United States
Environmental Protection
Agency
EPA 600/9-85-002
December 1984
Risk Assessment
and Management:
Framework for
Decision Making
Do not remove. This document
should be retained in the EPA
Region 5 Library Collection.
-------
PROPERTY OF THE
OFFICE OF SUPERFUND
PREFACE
The National Academy of Sciences, in its 1983 report
on Risk Assessment and Risk Management, provides a
great service to those of us in the business of protecting
public health and the environment. The report offered a
clear distinction between the role of science in helping us
assess the nature and extent of health and environmental
problems, and the role of government managers in
determining the most appropriate responses to those
problems. Perhaps even more important, the report
suggests a number of ways in which regulatory decision-
making can be made more consistent and rational and,
thus, more understandable and acceptable to the
American public.
In a speech to the National Academy of Sciences soon
after my return to EPA, I proposed that the Agency adopt
as many as possible of the report's risk assessment and risk
management goals. And we have. The Environmental
Protection Agency has initiated a very wide range of
activities designed to implement Academy recommenda-
tions. These activities are detailed in the following pages.
These new initiatives reflect our larger purpose to not
only strengthen the scientific basis upon which we take
regulatory action but to also make clear the necessary
distinction that must be made between the definition of
an environmental risk and what is done to reduce that
risk.
But our most fundamental belief is that the public
needs to understand fully how we intend to go about the
business of reducing risk in our society. This report is
meant to aid that understanding.
William D. Ruckelshaus
December, 1984 Envjronmental Protection Agency
Reaion 5, Library (PL-12J)
77 West Jackson Boulevacd, 12th Floor
Chicago, IL 60604-3590
-------
TABLE OF CONTENTS
PAGE
NUMBER
CHAPTER ONE
INTRODUCTION
A. PURPOSE OF THIS REPORT 1
B. PRACTICAL USE OF RISK ASSESSMENT
AND RISK MANAGEMENT CONCEPTS 3
C. ADVANTAGES OF THIS APPROACH 5
D. OUTLINE OF THIS REPORT 8
CHAPTER TWO
RISK ASSESSMENT: IMPROVING THE
SCIENTIFIC FOUNDATION
A. STRUCTURE OF THE ASSESSMENT PROCESS ... 13
B. UNCERTAINTY IN RISK ASSESSMENT 14
C. IMPORTANCE OF SCIENCE GUIDELINES 16
D. DEVELOPING RISK ASSESSMENT
GUIDELINES 19
E. ESTABLISHMENT OF A FORUM ON RISK
ASSESSMENT ISSUES 22
CHAPTER THREE
GOALS AND APPLICATIONS OF RISK MANAGEMENT
A. DEFINING RISK MANAGEMENT 23
B. ELEMENTS OF RISK MANAGEMENT: SETTING
PRIORITIES AND MAKING CHOICES 25
C. CURRENT RISK MANAGEMENT
INITIATIVES 30
-------
CHAPTER ONE
INTRODUCTION
A. PURPOSE OF THIS REPORT
This report is about risk assessment and risk management at
the Environmental Protection Agency. In recent months both terms
ha\e been used extensively in public statements by senior Agency
officials and by the press in reference to many aspects of EPA's
work. We hope to clarify here what these terms mean in the context
of current Agency policy and operations.
Perhaps the most important thing any organization does is to
reassess periodically how it can best fulfill its basic purposes and
goals. Whether public or private, no institution can be successful
unless it knows exactly what its business is, and how it intends to
accomplish its mission. For EPA, a large organization with main
complex responsibilities, being clear about both ends and means is
essential. Controlling toxic chemical pollution is one of our most
vital missions. This report presents our plans to enable the Agency
to make belter and more rapid decisions about environmental toxic
chemical problems. This is not a technical report about the details
of scientific research. Instead it discusses how we intend to manage
the application of scientific research most effectively for making
control decisions. Our audience is the general public, which needs
to understand as much as possible about how we arrive at decisions
to protect the environment and public health.
The goal of the Environmental Protection Agency is to improve
the condition of the environment — to reduce risks to human
health, and to protect and enhance the qualitv of natural ecosystems.
This much has always been clear, but the more recently understood
threat of toxic chemicals has greatly inc reased the difficulty of
carrying out that assignment.
The job of reducing the risks of toxic chemicals is simply larger
and more complex than anyone expected. For one thing, science
allows us to detect ever smaller amounts of pollution. Air and water
that seemed pure ten years ago are now revealed to be contaminated,
even if only to a very small degree. Problems we once regarded as
solved turn out not to be solved. Each new scientific revelation
seems only to lengthen the agenda of possible actions. The number
of issues to address is now so large that the first order of business is
simply to separate problems from non-problems and grounded
fears from ungrounded ones.
In the past there was widespread agreement over what the major
environmental problems were, and how they should be controlled.
-------
Municipal sewers were discharging untreated waste into the nation's
rivers and streams, industrial stacks were emitting millions of tons
of participates and sulfur into the atmosphere unconstrained, and
automobiles were discharging over ten times as much pollution per
mile oi travel as they do now. The total impact of this stress on the
em ironment was not only clearly visible, it was an unambiguous
threat to public health and well-being.
Congress responded by passing the Clean Air and Clean Water
Acts, among other pieces of landmark legislation. Progress in
treating the gross forms of pollution at which these laws were
principally aimed has been remarkable. By any standard of mea-
surement we have made vast improvements: the air and water are
much cleaner than the\ were when EPA was set up in 1970. Our
programs have set a standard that much of the rest of the world
follows.
But the administrative tools that provided the first substantial
measure of progress are not fully appropriate for addressing the
qualitatively different problem of making decisions about control-
ling toxic wastes. As this report describes, our remaining pollution
problems are numerous but more subtle. The economic conse-
quences of addressing them are also potentially very great —
perhaps as large or larger than those of the first round of
environmental controls.
We cannot procrastinate in our mission of cleaning up toxic
chemical pollution, but we cannot afford to make man) mistakes
either. Time wasted in chasing down the wrong chemicals is time
wasted in controlling risks of cancer, birth defects, and other feared
diseases. "1'he issue is not simply one of dollars. Toxicology is a
growing but still comparatively tiny field. We cannot be profligate
with the research of the few experts we have; their time must be
husbanded as the scarce commodity that it is.
Controlling toxic chemicals is ultimately a management problem.
The scientific and technical issues are difficult, but no matter how
much they are debated, the basic question is always what to do next.
We need a new strategy that allows us to act constructively, despite
the uncertainty that surrounds us.
These discussions describe the outlines of such a strategy. In the
short term it envisions no drastic departures from past practice, but
oxer the long term it should demonstrably impnne our rate of
progress in controlling environmental health risks. It will do this by
aiming resources at the worst problems first, reducing the number
oi false moves, and helping to arriu1 at more consistent and
practical courses of action.
-------
B. PRACTICAL USE OF RISK ASSESSMENT AND
RISK MANAGEMENT CONCEPTS
In a speech to the National Academy of "Sciences, Administrator
Ruckelshaus described the distinction between risk assessment and
risk management:
"Scientists assess a nsk to find out what the problems an>. The process of
(leading what to do about the problems is risk management. The second
puxedure involves a much btoader array of disciplines, and is aimed
toward a deci.swn about (onhol. Risk management assumes we have
assessed the health risks of a suspect (hemical. We must then factor in its
benefits, the costs of the vanou.s methods available foi its control, and the
statutory fiameu'ork for decision."
The distinction between the two activities has become an attractive
means for understanding and improving upon the two fundamen-
tal processes involved in emironmental decisionmaking. This
distinction was a major point in the NAS Report, Risk Assessment
in the Federal Goveinmenl, many recommendations of which are
being considered for use at EPA.
But as Agency management has concentrated more attention
on risk assessment and risk management during the past year, the
question that has arisen is: ho\v should our understanding of the
distinction between these tv\o concepts be used to guide the long-
term management of the Agencv?
The issue of Agencv management is complicated by the
peculiar nature of EPA's mandate: Implementation of eight major
emironmental statutes, each dealing with a different aspect of"
emironmental protection, and each carried out with a history of
considerable independence. Some of these statutes require or allow
EPA to base its regulatou decisions direct!} on risk reduction.
