United States Science Advisory EPA SAB-EC-90-021B
Environmental Protection Board September 1990
Agency (A-101) , ,
~~ '
&EPA The Report Of
The Human Health
. AGENCY
Subcommittee <»
LIBRARY
Relative Risk
Reduction Project
-------
NOTICE
This report has been written as a part of the activities of
the Science Advisory Board, a public advisory group providing
extramural scientific information and advice to the Administrator
and other officials of the Environmental Protection Agency. The
Board is structured to provide balanced, expert assessment of
scientific matters related to problems facing the Agency. This
report has not been reviewed for approval by the Agency and, hence,
the contents of this report do not necessarily represent the views
and policies of the Environmental Protection Agency, nor of other
agencies in the Executive Branch of the Federal government, nor
does mention of trade names or commercial products constitute a
recommendation for use.
-------
ABSTRACT
The Human Health Subcommittee of the Relative Risk Reduction
Strategies Committee (RRRSC) of the U.S. Environmental Protection
Agency's Science Advisory Board (SAB) reviewed the Agency's 1987
report entitled "Unfinished Business: A Comparative Analysis of
Environmental Problems." (UB ) The Subcommittee's goal was to
evaluate the report's methodology for ranking environmental health
problems, determine the extent to which the risk rankings for
different environmental problems should be revised or updated,
combine if possible, rankings for carcinogenic and non-carcinogenic
effects into a single aggregate ranking, and recommend approaches
for the improve methodologies for assessing and ranking environ-
mental risks to human health. The Subcommittee was critical of
the original EPA ranking of problem areas which included a mixture
of specific environmental pollutants, sources of pollutants,
exposure media, and exposure situations—and which appeared not to
have been selected on the basis of their relevance to environmental
and health hazard assessment, or on the basis of overall public
health significance. Most of the 31 categories are so broad, and
include so many toxic and non-toxic agents, that ranking of these
categories could not be performed with any rigor or confidence.
Problems areas in the UB report representing proximal human
exposure situations were assigned the highest relative risk
rankings for cancer and/or other adverse health effects. Of the
"high" relative risk rankings assigned in the UB report, those for
criteria air pollutants, hazardous air pollutants, indoor radon,
other indoor air pollution, drinking water pollutants, the
application of pesticides, and occupational exposure to chemicals
were considered to be supported more firmly by the available data
than were the rankings for the others:
Future efforts should focus on broad environmental problems,
without regard to internal organizational strictures or to ultimate
regulatory responsibility. The Subcommittee recommends a new
approach to the risk ranking process, using a matrix based on
sources, exposure situations, agents, and health outcomes. This
approach will identify specific agents and mixtures (and the
principal sources and exposure situations in which they are found)
that should receive priorities for applying risk reduction efforts.
The Subcommittee further recommends that the Agency assign a
specific management focal point for this effort to assure
accountability, establish a risk assessment framework for other
toxicants consistent with that used for carcinogens, establish a
formal mechanism for risk anticipation, expand long-range research
on the assessment of human exposure,and improve the relevant
toxicological science base.
Kev Words; environmental health risk assessment; exposure
assessment; risk ranking; toxicological assessment
ii
-------
Human Health Subcommittee Roster
Chairman
Dr. Arthur Upton, Director of the Institute of Environmental
Medicine, New York University, New York, New York
Members
Dr. Julian Andleman, Professor of Water Chemistry, University
of Pittsburgh, Pittsburgh, Pennsylvania
Dr. Patricia Buffler, Director of the Epidemiological Research
Unit, University of Texas, Houston, Texas
Dr. Paul Deisler, Visiting Executive Professor, University of
Houston, Houston, Texas
Dr. Howard Hu, Assistant Professor of Occupational Medicine,
Brigham & Women's Hospital, Harvard University Medical
Center, Boston, Massachusetts
Dr. Nancy Kim, Director of the Division of Environmental
Health Assessment, New York Department of Health
Albany, New York
Dr. Morton Lippmann, Professor, Institute of Environmental
Medicine, New York University, Tuxedo, New York
Dr. Roger McClellan, President, Chemical Industry Institute of
Toxicology, Research Triangle Park, North Carolina
Dr. Arno Motulsky, Professor of Medicine and Genetics,
University of Washington School of Medicine
Seattle, Washington
Dr. Frederica Perera, Associate Professor of Public Health
Columbia University, New York, New York
Dr. Jonathan Samet, Professor of Medicine, University of New
Mexico, Albuquerque, New Mexico
Dr. Ellen Silbergeld, Senior Scientist, Environmental Defense
Fund. Washington, DC
Dr. Bernard Weiss, Professor of Toxicology, University of
Rochester Medical Center, Rochester, New York
iii
-------
Dr. Hanspeter Witschi, Associate Director of the Toxics
Program, University of California, Davis, California
Designated Federal Official
Mr. Samuel Rondberg, Science Advisory Board Staff
Environmental Protection Agency, Washington, DC
IV
-------
TABLE OF CONTENTS
1.0 Executive Summary 1
1.1 Introduction 1
1.2 Evaluation of Methodology 1
1.3 Comments on the Risk Rankings 4
1.4 Developing An Aggregate Risk Ranking 6
1.5 Recommended Approaches 6
2.0 Introduction 11
2.1 Background 11
2.2. Charge to the Human Health Subcommittee 13
2.3 Format of this Report 13
3.0 Essential Elements in Assessment of Environmental Risks
to Health 14
3.1 Overview 14
3.2 Assessment of Exposure 14
3.2.1 Data Gaps and Uncertainties 15
3.2.1.1 Specific Chemicals 15
3.2.1.2 Concentrations 15
3.2.1.3 Nature of Exposures 17
3.2.1.4 Ranges and Variabilities of
Exposure 17
3.2.1.5 Exposure to Complex Mixtures .... 19
3.2.2 Summary and Recommendations 19
3.3 Assessment of Toxicity 20
3.3.1 Hazard Identification 20
3.3.2 Dose-effect Characterization 22
3.3.2.1 Defining the Dose 24
3.3.2.2 Defining the Response 25
3.3.2.3 Defining Dose-response
Relationships 27
3.3.2.4 Summary 30
3.3.3 Assessment of Severity of Impact 32
3.3.3.1 Introduction 32
3.3.3.2 Impacts on Individuals 33
3.3.3.2.1 Exposure Status 36
3.3.3.2.2 Disease Status 37
3.3.3.2.3 Functional Status 38
3.3.3.2.4 Welfare Effects 33
3.3.3.2.5 Functional Effects 38
3.3.3.3 Impacts on Populations 39
3.3.3.4 Synthesis 39
3.3.4 Susceptible/Critical Subgroups 40
3.3.4.1 Introduction 40
3.3.4.2 Types of Susceptibility Variations . 41
3.3.4.2.1 Biological Variations in
Susceptibility 41
-------
3.3.4.2.2 Susceptibility Variations Due
to Social or Behavioral
Factors 43
3.3.4.3 Identifying Susceptible Subgroups
According to Hazard of EPA "Risk
Area" 45
3.4 Treatment of Uncertainty 45
3.4.1 Parameter Uncertainty 46
3.4.2 Model Uncertainty 46
3.4.3 Uncertainty Due to Inter-individual
Variability 46
3.4.4 Uncertainty in Quantifying and Comparing
Measures of Risk 46
4.0 Reducibility of Environmental Risks to Health 48
5.0 Review of The Health Risk Rankings in The "Unfinished
Business" Report 49
5.1 Methodology 49
5.2 Rankings for Risks of Cancer 49
5.2.1 Criteria Air Pollutants 51
5.2.2 Hazardous Air Pollutants 51
5.2.3 Other Air Pollutants 52
5.2.4 Indoor Radon 52
5.2.5 Indoor Air Pollutants Other Than Radon ... 52
5.2.6 Drinking Water 52
5.2.7 Pesticide Residues on Foods 53
5.2.8 Application of Pesticides 53
5.2.9 Worker Exposure to Chemicals 54
5.2.10 Consumer Product Exposure 55
5.2.11 Radiation Other Than Indoor Radon 55
5.2.12 Depletion of Stratospheric Ozone 56
5.2.13 Hazardous Waste Sites 56
5.3 Rankings for Risks of Adverse Effects Other Than
Cancer 56
5.3.1 Criteria Air Pollutants 57
5.3.2 Hazardous Air Pollutants 58
5.3.3 Indoor Radon 58
5.3.4 Indoor Air Pollution Other Than Radon ... 58
5.3.5 Drinking Water 58
5.3.6 Pesticide Residues on Foods 59
5.3.7 Application of Pesticides 60
5.3.8 Worker Exposure to Chemicals 60
5.3.9 Consumer Product Exposure 61
5.3.10 Radiation Other Than Indoor Radiation ... 61
5.3.11 Depletion of Stratospheric Ozone 61
5.4 Merging of Cancer and Non-cancer Risk Rankings . . 62
6.0 Approaches for The Long-term Development of Improved
Risk Assessment Strategy 65
6.1 Alternative Models for Risk Reduction Targets ... 65
vi
-------
6.2 Identification and Assessment of Specific
Toxicants 73
6.2.1 Selection of Specific Pollutants 74
6.2.2 Addressing Exposure Parameters 74
6.2.3 Summaries and Lessons Learned from the Case
Studies 75
6.2.3.1 Ozone 75
6.2.3.2 Radon 78
6.2.3.3 Overall Lessons 78
6.3 Ranking Schemes 79
6.3.1 General Considerations on Ranking and
Severity 80
6.3.2 Producing a Merged Health Risk Ranking: the
Zero-Based Approach 83
6.3.3 Producing a Merged Health Risk Ranking:
Merging Separate Rankings into One .... 84
6.3.4 Further Comments and Recommendations .... 88
6.4 Development of Necessary Resources 88
7.0 Conclusions and Recommendations 91
8.0 Appendices
8.1 Case studies
8.1.1 Ozone Case Study—Dr. Morton Lippmann 97
8.1.2 Radon Case Study—Drs. Arthur Upton, J. Samet and
J.Andleman 118
8.2 Ranking Schemes
8.2. Detailed derivation of Rank-merging—Dr. Paul Deisler 132
9.0 References 161
vii
-------
LIST OF TABLES AND FIGURES
Table 2.1
Table 5.2
Table 5.3
Table 5.4
Table 6.1.1
Table 6.1.2
Table 6.1.3
Figure 6.1.1
Figure 6.1.2
Figure 6.3.1
Table 6.3.1
Figure 8.1.2.1
Table 8.1.2.1
Table 8.1.2.2
Table 8.1.2.3
Table 8.1.2.4
Figure 8.1.2.2
Figure 8.1.2.3
Figure 8.1.2.4
Figure 8.1.2.5
Figure 8.1.2.6
Figure 8.1.2.7
Table 8.1.2.5
Figure 8.1.2.8
Table 8.1.2.6
Table 8.1.2.7
Table 8.1.2.8
Figure 8.1.2.9
Figure 8.2.2.1
Figure 8.2.2.2
Figure 8.2.4.1
Figure 8.2.4.2
Table 8.2.6.1
Table 8.2.6.2
Table 8.2.7.1
Figure 8.2.7.1
Table 8.2.8.1
Table 8.2.9.1
Table 8.2.9.2
"UB" Problem Areas 12
"UB High/Medium cancer rankings 50
"UB" Non-cancer risk rankings 57
"UB" Problems grouped by exposure/source
and risk rankings 63
Source/Exposure matrix 66
Source Terms/Vectors 67
Exposure Terms, Table 6.1.1 68
Three Dimensional matrix 69
Three Dimensional matrix showing sources,
exposures, and agents 70
Risk ranking plot for cancer vs non-cancer 85
Rankings for 3X3 linear array 86
Distribution of radiation sources 119
Underground miner mortality 119
Radon concentrations and lung cancer 122
Radon concentrations and lung cancer 122
Distribution of radon in U.S. homes 122
Distribution of radon in U.S. homes 123
Distribution of radon in N.J. homes 123
Distribution of radon in homes by season
and geographical location 123
Radon concentration vs. ventilation rates 124
Radon concentration vs. ventilation rates
in the San Francisco area 124
Distribution of radon concentrations vs.
ventilation rates 124
Population-weighted averages for radon in
drinking water 125
Occurrence of radon in drinking water 126
Life-time lung cancer risk 128
Factors affecting tumorigenic potential
of radon daughters 130
Life-time lung cancer risks from radon
exposure 130
Lung cancer attributable to radon daughters 130
Projecting a grid square (linear) 136
Projecting a grid square (non-linear) 137
Linear array of nodes 141
Non-linear array of nodes 142
Possible rankings, linear 3x3 array 145
Possible rankings, non-1inea 3x3 array 146
High, Low, and Medium rankings for "UB"
Problem Areas 149
Actual Problems 150
Problems in range, by cancer risk 151
Selected rankings for consideration 152
Hypothetical merged risk ranking 153
viii
-------
1* o Executive Summary
1.1 Introduction
This is the report of the Human Health Subcommittee of the
Relative Risk Reduction Strategies Committee (RRRSC), convened by
the U. S. Environmental Protection Agency's Science Advisory Board
(SAB). The report was written as part of an overall effort by the
SAB to assist in developing strategic risk reduction options that
would be helpful to the Agency in assessing its research and
regulatory activities.
In conjunction with other studies undertaken by the RRRSC, the
Subcommittee was charged with reviewing EPA's report entitled
"Unfinished Business" (EPA, 1987) to: (1) evaluate its methodology
for ranking environmental problems in terms of their relative risks
to human health, (2) determine the extent to which the relative
risk rankings it had assigned to different environmental problems
should be revised or updated on the basis of methodological
limitations or newer data, (3) combine if possible into a single
aggregate ranking the risk rankings for carcinogenic effects and
the risk rankings for other adverse effects on human health, and
(4) recommend approaches for the further development of a long-term
strategy to improve the methodology for assessing and ranking
environmental risks to human health. Given the breadth of the
charge, the Subcommittee focused its attention on methodological
and research issues, with the intent of providing recommendations
to a future expert group convened specifically for the purpose of
ranking relative environmental health risks.
1.2 Evaluation of Methodology
Toxicants that may be encountered in air, water, food,
consumer products, the home, the workplace, and other environments,
can pose risks to human health. In some instances, the risks from
such toxicants have already been adequately controlled by limiting
human exposure to the agents in question, but in other instances
environmental toxicant-related risks to health continue to exist,
as reported in "Unfinished Business." The Subcommittee agrees that
it is important therefore, that all such risks be assessed in order
that appropriate measures for controlling them may be developed.
-------
In the "Unfinished Business report 'B), 31 environmental
problem areas were identified and ranked according to the relative
magnitude of the risk of cancer or other adverse health effects
associated with each. No attempt was made to combine the rankings
for cancer with those for other adverse health effects.
On reviewing the "Unfinished Business" report, the
Subcommittee recognized the Agency's need to compare the relative
risks of different environmental problems in order to set
appropriate priorities for the allocation of its resources. The
Subcommittee also recognized that the 31 specific environmental
problems considered in "Unfinished Business"—which included a
mixture of specific environmental pollutants, sources of
pollutants, exposure media, and exposure situations—had been
selected largely on the basis of their relevance to the Agency's
legislative history and programmatic organization rather than on
the basis of their relevance to environmental and health hazard
assessment, or on the basis of overall public health significance.
Consequently, most of the 31 categories in the UB taxonomy are so
broad, and include so many toxic and non-toxic agents, that ranking
of these categories cannot be performed with any rigor or
confidence.
Future EPA efforts should focus more on broad environmental
problems, without regard to internal organizational strictures or
to ultimate regulatory responsibility. To conceptualize risks
better, the Subcommittee recommends a new approach to the risk
ranking process, using a matrix based on sources, exposure
situations, agents, and health outcomes. This approach will
identify specific agents and mixtures (and the principal sources
and exposure situations in which they are found) that should
receive priorities for applying risk reduction efforts.
Among the most serious of the limitations in the risk
assessments in UB was the inadequacy of the exposure information on
which they had been based. Without more adequate characterization
of the human exposure relevant to the environmental agents or
situations in question, the corresponding risk assessments will
remain tenuous. Consequently, the UB report was based, per force.
on a foreshortened hazard identification process. Even today, the
relevant exposure information is fragmentary or lacking, for the
most part. Measures for improving the assessment of exposure
should be pursued vigorously.
-------
Human or animal data that can be extrapolated to the low dose
domain in order to support risk assessment is available for only a
relative few environmental agents. In these cases, moreover, the
extrapolations are often based on incomplete or inconsistent data
and therefore involve uncertain assumptions about the shapes of the
dose response curves, the influence of age and other factors on the
susceptibility of the exposed persons, and the extent to which the
effects of the agent or situation may be modified by other
environmental variables.
Other limitations noted in the UB methodology include:
a) The report was based on a fundamental and largely un-
defined hazard identification process, which relied
heavily on preexisting listings of candidate problems,
instead of a systematic and exhaustive effort to identify
all relevant hazards according to clearly stated criteria.
b) Lack of comparability in the risk estimates for different
exposure and source categories or "problem areas" (as
defined in the UB report), because the estimates were
frequently based on different models and/or assumptions.
c) The frequent use of only a few agents or exposures to
estimate risk for a problem area in which many agents or
exposures were involved
d) The exclusion of significant factors from the selection
of risk areas, e.g., economic or technical controllability
of the risk
e) As acknowledged in the UB report, the failure to state the
scope that specific problems would pose without the
continuation of in-place control and regulatory ac-
tivities. Consequently, some problem areas appeared to
pose relatively low risks precisely because of existing
high levels of effort devoted to their control.
f) The failure to incorporate the assessment of preclinical
and subclinical effects of environmental agents into the
relative risk rankings, which undercut the ultimate goal
of risk prevention.
-------
g) The failure to consider the relative magnitude of ad-
additional benefits to be gained from completing partial
programs to reduce risks of specific toxicants,
particularly when the major expense of changing production
or use patterns had already been incurred and the marginal
costs of further risk reduction were considerably reduced
(e.g., removing the last lead from gasoline; banning PBBs
as well as PCBs) .
1.3 qomflianta on the Risk Rankings
Although the UB report was an important initial effort to
systematize a comparison of environmental problems, the risk
rankings presented must, because of the limitations noted above, be
regarded as provisional.
For want of sufficient time, the Subcommittee did not attempt
to update or reassess the rankings. Rather, the Subcommittee
focused on methodological issues inherent in a risk comparison
exercise of this type, as well as on the need for updated and
expanded databases to improve relevant human exposure and toxicity
information. As shown in Table 6.1.1 and discussed below, the
Subcommittee recommends a restructuring of the environmental
problem areas in the UB report in a way that can more accurately
reflect the different risk factors represented in each area and the
interrelationships among them.
Given the limitations in the taxonomy of the environmental
problems areas in the UB report and in the toxicity and exposure
data on which their respective risk assessments were based, it is
not illogical that those problem areas representing proximal human
exposure situations were assigned the highest relative risk
rankings for cancer and/or other adverse health effects in the UB
report. Such problem areas included the following: criteria air
pollutants, hazardous air pollutants, the application of
pesticides, indoor air pollution (excluding radon) , indoor radon
exposure, drinking water, pesticide residues on food, consumer
product exposure, and occupational exposure to chemicals.
Of the "high" relative risk rankings assigned in the UB report
to the above nine problem areas, those for the following seven
areas were considered to be supported more firmly by the available
data than were the rankings for the others:
-------
• criteria air pollutants
• hazardous air pollutants
• indoor radon
• indoor air pollution (excluding radon)
• drinking water pollutants
9 application of pesticides
a occupational exposure to chemicals
The data for the other two problem areas—pesticides residues
on food, and consumer product exposure—were less robust, but the
"high" relative risk rankings for these problems also might prove
to be justified on the basis of further study.
Depletion of stratospheric ozone (problem No. 7) was ranked
high for cancer effects and medium for other adverse effects in the
UB Report. The Subcommittee considers the supporting data for the
categorization of this particular problem to be less robust than
for those noted just above, but still sufficient to support the
classifications given. It should be emphasized, however, that if
the methodology for assessing relative risk that is proposed in
this report were applied to all other problem areas (or their
component toxicants) identified in the CB Report, certain other
areas might also be classified as "high." Conversely, the
classification of some areas noted above as "high" in the UB report
might possibly be changed to "medium" or "low."
In addition to the relative magnitudes of the risks to health
posed by different environmental problem areas, the controllability
of the risks is another factor that must be considered in
evaluating alternative risk-reduction strategies. Hence it must
not be forgotten that the adverse health effects of certain
environmental toxicants—such as carcinogens—may not appear until
decades after exposure, with the result that termination of
exposure to the toxicants does not abolish the risk for those who
have been previously exposed. Also, certain environmental
toxicants—such as heavy metals, PCBs, and long-lived radionuclides
—tend to persist indefinitely in the environment, and may actually
become concentrated in certain components of the human food chain.
Such toxicants may, therefore, pose a continuing threat to human
health, primarily through the ingestion pathway, long after their
release into the environment has been reduced.
-------
1.4 Pavel OP ino An Aggregate Risk RanXincr
The development of a single aggregate risk ranking that would
combine the relative risks for cancer with the relative risks for
other types of adverse health effects was addressed by the
Subcommittee, which evaluated the data and methodology required for
the purpose. Such an aggregate ranking would provide additional
guidance to the Agency in setting priorities. Although possible in
principle, the development cannot be accomplished without comparing
the impacts of different types of health effects on the total
population as well as on the individuals directly affected. The
Subcommittee recognized that the development of any aggregate risk
ranking that attempts a single scaling requires resolution of many
implicit value judgments and ethical issues beyond the scope or
authority of this Subcommittee or the EPA. That is, to attempt a
relative ranking in terms of severity (or significance) of such
disparate health outcomes as birth defects in infants compared to
paralysis in older persons requires consideration on many
dimensions of the values we place on various members of society,
families, and the utility of specific physical and mental functions
for individuals and society. Such a comparison requires that the
impact of each effect be scored for severity, a process
necessitating selection of suitable measures and scales of
severity, as well as appropriate weighting factors. In addition,
the current disparity in risk assessment approaches for carcinogens
and systemic toxicants makes it exceedingly difficult to construct
a universally acceptable aggregate ranking. Although the data and
time needed for such a complex task were not available, the
Subcommittee described ways by which such an aggregated ranking
might be undertaken in the future, assuming that the important
value-laden issues can be equitably resolved.
1.5 Racommandad Approaches
In considering how risk areas might be better defined and
relevant information organized for ranking/assessment purposes, the
Subcommittee proposes as a possible approach the development of a
matrix, the principal dimensions of which include sources, exposure
situations, agents. and health endpoints. For example, a two-
dimensional array, with rows representing ultimate sources (such as
agriculture) and columns representing direct or proximate sources
impacting human health (such as drinking water), would help to
identify those intersections at which risk reduction initiatives
-------
would produce the greatest benefits (see Table 6.1.1). Expanding
the dimensions of such an array, by including specific agents (as
in Figure 6.1.2) and health endpoints, would allow an even more
detailed identification of where the Agency could act most
effectively.
Developing the matrix in usable form and entering information
into it would be no small task. In the final analysis, the task
will never be quite complete; whatever initial system is adopted
will undergo continual change, expansion, and development (as
distinct from maintenance) as it is used and as experience is
gained from cataloguing new information in it.
The Subcommittee recommends that the Agency undertake the
development of such a prototype matrix, beginning with a limited
pilot effort using a few, widely spread agents, and designed to
explore its feasibility. Existing relational data base software
would support such an effort, and the resulting four-dimensional
information system would itself be usable, and would also provide
information for the development of an "ultimate" system. This
approach would reveal complexities and practical difficulties at an
early stage. A later stage of development would expand by adding
a larger number of agents selected for potency and ubiquity. They
could be selected from preexisting lists (such as those developed
under Title 3 of the SARA "Community Right to Know Provisions") .
As the system is developed, it should be linked to existing
databases, such as the EPA's Integrated Risk Information System
(IRIS) .
Once the system were to become even partially functional, its
value would be great. Applying the concept of the interconnected
four-dimensional system as an aid to the thought process when human
health risk issues are addressed should improve the risk assessment
process at once; documentation of such applications would be a
source of information for insertion into the system itself.
The Subcommittee further recommends that the Agency assign a
specific management focal point for this effort to assure
accountability.
With the ultimate aim of improving the assessment and ranking
of environmental risks to human health, the Subcommittee recommends
the following additional actions:
-------
a) Establishment of a risk assessment framework for other
toxicants consistent with that used for carcinogens. The
recommendations in b) , c) , d) , e) , and f) below, while
useful in and of themselves, will also contribute directly
to achieving this goal.
b) Establishment of a formal mechanism for risk anticipation
(i.e., identification of emerging problems), as rec-
ommended in the Future Risk report (EPA, 1988), including
an expert in-house committee, peer oversight, and a means
of supporting long-term research on emerging problem
areas.
c) Expansion of long-range research on the assessment of
human exposure. Topics should include developing data and
models on the variation of exposure with time and place,
and obtaining detailed and comprehensive exposure
measurements (including data on: (1) ambient exposure
levels; (2) tissue burdens; (3) uptake, distribution,
metabolism, and excretion of the toxicants of interest,
and the extent to which these parameters may vary with
age, sex, diet, physiological state, and other variables;
and (4) relevant biological and molecular markers of
exposure.
d) Improvement of the relevant toxicological science base,
including more systematic data on the toxicity of
environmental agents for humans of different ages, more
comprehensive assessment of their toxicity in surrogate
toxicological test systems, and better understanding of
the appropriate dose-response and trans-species scaling
functions to be used in assessing their risks to human
health.
e) Development of the extensive exposure and toxicity data-
bases needed, through closer cooperation with other
federal (e.g., NCHS, NIH, NIOSH, FDA, and DOE), state and
local agencies, as well as with institutions in the
private sector.
f) Establishment of a long-term program to improve the cap-
ability for assessing and ranking environmental risks to
human health. The program should involve extramural peer
8
-------
review and should be organized in such a way as to deal
most effectively with the relevant research issues.
g) Further development of scientific capability in the re-
quisite disciplines; i.e., since assessment of the health
risks of environmental agents requires the coordinated
efforts of biologists, chemists, epidemiologists,
mathematicians, physicians, toxicologists, geneticists,
and scientists of other disciplines, and since few
institutions have the multidisciplinary teams required
for such research, there is a need to develop programs
for fostering such collaboration on a broader scale, for
focusing it on the key problems that deserve to be
pursued, and for the further training of scientists with
the necessary expertise, through long-term support of
graduate and postgraduate training in toxicology,
epidemiology, exposure assessment, and the other relevant
disciplines.
Future risk rankings should be based on risk assessments for
specific single toxic agents or definable mixtures, and on the
cumulative human exposure to such agents. In actually conducting
future risk ranking exercises, the following factors, discussed in
the Subcommittee's report, should be considered:
a) The effects of uncertainty in exposure estimates should be
stated explicitly and factored into any risk charac-
terizations, and possible interactions for exposures
involving complex mixtures should be addressed.
b) Consistent criteria should be developed for the asses-
sment of toxicity and the identification of hazards. To
accomplish this, the Agency should develop and apply
consistent criteria for hazard identification, include
sub-clinical and pre-clinical effects of pollutants as
endpoints of concern, and expand its assessments of
substances/agents within selected "problem areas" (however
defined) to encompass truly representative samples.
c) The distribution as well as the mean, should be evaluated
when considering the severity of health effects. In the
case of lead, an average decrease of five percent in IQ
scores for individuals would translate into a greater than
-------
fifty percent decrease in the number of individuals
scoring in the upper intelligence ranges, and a
quadrupling of the number of persons with IQ scores less
than 80.
d) Assessments should consider risks to individuals, as well
as risks to the general population and to susceptible sub-
groups .
e) The Agency should be cautious in using merged ranking sch-
emes for cancer and non-cancer endpoints. Difficulties
arise from the lack of a clear biological rationale,
divergent histories, and the absence of an acknowledged
scoring system for severity of effect. Approaches to a
merged ranking system are described in the Subcommittee
report (section 6.3) as well as an illustration of the
steps and problems involved in the complex process of
merging rankings of different types of risks to human
health.
f) Consideration should be given to the time period over
which different risk reduction strategies may be
effective when evaluating the risk posed by a given
toxicant, as well as to the persistence of risks if
uncontrolled.
g) It should be recognized that the assessment of rela-
tive risk is a value-laden process (particularly with
respect to relative severity and equity), which should
involve toxicologists, epidemiologists, exposure
assessors, medical experts, sociologists, ethicists, and
informed representatives of the general public.
h) Risk rankings should explicitly address the extent to
which existing control strategies effect risk reduction,
and conversely, the estimated risk in the event that
existing programs were not to be continued at the current
levels.
10
-------
2.0 introduction
2.1 Background
Broader use of the concept of risk reduction in EPA's planning
of research and regulatory strategies was recommended to the Agency
by its Science Advisory Board in 1988, in the "Future Risk" report
noted above. The recommendation was followed in 1989 by a request
from the EPA Administrator, William K. Reilly, for SAB's technical
assistance in developing strategic risk reduction options to aid
the Agency in assessing its activities. In response, the SAB
undertook to provide the requested assistance, forming the Relative
Risk Reduction Strategies Committee (RRRSC) to expedite the
process.
The SAB recognized at the outset that one of the first steps
to be taken was a review of the 1987 report entitled "Unfinished
Business: A Comparative Assessment of Environmental Problems" (UB)
which summarized EPA's evaluation of the relative risks of the
major environmental problems of concern to the Agency at the time.
That evaluation had assessed the comparative risks of some 31
environmental problems (Table 2.1), judged in terms of:
a) their risks of contributing to the occurrence of human
cancer
b) their risks of causing other adverse effects on human
health
c) their risks of causing damage to the ecosystem, and
d) their risks of causing adverse effects to societal
welfare
In light of these earlier assessments by the Agency, the SAB
charged the RRRSC to:
a) provide a critical review of the "Unfinished Business"
report, taking into account any significant new
information bearing on the evaluation of the risks
associated with specific environmental problems
b) provide, to the extent possible, merged evaluations of
cancer and non-cancer risks (i.e., health risks) and of
ecological and welfare risks (i.e., environmental risks)
11
-------
1 - Criteria air pollutants
2 - Hazardous/ tox>c air pollutants
3 - Otrer air pollutants, e g , flourides, total reduced su'fur
4 - Paaon C iindoor pollution onty}
5 - Indoor air DO Nut ion Bother than radon}
5 - Radiation Cither than radon}
7 - Substances suspected of depleting stratospheric ozone layer
B - Carbon dioxide and global warming
3 - Direct point-source discharges to surface waters d& q , industry}
13 - indirect point source discharges, e g . POTWs
* 1 - Non-point source discharges to surface water oI us in-place
toxics in sediments
"2 - Contaminated sludge C'nc'udes municipal and scrubber sludges}
"3 - Discharges to estuaries, oceans, ect £al! sources}
"a - D'Scharges to wetlands Ca1 ' sources}
"5 - Cringing water at t~e tap C ' nc i uces cnenvcals, lead from
pipe, biological contaminants, radiation, etc
'5 - Active hazardous waste sites C'ncIudes hazardous waste
tanks, inputs to groundwater and other media
17 - inactive hazardous waste sites C'nc'udes Superfund, inputs
to groundwater and other media}
"B - Municipal non-hazardous waste sites C ' nputs to groundwater & otne1'
"9 - inaustriai non-nazardous waste sites
2G - Mining wastes., e.g , oi I and gas extraction wastes
2i - Accidental releases of toxics Cal! media}
22 - Accidental oil spi I is
23 - Releases from storage tanks C'ncIudes product & petroleum tanks}
24 - Other groundwater contamination Cseptic tanks, road salt, injection *e
25 - Pesticide residues on food eaten by humans or wildlife
2E - Application of pesticides CincIudes risk to pesticide workers ana
consumers who apply pesticides}
27 - Other pesticides risks
28 - New toxic chemicals
29 - Biotechnology
30 - Consumer product exposure
31 - Worker exposure to chemicals
Table 2.1 Original EPA list of Environmental Problems considered in the
1987 "Unfinished Business" report (pages 10-11)
c) provide optional strategies for reducing major risks
d) develop a long-term strategy for improving the methodology
for assessing and ranking risks to human health and the
12
-------
environment and for assessing the alternative strategies
to reduce the risks.
In order to facilitate the accomplishment of these tasks, the
SAB formed three Subcommittees of the RRRSC: the Ecology and
Welfare Subcommittee, the Human Health Subcommittee, and the
Strategic Options Subcommittee. The report of Human Health
Subcommittee follows.
2.2 Charge to t^* gum an Health
The Human Health Subcommittee was charged with the following
tasks: a) to provide a critical review of the "Unfinished
Business" report in light of new information bearing on the
evaluation of the risks to human health attributable to specific
environmental problems; b) to provide, insofar as possible, updated
and merged evaluations of the relative risks of cancer and the
relative risks of other adverse effects on human health
attributable to specific environmental problems; and c) to
recommend approaches for the development of a long-term strategy to
improve the methodology for assessing environmental risks to human
health.
2.3 Format of this Report
Section Three of this report reviews the kinds of information
and analyses that must go into any assessment of environmental
risks to human health. These include evaluation of the toxicity of
the environmental agent (s) in question, as well as the degree (s) of
human exposure to the agent (s) . The next section appraises the
extent to which the data and methodology in "Unfinished Business"
were adequate for accomplishing the intended assessments. The
following section considers approaches for developing a long-term
strategy to improve the evaluation and ranking of environmental
risks to human health, including the merging of cancer and non-
cancer risk rankings. The final section, presenting the
Subcommittee's conclusions and recommendations, is followed by
appendices containing case studies to illustrate the difficulties
inherent in environmental risk assessments as well as detailed
discussions of suggested methods for ranking different risks.
13
-------
3 • 0 Essential Elements in Assessment of
Risks to Health
3.1 overview
The development of any risk assessment and risk ranking
process requires specification of the criteria for ranking. The
UB participants, especially when dealing with non-cancer health
effects, struggled to impose order on a heterogenous universe of
exposure scenarios, agents, and endpoints. They adopted the tactic
of focusing on a limited number of agents within each problem area,
selecting those for which a reasonable amount of data were
available. On the basis of estimates of the severity of health
endpoints, the sizes of the exposed populations, and the potencies
of the different agents (actually defined as a margin of safety) ,
they assigned rankings to each of the 31 problem areas.
As a preliminary strategy, the effort was commendable because it
clarified the difficulties posed by the absence of definitive
information. In fact, much of the exercise had to proceed in the
absence of sufficient information. Naturally, the first item in
any strategy for improving risk predictions is the acquisition of
adequate data.
3.2 Assessment of Exposure
The "Unfinished Business" (UB) report addressed the fact that
there was significant uncertainty in estimates of exposure, and
hence risk. However, the discussion of potential exposure was
limited.
In Appendix I, the report of the Cancer Work Group, it was noted
(p. 16) that "Ranking environmental problems was complicated by a
lack of information, uncertainties in estimating exposures, the
diversity of methods used to assess different problems and to
project national cancer incidence from smaller-scale studies, and
differences in the degree of coverage of potential carcinogens."
It also noted that "the quality of the human exposure for the 31
environmental problem areas varies greatly, making comparisons
difficult." It was pointed out (p. 14) that "various methods of
assessing exposure may also have biased comparisons of different
problem areas. Not all analyses made exposure assumptions with the
same degree of conservatism." These statements are persuasive in
14
-------
suggesting caution in using their quantitative risk assessments, as
well the relative ranking of categories as a basis for decision-
making in environmental regulation, particularly since the
potential limitations were emphasized in this fashion by the group
performing the assessments.
3.2.1 Data Gaps and Uncertainties
In assessing exposure the UB report was faced with the kinds
of data gaps relative to exposure assessment that are not unique to
its undertaking and which are frequently encountered in assessing
risks to environmental contaminants. These include:
3.2.1.1 Specific Chemicals
For some categories there is insufficient information about
the presence of specific chemicals due to the fact that the data
base was either limited, established for other purposes or may not
be recent. Thus, for example on p. 13 of Appendix I (Report of the
Cancer Work Group) it was noted that for pesticide residues on food
the Group "extrapolated from a few suspected carcinogens to the
universe of potential carcinogens..." Another example is the
omission of arsenic among the list of carcinogens in Problem Area
15, Drinking Water. It appears to have been omitted because it was
not a member of the three categories of water constituents that
were addressed. In the case of Problem Area 17, Hazardous Waste
Sites-Inactive, it was noted that the data for the 12 chemicals for
which the risks were estimated were based on 35 sample sites which
were chosen to represent thousands of such sites. It is indeed
understandable that in an undertaking of this magnitude omissions
and limitations must necessarily occur. The question arises as to
their impact on the estimated population risks and the relative
ranking of categories.
3.2.1.2 Concentrations
For the UB report various methodologies were used to establish
concentrations of chemicals. These included measurements from
surveys, both large scale and small, as well as modeling, such as
dispersion modeling applied in Problem Area 2, Hazardous/Toxic Air
Pollutants. The calculated risks were in many cases based on
skimpy concentration data. The sludge section (#12), for example,
lists contaminants in sludge in Table A-l and page B-70. However,
15
-------
no information is given about the levels. Also, in Problem Area
17, Hazardous Waste Sites-Inactive, the number of best-guess cancer
cases was extrapolated to an estimated potentially exposed
population of 6.8 million, based on concentrations of 6 chemicals
at 35 sites with an estimated exposure population of about 50,000.
Aside from the uncertainties of extrapolating to such a large
number of other sites, the question necessarily arises as to the
validity of the concentrations reported as a basis for these
exposure estimates. Often at hazardous waste sites the modeling of
risk is based on a wide range of assumptions and often very limited
data. Thus, it may not follow that the calculated exposures based
on such limited data and the application of groundwater modeling
are accurate even within orders of magnitude as expressions of the
concentrations to which people are exposed in their water supplies.
The document itself points out (pages B-44) that the data on
the occurrence of synthetic organic chemicals (SOCs) in drinking
water are severely limited. However, since the document was
developed, additional monitoring or survey data have become
available and should be examined. These data include Superfund
SARA Title III reports, the health assessments for inactive
hazardous waste sites prepared by the Agency for Toxic Substances
and Disease Registry (ATSDR), and monitoring data required for
newly regulated constituents in public water supplies. The risk
assessments should be updated to determine whether it would be
consistent with the expanded database that is now available.
Frequently the calculations of lifetime risks are based only
on the current concentrations and do not consider how these might
change over decades of time. Estimates at a specific site are
subject to great uncertainty. If extrapolations are to be made to
estimate national risks, the uncertainties are necessarily much
greater yet. Finally, it must be emphasized that exposure
concentrations based on very few measurements or modeling are not
likely to reflect accurately those to which a complete population
is exposed. For example, at water supply treatment plants the
concentrations may be substantially different than at various
points in the distribution system. Or, in the case of lead,
corrosion in the system can add substantially to its concentration.
In the case of volatile organics (VOCs), their very volatile nature
will affect exposures by both ingestion and inhalation. These
factors that affect the concentration at the point of actual
exposure are important in accurately determining risk.
16
-------
3.2.1.3 Nature of Exposures
The contaminated media to which people are exposed are
frequently assessed on the basis of only one mode of contact, e.g,
ingestion for water, inhalation for air, skin contact for soil.
The UB report does recognize that there may be multiple routes of
exposure, as well as intermedia transport. However, it is not
clear that these are sufficiently considered. For example, in the
case of Problem Area #15, Drinking Water at the Tap, the report
states (in Appendix I, p. B-44) that "if the chemical has been
shown to be carcinogenic through inhalation and not ingestion, it
will not be considered a potential carcinogen via drinking water."
This does not seem to recognize the inhalation exposures to
volatile chemicals that regularly occur from indoor uses of water.
At the same time, it does not appear that skin contact with such
carcinogens from bathing with contaminated water were considered as
well.
Recent exposure estimates suggest that the ingestion pathway
may be of much greater importance than that for inhalation for
persistent chemicals, such as lead and the polychlorinated
dibenzodioxins and furans. These chemicals can be taken up or
deposited on plant or forage crops which in turn can be eaten by
people or food-producing animals. These same chemicals are
deposited in rivers and lakes, or are transported to water bodies
by surface water run-off; they can accumulate in fish consumed by
people. A recent EPA report estimates that these ingestion
exposures are likely to be greater than those via inhalation of
emissions from municipal solid waste incinerators. Thus these
integrated postdeposition routes of exposure may be important in
assessing exposure and risk from originating sources that release
substances into the air, but impact upon land and surface waters.
3.2.1.4 Ranges and Variabilities of Exposure
The UB document doesn't provide sufficient perspective of the
range of exposures that can occur within a given problem area, or
how the exposure may vary over time. On page 17 of the overview
the document states that descriptions of aggregate populations and
individual risk were of interest. The differences that were
considered appear to be limited to differences in exposure between
groups, such as pesticide applicators and their exposures, in
17
-------
comparison to pesticide exposures of the general population from
food or in their homes.
Some of the smaller public water supplies or private wells may
be highly contaminated from waste sites and other sources,
especially those that are usually not tested for unusual or
esoteric contaminants. Some private water supplies have been found
to be contaminated with organic chemical concentrations greater
than 10 ppm, and some public supplies greater than 1 ppm. Whether
or how such unusually high exposures were considered is not clear.
