230-2-87-025C
United States Office of Policy Analysis February, 1987
Environmental Protection Office of Policy, Planning
Agency and Evaluation
Unfinished Business:
A Comparative Assessment
of Environmental Problems
Appendix II
Non-Cancer Risk
Work Group
-------
j
J
J
COMPARATIVE RISK PROJECT
Heport of the Non^ancer VfcrR Group
February» l987
U.S. Environmental Protection Agency
Region 5, Library (PL-1ZJ)
77 West Jackson Bwtevard, 12th Floor
Chicago, II 60604-3590
-------
Table of Contents
Chapters
1. INTRODUCTION 1-1
2. RELATIVE RANKING OF ENVIRONMENTAL
PROBLEM AREAS 2-1
Observations on the Ranking 2-3
3. METHODOLOGY FOR RANKING NON-CANCER
HEALTH RISKS 3-1
4. OBSERVATIONS AND RECOMMENDATIONS 4-1
Substantive Conclusions 4-1
Procedural Conclusions 4-3
Tables
2-1: Relative Ranking of Environmental
Problem Areas 2-2
2-2: Rationale for Ranking of
Environmental Problem Areas 2-4
Appendix
METHOD FOR ASSESSING NON-CANCER HEALTH RISKS A-l
-------
Chapter 1: Introduction
Ihe non-cancer work group was asked to devise a ranking methodology and
to rank the 31 environmental problems according to the relative magnitude of
the non-cancer health risks they pose. This task was exceedingly difficult.
There are thousands of different chemicals* in the environment that may
cause adverse human health effects. Many of the 31 environmental problems
involve large numbers of these chemicals: at least 129 priority pollutants are
of concern in surface water, there are some 600 registered pesticide active
ingredients, about 75 chemicals are under study as hazardous air pollutants,
and Appendix VIII lists 400 hazardous constituents of concern at hazardous
waste management units. Little is known about the toxicological properties of
most of these chemicals; only a few have good information on their health
effects and potency and on the extent of human exposure to them.
Even if such data did exist for all chemicals, reaching a judgment on
aggregate risks from a group of chemicals constituting an environmental problem
would be difficult. Different chemicals produce different adverse effects,
ranging from effects of lesser concern (e.g., dental mottling from fluoride in
drinking water) to severe ones (e.g., death from pesticide poisoning). Entirely
different health effects may arise from a single chemical when exposure occurs
at different levels or by different routes. Dose-response functions are
generally non-linear, with most, but not all, non-cancer health effects thought
to involve thresholds. Effects for a given dose may differ depending on whether
the exposure was acute, subchronic, or chronic. Different individuals may react
differently to the same chemical; some substances at typical ambient concentra-
tions are of concern only to sensitive subpopulations such as asthmatics or
infants. And, a health effect may even be more specific; for example, it may
show up only when the asthmatic is exercising.
In short, to the extent we do have knowledge about non-cancer health effects
from toxic chemicals, it is highly particularized and difficult to aggregate.
There is no accepted common denominator by which to compare different health
effects. (In contrast, when analyzing cancer effects, the generally accepted
method is to treat different sorts of cancers as equivalent and estimate aggre-
gat^e cancer incidence. In analyzing welfare effects, diverse problems can be
expressed in common terms as dollar losses.) We can make no simplifying linear
assumptions to allow us to aggregate non-cancer effects over time and across
populations.
EPA therefore has had great difficulty in analyzing non-cancer health
effects. In September 1986, EPA promulgated a series of risk assessment guide-
lines in the Federal Register. Conspicuously missing from the set (both proposed
and final) were the guidelines on how to assess risks from "systemic toxicants"
(i.e., non-cancer effects excluding reproductive, developmental and mutagenic
effects). One reason was that members of the work group on these guidelines
* In this report, we use the term "chemicals" or "substances" very loosely,
to refer generally to a wide variety of agents including chemicals, radiation,
pathogens, and other things that can cause adverse non-cancer health effects
through environmental exposure.
-------
1-2
from different offices within EPA approached assessment of these effects
differently. The work group has recently reconvened to complete work on
the guidelines.
Most program offices do not actually assess risks from non-carcinogens.
Instead, they identify a "safe" level for which no adverse effects are expected
to occur in humans. This is often determined by reducing the no observed
adverse effects level (NOAEL) seen in animals by one or more uncertainty factors.
Most programs aim to control levels down to this "safe" level, sometimes known
as an acceptable daily intake (ADI) or reference dose (RfD)*. Most programs
merely evaluate the extent to which a regulatory option prevents exposures
above the RfD without an explicit calculation of risk. This type of analysis
may not be well suited to comparisons of aggregate risks across environmental
problem areas because RfDs for different chemicals may protect against different
health effects that vary widely in their severity. In addition, the shape of
the dose-response function at levels above the RfD probably varies substantially
across chemicals.
A few EPA programs do use methods for assessing and/or aggregating non-cancer
risks. The Office of Air and Radiation's program for National Ambient Air Quality
Standards (NAAQS) uses a probabilistic approach to calculate the uncertainty
and expected incidence of various health effects associated with alternative
exposure levels. The Office of Policy, Planning and Evaluation's Integrated
Environmental Management Program and the Office of Solid Waste and Emergency
Response's WET and Liner-Location models use slightly varying approaches to"
establishing no-effects thresholds and dose-response functions for exposures
above the thresholds. But each of these approaches is controversial, and each
applies at most to only about 100 of the thousands of potentially toxic chemicals.
As a result, the non-cancer effects work group had to break new ground to
compare the non-cancer risks associated with major environmental problem areas.
There was no established procedure for doing this. And even if there were an
acceptable method for aggregating and comparing non-cancer health effects, the
data with which to do so was certainly not available for the vast majority of
toxic chemicals.
NWith established methods and data lacking, the work group has relied
extensively on its judgment. We have ordered this judgment by whatever assess-
ment methods we have been able to develop and by whatever information we have
been able to marshall. The conclusions reached by the work group therefore
represent judgment and not verifiable fact. In this report we are offering our
opinion, or our best bet, as to the relative magnitude of the non-cancer health
risks associated with the environmental problem areas examined, and we will lay
out the reasoning that has led us to this opinion.
Recognizing that we would be relying heavily on our judgments, we have
been careful to create conditions that would make these judgments as informed,
* The official EPA term for these levels is the reference dose. The term
RfD, however, implies Agency approval, which in fact many of the ADIs do not yet
have. In this paper, we will refer to these "safe" levels as RfDs whether or
not they are approved.
-------
1-3
expert and systematic as possible. We think we have been very successful in
this effort. The work group consists of senior agency health scientists and
program managers from each major EPA program office, all of whom are expert on
the data and/or techniques available for assessing non-cancer health risks.
The work group has gathered relevant data through extensive canvassing. Its
judgments have all been arrived at collegially, and most, after some discussion,
have been unanimous.
In sum, the results of this project represent the judgment of a knowledge-
able and careful group of EPA professionals. We are the first to admit that
critical data and methods necessary to accurately respond to our charge are
lacking. CXir methods combine qualitative and quantitative factors in rough and
probably non-replicable fashion. Despite these caveats, though, we are confi-
dent that there really are substantial relative differences in non-cancer risks
across major environmental problem areas, and that our relative rankings reflect
the gist of these differences. The work group participants feel satisfied with
the process they created to rank the problem areas and with the results of the
process.
At the same time, the work group emphasizes that none of the assessments,
opinions or judgments included in this report were developed for regulatory
decision-making/purposes, and that they should not be used for such purposes.
While the report is based on Agency information used in making scientific
assessments and regulatory decisions, this information was evaluated here for
different purposes, using different procedures. In particular, the scientific
assessments have not been reviewed by any of the Agency's formally constituted
groups such as the Office of Research and Development or the Steering Committee,
nor have they been peer reviewed by external experts. Also, there has been no
public process such as notice and comment. Finally, this report which covers
a vast subject area was developed relatively quickly by a selected inter-office
work group without full involvement of senior Agency management. For all of
these reasons, the work group believes that the report should be used only for
priority setting activities and not for other purposes, such as guidance or
regulation.
The remainder of this report is organized as follows. Chapter 2 summarizes
the results of our ranking of the non-cancer health risks associated with major
environmental problem areas. Chapter 3 reviews the process we used to produce
these rankings. Chapter 4 presents some overall observations on the project
and recommendations resulting from it. The appendix describes the steps and
issues in the ranking methodology in further detail.
-------
Chapter 2: Relative Ranking of Environmental Problem Areas
Table 2-1 shows our relative ranking of the non-cancer health effects asso-
ciated with the 31 environmental problems. We do not feel confident in suggesting
a more detailed ranking than into, the three categories of high, medium and low.
No inferences should be drawn from the order in which the problems are listed
within a category. Although the rankings are qualitative, they were developed
partly from a quantitative scoring system that suggests that there is a difference
in aggregate risk from one category to the next of about two or more orders of
magnitude. Thus we believe the differences in risk between the categories are
substantial.
The table also characterizes our degree of confidence in the ranking
assigned to each problem. The degree of confidence notation reflects the
extent and quality of data that were available to us in making this assignment.
In most cases, we have not let our degree of confidence affect our ranking of a
problem. If, for example, a problem appears to be high risk but on the basis
of only sketchy information, we leave it ranked as high risk. But we would not
be overly surprised if further information came to light at some point that
would cause its ranking to change. In two cases indoor radon and radiation
other than radon we moved a problem down from the ranking it otherwise would
have had because of some uncertainty about the data responsible for the initial
high placement. (And also because the major effect in question [mutagenesis]
is very closely related to carcinogenesis, which is covered by another group.)
Finally, Table 2-1 also shows the approximate percentage of the non-cancer
risks for a problem area that are associated with the specific chemicals we
studied. This is important because we ranked a problem area mostly by consider-
ing a few specific chemicals that we thought represented the problem area.
Where the chemicals we studied represent most of the problem, the entire problem
is not much different from the sum of the components we studied. For example,
for criteria air pollutants we studied six chemicals ozone, sulfur oxides,
particulate matter, acid aerosols, carbon monoxide and lead leaving out only
nitrogen dioxide. With the six studied chemicals, we are confident that we
have considered the bulk of the non-cancer risk associated with criteria air
pollutants. For pesticide residues on food, though, we studied only three chemi-
cals (EPN, aldicarb and diazinon) of the perhaps 160 that may show some problem
with dietary residues exceeding a level of concern for non-cancer health effects.
In this case, the sum of the chemicals we studied constitutes only a small
fraction of the total risks attributable to pesticide residues on food. In
another case, although we studied only six of the hundreds or thousands of
chemicals of concern in indoor air, we thought environmental tobacco smoke
was such a large component of the total risk that the chemicals we studied
represented perhaps half of the risks from indoor air pollutants.
Our final ranking of the 31 problem areas incorporates the proportion of
the problem we had covered with the chemicals we studied. In performing sensi-
tivity analyses involving different ways of scaling up from individual chemicals
to entire problems (see Chapter 3), four problem areas moved between the high
and medium risk categories depending on the approach chosen. These problem
areas included: hazardous air pollutants, pesticide residues in food, worker
exposures and consumer exposures. We decided ultimately to rank each of these
four in the high risk category; partly because the chemicals we had studied in
each were such a small fraction of the problem, and partly because some of the
different mathematical approaches suggested a high ranking in the first place.
-------
2-2
Table 2-1 Relative Ranking of Environmental Problem Areas
Problem Area
Level of
Confidence*
% of Problem
Covered*
High Non-Cancer Risks
Criteria air pollutants (#1)
Hazardous air pollutants (#2)
Indoor air pollutants - not radon (15)
Drinking water (#15)
Accidental releases - toxics (#21)
Pesticide residues on food (#25)
Application of pesticides (#26)
Consumer product exposure (#30)
Worker exposure to chemicals (#31)
High
Medium
Medium
High
High
Medium
High
Medium
High
30-100
<3
30-100
30-100
30-100
<3
3-10
3-10
<3
Medium Non-Cancer Risks
Radon - indoor air (#4) Low
Radiation - not radon (#6) Medium
UV radiation/ozone depletion (#7) Low
Indirect discharges (POTWs) (#10) Medium
Non-point sources (#11)
To estuaries, coastal waters, oceans (#13) Medium
Municipal non-hazardous waste sites (#18) Medium
Industrial non-hazardous waste sites (#19) Low
Other pesticide risks (#27) Medium
30-100
30-100
30-100
3-10
30-100
10-30
30-100
10-30
Low Non-Cancer Risks
Direct discharges (industrial) (#9)
Contaminated sludge (#12)
To wetlands (#14)
Active hazardous waste sites (#16)
Inactive hazardous waste sites (#17)
Mining waste (#20)
Releases from storage tanks (#23)
Medium
Medium
Medium
Medium
Low
3-10
30-100
10-30
10-30
30-100
Unranked
Other air pollutants (#3)
CO2 and global warming (#8)
Accidental releases - oil spills (#22)
Other ground-water contamination (#24)
New toxic chemicals (#28)
Biotechnology (#29)
* For some problem areas, the work group did not believe it had sufficient
information to fill out these columns.
-------
2-3
Although there are other problem areas where we studied only a small portion of
the problem (e.g./ direct and indirect discharges to surface water), we did not
believe that this alone provided a sufficient reason to move them to a higher
risk category.
As noted, our ranking of a problem area depends primarily on an evaluation
of a few chemicals we thought were representative of the problem area. For each
chemical we studied, we accumulated data on the health effects it can cause,
the potency of the chemical in causing these health effects, and the amount of
exposure to the chemical. We devised a scoring system to combine data for
multiple chemicals on severity of health effects, potencies and exposures into
a single ranking for an environmental problem area. In a few cases we used
data on the incidence of health effects from a chemical rather than data on
potency and exposure. And in some cases where data were lacking we proceeded
directly to ranking a problem area without detailed consideration of individual
chemicals. Table 2-2 summarizes our rationale for ranking each of the 31
environmental problem areas as we did.
Observations on the Ranking:
o Criteria air pollutants and indoor air pollutants other than radon
ranked high generally because of large numbers of people exposed and
because ambient levels are frequently well above levels of concern.
o Drinking water risks were high, because of the very large numbers of
people exposed to levels often above RfDs. However, the pollutants
that resulted in a high ranking for drinking water may surprise
the general public (although the drinking water program office has for
some time been aware of this pattern). In general high risks do not
seem to come from chemical contaminants entering drinking water
from waste disposal, but instead from disinfection by-products, lead
from pipes, and pathogens.
o Risks to applicators from pesticides were high because of serious health
effects, frequent high levels of exposure, and high documented incidence.
