects and dietary concentrations (toxic equivalency
quotient; TEQ) (Ross et al., 1996; Summer et al., 1996a, b; Giesy and Kannan, 1998; Restum et
al., 1998; Shipp et al., 1998a, b; Ross, 2000).
In summary, current experimental evidence shows that for PCDDs, PCDFs, and co-planar
PCBs, additivity adequately estimates the dioxin-like toxicity of both synthetic mixtures and
mixtures found in the environment. Interactions other than additivity have been observed with a
variety of effects in both binary combinations and complex synthetic mixtures of dioxin-like and
non-dioxin-like chemicals (commercial PCBs, PCB153). However, it appears that at these high-
dose exposures, multiple mechanisms of action not under the direct control of the AhR are
responsible for these non-additive effects.
4.5.4 What Is EPA's Experience With TEFs?
EPA and other organizations have estimated the risks of exposure to PCDD and PCDF
mixtures using TEFs for more than a decade. The TEF concept and supporting literature are
covered in detail in chapter 9 of the draft dioxin reassessment ('Toxic Equivalency Factors
[TEF] for Dioxin and Related Compounds") and also in the Part HI, Risk Characterization. Both
of these documents have undergone interagency review and other external peer review. The RPF
method was also extensively reviewed outside the Agency. The Regions used the "RPF-like"
approach developed for PAHs (USEPA, 1993a) to estimate risks posed by this group of
compounds. The Agency used the RPF method in its assessment of the cumulative risks
associated with exposures to organophosphorus pesticides (USEPA, 2002f, g); this effort
underwent extensive internal and external peer review.
-------
Page 88 Examination of EPA Risk Assessment Principles and Practices
In 1998, EPA and the Department of the Interior sponsored a workshop that examined
application of the TEF concept and supporting assumptions using ecological risk assessment case
studies. The resulting document, Workshop Report on the Application of 2,3,7,8-TCDD Toxicity
Equivalence Factors to Fish and Wildlife (USEPA, 200 le), provides general conclusions
regarding application of and uncertainties associated with the TEF methodology in ecological
risk assessment, as follows:
a) The TEF methodology is technically appropriate for evaluating risks to fish, birds,
and mammals associated with AhR agonists, and it can support risk analyses
beyond screening-level assessments.
b) The methodology entails less uncertainty and is less likely to underestimate risks
than are methods based on single compounds (i.e., 2,3,7,8-TCDD).
c) Because total PCBs in the environment can consist of many compounds that vary
in concentration and potency as AhR agonists, the TEF methodology provides a
means for accounting for their variable potency.
d) The uncertainties associated with using the methodology are not thought to be
larger than other sources of uncertainty within the ecological risk assessment
process.
For human health assessment, issues concerning TEFs are being addressed through the
dioxin reassessment, its review by the NAS, and the policy implementation document planned
for after the release of the reassessment. In the interim, the Agency is obliged in many instances
to use the best available science and risk assessment methods in regulatory decisions: hence the
movement to use the internationally accepted practice of TEQ calculations for PCBs. From an
international scientific perspective, the current TEQ concept and values are a necessary and
appropriate interim procedure in advance of the development of more refined methods (van den
Berg et al, 1998; USEPA, 2001e).
4.6 Model Use
4.6.1 Why Does EPA Use Environmental Models?
A wide variety of models have been used over the years to support environmental
decision making by the Agency. These models often involve environmental characterization,
environmental fate and transport, contact between stressors and receptor populations,
relationships between exposure and adverse human health and ecological effects, quantitative
estimation of risk, and economic impact models, all of which affect the manner in which the
Agency chooses to address a multitude of environmental questions. Frequently, modeled
-------
Examination of EPA Risk Assessment Principles and Practices Page 89
estimates become the basis for environmental cleanup, protection, or regulation. Environmental
models are vital to the Agency's mission because they allow us to (1) predict the effects of future
actions, (2) estimate exposures and risks where no measured data are available, and (3) attribute
risks to individual sources. Models, therefore, underlie how the Agency chooses to address a
multitude of environmental questions. At issue is whether the underlying assumptions,
algorithms, etc., used in the modeling and potentially the analytic structure of the models
themselves - are articulated to the public well enough for stakeholders to assess the validity of
modeling assessments. Further, analysis of the uncertainties in modeled estimates should be
clear and transparent to those unfamiliar with the models.
4.6.2 How Does EPA Approach Environmental Models?
As the Agency's scientific understanding of environmental issues has improved over the
years, the models we use to understand them have become increasingly complex, requiring the
involvement of a wide range of scientific disciplines in the development of the models as well as
in the interpretation of the results. As a consequence, the Agency and others have become
increasingly focused on consistency, quality, and duplication of effort in model development,
selection, and application. Such issues can, unfortunately, lead to risk characterizations that are
not "transparent." Transparency requires explicitness in the presentation of each step in the risk
assessment process. It ensures that any reader understands all the steps, logic, key assumptions,
limitations, and decisions in the risk assessment, and comprehends the supporting rationale that
lead to the outcome, including any modeling that is done for fate/transport and exposures.
Generally, EPA risk assessments address several categories of uncertainty. Each of these
merits consideration. For example, for models, there are uncertainties associated with the
selection of specific scientific models for each parameter in risk assessment, e.g., dose-response
models, models of environmental fate and transport, and exposure models. The selection of
parameters for these models creates even more sources of uncertainty. Ultimately, EPA guidance
recommends that risk assessors identify those uncertainties that, if changed, would substantially
affect the modeled outcomes.
4.6.3 What Is EPA Doing To Improve the Use of Environmental Models?
As early as 1989, EPA's SAB expressed its concerns on EPA's use of environmental
models in Resolution on Use of Mathematical Models by EPA for Regulatory Assessment and
Decision Making (USEPA, 1989c). Since that time, the SAB has continued to recommend that
EPA "establish a general model evaluation protocol, provide sufficient resources to test and
confirm models with appropriate field and laboratory data, and establish an Agency-wide task
group to assess and guide model use by EPA."
-------
Page 90 Examination of EPA Risk Assessment Principles and Practices
In response to these concerns, the Deputy Administrator established the Agency Task
Force on Environmental Regulatory Modeling (ATFERM) in 1992. In 1994, the ATFERM
published the Report of the Agency Task Force on Environmental Regulatory Modeling
Guidance, Support Needs, Draft Criteria and Charter (USEPA, 1994b). The report concludes
that (1) there is a need for training and technical support, (2) there is a need for model use
acceptability criteria, (3) there is a need for Agency guidance for conducting external peer review
of environmental regulatory modeling, and (4) there is a need for a Committee on Regulatory
Environmental Modeling (CREM). To this end, the report included a section entitled "Guidance
for Conducting External Peer Review of Environmental Regulatory Modeling."
In 1994, the EPA Risk Assessment Forum developed a document entitled Model
Validation for Predictive Exposure Assessments (USEPA, 1994c), a draft protocol for model
validation that defined a set of procedures for evaluating models for exposure assessments. In
1997, EPA's Officer of Research and Development (ORD) and program offices conducted an
Agency-wide conference (called the Models-2000 Workshop) to facilitate Agency adherence to
existing guidance on modeling, to define and implement improvements to the way in which the
Agency develops and uses models, and to recommend an implementation and improvement plan
for enhancing modeling within the Agency. As a follow-on to the workshop, 10 Action Teams
and a Steering/Implementation Team were established to develop further action plans for
improving the Agency's use of models.
In 1999, a White Paper on the Nature and Scope of Issues on Adoption of Model Use
Acceptability Guidance was developed (USEPA, 1999c). The paper, by discussing current
practices and case studies, reviewed progress in and identified issues associated with developing
model use acceptability guidance.
In February 2000, the Deputy Administrator established the CREM and developed the
framework of guiding principles for its activities. The CREM was established to promote
consistency and consensus within the Agency on mathematical modeling issues (including
modeling guidance, development, and application) and to enhance both internal and external
communications on modeling activities. The CREM supports and enhances existing modeling
activities by Agency program offices. The CREM provides the Agency with consistent yet
flexible modeling tools to support environmental decision making, in particular as they relate to
the development and implementation of programs with cross-Agency implications. Further, the
CREM provides EPA staff and the public with a central point for inquiring about EPA's use of
modeling.
In May 2000, the CREM initiated several cross-Agency activities that were designed to
enhance the Agency's development, use, and selection of regulatory environmental models. One
of these activities involved the ultimate development of modeling guidance. The CREM
determined that a workshop to facilitate discussion of good modeling practices among
-------
Examination of EPA Risk Assessment Principles and Practices Page 91
participants would constitute a viable starting point for developing modeling guidance. To
organize the workshop and develop issue papers for discussion, a cross-Agency Model
Evaluation Action Team was formed in June 2000. Over the next several months, the Action
Team (1) reviewed the White Paper on the Nature and Scope of Issues on Adoption of Model Use
Acceptability Guidance to identify model evaluation issues; (2) developed a list of components,
or elements, to follow in evaluating a regulatory environmental model; (3) identified several
general principles to support the use of model evaluation guidance; (4) developed an
organizational plan for the workshop that covered the identified model evaluation elements; (5)
held the workshop; and (6) developed a summary report of the workshop.
In February 2003, EPA's Administrator directed the CREM to draft Agency guidance on
environmental models and create a Web-accessible inventory of EPA's most frequently used
models. Both these efforts address the transparency and uncertainty with which environmental
models including fate and transport models are used to make decisions. First, in its
guidance on environmental models, EPA intends to recommend that the modeling community
document all steps taken to develop and evaluate models; the guidance will also recommend best
practices for evaluating models. Second, EPA intends to house this documentation of model
development and evaluation within the Web-accessible inventory of models.
In addition to the activities noted above, EPA issued Order 5360.1 A2 (USEPA, 2000c),
which covers collection and use of environmental data, including information produced from
models in the areas of (1) use of a systematic planning approach to develop acceptance or
performance criteria; (2) approved Quality Assurance Project Plans (QAPPs) or equivalent
documents; and (3) assessment of existing data used to support Agency decisions or other
secondary purposes (to verify that they are of sufficient quantity and adequate quality for their
intended use). Requirements for QAPPs are provided in EPA QA Manual 5360 Al (USEPA,
2000d) for EPA personnel and EPA QA/R-5 (USEPA, 200If) for extramural personnel.
The atmospheric dispersion models we use (e.g., the Industrial Source Complex model, or
ISC, and AERMOD) have been through rigorous scientific peer review. To improve our tools
for modeling fate and transport from air into other media, EPA developed the Fate, Transport and
Ecological Exposure Module of the Total Risk Integrated Methodology, or the TRIM.FaTE
model. TRIM.FaTE not only handles the partitioning of pollutants between environmental
media, but does so in an interactive way, allowing pollutants to move back and forth between
media throughout the simulation according to the simulation conditions. The input values, any
assumptions, and transfer algorithms are completely transparent and accessible to the user
because they are stored in an electronic "library" accessed by the model, not "hardwired" into the
model itself. This feature allows the Agency to easily replace algorithms or values as new
findings indicate they are appropriate, or as appropriate for different applications. A variety of
documents have been prepared on the model and algorithms compiled in the library for
applications to date (e.g., USEPA, 2002h, i, j); these documents are available along with the
-------
Page 92 Examination of EPA Risk Assessment Principles and Practices
model on the Agency's Web site. TRIM.FaTE also includes features for sensitivity and Monte
Carlo analyses to assist in developing analyses of uncertainty and variability for multimedia risk
assessments.
With regard to inhalation exposure models, EPA has developed methods that use human
time activity patterns and stochastic sampling to derive probabilistic activity-based estimates of
exposure. The current example of this is the Human Exposure Event module of TRIM (the
TRIM.Expo inhalation model, also known as the air pollutant exposure model or APEX). This
model is derived from the probabilistic National Ambient Air Quality Standards exposure model
(pNEM) and has been developed for both criteria and HAP applications. TRIM.Expo is
available on the EPA Web site and by its design it relies on data from input files that are
completely transparent to the user.
Regarding uncertainty analysis, each Agency risk assessment needs to include an analysis
of important uncertainties that, if reduced, might change the outcome of the assessment (see the
Agency's 1997 Policy for Use of Probabilistic Analysis in Risk Assessment and Guiding
Principles for Monte Carlo Analysis), The Agency believes that it is generally preferred that
quantitative uncertainty analyses be presented in each risk characterization, but there is no single
recognized guidance on how to conduct such uncertainty analyses. Even if the uncertainty
analysis is conducted qualitatively, the Agency still considers it to be of great value to a risk
manager.
4.6.4 How Does EPA Approach Fate and Transport Modeling and Account for
Uncertainty?
Once chemicals are released to the environment, they are subject to a wide range of
environmental processes that influence their fate and transport. These processes may affect their
location (e.g., meteorology, wind, water currents), their chemical structure (e.g., transformation,
reactivity), and the media in which they partition or reside. These processes' level of influence
varies significantly, often depending on the chemicals' physical and chemical properties as well
as the conditions they encounter in the environment.
This section describes the modeling conducted under OPP to illustrate EPA's approach to
fate and transport modeling. (Fate and transport modeling in OPP efforts typically encounters
multiple media and therefore is a good example of broad modeling practices used by EPA.) Fate
and transport models are used in OPP to estimate ambient concentrations for ecological risk
assessment and drinking water concentrations for human risk assessments. For ecological risk
assessments, estimated environmental concentrations are divided by a toxicity reference value to
yield an index that is compared to a level established to constitute an acceptable risk. For human
risk assessments, estimated drinking water concentrations are incorporated into a total dietary
assessment, which is then evaluated for risk.
-------
Examination of EPA Risk Assessment Principles and Practices
Page 93
EPA responds to uncertainty in these assessments depending on the type of decision to be
made. In estimating pesticide concentrations in water resources, OPP uses a tiered screening
approach (consistent with the general approach discussed in chapter 2) that involves a
combination of modeling and analysis of monitoring data. In OPP's tiered modeling system, the
program begins with a public-health-protective, national, highly vulnerable estimate and then, if
concerns exist, progresses to regional and eventually local, probabilistic, more resource-intensive
estimates. This triage approach is designed to provide a thorough analysis of each pesticide,
while focusing OPP's efforts on those pesticides that pose the greatest potential risk.
The intent of the screening approach is to estimate pesticide concentrations in water from
sites that are vulnerable to runoff or leaching so that the program can be confident that any
pesticide that passes the screening tiers poses a low possibility of harming human health,
wildlife, or the environment. Therefore, during these screening assessments, the intent is to
address the uncertainty regarding the most sensitive scenario that may exist in areas of pesticide
application.
These initial screening assessments establish the environmental setting and input
parameters that would result in a high-end exposure of a pesticide after application. Half-life
data from laboratory studies serve as inputs, as described in the publicly available input
parameter guidance. The initial screening levels models used by OPP are: for the surface water
pathway, GENEEC2 (GENeric Estimated Exposure Concentration) and FIRST (FQPA Index
Reservoir Screening Tool), and for the groundwater pathway, SCI-GROW (Screening
Concentration in GROund Water).
If a pesticide raises a concern after the initial screening tier, a set of higher-tiered surface
water models called PRZM/EXAMS are used to conduct a more refined assessment. These
models, developed by ORD, allow the Agency to focus on specific sites (by accounting for site-
specific properties including soil type, weather, and agricultural practices) or watersheds in its
analysis. At the present time, OPP does not have a higher-tiered groundwater model.
A tier II or PRZM/EXAMS refined water assessment uses a region-specific site with an
agriculturally relevant soil that is vulnerable to runoff and is of a large extent in the watershed
(i.e., a benchmark soil). The assessment also uses 30 years of monitored weather data for the site
and maximum pesticide use rates. The combination of site characteristics and fate property
inputs is designed to approximate a "high-end" exposure scenario. A tier II drinking water
assessments uses an index reservoir drinking water source (USEPA, 1998d), and ecological
exposures assessments are conducted using a small fixed-volume water body, the "farm pond."
The drinking water reservoir modeled is Shipman City Lake in Illinois; the farm pond is located
in Georgia. Drinking water estimates are adjusted to reflect the amount that a major crop,
specifically corn, wheat, cotton, or soybeans (USEPA, 1999d), may be grown in a watershed
using a national percent crop area factor. Crop area factors for other crops or combinations of
-------
Page 94 Examination of EPA Risk Assessment Principles and Practices
crops to which a pesticide may be applied in the same watershed are assigned a national default
value representing the highest percent of a watershed in agricultural production.
Recently, region-specific default percent crop areas were developed to reflect different
cropping practices across the United States. Estimates may be further refined by focusing on
additional region-specific use sites and typical usage and application practices. Estimated
concentrations from PRZM/EXAMS in the initial tiers are a 1 in 10 year occurrence frequency
and, for cancer risk assessments, a 30-year overall mean. Additional refinements are pesticide-
specific and maybe performed at the request of risk managers seeking to focus on risk
mitigation. These refinements may include a probabilistic dietary assessment using the total
PRZM/EXAMS output of 30 years of daily values.
If degradates are observed in the laboratory studies that are of toxicological significance,
have or are predicted to have pesticidal activity based on structure, or are a significant component
of the residues in an environmental compartment, modeling is conducted on these degradates.
The screening-level models are bypassed in favor of PRZM EXAMS to model degradates and are
assessed in separate modeling runs. The amount of data available to serve as inputs varies. Half-
lives can sometimes be obtained from the registrant-submitted studies if those studies were
conducted for a period greater than the half-life of the parent. Defaults are used in the absence of
data, as described in the input parameter guidance.
4.6.5 Does EPA Consider or Review Existing Data on Environmental
Concentrations?
Monitoring data are preferred when estimating concentrations; however, monitoring data
are generally either not available or inadequate. Monitoring data are not available for new,
never-registered chemicals. No or little monitoring data will be available for a chemical that has
little use in the United States or has a "niche" market. In some cases, after reviewing the
monitoring data, modeling outputs are selected because monitoring was conducted in areas in
which the chemical was not used or had little use. Generally, monitoring data in targeted
pesticide use areas are most useful for estimating long-term time-weighted concentrations and
potential chronic impacts. Low sample frequency and lack of statistically representative use sites
limits the usefulness of monitoring data for estimating peak concentrations and its potential acute
impacts. Sufficient data are sometimes available for widely used pesticides, pesticides that are in
multi-residue monitoring methods, or pesticides that have the interest of the public. In all risk
assessments, available monitoring results are compared to screening-level modeling estimates.
-------
Examination of EPA Risk Assessment Principles and Practices
Page 95
4.6.6 Does EPA Consider and Use Case-Specific Data in Models? (Or: How Does
EPA Decide To Modify Default Assumptions in Later Tiers Using Models?)
As in all aspects of risk assessment, EPA prefers to use specific data rather than default
assumptions when working with models. Default assumptions are used when there are data gaps.
When data of sufficient quality are available, EPA uses such data. For example, the pesticide
industry is responsible for submitting the data that serve as the inputs into these models. Data
requirements are described in 40 CFR Part 158, and the protocols for conducting these studies
are discussed in the Pesticide Assessment Guidelines. Registrants may use Organisation for
Economic Cooperation and Development (OECD) guidelines or other guidance provided the data
meets the purpose of the study. In selecting model parameters, OPP established publicly
available guidance for selecting inputs (Water Models, Guidance for Selecting Input Parameters).
The amount of data available determines how confident an input will become.
4.6.7 How Will EPA Improve Pesticide Fate and Transport Modeling in the
Future?
The Agency's goal is to develop realistic estimates of pesticide concentrations in ambient
and drinking water. The following efforts are underway to improve its tools and technology:
a) Ground water: A tier II model that generates daily or short-time-step
concentrations for use in a probabilistic dietary assessment will be in use in late
2004. Currently available groundwater models will be assessed jointly with the
Pest Management Regulatory Agency/Health Canada. In addition, a refined
version of SCI/GROW is being investigated that uses all of the data from
prospective groundwater studies submitted by the pesticide industry.
b) Surface water: Modifications are being made to PRZM/EXAMS to allow the size
of the receiving body to be adjusted to reflect the size of a typical pond or
reservoir in the area being modeled. A geographic information system is being
developed that maps all of the drinking water intakes in the United States and
their associated watersheds. Additional data layers will be added to allow the
Agency to model specific sites and their cropping patterns. Spray drift is a fixed
variable in PRZM/EXAMS. Spray drift models will be linked to PRZM/EXAMS
to generate realistic drift inputs (USEPA, 1997e, 1999e). Watershed-scale
regression models are being developed in cooperation with the U.S. Geological
Survey to estimate a distribution of concentrations according to watershed
properties and pesticide usage areas (Larson et al., in press; Larson and Gilliom,
2001).
-------
Page 96
Examination of EPA Risk Assessment Principles and Practices
c) Drinking water treatment: A workshop planned for January 2004 will address the
types of data needed to be able to factor drinking water treatment into estimates of
pesticide concentrations. Additional workshops and a Federal Insecticide,
Fungicide, and Rodenticide Act (FIFRA) Scientific Advisory Panel meeting are
expected to finalize any new data requirements that may be imposed on industry
to provide data.
d) Tiered water exposure assessment process: Interaction is conducted with the
human health risk assessors and risk managers in OPP to standardize the
modeling tiers and generate the data needed for conducting probabilistic exposure
assessments earlier in the process and to identify the tools and technologies to
routinely conduct probabilistic analyses. The program's goal is to account for
daily and seasonal variations in residues over time due to the time of application
and runoff/leaching events, year-to-year variations in weather patterns, and place-
to-place variability in residues caused by the water source and regional/local
factors (such as soil, geology, hydrology, climate, crops, pest pressures, and
usage) that affect the vulnerability of a source.
4.6.8 Does EPA Use Worst-Case Assumptions in Modeling That Are Unreasonable
and Do Not Account for Degradation, Partitioning, and "Sinks" in the
Environment?
Default assumptions are frequently used in modeling, specifically in exposure assessment
for each medium or cumulative assessment. Model inputs for fate and transport use the most
sensitive soil type, combined with the lowest rates of degradation, air exchange, or water
dispersion, and 100% of mass used in each scenario. These default assumptions are used to
address data gaps and to ensure that risks will not be underestimated given those uncertainties.
The degree to which defaults are incorporated into any modeling exercise depends on the
purpose of the assessment and the availability of data. In a screening assessment with little or no
available data, several default assumptions may be used, such as the assumption that the
chemical is stable and not subject to transformation or degradation. In fact, the potential for
degradation of pesticides in the environment is and has been incorporated into many aspects of
OPP risk assessments, as illustrated in the examples below.
a) Food: In dietary exposure and risk assessments, the use of monitoring data (e.g.,
the U.S. Department of Agriculture's Pesticide Data Program or market basket
surveys) incorporates potential degradation of pesticides from "farm gate"
residues to "dinner plate" residues. Dietary exposure estimates include
consideration of the impact of typical consumer practices such as washing and
peeling, and commercial practices during processing (juicing, blanching, freezing,
-------
Examination of EPA Risk Assessment Principles and Practices
Page 97
etc.), that may reduce or concentrate residues. For pesticides found largely on the
surfaces of fruits and vegetables, considerable reduction of residues has been
observed in processing studies, as well as in monitoring data when results for
fresh and processed commodities can be compared. Concentration of residues is
observed more often in dried commodities.
b) Water. In determining the potential concern for pesticide residues in drinking
water, OPP risk assessments incorporate modeled surface and groundwater values
generated using chemical-specific environmental fate inputs (e.g., hydrolysis,
anaerobic and aerobic metabolism, and microbial degradation). If sufficient data
are available, assessments also consider the impact of treatment on potential
residues in drinking water. Finally, monitoring data are evaluated for use in the
risk assessment. (If used quantitatively, then degradation in the environment is
inherently considered; if used qualitatively, a comparison between modeled and
monitored values is provided as characterization.) Currently, the impact of water
treatment on drinking water concentrations is being assessed qualitatively.
Following a recommendation from the FIFRA Scientific Advisory Panel meeting
during September 2000 (USEPA, 2000J), ambient concentrations are used in a
drinking water assessment unless there are data on the impact of water treatment.
c) Residential: Exposure and risk estimates for pesticides in residential settings are
generated using both default assumptions and chemical-specific data The default
assumptions are used largely in screening-level assessments. Chemical-specific or
activity-based information is incorporated as available. The timing and duration
of exposure are the most critical factors in estimating risk. For example, in
estimating toddlers' dermal exposure to pets, the highest residues on the pets' fur
(and the highest exposure to the child) are expected to occur on the day of
application. For longer-term exposure durations, the decline in residues on fur is
taken into consideration. In assessing dermal exposure to golfers from treated
turf, short-term non-cancer risks are estimated on the day of application at the
label rate, but cancer risks are estimated based on average application rates, and
take into account the frequency of golfing over a lifetime. Exposure assessments
that consider "sinks" (e.g., carpeting or soil) can be problematic due to data
deficiencies, but there is an assumption that less pesticide would be available for
dermal exposure, and an adjustment can be made in calculating the risks using, for
example, a "soft surface" exposure factor. In estimating indoor inhalation
exposures, the registered label is used as the basis for determining the re-entry
time and ambient concentrations, but breathing rate and level of activity
considerations can be included in the final risk estimates.
