-------
I
CD
in
UJ
3
THER CHARACTERISTICS OF METHOD |
o
i—
Q-
•5
O
1
^_
o
rr
Q.
a.
oc
o
5
— J
o
o
en
UJ
O
2
CC
0
C/3
Q
§
h—
UJ
S
If Hl<1, there is less likelihood on
noncancer effect
The HI is not a probability of a noncancer
effect and not a "bright line' of
acceptability
Assumes dose additivity, which is most
appropriate for compounds with similar
effects and mechanisms of action
Effecl-spedfic Hi's should be calculated by
a toxicologist or may understate risk
.
CD
1
ZS
.0
X
"*2
•o
1
X
3;
1 1
Q §" -5
?8 I
O -C X "O
o
i
QJ
•35 >p
C >
8|
x S.
•
2. ~~ ™ -o
f 1 1 f
2 « §•"
S. = s "S
§ §. 1 °
" §• f= g-
•^ C CO CO CO
O -55 CD cj O o ^
uj cu cu ro •— Q. 3
O S ^ " O T3 £
CC -3- 2 -2 iZ .£ 2 .2 -J co 0.
•
o
o" ^
LU O -^3*
*" a. -—
•*• 2. S-
O ^ c
_ ,._ CD
=t[ QJ w_
UJ 0) 8
illl
£ u c
~--_ ^ o
UJ 'J C
11 ca "o
x 1-3
CO
a.
J
O CO
"~ fl)
X -§
^, ^5
x 2
•o -o
^ i
i — ,— i
3 £
T3 O
X 0
• « •
C/I
ca
^
a.
"H
CU
cu
1 53
0 -^
**"* iU
= s»
-*- o
Ji
^ cO
O S
X
^^
ca
a.
^s
2
Output is a probability of (an) effect(s)
Requires sufficient epidemiologic or
toxicologic data
Scientific issues (such as spedes
differences, thresholds, pharmacokinetics,
mecharrisms of action) must be addressed
in selecting dose response model and
described as part of risk characterization
. .
3
Id c
-i „ 5 -2
« SO ^ OJ
* s ^^.
"in -a cu o
•= co 3= a.
gj — CD g
•— s
S s1 y g
<= J3 0 -S
3 2 §=5
O -S £ .E
•
1 - -5-2
"5 O (D (13 "^
° -^ S. 5 ^
111 11
<2 -§ co £ at
S" o> * "55 f; -o
2 -£ V 5 co !2
« S ¥ » 1 "i
S =5 .2 S ^ ~
Q .= "a -a 2 ,2
* <•
o t
-^ "o ^
"5 "3 "^
§- f-s.
f-20-2
O 3 CO "Jo
"S O 3 ^§
-a x"c» E
.> * *- c
1 § '1 --S
t5 S s. 1 -i
<>> c co •• _^ ^Q
S !s S 5 <"
•S co w ^ 2
O S:^ 2 O "5
. . .
o
i i
Is!
* £- «
a> r: -c
(n o a?
§*- Q, W
C X «D
8-i •'€
to § o '5
§ $ 1 1
O ca CL S
* " i
1
cn 53 -rr
Q. "aj
1 , Iff
•5 ~ a. c cu ^2
2=2^^3
o -g « .2 2 E
s 1 a s s i-
^= tl X =** ° T3 -.
^3 Q. OJ CO CD .3, ^
-^ 03 "o .- -2 .12 3
i. i S 2 ^ -^ «"
liIHl!
O CO OJ CO CO O_ "O
. 1
3 i
a
^
o
Ss
ea
c:
g
o
O
m
Q.
_OJ
to
o
O
JO
o
<
CO
5-28
-------
Q
O
M
CC
LU
co
cc
CC
LU
O
O
O
z
CC
o
cr
LU
H;
CC
O
Q
i—
O
UJ
CO
>-
cr
<
LU
CQ
<
LU
p2
LU fC
^— ^^
-£
Q oZ
P <
CO W
VJJ —
U- Q
O O
_j ^Z
UJ 1—
S 2
Q
g
}_
LU
S
i
O3
«
1
Q3
=3
0) <"
CO Q.
CO x
X3. OJ
•° 1 —
CC en ^3
o .2 9-
§ £ 03
Q O3 03
< 2 -2.
* *
03
_03
O
5 T^
• Individual with a 1
of risk assessmel
1
5
13
-•-'
O)
c:
'c
03
O3
O
CO
.
o
CJ
cr S
2 E
O3 _9;
5 °
§.-§
X .£
O3 CO
03 CO
^ O
O =i
O CC
|
03
_
O
~*
to IRIS to other sources of
information (see Chapter 3)
to generate exposure estimate
CO — 1 T3
to LLI 0
03 <; jz
0 O 3
*
^1
LU
13 O
e-f
3 .S
.£ S3
o "o .
•- €
o uj"
"D
O3
-2
~§
o
-*-
O3
C
'c
Q3
O
CO
TJ
Q
•*•"
u?
o
2,
2
55
8.8
LU 'E
^_ CD
O -C
<~~
o>-2
5 c
^^ '"
«£? V}
s
c
o
-a
1
u
c
o
CO
03
1
"ca
O3
JC
o
11
< .E
*
• Toxicologist
T3
Oj
3
O3
Q
Q
._. ^^
, cr
(U W3 O
= - o
— 'o
03 Q- 0
5 03 "~
Tf, o .«
"O T3
J3J 03 <=
§ |-s
5 •= o S
• Individual with a 1
of risk assessmei
• Toxicologist if cal
effect-specific ha,
TJ
3
O3
O
O3
"c
O3
03
^j
CO
_o
CJ
CO
2
O-
x
03
"3
~
-a 33
CO 3
CO —
x 2
±
o
E
1
2
« 1-
||
S> °-
ssf
o i E
S »- °
OJ **~
co E .E
cr R "c/5
— " as
o o 03
CO T3 3
(O O CO
03 JT o
lilt
• Toxicologist
-a
-03
c5
"a3
"^
,2
"«
•a
o
2
X
03
-a
CO
^j
i
_£Z
'to
g.
^
2
.c
•^
to epidemiologic or toxicologic
sponse model (computer)
or model to estimate exposure
to co S -o
(O 03 ' O
•• § 1 § i
< to Q 2
• Toxicologist
•o
.2>
'5
^3
t~—
. .
C
03
CO
03
CO
C/5
**-
1
UJ
OJ
OC
^
Q
"§,
^p
sponse model
or model to estimate exposure
) elicit expert characterization of
nties and present them as
ities
*r o "^1_ t ~S
(11 r— ^* ^ CO
% & -a g ^
o ^ = g o
Q 2 co =i o.
"O
03 Si £
CO 3 ra
oo c £•
Q. Q. — 0
co x -a <"
c 03 03 03 -S
ill It
§_ -S ^- — s_ "g
15 ,2 to co c o ^"
53 8 o3 a3 S :« 1
JT 'i^ CO C/3 Jrr O ^T
O3 55 (O CO O3 O3 iri
co S co ca Q. -o S
-a
O3
.2
'S
a :
o
CO
O
a.
Q.
o
"ca
c
0
10
"o
O3
Q ;
M
c
I
o
5-29
-------
relatively quick and straight-forward and does not require a toxicologist. The RfCs and RfDs,
along with information on their uncertainty and level of confidence, can be found in IRIS.
For exposure to mixtures of noncarcinogenic compounds, a hazard index approach is most
commonly used. This involves dividing the exposure concentration by the acceptable level of
exposure (e.g., RfC/RfD). To characterize the noncancer risk for a mixture, the hazard index can
be computed from the ratio of exposure to RfC/RfD for each compound. Because this approach
is based on the assumption of dose additivity, it is most appropriate for compounds with the same
effect or endpoint and similar mechanisms of action.
A screening approach might calculate one hazard index, not differentiated by the type of
noncancer effect caused; however, this may under- of overestimate risk. Any available evidence
as to whether effects of two or more compounds are additive, synergistic, or antagonistic, should
X-
be considered. Instead of calculating a hazard index for a mixture, effect-specific hazard indices
can be derived, but expert judgment of a toxicologist is required to avoid underestimating the
risks. Mixture risk assessment is more time consuming than single-compound risk assessment
because information must be gathered on a number of compounds.
Two recently developed approaches to noncancer risk assessment, dose-response modeling
and the decision analysis approach can also be used. The advantage of these approaches is that,
where they can be used appropriately, the risk characterization can provide a probability of
noncancer health effects occurring in the study population. However, these approaches usually
require detailed experimental data, expert judgement, an understanding of the exposure/RfD ratio
or hazard index approach and of mechanisms of action and pharmacokinetics, as well as the use
of computer modeling.
If no RfC or RfD is available in IRIS, there are several options for characterizing
noncancer risk. One is to perform a qualitative rather than quantitative risk characterization.
This involves presenting the evidence gathered during hazard identification on the potential
noncancer effects of the compound without attempting to determine the likelihood that the human
exposures modeied in the study will cause an adverse health effect. Another option ;s to
determine a NOAEL from the available epidemioiogic or toxicologic data and calculate a margin
of exposure (MOE). In this case, the selection of the critical study, critical effect, and NOAEL
would need to be justified. A third option is to use the available data and follow the procedures
for developing a new RfC or RfD (see Section 3 and EPA/ECAO, 1989). The risk
5-30
-------
characterization would ultimately compare exposure to this user-derived value. This would not
be an officially-approved RfC or RfD unless it went through the EPA peer review process. A
toxicologist would need to be consulted during the dose-response and risk characterization steps.
When exposure is compared to a user-derived NOAEL or user-derived reference dose, the risk
characterization step is somewhat more complex than when an EPA-approved RfC or RfD is
used, because a more detailed discussion of the derivation of the reference dose may need to be
presented.
5.5 CHARACTERIZATION OF UNCERTAINTY
Recent EPA guidance calls for a full characterization of risk, not just the single point
estimate which has become synonymous with risk characterization. Critical to full
characterization of risk is a frank and open discussion of the uncertainty in the overall assessment
and in each of its components. Numerical estimates should always be accompanied by
descriptive information carefully selected to ensure an objective and balanced characterization
of risk (EPA, 1992).
Uncertainty can be introduced into a health risk assessment .at every step in the process.
It occurs because risk assessment is a complex process, requiring the integration of:
• the fate and transport of pollutants in a variable environment by processes that are
often poorly understood or too complex to quantify accurately;
• the potential for adverse health effects in humans as extrapolated from animal
bioassays; and
• the probability of adverse effects in a human population that is highly variable
genetically, in age, in activity level, and in life style.
Even using the most accurate data with the most sophisticated models, uncertainty is inherent in
the process.
Finkel (1990) classified all uncertainty into four types which are summarized in Table 5-8.
The first two, parameter uncertainty and model uncertainty, are generally recognized by risk
assessors as major sources of uncertainty.
Parameter uncertainty occurs when variables cannot be measured precisely either because
of equipment limitations or because the quantity being measured varies spatially or temporally.
Random, or sample errors, are a common source of parameter uncertainty that is especially
5-31
-------
Ul
2
en
CO
LLJ
CO
CO
CO
cc
z
>
cc
LLI
O
CO
111
o
en
CD
<
Comments/Examples
C
r:
03
O
5
"o
fl3
3
o
CO
.0
'o
03
Q.
CO
03
Q.
h-
"m
0)
03
CD
|
03
• include limitations of equipment, methodology, and human
• some processes impossible to measure exactly
£
0
03
urement
C/)
O3
O
2
^^
£Z
'ra
•c
CD
Q
^
03
03
E
2
03
Q.
sampling errors
• can be minimized increasing sample size
<73
^_
o
03
P
O
-o
c
eg
CC
• nonrandom errors
• result of inherent flaw in data gathering processes
• minimize by exernal peer review
o
03
O
'5
03
~y3
•>,
co
CO
c
e.g., use of animal bioassays to determine effect on huma
.2
"^t
CO
03
"to
O)
O
^
~^
co
2j*
c
03
o
—
"3
-a
o
2
co
03
N
• may be a result of model simplification or failure to recogn
important variable
CO
0)
.a
to
to
•a
03
T3
73
X
LU
• for example, failure to recognize importance of episodic
meteorological events
CO
g
.±±
o
u
"5
o
c
<
CO
• for example, choice of dose-response model for carcinoge
E
,0
03
0
"o
03
^
0
o
c
,
:^
CO
CO
o
O3
cn
o>
JiT
C
LL
T3
03
B.
to
5-32
-------
critical for small sample sizes. More difficult to recognize are nonrandom or systematic errors
that result from bias in sampling, experimental design, or choice of assumptions. Since the risk
assessor is unlikely to detect his or her own bias, it is wise to use external peer reviewers on any
risk assessment protocol.
Model uncertainty is associated with a variety of models used in all phases of a risk
assessment. These include the animal models used as surrogates for testing human
carcinogenicity as well as the computer models used to predict the fate and transport of chemicals
in the environment. The use of rodents as surrogates for humans introduces uncertainty into the
risk factor since different species do not respond to toxins in exactly the same way. Computer
models are simplifications of reality and some variables are excluded. The risk assessor needs
to consider the importance of excluded variables on a case-by-case basis, because a given
variable may be important in some instances and not in others. A similar problem can occur
when a model that is applicable under average conditions is used for a case where conditions are
abnormal. Large bodies of water, for example, can cause meteorological conditions that are not
adequately modeled by air dispersion models such as ISC. Finally, choosing the correct model
form is often difficult because conflicting theories seem to explain a phenomenon equally well.
Cothern (1988) gives a good example of this in contrasting dose-response models; four equally
reasonable models give risk estimates that vary by four orders of magnitude.
The third type, decision-rule uncertainty, is probably of more concern to risk managers.
This type of uncertainty arises out of the need to balance different social concerns when
determining an acceptabie level of risk, for example. The risk assessor needs to understand the
rationale for setting certain acceptable levels because the rational can affect the choice of model,
data, or assumptions. Finkel (1990) provides a complete discussion of decision-rule uncertainty.
Variability, the fourth type of uncertainty, is often used interchangeably with the term
"uncertainty," but this is not strictly correct. The variability of a characteristic may be known
with absolute certainty. For example, the age distribution of a population may be known and
represented by the mean age and its standard deviation. The fact that ages do vary introduces
uncertainty into characterizing risk for that population. On the other hand, if the age distribution
may not be known, then the variability associated with the population's age is in itself an
uncertainty.
5-33
-------
The degree to which all types of uncertainty need to be quantified and the amount of
uncertainty that is acceptable varies. For a screening level analysis, a high degree of uncertainty
is often acceptable, provided that conservative assumptions are used to bias potential error toward
protecting human health. Similarly, a region-wide or nationwide study will be more uncertain
than a site-specific one. In general, the more detailed or accurate the risk characterization, the
more carefully uncertainty needs to be considered.
Figure 5-1 depicts the factors contributing to uncertainty and the orders of magnitude of
uncertainty associated with each factor. The bar by each uncertainty factor shows the likely
direction of uncertainty. For example under Unit Risk Estimate, if synergisms are present, the
maximum individual risk could be an order of magnitude higher than the EPA estimate.
However, if antagonistic interactions occur, the maximum individual risk could be an order of
magnitude lower than the EPA estimate.
Often, the sources of uncertainty in a risk assessment can be determined, but they cannot
be quantified. This can occur when a factor is known or expected to be variable, but no data are
available (e.g., the amount of time people at a specific site spend out of doors). In this case,
sometimes default data are available that can be useful for estimating a possible range of values.
Uncertainty often arises out of a complete lack of data. A process may be so poorly understood
that the uncertainty cannot be quantified with any confidence. When uncertainty can only be
presented qualitatively, the possible direction and orders of magnitude of the potential error
should be considered.
Knowledge of experimental or measurement errors can also be used to introduce a aegree
of quantitative information into a qualitative presentation of uncertainty. For example, standard
laboratory procedures or field sampling methods may have a known error level that can be used
to quantify uncertainty. Cothern (1988) describes how the well-known experimental and
laboratory procedures used in animal bioassays can be used in a practical way to estimate order-
of-magmtude uncertainties for toxicity studies.
In many cases, the uncertainty associated with particular parameter values or for the
estimated risks can be expressed quantitatively. Finkel (1990) identified a six-step process to
producing a quantitative uncertainty estimate. Initially, the measure of risk should be defined
(e.g., deaths, life-years lost, maximum individual risk, population above an "unacceptable" level
of risk). More than one measure of risk may result from a particular risk assessment: however.
5-34
-------
Probable Impact of Major Assumptions on
EPA's Risk Assessment—MIR-Model plant
-2
Orders of Magnitude
EPA
— 1 Estimate +1
L
Unit Risk
Estimate
HHAG/CAG
API/CMA
GARB
Sensi Sub-Pop
Synergism
Antagonism
Other D/R Models
Emissions &
Pource
arameters
Fugitive
Stock
Eff Release Ht
Pfont Life
Disp Model
Complex Terrain
Urban Release
Dev wl bldg wake
Bevno bldg woke
Exposure
Assessment
HEM Meteorology
Plant Property
Exp at Pop Cant
Indoor « Outdoor
Migration
Human Activity
Urban/Rural Met
Lot/Long Urban
Lot/Long Rural
Area Emis—Point
Figure 5-1. Example of a Semi-Quantitative Presentation of Uncertainty
5-35
-------
the uncertainty should be quantified for each individually. Second, one or more "risk equations"
should be specified which present the mathematical relationships that express the risk measure
in terms of its components. This step is used to identify the important parameters in the risk
estimation process. Third, for each parameter or equation component, an uncertainty distribution
is generated. These uncertainty distributions may be generated by the use of analogy, statistical
inference techniques, or elicitation of expert opinion, or some combination of these. These will
be described in more detail below. Fourth, the individual distributions should be combined into
a composite uncertainty distribution. To carry this out, Monte Carlo simulation is frequently
used; this method is discussed in greater detail below. Fifth, the uncertainty distributions should
be "recalibrated". Inferential analysis could be used to "tighten" or "broaden" particular
distributions to possibly account for dependencies among the variables and/or to truncate the
distributions to exclude extreme values. Finally, the output should be summarized in a manner
which is clear and highlights the important risk management implications. Specific aspects
should be addressed including: the implication of supplanting a point estimate produced without
considering uncertainty, the balance of the costs of under- or over-estimating risks, unresolved
scientific controversies, and implications for research.