These risks rank high despite the limited number of applicators
exposed.
o Risks from hazardous air pollutants, pesticide residues on food, and
consumer exposures were ranked high because of large numbers of
people exposed and potentially serious health effects. Although
exposure concentrations associated with these problems are not usually
high relative to RfDs, the work group thought that it was probably
observing only "the tip of the iceberg" among the thousands of poten-
tially toxic chemicals in these areas.
o Nearly all of the environmental problems that were ranked as low or
medium risk were characterized generally by indirect routes of human
exposure. Such problems included direct and indirect discharges to
surface water, non-point sources, pesticide runoff, estuaries,
wetlands, hazardous and non-hazardous waste disposal, sludge, mining
waste, and storage tanks. In all these areas, moderate to small
numbers of people were exposed to pollutants, at levels usually not
far above RfDs. In these areas, there is usually substantial
opportunity for pollutants released into the environment to degrade,
-------
Table 2-2 Rationale for Ranking of Environmental Problem Areas
Problem Area and Rank
1. Criteria air pollutants
[Ranked HIGH]
2. Hazardous air pollutants
[Ranked HIGH]
3. Other air pollutants
[Unranked]
4. Indoor radon
[Ranked MEDIUM]
5. Indoor Air - other
than radon
[Ranked HIGH]
Substances Studied
o Lead
o Carbon monoxide
o Sulfur dioxide
o Particulate matter
o Acid aerosols
o Ozone
o Benzene
o Carbon tetrachloride
o Chlorine
o Chromium
o Formaldehyde
o Hydrogen sulfide
Comments
o Radon
o Benzene
o Carbon tetrachloride
o Environmental
tobacco smoke
Problem area ranking dependent mostly on ozone and
acid aerosols. Ozone has large numbers of people
exposed at levels far above safe levels. Acid
aerosols has large numbers of people exposed, with
a severe health effect (increased mortality)
possible. In general, large populations exposed
to most criteria air pollutants with moderate to
severe health effects.
Exposed populations are again large, but endpoints
are less severe (generally pulmonary irritation).
Benzene has large population exposed, with a more
significant health effect possible (bone marrow
hypoplasia), but with typical ambient concentra-
tion only slightly above the RfD. Ranked high
partially because of low proportion of problem
covered by the substances studied.
Thought to be low risk. No serious health risks
suggested from noise, odor, fluorides.
Incidence modeling suggests high ranking; perhaps
200 cases per year of serious mutagenic and tera-
togenic effects. Ranking revised to medium because
of some uncertainty about this estimate and because
effects are closely related to cancer. Effects are
spread over many generations. Very large population
exposed, with severe endpoints but low probability
of effects.
Large populations exposed to these pollutants above
levels of concern. Endpoints are moderate to severe
(from jaundice to mortality and teratogenicity), and
ambient levels are often substantially above RfDs.
-------
Table 2-2 (Continued)
Problem Area and Rank
5. Continued
6. Radiation - other
than radon
[Ranked MEDIUM]
7. Ozone depletion
[Ranked MEDIUM]
Substances Studied
o Formaldehyde
o Nitrogen dioxide
o Xylene
o Occupational
o Consumer
Comnents
o Ultraviolet
radiation
8. CO2 and global warming
lUnranked]
9. Direct discharges to
surface water
[Ranked LOW]
Environmental tobacco smoke thought to contribute
the largest portion of total risk.
Incidence modeling suggests high ranking; perhaps
160 to 220 serious mutagenic and teratogenic
effects per year. Ranking revised to medium because
of some uncertainty about this estimate, and because
effects are so closely related to cancer. Very large
populations exposed to consumer radiation, with
severe endpoints but low probability of effects.
Assessment does not consider possible non^cancer
effects due to non-ionizing radiation such as
microwaves or powerlines.
Ozone depletion and increased UV radiation will
increase risks of moderately serious eye damage
(e.g., cataracts) to the entire population. One
percent ozone depletion estimated to increase
incidence of senile cataracts by 10,000-30,000 per
year. Other effects on immune systems possible
but not considered.
Considered mostly an ecological problem.
Ranked as an entire problem without data on
specific substances. Problem area defined to exclude
POTWs, which are included in problem area I 10.
Risks via consumption of fish and shellfish that have
bioaccumulated toxics or that are contaminated by
pathogens thought to be low. Risks via consumption
of drinking water contaminated by surface water
discharges thought to be minimal.
-------
Table 2-2 (Continued)
Problem Area and Rank
10. Indirect discharges
to surface water
[Ranked MEDIUM]
11. Nonpoint discharges
to surface water
[Ranked MEDIUM]
Substances Studied
Comments
12. Contaminated sludge
[Ranked LOW]
13. Discharges to estuaries,
coastal waters, oceans
[Ranked MEDIUM]
14. Discharges to wetlands
[Ranked LOW]
15. Drinking water
[Ranked HIGH]
o Lead
o Cadmium
o Lead
o Pathogens
o Legionella
o Nitrates
o Chlorine dis-
infectants
Ranked as an entire problem without data on specific
substances. Problem area defined to include POTWs
and indirect dischargers that contribute to them.
Moderate concern for bacteriological contamination
of fish and shellfish from inadeguate sewage treatment,
including combined sewer overflows.
Ranked as an entire problem without data on specific
substances. Moderate concern for bacteriological
contamination of fish and shellfish from agricultural
and urban runoff. Some concern for runoff and
bioaccumulation of pesticides in fish and shellfish.
Some concern for toxics in sediment and effects via
fish or drinking water.
Human exposure to contaminants in sludge thought
to be indirect and extremely limited.
Ranked as an entire problem without data on specific
substances. Primary exposure route is through con-
sumption of contaminated fish and shellfish. Both
pathogens and toxic chemicals are important concerns.
Ranked as an entire problem without data on specific
substances. Same concerns as for estuaries, but
much lower consumption of contaminated food or water
from wetlands.
Generally very large exposed population and serious
health effects (neurotoxicity, mortality) are pos-
sible, but exposures not often far above levels of
concern. Primary concerns were over disinfection
byproducts, lead, and pathogens.
-------
Table 2-2 (Continued)
Problem Area and Rank
16. Active hazardous
waste sites
[Ranked LOW]
Substances Studied
Comments
17. Inactive hazardous
waste sites
[Ranked LOW]
18. Municipal non-
hazardous waste
sites
[Ranked MEDIUM]
19. Industrial non-
hazardous waste
sites
[Ranked MEDIUM]
Ranked as an entire problem without reference to
specific substances. Very low number of humans
potentially exposed around active hazardous waste
sites. Exposure concentrations also thought to be
low relative to levels of concern. Substances
involved are generally of moderate toxicity.
Ranked as an entire problem without reference to
specific substances. Moderate number of people
potentially exposed around inactive hazardous waste
sites, but exposure concentrations thought to be
usually low relative to levels of concern. Substances
involved are generally of moderate toxicity.
Ranked as an entire problem without reference to
specific substances. Large number of people poten-
tially exposed, due to large number of such sites
and proximity to populations. Exposure concentra-
tions thought to be very low relative to levels of
concern because of low concentration of hazardous
constituents in such sites and indirect routes of
exposure. Substances involved are generally of
moderate toxicity.
Ranked as an entire problem without reference to
specific substances. Moderate number of people
potentially exposed. Exposure concentrations may
not always be low relative to levels of concern
because wastes are concentrated in these sources
and controls are often not extensive. Substances
involved are generally of moderate toxicity.
-------
Table 2-2 (Continued)
Problem Area and Rank
20. Mining waste
[Ranked LOW]
Substances Studied
Contents
21. Accidental releases
(toxics)
[Ranked HIGH]
22. Accidental releases
(oil spills)
[Unranked]
23. Releases from storage
tanks
[Ranked LOW]
24. Other ground-water
contamination
[Unranked]
Ranked as an entire problem without reference to
specific substances. Low number of people poten-
tially exposed due to distance of sites from popu-
lation. Low concentrations when exposure does
occur. Low toxicity substances.
Ranked as an entire problem without reference to
specific substances. Incidence data show very
substantial morbidity and mortality. Working
lifetime risks to chemical plant and transportation
workers estimated at about 8x10"^ for death and
2.7xlO~2 for morbidity. Chronic risks are believed to
be small compared with acute risks (most incidents
counted result from fires and explosions). Perhaps
1-5% of risks are borne by individuals other than
chemical workers.
Risks thought to be small and primarily of ecological
concern.
Ranked as an entire problem without reference to
specific substances. Risks thought to be small.
Relatively few health impacts reported, controls
fairly good, and people take averting behavior to
avoid chronic risks when motor fuel contaminates
drinking water.
Risks exclusive of those covered under other
source categories are generally thought to be
small. Difficult to assess magnitude of problem
involving bacteriological contamination of private
wells by septic systems.
-------
Table 2-2 (Continued)
Problem Area and Rank
25. Pesticide residues
on foods
[Ranked HIGH]
26. Application of
pesticides
[Ranked HIGH]
27. Other pesticide
risks
[Ranked MEDIUM]
28. New toxic chemicals
[Unranked]
29. Biotechnology
[Unranked]
30. Consumer product
exposure
[Ranked HIGH]
Substances Studied
o Aldicarb
o Diazinon
o EPN
Comments
o Dinoseb
o Ethyl parathion
o Paraquat
o Aldicarb
o Carbofuran
o Chlordane
o 2-ethoxyethanol
o Methylene chloride
o Formaldehyde
Large populations exposed. Potentially serious
health effects (acetylcholinesterase inhibition).
Level of exposure not often much higher than levels
of concern. Could have been ranked as medium risk
but for the fact that these three pesticides repre-
sent only a very small fraction of the problem.
Modest applicator populations exposed (10,000-
250,000). Potentially very serious health effects
(acute poisoning, fetotoxicity, teratogenicity).
Exoosures often far above levels of concern. Sub-
stantial incidence estimates: 350 annual poisonings
from ethyl parathion, 100 from paraquat.
Large populations exposed to pesticides in drinking
Water, very large number exposed to pesticides in
indoor air. Potential health effects moderate
(increased liver weight) to serious (acetylcholines-
terase inhibition). Exposures typically low relative
to levels of concern.
No satisfactory method for projecting what risks
will be. Risks may be low, as new chemical review
program weeds out many potential problems.
No satisfactory method for projecting risks.
Suspect risks to be low.
Large populations exposed to all these substances.
Ambient exposures can be at levels well above RfDs.
Serious health effects possible, including terato-
genicity and hepatotoxicity. Methylene chloride
seems to present greatest risks. Substances studied
represent very small portion of the problem.
-------
Table 2-2 (Continued)
Problem Area and Rank
31. Worker exposure to
chemicals
[Ranked HIGH]
Substances Studied
o 2-ethoxyethanol
o Methylene chloride
o Perchloroethylene
o Formaldehyde
Comments
Exposed population of workers somewhat smaller than
the consumer category, but still at least 300,000
for each substance. Workplace concentrations can
be extremely high, exceeding RfDs by over three
orders of magnitude in some cases. Substances studied
represent very small portion of the problem.
-------
2-11
dissipate, or be diluted before exposure occurs. The problems among
this group that ranked higher than the others (non-point sources,
indirect discharges to surface water, estuaries and non-hazardous
waste sites) did so primarily because of .greater proximity to potenti-
ally exposed populations and/or large volumes of pollutants.
o Worker exposures entailed very high risks, despite the limited
numbers of people exposed in occupational settings. Occupational
exposures appear to rank as high risk even on a population-risk
basis. This is because of the far higher concentrations at which
contaminants can be found in the workplace relative to those observed
in the environment.
o Three problem areas accidental releases of toxics, radiation
other than radon, and indoor radon were initially ranked as high
risk on the basis of incidence data. Observed incidence of mortality
and morbidity from accidental releases clearly place it in the high
risk category. Estimates of the effects from radiation exposures
accounted for their initially high ranking, and were considered by
the work group to be more uncertain. One reason these results were
surprising is that these sorts of risks (e.g., injuries from accidents
in transporting hazardous substances, risks from radiation in building
materials and televisions) are conceptually familiar and small from
the standpoint of personal risk. Sut this does not mean that such
risks are negligible; exposures to these risks occur with such
frequency in our society that they may add up to very substantial
problems. Ultimately, because of the uncertainty associated with
the estimates of genetic effects from radiation and because these
effects are so closely related to carcinogenic effects that will be
considered by another work group, the work group revised the initial
rankings of the two radiation problem areas from "high" to "medium".
Upon reflection, the work group continues to agree with the high rankings given
to occupational exposures and accidental releases of toxics.
-------
3-1
Chapter 3: Methodology for Ranking Non-Cancer Health Risks
In the absence of any conventional method for assessing and aggregating
non-cancer health risks, we developed our own procedure. This chapter briefly
describes the method used by the work group to rank the 31 problem.areas. A
full description of this process is included in the appendix to this report.
Shortly after starting this project, we realized that the 31 problem areas
involved numerous different substances with the ability to cause numerous
different health effects. There aopeared to be no strong pattern to the sorts
of health effects associated with a particular environmental problem; the
association was much stronger between health effects and chemicals, with a
problem representing the sum of the diverse effects caused by its component
chemicals. We made an early decision to focus on a limited number of substances
associated with each environmental problem that are representative of the
problem and are reasonably well understood. We would then try to scale up from
the representative substances to t;he entire problem.
The work group developed a format for recording existing data on represen-
tative substances. These "summary sheets" were prepared for nearly all of the
31 environmental problems. They included the following information:
o A selection of 3-6 substances to represent the environmental problem,
and a description of the rationale for selecting these substances.
o An estimate of the proportion of risk associated with the entire problem
that is accounted for by the selected substances.
o Data on health endpoints, levels of toxicological concern ( RfDs,
NOEIs, etc.), ambient concentrations, exposed populations, incidence,
and other information bearing on the magnitude and severity of the
risks from, each selected substance.
o Sources and methods for the data on the selected substances.
To assess the risks from each selected substance, we used a logic akin to that
used in calculating the number of cases expected from a chemical:
Exposure x Potency = Incidence
We could then aggregate the differing health effects caused by a single substance
into a total risk from that chemical through use of a severity index.
The data available on health effects from and exposures to toxic substances
were far from adequate to perform these calculations in a quantitatively precise
way. Exposure or potency data were frequently not available for the substances
of interest. When data were available, they were of highly variable quality.
They were often generated using different and incompatible procedures by different
programs, and they reflected very different degrees of conservatism.
Thus, the work group added its judgment to these data and developed a semi-
quantitative scoring system with which to represent key attributes for each
selected substance. Scores were developed to cover:
-------
3-2
o The severity of the health endpoints caused by the substance. A
subcommittee of health scientists developed a severity index that
assessed the extent to which any health effect threatens the
viability of the individual subjected to it. Over 100 health end-
points associated with the selected substances were scored as to
their relative severity.
o The population exposed to the substance.
o The potency of the substance at the ambient concentration or dose
to which this population is exposed. This potency was represented
by the ratio between the dose at which exposure occurs and the RfD
for the substance. The higher this ratio, the greater the probability
of the health effect occurring, or the greater the potency. This
ratio can also be thought of as an index of the individual risk at a
specific concentration of a substance. The work group debated
basing this index on the LOAEL or NOEL rather than on the RfD.
