4

chapter
Risk Assessment
Elaine M. Faustman

Introduction and Historical Integrating Qualitative Aspects of Exposure Assessment

Context Risk Assessment
Mode of Action and Adverse Outcome Risk Characterization
Definitions Pathways Variation in Susceptibility
Decision Making Information Resources
Dose–Response Assessment
Hazard Identification Integrating Quantitative Aspects of Risk Perception and
Assessing Toxicity of Chemicals— Risk Assessment Comparative Analyses of Risk
Introduction Threshold Approaches
Assessing Toxicity of Chemicals: Nonthreshold Approaches Emerging Concepts
Approaches Statistical or Probability Distribution Public Health Risk Management
Structure–Activity Relationships Models
In Vitro and Short-Term Tests Models Derived from Mechanistic Summary
Animal Bioassays Assumptions Acknowledgments
Use of Epidemiologic Data in Risk Toxicological Enhancements of the
Assessment Models References

INTRODUCTION AND HISTORICAL Risk assessment was defined as a specific activity in 1983 in
the National Research Council (NRC) publication Risk Assessment
CONTEXT in the Federal Government: Managing the Process (Red Book).
In the 1970s U.S. Congress established a basic plan for environ- This book detailed steps for hazard identification, dose–response
mental laws that authorized regulatory actions to protect public assessment, exposure analysis, and characterization of risks (NRC,
health and the environment. These science-based actions provided 1983). These four basic framework steps now form the foundation
the foundation for environmental and human health risk assess- framework for risk assessment and risk management approaches
ment. Recent updates and revisions of these laws including the for many contexts such as children’s health and ecological health,
2016 Frank R. Lautenberg Chemical Safety from the 21st Century and they provide a consistent context for risk assessment frame-
Act, which modified the 1976 Toxic Substance Control Act, have works across agencies nationally and worldwide. Risk assessment
reinforced the need for toxicology evaluation. Toxicological frameworks have been used by the Health and Safety Executive
research and toxicity testing constitute the scientific core of risk Risk Assessment Policy Unit of the United Kingdom (Great Britain
assessment, which is used for evaluating potential adverse health and Health and Safety Executive, 1999) and by the World Health
impacts from chemical exposures. Considerations for risk evalu- Organization (WHO) (NRC, 2009; WHO, 2010) and build from
ation were not new, since for decades the American Conference these common components. Fig. 4-1 illustrates a framework with
of Governmental Industrial Hygienists (ACGIH) set threshold bidirectional arrows demonstrating an ideal situation where mecha-
limit values for occupational exposures and the U.S. Food and nistic research feeds directly into risk assessments and critical data
Drug Administration (FDA) established acceptable daily intakes uncertainty drives research. The common elements of the basic risk
(ADIs) for pesticide residues and food additives. In 1958, the U.S. assessment framework are shown in red highlight. Initially, attention
Congress instructed the FDA in the Delaney Clause to prohibit in risk assessment was focused on cancer risks; in subsequent years,
the addition of all substances found to cause cancer in animals noncancer endpoints were examined with similar methods and risk
or humans to the food supply. Pragmatically, this policy allowed assessments were performed across life stages and for ecosystems.
food sources that had nondetectable levels of these additives to Continuing advances in toxicology, epidemiology, exposure assess-
be declared “safe.” As advances in analytical chemistry revealed ment, biologically based modeling of adverse responses, and mod-
that “nondetects” were not equivalent to “not present,” regulatory eling of variability and uncertainty, as well as the rapidly growing
agencies were forced to develop “tolerance levels” and “accept- areas of “-omics” including genomics, transcriptomics, proteomics,
able risk levels.” Risk assessment methodologies blossomed in the metablomics, etc., have contributed to refinements in risk assess-
1970s with the rising need to address these issues and to provide a ment. Interest continues in biomarkers of exposure, effect, and
common framework for considering human and ecological health susceptibility to link exposure to disease prediction (NRC, 1989a,
effects (Albert, 1994). 1989b, 1992, 2017). Risk assessment links with the public policy
128
Research Risk assessment Risk management

Laboratory and field Hazard identification Risk characterization Development of

measurements of regulatory options.
Does the agent cause • What is the nature
Unit I

exposures. adverse health effects? and estimated • Control

Evaluation of • Structure–Activity analysis incidence of adverse • Substitute
exposed populations • In vitro tests effects in a given • Inform
and observation of • Animal bioassays population?
adverse effects. • Epidemiology • How robust is the
evidence?
Research
General Principles of Toxicology

• How certain is the Evaluation of public

needs
Dose–Response assessment evaluation? health, economic,
New mechanistic • Are susceptible social, political
What is the relationship
understandings of populations context for risk
between dose and response?
toxicity. • Susceptibility
characterized? management
• Is there a relevant options.
Research Age
mode of action?
Gene–Environment

New genomic Exposure assessment

information. What types, levels, and Policy decisions
duration of exposures are and actions.
experienced or anticipated?

Figure 4-1.  Risk assessment/risk management framework. This framework shows, under the red highlight, the four key steps of risk assessment: hazard identifi-
cation, dose–response assessment, exposure assessment, and risk characterization. It shows an interactive, two-way process where research needs from the risk
assessment process drive new research, and new research findings modify risk assessment outcomes. At the start of the risk assessment process, a risk problem
is formulated and provides context for the assessment (Ecology RA). (Adapted from Calkins et al. [1980], NRC [1993, 1994a], Risk Commission [1997], and
Gargas et al. [1999]; reprinted with permission from the National Academies Press, National Academy of Sciences.)

objectives from risk management often require extrapolations that

go far beyond the observation of actual effects. Extrapolating data
to address risks continues to generate controversy.
The National Academy of Sciences (NAS) 1994 report entitled
Science and Judgment in Risk Assessment captured in its title the
combination of qualitative and quantitative approaches essential for Problem/
effective assessment of risks (NRC, 1994). The report discussed context
in detail the challenges and provided approaches for incorporat-
ing new scientific findings into the risk assessment process. It also Evaluation Risks
highlighted ways to deal with uncertainty when insufficient scien-
tific information was available. To address these challenges, the Engage
Presidential/Congressional Commission on Risk Assessment and stakeholders
Risk Management (Risk Commission, 1997) formulated a com-
prehensive framework that (1) placed each environmental problem Actions Options
or issue into public health and/or ecological context and (2) proac-
tively called for engaging relevant stakeholders, often affected or
potentially affected community groups, from the very beginning Decisions
of the six-stage process shown in Fig. 4-2. This report emphasized
that particular exposures and potential health effects must be evalu-
ated across sources and exposure pathways and in light of multiple
endpoints, not just one chemical, in one environmental medium
(air, water, soil, food, products), for one health effect at a time.
The single-chemical, single-exposure approach had been the gen-
eral approach up to this time. The importance of defining the risk Figure 4-2.  Risk management framework for environmental health from the
problem based on this initial input is critical to risk assessment. U.S. Commission on Risk Assessment and Risk Management. The framework
comprises six stages: (1) formulating the problem in a broad public health
Problem formulation and scoping has now been defined as Phase 1
context; (2) analyzing the risks; (3) defining the options; (4) making risk
for risk-based decision making and has been outlined by the U.S. reduction decisions; (5) implementing those actions; and (6) evaluating the
Environmental Protection Agency (EPA) (U.S. EPA, 1992a) and effectiveness of the actions taken. Interactions with stakeholders are critical
reinforced by Science and Decisions, the Silver Book (NRC, 2009). across all the stages and thus have been put at the center of the framework
(Charnley and Omenn, 1997; Risk Commission, 1997).

DEFINITIONS is defined as the probability of an adverse outcome based on the

Risk assessment is the systematic scientific evaluation of potential exposure and potency of the hazardous agent(s). Although histori-
adverse health effects resulting from human exposures to hazard- cally there have been differences in how the term hazard has been
ous agents or situations (NRC, 1983, 1994; Omenn, 2000). Risk used, international risk harmonization activities have allowed for
 management. Part of this analytical–deliberative process is the cru- 129
Table 4-1 cial, too-often neglected or delayed need for considering the fears,
Example Objectives of Risk Assessment perceptions, priorities, and proposed remedies of these stakehold-
1. Protect human and ecological health ers (Risk Commission, 1997; Drew et al., 2003, 2006; Judd et al.,

Chapter 4
Toxic substances 2005; Kramer et al., 2006; Cullen et al., 2008). Underestimating
2. Prioritize testing needs potential public and media attitudes toward local polluters, other
3. Balance risks and benefits responsible parties, and relevant government agencies may lead
Drugs to what has been labeled “the outrage factor” (Sandman, 1993),
Pesticides greatly influencing the communication process and the choices for
4. Set target levels of risk risk management. Sometimes the decision makers and stakeholders

Risk Assessment
Food contaminants simply want to know the “bottom line”: whether a substance or a
Water pollutants situation is “safe” or not. Others will be keenly interested in know-
5. Set priorities for program activities ing how and why the risk estimates are uncertain and may be well
Regulatory agencies prepared to challenge underlying assumptions about context and
Manufacturers methodology. Stakeholders can also be part of a risk management
Environmental/consumer organizations solution (Judd et al., 2005; Drew et al., 2006). Perception of risk is
6. Inform green chemistry, life cycle analysis, and clinical further discussed at the end of this chapter.
alternatives
7. Estimate residual risks and extent of risk reduction after DECISION MAKING
steps are taken to reduce risks
Risk management decisions are reached under diverse statutes in
the United States. Table 4-2 lists examples of major environmental
statutes and the year of initial enactment. Some statutes specify
current consensus on hazard as reference to intrinsic toxic proper- reliance on risk alone, whereas others require a balancing of risks
ties, whereas exposure becomes an essential consideration along and benefits of the product or activity. Risk assessments provide a
with hazard for risk. Risk assessment requires an integration of valuable framework for priority setting within regulatory and health
both qualitative and quantitative scientific information. For exam- agencies, in the chemical development process within companies,
ple, qualitative information about the overall evidence and nature and in resource allocation by environmental organizations. Similar
of the endpoints and hazards is integrated with quantitative assess- approaches for risk assessment have been developed in many
ment of the exposures, host susceptibility factors, and the magni- other countries and through such international organizations as the
tude of the hazard. A description of the uncertainties and variability International Programme on Chemical Safety (IPCS) within the
in the estimates is a significant part of risk characterization and WHO. Within the IPCS, there are significant efforts toward global
an essential component of risk assessment. Analogous approaches harmonization of risk assessment methodology (WHO, 2000,
are applied to ecological risks, as demonstrated by EPA’s ecologi- 2006; WHO, 2004). The WHO Human Health Risk Assessment
cal risk assessment guidelines (U.S. EPA, 1992a, 2004). Example Toolkit: Chemical Hazards is an example (WHO, 2010). Their
objectives of risk assessment are outlined in Table 4-1. goals include use of a common risk assessment framework with
The phrase “characterization of risk” reflects the combination acceptance of different risk assessment approaches through this
of qualitative and quantitative analysis. The term “hazard char- common understanding and agreement on basic principles of test-
acterization” versus “hazard identification” has also been seen in ing and evaluation. Table 4-2 lists the recent Occupational Safety
both U.S. EPA (2012b) and WHO (1999) documents. This addi- and Health Administration (OSHA) Hazard Communication
tion recognizes the difficulties in evaluating only hazard informa- Standard 2015, which now uses the Globally Harmonized System
tion without dose–response information. Unfortunately, many users of Classification and Labelling of Chemicals (GHS). This standard
equate risk assessment with quantitative risk assessment, generating represents an arduous yet highly meritorious effort to harmonize
a number for an overly precise risk estimate, while ignoring crucial global labeling of chemicals. This GHS labeling has provided a
information about the uncertainties of the risk assessment, mode of common language for countries to use in trade and manufacturing
action (MOA), and type of effect across species or context. that facilitates safety in the workplace and for chemicals in transit.
Risk management refers to the process by which policy actions An emphasis on hazard and not risk has been one of the more con-
are chosen to control hazards identified in the risk assessment stage troversial issues across nations.
of the framework (Fig. 4-2). Risk managers consider scientific A major challenge for risk assessment, risk communication,
evidence and risk estimates—along with statutory, engineering, and risk management is to work across disciplines to demonstrate
economic, social, and political factors—in evaluating alternative the biological plausibility and clinical significance of the conclu-
options and choosing among those options (Risk Commission, sions from epidemiologic studies, animal bioassays, short-term in
1997). vitro and in vivo tests, and structure–activity studies of chemicals
Risk communication is the challenging process of making thought to have potential adverse effects. Biomarkers of expo-
risk assessment and risk management information comprehen- sure, effect, or individual susceptibility can link the presence of a
sible to the public, community groups, lawyers, local elected offi- chemical in various environmental compartments to specific sites of
cials, judges, business people, labor, environmentalists, etc. (NRC, potential action in target organs and to host health impacts and dis-
1989a, 1996, 2004; Morgan, 1993; Sandman, 1993; Fischhoff et al., ease (NRC, 1989a, 1989b, 1989c, 1992, 2017; Adami et al., 2011).
1996). Such communication requires an understanding of the level Mechanistic investigations of the actions of specific chemicals
of risk response needed, how and which individuals and organi- can help us penetrate the “black box” approach of simply count-
zations should be informed, and what the audience hears versus ing tumors, for example, in exposed animals in routine bioassays.
what was intended as the risk communication message. Involving Greater appreciation of the mechanisms and extent of individual
stakeholders in the initial problem formulation as well as in the variation in susceptibility to adverse health impacts among humans
translation of the bottom line has become an essential part of risk can improve protection of susceptible populations and better relate
130
Table 4-2
Examples of Major Toxic Chemical Laws in the United States by Responsible Agency and Year of Initial Enactment
EPA Air pollutants Clean Air Act, 1970
Water pollutants Federal Water Pollution Control Act, 1948
Unit I

Drinking water Safe Drinking Water Act, 1974

Pesticides Fungicides, Insecticides & Rodenticides Act (FIFRA), 1947; Food Quality
Protection Act (FQPA), 1996
Ocean dumping Marine Protection, Research and Sanctuaries Act (MPRSA) or Ocean
Dumping Act, 1972
General Principles of Toxicology

Toxic chemicals Toxic Substances Control Act (TSCA), 1976; The Frank R. Lautenberg
Chemical Safety for the 21st Century Act (2016)
Hazardous wastes Resource Conservation and Recovery Act (RCRA), 1976
Abandoned hazardous wastes Superfund (CERCLA), 1980
CEQ Environmental impacts National Environmental Policy Act (NEPA), 1969
OSHA Workplace Occupational Safety and Health (OSH) Act, 1970; Hazard Communication
Standard (HCS) and Globally Harmonized Systems of Classification and
Labelling of Chemicals (GHS, 2015).
FDA Foods, drugs, and cosmetics Food and Drugs Act, 1906; Food, Drugs and Cosmetics Act (FDC), 1938;
FDA Modernization Act, 1997
CPSC Dangerous consumer products Consumer Product Safety Act (CPSA), 1972*
DOT Transport of hazardous materials Hazardous Materials Transportation Act (HMTA), 1975

Abbreviations: CEQ, Council for Environmental Quality (now Office of Environmental Policy); CPSC, Consumer Product Safety Commission; DOT,
Department of Transportation; EPA, Environmental Protection Agency; FDA, Food and Drug Administration; OSHA, Occupational Safety and Health
Administration.
*
Note that numerous states have enacted reporting requirements related to toxic chemicals in children’s consumer products. The Washington Children’s
Safe Product Act of 2008 is one example.

findings in animals to the characterization of risk for humans. Challenge Program (U.S. EPA, 2011b). Industry participants in this
Identifying individual behavioral and social risk factors can be program committed to filling the data gaps for HPV chemicals.
critically important both to the characterization of risk and to the Since its inception, there have been over 2200 chemicals sponsored
reduction of risk. for additional testing. International efforts such as the Organization
for Economic Cooperation and Development (OECD) and screen-
ing information data set (SIDS) program are also addressing data
HAZARD IDENTIFICATION needs related to HPV chemicals, highlighted by the publication of
Assessing Toxicity of Chemicals— their Manual for Investigation of HPV Chemicals (OECD, 2013).
Data requirements for specific chemicals can vary greatly by
Introduction compound type, use, and applicable regulatory statutes. Introduced
Throughout this book, a common toxicology paradigm (Fig. 4-3) in 2003 and approved in 2006, the European Union released a
forms the basis for our toxicity assessments. For risk assessment regulatory framework for Registration, Evaluation, Authorisation
this paradigm is a critical underpinning where delivery represents and Restriction of Chemicals (REACH) (REACH, 2011). Under
exposure and absorption, distribution, metabolism, and excretion the REACH framework, since 2007, all stakeholders that sell or
(ADME) determining the amount, rate, form, distribution, and fate produce compounds in the European Union submit dossiers that
of the chemical. The key initial impacts of the chemical interacting include physical, chemical, and toxicological data as well as envi-
with the target link the early cellular dysfunction with early bio- ronmental fate and transport information to inform risk assessment
markers effect. Dysrepair is identified as endpoint impacts across studies for all chemicals in use in Europe. The REACH data is avail-
levels of biological investigation with observations at the molecu- able via the European Chemicals Agency registered substances list
lar, cellular, and tissue levels. (ECHA, 2017). Chemical safety assessments are submitted prior
In order to assess the toxicity of chemicals, information from to approval in an approach similar to premanufacturing notices
four example types of studies is used: structure–activity relation- (PMNs) in the United States for the EPA (U.S. EPA, 2011c).
ships (SAR), in vitro or short-term studies, in vivo animal bioas- Table 4-3 shows requirements and costs for one example class
says, and human epidemiologic studies. In many cases, toxicity of chemicals, pesticides (Stevens, 1997; U.S. EPA, 1998b, 2000b),
information for chemicals is limited. For example, in 1998, the in the United States (40 CFR 158.340). It also illustrates current
EPA evaluated high production volume (HPV) chemicals (those international efforts to align these testing guidelines by listing
produced in excess of one million pounds per year) to ascertain examples of the harmonized 870 test guidelines (U.S. EPA, 2010a;
the availability of chemical hazard data. The EPA study found that Patlewicz et al., 2013). The emphasis in REACH is on non–in vivo
for 43% of these HPV chemicals, there were no publicly available animal tests. Also, REACH has utilized the international globally
studies for any of the basic toxicity endpoints (U.S. EPA, 1998a). In harmonized labeling laws for hazard identification (GHS). Some
response, the EPA established a voluntary program called the HPV of the successes from REACH have stimulated new changes in the
 Absorption 131

1 Distribution
Toxicology
DELIVERY Metabolism Paradigm

Chapter 4
Excretion

2
INTERACTION

Risk Assessment
WITH TARGET
Gene
Expression
Cell T
Signaling Electrically
3
Excitable Cells O
CELLULAR
DYSFUNCTION
X
Internal
Cell Maintenance I
Maintenance
External
Maintenance
C

Protein
I
Molecular Lipid T
4 DNA
Cellular
DISREPAIR Y
Apoptosis
Tissue
Proliferation

Figure 4-3.  Toxicology paradigm. A common paradigm forms the basis for our toxicology assessments and subsequent translation for assessment of health
risks. For risk assessment this paradigm is a critical underpinning where delivery represents exposure and absorption, distribution, metabolism, and elimination
(ADME) determining the amount, the rate, the form and distribution of the chemical and its metabolites. The key initial impacts of the chemical interacting
with the target links the early cellular dysfunction with early biomarkers effect. Disrepair is identified as endpoint which impacts across levels of biological
investigation and observation at the molecular, cellular, and tissue levels.

