Chemical Risk Assessment: Selected Federal Agencies' Procedures, 
Assumptions, and Policies (06-AUG-01, GAO-01-810).		 
								 
As used in public health and environmental regulations, risk	 
assessment is the systematic, scientific description of potential
harmful effects of exposures to hazardous substances or 	 
situations. It is a complex but valuable set of tools for federal
regulatory agencies to identify issues of potential concern,	 
select regulatory options, and estimate the range of a		 
forthcoming regulation's benefits. However, given the significant
yet controversial nature of risk assessments, it is important	 
that policymakers understand how they are conducted, the extent  
to which risk estimates produced by different agencies and	 
programs are comparable, and the reasons for differences in	 
agencies' risk assessment approaches and results. GAO studied the
human health and safety risk assessment procedures of the	 
Environmental Protection Agency, the Food and Drug		 
Administration, the Occupational Safety and Health		 
Administration, and the Department of Transportation's Research  
and Special Programs Administration. This report identifies and  
describes (1) the agencies' chemical risk assessment activities, 
(2) the agencies primary procedures for conducting risk 	 
assessments, (3) major assumptions or methodological choices in  
their risk assessment procedures, and (4) the agencies' 	 
procedures or policies for characterizing the results of risk	 
assessments.							 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-01-810 					        
    ACCNO:   A01530						        
    TITLE:   Chemical Risk Assessment: Selected Federal Agencies'     
             Procedures, Assumptions, and Policies                            
     DATE:   08/06/2001 
  SUBJECT:   Consumer protection				 
	     Environmental monitoring				 
	     Hazardous substances				 
	     Occupational safety				 
	     Regulatory agencies				 
	     Safety regulation					 
	     Safety standards					 
	     EPA Integrated Risk Information System		 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Testimony.                                               **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-01-810
     
A

Report to Congressional Requesters

August 2001 CHEMICAL RISK ASSESSMENT

Selected Federal Agencies? Procedures, Assumptions, and Policies

GAO- 01- 810

Letter 7 Results in Brief 9 Background 11 Context for Agencies? Chemical
Risk Assessments Is Important 18 Agencies? Risk Assessment Procedures Share
Common Features, But Substantive Differences Also Exist 23

Agencies? Risk Assessment Procedures Often Include Precautionary Assumptions
and Methods 30 Risk Characterization Policies and Practices Emphasize
Transparency 43 Conclusions 46 Agency Comments and Our Evaluation 48

Appendixes Appendix I: Objectives, Scope, and Methodology 50 Appendix II:
Chemical Risk Assessment at the Environmental Protection Agency 55

Context for EPA Chemical Risk Assessment 56 Risk Assessment Procedures 64

Guidelines 65 Agencywide Risk Assessment Procedures 67

Hazard Identification 68 Dose- response Assessment 71 Exposure Assessment 79
Ecological Risk Assessment 81 Program- specific Risk Assessment Procedures
84

Office of Pesticide Programs 85 Office of Pollution Prevention and Toxics 90
Office of Emergency and Remedial Response (Superfund

Program) 94 Office of Solid Waste 99 Chemical Emergency Preparedness and
Prevention Office 103 Office of Air and Radiation 107 Office of Water 112
Risk Assessment Assumptions and Methodological Choices 118 Risk
Characterization 151

Appendix III: Chemical Risk Assessment at the Food and Drug Administration
153 Context for FDA Chemical Risk Assessment 154 Risk Assessment Procedures
158

Differences in Risk Assessment Among FDA Product Centers 159 CFSAN Risk
Assessment Procedures 162

CVM Risk Assessment Procedures 166 CDRH Risk Assessment Procedures 167 Risk
Assessment Assumptions and Methodological Choices 170 Risk Characterization
182

Appendix IV: Chemical Risk Assessment at the Occupational Safety and Health
Administration 184 Context for OSHA Chemical Risk Assessment 184 Risk
Assessment Procedures 186

General Approach 186 Hazard Identification 187 Dose- response Assessment 188
Exposure Assessment 193 Risk Assessment Assumptions and Methodological
Choices 194 Risk Characterization 199

Appendix V: Chemical Risk Assessment at the Department of Transportation?s
Research and Special Programs Administration 201

Context for RSPA Chemical Risk Assessment 201 Risk Assessment Procedures 204

Guidance 204 Risk Assessment Approach 205

Identifying Hazards 206 Analyzing the Consequences and Probabilities of
Risks 212 Risk Assessment Assumptions and Methodological Choices 216 Risk
Characterization 218

Glossary 220 Tables Table 1: Comparison of Selected Major Assumptions or
Methods Used in EPA, FDA, and OSHA Risk Assessments 41

Table 2: Chemical Risk Statutes, Tasks, and Mandates for EPA Offices 57
Table 3: EPA?s 1986 Classification System for Characterization of
Carcinogenicity 68

Table 4: Summary of EPA?s Ecological Risk Assessment Process 83 Table 5: EPA
Risk Assessment Assumptions and Methodological

Choices 120 Table 6: Differences in FDA Chemical Risk Assessment Factors 160
Table 7: FDA Risk Assessment Assumptions and Methodological

Choices 171

Table 8: OSHA Risk Assessment Assumptions and Methodological Choices 194
Table 9: UN Classification System for Transport of Dangerous

Goods 207 Table 10: Hazard Classification System of RSPA?s Hazardous

Materials Regulations 208 Table 11: Example of a Hazardous Materials Risk
Assessment Matrix 215

Figures Figure 1: Typical Sequence of Risk Assessment and Risk Management
Processes 13 Figure 2: Full Toxicity Data Is Available for Only a Small
Portion of High- Production- Volume Chemicals 15

Figure 3: Extrapolation for Carcinogens 74 Figure 4: Derivation of a
Reference Dose Using the Benchmark

Dose Approach 78

Abbreviations

ADI acceptable daily intake AWQC ambient water quality criteria BAF
bioaccumulation factor BAT best available technology CAA Clean Air Act CAAA
Clean Air Act Amendments CEPPO Chemical Emergency Preparedness and
Prevention Office CERCLA Comprehensive Environmental Response, Compensation,
and

Liability Act CSFII Continuing Survey of Food Intakes by Individuals CWA
Clean Water Act DES diethylstilbestrol DOT Department of Transportation ED
effective dose EPA Environmental Protection Agency EPCRA Emergency Planning
and Community Right- to- Know Act FAA Federal Aviation Administration FDA
Food and Drug Administration FDAMA Food and Drug Administration
Modernization Act FFDCA Federal Food, Drug, and Cosmetic Act FHWA Federal
Highway Administration FIFRA Federal Insecticide, Fungicide, and Rodenticide
Act FQPA Food Quality Protection Act FRA Federal Railroad Administration FTA
Federal Transit Administration GHS globally harmonized system HAZMAT
hazardous materials HHS Department of Health and Human Services HMIS
Hazardous Materials Information System HPV high production volume IRIS
Integrated Risk Information System ISO International Organization for
Standardization LED lowest effective dose LOAEL lowest observed adverse
effect level LOEL lowest observed effect level MACT maximum achievable
control technology MCL maximum contaminant level MCLG maximum contaminant
level goal MEI maximally exposed individual MLE maximum likelihood estimate

NAAQS national ambient air quality standards NAS National Academy of
Sciences NCTR National Center for Toxicological Research NOAEL no observed
adverse effect level NOEL no observed effect level NRC National Research
Council OAQPS Office of Air Quality Planning and Standards OAR Office of Air
and Radiation OERR Office of Emergency and Remedial Response OHMS Office of
Hazardous Materials Safety OHMT Office of Hazardous Materials Technology OPP
Office of Pesticide Programs OPPT Office of Pollution Prevention and Toxics
OPPTS Office of Prevention, Pesticides and Toxic Substances ORD Office of
Research and Development OSHA Occupational Safety and Health Administration
OSW Office of Solid Waste OW Office of Water PBPK physiologically based
pharmacokinetic PEL permissible exposure limit PMN premanufacture
notification POD point of departure RCRA Resource Conservation and Recovery
Act RfC reference concentration RfD reference dose RSC relative source
contribution RSPA Research and Special Programs Administration SAR
structure- activity relationship SARA Superfund Amendments and
Reauthorization Act SDWA Safe Drinking Water Act SOPs standard operating
procedures STEL short- term exposure limit TI tolerable intake TSCA Toxic
Substances Control Act TWA time- weighted average UCL upper confidence limit
UN United Nations USCG United States Coast Guard USDA United States
Department of Agriculture

Lett er

August 6, 2001 The Honorable W. J. (Billy) Tauzin Chairman, Committee on
Energy and Commerce House of Representatives

The Honorable Paul E. Gillmor Chairman, Subcommittee on Environment

and Hazardous Materials Committee on Energy and Commerce House of
Representatives

As used in public health and environmental regulation, risk assessment is
the systematic, scientific description of potential adverse effects of
exposures to hazardous substances or situations. It is a complex but
valuable set of tools for federal regulatory agencies, helping them to
identify issues of potential concern, select regulatory options, and
estimate the range of a forthcoming regulation?s benefits. For example, the

Environmental Protection Agency (EPA) used risk assessment information in a
1998 final rule to conclude that disinfection byproducts (e. g., chloroform)
in drinking water could cause as many as 9, 300 bladder cancer cases a year,
and that a 24- percent reduction in those byproducts could result in
monetized health benefits of about $4 billion. 1 However, risk assessments
are also sometimes controversial, as evidenced by the fact that the
disinfection byproduct rule was successfully challenged in court over
whether the agency used the best scientific evidence available in

support of certain assumptions. 2 Given the significant yet controversial
nature of risk assessments, it is important that policymakers understand how
risk assessments are conducted, the extent to which risk estimates produced
by different agencies and programs are comparable, and the reasons for
differences in agencies? risk assessment approaches and results.

1 ?National Primary Drinking Water Regulations: Disinfectants and
Disinfection Byproducts? (63 FR 69390, Dec. 16, 1998). The Food and Drug
Administration published a related rule on disinfection byproducts in
bottled water on March 28, 2001 (66 FR 16858). 2 Chlorine Chemistry Council
v EPA, 206 F. 3d 1286 (D. C. Cir. 2000). In March 2001, EPA announced that
it would propose a new assessment for chloroform, the disinfection byproduct
that was the subject of the dispute, using an approach based upon different
assumptions.

You asked us to provide information on selected federal agencies? risk
assessment procedures and the similarities and differences in how the
agencies? personnel are directed to conduct risk assessments. As you
requested, our review focused on the human health and safety (and, to a
lesser extent, ecological) risk assessment procedures of the following four
agencies with primary responsibility for regulating or managing risks from
potential exposure to chemicals: (1) EPA; (2) the Food and Drug
Administration (FDA) within the Department of Health and Human

Services (HHS); (3) the Occupational Safety and Health Administration (OSHA)
within the Department of Labor; and (4) the Department of Transportation?s
(DOT) Research and Special Programs Administration (RSPA). These agencies
regularly conduct chemical risk assessments in support of regulatory
activities and/ or illustrate the diversity of risk

assessment procedures. Our primary objectives were to identify and describe
(1) the general context for the agencies? chemical risk assessment
activities; (2) what the agencies view as their primary procedures for
conducting risk assessments; (3) what the agencies view as their major

assumptions or methodological choices in their risk assessment procedures;
and (4) the agencies? procedures or policies for characterizing the results
of risk assessments. In addressing each of these objectives, we also
identified similarities and differences between and within the agencies. To
the extent feasible, we were also asked to identify as part of the third
objective, (a) at what stages of the risk assessment process the assumptions
are used, (b) the reasons given for their selection, (c) their likely
effects on risk assessment results, and (d) how they compare to the

assumptions and choices used by other agencies or programs in similar
circumstances.

We addressed these objectives by reviewing agencies? general guidance
documents or, if there were no such documents, specific examples of
agencies? risk assessment procedures. We also reviewed previous reports on
agencies? procedures, interviewed agency officials, and provided

detailed descriptions of the relevant procedures to agency officials for
their review and comment. Our review focused on chemical risk assessments in
selected agencies, and therefore did not cover all types of risk assessments
or even all agencies or programs that conduct chemical risk assessments.
Also, our review did not evaluate how the selected agencies? procedures and
policies are applied in individual risk assessments, or how risk assessment
results are used in making regulatory decisions (risk management). We
provided a draft of this report to five risk assessment experts to ensure
technical accuracy. We also provided a draft to officials in each of the
four agencies for their review and comment. The comments

that we received from both the experts and the agencies are reflected in the
?Agency Comments and Our Evaluation? section of this letter. We conducted
this review between February 2000 and March 2001 in accordance with
generally accepted government auditing standards. Details of our scope and
methodology are presented in appendix I. Results in Brief The context in
which chemical risk assessments are conducted plays an

important role in determining what type of assessments federal regulatory
agencies perform and why certain approaches are used. The statutory and
legal context determines the general focus and goals of an agency?s risk
assessment activities and also may shape how risk assessments for those
activities are supposed to be conducted. The specific tasks and purposes for
which an agency will use the results of a particular risk assessment
determine the questions that the assessment needs to address and its scope

and level of detail. For example, risk assessments used by OSHA to set
occupational health standards must demonstrate that a significant risk
exists and that the proposed standard would reduce that risk. However, in
different contexts, FDA and EPA might use risk assessment procedures to

estimate the dose of a chemical that people could consume daily without
harmful effect, and not necessarily need to estimate the actual risk
associated with exposure to that chemical. In other words, the focus of
federal agencies? ?risk? assessments can sometimes be characterized more
accurately as safety assessment (i. e., estimating an exposure level below

which no significant risk will occur) rather than as risk assessment (i. e.,
simply describing the likelihood of a risk). All four of the agencies
included in our review have standard procedures for conducting risk
assessments involving chemical agents, although the agencies vary in the
extent to which they have documented their procedures in written guidance.
There are more similarities than differences in the overall chemical risk
assessment procedures developed

by three of the agencies- EPA, FDA, and OSHA. These agencies? procedures
generally follow four- step process recommended by the National Academy of
Sciences (NAS) of hazard identification, doseresponse assessment, exposure
assessment, and risk characterization. However, there are variations both
among and within the agencies in the details of those steps, particularly
during the exposure assessment step because agencies? regulatory authorities
regarding chemical agents tend to vary according to the kinds or sources of
exposure. The risk assessment procedures in DOT?s RSPA are not based on the
NAS four- step process because of the particular regulatory context in which
RSPA operates.

Instead, a classification system that is harmonized with international
agreements defines what is to be considered a hazardous material for
transportation purposes according to the general physical characteristics of
the material (e. g., whether it is explosive, flammable, or toxic). RSPA?s
analyses of risks then focus on identifying the potential circumstances
under which unintentional releases of hazardous materials could occur during
transit (e. g., due to transportation accidents) and assessing their
consequences and probability of occurrence.

Assumptions are an unavoidable part of risk assessment because science
cannot always provide definitive answers to questions raised at various
stages of an assessment. Agency guidelines and officials we contacted

during our review identified a large number and wide variety of assumptions
that may be used, in the absence of adequate information, during the first
three steps of a risk assessment. The agencies frequently indicated that
particular assumptions were chosen on the basis of their

evaluation of available scientific information, precedents established in
prior assessments, or policy decisions related to the agencies? regulatory
missions or mandates. In about half of the assumptions and

methodological choices identified, the agencies described their likely
effects on risk assessment results, most commonly (particularly at EPA and
FDA) indicating that they were precautionary in nature. 3 Agencies use
precautionary assumptions to ensure that a risk assessment will not
underestimate risks. Consequently, they have the effect of raising the
agencies? estimates of risk, compared to less precautionary options, and
potentially lowering the chemical doses or exposure levels at which agencies
might take regulatory action. Precautionary assumptions are particularly
common in the agencies? procedures for initial screening risk assessments,
when the primary task is to determine whether a risk might exist and more
detailed analysis is needed. Agency guidelines and related

documents indicate that subsequent assessments should involve more rigorous
analyses and fewer precautionary assumptions. There are both similarities
and differences in the assumptions and methods identified by EPA, FDA, and
OSHA. RSPA, given its focus on analyzing transportation accident scenarios
rather than chemical toxicity, uses different 3 These are also referred to
as ?conservative? or ?public- health conservative? assumptions. For
consistency, we use the term precautionary throughout this letter. However,
in the

technical appendices on individual agencies we use the terms expressed in
agency documents or by agency officials.

assumptions and methods because it tends to deal with different analytical
issues. Both EPA and DOT have written, agencywide risk characterization
policies that emphasize clear, complete, and transparent disclosure of the
data, methods, assumptions, and limitations of their risk assessments. The
policies also encourage agency personnel to characterize their risk
estimates in terms of ranges or distributions rather than simply providing a
single point estimate of risks. Both agencies also encourage the use of peer
review to obtain the views of other scientists and experts on the agencies?
risk assessments. Although FDA and OSHA do not have written risk

characterization policies, officials of those agencies said that in practice
they tend to emphasize comprehensive characterizations of risk assessment
results, discussions of limitations and uncertainties, and disclosure of the
data and analytic methodologies on which the agencies relied.

The complexity and diversity of risk assessment policies, procedures,
assumptions, and other choices affecting risk estimates underscore the
importance of transparency in both individual risk assessments and agencies?
general guidance documents. That transparency is particularly important with
regard to disclosing why certain data, methods, and default

assumptions are selected, and under what conditions the agency would depart
from its default assumptions or methods. Prudent use of risk assessment
results in formulating public policy requires policymakers to be aware of
the assumptions and methods used in the preparation of the assessments.

Background Risk assessments are conducted to estimate whether and/ or how
much damage or injury can be expected from exposures to a given risk agent
and

to assist in determining whether these effects are significant enough to
require action, such as regulation. The effects of concern can be diseases
such as cancer, reproductive and genetic abnormalities, workplace injuries,
or various types of ecosystem damage. The risk agent analyzed in an
assessment can be any number of things, including chemicals, radiation,
transportation systems, or a manufacturing process. The product of a risk
assessment is a quantitative and/ or qualitative statement regarding the

probability that an exposed population will be harmed and to what degree.
Risk assessment, particularly quantitative risk assessment, is a relatively
new discipline, developed in the first half of the 20 th century to
establish

various health and safety codes and standards. The role of risk assessment
in the regulatory process was accelerated by the enactment of various
health, safety, and environmental statutes in the early 1970s. The
development of chemical risk assessment procedures has traditionally
followed two different tracks- one for assessments of cancer risks and

another for assessments of noncancer risks. The procedures associated with
cancer risks have historically assumed that there is no ?threshold? below
which an agent would not cause adverse effects. In contrast,

procedures for assessments of noncancer risks were largely developed under
the assumption that there is such a threshold- that exposures up to a
certain level would not be expected to cause harm.

In 1983, NAS identified four steps in the risk assessment process: (1)
hazard identification (determining whether a substance or situation could
cause adverse effects), (2) dose- response assessment (determining the
relationship between the magnitude of exposure to a hazard and the
probability and severity of adverse effects), (3) exposure assessment
(identifying the extent to which exposure actually occurs), and (4) risk

characterization (combining the information from the preceding analyses into
a conclusion about the nature and magnitude of risk). 4 This paradigm,
originally intended to address assessments of long- term health risks, such
as cancer, has become a standard model for conducting risk assessments, but
is not the only model (e. g., different models are used for ecological risk

assessments). According to NAS, the results of the risk assessment process
should be conceptually distinguished from how those results are used in the
risk management process (e. g., the decision on where to establish a
particular standard). As illustrated by figure 1, the risk management
decision considers other information in addition to the risk
characterization.

4 Risk Assessment in the Federal Government: Managing the Process (commonly
referred to as the ?Red Book?), National Research Council of the National
Academy of Sciences (National Academy Press, 1983).

Figure 1: Typical Sequence of Risk Assessment and Risk Management Processes
Risk assessment Risk management

Control Legal

Dose- response options considerations

assessment Risk Hazard

Risk management

identification characterization

decisions Exposure

Other economic assessment

and social factors Source: EPA Office of Research and Development.

More recent reports have updated and expanded on these original concepts. In
1996, NAS urged risk assessors to update the original concept of risk
characterization as a summary added at the end of a risk assessment. 5
Instead, the report suggested that risk characterization should be a
?decision- driven? activity directed toward informing choices and solving
problems and one that involves decision makers and other stakeholders from
the very inception of a risk assessment. In this updated

view, the nature and goals of risk characterization are dictated by the
goals of the risk management decisions to be made. Similarly, the
Presidential/ Congressional Commission on Risk Assessment and Risk
Management (hereinafter referred to as the Presidential/ Congressional
Commission) recommended in 1997 that the performance of risk assessments be
guided by an understanding of the issues that will be important to risk
management decisions and to the public?s understanding of what is needed to
protect public health and the environment. 6

5 Understanding Risk: Informing Decisions in a Democratic Society, National
Research Council of the National Academy of Sciences (National Academy
Press, 1996). 6 Risk Assessment and Risk Management in Regulatory Decision-
Making, The Presidential/ Congressional Commission on Risk Assessment and
Risk Management, (Final Report, Volume 2, 1997).

Data on Chemical Health Substantial numbers and amounts of chemical
substances and mixtures are

Effects and Exposures Are produced, imported, and used in the United States.
For example, there are

Limited over 70, 000 commercial chemicals in EPA?s Toxic Substances Control
Act (TSCA) Chemical Substances Inventory, and the agency receives about
1,500 petitions each year requesting the approval of new chemicals or new

uses of existing chemicals. 7 However, there is relatively little empirical
data available on the toxicity of most chemicals and the extent to which
people or the environment might be exposed to the chemicals. For example, we
previously reported that EPA?s Integrated Risk Information System (IRIS),
which is a database of the agency?s consensus on the potential health
effects of chronic exposure to various substances found in the environment,
lacks basic data on the toxicity of about two- thirds of

known hazardous air pollutants. 8 Furthermore, to the extent that data on
health effects are available, the data are more often from toxicological
studies involving animal exposures than from epidemiological studies
involving human exposures. As a consequence, chemical risk assessments must
rely often on extrapolation from animal studies and are quite different from
risk assessments that use epidemiological studies or actuarial data (such as
accident statistics). 7 Excluding polymers (which are considered unlikely to
present significant risk concerns), EPA?s TSCA inventory identified about
15, 000 chemicals produced or imported at levels above 10, 000 pounds per
year. There are also other categories of chemical substances (such

as drugs, cosmetics, food additives, and pesticides) that are exempt from
TSCA but subject to control under other federal statutes. The number of
chemicals actually in commerce varies as new chemicals are added and other
chemicals are withdrawn. 8 Major Management Challenges and Program Risks:
Environmental Protection Agency (GAO/ OCG- 99- 17, Jan. 1999).

The limited nature of information on chemical toxicity was illustrated in a
1998 EPA report on the data that were publicly available on approximately
3,000 high- production- volume (HPV) chemicals. 9 For each of these
chemicals, EPA examined the available data corresponding to six basic tests
that have been internationally agreed to as necessary for a minimum
understanding of a chemical?s toxicity. 10 As shown in figure 2, the agency
concluded that the full set of basic toxicity data was available for only

about 200 (7 percent) of the chemicals, and that 43 percent of the chemicals
did not have publicly available data for any of the six tests.

Figure 2: Full Toxicity Data Is Available for Only a Small Portion of High-
Production- Volume Chemicals

7% Have all six of basic tests for

minimum understanding of toxicity per the international Screening
Information Data Set

50% 43% Are missing all of these tests

Have one to five of these tests N= 2863 HPV chemicals Source: EPA, Chemical
Hazard Data Availability Study: What Do We Really Know About the Safety of
High Production Volume Chemicals? (April 1998).

9 High- production- volume chemicals are those imported or produced at
volumes of more than 1 million pounds per year. Note that, for regulatory
approval purposes, some offices within EPA have access to confidential
business information on commercial chemicals and pesticides that would not
be reflected in this study of ?publicly available? toxicity data. 10 The six
tests are acute toxicity, chronic toxicity, developmental/ reproductive
toxicity, mutagenicity, ecotoxicity, and environmental fate. Collectively,
these tests are known as the Screening Information Data Set program.

There are also significant gaps in the available data on the extent to which
people are exposed to chemicals. For example, last year we reviewed federal
and state efforts to collect human exposure data on more than 1, 400

naturally occurring and manmade chemicals considered by HHS, EPA, and other
entities to pose a threat to human health. 11 We reported that, taken
together, HHS and EPA surveys measured the degree of exposure in the

general population for only 6 percent of those chemicals. Even for those
chemicals that were measured, information was often insufficient to identify
smaller population groups at high risk (e. g., women, children, and the
elderly). Uncertainty Contributes to

There is an ongoing debate about the appropriate application of risk
Controversy about

assessment in federal regulation. In 1990, Congress mandated that a Chemical
Risk Assessment commission be formed to ?make a full investigation of the
policy implications and appropriate uses of risk assessment and risk
management in regulatory programs under various Federal laws to prevent
cancer and

other chronic human health effects which may result from exposure to
hazardous substances.? The Presidential/ Congressional Commission published
its final report in 1997, and noted that often ?the controversy arises from
what we don?t know and from what risk assessments can?t tell us.? 12 NAS has
also emphasized that science cannot always provide

definitive answers to questions raised during the course of conducting a
risk assessment, so risk assessors must use assumptions throughout the
process that reflect professional judgments and policy choices. 13

11 Toxic Chemicals: Long- Term Coordinated Strategy Needed to Measure
Exposures in Humans (GAO/ HEHS- 00- 80, May 2, 2000). See also Environmental
Information: EPA Needs Better Information to Manage Risks and Measure
Results (GAO- 01- 97T, Oct. 3, 2000). 12 Framework for Environmental Health
Risk Management, The Presidential/ Congressional Commission on Risk
Assessment and Risk Management, (Final Report, Volume 1, 1997), p. 23.

13 See Risk Assessment in the Federal Government: Managing the Process
(1983) and Science and Judgment in Risk Assessment (1994).

One focus of the risk assessment debate has been agencies? use of
precautionary assumptions and analytical methods. The term

?precautionary? refers to the use of methods and assumptions that are
intended to produce estimates that should not underestimate actual risks.
Some critics of federal risk assessment practices believe agencies use

assumptions that are unjustifiably precautionary in the face of new
scientific data and methods, thereby producing estimates that overstate
actual risks. The critics contend that this effect is compounded when
multiple precautionary assumptions are used. Others, however, criticize

agency practices for not being precautionary enough in the face of
scientific uncertainties, failing, for example, to adequately account for
the synergism of exposures to multiple chemicals or the risks to persons
most exposed or most sensitive to a particular toxic agent. 14 Other
observers, including NAS, have expressed concerns about whether the
agencies? procedures and assumptions are sufficiently transparent, thereby
providing decision makers and the public with adequate information about the
scientific and policy bases for agencies? risk estimates as well as the
limitations and uncertainties associated with those estimates.

14 Proposals have been introduced in Congress regarding this issue. For
example, H. R. 199, proposed in the 106 th Congress, would have required the
EPA Administrator to evaluate, among other things, environmental health
risks to vulnerable subpopulations (e. g., children, pregnant women, and the
elderly) and to ensure that all EPA standards protect such subpopulations
with an adequate margin of safety.

We have discussed these issues in several previous reports. For example, in
1993, we noted that EPA used precautionary assumptions throughout the
process that it used to assess risk at Superfund hazardous waste sites, and
that the agency had been criticized for overstating risk by combining
precautionary estimates. 15 In September 2000, we reported on EPA?s use of
precautionary ?safety factors? pursuant to the Food Quality Protection Act
of 1996. 16 In October 2000, we said that three factors influenced EPA?s use

of precautionary assumptions in assessing health risks: (1) the agency?s
mission to protect human health and safeguard the natural environment, (2)
the nature and extent of relevant data (e. g., animal versus human studies),
and (3) the nature of the health risk being evaluated (e. g., cancer versus
noncancer risks). 17 Context for Agencies? The context in which chemical
risk assessments are conducted plays an

Chemical Risk important role in determining what type of assessments federal
regulatory agencies perform and why certain approaches are used. Two
dimensions Assessments Is

seem particularly important to understanding the context for an agency?s
Important

chemical risk assessment activities: (1) the general statutory and legal
framework underlying the agency?s regulation of chemicals and (2) how the
agency plans to use the risk assessment information. 18 The statutory and
legal framework determines the general focus and goals of an agency?s
chemical risk assessment activities and also can shape how risk assessments
for those activities are supposed to be conducted. The specific tasks and
purposes for which an agency will use the results of a particular risk
assessment determine the questions that the assessment needs to address and
the scope and level of detail of the assessment.

15 Superfund: Risk Assessment Process and Issues (GAO/ T- RCED- 93- 74,
Sept. 30, 1993). 16 Children and Pesticides: New Approach to Considering
Risk Is Partly in Place (GAO/ HEHS- 00- 175, Sept. 11, 2000).

17 Environmental Protection Agency: Use of Precautionary Assumptions in
Health Risk Assessments and Benefits Estimates (GAO- 01- 55, Oct. 16, 2000).
18 Other contextual factors, such as the data limitations and scientific
uncertainty, are also important. On a practical level, the availability of
resources (e. g., staff, schedule, funding, data) also affects the scope and
level of detail that an agency can provide in any given risk assessment.
However, such factors are either so broadly applicable or so case specific
that they do not distinctively characterize the risk assessment procedures
of an agency or program.

Statutory and Legal A diverse set of statutes addresses potential health,
safety, and Framework

environmental risks associated with chemical agents. These statutory
mandates generally focus on different types and sources of exposure to
chemicals, such as consumption of pesticide residues in foods, occupational
exposures to chemicals, or inhalation of toxic air pollutants. Therefore,
different agencies (and different offices within those agencies)

have distinctive concerns regarding chemical risks. For example, each major
program office within EPA (e. g., the Office of Air and Radiation or the
Office of Water) is responsible for addressing the risk- related mandates of
one or more statutes (e. g., the Clean Air Act, the Clean Water Act, or the
Safe Drinking Water Act). Also, international agreements provide

important legal context for transportation risk assessment activities. For
example, criteria for classifying dangerous chemicals in transportation have
been internationally harmonized through the United Nations? Recommendations
on the Transport of Dangerous Goods.

The legal framework underlying chemical regulation influences both the
extent to which risk assessment is needed for regulatory decision making and
how risk assessments are supposed to be conducted. Some statutes require
regulatory decisions to be based solely on risk (considering only

health and environmental effects), some require technology- based standards
(such as requiring use of the best available control technology), and still
others require risk balancing (requiring consideration of risks, costs, and
benefits). For example, section 112 of the Clean Air Act (CAA), as amended,
has a technology- based mandate requiring the use of the

maximum achievable control technology to control emissions of hazardous air
pollutants. A risk assessment is not needed to determine such technology,
but would be used to evaluate residual risks that remain after that
technology is in use. Some statutes also place the primary responsibility
for conducting risk assessments and compiling risk- related data for a
particular chemical or source of exposure to chemical agents with industry,
states, or local entities, rather than with the federal regulatory agencies.
For example, industry petitioners have the primary

responsibility to provide the data needed to support registration and
tolerances from EPA for their pesticides, including information on the
toxicological effects of the pesticides. 19

19 Registration involves the licensing of pesticides for sale and use in
agriculture and extermination. No chemical may be sold in the United States
as a pesticide without such registration, which establishes the conditions
of legal use. Pesticide tolerances are the

concentrations permitted to remain in or on food, as it is available to the
consumer.

Statutes can also affect risk assessment by specifically defining what will
be considered a hazard, directing the agency to take certain methodological
steps, or specifying the exposure scenario of regulatory concern. For
example, in response to the ?Delaney Clause? amendments to the Federal Food,
Drug, and Cosmetic Act, FDA identifies any food additive

for which an adequately conducted animal cancer study indicates that the
additive produces cancer in animals as a carcinogen under the conditions of
the study. No further corroboration or weight- of- evidence analysis is
required. The Food Quality Protection Act of 1996 requires EPA to add an
additional 10- fold safety factor to protect infants and children when
deriving standards for allowable pesticide residues in foods, unless
reliable data show that a different factor will be safe. Provisions in the

Occupational Safety and Health Act focus OSHA?s risk assessments on
estimating the risks to workers exposed to an agent for a working lifetime.
However, in most cases the statutes simply provide a general framework
within which the agencies make specific risk assessment assumptions and
methodological choices. For example, section 109 of the CAA requires EPA to
set national ambient air quality standards that in the judgment of the EPA
Administrator- and allowing for an ?ample margin of safety?- are

requisite to protect the public health. 20 EPA risk assessors translate that
general requirement into specific risk assessment assumptions and methods
(e. g., whether to assume a threshold or no- threshold relationship between
dose and response at low doses).

Use of Risk Assessment The specific purpose or task of an assessment
determines the kinds of risk Results information needed for the agency to
make its risk management decisions, and can significantly influence the
scope and level of detail required of a risk assessment. For example,

 If the agency?s task is to set a specific health- based standard (e. g., a
national air quality standard), a rigorous and detailed estimate of risks at
particular exposure levels might be required.

 If the agency?s task is to decide whether to approve the production and
use of commercial chemicals or pesticides, risk assessors may initially
focus on potential upper- bound exposures (e. g., assuming that a chemical
agent will be used at the maximum level permitted by law or

20 An ambient air quality standard is a national target for an acceptable
concentration of a specific pollutant in air.

focusing on individuals who consume the greatest amounts of a food
containing residues of the agent at issue). If such upper- bound estimates
exhibit no cause for concern, the agency may have no need to complete a more
comprehensive and refined risk assessment.

 A decision on whether to add or remove a chemical from the list of
potential hazards might focus the risk assessors on determining whether the
potential risk is above or below a specific threshold level, such as the
risk of 1 extra cancer case over the lifetime of 1 million people.

The influence of the specific regulatory task at hand is illustrated by a
method commonly used by agencies for risk assessments of noncancer health
effects. Agencies such as EPA and FDA have historically attempted to
identify a dose level of a chemical associated with no observed adverse
effect level (NOAEL) in animal experiments- or the lowest observed

adverse effect level (LOAEL) in the study, if every tested dose exhibited
some effect. 21 They then divided that NOAEL or LOAEL dose by multiple
?safety? or ?uncertainty? factors to account for the possibility that humans

may be more sensitive to the chemical than animals and other uncertainties.
This procedure is designed to identify a dose not likely to result in harm
to humans, not to provide an explicit quantitative estimate of the risks
associated with a given chemical. In other words, sometimes the focus of
federal agencies? ?risk? assessments could more accurately be described as a
safety assessment (i. e., estimating a ?safe? level of exposure to chemical
agents or a dose below which no significant risk is expected to occur)
rather than a risk assessment (i. e., estimating the actual risks associated
with exposures to chemical agents).

Implications of Contextual Because of contextual differences, the risk
assessment procedures used, Differences

the resulting risk estimates (and regulatory actions based upon those
estimates), and even whether a substance would be subject to risk
assessment, can vary among different agencies and programs within the same
agency. The following examples illustrate how contextual differences affect
the conduct of risk assessments.

 Because regulation of certain wastes may be impractical or otherwise
undesirable, regardless of the hazards that the waste might pose, 21 FDA
often determines a no observed effects level (NOEL) rather than a NOAEL
because

many significantly altered, standard toxicological endpoints are assumed to
be adverse to animals and/ or humans even in the absence of data affirming
that assumption.

Congress and EPA exempted certain materials (e. g., agricultural or mining
and mineral processing wastes) from the definitions of hazardous wastes. If
a material meets one of the categories of exemptions, it cannot be
identified as a hazardous waste even if it otherwise meets the criteria for
listing as a hazardous waste. For example, according to EPA?s RCRA
Orientation Manual, wastes generated in raw material, product storage, or
process (e. g., manufacturing) units are exempt from EPA?s hazardous waste
regulation while the waste remains in such units. However, OSHA might assess
and regulate risks associated with such materials as part of its mission

to protect the health of employees in the workplace.

 FDA and EPA both assess potential human health risks associated with
ingestion of chemical substances. If a substance is being assessed by FDA as
a food additive and results from any adequate study indicate that the
substance produces cancer in animals, FDA labels that additive as a
carcinogen without considering other scientific evidence (per the Delaney
clause of the Federal Food, Drug, and Cosmetic Act, as

amended). However, when assessing the risks associated with consumption of
residues from animal drugs (FDA) and pesticides (EPA) the agencies may need
to consider many scientific studies in determining whether and under what
conditions an agent might cause cancer or other adverse health effects in
humans.

 EPA?s risk assessments of commercial chemicals under TSCA vary depending
on whether the chemical at issue is ?existing? or ?new.? For EPA to control
the use of an existing chemical, the agency must make a

legal finding that the chemical will present an unreasonable risk to human
health or the environment. EPA said this standard requires the agency to
have conclusive data on risks associated with that particular chemical. By
comparison, newly introduced chemicals can be regulated

based on whether they may pose an unreasonable risk, and this finding of
risk can be based on data for structurally similar chemicals, not just data
on that particular chemical. Because industrial chemicals in commerce were
?grandfathered? under TSCA into the inventory of existing chemicals more
than 20 years ago, without considering whether they were hazardous, there
are situations in which existing chemicals might not be controlled while, at
the same time, EPA would act to control a new chemical of similar or less
toxicity.

 Within EPA?s Office of Water, risk assessments vary depending on whether
the assessment is done to establish drinking water standards or standards
for ambient water (e. g., bodies of water such as lakes and rivers). Risk
assessments for drinking water standards focus solely on human health
effects, but assessments used to establish ambient water

quality criteria consider both human health and ecological effects. Even
when considering just the human health risks, an important difference
between the ambient and drinking water risk assessments is an additional
focus for ambient water on exposures to contaminated water through
consumption of contaminated fish or shellfish. This additional factor is a
primary reason for potential differences in drinking water and ambient water
risk estimates and standards for the same chemical.

Appendices II through V describe the relevant contextual factors for each of
the four selected agencies in greater detail.

Agencies? Risk All four of the agencies included in our review have standard
procedures

Assessment for conducting risk assessments, although the agencies vary in
the extent to which their procedures are documented in written guidance. In
general, Procedures Share there are more similarities than differences
across EPA, FDA, and OSHA

Common Features, But procedures, because each of these agencies generally
follows the four- step Substantive

NAS risk assessment process. The procedures address the same basic questions
regarding hazard identification, dose- response assessment, and

Differences Also Exist exposure assessment. The specific analytical methods
and approaches in those procedures are also very similar (e. g.,
extrapolating from animal study data to model dose- response relationships
in humans, and generally

using different procedures for assessing cancer and noncancer risks). The
most substantive differences across and within these agencies are related to
exposure assessment, reflecting the diversity in the agencies? regulatory
authorities regarding chemical agents across different kinds or sources of

exposure. For example, both OSHA and EPA consider methylene chloride (also
known as dichloromethane) to be a probable human carcinogen. However, this
same chemical can be identified as a significant hazard by one agency in one
exposure setting (OSHA for purposes of assessing health risks associated
with occupational exposures) but as a low hazard by another agency in a
different setting (EPA for purposes of Superfund

hazard ranking screening). 22 RSPA, although sharing a concern over
identifying risks and analyzing their consequences and probabilities of 22
EPA has taken other actions regarding exposures to methylene chloride. For
example, EPA requires that releases of methylene chloride of 1,000 pounds or
more be reported to the federal government. EPA also has guidelines on how
much of this chemical people can be exposed to without harming their health
(e. g., EPA recommends that children not drink water that contains more than
13. 3 parts of methylene chloride per million parts of water for longer than
1 day or with more than 1.5 parts per million for longer than 10 days).

occurrence, has a different structure to its risk assessments than the other
three agencies because of its focus on risks associated with unintentional
releases of hazardous materials during transportation. In general, all four
agencies are incorporating more complex analytical models and methods into
their risk assessment procedures. However, some of the advanced models
require much more detailed information than may be currently available for
many chemicals.

Risk Assessment EPA has extensive written internal risk assessment
procedures. For Procedures at EPA

example, EPA has agencywide guidelines, policy memoranda, and handbooks
covering the following aspects of risk assessment:

 carcinogen risk assessment,

 neurotoxicity risk assessment,

 reproductive toxicity risk assessment,

 developmental toxicity risk assessment,

 mutagenicity risk assessment,

 health risk assessment of chemical mixtures,

 exposure assessment,

 ecological risk assessment,

 evaluating risk to children,

 use of probabilistic analysis in risk assessment,

 use of the benchmark dose approach in health risk assessment, and

 use of reference dose and reference concentration in health risk
assessment. EPA also has numerous program- specific guidelines and policy
documents, such as the Risk Assessment Guidance for Superfund series and a
set of more than 20 science policy papers and guidelines from the Office of
Pesticide Programs in response to the Food Quality Protection Act of 1996.
Many of the agency?s guidance documents are draft revisions to earlier

documents or procedures or draft guidance on new issues that have not
previously been addressed by EPA. Although such drafts are not yet final,
official statements of agency policies or procedures, they may better
represent the current practice of risk assessment in EPA than earlier
?final? documents.

EPA generally follows the NAS four- step risk assessment process. (The major
exception is the agency?s Chemical Emergency Preparedness and Prevention
Office, which follows a different set of procedures because of its focus on
risks associated with accidental chemical releases from fixed facilities.
See app. II for a discussion of this office?s risk assessment procedures.)
EPA?s risk assessment activities generally involve both the program offices
(e. g., the Office of Air and Radiation or the Office of Solid Waste) and
the Office of Research and Development (ORD), which is the principal
scientific and research arm of the agency. ORD often does risk assessment
work for EPA program offices that focuses on the first two steps in the
four- step NAS process- hazard identification and doseresponse

assessment- in particular, the development of ?risk per unit exposed?
numbers. 23 Preparation of the final two steps in the process- exposure
assessment and risk characterization- tends to be the responsibility of the
relevant program offices. Several programs, for example, frequently use a
single hazard assessment, but for different exposure scenarios. There are,
however, exceptions to this generalization.

For example, ORD carries out all steps for highly complex, precedentsetting
risk assessments, such as those for dioxin and mercury. There are also
instances when EPA program offices carry out all four steps of the process.
In some situations, EPA agencywide procedures also depart slightly from the
NAS paradigm. For example, when assessing noncancer health effects, EPA?s
normal practice is to do hazard identification in conjunction with the
analysis of dose- response relationships, rather than as distinct steps.
According to EPA?s guidelines, this is because the

determination of a hazard is often dependent on whether a dose- response
relationship is present. In the case of ecological risk assessments, EPA?s
guidelines suggest a three- step process consisting of (1) problem
formulation, (2) analysis, and (3) risk characterization, rather than the
fourstep process used for health risk assessments.

23 ORD also manages EPA?s IRIS database that contains agency- consensus
information on human health effects that may result from exposure to various
chemicals in the environment.

EPA has identified several new directions in its approach to exposure
assessment. First is an increased emphasis on total (aggregate) exposure to
a particular agent via all pathways. EPA policy directs all regulatory
programs to consider in their risk assessments exposures to an agent from

all sources, direct and indirect, and not just from the source that is
subject to regulation by the office doing the analysis. 24 Another area of
growing attention is the consideration of cumulative risks, when individuals
are exposed to many chemicals at the same time. The agency is also
increasing its use of probabilistic modeling methods to analyze variability
and

uncertainty in risk assessments and provide better estimates of the range of
exposure, dose, and risk to individuals in a population than are provided by
single point estimates. EPA?s guidance on probabilistic methods outlines
standards that exposure data prepared by industry or other external analysts
must meet to be accepted by EPA.

Risk Assessment FDA and OSHA also generally follow the NAS risk assessment
paradigm,

Procedures at FDA and but neither FDA nor OSHA had written internal guidance
specifically on OSHA

conducting risk assessments at the time of our review. However, both
agencies? standard procedures are well documented in the records of actual
risk assessments and in summary descriptions that have appeared in
scientific and professional literature. In addition, FDA has published
volumes of guidance on risk assessments for use by external parties affected
by the agency?s regulations (e. g., animal drug manufacturers seeking FDA
approval for their products). According to FDA officials, the

documents are meant to represent the agency?s current thinking on the
scientific data and studies considered appropriate for assessing the safety
of a product, and sometimes include detailed descriptions of the risk
assessment methods deemed appropriate to satisfy FDA?s requirements under
various statutory provisions. However, these guidelines do not preclude the
use of alternative procedures by either FDA or external parties.

The responsibility for conducting risk assessments in FDA is divided among
the agency?s program offices. For example, FDA?s Center for Food

24 The Presidential/ Congressional Commission noted that, traditionally,
risk assessments have largely focused on assessing the risks of just one
chemical in one medium at a time. Although some EPA offices, such as the
Office of Pesticide Programs and Office of Water, conduct more comprehensive
risk assessments, the Commission pointed out that few other regulatory
agencies consider exposures or risks comprehensively, and EPA often does not
do so because of resource or statutory limitations.

Safety and Applied Nutrition (CFSAN) is responsible for assessing risks
posed by food additives and contaminants, while the Center for Veterinary
Medicine (CVM) is responsible for assessing risks posed by animal drug
residues in food. In addition, FDA?s National Center for Toxicological
Research conducts scientific research to support the agency?s regulatory
needs, including research aimed at understanding the mechanisms of toxicity
and carcinogenicity and at developing and improving risk assessment methods.
FDA officials said that there are variations in the risk

assessment approaches used among the agency?s different product centers and,
in some cases, within those centers. In general, those variations are
traceable to differences in factors such as the substances being regulated,
the nature of the health risks involved (particularly carcinogens versus
noncarcinogens), and whether the risk assessment is part of the process to
review and approve a product before it can be marketed and used

(premarket) or part of the process of monitoring risks that arise after a
product is being used (postmarket). For example, risk assessments by CFSAN?s
Office of Food Additive Safety and Office of Nutritional Products, Labeling
and Dietary Supplements are mandatory for new dietary ingredients (and are
used for premarket review of such ingredients) but discretionary for other
food (and are associated with postmarket review). A unique characteristic of
the hazard identification phase of risk assessment in FDA is that, by
statute, if there is an adequate study that indicates a food additive can
cause cancer in animals, that additive is labeled as a carcinogen under the
conditions of the study. No additional

corroboration or weight- of- evidence analysis is required in such cases,
and there is no need to complete the other three risk assessment steps
before proceeding to a regulatory decision. FDA?s CVM is permitted to allow
the use of carcinogenic drugs in food- producing animals under the DES
proviso of the Federal Food, Drug, and Cosmetic Act, as amended, provided
that ?no residue of such drug will be found.?

OSHA?s Directorate of Health Standards Programs is primarily responsible for
conducting the agency?s chemical risk assessments. Such assessments focus
specifically on the potential risks to workers associated with exposures to
chemicals in an occupational setting. In contrast to agencies regulating
environmental exposures to toxic substances, OSHA frequently has relevant
human data available on occupational exposures. Even when the agency
assesses risks based on animal data, OSHA said that the workplace exposures
of concern are often not far removed from levels tested in the animal
studies. Therefore, OSHA?s risk assessments do not extrapolate as far beyond
the range of observed toxicity as might be

necessary to characterize environmental exposure risks. OSHA?s risk
assessment procedures have also evolved to consider data from advanced
physiologically based pharmacokinetic (PBPK) models on the relationship
between administered doses and effective doses (i. e., the amounts that
actually reach a target organ or tissue). 25 However, PBPK models are

complicated and require substantial data, which may not be available for
most chemicals. OSHA therefore developed a set of 11 criteria to judge
whether available data are adequate to permit the agency to rely on PBPK
analysis in place of administered exposure levels when estimating human
equivalent doses.

25 Pharmacokinetics is the study of the absorption, distribution,
metabolism, and elimination of chemicals in humans and animals. It is the
basis for developing what are believed to be more realistic and accurate
models of the movement and interactions of a chemical with blood, tissues,
and organs once it enters the body, including consideration of the body?s

ability to repair damage caused by a chemical.

Risk Assessment The applicable risk assessment guidance for RSPA is
generally documented

Procedures at RSPA within broader DOT- wide guidance on conducting
regulatory analyses and also in materials describing the agency?s Hazardous
Materials Safety Program. Because of the particular regulatory context in
which it operates,

RSPA does not apply the NAS four- step paradigm for risk assessment used by
EPA, FDA, and OSHA. RSPA is primarily concerned with potential risks
associated with the transportation of hazardous materials. In particular, it
is concerned with short- term or acute health risks due to relatively high
exposures from unintentional release of hazardous materials. For its
purposes, RSPA identifies chemicals as hazardous materials according to a
regulatory classification system that is harmonized with internationally
recognized criteria and EPA- defined hazardous substances. This
classification system defines the type of hazard associated with a given

material according to chemical, physical, or nuclear properties (e. g.,
whether it is an explosive, a flammable liquid, or a poisonous substance)
that can make it dangerous in or near transporting conveyances.

Therefore, a chemical?s toxicity is only one of its characteristics of
concern to RSPA, rather than being the primary focus of analysis as in
assessments of the other three agencies. The risk analyses by RSPA focus on
identifying the potential circumstances under which unintentional releases
of hazardous materials could occur during transit (e. g., due to
transportation accidents) and assessing their consequences and probability
of occurrence. Analysis of different modes (e. g., via truck, rail, or
aircraft) and routes of transportation is an important component of RSPA?s
consequence and

probability analyses. 26 Through DOT databases, directly relevant data on
the incidence and severity of hazardous materials transportation accidents
are available to assist RSPA in identifying and analyzing hazard scenarios.
Appendices II through V provide more detailed descriptions of the standard
procedures for chemical risk assessments in each of the four selected
agencies.

26 Assessment and regulation of risks associated with substances transported
by bulk marine carriers are the responsibility of the United States Coast
Guard.

