Regulatory Toxicology and Pharmacology 80 (2016) 125e133

Contents lists available at ScienceDirect

Regulatory Toxicology and Pharmacology

journal homepage: www.elsevier.com/locate/yrtph

Managing risks of noncancer health effects at hazardous waste sites: A

case study using the Reference Concentration (RfC) of
trichloroethylene (TCE)
Michael L. Dourson a, Bernard K. Gadagbui a, *, Rod B. Thompson b, Edward J. Pfau c,
John Lowe d
a
Toxicology Excellence for Risk Assessment (TERA) Center, University of Cincinnati, OH, USA
b
Risk Options, Indianapolis, IN, USA
c
Hull & Associates, Inc., Dublin, OH, USA
d
CH2M, Spokane, WA, USA

a r t i c l e i n f o a b s t r a c t

Article history: A method for determining a safety range for non-cancer risks is proposed, similar in concept to the range
Received 24 March 2016 used for cancer in the management of waste sites. This safety range brings transparency to the chemical
Received in revised form specific Reference Dose or Concentration by replacing their “order of magnitude” definitions with a
3 June 2016
scientifically-based range. EPA’s multiple RfCs for trichloroethylene (TCE) were evaluated as a case study.
Accepted 14 June 2016
For TCE, a multi-endpoint safety range was judged to be 3 mg/m3 to 30 mg/m,3 based on a review of
Available online 22 June 2016
kidney effects found in NTP (1988), thymus effects found in Keil et al. (2009) and cardiac effects found in
the Johnson et al. (2003) study. This multi-endpoint safety range is derived from studies for which the
Keywords:
Trichloroethylene
appropriate averaging time corresponds to different exposure durations, and, therefore, can be applied to
Safety range both long- and short-term exposures with appropriate consideration of exposure averaging times. For
Risk management shorter-term exposures, averaging time should be based on the time of cardiac development in humans
Uncertainty factor during fetal growth, an average of approximately 20e25 days.
Non-cancer hazard © 2016 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND
Short-term exposure license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Reference dose/concentration
Vapor intrusion
Sensitive subpopulation
Indoor air

1. Introduction
Risk managers responsible for making choices about accept-
able oral and inhalation chemical exposures at hazardous waste
* Corresponding author. Toxicology Excellence for Risk Assessment (TERA) Cen-
ter, University of Cincinnati College of Medicine, 160 Panzeca Way, Kettering Lab-
oratory, Room G24, Cincinnati, OH 45267-0056, USA.
E-mail address: gadagbbd@ucmail.uc.edu (B.K. Gadagbui).
1
Waste sites throughout the United States are often managed using risk
assessment information and methods from the U.S. Environmental Protection
Agency (EPA). Risk-based, medium-specific concentrations such as Regional
Screening Levels (RSLs) (formerly known as risk-based concentrations (RBCs) or
preliminary remediation goals (PRGs)) have been derived using these EPA methods
(EPA, 1989, 1991b, 2004, 2009, 2012a) and have been widely used in the assessment
and management of site-related risks from environmental sources.
sites1 have typically focused first on concerns about cancer. As risk
managers became increasingly familiar with how to effectively
manage these cancer risks, it became routine to make screening
and closure decisions using the widely accepted 100-fold upper-
bound, excess, lifetime, cancer risk range of 10
4 to 10
6 (U.S.
Environmental Protection Agency (EPA), 1991a). This range
allowed managers the flexibility needed to address the risk at
waste sites that differed with respect to environmental setting,
history, and current and future uses, based on the needs of the
community.
Risk managers have also always considered non-cancer health
effects in this process, but such effects often did not drive man-
agement decisions, or when they did, the evaluation lacked a cor-
responding non-cancer risk range. For example, risk managers
generally applied a Hazard Quotient (HQ) of 1 as the criterion for a

http://dx.doi.org/10.1016/j.yrtph.2016.06.013
0273-2300/© 2016 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
126 M.L. Dourson et al. / Regulatory Toxicology and Pharmacology 80 (2016) 125e133

non-cancer effect, rather than applying a range of acceptable HQ should become an integral component of the IRIS process” (NAS,
values.2 The HQ concept is used by various agencies, each using its 2014; page 130).
allowable health protective level as the denominator in the equa-
tion (for examples of these levels please see: http://toxnet.nlm.nih.
In addition, risk managers must now contend with situations
gov/newtoxnet/iter.htm). Although EPA (Barnes and Dourson,
where risk management decisions at sites may be driven by non-
1988; EPA, 1994, 2002) and others such as Felter and Dourson
cancer endpoints, even when the cancer endpoint has also been
(1998) have shown that the information underlying the HQ, that
quantitatively defined for a given chemical. For example, the RfD
is, the Reference Dose (RfD) or Reference Concentration (RfC)
and RfC values for trichloroethylene (TCE) have been revised to
[collectively referred to here as the RfD(C)], possesses “uncertainty
lower levels, suggesting greater toxicity than before (EPA, 2011),
spanning perhaps an order of magnitude,” risk managers have
whereas the available cancer risk-specific doses have increased
generally not implemented decisions based upon this stated un-
slightly, suggesting less cancer risk. Therefore, in any given expo-
certainty. Consequently, non-cancer hazards have frequently been
sure setting, the resulting HQ for the non-cancer endpoints for TCE
evaluated and regulated as if a bright line existed when exposures
is correspondingly increased (i.e., indicative of greater toxicity). As a
were at, or very near, the RfD or RfC, i.e., where the hazard quotient
result, these lower RfD and RfC values are now within the 10
4 to
of one (1) or less was interpreted as without risk, and values greater
10
6 range of risk-specific doses or risk-specific concentrations
than 1 were associated with some unspecified, unacceptable risk.
associated with the cancer endpoint. Consequently, risk managers
It can be easily argued that the phrase in the definition of an
may encounter unfamiliar challenges in making effective risk
RfD(C) “with uncertainty spanning perhaps an order of magnitude”
management decisions at sites with TCE contamination, in part
is precautionary. This is because the RfD(C) derivation process
because the non-cancer endpoint does not have an associated risk
contains several uncertainties, and the application of a factor for
range.
each of these uncertainties provides a margin of safety based on the
These challenges to effective risk management are not simply
behavior of the “average” chemical (Dourson and Stara, 1983;
academic. For example, TCE exhibits significant toxicity from both
Dourson et al., 1996). Since most determinations of the RfD(C)
oral and inhalation exposures, and is frequently encountered at
have more than one source of uncertainty, and consequently
contaminated sites (ATSDR, 2015). Moreover, TCE has been re-
exhibit multiple successive precautionary adjustments (i.e., the
ported as part of “background” conditions in about half the resi-
application of multiple uncertainty factors), this implicit margin of
dential structures in the United States (Dawson and McAlary, 2009),
safety increases as the overall uncertainty factor gets larger, as does
with higher background concentrations approaching the range of
the implicit protectiveness of the resulting RfD(C), as demonstrated
EPA’s RfCs. For TCE specifically, risk managers are thus confronted
theoretically by Swartout et al. (1998).
with twin challenges: (1) the discernment of TCE concentrations
Since their origin by Barnes and Dourson (1988) and EPA (1994),
that are attributable to environmental (subsurface) contamination
RfDs and RfCs have increasingly been based upon more sophisti-
from TCE concentrations from those that are associated with con-
cated data, algorithms and models (e.g., Renwick, 1991, 1993;
founding, background sources and (2) determining a remedial
Renwick and Walker, 1993; Jarabek, 1995a, 1995b; Dourson et al.,
objective resulting in acceptable indoor air concentrations of TCE.
1996; Guth et al., 1997; Kalberlah and Schneider, 1998; Meek
The multiple variables affecting potential and actual human expo-
et al., 2001; Haber et al., 2001, 2002; Dourson and Younes, 2002;
sures to TCE are relevant and important to risk managers, and these
EPA, 2002; Dourson and Patterson, 2003; IPCS, 2005; Seed et al.,
variables amplify the need for a clearer understanding of the non-
2005; Dourson and Parker, 2007; Gentry et al., 2003; EPA, 2012b;
cancer hazards associated with environmental concentrations
Meek et al., 2014; Dourson et al., 2013). This increased sophisti-
above the RfD or RfC, represented by HQ values greater than 1.
cation has defined internal concentrations from exposure dosing,
In part because of these concerns, the Alliance for Risk Assess-
accounted for toxicokinetic differences between the test animal
ment (ARA) was petitioned to form a coalition to develop guidance
and human populations, characterized route-to-route internal dose
to facilitate effective risk management decisions at TCE-
extrapolations, and interpolated effects levels between the LOAEL
contaminated sites. Several open meetings and deliberations
and NOAEL in the experimental study, among others. These
were held in 2012 and 2013; these efforts culminated in a guidance
probability-based elements fundamentally change the RfD(C)
document posted at http://allianceforrisk.org/guidance-for-
development process and necessitate a clearer understanding of
contaminated-sites/. The analysis presented herein is based in
risk in practical application. These changes require the user to go
part on this work. Specifically, this paper proposes a process to
beyond the bright line decision process currently employed. Expert
determine a range that is consistent with the uncertainty in the RfD
bodies are also urging such movement (NAS, 2009, 2014; EPA,
or RfC, and that is similar in concept to the acceptable risk range of
2014). For example,
10
6 to 10
4 for upper-bound, excess, lifetime, cancer effects. This
“EPA should clearly present two dose-response estimates: a “safety range” would allow managers to have comparable flexibility
central estimate (such as a maximum likelihood estimate or a in the management and/or regulatory closure of waste sites,
posterior mean) and a lower-bound estimate for a POD from particularly where the evaluation of non-cancer effects drives the
which a toxicity value is derived. The lower bound becomes an risk assessment. The evaluation of the non-cancer effects associ-
upper bound for a cancer slope factor but remains a lower ated with TCE is used as an example to illustrate the determination
bound for a reference value.” And further on: “EPA should of this safety range and its subsequent application in risk man-
develop IRIS-specific guidelines to frame uncertainty analysis agement due its prevalence at waste sites. The choice of TCE is
and uncertainty communication. Moreover, uncertainty analysis particularly relevant because of its overlapping exposure concen-
trations associated with its cancer and non-cancer effects, and the
uncertainties associated with its confounding background sources
resulting in exposure concentrations near the RfC.
2
For ingestion exposures, the hazard quotient (HQ) is the ratio of an average
daily dose to the Reference Dose (RfD). For inhalation exposures, the HQ is the ratio
of the exposure concentration of a substance (EC) to its Reference Concentration
2. Methods
(RfC). The HQ associated with acceptable exposures is generally one (1), with a
precision of one significant figure (i.e., an implicit HQ range of 0.95e1.5). The information used to evaluate the TCE risk assessment values
 M.L. Dourson et al. / Regulatory Toxicology and Pharmacology 80 (2016) 125e133
were taken directly from EPA’s Integrated Risk Information System
(IRIS) report (2011). EPA identified three candidate RfC values for
non-cancer inhalation toxicity. These three candidates are
described briefly as:
 a candidate RfC of 3 mg/m3, based on toxic nephropathy in fe-
male rats (NTP, 1988);
 a candidate RfC of 2 mg/m3 based on decreased thymus weight in
female mice (Keil et al., 2009);
 a candidate RfC of 2 mg/m3 based on fetal heart malformations in
rats (Johnson et al., 2003); however, at least one letter to the
editor and 2 errata have been published clarifying errors in the
original report (Anonymous, 2005; Hardin et al., 2004; Johnson
et al., 2004); findings from five studies conducted from 1989 to
1995 (US EPA, 2011) using a novel dissection method, and have
not been confirmed using guideline compliant, GLP methods.
Each of the candidate RfC values may be evaluated with respect
to its underlying uncertainty, including the precision and the safety
inherent in its derivation.
The safety range for each candidate RfC was built from the actual
definition of the RfD(C) as having “uncertainty spanning perhaps an
order of magnitude.” We evaluated each of the candidate RfC values
with respect to its underlying uncertainty, including the precision
and the safety inherent in its development by focusing on the word
“perhaps” by Barnes and Dourson (1988). This word was used in the
definition because of the variation in the underlying toxicity da-
tabases of different chemicals (as per author Dourson). For
example, a chemical database may have sufficient information in
humans and experimental animals so that the resulting RfD(C) can
be developed with a one-fold or three-fold uncertainty factor.
Uncertainty in such an RfD(C) is smaller, that is, perhaps less than an
order of magnitude. In other cases, fewer data are available so that
the resulting RfD(C) is developed with an aggregate uncertainty
factor of 1000 or more. Uncertainty in these RfD(C) values is greater,
that is, perhaps more than an order of magnitude.
The uncertainty in each RfD(C) value includes its inherent pre-
cision and safety. Precision refers to the repeatability of the overall
process. What this means for a RfD(C) is how close a second RfD(C),
estimated for the same chemical given the same information,
would be if developed by a different expert or expert group. In such
cases, the precision might be best characterized as perhaps three-
fold on either side of the RfD(C) (Felter and Dourson, 1998; EPA,
2002), although this could be less than or greater than a three-
fold margin, as described above.
Within the uncertainty of each RfD(C), the concept of safety
refers to the determination of its degree of protectiveness. This
degree of protectiveness is expected to vary, because a RfD(C) is
developed using one or more uncertainty factors, each of which is
protective based on the observed behavior of the “average”
chemical (Dourson and Stara, 1983; Dourson et al., 1996). This
protectiveness provides assurance that any RfD(C) is an underes-
timate of the expected value of the actual safe dose, i.e., the No
Observed Adverse Effect Level (NOAEL) for a sensitive human
subpopulation.3 Furthermore, the use of multiple uncertainty fac-
tors results in even more protective RfD(C) values as theoretically
demonstrated by Swartout et al. (1998). Because of this, the portion
of the uncertainty in an RfD(C) associated with safety is best
characterized as a range above the RfD(C), where the latter is
3
This follows directly from the definition of the RfD or RfC, and is reflected in
several examples on EPA’s Integrated Risk Information System (IRIS) where NOAEL
values for a sensitive human subpopulation have been used as the basis of an RfD
with an aggregate uncertainty factor of 1.
127
considered as the floor to this safety range, i.e., the RfD(C) is the
lowest value within the range (Dourson and Stara, 1983; Felter and
Dourson, 1998).4
Based on this understanding of precision and safety within an
“order of magnitude,” we then developed a range associated with
each of the candidate RfCs for TCE. This range was developed with a
floor, a ceiling and an intermediate value, and because each of these
RfCs was based on different data, the ranges were anticipated to be
somewhat different. As further described in ARA (2013), criteria
chosen to reflect the judgment of this range included:
 overall confidence in the RfC as demonstrated by the size of the
overall uncertainty factor;
 determination of the steepness of the dose response slope for
the specified effect (Summary from ARA. 2013);
 confidence in the determination of the critical effect; and
 confidence in the determination of the point of departure (POD).
Furthermore, since risk managers might wish to see an overall
safety range based on all three RfCs, we determined a “Multiple-
endpoint range of safety” through an integration of the above four
criteria for each of the three RfCs using professional judgment.
Finally, the overall safety range was compared to three hypothetical
TCE exposure scenarios to show how such a range might be used at
a waste site.
3. Results
3.1. Defining the uncertainty range in the TCE RfCs
As described in Methods, the precision of each of the candidate
RfCs may be considered as a uniform or equal distribution around
the RfC (i.e., the RfC is the central value of the distribution). This
precision would be expected to vary among the RfD or RfC values
for different chemical substances, and among the three different
candidate RfC values. This is because the individual and aggregate
uncertainty factors are unique to each risk value, and because the
judgment of the critical effect for a chemical substance is not al-
ways straightforward. Thus, a second expert group might develop a
different value given the same information.
TCE itself exemplifies this variation in precision. Depending on
the choice of critical effect and uncertainty factor, EPA (2011)
judged RfCs to be either 2 mg/m3 for the two candidate RfCs
based on either Keil et al. (2009) and Johnson et al. (2003),5 or to be
3 mg/m3 for the candidate RfC based on NTP (1988). The precision of
any one of these candidate RfC values can be determined by com-
parison to any second one. Depending on which TCE RfC is devel-
oped first, this precision is either 30% lower or 50% higher.
4
For example, see Dourson and Stara (1983), which states, “A possible modifi-
cation to the standard approach would be to present a range for the ADI rather than
one value. The range could be based at the high end on the average reductions in
dose needed to estimate the ADI (from Figs. I and 3) and the body-surface area ratio
(Fig. 2), and at the low end on the standard 10-fold reductions.” (p. 234). The ADI is
the acceptable daily intake, a parameter that is similar or equivalent to the RfD,
both of which are usually represented in units of milligram per kilogram body
weight per day (mg/kg-d). This recommendation is similar to that suggested by NAS
(2014, vide supra).
5
Please note that the use of the Johnson et al. (2003) study is for demonstration
purposes only. During the course of this study, it became apparent that EPA’s use of
cardiac anomalies for developing an RfC is highly controversial and not universally
shared among government agencies and expert bodies (see supplemental mate-
rials, ARA, 2013, section 2). It may be that an analysis of this specific effect is
warranted by appropriate experts, but at the very least risk managers need to
understand that this endpoint cannot, by itself, be used with confidence to form the
basis of any quantitative hazard estimate.
128 M.L. Dourson et al. / Regulatory Toxicology and Pharmacology 80 (2016) 125e133
Table 1
Different safety ranges for candidate TCE RfCs. All values are in mg/m3. Shaded areas indicate best overall safety range for risk management purposes. See ARA (2013) for
additional discussion as described in supplemental materials.
Study IRIS UFa Steep Slopeb
Johnson et al. (2003) 10 Lower
NTP (1988) 10 Higher
Keil et al., 2009 100 NA
NA¼Not available.
a
Size of the uncertainty factor as on IRIS.
b
Steepness of the hazard slope (i.e., the slope of the line describing hypothetical population responses at concentrations above the RfC), as per ARA (2013)
c
Confidence in the choice of critical effect as per ARA (2013) and text.
d
Confidence in the point of departure as per ARA (2013) and text.
In contrast to precision, the safety in each of the three candidate
RfC values for TCE is associated with a unique range of values that is
expected to lie above each candidate RfC, because choices made at
each step in the process of its derivation incorporate a margin of
safety. These choices include judgments of critical effects of
differing severity, the use of an effect that may not relate to
humans, the use of a lower limit on the benchmark dose (BMD),
and/or the use of a 10-fold default uncertainty factor in lieu of data
suggesting that a smaller uncertainty factor may be more appro-
priate. The range of safety may not be the same among these
candidate RfC values because each was developed using a different
