THE AIR TOXICS PROBLEM IN THE UNITED STATES:
AN ANALYSIS OF CA.1CEK RISKS FOR SELECTED POLLUTANTS
U.S. Environmental protection Agency
Office of Air and Radiation
Office of Policy, PI a n n i ny and Evaluation
May 1985
Elai ne Haemi segger
Alan Jo nes
Bern St ei ge rwa1d
Vivian Thomson
-------
PREFACE
This report on the air toxics problem in the United States
is a final version of a September 1984 U.S. Environmental Protec-
tion Agency draft document entitled "The Magnitude and Nature of
the Air Toxics Problem in the United States." Simultaneously
with release of the draft report, EPA solicited comments on the
analysis from a peer review panel made up of non-EPA experts in
fields such as toxicology, air monitoring, and air pollution
control. In response to the panel's comments, as well as to
unsolicited comments received from several organizations, the
authors have substantially revised the Executive Summary and have
made changes in other parts of the report. The revised version
more clearly delineates the limitations and caveats of the analysis
Also, the title of the report has been revised to recognize the
limited scope of the analysis.
In addition, certain of the risk estimates have been changed
as new information has become available. In particular, the risk
estimates for nickel and ethylene dichloride have been revised
substantially downward and methyl chloroform has been dropped
from the analysis. The report has not been altered where the
Agency is- currently considering, but has made no final decision
regarding, changes in certain of the data used in this analysis
(e.g., potency information) or in the regulatory status of certain
chemicals examined.
Several additions were made to the report. A new section
has been added that delineates current activities within EPA
resulting from the report, including a brief discussion of the
national strategy for air toxics. Air quality data have been
evaluated for 1970 and the estimated risks compared to those
based on 1980 data to provide a more quantitative estimate of
progress under programs for criteria pollutants. Finally, data
from personal exposure monitoring have been converted to aggregate
national risk for several compounds to enhance comparison of the
indoor/outdoor air toxics problem.
-------
ACKNOWLEDGMENTS
Many individuals and organizations within EPA participated in
this study. The report is based primarily on a series of detailed
analyses and reviews done specifically for the study, often in
cooperation with private companies under contract to EPA. These
are listed below. We thank the authors for their efforts and for
contributing so much to this analysis.
Joe Bufalini, Bruce Gay, Basil Dimitriades. "Production of
Hazardous Pollutants through Atmospheric Transformations."
June 1984.
Elaine Haemisegger. "Hazardous Air Pollutants: An Exposure and
Risk Assessment for 35 Counties." September 1984. (Contractors:
Versar; American Management Systems, Inc.)
Jim Hardin. "Issue Paper--National Air Toxics Problem:
Radionuclides." August 1984. Revised verbally January 1985.
Bill Hunt, Bob Faoro, Tom Curran, Jena Muntz. "Estimated
Cancer Incidence Rates for Selected Toxic Air Pollutants
Using Ambient Air Pollution Data." July 1984. Revised March
1985. (Contractor: PEI)
Tom Lahre. "Characterization of Available Nationwide Air Toxics
Emissions Data." June 1984. (Contractor: Radian Corp.)
Nancy Pate. "Review of the Clement Associates Report on Evidence
for Cancer Associated with Air Pollution." June 1984.
Bob Schell. "Estimation of the Public Health Risks Associated
with Exposure to Ambient Concentration of 87 Substances."
July 1984. Revised February 1985.
Bob Schell. "Definition of the Air Toxics Problem at the State/Local
Level." June 1984. (Contractor: Radian Corp.)
Vivian Thomson. "Indoor Air Pollution: Ramifications for Assessing
the Magni-tude and Nature of the Air Toxics Problem in the United
States." July 1984.
-------
ACKNOWLEDGMENTS (Continued)
Donn Viviani, Doreen Sterling, Robert Kayser. "Acceptable
Risk Levels and Federal Regulations: A Comparison of National
Emission Standards for Hazardous Air Pollutants (NESHAP) with
Other Federal Standards Based on Quantitative Risk Asessment
(QRA)." May 1984.
Others within EPA provided assistance during the study.
We especially wish to recognize Carol Cox, Alan Ehrlich, Greg
Glahn, Joan O'Callaghan and Sue Perl in.
-------
TABLE OF CONTENTS
Page
Executive Summary i
I. Introduction 1
II. Scope of Study 4
A. Reasons for Assessing Only Cancer Risks 5
III. Methods Used to Estimate Cancer Risks 8
A. Estimating the Carcinogenic Potency of 8
Pol 1 utants
B. Estimating Exposure to Pollutants 10
C. Using Monitoring Data to Estimate Ambient 11
Concentrati ons
D. Using Emission Estimates and Dispersion 13
Modeling to Estimate Ambient Concentrations
IV. Magnitude of the Ambient Air Toxics Problem 16
A. I ntroduct ion 16
B. Summaries of Individual Analyses 17
1. Survey of State and Local Agencies, 17
Canada, and Europe
2. Evaluation of Cancer Associated with 20
Air Pollution Using Epidemiol ogi cal
Studi es
3. NESHAP Study 25
4. 35-County Study 31
5. Ambient Air Quality Study 38
6. Other Pollutants, Sources and Pathways 47
C. Summary of the Magnitude of the Air Toxics 68
Problem
D. Perspective and Context: Other Cancer Risks 73
-------
TABLE OF CONTENTS
conti nued
Page
V. Nature of the Air Toxics Problem 77
A. Pollutants 77
B. Sources 78
C. Geographic Variability 82
D. Indirect Control of Air Toxics 87
VI. Adequacy of Data Bases 89
VII. Conclusions 94
VIII. Current Activities 99
Attachment A - Summary Table: Pollutants Examined,
Upper-Bound Risk Values, Preliminary Approximations
of Incidence and Maximum Lifetime Risk
-------
LIST OF TABLES
Number Page
1 NESHAP Study: Preliminary Approximation 27
of Annual Incidence and Maximum Lifetime Risk
35-County Study: Preliminary Approximation 33
of Annual Incidence
Ambient Air Quality Study: Preliminary 40
Approximation of Annual Incidence
Ambient Air Quality Study: Preliminary 44
Approximation of Individual Lifetime Risks
Ambient Air Quality Study: Preliminary 46
Approximation of Additive Lifetime Risks
Estimates of Incidence and Individual Risk 50
Due to Radi onucl i des Emitted to Air
Preliminary Estimates of Incidence and 62
Individual Risks Associated with Air Releases
from One Treatment, Storage, and Disposal
Facility (TSDF)
Summary Table: Preliminary Approximation of 69
Annual Incidence Estimates per Million Popula-
tion from the NESHAP Study, the Ambient Air
Quality Study, and the 35-County Study
Perspective and Context: Statistics on Cancer 74
Risks
10 Sources of Selected Compounds Examined in 79
This Study
11 Percent of Incidence Associated With Point and 83
Area Sources Based on 35-County Study
12 Comparison of Measured Air Quality for Selected 84
Cities and Pollutants
13 Comparison of Sources of Risk in Several Counties 86
Selected from 35-County Study
-------
EXECUTIVE SUMMARY
Goals
This report summarizes the results of a study which will
be used by the U.S. Environmental Protection Agency (EPA) to
provide a basis for consideration of strategies to deal with the
ambient air toxics problem in the United States. The study
attempted to assess the magnitude and nature of the air toxics
problem by developing quantitative estimates of the cancer risks
posed by selected air pollutants and their sources. Four basic
questions were examined:
1. What is the approximate magnitude of the air toxics
problem, as measured by the estimated cancer risks
associated with air pollution?
2. What is the nature of the air toxics problem, that is,
what pollutants and sources appear to cause the problem
and what is their relative importance?
3. Does the problem vary geographically, and if so, in
what ways?
4. Are current air toxics data bases adequate? If not,
what are the significant data gaps?
Context and Limitations
Readers of this report should fully understand the study's
limitations so that its conclusions are interpreted correctly.
The analysis was undertaken to orient EPA to the problem of
airborne carcinogens, to stimulate policy discussions, and to
guide further studies. Despite the fact that quantitative
estimates of risk are presented in this report, the study was not
initiated to support specific regulatory decisions. Instead,
-------
-i i-
its goal was to obtain a quick assessment of the air toxics
problem in the United States, and as such should be regarded as
a "scoping" study only. Only readily available existing data
were used regarding compound potencies, emissions, and ambient
levels; no new data were collected. The only health effect
examined quantitatively was cancer; health data on such effects as
teratogenicity are not extensive enough to permit quantification.
Also, acute health effects related to short-term exposures were not
i ncluded.
Consideration of the limited scope of the study, as well as of
the caveats and assumptions that are discussed in the text of the
report, is an important responsibility of those reading and using
this report. Some of the important caveats to keep in mind are
identified later in the Executive Summary under "Sources of
Uncertai nty."
In summary, the risk estimates presented in this report should
be regarded as only rough approximations of total incidence and
individual risk, and should be used in a relative sense only.
Estimates for individual compounds are highly uncertain and should
be used with extreme caution. The reader is cautioned against
applying these risk estimates to specific geographic locations,
since the relative importance of particular pollutants and sources
varies considerably from one place to the next.
As more data become available, these risk estimates will
undoubtedly change. As such, the portrait of the air toxics
-------
-111-
problem depicted in this study should be regarded as a snapshot,
the form and substance of which will certainly change as new
data become available.
Analytical Methods
Three major analyses were undertaken to estimate cancer
incidence and individual lifetime risks: each analysis included
a separate set of compounds, with considerable overlap. The
Ambient Air Quality Study used ambient data for five metals, ten
organic compounds, and benzo(a)pyrene (BaP) to assess these risks.
Two other analyses — the NESHAP Study and the 35-County Study—used
emission estimates and exposure models to estimate incidence and
maximum individual risks associated with the pollutants selected.
In addition, a "BaP surrogate" approach was used to estimate cancer
incidence associated with products of incomplete combustion (PIC):
a dose-response coefficient relating lung cancer incidence and PIC
was generated from epidemiological studies, and cancer incidence
associated with PIC exposure was estimated by applying this dose-
response coefficient to current ambient BaP levels. Finally,
quantitative risk assessments available from other EPA activities
for asbestos, radionuclides, and gasoline marketing were incorporated
into the study.
Four additional reports were prepared to assist in inter-
preting the results of the study. One report reviewed national
emissions data for over 90 compounds and provided summaries by
source type, geographic location, growth trends, and data
quality. Other papers were prepared on the atmospheric trans-
-------
-i v-
formation of air pollutants, indoor/outdoor relationships for
air toxics, and risk estimates used by other program offices
within EPA in regulating selected toxic substances. Further,
statistics on annual cancer incidence, cancer deaths, and
estimates of the cancer cases associated with other causes
(e.g., diet, smoking, indoor exposures) were compiled. The study
team also analyzed and summarized the information available on
several source categories for which current data are insufficient
to perform a quantitative risk assessment. Finally, contacts
made with all 50 state air pollution agencies, 33 local air
pollution agencies, the Canadian government, and the Commission
of European Communities revealed that virtually no other comprehen-
sive studies are available that quantify estimated cancer incidence
related to air toxics.
The study examined the magnitude and nature of the air
toxics problem using existing data and, where possible, standard
EPA techniques for quantitative risk assessment. We did not
attempt to evaluate the validity of those techniques; rather, we
tried to apply them as comprehensively as possible. For example,
we relied o'n unit risk estimates generated by EPA's Carcinogen
Assessment Group (CA6) a nd »by Clement Associates, many of which
represent plausible upper-bound estimates of unit risk. These
estimates also assume 70 years of continuous exposure to outdoor
ambient levels. However, where appropriate, we point out the
-------
-V-
possible effects of considering non-traditional approaches, such
as examining the risks associated with indoor exposures to air
pol 1 utants.
Concl us ions
Given the scope, limitations, methods, and assumptions dis-
cussed above, the following conclusions may be drawn from this
study :
1. Both point sources (major industrial sources) and
area sources (smaller sources that may be widespread
across a given area, such as solvent usage and motor
vehicles) appear to contribute significantly to the air
toxics problem. Large point sources are associated with
many high individual risks, while area sources appear to
be responsible for the majority of aggregate cancer
i nci dence.
2. While there is considerable uncertainty associated
with the risk estimates for some substances, available
data indicated that the following pollutants may be
important contributors to aggregate cancer incidence
from air toxics: metals, especially chromium and
arsenic; asbestos; products of incomplete combustion;
formaldehyde; benzene; ethylene oxide; gasoline vapors;
and chlorinated organic compounds, such as chloroform;
carbon tetrachloride; perchloroethylene; trichloro-
ethylene; and vinylidene chloride.
3. A wide variety of sources contributes to individual
risk and aggregate incidence from air toxics. These
include: road vehicles; combustion of coal and oil;
woodstoves; metallurgical industries; chemical produc-
tion and manufacturing; gasoline marketing; solvent
usage; and waste oil disposal. As a broad category of
activities, combustion/incineration is probably the
largest single source of risk.
4. For those cities with sufficient data for analysis,
large city-to-city and neighborhood-to-neighborhood
variation in pollutant levels and sources was found.
However, our current air toxics data base is inadequate
to accurately characterize most local air toxics problems
5. Three analyses quantified estimated cancer risks due to
15 to 45 toxic air pollutants (the number of pollutants
-------
-VI -
examined varied with the different analyses). The
estimates from 'these analyses showed a range of 5 to
7.4 cases of cancer per million people per year (1,300
to 1,700 cases annually nationwide) for the pollu-
tants examined. These are not actual predictions
of incidence, but are instead a statistical way to
represent the potential health effects of human exposure
to airborne carcinogens.
The reader is reminded that these estimates are
highly uncertain, and is cautioned that the conver-
gence of the various analyses on a seemingly narrow
range (5 to 7.4 cases per million) is somewhat coinci-
dental, given that estimates for individual compounds
varied widely among the different analyses.
For perspective, estimated nationwide cancer cases
and cancer deaths for 1983 were 850,000 and 440,000,
respectively.
6. Maximum lifetime individual risks of 10~4 (1 in 10,000)
or greater in the vicinity of major point sources were
estimated for 21 pollutants, about half of those that
were studied. Maximum lifetime individual risks of 10-3
(1 in 1,000) or greater were estimated for 13 pollutants.
7. Additive lifetime individual risks in urban areas due
to simultaneous exposure to 10 to 15 pollutants ranged
from 10-3 to 1Q-4. These risks, which were calculated
from monitoring data, did not appear to be related to
specific point sources. Instead, they represent a
portion of the total risks associated with the complex
pollutant mixtures typical of urban ambient air.
8-. Some low-production organic chemicals appeared to
contribute little to aggregate incidence: 21 organic
chemicals were estimated to account for a total of
less than 1.0 cancer cases per year nationwide.
However, this conclusion may be due in part to the lack
of data concerning the emissions and toxicity of these
"exotic" chemicals.
Some of these low-production compounds did appear to
be associated with high individual risks. For example,
the maximum lifetime individual risk for 4,4,-methylene
dianiline was estimated at 1.5 X 10~3.
9. The study indicated that "non-traditional" sources of
air toxics--such as publicly owned treatment works (POTW's)
and hazardous waste treatment, storage and disposal
-------
-VI 1 -
facilities (TSDF1s)--may pose important risks in
some locations. For instance, preliminary findings
suggest that POTW's with industrial indirect dis-
charges may emit volatile organic compounds in excess
of 100 kkg/yr. Individual lifetime risks for a single
compound at one TSDF were estimated as high as 10~5.
10. EPA's criteria pollutant programs appear to have done
more to reduce air toxics levels than have regulatory
actions aimed at specific toxic compounds. An analysis
of 16 pollutants was completed using both monitoring and
emission data in order to evaluate progress made on air
toxics between 1970 and 1980. The estimated cancer inci-
dence rate for these air pollutants in 1980 was less
than half that for 1970, i.e., 6.8 per million per year in
1980, compared to 17.5 per million in 1970. This seems
reasonable considering the diverse array of air toxics
sources, the multipol1utant nature of the problem, and
the relative intensity of EPA's criteria and air toxics
programs.
11. Even after regulations are implemented under Section 112
of the Clean Air Act for benzene and arsenic, these
pollutants still appear to make significant contributions
to aggregate incidence due to air -toxics. This seems
to demonstrate that the base for the air toxics regulatory
programs needs to be broadened to include emissions from
small area sources, such as combustion, road vehicles,
and solvent use.
12. Major weaknesses and gaps characterize air toxics data
bases at the federal, state, and local levels. The
few air toxics emission inventories available generally
show inconsistencies and anomalies, the air quality data
available are often inadequate to develop population
exposure estimates, and few compounds have been tested
adequately for health effects. The data limitations
preclude performing specific comprehensive risk assessments
for most urban areas, for many compounds, and for many
potentially large sources of air toxics risks (such as
incineration, hazardous waste disposal, indoor exposures,
atmospheric transformation, and Superfund sites).
Sources of Uncertainty
Many assumptions and extrapolations are necessary to transform
ambient or modeled levels of air pollutants into exposure estimates.
-------
-viii-
Whether such assumptions introduce a high or low bias into the
results is difficult to assess. However, it is clear that the use
of such assumptions injects a considerable degree of uncertainty
into the analyses.
Some of the factors which may have led the analyses to under-
state the risk of cancer related to air toxics are as follows:
1. Urban ambient air is characterized by the presence of
dozens, perhaps hundreds of substances. Risk estimates
for most of these could not be calculated due to data
1imitati ons.
2. Indoor concentrations of certain pollutants (e.g., radon,
tobacco smoke, formaldehyde, and other volatile organic
compounds) are commonly several times higher than out-
door concentrations. The estimated cancer incidence
associated with indoor exposures to passive smoking
(500 to 5,000 annually), radon (1,000 to 20,000 annually),
and with 24-hour personal exposures to 5 organic compounds
(1,500 annually) indicate that indoor sources make an
important contribution to air toxics risks.
3. Risks due to compounds formed in the atmosphere could
not be quantified in the analyses using exposure models,
but there are indications that these risks may be
significant. For example, formaldehyde is formed in the
atmosphere by the breakdown of other organic compounds,
and some compounds (e.g., toluene) may be converted into
toxic substances through photochemical reactions.
4. Although it has been shown that certain combinations
of exposures may have synergistic effects (for instance,
smoking and asbestos exposure), all risks were assumed
to be additive.
Factors which may have caused the analysis to overestimate
cancer risks associated with air toxics are as follows:
1. Cancer unit risk values were obtained from EPA's
Carcinogen Assessment Group (CAG) and Clement
Associates. EPA unit risk values are generally
regarded as plausible, upper-bound estimates. That
is, the unit risks are not likely to be higher, but
could be considerably lower. In many cases, the unit
risk values are preliminary.
-------
-1 X-
2. The weight of evidence of carcinogenicity for the
compounds examined varies greatly, from very limited
to very substantial. Further, the extent of evaluation
and health review performed varies considerably among
compounds. For this report, a conservative scenario
(i.e., that all compounds included in the report are
human carcinogens) was assumed.
3. The risk assessments assume that people living in an
area are exposed to the estimated ambient levels for
70 years, 24 hours a day. This especially compromises
estimates of maximum lifetime individual risk. Few plants
operate for 70 years, most people change their homes
several times during their lives, and they also leave
their neighborhoods during the day.
4. The degree to which outdoor emissions of many pollutants
(e.g., trace metals) penetrate indoors is largely
unknown. If emissions of a pollutant from outdoor
sources do not penetrate completely indoors and if there
are no indoor sources of that pollutant, then we will
have over-stated risks, since we have assumed constant
exposure to levels equalling those of outdoor air.
5. Although certain combinations of exposures may have
antagonistic effects, all risks were assumed to be
additi ve.
Current Activities
This study was completed and a draft was sent by the Agency
through peer review late in 1984. Simultaneously, based on the
preliminary findings of the analyses, EPA initiated a series of
activities designed to examine the need for a new national strategy
for air toxics. These included:
1. Formation of an Agency-wide Air Toxics Group to follow
up on the study and to guide the development of any
needed changes in national strategy.
2. Additional analytical studies to examine in more detail
the controllability of the most important pollutants and
the impact on risk of the current programs for criteria
pollutants from now through 1995.
-------
-X-
3. Discussions of the results of the study and possible
new national strategies with all interested groups,
including industry, public interest organizations,
state and local governments, and legislative staffs.
Although these activities will not be completed until mid-1985,
EPA is beginning to explore several changes to its program for air
toxics. The focus of direct federal regulation is shifting from
an emphasis on isolated large point sources to more complex
situations that have greater potential for high national incidence
of cancer. Generally, this will mean increased emphasis on area
sources and those point sources that emit several potentially
toxic pollutants. In addition, several new activities are being
considered that will provide for a more comprehensive national
program. These include a federal partnership with state and
local agencies to evaluate and, if necessary, regulate large
point sources with emphasis on reduction of high individual risk
situations, and an examination of problems caused by concentrations
of sources within cities or industrial regions.
-------
I. INTRODUCTION
A number of air pollutants have been identified as having
the potential to cause cancer when inhaled by humans. Section
112 of the Clean Air Act requires EPA to protect the public
health from exposure to hazardous pollutants which includes such
carcinogens. Recently, state officials, Congress, environmentalists,
and EPA management have expressed concern about EPA's program for
hazardous air pollutants. On August 26, 1983, the General Accounting
Office released a report entitled "Delays in EPA's Regulation of
Hazardous Air Pollutants." As a result, Congressman John Dingell,
Chairman of the House Energy and Commerce Committee, called
hearings held on November 7, 1983, that examined issues surrounding
Section 112.