Agencies? Risk Assumptions and methodological choices are an integral and
inescapable

Assessment part of risk assessment. They are often intended to address
uncertainty in the absence of adequate scientific data. However, those
assumptions and

Procedures Often methods may also reflect policy choices, such as how to
address variability Include Precautionary in exposures and effects among
different individuals and populations, or

Assumptions and particular contextual requirements. To the extent that the
four agencies identified the specific reasons for selecting their major
assumptions or Methods

methods, they most often attributed their choices to an evaluation of
available scientific data, the precedents established in prior risk
assessments, or policy decisions related to their regulatory missions.
Agencies? statements regarding the likely effects of their preferred
assumptions and methods most often addressed the extent to which the default
options would be considered precautionary. Some of the major assumptions and
methodological choices of EPA, FDA, and OSHA address similar issues and
circumstances during the risk assessment process, especially regarding
assessment of a chemical?s toxicity.

Agencies? Assumptions and Agency procedural guidelines and officials we
contacted during our review

Methodological Choices identified a large number and wide variety of major
assumptions and

Vary methodological choices that they might use when conducting chemical
risk assessments, in the absence of information that would indicate the
particular assumption or method is not valid in a given case. Some of these

assumptions and methodological choices were very broad (e. g., the common
assumption that, in the absence of evidence to the contrary, substances that
produce adverse health effects in experimental animals

pose a potential threat to humans). Other assumptions and choices were more
specific, covering particular details in the analytical process (e. g.,
identifying the preferred options for extrapolating high dose- response
relationships to low doses). EPA and OSHA identified some of their choices

as the default assumptions and methods of their agencies. FDA officials said
that their agency does not require the use of specific default assumptions
or risk assessment methods, but there are assumptions and methods that
typically have been used as standard choices in FDA risk assessments.
Although assumptions are also needed in RSPA?s risk assessments, RSPA
officials said that they do not have any default assumptions. Instead, they
said that their assumptions are specific to, and must be developed as part
of, each risk assessment. Appendices II through V present detailed
information on some of what the agencies identified as their major
assumptions and methodological choices

in chemical risk assessments. The tables illustrate both the number and
variety of assumptions that agencies may use when conducting those
assessments.

The following sections summarize information that was available from the
four agencies? procedures and related documents on (a) when the agencies
employ major assumptions and methods, (b) their reasons for selecting these
options, (c) the likely effects on risk assessment results of these options,
and (d) how they compare to the assumptions and choices used by

other agencies or programs in similar circumstances. In some cases the
agencies? documents did not contain this information, but there is no
requirement that the agencies do so. Also, the reason for using a particular
assumption and its effect on risk assessment results can vary on a case-
bycase basis, and therefore might not be addressed in general risk
assessment guidance. Nevertheless, both NAS and the Presidential/
Congressional Commission recommended greater transparency regarding the
procedures,

assumptions, and results of agencies? risk assessments. Also, as will be
discussed more fully later in this report, the agencies? own risk
characterization policies and practices emphasize the value of such
transparency in communicating information about risk assessment procedures
and results. Recent regulatory reform proposals considered by Congress have
had provisions requiring transparency in the use of assumptions. 27

When Assumptions and As previously mentioned, NAS and the Presidential/
Congressional Methods Are Used

Commission have both emphasized that science cannot always provide
definitive answers to questions raised during a risk assessment. For
example, in 1983, NAS identified at least 50 points during the course of a
cancer risk assessment when choices had to be made on the basis of
professional judgment, not science. EPA?s guidelines similarly point out
that, because there is no instance in which a set of data on an agent or
exposure is complete, all risk assessments must use general knowledge and
policy guidance to bridge data gaps. Except in the case of RSPA, default or
standard assumptions and methods may be used by agencies to

27 For example, S. 746, proposed in the 106 th Congress, provided that when
a risk assessment involves a choice of assumptions the agency must (1)
identify significant assumptions and their scientific and policy bases, (2)
explain the basis for any choices among assumptions, and (3) describe
reasonable alternative assumptions not selected that would have had a
significant effect on the results of the assessment.

address these gaps in knowledge, and to encourage consistency in the efforts
of agencies? risk assessors to address such basic issues as:

 uncertainty in the underlying data, model parameters, or state of
scientific understanding of how exposure to a particular chemical could lead
to adverse effects;

 variability in the potential extent of exposure and probability of adverse
effects for various subgroups or individuals within the general population;
28 and

 statutory requirements (and the related general agency missions) to be
protective of public health and the environment (e. g., to set standards
with ?an adequate margin of safety?).

However, agency risk assessors have considerable flexibility regarding
whether to use particular assumptions and methods, even when the agency has
default or standard options. For example, EPA stated that its revised
guidelines for carcinogen risk assessment were intended to be both explicit
and more flexible than in the past concerning the basis for making
departures from defaults, recognizing that expert judgment and peer review
are essential elements of the process. The Executive Director of ORD?s Risk
Assessment Forum pointed out that, although EPA?s guidelines always
permitted such flexibility, without detailed guidance on departing from
default assumptions there had been a tendency for analysts to not do so. He
also stated that when determining whether to use a default, the decision
maker must consider available information on an underlying scientific
process and agent- specific data, and that scientific peer review, peer
consultation workshops, and similar processes are the principal ways of
determining the strength of thinking and the general acceptance of these
views within the scientific community. FDA officials emphasized that their
agency does not presume that there is a ?best way? of doing a risk
assessment and does not require the use of a specific risk assessment
protocol or of specific default assumptions, but they are continually
updating procedures and techniques with the goal of using the ?best

available science.? 28 There is a conceptual difference between uncertainty
and variability. Uncertainty is a property of a lack of knowledge and may be
reduced through study and additional information. Variability is a property
of a system or population (e. g., every person has different physical
characteristics) and can only be understood, not reduced, through further
study. See Adam Finkel, ?A Second Opinion on an Environmental Misdiagnosis,?
New York University Environmental Law Journal, Volume 3 (1995), p. 299.

Agencies identified assumptions and methodological choices throughout the
risk assessment process, and each of the first three steps in the process
can have its own set of issues and choices that risk assessors need to
address. During hazard identification, agencies must make choices about

which types of data to use and what types of adverse effects and evidence
will be considered in their analyses. For example, risk assessors need to
decide whether data on benign tumors should be used along with data on
malignant tumors as the basis for quantitative estimates of cancer risks, or
whether only data on malignant tumors should be used. During doseresponse

assessment, agencies may need to make assumptions when extrapolating effects
from animals to humans (e. g., how to determine equivalent doses across
different species). In particular, choices among assumptions and methods are
needed when estimating dose- response relationships at doses that are much
lower than those used in the scientific studies that provided the data for
quantitative analysis. During exposure assessments, assumptions might be
needed to address issues such as when exposures occur (e. g., in infancy or
childhood versus as an adult), how long exposures last (e. g., short versus
long term and continuous versus episodic), differences in exposures and
effects for the population as a whole versus those affecting subpopulations
and individuals, and questions about the concentration and absorption of
chemical agents. Assumptions about human behavior also affect the relative
likelihood of different

exposure scenarios. For example, in assessing children?s residential
exposures to a pesticide, risk assessors might need to make assumptions
about how long children play in a treated area, the extent to which they are
wearing clothing, and potential hand- to- mouth exposure to treated soil,
among other factors.

Why Particular Assumptions Agencies generally indicated that they use their
major assumptions and

and Methods Are Selected methodological choices in risk assessments when
professional judgments

or policy choices must substitute for scientific information that is not
available or is inconclusive. We examined risk assessment guidance documents
and procedures in the four agencies to determine whether the agencies stated
a specific scientific or policy basis for their choices, as recommended by
NAS and the Presidential/ Congressional Commission. In

approximately three- quarters of the choices that we reviewed, the agencies
provided at least some rationale for the use of particular assumptions or
methods. The reasons most commonly cited were (1) an evaluation of available
scientific data, (2) the precedents established in prior risk assessments,
and (3) policy decisions related to their regulatory mandates. In some
instances, the agencies cited more than one reason in support of their
choices. For example, officials from FDA?s Center for Veterinary Medicine
said they assume that an adult weighs 60 kilograms when

converting an acceptable daily intake (ADI) to an intake level of residues
in food because of historical precedent and because this assumption should
protect women, growing adolescents, and the elderly. 29

29 According to FDA officials, if there is a need to convert an ADI
(expressed as milligrams per kilogram body weight per day) to an intake
level (expressed as the number of milligrams of an additive that would be
acceptable on a daily basis), they multiply the ADI by the assumed weight of
a person. Officials from FDA?s Center for Food Safety and Applied Nutrition
said they assume values of 60 kilograms for adults and 15 kilograms for
children, based on historical precedents which were based on population-
based surveys.

Of the three reasons, the agencies most often cited their evaluation of
available scientific evidence as a reason for selecting particular
assumptions or analytical methods. For example, one of the default
assumptions in EPA?s carcinogen risk assessment guidance is that positive
effects in animal cancer studies indicate that the agent under study can
have carcinogenic potential in humans. EPA cited scientific research
supporting this assumption, such as the evidence that nearly all agents

known to cause cancer in humans are carcinogenic in animals in tests with
adequate protocols. Other EPA guidelines stated that, in general, a
threshold is assumed for the dose- response curve for agents that produce
developmental toxicity. EPA?s guidelines noted that this assumption is based
on the known capacity of the developing organism to compensate for

or repair a certain amount of damage at the cellular, tissue, or organ
level. OSHA cited scientific evidence and the views of the Office of Science
and Technology Policy on chemical carcinogenesis (the origin or production
of a tumor) to support its choice to combine data on benign tumors with the
potential to progress to malignancies with data on malignant tumors
occurring in the same tissue and the same organ site. 30

30 The National Science and Technology Policy, Organization, and Priorities
Act of 1976 created the Office of Science and Technology Policy within the
Executive Office of the President to provide advice to the President on
issues relating to science and to coordinate federal efforts in science and
technology.

Even when basing a choice upon available scientific studies and data,
professional judgment may still be required regarding which particular
method or assumption to choose among competing alternatives. The

scientific evidence might show a range of assumptions or methods that
provide plausible results and may, in specific cases, vary in terms of which
one best fits the available evidence. For example, different mathematical
models can be used for estimating the low- dose effects of exposure to
suspected carcinogens. A basic problem for risk assessors is that, while

the results produced by different models may be similar at higher doses, the
estimates can vary dramatically at the low doses that are of concern to
agency regulators. One study of 5 dose- response models showed that all of
the models produced essentially the same dose- response curves at higher

doses, but the models? estimates differed by 3 or 4 orders of magnitude
(values 1, 000 to 10, 000 times different) at lower doses. 31 Because the
mechanism of carcinogenesis is not sufficiently understood, none of the
mathematical procedures for extrapolation has a fully adequate biological
basis. 32 Furthermore, because of the limitations in the ability of
toxicologic or epidemiologic studies to detect small responses at very low
doses, doseresponse relationships in the low- dose range are practically
unknowable.

Agencies can encounter similar problems in attempting to determine how much
of a chemical will produce the same effect in humans that was observed in
animals. An interagency group of federal scientists that studied this issue
noted that, although many alternatives had been

developed for such cross- species scaling, and despite considerable study
and debate, ?no alternative has emerged as clearly preferable, either on
empirical or theoretical grounds.? 33 The group noted further that the
various federal agencies conducting chemical risk assessments therefore
developed their own preferences and precedents, and this variation ?stands

among the chief causes of variation among estimates of a chemical?s
potential human risk, even when assessments are based on the same data.? For
purposes of consistency in federal risk assessments, the group

recommended a method intermediate between the two methods most 31 See
?Criteria for Evidence of Chemical Carcinogenicity,? Interdisciplinary Panel
on Carcinogenicity, Science 225 (1984), pp. 682- 687. 32 See, for example,
General Principles for Evaluating the Safety of Compounds Used in Food-
Producing Animals, DHHS/ FDA/ CVM (revised July 1994). 33 ?Draft Report: A
Cross- Species Scaling Factor for Carcinogen Risk Assessment Based on
Equivalence of mg/ kg 3/ 4 /Day,? 57 FR 24152 (June 5, 1992). No final
report has been issued.

commonly used by federal agencies, but reiterated that methodologies in use
?have not been shown to be in error.?

Other reasons cited by the agencies for selecting assumptions or methods
included the precedents established in prior risk assessments and policy
decisions related to their regulatory missions and mandates. For example,
FDA officials said that their practice of using the most sensitive species
and

sex when calculating the ADI of animal drug residues in food was based on
historical precedents dating back to at least 1954. In other instances, FDA
said that its use of precautionary assumptions was based on the agency?s

statutory responsibility to ensure to a ?reasonable certainty? that the
public will not be harmed. Similarly, EPA guidelines pointed out that the
default assumptions used in the agency?s risk assessments were chosen to be
health protective because EPA?s overall goal is public health protection.
For example, EPA?s neurotoxicity guidelines said that a choice to use the
most sensitive animal species to estimate human risk ?provides a
conservative estimate of sensitivity for added protection to the public.?

Effects of Agencies? The agencies provided information in their guidelines
on the likely effects Assumptions and Methods of using particular
assumptions or methods in about half of the examples on Risk Estimates

that we reviewed. When that information was provided, it was usually in the
context of whether and to what extent the agencies? choices could be
considered precautionary. In a number of cases, EPA and FDA characterized
their assumptions and methods as precautionary in that they were intended to
avoid underestimating risks in the interest of protecting

public health. Such assumptions tend to raise an agency?s estimate of risk
and lower the levels of exposure that are of regulatory concern.
Precautionary assumptions and methodological choices were a common component
of programs that have ?tiered? approaches for conducting risk assessments
(e. g., EPA?s Superfund and pesticides programs). In these tiered risk
assessment approaches, agencies move from initial rough

screening efforts to increasingly more refined and detailed levels of
analyses. The initial screening assessments will typically involve very
precautionary ?upper- bound? or even ?worst- case? assumptions to determine
whether there is cause for concern. Successive tiers of assessment, if
deemed necessary, are characterized in agency documents as more detailed and
focused assessments that require more extensive data and rigorous analysis.
For example, EPA indicated that its screening assessments might well use
precautionary upper- bound point estimates of exposures (e. g., that a
chemical is used on 100 percent of the eligible crop and at the maximum
permissible limit). However, subsequent tiers of

assessments might refine those estimates through the use of probability
distributions of exposure parameters or the use of monitoring data on actual
exposures, when feasible. OSHA and RSPA also use precautionary assumptions
in certain parts of their risk assessment procedures. However, these
agencies identified few of their risk assessment assumptions and methods as
precautionary. In fact, OSHA sometimes selected assumptions or methods that
it explicitly characterized as less precautionary than those used by other
agencies in similar circumstances. For example, OSHA stated that its
standard

approach to low- dose extrapolation can be much less precautionary than
EPA?s or FDA?s approaches because it tends to use central estimates of
potency rather than upper- bound confidence limits. OSHA officials also
noted that the algorithm they use is less precautionary because it may lead
to models that are sublinear at low doses.

The effect on risk estimates of using any one assumption is likely to be
less significant than that of applying a series of assumptions while
conducting a risk assessment, particularly if the assessment is compounding
a string of largely precautionary assumptions. As we previously pointed out,

assumptions and choices may be needed at many points during each step of an
agency?s analysis. The agency?s policy may well be to use precautionary
choices at most, if not all, of those points, if adequate information is not

available to indicate that the precautionary choice is invalid in a specific
case. The potential for such a string of precautionary assumptions is
illustrated by the set of standard choices identified by FDA for risk

assessments of carcinogenic animal drug residues in foods consumed by
humans.

1. Regulation is based on the target tissue site exhibiting the highest
potential for cancer risk for each carcinogenic compound.

2. If tumors are produced at more than one tissue site, the minimum
concentration of the compound that produced a tumor is used.

3. Cancer risk estimates are generally based on animal bioassays, using
upper 95- percent confidence limits of carcinogenic potency. 34

4. Low- dose extrapolation is done using a nonthreshold, conservative,
linear- at- low- dose procedure (i. e., assuming that there is no dose that
would not cause cancer and that effects vary in proportion to the

amount of the dose). 5. It is assumed that the carcinogenic potency in
humans is the same as

that in animals. 6. The concentration of the residue in the edible product
is at the permitted concentration.

7. Consumption is equal to that of the 90 th percentile consumer. 8. All
marketed animals are treated with the carcinogen. 9. In the absence of
information about the composition of the total

residue in edible tissue, assume that the entire residue is of carcinogenic
concern.

FDA?s description of its risk assessment procedures acknowledged that these
assumptions ?result in multiple conservatisms? and stated that some of these
choices are likely to overestimate risk by an unknown amount (although the
fourth assumption could also underestimate risk by an order of magnitude).
However, the agency also said that these assumptions are prudent because of
the uncertainties involved and cited its statutory

responsibility to ensure to a reasonable certainty that the public will not
be harmed. It is important to keep in mind that the primary purposes for
preparing such assessments are to identify safe concentration levels in
edible tissues and residue tolerances (the amount permitted to remain on

food) for postmarket monitoring rather than to produce a general estimate of
the risk posed by use of the animal drug. 34 Bioassay refers to the use of
living organisms to measure the effect of a risk agent or condition- for
example, a test for carcinogenicity in laboratory animals that includes
nearlifelong exposure to the agent being tested.

Comparison of Agencies? Agency documents very rarely made direct comparisons
of their Assumptions

assumptions and methodological choices to those used by other agencies, and
there is no requirement that they do so. Our review indicated that EPA, FDA,
and OSHA risk assessment procedures have many basic assumptions in common-
for example, that one can use results of animal experiments to estimate
risks to humans, and that most potential carcinogens do not have threshold
doses below which adverse effects would not occur. There are other default
or standard assumptions and models in the three

agencies? risk assessment procedures that are similar, but not identical.
For example, all three agencies employ a linear mathematical model for low-
dose extrapolation (in the absence of information indicating that a

linear model is inappropriate in a particular case). However, the agencies
prefer different options in the details of fitting such models, such as the
point of departure to low doses. EPA and FDA also consider similar, but not
identical, sets of uncertainty or safety factors when using the NOAEL

approach for noncancer risk assessments. Finally, as the discussion above
regarding low- dose extrapolation illustrates, there are also instances in
which the agencies use different assumptions in similar circumstances.

Table 1 compares and contrasts some of the risk assessment assumptions or
analytical methods identified in the guidelines or other descriptive
documents of EPA, FDA, and OSHA for use under similar circumstances. 35
(Note that, for comparability, the examples in table 1 all focus on

carcinogen risk assessments based on animal studies, but the agencies? major
assumptions and methods are not limited to only carcinogen risk assessments.
Note also that the ?circumstances? listed in the table also include that the
assumption or method would be used in the absence of data to the contrary.)

35 RSPA is not included in this table because it uses a different process
for risk assessments, and its assessments do not focus on the carcinogen
risk assessment issues highlighted in the table. However, RSPA?s risk
assessment methods are similar to EPA?s Chemical Emergency Preparedness and
Prevention Office, which also focuses primarily on short- term risks
associated with accidental releases of chemicals.

Table 1: Comparison of Selected Major Assumptions or Methods Used in EPA,
FDA, and OSHA Risk Assessments Circumstance EPA FDA OSHA

Which species/ sex to use in Use most sensitive species/ sex. Use most
sensitive species/ sex. Use most sensitive species/ sex

animal studies. for tumor sites appropriate for routes of exposure
experienced by workers.

Whether to include data on Include data on benign tumors if Include data on
benign tumors if Combine data on benign tumors benign tumors in a cancer
they have the capacity to

they have the capacity to with the potential to progress to assessment.

progress to the malignancies progress to the malignancies malignancies with
data on with which they are associated. with which they are associated.

malignant tumors occurring in Benign tumors that are not Benign tumors that
are not the same tissue and the same observed to progress to observed to
progress to organ site. malignancy are assessed on a

malignancy are assessed on a case- by- case basis. case- by- case basis.

Preferred cancer low- dose Default depends on the agent?s Use a no-
threshold, linear

Use a particular no- threshold, extrapolation method if a

mode of action. For example, if extrapolation method. linear approach known
as the

mathematical model is used. mode is not understood or ?maximum likelihood
estimate in

believed to be linear, a linear the Crump- Howe approach is recommended.

reparameterization of the When data supports a nonlinear multistage model.?
(This mode of action, the default particular approach may lead to changes to
a margin of exposure models that are sublinear at low analysis. doses.)

Point of departure in preferred Use the lower 95- percent limit of

Use data at the upper 95- According to agency officials, approach to low-
dose

the doses that are estimated to percent confidence limit. OSHA does not use
a point of extrapolation (i. e., the data point

cause a 10- percent response departure. OSHA tends to use from which the
agency (i. e., an effect in 10 percent of central estimates of potency,
extrapolates to lower, exposed subjects). (This dose is such as the maximum
likelihood

unobserved dose- response referred to as the LED 10 .)

estimate of the parameterized relationships). dose- response function.

Preferred method for cross For oral exposure, recommends Recommends use of
scaling

Assumes that equivalent doses species scaling of equivalent use of a scaling
factor of body factor of body weight to the ï¿½ scaled by body weight would
doses (i. e., how the agency weight to the ï¿½ power.

power. (However, risk assessors lead to equivalent risks. converts data from
doses given may also use the default of body

(However, it may in the future to one species, such as rats in a

weight scaling.) move to ï¿½ -power scaling.)

toxicological study, to doses presumed to have an equivalent effect on
another species, such as humans). Source: GAO analysis

There appears to be some convergence in the agencies? risk assessment
assumptions in at least one area where there had been significant
differences- their methods for cross- species dose scaling. In the absence
of adequate information on differences between species, EPA?s standard
practice in carcinogenic risk assessments had been to scale daily
administered doses by body surface area, whereas FDA?s and OSHA?s standard
practice had been to scale doses by body weight. Recently, the agencies have
either adopted, or consider as one of their options, the expression of doses
in terms of daily amount administered per unit of body weight to the ï¿½
power. 36

All four of the agencies included in our review have also been incorporating
more complex analytical methods and models into their risk assessment
procedures. Some of these methods (such as the use of probabilistic analyses
to provide distributions of exposure parameters) help to address

issues of uncertainty and variability in risk assessments and lessen the
need for some precautionary assumptions. Other advances, such as the use of
PBPK models, can provide better insights into how and to what extent a
chemical might produce adverse effects in humans. One outcome of the

integration of these methods into agencies? procedures is a diminishing of
the traditional distinction between cancer and noncancer risk assessment
methods. EPA, in particular, has noted that it is less likely to consider
cancer and noncancer endpoints in isolation as it develops and incorporates
more advanced scientific methods to measure and model the biological events
leading to adverse effects. According to EPA, the science of risk assessment
is moving toward a harmonization of the methodology

for cancer and noncancer assessments. The use of newer, more complex models
and methods also opens up a new range of choices and assumptions in the
analysis- along with the potential for risk estimates to diverge because of
the different assumptions that might be used. For example, in its methylene
chloride final rule OSHA

reported on the results of its analyses as well as risk assessments
submitted to OSHA by other risk assessors. 37 Although most of the risk
assessments used a linearized multistage model to predict risk, there were
differences in the estimates produced by these assessments. OSHA 36 Across
the range of plausible values, the body weight approach is generally
considered the least precautionary, surface area scaling the most
precautionary, and (body weight) 3/ 4 the midpoint value.

37 ?Occupational Exposure to Methylene Chloride,? 62 FR 1494 (Jan. 10,
1997).

pointed out that the differences in risk estimates were not generally due to
the dose- response model used, but to whether the risk assessor used PBPK
modeling to estimate target tissue doses and what assumptions were used in
the PBPK modeling.

Appendices II through V present more detailed information on some of the
major assumptions and methodological choices in each of the four selected
agencies.

Risk Characterization In the risk characterization step of a risk
assessment, agencies bring

Policies and Practices together the results of the preceding analyses in the
form of estimates and

conclusions about the nature and magnitude of a potential risk. Agencies?
Emphasize risk characterizations play a crucial role in explaining to
decision makers Transparency and other interested parties what the agency?s
risk assessors have concluded and on what basis they reached those
conclusions. Both EPA

and DOT have agencywide written policies on risk characterization that
emphasize the importance of providing comprehensive and transparent
characterizations of risk assessment results. Although FDA and OSHA do

not have written risk characterization policies, officials of those agencies
pointed out that, in practice, they also tend to emphasize comprehensive
characterizations of risk assessment results, discussions of limitations and
uncertainties, and disclosure of the data and analytic methodologies on
which the agencies relied.

EPA?s program offices are generally responsible for completing risk
characterizations, and EPA?s agencywide guidance on this issue includes a
risk characterization policy, a guidance memorandum, and a handbook. EPA?s
policy stipulates that risks should be characterized in a manner that is
clear, transparent, reasonable, and consistent with other risk

characterizations of similar scope. EPA said that all assessments ?should

identify and discuss all the major issues associated with determining the
nature and extent of the risk and provide commentary on any constraints
limiting fuller exposition.? EPA?s policy documents also recommend that risk
characterization should (1) bridge the gap between risk assessment and risk
management decisions; (2) discuss confidence and uncertainties involving
scientific concepts, data, and methods; and (3) present several types of
risk information (e. g., a range of exposures and multiple risk descriptors
such as high- end estimates and central tendencies). It is also

EPA?s policy that major scientifically and technically based work products
related to the agency?s decisions normally should be peer- reviewed. 38 In
its guidelines for carcinogen risk assessment, EPA also suggests preparing
separate ?technical? characterizations to summarize the findings of the
hazard identification, dose- response assessment, and exposure assessment
steps. The agency?s risk assessors are then to use these technical
characterizations to develop an integrative analysis of the whole

risk case, followed by a less extensive and nontechnical summary intended to
inform the risk manager and other interested readers. EPA identified several
reasons for preparing separate characterizations of each analysis phase
before preparing the final integrative summary. One is that different

people often do the analytical assessments and the integrative analysis. The
second is that there is very often a lapse of time between the conduct of
hazard and dose- response analyses and the conduct of the exposure
assessment and integrative analysis. Thus, according to EPA, it is necessary
to capture characterizations of assessments as the assessments

are done to avoid the need to go back and reconstruct them. Finally, several
programs frequently use a single hazard assessment for different exposure
scenarios. 38 Peer review generally takes one of two forms: (1) internal
peer review by a team of relevant experts from within EPA who have no other
involvement with respect to the work product that is to be evaluated or (2)
external peer review by a review team that consists primarily of independent
experts from outside EPA.

DOT?s policy principles regarding how the results of its risk or safety
assessments should be presented are straightforward and encourage agency
personnel to:

 make public the data and analytic methods on which the agency relied (for
replication and comment);  state explicitly the scientific basis for
significant assumptions, models,

and inferences underlying the risk assessment, and explain the rationale for
these judgments and their influence on the risk assessment;

 provide the range and distribution of risks for both the full population
at risk and for highly exposed or sensitive subpopulations and encompass all
appropriate risk to health, safety, and the environment;

 place the nature and magnitude of risks being analyzed in context
(including appropriate comparisons to other risks); and

 use peer review for issues with significant scientific dispute. FDA does
not have a written risk characterization policy, but FDA officials said
that, in practice, the agency uses a standard approach that is similar to
EPA?s official policy. They said that FDA?s general policy is to reveal the
risk assessment assumptions that have the greatest impact on the results of
the analysis, and to state whether the assumptions used in the assessment
were conservative. FDA officials also said that their risk assessors attempt
to show the implications of different distributions and choices (e. g., the
results expected at different levels of regulatory intervention). FDA may

employ probabilistic methods, such as Monte Carlo analysis, to provide
additional information on the effects of variability and uncertainty on
estimates of risk, and there are some differences in FDA risk
characterization procedures depending on the products being regulated and
the nature of the risks involved. 39

Although OSHA does not have written risk characterization policies, in
recent rules the agency emphasized (1) comprehensive characterizations of
risk assessment results; (2) discussions of assumptions, limitations, and

uncertainties; and (3) disclosure of the data and analytic methodologies on
which the agency relied. The agency devoted considerable effort to
addressing uncertainty and variability in its risk estimates. Such efforts
39 Monte Carlo analysis involves a repeated random sampling from the
distribution of values for each of the parameters in a calculation (such as
average daily exposure) to derive a distribution of estimates of exposures
for a population. According to FDA, because Monte Carlo modeling is a
probabilistic technique that can use all the available data, it will result
in more accurate estimates at upper percentiles of exposure.

included performing sensitivity analyses and providing estimates produced by
alternative analyses and assumptions (including analyses by risk assessors
outside of OSHA). In its risk characterizations, OSHA provided

both estimates of central tendency and upper limits (such as the 95 th
percentile of a distribution). Appendices II through V provide more detailed
descriptions of the risk characterization policies or approaches of each of
the four selected agencies.

Conclusions Risk assessment is an important, but extraordinarily complex,
element in federal agencies? regulation of potential risks associated with
chemicals. The assessments can help agencies decide whether to regulate a
particular chemical, select regulatory options, and estimate the benefits
associated with regulatory decisions. Scientific studies in such areas as
toxicology

and epidemiology are often used to produce the information needed for risk
assessment decisions. However, assessors frequently must produce estimates
of risk without complete scientific information about the extent of
exposures to potentially hazardous substances and the effects of those
exposures on human health and safety or the environment. Therefore,
professional judgment with regard to assumptions and methodological choices
is an inherent part of conducting risk assessments. The

appendices to this report identify many of the major assumptions and methods
that can be used in risk assessments prepared for EPA, FDA, OSHA, and RSPA.
The number and variety of those assumptions and methods illustrate the range
of issues that risk assessors confront during the course of their analyses.

Although there were more similarities than differences in the general risk
assessment procedures of three of the four agencies, there were also some
notable differences in the agencies? specific approaches, methods, and
assumptions. These differences can significantly affect the results and
conclusions drawn from the assessments. Therefore, risk estimates prepared
by different agencies, or by different program offices within those
agencies, may not be directly comparable, even if the same chemical agent is
the subject of the risk assessment. In some cases, the reasons for those

differences are readily apparent, such as when agencies focus on different
types of adverse effects (e. g., cancer versus noncancer) or different types
and sources of exposure. For example, the same chemical (e. g., methylene
chloride) might be identified as a significant hazard by one agency in one
exposure setting (OSHA for occupational exposures) but as a low hazard

by another agency in a different setting (EPA for Superfund hazard ranking
screening). In other cases, the reasons for different estimates may be more
subtle and harder to discern within the many layers of analyses and
professional judgments used to prepare the risk assessment. Because of the
range of assumptions and methods that are scientifically plausible in a
given situation, the risk characterization phase of the risk

assessment process takes on added importance. In their risk characterization
policies or procedures, the four agencies acknowledge the importance of
clearly communicating not only their conclusions about the nature and
likelihood of a given risk but also disclosing (1) the assumptions, methods,
data, and other choices that had the greatest impact on risk estimates; (2)
why those choices were made; and (3) the effect that alternative choices
would have had on the results of a risk assessment.

Transparency is important with regard to both individual risk assessments
and in agencies? general procedures regarding how the assessments should be
conducted. Those procedures encourage consistency in how agencies conduct
risk assessments and provide insights into agencies? decision making when
analyzing risks. For example, frameworks delineated by EPA

and OSHA for departing from certain default assumptions inform both agency
personnel and external parties as to whether particular data or analyses are
acceptable to the agency. Our review focused on describing the framework for
agencies? chemical risk assessments. We did not evaluate how that framework
is applied in practice, or how risk assessment results affect risk
management decisions by agencies and other policymakers. Nevertheless, our
report highlights the value of policymakers and other interested parties
becoming aware of the underlying risk assessment context, procedures,
assumptions, and

policies when using risk assessment data for risk management and other
public policy decisions. For example, prudent use of risk data requires the
user to be aware of the extent to which the data:

 represent estimates from screening assessments (which may rely heavily on
precautionary assumptions) or estimates from subsequent, more rigorous
assessments (which are likely to rely on more detailed and case- specific
data and analyses);

 show the distribution of exposures and potential adverse effects across
the population, including the extent to which the data address risks of the
most exposed or sensitive subgroups of the population, or focus on only part
of that distribution;

 were produced using directly relevant scientific data that were available
or had to rely on general assumptions and models; and

 reflect the flexibility permitted in agencies? standard procedures or
guidelines to depart from past precedent and default choices to use
alternative assumptions and models, when appropriate.

In our review we also found that, although the underlying statutes specified
the use of particular methods or assumptions in only three instances, the
legal and situational context within which an agency is conducting a
chemical risk assessment has a major effect on the specific focus, scope,
and level of detail of the resulting assessment. Comparison of risk
assessment estimates from different agencies and programs therefore

requires careful consideration of these contextual differences. Because the
central purpose of our review was to describe the framework for selected
agencies? chemical risk assessments, rather than to evaluate and critique
how that framework is applied in practice, we are not making any
recommendations in this report.

Agency Comments and At the end of our review, we sent a draft of this report
to five experts in the

Our Evaluation field of risk assessment to ensure the technical accuracy of
the report. The

three experts who provided comments were (1) the Executive Director of the
Presidential/ Congressional Commission, (2) the individual who prepared the
Survey of Methods for Chemical Health Risk Assessment Among Federal
Regulatory Agencies for the Commission, and (3) an expert in risk assessment
at Resources for the Future. The experts generally indicated that the report
had no material weaknesses, but provided a number of technical suggestions
that we incorporated as appropriate. For example, two of the reviewers
suggested that the report?s

discussion of the NAS four- step risk assessment paradigm, although
reflecting the definitions generally relied upon by federal agencies, should
also identify an updated view regarding the concept of risk
characterization. The updated view is that risk characterization should be a
decision- driven activity performed as part of the risk management

decision making process rather than a stand- alone activity at the end of a
risk assessment. We included this perspective in the report?s background
section. During our review, we obtained technical comments from officials in
each of the four agencies on a draft of the appendices to this report, which
we incorporated as appropriate. On June 18, 2001, we sent a draft of the
full

report to the Secretaries of Health and Human Services, Labor, and
Transportation, and the Administrator of EPA for their review and comment.
None of the agencies provided formal comments on the report, but we received
additional technical comments and suggestions from all

four of the agencies, which we incorporated as appropriate. As arranged with
your office, unless you publicly announce the contents of this report
earlier, we plan no further distribution until 30 days after the date of
this report. At that time, we will send copies of this report to the Ranking
Minority Member, House Committee on Energy and Commerce; the Ranking
Minority Member, Subcommittee on Environment and Hazardous Materials, House
Committee on Energy and Commerce; the Secretaries of Health and Human
Services, Labor, and Transportation; and the Administrator of EPA. We will
also make copies available to others on request.

If you have any questions concerning this report, please call me or Curtis
Copeland at (202) 512- 6806. Key contributors to this assignment were
Timothy Bober and Aaron Shiffrin. Victor S. Rezendes Managing Director,
Strategic Issues

Appendi Appendi xes x I

Objectives, Scope, and Methodology Scope and Objectives As requested, our
review focused on the chemical risk assessment procedures, assumptions, and
policies of four federal agencies with responsibilities for regulating or
managing risks from potential exposure to chemicals- the Environmental
Protection Agency (EPA), the Food and Drug Administration (FDA) within the
Department of Health and Human Services (HHS), the Occupational Safety and
Health Administration (OSHA) within the Department of Labor, and the
Department of Transportation?s (DOT) Research and Special Programs
Administration (RSPA- in particular the Office of Hazardous Materials
Safety). Our specific objectives were to identify and describe (1) the
general context for the agencies? chemical risk assessment activities; (2)
what the agencies view as their primary procedures for conducting risk
assessments; (3) what the agencies view as the major assumptions or
methodological choices in their risk assessment procedures; and (4) the
agencies? procedures or policies for characterizing the results of risk
assessments. To the extent feasible, we were also asked to identify for the
assumptions and choices identified in the third objective (a) at what stages
of the risk assessment process they are used, (b) the reasons given for
their selection, (c) their

likely effects on risk assessment results, and (d) how they compare to the
assumptions and choices used by other agencies or programs in similar
circumstances.

Methodology To address our objectives, we relied primarily on a detailed
review and analysis of agencies? general guidance documents on chemical risk

assessment or, if there were no guidance documents, reviews of specific
examples of agency risk assessments. We supplemented that information with
material from secondary source reports on risk assessment and interviews
with agency officials. Among the secondary sources that we used were
relevant reports by the Congressional Research Service, National Academy of
Sciences (NAS), and the Presidential/ Congressional Commission on Risk
Assessment and Risk Management (hereinafter referred to as the Presidential/
Congressional Commission). In particular,

as a starting point for our review we used a report on federal agencies?
chemical risk assessment methods that was prepared by Lorenz Rhomberg for
the Presidential/ Congressional Commission. 1 That report provided the
baseline descriptions of some of the chemical risk assessment procedures at
EPA, FDA, and OSHA. 2 We asked officials of those agencies to review
Rhomberg?s report to identify information that was still relevant to
addressing the objectives of this report as well as information that they
felt should be revised or added to reflect the agencies? current procedures.

There are several important limitations to our review. First, chemical risk
assessment is just one of several types of risk assessment being conducted
in federal agencies. Therefore, our review cannot be used to characterize
other types of risk assessments (e. g., risks associated with radiation
exposure). In fact, FDA officials considered risk assessments related to the
human drug approval process to be outside the scope of our review because a
completely different protocol is used in those assessments. However,
limiting the scope of our review to chemical risk assessments

makes comparisons among the agencies included more relevant and meaningful.
Second, our review did not include all agencies or programs that conduct
risk assessments involving chemicals. For example, we did not include the
Consumer Product Safety Commission, which periodically assesses products
with potential risks from chemicals. Nor did we include

the Agency for Toxic Substances and Disease Registry, which prepares

?health assessments? that closely resemble risk assessments but has no
regulatory authority. We focused on the risk assessment procedures in four
federal agencies that regularly conduct chemical risk assessments in

1 A Survey of Methods for Chemical Health Risk Assessment Among Federal
Regulatory Agencies, Lorenz Rhomberg (1996). 2 RSPA was not included in the
scope of Rhomberg?s report.

support of regulatory activities and/ or could illustrate the diversity of
risk assessment procedures. However, the results of our review cannot be
considered representative of chemical risk assessments in all federal
agencies. Third, our review does not describe every chemical risk assessment
procedure or assumption used by the agencies we reviewed.

The material describing the agencies? procedures is both voluminous and
extremely complex. The detailed information that we provide on agency
assumptions is illustrative of the assumptions included in agencies?
procedures, but not a compendium of all such assumptions. In addition,

we concentrated primarily on the human health and safety risk assessment
procedures of the four agencies and, to a lesser extent, on ecological risk
assessment procedures. Fourth, this report describes agencies? general
procedures and policies, but it is not a compliance review of how well those
procedures and policies are applied with regard to individual

assessments. The agencies? guidelines represent suggested procedures and are
not binding, so the agencies? practices may justifiably vary from the
general frameworks we describe. In practice, risk assessments do not

follow a simple recipe or formula. Each assessment has unique issues or
characteristics that require case- specific resolutions. Finally, this
report does not address risk management issues- e. g., using the results of
a risk assessment to determine what level of exposure to a risk agent
represents an acceptable or an unacceptable risk and deciding what control
options should be used.

We conducted this review between February 2000 and March 2001 in the
Washington, D. C., headquarters offices of the selected agencies in
accordance with generally accepted government auditing standards. We
obtained technical comments on our descriptions of the agencies? procedures,
assumptions, and policies in the appendices from

knowledgeable agency personnel. We then provided the draft report to
external experts in risk assessment, including the Center for Risk Analysis
at the Harvard School of Public Health in Boston, MA; Resources for the
Future in Washington, D. C.; the Executive Director of the Presidential/
Congressional Commission; and Lorenz Rhomberg, the analyst who surveyed
federal agencies? chemical risk assessment procedures for the Commission.
After incorporating their comments, we provided a draft of this report to
the Secretaries of Health and Human Services, Labor, and Transportation; and
the Administrator of the Environmental Protection Agency for their review
and comment.

Organization of In the following appendices, we provide more detailed
information Appendices on

regarding the framework and methods applicable to chemical risk assessment
activities of EPA, FDA, OSHA, and RSPA. There is a separate Chemical Risk

technical appendix covering each of these four agencies, along with their
Assessment at Selected relevant offices, programs, or centers that are
involved in conducting Federal Agencies

chemical risk assessments. For consistency and ease of presentation, we have
generally organized the appendices on each agency according to a standard
format with four major sections.

1. We describe the general context for the chemical risk assessment
activities of each agency. This includes a summary of the primary risk
statutes, mandates, and tasks related to potential risks from exposure to
chemical agents.

2. We identify and summarize the standard risk assessment procedures of each
agency and, if applicable, each agency?s various offices, programs, or
centers. This section is generally organized by the major analytical steps
of the risk assessment process: hazard identification, doseresponse
assessment, and exposure assessment. These correspond to the first three
steps of the four- step paradigm for risk assessment as defined by NAS and
used by three of the four agencies covered by our review. (We address the
fourth step of the process, risk characterization, as a separate objective
in the final section of each

agency appendix.) Within the descriptions of those steps, we often
distinguish between the procedures used for assessing cancer and noncancer
effects. Given developments in risk assessment methods, these distinctions
are sometimes more artificial than real. 3

3. We present additional information about major assumptions and
methodological choices in the agencies? standard risk assessment procedures.
For EPA, FDA, and OSHA, the primary focus of this

section is a detailed table identifying some of the major agencywide or
program- specific assumptions that may be used in chemical risk assessments.
To the extent that such information was available, each 3 For example,
agencies may omit or combine some of the steps in certain situations. Also,
with increasing research attention on the modes of action of chemical
agents, the line between procedures for identifying carcinogenic and
noncancer effects is blurring.

of these tables also includes information on the agency?s reason( s) for
selecting a particular assumption, when in the risk assessment process the
agency would apply the assumption, and the likely effect of using the
assumption on risk assessment results. (Because agencies very rarely made
direct comparisons of their choices to those of other agencies in their risk
assessment guidelines or related documents, we have not included a separate
column on that topic in the appendix

tables. That objective is, however, addressed in the letter portion of this
report.) The appendix on RSPA does not include all of these elements because
of differences in its context and approach to chemical risk assessment.

4. The final section of each appendix addresses each agency?s approach or
policies for characterizing the results of risk assessments for agency
decision makers and other interested parties. In particular, we describe

the agency?s policies or practices with regard to the transparency of risk
assessment results, such as reporting the range and distribution of risks
and identifying the uncertainties in the risk analysis and underlying data.

To avoid repetition in the appendices on agencies? risk assessment
procedures, our most detailed descriptions of basic methods and issues
appear in the EPA appendix under the discussion of agencywide procedures.
Descriptions of procedures used by other agencies or programs, including the
individual program offices within EPA, then reference the EPA- wide
descriptions of those particular methods, if they are similar. Although we
provide much more detailed technical information in these

appendices than in the main body of the report, it is still important to
recognize that agencies? risk assessment methods are more involved and
complex than we have described in this report. In particular, the tables of
assumptions do not represent a comprehensive listing of all assumptions and
choices of the agencies. Agencies might use many different types and numbers
of assumptions in any given assessment, and the assumptions are being
altered over time to reflect scientific improvements and changes in risk
approaches and the regulatory context. However, the information

presented is intended to illustrate the types and diversity of procedures
and assumptions employed by the agencies we examined.

Chemical Risk Assessment at the

Appendi x II

Environmental Protection Agency Chemical risk assessment at the
Environmental Protection Agency (EPA) is a complex and diverse undertaking.
The variety and range of the relevant regulatory authorities and activities
has a major effect on the organization and conduct of risk assessment at the
agency. An expanding set of agency guidelines reflects the evolving nature
of EPA?s risk assessment procedures. EPA generally follows the four- step
risk assessment process identified by the National Academy of Sciences
(NAS). Changes are occurring in EPA?s approaches to cancer, noncancer, and
exposure assessments, with a general trend toward the development and
application of more complex and comprehensive methodologies. To a greater
extent than the other agencies we reviewed, EPA has established a set of
default assumptions (often precautionary in nature) and standard data
factors for use by its risk assessors. In the ?tiered? risk assessment
approaches commonly employed by EPA?s program offices, precautionary default
assumptions are most often used during initial screening assessments, when
the primary task generally is to determine whether a risk might exist

and more rigorous analysis is needed. However, the information necessary for
more detailed analysis is not always available, so for regulatory purposes
the agency may be limited to using results from its initial tiers of

risk assessments. In presenting the results of its risk assessments, it is
EPA?s policy that risk characterizations should be prepared in a manner that
is clear, transparent, reasonable, and consistent with other risk
characterizations of similar scope prepared across the programs in the
agency. The following sections describe for EPA and its component offices,
the context for chemical risk assessment, the general procedures for
conducting risk assessments, major assumptions and methodological choices in
those procedures, and the agency?s policy for risk characterization. Because
chemical risk assessment at EPA is such a complex and diverse activity, this
appendix can only summarize and

illustrate the range of contexts, procedures, assumptions and methods, and
policies that affect the conduct of EPA risk assessments. For example, as in
our report as a whole, this appendix focuses primarily on human health and
safety risk assessment and less on ecological risk assessment.

However, we have included a brief section on EPA?s ecological risk
assessment guidelines under our discussion of agencywide risk assessment
procedures and illustrated the role played by ecological risk assessment in
the risk assessment activities of some, but not all, of EPA?s program
offices under our discussion of program- specific procedures. As a practical
matter, this appendix reflects risk assessment topics that were addressed in
agencywide or program- specific guidelines or descriptions of chemical

risk assessment at EPA. To the extent that such activities were not
explicitly addressed in the agency?s risk assessment guidelines and related
documents, there may be little information on them in this appendix.

Context for EPA EPA is responsible for a wide range of regulatory- and
related risk

Chemical Risk assessment- activities pertaining to potential health, safety,
and

environmental risks associated with chemical agents. This range of
Assessment activities reflects an equally broad and diverse range of
underlying environmental statutes. According to EPA, close to 30 provisions
within the major environmental statutes require decisions based on risk,
hazard, or exposure assessment, with varying requirements regarding the
scope and depth of the agency?s analyses. In general, EPA?s regulatory
authority

regarding chemical agents is compartmentalized according to the various
kinds and sources of exposure- such as pesticides, drinking water systems,
or air- borne pollutants- and reflected in the agency?s organization

into various program offices- such as the Office of Air and Radiation,
Office of Solid Waste, and Office of Water. Table 2 summarizes the principal
statutes, regulatory tasks, and risk mandates associated with chemical risk
assessment activities of EPA?s offices.

Table 2: Chemical Risk Statutes, Tasks, and Mandates for EPA Offices Office
or program Major risk- related statute( s) Primary risk- related tasks
Primary risk- related mandate( s)

Chemical Emergency Emergency Planning and Under EPCRA, evaluates Under
EPCRA, the EPA list of

Preparedness and Community Right- to- Know Act substances for toxicity,
reactivity,

extremely hazardous substances Prevention Office (CEPPO) (EPCRA) in the
Superfund

volatility, dispersability, and their thresholds, along with Amendments and
combustibility, or flammability to reporting requirements, are used by
Reauthorization Act of 1986 develop and maintain a list of state and local
entities to manage

(SARA), Title III extremely hazardous substances the risks associated with
chemical and threshold quantities. Also

emergencies at the local level. Clean Air Act Amendments of develops
regulatory requirements 1990 (CAAA)

for reporting accidental releases Under Section 112( r) of the and for
emergency planning. amended CAA, EPA must develop a list of at least 100
substances that Under Section 112( r) of the

pose the greatest risk of causing amended Clean Air Act (CAA),

death, injury, or serious adverse evaluates substances for acute effects to
human health or the adverse health effects, likelihood environment from
accidental of accidental release, and releases. EPA must also develop
magnitude of exposure to regulations for preparation and

develop a list of substances for submission of risk management

prevention of accidental release. programs and plans by industrial Evaluates
accidental chemical facilities handling these listed release risk management
and

substances. prevention practices for development of accidentprevention Also
under the amended CAA, for

and risk- reduction clean air research, EPA shall develop

regulations at industrial facilities. methods and techniques necessary to
identify and assess the risks to Also under the CAAA,

human health from accidental investigates chemical incidents, exposures.
evaluates the risks associated with accidental releases, and conducts
research on risk analysis and assessment.

(Continued From Previous Page)

Office or program Major risk- related statute( s) Primary risk- related
tasks Primary risk- related mandate( s)

Office of Air and Radiation Clean Air Act; Clean Air Act

Regulation of emissions of airborne For criteria air pollutants, set (OAR
air quality side) Amendments of 1990 pollutants, including

standards to protect public health

 setting national ambient air with an adequate margin of safety.

quality standards (NAAQS) for six ?criteria? pollutants, and

For hazardous air pollutants, set  setting standards for regulating
standards using maximum emissions of hazardous air achievable control
technology pollutants (toxic chemicals other (MACT) for a specified list of
than the criteria pollutants).

chemicals identified in the CAAA. However, if applying the MACT is Under the
CAA, EPA is also found to lower risks insufficiently, required to review the
scientific

EPA may pursue further regulation to data upon which air quality control the
residual risk, applying the standards are based and revise pre- CAAA
standard of providing an the standards, if necessary, every ample margin of
safety. 5 years. Office of Emergency and

Comprehensive Environmental Remediation of hazardous waste Remedial actions
are authorized at a

Remedial Response Response, Compensation, and sites, including

site whenever any hazardous (OERR Superfund)

Liability Act of 1980 (CERCLA);

 defining hazardous substances substance is released or there is a
Superfund Amendments and and the amounts of release that substantial threat
of such a release Reauthorization Act of 1986

must be reported to EPA, into the environment, or when there

 screening and ranking risks is a release or substantial threat of posed by
hazardous waste sites release into the environment of any and identifying
action priorities pollutant or contaminant which may among them,

present an imminent and substantial

 evaluating need for action at danger to the public health or hazardous
waste sites, and

welfare.

 evaluating effectiveness of options for remediation.