-------
Page 98
Examination of EPA Risk Assessment Principles and Practices
d) Occupational: For occupational handlers (those who mix, load, and/or apply
pesticides), estimated exposures are driven by pesticide formulation type, the
application method (e.g., aerial or ground boom), and the number of acres treated
per day. Exposure and risk estimates are calculated assuming different levels of
protective clothing and using the geometric mean of exposures (unit exposures
based on real data conducted with a variety of formulation types and application
equipment). Dermal risks estimated for different clothing types adjust for
potential protection afforded by a second layer of clothing or gloves, and consider
how much of the pesticide may actually get on the skin. Inhalation risks
calculated using unit exposure data already take into account activity levels,
breathing rate, and the potential ambient air concentrations for various pesticide
formulations during application. For harvesting or other foliar-contact activities
undertaken following application, the degradation of pesticide residues on leaves
over time is considered and used in the calculations. For estimating cancer risks,
exposures are calculated based on typical application rates, then amortized over a
lifetime.
In all of the above, the best available data are used to estimate exposure and risk and, to the
extent possible, reflect the use pattern for specific pesticides. If risks exceed the level of concern,
more specific information about a particular pesticide/crop/pest combination may be obtained to
provide more refined estimates of exposure and risk.
-------
Examination of EPA Risk Assessment Principles and Practices Page 99
5. SITE-AND CHEMICAL-SPECIFIC ASSESSMENTS
5.1 Overview
This chapter addresses EPA's use of site- or chemical-specific information in risk
assessments and the default assumptions that are often used in the absence of such information.
Many comments that EPA received regarding site- and chemical-specific information pertain to
assessments concerning waste sites. Therefore, this chapter largely focuses on the comments in
the context of human health risk assessment guidance for Superfund, which is also followed by
the Resource Conservation and Recovery Act Corrective Action program, as appropriate. EPA
also received a number of comments directed at risk management decisions made at specific
Superfund sites. Indeed, many of the site-specific comments repeat comments filed with the
Agency either during the development of the proposed plans or during the public comment
period for proposed plans at those specific sites. The EPA officials responsible for implementing
the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) at
those sites either have already responded to those comments or will be responding to them
through the CERCLA process. Therefore, this document does not respond to comments on
specific sites. Further information about the role of site- and chemical-specific data in other
regulatory programs may be available from individual program offices (refer to the EPA home
page at http://www.epa.gov/epahome/programs.htm) and by other examples found elsewhere in
this document (e.g., the "porch potato").
Due to the uniqueness of individual waste sites, site-specific information plays an
important role in risk assessments and management decisions on a regional level. In general,
EPA considers site- and chemical-specific information in risk assessments when it is available
and appropriate, then uses default assumptions when data gaps exist. Default assumptions are
typically based on scientific, peer reviewed data and aid in the evaluation of aspects such as
various age groups and various populations (e.g., residents, workers, Native American tribes,
subsistence fishers).
In terms of program goals and protecting public health, it is appropriate for EPA to use
default assumptions when data are not available to use in a risk assessment. Note that the default
exposure values are not all high-end, are not "worst-case," and have data supporting their
derivation (in addition to being peer reviewed). The following table presents examples of
commonly used default exposure factors and their associated percentile distribution, which range
from median to 95th percentile.
-------
Page 100 Examination of EPA Risk Assessment Principles and Practices
Table 5-1: Examples of Default Exposure Values With Percentiles
Exposure Pathway
Drinking water
consumption:
2 liters/day
Soil ingestion rate for
children:
200 mg/day
Residence duration:
30 years
Body weight:
70kg
Percentile
90th
65th
90th
80th
90th-95th
50th
Source of Data
Approximately a 90th percentile value (USEPA,
2000k).
Analyses and distributions constructed by Stanek
and Calabrese (1995a, b, 2000) places the 200 mg
ingestion rate around the 65th percentile of average
daily intakes throughout the year. The Stanek and
Calabrese analyses suggests that ingestion rates for
children in the top 10% (i.e., the high end) of the
distribution would be greater than 1,000 mg/day.
For home owners, farms, and rural populations; 30
years is greater than the 95th percentile residence
time for renters and urban populations.
For males and females 18 to 75 years old (NCHS,
1987)
Risk assessments and the collection of data and other information are conducted on a site-
specific basis by the EPA regional offices, within the regulatory requirements of the waste
programs. Public participation is a critical component of the regional Superfiind process. For
example, public knowledge about the history of disposal practices at an abandoned waste site
often informs the conceptual site model design and development of sampling plans around the
site. Another example of site-specific information involves people's interaction with their
environment. Some populations may eat more fish than would be estimated using EPA default
exposure values based on differences in consumption patterns. If resources are available, and
considering the population and other factors, a site-specific survey may be conducted to assess
the consumption patterns for the specific population.
Often, when EPA actively seeks public participation, the Agency staff also need to clarify
what constitutes useful data for risk assessment. For example, a soil sample placed in a
household container does not provide useful or adequate data for a risk assessment. EPA would
explain that information about the extent of contamination is collected following the quality
assurance/quality control guidance (USEPA, 20001) using a standard protocol with quality
control and quality assurance procedures.
To address the concerns and comments about use of site- and chemical-specific data and
to fully understand assessments of cancer risks and non-cancer health hazards, it is important to
-------
Examination of EPA Risk Assessment Principles and Practices Page 101
understand their underpinnings. Among these are the regulatory requirements and purpose of the
programs for which the assessments are conducted, whom the programs seek to protect, the role
of default assumptions and site- and chemical-specific information in risk assessments, the types
of risk assessment performed, and factors considered in risk management decisions. The
following sections present information about these aspects of the Superfund program.
5.1.1 What Are the Superfund Program's Purpose and Regulatory Requirements?
CERCLA, 42 U.S.C. 9605 (CERCLA, 1980), as amended by the Superfund Amendments
and Reauthorization Act of 1986 (SARA), P.L. 99-499 (SARA, 1986), requires that actions
selected to remedy hazardous waste sites be protective of human health and the environment.
The National Oil and Hazardous Substances Pollution Contingency Plan, or NCP, is the
regulation that implements CERCLA (NCP, 1990a, b). The NCP establishes the overall
approach for determining appropriate remedial action at Superfund sites (NCP, 1990a, b).
The national goal of the Superfund program is to select remedies " ... that are protective
of human health and the environment, that maintains protection over time, and that minimizes
untreated waste"(NCP, 1990a). This means that the Superfund program needs to protect human
health and the environment from current and potential future threats posed by uncontrolled
hazardous substance releases. Pursuant to the NCP, decisions at Superfund sites involve
consideration of site-specific information and cancer risks and non-cancer health hazards
associated with both current and future land use conditions.
5.1.2 Whom Does the Superfund Program Seek To Protect? The Reasonable
Maximum Exposure Scenario
One of the policy goals of the Superfund program is to protect a high-end, but not worst-
case, individual exposure: the reasonable maximum exposure (RME). The RME is the highest
exposure that is reasonably expected to occur at a Superfund site. As described in the preamble
to the NCP (NCP, 1990a, b), the RME will
... result in an overall exposure estimate that is conservative but within a realistic range of exposure. Under
this policy, EPA defines "reasonable maximum" such that only potential exposures that are likely to occur
will be included in the assessment of exposures. The Superfund program has always designed its remedies
to be protective of all individuals and environmental receptors that may be exposed at a site; consequently,
EPA believes it is important to include all reasonably expected exposures in its risk assessments ...
In addition to evaluating the risks to the RME individual, EPA evaluates risks for the
central tendency exposure (CTE) estimate, or average exposed individual. This approach is
consistent with the Risk Characterization Policy and Handbook (USEPA, 1995a, 2000a). CTE
estimates give the risk manager additional information to consider while making decisions at a
-------
Page 102 Examination of EPA Risk Assessment Principles and Practices
site. Pursuant to the NCP, decisions at Superfund sites are based on cancer risks and non-cancer
health hazards associated with RME estimates under both current and future land use conditions.
5.1.3 Is the RME Overly Conservative?
EPA received comments implying that the RME represents a "worst-case" or "overly
conservative" exposure estimate. As described above, the RME is not a worst-case estimate
the latter would be based on an assumption that the person is exposed for his or her entire
lifetime at the site. As the preamble to the NCP (NCP, 1990a, b) states:
... The reasonable maximum exposure scenario is "reasonable" because it is a product of factors, such as
concentration and exposure frequency and duration, that are an appropriate mix of values that reflect
averages and 95thpercentile distribution ...
The RME represents an exposure scenario within the realistic range of exposure, since the
goal of the Superfund program is to protect against high-end, not average, exposures. The "high
end" is defined as that part of the exposure distribution that is above the 90th percentile, but
below the 99.9th percentile. The approach was developed with technical support from ORD and
is consistent with the peer reviewed EPA Exposure Assessment Guidelines (USEPA, 1992a).
Further, for some individual activity patterns, the RME may be modified based on
site-specific considerations (e.g., workers who may remain in one location, individuals who live
in a residence for their lifetimes, Native Americans remaining on tribal lands, children exhibiting
pica behavior). The Superfund program supports the development of reasonable risk assessments
that address the exposure and risk to all segments of the community, not only the "average"
individual. EPA seeks to protect "sensitive populations," segments of the general population that
are at greater risk, either because of particular sensitivity to the toxic effects of certain chemicals
or because they experience higher exposures than the general population, as children do. Under
the NCP (NCP, 1990a, b), the Superfund program achieves this goal of protecting public health
by using the RME approach.
5.2 Superfund Risk Assessment Guidance
5.2.1 What Are Some Types of Risk Assessment Under Superfund?
EPA conducts risk assessments that vary in length and degree of detail, from screening-
level assessments to comprehensive site-specific baseline assessments of hazardous waste sites.
Risk assessments may focus on chronic, long-term exposures and/or evaluation of acute
exposures that may require an emergency response. They are often conducted using a triage
approach, beginning with a screening-level assessment to determine if a more comprehensive
assessment is necessary (USEPA, 1988b, 1996a, 1999a, 2001b, 2002k).
-------
Examination of EPA Risk Assessment Principles and Practices Page 103
5.2.2 How Does Superfund Use Site-Specific Concentration and Maximum
Concentration Data in Risk Assessment?
When evaluating the information used in risk assessment, it is important to consider the
level of risk assessment being conducted. In screening-level assessments, risk assessors may
assume that receptors are in contact with the maximum concentration for the entire exposure
duration to place an upper bound on exposure estimates. This approach is used to ascertain
which exposure pathways or contaminants are most important, so that resources may be more
efficiently used in developing a comprehensive risk assessment for the most important
contaminants and exposure pathways.
The concentration term is a sensitive parameter in an exposure assessment. In most risk
assessments, an estimate of the average chemical concentration is used to represent the exposure
concentration or exposure point concentration (EPC) for assessing long-term or chronic exposure
scenarios. Because of the uncertainty associated with estimating the true average concentration,
the 95% upper confidence limit of the arithmetic mean (UCL95) is used to account for
uncertainties (USEPA, 1992b, 20021).
The UCL^ is highly dependent on the number of samples collected. The greater the
number of environmental samples, the closer the UCL95 will be to the true mean. Sometimes,
environmental data are limited or there is extreme variability in measured or modeled data.
When this is true, the upper confidence limit can be greater than the highest measured or
modeled concentration and neither the average nor UCL^ is a reliable estimate of the exposure
point concentration. If additional data cannot practicably be obtained, it is more appropriate to
use the highest measured or modeled value as the EPC. In cases where the true mean is actually
higher than this maximum value, especially if sampling data are very limited, the maximum
concentration is used in the calculation (USEPA, 1992b).
Site-specific contaminant data are collected for all sites. In some cases, the maximum
concentration is used when data are limited or the calculated UCL95 exceeds the maximum
concentration. However, the maximum concentration is rarely used for hazardous waste sites in
which a comprehensive baseline risk assessment is being conducted, since reliable site-specific
data are typically collected as part of the remedial investigation (RI) process. During the
planning and scoping phase, the minimum number of samples that is sufficient to characterize
environmental media is determined using EPA's data quality objectives process (USEPA, 20001).
The reality is that the number of samples collected from each medium, and the resulting certainty
achieved, need to be balanced by the resources available and the purpose of the assessment.
Lastly, it is important to note that it is appropriate to assume individuals are exposed to
the maximum concentration when acute exposure scenarios are being evaluated, such as those
conducted for emergency removal actions. Another example: when "hot spots" with relatively
-------
Page 104 Examination of EPA Risk Assessment Principles and Practices
high concentrations are present in surface soil, then acute toxicity is of most concern and the
focus is on measuring or estimating short-term, peak environmental concentrations.
5.2.3 Has the Superfund Guidance Been Updated and Externally Peer Reviewed
Since It Was First Released in 1989?
EPA continues to develop and peer review guidance to assist risk assessors conducting
various types of risk assessments in Superfund. Since the development of the Superfund Public
Health Evaluation Manual (USEPA, 1986c), and later in 1989 the RAGS Part A, Human Health
Evaluation Manual (USEPA, 1989a), the Superfund program developed appropriate guidance to
help risk assessors evaluate risks to the RME individual (USEPA, 1996a, 1999a, f, 2001 a, b, g,
2002k). The guidance consistently included internal review by risk assessors across the Agency,
external comment by stakeholders, and external peer review where appropriate based on our peer
review guidance.
External peer reviews have been conducted for the majority of guidance documents
developed since 1989. For example, as early as 1989, the RAGS Part A was reviewed by the
Agency's Science Advisory Board. In addition, since the issuance of the Agency policy on peer
review in 1994 (USEPA, 1994d) and subsequent handbook (USEPA, 2000b), where appropriate
and consistent with peer review policies, Superfund Risk Assessment guidance has undergone
external peer review. The external peer review meetings have included opportunities for the
stakeholders to participate through public comment. All guidance documents are made available
on the Internet for public comment and addressed in the final documents.
The probabilistic risk assessment guidance is one example of the process used in the
Superfund program to develop guidance. In that case, Superfund identified the emerging science,
developed an EPA workgroup to evaluate the available science and its application within the
Superfund program, released the draft guidance document for public comment, and conducted an
external peer review. The guidance document (USEPA, 200la) provides program-specific
information regarding the conduct of probabilistic risk assessments and supplements the earlier
policy on this issue (USEPA, 1997c). In addition, EPA has developed training courses on the
application of this methodology within the Superfund program. To date, probabilistic risk
assessment methods have been used or are being developed at several sites to evaluate cancer
risks and non-cancer health hazards (USEPA, 2000m, 2003f).
The risk assessment processes used at individual Superfund sites have evolved over time
based on new science and EPA's understanding of new potential exposure pathways. For
example, in the early days of the program, dermal exposure was not fully evaluated based on a
lack of dermal exposure information; this guidance was updated including an external peer
review (USEPA, 2001g). The Superfund program has updated other guidance documents,
including external peer review where appropriate to address the current scientific knowledge.
-------
Examination of EPA Risk Assessment Principles and Practices Page 105
5.3 Default Assumptions and Site-Specific Information
5.3.1 What Is the Role of Default Assumptions and Site-Specific Information in
Superfund Risk Assessments?
Risk assessments incorporate both default assumptions and site-specific information. The
supplemental guidance document, "Standard Default Exposure Factors" (OSWER Directive
9285.6-03, March 25, 1991), presents the Superfund program's default exposure factors for
calculating RME exposure estimates (USEPA, 1991c). This guidance was developed in response
to requests that EPA make Superfund risk assessments more transparent and their assumptions
more consistent. However, the guidance clearly states that the defaults should be used where
"there is a lack of site-specific data or consensus on which parameter to choose, given a range of
possibilities."
The table in section 5.1 presented examples of default exposure values and the percentile
of the population the values represent, as well as the peer reviewed studies supporting these
assumptions. Again, the RME approach uses default values designed to estimate the exposure of
a high-end individual in the 90th percentile of exposure or above (USEPA, 1992a). Consistent
with this guidance, relevant default assumptions for various activity levels and age groups are
used for drinking water consumption rates, soil ingestion rates, residence times, body weight, and
inhalation rates. The table illustrates the range of percentiles some defaults included the 50th
percentile (e.g., body weight), 80th, 90th, and 95th percentiles.
Although the Superfund program routinely uses default assumptions to assess the risk to
the RME individual at many sites, the characteristics of the surrounding population change from
site to site. For example, the distributions of individual residence times will vary depending on
whether the site is located in a rural or an urban area. Individuals in rural communities are likely
to have longer residence times than individuals in urban communities. Thus, a default value of
30 years may fall at the 80th percentile for farmers but above the 95th percentile for renters in an
urban setting. The extent to which a single default value will impact the final exposure estimate
depends on the values and variabilities of all the parameters used to estimate exposure. The goal
is to estimate an individual exposure that actually occurs and is above the 90th percentile. In
some cases, use of default assumptions may produce an estimate near the 90th percentile; in
others, the estimate may be higher in the range.
In general, Superfund's default factors are designed to be reasonably protective of the
majority of the exposed population. The assumptions used in Superfund's risk assessments are
consistent with the 90th percentile or above and the Agency's exposure assessment guidelines
(USEPA, 1992a). Default exposure factors used to assess the RME are a mix of average and
high-end estimates (see table 5-1). The use of these default exposure assumptions does not
-------
Page 106 Examination of EPA Risk Assessment Principles and Practices
automatically result in an overestimation of exposures. These following sections illustrate this
with several examples.
5.3.2 What Is the Drinking Water Consumption Rate?
EPA uses site-specific data on drinking water consumption patterns that have met
appropriate review criteria at individual sites, but these site-specific studies are generally limited
or absent. In these cases, to evaluate exposures to contaminants in drinking water, the Superfund
program applies a default ingestion rate of 2 liters per day for the RME individual (USEPA,
1989a, 1991c, 2000k). The 2 liters per day assumption represents a reasonable default
assumption. The results of the 1994 - 1996 Continuing Survey of Food Intake by Individuals
(CSFII) analysis indicate that the arithmetic mean, 75th, and 90th percentile values for adults 20
years and older are 1.1, 1.5, and 2.2 liters per day, respectively (USEPA, 2000k). The 2 liters per
day value represents the 86th percentile for adults.
In certain parts of the United States or, during summer seasons, throughout the country
people may drink 4 to 4.5 liters per day (USEPA, 1997d). Similarly, a report from OW on the
analysis of the CSFn data shows that the 99th percentile is 4.2 liters per day (the report provides
no maximum) (USEPA, 2000k). In some circumstances, then, this default value may
underestimate cancer risks and non-cancer health hazards for the reasonable high-end exposure
scenario.
5.3.3 What Is the Inhalation Rate?
The 1989 RAGS Part A describes 20 cubic meters per day as an average inhalation rate
and 30 cubic meters per day as an upper bound (USEPA, 1989a, 1997d). The inhalation rate
varies depending upon the activity levels of the exposed individual (USEPA, 1997d). The
default value of 20 cubic meters per day reflects a typical mix of activity levels; however, more
accurate, site-specific activity level data may be used when available. The default value is
consistent with the recommendations from the International Commission on Radiological
Protection (ICRP) data set for the "reference man" (ICRP, 1981; cited in USEPA, 1997d). The
ICRP used values of 20 to 23 cubic meters per day. These values assumed 16 hours of light
activity and 8 hours of resting. Daily inhalation rates for individuals performing activities at
levels other than resting or light activities are not presented. Thus, the values or the exposure
period may need to be adjusted for individuals (e.g., construction workers) with moderate and/or
heavy activity levels where inhalation rates would be higher.
5.3.4 What Is the Exposure Duration for Residences?
When site-specific data are not available or appropriate, the default residential exposure
durations are 30 years. This value represents the 90th percentile for home owners and the 90th to
-------
Examination of EPA Risk Assessment Principles and Practices Page 107
95th percentile for rural populations (USEPA, 1991c, 1997d). It is not reflective of the residence
times for farms, where 30 years represent an 80th percentile for this specific population (USEPA,
1997d).
Some site-specific exposures may continue even if the individual changes residences,
provided that he or she remains in the general area of the site. For example, Connelly et al.
reported that individuals are willing to travel up to 34 miles to continue fishing in a favorite area
(Connelly et al., 1992). At an existing Superfund site where more than 100 river miles have been
contaminated, it is thus possible for an angler to change his residence within the surrounding
counties and continue to fish from the impacted river. An assumption of an exposure duration
based only on residence time in this instance would underestimate the risks to the angler.
5.3.5 What Is the Exposure Duration for Workers?
It is Superfund practice to assume that workers are generally exposed for a duration of 25
years, unless site-specific data are available (USEPA, 1991c). The exposure duration is assumed
to be equivalent to job tenure for receptors in an occupational scenario. An analysis of Bureau of
Labor Statistics data shows that the projected job tenure varies from a few months to 50 years,
depending on the specific occupational category (Burmaster, 2000).
More specifically, the projected 95th percentile values for job tenure for men and women
in the manufacturing sector are 25 years and 19 years, respectively. The projected 95th percentile
job tenure values for workers in the transportation/utility and wholesale sectors are only
somewhat less than the values for manufacturing workers 22 years and 18 years for men and
women, respectively. Values are lower for other non-industrial sectors approximately 13
years for workers in the finance and service sectors, and 7 years for retail workers. Note that the
95th percentile is within the range for the high-end distribution, which is defined as that part of
the exposure distribution that is between the 90th and 99.9th percentiles (USEPA, 1997d).
Thus, the 25-year default value is an estimate of the 95th percentile exposure duration for
workers across a wide spectrum of industrial and commercial sectors. This value maybe an
overestimate or underestimate of the actual exposure duration, depending on the particular
circumstances of the employment, e.g., for future workers on a cleaned-up Superfund site. EPA
supports the use of alternative exposure durations, if they are based on adequate data on job
tenure and the anticipated industrial/commercial site use. For example, in the evaluation of a
construction worker scenario, the exposure duration is typically 1 to 2 years, representing the
time an individual is actually engaged in construction activities on the site (USEPA, 2002k).
As mentioned above, EPA also typically evaluates CTE scenarios, which is consistent
with EPA's guidance on risk characterization (USEPA, 1995a, 2000a). For example, 6.6 years is
often used as a CTE duration for workers. This value is based on the median occupational tenure
-------
Page 108 Examination of EPA Risk Assessment Principles and Practices
of all workers in the United States (USEPA, 1997d). Evaluation of CTE is usually limited to
comprehensive baseline risk assessments, whereas 25 years is the default exposure duration for
screening-level assessments in EPA's waste programs.
5.3.6 What Is the Ingestion Rate for Construction Workers?
Comments submitted implied that the Superfund program continues to use an ingestion
rate of 480 milligrams per day for construction workers. This ingestion rate was updated in 2000
with EPA's externally peer reviewed soil screening level guidance to 330 milligrams per day
(USEPA, 2002k).
5.3.7 What Are the Incidental Soil Ingestion Rates for Children?
A common question asked of EPA is why Superfund risk assessments evaluate "dirt-
eating kids": Why should Superfund sites be cleaned up to levels such that children can safely
"eat" the soil there? Actually, EPA does not typically assume that children are eating the dirt;
rather, EPA assumes that they are exposed to contaminants through the course of normal
activities of play on the ground, exposure to dust in the home, and incidentally through mouthing
behavior (USEPA 1996b, 200Id).
It is commonly observed that young children suck their thumbs or put toys and other
objects in their mouths. This behavior occurs especially among children from 1 to 3 years old
(Behrman and Vaughan, 1983; Charney et al., 1980). This "hand-to-mouth" exposure is well
documented in the scientific literature for children under 6, and is especially prevalent among
children 11A to 3 years old, a critical period for brain development This time period is of special
concern regarding potential exposure, since children may be at special risk of exposure to
specific chemicals, e.g., lead (ATSDR, 1991). Superfund experience has taught us that children
do incur exposures to contaminated soil, as is evident at lead-contaminated sites in which
elevated blood levels occur in children residing at those sites (USEPA 1996b, 2001d).
Scientists agree that because of this behavior, children may incidentally or accidentally
take in soil and dust (Calabrese et al., 1989; Davis et al., 1990; Van Wijnen et al., 1990). Where
children are likely to be exposed to contaminated soils (in residential areas, for example), it is
appropriate for EPA to evaluate potential risks and set cleanup levels that will protect children
for this widely recognized pathway of exposure, especially during this sensitive developmental
period in the child's lifetime.