When a detailed quantitative treatment of uncertainty is required, statistical methods are
employed. A statistician should be consulted, as many of the methods used require expert
knowledge. Two approaches to a statistical treatment of uncertainty with regards to parameter
values are described here and can be applied to any particular step in the risk assessment process.
The first is simply to express ail variables for which uncertainty is a major concern using an
appropriate statistic (see Table 5-9). For example, if a value used is from a sample (e.g., hourly
emissions from a stack), both the mean and standard deviation should be presented. If the
sample size is very small, it may be appropriate to give the range of sample values and use a
midpoint in the model; or, both the smallest and largest measured value could be used to get two
estimates that bound the expected true value. The appropriate statistic co use depends on the
amount of data available and the degree of detail required. The propagation of uncertainties can
be done using analytical or numerical methods. A common analytical method is first-order
analysis; it is appropriate when only a few parameters are of interest, their distributions are
known, and all relationships are linear.
5-36
-------
TABLE 5-9. SOME STATISTICS USEFUL FOR QUANTIFYING UNCERTAINTY'
Statistic
Sample
Variance
Sample
Standard
Deviation
Coefficient of
Variation
Probability
Density
Function (pdf)
Cumulative
Distribution
Function (cdf)
Expression
8«.rJ&ta£
^ (n-1)
s= yirf
N^(n-1)
rv - s ' 100
y
-(x-^1
1 2o2
fM et
rw • . e
v/2^2
F(x) = P(X < x)
Definition/Derivation
• mean square error
• a measure of the spread of values of a
sample
• square root of the variance
• also a measure of spread but has the
same units as the mean
• data are often summarized ± one standard
deviation
• the standard deviation expressed as a
percentage of the mean
• useful for comparing samples or
populations with different means,
particularly if the means are different
orders of magnitude
• specifies likelihood of occurrence of a
value within the distribution defined by the
specified mean (u.) and variance (cf)
• the "bell curve" for normal data
• specifies the likelihood of a result greater
than any value in the range of a
distribution j
'Formulae shown are for normally distributed data.
5-37
-------
A second approach is to use the probability distributions of major variables to propagate
parameter value uncertainties through the risk analysis. A probability distribution of expected
values is developed for each parameter value. These probability distributions are typically
expressed as either probability density functions (PDF) or as cumulative density functions (CDF).
The PDF presents the relative probability for discrete parameter values while the CDF presents
the cumulative probability that a value is less than or equal to a specific value. These probability
distributions may be developed using statistical methods or through the use of judgmental
probability methods.
The judgmental probability approach requires that experts use subjective judgments to
develop quantitative estimates (encoded probability) of uncertainty for a particular parameter
value. This approach may be applied to parameter values for either the exposure assessment or
dose-response assessment of risk. This method was described in some detail in Section 3.4.3.4.
As described in that section, a main distinguishing characteristic of the judgmental probability
approach is the emphasis on explicitly characterizing and representing uncertainty using
probability as the language to convey the degree of uncertainty. A variety of quantitative
schemes are possible. Experts may be asked to assign order-of-magnitude error bounds about
a data point or model estimate; or, they may be asked to describe the probability distribution of
a variable. Experts are selected to represent the range of credible scientific opinion and,
therefore, to implicitly represent the major uncertainties. As a result, experts may diverge,
sometimes widely, m their estimates. Disagreement among experts needs to be recognized by
the risk assessor in any discussion of uncertainty.
The propagation of uncertainties is accomplished by developing a composite uncertainty
distribution by combining the individual distributions. Numerical methods are often employed
for this phase with Monte Carlo simulations gaining wide acceptance for this purpose. In Monte
Cario simulations, a computer program is used to repeatedly solve the model equations to
calculate a distribution of exposure (or risk) values. Each time the equations are calculated,
values are randomly sampled from the specified distributions. The end result is a distribution
of exposure for risk). These can again be expressed as PDFs, or more appropriately as CDFs.
The distribution allows the risk assessor to choose the value corresponding to the appropriate
percentile in the overall distribution. For example, an exposure level or risk level can be selected
5-38
-------
which corresponds to the 95th percentile of the overall risk distribution rather than relying on a
point estimate of risk based on the 95th percentile values for each parameter.
Numerical methods require the use of a computer, and until recently, were generally too
slow and too difficult to program to be widely used. However, the use of Latin hypercube
sampling and Monte Carlo simulation has increased the speed of these models considerably.
Also, several computer packages are available that minimize (or eliminate) the programming
required (Salmento et al., 1989; Salhotra et al., 1988).
A complete discussion of these statistical techniques is beyond the scope of this document.
For more general information and references to more detailed discussions, refer to Section 8.4
of the Risk Assessment Guidance for Superfund (U.S. EPA/OSW, 1989). Various uncertainty
modeling techniques are compared in Iman and Helton (1988). Monte Carlo simulations have
been used in risk assessments related to exposures of cartion monoxide and ozone (Johnson et
al., 1992a and 1992b), and used to identify greatest sources of model uncertainty in
photochemical models (Derwent and Hov, 1988) and in food chain analysis (McKone and Ryan,
1989). The encoded probability approach has been used to estimate the risks and uncertainties
from exposure to ozone (Whitfield et al., 1993) and lead (Whitfield and Wallsten, 1989).
5-39
-------
5.6 REFERENCES
U. S. Environmental Protection Agency, Environmental Criteria and Assessment Office.
November 1988. Technical Support Document on Risk Assessment for Chemical Mixtures.
Final Draft, ECAO-CIN-572.
U. S. Environmental Protection Agency, Environmental Criteria and Assessment Office. April
1989. Interim Methods for Development of Inhalation Reference Doses. EPA 600 8/88-066F.
Cincinnati, OH.
U. S. Environmental Protection Agency, National Air Toxics Information Clearing House. June
1987. Qualitative and Quantitative Carcinogenic Risk Assessment. Research Triangle Park, NC.
U. S. Environmental Protection Agency, Office of Air Quality Planning and Standards. March
1990. HEM-II Users Guide. Draft. Research Triangle Park, NC.
U. S. Environmental Protection Agency, Office of Solid W^ste and Emergency Response. July
1989. Risk Assessment Guidance for Superfund, Human Health Evaluation Manual Part A.
Interim Final. Chapters 7 and 8. Washington, DC. EPA-540/1-89-002.
U. S. Environmental Protection Agency. Office of Health and Environmental Assessment,
Cincinnati, Ohio. Integrated Risk Information System.
U. S. Environmental Protection Agency, Environmental Criteria and Assessment Office. February
1991. General Quantitative Risk Assessment Guidelines for Noncancer Health Effects. Second
External Review Draft, ECAO-CIN-538. Cincinnati, Ohio.
U. S. Environmental Protection Agency. Guidelines for Developmental Toxicity Risk
Assessment. 1991, 56FR 63798 - 63826.
U. S. Environmental Protection Agency. Office of the Administrator, Guidance on Risk
Characterization for Risk Managers and Risk Assessors, 1992, memorandum from F. Henry
Habicht, II, Deputy Administrator.
Whitfield, R.G., H.M. Richmond, S.R. Hayes, A.S. Rosenbaum, T.S. Wallsten, R.L. Winkler.
M.L.G. Absil, and P. Narducci. (1993) Health Risk Assessment of Ozone, in Tropospheric
Ozone: Human Health ana Agricultural Impacts. David J. McKee (ed.) Lewis Publishers, Boca
Raton, Florida.
Cothern, C. R., W. A. Coniglio, and W. L. Marcus. 1984. Uncertainty in Population Risk
Estimates for Environmental Contaminants, pp. 265-286 in Covello, V. T., L. B. Lave, A.
Moghissi, and V. R. R. Uppuluri (eds). Uncertainty in Risk Assessment, Risk Management, and
Decision Making. Plenum Press, New York. 535 pages.
Derwent, R. and O. Hov. 1988. Application of Sensitivity and Uncertainty Analysis Techniques
to a Photochemical Ozone Model. J. Geophys. Res., 93: 5185-5199.
5-40
-------
Finkel, A.M. 1990. Confronting Uncertainty in Risk Management. A Guide for Decision
Makers. Center for Risk Management, Resources for the Future, Washington, D. C.
General Sciences Corporation. December 1988. PCGEMS Users Guide Release 1.0. Draft.
Prepared for U. S. Environmental Protection Agency, Office of Pesticides and Toxic Substances,
Washington, D.C. Contract No. 68-02-4281.
Iman, R. L. and J. C. Helton. 1988. An Investigation of Uncertainty and Sensitivity Analysis
Techniques for Computer Models. Risk Analysis 8: 71-90.
Johnson, T., J. Capel, R. Paul, and L. Wijnberg, 1992a. Estimation of Carbon Monoxide
Exposures and Associated Carboxyhemoglobin Levels in Denver Residnets Using A Probabilistic
Version of NEM, Prepared by International Technology Air Qulaity Services for US
Environmental Protection Agency, Office of Air Quality Planning and Standards, Research
Triangle Park, NC, Contract No. 68-DO-0062.
Johnson, T., J. Capel, E. Olaguer, and L. Wijnberg, 1992b. Estimation of Ozone Exposures
Experienced by Residents of the ROMNET Domain Using A Probabilistic Version of NEM,
Prepared by International Technology Air Qulaity Services for US Environmental Protection
Agency, Office of Air Quality Planning and Standards, Research Triangle Park, NC, Contract No.
68-DO-0062.
McKone, T. E. and P. B. Ryan. 1989. Human Exposures to Chemicals through Food Chains:
An Uncertainty Analysis. Environ. Sci. Technol., 23: 1154-1163.
Morgan, M. G., M. Herion, S. C. Morris, and D. A. L. Amaral. 1985. Uncertainty in Risk
Assessment. Environ. Sci. Technol., 19: 662-667.
Salhotra. A. M., R.Schanz, and P. Mineart. 1988. A Monte Carlo Simulation Shell for
Uncertainty Analysis. Prepared for U. S. EPA, Environmental Research Laboratory, Athens, GA,
under contract no. 68-03-6304.
Salmento, J. S., E. S. Rubin, and A. M. Finkei. 1989. A Review of @RISK. Risk Analysis.
9: 244-257.
U, S. Environmental Protection Agency. i986a. Guidelines for Carcinogen Risk Assessment.
51 FR 33998. September 24, 1986.
U. S. Environmental Protection Agency. I986b. Guidelines for the Health Risk Assessment of
Chemical Mixtures. 51 FR 34014 - 34025. September 24, 1986.
U. S. Environmental Protection Agency, Environmental Criteria and Assessment Office.
November 1988. Technical Support Document on Risk Assessment for Chemical Mixtures.
Final Draft, ECAO-CIN-572.
5-41
-------
U. S. Environmental Protection Agency, Environmental Criteria and Assessment Office. April
1990. Interim Methods for Development of Inhalation Reference Doses. EPA 600 8/90-066A.
Cincinnati, OH., August 1990
U. S. Environmental Protection Agency, National Air Toxics Information Clearing House. June
1987. Qualitative and Quantitative Carcinogenic Risk Assessment. Research Triangle Park, NC.
U. S. Environmental Protection Agency, Office of Air Quality Planning and Standards. March
1990. HEM-II Users Guide. Draft. Research Triangle Park, NC.
U. S. Environmental Protection Agency, Office of Solid Waste and Emergency Response. July
1989. Risk Assessment Guidance for Superfund, Human Health Evaluation Manual Part A.
Interim Final. Chapters 7 and 8. Washington, DC. EPA-540/1-89-002.
U. S. Environmental Protection Agency. Office of Health and Environmental Assessment,
Cincinnati, Ohio. Integrated Risk Information System.
•\»
U. S. Environmental Protection Agency, Environmental Criteria and Assessment Office. February
1991. General Quantitative Risk Assessment Guidelines for Noncancer Health Effects. Second
External Review Draft, ECAO-CIN-538. Cincinnati, Ohio.
U. S. Environmental Protection Agency. Guidelines for Developmental Toxicity Risk
Assessment, 1991, 56FR 63798 - 63826.
U. S. Environmental Protection Agency. Office of the Administrator, Guidance on Risk
Characterization for Risk Managers and Risk Assessors, 1992, memorandum from F. Henry
Habicht, II, Deputy dministrator.
Whitfieid. R.G. and T.S. Wallsten, 1989. A Risk Assessment for Selected Lead-Induced Health
Effects: An Example of a General Methodology, Risk Analysis. 9(2): 197-208
Whitfieid, R.G., H.M. Richmond, S.R. Hayes, A.S. Rosenbaum, T.S. Wailsten, R.L. Winkler,
M.L.G. Absil. and P. Narducci, 1993. Health Risk Assessment of Ozone, in Troposphenc Ozone:
Human Health and Agricultural Impacts, D.J. McKee (ed.), Lewis Publishers, Boca Raton, FL.
5-42
-------
6.0 EMERGING ISSUES
6.1 ECOLOGICAL RISK ASSESSMENT
The EPA's mandate requires that it protect human health and the environment.
Although a broad interpretation of that mandate would require protection of nonhuman
populations and ecosystems, legal requirements of legislation and funding realities have
resulted in the emphasis being placed on human health. Recently, several factors have led to
a renewed interest on the part of EPA in ecological risk assessment. In the past, it was
assumed that the environmental levels that were protective of human health would be
sufficient to protect the environment. This has not proved to be true. EPA's Science
Advisory Board (SAB) recognized the importance of ecological systems in their report
Reducing Risk: Setting Priorities and Strategies for Environmental Protection. The SAB
•v-
stated that there are strong linkages between human health and the health of wetlands, forests,
oceans, and estuaries. Most human activities that pose significant ecological risks ~ for
example, the effects of agricultural activities on wetlands pose direct or indirect human health
risks as well. Healthy ecosystems are a prerequisite to health humans and prosperous
economies. Ecosystems, and the plant and animal populations that compose them, are
important as food resources, as bioindicators of environmental degradation, and as an
influence on climate.
Guidelines on ecological risk assessment were not included in the original set of EPA
risk guidelines pubiisned in 1986. Subsequently, the EPA has developed guidelines m other
risk assessment areas (e.g.. developmental toxicity). To date, no EPA guidelines have been
developed for ecological risk assessment. However, individual EPA programs have generated
program-specific guidance for ecological effects and the EPA is in the process of developing
Agency-wide guidelines as well. iMany programs at EPA have always had an ecological risk
assessment component to some degree; (Norten et ai. 1988; Bascietto et ai., 1990). For
exampie. EPA programs in pesticides and water have traditionally been concerned with
impacts on aquatic nd terrestrial organisms. Much of their work has been used, in part, to
develop much of the methodology described below. Examples of program-specific guidance
related to ecological risk assessment include: Interim Report on Data and Methods for
Assessment of 2, 3, 7, 8-Tetrachlorodibenzo-p-dioxin Risks to Aquatic Life and Associated
-------
Wildlife (EPA, 1993a) Proposed Water Quality Guidance for the Great Lake System.(EPA,
1993b) Wildlife Criteria Portions of the Proposed Water Quality Guidance for the Great
Lakes System. (EPA, 1993c) Derivation of Proposed Human Health and Wildlife
Bioaccumulation Factors for the Great Lakes Initiative (EPA, 1993d), Great Lakes Water
Quality Initiative Criteria Documents for the Protection of Wildlife (Proposed) -- DDT,
Mercury; 2,3,7,8-TCDD, PCBs (EPA, 1993e), Bioaccumulation of Selected Pollutants in
Fish (EPA, 1990), Technical Support Document for the Determination of the Need to
Regulate Pulp and Paper Mill Sludge Landfills and Surface Impoundments, (EPA, 1991)
Preliminary work on Agency guidelines for ecological effects began in 1988. As part
of this work, EPA studied existing assessments and identified issues to help develop a basis
for articulating guiding principles for the assessment of ecological risks (EPA, 1991). The
V-
EPA's Science Advisory Board urged the EPA to expand the consideration of ecological risk
issues to include both chemical and nonchemical stressors under programs and laws
administered by the EPA (EPA, 1990). As a result, a new program was initiated to develop
guidelines for ecological risk assessment. The EPA Risk Assessment Forum (RAF),
responsible for EPA's Agency-wide risk assessment guidance initiated three ecological risk
guidance projects: (1) compilation of case studies to illustrate "state-of-the-practice" in
ecological assessments, (2) preparation of long-term plan for developing specific ecological
risk assessment guidelines, and (3) development of a framework to describe the basic
principles for ecological risk assessment and provide a flexible structure conducting and
evaluating ecological risk assessments. Several documents have been published as a result of
these efforts (EPA, 1992 a,b,c) although the case studies have yet to be published.
Section 6.1.1 will introduce the framework for ecological risk assessment. The
methodologies developed by these programs are, in part, the basis for the methodology
discussion in Section 6.1.2. Major issues facing the future development of ecological risk
assessment are also discussed in Section 6.1.3. The field of ecological risk assessment is
rapidly evolving. Therefore, the discussion here will be limited to the activities and products
of the RAF to provide a general overview. Additional references will be mentioned which
can be used to obtain additional details.
6-2
-------
6.1.1 FRAMEWORK FOR ECOLOGICAL RISK ASSESSMENT (EPA, 1992a)
As defined by the framework (EPA, 1992a) ecological risk assessment is a process
that evaluates the likelihood that adverse ecological effects may occur or are occurring as a
result of exposure to one or more stressors. A risk does not exist unless (1) the stressor has
the inherent ability to cause one or more adverse effects and (2) the stressor impacts an
ecological component1 (i.e., organisms, populations, communities, or ecosystems) long
enough and at sufficient intensity to elicit an adverse effect. Ecological risk assessment may
evaluate one or many stressors and/or ecological components.