This decision may have influenced the final rankings somewhat (see
Appendix), but no sensitivity analysis was performed on it.
We tried to develop these three scores consistently for all the selected sub-
stances. We used a different method when available data covered incidence of a
health effect from a substance rather than exposure and potency (see Appendix).
When data on individual substances were lacking, the work group used its best
judgement to score a problem area as a whole without reference to its component
substances. In a few other cases when available information was minimal, the
entire problem area was ranked without developing conponent scores. Finally,
we did not rank at all some problem areas where we could not develop any way to
estimate risks.
We combined these three scores representing the severity of the end-
points, the exposed population, and the likelihood of an effect given an expo-
sure-in various alternative ways to produce tentative risk rankings of sub-
stances and of problem areas. We paid special attention in sensitivity analysis
to ascertaining whether ranking by individual risk would yield results much
different from ranking by population risk. It did not. In addition, different
approaches were used to aggregate scores from selected substances into scores
for an entire problem. With different approaches, a few problems (hazardous
air pollutants, drinking water, worker and consumer exposures) moved between
the median and high risk categories. All were ultimately ranked high.
We reviewed the various tentative rankings of problem areas and developed
our own qualitative ranking of problem areas that was consistent with most of
the tentative rankings. We assigned problem areas to categories of high,
medium or low non-cancer risks. The available quantitative data underlying the
scores suggest that there is about a 2+ order of magnitude difference in risk
between each successive risk category. As a final step in the ranking, we
adjusted the rankings slightly to reflect the quality of data on each problem
and the proportion of each problem we had covered with the substances we studied.
The appendix includes a full discussion of the major methodological or
data problems we encountered, how we resolved them, and hew satisfied we are
with their resolution.
-------
Chapter 4: Observations and Recommendations
In this concluding chapter we present summary observations and recommenda-
tions from the project. Chapter 2 included our final rankings of non-cancer
health risks associated with the 31 environmental problem areas. In this
chapter, we present conclusions of a broader nature; ones that are not specific
to a particular problem area. We have divided our observations and recommenda-
tions into substantive and procedural categories.
Substantive Conclusions;
1. There is a wide disparity between the amount of risk we estimate for many
problem areas and the amount of attention EPA gives these areas. High-risk
areas to which EPA devotes relatively little attention include:
o Indoor air
o Accidental releases of toxics
o Consumer product exposure
o Worker exposure to chemicals
Low-risk areas to which EPA devotes relatively large amounts of attention
include:
o Direct point source discharges (industrial)
o Active hazardous waste sites (RCRA Subtitle C)
o Inactive hazardous waste sites (Superfund)
o Releases from storage tanks (UST)
There are undoubtedly some good reasons for this disparity between risk and
program attention:
o Ws have assessed only one.class of risks. A program that ranks high
or low in non-cancer risks may rank differently when other sorts of
risks are considered, such as ecological, welfare, or cancer risks.
o We have assessed the residual risks remaining now in each environ-
mental problem area. That is, we have taken credit for all of EPA's
previous activities and incorporated their effects into the baseline.
A large program effort may still be necessary (e.g., in enforcement)
to hold future risks to these residual levels.
o We have not considered the controllability of the risks in our ranking.
It may make sense to focus EPA resources where we can have the largest
impact in risk reduction. This may not correlate precisely with rela-
tive risk.
o Other factors besides risk are also important determinants of the
appropriate level of effort to devote to a problem area. These include
statutory mandates, public demands, and the responsibilities of other
agencies and governments.
One principle in allocating resources is that EPA should put resources
where they will "do the most good." More precisely, a marginal increment
of resources in dollars or manpower should be allocated to the activity
-------
4-2
where it will result in the greatest reduction in risk. We can evaluate
the potential for the greatest reduction in risk by determining: (1) the
existing risk in each problem area, (2) the availability of technical
methods to reduce that risk, and (3) the availability of regulatory tools
that will lead to the use of the technical solutions to reduce the risk.
Our ranking concentrates only on the first of these three components. An
evaluation of all three is essential in order to make a logical decision
on resources.
We believe another distinction is important regarding the relationship
between the non-cancer analysis (as well as this risk project as a whole)
and EPA resource allocation. Our work suggests areas to which additional
resources should be devoted, but does not suggest areas from which existing
resources should be withdrawn. We focused only on residual risks given
existing regulatory programs. We did not attempt to estimate the risks
that existed before application of these programs. Without estimates of
these risks, we cannot judge the value of these programs at current resource
levels, nor can we guess what the effect of reductions in these programs
would be.
In sum, we support organizing environmental protection more around the
fundamentalxgoal of reducing demonstrable risks. EPA management should
consider providing incremental program resources to areas promising higher
risk reduction. At a minimum, EPA should devote substantially more R&D
resources to the seemingly high risk areas that are not already getting
them, in order to confirm or refute judgments about these areas. Such
areas should include indoor air, accidental releases of toxics, and consumer
and worker exposures. EPA should also engage the public and the Congress
in a dialogue on this so that over time, environmental statutes will
represent risk priorities.
2. In many of the areas we identify as high risk, EPA has vague or non-existent
regulatory authority to prevent emissions or limit exposures. In fact, EPA
seems more likely to have explicit regulatory authority in areas that we
rank as medium or low risk than in areas of high risk. Often this occurs
because EPA's regulatory authorities in these other areas have helped
reduce their risks.
In important areas where EPA has only tenuous or indirect regulatory authority
indoor air, accidental releases, consumer product exposure, and occupational
exposures EPA should intensify its use of broad non-regulatory authorities.
Such authorities include research and development, information gathering
(TSCA Sections 4 and 8), technical assistance, referral to other agencies
(TSCA Section 9), and public education. All available statutes should be
used to solve problems. These authorities may be very effective in reducing
risks. In particular, EPA now conducts little public health education, and
should do more. Successful efforts in areas including radon, health advisor-
ies (on drinking water contaminants and on used oil) and misfueling suggest
the effectiveness of public education.
3. Workplace exposures to toxic chemicals were often extremely high relative
to ambient exposures. EPA should make much more extensive use of TSCA's
authorities to address this area.
4. A few chemicals ware cited as major concerns in multiple problem areas.
-------
4-3
These chemicals include lead most prominently, and also cadmium, carbofuran,
chromium, dinoseb and formaldehyde. EPA should consider developing cross-
program integrated strategies for dealing with these key toxics. The
integrated strategies might include comprehensive information gathering
on sources and total exposure, cross-program comparison of cost-effectiveness
of options to reduce exposure, and promulgation of regulatory programs
drawing upon authorities under multiple statutes.
5. Our analyses of exposures in different problem areas point out some important
relationships between programs. Many programs aim to prevent the release
of pollutants that can eventually find their way into drinking water and
cause health risks. These prevention programs include primarily the surface
water (direct, indirect, and non-point discharges), hazardous and non-
hazardous waste (RCRA C and D, CERCIA), storage tank, other ground-water
contamination and pesticide runoff programs. The drinking water program
provides summary information on the risks that are missed or not abated by
these prevention programs. These prevention programs, though, take little
advantage of the information available on contaminants ultimately in drinking
water. The information on drinking water contaminants should be critical
in establishing priorities within a prevention program (e.g., the contaminants
eventually causing highest risks should be of greatest priority) and between
prevention programs (e.g., the programs that can abate eventual drinking
water risks most effectively should be of greatest priority).
A similar observation might be made about indoor air as another program
with a growing data base on health risks. Prevention programs involving
consumer product exposures, pesticides (indoor), drinking water (VDCs), and
radon should use data on conparative risks of different indoor exposures
for priority setting.
6. Concern about health impacts from radiation has traditionally focused on
cancer. Modeling of mutagenic and teratogenic effects, primarily because
of the large populations exposed, results in substantial projected incidence.
Further research on these non-cancer effects of radiation, as well as on
effects from non-ionizing radiation, is appropriate.
Procedural Conclusions;
1. EPA should make a concerted effort to increase its knowledge about non-cancer
risks to human health as they relate to EPA's responsibilities. In general,
we possess only poor data and inadeguate methods for assessing non-cancer
risks.
2. All programs need higher-quality data on exposure to substances capable of
causing non-cancer health effects. Exposure information currently is
shamefully poor, even on the highest-priority chemicals that are objects of
major regulatory efforts. Exposure information, when it exists, is often
based upon inconsistent methods. Exposure estimates in different areas
exhibit fundamentally different degrees of conservatism. Even in areas when
EPA has a relatively greater amount of good information (e.g., criteria air
pollutants), there still remain major uncertainties with respect to dose-
response and exposure relationships for many of the pollutants examined.
EPA should dedicate resources to collect the data needed to assess non-cancer
risks. EPA should consider getting stronger statutory authority and using
existing authority more fully to have others '(e.g., manufacturers, dischargers)
-------
4-4
also generate exposure data.
3. EPA has no consistent methods for assessing non-cancer risk, making it
difficult to compare results across programs. A few general methods are
in limited use OSW's WET and Liner-Location Models and OPPE's Integrated
Environmental Management approach but they are not widely accepted. This
lack of a consistent methodology does not characterize assessment of cancer
risks. The Risk Assessment Forum should make development of general methods
of assessing non-cancer risks a high priority.
4. A particular methodological problem is the lack of a dose-based model of
non-cancer effects that is able to deal with both severity and incidence.
EPA's traditional focus on NOELs, LOELs, RfDs, and margins of safety is
insufficient. Characterizing a dose-response function at levels above the
RfD is important. Scientists should be encouraged to report differences in
severity of effect, as well as incidence, at varying doses in animal and
human studies.
5. The results of our ranking reflect great uncertainties about appropriate
methods, exposed populations, exposure levels, dose-response functions, and
comparison of health effects. The guality of data used also varies greatly.
Large amounts of qualitative judgment have gone into the rankings. While the
rankings represent the disciplined opinions of experts, they are still only
opinions. These rankings would not withstand rigorous scientific peer
review or judicial review. The Agency should be very careful about how the
non-cancer study is portrayed in the final report on this risk project, and
how the rankings are used.
6. Despite the extensive difficulties and uncertainties in assessing non-cancer
risks in the 31 problem areas, we are reasonably confident that there are
major differences in non-cancer risks between problem areas, and that our
rankings have accurately captured these differences. We believe that there
are generally about two orders of magnitude difference in risk between
problems that we have placed in different categories.
7. Very large amounts of valuable staff time have been spent on this project.
On the whole, we believe this investment was worthwhile. However, this
effort should not be repeated again within the next few years. It will
take a substantial period of time before data and methods for non-cancer
risk assessment can be meaningfully improved. And, many of our conclusions
seem robust enough to resist major changes, should we perform this project
again in the near future.
8. During the work group meetings, it became apparent that most of us knew
little about each others' program areas. One of the most effective ways
that EPA can move toward a more integrated view of environmental protection
is for staff and managers to have a comprehensive appreciation of environmental
problems and programs. Personnel transfers, training, and other means of
encouraging a cross-media and cross-program perspective should be increased.
9. The Risk Assessment Forum and the Risk Management Council should be directed
to monitor progress EPA is making in collecting necessary information,
developing methodologies, and using non-cancer risks in decision-making.
-------
APPENDIX
METHOD FOR ASSESSING NON-CANCER HEALTH RISKS
In the absence of any conventional method for assessing and aggregating
non-cancer health risks, the work group developed its own procedure. This
chapter describes the steps the work group went through.
1. Establishing the Work Group
The work group consisted of at least one representative from each of EPA's
program offices, the regions, the Office of Research and Development and the
Office of Policy, Planning and Evaluation. Most of these individuals are
health scientists with extensive experience at EPA. All have access to relevant
data, studies and other experts in their individual program areas. The chair-
person of the work group was Marcia Williams, the Director of the Office of
Solid Waste and formerly the Deputy Assistant Administrator of the Office of
Pesticides and Toxic Substances. In short, the work group was chaired and
staffed by individuals with broad and appropriate expertise for the task.
2. Initial Decisions on Project Methodology
The work group's assignment was to assess the non-cancer health risks asso-
ciated with the 31 environmental problem areas, and to produce a relative ranking
of the 31 areas by non-cancer risk. The 31 areas were defined in a common
fashion for this and the other three work groups (cancer, ecological and welfare).
In general, the work group planned to go as far as it could in assessing
non-cancer risks quantitatively. It would use the standard approach (exposure
x potency = incidence) to calculate cases of adverse health effects from expo-
sures to a chemical, and then sum across chemicals to obtain total risks for a
problem area. Cases of different health effects might be aggregated if an
appropriate severity index could be agreed upon. Gaps in guantitative knowledge
would be filled through qualitative analysis or Delphi procedures. The work
group was ultimately unable to get very far with the quantitative approach alone.
An initial effort was made to collect easily available data on non-cancer
health effects in major environmental program areas. After reviewing this
information, the work group reached three conclusions about further work:
o A common format should be used for acquiring and organizing existing
data on the different problem areas.
o Particular health effects are associated more with particular substances
than they are with broad problem areas. Information collection should
focus on the effects of a substance within a problem area (e.g., the
effects of lead in drinking water, the effects of pathogens in drinking
water, the effects of lead as a criteria air pollutant) rather than on
the problem area as a whole.
o Most environmental problems involve numerous toxic chemicals. Any
attempt to be comprehensive and assess the risks from every toxic
chemical for a problem area would necessarily fail such a task
would be too large, and data would be inadeguate for a large propor-
tion of the chemicals.
-------
A-2
The work group thus decided to focus for each environmental problem on a limited
number of chemicals that are representative of the problem and are reasonably
well understood. The work group would then try to scale up from the represen-
tative chemicals to the entire problem.
3. Summary Sheets
The work group developed "summary sheets" for recording existing data on
representative chemicals. These sheets were to be filled out by the program
office most knowledgeable about each environmental problem area.
Each summary sheet reguired the following information on an environmental
problem area:
o Select 3-6 substances representative of the environmental problem and
describe the rationale for selecting these specific substances.
o Estimate the proportion of risk associated with the entire problem that
is accounted for' by the selected substances. Describe the rationale.
o Describe the overall approach and sources of data for estimating health
effects and exposures in this problem area.