The basis of EPA's default soil ingestion rate is generally a point of contention. EPA has
developed soil ingestion rates that are used as "default exposure assumptions" for adults and
children. For young children (6 years or younger), the Superfund program default value is 200
milligrams of soil and dust ingested per day (USEPA, 1991c, 1996a). EPA's risk estimates
-------
Examination of EPA Risk Assessment Principles and Practices Page 109
address the "incidental" ingestion that might occur when a child puts a hand or toy in his or her
mouth, or eats food that has touched a dusty surface. Although this default assumption is often
presented as an overly conservative value, the amount (200 milligrams per day) represents a
small amount of soil ingested. It is less than 1/100 of an ounce (or one-fifth of the contents of a
single-serving packet of sugar, e.g., a small package of Equal sweetener) a day. This peer
reviewed value is applied in estimates of RME exposures (USEPA, 1989a, 1991c, 1997d).
In Superfund risk assessments, this soil ingestion rate for young children is combined
with site-specific assumptions about exposure frequency (days per year) to estimate an average
intake over the 6-year exposure period. Exposure frequency varies depending on site-specific
current and future land uses. Soil ingestion studies report daily averages; the amount of soil
ingested cannot be prorated on an hourly basis. Also, soil ingestion is episodic in nature and
dependent upon a child's activity patterns, so prorating by time is not always appropriate. This is
a common misapplication of soil ingestion rates in risk assessment.
Some children deliberately eat soil and other non-food items (a behavior known as pica).
Pica behavior has been identified in children at rates of up to 5,000 milligrams per day (ATSDR,
1996, 2000; Calabrese et al., 1991). The Agency for Toxic Substances and Disease Registry uses
this pica ingestion rate when calculating Environmental Media Evaluation Guides, which are
used to select contaminants of concern at hazardous waste sites (ATSDR, 1996). EPA itself does
not routinely address this form of exposure unless site-specific information is available. The
default soil ingestion rate of 200 milligrams per day applied in Superfund risk assessments is
intended to ensure reasonable protection of children in cases where they are likely to become
exposed to contaminated soils and dust associated with a Superfund site.
5.4 Site-Specific Information
5.4.1 When Does EPA Use Site-Specific Information?
Again, site-specific information is considered in risk assessment where available. EPA
risk assessment guidance supports the use of site-specific data, where feasible and appropriate
according to a careful evaluation of the information including the specifics of the study design,
numbers of individuals evaluated, representativeness of the exposures, etc. Examples of useful
information are location of residences (receptor location), current and future land uses, behavior
patterns and preferences (drinking water use, fish preparation), unique exposure pathways
(hunting of specific game), and residence time within specific geographic areas.
The nature of the data submitted by stakeholders varies from anecdotal information to
knowledge about creel surveys or other surveys conducted by universities or other groups. The
nature and amount of data submitted will vary based on the site, ranging from no submission to
-------
Page 110 Examination of EPA Risk Assessment Principles and Practices
data on specific work practices at a facility. When such information is submitted, EPA critically
evaluates it to determine its use in the risk assessment portion of the remedial investigation (RI).
The most accurate way to characterize potential site-specific exposures to populations
around Superfund sites would be to conduct a detailed census at each site considering both
current and future land uses. Theoretically, this would involve interviewing all potentially
exposed individuals regarding their lifestyles, daily patterns, water usage, cleanliness,
consumption of local fish and game and procedures, working locations, and exposure conditions,
while collecting environmental samples (e.g., indoor air, water, soil, food, and possibly more).
Although site-specific data are collected on environmental media (e.g., soil, groundwater, air,
etc.) as appropriate during the RI, such collection has significant limitations. The three almost
insurmountable difficulties are time, expense, and a major intrusion on privacy.
Discussed below are two examples of exposure parameters for which EPA has collected
and applied site-specific data in risk assessment: bioavailability and fish consumption rates.
5.4.2 How Is Bioavailability Addressed?
The Agency addresses bioavailability by using default values and, in some cases,
developing site-specific values supported by laboratory studies. In the absence of reliable
information, RAGS Part A recommends assuming that the relative absorption efficiency between
food or soil and water is 1. In other words, the assumption is that the bioavailability of the
contaminant on the site, regardless of exposure medium, is the same as the bioavailability in the
toxicity study used to derive the RfD or CSF. Such a default value is used to ensure that we do
not underestimate risk in the face of uncertainty. The bioavailability of a compound in the
exposure medium of concern at the site may actually be greater than in the exposure medium
used in the critical toxicity study that formed the basis of the RfD or CSF. If this is the case,
assuming a relative bioavailability of 1 for the medium of concern may result in an underestimate
of risk at the site, depending on how all the other parameters are evaluated.
Appendix A of RAGS Part A (USEPA, 1989a) specifically addresses the consideration of
site-specific bioavailability information in human health risk assessments. This information is
referred to as "adjustments for absorption efficiency." In particular, Appendix A states that site-
specific bioavailability adjustments may be appropriate:
Adjustments also may be necessary for different absorption efficiencies depending on the medium of
exposure (e.g., contaminants ingested with food or soil might be less completely absorbed than
contaminants ingested with water).
If the medium of exposure in the site exposure assessment differs from the medium of exposure assumed by
the toxicity value (e.g., RfD values usually are based on or have been adjusted to reflect exposure via
drinking water, while the site medium of concern may be soil), an absorption adjustment may, on occasion,
be appropriate. For example, a substance might be more completely absorbed following exposure to
-------
Examination of EPA Risk Assessment Principles and Practices Page 111
contaminated drinking water than following exposure to contaminated food or soil (e.g., if the substance
does not desorb from soil in the gastrointestinal tract).
For example, juvenile swine have been used as a model to derive site-specific
bioavailability adjustments at several mining and smelter sites across the country (Casteel et al.,
1997; USEPA, 1999g). These adjustments have been primarily limited to lead and arsenic.
Default values have been developed for some metals. For example, lead risks in children
are typically assessed by predicting blood lead levels using the Integrated Exposure Uptake
Biokinetic Model for Lead in Children (IEUBK) developed by EPA (USEPA, 200Id). The
IEUBK model assigns default absolute bioavailability factors to all lead exposure media. The
default values for air, water, and soil are 32%, 50%, and 30%, respectively (USEPA, 1999g,
2001d). For nonresidential scenarios, EPA developed the Adult Lead Methodology, which
assumes the absolute bioavailability of lead in soil is 12% (USEPA, 1996b).
EPA finds that the site-specific application of quantitative bioavailability adjustments in
risk assessments is not frequently supported by available scientific data. This occurs principally
because of a lack of validated data and models to assess bioavailability for common site
contaminants.
5.4.3 How Are Fish Ingestion Rates Evaluated?
Fish ingestion is evaluated at many Superfund sites with contaminated river systems,
streams, and lakes. These contaminated sites range in size from small creeks to large river
systems covering hundreds of miles. The contaminants of concern found at these sites include
PCBs, dioxins and furans, mercury, pesticides, and other chemicals identified as persistent
organic pollutants. State health departments have established fish consumption advisories at
many of these water bodies.
The 1989 RAGS Part A guidance (USEPA, 1989a) recognized the importance of
evaluating the potential cancer risks and non-cancer health hazards associated with consuming
contaminated fish associated with Superfund sites. Page 6-6 of RAGS Part A specifically states:
... Be sure to include potentially exposed distant populations, such as public water supply consumers and
distant consumers offish or shellfish or agricultural products from the site area. ...
Further, on page 6-7, the guidance states:
... Identify any site-specific population characteristics that might influence exposure. For example, if the
site is located near major commercial or recreational fisheries or shell fisheries, the potentially exposed
population is likely to eat more locally-caught fish and shellfish than populations located inland. ...
-------
Page 112 Examination of EPA Risk Assessment Principles and Practices
A variety of data sources are available to evaluate fish ingestion rates. Fish ingestion
studies can be either creel surveys, in which anglers are interviewed in person while fishing, or
mail surveys, in which anglers (often identified as people with fishing licenses) are sent
questionnaires in the mail (reviewed in USEPA, 1992c). Creel surveys typically involve
interviews with anglers at the dockside requesting information about their fishing activities (fish
preference, consumption rates, cooking methods, age, gender, frequency of fishing the specific
water body, etc.)- This survey method can provide information on both licensed and unlicensed
anglers, depending upon who is interviewed. Mail surveys, meanwhile, typically involve sending
questionnaires to licensed anglers requesting information on fishing practices; preferred rivers,
lakes, or streams; fish consumption; and other information. If mailing addresses are obtained
from lists of licensed anglers, unlicenced anglers are not represented. A third type of survey,
diary surveys in which participants are asked to record the frequency offish ingestion, the
types offish eaten, and the meal size require more effort from survey participants, but are
generally assumed to yield more accurate results because they minimize the potential recall bias
found in the other survey methods.
Any of these methods may be appropriate to evaluate the impact of Superfund site
contamination on individuals consuming fish from contaminated water bodies impacted by a site.
The approaches used in assessing fish consumption rates at sites vary based on the availability of
creel surveys, the size of the impacted water body (e.g., a creek as opposed to a large river
system), the available information on fishing practices within the specific water body, and the
time and resources necessary to conduct a site-specific survey of fish consumption patterns.
The risk assessments include an evaluation of the RME and CTE individuals. In the
absence of site-specific fish consumption rates, the default values from the 1997 Exposure
Factors Handbook (USEPA, 1997d) may be used. The handbook makes a number of
recommendations based on 95th and 50th percentile fish ingestion rates for specific water bodies,
geographic locations, and fish species (USEPA, 1997d). Table 5-2, below, provides some of
those recommendations.
-------
Examination of EPA Risk Assessment Principles and Practices Page 113
Table 5-2: Comparison of Default Fish Ingestion Rates
Default
1997 EPA Exposure Factors
Handbook
General population
Long-term intake
distribution
Total fish
Marine fish
Freshwater/estuarine
1997 EPA Exposure Factors
Handbook (recreational
marine anglers)
Atlantic
Gulf
Pacific
1997 EPA Exposure Factors
Handbook (freshwater)
Maine
New York
Michigan
Michigan
Ingestion
Rate
(grams/day)
63
20.1
14.1
6.0
18.0
5.6
26.0
7.2
6.8
2.0
13
5
18
5
39
12
17
Percentile
95th
50th
50th
50th
95th
50th
95th
50th
95th
50th
95th
50th
95th
50th
95th
50th
95th
50th
Data Source
USEPA, 1997d, based on
TRI (Javitz, 1980; Ruffle et
al., 1994)
USEPA Analysis of CSFII,
1989-1991
USEPA, 1997d;NMFS,
1993
EbertetaL, 1993
Connelly et al., 1996
Westetal., 1989
Westetal, 1993
-------
Page 114 Examination of EPA Risk Assessment Principles and Practices
Species-specific intake information maybe derived from a state-wide angler survey
(Connelly et al., 1992) to calculate the concentration of PCBs ingested in fish. For example, at a
PCB site, EPA determined that each species offish has a characteristic PCB concentration and
the average concentration an angler consumes will, in part, be based on the relative percentages
of the different fish species consumed. This analysis yields both central tendency and RME
estimates for each of the parameters needed to calculate the cancer risks and non-cancer health
hazards. A site-specific analysis using this database found that, for adults, the central tendency
fish ingestion rate was 4.0 grams per day, or about six half-pound meals per year; the RME fish
ingestion rate was 31.9 grams per day, or about 51 half-pound meals per year for adults. Fish
ingestion rates for adolescents and young children were reduced based on the ratio of adolescent
or child body weight to that of an adult.
Other risk assessments have used studies from the published literature that are
representative of the relevant site-specific consumption patterns. Risk assessors review the
literature to determine how well it represents the site they are evaluating. In determining the
literature's appropriate use, they may evaluate the study design, hypotheses being tested, number
of individuals interviewed within the survey, rates of consumption, and populations studied.
In several other examples, site-specific surveys have been conducted to evaluate the
consumption patterns for specific populations that the published surveys do not capture. These
surveys found considerably higher consumption rates among these populations than if the
standard default assumptions from the 1997 Exposure Factors Handbook were used (USEPA,
1997d). For example, a 314-year site-specific creel survey (Toy et al., 1996) included
information on whether or not adults harvested fish and shellfish from Puget Sound. The survey
included 190 adults and 69 children between the ages of 0 and 6. The study found that tribal
seafood consumption rates were considerably higher than Exposure Factors Handbook values.
Among the Squaxin, the average consumption rate was 72.8 grams per day and the 90th percentile
ingestion rate was 201.6 grams per day. Among the Tulalips, the average consumption rate was
72.7 grams per day and the 90th percentile was 192.3 grams per day. Other site-specific
consumption surveys found similar differences in consumption rates (APEN, 1998; USEPA,
2001h; Sechena et al., 2003).
In cases where EPA has conducted individual surveys to identify fish consumption rates,
EPA has found it important to include the community in the process (USEPA, 1999f). EPA and
other agencies (both private and governmental) have spent considerable resources and time to
plan and implement these studies. The surveys (APEN, 1998; USEPA 2001h; Sechena et al.,
2003) were all conducted using one-on-one interviews, as opposed to creel or mail surveys. The
people conducting the interviews were always specially trained members of the ethnic group or
community being surveyed. (EPA realized early on that creel and mail surveys do not necessarily
work for tribes and other such groups. One-on-one interviews involving the affected
communities are more appropriate.)
-------
Examination of EPA Risk Assessment Principles and Practices Page 115
The determination of the need to conduct a creel survey or use site-specific information is
dependent on a number of site-specific factors. These include available information on
consumption patterns, whether an existing consumption advisory may bias responses when
surveys are conducted, time, and resources.
5.5 Other Factors in Superfund Assessments
5.5.1 What Is the Role of Stakeholders in Superfund Assessments?
EPA works with members of the community to understand their concerns at the national
and site-specific levels. In 1996, EPA conducted meetings with the communities to discuss the
Superfund process and their concerns (USEPA, 1996c). Under the 1996 RAGS Reform process,
EPA met with community members at several meetings to discuss their concerns and identify
priorities for new guidance documents. At these meetings, the communities identified ideas for
improving communication between the regulators and the community. This resulted in guidance
for involving communities in the risk assessment process (USEPA, 1999f).
Within the Superfund program, EPA conducts extensive outreach with stakeholders at the
site. As described in the guidance on community involvement (USEPA, 1999f, 2000n), at each
site the remedial project manager, the community involvement coordinator, and other members
of the team including the risk assessor regularly communicate with the members of the
community, including the potentially responsible parties. Each site has a Community
Involvement Plan, developed following interviews with the community, that outlines
mechanisms by which the community will be involved in the process.
EPA guidance and educational materials help illustrate the ways that citizens can be
involved in the risk assessment process (USEPA, 1999f, h). For example: Community-specific
information on fishing preferences helped to identify exposure areas for sampling and fish
species consumed by people who fish in a contaminated bay. Information from farmers on
pesticide applications helped EPA determine why certain contaminants were present in an
aquifer. Discussions with farmers about certain harvesting practices helped EPA refine exposure
models and assumptions at another site (USEPA, 1999h).
EPA uses a range of communication tools to include the community in the Superfund
process. These include newsletters, fact sheets, site-specific home pages, public meetings, public
availability sessions, and 1-800- numbers to contact EPA staff. EPA strives to communicate
information about the RI, the results of the risk assessment, proposed actions at the site, and the
proposed and final decisions for remedial actions. The Record of Decision (ROD) includes a
responsiveness summary that addresses comments including those from the community. During
the period of the remedial action, communication with the community continues, including
updates during the 5-year review process.
-------
Page 118 Examination of EPA Risk Assessment Principles and Practices
Table 5-3: Nine Evaluation Criteria for Superfund Remedial Alternatives
Threshold Criteria
Overall protection of human health and the environment determines whether an alternative
eliminates, reduces, or controls threats to public health and the environment through
institutional controls, engineering controls, or treatment.
Compliance with ARARs evaluates whether the alternative meets federal and state
environmental statutes, regulations, and other requirements that pertain to the site, or whether
a waiver is justified.
Primary Balancing Criteria
Long-term effectiveness and permanence considers the ability of an alternative to maintain
protection of human health and the environment over time.
Reduction oftoxicity, mobility, or volume of contaminants through treatment evaluates an
alternative's use of treatment to reduce the harmful effects of principal contaminants, their
ability to move in the environment, and the amount of contamination present.
Short-term effectiveness considers the length of time needed to implement an alternative and
the risks the alternative poses to workers, residents, and the environment during
implementation.
Implementability considers the technical and administrative feasibility of implementing the
alternative, including factors such as the relative availability of goods and services.
Cost includes estimated capital and annual operation and maintenance costs, as well as present
worth cost. Present worth cost is the total cost of an alternative over time in terms of today's
dollar value. Cost estimates are expected to be accurate within a range of+50% to -30%.
Modifying Criteria
State acceptance considers whether the state agrees with the EPA's analyses and
recommendations, as described in the RI/FS and Proposed Plan.
Community acceptance considers whether the local community agrees with EPA's analyses
and preferred alternative. Comments received on the Proposed Plan are an important indicator
of community acceptance.
-------
Examination of EPA Risk Assessment Principles and Practices Page 119
5.6 Superfund Site-Specific Conclusions
In general, the information applied in EPA Superfund risk assessments and the level of
protectiveness are considered in the context of the regulatory program. From the previous
discussion, the following conclusions are highlighted.
a) Reasonable maximum exposures. For the Superfund program, EPA bases
decisions on current and future risks associated with reasonable high-end
exposures or RME, not only the average exposures. By definition, the RME is
within a realistic range of exposures, not a "worst-case" estimate.
b) Default assumptions. Risk estimates are based on a combination of site-specific
information, when available and appropriate, and general default exposure factors
developed for Superfund baseline risk assessments. The default factors are based
on published scientific data and are designed to be reasonably protective of the
majority of those exposed. The use of defaults also streamlines risk assessments,
reduces unwarranted variability in the exposure assumptions used to characterize
exposures, and provides reasonable risk estimates where there is a lack of
site-specific data or consensus on which value to choose. For some exposure
parameters, default values are used because the costs of collecting data for
specific behaviors in a population may be prohibitive.
c) Site-specific information. Information submitted to the EPA regions or collected
by EPA is critically evaluated to determine how to apply it in the risk assessment.
Anecdotal information is valuable but not sufficient for quantifying exposures. It
may, however, be useful to consider such information when developing a
conceptual site model and characterizing uncertainties. In some cases, site-
specific data would be preferable to using default exposure data that represent the
majority of the population at the high end of the distribution for specific activity
patterns. However, the costs of collecting data for specific behaviors in a
population may be prohibitive. For example, a site-specific creel survey may cost
$100,000 or more and take several years to conduct (Toy et al., 1996; USEPA
2001h; Sechena et al., 2003).
Specialized studies may reduce uncertainties in a site-specific risk assessment;
however, the region considers the sensitivity of the parameter in question and the
added value of the information in light of uncertainties of other parameters,
including the toxicity information. In some cases, a research study of exposure
will not significantly reduce the uncertainty of the risk assessment when the
toxicity information is very limited (USEPA, 1992a).
-------
Page 120 Examination of EPA Risk Assessment Principles and Practices
d) Peer review. Consistent with Agency guidance on peer review, external peer
reviews are conducted for specific risk assessments and guidance documents.
e) Probabilistic risk assessment. Superfund has developed specific guidance for
conducting site-specific probabilistic risk assessments, and this guidance was
externally peer reviewed. At the present time, site-specific probabilistic risk
assessments have been conducted at several sites to examine exposure
assessments.
f) Guidance documents. The Superfund program continues to update guidance to
include new science where appropriate. The guidance documents are made
available for public review and external peer review.
g) Transparency and clarity in risk assessments. Superfund developed specific
guidance for improving the clarity and transparency of risk assessments. This
guidance covers community involvement, standardization of reporting of risk
information, and standardization of ROD information (USEPA, 1999f, 2000n).
EPA's risk assessment practices and policies apply a mix of site-specific information and
science-based default assumptions that are consistent with the goals of the Superfund program.
Site-specific considerations are addressed at the EPA site-specific level. In turn, the regions
inform EPA program offices of the research needed to reduce uncertainties in risk assessments.
There will always be a need for additional research on exposure and toxicity parameters to
improve understanding of how people contact their environments, chemical toxicity, and
chemical fate and transport.
5.7 Chemical Mixtures Risk Assessment Methods and Practice
5.7.1 What Are the Issues Regarding Chemical Mixture Site Assessments?
EPA published guidance on the health risk assessment of chemical mixtures (USEPA,
1986b; 2000h) and its use in site assessments (USEPA, 1989a; 1999a). As part of the
implementation of the 1996 Food Quality Protection Act, EPA also developed guidance for
conducting cumulative risk assessments of chemicals that appear to act by a common mechanism
of toxicity (USEPA, 2002f). EPA (USEPA, 2000h) guides the risk assessor through an
examination of data relevant to a mixture of concern to facilitate selection of an appropriate
mixture risk assessment method. Suggested methods include both whole-mixture approaches
and component-based algorithms that include interaction data. A comment regarding EPA's
practice is that the Agency "assumes the toxicity of a chemical mixture is equal to the sum of the
toxicity of each individual chemical, regardless of the toxicity type, competition or antagonism
among chemicals." This comment apparently assumes that:
-------
Examination of EPA Risk Assessment Principles and Practices Page 121
a) EPA uses only models of mixture toxicity based on dose addition,
b) EPA inappropriately applies dose addition by ignoring the requirement that the
mixture's chemical components share a common toxic mode of action (MOA),
c) EPA ignores interactions data that may reduce the toxicity of a mixture through
competition or antagonism, and
d) EPA never determines the toxicity of a whole mixture directly.
These assumptions do not reflect current guidance, as shown by the Agency's mixtures
guidelines (USEPA, 1986b) and supplementary guidance (USEPA, 2000h). These documents
offer detailed descriptions of a variety of component-based and whole-mixture methods far
beyond dose addition, provide instructions on appropriate application of the methods, and
recommend consideration of interaction data whenever possible. Further, program guidance
(USEPA, 1989a) specifies the use of component-based approaches to evaluate chemical
mixtures. The following sections explain the theory behind these procedures and discuss current
practice.
5.7.2 What Are Dose Addition and Response Addition?
To address low exposure levels when no interaction information or whole-mixture
toxicity data are available, it is EPA's practice to use response addition and dose addition as the
recommended default methods when the component chemicals in a mixture show dissimilar
toxicity and similar toxicity, respectively (USEPA, 2000h). Dose addition and response addition
are fundamentally different methods, based on different assumptions about toxicity. Because
both methods are relatively easy to apply and use single chemical toxicity and exposure
information, they are the methods most commonly used in site-specific assessments (USEPA,
1989a). The two additivity assumptions are further described as follows:
a) Dose addition sums the doses of the components in a mixture after they have been
scaled for toxic potency relative to each other. The risk of a toxic effect is
determined from this summed dose. Dose addition requires the components to
share a common toxic MOA.
Superfund site assessments have applied dose addition in the form of a hazard
index (HI) to evaluate sites for indications of health risk (USEPA, 1989a). The HI
is calculated as the sum of hazard quotients (HQs) for the chemical components of
the mixture; thus, it indicates risk, but is not an explicit risk estimate. A HQ is
typically calculated as the ratio of a chemical's exposure level to its safe level,
such that values larger than 1 are of concern. For a group of n chemicals in a
-------
Page 122 Examination of EPA Risk Assessment Principles and Practices
mixture with exposure levels (E) and using the RfD as a safe level, the HI is
calculated as:
HI =
R
- V
!'=!
Equation 5.7-1
The HI is usually calculated for groups of chemicals whose effects are observed in
a common target organ. As an initial screening step, though, risk assessors
calculate His by combining across health endpoints, contradictory to the
assumption of common MOA. If the HI across all effects indicates no risk (i.e., is
less than 1), then the assessment can stop. If risk is indicated (i.e., the HI is
greater than 1), the risk assessor repeats the analysis by breaking the mixture into
groups of chemical components with a common target organ and recalculating the
HI for each smaller group.
b) Response addition first estimates the probabilistic risk of observing a toxic
response for each chemical component in the mixture. Then the component risks
are summed to estimate total risk from exposure to the mixture, assuming
independence of toxic action (i.e., that the toxicity of one chemical in the body
does not affect the toxicity of another chemical). This can be thought of as an
organism receiving two (or more) independent insults to the body, so that the risks
are added under the statistical law of independent events.