United States via amendments for the Toxic Substances Control Act alerts such as N-nitroso or aromatic amine groups, amino azo dye
(TSCA). Recent release of the Integrated Chemical Environment structures, or phenanthrene nuclei are clues to prioritize chemicals
(ICE) reference database provides a welcome integrated access to in for additional evaluation as potential carcinogens. SAR informa-
vivo and in vitro data to support Tox21 activities (Bell et al., 2017). tion for specific noncancer health endpoints can be challenging.
The database of known developmental toxicants limits SARs to a
few chemical classes, including chemicals with structures related
Assessing Toxicity of Chemicals: to those of valproic acid, retinoic acid, phthalate esters, and glycol
Approaches ethers (NRC, 2000). More recently, -omic technologies have been
Structure–Activity Relationships used to supplement SAR relationship databases, as seen with the
Given the cost of two to four million dollars and the 3 to 5 years creation of the National Cancer Institute’s (NCI) gene expression
required for testing a single chemical in a lifetime rodent carcino- database (NCI, 2017) and the U.S. EPA Computational Toxicity
genicity bioassay, initial decisions on whether to continue develop- Program Screening Database (ToxCastTM) (U.S. EPA, 2017b).
ment of a chemical, to submit a premanufacture notice (PMN), or SARs have been used for assessment of complex mixtures of
to require additional testing may be based largely on results from structurally related compounds. Prominent applications have been
structure–activity relationships (SARs) and limited short-term the assessment of risks associated with 2,3,7,8-tetrachlorodibenzo-
assays. p-dioxin (TCDD), related chlorinated and brominated dibenzo-
A chemical’s structure, solubility, stability, pH sensitivity, p-dioxins, dibenzofurans, and planar biphenyls, and chemicals
electrophilicity, volatility, and chemical reactivity can be impor- generally present as mixtures in the environment. Toxicity equiva-
tant information for hazard identification. Historically, certain key lence factors (TEFs) are used to evaluate health risks associated
molecular structures have provided regulators with some of the most with closely related chemicals. For the TCDD class, this is based
readily available information on the basis of which to assess hazard on a common mechanism of aryl hydrocarbon (Ah) receptor induc-
potential. For example, 8 of the first 14 occupational carcinogens tion (U.S. EPA, 1994a). The estimated toxicity of environmental
were regulated together by the OSHA as belonging to the aromatic mixtures containing these chemicals is calculated as the sum of
amine chemical class. The EPA Office of Toxic Substances relies on the product of the concentration of each chemical multiplied by its
SARs to meet deadlines to respond to PMN for new chemical manu- TEF value. The WHO has organized efforts to reach international
facture under the Toxic Substances Control Act (TSCA). Structural consensus on the TEFs used for polychlorinated biphenyls (PCBs),
132
Table 4-3
Example EPA/FIFRA Requirements for Hazard Evaluation of Pesticides
REVISED 870 TYPE OF APPROXIMATE
GUIDELINE TOXICITY STUDY TEST SYSTEM OBJECTIVE COST/STUDY* (US$)
Unit I

1100 Acute oral Rats Define toxic dose by ingestion 2,700

1200 Acute dermal Rabbits Define toxic dose by absorption through skin 2,000
1300 Acute inhalation Rats Define toxic dose by inhalation 6,800
General Principles of Toxicology

2400 Ocular Rabbits Assess eye irritation/injury 2,000

2500 Skin irritation Rabbits Assess skin irritation/injury 130
2600 Sensitization Guinea pigs Assess allergic potential 4,000
6100–6855 Neurotoxicity†,‡ Hens/rats Assess nervous system injury 34,000‡
5100–5915 Mutagenicity§ In vivo/in vitro Determine genotoxic potential; screen for 6,800
carcinogenicity
3050–3465 Range-finding§ Rats Determine effects following repeated doses; set dose level 95,000
Subacute Mice 95,000
(28- to 90-day*)
Dogs For longer studies 136,000
Rabbits 101,000
Rats Identify target organs; set dose 258,000
Mice Levels for chronic studies 258,000
4200–4300 Carcinogenicity/ Rats Determine potential to induce tumors; define 1,900,000
chronic toxicity dose–response relationships (lifetime)
Mice 1,087,000
Dogs Determine long-term toxic effects (1 year) 543,000
3550–3800 Reproduction and Rats Determine potential to cause fetal abnormalities and 686,000
teratogenicity effects on development, fertility, pregnancy, and devel-
opment of offspring over at least two generations
Rabbits
7485 Toxicokinetics Rats, mice Determine and quantitate the metabolic fate of a 135,000
pesticide
8600 Developmental Rats Determine potential to cause adverse developmental 700,000
neurotoxicity outcomes following exposure during gestation and lac-
tation on gross neurological behavioral abnormalities
and function, motor activity, neuropathology and brain
weight measurements throughout post-natal develop-
ment and adulthood
*
Indicates per assay cost using Consumer Price Index (CPI) conversion factors to determine 2010 value (oregonstate.edu/cla/polisci/sahr/sahr). Estimates
were provided using guideline 81 series.
†
Required for organophosphate insecticides only.
‡
Additional neurotoxicity tests 81-7, 81-8, 82-6, 82-7, and 83-6 have been added to requirements for certain materials and can include tests such as func-
tional observational battery, motor activity, developmental landmarks, and learning and memory assessments (Sette, 1991). Costs listed for this type
of study are only those for the initial study, not additional testing.
§
Range-finding studies are not required, but provide justification for setting dose levels in required studies. EPA-required studies can include reverse muta-
tion assays in Salmonella, forward mutation assays in mammalian cells—for example, Chinese hamster ovary cells and mouse lymphoma L5178Y (the
locus cells)—and in vivo cytogenetics (Dearfield, 1990).
SOURCES: Adapted from Stevens (1997), and updated with revised EPA 870 guideline information (U.S. EPA, 2010a, 2017). For details on changes in
the health effects test guidelines reflective of the harmonization of the toxicology guidelines between the Office of Pollution Prevention and Toxics
(OPPT) and the Office of Pesticide Programs (OPP), within the EPA and with the Organization for Economic Cooperation and Development (OECD)
guidelines, see EPA (U.S. EPA, 2010a, 2017) (https://www.epa.gov/test-guidelines-pesticides-and-toxic-substances/series-870-health-effects-test-
guidelines).
polychlorinated dibenzo-p-dioxins (PCDDs), and polychlorinated Overall, progress in developing and validating new in vitro 133
dibenzofurans (PCDFs) for both humans and wildlife, and has assays has been episodic with recent encouraging efforts especially
updated its values and published the supporting database (Van den due to U.S. EPA’s ToxCast program (Kavlock et al., 2012). In the late
Berg et  al., 1998, 2006; Haws et  al., 2006). Under the auspices 1990s, the Interagency Coordinating Committee on the Validation

Chapter 4
of the WHO, the dioxin-like PCB congeners have been assigned of Alternative Methods (ICCVAM) of the NTP reinvigorated the
TEFs reflecting their toxicity relative to TCDD, which itself has validation process in the United States as a result of Public Law 103-
been assigned a TEF of 1.0. 43 and coordinated cross-agency issues relating to development,
Computerized SAR methods have, in general, given disap- validation, acceptance, and national/international harmonization of
pointing results in the U.S. National Toxicology Program (NTP) toxicological test methods for use in risk assessments (ICCVAM,
rodent carcinogenicity prediction challenges (Ashby and Tennant, 2017). The committee has put forth recommendations for over 40

Risk Assessment
1994; Omenn et al., 1995; Benigni and Zito, 2004). More success- alternative safety testing methods using various short-term/in vitro
ful are the efforts of pharmaceutical companies using combinato- assays such as the cell-free corrosivity test, and the mouse local
rial chemistry and three-dimensional (3D) molecular modeling lymph node assay for assessing chemical potential to elicit allergic
approaches to design ligands (new drugs) that can sterically fit reactions (ICCVAM, 1999, 2011a; NIEHS, 1999a, 1999b). In 2006,
into the “receptors of interest.” However, for environmental pol- the committee released a document that reviewed in vitro acute
lutants where selective binding to specific receptors is rare, these toxicity methods.This review was led by the National Toxicology
applications of SAR have had limited success within risk assess- Program Center for the Evaluation of Alternative Toxicological
ment. Applications for endocrine disrupting chemicals and evalu- Methods (United States) and Interagency Coordinating Committee
ation of their binding to hormone receptors have been a recent on the Validation of Alternative Methods (United States) (NIEHS,
success from ToxCast, with a proposal for using receptor-bind- 2006). The European Centre on the Validation of Alternative
ing assays for screening endocrine disruptor potential and as an Methods (ECVAM) is now known as the European Union Reference
acceptable approach without the additional need for assays such Laboratory for alternatives to animal testing (EURL-ECVAM, avail-
as the Hershberger assay (U.S. EPA, 2015). A renewed interest in able at https://eurl-ecvam.jrc.ec.europa.eu/).
quantitative SAR (QSAR) approaches has also resulted from the EURL-ECVAM has been very active given the visibility of
need to evaluate engineered nanomaterials where the tremendous animal rights issues in the European Union, and this Center was
number of unique new products has highlighted the necessity of originally formed to “support the development, validation, and
using QSAR approaches to handle the avalanche of novel untested acceptance of methods that could reduce, refine, or replace [3 Rs]
materials (Maynard et al., 2006; Liu et al., 2011; Nel et al., 2013). the use of laboratory animals.” Early successes of ECVAM are
Efforts within REACH have also emphasized the potential for described in Hartung et al. (2003). However, a recent report for the
use of SAR as similar chemicals are collectively evaluated using Center for Alternatives for Animal Testing highlights the slow prog-
a concept of “read-across.” Substances whose physicochemical, ress on full replacement of animal testing, especially for repeated
toxicological, and ecotoxicological properties are similar can be dose toxicity testing, carcinogenicity, or reproductive toxicity testing
grouped as a “category” of substances when they have a common (Adler et al., 2011; Hartung et al., 2011). It does highlight numer-
functional group, common precursor or breakdown product, or a ous successes for evaluating sensitization and toxicokinetics (TK).
common pattern of potency. Most recent successes have been driven by the European Union–
conceived free legislation for cosmetic testing. International efforts
In Vitro and Short-Term Tests to reduce the number of animals required for chemical safety testing
The next level of biological information obtained within the hazard have been established between ICCVAM, Korea, Japan, Canada,
identification process includes assessment of the test chemical in in and ECVAM (Table 4-4).
vitro or short-term tests, ranging from bacterial mutation assays per-
formed entirely in vitro to more elaborate short-term tests, such as
skin painting studies in mice or altered rat liver foci assays conducted Table 4-4
in vivo. For example, EPA mutagenicity guidelines call for assess-
ment of reverse mutations using the Ames Salmonella typhimurium Example Assays Used for Evaluation of Cosmetics:
assay; forward mutations using mammalian cells, mouse lym- Multiple OECD Guideline Studies Reviewed and
phoma L5178Y, Chinese hamster ovary, or Chinese hamster lung Designed for These Endpoints
fibroblasts; and in vivo cytogenetics assessment (bone marrow OECD TYPES OF
metaphase analysis or micronucleus tests) (U.S. EPA,1986, 2005a). GUIDELINES TOXICITY STUDY TEST SYSTEM OBJECTIVE
Chap. 8 discusses uses of these assays for identifying chemical car-
cinogens and Chap. 9 describes in detail various assays of genetic OECD 439 Irritation Skin Ethic Skin
and mutagenic endpoints. Other assays evaluate specific health end- EpiSkin®
points such as developmental toxicity (Chap. 21) (Faustman, 1988; OECD 492 Irritation MatTek Epi Eye
Whittaker and Faustman, 1994; Brown et al., 1995; Lewandowski Ocular®
et al., 2000; NRC, 2000; Spielmann et al., 2006), reproductive tox-
OECD 437 Irritation Bovine Eye
icity (Chap. 10) (Gray, 1988; Harris et al., 1992; Shelby et al., 1993;
Opacity, Cornea and Cornea
Yu et  al., 2009), neurotoxicity (Chap.  18) (Atterwill et  al., 1992;
Permeability Opacity
Costa, 2000), and immunotoxicity (Chap.  12) (ICCVAM, 1999,
Permeability
2011a). Less information is available on the extrapolation of these
(BCOP)
test results for noncancer risk assessment than for the mutagenic-
ity or carcinogenicity endpoints; however, mechanistic informa- OECD 431 Corrosion Skin Ethic Skin
tion obtained in these systems has been applied to risk assessment EpiSkin®
(Abbott et al., 1992; U.S. EPA, 1994b; Leroux et al., 1996; NRC, OECD 435 Corrosion Corrositex® Skin
2000; Smith et al., 2016a, 2016b).
134 The validation and application of short-term assays are par- are those that test exposure and biological pathways of most rele-
ticularly important to risk assessment because such assays can be vance to predicted or known human exposure pathways. Bioassays
designed to provide information about mechanisms of effects, and, for reproductive and developmental toxicity and other noncancer
moreover, they are fast and inexpensive compared with lifetime bio- endpoints have a similar rationale. The NTP serves as a resource
assays (McGregor et al., 1999; Collins et al., 2008; NRC, 2017). for conducting, designing, and evaluating bioassays for cancer as
Unit I

Validation considerations for in vitro assays, like other kinds of well as noncancer evaluation. The NTP Office of Human Health
tests, require determination of their sensitivity (e.g., ability to iden- Assessment and Translation serves as a resource to the public and
tify true carcinogens), specificity (e.g., ability to recognize non- regulatory agencies regarding the interpretation and assessment of
carcinogens as noncarcinogens), and predictive value for the toxic adverse effects of chemicals (NTP, 2011b). The WHO International
endpoint under evaluation. However, the recent NRC publication Agency for Research on Cancer (IARC) has evaluated over 900
General Principles of Toxicology

proposes a series of modified validation considerations that can agents “of which more than 400 have been identified as carcino-
further define pathways to regulation acceptance (NRC, 2017). The genic, probably carcinogenic or possibly carcinogenic to humans”
societal costs of relying on such tests, with false positives (noncar- (WHO, 2011b).
cinogens classified as carcinogens) and false negatives (true car- Consistent features in the design of standard cancer bioassays
cinogens not detected), are the subject of a value-of-information include testing in two species and both sexes, with 50 animals per
model for testing in risk assessment and risk management (Lave dose group and near-lifetime exposure. Important choices include
and Omenn, 1986; Lave et al., 1988; Omenn and Lampen, 1988). the strains of rats and mice, the number of doses, and dose lev-
Efforts to improve our ability to utilize short-term tests for els (typically 90%, 50%, and 10–25% of the maximally tolerated
carcinogenicity prediction include increased attention to improv- dose [MTD]), and details of the required histopathology (number
ing the mechanistic basis of short-term testing. Examples of this of organs to be examined, choice of interim sacrifice pathology,
approach include the development and application of several knock- etc.). The NTP Web site lists details on study designs and protocols
out transgenic mouse models as shorter-term in vivo assays to iden- (NTP, 2011a, 2017a). Positive evidence of chemical carcinogenic-
tify carcinogens (Nebert and Duffy, 1997; Tennant et al., 1999). ity can include increases in number of tumors at a particular organ
Specific assays for evaluating mechanisms such as the function of site, induction of rare tumors, earlier induction (shorter latency) of
the estrogen receptor are an example where in vitro test methods commonly observed tumors, and/or increases in the total number
are used to predict endocrine disruptor actions (ICCVAM, 2011b, of observed tumors. The NTP has added a perinatal exposure phase
2012). The primary use of short-term tests in the United States con- that includes gestation (in utero exposure via placenta) and postnatal
tinues to be for mechanistic evaluations with the hope of inform- lactation (via the mother’s milk) to more directly evaluate the sig-
ing chemical-specific information or overall MOA. In that context, nificance of early-life exposures for cancer incidence (NTP, 2017b).
results from short-term assays have impacted risk assessments. For The cancer bioassay, originally designed for hazard identifica-
example, evidence of nonmutagenicity in both in vitro and in vivo tion, is frequently used to evaluate dose–response. The relatively
short-term assays have played an essential role, allowing regulators limited number of evaluated doses and the use of high doses have
to consider nonlinear cancer risk assessment paradigms for non- caused issues for low-dose extrapolations and have limited the use of
genotoxic carcinogens (U.S. EPA, 1999). Mechanistic information cancer bioassays as a “gold standard” for prediction of human car-
from short-term in vitro assays can also be used to extend the range cinogenicity risk (McClain, 1994; Cohen, 1995; Risk Commission,
of biological observations available for dose–response assessment. 1997; Rodericks et al., 1997; Capen et al., 1999; Rice et al., 1999).
In addition, for developmental toxicity assessment, assay methods Tumors may be increased only at the highest dose tested, which is
that acknowledge the highly conserved nature of developmen- usually at or near a dose that causes systemic toxicity (Ames and
tal pathways across species have accelerated the use of a broader Gold, 1990). Second, even without toxicity, the high dose may trig-
range of model organisms and assay approaches for noncancer ger different events than do low-dose exposures and high doses can
risk assessments (NRC, 2000). Toxicity Testing in the 21st Century saturate important metabolism and elimination pathways.
and Toxicity Pathway-Based Risk Assessment emphasized such Rats and mice give concordant positive or negative results
approaches for using model organisms and tiered toxicity strategies in approximately 70% of bioassays, so it is unlikely that rodent/
for risk assessment (NRC, 2007a, 2010, 2017). human concordance would be higher (Lave et al., 1988). Haseman
and Lockhart (1993) concluded that most target sites in cancer
Animal Bioassays bioassays showed a strong correlation (65%) between males and
Animal bioassays have been a key component of the hazard identi- females—especially for forestomach, liver, and thyroid tumors—
fication process. See example guidelines for organ/ system-specific so they suggested, for efficiency, that bioassays could rely on a
endpoints (U.S. EPA, 1991, 1988, 1996a, 1998c). A basic premise combination of male rats and female mice. Even when concordant,
of risk assessment is that chemicals that cause tumors in animals positive results are observed, there can still be large differences in
can cause tumors in humans. All human carcinogens that have been potency, as observed in aflatoxin-induced tumors in rats and mice.
adequately tested in animals produce positive results in at least In this example, an almost 100,000-fold difference in susceptibil-
one animal model. Thus, “although this association cannot estab- ity to aflatoxin B1 (AFB1)–induced liver tumors is seen between
lish that all chemicals and mixtures that cause cancer in experi- the sensitive rat and trout species and the more resistant mouse
mental animals also cause cancer in humans, nevertheless, in the strains. Genetic differences in the expression of cytochrome P450
absence of adequate data on humans, it is biologically plausible and glutathione S-transferases explain most of these species differ-
and prudent to regard chemicals and mixtures for which there is ences and suggest that humans may be as sensitive to AFB1-induced
sufficient evidence of carcinogenicity in experimental animals as liver tumors as rats (Eaton and Gallagher, 1994; Eaton et al., 1995,
if they presented a carcinogenic risk to humans” (IARC, 2000)—a 2001). These species differences have been supported by research
reflection of the “precautionary principle.” The U.S. EPA cancer results (Groopman and Kensler, 1999; Kensler et  al., 2011) and
guidelines (U.S. EPA, 2005a) also assume relevance of animal bio- have been extended within epidemiologic studies to demonstrate
assays unless lack of relevance for human assessment is specifi- the interaction of hepatitis C infection with AFB1 exposure to more
cally determined. In general, the most appropriate rodent bioassays fully explain elevated human liver cancer risks.
 135
Table 4-5
Examples of Mechanistic Considerations for Carcinogens: Explanation for Special Cases of Rodent Bioassay Data
Lacking Relevance for Human Risk Evaluation

Chapter 4
MECHANISM FOR SUSCEPTIBLE ILLUSTRATIVE CHEMICAL
SYSTEM TARGET ORGAN SPECIES SPECIES DIFFERENCES AGENTS
Urinary tract Renal tumors in Chemicals bind to α2U globu- α2U-Globulin male rat-specific Unleaded gasoline, 1,4-dichlo-
male rats lin, accumulation in target kid- low-molecular-weight pro- robenzene, d-limonene,
ney cells, increased necrosis, tein not found in female rats, isophorone, dimethyl-methyl

Risk Assessment
increased regenerative hyper- humans, mice, monkeys phosphonate, perchloro­
plasia, renal tubular calcifica- ethylene, pentachloroethane,
tion neoplasia hexachloroethane
Bladder Reactive hyperplasia from Rodent exposure levels exceed Saccharin, melamine, nitrilo-
cytotoxic precipitated solubility, not relevant for triacetic acid, fosetyl-A2
chemicals human exposure
Gastric Forestomach Direct oral gavage, local cyto- Rodent gavage treatment, BHA, propionic acid, ethyl
toxicity, hyperplasia exposure conditions not rel- acrylate
evant for human exposure
Endocrine Thyroid gland Alteration in thyroid homeo- Lack of thyroid-binding pro- Ethylene bisdithiocarbamate,
tumors stasis, decreased thyroid hor- tein in rodents versus humans, fungicides, amitrol, goitro-
mone production, sustained decreased t1/2 for T4, increased gens, sulfamethazine
increase in thyroid stimulat- TSH levels in rodents
ing hormone (TSH), thyroid
tumors
Respiratory Rat lung Overwhelming clearance High-dose effects seen with Various particles, titanium
mechanisms rodent models dioxide

SOURCES: McClain (1994), Neumann and Olin (1995), Oberdorster (1995), Omenn et al. (1995), Risk Commission (1997), Rodericks et al. (1997), Da
Rocha et al. (2014), and Yamada et al. (2017).