For each of the selected substances/ the summary sheet also described the:
o Most important health endpoints, including the:
- assumptions used in deciding on endpoints for example, whether
there are sensitive subpopulations, the nature and quality of the
studies providing evidence, etc;
- particular endpoint that drives regulatory concern for the sub-
stance. The particular endpoint for which the RfD is established,
or which is observed at the LC&EL; and
- level at which the substance is of toxicological concern, such
as an RfD, ADI, NOEL, LCAEL, potency/ NAAQS or other standard or
benchmark.
o Magnitude and severity of the problem, including the:
- populations exposed to different concentrations of the substance;
- ratio of the concentration at which environmental exposure occurs
to the level at which the substance is of toxicological concern,
(the higher this ratio is, the more likely there are to be adverse
health effects from actual ambient exposures);
- effects and severity expected from actual ambient exposures;
- time dimensions to these exposures: chronic, subchronic, one-
hour, 24-hour, etc; and
- available data on the incidence of the health effect.
-------
A-3
Backup pages to each summary sheet were also requested. They were to explain
the basis for all these estimates: whether concentration data were modeled or
monitored, which models were used, what type of monitoring occurred, where the
population data came from, what methods were used in exposure assessment, etc..
Summary sheets were produced for 22 of the 31 environmental problem areas.
The work group had decided not to request summary sheets for three problem
areas {other air pollutants, C02 and global wanning and oil spills) because it
thought non-cancer health risks in these areas were likely to be minimal and
virtually impossible to estimate. For another five problem areas (wetlands,
releases from storage tanks, other ground-water contamination, new toxic chemi-
cals and biotechnology) the work group requested summary sheets. However, the
program offices did not prepare them because they were unable to develop a
method for projecting substances of concern, health effects and exposures for
new toxic chemicals and biotechnology and they thought that the non-cancer risks
from the other three problem areas were low and simply were not worth the
effort of producing summary sheets.
The final data base in the summary sheets included information on 50
chemicals. Many chemicals were selected because they caused concern in more
than one problem area. For example, lead appeared in six problem areas, and
cadmium, chromium, carbofuran, dinoseb and formaldehyde in four areas. Table
A-l lists the chemicals selected. In all cases the health effects and exposure
information on a substance were specific to the problem area for which it was
chosen. Thus when lead was chosen to represent criteria air pollutants, the
summary sheet included information on effects via inhalation, concentrations in
ambient air, etc. When lead was selected to represent drinking water problems,
the information pertained to ingestion risks, concentrations in drinking water,
etc.
The summary sheets differed very widely in quality. This was largely a
function of the amount of pre-existing information available on concentrations,
exposures, health effects and risks in each program area. The extent to which
information already existed depended in turn, on the historical amount of
attention each program office had given to learning about exposure to and the
health impacts of substances under its purview. To generalize, the Office of
Airland Radiation has the most extensive information on the air pollutants that
have been assessed for regulatory purposes. The Office of Water, with the
exception of contaminants in drinking water, has very little information avail-
able. The Office of Solid Waste and Emergency Response also has little informa-
tion available, but has recently made a substantial effort to model exposures
and health effects via pathways within their purview. The Office of Toxic
Substances has good data available on particular existing chemicals with the
exception of limited available data on consumer exposures. The Office of
Pesticide Programs has extensive data on health effects (including incidence)
from pesticides, but is weaker on exposure data.
The summary sheets highlighted some data problems that the work group
always knew it would have to deal with. The nature of the data problems varied
widely:
o At one extreme were some large, insoluble problems. There is only a
snail proportion of the thousands of substances in our environment
about which we have any toxicological understanding. In the future,
numerous substances that we knew nothing about today will be found
-------
(a)
TABLE A-1 OCCURRENCE OF SUBSTANCES REPORTED IN SUMMARY SHEETS
Substance
Aldicarb
Benzene
Cadmium
Carbofuran
Carbon tetrachloride
Chlordane, Aldrin, Heptachlor
Chlorine
Chlorobenzene
Chromium
Oiazinon
Dinoseb
Environmental tobacco smoke
EPN
2-ethoxyethanol
Ethyl parathion
Fluoride
Formaldehyde
Hydrogen sulfide
Lead
Legionella
Mercury
Nitrate/Nitrite
Nitrobenzene
Nitrogen dioxide
Paraquat
Pathogens (Giardia/Vi ruses)
Phenol
Radiation
2, 4, 6- Trinitrotoluene
UV Radiation (Ozone depletion)
Xylene
Methylene chloride
Chlorine disinfectants
Perchloroethylene
Carbon monoxide
Sulfur dioxide
Paniculate matter
Acid aerosols
Ozone
Copper
Zinc
Radon/Radon daughters
Endothall
Oxamyl
Glycol ethers
Asbestos
Anhydrous ammonia
Hydrochloric acid
Sulfuric acid
Frequency
3
2
4
4
2
1
2
1
4
1
4
1
1
2
1
1
4
1
6
1
3
2
1
1
1
1
3
1
1
1
1
2
1
1
1
1
1
1
1
1
1
1
2
2
2
2
1
1
1
Problem Areas
11,25,27
2,5
12,16,17,18
9,10,11,27
2,5
27
2,21
16
2,16,17,19
25
9,10,11,26
5
25
30,31
26
20
2,5,30,31
2
1,11,12,15,17,18
15
9,10,17
11,15
16
5
26
15
16,17,19
6
16
7
5
30,31
15
31
1
1
1
1
1
11
11
4
9,10
9,10
30,31
30,31
21
21
21
(a) This list consists of chemicals that were mentioned in the summary sheets.
Because of significant data gaps in some problem areas, the workgroup
scored these entire problem areas rather than individual, representative
substances. Therefore, some chemicals listed here do not appear in Tables
2-2 and A-6 which summarize the final rankings.
-------
A-5
to have toxic effects. The work group's attitude about such issues
was not to try and assess risks that we know nothing about.
o For many substances on which there was sufficient toxicological
knowledge/ exposure information was inconplete. Gaps would have to
be guessed at.
o Where some data existed on a substance in all relevant areas health
effectSr levels of concern/ ambient concentrations and exposed popula-
tions the data frequently did not mesh well. For example, an
estimated ambient concentration and an estimated exposed population
might not have been generated under identical assumptions. The
ambient concentration might derive from monitoring in urban areas/
while the exposed population consists of some individuals subjected
to lower concentrations in rural areas as well as those subjected
to the higher urban ones.
o Even if a reasonably consistent data set could be obtained for a
single substance/ making the comparisons across substances that are
necessary in assessing relative risks can be very difficult. Data on
different substances have often been generated under different ground
rules. For some chemicals ADIs or RfDs were available, for others
there^were only NOELs or LC&ELs. Some exposure estimates were very
conservative upper-bound estimates, while some were maximum likelihood
estimates.
o A final data problem related to the variable quality of the estimates.
Some estimates were good, precise and generated by reliable methods,
while others were little more than guesses. How could they be compared?
We are not referring here to comparing estimates of different things
where we know there will be some bias in making the comparison (e.g.,
comparing a conservative worst-case estimate of exposure in one area
with an annual average sort of estimate in another area). The work
group instead was worried about putting side-by-side and comparing
two estimates that were generated under identical ground rules but
were nevertheless of substantially different precision. Should we
compare the two estimates and simply note the level of uncertainty
associated with each? Or should we downgrade the less certain estimate
to reflect our lack of confidence in it?
The most pointed example of this issue ultimately arose in the
final ranking of the 31 problems, when criteria air pollutants and
indoor radon both seemed to present high non-cancer risks. The data
underlying the ranking for criteria air pollutants were strong and
reliable. The data behind the radon ranking were sparse, the model
used to estimate radon effects was unfamiliar, and the results were
surprising. But our best, albeit imprecise, estimate of radon risks
were that they were high. The work group had a lenqthy argument in
this case about whether criteria air pollutants and radon had to be
ranked similarly, or could we account for the much lower certainty
of the radon estimate by moving radon down in the ranking.
Some of these data problems evidenced in the summary sheets could be miti-
gated through more effort trying harder to find appropriate data, reworking
existing estimates to make them more comparable, and developing some quantitative
-------
A-6
way for expressing the quality of data, for example. The work group took some
of these steps. Some summary sheets were written two or three times, and some
estimates were revised to reduce inconsistencies. But the work group decided
generally not to go to extreme lengths to improve the data that were available.
In general, time and available resources did not permit extensive reworking
of the data base. Instead, the work group adapted the nature of the process it
planned to use for ranking the 31 problem areas to the quality of the available
data. It was clear that the ideal of quantitatively estimating the number of
cases of adverse health effects for each problem area was not feasible. Judg-
ment and qualitative evidence had to be substituted frequently where data were
unavailable. As long as existing data were sufficiently accurate to give a
rough impression of the magnitude of the quantity being assessed, they were
used. If a piece of existing data was probably wrong in some way (e.g., "I'll
bet that really is an overestimate"), the work group decided not to spend the
time to rework the piece of data, but instead only to appreciate the nature of
the likely error. The nature of the ranking process the work group initially
wanted to pursue dictated the information that was requested via the summary
sheets. But once the summary sheets had been developed, it is fair to say that
the nature of the information they contained dictated the further development
of the ranking process.
Althouqh it appeared impossible to explicitly estimate the number of cases
of health effects as the workgroup had initially hoped, the steps involved in
trying to estimate the number of cases continued to provide the organizing
principles in the group's ranking process. Cases are a function of exposure to
a substance and the potency of the substance. Cases of different health endpoints
could be aggregated if they were weighted by some acceptable index of severity.
The work group proceeded to evaluate the risks from each representative substance
by assessing these three factors: health endpoints, exposure and potency.
4. Health Endpoints
The workgroup had decided to classify all health effects into one of 11
categories: cardiovascular, developmental, hematopoietic, immunological,
kidney, liver, mutagenic, neurotoxic/behavioral, reproductive, respiratory and
other. At one point the workgroup considered not making distinctions between
different effects within a category; counting, for example, all respiratory
effects as similar. The ultimate risk ranking might then assess how many
respiratory effects a particular problem might cause, how many cardiovascular
effects, etc. After some consideration, though, the workgroup decided that it
ought to preserve the distinctions among health effects within a category. It
made little sense to lump effects of such disparate severity as nasal irritation
and emphysema simply because they both involved the respiratory system.
In reviewing the summary sheets, there were 106 different health endpoints
listed as being caused by exposure to the selected representative chemicals.
These endpoints, cross-referenced to the substances causing them, are listed in
Table A-2. However, ranking each of the 31 problem areas by comparing their
impacts on 106 different classes of effects was clearly impossible. Some
common denominator or sorting procedure had to be developed for these disparate
health effects. With aggregating effects by organ or function having been
rejected, the work group decided to develop a severity index. Each health
endpoint would be assigned a score representing its relative severity.
-------
TABLE A-2
CHARACTERIZATION OF SPECIFIC ENDPOINTS (a)
Specific endpoints
= ~S= = SS===SS== ===£= ===£=========
Cardiovascular
unspecified
increased heart attacks
aggravation of angina
increased blood pressure
mort. from ischemic heart dis.
Developmental
fetotoxicity
low birth weight
teratogenicity
Hematopoietic
unspecified
decreased heme production
bone marrow hypoplasia
impaired heme synthesis
methemogIobinemia
leukopenia
anemia
thrombocytopenia
Problem Area (Substance)
30CMethytene chloride], 17CLead], 15[Lead], 31 [Hethylene chloride]
1[Carbon monoxide]
1[Carbon monoxide]
1 [Lead]
5[Environmental tobacco smoke]
2[Benzene], 5[Benzene]
5[Environmental tobacco smoke], 30[2-Ethoxyethanol], 31[2-Ethoxyethanol], 1[Carbon monoxide]
26[Dinoseb], 30[2-Ethoxyethanol], 17[Lead], 17[Cadmium], 6[Radiation], 31 [2-Ethoxyethanol], 5[Xylene],
2[Carbon tetrachloride], 5[Carbon tetrachloride], 4[Radon]
1&[Leadl, 16[Nitrobenzene]
17 [Lead]
2[Benzene], 5[Benzene]
2[Hydrogen sulfide], ULeadl
15 [Nitrate], 11[Nitrate]
2[Benzene], 5[Benzene]
2[Benzene], 5[Benzene]
2[Benzene], 5[Benzene]
Immunological
unspecified
herpes
allergic reactions
increased infections
Kidney effects
unspecified
tubular degeneration
dysfunction
hyperplasia
hypertrophy
atrophy
necrosis
histopathotogical alterations
Liver effects
unspec i f i ed
hepatitis A
jaundice
increased weight
increased enzymes
necrosis
histopathotogical alterations
15[Chlorine disinfectants], 1 [Paniculate matter]
7[UV Radiation/Ozone depletion]
2[Formaldehyde], 5[Formaldehyde]
5[Nitrogen dioxide], 1[Ozone]
mchromium], 19[Phenol], 30[Methylene chloride], 17[Chromium], 16[Chlorobenzene], 16[Chromium], 16[Phenol], 16[Nitrobenzene],
16[Cadmium], 31[Hethylene chloride], 30[Methylene chloride]
2[Carbon tetrachloride], 5[Carbon tetrachloride], 18[Cadmium], 12[Cadmium], 12[Lead]
17 [Cadmium], 17 [Mercury]
17[Phenol]
17[Mercury], 17[Phenol]
17[Hercury], 17[Phenol]
17 [Mercury]
31 [Perchtoroethylene]
16[Cadmium], 19[Chromium], 17[Chromium], 15[Chlorine disinfectants], 16[Chlorobenzene], 16[Chromium], 16[Phenol], 16[Nitrobenzene]
15[Pathogens (Giardja/Viruses)]
2[Carbon tetrachloride], 5[Carbon tetrachloride]
27 [ChIordane/A Idr i n/HeptachIor]
27[Chlordane]
27[Chlordane]
31[Perchloroethylene], 30[Methylene chloride], 31[Methylene chloride]
Note: Problem Area numbers are given in Tables 2-1 and 2-2.