There are a number of individual behavioral factors that can
affect exposure. They include the frequency and use of materials
containing contaminants, the behavior that causes release of
contaminants, and the time-location patterns of individuals. For
the most part these do not appear to have been addressed in the UB
report. There is a brief discussion (p. 14, Appendix I) that
refers to mitigating behavior. This is described as the extent to
which people reduce their exposure when they know that they are at
risk. As an example, it is mentioned that "people may stop
drinking water that tastes bad or is known to be polluted." Such
mitigating behavior was not, however, specifically evaluated with
respect to its effect on exposure. However, this is indeed a
difficult area to assess. More importantly, the frequency and
locations where people spend their time will necessarily have a
substantial impact on assessing inhalation exposures to air
pollutants. National and regional studies in this regard are now
being undertaken and will provide a valuable data-base on the range
and distribution of individual behavior patterns of peoples' uses
of time indoors and outdoors, with specific reference to the impact
on exposure to air pollutants. Data on the variability of the
ingestion of water have been developed that indicate that standard
reference intakes of 2-liters per day for a 70 Kg adult needs to be
reassessed when estimating the exposures to waterborne pollutants.
The ingestion pattern is quite variable. Average consumption of
tapwater by children is estimated to be higher than for adults on
a body-weight basis (1 liter per 10 Kg—NAS, 1986). In addition,
while for many adults the average consumption of tapwater may be
less than 2 liters per day, the results of a recent survey showed
that 5% of adults 20-64 years old have an average daily water
consumption of tapwater of 2.71 liters per day, and an average
total water intake of 3.79 liters per day (Ershow and Cantor,
1986). There is a large variability around the mean. Whether this
18
-------
is important in relation to the probably considerably greater
uncertainty in dose-response relationships, however, is a separate
question. Finally the uses of water, other than for ingestion,
which lead to human exposures are also highly variable. Bathing
and showering lead to inhalation and dermal exposures, and other
indoor-water uses release volatile chemicals, causing inhalation
exposures to all the inhabitants of the building. Thus the
interaction of the behavior causing the releases, and the time
spent within the various rooms of the building all influence the
final determination of indoor-inhalation exposure.
3.2.1.5 Exposure to Complex Mixtures
Many of the problem areas involve exposure to complex
mixtures, or the selection of an indicator chemical as a surrogate
for a mixture. In many situations the specific mixture of
chemicals to which people are exposed is characterized to only a
limited extent. In these situations very few data may be available
to assess adequately assess the risks. However, new exposure
surveys could be used to identify additional chemicals of concern.
3.2.2 sHTnn>ary and Recommendations
It is clear that there are a variety of factors that have not
been, and probably could not readily be, determined in establishing
exposure for the purpose of assessing risk in the framework of the
UB report. The question arises as to the extent to which these
deficiencies bias or invalidate the quantitative impacts that were
calculated and, hence, the relative rankings of risk for the
various problem areas. Although it may be difficult to improve the
precision of the calculations of quantitative risk for each of
these areas by considering in detail the deficiencies in the
various exposure factors cited above, it would be useful to attempt
to include their variabilities where they are known, and in any
case estimate their uncertainties. Thus, for example, in the case
of drinking water the range of ingestion factors and the possible
impacts of inhalation and dermal exposure should be considered,
since information is available in these areas. With respect to
uncertainties in exposure, there should be at least a semi-
quantitative assessment or judgement of the impact on the risk
calculations. Where these uncertainties are very great, i.e.,
orders of magnitudes, as they are likely to be in some cases, a
19
-------
good understanding of their effects is essential in ordering and
prioritizing the problem areas.
3.3 Assessment of Toxicity
3.3.1 Hazard Identification
The first step in risk assessment is the identification of a
hazard, i.e., potential risk). This involves detailing the
inherent toxicity (including carcinogenicity) of the substance or
agent in question regardless of the actual level of exposure.
Specifically, hazard identification is aimed at determining whether
exposure to an agent can cause an adverse health effect (National
Research Council/National Academy of Sciences, 1983) . Evidence of
inherent toxicity conventionally includes data on structure-
activity relationships to known toxicants, in vitro or whole-animal
short-term tests, chronic or long-term animal bioassays, human
biomonitoring data, clinical studies, and epidemiology. A complete
hazard identification process entails review of available informa-
tion in these six categories in order to determine whether the next
step—quantitative risk assessment—is warranted. The National
Academy of Sciences has estimated that there are at least 25
components—of both a scientific and policy nature—in complete
hazard identification (ibid).
By contrast, the Unfinished Business report was based on a
foreshortened and largely undefined hazard identification process.
Instead of carrying out complete hazard identification reviews
according to clearly stated criteria, the working group relied
largely on preexisting listings of candidate chemicals. Although
these lists appear to have been driven by the non-availability of
positive human and/or laboratory animal testing data, the criteria
for hazard identification were never explicitly stated in the
document. In any future attempt to rank risks of environmental
toxicants, the hazard identification criteria should be explicitly
stated. In line with the goal of disease prevention, they should
include evidence of preclinical or subclinical effects of pol-
lutants .
This lack of a consistent approach in selecting hazards is a
serious limitation of the document. Yet it is easily understan-
dable given the dearth of available toxicologic data on new and
existing chemicals. The HAS has estimated that no toxicity data
20
-------
are available for approximately 80% of the 48,000 chemicals in
commerce (National Research Council, Toxicity Testing, National
Academy Press, 1984) .
This "Achilles heel" in hazard identification is no less
evident when new chemicals are considered. Here, information on
toxicity is woefully deficient. As stated in the Unfinished
Business report, the Toxic Substances Control Act (TSCA) requires
that industry submit to EPA data related to the health effects of
new substance prior to its manufacture or importation. The data
are claimed to be confidential by the submitters in the great
majority of cases, however, so the premanufacturing notification
(PMN) process allows EPA (but not the public) to identify potential
risks presented by specific new chemicals (App.I, 8-63). EPA's
own review of ten years experience with the PMN process under the
TSCA indicates that only 60% of the new chemicals have any
toxicological data (Auer, et al., 1988). The Subcommittee views
the development of an adequate toxicological data base on existing
and new chemicals as a priority—and a prerequisite—to any
attempts to quantify comparative risks more precisely.
A third major weakness of the Report, flowing from the first,
was the frequent reliance on a few selected surrogate contaminants
to represent large categories of pollutants. For example, the
Cancer Risk work group selected 4 agents—formaldehyde, methylene
chloride, paradichlorobenzene, and asbestos as representative of
the vast category of consumer product exposures. For non-cancer
effects the Work Group relied on 3 pesticides to illustrate
"pesticide residues on food," despite their acknowledgement that
perhaps 160 pesticides may constitute potential risks. Similarly,
only 6 of the hundreds or thousands of chemicals of concern in
indoor air were evaluated (App. II, p. 2-1).
In summary, the present Subcommittee makes the following
recommendations:
a) The agency should develop and apply consistent criteria
for hazard selection, since this process is the critical
first step in risk assessment and determines the validity
of the final product.
b) Subclinical and preclinical adverse effects of pollutants
should be included as endpoints of concern.
21
-------
c) EPA must make a concerted effort to improve its toxico-
logical data base on both new and existing chemicals.
d) EPA should expand its assessments of substances/agents
within selected "problem categories" to encompass truly
representative samples.
3.3.2 Dose-effect Characterization
A fundamental and basic tenet in toxicology is the existence
of a dose-response relationship. To quote Paracelsus: "All
substances are poison; there is none which is not a poison. The
right dose differentiates a poison and a remedy." Dose-response
data have, therefore, long been considered to be the cornerstone of
risk assessment.
More recently, consideration of the dose-response relationship
has become complicated by the recognition of at least two alter-
native dose-response models, defined in operational terms: the
threshold dose-response model and the non-threshold dose-response
model. All carcinogens are now assumed to be biologically active
even at the lowest doses, without thresholds; thus there is no
"right" dose at which they are considered harmless. On the other
hand, for many effects other than cancer, dose-response relation-
ships are known or presumed to have thresholds, with the result
that the causative agents are considered to be ineffectual at
sufficiently low doses. This dichotomy was reflected in the risk
assessments presented in the UB report.
It should be emphasized that a conceptual problem with
thresholds is the difficulty of identifying "safe" levels for a
diverse human population expected to have significant inter-
individual variations in biological response to toxicants. In the
case of lead, neurodevelopmental effects are being observed at
increasingly low levels of exposure. Recently, an extrapolation or
a combined extrapolation/safety factor approach has been suggested
for non-carcinogens such as reproductive or developmental toxicants
(Gaylor and Kodell, 1980; Gaylor, 1989).
Another difficulty lies with our concepts of "threshold."
Actually, we can envision that, for any given chemical, we might
have to deal with several thresholds. One threshold can be defined
by our present capabilities to detect the presence of a given
22
-------
chemical. Progress in analytical techniques made over the last
several decades has pushed this threshold to lower and lower
levels, as documented several years ago in "The Case of The
Vanishing Zero," (Zweig, 1970). Another threshold may be defined
by limitations of our analytical capabilities with regard to access
to materials to be analyzed. For example, many analytical
procedures allow quantification of foreign compounds in easily
accessible compartments such as body fluids. The same procedures
are of much less, if any practical value to detect the same
chemical in critical internal targets such as the brain or the
kidneys without interfering seriously with normal organ structure
and function. A third category of defining a threshold is time-
dependent. Today's lesion often heals or is gone away tomorrow.
On the other hand,a recent follow-up study has indicated long-term
neurobehavioral effects from low-level exposures to lead (Needleman
et al., 1990). There are many biological processes involved in
repair and regeneration and reversibility vs. irreversibility is an
important, but not sufficiently studied problem and must be
considered whenever there is discussion of thresholds. There are,
also, individual vs. population thresholds as well as "threshold-
like" behavior. Finally, there is no clear-cut and generally
accepted definition of what constitutes an untoward or "adverse"
health effect. The only method for adequately judging if a
threshold exists is an understanding of mechanism and of the
biological system being affected.
If difficulties arise in the interpretation of dose-response
data for risk assessment, the lack of sufficient data for precisely
characterizing dose is often a limiting factor. Another problem
may arise in linking dose to response and arriving at a judgement
as to what the response means. Recent developments in the science
and technology of "biomarkers" illustrate conceptual and
practical problems in the Paracelsian approach to risk assessment.
In lead poisoning, for example, biomarkers provide good evidence of
exposure, and it is possible to link such specific biomarkers with
some of the more florid manifestations of lead poisoning. Some
years ago, a "threshold" could be defined, but more recent
studies suggest that a "sub-threshold" dose for one untoward effect
by no means constitutes a "sub-threshold" dose for another,
potentially more deleterious effect (e.g., consequences of acute
vs. chronic exposure? early vs. late signs of poisoning). Similar-
ly, a blood alcohol level above a certain limit is predictive of
impaired motor and sensory function, but of little value in
23
-------
predicting chronic nervous system impairment or cirrhosis of the
liver, to say nothing of fetal alcohol syndrome.
In view of present limitations in our ability to interpret and
integrate dose-effect relationships in the low-dose domain, it is
necessary to rely on informed assumptions for the purpose of risk
assessment. There are several alternatives. One may adopt a
conservative stance (i.e., to err on the side of being safe and to
assume that any amount of a toxicant can increase the risk of
disease in some individuals) or one can assume a human population
threshold. The former assumption is justified by the observation
of significant interindividual variability in response to toxi-
cants, including carcinogens (Marquis and Siek, 1988; Harris, 1985;
Perera et al., in press).
A few of the general problems that were inherent in the risk
assessments contained in the UB report may be addressed as follows.
3.3.2.1 pefininq the Dose
The dose of a chemical is often defined as the amount of the
substance that is administered under specific conditions; however
a problem in defining the dose arises when the amount of the
chemical is not known precisely as is the case with most environ-
mental agents. In this situation, the dose is often related to, or
equated with, the extent of exposure. For example, the concentra-
tion of a given chemical in air, water or food is equated roughly
with the "dose11. Epidemiological studies often implicitly rely
heavily on this type of operational definition of dose, although
there is always uncertainty about the extent to which exposure
conditions (or concentrations) result in a given quantity of a
chemical actually entering the body.
A second problem concerns estimation of the precise relation-
ship between intake of a given amount of a chemical and the
resultant effective dose. Every chemical entering the body is
subject to the process of uptake, metabolism and elimination. Many
chemicals are rapidly inactivated and eliminated, while others may
accumulate or be activated. Dose depends thus not only on exposure
conditions, but also on the interplay between intrinsic properties
of the chemical and the capability of the organism to deal with the
agent.
24
-------
The third problem can be defined as "target dose11 vs. "body
dose." Many chemicals have no untoward effects unless they reach
a critical biological target, i.e. the site where they can cause
harm, in sufficient concentration to do so. Whether a chemical
reaches its target or not is subject to many variables, such as
route of exposure, toxicokinetic parameters and the capability of
the exposed organ, tissue, or cell to deal with the agent.
Ideally, the target dose should be known for a rational assessment
of risk; however, in practically all instances this information
remains unavailable for humans with the result that human risk
assessment is correspondingly imprecise. Exposure "dose" is thus
usually the best surrogate now available. Acceptable approaches
for extrapolating from exposure conditions to "dose", be it total
body dose or critical target site dose, may be developed through
mathematical modeling based on appropriately designed laboratory
experiments with animals. New developments with "biomarkers"
applicable directly to human populations promise to yield ad-
ditional approaches.
3.3.2.2 Defining the Response
Response, or "endpoint," can be difficult to assess or to
define. While certain endpoints, such as death, acute tissue
injury, and cancer, are easily recognized, other responses may be
much more difficult to detect or evaluate. During recent years,
progress has been made in identifying so-called biomarkers of
effect. The conceptual approach and techniques used, coupled with
an understanding of the underlying biology (e.g., detection of DNA
adducts) holds great promise for refining our analytical capabili-
ties. The difficulty lies in answering the question: "What is
truly a valid indication of an untoward health effect?" For
neoplasia, any indication that an exposure may cause benign or
malignant neoplasms is an unacceptable response. It is even more
difficult to deal with non-cancer responses, that may include the
more than 90 specific non-cancer health endpoints in the UB report.
Some effort was made to classify these endpoints into various
categories, from those of lesser concern to those that are severe,
but the classification lacks logic and consistency. Some of the
listed endpoints are true disease entities (e.g. pneumonia, herpes,
increased heart attacks, mortality). Some are only signs of
disease (e.g., angina, irritability, jaundice) or symptoms (e.g.,
headaches, learning disabilities). Still others are clinical or
subclinical findings (e.g., decreased heme production, transient
25
-------
decrease in pulmonary function, reduced time to onset of exercise-
induced angina (heart pain) and even alterations in chromosomes and
oncogenes that in themselves may or may not signify disease. While
any of these effects may be produced by chemicals, such a list does
not address the fundamental criteria for deciding whether a par-
ticular finding truly constitutes an adverse health effect.
Moreover, the associations of agent or problem area with specific
endpoints, given in Table A-2 of the Non-Cancer Work Group report,
are at times inconsistent, if not specious; for example every item
in the "Neurotoxic/Behavioral list evokes questions about its
suitability. For instance, why is micromercurialism accorded a
major category? Why is chlordane, but not mercury (a much more
potent tremorigenic agent) associated with tremor? How are
convulsions and neuropathy distinguished? Why is retardation listed
here and absent from developmental effects? And, given the history
of EPA's basis for lead regulation, how could the list exclude the
subtle population shifts in intelligence test scores correlated
with mild lead exposure except by regressing to the outmoded
measure of "number of cases?"
Much of the confusion seems to stem from lack of a clear
conception of the differences between cancer and non-cancer risk
assessment. The current process of risk assessment is based upon
the simplifying assumptions that cancer is a unitary endpoint and
exposure is a single dimension. If the assumptions are widened, it
becomes possible to incorporate different exposure scenarios with
different consequences into the risk assessment process.
One scenario might encompass the effects of chronic exposure
and cumulative damage, and include examples such as depletion of
lung function with pulmonary toxicants such as ozone, diminished
renal function with substances such as cadmium, and neurodegenera-
tive impairment with agents such as methylmercury or certain
organic solvents. Each of these examples can serve as a model of
progressive toxicity which, at some arbitrary stage, is transformed
from a reversible to an irreversible process or an asymptomatic to
a clinically significant change; the latter, too, may be the
product of a convergence between natural aging and toxic damage.
The rate of progression should serve as a crucial risk parameter.
Other scenarios emerge from acute exposures. Persistent
deficits could be the outcome of brief fetal exposures to agents
such as heavy metals or ethanol. Or, apparently reversible
26
-------
impairment might follow from a single episode of insecticide
exposure, or exposure to high levels of a volatile solvent with
anesthetic properties. It is not possible to judge the validity of
the UB conclusions without a description of which exposure-effect
scenarios were envisaged.
At present, defining health effects depends on recognizing
deviations from normal structure and function, an approach that is
driven by our analytical and diagnostic capabilities. Although we
still lack an adequate understanding of the health significance of
certain signs and symptoms, we must acknowledge that in the
interest of disease prevention, validated early indices of risk
such as chromosomal aberrations, gene mutations, certain enzyme
alterations, reduction of lung function, and other preclinical
indicators should be evaluated as elements in the spectrum of
health endpoints of possible concern.
3.3.2.3 Defining Dose-response Relationships
It must be remembered that any "response" defined and assessed
in a dose-response analysis, represents the mean value of a set of
responses that often follow a log-normal distribution in the
exposed population. Within a large population there may exist
families (defined by host characteristics) of dose-response curves,
that are shifted to the right or left of the "ideal" curve and that
have different slopes. The consequence of this phenomenon would be
the inability to identify a "population threshold."
It may be concluded from the foregoing that ideal dose-
response data for a toxicant should meet at least the following
criteria: 1) the response should be a quantifiable endpoint and
should be known to represent, in a health framework, an interpre-
table observation; its implications should be well enough under-
stood for the making of meaningful predictions with reasonable
accuracy; 2) it should be known to what extent the response depends
on total (integrated) dose, single dose or multiple doses, and on
the dose-rate; 3) qualitative information on the target site of
action of the toxicant must be known, e.g. what organ, organ
system, tissue, cell, or cellular mechanism is affected; 4)
quantitative relationships between the amount of chemical at, and
its effect on, the target must be known, both with regard to
exposure conditions and with regard to the target/tissue dose; 5)
because there are different exposure scenarios for different
27
-------
toxicants, the interrelationships and correlations between
different scenarios must be known well enough to allow extrapola-
tion from one scenario to another; and 6) it should be known
whether the response may be modified, or be subject to modification
in subgroups of the population at risk, whether or not it is
reversible, and whether it may be modified by other agents or by
other biological circumstances (e.g. concomitant disease).
Although, in general, the above information is available on
the acute effects of many chemicals, including drugs, pesticides,
certain metals, inhalants (such as CO) , and other agents of
environmental concern, much less information is available on the
chronic toxicity of such agents. Evaluation of chronic dose-
response relationships entails additional problems as well, some of
which are discussed briefly in the following.
Chronic dose-response data have usually been obtained,
construed, or evaluated on the assumption that the relevant
exposure has occurred continuously at a more or less constant
level, and that the resulting effect has been cumulative and
irreversible. Most animal studies dealing with chronic toxicity
have been designed this way, and in the assessment of chronic
effects in humans, the dose is usually estimated from exposure
conditions and integrated over the presumed exposure time.
Exceptions however, include studies providing the basis for some of
the ambient air quality standards (e.g., ozone) where human dose-
response data derived from acute exposures have been used to
estimate the dose-response relationship and, more importantly, the
no-effect level for chronic exposure (Lippmann, 1989) . While this
approach has its uses, one must not forget that it ignores the
possible influence of the duration of exposure. Thus, estimation
of chronic dose-response relationships is extremely complex for
many reasons, not the least of which the influence of time.
There is a substantial body of knowledge on the pathogenesis,
evolution and eventual outcome of chronic diseases in man. Cases
in point include chronic obstructive lung disease, ischemic heart
disease, certain degenerative lesions of the central and peripheral
nervous systems, infectious diseases and the natural history of
many cancers. Understanding of the relationship of exposure to
environmental agents and causation of disease is fragmentary,
however. Even for experimental animals, there is a comparative
paucity of descriptive, let alone mechanistic, information on many
28
-------
of the relevant disease entities. This is paralleled by the
limited database on the toxicokinetics of most chemicals under
conditions of chronic, low-level exposure. Few if any chronic
studies address questions such as recovery of tissue damage or
cellular repair mechanisms or the effects of intermittent-versus-
continuous exposure conditions, factors that may well be critical
determinants in inducing chronic disease states. In animal models
generally, and in toxicology in particular, chronic disease is not
thoroughly studied. Cancer may be the exception, at least in the
pre-oncogene area.
Few experimental studies address the question of what happens
once exposure to a given chemical ceases. Yet we know from the
epidemiology of cigarette smoking that cessation of exposure may
dramatically alter the risk of developing what might be an
otherwise unavoidable outcome. Furthermore, chronic dose-effect
estimates often fail to consider the importance of dose rate. It
is generally assumed that chronic effects are proportional to the
cumulative dose integrated over time. It is conceivable, however,
that the rate at which exposure to a chemical occurs is more
important in determining effects than is total cumulative dose. In
low level ionizing radiation studies, dose-rate is an important
determinant of the induced effects (Upton, 1984). It may become
equally important to consider the role of dose-rate in assessing
risk from exposure to such environmental agents as, for example,
the criteria air pollutants.
Estimations of chronic dose-response relationships are usually
based on the assumption that the toxicants in question act alone.
Yet a given chemical may cause no untoward effects unless a second
insult is superimposed. Most human exposures involve complex
mixtures, but there are few data on the nature and magnitude of
toxicological interactions between individual components (Waters et
al., in press; Vainio et al., in press). Epidemiologic data on
interactions which may modify risk estimates for cancer are limited
to smoking in conjunction with asbestos, radon, and nickel (re-
spectively) . Animal experiments have shown significant interac-
tions (e.g., between carbon tetrachloride and certain alcohols and
between cancer initiating and promoting agents). However, the
database here is limited as well. Thus, although the NOEL's or
AID'S for the latter usually include a safety factor of 100 or
1000, it is not known whether the effect of interactions, in
combination with the many other variables theoretically "covered"
29
-------
by safety factors, will exceed those margins of safety. On the
other hand, some of the National Ambient Air Quality standards,
such as those for ozone, have little or no margin of safety (CASAC,
1989, Lippmann, 1989) and modest degrees of interaction may be very
important.
While efforts to estimate cumulative exposure through the
measurement of biomarkers constitute a promising approach, major
uncertainties in their utility continue to exist. Ideally,
measurement of a biomarker of exposure dose or effect should
provide an index of total exposure over a period of time. In
pharmacokinetic studies, by comparison, it is not possible to
estimate total exposure from one single measurement. Whether this
will become possible with biomarkers remains an open question. Two
major problems to be resolved concern the biological half-life of
each biomarker (there is little information on the relevant repair
and recovery mechanisms) and the extent to which a given biomarker
is predictive of a subsequent biological effect.
3.3.2.4 Summary
Some of the problems involved in the interpretation of dose-
effect or relationship in risk assessment can be summarized as
follows:
a) Cancer Although the database on dose-response relation-
ships for carcinogens that has been obtained from both animal and
human studies is comparatively strong, there is considerable
uncertainty on how to extrapolate from high doses to low doses and
from animals to humans. Currently accepted opinion recommends that
a non-threshold model be used for assessing the carcinogenic risk
that any chemical or physical agent may pose to the general pop-
ulation, but the validity of this model remains to be determined,
as does the particular form of the model and species scaling factor
that may be appropriate for a given form of cancer and for a given
carcinogen (Upton,1989). At intermediate-to-high dose levels,
effects on cell proliferation kinetics may "promote" or otherwise
enhance carcinogenesis in ways that do not occur at lower dose
levels, thus complicating extrapolation to the low-dose domain.
Unfortunately, it may never be possible to prove the absence or
existence of thresholds as demonstrated by the so-called "ED01"
study involving more than 20,000 rodents which was unable to
confirm the shape of the low dose-response curve below a one
30
-------
percent tumor incidence (Littlefield et al., 1979). In short, the
scientific community is divided in its views of the feasibility of
quantitative risk estimation for cancer, owing to the uncertainties
involved.
b) Non-cancer The database for non-carcinogenic effects is
extensive for acute exposures but much less extensive for chronic
exposures. No conceptual problem precludes recognition and/or
assumption of experimental thresholds for many such effects;
however, as discussed above, major gaps in knowledge exist, con-
cerning population thresholds. As far as animal studies are
concerned, few chronic studies have been designed to deal with
endpoints other than cancer. Also, there is a paucity of chronic
toxicity data on the reversibility of the reaction process and on
the importance of the dose rate in relation to the total cumulative
dose. Human studies have been useful in detecting and confirming
some types of health hazards, but the observations are often
difficult to interpret because of scanty information concerning
over-all dose, tissue dose, dose rate, existence of multiple
endpoints, exposure, to additional chemical or physical agents,
preexisting disease conditions, and other variables which contrib-
ute to interindividual variations. Given the existing gaps in our
knowledge, caution should be exercised in qualitative and quan-
titative risk estimates.
Finally, consideration must also be given to the question as
to how risks for non-cancer health effects are best elucidated. A
possible strategy has been suggested by Doll and Peto (1981) for
cancer. In their landmark paper, these authors discussed two
possible strategies to explore the etiology (and, hence, risk) of
cancer: the "mechanistic" strategy, that investigates the biology
of cancer in order to make predictions, and the "black box"
strategy that identifies the cancers that occur in the population
and then looks for epidemiological clues as to their etiology. In
the view of the authors, the "black box" approach was considered to
be more likely to yield important clues quickly. It might be
appropriate, therefore, to conduct a similar analysis of "environ-
mentally caused," non-cancerous diseases, although the vital
statistics for such diseases are relatively incomplete in com-
parison with those for cancer. Moreover, Doll and Peto themselves
acknowledge the uncertainty in their "guesstimates" of the percent
of human cancer attributable to various sources. In the absence of
more complete information, the contribution of environmental
31
-------
toxicants to the total burden of illness in the population will
remain highly uncertain, as will the corresponding risk as-
sessments, such as those presented in the UB report.
3.3.3 Assessment of Severity of Impact
3.3.3.1 Introduction
The 1987 UB report, addressed four major types of risks:
cancer risks, non-cancer health risks, ecological effects, and
welfare effects. The report provided only a brief rationale for
these categories; the four types of health and environmental risks
were considered to be "major" and to be risks that were in
existence at the time the report was prepared. No attempt was made
to rank these four types of risk qualitatively or quantitatively
against each other.
The cancer risk considered by the Cancer Risk Work Group was
apparently overt malignancy and not intermediate indicators of
carcinogenesis, such as dysplasia or metaplasia of epithelial
membranes. The coverage of the Non-Cancer Risk Work Group was
broad and included eleven types of effects: cardiovascular,
developmental, hematopoietic, immunological, kidney, liver,
mutagenic, neurotoxic/behavioral, reproductive, respiratory, and
"other." The effects addressed by this group were heterogeneous,
including indicators of exposure (e.g. mutagenicity), indicators of
injury (e.g. lung injury), and the presence of frank disease and
even death. Through the application of a ranking of organs with
regard to importance to life and of the severity of the endpoints,
an attempt was made to provide an overall ordinal grouping of the
endpoints. The Welfare Risk Work Group also addressed a variety of
effects, including aesthetic values.
In considering the approach used to develop the 1987 UB
report, the overall choice of the four major risk categories noted
earlier remains appropriate as does the decision to avoid a
comparative ranking of the four types of risk. A process for
establishing the ranking has not been developed, and appropriate
criteria would not be wholly scientific or medical but would in-
corporate prevailing social values. For the health risks, place-
ment of cancer and non-cancer risks within a single framework
appears theoretically feasible, using indices of severity common to
all diseases, such as extent of interference with function or
32
-------
probability of mortality, as indices for comparison. However, one
must question such functional and clinical criteria as not covering
the full spectrum of adverse effects (i.e., excluding subtle pre-
clinical manifestations of disease occurring at earlier time-points
along the continuum between exposure to toxic substances and
clinical diseases). As mechanistic information becomes available,
it is likely that earlier occurring molecular or biochemical
changes (such as alterations in oncogenes or enzyme inhibition in
neurologic disease) , will supplant conventional endpoints, allowing
a preventive approach in priority-setting. Therefore, "inte-
rference with function" should be broadly defined to include
biochemical or molecular alterations established as indicative of
the disease process. This section considers measures of the impact
of environmental risks on individuals and on populations.
3.3.3.2 Impacts on Individuals
The effects of environmental pollutants on individuals may be
assessed on distinct axes that measure effects such as comfort,
functional status, and exposure status. While these axes overlap
to an extent (e.g., the presence of disease necessarily signals the
presence of a disease process), they offer a multidimensional
framework for considering the impact of pollutants.
The relative risk of disease—that is, the rate of occurrence
of disease in exposed persons, as compared with that in non-exposed
persons—is the most widely applied measure of impact on individ-
uals. The risk associated with exposure may also be expressed as
the cumulative lifetime probability of disease, and contrasted with
the lifetime risk in the absence of exposure. For individuals, the
strength of the exposure-disease association is measured by the
divergence of the relative risk from the no-effect value of unity.
Small increments of risk, perhaps a few percent to about 20
percent, are not detectable in epidemiological studies because of
statistical uncertainty. Thus, epidemiological data have generally
provided direct evidence for adverse effects at increments of
relative risk of about 50 percent or more. The consequences of
exposures associated with lower levels of relative risk are often
estimated by extrapolation. In considering the risks for specific
individuals, factors determining susceptibility must be addressed,
as specific host characteristics or exposure to other agents may
have significant interactions with the exposure of interest from
the biological and public health perspectives.
33
-------
The occurrence of disease, or even death, in association with
exposure to an environmental pollutant provides an unarguable
indication of effect. On the other hand, a pathophysiological
process may be detectable, even though overt disease is not
present. For example, lung function decline over time in excess of
the usual loss associated with aging might be detected in an
individual who has no evidence of overt lung disease. Similarly,
bronchoalveolar lavage may show an inflammatory response in
asymptomatic subjects exposed to an inhaled pollutant (National
Research Council, 1989).
Biological markers of exposure dose and effect represent
another approach for characterizing the effects of pollutants on
individuals (National Research Council 1989) . Markers of exposure
indicate only that an agent has entered a physiological compartment
and their detection does not signal the presence of disease or
necessarily of injury. Some markers have sufficient sensitivity
and specificity to identify exposed persons with a high degree of
certainty. For example, cotinine, the major metabolite of
nicotine, can be readily measured in blood, saliva, and urine.
High levels are produced by active cigarette smoking, whereas low
levels may result from involuntary exposure to tobacco smoke;
nonsmokers without any involuntary exposure to environmental
tobacco smoke do not have detectable levels of cotinine in body
fluids. A particular level of cotinine does not imply that a
smoking-related disease has occurred or will occur. By contrast,
a certain level of blood lead not only is indicative of exposure
but likely predictive of disease.
Functional status provides an overall measure of the impact of
exposure; the potential dimension of effect spans from minimal
interference with performing one's job and activities of daily
living to severity disability and death. Effects on well-being
have become an increasingly prominent concern of the public. The
range of impacts is broad, potentially including concern over
aesthetic degradation of the environment, changes in behavior, and
fear over the potential consequences of real or perceived exposure.
For any of these dimensions, the distinction between "adverse"
and "non-adverse" effects needs to be made. The judgment concern-
ing adversity incorporates not only medical criteria, but the pre-
dominant societal values at the time the decision is made. Thus,
judgments as to the adversity of effects should not be regarded as
34
-------
fixed, but as subject to change with social, political, and
economic conditions.
For atmospheric pollutants, the language of the Clean Air Act
forces consideration of the nature of an adverse effect health.
For the criteria pollutants, the Administrator of the Environmental
Protection Agency must set national primary ambient air quality
standards that will protect the public health with "an adequate
margin of safety." Section 112 of the 1977 Amendments requires the
Administrator to regulate "hazardous air pollutants," those not
covered by the primary standards but "... which may reasonably be
anticipated to result in an increase in mortality or an increase in
serious irreversible, or incapacitating reversible illness." The
Clean Air Act does not explicitly define adverse effects on public
health and welfare.
This ambiguity in the Clean Air Act has prompted both in-
dividuals and organizations to consider criteria for determining
adverse health effects (Ferris 1978; Higgins 1983; American
Thoracic Society 1985). Ferris (1978) noted the judgmental nature
of this determination and the difficulty of achieving consensus for
many effects. Higgins (1983) defined an adverse health effect as
"... a biological change that reduces the level of well being or
functional capacity." The report of an American Thoracic Society
committee (1985) on adverse respiratory health effects turned to
"medical significance" as the criterion for determining the
adversity of an effect. The committee provided a hierarchical
listing of potential respiratory effects without making a specific
demarcation between adverse and non-adverse; the range of effects
was from increased mortality to odors.
The issue of separating adverse from non-adverse health
effects remains topical and arises throughout this report. The
development of increasingly sensitive markers of exposure dose,
preclinical effect and injury can result in the identification of
potential effects of uncertain biological significance. The
efforts of the American Thoracic Society and others to develop
methodologies for establishing the adversity of health effects need
to be continued. With a goal of prevention, there is a strong
rationale for using the most sensitive indicators of early response
that can be identified.
35
-------
Epidemiological study designs are the principal approach used
to directly characterize the effects of environmental pollutants on
individuals (National Research Council 1985). For environmental
pollutants, the most widely used designs are the descriptive study
or survey, the case-control study, and the cohort study. While
epidemiological studies have the advantage of directly examining
disease risks in human populations, epidemiology has potential
limitations that may constrain the interpretation of epidemiologi-
cal data on environmental pollutants. Exposures of individuals to
pollutants may be difficult to accurately measure; the resulting
misclassification of the exposures of individuals may bias the
results of studies towards not finding associations between
exposure and disease. Moreover, many of the acute and chronic
diseases of concern with regard to exposure to environmental
pollutants are multifactorial in etiology. To accurately describe
the effects of pollutant exposure, it is necessary to carefully
measure and control for the effects of the other factors, e.g.
cigarette smoking, and to consider interactions of the pollutant of
interest with other factors.
This section considers the effects of environmental pollutants
on individuals; it proposes dimensions along which these effects
can be gauged as a basis for merging the diverse health endpoints
along a single spectrum; and it considers the approach of the 1987
Unfinished Business report in this framework.
3.3.3.2.1 Exposure status
The continuing evolution of approaches for assessing exposure
has led to increasing accuracy and sensitivity in the estimation of
human exposures to environmental pollutants. Through the 1980s,
estimates of exposure were often based on questionnaires, on
measurements of pollution in large geographical regions, or on
other surrogates for personal exposure. However, the development
of new biological markers of exposure dose and effect, of personal
exposure monitors for some pollutants, and of methods for charac-
terizing the exposures of individuals in specific micro-environ-
ments has provided potentially more accurate measures that can be
used to complement the older approaches (National Research Council
1989; Spengler and Soczek 1984; Wallace and Ott 1982).
The Non-Cancer Risk Work Group mingled mutagenicity, an
indicator of exposure (or internal dose), with other health
36
-------
endpoints in the 1987 report, but did not consider other exposure
measures. More appropriately, markers of exposure should be
handled independently from disease indicators. However, as
discussed below, the consideration of risks for individuals and
populations should be broadened to include markers of exposure.
3.3.3.2.2 Disease Status
Many of the endpoints considered by the Non-Cancer Risk Work
Group were indicators of response to environmental pollutants, and
not of overt disease that would produce symptoms or lead to a
clinical diagnosis. For example, the list of endpoints included
increased levels of liver enzymes, reduced corneal sensitivity,
pulmonary irritation, nasal cellular irritation, and decreased
midexpiratory flow rates. Many of the endpoints were histopatho-
logical abnormalities: tubular degeneration, hyperplasia, and
hypertrophy of the kidney, histopathological alterations of the
liver, and giant cell formation in the testes. While these
endpoints provide clear indications of damage to target tissues,
interpretation must be placed in the context of the relationship
between each endpoint and the likelihood of developing disease.
Measures of disease process should be handled separately from frank
disease.
Indicators of disease status may be variably based on the
presence of a clinical diagnosis, a specific physiological
parameter (e.g., the diagnosis of anemia is based on reduction of
the hematocrit, or another test. For some endpoints, the degree of
diagnostic certainty is generally high, e.g. lung cancer or
myocardial infarction, and the implications of the diagnosis for
functional status and mortality are well characterized, e.g.
Legionnaires' disease. The Non-Cancer Risk Work Group also
addressed exacerbation of the status of persons with established
disease, such as ischemic heart disease and asthma. Many of the
endpoints considered by the Non-Cancer Risk Work Group were
measures of effects that represented the final outcome of exposure,
but were not disease states, e.g. low birth weight, oligospermia,
and mucosal atrophy of the respiratory tract.
The heterogeneity of disease effects considered in the
Unfinished Business Report was recognized by the Non-Cancer Risk
Work Group. An attempt was made through the Toxicity Test Endpoint
Severity Scores to rank the effects; an attempt was not made to
37
-------
establish boundaries between adverse and not adverse. This aspect
of the report would have been strengthened by a sharper separation
of the various types of endpoints with a clearer demarcation
between the causation of disease, the exacerbation of disease, and
other effects. Within the category of effects, however, it is
important to view individual endpoints as occurring along a
continuum and to select those representing early sensitive effects
of environmental toxicants.
3.3.3.2.3 Functional Status
The Non-Cancer Risk Work Group included measures of functional
status among the other endpoints: pulmonary impairment, retarda-
tion, and learning disabilities. Functional impairment is a con-
sequence of disease and the degree of impairment might have more
appropriately been linked to the causative disease. This aspect of
the 1987 report merits reconsideration and expansion.
3.3.3.2.4 Welfare Effects
The range of welfare effects is wide and reflective of
societal responses to environmental degradation by pollution. The
potential links between welfare and health effects should not be
dismissed; noticeable environmental changes resulting from
pollution or even the perception of exposure to pollution could
have adverse impact by forcing behavioral modification (e.g.
forgoing activities out-of-doors), altering mood, or causing stress
(Evans et al. 1988). The public's increasing expectations of
living in a risk-free environment undoubtedly fosters the potential
for welfare effects.
3.3.3.2.5 Functional Effects
The capability of performing one's work and leisure activi-
ties, as well as routine activities of daily living, integrates
both behavioral and non-behavioral consequences of pollutant
exposure. At the extreme of adverse effects, impairment or even
death obviously impact functional status; at the other extreme,
subtle psychological effects may interfere with performance of
routine tasks with consequences such as reduced productivity and
increased absenteeism. This dimension of pollutant effects was not
addressed in the 1987 Unfinished Business Report.
38
-------
3.3.3.3 Impacts on Populations
The risks for populations should be addressed separately from
the risks for individuals, although the dimensions of effect
considered above for individuals remain relevant to populations.
For individuals, concern focuses on the likelihood of developing
disease following exposure; the relative risk indicates the
strength of association at the individual level. The population's
burden of disease integrates the distribution of exposure, the
inherent susceptibility of the population, and the level of risk
associated with exposure. It provides a measure of impact
complementary to individual measures, such as the relative risk.
From the public health perspective, exposures associated with
relative risks that are of acceptable magnitude to many individuals
might yield unacceptable disease burdens for the population for the
population as a whole. Or, conversely, as in the case of benzene
air emissions, there may be a small number of excess cancer cases
nationwide accompanied by high individual risks.
Epidemiological data can be used to describe directly the
population's burden of disease associated with an exposure. The
population attributable risk estimates the proportion of disease in
a population resulting from exposure (Rothman 1986) ; its cal-
culation requires information on the distribution of exposure in
the population and on the excess relative risk associated with
exposure. Risk assessment techniques can also be used to project
the burden of disease caused by an environmental pollutant
(National Research Council, 1983).
The distinction between individual and population risks was
explicitly recognized by the Cancer and Non-Cancer Risk Work
Groups. Both emphasized the population perspective, an approach
that seems appropriate in light of the Environmental Protection
Agency's public health charge. However, agents placing a small
number of susceptible persons at particularly high risk also merit
emphasis.
3.3.3.4 Synthesis
The selection of endpoints to characterize the impact of
environmental pollutants on individuals and on populations poses a
panoply of difficult choices. Exposure-effect relations need to be
postulated and measurement approaches devised to validly detect the
39
-------
anticipated effects. Choices must also be made among the various
dimensions along which effects can be measured. For regulatory and
risk management purposes, it may be necessary to separate "adverse"
from "non-adverse" health effects or to rank effects in terms of
overall impact. It may also be necessary to make judgments
concerning the relative impacts of pollutant exposures on individu-
als and on populations. How can risks incurred by individuals be
balanced against the population's burden of disease? The overall
disease burden in a population may be the same for a rare exposure
associated with a high relative risk as for a prevalent exposure
associated with a low relative risk. [A definition of the term
"adverse" and judgments about severity of effects are not purely
scientific issues but involve consideration of social and ethical
factors as well as public perception.]