Other regulator) decisions, such as the control of toxic pollutants in
ihe Clean Water Act Effluent Guidelines Regulations, are to be
based on available technology and cost instead of risk reduction.
I'hus, while we are proposing risk reduction as the integrating
concept for Agency management, we clearly intend to apply this
approach only to the extent possible and reasonable within the
constraints of the various statutes.
Additionally, risk assessment and risk management goals must be
understood within the general context of the Agency's need to act.
It is usually possible for the Agency to obtain more data with which
to evaluate ha/aid, exposure, risk reduction efficiency ol control
strategies, and costs. This information, as is frequently said of
pollution controls themselves, max take longer and cost more to
-------
obtain, as precision and comprehensiveness increase. This in turn
may mean that environmental and health costs incurred during the
evaluation process increase. Thus, there must be a balance between
the incremental costs of improved information and the benefits of
regulating "more efficiently across program lines" especially where
statutory considerations or political considerations may foreclose
the more efficient options.
Program integration has always been a problem for EPA leader-
ship because of the diverse mandates of the Agency. Nevertheless,
such integration of environmental programs was one of the reasons
for establishing EPA in the first place. Failure to coordinate has
often led to management difficulties, including: duplicative research
on the same substance; different programs producing different
risk assessments for the same substance; unwitting transfer of
pollutants from one em ironmental medium to another via pollution
control technology; uncoordinated regulation of the same industry
or the same substance by different programs; and seemingly
different risk management decisions from different programs on
the same substances.
For several reasons, program integration problems tend to be
exacerbated by the increased attention the Agency must pay to
toxic substances. First, the number of potentially toxic materials is
extremely large. There are over 65,000 industrial chemicals listed
as having been in commercial production since 1945. While many
are not yet characterized as to their toxicological potential, a few
thousand have some demonstrated toxic effect and are encountered
in sufficient volume to be of concern. Many of these chemicals are
of concern to more than one program, so that scientists and
managers in different regulatory offices find themselves dealing
with the same substances more often than in the past.
Next, since hazardous substances are often not destroyed or
permanently isolated by actions meant to control them, it is
necessary both to keep track of their movements — preventing
pollution control from degenerating into an expensive shell game
— and to make decisions about the most acceptable endpoint for
persistent pollutants. This is important because some toxics may
have effects on health at exceedingly small concentrations, so that in
some cases it may be impossible to establish a level at which they
present no risk at all. Hence, regulation must depend on some
balancing approach.
These aspects of the toxics problem argue for a more integrated
and readily comprehensible way to make regulatory decisions for the
-------
Agency as a whole. More consistent risk assessments and improved
risk management have thus become attractive as integration tools.
Neither risk assessment nor risk management is new at EPA.
Nothing now planned is going to make these processes easier or less
complex. They are inherently difficult things to do. What we hope to
accomplish //v the current emphasis is to foster consistency using the best
science and the best judgment across actions and across programs in both risk
assessment and risk management, and to make the many judgments that lead
to regulatory decisions more explicit.
The essential requirement of a risk management approach is to
demonstrate, to the extent possible, what each regulation does in
terms of reducing risk. We wish to ask: when this regulation takes
effect, will there be fewer deaths, less sickness, better visibility, more
fish, safer drinking water? Risk management means that we will, for
reasons connected with the changing nature of pollution control,
try to make our actions more consistent across programs and more
explicit in terms of environmental risk reduction.
The application of this concept of risk management works best
when we can easily quantify, or express in defensible numerical
form, the pollution risks Agency program offices are seeking to
reduce through regulation. But we cannot, of course, quantify
every value mandated by ernironmental statute nor can we reliably
reduce to numerical form all risks of toxic pollution to health and
the cmironment. Use of analytic techniques in risk management
should not imply a false precision, and we will continue to take into
account the need for qualitative judgment.
C. AD VANTAGES OF THIS APPROACH
The risk assessment and risk management initiatives described
in this report are tools which will help make possible more efficient
protection of the environment and human health. We expect to
gain the following specific management advantages:
• Risk assessment and risk management help set priorities.
As already noted, there are now thousands of chemicals in
commerce and an unknown number of contaminants and
unintended by-products. Some of these could be important as
pollutants, and as such, are proper targets for regulation if they
pose significant risks to health or the environment. We do not have
the budget, nor will we ever have the time, to test each chemical
exhaustively.
-------
ELEMENTS OF RISK ASSESSMENT AND RISK MANAGEMENT.
-------
Some of" our priorities will always be set for us from outside,
either by Congress or through the press of emergencies, but it
would be foolish to suppose that the great underside of the iceberg
will somehow take care of itself. We cannot escape the need for an
analytic approach to setting our own agenda.
Screening by estimates of potential risk reduction is an attractive
basis for such an approach. We can use risk analytic methods to
help sort problems in terms of the likelihood that the Agency can
do something constructive and effective to improve public health
and the environment. It makes no sense to spend time and Agency
resources to write regulations on chemicals, even highly toxic ones,
to which no one will ever be exposed or for which there is no
capability for additional control. On the other hand, it may make a
great deal of sense to control chemicals that are only moderately
toxic, if we have evidence that many people are exposed and that
practical controls are possible.
• Risk management provides a context for balanced analysis and
decision-making.
Toxic chemicals are legitimately frightening: they can and do
cause cancer and other diseases. The trouble is that we are exposed
to a complex, highly dilute mixture of chemicals, taken in through
air, water, and food. When disease strikes, cause and effect are
seldom clearh1 linked. Often the regulatory situation, in which any
action may involve substantial health or economic impacts and in
which the scientific basis is highly uncertain, allows extreme points
of view to develop. This polarizes debate, sometimes bringing
public policy to an impasse. The Agency can contribute to rational
discussion b) exposing the scientific basis for the risk, including the
confidence we have in the estimate; placing the risk reduction
expected from the regulation in context with other risks and other
opportunities for risk reduction; and explaining the values on
which the balancing judgments have been made.
• Risk assessment and management produce more efficient and
consistent risk reduction policies.
EPA's patchwork of authorities for controlling pollution needs to
be woven together moie coherently, beginning at the analytical level
and continuing through to the regulatory decision. Some impor-
tant differences — defined by statute — in the ways the laws
manage risk will always remain, but a risk management approach
can use our remaining administrative flexibility to make more
efficient use of the Agency's and society's resources to reduce risk
and to make the AgencVs actions more consistent.
-------
8
Enhancing efficiency means, when appropriate, examining all
available regulatory opportunities across all programs and selecting
those that buy the most risk reduction for any given level of
resources. It is also important to have consistent methods of
assessing risk for the same substances across programs. Although
our various mandates ma\ require us to respond differently to risks
that our calculations show are similar, we still need a consistent
method for understanding the connection between our goals under
the law and the actions we take. Further, unless we treat risks and
their control consistently, we may fail to recognize opportunities to
reduce risks in one arena that may contribute more to public health
and the environment than further expenditures in another. Consis-
tent policy-making based on well articulated principles is perhaps
the most important element in creating a strong base of public
understanding of Agency actions.
D. OUTLINE OF THIS REPORT
During the summer of 1983, Administrator Ruckelshaus and
Deputv Administrator Aim began a number of efforts to investigate
the issues associated with the assessment and management of risk
and to start making necessary improvements to the regulatory-
process as a whole. One major initiative was the creation of the
Toxics Integration Task Force, a group of senior agency scientists
and managers. This group was asked to suggest practical improve-
ments that could be implemented within existing organizational
and statutory limits; its final proposals are all being implemented.
Some of this report is based on that work. The rest describes
either analytic practices which have been evolving in the Agency
over the past several years or other new initiatives in this area
started in the past year by the Administrator and Deputy
Administrator.
Chapter Two focuses on improving the scientific foundations
for using risk assessment in the EPA regulatory context. It briefly
describes the wide range of activities that the Agency uses to
describe and quantify the effects of pollutants on human health
and the environment. Assessments of risk are inherently imprecise
— because knowledge is incomplete, and because the results of the
process depend heavily on the procedures and assumptions used.