During internal discussions before the hearings, it became
clear that EPA had not defined well the size or the causes of the
health problems caused by exposure to air toxics. A preliminary
analysis suggested that routine air releases of a group of pollutants
being considered for regulation under 112 might account for no
more than a few hundred cases of cancer each year. This led to
some fundamental questions concerning the magnitude and nature of
risks caused by air toxics.
0 Do air toxics present a significant health problem?
0 If air toxics do pose a significant health problem, what
pollutants and sources emitting those pollutants are
responsi ble?
0 Is there an important part of the national air toxics problem
that Section 112 cannot effectively address?
-------
-2-
EPA's Deputy Administrator decided that a broad scoping study
of the air toxics problem was needed before management could begin
to consider changes in the national program. An ad_ hoc study,
called the Six Month Study because of its original intended dura-
tion, was started in November 1983. Many offices and individuals
within EPA contributed to this analysis, but it was primarily a
cooperative effort between the Office of Air and Radiation (OAR)
and the Office of Policy, Planning and Evaluation (OPPE). This
report summarizes the results of that study.
In the early days of the study, we decided to emphasize four
general issues that would be most useful to policymakers as they
attempted to define the scope and direction of changes that may
be needed to the national program for controlling toxic air
pol 1 utant s.
The magnitude of the airborne carcinogen problem.
We have attempted to characterize the significance
of the problem by presenting quantitative estimates of
the annual incidence of cancer that may be linked to air
pollution, and estimates of lifetime individual risks of
cancer associated with long-term (70 years) exposure to
toxic air pollutants.
The nature of the problem.
What air pollutants and sources emitting those
pollutants cause risks to health? What is their relative
s ig ni fi cance?
Geographic variability.
EPA's strategy for toxic air pollutants may be
influenced by geographic differences in the nature of the
problem. Some sources emitting these pollutants may be
relatively widespread and found in most areas of the nation.
Other sources vary a great deal from city to city and
controlling them may require considerable flexibility
in the pollutants and sources controlled. For instance, an
urban area's unique problems may not be addressed by a
national regulatory program.
-------
-3-
Adequacy of data bases.
This study is the most comprehensive attempt to
date to assemble and analyze available data on air
toxics. Therefore, it is a useful vehicle for evalu-
ating existing data bases, and identifying knowledge
gaps. It should help programs set priorities and plan
for future data gathering efforts, while providing
policymakers with some insight into data needs for
implementing a comprehensive national program.
The resources and time available required that the study be
limited, in most cases, to gathering, organizing, and evaluating
exi sti ng information. This suggested that the results would be
less than definitive, would include major data gaps and assumptions,
and would require a great deal of judgment to interpret properly.
The final report supports these expectations. Risk analysis for
carcinogens is very uncertain, and assessing air toxics is compli-
cated by the poor quality of much of the available data. For
several potentially significant issues, the lack of information
prevents any analysis, and even in those areas with relatively
good information, we had to make important assumptions. Because
of these limitations, consideration of caveats, disclaimers, and
assumptions is an important responsibility of those using this
report. Some of the major assumptions used in this risk analysis
are presented in the following section. Specific limitations
associated with individual analyses are presented in the summaries
of each of those studies.
This study was not conducted to support regulatory decisions
on carcinogenicity or source regulation. Instead, it was designed
to: (1) identify the potential and relative significance of the
risks caused by pollutants and sources from a national and regional
perspective; (2) set research and regulatory priorities; (3) identify
-------
-4-
those pollutants and sources that have not been well studied; and
(4) develop long-term goals and general strategies for air toxics.
Because of the long list of uncertainties associated with
such a study, we used several analytical approaches in attempting to
assess the air toxics problem. Chapter III discusses these
disparate analyses, follows with a discussion of sources and
pollutants not covered by the six analyses, and then concludes
with a summary of the magnitude of the problem. Chapter IV
examines the nature of the airborne carcinogen problem by looking
at relative contributions of pollutants and sources, geographic
variability, and indirect control of air toxics. Chapter V
discusses the quality and extent of available data and Chapter VI
presents the study's conclusions. A final chapter summarizes
Agency actions being taken as a result of the study.
II. SCOPE OF STUDY
The study has the following important characteristics:
0 It is based on existing information. No new data were
generated, even though existing data were used in new
ways, e.g., the treatment of products of incomplete
combustion. The study's value lies in assembling and
organizing the available data and applying quantitative
analysis to better understand an environmental problem.
0 Primarily due to data limitations, the study relied solely
on quantitative estimates of cancer risk associated with
inhalation of ambient air. No other potential health
effects were considered nor were environmental effects
such as crop damage or visibility impairment.
0 The only environmental pathway considered was inhalation.
The study did not examine potential health risks from
ingestion of air pollutants that ultimately reach humans
through the diet or that are directly ingested. The study
also does not address the potential environmental effects
of direct deposition and urban runoff of air pollutants
to surface water.
-------
0 The study focused on inhalation of ambient air.
Indoor exposures were examined only for perspective.
0 The study covered only 15-45 compounds and represents
only a fraction of the total number of compounds
present in the ambient air. The major factor
preventing analysis of more pollutants is the lack
of health data.
0 No quantitative estimates are available for many
potentially important source categories, e.g., Super-
fund sites, hazardous waste disposal, and pollutants
formed in the atmosphere.
0 The study focused primarily on expected, routine,
and continuous emissions of hazardous substances.
Accidental releases, such as those that were
responsible for the tragedy in Bhopal , India, as
well as smaller, more common releases, were not the
subject of our study. However, the ambient air
quality analysis may have included the accidental
releases that contribute to ambient concentrations of
some compounds.
0 The risk estimates in the report are based on
layers of assumptions concerning the health
effects of chemicals, the degree of human exposure,
and the way these substances interact inside the
human body. Many of them are conservative, while
others may tend to underestimate risk. The
result is a degree of uncertainty that we cannot
even begin to quantify. The numerical estimates
presented in this report should be viewed as a
rough indication of the potential cancer risks caused
by a few pollutants. Many of the absolute values are
almost certainly inaccurate, and the strongest and best
use of the numbers is in making relative comparisons
across pollutants and sources, and in setting priorities
and allocating resources.
A. Reasons for Assessing Only Cancer Risk
At the beginning of the study, we decided to rely solely on
estimates of cancer risk associated with air pollutants and
exposures. There were several reasons:
0 Cancer is a significant cause of death in the
United States: approximately 20 percent
(440,000 per year) of all deaths in the United
States are caused by cancer.
-------
-6-
0 Urban areas have higher lung cancer rates than
rural areas.
0 Several identified air pollutants are human
carcinogens, e.g., benzene, arsenic, and vinyl
chloride.
0 The public is concerned about the possible link
between environmental pollution and cancer
i nci dence.
0 The use of a nonthreshold assumption in estimating
cancer risk has broad scientific support. There-
fore, even the very low concentrations typical
of ambient air may still be a problem. Most
acute and subchronic health effects appear to
have a threshold.
0 Cancer incidence lends itself to probabilistic
analysis. There is a well-established mathe-
matical model, i.e., the Crump model, for
estimating risk at low doses. This is not the
case for other effects.
0 There are short-term indicators of mutagenicity/
carcinogenicity , e.g., the Ames test.
The ambient concentrations of most noncriteria pollutants
appear generally to be below the threshold levels associated with
most acute and subchronic health effects. Most such effects are
caused by exposures to concentrations in the parts per million
range, while ambient concentrations of most compounds are in the
parts per billion range. However, this may not always be the
case. The ambient concentrations near some sources may approach
the levels associated with acute health effects. The lack of
data on acute health effects is a serious omission in this report,
and these health effects should not be ignored when assessing an
air toxics source.
Compared to cancer and acute effects, there is even less
information available concerning mutagenicity and teratogeni city.
There are a few examples of compounds with data. Ethylene dibromide
-------
-7-
and ethylene oxide have been shown to be mutagenic in test systems,
and 2,3,7,8-TCDD has been shown to be a developmental toxin in
animals. Unfortunately, the data for most compounds are too
limited to qualitatively determine whether the substances are
potentially mutagenic or teratogenic. For those few substances
with sufficient weight of evidence, there is rarely enough infor-
mation to develop any reliable dose-response estimates. While it
is generally accepted that there are thresholds for some terato-
genic effects in test animals, the data are seldom sufficient to
calculate those threshold levels.
The uncertainty is compounded when animal data are used to
predict human teratogenic effects. For teratogens, there tend
to be multiple end points, and the timing of exposure is often
crucial. These parameters may not be the same for both animals
and humans.
In contrast, there is biological support for a nonthreshold
theory of carcinogenicity in both animals and humans. Furthermore,
it is generally accepted that a substance that causes cancer in
test animals is likely to be carcinogenic to humans as well.
This has not been established for other health effects. Given
these gaps in our knowledge, quantitative risk assessments are
most defensible for cancer. More work is needed to establish
models and methods for assessing quantitatively the risk of other
health effects.
This study did not analyze the impact of major accidental
releases of toxic substances such as MIC in Bhopal, India. These
clearly are a major issue, but were outside the scope of this report.
However, EPA recognizes the potential of such releases and is cur-
rently examining regul atory* opt ions .
-------
-8-
III. METHODS USED TO ESTIMATE CANCER RISK
Assessing the cancer risks of exposure to an environmental
pollutant requires three pieces of information: (1) an estimate of
the carcinogenic potency (the unit risk value) of the pollutant
being considered; (2) an estimate of the ambient concentration
that an individual or group of people may breathe; and (3) an esti-
mate of the number of people that are exposed to those concentrations,
This study is based on methods for assessing carcinogenic
potency and estimating ambient concentrations now in use throughout
EPA. We did not judge the appropriateness of these methods, nor
did we attempt to use other methods. We thought that a comprehensive
analysis of risk assessment techniques was beyond the scope of
this study, and using other methods would make risk comparisons
with other EPA programs more difficult. The following is a
discussion of the methods that we used and the assumptions upon
which our estimates are based.
A. Estimating the Carcinogenic Potency of Pollutants
Assessing the risk of cancer caused by exposure to toxic
substances in the environment is a complex, controversial, and
uncertain business. For most of the pollutants covered by this
analysis, the estimates of risk per unit dose were developed by
EPA's Carcinogen Assessment Group (CAG).1 To calculate such esti-
mates, CAG made the following major assumptions:
1 Quantitative estimates of carcinogenic potency (the unit risk
value) are expressed as the chance of contracting cancer from a
70-year lifetime exposure to a concentration of 1 ug/m3 of a
given substance. Generally, the unit risk value represents the
probability of cancer cases, not deaths. However, since the
epidemiological studies that generated the potency number for
PIC (products of incomplete combustion) are based on lung cancer
mortality, the PIC estimates used in this report imply lung
cancer deaths.
-------
-9-
0 CAG uses experimental data showing that a substance
is carcinogenic in animals to demonstrate that the
substance may be carcinogenic in humans as well.
0 In the absence of human data, CAG uses the results of
such animal bioassays to estimate the probability of
carcinogenic effects in humans. Such extrapolations
assume humans to be as sensitive as the most sensitive
animal species tested.
0 CAG uses a nonthreshold, multistage model that is
linear at low doses to extrapolate from high-dose
response data (either occupational studies or animal
bioassays) to the low doses typically caused by exposure
to ambient air. In other words, CAG assumes that carcino-
genic substances cause some risk at any exposure level.
These unit risk values represent plausible upper bounds--
i.e., they are unlikely to be higher, and could be
substantially lower.
0 CAG assumes that exposed individuals are represented
by a reference male having a standard weight, breathing
rate, etc. No reference is made to health, race,
nutritional state, etc.
Some people have charged that some of these assumptions
overstate risk. However, other factors may offset the conservatism
in the techniques.
0 People are exposed to complex mixtures of chemicals.
Data are not available to demonstrate or deny the
existence of either synergistic or antagonistic health
effects at low exposures.
0 Virtually all animal and human data are based on expo-
sure to adults. There may be enhanced risk associated
with fetal, child, and/or young adult exposures to some
agents.
0 There may be high susceptibility for some population
groups because of metabolic differences or inherent
differences in their response to the effects of
carci nogens.
The Administration recently took a position on some of the more
controversial assumptions above. On May 22, 1984, the White House
Office of Science and Technology Policy (OSTP) released its final
report, Review on the Mechanisms of Effect and Detection of Chemical
-------
-10-
Carci nog ens. The report's statement of principles concludes that
available information "does not allow one to define the existence
or location of a threshold" for carcinogenicity. Furthermore, the
principles state that "a model which incorporates low-dose linearity
is preferred when data and information are limited as is the usual
case and when much uncertainty exists regarding the mechanisms of
carcinogenic action."
B. Estimating Exposure to Pollutants
For most of the analyses summarized in this report, two
measures of risk were calculated: liftime individual risk and
estimated annual incidence. Lifetime individual risk is a
measure of the probability of an individual's developing cancer
as a result of exposure to an ambient concentration of an air
pollutant or group of air pollutants over a 70-year period.
Often, the maximum lifetime individual risk is also presented,
which usually applies to people living nearest the source.2 in an
attempt to gauge the significance of additive risks, we also
calculated mul ti pol 1 utant individual risks caused by many pollutants
measured in the same area. These multipol1utant risks were not
associated with a specific point source.
Aggregate or population risk estimates, on the other hand, are
estimates of the annual incidence of excess cancers for the entire
2 A maximum individual lifetime risk estimate of 3.0x10-4, for
example, near a point source implies that if 10,000 people breathe a
given concentration for 70 years then it is likely that three of
the 10,000 will develop cancer as the result of the exposure to
that pollutant from the source.
-------
-11-
affected population. These estimates are calculated by multiplying
the estimated concentrations of the pollutant by the unit risk value
and by the number of people exposed to different concentrations.
This calculation yields an estimate of the total number of excess
cancers that may occur over a 70-year period. The total must then
be divided by 70 to estimate annual incidence.
C. Using Monitoring Data to Estimate Ambient Concentrations
This study used two major techniques to estimate the ambient
concentrations of pollutants that people may inhale. One technique
used measurements of ambient air quality and the other relied on
emission estimates and dispersion modeling. Each technique has
advantages and drawbacks. Using direct measurements of ambient
concentrations to estimate risk avoids the problems of incomplete
emission inventories, incomplete knowledge on current control
status, a lack of knowledge concerning pollutants formed or
destroyed in the atmosphere, and the list of errors associated
with dispersion modeling. However, there is significant potential
for error in using monitoring data to estimate risk.
The most important potential source of error is the classic
problem of extrapolating measurements at a single site to a much
larger geographic area in order to estimate population exposure.
To estimate concentrations in a city, we were forced to average
measured values and assume that these values applied to the
entire area. The number of monitoring sites in a metropolitan
area ranges from one or two to a maximum of ten in Baltimore and
-------
-12-
Philadelphia. Because of this limited coverage and because monitors
are often intentionally located away from major sources, using
monitoring data probably is especially unsuitable for estimating
maximum individual risk.
In addition, estimating annual incidence forced us to extrap-
olate the available data for a relatively small number of areas
to the rest of the nation. For trace metals and organic particu-
lates, the National Air Monitoring System and State and Local Air
Monitoring Systems (NAMS/SLAMS) contain data for counties representing
a total of 25 million to 75 million people. Data on volatile
organics are available for areas with a total population of only
2 million to 25 million people. The ambient air quality study
made certain assumptions in extending these data to the rest of the
nation. We have no way to assess the accuracy of these methods
which used ambient data measured at certain spots to estimate
concentrations elsewhere.
Third, because cancer risk assessment assumes long-term
exposures, the most useful data are long-term average concentra-
tions, preferably annual averages. Very few studies have collected
ambient samples for toxics continuously for an entire year. For
this study, monitoring data for 20 days a year was labeled as
being sufficient for calculating an annual average. This was
available for most of the trace metals. For organics, annual
averages were calculated if monitoring data were available for ten
separate days spread over at least two quarters.
Finally, all air quality data are subject to errors in sampling
and analytical methods. These are serious problems for many non-
-------
-13-
criteria pollutants. For many compounds, good analytical methods
have yet to be developed.
D. Using Emission Estimates and Dispersion Modeling to Estimate
Ambient Concentrations
Several of the analyses presented in this report relied on
emission estimates and dispersion modeling to estimate ambient
concentrations. A major advantage of this method over ambient data
is the ability to characterize the contribution of various sources.
Also, emission modeling provides more comprehensive geographical
coverage and, therefore, can identify "hot spots" that are of
concern because of high individual risk. Finally, modeling generally
allows a larger number of pollutants to be considered, and it
avoids the problem of geographic extrapolation.
Emission estimates and dispersion modeling were used in most
of the analyses summarized in this report. The major ones are
the 35-County Study, the NESHAP Study, and work completed by the
Office of Radiation Programs on radionuclides. Conceptually, the
models all operate the same way. Emission estimates for area and
mobile sources are apportioned uniformly over the entire area being
considered, while point sources are located at a specific site.
Emission estimates for point sources are developed using available
sources of information, which may vary widely in quality. The
emission estimates are loaded into the computer dispersion model,
along with information on stack height and diameter, emission
velocity, and temperature. Meteorological data (wind speed, direction,
and stability) from the nearest of over 300 National Climate Center
-------
-14-
sites are entered into the model, along with population distribution
information from 1980 census data. Running the models results in
estimates of ambient concentrations at different distances from
the source. The dispersion models were run for 50 km in the 35-
County Study, 20-50 km in the NESHAP analysis depending on the
pollutant, and 80 km for the radionuclides analysis. The appro-
priate choice for the outer boundary when estimating pollutant
dispersion is a matter of considerable debate.
Some of the major issues surrounding the use of both monitoring
and dispersion modeling techniques used in this study to estimate
exposure are as follows:
0 The dispersion models assume flat terrain and
average meteorological conditions. Rough terrain in
the area surrounding a source, such as a valley,
would probably cause higher concentrations near
point sources and lower concentrations further
away from the source.
0 Although exposure estimates apply to a certain point
in time, our risk assessments assume that the people
who live in an area are exposed to the estimated
ambient concentrations for 70 years. In other words,
we assume that the plant operates for 70 years, that
no one moves in or out of an area, and that no one
moves around within the area. Few plants operate
for 70 years, and most people change homes several
times during their life. However, a person may
still be exposed to emissions of the same or different
toxic compounds after moving from an area.
0 A related issue is the assumption that people are
continually exposed to outdoor ambient concentrations
of pollutants. In fact, most Americans spend 80 percent
to 90 percent of their time indoors. Thus, a significant
part of total exposure to air toxics occurs indoors.
We were unable to quantify the risks from indoor exposures
to most of the substances examined in this study. However,
there are strong indications that indoor levels of many
volatile organic compounds are higher than outdoor levels,
since there are many indoor sources of organic compounds.
-------
-15-
No indoor/outdoor comparisons were found for the metals
examined in this study, but the limited data available for
other trace metals show that indoor air levels are sometimes
higher and sometimes lower than outdoor levels.
Dispersion modeling is often extended to only 20 km
from the source. This technique can lead to under-
stating risk if extending dispersion increases
significantly the number of people exposed. To see
what difference a 50 km boundary would make, five
organic substances were modeled to that distance.
This change increased annual cancer incidence by a
factor of 1.35.
Dispersion estimates are rarely based on site-specific
meteorology. Often, data from hundreds of kilometers
away must be used.
In running the dispersion models, we do not consider
increases in concentrations that could result from
reentrainment of toxic particles from streets, rooftops,
etc. With the exception of radionucl ides, we also do
not consider background concentrations and emissions
from other sources not explicitly included in the
analyses, including toxics formed in the atmosphere.
Emission estimates are generated using data and assump-
tions that could be in error. For example, although
the 35-County Study incorporates plant-specific emission
estimates whenever possible, the pollutant releases
for the remaining sources were developed by applying
speciating factors against the VOC and TSP data in the
National Emission Data System (NEDS). Unfortunately,
some of the information in NEDS is of questionable
consistency and quality for the purposes of quantita-
tive risk assessment.
For other analyses, estimates are based on plant capacity
and emission factors. These studies assume that plants
continuously operate at an assumed percentage of capacity
and that no changes in emission rates occur. Emissions
from malfunctions and upsets were not considered in this
study.
-------
-16-
IV. MAGNITUDE OF THE AMBIENT AIR TOXICS PROBLEM
A. Introduction
One of the major goals of this study was to improve our
understanding of the size of the overall public health problem
caused by air toxics, a task that has been colorfully character-
ized in the trade press as determining whether the air toxics
problem is "an elephant or a mouse." We identified several
analytical techniques for assessing the nature and magnitude of
the air toxics problem. Each method offered different advantages,
as well as varying degrees of resolution and uncertainty.
Rather than select one approach for analyzing such a complex
issue, we chose to complete several studies:
o An assessment of the hazardous air pollutant problem based
on state and local experience;
o An estimate of national exposure and risk from about 40
pollutants being considered for listing under Section 112
of the Clean Ai r Act;
o A more detailed estimate and analysis of exposure and risk
in 35 counties for about 20 pollutants, including consideration
of sources that were not considered sources of air emissions
in the past, such as municipal sewage treatment plants (POTWs)
and waste oil combustion;
o An analysis of existing ambient air quality data; and
0 Risk estimates for pollutants and sources either not covered
by the analyses above--e.g., radionuclides, asbestos, and
gasoline marketing--and a discussion of others not easily
quantified--e.g., dioxin and combustion of hazardous waste in
boilers.
In this chapter, we describe each of these studies in more
detail and express the magnitude of the problem in three ways:
annual national cancer incidence; annual incidence per million
people; and lifetime individual risk. We then summarize and com-
pare the results from each effort, and develop general conclusions.