Many of these issues were directly confronted in the Un-
finished Business Report. The Work Groups left unsolved the
difficult problem of grouping the various endpoints into a common
framework. As discussed elsewhere in this report, this committee
considers that this challenging task must be addressed. Potential
scales for qualitatively ranking the various endpoints include the
probability of developing disease (for exposure status or disease
process), the degree of associated impairment (for disease status
and welfare effects), and the probability of death (for exposure
status or disease status). Ideally, these should be calculated for
both the general population and for the most sensitive subgroups.
3.3.4 Susceptible/Critical Subgroups
3.3.4.1 Introduction
In principle, efforts to characterize comprehensively the
risks of environmental hazards to human health should consider the
potential effects of variations in susceptibility among individu-
als, an issue that was not addressed in the 1987 UB report.
If one could depict the response of the entire U.S. population
to each potentially hazardous environmental exposure, there would
be, for each, a distribution characterized by individuals at each
extreme of susceptibility. Because this report addresses public
health, it focuses on those individuals who are the most, rather
than least, susceptible.
40
-------
One explanation for the increased susceptibility of some
individuals, of course, is the "random" variation in physiologic
make-up that is inherent in all biological systems. "Random", in
this sense, refers to a quality of unpredictability.
Some variations, however, are predictable. They may be
associated with an identifiable physiologic perturbation (e.g. an
enzyme deficiency), and/or they may fall along distinct sexual,
racial, ethnic, or other lines. The consequences of such varia-
tions are that the susceptible individuals receive a greater burden
of risk in a risk area.
In an era in which we, as a society, are striving for racial
and sexual equality, as well as sensitivity to the needs of the
elderly and disabled, the prospect of achieving equal protection
for all calls for sensitive scrutiny. In effect, the EPA ack-
nowledged the need to focus on special populations at risk when it
designated a separate risk area for occupational diseases. This
designation constituted acknowledgment of the disproportionate
burden of risk placed on certain occupational groups because of
higher exposures to many hazards. It follows logically that we
should make a distinction for other groups that bear a dispropor-
tionate burden of risk because of an identifiable susceptibility or
consistent pattern of unequal exposure, as in the case of inner-
city residents and lead, or rural fish consumers downstream from
paper mills.
3.3.4.2 Types of Susceptibility Variations
We advance here the concept of biological susceptibility
versus susceptibility due to social/behavioral factors.
3.3.4.2.1 Biological Variations in susceptibility
This can be defined as susceptibility because of host factors
(endogenous factors) that heighten an individual's risk of
toxicologic injury to a given environmental exposure. Examples:
(a) Pregnant women, the fetus, and the nursing infant:
(1) The developing fetal and infant nervous system is
extremely sensitive to the effects of lead, which
41
-------
freely crosses the placental barrier and is secreted
into breast milk.
(2) Other exposures that have been associated with birth
defects include mercury/ and PCBs. Embryonal and
fetal tissues are extremely sensitive to ionizing
radiation, especially during critical stages of
organogenesis.
(3) A number of toxicants are actively or passively trans-
transferred from plasma to breast milk, including
mercury, cadmium, DDT, PCBs, and related halogenated
hydrocarbons.
(b) Race or Ethnicity Factors:
Light-skinned whites, particularly those who tan poor-
ly, are at greater risk for UV-induced skin cancer
(Silverstone et al., 1970).
(c) Elderly: By virtue of their physiology, the elderly are
more susceptible to factors that affect the immediate
physical surroundings. The higher prevalence of chronic
diseases experienced by the elderly also indirectly in-
creases their risk to many of the hazards listed earlier.
(d) Children:
(1) In general, children can be seen as being more sus-
ceptible to toxins that require an extended latency
time in order to express their effects, such as
carcinogens.
(2) The developing systems of children are generally view-
ed as more vulnerable than those of adults, as il-
lustrated by the exquisite sensitivity of children's
nervous systems to the toxic effects of lead.
(e) Chronic disease or other medical conditions:
(1) Asthmatics: a broad range of air pollutants adversely
affect persons with asthma.
42
-------
(2) Coronary Heart Disease: individuals with pre-existing
coronary heart disease have increased susceptibility
to exposure to carbon monoxide and noise (increas-
ed stress leading to increase in blood pressure,
heart rate, circulating catecholamine and lipids.
(3) Chronic liver disease: decreased ability to detoxify
and increased susceptibility to a number of toxins,
including chlorinated hydrocarbons, halogenated
aromatics, etc.
(4) Malnutrition: a diet deficient in calcium, magnesium,
iron, or protein leads to increased dietary lead
absorption (Mahaffey, 1981)
(5) G-6-PD deficiency: more prone to methemoglobine-
mia
(e.g. from nitrate-contaminated well water, food high
in nitrates or nitrites)
(6) Decreased delta aminolevulinic acid dehydrase enzyme
activity (d-ALA polymorphism): more prone to toxic
effects of lead
(7) Alpha, antitrypsin deficiency: more prone to the pul-
monary effects of tobacco smoke, and grain dust (Chan
Yeung, 1978).
(8) Decreased activity of N-acetyltransferase: increased
susceptibility to environmental bladder carcinogenesis
(Cartwright, 1982).
3.3.4.2.2 Susceptibility Variations Due to Social or
Behavioral Factors
For a given risk area, particular population sub-groups are at
increased risk because of social/behavioral factors that dispropor-
tionately increase their exposure. "Occupational groups" can be
considered a category within this framework.
Geographical factors are also extremely important with respect
to many hazards, such as living in high altitude or equatorial
43
-------
regions which increases UV radiation exposure from ozone depletion.
These distinctions are, in most cases, self-evident. In addition,
for many of the EPA-defined "Risk Areas", geographical considsider-
ations are implicit in the construction of the Risk Area. For
instance, individuals living on the coast in an industrial area are
obviously the most susceptible to "Direct discharges to surface
water"; likewise, exposure to "hazardous air pollutants" follows
well-defined geographical distributions. Therefore, geographical
factors are not considered directly within this framework unless
they are accompanied by other factors that help distinguish
particular population sub-groups (e.g., see section (2) on race
below).
Examples:
(1) Lifestyle factors (particularly with respect to cancer):
Alcohol Ingestion: alcohol has been shown to enhance the
toxicity (primarily heptotoxicity) of several halogenated
hydrocarbons, including carbon tertrachloride, chloroform, and
methylene chloride (Hills and Venable, 1982). A synergistic
interaction between alcohol ingestion and inhaled vinyl
chloride for the induction of angiosarcoma of the liver has
been reported in rats (Radike, et al., 1977).
Cigarette Smoking: increases cancer risk from radon and as-
bestos exposure (see Appendix section 8.1.2). It may also
increase cancer risk from arsenic exposure (Steenland and
Thun, 1986). There is a suggestion that heavy urban air
pollution can add to the risk for lung cancer in smokers
(Jadrychowski, 1983).
Dietary habits: (excluding malnutrition—see section 3.3.4.2.1
(e) (4)) in epidemiologic studies, intake of specific nutri-
ents has been associated with varying risks for cancer, e.g.,
intake of vegetables and fruits has been inversely related to
the risk of lung cancer in many studies, perhaps through the
protective effects of beta-carotene (Willet, 1990). Nutri-
tional intake may also modify the human response to environ-
mental carcinogens, but little data is yet available to
evaluate this possibility.
44
-------
Sun exposure habits: total sun exposure increases the risk of
non-malignant skin tumors and cataracts. "Bursts" of sun
exposure increases the risk of malignant melanoma.
(2) Socio-economic Factors:
(a) Rural Hispanics are disproportionately exposed to pesticides
due to their concentration in agricultural jobs, and residence
in areas heavily exposed to pesticide spraying.
(b) Due to their concentrated residence in urban areas with
deteriorating, old housing stock, African Americans and
Hispanics are disproportionately exposed to lead (from lead
paint).
(c) Some groups live in subsistence economies relying heavily on
fish. They would be disproportionately susceptible to hazards
involving discharge pollution of estuaries, coastal waters,
oceans, wetlands, surface water, etc.
3.3.4.3 Identifying Susceptible Subgroups According to Hazard
of EPA "Risk Area"
It is difficult to append sections on special susceptible
populations to the existing structure that EPA chose for organizing
this risk reduction exercise. Some of the EPA-defined "Risk Areas"
are very broad, and have a considerable amount of overlap with
regards to specific toxins. Other Risk Areas are poorly defined
with respect to specific substances.
3.4 Treatment of Uncertainty
From the foregoing, it is apparent that the assessment of
environmental risks to human health is complicated at virtually
every step by potentially large uncertainties in: 1) numerical
values of measurement or other quantities affecting the risks; 2)
the modeling of exposure and/or toxic responses; 3) temporal,
spatial, and inter-individual variations in susceptibility; and 4)
the quantification and comparison of societal and personal measures
of risk. To the extent that the utility of a risk assessment may
be limited by any or all of these uncertainties, each should be
addressed explicitly in the design, conduct, interpretation, and
reporting of the assessment. The relevant problems, many of which
45
-------
are discussed in other sections of this report, have been consid-
ered in further detail elsewhere (eg, Finkel, 1990; Zackhauser and
Viscusi,1990).
3.4.1 Parameter Uncertainty
Uncertainty in the numerical values of quantities affecting
risks may result from: 1) errors in measurement, owing to impreci-
sion in instruments or human mistakes; 2) misclassification of
data; 3) random or sampling error; and 4) systematic errors in
data-gathering or analytical techniques. Each of these sources of
uncertainty has its own causes, the remedies for which must be
addressed specifically.
3.4.2 Model Uncertainty
In modeling exposure patterns or response to toxicants, error
may result from: 1) failure to measure or include the correct
quantities (e.g., "surrogate" variables); 2) exclusion or faulty
treatment of significant (e.g., confounding) variables; 3) use of
a model that is not of the correct form or structure (a major
controversy in environmental risk assessment has concerned the
selection of the appropriate model for estimating the risks
attributable to low-level exposure to carcinogens; predictions
derived from different models may differ by many orders of
magnitude) (Krewski and Van Ryzin, 1981).
3.4.3 Uncertainty Due to Inter-individual Variability
As noted above, inter-individual variations in exposure
patterns and in susceptibility may be due to age, sex, occupation,
socio-economic status, dietary practices, smoking habits, life-
styles, and other influences. For most environmental toxicants,
knowledge of the effects of these variables on human susceptibili-
ty, and to a lesser extent on exposure, is still limited. As a
consequence, risk assessments applicable to human populations
involve uncertain assumptions about the distribution of differences
among individuals.
3.4.4 Uncertainty in Quantifying and Comparing Measures of Risk
Because risks can be expressed in different ways, which
determines how they are perceived, there is frequently uncertainty
46
-------
in the choice of the appropriate measure of risk to use in
comparing different risks. If, for example, years of life lost
were considered as a relevant measure of risk, then a fatal effect
in a young person might be given more weight than the same effect
in an older person. Similarly, a 10"4 lifetime risk of death would
predict no attributable fatalities in a population of 1,000
persons, but 25,000 attributable fatalities in the U.S. population
as a whole. When the comparison among risks involves different
kinds of health effects—e.g., cancer vs. mental retardation—the
problem is complicated even further. Because ambiguity in the
criteria for deciding which measure of risk is appropriate in a
given situation will lead to uncertainty in the assessment,
decision rules for addressing the problem have been proposed (e.g.,
Milvy, 1986; Machin, 1990).
47
-------
4.0
Although the risk to health from any given environmental
toxicant can be lowered by reducing the extent of exposure to the
toxicant, some toxicants—e.g., carcinogens—cause effects that may
not become manifest until years or decades after exposure. With
mutagens, likewise, the heritable damage to reproductive cells may
affect offspring of the exposed person many generations later. In
the case of toxicants causing such delayed health outcomes,
therefore, cessation of exposure does not abolish risk immediately.
By the same token, stopping the release of a toxicant at its
source does not suffice to prevent exposure to any levels of the
toxicant that may have been previously released to the environment.
In the case of long-lived toxicants, such as heavy metals, PCBs,
asbestos, and long-lived radionuclides, indefinite persistence in
the environment and the possibility of bioaccumulation in the food
chain further complicate current risk-reduction efforts, as does
persistence in the tissues of persons who may already have been
exposed.
In light of the foregoing, evaluation of the reducibility of
the environmental risk posed by a given environmental toxicant must
take into account the time over which different risk-reduction
strategies may be effective; the potential for long-term risk
reduction must be weighed along with that for short-term reduction.
48
-------
5.0 Reviav of The Health Ri?fc Ranfcinqa in
*yha "Unfinished Business Report"
5.1 Methodology
In the "Unfinished Business" Report, as noted above, various
environmental problems were examined for their risks to the health
of persons residing in the U.S. Two categories of risk were
considered: 1) risks of contributing to the occurrence of cancer
and 2) risks of causing other adverse health effects. The
information used in that report was not based on new research
undertaken expressly for the purpose, but was extracted from risk
assessments conducted previously by EPA in support of other Agency
activities.
The 31 environmental problems considered in the "Unfinished
Business" report were selected primarily on the basis of their
relevance to the Agency's regulatory mandates and programmatic
organization. Because they included various sources of pollution,
various pollutants themselves, various exposure media, and various
situations involving human exposure (Table 2.1), their diversity
complicated ranking them for their relative risks, as discussed
below. The ranking was also complicated by inconsistencies in the
methods and assumptions that had been used by the Agency in its
earlier assessments of the different problems.
For virtually all problem areas, the risk assessments were
severely limited by uncertainty about: 1) the relevant extent of
human exposure (in some instances, the assessments were based on
only a small percentage of pertinent chemicals); 2) the toxicity of
the agents in question (eg, NAS, 1984) ; 3) the appropriate dose-
response models to use for estimating the risk relevant to ambient
exposure levels; 4) the extent of variations in susceptibility with
species, age, sex, and other variables; 5) and the extent to which
the relevant dose response(s) may be modified by exposure to other
chemical or physical agents. All numerical estimates of numbers of
individuals harmed need more careful examination to determine
consistency and comparability for risk ranking purposes.
5.2 Rankings for Risks of Cancer
In spite of their large uncertainties, the estimated risks of
cancer posed by the different problem areas were ranked in
49
-------
Environaantat.
Probtea
Cateflorv 1 (Hit* Kfsk)
Worker exposure
(*3D
Indoor radon (ft)
Pesticide residues
in foods (125)
Indoor air
frm.* fttf^^n^ f-f^\
\non-raoon) \nj
Exposure to
consumer product* (*30)
Other hazardous
air pollutants (ffZ)
Category 2 (MediuM-to-Hiah
Depletion of
stratospheric ozone (7)
Hazardous waste
sites (inactive)
Application of
pesticides (*26)
Radiation other
than radon) (46)
Other pesticides
risks (*27)
Hazardous waste
sites (active) (f16)
Industrial waste
(non-hazardous) (f19)
Hew toxic chesricals (.KB)
Rank
Order
1
(Tied)
1
(Tied)
3
4
(Tied)
4
(Tied)
6
Risk)
7
8
9
10
11
12
13
14
15
Estimated Magnitude of Risk
250 cancers annually attributable to only four of the •any cheaical
carcinogens in question. Risks to individuals a*y be high.
5,000-20.000 lung cancers annually. Risks to individuals say be
high.
6,000 cancers annually, based on asinnnsynt of only 7 of 200
potentially oncogenic pesticides.
3,500-6,500 cancers annually (priawily fro* tobacco smoke). Risks
to individuals say be high.
100-135 cancers annually fro« only 4 of the sore than 10,000
cheaicals in consiBW products.
2,000 cancers annually fro* only 20 of the eany pollutants in air.
Risks to individuals say be high.
Possibly 10,000 cases annually by the year 2100.
More than 1,000 cases annually.
400 to 1,000 cancers annually
100 cancers annually in CM 1 1 population exposed. Risks to
individuals can be high.
360 cancers annually, largely fro» building Materials. Risks to
individuals can be high.
150 cancers annually. Estimate highly uncertain.
Probably fewer than 100 cases annually. Risks to individuals can
be high.
No quantitative estimate, but judged less severe than hazardous
waste sites.
Mo quantitative estimate possible, but judged to pose moderate
risks.
Table 5.2 "Unfinished Business" Report High and Medium-to High cancer
risk rankings for the identified environmental problem areas
numerical order, problems estimated to pose the highest risks being
assigned to category 1 and those judged to pose smaller risks being
assigned to lower categories. Table 5.2 displays the UB report's
category 1 "High Risk" and category 2 "Medium-to-High" assignments.
Although in-depth reassessment or updating of the rankings was not
50
-------
possible within the time that was available to the Subcommittee,
each of the rankings was reviewed with care. For reasons discussed
in previous sections, the rankings in the UB report were considered
by the Subcommittee to be tenuous in view of present limitations in
the methodology and databases needed for quantitative estimates of
the cancer and non-cancer risks attributable to each category.
Salient comments, primarily addressing the "high" rankings,
are summarized in the following section.
5.2.1 Criteria Air Pollutants
Criteria air pollutants were ranked comparatively "low" for
cancer risks in the "Unfinished Business" report, mainly because
the air pollutants that were known to be carcinogens had been
assigned to other problem areas. However, it should be noted that
the same photochemical reaction sequence that leads to ozone
formation in the atmosphere produces a broad range of vapors and
particulates that are known carcinogens. In addition, inhaled
nitrogen oxides contribute to nitrosamine formation in vivo, and
lead is classified as a B2 carcinogen. While no cancer risk has
yet been attributed conclusively to other criteria air pollutants,
mechanistic studies, many of them with in vitro systems, suggest
that ozone and perhaps other criteria air pollutants possess
mutagenic and/or carcinogenic potential (Witschi, 1988). As far as
ozone is concerned, the long-term National Toxicology Program
bioassay of ozone is not yet complete, and earlier studies on the
induction of lung tumors in mice are equivocal. Other studies have
implied that under appropriate experimental conditions carcinogene-
sis in the respiratory tract of the rodent may be enhanced by ozone
(Hasset et al., 1985; also see the Ozone Case Study in section
8.1.1) and, possibly also by S02, although a recent experiment has
failed to confirm the enhancing effects of the latter (Gunnison et
al., 1988).
5.2.2 Hazardous Air Pollutants
This problem area was ranked relatively high for cancer risk,
on the basis of the estimate that 20 of the known human and animal
carcinogens to which people may be exposed by inhalation can be
expected to cause some 2000 cancers annually in the U.S. [It was
also noted that individual risks can be high]. Although additional
51
-------
information has since become available, it does not appear to
affect the over-all qualitative assessment of high risk.
5.2.3 Other Air Pollutants
By definition in the UB report, this problem area, which
included fluorides, total reduced sulfur, and other air pollutants
not assigned elsewhere, excluded all substances posing known or
suspected risks to human health. For this reason, it was not
ranked for either cancer or non-cancer risks to health in the
"Unfinished Business" report. It is noteworthy, however, that
these air pollutants can exact health effects (e.g., sulfuric acid
aerosol, both by inhalation and by mobilizing toxic metals in
drinking water sources). Hence, they should be assessed.
5.2.4 Indoor Radon
A "high" cancer risk ranking was assigned to indoor radon in
the "Unfinished Business" report, based on the estimate that it may
cause 5,000-20,000 lung cancers annually in the U.S. This
assessment, although uncertain, was considered reasonable by the
Subcommittee (see Case Study on Indoor Radon, section 8.1.2).
5.2.5 Indoor Air Pollutants Other Than Radon
This problem area was ranked "high" for cancer risk, on the
basis of the estimate that only seven specific pollutants (tobacco
smoke, benzene, p-dichlorobenzene, chloroform, carbon tetrachlo-
ride, tetrachloro-ethylene, and trichloroethylene) may account for
3,500-6,500 cancers each year in the U.S. population. With the
possible exception of environmental tobacco smoke however, the
relevant exposure and exposure-response relationships are not well
characterized for such pollutants.
5.2.6 Drinking Water
The cancer risk ranking assigned to this problem area in
"Unfinished Business" was "moderate", on the basis of the estimate
that only 23 of the known pollutants, may cause 400-1000 cancers
annually in the U.S. population, most of which are attributable to
radon (30-600) and trihalomethanes (322) . Although the methodology
used to estimate the risks was judged by the Subcommittee to be
reasonable, the estimates must remain highly uncertain in the
52
-------
absence of adequate information about exposure and the relevant
exposure-response relationships. Furthermore, petroleum-related
chemicals, such as benzene, xylene, toluene, and many pesticides,
do not appear to have been considered, although they are found in
drinking water not infrequently, especially in private wells. The
risks should be reexamined using the new exposure data from EPA's
recent groundwater pesticide survey.
5.2.7 Pesticide Residues on Foods
A "high" cancer risk ranking was assigned to this problem area
in "Unfinished Business", based on the estimate that about 6000
cancers per year in the U.S. population were attributable to the
ingestion of pesticide residues on foods. This estimate, derived
from assessing the risks of seven pesticides with oncogenicity for
rodents, was extrapolated to cover all other pesticides in use, on
the assumption that roughly one-third (200) of them were potential-
ly oncogenic. The estimate, although not inconsistent with
independent estimates based on similar methodology (e.g., NAS,
1987), rests almost entirely on uncertain extrapolation of
carcinogenicity data from animal experiments, on fragmentary
information about the extent of human exposure to the pesticides in
question, and on uncertain assumptions about duration-of-life
levels of intake of such substances. The UB analysis contained a
number of simplifications. Limited data on a handful of pesticides
was used to represent the more than 300 pesticides in use on food
crops today. The report assumed that residues of pesticides in
various foods were present at the maximum permissible concentra-
tions (TMRC). It would have been preferable to use the TMRC times
the percentage of crops treated, times consumption, based on the
updated Tolerance Assessment System to indicate an upper bound on
exposure. One should also estimate exposures to both the average
and the most exposed populations (e.g., the infant and young
child). The risk assessment did not include carcinogens such as
methylene chloride, benzene, and vinyl chloride, which in some
cases represent a significant percentage by weight of the relevant
formulations.
5.2.8 Application of Pesticides
This problem area was assigned a "moderate" ranking for cancer
risk in the "Unfinished Business" report, based on the estimate
that 100 cancers in pesticide applicators each year could be
53
-------
attributed to their occupational exposure to carcinogenic pesti-
cides, judging from risk assessments on 6 pesticides found to be
oncogenic in rodents. Although the estimated number of cancers was
small, the risks to individual workers were considered to be high.
Because estimates of the carcinogenicity of pesticides for humans
are based almost solely on uncertain extrapolations from animal
data, the estimates are highly uncertain.
5.2.9 Worker Exposure to Chemicals
The ranking assigned in "Unfinished Business" to this problem
area was one of the highest, based on the estimate that 250 cancers
each year in occupationally exposed workers are attributable to
only four chemicals (formaldehyde, tetrachloroethylene, asbestos,
and methylene chloride) of the more than 20,000 chemicals to which
they may be exposed. Although the total number of all such
occupationally related cancers was not calculated, the risks to
some individual workers were judged to be high. The Subcommittee
considered the UB report's ranking to be reasonable, especially in
view of the fact that the workplace is a source of potentially
toxic agents, so that when exposures occur there, they will tend to
be higher, in general, than in environmental settings outside the
workplace. At the same time, however, the estimates were judged to
rest on relatively fragmentary exposure data and on inadequate
knowledge of the carcinogenicity and carcinogenic potency of most
of the chemicals and combinations of chemicals to which workers are
currently being exposed.
In considering this ranking, the Subcommittee was cognizant
of the previous estimate (Doll and Peto, 1981)1 that 2-8 percent
of all cancers in the U.S. population—namely, 10,000-40,000 fatal
cases per year (with a "best" estimate of 20,000)—may be at-
tributed to occupational exposure to carcinogens, with the
attributable risks of all cancer in the worker population per se
therefore approaching or exceeding 30 percent (Nicholson, 1984)2.
Improvements in worker protection since these analyses suggest that
1See Section 5 reference #2
2See Section 5 reference #16
54
-------
the figures may no longer be applicable. Recent data3 do not show
this high an attribution for all cancers; however, they refer to
only a small proportion of the worker population. While the data
noted above do raise the question as to where cancer deaths occur
(an important consideration for regulation) , they do not, by
themselves, contradict the overall Doll-Peto estimates. It would
be useful if the assessment of carcinogenic risks to workers were
to be updated.
5.2. 10 consumer Product Exposure
A "high" cancer risk ranking was assigned in "Unfinished
Business" to this problem area, based on the assessment that 100-
135 cancers each year in the U.S. population are attributable to
only four substances (formaldehyde, methylene chloride, p-dichloro-
benzene, and asbestos) of the 10,000 chemicals estimated to be
present in consumer products (many of which are also present in
indoor air and other exposure media) .
Neither detailed exposure data nor toxicological data were
provided to support the assessment.
5.2.11 Radiation Other Than Indoor Radon
A "medium" cancer risk ranking was assigned in "Unfinished
Business" to this problem area, based on the estimate that 360
cancers each yea.r in the U.S. may be attributed to ionizing
irradiation from occupational exposures, consumer products (chiefly
building materials) , and industrial emissions. The exposure data
on which the estimate was based are extremely limited, although
somewhat better than the data for most environmental chemical
toxicants. Similarly, the relevant dose-incidence relationship for
radiation carcinogenesis is uncertain. The estimate was based on
the National Academy's recommended risk models, that have been
derived from analysis of cancer rates in irradiated human popula-
tions (e.g., NAS/BEIR, 1980) and have since been updated (NAS/BEIR,
1990) . Depending on the assumptions employed, the new models would
yield risk estimates that are higher by a factor of 2-4. In spite
of these limitations, the Subcommittee considered the ranking to be
reasonable.
sSee Section 5 references (3)-(6), (10), (14), (17) and (23)
55
-------
5.2.12 Depletion of Stratospheric Ozone
A relatively high cancer risk ranking (seventh) was assigned
in "Unfinished Business" to this problem, based on the estimate
that an additional 10,000 deaths from skin cancer would result
annually in the U.S. by the year 2100 if the levels of stratospher-
ic ozone continue to be depleted at present rates during the
interim. Although the estimate is based on uncertain projections,
the Subcommittee considered the ranking to be appropriate at this
time. Continued surveillance of the situation is called for,
however, since a higher ranking would be warranted if the projec-
tions were to be supported by future trends in ozone depletion and
skin cancer rates.
5.2.13 Hazardous Waste Sites
A moderately high relative risk rating was assigned to this
problem area in "Unfinished Business, on the basis of the estimate,
extrapolated from 35 of the 25,000 sites nationwide, that 1,000
cancers per year in the U.S. population are attributable to only
six (trichloroethylene, vinyl chloride, arsenic, tetrachloroethyl-
ene, benzene, and 1,2-dichloroethane) of the many potentially
carcinogenetic substances known to be present at hazardous chemical
waste sites. The ranking was also based on the assessment that the
risks to some individuals can be high.
In the absence of data on the extent of human exposure to the
chemicals in question, which were not provided in the UB report and
which remain fragmentary, numerical assessment of any associated
risks to human health is fraught with great uncertainty. The
uncertainty is compounded by the fact that an increase in the
incidence of human cancer attributable to residence in the vicinity
of a waste site is yet to be demonstrated conclusively (Buffler, et
al., Upton et al., 1989).
5.3 Rankings for Risks of Adverse Effects other Than Cancer
The estimated relative risks for causing adverse effects other
than cancer, although even more uncertain than the estimated risk
for causing cancer, were ranked into three categories, 1) high, 2)
medium, and 3) low, as shown in Table 5.3, below. Salient comments
on the "high" rankings are summarized in the following.
56
-------
5.3.1 Criteria Air Pollutants
This problem area was ranked comparatively "high" for non-
cancer risk in the "Unfinished Business" report, and the Subcommit-
tee agreed with this ranking. It merits a "high" classification
even though levels of some of these pollutants have declined with
the implementation of National Primary Ambient Air Quality
Standards. While acute episodes are infrequent for SO2/ TSP and 03
short-term concentraions can be high, and chronic effects of the
criteria pollutants are still a concern. There was no discussion
of the neurotox-
ic/behav ioral
effects of lead
in the UB report,
although the lat-
ter are of far-
reaching public
health signifi-
cance in view of
evidence that the
development of
the human brain
may be impaired
by lead at levels
resulting from
concentrations
widely prevalent
in ambient air.
On this basis
alone, a "high"
ranking
risk
would
amply
have been
justified
for criteria air
pollutants. By
and large, the
human health ef-
fects of the oth-
er pollutants in
this problem area
are well known in
terms of the
Problem
Environmental Problai _
High Risk
Criteria air pollutants
Hazardous air pollutants
Other indoor air pollutants
Drinking water
Accidental releases of toxics
Pesticide residues in foods
Application of pesticides
Consumer product exposure
Worker exposure to chemicals
Hedim Risk
Indoor radon
O tt^ i a*- T ru-k fnrtn rnrlrm\
Radiation vnon-raoonj
Ozone depletion by UV radiation
Indirect discharges (POTVs)
Non-point discharges
Discharge to estuaries
Municipal waste sites
Industrial waste sites
Other pesticide risks
Low Risk
Direct discharges (industrial)
Contaminated sludge
Discharge to wetlands
Hazardous waste sites (active)
Hazardous waste sites (inactive)
Mining wastes
Releases from storage tanks
Unranked
Other air pollutants
New toxic chemicals
Biotechnology
Greenhouse gases
"o-
1
2
5
15
21
25
26
30
31
4
7
10
11
13
18
19
27
9
12
14
16
17
20
23
3
24
29
8
Proportion
of Problei
Covered (»)
30-1001
<3X
30-100X
30-1 OCX
30-100X
<3X
3-10X
3-10X
<3X
30-1001
M41UW
-10UX
30-1001
3-10X
7
30-1001
10-30X
30-100*
10-30X
3-10X
30-100X
7
10-30!
10-30%
30-1001
7
?
7
7
7
Level of
Confidence
in Ranking
High
Medium
Medium
High
High
Medium
High
Medium
High
Low
M^uJ* • ^
Medim
Low
Medim
7
Mediua
Mediui
Low
Mediui
Nedius
Mediui
7
Mediua
NediuM
Low
?
7
7
7
7
Table 5.3 "Unfinished Business" Report non-
cancer risk rankings of the various environmen-
"acute" effects tal problems
57
-------
that they may produce during episodes of heavy pollution, and
warrant a ranking of "high." Their chronic health effects are less
well characterized, but are potentially of major health conse-
quence. For example, inhibitory effects on pulmonary clearance
mechanisms also have been documented experimentally (Driscoll et
al., 1986; Schlesinger and Gearhart, 1987), and other experimental
data suggest that interactions between such air pollutants,
particularly acidic aerosols and oxidants such as ozone or N02, may
potentiate fibrogenesis and other long-term effects (Warren and
Last, 1987; also see the Ozone Case Study, section 8.1.1).
Adverse effects of the pollutants on the "quality of life",
that may result through the production of disagreeable odors, smog,
haze, or irritation, were apparently not considered in the
"Unfinished Business" report. Nevertheless, these effects,
although difficult to evaluate quantitatively, can be stressful and
can disturb mood and behavior.
5.3.2 Hazardous Air Pollutants
Although, in general, their relevant health effects were not
expected to be severe, this class of pollutants was ranked "high"
in relative risk, in view of the large population that may be
exposed to them and the projected non-cancer health impacts that
were judged to be attributable to only six substances (estimated to
be only 3 per cent) of the many potentially hazardous pollutants in
question. This ranking was not explained in detail.
5.3.3 Indoor Radon
A "medium" non-cancer risk ranking was assigned to indoor
radon in the "Unfinished Business" report, based on the estimate
that it may cause "200 cases per year of serious mutagenic and
teratogenic effects;" however, the estimated radiation doses on
which the assessment was based were not specified. The Subcommit-
tee seriously questioned whether the relevant doses to the gonads
and to the embryo are large enough to cause risks of the magnitudes
projected (see the Case Study on Radon, section 8.1.2).
5.3.4 indoor Air Pollution Other Than Radon
The problem was ranked "high" for non-cancer risk in the
"Unfinished Business" report on the basis of the large extent of
58
-------
population exposure and the moderate-to-severe health effects that
may be attributable to the types of agents in question.
However, with regard to risk assessment, the issue is still
problematic. For most indoor air pollutants, the needed data on
exposure are not available, the health effects are diverse, and the
exposure-response relationships are not well characterized. It is
noteworthy, however, that a possible exception to this generaliza-
tion is environmental tobacco smoke, for which epidemiological
investigations have described exposure-response relationships
linking illnesses of the lower respiratory system and effects on
lung development during infancy with maternal smoking.
5.3.5 Drinking water
A "high" non-cancer risk ranking was assigned to this problem
area in "Unfinished Business", on the basis of the serious health
effects that may be associated with ingestion of water pollutants
such as lead, microbial pathogens, nitrates, and chlorine disinfec-
tant by-products. Again, this ranking is based on limited exposure
data. Lead used in plumbing may contaminate drinking water at high
levels, and concern is increasing as more is learned about the
toxicity of lead, especially at lower exposure levels. Also, as
other sources of exposure to lead are eliminated, this source may
be of greater importance even though water contamination is usually
intermittent. Pathogens also continue to be a source of morbidity,
especially in smaller systems that do not chlorinate or adequately
filter surface water. The Subcommittee recommends that procedures
be put into place to enable a better assessment of illness from
this source.
5.3.6 Pesticide Residues on Foods
A "high" non-cancer risk ranking was assigned to this problem
area in "Unfinished Business", on the basis of assessments of the
potential health effects attributable to only three (aldicarb,
diazinon, and EPN) of the hundreds of pesticides to which large
segments of the population are potentially exposed. The exposure
and toxicological data necessary to support this ranking were not
provided. In future ranking attempts, it is important to consider
risks to children. As EPA recognizes, children are subject to
higher exposures and constitute a more vulnerable population than
do adults. A broad spectrum of effects should be considered,
59
-------
including neurotoxicity, fetotoxicity, immunotoxicity, and enzyme
alterations.
5.3.7 Application of Pesticides
The problem area was ranked "high" for non-cancer risk in the
"Unfinished Business" report, owing to the relatively large numbers
of persons exposed (estimated at 10,000-250,000), the numbers of
acute poisonings each year attributed to pesticides among pesticide
applicators (e.g., 350 poisonings from ethylparathion and 100 from
paraquat), and the risks of other severe toxic effects (e.g.,
fetotoxicity, teratogenicity) that may conceivably occur. Although
the estimates cannot be evaluated critically in the absence of more
detailed exposure data for the populations at risk, the Subcommit-
tee considered the ranking to be reasonable.
5.3.8 Worker Exposure to Chemicals
The non-cancer risk ranking assigned in "Unfinished Business"
to this problem area was "high", based on the large population (at
least 300,000 workers) estimated to be exposed to each of the four
substances considered (2-ethoxyethanol, methylene chloride,
perchlorethylene, and formaldehyde) , and the high concentrations of
the substances that may be encountered in the workplace; however,
detailed data on the relevant exposure patterns and associated
health consequences were not provided. On the basis of other
assessments of the incidence of occupational disease—approximately
190,000 cases were reported in 1987 by the Bureau of Labor
Statistics (Yancey, 1988; also see Levy and Wegman, 1988), and the
Occupational Safety and Health Administration expects its new
standards to reduce by 500,000 the number of workdays lost each
year as a result of exposure to hazardous and toxic substances
(King, 1989)—the Subcommittee considered the "high" risk ranking
to be reasonable.
Although the Subcommittee concurred in the above risk rankings
for occupational exposure, the uncertainties in its evaluation
pointed to the need for several lines of effort to improve such
assessments.
60
-------
5.3.9
A "high" non-cancer risk ranking was also assigned, based on
consideration of three such substances (2-ethoxyethanol, methylene
chloride,and formaldehyde), the large populations exposed, and the
relatively high concentrations that may be encountered under
certain circumstances. Neither detailed exposure data nor
toxicological data were provided to support the assessment.
5.3.10 Radiation other Than Indoor Radiation
A "medium" non-cancer risk ranking was assigned in "Unfinished
Business", based on the estimate that 160-220 of the serious
mutagenic and teratogenic effects occurring annually in the U.S.
could be attributed to ionizing irradiation from consumer products
and occupational sources; however, the radiation dose estimates and
risk models on which the assessment was based were not presented.
Without further documentation, the ranking cannot be evaluated
critically. There are large uncertainties involved in assessing
the genetic (heritable) and mutagenic risks attributable to low-
dose irradiation (NAS/BEIR, 1990).
Excluded from consideration in "Unfinished Business" were the
potential risks attributable to low-frequency electromagnetic
radiation. These risks, although as yet equivocal (OTA, 1987),
merit consideration in future assessments of the health hazards of
environmental radiation, in view of the large populations that are
exposed.
Noise was, similarly, excluded from consideration in "U-
nfinished Business." This form of energy, akin to non-ionizing
radiation may also deserve inclusion in future assessments of
environmental health effects insofar as it may, under appropriate
conditions, cause hearing loss, stress, and impairment in the
"quality of life," with consequent impacts on mood, behavior, and
productivity.
5.3.11 Depletion of Stratospheric ozone
A "medium" non-cancer risk ranking was assigned in "Unfinished
Business", based on the estimate that the projected depletion of
stratospheric ozone could eventually increase the incidence of
senile cataracts in the U.S. by 10,000-30,000 cases per year, and
61
-------
that it might also cause other adverse health effects, including
disturbances of immunity.
5.4 Merging of cancer and Non-cancer Risk Rankings
Any attempt to combine into a single aggregate rank order the
risk rankings for cancer (Table 5.2) and the risk rankings for
health effects other than cancer (Table 5.3) would require
appropriate weighting of the different risks for incidence and
severity, as discussed below (Section 6.3). Because of the
complexity of such a task, as well as the lack of the requisite
data, the development of an aggregate ranking was not attempted by
the Subcommittee. In Section 6.3 the Subcommittee proposes two
possible methods for producing such merged rankings.
It is noteworthy, however, that if the 31 problems were to
have been arranged merely on the basis of whether they represented
either sources of en-vironmental pollutants or environmental
situations (or agents) involving direct human exposure,, they would
have appeared in categories such as those shown in Table 5.4. The
order in which the rankings appear in Table 5.4. is not entirely
unexpected since the public health impact of any toxicant depends
not only on its toxicity but also on the relevant human exposure.
Thus those problems representative of proximal exposure situations
(Nos. 2, 4, 5, 15, 25, 26, 30, and 31) would logically be expected
to receive relatively high risk rankings for cancer and/or other
adverse health effects. It should be noted, however, that among
such problems risk rankings for some (Nos. 2, 4, 5, 7, 15, 26, and
31) were supported more firmly by the available data than were the
rankings for others.
Since the rankings shown in Table 5.4. are based on highly
uncertain risk assessments, as noted above, the Subcommittee viewed
the rank order with reservations. Sufficient time was not
available, however, for in-depth reassessment of the rankings, that
would probably have been of limited value in any case without more
adequate information about the relevant levels of exposure and
toxicity. For optimal refinement of the risk assessments, further
effort must be directed toward developing the necessary databases,
scientific understanding, and methodology, as recommended elsewhere
in this report. Other comments that should be kept in mind in
interpreting the table are as follows:
62
-------
1) Risk estimates
for different
exposure and source
categories or
"problem areas" were
not directly
comparable because
they were often based
on different models
and assumptions made
by the various
program offices
involved.
2) In many cases,
estimates of risks
for a problem area
were incomplete,
covering only a few
of the agents or
exposures comprising
the exposure or
source category.
3) The assumption
underlying the UB
ranking was that
existing programs
would continue.
Therefore, under that
assumption, some
problems appeared to
pose relatively low
risks precisely
because of the high
Situations and Agents Involving Problem
the Potential for Direct ExDosure n<
Ambient air pollutants
criteria air pollutants
hazardous air pollutants
other air pollutants-
Indoor air
radon
other indoor air
Drinking Meter
Pesticide residues in food
Occupational exposures
application of pesticides
worker exposures
Consuaer products
External radiation
radiation other than radon
Sources of Environmental Pollution
Atmospheric
substances depleting Strat. Oj
greenhouse gases, COj, etc.
Surface water discharges
direct point source discharges
indirect point source discharges
non-point source discharges
discharges to estuaries
discharges to wetlands
Multimedia discharges
contaminated sludge
hazardous waste sites (active)
hazardous waste sites (inactive)
nonhazardous waste sites (Municipal)
nonhazardous waste sites (industrial)
•ining wastes
accidental releases of toxics
accidental releases (oil spills)
releases from storage tanks
Miscellaneous
Other ground water contamination
Other pesticide risks
New toxic chemicals
Biotechnology
* Risk rankings assigned in UB report;
Assigned Bisks1
)T (Cancer) (Non-c
1
2
3
*
5
15
25
26
31
30
6
7
8
9
10
11
13
14
12
16
17
18
19
20
21
22
23
24
27
28
29
H * high.
L
H
-
H
M
N
H
N
H
H
N
H
-
L
L
L
-
-
N
N
H
N
N
L
I
L
L
L
N
N
-
N » mediu«;
ancer)
H
H
-
M
H
tt
H
M
H
H
N
N
-
L
N
N
M
L
L
L
L
N
M
L
H
-
L
.
N
-
-
L * low (see Tables 5.2, 5.3, and Section 8.2)
Table 5.4 Environmental Problems grouped by
exposure and source categories with the risk
rankings assigned in the UB Report
levels of effort that had been devoted to controlling them (UB, p.