The line separating "science" from "science judgments" is not a
sharp one, in that judgment must be used where firm data are
absent. In order to avoid the necessity (or ad hot judgments in indi-
vidual cases, it is desirable to develop generic inference guidelines
for different aspects of risk assessment. They also make the
Agency's actions more predictable.
-------
9
This chapter describes work underway to develop six new or
revised risk assessment guidelines, and the creation of an internal
risk assessment forum. The forum will resolve risk assessment
disputes and will provide a mechanism for updating guidelines as
new knowledge becomes available.
Chapter Three discusses current and new approaches for using
risk management concepts in the regulatory process.
EPA's programs have differing mandates with respect to public
health protection, and these bear significantly on the way the
Agency deals with pollution control. The variety and uncertainty of
risks from toxic substances makes exercising these mandates com-
plicated and difficult. Risk management involves statements about
values, and about the way that EPA interprets its statutory man-
dates. The public has a right to know what values the Agency is
applying. This chapter describes several initiatives we are taking to
articulate risk management within the regulatory development
process, work we are doing to develop better analytic models for
setting regulatory priorities, and recent progress in benefits
assessment and regulatory impact analysis.
The actions described are only a start, but we believe they will
bring significantly more clarity, structure, and predictability to the
management of risks than has been the case up to now. We also
hope they will be a step toward improving public trust in EPA's
ability to address controversial matters of" policy in a constructive
arid open way.
-------
CHAPTER TWO
RISK ASSESSMENT: IMPROVING THE SCIENTIFIC FOUNDATION
EPA regulatory decisions address, of course, a wide range of
possible pollutant effects. Human health concerns include genetic-
damage and neurological effects as well as cancer, and we consider
such adverse environmental impacts as ecosystem disruption,
crop damage, and atmospheric impairment. At present, however,
the Agency has progressed further in developing procedural
guidelines for human health risk assessment than we have for
environmental effects. Also, our quantitative analytic techniques are
most refined for cancer assessment. The focus of this chapter is
mainly on innovations in the narrower area of human health risk
assessment.
Health risk assessments are conducted by scientists, but they are
not "classical science" in the strictest sense. For regulatory purposes,
risk assessments represent a tool that can be used to analyze
scientific evidence in order to evaluate the relationship between
exposure to toxic substances and the potential occurrence of
disease. The risk assessment process involves, on one extreme,
scientifically verifiable findings, and, on the other extreme, judg-
ments about the use of various kinds of scientific information. No
one should be misled into believing that results using present
techniques have the status of incontrovertible scientific agreement.
Despite its uncertainties, however, risk assessment is the only tool
we have for discriminating among environmental health problems.
The central question we address here is the extent to which risk
assessment judgments can be made more consistent and more
reflective of the state of scientific understanding.
There is no constant formula for conducting a risk assessment.
Because this is an analytical tool, it can be argued that it must be
tailored to the needs of the program in which it is used. Given the
different mandates within the Agency, it is not surprising that there
are a variety of reasons for performing risk assessments and an
equal variety of methods used to conduct them. Some examples
follow to demonstrate the diverse nature of these assessments.
Risk assessments of carcinogens are assessments of risk in the
literal sense, i.e., an estimation of the probability of developing
cancer as a direct result of chemical exposure. Carcinogens are,
by EPA policy, assumed not to have thresholds, i.e., no level of
exposure is assumed to be without risk unless there is specific
evidence to the contrary in a particular case. The data underlying
-------
12
such assessments typically involve doses to animals much higher
than the exposure levels expected for the human populations of
interest. Therefore, the risk assessments use probabilistic models to
draw dose-response curves to extrapolate from the higher, experi-
mental, dose levels, down to the zero exposure point. The product
of this kind of assessment is an estimate of the probability (risk) that
additional cases will be associated with some given exposure level.
This probability is usually expressed as a unit dose value, or risk
per unit of exposure, such as risk per part per million (ppm) of a
substance in the air or in drinking water. This unit risk can than be
multiplied by exposure levels and by the number of people exposed
to generate estimates of excess cancer incidence associated with the
exposure.
This approach to risk assessment can be used for studying any
health effect where thresholds are assumed not to exist, provided
that there exist adequate data to construct a dose-response curve.
Note that efforts are underway (such as in the air program) to
develop probabilistic risk estimates for noncarcinogens.
In other types of risk assessments, where a threshold effect is
assumed for a toxic substance, the primary focus of the analysis may
be to determine the "safe" or no-effect level of exposure. Informa-
tion derived from animal experiments, or human data in cases
where such data are available, is used to establish a "no-observable
effect level" (NOEL) or "lowest observed effect level" (LOEL)
for the substance in question. Such levels are then divided by
uncertainty factors, which vary depending on the nature of the
supporting data, to produce an acceptable exposure level. These
resulting levels may be expressed as acceptable daily intakes (ADI).
We assume that persons exposed to levels of pollutants below these
acceptable exposure levels will not suffer adverse effects from the
exposure. ADIs are then combined with any other available data
regarding effects above the threshold level.
Risk assessments can also be used as a screening dev ice for setting
priorities, or as a method for quickly ranking the relative toxicities
of large numbers of chemicals. This type of approach is usually less
detailed, and typically involves some kind of health effects rating
system. It is important, because formal risk assessment is slow and
expensive, and quicker decisions are often needed on the handling
of many chemical or regulatory problems. Two examples of this
type of screening assessment are the Superfund hazard ranking
system, and the methodology used by the Integrated Environ-
mental Management Division in the Office of Policy, Planning and
Evaluation. Also, the Office of Toxic Substances uses abbreviated
risk assessment methods which relv heavilv on structure activity
-------
relationships to screen new chemicals that lack adequate
lexicological information.
Where specific geographical areas have been contaminated
with toxics, the Agency must respond with an assessment of the
potential danger to people in the area and, in some cases, with
estimates of the risks associated with various clean-up options.
These assessments do not necessarily differ in kind from those
mentioned above, but may involve more or less detail depending
on the exigencies of the situation.
A. STRUCTURE OF THE ASSESSMENT PROCESS
In the simplest sense, population risks from toxic pollutants
are a function of two measurable factors: hazard and exposure. To
cause a risk, a chemical has to be both toxic (present an intrinsic
hazard), and be present in the human environment at some
significant level (provide opportunity lor human exposure). Risk
assessment interprets the evidence on these two points, judging
whether or not an adverse effect will occur, and (if appropriate)
making the necessary calculations to estimate the extent of total
effects.
In a regulatory setting, risk assessment has one or more of the
following four steps. One usually starts an assessment by consider-
ing ha/ard identification or exposure. If either is negative, one does
not proceed.
/. Hazard identification
This exercise involves weighing the available evidence and
deciding whether a substance exhibits a particular adverse health
effect. Most attention has been focused on cancer, but we may also
want to regulate on the basis of other effects, such as damage to
fetuses (teratological effects), inherited conditions (mutational ef-
fects), and damage to specific organs such as the liver or kidneys.
The Drinking Water and Air programs, for instance, regulate
lead on the basis of its neurotoxic effects, and the Toxic Substances
program is considering regulation of glycol ethers on the basis of
their teratogenic effects.
2. Dose-response assessments
Once it is determined that a chemical is likely to cause a
particular human effect, we then determine its potency: how
strongly it elicits that response at various levels of exposure (dose).
-------
14
Chemical potency varies widely; for instance, both saccharin and
dioxin cause cancer in animals, but it takes literally millions of times
more saccharin than dioxin to produce equivalent effects in the
laboratory (fortunately, real world exposure to dioxin is also much
lower).
3. Exposure assessment
We then estimate the likely degree of human exposure to a
chemical of concern. The best method is direct measurement or
monitoring of ambient conditions, but this is often prohibitively
expensive. In practice, we must usually rely on estimates of
emissions and limited monitoring information, combined with
mathematical models that estimate resulting concentrations.
The degree of exposure of concern may vary from pollutant to
pollutant. For mam effects, we may be primarily interested in
lifetime exposures over the whole population; for others, we may be
concerned about maximum levels of exposure to people near the
emission source, or peak levels of short term exposure. We are also
concerned with unusually sensitive portions of the population
(children, the elderly, people suffering from respirator}- or other
particular illnesses).