Remedial action priorities are to be based on relative risk or danger to
public health or welfare or the environment, taking into account the
population at risk, the hazard potential of the hazardous substances, and
the potential for contamination of air and drinking water, among other
factors. Need for

action is determined by evaluation of risks to human health and the
environment; effectiveness is determined by meeting requirements of other
laws or risk- based goals.

(Continued From Previous Page)

Office or program Major risk- related statute( s) Primary risk- related
tasks Primary risk- related mandate( s)

Office of Pesticide Federal Insecticide, Fungicide,

Regulation of pesticides, For pesticide residues in all foods, Programs
(OPP) and Rodenticide Act (FIFRA); including

determine whether there is Federal Food, Drug, and

 approving the registration of reasonable certainty of no harm, with

Cosmetic Act (FFDCA); pesticides to set allowable uses consideration of

Food Quality Protection Act in agriculture and extermination,  an
additional safety factor to protect

(FQPA) and

infants and children,

 setting tolerances for pesticide  aggregate exposure to a pesticide
residues permitted to remain in (including all exposures for which or on
foods available to the there is reliable information), and consumer.

 cumulative exposures to pesticides with a common mechanism of toxicity.
For other exposures, determine whether use of the pesticide would present
any unreasonable risk to

man or the environment. Office of Pollution Toxic Substances Control Act

Evaluation and regulation of For chemicals listed on the inventory

Prevention and Toxics (TSCA) existing and new chemicals used of ?existing
chemicals,? determine (OPPT)

in manufacturing and commerce whether use of that chemical will to identify
any potentially present an unreasonable risk to dangerous products or uses.

human health or the environment. For newly introduced chemicals, or
significant new uses of existing chemicals, determine whether use

may present an unreasonable risk to human health or the environment.

(Continued From Previous Page)

Office or program Major risk- related statute( s) Primary risk- related
tasks Primary risk- related mandate( s)

Office of Solid Waste Resource Conservation and

?Cradle- to- grave? regulation of Hazardous waste is defined as a

(OSW) Recovery Act of 1976 (RCRA) hazardous waste management, solid waste,
or combination of solid

including the use of risk wastes, which because of its assessment
information in

quantity, concentration, or physical,  defining (and delisting) chemical,
or infectious substances as hazardous

characteristics may wastes,

 cause, or significantly contribute to,  evaluating the hazards posed an
increase in mortality or an by waste streams, increase in serious
irreversible, or  assessing the need for incapacitating reversible,
illness; or corrective action at disposal

 pose a substantial present or sites, and

potential hazard to human health or  granting waste disposal permits. the
environment when improperly treated, stored, transported, or

disposed of, or otherwise managed. Treatment, storage, or disposal of waste
is to be conducted so as to minimize the present and future

threat to human health and the environment.

(Continued From Previous Page)

Office or program Major risk- related statute( s) Primary risk- related
tasks Primary risk- related mandate( s)

Office of Water (OW) Clean Water Act (CWA); Safe Evaluation and regulation
of Under CWA, EPA is to establish Drinking Water Act (SDWA);

ambient water and drinking water criteria for ambient water quality on SDWA
Amendments of 1996

quality, including the basis of health and ecological  recommending water
quality

effects and accurately reflecting the criteria and establishing latest
scientific knowledge on the national minimum effluent

kind and extent of all identifiable standards, effects on health and
welfare.

 prohibiting discharge of toxic pollutants in toxic amounts, Also under
CWA, effluent standards  establishing national drinking for toxic
pollutants are to be at that water standards for public water level which
the EPA Administrator systems, and

determines will provide an ample  identifying subpopulations at margin of
safety, so standards more elevated risk of health effects stringent than
those based on the from exposure to contaminants best available technology
in drinking water and economically achievable (the normal conducting studies

basis) may be named at EPA characterizing health risk to discretion.
sensitive populations from contaminants in drinking water.

Under SDWA, the EPA Administrator is to promulgate national primary drinking
water regulations for each contaminant which may have any adverse effect on
the health of persons and which is known or anticipated to occur in public
water

systems. Such regulations specify two levels of contamination

 a maximum contaminant level goal (MCLG) set solely on health grounds at a
level at which no known or anticipated effects on the

health of persons occur and which allows an adequate margin of safety, and

 a maximum contaminant level (MCL) set as close as feasible to the MCLG.

The 1996 amendments to SDWA require EPA, when developing drinking water
regulations, to (1) use the best available, peer- reviewed science and
supporting studies and data; and (2) make publicly available a risk
assessment document that discusses estimated risks, uncertainties, and
studies used in the assessment.

(Continued From Previous Page)

Office or program Major risk- related statute( s) Primary risk- related
tasks Primary risk- related mandate( s)

Office of Research and (Not applicable; supports the No direct regulatory
authority, but (Not applicable; supports the efforts Development (ORD)
efforts of other EPA program supports risk- related activities of of other
EPA program offices.)

offices.) other EPA offices by:

 conducting the hazard identification and doseresponse assessment steps of
risk assessments for specific chemicals at the request of program offices,

 preparing or assisting in the development of EPA risk assessment
guidelines and policies, and

 conducting complex, precedentsetting risk assessments.

Source: EPA documents and comments provided by EPA officials.

A number of other contextual factors affect the extent of involvement by EPA
offices in assessing and using risk assessment information in support of the
various statutes, mandates, and tasks identified in table 2.

 Risk assessment information may not be the only, or even the primary,
basis for the ultimate risk management decision. EPA statutes vary
fundamentally by whether the basis for regulation is (1) risk (health and
environmental) only, (2) technology- based, or (3) risk balancing
(consideration of risks, costs, and benefits).

 For some chemical risk assessment activities, EPA has a secondary role.
Instead, the main responsibility for determining the relative risk of a
chemical, compiling and analyzing risk- related data, or completing other
tasks associated with a particular statute might lie with industry, states,
or local entities.

 In practical terms, the resources available for conducting a risk
assessment for a given chemical might limit the depth and scope of EPA?s (or
other parties?) analysis. Such resource limitations might include not only
schedule and staffing constraints, but often the amount and quality of
directly relevant scientific data available for analysis.

Risk assessment activities involve both EPA?s program offices and its Office
of Research and Development (ORD), which is the principal scientific and
research arm of the agency. ORD often does risk assessment work for EPA
program offices that focuses on the first two steps in the four- step NAS
process- hazard identification and dose- response assessment- in particular,
the development of ?risk per unit exposed? numbers. The

exposure assessment and risk characterization steps tend to be the
responsibility of the various regulatory programs at EPA. However, according
to agency officials, both program offices and ORD may conduct all of the
risk assessment steps in particular cases. For example, OW?s

Office of Science and Technology does all of the assessments for purposes of
the SDWA, and, because of their particular statutory mandates, OPP and OPPT
have developed the capability to conduct all steps of a risk assessment on
their own. 1 ORD carries out all steps of highly complex, precedent- setting
risk assessments of specific chemicals, such as dioxin

and mercury. 1 Also, some of the material submitted by industry petitioners
seeking regulatory approvals of chemicals is confidential business
information, which precludes OPP and OPPT from sharing all of their risk
assessment data with other parts of EPA.

ORD also helps to coordinate the development of EPA?s risk assessment
methods, tools, models, and policies. In particular, much of EPA?s
agencywide guidance on conducting risk assessments is developed and
disseminated through ORD, with input from EPA?s program offices, Science
Policy Council, and Science Advisory Board, as well as other external
parties. Coordination of risk assessment activities also occurs through

EPA?s Risk Assessment Forum and the agency workgroups that approve
information for entry into EPA?s Integrated Risk Information System (IRIS).
The Risk Assessment Forum is a standing committee of senior EPA scientists
that was established to promote agencywide consensus on difficult and
controversial risk assessment issues and to ensure that this consensus is
incorporated into appropriate EPA risk assessment guidance.

Managed by ORD, IRIS is a computerized database that contains information on
human health effects that may result from exposure to various chemicals in
the environment. IRIS was initially developed for EPA staff in response to a
growing demand for consistent information on chemical substances for use in
risk assessments, decision making, and regulatory activities. The entries in
IRIS on individual chemicals represent a consensus opinion of EPA health
scientists representing the program offices and ORD and have been subject to
EPA?s peer review policy since its issuance in 1994. 2 Risk Assessment

There are agencywide risk assessment procedures that EPA?s various
Procedures

program offices generally follow, but each office also has different
statutory mandates and risk assessment tasks associated with its regulatory
authority. These contextual differences contribute to some

program- specific variations in the conduct of chemical risk assessments. In
addition, EPA?s procedures are in transition from more simplistic
traditional methods for identifying and assessing risks to increasingly
complex models and methods. It is particularly important to recognize

that, while most EPA guidelines (and this appendix) distinguish between
cancer and noncancer procedures, this distinction is becoming increasingly
blurred as new scientific methods are being developed and applied. In
general, EPA follows the NAS four- step process for human health risk
assessments: (1) hazard identification, (2) dose- response assessment, (3)
exposure assessment, and (4) risk characterization. However, for

ecological risk assessment, EPA?s guidelines recommend a three- step 2 IRIS
does not, however, incorporate OPP or OPPT risk assessment data from
confidential business data.

process: (1) problem formulation, (2) analysis, and (3) risk
characterization.

Guidelines To a much greater extent than the other agencies we reviewed, EPA
has documented its risk assessment procedures and policies in a voluminous
and expanding set of guidelines, policy papers, and memoranda. These
documents are primarily intended as internal guidance for use by risk

assessors in EPA and those consultants, contractors, or other persons who
perform work under EPA contract or sponsorship. 3 However, the documents
also make information on the principles, concepts, and

methods used in EPA?s risk assessments available to other interested
parties. EPA?s guidelines undergo internal and external peer review.

Beginning in 1986, EPA published a series of risk assessment guidelines to
set forth principles and procedures to guide EPA scientists in the conduct
of agency risk assessments, and to inform agency decision makers and the

public about these procedures. In general, EPA adopted the guiding
principles of fundamental risk assessment works, such as the 1983 Red Book
by the NAS? National Research Council (NRC). 4 EPA?s guidelines supplement
these principles. Five sets of guidelines were finalized in 1986, including
guidelines for carcinogen risk assessment, mutagenicity risk

assessment, health risk assessment of chemical mixtures, health assessment
of suspect developmental toxicants, and estimating exposures. 5 In part to
respond to advances and changes in risk assessment methods- but also in
response to criticisms of its guidelines by NRC, among others- EPA has
revised most of these guidelines, in either proposed or final form, and
produced additional guidance documents.

3 Guidelines are not rules, are not binding on either EPA or any outside
parties, and do not alter applicable EPA statutes and regulations. 4 NRC is
the principal operating agency and research arm of NAS in advising and
providing services to the federal government, the public, and the scientific
community. 5 Mutagenicity risk assessment focuses on analysis of agents that
may cause genetic mutations. Developmental toxicity risk assessment focuses
on risk to human development, growth, survival, and function because of
exposure to environmental agents prior to conception, prenatally, or to
infants and children.

Statutory changes have also prompted revisions and expansions of EPA?s risk
assessment guidelines and policy papers. In the Clean Air Act Amendments of
1990, for example, Congress directed EPA to revise its

carcinogen risk assessment guidelines, taking into consideration the NAS
recommendations, before making any determinations of the ?residual risks?
associated with emissions of hazardous air pollutants. 6 The results

of the NAS study appeared in the 1994 NRC report, Science and Judgment in
Risk Assessment. Among other things, NRC recommended that EPA better
identify the inference (default) assumptions in its guidelines, explain the
scientific or policy bases for selecting them, and provide guidance on when
it would be appropriate to depart from the assumptions. The current set of
agencywide risk assessment guidelines and policies includes the following
major topics: 7

 carcinogen risk assessment,

 neurotoxicity risk assessment,

 reproductive toxicity risk assessment,

 developmental toxicity risk assessment,

 mutagenicity risk assessment,

 health risk assessment of chemical mixtures,

 guidelines for exposure assessment,

 guidelines for ecological risk assessment,

 other risk assessment tools and policies,

 probabilistic analysis in risk assessment,

 use of the benchmark dose approach in health risk assessment,

 reference dose (RfD) and reference concentration (RfC),

 evaluating risk to children, and

 EPA risk characterization program. In addition to these agencywide
documents, there are also numerous program- specific guidelines and policy
documents. For example, the Risk Assessment Guidance for Superfund series
covers various stages of human health evaluation as well as ecological risk
assessment and

probabilistic risk assessment. There are also guidelines and policy
memoranda at the headquarters and regional office level that supplement

6 Residual risks are those remaining after the maximum achievable control
technology is in effect. 7 There are other documents, but they are mostly on
narrow topics- the assessment of thyroid follicular cell tumors, for
example.

these general Superfund guidelines. Similarly, OPP, with input from ORD, has
developed a series of science policy papers specifically on issues related
to pesticide risk assessments, in response to provisions of the Food

Quality Protection Act of 1996. Describing EPA?s risk assessment procedures
with any certainty is a difficult task, given the sheer volume of EPA
guidance documents, the continuing evolution of risk assessment practices,
and the extent to which many of EPA?s revisions are currently draft in
nature. For example, the official guidelines for cancer risk assessment are
still the 1986 version, but

the agency published a proposed revision of those guidelines in 1996, and
continued to revise them in 1999, but the revised guidelines have not yet
been made final by EPA. 8 Although the various revisions since 1986 do not
represent official agency policy at this stage, the approaches that they
describe are likely to provide a more accurate reflection of current
practices and directions in EPA risk assessments. To some extent EPA is
already applying these newer approaches, for example in the Office of
Water?s revised methodology for deriving ambient water quality criteria for

the protection of human health and the Office of Pesticide Programs? Cancer
Peer Review Committee. Agencywide Risk

The following sections summarize the basic elements of EPA?s agencywide
Assessment Procedures

procedures for conducting risk assessments. Because most of EPA?s guidelines
focus on human health risks, this section also focuses primarily on health
assessments in describing EPA?s general approach. EPA

generally uses the NAS four- step process for those assessments. However, a
separate short section on EPA?s approach to ecological risk assessment
appears at the end of this agencywide summary. Also, while this appendix
(and most of the source material from which it was derived) discusses
procedures for assessing cancer and noncancer effects separately, this
distinction is increasingly artificial. As EPA noted in its Strategy for
Research on Environmental Risks to Children, the agency is less likely to

consider cancer and noncancer endpoints in isolation as it develops and
incorporates more advanced scientific methods to measure and model the
biological events leading to adverse effects. 9 According to EPA, the
science

8 A particular focus of changes in the 1999 version is additional attention
to issues of human variability, especially risks to children versus adults.
9 EPA/ 600/ R- 00- 068 (August 2000).

of risk assessment is moving toward a harmonization of the methodology for
cancer and noncancer assessments.

Hazard Identification Carcinogens

EPA?s approach to hazard identification changed significantly between the
agency?s 1986 guidelines and its proposed revision. In its 1986 guidelines,
EPA defined a hierarchical classification scheme for hazard identification
of chemical agents (see table 3). In this scheme, analysis of whether an
agent is a potential human carcinogen proceeds through distinct steps based
on the type of human, animal, or ?other? evidence available and its

quality (whether such evidence is sufficient, limited, or inadequate),
resulting in classification of the agent in one of six alphanumeric
categories.

Table 3: EPA?s 1986 Classification System for Characterization of
Carcinogenicity Group Description Weight of evidence for carcinogenicity

A Human carcinogen Sufficient evidence from epidemiologic (human) studies B1
Probable human carcinogen Limited evidence from epidemiologic studies B2
Probable human carcinogen Sufficient evidence from animal studies and
inadequate evidence or no data from epidemiologic studies C Possible human
carcinogen Limited evidence in animals and absence of adequate human data

D Not classifiable Inadequate or no data E Evidence of noncarcinogenicity
for No evidence in adequate studies in at least two species or in both
epidemiologic humans and animal studies

Source: Adapted from EPA?s 1986 Guidelines for Carcinogen Risk Assessment
and other agency documents that describe the agency?s classification system.

In response to further developments in the understanding of carcinogenesis,
and to address limitations of its 1986 scheme, EPA proposed a revised
approach that melds the separate human- animal- other processes into a
single comprehensive evaluation. In this approach, weighing the evidence and
reaching conclusions about the carcinogenic potential of an agent would be
accomplished in a single step after assessing all individual lines of
evidence. Compared to the 1986 guidelines, the proposed revision also
encourages fuller use of all biological information- instead of relying
primarily on tumor findings- and emphasizes analysis of the agent?s mode of
action in leading to tumor development. 10 EPA?s proposed revision replaces
the 1986 alphanumeric classifications with a ?weight of evidence? narrative
to provide more complete information not only on the likelihood of human
carcinogenic effects but also the conditions under which such effects may be
expressed. To provide

some measure of consistency in the narratives, standard descriptors are to
be utilized to express the conclusion regarding the weight of evidence for
carcinogenic hazard potential. These descriptors have also been undergoing
some changes, but, according to EPA?s July 1999 discussion draft, would
include:

1. carcinogenic to humans, 2. likely to be carcinogenic to humans, 3.
suggestive evidence of carcinogenicity but not sufficient to assess

human carcinogenic potential, 4. data are inadequate for an assessment of
human carcinogenic potential,

and 5. not likely to be carcinogenic to humans.

The narrative might also reflect more than one conclusion for a given agent.
For example, a narrative could say that an agent is likely to be
carcinogenic by inhalation exposure but not likely to be carcinogenic by

oral exposure.

Noncancer effects

10 ?Mode of action? is defined as a series of key events and processes,
starting with interaction of an agent with a cell and proceeding through
operational and anatomical changes resulting in cancer formation.

EPA starts with a review and assessment of the toxicological database to
identify the type and magnitude of possible adverse health effects
associated with a chemical. Exposure to a given chemical might result in a
variety of toxic effects, so EPA has produced separate guidelines for the
assessment of mutagenicity, developmental toxicity, neurotoxicity, and
reproductive toxicity. 11 However, assessments for these noncancer health
effects may also overlap. For example, developmental effects might be traced
to exposures and factors also covered by reproductive toxicity assessments,
and developmental exposures may result in genetic damage that would require
evaluation of mutagenicity risks. The EPA guidelines for noncancer effects
are not step- by- step manuals, and they do not

prescribe a hazard identification classification scheme. Instead, they focus
on providing general advice to risk assessors on different types of toxicity
tests or data and on the appropriate toxicological interpretation of test
results (e. g., which outcomes should be considered adverse effects). In
addition to considering the types and severity of potential adverse effects,
hazard identification would also consider and describe the nature of
exposures associated with these effects. A review of the full range of
possibilities would consider:

 acute effects- generally referring to effects associated with exposure to
one dose or multiple doses within a short time frame (less than 24 hours,
for example);

 short- term effects- associated with multiple or continuous exposure
occurring within a slightly longer time frame, usually over a 14- day to
28day time period;

 subchronic effects- associated with repeated exposure over a limited
period of time, usually over 3 months; and

 chronic effects- associated with continuous or repeated exposure to a
chemical over an extended period of time or a significant portion of the
subject?s lifetime.

Procedurally, there is an important variation from the distinct four steps
of the risk assessment paradigm. In its guidelines, EPA notes that its
normal practice for assessments of noncancer health effects is to do hazard

11 Neurotoxicity focuses on adverse changes in the structure or function of
the central or peripheral nervous system following exposure to a chemical,
physical, or biological agent. Reproductive toxicity focuses on toxic
effects on the male and female reproductive systems, including outcomes of
pregnancy and lactation. Mutagenicity and developmental toxicity were
described in a previous footnote.

identification in conjunction with the analysis of dose- response
relationships. This is because the determination of a hazard is often
dependent on whether a dose- response relationship is present. According

to EPA, this approach has the advantages of (1) reflecting hazards in the
context of dose, route, duration, and timing of exposure; and (2) avoiding
the potential to label chemicals as toxicants on a purely qualitative basis.

Dose- Response Assessment Carcinogens

Risk assessors conducting dose- response assessments must make basic choices
regarding which data to base analyses upon and which models and assumptions
to use for extrapolation of study results to the potential

human exposures of regulatory interest. Data choices focus on the
availability and quality of human or animal studies. Three of the more
important extrapolation tasks are estimation of low- dose relationships (i.
e.,

those that fall below the range of observation in the studies supporting the
agency?s analysis), calculation of toxicologically equivalent doses when
dose- response data from animal studies are applied to human exposures, and
extrapolating results from data on one route of exposure to another route.
12  Data choices

The two main types of studies that provide data useful in a quantitative
dose- response assessment are (1) epidemiological studies of human
populations and (2) toxicological laboratory studies using animals or,
sometimes, human cells. Epidemiological studies examine the occurrence of
adverse health effects in human populations and attempt to identify the
causes. At a minimum, such studies can establish a potential link between
exposures to chemical agents and the occurrence of particular adverse
effects by comparing differences in exposed and nonexposed populations. If
there is adequate information on the exposure levels associated with adverse
effects, these studies can also provide the basis for a doseresponse

assessment. Because such data obviate the need to extrapolate from animals
to humans, EPA (like other agencies) prefers to use data from
epidemiological studies, if available.

12 Risk assessors might also need to make other types of extrapolations (e.
g., when estimating effects for less than a full lifetime).

Often, however, the available data for dose- response assessment will come
from animal studies. A common assumption underlying risk assessments by EPA
(and other agencies) is that an agent that produces adverse effects in
animals will pose a potential hazard to humans. EPA?s guidelines emphasize
that case- specific judgments are necessary in considering the

relevance of particular studies and their data. However, in the absence of
definitive information to the contrary, EPA?s guidelines establish some
standard default choices to assist risk assessors in selecting which studies
and data to use. (See the section on assumptions in this appendix for more
information on such default choices and assumptions.)  Extrapolation to low
doses

Quantifying risks engenders another set of issues and choices. In
particular, some type of low- dose extrapolation is usually necessary, given
that the doses observed in studies tend to be higher than the levels of
exposure of regulatory concern. There are limits to the ability of both
epidemiological and toxicological studies to detect changes in the

likelihood of health effects with acceptable statistical precision,
especially at the low- dose exposures typical of most environmental
exposures and given practical limits to the sizes of research studies. A
number of different models might be used for extrapolation, all giving
plausible results. In its proposed revision of the carcinogen risk
assessment guidelines, EPA identifies use of a biological extrapolation

model as the preferred approach for quantifying risk. Such models integrate
events in the carcinogenic process throughout the dose- response range from
high to low doses and include physiologically based pharmacokinetic (PBPK)
and biologically based dose- response models. PBPK models address the
exposure- dose relationship in an organism taken

as a whole, estimating the dose to a target tissue or organ by taking into
account rates of absorption into the body, metabolism, distribution among
target organs and tissues, storage, and elimination of an agent.
Biologically based dose- response models describe specific biological
processes at the cellular and molecular levels that link target- organ dose
to the adverse event. These models are useful in extrapolation between
animals and humans and between children and adults because they allow
consideration of species- and age- specific data on physiological factors
affecting dose levels and responses. However, biological models require
substantial quantitative data and adequate understanding of the carcinogenic
process for a specific agent. EPA cautions that the necessary data for using
such

models will not be available for most chemicals. Therefore, the agency?s
guidelines describe alternative methods.

Dose- response assessment is a two- step process when a biologically based
model is not used. The first step is the assessment of observed data to
derive a point of departure, and the second step is extrapolation from that
point of departure to lower (unobserved) exposures. According to EPA

guidelines, the agency?s standard point of departure for animal studies is
the effective dose (ED) corresponding to the lower 95- percent confidence
limit on a dose associated with 10- percent extra risk (LED 10 ) compared to
the control group. 13 EPA may use a lower point of departure for data from
human studies of a large population or from animal studies when such data
are available. For the extrapolation step, EPA?s proposed guidelines provide
three default approaches which assume, respectively, that the

dose- response relationship is linear, nonlinear, or both. The choice of
which default approach to apply is to be based on the available information
on the mode( s) of action of the chemical agent.

In the absence of sufficient mode of action information, or if the available
mode of action information indicates that the dose- response curve at low
doses is expected to be linear, the default is to use a linear approach for
the extrapolation step. The assumption of linearity is generally considered
a conservative, public- health protective default, intended to avoid
underestimating risks at low doses. It is rooted in EPA?s traditional

assumption for suspected carcinogens that no threshold exists regarding
adverse effects (i. e., any exposure to carcinogenic substances, no matter
how small, poses some risk of developing cancer). A linear, no- threshold
model generally assumes that adverse health effects are proportional to
exposure for any dose above zero. The linear approach is to draw a

straight line between the point of departure from observed data- typically
the LED 10 -and the origin (zero incremental dose, zero incremental
response). (See fig. 3 below.) According to EPA?s guidance, this linear
default approach is thought to generally provide an upper- bound calculation
of potential risk at low doses. The agency also pointed out that it gives
numerical results about the same as a linearized multistage approach, which
is the default approach under the 1986 cancer guidelines.

13 According to EPA, the LED is chosen to account protectively for
experimental variability 10 and is an appropriate representative of the
lower end of the observed range, because the limit of detection in studies
of tumor effects is about 10 percent.

Figure 3: Extrapolation for Carcinogens Extrapolation range Observed range
Response

on dose)

x

limit (Confidence

x estimate)

Human

(Central

exposure of interest

x LED 10

ED 10 10%

Projected linear

0% MOE Dose

LED 10 = Lower 95 percent confidence limit on a dose associated with a 10
percent extra risk ED 10 = Dose associated with a 10 percent extra risk MOE
= Margin of exposure

= Central estimate of dose- response curve = Confidence limit = Linear low-
dose extrapolation

Source: EPA Proposed Guidelines for Carcinogen Risk Assessment.

The default changes to a ?margin of exposure? analysis when (1) adequate
data on mode of action show that linearity is not plausible, and (2) the
data provide sufficient evidence to support a nonlinear mode of action for
the general population and any sub- populations of concern. Rather than
estimating the probability of effects at low doses, a margin of exposure
analysis compares the point of departure from study data with the dose

associated with the environmental exposure( s) of interest by computing the
ratio between the two. (See fig. 3 for a graphical representation of these
two points.) If the available evidence indicates that the doseresponse may
be adequately described by both a linear and a nonlinear approach, EPA?s
default is to present both the linear and the margin of exposure analysis.

 Extrapolation from animal to human- equivalent doses When dose- response
relationships are being extrapolated from the observed results of animal
studies, it is also necessary to estimate what doses are of equivalent risk
in the experimental animals (usually mice or rats) and humans. The objective
of this interspecies ?scaling? is to define dose units that are presumed to
lead to equivalent risk across species. Not

only are rodents much smaller in size than humans, but they have shorter
life spans and quicker metabolisms, all of which affect the equivalent dose.
For carcinogens, EPA?s historical default had been to assume that end-
oflife

cancer risks will be equivalent across species when lifetime dosing is
proportional to each species? body surface area. In practice, this ?surface

area scaling? was calculated by using daily milligrams scaled by the 2/ 3-
power of the species? body weight. EPA?s draft revision of the carcinogen
guidelines presents a default for oral dose of scaling daily applied doses
in proportion to body weight raised to the 3/ 4- power. This would be

consistent with the 1992 recommendation from an interagency group. 14
Scaling by this method generally results in human risk estimates that are
slightly lower than those obtained from surface area scaling. For cross-
species scaling of inhalation exposures, EPA uses a different approach- its
reference concentration (RfC) methodology- to determine what is
toxicologically equivalent. In the RfC approach, the agency estimates the
respiratory deposition of particles and gases and the internal doses of
gases with different absorption characteristics. EPA uses this

approach because it is the concentration needed to produce an equivalent
rate of loading the target tissue with deposited or absorbed chemical agent
that is important in inhalation exposures, not simply the dose in the air.

14 FDA and OSHA traditionally used an alternative assumption that risks are
equivalent when dosing is proportional to each species? body weight but are
now also using or considering (body weight) 3/ 4 .

 Extrapolation to different routes of exposure Exposure to a chemical agent
may occur through inhalation, oral ingestion, or dermal contact. However,
the data being used for a risk assessment may only reflect one of these
routes of exposure, and that route may be different than the route of
concern to agency regulators in a particular case. For example, available
animal study data might indicate that ingestion of a chemical substance is
associated with cancer, but risk

assessors may need to consider the relevance of this ingestion data to the
prediction of risks associated with inhalation exposures. In general, it is
EPA?s position that adverse effects manifested through one route of exposure
are relevant to consideration of any other route of exposure.

However, EPA?s guidelines also caution that such route- to- route
extrapolation should be consistent with existing biological information. 15

Noncancer effects

In contrast to its assessments of cancer risks, EPA?s traditional view of
noncancer toxic effects has been that a threshold exists. In other words,
the agency typically assumes that noncancer adverse effects occur only after
a threshold level of exposure to an agent has been exceeded. 16 For
noncancer effects, the primary objective of the agency?s assessment is to

derive an RfD or, in the case of inhalation exposures, an RfC representing a
concentration of the agent in the air rather than a dose. This RfD or RfC is
an estimate of a daily exposure to an agent by the human population

(including sensitive sub- groups) that is expected to be without an
appreciable risk of deleterious effects during a lifetime. For most of its
history, EPA has relied on a no observed adverse effect level (NOAEL)
approach to estimate the RfD. However, the agency may also use other
approaches, such as the benchmark dose approach, depending on the quality
and type of data available. The NOAEL and benchmark dose approaches are
described in more detail below.

15 For example, a chemical that acts as a carcinogen because of the way the
body metabolizes it via one route- such as ingestion into the stomach
transforming the chemical into a toxin- may not be metabolized in the same
way by another route. Carcinogenic effects, therefore, would not be expected
via another route of exposure that does not produce this same metabolite.

16 There are, however, exceptions to this general rule, such as reproductive
toxicants that may act through genetic mutation mechanisms. In such cases,
EPA will use the nothreshold approach described for suspected carcinogens.

Similar to other agencies, EPA?s traditional procedure for addressing
noncancer effects has been to define a NOAEL from experimental data and then
apply uncertainty factors to estimate an RfD or RfC. In this approach,
toxicologists first seek to identify the top of the range of dose levels
without any observed adverse effect in animals. 17 Then, to estimate a dose
to humans that will be similarly without effect, this dose level is divided
by

a set of uncertainty factors, typically factors of 10. 18 These uncertainty
factors are used to account for the possibility of greater sensitivity among
humans than in experimental animals, of greater sensitivity in some humans
compared to average humans, and for other concerns. One such concern is if
there was no dose in a study at which harmful effects were not detected. In
such cases, extrapolation is instead based on the lowest observed adverse
effect level (LOAEL), but an additional uncertainty factor

would then be incorporated in the agency?s estimated dose. The choice and
size of factors (e. g., 10 or 3) can vary from case to case, and EPA?s
guidelines note that professional scientific judgment must be used in
assigning uncertainty and modifying factors. EPA typically uses two or three
factors (a division of the NOAEL dose by 100- to 1000- fold), and,

under current policy, generally will use a maximum uncertainty factor of no
greater than 3, 000. 19 However, EPA may also reduce the standard
uncertainty factors if it has more informative pharmacokinetic data about

variability among humans or across species. An additional ?modifying? factor
may also be used when the areas of scientific uncertainty addressed with
uncertainty factors do not represent all of the uncertainties in the
estimation of a reference dose or concentration.

EPA may also use the benchmark dose approach to identify a dose without
appreciable effect from an experimental study. Unlike the NOAEL approach,
the benchmark dose approach uses the entire set of available data on doses
and responses; it is not limited to only considering the specific dose
levels tested in the study. In this benchmark dose approach,

17 Sometimes, a no observed effect level (NOEL) is used instead of a NOAEL.
18 These are often referred to as ?safety factors? by other agencies and in
general risk assessment literature. EPA, however, uses the term uncertainty
factor instead of safety factor because of concerns that the latter implies
an absolutely safe level, an assurance that the agency does not believe it
can provide.

19 It is EPA?s opinion that toxicity databases that are weaker and would
result in uncertainty factors in excess of 3, 000 are too uncertain as a
basis for quantification. In such cases, EPA no longer estimates an RfD.
Before this policy was in place, factors of 10, 000 were applied for a few
chemicals.

researchers fit a dose- response curve using all of the data in the observed
experimental range. EPA would typically use the modeled LED 10 dose as a
point of departure to derive an RfD, and the agency would still apply
uncertainty factors to this dose, as in the NOAEL approach. (See fig. 4.)

Figure 4: Derivation of a Reference Dose Using the Benchmark Dose Approach
100

Increasing response 90 80 70 60

UFs

50 40 30 20 10

LED ED10 10

0 0 RfD

50 100 150 200 250 300 350 Dose

LED 10 = Lower 95 percent confidence limit on a dose associated with a 10
percent extra risk

ED 10 = Dose associated with a 10 percent extra risk RfD = Reference dose
UFs = Uncertainty factors

= Central estimate of dose- response curve = Confidence limit

Source: EPA Methodology for Deriving Ambient Water Quality Criteria for the
Protection of Human Health (2000); Technical Support Document Volume 1: Risk
Assessment.

Exposure Assessment EPA?s program offices usually perform the exposure
assessment step, given the different exposure scenarios of interest for the
separate regulatory programs. 20 However, EPA has published agencywide
guidelines for

exposure assessment that describe general principles and practices for
conducting such assessments. The focus of EPA?s guidelines is on human
exposures to chemical substances, but the agency noted that much of the
guidance also applies to wildlife exposure to chemicals or human exposure to
biological, physical (e. g., noise), or radiological agents. 21 EPA points
out, though, that assessments in these other areas must consider additional
factors that are beyond the scope of the exposure assessment guidelines.

EPA?s guidelines establish a broad framework for agency exposure assessments
by describing the general concepts of exposure assessment, standardizing the
terminology (such as defining concepts of exposure, intake, uptake, and
dose), and providing guidance on the planning and implementation of an
exposure assessment. The guidelines are not, however, intended to serve as a
detailed instructional guide. EPA?s guidance prescribes no standard format
for presenting exposure assessment results, but recommends that all exposure
assessments, at a minimum, contain a narrative exposure characterization
section that

 provides a statement of purpose, scope, level of detail, and approach used
in the assessment, including key assumptions;

 presents the estimates of exposure and dose by pathway and route for
individuals, population segments, and populations in a manner appropriate
for the intended risk characterization;

 provides an evaluation of the overall quality of the assessment and the
degree of confidence the authors have in the estimates of exposure and dose
and the conclusions drawn; 22

 interprets the data and results; and 20 As mentioned earlier in this
appendix, sometimes ORD may do the entire risk assessment, including the
exposure assessment step. 21 The guidelines also discuss the implications
for exposure assessment if the assessment is to be used for purposes other
than risk assessment (e. g., to determine whether exposure occurs, to
monitor status and trends, or to establish exposure- incidence in
epidemiologic studies).

22 The guidelines note that it is common for the single largest source of
uncertainty in an exposure assessment to be the estimate of the duration of
an individual?s contact with a chemical at a given concentration. The
concentration of the chemical in the media (e. g., air, soil, or water)
often is known with more certainty than the activities of the exposed
individual( s).

 communicates the results of the exposure assessment to the risk assessor,
who can then use this information with the results from other risk
assessment elements to develop the overall risk characterization. The
guidelines encourage agency staff to use multiple ?descriptors? of both
individual and population risks, rather than a single descriptor or risk
value. The exposure guidelines also emphasize the use of more realistic

estimates of high- end exposures than had been the case in some previous
practices. In the past, EPA sometimes relied on exposure estimates derived
from a hypothetical ?maximally exposed individual? who might spend, for
example, a 70- year lifetime drinking only groundwater with the highest
concentrations of contaminants detected. According to the 1997 report of the
Presidential/ Congressional Commission, this approach was

often based on such unrealistic assumptions that it impaired the scientific
credibility of risk assessments. Now, however, EPA has adopted the use of
distributions of individual exposures as the preferred practice. EPA?s
guidance indicates that risk assessments should include both central
estimates of exposure (based on either the mean or median exposure) and

estimates of the exposures that are expected to occur in small, but
definable, high- end segments of the population. EPA states that a high- end
exposure estimate is to be a plausible estimate of the individual exposure
for those persons at the upper end of an exposure distribution. The agency?s
intent is to convey an estimate of exposure in the upper range of the
distribution, but to avoid estimates that are beyond the true distribution.
23

EPA has identified several new directions in its approach to exposure
assessment. First is an increased emphasis on total (aggregate) exposure via
all pathways. EPA policy directs all regulatory programs to consider in
their risk assessments exposures to an agent from all sources, direct and

indirect, and not just from the source that is subject to regulation by the
office doing the analysis. Another area of growing attention is the
consideration of cumulative risks, when individuals are exposed to many
chemicals at the same time. The agency is also increasing its use of
probabilistic modeling methods, such as Monte Carlo analysis, to analyze
variability and uncertainty in risk assessments and provide better estimates

of the range of exposure, dose, and risk in individuals in the population.
23 The guidelines state that, conceptually, the high end of the distribution
means above the 90 th percentile of the population distribution, but not
higher than the individual in the population who has the highest exposure.

EPA policy directs regulatory programs to pay special attention to the risks
of children and infants. EPA has produced some reference documents for
exposure assessments, such as the Exposure Factors Handbook. 24 This
handbook is intended to provide parameter values for use across the agency
and to encourage use of reasonable exposure estimates by providing
appropriate data sources and suggested methods. The handbook provides a
summary of available

statistical data on various factors used to assess human exposure to toxic
chemicals. These factors include: drinking water consumption; soil
ingestion; inhalation rates; dermal factors including skin area and soil
adherence factors; consumption/ intake of fruits and vegetables, fish,

meats, dairy products, homegrown foods, and breast milk; human activity
patterns, such as time spent performing household tasks; consumer product
use; and residential characteristics. EPA provides recommended values for
the general population and also for various segments of the population who
may have characteristics different from the general population (e. g., by
age, gender, race, or geographic location). EPA

guidance cautions, though, that these general default values should not be
used in the place of known, valid data that are more relevant to the
assessment being done. The default values used in EPA risk assessments,
however, sometimes vary slightly from the recommended values appearing in
the handbook. For example, while the handbook?s mean recommended

value for adult body weight is 71. 8 kilograms (kg), the handbook also noted
that a value of 70 kg has been commonly assumed in EPA?s risk assessments.
Similarly, the recommended value to reflect average life expectancy of the
general population is 75 years, but 70 years also has been

commonly assumed in EPA risk assessments. Officials from EPA program offices
pointed out that they may use different exposure factors in their risk
assessments because they sometimes develop exposure assessment methods
specific to their programs using different data sources or

population characteristics than those used by ORD for the Exposure Factors
Handbook.

Ecological Risk Assessment Ecological risk assessment is different from
human health risk assessment in that it may examine entire populations of
species and measure effects on partial or whole ecosystems. Often, the focus
is on not just a single ecological entity, but on the potential adverse
effects on multiple species 24 The agency also has a Wildlife Exposure
Factors Handbook for ecological risk assessments, and a Children?s Exposure
Factors Handbook is also in development.

and their interactions (for example, on the food chain). While human health
risk assessment is primarily concerned with an agent?s toxicity to humans,
ecological risk assessment might consider a range of adverse

effects on natural resources (such as crops, livestock, commercial
fisheries, and forests), wildlife (including plants), aesthetic values,
materials or properties, and recreational opportunities. For example, a
chemical agent could be considered a risk to wildlife if exposure to it
caused death, disease, behavioral abnormalities, mutations, or deformities
in the members of a species or their offspring. It could be considered a
risk to aesthetic values if it affected the color, taste, or odor of a water
source. By EPA?s definition, ecological risk assessment is a process that
evaluates the likelihood that adverse ecological effects may occur or are
occurring as a result of exposure to one or more ?stressors.? In other
words, ecological risk assessments may be prospective or retrospective, and,
in many cases,

both approaches are included in a single risk assessment. Chemicals are only
one of the possible ecological stressors that EPA might consider, along with
physical and biological ones. 25 EPA?s guidance focuses on stressors and
adverse ecological effects generated or influenced by human activity, which
could be addressed by the agency?s risk management decisions.

In comparison to human health risk assessment procedures, the approaches for
ecological risk assessment are more recent and less well developed. However,
as these methods have changed to incorporate and

better characterize dynamic, interconnected ecological relationships, EPA
has updated its guidance documents on the subject, with input from multiple
interested internal and external parties. According to EPA, the solicitation
of input from an array of sources is based, in part, on the need

to establish a framework for characterizing risks based on numerous
stressors, interconnected pathways of exposure, and multiple endpoints
(adverse effects).

25 For example, the alteration of a wildlife habitat would be a physical
stressor and the introduction of a nonnative invasive species would be a
biological stressor.

The most recent version of EPA?s framework appears in Guidelines for
Ecological Risk Assessment, published in 1998. 26 EPA?s guidelines describe
an iterative three- phase process consisting of problem formulation,
analysis, and risk characterization. These guidelines incorporate many of
the concepts and approaches called for in human health risk assessments.
However, particularly in the addition of a problem formulation phase, the
ecological risk assessment framework deviates from the standard four- step

process used for human health risk assessments. EPA pointed out that, unlike
human health assessments where the species of concern and the endpoints (e.
g., cancer) have been predetermined, ecological risk assessments need a
phase that focuses on the selection of ecological

entities and endpoints that will be the subject of the assessment. Table 4
summarizes the activities and expected outcomes for each of the three phases
of an ecological risk assessment. Prior to these phases, according to EPA, a
planning stage occurs during which risk assessors, risk managers, and other
interested parties are to have a dialogue and scope the problem.

Table 4: Summary of EPA?s Ecological Risk Assessment Process Phase Actions
Products

Problem formulation  Articulate purpose for assessment

 Assessment endpoints

 Define problem

 Conceptual models

 Determine plan for analyzing and characterizing risk

 Analysis plan

 Integrate available information on sources, stressors, effects, and
ecosystem and receptor characteristics Analysis  Characterize exposure

 Exposure profile

 Characterize ecological effects  Stressor- response profile Risk
characterization  Estimate risk

 Risk description

 Summarize assumptions, scientific uncertainties, and strengths and
limitations of the analyses

Source: EPA Guidelines for Ecological Risk Assessment.

Among the things considered during problem formulation is the selection of
assessment endpoints, which are ?explicit expressions of the actual
environmental value that is to be protected.? This is unlike human health

26 EPA has published other, more detailed guidance documents addressing
specific ecological risk assessment topics relevant to its program offices.
For example, Ecological Risk Assessment Guidance for Superfund: Process for
Designing and Conducting Ecological Risk Assessments from 1997 is specific
to EPA?s Superfund Program requirement for an ecological risk assessment.

assessments, where the species of concern and the endpoints have been
predetermined. The selection of endpoints at EPA has traditionally been done
internally by program offices, but more recently, affected parties or
communities are assisting in the selection of endpoints with their selection

based on ecological relevance, susceptibility, and relevance to management
goals. Furthermore, conceptual models are developed during the problem
formulation phase. Such models contain risk hypotheses in the form of
written descriptions and visual representations, outlining predicted
relationships between ecological entities and the stressors to which they
may be exposed. According to EPA the hypotheses are in effect assumptions,
being based on theory and logic, empirical data, mathematical models,
probability models, and professional judgment.

Subsequently, during the analysis phase data are selected that will be used
on the basis of their utility for evaluating risk hypotheses. The major
items considered during this phase are the sources and distribution of
stressors in the environment, the extent of contact and stressor- response

relationships, the evidence for causality, and the relationship between what
was measured and the assessment endpoint( s). Field studies involving
statistical techniques (i. e., correlation, clustering, or factor analysis),
surveys, the formation of indices, and the use of models are approaches to

evaluating the determined risk hypotheses. (EPA?s guidance on the risk
characterization phase of an ecological risk assessment is discussed in the
final section of this appendix.)

Program- Specific Risk EPA?s various program offices generally follow the
agencywide risk

Assessment Procedures assessment procedures and guidelines described above.
The major exception to this is the Chemical Emergency Preparedness and
Prevention

Office, which does not follow the NAS four- step process for its risk
assessment procedures because of its focus on risks associated with
accidental chemical releases. Overall, there is great diversity in the
context for risk assessment activities across EPA?s program offices. Each
program has different statutory mandates and risk assessment tasks
associated with its specific regulatory authority, and these contribute to
variations in the way the offices conduct risk assessments. In particular,
there are differences in the exposure assessment step across, and sometimes
within,

EPA?s program offices. This is not surprising, given that EPA?s regulatory
authorities regarding chemical agents primarily vary according to types and
sources of exposure. Although there are overlaps in these various exposures
to chemicals, EPA?s program offices generally assess and regulate different
aspects of the risks associated with exposures to humans

and/ or the environment. There are also some variations in the conduct of
hazard identification and dose- response analysis. The following sections
summarize the risk assessment activities and procedures of those EPA

program offices that are most likely to conduct assessments involving
chemical risks. The descriptions highlight some of the major variations and
similarities across the program offices.

Office of Pesticide Programs OPP is part of EPA?s Office of Prevention,
Pesticides and Toxic Substances (OPPTS). The primary risk assessment-
related activities of OPP are the registration of pesticides and the setting
of tolerances for pesticide residues. 27 Registration involves the licensing
of pesticides for sale and use in agriculture and extermination. No chemical
may be sold in the United States as a pesticide without such registration,
which establishes the conditions of legal use. All uses within the scope of
the registration conditions and limits are permissible, although actual
practice may vary. Pesticide tolerances are the concentrations (maximum
pesticide residue

levels) permitted to remain in or on food, as it is available to the
consumer. Registrations and tolerances are obtained through petitions to
OPP. The petitioner has the primary responsibility to provide the data
needed to support registration and tolerances, including information on the
toxicological effects of the pesticide. There are three major risk statutes
affecting EPA?s actions regarding pesticides. Registration is carried out
under the Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA).
Tolerances are established under

the Federal Food, Drug, and Cosmetic Act (FFDCA). 28 In 1996, Congress
amended both FIFRA and FFDCA through the FQPA, which mandated some key
changes in risk assessment of pesticides. Major features and characteristics
of chemical risk assessment by OPP are summarized below.

 OPP conducts all steps of risk assessments. Because OPP generally follows
the NAS four- step process for human health risk assessment and 27 The term
?pesticide? includes many kinds of ingredients used in products, such as
insecticides, fungicides, rodenticides, insect repellants, weed killers,
antimicrobials, and swimming pool chemicals, which are designed to prevent,
destroy, repel, or reduce pests of any sort. 28 While EPA administers the
setting of pesticide tolerances, FDA has authority over enforcement of the
tolerances.

the EPA- wide risk assessment guidelines, most of its procedures mirror
those used elsewhere in the agency.

 OPP officials noted that, over the last three decades, their office has
developed a rigorous process to support the development of chemical risk
assessments. This process includes regulations to establish

baseline data requirements and published guidelines for conducting required
studies. OPP officials emphasized the transparency of the process used to
develop EPA?s risk assessment procedures and the transparency of the
procedures EPA uses to make decisions on the risk of individual pesticides.
As an example, they noted that their program has consulted with outside
experts and asked for public comment on its

guidelines for reviewing studies, science policies for assessing the
significance of study data, and standard operating procedures for
implementing these policies in the development of a hazard identification or
exposure assessment for a chemical. They also pointed out that OPP adopted a
public participation process for reregistration and tolerance reassessment
decisions on registered pesticides and that they publish for public comment
proposed tolerances for proposed new uses of pesticides. In some
circumstances, OPP consults with outside experts concerning a risk
assessment of an individual pesticide.

 Pesticide registration decisions are based primarily on OPP?s evaluation
of the test data provided by petitioners (applicants). EPA has established a
number of requirements, such as the Good Laboratory Practice Standards, to
ensure the quality and integrity of pesticide data. OPPTS has also developed
harmonized test guidelines for use in the testing of pesticides and toxic
substances and the development of test

data that must be submitted to EPA for review under federal regulations. 29
Depending on the type of pesticide, OPP can require more than 100 different
tests to determine whether a pesticide has the potential to cause adverse
effects to humans, wildlife, fish, and plants.

29 OPPTS developed these guidelines through a process of harmonization that
blended the testing guidance and requirements that existed in OPP, OPPT, and
guidelines published by the Organization for Economic Cooperation and
Development. The purpose of harmonizing

these guidelines into a single set of OPPTS guidelines was to minimize
variations among the testing procedures that must be performed to meet EPA
data requirements under FIFRA (for OPP) and TSCA (for OPPT). The pesticide
data requirements are set out in 40 CFR 158.

 The FQPA established a single, health- based standard-? reasonable
certainty of no harm?- for pesticide residues in all foods. 30 All existing
tolerances that were in effect when the FQPA was passed are to be
reevaluated by 2006 to ensure that they meet the new safety standard.

The law requires EPA to place the highest priority for tolerance
reassessment on pesticides that appear to pose the greatest risk. To make
the finding of ?reasonable certainty of no harm? OPP considers: 1. the
toxicity of the pesticide and its break- down products; 2. how much of the
pesticide is applied and how often; and 3. how much of the pesticide remains
in or on food by the time it is marketed and prepared (the residue).

 Among other key changes affecting OPP?s risk assessments when setting
tolerances, the FQPA requires the agency to:

1. Explicitly address risks to infants and children and to publish a
specific safety finding before a tolerance can be established. It also
requires an additional tenfold uncertainty factor (unless reliable data show
that a different factor will be safe) to account for the possibly greater
sensitivity and exposure of children to pesticides.

2. Consider aggregate exposure from a pesticide, including all anticipated
dietary and all other exposures for which there is reliable information.
These include exposures through food, drinking water, and nondietary

exposures encountered through sources in the home, recreational areas, and
schools. 31 3. Consider cumulative exposures to pesticides with a common
mechanism of toxicity, which previously had been considered separately. 32

30 This eliminated problems caused by prior differences in the standards
that applied to raw versus processed foods. 31 Occupational exposures are
regulated separately.

32 See Children and Pesticides: New Approach to Considering Risk Is Partly
in Place (GAO/ HEHS- 00- 175, Sept. 11, 2000) for more detailed information
on EPA?s progress in implementing the FQPA provisions.