-------
-17-
Again, we must caution against misuse of the results of this
scoping study. The analysis was not undertaken to lead directly
to decisions on carcinogenicity nor regulation. It was designed
to: (1) identify the potential significance of the risk caused by
air toxics from a national and regional perspective; (2) assist
the Agency in setting research and regulatory priorities; (3)
identify those pollutants and sources for which only scant data
exist and should therefore be explored in more detail; and (4)
assist in developing long-term goals and general strategies for air
toxics.
B. Summaries of Individual Analyses
1. Survey of State and Local Agencies, Canada, and Europe
The responsibility of dealing with air toxics is not unique to
EPA or to the United States. Many state and local agencies have
active air toxics programs; also, other industrialized nations have
the same public concern over environmentally related cancer as
the United States. We reasoned that they may have the same need as
EPA to define the risks from air toxics in order to justify programs
and to set priorities. Therefore, a portion of the study involved
communication with Canada, the European community, all states, and
33 major local air agencies regarding their risk assessment
activities. 3,4,5
3 Memorandum from B.J. Steigerwald (U.S. EPA) to Alan Jones et al . ,
(U.S. EPA), "Air Toxics Program in Canada," April 16, 1984.
4 Memorandum from Delores Gregory (OIA) to B.J. Steigerwald (U.S.
EPA), "E.G. Regulation of Hazardous Air Pollutants," May 3, 1984.
5 Radian Corporation, "Definition of the Air Toxics Problem at the
State/Local Level," EPA Contract 68-02-3513, Work Assignment 45,
June 1984.
-------
-18-
Of the agencies and organizations contacted, only California
has attempted to quantify public health risks from air toxics.
Officials in Canada believe that risk assessment will be increasingly
important in their toxics programs, but they have not yet developed
methods and do not apply risk assessment systematically.
They will evaluate in detail the results of this study. We sent
cables to the Commission of the European Communities through
EPA's Office of International Activities and discussions were held
with individuals involved in toxics programs in Europe. There is
much information available from the international community on the
potential toxicity of various compounds, but nothing seems to be
available on cancer incidence or individual risks from exposure
to ambient air pollution.
The California estimate was an isolated analysis published in
1982 to support proposed regulation on air toxics.6 It used Los
Angeles air quality data for nine specific compounds to calculate
excess lifetime cancer rates per million population. Potency for
each compound was determined in a unique way, using an air equivalent
of EPA's Water Quality Criteria, rather than the unit risk value
used in EPA's risk assessment procedures. Therefore, the results
are not directly comparable to the results we obtained in this
study. For the 9 compounds selected, the California analysis
estimated about 1,000 lifetime cancers per million people, or about
14 annual cases per million. The study was used by the California
6 Batchelder, J. et al., "Proposed Amendments to Chapter 1, Part III
of Title 17, California Administrative Code, Regarding the Emission
of Toxic Air Contaminants," California Air Resources Board,
September 1982.
-------
-19-
Air Resources Board for orientation purposes only and to show that
the problem deserved additional attention. The Board does not
recommend the study be given weight beyond its original purpose.
Since most state and local agencies included in the poll
expressed concern over air toxics but could not quantify their
concern directly, we explored other more subtle indicators of the
problem. Counting air "episodes," "incidents," or "complaints"
involving health scares produced no usable statistics. An evaluation
of source permits indicated that, at least for states with fenceline
ambient standards, air toxics programs could utilize substantial
Agency resources and often require controls beyond those needed for
criteria pollutants.
For example, Michigan issues about 1,000 new source permits a year
for emissions of toxic pollutants. New York reviews 36,000 operating
permits every 1 to 5 years under their air toxics regulation; this
number increased by 6,000 emission points in the past 2 years.
Each year, Illinois reviews 5,000 to 6,000 permits that involve
emissions of air toxics. In a recent detailed study of 42 permits
for source categories likely to emit toxics, Illinois found that 20
of the sources were required to apply controls beyond those needed
for criteria pollutants.
In summary, our analyses could not find any other study that
has attempted to comprehensively define risks from air toxics.
However, general concern about the problem is universal, and an
increasing number of states have begun to issue air toxics permits
to large numbers of new and existing sources. Although sometimes
based on use of best technology, these permits are generally based
on dispersion modeling and compliance with fenceline ambient standards
that are derived from occupational guidelines.
-------
-20-
2. Evaluation of Epidemiological Studies Linking Cancer
with Air Pollution
Background
The traditional way to demonstrate the effect of environmental
pollution on public health has been to perform an epidemiologica 1
study. A variety of such studies has been attempted for air
pollution. Our primary source of data on these studies was a
Clement Associates report for EPA that described and critically
evaluated the evidence for cancer associated with air pol1ution.7,8
The report assembled three main types of evidence linking
cancer incidence to air pollution: epi demi ol ogi cal studies, laboratory
studies on the mutagenicity of airborne materials, and ambient air
monitoring data for pollutants known to be carcinogens. Data from
the mutagenicity and monitoring studies confirmed other reports
that extracts of airborne material from polluted air and emissions
from motor vehicles and stationary sources are mutagenic or carcinogenic
in experimental bioassay systems.
The report also reviewed epidemiological studies linking
air pollution and lung cancer by using levels of benzo( a) pyrene
(BaP), a known potent carcinogen, as an indicator of air pollution.9
7 Clement Associates, Inc., "Review and Evaluation of Evidence for
Cancer Associated with Air Pollution," (EPA-450/5-83-006) Review
Draft, November 9, 1983.
8 Pate, Nancy, "Review of the Document 'Review and Evaluation of
the Evidence for Cancer Associated with Air Pollution1 and
Assessment of This Approach for Better Defining the Extent and
Magnitude of the Air Toxics Issue," June 1984.
9 BaP is a ubiquitous pollutant generally found in emissions from
incomplete combustion processes, especially of wood and coal in small
combustion units and in motor vehicle exhaust. BaP is one of the
literally hundreds of organic particulates known as polynuclear
organic compounds. Many polynuclear organics are carcinogenic,
many are not.
-------
-21-
Using selected of these studies, the Clement report presented
calculations of the number of lung cancer deaths which could be
associated with a given level of air pollution as characterized by
BaP concentrations. By combining lung cancer mortality statistics
from the 1960s with estimated levels of BaP in the 1930s and 1940s,
Clement estimated that roughly 10,000 cases of lung cancer per
year during the 1960s were attributable to air pollution.
Unfortunately, because of the long lag time between exposure and
onset of cancer, these findings are not directly relevant to the
hazard posed by current air pollution, particularly since BaP
concentrations have generally declined by a factor of 10 since the
1960s.10
Despite this limitation in the direct use of the results of
epidemiol ogi cal studies, we decided we could, not ignore the polynuclear
organics represented by BaP in this analysis. Even though overall
BaP emissions have decreased significantly since the 1930s and
1940s, BaP-related compounds are still present in the ambient air
and may still represent an important part of the air toxics problem.-
For example, a recent study completed in New Jersey examined ambient
BaP concentrations and mutagenicity of organics extracted from
inhalable particulate matter samples. BaP levels and the mutageni-
city of the particulate increased significantly during the winter
relative to the summer. H
Pate, Nancy, "Review of the Document."
Lioy, Paul J., and Daisey, Joan M., "The New Jersey Project on
Airborne Toxic Elements and Organic Substances (ATEOS): A Summary
of 1981 Summer and 1982 Winter Studies," Journal of the Air
Pollution Control Association, Volume 33, Number 7, July 1983.
-------
-22-
Thus, we decided to use a dose-response coefficient derived
from data cited in the Clement report, and to combine it with current
air quality data and estimates of BaP emissions to estimate cancer
incidence associated with the large category of BaP-related pollutants
which we will refer to in this study as Products of Incomplete
Combustion (PIC). The Clement report presented 14 estimates obtained
from 12 separate reports of the dose-response relationship between
air pollution levels as indexed by BaP concentrations and lung
cancer rates. Of these 14 estimates, 6 were derived from occupa-
tional epidemiological studies, while 8 were derived from general
population studies that related cancer deaths in the period
1959-1975 to ambient BaP levels from 1958-1969.
Clement Associates adjusted the dose-response coefficients
in these general population studies downward to account for the
decline in ambient BaP levels during the lag periods between
exposure and death from lung cancer. In accordance with recom-
mendations by research groups within EPA, certain of the occupa-
tional dose-response estimates presented in the Clement report were
revised (for example, the Carcinogen Assessment Group's latest
estimate for coke oven emissions was substituted for that appearing
in the Clement report). The final potency estimates (as expressed
by lung cancer deaths per year per ng/m3 BaP) for the occupational
studies varied from 0.09 to 0.80 x 10'5, whereas those for the
general population studies varied from 0.3 to 1.4 x 10~5. Averaging
the potencies for each of the two categories of studies yielded
estimates of 0.7 x 10"5 (general population) and 0.5 x 10~5
(occupational). The midpoint of these two values--0.6 X 10~5 deaths
-------
-23-
per ng/m3 Bap per year (or 0.42 deaths per 70 years per ug/m3 BaP) —
was selected and combined with estimates of population exposure to
BaP. Based on air quality data, 610 incidences of lung cancer per
year nationwide were estimated to be attributable to PIC, whereas
124 deaths per year were estimated using BaP emission data and the
more limited population studied in the 35-County Study.
There are several key limitations to using BaP levels as a
surrogate for exposure to a complex mixture of compounds, as we
have done in this analysis. A major weakness of using the potency
estimates derived from the occupational studies is that the mix of
PIC in the exposures studied (coke oven emissions, roofing tar
fumes, and gas fumes) almost certainly differs from that of the
ambient air. Limitations of general population studies are that
BaP in these studies is used as a surrogate for all air pollution
involved in lung cancer, not just PIC, and also that BaP ambient
levels in the 1930s and 1940s had to be estimated. In addition,
the proportion of carcinogenic activity attributable to BaP in PIC
mixtures is known to vary among source categories and sometimes
within a source category (e.g., among different automobiles). The
impact of this varying ratio of BaP to other compounds is further
complicated since synergistic and antagonistic effects between BaP
and other PIC compounds are known to occur, but at present are
unquantifiable. All of these factors indicate strongly that BaP is
almost certainly not a stable index of the carci nogeni city of
pol1uted ai r.
In spite of the limitations of the BaP-surrogate method, we
could find no better alternative for estimating risk due to PIC.
-------
-24-
Simply citing risk estimates for mixtures from specific sources of
PIC was not an option, since quantitative risk estimates are available
for only one source category of interest--coke oven emissions--which
comprises only a small fraction of total estimated PIC emissions.
Also, sufficient data on potency and emissions do not exist to
characterize PIC risks on a compound by compound basis. There are
precedents for using BaP as a surrogate in just this way. The
National Academy of Sciences (NAS) recently used BaP as a proxy to
estimate the cancer risk from polycyclic aromatic hydrocarbons (a
chemically defined analogue of our more loosely defined "PIC"). In
a 1983 report entitled "Polycyclic Aromatic Hydrocarbons: Evaluation
of Sources and Effects," the NAS estimated cancer risks as follows:
This appendix...assumes that benzo(a)pyrene (BaP) can be used
as a proxy for PAH's and that human exposure to BaP in the
ambient air at an average concentration of 1 ng/m3
over an entire lifetime has the effect of increasing by
0.02-0.06% the risk of dying prematurely (at or before
the age of 70) because of lung cancer. Although the
appropriateness of BaP as a surrogate for PAH's in general
has been questioned, it has been so used extensively in the
past, and much of the available information refers to it as
an indicator for exposure to PAH's. (p. D-l )
By way of comparison with the potency estimate used in our analysis
(0.6 x 10-5), the NAS report's estimates of lifetime potency
translate into 0.3 to 0.9 x lO'5 lung cancer deaths per year per ng/m3
BaP. The fact that the midpoint of this range was identical to the
potency chosen for our analysis gave us greater confidence in the
use of the potency.
-------
-24a-
The same NAS report presented estimates of cumulative lung-cancer
incidence due to lifelong exposure to 1 ng/m3 BaP from both gasoline-
and diesel-fueled vehicles (accompanied by other compounds in the
ratios produced by the source). These estimates varied from a low of
20 per 100,000 for the single gasoline-fueled vehicle examined to
a high of 787 per 100,000 for a diesel-fueled vehicle, compared to
that of 43 per 100,000 for coke oven emissions. The PIC-surrogate
approach used in the 35-County Study assumes all sources have the
same incidence per ng/m3 of BaP. In contrast, the lung-cancer
incidence for coke oven emissions is 10-200 times greater than that
for the gasoline- and diesel-fueled vehicles, when expressed on a
constant-weight-of-extract basis, rather than a constant-weight-of-BaP
basis. This indicates that BaP is not a yood surrogate for PIC
associated with particulate emissions from road vehicles.
Thus, we acknowledge that there are real analytical problems
associated with estimating risk due to PIC and that there is vari-
ation in the BaP-surrogate potency estimates. However, since this
report was intended to focus policy and planning activities and was
not meant to serve as the basis for regulatory acton, we decided
to include the incidence estimate for PIC as a preliminary estimate
of the magnitude of the PIC problem.
-------
-25-
3. NESHAP Study
Background
The NESHAP Study was one of two major analyses that employed
dispersion modeling to assess exposure and risk due to air toxics. H
EPA's Human Exposure Model was employed to convert point source
emission estimates (routine emissions, not accidental) into estimated
ambient levels. The study was designed to examine in more detail the
growing belief that sources covered in the past under EPA's NESHAP
regulatory program (i.e., industrial producers and major users of the
chemicals of concern) may be responsible for only a small part of the
air toxics problemi The risk estimates obtained in this study are
national in scope, and consider emissions obtained from traditional air
pollution inventories. The sources covered included mobile and area
sources, but the emphasis was on large point sources. This analysis
did not consider some potentially important pollutants, such as
radionucl ides, gasoline vapors and products of incomplete combustion
(PIC), and such non-traditional sources as POTWs and hazardous
waste disposal.
11 Schell, R.M. "Estimation of the Public Health Risks Associated
with Exposure to Ambient Concentrations of 87 Substances,"
OAQPS, U.S. EPA, July 1984. Revised February 1985.
-------
-26-
The original intent of this effort was to estimate exposure
and risk for 87 pollutants: the original 37 candidates for listing
under Section 112 and 50 additional substances identified by EPA's
Office of Air Quality Planning and Standards (OAQPS). OAQPS
identified this latter grouping of pollutants using the Hazardous
Air Pollutant Prioritization System (HAPPS) developed by Argonne
National Laboratories. OAQPS also considered ambient air monitoring
data and production data in developing the list. Unfortunately,
after a great deal of effort to gather all available dose-response
data on these pollutants, we were only able to quantitatively
analyze 42 compounds (see Table 1). The qualitative judgment
regarding the carcinogenicity of some of these compounds is still
an open question: such compounds are included here for analytical
purposes only. All of the unit risk values used in this report are
presented in Attachment A.
Emission estimates for 27 of the 42 compounds were developed
using OAQPS staff analyses and other OAQPS contract documents.
For the remaining 15 compounds little information was available.
Surrogate estimates of exposure were made for these using a
"best-fit" approach with known compounds based on physical
properties, uses, and production volumes.
Fi ndi ngs
For the 42 compounds included in the NESHAP analysis, a total
nationwide annual cancer incidence of 504 was calculated (see
Table 1). Roughly 90 percent of these can be attributed to the
following 8 compounds, ranked in descending order: chromium;
ethylene oxide; benzene; t ri chl oroethyl ene ; ethylene dibromide;
-------
NESHAP STUDY:
-27-
TABLE 1
PRELIMINARY APPROXIMATION OF ANNUAL INCIDENCE
AND MAXIMUM LIFETIME RISK
Pollutants Having
Some Evidence of
Carci nogenicity*
Preliminary Approx-
imati on of
Maximum Individual
Lifetime Risk**
Preliminary Approx-
imation of
Incidence**
Acryl ami de
Acryl onitri 1 e
Al lyl chl oride
Arsenic
Benzene
Benzyl chl oride
Beryl 1 i urn
1,3 Butadiene
Cadmi urn
Carbon tetrachl ori de
Chl orof orm
Chromi urn"*"
Coke oven emissions
Di ethanol ami ne
Dimethyl nitrosami ne
Dioctyl phthalate
Epi chl orohydri n
Ethyl acrylate
Ethylene
Ethylene dibromide
7.4x10-5
3.8xlO-3
l.SxlO-6
6.5xlO-3
S.OxlO-3
3.0xlO-s
1.0x10-4
9.7xlO-6
3.6x10-3
5.8xlO-4
3.0x10-3
1.6X10-1
2.0xlO-2
2.0xlO-7
5.4xlO-5
9.8x10-6
1.9x10-6
4.7x10-5
4.9xlO-4
1.6xlO-4
0.01
0.42
<0.01
4.70
32.30
<0.01
1.20
0.01
8.50
14.00
0.27
330.0
8.60
<0.01
0.05
<0.01
<0.01
<0.01
<0.01
26.70
* The weight of evidence of carcinogenicity for the compounds
listed varies greatly, from very limited to very substantial.
Further, the extent of evaluation and health review performed
varies considerably among compounds. However, for the purposes
of this report, a conservative scenario (i.e., that all
compounds examined could be human carcinogens) has been assumed.
** Because of the uncertainties in the data used to make these
estimates, they should be regarded as rough approximations of
total incidence and maximum lifetime individual risk. Estimates
of incidence f or i ndi vi dual compounds are much less certain.
These incidence and maximum risk estimates have been performed
to provide a rough idea of the possible total magnitude of the
air toxics problem, and will be used only for priority-setting
and to provide policy guidance.
1" Risk estimates assume that all species of chromium are carcinogenic,
although only certain species have evidence of carcinogenicity.
Current data do not allow differentiation among species.
-------
-28-
NESHAP STUDY:
TABLE 1 (cont.)
PRELIMINARY APPROXIMATION OF ANNUAL
AND MAXIMUM LIFETIME RISK
INCIDENCE
Pol 1 utants Havi ng
Some Evidence of
Carci nogenicity*
Pre1imi nary Approx-
imation of
Maximum Individual
Lifetime Risk**
Prelimi nary Approx-
imation of
Incidence**
Ethylene di chloride
Ethylene oxide
Formal dehyde
4,4 Isopropyl idenedi phenol
Mel ami ne
Methyl Chloride
Methylene chloride
4 ,4-methyl ene dianiline
Nickel subsulfide
Nitrobenzene
Nitrosomorphol i ne
Pentachl orophenol
Perchloro ethyl ene
PCBs
Propylene dichloride
Propylene oxide
Styrene
Terephthal i c aci d
Titani urn di oxide
Tri chl oroethyl ene
Vinyl chloride***
Vinylidene chloride
3.6xlO-3
6.8xlO-3
6.1x10-4
l.lxlO-6
1.5xlO-6
l.ZxlO-5
9.0x10-6
l.BxlO-3
8.3x10-5
l.ZxlO-6
6.0X10-9
1.7xlO-5
4.6xlO-4
S.OxlO-4
Z.lxlO-6
S.OxlO-2
3.3xlO-5
l.BxlO-6
3.2X10-7
l.OxlO-4
3.8x10-3
4.2xlO-3
0.92
47.80
1.60
0.03
<0.01
<0.01
1.0
0.02
0.02
<0.01
<0.01
0.12
2.90
0.21
<0.01
0.97
<0.01
<0.01
0.01
9.70
11.70
0.04
Total
503.8
* The weight of evidence of carcinogenicity for the compounds
listed varies greatly, from very limited to very substantial.
Further, the extent of evaluation and health review performed
varies considerably among compounds. However, for the purposes
of this report, a" conservative scenario (i.e., that all
compounds examined could be human carcinogens) has been assumed.
** Because of the uncertainties in the data used to make these
estimates, they should be regarded as rough approximations of
total incidence and maximum lifetime individual risk. Estimates
of incidence for individual compounds are much less certain.
These incidence and maximum risk estimates have been performed
to provide a rough idea of the possible total magnitude of the
air toxics problem, and will be used only for priority-setting
and to provide policy guidance.
*** NESHAP being applied to many vinyl chloride sources
-------
-29-
carbon tetrachloride; coke oven emissions; and cadmium.13 Maximum
individual risks of 10~3 or greater were estimated for 13 compounds:
acrylonitrile; arsenic; benzene; cadmium; chloroform; chromium;
coke oven emissions; ethylene dichloride; ethylene oxide; 4-4
methylene dianiline; propylene oxide; vinyl chloride; and vinylidene
chloride.
In addition to the usual uncertainties associated with risk
assessment, there are further complications with the risk estimates
for several compounds, including chromium, carbon tetrachloride,
and formaldehyde. These considerations demonstrate the need for
caution in interpreting such studies.
In the case of chromium, only the hexavalent form has been
proven to be carcinogenic, although it is a potent carcinogen.
There is now insufficient evidence to determine that the trivalent
form is also carcinogenic. The NESHAP analysis, however, assumes
that total chromium releases are carcinogenic and that trivalent
chromium is as potent as hexavalent. There is no information now
available on the ratio of trivalent to hexavalent for emissions or
ambient concentrations, but some occupational exposure studies
suggest that the trivalent form may dominate in some source cate-
gories. On the other hand, several important source categories are
known to emit at least some hexavalent chromium, and there is some
evidence that changes in the valence state can occur in the atmosphere,
The problem of speciation adds one more layer of uncertainty to the
risk estimates for chromium.
13 Approximate individual percentage contributions of some of the
more important compounds are: chromium (65%); nickel (70%);
ethylene oxide (10%); benzene (70%); ethylene dibromide (5%);
coke oven emissions (2%); cadmium (2%); and carbon tetrachloride
(3%).