4. Risk characterization
Finally, we estimate the risk associated with the particular
exposures in the situation being considered for regulation. While
the final calculations themselves are straightforward (exposure
times potency, or unit risk), the way in which the information is
presented is important. The final assessment should display all
relevant information pertaining to the decision at hand, including
sue h factors as the nature and weight of evidence for each step of
the process, the estimated unc ertaint\ of the component parts, the
distribution of risk across various sectors of the population, the
assumptions contained \\ithin the estimates, and so forth.
B. UNCERTAINTY IN RISK ASSESSMENT
(iixen the usual limitations in the nature and extent of
information available in this process, we can ne\er say exactly how
mam people will be affected b\ a particular pollutant, or how
severely. Not enough is known about how pollutants contribute to
certain diseases or about the nature of human exposure to make
definitive findings possible.
-------
Definable relationships occasionally exist between certain dis-
eases and certain substances, as they exist, in a more verifiable way,
between diseases and bacteria. The task of risk assessment is to
make the most credible possible statements about these relation-
ships, reducing uncertainty as much as possible, and making
explicit whatever uncertainty remains.
This section discusses several of the types of uncertainty that
occur at various stages in a risk assessment.
1. Weight of evidence problems in hazard identification
Most risk assessments depend on animal tests. These tests allow
rigorous control over main factors that contribute to uncertainty,
but some fundamental problems remain.
For instance, animal biological systems are different from human
ones. Some species appear more sensitive to certain substances than
humans, and less sensitive to other substances. Or we may find that
a chemical is a strong carcinogen in various test animals, but it
induces a t\ pe of tumor that humans do not get. Another chemical
mav be a carcinogen in only one animal species. Should we consider
the chemical a human carcinogen?
2. Uncertainties in dose-response assessment
Despite existing scientific practice, it is not entirely clear that vife
leveh or tlupshold1, truly exist for any toxic chemicals (or, indeed, that
no threshold exists for carcinogens) and if so, at what levels. While
immediate effects such as respiratory distress may not occur after
short exposures to low doses, subtle damage to health wwy occur
alter long-term, low-level exposure and could do enough damage
to the population as a whole to be worth controlling.
Also, as mentioned above, scientists must extrapolate dose-
response relationships from animals to humans. Dosages must be
corrected for human-animal differences in weight or metabolism.
Effects at low doses must be inferred from high-dose results in
laboratory or epidemiology studies. For cancer, results have typ-
ically been couched in terms of the maximum amount of excess
disease that a chemical is likely to produce. This is a complex
process, with uncertainties attached to every judgment and in-
ference made. Because of its complexity and its heavy reliance on
assumptions, dose-response estimation for carcinogens is a particu-
larly controversial aspect of risk assessment. The Agency is
continuing to explore new statistical approaches for more accurately
representing close-response evidence.
-------
16
The Agency is not alone in its concern that different assumptions
and different mathematical models used can significantly alter the
outcome of a risk assessment. When the Occupational Safety and
Health Administration (OSHA) published its cancer policy in 1980,
it did detailed comparisons ol how estimates of carcinogenic risk
can vary with the assumptions used in developing the estimates (45
FR 5198-5200). By varying the method of low dose extrapolation
used, and the toxicology or epidemiology study which formed the
basis of the risk assessment, commenters to the OSHA policy
developed risk estimates for exposure to 1 ppm of vinyl chloride
which ranged from 10"K (one in one hundred million) to 10 '
(one in ten, or 10%). A similar exercise with saccharin by NAS, and
reprinted in the OSHA policy (45 FR 5200), estimated the expected
number of cancer cases in the general population (exposed at 0.12
grams/day) at between (J.001 cases per million exposed, and 5200
cases per million exposed. These differing estimates were developed
by using different low-dose extrapolation models and different
animal-to-human extrapolation methods — all of which had some
credence in the scientific communitv.
3. Uncertainty in exposure assessment
Thus far, exposure assessment has attracted less controversy
than the other stages of risk assessment, but it also involves
uncertainty.
Exposure assessment is based on human monitoring, ambient
monitoring, modeling, or some combination of these. Human data
and monitoring are typically quite limited, because we do not have
the resources or time required to do scientifically valid studies for all
the pollutants of concern.
Modeling is used to fill the gaps. In these techniques, data on
pollutant releases, release characteristics, meteorology, hydrology,
terrain, etc. are arranged and interrelated in mathematical models.
Computers are used to calculate the distribution of pollutants in the
ambient air and water at various distances from the pollutant
sources. The population exposed to these chemicals is then esti-
mated using census data, and information about drinking water
sources and other exposure routes.
C. IMPORTANCE OF SCIENCE GUIDELINES
To make progress in the face of such uncertainties we must
develop what the National Academy of Sciences report calls
"inference guidelines." Despite scientific consensus endorsing such
an approach, there is still debate about the degree to which standard
-------
17
procedures can or should he defined and about the terminology
involved in the process. The NAS report recommends, and we
believe, that greater c onsistency in the risk assessment process is
not only possible, but essential.
/. Principles
Assessments have historically been based on a number of
underlying principles, such as the judgment that animal bioassays
are indicative of probable human response, or that no threshold of
response to carcinogens can fie confidently defined. A statement of
principles may include simple observations of fact, statements of
broad-based scientific consensus, or judgments about "science
policy." While debate continues in some areas, there is wide
scientific agreement on most of the principles driving the assess-
ment of cancer risk. The most comprehensive reference on the
subject is the recent draft document bv the Office of Science and
Technology Policy, entitled Chemical Caicmogeiis: Review of the
Sciente and its Associated Principles. This document, published for
public comment in the Federal Register on May 22, 1984, provides a
thorough and well-referenced review of current knowledge about
the physical, chemical, and biological events underlying the process
of tumor development. As yet, no similarly comprehensive refer-
ence on health effects other than cancer is available. Although
general principles exist for approaching other types of risk assess-
ment, there is still considerable uncertainty and disagreement
about the details of how to relate pollutant exposures to the
incidence of noncancer effects.
2. Guidelines
Statements of principle in themselves offer direction for con-
ducting specific risk assessments, but they should be augmented by
more specific technical guidance appropriate to regulatory needs.
Attempts at more practical common approaches have therefore
evolved in the \arious regulatory agencies over the last several years.
General procedures for estimating risk (mainly in reference to
cancer) have been codified in the form of guidelines, which
include:
1. Interim Procedures and Guidelines for Health Risk and
Economic Impact Assessments for Suspected Carcinogens,
EPA, 41 FR 24102 (May 25, 1976).
2. Scientific Bases for Identification of Potential Carcinogens
and Estimation of Risks, Report by the Work Group on Risk
Assessment of the Interagency Regulatory Liaison Group,
44 FR 39858 (July 6,1979).
-------
18
3. For a specific EPA program:
Guidelines and Methodology Used in the Preparation of
Health Assessment Chapters ot the Consent Decree Water
Criteria Documents, Appendix C of Water Quality Criteria
Documents: KPA, 45 FR 79347 (November 28, 1980).
See also:
Appendix E: Response to Comments on the Human Health
Effects Methodology foi Deriving Ambient Water Quality
Criteria, 45 FR 79308.
4. Mutagenicity Risk Assessments: Proposed Guidelines, EPA 45
FR 74984 (November 13, 1980).
5. Guidance for the Preparation of Exposure Assessments, EPA,
Draft, September 12, 1983.
Conference proceedings on reproductive risks are also available at
EPA, and we can use these proceedings to guide evaluations of risk
in the absence of formal guidelines.
To be usetul, guidelines must simultaneously balance the need to
be comprehensive, specific, and flexible. Comprehensive means they
should deal with each discrete step of a risk assessment. Spec if K
means tlie\ must describe the scientific basis for each step, and how-
to perform it. And flexible means they must allow, even encourage,
departure from the general approach if data are available suggest-
ing that an alternative is preferable, or if a single approach is not
appropriate to the case at hand.