-------
Hutagenicity
unspecified
cytogenetic
hereditary disorders
Neurotox i c/Behavioral
unspecified
retardation
reduced corneaI sensitivity
retinal disorders
visual aging
AChE inhibition
learning disabilities (red. cogn.)
neuropathy
irritability
tremors
convulsions
sensory irritation
micromecurialism
Reproductive
unspecified
post iirplantation losses
testicular degeneration
spermatocyte damage
decreased testicular weight
aspermi a
increased resorptions
giant cell formation
increased spontan.abortions
male sterility
oligospermia
decreased sperm motility
Respiratory
unspecified
emphysema
nasal irritation
pulmonary irritation
nasal ulceration
mucosal atrophy
bronchitis
pulmonary impairment
lung injury
pneumonia
pulmonary edema
pontiac fever
congestion
hemorrhage
alveolar collapse
fibresis
lung structure changes
aggravation of asthma
2[Carbon tetrachloride], 5[Carbon tetrachloride], 1[Particulate matter], 2[Chromium]
2[Benzene], 5[Benzene]
6[Radiation], 4[Radon]
2[Formaldehyde], 2SCEPN1. 30[2-Ethoxyethanol], 18[Lead], 17[Lead], 15[Chlorine disinfectants], 30[Methylene chloride],
3U2-Ethoxyethanol], 31 [Perchloroethylene], 31 [Methylene chloride], 9[Mercury], 10[Mercury]
15[Lead], 6[Radiation], 4[Radon]
2[Carbon tetrachloride], 5[Carbon tetrachloride]
26[Ethyl parathion], 7[UV Radiation/Ozone depletion]
7[UV Radiation/Ozone depletion]
27[Aldicarb], 27[Carbofuran], 26[Ethyl parathion], 25[EPN], 25CAldicarb], 25[Diazinonl, 11 [Aldicarbl, 11 [Carbof uranl
15[Lead], 1 [Lead], 1 [Carbon monoxide], 12[Lead]
17 [Mercury], 12 [Lead]
27[Chlordane]
27[Chlordane]
27[Chlordane)
30[Formaldehyde], 31[Formaldehyde]
9 [Mercury], 10 [Mercury]
17 [Mercury]
5[Xylene]
2[Carbon tetrachloride], 5[Carbon tetrachloride], 27[Carbofuran], 26[0inoseb]
30[2-Ethoxyethanol), 31[2-Ethoxyethanol]
31[2-Ethoxyethanol]
31 [2-Ethoxyethanol]
30[2-Ethoxyethanol]
27[Carbofuran]
30[2-Ethoxyethanol]
27[Carbofuran]
1[Carbon monoxide]
11[Dinosebl
26[Dinoseb]
26[0inoseb]
1[Particulate matter]
laiCadmium], 12 [Cadmium]
5[Environmental tobacco smoke]
5[Environmental tobacco smoke], 2[Chlorinel, 2[Formaldehyde], 2[Hydrogen sulfide], 5[Formaldehyde]
2[Chromium]
2 [Chromium]
2[Chromium], 5[Environmental tobacco smoke]
5[Nitrogen dioxide], 5[Environmental tobacco smoke]
5[Nitrogen dioxide], 1[Ozone]
5[Formaldehyde], 15[Legionella], 12[Cadmium], 5[EnvironmentaI tobacco smoke]
5[Formaldehyde], 26[Paraquat], 2[Formaldehyde], 2[Hydrogen sulfide], 2[Chlorine)
15[Legionella]
26[Paraquat]
26[Paraquat]
26[Paraquat]
26[Paraquat]
5[Nitrogen dioxide], 1 [Ozone], 1[Acid Aerosols), 1[Particulate matter]
1[Sulfur dioxide], 5[Environmental tobacco smoke]
Note: Problem Area numbers are given in Tables 2-1 and 2-2.
-------
increased resp. disease
bronchoconstrict ion
decreased mid-expiratory flow rates
increased respiratory infections
altered lung function
aggrav. of resp. diseases
decreas. small airway function
ciliary effects
infant hospital, for resp. disease
squamous metaplasia
Other
unspecified organ effects
unspecified acute effects
mortality
morbidity
eye irritation
dental erosion
cataracts
leishmaniasis
adrenal
gastrointestinal disease
bone damage
dental mottling
symptomatic effects(headache)
Legionnaires' disease
1[Acid Aerosols]
5[Environmental tobacco smoke]
5[Environmental tobacco smoke]
5[Environmental tobacco smoke], 1[Ozone], 5[Nitrogen dioxide]
1 [Part iculate matter], 5 [Nitrogen dioxide], 2[Chromium], 1 [Ozone]
UParticulate matter], 1 [Sulfur dioxide], 1 [Acid Aerosols]
5[Environmental tobacco smoke]
30[Formaldehyde], 31[Formaldehyde]
5[Environmental tobacco smoke]
30[Formaldehyde], 31[Formaldehyde]
16[2,4,6-Trinitrotoluene]
26[Dinoseb]
5[Environmental tobacco smoke], 15[Legionella], 15[Pathogens (Giardia/Viruses)], 21[Accidental Releases - Toxics (b)]
UParticulate matter], 1 [Ac id Aerosols], 1[Sulfur dioxide], 26 [Ethyl pa rath ion], 26[Paraquat]
21[Accidental releases - Toxics (b)
2[Formaldehyde], 2[Hydrogen sulfide], 5[Environmental tobacco smoke], 5[Formaldehyde]
2[Chlorine]
7[UV Radiation/Ozone depletion]
7[UV Radiation/Ozone depletion]
16 [Nitrobenzene]
15[Pathogens (Giardia/Viruses)], 15[Lead]
20[Fluoride]
20[Fluoride]
5[Environmental tobacco smoke]
15[Legionella]
NoterProblem Area numbers are given in Tables 2-1 and 2-2.
(a) Endpoints listed in this table represent all endpoints specified in the summary sheets; not just the driving endpoints. Health endpoints were not reported
for some substances listed in Table A-1. Thus, they are not included in this table. Because of significant data gaps in some problem areas, the
workgroup scored these entire problem areas rather than individual, representative substances. Therefore, some chemicals listed here
do not appear in Tables 2-2 and A-6.
(b) Examples include chlorine, anhydrous ammonia, hydrochloric acid, and sulfuric acid.
-------
A-10
Severity indexes have been developed in many ways, and they are invariably
controversial. One approach involves estimating severity in economic terms
how much does each different health effect cost for treatment, or how much would
each different health effect reduce an individual's expected lifetime earnings.
The first of these approaches tends to assign low severity scores to effects
that are quickly fatal, and high scores to illnesses needing protracted treat-
ment. The second approach assigns low scores to diseases affecting older people
and high scores to those affecting the young. Another approach to severity
scores involves polling people about which effects they would least like to
suffer. This has the disadvantage of asking laymen to speculate about effects
about which they have little technical understanding.
A subcommittee of the work group developed an approach to assess the dis-
parate health endpoints. The subcommittee began with a severity index that had
been developed for a contract report to the EPA Environmental Criteria and
Assessment Office in Cincinnati. The index basically ranked health effects
by how threatening they were to the viability of the organism. The index
involved two steps:
o A ranking of organs. Category I organs (most important) include those
whose impairment or loss is fatal. Gradations extend to Category IV
organs, which include those found in animals which have no counterparts
in humans.
o A seven-point endpoint severity scale. A score of 1 (lowest) is given
to functional impairment of Category IV organs. A score of 4 is given
to mild functional impairment in Category I organs or major impairment
in Category II organs, etc.
A full description of this severity index appears in Table A-3. This index
has in no way been approved by EPA; in fact, it is highly controversial. Many
reviewers have disagreed strongly with the idea of ranking or comparing organs
by importance. There is much less reluctance to compare different effects to
the same organ.
With misgivings, the subcommittee nevertheless used this index as the
starting point in assigning severity scores to the health endpoints. The
scores assigned by the subcommittee ranged from 1 (least severe) to 7 (most
severe). While the index provided a guide in developing scores, the assigned
score for a health endpoint depended primarily on the subcommittee's qualitative
judgments about the extent to which the health effect was life threatening,
permanent, reversible and manageable therapeutically. The severity scores
assigned by the subcommittee are listed in Table A-4. Table A-5 indicates how
the assigned severity scores are distributed. The distribution of scores looks
surprisingly like a normal distribution, with very few endpoints ranked as
either extremely benign (score of 1) or extremely severe (score of 7) and most
clustering around a score of 4.
One important point about the severity scores is they are ordinal but not
cardinal. A score of six is not twice as bad as a score of three. Although
it is difficult to be quantitative about such a qualitative concept as severity,
the subcommittee believes that there is a large difference in severity (perhaps
up to an order of magnitude difference) involved in a one point difference in
ranking.
-------
Table A-3
Ranking of Organs
Category I - Includes organs, impairment or loss of which is fatal and
cannot be compensated for at all, or only with heroic measures
(i.e., expensive mechanical devices, transplantation). Also
includes gonads, loss of which prevents reproduction.
Lung, Heart, Brain/Spinal Cord, Kidney, Liver, Bone Marrow, Gonads
Category II - Includes organs whose loss or impairment may be fatal, but
which can be compensated for by replacement therapy. Also
includes organs, impairment or loss of which indicates an
adverse effect on immune function or hematopoietic function
which may be life threatening.
Adrenal, Thyroid, Parathyroid, Pituitary, Pancreatic Islets, Pancreas,
Esophagus, Stomach, Small Intestine, Large Intestine, Lymph Node, Spleen,
Thymus, Trachea, Pharynx, 'Urinary Bladder, Skin
Category III - Impairment or loss of any of these organs is not life
threatening but may result in severe functional or emotional
handicaps.
Accessory reproductive organs (Oviduct, F.pirtidvmis, Uterus, Prostate,
Coagulating Gland, Seminal Vesicle. Ductus D^ffirens, Peais, Vagina), Eye,
Bone, Nose, Nerve, Muscle, Urinary Bladder, Blood Vessel, Ear, Gall
Bladder, Harderian and Lacrimal Gland, Larynx, Mammary Gland, Salivary
Gland, Tongue, Tooth, Ureter, Urethra
Category IV - These organs are not found in humans and toxic lesions
(noncarcinogenic) in these organs are not readily extrapolable
to humans.
Clitoral/Preputial Gland, Zymbal's Gland, Anal Glands
-------
Table A-3 (Continued)
Toxicity Test Endpoint Severity Scores
Toxicity Test Endpoint
Severity Score (T)
1.0
2.0
3.0
4.0
5.0
Toxicity Test Endpoints
Body wt. change, food and/or water
consumption changes, minor serum
electrolyte changes, minor clinical
changes, functional impairment of
category IV organs
Small hematological changes, functional
impairment in category III organs,
organ weight change in category II-IV
organs
Mild function impairment in category II
organs, severe impairment in category
III organ, minor organ weight changes
in category I organs
Mild functional impairment in category I
organs (small increases in urine con-
centration, proteniuria, enzymuria;
changes in conjugated and unconjugated
bilirubin, increases in sleeping time;
small changes in ECG; minor changes
in pulmonary function tests; mild
behavioral changes; mild alterations
in reproductive function), major
impairment in category II organs,
major organ weight changes in category
I organs
Definite functional impairment in
category I organs (moderate increases
in urine concentration, proteinuria,
enzymuria; serum increases in SGPT,
SCOT, LDH, ICOH; increases in BSP
retention; moderate changes in ECG;
moderate impairments of pulmonary and
reproductive function; definite
behavioral changes; developmental
toxicity with maternal toxicity
-------
Table A-3 (Continued)
6.0 Major degree of functional impairment in
category I organs (substantial
increases in proteinuria, enzymuria,
BUM; substantial increases in serum
levels of SGPT, SCOT, LDH, ICDH,
albumin; severe changes in ECG, and
pulmonary and reproductive function;
substantial behavioral alterations)
7.0 Severe central nervous system,
respiratory or cardiovascular
depression, mortality, developmental
toxicity without maternal toxicity
-------
TABLE A-4:
RANKING OF SPECIFIC NON-CANCER HEALTH ENDPOINTS
SPECIFIC SCORE
ENDPOINTS (1-7)
Cardiovascular
- unspecified
- increased heart
attacks 7
- aggravation of
angina 5-6
- increased blood
pressure 4
Developmental
- fetotoxicity 6
- abnormal ossi-
fication (see teratogenicity)
- low birth weight 4
- teratogenicity 7
HematopoiTetic
-unspecified
- decreased heme
production 4
- bone marrow
hypoplasia 5
- impaired heme
synthesis 4
- methemoglobinemia 5
Immunological
- unspecified
- herpes 1
- allergic reactions 3
- increased
infections 4
Kidney effects
- unspecified
- tubular
degeneration
- dysfunction
- hyperplasia
- hypertrophy
- atrophy
- necrosis
5
3
3
3
4
6
SPECIFIC SCORE
ENDPOINTS (1-7)
Liver effects
- unspecified
-hepatitis A 5
-jaundice 4
-increased weight 3
-increased enzymes 2
-necrosis 6
Mutagenicity
-unspecified
-cytogenetic 4
-heriditary
disorders 7
Neurotoxic/Behavioral
- unspecified
- retardation 7
- reduced corneal
sensitivity 2
- retinal disorders 4
- visual aging 2
- AChe inhibition 5
- learning disabili-
ties 6
- neuropathy 6
- decreased sensory
perception 3
- irritability 3
- tremors 4
- convulsions 6
- sensory irritation 2
Reproductive
- unspecified
- post implantation
losses 4
- testicular degen-
eration 4
- spermatocyte
damage 4
- decreased testi-
cular weight 3
- uterine hypoplasia 3
- aspermia 6
- increased resorp-
tions 4
- giant cell forma-
tion 2
- increased spontan.
abortions 5
-------
TABLE A-4 CONT'D:
RANKING OF SPECIFIC NON-CANCER HEALTH ENDPOINTS
SPECIFIC
ENDPOINTS
SCORE
(1-7)
Respiratory
- unspecified
- emphysema 6
- nasal irrita-
tion 2
- pulmonary irri-
tation 3
- nasal ulceration 3
- mucosal atrophy 3
- bronchitis 4
- pulmonary impair-
ment 4
- lung injury 4
- pneumonia 5
- pulmonary edema 6
- Pontiac fever 5
- congestion 3
- hemorrhage 4
- alveolar collapse 5
- fibrosis 5
- nasal cellular
irritation 2
- lung structure
changes 5
- aggravation of
asthma 4
- increased resp.
disease 4
- bronchoconstric-
tion 4
- decreased mid-
expiratory flow
rates 3
- increased respira-
tory infections 4
SPECIFIC
ENDPOINTS
SCORE
(1-7)
Other
- unspecified organ
effects
- unspecified acute
effects
- mortality 7
- eye irritation 2
- dental erosion 3
- cataracts 5
- leishmaniasis 3
- adrenal
- gastrointestinal
disease 4
- bone damage
dental mottling 2
- symptomatic
effects (headache) 3
- Legionnaires dis. 5
-------
TABLE A-5:
DISTRIBUTION OF RANKINGS TOR NCN-CMKER HEALTH QJDPOIOTS
SCORE:
herpes
(non-infectious)
liver-increased
enzymes
reduced comeal
sensitivity
sensory irritation
giant cell formation
nasal irritation
nasal cellular irri-
tation
eye irritation
dental nettling
allergic reactions
kidney-hyperplasia
kidney-hypertrophy
liver-increased
weight
decreased sensory
perception
irritability
decreased testicular
weight
uterine hypoplasia
pulmonary irritation
nasal ulceration
nucosal atrophy
> pulnonary congestion
decreased mid-expir-
atory flow rates
dental erosion
leishmaniasis
symptomatic effects
(headache)
increased blood
pressure
low birth weight
decreased home
production
impaired heme
synthesis
increased infections
(iimui.)
kidney-atrophy
jaundice
nutagenicity-cytogenetic
retinal disorders
trenors
post implantation losses
' testicular degeneration
spermatccyte damage
increased resorpticns
bronchitis
pulmonary inpairment
lung injury
respiratory-henorrhage
aggravation of asthma
increased respiratory
disease
bronchoconstriction
increased respiratory
infections
gastrointestinal disease
aggravation of
bone marrow hypoplasia
methenoglobinemia
kidney-tubular degener-
ation
hepatitis A
AChE inhibition
increased spontaneous
abortion
pneunonia
Pontiac fever
alveolar collapse
fibrosis
lung structure changes
cataracts
Legionnaires disease
angina -
- fetotoxicity
- kidney-necrosis
- liver-necrosis
- learning disabilitie
- neuropathy
- convulsions
- aspermia
- emphysema
- pulnonary edema
- increased heart attack
- teratogenicity
- nutagenicity-*ereditary
disorders
- retardation
- mortality
-------
A-17
The subcommittee members expressed strong reservations about the endpoint
severity scores they had developed. They were willing to go out on this shaky
limb and develop severity scores because completion of the non-cancer risk
project required making such rough judgments in many areas besides this one.