Superfund site assessments calculate total cancer risk by summing the individual
cancer risks for the carcinogens in the mixture (USEPA, 1989a). For example, the
mixture risk (Rm) for two chemicals is the sum of the risks for chemical one (r,)
and chemical two (r2) minus the probability that the toxic event from exposure to
chemical one would overlap in time with the toxic event from exposure to
chemical two, as follows:
= r\
Equation 5.7-2
Because cancer risks are typically very small (i.e., in the 10"4 to 10"6 range), the
amount of risk subtracted in Equation 5.7-2 (e.g., an amount in the 10"8 to 10"12
range) is insignificant and is therefore usually ignored. Risks are appropriately
aggregated for cancers across various target organs because the result is
interpreted as the risk of any cancer, and the cancers from each chemical
component are considered to be independent events in the body.
-------
Examination of EPA Risk Assessment Principles and Practices Page 123
5.7.3 How Are Dose Addition and Response Addition Applied in Practice?
To get a general idea of how dose addition and response addition are actually practiced in
EPA's site assessment work, a search of EPA's RODs in the Comprehensive Environmental
Response, Compensation, and Liability Information System (CERCLIS) database was conducted
in August 2003 ('http://www.epa.gov/superfund/sites/siteinfo.htni'). keying on the term "hazard
index." From that list, 10 RODs were selected from across the Agency. These RODs were
examined for tables and text relevant to the application of response addition and the Hazard
Index (HI). RODs were used as a surrogate for the complete risk assessments since only a
limited number of risk assessments are available on the internet. It is noted that, as stated in the
guidance for preparing RODs, the full risk assessment need not be included in the ROD
document (USEPA, 1999a). The risk assessments are available in the Administrative Record for
each site and, based on the time limits of this analysis, it was not possible to evaluate each of the
individual site records.
To estimate cancer risks using response addition, the procedure is to sum risk estimates
across all carcinogenic modes of action, assuming independence of action. There is no
requirement to demonstrate independence empirically; however, when component exposures are
low, such biological independence is likely. Most of the examples appear to have appropriately
applied response addition, although in several cases, this conclusion is inferred from the
available text. In one instance, the technique was not used because there was only one
carcinogen found at the site. In two other cases, additional information from the risk assessment
would be needed to determine exactly how the evaluations were completed.
To apply the HI, the procedure in the guidance is to first perform a screening-level
analysis, summing Hazard Quotients (HQs) across all target organs. If the HI is below 1 across
all target organs, then the analysis is complete; this was the case for tap water risks in one
example. If the HI is above 1 across all target organs, then the analysis is to be repeated by
summing the HQs for only those mixture components that affect the same target organ; this
occurred in the three of the examples. The purpose of the re-analysis is to comply with the dose
addition assumption of a common toxic mode of action. In one case, the HI was shown in the
ROD to change from greater than 1 in the screening analysis to less than 1 under the re-analysis.
In two instances, the HI was greater than 1 in the screening analysis, but the ROD did not show
the re-analysis by target organ; in these cases, however, the individual chemical HQs were much
larger than 1, making all HI calculations in these cases moot (i.e., no HI below 1 could be
calculated under any constraints). Finally, in four of the examples, additional information would
be needed to determine exactly how the evaluations were completed.
In conclusion, the results from this limited analysis are generally positive, illustrating that
EPA practice is consistent with its own guidance regarding chemical mixtures risk assessment.
The clarity of EPA's risk communications could be improved by transferring additional
-------
Page 124 Examination of EPA Risk Assessment Principles and Practices
information from the risk assessment in the Administrative Record for each site to its associated
ROD.
5.7.4 Are Interaction Data Used in Mixture Risk Assessment?
EPA defines interactions as observed effects greater than (synergism) or less than
(antagonism) those expected under a specified form of additivity. EPA (USEPA, 1986b, 2000h)
also states that mixture risk assessment should reflect known toxicological interactions. In
practice, sufficient data are seldom available to model interactions. Data on binary combinations
of chemicals are most prevalent, so for mixtures of more than two chemicals, the true nature of
joint toxic action may be speculative at best. For exposures at low doses with low component
risks, the likelihood of a significant interaction is usually considered to be low. Interaction
arguments based on saturation of metabolic pathways or competition for cellular sites usually
imply an increasing interaction effect with dose, so that the importance for most low
environmental exposures is probably small.
EPA mixture risk assessments have usually considered the information on interactions in
a qualitative sense. For example, a Superfund site may receive more scrutiny or its remediation
may proceed faster if there are indications of potential synergism among the detected chemicals.
The cleanup goals and the estimated risk would not change, though, because the synergism could
not be quantified. No standard methods or mathematical models are yet in common use in
regulatory agencies to incorporate interactions and to serve as defaults. There is one newly
developed procedure, the interaction-based HI (USEPA, 2000h), that allows for numerical
adjustments to the HI based on evidence of synergism or antagonism for pairs of chemicals in the
mixture. This method is currently undergoing verification using laboratory data.
Despite the difficulties in using interaction information, some regional assessments have
considered relevant data. An interesting assessment by Region III explicitly evaluated synergy
and antagonism of multiple metals in soil (USEPA, 2002m). Text from that ROD reads as
follows:
EPA (1994b; [this reference is in the ROD]) prepared an extensive review of interactions among cadmium,
lead, and zinc reported in the open literature. Available evidence does not support a change in assumed
absorption of metals due to interactions among metals following ingestion of contaminated soils. The
quantitative risk assessment does not alter the estimates of exposure based on co-exposure to cadmium,
lead, and zinc. Further, available information suggests that any potential impact of cadmium on
neurotoxicity of lead in young children is largely speculative. No modification to the assessment of lead
toxicity is thus justified based on co-exposure to cadmium.
5.7.5 Are Data on Whole Mixtures Used in Risk Assessment?
Risk assessments based on tests of whole mixtures or on epidemiologic data determine
combined effects empirically. Examples of these (USEPA, 2003g) include (1) RfDs on
-------
Examination of EPA Risk Assessment Principles and Practices Page 125
commercial PCB mixtures (Aroclors 1016 and 1254) based on primate data, and (2) a CSF for
coke oven emissions based on human occupational exposures. Drinking water disinfection
byproducts represent a complex mixture for which epidemiologic data suggest potential health
risks (USEPA, 2003b). An EPA study, called the 4 Lab project, is currently underway to
toxicologically and chemically characterize this complex mixture to produce data on
reproductive and developmental effects in rats exposed to concentrations of this complex mixture
for use in risk assessment (Simmons et al., 2002).
Whole-mixture studies are routinely used in ecological risk assessments. The Agency has
developed subchronic toxicity tests for whole aqueous effluents and for contaminated ambient
waters, sediments, and soils (USEPA, 1989d, 1991d, 1994e). Further, the effects of mixtures in
aquatic ecosystems are evaluated using bioassessment techniques that are equivalent to
epidemiology, but more readily employed (USEPA, 1999i). Similar bioassessment methods are
sometimes used at Superfund sites (USEPA, 1994f). These empirical approaches to assessing
ecological risks from mixtures are employed in National Pollutant Discharge Elimination System
permitting and the development of Total Maximum Daily Loads, and are often used in Superfund
baseline ecological risk assessments.
5.7.6 What Other EPA Applications Exist for Mixtures Risk Assessment?
EPA program offices and regional risk assessors have a great need for both assessment
information and risk assessment methods to evaluate human health and ecological risks from
exposure to chemical mixtures. Further, several pieces of legislation exist that require EPA to
consider the evaluation of chemical mixtures (see below). The Agency conducted some specific
mixture risk assessments in the past and continues to pursue additional evaluations. Currently,
an IRIS assessment is being developed for the PAHs (USEPA, 2002n). In addition, EPA used a
dose additive approach to assess PCBs (USEPA, 1996d), organophosphorus pesticides (USEPA,
2002g), and dioxins (USEPA, 1989b; 2000i). (See section 4.5 for a discussion of this approach.)
Examples of other mixtures of interest to the Agency include creosote, particulate matter,
drinking water disinfection byproducts (DBPs), brominated flame retardants, mixtures of metals,
and pesticide mixtures in soils. Current Agency interests and supporting legislation include:
a) OW is concerned with contaminant mixtures in drinking water in response to
requirements of the Safe Drinking Water Act Amendments of 1996, including
mixtures of DBPs and of Contaminant Candidate List chemicals (e.g., organotins,
pesticides, metals, Pharmaceuticals). Information and methods are being
developed to better evaluate the toxic mode of action, the risk posed by drinking
water mixtures, exposure estimates for mixtures via multiple routes, and the
relative effectiveness of advanced treatment technologies (USEPA, 2003e, h).
-------
Page 126 Examination of EPA Risk Assessment Principles and Practices
b) The Office of Air Quality Planning and Standards used a mixture risk assessment
approach for multiple route exposures in conducting the National Air Toxics
Assessment of 33 air pollutants (a subset of 32 air toxics from the Clean Air Act's
list of 188 air toxics plus diesel particulate matter) (USEPA, 2001J).
c) The Office of Air and Radiation issued a proposed rule to evaluate the national
emission standards for HAPs for stationary combustion turbines, suggesting the
application of His to evaluate potential health risks (USEPA, 2003i).
d) The Officer of Pesticide Programs conducted a multiple-route assessment of
organophosphorus pesticide mixtures (USEPA, 2002g)s in response to the FQPA
of 1996. Future assessments may be performed on additional pesticide classes
and other co-occurring substances for which a common toxic mode of action can
be identified.
e) The National Homeland Security Research Center is considering the potential
toxicological interactions with co-exposure to respiratory toxicants, dust, and
smoke.
f) The Office of Solid Waste and Emergency Response evaluates contaminant
mixtures at Superfund sites (USEPA, 1989a) under CERCLA (see sections 5.7.2
and 5.7.3 above). Examples of current needs and research areas are the
assessment of mixtures of metals and guidance on how to group the
environmental contaminants most commonly found together at sites into common
mode of action classes for use in developing His or assessments based on RPFs.
-------
Examination of EPA Risk Assessment Principles and Practices Page 127
6. ECOLOGICAL ASSESSMENT
6.1 Overview
The early 1980s saw both the emergence of risk assessment as a regulatory paradigm and
the first widespread use of ecological impacts to inform regulatory and policy decisions. The use
of ecological information in decision making has expanded slowly through the 1980s, as
illustrated by the cancellation of the pesticide diazinon based on impacts to birds (Bascietto,
1998), the reductions in sulfur dioxide emissions to lessen the adverse impacts of acid deposition
on lakes and forests and the setting of ozone standards to reduce the damaging effects of
tropospheric ozone on crops. In the middle to late 1980s, tools and methods for conducting
ecological risk assessments began to be standardized with the publication of EPA's Ambient
Water Quality Criteria (WQC) methodology, the pesticide program's Standard Evaluation
Procedures, and Superfund's Environmental Evaluation Manual (these are discussed in Bascietto
et al., 1990). More recently, EPA published the Ecological Risk Assessment Guidance for
Superfund (USEPA, 19971) and the Guidelines for Ecological Risk Assessment (USE? A, 1998a).
In 1992, the Agency published the Framework for Ecological Risk Assessment (USEPA,
1992d; Norton et al., 1992) as the first statement of principles for ecological risk assessment.
That document and the 1998 first draft Guidelines for Ecological Risk Assessment not only
describe methods for conducting the more conventional single-species, chemical-based risk
assessments, but also describe techniques for assessing risks to ecosystems from multiple
stressors and multiple endpoints. With the publication of these important documents came the
need to enhance EPA's ability to do better ecological assessments. These important publications
recognized the need to advance the science of multiple-scale, multiple-stressor, and multiple-
endpoint ecological assessments and, more important, called for the consideration of human
dimensions in the planning and conduct of ecological risk assessments (Bachmann et al., 1998;
Barton and Sergeant, 1998; Cooper, 1998; Linthurst et al., 2000).
Following EPA's paradigm, an ecological risk assessment evaluates the potential for
adverse ecological effects resulting from human activities. The risk assessment process provides
a way to develop, organize, and present scientific information so that it is relevant to
environmental decisions. When conducted for a particular place such as a watershed, the
ecological risk assessment process can be used to identify vulnerable and valued resources,
prioritize data collection activity, and link human activities with their potential effects. Risk
assessments can also provide a focal point for cooperation between local communities and state
and federal government agencies. Risk assessment results provide a basis for comparing
different management options, enabling decision makers and the public to make more informed
decisions about the management of ecological resources.
-------
Page 128 Examination of EPA Risk Assessment Principles and Practices
Ecological risk assessments have become a common subject of discourse at EPA. A
summary of the evolution of this discourse is described by Bascietto et al. (1990), Maki and
Slimak (1990), Kutz et al. (1992), Norton et al. (1992), Bachmann et al. (1998), and Linthurst et
al. (2000). Beginning in 1984, the concepts of risk assessment and risk management became a
common language for justifying regulatory proposals throughout EPA. In 1987 a major Agency
report stated that the "fundamental mission of EPA is to reduce risks," and another in 1990
recommended that EPA "target its environmental protection efforts on opportunities for the
greatest risk reduction" (USEPA 1987b, 1990). And, as described previously, the rhetoric of risk
became the Agency's primary language for justifying its decisions (Russell and Gruber, 1987).
6.1.1 What Is the EPA Ecological Risk Assessment Approach?
Ecological risk assessment evolved out of the human health risk assessment process, an
approach to understanding and protecting humans from the threats predominately cancer
from exposure to chemicals. The National Academy of Sciences' National Research Council's
seminal document on risk assessment (NRC, 1983) defined risk assessment as assessment of the
"probability that an adverse effect may occur as a result of some human activity." This report,
and the articulation by EPA Administrator Ruckelshaus (Ruckelshaus, 1983) that risk was to be
EPA's defining operational paradigm, ushered the Agency into a period in which risk assessment
and risk management became the primary discourse among its staff. Recognizing that its
definition of risk is human-oriented, the report recommended that the EPA develop a counterpart
for non-human or ecological endpoints. Thus began a lengthy process within EPA to
complement the human risk assessment process with procedures for wildlife, ecosystems, and
endangered species, all of which fall into the category of ecological risk assessment. The EPA
relied on scientific and technical deliberations among experts in the field to complete the
Framework for Ecological Risk Assessment (USEP A, 1992d; Norton etal, 1992). The
Framework defined ecological risk assessment as "a process that evaluates the likelihood that
adverse ecological effects may occur or are occurring as a result of exposure to one or more
stressors." EPA subsequently published case studies (USEPA, 1993b, 1994g) and a companion
document on issues relating to the elements of an ecological risk assessment (USEPA, 1994h).
Then (as mentioned above) EPA published an Ecological Risk Assessment Guidance for
Superfund in 1997 (USEPA, 1997Ff) and Guidelines for Ecological Risk Assessment in 1998
(USEPA, 1998a).
According to EPA (USEPA, 1998a) and others (Suter, 1993; Lackey, 1994), an ecological
risk assessment is a flexible process for organizing and analyzing data, information, assumptions,
and uncertainties to evaluate the likelihood of adverse ecological effects. Ecological risk
assessment provides a critical element for environmental decision making by giving risk
managers an approach for considering available scientific information along with the other
factors they need to consider (e.g., social, legal, or economic factors) in selecting a course of
action. An ecological risk assessment includes an initial planning step, problem formulation,
-------
Examination of EPA Risk Assessment Principles and Practices Page 129
analysis of stressors and effects, and risk characterization. Problem formulation, the initial step
in analysis, involves defining and specifying the issue under consideration. Analysis is separated
into exposure (stressor) and effects (hazard) analysis. Risk characterization includes both
quantitative estimation of risk and a qualitative description of risk. The final step is
communicating the results to a manager and stakeholders. Important properties of the
assessment include data gathering and review during and after each step and the possibility of
repeating steps if new data become available.
Ecological risk assessments at EPA are generally characterized by a particular emphasis
at the problem formulation phase. The interface among risk assessors, risk managers, and
interested parties during planning at the beginning of the risk assessment, as well as
communication of risk at its end, is critical to ensure that the results of the assessment can be
used to support a management decision. Because of the diverse expertise required, especially in
complex ecological risk assessments, risk assessors and risk managers frequently work in multi-
disciplinary teams during problem formulation. This approach has important implications for
how an ecological risk assessment is formulated, since it requires an up-front discussion on what
is at risk, what the assessment endpoints are, how they are measured, and what unacceptable risk
levels are. This discussion and its outcome are shaped by the values, beliefs, and attitudes of the
risk assessors, risk managers, and stakeholders. Those conducting ecological risk assessments
have learned that risk managers, risk assessors, and stakeholders bring valuable perspectives to
the initial planning activities for an assessment. Risk managers charged with protecting the
environment can identify information they need to develop their decisions, risk assessors can
ensure that science is effectively used to address ecological concerns, and stakeholders bring a
sense of realism and purpose. Together, all can evaluate whether a risk assessment can address
the identified problems.
6.2 Organism-Level Versus Population-Level Ecological Risk Assessments
The Agency received comments expressing concern about the use of organism
(individual) level attributes to assess ecological risk. An example comment follows:
EPA (1997) ecological risk assessment guidance recommends that potential ecological risks should be
assessed at the population-level for all but threatened and endangered species, [note added here: This
reference and accompanying citation appear to be incorrect, in that that document does not make such a
recommendation.] Although no explicit guidance is provided, this is typically accomplished through the use
of measurement endpoints that are related to population effects (e.g., using Toxicity Reference Values
based on growth or reproductive effects), However, in many EPA ecological risk assessments, the agency
has defaulted to assessing effects on individual animals, This metric has no significance scientifically and is
entirely useless as a basis for making risk management decisions.
The above statement cites text from EPA documents emphasizing the importance of
protecting populations or communities. The Office of Solid Waste and Emergency Response's
guidance for Superfund sites (USEPA, 1999J) states:
-------
Page 130 Examination of EPA Risk Assessment Principles and Practices
Ecological risk assessments incorporate a wide range of tests and studies to either directly estimate
community effects (e.g., benthic species diversity) or indirectly predict local population-level effects (e.g.,
toxicity tests on individual species), both of which can contribute to estimating ecological risk. Superfund
remedial actions generally should not be designed to protect organisms on an individual basis (the exception
being designated protected status resources, such as listed or candidate threatened and endangered species
or treaty-protected species that could be exposed to site releases), but to protect local populations and
communities of biota. Levels that are expected to protect local populations and communities can be
estimated by extrapolating from effects on individuals and groups of individuals using a lines-of-evidence
approach.
As an initial matter, the comment may be more properly considered as a risk management
decision to be made in the problem formulation stage of an ecological risk assessment. The cited
text notes the important role that measurements of organism-level effects can play in
extrapolations to population-level effects. Where possible, toxicity tests are performed using
measurement endpoints that are related to population effects (e.g., measurements of growth rate
or reproduction). We agree that it would be desirable to define the full range of ecological
benefits by considering effects on populations, communities, and ecosystem processes as well as
organisms. But it is also important to recognize that the use of organism-level attributes such as
mortality as endpoints for ecological risk assessments is well-supported by policy and precedent
(USEPA, 1994i, 2003J) and by the courts e.g., in the diazinon decision (USEPA, 1988c; Ciba
Geigy v. EPA, 874 F.2d 277 [5th Cir. 1989])1. In addition, some laws including the Marine
Mammal Protection Act, the Migratory Bird Treaty Act, The Bald Eagle Protection Act, and the
Endangered Species Act call for protection of organism properties and even individual
organisms. Nonetheless, the protection of organism-level attributes is generally interpreted as
occurring in a population or community context (USEPA, 2003J). That is, increased mortality or
decreased fecundity or growth of organisms in an assessment population or community is
assumed to be significant, even with no demonstration that a population- or community-level
property is affected.
That assumption is necessitated by the extreme difficulty of predicting effects at higher
ecological levels. Most population-level attributes, including abundance attributes, are
determined by the vital rates (births, deaths) of individuals within the population, as well as the
rates of migration into and out of the population. Linkages between effects on vital rates and
those on population dynamics, or between effects on the biochemical and physiological processes
that determine vital rates and effects on populations, can either be established empirically by
correlating responses at different levels of biological organization or by determining causal
On March 29, 1988, the Agency cancelled registrations for the pesticide diazinon unless amended to
prohibit use on golf courses and sod farms (USEPA, 198 8c). As noted in the remand decision (Ciba Geigy v. EPA,
874 F.2d 277 (5th Cir. 1989)), the circuit court rejected the industry's argument that a risk is unreasonable only if it
endangers bird populations, stating; "FIFRA gives the Administrator sufficient discretion to determine that recurring
bird kills, even if they do not significantly reduce bird population, are themselves an unreasonable environmental
effect."
-------
Examination of EPA Risk Assessment Principles and Practices Page 131
relationships and constructing mechanistic models (see Maltby et al, 2001, for a
conceptualization and discussion of these linkages). Thus, with some degree of understanding of
how changes in vital rates manifest into population consequences (i.e., mechanistic
understanding), effects measured at the organism level can be extrapolated to expected
population-level responses. Extrapolation of this latter type is usually accomplished using
models that integrate the effects of stressors on survivorship and fecundity. However, the
practicality of this approach is cast into question by experience which indicates that so many
unverifiable assumptions are involved that in real cases the very different outcomes of dueling
models cannot be resolved (Barnthouse et al., 1984). Also, it is impractical to estimate density
dependent responses in real populations, particularly when modeling the effects of toxicants.
The bottom line is that, although methods exist for predicting the effects of chemicals at
levels of organization higher than the organism, they are still in the development phase and have
not been shown to be reliable. On the other hand, assessment of ecological risks using measures
of organism-level effects is justified by experience, policy, and judicial decisions.
6.3 Conservatism and Ecological Risk Assessments
6.3.1 How Is Conservatism Addressed in Ecological Risk Assessments?
A number of ecological concerns expressed to the Agency revolve around the concept of
being overly conservative. In early tiers of a risk assessment (e.g., for screening), a high degree
of conservatism is sought. Definitive assessments strive to be as realistic as possible, replacing
conservative assumptions with best estimates of exposures and effects and associated
uncertainties. Three examples in which conservatism is addressed in ecological risk assessments
are discussed in the following sections.
6.3.2 How Are Toxicity Reference Values Developed in Ecological Risk
Assessments?
A number of comments have been received relating to the development of toxicity
reference values (TRVs) for use in ecological assessments. One of these notes that EPA
guidance recommends use of a ''weight-of-evidence" method that does not generate TRVs
appropriate for individual sites. The guidance that this comment refers to is the Ecological Soil
Screening Levels (Eco-SSLs) document (USEPA, 2003k). The intent of this document is
precisely to develop generic TRVs that are not site-specific: the TRVs are meant to be
conservative and to be used for screening purposes.
In the aquatic arena, the use of conservative threshold concentrations for ecological
protection is not new. EPA Water Quality Criteria (WQC), designed to protect aquatic
organisms and their uses specifically, 95% of the taxa from an appropriate variety of
-------
Page 132 Examination of EPA Risk Assessment Principles and Practices
taxonomic and functional groups (USEPA, 1985) have been employed in a regulatory context
for many years. The degree to which Eco-SSLs (derived from generic TRVs) may be more
conservative than chronic WQC can not be assessed, because of differences in their derivation.
Another comment was received relating to inconsistency in use of UFs in developing
avian TRVs for dioxin. At issue are the conditions under which a subchronic-to-chronic UF
should be used. Use of an UF where unwarranted would result in an overly conservative TRY,
Based on guidance in Sample et al. (1996), avian studies where exposure duration is 10 weeks or
less are generally considered to be subchronic. Sample et al., however, considered 10-week
exposure of ring-necked pheasant hens to TCDD in a study by Nosek et al. (1992a) to represent
chronic conditions, in part because exposure occurred through a critical life-stage (reproduction).
It must be noted, though, that TCDD's persistence and bioaccumulative properties make it an
exception to the rule that exposure during a critical life-stage is sufficient for consideration as
chronic exposure. Another study by Nosek et al. (1992b) found a half-life for whole-body
elimination of TCDD in non egg-laying pheasant hens of approximately 1 year. Based on this
information, only 13% of steady-state accumulation would be achieved from a 10-week
exposure. A truly chronic exposure could presumably have had nearly an order of magnitude
lower concentration in the food and still elicited the same tissue levels and effects (USEPA,
1993c). Thus, the 10-week exposure in the Nosek et al. study (1992a) can be considered to be
subchronic, and a UF of 10 used to derive a TRV. In summary, all available information is taken
into account when deciding on the use of an UF.
6.3.3 How Do Screening-Level and Definitive (Baseline) Assessments Differ?
The Agency has been criticized for using screening-level ecological risk assessments
rather than definitive "baseline" assessments in making remedial decisions. The former are
designed to be conservative, and are generally meant to be followed by more detailed assessment.