-------
-30-
Carbon tetrachloride is a very stable organic compound that
has a half-life of about 35 years,' compared with a half-life of
hours or days for most other common volatile organic compounds. As
a result, carbon tetrachloride is accumulating in the atmosphere.
Therefore, current emissions are associated with current and future
cancer risks. The NESHAP analysis covers only current risks and
estimates incidence at 14 per year. If current ambient levels
(rather than modeled levels) are used, the incidence estimate
increases to about 85 per year. Carbon tetrachloride also has the
potential to deplete stratospheric ozone and thereby to indirectly
increase the incidence of skin cancer. For example, preliminary
calculations estimate that by the year 2020 U.S. emissions of
carbon tet rachl on" de could be responsible for between 500 and
22,000 cases of skin cancer annually in the U.S., resulting in 3 to
220 deaths per year.14
Finally, formaldehyde is another example of the complexities
that exist in the analysis. It can be formed in large quantities
in the atmosphere, and the risks posed by the resulting ambient
concentrations cannot be considered in exposure analyses based on
emission estimates alone. Assessments based on ambient monitoring
data provide a more complete accounting of actual risk due to
formaldehyde, because they cover concentrations resulting from both
emissions and atmospheric formation. The NESHAP estimate based
on emissions was 1.6 per year; the ambient data resulted in an
estimate of 191 per year.
14 Zaragoza, L. "Calculating Effects of Carbon Tetrachloride and
Other Chiorocarbons on Increases in Skin Cancer from Stratos-
pheric Ozone Depletion," EPA, OAQPS Draft. July 25, 1984.
-------
-31-
4. 35-County Study
Background
In contrast to the national scope of the NESHAP study, the 35-
County Study was designed to address the air toxics problem from a
more local perspective.15 Building on the work of EPA's Integrated
Environmental Management Division (IEMD) in its geographic demonstra-
tion projects in Philadelphia, Baltimore and Santa Clara Valley, this
analysis explored:
0 the incidence of cancer resulting from exposure to
several pollutants and sources in specific localities;
0 the pollutants and sources that are the most significant
contributors to incidence; and,
0 the geographic variability of pollutants, sources, and exposures.
The analysis focused on traditional sources, i. e-. ,
large point sources such as power plant and industrial facilities,
and area sources, such as motor vehicles, space heating, gasoline
marketing, and solvent usage. However, it also included "nontradi-
tional" sources, such as wood stoves, waste oil combustion, and
sewage treatment plants. Because of data limitations we could not
make emission estimates or perform any extensive exposure modeling
for TSDFs (hazardous waste treatment, storage and disposal facilities),
Superfund sites, hazardous waste in boilers, municipal waste inciner-
ators, municipal landfills, and sewage sludge incinerators. The
Agency has initiated various studies to explore emissions and risks
for most of these sources in more detail. Information on these
Versar; American Management Systems, Inc., "Hazardous Air
Pollutants: An Exposure and Risk Assessment for 35 Counties,"
U.S. EPA Contract #68-01-6715, September 1984.
-------
-32-
efforts, as well as any preliminary findings, is provided in the
section on Other Sources, Pollutants and Pathways at the end of
this chapter.
The analysis characterized exposure and risk associated with
22 compounds (see Table 2). Most of these compounds were screened
using one or more of the following criteria:
0 Sufficient evidence of carcinogenicity;
0 Significant release rates; and
0 Readily available emissions information.
Emission estimates for routine emissions were developed using
several techniques. Whenever possible, the analysis relied on
plant-specific data and EPA documents on emissions from specific
source categories. Where this information was unavailable, surrogate
loadings were developed using the information in the National
Emissions Data System (NEDS), and apportioning factors that speciate
the volatile organic compound and particulate matter data into
individual toxic constituents. NEDS data vary a great deal in
quality, and some of the data are very poor. However, an extensive
effort was made to screen NEDS data for the 35 counties to correct
for any obvious inaccuracies in release rates, source locations and
stack specifications.
We developed special algorithms for the following sources:
POTWs, waste oil combustion, woodsmoke, and gasoline marketing.
To calculate releases for selected volatile compounds from sewage
treatment plants, we modeled thirteen prototype POTWs using information
provided by EPA's Industrial Facilities Discharge (IFD) file, the
NEDS survey, and a study conducted by the effluent guidelines program
-------
-33-
35-COUNTY STUDY:
TABLE 2
PRELIMINARY APPROXIMATION OF ANNUAL
INCIDENCE
Pollutants Having Some
Evidence of Cardnogenicity*
Preliminary Approximation of
I nci dence**
(20% of U.S. Population)
PIC***
Benzene
Chromi urn"1"
Formaldehyde
Vinyl chloride
Tri chloroethylene
Gasoline Vapors
Perchloroethyl ene
Acrylonitri 1 e
Coke oven emi ssi ons
Ethylene dichloride
Arseni c
Cadmiurn
Benzo( a) pyrene
Ethylene dibromide
124.3
18.5
13.4
10.0
8.2
6.8
6.8
6.7
.4.2
2.4
1.5
1.1
1.1
1.1
1.0
* The weight of evidence of carcinogenicity for the compounds listed
varies greatly, from very limited to very substantial. Further, the
extent of evaluation and health review performed varies considerably
among compounds. However, for the purposes of this report, a conser-
vative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
** Because of the uncertainties in the data used to make these estimates,
they should be regarded as rough approximations of total incidence.
Estimates for individual compounds are much less certain. These incidence
estimates have been performed to provide a rough idea of the possible
total magnitude of the air toxics problem, and will be used only for
priority-setting and to provide policy guidance.
*** "Products of Incomplete Combustion" (PIC) refers to a large number of
compounds, probably consisting primarily of polynuclear organics. The
PIC unit risk value was derived from dose-response data which use B(a)P
levels as a surrogate for PIC or total air pollution. There are many
limitations of using the B(a)P surrogate method to estimate PIC risks:
all PIC estimates presented in this report must be regarded as highly
uncertain. Refer to pp. 20-25 for a more detailed explanation of how
the PIC unit risk value was derived.
i" Risk estimates assume that all species of chromium and nickel are
carcinogenic, although only certain species have evidence of carcino-
genicity. Current data do not allow differentiation among species.
-------
35-COUNTY STUDY:
-34-
TABLE 2 (Cont.)
PRELIMINARY APPROXIMATION OF ANNUAL
INCIDENCE
Pollutants Having Some
Evidence of Carcinogenicity*
Preliminary Approximation of
I ncidence**
(20% of U.S. Population)
Carbon tetrachloride
Chloroform
Styrene
Beryl 1 i urn
1 ,3-Butadiene
Pentachlorophenol
Total
0.2
0.1
0.02
0.01
0.01
< 0.01
207.4
* The weight of evidence of carcinogenicity for the compounds listed
varies greatly, from very limited to very substantial. Further, the
extent of evaluation and health review performed varies considerably
among compounds. However, for the purposes of this report, a conser-
vative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
** Because of the uncertainties in the data used to make these estimates,
they should be regarded as rough approximations of total incidence.
Estimates for individual compounds are much less certain. These incidence
estimates have been performed to provide a rough idea of the possible
total magnitude of the air toxics problem, and will be used only for
priority-setting and to provide policy guidance.
-------
-35-
to determine the fate of priority pollutants in 50 POTWS.16»17
The sewage treatment plants in each of the 35 counties were assigned
to one of the model plants based on the following factors: the
percent of inflow to the POTW attributable to industrial dischargers;
the types of industries that discharge to the POTW; and the level
of treatment at the POTW. The sewage treatment plant emissions
were modeled as point sources in the exposure assessment.
Toxic emissions from waste oil combustion were characterized
using data from EPA Office of Solid Waste (OSW) documents on: the
typical contaminant concentrations found in used oil; the estimated
amount of waste oil burned in each state; the destruction efficien-
cies for metals and organic compounds burned in industrial and in
residential, institutional and commercial boilers; and the
percentage of total waste oil burned in each type of boiler.18
The study of waste oil focused on chromium, cadmium, beryllium,
arsenic, benzene, benzo(a)pyrene, perchloroethylene, and trichloro-
ethylene. Waste oil emissions were modeled as area sources.
Air toxics releases from woodsmoke were estimated for two
sources--fireplaces and wood stoves.19 using available information,
16 Fate of Priority Pollutants in Publicly Owned Treatment Works,
Vol. I., (EPA 440/1-82-303), September 1982.
17 For further explanation on the method for estimating POTW
volatilization, see Versar/American Management Systems, Inc.,
"Hazardous Air Pollutants: An Exposure and Risk Assessment for
35 Counties," Appendix F-2, September 1984.
18 For further explanation on the method for estimating
toxics emissions from waste oil combustion, see Versar
American Management Systems, Inc. "Hazardous Air Pollutants,"
Appendix C.
19 For further information on the method for estimating woodsmoke
emissions, see Versar/American Management Systems, Inc.,
"Hazardous Air Pollutants," Appendix B.
-------
-36-
we developed factors for five compounds (benzo(a)pyrene, formalde-
hyde, beryllium, cadmium, and arsenic) relating pollutant emissions
to the quantity of wood burned in each county. Data on wood consump-
tion in each county were obtained from NEDS, and the breakdown on
the amount of wood burned in woodstoves vs. fireplaces in each
area was provided by an industry association. We modeled wood
smoke as an area source.
Finally, air toxics emissions from gasoline marketing were
calculated using volatile organic compound data in NEDS and apportion-
ing factor developed from varied sources. The pollutants considered
were: gasoline vapors, benzene, ethylene dibromide, and ethylene
dichloride.
As to the choice of geographic sites, we decided to concentrate
on counties, as data are rarely disaggregated below this level. We
chose 35 counties to explore in detail, and each county fell into
one of three categories:
0 densely populated, highly industrialized;
0 densely populated, low industrial activity; or
0 low population density, highly industrialized.
The counties were chosen to represent a wide range of industrial
bases and geographic locations. Although they contain only about
one percent of the counties in the U.S., the 35 counties account
for roughly 20 percent of U.S. population (1980 Census Data), 20
percent of total national VOC emissions, and 10 percent of total PM
emi ssi ons.
-------
-37-
As with the NESHAP analysis and other Agency studies on
exposure, the 35-County Study employed dispersion modeling to
calculate dose and exposure. EPA's Office of Toxic Substances'
fate and transport model, GAMS, was used in this effort. To
facilitate running the model more quickly and efficiently, we used
an approach that only allowed us to calculate annual aggregate
incidence for the 35 counties.
Fi ndi ngs
Multiplying the results from the exposure modeling by the
appropriate unit risk values (Attachment A) resulted in the incidence
estimates presented in Table 2. The estimated aggregate incidence
of cancer for the 22 pollutants and 35 counties is 207 per year.
As shown, 8 substances account for roughly 95 percent of the total
risk. These pollutants, ranked in descending order, are as follows:
PIC; benzene; chromium; formaldehyde; vinyl chloride; trichloroethy-
lene; gasoline vapors; and perchloroethylene. PIC alone contributes
60 percent to total incidence.
Many of the basic problems discussed in the NESHAP analyses
are applicable to the 35-County Study (see pp. 28-33). Also,
the 35-County Study considered emissions of carbon tetrachloride
from only a limited number of sources. Background concentrations
due to the long half-life of carbon tetrachloride were not modeled,
although they may significantly contribute to cancer risks.
-------
-38-
5. Ambient Air Quality Study
Background
As part of the overall study, we used ambient air quality data
to estimate cancer incidence and individual risks.20 jwo basic
groupings of compounds were used in this analysis: those for which
fairly extensive data were avai1able--four metals and B(a)P--and
those for which less extensive data could be found — nine organic
compounds. The metals and B(a)P data were drawn from the National
Air Data Bank's Storage and Retrieval of Aerometric Data (SAROAD)
system. In contrast, the data for organic compounds came from a
variety of sources, principally from studies which used diverse
sampling and analytical methods and sampling periods.
Every attempt was made to gather all available ambient data
on air toxics. For example, for organic compounds the data base
incorporated data compiled from a variety of sources by Dr. Hanwant
Singh of SRI International and from more recent monitoring studies
performed in Baltimore, Los Angeles, Houston, Philadelphia, and in
northern New Jersey. As far as we know, this effort represents the
most comprehensive attempt yet to compile nationwide data for toxic
air pol lutant .and to perform risk assessments based on those data.
It is appealing to use ambient air quality data — as opposed
to modeled estimates--to estimate risks because these data represent
the actual ambient concentrations to which people are exposed.
Hunt, Bill, et al., "Estimated Cancer Incidence Rates from Selected
Toxic Air Pollutants Using Ambient Air Data," U.S. EPA, revised
20
March 1985.
-------
-39-
However, the reader is reminded of three cautions which have been
discussed previously. First, we must assume that data collected at
a limited number of sites can be extrapolated to represent city-wide
and county-wide levels, and that these data in turn can be extrapo-
lated to the national level. Second, we must often use data collected
over a short time period (e.g., 24 hours,) and assume that in the
aggregate they are representative of longer term concentrations
(e.g., annual averages). Finally, we assume that people are continu-
ously exposed to levels equal to those of ambient air.
National estimates of cancer incidence were calculated for
metals (see Table 3) by estimating county averages based on 1979-1982
data for the approximately 170 counties that had data, by using
these averages to extrapolate to those counties that lacked data,
and then by applying the unit risk values presented in Attachment A.
We estimated a national incidence for PIC by dividing the country
into 11 regions and using urban/rural B(a)P concentrations in
combination with urban/rural population figures for each region.
Estimating incidence for the volatile organic compounds was
somewhat more difficult, given that ambient data on these compounds
are scarce and often derived from short-term studies. To
provide at least minimal seasonal balance when computing annual
averages, we established a data completeness criterion21 for
organic compounds in urban areas. This greatly reduced the amount
of data that could be used. Only data from studies performed in
21 More than two sites per county, and at least ten samples over
two quarters in a single calender year.
-------
-40-
TABLE 3
AMBIENT AIR QUALITY STUDY: PRELIMINARY APPROXIMATION OF
ANNUAL INCIDENCE
Pollutants Having
Some Evidence of
Carci nogenicity*
Preliminary
Approximation of
Incidence**
Incidence per
Million Population**
Arsenic
Benzo(a) pyrene
PIC***
Benzene
Beryl 1 i urn
Cadmi urn
Carbon tetrachl ori de
Chi orof orm
60
5
610
234
-------
-41-
TABLE 3 (Cont.)
AMBIENT AIR QUALITY STUDY: PRELIMINARY APPROXIMATION OF
ANNUAL INCIDENCE
**
Pollutants Having
Some Evidence of
Card nogeni city*
Pre I imi nary
Approximation of
Incidence**
Incidence per
M i 1 lion P OD u 1 ati on**
Chromi urn"*"
Ethylene di chloride
Formal dehyde
Methyl chloride
Methylene chloride
Perehloroethylene
Tri chloroethyl ene
Vinylidene chloride
Total
242
11
191
1
1
22
18
62
1532
1.05
0.05
0.83
<0.01
<0.01
0.10
0.08
0.27
6.7
The weight of evidence of carcinogenicity for the compounds listed
varies greatly, from very limited to very substantial. Further, the
extent of evaluation and health review performed varies considerably
among compounds. However, for the purpose of this report, a conser-
vative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
Because of the uncertainties in the data used to make these estimates,
they should be regarded as rough approximations of total incidence.
Estimates for individual compounds are much less certain. These
incidence estimates have been performed to provide a rough idea of the
possible total magnitude of the air toxics problem, and will be used
only for priority-setting and to provide policy guidance.
Risk estimates assume that all species of chromium are carcinogenic,
although only certain species have evidence of carcinogenicity. Current
data do not allow differentiation among species.
-------
-42-
Baltimore, Philadelphia, Los Angeles, Houston, and northern New
Jersey met the criterion. For these cities, an average level was
calculated for each organic compound, and these averages were then
combined with population figures to calculate incidence. Next,
these estimates were extrapolated to the national level by using
urban population data. Nonurban risks were calculated by using
nonurban pollutant levels and population data, and these were
added to urban risks to obtain national estimates.
Fi ndi ngs
As Table 3 shows, seven compounds are associated with greater
than 50 cancers per year: arsenic, PIC, benzene, chromium, formalde-
hyde, and vinylidene chloride. The national incidence estimate
based on ambient data for the compounds shown in Table 3 is approxi-
mately 1,530 per year. The estimated incidence per million population
for those pollutants is about 6.7 per year.
Individual lifetime risks were also estimated for metals,
PIC, and organics (Table 4). Individual risks ranged up to 10~3
for some of the trace metals and PIC, whereas individual risks for
the organics tended to be in the range of 10~4 and lower. It should
be noted that the sites where these data were collected are generally
not located near points of expected maximum concentrations.
Therefore, the individual risk estimates for single pollutants
based on air quality data tended to be lower than those based on
exposure modeling of emissions from point sources.
However, to provide a better understanding of risks in
urban areas, we estimated individual risks not only on an
individual pollutant basis, but also for many pollutants measured
at the same site. The results of this analysis are presented in
-------
-43-
Table 5 for several urban areas in which extensive ambient moni-
toring has been performed. These multipol1utant individual risks
represent the summed individual risks at each site using monitoring
data that were available for 10 to 15 organics, metals, and PIC.
Table 5 shows that these multipol1utant individual lifetime risks
range around 1 X 10~3 for those areas with sufficient data for
analysis. Lifetime individual risks for single pollutants at these
sites varied from 10~3 to 10~9; pollutants causing risks in the
10~3 to 10-4 range included chromium, PIC, benzene, and formaldehyde,
To our knowledge, none of the monitoring sites were near major
point sources of the relevant compounds, although all sites were
centrally located in major urban areas.
It is important to note that the uncertainties associated with
extrapolating data collected at a few monitoring sites to an entire
urban area do not apply to these estimates of multipol1utant
individual risk. All that is involved is summing individual risks
from a pollutant mixture at a given urban location. Thus, with the
assumption that risks are additive, we can say that, even in neigh-
borhoods not located near major point sources, urban dwellers may
experience individual risks of 10~3 to 10-4 due to multi-pol1utant
air exposures.
-------
-44-
TABLE 4
AMBIENT AIR QUALITY STUDY: PRELIMINARY APPROXIMATION OF
INDIVIDUAL LIFETIME RISKS
Pollutants Having Some
Evidence of
Carci nogeni city*
Preliminary Approximation
of Maximum Lifetime
Individual Risk**
Arseni c
B(a)P
PIC***
Benzene
Beryl 1i urn
Cadmi urn
Carbon tetrachloride
Chioroform
3.99x10-3
2.47x10-5
3.15x10-3
1.54x10-4
2.40xlO-7
1.47x10-3
1.54x10-4
7.70x10-5
* The weight of evidence of carcinogenicity for the compounds listed
varies greatly, from very limited to very substantial. Further, the
extent of evaluation and health review performed varies considerably
among compounds. However, for the purpose of this report, a conser-
vative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
** Because of the uncertainties in the data used to make these estimates,
they should be regarded as rough approximations of maximum lifetime
individual risk. Estimates for individual compounds are very uncertain.
These risk estimates have been performed to provide a rough idea of
the possible total magnitude of the air toxics problem, and will be
used only for priority-setting and to provide policy guidance.
*** "Products of Incomplete Combustion" (PIC) refers to a large number of
compounds, probably consisting primarily of polynuclear organics. The
PIC unit risk value was derived from dose-response data which use B(a)P
* levels as a surrogate for PIC or total air polluton. There are many
limitations of using the B(a)P surrogate method to estimate PIC risks:
all PIC estimates presented in this report must be regarded as highly
uncertain. Refer to pp. 20-25 for a more detailed explanation of how
the PIC unit risk value was derived.
-------
-45-
TABLE 4 (Cont.)
AMBIENT AIR QUALITY STUDY: PRELIMINARY APPROXIMATION OF
INDIVIDUAL LIFETIME RISKS
Pollutants Having Some
Evidence of
Carci nogem'city*
Preliminary Approximation
of Maximum Li fet ime
Individual Risk**
Chromium"*1
Formaldehyde
Methyl chloride
Methylene chloride
Perchloroethylene
Tri chloroethylene
Vi nylidene chloride
1.44x10-3
4.91x10-5
4.60x10-7
8.28x10-7
1.88x10-5
2.59x10-5
8.06x10-5
* The weight of evidence of carcinogenicity for the compounds listed
varies greatly, from very limited to very substantial. Further, the
extent of evaluation and health review performed varies considerably
among compounds. However, for the purpose of this report, a conser-
vative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
** Because of the uncertainties in the data used to make these estimates,
they should be regarded as rough approximations of maximum lifetime
individual risk. Estimates for individual compounds are very uncertain.
These risk estimates have been performed to provide a rough idea of the
possible total magnitude of the air toxics problem, and will be used
only for priority-setting and to provide policy guidance.
t Risk estimates assume that all species of chromium are carcinogenic,
although only certain species have evidence of carcinogenicity.
Current data do not allow differentiation among species.
-------
-46-
TABLE 5
AMBIENT AIR QUALITY STUDY: PRELIMINARY APPROXIMATION OF
ADDITIVE LIFETIME RISKS*
Urban Area A
Monitoring Site 1
Monitori ng Site 2
2.3x10-3
2.3x10-3
Urban Area B
Mo nit on' ng Site 1
Monitori ng Site 2
0.7x10-3
0.7x10-3
Urban Area C
Monitoring Site 1
Monitori ng Site 2
1.1x10-3
1.0x10-3
Urban Area D
Monitori ng Site 1
Monitoring Site 2
0.8x10-3
0.8x10-3
* These estimates are based on a sum of estimated lifetime
individual risks for PIC. (products of incomplete combustion),
two to three metals and six to ten organic compounds for each
monitoring site. Because of the uncertainties in the data used
to make these estimates, they should be regarded as rough approxi-
mations of individual risk. Estimates for individual compounds
are much less certain. These incidence estimates have been
performed to provide a rough idea of the possible total magnitude
of the air toxics problem, and will be used only for priority-setting
and to provide policy guidance.