96) . It is therefore important that future analyses state the
scope of the problem without the control assumption.
In addition to these caveats it should be noted that the UB
ranking system did not adequately address the goal of prevention of
risk. This being the goal of EPA, future analyses should include
assessment of subclinical or preclinical effects of environmental
63
-------
agents and should give weight to effects that would affect future
generations. In this regard, it is also important that risks be
estimated both for the general population and for the most exposed
or most sensitive sub-populations (e.g., children, those with
preexisting disease, etc.) Also, certain factors were excluded
from the UB analysis, including economic or technical control-
lability of the risks, the degree to which risks are voluntary or
equitable, EPA's statutory or public mandate to deal with risks,
etc. Translation of risk rankings into public policy should
explicitly incorporate these factors in the future.
64
-------
6.0 Approaches for The Long-term Development of Improved Risk
Assessment Strategy
It has long been known that health risks can be associated
with exposures to specific agents and combinations of agents, and
that such risks can be lessened by reducing the exposures. It
follows that the extent of the risks, and the benefits derived from
risk reduction, can be determined from risk assessments based on
reliable information about the distributions of exposures among
population groups of interest and the exposure-response
relationships for such groups. Furthermore, when such information
is available for a variety of specific agents and mixtures, and for
the severity of the various responses they produce, then the variou
s risks can be ranked. The rankings can then be used in the
development of overall risk reduction strategies.
Unfortunately, the straightforward logic outlined above
requires more information than has previously been available.
Section 6.1 outlines the problems with the UB framework and
illustrates a conceptual plan to deconstruct the UB's 31 risk
categories through a source—exposure—agent—effect matrix so that
the information required for a more logical ranking scheme can be
related to the information needs of Agency programs. It is
followed by Section 6.2, which outlines the Subcommittee's
recommended approach for developing risk assessments for specific
toxicants. Using this approach, for example, the limited number of
specific toxicants having credible risks could be ranked. Such
rankings could then be used in developing optimal risk reduction
strategies.
6.1 Alternative Models for Risk Reduction Targets
The problem areas defined in the Unfinished Business report
are a mix of three very different types. The first typically
represents agents that constitute direct sources of human
exposure. These include indoor and outdoor air pollutants,
radiation, pesticides, and consumer products. The second
represents sources of emissions which in most cases must be
transported to an exposure situation. The third type of problem
area represents agents which must make contact with the human
receptor before a toxic exposure can occur. Typical problem areas
of this kind are worker exposure and drinking water. Also, in
every source area, its impact is felt via agents in the first
65
-------
EXPOSURES
-1- -r- -3- -«- -5- -6- -7-
PRtOH AMBIENT MICRO
SOURCES CATEGORIES SOIL FOOD WATER AIR ENVIRON INDIRECT OCCUPATIONAL
-A-
lATmii 900ICI3 t ' ». "-
ntoczascs n t*
-B-
UJB I HTI1 H« 11. 13 n
AUK UJiOCHHT
-c-
«c:icamu T. e. it.
13. 1«. 21
-0-
.iimic i 9 u. i».
KTMCTIOI 30
- -
!»C»CT 9 9, U.
11. :a
TUIIKmnof 7, I. 9
-G-
n*jnir»cml«o 7. ». 9 13.
1«. 23. II
- — T,_ », 9, 12.
run CTOUOI. 13', i«. 16.
»IS»3»L t TnkTBOT I? It 19
ACCIDCRTlkL 9. 13. 14
ULIUES 21. '22. 2>
-J-
t
Table 6.1.1 Source and exposure matrix
category. Therefore, in conceptualizing the risks associated with
the different problem areas it is important to understand their
interactions so that the priorities of their relative impacts do
not become confused or double-counted.
It is understood that the basis for the 31 risk categories in
the UB report is the regulatory mandates and the administrative
structure of the EPA. Nevertheless, in order to conceptualize the
risks better and to understand the sources, and exposures that
contribute to the risks, we find that the development of a matrix
approach may be useful. An example of such a two-dimensional
matrix is shown in Table 6.1.1. The vertical columns consist of
direct exposure terms, that are the closest connections to the
human exposures. These seven exposure terms or secondary vectors
for the most part represent the routes via which humans are
exposed, with the exception of Category 7, Occupational Exposure.
66
-------
A> •aturai Sources and Processes Includes constituents released naturally into the environment,
even though their rates of release may be influenced anthropogenically. Radon seepage into homes
or present in groundwater is an example; also, arsenic in groundwater from dissolution of bedrock.
Included are natural processes that modify the chemical nature (and toxicity) of materials, such
as methylation of mercury.
B) Land and Water Use and Management Examples are urban land use affecting constituents in water
run-off such as road salt; emissions from the application of herbicides to control growth along
highways; also, uncollected and untreated wastes from all non-commercial sources: e.g., homes,
military installations, schools and universities, etc.
C) Agriculture Examples includes point and non-point emissions from fertilizer and pesticides.
0) Hinina and Extraction Includes air and water emissions from mine tailings and on-site processing
of minerals.
E> Energy Emissions of wastes from processing and production of coal, oil and nuclear energy.
Includes petroleum refining, coal desulfurization, and stack emissions.
F) Transportation All waste emissions from transportation: includes air emissions from mobile
sources (cars, trucks, airplanes); releases to water from ships.
G) Manufacturing Wastewater and air effluents, treated or untreated; fugitive emissions; deposition
on land; injections to groundwater.
H) Waste Storage. Disposal, and Treatment Includes community, industrial, and individual owned
wastewater treatment systems; landfills for hazardous and non-hazardous wastes; waste incinerators;
ocean disposal; and industrial wastewater lagoons.
I) Accidental Releases All accidental releases, whether sudden or continuing: above or below ground
storage tank ruptures or leaks; releases from train derailments and collisions of tank trucks;
releases from explosions of chemical or power plants.
j) Consumer and Commercial Products Emissions from or contact with materials and products, other
than the human exposures for which the product was intended. Examples are inhalation of emissions
from products used in offices and homes, such as paint solvents and pesticides.
Table 6.1.2 Source terms or primary vectors (A through J) of
Table 6.1.1, representing the various activities, materials,
or processes that constitute the recognizable sources of
chemical and other emissions
This can constitute a variety of possible routes of exposure,
including dermal, inhalation, and oral.
A complete description of each of these categories, as well as
the source terms represented by the horizontal rows is presented in
Tables 6.1.2 and 6.1.3. These source categories or primary vectors
represent the various activities, materials, or processes that can
constitute recognizable sources of chemical and other emissions
that are transported by various processes to human receptors.
There are overlaps among these source vectors. For example,
accidental releases can arise from transportation, waste storage,
or manufacturing processes. However, because of their sporadic
67
-------
1. Soft Direct hunan contact with contaminated soil, such as children playing in such soil.
2. Food Unintended contamination of food by anthropogenic chemicals.
3. Uater Ground and surface water, potable or otherwise, contaminated with chemicals,
radionuclides. and microorganisms.
4. Ambient Kir All ambient
-------
AGENTS
EA
SA .-
SITUATIONS
SS
souacEs
The next step would be
to consider each of the
exposure constituents in
an element and assess
the risk to the total
U.S. population of the
releevant endpoints.
Only then could the
final judgment of the
health impact of each
element in the matrix be
addressed. The ultimate
purpose would be to use
this information to
judge either the impact
of a given source term
by moving horizontally
across a given row; or pigure €titi Expansion of the two-dimen-
alternatively, to judge 3iOnal matrix to include a third dimen-
the health impact for a sion—Agents
given exposure term by moving vertically through the matrix table
for that exposure.
Expanding the two dimensional source-exposure matrix to in-
clude a third factor, agents, is illustrated in Figure 6.1.1.
Here, source #10 contributes to exposure situation #7; the
intersection, corresponding to that of Table 6.1.1, is labeled ES.
The figure also shows that source #10 contains agent #8, and thus
source #10 contributes agent #8 to exposure situation #7 (interse-
ction EA in the exposure situation/agent plane. The three
dimensional intersection ESA brings all this information together.
Figure 6.1.2 expands further on the three dimensional concept,
showing the interactions of a number of sources, exposure situa-
tions, and agents. Thus, source #5, containing agents #s 2, 4, and
7, contributes to exposure situations #s 3, 5, 6, and 7. Exposure
situation #3 also receives agents #s 3, 5, 6, and 9 from other
sources. This three dimensional matrix quickly discloses interac-
tions and multiple contributions, and entering the matrix at any
element of a dimension (for example, at agent #6) allows one to
determine which other exposure situations are affected by agent #6
from which sources, etc... The three dimensional intersections,
such as intersection ESA in Figure 6.1.2, are not shown, for
69
-------
,
AGENTS
H
i
a
-7
e
.5
4
.3
.2
1
L
r
L
F^
1
*
P
* ]
V -
1
1
1
1
!
1
!
fc A. i
£1.2 Expansion of 3-dimensional
exposure situations as- matrix, showing interactions of sources,
sociated with each exposures/ and agents
agent. Depicting this lacking information is not simply and
directly possible on a two dimensional plane; three planes in
addition to those shown in Figure 6.1.2, would be needed: Agent-
endpoint, source/endpoint, and exposure situation/endpoint.
Given the four dimensional organization of information on
sources, agents, exposure situations and endpoints, the work of
rank-ordering different elements within each of these four
dimensions, separately, for risk would be simplified and given
consistency. It would also be possible to identify the three
dimensional (ESA-type) intersections involved with other defined
environmental problems, such as those in the UB report, and so to
assist in rank ordering them.
In considering, then, how risk areas might be better defined
and relevant information organized for ranking/assessment purposes,
the Subcommittee proposes as a possible approach the development of
a matrix the principal dimensions of which include sources.
exposure situations, agents. and endpoints. Such a matrix can be
entered via an element of any of the dimensions (for example, an
agent or a source) and, via the intersections of that dimension
with others, the appropriate relationship to the others can be
70
-------
determined. Given such a matrix in computerized form exposure
situations, sources or agents can each be ranked according to risk,
bringing order to the problem of determining the most important
steps to take to reduce health risks. Further, identifying the
intersections relating to a risk area of interest to the agency
would consistently identify the elements of each dimension relating
to the risk area. Developing the matrix in usable form and
entering information into it would be no small task; once even
partially available, however, its utility would be great.
Developing and putting into practice the full information
system required for insertion of relevant information now avail-
able, and new information as it is acquired, including the
capability of tying into other existing information systems which
already contain toxicity, physical and chemical property, exposure,
dose response, or other information on agents, endpoints, exposure
situations, and/or sources, is a very large task that would involve
information system design specialists working closely with
scientists, technologists and risk managers in the Agency, and
possibly, outside of it. In essence, however, the information
required can be structured as a relational data base design for
which many commercial software products are available. The
Subcommittee does not have an estimate of the numbers of workyears
involved other than to say that it is expected to be large. In the
final analysis, and, in a very real sense, the task will never be
quite complete: whatever initial system is designed and put in
place will undergo continual change, expansion, and development (as
distinct from maintenance) as it is used and as experience is
gained from cataloguing new information in it.
The Subcommittee recommends that that this strategy be
implemented in small increments. At this early, conceptual stage
the complexities and practical difficulties cannot be projected,
but they will surely be there. Rather than address the design and
implementation of the whole, ultimate system at the start, the
Subcommittee recommends that a specific four dimensional system be
developed and filled with information for a small number (three or
four) of different but relatively widespread agents (tying in to
pre-existing data bases (such as IBIS)), and used as a test case;
this effort would take the form of a limited pilot project, the
product of which would find immediate, practical utility, and it
would serve to give practical guidance to the design of a more
advanced version of the system suitable for the insertion of data
71
-------
on many agents, endpoints, sources and exposures. Typical agents
to be used in the pilot project would be, as examples from which to
pick the small number to be used: Benzene, TCE, Lead, Ionizing
Radiation, Arsenic, Chloroform, Dioxins, PCBs, Carbon Monoxide or
Ozone. A step-wise, pilot-project-guided approach, the Subcommit-
tee thinks, would produce a usable product even at the pilot stage,
uncover what is needed to progress to a further stage of design and
development, and increase the ultimate chances of achieving a
successful four dimensional information organizing system or
matrix. Progression to a further stage of development should also
be restricted to a manageable project. The next stage is visual-
ized as beginning by discovering what is encountered in the way of
new problems, and how these are to be solved, by adding a larger
but still limited number of agents, selected for ubiquity and
potency, to the pilot project matrix. Such agents might be
selected from already existing lists such as the list of agents for
Community-Right-to-Know reporting under SARA, as well as new
substances identified in the application of TSCA. Section 5.
The Subcommittee does not recommend a particular organiza-
tional approach to managing this project, either in the short or
long term; it points out, however, that both short term and long
term aspects should be considered in organizing for this under-
taking. The Subcommittee recommends that a single, clear focus of
responsibility be assigned at the start to provide planning
(including budgetary planning), continuity, coordination, progress
reporting and accountability for the project.
The four dimensions are not new to the risk assessor. The
Subcommittee believes that conceiving of them formally as an
interconnected system, as human health risk questions are ad-
dressed, will improve the risk assessment process, helping the risk
assessors to think broadly and holistically when considering any
particular problem. The Subcommittee believes that risk assessors
within the Agency should be encouraged to consider the four
dimensions as they pursue their work and to document, wherever
possible, the four dimensions, their relationships, and the
relevant risk information, as they do their work. Such documen-
tation can become a source of information for insertion into the
information system itself.
There is thus no need to wait to receive some benefit from
this concept until the ultimate, or even the pilot, system is
72
-------
established and implemented. As an aid to the thought process it
can be useful.
We believe that this approach, although difficult to execute,
would provide a perspective that could assist in prioritizing the
efforts of the EPA in reducing the risks to the U.S. population.
It could help identify the agents and activities that contribute to
the greatest risks, as well as the exposure media of greatest
concern.
6.2 Identification and Assessment of Specific Toxicants
From the foregoing it may be concluded that the initial step
to be taken in environmental health risk assessment is the
identification and tabulation of health effects of non-trivial
concerns that are associated with those particular environmental
pollutants which demonstrate both evidence of 1) toxicity following
exposures of environmental relevance and 2) evidence of widespread
or intense exposure to populations or to individuals. Most
pollutants that meet these criteria will be specific agents, such
as 03, chloroform, benzene etc, or mixtures containing a common
active agent or functional group, such as Pb and its salts or
nitrosamines. More complex mixtures, especially those that vary
considerably in composition from place-to-place and/or from time-
to-time are harder to rank. The possibility of synergism in such
mixtures should be considered. An NAS-NRC report (1988) concluded
that synergism is a relatively rare occurrence. However, there is
a paucity of experimental and epidemiological data bearing on this
question (Waters et al., in press; Vainio et al., in press). Based
upon both experience and theoretical modelling it found that
additivity of effect is the common rule, and that synergism
generally occurs only when one component of the stressor or a co-
stressor, is a potent toxicant present at a sub-threshold level or
level that produces only a small yield of responses on its own. In
some practical cases, the toxicity of a complex mixture can be
characterized by the toxicity of its most active component that.
In theory, this approach can lead to lists containing
hundreds, or possibly even thousands of agents. In practice, it is
unlikely that as many as one hundred pollutants can meet the two
entry criteria posed earlier. The pollutants that do cross the
threshold can then be evaluated for risk levels according to the
processes we have adopted for the following case studies. In this
73
-------
manner the risks, and their uncertainties can be quantified and, if
desired, ranked in order or by groups.
6.2.1 Selection of Specific Pollutants
The health impacts associated with environmental exposures can
usually be attributed largely to individual chemical agents, such
as ozone, most of which have been relatively well studied. Still,
as the following case study on 03 demonstrates4, there remain
critical unknowns about exposure and exposure-response
relationships which limit our ability to perform essential
quantitative risk assessments.
Other health impacts are associated with classes of agents
such as radon and its decay products, lead and its salts, the
various nitrosamines, PCBs, dioxins, trihalomethanes, etc. In
these groupings, there are variations in bioavailability and
metabolism that result in widely varying toxicity according to the
composition of the mixture and the influence of other materials in
the exposure environment. The influence of such factors is
demonstrated in the following case study on radon decay products5.
Even more difficult to evaluate are groupings such the
products of incomplete combustion (PIC), municipal waste-treatment
sludges, etc., where the materials are so diverse that they can
range from highly toxic to essentially innocuous.
6.2.2 Addressing Exposure Parameters
The process of risk assessment depends on both the toxicity
assessment and the exposure assessment. In most cases, the
exposure assessment will be the limiting factor in the overall
process. The increasing recognition of this limitation by EPA
needs to be followed by positive action to address it. We strongly
recommend that EPA continue to expand its capabilities for
quantitative exposure assessment so that it can effectively utilize
the growing data base of toxicity information.
4See Appendix, Section 8.1
5See Appendix, Section 8.1.2
74
-------
6.2.3 jifpmrt«« «*<* Lasaons Learned from the Case Studies
The two case studies presented in the Appendix (Sections 8.1.1
and 8.1.2 provide lessons that may help us deal with risk
assessment problems in the future. Some of the lessons are:
6.2.3.1 Ozone
1. Ozone in ambient air was not initially established as a
human health hazard. Rather, it was considered primarily as a
nuisance, as well as a plant pathogen. Ozone was grouped under the
category "community or criteria air pollutant," which included
demonstrated health hazards (e.g. the acidic pollutant mixtures
that produced excess mortality and morbidity in Donora and London).
Initially, concern with 03 was thought to be limited to specific
geographical locations, such as Los Angeles, rather than a
widespread problem. As our concern has shifted from responses in
terms of body-counts and clinical cases of disease, to risks of
accelerated loss of lung function and/or avoidance of coughing and
chest pain during outdoor exposure, 03 came to be regarded as an
important health hazard.
2. Early animal studies on 03 clearly showed that it was
capable of producing massive lung damage. However, it did so only
at rather high doses. When experiments were conducted at levels of
03 approaching ambient levels, animals no longer suffered detec-
table effects. As our ability to detect and quantify subtle
changes in function and localized damage to airway linings
improved, we began to recognize and appreciate the importance of
gradual changes leading to disability or premature death late in
life.
3. Although some early animal experiments indicated that 03
would produce acute lung injury, little evidence for this was found
in humans. The advent of sophisticated pulmonary function
measurements eventually produced some evidence that 03 would alter
pulmonary function in humans at high ambient levels. The effects
were modest and transitory, and were reduced in magnitude following
repetitive daily exposures ("adaptation"). The laboratory studies
in humans stimulated the undertaking of larger-scale field studies
which provided further indication that a significant problem
existed. They also stimulated the development and application of
more sophisticated tests in the laboratory, such as broncho-
75
-------
constrictor challenge, and assays of lung lavage fluids for cell
number and function, and release of mediators.
4. The huge volume of health effects research on 03 has not
been adequate to define the extent of the human health risks
associated with population exposures, demonstrating the need for a
strategic research plan to address the critical knowledge gaps.
These gaps include the role of repeated exposures over a season or
a lifetime on the pathogenesis of chronic lung disease and the role
of co-pollutants and other environmental factors on both short- and
long-term responses.
5. We know a great deal about transient functional responses
to single 1- and 2-hour exposures to O3 under controlled con-
ditions, including the enhancement of response due to increased
ventilation during exercise. However, we have only recently
learned that:
a The acute response syndrome involves other transient
responses such as: 1) influx of inflammatory cells and
mediators into the lung; 2) increased airway reactivity;
3) increased airway permeability; 4) altered rates of
mucociliary particle clearance from the lung airways.
a The responses increase with duration of exposure for at
least six hours, and dissipate with a similar time
constant. This is important for people who remain
outdoors, since O3 exposures in most heavily populated
regions have a broad daily plateau lasting 6-10 hours.
Furthermore, peak O3 exposures generally occur on many
successive days during the summers, and exposures are
often as high or higher in suburban and rural areas as in
urban centers.
a Responses among children and healthy non-smoking adults
engaged in normal outdoor recreational activities are
greater than those observed in the controlled exposure
studies with 03 alone at comparable concentrations,
suggesting that other constituents potentiate the
characteristic 03 responses, and that exposure-response
relationships based on chamber studies underestimate the
health impacts on natural populations.
76
-------
• Acute responses among both laboratory and field study
populations indicate large interindividual variations in
sensitivity to 03. There is little known about the
causes and correlates for this wide range of respon-
siveness. The data suggest that large numbers of
individuals have symptomatic responses, as well as
functional deficits large enough to constitute adverse
effects, following natural exposures even on days when
the current standard is not exceeded.
We clearly need to identify the constitutional factors that
account for large variations in response among the population, so
that the more susceptible people can know when to avoid outdoor
exposures, and so prophylactic therapies can be designed to help
susceptible individuals avoid the effects of excessive exposure.
6. We know relatively little about the long-term consequences
of repetitive daily exposures of humans to O3. However, there are
serious concerns based on the results of chronic exposure studies
in laboratory animals showing that:
a Successive daily exposures of rats leads to progressive
epithelial cell damage even when respiratory function
changes are transient.
a Chronic exposure studies in rats and monkeys at high
ambient 03 concentrations produce functional and struc-
tural changes in the lung consistent with stiffening
and/or premature aging of the lung.
a Rats are less sensitive than humans to 03 in terms of
acute functional response, and comparable to humans in
their functional adaptation to multi-day exposures. The
lesser functional responses are consistent with the
dosimetry models for O3 uptake along the airways of
humans and rodents.
With so many people chronically exposed to 03, it is important
to determine whether premature aging of the lungs is occurring, and
if so, how the effects can be ameliorated.
77
-------
6.2.3.2 Radon
A variety of important lessons emerge from an analysis of the
findings in the radon case study.
1. The relatively large health risk of 5,000 to 20,000 lung
cancer deaths per year from exposures to indoor radon was not well
defined until NCRP (1984), EPA (1986), ICRP (1987) and NAS (1988)
gave serious attention to general population risks as well as
occupational exposure risks. Risks of this magnitude, which are
larger than most regulated cancer risks, could have been predicted
much earlier if any responsible authority has used available data
from the uranium miner experience and the available evidence that
a linear, non-threshold exposure-response model was appropriate.
2. EPA's advisory to the public on the risks from residential
radon included advice on obtaining measurement kits and remediation
services, providing effective guidance for individual home owner
actions. This was made possible by EPA's prior research and
development, efforts in these areas.
3. Multiple sources of indoor radon may be important to
residential exposure. While permeation of radon from subsurface
soil is usually the dominant source, radon dissolved in potable
water from wells can also be a significant source.
4. The risks to smokers are 6-10 times greater than for
nonsmokers exposed to a given level of radon, a conclusion not
generally communicated to the general public to help individual
citizens decide about remediation.
5. The residual uncertainties about the risk of lung cancer
from exposure to radon and daughters are quite small (30-50%).
However, one major uncertainty is the contribution of exposure
during childhood to the subsequent risk of disease.
6.2.3.3 Overall Lessons
In order to use quantitative risk assessment approaches for
relative risk ranking, we will need to define the risks of concern
and their overall impact on public health. A rich data base does
not necessarily ensure that adequate risk assessments can be
performed. The case study on ozone demonstrates that our knowledge
78
-------
of the chronic health impacts of 03 is extremely limited. We know
virtually all we need to know about the acute functional responses
in laboratory settings to 03. How- ever, we also know that
exposures in natural settings often produce much greater responses,
limiting the applicability of the laboratory data for predicting
population responses in natural settings. One lesson is that
further research based on the use of conventional tests and assays
and convenient durations of exposure should have lesser priority,
while research focussed on the critical knowledge gaps should
receive greater priority.
The radon case study demonstrates the importance of exposure
assessment in complementing the well developed exposure-response
relationships in the overall risk assessment. It also illustrates
the importance of considering multiple sources, in this case the
soil gas and radon dissolved in the potable water supply, as well
as the strong role of cigarette smoking as a modifier of radon-
induced cancer risk. Finally, it demonstrates how EPA can play an
important and productive role in public health protection concer-
ning an agent for which it has no direct regulatory authority.
6.3 Ranking Schemes
In the EPA's "Unfinished Business" Report (UB) thirty-one
problem areas (Problems) were identified and ranked, separately,
according to the cancer and non-cancer population risks believed,
as a result of analysis and consensus, to be associated with each.
The two rankings were not combined into a single population health
risk ranking but were reported separately in the UB.
From the standpoint of providing inputs to a planning,
budgetary, or resource allocation process, producing a combined
health risk ranking to include cancer and non-cancer health effects
in a single ranking would be useful. How to produce such a single
ranking of Problems, of either the UB report's original thirty-one
or of whatever different set may result from this study or future
considerations, either (1) starting from scratch or (2) by
combining separately derived rankings by cancer and non-cancer
risks, is the question considered in this section. Both approaches
to the question are explored, and frameworks are suggested for
accomplishing the ranking task in each case. The second case is
described in some depth in Section 8.2, along with an illustrative
example of how the framework should be applied in merging the
79
-------
rankings by cancer and non-cancer risks into a single health risk
based ranking.
While two frameworks are described for accomplishing the
development of health risk rankings, in neither of the two cases is
the application of the frameworks a simple matter of applying a
formula, nor can it be. The qualitative nature of much of the
information used in ranking prevents this. At each major step
scientific judgment, and preferably a consensus of knowledgeable
scientists, is needed. The example given in Section 8.2 of merging
two separate rankings into a single health risk based ranking is
just that: an hypothetical example, an illustration of how the
framework might be applied. It is not a final result of applying
the framework in a consensus-generating fashion.
6.3.1 General Considerations on Ranking and Severity
In the original ranking of the thirty one problems for non-
cancer risks as presented in the UB report the key variables were
all considered to the extent possible: exposure, potency,
incidence as derived from these two, numbers exposed, and severity
of effect, and numerical estimates and scoring systems were
developed and used, where possible, in addition to qualitative
information and best guesses. Because quantitative information
relevant to ranking was sparse, especially in the case of non-
cancer effects, the basic factors to be considered in ranking had
often to be taken into account by reaching consensus on the weights
to be accorded to qualitative information combined with what
quantitative indicators there were. This same problem exists
today.
With the cancer and non-cancer risk rankings in the UB report
done by different consensus groups, the ways in which information
was considered, classified and used by each in arriving at their
separate rankings, based on cancer and non-cancer risks, are not
entirely consistent. More attention should be given to this factor
in undertaking any new rankings. Also, ranking as a means of
setting priorities for action is a common, well-used tool in many
fields, including health. It would be useful, therefore, in any
new undertaking, to review some of the ways in which rankings
have been done in the past to ensure that what is good or useful in
them is incorporated into the ranking method ultimately used. The
well known medical practice of triage, used ordinarily under such
80
-------
conditions of high demand for scarce resources as the battlefield
or major disasters, is a good example of ranking for the purpose of
allocating scarce resources in such a way as to save as many lives
as possible. One reference well worth reviewing in considering
possible improvements in the health risk ranking of Problems is the
1984 report of the National Research Council on strategies to
determine needs and priorities for toxicity testing (see referen-
ces) . In this volume, a number of schemes utilizing different
bases are reviewed in the course of reaching the conclusions of the
study.
In ranking for non-cancer risk, severity of effect was
considered by the participants in the UB project, and, with many
apologies and qualifications, they developed an evaluation and
scoring of the relative severities of a wide variety of non-cancer
health effects as they were defined in the UB report. The method
used was a technical one based on estimating the impacts that
different apparent diseases or endpoints would have on different
organs or systems and, in turn, the severities of those impacts on
the individuals afflicted with the endpoints in question. In
ranking the problems by cancer risk, severity of effect was not
considered in the UB report. All types of excess cancers were
considered to be of highly severe consequence to affected in-
dividuals. Whether "highly severe" meant more, or less, or of
equal severity to the most severe of the non-cancer effects rated
in the UB report is unknown; it is reasonable to assume that most
cancers would be included in the highest of the seven severity
levels (or possibly, some of them, in a new, higher level) defined
in the UB report for non-cancer endpoints along with the most
severe of the non-cancer endpoints, with only some falling into
somewhat lower brackets.
In developing a merged ranking for different health endpoint
risks, whether for a diverse set of non-cancer health effects or
for cancer and non-cancer health effects combined, some way to
consider severity is needed; otherwise, effects of low severity
will be ranked as highly as those of high severity when they occur
at the same frequency, a clearly unreasonable approach to main-
taining or improving public health. The participants in the UB
report effort recognized this and attacked the problem of severity,
fully cognizant of the difficulty of the problem in the first
place.
81
-------
There is no universally acknowledged scoring system for
severity of effect at the present time, certainly not for so broad
a spectrum of diseases as falls under the heading of "non-cancer
health effects," though the problems of establishing an index have
been addressed in various contexts such as in the development of an
Index of Harm for radiation induced effects [1,2]. There is little
question that different diseases are of different degrees of
severity of impact on the sufferers; it is only necessary to think
of one's own response if asked which of two diseases one would
prefer least to contract if that was the only choice available.
What factors to consider and how to weigh and quantify such
differences in ways satisfactory to most people presents major
problems, however.
The technical approach used in the UB report must be regarded
as a laudable effort to recognize the existence of differences in
severity, but it may not give sufficient weight, in arriving at the
severity scores, to either medical specialists, on the one hand, or
to sufferers or potential sufferers on the other, nor to the
process by. which such a table of severity indexes might best be
derived in the first place. In section 3.3.3 of this report, some
of the broad factors that need to be considered in arriving at a
characterization of severity are discussed in some depth. These
factors range from scientific/technical factors to sociological/-
psychological ones. From the viewpoint of the sufferer or the
potential sufferer, such factors as "loss of productive years of
life" may not be of compelling interest; "When might I get it?;"
"How bad is it—will I die, will I be in lifelong pain, or will I
find it to be just a kind of nuisance?;" "How will it affect my
family, my friends, my job, my finances?;" "Can it be cured or
alleviated; does it progress or is it reversible?;" and "How
distressing is the treatment?" are samplings of the kinds of
questions laymen might ask when considering the severities of
different diseases. Developing a translation of these kinds of
questions into meaningful medical and scientific terms, and vice
versa, may be a necessary first step in approaching severity from
both the medical/technical and the lay perspectives in an in-
tegrated way; one possible way to accomplish this is through the
use of lay and professional focus groups meeting separately and
then together. The process by which this is done, whatever it may
be, the way in which the views of informed potential sufferers (and
how they become informed) and of medically and technically trained
experts are brought together is critical to developing severity
82
-------
factors or indices with any validity or credibility. Moreover,
such factors must be reviewed and updated from time to time as new
knowledge becomes available or as diseases become more curable or
mitigatable.
6.3.2 Producing a Merged Health Risk Ranking; the Zero-Based
Approach
We will consider, first, the problem of ranking starting from
scratch, i.e., a zero-based approach.
An approach to the zero-based ranking for creating a single,
merged health risk ranking would be to develop severity factors
for both cancer and non-cancer effects, together, as was done in
the case of non-cancer effects, only, in the UB report, but using
groups of experts and lay persons as suggested in Section 8.2, both
to develop the best set of variables to use in this exercise and to
develop the relative severity factors. This amounts to starting
over and, given the severity scores, having one consensus group
then consider both types of effects as a single spectrum of health
effects, connected to each other by the single severity factor
table. To conduct this consensus ranking exercise most ef-
ficiently, it is suggested that expert individuals drawn partly
from the UB ranking group and partly from outside sources, would be
best suited to developing the new, merged, consensus ranking. This
would help ensure that considerations raised in the present
relative risk reduction project, new information, and new under-
standings or correlations of existing information would be fully
utilized to avoid a full, duplicative refamiliarization with the
information already utilized in the UB report.
The development of the severity table, the factors that need
to be considered in defining severity, and how to combine the
factors into severities, needs further thought and definition, as
discussed in Section 8.2. Peer review of the result would ensure,
to the maximum extent possible, its scientific quality and
credibility. Such a value-laden process should include medical
experts and ethicists, sociologists, and lay persons.
In the UB report, population risk appears to be the primary
consideration in ranking, individual risk being only briefly
mentioned. In a new zero-based aggregate health risk ranking
effort, if consideration is to be given both to population and to
83
-------
individual risk, the way these are to be weighted in reaching a
conclusion on ranking must be defined and applied consistently to
obtain a credible result; the same is true of any other particular
factors such as individual or population subset sensitivity.
Long-term advantages to expending the resources needed to
apply this full procedure is that (1) the most credible result
would be produced, (2) a framework into which new information can
be fitted to update the ranking would be brought into existence,
and (3) the ranking, kept up to date, would provide useful, ongoing
guidance for budgetary and resource allocation planning. The
difficulties involved in establishing an agreed-upon severity table
must not be underestimated; a method for merging pre-existing,
separate rankings may prove to be more practical, in the immediate
term, for producing a single aggregate health risk ranking.
6.3.3 Producing a Merged Health Risk Ranking: Merging separate
Rankings into One
An alternative approach to the complete, start-from-scratch,
zero-based approach is developed in some detail in Section 8.2. It
builds on whatever may already have been done in ranking a set of
Problems (the UB report Problems or another set of issues such as
elements of one of the four dimensions described in Section 6.1)
separately for cancer and non-cancer risk, copes with the lack of
much quantitative information of any degree of precision and,
starting from the two separate rankings, involves less total effort
than the zero-based method, producing a preliminary merged risk
ranking that may provide some assistance in considering planning
alternatives.
A brief description of the principles involved in the merging
of two qualitatively ranked separate rankings of problems or other
defined issues according to cancer risks, on the one hand, and non-
cancer risks, on the other, follows. For a fuller understanding,
the reader is referred to the more detailed development in Section
8.2.
Figure 6.3.1 Shows a hypothetical linear (or cartesian) plot
of items ranked for non-cancer risks versus the same items ranked
for cancer risks as it would appear if the quantitative weights for
each of the items, non-cancer and cancer, were known. In a real
situation, the items might simply be ranked according to relative
risk: High (H) , Medium (M) , or Low (L) ; in this situation, the
84
-------
Figure 6.3.1 Plot of hypothetical risk rankings—non-cancer
vs. cancer
precise locations of items within the grid squares in the figure
would not be known: which of the items (such as the "problems" in
the UB report) lie somewhere within which of the grid squares is
all that would be known. From the figure it is obvious, assuming
that the two qualitative rankings were meaningfully done in the
first place, that items lying within grid squares A, E and I rank,
for the two risks combined, as groups, in the order: A > E > I.
In effect, these three sets of items are easily ranked according to
the combined risks of cancer and non-cancer effects by a simple
inspection of Figure 6.3.1.
85
-------
MS*
t.
2.
3.
4.
5.
6.
7.
8.
9.
For*;
Non-cancer Predominant
Cancer * non-cancer
The merging
method suggested
here provides the
means for determin-
ing where the
groups of items
lying in the off-
diagonal grid
squares rank rel-
ative to those on
the diagonal and to
each other. Once
this is accomplish-
ed a good start has
been made on the
merged ranking of
the items them-
selves since those
individual items
which need to be Table 6.3.1 Rankings possible for a three-by-
compared to each three linear array
other to arrive at a final ranking have been clearly identified.
The ranking pattern is **;
ADG > BEH > CF!
A>D>G>B>E>U>C>f>I
A>D>BG>E>CX>F>I
A>D>B>G>E>C>H>F>I
A > BO > CE6 > FH > I
A>B>D>C>E>G>F>H>!
A>B>CD>E>FG>H>I
A>B>C>0>E>F>G>H>I
Cancer Predominant ABC > OEF > GMI
* Ueight iaplied, overall, by the rank orders given.
** Grid squares written together (e.g., BD or OEF) are
of the saw rank.
The method is based on the fact that for any of several models
in which severity factors, in principle, can provide the link for
comparing risks of different endpoints, there is only a limited
number of sets of rankings of the groups of items in different grid
squares to be compared with the risk information about the sets of
items to determine which ranking is most consistent with the risk
information. Models relevant to the items or problems of concern
to the Agency include those which rank by individual risk, by
population risk, or by combined individual and population risk,
with or without taking account of other factors such as the
sensitivities of individuals or of population subsets.
For three-by-three, linear arrays of risks such as the one
plotted in Figure 6.3.1 the set of all possible rankings of the
grid squares (and therefore of the items falling within them) is
shown in Table 6.3.1. As shown in section 8.2, it is not necessary
to determine which ranking is most compatible with the available
information by a laborious comparison with each of the rankings
shown; use can be made of the major rank reversals (for example, G
and C in rankings 2., 3. or 4., versus 8., 7., or 6., respectively)
86
-------
to reduce sharply the number or rankings where detailed comparison
is necessary.
As shown in Section 8.2, Table 8.2.6.2, the number of possible
rankings increases to seventeen if the array is symmetrical but
nonlinear (for example, if the ranking coordinates are logarithmic)
and also increases as the order of the array increases. Practical
arrays for the merging of qualitatively ranked items are the three-
by-three arrays, linear or nonlinear, in which the original
rankings fall only into three categories: high, medium, and low.
Four-by-four systems might work if the original information on the
separate rankings is sufficiently complete and descriptive, but a
more highly subdivided set rankings than that soon becomes
cumbersome or outruns the ability of the information to discrim-
inate. Generally speaking, too, when the final ranking has been
achieved, it is desirable to express it in no more than the number
of categories of the original two rankings; to use more would
outrun the content of the original information.
One key point should be borne in mind: it is as true of the
zero-based ranking method and of the separate ranking of items by
cancer and non-cancer risks as it is of the process required to
carry out rank merging that the various comparisons need to be made
by appropriately chosen consensus groups for the comparisons, and
the final result, to be as good in quality and as credible as
possible.
From a practical standpoint, once that the possible ranking
patterns are tabulated this rank merging process can be carried out
without having to know or to decide whether cancer of non-cancer
effects predominate, whether it has been explicitly determined what
the relative severities might be, or whether the risk ranking
scales the relative severities might be, or whether the risk
ranking scales are linear or nonlinear. Comparison of the possible
rankings with the available risk infirmation to determine which is
most in keeping with the information accomplishes this, accounting
for whatever conscious or unconscious decisions may have been made
by those doing the ranking. For this reason, as well as for its
relative simplicity, the rank merging method is a preferable way to
produce merged or aggregated health risk rankings until such time
as a zero-based method can be put into practice.
87
-------
6.3.4
For the long term use of merged cancer and non-cancer risk
ranking, the so-called zero-based procedure outlined early in this
section, is best. Doing it once can form a solid basis for
updating and revising it and, since it deals most directly with the
problem in a manner as close as possible to the flexible and
relatively inclusive models described in detail in Section 8.2, it
is likely to yield the most correct and credible, and therefore
reliable, result when it comes to budgeting and allocating
resources to risk management activities and to research. It is
recommended that this effort be undertaken as an investment in
facilitating better planning and allocation.
One of the key missing sets of variables for producing a
single, health risk based ranking of Problems is a single set of
severities for cancer and non-cancer endpoints together. The
experience already gained in attempting to grade the severities of
different non-cancer endpoints in the UB report should help in the
formulation of a method and a process for undertaking the task of
producing a consensus on a health risk severity table including
both cancer and non-cancer effects, and it is recommended that any
updating of the UB report include this activity.
The procedure for merging separately ranked Problems (for
cancer and non-cancer risk) is relatively easy to use, once the
main possible rankings are tabulated (as for example, in Tables
8.2.6.1 and 8.2.6.2) and once separate cancer and non-cancer risk
rankings are in hand. The consensus mechanism recommended is
particularly useful not only in narrowing down the possible
rankings to one best one but also in reaching the final merged
ranking while ensuring that information that might have been lost
along the way is utilized at the end.
6.4 Development of Necessary Resources
Valid assessment of the health risks associated with
environmental problems will require major improvements in the
relevant exposure and toxicity data, as well as substantial
strengthening of the underlying science base. To expedite the
desired improvements, the following needs should be addressed:
88
-------
Databases For most of the chemical and physical agents of
environmental concern, the relevant data on human exposure are not
sufficiently quantitative, comprehensive, or detailed to enable
precise assessment of the associated risks to human health. Far
more detailed and comprehensive exposure measurements are neces-
sary, including data on tissue burdens as well as ambient exposure
levels. Also needed are pertinent data on the uptake, distribu-
tion, metabolism, and excretion of the substances in question, as
well as on the extent to which these parameters may vary with age,
sex, diet, physiological state, and other variables. The data
should also include, insofar as possible, information on the
relevant biological and molecular markers of exposure, dose, and
preclinical effects.
In addition to better exposure data, more adequate toxicolog-
ical information also is needed, including more systematic data on
the toxicity of the relevant agents for humans of different ages,
more comprehensive assessment of their toxicity in surrogate
toxicological test systems, and better understanding of the
appropriate dose-response and trans-species scaling functions to be
used in assessing their risks to human health.
Institutional Arrangements In order to develop exposure and
toxicity databases of the richness needed, closer cooperation among
different governmental and private institutions will be necessary.
For example, development of the exposure databases should include,
in addition to the data gathered by EPA itself, relevant infor-
mation from other federal (e.g., NCHS, NIH, NIOSH, FDA, and DOE),
State, and local agencies, as well as from the private sector.