A screening-level ecological risk assessment provides a comparison of abiotic media
concentrations to ecotoxicological benchmarks, and does not include extensive site-specific
information. In a definitive assessment, exposure concentrations are derived for wildlife
receptors as well as for abiotic media. Definitive assessments require analysis of exposure-
response relationships for the chemicals of concern. The use of site-specific information to
derive appropriate exposure concentrations is an important element of a definitive assessment
and is discussed below.
6.3.4 How Is Site-Specific Information Used To Derive Exposure Concentrations
in Ecological Risk Assessments?
The Agency has been criticized for ignoring site-specific habitat information in a number
of ecological risk assessments for Superfund sites. The degree to which spatial and temporal use
of habitats by receptor species should be considered in the conceptual site model is an important
question. Distinct management areas, known as operable units or OUs, are often evaluated in the
-------
Examination of EPA Risk Assessment Principles and Practices Page 133
RI/FS process. It is important to have an understanding of the degree to which various habitats
within an OU are, or will be, used by receptor species. For example, if we know that fish move
throughout the area of an OU, exposure concentrations should be derived using data from the
entire OU. Exposure estimates are generated using point estimates of exposure parameters (e.g.,
upper 95 % confidence limits of the mean chemical concentrations in media). More realistic
estimates of exposure may be generated using distributions of exposure parameters and
uncertainty analysis (e.g., Monte Carlo simulation; see USEPA, 1997b, and the references in
chapter 7 of Suter et al., 2000). In the case of a large river system with tributaries that may be
"relatively" clean compared to the main channel, the decision as to whether or not to include
tributaries in the assessment typically depends on how contaminated they are and how they are
used by the receptor species.
6.4 Water Quality Criteria
6.4.1 What Is the Agency Standard Methodology for Deriving Aquatic Life Water
Quality Criteria?
Criteria are derived according on the principles set forth in the 1985 Guidelines for
Deriving Numerical National Water Quality Criteria for the Protection of Aquatic Organisms
and Their Uses (USEPA, 1985). These guidelines are ordinarily applied as follows;
a) Acute toxicity test data need to be available for species from a minimum of eight
diverse taxonomic groups. The diversity of tested species is intended to ensure
protection of various components of an aquatic ecosystem.
b) The Final Acute Value (FAV) is derived by extrapolation or interpolation to a
hypothetical genus more sensitive than 95% of all tested genera. The FAV, which
typically represents an LC50 or EC50, is divided by two in order to obtain an acute
criterion protective of nearly all individuals in such a genus.
c) Chronic toxicity test data (representing longer-term survival, growth, or
reproduction end points) need to be available for at least three taxa to be
calculated directly. When such data are not available, which is often the case, the
chronic criterion is set by determining an appropriate acute-chronic ratio (the ratio
of acutely toxic concentrations to chronically toxic concentrations) and applying
that ratio to the acute value of the hypothetical genus more sensitive than 95% of
all tested genera.
d) When necessary, the acute and/or chronic criterion may be lowered to protect
recreationally or commercially important species.
-------
Page 134 Examination of EPA Risk Assessment Principles and Practices
6.4.2 What Is the GLI Tier II Methodology?
The Tier II methodology for aquatic life is contained in the final Water Quality Guidance
for the Great Lakes System, or Great Lakes Initiative (GLI). The methodology was developed for
use in deriving translators for narrative criteria to protect aquatic life in waters of the Great Lakes
system when insufficient data are available to derive criteria. Tier II values may be based on
toxicity data from as little as a single taxonomic family, provided the data are acceptable. The
values are generally more stringent than criteria to reflect greater uncertainty in the absence of
additional toxicity data. As more data become available, the Tier II values tend to become less
conservative and more closely approximate the criteria. The general procedure is:
a) The lowest Genus Mean Acute Value in the database is divided by the Secondary
Acute Factor (an empirically derived adjustment factor that varies according to the
number of satisfied minimum data requirements listed in the criteria
methodology) to derive the Secondary Acute Value (SAV).
b) A Secondary Chronic Value (SCV) is calculated from either the FAV (from the
criteria) or the SAV divided by either the Secondary Acute to Chronic Ratio or
Final Acute to Chronic Ratio.
c) The Secondary Maximum Concentration is equal to half of the SAV, and the
Secondary Continuous Concentration is equal to the SCV or the Final Plant
Value, if available whichever is lower.
d) When necessary, the acute and chronic values can be adjusted to protect
recreationally or commercially important species.
The final rule allows and encourages dischargers, states, and authorized tribes to develop
additional data that could reduce the Tier n adjustment factors or allow development of WQC
(Tier I criteria) using the methodology described above. Additionally, in situations when
dischargers generate data to derive Tier I criteria or Tier H values, the final rule specifies that
permit authorities may grant a reasonable period of time for data generation before the criteria or
values become effective for the discharge permit.
6.4.3 Was the GLI Tier II Methodology Made Available for Public Comment or
Vetted Through Other Public Processes?
The Tier II methodology was first developed by a Steering Committee, consisting of
directors of water programs from EPA's national and regional offices and the Great Lakes states'
environmental agencies (as co-regulators of Clean Water Act water quality programs), and a
Technical Work Group consisting of technical staff from the Great Lakes states' environmental
agencies, EPA, the U.S. Fish and Wildlife Service, and the U.S. National Park Service. All of
-------
Examination of EPA Risk Assessment Principles and Practices Page 135
the work was reviewed and commented upon by a Public Participation Group consisting of
representatives from environmental groups, municipalities, industry, and academia who observed
the deliberations of the other two groups, advised them of the public's concerns, and kept their
various constituencies informed.
EPA's SAB reviewed the draft Tier H methodology in 1992. (See its final report,
EPA-SAB-EPEC/DWC-93-005, completed December 16, 1992.) The SAB commended EPA for
the interactions among the states, EPA, the private sector, and the scientific community during
the development of this proposed guidance and provided substantial comments on many
elements of the submitted draft documents. The Board provided specific comments on the Tier II
methodology. EPA responded to each of these comments in proposing and finalizing the
methodology. The preamble to the proposed rule (see below) describes these comments and
EPA's response to each.
The Tier II methodology was part of a 1995 rulemaking (USEPA, 1995c, d) and as such
received full public comment, response, and revision based on the comments, in accordance with
the Administrative Procedure Act. Federal Register notices requesting public comment on the
guidance were published on April 16, 1993, August 9, 1993, September 13, 1993, and August 30,
1994 (e.g., see USEPA,1993d). Over 26,500 pages of comments, data, and information from
over 6,000 respondents were received in response to these Federal Register notices and meetings
held with the public.
6.5 Uncertainty and Pesticide Ecological Assessment
6.5.1 How Does EPA Address Uncertainty Analysis in Pesticide Ecological
Assessments?
As a result of a dialogue with internal and external experts in the field of ecological risk
assessment and probabilistic methods, the Environmental Fate and Effects Division of the Office
of Pesticide Programs is currently developing methodology to routinely conduct probabilistic risk
assessments (http://www.epa.gov/oppefedl/ecorisk/index.htrn). The models were initially
presented to the FIFRA Scientific Advisory Panel in March 2001
(http://www.epa.gov/scipoly/sap/2001/index.htm#march). The Panel recommended a number of
improvements prior to implementation; these improvements will be discussed with the Panel in a
SAP meeting originally scheduled for February 2004 followed by implementation in fiscal year
2004. The models were used in an assessment of the risks of the use of carbofuran flowable
which is currently scheduled for completion in fiscal year 2004. In the interim, pesticide
registrants have submitted probabilistic exposure assessments for chlorpyrifos, chlorfenapyr,
diazinon, and atrazine. The chlorfenapyr assessment was reviewed by the FIFRA SAP, which
concluded that the registrants' conclusion of negligible risk is not supported for any geographic
scale. The scaling up of field exposures to the Cotton Belt scale was done incorrectly. The
probabilistic exposure assessment had several other flaws that require remediation (see
-------
Page 138 Examination of EPA Risk Assessment Principles and Practices
A comparative analysis model was used to rank and compare potential primary and
secondary risks. The underlying methodology is a simple multi-attribute rating technique, or
SMART (Goodwin and Wright, 1998). SMART was adapted for comparing potential risks
among rodenticides based on a number of measure-of-effect values for primary and secondary
risk to birds and mammals. It is similar to the technique used in the Agency's Comparative
Analysis of Acute Avian Risk From Granular Pesticides (USEPA, 1992e) and A Comparative
Analysis of Ecological Risks From Pesticides and Their Uses: Background, Methodology, and
Case Study (USEPA, 1998e); both were reviewed by a FIFRA Scientific Review Panel.
Concerning the latter analysis, the Panel noted the many scientific uncertainties in the method,
yet agreed that it was a useful screening tool that provides a rough estimate of relative risk. The
Panel made a number of helpful suggestions to improve the utility of the method, most of which
are included in this comparative assessment. In this analysis, a risk quotient (RQ), calculated as
the ratio of toxicant potentially ingested to the inherent toxicity of the rodenticide, was used to
compare potential primary risks to birds and non-target mammals. RQs are compared among
rodenticide baits based on the amount of bait and number of bait pellets that birds or non-target
mammals of various sizes would need to eat to ingest an acute oral (LD50) dose. Dietary data
(LC50) also are available for birds (but not for mammals), and RQs based on bait concentration
and avian dietary toxicity are compared among the rodenticides. As noted by the Ecological
Committee on FIFRA Risk Assessment Methods (ECOFRAM, 1999), RQs do not quantify risk
but are useful for comparisons among alternative compounds. EPA's Guidelines for Ecological
Risk Assessment (USEPA, 1998a) also notes that quotients provide an efficient, inexpensive
means of identifying high- or low-risk situations that can allow risk management decisions to be
made without the need for further information.
A major uncertainty has been how and where these pesticides are used. In estimating
pesticide exposure, risk assessors rely upon the label instructions, since the label prescribes the
level and manner of use. Use other than that specific on the label is illegal. The commensal use
is common to all nine rodenticides. The terminology "in and around buildings" appears on
product labels registered for commensal use. This statement does not limit bait placements to
any specified distance from buildings, and in many non-urban areas bait applications might pose
an exposure scenario comparable to some field uses. Only two of the nine rodenticide labels
limit the "in and around buildings" use to urban areas; applications in non-urban areas must be
indoors. Because of this label language, exposure to non-targets has been difficult to estimate:
the phrase "in and around buildings" i.e., how far from a building bait may be applied is
subject to interpretation. Further, many of these products are sold to the general public and
available in outlets ranging from grocery to hardware stores. Homeowners may or may not
follow the label instructions. How and where the public and pest control operators use these
products is largely unknown. Currently, general marketing data are available. At best, these data
report sales on a regional basis, and thus maybe used to estimate volume of use. California does
require that reports be maintained of pesticide use, but homeowners are exempt.
-------
Examination of EPA Risk Assessment Principles and Practices Page 139
For agricultural uses, the Agency obtains information on how the pesticide is typically
used: application rates, number of applications, areas in the United States where it is used, etc.
The registrants, the Rodenticide Registrants Task Force, have volunteered to provide usage
information for the rodenticide of most concern which will allow the Agency to further refine its
risk assessment.
6.6 Summary and Conclusions
This chapter examines ecological risk assessment approaches in a few areas within the
Superfund, water, and pesticides programs in order to address issues raised by comments
received by EPA, Within the Superfund program, we discuss EPA's approach regarding
organism- versus population-level ecological risk assessments and provide three examples in
which conservatism is addressed. Within the water program, we discuss the standard
methodology for deriving aquatic life WQC and the Great Lakes Initiative Tier n Methodology.
In the pesticides program we discuss how EPA addresses uncertainty analysis in pesticide
ecological assessments and rodenticide use patterns. To paraphrase a statement made earlier in
this document, we attempt to use all information available in an objective, realistic, scientifically
balanced way to make decisions.
Ecological risk assessment "evaluates the likelihood that adverse ecological effects may
occur or are occurring as a result of exposure to one or more stressors." It is a flexible process
for organizing and analyzing data, information, assumptions, and uncertainties to evaluate the
likelihood of adverse ecological effects. Ecological risk assessment provides a critical element
for environmental decision making by giving risk managers an approach for considering
available scientific information along with the other factors they need to consider (e.g., social,
legal, political, or economic) in selecting a course of action.
Ecological risk assessment includes three primary phases: problem formulation, analysis,
and risk characterization. In problem formulation, risk assessors evaluate goals and select
assessment endpoints, prepare the conceptual model, and develop an analysis plan. During the
analysis phase, assessors evaluate exposure to stressors and the relationship between stressor
levels and ecological effects. In the third phase, risk characterization, assessors estimate risk
through integration of exposure and stressor-response profiles, describe risk by discussing lines
of evidence and determining ecological adversity, and prepare a report.
The interface among risk assessors, risk managers, and interested parties during planning
at the beginning of the risk assessment, and communication of risk at its end, is critical to ensure
that the results of the assessment can be used to support a management decision. Because of the
diverse expertise required (especially in complex ecological risk assessments), risk assessors and
risk managers frequently work in multidisciplinary teams. Both risk managers and risk assessors
bring valuable perspectives to the initial planning activities for an ecological risk assessment.
Risk managers charged with protecting the environment can identify information they need to
-------
Page 140 Examination of EPA Risk Assessment Principles and Practices
develop their decisions, risk assessors can ensure that science is effectively used to address
ecological concerns, and together they can evaluate whether a risk assessment can address
identified problems. However, this planning process is distinct from the scientific conduct of an
ecological risk assessment. This distinction helps ensure that political and social issues, while
helping to define the objectives for the risk assessment, do not introduce undue bias.
Problem formulation, which follows these planning discussions, provides a foundation
upon which the entire risk assessment depends. Successful completion of problem formulation
depends on the quality of three products: assessment endpoints, conceptual models, and an
analysis plan. Since problem formulation is an interactive, nonlinear process, substantial
reevaluation is expected to occur during the development of all problem formulation products.
The analysis phase includes two principal activities: characterization of exposure and
characterization of ecological effects. The process is flexible, and interaction between the two
evaluations is essential. Both activities evaluate available data for scientific credibility and
relevance to assessment endpoints and the conceptual model. Exposure characterization
describes sources of stressors, their distribution in the environment, and their contact or co-
occurrence with ecological receptors. Ecological effects characterization evaluates stressor-
response relationships or evidence that exposure to stressors causes an observed response. The
bulk of quantitative uncertainty analysis is performed in the analysis phase, although uncertainty
is an important consideration throughout the entire risk assessment. The analysis phase products
are summary profiles that describe exposure and the stressor-response relationships. During risk
characterization, the final phase of an ecological risk assessment, risk assessors estimate
ecological risks, indicate the overall degree of confidence in the risk estimates, cite evidence
supporting the risk estimates, and interpret the adversity of ecological effects. To ensure mutual
understanding between risk assessors and managers, a good risk characterization will express
results clearly, articulate major assumptions and uncertainties, identify reasonable alternative
interpretations, and separate scientific conclusions from policy judgments.
The Agency's Ecological Risk Assessment Guidelines are just that guidelines, meant
not to be overly prescriptive. Each program office administers different statutes that in many
cases are silent on assessment endpoints or that stipulate endpoints that can be difficult to
quantify. As such, programs have the responsibility to develop assessment procedures that are
consistent with their enabling legislation. For the most part, the various programs' practice of
ecological risk assessment is consistent across the Agency, but there are enough differences to
raise questions of consistency, as illustrated by the comments this document seeks to address.
EPA's goal is to apply ecological risk assessment methods in as consistent and transparent a
process as possible, enabling our stakeholders to engage in a fully open debate on the risks and
the management of those risks for the overall protection of the environment.
-------
Examination of EPA Risk Assessment Principles and Practices Page 141
7. SUMMARY AND RECOMMENDATIONS
7.1 Summary
EPA's mission and statutes require that EPA protect human health and the environment,
and ensure that EPA's risk management decisions provide some "margin of safety" when risk
assessment answers are uncertain or variable. The risk estimates provided in risk assessments
and ultimately addressed in EPA decisions are driven by the uncertainty and variability inherent
in practically all the information and methodologies EPA uses. This uncertainty stems to a large
extent from the complex nature of the environment and the processes that interact with the
stressor(s) to cause risk. It is clear that risk assessors best understand the basic elements of a risk
assessment, in particular how to analyze the uncertainty and variability found in individual risk
assessments. Risk managers, by taking into consideration the characterization of all the elements
of a risk assessment including the uncertainty and variability, provided by the risk assessors
should determine the appropriate degree of protection to ensure that risk is not underestimated
and conversely, is not appreciably overestimated.
Comments on EPA risk assessment principles and practices have been received over the
years, more recently in response to the OMB Federal Register notice asking for such comments.
An overarching position in the comments suggests that EPA inappropriately mixes policy and
science; that, in generating risk assessments, risk assessors "decide" the degree of protection in a
way that is not transparently characterized. Further, in more than a few instances, it is suggested
that EPA's current risk assessment practice compounds risk estimates to a protective position
that is beyond reasonable. Also, in many instances, risk managers apparently are not aware of
the degree of protection inherent in each risk assessment. Some other comments suggest that
EPA risk estimates tend to underestimate risk, since a majority of the assessments tend to be
single-stressor assessments and do not adequately take into account multiple stressors, multiple
exposures, and susceptible populations.
EPA risk assessment practices are meant to produce credible, science-based risk
assessments that provide reasonable risk estimates. The assessments are designed to ensure that
risk is not underestimated or grossly overestimated. EPA views a critical analysis of the
available data specific to the relevant chemical (stressor) and/or site as the starting point for a
risk assessment. The derivation of risk estimates improves continually with the addition of
newer techniques and relevant data. In the absence of relevant data, EPA uses assumptions in
developing estimates of risk. The assumptions (defaults and extrapolations) used in the
assessments are necessary components of the assessment, as they help bridge the data gaps
encountered during the assessment. These assumptions are also focused on providing a
reasonable default position and usually have scientific data and scientific support. They are also
publicly vetted and peer reviewed. For the conduct of risk assessment, EPA uses a triage
approach (from screening to "in-depth" assessments) to determining how much time and
-------
Page 142 Examination of EPA Risk Assessment Principles and Practices
resources need to be expended. This is a practical approach for dealing with the generation of
different "depths" of assessment.
Another major theme in the comments regarding EPA risk assessment principles and
practices suggests that transparency and clarity are still absent in characterizing the elements of
the risk assessment, particularly the uncertainty and variability as well as the choices made in the
assessment. This may be an aspect of EPA's practices that needs strengthening. It should be
clear that the development of risk assessments for risk management purposes is a mix of science
data, science policy, judgments, guidelines, and best professional judgement. Transparency
about all these matters is crucial so that nothing appears "hidden" or "buried" in an assessment
so that nothing keeps one from understanding the impact of the elements that go into
estimating and characterizing a risk.
7.2 Several Current EPA Activities That Will Enhance EPA Risk Assessment Principles
and Practices
EPA works on many fronts to review our practices and methodologies as part of an
ongoing evaluation of risk assessment principles and practices. Several current examples
demonstrate our commitment to enhancing EPA risk assessment principles and practices:
a) EPA is currently finalizing its proposed Guidelines for Cancer Risk Assessment.
Included in this effort is the generation of a supplemental guidance focused on
cancer risk to children, particularly from mutagenic carcinogens. This is a major
effort that explores many newer issues, such as examining data before invoking
defaults, extensive use of mode of action to inform the assessment, more clearly
defining default positions that can be used when data are unavailable, and making
the cancer risk assessment practice more transparent and clear.
b) EPA is engaged in a long-term examination to "harmonize" cancer and non-
cancer health assessments into a unified approach more geared toward examining
how health endpoints, via a mode of action, present linear or nonlinear dose-
response curves that can be used for quantitation. Other major aspects of this
effort are the use of chemical-defined uncertainty factors, guidance for selecting a
point of departure from observed data to extrapolate to lower doses of expected
exposures, and clarification of dosing and scaling factors to use in health
assessments.
c) The use of environmental models to support risk assessments is constantly
increasing. EPA promotes consistency and consensus on mathematical modeling
issues including modeling guidance, development, and application, and enhances
both internal and external communications on modeling activities through
organizations such as the EPA Council on Regulatory Environmental Modeling.
-------
Examination of EPA Risk Assessment Principles and Practices Page 143
d) The better the data and information are, the better the risk assessment will be in
terms with being more reasonable with more certainty. To that end, EPA initiated
an effort to examine the usefulness of the toxicity tests needed, and required under
several statutes, to support EPA assessments and decisions. As the field of
toxicology evolves, particularly recently with the Human Genome Project and the
burgeoning area of genomics and related technologies, it is appropriate to examine
what will be the best tests we can use to generate data useful for risk assessments.
EPA is working with the NAS to examine these toxicity testing needs as projected
into the next two decades.
e) EPA is undertaking a major effort to upgrade and improve its process for creating
assessments to enter onto IRIS. As a result of an increase in IRIS staff and
funding, the number of new and updated assessments completed each year will be
increased. A more open external peer review component will provide more
transparency and interaction for external stakeholders.
f) EPA focuses much attention on its research program to help reduce major
uncertainties in risk assessment by understanding and elucidating the fundamental
determinants of exposure and dose and the basic biological changes that follow
exposure to stressors leading to a toxic response. For example, in EPA's recent
Human Health Research Strategy (USEPA, 2003rn), EPA will focus on ways to
improve the information supporting its risk assessments. These include
approaches to harmonizing the use of mechanistic data in human health
assessments, predicting the effects of aggregate and cumulative exposure, and
protecting susceptible populations such as children, older adults, and those with
preexisting disease or genetic predispositions for different responsiveness to
environmental stressors. Many of these areas are discussed in this document.
7.3 Recommendations for Further Improvements to Risk Assessment Principles and
Practices
Based on the evaluation in this document, the staff provides several recommendations for
EPA to consider as part of an agenda for further improvements to risk assessment principles and
practices.
a) Confidence in our risk assessments is critical. This confidence is increased by the
use of scientifically sound, chemical- and site-specific information. In the
absence of such data, EPA should continue to use default assumptions to fill data
gaps.
-------
Page 144 Examination of EPA Risk Assessment Principles and Practices
1) We need to continue to encourage the development of the specific data
necessary to more accurately assess potential risks, particularly with those
responsible for generating appropriate data (e.g., pesticide registrants). In
the absence of strong data requirements, EPA should continue to work
with industry and other partners to implement voluntary programs, such as
the High Production Volume Challenge, to generate such data, thus
lessening our reliance on defaults in developing risk assessments.
2) Since defaults assumptions are used in risk assessments to address
uncertainty when chemical- and/or site-specific data are not available or
useful, we need to ensure that the defaults themselves are supported by the
best available data. For example, the draft supplemental children's
guidance accompanying the EPA draft cancer guidelines suggests a default
position for children exposed to mutagenic carcinogens that is based on
data, albeit with a small sample number, that supports the default
assumption (USEPA, 2003d). We should continually look for
opportunities to increase our certainty and confidence in the defaults and
extrapolations we use. As more data become available, we develop and/or
refine the default assumptions and extrapolations to use; for example,
defaults might be developed for different life-stages or for chemicals with
a specific mode of action. EPA should encourage and support research
needed to develop and refine assumptions for use in risk assessments.
Further, to help increase our confidence in risk assessments, EPA might
consider the use of sensitivity analysis to examine what data points and/or
defaults involved in the assessment are the most critical.
b) Transparency in our risk assessment practice and our risk management process is
an issue that needs to be addressed. Initiatives for EPA to improve our
transparency in risk assessments and risk management include:
1) Continued concerted and conscious use of planning and scoping with risk
assessors and risk managers before a risk assessment is started. The intent
is not to point to a specific decision choice, but to outline what needs to be
addressed and how to do so. Just as important is to be clear about what
will not be addressed (i.e., the scope of the assessment effort needs to be
defined).
2) Continued use of the triage approach to decide how much time and
resources are necessary for a risk assessment. However, we need to be
more clear about why the risk assessment is being performed. This would
help reduce potential confusion about why EPA decides upon particular
risk estimates in one decision and others in another decision (e.g., a
-------
Examination of EPA Risk Assessment Principles and Practices Page 145
screening assessment versus a detailed assessment; different circumstances
at different sites).
3) We need to focus on better communication of the data and assumptions
and choices used in our risk assessments. We already have the guidance in
place in the form of our Risk Assessment Guidelines and Agency Risk
Characterization Policy and Handbook. Also, we have Planning and
Scoping Guidance as the first part of our Cumulative Risk initiative, and
this is incorporated into the new Cumulative Risk Framework, Close
attention to our guidances will ensure transparency and clarity in our risk
assessments. This will also help ensure that risk assessments will
appropriately support and inform decision making and risk management.