Ecological risk may be expressed in a variety of ways. While some ecological risk
assessments may provide true probabilistic estimates of both adverse effects and exposure
elements, others may be deterministic or even qualitative in nature. In these cases, the
likelihood of adverse effects is expressed through a semiqiiantitative or qualitative comparison
of effects and exposure.
The distinctive nature of the ecological risk framework results from three differences
in emphasis relative to previous risk assessment approaches. First, ecological risk assessment
may consider effects beyond those on individuals of a single species and may examine
population, community, or ecosystem impacts. Second, there is no single set of assessment
endpoints (i.e., environmental elements to be protected) that can be generally applied in all
ecological risk assessments. Instead the assessment endpoints are selected from a large
number of possibilities based on both scientific and policy considerations. Finally, a
comprehensive approach may go beyond the traditional emphasis on chemical effects to
consider the effects of nonchemical stressors. The ecological risk assessment framework as
proposed by the RAF is shown in Figure 6-1. The framework consists of three major
elements, problem formulation, analysis and risk characterization.
6.1.1.1 Problem Formulation. The first phase is problem formulation, a preliminary
characienzation of exposure and effects. In this phase, a wide range of data and policy issues
must be considered including the examination of scientific data and data needs, policy and
lRecent trends have indicated support and consensus on the use of the term
ecological receptor to replace the term ecological component as proposed by the
RAF.
6-3
-------
Discussion
Between the
Risk Assessor
and
Risk Manager
(Planning)
Ecological Risk Assessment
PROBLEfv
1 FORMULATION
A
N
A Characterization Characterization
L of o'
Y Exposure Ecological
' Effects
1
S,,._
^
RISK CHA
\7 \7
RACTERIZATION
4
IP
o
&
|
55°
=v
on: Verification and Monitoring
4
Discussion Between the
Risk Assessor and Risk Manager
(results)
4
ir
Risk Management ^
Figure 6-1. FRAMEWORK FOR ECOLOGICAL RISK ASSESSMENT
Source: EPA. 1992
6-4
-------
regulatory issues, and site-specific factors to define the feasibility, scope, and objectives of
the ecological risk assessment. The level of detail and the information needed to complete
the assessment are also determined. Ecological risk assessments tend to be of higher
complexity than other traditional risk assessment where data needs are clearly defined. The
high degree of complexity and variability in ecological risk assessments are due to site-
specific differences in environmental conditions, species, nutrient cycling, and community
structure. As a result, ecological component response may vary dramatically among sites.
This systematic planning phase is proposed because ecological risk assessments often address
the risks of stressors to many species as well as risks to communities and ecosystems. In
addition, there may be many ways a stressor can elicit adverse effects (e.g., direct effects on
mortality and growth, and indirect effects such as decreased food supply). Therefore, this
>-
planning phase is critical in defining the approach and to consider all potential issues, inputs,
and effects.
The major elements of the problem formulation phase are characterizing the stressors,
defining the ecosystem potentially at risk, defining the range of ecological effects and
selection of the endpoints to be analyzed. Stressors should be characterized in terms of type
(chemical or nonchemical), intensity (concentration or magnitude), duration, frequency (single
event, episodic, or continuous), timing (occurrence relative to biological cycles), and scale
(spatial heterogeneity and extent). In this context, problem formulation is used in place of the
analogous hazard assessment in human health risk assessments to define the range of potential
ecological effects wnich may occur as a result of the stressor. A wide range of endpoints
may be considered, including the death of a population of one species or a disruption of
nutrient cycling in an ecosystem. The appropriate end points will vary depending on the site
and the chemicals of interest. A useful end point, however, should have the following
characteristics:
i i) have ecological relevance.
(2) be of importance to society, and
(3) have demonstrated susceptibility to the stressor
Endpoints can be divided into two basis categories, measurement endpoints and
assessment endpoints. Ecological effects are difficult to measure directly and would require
6-5
-------
extensive field studies. As a result, measurement endpoints are selected because they are
easily defined and quantified. Ideally these measurement endpoints would be related to some
environmental impact, which is represented by an assessment endpoint. The relationship
between measurement endpoints and an assessment endpoint (ecological impact) should be
evaluated during the problem formulation phase and the most efficient and ecologically
plausible endpoints selected. The selection of endpoints would likely vary from site to site
and depend on both the type of stressor under consideration and the environmental of
concern. Some types of endpoints and examples are given in Table 6.1 for three levels of
biological organization the population, community, and ecosystem levels. The lower levels--
cells or individual organisms—are generally not relevant to an ecological risk assessment,
though few individuals of an endangered species may present legitimate concern.
6.1.1.2 Analysis. The second phase of the framework is termed analysis and consists
of two activities, characterization of exposure and characterization of ecological effects. The
purpose of exposure characterization is to predict or measure the spatial and temporal
distribution of a stressor and its co-occurrence with the ecological components of concern.
The purpose of ecological effect characterization is to identify or quantify the adverse effects
elicited by a stressor and, to the extent possible, evaluate cause-and-effect relationships.
Exposure characterization also has two major components, stressor characterization and
ecosystem characterization. Stressor characterization involves determining the stressor's
distribution or pattern of change. Many techniques can be applied to stressor characterization
including fate and transport modeling similar to that used in human risk assessments.
Ecosystem characterization defines the spatial and temporal distributions of the ecological
component, and the ecosystem attributes that influence the distribution and nature of the
stressor. Characteristics of the ecosystem can greatly modify the ultimate nature and
distribution of a particular stressor through biotransformation by microbial communities or
through other environmental fate processes, such as photolysis, hydrolysis, and sorption.
Characteristics of ecological components may influence their exposure to and response due to
a particular stressor. These characteristics should be defined and may include habitat needs.
food preferences, reproductive cycles, and seasonal activities.
6-6
-------
TABLE 6-1. SOME POTENTIAL ENDPOINTS OF AN ECOLOGICAL RISK
ASSESSMENT
Level of
Organization
Population
Community
Ecosystem
Characteristic
Density/Natality, Mortality
B amboos/Productivity
Genetic Composition/Evaluation
Spatial Pattern
Diversity
Physical Structure
Trophic Structure/Food Webs
Bamboos/Energy Flow
Nutrient Pools/Biogeochemical
Cycling
Potential Adverse Effect
Decrease
Increase (if undesirable species)
Decline in yields of a crop
Selection of resistant species (e.g.,
insects resistant to pesticides)
Change in range or dispersion pattern
Loss of species
Decrease in the number of animals
Decrease or change in complexity
resulting in loss of habitat
Loss of top predators
Decrease in vegetation
Decrease in bamboos per unit area
Disruption of nutrient cycles
6-7
-------
The characterization of ecological effects describes the relationship between the
stressor and the assessment (and/or measurement endpoints) identified during the problem
formulation phase. The characterization begins with the evaluation of effects data that are
relevant to the stressor, ranging from mortality and reproductive impairment in individuals
and populations to disruptions in community and ecosystem function such as primary
productivity. During ecological response analysis, the relationship between the stressor and
endpoints of concern is quantified and the ecological effects elicited are evaluated. In
addition, extrapolations from measurement endpoints to assessment endpoints are conducted
during this phase. The product is a stressor-response profile that quantifies and summarizes
the relationship between (1) the magnitude, frequency, or duration of the stressor in an
observational or experimental setting and (2) the magnitude of response. The stressor-
response profile is then used as input to risk characterization.
In the past, most efforts of evaluating environmental toxicity have focused on
determining the toxicity of an agent for particular species. This usually relied on laboratory
(acute) toxicity studies which yield little information on how that toxicity would be expressed
in natural ecosystems (e.g., community structure, species population, diversity). Furthermore,
these efforts usually focused on direct exposure to a particular contaminated media and did
not address inter-media transfer and more importantly bioaccumulation and food chain
pathways. Therefore, there is a paucity of existing data by which to evaluate ecological
effects of particular stressor. Ideally, fieid testing or chronic testing provides more useful
information.
6.1.1.3 Risk Characterization. As with human health risk assessment, risk
characterization is the final phase of ecological risk assessment. During this phase, the
likelihood of adverse effects occurring as a result of exposure to a stressor are evaluated.
Risk characterization contains two major steps: risk estimation and risk description.
The stressor-response profile and the exposure profile from the analysis phase serve as
input to risk estimation. Three general approaches are suggested to integrate the stressor-
response and exposure profiles: d) comparing the single effect and exposure values ('also
known as the quotient approach discussed below); (2) comparing distributions of effects and
exposure; and (3) conducting simulation modeling. The final choice as to which approach.
6-8
-------
will be selected depends on the original purpose of the assessment as well as time and data
constraints. The uncertainties identified during all phases of the risk assessment should be
examined for the four major areas of uncertainty: conceptual model formulation, information
and data, natural variability, and error. To conform to recent EPA guidance on risk
characterization (discussed in Section 5), uncertainty analyses should be explicit in all
ecological risk assessments.
Risk description has two primary elements: ecological risk summary and ecological
significance. The ecological risk summary can be divided into three components: summary of
risk estimation and uncertainty, weight of evidence, and identification of additional analyses.
First, the results of the risk estimation are summarized in a quantitative statement and the
uncertainties associated with problem formulation, analysis, and risk characterization are
discussed quantitatively and/or qualitatively. Secondly, the confidence in the risk estimates is
expressed through a weight-of-evidence discussion. The weight-of-evidence discussion should
consider the sufficiency and quality of data, corroborative information, and evidence of
causality, and the need for additional analysis.
The interpretation of ecological significance evaluates risk estimates in the context of
the types and extent of anticipated effects. Ecological significance may be defined in terms
of the nature and magnitude of effects, the spatial and temporal patterns of effects, and the
potential for recovery once a stressor is removed. These three aspects of ecological
significance are inextricably linked. A single stressor may cause several effects, or a
particular effects may be affected be several stressors. Therefore it is important ro define the
nature of the assessment or measurement endpoints which are of concern. The magnitude of
effects may be influenced both by ecological context (e.g., reduction in reproductive rate in a
population that reproduces rapidly or slowly) or by spatial or temporal patterns. For example,
a spotted owl requires old growth forests for habitat, and loss of that habitat would have
much greater magnitude for the spotted own than for other species which may have the ability
to adapt to other habitats. Spatial patterns may also be influenced by environmental fate and
transport, and bioaccumulation. Temporal patterns may be influence magnitude of effects due
to the persistence of the stressor as well as how often the stressor occurs, and how it relates
to critical life stages of organisms. The recovery potential for a particular ecological effect
6-9
-------
may be dependent on the nature, duration and, extent of the stressor and may require
additional analyses which should be explicitly discussed in the final risk assessment. As
indicated above, there may be a higher degree of complexity than traditionally found in
human health risk assessment and may require more qualifications in the presentation of risk.
6.1.2 DEFINING ECOLOGICAL RISK ASSESSMENT METHODS
The definition of ecological risk assessment used in this discussion is the qualitative
and/or quantitative appraisal of the potential effects of a pollutant on ecological receptors.
This definition does not include damage assessment which generally applies when the
pollutant has been released in the environment. While some the references used in this
discussion include damage assessment as a part of ecological risk assessment (e.g.,
USEPA/OERR 1989) there are different in that ecologicarrisk assessment is predictive while
damage assessments are after-the-fact assessments of environmental damage. The reader who
refers to those references for more information needs to bear this distinction in mind.
Fate and transport of chemicals can be determined using the same models and methods
used for estimating human exposure in the food chain. The main differences are in the end
points chosen and the methods used to determine risk. The methods used to assess effects are
somewhat dependent on the end points chosen. Unlike human health risk assessment, cancer
is not generally an effect of concern. Acute effects on nonhuman populations include the
death of all or part of a population, increases in undesirable species (such as blooms of blue-
green algae), or reductions in yield of agricultural crops or umber. Two types of assessment
methods exist. Qualitative methods are useful in situations where data and/or resources are
limited. Qualitative methods may be used for a screening level analysis, however, they are
rarely sufficient to support regulatory development.
Most quantitative methods can be classified as either quotient (or ratio) methods or
continuous (or exposure-response methods) (Morten et al. 1988). Barnthouse and Suter (1986)
also include ecosystem uncertainty analysis as a quantitative method. Although
characterization of uncertainty should be a part of any risk assessment, in some cases (e.g.,
where ecosystem effects are the endpomts and a suitable model is available) the uncertainty
analysis itself may be an appropriate assessment method.
6-10
-------
6.1.2.1 Qualitative Methods. A qualitative or semi-quantitative approach is
often used at the screening level of an assessment. At the least, some inventory of the site is
made (or existing studies on the site are used) and important species or site characteristics are
identified. The presence of rare or endangered species or commercially important species
should be noted. Any relevant studies that might help estimate the potential for adverse
effects should be consulted. If sufficient data are available, a hazard assessment should be
conducted. Fate and transport models could then be used to estimate the potential for
exposure. Risk characterization may consist of classifying each chemical into categories
based in a subjective appraisal of the likelihood of an adverse effect occurring. An example
of a categorization scheme is (1) no effect on any species, (2) some effect on sensitive
species, (3) effects on most species, or (4) effects on all species. Categorization schemes may
also be developed for ecosystems depending on the sensitivity of those ecosystems to
stressors. Several qualitative approaches are described in more detail in Norten et al., 1988,
and USEPA/OERR, 1989.
Semiquantitative methods have been adapted to ecological risk assessment from other
disciplines (Barnthouse and Suter 1986; Suter et al., 1987). One of these, fault tree analysis,
is used in engineering safety assessments to determine events and system states that have the
potential to cause the system to fail. If ecological communities are treated as systems, it is
possible to identify potential ecological failures, such as population extinctions, or failure of a
crop to set fruit. However, it is usually impossible even to assign probabilities to the
likelihood of an event occurring.
*
Even when qualitative methods are used, the assessment needs to be performed by
someone who is trained in a relevant discipline. In fact, the absence of good quantitative data
makes the use of an expert even more important when important decisions depend on the
outcome of professional judgment.
6.1.1.2 Quantitative Methods.
Quotient Methods. Although these are considered quantitative methods, they are still
generally used as screening level methods. A benchmark concentration is used to compare
against expected environmental concentrations to assess the possibility of an effect. These
benchmark concentrations are indicative of some toxic endpoint or stressor. The LD50 (the
6-11
-------
dose lethal to 50% of the test population), the LC50 (the concentration lethal to 50% of the
test population), or the chronic no-effect level for a sensitive indicator species are typical
ecotoxicological hazard data. If the ratio of the estimated environmental concentrations
(EEC) to one of these hazard data exceeds certain fixed criteria, a potential adverse effect is
inferred (Bascietto et al. 1990).
Quotient methods are useful for setting standards or priorities, or for identifying the
potential adverse effects for further study. However, quotient methods give only a "risk" or
"no risk" result; they do not give the degree of risk for quotients greater than one. They also
can not be used to estimate the magnitude of effects associated with pollutant concentrations.
A good discussion of the application of the quotient method is found in Barnthouse and Suter
(1986).
Exposure-Response Methods. When an estimate of the magnitude of the risk is
needed, exposure-response or continuous methods are used. These methods use a continuous
curve relating an effect (e.g., reductions in crop yield) to an estimated exposure concentration.
These curves are analogous to dose-response curves used in toxicological studies in human
health risk assessments.
Continuous methods provide much more information to the policymaker or risk
manager. This allows the decision-maker to weigh the costs and benefits associated with
different levels of pollution. The limitation of this method is that it requires more data than
other methods, and that data are often unavailable for many species and ecosystems. In some
cases, concentration-response functions for different life stages of both tested and untested
species can be extrapolated from laboratory tests by analysis of extrapolation uncertainty
(Barnthouse and Suter, 1986).
Ecosystem Uncertainty Analysis. Characterization of uncertainty is especially
important in ecological risk assessment because of the compiexity and variability of the
system being analyzed. The number of unknowns in an ecological risk assessment is
generally much greater than in a human risk assessment. The methods of characterizing
uncertainty used in human risk assessments apply to ecological ones as well.
If reliable ecological models ('See Suter, 1993V exist for the situation of interest, those
models can be used to generate probabilistic estimates of risk. The effects of pollutants or
other stressors on individual organisms (as determined by laboratory or field experiments) can
be extrapolated to populations, trophic levels, or ecosystems by use of ecosystem simulation
6-12
-------
models. Monte Carlo (or other numerical methods) are used to propagate the uncertainties in
the baseline data as well as any uncertainties associated with the model to give probabilistic
estimates of risk for the desired endpoints.
One advantage of this approach is that it allows for assessment of risk on endpoints
for which data may be difficult or costly to obtain. These models allow for the extension of
analysis to the ecosystem population or community levels. Models may also be useful for
assessing the interaction of natural processes such as species competition or predation with
the contaminant effects. The results are expressed as risk with attendant uncertainty, which is
more realistic than single number risk estimates. However, the results are only as good as the
model, and the confidence that can be placed in the model is very important to consider if
this method is chosen. The application of ecosystem uncertainty analysis is described in
detail in Barnthouse and Suter (1986).
6.1.3 MAJOR ISSUES
As scientists and policymakers attempt to define ecological risk assessment and
develop acceptable methods, many issues have been raised (Norten et al. 1988). For the most
part, these issues relate to the lack of information on natural populations and ecosystems
suitable for making important risk management or regulatory decisions.
Most ecological risk assessors choose end points related to population dynamics for
only one of a few of the species that are potentially affected. This is due, in part, to the
significance attached to commercially-important species. However, the choice is frequently
determined by the paucity of data available on community- or ecosystem-level effects.
Scientists recognize that ecosystem response may ultimately be the most important er'fect to
assess, as ecosystem integrity affects ultimately both human and ecological receptors. The
complexity of ecosystems, the difficulty of defining and delimiting an ecosystem, and the
long delays associated with ecosystem responses (decades or centuries) ail contribute to the
difficulty m assessing ecological risks.
Another issue is the lack of sufficient exposure-response data for most species. This
necessitates the use of quotient methods in quantitative risk assessments which provide very
limited information on which to base important regulatory or risk management decisions.
Also, the information needed to treat uncertainty quantitatively is often lacking.