 Title III of the FQPA also requires certain data collection activities of
the Secretary of Agriculture, in consultation or cooperation with the
Administrator of EPA and the Secretary of Health and Human Services,
regarding food consumption patterns, pesticide residue levels, and pesticide
use that, according to EPA, affect its risk assessments when setting
tolerances.

 Also as a result of the FQPA, OPP uses a population adjusted dose (PAD),
which involves dividing the acute or chronic reference dose by the FQPA
uncertainty factor. According to OPP officials, this allowed OPP to be
consistent with the rest of the agency regarding setting RfDs, but still use
the FQPA factor for regulating pesticides.

 OPP is concerned with both cancer and noncancer toxicity. However, for
noncancer effects, OPP has paid special attention to neurotoxicity (because
many pesticides work through this mechanism) and, more recently, to
endocrine disrupting effects (those affecting the body?s hormone system). 33

 OPP officials noted that, while their agency has made important use of
?real life? monitoring or incident data, it primarily relies on studies
conducted in laboratory animals and on laboratory or limited field studies.
They stated that, in their experience, ?real life? data have profound
limitations and that such data are inconsistent, expensive, inconclusive,
and are not available for premarket decision making.

They said that, most importantly, by the time there are observable health or
environmental effects, it is too late to prevent the harm that could have
been predicted from judicious use of animal or environmental fate studies
conducted in the laboratory.  During the exposure assessment step, OPP is
concerned with a variety of routes, sources, and types of exposure. The
three routes by which

people can be exposed to pesticides are inhalation, dermal (absorbing 33
Because of the potentially serious consequences of human exposure to
endocrine disrupting chemicals, Congress included specific provisions on
this topic in both the FQPA and the 1996 amendments to the Safe Drinking
Water Act. In response to the FQPA language, EPA developed its Endocrine
Disruptor Screening Program, which focuses on providing methods and
procedures to detect and characterize endocrine activity of pesticides,
commercial chemicals, and environmental contaminants. EPA uses a tiered
approach for these risk assessments, sorting chemicals into four categories
on the basis of the existing scientific data.

pesticides through the skin), and oral (getting pesticides in the mouth or
digestive tract). Depending on the situation, a pesticide could enter the
body by any one or all of these routes. Typical sources of pesticide

exposure include food, home and personal use of pesticides, drinking water,
and work- related exposure to pesticides (in particular, to pesticide
applicators or vegetable and fruit pickers). In its approach to exposure
assessment, OPP distinguishes between residential and occupational types of
exposures. OPP officials noted that their program is further developing
procedures to conduct drinking water exposure assessments and residential
exposure assessments and that they have new procedures for ecological risk
assessments.

 OPP calculates estimates of acute (i. e., short- term) pesticide exposure
slightly differently from those for chronic (i. e., longer- term) exposures.
This is because an acute assessment estimates how much of a pesticide

residue might be consumed in a single day, while a chronic assessment
estimates how much might be consumed on a daily basis over the course of a
lifetime. In an important difference, acute assessments are based on high-
end individual exposure assumptions, while chronic

assessments use average exposure assumptions.

 In assessing both acute and chronic risks, OPP uses a tiered approach,
starting with an initial screening tier and proceeding through progressively
more elaborate risk assessments, if needed. The analytical tiers proceed
from more conservative to less conservative assumptions. For the first- tier
risk assessment, OPP uses ?worst- case? assumptions (e. g., that pesticide
residues are at tolerance levels and that

100 percent of the food crop is treated with the pesticide) that give only
an upper- bound estimate of exposure. For more refined analyses, OPP
officials noted that they have new procedures for conducting probabilistic
dietary exposure assessments. Generally, the level of resources and the data
needed to refine exposure estimates increase with each tier. Typically, if
risks from pesticide residues are not of concern using lower- tier exposure
estimates, OPP does not make

further refinements through additional tiers. However, with the aggregate
and cumulative exposure assessments now required by the FQPA, EPA notes that
it is likely that higher- tier exposure estimates will be needed.

 The agency has developed procedures for modeling the environmental fate of
pesticides. OPP officials said that these models use real data on the
physical and chemical properties of the pesticide, information on the

proposed or actual uses of the pesticide, and real data on the movement of
pesticides or other materials through soil, air, water, skin, textiles, or
other media to predict potential exposures to a pesticide. These models are
guided by scientific judgments that are based upon data and

scientists? experience in drawing inferences from these data. Office of
Pollution Prevention OPPT (formerly the Office of Toxic Substances) is also
part of OPPTS. and Toxics

OPPT was established to implement the Toxic Substances Control Act (TSCA),
which authorizes EPA to screen existing and new chemicals used in
manufacturing and commerce to identify potentially dangerous products or
uses. 34 TSCA focuses on the properties of a chemical and paths of exposure
to that chemical. Risk assessment activities are primarily related

to four sections of TSCA:

 Section 4 directs EPA to require manufacturers and processors to conduct
tests for existing chemicals when: (1) their manufacture, distribution,
processing, use, or disposal may present an unreasonable risk of injury to
health or the environment; or (2) they are to be produced in substantial
quantities and the potential for environmental release or human exposure is
substantial or significant. Under either condition, EPA must issue a rule
requiring testing if existing data are insufficient to predict the effects
of human exposure and environmental releases and testing is necessary to
develop such data. Rhomberg pointed out that these conditions require OPPT
to do some preliminary

risk assessment and that, unlike testing mandates under other statutes (e.
g., regarding pesticides), the agency has the burden of showing that such
testing is necessary. 35

 Section 5 addresses future risks through EPA?s premanufacture screening-
the premanufacture notification (PMN) process. This also applies to a
?significant new use? of an existing chemical.

 Section 6 directs EPA to control unreasonable risks presented or that will
be presented by existing chemicals. 34 15 U. S. C. 2601 et seq. In addition
to its original role to implement TSCA, OPPT has been given responsibility
for pollution prevention programs; regulation of specific toxic substances,
including asbestos, radon, and lead; and administration of the Toxics
Release Inventory. 35 Lorenz Rhomberg, A Survey of Methods for Chemical
Health Risk Assessment Among Federal Regulatory Agencies, a report prepared
for the Presidential/ Congressional

Commission on Risk Assessment and Risk Management (1996).

 Section 8 requires EPA to gather and disseminate information about
chemical production, use, and possible adverse effects to human health and
the environment. This section requires EPA to develop and maintain an
inventory of all chemicals, or categories of chemicals, manufactured or
processed in the United States. All chemicals not on the inventory are, by
definition, ?new? and subject to the notification provisions of section 5.
Once a chemical enters commerce through the section 5 process, it is listed
as an existing chemical.

Although TSCA gives EPA general authority to seek out and regulate any
?unreasonable risk? associated with new or existing chemicals, there are two
major limitations on the agency?s regulatory actions. First, as implemented
by EPA, regulation under TSCA involves consideration of both risks and
applying the least burdensome requirement needed to

regulate the risk. The term ?unreasonable risk? is not defined in TSCA.
However, according to EPA, the legislative history indicates that
unreasonable risk involves the balancing of the probability that harm will
occur, and the magnitude and severity of that harm, against the effect of a
proposed regulatory action on the availability to society of the expected
benefits of the chemical substance. The second major limitation on EPA?s
authority under TSCA is a requirement to defer to other federal laws.
Generally, if a risk of injury to health or the environment could be
eliminated or reduced to a sufficient extent by actions taken under another
federal law, that other law must be deferred to unless it can be shown to be
in the public interest to regulate under TSCA.

The major distinction in the procedures that apply to OPPT risk assessments
is between the evaluation of potential risks associated with exposures to
new versus existing chemicals. For EPA to control the use of a chemical
listed on the inventory of existing chemicals, according to

OPPT, a legal finding has to be made that the chemical will present an
unreasonable risk to human health or the environment. According to OPPT,
this standard requires the agency to have conclusive data on that particular
chemical. The agency noted, in comparison, that newly introduced chemicals
(or uses) can be regulated under TSCA based on whether they may present an
unreasonable risk, and this finding of risk can

be based on data for structurally similar chemicals. Because industrial
chemicals in commerce in 1975- 1977 were ?grandfathered? into the inventory
without considering whether they were hazardous, there are situations in
which existing chemicals might not be controlled, while EPA would act to
control a new chemical of similar or less toxicity under the PMN program.
Additional information on the major features and

characteristics of assessments for new versus existing chemicals is
presented below.

Premanufacture notification for new chemicals or significant new uses

 TSCA requires manufacturers, importers, and processors to notify EPA at
least 90 days prior to introducing a new chemical into the U. S. or
undertaking a significant new use of a chemical already listed on the TSCA
inventory. If available, test data and information on the chemical?s
potential adverse effects on human health or the

environment are to be submitted to EPA. Much of this submission must be kept
confidential by OPPT. However, there is no defined toxicity data set
required before PMN, and, unless EPA promulgates a rule requiring the
submission of test data, TSCA does not require prior testing of new
chemicals. Consequently, according to EPA, less than half of the PMNs
submitted include toxicological data. OPPT reviews approximately 1,500 PMNs
annually.

 EPA has 90 days after notification to evaluate the potential risk posed by
the chemical. EPA must then decide whether to (1) permit manufacture and
distribution (the default if EPA takes no action), (2) suspend manufacture
and distribution or to restrict use pending the development of further data,
or (3) initiate rulemaking to regulate manufacture or distribution.

 OPPT typically has very limited chemical- specific data on toxic effects
and exposure associated with new chemicals. When no data exist on the
effects of exposure to a chemical, EPA may make its determination on what is
known about the chemical?s molecular structure (called the structure-
activity relationship, or SAR) and the effects of other chemicals that have
similar structures and are used in similar ways.

OPPT?s New Chemicals Program has issued a document entitled

Chemical Categories that describes information for numerous classes of
chemicals. In assessing exposures for new chemicals where exposure
monitoring data are unavailable, OPPT uses several screeninglevel
approaches, including (1) estimates based on data on analogous chemicals;
(2) generic scenarios (i. e., standardized approaches for

assessing exposure and release for a given use scenario); (3) mathematical
models based on empirical and theoretical data and information; and (4)
assumptions of compliance with regulatory limits, such as OSHA Permissible
Exposure Limits (PELs).

 Rhomberg noted that OPPT cannot require full testing for all chemicals,
because of statutory limitations under TSCA. He therefore characterized
OPPT?s assessments as ?rough screens? designed to flag situations in which
further testing should be required.

Assessments of existing chemicals

 Chemicals that OPPT assesses for regulation under sections 4 or 6 of TSCA
are subject to a more rigorous risk assessment process. Compared to PMN
reviews, such assessments are much more similar to those conducted elsewhere
in EPA, so the EPA- wide guidelines generally apply.

 For hazard identification and dose- response assessment of carcinogens and
noncancer effects, OPPT follows EPA- wide procedures. Because TSCA focuses
on the properties of a chemical, rather than on a specific pathway or mode
of exposure, OPPT considers the potential hazards posed through multiple
routes of exposure. In lieu of information to the contrary, OPPT typically
presumes that the results for one route are applicable to other routes.

 Similarly, in exposure assessment OPPT considers a variety of types and
routes of exposure. Unlike other programs that focus on exposure through one
medium, assessments under TSCA must assess all potential exposures to a
chemical that may lead to unreasonable risk, considering, for example, both
residential and occupational exposures. These risks may be assessed
separately for each mode of exposure, even

if occurring in the same setting. Overall, OPPT aims to provide both central
estimates and upper- bound estimates of exposure, and it considers
population risks as well as individual risks.

 OPPT shares overlapping concerns about a number of different kinds of
exposure with other federal regulatory agencies. However, some aspects of
OPPT?s exposure assessments may differ from those of other

programs or agencies concerned with similar exposures. For example, with
regard to occupational exposures OPPT assumes that a working lifetime is 40
years, rather than the 45 years assumed by OSHA. Another example is the
assumption of body weight; OPPT uses 70kg, whereas ORD recommends a value of
71.8 kg in its Exposure Factors Handbook.

In addition to the assessment of chemicals for regulation under sections 4
and 6 of TSCA, OPPT has recently launched a new program to voluntarily add
screening- level hazard information on approximately 2, 800 highproduction-
volume industrial chemicals and has proposed a second new voluntary program
to address the risks of certain industrial chemicals to which children may
be exposed. These two new programs operate under the same risk assessment
processes used in the other OPPT programs noted above.

Office of Emergency and OERR is part of EPA?s Office of Solid Waste and
Emergency Response Remedial Response (Superfund (OSWER). Risk assessments
are a required component of a larger Program)

remediation process established by the Comprehensive Environmental Response,
Compensation, and Liability Act of 1980 (CERCLA or Superfund), as amended by
the Superfund Amendments and Reauthorization Act of 1986 (SARA). 36 Congress
enacted CERCLA to facilitate the cleanup of hazardous waste sites. The act
gave EPA broad

authority to respond to releases of hazardous substances. SARA requires EPA
to emphasize cleanup remedies that treat- rather than simply contain-
contaminated waste to the maximum extent practicable and to

use innovative waste treatment technologies. Hazardous substances are
defined by CERCLA to include substances identified under the Solid Waste
Disposal Act, the Clean Water Act, the Clean Air Act, and the Toxic
Substances Control Act, or designated by EPA. After investigating
potentially hazardous sites, EPA ranks them according to the severity of
their waste problems and places the worst on its National Priorities List
for Superfund cleanup. Under CERCLA section 105, EPA uses a Hazard Ranking
System to decide which sites to include on the list. Section 105 states that
priorities are to be based upon relative risk or

danger to public health or welfare or the environment, taking into account
the population at risk, the hazard potential of the hazardous substances,
and the potential for contamination of air and drinking water, among other
factors.

OERR has developed a human health and environmental evaluation process as
part of its remedial response program. Major features and characteristics of
the Superfund risk assessment procedures are summarized below.

36 42 U. S. C. 9601- 9675.

 Overall, the risk scenarios for Superfund can be very complex. Superfund
sites are often associated with multiple potential pathways and routes of
exposure, and mixtures of chemicals at Superfund sites are common. In
addition, the Superfund program is required to consider ecological as well
as human health risks.

 A risk assessment is performed after a particular site has been identified
according to the National Contingency Plan, EPA?s regulation outlining
requirements relevant to response action( s) for hazardous substances. The
remedial response process under the National Contingency Plan- and the role
of risk information in the process- is summarized in the following seven
steps:

1. Site discovery or notification:

 report determinations about which substances are hazardous. 2. Preliminary
assessment and site inspection:

 collect and review all available information to evaluate the source and
nature of hazardous substances. 3. Hazard ranking system:

 compile data from steps one and two in a numerical scoring model to
determine a relative risk measure. 4. Possible inclusion of site on the
National Priorities List based on one of the following criteria:

 the release scores sufficiently high pursuant to the Hazard Ranking
System,

 a state designates a release as its highest priority, or

 the release satisfies all of the following criteria:

 the Agency for Toxic Substances and Disease Registry has issued a health
advisory that recommends dissociation of individuals from the release, 37

 EPA determines that the release poses a significant threat to public
health, and 37 The Agency for Toxic Substances and Disease Registry within
the Department of Health and Human Services has an oversight role but no
regulatory authority. It prepares ?health assessments? rather than risk
assessments that tend to be qualitative in nature, site specific, and
focused on medical and public health perspectives. Exposures to site
contaminants are presented in terms of sensitive populations, mechanisms of
toxic chemical action, and possible disease outcomes. In contrast, EPA?s
human health evaluation, which is more quantitative, is a characterization
of the potential for adverse effects from human exposures

to environmental hazards.

 EPA anticipates that it will be more cost effective to use its remedial
authority than to use removal authority to respond to the release.

5. Remedial investigation and feasibility study:

 characterize the contamination at site where data is obtained to identify,
evaluate, and select cleanup alternatives. 6. Selection of a remedy:

 choose remedy that is protective of human health and the environment by
eliminating, reducing, or controlling risks posed through each pathway, and

 utilize risk information obtained during step five. 7. 5- year review.

One intended result of the remedial steps is the facilitation of a
sitespecific baseline risk assessment, designed to support risk management
decision making. Human health and ecological risk assessments occur during
step five, the Remedial Investigation/ Feasibility Study stage.

 For human health risk assessments, Superfund procedures approximate the
NAS paradigm, using the following four stages. 1. A data collection and
evaluation stage that involves:

 gathering and analyzing site data relevant to the human health evaluation,
and

 identifying substances present at the site that are the focus of the risk
assessment process. 2. An exposure assessment that involves:

 analyzing contaminant releases,

 identifying exposed populations,

 identifying potential exposure pathways and estimating exposure
concentrations for pathways, and

 estimating contaminant intakes for pathways. 3. A toxicity assessment
stage that considers:

 types of adverse health effects associated with chemical exposures,

 relationships between magnitude of exposure and adverse effects,

 related uncertainties such as the weight evidence of a particular
chemical?s carcinogenicity in humans, and

 existing toxicity information developed through hazard identification and
dose- response assessment. 4. A risk characterization that involves:

 characterizing potential for adverse health effects (cancer or noncancer)
to occur,

 evaluating uncertainty, and

 summarizing risk information.

 For ecological risk assessments, EPA?s guidelines suggest that Superfund
remedial actions generally should not be designed to protect organisms on an
individual basis, but should protect local populations and communities of
biota. 38 Furthermore, except for a few very large sites, Superfund
ecological risk assessments typically do not address effects on entire
ecosystems. Instead, they gather data regarding the effects on individuals
in order to predict or postulate potential effects on local wildlife, fish,
invertebrate, and plant populations and communities that occur or that could
occur in specific habitats at sites (e. g., a wetland, floodplain, stream,
estuary, or grassland). Specifically, the guidelines recommend that
ecological risk assessments performed

at every site follow an eight- step process: 1. Screening- level problem
formulation and ecological effects evaluation:

 site history,

 site visit,

 problem formulation, and

 ecological effects evaluation. 2. Screening- level exposure estimate and
risk calculation:

 exposure estimate, and

 risk calculation. 3. Baseline risk assessment problem formation:

 ecotoxicity literature review,

 exposure pathways,

 assessment endpoints and conceptual model, and

 risk questions. 4. Measurement endpoints and study design. 5. Verification
of field sampling design. 6. Site investigation and data analysis. 7. Risk
characterization. 8. Risk management.

 OERR uses a tiered approach for Superfund risk assessments, in which the
agency employs more conservative methods and assumptions in the initial
screening phases, followed by a more rigorous, multistage risk assessment if
screening results indicate the need. Under Superfund,

38 An exception would be actions affecting designated protected status
resources that could be exposed to site releases. Such resources include
treaty- protected species and species that are listed as or candidates for
threatened or endangered status.

decisions generally are made on a site- by- site basis. According to agency
officials, early activities at Superfund sites are often based on initial
tier screening. However, they pointed out that the remedial cleanup decision
is supported by a site- specific risk assessment that is usually quite
detailed with either site- specific exposure assumptions or

national default assumptions appropriate to the site which result in ?high-
end? reasonable risk estimates. Although the Superfund program initially
employed an approach of using a hypothetical ?worst case? scenario for
exposure assessments, EPA?s exposure assessment guidance now emphasizes use
of a more realistic upper- bound exposure

scenario. The EPA guidelines emphasize that this exposure scenario should be
in the range of plausible real exposures, and also call for a central
tendency case. In addition, guidelines put forth by the Superfund program
office emphasize streamlining the process and reducing the cost and time
required, focusing on providing information necessary to justify action and
select the best remedy for a Superfund site. In doing so, Superfund
guidelines suggest using standardized assumptions, equations, and values
wherever appropriate.

 The Superfund program uses extensive additional program- specific guidance
documents addressing human health and ecological risk assessments, as well
as analytical tools, such as probabilistic analysis. These documents
supplement applicable EPA- wide guidelines. The Superfund guidelines for
human health risk assessment, for example, cover developing a baseline risk
assessment (Part A), developing or refining preliminary remediation goals
(Part B), performing a risk evaluation of remedial alternatives (Part C),
and standardizing, planning, reporting, and completing a review (Part D).
There are also

other headquarters and regional office documents that further supplement the
program- specific guidelines and manuals.

Office of Solid Waste The Office of Solid Waste, like OERR, is part of
OSWER. OSW regulates the management of solid waste and hazardous waste
through federal programs established by the Resource Conservation and
Recovery Act of 1976, as amended (RCRA). 39 Congress enacted RCRA to protect
human health and the environment from the potential hazards of waste
disposal, conserve

energy and natural resources, reduce the amount of waste generated, and
ensure that wastes are managed in a manner that is protective of human
health and the environment. The act defines solid and hazardous waste,

39 42 U. S. C. 6901- 6991k.

authorizes EPA to set standards for facilities that generate or manage
hazardous waste, and establishes a permit program for hazardous waste
treatment, storage, and disposal facilities. The RCRA hazardous waste
program has a ?cradle to grave? focus, regulating facilities that generate,
transport, treat, store, or dispose of hazardous waste from the moment it is
generated until its ultimate disposal or destruction.

RCRA regulations interact closely with other environmental statutes,
especially CERCLA. EPA notes that both programs are similar in that they are
designed to protect human health and the environment from the dangers of
hazardous waste, but each has a different regulatory focus.

RCRA mainly regulates how wastes should be managed to avoid potential
threats to human health and the environment. On the other hand, according to
EPA, CERCLA is relevant primarily when mismanagement occurs or has occurred,
such as when there has been a release or a substantial threat of a release
in the environment of a hazardous substance. Regulatory activity under RCRA
focuses primarily on specifying procedures and technology to be used to
ensure proper handling and disposal of wastes, but risk assessments play a
role in several supporting tasks, particularly those involving hazardous
waste regulation under RCRA Subtitle C. For example, risk assessment
information may be used in the processes for defining (and delisting)
substances as hazardous wastes, evaluating the hazards posed by waste
streams, assessing the need for

corrective action at disposal sites, and granting waste disposal permits
(such as incinerator permits). In its RCRA Orientation Manual, OSW expressed
an increasing emphasis on making the RCRA hazardous waste program more risk
based (with the intention of ensuring that the

regulations correspond to the level of risk posed by the hazardous waste
being regulated). Major features and characteristics of risk assessment for
hazardous waste regulation are summarized below.

 Making the determination of whether a substance is a hazardous waste is a
central component of the waste management program. The Subtitle C program
includes procedures to facilitate this identification and classification of
hazardous waste. Under the RCRA framework, hazardous wastes are a subset of
solid wastes. 40 In RCRA sect.1004( 5), Congress defined hazardous waste as a
solid waste, or combination of 40 Despite the name, solid wastes actually
may be solids, semi- solids, liquids, or sludges.

solid wastes, which because of its quantity, concentration, or physical,
chemical, or infectious characteristics may:

 cause, or significantly contribute to, an increase in mortality or an
increase in serious irreversible, or incapacitating reversible, illness; or

 pose a substantial present or potential hazard to human health or the
environment when improperly treated, stored, transported, or disposed of, or
otherwise managed.

EPA developed more specific criteria for defining hazardous waste using two
different mechanisms: (1) listing certain specific solid wastes as hazardous
and (2) identifying characteristics (physical or chemical properties) which,
when exhibited by a solid waste, make it hazardous.

The agency has done so, and risk assessment information may be used to
support both mechanisms. 41

? Listed wastes? are wastes from generic industrial processes, wastes from
certain sectors of industry, and unused pure chemical products and
formulations. EPA uses four criteria to decide whether or not to list a
waste as hazardous.

1. The waste typically contains harmful chemicals (and exhibits other
factors, such as risk and bioaccumulation potential) which indicate that it
could pose a threat to human health and the environment in the

absence of special regulation. Such wastes are known as toxic listed wastes.

2. The waste contains such dangerous chemicals that it could pose a threat
to health or the environment even when properly managed. These wastes are
fatal to humans and animals even in small doses and are known as acute
hazardous wastes.

3. The waste typically exhibits one of the four characteristics of hazardous
waste: ignitability, corrosivity, reactivity, and toxicity. 4. EPA has cause
to believe that, for some other reason, the waste

typically fits within the statutory definition of hazardous waste. 41 In
addition, there are some wastes that are specifically excluded from Subtitle
C regulation and wastes that may be exempted when recycled.

Listed hazardous wastes can exit Subtitle C regulation through a
sitespecific delisting process initiated by a petition from a waste handler
to an EPA region or a state. The petition must demonstrate that, even though
a particular waste stream generated at a facility is a listed hazardous
waste, it does not pose sufficient hazard to merit RCRA regulation.

? Characteristic wastes? are wastes that exhibit measurable properties that
indicate they pose enough of a threat to deserve regulation as hazardous
wastes. EPA established four hazardous waste characteristics.

1. Ignitability identifies wastes that can readily catch fire and sustain
combustion.

2. Corrosivity identifies wastes that are acidic or alkaline. Such wastes
can readily corrode or dissolve flesh, metal, or other materials. 3.
Reactivity identifies wastes that readily explode or undergo violent

reactions (e. g., when exposed to water or under normal handling
conditions).

4. Toxicity is used in a rather narrow and specific sense under this program
to identify wastes that are likely to leach dangerous concentrations of
chemicals into ground water if not properly managed (and thus expose users
of the water to hazardous chemicals and constituents). EPA developed a
specific lab procedure, known as the Toxicity Characteristic Leaching
Procedure, to predict whether any particular waste is likely to leach
chemicals into ground water at dangerous levels. In this procedure, liquid
leachate created from hazardous waste samples is analyzed to determine
whether it contains any of 40 different common

toxic chemicals in amounts above specified regulatory levels. The regulatory
levels are based on ground water modeling studies and toxicity data that
calculate the limit above which these toxic compounds and elements will
threaten human health and the environment.

 For OSW, the task of identifying and assessing hazardous wastes is made
more difficult because waste may be in the form of a mixture of
constituents, some of which may be hazardous and some not. (This is also a
common issue for the Superfund program.) The EPA- wide guidelines on
assessments of chemical mixtures therefore could come

into play in OSW risk assessments.

 For dose- response data on the toxicity and potency of hazardous
substances, OSW largely relies on information from other EPA sources. For
example, OSW may use the chemical- specific assessments prepared by ORD,
data in EPA?s IRIS database, and regulatory standards from

other EPA program offices, in particular the Office of Water. However, OSW
combines this information with its own exposure analyses.

 Rhomberg categorized exposure assessment by OSW as either hypothetical or
site specific. He noted that hypothetical exposures principally come into
play when the agency is defining hazardous wastes and evaluating disposal
options. These exposure analyses cover hypothetical waste- handling and
disposal practices anywhere in the

nation, and OSW focuses on the question of whether such practices might
cause undue risks to individuals, not on characterizing the actual
distribution of exposures across the population. One of the principal
concerns in OSW exposure assessments is leaching to groundwater, but

OSW evaluates other exposure pathways from virtually all treatment and
disposal practices, with the specific pathways for any particular analysis
being decided on a case- by- case basis. Site- specific exposure assessments
might be needed when OSW is making regulatory decisions regarding actual
waste disposal facilities, as when assessing the need for remedial action at
a given site or permitting incineration or other disposal activities. In
such cases, the office can focus exposure estimates on the off- site
migration of the particular toxic compounds associated with that location.
In general, an important part of OSW?s

exposure assessments is evaluating the ?relative contribution? of hazardous
wastes to the overall exposure to a hazardous chemical (which is very
similar to assessments by EPA?s Office of Water).

 In exposure assessments, OSW?s deterministic analyses follow EPA?s risk
characterization guidance by setting only two sensitive parameters at high-
end values, with the rest of the parameters being set at their central
tendency values. According to OSW, this approach is meant to produce a risk
estimate above the 90 th percentile of the risk distribution

but still on the actual distribution. Chemical Emergency

CEPPO is also part of OSWER. It provides leadership, advocacy, and
Preparedness and Prevention assistance to: (1) prevent and prepare for
chemical emergencies; (2) Office

respond to environmental crises; and (3) inform the public about chemical
hazards in their community. To protect human health and the environment,

CEPPO develops, implements, and coordinates regulatory and nonregulatory
programs. It carries out this work in partnership with EPA regions, domestic
and international organizations in the public and private sectors, and the
general public.

CEPPO is responsible for the risks associated with accidental chemical
releases. Under the Emergency Planning and Community Right- to- Know Act
(EPCRA) in Title III of the Superfund Amendments and Reauthorization Act of
1986, CEPPO must evaluate, develop, and maintain a list of chemicals and
threshold quantities that are subject to reporting for emergency planning.
In addition, CEPPO develops the emergency reporting and planning
requirements, guidance for industry, and guidance

and tools for use of the reporting information by Local Emergency Planning
Committees. These reporting and planning requirements serve to provide the
necessary information to be used at the local level to manage the risks

associated with accidental chemical releases. CEPPO is also responsible for
accidental chemical release prevention. Under Section 112( r) of the Clean
Air Act, as amended by the Clean Air Act Amendments of 1990, CEPPO must
evaluate chemicals for acute adverse health effects, likelihood of
accidental release, and magnitude of exposure to develop a list of at least
100 substances that pose the greatest risk of causing death, injury, or
serious adverse effects to human health or the

environment from accidental releases. Each listed substance must have a
threshold quantity that takes into account the chemical?s toxicity,
reactivity, volatility, dispersability, combustibility, or flammability.

Facilities handling a listed substance above its threshold quantity must
implement a risk management program and develop a risk management plan. The
risk management program must address a hazards analysis, prevention program,
and emergency response program. According to CEPPO officials, they scaled
these regulatory requirements according to the risk posed by the wide range
of facilities subject to the requirements- the greater the risk, the greater
the risk management responsibilities. The facilities submit their risk
management plans to EPA and to state and local officials for use in
emergency planning and local risk management and reduction. CEPPO
investigates chemical accidents, conducts research, and collects information
about chemical and industrial process hazards to issue Chemical Safety
Alerts and other publications to raise awareness about chemical accident
risks. CEPPO also develops tools, methods, and

guidance necessary to identify and assess the risks to human health from
accidental releases.

Major features and characteristics of CEPPO?s risk assessment procedures are
summarized below.

 The chemical risk assessments conducted by CEPPO are unique from the risk
assessments conducted by other EPA offices. CEPPO?s procedures do not follow
the NAS four- step risk assessment approach, but are similar to the chemical
risk assessment approach used by the Department of Transportation?s (DOT)
Research and Special Programs

Administration (RSPA) in that hazards are identified and a measure of
exposure (or consequence) is determined to yield a ?threat? associated with
an accidental release. While RSPA focuses on risks associated with accidents
involving unintentional releases of hazardous materials during
transportation, CEPPO focuses on risk associated with accidental releases
from a fixed facility. According to CEPPO, for accidental release risks,
because these events are high consequence and low probability, the hazard
and exposure typically can be estimated with some degree of confidence.
However, the likelihood or probability of an

accidental release is very uncertain. Consequently, likelihood is addressed
only in a limited way and the ?threat? is judged to be a surrogate for risk.

 CEPPO?s approaches with respect to chemical accident risk are published
mainly in two rulemakings-? List of Regulated Substances and Thresholds for
Accidental Release Prevention and Risk Management Programs for Chemical
Accident Release Prevention,? 59 FR 4478 (Jan. 31, 1994) and ?Accidental
Release Prevention Requirements: Risk Management Programs under the Clean
Air Act, Section 112( r)( 7),? 61 FR 31668 (June 20, 1996)- and in
guidelines,

especially ?Technical Guidance for Hazards Analysis, Emergency Planning for
Extremely Hazardous Substances,? which was issued jointly by EPA, DOT, and
the Federal Emergency Management Agency (Dec. 1987).

 For hazard identification, CEPPO identifies the hazards that pose a risk
to human health and the environment from an analysis of chemical accidents
and of the physical/ chemical properties of substances that make them more
likely to cause harm as a result of an accidental chemical release. For
example, the catastrophic chemical release in Bhopal, India, in December
1984 involved methyl isocyanate, a chemical that is toxic when inhaled.
CEPPO identified the criteria necessary to

identify those substances that are so toxic that, upon exposure (i. e.,
inhalation, dermal contact, or ingestion) to a small amount, they cause

death or serious irreversible health effects in a short time (acute
toxicity). CEPPO also has developed criteria to identify other substances,
such as highly flammable substances that can trigger a

vapor cloud explosion harming the public and environment. CEPPO is also
working to understand the long- term (chronic) effects that might be
generated by a single acute exposure.

 As part of its identification of hazards, CEPPO also evaluates the
quantity of a chemical that would need to be released and travel off- site
to establish a threshold quantity. 42 If a facility handles more than this
quantity, there is a presumption of risk triggering some action by the
facility?s owner( s) and operator( s). The hazardous chemicals and threshold
quantities identified by CEPPO are published in rulemakings.

 According to CEPPO, the exposure assessment (or consequence analysis)
phase of a chemical accident release assessment is somewhat unique from the
classical risk assessment approaches and procedures. The actual exposure to
humans after an accidental release is often not known. In addition, the
amount and rate of chemical released and the

precise conditions (e. g., weather) are usually not known. However, these
parameters can be estimated using engineering calculations and mathematical
models to generate the concentration likely to have been present or that
could be present in a certain type of accidental release. Using these
techniques, chemicals that possess the physical/ chemical properties most
likely to harm the public or the environment can be

evaluated to estimate the degree of ?threat? that they may pose in an
accidental release.

 CEPPO uses these exposure assessment (consequence analysis) techniques to
understand the potential magnitude of exposure associated with a variety of
hazardous chemicals. In addition, CEPPO publishes the techniques in
guidelines and as software to assist facilities in their assessment of
accidental release risk. According to CEPPO, industry has a fundamental
responsibility to understand the risks associated with chemical accidents.
In addition, the Risk Management

Plan requirements under section 112( r) of the Clean Air Act require that
this information be made available to the public so that industry and the 42
A CEPPO official pointed out that his agency?s focus is the general public
and environment outside the bounds of an industrial facility. The
Occupational Safety and Health Administration is responsible for risk within
the bounds of an industrial facility.

community can work together to manage the risks that might be present.

 CEPPO may characterize the risks associated with accidental releases using
a number of parameters, such as the presence of a large quantity of a highly
hazardous substance in proximity to a large facility that has had a number
of accidental releases in the past. CEPPO uses these

parameters to place more responsibility on such facilities (e. g., greater
accidental release prevention measures under the Risk Management Program
requirements), to investigate the underlying reasons for their accidental
releases, or to assist in audits and inspections of their accident
prevention programs.

Office of Air and Radiation OAR oversees the air and radiation protection
activities of the agency. Radiation risk assessments conducted by OAR are
outside the scope of this report, but chemical risk assessments do have a
part in OAR?s efforts to preserve and improve air quality in the United
States. 43 Such air quality concerns are the primary mission of OAR?s Office
of Air Quality Planning and Standards (OAQPS), which, among other
activities, compiles and reviews air pollution data and develops regulations
to limit and reduce air pollution. The Risk and Exposure Assessment and the
Health and Ecosystem Effects Groups within OAQPS provide the scientific and
analytical expertise to conduct and support human health and ecological risk
assessments in this area, in coordination with ORD.

The Clean Air Act, as amended, provides the statutory basis for air- related
risk assessments by OAR. 44 The CAA requires EPA to establish national
standards for air quality, but it gives states the primary responsibility
for assuring compliance with the standards. Chemical risk assessments are

primarily associated with regulation of (1) criteria air pollutants and (2)
hazardous air pollutants, also referred to as ?air toxics.? 45 43 See
Radiation Standards: Scientific Basis Inconclusive, and EPA and NRC

Disagreement Continues (GAO/ RCED- 00- 152, June 30, 2000) for information
regarding some of the procedures used by EPA in setting radiation standards.

44 42 U. S. C. 7401- 7626. 45 There are other distinct programs within OAR-
for example, focusing on the regulation of mobile sources (including fuels
and fuel additives), acid rain, and global climate change- that also include
consideration of chemicals.

The CAA requires EPA to set health- based air quality standards (National
Ambient Air Quality Standards, or NAAQS) for criteria pollutants, which are
common throughout the United States and mostly the products of combustion.
46 Under the CAA, EPA is also required to review the scientific data upon
which the standards are based and revise the standards, if necessary, every
5 years. The criteria pollutants are particulate matter,

carbon monoxide, sulfur oxides, nitrogen dioxide, ozone, and lead. Of these
pollutants, ozone is not directly emitted by a source, but rather is the
product of the interaction of nitrogen oxide, volatile organic compounds,
and sunlight. Therefore, regulations targeting ozone focus on controlling
emissions of nitrogen oxide and volatile organic compounds. The CAA requires
EPA to set health- based standards with an ?adequate margin of safety,? but
according to EPA it is not required to set air quality standards at a zero-
risk level to achieve an adequate margin of safety, but simply at a

level that avoids unacceptable risks. EPA therefore sets the standards to
protect the substantial part of the national population, including sensitive
or at- risk populations, but not necessarily the most sensitive or exposed

individuals. The CAA also contains provisions, first added in 1970, for the
regulation of emissions to the atmosphere of hazardous air pollutants- toxic
chemicals other than the six criteria pollutants. The 1970 amendments to the
CAA

required EPA to identify and control hazardous air pollutants so as to
achieve ?an ample margin of safety.? However, Congress passed another major
set of amendments, the Clean Air Act Amendments of 1990 (CAAA), which
revised the hazardous air pollutant provisions and substantially affected
the application of risk assessment regarding air toxics. The amendments
explicitly wrote into the act a list of 189 hazardous air pollutants to be
regulated. In addition, the amendments replaced the

former health- based criterion for standards with a criterion that is
primarily technology based, mandating the maximum achievable control
technology (MACT) for the specified list of chemicals. 47 The act further
mandates that EPA evaluate residual risks remaining after implementation

46 The states are responsible for establishing procedures to attain and
maintain the standards. They prepare State Implementation Plans (SIPs) on
these procedures that are submitted to EPA for approval.

47 All new or existing sources of these pollutants are to require the use of
MACT, which is judged to be ?the best of the best? for new sources and at
the top end (best 12 percent) of current emissions control performance for
existing sources.

of the MACT standards to determine if additional standards are needed to
protect the public health with an ample margin of safety.

Additional information on the major features and characteristics of chemical
risk assessments related to these air quality protection activities is
presented below.

Criteria air pollutants

 There are several unique features that affect risk assessments for
criteria air pollutants. Compared to many other agents assessed by EPA, the
agency generally has extensive human data available on health effects at
relevant exposure levels. Therefore, risk assessments for criteria air
pollutants require little extrapolation across species or to low doses and
few default assumptions. These are the least likely of EPA?s risk
assessments to use precautionary or conservative methods and assumptions,
and the results are intended to be unbiased estimates without any built- in
conservatism.

 For criteria air pollutants, ?hazard identification? information on health
effects appears primarily in air quality criteria documents prepared by ORD
and staff papers prepared by OAQPS to support the review and development of
national ambient air quality standards. These documents are intended to
reflect the available scientific evidence on toxicity endpoints of concern.
The definition of what responses constitute ?adverse? outcomes is ultimately
left to the Administrator?s judgment, informed by staff recommendations,
advice from the Clean Air Scientific Advisory Committee (part of EPA?s
Science Advisory Board), and public comments.

 EPA?s principal concern regarding criteria pollutants is for noncancer
health effects. In contrast to most other EPA noncancer risk assessments,
however, EPA does not apply a threshold approach in the case of criteria
pollutants. Instead, the agency models response curves as though they have
no threshold, recognizing that, as a practical matter,

at least some members of the general population will have their thresholds
exceeded at or near the lowest exposure levels. EPA characterizes these
response relationships without any conservative upper- bound methods.
However, probabilistic methods are used to characterize uncertainty in the
fitted exposure- response relationships. In addition, there is temporal
variation in pollution concentrations, so characterization of exposure- time
relationships is also an important

component of EPA?s assessments of criteria pollutants.

 Although EPA?s exposure assessments (and risk characterization) for
criteria pollutants focus on population risks, rather than individual risks,
the agency does consider effects on more sensitive or exposed populations.
Exposure assessments are also affected by the need to establish air quality
standards for both annual and daily concentrations for some pollutants. The
annual standards are intended to provide

protection against typical day- to- day exposures as well as longer- term
exposures, while the daily standards are intended to provide protection
against days with high peak concentrations of pollutants. EPA?s exposure
assessments therefore need to address these types of variations. Rhomberg
noted that, because of the long history of analysis of standard pollutants,
EPA?s exposure modeling has been continually improved and expanded,
resulting in sophisticated models with

capabilities well beyond models used in other situations that do not have
the benefit of decades of experience and application.

 Finally, it is important to recognize that one of the most important uses
of risk assessments regarding criteria air pollutants is to characterize the
population exposure levels and health effects that would be expected given
various specified air quality criteria. In other words, one of the primary
uses of risk assessment is to estimate what the effects would be if
standards were set at various specified levels, rather than using the tool
simply to estimate what health risks these pollutants pose.

Hazardous air pollutants (air toxics)

 Although the Clean Air Act Amendments of 1990 shifted the focus in
hazardous air pollutant regulation to technology- based controls, several
activities may still involve risk assessments, including

 listing and delisting of hazardous air pollutants, which depends on
whether a chemical may present a threat of adverse effects to humans and the
environment;

 de minimis delisting of source categories, which requires sources be
listed unless they pose less than a 10 -6 risk to the maximally exposed
individual (MEI);

 triggering the consideration of further regulation to address residual
risks that remain after applying MACT standards (triggered if the MEI
suffers a 10 -6 or greater lifetime risk); and

 offset trading of one pollutant for another based on whether the increase
in emissions is offset by an equal or greater decrease in a

?more hazardous? air pollutant.

 According to section 112( o) of the amended CAA, prior to the promulgation
of any residual risk standard, EPA shall revise its guidelines for
carcinogen risk assessment or provide an explanation of the reasons
regarding any NAS report section 112( o)( 4) recommendations that have not
been implemented.

 The amended act also had a major impact on hazard identification for air
toxics. The amendments defined hazardous air pollutants as air pollutants
listed pursuant to section 112( b) of the act. Section 112( b) included an
initial list of 189 compounds incorporated by reference into the law.

 Dose- response analysis for air toxics has in the past been done largely
through Health Assessment Documents produced by ORD for the air office,
according to the methods discussed in the earlier section on EPAwide risk
assessment procedures. Carcinogen potency calculations for

de minimis delisting and residual risk determination will be done under the
revised carcinogen assessment guidelines, once they are finalized. EPA
addresses noncancer risks for hazardous air pollutants with its usual
methodologies (e. g., NOAEL/ LOAEL, benchmark dose, or others).

 With the 1990 amendments, exposure assessments for air toxics will focus
on assessing the residual risk for the most exposed individual after MACT
has been applied. OAR uses a population- based risk assessment to generate
estimates of how risks are distributed within the population, not just for
specific conservative scenarios. According to Rhomberg (and confirmed by OAR
officials), OAR?s intent is to define the actual most exposed person in the
population, rather than a

hypothetical person with an unrealistically high estimated exposure. EPA has
adopted a tiered approach to analyzing residual risk consistent with the
recommendations from NAS and the Presidential/ Congressional Commission. 48
In the screening phase, default conservative assumptions are used to ensure
that risks will not be underestimated. Sources and hazardous air pollutants
that exceed

some benchmark in the screening analysis will be evaluated further.
According to OAR, the more refined assessments will utilize more
sitespecific information and more realistic assumptions, especially as they
relate to exposure. EPA estimates exposures to air toxics using a general-
purpose model largely based on fate and transport considerations for stack
emissions. 49 OAR officials noted that they are updating their modeling
methodology, updating their Human Exposure

Model with the current state- of- the- art dispersion model (ISCST3), and
will be updating the census data they use with the 2000 Census numbers when
they become available.

Office of Water OW is responsible for the agency?s water quality activities,
including development of national programs, technical and science policies,
regulations, and guidance relating to drinking water, water quality, ground
water, pollution source standards, and the protection of wetlands, marine,
and estuarine areas. Chemical risk assessments are associated, in
particular, with EPA?s ambient water quality criteria, under the CWA, and
drinking water quality regulations, under the SDWA. 50

The goal of CWA is to maintain and improve the cleanliness and biological
integrity of the nation?s waters, including lakes, rivers, and navigable
waters. Under CWA, EPA publishes water quality criteria defining the 48 See
Residual Risk Report to Congress, U. S. EPA (March 1999). 49 Fate and
transport models are mathematical descriptions of the movement and
transformation of substances through various media, such as air, soil, and
water.

50 CWA is codified generally as 33 U. S. C. 1251- 1387 and SDWA as 42 U. S.
C. 300f- 300j.

degree of water quality that is compatible with intended uses and states of
different water bodies. The criteria are health based, but they are not
rules and are themselves unenforceable. States use these criteria as
guidance for developing state water quality standards and setting
enforceable limits in permits for facilities that discharge pollutants into
surface waters. CWA distinguishes ?conventional? from ?toxic? pollutants.
Toxic water pollutants are evaluated as exposures to toxic chemicals
(similar to EPA?s

treatment of hazardous air pollutants). The goal of SDWA is to protect the
quality of public drinking water systems. The law focuses on all waters
actually or potentially designed for drinking use, whether from above ground
or underground sources. SDWA requires

EPA to set drinking water standards to control the level of contaminants in
drinking water provided by public water systems, which the water systems are
required to meet. 51 Congress passed extensive amendments to SDWA through
the Safe Drinking Water Act Amendments of 1996 (PL 104- 182). Among other
key changes, the amendments increased regulatory flexibility, focused
regulatory efforts on contaminants posing the greatest health risks, and
added risk assessment and risk communication provisions to

SDWA. There are several risk- related mandates in these acts.

 Under CWA, EPA is to establish criteria for water quality solely on the
basis of health and ecological effects and ?accurately reflecting the latest
scientific knowledge on the kind and extent of all identifiable effects on
health and welfare.? CWA defines a toxic pollutant as one that after
discharge and upon exposure, ingestion, inhalation, or

assimilation into any organism, either directly from the environment or
indirectly by ingestion through food chains, will, on the basis of
information available to the Administrator, cause death, disease, behavioral
abnormalities, cancer, genetic mutations, physiological malfunctions
(including malfunctions in reproduction), or physical

deformities in such organisms or their offspring. Federal water quality
criteria are unenforceable, but states develop enforceable permit limits 51
Drinking water standards apply to public water systems that provide water to
at least 15 connections used by year- round residents or regularly serve at
least 25 year- round residents. Private water wells serving fewer than 25
persons are not covered by these federal standards. EPA also sets Secondary
Drinking Water Regulations, which are nonenforceable guidelines for
contaminants that may cause cosmetic or aesthetic effects.

based on them.

 In contrast to the unenforceable federal water quality criteria, CWA also
provides for the promulgation of enforceable federal performance standards
for sources of effluent (waste discharged into a river or other water body)
that do include consideration of technological and economic feasibility.
Since 1977, establishment of effluent standards for toxic pollutants has
been based on the best available technology (BAT)

economically achievable by particular source category. The compounds to be
regulated are specified in a list, and there are provisions for additions
and deletions to the list. Standards must be at that level which the
Administrator determines provides ?an ample margin of safety,? so that
standards more stringent than BAT may be named at EPA discretion.

 Under SDWA, the EPA Administrator is to ?promulgate national primary
drinking water regulations for each contaminant which may have any adverse
effect on the health of persons and which is known or anticipated to occur
in public water systems.? An important feature of such regulations, however,
is that a standard specifies two levels of

contamination. First, a maximum contaminant level goal (MCLG) is set solely
on health grounds ?at a level at which no known or anticipated adverse
effects on the health of persons occur and which allows an adequate margin
of safety.? For each such goal there is also a maximum contaminant level
(MCL). This MCL is to be as close to the MCLG ?as is feasible,? where
feasible means ?with the use of the best technology, treatment techniques
and other means which are available (taking cost into consideration).? The
MCL is the enforceable standard.

 The 1996 amendments to SDWA added several provisions that increased the
importance of risk assessment and risk communication in EPA?s regulation of
drinking water quality. For example, the amendments

 Require EPA, when developing regulations, to (1) use the best available,
peer- reviewed science and supporting studies and data and (2) make publicly
available a risk assessment document that discusses estimated risks,
uncertainties, and studies used in the assessment.

 Require EPA to conduct a cost- benefit analysis for every new standard to
determine whether the benefits (health risk reduction) of a drinking water
standard justify the costs.

 Permit consideration of ?risk- risk? issues by authorizing EPA to set a
standard other than the feasible level if the feasible level would lead

to an increase in health risks by increasing the concentration of other
contaminants or by interfering with the treatment processes used to comply
with other SDWA regulations.

 Require EPA to review and revise, as appropriate, each national primary
drinking water regulation promulgated by the agency at least every 6 years.
Of particular relevance to the use of risk assessment information, any
revisions must ?maintain, or provide for greater, protection of the health
of persons.?

 Require EPA to identify subpopulations at elevated risk of health effects
from exposure to contaminants in drinking water and to conduct studies
characterizing health risk to sensitive populations from contaminants in
drinking water.

Additional information on major features and characteristics of chemical
risk assessments related to water quality protection activities is presented
below.

 The various offices within OW- the Office of Ground Water and Drinking
Water; Office of Science and Technology; Office of Wastewater Management;
and Office of Wetlands, Oceans, and

Watersheds- have developed extensive technical and analytical guidance on
water quality monitoring and the development of water quality criteria. One
recently finalized document particularly relevant for describing OW?s
current risk assessment procedures is the revision to the methodology for
deriving ambient water quality criteria (AWQC) for the protection of human
health. 52 Published pursuant to section 304( a)( 1) of the CWA, OW noted
that this revised methodology supersedes EPA?s 1980 guidelines and
methodology on this subject. In addition to describing OW?s approach to
developing new and revising existing AWQC, it defines the default factors
that EPA will use in evaluating and determining consistency of state water
quality standards with the requirements of the CWA.