Personnel Also in need of further development is scientific
capability in the requisite disciplines. Furthermore, since
assessment of the health risks of environmental agents requires the
coordinated efforts of biologists, chemists, epidemiologists,
mathematicians, physicians, toxicologists, and scientists of other
disciplines, few institutions have the multidisciplinary teams
needed for such research. Measures to develop such collaboration
on a broader scale and to focus it on key problems deserve to be
pursued. Inherent in the development of the needed scientific
capability is the training of scientists with the necessary
expertise. For this purpose, there is need for more long-term
support of graduate and postgraduate training in toxicology,
89
-------
Toxicants that may pose significant risks to human health can
be encountered in air, water, food, consumer products, the home,
the workplace, and other environments. Although in some instances
the risks from such toxicants have been adequately controlled by
limiting human exposure to the agents in question, other
environmental toxicant-related risks to health continue to exist,
as reported in "Unfinished Business." It is important, therefore,
to assess any such risks and to develop measures for controlling
them.
In order to set appropriate priorities for allocating
resources to different environmental risk problems, the relative
importance of each problem must be evaluated. For this purpose,
some sort of comparative risk assessment is required. At present,
however, such assessments must be interpreted with caution, in view
of their large uncertainties.
Among the most serious sources of uncertainty is the inade-
quacy of available data on the extent of human exposure to the
toxicants in question. In few cases has the concentration of a
given toxicant in the relevant exposure media been characterized
well enough in time and space to enable precise estimation of the
patterns and extent of human exposure to the agent in question. In
even fewer cases have environmental exposure measurements of a
toxicant been accompanied by systematic analyses of its uptake,
distribution, metabolism, and retention in the tissues of persons
differing in age, sex, dietary habits, lifestyle, occupation, and
other potentially important variables. In the absence of such
information, quantitative estimation of the extent of human
exposure to most toxicants, and of the exposure-dose relationships
relevant to assessment of their risks to human health, must remain
highly tenuous.
To provide exposure-dose data of the quantity and quality
needed for more adequate assessment of environmental risks to human
health, there is need for far more systematic monitoring of the
environment and of human tissues, including the use of biomarkers
and other newly-developing measures of exposure and effects.
Toward this end, expanded research and data collection are
91
-------
recommended, including closer interagency cooperation and data-
linkages to facilitate development of the requisite networks and
databases.
Another serious limitation in risk assessment results from the
uncertainty inherent in evaluating the toxicity of virtually any
environmental toxicant under conditions of chronic low-level
exposure. For relatively few environmental agents has toxicity for
humans been observed directly, even at relatively high doses, and
in these instances the relevant dose-response relationships and
mechanisms of toxicity have not been defined well enough to enable
risk assessment without reliance on uncertain dose-effect models
for extrapolation to the low dose domain. In these cases the
assessments also involve uncertain assumptions about the influence
of age, sex, and other factors on the susceptibility of the exposed
persons, as well as the extent to which the effects of a given
toxicant may be modified by the action of other environmental
agents. For the majority of environmental toxicants, human data
are lacking altogether, with the result that assessment of their
potential risks must be based on extrapolation from studies of
laboratory animals and other surrogate test systems, that involves
uncertainty about species differences as well as the other
uncertainties mentioned above. For thousands of additional
chemicals to which humans may be exposed, no toxicological data of
any kind are available as yet, precluding even the most rudimentary
assessment of their potential impacts on human health.
In order to improve the assessment of environmental risks to
human health, the following steps must be taken to strengthen the
underlying toxicological science, methodology, and database: 1)
further research on the development and validation of toxicological
testing methods, including analyses of structure-activity relation-
ships and other correlational techniques, short-term in vitro and
in vivo tests, and long-term and inter-generational animal
bioassays; 2) use of these testing methods to screen new chemicals
before they enter commercial use and to test expeditiously existing
chemicals identified as possible hazards; 3) expanded epidemiologi-
cal study of human populations, with particular reference to
populations at increased risk because of elevated levels of
exposure or heightened susceptibility; 4) studies to elucidate the
mechanisms and dose-response relationships of the various types of
health effects that may be associated with low-level exposure to
different toxicants and combinations of toxicants; and 5) inter-
92
-------
national, national, and local interagency cooperation in the
collection of vital statistics and other data, record-linkage, and
networking, so as to enlarge the toxicological database as rapidly
as possible.
In view of the above limitations in the available exposure and
toxicity data, the risk rankings that were assigned in "Unfinished
Business" must be regarded as provisional. Whether the rankings
could be improved greatly in the absence of more adequate data is
problematic. Pending better data and scientific knowledge, it may
be inferred that those environmental problem areas involving the
highest probability of proximal human exposure to toxicants are
likely to pose the largest potential risks to human health. Such
situations include those encountered by the general population
through exposure to pollutants in ambient outdoor air, indoor air,
drinking water, food, and consumer products, and those encountered
by workers in the workplace. It is not illogical, therefore, that
the environmental problems assigned the highest relative risk
rankings for cancer and/or other adverse health effects in
"Unfinished Business" were representative of such exposure
situations; i.e., criteria and hazardous air pollutants, indoor air
pollution and indoor radon exposure, drinking water, pesticide
residues on food, pesticide application, consumer product exposure,
and occupational exposure to chemicals.
Among the latter problems, however, it should be noted that
the "high" risk rankings for the following problems are supported
more firmly by the available data than are the rankings for others:
• criteria air pollutants
• hazardous air pollutants
• indoor air pollutants (excluding radon)
• indoor radon
• drinking water
• pesticide application
• occupational exposure to chemicals)
Another factor seriously complicating the comparative ranking
of environmental risks to health is the diversity of the health
outcomes that are involved. While cancer is clearly a serious
health outcome, a cancer occurring in a 90-year-old man could be
considered less serious than mental retardation in a newborn
infant. In any case, however, quantification of the health impacts
93
-------
of different types of toxicant-induced effects is complicated,
since it must take into account both the aggregate numbers of all
persons who are affected, including those affected indirectly as
well as those affected directly, and the severity of the effects
judged in terms of their physical, psychological, social, and
economic impacts. Detailed consideration of these ramifications
calls for more detailed analyses than have been conducted thus far
and may not be feasible without further refinement in the data.
In addition to the relative magnitudes of the health impacts
of different environmental risks, their controllability must also
be considered in evaluating alternative risk-reduction strategies.
It must not be forgotten, therefore, that the adverse health
outcomes caused by certain environmental toxicants—such as
carcinogens—may not appear until decades after exposure, with the
result that termination of exposure to the toxicants does not
suffice to abolish risk in those who have already been exposed. It
is also noteworthy that certain environmental toxicants—such as
heavy metals, PCBs, and long-lived radionuclides—tend to persist
indefinitely in the environment and may gradually become
concentrated in certain components of the human food chain.
Consequently, such toxicants may continue to pose a threat to human
health long after their release into the environment has been
halted.
It must also be recognized that, in many instances over the
past 20 years, EPA has undertaken programs to reduce risks
attributable to specific substances in the environment, either
through legislative mandate (as in the case of PCBs under TSCA) or
through utilizing regulatory powers (as in the case of lead in
gasoline). However, none of these risk reduction programs has been
complete in terms of absolutely banning all production and use
(including in situ uses as with PCB containing electrical
equipment) of these substances. Residual risks remain associated
with these continued uses, including waste disposal. Nevertheless,
EPA has already devoted considerable efforts to identify the risks
of these substances, through epidemiological studies, other
research on toxicity, and exposure assessments. Similarly, the
private sector has already invested in partial control technologies
or substitute materials. Thus the major expenses of risk reduction
may in these cases have already been incurred (e.g., capital
investment in catalytic cracking units at oil refineries to produce
additives for unleaded gasoline). In these cases, the cost of
94
-------
further risk reduction—even risk elimination—may be relatively
small, as compared to undertaking risk reduction 3s novo for
substances and exposures not previously addressed in any
substantial fashion. EPA should consider these factors in
evaluating strategies for relative risk prioritization and for
implementing risk reduction measures.
Limited as the existing data may be for assessing recognized
risks to health, our capacity to predict future risks and to
respond to emerging problems is even more severely limited. There
is need, therefore, for the establishment of a formal mechanism for
risk anticipation, including an in-house expert committee, peer
oversight, and a means of supporting long-range research on
emerging problem areas. Emerging problems that merit attention at
this time would appear to include the potential risks associated
with low-level exposure to 60 Hz magnetic fields.
Finally, the development of any long-range strategy to improve
environmental risk assessment and risk reduction will require
provision for developing and sustaining the needed scientific
capability and workforce. This will necessitate programs for
graduate and postgraduate training in the relevant disciplines, as
well as the development of measures to enlist and nurture the
participation of the scientific community in the kinds of
interdisciplinary research that are required.
95
-------
8.1 case studies
8.1.1 Ozone Case Study
Dr. Morton Lippmann
New York University
Institute of Environmental Medicine
97
-------
8.1.1.1 Introduction and Background
Ozone (O3) was recognized by Schonbein (1851) as a powerful
lung irritant soon after its initial synthesis (Bates, 1989). It
was first listed among the American Conference of Governmental
Industrial Hygienists (ACGIH) list of Threshold Limit Values (TLVs)
for occupational exposure in 1946, with an eight-hour time weighted
average (TWA) concentration limit of 1 ppm. In 1954, the TLV was
reduced to 0.1 ppm TWA. The current ACGIH TLV of 0.1 ppm, as a
ceiling value, was adopted in 1989.
Health effects among the general community were first reported
among high school athletes in California, in terms of lesser
performance on high exposure days (Wayne et al., 1967). The
initial National Ambient Air Quality Standard (NAAQS) of 1971 was
0.08 ppm of total oxidant. The NAAQS was revised in 1979 to 0.12
ppm of O3, and was based upon clinical studies by DeLucia and Adams
(1977) showing that exercising asthmatic adults exposed for 1 hr to
0.15 ppm in a test chamber had increased cough, dyspnea, and
wheezing, along with small, but nonsignificant reductions in
pulmonary function (U.S. EPA, 1986). A small margin of safety was
applied to protect against adverse effects not yet uncovered by
research and effects whose medical significance is a matter of
disagreement. In its May 1, 1989 closure letter to the EPA
Administrator on its reviews of the 1986 Ozone Criteria Document
(CD), the 1988 CD Supplement, and the Agency Staff Paper of 1988,
the Clean Air Scientific Advisory Committee (CASAC) split on its
recommendation to the Administrator concerning a scientifically
supportable upper bound to the range for a revised 1 hr NAAQS, with
half the members accepting 0.12, and the other half recommending a
reduced upper bound (CASAC-1989).
The effects of concern with respect to acute response in the
population at large are reductions in lung function and increases
in respiratory symptoms, airway reactivity, airway permeability,
and airway inflammation. For asthmatics, there are increased rates
of medication usage and restricted activities. Margin of safety
considerations include: 1) the influence of repetitive elicitation
of such responses in the progression of chronic damage to the lung
of the kinds seen in chronic exposure studies in rats and monkeys;
and 2) evidence from laboratory and field studies that ambient air
co-pollutants potentiate the responses to 03
98
-------
03 is almost entirely a secondary air pollutant, formed in the
atmosphere through a complex photochemical reaction sequence
requiring reactive hydrocarbons, nitrogen dioxide (N02) and
sunlight. It can only be controlled by reducing ambient air
concentrations of hydrocarbons, N02, or both. NO and N02 are
primary pollutants, known collectively as NOx. In the atmosphere,
NO is gradually converted to N02. One of the major sources of
hydrocarbons and N0x, i.e., motor vehicles, has been the major
target of control efforts, and major reductions (> 90%) have been
achieved in hydrocarbon emissions per vehicle. NOX from stationary
source combustion has increased, and there has also been major
increases in vehicle miles driven. The net reduction in exposure
has been modest at best. In 1986-1988, there were high levels of
ambient 03 with exceedances of the current NAAQS recorded in 101
communities with a total population of 112 million people.
The risks remain very high for demonstrable acute responses,
and potentially very high for the still poorly defined chronic
health risks, especially premature aging of the lungs.
8.1.1.2. Current Knowledge on Exposure and Sources
A. Exposures
1. Personal Air No personal monitors have been
available; hence there are no data.
2. Microenvironmental Air
a. Ambient Air Extensive data are available from
continuous monitors at many urban and some rural sites since the
early 1970's. Most readily available data are on one hour maximum
concentrations and numbers of exceedances of the one hour NAAQS of
0.12 ppm. Data on distributions of concentrations over various
averaging times are not normally reported.
b. Indoor Air Relatively few data are available.
A recent review by Weschler et al. (1989) indicates that
indoor/outdoor ratio (I/O) varies from 0.2 to 0.8, averaging about
0.5.
99
-------
c. Transportation 03 within motor vehicles is
generally lower than in outdoor air because of efficient scrubbing
by tailpipe NO in transportation corridors. 03 in cabins of
jetliners flying in the stratosphere can be quite high due to high
concentrations in compressed stratospheric air used to ventilate
the cabins (NRG, 1988).
d. Other Electrostatic air cleaners generate 03
that can be distributed widely through ducts to occupied spaces.
Xerographic copying machines can elevate 03 in rooms containing
them. A major source of occupational exposure is arc welding.
3. Ingestion Not applicable to 03
4. Dermal Not applicable to 03
5. Overall Exposure Biomarkers Not applicable to 03
B. Populations Exposed
1. Healthy Adults With children, healthy young adults
are the most sensitive to the acute effects of 03) especially those
engaged in active exercise out-of-doors (McDonnell et al., 1983;
McDonnell et al., 1985).
2. Infants and Children No data on infants. Children
and adolescents are, with young adults the most responsive to acute
effects. Children may be at greater risk because of more time out-
of-doors.
3. Elderly Healthy elderly adults are less responsive
than younger people to 03 in terms of acute effects (Drechsler-
Parks et al., 1987; Reisenauer et al., 1988).
4. Susceptible Subgroups Healthy children and young
adults are the most responsive to 03 in terms of acute functional
decrements, and no biomarkers of susceptibility have yet been
identified. Since exercise during exposure potentiates acute
responses, healthy individuals exercising out-of-doors are an
especially susceptible group.
Another susceptible subgroup are asthmatics, based on reports
of increased medication usage and restricted activities during high
100
-------
03 days. Whittemore and Korn (1980) reported that daily asthma
attack rates were increased on days with high oxidant levels in Los
Angeles area communities. Holguin et al. (1985) reported a similar
association for asthmatics in Houston.
C. Factors Modifying Effective Dose
1. Activity Level The effect of ventilation rate on
acute functional response has been summarized by Hazucha (1987).
Response increases progressively with minute ventilation over the
range of available data (0-68 L/min). However, at levels above 80
L/min, ^creasing ventilation reduces the response (Spektor et
al., 19S t .
2. Pre-existing Disease No data available.
3. Constitutional Factors Affecting Uptake and Retention
Studies of regional particle deposition in healthy humans show a
large degree of variability in conductive airway caliber, affecting
the distribution and depth of penetration of tidal air (Bohning et
al., 1975; Chan et al., 1980). Combined with O3 dosimetry models
(Miller et al., 1978; Overton et al., 1987). These differences
could account for unexplained variability in acute responsiveness
to 03 among healthy persons.
4. Constitutional Factors Affecting Metabolic
Transformation Not applicable to 03
D. Sources
1. Energy Production Sources of hydrocarbons and
nitrogen oxides vary greatly by region, season, and time of day.
Stationary fossil fuel combustion accounts for almost half of
ambient NOX.
2. Transportation Motor vehicles account for almost
31% of N0x emissions and some 26% of the hydrocarbons.
Transportation in total accounted for 41% of the NOX and 33% of
hydrocarbons.
101
-------
3. Other Sources of Hydrocarbons Other sources of
hydrocarbons vary greatly according to region and season. In the
southeastern U.S. in the summer the transpiration of trees and
shrubs can be the dominant source. Other significant sources are
fugitive emissions from petrochemical plants, sewage treatment
plants, agricultural operations and consumer product usage.
8.1.1.3 Toxicitv and Health Effects
A. Human - Clinical Studies
1. Laboratory Studies The major focus of the extensive
body of data on the health effects of a single day's maximum hourly
exposure to ambient 03 The 1971 and 1979 NAAQS for photochemical
oxidants were based on the maximum 1 hr concentrations as the
relevant index of exposure, and this, in turn, has focused most of
the clinical research on exposure protocols involving either 1 or
2 hours of exposure. However, recent research has shown that
effects can be produced with exposures as short as 5 minutes (Fouke
et al., 1988) , and that various effects become progressively larger
as exposures at a given concentration are extended out to 6.6 hours
(Folinsbee et al., 1988; Horstman et al., 1989).
There are more data on respiratory function responses than on
any of the other coincident responses to short-term 03 inhalation.
The major debate about very small, but statistically significant,
decrements in function from such studies is how to interpret their
health significance (Lippmann, 1988).
The inhalation of 03 causes concentration dependent mean
decrements in exhaled volumes and flow-rates during forced
expiratory maneuvers, and the decrements increase with depth of
breathing (Hazucha, 1987). There is a wide range of reproducible
responsiveness among healthy subjects (McDonnell et al., 1985), and
functional responsiveness to O3 is no greater, and usually lower,
among cigarette smokers (Kagawa, 1984; Shephard et al., 1983),
older adults (Drechsler-Parks et al., 1987; Reisenauer et al.,
1988), asthmatics (Koenig et al., 1987; Linn et al., 1983), and
patients with chronic obstructive pulmonary disease (COPD) (Linn et
al., 1983; Solic et al., 1982). The only exception is that
patients with allergic rhinitis had greater changes in airway
resistance (McDonnell et al., 1987).
102
-------
The effects of 03 on respiratory function accumulate over
time. Folinsbee et al. (1988) undertook a chamber exposure study of
10 adult male volunteers involving 6.6 hours of 03 exposure at 120
ppb. Moderate exercise was performed for 50 min/h for 3 hours in
the morning, and again in the afternoon. They found that the
functional decrements become progressively greater after each hour
of exposure, reaching average values of 400 Ml for forced vital
capacity (FVC) and 540 Ml for forced expiratory volume in one
second (FEV1) by the end of the day. Follow-up studies by Horstman
et al. (1989) were done on 21 adult males with 6.6 hour exposures
at 80, 100, and 120 ppb. The exposures at 120 ppb produced very
similar responses, e.g., a mean FEV1 decline of 12.3 percent while
those at 80 and 100 ppb showed lesser changes that also became
progressively greater after each hour of exposure.
The time scale for the biological integration of 03 exposure
can also be deduced from the rate at which the effects dissipate.
Folinsbee and Hazucha (1989) studied 18 young adult females exposed
to 350 ppb 03 for 70 min, including two 30 min periods of treadmill
exercise at 40 L/min. The responses were highly variable, from
zero to 40%. Their mean decrement in FEVl at the end of the
exposure was 21 percent. After 18 hours, their mean decrement was
4 percent, while at 42 hours it was 2 percent.
In summary, it is now clear that the respiratory function
effects can accumulate over many hours, and that an appropriate
averaging time fpr transient functional decrements caused by 03 is
6 hours. Thus, there is less scientific basis for the current
health based exposure limit with an averaging time of 1 hour than
previously believed. Since 03 exposures in ambient air now can
have broad peaks with 8 hour averages equal to 90 percent of the
peak 1 hour averages (Rombout et al., 1986), the functional
decrements associated with ambient concentrations are likely to be
much greater than those predicted on the basis of the responses in
the chamber studies following 1 to 2 hour exposures.
Respiratory symptoms have been closely associated with group
mean pulmonary function changes in adults acutely exposed in
controlled exposures to 03 However, Hayes et al. (1987) found only
a weak-to-moderate correlation between FEVl changes and symptoms
severity when the analysis is conducted using individual data.
103
-------
Exposure to 03 can also alter the responsiveness of the
airways to other bronchoconstrictive challenges as measured by
changes in respiratory mechanics. For example, Folinsbee et al.
(1988) reported that airway reactivity to the bronchoconstrictive
drug methacholine for the group of subjects as a whole was
approximately doubled following 6.6 hour exposures to 120 ppb. O3.
On an individual basis, Folinsbee et al. (1988) found no apparent
relationship between the 03-associated changes in methacholine
reactivity and those in FVC or FEV1. The follow up tests by
Horstman et al. (1989), involving 6.6 hour exposures to 80, 100 and
120 ppb indicated 56, 89 and 121 percent increases in methacholine
responsiveness respectively.
Koren et al. (1989) reported that an inflammatory response, as
indicated by increased levels of PMN, was also observed in BAL
fluid from subjects exposed to 100 ppb 03 for 6.6 hours. The 6.6
hours at 100 ppb 03 produced a 4.8x increase in PMNs at 18 hours
after the exposure. Since the amount of O3 inhaled in the 100 ppb
protocol was -2.5 ^q, while it was -3.6 jig in a prior 400 ppb
protocol (Koren et al., 1989), we might have expected a 2.5/3.6 x
8.2 = 5.7 times increase in PMNs. The close correspondence of the
observed to expected ratio suggests that lung inflammation from
inhaled 03 also has no threshold down to ambient background 03
levels.
Foster et al. (1987) studied the effect of 2-hour exposures to
200 or 400 ppb 03 with intermittent light exercise on the rates of
tracheobronchial mucociliary particle clearance in healthy adult
males. The 400 ppb 03 exposure produced a marked acceleration in
particle clearance from both central and peripheral airways, as
well as a 12 percent drop in FVC. It is of interest that the 200
ppb O3 exposure produced a significant acceleration of particle
clearance in peripheral airways, but failed to produce a
significant reduction in FVC, suggesting that significant changes
in the ability of the deep lung to clear deposited particles take
place before significant changes in respiratory function take
place.
The weight of the evidence from these results, showing both
functional and biochemical responses that accumulate over multiple
hours and persist for many hours or days after exposure ceases, is
clear and compelling.
104
-------
2. Field Studies Spektor et al. (1988a) found that
children at summer camps with active outdoor recreation programs
had greater decrements in lung function than children exposed to 03
at comparable concentrations in chambers for 1 or 2 hours.
Furthermore, their activity levels, although not measured, were
known to be considerably lower than those of the children exposed
in the chamber studies while performing very vigorous exercise.
Since it is well established that functional responses to 03
increase with levels of physical activity and ventilation (Hazucha,
1987) , the greater responses in the camp children had to be caused
by other factors, such as greater cumulative exposure, or to the
potentiation of the response to 03 by other pollutants in the
ambient air. Cumulative daily exposures to 03 were generally
greater for the camp children, since they were exposed all day long
rather than for a 1 or 2-hour period preceded and followed by clean
air exposure.
A follow-up (Spektor et al. 1988b) study addressed the issue
of the potentiation of the characteristic functional response to
inhaled O3 by other environmental cofactors. It involved healthy
ad It nonsmokers engaged in a daily program of outdoor exercise
with exposures to an ambient mixture containing low concentrations
of acidic aerosols and N02 as well as 03. Each subject did the same
exercise each day, but exercise intensity and duration varied
widely between subjects, with an average minute ventilation of 79
liters, and with duration of daily exercise averaging 29 min.
Respiratory function measurements were performed immediately before
and after each exercise period. 03 concentrations during exercise
ranged from 0.021 to 0.124 ppm. All measured functional indices
showed significant (p<0.01) 03 associated mean decrements. It was
concluded that ambient cofactors potentiate the responses to 03.
B. Human - Epidemiology
1. Acute Effects Kinney et al. (1988) studied school
children in Kingston and Harriman, Tennessee, whose lung function
was measured in school on up to six occasions during a 2-month
period in the late winter and early spring. Child specific
regressions of function versus maximum 1-hour 03 during the
previous day indicated significant associations between 03 and
function, with coefficients similar to those seen in the summer
camp studies of Lippmann et al. (1983) and Spektor et al. (1988a).
Since children in school may be expected to have relatively low
105
-------
activity levels, the relatively high response coefficients may be
due to potentiation by other pollutants or to a low-level of
seasonal adaptation. Kingston-Harriman is notable for its
relatively high levels of aerosol acidity. As shown by Spengler et
al. (1989), Kingston-Harriman has higher annual average and higher
peak acid aerosol concentrations than other cities studied, i.e.,
Steubenville, Ohio; St. Louis, Missouri; and Portage, Wisconsin.
Alternatively, the relatively high response coefficients could have
been due to the fact that the measurements were made in the late
winter and early spring. Linn et al. (1988) have shown evidence
for a seasonal adaptation, and children studied during the summer
may not be as responsive as children measured earlier in the year.
2. Chronic Effects Epidemiologic studies of populations
living in Southern California suggest that chronic oxidant
exposures do affect baseline respiratory function. Detels et al.
(1987) compared respiratory function at two points in time five
years apart in Glendora (a high oxidant community) and in Lancaster
(a lower oxidant community-but not low by national standards).
Baseline function was lower in Glendora, and there was a greater
rate of decline over 5 years. The annual change in lung function
in Glendora was much greater than that in Lancaster, that, in turn,
was much greater than that in Tucson, Arizona (Knudson et al.,
1983) for a comparable population of Caucasian non-smokers. The
second highest 1 hour 03 concentrations in Tucson in all of 1981,
1982, and 1983 were 100, 120 and 110 ppb (EPA, 1986). In Lancaster
there were 58 days in 1985 with 1 hr 03 maxima greater than 120
ppb, while in Azusa, adjacent to Glendora, there were 117 days in
1985 with 1 hr 03 maxima greater than 120 ppb. Thus, the three
different rates of function decline appear to suggest an exposure-
response relationship with potentially significant health
importance.
Further evidence for chronic effects of 03 were recently
reported by Schwartz (1989) based upon an analysis of pulmonary
function data in a national population study in 1976-80, i.e., the
second National Health and Nutrition Examination Survey (NHANES
II) . Using ambient O3 data from nearby monitoring sites, he
reported a highly significant 03 associated reduction in lung
function for people living in areas where the annual average 03
concentrations exceeded 40 ppb. On the other hand, there were no
significant correlations with other indices of 03 exposure, and the
results should be interpreted cautiously at this time.
106
-------
C. Animal Toxicology
1. Acute Effects studies in laboratory animals have
examined the roles of O3 concentration and exposure time on
biochemical and cellular responses. Rombout et al. (1989) exposed
mice and rats to 380, 750, 1250, and 2,000 ppb 03 for 1, 2, 4, and
8 hours, and measured broncho-alveolar lavage (BAL) protein with
both daytime and nighttime exposures. Observation times extended
from 1 to 54 hours. The responses varied with 03 concentration,
duration of exposure, time after the start of the exposure, and
minute volume, with time of exposure having a greater than
proportional influence. For 4 and 8-hour exposures, the protein
content of BAL peaked at 24 hours, and remained at elevated levels
even at 54 hours. As indicated previously, Koren et al. (1989)
found increased BAL protein in humans 18 hours after an exposure to
100 ppb O3 for 6.6 hours.
The effects of O3 on mucociliary particle clearance have been
studied in rats and rabbits. Rats exposed for 4 hours to 03
exhibited a slowing of particle clearance at 800 ppb (Frager et
al., 1979; Kenoyer et al., 1981). Rabbits exposed for 2 hours at
100, 250 and 600 ppb 03 showed a concentration dependent trend of
reduced clearance rate with increasing concentrations, with the
change at 600 ppb being - 50 percent and significantly different
from control (Schlesinger and Driscoll, 1987).
Phipps et al. (1986) examined the effects of acute exposure to
03 on some of the factors that affect mucociliary transport rates
in studies in which sheep were exposed to 500 ppb 03 for 2 hours
on two consecutive days. The exposures produced increased basal
secretion of sulfated glycoproteins, but had no effect on ion
fluxes. Their histological examination indicated a moderate
hypertrophy of submucosal glands in the lower trachea, and they
concluded that the exposure caused airway mucus hypersecretion.
Studies of the effects of O3 on alveolar macrophage mediated
particle clearance during the first few weeks have also been
performed in rabbits. Rabbits exposed to 100, 600, or 1200 ppb 03
once for 2 hours had accelerated clearance at 100 ppb and retarded
clearance at 1200 ppb. Rabbits exposed for 2 hours/day for 13
consecutive days at 100 or 600 ppb 03 had accelerated clearance
for the first 10 days, with a greater effect at 600 ppb (Driscoll
et al., 1986).
107
-------
The responses of the alveolar macrophages to these exposures
was examined by Driscoll et al. (1987). These studies demonstrated
significant alterations in the numbers and functional properties of
alveolar macrophages as a result of single or repeated exposure to
100 ppb ozone, a level frequently encountered in areas of high
photochemical air pollution.
Both in vivo and in vitro studies have demonstrated that 03
can affect the ability of the immune system to defend against
infection. Increased susceptibility to bacterial infection has been
reported in mice at 80 to 100 ppb 03 for a single 3 hour exposure
(Coffin et al., 1967; Ehrlich et al., 1977; Miller et al., 1978).
Related alterations of the pulmonary defenses caused by short-term
exposures to 03 include: impaired ability to inactivate bacteria
in rabbits and mice (Coffin et al., 1968; Coffin and Gardner, 1972;
Goldstein et al., 1977; Ehrlich et al., 1979), and impaired
macrophage phagocytic activity, mobility, fragility and membrane
alterations, and reduced lysosomal enzymatic activity (Witz et al.,
1983; Dowell et al., 1970; Hurst and Coffin, 1971; Hurst et al.,
1970; Goldstein et al., 197la; Goldstein et al., 1971b; McAllen et
al., 1981; Amoruso et al., 1981). Some of these effects have been
shown to occur in a variety of species including mice, rats,
rabbits, guinea pigs, dogs, sheep, and monkeys.
Other studies indicate similar effects for short-term and
subchronic exposures of mice to O3 combined with pollutants such
as 502, N02, H2S04 and particles (Gardner et al., 1977; Aranyi et
al. . 1983; Ehrlich, 1980; Grose et al., 1980a; Grose et al., 1980b;
Phalen et al., 1980). Similar to human pulmonary function response
to O3 activity levels of mice exposed to 03 has been shown to play
a role in determining the lowest effective concentration that
alters the immune defenses (Illing et al., 1980). In addition, the
duration of exposure must be considered. In groups of mice exposed
to 200 ppb 03 for 1, 3, or 6 hours, superoxide anion radical
production decreased 8, 18, and 35%, respectively, indicating a
progressive decrease in bacteriocidal capacity with increasing
duration of exposure (Amoruso and Goldstein, 1988).
The major limitation of this large body of data on the
influence of inhaled 03 on lung infectivity is that it
requires uncertain interspecies extrapolating in order to estimate
the possible effects of 03 on infectivity in humans.
108
-------
2. Chronic Effects It is well established that
repetitive daily exposures, at a level which produces a functional
response upon single exposure, result in an enhanced response on
the second day, with diminishing responses on days 3 and 4, and
virtually no response by day 5 (Farrell et al., 1979; Folinsbee et
al., 1980; Hackney et al., 1977). This functional adaptation to
exposure disappears about a week after exposure ceases (Horvath et
al., 1981; Kulle et al., 1982). The adaptation phenomenon has led
some people to conclude that transient functional decrements are
not important health effects. On the other hand, recent research
in animals has shown that persistent damage to lung cells
accumulates even as functional adaptation takes place. Tepper et
al. (1987) exposed rats to 350, 500, or 1000 ppb O3 for 2.25 hours
on five consecutive days. Carbon dioxide (8%) was added to the
exposure during alternate 15 min periods to stimulate breathing and
thereby increase 03 uptake and distribution. Tidal volume,
frequency of breathing, inspiratory time, expiratory time and
maximal tidal flows were affected by 03 during day 1 and 2 at all
03 concentrations. By day 5, these 03 responses were completely
adapted at 350 ppb, greatly attenuated at 500 ppb, but showed no
signs of adaptation in the group exposed to 1000 ppb. Unlike the
pulmonary function data, light microscopy indicated a pattern of
progressive epithelial damage and inflammatory changes associated
with the terminal bronchiole region. These data suggest that
attenuation of the pulmonary functional response occurs while
aspects of the tissue response reveal progressive damage.
The effects of multi-day 03 exposures of laboratory animals on
particle clearance from the lungs and on lung infectivity were
reviewed previously. They also show that 03 -induced transient
effects often become greater with repetitive exposures.
Last (1989) reported synergistic interaction in rats, in terms
of a significant increase in lung protein content, following 9 day
exposures at 200 ppb O3 with 20 or 40 Mg/ro3 H2S04, and a non-
significant increase for 9 days at 200 ppb 03 with 5 vg/m3 H2S04.
The highest 03 dose is received at the acinus, where the
terminal bronchioles lead into alveolar ducts, and a series of
studies has shown that the effects of inhaled 03 on lung structure
is also greatest in this region. Using morphometric techniques
selectively focussed on this limited region of the lung. Barry et
al. (1985) showed that significant changes occurred in the alveoli
109
-------
just distal to the terminal bronchioles in rats exposed for 12
hr/day for 6 or 12 weeks to 120 or 240 ppb O3 From physiological
studies of rats that were simultaneously exposed Raub et al. (1983)
reported that there were significant increases in the vital
capacity and end expiratory volume that suggested alterations in
distensibility of the lung tissue.
The plausibility of accelerated aging of the human lung due to
chronic 03 exposure is greatly enhanced by the results of recent
chronic animal exposure studies in rats and monkeys, especially
those in rats of Huang et al. (1988) and Grose et al. (1989) using
a daily cycle with a 180 ppb average over 9 hrs superimposed on a
13 hr base of 60 ppb, and those in monkeys of Hyde et al. (1989)
and Tyler et al. (1988) using 8 hr/day of 150 and 250 ppb. The
persistent cellular and morphometric changes produced by these
exposures in the terminal bronchioles and proximal alveolar region,
and the functional changes consistent with a stiffening of the lung
reported by Raub et al. (1983) and Tyler et al. (1988) are
certainly consistent with the results of the epidemiological
studies.
There has long been interest in the possible role of 03 in
lung cancer because of its radiomimetic properties. A
comprehensive review of these issues has recently been prepared by
Witschi (1988). His analysis indicated that there is, to date, no
epidemiological or experimental evidence to support the hypothesis
that O3 is a pulmonary carcinogen. There are data that show that
03 increases the incidence of lung tumors in strain A mice, but
the tumor yield can be either increased or decreased depending on
the exposure protocol. Also, the proliferation of pulmonary
neuroendricrine cells, the precursor cells for small cell lung
cancer can be altered by O3 exposure. Witschi concluded that
there is little evidence to implicate 03 as a pulmonary carcinogen,
but that it might modify and influence the carcinogenic process in
the lung.
3. Mechanistic Studies investigating mechanisms of 03
toxicity in animals have been included in the previous discussions.
The best discussion on the mechanisms of the acute functional
responses in humans was recently presented by Hazucha et al (1989).
110
-------
D. In-Vitro Assay^
1. Genotoxic Effects An EPA Criteria Document (Air Quality
Criteria for Ozone and Other Photochemical Oxidants. 1986) reviewed
the genotoxic effects of ozone. It noted that "the mutagenic
properties of 03 have been demonstrated in procaryotic and
eucaryotic cells. Only one study, however, (Hamelin and Chung,
1975a, with £. colil investigated the mutagenic effect of 03 at
concentrations of less than 1 ppm. The results clearly indicate
that if cells in cultures are exposed to sufficiently high
concentrations of 03 for significantly long periods, mutations will
result. The relevance of the presently described investigations to
human or even other mammalian mutagenicity is not apparent.
Additional studies with human and other mammalian cells will be
required before the mutagenic potency of 03 toward these species
can be determined."
A more recent review of the pathobiology of 03-induced damage
at the cellular and molecular levels by Steinberg et al., (1990)
concluded that 03 linearizes circular DNA and induces 03 sensitive
pneumocytes to repair its DNA. DNA adducts from 03 exposure free
radical damage effect—aging, cellular transformation, mutagenesis,
carcinogenesis, and cell death. DNA-binding proteins are potent
positive and negative regulators, enhancers, or silencers of gene
expression. Part of their action ,may be related to their ability
to initiate the binding sequence of DNA transcription proteins and
thus form complexes. Alteration of DNA-binding sites by 03 adducts
may affect mRNA transcription due to altered binding by DNA-binding
proteins.
In a recent study by Harder et al., (1990) the effect of in
vitro O3 exposure on human peripheral blood natural killer (NK)
cell activity was measured using K562 tumor target cells. The NK
activity was inhibited in a time-dependant manner with marked
suppression observed after 6 hours at three different levels of 03
exposure (1.0, 0.5, and 0.18 ppm) and effector cell:target (E:T)
ratios (50:1, 25:1, and 12.5:1) compared to air controls (p <
0.05). The capacity of O3 exposed NK cells to kill tumor cells
decreased in a linear fashion as the level of 03 increased from
0.18 to 1.0 ppm (p - 0.006 at 50:1; 0.004 at 25:1). Unexposed
cells treated with supernatant from 03 exposed cells showed no
decrease in NK activity.
Ill
-------
2. Cellular Function Leikauf et al. (1988) investigated
the hypothesis that oxidant damage to the tracheal epithelium may
result in elaboration of various eicosanoids. To examine
eicosanoid metabolism after exposure to 100 ppb to 10.0 ppm ozone,
epithelial cells derived from bovine trachea were isolated and
grown to confluency. Monolayers were alternately exposed to ozone
and culture medium for 2 hours. There were 03-induced increases in
cyclooxygenase and lipoxygenase product formation. Ozone
concentrations as low as 100 ppb produced an increase in
prostaglandin F2a. Thus, ozone can augment eicosanoid metabolism
in airway epithelial cells.
In a study focussed on the effects of the 6 week exposures at
250 ppb on the terminal bronchioles, Barry et al. (1988) reported
that exposure to 03 produced alterations in the surface
characteristics of ciliated and nonciliated (Clara) cells in rats.
Rats were also exposed to 03 in tests in which there was a
daily cycle with a baseline of 60 ppb for 13 hr with a 5 day/week
broad peak for 9 hr averaging 180 ppb and containing a 1 hr maximum
of 250 ppb for a period of 3 or 12 weeks. Combining the results of
all these tests, Huang et al. (1988) reported that hyperplasia of
type I alveolar cells in the proximal alveoli was linearly related
to the cumulative 03 exposure. Thus, there is no threshold for
cumulative lung damage and any future standard to protect against
chronic health damage from O3 should have a seasonal or annual
averaging time.
Rats exposed for 6 weeks to clean air or to 03 using the
daily cyclic exposure regimen used by Huang et al. (1988) were
exposed once for 5 hr to an asbestos aerosol by Pinkerton et al.
(1988) . When sacrificed 30 days later, the fiber count in the
lungs of the O3 exposed animals were 3 times greater than in the
sham exposed animals. Thus, subchronic O3 exposure can increase
the effective dose of insoluble particles that may have toxic
and/or carcinogenic effects.
One year of 03 exposure to the same daily cycle caused: (1)
functional lung changes indicative of a "stiffer" lung; (2)
biochemical changes suggestive of increased antioxidant metabolism;
and (3) no observable immunological changes (Grose et al., 1989).
112
-------
Studies at relatively low 03 concentrations have also been
done in monkeys. Hyde et al. (1989) exposed them to O3 for 8
hr/day for 6 or 90 days to 150 or 300 ppb. Responses included
ciliated cell necrosis, shortened cilia, and secretory cell
hyperplasia with less stored glycoconjugates in the nasal region.
Respiratory bronchiolitis observed at 6 days persisted to 90 days
of exposure. Even at the lower concentration of 150 ppb 03,
nonciliated bronchiolar cells appeared hypertrophied and increased
in abundance in respiratory bronchioles.
For some chronic effects, intermittent exposures can produce
greater effects than those produced by a continuous exposure regime
that results in higher cumulative exposures. For example, Tyler et
al. (1988) exposed two groups of 7 month old male monkeys to 250
ppb 03 for 8 hr/day either daily or, in the seasonal model, on
days of alternate months during a total exposure period of 18
months. A control group breathed only filtered air. Monkeys from
the seasonal exposure model, but not those exposed daily, had
significantly increased total lung collagen content, chest wall
compliance, and inspiratory capacity. All monkeys exposed to 03
had respiratory bronchiolitis with significant increases in related
morphometric parameters. Even though the seasonally exposed
monkeys were exposed to the same concentration of O3 for only half
as many days, they had larger biochemical and physiological
alterations and equivalent morphometric changes as those exposed
daily. Lung growth was not completely normal in either exposed
group. Thus, long-term effects of oxidant air pollutants that have
a seasonal occurrence may be more dependent upon the sequence of
polluted and clean air than on the total number of days of
pollution, and estimations of the risks of human exposure to
seasonal air pollutants from effects observed in animals exposed
daily may underestimate long-term pulmonary damage.
The preceding chronic animal exposure studies were performed
at concentrations that occur frequently in ambient air, at least in
Southern California. Thus, the effects observed may be considered
directly relevant to human health, especially in view of our
knowledge that humans receive even greater local doses of 03 in
the vicinity of the acinus than do rats.
A number of other interesting chronic exposure studies have
been done in animals with 03 concentrations in the range of 300 to
113
-------
1000 ppb. Those of them that appear to provide useful insights into
mechanisms of toxic action have been reviewed by Lippmann (1989).