critical effect, POD and aggregate uncertainty factor.
For risk management decisions, the range of safety associated
with each RfD or RfC is generally more important than the
respective range of precision. This is because risk managers are
interested in making decisions that are protective of public health
and it is an understanding of the range of safety in each RfD or RfC
that is generally more informative. With this risk management
mindset, we determined a unique safety range associated with each
candidate RfC for TCE as follows.
For each candidate RfC, the safety range is defined by a floor
value based upon the actual candidate RfC point value, as described
on IRIS. This choice of the individual candidate RfC as the floor of
the range for each non-cancer endpoint is reasonable from a
practical point of view, because managers are unlikely to take
regulatory action at or below these values, due to the protective
nature implicit in the derivation of each candidate RfC, as described
above. However, using the RfD or RfC as the floor to a range in its
value also has theoretical support where it is shown that uncer-
tainty factors are protective based on the behavior of the average
chemical (see Methods). These floor values for each endpoint-
specific safety range are shown in Table 1.
For each candidate RfC, the safety range is also defined by a
ceiling value based upon the POD for each candidate RfC, as
described on IRIS. This ceiling value is then further adjusted
downward, if needed, to reflect the known toxicokinetic differences
between the test organism and the human population in order to
determine the human equivalent concentration, and/or known
uncertainties in the overall database, such as the lack of NOAEL, a
study of insufficient duration, or the lack of a study investigating
important endpoints. These reductions are based on available data,
or a default factor of three as per EPA (2002). The intent of these
reductions, if needed, is to estimate the likely human equivalent
NOAEL from a chronic experimental animal study (or other study
duration, if appropriate for the endpoint).
Specifically, the unadjusted ceiling value for each candidate
RfC’s safety range is the POD of 20 mg/m3 for cardiac effects from the
controversial Johnson et al. (2003) study, of 30 mg/m3 for kidney
effects from NTP (1988), and of 190 mg/m3 for immune effects from
the Keil et al. (2009). No adjustments were needed to the first two
Confidence Safety ranges
Critical Effectc Point of Departured Floor Intermediate Ceiling
Low Low 2 10 20
Medium Medium to Low 3 9 30
Medium Medium to Low 2 20 63
PODs because both of them represent the human equivalent NOAEL
for the duration of interest; however, an adjustment to the final
POD from Keil et al. (2009) was considered necessary because it
was based on a LOAEL, rather than a NOAEL. The adjusted POD was
63 mg/m3 found by reducing the LOAEL of 190 mg/m3 by an uncer-
tainty factor of three. The use of three-fold uncertainty factor
represents the midpoint of the uncertainty factor of 10 for use of a
LOAEL. Each of these ceiling values is reasonable from a practical
point of view, because risk managers are likely to take regulatory
action at or above these values due to the fact that specific toxic
effects can sometimes be associated, or be anticipated, with them.
These ceiling values for each endpoint-specific safety range are also
shown in Table 1.
Because the range between the floor and ceiling varies for each
RfC, we also developed an intermediate value within each safety
range. This intermediate value might enable risk managers to gauge
whether to take regulatory action when exposures fall within the
range. The determination of the intermediate value of the safety
range for each RfD or RfC may encompass many attributes. As a
start, we mesh four considerations:
 size of the total uncertainty factor on EPA’s IRIS as a crude
measure of the overall uncertainty in the database;
 steepness of the hazard slope (i.e., the slope of the line
describing hypothetical population responses at concentrations
above the RfC, see supplemental materials, ARA, 2013, section
3);
 confidence6 in the choices of critical effect; and
 confidence in the POD.
The intermediate values of those safety ranges that are closer to
their respective candidate RfC (i.e., floor value) are generally asso-
ciated with a smaller aggregate uncertainty factor, a steeper hazard
slope, a higher confidence in the critical effect, and a higher con-
fidence in the POD. The intermediate values of those safety ranges
that are further from their respective RfC (i.e., floor value) are
generally associated with a larger aggregate uncertainty factor, a
shallower hazard slope, a lower confidence in the critical effect, and
a lower confidence in the POD.
For the controversial fetal malformation endpoint based on
Johnson et al. (2003), we judged the intermediate value of the
endpoint-specific uncertainty range to be 10 mg/m3, or five-fold
above its respective candidate RfC. This is due to its small aggre-
gate uncertainty factor of 10 (which argues for a value closer to the
candidate RfC), shallower hazard slope (which argues for a value
6
Our judgment of confidence follows the EPA practice of defining low confidence
as the likelihood of new data changing the critical effect, point of departure, and/or
uncertainty factor (greater likelihood ¼ lower confidence) (Dourson, personal
experience as RfD(C) Work Group Co-Chair 1986 to 1994).
 M.L. Dourson et al. / Regulatory Toxicology and Pharmacology 80 (2016) 125e133
farther from the candidate RfC), low confidence7 in the critical ef-
fect (which argues for a value farther from the candidate RfC), and
low confidence in the choice of a benchmark response of 1% (which
argues for a value farther from the candidate RfC). See also sup-
plemental materials.
For the toxic nephropathy endpoint based on the NTP (1988)
study, we judged the intermediate value of the endpoint-specific
uncertainty range to be 9 mg/m3, or three-fold above its respec-
tive candidate RfC. This is due to its small aggregate uncertainty
factor of 10 (which argues for a value closer to the candidate RfC),
steeper hazard slope (which argues for a value closer to the
candidate RfC), medium confidence in the critical effect (which
argues for a value neither closer to nor farther from the candidate
RfC), and medium to low confidence in the choice of a benchmark
response of 5% (which argues for a value farther from the candidate
RfC). See also supplemental materials.
For the decreased thymus weight endpoint based on the Keil
et al. (2009) study, we judged the intermediate value of the
endpoint-specific uncertainty range to be 20 mg/m3, or 10-fold
above its respective candidate RfC due to its larger uncertainty
factor of 100 (which argues for a value farther from the candidate
RfC), medium confidence in the critical effect (which argues for a
value neither closer to nor farther from the candidate RfC), and
medium to low confidence in its choice of a lowest observed
adverse effect level (LOAEL) as the POD (which argues for a value
farther from the candidate RfC). The effect shown by Keil et al.
(2009) does not lend itself to dose response modeling, so a judg-
ment of steepness of hazard slope is not possible (see supplemental
materials.
The safety range for each candidate RfC is shown in Table 1.
3.2. Defining the multi-endpoint range of safety
Since EPA developed three candidate RfCs for non-cancer ef-
fects, the endpoint-specific safety range of each of the candidate
RfCs may be considered individually, or collectively, in risk man-
agement decisions. From the collective evaluation of the endpoint-
specific safety ranges of all three candidate RfCs, a “total safety
range” of 2 mg/m3 to 63 mg/m3 may be inferred. However, extraction
of a “multi-endpoint safety range” from the broader total safety
range may be more useful for risk management decision-making.
The multi-endpoint safety range may be defined here as an esti-
mate of “a daily exposure to the human population (including
sensitive subgroups) that is likely to be without an appreciable risk
of deleterious effects during a lifetime.” This is based upon the
definition of the RfD(C) (Barnes and Dourson, 1988; EPA, 1994,
2002), but where the phrase “with uncertainty spanning perhaps
an order of magnitude” is absent, because the range is now speci-
fied for each RfD(C) value.
The POD for the RfD(C) is typically (although not always) based
upon a NOAEL from a single study for a single critical effect. The
assessment of confidence levels in the study design, critical effect
and POD of a single study and critical effect enables the assignation
of uncertainty factors in a relatively straightforward manner. The
use of the definition of the RfD(C) to also define the multi-endpoint
safety range is a recognition of the complexity associated with the
prediction of a “safe dose” or “safe concentration” associated with
multiple effects, observed in multiple studies and species, and
multiple PODs, each variously based on a BMDL01, BMDL05 or LOAEL
value, appropriately averaged over two different exposure periods,
7
Low confidence argues for an intermediate value that is farther from the RfC
because choices of critical effect and POD in such situations are generally more
protective.
129
as relevant for developmental or chronic effects. Therefore, the
concept of a “safe dose” for the non-cancer effects of TCE has been
applied here to a range of values (i.e., the multi-endpoint safety
range) which represents a “safe concentration” for multiple end-
points and multiple studies, with various degrees of confidence in
study design, critical effect and POD, as illustrated by the shaded
cells in Table 1.
Specifically, the multi-endpoint safety range of the RfC for TCE
was determined by careful discernment of the confidence, preci-
sion and safety associated with each endpoint-specific floor, in-
termediate and ceiling value. The floor of the multi-endpoint
uncertainty range of the RfC for TCE was determined by comparing
the candidate RfC values from each of the three studies (i.e., 2 mg/m3
for both the decreased thymus weight and controversial fetal car-
diac malformation endpoints, and 3 mg/m3 for the toxic nephrop-
athy endpoint). These three floor values are so closely clustered
that, based on the inherent precision, the values are toxicologically
indistinguishable. The endpoint-specific floor value of 3 mg/m3
based on toxic nephropathy represents the endpoint-specific floor
value of the highest overall confidence from among the three
endpoint-specific floor values (see Table 1). Therefore, the value of
3 mg/m3 was selected to represent the floor value of the multi-
endpoint uncertainty range.
Similarly, the intermediate value of the multi-endpoint uncer-
tainty range of the RfC for TCE is determined by comparing the
endpoint-specific intermediate values from each of the three
studies (i.e., 20 mg/m3 for the decreased thymus weight endpoint,
10 mg/m3 for the controversial fetal cardiac malformation endpoint,
and 9 mg/m3 for the toxic nephropathy endpoint). The intermediate
values for the fetal cardiac malformation and toxic nephropathy
endpoints are so closely clustered that the values are toxicologically
indistinguishable. The endpoint-specific intermediate value of 9 mg/
m3 based on toxic nephropathy not only represents the endpoint-
specific intermediate value of the highest overall confidence from
among the three endpoint-specific intermediate values (see
Table 1), but is lower than or equivalent to the other two endpoint-
specific intermediate values. Therefore, the endpoint-specific in-
termediate value of 9 mg/m3 was selected to represent the inter-
mediate value of the multi-endpoint uncertainty range.
The ceiling of the multi-endpoint uncertainty range of the RfC
for TCE is determined by comparing the endpoint-specific ceiling
values from each of the three studies (i.e., 63 mg/m3 for the
decreased thymus weight endpoint, 20 mg/m3 for the controversial
fetal cardiac malformation endpoint, and 30 mg/m3 for the toxic
nephropathy endpoint). The endpoint-specific ceiling values for the
fetal cardiac malformation and toxic nephropathy endpoints are
closely clustered, and the mathematical precision ranges of the two
values overlap8; the endpoint-specific ceiling value for decreased
thymus weight is substantially higher. Therefore, the endpoint-
specific ceiling value for toxic nephropathy (30 mg/m3) has been
selected here as the ceiling value for the multi-endpoint uncer-
tainty range, because it is the endpoint with higher degree of
confidence than the controversial fetal cardiac malformations
endpoint.
In summary, the confidence in each of the endpoint-specific
safety range was subsequently considered in the determination of
the multi-endpoint safety range for risk management purposes (i.e.,
3 mg/m3 to 30 mg/m3). The higher-confidence results of the NTP
8
The mathematical precision (at HQ ¼ 1) associated with the endpoint-specific
ceiling value of 20 mg/m3 corresponds to a range of 19 mg/m3 to 30 mg/m3; the
mathematical precision (at HQ ¼ 1) of the endpoint-specific ceiling value of 30 mg/
m3 corresponds to a range of 28 mg/m3 to 45 mg/m3. Thus, the HQ ranges associated
with the implicit precision of each endpoint-specific ceiling value overlap.
130 M.L. Dourson et al. / Regulatory Toxicology and Pharmacology 80 (2016) 125e133
study were used to determine the floor, intermediate and ceiling of
this safety range. The highly controversial results from the Johnson
et al. (2003) study, while associated with low confidence (see
Supplemental materials, section 2), were nevertheless used to
support this safety range. This multi-endpoint safety range is
embedded within the individual safety range from Keil et al.
(2009); therefore, this study was also considered to be
confirmatory.
3.3. Risk management use of “uncertainty spanning perhaps an
order of magnitude”
Toxicologists are not able to distinguish the absence of health
risk between any two or more values within the uncertainty range
determined for any RfD(C) by its inherent precision and safety. For
TCE specifically, toxicologists cannot differentiate the “safety” of
any value within a range of 3 mg/m3 to 30 mg/m3, nor can they
differentiate among hazard quotients (HQs) developed from any
value within this range. Because of this, managers may use different
values within the multi-endpoint safety range, along with site-
specific exposure assessments, and other site risk management
considerations to make different decisions on a case-by-case basis.
Fig. 1 shows three hypothetical exposure scenarios overlaid on
the TCE safety range of 3 mg/m3 to 30 mg/m3. For example, when a
site-specific exposure assessment defines a range of exposures that
is primarily below the multi-endpoint safety range of 3 mg/m3 to
30 mg/m3, then the probability of inducing any non-cancer effects in
the exposure population is lower and the priority for any risk
management action is reduced (see Fig. 1a). In this case, a risk
manager may decide to take no action, or to delay action pending
further information. Such action would be readily seen as the cur-
rent practice at waste sites throughout the country.
In contrast, when the exposure assessment defines a range in
exposures that also exceeds the multi-endpoint safety range of
3 mg/m3 to 30 mg/m3, then the probability of inducing non-cancer
effects in the exposure population is higher and the priority for
risk management action is increased (see Fig. 1c). In this case, a risk
manager may decide to take action, or to ask for specific informa-
tion that would refine the estimates of health risk and/or exposure.
Likewise, such action would be readily seen as the current practice
at waste sites throughout the country.
When the exposure assessment defines a range in exposures
that is primarily in the multi-endpoint safety range of 3 mg/m3 to
30 mg/m3, then risk managers can use the intermediate value in this
safety range, that is 9 mg/m3 and other site considerations, to gauge
whether a management action is needed or if further information
should be sought (see Fig. 1b).
4. Discussion
In this paper, a “safety range” is proposed for each RfD(C). This
safety range provides further clarification to the phrase “with un-
certainty spanning perhaps an order of magnitude” in the defini-
tion of the RfD(C). This safety range can also be used with exposure
measurements at waste sites to develop a “hazard range”9 to
facilitate risk management decisions for non-cancer health effects.
This hazard range is similar in concept to the upper bound, excess,
lifetime, cancer risk range of 10
6 to 10
4.
The lower end of this safety range is the value of the RfD(C); the
lower end for the corresponding hazard range is a HQ value of 1.
These lower bounds are not only consistent with the definition and
9 10
The term “hazard range” is introduced here since this incorporates the safety
range of the RfD(C) and also exposure measurements.
intent of any RfD(C) value, but also reflect typical risk management
decisions about HQs at waste sites. The upper end of either of these
ranges can vary depending in part on the basis of choice of and
confidence in the critical effect, the relevant POD and any of its
necessary adjustments, the aggregate uncertainty factor, and, for
the hazard range, the uncertainty inherent in the estimates of
exposure. For the safety ranges derived specifically for the three
TCE RfCs, the upper end varies from 10 to ~30 times above the
RfD(C). Fortunately, the information needed to determine the lower
and upper ends of these safety ranges are readily available for any
chemical on EPA’s IRIS or the International Toxicity Estimates for
Risk (ITER) on the National Library of Medicine’s Toxnet. The
resulting hazard ranges, expressed as the range of the HQ, can then
be determined from these safety ranges and available information
related to site-specific exposures.
Risk management decisions about exposure levels wholly below
the lower end of a safety range may be straightforward. For
example, when a site-specific exposure assessment defines TCE
exposures that are primarily below the multi-endpoint safety range
of 3 mg/m3 to 30 mg/m3, the risk manager may decide to take no
action, or to delay action pending further information. Risk man-
agement decisions may also be likewise straightforward when the
exposure levels are wholly above the upper end of this safety range
of 3 mg/m3 to 30 mg/; in such cases, a risk manager may decide to
take action, or to ask for specific information that would refine the
estimates of health risk and/or exposure. Risk management de-
cisions about exposure levels wholly or partially within the range
are more complex; therefore, it may be helpful to have an inter-
mediate value within the safety range to aid in these decisions. The
intermediate values we develop are based on information that is
often available and reflects our best collective judgment. Other
criteria and judgments might also be reasonable. For example, a
recent publication notes five different ways to characterize the
uncertainty inherent in risk assessment values such as RfCs, with
the suggestion to develop a range being one of them (Beck et al.,
2016). In addition, the science panel of the ARA project “Beyond
Science and Decisions: From Problem Formulation to Dose
Response” recommended changes incorporating some of the
findings of Beck et al. (2016) into our proposed method.10 Some of
these suggestions have been adopted here.
Additional discussion is warranted with respect to establishing the
floor of our multi-endpoint safety range for TCE. The mathematical
precision of the cardiac or immune endpoint-specific floor value of
2 mg/m3 corresponds to a range of 1.5 mg/m3 to 2.5 mg/m3; the math-
ematical precision of the nephropathy endpoint-specific floor value of
3 mg/m3 corresponds to a range of 2.5 mg/m3 to 3.5 mg/m3. Thus, the
ranges associated with the implicit precision of the three endpoint-
specific floors coincide at a precision of one significant figure. Since
the endpoint-specific value of 3 mg/m3 based on toxic nephropathy
has the highest overall confidence from among the three endpoint-
specific floor values (see Table 1), and since its precision coincides
with the other two endpoint-specific values, its choice of the floor of
the multi-endpoint range seems most reasonable.
Some discussion is also warranted with respect to establishing the
ceiling value for this multi-endpoint safety range for TCE. For
example, the toxic nephropathy endpoint (30 mg/m3) is a preferred
representation of the ceiling value based on its overall confidence,
whereas an alternative value of 20 mg/m3 based on cardiac effects is a
less plausible representation of the ceiling value, based on the high
uncertainty regarding the frankness and severity of the observed ef-
fect and the quantification of the response, and the ongoing
See http://allianceforrisk.org/wp-content/uploads/2015/07/DR8_Meeting_
Report_and_Appendices.pdf).
 M.L. Dourson et al. / Regulatory Toxicology and Pharmacology 80 (2016) 125e133 131

