Critical Reviews in Toxicology, 2011; 41(5): 369383

2011 Informa Healthcare USA, Inc.

ISSN 1040-8444 print/ISSN 1547-6898 online
DOI: 10.3109/10408444.2010.543655

Review Article

Critical analysis of literature on low-dose synergy for use in

screening chemical mixtures for risk assessment
Alan Boobis1, Robert Budinsky2, Shanna Collie3, Kevin Crofton4, Michelle Embry5, Susan Felter6,
Richard Hertzberg7, David Kopp7, Gary Mihlan8, Moiz Mumtaz9, Paul Price2, Keith Solomon10, Linda
Teuschler11, Raymond Yang12, and Rosemary Zaleski13,
Imperial College London, London, UK, 2The Dow Chemical Company, Midland, Michigan, USA, 3Synergy Toxicology,
Boerne, Texas, USA, 4US Environmental Protection Agency, Research Triangle Park, North Carolina, USA, 5ILSI Health
and Environmental Sciences Institute, Washington, DC, USA, 6The Procter & Gamble Company, Cincinnati, Ohio, USA,
7
Emory University, Atlanta, Georgia, USA, 8Bayer CropScience, Research Triangle Park, North Carolina, USA, 9Centers
for Disease Control, Agency for Toxic Substances and Disease Registry, Atlanta, Georgia, USA, 10University of Guelph,
Guelph, Ontario, Canada, 11US Environmental Protection Agency, Cincinnati, Ohio, USA, 12Colorado State University,
Fort Collins, Colorado, USA, and 13ExxonMobil Biomedical Sciences, Inc., Annandale, New Jersey, USA
1

Abstract
There is increasing interest in the use of tiered approaches in risk assessment of mixtures or co-exposures to
chemicals for prioritization. One possible screening-level risk assessment approach is the threshold of toxicological
concern (TTC). To date, default assumptions of dose or response additivity have been used to characterize the toxicity
of chemical mixtures. Before a screening-level approach could be used, it is essential to know whether synergistic
interactions can occur at low, environmentally relevant exposure levels. Studies demonstrating synergism in
mammalian test systems were identified from the literature, with emphasis on studies performed at doses close to
the points of departure (PODs) for individual chemicals. This search identified 90 studies on mixtures. Few included
quantitative estimates of low-dose synergy; calculations of the magnitude of interaction were included in only 11
papers. Quantitative methodology varied across studies in terms of the null hypothesis, response measured, POD
used to test for synergy, and consideration of the slope of the dose-response curve. It was concluded that consistent
approaches should be applied for quantification of synergy, including that synergy be defined in terms of departure
from dose additivity; uniform procedures be developed for assessing synergy at low exposures; and the method for
determining the POD for calculating synergy be standardized. After evaluation of the six studies that provided useful
quantitative estimates of synergy, the magnitude of synergy at low doses did not exceed the levels predicted by
additive models by more than a factor of 4.
Keywords: Chemical mixtures, low dose, risk assessment, synergy, TTC

Contents
Abstract 369
1. Introduction and background 370
2. Methods 370
2.1. Literature search 370
2.2. Database 372
2.3. Evaluation of synergy 372
3. Results 373
3.1. General findings 373
Address for Correspondence: Michelle Embry, ILSI Health and Environmental Sciences Institute, Washington, DC 20005, USA. E-mail:
membry@ilsi.org
(Received 30 August 2010; revised 24 November 2010; accepted 25 November 2010)

369

370 A. Boobis etal.

3.2. Magnitude of interaction 373
3.2.1. Studies reporting synergy using Method A, the ratio of observed to predicted dose for a fixed response 373
3.2.2. Studies reporting synergy using Method B, the ratio of observed to predicted response at a specific dose377
4. Discussion 379
4.1. Critical evaluation of reported magnitude of synergy 379
4.2. Overall conclusions/future directions 379
Acknowledgements 382
Declaration of interest 382
References 382

1. Introduction and background

Regulatory and public concern over the potential risk
from exposure to combinations of chemicals is increasing given the multitude of combinations possible, either
in the form of a preformed mixture, through co-exposures
to multiple sources, or accumulation of multiple compounds in the body over time. Current methodologies for
assessing the risks from mixtures include whole-mixture
and component-based approaches (ATSDR, 2004a; US
EPA 2000). Whole-mixture approaches are often preferred because these inherently account for unidentified materials in the mixture and for any interactions
among component chemicals. However, whole-mixture
approaches are not always feasible, due to the diversity of
potential mixtures and temporal variation in real-world
exposures. This has required risk managers to rely on
component-based approaches using, for example, doseor response-additivity models.
In general, component-based approaches require
toxicity information on each component of a mixture.
The availability of information on the toxicity of the components of mixtures is inconsistent; some compounds
have extensive data on toxicity whereas others have
never been tested, and these data gaps limit the use of
component-based approaches. In addition, the present
use of additivity models in mixture toxicology is based
on the assumption that no significant toxicodynamic or
toxicokinetic interactions occur at those exposures (i.e.,
requires the assumption that synergy does not occur
at the exposures of interest). Synergy is defined in this
article as a mixture response that significantly exceeds
that predicted from the no-interaction model, further
discussed under Section 2.3.
To address these issues, the International Life Sciences
Institute (ILSI) Health and Environmental Sciences
Institute (HESI) Risk Assessment Methodologies
Technical Committee convened a working group of
academic, government, and industry representatives to
explore and to improve methodologies currently available for assessing human health risks posed by exposure to chemical mixtures. The team elected to initially
explore screening-level risk assessment methodologies
that could address risks from real world exposures to
mixtures. Ideally, a successful screening-level methodology for risk assessment of mixtures would differentiate