E. Structure-Activity Relationship Not applicable to 03
F. Biomarkers of Response Not applicable to 03
G. Overall Toxicity Assessment In terms of functional
effects, single 03 exposures to healthy non-smoking young adults
at concentrations in the range of 80-200 ppb produce a complex
array of pulmonary responses including decreases in respiratory
function and athletic performance, and increases in symptoms,
airway reactivity, neutrophil content in lung lavage, and rate of
mucociliary particle clearance. Responses to 03 in purified air
in chambers occur at concentrations of 80 or 100 ppb when the
exposures involve moderate exercise over 6 hr or more and require
concentrations of 180 or 200 ppb when the duration of exposure is
2 hr or less. On the other hand, mean FEV1 decrements 5% have been
seen at 100 ppb of O3 in ambient air for children exposed all day
at summer camps and for adults engaged in outdoor exercise for only
1/2 hr. The apparently greater responses to peak 03 concentrations
in ambient air may be due to the presence of, or prior exposures
to, acidic aerosol, but further investigation of this tentative
hypothesis is needed.
Further research is also needed to establish the
interrelationships between small transient functional decrements,
such as FEV1, PEFR, and mucociliary clearance rates, that may not
in themselves be adverse effects, and changes in symptoms,
performance, reactivity, permeability and neutrophil counts. The
latter may be more closely associated with adversity in themselves
or in the accumulation or progression of chronic lung damage.
Successive days of exposure of adult humans in chambers to 03
at current high ambient levels leads to a functional adaptation in
that the responses are attenuated by the third day, and are
negligible by the fifth day. On the other hand, a comparable
functional adaptation in rats does not prevent the progressive
damage to the lung epithelium. Daily exposures of animals also
increase other responses in comparison to single exposures, such as
a loss of cilia, a hypertrophic response of Clara cells,
alterations in macrophage function, and alterations in the rates of
particle clearance from the lungs.
114
-------
For children exposed to 03 in ambient air there was a week-
long baseline shift in peak flow following a summer haze exposure
of four days duration with daily peak 03 concentrations ranging
from 125 to 185 ppb (Lioyatch, 1985). Since higher concentrations
used in adult adaptation studies in chambers did not report such
effects, it is possible that baseline shifts require the presence
of other pollutants in the ambient air.
Chronic human exposures to ambient air appear to produce a
functional adaptation that persists for at least a few months after
the end of the 03 season, but which dissipates by the spring.
Several population-based studies of lung function indicate that
there may be an accelerated loss of lung function associated with
living in communities with persistently elevated ambient 03, but
the limited ability to accurately assign exposure classifications
of the various populations in these studies makes a cautious
assessment of these provocative data prudent.
The plausibility of accelerated aging of the human lung due to
chronic 03 exposure is greatly enhanced by the results of a series
of chronic animal exposure studies in rats and monkeys. There is
little reason to expect humans to be less sensitive than rats or
monkeys. On the contrary, humans have a greater dosage delivered
to the respiratory acinus than do rats for the same exposures.
Another factor is that the rat and monkey exposures were to
confined animals with little opportunity for heavy exercise. Thus
humans who are active outdoors during the warmer months may have
greater effective 03 exposures than the test animals. Finally,
humans are exposed to O3 in ambient mixtures. The potentiation of
the characteristic 03 responses by other ambient air constituents
seen in the short-term exposure studies in humans and animals may
also contribute toward the accumulation of chronic lung damage from
long term exposures to ambient air containing 03.
8.1.1.4 Riste Characterization
A. Combining Exposure and Toxicity Assessments
1.Individual Risks Individual risks are highly
variable. In terms of acute functional and symptomatic responses
they vary enormously among healthy individuals for reasons that are
currently unknown. Prolonged daily exposures to some healthy
individuals engaged in moderate exercise at concentrations within
115
-------
the current NAAQS produce as much as 40% loss in forced vital
capacity, while others show little, if any response. Since
function decrements greater than 10% are considered adverse
(Lippmann, 1988), and since many millions of people are subject to
such exposures while exercising one or more times each year, there
is a very high, but unquantitated risk of a marginally significant
acute response to large numbers of people.
The concomitant changes in lung reactivity and inflammation
that these widespread exposures also produce are potentially quite
important in terms of an accelerated aging of the lung. However,
the risk of such an effect cannot be quantitated.
2. Population Risks The population risks are the
summation of the individual risks. Since the latter cannot be
quantitated at this time, neither can the former.
B. Descriptions of Risk
1. Absolute risk levels for acute responses have been
calculated in the 1988 03 Staff Paper. However, these risk levels
are undoubtedly too low since they wer^ based largely upon the
results of 1 and 2 hour chamber exposures co 03 in purified air.
This is due to the .greater cumulative outdoor exposures and the
likelihood that outdoor air contains factors that potentiate the
characteristic 03 response.
2. Relative and Marginal Relative and marginal risks
cannot be determined for a ubiquitous pollutant such as 03 There
is no evidence for a threshold exposure for acute response, and no
population which can be considered unexposed.
C. Risk Projections
1. WitlL Current Controls Exposures are not likely to
decline significantly. New motor vehicles emit less hydrocarbons
and N0x than those being scrapped, but the projected increases in
vehicle miles travelled should at least partially balance the
reduced unit rate of emissions.
2. With Enhanced Controls The effects will depend on
the kinds of controls implemented. Further hydrocarbon emission
controls on anthropogenic sources can have only a modest effect of
116
-------
ambient 03, unless there are also major reductions in NOX emissions.
Natural hydrocarbons, combined with uncontrollable small sources
will still combine with NOX to produce 03 at levels that produce
measurable acute responses. On the other hand, tight controls on
tailpipe and power plant stack emissions of NOX could substantially
reduce ambient 03 concentrations.
3. With Relaxed Controls Exposures and effects would
rapidly increase.
117
-------
8.1.2 Radon Case Study
Dr. Arthur Upton
New York University
Dr. Jonathan Samet Dr. Julian Andleman
University of New Mexico University of Pittsburgh
118
-------
8.1.2.1
Introduction
A.
Of the various sources
general population is
exposed, indoor radon
contributes the greater
part of the average dose
(NCRP, 1987) , and is
thought to be the most
important from the
standpoint of risk to
human health (Table
8.1.2.1, Figure
8.1.2.1). The role of
radon in the causation
of lung cancer in
underground miners has
been recognized for
decades. Although iron,
zinc, silver, and
uran ium mines
contain other
potential carcin-
nogens, the high
radon levels in the
air of such mines
have been impli-
icated as the main
cause of the in-
creased rates of
lung cancer in the
miners (NAS/BEIR,
1989) .
of ionizing radiation to which the
internal n%
Terrestrial 8%
Nuclear Mea 2%
Consumer °roa 2
Radon 56%
Figure 8.1.2.1 The percentage
contribution of different sources of
radiation to the average total effective
dose equivalent to members of the U.S.
population (From NCRP, 1987)
Study
Population
Average
emulative
Exposure
(VUO*
U.S. uraniui
•iners
Czech uraniuM
•iners
Ontario uraniui
•iners
Saskatchewan
uraniua Miners
Malaterget iron
•iners
Newfoundland
fluorspar Miners
313
226
40-90
20.2
81.4
382.8
Excess
Relative Risk
(WHO
0.45
0.6
1.92
1.5
0.15-1.3
1.4
3.28
2.6
3.6
1.4
0.9
Reference
Thomas, gj aU, 1985
NAS/BEIR. 1988
Thosns, et al., 1985
Svec, et al., 1988
Nuller, 1985
MAS/BEIR, 1988
Howe, et at.. 1986
MAS/BEIR. 1988
Radford and Renard,1984
NAS/BEIR, 1988
Morrison, ej fit-. 1988
1 WIN -
3.4 x
1 UL for 170 Mr - 2 x 10
0"* J Hr m~*
-5 , -3
x 170 Hr
In •iners. exposure to radon decay products are expressed
in units of working levels (WC), which are Measures of th«
concentration of decay products in air recorded in working
level Months (ULM), one ULN representing exposure to an air
concentration of 1 UL for a working Month of 170 hours.
Subsequently reported as 834 wla.
While the
average levels of
radon in the air
inside buildings
tend to be only a TaJ)le 8sl>2-1 Mortality from lung cancer in
fraction of the major cohorts of underground miners (from
current puskin and Nelson, 1989)
levels in
119
-------
underground mines, the epidemiological data on miners imply that
radon may pose some risk of lung cancer even at the low levels
customarily encountered in private houses (NAS/BEIR, 1988) .
Moreover, many homes have been identified with concentrations
comparable to those in mines where workers have been found to be at
increased risk of lung cancer. Thus, the recent recognition that
radon is present in all homes and at unacceptably high
concentrations in many houses and other buildings has prompted
concern about the health hazard that radon may pose to the public
(NCRP, 1984a, 1984b).
B. History of Regulation/Guidelines
Radiation protection guidelines for radon, established
initially to prevent the excessive occupational exposure of
underground miners, were extended to the general U.S. population in
1984, when the National Council on Radiation Protection and
Measurements (NCRP, 1984a) recommended that the annual exposure of
members of the public not exceed 2 WLM per year6. This was
followed by the recommendation from EPA, in 1986, that the average
annual concentration of radon in the indoor air of houses not
exceed 4 pCi I"1 (EPA, 1986), and by the recommendation from ICRP,
in 1987, that the concentration of radon in the indoor air of new
and existing houses not exceed 7 and 14 Pci I"1 (100 and 200 Bq in"3
EEC), respectively (ICRP, 1987). In parallel with these
developments at the national and international levels, similar
attempts to limit exposure to radon have been made by agencies at
the State and local levels (e.g., Reilly, 1988; Nichols and
Stearns, 1988; Roessler, 1988).
8.1.2.2 Current Knowledge Of Exposures
The radiation dose from radon is delivered by short-lived,
alpha-emitting decay products, a large fraction of which is
attached to the inhaled background aerosol. Both attached and
unattached decay products deposit in the respiratory tract. The
resulting radiation dose, delivered to critical sites along the
lining of the respiratory tract, is highest in the bronchial
airways, the sites at which most lung cancers arise.
6 1 WLM = 1 WL for 170 Hr = 2 x 10"5 J m'3 x 170 Hr -
3.4 x 10'3 J Hr m"3.
120
-------
Exposure of the U.S. population to radon first became a matter
of public concern in Grand Junction, Colorado, where uranium
milling wastes containing radium were used as fill. Concern later
developed in areas of Florida and Montana where phosphate rock was
mined. Subsequently, surveys in other parts of the country made it
evident that homes in many areas contained elevated levels of
naturally occurring radon. Although the Reading Prong area in
Pennsylvania, New Jersey, and New York has received special
attention, many other areas have a significant proportion of homes
in excess of the 4 Pci I"1 action guideline recommended by EPA.
Radon is the immediate decay product of radium, that is
present at low concentrations (40 Bq kg"1; 1 Pci g'1) in most soils
and rocks. The average rate of release of radon from the soil—
about 0.2 Bq m'2 (0.5 Pci m"2) per second—can be calculated to cause
an average concentration of radon in the overlying outdoor air of
about 8 Bq m"3 (0.2 pCi I"1). Radon, with a half-life of 3.8 days,
is released from soil near the ground surface and is dispersed
upward by convection. Under inversion conditions, however, the
upward dispersion of radon is limited, so that most locations show
concentrations rising at night and falling in the morning.
Seasonal cycles also occur, depending on location, freezing of the
ground, rainfall, and other factors. The radon decay products are
at about 70 percent of equilibrium outdoors, the unattached
fraction being somewhat below 10 percent of the total radon
daughter concentration. With this degree of equilibrium, the
estimated average outdoor concentration (8 Bq m"3, or 0.2 pCi I"1)
corresponds to a WL of about 0.001 and thus an annual exposure of
0.05 WL7 for anyone remaining outdoors all of the time.
Radon released into an enclosed space, as in a mine or
building, cannot disperse into the atmosphere and therefore
gradually increases in concentration. The major source of radon in
the air inside a building and the soil beneath and adjacent to the
building, although release from the water supply may also be
significant in.some locations. The observed values of indoor radon
show a log-normal distribution (Tables 8.1.2.2 and 8.1.2.3), the
numbers of buildings with concentrations 10-100 times the average
value being disproportionately large compared with the numbers
expected from a normal distribution (NCRP, 1984b), and the
percentages of buildings with high concentrations varying among
(2 x 10'4 Jh m'3)
121
-------
surveys (Table 8.1.2.4)
The highest indoor
radon concentrations
have been measured in
the basements of
single-family houses,
concentrations on
higher floors de-
creasing somewhat. The
concentrations in high-
rise apartments and
public buildings have
generally been much
lower, largely because
of their greater
ventilation and more
substantial foun-
dations, and physical
separation from
basement air.
As discussed by
Robkin (1987), radon
concentrations in the
air of homes in the
U.S. have been measured
widely. Geometric mean
air concentrations for
single-family homes are
about 1 Pci/L. This
has been estimated to
be about equivalent to
0.005 working levels
(WL) . However, some
homes have been found
to have concentrations
greater than or equal
to one WL.
Radon Level Portion of
X Houses
tPcf/L) Above X*
0 1.0 10°.
1 4.6 10";
2 2.2 10";
4 7.4 10";
10 9.7 10"*
20 1.3 10"|
50 4.8 10"*
100 2.4 10"*
* Based on log-noreal
estieated by Hero,
(CM * 0.9 Pci/L. OS
Average Radon
Level in Houses
Above X* (Pci/t)
1.5
2.7
4.2
7.0
15
28
65
130
distribution of ra
Si el.. 1986.
> « 2.8).
Percent of Risk
Associated tiith
Houses Above X*
100
82
60
33
9
2
0.2
0.01
don levels
Table 8.1.2.2 Distribution of houses and
radon-induced lung cancer risk with
respect to radon concentration
(Distribution I, Puskin and Nelson, 1989)
Radon Level Portion of Average Radon Percent of Risk
X Houses Level in Houses Associated with
(Pci/L). Above X* Above X* (Pci/L) Houses Above X*
0
1
2
4
10
20
50
100
1.0 10°.
4.6 10"]
2.5 10"!
i.o 10";
1.9 10";
3.8 10"f
2.8 10"*
2.6 10~5
* Based on log-nonst
estimated by Hero
1.8
3.3
4.9
8.0
17
31
70
130
il distribution of
. ej si.. 1986.
100
86
68
44
6
0.9
0.1
radon levels
(CM * 0.9 Pci/L. GSO * 3.2).
Table 8.1.2.3 Distribution of houses and
radon-induced lung cancer risk with
respect to radon concentration
(Distribution II, Puskin and Nelson, 1989)
Average Radon Concentration Percent-Greater Than
BqV3 fDCi 1 » «0 Bd •"* (4 uCf 1"1)
HOP (19845)
Hero et fll., (1986)
Alter I Oswald (1987)
Cohen (1988)
37
55
260
120
(1.0)
(1-5)
(7.1)
(3.3)
3
7
23
19
Table 8.1.2.4 Reported distribution of
radon in U.S. living areas.
Generally surveys show that the concentrations are distributed
approximately log-normally (see Fig. 8.1.2.2 from the Robkin
article). Such distributions from surveys in the eastern part of
122
-------
Percemt of nouses
the U.S. show that the
means and slope of the
distribution curves
can, however, vary
considerably regionally
(George and Hinchliffe,
1987) , as shown in
Figure 8.1.2.3. Even
in a given community
indoor air concen-
trations vary, as do
concentrations in Figure 8.1.2.2 Distribution of radon 222
different parts of the concentrations in single-family homes for
552 sites (after Robfcin, 1987, based on
4-w • r,- Nerof 1984)
(see their Figure
232 fin Concentration
and by
their
8.1.2.4) .
One would expect
that air-infiltration
rates would have a
substantial effect on
indoor-air concen-
trations of radon.
However, because source
rates and other
controlling factors
operate as well, the
effect of ventilation
rates may not be great,
as shown by Nero et al.
(1983). Figures
8.1.2.5 and 8.1.2.6
show a wide scatter of
radon concentration
among homes when
plotted against
ventilation rate.
However, when a
frequency distribution
plot is made of the
product of air
concentrations and
ventilation rates
Radon Concentration pCi/L
28
38 46 66 66 76
% Home* < Olv«n Concentration
86
Living ArM-Sumiwr
Bn«»iii.nt-Sum.r
Living ATM-Winter
Figure 8.1.2.3 Distribution of radon
concentration in residential buildings in
Morris County, New Jersey (after George
nd Hinehliffe. 19871
noon concentration pCI
20
30
40 60 60 70 80
% Homea < Qlven Concentration
90
TOO
Lang Idtnd. N.Y.
Albiny. N.Y.
-*- Savannah River. 8.C.
*«** N. V» -O.C.-MO
Luatffw County, PA
Figure 8.1.2.4 Distribution of radon
concentrations in living areas during
winter in different geographical locations
(after George & Hinchliffe/ 1987)
123
-------
Radon InCI/cubie i
0.01
0.1 1
Air change rate (hr/10)
Hooon 223 conctmrilMM (
0.01
0.1 1
Air change rate (hr/10)
Figure 8.1.2.5 Radon-222 Figure 8.1.2.6 Radon-222
concentrations vs. ventilation concentrations vs ventilation
rates in 17 "energy-efficient" rates in 29 houses in the San
houses (after Nero/ 1983) Francisco area (after Nero,
(Figure 8.1.2.7), there seems to be 1983)
a distribution around a mean. Such variations could result from
differences in the nature of the sources from house to house, and
from differences in design and construction, as well as temporal
fluctuations in the source.
28
20
Pwewitag* of hout«*
0.01 0.06 0,076 10 0.2 0.8 0.7S 1.0 2 8 10 W.I
Radon «ourc« magnitude
Radon has been surveyed
in groundwaters of the U.S.
(Longtin, 1988). Table
8.1.2.5 summarizes the
population-weighted averages
for radon concentrations
(Pci/L) in various states.
This displays data from two
sources, one an EPA National
Inorganics and Radionuclides
Survey (NIRS) of 1000 U.S.
public groundwater supply
systems randomly selected from Figure 8.1.2.7 Frequency
four population categories, distribution of radon source
The two surveys indicate that magnitudes calculated from the data
the U.S. state-average jn Figs. 8.1.2.5 and 8.1.2.6 by
*. ..- ,f~~ nnn * • ,r, taking the product of ^Rn
concentrations (600-800 Pci/L) concentrations and ventilation rate
for sites with populations (After Nero/ 1983)
<1000 were higher than those for the sites >1000 (about 200 Pci/L) .
There was a large range in average concen- trations among the
states. As shown in Figure 8.1.2.8, a very small number of
supplies have radon concentrations greater than 10,000 Pci/L in
water, and about 80% have less than 500 Pci/L.
124
-------
1 Sites with <1.000 Peoole
SJAJE
Alabasa
Alaska
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
Florida
Georgia
Hawaii
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New Tork
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vemont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
Puerto Rico
US average
•Based on data
Cothern*
160 (40)+
100 (47)
120 (44)
75 (51)
500 (18)
380 (23)
1,500 (3)
100 (48)
1.000 (9)
1,100 (6)
50 (52)
256 (30)
100 (49)
105 (45)
250 (31)
250 (32)
250 (33)
180 (39)
10,000 (1)
700 (15)
1,500 (4)
105 (46)
210 (36)
150 (4)
300 (24)
500 (19)
300 (25)
550 (17)
1,400 (5)
150 (42)
200 (37)
500 (20)
1,100 (7)
300 (26)
200 (38)
250 (34)
300 (27)
1,000 (10)
3,400 (2)
1,100 (8)
300 (28)
100 (50)
150 (43)
500 (21)
250 (35)
700 (16)
300 (29)
1,000 (11)
750 (14)
880 (12)
500 (22)
780 (13)
MIES
2,025 (5)
129 (44)
1,302 (7)
75 (50)
538 (18)
336 (29)
3.328 (1)
116 (48)
393 (25)
419 (24)
431 (22)
136 (40)
136 (41)
166 (35)
365 (27)
148 (39)
116 (49)
1,228 (9)
2,161 (4)
253 (33)
370 (26)
342 (28)
133 (32)
125 (46)
535 (19)
291 (31)
743 (12)
2.674 (3)
737 (13)
423 (23)
647 (14)
2,876 (2)
125 (47)
164 (36)
164 (37)
130 (43)
467 (20)
1,170 (10)
1,260 (8)
334 (30)
128 (45)
264 (32)
157 (38)
1,533 (6)
952 (11)
238 (34)
459 (21)
540 (17)
558 (16)
602 (15)
Sites With >1.000 Peoole
Cothern
160 (35)
100 (47)
320 (17)
100 (42)
500 (10)
380 (14)
770 (4)
126 (42)
148 (40)
150 (37)
50 (51)
256 (25)
167 (34)
105 (45)
200 (29)
106 (44)
10 (43)
180 (31)
2,000 (1)
450 (11)
770 (5)
105 (46)
210 (28)
82 (49)
100 (48)
328 (16)
290 (19)
550 (9)
1,183 (2)
300 (18)
180 (32)
132 (41)
278 (21)
150 (38)
169 (33)
160 (36)
264 (23)
720 (6)
1,151 (3)
276 (22)
290 (20)
24 (52)
150 (39)
360 (15)
656 (8)
450 (12)
264 (24)
720 (7)
234 (27)
415 (13)
240 (26)
Nigs,
171 (26)
1,610 (1)
161 (28)
317 (12)
646 (2)
126 (33)
118 (35)
583 (4)
438 (9)
198 (20)
195 (22)
130 (32)
370 (11)
220 (19)
107 (41)
112 (36)
596 (3)
164 (27)
397 (10)
100 (43)
148 (30)
112 (37)
444 (8)
125 (34)
250 (16)
173 (25)
100 (44)
109 (40)
177 (24)
158 (29)
112 (38)
535 (5)
196 (21)
273 (15)
112 (39)
138 (31)
238 (18)
497 (7)
313 (13)
520 (6)
240 (17)
300 (14)
200 (30)
194 (23)
of Hess et al.
•Hluabers in parentheses are
relative rankings.
Table 8.1.2.5 Population-weighted averages for
radon activity (Pci/L) (After Longtin, 1988)
Prichard and Gesell
(1981) have es-
timated population
exposures to radon
volatilized indoors
from water. They
estimated that the
average radon
indoor air con-
centration em-
anating from 1000
Pci/L in water
might vary from
0.01 to 0.1 Pci/L
in air, depending
on the nature of
the dwe 11 i ng. The
water use-weighted
volatilization rate
of radon from water
is typically 50%.
Others have es-
timated that the
ratio of the air-
to-water con-
centrations in U.S.
homes, CA/CW, would
be typically 10"4,
consistent with the
high estimate of
Prichard and
Gesell.
Andelman and co-
workers have
measured the
volatilization of
chemicals from
indoor uses of
water, and have
shown that these
inhalation
exposures from
125
-------
Occurrence-Percent
HES3
NIRS
o-wo wo-aoo (oo-iooo KJOO-MOO LOOO-KJOOO* >K>OOO
32.4
28.3
42.6
48.8
10.1
11.7
11.8
8.9
2.2
1.2
2.2
1
Activity-pCi/L
I HESS
iNIRS
showers, baths and other water
uses is at least comparable to
those from the direct
ingestion of water (Andelman,
1985). In a recent assessment
of such exposures it was
judged that a whole house
inhalation exposure for an
adult spending 24 hours in a
home would be given by
E = (0.2 to 10) Cu
Figure 8.1.2.8 Occurrence of radon
where E is the inhalation in drinking water
exposure (in the case of radon, Pci/day) and Cy the concentration
in water, Pci/L (Andelman, 1990). Thus, for example, if the water
supply concentration contains 1000 Pci/L, this would constitute a
24-hr predicted inhalation exposure expected to range from 200 to
10,000 Pci. For comparison, a typical U.S. indoor-air
concentration of 1 pCi/L would lead to an inhalation exposure of
about 20,000 pCi/day. It should be emphasized that "inhalation
exposure" refers to the quantity of radon inhaled.
It has also been shown by Andelman (1990) , that the inhalation
exposure from a shower alone is substantial, such that E is about
equal to Cu. A shower using water containing 1000 pCi/L radon
would lead to an inhalation exposure of about 1000 pCi. This is in
addition to the exposure in the home from all water uses.
One can conclude that radon is ubiquitous in U.S. homes, the
concentrations vary considerably regionally, locally, seasonably,
and temporally. Also ventilation and other individual home
characteristics, and location within the home will affect the
concentrations. Water as a source can add to the exposures, and
localized point source exposures, such as showering, can be
important.
Although the overall risks on a national basis may be
estimated from national survey data, the variability of exposure
and, therefore, risk can be expected to be substantial.
The average exposure of members of the U.S. population has
126
-------
been estimated to range from 0.2 WLM per year (NCRP, 1984b) to 0.25
WLM per year (Puskin and Nelson, 1989), or possibly higher (NCRP,
1987) . The many radon measurements that have been made are of only
limited value for exposure estimation, however, since they have
been designed primarily to determine the maximum potential
concentrations of radon in houses rather than the actual levels to
which occupants are exposed. The commercial measurements are also
biased by the fact that the customers requesting them usually have
had reasons to suspect high concentrations of radon in their homes
(Cohen, 1988). Hence there is a need for a statistically
stratified program of radon sampling to estimate the average level
of exposure in the United States (NCRP, 1984b, 1989) . EPA is
currently implementing such a survey. Without a more accurate
estimate of the average exposure of the U.S. population, a precise
assessment of the magnitude of the health risks from radon is not
possible.
8.1.2.3 Toxicitv And Health Effects
A. Human Epidemiology The major studies of underground
miners reported thus far are listed in Table 8.1.2.1; however, a
number of problems complicate the interpretation of these studies.
First, the exposures of the miners were documented to a varying
extent and the estimates are subject to misclassification.
Exposures were not measured at all for many of the early miners.
Second, the contribution of smoking to the observed excess of lung
cancer is difficult to evaluate, especially since smoking histories
of the miners were not available in most of the studies. Third,
selection of an appropriate control population is subject to
uncertainty, although internal analyses are most appropriate for
estimating exposure-response relationships.
Many epidemiological studies are under way in the general
population to estimate directly the risk of indoor radon, they are
also subject to limitations from exposure misclassifications,
inadequate sample size, and the possible confounding effects of
extraneous risk factors. Because the general population has had
lower levels of exposure than the miners, and, consequently,
smaller effects are anticipated, the statistical power of the
studies may be inadequate for the detection of small effects. At
present, therefore, estimates of the risks to the general
127
-------
population from exposure to radon have been based on extrapolation
from the data on miners.
From the lung cancer mortality reported in various cohorts of
miners, the exposure-response relationship for lung cancer appears
to be linear in the low-to-intermediate dose range. On the basis
of this epidemiological evidence, supporting animal studies, and
biological considerations, the frequency of lung cancer is assumed
to increase linearly with exposure below 50 WLM. To assess the
total magnitude of the radon risk, however, it is necessary to
predict the lifetime lung cancer mortality in the various mining
populations, many members of which still survive. For this
purpose, neither the simple absolute risk model (which predicts a
constant additional risk of death per year following a given
exposure) nor the simple relative risk model (which predicts a
constant percentage increase in the annual age-dependent baseline
risk following a given exposure) adequately describe the observed
patterns of mortality. Instead, either a modified absolute risk
model, in which the
risk is reduced with
time after exposure
(NCRP, 1984a), or a
modified relative risk
model in which the risk
varies as a function of
age and time after
exposure, (NAS/BEIR,
1988), would seem
preferable (Table
8.1.2.6). A model of
the latter type has
Source of Estimate Lifetime Risk(X) Projection Model
MCRP (19846) 0.9 Modified absolute risk
ICRP (1987) 1.6 Constant relative risk
1.1 Absolute risk
BEIR IV (HAS, 1988} 3.4
1.4
EPA (1989)* 2.0
Modified relative risk
Relative risk
• Puskin and Nelson (1989)
Table 8.1.2.6 Estimated lifetime risk of
been adopted by EPA for lung cancer attributable to 0.02 WL (4 pci
its radon risk i'1) exposure to radon/ assuming the short
assessment (Puskin and half-life decay products are in 50%
Nelson, 1989).
and half-life decay products
equilibrium with the radon
The use of risk models for estimating risks to the general
population from the data on miners involves additional
uncertainties owing to differences in age-and sex-distribution, and
potential differences between continuous exposure over a lifetime
and short-term occupational exposure during working-hours only.
Other uncertainties complicating the assessment relate to
estimation of the actual dose delivered to the lung, owing to
128
-------
differences in breathing rate and to differences in aerosol
particle size, degree of radioactive equilibrium of the decay
products in the atmosphere, and other variables (NCRP, 1984a;
Harley and Cohen, 1987). Also uncertain is the form of the
interaction between the effects of smoking and those of radon;
assessment of this interaction is possible in only a few studies.
The strongest evidence is available from the study of Colorado
plateau uranium miners, that suggests a somewhat less than
multiplicatve interaction (NAS,1988). If the multiplicative
interaction model is correct (e.g., NAS, 1988), the absolute
lifetime risk for a given level of radon exposure would be 6-10
times higher in smokers than in non-smokers.
The apparent decrease in risk with time after cessation of
radon exposure has not been precisely established. Since lung
cancer is rare before the age of 40, exposure during childhood may
possibly con- tribute little to the subsequent risk of the disease
(BEIR, 1990); however the ICRP (1987) has considered risks to be
greater for exposure during childhood.
B. Animal Toxicology Radon and radon decay products have
been shown to increase the incidence of benign and malignant tumors
of the respiratory tract in rats exposed to these radionuclides by
chronic inhalation (Cross et al., 1982; Chameaud et al., 1984), the
magnitude of the increase varying, depending on the dose and on
the influence of other factors, such as inhalation of dusts or
cigarette smoke (Table 8.1.2.7). The lifetime risk of lung cancer
has been calculated from such experiments to approximate 1-5 10"4
WLM"1 (Bair, 1986; Cross, 1988).
8.1.2.4 Risk Characterization
The average level of exposure to radon in members of the U.S.
population has not been characterized in a large nationwide survey.
However, data from diverse sources suggest a mean concentration in
U.S. homes of about 1.5 pCi I'1. If annual exposure is assumed to
approximate 0.25 WLM per year (Puskin and Nelson, 1989), as noted
above, the lifetime risk of mortality from lung cancer can be
calculated with the use of the risk models cited (Table 8.1.2.8,
Figure 8.1.2.9). With the use of such models, the lifetime risk of
lung cancer from exposure to radon in the U.S. population can be
estimated to range from roughly 0.4 to 1.8 percent. By the same
token, exposure to radon can be estimated to account for some
129
-------
factor
Radon-daughter
cumulative exposure
Radon-daughter
exposure rate
Radon-daughter
unattached fraction
Radon-daughter
disequilibrium
Concomitant exposure
to cigarette smoke
Increases approximately linearly with exposure
Increases with decreete in exposure rate
(approximately 200 to 400X increase from about
500 to 50 WIM/waek. The 500-, the 50-,and the
5-ULM/week data are not significantly different
at approximately 300-MJI exposures.
Increeses with increase in unattached fractior
(approximately 50X increase per WIN exposure
from 2 to 1«f0)D
Increeses with increase in disequilibrium
(approximately 30X increase per WLN exposure
(borderline significance) from 0.4 to 0.1F)C
Decreases if smoking alternates on same
day with radon-daughter exposure. Increases if
smoking following cumulative radon-daughter
exposures. Mo effect if smoking precedes
cumulative radon-daughter exposures
Data pertain to raw tumor-incidence date uncorrected for time
related factors and life span differences from control animals.
b f is the percentage of 218Po that is unattached. When ex-
pressed as percentage of radon concentrations, they are 1.3
and 5.2X, respectively.
c Equilibrium factor (F) is the ratio of the non-equilibrium
concentration of short-lived daughters in air to the equilibrium
equivalent concentration.
Study
Cancer Deaths/10"
um'1
UMSCEAR, 1977
8EIR III, 1980
MOtP, 1984
ICRP, 1987
KIR IV, 1988
EPA, 1989
200-450
750
130
170-230*
360"
350
360C
* Relative risk with ICRP Ref.
. population.
0 Relative risk with BEIR IV U.S
Ref population.
c Based on average of BEIR IV ant
IRCP 50 risk models (confident
interval of 140-730)
RK approximately constant wit)
age at exposure (but decreesec
with time after exposure)
Table 8.1.2.7 Salient factors influencing
the tumorigenic potential of radon-daughter
exposures in rats (from Cross, 1988)
Table 8.1.2.8 Life
time risk, estimates
for lung cancer due to
lifetime exposures to
radon
Relative frict. of tot. I.e. fr«q (%)
100 c
.01 .ot o« 1 t
Annual Rn-daughter exp. at home (WLM)
MM* conlrlMilMn Iron Mdeof*. 11 neuM
Figure 8.1.2.9 Expected relative percentage
of the total lung cancer attributable to
indoor exposure to Rn daughters, as a func-
tion of the mean level of Rn daughters in
indoor air at home (after Jacobi, 1986)
130
-------
5,000-40,000 deaths from lung cancer each year in the U.S., or
about 4-30 per cent of all lung cancer deaths in the U.S.
population (Puskin and Nelson, 1989).
The above estimates strongly suggest that radon exposure
presents a significant public health problem. The uncertainties in
the exposure levels and in the risk estimates are large,
however, and vigorous efforts to refine the levels and the risk
estimates are needed.
These analyses illustrate that risk assessment techniques can
be used, even when definitive data are not available, to estimate
the extent of an environmental disease risk. In the case of radon,
a number of uncertainties affect the projected risk. The range of
these uncertainties can be specified however; most analyses
indicate that extrapolation from the studies of miners to the
indoor environment introduces only a relatively small degree of
uncertainty, ranging up to 30 percent. Thus, even in the face of
uncertainty, radon must be considered an important public health
problem. The use of risk assessment can provide an indication of
harm (numbers of cancer deaths), which is useful for a ranking
process.
131
-------
8.2 RanXincr Schemes; Detailed derivation of RanK-merainq
Dr. Paul Deisler
University of Houston
132
-------
8.2.1 Formulation of the Basic Model
Considering the total possible set of endpoints (both cancer
and non-cancer) that may be caused by agents in the environment, E
in number, and the total set of agents in the total environment
(specific substances and types of radiation) , A in number, that may
cause, individually, anywhere from none to many of the endpoints,
the weight to be accorded the jth of the Problems (thirty-one in
number in the case of the UB report) in ranking that Problem in
comparison to the others, based on population risk, is proportional
to W , where
E A
Wj = N X Z S, fjjk (1)
i=l k=l
W. may also be written in the form,
E A
Wj = Nj Z S S, Djk Pik(Djk) (2)
i=l k=l
In these equations N, is the number of individuals comprising the
population relevant to the jth Problem, S is severity, P is
potency, f is the fraction of the population that exhibits an
endpoint (the response to the exposure to an agent) at exposure D,
and the subscripts i and k designate, respectively, the ith
endpoint and the kth agent. E, to reiterate, includes all possible
endpoints that might be considered, caused by whichever agents, and
A includes any and all agents (not just those known to be
associated with the j-th Problem) present in any way and in any and
all parts of the environment. In the case of endpoints that
respond proportionally to exposure/dose (that are said to be
"linear" in dose), P is independent of D; in the case of endpoints
whose responses are curvilinear or that exhibit thresholds or
threshold-like behavior, P is a function of D (as shown here in the
general case) . Thus, ffj.k is the fraction of the population
relevant to the jth Problem affected by the ith endpoint if caused,
in turn, by the kth agent; the product of Nj and fjjk is the number
of individuals affected by that (ith) endpoint as caused by the one
(kth) agent, and is thus a measure of the excess population risk of
that endpoint from that agent. If the kth agent does not exist in
the jth Problem, or if it does not cause the ith endpoint, or both,
133
-------
then frk, corresponding to that particular agent and/or endpoint,
is zero. The same endpoint may also be caused by other agents and
the same agent may cause other endpoints. Multiplying by S( weights
the population risk according to severity, for the ith endpoint,
and the summations over i and k give the weighted sum of the excess
risks of all endpoints of every kind for the jth Problem, VT.
Equation (1) is for cases when fjjk can be obtained directly from
epidemiologic information when available at appropriate exposure
levels. Equation (2) is the form of equation (1) necessary when
such direct data may not be adequate and when estimates of dose
response may have to be used (from human or animal data); this is
the more usual case. If one knew all of the factors in either
equation (1) or (2) , then ranking would be easy: the Problem with
the highest value of the weighted sum would be the highest ranking,
and so forth. Independent action by agents is assumed.
As written, the two equations suppose that in those cases
where the same individuals exhibit more than one endpoint the
aggregate severity is the sum of the Sjf..k products, with no special
allowance for the fact that some individuals may exhibit more than
one endpoint. In many instances this is probably a reasonable
assumption; however, there may be instances in which the true
severity of affliction by two endpoints is greater than would be
indicated by the sum and others in which it is less. An example of
the latter case would be if the result of each endpoint is certain
death in about the same time period and under similar
circumstances: two such deaths, for the same individual, are no
more severe than one. Because of the smallness of the values of
the f^ for the usually encountered levels of human exposure and
the smallness of the fraction of individuals involved, out of the
total, with endpoints such as those described, W;. will be only
slightly overestimated using this model. It should also be noted
that S is not considered to be a function of exposure in the first
two equations whereas it may be so in real cases. In the case of
carcinogens, for example, not only does the number of subjects
exhibiting at least one tumor increase with exposure but so,
usually, does the number tumors per individual, on average, a
factor that may be deemed, in different instances, to impact S8.
While the above equations represent real simplifications of
the actual situation (exposures, for example, are not represented
8If S^ is a function of exposure, equation (3), derived later,
still applies and the merging method proposed is still valid.
134
-------
in actual fact, for any one agent, by a single value, D,
independent of time, place, individual circumstance, or other
factors), they contain the main variables of importance in an
appropriate relationship and they would provide, from the outset,
given reasonable estimates of the levels of the variables,
consistent, merged rankings for both cancer and non-cancer risks
taken as a spectrum of health risks. Thus, if all of the
quantitative values of the variables in either equation (1) or
equation (2) were available, developing a merged health risk
ranking would be a simple matter of calculation; indeed, the
ranking, itself, would be directly quantitative and not merely a
listing of rank order.
One of the key missing sets of variables for producing a
single, health risk based ranking of Problems is a single set of
severities for cancer and non-cancer endpoints together. The
experience already gained in attempting to grade the severities of
different non-cancer endpoints in the UB report should help in the
formulation of a method and a process for undertaking the task of
producing a consensus on a health risk severity table including
both cancer and non-cancer effects, and it is recommended that any
updating of the UB report include this activity.
It is highly desirable to utilize the above two equations and
the operations they depict to the maximum extent possible when
developing a merged health risk ranking procedure because of their
scientific basis and the mutual consistency across the different
kinds of endpoints that they therefore automatically provide.
Although the same lack of information that prevented a more
rigorous approach in the UB report prevents the straightforward
utilization of the above equations, any approach should approximate
as nearly as possible the above equations so as to provide the best
basis for merging the rankings.
8.2.2 Merging Separately Established rankings; General
Discussion
The "merging of separate rankings" procedure depends, as
discussed below, on certain characteristics of three-by-three grid
arrays (see Figures 8.2.2.1 and 8.2.2.2) when combined with an
algebraic expression described below that, in turn, is based on
equation (1) . These characteristics lead to the conclusion that
there are only a finite number of ranking patterns that need to be
considered in the merging process, a fact that reduces the problem
135
-------
of selecting sets of Problems for consideration of their combined
risks. The same is true of larger arrays but, while four-by-four
or larger arrays of grids might be used, for example, the procedure
rapidly becomes cumbersome because of the increase in the number of
ranking patterns that must be considered as the order of the grid
array used increases.
Figure 8.2.2,1 Projecting a grid square—linear array
The use of a three-by-three grid array means that the rankings
of Problems for cancer and non-cancer risks must first be grouped
into three qualitative risk categories: high (H), medium (M) and
low (L) . Each of these levels may be thought of as bounded by
quantitative risk values, h, m, and 1, as shown in Figure 8.2.2.1,
136
-------
where h > m > 1. Thus, a Problem judged to be of high risk for
cancer, and so categorized as H, would have a quantitative risk
value lying between m and h on the cancer axis, if its risk could
be quantified in some manner; and similarly for Problems
categorized as H, M or L, on either axis. Plotted, each Problem
Figure 8.2.2.2 Projecting a grid square—nonlinear array
would appear as a point within an appropriate grid square; thus a
Problem categorized as H for cancer and M for non-cancer would fall
somewhere within grid square B in Figure 8.2.2.1 or 8.2.2.2. As
shown in the Figures, each of the grid squares is labeled A through
I, for identification, and one of nine nodes (denoted by circles)
is associated with each of the grid squares by being given the same
letter designation as its corresponding grid square.
137
-------
With three risk categories for each of the two sets of
rankings, nine pairs of categories, nine grid squares, and up to
nine risk levels, are possible for merging the rankings of those
Problems ranked for both cancer and non-cancer risks as seen in
Figures 8.2.2.1 and 8.2.2.2. In this section, the type of array in
Figure 8.2.2.1 in which h-m = m-lis called a linear array.
Depending on how the individual risk factors are taken into account
in the separate rankings by cancer and non-cancer risks, the actual
array of risks may be linear or it may be nonlinear; in this
section the only type of nonlinear array to be considered is the
one in which h - m > m - 1 and in which the array is symmetrical
around the diagonal (see Figure 8.2.2.2); linear arrays, by nature,
are symmetrical about the diagonal.