Fig. 1. Three Hypothetical Exposure Scenarios Overlaid on the TCE Safety Range. a. Hypothetical probability density function of exposures of TCE in indoor air lies primarily below
the 3 mg/m3 to 30 mg/m3 safety range. A relatively small proportion of exposures is higher than 3 mg/m3. Nominal actions or no further action may be warranted for risk man-
agement. b. Hypothetical probability density function of exposures of TCE in indoor air falling within the 3 mg/m3 to 30 mg/m3 safety range. A relatively small proportion of
exposures is higher than 9 mg/m3. Limited action may be warranted for risk management. c. Hypothetical probability density function of exposure of TCE in indoor air falling
within the 3 mg/m3 to 30 mg/m3 safety range. A relatively small proportion of exposures is higher than 30 mg/m3. Actions to reduce exposures may be warranted for risk
management.

controversy regarding the use of this study in dose-response assess-
ment. Specifically, we favor the endpoint-specific ceiling value for
toxic nephropathy over the endpoint-specific ceiling value for fetal
cardiac malformation on the basis of higher confidence not only in the
critical effect (high versus low) but also in the POD (medium versus
low); the latter comparison is of particular relevance since the POD
serves as the endpoint-specific ceiling value for each endpoint.
Nevertheless, the mathematical precision ranges of the HQs based on
either of the endpoint-specific ceiling values also coincide. In this case,
the values may be considered to be approximately equivalent,
enabling risk manager to conservatively account for potential devel-
opmental endpoints when the multi-endpoint safety range is used as
the basis for risk management decisions.11
11
Significant science controversy surrounds the use of the Johnson et al. (2003)
study in regulating TCE exposures, and if the science community had a higher
level of confidence in the Johnson et al. (2003) study, we would have used it to set
the ceiling of the range at 20 mg/m3, as opposed to 30 mg/m3. However, it is
important to recognize that the regulatory community, USEPA and States, are likely
to continue to support the use of this study, at least in the interim. Thus, risk
managers should consider whether to use 20 mg/m3 as the ceiling of the range
because it represents a POD, which can be, and has been (ATSDR, 2013a; 2103b),
interpreted as an “effects level”, or a level at which expression of the toxic effect is
often measurable or “real”. The risk of a toxic effect below these levels, especially in
acute or subchronic exposure, is minimal. Because of continued regulatory support
for the Johnson et al. study, at least over the interim, this may be an important
guidance issue for anyone making day to day risk based decisions for immediate
action.
For TCE, the multi-endpoint safety range is based on floor, in-
termediate and ceiling values for effects of different exposure du-
rations (i.e., developmental or chronic exposure periods) with
expected different averaging times. Therefore, this range for TCE
can be applied to both long-term and short-term exposures, with
the associated differences in exposure averaging times. For shorter-
term exposures, the results from the Johnson et al. (2003) might
also be used to describe the best exposure averaging time, but if so,
this exposure averaging time should be based on the average time
of cardiac development in human fetal growth, approximately
20e25 days (based on Marcela et al., 2012; Hood, 2011;
Dhanantwari et al., 2009; Martin et al., 2002; Schleich, 2002),
rather than a specific window of time. This is because the use of
only a narrow window of exposure during the cardiac development
in humans would be inconsistent with the results of the Johnson
et al. (2003).12
It should be recognized that risk management decisions at
12
Based on the available information, the human heart starts developing between
days 21 and 23 after gestation or later and may be completely formed by 43e46
days into gestation or later (Hood, 2011), indicating the length of cardiac devel-
opment in humans during fetal growth to be in the approximate range of 20e25
days. Since EPA’s RfC is based on all effects that occurred during the whole time of
cardiac development in the rat by Johnson et al. (2003), it is reasonable to use the
whole time of cardiac development in humans, 20e25 days, as the averaging time
for risk management decisions. This range falls within the previously stated range
of 21e30 days (see supplemental materials), and for similar reasons.
132 M.L. Dourson et al. / Regulatory Toxicology and Pharmacology 80 (2016) 125e133
different waste sites may differ, even for the same set of chemicals
with the same underlying hazard information, because other as-
pects of the site problem formulation differ. In fact, differences in
risk management decisions, and in the products of the individual
components of hazard characterization, doseeresponse assess-
ment, exposure assessment, and risk characterizations, should be
expected based on different problem formulations (Dourson et al.,
2013; Ethridge et al., 2015). Such differences in risk management
outcomes do not mean that the science behind the decision has
been tampered, but rather that these situations should be
addressed on a case-by-case basis.
As risk managers struggle with waste site decisions that inte-
grate various aspects of hazard identification, dose response
assessment, exposure assessment and risk characterization, un-
derstanding the uncertainties underlying each component will
increasingly be important. Our approach is an attempt to put un-
certainties in perspective and advance transparency in the chemi-
cal risk assessment process. Such perspective and transparency
should promote more confident use of allowable levels, such as
RfDs and HQs, and their associated safety and hazard ranges, in
waste site management decision-making.
Acknowledgements
The authors acknowledge a gift from the American Chemistry
Council to prepare this manuscript from the previous work of the
Alliance for Risk Assessment (ARA) coalition entitled “Practical
Guidance for Contaminated Sites: Case Study: Trichloroethylene
(TCE) Risk Assessment and Management.” See: http://
allianceforrisk.org/guidance-for-contaminated-sites/. The authors
also acknowledge the insights and thoughtful comments of the
Science Panel of the “Beyond Science and Decisions: From Problem
Formulation to Dose Response” project at its May 2014 meeting.
See: http://allianceforrisk.org/wp-content/uploads/2015/07/DR8_
Meeting_Report_and_Appendices.pdf. However, the research hy-
potheses addressed, findings, and conclusions expressed in this
paper are solely those of the authors and not necessarily those of
supporting groups. The authors thank Mr. David R. Gillay (Barnes &
Thornburg LLP’s Environmental Department, Indianapolis, Indiana)
and Dr. Lorenz R. Rhomberg (Gradient Corporation) for their
insightful comments in an area where discernment is at a premium.
The authors accept responsibility for any errors or omissions.
Transparency document
Transparency document related to this article can be found
online at http://dx.doi.org/10.1016/j.yrtph.2016.06.013.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.yrtph.2016.06.013.
References
Anonymous. 2005. Erratum: Erratum for Johnson, et al. [Environ Health Perspect
113:A18 (2005)]. Environ. Health Perspect. 122(4): A94. http://ehp.niehs.nih.
gov/122-A94/.
ARA (Alliance for Risk Assessment), 2013. Guidance for Contaminated Sites:
Trichloroethylene (TCE) Risk Assessment Case Study. April 15. www.
allianceforrisk.org.
ATSDR (Agency for Toxic Substances and Disease Registry), 2013a. Health Consul-
tation, Millsboro TCE, Millsboro, Delaware. U.S. Department of Health and Hu-
man Services, Agency for Toxic Substances and Disease Registry, Atlanta,
Georgia. February 13.
ATSDR (Agency for Toxic Substances and Disease Registry), 2013b. Public Health
Assessment. Public Health Implications of Site-related Exposures to Tetra-
chloroethylene and Trichloroethylene, Pohatcong Valley Groundwater
Contamination Superfund Site, Warren County, New Jersey. EPA Facility ID
NJD981179047. Prepared under a Cooperative Agreement with the U.S.
Department of Health and Human Services, Agency for Toxic Substances and
Disease Registry, Atlanta, Georgia and New Jersey Department of Health. Final
Release. May 8.
ATSDR (Agency for Toxic Substances and Disease Registry), 2015. Draft Toxicological
Profile for Trichloroethylene. Available at: http://www.atsdr.cdc.gov/
toxprofiles/tp19.pdf.
Barnes, D.G., Dourson, M.L., 1988. Reference dose (RfD): description and use in
health risk assessments. Regul. Toxicol. Pharmacol. 8, 471e486.
Beck, N.B., Becker, R.A., Erraguntla, N., Farland, W.H., Grant, R.L., Gray, G., Kirman, C.,
LaKind, J.S., Lewis, R.J., Nance, P., Pottenger, L.H., Santos, S.L., Shirley, S.,
Simon, T., Dourson, M.L., 2016. Approaches for describing and communicating
overall uncertainty in toxicity characterizations: U.S. Environmental Protection
Agency’s Integrated Risk Information System (IRIS) as a case study. Environ. Int.
89e90, 110e128.
Dawson, H.E., McAlary, T., 2009. A compilation of statistics for VOCs from post-1990
indoor air concentration studies in North American residences unaffected by
subsurface vapor intrusion. Ground Water Monit. Remediat. 29 (1), 60e69.
Dhanantwari, P., Lee, E., Krishnan, A., Samtani, R., Yamada, S., Anderson, S.,
Lockett, S., Donofrio, M., Shiota, K., Leatherbury, L., Lo, C.W., 2009. Human
cardiac development in the first trimester a high-resolution magnetic reso-
nance imaging and episcopic fluorescence image capture atlas. Circulation 120,
343e351.
Dourson, M., Becker, R.A., Haber, L.T., Pottenger, L.H., Bredfeldt, T., Fenner-Crisp, P.,
2013. Advancing human health risk assessment: integrating recent advisory
committee recommendations. Crit. Rev. Toxicol. 43 (6), 467e492.
Dourson, M.L., Stara, J.F., 1983. Regulatory history and experimental support of
uncertainty (safety) factors. Regul. Toxicol. Pharmacol. 3, 224e238.
Dourson, M.L., Felter, S.P., Robinson, D., 1996. Evolution of science-based uncertainty
factors in noncancer risk assessment. Regul. Toxicol. Pharmacol. 24, 108e120.
Dourson, M.L., Parker, A., 2007. Past and future use of default assumptions and
uncertainty factors: default assumptions, misunderstandings, and new con-
cepts. Hum. Ecol. Risk Assess. 13 (1), 82e88.
Dourson, M.L., Patterson, J., 2003. A 20-Year perspective on the development of
Non-Cancer risk assessment methods. Hum. Ecol. Risk Assess. 9, 1239e1252.
Special issue of the Journal of human and ecological risk assessment
commemorating 20th anniversary of the NRC’s Red Book on risk assessment
and risk management.
Dourson, M.L., Younes, M., 2002. Evolution in noncancer risk assessment - current
practice, controversies, and challenges. Comments Toxicol. 7 (5e6), 399e414.
EPA (U.S. Environmental Protection Agency), 1989. Risk assessment guidance for
superfund. In: Volume I: Human Health Evaluation Manual (Part A)Office of
Solid Waste and Emergency Response. EPA/540/1e89/002.
EPA (U.S. Environmental Protection Agency), 1991a. Role of the Baseline Risk
Assessment in Superfund Remedy Selection Decisions. Office of Solid Waste and
Emergency Response. OSWER Directive 9355.0e30.
EPA (U.S. Environmental Protection Agency), 1991b. Risk Assessment Guidance for
Superfund Volume I: Human Health Evaluation Manual (Part B, Development of
Risk-based Preliminary Remediation Goals). Interim. Publication 9285.7e01B.
EPA/540/R-92/003.
EPA (U.S. Environmental Protection Agency), 1994. Methods for Derivation of
Inhalation Reference Concentrations and Application of Inhalation Dosimetry.
Office of Health and Environmental Assessment, Washington, DC.. EPA/600/8-
90-066F.
EPA (U.S. Environmental Protection Agency), 2002. A Review of the Reference Dose
(RfD) and Reference Concentration (RfC) Processes. Risk Assessment Forum.
EPA/630/P-02/002F.
EPA (U.S. Environmental Protection Agency), 2004. Risk Assessment Guidance for
Superfund Volume I: Human Health Evaluation Manual (Part E, Supplemental
Guidance for Dermal Risk Assessment) Final. OSWER 9285.7e02EP.
EPA (U.S. Environmental Protection Agency), 2009. Risk Assessment Guidance for
Superfund Volume I: Human Health Evaluation Manual (Part F, Supplemental
Guidance for Inhalation Risk Assessment) Final. OSWER 9285.7e82. January
2009.
EPA (U.S. Environmental Protection Agency), 2011. Toxicological review of trichlo-
roethylene. In: Support of Summary Information on the Integrated Risk Infor-
mation System (IRIS). EPA/635/R-09/011F.
EPA (U.S. Environmental Protection Agency), 2012a. Human Health Risk Assessment
Bulletins. Supplement to RAGS, USEPA Region 4 Front Page. Available at: http://
www.epa.gov/region4/superfund/programs/riskassess/healtbul.html.
EPA (U.S. Environmental Protection Agency), 2012b. APTI Atmospheric Sampling
Course Appendix G: Significant Figures and Rounding. Available at: http://
www.epa.gov/apti/Materials/APTI%20435%20student/Student%20Manual/
Appendix_G_no%20TOC-cover_MRpf.pdf.
EPA (U.S. Environmental Protection Agency), 2014. Framework for Human Health
Risk Assessment Inform Decision-making. EPA/100/R-14/001. April. Available
at: http://www.epa.gov/risk/framework-human-health-risk-assessment-
inform-decision-making.
Ethridge, S., Bredfeldt, T., Sheedy, K., Shirley, S., Lopez, G., Honeycutt, M., 2015. The
Barnett Shale: from problem formulation to risk management. J. Unconv. Oil
Gas Resour. 11, 95e110.
Felter, S.P., Dourson, M.L., 1998. The inexact science of risk assessment (and im-
plications for risk management). Hum. Ecol. Risk Assess. 2, 245e251.
Gentry, P.R., Covington, T.R., Clewell 3rd, H.J., Anderson, M.E., 2003. Application of a
 M.L. Dourson et al. / Regulatory Toxicology and Pharmacology 80 (2016) 125e133
physiologically based pharmacokinetic model for reference dose and reference
concentration estimation for acetone. J. Toxicol. Environ. Health A 66 (23),
2209e2225.
Guth, D.J., Carroll, R.J., Simpson, D.G., Zhao, H., 1997. Categorical regression analysis
of actue exposure to tetrachloroethylene. Risk Anal. 17, 321e332.
Haber, L.T., Dollarhide, J.S., Maier, A., Dourson, M.L., 2001. Noncancer risk assess-
ment: principles and practice in environmental and occupational settings. In:
Bingham, E., Cohrssen, B., Powell, C.H. (Eds.), Patty’s Toxicology, fifth ed. Wiley
and Sons, Inc., pp. 169e232
Haber, L.T., Maier, A., Gentry, P.R., Clewell, H.J., Dourson, M.L., 2002. Genetic poly-
morphisms in assessing inter-individual variability in delivered dose. Reg.
Toxicol. Pharmacol. 35, 177e197.
Hardin, B.D., Kelman, B.J., Brent, R.L., 2004. Trichloroethylene and cardiac malfor-
mations. Environ. Health Perspect. 112 (11), A607eA608.
Hood, R., 2011. Developmental and Reproductive Toxicology: a Practical Approach,
third ed.
IPCS (International Programme on Chemical Safety), 2005. Chemical-specific
Adjustment Factors for Interspecies Differences and Human Variability: Guid-
ance Document for Use of Data in Dose/concentration-response Assessment.
Available at: http://whqlibdoc.who.int/publications/2005/9241546786_eng.pdf.
Jarabek, A.M., 1995a. Interspecies extrapolation based on mechanistic determinants
of chemical disposition. J. Hum. Ecol. Risk Assess. 1 (5), 641e662.
Jarabek, A.M., 1995b. The application of dosimetry models to identify key processes
and parameters for default dose response assessment approaches. Toxicol. Lett.
79, 171e184.
Johnson, P., Goldberg, S., Mays, M., Dawson, B., 2003. Threshold of trichloroethylene
contamination in maternal drinking waters affecting fetal heart development in
the rat. Environ. Health Perspect. 111, 289e292.
Johnson, P.D., Dawson, B.V., Goldberg, S.J., Mays, M.S., 2004. Trichloroethylene:
Johnson et al.’s response. Environ. Health Perspect. 112 (11), A608eA609.
Kalberlah, F., Schneider, K., 1998. Quantification of Extrapolation Factors. Federal
Environmental Agency, Germany. Final report of the research project No. 116 06
113.
Keil, D., Peden-Adams, M., Wallace, S., Ruiz, P., Gilkeson, G., 2009. Assessment of
trichloroethylene (TCE) exposure in murine strains genetically-prone and non-
prone to develop autoimmune disease. J. Environ. Sci. Health A Tox. Hazard
Subst. Environ. Eng. 44, 443e453.
Marcela, S.G., Monsalve, M.C.R., Garibay, M.A.P., Martinez, M.A., Diaz-Cintra, S., De
La Rosa-Santander, P., Bladimir, R.-R., Gomez, C.S., 2012. Chronological and
133
morphological study of heart development in the rat. Anat. Rec. 295,
1267e1290.
Martin, R.J., Fanaroff, A.A., Walsh, M.C., 2002. Diseases of the Fetus and Infant, ninth
ed. Fanaroff and Martin’s Neonatal-Perinatal Medicine.
Meek, M., Renwick, A., Ohanian, E., Dourson, M., Lake, B., Naumann, B., Vu, V., 2001.
Guidelines for application of compound specific adjustment factors (CSAF) in
dose/concentration response assessment. Comments Toxicol. 7 (5e6), 575e590.
Meek (Bette), M.E., Palermo, C.M., Bachman, A.N., North, C.M., Lewis, J.R., 2014.
Mode of action human relevance (species concordance) framework: evolution
of the Bradford Hill considerations and comparative analysis of weight of evi-
dence. J. Appl. Toxicol. 34 (6), 595e606.
NAS (National Academy of Science), 2009. Science and Decisions: Advancing Risk
Assessment. Available at: http://www.nap.edu/catalog/12209/science-and-
decisions-advancing-risk-assessment.
NAS (National Academy of Science), 2014. Review of EPA’s Integrated Risk Infor-
mation System (IRIS) Process. National Academy of Sciences Press. BN 978-0-
309-30414-6. Available at: http://www.nap.edu/catalog.php?record_id¼18764.
NTP (National Toxicology Program), 1988. Toxicology and Carcinogenesis Studies of
Trichloroethylene (CAS No. 79-01-6) in Four Strains of Rats (ACI, August,
Marshall, Osborne-Mendel) (Gavage Studies). Research Triangle Park, NC: U.S.
Department of Health and Human Services, Public Health Service, National
Institutes of Health. Available at: http://ntp.niehs.nih.gov/ntp/htdocs/LT_rpts/
tr273.pdf.
Renwick, A.G., 1991. Safety factors and establishment of acceptable daily intake.
Food Add. Contam. 8 (2), 135e149.
Renwick, A.G., 1993. Data derived safety factors for the evaluation of food additives
and environmental contaminants. Food. Addit. Contam. 10 (3), 275e305.
Renwick, A.G., Walker, R., 1993. An analysis of the risk of exceeding the acceptable
or tolerable daily intake. Regul. Toxicol. Pharmacol. 18, 463e480.
Schleich, J.-M., 2002. Miscellanea: development of the human heart: days 15e21.
Heart 87, 487.
Seed, J., Carney, E.W., Corley, R.A., Crofton, K.M., DeSesso, J.M., Foster, P.M.D.,
Kavlock, R., Kimmel, G., Klaunig, J., Meek, M.E., Preston, R.J., Slikker Jr., W.,
Tabacova, S., Williams, G.M., Wiltse, J., Zoeller, R.T., Fenner-Crisp, P., Patton, D.E.,
2005. Overview: using mode of action and life stage information to evaluate the
human relevance of animal toxicity data. Crit. Rev. Toxicol. 35 (8e9), 664e672.
Swartout, J., Price, P., Dourson, M., Carlson-Lynch, H., Keenan, R., 1998.
A probabilistic framework for the reference dose. Risk Anal. 18 (3), 271e282.
 Guidance for Contaminated Sites:
Trichloroethylene (TCE) Risk Assessment Case Study
 Executive Summary

The overall purpose of this effort was to develop a range in noncancer risks, similar to the
range used for cancer risks in management of waste sites. This range would enable
noncancer risk to be evaluated during site decisions in a way that reflects the uncertainty
of the noncancer benchmark. Based on readily available information from U.S. Environmental
Protection Agency (EPA) and elsewhere, we adopted a general approach to develop a range for
noncancer risk values, such as Reference Concentrations (RfCs). The recent evaluation of EPA’s
RfC for trichloroethylene (TCE) was used as a case study. The confidence in this range can be
considered in the determination of risk management choices.
For TCE, this range was judged to be 3 μg/m3 to 20 μg/m3. The results of the NTP study-based
RfC were used to determine the floor and midpoint of this uncertainty range. The highly
controversial results from the Johnson et al. (2003) study-based RfC, while associated with low
confidence, were nevertheless used to determine the ceiling level of this uncertainty range. This
3 μg/m3 to 20 μg/m3 range was entirely within the wider individual uncertainty range from the
Keil et al. (2009) study; therefore, this latter study was considered to be confirmatory.
Toxicologists are not able to distinguish the absence of health risk between any two or more
values within this overall uncertainty range of 3 µg/m3 to 20 µg/m3. That is to say, toxicologists
cannot differentiate the “safety” of 4 µg/m3 versus 17 µg/m3. Because of this, managers may use
different values within the multi-endpoint uncertainty range of 3 µg/m3 to 20 µg/m3, along with
site-specific exposure assessments, and other site considerations to make different management
decisions on a case-by-case basis.
For example, when a site-specific exposure assessment defines a range of exposures that are
primarily below the multi-endpoint uncertainty range of 3 µg/m3 to 20 µg/m3, then the
probability of inducing any noncancer effects in the exposure population is lower and the priority
for any management action is reduced (see ES Figure 1a). In this case, a risk manager may
decide to take no action, or to delay action pending further information.
In contrast, when the exposure assessment defines a range in exposures that is primarily above
the multi-endpoint uncertainty range of 3 µg/m3 to 20 µg/m3, then the probability of inducing
noncancer effects in the exposure population is higher and the priority for risk management
action is increased (see ES Figure 1c). In this case, a risk manager may decide to take action, or
to ask for specific information that would refine the estimates of health risk and/or exposure.
When the exposure assessment defines a range in exposures that are primarily in the multi-
endpoint uncertainty range of 3 µg/m3 to 20 µg/m3, then risk managers can use the intermediate
value in this uncertainty range (i.e., 9 µg/m3) and other site considerations, to gauge whether a
management action is needed or if further information should be sought (see ES Figure 1b).
The multi-endpoint uncertainty range can also be used to develop a comparable range of hazard
quotient estimates for single-chemical exposures to TCE alone, or hazard index estimates for
exposures to mixtures of TCE and other chemicals (e.g., solvent impurities such as 1,1,1-
trichloroethane, 1,2-dichloroethane, 1,1-dichloroethene; or degradation products, such as cis-1,2-
dichloroethene and vinyl chloride), when only a point estimate of exposure is available for
comparison. This adaptation is consistent with EPA’s definition of the RfD/RfC, and is akin to

8/22/13 2
the range developed for cancer risk (i.e., an excess lifetime cancer risk of 10-6 to 10-4). When
considering exposures to mixtures, risk managers should be cognizant of the target organ(s),
mode(s) of action and mechanism(s) of action of the various chemicals in the mixture, based on
critical effects and other effects that may be elicited at environmentally relevant concentrations.
Additionally, the role of co-exposures and interactions of chemicals may be considered and/or
ruled out in developing the range in the hazard index for actions that are recommended by risk
managers.
The multi-endpoint uncertainty range is composed of floor, midpoint and ceiling values for
which the appropriate averaging time corresponds to different exposure durations (i.e.,
developmental or chronic exposure periods). Therefore, this range can be applied to both long-
and short-term exposures, with the associated differences in exposure averaging times. For
shorter-term exposures, the results from the Johnson et al. study (2003) might also be used to
describe the best averaging time, but if so, this averaging time should be based on the average
time of cardiac development in humans during fetal growth, approximately 24 days. The use of
a 24-day average time for cardiac development in humans is consistent with the fact that the
dosing in the Johnson et al. study (2003) occurred during the whole time of cardiac development
in the rat.
During the course of this study, it became apparent that EPA’s use of cardiac anomalies for
developing an RfC is highly controversial and not universally shared among government
agencies and experts bodies. It may be that an analysis of this specific effect is warranted by
appropriate experts, but at the very least risk managers need to understand that this endpoint
cannot be used with confidence.

8/22/13 3
 ES Figure 1a. Exposure distribution of indoor air concentrations
primarily below the 3 µg/m3 to 20 µg/m3 hazard range. Relatively small
proportion of exposures is higher than 3 µg/m3. Nominal actions or no
further action may be warranted for risk management.

ES Figure 1b. Exposure distribution of indoor air concentrations falling

within the 3 µg/m3 to 20 µg/m3 hazard range. Relatively small proportion
of exposures is higher than 9 µg/m3. Limited action may be warranted for
risk management

ES Figure 1c. Exposure distribution of indoor air concentrations primarily

above the 3 µg/m3 to 20 µg/m3 hazard range. Actions to reduce exposures
may be warranted for risk management.