-------
-47-
6. Other Pollutants, Sources and Pathways
One of the principal findings of this study of air toxics is
that there are important gaps in our knowledge of this problem. This
study estimates cancer risks caused by 15-45 substances, when there
may be many more potential carcinogens in the ambient air. The
International Agency for Research on Cancer (IARC), the National
Toxicology Program, and EPA's Carcinogen Assessment Group have each
identified over 100 compounds as carcinogenic. Many of these
compounds are probably not air pollutants, but it is clear that this
study does not quantitatively address a large number of pollutants
that exist in significant quantities in the ambient air. This
study attempted also to address all known or suspected sources of
air toxics, as well as known pollutants. Unfortunately, we were
unable to quantify the risks caused by several source categories,
including several nontraditional sources. In addition, each of the
individual analyses missed some sources or pollutants.
However, some of the sources and pollutants not included in
the major analyses have been subjects of quantitative analysis by
others. The following section summarizes available information on
the pollutants and sources that (1) were not covered by the individual
analyses; or (2) could not be quantitatively assessed because of
data .limitations. As with most of the analyses these represent the
situation early in 1985 and changes are inevitable over time.
-------
-48-
POLLUTANTS
Dioxi n
Only isolated estimates are provided for individual risks from
emissions of dioxin, and these are limited to municipal incinerators.
The exposure pattern for dioxin appears to be complex and available
data are inconsistent. However, this is true for many compounds
that we have included in the study. Dioxin is unique because
exposure and risk are being studied in great detail by EPA's Dioxin
Task Force. The study team believed that there was little value at
this time in attempting an estimate of the aggregate risk from air
exposure for a pollutant that is currently being analyzed elsewhere
i n such detai1 .
Asbestos
EPA's Office of Pesticides and Toxic Substances is actively
considering regulations for asbestos, since past use of asbestos-
containing building materials can lead to indoor contamination.
Asbestos is also commonly found in the ambient air, although at
much lower levels than -i ndoors, and selected sources are already
covered by federal emission standards under Section 112 of the
Clean Air Act. Sampling and analysis for asbestos in the atmosphere
presents significant problems and concentrations vary by several
orders of magnitude. The available data suggest an average of
three nanograms/m3 and 30 fibers per nanogram.22 Coupling this
22 "Guidance for Controlling Friable Asbestos-Containing Materials
in Buildings." U.S. EPA Office of Pesticides and Toxic Substances,
EPA 560/5-83-002, March 1983.
-------
-49-
with an average risk factor for lung cancer and mesothelioma ,23
gives a national estimate of over 100 excess cancers per year, or
about 0.5 per million population per year. This estimate covers
outdoor exposures only.
Radionuclides
EPA's Office of Radiation Programs (ORP) has evaluated
radionucl ides as a hazardous pollutant, based on the widespread
human exposure to radionuclides in the ambient air, and on numerous
studies that document the incidence of cancer resulting from exposure
to ionizing radiation in many species of animals and human populations.
ORP has recently summarized its exposure and risk assessment for
radionucl ides.24 As shown in Table 6, the total national es-timated
incidence for radionuclides is 16 per year; maximum lifetime individual
risks range from 5 x 10~2 to 4 x 10~7. The major sources of radionu-
clides include nuclear power plants, national defense weapon facilities,
industrial plants, coal-fired boilers and natural sources. The
incidence calculation does not consider exposure to indoor concentra-
tions of radon.
23 Schneiderman, Nisbet, and Brett, "Assessment of Risks Posed by
Exposure to Low Levels of Asbestos in the General Environment,"
Berichte. Bundes-Gesundheits-AMT, pp. 3-1 to 3-28, April 1981.
24 Hardin, J. "Issue Paper. National Air Toxics Problem: Radio-
nuclides." EPA, Office of- Radiation Programs, August 1984.
Update provided verbally on January 31, 1985.
-------
-50-
TABLE 6
ESTIMATES OF INCIDENCE AND INDIVIDUAL RISK DUE TO
RADIONUCLIDES EMITTED TO AIR*
Source An
Dept. of Energy
Faci 1 itles
Nuclear Regulatory
Commission (NRC)
Licensed Facilities
Federal Facilities
Uranium Fuel Cycle
Maximum Individual
nual Cancer Incidence Lifetime Risk
0.07
<0.01
<0.01
5
8 x
2 x
4 x
1 x
10
10
10
10
-4
-5
-7
-4
Federal Facilities <0.01
Uranium Fuel Cycle 5
Faci 1 ities
Uranium Mill Tailings 7
Piles
Urani urn Mi nes 1 .0
Phosphorus Plants 0.06
Coal-FiredBoilers 3
Sources of Natural Radio- <0.1
nuclides to Air
4 x 10-7
1 x 10-4
4 x 10-2
5 x 10-2
1 x 10-3
3 x 10-5
2 x TO'4
TOTAL 16
* Because of uncertainties in underlying data, the values presented
in this table should be regarded as estimates of incidence and
maximum lifetime risk. This table was provided by EPA's Office
of Radiation Programs. Please refer to footnote 24 for a more
detailed explanation of the methodology.
-------
-51-
Recent studies have indicated that indoor air concentrations
of various pollutants can greatly exceed ambient conditions. As
a result, risk assessments based on ambient levels may understate
the actual risk. For radionuclides, recent estimates place the
annual incidence of cancer due to indoor radon exposure at between
1,000 and 20,000. A more detailed discussion of the ramifications
of indoor air on the hazardous air pollutant problem is provided
in the section of this report on Perspective and Context.
Other Pollutants
It is apparent that urban ambient air is characterized by the
presence of hundreds of organic compounds; fine particulate matter,
including metals and organic part i cul ates; and significant concentra-
tions of the other criteria pollutants, including sulfur and nitrogen
oxides, and carbon monoxide. Relatively few data are available
on how all of these substances may interact once they enter the
human body.
An example of the complexity of urban air is shown in Figure I.
a gas chromatogram from EPA's Integrated Environmental Management
Division's monitoring program in a major metropolitan area. It
represents the concentrations and number of gaseous organics in the
ambient air as detected by gas chromatography/mass spectroscopy.
Each peak represents a separate organic compound. The peaks corre-
sponding to some compounds are labeled. Almost 50 tentatively
identified compounds added up to the following totals:
-------
RIC
03/12/84
14:50:60
OATH:
CALI:
1510 tfl
1513 112
SCAMS 1 TO 1400
SAMPLE: SITEtt2 Ptt27 URE466A 38.9L TAGIJ84&2A
CGI IDS.: FSCC 30H DB-5 8 FOR 6 TO 120 818
RANGE: G
100.0-1
1*1488
229
LABEL: II 0, 4.0 QUAN: A 0. 1.0 J 0 BASE: U 2Q, 3
< z
(J n
•- U.
RIC
1077
65243,
1236
o
CD
S
°
1000
12:30
1200
15:00
1400 SCAN
17:30 TIME
-------
-53-
Alkanes 39.1 ug/m3
Aromatics 34.8 u'g/m3
Halogenated compounds 9.8 ug/m3
Oxygenated compounds 7.5 ug/m3
Alkenes 3.4 ug/m3
SOURCES
Atmospheric Transformation
Most population exposure models begin with estimates of emis-
sions, and thus inherently cannot handle toxic compounds that may be
formed or rapidly destroyed in the atmosphere. The exposure models
used in the NESHAP and 35-County studies assume that all exposures
occur within several hours of emission (within 20 km to 50 km of the
source) and no corrections are made for transformation of pollutants
i n the atmosphere.
As part of the study, EPA's Office of Research and Development
was asked to review the possibility that chemical reactions in the
atmosphere could form toxic compounds or increase the potency of
emitted pol1utants.25 Ozone is the prime example of this phenomenon
for criteria pollutants. Although work in this area has not been
extensive, the study identified several potentially significant
examples of atmospheric transformation. A few of the situations
discussed in the review are summarized below.
25 Bufalini, J., B. Gay and B. Dimitriades, "Production of Hazardous
Pollutants Through Atmospheric Transformation," U.S. EPA Office
of Research and Development, June 1984.
-------
-54-
Formaldehyde and acrolein are formed readily in a variety of
photochemical reactions involving emissions from many types of
natural and man-made hydrocarbon emissions. For formaldehyde, an
important contributor to total risk in this study, atmospheric
formation may produce several times the amount directly emitted
from all sources. This may explain some of the major differences
between the risks estimates obtained by using exposure models
versus measured data.
Experimental evidence is also available that photooxidation
of compounds with little evidence of carcinogenicity, such as toluene
and propylene, produce substances with significant mutagenicity.
The compounds responsible have not been fully identified. In
other experiments, phosgene has been produced photochemically from
chlorinated hydrocarbons, such as solvents. The studies suggest
that a hundred times more phosgene may be formed in the atmosphere
than is emitted directly.
As a final example, studies of the mutagenic activity of
polycyclic organic particulates show large increases in activity
when the material is subjected to mixtures of ozone and nitrogen
oxides. Organic particulates are a ubiquitous group of pollutants
generally associated with incomplete combustion (mobile sources,
small units burning wood, coal, and oil). They are represented by
PIC in this report and may be a major contributor to risks from air
toxics in many communities.
Gasoline Marketing
Gasoline marketing includes a series of emission points ranging
from major bulk terminals to filling of individual vehicles at self-
service stations. These sources are receiving special attention within
EPA because of the significance of their emissions, the potential
-------
-55-
economic impact of control on thousands of service stations, the
alternative of onboard controls, litigation on benzene under Section
112, and the importance of gasoline marketing for ozone attainment
strategies. EPA's Gasoline Marketing Task Force has developed detailed
estimates of the risk from these facilities that cover benzene,
ethylene dibromide, ethylene dichloride, and gasoline vapors. The
Task Force estimated an aggregate incidence of 43 excess cancers
per year from all gasoline marketing sources. This estimate was
used in portions of this study.
Moodstoves
As indicated in the Ambient Air Quality and 35-County studies,
products of incomplete combustion may be a significant hazardous
air pollutant problem. Recent studies suggest that residential
wood combustion contributes about 40 percent of total national
emissions of polycyclic organic matter (POM). POM compounds found
in wood smoke include BaP and polycyclic organic ketones. In
addition, one EPA study suggests that the emission rate of mutagenic
and carcinogenic substances from woodstoves is at least several
orders of magnitude greater than from other combustion sources used
to heat homes. The results of the 35-County Study supported this
concern: roughly 80 percent of the annual estimated cancer incidence
for BaP from heating in the 35 counties was attributable to wood
combusti on.
There are currently no data on the human health risks attribu-
table specifically to wood smoke. As a result, the 35-County Study
-------
-56-
assessed the potential human health hazard posed by wood com-
bustion, considering the health effects associated with only a
few individual compounds (BaP, formaldehyde, cadmium, beryllium,
and arsenic). The estimated annual cancer incidence in the 35
counties resulting from exposure to these compounds is 27, including
the use of BaP exposure as a surrogate for PIC. However, it should
be noted that the 35 counties analyzed are not representative of
those areas where wood combustion is of greatest concern, such as
parts of New England, Montana, and Colorado. Thus, the estimates
for woodstoves understate risks for such areas.
EPA has established a committee that soon will recommend
research and regulatory initiatives for woodstoves to the Agency.
These recommendations will include: a comprehensive research
program on health effects, emission testing procedures, and control
techniques; establishment of a variety of technical assistance
programs on wood smoke; and consideration of a new source performance
standard for woodstoves. The Integrated Cancer Assessment Project,
which is scheduled to begin this fall, also plans to assess the
contribution of woodstove emissions to the total organics, POM, and
mutagenic activity in the airsheds to be studied.
Sewage Treatment Plants
Sewage treatment plants have become a source of interest for
air releases primarily because of work undertaken by EPA's Integrated
Environmental Management Division (IEMD). Preliminary findings
suggest that many Publicly Owned Treatment Works (POTW's) located
in urban areas with indirect industrial dischargers may emit
volatile organic compounds in excess of 100 kkg/year. Using a POTW
algorithm developed for the 35-County Study, we estimated an annual
-------
-57-
cancer incidence in the 35 counties of 2.3 for the nine pollutants
that we were able to consider.
Given the paucity of data on air releases from sewage treat-
ment plants, there is a need to explore this area in more detail.
The IEMD will continue to monitor and model POTW's as part of its
activities in future work on geographic sites; EPA's Pretreatment
Task Force may also explore potential air emissions from sewage
treatment plants.
Hazardous Waste Combustion in Boilers
Although insufficient data were available to quantify the
problem of disposal of hazardous waste in boilers, EPA's Office of
Solid Waste (OSW) has attempted to assess the risk resulting from
the burning of hazardous waste using a model boiler approach. This
study considered three boiler sizes and estimated exposure and risk
for three metropolitan areas: New York, Cleveland, and Los Angeles.
These cities were chosen because they represent a wide variety of
exposure characteristics for densely populated, highly industrialized
areas. As information on the quantity, distribution, and toxic
content of the hazardous material burned was limited at the time
OSW initiated this analysis, this study tends to depict a worst-case
scenario. The study findings suggest that:
0 Lifetime individual risks for the most exposed individual
in these three regions range from 5xlO~6 to 1.4x10-5,
depending on the boiler type.26
26 "Draft Preliminary Risk Assessment for Burning Hazardous Waste
in Boilers." Office of Solid Waste, EPA. February 16, 1984,
p. 2.
-------
-58-
0 Lifetime risks to the average exposed individual in these
three regions ranged from l.ZxlO'7 to SxlO'7, depending on
the boiler type.
0 Estimated annual cancer incidence in these three regions
ranges from 0.01 to 0.2, depending on the boiler type.
0 The risk associated with metals is potentially much higher
than that for organics. Using metal concentrations found
in virgin fuel, the analysis shows that metals contribute
roughly 52 percent to the total estimated incidence.
Burning hazardous material with metal concentrations higher
than these could increase the problem.
OSW has also just completed the Survey of Handlers and Burners
of Used or Waste Oil and Waste-Derived Fuel Materials (Track II)
which should provide useful information for future studies on risk.
Although OSW has only begun to analyze the survey results, some
preliminary findings on the burning of waste-derived fuel material
(WDFM)27 are as follows:
0 924 million gallons of WDFM are burned each year;
0 About 200 million gallons of this material are estimated to
be hazardous, as defined by the Resource Conservation and
Recovery Act (RCRA); and,
0 Chemical manufacturing, pulp and paper, lumber, primary metals,
and petroleum refining industries burn about 90% of total WDFM.
27 "Status of the Data Collection Effort for the U.S. EPA:
Survey of Handlers and Burners of Used or Waste Oil and Waste-
Derived Fuel Material: Track II." December 1983, pp. 3-4. It
should be noted that WDFM is a broader category than hazardous
waste. For the purposes of the survey, WDFM was defined as "any
material that is a constituent of a fuel, or is destined to be
burned as a fuel, that is not a conventional fuel material."
Examples of conventional fuel are: distillate fuel oil; residual
fuel oil; natural gas; coal; liquified petroleum gas; and refuse-
derived fuels.
-------
-59-
OSW is initiating analyses to identify boiler operating prac-
tices, to characterize the specific wastes being burned, and to
determine the quantity and geographic distribution of these hazardous
wastes. This information will be used to complete an exposure and
risk assessment that will support the Regulatory Impact Analysis
for the regulation of burning hazardous waste and used oil fuels.
The tentative schedule for completing this analysis is September
1985.
Haste Oil Combustion
The Office of Solid Waste (OSW) estimates that 500 to 550
million gallons of used oil are recycled as fuels each year.28
While most of these fuels are burned in boilers, they may also be
burned in kilns, space heaters, and diesel engines. Because of
contamination during use and because of mixing, used oils typically
contain elevated levels of toxic metals--such as arsenic, cadmium,
and chromium--and organics, such as BaP and PCBs. Burning used
oils may result in elevated ambient concentrations of some of
these contaminants, particularly when the combustion sources are
clustered.29 The potential emissions of metals appear to contribute
more to risk than organic emissions. The 35-County Study corrobor-
ated the importance of metals: of a total annual cancer incidence
28 U.S.EPA, "Composition and Management of Used Oil Generated in
the U.S." December 1983.
29 U.S.EPA, "A Risk Assessment of Waste Oil Burning in Boilers
and Space Heaters." Draft, January 1984.
-------
-60-
of 6.7 from waste oil combustion in the 35 counties, 90 percent
was accounted for by chromium and 9.5 percent by arsenic.
OSW is currently developing emission standards for waste oil
combustion and will evaluate these risks more closely, for a
variety of exposure pathways.
Operational Hazardous Waste Facilities
Over the past several years, there has been an increasing
concern that operational treatment, storage, and disposal facili-
ties (TSDF's) for hazardous wastes may be an important source of
air emissions. There have been many efforts to quantify releases
of volatile organic compounds from TSDF's. In general, these
analyses have either focused on individual facilities, using
ambient monitoring to estimate atmospheric pollutant concentrations,
or on national estimates, employing emission models to assess air
releases. In addition, Westat, Inc. recently completed an extensive
survey for the Office of Solid Waste (OSW) that provides a great
deal of background information on the quantity, constituency, and
distribution of hazardous waste generated and managed by TSDF's
throughout the country. The survey estimates that a total of
71.3xl09 gallons (264xl06 metric tons) of waste are managed by
hazardous waste facilities, and that over 50 percent of this
quantity is treated, stored, and/or disposed of in impoundments
and landfills. In addition, the survey indicates that the chemical
industry generates over 70 percent of total hazardous waste. If
-------
-61-
we assume that a substantial amount of the chemical•tndustry's
waste consists of volatile organic compounds, there is a clear
potential for significant air emissions from TSDF's.
Although the survey yields some interesting findings on the
types and quantity of hazardous waste managed at TSDF's, it is
nonetheless one step removed from actual emission estimates. There
have been several recent attempts to estimate releases from TSDF's
at the national level using emission modeling. Unfortunately,
these studies have been severely criticized. It is apparent that
estimating volatilization from TSDF's is still in its infancy, and
that these models generally require further refinement and validation.
The monitoring data on ambient concentrations around specific
TSDF's are probably more persuasive in making the case that TSDF's
are potentially significant sources of air toxics. To explore the
potential hazard from the volatilization of organic compounds, we
used air toxics concentration data from a study of one TSDF, the
BKK landfill in California.30 This was the only data set found that
attempted to measure actual ambient concentrations to which people
living around the TSDF would most likely be exposed. The results
are presented in Table 7. It is important that these estimates be
30 California Department of Health Services, California Air
Resources Board and South Coast Air Quality Management District.
"Ambient Air Monitoring and Health Risk Assessment for Suspect
Human Carcinogens around the BKK Landfill in West Covina,"
1983.
-------
-62-
TABLE 7
PRELIMINARY ESTIMATES OF INCIDENCE AND INDIVIDUAL RISKS ASSOCIATED WITH
AIR RELEASES FROM ONE TREATMENT, STORAGE, AND DISPOSAL FACILITY (TSDF)
Pol 1utants Havi ng Preliminary Approximation
Some Evidence of Concentration**(ug/m3) of Individual Lifetime Risk***
Carci nogenicity* Max Min Max Min
Benze
ne 3.8
Chi oroform 1 .0
Vi nyl
Perch
ethy
1
1
Trichl
ethyl
Ethyl
dich
Total
e
1
chloride 12.1
oro- 6.8
ene
oro- 5.4
ene
ne 6.3
oride
Additive Lifetime Risk
0.0 2
0.0 1
0.0 3
0.0 1
2.1 2
0.8 4
1
.6x
.Ox
.2x
.2x
.2x
,4x
.4x
1
1
1
1
1
1
1
o-
o-
o-
o-
o-
o-
o-
5
6
5
5
5
5
4
0.
0.
0.
0.
8.6x
b.6x
1.4x
0
0
0
0
1
1
1
0-6
0-6
0-5
* The weight of evidence of carcinogenicity for the compounds listed
varies greatly, from very limited to very substantial. Further, the
extent of evaluation and health review performed varies considerably
among compounds. However, for the purposes of this report, a conser-
vative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
** Concentration data source: California Department of Health
Services, California Air Resources Board and South Coast
Air Quality Management District, "Ambient Air Monitoring and
Health Risk Assessment for Suspect Human Carcinogens Around
the BKK Landfill in West Covina," 1983.
*** Because of the uncertainties in the data used to make these estimates,
they should be regarded as rough approximations of lifetime individual
risk. These estimates are drawn from measurements made at one TSDF,
and should not be considered representative of usual TSDF emissions,
but rather illustrative of potential TSDF emissions.
-------
-63-
interpreted as an isolated example, providing only a rough indica-
tor of potential risk from TSDF's. However, the estimates
exhibited in this table suggest that risks around this landfill
are similar to those near major point sources. The lifetime
individual risks for the highest observed values range from
10~5 to 10-6, and the maximum additive lifetime individual risk for
the six compounds is 1.4 x 10~4.
Superfund Sites
There is evidence suggesting that uncontrolled or abandoned
hazardous waste facilities, e.g., Superfund sites, may be significant
sources of air toxic releases. Information provided by the Hazard
Ranking System (HRS) [40 CFR Part 300: Appendix A], is one indication
of this potent i al.