Lifetime bioassays have been enhanced with the collection of Table 4-5 also illustrates quantitative considerations important
such additional mechanistic data and with the assessment of mul- for determining human relevance of animal bioassay information.
tiple noncancer endpoints. It is feasible and desirable to integrate For example, doses of compounds so high as to exceed solubility
such information together with data from mechanistically oriented in the urinary tract outflow lead to tumors of the urinary bladder in
short-term tests and biomarker and genetic studies in epidemiology male rats following crystal precipitation and local irritation lead-
(Perera and Weinstein, 2000). In the example of AFB1-induced liver ing to hyperplasia. Such precipitates are known to occur following
tumors, AFB1–DNA adducts have proved to be an extremely useful saccharin or nitrilotriacetic acid bioassay exposures (Cohen et al.,
biomarker. A highly linear relationship was observed between liver 2000). The decision to exclude saccharin from the NTP list of sus-
tumor incidence (in rats, mice, and trout) and AFB1–DNA adduct pected human carcinogens reaffirms the nonrelevance of such high-
formation over a dose range of five orders of magnitude (Eaton and dose responses for likely human exposure considerations (Neumann
Gallagher, 1994). Such approaches may allow for an extension of and Olin, 1995; NTP, 2016). A gross overloading of the particle
biologically observable phenomena to doses lower than those lead- clearance mechanism of rat lungs via directly administered parti-
ing to frank tumor development, and help to address the issues of cles, as was seen in titanium dioxide (TDO) exposures, resulted in
extrapolation over multiple orders of magnitude to predict response the EPA’s delisting of TDO as a reportable toxicant for the Clean Air
at environmentally relevant doses. Act Toxic Release Inventory (Oberdorster, 1995; U.S. EPA, 2016).
Table 4-5 presents some mechanistic details about rodent Other rodent responses not likely to be predictive for humans
tumor responses that are no longer thought to be predictive of can- include localized forestomach tumors after gavage. Ethyl acrylate,
cer risk for humans. This table lists examples of both qualitative which produces such tumors, was delisted on the basis of extensive
and quantitative considerations useful for determining relevance of mechanistic studies (NTP, 2016). In general, for risk assessment,
rodent tumor responses for human risk evaluations. An example of it is desirable to use the same route of administration as the likely
qualitative considerations is the male rat kidney tumors observed exposure pathway in humans to avoid such extrapolation issues.
following exposure to chemicals that bind to α2u-globulin (e.g., Despite the example of forestomach tumors, tumors in unusual
unleaded gasoline, 1,4-dichlorobenzene, d-limonene). The α2u- sites—such as the pituitary gland, the eighth cranial nerve, or the
globulin is a male-rat-specific low-molecular-weight protein not Zymbal gland—should not be immediately dismissed as irrelevant,
found in female rats, humans, or other species, including mice and because organ-to-organ correlation is often lacking (NRC, 1994).
monkeys (McClain, 1994; Neumann and Olin, 1995; Oberdorster, The EPA cancer guidelines provide a good list of considerations
1995; Omenn et  al., 1995; Risk Commission, 1997; Rodericks for evaluating relevance in the sections on evaluating weight of evi-
et al., 1997). dence (U.S. EPA, 2005a).
136
Table 4-6
Example of Three Types of Epidemiologic Study Designs
METHODOLOGICAL TYPE OF STUDY
Unit I

ATTRIBUTES COHORT CASE–CONTROL CROSS-SECTIONAL

Initial classification Exposure–nonexposure Disease–nondisease Either one
Time sequence Prospective Retrospective Present time
Sample composition Nondiseased individuals Cases and controls Survivors
General Principles of Toxicology

Comparison Proportion of exposed with dis- Proportion of cases with exposure Either one
ease
Rates Incidence Fractional (%) Prevalence
Risk index Relative risk–attributable risk Relative odds Prevalence
Advantages Lack of bias in exposure; yields Inexpensive, small number of Quick results
incidence and risk rates subjects, rapid results, suitable for
rare diseases, no attrition
Disadvantages Large number of subjects Incomplete information, biased Cannot establish causation (ante-
required, long follow-up, attrition, recall, problem in selecting con- cedent consequence), population
change in time of criteria and trol and matching, yields only of survivors, inadequate for rare
methods, costly, inadequate for relative risk—cannot establish diseases
rare diseases causation, population of survivors

SOURCES: Gamble and Battigelli (1978, 1991).

In an attempt to improve the prediction of cancer risks to Environmental and occupational epidemiologic studies are fre-
humans, transgenic mouse models have been developed as possible quently opportunistic. Studies begin with known or presumed
alternative to the standard 2-year cancer bioassay. Transgenic mod- exposures, comparing exposed with nonexposed individuals, or
els use knockout or transgenic mice that incorporate or eliminate with known cases, compared with persons lacking the particular
a gene that has been linked to human cancer. The NTP evaluated diagnosis.
some of these models and found the p53-deficient (p53+/− het- Table 4-6 shows examples of epidemiologic study designs
erozygous) and Tg.AC (v-Ha-ras transgene) models to be particu- and provides clues on types of outcomes and exposures evaluated.
larly useful in identifying carcinogens and mechanisms of action Although convincing, there are important limitations inherent in
(Bucher, 1998; Chhabra et al., 2003). The use of transgenic models epidemiologic studies. Robust exposure estimates are often difficult
has the power to improve the characterization of key cellular and to obtain as they are frequently done retrospectively (e.g., through
MOA of toxicological responses (Mendoza et al., 2002; Gribble retrospective job history records). Also, because many important
et al., 2005). However, these studies have been used primarily for health effects have long latency before clinical manifestations
mechanistic characterization rather than for hazard identification. appear, reconsideration of relevant populations can be challenging.
Transgenic models have been shown to reduce cost and time as Another challenge for interpretation is that there are often exposures
compared with the standard 2-year assay, but they have also been to multiple chemicals, especially when a lifetime exposure period is
shown to be somewhat limited in their sensitivity (Cohen, 2001). considered. There is frequently a trade-off between detailed infor-
As stated in the current EPA cancer guidelines, transgenic models mation on relatively few persons and very limited information on
should not be used to replace the standard 2-year assay, but can be large numbers of persons. Contributions from lifestyle factors, such
used in conjunction with other types of data to assist in the inter- as smoking and diet, are important to assess as they can have a
pretation of additional toxicological and mechanistic evidence (U.S. significant impact on cancer development. Human epidemiologic
EPA, 2005a). An especially promising emerging area of research studies can provide both very useful information for hazard assess-
has been with the Collaborative Cross. A series of genetically ment and quantitative information for data characterization. Good
defined (fully sequenced genomes, 20× coverage) mice referred illustrations of epidemiologic studies and their interpretation for
to as the Collaborative Cross have been established to investigate toxicological evaluation are available (Regalado et al., 2006; Gill
genetic and environmental influences on toxicological response in et al., 2011).
mice (Chesler et al., 2008) and these strains have already improved Three example types of epidemiologic study designs—
our understanding of mammalian genes that predispose to cancer cross-sectional studies, cohort studies, and case–control studies—are
(Koturbash et al., 2011). In a similar manner, human cells have also detailed in Table 4-6. Cross-sectional studies survey groups of
provided a method to explore human diversity in response directly humans to identify risk factors (exposure) and disease, but are not
in vitro (Eduati et al., 2015). useful for establishing cause and effect. Cohort studies can evalu-
ate individuals selected on the basis of their exposure to a chemi-
Use of Epidemiologic Data in Risk Assessment cal under study. Thus, based on exposure status, these individuals
The most convincing lines of evidence for human risk are well- are monitored for development of disease. These prospective stud-
conducted epidemiologic studies in which a positive association ies monitor over time individuals who initially are disease-free to
between exposure and disease has been observed (NRC, 1983). determine the rates at which they develop disease. In case–control
studies, subjects are selected on the basis of disease status: disease Integrating Qualitative Aspects of 137
cases and matched cases of disease-free individuals. Exposure his-
tories of the two groups are compared to determine key consistent Risk Assessment
features in their exposure histories. All case–control studies are ret- Qualitative assessment of hazard information should include a con-

Chapter 4
rospective studies. sideration of the consistency and concordance of findings, includ-
In risk assessment, epidemiologic findings can be judged by ing a determination of the consistency of the toxicological findings
the following criteria: strength of association, consistency of obser- across species and target organs, an evaluation of consistency
vations (reproducibility in time and space), specificity (uniqueness across duplicate experimental conditions, and a determination of
in quality or quantity of response), appropriateness of temporal the adequacy of the experiments to consistently detect the adverse
relationship (did the exposure precede responses?), dose–response, endpoints of interest.

Risk Assessment
biological plausibility and coherence, verification, and analogy Qualitative assessment of animal or human evidence is done
(biological extrapolation) (Hill, 1965; Faustman et al., 1997; WHO, by many agencies, including the EPA and IARC. Similar evidence
1999; Adami et al., 2011). These same criteria have been used for classifications have been used for both the animal and human evi-
evaluating MOAs where integrated considerations of both human dence categories by both agencies. These evidence classifications
and animal studies are done. have included levels of “sufficient, limited, inadequate, and no evi-
Epidemiologic study designs should also be evaluated for their dence” (U.S. EPA, 1999, 1994b, 2005a), or “evidence suggesting
power of detection, appropriateness of outcomes, verification of lack of carcinogenicity” (IARC, 2000). For both agencies, these
exposure assessments, completeness of assessing confounding fac- classifications are used for an overall weight-of-evidence approach
tors, and general applicability of the outcomes to other populations for carcinogenicity classification. Current categories expand on
at risk. Power of detection is calculated using study size, variability, these initial categories including “Group 4 probably not carcino-
accepted detection limits for endpoints under study, and a speci- genic to humans” (IARC, 2017) or “not likely to be carcinogenic to
fied significance level (Healey, 1987; EGRET, 1994; Dean et al., humans” (U.S. EPA, 2005a).
1995). Meta-analysis is used with epidemiologic studies to combine Weight of evidence is an integrative step used by the EPA
results from different studies using weighting of results to account to “characterize the extent to which the available data support the
for sample size across studies. The importance and complexity of hypothesis that a chemical causes cancer in humans” (U.S. EPA,
human studies for risk assessment are shown in evaluations with 1999, 2005a). It is the process of “weighing” all of the evidence to
arsenic and dioxin (U.S. EPA, 2010b, 2010d). reach a conclusion about carcinogenicity. The weight of evidence can
Advances from the human genome project have increased consider both the quality and quantity of data as well as any underly-
sophistication of molecular biomarkers and have improved the mech- ing assumptions. The evidence includes data from all of the hazard
anistic bases for epidemiologic hypotheses. This has allowed epide- assessment and characterization studies such as SAR data, in vivo and/
miologists to get within the “black box” of statistical associations or in vitro studies, and epidemiologic data. Using this type of infor-
and forward our understanding of biological plausibility and clinical mation and weight-of-evidence approach, the EPA includes hazard
relevance. “Molecular epidemiology,” the integration of molecular descriptors to define carcinogenic potential and to provide a measure
biology into traditional epidemiologic research, is an important of clarity and consistency in the characterization narrative: “carcino-
focus of human studies where improved molecular biomarkers of genic to humans,” “likely to be carcinogenic to humans,” “sugges-
exposure, effect, and susceptibility have allowed investigators to tive evidence of carcinogenic potential,” “inadequate information to
more effectively link molecular events in the causal disease path- assess carcinogenic potential,” and “not likely to be carcinogenic to
way. Epidemiologists can now include the contribution of potential humans.” In this section, approaches for evaluating cancer endpoints
genetic factors with environmental risk factors for the determination are discussed for carcinogens. Similar weight-of-evidence approaches
of the etiology, distribution, and prevention of disease. Highlighting have been proposed for reproductive risk assessment (refer to suf-
the potential power of genetic information to epidemiologic studies, ficient and insufficient evidence categories in the EPA’s guidelines
the Human Genome Epidemiology (HuGE) Network was launched for reproductive risk [U.S. EPA, 1996a] and considerations by NTP).
in 1998, providing a literature database of published, population- The Institute for Evaluating Health Risks defined an “evaluation
based epidemiologic studies of human genes (Khoury, 1999). process” by which reproductive and developmental toxicity data can
With the advance of genomics, the range of biomarkers has be consistently evaluated and integrated to ascertain their relevance
grown dramatically and includes identification of single-nucleotide for human health risk assessment (Moore et al., 1995; Faustman
polymorphisms (SNPs), genomic profiling, transcriptome analysis, et al., 2011). This evaluation process supported the implementation
and proteomic analysis (Simon and Wang, 2006). Implications of of EPA’s guidelines for developmental toxicity risk assessment (U.S.
these improvements for risk assessment are tremendous, as they EPA, 1991) and for the NTP’s Office of Health Assessment and
provide an improved biological basis for extrapolation across the Translation (formerly CERHR) (NTP, 2011b). Application of such
diversity of human populations and allow for improved cross- carefully deliberated approaches for assessing noncancer endpoints
species comparisons with rodent bioassay information because of has helped avoid the tendency to list chemicals as yes or no (positive
evolutionarily conserved response pathways and ontologies (NRC, or negative) without human relevancy information. Application of
2000). In addition, genomics allows for “systems-based” under- systematic review methods for evaluation of chemicals at the NTP
standing of disease and response, moving risk assessment away (Birnbaum et al., 2013; Rooney et al., 2014; NTP, 2017), EPA, and
from a linear, single-event-based concept and improving the bio- WHO is under assessment and modification to facilitate the method-
logical plausibility of epidemiologic associations (Toscano and ological review of the ever-widening body of literature on chemical
Oehlke, 2005; NRC, 2010). Human databases such as SNP Track substructures and nonhuman studies (NTP, 2017).
from the FDA provide current information on genes and proteins Mode of Action and Adverse Outcome Pathways
that contain polymorphisms that can affect human response, espe- In the 2005 EPA cancer guidelines the importance of using “weight
cially for drug response (U.S. FDA, 2017a, 2017b). An extensive of evidence” to arrive at insights for possible mode of action (MOA)
battery of tools to link phenotypic and functional changes as well was identified (U.S. EPA, 2005a). MOA information describes key
as drug toxicity and interaction information to genetics is available. events and processes leading to molecular and functional effects
138 Adverse Outcome Pathways (AOPs) and Biomarkers Link Responses and Disease Pathogenesis

Macro-Molecular Cellular Organ Organism Population

Toxicant
Interactions Responses Responses Responses Responses

Chemical Receptor–Ligand Gene Activation Altered Physiology Lethality Structure

Unit I

Endpoints Properties Interaction Disrupted

Protein Impaired Extinction
DBA Binding Production Homeostasis Development
Protein Oxidation Altered Altered Tissue Impaired
Signaling Development/ Reproduction
Function
General Principles of Toxicology

Key Initiating
Biomarkers Exposure Biological Response Disease
Events

Susceptibility

Assessment Tools/ Toxicity Pathways Regulatory Endpoints

Chemical Properties
Methods

QSAR: HTS MTS Bioassays Adverse

Quantitative High Medium In Vitro Outcomes
Structure–Activity Throughput Throughput and
Relationship Systems Systems In Vivo

Figure 4-4.  Adverse outcome pathways (AOPs) and biomarkers link responses and describe disease pathogenesis. Relationships between AOPs and biomarkers
of exposure, effect, susceptibility, and disease are illustrated across disease response. Methods used to collect information along the disease continuum include
quantitative SAR (QSAR), high and medium throughput in vitro systems, and bioassays (other in vivo and in vitro assessments).

that would in general explain the overall process of cancer devel- is linked through a series of events including early molecular events
opment. In many cases these could be plausible hypothesized known as molecular initiating events (MIEs) through responses
MOAs for specific toxicity endpoints, but the detailed mechanistic at the cellular, organ, and ultimately to organism and population
nuances of the pathway might not yet be fully known. The EPA has level adverse outcomes. By linking these events the AOP approach
used such MOA information to suggest nondefault approaches for defines key event relationships that resemble the causal relation-
cancer risk assessments and for evaluating toxicity of compounds ships identified using MOA approaches. Perhaps most importantly
with common MOAs in cumulative risk assessments (U.S. EPA, for risk assessors is that these approaches allow for the identifica-
1996b, 1998a). tion of a temporal series of events that can be evaluated in vivo and
Within the EPA’s carcinogenic risk assessment guidelines, the in vitro and across species using related biomarkers of exposure,
MOA framework considers evidence from animal studies, relevance biological effect, and response. It also can incorporate biomarkers
to humans, and life stage or population susceptibility (U.S. EPA, of susceptibility. Fig. 4-4 shows these linkages and illustrates how
2005a). Chemical-specific adjustment factors for interspecies dif- information from methods such as structure–activity relationship
ferences and human variability have been proposed and build upon assessments can link high- and medium-throughput assays with
guidance developed by the WHO’s International Programme on high content assay outcomes across in vitro assessments with bio-
Chemical Safety Harmonization Project (WHO, 2000). Critical to markers of effect seen at the organ, organism, and population levels
the MOA development is the use of “criteria of causality” consid- with adverse outcomes and disease. These useful concepts can pro-
erations, which build upon the Hill criteria used in epidemiology vide both a qualitative and quantitative frame for risk assessment.
(Hill, 1965; Faustman et al., 1997; U.S. EPA, 1999; Klaunig et al.,
2003), and considerations of dose–response relationships and tem-
poral associations, as well as the biological plausibility, coherence, DOSE–RESPONSE ASSESSMENT
strength, consistency, and specificity of the postulated MOA.
Building on concepts from MOA analyses, Ankley proposed
Integrating Quantitative Aspects of
a structured framework for ecological risk assessment entitled Risk Assessment
Adverse Outcome Pathways that links biological events with Quantitative considerations in risk assessment include dose–
adverse outcomes and disease (Ankley et al., 2010). This has gained response assessment, exposure assessment, variation in suscep-
applicability for human risk assessment as well and the OECD has tibility, and characterization of uncertainty. For dose–response
developed a program to facilitate its development and distribution assessment, varying approaches have been proposed including
using a knowledge base and wiki facilitated IT site (OECD, 2017). threshold versus nonthreshold endpoints. Traditionally, in the
Fig. 4-4 shows how biomarkers and adverse-outcome path- United States, threshold approaches have been applied for assess-
ways (AOPs) can be linked to describe adverse outcomes and dis- ment of noncancer endpoints, and nonthreshold approaches have
ease pathogenesis. In this figure the initial exposure to a chemical been used for cancer endpoints. As we have learned more about
nongenotoxic mechanisms of carcinogenicity, these processes have 60 139
been evaluated using threshold approaches. Each approach and its
inherent assumptions are discussed below, as are efforts to include *
50 I
more detailed mechanistic considerations to harmonize these

Chapter 4
approaches (Bogdanffy et al., 2001).
In general, human exposure data for prediction of human 40 * H
response to environmental chemicals are quite limited; thus, animal

% response
bioassay data have primarily served as the basis for most quantita-
tive risk assessments and have required extrapolation for human 30
health risk prediction. The risk assessor, however, is normally inter-

Risk Assessment
ested in low environmental exposures when considering human risk, *
20 G
exposures that are well below the experimentally observable range
of responses in most animal assays. Thus, methods for extrapolating
F
from high dose to low dose as well as extrapolating from animal 10
A E
risk to human risk are required and comprise a major emphasis of
dose–response assessment. B C D
The fundamental basis of the quantitative relationships between
exposure to a chemical and the incidence of an adverse response 1 2 3 4 5
is the dose–response assessment. Analysis of dose–response rela- T Dose (mg/kg BW)
tionships must start with the determination of the critical effects to
be quantitatively evaluated. It is usual practice to choose the most Figure 4-5.  Dose–response curve. This figure is designed to illustrate a typi-
cal dose–response curve with points F to I indicating the biologically deter-
robust data sets with adverse effects occurring at the lowest levels mined responses. Statistical significance of these responses is indicated
of exposure from studies using the most relevant exposure routes. with the symbol “*.” Point E (■) represents a dose near the lower end of the
The “critical” adverse effect is defined as the significant adverse observed dose–response range, below which extrapolation to lower doses
biological effect that occurs at the lowest exposure level (Barnes can occur for cancer risk estimates (U.S. EPA, 2005a). Point F is the high-
and Dourson, 1988). The EPA has issued toxicity-specific guide- est nonstatistically significant response point; hence, it is the “no observed
lines that are useful in identifying such critical effects (for devel- adverse effect level” (NOAEL) for this example. Point G is the “lowest
observed adverse response level” (LOAEL). Point T represents a threshold
opmental toxicity [U.S. EPA, 1991], reproductive toxicity [U.S.
for response curve D. Curves A to D show some options for extrapolating
EPA, 1996a], neurotoxicity [U.S. EPA, 1998c], and cancer [U.S. the dose–response relationship below the range of biologically observed
EPA, 1994b, , 1999, 2005a]). International Programme on Chemical data points and POD.
Safety (IPCS) has compiled a document cited as the Principles for
Modeling Dose–Responses for the Risk Assessment of Chemicals
the responses being evaluated in order to place statistical observa-
(WHO, 2006b). It outlines key concepts and considerations for
tions in context. Significance thus usually refers to both biological
dose–response evaluations within the context of a risk assessment.
and statistical criteria (Faustman et al., 1994) and is dependent on
the number of dose levels tested, the number of animals tested at
Threshold Approaches each dose, and background incidence of the adverse response in the
Approaches for characterizing threshold dose–response relation- nonexposed control groups. The NOAEL should not be perceived
ships include identification of “no observed adverse effect level” as risk-free, as several reports have shown that the response of
(NOAEL) or “lowest observed adverse effect level” (LOAEL). On NOAELs for continuous endpoints averages 5% risk, and NOAELs
the dose–response curve illustrated in Fig. 4-5, the doses tested based on quantal endpoints can be associated with risk of greater
in the bioassay are given as F, G, H, and I. The statistical signifi- than 10% (Allen et al., 1994a; Faustman et al., 1994).
cance of points G, H, and I is indicated using an asterisk (*). The As described in Chap. 2, approaches for characterizing dose–
NOAEL (F) is identified as the highest nonstatistically significant response relationships include identification of effect levels such as
dose tested; in this example the NOAEL occurs at approximately LD50 (dose producing 50% lethality), LC50 (concentration produc-
2 mg/kg body weight. Point G is the LOAEL (~2.3 mg/kg body ing 50% lethality), ED10 (dose producing 10% response), as well as
weight), as it is the lowest dose tested with a statistically significant NOAELs. NOAELs have traditionally served as the basis for risk
effect. Lines A to D represent possible extrapolations below the assessment calculations, such as reference doses (RfDs) or accept-
point of departure (POD), which is represented on this figure as a able daily intake (ADI) values. RfDs or concentrations (RfCs) are
square (■) and is labeled as point E. The POD is used to specify estimates of daily exposure (oral or inhalation, respectively) to a
the estimated dose near the lower end of the observed dose range, chemical that is assumed to be without adverse health impact in
below which extrapolation to lower exposures is necessary (U.S. humans. The ADIs are used by the WHO for pesticides and food
EPA, 2005a). In Fig. 4-5, the POD occurs at 10% effective dose or additives to define “the daily intake of chemical, which during an
ED10. The type of extrapolation below the POD is depending on the entire lifetime appears to be without appreciable risk on the basis
type of data available. The importance of choosing an appropriate of all known facts at that time” (WHO, 1962; Dourson et al., 1985).
POD and extrapolation type is discussed further in the following RfDs (first introduced in Chap. 2) and ADI values typically are cal-
sections of this chapter. From these modeled approaches, model D culated from NOAEL values by dividing by uncertainty factor (UF)
shows a model where the threshold (T) represents the dose below and/or modifying factor (MF) (Dourson and Stara, 1983; Dourson
which no additional increase in response is observed. and DeRosa, 1991; U.S. EPA, 1991):
In general, most animal bioassays are constructed with suf-
ficient numbers of test animals to detect biological responses at the NOAEL
RfD = ,
10% response range; however, this is also dependent on endpoint UF × MF
and the background rate of the endpoint in control animals. The risk NOAEL
ADI = .
assessor should always understand the biological significance of UF × MF
140
Species Human
differences variability
Unit I

10 10
General Principles of Toxicology

Kinetics Dynamics Kinetics Dynamics

Figure 4-6.  Toxicokinetic (TK) and toxicodynamic (TD) considerations inherent in interspecies and interindividual extrapolations. TK refers to the processes of
absorption, distribution, elimination, and metabolism of a toxicant. TD refers to the actions and interactions of the toxicant within the organism and describes
processes at organ, tissue, cellular, and molecular levels. This figure shows how uncertainty in extrapolation both across and within species can be considered
as being due to two key factors: a kinetic component and a dynamic component. Challenges remain for extrapolating information on human variation to specific
populations (NRC, 2009). Refer to the text for detailed explanations.