Introduction
Risk managers responsible for making choices about acceptable oral and inhalation exposures at
hazardous waste clean-up sites may benefit from an in-depth understanding of the imprecision
and uncertainty in the process used to develop non-cancer toxicity criteria, specifically, non-
cancer Reference Doses (RfDs) and Reference Concentrations (RfCs). This understanding will
allow a higher level of confidence in decisions regarding site screening and closure, with respect
to the evaluation of the non-cancer effects of chemicals present at the site. Although screening
and closure exposure levels based upon the non-cancer endpoint (i.e., “non-cancer driven”) are
now common, this was generally not the case in the past. Formerly, cancer risk was typically the
driver in site decisions for any chemical evaluated with respect to both cancer risk and non-
cancer hazard, particularly when risk management decisions emphasized the lower end of the
excess lifetime cancer risk range. As risk managers became increasingly familiar with how to
effectively manage the cancer risks associated with site chemicals, it is now routine to make
screening and closure exposure level decisions using the widely-accepted 100 fold, cancer risk
range (i.e., 10-4 to 10-6).
However, recent experience indicates that risk managers must contend with situations where risk
management decisions at sites are driven by non-cancer endpoints. Recently, the RfD and RfC
values for some chemicals (e.g., trichloroethylene, or TCE) have been revised to lower levels
(i.e., indicative of greater toxicity), thereby becoming the subject of much attention. For

8/22/13 4
example, the RfD and RfC values for TCE published in September 2011 (U.S. EPA, 2011a) are
lower than the values previously in use. The lower RfD and RfC values resulted from the
evaluation of critical effects not previously quantified, and the use of dose modeling that enabled
the estimation of effects resulting from low-dose exposures (i.e., at levels between the lowest
observed adverse effect level (LOAEL) and the no observed adverse effect level (NOAEL).
The hazard quotient (HQ) is a quantitative indicator of the magnitude of hazard associated with
an exposure to a chemical by a receptor, based on non-cancer effects. The HQ may be defined as
the ratio of the exposure dose to the RfD, or the ratio of the exposure concentration to the RfC.
Multiple exposures (by multiple routes of exposure and/or chemicals) are characterized by a
hazard index (HI). When one or more non-cancer endpoints drive risk management decisions for
chemical exposures, risk managers generally apply a point value (i.e., an HQ = 1) as the criterion
of safety; risk managers do not generally apply a range of acceptable HQ values (i.e., a “hazard
range”). This conventional practice is in contrast to risk management decisions driven by the
cancer endpoint, wherein such decisions are made on the basis of the cancer risk range described
above. Although U.S. EPA (Integrated Risk Information System) has defined the RfD/RfC as
possessing “uncertainty spanning perhaps an order of magnitude”, risk managers have generally
not implemented decisions based upon this implicit uncertainty. Consequently, non-cancer
hazards have frequently been evaluated and regulated with sort of an “off/on switch” that is at, or
very near, the RfD/RfC, wherein the hazard quotient of one (1) is interpreted as a “bright line”
for risk management decision-making.
Risk management tools that incorporate the uncertainty associated with each RfD/RfC value
have not been widely implemented. The statement “uncertainty spanning perhaps an order of
magnitude” is sometimes interpreted as a precautionary statement addressing the implicit health-
protectiveness of the RfD/RfC value. This interpretation is perhaps an intuitive one, because the
derivation process is considered so uncertain. This perspective heightens the health concerns
associated with exposures exceeding a hazard quotient of one. However, the process by which
uncertainty is accounted for in the derivation of the RfD/RfC is implicitly precautionary; the
application of an uncertainty factor provides a margin of safety in response to each source of
uncertainty. Each uncertainty factor (greater than one) serves to reduce the RfD/RfC by
increasing the gap between the RfD/RfC and the protective value (i.e., the “safe dose” based on
experimental observation, dose adjustment and toxicokinetic extrapolation). Most
determinations of the RfD/RfC have more than one source of uncertainty, and consequently
exhibit multiple successive precautionary adjustments (i.e., the application of multiple
uncertainty factors). The use of multiple sequential uncertainty factors significantly increases
the margin of safety and the implicit protectiveness of the RfD/RfC.
It is not surprising that many risk mangers view the evaluation of non-cancer hazards as a “bright
line”. Throughout most of the 1980s and 1990s, the NOAEL (the presumptive “safe dose”
below the “effect threshold”) for the critical effect served as the “point of departure” (POD) for
the derivation of the RfD/RfC by the application of one or more uncertainty factors (each
uncertainty factor generally equal to three or ten). In the past 15 years, however, the RfD/RfC
has increasingly been based upon sophisticated models, which define internal concentrations
from exposure dosing, account for toxicokinetic differences between the test animal and human
populations, perform route-to-route internal dose extrapolations, and interpolate effects levels
between the LOAEL and NOAEL in the experimental study. These probability-based elements

8/22/13 5
are now commonly incorporated into the RfD/RfC derivation process, thereby uncoupling the
POD from the threshold dose (or, at least, obscuring the association between them).
In this paper, the authors propose the development of a “hazard range” as a risk management tool
to facilitate informed decision-making about exposures and likely effects to humans (including
sensitive sub-populations). This hazard range accounts for the imprecision and uncertainty
implicit in the derivation of the RfD/RfC value. The lower end of the range, commonly referred
to as the “floor”, is the RfD/RfC. A distinguishable upper end of the range is established,
generally associated with a specific toxic effect level. Thus, the lower and upper ends of the
hazard range can be clearly established from information available on the derivation of the
RfD/RfC. Risk management decisions about exposure levels wholly above or below the range
may be straightforward. However, risk management decisions about exposure levels wholly or
partially within the range are more complex; therefore, it is helpful to have a “midpoint” within
the range to aid in these decisions.
The use of many uncertainty factors to derive the RfD/RfC, and the inherent imprecision of the
RfD/RfC value, each contribute to how the midpoint is established. These considerations include
the level of confidence in the dose-response curve, and the completeness and representativeness
of the animal study(ies) used as the basis of the RfD/RfC derivation. Thus, a comprehensive
analysis of the imprecision and uncertainty used in the derivation process is necessary and results
in a hazard range with a protective floor, a midpoint, and an effects level ceiling.
How a responsible risk manager addresses exposure levels within that range is dependent on both
the imprecision and uncertainty analysis, and two additional risk management evaluations. The
first evaluation takes a holistic look at the margin of safety applied in response to each source of
uncertainty in the RfD/RfC derivation process. This risk management evaluation is often more
semi-qualitative, relying on an understanding of how three-fold or ten-fold adjustments (i.e.,
equivalent to uncertainty factors of one-half or one full order of magnitude) were applied and
how these adjustments impact the level of confidence in the “health effects level” and the
RfD/RfC derived from it. The second risk management evaluation addresses the sum of the
animal and human evidence linking the toxic effect to the exposure. Here consideration of the
strength of epidemiology evidence, repeatability of the animal results and the general quality of
the studies and extrapolations used are examined for important strengths and weaknesses that
may affect level of confidence in decisions.
Responsible risk management considers both health protection and practical implementation; an
enhanced understanding of the inherent imprecision, uncertainty and margin of safety built into
the RfD/RfC facilitates and enables responsible risk management. In this paper, the authors have
proposed a method for determining a hazard range to inform such an understanding. The
establishment of a hazard range for the RfC for TCE has been used as an example.
Risk management concerns resulting from widespread TCE exposure are considerable. Its use as
a common degreaser and its prevalence as a breakdown product of the dry cleaner agent
tetrachloroethylene (TCE) have resulted in widespread exposure. The volatility of TCE creates
an indoor air inhalation concern via volatilization from soil or groundwater through the
subsurface. Common disposal practices with older dry-cleaning facilities, automotive repair
shops and machining operations have resulted in groundwater and soil contamination, and
potential inhalation exposure to indoor air at many commercial and industrial settings.

8/22/13 6
Additionally, an understanding of non-cancer hazard is important because common household
products have historically contained TCE. This results in indoor air “background” in about half
the residential structures (Dawson and McAlary, 2009). The indoor air background levels may
approximate the RfC in many structures and may confound or complicate determinations of
whether subsurface contamination is the source of TCE in indoor air through vapor intrusion.
These important variables regarding potential and actual human exposures amplifies the
importance of a clearer understanding of the hazards associated with environmental
concentrations above the RfC.
The analysis of TCE is complicated by the fact that TCE has both long term and short term
exposure concerns. The chronic RfC was determined using an average of two critical effects
studies (Keil et al., 2009; Johnson et al., 2003), with support from a third studies (NTP, 1988),
all of which were incorporated into a range of candidate RfC values that are reasonably close in
magnitude. Chronic exposure concerns exist for developmental, immune system and kidney
toxic effects. Based on U.S. EPA’s use of one study (which examined exposure across the 21-
day fetal developmental period in the rat and identified developmental fetal heart
malformations), the RfC incorporates effects based on short term exposures. Thus, we examine
the hazard associated with both short term and chronic exposures to environmental
concentrations above the RfC.
To provide guidance on the practical understanding of the uncertainty and other non-cancer
issues associated with TCE, three charge questions were formulated and addressed for this
Alliance for Risk Assessment (ARA) project (http://www.allianceforrisk.org/Projects/TCE.html):
1. Develop additional risk assessment guidance on how to interpret the non-cancer endpoint
when it is used for deciding clean-up standards or acceptable exposure levels when
closing sites.
2. Clarify the issues surrounding the potential developmental cardiac malformations for use
in understanding clean-up standards and short term exposure levels.
3. Explore the margin of safety measures used to set the TCE RfC and evaluate if these
measures are consistent with the baseline principles developed for determining the RfC.
The following 3 sections of text give a general, and sometimes overlapping, response to each of
these questions. Section 4 is a synthesis of significant information from the first 3 sections for
use in risk management decisions.

Section 1:
Additional Risk Assessment Guidance On How To Interpret The Non-Cancer Endpoint

Trichloroethylene Non-Cancer Exposure

The evaluation of low-level inhalation exposures to TCE has become a subject of increased
concern for regulators and risk managers for contaminated site managers over the past year. This
concern has been largely prompted by the final TCE assessment conducted by the U.S.

8/22/13 7
Environmental Protection Agency (U.S. EPA, 2011a) through its Integrated Risk Information
System (IRIS) database. A Reference Concentration (RfC) for TCE of 0.002 mg/m3 (i.e., 2.0
µg/m3) was posted on IRIS on September 28, 2011.
Prior to the final TCE assessment regulators either relied upon other sources for an estimate of
the safe level for chronic or lifetime human exposures (e.g., the provisional reference
concentration, RfC, values of 40 µg/m3 from EPA’s National Center for Environmental
Assessment [U.S. EPA, 2001] and the Chronic Inhalation Reference Exposure Level [REL] of
600 µg/m3 from California EPA’s Office of Environmental Health Hazard Assessment
[OEHHA, 2007]), or did not quantitatively evaluate inhalation exposures with respect to non-
cancer human health endpoints. Consequently, regulatory and public health agencies often
evaluated TCE exposures on the basis of the cancer endpoint only. Even when TCE exposures
were evaluated on the basis of both non-cancer and cancer endpoints, medium-specific
concentrations1 were typically based upon the cancer endpoint, particularly if the acceptable
excess lifetime cancer risk (ELCR) was established at the lower end of the acceptable cancer risk
range (i.e., 1 x 10-4 to 1 x 10-6).
A comparison of the California EPA REL (600 µg/m3) and the IRIS RfC (2 µg/m3) shows that
the former value is 300 times greater than the latter. Alternatively stated, the estimate of the safe
level of chronic inhalation exposures from IRIS is two and one-half orders of magnitude lower
than the analogous estimate from OEHHA. An evaluation of TCE exposures using the final IRIS
RfC in lieu of the OEHHA REL would result in an estimate of non-cancer hazard (i.e., the
hazard quotient, or HQ) that is 300 times greater.
As a result of this change in the RfC value, regulatory and public health agencies using an
acceptable ELCR greater than 1 x 10-6 for the development of medium-specific concentrations
(e.g., ambient air, sub-slab vapor or soil gas) will generally find the non-cancer endpoint (i.e., the
RfC) driving these values; at an ELCR of 1 x 10-6 (i.e., a lower level of acceptable cancer risk), it
is likely that the screening level or RAO will be based upon the inhalation unit risk. For
example, the U.S. EPA RSL for residential air is 0.43 µg/m3, based on the cancer endpoint and a
target ELCR of 1 x 10-6; a value that is lower than the RfC of 2 µg/m3. A residential air
screening level based on the cancer endpoint and a target ELCR of 1 x 10-5 would be 4.3 µg/m3;
a value greater than the RfC of 2 µg/m3. Therefore, the acceptable ELCR that regulatory and
public health agencies adopt will determine which endpoint drives the assessment, particularly
when as single numerical point value of ELCR is identified as the target ELCR for the
calculation of a medium-specific concentration.
The selection of a numerical ELCR within the acceptable risk goal range may be established by
statute or regulation, or may be selected on a case-by-case basis by risk managers. Risk
managers with regulatory and public health agencies are generally experienced with the selection
of one or more target ELCR values within the acceptable risk goal range. If, for example, the
calculation of a medium-specific RAO based on an ELCR of 1 x 10-6 proved to be problematic
(e.g., the proposed concentration was below laboratory quantitation limits, near or below
background levels, or too impractical to achieve), the risk manager could elect to base the RAO
on a higher target ELCR within the acceptable risk goal range.

1
For example, screening levels concentrations above which further investigation is justified; or remedial
action objectives [RAOs], concentrations above which a remedial action is justified.

8/22/13 8
However, non-cancer assessments have typically been conducted with a fixed target HQ of one.
As a consequence, risk managers have not exercised the same flexibility with respect to the
establishment of medium-specific concentrations based upon the non-cancer endpoint (i.e., a
target HQ of one). This has proven problematic with respect to the setting of screening levels,
exposure limits or RAOs for TCE. Risk managers have been compelled to establish medium-
specific concentrations based on an HQ of one, or even more conservatively, establish the RfC
value itself as the medium-specific concentration. To do otherwise, risk managers would need a
comprehensive understanding of the sources of uncertainty associated with the derivation of the
medium-specific concentration, and consequently, the relevance of the hazards posed from
exposures to concentrations exceeding the RfC.
Three sources of uncertainty associated with implementation of a medium-specific concentration
(e.g., an indoor air exposure limit for TCE) may be identified:
1. the uncertainty associated with the derivation of the RfC for TCE;
2. the uncertainty associated with the hazard estimate based on the RfC for TCE;
3. the uncertainty associated with the determination of the exposure concentration; and
4. the uncertainty associated with the determination of the exposure concentration with
specific regard to the RfC for TCE.
Each of these sources of uncertainty is discussed below.

Uncertainty Associated with the Derivation of the RfC

Understanding the risk of exposures to TCE at concentrations above the RfC will require a
comprehensive understanding of the science and science policy used to set the RfC. This science
and science policy evaluation is best understood if one divides the issues into two interdependent
but distinct perspectives. The first perspective is toxicology-based, and examines the process
and methods used to extrapolate from the animal results to safe human exposure levels. The
second perspective focuses more on understanding the actual assessment of risk, how risk is
measured and the uncertainty and precision with which numerical expressions of risk can be
practically applied.
Examining the toxicology science and science policy used to develop the RfC is generally
beyond the scope of most regulatory risk managers’ capabilities. The complexity of the process
by which the RfC was derived from experimental data from animal assays (including benchmark
dose derivation, allometric scaling and physiologically based pharmacokinetic modeling for
equivalent human exposure concentration) precludes most risk managers from being able to
make confident decisions regarding the human health effects that may be expected at TCE
concentrations above the RfC.
In the absence of a toxicology analysis, risk managers should approach an understanding of the
exposure from the risk assessment/management side. If a risk based analysis is made, then most
risk managers will find some flexibility for making risk based decisions. First, risk managers

8/22/13 9
should attempt to characterize and have some practical understanding of the range of imprecision
and uncertainty in the RfC estimate. This understanding begins with characterizing how precise,
or more accurately, how “imprecise” is the numerical value of the RfC (and any screening level
that is calculated on the basis of the RfC) that the risk manager is using to assess risk. Numbers
are either exact or an approximation. Approximate numbers have some degree of uncertainty
associated with them and precision, generally expressed as significant figures, is commonly used
to determine when to truncate numerical values (e.g., the RfC or any screening value calculated
on the basis of the RfC) (U.S. EPA, 2012b). In general, when performing calculations with
approximate numbers, one uses as many digits as possible in calculations until the final result.
U.S. EPA considers the precision of a hazard estimate to be one significant figure (U.S. EPA,
1989, Exhibit 8-4, p. 8-9). When determining the final result the rules of truncating and
rounding would apply (U.S. EPA, 2012b). Accordingly, any representation of the hazard index
(HI) should be one significant figure. For example, an HI calculated to be 0.94 would be
rounded to 0.9, and HI calculated to be 1.6, the HI estimate would be rounded to 2.

Uncertainty Associated with the Estimates of Hazard Based on the RfC

The U.S. EPA (2012b) generally represents the RfC as a value rounded to one significant figure.
Similarly, U.S. EPA (2012a) defines the use of the HI [or HQ] as rounded to one significant
figure. Thus, the precision of any hazard estimate, including any screening level or RAO derived
from a target HQ, is limited to one significant figure. In practical application, the hazards
associated with an HQ of 0.95 and an HQ of 1.49, and any value which lies between them, is
represented as an HQ of 1 (one); differences in hazard between any two estimates within this
range cannot be discerned. U.S. EPA (1993) acknowledges and demonstrates practical
application of this imprecision when defining the risk in a Site-Specific Record of Decision from
two separate HQs of 0.7 and 1.2; here, U.S. EPA (1993, Appendix B) defines both hazard
estimates as essentially equal to an HQ of 1 (one), stating the reason is the imprecision in the
reference doses for nickel and copper, respectively.
The uncertainty must also be considered. A major portion of uncertainty involved in calculating
a HQ is imprecision, but uncertainty enters into any evaluation when one considers that certainty
is true knowledge about the meaning of the data and this is rarely the case when U.S. EPA
determines a value for the RfC. U.S. EPA (IRIS, 2012) defines the TCE RfC as “possibly
having up to” an order of magnitude uncertainty. There are many ways to characterize an order
of magnitude of uncertainty. Guidance is taken from U.S. EPA citing that the order of
magnitude is “centered” on the RfC (U.S. EPA, 2002). Further guidance on how to apply the
centered order of magnitude is taken from Felter and Dourson (1998), who state many on the
EPA RfD/RfC workgroup believe the range of uncertainty amounts to one-half an order of
magnitude on either side of the RfC.
Therefore, the uncertainty and imprecision in the RfC can vary from one significant figure up to
an order of magnitude. This imprecision and uncertainty must be defined relative to the issues of
most concern to the risk manager as demonstrated by the following:
1. U.S. EPA screening levels are not RAOs (Regional Screening Level Table, 2012) but are
meant to define exposures under which further investigation is warranted. Therefore

8/22/13 10
 screening levels are not meant to define effects level, e.g., levels at which a toxic effect is
expected to occur. Screening levels are meant to define a bright line for further
investigation. Defining a bright line for further investigation is much different than
defining a bright line for actual health effects. Thus, the risk managers’ question
becomes: at what exposure level above screening does the risk manager become
concerned with a significant or measurable increase in potential for a toxic effect?
2. The use of the RfC term in various equations to determine screening levels or assess the
risk of exposure must consider the precision with which such equations or methods can
be applied and understood. U.S. EPA’s designation for the RfC as rounded to one
significant figure (IRIS, 2012) is a clear acknowledgement of the limitations implicit in
the precision of the RfC. Therefore, it is not possible to determine the difference in
hazard between an HQ of 0.95 and an HQ of 1.49, both of which would round to 1 (one).
3. RAOs are distinct from screening levels. RAOs should therefore be indicative of
concentrations that are protective of human health, i.e., concentrations that do not pose a
“measurable and quantifiable hazard or risk”. The RAOs (like screening levels) should
be below effects levels, although RAOs may be at concentrations that are above the
screening level. If the risk manager cannot demonstrate that a concentration exceeding a
screening poses a measurable and quantifiable hazard or risk, then there would be no
need for remedial action.

Thus, due to the imprecision of the methods used to develop the RfC, the risk manager cannot
reliably show a measurable risk-based difference between an HQ value of 0.95 and 1.49. If this
were true, then the difference in hazard estimates at levels of precision beyond the single
significant figure has little meaning. Using the common equation for screening and remedial
action (RSLTs, 2012) and calculating exposure levels associated with a target HQ (THQ) of 1 up
to 1.49 (which rounds to 1) yields:

( )

( ) ( )

where THQ = 1.0, Screening Level = 2.1 µg/m3, or 2 µg/m3 when rounded to one significant
figure; and
where THQ = 1.49, Screening Level = 3.1 µg/m3 , or 3 µg/m3 when rounded to one significant
figure.

8/22/13 11
Thus, it is not possible to discern a discrete difference in the hazard associated with an air
concentration of 2 µg/m3and an air concentration of 3 µg/m3. A similar analysis can be
undertaken by calculating the order of magnitude risk range centered on the RfC. While there
can be different methods used to assess “half an order of magnitude” around the RfC, the use of
the following, after Felter and Dourson (1998), seems the most reasonable and is expressed using
the practical methods of assessing exposure in µg/m3 as:

, rounding to one significant figure, and a factor of 3

Thus, the RfC is simply the center that sets the boundaries of the order of magnitude range. U.S.
EPA (2002) states how the range should be applied:
“….. [The] reference value is intended to provide an estimate that is centered within an order of
magnitude, further emphasizing that the estimate is not a bright line, but has some range of
variability that may be considered by risk managers in decision making.”
Recognizing there is some flexibility, the risk manager is still in the position of trying to
understand what the range of uncertainty around the RfC means in practical reality. Here is
where information is used on how the range is “weighted.” That is, if the magnitude of the
composite uncertainty factor used to determine the RfC is large, then it is likely that the hazard
associated with toxic effects above the RfC would shift to the high side of the RfC range.
Similarly, use of a composite uncertainty factor that is somewhat marginal would indicate the
risk of toxic effects could shift to the central RfC value or the low side of the range.
Toxicology Excellence for Risk Assessment (TERA, 2012) clarifies this concept stating:
“Moreover, the precision of the RfD or RfC depends in part on the overall magnitude of the
composite uncertainty and modifying factors used in its calculation. The precision at best is
probably one significant figure and more generally an order of magnitude, base 10. As the
magnitude of this composite factor increases, the estimate becomes even less precise.”
Another supporting perspective can be determined using U.S. EPA guidance on how the RfC for
TCE was determined (U.S. EPA, 2011a). Here U.S. EPA used a set of three studies at the lowest
end of the response spectrum. Three candidate RfC values (i.e., cRfCs) at the lower end of the

8/22/13 12
responses were grouped in the range of 0.0003-0.0006 ppm and were selected to derive and
support the final RfC. The grouping establishes a narrow range of candidate RfC values. U.S.
EPA used the lower portion of this range of values to set the RfC. U.S. EPA states the reasoning
and justification for this approach (U.S. EPA, 2011a; Section 5.1.5.2, page 5-95):
“One approach to selecting an RfC would be to select the lowest calculated value of 0.0003 ppm
for decreased thymus weight in mice. However, as can be seen in Table 5-24, three p-cRfCs are
in the relatively narrow range of 0.0003–0.0006 ppm at the low end of the overall range. Given
the somewhat imprecise nature of the individual candidate RfC values, and the fact that multiple
effects/studies lead to similar candidate RfC values, the approach taken in this assessment is to
select an RfC supported by multiple effects/studies. The advantages of this approach, which is
only possible when there is a relatively large database of studies/effects and when multiple
candidate values happen to fall within a narrow range at the low end of the overall range, are that
it leads to a more robust RfC (less sensitive to limitations of individual studies) and that it
provides the important characterization that the RfC exposure level is similar for multiple
noncancer effects rather than being based on a sole explicit critical effect.”
IRIS agreed with this interpretation and concluded similarly (U.S. EPA IRIS, 2012; Section
I.B.2). U.S. EPA did not signal that a developmental endpoint was used to set the RfC. Instead,
they cite a group of studies at the low end of the response spectrum and treat them as a group.
The U.S. EPA used 0.0004 ppm from the range of 0.0003 to 0.0006 ppm to set the RfC. Given
adequate site-specific exposure and source information, it is reasonable to use the higher portion
of the range to set remedial objectives. Using the 0.0006 ppm value (i.e., the upper value from
the range of candidate RfC values, or 3.26 µg/m3), and a target HQ of one, a screening level of
3.3 µg/m3 (or 3 µg/m3 when represented at one significant figure) may be calculated by using the
equation presented above. This estimate is equivalent to the screening level calculated on the
basis of a target HQ of 1.49 (discussed above).
An RAO may be calculated as shown below (i.e., by the same equation as the screening levels
were calculated above):

( )

( ) ( )

as an indicator of an air concentration that is protective of human health, i.e., concentrations that
do not pose a “measurable and quantifiable hazard or risk”. As stated above, the RAOs (like
screening levels) should be below effects levels, although RAOs may be at concentrations that
are above the screening level. For purposes of illustration, a value for an RAO may be
calculated using the upper-bound value on the range of uncertainty centered on the RfC (i.e., 6
µg/m3, discussed above) and a target HQ of 1.49 (to account for the implicit precision of the

8/22/13 13
hazard estimate. This calculation would result in an RAO for residential air of 9.3 µg/m3, or 9
µg/m3 when rounded to one significant figure. This analysis suggests that a TCE air
concentration of 9 µg/m3 may not be any less protective of lifetime exposures than the Screening
Level of 2.1 µg/m3, based on both the uncertainty associated with the RfC and the precision of
the target HQ. Therefore, TCE concentrations above the screening level of 2.1 µg/m3 merits
further investigation; this evaluation suggests that. In a particular circumstance, TCE
concentrations as high as 9.3 µg/m3may not require remedial action.
Another way to represent the uncertainty that is implicit in the RfC is to evaluate the alternate
points of departure that may be used to derive the RfC based on the fetal malformation endpoint
(using the BMDL01, BMD01 as alternate points of departure); and the candidate RfC based on the
toxic nephropathy endpoint (using the BMDL05, BMD05 as alternate points of departure). For
each alternate POD, a Screening Level based on the equation above was derived for residential
and industrial exposures. Table 1 summarizes the range of TCE screening levels that are
calculated based on alternate representations of the fetal cardiac malformation and toxic
nephropathy endpoints from section 3, for comparison with indoor air data. The values
presented in Table 1 represent point estimates corresponding to a HQ of 1 and centered on the
order-of-magnitude uncertainty associated with an RfC, and the imprecision associated with
rounding of the HQ value to one significant figure.