For an abandoned hazardous waste site to be listed as a Superfund
site and placed on the National Priorities List (NPL), the site must
receive a specified score using the HRS. In the HRS, air releases
must be both significantly above background concentrations and also
"observed" (that is, measured) to receive a score. In contrast,
only a "potential" for release to surface or ground water is required
in the HRS. The requirement for an observed release for air resulted
from a lack of any better method for considering the air route: no
good, consistent correlation has been found between physical and
chemical properties of wastes and their potential for air migration.
To date, 109 sites have been placed on the NPL due to high air
scores. Of these, 43 were listed for particulate, heavy metal, or
radium releases. The remaining 67 sites were listed because of
-------
-64-
volatile organic compound emissions. These 109 facilities repre-
sent a total of 16 percent of all currently listed NPL sites.
Municipal Waste Disposal: Incinerators and Landfills
Few attempts have been made to assess the risks that may be
attributable to air toxics emissions from municipal incinerators and
municipal landfills. Our search for risk assessments on municipal
waste treatment led to only one study designed specifically for
assessing risks. In this study, dioxin emissions from six municipal
incinerators were measured, and maximum individual risks estimated
at levels varying from 10~5 to 10-6.31 The investigators concluded
that the levels monitored did not present a public health hazard
for the residents living in the immediate vicinity.
In another EPA-sponsored analysis, very preliminary estimates
were made of emissions of several metals and organic compounds from
municipal incinerators. These estimates indicated that maximum
individual risks from poorly run facilities may in certain cases
exceed those measured in the dioxin risk assessment described above:
well-run facilities appear to pose risks approximately 10 to 100
times less than those of poorly run facilities.32 These latter
estimates could be made only by using a variety of assumptions,
31 Memorandum from Michael Cook (U.S. EPA Office of Solid Waste
and Emergency Response) to EPA's Regional Dioxin Coordinators,
"TCOD Emissions from Municipal Waste Combustors," December 16,
1983.
32 Personal communication from David Sussman, U.S. EPA Office of
Solid Waste, June 1984.
-------
-65-
since no systematic program has been undertaken to monitor stack
emissions from municipal incinerators for the purpose of risk
analysi s.
We were unable to identify any broad-based studies character-
izing risks due to air toxics emissions from municipal landfills.
However, there is speculation that emissions may in some cases be
high due to decomposing plastics, discarded solvents, and mobiliza-
tion of volatile organics to the atmosphere by methane gas. Two ad
hoc studies performed at municipal landfills on Long Island and
in the Los Angeles area provide preliminary confirmation of such
speculation. At the Long Island landfill, vinyl chloride was
detected in the landfill gases at 90 ppm.33 At the Los Angeles
landfill, landfill gas concentrations of vinyl chloride reached
20 to 30 ppm, and ambient levels near the landfill exceeded those
found away from the landfill.34 In addition, stack emissions
of vinyl chloride from a gas collection facility at this same Los
Angeles landfill exceeded the vinyl chloride NESHAP emission
limit (10 ppm) established for other source categories. Since
their initial detection, these emissions have been abated. The
Los Angeles air pollution control authorities are currently
conducting a monitoring program near selected Los Angeles landfills
to evaluate the need for air emissions controls.
Drinking Water Treatment Facilities
EPA's Office of Drinking Water and the Office of Policy Analysis
are studying air emissions from aeration facilities at drinking
33 Personal communication from Marcus Kantz, EPA Region 3, May 1984.
34 Personal communication from Edward Camarena, South Coast Air
Quality Management District, June 1984.
-------
-66-
water treatment plants. Aeration is used to remove volatile
organics from surface water before it is pumped to residential
communities for use.
A second issue regarding these facilities concerns potential
air emissions of chloroform from chlorination of drinking water
supplies. In a monitoring program conducted in Philadelphia by
EPA's Integrated Environmental Management Division, the highest
ambient concentrations of chloroform found in the city were
measured on the grounds of the drinking water treatment plant.
However, these findings are still preliminary and must be examined
i n greater detai1.
Sewage Sludge Incineration
EPA's Office of Water Regulations and Standards and the Office
of Policy Analysis are examining the issue of air emissions from
sewage sludge incineration. The Water Office is specifically
interested in whether the New Source Performance Standard (NSPS)
for sewage sludge incinerators promulgated under the Clean Air Act
is adequate. The NSPS regulates emissions of particulate matter, but
does not consider the potential health effects of the toxic constituents
of those emi ssi ons.
PATHWAYS
I ngestion
This study considers only the effects of inhaling toxic air
pollutants. The quantitative risks due to human ingestion of air
pollutants are not covered, although there are several examples of
-------
-67-
the ways that toxic air emissions may be ingested. In Tacoma,
Washington, researchers discovered that children living near the
ASARCO copper smelter have elevated levels of arsenic in their
urine; one possible exposure route is by ingestion of contaminated
soil. Fish in Lake Superior contain toxaphene that was deposited
in the lake after being carried by the wind from areas where toxaphene
was used as a pesticide. In Maryland, some analyses suggest that
as much as 30 percent of the metals present in the Baltimore Harbor
may have been air-deposited, either by direct deposition from the
air or indirectly through urban runoff. Finally, half of the 1,000+
chemicals inventoried in the Great Lakes appear to result at least
in part from air deposition.
Stratospheric Ozone Depletion and Skin Cancer
The analysis did not consider the possible health effects
caused by a reduction in the stratospheric ozone layer. Carbon
tetrachl ori de, and other chlorinated organics with long atmospheric
lifetimes, have the potential to affect the ozone layer, and could
indirectly increase the incidence of skin cancer. For example, it
is estimated that by the year 2020, U.S. emissions of carbon tetra-
chloride could be responsible for between 500 and 22,000 excess
cases of skin cancer annually in the U.S., resulting in 3-220
excess deaths per year-35
35 Zaragoza, L. "Calculating jEffects of Carbon Tetrachloride
and Other Chiorocarbons on Increases in Skin Cancer from
Stratospheric Ozone Depletion," EPA, OAQPS Draft. July 25, 1984
-------
-68-
C. Summary of the Magnitude of the Air Toxics Problem
Estimated risks from air toxics have been presented for each
major analytical study: the NESHAP Study, the 35-County Study, and
the Ambient Air Quality Study. The results differ among the three
studies because of differences in technical approaches, pollutants
and sources covered, and emissions estimates, making interpretation
and integration of the disparate results difficult. A useful
statistic for summarizing the results of all three studies seems to
be annual incidence per million population. Table 8 summarizes this
statistic for the 17 pollutants and pollutant groups for which size-
able risks were estimated in any of the analyses.
It should be noted that estimates were derived differently in
each of the studies: those from ambient air data weighted urban and
rural population and concentrations to arrive at a national average;
the national aggregate values calculated for the NESHAP Study and
for asbestos, radionuclides , and gasoline marketing were spread
over the total national population of 230 million; and the population
living in the 35 counties was used to calculate incidence per
million for the 35-County Study.
The estimated annual incidences per million people for the
pollutants included in this report were 5.6 for the NESHAP analysis,
7.4 for the Ambient Air Quality Study, and 4.9 for the 35-County
Study. These totals are surprisingly close. However, this closeness
is somewhat coincidental and disguises large inconsistencies in the
pol1utant-by-pol1utant estimates. For instance, chromium accounts
for only 0.29 cases per million in the 35-County study and 1.43
in the NESHAP analysis. Volatile organic compounds contribute a
total of 2.6 per million based on the ambient measurements and only
0.6 for the NESHAP data.
-------
-69-
TABLE 8
SUMMARY TABLE: PRELIMINARY APPROXIMATION OF ANNUAL INCIDENCE
ESTIMATES PER MILLION POPULATION FROM THE NESHAP STUDY, THE
AMBIENT AIR QUALITY STUDY AND THE 35-COUNTY STUDY**
Pollutants Having
Some Evidence of
Carci nogenicity*
NESHAP
Study
Ambient Ai r
Quality Study
35-County
Study
Six Month Study Risk Estimates
Formaldehyde 0.01
Benzene 0.14
Chromi urn'*' 1 .43
Cadmium 0.04
Arsenic 0.02
Trichloro-
ethylene 0.04
Perchloro-
ethylene 0.01
Ethylene oxide 0.21
Carbon tetra-
chloride 0.06
0,
1,
1,
0,
0,
83
02
05
06
26
0.08
0.10
N/A
0.19
0.21
0.39
0.29
0.02
0.02
0.15
0.14
N/A
0.004
**
The weight of evidence of carcinogenicity for the compounds listed
varies greatly, from very limited to very substantial. Further, the
extent of evaluation and health review performed varies considerably
among compounds. However, for the purposes of this report, a conser-
vative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
Because of the uncertainties in the data used to make these estimates,
they should be regarded as rough approximations of incidence. Estimates
for individual compounds are much less certain. These incidence esti-
mates have been performed to provide a rough idea of the possible total
magnitude of the air toxics problem, and will be used only for priority-
setting and to provide policy guidance.
Risk estimates assume that all species of chromium are carcinogenic,
although only certain species have evidence of carcinogenicity.
Current data do not allow differentiation among species.
-------
-70-
TABLE 8 (Cont.)
SUMMARY TABLE: PRELIMINARY APPROXIMATION OF ANNUAL INCIDENCE
ESTIMATES PER MILLION POPULATION FROM THE NESHAP STUDY, THE
AMBIENT AIR QUALITY STUDY AND THE 35-COUNTY STUDY**
Pollutants Having
Some Evidence of
Carci nogenicity*
Ethyl ene di -
bromi de
Chi orof orm
Vi nyl idene
chloride
Gasoline vapors
Al 1 other
NESHAP
Study
0.12
< 0.01
< 0.01
N/A
0.11
Ambi ent Air
Quality Study
N/A
0.07
0.27
N/A
0.01
35-County
Study
0.02
0.002
N/A
0.15
0.38
Risk Estimates from Other EPA Efforts
Radi onucl i des
Asbestos
PIC***
Gasol i ne Market i ng
0.07
0.50
2.65
0.20
0.07
0.50
2.65
0.20
TOTAL
5.6
0.07
0.50
2.60
4.9
* The weight of evidence of carcinogenicity for the compounds listed
varies greatly, from very limited to very substantial. Further, the
extent of evaluation and health review performed varies considerably
among compounds. However, for the purposes of this report, a conser-
vative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
** Because of the uncertainties in the da'ta used to make these estimates,
they should be regarded as rough approximations of incidence. Estimates
for individual compounds are much less certain. These incidence esti-
mates have been performed to provide a rough idea of the possible total
magnitude of the air toxics problem, and will be used only for priority-
setting and to provide policy guidance.
*** "Products of Incomplete Combustion" (PIC) refers to a large number of
compounds, probably consisting primarily of polynuclear oryanics. The
PIC unit risk value was derived from dose-response data that use BaP
levels as a surrogate for PIC or total air pollution. There are many
limitations of using the BaP surrogate method to estimate PIC risks.
Thus, all PIC estimates presented in this report must be regarded as
highly uncertain. Refer to pp. 20-25 for a more detailed explanation
of how the PIC unit risk value was derived.
-------
-71-
A major contributor to these estimates is the pollutant cate-
gory we have labeled products of incomplete combustion (PIC). It
is unique among the pollutants examined and deserves special mention.
PIC is used in this study to represent a large number of air pollu-
tants associated with lung cancer in epidemiological studies of
people exposed to those pollutants in the 1940's and 1950's. We
assumed that these exposures were dominated by PIC. The unit risk
factor was derived by using BaP as a surrogate for PIC, and is
based on these epidemiological studies. This method of quantifying
risk is unusual, and the fact that major risks are estimated for
PIC makes the calculation controversial. The alternative is to
exclude PIC and to ignore the implications of the epidemiological
studies and the contribution of these compounds, some of which are
proven carcinogens. More detail on the derivation of the unit risk
value for PIC is provided on pages 20 to 2b.
Although incidence per million population is an important
statistic, aggregate national totals also provide perspective and
allow comparison with other cancer statistics. The annual inci-
dence estimates derived from the incidence rate for the major analyses
(Table 8) are:
NESHAP Study - 1,300 (national estimate)
Ambient Air Quality Study - 1,700 (national estimate)
35-County Study - 230 (for 35 counties only)
Most of our analyses also estimated individual lifetime risk.
As opposed to aggregate incidence, which applies to an entire popu-
lation, individual lifetime risk describes the risk to a specific
individual at a specific location (usually the worst-case site).
It almost always occurs within 0.1 km and 0.3 km from the fenceline
-------
-72-
of major sources. The values are very susceptible to errors in
modeling assumptions, population location, and emission estimates,
and it is difficult to interpret the results of national studies.
In our analysis, maximum risks near point sources frequently reached
one in a thousand (10~3) or greater and were routinely greater than
10-4. For example, in the NESHAP study, 13 pollutants presented a
maximum individual risk of IxlO'3 or greater in at least one location,
and 21 pollutants (nearly half of those studied in the NESHAP
analysis) presented risks greater than IxlO'4.
The ambient air data were used to calculate an aggregate
individual risk for multi-pollutant exposures. Since these
aggregate individual risks were based on measured data for a
specific sampling site, they were subject to less uncertainty than
most of the risk estimates in this report and may be used as an
important indicator of the general magnitude of the urban air
toxics problem. However, the amount of data available falls short
of that needed for a comprehensive analysis of any of the urban
areas, and the results should not be used for city-to-city comparison.
Since reasonably complete monitoring data were needed to
estimate these aggregate risks, only a few urban areas with the best
data bases could be included. Generally, these were large cities
with medium to heavy industrialization. The additive risks ranged
from 0.7xlO-3 to 2.3xlO'3, based on measurements of two to three metals,
BaP as an indicator for PIC's, and 6 to 10 volatile organics
monitored at the same or very proximate locations (Table 5). These
locations generally were in city centers and were not associated
with specific point sources.
It is not possible to estimate the number of people exposed
to such multi-pollutant risks. However, it is interesting to
compare them to the estimates of annual incidence per million
-------
-73-
reported earlier. A lifetime risk of 2.3x10-3 equals 2,300
excess cancer cases per million population for a 70-year period,
or 33 per million per year; a lifetime risk of 0.7x10-3 equals
about 10 per million cancer cases per year.
D. Perspective and Context; Other Cancer Risks
One way to evaluate the importance of the air toxics risks
described above is to compare them with risks linked to other
factors. For example, Doll and Peto estimate that about 65 percent
(286,000) of annual cancer deaths appear to be related to smoking
(30 percent) or diet (35 percent), and that about 2 percent of total
cancer deaths (8,800) are associated with environmental pollution.36
The magnitude of the air toxics problem presented in this study
is given for PIC in terms of cancer deaths, and as cancer cases for
other pollutants. Therefore, these risks should be compared both
to statistics regarding both total cancer cases and cancer deaths.
Table 9 presents projected estimates of 1983 cancer mortality and
morbidity made by the American Cancer Society (ACS).37 This table
shows that about 850,000 cancer cases and 440,000 cancer deaths
were projected for 1983. The ACS reports also that 135,000 lung
cancer cases and 117,000 lung cancer deaths were projected for 1983.
If indoor air exposures are considered, this study may not
accurately estimate the potential number of cancers associated
36 Doll, Richard, and Richard Peto, "The Causes of Cancer:
Quantitative Estimates of Avoidable -Risks of Cancer in the United
States Today," Journal of the National Cancer Institute. June 1981
37 American Cancer Society, 1982. Cancer facts and figures, 1983.
-------
-74-
TABLE 9
PERSPECTIVE AND CONTEXT: STATISTICS ON CANCER RISKS
TOTAL ESTIMATED CANCER CASES (1983)1 850,000
TOTAL ESTIMATED CANCER DEATHS (1983)1 440,000
Diet2 154,000
Smoking2 132,000
Environmental pollution? 8,800
CANCER CASES ASSOCIATED WITH INDOOR AIR EXPOSURES
Radon3
Passive smoki ng^
Formaldehyde^
Other organic compounds^
Carbon tetrachloride
Benzene
Chioroform
Tetrachloroethylene
Trichloroethyl ene
1 ,000
500
to 20,000
to 5,000
160
(3,700/million)
(1 ,900/mil lion)
(670/million)
(570/million)
(38/mi 11 ion)
(4 to 91/million)
(2.2 to 22/million)
(0.7/million)
340
500
240
200
220
(1 .5/mil
(2.2/mil
(1 .I/mil
(0.9/mil
(1 .0/mil
ion)
i on)
i on)
i on)
i on)
Source: American Cancer Society, 1982.
Figures, 1983.
Cancer Facts and
2 These estimates are presented for illustrative purposes only,
since many consider that such attribution of cancer cases to
a particular exposure oversimplifies the multi-causal nature
of cancer. The estimates were derived by combining the estimated
percent of cancer deaths attributed to diet, smoking, and pollu-
tion presented in Doll and Peto (reference 35) with the American
Cancer Society estimates of total 1983 cancer deaths.
3 These estimates were made by the relevant EPA program offices.
For specific references and a discussion of these estimates,
refer to Thomson, Vivian, "Indoor Air Pollution: Ramifications
for Assessing the Magnitude and Nature of the Air Toxics Problem
in the United States," U.S. EPA Office of Policy Analysis,
September 1984.
Repace, J. L
No n-smokers'
, , and A. H.
Lung Cancer
Lowrey, "A Quantitative Estimate of
Risk from Passive Smoking," EPA, in
press
These values are calculated from personal exposure monitoring data
collected in EPA's Total Exposure Assessment Methodology Study (TEAM),
which weights them heavily toward the impact of indoor exposures.
They are taken from: Wallace, Lance, "Review of Air Toxic Docume
Memorandum to Bern Steigerwald, EPA, October 30, 1984.
nt,"
-------
-75-
with exposures to toxic air pollutants. Historically, indoor,
nonoccupational air quality has been virtually ignored by EPA and
other federal agencies, despite the fact that average Americans
spend about 80 to 90 percent of their time indoors. Recent data
show that indoor radon exposures may cause from 1,000 to 20,000
lung cancer cases annually, and EPA estimates show that 500 to
5,000 cancer cases may be caused by passive smoking.38 In addition,
indoor levels of formaldehyde routinely exceed outdoor levels by an
order of magnitude, while indoor levels of other organics--such as
benzene, trichloroethylene , and tetrachloroethylene--may exceed
outdoor levels by 2 to 5 times for the median-exposed individual
and up to 50 times for the most-exposed individual.39 Preliminary
risk estimates (Table 9) for indoor plus outdoor exposures to five
organic compounds greatly exceed those based on ambient levels only
(Table 5). Combined with the large amount of time that Americans
spend indoors, these data indicate that our estimates of the magnitude
of the air toxics probl em--based only on outdoor ambient levels — may
understate the extent of the air toxics problem for those compounds
that can be emitted indoors.
It is also possible that our analysis has somewhat overstated
risks due to" the metals examined in the study. No indoor versus
outdoor data could be found for the specific metals examined in
this study. However, there are limited data indicating that other
trace metals (e.g., vanadium, manganese) show indoor/outdoor ratios
somewhat less than l.O.38
38 Thomson, Vivian, "Indoor Air Pollution: Ramifications for
Assessing the Magnitude and Nature of the Air Toxics Problem
in the United States," U.S. EPA Office of Policy Analysis,
September 1984.
39 Wallace, Lance et al. "Total Exposure Assessment Methodology
(TEAM) Study: First Season - Northern New Jersey." Interim
Report. U.S. EPA, Office of Research and Development.
-------
-76-
Limited data are available characterizing the cancer risks
due to ambient environmental exposures other than air pollution.
As part of this study, EPA's Chemical Coordination Staff attempted
to compare risk levels triggering regulation across several of
EPA's program offices. The staff concluded that such comparisons
are difficult to make, since EPA has in fact made few regulatory
decisions for carcinogens based on quantitative risk assessment.
However, a few examples of risk-assessment based decisions were
found. For instance, EPA recently banned most uses of the pesticide
ethylene dibromide after estimating that EDB exposures might
cause as many as 13,000 cancer cases per year. EPA has also
banned most uses of chlordane/heptachlor, based on estimates of
500 cancer cases caused annually, and the asbestos school inspection
program was started after risks were estimated at approximately
60 cancer cases annually.40
As previously discussed, the maximum individual risks estimated
in this study ranged widely, from 10'1 to less than 10'5- Risks of
10~3 and greater were commonly estimated for major point sources, and
the combined lifetime individual risks based on ambient data were
in the 10"3 range. The Chemical Coordination Staff's analysis
shows that, on average, EPA has taken regulatory action based on
Viviani, Donn et al., "Acceptable Risk Levels and Federal Regula-
tions: A Comparison of National Emission Standards for Hazardous
Air Pollutants (NESHAP) with Other Federal Standards Based on
Quantitative Risk Assessment," U.S. EPA Office of Pesticides
and Toxic Substances, May 1984.
-------
-77-
maximum individual risks in the 10~3 to 10-4 range, although there
may be differences among program offices:
Although the data are somewhat limited, the Office of Air and
Radiation generally appears to use a marginally higher level
of individual risk (both before and after regulation) than
other offices. However, when viewed from an aggregate risk
perspective, risks to the total population are not much
different from those of other offices.39
V. NATURE OF THE AIR TOXICS PROBLEM
Whereas previous sections of this report focused on the
magnitude of the national air toxics problem, the following section
will discuss the causes of air toxics exposures and risks. Four
questions will be addressed, using the results of the studies and
analyses previously discussed:
1. What pollutants appear to cause most of the air toxics problem
as we understand it now?
2. What sources appear to be major contributors to air toxics
risks?
3. Do air toxics problems vary geographically?
4. Can we estimate the degree to which indirect control of
air toxics is affected through the criteria pollutant programs?