between mixtures with insignificant risk from those with

potentially significant risk, while still being conservative and protective. Such methodology would allow
toxicity testing and risk assessment efforts to focus on
those mixtures with the greatest potential risk to human
health. Our initial objective was to determine the feasibility of applying a screening-level approach, such as the
threshold of toxicological concern (TTC) (Cramer etal.,
1978; Munro et al., 1996; Kroes, 2004; Barlow, 2005), to
mixtures. However, after investigating various screening approaches, including the TTC, it became clear that
in order to ensure that any screen developed was sufficiently conservative (i.e., very few false negatives), there
was a need to address whether synergy was of concern at
real world exposures, and if so, its magnitude.
In order to determine the likelihood that synergy would
occur at exposure levels relevant to the application of the
TTC approach and cause the toxicity of a mixture to be
underestimated by the use of additivity models, the committee commissioned a literature review of the published
information on empirical findings of synergy. This began
with the following aims:
Identify papers reporting synergy using doses at
or near the respective points of departure (PODs),
for example, no observed adverse effect levels
(NOAELs) or benchmark dose levels (BMDLs) from
which chronic health-based guidance values may be
derived for individual mixture components (e.g., reference doses [RfDs], tolerable daily intakes [TDIs]).
Determine those studies in which the magnitude of
synergy could be quantified.

2. Methods
2.1. Literature search
A literature search was conducted to identify key studies
from 1990 to 2008 relating to synergism in mammalian test
systems at low doses. Since the focus of this exercise was
on the potential underestimation of risks from mixtures
due to synergy, this search did not consider antagonism
(i.e., interactions less than additive), as this would result
in an overestimation of risk. Key references prior to 1990
were also identified from reviews and relevant databases.
The emphasis of the search was on studies deemed low
dose in order to be consistent with the HESI projects
Critical Reviews in Toxicology

Literature analysis of low-dose synergy 371

No

Interaction
mentioned
in abstract?

No

Assess as discussion
reference

Evaluate citations in
bibliography

Yes

Specific
secondary
info?

No

Extract data to
populate fields

Original
research?

Yes

Yes

Reject paper as
conditions for
inclusion are not met

Compare Against Inclusion/Exclusion Rules

Mammalian systems only: reptile, fish excluded
Epidemiology excluded without validated exposures
Chemical stressors preferred: eg., noise excluded
In vivo data preferred over in vitro/in silico studies
Conform > additive (synergistic) interaction
Work is peer reviewed or agency produced

No

Quality
apparent?

Yes

No

Unique
features?

No

Yes

Work
since
1990?

Yes
Discuss test in
relation to low dose

Figure 1. Decision logic for inclusion of studies.

goal of developing screening-level risk assessment methodologies for low to moderate environmental exposures.
As discussed above, low doses were defined as at or near
the respective PODs (e.g., NOAELs, BMDLs, etc.) that
could be used to derive health-based guidance values
for chronic exposures to individual mixture components.
However, very few studies were found in which such dose
levels had been used. The survey was therefore expanded
to include shorter-term studies of nonlethal endpoints,
resulting in the inclusion of acute studies with doses well
above the chronic PODs.
Figure 1 depicts the scheme for conducting the literature search and selecting papers for follow-up evaluation. A range of scientific journal databases (PubMed/
MEDLINE, TOXLINE, Google Scholar, and SCIRUS) was
searched to find the key articles published after 1990.
2011 Informa Healthcare USA, Inc.

To ensure consideration of as many key studies as possible, the initial selection included an assessment of
synergistic interactions listed in the Interaction Profiles
by the Agency for Toxic Substances and Disease Registry
(ATSDR), chemicals evaluated as synergistic in the US
Environmental Protection Agency (US EPA) MIXTOX
interaction database, and synergy identified with specific
pesticides (Carpy etal., 2000). Additional activities were
undertaken to widen the breadth of the search, including an aggressive investigation of the gray literature via
Google search of the Worldwide Web, contacting key
scientists in toxicology and risk assessment, and posting
a request for papers on online discussion communities.
Combinations of chemicals of potential interest were further identified through authorship or other links to known
interaction studies. Finally, to aid the search and capture

372 A. Boobis etal.

of relevant data, papers on agents that were judged likely
to modulate metabolic pathways (e.g., enzyme inducers)
were examined, and search terms based on studies of
known interactions were used.
Additional search limitations included the following:
Data quality assessment: Exclusion of studies that
lacked defined standards for data quality.
Novel findings: Inclusion of well-known synergistic
interactions if they extended current knowledge;
otherwise, there was only limited inclusion of studies
on well-established synergistic phenomena (such as
asbestos and cigarette smoking) that occur at higher
doses.
True synergy only: Exclusion of epidemiology studies without adequate exposure measurements and
studies that described idiosyncratic findings, such
as those that appeared to be restricted to specialized
conditions and that have been questioned or contradicted by related studies.
A list of literature databases and keywords searched as
part of this review are available in Appendix A of the
Supplemental Information.