Tolerable daily intakes (TDIs) can be used to describe intakes clinical conditions (Renwick, 1991, 1999; Johnson et al., 1997).
for chemicals that are not “acceptable” but are “tolerable” as they Toward this goal, Renwick has separated the intraspecies and inter-
are below levels thought to cause adverse health effects. These are species UFs into two components: toxicokinetics (TK) and toxico-
calculated in a manner similar to ADI. Historically, dividing by the dynamic (TD) aspects (Renwick, 1991, 1999; Johnson et al., 1997).
UFs allows for interspecies (animal-to-human) and intraspecies Fig. 4-6 shows these distinctions. A key advantage of this approach
(human-to-human) variability with default values of 10 each. An is that it provides a structure for incorporating scientific information
additional UF is used to account for experimental inadequacies— on specific aspects of the overall toxicological process into the RfD
for example, to extrapolate from short-exposure-duration studies to calculations; thus, relevant data can replace a portion of the overall
a situation more relevant for chronic study or to account for inad- “uncertainty” surrounding these extrapolations. Current WHO guid-
equate numbers of animals or other experimental limitations. If only ance uses a 4.0- and 2.5-fold factor for the TK and TD interspecies
a LOAEL value is available, then an additional 10-fold factor is components, respectively (WHO, 2005), and interindividual TK and
commonly used to arrive at a value more comparable to a NOAEL. TD factors of 3.16 (Renwick and Lazarus, 1998). The NAS publi-
For developmental toxicity endpoints, it has been demonstrated cations (NRC, 2009, 2017) have emphasized the need to consider
that the application of the 10-fold factor for LOAEL-to-NOAEL population variability and context in risk assessment. In examples
conversion is too large (Allen et al., 1994a). Traditionally, a safety described in this NRC report, high background incidence of cancers
factor of 100 would be used for RfD calculations to extrapolate or other adverse health outcomes in the population should be con-
from a well-conducted animal bioassay (10-fold factor for animal- sidered before setting “acceptable” levels of additional risks (NRC,
to-human variability) and to account for human variability in 2009).
response (10-fold factor for human-to-human variability). NOAEL values have also been utilized for risk assessment by
MFs are values that can be used to adjust the UFs if data on evaluating a “margin of exposure” (MOE), where the ratio of the
mechanisms, pharmacokinetics, or relevance of the animal response NOAEL determined in animals and expressed as mg/kg per day is
to human risk are available. For example, if there is kinetic infor- compared with the level to which a human may be exposed. For
mation suggesting there is very similar metabolism for a particular example, consider the case where human exposures to a specific
compound in rats and humans, producing the same active target chemical are calculated to be solely via drinking water, and the
metabolite, then—rather than using a 10-fold UF to divide the total daily intake of the compound is 0.04 mg/kg per day. If the
NOAEL from the animal toxicity study to obtain a human relevant NOAEL for neurotoxicity is 100 mg/kg per day, then the MOE
RfD—a factor of 3 for that UF might be used. Of particular interest would be 2500 for the oral exposure route for neurotoxicity. Such
is the addition of an extra 10-fold factor through the Food Quality a large value is reassuring to public health officials. Low values of
Protection Act (FQPA) to ensure protection of infants and children MOE indicate that the human levels of exposure are close to levels
(U.S. EPA, 1996b). Under this law an additional UF is added to for the NOAEL in animals. There is usually no factor included in
ensure protection of children’s health and it is currently being used this calculation for differences in human or animal susceptibility
for determining allowable pesticide chemical residues. This factor or animal-to-human extrapolation; thus, MOE values of less than
is designed to take into account potential prenatal and postnatal tox- 100 have been used by regulatory agencies as flags for requiring
icity and to overcome the incompleteness of toxicity and exposure further evaluation.
data (FQPA, PL 104-170). Illustrative discussions on how such a The NOAEL approach has been criticized on several points,
legislatively mandated UF might be applied are available for chlor- including that (1) the NOAEL must, by definition, be one of the
pyrifos (Schardein and Scialli, 1999; U.S. EPA, 2000b, 2016; Zhao experimental doses tested and (2) once this is identified, the rest of
et al., 2006). the dose–response curve is ignored. Because of these limitations,
To reduce uncertainty in calculating RfDs and ADIs, there an alternative to the NOAEL approach, the benchmark dose (BMD)
has been a transition from the use of traditional 10-fold UFs to the method, was proposed (Crump, 1980). In this approach, the dose–
use of data-derived and chemical-specific adjustment factors. Such response is modeled and the lower confidence bound for a dose at a
efforts have included reviewing the human pharmacologic literature specified response level (benchmark response [BMR]) is calculated.
from published clinical trials (Silverman et al., 1999) and develop- The BMR is usually specified at 1%, 5%, or 10%. Fig. 4-7 shows
ing human variability databases for a large range of exposures and the BMD using a 10% BMR (BMD10) and a 95% lower confidence
 60 widely spaced test doses will limit the utility of these assays for any 141
type of quantitative assessments, whether NOAEL- or BMD-based
approaches.
Upper confidence limit
50 on estimated risk Nonthreshold Approaches

Chapter 4
As Fig. 4-5 shows, numerous dose–response curves can be pro-
posed in the low-dose region of the dose–response curve if a thresh-
40
old assumption is not made. Because the risk assessor may need to
extrapolate beyond the region of the dose–response curve for which
% Response

experimentally observed data are available, the choice of models

30 Dose–response fitted to to generate curves in this region has received lots of attention. For

Risk Assessment
experimental data nonthreshold responses, methods for dose–response assessments
have models for extrapolation to de minimus (10−4 to 10−6) risk lev-
20
els at very low doses, far below the biologically observed response
range and far below the effect levels evaluated for threshold
responses.
10 The EPA cancer guidelines define the dose–response methods
as requiring two steps: (1) defining the “POD” or the lowest dose
associated with adverse effects within the range of the experimen-
tal data and (2) the extrapolation from the POD to low environ-
2 3 4 5 mentally relevant exposure levels based on experimental data. The
Dose extrapolation can be done with a linear model or a nonlinear model
BMDL10 BMD10
with this choice dependent on the amount and type of experimen-
Figure 4-7. Illustration of benchmark dose (BMD) approach. This figure tal data available. Risk estimates using the linear model (biologi-
shows the BMDL10 and BMD10, the lower confidence limit and dose (ED10) cal response increases proportionally with level of exposure) are
associated with a 10% incidence of adverse response, respectively (Kavlock generally higher than those using nonlinear models. For example,
et al., 1995; U.S. EPA, 2005a, 2012a). For reference, points are labeled as in 2003, the EPA released the Dioxin Reassessment and used the
described in the legend for Fig. 4-5. BMD dose–response method. The reassessment used a POD based
on a 1% response with a linear extrapolation model based on the
position that the scientific data were inadequate to rule out the EPA’s
standard default linear assumption. The EPA Dioxin Reassessment
bound on dose (BMDL10). The BMDx (with x representing the per-
was evaluated in a National Academies Report in 2006 (National
cent BMR) is used as an alternative to the NOAEL value for RfD
Academy of Science, 2006) and this review discussed remaining
calculations. Thus, the RfD would be:
extrapolation issues. In the EPA’s recent draft of dioxin risk assess-
BMD x or BMDL x ment (U.S. EPA, 2010b) a revised TCDD cancer risk assessment is
RfD = . presented using the upper bound on the regression slope for esti-
UF × MF •
mating cancer mortality risk from the epidemiology evaluation by
The EPA has developed software for the application of BMD Cheng et al. (2006). The top 5% of exposure estimates from this
methods and a technical guidance document to provide guidelines epidemiologic study were excluded as the dose-related changes in
for application of BMDs for both cancer and noncancer endpoints cancer mortality plateaued at high doses. The EPA lagged exposures
(U.S. EPA, 2000a, 2012a). The final guidance can be found at by 15 years and adjusted cancer risk based on cumulative TCDD fat
https://www.epa.gov/risk/benchmark-dose-technical-guidance. The concentrations. Their analysis included an estimate of TCDD can-
EPA recently released an updated version (2.7) of its BMD software cer risks above the TCDD background levels found in the NIOSH
(U.S. EPA, 2017a). worker cohort analyzed by Cheng et al. (2006). Below the POD, the
The 2009 WHO guidelines for dose–response modeling for EPA assumed a default linear slope value. This document represents
risk assessment also discuss BMD approaches (WHO, 2006b). an example recent interpretation of cancer risk assessment by the
Both the EPA and WHO guidelines distinguish NOAEL versus EPA for data-rich compounds with multiple cancer risk estimates
BMD-based approaches. Harmonization of approaches available for animal and human studies as well as how extensive mechanistic
for assessments is discussed in the WHO documents. and kinetic information is considered. This draft report provides
The BMD approach has been applied to study noncancer end- insight into recent interpretations of the EPA cancer guidelines
points, including developmental (Allen et al., 1994a) and repro- (U.S. EPA, 2005a). Increased attention has been given to the use
ductive toxicity (Auton, 1994). The most extensive studies with of mechanistic data (such as biomarker data) to extend the POD to
developmental toxicity have shown that BMD05 values were similar lower regions of the dose–response curve.
to a statistically derived NOAEL for a wide range of developmental
toxicity endpoints and that results from using generalized dose– Statistical or Probability Distribution Models
response models were similar to biologically based dose–response Two general types of dose–response models exist for extrapolation:
(BBDR) models designed specifically to represent unique features statistical (or probability distribution) and mechanistic models
of developmental toxicity testing (Faustman et al., 1999). (Krewski and Van Ryzin, 1981). The distribution models are based
Advantages of the BMD approach include (1) the ability to on the assumption that each individual has a tolerance level for a
take into account the dose–response curve; (2) the inclusion of a test chemical and that this response level is a variable following a
measure of variability (confidence limit); and (3) the use of a con- specific probability distribution function. These responses can be
sistent BMR level for RfD calculations across studies. Limitations modeled using a cumulative dose–response function. Chap. 2 dis-
in the animal bioassays include minimal number of test doses for cusses the common normal distribution pattern (see Fig. 2-5). A log-
evaluation, shallow dose responses, and use of study designs with probit model estimates the probability of response at a specified
142 dose (d); thus, P(d) = ϕ [a + β log d], where ϕ is the cumulative If the true value of λ1 is replaced with λ1* (the upper confidence
function for a standard normal distribution of the log tolerances limit of λ1), then a linearized-multistage model can be derived
with standard deviations σ and mean μ, a equals μ/σ, and β equals where the expression is dominated by d at low doses. The slope on
the slope of the probit line (−1/σ). The probit curve at low doses this confidence interval, q1* , has been used by the EPA for quantita-
usually assumes an S shape. Chap. 2 discusses determination of tive cancer assessment. To obtain an upper 95% confidence interval
Unit I

the LD50 value from such a curve. However, extrapolation of the on risk, the q1* value (risk/Δ dose in mg/kg per day) is multiplied
experimental data from 50% response levels to a “safe,” “accept- by the amount of exposure (mg/kg per day). Thus, the upper bound
able,” or “de minimus” level of exposure—for example, one in a estimate on risk (R) is calculated as:
million risk above background—illustrates the huge gap between
scientific observations and highly protective risk limits (sometimes R = q1* [ risk (mg / kg per day )−1 ] × exposure(mg / kg per day ).
General Principles of Toxicology

called virtually safe doses [VSDs] or those corresponding to a 95%

upper confidence limit on adverse response rates). This relationship has been used to calculate a “VSD,” which
The log-logistic model was derived from chemical kinetic represents the lower 95% confidence limit on a dose that gives an
theory. The probability of response at dose d is defined as P(d) = “acceptable level” of risk (e.g., upper confidence limit for 10–6
[1 exp(a + β log d)]−1. Like the probit model, this model defines excess risk). In the past the Integrated Risk Information System
sigmoidal curves that are symmetrical around the 50% response (IRIS) (U.S. EPA, 2011d) developed by the EPA has given q* values
level; however, the log-logistic curves approach the 0% and 100% for many environmental carcinogens (U.S. EPA, 2011d). Because
response levels with a shallow curve shape. The logit and probit both the q1* and VSD values are calculated using 95% confidence
curves are indistinguishable in fitting the data in the region of the intervals, the values are believed to represent conservative, protec-
response curve where experimentally derived data are present tive estimates.
(Brown, 1984; Hartung, 1987; Allen et al., 1994b). The EPA has utilized the LMS model to calculate “unit risk
estimates” in which the upper confidence limit on increased indi-
Models Derived from Mechanistic Assumptions vidual lifetime risk of cancer for a 70-kg human breathing 1 μg/m3
This modeling approach designs a mathematical equation to of contaminated air or drinking 2 L per day of water containing 1
describe dose–response relationships that are consistent with pos- ppm (1 mg/L) is estimated over a 70-year life span. The example
tulated biological mechanisms of response. These models are based given in Fig. 4-8 shows the calculation of incremental lifetime can-
on the idea that a response (toxic effect) in a particular biological cer risk (ILCR) of skin cancer using soil exposure and q* values
unit (animal, human, pup, etc.) is the result of the random occur- for inorganic arsenic. If a POD-based approach is used for low-
rence of one or more biological events (stochastic events). dose extrapolation, the model would be used to extrapolate from
Radiation research spawned a series of such “hit models” for the POD or upper confidence limit and these slopes would be con-
cancer modeling, where a hit is defined as a critical cellular event sidered for the q* values.
that must occur before a toxic effect is produced. The simplest Toxicological Enhancements of the Models
mechanistic model is the one-hit (one-stage) linear model in which Three exemplary areas of research that have improved the models
only one hit or critical cellular interaction is required for a cell to used in risk extrapolation are time to tumor information, physio-
be altered. For example, based on somatic mutation theory, a single logically based TK modeling, and BBDR modeling (Albert, 1994).
mutational change would be sufficient for a cell to become cancer- Chap. 7 discusses in detail improvements in our estimation of expo-
ous through a transformational event and dose-independent clonal sure and offers approaches on how to model “target internal effec-
expansion. The probability statement for these models is P(d) = tive dose” in risk assessment rather than just using single-value
1 − exp(−λd), where λd equals the number of hits occurring during “external exposure doses.” In this chapter we discuss the BBDR
a time period. Using this approach, a single molecule of a geno- modeling.
toxic carcinogen would have a minute but finite chance of causing BBDR modeling aims to make the generalized mechanistic
a mutational event. models discussed in the previous section more clearly reflect spe-
As theories of cancer have grown in complexity, so too cific biological processes. Measured rates are incorporated into the
have these hit-based mechanistic models. Multihit models have mechanistic equations to replace default or computer-generated
been developed that can describe hypothesized single-target values. For example, the Moolgavkar–Venson–Knudson (MVK)
multihit events, as well as multitarget, multihit events in car- model is based on a two-stage model for carcinogenesis, where two
cinogenesis. The probability statements for these models are mutations are required for carcinogenesis and birth and death rates
λd
P (d ) = ∫ x k −1 exp(−x ) / Γ(k )dx , where Γ(k) denotes the gamma of cells are modeled through clonal expansion and tumor formation.
function with k being the critical number of hits for the adverse This model has been applied effectively to human epidemiologic
response. The Weibull model has a dose–response function with data on retinoblastoma. In animal studies, kidney and liver tumors
characteristics similar to those of the multihit models, where the in the 2-acetylaminofluorene (2-AAF) “mega mouse” study, rat
response equation is P(d) = 1 − exp[–λdk]. Here again, k is the lung tumors following radiation exposure, rat liver tumors follow-
critical number of hits for the toxic cellular response. ing N-nitrosomorpholine exposure, respiratory tract tumors follow-
Armitage and Doll (1957) developed a multistage model ing benzo[a]pyrene exposure, and mouse liver tumors following
for carcinogenesis that was based on these equations and on the chlordane exposure have been modeled (Cohen and Ellwein, 1990;
hypothesis that a series of ordered stages was required before a cell Moolgavkar and Luebeck, 1990).
could undergo mutation, initiation, transformation, and progression There have been concerted research efforts to improve our
to form a tumor. This relationship was generalized by Crump (1984) mechanistic understanding of response at lower exposures. Two
by maximizing the likelihood function over polynomials, so that the excellent examples of the use of biomarkers to supplement our
probability statement is: exploration of these dose–response relationships are seen in Bailey
et  al. (2009) and Vlaanderen et  al. (2011). Hormetic responses
P (d ) = 1 − exp[−(λ0 + λ1d 1 + λ2 d 2 +  + λk d k )]. are U-shaped dose–response relationships where small doses of
A. Ingestion of arsenic from soil—point estimation method 143
Soil concentration × Ingestion rate × Exposure duration × Exposure frequency
× Bioavailability × q* = Lifetime cancer risk (ILCR)
Body weight × Averaging time from skin cancer

Chapter 4
2300 mg/kg × 100 mg per day × 30 years × 350 days per year
× 0.09 × 1.50 (mg/kg per day)–1 = 1.8 × 10–4
70 kg × 25,550 days

B. Ingestion of arsenic from soil—probabilistic methods

Risk Assessment
Soil concentration Ingestion rate Exposure duration Exposure frequency
0.040 0.025 0.110 0.014

0.030 0.019 0.083 0.011

Probability

0.020 × 0.013 × 0.055 × 0.007

0.010 0.006 0.028 0.004

0.000 0.000 0.000 0.000

0 1750 3500 5250 7000 0 75 150 225 300 0 25 50 75 100 0 100 200 300 400
As concentration, mg/kg Ingestion rate, Exposure duration, Exposure frequency,
mg per day years days per year
Bioavailability Slope factor, q* = ILCR skin cancer
0.025 1.0 0.298
95th
0.019 0.75 0.223 percentile
Body weight Averaging time
0.016 1.0
× 0.013 × 0.50 0.149
0.012 0.75 0.006 0.25 0.075
Probability

0.008 × 0.50 0.000

0.0 0.125 0.25 0.38 0.50
0.00
1.50
0.000
0E+0 7.5E–6 1.5E–5 2.3E–5 3E–5
0.004 0.25 Bioavailability (–) q* (mg/kg per day) 95th percentile = 2.3 × 10−5
0.000 0.00
30 60 90 110 140 2555
Body weight, kg Averaging time, days

Figure 4-8.  Example of risk calculations for incremental lifetime cancer risk (ILCR) of skin cancer due to ingestion of arsenic in soil. (A) Point exposure esti-
mation method for calculation of ILCR. Point estimates for arsenic exposure input parameters are used in this example to calculate the ILCR. This exposure
estimate is multiplied by the bioavailability of arsenic in soil to calculate the absorbed dose. Multiplication of the dose by the slope factor (q*) yields the lifetime
risk. (B) Probabilistic exposure methods for calculating the ILCR from arsenic ingestion. In this example, the soil concentration, ingestion rate, exposure dura-
tion and frequency, body weight, and bioavailability are modeled as distributions. Note that q* and averaging time (years) are given as single-point estimations.
This method yields a distribution of ILCR, with a 95th percentile upper confidence interval of 2.3 × 10−5. (Data from Calabrese et al. [1989], U.S. EPA [1989,
1992b, 1999], Davis et al. [1990], Israeli and Nelson [1992], Brorby and Finley [1993], and Agency for Toxic Substances and Disease Registry [2007].)