 Although there are different statutory bases and risk mandates for the
regulation of ambient and drinking water, OW?s risk assessment procedures in
support of CWA and SDWA are mostly similar. 52 Methodology for Deriving
Ambient Water Quality Criteria for the Protection of Human

Health, also referred to as the ?2000 Human Health Methodology,? EPA- 822-
B- 00- 005 (October 2000). The agency is also in the process of revising its
methodology for deriving AWQC for the protection of aquatic life.

However, risk assessments in support of CWA consider not just human health
effects but also the ecological effects associated with exposure to
pollutants. With regard to human health risks, perhaps the most notable
difference between the ambient water and drinking water parts of OW is the
additional focus, during exposure assessments for CWA purposes, on exposures
to contaminated water through consumption of contaminated fish or shellfish.
(This is a primary reason for potential differences in the resulting
drinking water and ambient water quality criteria or standards for the same

chemical.)

 OW?s Office of Science and Technology does all of the risk assessments for
SDWA maximum contaminant level goals and CWA?s AWQC. For cancer risk
evaluation, OW has been applying the principles in EPA?s proposed revision
of the carcinogen guidelines.

 For hazard identification purposes, SDWA originally had specified a list
of compounds to be regulated as toxic pollutants and required EPA to
regulate an additional 25 contaminants every 3 years. However, the 1996
amendments eliminated that requirement and

revised OW?s approach for listing, reviewing, and prioritizing the drinking
water contaminant candidate list. The new risk- based contaminant selection
process provides EPA the flexibility to decide whether or not to regulate a
contaminant after completing a required review of at least five contaminants
every 5 years. EPA must use three risk- related criteria to determine
whether or not to regulate:

(1) that the contaminant adversely affects human health; (2) it is known or
substantially likely to occur in public water systems with a frequency and
at levels of public health concern; and (3) regulation of the contaminant
presents a meaningful opportunity for health risk reduction. The 1996
amendments also included specific requirements to assess health risks and
set standards for arsenic,

sulfate, radon, and disinfection byproducts.

 There are a number of important features regarding OW?s exposure
assessments in support of CWA and SDWA regulations.

 OW?s primary exposure question during the criteria/ standard- setting
process for drinking or ambient water is hypothetical: What health effects
might be expected if people consumed water and/ or finfish and shellfish
contaminated at the level of a candidate standard? The main function of
exposure assessment is to link criteria or water

concentrations to doses of chemicals and the associated health effects that
might be projected.

 For its exposure assessments, OW uses estimates of water and food
ingestion in the United States based on a variety of surveys and studies.
One of the major sources of per capita water and fish ingestion is the
Department of Agriculture?s Continuing Survey of Food Intakes by Individuals
(CSFII), which presents results for the general population and for certain
subpopulations (e. g., pregnant and lactating women, children). 53

 For assessing standards under SDWA, the linking of water concentration to
dose is conducted through standardized consumption values. For example, the
default exposure scenario of lifetime consumption by individuals is 2 liters
of water per day. However, OW uses other default values to address
consumption by sensitive subpopulations, especially children and infants.
For assessing AWQC under the CWA, EPA uses the same water consumption rate
as under SDWA. In addition, though, the agency adds the dose resulting from
the daily average consumption of 17.5

grams of fish.  An important change in EPA?s approach for developing AWQC,
reflected in the 2000 Human Health Methodology, has been the move toward use
of a bioaccumulation factor (BAF) to estimate potential human exposure to
contaminants via the consumption of contaminated fish and shellfish. BAFs
reflect the accumulation of chemicals by aquatic organisms from all
surrounding media (e. g., water, food, and sediment). EPA?s 1980 method used
a bioconcentration factor that reflected only absorption directly out of the
water column, and therefore tended to underestimate actual contaminant
levels in fish and shellfish. EPA?s revised methodology also gives
preference to the use of high- quality field data over laboratory or model-
derived estimates of BAFs.

 OW considers indirect exposures to a substance from sources other than
drinking water (e. g., food and air) when establishing AWQC. This is
particularly important for noncarcinogens, where the fact that 53 The CSFII
is a complex multistage probability sample of the entire United States
conducted to survey food and beverage intake.

several exposure sources might individually be below the RfD level does not
mean that collectively the exposure is below this presumably safe level. OW
has revised and expanded its policy on accounting for nonwater sources of
indirect exposures known as the ?relative source contribution.? The
procedures for calculating the

relative source contribution vary depending on the adequacy of available
exposure data, levels of exposure, sources and media of exposure relevant to
the pollutant of concern, and whether there are multiple health- based
criteria or standards for the same pollutant.

(See table 5 in the next section for a more detailed description of these
assumptions.) Risk Assessment

EPA?s risk assessment guidelines and other related documents identify
Assumptions and many default assumptions, standardized data factors, and
methodological choices that may be used in chemical risk assessments. As
pointed out by Methodological NAS, assumptions and professional judgment are
used at every stage of a

Choices risk assessment, because there are always uncertainties in risk

assessments that science can not directly answer. For the most part, these
assumptions and choices are intended to address various types of
uncertainties- such as an absence or limited amount of available data,

model uncertainty, and gaps in the general state of scientific knowledge- or
variability in the population. They are also intended to provide some
consistency and transparency to agency risk assessments. Defaults are
generally used in the absence of definitive information to the contrary, but
also reflect policy decisions.

In its guidelines, EPA characterizes many of its choices as conservative or
public- health protective in that they are intended to help the agency avoid
underestimating possible risks. Agency guidelines often cited the scientific
studies and other evidence that supported the agency?s choice and the
plausibility of the resulting risk estimates. In our recent report on EPA?s
use of precautionary assumptions, we identified three major factors
influencing the agency?s use of such assumptions: (1) EPA?s mission to
protect human health and safeguard the natural environment (including
specific requirements in some of the underlying environmental statutes),

(2) the nature and extent of relevant data, and (3) the nature of the risk
being evaluated. 54 54 Environmental Protection Agency: Use of Precautionary
Assumptions in Health Risk Assessments and Benefits Estimates (GAO- 01- 55,
Oct. 16, 2000).

EPA?s program offices commonly employ tiered risk assessment approaches that
progress from rough screening assessments (for which only limited data may
be available) through increasingly detailed and rigorous analyses, if
needed. EPA?s guidelines and program- specific documents indicate that
conservative default assumptions are most often

used during initial screening assessments, when the primary task is to
determine whether a risk might exist and further analysis is called for.
Such screening assessments may use ?worst case? assumptions to determine
whether, even under those conditions, risk is low enough that a potential
problem can be eliminated from further consideration. According to
guidelines and related descriptive materials from the program offices,
conservative assumptions are used less often in later tiers, as the agency
attempts to gather and incorporate more detailed data into its analyses.

Several circumstances may lead to conservative choices playing a less
prominent role in EPA risk assessments. For example, the development of more
complex and sophisticated models for cancer and noncancer effects places
more emphasis on using the full range of available data and characterizing
the full range of potential adverse outcomes and effects. Similarly, the
increased use of probabilistic analytical methods to derive

parameter values will tend to reduce the ?compounding? effect of picking
conservative point values for each factor. As noted above, the use of tiered
risk assessment approaches may also limit the use of default assumptions if
more rigorous and case- specific analysis is done beyond initial screening
assessments. However, all of these developments may require substantial
additional effort and the availability of considerable data, which might not
be possible in many cases.

Although not intended to be comprehensive, table 5 illustrates in detail
some of the specific assumptions, default data values, or methodological
choices that are used in EPA chemical risk assessments. The table
concentrates primarily on default choices from EPA?s various agencywide risk
assessment guidelines. However, to also provide a sense of how default
choices are used at the program level, we have included examples of standard
assumptions and values employed by two of EPA?s program offices. One set of
examples illustrates assumptions and choices used by OPP. The second set
presents more detailed descriptions of the standard assumptions and choices
identified in OW?s risk assessment methodology

for deriving AWQC for the protection of human health. OW?s policy reflects
many of the same basic choices that would apply to assessments conducted
across the agency, such as the use of uncertainty factors when estimating an
RfD.

To the extent that EPA?s documents identified for each of these assumptions
or choices a reason for its selection, when it would be applied in the risk
assessment process, and its likely effect on risk assessment results, we
have reported that information. However, it is important to recognize that
there is no requirement that agencies provide such

information in their guidelines (or even that they have guidelines). In
particular with regard to the ?likely effects? column, EPA officials
cautioned that it is not always appropriate to characterize a single
assumption separate from the rest and that it is not always possible to

quantify the effect of each default assumption. They noted that, in general,
their default assumptions are intended to be public- health protective.

The information presented in table 5 was taken primarily from EPA risk
assessment guidelines and related documents but also reflects additional
comments provided by EPA officials. (GAO notes and comments appear in
parentheses.)

Table 5: EPA Risk Assessment Assumptions and Methodological Choices When the
assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

1. Agencywide (ORD) proposed carcinogen risk assessment guidelines

A carcinogen is a substance or agent that produces or incites cancerous
growth. The following items reflect major assumptions or choices identified
in the 1999 version of EPA?s proposed revision to its 1986 carcinogen risk
assessment guidelines.

1.1 (The discussion setting up this (General framework outlined for

(Not identified in the The guidelines use a combination of

framework refers to reports by the determining whether or not to use
proposed guidelines.)

principles and process to describe a National Academy of Sciences? default
assumptions.)

general framework for the application NRC. In particular, the guidelines of,
or departure from, default note that, in 1994, NRC supported assumptions.

continued use of default assumptions as a reasonable way The proposed
guidelines state that the to deal with uncertainty about decision to use a
default, or not, is a

underlying mechanisms in choice considering available selecting methods and
models for information on an underlying scientific use in risk assessment
but also process and agent- specific data, recommended that EPA should
depending on which kind of data it is.

consider attempting to give Generally, greater formality to its criteria for
a departure from default options.)

 If a gap in basic understanding exists, or if agent- specific data are
missing,

The framework of default the default is used without pause.

assumptions allows risk assessment to proceed when

 If data are present, their evaluation current scientific theory or may
reveal inadequacies that also available case- specific data do not lead to
use of the default.

provide firm answers in a particular case.

 If data support a plausible alternative to the default, but no more
strongly than they support the default, both

the default and its alternative are carried through the assessment and
characterized by the risk manager.

 If the alternative to the default is strongly supported by data, the
alternative may be used in place of the default.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

1.2 (Not identified in the proposed

When determining whether the The guidelines caution When cancer effects in
exposed guidelines, but this is a standard

presence or absence of effects that studies either humans are attributed to
exposure to assumption in risk assessment.)

observed in a human population is attributing cancer an exogenous agent, the
default predictive of an agent posing a effects in humans to assumption is
that such data are

carcinogenic hazard to other exogenous agents or

predictive of cancer in any other exposed humans. reporting no effects
exposed human population.

are often studies of occupationally exposed humans and,

therefore, not representative of the general population exposed
environmentally to the same agents. Therefore, the guidelines state that

this assumption does not err on the side of public- health conservatism,
because it could still underestimate the response of certain sensitive human
subpopulations.

1.3 Epidemiologic studies usually

When determining whether the (Not identified in the When cancer effects are
not found in have low power to detect and presence or absence of effects
proposed guidelines.) an exposed human population, this attribute responses
and typically observed in a human population is information by itself is not
generally

evaluate cancer potential in a predictive of an agent posing a sufficient to
conclude that the agent restricted population (e. g., by age, carcinogenic
hazard to other poses no carcinogenic hazard to this or occupation, etc.).

exposed humans. other populations of potentially exposed humans, including
sensitive subpopulations.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

1.4 The assumption is supported by When determining whether the This
assumption is a Positive effects in animal cancer

the fact that nearly all of the presence or absence of effects public-
health

studies indicate that the agent under agents known to cause cancer in
observed in an animal population

conservative policy, study can have carcinogenic potential humans are
carcinogenic in is predictive of an agent posing a and it is both in humans.

animals in tests with adequate carcinogenic hazard to exposed appropriate
and protocols [citations provided].

humans, if no adequate human necessary given that  To demonstrate that a
response in

Moreover, almost one- third of data are present. we do not test for

animals is not relevant to any human human carcinogens were carcinogenicity
in situation, adequate data to assess identified subsequent to animal

Available mode of action humans. the relevancy issue must be testing
[citations provided]. information is studied for its available.

Further support is provided by implications in both hazard and research on
the molecular biology

dose- response assessment and its of cancer processes, which has effect on
default assumptions. shown that the mechanisms of control of cell growth and
differentiation are remarkably homologous among species and

highly conserved in evolution.

 (Relevancy issue) There may be instances in which the use of an animal
model would identify a

hazard in animals that is not truly a hazard in humans [citation provided].
1.5

Animal studies are conducted at When determining whether the (Not identified
in the Effects seen at the highest dose tested high doses in order to
provide presence or absence of effects guidelines.) are appropriate for
assessment, but it statistical power, the highest dose observed in an animal
population is necessary that the experimental being one that is minimally
toxic. is predictive of an agent posing a conditions be scrutinized.

Consequently, the question often carcinogenic hazard to exposed arises
whether a carcinogenic humans, if no adequate human

 If adequate data demonstrate that the effect at the highest dose may be
data are present. effects are solely the result of a consequence of cell
killing with excessive toxicity rather than compensatory cell replication or
of This is a matter of expert carcinogenicity of the tested agent general
physiological disruption, judgment, considering all of the

per se, then the effects may be rather than inherent data available about
the agent, regarded as not appropriate to carcinogenicity of the tested
including effects in other toxicity include in assessment of the potential

agent. There is little doubt that studies, structure- activity for human
carcinogenicity of the

this may happen in some cases, relationships, and effects on agent. but
skepticism exists among some growth control and differentiation. scientists
that it is a pervasive problem [citations provided].

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

1.6 (Not identified in the guidelines. When determining whether the The
guidelines note When cancer effects are not found in Instead, the discussion
focuses on

presence or absence of effects that this default well- conducted animal
cancer studies

the limitations of this default observed in an animal population assumption
about lack in two or more appropriate species and assumption. EPA notes, for

is predictive of an agent posing a of cancer effects has other information
does not support the

example, that it is recognized that carcinogenic hazard to exposed
limitations. For

carcinogenic potential of the agent, animal studies and epidemiologic
humans, in the absence of human example, in some these data provide a basis
for

studies have very low power to data to the contrary. situations, the tested
concluding that the agent is not likely detect cancer effects and, in some
animal species may to possess human carcinogenic

situations, the tested animal not be predictive of potential, in the absence
of human

species may not be predictive of effects in humans. data to the contrary.

effects in humans (e. g., arsenic).) Therefore, the guidelines discuss the
importance of using supplementary data to support conclusions

that negative results in animal studies indicate a lack of human hazard. 1.7

Target organs of carcinogenesis When determining whether the This is a
public- health Target organ concordance is not a for agents that cause
cancer in presence or absence of effects conservative science prerequisite
for evaluating the

both animals and humans are observed in an animal population policy.

implications of animal study results for most often concordant at one or

is predictive of an agent posing a humans.

more sites [citations provided]. carcinogenic hazard to exposed However,
concordance by site is humans.

 An exception to the basic default of not uniform. not assuming site
concordance exists ?It is appropriate under these in the context of
toxicokinetic

guidelines to consider the modeling. Site concordance is influences of route
of exposure, inherently assumed when these

metabolism, and, particularly, models are used to estimate hormonal modes of
action that delivered dose in humans based on may either support or not
support animal data.

target organ concordance between animals and humans. When data allow, these
influences are considered in deciding whether the default remains
appropriate in individual instances.?

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

1.8 This default is consistent with the When determining whether the This is
a science policy The default is to include benign tumors

approach of the National presence or absence of effects decision that is
observed in animal studies in the

Toxicology Program and the observed in an animal population somewhat more

assessment of animal tumor incidence International Agency for Research is
predictive of an agent posing a conservative of public if they have the
capacity to progress to

on Cancer and is somewhat more carcinogenic hazard to exposed health than
not

the malignancies with which they are protective of public health than not

humans. including benign

associated. including benign tumors in the

tumors in the assessment. This treats the In assessing findings from animal
assessment. benign and malignant tumors as studies, a greater proportion of
representative of related

malignancy is weighed more responses to the test agent, which heavily than a
response with a is scientifically appropriate greater proportion of benign

[citation provided]. tumors.

1.9 (Not identified in the proposed

When determining whether the (Not identified in the Benign tumors that are
not observed to guidelines.) presence or absence of effects proposed
guidelines.) progress to malignancy are assessed observed in an animal
population on a case- by- case basis.

is predictive of an agent posing a carcinogenic hazard to exposed humans.

1.10 (Not identified in the proposed

When extrapolating from animal (Not identified in the There is a similarity
of the basic

guidelines.) studies and considering how proposed guidelines.) pathways of
metabolism and the metabolic pathways relate across occurrence of
metabolites in tissues in species.

regard to the species- to- species extrapolation of cancer hazard and risk.
1.11

This adjustment factor is used When estimating human (Not identified in the
For oral exposure, a human equivalent because it represents scaling of

equivalent doses in extrapolating proposed guidelines. dose for adults is
estimated from data metabolic rate across animals of

from animal studies. However, from other

on another species by an adjustment different sizes. (Also see reason
sources, this is of animal applied oral dose by a

cited under next assumption, and Because the factor adjusts for a

generally considered scaling factor of body weight to the note that an
interagency parameter that can be improved on to provide the midpoint 0. 75
power. committee recommended this and brought into more

of plausible values.) scaling factor as a default sophisticated
toxicokinetic

 The same factor is used for children. approach for federal agencies.)

modeling, when such data become available, the default assumption

 The same factor is used for of 0. 75 power can be refined or

children because it is slightly replaced. more protective than using
children?s body weight.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

1.12 This default assumption, like the

When extrapolating from animal Health conservatism is For inhalation
exposure, a humanequivalent one with oral exposure, is selected studies and
considering how not an element in dose for adults is estimated in part
because it lays a

toxicokinetic processes relate choosing the default. by default
methodologies that provide

foundation for incorporating better across species. estimates of lung
deposition and of data. The use of information to internal dose.

improve dose estimation from applied, to internal, to delivered  Because of
the differences for infants dose is encouraged, including use and children,
for gases and aerosols,

of toxicokinetic modeling instead an adjustment is made for their of any
default, where data are breathing rate and their body weight.

available.

 The guidelines point out that the processes of absorption, distribution,
and elimination have important differences among infants and adults
[citation provided].

1.13 The rationale is that for internal When extrapolating from one route
This is a qualitative

For a route- to- route exposure tumors an internal dose is of exposure to
another route and

assumption and is extrapolation, the default assumption is significant no
matter what the considering how toxicokinetic

considered to be that an agent that causes internal route of exposure.
Additionally, processes relate across species.

public- health tumors by one route of exposure will be the metabolism of the
agent will be conservative. carcinogenic by another route if it is
qualitatively the same for an This is a qualitative assumption. absorbed by
the second route to give internal dose.

The issue of quantitative an internal dose. extrapolation of the dose-
response relationship from one route to another is addressed case by case.
Adequate data are

necessary to demonstrate that an agent will act differently by one route
versus another route of exposure.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

1.14 While no standard biologically When assessing the correlation of (See
more specific If sufficient data are available, a

based models are in existence, the observed dose- response assumptions
biologically based model for both the

one may be developed if extensive relationship to the relationship at
described below.)

observed range and extrapolation data exist in a particular case and lower
doses. [Low- dose

below that range may be used. the purpose of the assessment extrapolation
during doseresponse

justifies the investment of assessment.] When a biologically based model is
not

resources needed. used,

In the absence of data supporting

 The default procedure for the (See more specific assumptions a
biologically based model for observed range of data is to use a described
below for rationales extrapolation outside of the curve- fitting model for
incidence data.

regarding extrapolation when a observed range, the choice of  For
extrapolation outside of the

biologically based model is not approach is based on the view of

observed range, the choice of used.) the mode of action of the agent

approach is based on the view of arrived at in the hazard mode of action of
the agent arrived at assessment. in the hazard assessment (covered by items
1.15 through 1.17 below).

1.15 (Although not mentioned in the

Low- dose extrapolation during This approach is When the mode of action
information is proposed guidelines, the dose- response assessment, in the

generally considered supportive of linearity or mode of assumption of
linearity for absence of data supporting a to be public- health action is
not understood, the basic suspected carcinogens biologically based model.

conservative. The default is to assume linearity and use a

traditionally has been a standard linear default is linear default approach.
default assumption for EPA.)

thought to generally produce an upper The linear approach is to draw a The
LED 10 is the lower 95- percent

bound on potential risk straight line between a point of limit on a dose
that is estimated to at low doses. This departure from observed data,

cause a 10- percent response. upper bound is thought generally, as a
default, the LED 10 , and

This level is chosen to account to cover the range of the origin (zero
incremental dose, zero (conservatively) for experimental human variability
incremental response). Other points of variability. Additionally, it is
although, in some departure may be more appropriate for chosen because it
rewards

cases, it may not certain data sets; these may be used experiments with
better designs in completely do so. EPA instead of the LED 10 .

regard to number of doses and considers the linear

dose spacing, since these default to be inherently generally will have
narrower

conservative of public confidence limits. It is also an health, without
appropriate representative of the addition of another lower end of the
observed range factor for human because the limit of detection of
variability. studies of tumor effect is about 10 percent.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

1.16 In the nonlinear approach, the

Low- dose extrapolation during (Not identified in the When adequate data on
mode of

margin that exists between a dose- response assessment, in the

proposed guidelines.) action show that linearity is not

human exposure of interest and absence of data supporting a plausible, and
provide sufficient

the point of departure is examined biologically based model. evidence to
support a nonlinear mode

for adequacy to protect public of action for the general population and
health. A margin of exposure As noted in the first column, this any
subpopulations of concern, the analysis may be used as the basis

default is to be used when default changes to a different to consider the
protectiveness of a adequate data on mode of action approach- a margin of
exposure possible environmental criterion

(1) show that linearity is not analysis- which assumes that

for regulation or to judge whether plausible and (2) provide sufficient
nonlinearity is more reasonable. (A an existing exposure might

evidence to support a nonlinear margin of exposure analysis compares present
risk.

mode of action. The guidelines the dose at the point of departure with also
state that a sufficient basis to the dose associated with the According to
the guidelines, the support this nonlinear procedure environmental exposure(
s) of interest NOAEL/ LOAEL procedures are will include data on responses
that by computing the ratio between the used with continuous data

are key events integral to the two.) because modeling approaches for

carcinogenic process. This means deriving a point of departure from that the
point of departure mostly The departure point is again generally continuous
data are not yet

will be from these precursor data the LED 10 when incidence data are

available. rather than tumor incidence data. modeled. When the data
available are continuous data, such as blood levels of hormones or organ
weight, a

NOAEL/ LOAEL procedure is typically used. 1.17 (No additional discussion in
the

Low- dose extrapolation during (No additional When the mode of action
information proposed guidelines. See

dose- response assessment, in the discussion in the

indicates that the dose response may previous assumptions regarding absence
of data supporting a proposed guidelines. be adequately described by both a
linear and nonlinear approaches.)

biologically based model. See previous linear and a nonlinear approach, then
assumptions regarding the default is to present both the linear

As noted in the first column, this linear and nonlinear and margin of
exposure analyses.

default is to be used when mode of approaches.)

action information indicates that both a linear and a nonlinear approach may
adequately describe the dose response. 1.18

This assumes that a high dose of When assessing the correlation of This is
thought to be a A default assumption is made that

such an agent over a shorter the observed dose- response

relatively public- health cumulative dose received over a period of time is
equivalent to a relationship to the relationship at conservative lifetime,
expressed as a lifetime low dose spread over a lifetime.

lower doses. [Low- dose assumption.

average daily dose, is an appropriate This assumption has empirical
extrapolation during doseresponse

measure of dose (exposure to a support [citation provided].

assessment.] carcinogen).

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

2. Agencywide (ORD) guidelines for neurotoxicity risk assessment
Neurotoxicity is an adverse change in the structure or function of the
central and/ or peripheral nervous system following exposure to a chemical,
physical, or biological agent. 2.1

This assumption is based on the When extrapolating data from These
assumptions

It is assumed that an agent that comparisons of data for known animal
studies to humans.

are ?plausibly

produces detectable adverse human neurotoxicants [citations conservative? in
that neurotoxic effects in experimental

provided], which indicate that Generally applied in the absence they are
protective of animal studies will pose a potential

experimental animal data are of data on the relevance of effects public
health and are hazard to humans. frequently predictive of a

to potential human risk. also well founded in neurotoxic effect in humans.

scientific knowledge about the effects of concern. 2.2

In the past, the tendency has been When extrapolating data from (See general
It is assumed that behavioral, to consider only neuropathological animal
studies to humans.

statement about these neurophysical, neurochemical, and

changes as endpoints of concern. assumptions cited neuroanatomical
manifestations are of Based on data on agents that are

Generally applied in the absence above.) concern.

known human neurotoxicants of data on the relevance of effects [citations
provided], there is at to potential human risk. A biologically significant
increase in

least one experimental species any of the manifestations is considered that
mimics the types of effects indicative of an agent?s potential for seen in
humans, but in other disrupting the structure or function of

species tested, the neurotoxic the human nervous system. effect may be
different or absent.

2.3 The fact that every species may

When extrapolating data from (See general It is assumed that the neurotoxic
not react in the same way is animal studies to humans.

statement about these effects seen in animal studies may not probably due to
species- specific assumptions cited always be the same as those produced
differences in maturation of the

Generally applied in the absence above.) in humans. Therefore, it may be
nervous system, differences in of data on the relevance of effects difficult
to determine the most timing of exposure, metabolism, or to potential human
risk.

appropriate species in terms of mechanisms of action. predicting specific
effects in humans. 2.4

This is based on the assumption When extrapolating data from Provides a

It is also assumed that, in the absence that humans are as sensitive as
animal studies to humans. conservative estimate of data to the contrary, the
most the most sensitive animal species of sensitivity. (Also

sensitive species is used to estimate tested. This provides a Generally
applied in the absence see general statement human risk.

conservative estimate of of data on the relevance of effects about these
sensitivity for added protection to to potential human risk.

assumptions cited the public. above.)

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

2.5 Although there may be a threshold

When extrapolating data from (See general As with other noncancer endpoints,
it is for neurotoxic effects, these are animal studies to humans.

statement about these assumed that there is a nonlinear

often difficult to determine assumptions cited dose- response relationship
for

empirically. Therefore, a nonlinear Generally applied in the absence above.)
neurotoxicants.

relationship is assumed to exist for of data on the relevance of effects
neurotoxicants.

to potential human risk.

3. Agencywide (ORD) guidelines for reproductive toxicity risk assessment

Reproductive toxicity focuses on toxic effects regarding the male and female
reproductive systems, including outcomes of pregnancy and lactation.

3.1 This assumption is based on

When extrapolating data from (Not identified in the An agent that produces
an adverse

comparisons of data for agents experimental animal studies to

guidelines.) reproductive effect in experimental that are known to cause
human humans, in the absence of animals is assumed to pose a potential

reproductive toxicity [citations adequate human data threat to humans.

provided]. In general, the experimental animal data indicated adverse
reproductive effects that are also seen in humans.

3.2 Because similar mechanisms can

When extrapolating data from (Not identified in the Effects of xenobiotics
on male and

be identified in the male and experimental animal studies to guidelines.)

female reproductive process are female of many mammalian humans, in the
absence of

assumed generally to be similar unless species, effects of xenobiotics on
adequate human data. demonstrated otherwise.

male and female reproductive processes are assumed generally For
developmental outcomes, the

to be similar across species, specific effects in humans are not unless
demonstrated otherwise. necessarily the same as those seen in the
experimental species. However, The assumption for developmental adverse
developmental outcomes in outcomes is made because of the laboratory
mammalian studies are possibility of species- specific presumed to predict a
hazard for

differences in timing of exposure adverse developmental outcome in relative
to critical periods of humans. development, pharmacokinetics (including
metabolism), developmental patterns, placentation, or modes of action.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

3.3 It is assumed that the most

When extrapolating data from (Not identified in the In the absence of
information to

sensitive species is most experimental animal studies to guidelines.)

determine the most appropriate appropriate because, for a humans, in the
absence of

experimental species, data from the majority of agents known to cause

adequate human data, and in the most sensitive species should be used.

human reproductive toxicity, absence of sufficient information humans appear
to be as or more (e. g., pharmacokinetic data) to sensitive than the most
sensitive

determine the most appropriate animal species tested, based on experimental
species. data from studies that determined dose on a body weight or air

concentration basis [citations provided]. 3.4

This assumption for reproductive When extrapolating data from (Not
identified in the In the absence of information to the risk assessment is
based on three experimental animal studies to

guidelines.) contrary, an agent that affects considerations:

humans, in the absence of reproductive function in one sex is 1. For most
agents, the nature of adequate human data. assumed to adversely affect the
testing and data available reproductive function in the other sex.

are limited, reducing In the absence of information to confidence that the
potential the contrary.

for toxicity to both sexes and their offspring has been examined equally. 2.
Exposures of either males or

females have resulted in developmental toxicity. 3. Many of the mechanisms
controlling important aspects of reproductive system function are similar in
females

and males, and therefore could be susceptible to the same agents.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

3.5 This is based on known

When extrapolating data from (Not identified in the A nonlinear dose-
response curve is homeostatic, compensatory, or experimental animal studies
to

guidelines.) assumed for reproductive toxicity.

adaptive mechanisms that must humans, in the absence of be overcome before a
toxic adequate human data. endpoint is manifested and on the rationale that
cells and organs of In a quantitative dose- response the reproductive system
and the analysis, mode of action, developing organism are known to
pharmacokinetic, and

have some capacity for repair of pharmacodynamic information

damage. should be used to predict the

shape of the dose- response curve Although a threshold may exist for

when sufficient information of that endpoints of reproductive toxicity,
nature is available. When that it usually is not feasible to information is
insufficient, it has distinguish empirically between a

generally been assumed that there true threshold and a nonlinear

is a nonlinear dose response for low- dose relationship.

reproductive toxicity.

4. Agencywide (ORD) guidelines for developmental toxicity risk assessment
Developmental toxicity risk assessment focuses on risk to human development,
growth, survival, and function because of exposure to environmental agents
prior to conception, prenatally, or to infants and children. 4.1

This assumption is based on the When extrapolating data from (Not identified
in the

It is assumed that an agent that comparisons of data for agents

animal studies to humans for guidelines.)

produces an adverse developmental known to cause human hazard
identification/ doseresponse

effect in experimental animal studies developmental toxicity [citations
analysis.

will potentially pose a hazard to provided] which indicate that, in humans
following sufficient exposure almost all cases, experimental Generally
applied in the absence during development. animal data are predictive of a

of adequate human data. developmental effect in humans.

4.2 In the past, there has been a

When extrapolating data from (Not identified in the It is assumed that all
of the four tendency to consider only

animal studies to humans for guidelines.)

manifestations of developmental malformations or malformations hazard
identification/ doseresponse

toxicity (death, structural and death as endpoints of

analysis. abnormalities, growth alternations, and

concern. From data on agents functional deficits) are of concern. A that are
known to cause human Generally applied in the absence biologically
significant increase in any developmental toxicity [citations of adequate
human data.

of the four manifestations is considered provided], there is usually at
least indicative of an agent?s potential for one experimental species that
disrupting development and producing

mimics the types of effects seen in a developmental hazard. humans, but in
other species tested, the type of developmental perturbation may be
different.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

4.3 This assumption is made because

When extrapolating data from (Not identified in the It is assumed that the
types of

it is impossible to determine which animal studies to humans for

guidelines.) developmental effects seen in animal

will be the most appropriate hazard identification/ doseresponse studies are
not necessarily the same species in predicting the specific

analysis. as those that may be produced in types of effects seen in humans.
humans.

The fact that every species may Generally applied in the absence not react
the same way could be of adequate human data.

due to species- specific differences in critical periods, differences in
timing of exposure, metabolism, developmental patterns, placentation, or
mechanisms of action.

4.4 This assumption is based on

When extrapolating data from (Not identified in the The most appropriate
species is used

observations that humans are as animal studies to humans for

guidelines.) to estimate human risk when data are

sensitive or more so than the most hazard identification/ doseresponse
available (e. g., pharmacokinetics). In sensitive animal species tested for

analysis. the absence of such data, it is the majority of agents known to
assumed that the most sensitive

cause human developmental Generally applied in the absence species is
appropriate for use. toxicity [citations provided].

of adequate human data. 4.5

This is based on the known When evaluating the doseresponse (Not identified
in the

In general, a threshold is assumed for capacity of the developing
relationship. guidelines.)

the dose- response curve for agents organism to compensate for or to that
produce developmental toxicity.

repair a certain amount of damage at the cellular, tissue, or organ level.
In addition, because of the multipotency of cells at certain

stages of development, multiple insults at the molecular or cellular level
may be required to produce an effect on the whole organism.

4.6 (Not identified in the guidelines.) When doing dose- response (Not
identified in the

For developmental toxic effects, a evaluation. guidelines.)

primary assumption is that a single exposure at a critical time in
development may produce an adverse developmental effect, i. e., repeated
exposure is not a necessary prerequisite for developmental toxicity to be
manifested.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

4.7 (Not identified in the guidelines, When doing dose- response (Not
identified in the

If absorption [of the administered dose but this assumption is a variation

evaluation using data from animal guidelines.)

of an agent] in the experimental on the inference assumption studies.
species has been determined, but

identified under item 4. 4, and human absorption is not known, other related
EPA guidelines are human absorption is generally referenced as well.)
assumed to be the same as that for the species with the greatest degree of
absorption.

4.8 (Not identified in the section on When determining the reference

(Not identified in the When determining the reference dose RfD and RfC, but
the assumptions dose or concentration for

guidelines.) or reference concentration for regarding use of the most
developmental toxicity (the RfD DT developmental toxicity:

sensitive effect and most or RfC DT ) a level at or below which

appropriate and/ or sensitive it is assumed that no significant  The most
sensitive developmental species are variations on risk occurs.

effect from the most appropriate inference assumptions described and/ or
sensitive mammalian species above.) is used for determining the NOAEL,
LOAEL, or benchmark dose.

 Uncertainty factors for developmental and maternal toxicity generally
include a 10- fold factor for interspecies variation and a 10- fold factor
for intraspecies variation. Additional factors may be applied to account for
other uncertainties or additional information that may exist in the
database. In general, an uncertainty factor is not applied to account for
duration of exposure.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

5. Agencywide (ORD) guidelines for health risk assessment of chemical
mixtures (1999 draft of supplemental guidance)

5.1 Several studies have Used in dose- response (Not identified in the

For the component chemicals that demonstrated that dose (or

assessment. guidelines.)

show similar toxicity, dose addition is concentration) addition often
recommended. (When the effect of the

predicts reasonably well the Dose addition is the default combination is the
effect expected from toxicities of mixtures composed of

approach in situations where the the equivalent dose of an index a
substantial variety of both similar dose for each individual chemical, the
equivalent dose is the and dissimilar compounds component is at a level at
which sum of the component doses scaled by (citations provided).

effects are not expected to occur, their potency relative to the index be
observable, or be of concern. chemical. This sum of the exposure The
assessment of multiple However, when the doses are levels is a weighted
sum.) toxicant exposure has been

combined, effects of concern are addressed by the American then expected or
observed in Dose addition is different from Conference of Governmental
response to the higher dose level response addition because two Industrial
Hygienists, OSHA, the of the mixture.)

assumptions are made: that all of the World Health Organization, and
components have similar uptake,

the NRC (citations provided). pharmacokinetics, and toxicologic

Although the focus and purpose of processes, and that the (log probit) each
group was somewhat dose- response curves of the different, all of the
recommended components are parallel.

approaches included some type of dose- additive model.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

5.2 Dose- additive models may not be Used in dose- response (Not identified
in the

For the component chemicals in a the most biologically plausible assessment.

guidelines.) mixture that show dissimilar toxicity, approach if the
compounds do not response addition is recommended.

have the same mode of toxicologic Response addition is the default

(Under response addition, the general action. approach when the component

procedure is to first determine the risks chemicals are functionally per
exposure for the individual independent. It is most often components; the
mixtures risk is then applied when an effect that is of estimated by adding
the individual concern is expected to be present

risks together.) at low- dose levels for each of the

component chemicals, even Response addition is different from though it is
highly unlikely to be dose addition in that it does not observable at these
low levels in assume similar kinetics or a similar either epidemiologic or
toxicologic mode of action and does not assume studies. parallel dose-
response curves. It assumes that the components of the

Because response addition does mixture are considered to be not require a
similar mode of action functionally independent of one

across the chemicals in the another at low- exposure levels, so that
mixture, it allows for combining the risks may be added together. (This

risks across different types of sum of the effects of the individual

endpoints. chemicals is a conditional sum.)

5.3 (Not identified in the guidelines.) Used in dose- response (Not
identified in the

If interactions data are available, the assessment. guidelines.) default
recommendation is that they be incorporated into the risk assessment either
by using the interactions based hazard index or by including a qualitative
assessment of the direction and magnitude of the impact of the interaction
data.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

6. Agencywide (ORD) guidelines for ecological risk assessment

These guidelines focus on exposures to one or more anthropogenic chemical,
physical, or biological stressors that may result in adverse ecological
effects. In comparison to the NAS risk paradigm, the information here
reflects the novel approaches and methodological choices advocated by EPA in
ecological risk assessments. Only a subset of the primary methodological
choices are presented- rather than specific default assumptions, reflecting
the level of specificity in EPA?s ecological risk assessment guidance
document.

6.1 Allows determinations regarding

Applied in most ecological risk Precautionary EPA advocates tiering
ecological risk how extensive a risk assessment assessments. approach
maintaining assessments such that conservative should be, taking into a
higher degree of approaches are first employed in both

consideration risk management conservatism in the data and model use,
followed by a

goals. risk assessment.

more detailed assessment process (if warranted).

The underlying assumptions and the risk scenarios can be carried When a risk
has been identified,

through to risk characterization subsequent tiers use additional data to
phase in a first- tier risk address the uncertainties of the initial
assessment, allowing their

assessment( s). plausibility to be discussed and

reevaluated. Examples of methodological choices in first tier of a
ecological risk assessment:

 use simple rather than complex models;

 analyze uncertainty propagation and how uncertainty in individual
parameters can affect the overall

uncertainty in the results (e. g., calculate error bounds on a point
estimate); and

 conduct tests designed to evaluate effects such as lethality and
immobility.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

6.2 Better characterization of stressorassessment General data
considerations Data choice( s) can Methodological choices regarding data

endpoint during early first- and subsequenttier effect a risk issues:

relationships and better ability to risk assessment stages of assessment in
a accurately formulate and address problem formulation phase-

variety of ways, for  When existing data are few and new risk hypotheses:
particularly when data or example: data cannot be collected, consider
relationships cannot be defined in extrapolation from existing data when 
Extrapolation of data collected a traditional laboratory setting or

 More accurate characterizing effects of stressors on from other locations
or on

where assessments of multiple conclusions in assessment endpoints. organisms
where similar

stressors or site- specific factors laboratory settings

 Employ field- observational studies circumstances exist can be a
significantly influence exposure.

may be possible; that represent exposures and effects useful proxy-
particularly when however, laboratory better than estimates generated from
obtaining original data is Studies that minimize the amount controls may
limit the laboratory studies or theoretical unachievable.

of extrapolation are preferred. range of responses

models. and/ or may not  Use statistical tests (e. g., correlation,

 Field- observational studies reflect responses in clustering, factor
analysis) or indices

(surveys) are necessary to the environment. to measure and evaluate effects.

measure biological changes in  Directly measure environmental uncontrolled
situations that best

 Results can be media (or a combination of modeling mimic the relationship
between

presented as a and direct measurement) if stressors assessment endpoint and

series of point have already been released into the

stressor [citations provided]. estimates with environment, if possible.

different aspects of  Use point estimate/ descriptor

 Large- scale ecological uncertainty reflected approach when there is not
enough

processes are difficult to detect in each (e. g., information to describe a
distribution. in laboratory settings.

classical statistical methods such as confidence limits, or percentiles, can
be employed).

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

6.3 Allows for the establishment of a

Determination of dose in analysis (Not indicated in Generally, potential
dose of a chemical standardized approach in deriving phase. guidelines.) is
quantified as the amount of chemical values of exposure for ingested
ingested, inhaled, and/ or applied media (food or soil).

dermally. Specifically, potential average daily dose for ingested media is a
function of:

 average contaminant concentration in the type of food (using modeled or
measured values);

 fraction of intake of the food type that is from the contaminated area;

 normalized ingestion rate of the food type on a wet- weight basis;

 number of contaminated food types.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

6.4 EPA advocates such Production of a summary exposure

(Not indicated in In creating an exposure profile, methodological choices
regarding profile during analysis phase that:

guidelines.) estimates of exposure should be

exposure profiles in order to allow considered, at a minimum, along three a
risk assessor to best estimate

 identifies the receptor (i. e., the dimensions: intensity, time, and
space. risks. The exposure profile, in exposed ecological entity); These,
and other exposure guidelines,

turn, is combined with an effects are outlined below: profile to estimate
risks.

 describes the course a stressor takes from the source to the

 Intensity may be expressed as the receptor (i. e., the exposure amount of
chemical contacted per

pathway); and day or the number of pathogenic organisms per unit area.

 describes the intensity and spatial and temporal extent of cooccurrence

 The temporal dimension of exposure or contact. is best addressed with the
parameters of duration, frequency, and timing.

 In its simplest form, a stressor is quantified as a concentration, with
the assumption that the chemical is well mixed or that the organism moves
randomly through the medium.

? High end? exposure should refer to estimates that are expected to fall
between the 90 th and 99. 9 th percentile of the exposure distribution.

 Bounding estimates should refer to those higher than any actual exposure.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

6.5 Quotient method is commonly When performing activities of risk (Not
indicated in

When sufficient data are available to used for chemical stressors,
characterization phase (i. e., risk guidelines.) quantify exposure and
effects where reference or benchmark estimation, risk description,
estimates, the simplest approach for

toxicity values are widely reporting risk).

comparing the estimates is a ratio (or available.

quotient). Principal reasons for selecting A quotient addition approach
assumes quotient method: that toxicities are additive or

 It is simple and quick to use and approximately additive (this risk
assessors and managers assumption may be mostly applicable

are familiar with its application. when the modes of action of chemicals

 It provides an efficient, in a mixture are similar).

inexpensive means of identifying high- or low- risk situations that can
allow risk management decisions to be made without the need for further
information.

 Quotients can be used to integrate risks of multiple chemical stressors.

The following sections illustrate the types of additional, program- specific
assumptions and choices identified by individual program offices within EPA.
We only provide examples from two of EPA?s offices- the Office of Pesticide
Programs and the Office of Water. However, other offices can also have their
own preferences with regard to risk assessment assumptions and methods tied
to the particular exposures of concern to their programs.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

7. Office of Pesticide Programs science policy papers and guidance documents
regarding risk assessments for pesticides

OPP has published an expanding series of over 20 science policy papers- many
still at the draft stage- on various issues related to regulation of
pesticides in response to the Food Quality Protection Act of 1996. The
following rows illustrate some of the major assumptions or choices described
within those science policy papers or guidance documents.

7.1 Statutory provision of the FQPA, in (This is primarily a ?risk (The
intent of this Applying the FQPA tenfold safety factor

section 408( b)( 2)( C). management? issue when EPA is additional ?safety
?establishing, modifying, leaving in factor? is The FQPA requires EPA to
apply, in the (In draft policy papers on this effect, or revoking a
tolerance or precautionary- to give case of threshold effects, an additional

subject, EPA cited a number of exemption for a pesticide chemical special
consideration tenfold margin of safety for the reasons regarding the
agency?s

residue.? However, it affects risk with respect to pesticide chemical
residue and other preference for making a case- bycase assessment by OPP in
that it calls exposure and toxicity

sources of exposure for infants and determination of an for a determination
of whether the

to infants and children to take into account potential

appropriate safety factor. The available data are reliable and a children.)

pre- and postnatal toxicity and agency pointed out, in particular,

different ?margin of safety? will be completeness of data with respect to
that because OPP?s approach to safe for infants and children.) exposure and
toxicity to infants and

estimating exposure in the children. The Administrator of EPA absence of
extensive, specific may use a different margin of safety for data is
typically very conservative, the pesticide chemical residue only if, OPP can
usually conclude, with a on the basis of reliable data, such a high degree
of confidence, that its margin will be safe for infants and

approach adequately protects children. infants and children. EPA is also
revising its toxicology and Where reliable data are available, OPP exposure
data requirements in favors an approach that attempts to response to the
FQPA.) make a specific case- by- case determination as to the size of the
additional factor rather than rely on the tenfold default value.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

7.2 The tiered approach is used to When assessing dietary exposures (EPA
characterizes the Dietary exposure estimates (draft):

conserve limited resources. to pesticides. Analysis proceeds initial tiers,
especially through more rigorous tiers, as the

tier 1, as producing For regulatory purposes, acute and overall risk
assessment situation

conservative exposure chronic dietary exposure to pesticides requires, when
data exist. estimates. The higher in foods are estimated using indirect
level tiers are modeling approaches that consider characterized as pesticide
residues in food and the producing more

amount of food consumed. EPA accurate, and less

assesses dietary exposure using a conservative, tiered approach, proceeding
from exposure conservative to more refined assessments.)

assumptions and varying by acute (short- term) and chronic (long- term)
assessments. Initially, EPA uses deterministic assessments based on

various assumptions about the concentration of pesticide residue in the
food. For more refined dietary

assessments, EPA uses probabilistic exposure assessments. (See details for
these various tiers in rows 7.2.1 through 7. 2.4 below.] 7.2. 1

The tiered approach is used to When assessing dietary exposures Tolerance
levels for Tier 1 approach to estimating dietary

conserve limited resources. to pesticides. Analysis proceeds residues used
in Tier 1

risk from pesticides in foods (draft): through more rigorous tiers, as the

dietary exposure overall risk assessment situation estimates are not  Both
acute and chronic assessments requires, when data exist.

expected to accurately use deterministic values (point reflect actual
residues estimates) for exposure, assuming in ready- to- eat foods; residues
on foods to be at maximum rather, they are legal tolerance levels and that
100 intended to provide percent of the crop is treated with the inputs for
?worst- case? pesticide.

exposure estimates. Residue data for Tier 1 assessments meet the criterion
for conservative exposure factors.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

7.2. 2 The tiered approach is used to When assessing dietary exposures
Assessments are

Tier 2 approach to estimating dietary conserve limited resources. to
pesticides. Analysis proceeds

refined in Tier 2 using risk from pesticides in foods (draft): through more
rigorous tiers, as the

more realistic values overall risk assessment situation for pesticide
residues.

 Acute assessments use deterministic requires, when data exist. values for
exposure and assume that (In its guidance for residues on single- serving
size items

performing aggregate are at maximum legal tolerance

exposure levels, residues for blended assessments, OPP commodities are based
on average noted that field trial residue from field trials or monitoring
data, which are data, and 100 percent of the crop is traditionally the
treated with the pesticide. primary source of residue data in foods, 
Chronic assessments use

overestimate the deterministic values and assume residues that are likely
residues are at tolerance levels, but to occur in foods as use actual
percentage of the crop actually consumed. treated. According to the agency,
this is because they reflect

the maximum application rate and shortest pre- harvest interval, and
represent residue levels ?at the farm gate.? OPP

pointed out that data more reflective of residues on foods as consumed are
often available from monitoring data in which food samples are obtained
closer to the dinner table in the chain of commerce and analyzed.)

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

7.2. 3 The tiered approach is used to When assessing dietary exposures In
Tiers 3 and 4,

Tier 3 approach to estimating dietary conserve limited resources. to
pesticides. Analysis proceeds

pesticide residue data risk from pesticides in foods (draft):

through more rigorous tiers, as the are combined with overall risk
assessment situation

assumptions on actual  Acute assessments use probabilistic requires, when
data exist.

pesticide application values (distributional estimates), an rates,
stability, etc., to empirical distribution frequency from refine further
residue field trials for single- serving items, estimates in foods as
average residue data from field trials they are consumed.

or monitoring data for blended commodities, and actual percentage (See note
on field trial of the crop treated. These and monitoring data in assessments
also consider food row 7.2.2 above.)

processing factors.

 Chronic assessments use deterministic values, average residue data from
field trials or monitoring data for both blended and singleserving

commodities, and actual percentage of the crop treated. Again, food
processing factors are

considered. 7.2. 4

The tiered approach is used to When assessing dietary exposures In Tiers 3
and 4, Tier 4 approach to estimating dietary

conserve limited resources. to pesticides. Analysis proceeds pesticide
residue data

risk from pesticides in foods (draft): through more rigorous tiers, as the

are combined with overall risk assessment situation assumptions on actual 
Acute assessments use probabilistic requires, when data exist.

pesticide application values, market basket survey data on rates, stability,
etc., to consumption, actual percentage of

refine further residue the crop treated, and consider

estimates in foods as cooking, residue decline, residue they are consumed.
degradation, and other factors that may affect residues in foods as they are
consumed.

 Chronic assessments use the same types of data as acute assessments, but
with deterministic values.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

7. 3 Assessing Residential Exposures This conservatism adds an extra

When assessing risks associated EPA said that when measure of safety when
regulating with residential exposures to

scientists have studied EPA notes that although its residential pesticides.

(residential uses of) pesticides. people in the real exposure assessments
?are designed world (including the

to be as realistic as possible,? the Explanations for the specific children
of assessments also are ?generally

assumptions include: farmworkers) they conservative.? Specific assumptions
have generally found a

include: a. The highest reasonably person?s exposure to

a. Assume high amounts of pesticide possible transfer rate must be be less
than that residues will transfer to a person. assumed for safety.

predicted by EPA?s (EPA generally assumes 20- 50 exposure percent of the
residues will b. Dissipation rate is based on assessments.

transfer.) many factors (heat, sunlight,

and rain, etc.) so EPA says it (In other guidance b. Assume no residue
dissipation. In must include the conservative documents, EPA other words,
all the residues prospect that in a given case points out that its available
initially are available there is no residue

residential SOPs are throughout the time a person is dissipation. by nature
designed to exposed. produce screeninglevel c. Assumed because little or no
assessments that c. Assume that a person has no clothing is a possible
realistic

are intentionally clothing on to protect from scenario in some conservative
in exposure. circumstances.

nature.) d. Assume 2 to 8 hours of continuous

d. (No additional explanation contact. provided.)