2.2. Database
A database was developed based upon the identified
studies (Supplemental Information, Appendix B). For
each study, the database includes detailed information
on substances evaluated and experimental methodology
(species, study duration, endpoints assessed, dose regime
including levels, route, and timing). Outcomes relevant
to synergy are capturede.g., whether or not synergy
was found; if found and quantified, the magnitude of
synergy as reported in the study. In one case (Korsak
etal., 1988), the specific method used for calculating the
reported increase in potency could not be determined, so
the synergy magnitude was recalculated for the purposes
of this evaluation, based on reported EC50s and dose
addition. Other relevant information including statistical
significance and general comments are noted. Chemical
1
0.9

Whole Mixture
Observed Dose
Response Curve

Mixture response

0.8

Dose response of the

mixture predicted using
an additive model

0.7
0.6

Synergy factor (the ratio of

effective doses at equivalent
mixture response)
A = EDx(ad)/EDx(m)

0.5
0.4

R(m)

Synergy factor (the ratio of

responses at an equivalent
total mixture dose)
B = R(m)/R(ad)

0.3
0.2
0.1
0

R(ad)

EDx(m)

EDx(ad)
Mixture dose

Figure 2. Two approaches for defining synergy. R is response,

ED is dose producing a defined response (x% of maximum), (m)
refers to observed results with mixture, and (ad) refers to predicted
results with mixture.


Abstracts Service (CAS) registry numbers are used as

unique identifiers for individual compounds entered
into the database.

2.3. Evaluation of synergy

Several definitions of toxicological interaction consistent with the term synergy can be found in the
published literature (Hertzberg and MacDonell,
2002; Kodell and Pounds, 1991; Knemann, 1996). In
general, synergy occurs when co-exposure to two or
more compounds leads to an enhancement of toxicity
beyond that which would occur under a specified definition of additivity. In this paper, dose additivity is the
preferred assumption, which is usually applied when
mixtures comprise combinations of toxicologically
similar chemicals; however, if a group of chemicals act
by different toxic modes of action so that toxicological
independence is plausible, then response addition is
a reasonable assumption. Synergy is generally determined empirically and the mechanism by which it
occurs varies. In addition, interaction is relative, i.e.,
all interactions are judged by comparison to a nointeraction model.
The approach used in this review to define, evaluate,
and characterize synergy is based on the use of this nointeraction null hypothesis, i.e., the determination of
both the likelihood that an additive model of effects will
underestimate the toxicity of a mixture and the magnitude of the underestimation. Synergy is therefore defined
as a mixture response that significantly exceeds that predicted by a no-interaction model.
Two methods of characterizing the magnitude
of synergy are commonly used (Figure 2). The sigmoidal response (solid curve) at the left is the observed
response for the mixture, and the sigmoidal response
(dashed curve) at the right represents the predicted
response based on an assumption of additivity. Ideally,
this prediction is established based on a comparison
of results from a well-conducted study of the doseresponse of the components as a mixture to the results
of well-conducted studies of the dose-responses of
the individual mixture components. Method A (ovals)
expresses the magnitude of synergy as the ratio of doses
causing a specified response (i.e., how far does the doseresponse curve for the mixture shift left). This ratio has
been called the interaction magnitude (US EPA, 2000).
In Figure 2, the observed ED20 of the mixture is 20 and
the predicted ED20 based on an additivity model is 25.
The synergistic magnitude using Method A is therefore
1.25 (25/20).
Method B (triangles) defines the magnitude of synergy
as the ratio of responses at a specific dose (i.e., how far
does the dose-response curve for the mixture shift up).
In this example, the mixture induces a 40% response at
a given dose level, whereas the simple additivity model
would predict a 10% response at this dose level. The magnitude of the synergistic effect by Method B is therefore 4
(40%/10%).
Critical Reviews in Toxicology