a chemical can be beneficial but high doses are toxic. These are as risk does not occur in the absence of exposure. However, it is
discussed in Chap. 2. These responses pose challenges to risk also frequently identified as the key area of uncertainty in the over-
assessors. In general, science-based mechanistic considerations of all risk determination. Here, the primary focus is on uses of expo-
low-dose response are always critical. (See approaches used by the sure information in quantitative risk assessment.
Institute of Medicine for vitamins and essential elements as these The primary goal of exposure calculations is not only to deter-
compounds consistently exhibit hormesis [Institute of Medicine, mine the type and amount of total exposure but also to find out
2008]). specifically who may be exposed and how large a dose may be
Development of BBDR models for endpoints other than can- reaching target tissues. A key step in making an exposure assess-
cer is limited; however, several approaches have been explored in ment is determining what exposure pathways are relevant for the
developmental toxicity, utilizing MOA information on cell cycle risk scenario under development. Fig. 4-9 shows an example expo-
kinetics, enzyme activity, and cytotoxicity as critical endpoints sure diagram used to illustrate possible exposure pathways from a
(Faustman et  al., 1989, 2005; Shuey et  al., 1994; Leroux et  al., hazardous chemical release. The subsequent steps entail quantita-
1996; Gohlke et al., 2002, 2004, 2005, 2007; Daston et al., 2004). tion of each pathway identified as a relevant exposure and then sum-
Approaches have been proposed that link pregnancy-specific TK marizing these pathway-specific exposures for calculation of overall
models with temporally sensitive TD models for developmental exposure. Such calculations can include an estimation of total expo-
impacts (Faustman et al., 1999). Unfortunately, there is a lack of sures for a specified population as well as calculation of exposure
specific, quantitative biological information on both kinetics and for highly exposed individuals. The EPA has published numerous
dynamics for most toxicants and endpoints (NRC, 2000; Faustman documents that provide detailed definitions and guidelines for
et al., 2005). The new NRC report titled Using 21st Century Science determining such exposures (U.S. EPA, 1992b, 2008, 2011a), with
to Improve Risk-related Evaluations (NRC, 2017) may provide new the recognition that special considerations need to be adopted when
ideas on how to identify mechanistic findings that link epidemiol- assessing childhood exposures (U.S. EPA, 2005b, 2005c, 2008).
ogy with mechanistic observations in in vitro and in vivo assays, Conceptually, calculations are designed to represent “a plau-
hence new attention to mechanistic models may be forthcoming. sible estimate” of exposure of individuals in the upper 90th to 95th
percentile of the exposure distribution. Upper bound estimations
would be “bounding calculations” designed to represent exposures
EXPOSURE ASSESSMENT at levels that exceed the exposures experienced by all individuals in
The primary objectives of exposure assessment are to determine the exposure distribution and are calculated by assuming upper lim-
source, type, magnitude, and duration of contact with the chemical(s) its for all exposure variables. A calculation for individuals exposed
of interest. This is a critical element of the risk assessment process, at concentrations near the middle of the exposure distribution is a
144 Gaseous and Particulate Effluents

Deposition Deposition Inhalation Air

Unit I

to crops to ground and submersion Chemical plant

Transpiration
General Principles of Toxicology

Crop
Ground-water Liquid
ingestion
ingestion effluents
Exposure
to deposited
Irrigation materials
Milk ingestion Shoreline exposure

Water ingestion Water immersion and

Water surface

Irrigation Uptake by
aquatic plants
Ingestion Meat

Aquatic food
ingestion

Figure 4-9.  Exposure pathway schematic. Within the risk assessment process, a critical early step is the identification of the exposure pathway(s). To fully char-
acterize exposure, the site- and chemical-specific exposure pathways must be identified. This hypothetical exposure pathway schematic illustrated the various
ways in which contaminants can move from the source through media to points of exposure. (Adapted from Sexton et al. [1992]; reproduced with permission
from Taylor & Francis Group, LLC, http://www.taylorandfrancis.com.)

central estimate. Fig. 4-8 gives example risk calculations using two factors within the LADD equation (Finley et al., 1994; Cullen and
types of exposure estimation procedures (U.S. EPA, 1989, 1992b, Frey, 1999). Such approaches can provide a reality check and can
2011a). Part A shows an example point estimation method for the be useful for generating more realistic exposure profiles reflective
calculation of arsenic (As) exposure via a soil ingestion route. In of population exposures. Good sources for U.S. exposures can be
this hypothetical scenario, As exposure is calculated using point obtained from the EPA Exposure Factors Handbook (U.S. EPA,
estimates for the upper 95th percent confidence limit of the arithme- 2011a). Fig. 4-8 (Part B) shows how this is done using an example
tic mean for each value in the lifetime average daily dose (LADD) As risk scenario with soil As concentration, ingestion rate, exposure
equation. The LADD is calculated as follows: duration, frequency, body weight, and bioavailability modeled as
distributed variables. Using Monte Carlo simulation techniques, an
Concentration of the Contact Exposure overall ILCR distribution can be generated and a 95th percentile for
× × population risk obtained.
toxicant in soil rate duration
LADD = . The FQPA of 1996 has highlighted the need for considerations
Bodyweight × Lifetime
of potential special risks for children and the need to evaluate child-
In this scenario, the average time (AT) is lifetime (25,550 specific exposures as well as cumulative exposures and risks (U.S.
days) and the contact or ingestion rate (CR) is equal to 100 mg soil EPA, 1996b). The EPA released a supplemental document with
per day. The exposure frequency and duration (EDF) is equal to the child-specific exposure factors to address the special considerations
exposure frequency (daily exposure = 365 days per year) times the necessary when evaluating childhood exposures (U.S. EPA, 2008).
exposure duration (30 years). To obtain the ILCR, the LADD must Table 4-7 shows example exposure factors for drinking water intake
be multiplied by the bioavailability to obtain an absorbed dose and for children and adults. The EPA Child-Specific Exposure Factors
q* where the upper 95% CLM slope (q*) is 1.5× the change in risk Handbook, which is online, provides useful information about such
per mg/kg body weight per day in this example. exposure distributions. The FQPA also identified the need to evalu-
Many exposures are now estimated using probability distri- ate total exposures by determining aggregate exposure measures for
butions for exposures rather than single-point estimates for the all exposures to a single substance. Cross-media exposure analyses,
 145
Table 4-7
Example Exposure Factor Handbook Information: Drinking Water Intake for Children
INTAKE (mL PER DAY), INTAKE (mL/kg PER DAY),

Chapter 4
AGE GROUP MEAN 95TH PERCENTILE MEAN 95TH PERCENTILE
Newborn (<1 month) 184 839 52 232
Infants (1 to <3 months) 227 896 48 205
Infants (3 to <6 months) 362 1056 52 159

Risk Assessment
Infants (6 to <12 months) 360 1055 41 126
Child (1 to <2 years) 271 837 23 71
Child (2 to <3 years) 317 77 23 60
Child (3 to <6 years) 380 1078 22 61
Child (6 to <11 years) 447 1235 16 43
Child (11 to <16 years) 606 1727 12 34
Child (16 to <18 years) 731 1983 11 31
Adult (18 to <21 years) 826 2540 12 35

SOURCES: Ershow and Cantor (1989) and U.S. EPA (2008, 2011a).

such as those conducted for lead and mercury, are good examples of the conclusions from the hazard assessment, the dose–response
of the value of looking at such total exposures in evaluating human assessment, and the exposure assessment. The EPA outlines the sci-
risks, and complex models are available for some compounds (see ence policy and elements for completing a risk characterization in
the EPA’s integrated exposure uptake biokinetic [IEUBK] model its Risk Characterization Handbook (U.S. EPA, 2000d). As shown
for lead in children; U.S. EPA, 2010c). Cumulative exposures and in Fig. 4-1, risk characterization considers the nature, estimated
cumulative risk refer to the total exposure and risks from a group of incidence, and reversibility of adverse effects in a given population;
compounds. In the FQPA, chemicals with similar modes of toxic- how robust the evidence is; how certain the evaluation is; if suscep-
ity are evaluated using cumulative exposure and risk estimates. For tible populations are characterized; and if there is a relevant MOA.
example, the EPA is identifying and categorizing pesticides that act For many years there has been an information sharing process
by a common MOA, and cumulative exposures to organophosphates aimed at harmonization of chemical testing regimes and clinical trial
with similar MOAs (cholinesterase inhibition) have been used as methodologies, so that data might be accepted in multiple countries.
an example of a class of pesticides for which cumulative exposure Efforts by the WHO and IPCS have included the Harmonization of
and cumulative risk estimates are needed (U.S. EPA, 1996b, 1998b; Approaches to the Assessment of Risk from Exposure to Chemicals
ILSI, 1999; Ryan, 2010). The WHO has recently published a frame- Project (WHO, 2011a). This project has the goal of globally har-
work for combined exposures (Meek et al., 2011). monizing but not standardizing the approaches to risk assessment
Additional considerations for exposure assessments include by increasing understanding and developing basic principles and
how time and duration of exposure are evaluated in risk assess- guidance on specific chemical risk assessment issues. The WHO
ments. In general, estimates for cancer risk use average exposure has worked to harmonize dose–response methodologies (DRM) and
over a lifetime, as lifetime average daily dose (see LADD example released a report providing descriptive guidance for risk assessors in
above). In a few cases, short-term exposure limits (STELs) are using DRM in hazard characterization (WHO, 2006b). Examples of
required (e.g., ethylene oxide and occupational exposure limits) and other reports include those on terminology (WHO, 2004) and expo-
characterization of brief but high levels of exposure is significant. sure assessment (WHO, 2006a) and Uncertainty and Data Quality
In these cases exposures are not averaged over the lifetime and the in Exposure Assessment (WHO, 2008).
effects of high, short-term doses are directly estimated. With devel- Uncertainty analysis is an essential component in our final risk
opmental toxicity, a single exposure can be sufficient to produce an characterization and includes factors such as variability and lack of
adverse developmental effect if exposures occur during a window of knowledge (NRC, 1994). Variability refers to true differences such
developmental susceptibility; thus, daily doses are used, rather than as is reflected in temporal, spatial, or interindividual differences. It
lifetime weighted averages (U.S. EPA, 1991; Weller et al., 1999). usually cannot be reduced with further study. Lack of knowledge can
ExpoCast, a part of the EPA’s suite of computational tools, provides also be discussed as a source of uncertainty and these factors may be
new insights on how to estimate population exposures and charac- reduced with further study. For example, if only a few environmental
terize exposure variability (Wambaugh et al., 2013). samples are taken from a contaminated site for analysis, then uncer-
tainty can be reduced by taking further samples. Formal methods
for dealing with uncertainty analysis are described in several refer-
RISK CHARACTERIZATION ences including IPCS (WHO, 2008) and Bogen et al. (2009). Using
Risk characterization is a summary of the risk assessment com- probabilistic-based approaches (PRA) for exposure assessment and
ponents and serves to outline the key findings and inform the risk assuming distributional versus dichotomous data for exposure fac-
manager in public health decisions. It is an analysis and integration tors are one way of addressing uncertainty that is commonly applied.
146 Variation in Susceptibility One of the key challenges for toxicologists doing risk assess-
ments will be the interpretation and linking of observations from
Risk assessment methodologies incorporating human variability
highly sensitive molecular and genome-based methods with the
have been slow to develop. Generally, assay results utilize means
overall process of toxicity (Eisen et al., 1998; Limbird and Taylor,
and standard deviations to measure variation, or even standard
1998; Andersen and Barton, 1999; NRC, 2000). The basic need for
errors of the mean. This ignores variability in response due to spe-
Unit I

linkage of observations was highlighted in early biomarker work.

cific differences in age, sex, health status, and genetics. Default fac-
The NRC reports on biomarkers (NRC, 1989a, 1989b, 1992) drew
tors of 10× are overutilized to describe cross- and between-species
distinctions for biomarkers of effect, exposure, and susceptibil-
differences and TK and TD data are rare. Nevertheless, EPA and
ity across a continuum of exposure, effect, and disease/toxicity.
OSHA are expected under the Clean Air Act and the Occupational
Biomarkers of early effects, such as frank clinical pathology, arise
Safety and Health Act to promulgate standards that protect the most
General Principles of Toxicology

as a function of exposure, response, and time. Early, subtle, and pos-

susceptible subgroups or individuals in the population (Omenn
sibly reversible effects can generally be distinguished from irrevers-
et al., 1990; Cullen and Frey, 1999; Faustman and Omenn, 2006;
ible disease states. Chemical-specific biomarkers have the potential
Kramer et al., 2006; Cullen et al., 2008). Sensitive or susceptible
to provide critical information, but there is an inherent complexity
populations have been defined and interpreted for the Clean Air Act
in pulling the information together. If biomarkers are considered a
and in this context “particularly sensitive citizens, such as bron-
reflection of exposure or disease state, then considerations have to
chial asthmatics and emphysematics” are to be protected by the
be made regarding the interactions between genes and the environ-
standards. This illustrates not only the importance but also the chal-
ment and the difference between a population-based assessment and
lenge to risk assessors to ensure protection for human populations.
individual assessment (Groopman and Kensler, 1999; NRC, 2017).
Ecogenetics has been defined as the study of critical genetic
Nowhere is the challenge for interpretation of early and highly
determinants that define susceptibility to environmentally influ-
sensitive response biomarkers clearer than in the complicated data
enced adverse health effects (Costa and Eaton, 2006). Ecogenetic
from gene expression arrays (toxicogenomics). Our continued
variation can affect biotransformation systems that activate and
ability to monitor changes in response in thousands of genes has
detoxify chemicals or alter the response in target tissues. With the
proven challenging for toxicologists. Emphasis on toxicity-based
completion of the Human Genome Project in 2003, the identifica-
risk assessment methods and analysis has been strongly encour-
tion of human polymorphisms has greatly expanded our potential
aged by several significant NAS reports (NRC, 2000, 2007a, 2007b,
for understanding how genetic variability can impact biological
2010), as well as development and application of pathway-based
response and susceptibility. There have been numerous activities
methods (Yu et al., 2006). Toxicogenomics projects have confirmed
initiated with the goal of understanding the linkage between genes
the repeatability and cross-platform concordance of microarray
and the environment. The following databases and initiatives are
data (Bammler et al., 2005; Guo et al., 2006; Nature Biotechnology
designed to identify polymorphisms, including single-nucleotide
Editorial, 2006; Patterson et al., 2006) and standards for data inter-
polymorphisms (SNP), insertions, deletions, amplifications, and
pretation (minimum information about a microarray experiment
alternative splicing. These include the International HapMap
[MIAME]) have been accepted (Brazma et al., 2001).
Project (HapMap, 2009), the DNA Polymorphism Discovery
Microarray analysis for risk assessment requires analyses
Resource (National Human Genome Research Institute, 2010), the
beyond basic cluster analysis (Eisen et al., 1998) to arrive at a func-
National Center for Biotechnology Information (NCBI) SNP data-
tional interpretation and linkage to conventional toxicological (api-
bases (NCBI, 2011a, 2011b), and the NIH GAIN program (National
cal) endpoints. Because of the vast number of measured responses
Human Genome Research Institute, 2010). Using the information
with gene expression arrays, pattern analysis techniques are being
from such databases, researchers can identify genetic similarities
used through pathway mapping platforms such as MAPPFinder
and differences in human beings with the goal of identifying genes
and GenMAPP (Moggs et al., 2004; Currie et al., 2005) as well
that affect health, disease, and individual responses to medications
as big data approaches (Conesa et al., 2016). The Gene Ontology
and environmental factors (Omenn, 2012). Recent observations
Consortium has developed a controlled vocabulary (ontology) for
now demonstrate that not only genetics but also epigenetics is sig-
sharing biological information across species. Database users can
nificant (Sartor et al., 2012). Databases such as ENCODE provide
then annotate their gene products with gene ontology (GO) terms,
detailed information on cross-species epigenetic signals for a range
establishing a consistent description and definition across databases
of conditions (Fingerman et al., 2010; NHGRI, 2017). SNP arrays
and studies, and this facilitates risk assessment. Through the power
are available that can easily identify and characterize approxi-
of GO terms, analysis methods are being developed to allow for
mately 1 × 106 SNPs in the human population. Arrays also exist
a quantitative time- and dose-dependent interpretation of genomic
that evaluate SNPs in pathways significant for drug metabolism
response using systems-based approaches to integrate responses
enzymes and transporters. These are in FDA-approved pathways
at the biological and cellular levels (Yu et al., 2006). An example
and have been used for evaluation of appropriate drug therapies.
of integrated analysis of multiple -omics data sets is Conceptgen
However, they are just starting to be used for environmental risk
(Sartor et al., 2010). The 2017 NRC report provides a new frame-
assessment (Guerrette et  al., 2011). Next-generation sequenc-
work for incorporating this information across various levels of
ing, where coverage of the genome can be extended to 40×, has
biological observation.
revealed rare copy number variants that can be linked to disease and
Both the EPA and FDA have formally recognized the power of
provide another level of understanding concerning human variation
genomic information as well as its potential limitations. Both agen-
(Sudmant et al., 2010). It should be noted that this genetic infor-
cies have issued interim policies guiding how to include genomic
mation on individuals can raise ethical, social, and legal concerns
information into risk assessments and they have agreed to use
regarding protections of individuals and their rights (Hsieh, 2004).
genomic information in conjunction with (but not in isolation of)
Recent legislation that protects individuals from discrimination on
standard risk assessment data (U.S. EPA, 2011e; U.S. FDA, 2011).
the basis of genetic variation helps to ensure protection and not
Extensive research efforts by the EPA Computational Toxicology
exclusion for all workers (Human Genome Program, 2008; United
program will facilitate future risk assessment uses (Kavlock et al.,
States Act, 2008).
2012; U.S. EPA, 2012c).
INFORMATION RESOURCES and chemical–disease and gene–disease relationships to illuminate 147
molecular mechanisms underlying variable susceptibility and envi-
There are numerous information resources available for risk assess- ronmentally influenced diseases. These data provide insight into
ment and a few are listed below to provide the reader with exam- complex chemical–gene/protein interaction networks (Davis et al.,

Chapter 4
ples, but not a comprehensive list of risk assessment resources and 2013). These databases can be especially useful for hazard iden-
databases. Such resources include the Toxicology Data Network tification and mechanistic information; few emphasize exposure
(NLM, 2017) from the U.S. National Library of Medicine; the information.
NTP (NTP, 2011a), including the Report on Carcinogens and
the Office of Health Assessment and Translation; the WHO IPCS
(WHO, 2010); and the IARC (WHO, 2011b). The WHO Toolkit for RISK PERCEPTION AND COMPARATIVE
Risk Assessment is especially user friendly for new risk assessors

Risk Assessment
(WHO, 2010). The WHO, through its IPCS, produces useful risk
ANALYSES OF RISK
documents entitled Concise International Chemical Assessment It is well known that individuals respond differently to informa-
Documents that cover a variety of chemical substances (WHO, tion about hazardous situations and products (Fischhoff, 1981;
2011c). The Codex Alimentarius Commission created by the Food Fischhoff et  al., 1993, 1996; Sandman, 1993; NRC, 1996; Risk
and Agriculture Organization of the UN (FAO) and the WHO pro- Commission, 1997; Institute of Medicine, 1999; Slovic, 2010).
duces guidelines and reports relevant for food safety and standards Understanding these behavioral responses at the individual, com-
(FAO and WHO, 2011). The Integrated Risk Information Service munity, and population levels is critical in stimulating construc-
(IRIS) is the EPA’s human health assessment program that evaluates tive risk communication and evaluating potential risk management
health risks for over 500 chemical substances (U.S. EPA, 2011d). options for risk assessment issues. In a classic study, students,
IRIS lists specific risk values used by the U.S. EPA. Additional League of Women Voters members, active club members, and sci-
databases are available from REACH. entific experts were asked to rank 30 activities or agents in order
There is a need for a note of caution for users of these risk- of their annual contribution to deaths (Slovic et al., 1979, 2005).
relevant databases, as with most databases there is a need to ­consider Club members ranked pesticides, spray cans, and nuclear power
multiple contributions to differences between recommended values. as safer than did other laypersons. Students ranked contraceptives
These differences can be due to variations in how and when new and food preservatives as riskier and mountain climbing as safer
scientific information is incorporated in the evaluations. Also, dif- than did others. Experts ranked electric power, surgery, swimming,
ferent regulatory drivers for national values or issues related to legal and x-rays as more risky, but nuclear power and police work as
priorities or delays in implementing draft to final recommended less risky than did laypersons. From studies like these, we now
values can occur. Such factors have contributed to differences in know that there are cultural and gender differences in perception
arsenic and dioxin risk estimates in the United States, for example, of risks. There are also group differences in perceptions of risk
versus other observational values. from chemicals among toxicologists, correlated with their employ-
Recently, new toxicogenomic databases that identify and, in ment in industry, academia, or government (Neal et  al., 1994).
some cases, provide characterization of chemicals have become avail- Recent risk perception research has emphasized the importance
able. The NCBI (2011a, 2011b) provides access to an enormous set of knowing the balance between analytical thinking and “affect.”
of biomedical and genomic information. The NCBI portfolio of com- Our immediate emotional reaction can shape our initial risk per-
prehensive data and tools can be valuable for risk assessment and the ceptions (“affect”), while our thoughtful, deliberate “analytical”
aim to integrate with toxicologically relevant endpoints and disease thinking also impacts our risk understanding but the latter is a
is laudable. It contains information on chemicals and bioassays, data slower response. Understanding this balance can help us explain
and software, DNA and RNA, protein domains and structures, genes why there is a complex relationship between perceived risk and
and expression, genetic information databases linked with decision benefits (Slovic, 2010).
outcomes, genomes and gene maps, and cross-species homology. Psychological factors such as dread, perceived uncontrollabil-
Resources covering databases and tools help in the study of SARs. ity, and involuntary exposure interact with factors that represent the
A publicly available functional genomics data repository (Gene extent to which a hazard is familiar, observable, and “essential” for
Expression Omnibus [GEO] database) supports MIAME-compliant daily living (Lowrance, 1976; Morgan, 1993). Fig. 4-10 presents a
data submission, and allows for easily accessible microarray- and grid on the parameters controllable/uncontrollable and observable/
sequence-based data, which are searchable for specific profiles of not observable for a large number of risky activities; for each of the
interest based on gene annotation or precomputed profile character- two-paired main factors, highly correlated factors are described in
istics. Also, variations such as data on genomic variation, interaction the boxes.
of genotypes and phenotypes, and the single-nucleotide polymor- Public demand for government regulations often focuses on
phisms (dbSNP) can be obtained through the NCBI. ACToR (U.S. involuntary exposures (especially in the food supply, drinking
EPA, 2017b), the EPA’s online database on chemical toxicity data water, and air) and unfamiliar hazards, such as radioactive waste,
and potential chemical risks to human health and the environment, is electromagnetic fields, asbestos insulation, and genetically modi-
another useful resource for risk assessments (Judson et al., 2012). It fied crops and foods. The public can respond negatively when they
allows users to search and query data including Toxicity Reference perceive that information about hazards or new technologies has
Database (ToxRefDB) (30 years and $2 billion worth of animal toxic- been withheld or under-rated. This can explain some of the very
ity studies), ToxCastDB (data from screening 1000 chemicals in over strong responses to genetically modified foods, HIV-contaminated
500 high-throughput assays), ExpoCastDB, and DSSTox (provides blood transfusions in the 1980s, or hazardous chemical and radio-
high-quality chemical structures and annotations). The ToxRefDB active wastes. Loss of trust is exemplified by Japanese reactions
captures thousands of in vivo animal toxicity studies on hundreds following the Fukushima Daiichi nuclear reactor meltdown (March
of chemicals. Another curated Comparative Toxicogenomics 2011), where initial perceived benefits of nuclear power were trans-
Database (Comparative Toxicogenomics Database, 2017) includes formed to distrust during follow-up actions and responses to the
data describing cross-species chemical and gene/protein interactions earthquake and tsunami.
148
Not observable
Unknown to those exposed,
effect delayed, new risk, risks
unknown to science.
Unit I

DNA technology
Microwave ovens Electric fields
Water fluoridation DES
Nitrites Nitrogen fertilizers
Saccharin
Water chlorination Radioactive waste
Polyvinyl chloride
General Principles of Toxicology

Oral contraceptives Nuclear reactor accidents

Diagnostic
Uranium mining
Valium x-rays Pesticides
IUDs Mercury
Antibiotics PCBs Nuclear weapons fallout

Asbestos Satellite crashes

Lead (autos)
Controllable Aspirin insulation Uncontrollable
Lead paint
Not dread, not global catastrophic, Vaccines Coal-burning pollution Dread, global catastrophic,
consequences not fatal, equitable, consequences fatal, not equitable,
low risk to future generations, Skateboards Carbon monoxide high risk to future generations, not
easily reduced, risk decreasing, (autos) Storage and easily reduced, risk increasing,
voluntary. Smoking (disease) transport of liquified involuntary.
Power mowers Snowmobiles Black lung
natural gas
Trampolines Large dams
Tractors Nerve gas accidents
Skyscraper fires
Chain saws
Elevators Nuclear weapons (war)
Home swimming pools Underwater construction
Downhill skiing Sport parachutes Coal mining
General aviation
Recreational boating
High construction
Motorcycles
Bicycles Railroad collisions
Alcohol-related accidents Commercial aviation
Auto accidents
Fireworks Auto racing

Dynamite Handguns

Observable
Known to those exposed, effect
immediate, old risk, risks known
to science.