During a recent brief review of current risk assessment practices
throughout the Agency, we became concerned about two issues:
• Are there inconsistencies in risk assessments now performed by
different groups within EPA, and ii so, why?
• Should we have a more explicit, formal means for resolving
scientific controversy, and tor updating our practice as regards
assessment procedures and assumptions?
These concerns have prompted two decisions: an Agency com-
mitment to develop six risk assessment guidelines: and the creation
of an internal Risk Assessment Forum. The remainder of this
chapter describes work EPA has underway on those guidelines and
the role of the new Forum.
-------
19
D. DEVELOPING RISK ASSESSMENT GUIDELINES
In light of the NAS recommendations for developing risk
assessment guidelines and procedures, we reviewed many of the
technical issues that constitute the components of a risk assessment.
These issues are numerous, diverse and cover a broad spectrum of
potential problems.
To deal with problems like these, the Agency plans to complete
new (or revise existing) guidelines on the following topics:
1. Carcinogenicity
The 1976 Interim Guidelines for Cancer Risk Assessment are
being updated in order to reflect more recent developments in the
scientific data base supporting the guidelines and the cancer
principles in the OSTP documents, and to clarify points that are
unclear or have been unresolved in the existing guidelines.
2. Mutagenicity
Mutagenicity refers to the potential of an agent to induce
alteration in the genetic material of living organisms. These altera-
tions may include point mutations (such as changes in the base
sequence oi'DNA) and structural or numerical chromosomal
aberrations.
The Agency currently evaluates mutagenicity data as a basis for
possible chemical regulation. While this information is most often
used to predict carcinogenic potential, increasingly such data are
being used to predict potential tor the induction of heritable
mutations.
The Agency's existing draft guidelines for mutagenicity risk
assessments describe the tv pes of evidence to be weighed in
determining the potential mutagenicity of a chemical, as well as
quantitative approaches that may be appropriate to the estimation
of human heritable mutation and disease. In preparing the revised
mutagenicity guidelines, the Agency will review and update the
proposed guidelines.
3. Reproductive effects
The male and female reproductive systems and the developing
fetus are potentially sensitive targets for the action of toxic agents.
Developmental toxicity is included in the category of reproductive
effects and pertains to teratogenicity as well as other effects such as
-------
20
to resorbed fetuses, stillbirths, spontaneous abortions and other
congenital dysf unc lions.
Teratogenic effects encompass an extremely diverse set of
impacts that harm the developing fetus and are manifested as
congenital malformations (such as cleft palate), developmental
malformations, or functional malformations (such as nervous
system dysfunction). The Agency has a preliminary document
("Assessment of Risks to Human Reproduction and to Develop-
ment of the Human Conceptus from Exposure to Environmental
Substances," EPA, 1982) outlining major issues for consideration in
assessing both teratogenic and reproductive effects resulting from
exposure to environmental agents. However, this document does
not provide definitive guidelines for assessing such risks to the
developing organism. At least two major issues need resolution:
• Existence of a threshold
Unlike carcinogens or mutagens, which may be presumed to
ha\e some finite probability of an effect at any dose, no matter how
small, some teratogens mav have an effect only if they exceed a
certain level of exposure. Where thresholds are believed to exist, we
want to be able to define a no observed adverse effects level, with
appropriate safety factors. On the other hand, we also want to
provide technical guidance for determining which assumptions to
make.
• Extrapolation between species
Inherent interspecies differences complicate extrapolation of
animal test data to direct determinations of human risk. For
example, aspirin is a fair!) powerful rodent teratogen and thalido-
mide is a fairly weak animal teratogen. The situation is, of course,
reversed in humans.
4. Systemic effects
In addition to the effects previously discussed in this report,
exposure to toxic substances can lead to adverse effects on various
organs, such as the liver, the kidneys, or the lungs. These effects
cover a broad spectrum of graded tissue responses ranging from
small changes in enzyme levels to severe organ dysfunction and
debilitation or death of the individual. Thus, we need some
mechanism for interpreting data pertaining to the graded effects
observed in a specific organ from exposure to a particular toxic
substance.
-------
21
Available scientific evidence indicates that many systemic toxic
substances may have thresholds, which further complicates risk
assessment. Moreover-, the results of"the calculations for estimating
the threshold can vary depending on which one of the graded
responses is used to represent the disease state. For example, if one
wants to calculate a threshold for kidney effects due to cadmium
exposure, it will make a difference if one looks at the early stages of
the disease (i.e., mild cellular changes) or, assuming the disease
progresses, at the later stages of the disorder (i.e., severe cellular
changes and kidney dysfunction).
Current Agencv draft guidelines for risk assessment of systemic
toxicants onlv address the estimation of "safe" exposure levels, such
as ADIs. Further development of these guidelines will focus on
these key issues:
• Extrapolation of health effects in test animals to health effects in
humans.
• Extrapolation of dose in test animals to equivalent exposure in
humans.
• Comparison between different routes of exposure in test animals
and humans.
• Ranking ot graded toxic effects.
5. Assessment methods for chemical mixtures
Most risk assessments address the health impact of individual
chemicals; the Agency has an increasing need to know the risks
associated with chemical mixtures. The hazardous waste and Super-
fund programs, for instance, deal with exposure to mixtures more
frequently than they do with exposure to single chemicals. We also
know that people are typically exposed not to isolated pollutants,
but rather to a complex, dilute mixture of many substances.
Considering how man) chemicals there are in the environment,
there is a virtually infinite number of combinations that could
constitute potential synergisms or antagonisms. In only a few cases
do we have concrete evidence of what these interactive effects might
be. In the absence of such evidence, we use an additive method that
simply sums individual chemical effects on a target organ. As we
learn more about the effects of chemical mixtures, we may have to
modify the additivity approach. The new guidelines will describe a
hierarchy of procedures for estimating health risk, based on the
-------
22
nature of the available data. The possible data types include: (1)
specific toxicity data on the mixture itself, (2) toxicity data on a
chemically similar mixture, (3) data on interactions of some of the
components of the mixture, and (4) quantitative toxicity data on
single chemicals within the mixture.
6. Exposure assessment
Given the absence of complete environmental data, we must
estimate ambient levels of pollutants or chemical exposures using
the following information sources, singly or in combination:
• measurements of actual tissue concentrations of pollutants;
• field monitoring;
• laboratory modeling data;
• mathematical modeling.
To estimate human uptake of some substance, the first step is
to determine how many people are exposed through the various
relevant environmental pathways — in the air, soil, water, drinking
water, or food. The next step is to calculate the rate of uptake
through breathing, eating, drinking, or absorption through the
skin.
The new exposure guidelines will review, and modify as needed,
the standard factors used in estimating exposure; for humans these
factors include standard body weights, breathing rates, and fluid
and food intakes. In the past, there have been some differences
among programs; these have not been great, but they do contribute
unwanted variation in the results of risk assessments. The guide-
lines will address statistical approaches designed to estimate
the degree of uncertainty associated with different modelling
assumptions.
E. ESTABLISHMENT OF A FORUM ON RISK ASSESSMENT
ISSUES
We have established a Risk Assessment Forum to provide an
institutional locus for the resolution of significant risk assessment
issues as they arise, and to insure that Agency consensus on such
issues is incorporated into the appropriate risk assessment guide-
lines. The Forum will also provide Agency scientists with a regular
time and place to discuss problems of risk assessments in produc-
tion. Peer advice and comment of this type will help improve the
quality of risk assessments, with associated savings in time and
resources.
-------
CHAPTER THREE
GOALS AND APPLICATIONS OF RISK MANAGEMENT
A. DEFINING RISK MANAGEMENT
The NAS has defined the term risk management as the
complex of judgment and analysis that uses the results of risk
assessment to produce a decision about environmental action. The
term was originally meant to distinguish the political, economic,
and social aspects of decision-making from the scientific exercise
involved in the assessment of risk. It has come in the last year or so
to stand for a wider and potentially more useful concept as the
Agency has begun to implement the National Academy of
Sciences definition.