Since most of the information on lethal effects comes from laboratory studies, very
little is known about the interaction of physical and biological processes in nature with
6-13
-------
pollutants. Bioaccumulation plays a critical role in determining the extent of stressor impacts
and introducing potential impacted species not previously considered in laboratory
evaluations. A certain amount of error is associated with extrapolation from laboratory to real
ecosystems and that error is potentially very large. The use of mesocosms (artificial
assemblages of organisms in chambers used to study ecosystem processes such as nutrient
cycling) may help bridge this gap.
A critical issue in ecological risk assessment is what are relevant endpoints of concern
and what level of risks are significant. In human health risk assessment, endpoints (e.g.,
cancer) and level or measure of risk (e.g., 10"6) are usually clearly defined. However, given
our inexperience in ecological risk assessment, the endpoints (e.g., death v. reduced fertility)
and measure of risks (e.g., number of deaths, single deaths v. 150 deaths in a population) are
not clearly defined.
6-14
-------
6.2 CHEMICAL MIXTURES
While some potential environmental hazards involve significant exposure to only a
single compound, most instances of environmental contamination involve concurrent or
sequential exposures to a mixture of compounds that may induce similar or dissimilar effects
over exposure periods ranging from short-term to lifetime (EPA, 1986). The EPA has long
recognized the importance of chemical mixtures in environmental risk assessment. When
EPA developed their first set of risk assessment guidelines in 1986, chemical mixtures was
one of five topics for which guidelines were developed (EPA, 1986).
Mixtures have been defined in the EPA Guidelines for the Health Risk Assessment of
Chemical Mixtures as "any combination of two or more chemical substances regardless of
source or spatial or temporal proximity." In some cases, mixtures may be highly complex
consisting of scores of compounds that are generated simultaneously from a single source or
process (e.g., coke oven emissions and diesel exhaust). In other cases, mixtures may consist
of related compounds produced by commercial products (e.g., PCBs, gasoline, and pesticide
formulations). Other mixtures may consist of unrelated chemicals (chemically or
commercially) which are placed in proximity to one another and eventually may mix and
released into the environment (e.g., hazard waste disposal sites).
The quality and quantity of pertinent information available for risk assessment varies
considerably for different mixtures. Some mixtures are well characterized or have been well
studied, and have well-defined levels of exposure and toxicoiogical properties well defined.
However, in most cases there is limited information on the mixture especially if the
composition of the mixture is in question or toxicologic data on its constituents are iirmted.
As a result, how risk assessments are conducted can be anticipated to vary widely from
mixture to mixture with a high degree of case-by-case decision-making. The EPA guidelines
present a scheme for overall method selection and analysis depending on the nature and
quality of the data. In addition, the EPA has developed the Technical Support Document on
Risk Assessment of Chemical Mixtures (EPA, 1988) which contains a thorough review of
available information on the toxicity of chemical mixtures and a discussion of research needs.
The major points of these documents are summarized below.
6.2.1 Overall Approach
No single approach can be recommended for risk assessments of multiple chemical
exposures. However, general guidelines can be recommended depending on type of mixture,
6-15
-------
known toxic effects of mixture components, availability of toxicity data on the mixture or
similar mixtures, interactions among components, and exposure data. Given the variety of
data available for mixtures, emphasis must be placed on flexibility, judgement, and a clear
articulation of the assumptions and limitations in any risk assessment (EPA, 1986). As a
result, the EPA has proposed an approach to method selection, summarized in Figure 6-2,
A general hierarchy of analysis has been suggested when conducting risk assessments
on chemical mixtures. When possible, the risk assessment should be based on the mixture of
concern. Some highly complex mixtures that are generated in large quantities (e.g., coke
oven emissions, diesel exhaust) have been well studied using toxicity tests conducted directly
on the mixture. If no data are available for a particular mixture, the risk assessment should
focus on data available from similar mixtures. Mixtures can be considered similar if they
have the same components but in slightly different ratios, or if they have several common
components but lack one or more additional components, or have one or more additional
components. Whether mixtures are "sufficiently similar" to justify using the data must be
decided on a case-by-case basis, considering not only the uncertainty of using a data from a
dissimilar mixture but also the uncertainties using other approaches such as component
additivity. Consideration should be given to any information on the components that differ or
are contained in markedly different proportions. Particular emphasis should be placed on any
toxicologic or pharmacokinetic data on the components of the mixture that would be useful in
assessing the significance of any chemical differences between the mixtures. If no data are
available for the mixture of concern or for any similar mixtures, then the risk assessment
should focus on evaluating the mixture components. Consideration should be given to
potential interactions between components (see Section 6.2.3). If no data are available for
any of the above approaches, then no quantitative risk assessment can be conducted.
An alphanumeric classification scheme has also been developed for ranking the quality
of data used in risk assessment and the overall quality of the risk assessment. This scheme is
outlined in Table 6-2.
6.2.2 Existing Methods
Typically, data are not available on an identical or reasonably similar mixture.
Therefore, most of the commonly used methods focus on analysis of individual constituents
and assume dose additivity. Dose additivity is based on the assumption that the components
in the mixture have the same mode of action and elicit the same effects. In actuality,
6-16
-------
1. Assess quality of data using Table 2.
I. Interactions
((.Health Effects
III. Exposure
Adequate
2. Data on mixture of concern?
3. Risk assessment using data
on mixture of concern.
Inadequate
4. Data on similar mixture?
5.Mixtures sufficiently similar?
6. Risk assessment using data
on similar mixtures
12. Compare risk assessment from
steps 3.6. 10.11 as appropriate
Identify preferred assessment
14 Qualitatively assess hazard
No quantitative risk assessment.
7 Data cm mixture components
8. Indices of acceptability and
risk based on component data.
9 Sufficient information to
quantify interactions?
1! Y
10 Risk assessment with interactions
quantified where appropriate
Use addrtivity for all components.
Optional
11. Risk assessment
for all components
using additivity
13. Develop integrated summary including
discussion of uncertainties.
Figure 6-2. FLOW CHART FOR RISK ASSESMENT OF CHEMICAL MIXTURES
6-17
-------
Table 6-2. CLASSIFICATION SCHEME FOR THE QUALITY OF THE RISK ASSESSMENT OF THE MIXTURE
Classification Description
Information on Interactions
I Assessment is based on data on the mixture of concern.
II Assessment is based on data on a sufficiently similar mixture.
Ill Quantitative interactions of components are well characterized.
IV The assumption of additivity is justified based on the nature of the health effects and on the number of
component compounds.
V An assumption of additivity cannot be justified and no quantitative risk assessment can be conducted.
Health Effects Information
A Full health effects data are available and relatively minor extrapolation is required.
B Full health effects data are available but extensive extrapolation is required for route or duration of
exposure for species differences. These extrapolations are supported by pharmacokinetic considerations,
empirical observations, or other relevant information.
C Full health effects data are available but extensive extrapolation is required for route or duration of
exposure for species differences. These extrapolations are not directly supported by information available.
D Certain important health effects data are lacking and extensive extrapolations are required for route or
duration of exposure of for species differences.
E A lack of health effects information on the mixture and its components in the mixture precludes a
quantitative risk assessment.
Exposure Information
1 Monitoring information either alone or in combination with modeling information is sufficient to accurately
characterize human exposure to the mixture or its components.
2 Moaenng information is sufficient to reasonably characterize human exposure to the mixture or its
components.
3 Exposure estimates for some components are lacking, uncertain, or variable. Information on health
effects or environmental chemistry suggest that this limitation is not likely to substantially affect the risk
assessment.
4 Not all components in the mixture have been identified or levels of exposure are highly uncertain or
variable. Information on health effects or environmental chemistry is not sufficient to assess the effect of
this limitation on the risk assessment.
5 The available exposure information is insufficient for conducting a risk assessment.
6-18
-------
compounds have the ability to interact or have different receptors or modes of action.
Interactions will be discussed in greater detail in Section 6.2.3. While dose additivity may
not be the most biologically plausible approach, several studies have demonstrated that dose-
additive models often predict reasonably well the toxicities of mixtures composed of a
substantial variety of both similar and dissimilar compounds. Basic methods to assess
mixtures have been developed for both noncarcinogens (systemic toxicants) and carcinogens.
Most methods used in the Agency focus on the dose additivity of individual constituents. The
Office of Emergency Response and Remediation (OERR), responsible for Superfund cleanups,
has modified this approach to include only a subset of the chemicals in the mixture to act as
indicator compounds. For each mixture the analysis if noncarcinogen and carcinogens is
carried out independently.
6.2.2.1 Noncarcinogens. The approach to assess noncarcinogenic mixtures is based
•s*
on the current methodology used by the Agency for single compounds, i.e., the derivation of
an exposure level that is not anticipated to cause significant adverse effects. The level may
be expressed in a variety of ways depending on the route of exposure, media of concern, and
legislative mandate. As described in Section 3.0, the noncancer risk standard for the
inhalation route is the Reference Concentration (RfC). As described in Section 5.0, risks for
noncarcinogens are evaluated by dividing the exposure level by the RfC, with values greater
than one indicating an exceedance of the Hazard Index (HI). The HI has been developed to
accomplish the same function for mixtures that the RfC does for single compounds. The HI
of a mixture, based on the assumption of dose addition, is defined as:
HI = E,/RfC, + Ej/RfC, + ... + E/RfC,
Where:
E, = exposure level of the ith toxicant, and
RfC. = Reference concentration for the ith toxicant
The HI provides a rough measure of likely toxicity, not a direct measure of incidence. -As
with single chemicals, an HI below 1 is not anticipated to result in adverse health effects. As
the HI approaches 1 concern for the potential hazard of the mixture increases. If the HI
exceeds I, the concern is the same as if an individual chemical exposure exceeded the RfC by
the same level. His exceeding 1 do not necessarily imply that adverse health effects will
occur, only that the potential exists and the likelihood of an effect increases as the HI
increases. The HI is not a direct estimate of risk as it does not define a dose-response
6-19
-------
relationship, or yield risk estimates as probability, nor is there, a strict delineation of "safe"
and "unsafe" levels.
Dose addition, that mixture components induce the same effect by similar modes of
action. Ideally, a separate HI should be generated for each endpoint to support this
assumption. In cases where data clearly indicate that dissimilar modes of action and effects
exist, then the use of the HI should be questioned. In any event, if a hazard index is
generated, the quality of the experimental evidence supporting the assumption of dose
addition should be explicitly discussed as part of the risk estimate.
6.2.2.2 Carcinogens. For carcinogens, the increase in risk can be assumed to be
additive whenever linearity of the individual dose-response curves has been assumed. This is
usually restricted to low doses such as those used with the linearized multistage model. The
risk from simultaneous exposures to several carcinogens in such a mixture can be estimated
>••
from the following equation:
P = Idft
where:
P = excess cancer risk
' dj = exposure level of the i* component
B; = carcinogenic potency of the im component
This equation assumes an independence of action and is assumed to be equivalent to
assumption of dose addition. The equation is an approximation of a more precise equation
for combining risks described in detail in 51 FR 34014. The precise equation is consistent
with the assumption of dose additivity and accounts for the joint probabilities of the same
individual developing cancer as a result of exposure to two or more carcinogens (See EPA.
1986, 51 FR 34014). The difference between equation 5-1 and the precise equation are
negligible for total individual cancer risks less than 1 x 10"1 (or 0.1). (EPA/OSW, 1989) The
•EPA Guidelines for Chemical Mixtures (EPA, 1986, 51 FR 34014) and EPA/ECAO (1988)
provide further detail on mathematical models for multiple chemical risk estimation.
The risk summation technique assumes exposures are in the low-dose range where
responses are linear. At higher risk levels, nonlineanty may need to be considered. The
approach also assumes independence of action by the compounds involved (i.e., that there are
no synergistic or antagonistic chemical interactions and that all chemicals produce the same
effect, i.e., cancer through independent mechanisms of action). If these assumptions are
6-20
-------
incorrect, over- or under-estimation of the actual multiple-substance risk could result
(EPA/OSW, 1989).
There are several other limitations to this approach that must be acknowledged. First,
because each slope factor or URE is typically an upper 95th percentile estimate of potency,
and because upper 95th percentiles of probability distributions are not strictly additive, the
total cancer risk estimate might become artificially more conservative as risks from a number
of different carcinogens are summed. If one or two carcinogens drive the risk, however, this
problem is not of concern. Second, it often will be the case that substances with different
weights of evidence for human carcinogenicity are included. The cancer risk equation for
multiple substances sums all carcinogens equally, giving as much weight to class B or C as to
class A carcinogens. In addition, slope factors or URE's derived from animal data will be
given the same weight as slope factors derived from human data (EPA/OSW, 1989).
6.2.3 Interactions
Those methods described above assume dose additivity, that the toxic action of the
mixture as a whole is equal to the sum of the toxicity of its individual components.
However, the possibility exists that interactions may occur. Compounds may interact and
result in synergistic or antagonistic interactions. A synergistic interaction is a pharmacologic
or toxicologic interaction in which the compounds reinforce or magnify the toxic effect. The
resulting toxicity is much greater than would be predicted by the sum of the individual
toxicities. In an antagonistic interactions are those where a compound may interfere with or
cancel the toxic activity of another compound and the resulting toxicity is much less than
would be predicted by the sum of the individual component toxicities. Data should be
reviewed to determine if there are known interactions between any of the compounds in the
mixture.
Most of the data available on toxicant interactions are derived from acute toxicity
studies using experimental animals exposed to mixtures of two compounds, often in only a
single combination. The use of information from two-component mixtures to assess
interactions in mixtures containing more than two components is difficult from a mechanistic
perspective. Studies of mixtures with more than two chemicals are rare and difficult to
interpret. The EPA's Environmental Criteria and Assessment Office in Cincinnati has
developed a computerized database and data retrieval system called MIXTOX, which contains
summaries of available literature on toxicologic interactions between environmental chemicals
6-21
-------
(EPA, 1992). The database reflects the literature published prior to 1988. Once a
determination of potential interactions has been made, the relevance of the data to chronic
exposures should be assessed, especially with regard to the use of acute data and the presence
of other compounds in the mixture.
6.2.4 Other Mathematical Models
The common methods described above use simple additivity to assess the risks from
chemical mixtures. More detailed models can be used to describe or quantify toxicity of
chemical mixtures. Two basic types of models are available, the dose addition model and the
response addition model. The dose addition model assumes that toxicants in a mixture
behave as if they were dilutions of one another, that the true slopes of the dose-response
curves for individual compounds are identical, and that the response elicited by the mixture
can be predicted by summing the individual doses after adjusting for differences in potency.
The adjustment for differences in potency is defined as the ratio of equitoxic doses. Probit
transformation, introduced in Section 3, typically makes this ratio constant at all doses when
parallel straight lines are obtained. This model is based on the assumption that each
compounds behaves similarly in terms of mechanism of action and target site.
The other model, response addition, assumes that the two toxicants act on different
receptor systems and that the correlation of individual tolerances may range from completely
negative to completely positive. Response addition assumes that the response of a given
concentration of a mixture of toxicants is completely determined by the response to the
components and the pairwise correlation coefficient.
Detailed descriptions and mathematical forms can be found in the Guidelines for
Chemical Mixtures, 51 FR 34021-34022. Each of the above models assume no interactions.
However, in cases where quantitative data are available for interactions, any of the
mathematical models described above can be modified to reflect the interaction, either
synergisms or anatagomsms. Descriptions of mathematical modifications can also be found in
the Guidelines.
6.2.5 Uncertainties
As with ail risk assessments, the uncertainties should be clearly discussed and the
overall quality of the risk assessment should be classified. As described above in Table 6-1,
an overall classification scheme has been developed to classify the quality of a risk
assessment on chemical mixtures. There are several categories for sources of uncertainty,
6-22
-------
including health effects, exposure, and the composition of the.mixture. In many cases, data
will not be available for mixtures of concern or similar mixtures and the assessment must rely
on data for individual compounds. An assumption of dose additivity is typically used unless
there are data on interactions. Interactions among chemicals are likely though information on
the type and magnitude are typically not available. This has been described above in some
detail and it should be explicitly discussed. Other sources of uncertainty include exposure
and composition uncertainties. General uncertainties related to exposure have been described
in Section 4. However, for chemical mixtures these uncertainties may be increased as the
number of compounds of concern increases. If the chemicals in a mixture are not from a
single source or have different chemical and physical properties, then environmental fate,
transport, and environmental concentrations may vary with time. Therefore, care must be
taken in estimating exposure levels for each component and any differences should be
incorporated into the dose-response assessment as well. Furthermore, in many cases the
identity of all constituents in a mixture may not be known or may change with time due to
changes in the source. As a result, unidentified individual components or those without
quantified exposure levels may represent significant risk. When possible, efforts should be
made to identify all components, and conduct a hazard assessment on each of these
compounds. However, when data are limited, major constituents can be used in the risk
assessment, as long as the uncertainties of unidentified components should be explicitly
discussed.
6-23
-------
6.4.1 References For Section 6.1. Ecological Risk Assessment
Barnthouse, L. W. and G. W. Suter, II (eds.). 1986. User's Manual for Ecological Risk
Assessment. Oak Ridge National Laboratory, Oak Ridge, TN. ORNL-6251.
Bascietto, J., D. Hinckley, J. Plafkin, and M. Slimak. 1990. Ecotoxicity and ecological risk
assessment. Environ. Sci. Technol. 24:10-15.
Norten, S., M. McVey, J. Colt, J. Durda, and R. Hegner. 1988. Review of Ecological Risk
Assessment Methods. Prepared for Office of Policy Planning and Evaluation, USEPA,
Washington, D.C. EPA/230-10-88-041.
Suter, G.W., II,L.W. Barnthouse, S.M. Bartell, T. Mill, D. Mackay and S. Paterson.
Ecological Risk Assessment. Lewis Publishers, Chelsea, MI
Suter, G. W., II, L. W. Barnthouse, and R. V. O'Neill. 1987. Treatment of Risk in
Environmental Impact Assessment. Environ. Management, 11:295-303.
US Environmental Protection Agency, 1989. Risk Assessment Guidance for Superfund,
Volume II. Environmental Evaluation Manual. Office of Emergency and Remedial
Response, Washington, D.C. EPA/540/1-89/001.