Developing severity scores seemed no more far-fetched or unfounded than other
steps in assessing non-cancer risks. But the subcommittee cannot support use
of these severity scores in other contexts or for other purposes. Ihe scores
are highly subjective, and have been reviewed and approved by no one. The base
index has not been peer-reviewed or in any way approved by EPA. The subcommittee
developed the scores for the endpoints in many cases without a detailed
understanding of the endpoint and without reference to the studies reporting
observation of the endpoint. In many cases the endpoint has been observed only
under experimental conditions in animals; the subcommittee could only assume
generally similar responses in humans.
In addition to developing severity scores, the endpoints subcommittee
resolved another difficult issue for the non-cancer work group. Many toxic
chemicals can produce more than one type of health effect. Cadmium, for example,
can cause renal disfunction at low doses, but can also cause teratogenic effects
at higher doses. The severity of the health effect may also vary with the
dose: as cadmium dosage increases, renal disfunction may instead become renal
tubular degeneration. The instructions to the program offices on generating
the summary sheets were not specific on which of multiple endpoints to report
and how to describe gradations of effects. Some summary sheets in fact reported
multiple endpoints with experimental data for each, while others reported data
only on a single driving endpoint (the endpoint that drives regulatory concern,
or the significant endpoint that occurs at the lowest dose). Some summary
sheets reported only a single gradation of effect, while others reported ranges.
The endpoints subcommittee suggested two simplifying rules that the
workgroup agreed to try to follow in assessing health effects:
o Deal with the driving endpoint only. All summary sheets have consis-
tently reported on driving endpoints, while inclusion of data on other
endpoints is erratic.
o Focus on the grade of the effect as it would occur at the dose for
which ambient exposures occur. If a dose at a level far above the LOEL
is being considered, for example, grade the effect as it would occur at
that high dose.
5. Exposure
In theory, for any chemical there is an entire distribution that relates
the size of the exposed population to the ambient concentration or dose. In
general there will be large numbers of people exposed to small amounts of the
chemical, and smaller numbers of people exposed to larger amounts of the chemical.
Such an exposure distribution is illustrated in Figure A-l.
For the substances covered by the summary sheets, there was rarely enough
exposure information available to specify much of the exposure distribution.
For most substances, only one data point was available; the exposed population
had been estimated for only one ambient concentration or corresponding dose.
For a few substances (mostly air pollutants), multiple data points were avail-
able. For many substances, no data points were available, or estimates were
-------
A-18
very rough (e.g., somewhere between 1 and 10 million people are exposed to a
specified ambient concentration of the pollutant).
Figure A-l: Example of an Exposure Distribution
I V
Number of
people
exposed
Concentration or dose
A further problem was that the ambient concentrations and the exposed
populations reported on the summary sheets frequently had been estimated under
inconsistent assumptions. For example, a summary sheet might claim that the
entire U.S. population of 240 million is exposed to ambient levels of the
pollutant of concern. The concentration at which this exposure is assumed to
occur is then estimated by reference to annual average monitored levels of the
pollutant in urban areas, the only places where the pollutant is routinely
monitored for. In fact, though, urban ambient levels of the pollutant are
probably much higher than rural levels. In reality the entire U.S. population
of 240 million is actually exposed to ambient levels lower than the urban
average, or only the urban fraction of the 240 million people are actually
exposed to the urban average monitored concentration.
More generally, a problem in reviewing exposure estimates across different
chemicals was use of widely differing levels of conservatism. For example, the
Superfund estimate was initially generated under a conservative assumption that
all people served by ground water within a three mile radius of a Superfund
site drank contaminated water. By contrast, the RCRA estimate was generated by
modeling contaminant transport and potential exposure at actual RCRA sites,
allowing contaminants to affect only downgradient water users, and providing
for degradation, retardation, dilution, etc.. The result was a RCRA estimate
of exposed population many orders of magnitude smaller than Superfund's. This
result was due primarily to differences in conservatism of the modeling
assumptions, and Superfund's approach was eventually revised.
Another sort of problem involved the distinction between the population
exposed to a substance and the population at risk of suffering adverse health
effects from the substance. If only sensitive subgroups (e.g., infants,
asthmatics) are at risk from the substance, the population at risk will only
be a fraction of the exposed population.
The workgroup adopted some principles for dealing with these problems and
to provide a consistent approach to exposure assessment.
o All available concentration/exposed population data pairs should be
used. Substances with more than one data pair had all the available
data pairs carried forward into the next steps of the work group's
methodology (scoring and ranking).
-------
A-19
o If a concentration or an exposed population can be estimated only
within a range, the range will be converted to a single number by
taking the geometric mean.
o A concentration and its corresponding exposed population should be
generated under consistent ground rules.
o Exposure estimates across the different substances should exhibit a
similar degree of conservatism. Excessively conservative estimates
should either be revised, or they will mentally be scaled down in
the final ranking stage.
o The population estimated should be the population at risk for the
particular effect. For example, if the risk applies only to infants,
then the estimate should be of the infant population.
The workgroup did not, however, rework and revise all the exposure estimates
to comply with these principles. For the most part the general direction and
magnitude of any error in an exposure estimate could be guessed at, and if this
error was not likely to be large enough to affect the work group's conclusions,
the estimate was left alone. The work group ultimately converted the estimated
population exposed to a substance into a score from 1 to 4. One point of change
in the score represented a difference in exposed population of two orders of
magnitude. Thus, it was not likely to matter if an estimated exposure varied
from the actual by a factor of 2, 5, or even 10. Most of the problems in
exposure estimates seemed to be within this range, and the work group thus
believed most of the population estimates to be sufficiently accurate for the
purposes of this project.
6. Potency
The work group was interested in developing a mathematical representation
of the potency of a substance in inducing its adverse health effects. Potency
is generally the relationship between the dose of a substance and the likelihood
that the dose will produce the health effect. A dose-response function for
a substance is a representation of its potency. A typical dose-response function
for a non-carcinogenic effect is shown in Figure A-2 below.
Figure A-2: Typical Dose-Response Function for Non^Carcinogen
Probability of 1
the health
effect
0
Most non-carcinogenic dose-response functions are thought to involve a threshold
(T on the graph). The threshold is the minimum level of dose that is necessary
-------
A-20
before there is any significant probability of the health effect occurring. Or
alternatively, the threshold is the highest level of dose for which there is no
probability of the health effect occurring. Noncarcinogenic health effects
that are thought not to involve thresholds include nutagenicity and in some
cases teratogenicity.
Research on responses to different doses of a substance will yield data on
the highest dose at which no effect has been observed (the "No observed effect
level" or NOEL) and the lowest dose at which some effect has been observed (the
"Lowest observed effect level" or LOEL). These two levels presumably bracket
the threshold level. When regulating a substance and establishing a level of
dose that is safe, EPA typically applies uncertainty factors to the NOEL. The
Acceptable Daily Intake (ADI) or the Reference Dose (RfD) of a substance is the
NOEL divided by an uncertainty factor ranging usually from 10 to 1,000. The
magnitude of the uncertainty factor is related to the degree of confidence we
have in our knowledge about the health effects of the substance in question. If
we are dealing with a well-studied chemical with results from humans we can be
reasonably confident of where the threshold is. We can set the RfD only slightly
below the LOEL (using a small uncertainty factor of 10 or so) and still be
confident that the RfD really is a safe level. But if we are dealing with a
poorly understood chemical, with only experimental evidence from an animal
study, we cannot be very sure of where the threshold is. In this case, we must
use a larger uncertainty factor (1,000 or so) and set the RfD well below the LOEL
in order to have the same level of confidence that the RfD represents a level
that really is safe.
Although the work group might have liked to use real dose-response functions
to represent the potency of the selected substances, acquiring the data would
have been very difficult. Whereas a LOEL, NOEL, ADI or RfD is easily available
for many substances, the dose-response functions themselves are not. In some
cases they could be estimated by reviewing the original studies, but this would
have been very time consuming. In EPA practice, dose-response functions are
rarely used to assess risks of non-carcinogenic effects. Instead, a margin of
safety (MOS) approach is employed by many programs. For most programs, the MOS
is the RfD divided by the dose of the chemical actually received. If the MOS
is low, the dose is at levels near the RfD and the exposure may be of regulatory
concern; if the MOS is high the problem is minimal.
The work group decided to use an inverse MOS approach to reflect potency.
The MOS approach is generally used in cases where low doses may be high enough
to threaten to cause a problem. The work group inverted this concept to provide
a notion of how large a problem there is when doses are high. A ratio which
the work group termed the "individual exposure ratio" was calculated as the
concentration at which exposure occurs divided by the RfD. The more a concen-
tration level is above the RfD, the higher this ratio becomes. The individual
exposure ratio correlates roughly with the probability that an individual
encountering the substance at a given concentration will suffer the health
effect. At concentrations near the RfD we expect little likelihood of the
effect; the ratio signifies this with a value near 1. At concentrations far
above the RfD, the likelihood of the health effect will be much greater, and
the ratio signifies this with values that are much higher. A few points relating
to the individual exposure ratio should be noted:
o The concentration and the RfD must be expressed in similar units so
that the ratio is dimensionless. This can be done either by convert-
-------
A-21
ing the concentration to an eguivalent dose (by using standard uptake
assunptiol*) and dividing by a RfD in traditional units of mg/kg/day,
or by leaving th% concentration as is but expressing the RfD in
concentration tetmt.
o The work group decided that there was negligible risk at levels below
the RfD (i.e., when the ratio was less than one). The work group
thus did not consider any substances in exposure settings where
concentrations were less than RfDs.
o In order to calculate this individual exposure ratio, RfDs were
needed for all substances selected to represent the 31 problems.
Approved ADIs or RfDs are available for only some of the substances
of concernf so the workgroup had to generate a large number of unoffi-
cial RfDs. This was done in the traditional way by applying uncer-
tainty factors to LOELs or NOELs in the experimental literature.
o The work group debated whether to use the RfD, NOEL, or LOEL in the
denominator of the individual exposure ratio. The decision to use the
RfD was made for two reasons: to be consistent with general EPA
practice in focusing on RfDs when making risk management decisions,
and to base the ratio on a level we know presents negligible risks.
(If the denominator were the NOEL or the LOEL, an ambient exposure
below that level -a ratio of less than one might nevertheless be
above the threshold and be unsafe.)
The choice of RfD rather than NOEL or LOEL perhaps has an
important influence on the eventual rankings of the problem areas.
The NOEL is converted to an RfD by dividing by an uncertainty factor,
the magnitude of which depends basically on how poorly understood the
health effects of the substance are. For well-understood chemicals
such as certain criteria air pollutants the uncertainty factors are
small. For more unusual chemicals such as many pesticides, the un-
certainty factors are much larger. In general, the RfDs for criteria
air pollutants are slightly lower than their NOELs, while the RfDs
for pesticides are far lower than their NOELs. A pesticide exposure
at the NOEL for that pesticide will tend to get an individual exposure
ratio score 10 or 100 times higher than a criteria air pollutant will
get for an exposure at its NOEL. If NOELs or LOELs were used as
the denominator, the workgroup's estimated "potency" for chemicals
such as pesticides would be reduced sharply, while the estimated
potency for chemicals like criteria air pollutants would be reduced
only minimally.
The work group realizes that there are some major conceptual problems in
representing potency or the probability of an effect at some exposure level
through this individual exposure ratio. First, the real probability of an
effect can actually range only between zero and one. But our individual expo-
sure ratio can range from zero to extremely large. Whereas a real dose-response
function will asymptotically approach a probability of effect of one as the
dose increases, our individual exposure ratio continues to grow indefinitely.
Secondly, using our ratio implicitly assumes an identical shape to the dose-
response function for every substance. We implicitly assume that at a dose of
twice the RfD for a chemical there will be the same probability of an effect no
matter what the chemical. This clearly is not true? at least in part because
-------
A-22
of the differential uncertainty factors intervening between the RfD and the
LOEL for different chemicals. We also implicitly assume for all substances
that a dose of ten times the ADI is twice as bad as a dose of five times the
RfD. There is no reason, in reality, that dose-response functions have to
exhibit these properties. Real dose-response functions may have very different
shapes and slopes for different chemicals.
Despite these problems, the work group believes that for the purposes of
this project the "individual exposure ratio" is a reasonable and practical
representation of the potency of a substance, or of the likelihood of an effect
resulting from exposure to the substance at a given ambient level. The individual
exposure ratio is not a precise measure of potency, but the two are roughly
correlated.
7. Combining Data on Endpoints, Exposure and Potency
At this point, the work group had accumulated data for representative sub-
stances on:
o The driving health endpoint expected from exposure to the substance,
and the severity of that endpoint. The severity of the endpoint was
expressed by a score ranging from 1 to 7.
o The amount of exposure to the substance. This was expressed as the
population exposed to a given concentration of the substance.
o The potency of the substance in causing the adverse health effect for
exposures at the given concentration. This was expressed by an
"individual exposure ratio" derived by dividing the dose at ambient
concentrations by the RfD for the substance (or by using the equivalent
concentrations).
These data were organized into a large matrix, in the format of Figure A-3.
A subgroup of the work group recommended a procedure for converting the
data in this matrix into a basis for assessing risks from particular substances.
The subgroup suggested multiplying the individual exposure ratio (akin to
individual risk) by the exposed population to derive a "population exposure
ratio" (akin to population risk or number of cases), as in Column 9 of the
matrix. This population risk would then be multiplied by the severity index in
order to convert cases of disparate health effects into common terms. The
severity-weighted number of cases (Column 10 of the matrix) would then provide
the overall score for assessing the non-cancer risk from a particular substance
within a problem area.