Under whatever system is employed, the merged ranking of the
Problems lying within grid square A (or (H,H)) and of those within
grid square I (or (L,L)) is clear enough: grid square A contains
the Problems of the highest merged risks and grid square I contains
those of the lowest; moreover, grid square E (or (M,M)), and its
Problems, falls unambiguously between them. Geometrically, as seen
in the Figures, these three grid squares are rank ordered as they
fall along the diagonal, A > E > I; the question is, then, how to
project the off-diagonal grid squares, and their corresponding
problems, such as grid square D (or (M,H)), onto the diagonal so as
to know where they fit in the resulting ranking against the three
grid squares already athwart (or, "on") the diagonal. This is best
seen by considering the projection of a single, off-diagonal
Problem (or point) onto the diagonal.
8.2.3 The Principle of Projection onto the Diagonal
An equation of the following form may be derived, starting
from equation (1) (see Section 8.2.11):
Wj = wcj + vwNj (3)
M O V
1 NH ''MM rNH
where v = WNH/WCH = (4)
\J c P
"CH °CH rCH
In equation (3) wcj. is the weight for ranking purposes of the jth
Problem that may fall into one of the three categories, H, M or L,
for cancer, and WMJ. is the same, separately, for non-cancer; and all
138
-------
weights are scaled so that the highest quantitative level is h on
each axis. In other words, the two weights represent the
quantitative risk rankings, for cancer risk only, on the one hand,
and for non-cancer risk only, on the other. The two weights are
therefore the coordinates of points plotted within one or another
of the grid squares and the sum, wjt, is just the merged ranking
score, for the point in question. AS shown in Section 8.2.11, the
coefficient v takes into account the differences in number exposed,
the fractions of those exposed who suffer harm (potencies of and
exposures to agents) and their relative <~2verities, cancer versus
non-cancer; it represents the weight gi/an to non-cancer versus
cancer risks. In equation (4) , WMH and WCH are the weights (see
equation 7) for cancer risk and non-cancer risk, respectively, of
the problems having the highest such weights without respect to j
(that is, the two weights need not correspond to the same Problem)
and the S and F values are the mean values of severity and of the
fraction of the relevant population affected corresponding to these
same highest weights; the N values are the numbers exposed, also
corresponding to the same highest weights. Note that v is a
constant for the ranking of a particular set of problems; for a
different set (or subset) in which the highest weights correspond
to different Problems and are therefore likely to be different, a
different value of v is likely to obtain.
The way in which equation (3) governs the slopes of the
projection vectors, and the sets of possible rankings that can
result, is described in more detail below. A brief description is
given here for convenience.
Referring to Figure 8.2.2.1, if a point on the diagonal, with
coordinates (x,x), is the projection of an off-diagonal point, with
coordinates (y,z), that means that the value of w at (x,x) is equal
to the value at (y,z). That is, by equation (3),
(1 + v)x = y + vz (5)
from which the value of v required to yield the projection of (y,z)
onto the diagonal at (x,x) is obtained in terms of x, y and z.
Conversely, given a value of v, the projection of any point or node
onto the diagonal is known, where points or nodes already on the
diagonal are their own projections. The order in which such
projected points or nodes appear on the diagonal is therefore their
rank order in terms of the combination of both cancer and non-
cancer risks.
139
-------
It is shown below that the slopes of the projection vectors
are all the same for projections from the off-diagonal points or
nodes onto the diagonal for the value of v pertaining to the
particular set of ranked groupings, whatever that value of v may
be. Moreover, the slopes are all equal to -1/v; that is, the
slopes are all negative (since v is not less than zero), and the
projection vectors, for a given v, are all parallel. Thus,
continuing with our example in Figure 8.2.2.1, any off-diagonal
point within any grid square will project onto the diagonal along
a vector parallel to that joining (y,z) and (x,x), so long as (x,x)
is the projection of (y,z), and vectors (1) and (2) thus define the
projections of node G and of the vertex diagonally opposite to node
G in grid square G. Moreover, every point or Problem contained
within grid square G lies on a line segment or range on the
diagonal lying between the intersections of vectors (1) and (2)
with the diagonal. Generalizing, the projection of any grid square
is such a line segment or range. and the projections of all grid
squares constitute a set of overlapping ranges, the positions of
which on the diagonal with respect to each other (or their rank
orders), and degrees of overlap, are dependent on the value of v.
In the case of Problems that are qualitatively but not
quantitatively rank ordered, as in the UB report, the coordinates
of the Problems within a particular grid square are not known;
however, the projection of the grid square itself onto the diagonal
gives the range within which those problems must lie, narrowing the
range of comparisons that must be made to arrive at an ultimate
rank ordering of Problems on the basis of total health risk.
Because of the overlaps, some of the Problems within individual
ranges may ultimately be rank ordered oppositely to the rank
ordering of their ranges. How this is accounted for in achieving
the final rank order is described further on.
8.2.4 Derivation of the Possible Ranking Patterns for Grid
scroares and Ranges
Consideration of the projections of all the grid squares onto
the diagonal as was done in the case of grid square G in Figure
8.2.2.1 shows that the rank orders of the resulting ranges,
including those of grid squares lying athwart the diagonal in the
first place, are the same as the rank orders of the projections of
their corresponding nodes (or of any other conveniently defined
point within the grid squares). To derive the possible rank orders
of the ranges for different values of v, it is possible and
140
-------
w
CJ
h
H
m
M
1
L
0
0
L
1 M
CANCER
m
H
Figure 8.2.4.1 Linear array of nodes
convenient to do so by considering those of the nodes. The plot of
nodes only, corresponding to the grid squares and their nodes in
Figure 8.2.2.1, is shown in Figure 8.2.4.1.
In Figure 8.2.4.1, two kinds of vectors can be distinguished:
(1) those that project more than one off-diagonal node onto the
same point on the diagonal (vectors drawn with continuous black
lines are examples of these such as the vectors connecting nodes H
and C, D and C, etc...) and (2) the dashed-line vectors that
project only one node onto the diagonal. Imagining the dashed-line
vectors to rotate around their off-diagonal nodes so as to pass
through points where more than one node is projected (common
projections) , the order (the rank order) of the projections (of the
merged risk rankings) changes: one order occurs when a dashed-line
141
-------
h
H
o
s
<
u
s
o
ni
11
1
L
L 1
m
CANCER
H
h
Figure 8.2.4.2 Nonlinear array of nodes
vector is on one side of a common projection, a reversal of order
occurs on the other side, and an order unique to the common point
occurs at the common point. The same observation pertains to
nonlinear arrays, an example of which is shown in Figure 8.2.4.2.
For linear arrays, it is found that, in addition to the
physically trivial cases where v is equal to either zero or
infinity, there are three values of v that yield common points and
four ranges of v that do not; these yield seven different rankings,
all that are possible for a three-by-three, linear array: three for
v > 1, three for v < 1, and one for v = 1. For a three-by-three
nonlinear array of the type considered in this report regardless of
the values of h, m and 1, again excluding the physically trivial
142
-------
cases, there are at most fifteen possible rankings: seven for v >
1, seven for v < i, and one for v = 1. Two such rankings are
listed in Tables 8.2.6.1 and 8.2.6.2, including the trivial cases
for completeness; only the non-trivial cases are numbered for
reference in each table. Note that for a four-by-four linear
array, the number of rankings that must be considered jumps to
fifteen; hence the practical importance of using the three-by-three
array.
The practical meaning of all of this is that even if it is not
known whether the rankings are linear or nonlinear, what the values
of h, m and 1 are, or what the value of v is, the nurber of
possible rankings of ranges that need to be considered and compared
for consistency with the information available on their separately
ranked Problems is no more than seven for linear three-by-three
arrays and fifteen for nonlinear ones as defined here.
8.2.5 Comparing Range Rankings With Data
If it can be decided whether v > 1 or v < 1 (the most
important considerations) or whether v = 1 (or close to it), then
the number of possible range rankings that must be considered to
obtain a first rough ranking of Problems associated with the
rankings is further reduced. Since there are certain features
among the possible rankings such as specific reversals of ranking
of ranges between pairs of rankings for different values of v, or
cases in which certain ranges are of equal rank, there are
additional ways in which the number of rankings that must be
compared with the information in any detail in any given case can
be reduced; moreover, these types of comparisons yield an answer to
the question of the value of v relative to unity without requiring
any direct attempt to evaluate v.
A final ranking of ranges (each with its contained Problems)
chosen by using the properties of three-by-three arrays, as
governed by equation (3), becomes the basis for further, detailed
comparison of Problems contained within overlapping rankings, using
the information available, to introduce changes in the rankings of
individual Problems if these seem necessary.
The actual comparison of any two ranges to determine their
relative ranking requires that available data or information on the
risks associated with the Problems associated with (contained
within) one of the two ranges be compared with the information on
143
-------
the Problems in the other to determine where the two sets of
Problems, on balance, appear to lie in terms of relative rank
order: one generally above other (recognizing that some individual
Problems may rank differently than their ranges because of
overlap), one generally below the other, or that the two sets are
generally similar in rank (the ranges have roughly equal rank).
There will generally be small (and different) numbers of Problems
associated with any two ranges, a fact that does not make the
comparison easy since the ranges are relatively broad (see, for
example, Figure 8.2.2.1) and there is no way of knowing,
quantitatively, where the Problems lie within ranges. Of some
small help is the fact that the projections of Problems will tend
to be grouped centrally within the ranges rather than uniformly,
even if one supposes the Problems to be drawn from a uniform
distribution of Problems over the area of the grid squares, except
as v becomes either very large or very small. For v equal to zero
or infinity, the distribution of projections of Problems within
ranges will be uniform.
8.2.6 Steps in the Process for Producing a Merged Health Risk
Ranking
Given that the possible ranking patterns of ranges for three-
by-three linear and nonlinear arrays are now established (Tables
8.2.6.1 and 8.2.6.2), the following are the steps to be taken in
arriving at a merged health risk ranking for a set of Problems that
have been ranked separately according to the risks associated with
two different classes of health effects (cancer and non-cancer
effects, in this case):
(1) List the Problems that have been ranked for both cancer
and non-cancer risks.
(2) For those Problems that have been ranked for both, group
the cancer and non-cancer rankings separately into three
qualitative risk levels: high (H), medium (M) and low (L)
if this is not the way they have been ranked already.
This is best done by an appropriately selected,
knowledgeable, consensus group.
(3) List the Problems that lie within each of the nine
possible grid squares of the three-by- three risk array;
plotting them helps visualize the information.
144
-------
(4) Make an initial
decision as to how
v relates to unity
if the information
available permits;
in any case,
whether this is
possible or not,
check major rank
reversals between
pairs of rankings
for different
values of v (see
Tables 8.2.6.1 and
8.2.6.2 for
examples of major
rank reversals as
v changes;
particularly re-
versals of C and G,
v or con- firm the
the same consensus
Mo.* For; The ranking pattern it;
v * infinity AOC> KM > CFI
11. V > 1 A>D>G>B>E>N>C>F>I
12. • A>D>BG>E>CM>F>I
L3. " A>D>B>G>E>C>H>F>!
L4. V « 1 A > 80 > CEG > FH > I
L5. v < 1 A>B>D>C>E>G>F>H>I
L6. • A>B>CO>E>FG>H>I
L7. • A>B>C>D>E>F>G>H>I
V » 0 ABC > DEF > GHI
* Only the physically non-trivial rankings are
nunbered.
Table 8.2.6.1 Rankings possible for a
linear three-by-three array
B and D, and F and H) to either select
selection made. This is best done by
group.
(5) Using the result of step (4),select from the possible
range rankings for the linear array (Table 8.2.6.1) the
rankings in keeping with that result.
(6) Compare the rankings in step 5) with the information
available on individual Problems, as described above, to
conclude, on balance, which ranking is most in keeping
with the information. This is best done by the same
consensus group.
(7) Use the result of step (6) as guidance to select rankings
for nonlinear arrays (e.g., Table 8.2.6.2) for comparisons
such as have been made for linear arrays in step (6)
(Check the selection of v against this array of rankings,
also).
(8) Of the sets of rankings now in hand, select the best one,
overall, from the two types of arrays. This is best done
by the same consensus group.
145
-------
(9) Rank order the prob-
lems within each of
the ranking groups
obtained in step (8)
if, and to the
extent which, this
is possible. This is
best done by the
same consensus
group.
(10) Check the nearly
final ranking, Prob-
lem by Problem,
against the infor-
rmation available on
each Problem to see
if any specific
Problems need to be
moved upward or
downward in rank in
the nearly final
ranking. If step
(9) has been car-
ried out, step (10)
can be made easier
N1.
M2.
M5.
M.
MS.
N6.
M7.
N8.
H9.
MO.
»11.
M12.
H13.
NU.
N15.
Fort The ranking pattern i«:
v * infinite ADG > BEN > CF1
v > 1 A>D>6>B>E>H>OF>1
• A>0>G>B>E>HC>F>I
• A>0>G>B>E»C>»»F>I
• A>D>G>B>EC>H>F>I
• A>D>G>B>C>E>H>F>I
• A>0>GB>C>E>H>F>I
• A>0>B>G>C>E>H>F>I
v a 1 A>08>GC>E>HF>I
v < 1 A>8>0>C>G>E>F>M>1
• A>B>DC>6>E>F>H>I
• A>B>C>D>G>E>F>M>I
• A>B>C>D>GE>F>H>I
• A>B>C>0>E>G>F>H>I
• A>B>C>D>E>GF>H>I
• A>B>C>0>E>F>G>H>I
v * 0 ABC > DEF > GHI
• Only the physically non-trivial rankings are
nuabered.
by comparing Prob-
lems at the high Table 8.2.6.2Rankings possible for
end of one range a non-linear three-by-three array
with those at the low end of the
range first, and vice versa. This,
the same consensus group.
next higher ranking
too, is best done by
When step (10) is completed, the final ranking is in
hand. This final step, not taken in the illustrative example, is
very important; it is the final opportunity to correct the joint
ranking, exposing and correcting not only the overlaps already
described but even, possibly, any errors made in the original
rankings.
A note of caution: this discussion should not imply that one
would require a high ranking for both cancer and non-cancer health
effects to consider an exposure to be of high priority.
146
-------
8.2.7 An Illustration of tha Merging of Separate Rankings
into One
The UB report represents an extensive study in which a set of
Problems, thirty-one in number, significant to the U.S. EPA, was
defined and ranked in two separate rankings according to their
cancer and non-cancer effects population risks. In this section
the UB report problems are referred to using the same numbers, from
1 to 31, as used in the report. Although the Human Health
Subcommittee of the Relative Risk Reduction Steering Committee of
the Science Advisory Board has reservations about the definitions
and rankings of the Problems in the UB report (see elsewhere in
this report for discussions and recommendations), it was concluded
that the thirty-one Problems in the UB report, with some
modification, and the information in the UB report relative to
those Problems, could be used to illustrate how the risk merging
procedure is applied.
The modifications involve a regrouping of the thirty-one
Problems and one additional one (electromagnetic fields) under
three main headings: Situations and Agents Involving the Potential
for direct Exposure, Sources of Environmental Pollution, and
Miscellaneous (see section 5.4 and Table 5.4.1). Further sub-
groupings within these categories were proposed. Thus, for
example, Occupational Exposures included Worker Exposures (Problem
# 31) and Application of Pesticides (#26). In the proposal a
number of individual Problems as defined in the UB report appeared
to be better combined as new Problems (an example is the possible
combination, for purposes of health risk ranking of Discharges to
Estuaries, #13, and Discharges to Wetlands, #14). In the case of
Occupational Exposures, although the proposal groups them together,
the two types of exposures Application of Pesticides and Worker
Exposures) are so very different (different populations, different
physical conditions, different kinds of remedial actions possible,
etc...) that it would not be useful to consider them as one
Problem. However, Indoor Air-Radon (#4) and Indoor Air-Other (#5)
are readily redefined as a single Problem, Indoor Air (#4/5); here
the same population is affected, the exposure situation is
physically well defined, and many of the remedial methods apply to
more than one agent present.
In this illustrative example of how the merging process is
carried out, only one pair of Problems, Indoor Air-Radon (#4) and
147
-------
Indoor Air-Other (#5) will be combined into a new Problem, Indoor
Air (#4/5), leaving the other Problems as in the UB report.
In the UB report, Problems are already grouped as H, M or L
for non-cancer risk, the consensus group that created that ranking
having concluded that no finer subdivision was possible. In the
case of cancer risks, all but five of the thirty-one problems were
ranked, qualitatively, one above the other, a few pairs being given
the same ranking; here, the existing ranking had to be reduced to
three levels as is already the case for non-cancer risk ranking.
8.2.7 Grouping the Cancer Risk Ranked Problems into Three
RisX Groupings
An examination of the ranking on the basis of cancer risks and
of the factors considered in the UB report leads to the conclusion
that a reasonably natural boundary between the "high" and the
"medium" levels for cancer risks lies between the eighth and ninth
ranked Problems; similarly, the boundary between the "medium" and
"low" levels lies reasonably naturally between the seventeenth and
eighteenth ranked Problems. The three rankings for cancer risk
that result are as follows (with Indoor Air combined as #4/5, as
above):
Rank Level Problem Numbers
High (H) 2,4/5,7,17,25,30,31
Medium (M) 6,12,15,16,18,19,26,27,28
Low (L) 1,9,10,11,20,21,22,23,24
Among the H-ranked Problems, nos. 31 and 4/5 stand out
relative to the rest as, in effect, "extra high." Please note that
this is a tentative grouping into H. M. and L categories for
illustrative purposes only; this grouping in no way represents a
consensus of the current Subcommittee as to the relative risk to
human health from any of the "Problem Areas" as set forth in the
original UB Report. In any actual case of ranking Problems,
however defined, or by using any methodologyy, such a ranking
should be subjected to a broad consensus process.
Table 8.2.7.1 shows the above Problems, ranked H, M or L for
cancer risk together with the H, M or L rankings for non-cancer
risks. Only twenty-two Problems, as defined here, were ranked
148
-------
simultaneously for both cancer
and non-cancer risks. The
rest of this example considers
how to merge the rankings for
these twenty-two Problems,
only, to produce a rank order
list based on both cancer and
non-cancer risks.
The information in Table
8.2.7.1 is plotted in Figure
8.2.7.1; the numbers near each
of the nodes indicate which
Problems lie within the
corresponding grid squares,
their actual locations within
the grid squares being
unknown. Table 8.2.8.1 shows
the same information, with the
problems listed in the same
order, from left to right, as
in their cancer risk ranking
so as not to lose sight of
this information.
8.2.8 The Value of v
Relative to 1.0
Probloi iHntH
2
4/5
7
17
25
30
31
6
12
15
16
18
19
26
27
28
1
9
10
11
20
21
22
23
24
3
8
13
14
29
T Cancer Rant; IK
H
H
H
H
H
H
H
N
N
N
N
H
N
N
N
N
L
L
L
L
L
L
L
L
L
.
-
-
-
^n-cmer Rank
H
H
N
L
M
H
H
N
L
H
L
N
N
H
N
-
H
L
N
N
L
H
-
L
-
.
-
N
L
Table 8.2.7.1 High/ Low/ and
Medium rankings for UB problem
areas
A direct estimate of v, even to the extent of concluding
whether it is above or below one, cannot be made with any
reasonable degree of certainty. The severities of cancers, on
average, are well above those of the aggregation of non-cancer
endpoints considered in the UB report so that, unless the fraction
of the population affected by non-cancer endpoints is very much
higher than that for all cancers, as related to relevant agents in
the environment, v is more probably less than one than it is above
one. The consideration of the consistency of the information in
the UB report on individual Problems with respect to reversals of
rankings of ranges from one possible ranking to another for
different values of v (Tables 8.2.6.1 and 8.2.6.2) is a surer
indicator of where v lies in this case. These comparisons are
described in the next subsection. In any case, where v lies with
respect to one is not to be chosen arbitrarily but, rather, on the
basis of what the information itself demonstrates it to be.
149
-------
cr
LLJ
u
o
M —^^- m -^B— H •
CANCER
Figure 8.2.7.1 Actual Problems
8.2.9 Consideration of Possible Rankings for consistency
with Available Informations Selection of a Ranking
of Ranges
The simplest approach is to examine the linear array rankings
in Table 8.2.6.1 before passing on to considering the nonlinear
rankings in Table 8.2.6.2. The results of considering linear
rankings first, despite the fact that it would seem more reasonable
to choose the nonlinear ones first, can serve as guides when
nonlinear rankings are examined; moreover, it is not in fact known
what the original ranking teams had in mind insofar as linearity or
non-linearity is concerned, nor what their unconscious choices
150
-------
might have been, as they ranked the
Problems for cancer and non-cancer risks.
The assumption that v < l, for a
linear array was first tested by examining
rankings where large reversals of rank of
pairs of ranges in Table 8.2.6.1 occur.
The reversals of B and D, F and H, and C
and G between the rankings for v < 1 and v
> 1 are striking. Examining the
descriptive information in the UB report
on the Problems contained in these three
ranges (see Table 8.2.8.1) shows clearly
that B > D, F > H, and C > G is more
consistent with the information than the
reverse and thus the conclusion that v < contained
1 is the reasonable one. range, in
cancer risk
With v < 1, Table 8.2.6.1 gives the ri$ht>
following possible rankings for further consideration:
SfH TTIfTf Probta»
A 31. 4/5, 25, 30, 2
I 7
C 17
0 15. 26
E 6. 27, 19, 18
f 16. 12
C 1. 21
H 11. 10
1
20. 23. 9
in each
order of
(left-to-
L5
A>B>D>C>E>G>F>H>I
L6
A>B>CD>E>FG>H>I
L7
A>B>C>D>E>F>G>H>I
Ranges shown grpuped together are of equal rank.
Examining the information given in the UB report for the
Problems contained in ranges C and D, the information is
inconsistent with the order C > D; D > C is only weakly supported;
and lumping C and D is the best choice. Moreover, on balance, the
arguments for lumping F and G appeared to be better than those for
keeping them separate; thus, merged risk ranking number L6 was
selected of the three possible ones given above for a linear array.
Other inequalities in L6 appear consistent with the information in
the UB report, although B and D appear to be closely ranked; H and
I also appear to be close together, though not so close as B and D;
one has to bear in mind the overlapping nature of the ranges. For
the moment, pending examination of possible rankings for nonlinear
arrays (e.g., as in Table 8.2.6.2), L6 will be the ranking of
choice.
151
-------
Considering the nonlinear array rankings against L6, rankings
N9 and N10 appear to be the best to examine first, in particular
the relationships of E, G and F that, in L6, are in the order E >
FG but in N9 and N10 are in the order G > E > F. It is found that
E > F and E > G is consistent with the information in the UB
report; and, since it was already determined that lumping F and G
was more reasonable than not, L6 appears as the best choice of all
for the ranking to be examined in detail, problem by problem. Note
that N14 is not a good choice since lumping C and D, and F and G,
is preferable. Table 8.2.9.1 shows this ranking with the Problems
within each of the ranges listed, from left to right in their
cancer risk order and with their non-cancer rankings shown in
parentheses. The fact that the linear rather than the non- linear
array produced the most consistent result does not force the
conclusion that the array is indeed linear. It may be not too
highly nonlinear, the uncer-
tainty of the inforrmation and
its gualitative nature being
such as to not make too close
a discrimination possible; or
the array may be nonlinear but
not exactly symmetrical about
the diagonal; or other
deviations of the real example
from the theoretical model may
cause the appearance of near
linearity.
Ranges)
A
ft
DC
E
6F
H
I
Problaa
3UH). 4/5CH), 25(H), 3Q(H), 2. 12(L)
1KN). 10(N)
20(L). 23(L), 9(L>
Table 8.2.9.1 Selected rankings for
further expansion and consideration
In Table 8.2.9.1, in all (in order of cancer risk/ left-to-
but DC and GF the, non-cancer right; non-cancer risk ranking in
rankings are the same; thus, parentheses)
recognizing that with new information the non-cancer rankings may
alter this conclusion, for the moment the cancer ranking order
appears to prevail within each of these ranges. In DC and GF,
consideration of the information in the UB report on each of the
individual Problems shows that ranking them in cancer risk order is
consistent with that information. In GF, the fact that the two
highest in cancer order are both ranked H for non-cancer risk and
the lower two, in cancer order, are ranked L for non-cancer risk is
consistent with this finding. Table 8.2.9.2 shows the information
in Table 8.2.9.1, with the Problem descriptions included. In this
form the tabulation is, to all intents and purposes, a nearly final
illustrative merged health risk ranking, by problem, of the twenty-
two original problems, as previously ranked in the UB Report,
152
-------
separately by
cancer and non-
cancer population
risk. This ranking
is based on the
original UB report
scores for the 31
categories
of
problem areas, and
in no way reflects
the view of this
Subcommittee. It
serves
only
to
illustrate
the
application of a
theoretical
approach to merged
risk ranking.
Th is last
ranking is called
nearly final
because this is the
ranking that now
should be examined,
Problem by Problem,
information to see
order or rank
EanaeCsJ Probleat D^criotion of Probleq
A 31 * Worker exposure to cheaicals
4/5 • Indoor air
25 Pest. re*, on food eaten by huBins/wildlife
30 Consider product exposure
2 Hazardous/toxic air pollutants
B 7 Stratospheric ozone depleting substances
DC 17 Hazardous waste sites — inactive
15 Drinking water at the tap
26 App. of past, (applicators, consumers, etc)
E 6 Radiation — other than radon
27 Other pest, risks (leaching, runoff, etc)
19 Honhazardous wast* sites -- industrial
18 Honhazardous waste sites — Municipal
GF 1 Criteria air pollutants (stat.ft anbile src.
21 Accidental releases -- toxics (all Media)
16 Hazardous waste sites — active
12 Contaminated sludge (Municipal and scrubber)
H 11 Non-point surface discharges to surface uat.
10 Indir. pt. src. disch. (POTUs) to surf. wat.
I 20 Mining waste (inc. oil 4 gas extraction)
23 Rel. froai stor. tanks (on/above/underground)
9 Direct point discharges to surface waters
•Essentially of equal rank, high relative to the others in
this range.
Table 8.2.9.2 Hypothetical "Nearly Final"
merged risk rankings (illustrative), based on
the unmodified UB Report information
for overall consistency with the available
if any Problems need to be exchanged in rank
ordered equally because of the overlap of ranges
already discussed. A cursory examination does not indicate the
need for changes, though a few pairs of Problems are probably better
shown to be of equal rank as opposed to their ranking in Table
8.2.9.2. The examinations of the pairs of ranges above, and of
this nearly final ranking for consistency with the available
information should be by consensus of experts for the soundest
results. When completed, the results, to be consistent with the
input information, should be reported as "high," "medium," and
"low" risk groups (though some may stand out within these groupings
as, for example, nos. three and four/five).
8.2.10
Further Comments and Recommendations
For the long term use of merged cancer and non-cancer risk
ranking, the so-called zero-based procedure outlined early in this
section is best. Doing it once can form a solid basis for updating
153
-------
and revising it and, since it deals most directly with the problem
in a manner as close as possible to equation (l) and its alternate
form in terms of P and D, equation (2), it is likely to yield the
most correct and credible, and therefore reliable, result when it
comes to budgeting and allocating resources to risk management
activities and to research. It is recommended that this effort be
undertaken as an investment in facilitating better planning and
allocation.
The procedure for merging separately ranked Problems (for
cancer and non-cancer risk) is relatively easy to use, now that the
possible rankings are tabulated in Tables 8.2.6.1 and 8.2.6.2 and
once separate cancer and non-cancer risk rankings are in hand. The
consensus mechanism recommended is particularly useful not only in
narrowing down the possible rankings to one best one but also in
reaching the final merged ranking while ensuring that information
that might have been lost in reducing the cancer risk based
rankings to three levels is utilized at the end.
8.2.11 Derivation of Wj s vcj + vw,,j and of v
Numbering the cancer endpoints from 1 to C and the non-cancer
endpoints from C + 1 to E, equation (1) may be rewritten in the
following form:
Wj = Nj X Z Sj fljk + Nj Z Z S. fjjk (6)
1=1 k=l i=C+l k=l
that in turn may be written in the form
Wj = Wcj + WM. (7)
Here, the three terms correspond to the parallel three terms
in equation (Al) . The first term on the right represents the
aggregate weight of cancers and the second term that for all non-
cancer endpoints, all appropriately scaled for severity, potency
and exposure (see, too, equation (2) in the body of the report) .
Wj, as before, then represents the weight to be used for ranking
purposes to rank the jth problem with respect to the other problems
being considered.
The terms in these equations are not directly known, in the
present case, cancer risks having been ranked against cancer risks,
154
-------
only, non-cancer risks against non-cancer risks, only, and the
relative scaling of the two types of risks not having been
addressed. The rankings are qualitative, too, not quantitative.
Equation (7) , to be useful here, must be recast in terms relevant
to the present case.
Defining wcj as the weight to be used for ranking the jth
Problem against other Problems on the basis of cancer risk only,
scaled so that the weight of the problem of maximum cancer risk is
h, then
wcj = hWcj/WCH (8)
Similarly, for non-cancer risk ranking,
wNj = hWMj/WNH (9)
Here WCH is the weight for cancer risk, as defined for
equations (6) and (7) , of the Problem in the set of Problems
considered for ranking according to cancer risk that has the
highest weight (and therefore would be ranked first for cancer if
the weight were known, quantitatively), and WNH is the same but for
the Problem that ranks highest for non-cancer risk. The Problems
need not be the same in the two cases.
Equation (7), combined with equations (8) and (9) and
multiplied by h/WCH becomes
(h/WCH)W. = wcj + vwNj (10)
where v = WMH/WCH (11)
a constant for the particular set of problems being ranked; note
that v may take a different value for another set of problems or
for subsets of the original set of problems if WMH and WCH or their
ratio is not the same from one set to the other.
Since h is a constant, and since WCH is a constant for the
particular set of ranked problems under consideration, then the
left hand side of equation (10) is the weight to be accorded the
combination of cancer and non-cancer risks in ranking the jth
problem against the other problems in its set, Wj. Thus equation
(10) becomes
155
-------
W. = Wcj + VWNj (12)
This is the key equation in the rank merging method. Though
it is derived in this instance for the case in which the separate
rankings are made on the basis of population risk (as is the
derivation of v which follows) , the same form of equation is
obtained if either individual risk or a mixture of individual and
population risk is used. In these latter two cases the definition
of v is different, in each case, from the one derived below, but
this has no impact on the number and nature of the possible
rankings derived later on. Special factors, suitably formulated,
such as for individual or population sensitivity, may also be
included in the derivation without altering the form of the key
equation.
From equations (6) and (7) ,
WCH = NCH S X S,f(Jk (13)
1=1 k=l
where here j is for the Problem with the highest cancer risk
E A
and WMH = N S Z Sff(jk (14)
i=c+l k=l
where here j for the Problem with the highest non-cancer risk.
The average severity of cancers in the Problem with weight WCH is
SCH, where
1=1
S
CH
C A
s r ffjk
156
-------
and a similar expression is obtained for the average severity of
non-cancer endpoints, SMH, in the Problem with weight WMH.
Substituting these terms into equations (13) and (14) ,
A
S
k=l
WCH - scHNCH . * 2 fijk (16)
W_ O VT r* V -f
iitj *^UU^klU ""* "™* "^ i lU
NH NH NH i=c+i k=i ,jk
The double summation in equation (16) is an estimate of the
fraction of all exposed subjects in the highest cancer risk Problem
who exhibit cancerous endpoints, FCH (note that the fraction of
those who exhibit at least one endpoint—those showing any effect -
-is slightly less, but the difference is small because the
individual f-values are small). Similarly, fraction FMH is defined
for the Problem with the highest non-cancer risk. Then
NH
(18)
The two fractions, F, are functions of the potencies of the agents
and of exposures to them; equation (18) thus indicates the factors,
and their relationships, that determine the value of v for a
particular set of points being ranked.
8.1.12 Derivation of Some of the Characteristics of Three-By-
Three Arrays as Governed by Equation (3)
Suppose that for some value of v both nodes C and G project
onto the diagonal at E (see the solid arrows indicating this in
Figure 8.2.4.1. This means that the WCMJ-values for C, E and G are
equal. Similarly to equation (5),
h+vl=m+vm=l+vh (19)
from which it follows that v = 1 for this case. Similarly, if F
and H project onto the diagonal at some common point and B and D do
so at another common point, then
157
-------
m + vl = 1 + vm (20)
and h + vm = m + vh (21)
and in each case v = l. Thus for v = 1, the merged risk rank order
is, for a linear array:
A > BD > CEG > FH > I (22)
giving a total of five risk levels since some of the nodes are of
the same relative rank, as indicated. For a. nonlinear array, the
v - 1 rank order becomes: A > BD > CG > E > FH > I; here, C, E and
G are not co-linear (see Figure 8.2.4.2).
Consider nodes F and G in Figure 8.2.4.1, projected onto the
diagonal at a common point. Here,
m + vl - 1 + vh (23)
from which
m - 1
< 1 (24)
h - 1
Designating the value of v by the letters corresponding to the
off-diagonal nodes projected onto common points oh the diagonal,
referring to Figure 8.2.7.1,
v(FH) * v(CG) - v(BD) - 1 (25)
m - 1
and v(FG) * < 1 (26)
h - 1
By similar reasoning,
h - m
v(CD) * 1 - v(FG) <1 (27)
h - 1
158
-------
h - 1
V(BG) = = 1/V(CD) > 1 (28)
h - m
h - 1
and v(CH) l/v(FG) > 1 (29)
m - 1
This kind of treatment gives the values of v, and the relationships
between them, for which off-diagonal nodes can be projected onto
common points on the diagonal for linear and nonlinear arrays. For
linear arrays, regardless of the values of h, m or 1, it turns out
that not only does v(FH) = v(CG) - v(BD) - 1, but v(CD) = v(FG) =
0.5 and v(BG) = v(CH) = 2.0, and these are the three values of v
leading to common projections. For nonlinear arrays, the same
common projections lead to v = 1; however v(CD) and v(FG) are not
equal, and nor are the pairs v(BG) and v(CH) or v(CG) and v(EG).
In the nonlinear case, there are therefore seven values of v that
lead to common projections.
For projections onto any point on the diagonal, with
coordinates (x,x), from an off-diagonal node with coordinates C1
(for cancer) and N1 (for non-cancer), the equality of the two WCMJ.
values at the two points requires that
x+vx=C'+vN' (30)
C1- x
or, v = (31)
x - N1
The slope of the vector connecting (C',N') with (x,x) is
N1 - x - N1
slope = (32)
x - C1 C1- x
Thus, for any such vector,
slope = -l/v (33)
159
-------
Considering the above, it is expected that for linear arrays
there will be seven possible rankings (three for the values of v
leading to common projections and four more corresponding to values
of v between and on each side of the highest and lowest of the
three common projection values, excluding the physically trivial
cases of v equal either to zero or infinity) and fifteen possible
rankings for the nonlinear case (seven for the common projections
and eight other, again excluding the two trivial cases). These
are tabulated in Tables 8.2.6.1 and 8.2.6.2. An alternative to
deriving the rankings by considering, geometrically, the rotations
of the vectors around the nodes so as intersect the diagonal is to
assign values to v. For either the linear or nonlinear arrays,
values of v for the common points plus arbitrarily chosen values of
v between and to either side of these makes it possible to
calculate the weights, and therefore to derive the rank orders,
corresponding to each of these values of v; thus the possible sets
of rank orders are derived. For the linear case, the common-point
values of v are 0.5, 1.0 and 2.0, whereas for the nonlinear case
these values depend on the h-m/m-1 ratio; however, common-point
values of v based on any arbitrarily chosen value of this ratio,
(or of the values of h, m, or 1) plus other values of v falling
between and to either side of the common-point values, may be used
to make the weight calculation with the same result regardless of
the choices; in the case of three-by-three nonlinear arrays,
whatever the value chosen for h-m/m-1, the rank order patterns
derived will be the same. For four-by-four and higher nonlinear
arrays, more than one set of rankings will be obtained depending on
the specific values of v or ranges of values of h, m, or 1.
160
-------
REFERENCES
Section 1
1. EPA (U.S. Environmental Protection Agency) , Unfinished Business:
A Comparative Assessment of Environmental Problems. Washington,
DC. 1987.
2. EPA (U.S. Environmental Protection Agency), Future Risk:
Research Strategies for The 1990s. Washington, D.C. 1988.
Section 3
1. Cartwright R.A., Glashan R.W., Rogers H.J., et al: Role of N-
acetyltransferase phenotype in bladder carcinogenesis: A phar-
macogenetic epidemiological approach to bladder cancer. Lancet,
1982; ii:842-846.
2. CASAC (1989, Report of the Clean Air Scientific Advisory
Committee—Review of the NAAQS for Ozone; Closure on the OAOPS
Staff Paper (1988) and the Criteria Document Supplement (1988).
U.S. E.P.A., Washington D.C.
3. Chan-Yeung M, Ashley M. J., Corey P., et al: Pi phenotypes and
the prevalence of chest symptoms and lung function abnormalities in
workers employed in dusty industries. Am Rev Respir Pis 1978;
117:239-245.
4. Doll R. and Peto R. (1981). The causes of cancer. Quantitative
estimates of avoidable risks of cancer in the United States today.
JNCI 66, 1192-1308.
5. Evans, G.W., Colome, S.D., Shearer, D.F., Psychological
Reactions to Air Pollution. Environmental Research. 1988, Vol 45:
1-15.
6. Ershow, A.G. and Cantor, K.P. Population-based estimate of water
intake. Fed. Proc. 45:706, 1986.
7. Ferris, B.C., Health Effects of Exposure to Low Levels of
Regulated Air Pollutants: A Critical Review. Air Pollution Control
Assoc.. Vol 28, No. 5, May 1978
8. Finkel, A.M. Confronting uncertainty in risk management. A guide
for decision-makers. Center for Risk Management, Resources for the
Future, Washington D.C. 1990
9. Gaylor, D.W. and Kodell, R.L. Linear interpolation algorithm for
low dose risk assessment of toxic substances. J. Environ. Pathol.
and Toxicol. 4:305-312, 1980
161
-------
10. Gaylor, D.W. Quantitative risk analysis for quantal
reproductive and developmental effects. Environ. Health Persp.
79:243-246, 1989
11. Harris, C.C. Future directions in the use of DNA adducts as
internal dosimeters for monitoring human exposure to environmental
mutagens and carcinogens. Environ. Health Persp. 62:185-91, 1985.
12. Higgins I.T.T., What Is An Adverse Health Effect? J Air
Pollution Control Assoc. 1983, Vol 33:661-663.
13. Hills, B.W., Venable, H.L. The interaction of ethyl alcohol and
industrial chemicals. Am Jour Ind Med. 1982; 3:331-333.
14. Jadrychowski, W. Interaction of smoking and urban air pollution
in the etiology of lung cancer. Neoplama. 1983 30:603-609.
15. Krewski, D. and Van Ryzin, J. Dose response models for quantal
response toxicity data, pp 201 in Statistics and Related Topics.
1981, Ed. .Sxorgo, J., Dawson, D., Rao, J.N.K., and Saleh, E.,
North Holland, New York.
16. Lippmann M. (1989). Health effects of ozone. A critical review.
JAPCA. 39:671-695.
17. Littlefield, N.A., Farmer, J.H., Gaylor, D.W., and Sheldon,
W.G. Effects of dose and time in a long-term, low-dose carcinogenic
study. J. Environ. Patho}. and Toxicol. 3:17-34, 1979.
18. Machina, M.J. Choice Under Uncertainty: Problems Solved and
Unsolved. In: Valuing Health Risks. Costs, and Benefits for
Environmental Decision Making - Report of a Conference. Ed. P. B.
Hammond and R. Coppock. National Academy Press, Washington, 1990,
pp. 134-188
19. Mahaffey, K.R. Nutritional factors in lead poisoning. Nutr Rev
1981; 39:353-365
20. Marquis, J.K. and Siek, G.C. Sensitive populations and risk
assessment in environmental policy-making. In: Saxena, J. (ed.),
Hazard Assessment of Chemicals. Vol. 6, Washington D.C. Hemisphere,
1988.
162
-------
21. Milvy, P. 1986. "A General Guideline for Management of Risk
from Carcinogens." Risk Analysis 6(1), 81-94.
22. NAS, Committee on the Epidemiology of Air Pollutants.
Epidemiology and Air Pollution. Washington D.C.: National Academy
Press, 1985.
23. NAS, Committee on the Institutional Means for Assessment of
Risks to Public Health. National Research Council. Risk Assessment
in the Federal Government; Managing the Process. Washington D.C.:
National Academy Press, 1983
24. NAS, Drinking Water and Health. Safe Drinking Water Committee.
Wash. D.C. National Academy of Sciences, Vol. 6, 1986
25. National Research Council. Biologic Markers in Pulmonary
Toxicology. Washington D.C.: National Academy Press, 1989.
26. National Research Council, Risk Assessment. In: The Federal
Government; Managing the Process. National Academy Press, 1983.)
27. Needleman, H., Schell, A., Bellinger, D., Leviton, A., Allred,
E.N. The Long-term Effects of Exposure to Low-doses of Lead in
Children: 11 Year Follow-up Report. New Ena. J. Med.. 1990, Vol
322, pp 83-88.