Literature analysis of low-dose synergy 373

It is clear that the magnitude of synergy calculated
using either method is highly dependent on the dose or
response level at which synergy is evaluated.
The US EPA recommends Method A for defining the
interaction (synergy or antagonism) between chemicals
when estimating this for its interaction-based hazard
index formula (ATSDR, 2004a; Hertzberg and Teuschler,
2002; US EPA, 2000; Mumtaz and Durkin, 1992). Method
A better relates to the risk assessment process where
changes in the POD would be reflected directly in the setting of corresponding guidance values, unlike Method B,
which reflects changes in response rates at overtly toxic
effect levels. However, Method A requires the assumption of a common mode of action or, at a minimum, a
common target organ for toxicity amongst the chemicals.
Method B has several other characteristics:
The response ratios are limited at higher responses,
as well as below the no-effect level (the value of the
ratio becomes infinite when the additive model predicts a zero response).
Response ratios may also be limited by the dynamic
range of the response, i.e., a small range for induction
of liver weight versus a large range for induction of
cytochrome P450 (CYP) mRNAs.
The examination of observed versus predicted
response ratios is problematic when there are differences in the efficacies (maximal response) of the
mixture versus the predicted additive response. For
example, the mixture may show antagonism that
blunts its maximal response.
Method B is appropriate only for monotonic doseresponses curves; not all biological responses have
this characteristic (e.g., dose-dependent transitions
in mode of action).
Method B is appropriate for cases where the chemicals in the mixture have independent modes of
action.
For the purposes of this project, it was decided that the
determination of the occurrence of synergy would be
best evaluated in terms of the plausibility of the authors
determination of a departure from the no-interaction
estimate. This finding could have been made based on
either of the two approaches described above. In addition to the studies identified in the literature search, two
recent studies were included. Because many authors do
not provide raw data, we did not attempt to reanalyze the
published data. The exception, stated previously, was
Korsak et al. (1988), where EC50s were provided so that
it was possible to calculate interaction magnitude based
on dose addition. When available, details were included
in the database on the basis of the interaction, such
as the underlying interactive process (e.g., metabolic
interference).
As discussed above in this section, the two approaches
for calculating the magnitude of synergy (Figure 2) can
produce different estimates from the same data. Therefore,
when comparing estimates of the magnitude of synergy
across different mixtures in different publications, those
2011 Informa Healthcare USA, Inc.

studies using Method A were distinguished from those

using Method B.

3. Results
3.1. General findings
The search strategy identified 90 papers, with data for 204
unique chemicals where a mixture effect had been studied. Data from all 90 papers were entered into the database
(included in Appendix B, Supplemental Information).
These publications were critically reviewed, and studies
that did not provide novel findings on synergy and studies where synergy was not supported by the information
provided in this article were excluded from further consideration. Comments related to this critical review are
included in Appendix B. This left 43 papers on mixtures
with original mammalian data from which synergy could
be examined.

3.2. Magnitude of interaction

The magnitude of the interaction (observed compared
to predicted from a no-interaction model) was rarely
included in the publications. Only 11 studies identified
in this project actually reported the magnitude of the
interaction (Table 1). The magnitudes of the interactions
reported were calculated using Methods A or B described
in Section 2.3 above. Each of the studies that reported a
quantified magnitude for synergy is discussed in greater
detail below. The magnitude of synergy included in our
database is the value the authors reported; with the
exception of Korsak etal. (1988), none of the values was
recalculated for the database.
3.2.1. Studies reporting synergy using Method A, the ratio of
observed to predicted dose for a fixed response
Moser etal. (2005) Moser etal. (2005) tested the individual and combined doses of five organophosphate
pesticides (acephate, diazinon, chlorpyrifos, malathion,
and dimethoate). The ratio of pesticides in the mixture reflected that found in estimates of food intake in
humans. The lowest tested doses used by Moser et al.
(2005) for two of the organophosphates were approximately equivalent to their acute oral NOAELs, but in two
instances (chlorpyrifos and malathion) exceeded acute
oral NOAELs by an order of magnitude or more. Note
that no comparison can be made with acephate because
there is no available reported NOAEL. Two mixtures
were tested: the first contained all five chemicals (full
ray), the second mixture was the same with the exception that malathion was not included. At the highest test
concentrations, doses of individual constituents were
1.5 (diazinon)1614 (malathion) times their respective
NOAELs in the full ray mixture and 1.5466 (chlorpyrifos) times their respective NOAELs in the reduced ray
mixture.
The study examined the acute effects of the individual chemicals and of the mixtures on neurobehavioral function in adult male Long-Evans rats following

Serum total thyroxine,

ED30

Rats, oral, 4 days

18 thyroid disruptors
(2 dioxins, 2
dibenzofurans,
12 PCBs)

Toluene: 10304850ppm
1.5 (>1000ppm
doses)
Xylene: 10304970ppm
Toluene + xylene: 1050
4700ppm (50/50)
2.5 (at 3 highest
In highest mixture dose,
doses)
individual components
ranged from 0.006382 g/
kg/day [environmental
background to 100
background]

1.33.5
Full ray dose 10165mg/
kg/day; reduced ray
(no malathion, other
components at same as full
ray) 1.7528.9mg/kg/day

Exposure

Species, route,
duration

Mixture

Endpoint

Hydroquinone: 60mg/kg
Phenol: 50160mg/kg

Mouse, i.p. Injection, Total micronuclei

3 days
frequency

Hydroquinonephenol

Synergy magnitude
reported in study

4 (at 3 highest
doses)