When the work group as a whole met to begin ranking the 31 problems, they
found in practice that they did not like this highly quantitative approach the
subgroup had developed. Three problems arose:
1. For some substances, there was not a complete set of data available
on the severity score, the exposed population and the individual
exposure ratio. For substances missing some of this data, an overall
score could not be calculated.
-------
FIGURE A-3
SUMMARY DATA ON ENDPOINTS, EXPOSURE. AND POTENCY
(1)
Problem
Area
(2)
Chemical
(J)
Population
Number
(4)
Exposure
Concentration
(5)
RfD or
ADI
(6)
Effects
(7)
Nature
Score
(8)
Individ. Exp.
Ratio
(9)
Pop. Exp.
Ratio
(3) x (8)
(10)
Overal1
Score
(9) x (7)
-------
A-24
2. In concept the overall score represented something like the nationwide
number of severity-weighted cases. Many members of the work group
felt that this score overemphasized population risk, and that individual
risk should be an equally important determinant of the final ranking.
A problem area which caused very high risks for the small number of
exposed individuals should perhaps be scored as high even if population
risks from this problem-were small.
This conceptual problem with the scoring process the subgroup
had tentatively designed was exacerbated by a mathematical problem.
The variance across substances in the severity score was small
(scores ranged from 1 to 7). The variance in the individual exposure
ratios was much larger, with scores ranging across about three
orders of magnitude. But the variance in the exposed population
numbers was huge, with the range covering about seven orders of
magnitude. The result, when severity, individual exposure ratio and
population were multiplied together to determine the overall score,
was that the overall score depended most heavily on the exposed
population. In statistical terms, the variance in the overall score
was determined largely by the variance in population. Mathematically,
the other factors didn't matter much in determining the overall score.
3. Perhaps most importantly, -the work group did not feel that the level
of precision implied by these mathematical operations and by the
quantitative overall score matched the degree of confidence the
work group had in the data. The severity index represented only the
rough judgment of the endpoints subcommittee. The individual exposure
ratio was subject to all of the theoretical misgivings discussed in
the potency section. The exposed population figures were known in
most cases to be shaky; generated by a variety of incompatible
techniques and frequently biased by differing degrees of conservatism.
The work group responded to these problems by turning to a much less
precisely quantitative scoring scheme. The three factors severity index,
individual exposure ratio, and exposed population were turned into three
scores for each substance, with each score ranging from 1 to 4:
o A severity score. The severity index was converted to a score
as follows:
Severity score Severity index from endpoints group
1 1-2
2 3-4
3 5-6
4 7
o A ratio score. The individual exposure ratio was converted to a
score as follows:
Ratio score Individual exposure ratio
1 1-10
2 10 - 100
3 100 - 1,000
4 XL, 000
-------
A-25
o A population score. The exposed population was converted to a score
as follows:
Population score Number of people exposed
1 <1,000
2 1,000 - 100,000
3 100,000 - 10,000,000
4 >10,000,000
Converting the data on individual substances into these scores solved many
of the work group's problems. First, where data were missing, the work group
felt reasonably confident in its ability to assign an appropriate score between
1 and 4 to fill the gap. Secondly, with severity, exposed population and
individual risk new represented by scores ranging equally from 1 to 4, much of
the statistical domination of the final ranking by population could be avoided.
The influence of population was also diluted somewhat by requiring a two order
of magnitude increase in order to gain a point in the population score, while
only a one order of magnitude increase was needed to gain a point in the ratio
score. Also, the three scores could be combined in various different ways
intended to emphasize or deenphasize the importance of any of the factors.
Finally, the work group felt that the 1 through 4 score signified about the
right level of precision in the data that had been generated. For example,
although the absolute number of people exposed to a substance at a given concen-
tration might easily have been mis-estimated by a factor of two or five, the
work group felt it unlikely that a mis-estimate was by two orders of magnitude,
the amount necessary to produce a change in the population score.
A few more issues had to be resolved before the work group had a full set
of scores for all the representative substances in all problem areas.
First, some programs were able to produce data on incidence of adverse
health effects rather than data on concentrations and exposed populations. For
example, the drinking water program could produce data on the number of reported
cases of giardiasis and legionnaire's disease, but had no data on typical concen-
trations of or exposures to these microbial contaminants. Similarly, data were
available on mortality and morbidity from accidental chemical releases, and on
the number of pesticide poisonings attributable to different agents. In these
areas again it would have been extremetly difficult to estimate typical concen-
trations and exposed populations.
In these cases, the incidence data were converted to a population score
and a ratio score in order to be consistent with the remainder of the data.
The population at risk of suffering the health effect measured by incidence
was estimated roughly and scored using the 1-4 population score. A ratio score
was then calculated as follows:
Average individual risk where annual incidence
data are on incidence = estimated population at risk
-------
A-26
Ratio Score
1
2
3
4
Avq. Individual Risk
<1.00E-06
l.OOE-06 - l.OOE-04
l.OOE-04 - l.OOE-02
>1.00E-02
Individual Exposure Ratio
1-10
10 - 100
100 - 1,000
>1,000
For comparison, the individual exposure ratios that would result in the same
scores if data were available on concentrations are shown also.
In effect/ we broke apart the incidence data into a population score to
represent the size of the population at risk, and into a ratio score to reflect
the degree of risk faced by an individual in that population. In one respect,
this process is unusual. Incidents,of harm are what we really care about in this
entire project. Our basic methodology is designed to take data on the components
of harm (exposure concentrations; population exposed; likelihood of harm at
those concentrations) and implicitly combine them. The paradox is that when we
have fairly good data on incidence, our methodology requires us to break that
data apart into some probably poorer estimates of exposure and likelihood of
harm, in order to be able to describe that environmental problem in terms
consistent with all the others. Nevertheless, the bulk of our data on chemicals
was in the form of exposed populations, ambient concentrations and likelihood
of harm rather than on incidence, and the work group believed that that a con-
sistent ranking scheme demanded that all the data be converted into one format.
The final step taken in scoring was to develop scores for the problem
areas for which we had inadequate data on representative substances. It
had not been possible to produce data sets on exposure concentrations and
exposed populations for substances representing several of the 31 environmental
problems. For problem areas like this, the work group scored the entire problem
area rather than individual representative substances. A representative of the
cognizant program described the problem area to the work group, and the work group
as a whole then scored the problem area. For inactive hazardous waste sites
(problem area t 17) for example, the work group thought that perhaps several
million people could be exposed to potentially harmful levels of toxic chemicals
in drinking water or air around such sites. But we did not think this number
was likely to be as high as 10 million people, so we gave inactive hazardous
waste sites a 3 for a population score. A similar sort of process was used to
give inactive hazardous waste sites a severity score and a ratio score. The
problem areas that were scored in this way as a whole, without scoring specific
representative substances, included sludge, accidental releases and all the
waste site categories (active and inactive hazardous, municipal and industrial
non-hazardous industrial, and mining). In some of these cases, the work group
decided to score the problem as a whole despite having some data on represent-
ative substances. (The particularly attentive reader may note that Table A-l
thus lists some substances that were ultimately not scored. For example,
although we had data on sulfuric acid, hydrochloric acid, etc. involved in
accidental releases, we chose to score the accidental release problem area
directly rather than scoring its representative substances.)
For certain problem areas, the work group lacked the information needed
to estimate scores for specific factors (i.e., population, health endpoint
and individual exposure ratio). Instead, the work group discussed what was
known about the problem with representatives from the relevant program office
-------
A-27
and agreed upon an overall ranking of high, median or low. This approach was
used for direct discharges to surface water (industrial), indirect discharges
to surface water (POTWs), non-point source discharges to surface water, estuaries,
wetlands and releases from storage tanks.
The final scores developed by the work group for the representative substances
and the problems scored as a whole are shown in Table A-6.
8. Producing the Final Ranking
The work group had produced a set of scores that generally represented the
exposed population, individual risk and the severity of the health endpoint for
each representative substance or problem area. Two steps remained before producing
a final ranking:
o For problem areas represented by multiple substances, deciding how to
combine the scores on individual substances to produce scores for the
problem as a whole.
o Deciding how to combine the three scores for a problem area into a
single ranking for the problem area.
The work group decided to look at some options on the second issue first. The
group was most concerned that they did not want to make a judgment that either
population risk or individual risk was more important. If the ranking for a
problem area would differ substantially depending on whether individual or
population risk were emphasized, the work group wanted to display two sets of
rankings. To test the importance of this issue, the work group developed a
concept called a scatter plot. The scatter plot would graph the individual risk
of each substance or a problem area against the population risk of this substance
or problem area. If most of the points in the scatter plot are on or near the
diagonal for which individual and population risk were equal, the issue of what
sort of risk to focus on is not terribly important. But if most of the points
are off this diagonal (i.e., if individual risk for a substance or problem area
usually does not look much like population risk for that substance or problem
area), the issue is an important one.
Thus, work group developed the matrix shown in Figure A-4. On the
horizontal axis is the population score given to the substance or problem area.
On the vertical axis is a representation of the severity of individual risk
associated with the substance or problem area (called the "individual score,"
developed by adding the severity score and the ratio score, dividing by two,
and rounding upward). Most of the substances and problem areas are either on
the diagonal where the population score and the individual score are equal or in
adjacent boxes (where one score differs from the other by 1). Relatively few
substances and problem areas have scores differing by 2 or more. The work group
concluded that the individual risk vs. population risk issue was not likely to
be critical in practice.
More interesting conclusions can be drawn from the matrix. The items that
rank the highest (scores of 4,4; 4,3; or 3,4) are associated with accidental
releases (Problem 21), air and radiation (Problems 1, 2, 4, 5, and 6), drinking
water (Problem 15) and consumer exposures (Problem 30). More items are given
high population scores than are given high individual scores. The work group
interpreted this observation as consistent with the fact that EPA generally
deals with broad exposure environmental problems, and not so much with narrow
-------
TABLE A-6
SCORES AND ACTUAL NUMBERS FOR ALL ENVIRONMENTAL PROBLEM AREAS
Problem
Area
Substance or Problem Area
Population
Exposed
Pop.
Score a
Driving Health
Epdpoint
1 Lead(children)
1 Carbon monoxide
1 Sulfur dioxide
1 Particulate matter(elderly, diseased)-Acute
1 Particulate matter - Chronic
1 Acid Aerosols
1 Ozone Acute
1 Ozone - Chronic
2 Benzene
2 Carbon tetrachloride
2 Chlorine
2 Chromium
2 Formaldehyde
2 Hydrogen sulfide
3 Other air pollutants
4 Radon (indoor air only)**
5-a Benzene
5-b Benzene
5-c Benzene
5-a Carbon tetrachloride
5-b Carbon tetrachloride
5-c Carbon tetrachloride
5 Environmental tobacco smoke(adults) >
5 Environ.tobac. smoke(infants & children) >
5-a Formaldehyde
5-b Formaldehyde
5 Nitrogen dioxide - Acute
5 Nitrogen dioxide Chronic >
5-a Xylene
5-b Xytene
6 Radiation - occupational
6 Radiation - consumer
7 UV Radiation/Ozone depletion **
& Carbon dioxide and global warming
9 Direct indus. disch. to surf, water
10 Indirect disch. to surf, waterd'ncl. POTUs)
11 Nonpoint source discharges to surface water
12 Sludge **
13 Discharges to estuaries, coastal waters, oceans
14 Discharges to wetlands
15 Lead (children)
15 Legionella **
15 Nitrate (infants)
15 Pathogens (Giardia/Viruses)**
15 Chlorine disinfectants
16 Active hazardous waste sites**
17 Inactive hazardous waste sites**
18 Municipal non-hazardous waste sites**
19 Industrial non-hazardous waste sites**
20 Mining waste**
2.68E+03
3.23E+06
2.75E+05
3.00E+06
1.20E+07
1.OOE+07
1.08E+08
2.37E+08
1.36E+07
8.50E+06
2.31E+05
2.70E+06
1.70E+08
9.90E+05
1.14E+08
6.75E+07
4.50E+06
4.76E+07
8.60E+06
40E+06
OOE+07
OOE+07
OOE+08
6.00E+06
OOE+06
OOE+07
76E+07
40E+06
95E+06
2.38E+08
1.67E+07
1.50E+04
1.50E+08
2
3
3
3
4
4
4
4
4
3
3
3
4
3
4
4
4
3
4
3
3
4
4
4
3
3
4
4
3
3
4
4
learning disabilities
aggravation of angina
aggravation of asthma
premature mortality
respiratory symptoms
increased mortality
increased resp. infections
lung structure changes
bone marrow hypoplasia
jaundice
pulmonary edema
bronch i t i s
pulmonary irritation
pulmonary irritation
Group consensus -- dropped
bone marrow hypoplasia
bone marrow hypoplasia
bone marrow hypoplasia
jaundice
jaundice
jaundice
inc. mort. from heart disease
hosp. for bronchitis/pneumonia
pulmonary irritation
pulmonary irritation
lung function changes
lung injury/structure changes
teratogenicity
teratogenicity
mutagen. (hereditary disorders)
mutagen. (hereditary disorders)
cataracts
Endpoint
Score b
3
3
2
4
2
4
2
3
3
2
3
2
2
2
4
3
3
3
2
2
2
4
3
2
2
2
3
4
4
4
4
3
Individual exposure ratio
or Incidence/population *
1.70E+00
4.80E+00
2.50E+01
2.50E+01
2.60E+01
1.00E+02
1.00E+02
2.92E+02
1.20E+01
1.00E+01
1.70E+02
3.00E+00
1.33E+04
2.32E+02
16E+00
16E+01
16E+02
17E+00
17E+01
17E+02
70E+01
25E+01
20E+02
9.75E+02
2.00E+01
1.14E+01
3.16E+00
3.16E+01
1.65E-05 *
6.50E-07 *
Group consensus
Group consensus
Group consensus
Group consensus
Group consensus
Group consensus
learning disabilities
Legionnaires' disease
methemoglobinemia
hepatotoxicity
dropped
low rating; but no
medium rating; but
medium rating; but
1
medium rating; but
low rating; but no
3
3
3
2
2
2
2
2
2
1
specific scores
no specific scores
no specific scores
no specific scores
specific scores
2.11E+00
1.00E+00
Ratio
Score c
1
1
2
2
2
3
3
3
2
2
3
1
4
3
1
1
2
3
1
2
3
2
2
3
3
2
2
1
2
2
1
1
1
2
1
3
1 **
1
1
1
2
1
-------
TABLE A-6 (CONTINUED)
Problem
Area
Substance or Problem Area
Population
Exposed
21 Accidental releases - Toxics (mortality)**
21 Accidental releases - Toxics (morbidity)**
22 Accidental releases -- oil spills
23 Releases from storage tanks
24 Other groundwater contamination
25 Aldicarb**
25 Diazinon**
25 EPN**
26 Dinoseb - males**
26 Dinoseb - females**
26 Paraquat**
26 Ethyl parathion**
27 Pesticides (Ground/Surface water) -- carbamates**
27 Pesticides (Indoor Air) -- cyclodienes**
28 New toxic chemicals
29 Biotechnology
30 2-Ethoxyethanol 2
30 Methylene chloride 2
30 Formaldehyde 1
31 2-Ethoxyethanol 4
31-a Methylene chloride 2
31-b Methylene chloride 3
31 Perchloroethylene**
31 Formaldehyde 1
Pop. Health Endpoint Individual exposure ratio
Score a Endpoint Score b or Incidence/population *
.50E+06
.50E+06
2.50E+05
OOE+05
.50E+07
OOE+06
50E+05
70E+05
.OOE+04
10E+06
death
injuries
Group consensus
Group consensus
Group consensus
AcHE inhibition
AcHE inhibition
AcHE inhibition
dropped
low rating;
dropped
8.00E-04*
2.70E-02*
fibrosis
AcHE inhibition
AcHE inhibition
increased liver weight
No information available
No information available
teratogenicity 4
liver histopath. alterations 2
sensory irritation 1
teratogenicity 4
liver histopath. alterations 2
liver histopath. alterations 2
renal/liver histopath. changes 2
sensory irritation 1
but no specific scores
4.00E-04
3.20E+00
8.81E+02
2.10E+03
1.26E+00
4.00E+02
4.76E+03
3.20E+03
Ratio
Score c
=========
3
4
Incidence/Population = number of cases / number of people exposed; calculated when the number
of cases were available and no Individual Exposure Ratio was available.