(Note that EPA was reviewing and considering whether to update some of the
detailed exposure scenario assumptions in its standard operating procedures
(SOPs) for residential assessments.)

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

7.4 Guidance for Performing (EPA revised its approach to When performing
aggregate (Not directly Aggregate Exposure and Risk aggregate exposure and
risk exposure and risk assessments addressed in the

Assessment assessment in response to the

general discussion in FQPA?s requirement that the the revised guidance.
Aggregate exposure and risk agency consider aggregate The inference from
assessment involves the analysis of a exposure in its decision making.)

EPA?s comparison to single chemical exposure by multiple the interim
guidance is pathways of exposure (i. e., food, EPA noted in the guidance
that that the revised drinking water, and residential, aggregate exposure
assessments

practices should result nonoccupational), and all relevant built individual
by individual,

in less conservative routes of exposure (i. e., oral, dermal, culminating in
a total exposure to estimates than the and inhalation). EPA?s revised the
population, may allow for interim practices. guidelines support an approach
in probabilistic treatment of data

However, the revised which an analyst assesses exposure

incorporating all pathways of guidance also on an individual- by- individual
basis, exposure, (i. e., food, drinking emphasizes the use of culminating in
a representative water, and residential).

?reasonable population of interest. This approach assumptions that do
differs from EPA?s interim guidelines. Distributional data analysis is not
underestimate (According to the interim guidelines, preferred as this tool
allows an exposure? when data aggregate assessments most aggregate exposure
assessor to are not available, so frequently added the ?high- end? or more
fully understand the

there still may be a upper- bound point estimates from the uncertainty and
variability inherent

precautionary drinking water and residential

in the data set. element.)

exposure pathways to a point on the distribution of food ingestion exposure,
e. g., the 99. 9 th percentile.) Aggregate exposure and risk assessments
will be more realistic to the extent that the appropriate

temporal, spatial, and demographic factors that affect exposure to an
individual are understood and accounted for. When these data are not
available, reasonable assumptions that do not underestimate exposure can be
used. Once an aggregate exposure and risk assessment is completed for one
individual and repeated for many individuals, population and subpopulation
distributions of total exposures and risk may be constructed by
probabilistic techniques. Distributional data analysis is preferred.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

8. Office of Water Risk Assessments

OW has published a collection of guidance documents relating to water
quality issues, based primarily on mandates derived from the Clean Water Act
(CWA) and Safe Drinking Water Act (SDWA). For illustrative purposes,
emphasis here is on assumptions and methodological choices taken from
Methodology for Deriving Ambient Water Quality Criteria for the Protection
of Human Health, EPA822-

B- 00- 005 (October 2000). This document provides detailed guidance on
developing Ambient Water Quality Criteria (AWQC), with organized procedures
for evaluating cancer risk, noncancer health effects, human exposure, and
bioaccumulation potential in fish. In this guidance, EPA states that
protective assumptions are made regarding potential human exposure intakes
at an appropriate level of conservatism where uncertainties exist.
Furthermore, criteria are derived to be protective, not predictive, of an
exact percentile of the

total population that is protected. 8.1

Body weight, water intake, and Used in the derivation of AWQC,

Conservative EPA has identified a number of fish intake are default
parameters and also, to be consistent with

assumptions in standard or default exposure factors: specific to target
populations that

Section 101( a) of CWA which choosing exposure  All drinking water consumed
is are considered important when specifies possible water body uses
parameters support contaminated at the criteria level.

determining AWQC values. (e. g., consumption of fish).

the goal of protecting  All fish consumed is contaminated at the majority
of the the criteria level and all fish may Drinking water and fish intake
population. come from one water body.

values are 90 th percentile  Body weight of a child ages 1 to 3 is
estimates, based on most recent 13 kg (when the age group is ages 1

USDA survey data reflecting the to 14, body weight is 30 kg).

90 th percentile of the general

 Body weight of an adult is 70 kg. population (1994- 96 CSFII data),  Body
weight of a woman of and are percentiles selected to

childbearing age is 67 kg. ensure protection of the majority

 Daily untreated surface water of consumers of drinking water consumed is 2
liters.

and fish.

 Daily fish consumption is 17. 5 g for general adult population and
(average) sport fishers; 142. 4 g for

subsistence fishers.

 Criteria generally represent ambient pollutant concentrations that are
acceptable based on a lifetime (70 years) of exposure.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

8.2 Uncertainty factors listed in the Deriving water quality criteria for
Conservative In setting RfDs for noncarcinogens,

previous column are used to more the protection of human health

approach resulting in EPA advocates the use of uncertainty

accurately determine an RfD by from noncancer effects (i. e., lower RfD
values. factors and, if needed, a modifying more closely identifying and

noncarcinogenic chemicals) via factor to account for areas of scientific
incorporating, respectively:

the determination RfDs. uncertainty in toxicity databases. EPA recommends a
default of 10 for

a. interhuman variability (to Specifically, the RfD is used with

uncertainty factors and a default of 1 account for variation in additional
information regarding

for modifying factors when lacking sensitivity among the exposure and the
bioaccumulation

other information. members of the human potential of the substance to derive
population);

an AWQC for noncancer effects. Specifically, EPA recommends b. experimental
animal- tohuman applying a 1-, 3-, or 10- fold uncertainty

extrapolation According to EPA, the ?[ c] hoice of factor when:
(interspecies variation); appropriate uncertainty factors and a.
extrapolating from valid data in c. subchronic to chronic modifying factors
must be a caseby- studies using long- term exposure extrapolation (to
account for

case judgment by experts and to average healthy humans; uncertainty in
extrapolating

should account for each of the b. extrapolating from valid results of from
less- than- chronic applicable areas for uncertainty long- term studies on
experimental NOAELs or LOAELs to and nuances in the available data animals
when results of studies of chronic NOAELs);

that impact uncertainty.? human exposure are not available

d. database completeness (to or inadequate;

account for the inability of any c. extrapolating from less- thanchronic
single study to adequately

results on experimental address all possible adverse

animals when there are no useful outcomes); and long- term human data; or

e. LOAEL- to- NOAEL d. deriving an RfD from an extrapolation.

?incomplete? database. A modifying factor is to be used EPA recommends a 3-
or 10- fold when the areas of scientific uncertainty factor when:

uncertainty addressed with e. deriving an RfD from a LOAEL, uncertainty
factors do not instead of a NOAEL.

represent all of the uncertainties in the estimation of a RfD. EPA advocates
using professional judgment in determining the modifying factor, the
magnitude of which is greater than zero and less than or equal to 10, being
dependent on assessment of uncertainties of relevant studies and databases.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

8.3 Such values reflect appropriate (None given.) (None given.)

EPA identified both 10 -6 and 10 -5 as target risk levels for health
appropriate cancer target risk levels, protection of the general and said
that highly exposed

population. populations should not exceed a 10 -4 cancer risk level.

The cancer risk levels increase the degree of consistency between Cancer
risk level of 10 -6 is based on a the drinking water program, fish intake
rate of 17. 5 g/ day for

ambient water program, and other general population and sport anglers, a EPA
programs. fish intake rate of 142.4 g/ day for subsistence fishers, and a
drinking water intake rate of 2 liters per day for each of these groups.

8.4 This methodological choice offers

Where either new risk assessment AWQC may increase Under some circumstances,
it may be some flexibility for site- specific or studies or site- specific
information

or decrease. appropriate to use a point within an

contaminant- specific situations but support derivation of an AWQC RfD range
as the basis for deriving remains protective of public from a point within
RfD range.

AWQC, rather than the single point health. default estimate of the RfD, when
the uncertainty factor is 100 or greater and the range is either a quarter
or half log unit to either side of the calculated RfD.

8.5 In deriving RfD, use of RSC allows Derivation or revision of an AWQC
Conservative and

Consideration of nonwater sources of for assessing total human based on RfDs
or nonlinear

protective approach exposure (e. g., ingestion and/ or exposure to a
contaminant and

carcinogen PODs. when known or inhalation exposure) is recommended
apportions the RfD among the

anticipated nonwater when determining a relative source media of concern.
sources of exposure contribution (RSC) factor to apply to a are anticipated.
nonlinear carcinogen point- ofdeparture

 Use of ?Exposure Decision Tree? (POD) or a RfD. approach allows for use of
either subtraction or percentage Apply an estimate of the RSC factor
methods, depending on (between 20 and 80 percent) to the

chemical circumstances, within RfD when adequate exposure data do the 20- to
80- percent range. not exist, using an exposure decision tree approach.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the Assumption or
methodological

risk assessment process or Likely effect on risk choice Reason( s) for
selection circumstances) assessment results

8.6 This change was made in order to BAF values are used in

The ambient water EPA recommends using a

reflect the uptake of a contaminant calculations of AWQC for

quality criteria for bioaccumulation factor (BAF) rather by aquatic
organisms from all

carcinogens (in both linear and highly bioaccumulative than a
bioconcentration factor as sources (e. g., water, food, nonlinear
approaches) and for

pollutants may be up previously set forth in the 1980 sediment) rather than
just from the noncarcinogens. to 2 orders of Methodology.

water column. magnitude lower than

criteria derived with bioconcentraton factors.

Source: Compiled from GAO review of EPA risk assessment guidelines and
related documents and from additional comments provided by agency officials.

Risk Characterization As with exposure assessment, the program offices
typically are responsible for completing the risk characterization. EPA
does, however, have several documents that provide agencywide guidance on
how such characterization is to be done. The guidance includes a February
26, 1992, memorandum from the EPA Deputy Administrator entitled, ?Guidance
on Risk Characterization for Risk Managers and Risk Assessors,? and a March
21, 1995, document issued by the EPA Administrator entitled, ?Policy for
Risk Characterization at the U. S. Environmental Protection Agency.? EPA
also has developed a Risk Characterization Handbook to provide more detailed
guidance to agency staff. In the statement accompanying its 1994 report
Science and Judgment in

Risk Assessment, NRC said that although EPA?s overall approach for assessing
risks was fundamentally sound, the agency ?must more clearly establish the
scientific and policy basis for risk estimates and better describe the
uncertainties in its estimates of risk.? In March 1995, the EPA
Administrator issued the agency?s risk characterization policy and guidance,
which reaffirmed the principles and guidance in the agency?s 1992 policy.
EPA?s guidance document defined risk characterization as the final step in
the risk assessment process that (1) integrates the individual
characterizations from the hazard identification, dose- response, and

exposure assessments; (2) provides an evaluation of the overall quality of
the assessment and the degree of confidence the authors have in the
estimates of risk and conclusions drawn; (3) describes the risks to
individuals and populations in terms of extent and severity of probable
harm; and (4) communicates the results of the risk assessment to the risk

manager. Discussing ?guiding principles? for risk characterization, EPA
emphasized that the integration of information from the three earlier stages
of risk assessment, discussion of uncertainty and variability, and
presentation of information to risk managers requires the use of both
qualitative and quantitative information. For example, when assumptions are
made in exposure assessment, EPA said that the source and general

logic used to develop the assumptions should be described, as well as the
confidence in the assumptions made and the relative likelihood of different
exposure scenarios. In the 1995 policy statement, EPA said that risks should
be characterized in a manner that is clear, transparent, reasonable, and
consistent with other risk characterizations of similar scope. EPA said that
all assessments ?should identify and discuss all the major issues associated
with determining the nature and extent of the risk and provide commentary on

any constraints limiting fuller exposition.? The policy also said risk
characterization should (1) bridge the gap between risk assessment and risk
management decisions; (2) discuss confidence and uncertainties involving
scientific concepts, data, and methods; and (3) present several types of
risk information (i. e., a range of exposures and multiple risk

descriptors such as high ends and central tendencies). The policy stated
that each risk assessment used in support of decision making at EPA should
include a risk characterization that follows the principles and reflects the
values outlined in the policy. However, the policy statement went on to say
that it and the associated guidance did not establish or affect legal rights
or obligations.

Some of EPA?s other risk assessment guidelines also discuss and recommend
certain approaches to the risk characterization phase. For example, EPA?s
proposed guidelines for carcinogen risk assessment call for greater emphasis
on the preparation of ?technical? characterizations to summarize the
findings of the hazard identification, dose- response assessment, and
exposure assessment steps. The agency?s risk assessors

are then to use these technical characterizations to develop an integrative
analysis of the whole risk case. That integrative analysis is in turn used
to prepare a less extensive and nontechnical Risk Characterization Summary
intended to inform the risk manager and other interested readers. EPA
identified several reasons for individually characterizing the results of
each analysis phase before preparing the final integrative summary. One is
that the analytical assessments are often done by different people than
those who do the integrative analysis. The second is that there is very
often a lapse of time between the conduct of hazard and dose- response
analyses

and the conduct of the exposure assessment and integrative analysis. Thus,
according to EPA, it is necessary to capture characterizations of
assessments as the assessments are done to avoid the need to go back and
reconstruct them. Finally, several programs frequently use a single hazard
assessment for different exposure scenarios. The guidelines also point out
that the objective of risk characterization is to call out any significant

issues that arose within the particular assessment being characterized and
inform the reader about significant uncertainties that affect conclusions,
rather than to recount generic issues that are covered in agency guidance
documents.

In another example, EPA?s ecological risk guidelines emphasize that risk
characterization is a means for clarifying relationships between stressors,
adverse effects, and ecological entities. In addition, this phase of the
risk assessment process is a time to reach conclusions regarding the
occurrence of exposure( s) and the adversity of existing or anticipated
effects. Specifically, EPA guidance describes three ecological risk
characterization activities: (1) risk estimation (i. e., integrating
exposure and effects data and evaluating uncertainties); (2) risk
description (i. e.,

interpreting and discussing available information about risks to the
assessment endpoints); and (3) risk reporting (i. e., estimating risks
indicating the overall degree of confidence in such estimates, citing lines
of

evidence to support risk estimates, and addressing assumptions and
uncertainties). Similar to EPA- wide guidance on risk characterization,
EPA?s ecological risk characterization guidelines emphasize open

communication with risk managers and other interested parties to clearly
convey information needed for decision making in a risk management context.

It is also EPA?s policy that major scientifically and technically based work
products related to the agency?s decisions normally should be peer reviewed
to enhance the quality and credibility of the agency?s decisions.

With regard to EPA?s chemical risk assessments, peer review can be used for
evaluating both specific assessments and the general methods EPA uses in its
risk assessments. Peer review generally takes one of two forms: (1) internal
peer review by a team of relevant experts from within EPA who have no other
involvement with respect to the work product that is to be evaluated or (2)
external peer review by a review team that consists primarily of independent
experts from outside EPA. In December 2000,

EPA released a revised edition of its Peer Review Handbook for use within
the agency.

Chemical Risk Assessment at the Food and

Appendi x II I Drug Administration The Food and Drug Administration (FDA)
within the Department of Health and Human Services regulates the safety of a
large number and wide variety of consumer products, including foods,
cosmetics, human and animal medicines, medical devices, biologics (such as
vaccines and blood products), and radiation- emitting products (such as
microwave ovens). Chemical risk assessments are primarily conducted by three
of FDA?s five product- oriented centers- the Center for Food Safety and
Applied Nutrition (CFSAN), the Center for Veterinary Medicine (CVM), and the
Center for Devices and Radiological Health (CDRH). The chemical risk
assessment activities of these centers vary depending on factors such as

the underlying statutory requirements, the substances being regulated,
whether cancer or noncancer effects are of concern, and whether a product is
under pre- or postmarket scrutiny. FDA officials said that the agency
generally follows the National Academy of Sciences? (NAS) fourstep risk
assessment process, although it has not developed written internal
guidelines. FDA often incorporates conservative assumptions into its
assessments when information essential to a risk assessment is not known,
but such assumptions are supposed to be scientifically plausible and
consistent with agency regulations or policies. For example, CFSAN

assumptions are expected to be reasonably protective of human health. FDA
does not have an official policy on how risk assessment results should be
characterized and communicated to policymakers and the public. However, FDA
officials said that, in practice, they use a standard approach

that typically highlights the assumptions with the greatest impact on the
results of an analysis, states whether the assumptions used were
conservative, and shows the implications of different choices.

Context for FDA FDA?s regulatory authority is primarily derived from the
Federal Food, Chemical Risk

Drug, and Cosmetic Act, as amended (FFDCA), although several related public
health laws (e. g., the Food and Drug Administration Modernization
Assessment

Act of 1997, or FDAMA) provide additional authority. FDA administers its
regulatory responsibilities through its five product- oriented centers: (1)
CFSAN, (2) CVM, (3) CDRH, (4) the Center for Drug Evaluation and Research,
and (5) the Center for Biologics Evaluation and Research. FDA officials said
that, although each of these five product centers conducts some type of risk
assessments, the first three primarily conduct the

chemical risk assessments that are the focus of this report. 1 Each of these
centers has different responsibilities, authorities, and constraints on its
regulatory and risk assessment activities. 1 Some of the offices focused on
in this appendix also conduct other, nonchemical, types of risk assessments.
For example, the Center for Devices and Radiological Health also uses risk
assessments to determine the carcinogenic, genetic, and/ or reproductive
health risks associated with radiation- emitting products. The Center?s
reviews of medical devices are also likely to consider engineering risk
assessments and the potential adverse effects of exposure to microbial
contamination. The Center for Veterinary Medicine and the Center for Food
Safety and Applied Nutrition also carry out microbiological risk
assessments.

 CFSAN is responsible for the regulation of food additives, color additives
used in food, and cosmetic additives. 2 Under the FFDCA, the regulation of
substances intentionally added to food or used in contact with food must be
based solely on the safety of the substances for their

intended uses (i. e., consideration of benefits and costs is not allowed). A
food containing an unapproved food or color additive is considered

?unsafe? unless FDA issues a regulation approving its use or, in the case of
a food contact substance, there exists an effective notification. 3 To
obtain an authorizing regulation or an effective notification, the sponsor
of a food or color additive must show that it is safe for its intended use.

FDA regulations under the FFDCA define a product as safe if there is ?a

reasonable certainty in the minds of competent scientists that the substance
is not harmful under the intended conditions of use.? 4 For food additives
and color additives that are not themselves carcinogenic but contain
carcinogenic impurities, CFSAN uses a quantitative risk

assessment to determine whether the risk posed by a carcinogenic impurity is
acceptable (i. e., a lifetime risk below one per million) under the FFDCA?s
general safety clause of ?reasonable certainty of no harm.? Nevertheless, if
the food or color additive itself is a known carcinogen, under the ?Delaney
Clause? amendments to FFDCA, it cannot be deemed safe and is prohibited from
use in food. 5 CFSAN is also involved with substantial activities in the
area of postmarket concerns with contaminants and naturally occurring
toxicants. For example, in

the past year, CFSAN participated in a number of major, international
chemical risk assessments in the areas of dioxins and various mycotoxins.

2 FDA?s responsibility in the food area covers all food except meat,
poultry, and egg products, which are under the authority of the U. S.
Department of Agriculture. Generally, FDA regulates food products sold in
interstate commerce, whereas products made and sold entirely within a state
are regulated by that state.

3 Section 309 of FDAMA of 1997. 4 21 CFR 170.3( i). 5 21 U. S. C. 348 (c)(
3)( A) on food additives and 379e( b)( 5)( B) on color additives. The Food
Quality Protection Act of 1996 eliminated application of the Delaney proviso
to pesticides.

 CVM?s primary role is to implement the FFDCA requirement that animal drugs
and medicated feeds are safe and effective for their intended uses and that
food from treated animals is safe for human consumption. Under the FFDCA,
the regulation of residues of animal drugs that become a part of food
because of the use of the animal drug must be based solely on health factors
(i. e., consideration of benefits and costs is not allowed). A carcass or
any of its parts that contain residues of an unapproved drug, or residues of
an approved drug above approved levels, is considered to be unsafe and the
carcass is considered adulterated. CVM uses risk assessment to help develop
safe

concentration levels in edible tissues, residue tolerances for postmarket
monitoring, and withdrawal periods for slaughter following drug treatment.
For noncancer effects, the applicable safety standard under

FFDCA is that that these concentrations, tolerances, and withdrawal periods
should represent a ?reasonable certainty of no harm.? FFDCA includes
provisions that permit FDA to authorize extralabel uses of an animal drug
that would pose a ?reasonable probability? of risk to human health if
residues of the drug are consumed. The agency may establish a safe level for
the residue and require that the drug sponsor provide an analytical method
for detecting residues of such a compound. 6 However, the act prohibits use
in food- producing animals of any

compound found to induce cancer when ingested by people or animals unless it
can be determined that ?no residue? of that compound will be found in the
food produced from those animals under conditions of use

reasonably certain to be followed in practice. 7 FDA has interpreted the
intention of the ?no residue? language in the statute as meaning that any
remaining residues should present an insignificant risk of cancer to people.
As a matter of policy, FDA accepts a lifetime risk below one per million as
an insignificant level.

6 21 U. S. C. 360b( a)( 4)( B). 7 21 U. S. C. 360b( d)( 1)( I) on new animal
drugs. This is known as the DES (diethylstilbestrol) proviso to the Delaney
Clause. The Center for Veterinary Medicine?s regulation found in 21 CFR
500.80 outlines how the center regulates carcinogenic drugs and feed
additives under the DES proviso.

 CDRH administers the medical device provisions of FFDCA, and assesses
risks posed by chemicals that might leach out from medical devices (e. g.,
breast implants) into surrounding tissue. The center?s

basic mission is to protect the public health by ensuring that there is
reasonable evidence of the safety and effectiveness of medical devices
intended for human use. CDRH usually evaluates risks in the context of a
premarket review system, and the decision to clear or approve a product to
treat a specific condition is based on a benefit- risk analysis for the
intended population and use (not just on the basis of safety or

human health as in the case of food regulation). Because all medical
products are associated with risks, CDRH considers a medical product to be
safe if it has reasonable risks given the magnitude of the benefit

expected and the alternatives available. 8 Another unit of FDA, the National
Center for Toxicological Research (NCTR), has an important supporting role
in the risk- related activities of the product centers. NCTR conducts much
of the agency?s methodological research on risk assessment methods and helps
to develop and modify FDA?s quantitative methods, in conjunction with
experts from the various product centers. NCTR also provides toxicology
research supporting all components of FDA. It performs fundamental and
applied research designed specifically to define biologic mechanisms of
action underlying

the toxicity of products regulated by FDA. 8 This applies to all medical
products, including drugs and biological products, not just to devices.

Risk Assessment Although FDA has long been a pioneer in the development of
risk assessment methods, the agency has not developed written internal
Procedures

guidance specifically on conducting risk assessments. 9 FDA officials noted
that much of their work is done before products are placed on the market
and, in those instances, the burden of proof is on sponsors seeking FDA
approval for new products. 10 In keeping with this requirement, FDA produces
extensive external guidance documents that are primarily directed at those
sponsors. 11 The documents are meant to represent the agency?s current
thinking on the scientific data and studies considered appropriate for
assessing the safety of a product. However, the guidance documents are not
legal requirements and do not preclude the use of

alternative procedures or practices by either FDA or external parties. Some
of these guidelines include detailed descriptions of risk assessment methods
deemed appropriate to satisfy FDA?s reviews under various statutory
provisions. FDA has also adopted a number of domestic and international
consensus standards that prescribe certain risk assessment methods (e. g.,
approaches for assessing the safety of medical devices and default
consumption values for meat products).

9 FDA published ?Procedures for the Appraisal of Toxicity of Chemicals in
Food,? the agency?s first guidance to industry, in 1949. In the 1950s, FDA
initially developed the general NOAEL (or NOEL)/ safety factor approach to
noncancer risk assessment that has been used by most regulatory bodies (e.
g., EPA?s reference dose approach). 10 This is not true with regard to
dietary supplements. The Dietary Supplement Health and Education Act of 1994
created a new framework for FDA?s regulation of dietary supplements, which
do not have to undergo preapproval by FDA to determine their safety or
efficacy. FDA officials said they currently have no standard procedures for
dietary supplement risk assessment. 11 Examples of such guidelines include
?Redbook 2000: Toxicological Principles for the Safety of Food Ingredients,?
?Estimating Exposure to Direct Food Additives and Chemical Contaminants in
the Diet,? and ?General Principles for Evaluating the Safety of Compounds
Used in Food- Producing Animals.?

FDA risk assessment procedures have also been described by individuals and
organizations from within and outside of the agency in scientific and
professional journal articles. For example, a 1997 journal article written
by a panel of officials from across FDA summarized the risk assessment
approaches of each of FDA?s product centers. 12 A 1996 report on federal
agencies? chemical risk assessment methods described CFSAN?s methods, but
did not describe the approaches used by the other centers within FDA. 13
FDA?s food safety risk assessment procedures were also described in
?Precaution in U. S. Food Safety Decisionmaking: Annex II to the United
States? National Food Safety System Paper,? which was prepared for the
Organization for Economic Cooperation and Development in March 2000. 14 FDA
officials said that the agency generally follows the four- step risk

assessment process identified by NAS: hazard identification, dose- response
assessment (which FDA prefers to call ?hazard characterization?), exposure
assessment, and risk characterization. They said that they also rely on past
precedent and other seminal works on risk assessment, such

as the 1985 Office of Science and Technology Policy guidance document on
cancer risk assessment. However, they emphasized that FDA does not presume
there is a ?best way? of doing a risk assessment and is continually updating
its procedures and techniques with the goal of using the ?best available
science.? Differences in Risk

FDA officials also said that there are variations in the risk assessment
Assessment Among FDA approaches used among the agency?s different product
centers and, in Product Centers

some cases, within those centers. In general, those variations are traceable
to differences in the following factors:

 the substances being regulated, 12 D. W. Gaylor, J. A. Axelrad, R. P.
Brown, J. A. Cavagnaro, W. H. Cyr, K. L. Hulebak, R. J. Lorentzen, M. A.
Miller, L. T. Mulligan, and B. A. Schwetz, ?Health Risk Assessment Practices
in the U. S. Food and Drug Administration,? Regulatory Toxicology and
Pharmacology 26 (1997).

13 Lorenz R. Rhomberg, A Survey of Methods for Chemical Health Risk
Assessment Among Federal Regulatory Agencies, a report prepared for the
National Commission on Risk Assessment and Risk Management (1996).

14 Hereinafter referred to as Precaution in U. S. Food Safety
Decisionmaking. This paper also addresses risk assessment practices of other
federal agencies involved in food safety, such as the Department of
Agriculture and EPA.

 the nature of the health risks involved (particularly carcinogens versus
noncarcinogens),

 statutory and regulatory requirements,

 whether the risk assessment is part of the process to review and approve a
product before it can be marketed and used (premarket) or whether the
assessment is for risks that might arise during monitoring

of a product once it is being used (postmarket), and

 the nature and extent of the scientific information available. The nature
and extent of scientific information varies on a case- by- case basis. The
other factors, however, are more generic, and table 6 illustrates how they
are similar or different across CFSAN, CVM, and CDRH. The subsections
following the table describe more specifically how CFSAN, CVM, and CDRH
conduct the first three stages of risk assessment.

Table 6: Differences in FDA Chemical Risk Assessment Factors Nature of
chemical Stage of product health risk Statutory/ regulatory requirements
review

(Endpoint of Substances being Risk assessment

(Premarket/ Product center concern) regulated Safety standard mandated?
Postmarket review)

CFSAN Office of Noncancer Food contact

Reasonable Mandatory Primarily premarket

Food Additive Safety toxicological substances, direct certainty of no harm
endpoints and indirect food additives, color

additives, and impurities

CFSAN Office of Various toxicological

Industrial and May render injurious

Mandatory for new Postmarket for foods Food Additive Safety endpoints,
including naturally occurring to health

dietary ingredients and Office of cancer

chemical Premarket for new Nutritional Products, contaminants,

In addition, for Discretionary for dietary ingredients in Labeling and
Dietary dietary supplements

dietary supplements, other food dietary supplements

Supplements significant or

unreasonable risk of illness or injury

Nature of chemical Stage of product

health risk Statutory/ regulatory requirements review (Endpoint of
Substances being

Risk assessment (Premarket/

Product center concern) regulated Safety standard mandated? Postmarket
review)

CFSAN Cancer Cancer Direct and indirect For additives, Assessment for
Primarily premarket, Assessment

food additives, food Delaney clause carcinogenicity is but can (and has
Committee and

contact substances, applies (carcinogens

mandatory as part of been) used for Quantitative Risk color additives, and
prohibited)

the premarket postmarket issues Assessment

impurities approval process, Committee

For others, but a risk reasonable certainty assessment is of no harm
standard

discretionary applies

CVM Various noncancer Residues of new Reasonable

Mandatory Primarily premarket, toxicological animal drugs and certainty of
no harm but can (and has endpoints

feed additives been) used for

postmarket issues CVM Cancer Residues of new Reasonable

Mandatory Primarily premarket, animal drugs, feed certainty of no harm but
can (and has additives, or their

with inclusion of been) used for impurities

Delaney clause and postmarket issues DES proviso

CDRH Cancer and Chemical Reasonable Mandatory for

Primarily premarket, noncancer endpoints compounds released assurance that
the certain types of but also could occur (leaching) from device is safe
under devices (Class III), per postmarket medical devices the conditions of
use

discretionary for surveillance

prescribed, others

recommended, or suggested in the labeling thereof In addition, regarding
Class III premarket approval, unreasonable risk of illness or injury

Source: Compiled from information provided by FDA.

CFSAN Risk Assessment CFSAN?s procedures for hazard identification and dose-
response Procedures assessment vary depending on whether noncancer or cancer
risks are at issue. For noncancer effects, CFSAN starts with the largest
dose in a

chronic animal study that did not appear to lead to an increase in toxic
effects above the level measured in unexposed control animals- the ?no

observed adverse effect level? or NOAEL. 15 CFSAN then divides this NOAEL by
one or more safety factors to arrive at an ?acceptable daily intake? (ADI)
intended to be an amount that can be ingested daily for a lifetime without
harm. For example, CFSAN typically divides the NOAEL by 10 to allow for the
possibility that humans might be more sensitive to a chemical than the
experimental animals and then by another 10 to account

for the possibility that some individuals might have greater sensitivity
than others might. Therefore, for ADIs derived from long- term animal
studies, CFSAN commonly uses a combined safety factor of 100. Additional
safety factors may also be applied to account for long- term effects versus
shortterm experiments, inadequacies of the experimental data, or other
factors. 16 For cancer effects, CFSAN uses two different hazard assessment/
doseresponse

approaches, depending on the nature of the products being regulated.

15 FDA often determines a no observed effects level (NOEL) rather than a
NOAEL because many significantly altered, standard toxicological endpoints
are assumed to be adverse to animals and/ or humans even in the absence of
data affirming that assumption.

16 For example, Lorenz Rhomberg noted that, for developmental toxicity, the
Center for Food Safety and Applied Nutrition may use a combined safety
factor of 1, 000 for severe, irreversible health effects.

 For food and color additives that are themselves known carcinogens, the
Delaney provisions in FFDCA make risk assessment rather straightforward. If
a petition to market a food ingredient contains an adequately conducted
animal cancer study, and if results of that study indicate that the food
ingredient produces cancer in animals, CFSAN

identifies the substance as a carcinogen under the conditions of the study.
No further corroboration or weight- of- evidence analysis is required, and
there is no need for a detailed dose- response assessment, exposure
assessment, or risk characterization for the purpose of determining a
specific level of the carcinogenic substance in food that may be considered
to be safe. 17

 CFSAN uses more elaborate procedures for known or suspected carcinogenic
impurities in food additives. The center?s method for lowdose cancer risk
estimation is similar to EPA?s method (presented in app. II) on
extrapolation for carcinogens (see fig. 3). On the doseresponse

curve of tumor incidence versus dose for a chemical, CFSAN chooses a point
below which the data are no longer considered reliable, usually in the range
of a tumor incidence of 1 percent to 10 percent. A straight line is drawn
from the upper- confidence limit on the estimated risk at that point to the
origin (i. e., zero incremental dose/ zero incremental response). This
provides the slope of the line used to provide upper- bound estimates of
cancer risk at low doses. CFSAN

does not specify a particular mathematical form for the dose- response
relationship in the experimental dose range; the only requirement is an
adequate fit to the data. According to FDA officials, CFSAN risk assessors
use one of two different methods in animal- to- human scaling when
extrapolating this dose- response curve to the estimation of upper bounds on
human risk. In one of the methods, CFSAN assumes that cancer risks are equal
in animals and humans when doses are similar on a lifetime- averaged
milligram/ kilogram/ day basis (i. e., body weight scaling). In the other
method, CFSAN bases its interspecies dose scaling on body weight to the ï¿½
power (in the absence of information to the contrary). Although the

literature suggests that scaling methods can have a significant impact on
risk assessment results, FDA officials said that using one approach versus
the other makes relatively little difference. Also, because tumor rates can
be biased by intercurrent mortality in animal studies (i. e., some animals
die 17 FDA may do a detailed assessment for other purposes, such as helping
to determine

regulatory priorities.

during the study from causes other than the tumor type being investigated),
CFSAN uses a statistical procedure to make adjustments for intercurrent
mortality in testing and estimating tumor rates. CFSAN procedures for
exposure assessments to food and color additives are largely driven by the
FFDCA requirement that the safety of a chemical compound be assessed in
terms of the total amount of the compound in the diet. Therefore, to
determine exposure, CFSAN risk assessors must consider all potential uses of
the compound being reviewed. Similarly, in defining the allowable limits,
the assessors must conclude that the sum total of all of these uses is
within safe limits. CFSAN generally assumes in its exposure assessments that
the compound is present at its maximum proposed use level in all foods in
which it may be used, that any contaminants are present at residue levels
established through chemical analysis, and that consumers are exposed to the
additive every day. Although most of the agency?s focus is on chronic (long-
term) exposures,

the agency must also sometimes focus on very short- term, or even single,
exposures, especially for contaminants associated with acute toxic effects.

The first component in CFSAN?s exposure assessment for food safety is the
determination of the concentrations (i. e., use levels or residue levels, in
the case of a chemical contaminant) of a chemical in foods. In the premarket
approval process, the sponsor of the petition or notification provides this
information. For postmarket assessments, information may come from

focused field surveys or from established monitoring programs such as the
Total Diet Study, which has provided data since 1961 on dietary intakes of a
variety of food contaminants, including pesticides, industrial chemicals,
toxic and nutritional elements, and vitamins and radionucleides. Analyses
are performed on foods prepared for consumption in order to provide a

realistic measure of human intake.

The second component of CFSAN?s exposure assessment is determining the
extent of consumption of different foods. In this process, CFSAN primarily
relies on multiple- day national food consumption surveys, and focuses on
the upper end of the food intake distribution (i. e., the heaviest consumers
of particular foods). CFSAN assumes that, within demographic

subgroups, all variation in the survey data represents variation among
individuals. That is, the average daily consumption of a food during the
survey period is assumed to apply to that person for his or her whole life,
and the intakes for different survey participants are assumed to reflect
differences from one person to the next in each person?s lifetime
consumption. This default assumption has acknowledged biases that result in
both overestimating high- end chronic exposures and

underestimating the proportion of the population ever consuming particular
foods. 18 To complete the exposure assessment, levels of an additive or
contaminant

in each food type are combined with estimates of daily consumption of each
food type to give a total estimated daily intake. FDA may calculate
exposures for various demographic groups, attempting to characterize both a
mean exposure and an exposure for the heavy consumer (typically consumers at
the 90 th percentile of the intake distribution). FDA officials

also pointed out that the exposure models they use for direct food additives
are very different from those for food- contact substances (e. g.,
packaging). 19 For the latter, they said that the bottom line is usually a
mean exposure.

18 EPA?s pesticide office uses an alternative assumption- that surveyed
variation represents day- to- day variation. Using this assumption, the
exposure estimate represents average chronic consumption, but fails to
estimate high- end exposures. 19 See CFSAN?s ?Guidance for Industry:
Preparation of Premarket Notifications for Food Contact Substances:
Chemistry Recommendations.?

FDA officials said that for risk management purposes they may attempt to
show the implications of different scenarios used to estimate risk. FDA
noted that a computer program that employs Monte Carlo techniques has been
developed to study the effects of variability and uncertainty of potency and
exposure estimates on estimates of risk. 20 Such complex

analyses have been applied principally to contaminants rather than in the
premarket evaluations for food and color additives.

CVM Risk Assessment CVM uses risk assessment in both the premarket approval
process and

Procedures postmarket surveillance. Risk assessments support risk management
decisions such as the development of safe concentration values and

residue tolerances for these drugs in foods. The primary human health
concern in chemical risk assessment for CVM is animal drug residues in food.
Residue is defined as any compound present in edible tissues (including milk
and eggs) of the food- producing animal that results from the use of the
chemical compound, including the compound, its metabolites, or other
substances formed in or on food because of the use of the compound. Like
CFSAN, CVM?s risk assessment procedures vary based on whether noncancer or
cancer risks are at issue. According to FDA officials, the center?s risk
assessment procedures for noncarcinogens are similar to those used by the
rest of FDA, and are based on laboratory animal data, estimated daily food
consumption, drug and metabolite residue data, and

appropriate safety factors. CVM?s guidelines for industry note that the
agency will calculate the ADI from the results of the most sensitive study
in the most sensitive species. The center will normally use different safety
factors depending on the type of study supporting the ADI calculation. When
using the ADI to calculate the ?safe concentration? for an animal drug
product, CVM uses standard values for residues of veterinary drugs in edible
tissues for the weight of an average adult and the amount and

proportion of meat products, milk, and eggs consumed per day. CVM officials
pointed out that the consumption values in their guidelines for industry are
standard values used by the Joint Expert Committee on Food Additives,
sponsored by the World Health Organization and Food and

20 Monte Carlo analysis involves a repeated random sampling from the
distribution of values for each of the parameters in a calculation (such as
average daily exposure) to derive a distribution of estimates of exposures
for a population. According to FDA, because Monte Carlo modeling is a
probabilistic technique that can use all the available food intake and
concentration data, it will result in more accurate estimates at upper
percentiles of exposure than those obtained using point values from consumer
surveys.

Agriculture Organization, that provides food safety recommendations to the
Codex Committee on Residues of Veterinary Drugs in Foods. For carcinogen
risk assessments, CVM uses a nonthreshold, conservative, linear- at- low-
dose extrapolation procedure to estimate an upper limit of low- dose risk
(as described under CFSAN). Cancer risk estimates are

generally based on animal bioassays, and upper 95- percent confidence limits
of carcinogenic potency are used to account for inherent experimental
variability. FDA officials noted that some elements and assumptions of its
dose- response analysis procedures are likely to overestimate risk by an
unknown amount. Similarly, some of its assumptions on exposure may also
overestimate cancer risks. For example, CVM?s risk assessment procedures
assume that the concentration

of residue in the edible product is at the permitted concentration and that
consumption is equal to that of the 90 th percentile consumer. In addition,
the agency assumes that all marketed animals are treated with the
carcinogen. While acknowledging that all of these assumptions result in
multiple conservatisms, FDA also states that they are prudent because of the
uncertainties involved.

CDRH Risk Assessment Medical devices, supplies, and implants may contain
chemicals that can Procedures

leach out of the devices into surrounding tissues. Risks from these types of
chemical contaminants are considered during the premarket review of the
material safety of a device, but concerns may also arise during CDRH?s
postmarket surveillance activities. 21 According to FDA officials, the
concentrations of such leachants in human tissues are generally small and
amenable to typical safety risk assessment procedures.

21 As an example of a postmarket quantitative risk assessment, FDA cited its
assessment of cancer risk posed by patient exposure to 2, 4- toluenediamine
released from polyurethane foam- covered breast implants.

CDRH has issued guidance for the preclinical (premarket) biological safety
evaluation of medical devices. 22 In that guidance, CDRH recognizes and uses
a number of domestic and international consensus standards that have been
developed to address aspects of medical device safety, including risks posed
by exposure to compounds released from medical devices. However, CDRH
officials pointed out that they and medical device approval applicants may
use approaches other than those described in the

consensus standards to conduct risk assessments. They said the standard that
comes closest to describing CDRH?s approach for chemical risk assessment is
International Organization for Standardization (ISO)/ FDIS 10933- 17. 23
CDRH officials noted that, although this international standard is still in
draft and has not been formally recognized by the center, the

methods that it describes represent the primary procedures used by CDRH to
assess the risk posed by patient exposure to compounds released from medical
devices. They also pointed out that this standard is unique among risk
assessment guidelines in that it provides methods to derive healthbased

exposure levels for local effects such as irritation, which often ?drive?
the risk assessment for compounds released from implanted devices. According
to CDRH, hazards posed by patient exposure to a device are typically
determined after subjecting the device to a series of tests defined by the
preclinical evaluation guidance. Evaluation of potential toxicity is
supposed to cover a number of adverse effects, including local or systemic
effects, cancer, and reproductive and developmental effects. Unless
justification is otherwise provided, CDRH assumes that the results

obtained in animal studies are relevant for humans. One notable exception
for medical device risk assessment, according to CDRH, is that implantation-
site sarcomas (malignant tumors) found in rodents are not assumed to be
relevant for humans.

One option available to applicants is to use a risk assessment approach
involving: (1) characterization of the chemical constituents released from a
22 The Center for Devices and Radiological Health?s Office of Device
Evaluation ?Blue Book? memorandum #G95- 1, ?Use of International Standard
ISO- 10993, ?Biological Evaluation of Medical Devices Part- 1: Evaluation
and Testing. ??

23 Biological evaluation of medical device Part 17: Methods for the
establishment of allowable limits for leachable substances. The
International Organization for Standardization is a nongovernmental
federation of national standards bodies from about 140 countries. Its work
results in international agreements that are published as international
standards.

device; (2) derivation of a tolerable intake (TI) value for the compound;
and (3) comparing the dose of each constituent received by a patient to its
respective TI value. A TI value is a dose of a compound that is not expected
to produce adverse effects in patients following exposure to the compound
for a defined period. According to CDRH, it is conceptually similar to EPA?s
reference dose, but different TI values can be derived for a

compound depending on the route and duration of exposure to the medical
device. CDRH?s procedures recommend establishing TI values for noncancer
adverse effects using standard uncertainty factors in order to account for
interspecies and inter- individual differences in sensitivity. However, CDRH
permits flexibility in the event that data are available to characterize
these

uncertainties more accurately. CDRH also uses a lumped uncertainty factor to
adjust for limitations in data quality such as (1) the use of short- term
studies in the absence of long- term studies, (2) the absence of supporting
studies, and (3) use of studies involving different routes or rates of
exposure. According to CDRH, this lumped uncertainty value typically

does not exceed 100, but can exceed 100 when acute (short- term) toxicity
data are the only basis of the calculation of a TI value for permanent
exposure. CDRH considers this provision especially important for medical
device risk assessment because of the paucity of long- term toxicity data
for

many of the compounds released from medical devices. For carcinogenic
leachants, FDA often uses low- dose linear extrapolation techniques. For a
device- released compound that has been determined to be a carcinogen, CDRH
uses a weight- of- the- evidence approach to

determine the likelihood that it exerts its carcinogenic effect via a
genotoxic mechanism. 24 If the evidence suggests that the compound is
genotoxic, then CDRH uses quantitative risk assessment to estimate a TI
consistent with a risk level of 1 per 10,000. No specific quantitative risk
assessment approaches have been identified as better than others for

conducting the cancer risk assessment. If, however, the weight- of-
theevidence test suggests that the compound is a nongenotoxic carcinogen,
the uncertainty factor approach described above should be employed to derive
the TI.

24 A genotoxic carcinogen is one that initiates cancer through a direct
effect on genetic material. It is capable of causing heritable changes or
damage leading to heritable changes in genetic material.

Once the TI is derived for each compound released from a device, it is then
converted to a tolerable exposure value by taking into account the body
weight of the patient and the usage patterns of the device that releases the
compound. Overall, the agency noted that one of the most challenging
problems in risk assessments for devices is determining the level of
exposure to leached chemicals.

Risk Assessment As previously noted, FDA does not require the use of a
specific risk Assumptions and assessment protocol or of specific default
assumptions. However, the

summary of FDA procedures also demonstrated that assumptions and
Methodological methodological choices are an integral part of a risk
assessment. FDA Choices

officials noted that they employ many default assumptions or choices by
precedent. In particular, FDA officials and several reference documents on
FDA risk assessment procedures pointed out that the agency routinely
incorporates conservative assumptions into its assessments in the face of
uncertainty. The report on the U. S. food safety system emphasized that
precaution is embedded in the underlying statutes and the actions of
regulatory agencies to ensure acceptable levels of consumer protection. 25
Therefore, precautionary approaches are very much a part of the agency?s

risk analysis policies and procedures. Although not intended to be
comprehensive, the following table illustrates in detail some of the
specific assumptions or methodological choices that are used in FDA as a
whole and within particular FDA product centers.

The information in the table was taken primarily from FDA documents, but
also reflects additional comments provided by FDA officials. (GAO notes and
comments appear in parentheses.)

25 Precaution in U. S. Food Safety Decisionmaking.

Table 7: FDA Risk Assessment Assumptions and Methodological Choices When the
assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

(Methodological assumptions According to FDA officials, According to FDA
officials, the

(Not applicable to this generic and choices in general.) ?reasons? always
include the circumstances under which

information.) following three general items:

these assumptions would be used always include that data

 information essential to a risk indicating the assumption may assessment
is not known; be invalid for a particular  the assumption, selected by
circumstance are not available.

FDA to substitute for that information, is scientifically plausible; and

 the effect of using the assumption is consistent with agency regulations
or policy (e. g., for CFSAN, an assumption is expected to be

reasonably protective of human health).

1. CFSAN general procedures regarding food safety 1.1

This is a fundamental When there are no data or For chemicals that are human
Animal studies are useful for assumption of quantitative risk inadequate
data on human

carcinogens, animal data may human risk assessment, with assessment,
accepted as a

outcomes. underestimate human risk by up appropriate uncertainty factors

basic precedent according to to one order of magnitude and (i. e., you can
extrapolate human FDA officials. They noted that, in overestimate risk by up
to three effects from animal data). general, adequate information

orders of magnitude [citations from human studies is preferred provided]. to
adequate data from animal studies. However, when human data are not
available, or when such data are judged to be

relatively insensitive to moderate- or low- level risks and difficult to
quantify, FDA makes the assumption that animal studies can be used for human
risk assessment. FDA officials also said that the application of appropriate
uncertainty factors

to data from animal studies makes this assumption reasonably protective of
human health.

When the assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

1.2 (Not identified in the FDA

When choosing which animal (Not specifically identified, but Use the most
sensitive animal

documents that we reviewed study data to use for hazard

FDA officials commented that species, unless you have describing risk
assessment identification and dose- response risk assessment literature
information to the contrary. procedures. However, FDA

estimation, when the most identifies the ?usual? range of officials
explained that appropriate animal model for

toxicological responses among comparative absorption,

humans is not known. animal models for this metabolism, distribution, and

assumption.) elimination data could identify the most appropriate animal
model for extrapolating results

from animals to humans. They also stated that FDA has determined that use of
this assumption is reasonably protective of human health.)

1.3 FDA officials identified this When doing low- dose

(Not explicitly identified. FDA You can extrapolate low- dose

assumption as another extrapolation because low- dose officials explained
that the effects from high- dose effects. fundamental assumption, similar

effects were not directly literature on this subject shows to 1. 1 above,
noting that, in measured in available human or

?very large? variability among general, low- dose effects cannot animal
studies and the

predictions of low- dose effects be directly measured in human

mechanism of action at low using different modeling epidemiology studies or
in doses is not known to be assumptions. They also stated animal toxicity
and

different than the mechanism at that, because they do not know
carcinogenicity studies. They higher doses. the correct relationship between
said that, in such circumstances,

high- and low- dose effects for a the mechanism of action at low substance,
they do not know doses is assumed to be the how much conservatism is same as
that operating at higher associated with the use of this doses, which makes
the related assumption.) assumption that one can extrapolate effects at low
doses from effects at high doses scientifically plausible. FDA has

determined that such an assumption is reasonably protective of human health.

When the assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

1.4 Not specifically identified, but When doing low- dose

This linear extrapolation provides Low- dose cancer risk estimation FDA
noted in general that, extrapolation for carcinogens. upper- bound estimates
of cancer is done using a linear, nothreshold because it is not possible to
risk at low doses.

approach. determine the accuracy and

precision of low- dose cancer The point of departure for this estimates, the
agency employs linear extrapolation is the upperconfidence conservative risk
assessment

limit on risk at a point procedures to compensate for on the dose- response
curve

weaknesses in scientific rigor. below which the data no longer appear to be
reliable. However, The agency did provide specific the agency does not
specify any

citations for various low- dose particular mathematical form for
extrapolation procedures. the dose- response relationship in the
experimental dose range,

just that there is an adequate fit to the data.

A more specific, but related, string of assumptions is that genotoxic
carcinogens do not have thresholds, and the appropriate low- dose
extrapolation is linear.

1.5 The ï¿½ power approach was When extrapolating equivalent (Not addressed in
FDA

Dose scaling across species is recommended in a 1992 doses of a carcinogen
from materials, but other sources based on body weight to the ï¿½ proposed
rule by an interagency

animals to humans, in the indicated that the ?body weight ï¿½ power (an
adjustment factor for committee [citation provided].

absence of information to the power? approach is considered calculating the
dose at which contrary.

to produce the midpoint of cancer risks are equal in rodents plausible
values from among the and humans)

common alternative approaches for interspecies dose scaling, (In meetings
with CFSAN while ?body weight? scaling is officials, we were informed that
considered to produce the least the agency?s risk assessors may

conservative values. A CFSAN also use the default that cancer official
stated that this choice risks are presumed equal when makes little
difference in the daily amounts of a chemical results.) agent are scaled in
proportion to a species? body weight.)