8/22/13 14
Table 1. TCE Indoor Air Screening Levels Based on Alternate Representations of
Endpoints

Endpoint Value Value Screening Screening

(ppm) (µg/m3) Level - Level -
Residential Industrial
(µg/m3) (µg/m3)

RfC, fetal malformation endpoint 0.00037 2 2.1 8.7

BMDL, 1% increased risk on the 0.004 21 22.3 93.9

fetal malformation endpoint

BMD, 1% increased risk on the 0.011 59 61.5 258.1

fetal malformation endpoint

Candidate RfC, toxic nephropathy 0.00056 3 3.1 13.1

BMDL, 5% increased risk on toxic 0.014 75 78.2 328.5

nephropathy

BMD, 5% increased risk on toxic 0.02 107 111.7 469.3

nephropathy

In conclusion, what is needed is a clearer toxicological understanding of the range, how it should
be weighted and any other insight toxicology experts can provide for risk managers regarding
exposure levels and risk of a toxic effect above the RfC. However, any understanding of risk
estimates must be interpreted in light of the imprecision and the uncertainty that are a part of the
RfC derivation process. The use of significant figure recommendations and uncertainty
estimates are parameters that risk managers can consider now, in the absence of any other
guidance. At a minimum, without any further toxicology understanding, the differences between
exposure levels for TCE between 2 and 3 µg/m3 are really not measurable and an increase in
calculated risk is really not observable until exposure levels reach at least 3 µg/m3.

Uncertainty Associated with the Determination of the Exposure Concentration: General

Considerations

Inhalation exposures to TCE are a concern in indoor air, particularly when the TCE
concentrations may be largely or partially attributable to vapor intrusion. The indoor air
concentrations of TCE may be predicted by modeling or directly measured by air sampling. Air
sampling reflects all of the factors in the vapor intrusion pathway (source strength, subsurface
characteristics and building conditions), and, of the types of vapor intrusion sampling that can be
performed, most effectively characterizes the variability in potential human exposure inherent in
the vapor intrusion pathway. Air sampling is a more direct indication of potential exposure and
risk to building occupants from VOCs in indoor air arising from vapor intrusion, when compared

8/22/13 15
to modeled indoor air concentrations predicted from sub-slab vapor, soil gas, or groundwater
concentrations. This last point is critical, because the collection of air sampling data often is
required by a regulatory agency to assess risk and make a determination of whether or not a
vapor intrusion pathway in a building is of regulatory concern and warrants action to reduce
exposure.
It is well-recognized that indoor air concentrations and potential exposures vary, whatever
measurement period is chosen and whatever measures are undertaken to keep conditions
constant. This variation is usually much greater than can be accounted for by sampling and
analytical error. Typically, air sampling results peak at a relatively low concentration, with a few
results several times the most frequently detected concentration. This has been observed in
multiple fields such as indoor radon assessment and industrial hygiene investigations. This
pattern of results can also be observed in indoor air sampling data collected from vapor intrusion
sites.
Typical indoor air sampling strategies at vapor intrusion sites typically measure indoor air
concentrations during a few days throughout a year, usually at times selected based on practical
concerns or under assumed worst-case conditions. These worst-case conditions typically occur
under winter conditions, where higher indoor air temperatures produce a stack effect that
enhances advective transport of soil vapor indoors, and where the outdoor air exchange rate is
minimized to reduce heating costs. Sampling may also be performed in the other seasons to
provide some understanding of variability in indoor air concentrations. Evaluation of the data
typically involves comparison of individual measured concentrations with risk-based screening
levels. Reported concentrations higher than the screening level represents a trigger for further
investigation or implementation of mitigation measures; depending on the regulator, a single
result exceeding a screening level may be sufficient to trigger further action.
Statistical evaluation of indoor air sampling data becomes possible when multiple samples can
be collected from a building (or from a group of buildings judged to have similar characteristics).
Statistical methods can provide a more refined assessment of potential exposure than a “worst-
case” evaluation. Performing statistical analysis on air sampling data requires adopting a model
of the distribution of concentrations and associated exposures. Historically, a lognormal
distribution is used to summarize air concentration and exposure data. Lognormal distributions
are generally leptokurtic, or positively skewed (a long “tail” to the right). Presenting indoor air
sampling data in this manner allows the variability in exposures to be combined with dose-
response variability for purposes of risk characterization.
Some data are available regarding the expected variability in indoor air concentrations associated
with vapor intrusion (VI). U.S. EPA has developed a database of VI sampling results,
principally from residential structures and including indoor air sampling data (U.S. EPA, 2012b).
The indoor air sampling data can be used to develop an understanding of between-structure
variability. For example, U.S. EPA’s report states that residential structures at Lowry Air Force
Base exhibit almost two orders of magnitude variability in indoor air concentrations measured
monthly over a period of six months. Folkes et al. (2009) determined temporal variability using
up to 10 years of data at 45 unmitigated structures and concluded that concentrations vary by a
factor of two to three from an annual average concentration. Folkes et al. indicated that seasonal
variations explain a portion of the variability but the majority of the variation at the study site
appeared to be due to factors such as variation in air exchange rates and differential pressure

8/22/13 16
caused by wind, barometric pressure, temperature and the use of exhaust fans, HVAC and open
windows. This is consistent with the conceptual understanding of factors that can impact
variability.

Uncertainty Associated with the Determination of the Exposure Concentration: Specific

Implications for Comparisons of Measured Air Concentrations to the RfC for TCE (or to
Screening Levels or RAOs Based on the RfC)

The RfC for TCE on IRIS was set by U.S. EPA (2011a), in part, using the critical effect of the
Johnson et al. (2003) study. This study described a developmental endpoint of fetal heart
malformations. The critical effect in this study occurred across a short-term exposure period
during gestation from 1-21 days in Sprague-Dawley rats. Because the toxic effect occurs across
a short dosing period, some states (e.g., New Jersey, Massachusetts) and U.S. EPA regions (e.g.,
Region IX, Region X) have developed exposure levels and sampling guidance commensurate
with the short exposure (i.e., developmental) time frame. This has raised many questions and
concerns regarding how to assess indoor air exposure levels (based on the estimate of reasonable
maximum exposure, or RME), given the inherent spatial and temporal variability of indoor air
concentrations. If U.S. EPA and the state environmental and public health agencies are going to
base compliance and remedial decision-making regarding TCE concentrations in indoor air using
a developmental endpoint across a short period of time varying from days to months, then a
conceptual understanding and characterization of short term spatial and temporal variability is
critical.
Given the range of the expected variability in indoor air concentrations, it now becomes
important to design a sampling program that will adequately assess TCE indoor air
concentrations in the relevant time frame (i.e., the exposure concentration). Since the exposure
period used to determine the fetal heart malformation critical effect was 21 days, it seems
reasonable to assess short term variation across a corresponding time window, (i.e., 21 days, for
which average concentrations of a month or less may be reasonably representative). The use of
indoor air samples taken in the main living areas, and spatial averaging may be appropriate for
assessing spatial variability in indoor air concentrations. However, accounting for temporal
variability may pose a much more difficult problem; temporal variability across a 21-30 day time
window is superimposed upon other cycles of temporal variability (e.g., diurnal, seasonal, and
annual variability). Alternatively stated, the indoor air exposure concentration that accounts for
temporal variability across 21-30 days in April may not be representative of the indoor air
exposure concentrations for the same receptor population in the same structure across 21-30 days
in July, October or January. Thus, determination of the indoor air exposure concentration must
account for spatial variability and multiple cycles of temporal variability (i.e., diurnal, seasonal,
annual) superimposed upon the relevant 21 day window.
The common method of assessing variability has been to focus indoor air sampling at times that
are expected to be representative of worst-case conditions. Many regulatory programs assume
worst case conditions occur during cold weather periods when HVAC systems provide a
negative pressurization of indoor air compared to ambient pressures, and this pressure
differential then facilitates the transport of vapor from the subsurface to the indoor air. Other
regulatory programs do not assume cold weather conditions will provide worst case conditions

8/22/13 17
and require sampling during at least two seasons. Generally, this type of sampling entails a
single 24 hour “snapshot” sample which does little to assess the temporal variability associated
with seasonal or annual cycles, unless one assumes that the one-day sampling event was
positively biased (i.e., conducted under expected worst case conditions). The problem with
determining any “variability” in indoor air is that it is difficult to determine when all of the
factors impacting variability will combine to provide a scenario that may be construed to be
“worst case conditions”. Thus, indoor air exposure concentrations to represent the 21-day
exposure period ideally should be based upon multiple sampling events which are representative
of variations in temperature, water table elevation, and building-specific parameters such as
pressure differential and ventilation rate. Adequately accounting for this variability may likely
require that indoor air sampling be conducted on two or more 21-30 day events (e.g., winter,
summer) throughout a single year. As discussed previously, statistical evaluation of indoor air
sampling data becomes possible when multiple samples can be collected from a building (or
from a group of buildings judged to have similar characteristics).
Given the sources of spatial and temporal variability discussed above, there are some possible
approaches to determine exposure levels across a 21-30 day window within a structure include:
1. Collect multiple indoor air 24-hour TO-15 type composite samples randomly selecting
the 24-hour periods within a 21-30 day time window. Then use statistical software to
determine an exposure point concentration for that window.2
2. Use passive diffusion samplers for a 21-30 day time window that provide the “average”
exposure concentration over that window.
3. Use real-time sampling such as an onsite GC or GC/MS recording readings at some
frequency throughout each 24 hour period for a 21-30 day interval and then determine the
UCL as above.
4. Take a one-time 24 hour sample and apply some pre-determined variability factor such as
a factor of three or 10 (note: the concept of a variability factor has a parallel with
seasonal adjustment factors used in indoor radon sampling).
There are advantages and disadvantages to each of the four methods listed above. The first three
approaches are presented with respect to one sampling events; each of these methods would need
to be repeated at least during the year to account for expected seasonal variation. The cost of
multiple 24-hour composite samples is likely to be prohibitive in many applications. The use of
passive or active samplers does not distinguish between indoor air vapor intrusion and
background contributions. As noted by U.S. EPA (2011b) indoor air background can often be
similar in concentration to the regulatory action level. Cost of real-time on-site GC or GC/MS
sampling may also be prohibitive, however, it should be noted that this method has the real
advantage of identifying peaks and also providing valuable information about potential indoor air
background. It appears the most cost effective way to address temporal variability may be to use

2
For example, U.S. EPA Pro-UCL (U.S. EPA, 2013) can be used to determine the 95% upper confidence
limit on the mean concentration (i.e., the 95% UCL, representative of the reasonable maximum exposure
(RME) concentration, as defined by U.S. EPA, 1989). Other software packages also may be used to
develop a representative concentration.

8/22/13 18
passive indoor air samplers over a few 21-30 day periods throughout the year. The objective of
this approach is to approximate the worst case 21-day indoor air exposure concentration and
account for temporal variability.
There are continuing efforts addressing the questions about how to best address variability when
devising strategies for assessing vapor intrusion. A recent EPA workshop, “Looking Beyond
Natural Variation in Vapor Intrusion: Understanding, Controlling and Addressing Site Variables
for Improved Practices and Effective Protection Strategies”
(https://iavi.rti.org/WorkshopsAndConferences.cfm), was convened in March 2013 to examine
different aspects of spatial and temporal variability based on research in single buildings and
studies of data from multiple buildings and sites. Efforts such as this workshop will contribute to
evolving best practices for sampling and investigating indoor air at vapor intrusion sites.

Section 2:
Issues Surrounding The Potential Developmental Cardiac Malformations
Risk Management Considerations for the Evaluation of a Developmental Toxic Effect for
TCE at Exposure Levels above the RfC

The intent of this section is to gain a broader understanding of the level of confidence in the
weight of evidence linking TCE exposure with fetal heart malformations and the overall margin
of safety built into the process that defined the RfC. This information will provide a more
qualitative understanding of the risk from various TCE exposure levels and provide a higher
level of confidence when making site-specific decisions on acceptable exposure levels.
U.S. EPA (2011a) evaluated a large number of studies for derivation of the RfC. For each study
they developed a “candidate RfC”, as if each study were being used as the principal RfC study.
Three candidate RfCs were close in magnitude and as a group, well below other candidate RfCs.
The low grouped studies were the Johnson et al., (2003); Keil et al., (2009); and, NTP , 1988
studies. EPA (2011a) then decided two studies were the main critical effects studies (Johnson et
al., 2003 and Keil et al., 2009) and used the average of the two main critical effects studies to
derive the RfC, citing the third study (NTP, 1988) as support for the two main critical effects
studies.
One critical effects study was linked with developmental effects (i.e., fetal heart malformations
as described by Johnson, et al., 2003). The other two studies were longer-term (Keil et al., 2009)
and chronic (NTP, 1988) exposure studies and the critical effects, decreased thymus weight and
kidney toxicity, were associated with long-term exposure. Although no short term or
developmental effects were reported in either the Keil et al. or the NTP studies, short-term
exposures are of concern with respect to the developmental effect reported by Johnson et al.
(2003). The nature of this toxic effect is serious and many risk managers nationally have
become concerned about the risk for fetal heart malformations at exposures above the RfC.
There are two major methods used to evaluate the risk of a developmental effect. The first
method is quantitative and addresses the magnitude of acceptable exposure levels for both acute
and chronic exposures. These responses are contained in the RfC range discussion included in
the “Management Considerations of Noncancer Risk Value Ranges Using Trichloroethylene as
an Example” section of this document as response to Charge Question 1.

8/22/13 19
The second evaluation method is more qualitative or semi-quantitative in nature, addressing four
main considerations:
1. The weight of evidence that links exposure with the toxic effect examining both the
strength of the epidemiology and animal evidence linking fetal heart malformations to
TCE exposure
2. Information on how other countries regulate TCE and how they address the risk of a
developmental effect
3. The margin of safety incorporated into entire RfC derivation process.
4. The use of the critical effects study: Johnson et al. (2003).

Weight of Evidence

U.S. EPA in their 2011 Toxicological Profile provides considerable information on the evidence
for TCE exposure and cardiac malformations. U.S. EPA concludes, based on weakly suggestive
evidence, but overall consistent epidemiological data in combination with evidence from
experimental and mechanistic studies that TCE exposure poses a potential hazard for congenital
malformations (U.S. EPA, 2011a: pgs. 6-10, 11). EPA’s conclusions were based on the
following:.
1. EPA cites three principal human studies indicating an increased risk of cardiac
malformations: ATSDR (2006a; 2008 two studies same population) and Yauck et al. (2004).
o ATSDR concluded that there was an increased risk of heart malformations in the
exposed population among Endicott, New York residents living in the area where
volatile organic compounds (VOCs) were found in soil vapor.
o The Yauck et al. (2004) study was conducted on children born between 1997-99 with
congenital heart defects in Milwaukee, Wisconsin, using TCE emissions data from
industrial sources and distance between the emission source and maternal residence
within 1.3 miles of the source. Yauck et al. concluded that offspring of women >38
years of age showed an increased risk of congenital heart defects while offspring of
women <38 years of showed no increased risk.
2. A series of animal studies was used to support EPA’s opinion. EPA (2011a) cited a number
of animal studies where fetal cardiac malformations occurred after oral dosing of TCE during
gestation (Dawson et al., 1993, 1990; Johnson et al., 2005, 2003) and fetal heart
malformations from oral dosing during gestation with the metabolites of TCE (Johnson et al.,
1998b; Johnson et al., 1998a; Epstein et al., 1992; Smith et al., 1992; Smith et al., 1989).
Additionally, EPA (2011a) cites support for fetal heart malformations from avian studies
(Rufer et al., 2008; Drake et al., 2006a; Drake et al., 2006b; Mishima et al., 2006; Boyer et
al., 2000; Loeber et al., 1988; Bross et al., 1983). The Rufer et al. (2008), Drake et al.
(2006a, 2006b), Loeber et al. (1988), and Bross et al. (1983) are all studies where TCE was
injected into the developing egg and cardiac malformations observed. Mishima et al. (2006)

8/22/13 20
 and Boyer et al. (2000) were studies presenting a plausible mechanism for fetal cardiac
malformations in chick embryos
3. EPA concludes that although the study used to set the developmental endpoint has significant
limitations, there is insufficient evidence to dismiss their findings. The EPA Science
Advisory Board (2011a) recommended the two endpoints for immune effects from Keil et al.
(2009) and the cardiac malformations from Johnson et al. (2003) to be considered the
principal studies supporting the RfC.
It is important to evaluate the supporting evidence in more detail and also to consider many
additional lines of evidence.
1. ATSDR (2006a, 2008) used strongly cautionary language in the interpretation of the
results of their two studies; stating that while risk was significantly elevated, an
extremely small number of infants born with birth defects was observed. ATSDR notes
(2008) that smoking and women working in the electronics industry confounds the
results. The elevated smoking rates likely contributed to some of the outcomes and that
workplace chemical exposure was also likely.
2. U.S. EPA also concluded that the evidence for an association between TCE exposures in
the human population and the occurrence of congenital cardiac defects is not particularly
strong. Many of the epidemiological study designs were not sufficiently robust to detect
exposure-related birth defects with a high degree of confidence.
3. The only animal studies conducted on TCE exposure and cardiac malformations came
from one study facility. Other studies were done with the metabolites of TCE or were
direct injections into developing avian embryos and did not include maternal exposure. It
is difficult to determine the relevance of direct embryo injection to the weight of
evidence.
It is apparent that the significant weight of evidence comes from animal studies done by one
facility. There is a considerable weight of evidence that the fetal cardiac malformations are not a
relevant inhalation critical effect. In particular:
1. Studies evidencing fetal heart malformations (Johnson et al., 2005, 2003; Dawson et al.,
1993, 1990) were limited to one testing facility. Following these results, two studies
were designed to duplicate the fetal cardiac malformations results of Johnson et al. (2005,
2003) and Dawson et al.(1993, 1990) but were unable to detect fetal cardiac anomalies.
In one study by Carney et al. (2006), TCE was administered via inhalation and in the
second study (Fisher et al., 2001), TCE was administered via oral dosing. Johnson
collaborated on the Fisher et al. (2001) study. The inability of the guideline, GLP-
quality study (Carney et al., 2006) and the study by Fisher et al. (2001) both designed to
reproduce the Johnson et al. (2003) results is significant. Despite considerable effort to
reproduce the results of the Johnson work, the fetal heart malformation critical effect
does not appear to be reproducible outside one testing facility.
2. EPA (2011a) concludes that a number of well conducted epidemiology studies did not
report a significant increase in fetal cardiac malformations (Bove, 1996; Bove et al.,
1995; Goldberg et al., 1990; Lagakos et al., 1986).