4) Regarding transparency in the risk management process: since we do not
have much guidance here, we should encourage work on a decision
making framework. This framework would help everyone, from those
inside EPA to all interested parties outside EPA, understand the process
risk managers (decision makers) go through to arrive at a decision. It
should not be a checklist of steps, but a set of actions (e.g., peer review)
and factors (e.g., cost/benefit analysis) risk managers should consider
before making a decision. While it is certainly feasible to produce such a
framework explaining our methods of procedure in general terms, it
cannot explain relevant "weights" given in risk management decisions to
the various factors (costs, benefits, treatment options, etc.) as each
decision will be made on a case-by-case basis. So most importantly, this
transparency will clarify how a risk assessment is incorporated into a
decision and how it informs the decision. Places to look for information to
construct such a framework include the NAS reports and the Commission
on Risk Assessment and Risk Management.
c) A major method to help address uncertainty in risk assessment is probabilistic
analysis. We can use that technique for exposure assessments, but even for these
assessments, we should encourage greater use and reliance on it when appropriate.
Further, we need to explore the feasibility of using probabilistic analyses in any
phase of the risk assessment. For example, we need to encourage the
development of Monte Carlo and other probabilistic analyses in the dose-response
assessment. If probabilistic analyses are identified for particular uses and it is
generally agreed they are appropriate, we should start incorporating the relevant
analyses where the risk assessor feels it is appropriate to do so.
1) It should be noted that the seemingly slow implementation of probabilistic
analysis is not strictly an internal EPA issue. Much environmental risk
-------
Page 146 Examination of EPA Risk Assessment Principles and Practices
assessment is performed by affected parties rather than EPA itself, and
outside parties' pace in implementing and/or accepting
probabilistic/quantitative analysis of uncertainty is seemingly slow as well.
Notwithstanding, we recommend that EPA examine a few recent or
ongoing risk assessments to determine what the obstacles are to use of
probabilistic analysis to describe uncertainty and how to overcome those
obstacles more frequently in the future. A major area that needs work is
the development of probability distributions for many endpoints in the
toxicity part of risk assessment, as well as putting the toxicity and
exposure components together in an overall probabilistic risk assessment.
Preliminary work in this area has been published, with some key papers
concerning the development of probability distributions for uncertainty
factors (Dourson et al., 1992; Baird et al., 1996; Swartout et al., 1998).
d) The staff feels greater harmonization with our partners states, other national
governments (as in the case of OPP's interactions with Canada), OECD as well
as greater cross-program interaction inside the Agency to ensure consistency and
scientific consensus for practices and methodologies (e.g., probabilistic analyses)
will help avoid a duplication of effort or differences in approaches and
interpretations.
e) EPA has historically assessed risks based on individual chemicals/stressors and
often focused on one source, pathway, or adverse effect. However, people do not
live in a "single chemical" world: in reality, they are exposed to multiple
chemicals/stressors from a variety of sources. Because people are exposed to
multiple stressors, this suggests that the total risk they may experience in their
daily lives may be underestimated by assessments of single stressors. Tools are
needed to understand combined risks. While the Agency has moved forward with
emphasis on evaluating cumulative risks (e.g., planning and scoping activities for
cumulative risk, the Cumulative Risk Framework, OPP examination of aggregate
risk), EPA needs to continue moving aggressively to flesh out approaches for
cumulative risk and to produce the most scientifically rigorous and realistic
evaluation of cumulative risk that the state-of-the-science can accommodate.
f) Overarching guidance is needed to improve the ease and consistency of risk
assessment for susceptible populations and life-stages. Such guidance should
provide more detail, methods, and advice for considering the issues regarding
susceptibility. As a part of such efforts, EPA needs to continue developing a
strategy and methodologies to address the cumulative risks of unique or
disproportionate exposures.
-------
Examination of EPA Risk Assessment Principles and Practices Page 147
g) Peer review is fundamental to ensuring the quality and integrity of all
assessments. We need to continue with constant vigilance to peer review all our
major scientific and technical work products.
h) Many practices EPA is criticized for are practices we no longer follow (e.g.,
compound uncertainty factors up to 10,000). However, there are older risk
assessments supporting standing decisions that may need reexamination. This
would need to be done in light of current resources and priorities, e.g., looking at
older IRIS values for chemicals that are still EPA priorities.
-------
Page 148
Examination of EPA Risk Assessment Principles and Practices
LIST OF USEFUL ABBREVIATIONS AND ACRONYMS
ACC American Chemistry Council
ADI Acceptable Daily Intake
AEGL acute exposure guideline level
ARAR applicable or relevant and appropriate requirement
ARE acute reference exposure
AQI Air Quality Index
ATFERM Agency Task Force on Environmental Regulatory Modeling
ATSDR Agency for Toxic Substances and Disease Registry
AUC area under the curve
BAF bioaccumulation factor
BAT best available technology
BMC benchmark concentration
BMCL benchmark concentration lower confidence limit
BMD benchmark dose
BMDL benchmark dose lower confidence level
BMR benchmark response
CAA Clean Air Act
CAAA Clean Air Act Amendments
CBD chronic beryllium disease
CDC Centers for Disease Control and Prevention
CEPPO Chemical Emergency Preparedness and Prevention Office
CERCLA Comprehensive Environmental Response, Compensation, and Liability
Act
CERCLIS Comprehensive Environmental Response, Compensation, and Liability
Information System
CFSAN Center for Food Safety and Nutrition
CNS central nervous system
CPSC Consumer Product Safety Commission
CRARM Congressional/Presidential Commission on Risk Assessment and Risk
Management
CREM Committee on Regulatory Environmental Modeling
CSAF chemical-specific adjustment factor
CSF cancer slope factor
CSFII Continuing Survey of Food Intakes by Individuals
CTE central tendency exposure
CWA Clean Water Act
DAF dosimetric adjustment factor
DBP disinfection byproduct
DENR Department of Environment and Natural Resources
DHHS Department of Health and Human Services
-------
Examination of EPA Risk Assessment Principles and Practices Page 149
DHS Department of Homeland Security
DNA deoxyribonucleic acid
DNT developmental neurotoxicity
DOD Department of Defense
DOE Department of Energy
DO J Department of Justice
DOT Department of Transportation
ECC Environmental Clearance Committee
Eco-SSLs Ecological Soil Screening Levels
ED effective dose
EDC endocrine disrupting chemical
EMAP Environmental Monitoring and Assessment Program
EPA Environmental Protection Agency
EPCRA Emergency Planning and Community Right-To-Know Act
ETV Environmental Technology Verification
FAV Final Acute Value
FDA Food and Drug Administration
FDAMA Food and Drug Administration Modernization Act
FFDCA Federal Food, Drug, and Cosmetic Act
FIELDS Field Environmental Decision Support
FIFRA Federal Insecticide, Fungicide, and Rodenticide Act
FQPA Food Quality Protection Act
FR Federal Register
FS feasibility study
GAO General Accounting Office
GHS globally harmonized system
GIS geographic information system
GLI Great Lakes Initiative
GLNPO Great Lakes National Program Office
GPRA Government Performance and Results Act
GPS Global Positioning System
GSA General Services Administration
HA Health Advisory
HAP hazardous air pollutant
HAZMAT hazardous materials
HEC human equivalent concentration
HED human equivalent dose
HEPA high-efficiency particulate air
HI hazard index
HMIS Hazardous Materials Information System
HQ hazard quotient
HVAC heating, ventilation, and air conditioning
-------
Page 150 Examination of EPA Risk Assessment Principles and Practices
IBI
1C
ICRP
IEUBK
1MB
IPCS
IRIS
ISO
LED
LOAEL
LOEL
MACT
MCL
MCLG
MEI
MF
MIMS
MLE
MOA
MOE
MOU
MRL
MTD
NAAQS
NAS
NATO/CCMS
NCEA
NCER
NCP
NCTR
NEIC
NERL
NESHAP
NHEERL
NHSRC
NIEHS
NIOSH
NNI
NOAA
NOAEL
NOEL
Index of Biological Integrity
index compound
International Commission on Radiological Protection
Integrated Exposure Uptake Biokinetic Model for Lead in Children
individuals most exposed
International Programme on Chemical Safety
Integrated Risk Information System
International Organization for Standardization
lowest effective dose
lowest-observed-adverse-effect level
lowest-observed-effect level
maximum achievable control technology
maximum contaminant level
maximum contaminant level goal
maximally exposed individual
modifying factor
Multimedia Integrated Modeling System
maximum likelihood estimate
mode of action
margin of exposure
Memorandum of Understanding
minimal risk level
maximum tolerated dose
National Ambient Air Quality Standards
National Academy of Sciences
North Atlantic Treaty Organization Committee on the Challenges of
Modern Society
National Center for Environmental Assessment
National Center for Environmental Research
National Oil and Hazardous Substances Pollution Contingency Plan
National Center for Toxicological Research
National Enforcement Investigation Center
National Environmental Research Laboratory
National Emissions Standard for Hazardous Air Pollutants
National Health and Environmental Effects Research Laboratory
National Homeland Security Research Center
National Institute for Environmental Health Sciences
National Institute for Occupational Safety and Health
National Nanotechnology Initiative
National Oceanic and Atmospheric Administration
no-observed-adverse-effect level
no-observed-effect level
-------
Examination of EPA Risk Assessment Principles and Practices Page 151
NPDES
NRC
NRMRL
NSF
NIP
OAQPS
OAR
OECD
OERR
OMB
OPEI
OPP
OPPT
OPPTS
ORD
OSHA
OSTP
OSW
OSWER
OU
OW
OWOW
PAD
PAH
PBPK
PCB
PCDD
PCDF
PEL
PM
PMN
POD
PPE
ppm
PRO
QA
QAPP
QA/QC
QSAR
RAF
RAGS
RCRA
National Pollutant Discharge Elimination System
National Research Council
National Risk Management Research Laboratory
National Science Foundation
National Toxicology Program
Office of Air Quality, Planning, and Standards
Office of Air and Radiation
Organisation for Economic Cooperation and Development
Office of Emergency and Remedial Response
Office of Management and Budget
Office of Policy, Economics, and Innovation
Office of Pesticide Programs
Office of Pollution Prevention and Toxics
Office of Prevention, Pesticides and Toxic Substances
Office of Research and Development
Occupational Safety and Health Administration
Office of Science and Technology Policy
Office of Solid Waste
Office of Solid Waste and Emergency Response
operable unit
Office of Water
Office of Wetlands, Oceans, and Watersheds
population-adjusted dose
polycyclic aromatic hydrocarbon
physiologically based pharmacokinetic
polychlorinated biphenyl
polychlorinated dibenzo-/?-dioxin
polychlorinated dibenzofurans
permissible exposure limit
paniculate matter
premanufacture notification
point of departure
personal protective equipment
parts per million
Preliminary Remediation Goal
quality assurance
Quality Assurance Project Plan
quality assurance/quality control
quantitative structure-activity relationship
Risk Assessment Forum
Risk Assessment Guidance for Superfund
Resource Conservation and Recovery Act
-------
Page 152 Examination of EPA Risk Assessment Principles and Practices
RDDR
ReVA
RGDR
RfC
RiD
RI
RME
RNA
ROD
RPF
RQ
RSC
SAB
SAP
SAR
SAV
SARA
SCV
SDWA
SDWAA
SEER
SEQL
SIP
SMART
SMCL
SOPs
STEL
STAR
TCCR
TCDD
TEF
TEQ
TI
TIO
TRC
TRI
TRY
TSCA
TWA
UCL
UCL95
regional deposited dose ratio
Regional Vulnerability Assessment
regional gas dose ratio
reference concentration
reference dose
remedial investigation
reasonable maximum exposure
ribonucleic acid
Record of Decision
relative potency factor
risk quotient
relative source contribution
Science Advisory Board
Scientific Advisory Panel
structure-activity relationship
Secondary Acute Value
Superfund Amendments and Reauthorization Act
Secondary Chronic Value
Safe Drinking Water Act
Safe Drinking Water Act Amendments
Surveillance, Epidemiology, and End Results
Sustainable Environment for Quality of Life
State Implementation Plan
simple multi-attribute rating technique
secondary maximum contaminant level
standard operating procedures
short-term exposure limit
Science To Achieve Results
transparency, clarity, consistency, reasonableness
tetrachlorodibenzo-p-dioxin
Toxicity Equivalency Factor
Toxic Equivalency Quotient
tolerable intake
Technology Innovation Office
Toxicogenomics Research Consortium
Toxics Release Inventory
toxicity reference value
Toxic Substances Control Act
time-weighted average
upper confidence limit
95% upper confidence limit of the arithmetic mean
-------
Examination of EPA Risk Assessment Principles and Practices Page 153
UF uncertainty factor
USDA United States Department of Agriculture
WHO-ECEH World Health Organization European Centre for Environmental Health
WQC water quality criteria
-------
Page 154 Examination of EPA Risk Assessment Principles and Practices
GENERAL REFERENCES
Agency for Toxic Substances and Disease Registry (ATSDR). (1991). Preventing lead poisoning
in young children. Atlanta, GA.
Agency for Toxic Substances and Disease Registry (ATSDR). (1996). ATSDR public health
assessment guidance manual. Atlanta, GA.
Agency for Toxic Substances and Disease Registry (ATSDR). (2000). Summary report for the
ATSDR Soil-Pica Workshop, June 2000. Atlanta, GA.
Ahlborg, M.G., G.C. Becking, L.S. Birnbaum, A. Brouwer, HJ.G.M. Derks, M. Feeley, G.
Golog, S. Hanberg, J.C. Larsen, A.K.D. Liem, S. Safe, C. Schlatter, F. Waern, M.
Younes, and E. Yrjanheikki. (1994). Toxic equivalency factors for dioxin-like PCBs:
Report on a WHO-ECEH and IPCS consultation, December 1993. Chemosphere
26(6):1049-1067.
American Academy of Pediatrics (AAP). (2003). Pediatric environmental health, 2nd ed. Ruth A.
Etzel, ed. Elk Grove Village, IL.
American Conference of Governmental Industrial Hygienists (ACGIH). (1991). Guidelines for
the classification of occupational carcinogens. In: Documentation of the threshold limit
values and biological exposure indices, 6th ed. Cincinnati, OH.
Ames, B.N., and L.S Gold. (1990). Chemical carcinogenesis; too many rodent carcinogens. Proc
NatlAcadSci USA %1-.1112-1116,
Asian Pacific Environmental Network (APEN). (1998). A seafood consumption survey of the
Laotian community of West Contra Costa County, CA. Oakland, CA.
Bachmann, R., A.L. Barton, J.R. Clark, P.L. deFur, S.J. Ells, C.A. Pittinger, M.W. Slimak, R.G.
Stahl, and R.S. Wentsel. (1998). A multi-stakeholder framework for ecological risk
management; Summary of a SETAC technical workshop. Supplement to Environ Toxicol
C/iewl8(2).24pp.
Baird, S., J, Cohen, J. Graham, A. Shlyakhter, and J. Evans. (1996). Noncancer risk assessment:
A probabilistic alternative to current practice. Hum Ecol Risk Assess 2(1):79-102.
Barnes, D., A. Alford-Stevens, L. Birnbaum, F.W. Kutz, W. Wood, and D. Patton. (1991).
Toxicity equivalency factors for PCBs? QualAssur 1(1):70-81.
-------
Examination of EPA Risk Assessment Principles and Practices Page 155
Barnthouse, L.W., J. Boreman, S.W. Christensen, C.P. Goodyear, D.S. Vaughan, and W. van
Winkle. (1984). Population biology in the courtroom: The Hudson River controversy.
Bioscience 34:14-19.
Barton, A., and A. Sergeant. (1998). Policy before the ecological risk assessment: What are we
trying to protect? Hum Ecol Risk Assess 4(4):787-795.
Bascietto, J.J. (1998). A framework for ecological risk assessment: Beyond the quotient method.
In: M.C. Newman and C.L. Strojan, eds. Risk assessment: Logic and measurement. Ann
Arbor Press. 352 pp.
Bascietto, J., D. Hinckley, J. Plafkin, and M. Slimak. (1990). Ecotoxicity and ecological risk
assessment. Environ Sci Technol 24(1):10-15.
Behrman, L.E., and V.C. Vaughan, HI. (1983). Textbook of Pediatrics. Philadelphia, PA: W.B.
Saunders Co.
Bickis, M., and D. Krewski. (1989). Statistical issues in the analysis of the long-term
carcinogenicity bioassay in small rodents: an empirical evaluation of statistical decision
rules. Fundam Appl Toxicol 12:202-221.
Burmaster, D.E. (2000). Distributions of projected job tenure for men and women in selected
industries and occupations in the United States, February 1996. Risk Anal 20(2):205-224.
Calabrese, E.J., R. Barnes, E.J. Stanek 3rd, H. Pastides, C.E. Gilbert, P. Veneman, X.R. Wang,
A. Lasztity, and P. T. Kostecki. (1989). How much soil do young children ingest: An
epidemiologic study. Regul Toxicol Pharmacol 10:123-137.
Calabrese E.J., E.J. Stanek, and C.E. Gilbert. (1991). Evidence of soil-pica behavior and
quantification of soil ingestion. Hum Exp Toxicol 10:245-249.
Casteel, S., R. Coward, C. Weis, G. Hermingsen, E. Hoffman, W. Brattin, R. Guzman, M.
Starost, J. Payne, S. Stockharm, S. Becker, J. Drexler, and J. Turk. (1997). Bioavailability
of lead to juvenile swine dosed with soil from the Smuggler Mountain NPL site of Aspen,
Colorado. Fundam Appl Toxicol 36(2):177-187.
Centers for Disease Control and Prevention (CDC). (2003). Second national report on human
exposure to environmental chemicals. NCEH Publication No. 02-0716. National Center
for Environmental Health (NCEH), Centers for Disease Control and Prevention, Atlanta,
GA. (http://www.cdc.gov/exposurereport/)
-------
Page 156 Examination of EPA Risk Assessment Principles and Practices
Charney, E., J. Sayer, and M. Coulter. (1980). Increased lead absorption in inner city children:
Where does the lead come from? Pediatrics 65:226.
Cogliano, V.J. (1997). Plausible upper bounds: Are their sums plausible? Risk Anal 17:77-84.
Columbia River Inter-Tribal Fish Commission (CRITFC). (1994). A fish consumption survey of
the Umatilla, Nez Perce, Yakama, and Warm Springs Tribes of the Columbia River
Basin. Technical Report 94-3. CRITFC, Portland, OR.
Commission on Risk Assessment and Risk Management (CRARM). (1997a). Framework for
environmental health risk management, Final report, Volume 1. The
Presidential/Congressional Commission on Risk Assessment and Risk Management,
Washington, DC.
Commission on Risk Assessment and Risk Management (CRARM). (1997b). Risk assessment
and risk management in regulatory decision-making, Final report, Volume 2. The
Presidential/Congressional Commission on Risk Assessment and Risk Management,
Washington, DC.
Comprehensive Environmental Response, Compensation, and Liability Act (Superfund)
(CERCLA). (1980). 42 U.S.C. 9601-9627.
Connelly, N.A., B.A. Knuth, and C.A. Bisogni. (1992). Effects of the health advisory changes on
fishing habits and fish consumption in New York sport fisheries. Report for the New
York Sea Grant Institute, Project No. R/FHD-2-PD. Dept. of Natural Resources, Cornell
University, Ithaca, NY. (Raw survey data also received electronically from study authors.)
Connelly, N.A., B.A. Knuth, and T.L. Brown. (1996). Sportfish consumption patterns of Lake
Ontario anglers and the relationship to health advisories. NAmer J Fish Man 16:90-101.
Cooper, W.E. (1998). Risk assessment and risk management: An essential integration. Hum Ecol
Risk Assess 4(4):931-937.
Corish, J., and D. Fitzpatrick. (2002). Theoretical models of percutaneous absorption. In:
Proceedings of the International Conference on Occupational and Environmental
Exposures of Skin to Chemicals, Crystal City, VA. Sept. 8-11. National Institute for
Occupational Safety and Health, Washington, DC,
Davis, S., P. Waller, R. Buschbon, J. Ballou, and P. White. (1990). Quantitative estimates of soil
ingestion in normal children between the ages of 2 and 7 years: Population-based
estimates using aluminum, silicon, and titanium as oil tracer elements. Arch Environ
Health 45:112-122.
-------
Examination of EPA Risk Assessment Principles and Practices Page 157
Dourson, M., S. Felter, and D. Robinson. (1996). Evolution of science-based uncertainty factors
in noncancer risk assessment. Regul Toxicol Pharmacol 24:108-120.
Dourson, M., L. Knauf, and J. Swartout (1992). On reference dose (RiD) and its underlying
toxicity data base. Toxicol IndHealth 8:171-189.
Dourson, M., and J. Stara. (1983). Regulatory history and experimental support of uncertainty
(safety) factors. Regul Toxicol Pharmacol 3:224-238.
Duguay, A.B., and G.L. Plaa. (1995). Tissue concentrations of methyl isobutyl ketone, methyl n-
butyl ketone and their metabolites after oral or inhalation exposure. Toxicol Lett
75(l-3):51-58.
Eadon, G,, L. Kaminsky, J. Silkworth, K. Aldous, D. Hilker, P. O'Keefe, R. Smith, J. Gierthy, J.
Hawley, N. Kin, and A. DeCaprio. (1986). Calculation of 2,3,7,8-TCDD equivalent
concentrations of complex environmental contaminant mixtures. Environ Health Perspect
70:221-227.
Ebert, E., N. Harrington, K. Boyle, J. Knight, and R. Keenan. (1993). Estimating consumption of
freshwater fish among Maine anglers. NAmer J Fish Man 13:727-745.
Ecological Committee on FIFRA Risk Assessment Methods (ECOFRAM). (1999). Terrestrial
draft report. Report by ECOFRAM Terrestrial Workgroup. May 10. 677 pp.
Executive Office of the President of the United States (EOP). (1991). Regulatory Program of the
United States Government. Current Regulatory Issues in Risk Assessment and Risk
Management, 1 April 1990-31 March 1991.
Executive Order (EO) 12898. (1994). Federal Actions to Address Environmental Justice in
Minority Populations and Low-Income Populations. Fed Reg 59(32):7629.
Executive Order (EO) 13045. (1997). Protection of children from environmental health and
safety risks. Fed Reg 62(78):19883-19888. April 23.
Food Quality Protection Act (FQPA). (1996). U.S. Public Law 104-170. 104th Cong., 3 August
1996. Section 408(b)(2)(C).
Foureman, G.L and HJ. Clewell. (1999). Route-to-route extrapolation with a physiologically-
based pharmacokinetic (PBPK) model for cumene. Toxicol Sci 48(1 -S):395 (abstract).
-------
Page 158 Examination of EPA Risk Assessment Principles and Practices
Gansewendt, B., U. Foest, D. Xu, E. Hallier, H.M. Bolt, and H. Peter. (1991). Formation of DNA
adducts in F-344 rats after oral administration or inhalation of [14C]methyl bromide.
Food Chem Toxicol 29:557-563.
Gaylor, D.W., and R.L. Kodell. (2000). Percentiles of the product of uncertainty factors for
establishing probabilistic reference doses. Risk Anal 20:245-250.
Giesy, J.P., and K. Kannan. (1998). Dioxin-like and non-dioxin-like toxic effects of
polychlorinated biphenyls (PCBs): Implications for risk assessment. Crit Rev Toxicol
28:511-569.
Gillies, M.E., and H.V. Paulin. (1983). Variability of mineral intakes from drinking water: A
possible explanation for the controversy over the relationship of water quality to
cardiovascular disease. Int JEpid 12(1):45-50.
Goodwin, P., and G. Wright. (1998). Decision analysis for management judgement, 2nd ed.
Hoboken, NJ: John Wiley & Sons, Inc. 454 pp.
Gospe, S.M., and M.A.S. Albayati. (1994). Comparison of oral and inhalation exposures to
toluene. JAm Coll Toxicol 13:21-32.
Hamm, J.T., C.-Y. Chen, and L.S. Birnbaum. (2003). A mixture of dioxins, fiirans and non-ortho
PCBs based upon consensus toxic equivalency factors produces dioxin-like reproductive
effects. Toxicol Sci 74:182-191.
Haseman, J.K. (1985) Issues in carcinogenicity testing: Dose selection. Fundam Appl Toxicol
5:66-78.
Haseman, J.K., and A.M. Lockhart. (1993). Correlations between chemically related site-specific
carcinogenic effects in long-term studies in rats and mice. Environ Health Perspect
101:50-54.