2. Studies that report synergy magnitude using Method B (ratio of observed to predicted response at a specific dose)
MirexTPA
Mouse, dermal,
No. of skin tumors
200 nmol mirex
1.9
20 wks
2 nmol TPA
200 nmol DMBA
1.6 - 2.5
Ascigarette smoke Epidemiology study No. of lung cancer cases Arsenic level in drinking
water <10 to >700 g/L;
cigarette smoke nonsmoker,
<25 or 25 pack-years

EC50 rotarod
performance

Rats, inhalation, 4h

toluene/xylene

ED20s; motor activity,

blood ChE, brain ChE,
gait score, tail pinch
response

Rats, oral, 4h

5 pesticides

Quantification method

Excluded
quantitative

Chen etal.,
1995

Table 1. continued on next page

Committee
evaluation

Included

Chen etal.,
2004

Reference

Included

Meyer etal.,
1994

Included

Departure from additivity Crofton etal.,

model
2005

Ratio of no. of tumors

for mixture to combined
response of DMBAinitiated mirex or TPA
Ratio of observed excess
alone
risk in those exposed to
both As and cigarette
smoke and the excess risk
predicted under additivity
Ratio of mixture response
to combined responses of
individual doses

Included

Included

Included

Committee
evaluation

Korsak etal.,
1988

EC50 ratio based on dose

addition

Departure from additivity Moser etal.,

model
2006

Table 1. Summary of studies reporting interaction magnitudes.

Species, route,
Synergy magnitude
Mixture
duration
Endpoint
Exposure
reported in study
Quantification method
Reference
1. Studies that report synergy magnitude using Method A (ratio of observed to predicted dose associated with a fixed response)
1.22.1
Departure from additivity Moser etal.,
5 pesticides
Rats, oral, 4h
ED20 for blood ChE, EC20s Full ray dose 10450mg/
model
2005
kg/day; reduced ray
for motor activity, gait
(no malathion, other
score, brain ChE
components at same as full
ray) 1.7578.7mg/kg/day

374 A. Boobis etal.

Critical Reviews in Toxicology

 2011 Informa Healthcare USA, Inc.

10HcAM

No. of liver cell of foci

No. of liver cell of foci

Rats, oral, 6 wks

initiated 2 wks post
DEN i.p. injection

5HcAM

AUC ratio

Modeled peak
concentration

Rats, inhalation, 4h

Toluene/Ethylbenzene
Xylene/Ethylbenzene
Toluene/xylene/
Ethylbenzene
Toluene/xylene/
N/A
Ethylbenzene

Table 1. Continued.
Toluene/xylene

1.82.1
2.02.1
3.2

1.72.2

Toluene: 17200ppm
1316
Ethylbenzene: 33100ppm
Xylene: 33100 ppm
1.1110
Total HcAM dose 55 or
330ppm, individual HcAM
at 6100ppm (representing
1/5th or 1/25th carcinogenic
dose each)
42
Total HcAM dose 450ppm,
individual HcAM at
1580ppm representing
1/10th carcinogenic dose
each (no effect with each at
1/100th carcinogenic dose)

100ppm of each solvent

200ppm of one solvent and

100ppm of another

Excluded due
to issues with
quantitative
methodology

Excluded from
summary
of empirical

Tardif etal., 1996 Excluded

does not
measure a
toxicological
endpoint

Dennison etal.,
Unity calculation;
2005
interaction magnitude
calculated using
PEL values
Ito etal., 1991
Net value of mixture
compared to sum totals
for individual chemicals
tested at the same conc. as
present in the mixture
Ito etal., 1998
Results for mixture
compared with those for
each chemical at 1/10th the
dose level