8/22/13 21
 3. A significant number of animal studies did not report induction of cardiac effects in rats
by inhalation (Carney et al., 2006; Healy et al., 1982; Dorfmueller et al., 1979; Schwetz
et al., 1975), in rabbits by inhalation (Hardin et al., 1981), in rats by gavage (Fisher et al.,
2001; Narotsky and Kavlock, 1995; Narotsky et al., 1995) or mice by gavage (Cosby and
Dukelow, 1992).
4. EPA (2011a: pg 4-560) indicated that the only identified cardiac “anomalies” following
gestational exposure to TCE or its metabolites were conducted in rats and dosed by the
oral route of exposure (gavage or drinking water).

Worldwide TCE Regulation

It is also important to consider how other countries and regulatory entities regulate TCE as
concerns cardiac malformations. The following table considers some national and global
perspectives on TCE exposure and Fetal Cardiac Malformations as a viable critical effect.

Table 2: Regulation of TCE for a Developmental Critical Effect

Regulatory Entity Year Fetal Cardiac Malformations (FCM) as critical

effect
European Union 2009 Did not define FCM as critical effect
National Research 2006 Notes FCM came from single laboratory, flat dose-
Council response, study needs to be repeatable in another
laboratory
National Institute of 1999, 2001 Did not consider cardiac effect or Dawson et al.
Public Health and (1993) study
Environment (RIVM)
Cal-EPA 2009 Johnson FCM was not a meaningful or
interpretable dose-response relationship, not
consistent with other work. Re-evaluating in
consideration of 2011a U.S. EPA RfC
WHO 2005, 2008, Notes Johnson et al. did not show a clear dose-
2009, 2010 response as might first appear on closer
examination of the data, but corroborated the
preferred Dawson et al. that showed a clearer
dose-response relationship.
Health Canada 2005, 2007 Notes cardiac anomalies reported in Dawson et al.
(1993) were corroborated by Johnson et al. (2003),
but dose–response from Johnson et al. is not as
clear as might first appear on closer examination
of the data; endpoint deserves close scrutiny, but
deemed Dawson et al. more appropriate as the key
study, because it showed a clearer dose–response
relationship

8/22/13 22
NICNAS (Australian 2010 Adopted WHO position on the developmental
National Industrial effects of TCE
Chemicals Notification
and Assessment Scheme)

Margin of Safety Analysis

An understanding of the risk from a developmental effect from exposure above the RfC can be
supported with an analysis of the margin of safety built into the entire process of the derivation
of the fetal health malformation critical effect candidate RfC. At each decision point in the
process for the candidate RfC there was uncertainty. EPA responded to this uncertainty by
incorporating a safety factor. For instance, when extrapolating from the animal study to human
exposure, EPA used a 1% response rate as opposed to the more common 5% or 10% response
rates. Each decision point incorporates a similar safety factor whose sum, considered as a whole,
provides an overall awareness of the entire margin of safety.
EPA used a considerable margin of safety in deriving the candidate RfC from the Johnson et al.
(2003) evidencing the cardiac malformations.
1. Physiologically based pharmacokinetic (PBPK) modeling was used to determine the
internal or absorbed dose in the animal resulting from the administered dose. It is
reasonable to assume the model outputs were conservative and provided an
underestimation of the true value. The output of this model is used as the input to a
second model.
2. The second model, a dose-response model, was used to determine the response rate
protective of 99% of the dosed population; termed the Benchmark Dose for a 1%
Response rate (BMR 1% or BMD01). The BMR used was 1%; the most protective
response rate. It is common to use a BMR of 10% (U.S. EPA, 2000). Response rates are
tied to study sensitivity and a 1% BMR is reserved for the most sensitive. By
comparison, nested reproductive or developmental studies are considered to be sensitive
and they generally use a BMR of 5% (U.S. EPA, 2000). Casarret and Doull (2001) point
out that the most extensive studies with developmental toxicity indicate that BMD05 (a
BMR of 5%) is similar to a statistically derived NOAEL for a wide range of
developmental toxicity endpoints. Thus, the 1% response rate would add a considerable
margin of safety above what is normally used.
3. Confidence bounds were placed on the BMR1% value for a BMDL or Benchmark Dose
Lower Confidence Limit. That is, the 95% lower confidence bound was placed on the
BMD of 1% to obtain a BMDL. The ratio of the BMD to the BMDL was about three
(EPA, 2011a: pg 5-45). Thus, use of the lower confidence limit on the BMD introduces a
factor of three margin of safety
4. Next, the animal BMDL is converted to a Human Equivalent Concentration (HEC99) at
the 99th percentile using a third model. That is, the BMDL modeling projects the lower

8/22/13 23
 confidence bound on a level at which 99% of exposed study population would not have a
sufficient internal dose to elicit a toxic response; then, a third model was used to
determine the HEC of a human dose that is protective of 99% of the human population.
The difference between the HEC50, what the average population would experience, and
the HEC99 which is protective of 99% of the population, was also about a factor of three.
5. It is important to recognize that the output of each of the models is the input for the next.
It is reasonable to assume each model is developed with a margin of safety that under-
predicts the true value. If the first model under-predicts the true value and its output is
used as an input into the next successive model, which also under predicts the true value,
then the margin of safety is compounded. This occurs for three successive models.
6. An additional factor of 10 margin of safety is added.

Section 3:
Margin Of Safety Measures Used To Set The TCE RfC: Hazard Ranges Above the
Reference Concentration (RfC)

Introduction

Waste sites around the world often contain chemicals for which target safe levels are
exceeded. In cases where the target safe levels are established for chemicals identified
as known or potential carcinogens, the acceptable cancer risks are generally presumed
to fall within an upper bound, excess cancer hazard range of 10-6 to 10-4. Screening
levels are typically set at a point value at the lower end of this range (e.g., 10-6) and risk
management decisions regarding the implementation of remedial actions (if any) are
made within this range based on various considerations (e.g., long-term effectiveness,
mass reduction of contaminant, cost, community acceptance of the proposed remedial
action). In the case of noncancer toxicity, few if any hazard ranges are determined,
primarily because the evaluation of noncancer hazards is predicated upon the
determination of a “safe dose” below which no toxic effect is expected, rather than
upon an acceptable (if non-zero) probability that a cancer-causing event may occur.

The National Academy of Sciences (2009) suggested the development of methods for
noncancer toxicity that have the capability of determining hazard ranges. The Alliance
for Risk Assessment (ARA) project entitled "Beyond Science and Decisions: From
Problem Formulation to Dose Response” is a response to this NAS suggestion, and six
of its case studies attempt to do this. These case studies are:
 Comparison of Hattis strawman approach and BMDs/UFs for noncancer
endpoints (carbonyl sulfide and tetrachlorobenzene) by Greco et al.
 Implications of Linear Low-Dose Extrapolation from Benchmark Dose for
Noncancer Risk Assessment by Kroner and Haber.
 Review and application of data fusion methodologies for toxicological dataset
analysis to resolve data quality issues in predictive toxicology and contaminated
sites risk assessment by Mohapatra et al.

8/22/13 24
  Estimate Risk Above the RfD Using Uncertainty Factor Distributions by Spalt
and Kroner.
 Use of biomarkers in the benchmark dose method Gentry et al.
 Use of Categorical Regression – Risk Above the RfD by Danzeisen et al.

The Science Panel (http://www.allianceforrisk.org/Workshop/Panel.htm) of this

workshop series vetted each of these case studies and found five of them to have merit.
3
Of these five, we focus on 2 for the purposes of this work, specifically:
 Use of biomarkers in the benchmark dose method; and
 Estimate Risk Above the RfD Using Uncertainty Factor Distributions

The remaining three case studies might be used at a future time.

Use Of Biomarkers In The Benchmark Dose Method

Method

The case study has been adapted to use with experimental animal data, and takes EPA
(2011a) determinations of benchmark dose (BMD) and the lower 95% confidence
interval on the benchmark dose (BMDL) for critical endpoints of TCE at face value.
The original case study is an extension of the BMD method that allows the
development of a hazard range at doses above the Reference Dose (RfD) when the
existing data are based on human responses.

The appropriate BMD is chosen in the usual fashion using existing EPA software and
criteria, including p-values, visual fit, residuals, BMD to BMDL ratios and Akaike’s
Information Criterion (AIC). The data are modeled to an appropriate POD using the
usual judgment, and then four different procedures were investigated to determine the
risk range. These procedures are:

1. A straight line is drawn from both the chosen BMDL and BMD to the RfC, where
the RfC is considered to be without risk;
2. The chosen model, rather than a straight line from the BMD/L, is extrapolated to
the RfC and then the risk at the RfC is considered to be zero;
3. The appropriate BMD model is extrapolated to the RfC and the risk at the RfC is
considered to be no greater than its upper bound; and
4. The appropriate BMD model is extrapolated using a threshold term, where the
threshold value is judged to be the RfC, or some higher value.

3
The Science Panel discourage the use of Implications of Linear Low-Dose Extrapolation from
Benchmark Dose for Noncancer Risk Assessment for purposes other than for comparison amongst
chemicals, because the modeled results did not appear to be consistent with a general understanding of
noncancer toxicity.

8/22/13 25
Results

Results of modeling procedure 1 for fetal heart malformations are shown in Table 3 and
Figures 1 and 2. Here the hazard range is determined from a straight line drawn from
each of two candidate PODs, i.e., the BMDL01 (the BMDL at a predicted response 1%
above background) and BMD01 (the BMD at a predicted response 1% above
background) to the RfC, where the RfC is considered to be without risk (i.e., 100%
probability that a safe human dose has not been overestimated). The BMDL01 and the
BMD01 are the median human equivalent concentrations (HECs) taken from EPA (IRIS
Toxicological Review of Trichloroethylene, Appendix F, p. F-35, 2011a) and adjusted
by a three-fold uncertainty factor to account for the expected differences in
toxicodynamics when extrapolating from the experimental animal (rats) to the human,
thereby accounting for the potential overestimation of the safe human dose based on
observed effects in test animals. Using the BMD methodology, a dose-response curve
is associated with the derivation of each POD (i.e., the BMD01 and BMDL01); each
dose-response curve represents the range of the expected human response, with the low
dose region of each curve reflecting sensitive individuals. EPA’s HEC99 value (i.e., the
human equivalent concentration based upon an individual at the 99th percentile of the
range of predicted human exposures, based on toxicokinetic variability in the human
population) is specifically not used in this procedure, since its use conflates the results
of procedure 1, specifically the use of the HEC99 would lower the BMD/L01 by an
undetermined amount.

Modeling procedures 2 and 3 were also attempted for fetal heart malformations.
Unfortunately, only one of EPA’s nested models fit the available data, so that a
comparison of results from different models was not possible. Furthermore, EPA’s
model gave a linear response in the low dose region, similar to those found with
procedure 1, and thus the results would be similar. Future work with either procedure 2
or 3 would be to try different models, since EPA’s sole fitting model does not reflect
the low dose data, that is, the response looks to be hormetic at the lowest, non-zero
dose (see Table 4).

Table 3. Modeling results for fetal heart malformations using procedure 1. All units are
in ppm, but convertible to µg/m3 by multiplying by 5.8.
Johnson et al. (2003): Fetal heart malformations
Concentration Hazard up to
RfC 0.00037 0
HEC, BMDL/3UFad 0.0040 0.01
EPA, 2011a, Figure F-15 (page F-35)

RfC 0.00037 0
HEC, BMDL/3UFad 0.011 0.01
EPA, 2011a, Table F-7 (page F-12) and EPA, 2011a, Figure F-15 (page F-35)

8/22/13 26
 Notes with the cardiac analysis:
 erudite developmental toxicologists think that the effect is spurious
 assessment is based on pup statistics, against EPA guidelines
 BMDL01 is used; not typical choice and several models failed
 High dose group dropped
3UFad = three-fold uncertainty factor for toxicodynamic uncertainty when extrapolating from
the experimental animal (rats) to the human

3UFad = three-fold uncertainty factor for toxicodynamic uncertainty when extrapolating from
the experimental animal (rats) to the human

Figure 1. Hazard Range of Heart

Malformations POD = BMDL
0.05
R
e 0.04
s 0.03
p BMDL risk up to
o 0.02
0.0040
n Linear (BMDL risk up
0.01
0.00037 to)
s
0
e
0 0.005 0.01 0.015 0.02
Concentration ppm

Figure 2. Hazard Range of Heart

Malformations POD = BMD
0.05
R
e 0.04
s 0.03
p
0.02 BMD risk up to
o 0.011
n Linear (BMD risk up to)
0.01
0.00037
s
0
e
0 0.005 0.01 0.015 0.02
Concentration ppm

8/22/13 27
Table 4. Data on fetuses and litters with abnormal hearts from Johnson et al. (2003)

Dose (mg/kg/d): 0 0.00045 0.048 0.218 129

Number of pups: 606 144 110 181 105

Abnormal heart 13 0 5 9 11

Litters 55 12 9 13 9
Abnormal heart 9 0 4 5 6

From EPA, Table F-4, Page F-9

Results of modeling procedure 1 for nephropathy are shown in Table 5 and Figures 3 and 4. As
before, hazard ranges above the RfC are determined from a straight line drawn from both the
BMDL05 (the BMDL at a predicted response 5% above background) and BMD05 (the BMD at a
predicted response 5% above background) to the RfC, where the RfC is considered to a
concentration associated with be without risk, and the BMDL05 and the BMD05 are the HECs
taken from EPA (IRIS Toxicological Review of Trichloroethylene, Appendix F, p. F-30, 2011a)
and adjusted by a three-fold uncertainty factor for toxicodynamic uncertainty when extrapolating
from the experimental animal (rats) to the human. Again, the use of the HEC99 is contra-
indicated because its use conflates with the results of procedure 1.
Modeling procedures 2 and 3 were also attempted for nephropathy. Unfortunately, the lowest
data point in these data is greater than 60% response, and, thus, these data are insufficient in the
low dose region to confidently predict a response with any model. In such cases, responses are
solely dependent on the choice of model, and would vary widely. Future work with either
procedure 2 or 3 is not possible on this data set.

Table 5. Modeling results for nephropathy using procedure 1. All units are in ppm, but
convertible to ug/m3 by multiply by 5.8.

NTP (1988): toxic nephropathy

Concentration Hazard Range
RfC 0.00056 0
HEC, BMDL/3UFad 0.014 0.05
(EPA, 2011a, Figure F-11 (page F-30)

RfC 0.00056 0
HEC, BMD/3UFad 0.020 0.05
EPA, 2011a, Figure F-10 (page F-29) and
EPA, 2011a, Figure F-11 (page F-30)

8/22/13 28
 Figure 3. Hazard Range of
Nephropathy POD=BMDL
0.014
0.05
R
e 0.04
s 0.03
p BMDL risk up to
o 0.02
n Linear (BMDL risk up
0.01
0.00056 to)
s
0
e
0 0.005 0.01 0.015 0.02
Concentration ppm

Figure 4. Hazard Range of

Nephropathy POD=BMD
0.020
0.05
R
e 0.04
s 0.03
p
0.02 BMD risk up to
o
n Linear (BMD risk up to)
0.01
0.00056
s
0
e
0 0.005 0.01 0.015 0.02
Concentration ppm

An interesting finding when comparing the results of both endpoints is that when hazard ranges
above the different RfCs are compared, the hazard ranges above the RfC for nephropathy are
more severe than for fetal heart malformations. For example, compare the BMD results in
Figures 4 and 2. Here, the risks associated with nephropathy at a concentration of 0.01 ppm
approximate may range up to approximately 2.5%, whereas the risks associated with fetal heart
malformations at the same concentration may range up to approximate 1%. This difference in
hazard ranges is also true when risks at the BMDL are compared in Figures 3 and 1.
Discussion
This development of hazard ranges for two these noncancer endpoints, considered as critical
effects in the EPA (2011a) TCE IRIS text, is in keeping with the NAS (2009) suggestion to

8/22/13 29
explore quantitative aspects of noncancer dose response assessment. The stated risks can be
confidently expected to not exceed the given values, but the corollary is that the risks are then
likely to be less and could even be zero.
In fact, since uncertainty factors are generally considered to be protective from the standpoint of
the expected average value (Dourson and Stara, 1983; Swartout et al., 1998), and since
noncancer effects are thought to be associated with thresholds (EPA, 2013), the actual risks at or
around the RfC are likely to be much less than that determined from a linear response. Thus, the
proper interpretation of these stated values is as a risk range, and not as a single value. This
same reason also leads to considering the hazard range associated with the BMD extrapolation to
be more appropriate than the hazard range associated with the BMDL extrapolation. This is
because the BMDL is used to estimate the RfC in a protective sense; i.e., it works with the
protective uncertainty factor to drive the RfC to an expected NOAEL in a sensitive human
population, or below a population threshold dose. The use of the BMDL extrapolation to project
an accurate hazard range is correspondingly diminished.
The choice of hazard range associated with the two different critical effects depends in part on
the level of confidence assigned to each RfC, and in part on the severity of the associated hazard
ranges developed in this analysis. EPA’s confidence in the RfC is “high” with nephropathy and
“medium” with heart malformations. The hazard range is more severe with nephropathy when
compared with heart malformations. In either case, the more appropriate choice of hazard range
is that associated with nephropathy.
Thus, the most reasonable hazard range to use is the one based on the BMD of nephropathy. This
choice can then be compared with other methods developed in this Alliance for Risk Assessment
(ARA) project.

Estimate RfD Probabilities Using Uncertainty Factor Distributions

Methods

This method uses theoretical distributions of EPA’s five uncertainty factors (i.e., intrahuman,
experimental animal to human, subchronic to chronic, LOAEL to NOAEL and lack of database),
based on the work of Swartout et al. (1998). Three of these factors are based on estimating a
NOAEL of one dose group to a comparable NOAEL of another (i.e., the expected same
response; e.g., experimental animal NOAEL to human NOAEL, subchronic NOAEL to chronic
NOAEL, and lack of database), and not estimating a different population response (i.e.,
interhuman and LOAEL to NOAEL). Thus, this method cannot develop a dose response
relationship in humans for the range of interest. Rather the probabilities are interpreted as the
likelihood that the stated RfD is a sensitive human NOAEL, which is the intent of the RfD’s
definition.4 Furthermore, if uncertainty factors are seen, individually, as "upper bounds" on the
dose-scaling factor for sources of uncertainty, then determining comparable upper bounds for
combinations of uncertainty factors can be accomplished by treating uncertainty factors as

4
The RfD is an estimate (with uncertainty spanning perhaps an order of magnitude) of a daily exposure to
the human population (including sensitive subgroups) that is likely to be without an appreciable
risk of deleterious effects during a lifetime (Barnes and Dourson, 1988).

8/22/13 30
distributions, which can be combined by probabilistic techniques. The Swartout et al. (1998)
approach does not attempt to distinguish one uncertainty factor from another based on empirical
data or biological mechanisms, but rather uses a simple displaced lognormal distribution as a
generic representation of all uncertainty factors.
Determining the probabilistic limits for the uncertainty factors used in the derivation of the RfD
is an important step toward the goal of characterizing the risk of noncarcinogenic effects from
exposure to environmental pollutants. The probabilities developed have applicability in
comparisons among RfDs and/or for determining different RfDs based on different choices of
probability.
Results
Table 5 shows the modeling results for EPA’s three different TCE IRIS RfCs at probabilities of
50%, 95% and 99%. The candidate RfC value associated with each of these probabilities reflects
the level of confidence that the underlying RfC truly reflects a sensitive human NOAEL (i.e, the
probability that the candidate RfC value does not exceed the NOAEL for any sensitive human
individual or subpopulation). Modeling results for two of these RfC values, Johnson et al.
(2003) and NTP (1998) are based on identical probabilities, since the uncertainty factor of 10-
fold is the same in both cases. The modeling results for Keil et al. (2009) are based on a
different probability distribution, since the uncertainty factor for this RfC is larger.
BOLD printed entries, shown in either black or red text, indicate matches to the IRIS RfC. Note
that whereas the IRIS RfCs for Johnson et al. (2003) and NTP (1998) are both associated with a
95% probability, the IRIS RfC for Keil et al. (2009) is associated with a 99% probability. RED
printed entries show consistency in 95% probability. Of note here is that the RfC based on Keil
et al. (2009) would be 2-fold higher than its IRIS equivalent if based on a 95% probability. Also
notable is that the RfCs based on Keil et al. (2009) and NTP (1998) are identical when based on
the same probability, whereas they are different when seen on EPA’s IRIS.
Table 6. 50th, 95th, & 99th % for TCE IRIS RfCs with Uncertainty Factors of 10 & 100; units in ug/m3.
Study IRIS UF Confidence IRIS RfC 50th % 95th % 99th %
Johnson et al. (2003) 10 Medium 4E-04 1E-03 4E-04 2E-04
NTP (1988) 10 High 6E-04 2E-03 6E-04 4E-04
Keil et al. 2009 100 Medium 3E-04 3E-03 6E-04 3E-04

Discussion
This development of probabilities with this method is not a dose response relationship in humans
for the effect of interest. Rather the probabilities are interpreted as the likelihood that the stated
RfC is a sensitive human NOAEL, which as stated before, is the intent of the RfC’s definition.
This method is also in keeping with the NAS (2009) suggestion to explore quantitative aspects of
noncancer dose response assessment.
The stated probabilities can be confidently expected to approximate the given values, but with
the realization that the Swartout et al. (1998) approach does not attempt to distinguish one

8/22/13 31
uncertainty factor from another based on empirical data or biological mechanisms, but rather
uses a simple displaced lognormal distribution as a generic representation of all uncertainty
factors. Specific, refined distributions could be used to replace these generic ones (see for
example, Baird et al., 1996).
If managers wish to focus on RfCs with a consistent probability, then the comparison among
RfCs would be on the basis of the values associated with the chosen probability, and not the
value as stated on IRIS. For example, in the choice of a 95% probability, the associated RfCs
would be 4E-04 for heart malformations, 6E-04 for nephropathy and 6E-04 for immune effects.
In this case, the RfCs for immune and nephropathy are the same and the value for heart
malformations is different, whereas the values on IRIS suggest that the RfCs for immune and
heart malformations are similar and that the RfC for nephropathy is different. Different choices
of probability lead to correspondingly different RfCs.