This chapter is about the Agency's application of that wider
meaning. It defines what we hope to gain from the risk manage-
ment approach to environmental protection: what current
problems we are addressing with it and what our goals are. It also
describes some of the changes in Agency procedures, operations,
or objectives that have developed over the past several years or
more recently, as a result of adopting the risk management
approach suggested by the NAS.
If we regard risk reduction as one of EPA's main reasons for
being, then we can define risk management as determining and
accomplishing those actions that will reduce risk to the greatest
degree given any particular level of resources, meaning Agency
resources and those of society in general. The resource considera-
tion is vital here. One can argue about how much should be spent
on environmental protection, but at some point everyone must
accept that the commitment of resources for any social purpose has
a finite limit. If the number of potential risk targets is very large in
comparison to the number we can realistically pursue, which seems
now to be the case, then some rational method of choosing which
risks to reduce and deciding how far we should try to reduce them
is indispensable.
It is important to keep in mind that while individual mk manage-
ment decisions may be seen as balancing risk reduction against resources, the
system as a whole is designed to balance risk against risk. In other words, it
is essential that we address the worst and most controllable risks
first; failure to do so means that the total amount of harm that we
prevent is smaller than the amount we might have prevented.
Making incorrect priority choices, saving one where we might have
saved two, represents a profound failure of the Agency's basic
protective mission.
-------
24
In making such balances, the risk management approach
regards risks of the same type (e.g., risks of a particular disease) as
comparable regardless of the route through which people are
exposed to them. This makes sense because we know that risk may
be transferred around the environment and among environmental
media by natural processes or by pollution control itself, and the
idea is, of course, to reduce the total risk in the whole environment.
In practice, however, this is extremely difficult to do, as EPA
operates under eight major statutes, each directed at a different
form or locus of pollution. These statutes not only establish the
values that the Agency must protect (and these naturally differ
among the statutes), but in the case of risk to human health, they
often appear to direct different approaches to risk reduction.
Briefly, there are two broad groups of statutory mandates to which
any risk management approach must be adapted. In the first (e.g.,
Toxic Substances Control Act), explicit balancing of risks against
benefits or costs of control is authorized or required. When applied
in reference to programs under such laws, risk management is the
analysis and exposition of the balancing considerations.
In the second group (e.g., the Clean Air Act), a standard that
protects human health or some other value must be established or
some particular level of technical control must be applied. Cost
considerations may be specifically prohibited during the develop-
ment of the protective standard. Here risk management means
finding the most efficient way of achieving the standard, while at
the same time assuring that policies designed to remove specific
pollutants under these laws do not have perverse effects, such as
transferring an ecjual or increased risk to another environmental
medium.
Another difficulty arises because we are called upon by most of
our statutes to protect a variety of environmental values, as well as
human health. In general, it is more difficult to quantify risks to
these environmental values (and reductions in such risks by regula-
tory action) than it is to come up with comparable estimates for
human health risks. Protection of environmental values is of
especially great importance in statutes such as the Clean Water Act.
The special attention given to human health risk reduction in parts
of this report is not intended to indicate any less concern for
protection of environmental values, or any diminished intention to
act to protect them; rather, it recognizes that risk assessment and
the analytical approaches for risk management discussed here have
progressed considerably further in their application to human
health risk.
-------
25
B. ELEMENTS OF RISK MANAGEMENT: SETTING
PRIORITIES AND MAKING CHOICES
In operation within the Agency, the risk management ap-
proach has two major ends: setting priorities among the risks
presented by pollution that are amenable to control by EPA; and
choosing the appropriate reduction actions for the risks so selected.
EPA's current risk management approach requires that these
traditional activities be expressed, where feasible, in terms of risk
reduction.
In the case of priority setting, this risk-based management ideally
would enable us to insure that the Agency as a whole had an
agenda of potential activities directed against the worst set of risks
susceptible to its control. Priority-setting, in turn, is important
because historically the Agency's agenda has been set less by
systematic analysis than by direct public pressure in response to the
environmental issues of the day, often embodied in court orders;
diverse legislative mandates; or merely random action. To a certain
degree, this reactive mode of behavior is inevitable. But it would be
of tremendous advantage to the cause of real risk reduction if the
Agency were able to make the case that its assemblage of proposed
risk-reduction targets was a clemonstrably more important one than
any other set.
In choosing control actions, the Agency's discretion regarding the
balancing of risks with other factors varies with the applicable
statutes. Nevertheless, a certain amount of balancing goes on in
virtually every important Agency control decision. Historically, this
kind of judgment has taken place at many different levels and
sections of the Agency; when policies or regulations reached the
final stages it was often impossible to establish in any meaningful
way the nature of the judgments that had gone into them. In
contrast, in the risk management approach it is of the essence that
such judgments be made as early and as explicitly as possible, and
that the whole array of considerations that establish an Agency
decision about controlling some risk be presented in a comprehen-
sible fashion to senior EPA management.
In general, the balancing that goes into such risk management
decisions includes consideration of at least three major compo-
nents. The first is the harmful effect of the pollutant(s) proposed
for control. When the effect is on human health this factor may be
expressed as a numerical risk estimate, but EPA must control man}'
harmful effects that cannot be so expressed. It is important to
remember that the term "risk management" is used broadly
enough to apply to these non-quantifiable effects as well.
-------
26
Particularly difficult issues of value arise in connection with non-
health effects. While we can quantify certain "benefits" that accrue
when such effects are reduced (e.g., increased fishing days, reduced
materials damage), clearly there are some values that defy this
approach, yet are obviously important and, indeed, are built into
the language of much of our legislation. The values include such
considerations as the value of an unused aquifer or of the preserva-
tion of pristine wilderness areas. It is important to remember that
risk management includes makingjudgments about values that do
not involve human health risk and can not be quantified under the
present state of the art.
The effects factor is therefore not a simple one. Besides the many
complexities involved in assessing the extent of exposure and the
severity of hazard (which have been discussed elsewhere in this
report), the balancing decision should consider the distribution of
the effect in terms of how man) people it affects over how wide a
geographic area, the reversibility or persistence of the effect, and
the impact of the decision on the long-term health of ecological
systems.
The second factor may be called "cost," although it is not simple
either. It may include the cost of pollution controls, consideration of
the effects of alternative practices, the relinquished benefits of using
a pesticide or other toxic chemical, the danger of displacing private
sector initiatives, or the impact of some control option on employ-
ment, firms, or communities.
The third factor is essentially a measure of confidence. The
Agency almost always acts under conditions of uncertainty, but that
uncertainty has an enormous range. Similar cost-effect relation-
ships may look very different to the risk manager if they differ
substantially in the weight of evidence tying pollutant to effect, or
control strategy to reduction in risk.
Cost and effect as we have defined them are, of course, related;
examination of that relationship is at the heart of risk management.
Greater reductions in the harmful effect are usually associated with
higher control costs, typically along a curve of declining efficiency.
That is, the last increments of pollution control are far more
expensive than the first. We may find, for example, that it costs as
much to get from 95 to 99 per cent removal of some toxicant as it
did to get from zero to 95 per cent.
There are a number of analytic tools that are aimed at exploring
this relationship, which may be briefly distinguished:
-------
27
Benefit/cost analysis
This approach weighs the costs of control, explicitly and directly,
against the monetized benefits of control — the avoidance of
disease and the attainment of other social goods (e.g., increased
visibility, reduced soiling and materials damage, etc.). Optimal use
of benefit/cost analysis occurs when all factors affected by a decision
can be accurately represented in dollars. This is often difficult to do,
since the Agency is frequently concerned with protecting such
things as human life and the stability of ecosystems, social values for
which there is no market price, or for which current procedures for
finding "shadow prices" are bitterly controversial. In areas where
this is not a constraint (cost of control vs. avoided crop or materials
damage, for example) benefit/cost analysis provides a structured
way to balance effects and costs directly.
• Risk-benefit analysis
Risk-benefit analysis balances the economic benefits of a
polluting activity against the associated risks to health and the
environment. For example, the benefits ofmmga pesticide (e.g.,
the value of the increased crop yield minus the application cost) are
explicitly weighed against the risks generated by the pesticide's use.