Ratio of total AUC for

mixture exposure to
sum of AUC for single
exposures

Literature analysis of low-dose synergy 375

376 A. Boobis etal.

oral gavage exposure. Acetylcholinesterase (brain and
blood) activity and functional observations (cholinergic signs, gait, tail pinch, and motor activity) were
measured. The dose response for each of the individual
OPs was established in order to make predictions for
the mixture response, and the mixture was then tested
using a fixed ray design in which a percentage of the OP
in the mixture stayed constant but the total mg/kg of
the overall mixture increased. The data were modeled
using a threshold additivity model (Gennings et al.,
1997) for motor activity and brain cholinesterase (ChE)
acitivity, whereas for blood ChE activity the data fit a
threshold outside the experimental range, and a generalized linear model was used.
For the full ray treatments (mixtures of all five chemicals), all endpoints (brain and blood cholinesterase
activities, motor activity, and gait score) demonstrated
synergy by a factor that ranged from 1.2 to 2.1. With
malathion omitted, all but the gait score demonstrated
synergy magnitudes that ranged from 1.2 to 2.0. In all
cases synergy was observed even at the lower mixture
doses.
Moser etal. (2006) Moser etal. (2006) is an extension
of their 2005 study, using very young male pups dosed
on postnatal day 17. This work tested the hypothesis
that the synergy observed in adult rats would be greater
in very young rats, likely due to kinetic differences. All
measured endpoints were as in the previous study. The
analyses used included the single chemical required
method (Casey et al., 2004) to test the hypotheses of
additivity, or the flexible single chemicals required
method (Gennings et al., 2004a) to determine if the
response asymptotes were similar. At the lowest test
concentrations, doses of individual constituents were
0.03 (diazinon) to 36 (malathion) times their respective
NOAELs in the full ray mixture and 0.03 to 10 (chlorpyrifos) times their respective NOAELs in the reduced ray
mixture.
For the full-ray treatment (mixture of all five chemicals), all of the endpoints demonstrated synergy (synergy
magnitude ranging from 1.3 to 3.5). With malathion
omitted, some endpoints did not show synergy and the
remaining synergy magnitudes ranged from 1.3 to 2.3.
Crofton etal. (2005) Crofton etal. (2005) used a fix-ray
approach to evaluate the interaction among a mixture
of 18 thyroid-disrupting chemicals given to rats by oral
gavage. The objective of this work was to determine
whether the effects of a mixture of chemicals with
multiple mechanisms of action could be predicted
by dose-addition theory. The mixture consisted of 18
polyhalogenated aromatic hydrocarbons (2 dioxins, 4
furans, and 12 dioxin and nondioxin-like polychlorobiphenyls [PCBs]) in a ratio that reflected that found
in human breast milk and food. At the highest mixture
doses, the exposure of rats to each constituent chemical was at or below the known no observed effect level


(NOEL) for that chemical on serum thyroxine (T4) concentration. Tested dilutions of the mixture ranged to
100-fold less than this. The mixture was administered
daily for 4 days and 24 hours after the last dose, serum
T4 concentrations were measured. Using the flexible single-chemical required model (Gennings et al.,
2004), the single-chemical data, without and with the
mixture data, were modeled to determine both the
expected mixture response (the dose-additivity model)
and the experimentally observed mixture response (the
empirical model).
There was no evidence of synergy or antagonism
at the lowest doses of the tested mixture. At the three
highest doses, the empirical data were underpredicted
by the dose additivity models, i.e., there was a 2.5-fold
synergy. These results show that synergy can occur with
low doses (near the NOELs) of mixtures of thyroid hormone disrupting chemicals and that dose addition can
exist at low doses of chemicals that may act via different
mechanisms.
Korsak et al. (1988) Rotarod performance (an indicator of neuromuscular function and balance) was
evaluated in groups of 10 male Wistar rats that were
exposed by inhalation to toluene (10304850ppm),
xylene (10304970ppm), and a toluene/xylene mixture
(50% v:v of each, total concentration 10504700ppm)
for a 4-hour period. The investigators also evaluated
depression of respiratory rate secondary to respiratory irritation. Comparison with pre-exposure test
values showed concentration-related disturbances
of performance on the rotarod following exposure
to toluene and xylene individually and in combination. Probit analysis of exposure versus response
and statistical comparison of relative potency values
determined from the slopes of the exposure-response
curves were used to suggest a magnitude of synergy,
of about 1.5-fold, and the authors reported a statistically significant difference from the additive model
(see Table 1; 95% confidence interval [CI]: 1.22.0).
They did, however, provide sufficient information to
calculate the magnitude of synergy based on dose
(concentration) addition: the EC50s for the single
chemicals and for the 50:50 mixture. By comparing
the mixture EC50 (2770ppm) to that estimated by concentration addition (the harmonic mean, 4270ppm,
of the component values), an interaction magnitude
of 1.5 is obtained. No effect on rotarod performance
was detected for any of the 1000ppm exposures. At
2000ppm, the authors conclude that altered rotarod
performance was observed for the mixture, but not
for exposures to the individual chemicals at this concentration. ATSDR (2004b) concluded that this study
demonstrated a greater than additive response, while
recognizing that it utilized high exposure levels.
Subsequent to this study, Korsak etal. (1992) reported
the results of a subchronic study, in which rats were
exposed for 3 or 6 months via inhalation to toluene and
Critical Reviews in Toxicology

Literature analysis of low-dose synergy 377

m-xylene at concentrations of 1000ppm or 100ppm,
alone or in a 1:1 mixture. Endpoints assessed included
rotarod performance, spontaneous motor activity, and
counts of red blood cells and rod neutrophils. Rotarod
performance and spontaneous motor activity were statistically different from controls for the individual and
mixed exposures. Changes were more pronounced but
not significantly different for the mixed exposure as compared to single solvents. Significantly decreased counts
of red blood cells and significantly increased counts of
rod neutrophils were observed for the mixed exposure
but not individual exposures at 3 months only. No significant differences in final body weights, absolute and relative weights of organs, and clinical chemistry parameters
were observed as compared to controls. ATSDR (2004b)
concluded that the mixed exposure results of this study
were consistent with an additive effect of the individual
compounds.

obtained ranged from 1.7 to 2.2 for the binary mixtures

and was 3.2 for the ternary mixture.
The authors indicated that the greater interactive
effect of the ternary mixture as compared to the binary
mixtures may be related to one component in the binary
mixture being near or above its metabolic saturation
level. Individual chemicals were tested at exposures of
100 and 200ppm. Tested individually, the AUCs for the
200ppm concentrations were 45-fold greater than that
of the 100ppm concentration: toluene 3.17 vs. 0.69;
xylene 4.43 vs. 0.9; ethyl benzene 5.58 vs. 1.3. The authors
indicate that these data suggest that the metabolism of
these solvents may reach saturation at levels approaching 200ppm, resulting in higher blood concentrations
of unchanged solvents. For 300ppm mixed exposures,
the total AUC for the ternary mixture (9.2) fell within the
range of AUCs observed for the binary mixtures (range of
713.4, average 10.6).