Section 4.
Risk Management Synthesis
Introduction

Waste sites throughout the United States are often managed using risk assessment information
and methods from EPA. Risk values such as Regional Screening Levels (RSLs) that are derived
with these EPA methods (U.S. EPA, 2012b) are used as part of the process of managing site
risks. For chemicals that cause cancer, acceptable cancer risks are typically defined as a risk
range that spans two orders of magnitude, or a factor of 100-fold, that is, a range from 10-6 to 10-
4
(U.S. EPA, 1991). For noncancer toxicity, the RfD or its inhalation equivalent, the RfC is used,
without specifying any risk or hazard range. However, the RfD and RfC as defined by EPA
(Barnes and Dourson, 1988; U.S. EPA, 1994; and U.S. EPA, 2002) are associated with
“uncertainty spanning perhaps an order of magnitude.” So can a range be determined from this
definition, and if so, might a risk manager be able to use it?
The reasoning behind this definition of RfD/RfC, and, specifically, with the use of the word
“perhaps” by Barnes and Dourson (1988) had everything to do with the variation in the
underlying toxicity databases of different chemicals. In some cases, the database has sufficient
information in humans and experimental animals so that the resulting RfD/RfC can be developed
with a one-fold or 10-fold uncertainty factor. Uncertainty or imprecision in such an RfD/RfC is
smaller, that is, perhaps less than an order of magnitude. In other cases, fewer data are available
so that the resulting RfD/RfC is developed with a 1000-fold or larger uncertainty factor.
Uncertainty or imprecision in these RfD/RfCs is greater, that is, perhaps more than an order of
magnitude.
Each RfD/RfC value has both imprecision and uncertainty. The imprecision part comes from the
repeatability of the overall process; i.e., how close to the first RfD/RfC would a second expert
body come, if it were to repeat the process. In such cases, the imprecision might be best
characterized as perhaps three-fold on either side of the RfD/RfC (Felter and Dourson, 1998;
U.S. EPA, 2002).

8/22/13 32
The uncertainty part comes from the fact that each RfD/RfC is protective, because it is developed
using uncertainty factors that are each protective based on the observed behavior of the
“average” chemical (Dourson and Stara, 1983; Dourson et al., 1996). This protectiveness
assures that any RfD/RfC is an underestimate of the expected value. Furthermore, the use of
multiple uncertainty factors results in even more protective RfD/RfC values (Swartout et al.,
1998). Because of this, the uncertainty in an RfD/RfC might be best characterized as perhaps
10-fold above the RfD/RfC, where the latter is considered as the floor to this range, i.e., the
RfD/RfC is the lowest value within the range (Dourson and Stara, 1983; Felter and Dourson,
1998).
Risk managers who need to manage waste sites throughout the U.S. often address risk for
carcinogenicity using a risk range of 10-6 to 10-4 and for noncancer toxicity using a hazard index
of 1 for mixtures, or a Hazard Quotient of 1 for single chemicals. The purpose of this paper is to
determine a range in the hazard index or quotient that is consistent with the range of uncertainty
in the RfD/RfC, and that is similar in concept to the risk range of 10-6 to 10-4 for carcinogenicity.
This range in the hazard index or quotient might allow managers to have comparable flexibility
in closing or otherwise managing waste sites where the evaluation of noncancer effects drives the
risk assessment. The evaluation of the noncancer effects associated with the chlorinated solvent
trichloroethylene (TCE) is used as an example to illustrate the determination of, and application
of, this hazard index range.

Defining the Range in the RfD/RfC and Hazard Index or Quotient

In the IRIS Summary for TCE, U.S. EPA (2011a) has identified three candidate RfC values in its
evaluation of the noncancer inhalation toxicity of TCE. These three candidates are described
below:

 a candidate RfC of 2 µg/m3 based on decreased thymus weight in female mice (Keil et
al., 2009);

 a candidate RfC of 2 µg/m3 based on fetal heart malformations in rats (Johnson et al.,
2003); and

 a candidate RfC of 3 µg/m3, based on toxic nephropathy in female rats (NTP, 1988).
Each of the candidate RfC values may be evaluated with respect to the imprecision and the
uncertainty inherent in its derivation.
As further described in Section 1, the imprecision of each of the candidate RfCs may be
considered as a uniform or equal distribution around the RfC as its central value, that is, as 2
µg/m3 for Keil et al. (2009) or Johnson et al. (2003), or 3 µg/m3 for NTP (1988). A floor and
ceiling to this uniform distribution would be a value lower and higher than the candidate RfC,
perhaps two-fold in each direction for the Johnson et al. (2003) or NTP (1988) RfCs, and three-
fold in either direction for the Keil et al. (2009) RfC.
The uncertainty in each of the candidate RfCs may also be evaluated, but in contrast to
imprecision, these uncertainties are associated with ranges of values that are not uniformly

8/22/13 33
distributed around the RfC as its central value. This is because these three candidate RfCs are
developed using different PODs and different uncertainty factors that are known to be protective
from the behavior of the “average” chemical. For risk management decisions, it is generally the
range of the uncertainty that is more important when compared with the range of imprecision.
This is because managers are interested in making decisions that are protective of public health
and it is an understanding of the uncertainties in the range of RfCs that is generally more
informative.
These non-uniform ranges of uncertainty have as a floor the individual candidate RfCs as on
IRIS. This decision to use the individual candidate RfC as the floor of the range for each non-
cancer endpoint is reasonable from a practical point of view, because managers are unlikely to
take regulatory action below these values, due to the protective nature implicit in the derivation
of each candidate RfC, as described above. However, using the RfD/RfC as the floor to a range
in its value also has theoretical support (Dourson and Stara, 1983).

Because the uncertainty range varies for each study and each candidate RfC, it may be helpful to
develop a midpoint so as to guide managers when exposures fall within the range. The
determination of the midpoint of this non-uniform hazard range is a judgment that meshes four
considerations for each candidate RfC; specifically:
 the size of the uncertainty factor,
 the steepness of the hazard slope (i.e., the slope of the line describing hypothetical
population responses at concentrations above the RfC, see section 3),
 the confidence in the choices of critical effect, and
 the confidence in the PDD.

Midpoints that are closer to their RfCs are associated with a smaller uncertainty factor, a steeper
hazard slope, a higher confidence in the critical effect, and a higher confidence in the POD.
Midpoints that are further from their RfCs are associated with a larger uncertainty factor, a
shallower hazard slope, a lower confidence in the critical effect, and a lower confidence in the
POD.

For the fetal malformation endpoint based on Johnson et al. (2003), the midpoint of the
endpoint-specific uncertainty range (10 µg/m3) is judged to be five-fold above the candidate RfC
due to its small uncertainty factor of 10, shallower hazard slope (Section 3), low confidence in
the critical effect (Section 2), and low confidence in the choice of a benchmark response of 1%
(BMDL01). For the toxic nephropathy endpoint based on the NTP (1988) study, the midpoint of
the endpoint-specific uncertainty range (9 µg/m3) is judged to be three-fold above the candidate
RfC due to its small uncertainty factor of 10, steeper hazard slope (Section 3), medium
confidence in the critical effect, and medium to low confidence in the choice of a benchmark
response of 5% (BMDL05). For decreased thymus weight endpoint based on the Keil et al.
(2009) study, the midpoint of the endpoint-specific uncertainty range (20 µg/m3) is judged to be
10-fold above the candidate RfC due to its larger uncertainty factor of 100, medium confidence
in the critical effect, and medium to low confidence in its choice of a LOAEL as the POD. The
effect shown by Keil et al. (2009) does not lend itself to dose response modeling, so a judgment
of steepness of hazard slope is not possible.

8/22/13 34
The ceiling value for the endpoint-specific uncertainty range for each RfC is defined as the POD
(i.e., the HEC99) for each candidate RfC (i.e., for each non-cancer endpoint), as shown on EPA’s
IRIS. Therefore, the ceiling value of each endpoint-specific uncertainty range is 20 µg/m3 from
the Johnson et al. (2003) study), 30 µg/m3 from the NTP (1988) study and 190 µg/m3 from the
Keil et al. study (2009). This decision is reasonable from a practical point of view, because
managers are likely to take regulatory action above these values due to the fact that specific toxic
effects can sometimes be associated with their values. The floor, midpoint and ceiling values for
each endpoint-specific uncertainty range are shown in Table 7.

Table 7. Different uncertainty ranges for different TCE RfCs. All values are in µg/m3. Shaded
areas indicate best overall uncertainty range for risk management purposes.
Confidence Uncertainty Ranges
Study IRIS Steep b Critical c Point of d Floor Intermediate Ceiling
UF a Slope Effect Departure
Floor

Johnson et al. Lower Low Low 2 10 20

10
(2003)

Higher Medium Medium to 9 30

NTP (1988) 10 3
Low

Medium to
Keil et al. 2009 100 NA Medium 2 20 190
Low

a. Size of the uncertainty factor as on IRIS

b. Steepness of the hazard slope (i.e., the slope of the line describing hypothetical population
responses at concentrations above the RfC), as per Section 3.
c. Confidence in the choices of critical effect, as per Section 4.
d. Confidence in the POD, as per Section 4.

Since EPA selected these three studies (two principal and one supporting) as the basis for the
derivation of the RfC for non-cancer effects, the endpoint-specific uncertainty range of each of
the three studies may be considered in the implementation of risk management decisions. From
the collective evaluation of the endpoint-specific uncertainty ranges of all three studies, a “total
uncertainty range” of 2 µg/m3 to 190 µg/m3 may be inferred. However, extraction of a “multi-
endpoint uncertainty range” from the broader total uncertainty range is more useful for risk
management decision-making. The multi-endpoint uncertainty range may be defined as an
estimate of “a daily exposure to the human population (including sensitive subgroups) that is

8/22/13 35
likely to be without an appreciable risk of deleterious effects during a lifetime”, based upon the
definition of the RfD (Barnes and Dourson, 1988)5.
The multi-endpoint uncertainty range of the RfC for TCE (as defined by multi-endpoint floor,
midpoint and ceiling values) was determined here by careful discernment of the confidence,
imprecision and uncertainty associated with each endpoint-specific floor, midpoint and ceiling
value. The floor of the multi-endpoint uncertainty range of the RfC for TCE was determined by
comparing the candidate RfC values from each of the three studies (i.e., 2 µg/m3 for both the
decreased thymus weight and fetal cardiac malformation endpoints, and 3 µg/m3 for the toxic
nephropathy endpoint). These three floor values are so closely clustered that, based on the
imprecision inherent in non-cancer hazard estimates, the values are essentially indistinguishable6.
The endpoint-specific floor value of 3 µg/m3 based on toxic nephropathy not only represents the
endpoint-specific floor value of the highest overall confidence from among the three endpoint-
specific floor values (see Table 7), but is equivalent to a concentration that is found within the
mathematical precision range of each of the three endpoint-specific floor values7. Therefore, the
5
Barnes and Dourson (1988) defined the RfD as a point value “with an uncertainty spanning perhaps an
order of magnitude”. The point of departure (POD) for the RfD as defined by Barnes and Dourson was
typically (although not always) based upon a no observed adverse-effects level (NOAEL) from a single
study for a single critical effect. The assessment of confidence levels in the study design, critical effect
and POD of a single study and single critical effect enables the assignation of uncertainty factors in a
relatively straightforward manner. The multi-endpoint uncertainty range of the RfC for TCE has been
defined here (in 2013) in the same terms as the RfD (as stated in 1988). The use of the definition of the
RfD to also define the multi-endpoint uncertainty range is a recognition of the complexity associated with
the prediction of a “safe dose” or “safe concentration” associated with multiple effects, observed in
multiple studies and species, and multiple points of departure (each variously based on a BMDL01,
BMDL05 or lowest observed adverse effect level (LOAEL) value, appropriately averaged over two
different exposure periods, as relevant for developmental or chronic effects. Therefore, the concept of a
“safe dose” for the non-cancer effects of TCE has been applied here to a range of values (i.e., the multi-
endpoint uncertainty range) which represents a “safe concentration” for multiple endpoints and multiple
studies, with various degrees of confidence in study design, critical effect and POD, as illustrated in Table
7.
6
With respect to non-cancer endpoints, the safety of human exposures to environmental concentrations is
evaluated on the basis of the hazard quotient (HQ), i.e., the ratio of the environmental concentration of a
substance (Cair) to its RfC (i.e., Cair/RfC). The HQ associated with acceptable exposures is one (1), with a
precision of one significant figure (i.e., an implicit HQ range of 0.95 to 1.5). The mathematical precision
of the endpoint-specific floor value of 2 µg/m3 corresponds to a range of 1.9 µg/m3 to 3 µg/m3; the
mathematical precision of the endpoint-specific floor value of 3 µg/m3 corresponds to a range of 2.5
µg/m3 to 4.5 µg/m3. Thus, the ranges associated with the implicit precision of the three endpoint-specific
floor values overlap; at a precision of one significant figure, the value of 3 µg/m3 falls within the
mathematical precision ranges of all three endpoint-specific floor values. The endpoint-specific floor
value of 3 µg/m3 based on toxic nephropathy not only represents the endpoint-specific floor value of the
highest overall confidence from among the three endpoint-specific floor values (see Table 7), but is
equivalent to a concentration that is found within the mathematical precision range of each of the three
endpoint-specific floor values; the value of 3 µg/m3 was thus selected to represent the floor value of the
multi-endpoint uncertainty range.
7
Ibid.

8/22/13 36
value of 3 µg/m3 was selected to represent the floor value of the multi-endpoint uncertainty
range.

Similarly, the midpoint of the multi-endpoint uncertainty range of the RfC for TCE is determined
by comparing the endpoint-specific midpoint values from each of the three studies (i.e., 20 µg/m3
for the decreased thymus weight endpoint, 10 µg/m3for the fetal cardiac malformation endpoint,
and 9 µg/m3 for the toxic nephropathy endpoint). The midpoint values for the fetal cardiac
malformation and toxic nephropathy endpoints are so closely clustered that the values are
essentially indistinguishable. The endpoint-specific midpoint value of 9 µg/m3 based on toxic
nephropathy not only represents the endpoint-specific midpoint value of the highest overall
confidence from among the three endpoint-specific midpoint values (see Table 7), but is lower
than or equivalent to the other two endpoint-specific midpoint values. Therefore, the value of 9
µg/m3 was selected to represent the midpoint value of the multi-endpoint uncertainty range.

The ceiling of the multi-endpoint uncertainty range of the RfC for TCE is determined by
comparing the endpoint-specific ceiling values from each of the three studies (i.e., 190 µg/m3 for
the decreased thymus weight endpoint, 20 µg/m3 for the fetal cardiac malformation endpoint, and
30 µg/m3 for the toxic nephropathy endpoint). The endpoint-specific ceiling values for the fetal
cardiac malformation and toxic nephropathy endpoints are closely clustered, and the
mathematical precision ranges of the two values overlap8; the endpoint-specific ceiling value for
decreased thymus weight is substantially higher. Therefore, the endpoint-specific ceiling value
for fetal cardiac malformations (20 µg/m3) has been selected here as the ceiling value for the
multi-endpoint uncertainty range. Although the POD is of low confidence, the value is based on
the HEC99 (calculated from the BMDL01), and is a plausible yet conservative estimate of the
“safe concentration” for human exposures9.
In summary, the confidence in each of the endpoint-specific uncertainty ranges was subsequently
considered in the determination of the multi-endpoint uncertainty range for risk management
purposes (i.e., 3 µg/m3 to 20 µg/m3). The higher-confidence results of the NTP study were used

8
The mathematical precision (at HQ = 1) associated with the endpoint-specific ceiling value of 20 ug/m3
corresponds to a range of 15 ug/m3 to 30 ug/m3; the mathematical precision (at HQ = 1) of the endpoint-
specific floor value of 30 ug/m3 corresponds to a range of 25 ug/m3 to 45 ug/m3. Thus, the ranges
associated with the implicit precision of each endpoint-specific floor value overlap.
9
The ceiling value for the toxic nephropathy endpoint (30 ug/m3) is a plausible alternative representation
of the ceiling value for the multi-endpoint uncertainty range. Since the mathematical precision ranges (at
HQ=1) of the endpoint-specific ceiling values for the fetal cardiac malformation and toxic nephropathy
endpoints overlap, the values may be considered to be approximately equivalent. In this case, the
endpoint-specific ceiling value for toxic nephropathy would be favored over the endpoint-specific ceiling
value for fetal cardiac malformation on the basis of higher confidence not only in the critical effect (high
versus low) but also in the point of departure (medium versus low); the latter comparison is of particular
relevance since the point of departure serves as the endpoint-specific ceiling value for each endpoint.
Nevertheless, the endpoint-specific ceiling value for fetal cardiac malformations was selected as the
multi-endpoint ceiling value, since the amount of overlap between the two ranges of mathematic precision
was not substantial. In addition, the selection of the ceiling value of 20 ug/m3 based on the HEC99 for
fetal cardiac malformations, enables the risk manager to conservatively account for developmental
endpoints when the multi-endpoint uncertainty range is used as the basis for risk management decisions.

8/22/13 37
to determine the floor and midpoint of this uncertainty range. The highly controversial results
from the Johnson et al. (2003) study, while associated with low confidence (see Section 2), were
nevertheless used to determine the ceiling level of this uncertainty range. The widest endpoint-
specific uncertainty range is embedded within the individual uncertainty range from the Keil et
al. (2009) study; therefore, this study was considered to be confirmatory of the uncertainty
ranges associated with the other two endpoints.
Toxicologists are not able to distinguish the absence of health risk between any two or more
values within this overall uncertainty range. That is to say, toxicologists cannot differentiate, for
example, the “safety” of a value of 4 µg/m3 versus a value of 17 µg/m3. Because of this,
managers may use different values within the multi-endpoint uncertainty range of 3 µg/m3 to 20
µg/m3, along with site-specific exposure assessments, and other site considerations to make
different decisions on a case-by-case basis.
For example, when a site-specific exposure assessment defines a range of exposures that are
primarily below the multi-endpoint uncertainty range of 3 µg/m3 to 20 µg/m3, then the
probability of inducing any noncancer effects in the exposure population is lower and the priority
for any management action is reduced (see Figure 3a). In this case, a risk manager may decide to
take no action, or to delay action pending further information.
In contrast, when the exposure assessment defines a range in exposures that is primarily above
the multi-endpoint uncertainty range of 3 µg/m3 to 20 µg/m3, then the probability of inducing
noncancer effects in the exposure population is higher and the priority for risk management
action is increased (see Figure 3c). In this case, a risk manager may decide to take action, or to
ask for specific information that would refine the estimates of health risk and/or exposure.
When the exposure assessment defines a range in exposures that are primarily in the multi-
endpoint uncertainty range of 3 µg/m3 to 20 µg/m3, then risk managers can use the intermediate
value in this uncertainty range, that is 9 µg/m3 and other site considerations, to gauge whether a
management action is needed or if further information should be sought (see Figure 3b).
The multi-endpoint uncertainty range can also be used to develop a comparable range of hazard
quotient estimates for single-chemical exposures to TCE alone, or hazard index estimates for
exposures to mixtures of TCE and other chemicals (e.g., solvent impurities such as 1,1,1-
trichloroethane, 1,2-dichlroethane, 1,1-dichloroethene; or degradation products, such as cis-1,2-
dichloroethene and vinyl chloride), when only a point estimate of exposure is available for
comparison. This adaptation is consistent with EPA’s definition of the RfD/RfC, and is akin to
the range developed for cancer risk (i.e., an excess lifetime cancer risk of 10-6 to 10-4). When
considering exposures to mixtures, risk managers should be cognizant of the target organ(s),
mode(s) of action and mechanism(s) of action of the various chemicals in the mixture, based on
critical effects and other effects that may be elicited at environmentally relevant concentrations.
Additionally, the role of co-exposures and interactions of chemicals may be considered and/or
ruled out in developing the range in the hazard index for actions that are recommended by risk
managers.
The multi-endpoint uncertainty range is composed of floor, midpoint and ceiling values for
which the appropriate averaging time corresponds to different exposure durations (i.e.,
developmental or long-term/chronic exposure periods). Therefore, this range can be applied to

8/22/13 38
both long- and short-term exposures, with the associated differences in exposure averaging
times.10 For shorter-term exposures, the results from the Johnson et al. study (2003) might also
be used to describe the best averaging time, but if so, this averaging time should be based on the
average time of cardiac development in humans during fetal growth, approximately 24 days
(based on Marcela et al., 2013; Fanaroff and Martin, 2003; Schleich, 2003). The use of a 24-day
average time for cardiac development in humans is consistent with the fact that the dosing in the

Figure 3a. Exposure distribution of indoor air concentrations primarily

below the 3 µg/m3 to 20 µg/m3 hazard range. Relatively small proportion
of exposures is higher than 3 µg/m3. Nominal actions or no further action
may be warranted for risk management.