A. Pol 1 utants
Table 8 shows that approximately 15 pollutants and pollutant
groups account for most of the cancer risks examined in this study:
PIC, chromium, benzene, arsenic, cadmium, carbon tetrachloride,
chloroform, ethylene dibromide, ethylene oxide, formaldehyde, gaso-
line vapors, perchloroethylene, trichloroethylene, asbestos, and
-------
-78-
radionuclides. Thus, it appears that the pollutants responsible for
most of the cancer cases associated with air toxics consist of a
mixture of metals, volatile organic compounds, and products of incom-
plete combustion. Many of these same pollutants (for example,
chromium, benzene, ethylene oxide, and arsenic) also show maximum
individual risks in the 10"1 to 10~3 range.
An interesting feature of the analysis is the relatively low
aggregate risk estimated for many of the synthetic organic chemicals:
national incidence totalled less than 1.0 cancer cases per year for
21 such compounds. This fact is noteworthy, since it has been
speculated that such "exotic" chemicals may be major sources of air
toxics risks. However, these low incidence estimates are based
on exposure modeling, and have not been verified by ambient data.
In addition, maximum individual risks associated with some of these
chemicals ranged up to 10~3.
B. Sources
Not surprisingly, an examination of emissions associated with
the pollutants listed above shows a diverse and complex group of
sources. Table 10 gives a source breakdown for several of the more
important pollutants examined in the study. For example, chromium
is emitted from such major point sources as steel and refractory
manufacturing facilities, as well as from fuel combustion. Formalde-
hyde is emitted from mobile sources, chemical plants, fuel combustion,
indoor sources (such as part i cl eboard) , and is formed photocherni cal 1 y
in the atmosphere.
-------
-79-
TABLE 10
SOURCES OF SELECTED COMPOUNDS EXAMINED IN THIS STUDY
Pollutant
Sources
Arseni c
Benzene
Chioroform
Chromi urn
Ethylene oxide
Formaldehyde
Perchloroethyl ene
PIC*
Tri chloroethyl ene
Combustion sources such as waste oil
burning, utility boilers (coal-fired) ,
wood smoke, smelters, glass manufacturing
Road vehicles, gasoline marketing,
petrol eum ref i ni ny
Solvent usage, water treatment
Waste oil burning, steel manufacturing,
refractory manufacturing, metals
manufacturing, combustion sources
Chemical industry, sterilant
Road vehicles, formaldehyde manufacturing,
petroleum refining, oil and gas combustion
Solvent usage, dry cleaning facilities
BaP sources include use of wood and coal
in small combustion units, coke operations,
internal combustion engines
Solvent usage
* "PIC" is shorthand for Products of Incomplete Combustion, a broad
and ill-defined group of compounds represented in this study by
BaP, an organic particulate. The mix of compounds present will
vary from different combustion processes.
-------
-80-
The complexity and diversity of air toxics sources are under-
scored by the following observations concerning emissions of the
most significant pollutants listed in Table 8.41
- Manufacturing facilities for synthetic organic chemicals are
responsible for greater than 20 percent of total national
emissions for only 3 of the major pollutants.
- Mobile sources account for greater than 20 percent of emissions
for only 3 of the major pollutants.
- Solvent usage is responsible for greater than 20 percent of
emissions for only 3 of the major pollutants.
- Fuel combustion in stationary sources accounts for greater than
20 percent of emissions for only 4 of the major pollutants.
Another perspective on which source types appear to be important
contributors to the air toxics problem can be had by using the
individual risk or incidence estimates from the NESHAP and the 35-
County Studies. For pollutants that were evaluated directly, area
and point sources each accounted for about half of the aggregate
incidence in both the NESHAP and 35-County Study. When PIC is
included (using BaP as a surrogate) area sources become more dominant,
accounting for over 75 percent of the incidence in both the 35-County
and NESHAP studies. This result is consistent with the fact that PIC
is estimated to account for a large portion of aggregate incidence,
and that nearly all BaP emissions appear to come from area sources
(principally motor vehicles, and combustion of wood, coal, and oil
in small heating units). The contribution of the most significant
41 Lahre, Tom, "Characterization of Available Nationwide Air Toxics
Emissions Data," EPA Contract No. 68-02-3513, Task No. 46,
June 1984.
-------
-81-
source types based on cancer incidence as determined by the 35-County
Study are shown in Table 11.
The high proportion of total incidence that apparently is due to
road vehicles as shown in Table 11 merits additional discussion. To
estimate incidence attributable to PIC, the 35-County Study used
emission estimates to model ambient levels of BaP and then applied
the PIC unit risk factor to those modeled ambient concentrations.
An alternative method recently brought to our attention applied a
diesel emission potency that relates expected cancer incidence
specifically to diesel particulate emissions.42»42a EPA's Office of
Mobile Sources has applied this alternative technique and estimated
that after implementation of EPA's recent rule on heavy duty diesel
particulate emissions, the incidence rate in the year 2000 would be
1-4 per million population for urban areas. By comparison, the
35-County Study estimated a 1982 incidence of 120 for all mobile
vehicle emissions (including gasoline engines) for 45 million people
or a rate of 2.7 per million.
The second measure of risk used in this study is maximum individual
risk. The NESHAP Study indicates that the highest individual risk
for a pollutant is generally associated with large point sources.
42 "Control of Air Pollution From New New Motor Vehicles and New
Motor Vehicle Engines; Gaseous Emission and Particulate Emission
Regulations," 40 CFR Parts 86 and 600, Vol. 50, No. 51, Friday,
March 15, 1985.
42a "Diesel Particulate Study," U.S. Environmental Protection Agency,
Office of Mobile Sources, Emission Control Technology Division,
October 1983.
-------
PERCENT OF I
SOURC
Point Sources
Chemicals Production
Metals Manufacturing
Petrol eum Ref i ni ng
Rubber Production
Utilities
POTWs
All Other
TOTAL PERCENT: POINT
Area Sources
Road Vehicles
Solvent Usage
Gasol i ne Market i ng
Waste Oi 1 Bur ni ng
Heati ng
Wood smoke (stoves/
Al 1 other
TOTAL PERCENT: AREA
-82-
TABLE 11
NCIDENCE ASSOCIATED WITH POINT
ES BASED ON THE 35-COUNTY STUDY
% Total
I nci dence
(w/o PIC)
11
8
5
5
4
3
11
SOURCES 47
23
1 1
9
9
fireplaces) 0.5
1.5
SOURCES 53
AND AREA
*
% TotfaHF
I nci dence
(w/PIC)
4
3
2
2
1
1
4
15
60
4
3
3
12
3
85
* Because of the uncertainties in the incidence estimates used to
derive these estimates, they should be regarded as rough indicators
only. These computations have been performed to provide a rough
idea'of the nature of the air toxics problem, and will be used only
for priority-setting and to provide policy guidance.
-------
-83-
C. Geographic Variability
A final method of characterizing the nature of the air toxics
problem is to examine geographic variability in ambient air quality
and in cancer risks from air toxics. Mean ambient concentrations
for selected metals and organic compounds are shown for several
cities in Table 12. These data may be for different years and are
not for matched sites; therefore, detailed comparison is not war-
ranted. However, they do indicate that ambient levels of toxic air
pollutants can vary widely from city to city, with ratios commonly
ranging from 5/1 to 10/1.
Although information is scarce, the data available suggest
that intercity variation of risk also may be significant. Even
with sparse monitoring networks, limited geographical areas within
a city are observed with air quality for many air toxics 3 to 10
times the urban average. Two peer reviewers noted the existence of
such air quality variation in metropolitan areas and commented on
their possible importance in evaluating and regulating the problem
of air toxics.43,44
43 Ferrand, E.F., Department of Environmental Protection, City of
New York, letter to Dr. Terry Yosie, Director EPA Science
Advisory Board; December 18, 1984.
44 Lioy, P.J., New York University Medical Center, letter to Dr.
Terry Yosie, EPA; October 23, 1934.
-------
-84-
TABLE 12
COMPARISON OF MEASURED AIR QUALITY FOR SELECTED CITIES
AND POLLUTANTS
City
Pollutant A
Arsenic* 7.4
Benzo(a)pyrene* 1.7
Chromium* 93.5
B C D E F G
3.7 3.2 33.5 7.0 6.0
0.5 0.2 0.3 0.2 0.4
9.3 25.3 13.4 17.0 60.0
Benzene**
11.0
14.8
15.7
9.5
Carbon tetra-
chloride** 4.2
0.3
2.4
2.6
Chloroform1*
9.9
0.4
1.5
7.9
Trichloro-
ethylene**
1.4
2.0
0.4
2.8
* Concentrations expressed in nanograms/m3.
** Concentrations expressed in micrograms/m3.
-------
-85-
The 35-County Study also allowed us to examine the ways in
which risks vary among counties. The results are shown in Table
13 (PIC was excluded from this data set because the uncertainty
in the emission estimates for BaP make detailed city-specific
comparisons especially unreliable). For example, the percent of
risk from point sources varies from 52 percent in County 4 to
25 percent in County 2. Similarly, petroleum refining accounts
for 22 percent of total risk in County 2, but 0 percent in Counties
3 and 4. There are, however, source categories (road vehicles
and waste oil burning) that account for approximately the same
percent of risks across counties, primarily because these risks
are strongly linked to population. Thus, two main types of
sources appear to emerge from the analysis: sources accounting
for approximately equal portions of risk from one area to the
next, and sources peculiar to a particular area. While the
data bases used in these analyses are inadequate to accurately
define most areas' air toxics problems, the data do support the
intuitive prediction that reducing cancer risks from air toxics
will necessitate dealing with certain types of problems at the
1ocal 1evel.
If we consider air toxics emissions data, we also find regional
variation. For example, of the 93 compounds covered in the emissions
study45, a large concentration of organic substances were found
to be produced in an area stretching from Corpus Christi, Texas to
New Orleans, Louisiana. Eighteen organic compounds are produced
45 Lahre, Tom, "Characterization of Available Nationwide Air
Toxics Emissions Data," EPA Contract No. 68-02-3513, Task No.
46, June 1984.
-------
TABLE 13
COMPARISON OF SOURCES OF RISK IN SEVERAL COUNTIES SELECTED FROM 35-COUNTY STUDYl,2
All 35
County 1 County 2 County 3 County 4 County 5 Counties Combined
Percent of incidence from j*rea sources, point sources, and POTW's
Area
Point
POTW's
Percent of incidence
Road vehicles
Petroleum refining
61
38
1
from given
31
13
Chemical production 5
Solvent usage
Waste oi 1 burni ng
Percent of incidence
Formaldehyde
Chromium
Be nze ne
Vinyl chloride
Perchloroethylene
1 For pollutants
2 Because of the
8
8
from given
18
9
30
2
10
evaluated
66
25
9
source categories
26
1
3
18
11
pollutants
7
14
24
0
1U
directly; excludes PIC.
48
50
2
23
22
21
5
9
29
8
24
2
3
uncertainties in the incidence estimates
41
52
7
14
0
24
10
12
5
10
20
25
6
used to derive
67
32
1
31
0
2
17
10
30
12
25
0
11
these estimates, tt
51
46
3
23
5
10
10
8
12
17
23
11
8
ley should
be regarded as rough indicators only. These computations have been performed to provide a rough
idea of the nature of the air toxics problem, and will be used only for priority-setting and to
provide policy guidance.
oo
-------
-87-
entirely in Texas and Louisiana, and almost 50 percent of the
remaining organic compounds examined in the emissions study are
manufactured in those two states. As noted earlier in the report,
emissions of many of the synthetic organics are associated with
only very low annual incidence.
D. Indirect Control of Air Toxics
Toxic compounds are emitted into the atmosphere from many
sources that are controlled for criteria pollutants (EPA's criteria
pollutants are: carbon monoxide, ozone, lead, total suspended
particulates, oxides of nitrogen, and sulfur dioxide). Metals and
polynuclear compounds usually are emitted as particulate matter and
most of the volatile organic compounds are ozone precursors. As
such, they are regulated under State Implementation Plans (SIP's),
New Source Performance Standard (NSPS) program, and Title II for
motor vehicles. Also, emissions of some of the compounds--especi-
ally solvents--are accomplished for economic reasons to recover
lost product or energy.
In attempting to evaluate available analyses on the effects of
such indirect control of toxic air compounds, we found two studies.
One focused on nine potential air toxics (including benzene,
chloroform, and chromium) and evaluated the impact of existing
regulations on major point sources. Control of metals from point
sources was generally high, ranging from 80 to 98 percent. Much
more variation and less control was apparent for organics, with the
percentage control ranging from 30 to 90 percent.
46 Lahre, Tom. Op. cit.
-------
-88-
A second study was less quantitative but provided estimates
for 37 compounds and included area sources and- motor vehicles. Air
quality trends, rather than control regulations, were evaluated to
estimate the indirect control of metallic part i cul ates. Generally,
reductions of 30 to 70 percent have been observed since the 1960's.
In addition, SIP's and NSPS are credited with reducing emissions of
15 chemicals from the chemical industry by 10 to 80% and 8 solvents
by 30% nationwide. Motor vehicle controls remove up to 90% of
several potentially toxic compounds from exhaust gases.
A more recent analysis compared air quality and emissions data
for 1970 with the estimates of incidence for 1980 provided for this
report in the Ambient Air Quality Study analysis.47 Methods, assump-
tions, and pollutants included were held constant over the period.
The calculations showed a significant decrease in incidence during
the decade due to improvements in air quality, presumably related to
general regulatory programs. The annual incidence rate for the 16
pollutants studied dropped from 17.5 per million using 1970 data to
6.8 per million in 1980. Estimated nationwide incidence decreased
from 3600 in 1970 to 1600 in 1980.
Even from these cursory analyses, it is apparent that indirect
control can be very significant for toxic compounds. At this time,
it appears that control under criteria pollutant provisions of the
Clean Air Act far exceeds the impact of Section 112 regulations.
Also, since sources are already controlled by criteria pollutant
programs, the remaining risks will probably be more difficult to
control.
47 Hunt, U. F., Faoro, R. B. and Curran, T. C., "Estimation of
Cancer Incidence Cases and Rates for Selected Toxic Air Pol-
lutants Using Ambient Air Pollution Data, 1970 vs. 1980,"
U.S. EPA. April 1985.
-------
-89-
VI. ADEQUACY OF DATA BASES
Quantitatively assessing risks from air toxics exposures
poses two principal informational problems. The first involves
basic health factors, such as evidence of carcinogenicity, potency,
the presence or absence of thresholds, and synergism. These are
well-known knowledge gaps basic to cancer risk assessment and
strategic discussions on air toxics will not influence their resolu-
tion. No attempt was made in this study to use new assumptions or
procedures regarding health effects; we relied on techniques and
methods in use across EPA.
In the short term, the more relevant problem to understanding
the air toxics issue is lack of information on emissions and air
quality. These data gaps make it difficult to clearly define
problems for many situations and impede policy discussions on risk
assessment. The problem is widely recognized and universally
frustrating. In the poll of state and local agencies, we interviewed
10 agencies in depth on their air toxics problems. All perceived a
need for better emissions data. The contractor who conducted the
interviews concluded that "The agencies do not seem to have adequate
data that would enable them to perform risk assessments for the
toxic pollutants emitted."48
With the exception of radonuclides, the study consistently
found major weaknesses in the data base for air toxics, both in the
48 Radian Corp., "Definition of the Air Toxics Problem at the
State/Local Level," EPA Contract No. 68-02-3513; Work
Assignment 45, June 1984.
-------
-90-
coverage and in the quality of information available. If more
than one source of data existed, inconsistencies were the norm.
Most of the air quality data could not be used for population
exposure and were clearly not obtained for risk assessment purposes.
Many potentially large source categories could not even be included
in the study due to a lack of data. These sources included
incineration, hazardous waste disposal, atmospheric formation,
and Superfund sites.
Today, air quality data are generally collected to determine
trends for criteria pollutants; very few data are available for
developing population exposure estimates for toxic air pollutants.
Despite significant efforts to assemble monitoring data for all
sources, this analysis could only cover about 18 pollutants.
Several observat i.ons regarding the air quality data are as follows:
0 More air quality data were found for metals than for BaP
or volatile organics. However, while 170 counties with a
total population of about 60 million had monitoring data,
only 30 counties had data for more than one site, and essenti-
ally no measurements were optimal for exposure assessment.
0 Data for BaP were found for about 50- counties. However,
most of the measurements were taken three to five years ago
and only two areas had data for more than one site.
0 For volatile organic compounds, EPA's Office of Air Quality
Planning and Standards evaluated over 250 references with
thousands of entries for over 40 pollutants. However, even
with the most relaxed criteria for data completeness, only
five cities had data that allowed estimates of annual averages
for more than one site, and two of those five had data only
because of the monitoring programs conducted as part of multi-
media studies by EPA's Integrated Environmental Management
Division.
EPA does not routinely measure ambient levels of potentially
toxic VOCs, and only a few states--e.g . , California--routinely
gather such data. Of the available reports examined for this
analysis, most involve spot measurements for 24 hours or less
as part of a special project. Only 45 areas in the nation had
one valid calendar quarter's worth of data for any toxic VOC,
and only 12 areas had two valid quarters of 5 days each.
-------
-91-
Emission inventories for toxic compounds also have major
problems. About 250 references were evaluated in this study.
Based on this analysis, the most significant concerns were:49
0 inconsistent coverage of sources;
0 highly variable emission estimates;
0 poorly defined source categories;
0 obvious anomalies and gaps;
0 form of metals not shown (speciation);
0 poor coverage of dispersive end uses, e.g., solvents; and
0 changing data base with time.
To quantify the quality of the emission data available, the
reviewers assigned a Confidence Score to each of the 93 pollutants.
This subjective rating system is commonly used in evaluating
emission inventories. The reviewers' scores are summarized
below.
0 5 pollutants, "A" (consistent among
information sources; recent detailed study);
22, "B" (reasonable agreement among several
information sources);
0 59, "C" (sketchy data or significant variability
in the estimates);
° 7, "D" (virtually no information found).
The detailed report on emissions also discusses some examples
of inconsistencies found in the data. For example, five references
|_ahre Tom, "Characterization of Available Nationwide Air Toxics
Emissions Data," EPA Contract No. 68-02-3513, Task No. 46,
June 1984.
-------
-92-
were found for chloroform with emissions ranging from 3,999
kkg/year to 11,800 kkg/year (kkg = 1,000 kilograms). For chloro-
form, the subcategory of solvent use accounted for percentages of
total emissions ranging from 6.2 to 92 percent in the various
studies and production emissions varied from 1.7 to 11.7 percent.
Water chlorination was mentioned as a source of chloroform emissions
i n only one study.
Not only are emissions data scarce and often inconsistent, but
systems and institutions are not in place to collect, store, or
retrieve data that may become available. There is an almost complete
lack of standardization, definition, and data systems. If data are
collected, they are collected for a single, short-term purpose.
For monitoring programs, there are no standard methods or
guidance available on network design, siting of monitors, and
averaging times. The Aerometric Information Retrieval System is
being developed by EPA, but until it becomes available in 1987,
there is no central repository for air toxics monitoring data.
A comparison with criteria pollutants helps explain why the
data base for toxics is relatively inadequate. There are eight
pollutants or pollutant categories tracked or regulated under SIPs,
while toxic compounds of interest number from 50 to 100. About $30
million per year of EPA grants to state and local agencies are used
for gathering data on criteria pollutants, while only about $1 mil-
lion is used for air toxics. In addition, ambient concentrations of
toxics are almost always 100 times less than those of the criteria
pollutants. Metals, such as chromium and cadmium, are rarely seen at
0.01 ug/m3, whereas TSP is measured in tens of ug/m3. The TSP primary
-------
-93-
annual ambient standard is set at 75 ug/m3. Regulation of criteria
pollutants is based simply on attainment of a uniform ambient
level everywhere. However, toxics regulation often is driven by
risk analysis, which requires population exposure estimates and,
therefore, a more comprehensive data base. Institutional support
has been developed for criteria pollutants over a period of two
decades. This infrastructure includes data systems for ambient
and emissions data, regulations requiring monitoring networks and
comprehensive emission inventories, standard methods of sampling
and analysis, and formal quality assurance programs. None of
these are yet available for air toxics.
-------
-94-
VII. CONCLUSIONS
Given that this analysis was a scoping effort undertaken for
purposes of orientation and not to directly support regulation, and
considering the omissions and uncertainties discussed in this
report, the Study Team believes that the following conclusions
can be drawn from this study:
1. Both point sources (major industrial sources) and
area sources (smaller sources that may be widespread
across a given area, such as solvent usage and motor
vehicles) appear to contribute significantly to the air
toxics problem. Large point sources are associated with
many high individual risks, while area sources appear to
be responsible for the majority of aggregate cancer
i ncidence.
2. While there is considerable uncertainty associated
with the risk estimates for some substances, available
data indicated that the following pollutants may be
important contributors to aggregate cancer incidence
from air toxics: metals, especially chromium and
arsenic; asbestos; products of incomplete combustion;
formaldehyde; benzene; ethylene oxide; gasoline vapors;
and chlorinated organic compounds such as chloroform,
carbon tetrachloride , perchloroethylene, and trichloro-
ethylene; and vinylidene chloride.
3. A wide variety of sources contributes to individual
risk and aggregate incidence from air toxics. These
include: road vehicles; combustion of coal and oil;
woodstoves; metallurgical industries; chemical produc-
tion and manufacturing; gasoline marketing; solvent
usage; and waste oil disposal. As a broad category of
activities, combustion/incineration is probably the
largest single source of risk.
4. For those cities with sufficient data for analysis,
large city-to-city and neighborhood-to-neighborhood
variation in pollutant levels and sources was found.
However, our current air toxics data base is inadequate
to accurately characterize most local air toxics problems.