Hill, A.B. (1965). The environment and disease: Association or causation? Proc R Soc Med
58:295-300.
Huff, J., J. Cirvello, J. Haseman, and J. Bucher. (1991). Chemicals associated with site-specific
neoplasia in 1394 long-term carcinogenesis experiments in laboratory rodents. Environ
Health Perspect 93:247-270.
International Commission on Radiological Protection (ICRP). (1981). Report of the task group
on reference man. New York: Pergammon Press.
-------
Examination of EPA Risk Assessment Principles and Practices Page 159
Javitz, H. (1980). Seafood consumption data analysis. SRI International. Final report prepared for
EPA Office of Water Regulations and Standards. EPA Contract 68-01-3887.
Kadry, A.M., G.A. Skowronski, and M.S. Abdel-Rahman. (1995). Evaluation of the use of
uncertainty factors in deriving RfDs for some chlorinated compounds. J Toxicol Environ
Health 45:83-95.
Kalbfleisch, J.D., and R.L. Prentice. (1980). The statistical analysis of failure time data. New
York: Wiley.
Kaneko, T., P.Y. Wang, H. Tsukasa, and A. Sato. (1995). m-xylene toxicokinetics in
phenobarbital-treated rats comparison among inhalation exposure, oral administration,
and intraperitoneal administration. Toxicol Appl Pharmacol, 131(1):13-20.
Kutz, F.W., D.G. Barnes, E.W. Bretthauer, D.P. Bottimore, and H. Greim. (1990). The
international toxicity equivalency factor (I-TEF) method for estimating risks associated
with exposures to complex mixtures of dioxins and related compounds. Toxicol Environ
Chem 26:99-109.
Kutz, F.W., R.A. Linthurst, C. Riordan, M.W. Slimak, and R. Frederick. (1992). Ecological
research at EPA: New directions. Environ Sci Technol 26:861-866.
Kuzmack, A.M., andR.E. McGaughy. (1975). Quantitative risk assessment for community
exposure to vinyl chloride. EPA Office of Planning and Management and Office of
Health and Ecological Effects, December 1975.
Lackey, R.L. (1994). Ecological risk assessment. Fisheries 19(9):14-18.
Larson, S.J., C.G. Crawford, and R.J. Gilliom. (In press). Development and application of
watershed regression for pesticides (WARP) methods to estimate atrazine concentration
distributions in streams. U.S. Geological Survey Water Resources Investigations Report
03-4047. In press. (Draft version:
(http://in.water.usgs.gov/nawqa/mmwg/warp.atrazine. 110502.pdf)
Larson, S.J., and RJ. Gilliom. (2001). Regression models for estimating herbicide concentrations
in U.S. streams from watershed characteristics. JAmer WatResAssoc 37:1349-1367.
Leece, B., M.A. Denomme, R. Towner, S.M.A. Li, and S. Safe. (1985). Polychlorinated
biphenyls: Correlation between in vivo and in vitro quantitative structural-activity
relationships (QSARs). J Toxicol Environ Health 16:379-388.
-------
Page 160 Examination of EPA Risk Assessment Principles and Practices
Linthurst, R.A., L.A. Mulkey, M.W. Slimak, G.D. Veith, and B.M. Levinson. (2000). Ecological
research in the Office of Research and Development at the U.S. Environmental Protection
Agency: An overview of new directions. Environ Toxicol Chem 19(4): 1222-1229.
Maki, A.W., and M.W. Slimak. (1990). The role of ecological risk assessment in environmental
decision making. In: Grodzinski, W., ed. Ecological risks: Perspectives from Poland and
the United States. National Academy Press: Washington, DC. pp. 77-91.
Maltby, L., T.J. Kedwards, V.E. Forbes, K. Grasman, J.E. Kammenga, W.R. Munns, Jr., A.H.
Ringwood, J.S. Weis, and S.N. Wood. (2001). Linking individual-level responses and
population-level consequences. In: Baird, D.J., and H.A. Burton, Jr., eds. Ecological
complexity: New directions for assessing responses to stress. SETAC Special Publication
Series, Society of Environmental Toxicology and Chemistry, Pensacola, FL. pp. 27-82.
McEuen, S.F., C.F. Jacobson, C.D. Brown, and M.G. Miller. (1995). Metabolism and testicular
toxicity of 1,3-dinitrobenzene in the rat: Effect of route of administration. Fundam Appl
Toxicol 28(l):94-9.
National Center for Health Statistics (NCHS). (1987). Anthropometric reference data and
prevalence of overweight, United States, 1976-1980, Data from the National Health and
Nutrition Examination Survey. Series 11, No. 238. U.S. Department of Health and
Human Services, Public Health Service, National Center for Health Statistics, Hyattsville,
MD. DHHS Publication No. (PHS) 87-1688.
National Contingency Plan (NCP). (1990a). National Oil and Hazardous Substances Pollution
Contingency Plan. 40 CFR Part 300, Fed Reg 55:8666. March 8.
National Contingency Plan (NCP). (1990b). Preamble to the National Oil and Hazardous
Substances Pollution Contingency Plan. Fed Reg 53:51394. March 8.
National Environmental Justice Advisory Committee (NEJAC). (2002). Fish consumption and
environmental justice: A report on the public meeting convened by the National
Environmental Justice Advisory Council, December 3-6, 2001. Seattle.
(http://www.epa.gov/compliance/resources/publications/ej/fish_consump_report_1102.pd
f)
National Marine Fisheries Service (NMFS). (1993). Data tapes for the 1993 NMFS provided to
U.S. EPA National Center for Environmental Assessments. Cited in USEPA, 1997f.
National Research Council (NRC). (1983). Risk Assessment in the federal government:
Managing the process. National Academies Press: Washington, DC.
-------
Examination of EPA Risk Assessment Principles and Practices Page 161
National Research Council (NRC). (1993). Pesticides in the diets of infants and children.
National Academies Press: Washington, DC,
National Research Council (NRC). (1994). Science and judgment in risk assessment. National
Academies Press; Washington, DC.
National Research Council (NRC). (1996). Understanding risk: Informing decisions in a
democratic society. Stern, P.C., and H.V. Fineberg, eds. National Academies Press:
Washington, DC,
National Research Council (NRC). (2001). Arsenic in drinking water: 2001 update. National
Academies Press: Washington, DC.
North Atlantic Treaty Organization (NATO). (1988a). International toxicity equivalency factor
(I-TEF), method of risk assessment for complex mixtures of dioxins and related
compounds. Report No. 176. NATO.
North Atlantic Treaty Organization (NATO). (1988b). Scientific basis for the development of
international toxicity equivalency factor (I-TEF), method of risk assessment for complex
mixtures of dioxins and related compounds. Report No. 178. NATO Committee on the
Challenges of Modern Society.
Norton, S.B., DJ. Rodier, J.H. Gentile, W.H. van der Schalie, W.P. Wood, and M.W. Slimak.
(1992). A framework for ecological risk assessment at the EPA. Environ Toxicol Chem
11:1663-1672.
Nosek, J.A., S.R. Craven, J.R. Sullivan, S.S. Hurley, andR.E. Peterson. (1992a). Toxicity and
reproductive effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin in ring-necked pheasant hens.
J Toxicol Environ Health 35(3):187-198.
Nosek, J.A., S.R. Craven, J.R. Sullivan, J.R. Olson, and R.E. Peterson. (1992b). Metabolism and
disposition of 2,3,7,8-tetrachlorodibenzo-p-dioxin in ring-necked pheasant hens, chicks,
and eggs. J Toxicol Environ Health 35(3): 153-164.
Office of Science and Technology Policy (OSTP). (1985). Chemical carcinogens: A review of
the science and its associated principles. OSTP, Executive Office of the President. Fed
Reg 50:10371-10442. March 14.
Office of Science and Technology Policy (OSTP). (1995). Science, risk, and public policy.
OSTP, Executive Office of the President, Washington, DC.
-------
Page 162 Examination of EPA Risk Assessment Principles and Practices
Owen, B.A. (1990). Literature-derived absorption coefficients for 39 chemicals via oral and
inhalation routes of exposure. Regul Toxicol Pharmacol 11(3):237-252.
Pacala, S.W., E. Bulte, J.A. List, and S.A. Levin, (2003). False alarm over environmental false
alarms. Science 301:1187-1188.
Ramesh, A., D.B. Hood, F. Inyang, M. Greenwood, A.M. Nyanda, A.E. Archibong, and M.E.
Knuckles. (2002). Comparative metabolism, bioavailability, and toxicokinetics of
benzo[a]pyrene in rats after acute oral, inhalation, and intravenous administration.
Polycydic Aromatic Hydrocarbons 22(3-4):969-980.
Restum, J., S. Bursian, J. Giesy, J. Render, W. Helferich, E. Shipp, D. Verbrugge, and R.
Aulerich. (1998). Multigenerational study of the effects of consumption of PCB-
contaminated carp from Saginaw Bay, Lake Huron, on mink. 1. Effects on mink
reproduction, kit growth and survival, and selected biological parameters. J Toxicol
Environ Health 54:343-375.
Rhomberg, L.R. (1997). A survey of methods for chemical health risk assessment among federal
regulatory agencies. Supporting document for Framework for Environmental Health Risk
Management Report, National Commission on Risk Assessment and Risk Management.
(http://www.riskworld.CQm/nreports/1996/risk rpt/)
Ross, P. (2000). Marine mammals as sentinels in ecological risk assessment. Hum EcolRisk
Assess 6:29-46.
Ross, P., R. De Swart, R. Addison, H. Van Loveren, J. Vos, and A. Osterhaus. (1996).
Contaminant-induced immunotoxicity in harbour seals: Wildlife at risk? Toxicology
112:157-169.
Ruckelshaus, W.D. (1983). Science, risk, and public policy. Science 221:1026-1028.
Ruffle, B., D. Burmaster, P. Anderson, and D. Gordon. (1994). Lognormal distributions for fish
consumption by the general U.S. population. Risk Anal 14(4):395-404.
Russell, M. and M. Gruber. (1987). Risk assessment in environmental policy-making. Science
236:286-290.
Safe Drinking Water Act Amendments (SDWAA). (1996). U.S. Public Law 104-182. 104th
Cong. 6 August 1996. Sections 102(b)(l)(C) and 103(b)(3)(C)(i)(V).
-------
Examination of EPA Risk Assessment Principles and Practices Page 163
Safe, S. (1990). Polychlorinated biphenyls (PCBs), dibenzop-dioxins (PCDDs), dibenzofurans
(PCDFs), and related compounds: Environmental and mechanistic considerations which
support the development of toxic equivalency factors (TEFs). Crit Rev Toxicol
Safe, S. (1994). Polychlorinated biphenyls (PCBs); Environmental impact, biochemical and toxic
responses, and implications for risk assessment. Crit Rev Toxicol 24(2):87-149.
Sample, B.E., D.M. Opresko, and G.W. Suter, II. (1996). Toxicological benchmarks for wildlife.
1996 Revision. Risk Assessment Program, Health Science Research Division, Oak Ridge
National Laboratory. Oak Ridge, TN. ES/ER/TM-86/R3.
Schoeny, R., P. Muller, and J. Mumford. (1998). Risk assessment for human health protection
Applications to environmental mixtures. In: Douben, P., ed. Pollution risk assessment and
management. New York: John Wiley & Sons. pp. 205-234.
Sechena, R., S. Liao, R.M. Lorenzana, C. Nakano, N. Polissar, and R.A. Fenske. (2003). Asian
American and Pacific Islander seafood consumption A community-based study in
King County, Washington. J Expo Anal Environ Epidemiol 13(4):256-66.
Sexton, K., J.L.Adgate, G. Ramachandran, G.C. Pratt, S.J. Mongin, T.H. Stock, and M.T.
Morandi. .(2004). Comparison of personal, indoor and outdoor exposures to hazardous air
pollutants in three urban communities. Environ Sci Technol 38:423-430.
Shipp, E., J. Restum, J. Giesy, S. Bursian, R. Aulerich, and W. Helferich. (199.8a).
Multigenerational study of the effects of consumption of PCB-contaminated carp from
Saginaw Bay, Lake Huron, on mink. 2. Liver PCB concentration and induction of hepatic
cytochrome P-450 activity as a potential biomarker for PCB exposure. J Toxicol Environ
Health 54:377 '-401.
Shipp, E., J. Restum, J. Giesy, S. Bursian, R. Aulerich, and W. Helferich. (1998b).
Multigenerational study of the effects of consumption of PCB-contaminated carp from
Saginaw Bay, Lake Huron, on mink. 3. Estrogen receptor and progesterone receptor
concentrations, and potential correlation with dietary PCB consumption. J Toxicol
Environ Health 54:403-420.
Simmons, J.E., S.D. Richardson, T.F. Speth, R.L Miltner, G. Rice, K. Schenck, E.S. Hunter, III,
and L.K. Teuschler. (2002). Development of a research strategy for integrated
technology-based toxicological and chemical evaluation of complex mixtures of drinking
water disinfection byproducts. Environ Health Perspect 1 10(6):1013-1024.
-------
Page 164 Examination of EPA Risk Assessment Principles and Practices
Stanek, E.J., and EJ. Calabrese. (1995a). Daily estimates of soil ingestion in children. Environ
Health Perspect 103(3):276-285.
Stanek, E.J., and EJ. Calabrese. (1995b). Soil ingestion estimates for use in site evaluations
based on the best tracer method. Hum Ecol Risk Assess 1:133-156.
Stanek, E.J., in, and E.J. Calabrese. (2000). Daily soil ingestion estimates for children at a
Superfund site. Risk Anal 20:627-635.
Summer, C., J. Giesy, S. Bursian, J. Render, T. Kubiak, P. Jones, D. Verbrugge, R. Aulerich.
(1996a). Effects induced by feeding organochlorine-contaminated carp from Saginaw
Bay, Lake Huron, to laying white leghorn hens. I. Effects on health of adult hens, egg
production, and fertility. J Toxicol Environ Health 49:389-407.
Summer, C., J. Giesy, S. Bursian, J. Render, T. Kubiak, P. Jones, D. Verbrugge, R. Aulerich.
(1996b). Effects induced by feeding organochlorine-contaminated carp from Saginaw
Bay, Lake Huron, to laying white leghorn hens. n. Embryotoxic and teratogenic effects. J
Toxicol Environ Health 49:409-438.
Superfund Amendments and Reauthorization Act (SARA). (1986). 42 U.S.C. 9605.
Suter, G. (1993). Ecological risk assessment. Boca Raton, FL: Lewis Publishers, Inc.
Suter, G.W., II, R.A. Efroymson, B.E. Sample, andD.S. Jones. (2000). Ecological risk
assessment for contaminated sites. Boca Raton, FL: CRC Press LLC.
Swartout, J.C., P.S. Price, M.L. Dourson, H.L. Carlson-Lynak, and R.E. Keenan. (1998). A
probabilistic framework for the reference dose (probabilistic RiD). Risk Anal 18:271-282.
Tillitt, D., and P. Wright. (1997). Dioxin-like embryotoxicity of a Lake Michigan lake trout
extract to developing lake trout. Organohalogen Cmpds 34:221-225.
Timchalk, C., M.D. Dryzga, F.A. Smith, and M. Bartels. (1991). Disposition and metabolism of
[14C]l,2-dichloropropane following oral and inhalation exposure in Fischer 344 rats.
Toxicology 68(3):291-306.
Toy, K., G. Gawne-Mittelstaedt, N. Polissar, and S. Liao. (1996). A fish consumption survey of
the Tulalip and Squaxin Island tribes of the Puget Sound region. Tulalip Tribes, National
Resources Department, Marysville, WA.
Train, R. (1976). Interim procedures and guidelines for health risk and economic impact
assessments of suspected carcinogens. EPA Office of the Administrator, May 1976.
-------
Examination of EPA Risk Assessment Principles and Practices Page 165
U.S. Environmental Protection Agency (USEPA). (1980). Water quality criteria documents:
availability. Fed Reg 45: 79318-79379.
U.S. Environmental Protection Agency (USEPA). (1984). Risk assessment and management:
Framework for decision making, EPA 600/9-85-002, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1985). Guidelines for deriving numerical
National Water Quality Criteria for the protection of aquatic organisms and their uses.
EPA 822/R-85-100. Office of Research and Development, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1986a). Guidelines for carcinogen risk
assessment. EPA/600/8-87/045. Risk Assessment Forum, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1986b). Guidelines for the health risk
assessment of chemical mixtures. EPA/600/8-87/045. Risk Assessment Forum,
Washington DC.
U.S. Environmental Protection Agency (USEPA). (1986c). Superfund public health evaluation
manual. EPA/540/1-86/060. Office of Emergency and Remedial Response, Washington,
DC.
U.S. Environmental Protection Agency (USEPA). (1987a). Interim procedures for estimating
risks associated with exposures to mixtures of chlorinated dibenzo-p-dioxins and -
dibenzofurans (CDDs and CDFs). EPA/625/3-87/012. Risk Assessment Forum,
Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1987b). Unfinished business: A comparative
assessment of environmental problems. Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1988a). Technical support document on risk
assessment of chemical mixtures. EPA/600/8-90/064. Risk Assessment Forum,
Washington DC.
U.S. Environmental Protection Agency (USEPA). (1988b). Guidance for conducting remedial
investigations and feasibility studies. OSWER Directive 9355.3-01. EPA/540/G-89/004.
Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1988c). Special review final decision relating
to diazinon in the matter of Ciba-Geigy Corporation et al. FIFRA Docket Nos. 562 et seq.
March 29.
-------
Page 166 Examination of EPA Risk Assessment Principles and Practices
U.S. Environmental Protection Agency (USEPA). (1989a). Risk assessment guidance for
Superfund, Volume I: Human health evaluation manual (Part A), Interim Final.
EPA/540/1-89/002. Office of Emergency and Remedial Response, Washington, DC.
(http://www.epa.gov/superfund/programs/risk/ragsa)
U.S. Environmental Protection Agency (USEPA). (1989b). Interim procedures for estimating
risks associated with exposures to mixtures of chlorinated dibenzo-p-dioxins and -
dibenzofurans (CDDs and CDFs) and 1989 update. EPA/625/3-89/016. Risk Assessment
Forum, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1989c). Resolution on use of mathematical
models by EPA for regulatory assessment and decision making. SAB-EEC-89-012,
Science Advisory Board, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1989d). Short-term methods for estimating
the chronic toxicity of effluents and receiving waters to freshwater organisms. EPA
600/4-89/001. Cincinnati, OH.
U.S. Environmental Protection Agency (USEPA). (1990). Reducing risk: Setting priorities and
strategies for environmental protection. EPA-SAB-EC-90-021. Science Advisory Board,
Washington, DC. 26 pp.
U.S. Environmental Protection Agency (USEPA). (1991a). Alpha-2u-globulin: association with
chemically induced renal toxicity and neoplasia in the male rat. EPA/625/3- 91/019F.
Risk Assessment Forum, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1991b). Workshop report on toxicity
equivalency factors for polychlorinated biphenyl congeners. EPA/625/3-91/020. Risk
Assessment Forum, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1991c). Risk assessment guidance for
Superfund, Volume I: Human health evaluation manual, Supplemental Guidance,
"Standard default exposure factors." OSWER Directive 9285.6-03. Office of Emergency
and Remedial Response, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1991d). Methods for measuring the acute
toxicity of effluents to freshwater and marine organisms, 4th ed. EPA-600/4-90/027.
Cincinnati, OH.
U.S. Environmental Protection Agency (USEPA). (1992a). Guidelines for exposure assessment.
EPA 600Z-92/001. Risk Assessment Forum, Washington, DC. 170 pp.
(http://cfpub.epa.gov/ncea/raf/recordisplay.cfrn?deid=559070)
-------
Examination of EPA Risk Assessment Principles and Practices Page 167
U.S. Environmental Protection Agency (USEPA). (1992b). Supplemental guidance to RAGS:
Calculating the concentration term. OSWER Directive 9285.7-081. Office of Solid Waste
and Emergency Response, Washington, DC,
U.S. Environmental Protection Agency (USEPA). (1992c). Consumption surveys for fish and
shellfish: A review and analysis of survey methods. EPA 822/R-92-001. Office of Water,
Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1992d). Framework for ecological risk
assessment. EPA/630/R-92/001. Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1992e). Comparative analysis of acute avian
risk from granular pesticides. Office of Pesticide Programs, Washington, DC. 71 pp.
U.S. Environmental Protection Agency (USEPA). (1993a). Provisional guidance for quantitative
risk assessment of polycyclic aromatic hydrocarbons. EPA/600/R-93/089. Office of
Research and Development, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1993b). A review of ecological assessment
case studies from a risk assessment perspective. EPA/630/R-92/005. Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1993c). Interim report on data and methods
for assessment of 2,3,7,8-tetrachlorodibenzo-p-dioxin risks to aquatic life and associated
wildlife. EPA/600/R-93/055. Office of Research and Development, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1993d). Water quality guidance for the great
lakes system, Proposed rule. Fed Reg 58:20802.ff April 16.
U.S. Environmental Protection Agency (USEPA). (1994a). Methods for Derivation of Inhalation
reference concentrations and application of inhalation dosimetry. EPA/600/8-90/066F.
Office of Research and Development, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1994b). Report of the Agency Task Force on
environmental regulatory modeling Guidance, support needs, draft criteria and charter.
EPA 500-R-94-001. Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1994c). Model validation for predictive
exposure assessments. Risk Assessment Forum, Washington, DC.
(http://cfpub.epa.gov/crem/cremlib.cfin)
-------
Page 168 Examination of EPA Risk Assessment Principles and Practices
U.S. Environmental Protection Agency (USEPA). (1994d). Peer review program. Memorandum
from Administrator Carol Browner to Assistant Administrators, General Counsel,
Inspector General, Associate Administrators, Regional Administrators, and Staff Office
Directors. Science Policy Council, Washington, DC. June 7.
(http://www.epa.gov/osp/spc/memoQ607.htm)
U.S. Environmental Protection Agency (USEPA). (1994e). Catalog of standard toxicity tests for
ecological risk assessment. EPA 540-F-94-013, Office of Emergency and Remedial
Response, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1994f). Field studies for ecological risk
assessment. EPA 540-F-94-014. Office of Emergency and Remedial Response,
Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1994g). A review of ecological assessment
case studies from a risk assessment perspective, Vol. n. EPA 630-R-94-003. Washington,
D.C.
U.S. Environmental Protection Agency (USEPA). (1994h). Ecological risk assessment issue
papers. EPA/630/R-94/009. Risk Assessment Forum, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1994i). Managing ecological risks at EPA:
Issues and recommendations for progress. EPA/600/R-94/183. Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1995a). Policy for risk characterization.
Science Policy Council, Washington, DC. fhttp://www.epa.gov/osp/spc/rcpolicv.htm)
U.S. Environmental Protection Agency (USEPA). (1995b). Policy on evaluating health risks to
children. Science Policy Council, Washington, DC.
(http://www.epa.gov/osp/spc/memohlth.htm)
U.S. Environmental Protection Agency (USEPA). (1995c). Water quality guidance for the Great
Lakes system, Final Rule. Fed Reg 60(56): 15366-15425.
U.S. Environmental Protection Agency (USEPA). (1995d). Water quality guidance for the Great
Lakes system: Supplementary information document (SID). EPA-820-B-95-001. Office
of Water, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1996a). Soil screening guidance: User's
guide. 2nd ed. OSWERPub. 9355.4-23. Office of Emergency and Remedial Response,
Washington, DC.
-------
Examination of EPA Risk Assessment Principles and Practices Page 169
U.S. Environmental Protection Agency (USEPA). (1996b). Recommendations of the technical
review workgroup for lead for an approach to assessing risks associated with adult
exposures to lead in soil. EPA-540-R-03-001. Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1996c). Risk assessment guidance reforms.
Office of Emergency and Remedial Response, Washington, DC.
(http://wAvw.epa.gov/superrund/programs/refonns/refonns/3-5a.htrn#des^
U.S. Environmental Protection Agency (USEPA). (1996d). PCBs: Cancer dose-response
assessment and application to environmental mixtures. EPA/600/P-96/001F. National
Center for Environmental Assessment, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1997a). Guidance on cumulative risk
assessment. Part 1: Planning and scoping. Science Policy Council, Washington, DC.
(http://www.epa.gpv/osp/spc/cumrisk2.htm')
U.S. Environmental Protection Agency (USEPA). (1997b). Guiding principles for Monte Carlo
analysis. EPA/630/R-97/001. Risk Assessment Forum, Office of Research and
Development, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1997c). Policy for use of probabilistic
analysis in risk assessment at the U.S. Environmental Protection Agency. Fred Hansen,
Deputy Administrator. Science Policy Council, Washington, DC.