Note that benefits do not refer to the benefits from regulation, but
rather to the benefits from the use of the chemical. Because risks
are not reduced to commensurate units (dollars), risk-benefit
analysis is most appropriate when the Agency must balance one or
two types of risk against the economic benefits of an economic
activity.
• Cost-effectiveness analysis
Unlike the other two approaches, cost-effectiveness analysis
begins by accepting the desirability of a particular control action. It
does not weigh risks against benefits, or monetize benefits; it only
looks for the least-cost path to achieve a given goal, such as the
achievement of a protective standard. For example, if a number of
controls are available to remove some pollutant from the atmo-
sphere down to a certain pre-determined level, the cost-effective
solution is the one that does this most cheaply.
Cost-effectiveness analysis is at present the most frequently used
risk management tool at EPA, since so much of the Agency's work
is involved with implementing pollution standards. It is straight-
forward in application: in a simplified version, for example, one
calculates the cost-per-ton removed associated with the available
-------
28
options and, all other things being equal, picks the lowest. But cost-
effectiveness analysis can also be used to compare different ways of
obtaining some specified degree of risk reduction. An integrated
method for doing this is summarized later in this chapter.
While these types of analysis could be part of any particular risk
management exercise, it is important to note that risk management
does not, as some critics have implied, demand the inappropriate
monetization of the social values the Agency is charged with
protecting. Risk management in the EPA sense is the expression of
the value of the societal and governmental expenditure represented
by an environmental control action. The value expressed could be
relatively easy to quantify (e.g., reduction in materials damage or
cases of particular diseases) or difficult — protection of sensitive
ecosystems or future groundwater use. Risk management is a way
of explaining the logical connections between a body of research,
the application of certain economic, political and social values, and
the achievement of some environmental result.
Inherent in risk management is the idea of comparability. The
Agency has a number of goals, some of which may conflict. For
example, deep ocean dumping of sewage sludge may reduce
human health risk in comparison with incineration or land spread-
ing, but may have adverse effects on marine ecosystems, which are
valued in their own right, and on the human food chain. Assigning
resources on the basis of the varying importance attached to the
attainment of different goals, and coordinating efforts that are
driven by apparently conflicting goals, are both susceptible to a risk
management approach. Indeed, it is hard to see how they could
effectively be done otherwise.
Also inherent in risk management is the principle of consistency.
Since pollution control (and hence risk reduction) is an incremental
process, with the later increments typically costing more to achieve
than the earlier ones, the Agency may be faced with a number of
potential actions with widely differing marginal costs for the same
or similar risk reduction. It does not make sense to buy dear what
you can get cheap.
Marginal cost consistency, however, may conflict with another
sort of consistency of result. We may want to please an absolute limit
on the risk experienced by any particular individual from environ-
mental contaminants, or, in cases where residual risks are unavoid-
able, we may want to achieve a consistent level of post-regulatory
risk in all control actions.
-------
29
But proposals to establish absolute regulatory levels, however
attractive they may be in terms of rational management, are
constrained by a sense of the limits of quantification methods.
Some important things cannot be quantified, but are nonetheless
real. The Agency will not be allowed to undertake risk manage-
ment if the public does not trust our response to their perceived
concerns. Strictly quantitative models usually do not make allow-
ances for such imponderables as public confidence; this is why risk
management at EPA is not just numbers. And, of course, we will
always be limited by the uncertain nature of environmental re-
search and economic data, and we wish to take care not to read into
such numbers more precision than their origins warrant.
Although we feel that the movement toward greater quantifica-
tion of environmental decision variables is a good trend, the
limitations noted above make it unwise to establish formal trigger
points for Agency actions. It is not appropriate, for example, to
settle on a single level of risk that would be required before we
would consider regulatory action. Measures of consistency of result,
such as marginal cost per case avoided, are useful guides; they
should not be made into rigid grooves that might deprive the
Agency of the flexibility it needs to carry out its complex missions.
On the other hand, consistency oCajtpwafh in making decisions
based on risk, cost, and uncertainty is essential. The management
of EPA needs to know hew the actions of the various programs
actually work all togel/in to reduce the harm clone b) pollutants.
Management should also have the opportunity to compare the
relative impact of the programs.
In order to determine current risk management approaches
the Toxics Integration Task Force carried out a study of how the
\arious considerations that make up risk management were used in
27 recent Agency regulatory actions. The study showed significant
variations across programs in the way risk management was actually
carried out. As mentioned earlier, risk management implies some
balancing of values; the statutes differ in the way they direct us to
balance values, partic ularly in the extent to which control costs may be
considered in establishing allowable or "safe" levels of a pollutant.
Part of the difficulty in comparing risk management across
programs arises because risk reduction does not even appear as an
explicit concept in several of the Agency's statutes. Thus, when
applied to a program that protects mainly environmental values via
technology-based standards, such as the Effluent Guidelines Pro-
-------
gram, risk management means something different from what it
means in connection with a program (like the Hazardous Air
Pollutants Program) focused on human health protection. Similarly,
it is easier to present risk management decisions in the case of a
national program to control a single substance than when, as in the
Solid Waste Program and Superfund, we control complex waste
streams. Further, programs responsible for cleaning up wastes (e.g.,
Superfund) or for controlling useful substances that are poisonous
(pesticides and toxic industrial chemicals) have ways of expressing
risk management information that are different from those used by
programs that impose pollution controls in the usual sense.
Differences in mandate and program structure, however, do not
excuse the Agency from developing consistent approaches in the
areas of risk management to which the statutes are silent. The
retrospective study found that the Agency as a whole had no
generally accepted way of expressing the degree of confidence in
the pollutant-effect connection or of dealing with intermedia
transfer of risk. Moreover, despite their historic differences in
approach, EPA programs are part of a single national effort
embodied in a single Agency. The Agency in turn must respond to
a basic requiiement of good public policy: to establish the contieitwn
between wme e\/)eiifliline tnitl wme mogiiized fniblu good. Most regula-
tions cost something, whether expressed in dollars spent by
industry or in terms of the consequences of doing without a useful
product. All regulations use up Agency resources. Agency manage-
ment needs to assure that the total of Agency and societal resources
de\oted to the prevention of harm to human health and the
environment is being applied efficiently.
C. CURRENT RISK MANAGEMENT INITIATIVES
We understand the difficulties ol applying risk management
principles in an Agency as complex and as variously mandated as
F.PA. The risk management effort at the Agency therefore consists
of a wide range of activities affecting nearly every aspect of the
Agency's work. These activities fall into three broad classes. First, we
want to obtain a better and more consistent information base for
making dec isions about the control of risk. Second, we want to use
the various analytic methods associated with risk management
whenever appropriate in developing environmental policy; we also
want to plac e more emphasis on figuring out what we have achieved
in terms of risk reduction through past efforts and on locating and
efficiently managing the serious risks remaining. Third, we must
communicate to the public what we are doing. v\hv we are doing it
in risk management terms, and how the risk management ap-
proat h \\ ill improve the vvav that F.PA carries out its mission.
-------
31
/. Building the Information Base
No matter what risk management techniques are appropriate
to a particular program (and these may legitimately vary), there
should be a uniform way of reporting and recording these
decisions. We have therefore developed a uniform format for
presenting a summary of risk management information to KPA
management, including the Steering Committee (which oversees
the development of major rules), staff and program Assistant
Administrators, the Deputy Administrator, and the Administrator.
The risk management format solicits information regarding expo-
sure, intermedia transfer, and the benefits and costs of regulation.
Space is provided in the form to record both health risk reduction
and "benefits other than risk reduction," sue h as prevention of
damage to ecological systems, preservation of endangered species,
or other factors.
We believe that using the form within the regulatory develop-
ment process offers the following advantages:
— It will provide data for analyses of the variations in risk
management practices across programs much more clearly;
it will also note where these variations are necessary and
appropriate.
— Using the form may help sharpen our perception of what the
underlying technical problems are in evaluating complex
risks, and how they ought to be presented for decisions.
— Bv documenting the expected nature and extent of pre- and
posl-regulatorv risks, it should be possible to increase consis-
tency here. Uniformity, however, is not the goal. Variations
among statutes and individual circumstances require that this
area be kept flexible.