US Environmental Protection Agency, 1990. Reducing Risks: Setting Priorities and Strategies
for Environmental Protection., Science Advisory Board, Washignton, DC.
U.S. EPA, 1990: Bioaccumulation of Selected Pollutants in Fish. A National Study, Vol. II.
Office of Water Regulations and Standards (WH-552), Washington, DC. EPA 506/6-90/00 Ib.
US Environmental Protection Agency, 1991, Summary Report on Issues in Ecological Risk
Assessment, Risk Assessment Forum, Washington, DC., February 1991, EPA/625/3-91/018.
U.S. Environmental Protection Agency, 1991: Technical Support Document for the
Determination of the Need to Regulate Pulp and Paper Mill Sludge Landfills and Surface
Impoundments.
US Environmental Protection Agency, 1992a, Framework for Ecological Risk Assessment,
Risk Assessment Forum, Washington, DC., February 1992, EPA/630/R-92/001.
US Environmental Protection Agency, 1992b, Report on the Ecological Risk Assessment
Guidelines Strategic Planning Workshop, Risk Assessment Forum, Washington, DC., February
1992, EPA/630/R-92/002.
US Environmental Protection Agency, 1992c, Peer Review Workshop Report on a Framework
for Ecological Risk Assessment, Risk Assessment Forum, Washington, DC., February 1992,
EPA/625/3-91/022.
6-24
-------
U.S. Department of Commerce, NTIS, 1993: Derivation of Proposed Human Health and
Wildlife Bioaccumulation Factors for the Great Lakes Initiative. PB93-154672.
U.S. EPA, 1993: Interim Report on Data and Methods for Assessment of 2, 3, 7, 8-
Tetrachlorodibenzo-p-dioxin Risks to Aquatic Life and Associated Wildlife. Office of
Research and Development Washington, DC. EPA/600/R-93/055.
U.S.EPA, 1993: Great Lakes Water Quality Initiative Criteria Documents for the Protection
of Wildlife (Proposed) - DDT, Mercury; 2,3,7,8-TCDD, PCBs. Office of Water (WH-586),
Office of Science and Technology, Washington, DC. EPA-822-R-93-007.
U.S.EPA, 1993: Proposed Water Quality Guidance for the Great Lake System. Federal
Register, Vol 58, No. 72. RIN 2040-AC08 [FRL 4205-5] Proposed Rules.
U.S.EPA, 1993: Wildlife Criteria Portions of the Proposed Water Quality Guidance for the
Great Lakes System. Office of Water (WH-586), Office of Science and Technology,
Washington, DC. EPA--822-R-93-006.
6.4.2 References for Section 6.2, Chemical Mixtures
U.S. Environmental Protection Agency, 1986. Guidelines for the Health Risk Assessment of
Chemical Mixtures, Office of Research and Development, 51 FR 34014-34025.
U. S. Environmental Protection Agency, 1992. MDCTOX Version 1.5, An Information
System on Toxicologic Interactions for the MS-DOS Personal Computer, Environmental
Criteria and Assessment Office. Cincinnati, Ohio.
U. S. Environmental Protection Agency, Environmental Criteria and Assessment Office.
November 1988. Technical Support Document on Risk Assessment for Chemical Mixtures.
Final Draft, ECAO-CIN-572.
6-25
-------
-------
7.0 USES OF RISK ASSESSMENT METHODOLOGIES
7.1 INTRODUCTION
The objective of this chapter is to illustrate the selection and use of the risk
assessment methodologies described in this report. Performing a risk assessment may be
fairly simple or it may be extremely complicated. Risk assessments differ greatly from one
another in scope, depth, approach, and level of uncertainty. The approaches chosen will
depend on the goals of the risk assessment, the level of uncertainty that is acceptable, and the
time, resources, and expertise available. As described in Chapters 2 through 5, there is a
wide range of methods for performing each step of a risk assessment. It would be impossible
to describe exactly when and how one might use each of the methods presented. Therefore,
this chapter illustrates the selection and use of various methods by presenting three example
case studies.
The case studies are hypothetical, but were designed to reflect realistic situations in
which States may use risk assessments to reach decisions on control of toxic air emissions.
The example case studies include a range of methodologies from simple screening approaches
to very detailed approaches. Case I describes a screening study to create a relative ranking of
industries in order to prioritize further risk assessment and regulatory activities. The ranking
is based on a rough estimate of the maximum cancer risks posed by the higher emitting plants
within each industry group. Case II presents a more refined estimate of risk which uses more
site-specific information and estimates aggregate population risk as well as maximum
individual risk. Noncancer risks are also addressed. Case III illustrates a very detailed site-
specific risk assessment considering multiple pollutants and indirect as well as inhalation
exposure pathways. This type of assessment might be used in a permitting decision. The text
of each case study discusses why particular methods were chosen and how the goals of the
risk assessment and the level of uncertainty deemed acceptable influenced the choice of
methods.
Each case study is organized into several sections. The first section describes the
objectives and scope of the study. The sections that follow describe each of the four steps of
risk assessment — hazard identification, dose-response assessment, exposure assessment, and
7-1
-------
risk characterization. Within each of these sections there is a discussion of the methods used
and why they were selected, the type of output obtained, and the level of expertise and
resources needed. Finally, there is a short section identifying a few of the possible extensions
or alternative methods that could have been used in the case study. It should be stressed that
there is no single correct way to conduct a risk assessment, and these case studies are not
intended to prescribe the specific approach that should be used in a given situation. Rather
they are to illustrate how methods could be selected and used.
7-2
-------
7.2 CASE I
7.2.1 Study Objectives
The purpose of this example case study was to rank the relative cancer risks posed by
various industrial source categories within a particular state. The results of the ranking were
intended to provide a conservative basis for prioritizing the industrial source categories for
further, detailed risk assessment and potential state regulation. Thus, the level-of-effort and
resource expenditure for this study were desired to be kept at a minimum. The result of the
study was a table showing the source categories ranked relative to one another by scores of
estimated maximum cancer risk. To additionally aid decision-makers in setting priorities, the
number of facilities in and total emissions for each source category were also presented.
Since the objective of the study was to estimate the source category rank order,
emphasis was placed on using a consistent, conservative methodology for assessing the
relative risks, without regard to the actual values predicted. The actual values generated
incorporate a high degree of uncertainty, due to the screening nature of the study. This level
of uncertainty was judged to be acceptable, since the study would only be used to prioritize
source categories. More detailed risk assessments with lower uncertainties would be
conducted for high priority sources before developing any potential regulations.
7.2.2 Scope
A simple screening technique for estimating the relative risks associated with the
various source categories would achieve the study objectives without requiring large amounts
of time, resources and expertise. Information on source category emissions and emission
release characteristics and on pollutant toxicity was obtained from resources that are readily
available. Given that the ranking was only intended to provide a basis for establishing
priorities, a high level of accuracy was not needed. For the purpose of this case example, it
was decided that the dispersion technique and emissions estimates should be conservative in
order to avoid missing source categories that could impact public health. Screening
dispersion modeling was conducted to differentiate the impacts of point and fugitive (area)
sources.
The source categories involved in the study were identified as those of primary
Standard Industrial Classification (SIC) codes 20 through 39. Facilities with these SIC codes
7-3
-------
are those for which toxic chemical release inventory data are reported under Section 313 of
the Emergency Planning and Community Right-to-Know Act. Only facilities located within
the state boundary were included in the assessment. Emissions estimates for both point and
fugitive sources were used, based on the 1989 reporting year only.
The exposure media was assumed to be ambient air. Indirect exposure pathways were
not considered. Chemicals identified as carcinogens by EPA were included in the study.
7.2.3 Selection and Use of Risk Assessment Methods
7.2.3.1 Hazard Identification .
Methods. This study considered only those chemicals for which EPA had already
performed a hazard identification. Only those chemicals classified as human carcinogens
(Class A) or probable human carcinogens (Classes B1 and B2) were included. Carcinogenic
classification was obtained from the EPA Integrated Risk information System (IRIS). The
IRIS data base can be searched to isolate those chemicals for which some carcinogenicity
information is available. Since the objective of the study was to rank source categories by
maximum cancer risk, noncancer health effects were not considered.
It was decided not to search other literature sources or perform hazard identifications
for chemicals not yet reviewed by EPA or included in IRIS. Further information gathering
and assessment would have required more level of resources and scientific expertise than
were available for the study. Also, the State considered it reasonable to concentrate first on
regulation of those chemicals already identified as carcinogens.
Output. The output of the hazard identification step was a list of chemicals along with
their weight-of-evidence classification (i.e.. Class A, Bl or B2, as defined in Chapter 2 of this
report).
Expertise and Resources. Since a State computer account with the National Library of
Medicine already existed, the IRIS data base was readily accessed via personal computer
communications software. (The information contained in the data base is also available on
hard copy through the National Technical Information Service.) This step and the dose-
response step that follows were performed by a health scientist in approximately one day.
7-4
-------
7.2.3.2 Dose-Response Assessment.
Methods. To determine maximum cancer risk, dose-response information for the
carcinogenic chemicals was required. The IRIS data base was accessed to obtain the
inhalation unit risk estimates for each chemical. The unit risk estimate (URE) represents the
increased cancer risk from a lifetime (70 year) exposure to an ambient air concentration of 1
ug/m3. The IRIS data base contains updated UREs and hazard identification information and
is part of the National Library of Medicine's Toxicology Data Network (TOXNET).
Output. The output of the dose-response step was a listing of carcinogenic chemicals
and their associated UREs and weight-of-evidence classifications. A sample of IRIS output
pertaining to the carcinogenicity of a given pollutant is provided in Appendix A.
Expertise and Resources. As mentioned in the hazard identification step, the IRIS data
base was readily accessed via personal computer communications software, since a State
computer account with the National Library of Medicine already existed. (Information
contained in the data base is also available on hard copy through the National Technical
Information Service.) This step and the hazard identification step were performed by a health
scientist in approximately one day.
7.2.3.3 Exposure Assessment.
Emission Characterization. Point and area (fugitive) source emission rates were
obtained through the Toxic Chemical Release Inventory System (TRIS), which is also
available through the National Library of Medicine's Toxicology Data Network (TOXNET).
Data contained in the inventory are submitted to EPA by industrial facilities on a yearly basis,
in compliance with Section 313 of the Emergency Planning and Community Right-to-Know
Act (Title III of the Superfund Amendments and Reauthorization Act of 1986), Public Law
99-499. Non-point (area) and point (stack) air emissions are required to be reported under the
rule.
Point source emissions in TRIS are reported as the total of all releases to the air that
are released through stacks, vents, ducts, pipes, or any other confined air stream. Air
emissions not so released are considered fugitive emissions, and include: fugitive equipment
leaks, evaporative losses from surface impoundments and spills, releases from building
ventilation systems. Each release estimate is accompanied by a letter indicator of the
7-5
-------
principal method used to determine the amount of release reported. The four codes used
represent the following methods: monitoring, mass balance calculations, emission factors and
engineering calculations. Only a single code is entered per emissions estimate. Therefore,
the code identifies the method applying to the largest portion of the total estimated release
quantity. The emission estimation techniques (and associated uncertainties), while varying
from plant-to-plant, were assumed to be consistent between standard industrial source
categories as a whole.
The Toxic Chemical Release Inventory was searched to isolate the facilities located
within the State of interest. Reports submitted for calendar years 1987, 1988 and 1989 may
exhibit reporting of emissions in ranges, for releases to an environmental medium (e.g., air)
that are less than 1,000 pounds for the year. Three ranges are specified: 0, 1-499 and 500-
999 pounds/year. Emissions for those facilities reporting a range in emissions and not a
specific emissions estimate for the year were assumed to be equal to the midpoint of the
range indicated.
Facility emissions were entered into a computerized spreadsheet and sorted by
emission rate within each SIC. The facility with the highest point emissions and with the
highest area emissions of each chemical were identified within each SIC code. Since the
ranking was to be based on maximum risk by SIC, assessments of every facility within an
SIC was not performed. Given the methodology used in the exposure modeling (see the
following steps), facilities emitting smaller amounts would be predicted to produce lower
exposures than those emitting larger amounts. Therefore, the higher emitting sources were
selected for dispersion modeling. It should be noted, however, that emissions alone are not
an accurate measure of risk. Ground level concentrations may be affected by local
meteorology and dispersion characteristics. Exposure may be affected by the proximity of
populations to the facility fenceline or maximum ground level concentration. It may be
possible that the highest emitting facility may not result in the maximum risk. Therefore.
several of the top emitters should be assessed and then compared.
In order to model the facilities, as described in the fate and transport step,
specification of the following point and area emission release parameters was required.
7-6
-------
Emission Release Parameters
Point Sources Area Sources
Stack height Release height
Stack inner diameter Length of side of square area
Stack gas exit velocity
Stack gas exit temperature
Identifying a single representative set of emission release parameters for each SIC
code is a difficult undertaking, and unquestionably introduces more uncertainty in the study
results. The emission release parameters indicated above differentiate and dictate how the
pollutant initially disperses in the atmosphere, and greatly influence the prediction of ambient
concentration to which people are assumed to be exposed. Determining a representative set
of release parameters is complicated by the fact that a wide range of source characteristics
exists within a given SIC code. For example, the majority of facilities within SIC code "A"
could be typified by short stacks and iow-to-moderate gas exit velocities. The average value
of stack height and gas exit velocity calculated over the population of facilities within that
SIC code would reflect these characteristics. While facilities with these stack parameters
could be the most numerous, it is possible that they may account for only a small percentage
of the total emissions within the SIC. If the stack heights of facilities with high emissions are
much taller than the average value selected, the impacts of these facilities may be greatly
overestimated, and consequently bias the source category ranking. Essentially, if dispersion
characteristics are to be introduced into the assessment, representative emission release
parameter data of good quality are required in order to produce meaningful results.
For the purpose of this case study, it was assumed that source parameters by SIC code
were obtained from the State's computerized emission inventory system. Since it was not
7-7
-------
intended that the unique characteristics of each facility be modeled in this case study, a single
set of worst case (i.e., producing the worst atmospheric dispersion) release parameters per SIC
code was determined.
Another product of the emissions characterization step was the total number of
facilities for the State, and the total emissions by chemical for each SIC. While this
information was not used in the modeling, it was presented in the final report.
Fate and Transport Analysis. Pollutant transport was estimated with the EPA
SCREEN model. As its name implies, the SCREEN model is a screening technique that is
used for estimating the air quality impact of stationary sources conservatively. Selection of
the SCREEN model was therefore consistent with the objectives of the study; furthermore it
is a model that is readily available and easy to use.
For receptors in simple terrain (terrain below stack top), SCREEN predicts ambient air
concentrations using standard bi-variate Gaussian dispersion model assumptions. No chemical
fate mechanisms are employed. Facility emissions were converted to units of grams/second
for input to the model.
The highest emitting point and area sources associated with each SIC were modeled in
SCREEN. For the purpose of this study, the following set of options were selected for
modeling each SIC facility.
SCREEN Model Options
Simple Terrain
Full Meteorology
Rural Mode
No Building Downwash
Automated Receptor Array: 200 m to 2000 m
In actuality, the terrain and land use classifications and occurrence of building
downwash will vary from facility to facility. Specification of these variables on a facility-by-
facility basis was beyond the scope of this study; hence they were kept constant. Predicted
7-8
-------
concentrations therefore reflect the magnitude of the emission rate and the emission release
parameters.
Population Characterization. No population characterization was required for this
study, as the source category ranking was based on maximum risk to the individual.
Exposure Calculation. For the modeled facilities within each SIC, the highest
predicted concentration was identified. An individual was simply assumed to be exposed to
this maximum ambient air concentration.
Output. The final products of the exposure assessment step were the rough estimates
of maximum carcinogenic pollutant concentration for the facilities with the highest point and
area emissions within each SIC. In the risk characterization step, these values were used to
estimate maximum risks to individuals for each SIC.
Other outputs included information on the number" of facilities and the estimated
quantity of emissions for each facility, sorted by SIC. The total emissions for each SIC were
also calculated.
Expertise and Resources. Accessing the TRIS data base required a personal computer
with communications software and a computer account with the National Library of Medicine.
The selection and determination of facilities with the highest emissions within each SIC was
accomplished by a permit engineer/data base specialist within a couple of days.
Determination of the representative emission release parameters for each SIC was performed
by a permit engineer/data base specialist and a scientist familiar with air quality modeling and
was performed within approximately two weeks. The SCREEN modeling and summarization
of SCREEN results was performed within a week by a scientist familiar with air quality
modeling. Modeling was performed on a personal computer.
7.2.3.4 Risk Characterization.
Methods. The maximum predicted concentrations from SCREEN were multiplied by
the respective pollutant UREs to obtain indicators of maximum cancer risk. Although UREs
are associated with long-term (annual) average concentrations, they were multiplied by the
hourly average concentrations from SCREEN for the purpose of this case study. Since the
objective of the study was to provide a relative ranking of the source categories, the actual
7-9
-------
value of the risk estimate was not important. Uniformly adjusting the hourly average
concentrations to annual averages would not change the relative ranking.
To produce a maximum cancer risk score for each SIC, the maximum point and area
source concentrations were summed and multiplied by the appropriate UREs. This provides
an indication of maximum risk, but due to the many simplifying assumptions and high degree
of uncertainty in the exposure assessment, this value should not be used as an estimate of
actual risk, but only be used for the purpose of ranking. This value was then scaled by the
total of all SIC maximum cancer risk estimates, as follows.
For each carcinogen:
Indicator of
1.1 (Max Pt Src Cone) + (Max Area Src Conc)l x URE = Maximum Risk
1 (ug/m3) (ug/m3) J (ug/m3)-1 for SIC A
Indicator of 1 Sum of Maximum Risk
2. I Maximum Risk J -^ Indicators for all SICs = Cancer Risk Score
for SIC A
Example calculations for arsenic and benzene are shown in Tables 7-1 and 7-2 at the back of
this case study. Note that generic SIC codes were used in the tables, which are for
illustrative purposes only.