Figure 4-10.  Perceptions of risk illustrated using a “risk space” axis diagram. Risk space has axes that correspond roughly to a hazard’s perceived “dreaded-
ness” and to the degree to which it is familiar or observable. Risks in the upper right quadrant of this space are most likely to provoke calls for government
regulation.

Perceptions of risk led to the addition of an extra safety factor Well-being goes beyond “disease-free” existence to freedom from
(default value 10) for children in the FQPA of 1996. Engineering- want (including food and water security) and fear (personal safety)
based “as low as reasonably achievable” (ALARA) approaches and sustainable futures. Concepts such as food security (abundance
also reflect the general “precautionary principle,” which is strongly and quality of foods), water security (plentiful supplies and high
favored by those who, justifiably, believe we are far from knowing quality of water), and sustainability form internationally recognized
all risks given frequently limited toxicity testing data (Roe et al., environmental and developmental goals. Sustainability embraces
1997). the risk management concept that “development that meets the
needs of the present, without compromising the ability of future
generations” to thrive and hence well-being is one of the goals of
EMERGING CONCEPTS environmental actions and decisions. Recognition that environmen-
In order for risk assessment to inform environmental risk man- tal problems are global is essential to our understanding of how
agement, there is a need to ensure that the initial “problem for- we manage risks and how we address sustainability (Leiserowitz
mulation” of the risk question(s) is succinctly framed to answer et al., 2006; Gohlke et al., 2008; Kite-Powell et al., 2008). Ocean
questions in the real world. Environmental health is very dynamic health and air pollution are excellent examples of the need for
and many divergent emerging environmental challenges such as cli- understanding the global context where pollutants do not honor
mate change, energy shortages, and engineered nanoparticles will country and national borders. World Trade Organization (WTO)
require an expansion of our context well beyond single-chemical, agreements have also placed risk assessment approaches in the
single-exposure scenarios. In order to accomplish this goal, several forefront as tools for managers working to meet the global goals
factors need to be considered. These factors include defining not for sustainable trade and economic health. At the most basic level
only health but also well-being and sustainability and will require a new risk assessments are being done with a new global empha-
context of global and international scale. sis (Carruth and Goldstein, 2004; EPA, 2011f). Reorganization of
Well-being is increasingly being used to describe human the EPA along management lines of sustainability now provides
health and the goal of sustainable environmental risk management. an integrative framework for working toward sustainability goals.
The WHO has championed the use of well-being rather than health As a response for this broader framing of health impacts, the EPA
to describe the Millennium Development Goals (WHO, 2011d). Office of Research and Development realigned its research within
a structure of Chemical Safety for Sustainability; Sustainable and public health concepts for sustainability, environmental disaster 149
Healthy Communities; Safe and Sustainable Water Resources; and response, and life-cycle systems analysis than many of our tradi-
Air, Climate and Energy. Risk assessment is viewed as one of the tional frameworks for environmental chemical risk assessment,
critical approaches for integrating these goals and toxicology is one which can be done in relative isolation.

Chapter 4
of the tools used within that approach (Teichman and U.S. EPA,
2011). Further applications for alternatives assessment, life cycle
analysis, and green chemistry open up new opportunities to enlarge SUMMARY
the framing of our assessments (NRC, 2014a, 2014b). The NRC and Risk Commission frameworks for risk assessment
and risk management continue to provide a consistent framework-
based approach for evaluating risks and taking action to reduce
PUBLIC HEALTH RISK MANAGEMENT

Risk Assessment
risks. The objectives of risk assessments vary with the issues, risk
Associated with concepts of well-being and sustainability is a management needs, and statutory requirements. Hence, setting
public health orientation to risk assessment and risk management. the context and problem formation for risk evaluation is essential
Public health as a discipline is very compatible with toxicology, (NRC, 2009). The frameworks are sufficiently flexible to address
where toxicological tests are performed to identify and characterize various objectives and to accommodate new knowledge while also
potential health risks and to prevent the unsafe use of such agents. providing guidance for priority setting in industry, environmental
Public health also has an emphasis on approaches for identifying, organizations, and government regulatory and public health agen-
characterizing, and preventing risks. cies. Risk assessment analyzes the science and, if incomplete,
Within public health risk management, there are three stages identifies uncertainty and provides approaches for moving forward
of prevention: primary, whose goal is prevention and risk or haz- with decisions (NRC, 2009). Toxicology, epidemiology, exposure
ard avoidance; secondary, whose goal is mitigation or prepared- assessment, and clinical observations can be linked with biomark-
ness including risk or vulnerability reduction and risk transfer; and ers, cross-species investigations of mechanisms of effects, and sys-
tertiary, where prompt response or recovery is an approach for tematic approaches to risk assessment, risk communication, and
decreasing residual risk or risk reduction (Frumkin, 2016). These risk management. New insights from the NRC “Using 21st century
three prevention stages frame many risk management actions. For science to improve risk-related evaluations” provide suggestions
example, in occupational health, a prevention hierarchy is applied on how to achieve integration (NRC, 2017). Advances in toxicol-
that is related to these stages where initial preferred risk manage- ogy are certain to improve the quality of risk assessments for a
ment actions are substitution where the hazardous chemical or broad array of health endpoints as scientific findings substitute data
activities is removed and substituted with a safer chemical or pro- for assumptions and help to describe and model uncertainty more
cess option. Engineering controls with a focus on reducing work- credibly.
place exposures is a subsequent action. Administrative controls
can be used to change job tasks and methods. The use of personal
protective equipment (PPE) such as gloves, hats, protective gloves,
and work clothes is usually considered after all other management
ACKNOWLEDGMENTS
efforts fail to meet occupationally safe levels. In a similar manner, The editor and author recognize the contributions of previous
pollution prevention can be related to these stages of prevention authors of this chapter: Morton Corn (2nd edition), Robert A. Scala
with initial elimination of the source of hazardous chemical. (4th edition), and Gilbert S. Omenn (5th to 8th editions).
The context for prevention actions can be supported by source
reduction or a life-cycle analysis or alternatives analysis. If the for-
mer, the prevention activities would consider potential for adverse REFERENCES
outcomes along the life of the product including manufacture as Abbott BD, Harris MW, Birnbaum LS. Comparisons of the effects of
well as end-of-life disposal options. Sometimes when an alterna- TCDD and hydrocortisone on growth factor expression provide insight
tives analysis is performed, only one stage of the life of a product into their interaction in the embryonic mouse palate. Teratology.
is examined and a “regrettable substitution” can occur where the 1992;45:35–53.
alternative is either less well studied or knowledge about use and Adami HO, Berry SC, Breckenridge CB, et al. Toxicology and epidemiol-
fate and transport is missing (Goldman, 2016). ogy: improving the science with a framework for combining toxicologi-
Hazard analysis frequently used in public health framework cal and epidemiological evidence to establish causal inference. Toxicol
Sci. 2011;122:223–234.
refers to both hazard identification and probability-based frequency
Adler S, Basketter D, Creton S, et  al. Alternative (non-animal) methods
of anticipated events. Capacity assessment has been used for iden- for cosmetics testing: current status and future prospects—2010. Arch
tifying strengths and resiliency of a system to impact. Recent Toxicol. 2011;85:367–485.
disasters such as Hurricane Katrina, the Gulf Oil Spill, and the Agency for Toxic Substances and Disease Registry. Toxicological Profile
Fukushima Daiichi nuclear reactor meltdown all point to the need for Arsenic. Atlanta, GA: Department of Health and Human Services;
for environmental risk assessment to be considered as a part of an 2007.
even larger context in order to understand critical infrastructure for Albert RE. Carcinogen risk assessment in the US Environmental Protection
determining human and environmental risks. Life-cycle analysis, Agency. Crit Rev Toxicol. 1994;24:75–85.
green chemistry, and built environment all require such a broad Allen BC, Kavlock RJ, Kimmel CA, Faustman EM. Dose-response assess-
context for evaluation (Keim, 2002; NRC, 2009). Understanding ments for developmental toxicity: II. Comparison of generic benchmark
dose estimates with no observed adverse effect levels. Fundam Appl
population variability and vulnerability in a context of “built envi-
Toxicol. 1994a;23:487–495.
ronment” is also important. The “built environment” encompasses Allen BC, Kavlock RJ, Kimmel CA, Faustman EM. Dose-response assess-
all man-made resources and infrastructure, including buildings, ment for developmental toxicity. 3. Statistical models. Fundam Appl
spaces, and transportation systems, which support human activity Toxicol. 1994b;23(4):496–509.
and have a strong influence on public health (Perdue et al., 2003). Ames BN, Gold LS. Too many rodent carcinogens: mitogenesis increases
These types of frameworks can allow for easier consideration of mutagenesis. Science. 1990;247:970–971.
150 Andersen ME, Barton HA. Biological regulation of receptor–hormone Cheng H, Aylward L, Beall C, et al. TCDD exposure-response analysis and
complex C in relation to dose–response assessments of endocrine-active risk assessment. Risk Anal. 2006;26(4):1059–1071.
compounds. Toxicol Sci. 1999;48:38–50. Chesler EJ, Miller DR, Branstetter LR, et  al. The Collaborative Cross at
Ankley GT, Bennett RS, Erickson RJ, et  al. Adverse outcome pathways: Oak Ridge National Laboratory: developing a powerful resource for sys-
a  conceptual framework to support ecotoxicology research and risk tems genetics. Mamm Genome. 2008;19:382–389.
assessment. Environ Toxicol Chem. 2010;29(3): 730-741. Chhabra RS, Bucher JR, Wolfe M, Portier C. Toxicity characterization of
Unit I

Armitage P, Doll R. A two-stage theory of carcinogenesis in relation to the environmental chemicals by the US National Toxicology Program: an
age distribution of human cancer. Br J Cancer. 1957;11:161–169. overview. Int J Hyg Environ Health. 2003;206:437–445.
Ashby J, Tennant RW. Prediction of rodent carcinogenicity of 44 chemicals: Cohen SM. Human relevance of animal carcinogenicity studies. Regul
results. Mutagenesis. 1994;9:7–15. Toxicol Pharmacol. 1995;21:75–80 [discussion 81–86].
Atterwill CK, Johnston H, Thomas SM. Models for the in vitro assess- Cohen SM. Alternative models for carcinogenicity testing: weight of
General Principles of Toxicology

ment of neurotoxicity in the nervous system in relation to xenobiotic evidence evaluations across models. Toxicol Pathol. 2001;29(suppl):
and neurotrophic factor-mediated events. Neurotoxicology. 1992;13: 183–190.
39–54. Cohen SM, Arnold LL, Cano M, Ito M, Garland EM, Shaw RA. Calcium
Auton TR. Calculation of benchmark doses from teratology data. Regul phosphate-containing precipitate and the carcinogenicity of sodium salts
Toxicol Pharmacol. 1994;19:152–167. in rats. Carcinogenesis. 2000;21:783–792.
Bailey GS, Reddy AP, Pereira CB, et  al. Nonlinear cancer response at Cohen SM, Ellwein LB. Proliferative and genotoxic cellular effects in
ultralow dose: a 40800-animal ED(001) tumor and biomarker study. 2-acetylaminofluorene bladder and liver carcinogenesis: biological mod-
Chem Res Toxicol. 2009;22:1264–1276. eling of the EDO1 study. Toxicol Appl Pharmacol. 1990;104:79–93.
Bammler T, Beyer RP, Bhattacharya S, et  al. Standardizing global gene Collins FS, Gray GM, Bucher JR. Toxicology. Transforming environmental
expression analysis between laboratories and across platforms. Nat health protection. Science. 2008;319(5865):906-907.
Methods. 2005;2:351–356. Comparative Toxicogenomics Database. Available at: http://ctd.mdibl.org/.
Barnes DG, Dourson MJ. Reference dose (Rfd): description and use in 2017.
health risk assessment. Regul Toxicol Pharmacol. 1988;8:471–486. Conesa A, Madrigal P, Tarazona S, et  al. A survey of best practices for
Bell SM, Phillips J, Sedykh A, et  al. An integrated chemical environ- RNA-seq data analysis. Genome Biol. 2016; 17: 13.
ment to support 21st-century toxicology. Environ Health Perspect. Costa L, Eaton D. Introduction. In: Costa L, Eaton D, eds. Gene–
2017;125(5):054501. Environment Interactions: Fundamentals of Ecogenetics. Hoboken, NJ:
Benigni R, Zito R. The second National Toxicology Program comparative Wiley; 2006.
exercise on the prediction of rodent carcinogenicity: definitive results. Costa LG. Biochemical and molecular neurotoxicology. In: Maines MD,
Mutat Res. 2004;566:49–63. Costa LG, Reed DJ, eds. Current Protocols in Toxicology. New York:
Birnbaum LS, Thayer KA, Bucher JR, Wolfe MS. Implementing system- Wiley; 2000.
atic review at the National Toxicology Program: status and next steps. Crump KS. An improved procedure for low-dose carcinogenic risk assess-
Environ Health Perspect. 2013;121(4):a108–a109. ment from animal data. J Environ Pathol Toxicol. 1980;5:339–348.
Bogdanffy MS, Daston G, Faustman EM, et al. Harmonization of cancer Crump KS. A new method for determining allowable daily intakes. Fundam
and noncancer risk assessment: proceedings of a consensus-building Appl Toxicol. 1984;4:854–871.
workshop. Toxicol Sci. 2001;61:18–31. Cullen AC, Corrales MA, Kramer CB, Faustman EM. The application of
Bogen KT, Cullen AC, Frey HC, Price PS. Probabilistic exposure analysis genetic information for regulatory standard setting under the Clean Air
for chemical risk characterization. Toxicol Sci. 2009;109:4–17. Act: a decision-analytic approach. Risk Anal. 2008;28:877–890.
Brazma A, Hingamp P, Quackenbush J, et al. Minimum information about Cullen AC, Frey HC. Probabilistic Techniques in Exposure Assessment: A
a microarray experiment (MIAME)—toward standards for microarray Handbook for Dealing with Variability and Uncertainty in Models and
data. Nat Genet. 2001;29:365–371. Inputs. New York: Plenum Press; 1999.
Brorby G, Finley G. Standard probability density functions for routine use Currie RA, Bombail V, Oliver JD, et  al. Gene ontology mapping as an
in environmental health risk assessment. In: Society for Risk Analysis unbiased method for identifying molecular pathways and processes
Annual Meeting. Savannah, GA; 1993. affected by toxicant exposure: application to acute effects caused by
Brown CC. High-to-low-dose extrapolation in animals. In: Rodricks the rodent non-genotoxic carcinogen diethylhexylphthalate. Toxicol Sci.
JV, Tardiff RG, eds. Assessment and Management of Chemical Risks. 2005;86:453–469.
Washington, DC: American Chemical Society; 1984:57–79. Da Rocha MS, Arnold LL, De Oliveira ML, et al. Diuron-induced rat urinary
Brown NA, Spielmann H, Bechter R. Screening chemicals for reproductive bladder carcinogenesis: Mode of action and human relevance evaluations
toxicity: the current alternatives: the report and recommendations of an using the International Programme on Chemical Safety framework. Crit
ECVAM/ETS workshop. Altern Lab Anim. 1995;23:868–882. Rev Toxicol. 2014;44(5):393-406.
Bucher JR. Update on National Toxicology Program (NTP) assays with Daston G, Faustman EM, Ginsberg G, et  al. A framework for assessing
genetically altered or “transgenic” mice. Environ Health Perspect. risks to children from exposure to environmental agents. Environ Health
1998;106:619–621. Perspect. 2004;112:238–256.
Calabrese EJ, Pastides H, Barnes R. How much soil do young children ingest: Davis AP, Murphy CG, Johnson R, et  al. The comparative toxicoge-
an epidemiologic study. In: Kostecki PT, Calabrese EJ, eds. Petroleum nomics database: update 2013. Nucleic Acids Res. 2013;41(database
Contaminated Soils. Chelsea, MI: Lewis Publishers; 1989:363–397. issue):D1104–D1114.
Calkins DR, Dixon RI, Gerber CR. Identification, characterization, and Davis S, Waller P, Buschborn R. Quantitative estimates of soil ingestion
control of potential human carcinogens: a framework for federal deci- in normal children between the ages of 2 and 7 years: population-based
sion-making. J Natl Cancer Inst. 1980;61:167–175. estimates using aluminum, silicon, and titanium as soil tracer elements.
Capen CC, Dybing E, Rice JM, Wilbourn JD, eds. Species Differences in Arch Environ Health. 1990;45:112–122.
Thyroid, Kidney and Urinary Bladder Carcinogenesis. IARC Scientific Dean AG, Dean JA, Coulombier D. Epi Info, Version 6: A Word Processing,
Publications No. 147. International Agency for Research on Cancer and Database, and Statistics Program for Public Health on IBM-Compatible
World Health Organization; 1999. Microcomputers. Atlanta, GA: Centers for Disease Control and
Carruth RS, Goldstein BD. The asbestos case: a comment on the appoint- Prevention (CDC); 1995.
ment and use of nonpartisan experts in World Trade Organization dispute Dearfield KL. Pesticide Assessment Guidelines. Subdivision F, Hazard
resolution involving health risk. Risk Anal. 2004;24:471–481. Evaluation: Humans and Domestic Animals. Series 83-3. Washington,
Charnley G, Omenn GS. A summary of the findings and recommendations DC: Environmental Protection Agency; 1990.
of the commission on risk assessment and risk management (and accom- Dourson MJ, Hertzberg RC, Hartung R, Blackburn K. Novel methods for
panying papers prepared for the commission). Hum Ecol Risk Assess. the estimation of acceptable daily intake. Toxicol Ind Health. 1985;1:
1997;3:701–711. 23–41.
Dourson ML, DeRosa CT. The use of uncertainty factors in establishing Finley G, Proctor D, Scott P. Recommended distributions for expo- 151
safe levels of exposure. In: Krewski D, Franklin C, eds. Statistics in sure factors frequently used in health risk assessment. Risk Anal.
Toxicology. New York: Gordon & Breach; 1991:613–627. 1994;14:533–553.
Dourson ML, Stara JF. Regulatory history and experimental support of Fischhoff B. Cost–benefit analysis: an uncertain guide to public policy. Am
uncertainty (safety factors). Regul Toxicol Pharmacol. 1983;3:224–238. N Y Acad Sci. 1981;363:173–188.