— Bv documenting intermedia transfer, the format should im-
prove our abilitv to track risk in the environment as a whole.
We have also instituted an Options Selection Process as a way of
providing consistent polio guidance from senior Agency manage-
ment throughout the entire course of regulators development. For
this process, even important regulation or other policy-setting
action the Agency takes is evaluated at an earlv stage of its
development by the Deputv Administrator. Review criteria include
level of environmental risk reduction, net benefit to society, flexibil-
ity, and propensity to encourage emiionmentally acceptable
innovation. The Deput\ Admimstiator and other senior EPA
-------
managers can review not only the particular regulatory option that
a program recommends, hut also the other options considered, the
applicable decision rules, the data on which the recommendation is
based, and the rationale for selection. The Options Selection
Process should ultimate!) increase the efficiency of EPA's operations
In involving senior EPA management from the very beginning of
regulator) development.
2. Using Risk Management Tools
The purpose of the analytic tools used in risk management is,
naturally, to help determine the most efficient way to reduce risk,
which implies the establishment of the risk (or effect) measure and
analvsis of the impact some risk-reducing action has on other social
values. This often represents a departure from the past analytic
practices.
• Benefits Assessment
As the Agency turns its attention increasingly toward controlling
toxic chemicals, it is important to do more than simply calculate the
magnitude and distribution of costs for each alternative considered.
We have to evaluate benefits as \\ell as the costs, identifying what
new controls are buying in the way of additional health or environ-
mental improvements. Onh b\ expressing all potential actions in
terms of what will result from them — broadly speaking, their
risk-reduction effect — can we select those actions thai benefit the
environment to the greatest degree. Choosing intelligently among
risk targets in this way is not possible without some way of
estimating benefits.
Estimating benefits is generalK more difficult than estimating
costs, ft requires more extenshe data and more sophisticated
techniques. \Ve have to reh on often sparse information from such
areas as risk assessment, dispersion modeling, and other scientific
fields, coupling this with imperfect information about how the
public values the main- factors that figure in environmental deci-
sions. Despite the difficulties, however, EPA has made significant
progress in this area over the last several years, and we believe that
the effort pa\s off in more consistent and defensible decisions.
Some of the advantages of using benefits assessment are made
clear by the recent example of work done on lead in gasoline. Our
anahsis included a comprehensive assessment of the benefits that
further regulation would provide.
-------
The cost/benefit analysis produced the following findings:
• Our analysis and other major studies both in the U.S. and abroad
indicated that the amount of lead in blood is directly related to
the amount of lead in gasoline. Children with elevated blood
levels require medical monitoring and sometimes treatment. The
costs of these, not to mention the pain and suffering incurred,
are substantial.
• Recent EPA surveys indicate that over 12% of all cars equipped
with catalytic converters to control auto emission are currently
being "misfueled" with leaded gasoline to take advantage of
cheaper leaded gas or to obtain higher octane. Misfueling
poisons catalysts and substantially increases other auto pollutant
emissions. Given current misfueling rates, misfueled vehicles will
account for one-third of leaded gasoline demand in 1988,
significantly increasing our estimates of future lead and other
pollutant emissions. The impact of these unanticipated emissions
on public health and welfare is substantial, and can be quantified.
• Lead forms corrosive compounds that increase automotive main-
tenance costs. Cars that use leaded gasoline need more frequent
tune-ups, exhaust system replacements, and oil changes. These
too involve costs that are measurable for the country as a whole.
The analysis showed that economic impacts described above are
substantially larger than the increased costs of producing gasoline
with reduced lead content, even though many of the projected
health impairment effects were not monetized.
• Cost-effectiveness analysis
At a more general level, we have begun to develop an integrated
methodology for applying risk-based cost-effectiveness analysis to
regulatory decisions. This Integrated Environmental Management
approach is the first attempt to analyze, in quantitative terms, the
regulatory inefficiencies created by the Agency's decentralized
structure, and to propose specific remedies. To do this it estimates
the amount of health and environmental risk reduction that has
been bought for each regulatory dollar that an industry has spent.
It keeps track of the cumulative costs of regulations of specific-
industries, the relative importance of different regulations in
reducing risks, and shifts of pollutants among environmental
media. The ultimate goal of these integration studies is to find the
combination of controls that achieves the most risk reduction for
any specified cost or the leastcost way of arriving at any specified
level of risk reduction.
-------
34
Integration studies attempt to bring together all the useful
information (e.g., data on production and processes, emissions and
effluent levels, and costs of controls) that EPA possesses on a
particular pollution source or set of sources. These could be
industrial sec tors, publicly-owned sewage plants, or combinations of
sources in particular geographic areas. The methodology then uses
exposure models to trace pollutants through air, water, ground
water, and land, so as to estimate human and environmental
exposures in particular places or in modelled environments repre-
sentative of where plants in particular industries are located.
Using research on health and environmental effects associated
with these pollutants, it then estimates the effects that may be
generated as a result of such epxosures, and combines these data in
a decision model similar to those used elsewhere in industry and
government to solve complex strategic problems. The transfer of
pollutants among media is tracked and calculated throughout the
analysis. The result is a set of specific cost-effective control strategies
for a geographic area or for some c haracterizable source such as an
industry, based on particular environmental conditions.
One of the major findings of this work is how radically the
relative cost-effectiveness of many controls changes when one
modifies important environmental variables, such as population
distribution and density, topography/meteorology, the volume of
receiving waters and so forth. The lesson here is that risk manage-
ment is eventually going to require looking more closely at the
projected effects of pollutants where the pollutants are actually found.
• Managing for environmental results
EPA has always experienced some difficulty in relating its actions
to the actual reduction of risk, (liven the language of the statutory
mandates, programs have typically placed more emphasis on
mandating emission or effluent controls than on determining what
effect such controls had on the actual environment. This approach
was, of course, indispensible in the early days of environmental
protection. The need for controlling gross pollution was so pressing
that the Congress opted for an easily administered and implemen-
table set of regulatory tools.
At present, however, with ten or so years of pollution control
behind us, fixing the precise nature of the remaining problems is
much more difficult. Also, we have created an enormous apparatus
tor controlling pollution; as we add more increments to it, it
would be well to determine its impact on the values the Agency is
mandated to protect. These reali/ations have led to a general policy
-------
we call managing for environmental results, which may be consid-
ered the business end of the risk management approach.
An immediate outcome of this policy is the requirement that
programs include in their budget submissions statements about
what environmental results requests for funding are designed to
achieve. In addition, the Agency is implementing a national
monitoring strategy designed to meet the full range of current and
future needs for environmental data, so that we can effectively tie
our policies to actualimprovements in environmental quality. We
are also beginning to shift our enforcement priorities toward
compliance measures based on actual risk reduction, rather than
simply examining the numbers of court referrals or administrative
orders.
3. Strengthening the Role of Communication in Risk Management
This document has touched in a number of places on the
importance of communication in risk management. The point
cannot be made too often. In one sense, risk management w a form
of communication. Technical analysis of the costs and benefits of a
proposed action is not a device for coming up with the "right" or
"rational" answer; all such analyses are far too sensitive to subjective
values and far too dependent on uncertain data for us to pretend
that they are. Risk management, and the technical analysis that
contributes to it, is largely the exposition of the information we
believe is reliable, the values we wish to apply and the way that
these two are linked to produce a set of policies. The information
is derived from the best science we can find or commission although
it is important to note that even the scientific parts of a risk man-
agement decision are touched with value considerations and are
generally not testable under ordinary scientific canons. The values
are derived first of all from our statutory charters and the judicial
interpretations that have grown up around them and from the
legitimate exercise of judgment and choice by politically appointed
Agency managers.
Obviously, not everybody will agree with the values so expressed,
but in order for the debate about values to begin and for the
democratic processes that ultimately establish values to take place,
everyone has to know what the values underlying our decisions
really are. If they are hidden behind a facade of "rationality" the
debate is never joined. The frank expression of our values, of our
uncertainties, and of the trade-offs involved in every important
decision about environmental protection is itself a central value of
EPA. It is this value that drives the set of current practices and
policy initiatives we are calling risk management.
------- |