28. Perera, F., Mayer, J., Santella, R.M., Brenner, D., Jeffrey,
A., Latriano, L., Smith, S., Warburton, D., Toung, T.L., Tsai,
W.Y., Hemminki, K., and Brandt-Rauf, P., Biologic markers in risk
assessment for environmental carcinogens. Environ. Health Persp.
(in press), 1990.
29. Radike, M.K., Stemmer, K.L., Brown, P.G., Larson E., Bingham,
E. Effect of ethanol and vinyl chloride on the induction of liver
tumors: Preliminary Report. Environ. Health Perspect. 1977, 21:153-
155.
30. Rothman K.J., Modern Epidemiology. Little Brown and Company.
Boston, 1986.
163
-------
31. Silverstone H., Searle J.H.A.: The epidemiology of skin cancer
in Queensland: the influence of phenotype and environment. Br J
Cancer 1970; 24:235-252.
32. Spengler J.D., Soczek M.L., Evidence for improved ambient air
quality and the need for personal exposure research. Environ. Sci
Technol 1984; 18:286-280
33. Steenland, K, Thun, M. Interaction between tobacco smoking and
occupational exposures in the causation of lung cancer. Jour Occup
Med 1986; 28:110-118
34. Upton A.C. Biological aspects of radiation carcinogenesis. In:
Radiation Carcinogenesis: Epidemiology and Biological Significance
Ed. J.D. Boice Jr., J.F. Fraumeni Jr.), pp. 9-19. Raven Press New
York, 1984.
35. Upton A.C. The question of thresholds for radiation and
chemical carcinogenesis. Cancer Investigation. 1989, 7: 267-276.
36. Vainio, H., Sorsa, M., and McMichael, A.J. (eds.), Complex
Mixtures and Cancer Risk. IARC Scientific Publications, No. 104,
Lyon (in press).
37. Wallace L.A., Ott W.R., Personal Monitors: a state-of-the-art-
survey J Air Pollut Control Assoc 1982; 32:601-10be
38. Waters, M.D., Nesnow, S., Lewtas, J., Moore, M.M., and Daniel,
F.B. (eds.), Genetic Toxicoloo^ of Complex Mixtures. Plenum
Publishing Corp. NY (in press).
39. Willet, W. Vitamin A and Lung Cancer. In Nutritional
Epidemiology. W. Willet Ed. New York; Oxford University Press,
1990.
40. Witschi H.P. Altered tissue reactivity and interactions between
chemicals. In: Assessment of Multichemical Contamination. Proceed-
ings of an International Workshop. Milan, Italy. National Academy
Press, 1982.
164
-------
41. witschi H.P. Toxicological interactions and the pathogenesis of
acute and chronic lung injury. In: Concents in Inhalation Toxicol-
ogy (eds. R.O. McClellan and R.F. Henderson), pp. 475-492.
Hemisphere Publishing Corp., Washington DC, 1988.
42. Zeckhauser, R.J. and Viscusi, W.K., Risk Within Reason. 1990,
Science 248:559-564.
43. Zweig, G.A. The Case of The Vanishing Zero: Evolution of
Pesticides Analysis, Essays in Toxicology. Vol. II, 1970, pp 155-
198.
Section 5
1. Huffier, P.A.,Crane, M., Key, M.M., Possibilities of Detecting
Health Effects By Studies of Populations Exposed to Chemicals from
Waste Disposal Sites. Env. Hlth. Persp., 62:423-456, 1985.
2. Doll, R. and Peto, R., JNCI 66:1191-1308. 1981
3. Bond, G. G., Shellenberger, R. J., Fishbeck, W. A.,
Cartmill, J. B., Lasich, J. B., Wymer, K. T., and Cook, R.
R., JNCI 75(5):859-869
4. Bond, G. G., McLaren, E. A., Cartmill, J. B., Wymer, K. T.,
Sobel, W., Lipps, T. E., and Cook, R. R., Am. J. Indust.
Med. 12:563-578, 1987
5. Bond, G. G., McLaren, E. A., Cartmill, J. B., Wymer, K. T.,
Sobel, W., Lipps, T. E., and Cook, R. R., Am. J. Indust.
Med. 12:353-383, 1987
6. Delzell, E., Austin H. and Cole, P., Occup. Med. l(No.3):455-
474, 1988
7. Driscoll, K.E., Vollmuth, T.A., and Schlesinger, R.B. Early
Alveolar Clearance of Particles in rabbits Undergoing Acute and
Sub-chronic exposure to Ozone. Fundamental Appl. Toxicol. 7:264-
271, 1986.
165
-------
8. Gunnison, A.F., Sellakumar, A., Synder, E.A., and Currie, D. The
Effect of Inhaled sulfur Dioxide And Systemic Sulfite on The
Induction of Luna carcinoma in Rats Bv Benzofa)pyrene. Environ.
Res. 46:59-73, 1988.
9. Hassett, C., Mustafa, M.G., Coulson, W.F., and Elashoff, R.M.
Murine Lung carcinoaenesis Following Exposure to Ambient Ozone
Concentrations. JNCI 75:771-777, 1985.
10. Kaplan, S. D., J. OCCUP. Med. 28(7);514-516. 1986
11. King, J. OSHA Dramatically Re-writes Exposure Limits to Protect
Workers. AP Wire Service, January 13, 1989.
12. Levy, B.S., and Wegman, D.H. "Occupational Health in The United
States," in: Occupational Health. Recognizing And Preventing Work-
related Disease, pp. 3-4, Ed: Levy, B.S. and Wegman, D.H. Little
Brown and Co., Boston, 1988.
13. National Academy of Sciences-National Research Council
(NAS/BEIR). The Effects on Populations of Exposure to Low Levels
of Ionizing Radiation. (BEIR III). National Academy Press,
Washington, 1980.
14. National Academy of Science-National research Council.
Regulating Pesticides in Food. National Academy Press, Washington
D.C., 1987.
15. National Academy of Sciences-National Research Council
(NAS/BEIR). Health Effects of Low Level Ionizing Radiation. BEIR V,
National Academy Press, Washington D.C. 1990.
16. Nicholson, W.J., Arch, of Environ. Health 39(No.3);190-202
(1984)
17. Pifer, J. W., Hearne, F. T., Friedlander, B. R., and
McDonough, J. R., J. OCCUP. Med. 28(6);438-444 (1986)
18. Schlesinger, R.B. and Gearhart, J.M. Intermittent Exposures to
Mixed Atmospheres of Nitrogen Dioxide And Sulfuric Acid; Effect on
Particle Clearance from The Respiratory Region of Rabbj.fr Luncrs.
Toxicology 44:309-319, 1987.
166
-------
19. Speers, M.A., Dobbins, J.G, and Miller, Van s. Occupational
exposures and brain cancer mortality; A preliminary study of
East Texas resident?. Am. J. Indstr. Med. 13:629-638, 1988.
20. Upton, A.C., Kneip, E., and Toniolo, P. Public Health Aspects
of Toxic Chemical Disposal Sites. Ann. Rev. of Pub. Hlth, 10:1-25,
1988.
21. Warren, D.L., and Last, J.A. Synergistic Interaction of Ozone
and Respiratory Aerosols in rat Luna. III. Ozone and sulfuric acid
aerosol. Toxicol. Appl. Pharmacol. 88:203-216, 1987.
22. Witschi, H. Ozone, nitrogen dioxide and lung cancer: A review
of some recent issues and problems. Toxicol. 48; 1-20, 1988.
23. Wong, 0. and Raabe, G. K., Am. J. Indust. Med. ^5:283-310
1989
24. Yancey, M. Rate of Job Injuries. Disease, on Rise. AP Wire
Service, November 11, 1988.
Section 6
1. International Commission on Radiological Protection, Problems
Involved in Developing an Index of Harm. ICRP Publication 27,
Pergamon Press, New York, 1977.
2. International Commission on Radiological Protection, Quan-
titative Bases for Developing a Unified Index of Harm. ICRP
Publication 45, Pergamon Press, New York, 1985.
3. National Research Council. Complex Mixtures - Methods for In
Vivo Toxicity Testing. National Academy Press, Washington, DC
(1988).
4. National Research Council, Toxicity Testing — Strategies to
Determine Needs and Priorities. National Academy Press, Washington,
D.C. 1984.
167
-------
Section 8.1.1—Appendix
OZONE CASE STUDY
1. Abraham, W.M., Delehunt, J.C., Yerger, L. , Marchete, B. and
Oliver, W., Jr. Changes in airway permeability and responsiveness
after exposure to ozone. Environ. Res. 34; 110-19, 1984.
2. Amoruso, M.A., Witz, G. and Goldstein, B.D. Decreased
superoxide anion radical production by rat alveolar macrophages
following inhalation of ozone or nitrogen dioxide. Life Sci. 28;
2215-21, 1981.
3. Amoruso, M.A. and Goldstein, B.D. Effect of 1, 3, and 6 hour
ozone exposure on alveolar macrophages superoxide production.
Toxicolocrist 8: 197 1988.
4. Aranyi, C., Vana, S.C., Thomas, P.T., Bradof, J.N., Fenters,
J.D., Graham, J.A. and Miller, F.J. Effects of subchronic exposure
to a mixture of 03 , S02, and (NH4)2S04 on host defenses of mice.
J. Toxicol. Environ. Health 12: 55-71, 1983.
5. Barry, B.E., Miller, F.J. and Crapo, J.D. Effects of inhalation
of 0.12 and 0.5 ppm ozone on the proximal alveolar region of
juvenile and adult rats. Lab. Invest. 53; 692-704 (1985).
6. Barry, B.E., Mercer, R.R., Miller, F.J. and Crapo, J.D.
Effects of inhalation of 0.25 ppm ozone on the terminal bronchioles
of juvenile and adult rats. Exp. Luna. Res. 14: 225-45 (1988).
7. Bates, D.V. Ozone- myth and reality. Environ. Res. 50; 230-7
(1989) .
8. Bohning, D.E., Albert, R.E., Lippmann, M. and Foster, W.M.
Tracheobronchial particle deposition and clearance. A study of the
effects of cigarette smoking in monozygotic twins. Arch. Environ.
Health 30: 457-62 (1975).
9. tASAC. Review of the NAAQS for Ozone. EPA-SAB-CASAC-89-019,
U.S. EPA, Washington, DC (1989).
168
-------
10. Chan, T.L. and Lippmann, M. Experimental measurements and
empirical modelling of the regional deposition of inhaled particles
in humans. Am. Ind. Hva. Assoc. J. 41; 447-59 (1980).
11. Coffin, D.L., Blommer, E.J., Gardner, D.E. and Holzman, R.
Effect of air pollution on alteration of susceptibility to
pulmonary infection. In: Proceedings of the Third Annual Conference
on Atmospheric Contamination in Confined Spaces. Wright-Patterson
Air Force Base, OH: Aerospace Medical Research Laboratories, pp.
71-80, Report No. AMRL-TR-67-20O. Available from: NTIS,
Springfield, VA, AD-8350O8 (1967).
12. Coffin, D.L., Gardner, D.E., Holzman, R.S. and Wolock, F.J.
Influence of ozone on pulmonary cells. Arch. Environ. Health 16:
633-36 (1968).
13. Coffin, D.L. and Gardner D.E. Interaction of biological agents
and chemical air pollutants. Ann. Occup. Hyg. 15; 219-35 (1972).
14. DeLucia, A.J. and Adams, W.C. Effects of 03 inhalation during
exercise on pulmonary function and blood biochemistry. J. Appl.
PhVSJQl. 43: 75-81 (1977).
15. Detels, R., Tashkin, D.P., Sayre, J.W., Rokaw, S.N., Coulson,
A.H., Massey, F.J. and Wegman, D.H. The UCLA population studies
of chronic obstructive respiratory disease. 9. Lung function
changes associated with chronic exposure to photochemical oxidants;
a cohort study among never-smokers. Chest 92; 594-603 (1987).
16. Dowell, A.R., Lohrbauer, L.A., Hurst, D. and Lee, S.D. Rabbit
alveolar macrophage damage caused by in vivo ozone inhalation.
Arch. Environ. Health 21; 121-27 (1970).
17. Drechsler-Parks, D.M., Bedi, J.F. and Horvath, S.M. Pulmonary
function responses of older men and women to ozone exposure.
Geront. 22: 91-101 (1987).
18. Driscoll, K.E., Vollmuth, T.A. and Schlesinger, R.B. Early
alveolar clearance of particles in rabbits undergoing acute and
subchronic exposure to ozone. Fund. Appl. Toxicol. 7: 264-71
(1986).
169
-------
19. Driscoll, K.E., Vollmuth, T.A. and Schlesinger, R.B. Acute and
subchronic ozone inhalation in the rabbit: Response of alveolar
macrophages. J. Toxicol. Environ. Health 21; 27-43 (1987) .
20. Ehrlich, R. , Findlay, J.C., renters, J.D. and Gardner, D.E.
Health effects of short-term inhalation of nitrogen dioxide and
ozone mixtures. Environ. Res. 14; 223-31 (1977) .
21. Ehrlich, R., Findlay, J.C. and Gardner, D.E. Effects of
repeated exposures to peak concentrations of nitrogen dioxide and
ozone on resistance to streptococcal pneumonia. J. Toxicol.
Environ. Health 5: 631-42 (1979).
22. Ehrlich, R. Interaction between environmental pollutants and
respiratory infections. Environ. Health Perspec. 35: 89-100
(1980).
23. Farrell, B.P., Kerr, H.D., Kulle, T.J., Sauder, L.R. and
Young, J.L. Adaptation in human subjects to the effects of inhaled
ozone after repeated exposure. Am. Rev. Pis. 119; 725-30 (1979).
24. Folinsbee, L.J., Bedi, J.F. and Horvath, S.M. Respiratory
responses in humans repeatedly exposed to low concentrations of
ozone. Am. Rev. Resp. Pis. 121: 431-39 (1980).
25. Folinsbee, L.J., McDonnell, W.F. and Horstman, D.H. Pulmonary
function and symptom responses after 6.6 hour exposure to 0.12 ppm
ozone with moderate exercise. JAPCA 38: 28-35 (1988).
26. Folinsbee, L.J. and Hazucha, M.J. Persistence of ozone-induced
changes in lung function and airway responsiveness In: Atmospheric
Ozone Research and its Policy Implications. T. Schneider, S.D.
Lee, G.J.R. Wolters, L.D. Grant, Eds., Elsevier, Amsterdam, The
Netherlands, pp. 483-92 (1989).
27. Foster, W.M., Costa, D.L. and Langenback, E.G. Ozone exposure
alters tracheobronchial mucociliary function in humans. J. Appl..
Phvsiol. 63: 996-1002 (1987).
28. Fouke, J.M, Delemos, R.A., McFadden, E.R., Jr. Airway response
to ultra short-term exposure to ozone. Am. Rev. Resp. Pis. 137;
326-30 (1988) .
170
-------
29. Frager, N.B., Phalen, R.F. and Kenoyer, J.L. Adaptation to
ozone in reference to mucociliary clearance. Arch. Environ. Health
34; 51-57 (1979).
30. Gardner, D.E., Miller, F.J., Illing, J.W. and Kirtz, J.M.
Increased infectivity with exposure to ozone and sulfuric acid.
Toxicol. Lett. 1: 59-64 (1977).
31. Goldstein, E., Tyler, W.S., Hoeprich, P.O. and Eagle, C. Ozone
and the antibacterial defense mechanisms of the murine lung. Arch.
Intern. Med. 128: 1099-1102 (1971).
32. Goldstein, E., Tyler, W.S., Hoeprich, P.O. and Eagle, C.
Adverse influence of ozone on pulmonary bactericidal activity of
murine lungs. Nature 229; 262-63 (1971).
33. Goldstein, B.D., Hamburger, S.J., Falk, G.W. and Amoruso, M.A.
Effect of ozone and nitrogen dioxide on the agglutination of rat
alveolar macrophages by concanavalin A. Life Sci. 21; 1637-44
(1977).
34. Grose, E.G., Gardner, D.E. and Miller, F.J. Response of
ciliated epithelium to ozone and sulfuric acid. Environ. Res. 22:
377-85 (1980).
35. Grose, E.G., Richards, J.H., Illing, J.W., Miller, F.J.,
Davies, D.W., Graham, J.A. and Gardner, D.E. Pulmonary host
defense responses to inhalation of sulfuric acid and ozone. J.
Toxicol. Environ. Health 10: 351-62 (1980).
36. Grose, E.G., Stevens, M.A., Hatch, G.E., Jaskot, R.H. ,
Belgrade, M.J.K., Stead, A.G., Costa, D.L. and Graham, J.A. The
impact of a 12 month exposure to a diurnal pattern of ozone on
pulmonary function, antioxidant biochemistry and immunology. In:
Atmospheric Ozone Research and its tic Implications. T. Schneider,
S.D. Lee, G.J.R. Wolters, L.D. Grant, Eds., Elsevier, Amsterdam,
The Netherlands, pp. 535-44 (1989).
37. Hackney, J.D., Linn, W.S., Mohler, J.G. and Collier, C.R.
Adaptation to short-term respiratory effects of ozone in men
exposed repeatedly. J. Phvsiol. 43; 82-85 (1977).
171
-------
38. Harder, S.D., Harris, D.T., House, D.E., and Koren, H.S.
(1990) . Inhibition of human natural killer cell activity following
in vitro exposure to ozone. Inhalation Toxicology 2.: 161-173
39. Hayes, S.R., Moezzi, M., Wallsten, T.S. and Winkler, R.L. An
analysis of symptom and lung function data from several human
controlled ozone exposure studies (Draft Final Report), San Rafael,
CA: Systems Applications, Inc. (1987).
40. Hazucha, M.J. Relationship between ozone exposure and
pulmonary function changes. J. Appl. Physiol. 62; 1671-80 (1987) .
41. Hazucha, M.J., Bates, D.V. and Bromberg, P.A. Mechanism of
action of ozone on the human lung. J. Appl. Physiol. 67: 1535-41
(1989) .
42. Holguin, A.H., Buffler, P.A., Contant, C.F., Jr., Stock, T.N.,
Kotchmar, D., Hsi, B.P., Jenkins, D.E., Gehan, B.M., Noel, L.M. and
Mei, M. The effects of ozone on asthmatics in the Houston area.
Trans. APCA TR-4; 262-80 (1985).
43. Horstman, D., McDonnell, W., Abdul-Salaam, S., Folinsbee, L.
and Ives, P. Changes in pulmonary function and airway reactivity
due to prolonged exposure to near ambient ozone levels. In:
Atmospheric Ozone Research and its tic Implications. T. Schneider,
S.D. Lee, G.J.R. Wolters, L.D. Grant, Eds., Elsevier, Amsterdam,
The Netherlands, pp. 755-62 (1989).
44. Horvath, S.M., Gliner, J.A. and Folinsbee, L.J. Adaptation to
ozone: Duration of effect. Am. Rev. Resp. Pis. 123; 496-99 (1981).
45. Huang, Y., Chang, L.Y., Miller, F.J., Graham, J.A., Ospital,
J.J. and Crapo, J.D. Lung injury caused by ambient levels of
oxidant air pollutants: Extrapolation from animal to man. Am. J.
Aerosol Med. 1: 180-83 (1988).
46. Hurst, D.J. and Coffin, D.L. Ozone effect on lysosomal
hydrolases of alveolar macrophages in vitro. Arch. Intern. Med.
127: 1059-63 (1971).
172
-------
47. Hurst, D.J., Gardner, D.E. and Coffin, D.L. Effect of ozone on
acid hydrolases of the pulmonary alveolar macrophage. J.
Reticuloendothel. Soc. 8: 288-300 (1970).
48. Hyde, D.M., Plopper, C.C., Harkema, J.R., St. George, J.A.,
Tyler, W.S. and Dungworth, D.L. Ozone-induced structural changes
in monkey respiratory system. In: Atmospheric Ozone Research and
its Policy Implications. T. Schneider, S.D. Lee, G.J.R. Wolters,
L.D. Grant, Eds., Elsevier, Amsterdam, The Netherlands, pp. 535-44
(1989) .
49. Illing, J.W., Miller, F.J. and Gardner, D.E. Decreased
resistance to infection in exercised mice exposed to N02 and 03 J
Toxicol. Environ. Health 6: 843-51 (1980).
50. Kagawa, J. Exposure-effect relationship of selected pulmonary
function measurements in subjects exposed to ozone. Int. Arch.
OCCUP. Environ. Health 53; 345-58 (1984).
51. Kenoyer, J.L., Phalen, R.F. and Davis, J.R. Particle clearance
from the respiratory tract as a test of toxicity: Effect of ozone
on short and long term clearance. Exp. Lung. Res. 2: 111-20
(1981).
52. Kinney, P.L., Ware, J.H. and Spengler, J.D. A critical
evaluation of acute ozone epidemiology results. Arch. Environ.
Health 43! 168-73 (1988).
53. Knudson, R.J., Lebowitz, M.D., Holberg, C.J. and Burrows, B.
Changes in the normal maximal expiratory flow-volume curve with
growth and aging. Am. Rev. Resp. Pis. 127; 725-34 (1983).
54. Koenig, J.Q., Covert, D.S., Marshall, S.C., van Belle, G. and
Pierson, W.E. The effects of ozone and nitrogen dioxide on
pulmonary function in healthy and in asthmatic adolescents. Am.
Rev. Respir. Pis. 136; 1152-57 (1987).
55. Keren, H.S., Devlin, R.B., Graham, D.E., Mann, R., McGee, M.P.,
Horstman, D.E., Kozumbo, W.J., Becker, S., House, D.E., McDonnell,
W.F. and Bromberg, P.A. Ozone-induced inflammation in the lower
airways of human subjects. Am. Rev. Respir. Pis. 139; 407 (1989).
173
-------
56. Keren, H., Graham, D., Becker, S. and Devlin, R. Modulation of
the inflammatory response in the human lung exposed to ambient
levels of ozone. In: Atmospheric Ozone Research and its Policy
Implications. T. Schneider, S.D. Lee, G.J.R. Wolters, L.b. Grant,
Eds., Elsevier, Amsterdam, The Netherlands, pp. 745-53 (1989).
57. Kulle, T.J., Sauder, L.R., Kerr, H.D., Farrell, B.P., Bermel,
M.S. and Smith, D.M. Duration of pulmonary function adaptation to
ozone in humans. Am. Ind. Hyg. Assoc. J. 43; 832-37 (1982).
58. Last, J.A. Effects of inhaled acids on lung biochemistry.
Environ. Health Persoect. 79; 115 (1989).
59. Leikauf, G.D., Driscoll, K.E. and Wey, H.E. Ozone-induced
augmentation of eicosanoid metabolism in epithelial cells from
bovine trachea. Am. Rev. Resp. Pis. 137; 435-42 (1988).
60. Linn, W.S., Jones, M.P., Bachmeyer, E.A., Spier, C.E., Mazur,
S.F., Avol, E.L. and Hackney, J.D. Short-term respiratory effects
of polluted air: A laboratory study of volunteers in a high-oxidant
community. Am. Rev. Resp. Pis. 121; 243-52 (1980).
61. Linn, W.S., Shamoo, D.A., Venet, T.G., Spier, C.E., Valencia,
L.M., Anzar, U.T. and Hackney, J.D. Response to ozone in
volunteers with chronic obstructive pulmonary disease. Arch.
Environ. Health 38; 278-83 (1983).
62. Linn, W.S., Avol, E.L., Shamoo, D.A., Peng, R.C., Valencia,
L.M., Little, D.E. and Hackney, J.D. Repeated laboratory ozone
exposures of volunteer Los Angeles residents: An apparent seasonal
variation in response. Toxicol. Indust. Health 4: 505-20 (1988).
63. Lippmann, M. Health significance of pulmonary function tests.
JAPCA 38; 881-87 (1988).
64. Lippmann, M., Lioy, P.J., Leikauf, G., Green, K.B., Baxter, D.,
Morandi, M., Pasternack, B., Fife, D. and Speizer, F.E. Effects of
ozone on the pulmonary function of children. Adv. in Modern
Environ. Toxicol. 5: 423-46 (1983).
65. Lippmann, M. Health effects of ozone: A critical review.
JAPCA 39: 672-695 (1989).
174
-------
66. McAllen, S.J., Chiu, S.P., Phalen, R.F. and Rasmussen, R.E.
Effect of in vivo ozone exposure on in vitro pulmonary alveolar
macrophage mobility. J. Toxicol. Environ. Health 7: 373-81 (1981).
67. McDonnell, W.F., Horstman, D.H. , Hazucha, M.J. , Seal, E., HaaJc,
E.D., Salaam, S.A. and House, D.E. Pulmonary effects of ozone
exposure during exercise: Dose-response characteristics. J. Appl.
Phvsiol. 54; 1345-52 (1983).
68. McDonnell, W.F., Chapman, R.S., III, Leigh, M.U., Strope, G.L.
and Collier, A.M. Respiratory responses of vigorously exercising
children to 0.12 ppm ozone exposure. Am. Rev. Resp. Pis. 132; 875-
79 (1985).
69. McDonnell, W.F., Horstman, D.H., Abdul-Salaam, S. and House,
D.E. Reproducibility of individual responses to ozone exposure.
Am. Rev.Resp. Pis. 131; 36-40 (1985).
70. McDonnell, W.F., Horstman, D.H., Abdul-Salaam, S., Raggio, L.J.
and Green, J.A. The respiratory responses of subjects with
allergic rhinitis to ozone exposure and their relationship to
nonspecific airway reactivity. Toxicol. Indust. Health 3: 507-17
(1987) .
71. Miller, F.J., Illing, J.W. and Gardner, D.E. Effect of urban
ozone levels on laboratory-induced respiratory infections.
Toxicol. Lett. 2; 163-69 (1978).
72. Miller, F.J., Menzel, D.B. and Coffin, D.L. Similarity between
man and laboratory animals in regional pulmonary deposition of
ozone. Environ. Res. 17; 84-101 (1978).
73. National Research Council. Complex Mixtures - Methods for In
Vivo Toxicity Testing. National Academy Press, Washington, DC
(1988) .
74. Overton, J.H. and Miller, F.J. Modelling ozone absorption in
lower respiratory tract. Preprint 87-99.4. 1987 Annual Meeting of
Air Pollut. Contr. Assoc., New York (June 1987).
175
-------
75. Phalen, R.F., Kenoyer, J.L., Crocker, T.T. and McClure, T.R.
Effects of sulfate aerosols in combination with ozone on
elimination of tracer particles inhaled by rats. J. Toxicol.
Environ. Health 6: 797-810 (1980).
76. Phipps, R.J., Denas, S.M., Siekzak, M.U. and Wanner, A.
Effects of 0.5 ppm ozone on glycoprotein secretion, ion and water
fluxes in sheep trachea. J. Appl. Phvsiol. 60; 918-27 (1986).
77. Pinkerton, K.E., Brody, A.R., Miller, F.J. and Crapo, J.D.
Exposure to a simulated "ambient" pattern of ozone results in
significant pulmonary retention of asbestos fibers. Am. Rev. Resp.
Pis. 137; A166 (1988).
78. Raub, J.A., Miller, F.J. and Graham, J.A. Effects of low level
ozone exposure on pulmonary function in adult and neonatal rats.
Adv. Mod. Environ. Toxicol. 5: 363-367 (1983).
79. Reisenauer, C.S., Koenig, J.Q., McManus, M.S., Smith, M.S.,
Kusic, G. and Pierson, W.E. Pulmonary response to ozone exposures
in healthy individuals aged 55 years or greater. JAPCA 38; 51-55
(1988) .
80. Rombout, P.J.A., Lioy, P.J. and Goldstein, B.D. Rationale for
an eight-hour ozone standard. JAPCA 36; 913-17 (1986).
81. Rombout, P.J.A. U.S.-Dutch collaboration. In: Atmospheric
Ozone Research and its Policy Implications. T. Schneider, S.D. Lee,
G.J.R. Wolters, L.D. Grant, Eds., Elsevier, Amsterdam, The
Netherlands, pp. 710-10 (1989).
82. Schlesinger, R.B. and Driscoll, K.E. Mucociliary clearance
from the lungs of rabbits following single and intermittent
exposures to ozone.J.Toxicol. Environ. Health 20: 125-34 (1987).
83. Schwartz, J. Lung function and chronic exposure to air
pollution: A cross-sectional analysis of NHANES II. Environ. Res.
50; 309-21 (1989).
84. Shephard, R.J., Urch, B., Silverman, F. and Corey, P.N.
Interaction of ozone and cigarette smoke exposure. Environ. Res.
31: 125-37 (1983).
176
-------
85. Solic, J.J., Hazucha, M.J. and Bromberg, P.A. The acute
effects of 0.2 ppm ozone in patients with chronic obstructive
pulmonary disease. Am. Rev. Resp. Pis. 125: 664-69 (1982).
86. Spektor, D.M., Lippmann, M., Lioy, P.J., Thurston, G.D., Citak,
K., James, D.J., Bock, N., Speizer, F.E. and Hayes, C. Effects of
ambient ozone on respiratory function in active normal children.
Am. Rev. Resp. Pis. 137: 313-20 (1988a).
87. Spektor, D.M., Lippmann, M., Thurston, G.D., Lioy, P.J.,
Stecko, J., O'Connor, G., Garshick, E., Speizer, F.E. and Hayes, C.
Effects of ambient ozone on respiratory function in healthy adults
exercising outdoors. Am. Rev. Resp. Pis. 138: 821-28 (1988b).
88. Spengler, J.P., Keeler, G.J., Koutrakis, P., Ryan, P.B.,
Raizenne, M. and Franklin, C.A. Exposures to acidic aerosols.
Environ. Health Perspect. (in press).
89. Steinberg, J.J., Gleeson, J.L., and Gil, P. (1990) The
pathobiology of ozone-induced damage. Arch. Environ. Health 45;
80-87
90. Tepper, J.S., Costa, P.L., Weber, M.F., Wiester, M.J., Hatch,
G.E. and Selgrade, M.J.K. Functional and organic changes in a rat
model of ozone adaptation. Amer. Rev. Resp. Pis. 135; A283 (1987) .
91. Tyler, W.S., Tyler, N.K., Last, J.A., Gillespie, M.J. and
Barstow, T.J. Comparison of daily and seasonal exposures of young
monkeys to ozone. Toxicology 50; 131-44 (1988).
92. Wayne, W.S., Wehrle, P.F. and Carroll, R.E. Oxidant air
pollution and athletic performance. J. Am. Med. Assoc. 199; 901-
904 (1967).
93. Wayne, W.S., Wehrle, P.F. and Carroll, R.E. Oxidant air
pollution and school absenteeism. Arch. Environ. Health 19; 315-22
(1967).
94. Weschler, C.J., Shields, H.C. and Naik, P.V. Indoor ozone
exposures. JAPCA 39; 1562-8 (1989).
95. Whittemore, A.S. and Korn, E.L. Asthma and air pollution in
the Los Angeles area. Am. J. Public Health 70: 687-96 (1980).
177
-------
96. Witschi, H. Ozone, nitrogen dioxide and lung cancer: A review
of some recent issues and problems. Toxlcol. 48j 1-20 (1988).
97. Witz, G., Amoruso, M.A. and Goldstein, B.D. Effect of ozone on
alveolar macrophage function: Membrane dynamic properties. Adv.
Mod. Environ. Toxicol. 5: 263-72 (1983).
98. U.S E.P.A. Air Quality Criteria for Ozone and Other
Photochemical Oxidants. Vol. II, EPA/600/8-84/0206F, ECAO.
Research Triangle Park, NC (August 1986).
Section 8.1.2—Appendix
RADON CASE STUDY
1. Alter, H.W. and Oswald, R.A. Nationwide distribution of indoor
radon measurements: A preliminary data base. J. Air Pollution
Control Assoc. 3_7:227, 1988.
2. Andelmah, J.B. Science of Total Environ.f 47:443-460, 1985)
3. Andelman, J.B. Total Exposure to volatile organic compounds in
potable water. In: Significance and Treatment of Volatile Organic
Compounds in Water Supplies. (N. Ram, R. Christman, and K. Cantor,
Eds.) Lewis Publishers, Inc., Ann Arbor, MI, Ch. 20, pp. 485-504,
1990.
4. Bair, W.J. Experimental carcinogenesis in the respiratory
tract. In: Radiation Carcinoaenesis. edited by A.C. Upton, R.E.
Albert, F.J. Burns, and R.E. Shore. Elsevier, New York, 1986, pp.
151-167.
5. Chameaud, J., Masse, R., LaFuma, J. Influence of radon daughter
exposure at low doses on occurrence of lung cancer in rats. In:
Radiation Protection Dosimetrv; Indoor Exposure to Natural
Radiation and Associated Risk Assessment. Vol. 7, edited by G.F.
Clemente, H. Eriskat, M.C. O'Riordan, and J. Sinnaeve. Proceedings
of an International Seminar held at Anacapri, Italy, October 3-5,
1983, 1984.
178
-------
6. Cohen, B.L. Radon Levels by States and Counties. Report of the
Radon Project data through February, 1988. Radon Project,
Pittsburgh, 1988.
7. Cothern, C.R. Indoor Air Radon, in "Reviews of Environmental
Contamination and Toxicology," Springer-Verlag, New York, 1990.
8. Cross, F.T. Evidence of lung cancer risk from animal studies.
In: Radon. Proceedings No. 10. Proceedings of the Twenty-fourth
Annual Meeting of the National Council on Radiation Protection and
Measurements, March 30-31, 1988. National Council on Radiation
Protection and Measurements, Bethesda, Maryland, 1989, pp. 129-139.
9. Cross. F.T., Busch, R.H., Buschbom, R.L., Dagle, G.E., Filipy,
R.E., Loscutoff, S.M., Mihalko, P.J., and Palmer, R.F. Inhalation
hazards to uranium miners. In: Pacific Northwest Laboratory
Annual Report for 1981 to the DoE Office of Energy Research. Part
1. Biomedical Sciences, Battelle, Pacific Northwest Laboratories,
Richland, Washington, 1982.
10. Environmental Protection Agency (EPA). A Citizens Guide to
Radon. U.S. Environmental Protection Agency, Washington, 1986.
11. Environmental Protection Agency (EPA). Draft Environmental
Impact Statement for Proposed NESHAPS for Radionuclides.
Background Information Document Volume 1. U.E. Environmental
Protection Agency, Washington, D.C., 1989.
12. George, A.C. and Hinchcliffe, L.E. Measurements of radon
concentrations in residential buildings in the eastern United
States, in Radon and Its Daughter products. P. Hopke, Ed., American
Chemical Society, Washington DC, 1987, Ch 4, pp 42-69.
13. Harley, N.H. and Cohen, B.S. Updating radon daughter
dosimetry. In: American Chemical Society Symposium on Radon and
Its Decay Products. edited by P.K. Hope. American Chemical
Society, Washington, 1987, p. 419.
14. Hess, C.T. The occurrence of radioactivity in public water
supplies of the United States, Health Phvsics. 1985, 48:553.
179
-------
15. Howe, G.R., Nair, R.D., Newcombe, H.B., et al. Lung cancer
mortality (1950-1980) in relation to radon daughter exposure in a
cohort of workers at the Eldorado Beaverlodge uranium mine. JNCI
71:357-362, 1986.
16. International Commission on Radiological Protection (ICRP).
Lung Cancer Risk from Indoor Exposures to Radon Daughters.
International Commission on Radiological Protection Publication No.
50. Pergamon Press, New York, 1987.
17. Jacobi, W. Carcinogenic effects of radiation on the human
respiratory tract. In: Radiation Carcinoaenesis. edited by A.C.
Upton, R.E. Albert, F.J. Burns, and R.E. Shore. Elsevier, New
York, 1986, pp. 261-278..
18. Longtin, J.P. Occurrence of radon, radium, and uranium in
groundwater. J. American Water Works Assoc.. 1988, 80:84-93.
19. Morrison, H.I., Semenciw, R.M., Mao, Y. , and Wigle, D.T.
Cancer mortality among a group of fluorspar miners exposed to radon
progeny. Am. J. Epidemiol. 128:1266. 1988.
20. Muller, J., Wheeler, W.C., Gentlemean, J.F., Suranyi, G. , and
Kusiak, R.A. Study of mortality of Ontario miners. In:
Proceedings of the International Conference on Occupational Safety
in Mining, edited by H. Stocker. Canadian Nuclear Association,
Toronto, 1985, p. 335.
21. National Academy of Sciences-National Research Council
(NAS/BEIR). The Effects on Populations of Exposure to Low Levels
of Ionizing Radiation. (BEIR III). National Academy Press,
Washington, 1980.
22. National Academy of Sciences-National Research Council
(NAS/BEIR). Health Risks of Radon and Other Internally Deposited
Alpha-Emitters. (BEIR IV). National Academy Press, Washington,
1988.
23. National Academy of Sciences-National Research council
(NAS/BEIR). Health Effects of Exposure to Low Levels of Ionizing
Radiation. (BEIR V). National Academy Press, Washington, 1989.
180
-------
24. National Council on Radiation Protection and Measurements
(NCRP). Evaluation of Occupational and Environmental Exposures to
Radon and Radon Daughters in the United States. National Council
on Radiation Protection and Measurements Report No. 78. National
Council on Radiation Protection and Measurements, Bethesda, I984a.
25. National Council on Radiation Protection and Measurements
(NCRP). Exposures from the Uranium Series with Emphasis on Radon
and Its Daughters. National Council on Radiation Protection and
Measurements Report No. 77. National Council on Radiation
Protection and Measurements, Bethesda, 1984b.
26. National Council on Radiation Protection and Measurements
(NCRP). Ionizing Radiation Exposure of the Population of the
United States. National Council on Radiation Protection and
Measurements Report No. 93. National Council on Radiation
Protection and Measurements, Bethesda, 1987.
27. National Council on Radiation Protection and Measurements
(NCRP). Measurement of Radon and Radon Daughters in Air. National
Council on Radiation Protection and Measurements Report No. 97.
National Council on Radiation Protection and Measurements,
Bethesda, 1989.
28. Nero, A.V., Schwehr, M.B., Nazaroff, W.W., and Revzan, K.L.
Distribution of airborne radon-222 concentrations in U.S. homes.
Science 231:992-996, 1986.
29. Nichols, G.P. and Stearns, R.J. In: Radon, Proceedings No.
10. Proceedings of the Twenty-fourth Annual Meeting of the
National Council on Radiation Protection and Measurements, March
30-31, 1988. National Council on Radiation Protection and
Measurements, Bethesda, Maryland, 1989, pp. 275-282.
30. Prichard, H.M. and Gesell, T.F. An estimate of population
exposures due to radon in public water supplies in the area of
Houston, Texas, Health Physics. 1981, 41:599-606.
31. Puskin, J.S. and Nelson, C.B. EPA's perspective on risks from
residential radon exposure. JAPCA 3_9:915-920, 1989.
181
-------
32. Puskin, J.S. and Yang, Y. A retrospective look at Rn-induced
lung cancer mortality from the viewpoint of a relative risk model.
Health. Phvs. j>4:635, 1988.
33. Radford, E.P. and Renard, K.G.S. Lung cancer in Swedish iron
miners exposed to low doses of radon daughters. New Enal. J. Med.
: 1485-1494, 1984.
34. Reilly, M.A. The Pennsylvania radon program. In: Radon.
Proceedings No. 10. Proceedings of the Twenty-fourth Annual
Meeting of the National Council on Radiation Protection and
Measurements, March 30-31, 1988. National Council on Radiation
Protection and Measurements, Bethesda, Maryland, 1989, pp. 270-274.
35. Robkin, M.A. Indoor Radon levels, in Indoor Radon and Its
Hazards. Eds. D. Bodansky, M.A. Robkin, D.R. Stadler, U. of
Washington Press, Seattle, 1987, Ch 5, pp 50-66.
36. Roessler, C.E. The Florida radon program. In: Radon.
Proceedings No. 10. Proceedings of the Twenty-fourth Annual
Meeting of the National Council on Radiation Protection and
Measurements, March 30-31, 1988. National Council on Radiation
Protection and Measurements, Bethesda, Maryland, 1989, pp. 283-291.
37. Sevc, J., Kunz, E., Tomasek, L. , Placek, V., and Horacek, J.
Cancer in man after exposure to Rn daughters. Health Phys. 54:27-
46, 1988.
38. Thomas, D.C., McNeill, K.G., Dougherty, C. Estimates of
lifetime lung cancer risk resulting from Rn progeny exposures.
Health Phys. 4J9:825, 1985.
39. United Nations Scientific Committee on the Effects of Atomic
Radiation (UNSCEAR) . Sources and Effects of Ionizing Radiation.
Report to the General Assembly, with annexes. Thirty-second
Session. Supplement No. 40. United Nations, New York, 1977.
182
-------
EPASAB-EC-90-021B
EPASAB-EC-90-021B
Thp pppnrt. nf Thp Human Health
AUTH05ubcotnmittee. Relative Risk
Rpdurt.ini n Project..
TIE£ducing Risk Appendix b
DATE
LOANED
^•/B> -f /
BORROWER'S NAME
^£,~~cc. /^J
DATE
RETURNED
\!£&V
6NVIRONMENTM
PROTECTION
AGENCY
DALLAS. TEXA8
-------
(D
(0
------- |