Chapter 4
Drew CH, Grace DA, Silbernagel SM, et al. Nuclear waste transportation: Fischhoff B, Bostrom A, Quandrel MJ. Risk perception and communica-
case studies of identifying stakeholder risk information needs. Environ tion. Annu Rev Public Health. 1993;14:183–203.
Health Perspect. 2003;111:263–272. Fischhoff B, Bostrom A, Quandrel MJ. Risk perception and communication.
Drew CH, Kern M, Martin T, Blozek ML, Power M, Faustman EM. The In: Detels R, Holland W, McEwen J, Omenn GS, eds. Oxford Textbook
Hanford Openness Workshops: fostering open and transparent long-term of Public Health. New York: Oxford University Press; 1996:987–1002.
decision making at the department of energy. In: Leschine TM, ed. Long-Term Frumkin H. Introduction to environmental health. In: Frumkin H, ed.

Risk Assessment
Management of Contaminated Sites (Research in Social Problems and Public Environmental Health: From Global to Local. 3rd ed. San Francisco, CA:
Policy, Volume 13). Emerald: Bingley, West Yorkshire UK; 2006;13–48. Wiley; 2016:3–25.
Eaton DL, Bammler TK, Kelly EJ. Interindividual differences in response Gamble JF, Battigelli MC. Epidemiology. In: Clayton GD, Clayton FE,
to chemoprotection against aflatoxin-induced hepatocarcinogenesis: eds. Patty’s Industrial Hygiene and Toxicology. New York: Wiley;
implications for human biotransformation enzyme polymorphisms. Adv 1978:113–127.
Exp Med Biol. 2001;500:559–576. Gamble JF, Battigelli MC. Occupational epidemiology: some guide-
Eaton DL, Gallagher EP. Mechanisms of aflatoxin carcinogenesis. Annu posts. In: Clayton GD, Clayton FE, eds. Patty’s Industrial Hygiene and
Rev Pharmacol Toxicol. 1994;34:135–172. Toxicology. New York: Wiley; 1991.
Eaton DL, Gallagher EP, Bammler TK, Kunze KL. Role of cytochrome Gargas ML, Finley BL, Paustenback DJ, Long TF. Environmental health
P4501a2 in chemical carcinogenesis—implications for human variability risk assessment: theory and practice. In: Ballantyne B, Marrs T, Syversen
in expression and enzyme-activity. Pharmacogenetics. 1995;5:259–274. T, eds. General and Applied Toxicology. New York: Grove’s Dictionaries;
ECHA (European Chemicals Agency). Registered substances database. 1999:1749–1809.
2017. Available at: https://echa.europa.eu/information-on-chemicals/ Gill EA, Curl CL, Adar SD, et al. Air pollution and cardiovascular disease
registered-substances. Accessed 23 January, 2017. in the Multi-Ethnic Study of Atherosclerosis. Prog Cardiovasc Dis.
Eduati F, Mangravite LM, Wang T, et  al. Prediction of human popula- 2011;53(5):353–360.
tion responses to toxic compounds by a collaborative competition. Nat Gohlke JM, Griffith WC, Bartell SM, Lewandowski TA, Faustman
Biotechnol. 2015;33(9):933-40. EM. A computational model for neocortical neuronogenesis predicts
EGRET. Statistics and Epidemiology Research Corporation (SERC). ethanol-induced neocortical neuron number deficits. Dev Neurosci.
Seattle, WA: Statistics and Epidemiology Research Corporation; 1994. 2002;24:467–477.
Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and dis- Gohlke JM, Griffith WC, Faustman EM. The role of cell death dur-
play of genome-wide expression patterns. Proc Natl Acad Sci U S A. ing neocortical neurogenesis and synaptogenesis: implications from a
1998;95:14863–14868. computational model for the rat and mouse. Brain Res Dev Brain Res.
Ershow AG, Cantor KP. Federation of American Societies for Experimental 2004;151:43–54.
Biology. Life Sciences Research Office, and National Cancer Institute Gohlke JM, Griffith WC, Faustman EM. A systems-based computational
(US). Division of Cancer Etiology. Epidemiology and Biostatistics model for dose–response comparisons of two mode of action hypoth-
Program. Total Water and Tapwater Intake in the United States: eses for ethanol-induced neurodevelopmental toxicity. Toxicol Sci.
Population-based Estimates of Quantities and Sources. Bethesda, MD: 2005;86:470–484.
Life Sciences Research Office Federation of American Societies for Gohlke JM, Griffith WC, Faustman EM. Computational models of neocorti-
Experimental Biology; 1989. cal neuronogenesis and programmed cell death in the developing mouse,
FAO, WHO. Codex Alimentarius Commission—International Food monkey, and human. Cereb Cortex. 2007;17:2433–2442.
Standards. 2011. Available at: http://www.codexalimentarius.org/. Gohlke JM, Hrynkow SH, Portier CJ. Health, economy, and environ-
Faustman E, Omenn G. Risk assessment and the impact of ecogenetics. In: ment: sustainable energy choices for a nation. Environ Health Perspect.
Costa L, Eaton D, eds. Gene–Environment Interactions: Fundamentals of 2008;116:A236–A237.
Ecogenetics. Hoboken, NJ: Wiley; 2006. Goldman L. Environmental public health: from theory to practice. In:
Faustman EM. Short-term test for teratogens. Mutat Res. 1988;205:355–384. Frumkin H, ed. Environmental Health: From Global to Local. 3rd ed.
Faustman EM, Allen BC, Kavlock RJ, Kimmel CA. Dose–response assess- San Francisco, CA: Wiley; 2016:725–746.
ment for developmental toxicity: I. Characterization of data base and Gray TJB. Application of in vitro systems in male reproductive toxicology.
determination of no observed adverse effect levels. Fundam Appl Toxicol. In: Lamb JC IV, Foster PMD, eds. Physiology and Toxicology of Male
1994;23:478–486. Reproduction. San Diego, CA: Academic Press; 1988:250–253.
Faustman EM, Gohlke J, Judd NL, et al. Modeling developmental processes Great Britain, Health and Safety Executive. Reducing Risks, Protecting
in animals: applications in neurodevelopmental toxicology. Environ People: Health & Safety Executive. Suffolk, UK: HSE Books; 1999.
Toxicol Pharmacol. 2005;19:615–624. Gribble EJ, Hong SW, Faustman EM. The magnitude of methylmercury-
Faustman EM, Gohlke JM, Ponce RA, et  al. Experimental approaches induced cytotoxicity and cell cycle arrest is p53-dependent. Birth Defects
to evaluate mechanisms of developmental toxicity. In: Hood R, ed. Res A Clin Mol Teratol. 2005;73:29–38.
Developmental and Reproductive Toxicology: A Practical Approach. 3rd Groopman JD, Kensler TW. The light at the end of the tunnel for chem-
ed. Boca Raton, FL: CRC Press; 2011. ical-specific biomarkers: daylight or headlight? Carcinogenesis.
Faustman EM, Lewandowski TA, Ponce RA, Bartell SM. Biologically 1999;20:1–11.
based dose–response models for developmental toxicants: lessons from Guerrette Z, Moreira EG, Griffith WC, et al. Cytochrome P450 3A5 geno-
methylmercury. Inhal Toxicol. 1999;11:101–114. type is correlated with acetylcholinesterase inhibition levels after expo-
Faustman EM, Ponce RA, Seeley MR, Whittaker SG. Experimental sure to organophosphate pesticides. Abstract 2111. The Toxicologist
approaches to evaluate mechanisms of developmental toxicity. In: Hood CD—An Official Journal of the Society of Toxicology. 2011;120(March).
R, ed. Handbook of Developmental Toxicology. Boca Raton, FL: CRC Guo L, Lobenhofer EK, Wang C, et  al. Rat toxicogenomic study reveals
Press; 1997;13–41. analytical consistency across microarray platforms. Nat Biotechnol.
Faustman EM, Wellington DG, Smith WP, Kimmel CS. Characterization of 2006;24:1162–1169.
a developmental toxicity dose response model. Environ Health Perspect. HapMap. International HapMap Project. 2009. Available at: http://snp
1989;79:229–241. .cshl.org/index.html.en.
Fingerman IM, McDaniel L, Zhang X, et al. NCBI epigenomics: a new pub- Harris MW, Chapin RE, Lockhart AC. Assessment of a short-term repro-
lic resource for exploring epigenomic data sets. Nucleic Acids Research. ductive and developmental toxicity screen. Fundam Appl Toxicol.
2010;39(suppl 1):D908–D912. 1992;19:186–196.
152 Hartung R. Dose–response relationships. In: Tardiff RG, Rodricks JV, eds. Kavlock RJ, Allen BC, Faustman EM, Kimmel CA. Dose response assess-
Toxic Substances and Human Risk: Principles of Data Interpretation. ments for developmental toxicity: IV. Benchmark doses for fetal weight
New York: Plenum Press; 1987:29–46. changes. Fundam Appl Toxicol. 1995;26:211–222.
Hartung T, Blaauboer BJ, Bosgra S, et  al. An expert consortium review Kavlock R, Chandler K, Houck K, et al. Update on EPA’s ToxCast program:
of the EC-commissioned report “alternative (non-animal) methods for providing high throughput decision support tools for chemical risk man-
cosmetics testing: current status and future prospects—2010”. ALTEX. agement. Chem Res Toxicol. 2012;25(7):1287–302.
Unit I

2011;28:183–209. Keim M. Intentional chemical disasters. In: Hogan D, Burstein J, eds.

Hartung T, Bremer S, Casati S, et  al. ECVAM’s response to the chang- Disaster Medicine. Philadelphia, PA: Lippincott Williams & Wilkins;
ing political environment for alternatives: consequences of the 2002:340–348.
European Union chemicals and cosmetics policies. Altern Lab Anim. Kensler TW, Roebuck BD, Wogan GN, Groopman JD. Aflatoxin: a 50-year
2003;31:473–481. odyssey of mechanistic and translational toxicology. Toxicol Sci.
General Principles of Toxicology

Haseman JK, Lockhart AM. Correlations between chemically related 2011;120(suppl 1):S28–S48.
site-specific carcinogenic effects in long-term studies in rats and mice. Khoury MJ. Human genome epidemiology: translating advances in human
Environ Health Perspect. 1993;101:50–54. genetics into population-based data for medicine and public health.
Haws LC, Su SH, Harris M, et  al. Development of a refined database of Genet Med. 1999;1:71–73.
mammalian relative potency estimates for dioxin-like compounds. Kite-Powell HL, Fleming LE, Backer LC, et al. Linking the oceans to pub-
Toxicol Sci. 2006;89:4–30. lic health: current efforts and future directions. Environ Health. 2008;
Healey GF. Power calculations in toxicology. Altern Lab Anim. 7:S6.
1987;15:132–139. Klaunig JE, Babich M, Baetcke K, et  al. PPARα agonist-induced rodent
Hill AB. The environment and disease: association or causation. Proc R Soc tumors: modes of action and human relevance. Crit Rev Toxicol.
Med. 1965;58:295–300. 2003;33:655–780.
Hsieh A. A nation’s genes for a cure to cancer: evolving ethical, social and Koturbash I, Scherhag A, Sorrentino J, et  al. Epigenetic mechanisms of
legal issues regarding population genetic databases. Columbia J Law Soc mouse interstrain variability in genotoxicity of the environmental toxi-
Probl. 2004;37:359–411. cant 1,3-butadiene. Toxicol Sci. 2011;122:448–456.
Human Genome Program. Breaking News: GINA Becomes Law May 2008. Kramer CB, Cullen AC, Faustman EM. Policy implications of genetic infor-
Genetics Privacy and Legislation. 2008. Available at: http://www.ornl.gov/ mation on regulation under the Clean Air Act: the case of particulate mat-
sci/techresources/Human_Genome/elsi/legislat.shtml. Accessed 2011. ter and asthmatics. Environ Health Perspect. 2006;114:313–319.
IARC. IARC Monographs on the Evaluation of Carcinogenic Risks to Krewski D, Van Ryzin J. Dose-response models for quantal response toxic-
Humans. Lyon, France: World Health Organization; 2000. ity data. In: Csorgo M, Dawson D, Rao J, Seleh A, eds. Statistics and
IARC. Monographs on the Evaluation of Carcinogenic Risks to Humans. Related Topics. Amsterdam: North-Holland; 1981:201–229.
Lyon, France: International Agency for Research on Cancer; 2017. Lave LB, Ennever F, Rosenkranz HS, Omenn GS. Information value of the
ICCVAM. The Murine Local Lymph Node Assay: A Test Method for rodent bioassay. Nature. 1988;336:631–633.
Assessing the Allergic Contact Dermatitis Potential of Chemicals/ Lave LB, Omenn GS. Cost-effectiveness of short-term tests for carcinoge-
Compounds: The Results of an Independent Peer Review Evaluation. nicity. Nature. 1986;334:29–34.
Research Triangle Park, NC: National Institute of Environmental Health Leiserowitz AA, Kates RW, Parris TM. Sustainability values, attitudes, and
Sciences; 1999. behaviors: a review of multinational and global trends. Annu Rev Environ
ICCVAM. ICCVAM Test Method Evaluation Report on using the LLNA for Resour. 2006;31:413–444.
Testing Pesticide Formulations, Metals, Substances in Aqueous Solutions, Leroux BG, Leisenring WM, Moolgavkar SH, Faustman EM. A bio-
and Other Products. Research Triangle Park, NC: National Institute of logically based dose–response model for development. Risk Anal.
Environmental Health Sciences; 2011a. 1996;16:449–458.
ICCVAM. Independent Scientific Peer Review Panel Report Evaluation of Lewandowski TA, Ponce RA, Whittaker SG, Faustman EM. In vitro models
the LUMI-CELL®ER (BG1Luc ER TA) Test Method. Research Triangle for evaluating developmental toxicity. In: Gad S, ed. Vitro Toxicology.
Park, NC: National Institute of Environmental Health Sciences; 2011b. New York: Taylor & Francis; 2000:139–187.
ICCVAM. Validation of the BG1Luc Estrogen Receptor Transcriptional Limbird LE, Taylor P. Endocrine disruptors signal the need for receptor
Activation Test Method. Research Triangle Park, NC: National Institute models and mechanisms to inform policy. Cell. 1998;93:157–163.
of Environmental Health Sciences. 2012. Available at: http://iccvam Liu R, Rallo R, George S, et al. Classification NanoSAR development for
.niehs.nih.gov/methods/endocrine/end_eval.htm. cytotoxicity of metal oxide nanoparticles. Small. 2011;7:1118–1126.
ICCVAM. Available at: https://ntp.niehs.nih.gov/iccvam/docs/about_docs/ Lowrance WW. Of Acceptable Risk: Science and the Determination of
icatm-moc-mar11.pdf. Safety. Los Altos, CA: William Kaufmann; 1976.
ILSI. A Framework for Cumulative Risk Assessment. Washington, DC: Maynard AD, Aitken RJ, Butz T, et al. Safe handling of nanotechnology.
International Life Science Institute; 1999. Nature. 2006;444:267–269.
Institute of Medicine. Toward Environmental Justice. Research, Education, McClain RM. Mechanistic considerations in the regulation and classifica-
and Health Policy Needs. Washington, DC: National Academies Press; tion of chemical carcinogens. In: Kotsonis F, Mackey M, Hjelle J, eds.
1999. Nutritional Toxicology. New York: Raven Press; 1994:278–304.
Institute of Medicine. Development of DRIs 1994–2004 Lessons Learned McGregor DB, Rice JM, Venitt S. The Use of Short- and Medium-Term
and New Challenges: Workshop Summary. Washington, DC: National Tests for Carcinogens and Data on Genetic Effects in Carcinogenic
Academies Press; 2008. Hazard Evaluation. Lyon, France: IARC; 1999.
Israeli M, Nelson CB. Distribution and expected time of residence for Meek ME, Boobis AR, Crofton KM, et al. Risk assessment of combined
United States households. Risk Anal. 1992;12:65–72. exposure to multiple chemicals: a WHO/IPCS framework. Regul Toxicol
Johnson DE, Wolfgang GHI, Gledin MA, Braeckman RA. Toxicokinetics Pharmacol. 2011;60(2):S1–S14.
and toxicodynamics. In: Sipes I, McQueen C, Gandolfi A, eds. Mendoza MA, Ponce RA, Ou YC, Faustman EM. p21(WAF1/CIP1) inhibits
Comprehensive Toxicology. Oxford, UK: Pergamon; 1997:169–181. cell cycle progression but not G2/M-phase transition following methyl-
Judd NL, Drew CH, Acharya C, et al. Framing scientific analyses for risk mercury exposure. Toxicol Appl Pharmacol. 2002;178:117–125.
management of environmental hazards by communities: case studies with Moggs JG, Tinwell H, Spurway T, et  al. Phenotypic anchoring of gene
seafood safety issues. Environ Health Perspect. 2005;113:1502–1508. expression changes during estrogen-induced uterine growth. Environ
Judson RS, Martin MT, Egeghy P, et al. Aggregating data for computational Health Perspect. 2004;112:1589–1606.
toxicology applications: the US Environmental Protection Agency (EPA) Moolgavkar SH, Luebeck G. Two-event model for carcinogenesis:
aggregated computational toxicology resource (ACToR) system. Int J biological, mathematical, and statistical considerations. Risk Anal.
Mol Sci. 2012;13(2):1805–1831. 1990;10:323–341.
Moore JA, Daston GP, Faustman EM. An evaluative process for assess- NRC. Toxicity Pathway-Based Risk Assessment Preparing for Paradigm 153
ing human reproductive and developmental toxicity of agents. Reprod Change: A Symposium Summary. Washington, DC: National Academies
Toxicol. 1995;9:61–95. Press; 2010.
Morgan GM. Risk analysis and management. Sci Am. 1993;269:32–41. NRC. Sustainability Concepts in Decision-making: Tools and Approaches
National Academy of Science. Health Risks from Dioxin and Related for the US Environmental Protection Agency. Washington, DC: National

Chapter 4
Compounds: Evaluation of the EPA Reassessment. Washington, DC: Academies Press; 2014a.
National Academies Press; 2006. NRC. A Framework to Guide Selection of Chemical Alternatives.
National Human Genome Research Institute. A DNA Polymorphism Discovery Washington, DC: National Academies Press; 2014b.
Resource. 2010. Available at: http://www.genome.gov/10001552. NRC. Using 21st Century Science to Improve Risk-related Evaluations.
Nature Biotechnology Editorial. Making the most of microarrays. Nat Washington, DC: National Academies Press; 2017.
Biotechnol. 2006;24:1039–1039. NTP. Available at: http://ntp.niehs.nih.gov/. 2011a.