This scheme produced a source category ranking for each pollutant, as illustrated in
Table 7-3. To obtain a cumulative ranking, the maximum cancer risk indicators for all
carcinogens were summed for each SIC and then scaled by the sum of maximum cancer risk
estimates over ail carcinogens and SIC codes, as follows.
Sum of Maximum Risk Sum of Maximum Risk Cumulative
Indicators for all I ^ { Indicators for all 1 = Cancer Risk
carcinogens for J I carcinogens and I Score
SIC A ail SICs
A sample format of the cumulative ranking table is given in Table 7-4.
7-10
-------
Output. A report of the results of this case study would include a summary table
indicating the top ranked SIC for each individual pollutant and for all pollutants collectively.
The entire focus would be on ranking and hence on the cancer risk scores produced, rather
than on the estimates themselves. The accompanying discussion would clearly state the scope
and limitations of the study and would indicate how the results might be used to define a
strategy for further, more refined risk assessment. It should be emphasized, for example, that
the ranking was based only on pollutants classified as either human carcinogens or probable
human carcinogens. A particular SIC may rank low with regard to its emissions of
carcinogenic pollutants, but may rank high with respect to its potential emission of highly
toxic (noncarcinogenic) pollutants.
Aside from the ranking based on maximum cancer risk, the study report should also
indicate, on an SIC basis, the total number of facilities and total emissions of each chemical.
The UREs of each chemical should be listed to indicate their relative toxicities. A State
could decide to regulate sources with high total emissions (statewide) or to regulate all
sources of a chemical with UREs of a certain magnitude, even if those sources did not rank at
the top based on the modeled maximum individual risk.
The major sources of uncertainty in the case study should be reported. In this
example, a high degree of uncertainty was considered acceptable, given the study objectives.
Generally, any assumption made in the methodology will introduce a measure of uncertainty.
Uncertainty in the chemical carcinogemcity is indicated by the weight-of-evidence
classification as is given in the hazard identification step. Uncertainties associated with the
UREs are discussed in Chapters 2 and 5 and include, for example, extrapolation of toxicity
data from animais to humans and from high experimental doses to low doses encountered in
the ambient air. In this case, uncertainty exists in the accuracy of the emissions estimates
from TRIS and in the assumption of emissions from facilities reporting their emissions in
ranges.
Uncertainty is introduced in the modeling by the selection of model options and by
selecting only the facility, from each SIC, with the highest point and area source emissions
for input. The degree to which facility emissions vary within an SIC should be identified
with simple statistics. A high degree of uncertainty is associated with the screening model
7-11
-------
results, given the selection and use of a single set of point and area source emission release
parameters for all SICs in the study. The available source parameter data for each SIC
should be reviewed to assess the variability within and between each SIC.
This report would be given to decision-makers within the State agency for their use in
prioritizing sources for further risk assessment and regulation.
Expertise and Resources. Risk calculations (exposure x URE) were done by a scientist
on a personal computer with spreadsheet software. Chemical specific and summary tables
were also generated by computer spreadsheet. Report preparation, including drafting
discussions of the methodologies used, results and uncertainties, was performed by the various
persons responsible for the study. The report was subsequently reviewed and edited by
supervisory personnel. The total time for performing the risk characterization, reporting and
review was approximately two months.
7.2.4 Other Considerations
Several possible variations to the methods described in this case study could be
implemented. These would generally require additional resources and level-of-effort. Some
of the potential options are given below.
• Instead of conducting screening dispersion modeling, emissions could be multiplied
directly by the respective UREs to produce an alternate indicator ranking based
solely on the magnitude of emissions and the chemical carcinogenicity. This
procedure would save resources, if performed alone, and would provide an
alternative rank order if done in combination with our approach.
• Emissions couid be estimated based on production data and established emission
factors for the SIC as opposed to data from TRIS.
• A more refined method for assigning emission release parameters couid be adopted
by incorporating facility specific information.
• This ranking scheme does not consider whether sources are located in high-
population areas. Population distribution information could be gathered to
supplement the ranking.
• The ranking could be based on a greater number of modeled facilities within each
SIC.
7-12
-------
I
X
CO
en
uu
c/3
cc
O
r~-
LU
1
DC
E »— fc—
| g 8
c§ <5 -*
s if
_ = 75 I"
0 P =1 OC
75 -| s i:
.0 (§ .> cu
— 75 ~" o
E c
g -2 — .
.£ 75 <=P* CD_
JS 7z -~, +
!5 CU 2> ^ ^
&—
03 C
en 2
2 75
> 7=
< i
•c c
o 0 S"
X 03
||
S||
'o
Q.
03
>. °C .
^C O to
*= f— , — .
-— * ^ ~f^ <^
1 I 3="
X o
E §
§ 0
S CD
x o
TJ 3
O
CO
CO
CM
O
O
co
g
3
C\J
o
o
IO
<
CM
O
in
CO
o
o
o
C\4
CD
CO
o
§
s
o
o
CO
00
o
CM
CM
O
CO
g
CO
CD
O
o
in
a
S
75
CO
75
-C
o
Q.
_>,
QJ
5
a.
CD
ra
CO
cu
75
>
"33
T3
o
'IJ
LU
QC
0
co
03
.£
(=
O
In
o
75
c
Q
o
O
ca
E
in
cu
._
^i—
'E
CA
c
~
S
c
o
o
£
^
E
"x
ca
S
75
II
^
>—
^
o
T3
O
75
-g
^
—
E
£
'§
E
o
73
o
-5
_c
en
CO
CM
JD
-a
cu
•a
^
C/3
S
ca
o
15
T3
^
C
' =
5
E
X
£
"o
^
O
ca
o
•o
.S
11
8
en
c/3
•~
0
ca
o
|
1
x
ca
^
"o
a.
1
^
,0
en
o
u
I
^
1
i
X
ca
E
=
CO
"o
JH
J2
7-13
-------
x
CO
CO
II
co
LLJ
co
DC
LLJ
LLJ
LLJ
CQ
O
a_
CM
r>»
LLJ
<
1
53
fill
— o
C 'f ^ , - — -
'x .fc "g • — •
3> co §>
is = — <
.So ~"
£0
Si
CO -J2
a> i=
< 0)
f|is
m CO
« o>
o> "=
en 2
2 2
> c
,^ O ^ |
"§ o "c .
X ^ "1^
11
x — -
CO C
?S O
Q.
O
CO
CO
CM
O
O
O
O
g
CM
§
CM
0
in
o
<
*f
0
o
§
in
CM
o
g
0
o
0
C\J
m
•^
fl^
S
o
g
CO
0
o
o
03
0
CM
CM
CO
O
tfi
O
C3
O
O
CO
g
CO
o
o
Q
r—
S
0
1
CO
.a
"3
o.
>,
O)
5
a.
CO
a>
15
">
"a3
o
Lti
LU
QC
o
co
CD
£
0
"•—
C/l
_o
2
"H
CD
CJ
3
O
.2
2
• S
CO
CD
^£
II
'c
CO
.1
c
.0
"co
J=
C
cu
g
o
o
£
3
£
X
ca
£
"5
it
^
•=
CD
8
03
CJ
"ca
"2
-a
|
c
'x
CO
E
"o
_a
ta
CJ
~o
c:
o
o.
OB
5
2
^5
CJ
1
^.
.<£
E
E
15
"o
la
o
^r
o
o
XI
T3
a>
'S
=0
CO
^_
cu
o
tz
CO
CJ
"S5
73
'>
"•^
g
1
X
CO
g
"3
0
"ca
o
-a
.c
n
CD
3
o
en
^"
CD
O
C
CO
CJ
1
i
X
ca
2
7-14
-------
TABLE 7-3. SAMPLE OUTPUT OF SOURCE CATEGORY RANKING FOR SINGLE POLLUTANT
Pollutant: A
Rank Based on
Maximum Individual
Risk
1
2
3
4
SIC
c
D
A
B
Number of
Facilities in SIC
35
20
12
6
Total Pollutant A
Emissions for SIC
(Ibs/year)
90,000
78,000
52,000
50,000
7-15
-------
TABLE 7-4. SAMPLE OUTPUT OF SOURCE CATEGORY RANKING BASED ON ALL POLLUTANTS
Pollutant: Alt
Rank Based on
Maximum Individual
Risk
1
2
3
4
SIC
c
D
A
B
Number of
Facilities
in SIC
35
20
12
6
Total Emissions of
all
Carcinogenic
Chemicals
210,000
170,000
120,000
100,000
Chemicals
Emitted
A, B, C, D
B, C, D, E
A, B,C
A, C, D
7-L6
-------
7.3 CASE II
7.3.1 Study Objectives
The purpose of this case study is to evaluate the impact of a proposed air toxics
regulation that would require control of carcinogenic pollutants to a risk level of 1 x 10"5
maximum individual cancer risk, and control of noncarcinogenic pollutants to exposures
below their respective reference concentrations for chronic inhalation exposure (RfCs). The
primary objective of the study is to identify which source categories, as defined by Standard
Industrial Classification (SIC) code, would be affected by the proposed rule. Sources affected
are identified as those whose ambient impacts, based on source specific dispersion modeling,
are above the proposed carcinogenic or noncarcinogenic risk levels.
The study will provide an indication of the expected permit work load that may result
from adoption of the rule. It will also provide the necessary basis for evaluating the potential
economic impacts of the proposed regulation and for identifying possible compliance options.
Implementation of these additional tasks is not discussed in this case study.
7.3.2 Scope
The proposed regulation would require all facilities within the state boundary that emit
an toxic air pollutant to obtain a permit. In order for a permit to be issued, a facility owner
or operator would have to demonstrate that the ambient air concentrations predicted in the
area surrounding their facility meet the health-based risk criteria of 1 x 10° maximum
individual cancer risk, for carcinogenic air toxics, and that the predicted annual average
concentrations be less than the inhalation RfCs, for noncarcinogenic air toxics. Those air
toxics covered by the proposed regulation are thus those for which EPA has established unit
risk estimates (UREs) and inhalation RfCs.
The exposure media is assumed to be only ambient air, since the regulation is
designed to control the level of toxic pollutants in air. Since the estimates of exposure and
risk will actually be used for comparison with the health risk criteria, uncertainties associated
with the study should be minimized where possible; however, it is accepted that wide
variability in the accuracy of the estimates may exist, depending on the accuracy of facility
emissions and model input parameters. Any assumptions made in the course of the study
7-17
-------
should be conservative. Although the exposure and risk estimates are based on facility-
specific information, representative of a given source category, they are not intended to be
conclusive of the impacts posed by any particular facility. This analysis would not be the
final determination of whether permits would be issued to specific facilities. If the regulation
were promulgated, an affected source could do a more detailed site-specific risk assessment to
verify whether estimated risks were above or below
1 x 10'5.
In addition to maximum individual risks, decision-makers also requested information
on the distribution of risk and the predicted annual cancer incidence in the population
surrounding facilities that could be affected, if such information could be easily generated.
To obtain this information and to calculate the exposure and risk estimates for comparison
with the health risk criteria, the EPA Human Exposure Model II (HEM-II) was selected.
HEM-II is a combination atmospheric dispersion and population exposure/risk model, and has
been used by EPA in setting natipnal emission standards for hazardous air pollutants
(NESHAPs) under Section 112 of the Clean Air Act. Within HEM-II, concentrations
produced from the dispersion model are matched with the population distributed throughout
the modeled region to produce cumulative or aggregate estimates of exposure and risk.
7.3.3 Selection and Use of Risk Assessment Methods
7.3.3.1 Hazard Identification.
Methods. This study considered only those chemicals for which EPA has already
performed a hazard identification. Botfi carcinogenic and noncarcinogenic pollutants were
•
included. Carcinogenic classification was obtained from the EPA Integrated Risk Information
System (IRIS). IRIS contains updated-dose-response information and is available through the
National Library of Medicine's Toxicology Data Network (TOXNET). Noncarcinogenic
pollutants were limited to those for which a Reference concentration for chronic inhalation
exposure (RfC) existed. Chemicals classified as human carcinogens (Class A) or probable
human carcinogens (Classes Bl and B2) were included. The IRIS data base can be searched
to isolate those chemicals for which carcinogenicity information is available.
It was decided not to search other literature sources or perform hazard identifications
for chemicals not yet reviewed by EPA or included in IRIS. Further information gathering
-------
and assessment would have required more resources and scientific expertise than were
available for the study. Also, the State considered it reasonable to concentrate initially on
regulation of those chemicals for which EPA had already performed a hazard identification.
Output. The output of the hazard identification step was a list of carcinogenic and
noncarcinogenic chemicals, with weight-of-evidence classifications (i.e., Class A, Bl or B2,
as defined in Chapter 2 of this report) indicated for carcinogens.
Expertise and Resources. Since a State computer account with the National Library of
Medicine already existed, the IRIS data base was readily accessed via personal computer
communications software. (The information contained in the data base is also available on
hard copy through the National Technical Information Service, and may be purchased on
floppy diskette.) This step and the dose-response step that follows were performed by a
health scientist in approximately two days.
7.3.3.2 Dose-Response Assessment.
Methods. To determine maximum individual and aggregate cancer risk, inhalation unit
risk estimates (UREs) of the carcinogenic chemicals were used. The URE represents the
increased cancer risk from a lifetime (70 year) exposure to an ambient air concentration of 1
ug/m3.
IRIS was accessed to obtain the unit risk estimates. Inhalation RfCs were also
obtained through IRIS and represent ambient air concentrations below which adverse health
effects are generally not expected to occur.
Output. The output of the dose-response step was a table listing the carcinogenic
chemicals, their associated UREs and weight-of-evidence classifications, and the
noncarcinogenic chemicals with inhalation RfCs. A sample of IRIS output pertaining to the
inhalation RfC summary is provided in Appendix A.
Expertise and Resources. As mentioned in the hazard identification step, the IRIS data
base was readily accessed via personal computer communications software, since a State
computer account with the National Library of Medicine already existed. (Information
contained in the data base is also available on hard copy through the National Technical
Information Service.) This step and the hazard identification step were performed by a health
scientist in approximately two days.
7-19
-------
7.3.3.3 Exposure Assessment.
Emission Characterization. To compile a sample set of sources to be used in the
study, regional offices within the State .were asked to identify three major and three minor
facilities in industrial categories as defined by the primary SIC codes. Major sources were
defined as those emitting more than 100 tons per year of any of the six pollutants for which
National Ambient Air Quality Standards exist (NAAQS). Minor sources were considered to
be those emitting less than 100 tons per year of any of these pollutants. The SIC codes
evaluated in this study were limited, based on a previous determination of the SIC codes for
facilities whose air toxic emissions posed the greatest potential risk to human health within
the State.
The facilities identified by the regional offices were mailed survey forms requesting
information on plant location (latitude and longitude coordinates), plant operating schedule,
process descriptions, point and fugitive emissions, control equipment and stack parameters.
Emission rates were requested for only those pollutants identified in the hazard identification
step. Furthermore, the rates requested were maximum hourly and annual average emission
rates. To conservatively estimate exposure, the hourly average rates were modeled as annual
average emission rates in the fate and transport step.
In order to model the facility emissions, specification of the following point and area
emission release parameters was required.
Point Sources Area Sources
Stack height Release height
Stack inner diameter Length of side of square area
Stack gas exit velocity
Stack gas exit temperature
To assess whether building wake effects might be significant, facilities were requested
to indicate if stack heights were exceeded by any building height within the facility. In
7-20
-------
addition, respondents were requested to indicate whether the predominant land use
surrounding their facility was urban (mostly commercial, industrial and compact residential)
or rural (primarily agricultural and residential) for input to the air dispersion modeling.
Approximately 80 percent of the facilities identified responded to the survey. Those
not responding had either closed or claimed no emissions of the pollutants identified. To
limit the amount of modeling required, the facilities were screened by emission rate.
Screening emission levels had previously been determined by the State. The selection of
these screening levels and the conservative modeling analysis conducted to determine them is
not included in the scope of this study. Those sources exceeding the trace levels of emission
for any pollutant were selected for input to the following steps of the study.
Fate and Transport Analysis. Pollutant transport and dispersion of facility emissions
within each SIC were simulated directly using the EPA Human Exposure Model-II (HEM-II).
HEM-II was selected because it is a model that is consistent with EPA guidelines and because
it has the ability to estimate both maximum and aggregate ambient exposure and risk using
built-in meteorological and population data bases. This simplifies the exposure and risk
calculations for the model user.
HEM-II contains the long-term version of the EPA Industrial Source Complex
(ISCLT) model. ISCLT is a Gaussian plume model that calculates annual average
concentrations resulting from continuously emitting point and area sources, such as those
located within facilities included in this assessment. The ISC regulatory default option, which
assigns several model options in accordance with the EPA "Guideline on Air Quality
Modeling" recommendations for regulatory air quality modeling analyses, was used for
modeling all facilities. The urban or rural land use option, which determines the algorithms
used to predict plume dispersion, was selected based on the information provided in the
facility survey.
The default receptor grid was selected in executing ISCLT within HEM-II. With the
default grid, annual average concentrations are predicted at 22.5 degree intervals on
concentric rings of radial distances, in meters: 200, 500, 1000, 2000, 5000, 10000. 20000.
30000, 40000 and 50000. The grid used in the modeling must include receptors near the
facility to identify the maximum concentration and include enough receptors spatially
7-21
-------
distributed throughout the entire modeled region to adequately determine aggregate population
exposure. The meteorological STAR data required to run ISCLT for the selected facilities is
included in an internal HEM-II data base.
Since the majority of facilities in the inventory reported stacks below building height,
building wake effects were assumed to be a potential problem at every facility. Building
heights used as input to ISCLT were conservatively assumed to equal stack height, subject to
the constraint that all buildings were less than 100 feet tall. Building lengths and widths were
calculated by the formula:
D2
4
<+
where Bw is the building width, BL is the building length, BH is the building height and D is
the effective diameter, set equal to the building height to simulate worst-case building
downwash.