3.2.2. Studies reporting synergy using Method B, the ratio of

observed to predicted response at a specific dose
Tardif et al. (1996) Tardif et al. (1996) evaluated the
toxicokinetic interaction magnitude for a combination of
toluene, ethylbenzene, and xylene by measuring concentrations of the parent chemicals in blood using the area
under the concentration time curve (AUC; also termed
integrated concentration) as the response measure.
Rats were exposed via inhalation for 4 hours to toluene,
xylene, and ethylbenzene individually or in binary or
ternary mixtures. For individual chemicals, exposures
of 100 and 200ppm were used. In the binary and ternary mixtures, the sum of component concentrations
was 300ppm. Individual compounds were not tested at
300ppm.
The magnitude of any interaction was determined
from the ratios of the expected AUC to the observed AUC
for the mixture. The expected AUC represented the sum
of the AUCs for the individual chemicals administered
alone, whereas the observed AUC for the mixture was
the sum of the individual AUCs when the chemicals were
administered together. If AUCTOL is the AUC for toluene
when administered alone, and AUCTOLm is the AUC for
toluene when administered as part of the mixture (and
similarly for the other two chemicals), then the expected
AUC under effect addition, AUCADD is

Dennison et al. (2005) Dennison et al. (2005) applied

the human pharmacokinetic model of Tardif et al.
(1997) for toluene, xylene, and ethylbenzene to occupational exposures for lightly exercising adults at
Occupational Safety and Health Organization (OSHA)
permissible exposure levels (PELs) (200, 100, and
100ppm, respectively). The response measure was
the peak chemical concentration in blood (modeled),
reached at the end of the 8-hour exposure period. The
mixture formula they used was the pharmacokineticsbased unity calculation, also termed the biological
hazard index (Haddad etal., 1999), which they denote
EMPK. The authors note that the EMPK is equivalent in
concept to the hazard index (US EPA, 2000), where the
8-hour peak blood level at the threshold limit value
(TLV) is used as the single chemical acceptable level.
For n chemicals in the mixture,

AUC ADD = AUCTOL + AUC EBZ + AUC XYL

The observed AUC for the mixture, AUCm, is then
AUCm = AUCTOLm + AUC EBZm + AUC XYLm
A ratio AUCm/AUCADD that is greater than 1 indicates that
the internal doses (AUC) of the parent chemicals are
greater from co-exposure than indicated by the single
chemical measurements. One-way analysis of variance
(ANOVA) and the Student- Newman-Keuls test were used
to determine significance (p <.05) of treatment differences for the solvent AUCs. The interaction (ratio) values
2011 Informa Healthcare USA, Inc.

Ci ELi
C
i =1 i OELi
n

EM PK =

where C is the blood concentration of the chemical, EL

is the external exposure level, and OEL is the permissible occupational external exposure level. Thus each C
is the concentration in blood calculated from the physiologically based pharmacokinetic model (PBPK) model
at the given external exposure level. The resulting EM
value is the interaction magnitude. When exposures were
controlled by a unity calculation based on PELs,1 internal exposure was calculated to be 2.9 and 4.6 times the
exposures at the TLVs at rest and workload, respectively.
When exposure was set at the PELs outright, the modeled
interaction magnitude for resting was 12.5 and for light
exercise was 16.
The authors cited several references for modeled
concentrations in blood that suggest that the potential
for interaction is dose dependent. They indicated the
Haddad etal. (1999, 2001) evaluations of toluene, ethylbenzene, and xylene mixtures at lower concentrations