Figure 3b. Exposure distribution of indoor air concentrations falling

within the 3 µg/m3 to 20 µg/m3 hazard range. Relatively small proportion
of exposures is higher than 9 µg/m3. Limited action may be warranted for
risk management

Figure 3c. Exposure distribution of indoor air concentrations primarily

above the 3 µg/m3 to 20 µg/m3 hazard range. Actions to reduce exposures
may be warranted for risk management.

Johnson et al. study (2003) occurred during the whole time of cardiac development in the rat, as
described in the next section.

Cardiac development in various species

Cardiac development has been described in several species including humans (Marcela et al.,
2012; Moorman et al., 2003; Zaffran and Frasch, 2002; Srivastava, 2000). The heart is reported
to be the first internal organ to form and become functional in the vertebrate embryo (Srivastava,
2000; Richtsmeier, 1999). It develops in several sequential steps, beginning from an epithelium
to mesodermal tissue, to a tube, to a looped, two-chambered structure, and finally to the mature
four-chambered organ (Fanaroff and Martin, 2002). In the rat, the first primitive cardiac
segment appears on embryonic day 9 (ED 9) (Marcela et al., 2012). A comparable
developmental stage has been described in stage 10 mouse embryo (7 days post coitum) and
Carnegie Stage VI-VII human embryos (corresponding to second week of gestation) (Marcela et
al., 2012) or beginning of third week of gestation) (Fanaroff and Martin, 2003). Schleich (2003)

10
Use of the RfC values for assessing long-term exposures is well-established in EPA guidance.

8/22/13 39
lists the following stages as occurring during the third week of gestation in humans: development
of the primitive streak—gastrulation (day 15); formation of intra-embryonic mesoderm (day 16);
mesoblast differentiation—somatopleura and splanchnopleura (day 17); development of blood
islets, the cardiogenic region, and primitive heart tubes (day 19); and formation of the primitive
endocardial tube (day 21). A series of internal and external changes occurs, leading to a fully
formed heart (i.e., a mature heart, with interventricular foramen completely closed and almost
compact interventricular septum) in the rat ED 16 embryo. The heart is also reported to be fully
formed in the stage 30 Hamburger–Hamilton (HH) chick embryo (Contreras Ramos et al., 2008)
and Carnegie Stage XIX human embryo (Goor and Lillehei, 1975) (corresponding to 38 ± 1
days [Marcela et al., 2012]). According to Marcela et al. (2012), the more relevant
developmental events in rat fetuses from 17 to 20 days and 21 days newborn include an increase
in weight and crown to tail length. Aside from these changes, the fetus morphology remained
almost the same, including the still-closed eyes. Histological maturation of all the organs
appears to be almost fully in the ED 16 embryo.
Based on the available information, the human hearts starts developing between days 13 and 15
after gestation or later and may be completely formed by 37-39 days into gestation (Marcela et
al., 2012) or later, indicating the length of cardiac development in humans during fetal growth to
be 24-26 days. Since exposure to TCE in Johnson et al. (2003) occurred during the whole time
of cardiac development in the rat, it is reasonable to use the time of cardiac development in
humans, 24-26 days, as the averaging time for risk management decisions.

References
ATSDR (Agency for Toxic Substances and Disease Registry). (2006a). Health statistics review:
Cancer and birth outcome analysis, Endicott area investigation, Town of Union, Broome County,
New York. U.S. Department of Health and Humans Services. Atlanta, GA. Available at:
http://www.atsdr.cdc.gov/HAC/pha/EndicottAreaInvestigation/EndicottHealthStatsReviewHC05
2606.pdf
ATSDR (Agency for Toxic Substances and Disease Registry). (2006b). Health consultation:
Public health implications of exposures to low-level volatile organic compounds in public
drinking water. Endicott area investigation, Broome County, New York. U.S. Department of
Health and Human Services. Atlanta, GA. Available at:
http://www.atsdr.cdc.gov/HAC/pha/EndicottAreaInvestigation113006/EndicottAreaInvestigation
HC113006.pdf .
ATSDR (Agency for Toxic Substances and Disease Registry). (2008). Health consultation:
Health statistics review follow-up: Cancer and birth outcome analysis. Endicott area
investigation, Town of Union, Broome County, New York. U.S. Department of Health and
Human Services. Atlanta, GA. Available at:
http://www.atsdr.cdc.gov/hac/pha//EndicottAreaInvestigationFollowUp/EndicottAreaHC051508.
pdf .

8/22/13 40
Baird, S.J.S., Cohen, J.T., Graham, J.D., Shlyakhter, A.I., and Evans, J.S. (1996). Noncancer
Risk Assessment: A Probabilistic Alternative to Current Practice. Human Ecol Risk Assess. 2(1):
78-99.
Barnes, D.G. and Dourson, M.L. (1988). Reference dose (RfD): Description and use in health
risk assessments. Regul Toxicol Pharmacol. 8: 471-486.
Blossom, S.J., Doss, J.C., Hennings, L.J., Jernigan, S., Melnyk, S., and James, S.J. (2008).
Developmental exposure to trichloroethylene promotes CD4+ T cell differentiation and
hyperactivity in association with oxidative stress and neurobehavioral deficits in MRL+/+ mice.
Toxicol Appl Pharmacol. 231: 344–353.
Bove, F., Fulcomer, M., Klotz, J., Esmart, J., Dufficy, E., and Savrin, J. (1995). Public drinking
water contamination and birth outcomes. Am J Epidemiol. 141: 850-862.
Bove, F. (1996). Public drinking water contamination and birthweight, prematurity, fetal deaths,
and birth defects. Toxicol Ind Health. 12: 255-266.
Boyer, A., Finch, W., and Runyan, R. (2000). Trichloroethylene inhibits development of
embryonic heart valve precursors in vitro. Toxicol Sci. 53: 109-117.
Bross, G., DiFranceisco, D., and Desmond, M. (1983). The effects of low dosages of
trichloroethylene on chick development. Toxicology. 28: 283-294.
CalEPA (California Environmental Protection Agency). (2009). Public Health Goal for
Trichloroethylene in Drinking Water. Pesticide and Environmental Toxicology Branch, Office of
Environmental Health Hazard Assessment. Available online at:
http://oehha.ca.gov/water/phg/pdf/TCE_phg070909.pdf.
Carney, E., Thorsrud, B., Dugard, P., and Zablotny, C. (2006). Developmental toxicity studies in
Crl:CD (SD) rats following inhalation exposure to trichloroethylene and perchloroethylene. Birth
Defects Res B Dev Reprod Toxicol. 77: 405-412. Available at:
http://dx.doi.org/10.1002/bdrb.20091.
Casarett, L.J., & Doull, J. (2001). Casarett and Doull's Toxicology: The Basic Science of
Poisons, 6th Ed. McGraw-Hill Pub. Division, New York, NY.
Contreras-Ramos, A., Sanchez Gomez, C., Garcia Romero, H., and Cimarosti, L. (2008). Normal
development of the muscular region of the inter- ventricular septum. I. The significance of the
ventricular trabecu- lations. Anat Histol Embryol. 37: 344–351.
Cosby, N., and Dukelow, W. (1992). Toxicology of maternally ingested trichloroethylene (TCE)
on embryonal and fetal development in mice and of TCE metabolites on in vitro fertilization.
Toxicol Sci. 19: 268-274.
Dawson, B; Johnson, P; Goldberg, S; Ulreich, J. (1990). Cardiac teratogenesis of
trichloroethylene and dichloroethylene in a mammalian model. J Am Coll Cardiol 16: 1304-
1309.

8/22/13 41
Dawson, B., Johnson, P., Goldberg, S., and Ulreich, J. (1993). Cardiac teratogenesis of
halogenated hydrocarbon-contaminated drinking water. J Am Coll Cardiol. 21: 1466-1472.
Available at: http://dx.doi.org/10.1016/0735-1097(93)90325-U.
Dawson, H.E., and McAlaray, T. (2009). A compilation of statistics for VOCs from post-1990
indoor air concentration studies in North American residences unaffected by subsurface vapor
intrusion. Ground Water Monit R. 29: 60-69.
Dorfmueller, M., Henne, S., York, R., Bornschein, R., and Manson, J. (1979). Evaluation of
teratogenicity and behavioral toxicity with inhalation exposure of maternal rats to
trichloroethylene. Toxicology. 14: 153-166.
Dourson, M.L., and Stara, J.F. (1983). Regulatory history and experimental support of
uncertainty (safety) factors. Regul Toxicol Pharmacol. 3: 224-238.
Dourson, M.L., Felter, S.P., and Robinson, D. (1996). Evolution of science-based uncertainty
factors in noncancer risk assessment. Regul Toxicol Pharmacol. 24: 108-120.
Drake, V., Koprowski, S., Hu, N., Smith, S., and Lough, J. (2006a). Cardiogenic effects of
trichloroethylene and trichloroacetic acid following exposure during heart specification of avian
development. Toxicol Sci. 94: 153-162. Available at: http://dx.doi.org/10.1093/toxsci/kfl083.
Drake, V., Koprowski, S., Lough, J., Hu, N., and Smith, S. (2006b). Trichloroethylene exposure
during cardiac valvuloseptal morphogenesis alters cushion formation and cardiac hemodynamics
in the avian embryo. Environ Health Perspect. 114: 842-847.
Epstein, D., Nolen, G., Randall, J., Christ, S., Read, E., Stober, J., and Smith, M. (1992).
Cardiopathic effects of dichloroacetate in the fetal Long-Evans rat. Teratology. 46: 225-235.
Available at: http://dx.doi.org/10.1002/tera.1420460306.
EU (European Union). 2009. Recommendation from the Scientific Committee on Occupational
Exposure Limits for Trichloroethylene. SCOEL/SUM/142. Available at:
http://ec.europa.eu/social/BlobServlet?docId=6405&langId=en
Faranoff, A.A., and Martin, R.J. (2002). Neonatal-Perinatal Medicine, Diseases of the Fetus and
Infant, 7th Ed. Vol. 2. Mosby Inc. pp. 1095-1103.
Felter, S.P., and Dourson, M.L. (1998). The Inexact Science of Risk Assessment (and
Implications for Risk Management). Human Ecol Risk Assess. 2: 245-251
Fisher, J., Channel, S., Eggers, J., Johnson, P., MacMahon, K., Goodyear, C., Sudberry, G.,
Warren, D., Latendresse, J., and Graeter, L. (2001). Trichloroethylene, trichloroacetic acid, and
dichloroacetic acid: Do they affect fetal rat heart development. Int J Toxicol. 20: 257-267.
Folkes, D., Wertz, W., Kurtz, J., and Kuehster, T. (2009). Observed spatial and temporal
distributions of CVOCs at Colorado and New York vapor intrusion sites. Ground Water Monit
R. 29: 70-80.

8/22/13 42
Goldberg, S., Lebowitz, M., Graver, E., and Hicks, S. (1990). An association of human
congenital cardiac malformations and drinking water contaminants. J Am Coll Cardiol. 16: 155-
164.
Goor, D.A., and Lillehei, C.W. (1975). Congenital malformations of the heart. Embryology,
anatomy, and operative considerations. Grune & Strattion, New York, NY.
Hardin, B., Bond, G., Sikov, M., Andrew, F., Beliles, R., and Niemeier, R. (1981). Testing of
selected workplace chemicals for teratogenic potential. Scand J Work Environ Health. 7: 66-75.
Health Canada (2005). Guidelines for Canadian drinking water quality: supporting
documentation. Trichloroethlyene. Ottawa, Ontario.
Healy, T., Poole, T., and Hopper, A. (1982). Rat fetal development and maternal exposure to
trichloroethylene 100 ppm. Br J Anaesth. 54: 337-341.
IRIS (Integrated Risk Information System). (2012). Trichloroethylene (CASRN 79-01-6). U.S.
Environmental Protection Agency. Available at: http://www.epa.gov/iris/subst/0199.htm
Johnson, P., Dawson, B., and Goldberg, S. (1998a). Cardiac teratogenicity of trichloroethylene
metabolites. J Am Coll Cardiol. 32: 540-545. Available at: http://dx.doi.org/10.1016/S0735-
1097(98)00232-0.
Johnson, P., Dawson, B., and Goldberg, S. (1998b). A review: trichloroethylene metabolites:
potential cardiac teratogens. Environ Health Perspect. 106(Suppl 4): 995-999.
Johnson, P., Goldberg, S., Mays, M., and Dawson, B. (2003). Threshold of trichloroethylene
contamination in maternal drinking waters affecting fetal heart development in the rat. Environ
Health Perspect. 111: 289-292.
Johnson, P., Goldberg, S., Mays, M., and Dawson, B. (2005). Correction: Threshold of
trichloroethylene contamination in maternal drinking waters affecting fetal heart development in
the rat. Environ Health Perspect. 113: A18.
Keil, D., Peden-Adams, M., Wallace, S., Ruiz, P., and Gilkeson, G. (2009). Assessment of
trichloroethylene (TCE) exposure in murine strains genetically-prone and non-prone to develop
autoimmune disease. J Environ Sci Health A Tox Hazard Subst Environ Eng. 44: 443-453.
Available at: http://dx.doi.org/10.1080/10934520902719738.
Lagakos, S., Wessen, B., and Zelen, M. (1986). An analysis of contaminated well water and
health effects in Woburn, Massachusetts. J Am Stat Assoc. 81: 583-596.
Loeber, C., Hendrix, M., Diez De Pinos, S., and Goldberg, S. (1988). Trichloroethylene: a
cardiac teratogen in developing chick embryos. Pediatr Res. 24: 740-744.
Marcela, S.G., Monsalve, M.C.R., Garibay, M.A.P., Martinez, M.A., Diaz-Cintra, S., De La
Rosa-Santander, P., Bladimir, R-R., and Gomez, C.S. (2012). Chronological and morphological
study of heart development in the rat. The Anatomoical Record. 295: 1267-1290.

8/22/13 43
Mishima, N., Hoffman, S., Hill, E., and Krug, E. (2006). Chick embryos exposed to
trichloroethylene in an ex ovo culture model show selective defects in early endocardial cushion
tissue formation. Birth Defects Res A Clin Mol Teratol. 76: 517-527. Availabe at:
http://dx.doi.org/10.1002/bdra.20283.
Moorman, A., Webb, S., Nigel, A., Brown, N.A., Lamers, W., and Anderson, R.H. (2003).
Development of the heart: (1) Formation of the cardiac chambers and arterial trunks. Heart. 89:
806-814.
Narotsky, M., and Kavlock, R. (1995). A multidisciplinary approach to toxicological screening:
II. Developmental toxicity. J Toxicol Environ Health. 45: 145-171. Available at:
http://dx.doi.org/10.1080/15287399509531987.
Narotsky, M., Weller, E., Chinchilli, V., and Kavlock, R. (1995). Nonadditive developmental
toxicity in mixtures of trichloroethylene, di(2-ethylhexyl) phthalate, and heptachlor in a 5 x 5 x 5
design. Fundam Appl Toxicol. 27: 203-216. Available at:
http://dx.doi.org/10.1093/toxsci/27.2.203.
NRC (National Research Council/ Committee on Human Health Risks of Trichloroethylene).
(2006). Assessing the Human Health Risks of Trichloroethylene: Key Scientific Issues.
Washington, DC, National Academies Press, Available at:
http://www.nap.edu/catalog/11707.html
NTP (National Toxicology Program). (1988). Toxicology and carcinogenesis studies of
trichloroethylene (CAS No. 79-01-6) in four strains of rats (ACI, August, Marshall, Osborne-
Mendel) (gavage studies). Research Triangle Park, NC: U.S. Department of Health and Human
Services, Public Health Service, National Institutes of Health. Available at:
http://ntp.niehs.nih.gov/ntp/htdocs/LT_rpts/tr273.pdf.
OEHHA (Office of Environmental Health Hazard Assessment). (2007). Trichloroethylene.
Chronic Toxicity Summary. Available at:
http://oehha.ca.gov/air/chronic_rels/pdf/79016.pdfOEHHA (Office of Environmental Health
Hazard Assessment). (2012). Initiation of Process to Update Public Health Goals for Three
Chemicals in Drinking Water , September 2012. Available at:
http://oehha.ca.gov/water/phg/092512InitRep.html
RSLTs (Regional Screening Level Tables). (2012) Regional Screening Level Tables User’s
Guide. Available at: http://www.epa.gov/reg3hwmd/risk/human/rb-
concentration_table/usersguide.htm
Rufer, E., Hacker, T., Lough, J., and Smith, S. (2008). Low-dose trichloroethylene exposure
during valvuloseptal morphogenesis causes ventricular septal defects in hatched chicks.
Toxicologist. 102: 314.
Schwetz, B., Leong, B., and Gehring, P. (1975). The effect of maternally inhaled
trichloroethylene, perchloroethylene, methyl chloroform, and methylene chloride on embryonal
and fetal development in mice and rats. Toxicol Appl Pharmacol. 32: 84-96. Available at:
http://dx.doi.org/10.1016/0041-008X(75)90197-0.

8/22/13 44
Smets, P. (1999). Imperfect Information-Imprecision-Uncertainty. IRIDIA Université Libre de
Bruxelles. Available at: http://sites.poli.usp.br/d/pmr5406/Download/papers/Imperfect_Data.pdf
Smith, M., Randall, J., Read, E., and Stober, J. (1989). Teratogenic activity of trichloroacetic
acid in the rat. Teratology. 40: 445-451. Available at: http://dx.doi.org/10.1002/tera.1420400506.
Smith, M., Randall, J., Read, E., and Stober, J. (1992). Developmental toxicity of dichloroacetate
in the rat. Teratology. 46: 217-223. Available at: http://dx.doi.org/10.1002/tera.1420460305.
Srivastava, D., and Olson, E.N. (2000). A gene tic blueprint for cardiac development. Nature.
407: 221-226.
Swartout, J., Price, P., Dourson, M., Carlson-Lynch, H., and Keenan, R. (1998). A probabilistic
framework for the reference dose. Risk Anal. 18(3): 271-282.
TERA (Toxicology Excellence for Risk Assessment). (2012). Risk Methods. Available at:
http://www.tera.org/iter/riskmethods.html
U.S. EPA (U.S. Environmental Protection Agency). (1989). Risk Assessment Guidance for
Superfund. Volume I: Human Health Evaluation Manual (Part A). Office of Solid Waste and
Emergency Response. EPA/540/1-89/002. Available at:
http://www.epa.gov/oswer/riskassessment/ragsa/index.htm
U.S. EPA (U.S. Environmental Protection Agency). (1991). Role of the Baseline Risk
Assessment in Superfund Remedy Selection Decisions. Office of Solid Waste and Emergency
Response, OSWER Directive 9355.0-30, April 22, 1991.
U.S. EPA (U.S. Environmental Protection Agency). (1993). EPA ROD EPA/ROD/R10-93/059
Allied Plating, Inc., EPA ID: ORD009051442 OU 01 Portland OR
U.S. EPA (U.S. Environmental Protection Agency). (1994). Methods for Derivation of
Inhalation Reference Concentrations and Application of Inhalation Dosimetry. Office of Health
and Environmental Assessment. Washington, DC. EPA/600/8-90-066F. October.
U.S. EPA (U.S. Environmental Protection Agency). (2001), Trichloroethylene Health Risk
Assessment: Synthesis and Characterization - External Review Draft, Office of Research and
Development, EPA/600/P-01/002A.
U.S. EPA (U.S. Environmental Protection Agency). (2002). A Review of the Reference Dose
(RfD) and Reference Concentration (RfC) Processes. Risk Assessment Forum. EPA/630/P-
02/002F, December.
U.S. EPA (U.S. Environmental Protection Agency). (2011a). Toxicological Review of
Trichloroethylene, in support of summary information on the Integrated Risk Information
System (IRIS). September 2011. EPA/635/R-09/011F.
U.S. EPA (U.S. Environmental Protection Agency). (2011b). Background Indoor Air
Concentrations of Volatile Organic Compounds in North American Residences (1990 – 2005): A
Compilation of Statistics for Assessing Vapor Intrusion. June 2011. EPA 530-R-10-001.

8/22/13 45
U.S. EPA (U.S. Environmental Protection Agency). (2012b). APTI Atmospheric Sampling
Course Appendix G: Significant Figures and Rounding. Available at:
http://www.epa.gov/apti/Materials/APTI%20435%20student/Student%20Manual/Appendix_G_n
o%20TOC-cover_MRpf.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2012a). Human Health Risk Assessment
Bulletins, Supplement to RAGS, USEPA Region 4 Front Page. Available at:
http://www.epa.gov/region4/superfund/programs/riskassess/healtbul.html
WHO DWG (World Health Organization Drinking Water Guidelines). (2008, 2009). Guidelines
for drinking water quality, 3rd Ed., incorporating first and second addenda (2008) and rolling
revisions current to 2009. World Health Organization. Available at:
<http://www.who.int/water_sanitation_health/dwq/chemicals/en/index.html>.
Williams, A.L., and DeSesso, J.M. (2008). Trichloroethylene and ocular malformations: analysis
of extant literature. Int J Toxicol. 27: 81–95.
Yauck, J., Malloy, M., Blair, K., Simpson, P., and McCarver, D. (2004). Proximity of residence
to trichloroethylene-emitting sites and increased risk of offspring congenital heart defects among
older women. Birth Defects Res A Clin Mol Teratol. 70: 808-814. Available at:
http://dx.doi.org/10.1002/bdra.20060.
Zaffran, S., and Frasch, M. (2002). Early signals in cardiac development. Circ Res. 91: 457-469.

8/22/13 46