In actuality, the influence of terrain and building downwash will vary from facility to
facility. Terrain effects were not considered in the modeling, given the infeasibility of
collecting receptor terrain data and implementing the complex terrain modeling approach
within the study schedule and budget. Terrain can have a significant impact on the predicted
ambient concentrations, as it can bring receptors to heights closer to plume centerline.
To use HEM-II, access to the EPA National Computer Center VAX computer was
required. Since no previous account had been established, an account for invoicing the
computer charges was opened through the National Technical Information Service (NTIS),
and the Pollutant Assessment Branch of EPA, which manages. HEM-II, was contacted on
proper access procedures (the Pollutant Assessment Branch is located in Research Tnangie
Park, NC). Once the account was authorized, HEM-II was readily accessed via personal
computer communications and VAX terminal emulator software.
Population Characterization. Within HEM-II, people are spatially distributed within
the modeled area based on the latest available U.S. Census Bureau Block Group/Enumeration
District (BGED) data. (Note that a Block Group is essentially a combination of contiguous
7-22
-------
city blocks having an average population of near 1,000.) The population data base contains
the number of people associated with the BGED and the latitude and longitude coordinates of
the BGED for mapping with the modeled concentration data.
Exposure Calculation. The maximum individual exposure was estimated for each
source selected for modeling. The maximum individual exposure is simply the maximum
predicted concentration. The presumption in the calculation is that an individual is located at
the point of maximum predicted concentration. Aggregate exposure was estimated within
HEM-II by matching the BGED population data with the modeled concentrations.
Population exposures for given concentration intervals are calculated as the product of the
BGED population exposed to a concentration within the interval and the concentration to
*
which that BGED population is exposed. The sum of these population exposures produces
the overall aggregate or cumulative population exposure.
Output. Within the exposure assessment step, the facility inventory was compiled and
screened by previously determined, trace emission rates to obtain a final source inventory for
the dispersion modeling. A listing of the facilities selected for modeling and of those that
were screened out was also made, sorted by SIC.
The maximum individual exposure and aggregate population exposures from HEM-II
were recorded for each SIC facility modeled. A sample of HEM-II output is provided in
Appendix B. The maximum concentration (maximum individual exposure) for each chemical
was compared to the inhalation RfC for that chemical to identify any exceedances. Results of
this comparison were recorded in a separate table, a sample of which is provided as Table 7-
~j.
Expertise and Resources. Accessing HEM-II required a personal computer with
communications software and VAX terminal emulator software. An account for invoicing the
computer charges was established with NTIS, and the Pollutant Assessment Branch within the
EPA Office of Air Quality Planning and Standards was contacted regarding user support for
HEM-II.
Final determination of the emissions inventory for modeling was accomplished by a
permit engineer and an air quality modeler, working with the State regional contacts, within a
7-23
-------
few weeks, not counting the time elapsed while the completed survey forms were being
returned. The inventory was compiled on a computerized spreadsheet.
The HEM-II modeling of each facility was performed by an air quality modeler and
was completed in approximately two weeks (the amount of time required to input the source
parameters, execute the model and produce the summary reports will naturally vary with the
number of sources involved in the study). Summarization of the HEM-II exposure results
was performed within a few days by the same air quality modeler. Comparison of maximum
concentrations with RfCs was done with a personal computer spreadsheet program.
7.3.3.4 Risk Characterization.
Methods. HEM-II calculates maximum individual risk (MIR) by multiplying the
»
maximum predicted annual average concentration (in ug/m3) by the pollutant unit risk
estimate (risk of cancer incidence per 1 ug/m3). As described in Chapter 5, maximum
individual risk represents the probability that an individual exposed continuously to the
maximum predicted annual average concentration for a 70 year lifetime will develop the
cancer related to that pollutant within his or her lifetime.
Aggregate population risk is also estimated within HEM-II. This aggregate risk is
often expressed as the annual cancer incidence, which is calculated by multiplying the
cumulative exposure (the cumulative product of the number of people exposed to a predicted
pollutant concentration) by the unit risk factor for the pollutant, and dividing by 70 years
[(cumulative exposure x unit risk)/70]. Annual incidence represents the number of cancer
cases expected per year due to emissions from the facility under study, if the population
within the modeled area are exposed continuously to the pollutant for 70 years and the facility
emissions of that pollutant remain unchanged during that period. Note that the exposure and
risk estimates generated for the facilities in this case study are not considered to be absolute
estimates for the modeled facilities, but are to indicate which source categories may oe
affected by the proposed regulation.
HEM-II also generates risk distributions. The output shows the number of indiviauals
exposed to various lifetime risk leveis. as well as the aggregate cancer cases expected in the
segment of the population exposed at each level. The cancer risk is calculated by multiplying
the pollutant unit risk estimate by the exposure for each BGED. Risks for BGEDs within an
7-24
-------
interval are summed to obtain the population risk for that interval. A sample of HEM-II
output is provided in Appendix B.
Output. Results from the HEM-II output were summarized in tables. A sample
format of the table for carcinogens is shown in Table 7-6 to this case study. The tables
shown in Tables 7-5 and 7-6 serve to consolidate the HEM-II output. A final table isolating
those source categories (SICs) which exceeded either the carcinogenic or noncarcinogenic risk
levels was then developed. A sample format is shown in Table 7-7. For carcinogens, the
predicted MIR is expressed as the ratio of the modeled risk to the standard for carcinogens
(10~5 maximum individual risk):
(modeled MIR/10'5).
Similarly, for noncarcinogens, the ratio of the predicted maximum pollutant
concentration (exposure) to the appropriate RfC is calculated as:
(maximum annual average concentration/pollutant RfC).
These values give a rough indication of the amount by which the standards are exceeded (a
ratio of 2, for example, indicates that the estimated value is double the standard).
The results of this case study estimate very generally whether or not facilities within a
given a source category, as defined by SIC code, will be affected by the proposed reguiation.
Although based on actual facility data, the assessment reflects only a sample of the state' s
facility inventory and cannot be conclusive for the entire state-wide population of permitted
air pollution facilities. As in Case I, a full report of the study, including a description of the
methodology, assumptions, and uncertainties was produced as part of the risk characterization.
In addition, an evaluation of the expected work load that may result from adoption of the rule
was presented, along with recommendations for further analysis of the possible compliance
options and associated economic impacts for the facilities affected.
In this example, an attempt was made to limit the amounr of uncertainty introduced by
using actual data, where possible. A moderate degree of uncertainty was considered
7-25
-------
acceptable, however, given the variability in accuracy of model inputs and study objectives
overall. Any assumption made in the methodology will introduce a degree of uncertainty.
Uncertainty is introduced in the modeling by the selection of model options (including
the omission of such items as terrain), and by selecting only a subset of the facilities from
each SIC for input. The degree to which facility emissions vary within an SIC should be
identified with simple statistics. Uncertainty in the chemical carcinogenicity is indicated by
the weight-of-evidence classification it is given in the hazard identification step.
Uncertainties associated with the UREs are discussed in Chapters 2 and 5 and include, for
example, extrapolation of toxicity data from animals to humans and from high experimental
doses to low doses encountered in the ambient air.
Expertise and Resources. Risk calculations from HEM-II were compiled and
summarized by an air quality scientist. Report preparation", including drafting discussions of
the methodologies used, results and uncertainties, was performed by the various persons
responsible for the study. The report was subsequently reviewed and edited by supervisory
personnel. The total time for evaluating the risk results, reporting the outcome for each SIC,
and circulating the study report for peer review was approximately two months.
7.3.4 Other Considerations
Possible variations of the methods described in this case study could be implemented.
These variations are indicated below.
• The study couid be expanded to include a determination of possible compliance
options for the sources affected. For example, if a control device would decrease
emissions by a certain percent, but would also affect the emissions release
characteristics, such as gas exit temperature, the predicted emission conditions
could be modeled to determine the controlled maximum risk level.
• If facility-specific information were not available, emissions could be based on
established production data and emission factors for the SIC.
• The exposure and risk estimation couid be performed through use of the Graphical
Exposure Modeling System (GEMS) instead of HEM II. GEMS also incorporates
the long-term version of the Industrial Source Complex (ISC) Model.
*
• Annual average ambient concentrations couid be estimated using an alternate EPA-
approved air dispersion model. This model output can be used as input to HEM
7-26
-------
II. Possible model options could include receptor terrain heights and building
cavity impacts.
All the facilities in one or more of the source categories could be modeled instead
of just a sample. This would greatly increase the level of effort, but would
indicate how many and which facilities are likely to be affected by the proposed
regulation.
7-27
-------
CO
z
LU
CD
O
O
cc
<
o
z
o
a:
2
CO
CO
LU
QC
LU
X
LL
O
>-
CC
CO
in
K-
LU
_l
m
co "c 5
283
c
o
O
g
n~*
c E
3 o)
"o
Q.
0)
O)
co
OJ
> c
llf
3 CD O)
< o a.
E o
1°
X
CO
IE
« CT3
O Z
a.
11
.=: c
o .5
Q. z
w§
IQ
CO
u_
o
o
•^
s
d
CD
2
j^
"o
Q.
CO
<
O
o
^
co
d
CD
"c
3
1
-------
CO
LU
o
CC
CO
OJ
CC
CC
<
CO
cci
o-
fc>
A
i»
CC
15
_-a
'•5
'x
CO
.la
CC
"75
«
'>
1
^""
1
'x
CO
03
ca
^
c
•5
"o
0.
TJ
-O
E
— j
•^
c
^o
"o
Q-
1)
T3
O
0
CO
Q
><
^^
o
CO
1
o
z
CD
0
X
CJ
'-
o
^
s
"o
Q-
^_
CO
>-
HI
o
X
r--
cg
O
"c
"o
o.
CO
>-
ir
O
x
CO
•*
Q
^
"3
"o
Q.
j
!
.,_
-
T
O
x
o
co
O
^
•2
"o
Q.
T_
CD
in
OJ
I
o
'z-
<*?
o
x
Tf
CO
Q
"c
•i
"o
Q.
CVJ
0
CO
7-29
-------
TABLE 7-7. SOURCE CATEGORIES AFFECTED BY THE PROPOSED REGULATION
SIC
Code
A
B
C
SIC
Description
X Manufacturing
Y Manufacturing
Z Manufacturing
Facility
ID
11
25
18
Pollutant
Pollutant C
Pollutant C
Pollutant D ,
Pollutant I
Pollutant H
Ratio of Maximum Risk
to Proposed
Carcinogen Standard3
2.7
60
43
NA
NA
Ratio of
Maximum
Exposure to
Proposed
Noncarcmogen
Standard
NA
NA
NA
1.4
1.4
aFor carcinogens, the ratio of the modeled risk TO the standard of (10~5 maximum
individual cancer risk is determined. A value of. 2%, for example, means that the
predicted MIR is twice the standard. For noncarcinogens, the ratio of the maximum
exposure to the pollutant RfC is determined.
7-30
-------
7.4 CASE III
7.4.1 Study Objectives
This case study examined the health risks associated with emissions from a waste-to-
energy facility for the purpose of obtaining an air permit. The results of this comprehensive
health risk assessment provided the State agency with the potential health risks (both
carcinogenic and noncarcinogenic) occurring in the surrounding community due to the
operation of the proposed facility. The level of detail required for this assessment was much
greater than that in the first two case studies. As a result, the level-of-effort and resource
expenditure for this type of analysis was high.
Since the degree of accuracy for this type of risk assessment needed to be as high as
technically possible, site-specific estimates for all parameters were required. These estimates
used should be as close to actual or "real life" as possible.
7.4.2 Scope
The risk assessment covers only one plant, but it is to be very comprehensive in
scope. The hazard potential of all pollutants to be emitted from the proposed facility must be
determined in the hazard identification process. Dose-response information was identified
from available sources for all pollutants that were potentially toxic. For compounds where
EPA-approved dose-response parameters (such as UREs and slope factors for carcinogens and
inhalation RfCs and RfDs for noncarcinogens) were not available, review of information
source and derivation of values were necessary.
A comprehensive health risk assessment required assessment of exposure by all
pathways including ingestion, inhalation, and dermal contact. Multipathway exposures were
determined using various fate and transport techniques including air dispersion and deposition
modeling and food chain modeling. The parameters utilized in these models (e.g., panicle
mass fractions for deposition modeling) should be derived from available data and be as close
to the actual value as possible.
The final step in the comprehensive health risk assessment was the quantification of
the overall risk to the exposed population. Maximum cancer and noncancer risks and hazard
indices were calculated in order to evaluate the overall risk to the surrounding population.
Aggregate cancer risk was also calculated to determine the number of people exposed to
7-31
-------
various risk levels and the total expected cancer incidence in the exposed population. This
information is often useful for risk communication with the public during the permitting
process.
7.4.3 Selection and Use of Risk Assessment Methods
7.4.3.1 Hazard Identification.
Methods. This case study considered all the pollutants to be emitted from the waste-
to-energy facility. The potential of each chemical to cause both carcinogenic and
noncarcinogenic health effects was evaluated. Initially, readily available data sources such as
IRIS, NTP annual reports, IARC monographs, ATSDR toxicologic profiles, and EPA Health
Effects Assessment Use documents were reviewed to determine the hazard potential of the
emitted pollutants. If the data search process revealed that a pollutant exhibits insignificant or
low toxicity (e.g., calcium, sodium) the compounds were eliminated from further evaluation in
the health risk assessment.
For compounds which have not been evaluated by an agency or group, the general
scientific literature was reviewed by a toxicologist to determine carcinogenic and
noncarcinogenic potential. However, instances where a compound had not been evaluated by
some group or agency were rare. Toxicologists also made a determination on the weight-of-
evidence for compounds the general literature identified as potential carcinogens. A scheme
similar to the EPA's weight-of-evidence classification system was used.
Output. The output of the hazard identification step was a series of tables listing the
chemicals emitted, the weight-of-evidence classification for carcinogens, and the
noncarcinogenic health effects endpoints of each chemical. References were also cited on the
tables and kept for use in writing the final report.
Expertise and Resources. Hazard identification was done by a toxicologist and a
scientist. The scientist searched the IRIS data base and some of the other readily available
sources of information. A personal computer was used for searches of the IRIS data bases
and for literature searches.
The toxicologist reviewed the general scientific literature and made the weight-of-
evidence determinations for compounds not previously reviewed by EPA or another Agency.
7-32
-------
This step required a thorough knowledge of toxicology and epidemiology. Another
toxicologist served as a peer reviewer for this step in the risk assessment.
The hazard identification step of Case in involved a lot more time and effort than
were required in Cases I and II because of the large number of sources that were reviewed.
Depending on the number of compounds emitted from the source in question, a thorough
hazard identification could take from one to several weeks.
7.4.3.2 Dose-Response Assessment.
Methods. In order to determine the cancer and noncancer risks associated with the
emissions from the proposed facility, dose-response parameters were required. For
determining carcinogenic risks, inhalation and oral unit risk estimates [(ug/m3)"1] or slope
factors [(mg/kg/day)"'] were necessary. The dose-response parameters required for the
noncarcinogenic risk assessment were RfCs for inhalation'exposures and RfDs for oral
exposures.
The primary source for both carcinogenic and noncarcinogenic dose-response
parameters was the IRIS data base. IRIS contains updated dose-response information and is
available on the National Library of Medicine's Toxicology Data Network through the
TOXNET directory. Once a computer account has been established through the National
Technical Information Service (NTIS), the data base is readily accessible via a personal
computer. The information contained in the data base is also available on diskette through
the NTIS. An example of the output from IRIS is given in Appendix A.
Since a number of compounds have not been reviewed or the dose-response
parameters verified by the EPA CRAVE or RfD work groups, additional data sources may
have to be searched to identify the dose-response values. A common source of dose-response
parameters is the EPA Health Effects Assessment Summary Tables (HEAST) issued quarterly
by the Office of Solid Waste and Emergency Response (OSWER). The HEAST contains
LJREs, slope factors, RfDs. and RfCs that are listed in the following EPA documents: Health
Effects Assessment documents. Health and Environmental Effects profiles. Health and
Environmental Effects Documents. Health Assessment Documents, and Air Quality Criteria
Documents. This quarterly publication also includes the most recent dose-response
parameters verified by either the CRAVE or RfD work groups.
7-33
-------
Once all the relevant EPA documents were reviewed, it was necessary to develop
dose-response parameters from the primary scientific literature for a few chemicals. The
derivation of dose-response parameters required an expert knowledgeable in dose-response
models and interpretation of toxicologic and epidemiologic data. Determination of the study
used to generate an RfD or RfC was of particular importance. Once a study was identified,
the methods used by EPA to develop an RfC or RfD were followed, but the resulting values
were not official RfCs or RfDs because they had not been through the EPA review process.
Development of the RfCs and RfDs required the identification of the No Observed Adverse
Effect Level or the Lowest Observed Adverse Effect Level from the critical study and
application of the uncertainty and modifying factors described in Chapter 3. Particular care
was taken in reviewing the literature for metabolic and pharmacokinetic information to
determine whether studies conducted for non-inhalation pathways could be used to develop
inhalation reference concentrations.
Derivation of UREs and slope factors for carcinogens is more difficult because dose-
response models have to be utilized. For this study it was determined that there was adequate
data in the literature to develop a URE for one compound not previously reviewed by EPA.
Procedures similar to those used in developing EPA's UREs were chosen to provide
consistency. Dose conversions from animals to humans was done using the surface area
approach described in Chapter 2. The linearized multistage model was used to extrapolate
responses for high experimental doses to low environmental doses, after checking to see that
there was no scientific evidence that a different model was more appropriate. See Chapter 2
for further discussion of dose-response modeling. It should be noted that for most
comprehensive risk assessments adequate dose-response parameters are available through the
previously mentioned sources to quantitate human health risks.
Output. The output of the dose-response assessment was a list of inhalation UREs,
slope factors, and drinking water UREs for the carcinogens along with the weight-of-evidence
classification for each compound. For the noncarcinogens, a list of RfCs and RfDs for each
compound. References and descriptions of methods, assumptions, and uncertainties were also
developed for inclusion in the final report.
7-34
-------
Expertise and Resources. To complete the dose-response assessment, a team
comprised of scientists, toxicologists, epidemiologist, and computer modelers were utilized.