When the assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

1.6 (Note that the underlying statute

Risk assessments for noncancer The resulting ADI is an estimate FDA/ CFSAN
uses a basic no specifically directs the use of endpoints estimating an ADI

with uncertainty spanning observed effect level (NOEL) safety factors.) from
the results of animal

perhaps an order of magnitude safety factor approach for experiments.

of a daily exposure that is likely noncancer effects. The agency to be
without appreciable risk of generally uses standard safety

deleterious effects during a factors to account for lifetime. uncertainty:
a. A safety factor of 10 is used to account for the possible increased
sensitivity of humans compared to test animals.

b. Another safety factor of 10 is used to account for the increased
sensitivity of some humans. c. Additional safety factors are

often employed to account for long- term effects based on short- term
experiments, use of a LOAEL instead of a NOEL/ NOAEL, and other

inadequacies of the experimental data.

When the assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

1.7 a. Empirical: based on food Assumptions about consumption FDA officials
told us that these CFSAN uses a series of consumption survey data.

used when estimating human assumptions provide a publichealth

assumptions about food b. Concern that the mean

exposure to chemical conservative estimate of consumption:

exposure is not sufficiently substances (i. e., additives, exposure for
comparison with

public- health conservative. residues) in foods.

the ADI. a. The agency assumes that c. The method for estimating the average
total diet of a

the 90 th percentile in the 60- kg adult generally absence of other data is
consists of 1500 grams (g)

based on observations from of solid food and 1500 g of dietary survey
distribution liquid per day.

data for a variety of b. For calculating exposure to products. additives or
contaminants, CFSAN frequently uses the 90 th percentile of consumption to
represent

intake of heavy consumers. c. In the absence of the distribution of intakes
among individuals in a

population and a direct measure of the 90 th percentile, the 90 th
percentile is estimated to be

two to three times the average intake.

1.8 FDA officials said they use this When basing exposure

According to the Rhomberg The agency usually assumes assumption because only
shortterm estimates on food intake report, the effect of this that, within
demographic groups, food intake surveys are

surveys. assumption is an acknowledged food intake survey variation

available. overestimation of high- end

represents differences in chronic chronic exposures and an consumption among
individuals. underestimation of the

The average daily consumption proportion of the population ever by an
individual of a food during

consuming particular foods. the survey period is assumed to apply to that
person for his or her

whole life.

When the assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

1.9 FDA officials said that they use

When doing exposure FDA officials said these The agency generally assumes

these public- health conservative assessments. assumptions were public-
health that all foods in which a chemical assumptions because of the lack
conservative. additive is proposed or permitted of scientifically sound for
use will bear it at the information that would permit maximum proposed or

their relaxation. permissible use level.

The agency also assumes that consumers are exposed to an additive every day
for a lifetime.

1.10 The agency noted that its intent

When doing risk assessments The intent is to be conservative.

When dealing with mixtures of is to be conservative, and it cited

involving mixtures of chemicals carcinogens, the agency usually research on
the estimation of

(with dose levels for each considers that interactions are upper- confidence
limits on component below those having not likely and uses some estimates of
risk for mixtures.

measurable effects for that standard assumptions:

compound). a. It is assumed that

carcinogens are acting independently. Therefore, the risk of cancer from a
mixture may be obtained by summing the individual risk. b. It is generally
assumed that

all carcinogenic components are at their tolerance concentrations.

When the assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

2. CVM procedures regarding animal drug residues in food 2.1

In general, FDA noted that all of When conducting cancer risk This set of
assumptions was As a matter of policy regarding

the assumptions identified in this assessments. identified generically as
potential risks from residues of section on the DES Proviso (see ?resulting
in multiple carcinogenic animal drugs in following rows 2.2 through 2.5),

conservatisms.? foods consumed by humans (per

?result in multiple conservatisms, the DES Proviso to the Delaney but are
prudent because of the Clause), FDA accepts a lifetime uncertainties.? risk
below one per million as an insignificant level.

?Thus, adoption of a [residue] concentration associated with a cancer risk
of one per million is likely to be well below that level

of risk and to satisfy the FDA?s responsibility under the statute [FFDCA] to
ensure to a reasonable certainty that the public will not be harmed.?

2.2 (No additional information

Choosing data sets to use for (See previous row.)

Regulation is based on the target provided, but see row 2.1 for cancer risk
assessments. tissue site exhibiting the highest

general rationale.) potential for cancer risk for each carcinogenic
compound.

If tumors are produced at more than one tissue site, the minimum
concentration that produced a tumor is used.

When the assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

2.3 Upper 95- percent confidence

When doing low- dose The process of linearly Cancer risk estimates are

limits are used to account for extrapolation from animal

extrapolating from the high generally based on animal inherent experimental
variability.

bioassay data in the absence of doses used in animal bioassays bioassays,
using upper 95percent information establishing the

to concentrations of residues is confidence limits of The low- dose
extrapolation

mechanism of carcinogenesis of likely to overestimate risk by an
carcinogenic potency.

procedure estimates an upper a compound.

unknown amount. limit on low- dose risk. Low- dose extrapolation is done
using a nonthreshold,

In its general principles for conservative, linear- at- low- dose industry
(guideline #3), CVM procedure (modified GaylorKodell). notes that none of
the mathematical procedures for

extrapolation has a fully adequate biological rationale, because the
mechanism of carcinogenesis is not sufficiently understood.

2.4 (No additional information

In cancer risk assessments From rather limited data It is assumed that the

provided, but see row 2.1 for when doing low- dose comparing human and
animal

carcinogenic potency in humans general rationale.) extrapolation from animal

cancer potencies of compounds, is the same as that in animals.

bioassay data. animal results are likely to

overestimate human cancer risk but could underestimate risk by an order of
magnitude (citations provided).

When the assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

2.5 (See row 2.1 for general Used for exposure assessments These assumptions
on Multiple assumptions are used in rationale. Note that item d is for
cancer risks. exposures also overestimate the

carcinogen exposure taken from CVM?s guidance for risk. assessments:

industry.) a. The concentration of

residue in the edible product is at the permitted concentration. b.
Consumption is equal to that of the 90 th percentile consumer.

c. All marketed animals are treated with the carcinogen. d. In the absence
of information about the

composition of the total residue in edible tissue, assume that the entire

residue is of carcinogenic concern.

2.6 FDA officials said this In noncancer risk assessment, (Not identified,
but the intent of For noncancer toxicological

assumption was used because when calculating the ADI based

this calculation is to provide endpoints, the agency will of historical
precedent back to at upon animal studies and in the

protective, public- health calculate the ADI from the NOEL least 1954,
because, in the

absence of data suggesting that conservative estimates.)

of the most sensitive effect in an absence of data (e. g., an alternative
approach is animal study of the most

physiologically based scientifically justified.

sensitive sex and species. pharmacokinetic or other data) suggesting that
another species is more appropriate to use for

extrapolating to humans, using the most sensitive endpoint in the most
sensitive species and sex is still considered to be protective of public
health.

When the assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

2.7 These are identified as In noncancer risk assessment

(Not identified, but the intent of The agency will normally use the
?appropriate? safety factors. when calculating the ADI by such factors is
usually to provide following safety factors, dividing the NOEL from an
protective, public- health depending on the type of study animal study by
safety factors.

conservative estimates.) supporting the ADI estimate:

a. chronic study (factor of 100), b. reproduction/ teratology study (100 for
a clear

indication of maternal toxicity; 1000 for other effects), and c. 90- day
study (1000).

2.8 CVM officials noted that the

Exposure analysis for noncancer (Not identified, but the intent of The
agency uses a series of

standard consumption factors risk assessments. Consumption such factors is
usually to provide standard values and (e. g., 300 grams of muscle) were

values applied to determine the protective, public- health assumptions to
estimate an established by the Joint Expert safe concentration for most new
conservative estimates.)

individual?s daily food Committee on Food Additives animal drug products,
unless an consumption: (sponsored by the World Health appropriate scientific
justification Organization and the Food and

supports alternative a. edible muscle (300 grams); Agriculture Organization
of the

consumption values. b. liver (100 g);

United Nations). c. kidney (50 g); d. fat (50 g); e. assume that when an

individual consumes a full portion of a meat product from one species, he or
she will not consume a full portion of a meat product from another species;
f. assume a person consumes a full portion of milk (1. 5

liters) in addition to the full portion of edible muscle or organ tissue;
and g. assume that a person consumes a full portion of eggs (100 g) in
addition to

the consumption of muscle or organ tissue.

When the assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

2.9 CVM officials said this CVM officials said that this CVM officials said
the

The agency assumes the weight assumption is used because of assumption is
applied in all assumption should protect of an average adult is 60 kg
historical precedent and

circumstances. women, growing adolescents, when calculating the safe because
the 60- kg person is and the elderly.

concentration of a compound in likely to cover the exposure of edible tissue
of a food- producing women, growing adolescents, animal. and the elderly. 3.
CDRH procedures regarding medical devices (For purposes of this report,
limited to chemical leachants. Does not include engineering or radiation
risk assessments.)

3.1 (Not specifically addressed in During hazard identification. (Not
mentioned.)

CDRH assumes that the results the material provided by CDRH, obtained in
animal studies are although the first part is a relevant for humans, but
with one

standard assumption of notable exception for medical quantitative risk
assessment.) device risk assessment implantation- site malignant

tumors found in rodents are not assumed to be relevant for humans.

3.2 To provide a safety factor. During hazard identification. (Not
specifically identified, but Extractions of medical devices

the intent of such factors is (to test for adverse effects) are usually to
provide protective, typically carried out using

public- health conservative rigorous but not exhaustive estimates.)
conditions to provide a ?safety factor? for the hazard identification phase.

When the assumption/ choice would be applied (step in the Assumption or risk
assessment process or Likely effect on risk methodological choice Reason( s)
for selection circumstances) assessment results

3.3 The first two factors are standard

Used when calculating TI values (Not specifically identified, but The
following uncertainty factors uncertainty factors for noncancer effects of
chemical the intent of such factors is are typically used for adverse

recommended in the scientific compounds, unless case data usually to provide
protective,

effects other than cancer: literature.

are available to characterize public- health conservative these
uncertainties more estimates.) a. a factor to adjust for

The lumped uncertainty factor accurately. interspecies differences in takes
into account: sensitivity to toxic compounds;

1. use of short- term studies in b. a factor to adjust for the absence of
long- term interindividual differences in studies;

sensitivity to toxic 2. having only a LOAEL

compounds; and instead of a NOAEL;

c. a ?lumped? factor to adjust 3. absence of supporting for limitations in
data quality.

studies; 4. route- to- route extrapolation of dose, when needed;

5. rate of exposure; and 6. confidence in the data base.

CDRH noted that the lumped factor typically does not exceed 100, but can
exceed this value when acute toxicity data are the only basis for the
calculation of a TI value for permanent exposure.

Source: Compiled from GAO review of FDA risk assessment guidelines and
related documents and from additional comments provided by agency officials.

Risk Characterization Unlike EPA, FDA does not have an official policy on
how the results of the agency?s risk assessments should be characterized to
decision makers and

the public. However, FDA officials said that, in practice, the agency uses a
standard approach for risk characterization that is similar to EPA?s
official policy. They said that FDA?s general policy is to reveal the risk
assessment assumptions that have the greatest impact on the results of the
analysis, and to state whether the assumptions used in the assessment were
conservative. FDA officials also said that their risk assessors attempt to

show the implications of different distributions and choices (e. g., the
results expected at different levels of regulatory intervention). As noted
earlier, FDA may employ methods such as Monte Carlo techniques to provide
additional information on the effects of variability and uncertainty on
estimates of risk.

There are some differences in FDA risk characterization procedures depending
on the products being regulated and the nature of the risks involved. For
food ingredients (direct and indirect food additives, color additives used
in food, and substances generally recognized as safe) and animal drug
residues that are not carcinogenic, risk characterization under the FFDCA
focuses on whether the mandate of reasonable certainty of no harm will be
achieved given the proposed limits on use and permissible residues. The main
issue is whether the higher end (the 90 th percentile) of

the distribution of estimated daily intakes is below the ADI calculated from
toxicity data. The statutory mandate is interpreted as requiring that, for a
food additive to be declared safe, heavy consumers of particular foods

should be reasonably assured of protection even if residues were at the
maximum level allowed. For carcinogenic impurities, FDA?s focus is also on
characterizing whether there is reasonable certainty of no harm. However,
because of the Delaney clause, risk characterization is not needed for
carcinogenic food ingredients. Residues of carcinogenic animal drugs are
also evaluated separately under the DES proviso.

CDRH officials pointed out that the draft ISO/ FDIS 10933- 17 international
standard explicitly addresses one risk characterization issue- how sensitive
subpopulations should be taken into account when setting allowable limits
for compounds released from devices. Although it states that ?idiosyncratic
hypersusceptibility? should not normally be the basis of the tolerable
exposure or allowable limit, the ISO standard does not preclude setting
standards in this manner. Furthermore, the standard says that limits should
be based on the use of the device by the broadest segment of the anticipated
user population. Therefore, if a device is intended for a specific
population, such as pregnant women, estimates

should be based on that population.

Chemical Risk Assessment at the Occupational Safety and Health

Appendi x V I Administration Although the Occupational Safety and Health
Administration (OSHA) generally follows the standard four- step National
Academy of Sciences? (NAS) paradigm for risk assessment, there are several
distinguishing characteristics of its assessments. Under its statutory
mandate, OSHA has a specific and narrow focus on the potential risks to
workers in an occupational setting. Further, the underlying statute and
court decisions interpreting the statute have required the agency to focus
on

demonstrating, with substantial evidence, that significant risks to workers
exist before it can regulate. In addition to presenting its own best
estimates of risk, OSHA may present estimates based on alternative methods
and assumptions.

Context for OSHA Much of what is distinct about risk assessment at OSHA can
be traced to

Chemical Risk statutory provisions, court decisions, and the nature of
workplace

exposures to chemicals. OSHA, an agency within the Department of Labor,
Assessment

was created by the Occupational Safety and Health Act of 1970 (the OSH Act).
1 The central purpose of the act is to ensure safe and healthful working
conditions. As one of the primary means of achieving this goal, the act
authorizes the Secretary of Labor to promulgate and enforce mandatory
occupational safety and health standards. 2 Certain provisions

in the act stipulate both the nature and the manner in which these standards
should be established. For example:

 Under section 3( 8) of the OSH Act, a safety or health standard is defined
as a standard that requires conditions, or the adoption or use of one or
more practices, means, methods, operations, or processes, reasonably

necessary or appropriate to provide safe or healthful employment or places
of employment.

 According to OSHA, a standard is reasonably necessary or appropriate
within the meaning of section 3( 8) if it eliminates or substantially
reduces significant risk and is economically feasible, technologically
feasible, cost effective, consistent with prior OSHA action or supported by
a reasoned justification for departing from prior OSHA actions, supported by
substantial evidence on the record as a whole, and is 1 29 U. S. C. 651 et
seq.

2 Safety standards are generally designed to reduce on- the- job injuries.
Health standards are usually directed at limiting the risk of workers
developing occupational diseases from exposure to hazardous chemical or
physical agents.

better able to effectuate the act?s purposes than any national consensus
standard it supersedes.

 Section 6( b)( 5) of the act states that ?The Secretary, in promulgating
standards dealing with toxic materials or harmful physical agents shall set
the standard which most adequately assures, to the extent feasible, on the
basis of the best available evidence, that no employee will suffer material
impairment of health or functional capacity even if such employee has
regular exposure to the hazard dealt with by such standard for the period of
his working life.? A significant factor influencing the interpretation of
the OSH Act provisions and OSHA?s approach to risk assessment is the Supreme
Court ruling in its 1980 ?Benzene? decision that, before issuing a standard,
OSHA must demonstrate that the chemical involved poses a ?significant risk?
under workplace conditions permitted by current regulations and that the new

limit OSHA proposes will substantially reduce that risk. 3 This decision
effectively requires OSHA to evaluate the risks associated with exposure to
a chemical and to determine that these risks are ?significant? before
issuing a standard. However, the court provided only general guidance on
what level of risk should be considered significant. The court noted that a

reasonable person might consider a fatality risk of 1 in 1000 (10 -3 ) to be
a significant risk and a risk of one in one billion (10 -9 ) to be
insignificant. Thus, OSHA considers a lifetime risk of 1 death per 1, 000
workers to

represent a level of risk that is clearly significant. The court also stated
that ?while the Agency must support its findings that a certain level of
risk exists with substantial evidence, we recognize that its determination
that a particular level of risk is significant will be based largely on
policy considerations.? 4

Later Court of Appeals decisions have interpreted the Supreme Court?s

?Benzene? decision to mean that OSHA must quantify or explain the risk for
each substance that it seeks to regulate unless it can demonstrate that a
group of substances share common properties and pose similar risks. 5
Although this decision does not require OSHA to quantitatively estimate the

risk to workers in every case, it does preclude OSHA from setting new 3
Industrial Union Dept. v. American Petroleum Inst., 448 U. S. 607, 642
(1980). 4 448 U. S. at 655- 56 n. 62. 5 AFL- CIO v. OSHA, 965 F. 2d 962 (11
th Cir. 1992), International Union, UAW v. Pendergrass,

878 F. 2d 389 (D. C. Cir. 1989).

standards without explaining how it arrives at a determination that the
standard will substantially reduce a significant risk. According to OSHA
officials, the other important contextual influence on OSHA risk assessment
is the very nature of workplace exposures to chemicals. Generally, workplace
exposures to chemicals are at higher

levels than most environmental exposures to chemicals experienced by the
general public. Workers are often exposed to many chemical agents at levels
not much lower than those used in experimental animal studies.

According to agency officials, this is one of the unique features of OSHA?s
chemical risk assessments. Also, OSHA frequently has relevant human data
available on current exposures, in contrast to most other agencies
regulating toxic substances.

Risk Assessment Procedures

General Approach OSHA currently has no formal internal risk assessment
guidance. Instead, OSHA has primarily described its general risk assessment
methods, as well as the rationale for specific models and assumptions
selected, in the record

of each risk assessment and regulatory action. One reason for this,
according to agency officials, is that OSHA performs risk assessments only
for its standards. Overall, they said the agency only publishes two or three
proposed or final rules per year, and not all of these rules involve a
chemical risk assessment. The officials also emphasized the incremental
nature of advances in risk assessment methods and science, with successive
assessments establishing precedents for methods that may be used for
succeeding analyses.

Like EPA and FDA, OSHA uses the basic NAS four- step process for risk
assessment. Another fundamental source for OSHA?s (and EPA?s and FDA?s)
methods was the 1985 document on chemical carcinogens produced by the Office
of Science and Technology Policy. 6 OSHA often refers to the reference
sources of other entities, including other federal agencies, in 6 Office of
Science and Technology Policy, Executive Office of the President, ?Chemical
Carcinogens: A Review of the Science and Its Associated Principles, February
1985,? 50 FR 10372 (Mar. 14, 1985).

both specific rulemakings and as general technical links to its on- line
information on occupational risks. Despite these common elements and
procedures, several features of OSHA?s approach differ from those of other
federal agencies. Because OSHA does not currently have written internal
guidance on its risk

assessment procedures, the information in the following sections is derived
primarily from an examination of OSHA?s chemical risk assessments. 7 We also
relied on secondary sources, such as Lorenz Rhomberg?s report on

federal agencies? risk assessment methods. 8 Hazard Identification In OSHA?s
risk assessments, the hazard identification step results in a determination
that an exposure to a toxic substance causes, is likely to cause, or is
unlikely or unable to cause, one or more specific adverse health effects in
workers. According to OSHA, this step also shows which studies have data
that would allow a quantitative estimation of risk. OSHA defines hazardous
and toxic substances as those chemicals present in the workplace that are
capable of causing harm. In this definition, the term chemicals includes
dusts, mixtures, and common materials such as paints, fuels, and solvents.
OSHA currently regulates exposure to approximately 400 such substances. In
the workplace environment, chemicals pose a wide range of health hazards (e.
g., irritation, sensitization, carcinogenicity, and noncancer acute and
chronic toxic effects) and physical hazards (e. g., ionizing and nonionizing
radiation, noise, and vibration).

7 Because OSHA?s methylene chloride standard is the most recent hazardous
chemical rulemaking, most of the specific examples cited come from OSHA?s
methylene chloride rulemaking: ?Occupational Exposure to Methylene
Chloride,? 62 FR 1494 (Jan. 10, 1997). 8 Lorenz R. Rhomberg, A Survey of
Methods for Chemical Health Risk Assessment Among Federal Regulatory
Agencies, a report prepared for the National Commission on Risk Assessment
and Risk Management (1996).

Most of OSHA?s chemical risk assessments have addressed occupational
carcinogens. In assessing potential carcinogens, OSHA may consider the
formal hazard classification or ranking schemes of other entities as part of
the available evidence on a particular chemical. Ultimately, though, OSHA

makes its own determinations on the risk posed by particular compounds and
their classification as potential occupational carcinogens. OSHA?s chemical
risk assessments may also discuss noncancer hazards. For example, in the
final rule on methylene chloride the agency discussed the evidence regarding
central nervous system, cardiac, hepatic (liver), and reproductive toxicity,
as well as carcinogenicity. Similarly, the agency?s rulemaking on 1,3-
butadiene addressed adverse health effects such as developmental and
reproductive toxicity and bone marrow effects in addition to the evidence on
carcinogenicity. 9 OSHA quantifies the risks of

noncancer effects if it determines that there are adequate data on exposure
and response for the substance of interest. OSHA officials also noted that
OSHA has a hazard communication standard, which requires manufacturers,
shippers, importers, and employees to inform their employees of any
potential health hazard when handling these chemicals. This is usually done
through container labeling and material safety data sheets. Although this
standard does not address

specific risks posed by individual chemicals, it is a comprehensive hazard
information standard for nearly all chemicals in commerce.

Dose- Response Assessment Carcinogens OSHA?s general procedures for dose-
response assessment are similar to

those of EPA and FDA, especially in the choice of data sets to use for
quantitative assessments. However, OSHA probably uses a linear low- dose
extrapolation model less often than is the case for other agencies. OSHA
differs from the other federal regulatory agencies also in being less
conservative in setting its target risk levels when conducting low- dose
extrapolation. As previously noted, the main points of OSHA?s risk

assessments for regulatory purposes are to determine whether significant
risks exist and to demonstrate in a broad sense the degree to which the
standard would reduce significant risk. The specific choice of where to set

9 ?Occupational Exposure to 1, 3- Butadiene? 61 FR 56746 (Nov. 4, 1996).

the standard is tempered by the statutory mandate that standards must be
technologically and economically feasible.

Like other agencies, OSHA states that, all things being equal,
epidemiological data are preferred over data from animal studies whenever
good data on human cancer risks exist. More often than some other agencies
regulating exposures to toxic substances, OSHA may have relevant human data
on adverse health effects available for consideration in its risk
assessments. However, the rulemaking examples we reviewed also illustrate
that these epidemiological data may be considered inadequate for
quantitative dose- response assessment, while animal data may provide more
precise and useful dose- response information. In both

the methylene chloride and 1,3- butadiene dose- response assessments, for
example, OSHA had both epidemiological and animal data available, but based
its quantitative estimates on data from rodent models. However,

OSHA did use its analysis of the epidemiological data when examining the
consistency of the results derived from animal studies.

When faced with the choice of several animal data sets, OSHA tends not to
combine tumor sites but to choose the data set showing the highest
sensitivity (i. e., most sensitive sex, species, and tumor site). The agency
will, however, frequently present information from alternative data sets and
analyses. The agency is likely to include benign tumors with the potential
to progress to malignancy along with malignant tumors in the data set used
for its quantitative assessments. OSHA cited the Office of Science and
Technology Policy?s views on chemical carcinogens in support of this
practice, as well as noting that other federal agencies, including EPA and

FDA, have also included benign responses in their assessments. Because
occupational exposures tend to be closer to the range of experimentally
tested doses in animal studies, extrapolation may pose less of a challenge
for OSHA than for other regulatory agencies. OSHA?s

preferred model for quantitative analysis of animal cancer dose- response
data and for extrapolation of these data to low doses is the ?multistage
model,? which is based on the biological assumption that carcinogens

induce cancer through a series of independent ordered viable mutations, and
that cancer develops through stages. Unlike EPA and FDA, however, OSHA tends
to focus on the maximum likelihood estimate (MLE) of the fitted dose-
response curve rather than on an upper bound, although the

agency also provides estimates for the 95- percent upper confidence limit
(UCL) of the dose- response function. This procedure generally leads to a
less conservative risk estimate than the procedures used by EPA or FDA.

Like EPA and FDA, OSHA generally assumes no threshold for carcinogenesis. In
contrast to the other agencies, OSHA?s default dosemetric for interspecies
extrapolation is body weight scaling (mg/ kg/ day i. e., risks equivalent at
equivalent body weights). According to OSHA, this default is used to be
consistent with prior chemical risk assessments, but it also reflects a
conscious policy decision that its methodology should not be

overly conservative. OSHA says it may in the future move to ï¿½ -power
scaling, as agreed to by EPA, FDA, and the Consumer Product Safety
Commission some years ago. OSHA also says it is currently considering
developing a different form of the multistage model, which will provide more
stable MLE estimates than does the current form. OSHA also considered data
from physiologically based pharmacokinetic (PBPK) models in the risk
assessment examples we reviewed. 10 PBPK models provide information on
target organ dose by estimating the time distribution of a chemical or its
metabolites through an exposed subject?s system. 11 OSHA noted that PBPK
modeling can be a useful tool for describing the distribution, metabolism,
and elimination of a compound of interest under conditions of actual
exposure and, if data are adequate, can allow extrapolation across dose
levels, routes of exposure, and species. In particular, pharmacokinetic
information is useful in modeling the relationship between administered
doses and effective doses as a function of the exposure level. 12 However,
PBPK models are complicated and require substantial data, which may not be
available for most chemicals. OSHA pointed out in the methylene chloride
rule that differences in the risk

estimates from alternative assessments (including those submitted by outside
parties) were not generally due to the dose- response model used, but to
whether the risk assessor used pharmacokinetic modeling to 10
Pharmacokinetics is the study of the absorption, distribution, metabolism,
and elimination

of chemicals in humans and animals. It is the basis for developing more
realistic and accurate models of the movement and interactions of a chemical
with blood, tissues, and organs once it enters the body, including
consideration of the body?s ability to repair damage caused by a chemical.
PBPK models are based on the physiology of the exposed subjects, in

contrast to more general compartmental pharmacokinetic models that do not
necessarily represent effects on real anatomic regions/ compartments of the
body. 11 Once in the body, a chemical may be chemically altered to form
metabolites. Either the chemical itself or its metabolites may produce toxic
effects. Therefore, both may need to be considered in assessing the
potential harm associated with a given chemical. 12 The administered dose is
the amount of a substance given to an animal or human (e. g., through diet,
drinking water, or ambient air). The effective dose is the amount that
actually reaches a target organ or tissue.

estimate target tissue doses and what assumptions were used in that
modeling. In the methylene chloride standard, OSHA developed a set of 11
criteria to judge whether available data are adequate to permit the agency
to rely on PBPK analysis in place of administered exposure levels when
estimating human equivalent doses. Although it is beyond the scope of this
appendix to provide a full technical explanation of the following criteria,
they do illustrate the complex nature of PBPK analysis and, more generally,
the types of issues that risk assessors consider in weighing the available
data.

1. The predominant as well as all relevant minor metabolic pathways must be
well described in several species, including humans.

2. The metabolism must be adequately modeled. 3. There must be strong
empirical support for the putative mechanism of

carcinogenesis. 4. The kinetics for the putative carcinogenic metabolic
pathway must

have been measured in test animals in vivo and in vitro and in corresponding
human tissues at least in vitro. 13 5. The putative carcinogenic metabolic
pathway must contain metabolites

that are plausible proximate carcinogens. 6. The contribution to
carcinogenesis via other pathways must be

adequately modeled or ruled out as a factor. 7. The dose surrogate in target
tissues used in PBPK modeling must correlate with tumor responses
experienced by test animals.

8. All biochemical parameters specific to the compound, such as blood: air
partition coefficients, must have been experimentally and reproducibly
measured. This must especially be true for those parameters to which the
PBPK model is sensitive.

13 The term in vivo refers to tests carried out within living organisms,
while in vitro refers to tests outside the organism (e. g., using cells
taken from an animal or human).

9. The model must adequately describe experimentally measured physiological
and biochemical phenomena.

10. The PBPK models must have been validated with other data (including
human data) that were not used to construct the models. 11. There must be
sufficient data, especially data from a broadly

representative sample of humans, to assess uncertainty and variability in
the PBPK modeling. In the 1,3- butadiene standard, which came out after the
methylene chloride standard, OSHA used these same 11 criteria to judge the
adequacy of the 1,3- butadiene PBPK models for dose- response assessment. In
the butadiene case, the PBPK models did not meet all of these criteria.

For dose- response analyses from human cancer data, OSHA tends to use
similar methodologies to the other regulatory agencies. Mostly these are
simple linear models, such as relative risk models, and estimates of risk
are based on the MLE.

Noncancer Effects No specific approach or procedure for the assessment of
noncancer effects was evident in the examples of OSHA rulemakings we
reviewed. However, OSHA clearly considered a range of noncancer toxic
effects in its analyses.

In its rulemakings, OSHA focused on describing and analyzing a variety of
relevant studies, case reports, and other information found in the
scientific literature. Rhomberg noted that, in the past, OSHA used methods
that were

comparable to those of other agencies. However, the federal court in the

AFL- CIO v. OSHA case questioned the use of standard safety factors, noting
that ?application of such factors without explaining the method by which
they were determined is clearly not permitted.? 14 OSHA has produced
quantitative risk estimates for reproductive and

developmental effects (glycol ethers, 1993), heart disease and asthma
(environmental tobacco smoke, 1994), Hepatitis B virus infection (bloodborne
pathogens, 1992), tuberculosis, and kidney toxicity from cadmium exposure.
OSHA is currently working on quantitative risk assessments for such adverse
health effects as cardiovascular disease

mortality, neural effects, asthma, and respiratory tract irritation for a
number of substances. OSHA states that new methodology has been used 14 965
F. 2d at 978.

for these assessments, but review drafts were not yet ready and we cannot
comment further. Exposure Assessment Under the OSH Act, OSHA has a
relatively specific and narrow focus on exposure assessment. OSHA?s primary
focus is estimating the risk to workers exposed to an agent for a working
lifetime. This risk is calculated

in terms of a person exposed at a constant daily exposure level for 45 years
at 5 days per workweek and 8 hours per workday. The goal is to set
standards, in the form of permissible exposure limits (PELs), so workers
would suffer no impairment during the course of their lifetime under a
continuous exposure scenario. Although this is a hypothetical exposure
scenario, Rhomberg observed that it is not conservative compared with the
actual distribution of exposures in the workplace. He also noted that, in

assessing the exposures and risks associated with the new proposed standard,
OSHA assumes that the standard is applied to newly exposed workers who will
work under the new standard for their entire working lives. No allowance is
made for the fact that current workers may already have had exposures higher
than the new standard.

Despite the primary focus on long- term working lifetime exposures, there
may also be some risks posed by acute, short- term exposures. Therefore,
although part of OSHA?s risk assessment could focus on longer- term risks
and deal with 8- hour time- weighted average (TWA) exposure, the agency?s
analysis may also cover short- term exposure effects. In the methylene
chloride rule, for example, OSHA set the 8- hour TWA PEL primarily to reduce
the risk of employees developing cancer, while the 15- minute shortterm
exposure limit (STEL) was primarily designed to protect against noncancer
risks, such as negative effects on the central nervous system.

Finally, Rhomberg pointed out the following distinct features of
occupational exposure assessments: Compared to environmental exposures,
exposures in the workplace tend to be much better defined. The workplace is
a confined setting within which practices and behaviors tend to be
standardized. Exposure levels are often high enough to be easily measured,
and many workplaces have ongoing monitoring of environmental levels of
compounds. 15

15 Rhomberg (1996), p. 36.

Risk Assessment As previously noted, OSHA?s risk assessment procedures,
including its

Assumptions and default assumptions and methodological preferences, tend to
be established through the precedents of prior rulemakings. In contrast to
Methodological EPA and FDA, OSHA also appears to choose somewhat less
conservative

Choices options, even though the agency notes that Congress and the courts
have

permitted and even encouraged it to consider ?conservative? responses to
both uncertainty and human variability. The Supreme Court?s Benzene
decision, in particular, affirmed that ?the Agency is free to use
conservative assumptions in interpreting the data with respect to
carcinogens, risking error on the side of over- protection rather than under
protection.? 16 On the other hand, OSHA explicitly stated in rulemakings
that it takes various

steps to be confident that its risk assessment methodology is not designed
to be overly conservative (in the sense of erring on the side of
overprotection). Although not intended to be comprehensive, table 8

illustrates some of the specific assumptions or methodological choices used
by OSHA. It also illustrates the overt balancing of more and less
conservative choices that characterizes OSHA?s approach to risk assessment.
The information presented in table 8 was taken primarily from OSHA risk
assessment documents but also reflects additional comments provided by OSHA
officials. (GAO notes and comments appear in parentheses.)

Table 8: OSHA Risk Assessment Assumptions and Methodological Choices When
the assumption/ choice would be applied (step in the risk assessment process
or Likely effect on risk Assumption or

circumstances) assessment results methodological choice Reason( s) for
selection

1. Avoids the uncertainty of crossspecies Choice of data set for

(Not identified.) Most things being equal, extrapolation. Also,

quantitative cancer risk epidemiologic data are preferred most human studies
on nondrug assessment (hazard to data from animal studies

chemicals come from identification and dose- response whenever good data on
human occupational exposures. assessment) risks exist.

16 448 U. S. at 656.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the risk assessment
process or Likely effect on risk Assumption or circumstances) assessment
results methodological choice Reason( s) for selection

2. Virtually all of the toxic

Choice of data set for qualitative (Not specifically identified, but It is
reasonable to suspect that substances that have been

and quantitative cancer risk OSHA did note that it is possible substances
that cause cancer in demonstrated to be carcinogenic assessment (hazard

that a substance may be multiple animal species and at in humans are also
carcinogenic identification and dose- response carcinogenic in a laboratory
multiple target organ sites would

in laboratory animals. assessment) in the absence of species but not in
humans.

be carcinogenic to humans. sufficiently powerful negative OSHA officials
also pointed out

Therefore, OSHA relies on wellconducted, epidemiological studies or that, as
part of its risk high- quality animal mechanistic studies assessment, OSHA
examines all

bioassays as the primary basis demonstrating that the purported

relevant toxicity data to for cancer hazard identification carcinogenic
mechanism of determine the appropriateness and often for quantitative risk
action of the substance is of relying on extrapolation from assessment.

irrelevant to humans. animal studies.)

3. To protect workers from Analysis of epidemiological and

(Not identified.) If human (epidemiological) data

significant risk. animal data for quantitative are equivocal, or the cancer
risk assessment (hazard epidemiologic study is not identification and dose-
response sufficiently sensitive to identify assessment) when animal an
increased risk predicted by a studies indicate a positive well- conducted
animal bioassay, response to a particular it is necessary to consider the

chemical and epidemiological animal data to protect workers

studies of exposures to the same from significant risk. chemical fail to
exhibit a statistically significant increase

in risk. 4.

(Not identified.) Choice of animal- to- human dose (Not identified, but this
is In the absence of equivalence for quantitative risk generally considered
to be a pharmacokinetic information

assessment (dose- response conservative approach.)

satisfying OSHA?s criteria for assessment). acceptance of PBPK models, OSHA
relies on a default mg/ kg/ day species conversion factor. 5.

(Not explicitly identified. Per Choice of data set for

OSHA cited this as an instance OSHA uses site- specific tumor comments from
OSHA it reflects,

quantitative cancer risk where the agency does not use incidence, rather
than pooled

in part, a policy choice to be assessment (dose- response the most
conservative approach.

tumor response, in determining conservative, but not overly assessment).

the dose- response function for a conservative.)

chemical agent. OSHA estimates excess risks to humans based on the most
sensitive species- sex- tumor site.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the risk assessment
process or Likely effect on risk Assumption or circumstances) assessment
results methodological choice Reason( s) for selection

6. Evidence suggests that such Choice of data set for

(Not specifically identified in the OSHA combines benign tumors

tumors should be interpreted as quantitative cancer risk risk assessments we
reviewed, with the potential to progress to

representing a potentially assessment (dose- response

but according to OSHA officials malignancies with malignant

carcinogenic response. assessment).

is almost always conservative.) tumors occurring in the same tissue and the
same organ site.

(In support of this position, OSHA cited the views of the Office of Science
and Technology Policy on chemical carcinogenesis [citation provided]. OSHA
also pointed out that other federal agencies-

EPA, FDA, the Consumer Product Safety Commission, and the National Institute
for Occupational Safety and Health- have also included benign responses in
their

assessments.) 7.

(Not explicitly identified, but the Dose- response assessment. (Not
identified.)

OSHA relies on low- dose assumption that you can extrapolation to estimate
risks

extrapolate low- dose effects associated with exposure levels

from high- dose effects is a of interest; however, because standard
assumption of risk occupational exposures are assessment.) typically much
higher than those encountered in the general environment, OSHA?s risk
assessments do not extrapolate

as far beyond the range of observed toxicity as might be necessary to
characterize environmental risks.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the risk assessment
process or Likely effect on risk Assumption or circumstances) assessment
results methodological choice Reason( s) for selection

8. OSHA stated that it believes that

Cancer dose- response The multistage model is For low- dose animal- to-
human the multistage model conforms assessment. generally considered to be a
cancer risk extrapolation, most closely to what is known

conservative model because it is OSHA?s preference is to use the about the
etiology of cancer,

approximately linear at low maximum likelihood estimate including the fact
that linear- atlow- doses and because it assumes (MLE) in the Crump- Howe
dose behavior is expected no threshold for carcinogenesis,

reparameterization of the for exogenous agents, which

although there are other

?multistage model.? increase the risk of cancer

plausible models of already posed by similar carcinogenesis which are more
This model is based on the "background? processes. OSHA conservative at low
doses. biological assumption that noted that there is no evidence
carcinogens induce cancer that the multistage model is (OSHA officials also
pointed out through a series of independent

biologically inappropriate, that the algorithm that OSHA viable mutations in
a series of especially for genotoxic uses to compute MLE estimates stages,
and that each mutation carcinogens, and that the

is less conservative because it rate is linearly related to dose.

overwhelming scientific may lead to models that are consensus is that
genotoxins sublinear at low doses.) The multistage model used by follow low-
dose linear functions. the agency also assumes no However, OSHA officials
also threshold for carcinogenesis.

pointed out that the CrumpHowe algorithm that OSHA uses can yield nonlinear
models.

OSHA?s preference is consistent with the position of the Office of Science
and Technology Policy, which recommended that ?when data and information are
limited, models or procedures that

incorporate low- dose linearity are preferred when compatible with limited
information? [citation provided].

9. In part, this appears to reflect a Cancer dose- response OSHA cited this
as an instance OSHA?s default choice is to policy choice. OSHA cited this
assessment. where the agency does not use

select the MLE of the choice as one of the steps it has a conservative (or
the most parameterized dose- response taken that make it fairly confident

conservative) approach. function, rather than the upper

its risk assessment methodology 95- percent confidence limit. is not
designed to be overly ?conservative,? in the sense of erring on the side of

overprotection.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the risk assessment
process or Likely effect on risk Assumption or circumstances) assessment
results methodological choice Reason( s) for selection

10. In its risk assessments, OSHA During dose- response

OSHA notes that the body For interspecies dose scaling, points out that
there are several assessment, when estimating weight extrapolation approach
OSHA assumes that equivalent plausible options for

the equivalent human dose that it generally uses tends to be doses in mg/
kg/ day (body weight extrapolating human risks from based upon an
experimental significantly less conservative scaling) would lead to
equivalent animal data via interspecies

dose in animals. than other plausible risks.

scaling factors [citations methodologies and most likely is provided]. OSHA
states that its less conservative even than the (OSHA?s Director of Health
selection of body weight scaling central tendency of the plausible Standards
noted that the agency

is one of the steps it takes that values. might also move to consideration
make the agency fairly confident of ï¿½- power scaling, as agreed to that its
risk assessment The agency also notes that,

by EPA, FDA, and the Consumer methodology is not

across the series of plausible Product Safety Commission, or

?conservative? in the sense of values, its body weight

to develop a probability erring on the side of

extrapolation approach is distribution for the power.) overprotection.

generally considered the least conservative, (body weight) 2/ 3 In addition,
to convert mg/ kg/ day (No particular basis cited for [surface area scaling]
the most doses to parts per million (ppm), using the specific breathing rate
conservative, and (body OSHA uses a human breathing and body weight figures,
just that weight) 3/ 4 the midpoint value.

rate of 9.6 m 3 /workday and they are OSHA?s preferred

human body weight of 70 kg. values.)

11. The focus on working lifetime Exposure assessment. OSHA notes that this
reflects a

OSHA assumes that workers will exposure comes from the ?more conservative?
choice. be exposed to a chemical at the statutory mandate under the maximum
permissible level for

OSH Act to protect an employee 45 years.

?even if such employee has regular exposure to the hazard The standard
values used for

for the period of his working life.? assessing exposures over a working
lifetime are:

The choice of 45 years is based a. 45 years per working

on a worker beginning work at lifetime, age 20 and retiring at age 65. b. 5
workdays per week, and c. 8 hours per workday.

(Continued From Previous Page)

When the assumption/ choice would be applied (step in the risk assessment
process or Likely effect on risk Assumption or circumstances) assessment
results methodological choice Reason( s) for selection

12. The general boundary is directly Policy for evaluating ?significant (No
direct effect on the risk The general boundary within attributed to the
Supreme

risk.? estimates, but this general policy which acceptable versus Court?s
1980 Benzene decision.

does serve as an underlying unacceptable risk falls is focus in conducting
risk between an insignificant fatality assessments.)

risk of one in one billion (10 -9 ) and a significant risk of 1 in one
thousand (10 -3 ).

More explicitly, OSHA stated in one of its rulemakings that risks at or
above 10 -3 (1 per 1000) are always significant by any empirical, legal, or
economic

argument available. Source: Compiled from GAO review of OSHA risk assessment
documents and from additional comments provided by agency officials.

Risk Characterization Although OSHA does not have written risk
characterization policies, recent OSHA rulemakings showed that the agency
emphasized (1) comprehensive characterizations of risk assessment results;
(2) discussions of assumptions, limitations, and uncertainties; and (3)
disclosure of the data and analytic methodologies on which the agency
relied. Rhomberg noted

that OSHA?s usual practice is to present the results and methodological
bases of outside parties? risk assessments for a chemical in addition to
OSHA?s own assessment, and to feature several possible bases for risk
calculation in its characterization of risks. In checking examples of recent
OSHA rulemakings, we also observed this emphasis on showing a range of
alternative assessments, both those of external parties and OSHA?s own
sensitivity analyses.

At least three factors help to explain this proclivity to characterize risks
using different data sets, assumptions, and analytical approaches, all of
which are rooted in the statutory context for OSHA standards setting. First,
the agency?s statutory mandate, reinforced by the Supreme Court?s Benzene
decision, is that it must demonstrate ?significant? risk from workplace
exposure to a chemical with ?substantial evidence.? Second, the OSH Act
directs OSHA to base health standards on the ?best available

evidence? and consider the ?latest scientific data.? The third factor is
that the standard selected will be limited by consideration of its
technological and economic feasibility and cost effectiveness. Together,
these provisions

provide ample incentive to show that a compound presents a significant risk
even when using a range of alternative estimates and scientific evidence.
(This does not preclude the agency from focusing on one analysis as the most
appropriate to support its final estimate of risk at a particular level of
exposure.) The bottom line is that OSHA uses risk assessment to justify a
standard by showing, in general, that significant risks exist and that
reducing exposure as proposed in the agency?s standard will reduce those
risks.

In recent OSHA rulemakings, the agency devoted considerable effort to
addressing uncertainty and variability in its risk estimates. Such efforts
included performing sensitivity analyses, providing the results produced by

alternative analyses and assumptions, and using techniques such as Monte
Carlo and Bayesian statistical analyses. In its risk characterizations, OSHA
provided both estimates of central tendency (such as the mean) and upper
limits (such as the 95 th percentile of a distribution). In the methylene

chloride rule, OSHA noted that, in its past rulemakings, it had frequently
estimated carcinogenic potencies via the MLE of the multistage model
parameters. However, in this particular rule it chose for its final risk
estimate to couple one measure of central tendency (the MLE of the
doseresponse parameters) with a somewhat conservative measure of its PBPK
output (the 95 th percentile of the distribution of human internal dose).
OSHA concluded that this combination represented ?a reasonable attempt

to account for uncertainty and variability.?

Chemical Risk Assessment at the Department of Transportation?s Research and
Special

Appendi x V

Programs Administration The chemical risk assessments conducted by the
Department of Transportation?s (DOT) Research and Special Programs
Administration (RSPA) focus primarily on acute (short- term) risks
associated with potential accidents involving unintentional releases of
hazardous materials (HAZMAT) during transportation. 1 As such, they are very
different from risk assessments that focus on chronic health risks.
According to agency officials, RSPA?s assessments are done using a flexible,
criteria- based system. RSPA?s HAZMAT transportation safety program begins
with a hazard analysis that results in material classification. There are
international standards on the transportation and labeling of dangerous

goods that classify the type of hazard associated with a given substance (e.
g., whether it is flammable, explosive, or toxic) and the appropriate type
of packaging. Once a hazard is classified, RSPA?s analysis focuses on
identifying the potential circumstances, probability, and consequences of

unintentional releases of hazardous material during its transportation. DOT
has written principles on how the results of its risk or safety assessments
should be presented. Those principles emphasize

transparency regarding the methods, data, and assumptions used for risk
assessments and encourage DOT personnel to not only characterize the range
and distribution of risk estimates, but also to put the risk estimates into
a context understandable by the general public.

Context for RSPA According to DOT officials, chemical risks may be an
element of almost Chemical Risk any departmental risk assessment. For
example, they said that one of the

alternatives they explored regarding air bags involved potential exposure
Assessment

to chemicals used in the inflation mechanism. They also noted that Federal
Aviation Administration (FAA) safety analyses include some elements related
to potential exposures to the chemicals that are always found in aircraft
mechanisms. However, DOT?s risk assessment most commonly focus on chemical
risks when considering the transportation of hazardous materials.
Unintentional releases of hazardous materials during transportation, whether
due to packaging leaks or transportation accidents, may pose risks to human
health and safety, the environment, and property. The potential consequences
of such incidents include deaths or 1 DOT uses the term ?risk assessment?
narrowly to refer to risk characterization, specifically the determination
of risk context and acceptability, often by comparison to other similar
risks. However, to be consistent with the rest of this report, we are using
the term in this appendix to refer to the entire process of identifying,
analyzing, and characterizing risks.

injuries caused by an explosion, fire, or release of gases that are toxic
when inhaled. Under the Federal Hazardous Materials Transportation Act, as
amended, the Secretary of Transportation has the regulatory authority to
provide adequate protection against risks to life and property inherent in
transporting hazardous materials in commerce. 2 DOT officials pointed out
that, because this act tends to be more general than those relevant to other

agencies? regulation of risks from chemicals, it gives DOT more flexibility
to define what is ?adequate? to address potential risks. The statute directs
the DOT Secretary to designate a material or group or class of materials as
hazardous when he or she decides that transporting the material in commerce
in a particular amount and form may pose an unreasonable risk to health and
safety or property. The Secretary is also directed to issue regulations for
the safe transportation of such materials. The hazardous materials
regulations apply to interstate, intrastate, and foreign transportation in
commerce by aircraft, railcars, vessels (except most bulk

carriage), and motor vehicles. The Secretary has delegated authority for
implementing these hazardous materials responsibilities to various
components within DOT. In particular, RSPA issues the Hazardous Materials
Regulations and carries out related regulatory functions, such as issuing,
renewing, modifying, and terminating

exemptions from the regulations. The Superfund Amendments and
Reauthorization Act of 1986 mandated that RSPA also list and regulate under
the Hazardous Materials Regulations all hazardous substances designated by
EPA. According to DOT officials, RSPA conducts most of the department?s risk
assessments regarding the transportation of chemical hazardous materials.
RSPA and the modal administrations in DOT- FAA,

the United States Coast Guard, the Federal Motor Carrier Safety
Administration, and the Federal Railroad Administration- share enforcement
authority for hazardous materials transportation. RSPA?s Office of Hazardous
Materials Safety (OHMS) has the primary responsibility for managing the
risks of hazardous materials transportation

within the boundaries of the United States, unless such materials are being
transported via bulk marine mode (in which case the Coast Guard is
responsible). Overall, OHMS notes that its Hazardous Materials Safety
Program and resulting regulations (1) are risk based; (2) use data,

2 49 U. S. C. 5101 et seq.

information, and experience to define hazardous materials and manage the
risk hazardous materials present in transportation; and (3) are prevention
oriented. Therefore, the analysis of risk is an important element of OHMS?
responsibilities. Within OHMS, the Office of Hazardous Materials Technology
(OHMT) provides scientific, engineering, radiological, and risk analysis
expertise. Other entities may also be involved in conducting transportation-
related chemical risk and safety assessments. For example, OHMS sponsored a
quantitative threat assessment by the John A. Volpe National Transportation
Systems Center (the Volpe Center), which is operated by RSPA, to determine
the probability that a life- threatening incident would

occur as a result of transporting hazardous materials in aircraft cargo
compartments. 3 OHMS also sponsored a multiyear research effort by the
Argonne National Laboratory to characterize the risk associated with
transportation of selected hazardous materials on a national basis. 4

One of the most distinctive aspects regarding the regulation of hazardous
materials transportation is the role that is played by international
agreements and definitions. Criteria for classifying and labeling dangerous
chemicals being transported have been internationally harmonized through the
United Nations Recommendations on the Transport of Dangerous Goods. 5 This
UN classification system is internationally recognized, and

RSPA has essentially adopted the UN recommendations into the domestic
hazardous materials regulations. (A more detailed description of this
classification system appears in the following section.) 3 Threat Assessment
of Hazardous Materials Transportation in Aircraft Cargo Compartments, U. S.
Department of Transportation, Research and Special Programs Administration,
John A. Volpe National Transportation Systems Center, Final Report (December
1999). 4 A National Risk Assessment for Selected Hazardous Materials
Transportation, Decision and Information Sciences Division, Argonne National
Laboratory (December 2000). The Argonne National Laboratory is operated by
the University of Chicago under contract for the U. S. Department of Energy.
5 The United States and other countries are attempting to develop a Globally
Harmonized System (GHS) for the classification and labeling of chemicals.
However, while criteria have been internationally harmonized for purposes of
transportation, harmonized requirements have not yet been established for
purposes of environmental, worker, or consumer safety regulations. RSPA
participates, along with other federal agencies, in these GHS activities.