Group consensus.
-------
TABLE A-6 (CONTINUED)
a Pop. Score derived as follows:
Score Number of people exposed
1 < 1.000
2 1,000 - 100,000
3 100,000 - 10,000,000
4 > 10,000,000
b Health Endpoint Score derived as follows:
Score Score given by Endpoints Group
1 1-2
2 3-4
3 5-6
4 7
Ratio Score derived as follows:
Score Ind. Exp. Ratio # cases / # people exposed*
1 1-10 < 1.00E-06
2 10 - 100 1.00E-06 - 1.00E-04
3 100 - 1000 1.00E-04 - 1.00E-02
4 > 1000 > 1.00E-02
* Incidence/Population = number of cases / number of people exposed; calculated when only the number
of cases were available and no Individual Exposure Ratio was available.
-------
FIGURE A-4 MATRIX OF INDIVIDUAL SCORES AND POPULATION SCORES FOR ALL ENVIRONMENTAL PROBLEM AREAS
26 Dinoseb - females
16 Active hazardous waste sites
12 Sludge
26 Ethyl parathion
26 Dinoseb - males
31 Methylene chloride-b
1 Lead
15 Nitrate (infants)
20 Mining waste
21 Accidental releases Toxics (mortality)
21 Accidental releases - Toxics (morbidity)
1 Particulate matter - Acute
2 Chlorine
2 Hydrogen sulfide
5 Benzene- c
5 Carbon tetrachloride-c
5 Formaldehyde-b
5 Xylene-b
6 Radiation occupational
15 Legionella
25 EPN
26 Paraquat
30,31 2-Ethoxyethanol
31 Methylene chloride-a
31 Perchloroethylene
30,31 Formaldehyde
1 Carbon monoxide
1 Sulfur dioxide
2 Carbon tetrachloride
2 Chromium
5 Carbon tetrachloride-b
5 Nitrogen dioxide - Acute
17 Inactive hazardous waste sites
19 Industrial non-hazardous waste sites
25 Aldicarb
27 Pesticides (Ground/Surface Uater)
1 Acid Aerosols
1 Ozone - Acute
1 Ozone - Chronic
2 Benzene
2 Formaldehyde
4 Radon (indoor air only)
5 Benzene -b
5 Environmental tobacco smoke (adults)
5 Environmental tobacco smoke (children)
5 Formaldehyde-a
5 Nitrogen dioxide - Chronic
5 Xylene-a
6 Radiation - consumer
15 Pathogens (Giardia/Vi ruses)
30 Methylene chloride
1 Particulate Matter - Chronic
5 Benzene -a
5 Carbon tetrachloride-a
7 UV Radiation/Ozone depletion
15 Lead (children)
15 Chlorine disinfectants
18 Municipal non-hazardous waste sites
25 Diazinon
27 Pesticides (Indoor air)
POPULATION SCORE
* INDIVIDUAL SCORE was calculated by adding the Health Endpoint Score (Table A-6) and the Individual Exposure Ratio Score (Table A-6), and dividing that sum by 2.
All values were rounded upwards.
a,b,c - See Table A-6 for exposure distribution.
-------
A-32
high risk exposures to small subgroups such as OSHA might be concerned with.
The work group had hoped that the matrix would show clear trends for
all representative substances within a problem area. If, for example, all the
criteria air pollutants ranked quite high (4,4; 4,3; 3,4; or 3,3), then the work
group could easily rank criteria air pollutants as a whole very high. Such was
not the case, however. Disaggregated matrices concentrating on particular
areas (air and radiation, water, waste, pesticides and toxics) show a few trends,
but present no obvious conclusions. The work group had to return to the question
of how to aggregate scores for individual representative substances into scores
for a whole problem area.
A few members of the workgroup convened to perform various mathematical
operations on the scores, with the aim of producing different rankings of the
31 problem areas that resulted under different mathematical ground rules. The
different approaches included:
o Different ways for aggregating population, endpoint and ratio scores
into a total score for a problem area or substance. Methods were
used that weighted the three scores equally or that weighted one or
another factor more heavily.
o Different ways of aggregating scores (population score, endpoint
score and ratio score) for substances into scores for whole problem
areas. Three different methods were used:
1. Select the substance with the highest total score to represent
the problem. The rationale was that no problem should get a
lower score than any single component of the problem.
2. Represent the problem by the highest population score for any
substance, and with the average endpoint and ratio scores for
all the substances.
3. Represent the problem by the average population, endpoint and
ratio scores for all the substances.
Different combinations of these approaches were pursued, and resulting mathematical
rankings of the problem areas were shown to the entire workgroup. In general,
as suggested by the analysis, shifting the weights among population and indivi-
dual risk made relatively little difference to the final results. The process
for aggregating multiple substances had slightly more effect on the ultimate
mathematical ranking, but again there was substantial stability to whether a
particular problem area ranked at the top, in the middle, or low. Table A-7
shows a few of the alternative mathematical rankings that came from varying the
aggregation process.
The work group adopted these mathematical rankings as the starting point
for its final ranking of problem areas. A problem area was grouped by whether
it generally showed up high in the mathematical rankings, medium, or low. The
qualitative logic for such a ranking was then reviewed for each problem area.
Substantial discussion ensued on the following problem areas:
o The initially high rankings for indoor radon and radiation/not radon.
Data supporting these rankings came from modeled incidence of nutagenic
-------
TABLE A-7 COMPARISON OF RANKING RESULTS BY METHODS A.I, A.2, AND A.3
Problem
Area No.
=======S=
1
21
5
2
15
31
26
4
6
30
7
25
19
18
27
17
20
16
12
Problem Area
A.I*
Problem
Area No.
Problem Area
Problem
A.2** Area No.
Problem Area
Criteria air pollutants 11.0 21
Accidental releases - Toxics 10.0 6
Indoor air pollution - other than radon 10.0 26
Hazardous/toxic air pollutants 10.0 30
Drinking water 9.0 4
Worker exposure to chemicals 9.0 1
Application of pesticides 9.0 5
Radon (indoor air only) 9.0 2
Radiation - other than radon 9.0 31
Consumer product expsoure 9.0 25
UV Radiation/Ozone depletion 8.0 15
Pesticides residues on food 8.0 7
Industrial non-hazardous waste sites 7.0 27
Municipal non-hazardous waste sites 7.0 19
Other pesticide risks 7.0 18
Inactive hazardous waste sites 6.0 17
Mining waste 4.0 20
Active hazardous waste sites 4.0 16
Sludge 3.0 12
Accidental releases - Toxics
Radiation - other than radon
Application of pesticides
Consumer product expsoure
Radon (indoor air only)
Criteria air pollutants
10.0
9.5
9.5
9.0
9.0
9.0
Indoor air pollution - other than radon 8.9
Hazardous/toxic air pollutants 8.8
Worker exposure to chemicals 8.4
Pesticides residues on food 8.3
Drinking water 8.2
UV Radiation/Ozone depletion 8.0
Other pesticide risks 7.5
Industrial non-hazardous waste sites 7.0
Municipal non-hazardous waste sites 7.0
Inactive hazardous waste sites 6.0
Mining waste 4.0
Active hazardous waste sites 4.0
Sludge 3.0
21 Accidental releases - Toxics
6 Radiation - other than radon
4 Radon (indoor air only)
5 Application of pesticides
26 Indoor air pollution - other than radon
1 Criteria air pollutants
30 Consumer product expsoure
31 Worker exposure to chemicals
2 Hazardous/toxic air pollutants
7 UV Radiation/Ozone depletion
25 Pesticides residues on food
15 Drinking water
18 Municipal non-hazardous waste sites
19 Industrial non-hazardous waste sites
27 Other pesticide risks
17 Inactive hazardous waste sites
20 Mining waste
16- Active hazardous waste sites
12 Sludge
A.3**
=-====:
10.0
9.0
9.0
8.5
8.4
8.4
8.3
8.2
8.2
8.0
7.7
7.6
7.0
7.0
7.0
6.0
4.0
4.0
3.0
*A.1 Population Score + Endpoint Score + Ratio Score = Total Score.
When multiple chemicals exist for a Problem Area the one with the highest Total Score was selected.
**A.2 Highest Population Score + Average Endpoint Score + Average Ratio Score = A.2
***A.3 Average Population Score + Average Endpoint Score + Average Ratio Score = A.3
-------
A-34
and teratogenic effects. The work group was concerned about the
uncertainty of these estimates, and also believed that these effects of
radiation are very close to cancer effects and perhaps should more
properly be included with cancer risks. Accordingly, the rankings for
the two radiation problem areas were downgraded to medium risk.
The high ranking for indoor air pollution. This was due primarily to
environmental tobacco smoke, and substantial discussion took place about
whether the risk and exposure estimates here were believable. The work
group eventually decided they were persuasive.
The high ranking given to drinking water. Data were cited on levels of
disinfection byproducts widely found in drinking water and the possible
health effects from them. Recent findings on lead were also discussed.
It was suggested that there may be no threshold for cardiovascular
effects associated with exposure to lead. In addition, the group
discussed the more certain data on learning disabilities in children,
the driving endpoint used for lead. Some concern was expressed that
the scores for lead did not accurately reflect the seriousness of it as
a problem. However, the high risk ranking for drinking water seemed
clear, and the lead scores were not revised.
No ranking given to new toxic chemicals. Various approaches and rationales
were discussed in an attenpt to rank this problem area. Arguments were
made that it should rank low because: the new chemical review program
catches many of the potentially harmful chemicals, new chemicals often
replace riskier existing chemicals, and most new chemicals are low volume
specialty chemicals not resulting in any widespread exposure. An argument
was made that it might rank high because unanticipated adverse effects
are often discovered long after a chemical has been put in commerce.
Ultimately, the work group decided it could not develop a satisfactory
approach to analyzing the effects of new chemicals and ranking them as
a problem area.
Detailed consideration was given to the relative rankings among five
water-related problem areas: direct point source dischargers, indirect
point sources, non-point sources, discharges to estuaries, and discharges
to wetlands. Representatives of the Office of Marine and Estuarine Pro-
tection and the Office of Water Regulations and Standards were asked to
join the work group and present relevant data. The work group eventually
concluded that the most significant non-cancer risk in this area was
from consumption of contaminated fish and shellfish, with pathogens
being the primary contaminants of concern and metals and pesticides
of slightly lesser importance. POTWs (defined as constituting the
"indirect point source" problem area) and non-point sources are the
major contributors of these contaminants, with direct industrial dis-
chargers far less significant. The estuaries and the wetlands problem
areas were distinguished from each other on the basis of the much
greater amount of fish and shellfish taken from estuaries and marine
waters than from wetlands. Ultimately, all five of these water-related
problem areas were ranked on the basis of this group consensus rather
than through the scoring process.
The work group reviewed the areas ranked as low risk, and asked if anyone
had a rationale for any of them entailing substantial risks. No one did.
-------
A-35
Two final issues discussed by the work group during the ranking process were the
treatment of uncertainty and the guestion of how the percentage of the problem
captured by the selected substances should affect the ranking.
The discussion of uncertainty was prompted by the rankings of radiation
problems. Some work group members felt that a problem should be ranked lower
if the data available for ranking it are qualitatively weak. The highest .rankings
should be given only to problems that seem to entail a large amount of risk and
that we are confident about. Others argued that the non-cancer group had been
asked to give its best guess about the relative risks across problem areas, and
that a best guess should be simply that whatever we can say about a problem
area given existing information, whether that information is good or bad.
Ultimately the work group agreed that the quality of available data had two
aspects to it: bias and inprecision. To the extent that the workgroup felt some
data were biased and there was reason to believe that a problem was really larger
than the data indicated or really smaller than the data indicated, this judgment
should be reflected in the rankings. Qualitative corrections for biased estimates
were appropriate; in fact the work group had been making them frequently in the
process of assigning scores. But qualitative corrections should not be made for
imprecision when there is no suggestion of bias. The work group agreed to note
the problem areas where rankings were more and less confident, but the ranking
itself was not to be affected by the judgment about precision.
The work group also considered the guestion of how to deal with problem
areas where the ranking had been based on substances representing a small
proportion of the total problem. The work group spent some time developing
estimates of the proportion of the problem captured by the representative
substances; how should they use the estimates? The argument that was ultimately
persuasive was this. The problems that had been ranked as high risk had an
average total score about two points higher than the average total score for
those that had been ranked as medium risk. Similarly, medium risk problems had
an average total score about two points higher than low risk problems. Because
of the highly logarithmic nature of the point scoring system, a two point score
difference entailed more than a two order of magnitude difference in the data
underlying the scores. It thus seemed appropriate to move a problem up in the
ranking (in effect to give its score two more points) only if the entire problem
was about two or more orders of magnitude larger than the portion of the problem
covered by the selected substances. Four of these "tip of the iceberg" type
problem areas hazardous air pollutants, pesticide residues on food, worker
exposure, and consumer exposures ranked in either the high or medium categories,
depending on the particular mathematical ranking process chosen. The work
group decided to give all four a final ranking of high, combining judgments
that substances in these problem areas- ranked fairly high and that the substances
studied represented a small fraction of the total problems.
The final rankings of problem areas by the work group are in Tables 2-1
and 2-2 in Chapter 2.
------- |