Risk Assessment
NCBI. dbSNP Short Genetic Variations. Available at: http://www.ncbi.nlm. NTP, Office of Health Assessment and Translation (OHAT, Formerly
nih.gov/SNP/. CERHR). Available at: http://ntp.niehs.nih.gov/?objectid=497C419D-
NCBI. Available at: http://www.ncbi.nlm.nih.gov/. 2011b. E834-6B35-8AF15D389859AF07. 2011b.
NCI. Available at: http://dtp.nci.nih.gov. 2017. NTP. 14th Report on Carcinogens; 2016.
Neal N, Malmfors T, Slovic P. Intuitive toxicology: expert and lay judg- NTP, Office of Health Assessment and Translation. OHAT Systematic
ments of chemical risks. Toxicol Pathol. 1994;22:198–201. Review. 2017a. Available at: https://ntp.niehs.nih.gov/pubhealth/hat/
Nebert DW, Duffy JJ. How knockout mouse lines will be used to study noms/index-2.html.
the role of drug-metabolizing enzymes and their receptors during repro- NTP, Guidance document for the developmental neurotoxicity testing arm
duction, development, and environmental toxicity, cancer and oxidative of the modified one-generation study. 2017b. Available at https://ntp
stress. Biochem Pharmacol. 1997;53:249–254. .niehs.nih.gov/ntp/test_info/mog_guidance_508.pdf
Nel AE, Nasser E, Godwin H, et al. A multi-stakeholder perspective on the Oberdorster G. Lung particle overload: implications for occupational expo-
use of alternative test strategies for nanomaterial safety assessment. ACS sure to particles. Regul Toxicol Pharmacol. 1995;12:123–135.
Nano. 2013;7(8):6422–6433. OECD (Organisation for Economic Co-Operation and Development). Manual
Neumann DA, Olin SS. Urinary bladder carcinogenesis: a working group for Investigation of HPV Chemicals; 2013. Available at: www.oecd.org.
approach to risk assessment. Food Chem Toxicol. 1995;33:701–704. OECD. Adverse Outcome Pathways. Molecular Screening and Toxicoge-
NHGRI. The ENCODE project: encyclopedia of DNA elements. 2017. nomics; 2017. Available at: http://www.oecd.org/chemicalsafety/testing/
Available at: https://www.encodeproject.org. adverse-outcome-pathways-molecular-screening-and-toxicogenomics.htm.
NIEHS. Corrositex: An In Vitro Test Method for Assessing Dermal Corrosivity Omenn GS. The genomic era: a crucial role for the public health sciences.
Potential of Chemicals. Washington, DC: ICCVAM; 1999a:33109–33111. Environ Health Perspect. 2000;108:A204–A205.
NIEHS. The Murine Local Lymph Node Assay: A Test Method for Assessing Omenn GS. Gene–environment interactions: eco-genetics and toxicoge-
the Allergic Contact Dermatitis Potential of Chemicals/Compounds. nomics. In: Ginsburg GS, Willard HF, eds. Genomic and Personalized
Washington, DC: ICCVAM; 1999b:14006–14007. Medicine. 2nd ed. Elsevier: New York; 2012.
NIEHS. ICCVAM Test Method Evaluation Report: In Vitro Cytotoxicity Omenn GS, Lampen A. Scientific and cost-effectiveness criteria in selecting
Test Methods for Estimating Starting Doses for Acute Oral Systemic batteries of short-term tests. Mutat Res. 1988;205:41–49.
Toxicity Tests. National Toxicology Program (NTP) Interagency Center Omenn GS, Omiecinski CJ, Eaton DE. Eco-genetics of chemical carcino-
for the Evaluation of Alternative Toxicological Methods (NICEATM), gens. In: Cantor C, Caskey C, Hood L, eds. Biotechnology and Human
Interagency Coordinating Committee on the Validation of Alternative Genetic Predisposition to Disease. New York: Wiley-Liss; 1990:81–93.
Methods (ICCVAM). Bethesda, MD: National Institutes of Health; 2006. Omenn GS, Stuebbe S, Lave LB. Predictions of rodent carcinogenicity test-
NLM. Available at: http://toxnet.nlm.nih.gov/. 2017. ing results: interpretation in light of the Lave–Omenn value-of-informa-
NRC. Risk Assessment in the Federal Government: Managing the Process. tion model. Mol Carcinogen. 1995;14:37–45.
Washington, DC: National Academies Press; 1983. Patlewicz G, Roberts DW, Aptula A, Blackburn K, Hubesch B. Workshop:
NRC. Improving Risk Communication. Washington, DC: National use of “read-across” for chemical safety assessment under REACH.
Academies Press; 1989a. Regul Toxicol Pharmacol. 2013;65(2):226–228.
NRC. Biological Markers in Reproductive Toxicology. Washington, DC: Patterson TA, Lobenhofer EK, Fulmer-Smentek SB, et  al. Performance
National Academies Press; 1989b. comparison of one-color and two-color platforms within the MicroArray
NRC. Biological Markers in Pulmonary Toxicology. Washington, DC: Quality Control (MAQC) project. Nat Biotechnol. 2006;24:1140–1150.
National Academies Press; 1989c. Perdue WC, Stone LA, Gostin LO. The built environment and its relation-
NRC. Biological Markers in Immunotoxicology. Washington, DC: National ship to the public’s health: the legal framework. Am J Public Health.
Academies Press; 1992. 2003;93:1390–1394.
NRC Committee on Risk Assessment Methodology. Issues in Risk Perera FP, Weinstein IB. Molecular epidemiology: recent advances and
Assessment: Use of Maximum Tolerated Dose in Animal Bioassays for future directions. Carcinogenesis. 2000;21:517–524.
Carcinogenicity. Washington, DC: National Academies Press; 1993. REACH (Registration, Evaluation, Authorisation and Restriction of
NRC. Science and Judgement in Risk Assessment. Washington, DC: Chemicals). 2011. Available at: http://ec.europa.eu/environment/
National Academies Press; 1994 chemicals/reach/reach_en.htm.
NRC. Understanding Risk: Informing Decisions in a Democratic Society. Regalado J, Perez-Padilla R, Sansores R, et al. The effect of biomass burn-
Stern PC, Fineberg HV, eds. Washington, DC: National Academies Press; ing on respiratory symptoms and lung function in rural Mexican women.
1996. Am J Respir Crit Care Med. 2006;174(8):901–905.
NRC. Scientific Frontiers in Developmental Toxicology and Risk Renwick AG. Safety factors and the establishment of acceptable daily
Assessment. Washington, DC: National Research Council; 2000. intakes. Food Addit Contam. 1991;8:135–150.
NRC. Effective Risk Communication, The Nuclear Regulatory Commission’s Renwick AG. Toxicokinetics. In: Ballantyne B, Marrs T, Syversen T,
Guidelines for External Risk Communication. Washington, DC: National eds. General and Applied Toxicology. New York: Grove’s Dictionaries;
Academies Press, 2004. 1999:67–95.
NRC. Toxicity Testing in the 21st Century: A Vision and a Strategy. Renwick AG, Lazarus NR. Human variability and noncancer risk
Washington, DC: National Academies Press; 2007a. ­assessment—an analysis of the default uncertainty factor. Regul Toxicol
NRC. Applications of Toxicogenomic Technologies to Predictive Toxicology Pharmacol. 1998;27:3–20.
and Risk. Washington, DC: National Academies Press; 2007b. Rice JM, Baan RA, Blettner M. Rodent tumors of urinary bladder, renal
NRC. Science and Decisions: Advancing Risk Assessment. Washington, cortex, and thyroid gland. IARC monographs: evaluations of carcino-
DC: National Academies Press; 2009. genic risk to humans. Toxicol Sci. 1999;49:166–171.
154 Risk Commission. The Presidential/Congressional Commission on Risk Teichman K, U.S. EPA. Presentation to the Science Advisory Board and
Assessment and Risk Management. Washington, DC: Government Board of Scientific Counselors (Overview). 2011, June. Available at:
Printing Office; 1997. http://my.brainshark.com/Dr-Kevin-Teichman-SAB-BOSC-Presentation-
Rodericks JV, Rudenko I, Starr TB, Turnbull D. Risk assessment. In: Sipes 720933584.
I, McQueen C, Gandolfi A, eds. Comprehensive Toxicology. Oxford, UK: Tennant RW, Stasiewicz S, Mennear J. Genetically altered mouse models
Pergamon; 1997. for identifying carcinogens. In: McGregor DB, Rice JM, Venitt S, eds.
Unit I

Roe D, Pease W, Florini K, Silbergeld E. Toxic Ignorance: The Continuing The Use of Short- and Medium-Term Tests for Carcinogens and Data
Absence of Basic Health Testing for Top-Selling Chemicals in the United on Genetic Effects in Carcinogenic Hazard Evaluation. Vol. 146. Lyon,
States. Washington, DC: Environmental Defense Fund; 1997. France: IARC; 1999:123–150.
Rooney AA, Boyles AL, Wolfe MS, Bucher JR, Thayer KA. Systematic Toscano WA, Oehlke KP. Systems biology: new approaches to old environ-
review and evidence integration for literature-based environmen- mental health problems. Int J Environ Res Public Health. 2005;2:4–9.
General Principles of Toxicology

tal health science assessments. Environmental Health Perspectives. United States Act. The Genetic Information Nondiscrimination Act of 2008
2014;122(7):711–718. (GINA). Public Law 110-233; May 21, 2008.
Ryan PB. Exposure assessment, industrial hygiene, and environmental U.S. EPA. Guidelines for mutagenicity risk assessment. Fed Reg.
management. In: Frumkin H, ed. Environmental Health: From Global to 1986;51:34006–34012.
Local. 2nd ed. San Francisco, CA: Jossey-Bass; 2010. U.S. EPA. Proposed guidelines for assessing female reproductive risk. Fed
Sandman PM. Responding to Community Outrage: Strategies for Effective Reg. 1988;53:24834–24847.
Risk Communication. Fairfax, VA: American Industrial Hygiene U.S. EPA. Risk Assessment Guidance for Superfund. Vol 1: Human Health
Association; 1993. Evaluation Manual, Part A. Washington, DC: Office of Policy Analysis
Sartor MA, Dolinoy DC, Rozek LS, Omenn GS. Bioinformatics for high- and Office of Emergency and Remedial Response; 1989.
throughput toxico-epigenomics studies. In: Sahu SC, ed. Toxicology and U.S. EPA. Guidelines for developmental toxicity risk assessment. Fed Reg.
Epigenetics. New York: Wiley; 2012. 1991;56:63798–63826.
Sartor MA, Mahavisno V, Keshamouni VG, et  al. ConceptGen: a gene U.S. EPA. Frameworks for Ecological Risk Assessment. EPA\630\R-92\001;
set enrichment and gene set relation mapping tool. Bioinformatics. 1992a.
2010;26:456–463. U.S. EPA. Guidelines for exposure assessment. Fed Reg. 1992b;57:
Schardein JL, Scialli AR. The legislation of toxicologic safety factors: the 22888–22938.
Food Quality Protection Act with chlorpyrifos as a test case. Reprod U.S. EPA. Health Assessment Document for 2,3,7,8-Tetrachlorodibenzo-
Toxicol Rev. 1999;13:1–14. p-Dioxin (TCDD) and Related Compounds. Washington, DC: Office of
Sette WF. Pesticide Assessment Guidelines, Subdivision D, Hazard Research and Development; 1994a.
Evaluation, Human and Domestic Animals, Addendum 10. Neurotoxicity U.S. EPA. Guidelines for Carcinogen Risk Assessment (Draft Revisions).
Series 81, 82, and 83 PB91-154617. Washington, DC: National Technical Washington, DC: Office of Health and Environmental Assessment;
Information Services; 1991. 1994b.
Sexton K, Selevan SG, Wagener DK, Lybarger JA. Estimating human expo- U.S. EPA. Guidelines for reproductive toxicity risk assessment. Fed Reg.
sures to environmental pollutants: availability and utility of existing data- 1996a;61:56278–56322.
bases. Arch Environ Health. 1992;47:398–407. U.S. EPA. Food Quality Protection Act (FQPA). Washington, DC: Office of
Shelby MD, Bishop JB, Mason JM, Tindall KR. Fertility, reproduction and Pesticide Programs; 1996b.
genetic disease: studies on the mutagenic effects of environmental agents U.S. EPA. Chemical Hazard Availability Study. Washington, DC: Office of
on mammalian germ cells. Environ Health Perspect. 1993;200:283–291. Pollution Prevention and Toxics; 1998a.
Shuey DL, Lau C, Logsdon TR. Biologically based dose–response model- U.S. EPA. Guidance for identifying pesticides that have a common mecha-
ing in developmental toxicology: biochemical and cellular sequelae of nism of toxicity: notice of availability and solicitation of public com-
5-fluorouracil exposure in the developing rat. Toxicol Appl Pharmacol. ments. Fed Reg. 1998b; 63:42031–42032.
1994;126:129–144. U.S. EPA. Guidelines for Neurotoxicity Risk Assessment. EPA Technical
Silverman KC, Naumann BD, Holder DJ. Establishing data-derived adjust- Report. Risk Assessment Forum. Washington, DC: U.S. Environmental
ment factors from published pharmaceutical clinical trial data. Hum Ecol Protection Agency; 1998c.
Risk Assess. 1999;5:1059–1089. U.S. EPA. Guidelines for Carcinogen Risk Assessment, Draft.
Simon R, Wang SJ. Use of genomic signatures in therapeutics development NCEA-F-0644. Washington, DC: EPA; 1999.
in oncology and other diseases. Pharmacogenomics J. 2006;6:166–173. U.S. EPA. Benchmark Dose Technical Guidance Document. External
Slovic P. The Feeling of Risk: New Perspectives on Risk Perception. The Review Draft. Washington, DC: Risk Assessment Forum; 2000a.
Earthscan Risk in Society Series. Washington, DC: Earthscan, in associa- U.S. EPA. Chlorpyrifos: Re-evaluation Report of the FQPA Safety Factor
tion with the International Institute for Environment and Development; Committee. HED Doc. No. 014077. Washington, DC: U.S. Environmental
2010. Protection Agency; 2000b.
Slovic P, Baruch F, Lichtenstein S. Rating the risks. Environment. U.S. EPA. Risk Characterization Handbook. Washington, DC: U.S.
1979;21:1–20, 36–39. Environmental Protection Agency Science Policy Council; 2000c.
Slovic P, Peters E, Finucane ML, MacGregor DG. Affect, risk, and decision U.S. EPA. Generic Ecological Assessment Endpoints (GEAE) for
making. Health Psychol. 2005;24:S35–S40. Ecological Risk Assessment. Washington, DC: U.S. Environmental
Smith MN, Grice J, Cullen, Faustman EM. A toxicological framework for Protection Agency; 2004.
the prioritization of children’s safe product act data. Int J Environ Res U.S. EPA. Guidelines for Carcinogen Risk Assessment (Cancer Guidelines).
Public Health. 2016a;13(4):431. Federal Register Notice. Washington, DC: U.S. Environmental Protection
Smith MT, Guyton KZ, Gibbons CF, et al. Key characteristics of carcino- Agency; 2005a.
gens as a basis for organizing data on mechanisms of carcinogenesis. U.S. EPA. Policy on Evaluating Heath Risks to Children. Washington, DC:
Environ Health Perspect. 2016b;124(6):713–721. U.S. Environmental Protection Agency; 2005b.
Spielmann H, Seiler A, Bremer S, et al. The practical application of three U.S. EPA. Supplemental Guidance for Assessing Susceptibility from Early-
validated in vitro embryotoxicity tests—the report and recommendations Life Exposure to Carcinogens. Federal Register Notice. EPA/630/R-
of an ECVAM/ZEBET workshop (ECVAM workshop 57). Altern Lab 03/003F. Washington, DC: U.S. Environmental Protection Agency;
Anim. 2006;34:527–538. 2005c.
Stevens JT. Risk assessment of pesticides. In: Williams P, Hottendorf G, U.S. EPA. Child-Specific Exposure Factors Handbook (Final Report).
eds. Comprehensive Toxicology. Oxford, UK: Pergamon; 1997:17–26. EPA/600/R-06/096F. Washington, DC: Office of Research and
Sudmant PH, Kitzman JO, Antonacci F, et  al. Diversity of human copy Development and the National Center for Environmental Assessment;
number variation and multicopy genes. Science. 2010;330:641–646. 2008.
U.S. EPA. The Office of Chemical Safety and Pollution Prevention (OCSPP) Vlaanderen J, Portengen L, Rappaport SM, et al. The impact of saturable 155
Harmonized Test Guidelines: Series 870—Health Effects Test Guidelines. metabolism on exposure–response relations in 2 studies of benzene-
870 Series Final Guidelines. 2010a. Available at: http://www.epa.gov/ induced leukemia. Am J Epidemiol. 2011;174:621–629.
ocspp/pubs/frs/publications/Test_Guidelines/series870.htm. Wambaugh JF, Setzer RW, Reif DM, et  al. High-throughput models for
U.S. EPA. EPA’s Reanalysis of Key Issues Related to Dioxin Toxicity and exposure-based chemical prioritization in the ExpoCast project. Environ

Chapter 4
Response to NAS Comments Draft. 2010b. NAS comments are published Sci Technol. 2013; 47(15):8579–8588.
by the National Research Council of the National Academies and avail- Weller E, Long N, Smith A. Dose–rate effects of ethylene oxide exposure
able from the National Technical Information Service, Springfield, VA, on developmental toxicity. Toxicol Sci. 1999;50:259–270.
and available at: http://www.epa.gov/ncea. Whittaker SG, Faustman EM. In Vitro Assays for Developmental Toxicity.
U.S. EPA. Integrated Exposure Uptake Biokinetic Model for Lead in New York: Raven Press; 1994.
Children. Windows® version (IEUBKwin v1.1 build 11). Washington, WHO. Principles in Governing Consumer Safety in Relation to Pesticide

Risk Assessment
DC: U.S. Environmental Protection Agency; 2010c. Available from the Residues: Report of a Meeting of a W.H.O. Expert Committee on
National Technical Information Service. Available at: http://www.epa Pesticide Residues, held jointly with the F.A.O. Panel of Experts on the
.gov/superfund/lead/products.htm. Use of Pesticides in Agriculture. World Health Organization Technical
U.S. EPA. IRIS Toxicological Review of Inorganic Arsenic (Cancer) (2010 Report Series No. 240. Geneva: WHO; 1962.
External Review Draft). EPA/635/R-10/001. Washington, DC: U.S. WHO. Principles for the Assessment of Risks to Human Health from
Environmental Protection Agency; 2010d. Exposure to Chemicals. Environmental Health Criteria. World Health
U.S. EPA, Office of Research and Development. Presentations for the Organization, International Programme on Chemical Safety. Geneva:
Science Advisory Board and Board of Scientific Counselors. In: World Health Organization; 1999.
Preparation for the June 29–30, 2011 Joint Meeting; 2001f. Available at: WHO. Joint IPCS/OECD Project on the Harmonization of Chemical
https://yosemite.epa.gov/sab/sabproduct.nsf/F53952681221F76F852578 Hazard/Risk Assessment Terminology (Harmonization Project). Geneva:
AF007EE972/$File/SABBOSC+Presentations.pdf. World Health Organization; 2000.
U.S. EPA. Exposure Factors Handbook: 2011 Edition. Final Report. WHO. IPCS Risk Assessment Terminology. Harmonization Project
Washington, DC: Office of Health and Environmental Assessment; Document. Geneva: World Health Organization, International Program
2011a. on Chemical Safety; 2004.
U.S. EPA. High Production Volume (HPV) Challenge. 2011b. Available at: WHO. Chemical-Specific Adjustment Factors for Interspecies Differences
http://www.epa.gov/hpv/. and Human Variability: Guidance Document for Use of Data in
U.S. EPA. PMN Forms and Information. New Chemicals. 2011c. Available Dose/Concentration–Response Assessment. Geneva: World Health
at: http://www.epa.gov/oppt/newchems/pubs/pmnforms.htm. Organization; 2005.
U.S. EPA. Integrated Risk Information System (IRIS). 2011d. Available at: WHO. Principles of Characterizing and Applying Human Exposure
http://www.epa.gov/IRIS/. Models. IPCS Harmonization Project Document No. 3. Geneva: World
U.S. EPA. Interim Genomics Policy. 2011e. Available at: http://www.epa Health Organization; 2006a.
.gov/spc/genomics.htm. WHO. Principles for Modeling Dose-response for the Risk Assessment
U.S. EPA. Benchmark Dose Technical Guidance. EPA/100/R-12/001. 2012a. of Chemicals. Environmental Health Criteria Series, No. 234. Geneva:
Available at: https://www.epa.gov/risk/benchmark-dose-technical-guidance. World Health Organization, International Program on Chemical Safety;
U.S. EPA. Integrated Risk Information System (IRIS): Glossary of IRIS 2006b.
Terms. 2012b. Available at: http://www.epa.gov/iris/help_gloss.htm. WHO. Uncertainty and Data Quality in Exposure Assessment. Geneva:
U.S. EPA. Computational Toxicology Research. 2012c. Available at: World Health Organization, International Program on Chemical Safety;
http://www.epa.gov/ncct/. 2008.
U.S. EPA. Use of High-Throughput Assays and Computational Tools. WHO. WHO Human Health Risk Assessment Toolkit: Chemical Hazards.
Endocrine Disruptor Screening Program. Notice of Availability and IPCS Harmonization Project Document No. 8. International Programme
Opportunity for Comment; 2015. Available at: https://www.federalregister. on Chemical Safety, Inter-Organization Programme for the Sound
gov/documents/2015/06/19/2015-15182/use-of-high-throughput-assays- Management of Chemicals. Geneva: World Health Organization; 2010.
and-computational-tools-endocrine-disruptor-screening-program-notice. WHO. IPCS Harmonization Project. Programmes and Projects. 2011a.
U.S. EPA. TRI Chemical List Changes; 2016. Available at: http://www.who.int/ipcs/methods/harmonization/en/.
U.S. EPA. Benchmark Modelling Software, Version 2.7. 2017a. Available WHO. The International Agency for Research on Cancer (IARC). 2011b.
at: https://www.epa.gov/bmds/what-benchmark-dose-software-bmds. Available at: http://www.iarc.fr/.
U.S. EPA. ACToR. Washington, DC: Office of Health and Environmental WHO. Concise International Chemical Assessment Documents.
Assessment; 2017b. Available at: https://actor.epa.gov/actor/home.xhtml. International Programme on Chemical Safety (IPCS); 2011c. Available
U.S. FDA. Critical Path Initiative. Science & Research. Silver Spring, MD: at: http://www.who.int/ipcs/publications/cicad/en/.
U.S. Food and Drug Administration; 2011. Available at: http://www.fda WHO. Millennium Development Goals (MDGs). Health Topics. 2011d.
.gov/ScienceResearch/SpecialTopics/CriticalPathInitiative/default.htm. Available at: http://www.who.int/topics/millennium_development_goals/en/.
U.S. FDA. Pharmacogenetics. 2017a. Available at: https://www.fds.gov/ Yamada T, Kondo M, Miyata K, et al. An evaluation of the human relevance
drugs/scienceresearch/researchareas/pharmacogenetics/ucm083378. of the lung tumors observed in female mice treated with permethrin
html. based on mode of action. Toxicol Sci. 2017;157(2):465–486.
U.S. FDA. SNP track. 2017b. Available at: https://www.fds.gov/sciencere- Yu X, Griffith WC, Hanspers K, et al. A system based approach to interpret
search/bioinformaticstools/snptrack/default.htm. dose and time-dependent microarray data: quantitative integration of GO
Van den Berg M, Birnbaum L, Bosveld AT, et  al. Toxic equivalency fac- ontology analysis for risk assessment. Toxicol Sci. 2006;92:560–577.
tors (TEFs) for PCBs, PCDDs, PCDFs for humans and wildlife. Environ Yu XZ, Hong S, Moreira EG, Faustman EM. Improving in vitro Sertoli cell/
Health Perspect. 1998;106:775–792. gonocyte co-culture model for assessing male reproductive toxicity: les-
Van den Berg M, Birnbaum LS, Denison M, et al. The 2005 World Health sons learned from comparisons of cytotoxicity versus genomic responses
Organization reevaluation of human and mammalian toxic equiva- to phthalates. Toxicol Appl Pharmacol. 2009;239:325–336.
lency factors for dioxins and dioxin-like compounds. Toxicol Sci. Zhao Q, Dourson M, Gadagbui B. A review of the reference dose for chlor-
2006;93:223–241. pyrifos. Regul Toxicol Pharmacol. 2006;44:111–124.