A health scientist was used to search the IRIS data base and retrieve the applicable dose-
response parameters. A toxicologist derived the RfCs, RfDs, UREs and slope factors for the
compounds that had not been previously evaluated, with assistance from a computer modeler
in completing the computerized dose-response modeling. The peer review duties were
completed by a toxicologist and an epidemiologist.
The time expenditure for this task was large due to the level of detail and research
required for the dose-response assessment. The IRIS data base research required a day while
the literature review (select critical studies, identify critical endpoints, etc.) took a couple of
weeks to complete. A couple of weeks were also required to assemble the data, make
-v.
judgements on the methodologies and assumptions, and perform the dose-response modeling.
Peer review and revisions to the draft document required approximately one month.
7.4.3.3 Exposure Assessment.
Methods. The exposure assessment for evaluating the waste-to-energy facility
included a multiple pathway analysis with inhalation, ingestion, and dermal absorption
considered. This component of the risk assessment incorporated facility- and area-specific
data to accurately assess exposure in the communities surrounding the facility.
Emission Characterization. Point source emission rates were obtained by direct
measurement of the pollutants in trial burns from the specific waste-to-energy facility. These
trial burns were used specifically to quantitate the organic and inorganic pollutants being
emitted from the stack. EPA test methods were used for measurement of both criteria and
noncriteria pollutants from stacks. These test methods specify both the sampling protocol and
the analytical technique for qualitatively and quantitatively characterizing emissions. The test
burn approach was chosen because it would more accurately predict actual emissions from the
facility than the use of emission factors or data from other facilities.
Since air permits typically require a muitipathway exposure analysis, emissions from
any wastewater effluent must also be determined. Concentrations of pollutants in wastewater
effluent were determined from grab samples during the trial burns. The effluent resulted from
the blow down of the cooling tower.
7-35
-------
Exposure Pathways. Some State agencies require determination of exposure from all
pathways which contribute a significant amount to an individual's overall exposure. Exposure
pathways considered in this case study included inhalation, dermal absorption, soil ingestion,
ingestion of contaminated crops, meat, and dairy products, direct surface water consumption,
and consumption of contaminated fish. Other exposure pathways such as swimming and
showering were excluded since their contribution to overall exposure is generally negligible
compared to the other pathways. Table 7-8 lists the exposure scenarios considered in the
comprehensive health risk assessment.
Fate and Transport Analysis. Air pollutant transport was estimated with the ISCLT air
dispersion model. The ISCLT model was used with representative National Weather Service
meteorological data to estimate long-term (annual average) pollutant concentrations. ISCLT
is a Gaussian plume model that calculates annual concentrations resulting from continuously
emitting point and area sources. For the waste-to-energy facility exposure and risk
assessments, only annual averages were calculated.
Initially, a screening receptor grid was established for use in the ISCLT model to
predict annual concentrations at 22.5 degree intervals on concentric rings of radial distances,
in meters: 200, 500, 1000, 2000, 5000, 10000, 20000, 40000 and 50000. The screening grid
used in the modeling must include receptors near the facility to identify the maximum
concentration and include enough receptors spatially distributed throughout the entire modeled
region to adequately determine aggregate population exposure.
Based on the screening analysis, a refined grid was developed to pinpoint the
maximum ground level concentration for the annual averaging period. For this case study,
the refined grid consisted of a square cartesian grid (1000 meters on a side) with 100
receptors spaced 100 meters apart. Such grids were centered on the maximum predicted
ground level concentrations identified in the screening analysis. Discrete receptors were also
placed at sensitive receptors, such as schools and hospitals, to determine the exposure of these
sensitive populations.
Surface water was another potential source of exposure that had to be addressed.
Pollutants may enter surface waters by two pathways: transport of airborne pollutants to the
water body and effluent discharges to the water body. Airborne pollutants may enter the
7-36
-------
water body through atmospheric transport and deposition directly onto the surface water or
indirectly through soil deposition adjacent to the water body with subsequent transport via
surface runoff.
Since a lake located near the facility was a source of drinking water, estimates of
concentration attributable to the waste-to energy facility were calculated. The rate of
deposition of pollutants on the lake was calculated with the ISCLT dispersion model.
Concentrations of pollutants in the wastewater effluent were determined via grab samples
obtained during the trial burns.
Another route of exposure that was considered was the consumption of local fish, soil,
crops, and livestock which take up pollutants from their environment. For the evaluation of
exposure from fish consumption, chemical-specific bioconcentration factors were required
since fish tend to accumulate pollutants in their tissues. The bioconcentration factors were
used to estimate the equilibrium partitioning of pollutant between the surface water and fish
tissues. The following equation was used to estimate pollutant concentrations in fish:
Cf = Cw x BA
where:
Cf = Contaminant concentration in fish (ug/Kg)
Cw = Contaminant concentration in water (ug/L)
BA = Bioconcentration factor
Pollutant deposition and concentration in the uppermost soil layer were determined to
evaluate exposure from direct soil ingestion and root uptake in crops. Since pollutants may
be !ost from the soil by leaching, chemical and biological degradation, and volatilization, soil
loss constants can be applied. However, insufficient data existed to estimate loss for most
pollutants. For exposure via direct soil consumption and dermal contact, a mixing depth of 1
cm was assumed: for root uptake by crops, it was assumed thai the soil is tilled so that the
mixing depth is 20 cm.
Human exposure to pollutants by consumption of contaminated crops was estimated
for three types of produce grown locally in the surrounding area: root crops, leafy vegetables,
and fruits. Contamination of produce occurs via root uptake and by direct deposition onto
exposed plant parts. For root crops, only root uptake is a factor. To calculate a concentration
7-37
-------
of a pollutant in crops, plant-soil bioconcentration factors were used to measure the plant's
ability to accumulate pollutants within its tissue. Direct deposition was calculated as a
function of interception fraction, weathering, and duration of exposure. Equations for
calculating plant concentrations from both root uptake and direct deposition can be found in
"Methodology for Assessing Health Risks Associated with Indirect Exposure to Combustor
Emissions" (EPA, 1990).
Food chain models have been developed for predicting the concentration of pollutants
in animal tissues by considering the pollutant concentrations in plants and soils, the quantity
of plants and soils that animals consume, and the biotransfer factors for each type of animal
tissue. A biotransfer factor is defined as the ratio of pollutant concentration in animal tissue
to the daily intake of pollutant by the animal. These factors describe the extent to which
contaminants are transferred from the environment to specific animal tissues.
Pollutant levels in livestock were determined based on dietary intake and pollutant
levels in the crops eaten by the livestock. Concentration of pollutants in crops, as described
above, were calculated using pollutant-specific bioconcentration factors for root uptake, and
interception fractions to estimate direct deposition on plant surfaces. Typical dietary
compositions for livestock were obtained from State and county agricultural authorities.
Population Characterization. The locations of nearby communities and residences
were determined by reviewing U.S. Geological Survey topographic maps and local street and
land use planning maps of the area surrounding the facility. Sensitive receptors (e.g., schools
and hospitals) were of most importance since individuals located at these facilities are those
particularly susceptible to the effects of pollutants. After the location of the sensitive
receptors was identified, air modeling was conducted to determine annual pollutant
concentrations at these discrete points. The HEM-II model was used for estimating aggregate
population exposure, [n HEM-II people within the modeled region are distributed based on
the latest available U.S. Census Bureau Block Group/Enumeration District (BGED) data.
Exposure Calculation. Ambient air concentrations of pollutants emitted from the
waste-to-energy facility were estimated using the ISCLT model. Maximum individual
exposure for the annual average was determined for each chemical assuming an inhalation
rate of 22 nrVday. This inhalation rate is based upon 16 hours of light activity and 3 hours of
7-38
-------
rest per day. The dose rates of each chemical were determined by multiplying the inhalation
rate times the modeled concentration. Concentrations of pollutants at sensitive receptor sites
were also estimated with the ISCLT model. Aggregate population exposure was estimated
with the HEM-II model by matching the BGED population data with modeled concentrations.
It should be noted that aggregate exposure was estimated for the inhalation route only, since
HEM-II does not indicate exposure via other pathways.
Estimates of human exposures from drinking water, soil, fish, fruits, vegetables, and
livestock were made using behavioral characteristics from the U.S. EPA "Exposure Factors
Handbook" (1989). For drinking water, it was assumed that an individual consumes 2L/day.
Soil ingestion was assumed to occur at a rate of 0.1 g/day for adults and 0.2 g/day for
children. The fish consumption rate was conservatively estimated at 140g/day. Fruit and
vegetable consumption were assumed to occur at 142 and "201 g/day, respectively. Beef and
dairy product consumption rates were estimated at 100 and 400 g/day, respectively. And,
lastly, dermal exposure estimates were based on an estimated percentage of total body surface
area exposed. It was conservatively assumed that 100% absorption of the pollutants occurred.
All of the consumption rates given above were assumed to apply uniformly across the entire
population.
Output. The annual average exposure levels of the compounds emitted from the
waste-to-energy facility were estimated in the exposure assessment steps as well as the
aggregate population exposure. Through use of the refined air modeling analysis, the points
of maximum exposure were identified. Exposure levels resulting from consumption of
drinking water, soil, fish, fruits, vegetables, and livestock were also determined. Another
output was a list of references used in the exposure assessment and the assumptions and
uncertainties associated with the analysis.
Expertise and Resources. The exposure assessment was performed by an air quality
scientist and an ecoiogist. The air quality scientist estimated the dispersion and deposition of
the pollutants using ISCLT and HEM-II. A personal computer with communications software
was required for the air modeling task. A computer account with EPA was required for
assessing HEM-II. The ecoiogist evaluated the other relevant exposure pathways (drinking
7-39
-------
water, soil, etc.) by obtaining the proper EPA-approved algorithms for estimating exposure.
A personal computer was required for developing spreadsheets to calculate human exposure.
Fate and transport analysis for Case HI involved considerably more time and effort
than for Cases I and II. The multipathway nature of this comprehensive risk assessment
required more details and more actual or "real-life" values. As a result,the exposure
assessment could require anywhere from 1 to 3 months to complete depending upon the
number of compounds and pathways to be considered.
7.4.3.4 Risk Characterization.
Methods. Both cancer and noncancer health risks were estimated for the community
surrounding the waste-to-energy facility. Cancer risk was expressed as a probability and
derived by multiplying the URE or slope factor by the maximum exposure estimate. In order
to determine the overall cancer risk, risk estimates from all pathways and compounds were
added. Table 7-9 presents the results from such an analysis. This approach assumes dose
additivity and that synergistic and antagonistic effects do not occur. The resultant maximum
individual risk reflects the risk associated with inhabiting the point of maximum impact for a
lifetime and consuming all foodstuffs from local sources.
Aggregate population risk was estimated within HEM-II and is expressed as an annual
cancer incidence for the population exposed to the facility emissions. This value is calculated
by multiplying the cumulative exposure (the cumulative product of the number of people
exposed to a predicted pollutant concentration) by the pollutant URE, and dividing by 70
years [(cumulative exposure x unit risk)/70]. Annual incidence represents the number of
cancer cases expected per year due to emissions from the facility under study, if the
population within the modeled area is exposed continuously to the pollutant for 70 years and
the facility emissions of that pollutant remain unchanged during that period.
HEM-II also generates risk distributions. The output shows the number of individuals
exposed to various lifetime risk levels, as well as the aggregate cancer cases expected in ihe
segment of the population exposed at each level. The cancer risk is calculated by multiplying
the pollutant unit risk estimate by the exposure for each BGED. Risks for BGEDs within an
interval are summed to obtain the population risk for that interval.
7-40
-------
In addition to the carcinogenic risks, the noncancer risks associated with facility
emissions were estimated. For assessing noncancer risks, hazard indices were generated by
dividing the exposure estimates by either the RfC or RfD. If the resultant value is less than
unity, it is assumed that the exposure is not likely to result in any adverse cancer effects. As
with the cancer risks, the noncancer risks were summed across all exposure pathways and
chemicals to arrive at an overall noncancer risk estimate. Table 7-10 shows the output that
results from such an analysis.
The results of this case study estimate the potential cancer and noncancer health risks
in the surrounding community due to emissions from the waste-to-energy facility. Cancer
risks were estimated for the point of maximum impact, for the general population as a whole,
and at sensitive receptor sites (schools, hospitals, etc.). Noncancer risk was estimated at the
point of maximum impact only. As in Case I and II, a full report of the study, including a
description of the methodology, assumptions, and uncertainties was produced as part of the
risk characterization.
Uncertainties. A number of uncertainties were associated with each step of this
comprehensive health risk assessment. These uncertainties are introduced for steps in which
assumptions have to be made for particular parameters. In the hazard identification step,
uncertainty exists in the determination of the compound's carcinogenic potential.
Uncertainties associated with the derivation of UREs are discussed in Chapters 2 and 5 and
includes extrapolation of animal data to humans and extrapolation from high to low doses.
Uncertainty is introduced in the modeling by the selection of various model option which
affect the pollutant dispersion and deposition. In the multipathway exposure analysis,
uncertainty arises from the selection of consumption rates via the various pathways. There is
also uncertainty associated with the estimation of uptake by biota in the exposure assessment.
Output. A detailed report presenting the exposure and risk estimates for the waste-to-
energy facility and a summary of the methodology, assumptions, and uncertainties associated
with the health risk assessment.
Expertise and Resources. Cancer risk calculations from HEM-II were compiled and
summarized by an air quality scientist. The multipathway cancer and noncancer risk
calculations were performed by a toxicologist on computerized spreadsheets. The final report
7-41
-------
was reviewed and edited by a senior level toxicologist and air quality scientist. The total
time required for generating the risk estimates and describing the methodology, assumptions,
and uncertainties was two months.
7.4.4 Other Considerations
Alternative methods are available for conducting a comprehensive health risk
assessment than described in the above case study. The following is a list of some of these
alternatives.
• For both qualifying and quantifying emissions from the waste-to-energy facility,
test data from a similar facility could be used if available in place of actual data.
AP-42 emission factors may also be used if available.
• The exposure and risk estimation could be performed through use of the Graphical
Exposure Modeling System (GEMS) instead of HEM-II.
• A multipathway exposure analysis may not be required by all State agencies.
• Dose-response models other than the linearized multistage model, such as the one-
hit and Weiiull models, could be used if the data fit.
• Short-term (24-hr average or less) ambient air concentrations could be modeled to
evaluate short-term health effects for chemicals for which acute or subchronic
dose-response parameters have been derived.
7-42
-------
D
o
o
cc
LJJ
UU
LU
CO
I
Q
LU
CO
O
Q.
O
cc
Q.
CC
O
LL
CO
o
cc
<
LU
O
co
LU
CC
rj
CO
O
CO
rv.
LU
—I
CQ
<
Assumptions
-
05
•ss
o
CC
CD
21
CO
O
Q.
x
LU
"3
73
CD
•^
73
CD
O
CD
^
"c
o
"ca
O
5
"5.
CD
CD
CC
E
3
• Worst-case annual average emission
rates
• Normal meteorologic conditions
• Impacted individual resides at maxim
impact point
• Assumed crops not washed before
consumption
CO
Q.
O
O
"^
Si
co
c
g I
•— c
1 § 1
n — O
C « CO _
O ra CD S
•.= -^ O) 2
•5 £ ••= ^
CO t -= CD
-C CD O ^^
£ Q co S
* * * *
.
o
• Beef/dairy farms assumed to be
located at point of highest concentrat
outside city limits
co
0)
£
73
0)
"55
_c
S. 1
^ Q.
It
V3 *O
^ "O
O) ^
-£ co
•
^-
Q
a.
o
.-a
a.
P
w ! ™
0
Q.
CB
Q.
.i -5
E c
3 CD
£ la _
x £ o
^ < CO
^
CO
ii
3
"5
o
o) 5
1 1
111
?B < CO
^ • •
• Concentration at maximum immpact
point assumed for entire watershed
• All deposited particulate assumed to
reach surface water
CO
73
_0)
CO
c
co
1 I
P °
W 0
§ 1- '
o E
"£3 (/j
0) C
.i O
Q O
•
CD
-X
"CD
£ l_
"3 —
O CO
i 8
1 =
"3 co
5 •
7-43
-------
>•
p
o
CD
OC
LU
LJJ
6
111
LU
1
UL
C/3
^i
CO
cc
cc
LU
O
1
UJ
tr
en
ri
LU
—i
CQ
1
u.
1
en
^
0.
cr
CO
O
X
UJ
„
c.
CO
a_
c/1
|
O_
^
"c5
0
__!
CD
05
(U
CO
O)
CO
"o
X
<-r)
oo
<•>
o
X
O5
0
X
CO
o
X
^f
m
o
X
•^r
CM
O
^D
X
CO
m
O
X
CO
OO
O
X
CO
o
•»~
X
~
o
X
in
"^
0
O
X
CO
CO
0
X
o
I —
0
X
n-
CD
0
X
,_
Q
0
X
CO
evi
o
X
^~
cd
n
O
X
in
CM
O
•r—
X >
CO
at
O
X
»—
CO
CO
o
X
0
X
CO
CO
o
X
co
•—
0
0
X
m
LU
^.
X
CO
o
X
^~
h-
o
X
••—
o
•^~
X
CO
ID
X
0
""
.-,
o
X
o
X
o
X
^
h-« i
o
X
-------
I
LU
g
CO
LU
CO
^
CO
cc
cr
LU
o
a
LU
SS
LU
CC
o
X
<
01
CO
LU
CO
s
o
i
o
o
o
o
o
1
o
o
Si
CO
0
i
o
00
o
o
o
.
o
o
O
CO
8
0
CO
o
s
o
o
o
i
0
55
0
o
o
o
o
o
m
8
o o
1
I
0
CD
8
o
0
s
0
o
§
o
Q
s
Si
o
0
0
§
o
•v-
1
0
o
o
o
8
0
CNJ
o
e\j
cr>
S
o
0
o
CJ
c>
ao
O
CD
O
o
ao
0
0
'
CO
o
0
i
0
at
i
1
1
s
O |
o
C\J
o'
o
I
7-45
-------