Risk Assessment Because of the particular regulatory context in which it
operates- in

Procedures particular, its focus on acute (short- term) risks associated
with transportation accidents- RSPA does not follow the four- step risk

assessment paradigm identified by NAS and used by EPA, FDA, and OSHA.
However, RSPA?s procedures do address similar generic questions, such as
whether a particular material or activity poses a threat and the likelihood

and consequences of potential accidents. The agency uses a criteria- based
system to assess the hazards to human health and safety, property, and the
environment that are associated with potential accidents during hazardous
materials transportation. Chemicals are identified and classified as hazards
according to a classification system in the Hazardous Materials

Regulations that is largely harmonized with internationally recognized
criteria. The risk analyses by RSPA then focus on assessing the potential
circumstances under which exposure could occur during transportation,

their causes, consequences, and probability of occurrence. Guidance The
general risk assessment procedures applicable to RSPA are found within DOT-
wide policies on conducting regulatory analyses and also in descriptive
materials about the agency?s Hazardous Materials Safety

Program. DOT included general guidelines for conducting a risk assessment as
part of its broader Methods for Economic Assessment of Transportation
Industry Regulations (Office of the Assistant Secretary for Policy and
International Affairs, June 1982). The DOT guidelines for risk assessment
are grouped under three major topics:

 procedural guidelines that recommend formats for presentation of risk
analyses, formats for conducting risk analyses, and reporting of assumptions
and limits of analyses;

 methodology guidelines that discuss some of the more frequently used risk
methods and their applicability; and

 data guidelines that discuss data sources, collection and presentation of
data, and raw and derived statistics.

The primary focus of the DOT- wide risk assessment methodology and
guidelines is on estimating the risk reduction attributable to proposed
transportation safety regulations. DOT?s guidelines are intended to be

applicable to risk assessment of hazardous material transport by any mode as
well as assessment of other types of transportation risk. However, DOT
stated that the guidelines are not intended to be a ?cookbook,? or a
prescriptive methodology, specifying each step in a risk assessment. DOT

pointed out in the guidelines that such an approach is not desirable,
because there is no single ?correct? set of methods for assessing
transportation risk.

In addition to the DOT- wide guidelines, RSPA has produced written materials
specifically on the Hazardous Materials Safety Program. These materials
describe the role of risk assessment in the management of risks associated
with transportation of hazardous materials and the general

process used for analysis of risks, and they define risk assessment and
management terms for purposes of hazardous materials safety. 6 There also
are a number of general guidance documents and reports on various aspects of
hazardous materials transportation safety that provide additional insights
into the identification and assessment of risks.

Risk Assessment Approach RSPA does not apply the same NAS four- step
paradigm for risk assessment as generally used by EPA, FDA, and OSHA.
According to RSPA officials, the main reason for this difference between
their risk assessments and most of those conducted by the other three
agencies is the focus of RSPA?s assessments. RSPA?s concerns relative to
hazardous materials transportation are primarily directed at short- term or
acute health risks due to relatively high exposures from the unintentional
release of hazardous materials. The officials said that, in contrast, the
four basic steps of the NAS paradigm were intended to focus on chronic
health risks due to longterm,

low- level background chemical exposure. 7 The main exceptions to this
difference in general risk assessment procedures occur when other agencies?
assessments are similarly directed at risks associated with unintentional
releases of chemicals. In particular, RSPA officials said that there are
parallels between their risk assessment and management efforts and those of
EPA and OSHA programs that are directed at chemical accidents. (See, for
example, the description of the risk assessment procedures for EPA?s
Chemical Emergency Preparedness and Prevention Office in app. II.)

6 DOT?s definitions sometimes vary from the definitions of those terms in
other risk assessment settings. Items in the glossary of this report
identify some of these different definitions. 7 Note, however, that the
general risk assessment procedures of the other three agencies cover both
chronic and acute risks.

In sharp contrast to most of the risk assessment procedures we described for
EPA, FDA, and OSHA, toxicity is simply one of several potentially dangerous
properties of a hazardous material of concern to RSPA. Where

toxicity is a factor, RSPA?s risk assessments tend to center on exposure
levels that pose an immediate health hazard. This focus is reflected in the
types of chemical toxicity information that RSPA helps develop. For example,
RSPA actively participates on a National Advisory Committee developing Acute
Exposure Guideline Levels for chemicals. In specific

cases where chronic toxicity or environmental values play a role in RSPA
analyses, agency officials said that they rely on what EPA, FDA, OSHA, and
other agencies have developed. Despite such differences, RSPA?s risk
assessments address similar basic

issues as the chemical risk assessments of the other three agencies (e. g.,
whether a particular material or activity poses a threat and the severity
and likelihood of potential exposures). The DOT- wide risk assessment
guidelines primarily discuss ?consequence? and ?probability? analyses, but
also describe a preliminary step for defining scenarios of concern

(essentially part of a hazard identification step) and a final step to
summarize results and conclusions from the preceding analyses (essentially a
risk characterization). The Hazardous Materials Safety Program materials
outline a similar risk assessment process that progresses from the
identification of hazards to an evaluation of incident causes, frequencies,
and consequences.

Identifying Hazards RSPA begins with a hazard analysis that results in
material classification. In RSPA risk assessments, hazardous materials are
chemical, radioactive, or infectious substances or articles containing
hazardous materials that can

pose a threat to public safety or the environment during transport.
Hazardous materials pose this threat through chemical, physical, nuclear, or
infectious properties that can make them dangerous to transport workers or
the public. For example, RSPA is concerned with the potential for the
unintentional release of hazardous materials to lead to adverse

outcomes such as explosions, fires, or severely enhanced fires that can
cause deaths, injuries, or property damage. The agency is also concerned
with the potential toxic, corrosive, or infectious effects of released
materials on humans and the environment.

According to DOT officials, their hazard classification approach is a
criteria- based system that provides them considerable flexibility in their
analysis and regulation of potential hazards. They noted that their criteria

are geared more toward the hazard a material may pose in an accident

scenario than toward a chronic health risk. The Director of the Office of
Hazardous Materials Technology characterized this hazard classification
approach as a more open system than used in other agencies (e. g., EPA). He
explained that, in this system, any new chemical or substance that fits
within RSPA?s matrix of hazard criteria falls under the hazardous materials

transportation regulations. Hazard identification for these assessments is
based largely on international agreements regarding transportation of
dangerous goods. Of particular importance, there is an internationally
recognized system for the classification, identification, and ranking of all
types of hazardous

materials that was created by the UN Committee of Experts on the Transport
of Dangerous Goods. This system is revised biennially and published as the
?United Nations Recommendations on the Transport of Dangerous Goods.? Under
this classification system, all hazardous materials are divided into nine
general classes according to physical,

chemical, and nuclear properties. The system also specifies subdivisions and
packing group designations (that indicate a relative level of hazard) for
some classes. (See table 9.)

Table 9: UN Classification System for Transport of Dangerous Goods Packaging
performance Hazard class Description Subdivisions specified requirements
specified

1 Explosives and pyrotechnics X 2 Compressed and liquefied gases X 3
Flammable liquids X 4 Flammable solids (including self- reactive liquids) X
X 5 Oxidizers and peroxides X X 6 Toxic (poisonous) and infectious materials
X X 7 Radioactive materials 8 Corrosive materials (acidic or basic) X 9
Miscellaneous dangerous substances and articles

Packing groups

1 Great danger 2 Medium danger 3 Minor danger

Source: Research and Special Programs Administration, Department of
Transportation.

These are broad categories that may include large numbers of diverse
materials. For example, the air cargo threat assessment noted that there
were 535 different flammable liquid entries in the hazardous materials table
and more than 700 toxic material entries. Because there are hazardous
materials with multiple dangerous properties, these classes and

subdivisions are not mutually exclusive. Compressed or liquefied gases, for
example, also may be toxic or flammable. The UN Committee of Experts created
more than 3, 400 possible identification numbers, proper

shipping descriptions, and hazard classes to be assigned to various
hazardous material compounds, mixtures, solutions, and devices. There are
also generic ?not otherwise specified? identification numbers and

shipping descriptions that allow the material to be classed by its defined
properties.

RSPA uses essentially the same framework as the UN recommendations for the
hazard classes and packing requirements of its Hazardous Materials
Regulations. Table 10 shows the hazard classification system in the
regulations.

Table 10: Hazard Classification System of RSPA?s Hazardous Materials
Regulations Class/ division number Description

None Forbidden materials a None Forbidden explosives b Class 1

Explosives

 Division 1.1

 Explosives with a mass explosion hazard

 Division 1.2

 Explosives with a projection hazard

 Division 1.3

 Explosives with predominantly a fire hazard

 Division 1.4

 Explosives with no significant blast hazard

 Division 1.5

 Very insensitive explosives; blasting agents

 Division 1.6

 Extremely insensitive detonating articles Class 2

Gases

 Division 2.1

 Flammable gas

 Division 2.2

 Nonflammable compressed gas

 Division 2.3

 Poisonous (toxic by inhalation) gas Class 3 Flammable and combustible
liquid Class 4

Flammable solids; spontaneously combustible materials; and dangerous when
wet materials

 Division 4.1

 Flammable solid

 Division 4.2

 Spontaneously combustible material

 Division 4.3

 Dangerous when wet material

(Continued From Previous Page)

Class/ division number Description

Class 5 Oxidizers and organic peroxides

 Division 5.1

 Oxidizer

 Division 5.2

 Organic peroxide Class 6

Poisonous (toxic) materials and infectious substances

 Division 6.1

 Poisonous (toxic) materials

 Division 6.2

 Infectious substance (etiologic agent) Class 7 Radioactive material Class
8 Corrosive material Class 9 Miscellaneous hazardous material None Other
regulated material a 49 CFR Section 173.21 defines the materials that shall
not be offered for transportation or

transported. b 49 CFR Section 173.54 defines the explosives that shall not
be offered for transportation or

transported. Source: 49 CFR section 173.2.

The classification system in these regulations can be very detailed for some
subjects. For example, the regulations specifically identify the types of
toxicity tests and data that should be used to determine whether something

would be classified as poisonous material (class 6, division 6. 1). The
regulations define poisonous material as a material, other than a gas, which
is known to be so toxic to humans as to afford a hazard to health during
transportation, or which, in the absence of adequate data on human toxicity,

 is presumed to be toxic to humans because it falls within one of several
specified categories for oral, dermal, or inhalation toxicity when tested on
laboratory animals; or  is an irritating material, with properties similar
to tear gas, which causes extreme irritation, especially in confined spaces.

Of particular relevance to comparisons with chemical risk assessments of
other agencies, the regulations contain precise definitions of what
constitutes oral, dermal, or inhalation toxicity for purposes of the
Hazardous Materials Regulations. For example, one threshold for

inhalation toxicity is defined as a dust or mist with an LC 50 for acute
toxicity on inhalation of not more than 10 mg per liter of air. 8 (A
different definition applies to the inhalation toxicity of a vapor.) The
regulations also address other testing requirements and conversion factors.
The regulations state that, whenever possible, animal test data that have
been reported in the chemical literature should be used.

The Hazardous Materials Regulations include an extensive Hazardous Material
Table with itemized information about specific hazardous materials. The
number of HAZMAT table entries corresponds closely with the number created
by the UN. RSPA officials noted that the number of specific chemicals
covered by the regulations is many multiples of the

more than 3,400 entries, though, because of the generic nature of the ?not

otherwise specified? descriptions. The table includes, but is not limited
to, information such as the material?s description, hazard class or
division, identification number, packing group, label codes, limits to the
quantity of the material permitted in a single package, and special
provisions

concerning its transportation. Allyl chloride, for example, is identified as
a class 3 material (flammable and combustible liquid), is in packing group I
(indicating great danger), is forbidden on passenger aircraft and rail, and

has two special provisions regarding the tanks used for transporting this
substance. A material that meets the definition of more than one hazard
class or division, but is not specifically listed in the table, is to be
classed according to the highest applicable hazard class or division
according to a descending order of hazard. For example, the division of
poisonous gases is ranked as a greater hazard than the division of flammable
gases.

According to OHMS, the process of classifying a material in accordance with
these hazard classes and packing groups is itself a form of hazard analysis.
Another important feature of this process is that the regulations require
the shipper to communicate the material?s hazards through the use of the
hazard class, packing group, and proper shipping name on the shipping paper
and the use of labels on packages and placards on the

8 In this case, LC for acute toxicity on inhalation means that concentration
of dust or mist 50 which, administered by continuous inhalation for 1 hour
to both male and female young adult albino rats, causes death within 14 days
in half of the animals tested.

transport vehicle. Therefore, the shipping paper, labels, and placards
communicate the most significant findings of the shipper?s hazard analysis
to other parties. This communication aspect is particularly important in

emergency response situations if an accident occurs during transport of
these materials.

The classification system, by itself, is not sufficient for all risk
assessment purposes. For example, RSPA and OHMS still need to identify
potential scenarios in which transportation accidents, spills, and leaks
could occur. As evidenced by the air cargo threat assessment, such scenarios
include the possibility that hazardous materials might be transported in a
manner not in compliance with current regulations. 9 Also, as emphasized in
a November 2000 report for RSPA, the hazardous materials transport system is
highly heterogeneous and complex. 10 The report pointed out that this system
involves not only many different materials posing a variety of

hazards (as reflected in the classification system outlined in table 9) but
also:

 a chain of events involving multiple players having different roles in the
process of moving hazardous materials (such as shippers, carriers, packaging
manufacturers, freight forwarders, and receivers of shipments) and the
possibility of multiple handoffs of a material from one party to another
during transport;

 several different modes of transport (principally highway, rail, waterway,
and air), with some shipments that switch from one mode to another during
transit; and

 multiple possible routes of transit. 9 The threat assessment project was,
in fact, initiated following the crash of ValuJet Flight 592 near Miami, FL
in May 1996, which had been linked to hazardous material devices (chemical
oxygen generators) shipped in violation of DOT regulations.

10 ?Risk Management Framework for Hazardous Materials Transportation,?
prepared by ICF Consulting for the U. S. Department of Transportation,
Research and Special Programs Administration (Nov. 1, 2000). RSPA officials
noted that this framework provides a structure for their efforts and serves
as a tool for all parties involved in hazardous materials

transportation to consider in fostering continuous safety improvement.

All of these complex features might need to be considered in identifying
hazard scenarios. However, in identifying (and analyzing) potential hazard
scenarios, RSPA and OHMS benefit from being able to use data, information,
and experience on hazardous materials transportation

incidents. For example, risk assessors can review data from sources such as
the DOT Hazardous Materials Information System (HMIS) that catalogues
transportation- related incidents that involve a release of

hazardous materials. An OHMS official pointed out that the agency also uses
fairly sophisticated models in analyzing various scenarios. He said that
such models were used, for example, to provide a scientific basis for

determining evacuation zones when developing the 2000 Emergency Response
Guidebook. 11 Analyzing the Consequences and In contrast to the other
agencies covered in this report, determining the

Probabilities of Risks toxicity of a particular chemical (dose- response
assessment) is not a

central focus of risk assessment in RSPA and OHMS. Toxicity is only one of
many risk factors under consideration (and should already be addressed
through the hazard classification system). Instead, the primary focus of
analysis is on the potential for hazardous materials to (1) spill or leak
while in transit or (2) cause, contribute to, or multiply the consequences
of a

transportation- related accident. Analysis regarding the first item is
primarily concerned with the packaging and containers used for
transportation of hazardous materials, while analysis of the second item
also considers other elements, such as the modes and routes of
transportation for these materials. As the DOT risk assessment guidelines
state, ?Hazardous materials accidents generally are transportation accidents
in which hazardous materials happen to be present.?

DOT documents use a variety of terms to describe and refer to the analysis
of hazards or risks of concern to the department and its component offices
(e. g., hazard analysis, risk analysis, threat assessment). However, the
core of the analysis remains the same- an evaluation of the causes, 11 The
2000 Emergency Response Guidebook was developed jointly by DOT, Transport

Canada, and the Secretariat of Communications and Transportation of Mexico.
It is intended primarily to guide firefighters, police, and other emergency
personnel who may be the first to arrive at the scene of a transportation
incident involving a hazardous material in (1) quickly identifying the
specific or generic classification of the material( s) involved in the
incident, and (2) protecting themselves and the general public during the
initial response phase of the incident. According to RSPA, the 2000 revision
of this guidebook was based on risk principles and analyses, and the
technical basis and derivation of the values in the guidebook?s Table of
Initial Isolation and Protective Action Distances are available on the RSPA
website.

consequences, and likelihood of transportation incidents involving hazardous
materials. The general model in DOT?s guidelines for risk assessment of
transportation activities or operations partitions the analysis of risk into
two main parts:

 prediction of possible consequences in terms of loss from accidents (or,
more broadly, incidents) while transporting materials in a specified way;
and

 estimation of the probabilities or frequencies of occurrence of the
consequences of such accidents (e. g., the likelihood or expected number of
accidents occurring that would result in the above loss).

For purposes of estimating the risk reduction attributable to transportation
safety regulations, the expected loss or ?risk? is computed by summing the
products of each possible loss multiplied by its probability. (In other

words, risk in this context is the probability- weighted average loss.)
According to DOT definitions, consequence analysis is the evaluation of the
severity and magnitude of impacts associated with the occurrence of
postulated accident scenarios. For purposes of analysis, the DOT

guidelines recommend partitioning this evaluation into three segments: (1)
initiating events (i. e., causes of an accident that can result in loss),
(2) effects (i. e., the possible mechanisms by which an initiating event
might result in injury or damage); and (3) consequences (i. e., the loss of
life,

injuries, property damage, or other losses expected from the effects). The
evaluation of consequences reflects many factors, including the
characteristics of the agent involved, the type of packaging or container

used, the amount of material being transported, and the particular modes and
routes of transportation (which also affect the extent of potential exposure
by the public and environment). DOT defines probability analysis as the
evaluation of the likelihood of individual accident scenarios and outcomes
of adverse events. The likelihood of a particular hazard might be expressed
either as a frequency or probability.

The analyses of consequences and probabilities are based on a variety of
data sources, including, to the extent possible, ?experience? data. Among
the sources of information identified in OHMS materials to address
consequences and probabilities are:

 data from the Hazardous Materials Information System (HMIS);

 commodity flow surveys;

 chemical substance manufacturing, use, and transportation studies;

 special analyses (such as the National Transportation Risk Analysis and
Air Cargo reports mentioned earlier in this appendix, as well as shipment
counts); and

 public comments on rulemakings. Such sources can provide valuable
information for risk assessment in general and the statistical analysis of
hazardous material transportation incidents in particular. The HMIS database
provides a good illustration of the types of baseline data available. This
database provides incident counts according to time, transportation phase
(i. e., en route from origin to destination, loading or unloading, and
temporary storage), and

transportation mode (e. g., air, highway, and rail). For each incident, the
database includes information on the hazardous materials involved, including
the name of the chemical shipped, container type and capacity, number of
containers shipped, number of containers that fail, and the

amount of material released. The database also contains information
concerning the occurrence of fire, explosion, water immersion, environmental
damage, and the numbers of deaths, major and minor injuries, and persons
evacuated.

However, because DOT?s risk assessments are often used to estimate the

?risk impact? of proposed regulations, the DOT guidelines caution that lack
of directly applicable experience data for assessing the impacts is probably
the rule rather than the exception. This is because the controls provided by

the proposed regulations constitute changes from present conditions, and
experience data, by definition, relate to present conditions. The guidelines
also emphasize that, to evaluate the impact on risk of a proposed regulation

or its alternatives, it is necessary to perform a ?with and without? type of
assessment, considering the potential effects on any or all of the elements
of the risk model.

As was the case with the classification of hazardous materials and
packaging, the agency may employ criteria- based classifications of the
consequences of potential adverse events and their likelihood of occurrence.
A 1995 guidance document illustrates how consequence and

frequency categories were combined into a ?risk assessment matrix? to assist
decision makers in their risk management decisions. 12 (See table 11 below.)

Table 11: Example of a Hazardous Materials Risk Assessment Matrix
Consequence of occurrence categories Frequency of occurrence categories
Catastrophic Critical Marginal Minor Negligible

Frequent U U U C- MDR A- MRR Probable U U C- MDR A- MRR A- MRR Occasional U
C- MDR A- MRR A- MRR A Remote C- MDR A- MRR A- MRR A A Improbable A- MRR A-
MRR A A A

Risk index: U - Unacceptable C- MDR - Conditional management decision
required A- MRR - Acceptable management review required A - Acceptable
Source: Adapted from ?Procedure for Removal of Nonconforming Hazardous
Materials Packagings from Service, 7- 13- 95,? Office of Hazardous Materials
Safety.

12 In a nonchemical risk assessment context, DOT?s Federal Transit
Administration (FTA) and the American Public Transit Association recommend
the use of very similar matrices for analysis of rail systems? safety. See
FTA?s State Safety Oversight, Issue 6 (Fall 1999).

As was the case with the three other agencies covered by our review, some of
the chemical risk assessments produced by or for DOT have begun using more
sophisticated methods and models. For example, the Director of OHMT
characterized the National Transportation Risk Assessment study

prepared for OHMS by the Argonne National Laboratory as using state- ofthe-
art risk assessment techniques to characterize risks associated with the
transportation of selected hazardous materials on a national basis. The
consequence assessments in this study employed the Chemical Accident
Statistical Risk Assessment Model that predicts distributions of hazard

zones (i. e., areas in which a threshold chemical concentration is exceeded)
resulting from hazardous material release. 13 That model, in turn, reflected
the input of other physical models on subjects such as hazardous material
release rates of toxic- by- inhalation materials. The Director noted that
his

office believed this study to be the first comprehensive application of
these techniques in this arena for this purpose. Risk Assessment

Although generally very structured and criteria based, RSPA?s risk
Assumptions and assessments for hazardous materials transportation also use
assumptions. DOT- wide guidance documents provide a general framework for
the use of Methodological assumptions. In general, DOT guidance recognizes
that assumptions may Choices

be made when data are lacking or uncertain, or when it is necessary to limit
the scope of an analysis. However, the assumptions, while not empirically
verifiable, are supposed to be reasonable, logically credible, and
supportable in comparison with alternative assumptions. The DOT risk
assessment methodology guidance specifically states that every assessment
should include a list of the major assumptions, conditions, and limitations
of the risk analysis, as well as the reasons why the assumptions

were made. As noted earlier in this appendix, RSPA has access to a number of
sources of directly relevant data and statistics on the transportation of
hazardous materials. However, there are limitations to these systems and
data. For example, the authors of the national transportation risk
assessment for selected hazardous materials cautioned that the information
in DOT?s data systems was not always sufficient or detailed enough to
directly support a

quantitative risk assessment. For example, incidents involving most 13
Rather than specifying a deterministic measure of risk, this model
determines the distribution of possible outcomes, allowing the probability
of a particular consequence to be identified within the limits of the
statistical data used.

hazardous materials (other than gasoline- truck accidents) typically occur
too infrequently to provide statistically reliable data for directly
projecting future risks. In his introduction to the study, the Director of
OHMT also

stated that the quantitative results of this study should be used with
caution. Specifically, he noted, ?While the model of the hazardous materials
transportation system employed in this study is sophisticated, the accuracy
of the data used in the model is often less precise. Estimates, assumptions,
and aggregate numbers have been used in many cases.?

Some of the topics that might require assumptions or choices during a
hazardous materials transportation assessment include:

 the probability of the release of a hazardous material, depending on the
nature of the accident, type of material being transported, and the
containers used;

 the amount of material released in an accident, depending again on factors
such as the severity of the accident, nature of the material, and type of
container, but also depending on assumptions about the size of holes in
containers;

 commodity flows of the materials (e. g., modes of transportation used,
classes of rail tracks, types of highways, routing through urban and rural
areas and related population density);

 the dispersion of released hazardous material, including assumptions about
climate and meteorological conditions and the type of surface that a liquid
might ?pool? on if spilled;

 the probability of a fire or explosion being ignited (both as a
consequence of a release or as a cause of a release); and

 the extent to which humans potentially exposed to released materials would
be sheltered or protected (both within a given mode of transportation, such
as an aircraft, or external to the carrier).

In addition to these topics, RSPA sometimes uses a factor to adjust data in
the HMIS database to address underreporting. However, RSPA officials noted
that, for certain purposes, it might be inappropriate to extrapolate

information in the database. Although assumptions may be needed in RSPA
assessments, RSPA officials said that they do not have default assumptions
for their risk assessments. According to the officials, assumptions must be
developed and described as part of each risk assessment and are specific to
the risk assessment. RSPA officials also noted that they do not use ?safety
factors? in risk assessments, but rather base their assessments on expected
levels or

ranges of performance. Therefore, unlike in the appendices on EPA, FDA, and
OSHA, we have not included a table in this DOT appendix to identify major
default choices, the reasons for their selection, when they would be used in
the process, and their likely effects on risk assessment results. However,
with regard to some of the case- specific assumptions or choices we
identified during our review, we did observe that DOT?s assessments
typically discussed the reasons for particular choices (as with the other
agencies, often citing an interpretation of related research studies). In
some instances, information was also provided on the likely effect (e. g.,
that a particular value represented a conservative estimate or an upper
limit) or level of uncertainty (e. g., that a particular parameter value
might be high by a factor of 3 to 10 times the results from another study)

associated with choices made by the analysts. Risk Characterization DOT has
explicit, written principles regarding how the results of its risk or safety
assessments should be presented. The department?s policies

emphasize the principle of transparency and encourage agency personnel to
not only characterize the range and distribution of risk assessment
estimates, but also to put risk estimates into a context understandable by
the general public. For example, DOT?s ?risk assessment principles? state
that the risk assessment should:

 make available to the public data and analytic methodology on which the
agency relied in order to permit interested entities to replicate and
comment on the agency?s assessment;

 state explicitly the scientific basis for the significant assumptions,
models, and inferences underlying the risk assessment, and explain the
rationale for these judgments and their influence on the risk assessment;

 provide the range and distribution of risks for both the full population
at risk and for highly exposed or sensitive subpopulations, and encompass
all appropriate risks, such as acute and chronic risks, and cancer and
noncancer risks, to health, safety, and the environment;

 place the nature and magnitude of risks being analyzed in context,
including appropriate comparisons with other risks that are regulated by the
agency as well as risks that are familiar to, and routinely encountered by,
the general public, taking into account, for example, public attitudes with
respect to voluntary versus involuntary risks, wellunderstood

versus newly discovered risks, and reversible versus irreversible risks; and

 use peer review where there are issues with respect to which there is
significant scientific dispute to ensure that the highest professional
standards are maintained. The DOT risk assessment guidelines also state that
every risk analysis should present information on (1) quantitative estimates
of risk (over the entire range of plausible values of the developed
variables, and with a ?base case? loss to provide a point of reference); (2)
insights gained from performing the analysis into the factors that most
affect risk assessment

results; and (3) assumptions, conditions, and limitations of the analysis.
With regard to the third item, the guidelines specifically state that
reasons why the assumptions were made, and why the limitations of the
analysis do not significantly impact the risk estimate, should be provided.
The guidelines also suggest two methods for treating uncertainty in a risk
analysis:

 sensitivity analysis (DOT?s preferred method for treating and reporting
the impact of uncertainty), which should be conducted for each scenario in a
risk analysis; and

 bounding analysis involving error propagation (requiring that each model
parameter be expressed as a distribution, or at least a variance, to trace
the implication of uncertainty for the risk estimate).

Glossary Acceptable daily intake

The maximum dose of a hazardous substance that can be consumed daily (ADI)

without causing adverse health effects over a lifetime. Acute effect An
effect that results from a brief exposure or shortly after an acute exposure
(possibly at high levels) of duration measured in minutes, hours,

or days. Additive dose- response

A dose- response model in which the health effects attributable to exposure
model to particular levels of two or more risk agents are equal to the sum
of the responses predicted for each agent alone.

Aggregate risk The risk resulting from the combined exposure of an
individual or defined population to a single agent or stressor via all
relevant routes, pathways, and sources.

Ambient concentration The average amount of a substance in a particular
environmental medium, such as air or water.

Ambient water quality Numeric values limiting the amount of chemicals
present in our nation?s

criteria (AWQC) waters. Developed under section 304( a) of the Clean Water
Act (CWA) and periodically updated, they are determined by assessing the
relationship

between pollutants and their effect on human health and the environment.
These recommended criteria provide guidance for states and tribes in
adopting water quality standards under section 303( c) of the CWA and to
ultimately provide a basis for controlling discharges or releases of
pollutants. Benchmark dose An exposure level that corresponds to a
statistical lower bound on a

standard probability of an effect, such as 10 percent of people affected.
Bioaccumulation factor Bioaccumulation is a process whereby the
concentration of certain (BAF) substances in organisms increases as the
organisms breathe contaminated air, drink contaminated water, or eat
contaminated food. The BAF is the

ratio of a substances' concentration in an organism?s tissue to its
concentration in the medium where the organism lives. BAFs measure a
chemical's potential to accumulate in tissue through exposure to both food
and water.

Bioassay The use of living organisms to measure the effect of a risk agent
or condition- for example, a test for carcinogenicity in laboratory animals
that includes near- lifelong exposure to the agent under test. Used
interchangeably with animal test.

Biologically based dose A model that describes biological processes at the
cellular and molecular

response model level linking the target organ dose to the adverse effect.

Cancer A group of diseases characterized by abnormal, disorderly, and
potentially unlimited new tissue growth.

Carcinogen Any chemical or physical agent possessing the ability to induce
cancer in living organisms. Chronic health effects Diseases occurring as a
result of repeated or persistent exposures. Chronic exposure An exposure
(usually of low concentration) of long duration, e. g., months

or years. Consequence The direct effect of an event, incident, or accident.
It is expressed as a health effect (e. g., death, injury, exposure),
property loss, environmental effect, evacuation, or quantity spilled.

Contaminant( s) Chemicals, microorganisms, or radiation found in air, soil,
water, or food that are not normally constituents of these environmental
media.

Cumulative risk The combined risk from aggregate exposures to multiple
agents or stressors.

Dose The amount of a substance available for interactions with metabolic
processes or biologically significant receptors after crossing the outer
boundary of an organism. A potential dose is the amount ingested, inhaled,
or applied to the skin. An applied dose is the amount presented to an
absorption barrier and available for absorption (although not necessarily
having yet crossed the outer boundary of the organism). An

absorbed dose is the amount crossing a specific absorption barrier (e. g.,
the exchange boundaries of the skin, lung, and digestive tract) through
uptake processes. An internal dose denotes the amount absorbed without
respect to specific absorption barriers or exchange boundaries. A

delivered or biologically effective dose is the amount of the chemical
available for interaction by any particular organ or cell.

Dose- response assessment The determination of the relationship between the
magnitude of administered, applied, or internal dose and a specified
biological response.

Dose- response relationship The relationship between exposure level and the
incidence of adverse effects. Ecological risk assessment A process used to
estimate the likelihood of adverse effects on plants or animals from
exposure to stressors, such as chemicals or the draining of wetlands. The
process includes problem formulation, characterization of exposure,
characterization of ecological effects, and risk characterization.

Effluent Treated or untreated waste material discharged into the
environment. Generally refers to wastes discharged to surface waters.
Emission Pollution discharged into the atmosphere from smokestacks, other
vents,

and surface areas of commercial or industrial facilities; residential
chimneys; and motor vehicle, locomotive, or aircraft exhausts.

Environmental fate The distribution and transformation of a chemical from
its first release until its ultimate removal from or recycling through the
environment.

Epidemiology The study of diseases as they affect populations, including the
distribution of disease, or other health- related states and events in human
population;

the factors (e. g., age, sex, occupation, economic status) that influence
this distribution; and the application of this study to assess and control
health risk. It involves investigating the causes and risk factors of
disease and

injury in populations and the potential to reduce such disease burdens.
Exposure Contact of a chemical, physical, or biological agent with the outer
boundary of an organism. Exposure is quantified as the concentration of the
agent in the medium over time.

Exposure assessment The process of developing a description of the relevant
conditions and characteristics of human and other exposures to risk agents
produced or released by a specified source of risk. This usually involves
the determination or estimation of the magnitude, frequency, duration, and
route of exposure.

Exposure pathway( s) The physical course or means by which risk agents are
transmitted- e. g., the route by which a given population, individual,
species, or setting is exposed to a toxic substance (e. g., via drinking
water, air, or dermal

contact). Exposure route The way an environmental agent enters an organism
(e. g., ingestion,

inhalation, or dermal absorption). Extrapolation Prediction of the value of
a variable outside the range of observation. This involves making inferences
about the unknown by projecting or extending

known information, using models and assumptions. Genotoxic Capable of
causing heritable changes or damage leading to heritable changes in genetic
material (i. e., altering the structure of DNA). A

genotoxic carcinogen is one that initiates cancer through a direct effect on
genetic material.

Hazard A (potential) source of risk that does not necessarily produce risk.
A hazard is the inherent characteristic of a material, condition, or
activity that has the potential to cause harm to people, property, or the
environment and

produces risk only if an exposure pathway exists and if exposures create the
possibility of adverse consequences. Hazard analysis The identification of
material properties, system elements, or events that lead to harm or loss.
The term hazard analysis may also include evaluation of consequences from an
event or incident. Incidence The rate at which an event occurs. In
toxicology, the number of new cases

of a disease within a specified period of time, often expressed per 100,000
individuals per year. Interspecies extrapolation The act of applying a set
of data or an individual test result from one

species, under certain conditions and subject to particular dose levels of a
toxic substance and application method, to another population of a different
species under perhaps different conditions, dose levels, and application
method.

LC 50 The concentration of a substance in air that when administered by
inhalation to all animals in a test (over a specified time period) is lethal
to 50 percent of the animals. LD 50 The dose that when administered to all
animals in a test (over a specified

time period) is lethal to 50 percent of the animals. Likelihood Expressed as
either a frequency or a probability. Frequency is a measure of the rate at
which events occur over time (e. g., events per year, incidents per year,
deaths per year). Probability is a measure of the rate of a possible

event expressed as a fraction of the total number of events (e. g. 1/
1,000,000).

Lowest effective dose (LED) The lowest dose of a chemical that produced a
specified level of an adverse effect when it was administered to animals in
a toxicity study. For

example, the LED 10 is the lowest effective dose that produced an effect in
10 percent of the exposed animals. Lowest observed adverse

The lowest exposure at which there is a statistically or biologically effect
level (LOAEL)

significant increase in the frequency of an adverse effect when compared
with a control group.

Margin of exposure A ratio defined by EPA as a dose derived from a tumor
bioassay, epidemiologic study, or biologic marker study, such as the dose
associated

with a 10- percent response rate, divided by an actual or projected human
exposure.

Mechanism of action/ mode The mechanism of action is the complete sequence
of biological events of action

that must occur to produce the toxic effect. The mode of action is a
lessdetailed description of the mechanism of action in which some but not
all of the sequence of biological events leading to a toxic effect is known.

Monte Carlo analysis The computation of a probability distribution over
consequences by means of a random sampling method analogous to the game of
roulette. Combinations of events and outcomes that yield possible
consequences are randomly selected according to a specified probability
distribution. The resulting consequences are counted and used to estimate
other probability distributions.

Multistage models Dose- response models that assume a given number of
biological stages occur following exposure to a risk agent (e. g.,
metabolism, covalent

binding, DNA repair) before manifestation of the effect in question is
possible.

Mutagen Any substance that can cause a change in genetic material. Mutagens
have the ability to induce adverse, heritable changes in the genetic
material of

living organisms. No observed adverse effects

The highest dose at which there is no statistically or biologically
significant level (NOAEL) increase in the frequency of an adverse effect
when compared with a control group. No observed effects level The highest
dose at which there is no statistically or biologically significant (NOEL)
increase in the frequency of any effect, adverse or not, compared with a

control group. Order of magnitude An expression often used in reference to
calculations of environmental quantities of risk. Order of magnitude means a
factor of 10. For example,

20 (2 x 10) is 1 order of magnitude greater than 2; 200 (2 x 10 x 10) is 2
orders of magnitude greater than 2; and so forth.

Permissible exposure limit PELs are time- weighted average (TWA) air
concentrations that must not be (PEL) exceeded during any 8- hour work shift
of a 40- hour workweek as defined by OSHA.

Pharmacokinetics Study of the absorption, distribution, metabolism, and
excretion of chemicals and the genetic, nutritional, behavioral, and
environmental factors that modify these parameters.

Pharmacokinetic models Dose- response models based on the principle that
biological effects are the result of biochemical interaction between foreign
substances or metabolites and parts of the body.

Physiologically based A model that estimates the dose to a target tissue or
organ by taking into pharmacokinetic (PBPK)

account the rate of absorption into the body, distribution among target
model organs and tissues, metabolism, and excretion.

Probabilistic approaches Evaluating a range of possible risk estimates and
their likelihood, tied to various mathematical models of the likely
distribution of potential values, instead of relying on single numbers or
point estimates.

Reference concentration A reference concentration is an estimate of a
continuous inhalation

(RfC)/ reference dose (RfD) exposure to the human population (including
sensitive subgroups) that is likely to be without an appreciable risk of
deleterious noncancer effects during a lifetime. EPA uses a reference dose
to express a conservative threshold value for a dose- response relationship
for noncarcinogenic effects. It is an estimate of a daily dose to the human
population (including

sensitive subgroups) that is likely to be without an appreciable risk of
deleterious noncancer effects during a lifetime. Residual risk The health
risk remaining after risk- reduction actions are implemented, such as risks
associated with sources of air pollution that remain after the
implementation of maximum achievable control technology. Risk The
probability that a substance or situation will produce harm under

specified conditions. Risk is a combination of two factors: the probability
that an adverse event will occur and the consequences of its occurrence (e.
g., a specific disease or type of injury).

Risk assessment The systematic, scientific description of potential adverse
effects of exposures to hazardous agents or activities. According to NAS, it
involves the four steps of hazard identification, dose- response assessment,
exposure assessment, and risk characterization. The product of the risk
assessment is a statement regarding the probability that populations,
individuals, or environmental entities so exposed will be harmed and to what
degree.

Risk characterization The process of organizing, evaluating, and
communicating information about the nature, strength of evidence, and the
likelihood of adverse health or ecological effects from particular
exposures.

Risk management The process of analyzing, selecting, implementing, and
evaluating actions to reduce risk.

Screening risk assessment A risk assessment performed using few data and
many assumptions to identify exposures that should be evaluated more
carefully for their

potential risks. Sensitivity analysis A method used to examine the behavior
of a model by systematically measuring the deviation in its outputs produced
as each input, parameter, or assumption is varied from its nominal or base-
case value.

Synergistic effects A term used to describe when the combined biological
effects of two risk agents are greater than the sum of the effects of each
agent acting alone.

Target organ The specific organ affected by a dose of a toxic substance,
which is not necessarily the organ receiving the highest concentration.

Threshold The level of exposure above which adverse health effects are
thought to occur, and below which no adverse effect is thought to occur.

Threshold dose Minimum application of a given risk agent required to produce
a measurable response.

Threshold effect An effect for which there is some dose below which the
probability of an individual's responding is zero. Tiered approach A series
of assessments of increasing complexity. Toxicity A measure of the degree of
harm caused by a specified exposure of human,

animal, or plant life to a chemical substance.

Toxicokinetics The mechanism by which chemical or physical change causes
toxic effects. Toxicology The study of adverse effects of chemicals on
living organisms. Tumor Any abnormal growth of tissue in which growth is
uncontrolled and

progressive. Uncertainty factor One of several factors used to calculate an
exposure level that will not

cause toxicity from experimental data. Uncertainty factors are used to
account for the variation in susceptibility among humans, the uncertainty in
extrapolating from experimental animal data to humans, the uncertainty in
extrapolating from data from studies in which agents are given for less than
a lifetime, and other uncertainties such as using LOAEL data instead

of NOAEL data. Variability A population's natural heterogeneity or
diversity, particularly that which contributes to differences in exposure
levels or in susceptibility to the

effects of chemical exposures.

(410543) Lett er

Ordering Information The first copy of each GAO report is free. Additional
copies of reports are $2 each. A check or money order should be made out to

the Superintendent of Documents. VISA and MasterCard credit cards are
accepted, also. Orders for 100 or more copies to be mailed to a single
address are discounted 25 percent.

Orders by mail: U. S. General Accounting Office P. O. Box 37050 Washington,
DC 20013

Orders by visiting: Room 1100 700 4th St. NW (corner of 4th and G Sts. NW)
U. S. General Accounting Office Washington, DC

Orders by phone: (202) 512- 6000 fax: (202) 512- 6061 TDD (202) 512- 2537

Each day, GAO issues a list of newly available reports and testimony. To
receive facsimile copies of the daily list or any list from the past 30
days, please call (202) 512- 6000 using a touchtone

phone. A recorded menu will provide information on how to obtain these
lists.

Orders by Internet: For information on how to access GAO reports on the
Internet, send an e- mail message with ?info? in the body to:

info@ www. gao. gov or visit GAO?s World Wide Web home page at: http:// www.
gao. gov

To Report Fraud, Contact one:

Waste, or Abuse in  Web site: http:// www. gao. gov/ fraudnet/ fraudnet.
htm

Federal Programs

 e- mail: fraudnet@ gao. gov

 1- 800- 424- 5454 (automated answering system)

GAO United States General Accounting Office

Page 1 GAO- 01- 810 Chemical Risk Assessment

Contents

Contents Page 2 GAO- 01- 810 Chemical Risk Assessment

Contents Page 3 GAO- 01- 810 Chemical Risk Assessment

Contents Page 4 GAO- 01- 810 Chemical Risk Assessment

Contents Page 5 GAO- 01- 810 Chemical Risk Assessment

Contents Page 6 GAO- 01- 810 Chemical Risk Assessment

Page 7 GAO- 01- 810 Chemical Risk Assessment United States General
Accounting Office

Washington, D. C. 20548 Page 7 GAO- 01- 810 Chemical Risk Assessment

Page 8 GAO- 01- 810 Chemical Risk Assessment

Page 9 GAO- 01- 810 Chemical Risk Assessment

Page 10 GAO- 01- 810 Chemical Risk Assessment

Page 11 GAO- 01- 810 Chemical Risk Assessment

Page 12 GAO- 01- 810 Chemical Risk Assessment

Page 13 GAO- 01- 810 Chemical Risk Assessment

Page 14 GAO- 01- 810 Chemical Risk Assessment

Page 15 GAO- 01- 810 Chemical Risk Assessment

Page 16 GAO- 01- 810 Chemical Risk Assessment

Page 17 GAO- 01- 810 Chemical Risk Assessment

Page 18 GAO- 01- 810 Chemical Risk Assessment

Page 19 GAO- 01- 810 Chemical Risk Assessment

Page 20 GAO- 01- 810 Chemical Risk Assessment

Page 21 GAO- 01- 810 Chemical Risk Assessment

Page 22 GAO- 01- 810 Chemical Risk Assessment

Page 23 GAO- 01- 810 Chemical Risk Assessment

Page 24 GAO- 01- 810 Chemical Risk Assessment

Page 25 GAO- 01- 810 Chemical Risk Assessment

Page 26 GAO- 01- 810 Chemical Risk Assessment

Page 27 GAO- 01- 810 Chemical Risk Assessment

Page 28 GAO- 01- 810 Chemical Risk Assessment

Page 29 GAO- 01- 810 Chemical Risk Assessment

Page 30 GAO- 01- 810 Chemical Risk Assessment

Page 31 GAO- 01- 810 Chemical Risk Assessment

Page 32 GAO- 01- 810 Chemical Risk Assessment

Page 33 GAO- 01- 810 Chemical Risk Assessment

Page 34 GAO- 01- 810 Chemical Risk Assessment

Page 35 GAO- 01- 810 Chemical Risk Assessment

Page 36 GAO- 01- 810 Chemical Risk Assessment

Page 37 GAO- 01- 810 Chemical Risk Assessment

Page 38 GAO- 01- 810 Chemical Risk Assessment

Page 39 GAO- 01- 810 Chemical Risk Assessment

Page 40 GAO- 01- 810 Chemical Risk Assessment

Page 41 GAO- 01- 810 Chemical Risk Assessment

Page 42 GAO- 01- 810 Chemical Risk Assessment

Page 43 GAO- 01- 810 Chemical Risk Assessment

Page 44 GAO- 01- 810 Chemical Risk Assessment

Page 45 GAO- 01- 810 Chemical Risk Assessment

Page 46 GAO- 01- 810 Chemical Risk Assessment

Page 47 GAO- 01- 810 Chemical Risk Assessment

Page 48 GAO- 01- 810 Chemical Risk Assessment

Page 49 GAO- 01- 810 Chemical Risk Assessment

Page 50 GAO- 01- 810 Chemical Risk Assessment

Appendix I

Appendix I Objectives, Scope, and Methodology

Page 51 GAO- 01- 810 Chemical Risk Assessment

Appendix I Objectives, Scope, and Methodology

Page 52 GAO- 01- 810 Chemical Risk Assessment

Appendix I Objectives, Scope, and Methodology

Page 53 GAO- 01- 810 Chemical Risk Assessment

Appendix I Objectives, Scope, and Methodology

Page 54 GAO- 01- 810 Chemical Risk Assessment

Page 55 GAO- 01- 810 Chemical Risk Assessment

Appendix II

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 56 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 57 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 58 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 59 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 60 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 61 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 62 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 63 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 64 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 65 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 66 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 67 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 68 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 69 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 70 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 71 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 72 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 73 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 74 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 75 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 76 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 77 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 78 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 79 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 80 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 81 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 82 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 83 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 84 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 85 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 86 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 87 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 88 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 89 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 90 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 91 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 92 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 93 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 94 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 95 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 96 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 97 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 98 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 99 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 100 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 101 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 102 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 103 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 104 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 105 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 106 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 107 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 108 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 109 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 110 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 111 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 112 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 113 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 114 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 115 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 116 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 117 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 118 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 119 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 120 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 121 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 122 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 123 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 124 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 125 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 126 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 127 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 128 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 129 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 130 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 131 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 132 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 133 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 134 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 135 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 136 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 137 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 138 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 139 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 140 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 141 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 142 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 143 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 144 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 145 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 146 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 147 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 148 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 149 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 150 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 151 GAO- 01- 810 Chemical Risk Assessment

Appendix II Chemical Risk Assessment at the Environmental Protection Agency

Page 152 GAO- 01- 810 Chemical Risk Assessment

Page 153 GAO- 01- 810 Chemical Risk Assessment

Appendix III

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 154 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 155 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 156 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 157 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 158 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 159 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 160 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 161 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 162 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 163 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 164 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 165 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 166 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 167 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 168 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 169 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 170 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 171 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 172 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 173 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 174 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 175 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 176 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 177 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 178 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 179 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 180 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 181 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 182 GAO- 01- 810 Chemical Risk Assessment

Appendix III Chemical Risk Assessment at the Food and Drug Administration

Page 183 GAO- 01- 810 Chemical Risk Assessment

Page 184 GAO- 01- 810 Chemical Risk Assessment

Appendix IV

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 185 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 186 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 187 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 188 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 189 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 190 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 191 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 192 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 193 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 194 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 195 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 196 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 197 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 198 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 199 GAO- 01- 810 Chemical Risk Assessment

Appendix IV Chemical Risk Assessment at the Occupational Safety and Health
Administration

Page 200 GAO- 01- 810 Chemical Risk Assessment

Page 201 GAO- 01- 810 Chemical Risk Assessment

Appendix V

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special Programs Administration

Page 202 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 203 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 204 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 205 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 206 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 207 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 208 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 209 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 210 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 211 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 212 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 213 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 214 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 215 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 216 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 217 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 218 GAO- 01- 810 Chemical Risk Assessment

Appendix V Chemical Risk Assessment at the Department of Transportation?s
Research and Special

Programs Administration Page 219 GAO- 01- 810 Chemical Risk Assessment

Page 220 GAO- 01- 810 Chemical Risk Assessment

Glossary Page 221 GAO- 01- 810 Chemical Risk Assessment

Glossary Page 222 GAO- 01- 810 Chemical Risk Assessment

Glossary Page 223 GAO- 01- 810 Chemical Risk Assessment

Glossary Page 224 GAO- 01- 810 Chemical Risk Assessment

Glossary Page 225 GAO- 01- 810 Chemical Risk Assessment

Glossary Page 226 GAO- 01- 810 Chemical Risk Assessment

Glossary Page 227 GAO- 01- 810 Chemical Risk Assessment

Glossary Page 228 GAO- 01- 810 Chemical Risk Assessment

Glossary Page 229 GAO- 01- 810 Chemical Risk Assessment

United States General Accounting Office Washington, D. C. 20548- 0001

Official Business Penalty for Private Use $300

Address Correction Requested Presorted Standard

Postage & Fees Paid GAO Permit No. GI00
*** End of document. ***