5. Three analyses quantified estimated cancer risks due to
15 to 45 toxic air pollutants (the number of pollutants
-------
-95-
examined varied with the different analyses). The
estimates from these analyses showed a range of 5 to
7.4 cases of cancer per million people per year (1,300
to 1,700 cases annually nationwide) for the pollu-
tants examined. These are not actual predictions
of incidence, but are instead a statistical way to
represent the aggregate risks of pollutants and
sources.
The reader is reminded that these estimates are
highly uncertain, and is cautioned that the conver-
gence of the various analyses on a seemingly narrow
range (5 to 7.5 cases per million) is somewhat coinci-
dental, given that estimates for individual compounds
varied widely among the different analyses.
For perspective, estimated nationwide cancer cases
and cancer deaths for 1983 were 850,000 and 440,000,
respectively.
6. Maximum lifetime individual risks of 10~4 (1 in 10,000)
or greater in the vicinity of major point sources were
estimated for 21 pollutants, about half of those that
were studied. Maximum lifetime individual risks of 10~3
(1 in 1,000) or greater were estimated for 13 pollutants.
7. Additive lifetime individual risks in urban areas due
to simultaneous exposure to 10 to 15 pollutants ranged
from 10~3 to 10~4. These risks, which were calculated
from monitoring data, did not appear to be related to
specific point sources. Instead, they represent a
portion of the total risks associated with the complex
pollutant mixtures typical of urban ambient air.
8. Some low-production organic chemicals appeared to
contribute little to aggregate incidence: 21 organic
chemicals were estimated to account for a total of
less than 1.0 cancer cases per year nationwide.
However, this conclusion may be due in part to the lack
of data concerning the emissions and toxicity of these
"exoti c" chemi cals.
Some of these low-production compounds did appear to
be associated with high individual risks. For example,
the maximum lifetime individual risk for 4,4,-methylene
dianiline was estimated at 1.5 X 10~3.
9. The study indicated that "non-traditional" sources of
air toxics--such as publicly owned treatment works (POTW's)
and hazardous waste treatment, storage and disposal
-------
-96-
facilities (TSDF's)--may pose important risks in
some locations. For instance, preliminary findings
suggest that POTW's with industrial indirect dis-
charges may emit volatile organic compounds in excess
of 100 kkg/yr. Individual lifetime risks for a single
compound at one TSDF were estimated as high as 10~5.
10. EPA's criteria pollutant programs appear to have done
more to reduce air toxics levels than have regulatory
actions aimed at specific toxic compounds. An analysis
of 16 pollutants was made using both monitoring and
emission data in order to evaluate progress between 1970
and 1980. The estimated cancer incidence rate for these
air pollutants in 1980 was less than half that for 1970:
6.8 per million per year in 1980, compared to 17.5 per
million in 1970. This seems reasonable considering the
diverse array of air toxics sources, the multi-pollutant
nature of the problem, and the relative intensity of
EPA's criteria and air toxics programs.
11. Even after regulations are implemented under Section 112
of the Clean Air Act for benzene and arsenic, these
pollutants still appear to make significant contributions
to aggregate incidence due to air toxics. This seems
to demonstrate that the base for the air toxics regulatory
programs needs to be broadened to include em-issions from
small area sources, such as combustion, road vehicles,
a nd sol vent use.
12. Major weaknesses and gaps characterize air toxics data
bases at the federal, state, and local levels. The
few air toxics emission inventories available generally
show inconsistencies and anomalies, the air quality data
available are often inadequate to develop population
exposure estimates, and few compounds have been tested
adequately for health effects. The data limitations
preclude performing specific comprehensive risk assessments
for most urban areas, for many compounds, and for many
potentially large sources of air toxics risks (such as
incineration, hazardous waste disposal, indoor exposures,
atmospheric transformation, and Superfund sites).
Sources of Uncertainty
Many assumptions and extrapolations are necessary to transform
ambient or modeled levels of air pollutants into exposure estimates.
-------
-97-
Whether such assumptions introduce a high or low bias into the
results is difficult to assess. However, it is clear that the use
of such assumptions injects a considerable degree of uncertainty
i nto the analyses.
Some of the factors which may have led the analyses to under-
state the risk of cancer related to air toxics are as follows:
1. Urban ambient air is characterized by the presence of
dozens, perhaps hundreds of substances. Risk estimates
for most of these could not be calculated due to data
1 imit at i ons .
2. Indoor concentrations of certain pollutants (e.g., radon,
tobacco smoke, formaldehyde, and other volatile organic
compounds) are commonly several times higher than out-
door concentrations. The estimated cancer incidences
associated with indoor exposures to passive smoking
(5,000 annually) and radon (1,000 to 20,000 annually),
and 24-hour personal exposures to six organic compounds
(1,700 annually) indicate that indoor sources make an
important contribution to air toxics risks.
3. Risks due to compounds formed in the atmosphere cou.ld
not be quantified in the analyses using exposure models,
but there are indications that these risks may be
significant. For example, formaldehyde is formed in the
atmosphere by the breakdown of other organic compounds,
and some compounds (e.g., toluene) may be converted into
toxic substances through photochemical reactions.
4. Although it has been shown that certain combinations
of exposures may have synergistic effects (for instance,
smoking and asbestos exposure), all risks were assumed
to be additive.
Factors which may have caused the analysis to overestimate
cancer risks associated with air toxics are as follows:
1. Cancer unit risk values were obtained from EPA's
Carcinogen Assessment Group (CAG) and Clement
Associates. EPA unit risk values are generally
regarded as plausible, upper-bound estimates. That
is, the unit risks are not likely to be higher, but
could be considerably lower. In many cases, the unit
risk values are preliminary.
-------
-98-
2. The weight of evidence of card nogeni city for the
compounds examined varies greatly, from very limited
to very substantial. Further, the extent of evaluation
and health review performed varies considerably among
compounds. For this report, a conservative scenario
(i.e., that all compounds included in the report are
human carcinogens) was assumed.
3. The risk assessments assume that people living in an
area are exposed to the estimated ambient levels for
7U years, 24 hours a day. This especially compromises
estimates of maximum lifetime individual risk. Few plants
operate for 70 years, most people change their homes
several times during their lives, and leave their neigh-
borhoods during the day.
4. The degree to which outdoor emissions of many pollutants
(e.g., trace metals) penetrate indoors is largely
unknown. If emissions of a pollutant from outdoor
sources do not penetrate completely indoors and if there
are no indoor sources of that pollutant, then we will
have over-stated risks, since we have assumed constant
exposure to levels equalling those of outdoor air.
5. Although certain combinations of exposures may have
antagonistic effects, all risks were assumed to be
additi ve.
-------
-99-
VIII. CURRENT ACTIVITIES
The study was completed during the summer of 1984 and a draft
report was circulated for peer review to approximately 20 experts
in the field in October 1984. Their technical suggestions are
reflected in this final version of the report. Many of the reviewers
recommended a final section that would provide information on EPA's
response to the principal findings of this analysis. This section
responds to that suggestion.
Based on the conclusions in the draft report, EPA initiated
a series of intense activities to reexamine its program for air
toxics and to evaluate alternative national strategies. Some of
these activities include:
0 Formation of an Agency-wide Air Toxics Group to guide
the review of the study and to assist in the development
and implementation of changes to the national program
0 Initiation of several additional analyses of the air
toxics problem to study 1) the controllability of the
most important toxic pollutants, including cost of control
to various levels and impact on cancer incidence in
several representative cities; 2) the effectiveness of
current programs for criteria pollutants and air toxics
in reducing risk in several representative cities over
the next decade; 3) the existence, intensity, and
controllability of high risk areas in several cities
caused by concentrations of sources; and 4) an analysis
of the feasibility of improving data on emissions and air
quality for major urban areas. These studies are scheduled
for completion in May 1985.
0 Discussion of the results of the study and of possible
strategic implications in over 20 presentations to groups
representing industry, environmentalists, State and local
governments, Congressional staff, and professional organi-
zations.
-------
-100-
0 Review of the report by groups within EPA who are respon-
• sible for implementing programs related to air toxics,
including monitoring, emission inventories, methods
development, and regulatory analyses; program changes
are being made as appropriate;
0 Giving increased priority is being given to a pilot
program to evaluate and possibly regulate large point
sources through a cooperative effort of Federal, State,
and local agencies. Acryl onitri 1 e is the pollutant
selected and it is typical of many pollutants with aggregate
incidence too small to justify national regulation but
with high individual risks in the vicinity of some plants.
Although most of the ongoing analyses and discussions will
not be completed until mid-1985, it already appears that signifi-
cant changes to the current national strategy will be recommended
to respond to some of the findings in this report. These include
shifting the focus of the direct Federal regulatory program from
point sources to more complex situations, including area sources
that may be responsible for high aggregate incidence. In addition,
two new programs are being evaluated: a formal partnership with
State and local agencies on regulation of large point sources to
provide better coverage for areas of high individual risk; and an
initiative directed at larger geographic areas of high risk
resulting from the interaction of many sources either in an urban
area or in an isolated industrial region.
-------
ATTACHMENT A
SUMMARY TABLE
POLLUTANTS EXAMINED, UPPER-BOUND RISK VALUES, PRELIMINARY
APPROXIMATIONS OF INCIDENCE AND MAXIMUM LIFETIME RISK
-------
POLLUTANTS EXAMINED, UPPER-BOUND RISK VALUES, PRELIMINARY APPROXIMATIONS OF INCIDENCE AND MAXIMUM LIFETIME RISK
Pollutants Having Some
Evidence of
Carcinogenicity*
Ac ryl amide
Acrylonitrile
Allyl Chloride
Arsenic
Asbestos
Benzene
Benzo-a-Pyrene
Benzyl Chloride
Beryl lium
1,3 Butadiene
Cadmi urn
I/
Unit Risk
Value
1.7x10-5
6.8x10-5
5.5x10-8
4.3x10-3
I/
6.9x10-6
3.3x10-3
1.2x10-5
4.0x10-4
4.6x10-7
2.3x10-3
Source
CLEM
CA6
CAG
CAG
CLEM
CAG
CAG
CLEM
CAG
CLEM
CAG
Preliminary Approximation
of Annual Incidence**
NESHAP
0.01
0.42
<0.01
4.7
32.3
<0.01
1.2
0.01
8.5
35 21
County
4.2
1.1
18.5
1.1
0.01
0.01
1.1
Air
Quality
60
234
5.4
0.1
14.6
Other
Preliminary Approxima-
tion of Incidence Per
106 Population**
35
County
0.02
0.5
0.39
0.02
<0.001
<0.001
0.02
Air
Quality
0.26
0.5
1.02
0.02
<0.001
0.06
NESHAP
<0.01
0.002
<0.01
0.02
0.5
0.14
<0.01
0.01
<0.01
0.04
Preliminary Approx-
imation of Maximum
Lifetime Individual
Risk** (xlO4)
NESHAP
0.74
38
0.01
65
80
0.3
1.0
0.1
36
Air
Quality
40
1.5
0.25
0.002
14.7
* The weight of evidence of carcinogenicity for the compounds listed varies greatly, from very limited to very substan-
tial. Further, the extent of evaluation and health review performed varies considerably among compounds. However,
for the purposes of this report, a conservative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
** Because of the uncertainties in the data used to make these estimates, they should be regarded as rough approxima-
tions of total incidence and maximum lifetime individual risk. Estimates for individual compounds are very uncertain.
These incidence and maximum risk estimates have been performed to provide a rough idea of the possible total magnitude
of the air toxics problem, and will be used only for priority-setting and to provide policy guidance.
-------
-2-
POLLUTANTS EXAMINED, UPPER-BOUND RISK VALUES, PRELIMINARY APPROXIMATIONS OF INCIDENCE AND MAXIMUM LIFETIME RISK
Pollutants Having Some
Evidence of
Carcinogenicity*
Carbon Tetrachloride
Chloroform
Chromium''"
Coke Oven Emissions
Diethanolami ne
Dimethyl nitrosami ne
Dioctyl Phthalate
Epichlorohydri n
Ethyl Aery late
I/
Unit Risk
Value
1.5x10-5
1.0x10-5
1.2x10-2
6.2x10-4
1.1x10-7
5.4x10-3
1.3x10-7
2.2x10-7
5.0x10-7
Source
CAG
CAG
CAG
CAG
CLEM
CAG
CLEM
CAG
CLEM
Preliminary Approximation
of Annual Incidence**
NESHAP
14
0.27
330.0
8.6
<0.01
0.05
<0.01
<0.01
<0.01
35 2J
County
0.2
0.1
13.4
2.4
Air
Quality
43
17
242
Other
Preliminary Approxima-
tion of Incidence Per
106 Population**
35
County
0.004
0.003
0.29
0.05
Air
Quality
0.19
0.07
1.05
NESHAP
0.06
<0.01
1.43
0.04
<0.01
<0.01
<0.01
<0.01
<0.01
Preliminary Approx-
imation of Maximum
Lifetime Individual
Risk** (xlO4)
NESHAP
5.8
30
1600
200
<0.01
0.54
0.1
0.02
0.47
Air
Quality
1.54
0.77
14.4
* The weight of evidence of carcinogenicity for the compounds listed varies greatly, from very limited to very substan-
tial. Further, the extent of evaluation and health review performed varies considerably among compounds. However,
for the purposes of this report, a conservative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
** Because of the uncertainties in the data used to make these estimates, they should be regarded as rough approxima-
tions of total incidence and maximum lifetime individual risk. Estimates for individual compounds are very uncertain.
These incidence and maximum risk estimates have been performed to provide a rough idea of the possible total magnitude
of the air toxics problem, and will be used only for priority-setting and to provide policy guidance.
t Risk estimates assume that all species of chromium are carcinogenic, although only certain species have evidence of
carcinogenicity. Current data do not allow differentiation among species.
-------
-3-
POLLUTANTS EXAMINED, UPPER-BOUND RISK VALUES, PRELIMINARY APPROXIMATIONS OF INCIDENCE AND MAXIMUM LIFETIME RISK
Pollutants Having Some
Evidence of
Carcinogenicity*
Ethylene
Ethyl ene Di bromide
Ethylene Di chloride
Ethylene Oxide
Formaldehyde
Gasoli ne Vapors
Gasoli ne Marketing
4,4 150 Propylidene
Di phenol
Mel ami ne
Methyl Chloride
I/
Unit Risk
Value
2.7x10-6
5.1x10-4
2.6x10-5
3.6x10-4
6.1x10-6
7.5x10-7
7.5x10-7
1.4x10-6
4.1x10-7
1.4x10-7
Source
CLEM
CAG
CAG
CAG
CAG
CAG
CAG
CLEM
CLEM
CLEM
Preliminary Approximation
of Annual Incidence**
NESHAP
<0.01
26.7
0.9
47.8
1.6
0.03
<0.01
<0.01
35 21
County
1.0
1.5
10.0
6.8
Air
Quality
11.0
191.3
0.9
Other
—
43
Prel imi nary Approxima-
tion of Incidence Per
106 Population**
35
County
0.02
0.03
0.21
0.15
Air
Quality
0.05
0.83
<0.01
NESHAP
<0.01
0.12
<0.01
0.21
0.01
<0.01
<0.01
<0.01
Prel imi nary Approx-
imation of Maximum
Lifetime Individual
Risk** (xlO^)
NESHAP
4.9
1.6
36
68
6.1
<0.01
<0.01
0.12
Air
Quality
0.73
0.49
<0.01
* The weight of evidence of carcinogenicity for the compounds listed varies greatly, from very limited to very substan-
tial. Further, the extent of evaluation and health review performed varies considerably among compounds. However,
for the purposes of this report, a conservative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
** Because of the uncertainties in the data used to make these estimates, they should be regarded as rough approxima-
tions of total incidence and maximum lifetime individual risk. Estimates for individual compounds are very uncertain.
These incidence and maximum risk estimates have been performed to provide a rough idea of the possible total magnitude
of the air toxics problem, and will be used only for priority-setting and to provide policy guidance.
-------
-4-
POLLUTANTS EXAMINED, UPPER-BOUND RISK VALUES, PRELIMINARY APPROXIMATIONS OF INCIDENCE AND MAXIMUM LIFETIME RISK
Pollutants Having Some
Evidence of
Carcinoyenicity*
Methyl ene Chloride
4,4 Methyl ene Di aniline
Nickel (Subsulfide)
Nitrobenzene
Nitrosomorpholine
Pentachlorphenol
Perchl oroethyl ene
Products Incomplete
Combustion'
PCBs
I/
Unit Risk
Value
1.8x10-7
2.1x10-5
3.3x10-*
1.2x10-7
2.5x10-5
3.9xlO-7
1.7x10-6
0.42x10°
(1.2x10-3
Source
CAG
CLEM
CAG
CLEM
CLEM
CLEM
CAG
I/
CLEM
Preliminary Approximation
of Annual Incidence**
NESHAP
1.0
0.02
0.02
<0.01
<0.01
0.12
2.9
0.21
35 2/
Cou nty
<0.01
6.7
124
Air
Quality
7.4
22
610
Other
Preliminary Approxima-
tion of Incidence Per
106 Population**
35
County
0.14
2.6
Air
Quality
0.03
0.10
2.65
NESHAP
0.004
<0.01
<0.01
<0.01
<0.01
0.001
0.01
2.65
0.001
Prel imi nary Approx-
imation of Maximum
Lifetime Individual
Risk** (xlO*)
NESHAP
0.1
15.0
0.8
<0.01
<0.01
0.17
4.6
3.0
Air
Quality
<0.01
0.19
31.5
* The weight of evidence of carcinogenicity for the compounds listed varies greatly, from very limited to very substan-
tial. Further, the extent of evaluation and health review performed varies considerably among compounds. However,
for the purposes of this report, a conservative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
** Because of the uncertainties in the data used to make these estimates, they should be regarded as rough approxima-
tions of total incidence and maximum lifetime individual risk. Estimates for individual compounds are very uncertain.
These incidence and maximum risk estimates have been performed to provide a rough idea of the possible total magnitude
of the air toxics problem, and will be used only for priority-setting and to provide policy guidance.
-------
-5-
POLLUTANTS EXAMINED, UPPER-BOUND RISK VALUES, PRELIMINARY APPROXIMATIONS OF INCIDENCE AND MAXIMUM LIFETIME RISK
Pollutants Having Some
Evidence of
Carcinogenicity*
Propylene Dichloride
Propylene Oxide
R ad ionuc Tides
Styrene
Terephthalic Acid
Titanium Dioxide
Trichloroethylene
Vinyl Chloride
Vinyl idene Chloride
I/
Unit Risk
Value
7.2xlO-7
1.2xlO-4
varies
2.9x10-7
1.8x10-8
5.6xlO-7
4.1x10-6
2.6x10-6
4.2x10-5
Source
CLEM
CLEM
I/
CLEM
CLEM
CLEM
CAG
CAG
CAG
Preliminary Approximation
of Annual Incidence**
NESHAP
<0.01
0.97
<0.01
<0.01
0.01
9.7
11.7
0.04
35 21
County
0.02
6.8
8.2
Air
Quality
18
62
Other
16
Preliminary Approxima-
tion of Incidence Per
106 Population**
35
County
<0.01
0.15
0.2
Air
Quality
0.08
0.27
NESHAP
<0.01
0.004
<0.01
<0.01
<0.01
0.04
0.05
<0.01
Prel imi nary Approx-
imation of Maximum
Lifetime Individual
Risk** (xlO4)
NESHAP
0.02
300
0.33
<0.01
<0.01
1.0
38
42
Air
Qual Ity
0.07
0.26
0.8
* The weight of evidence of carcinogenicity for the compounds listed varies greatly, from very limited to very substan-
tial. Further, the extent of evaluation and health review performed varies considerably among compounds. However,
for the purposes of this report, a conservative scenario (i.e., that all compounds examined could be human
carcinogens) has been assumed.
** Because of the uncertainties in the data used to make these estimates, they should be regarded as rough approxima-
tions of total incidence and maximum lifetime individual risk. Estimates for individual compounds are very uncertain.
These incidence and maximum risk estimates have been performed to provide a rough idea of the possible total magnitude
of the air toxics problem, and will be used only for priority-setting and to provide policy guidance.
-------
FOOTNOTES - ATTACHMENT A, SUMMARY TABLE
"Pollutants Examined, Upper-Bound Risk Values, Preliminary
Approximations of Incidence and Maximum Lifetime Risk"
_!/ The unit risk value is the estimated probability of contracting cancer as the
result of a constant exposure over 70 years to an ambient concentration of
one microgram per cubic meter (ug/m3). "CAG" denotes risk values obtained
from EPA's Carcinogen Assessment Group; "CLEM" denotes risk values obtained
from Clement Associates.
27 The population of the counties covered in the 35 County Study (about 47.3 million)
represents approximately 20% of the national population.
_3_/ The unit risk value used for asbestos was that a lifetime risk of 10~6 for lung cancer
would result from an exposure to 10 fibers/cc and that a lifetime risk of 10~6 for
mesothelioma would result from an exposure to 5 fibers/cc; 30 fibers per nanogram
were assumed.
4/ "Products of Incomplete Combustion" (PIC) refers to a large number of compounds,
probably consisting primarily of polynuclear organics. The PIC unit risk value was
derived from dose-response data which use Benzo( a) Pyrene (BaP) levels as a surrogate
for PIC or total air pollution. There are many limitations of using the B(a)P
surrogate method to estimate PIC risks: all PIC estimates presented in this report
must be regarded as highly uncertain. Refer to pp. 20-25 for a more detailed explana-
tion of how the PIC unit risk value was derived.
_5/ Estimates of cancer and genetic risks are based on those found in the 1980 National
Academy of Science Report, "Effects on Population of Exposures to Low Levels of
Ionizing Radiation" (BEIR - 3 reports).
------- |