(http://www.epa.gov/osp/spc/nrobpol.htm')
U.S. Environmental Protection Agency (USEPA). (1997d). Exposure factors handbook.
EPA/600/P-95/002F(a,b,c). Office of Research and Development, Washington, DC.
(http://cfpub.epa.gov/ncea/cfm/recordisplav.cfrn?deid=12464')
U.S. Environmental Protection Agency (USEPA). (1997e). Aerial spray drift review. FIFRA
Scientific Advisory Panel, Washington, DC. December 10-11.
(http://www.epa. gov/scipolv/sap/1997/december/spravdrift.htm^
U.S. Environmental Protection Agency (USEPA). (1997f). Ecological Risk Assessment
Guidance for Superfund: Process for Designing and Conducting Ecological Risk
Assessments - Interim Final. Office of Solid Waste and Emergency Response, OSWER
9285.7-25, EPA 540-R-97-006, PB97-963211, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1998a). Guidelines for ecological risk
assessment. EPA/630/R-95/002F. Risk Assessment Forum, Washington, DC. 171 pp.
(http ://cfbub .et>a. go v/ncea/cfrn/recordisplav.cfnrfdeid^ 1246Q~t
-------
Page 170 Examination of EPA Risk Assessment Principles and Practices
U.S. Environmental Protection Agency (USEPA). (I998b). Beryllium and compounds. Integrated
Risk Information System, Office of Research and Development, Washington, DC.
(http://www.epa.gov/iris/subst/QQ12.htm')
U.S. Environmental Protection Agency (USEPA). (1998c). Assessment of thyroid follicular cell
tumors. EPA/630/R-97/002. Science Policy Council, Washington, DC.
(http://www.epa.gov/osp/SDC/2thvfoll.htm')
U.S. Environmental Protection Agency (USEPA). (1998d). Proposed methods for basin-scale
estimation of pesticide concentrations in flowing water and reservoirs for tolerance
reassessment, Chapter 4: An index reservoir for use in assessing drinking water exposure.
FIFRA Scientific Advisory Panel, Washington, DC. July.
(http://www.epa.gov/oscpmont/sap/1998/index.htari
U.S. Environmental Protection Agency (USEPA). (1998e). A comparative analysis of ecological
risks from pesticides and their uses: Background, methodology & case study. Office of
Pesticide Programs, Washington, DC.
(http://www.epa.gov/scipolv/sap/1998/index.htm#december8^
U.S. Environmental Protection Agency (USEPA). (1999a). Guide to preparing Superfund
proposed plans, records of decision, and other remedy selection decision documents. EPA
540/R-98/031. Office of Solid Waste and Emergency Response, Washington, DC.
(http ://www. epa. gov/superfund/resources/remedv/rods'l
U.S. Environmental Protection Agency (USEPA). (1999b). Guidelines for carcinogen risk
assessment, Revised draft. NCEA-F-0644. Office of Research and Development,
Washington, DC.
(http://cfpub.epa.gov/ncea/raf/cancer gls.pdf)
U.S. Environmental Protection Agency (USEPA). (1999c). White paper on the nature and scope
of issues on adoption of model use acceptability guidance. Science Policy Council.,
Washington, DC.
(http://cfpub.epa.gov/crem/cremlib.cfm1
U.S. Environmental Protection Agency (USEPA). (1999d). Proposed methods for determining
watershed-derived percent crop areas and considerations for applying crop area
adjustments to surface water screening models. FIFRA Scientific Advisory Panel,
Washington, DC. May 25-27.
(http://www.epa.gov/oscpmont/sap/1999/index.htm')
-------
Examination of EPA Risk Assessment Principles and Practices Page 171
U.S. Environmental Protection Agency (USEPA). (1999e). Two documents: 1) Downwind
deposition tolerance bounds for ground hydraulic boom sprayers; 2) Downwind
deposition tolerance bounds for orchards. FIFRA Scientific Advisory Panel, Washington,
DC. July 20-23.
(http://www.epa. gov/scipolv/sap/1999/index.htrri)
U.S. Environmental Protection Agency (USEPA). (1999f). Risk assessment guidance for
Superfund, Volume 1, Human health evaluation manual (Supplement to Part A):
Community involvement in Superfund risk assessments. EPA 540-R-98-042. Office of
Solid Waste and Emergency Response, Washington, DC. March.
U.S. Environmental Protection Agency (USEPA). (1999g). Short Sheet: IEUBK Model
Bioavailability Variable. EPA 540-F-00-006. Office of Solid Waste and Emergency
Response, Washington, DC. October.
U.S. Environmental Protection Agency (USEPA). (1999h). Superfund risk assessment and how
you can help (videotape). EPA-540-V-99-002. Office of Solid Waste and Emergency
Response, Washington, DC. September.
U.S. Environmental Protection Agency (USEPA). (1999i). Rapid bioassessment protocols for use
in streams and wadeable rivers: Periphyton, benthic macroinvertebrates and fish, 2nd
edition. EPA 841-B-99-002. Office of Water, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (1999J). Issuance of final guidance: Ecological
risk assessment and risk management principles for Superfimd Sites. OSWER Directive
9285.7-28 P. Office of Solid Waste and Emergency Response, Washington, DC. October
7.
U.S. Environmental Protection Agency (USEPA). (2000a). Science Policy Council Handbook:
Risk Characterization Handbook. EPA 100-BOO-002. Science Policy Council,
Washington, DC. December, (http://www.epa.gov/osp/srjc/rchandbk.rjdfi
U.S. Environmental Protection Agency (USEPA). (2000b). Science Policy Council Handbook:
Peer Review Handbook, 2nd edition. EPA 100-BOO-001. Science Policy Council,
Washington, DC. December. (http://www.epa.gov/osp/spc/prhandbk.pdf)
U.S. Environmental Protection Agency (USEPA). (2000c). Policy and program requirements for
the mandatory agency-wide quality system. EPA Order 5360.1 A2. Office of
Environmental Information, Washington, DC.
(http://www.epa.gov/qualitv/qs-docs/5360-l.pdf)
-------
Page 172 Examination of EPA Risk Assessment Principles and Practices
U.S. Environmental Protection Agency (USEPA). (2000d). EPA quality manual for
environmental programs. EPA Order 5360 Al. Office of Environmental Information,
Washington, DC. (http://www.epa.gov/qualitv/qs-docs/5360.pdf)
U.S. Environmental Protection Agency (USEPA). (2000e). An SAB advisory on the USEPA's
draft case study analysis of the residual risk of secondary lead smelters, EPA Science
Advisory Board, May 2000, pp. 20-21.
(http://www.epa.gov/sab/pdfyecadvQ5.pdf)
U.S. Environmental Protection Agency (USEPA). (2000f). Estimated per capita fish
consumption in the United States. EPA-821-R-00-025.
U.S. Environmental Protection Agency (USEPA). (2000g). Benchmark dose technical guidance
document EPA/630/R-00/001. Risk Assessment Forum, Washington, DC.
(http://cfpubl.epa.gov/ncea/cfm/recordisplav.cfm?deid=22506'l
U.S. Environmental Protection Agency (USEPA). (2000h). Supplementary guidance for
conducting health risk assessment of chemical mixtures. EPA/630/R-00/002. Risk
Assessment Forum, Washington DC.
U.S. Environmental Protection Agency (USEPA). (2000i). Exposure and human health
reassessment of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and related compounds
(External Review Draft). EPA/600/P-00/001B(a-f). National Center for Environmental
Assessment, Office of Research and Development, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (2000J). Session VI Progress report on
estimating pesticide concentrations in drinking water and assessing water treatment
effects on pesticide removal and transformation; A consultation. FIFRA Scientific
Advisory Panel, Washington, DC. September, 26-29.
(http://www.epa.gov/oscpmont/sap/2000/index.htm)
U.S. Environmental Protection Agency (USEPA). (2000k). Methodology for deriving ambient
water quality criteria for the protection of human health. EPA-822-B-00-004. Office of
Water, Office of Science and Technology, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (20001). Data quality objectives process for
hazardous waste site investigations Data quality objectives guidance. EPA/600/R-
00/007. Office of Environmental Information, Washington, DC. January.
-------
Examination of EPA Risk Assessment Principles and Practices Page 173
U.S. Environmental Protection Agency (USEPA). (2000m). Phase 2 report: Further site
characterization and analysis, Volume 2F. Revised human health risk assessment Hudson
River PCBs reassessment RI/FS. Region 2, New York, NY. November.
(http://www.epa.gov/hudson/revisedhhra-text.pdf)
U.S. Environmental Protection Agency (USEPA). (2000n). Superfund community involvement
guidance. EPA 540-K-01-003. Office of Solid Waste and Emergency Response,
Washington, DC. April.
U.S. Environmental Protection Agency (USEPA). (2001a). Risk assessment guidance for
Superfund: Volume in - Part A, Process for conducting probabilistic risk assessment.
EPA 540-R-02-002. Office of Emergency and Remedial Response, Washington, DC.
December.
(http://www.epa.gov/superfund/prograrns/risk/rags3a/index.htm)
U.S. Environmental Protection Agency (USEPA). (2001b). Risk assessment guidance for
Superfund, Volume I, Human health evaluation manual (Part D, Standardized planning,
reporting and review of Superfund risk assessments) Final. Office of Solid Waste and
Emergency Response, Washington, DC. December.
(http://www.epa.gov/superfund/proErams/risk/ragsd')
U.S. Environmental Protection Agency (USEPA). (2001c). Chloroform. Integrated Risk
Information System, Office of Research and Development, Washington, DC.
(http://www.epa.gov/iris/subst/0025.htm)
U.S. Environmental Protection Agency (USEPA). (2001d). Integrated exposure uptake biokinetic
model for lead in children (lEUBKwin vl.O). Washington, DC.
(http://www.epa.gov/superfund/programs/lead/ieubk.htm')
U.S. Environmental Protection Agency (USEPA). (2001e). Workshop report on the application
of 2,3,7,8-TCDD toxicity equivalence factors to fish and wildlife. EPA/630/R-01/002.
Risk Assessment Forum, Washington, DC.
(http://cfbub.epa.gov/ncea/raf/recordisplav.cfm?deid=23763')
U.S. Environmental Protection Agency (USEPA). (2001f). EPA requirements for quality
assurance project plans, EPA QA/R-5. EPA/240/B-01/003. Office of Environmental
Information, Washington, DC.
(http://www.epa.gov/QUALITY/qs-docs/r5-fmal.pdf)
-------
Page 174 Examination of EPA Risk Assessment Principles and Practices
U.S. Environmental Protection Agency (USEPA). (2001g). Risk assessment guidance for
Superfund, Volume I, Human health evaluation manual (Part E); Supplemental guidance
for dermal risk assessment, Interim. PB99-963312. Office of Solid Waste and Emergency
Response, Washington, DC. May.
U.S. Environmental Protection Agency (USEPA). (2001h). Record of decision: Alcoa (Point
Comfort)/Lavaca Bay Site Point Comfort, XX. CERCLIS #TXD008123168. Region VI,
Dallas, TX. December.
U.S. Environmental Protection Agency (USEPA). (200 li). Comprehensive five-year review
guidance. EPA 40-R-01-007. Office of Emergency and Remedial Response, Washington,
DC. June.
U.S. Environmental Protection Agency (USEPA). (200 Ij). National-scale air toxics assessment
for 1996. Science Advisory Board Preliminary Draft. EPA-453/R-01-003, Office of Air
Quality Standards and Planning, Research Triangle Park, NC.
(http://www.epa.gov/ttn/atw/nata/natsaov.html')
U.S. Environmental Protection Agency (USEPA). (2002a). Guidelines for ensuring and
maximizing the quality, objectivity, utility, and integrity, of information disseminated by
the Environmental Protection Agency. EPA/260R-02-008, Office of Environmental
Information, Washington, DC. October.
(http://www.epa.gov/oei/Qualitvguidelines')
U.S. Environmental Protection Agency (USEPA). (2002b). Lessons learned on planning and
scoping for environmental risk assessments. Science Policy Council. January.
(http://www.epa.gov/osp/spc/2cumrisk.htm)
U. S. Environmental Protection Agency (USEPA). (2002c). A review of the reference dose and
reference concentration processes. EPA/630/P-02/002F. Risk Assessment Forum,
Washington, DC. December. 192 pp.
(http://cfpub.epa.gov/ncea/cfm/recordisplav.cmi?deid=55365^
U.S. Environmental Protection Agency (USEPA). (2002d). Child-specific exposure factors
handbook (Interim report). EPA-600-P-00-002B. Office of Research and Development,
Washington, DC. September.
U.S. Environmental Protection Agency (USEPA). (2002e). 1,3-Butadiene Integrated Risk
Information System, Office of Research and Development, Washington, DC.
(http://www.epa.gov/iris/subst/0139.htm')
-------
Examination of EPA Risk Assessment Principles and Practices Page 175
U.S. Environmental Protection Agency (USEPA). (2002f). Guidance on cumulative risk
assessment of pesticide chemicals that have a common mechanism of toxicity. Office of
Pesticide Programs, Washington, DC.
(http://www.epa.gQv/oT>pfeadl/trac/science/cumulative guidance.pdf)
U.S. Environmental Protection Agency (USEPA). (2002g). Organophosphate pesticides; Revised
OP cumulative risk assessment. Office of Pesticide Programs, Washington, DC.
(http://www.epa.gov/pesticides/cumulative/rra-ops)
U.S. Environmental Protection Agency (USEPA). (2002h). Total risk integrated methodology:
TRIM.FaTE technical support document, Volume I: Description of module. EPA-453/R-
02-01 la. Office of Air Quality Planning and Standards, Research Triangle Park, NC.
U.S. Environmental Protection Agency (USEPA). (2002i). Total risk integrated methodology,
TRIM.FaTE technical support document, Volume II: Description of chemical transport
and transformation algorithms. EPA-453/R-02-01 Ib. Office of Air Quality Planning and
Standards, Research Triangle Park, NC.
U.S. Environmental Protection Agency (USEPA). (2002J). Evaluation of TRIM.FaTE, Volume I:
Region ffl, Philadelphia, PA. Approach and initial findings. EPA-453/R-02-012. Office
of Air Quality Planning and Standards, Research Triangle Park, NC.
U.S. Environmental Protection Agency (USEPA). (2002k). Supplemental guidance for
developing soil screening levels for Superfund sites. OSWER 9355.4-24. Office of Solid
Waste and Emergency Response, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (20021). Calculating upper confidence limits
for exposure point concentrations at hazardous waste sites. OSWER 9285.6-10. Office of
Emergency and Remedial Response, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (2002m). Record of decision: Palmerton Zinc
Pile, Palmerton, PA. EPD ID: PAD002395887.OU 03. EPA/ROD/R03-02/007.
U.S. Environmental Protection Agency (USEPA). (2002n). Peer consultation workshop on
approaches to polycyclic aromatic hydrocarbon (PAH) health assessment. EPA/635/R-
02/005. Office of Research and Development, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (2003a). A summary of general assessment
factors for evaluating the quality of scientific and technical information. EPA 100/B-
03/001. Science Policy Council, Washington, DC. June.
(http://www.epa.gov/oei/qualityguidelines/af home.htnfl
-------
Page 176 Examination of EPA Risk Assessment Principles and Practices
U.S. Environmental Protection Agency (USEPA). (2003b). Framework for cumulative risk
assessment. EPA/630/P-02/001F. Office of Research and Development, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (2003c). Draft final guidelines for carcinogen
risk assessment. EPA/630/P-03/001A. NCEA-F-0644A. Risk Assessment Forum,
Washington, DC.
U.S. Environmental Protection Agency (USEPA). (2003d). Supplemental guidance for assessing
cancer susceptibility from early-life exposure to carcinogens (external review draft).
EPA/630/R-03/003. Risk Assessment Forum, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (2003e). Developing relative potency factors
for pesticide mixtures; Biostatistical analyses of joint dose-response. EPA/600/R-03/052.
Office of Research and Development, National Center for Environmental Assessment,
Cincinnati, OH.
U.S. Environmental Protection Agency (USEPA). (2003f). Human health risk assessment for the
Housatonic River Rest of the river. Region I, Boston, MA. June.
(http://www.epa.gov/NE/ge/thesite/restofriver-reports.htmB
U.S. Environmental Protection Agency (USEPA). (2003g). Integrated risk information system
(IRIS). Office of Research and Development, National Center for Environmental
Assessment, Washington, DC.
(http://www.epa.gov/iris')
U.S. Environmental Protection Agency (USEPA). (2003h). The feasibility of performing
cumulative risk assessments for mixtures of disinfection by-products in drinking water.
EPA/600/R-03/051. Office of Research and Development, National Center for
Environmental Assessment, Cincinnati, OH.
U.S. Environmental Protection Agency (USEPA). (2003i). National emission standards for
hazardous air pollutants for stationary combustion turbines; Proposed rule. 40 CFR Part
63. Fed Reg 68(9) 1888-1929. January 3.
U.S. Environmental Protection Agency (USEPA). (2003J). Generic endpoints for ecological risk
assessment. EPA/630/P-02/004A. Risk Assessment Forum, Washington, DC.
U.S. Environmental Protection Agency (USEPA). (2003k). Guidance for developing ecological
soil screening levels (Eco-SSLs). OSWER Directive 92857-55. Office of Solid Waste and
Emergency Response, Washington, DC.
-------
Examination of EPA Risk Assessment Principles and Practices Page 177
U.S. Environmental Protection Agency (USEPA). (20031). Potential risks of nine rodenticides to
birds and nontarget mammals: A comparative approach. January.
(http://www.epa.gov/pesticides/rodenticidecluster/)
U.S. Environmental Protection Agency (USEPA). (2003m). Human health research strategy.
EPA 600/R-02/050. Office of Research and Development, Washington, DC. July.
U.S. General Accounting Office (USGAO). (2001). Chemical risk assessment: Selected federal
agencies' procedures, assumptions, and policies. GAO-01-810. Washington, DC. August.
(http://www.gao.gov/new.items/d01810.pdfl
U.S. Office of Management and Budget (USOMB). (2002). Guidelines for ensuring and
maximizing the quality, objectivity, utility, and integrity of information disseminated by
federal agencies. Fed Reg 67:8452-8460. February 22.
(http://www.whitehouse.eov/omb/fedreg/reproducible2.pdf)
U.S. Office of Management and Budget (USOMB). (2003a). Draft 2003 report to congress on the
costs and benefits of federal regulations. Fed Reg 68:5492-5527. February 3.
(http://www.whitehouse.gov/onib/fedreg/2Q03draft_cost-beneiit_rpt.pdf)
U.S. Office of Management and Budget (USOMB). (2003b). Informing regulatory decisions:
2003 report to congress on the costs and benefits of federal regulations and unfunded
mandates on state, local, and tribal entities. Office of Information and Regulatory Affairs,
Washington, DC. (http://www.whitehouse.gov/omb/inforeg/2003_cost-ben final_rpt.pdf)
U.S. Office of Management and Budget (USOMB). (2003c). Informing regulatory decisions:
2003 report to congress on the costs and benefits of federal regulations and unfunded
mandates on state, local, and tribal entities, Appendix D; OMB Circular A-4. Regulatory
Analysis. Office of Information and Regulatory Affairs, Washington, DC.
van Birgelen, A.P., J. van der Kolk, K.M. Fase, I. Bol, H. Poiger, A. Brouwer, and M. van den
Berg. (1994a). Toxic potency of 3,3',4,4',5-pentachlorobiphenyl relative to and in
combination with 2,3,7,8-tetrachlorodibenzo-p-dioxin in a subchronic feeding study in
the rat. Toxicol Appl Pharmacol, 127(2):209-221.
van Birgelen, A.P., J. van der Kolk, K.M. Fase, I. Bol, H. Poiger, A. Brouwer, and M. van den
Berg. (1994b). Toxic potency of 2,3,3',4,4',5-hexachlorobiphenyl relative to and in
combination with 2,3,7,8-tetrachlorodibenzo-p~dioxin in a subchronic feeding study in
the rat. Toxicol Appl Pharmacol 126(2):202-213.
-------
Page 178 _ Examination of EPA Risk Assessment Principles and Practices _
van den Berg, M., L. Birnbaum, A.T.C. Bosveld, B. Brunstrom, P. Cook, M. Feeley, J.P. Giesy,
A. Hanberg, R. Hasegawa, S.W. Kennedy, T. Kubiak, J.C. Larsen, F. X. van Leeuwen, A.
K. Liem, C. Nolt, R.E. Peterson, L. Poellinger, S. Safe, D. Schrenk, D. Tillitt, M.
Tysklind, M. Younes, F. Waern, and T. Zacharewski. (1998). Toxic equivalency factors
(TEFs) for PCBs, PCDDs, PCDFs for humans and wildlife. Environ Health Perspect 106
(12):775-792.
van der Plas, S.A., M. Haag-Gronlund, G. Scheu, L. Warngard, M. van der Berg, P. Wester, J.H.
Koeman, and A. Brouwer. (1999). Induction of altered hepatic foci by a mixture of
dioxin-like compounds with and without 2,2',4,4',5,5'-hexachlorobiphenyl in female
Sprague-Dawley rats. Toxicol Appl Pharmacol 156(l):30-39.
van der Plas, S.A., J. Sundberg, H. van den Berg, G. Scheu, P. Wester, S. Jensen, A. Bergman, J.
de Boer, J.H. Koeman, A. Brouwer. (2000). Contribution of planar (0-1 ortho) and
nonplanar (2-4 ortho) fractions of Aroclor 1260 to the induction of altered hepatic foci in
female Sprague-Dawley rats. Toxicol Appl Pharmacol 169(3):255-68.
van Wijnen, J.H., P. Clausing, and B. Brunekreff. (1990). Estimated soil ingestion by children.
Environ Res 51:147-162.
Walker, M.K., P.M. Cook, B.C. Butterworth, E.W. Zabel, and R.E. Peterson. (1996). Potency of
a complex mixture of polychlorinated dibenzo-p-dioxin, dibenzofuran, and biphenyl
congeners compared to 2,3,7,8-tetrachlorodibenzo-p-dioxin in causing fish early life stage
mortality. Fundam Appl Toxicol 30:1 78- 1 86.
Wang, P.Y., T. Kaneko, H. Tsukada, M. Nakano, and A. Sato. (1997). Dose- and
route-dependent alterations in metabolism and toxicity of chemical compounds in
ethanol-treated rats: Difference between highly (chloroform) and poorly (carbon
tetrachloride) metabolized hepatotoxic compounds. Toxicol Appl Pharmacol 142
West, P.C., M.J. Fly, R. Marans, and F. Larkin. (1989). Michigan sport anglers fish consumption
survey. A report to the Michigan Toxic Substance Control Commission. Contract No. 87-
20141. Michigan Department of Management and Budget, Lansing, MI.
West, P.C., M.J. Fly, R. Marans, F. Larkin, and D. Roesnblatt. (1993). 1991-92 Michigan sport
anglers fish consumption study. Technical Report No. 6. Michigan Department of Natural
Resources, Lansing, MI. May.
Woutersen, R.A., H.P. Til, and V.J. Feron. (1984). Sub-acute versus sub-chronic oral toxicity
study in rats: comparative study of 82 compounds. J Appl Toxicol 4:277-280.
-------
Examination of EPA Risk Assessment Principles and Practices Page 179
Yrjanheiki, E.J. (1992). Review of the models for TEFs in assessing health risks of PCDDs and
PCDFs. Toxic Sub J 12:283-288.
Zabel, E.W., P.M. Cook, and R.E. Peterson. (1995). Toxic equivalency factors of polychlorinated
dibenzo-p-dioxins, dibenzofuran and biphenyl congeners based on early life stage
mortality in rainbow trout (Oncorhynchus mykiss). Aquat Toxicol 31:315-328.
-------
Page 182 Examination of EPA Risk Assessment Principles and Practices
ADDITIONAL USEFUL WEB SITES
EPA Risk Assessment Guidelines Web site:
http://cipub,epa.gov/ncea/raf/recordisplay.cfm?deid=55907
EPA Quality System Web site: http://www.epa.eov/qualitv
EPA Science Policy Council Web site: http://www.epa.gov/osp/spc
EPA Information Quality Guidelines Web site: http://www.epa.gov/oei/qualitvguidelines
-------
-------
vvEPA
United States
Environmental Protection
Agency
Cincinnati, OH 45268
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT No. G-35
Official Business
Penalty for Private Use
$300
EPA/100/B-04/001
March 2004
www.epa.gov
Recyclad/R«yclable
Printed with vegetable-based ink on
paper that contains a minimum of
50% post-consumer fiber content
processed chlorine free
------- |