378 A. Boobis etal.

predicted internal doses of 411% over unity using the
EM mixture formula. These predictions were for exposures in which each chemical was at approximately
one-third of its TLV (toluene TLV=50, ethylbenzene
TLV=100ppm, xylenes TLV =100ppm). Dennison etal.
also cited the Tardif etal. (1997) toluene, ethylbenzene,
m-xylene (TEX) model analysis that predicted that at
exposures of 2030ppm, the additivity assumption
holds, but at higher exposures a greater than additive
tissue dose occurs.
Chen and Eastmond (1995) Chen and Eastmond (1995)
evaluated a combination of hydroquinone and phenol
on the formation of bone marrowderived erythrocyte
micronuclei in mice. Mice were dosed via intraperitoneal (i.p.) injection at 24-hour intervals over 3 consecutive days. Femoral bone marrow cells were harvested 24
hours after the final dose. The frequency of bone marrow
erythrocytic micronuclei (MN) observed with the binary
combination was compared to the sum of the frequencies observed for the two single-chemical dose-response
results, i.e., effect addition. Statistical significance of
treatment effects was determined using linear and multiple regression on the square root transformed number
of MN (+0.5) per 2000 cells scored per mouse (critical
value set at 0.05 probability of Type I error). At the highest
dosage (100mg/kg phenol and 60mg/kg hydroquinone),
a 4-fold increase in the frequency of MN was observed
with the mixture. The interactive effect was most pronounced at higher doses of phenol (100 and 160mg/kg)
and hydroquinone (60mg/kg). Lesser degrees of interaction were observed at lower doses, with no interaction at
the lowest combined dose (50mg/kg phenol and 60mg/
kg hydroquinone). The authors also noted that there was
a significant decrease in the ratio of polychromatic/normochromatic erythrocytes (PCEs:NCEs) following coadministration. The frequency of MN appeared similar
to that with the combined administration of the higher
doses of hydroquinone and phenol, indicating myelotoxic
effects. The authors also reported that they attempted to
use higher doses, but this resulted in high animal mortality. The relevance of these high dose injection studies
to environmental exposures is therefore questionable
(statistical significance not provided).
Chen etal. (2004) Chen etal. (2004) examined the effects
of ingested arsenic and cigarette smoke on lung cancer
risk in >10,000 residents in Taiwan followed for 8 years.
Two approaches were used to calculate magnitude of
synergy: an etiologic fraction, and a synergy index. The
etiologic fraction is defined in their article as [RR11 RR01
RR10 + RR00]/RR11), with RR11 representing the relative
risk of the joint exposure, RR01 the relative risk of arsenic
exposure only, RR10 the relative risk of cigarette smoke
1
It is noted that the PEL is equivalent to the TLV for ethylbenzene and
xylenes (100ppm for all values); it is 4-fold higher than the TLV for toluene (200ppm vs. 50ppm).

exposure alone, and RR00 no exposure to either agent.

This fraction indicates the percentage of cases with both
exposures (arsenic and cigarette smoke) that was due
to synergy. Range of departure from additivity was estimated using the 95% confidence interval of the etiologic
fraction.
Three exposure categories were established for arsenic (<10, 10699, and 700 g/L) and for cigarette smoking (nonsmokers, <25 pack-years, or 25 pack-years). For
comparison, the drinking water standard for arsenic in the
United States is 10 g/L. Lung cancer risk was elevated by
either exposure: the RR was 2.21 (95% CI: 0.716.86) for
the highest exposure to arsenic in nonsmokers, while the
RR was 3.80 (95% CI: 1.2911.2) for the highest smoking
exposure in the low-arsenic area. The RR for the population in the two highest exposure categories was 11.1 (95%
CI: 3.3237.2). The synergy index was defined by the ratio
of excess risk with joint exposure (RR11 1) and the risk
predicted when assuming response additivity (the sum
of 2 excess risks with only exposure to a single risk factor [RR10 1] + [RR01 1]). The reported synergy indices
for the combinations of exposures (mid and high levels
of arsenic and two levels of cigarette smoke as compared
to the lowest arsenic level and nonsmokers) range from
1.62 to 2.52. This study is included due to the paucity of
low-dose synergy data, but it is noted that the exposure
levels for smoking alone exceeded PODs (i.e., RR > 1 for
either smoking exposure level).
Ito etal. (1991, 1998) Ito etal. studied the promotional
effect of five heterocyclic aromatic amines alone or in
one of two mixtures, in a DEN initiation/two-thirds
partial hepatectomy model. In the mixture exposures,
each component was either at 1/5 or 1/25 of its individual carcinogenic dose. Both the number and area of
hepatic GSTP+ foci were measured. To test additivity,
the foci values in controls were subtracted from each test
result before adding to determine predicted combined
effect. Magnitude of synergy was calculated as observed
response for a mixture (corrected for background)
divided by the sum of the responses for the individual
compounds in the mixture (again corrected for background). Ratios reported by the investigators were 1.1
for the 1/5 carcinogenic dose mixture and 107 for the
1/25 carcinogenic dose mixture. A number of concerns
have been expressed about the analysis and interpretation of these data, for example by the UK Committee
on Carcinogenicity (http://www.iacoc.org.uk/papers/
documents/CC0912COCSTATEMENTONMIXTURES.
pdf ). On the basis of the information available, the implications of these findings for toxicity or cancer risk are
unclear (as discussed in Appendix C).
Meyer etal. (1994) Meyer etal. (1994) investigated mouse
skin tumor promotion with a combination of mirex (an
organochlorine pesticide) and 12-O-tetradecanoylphorbol-13-acetate (TPA) applied to 7,12-dimethylbenz(a)
anthracene (DMBA)-initiated female CD-1 mouse skin.
Critical Reviews in Toxicology

Literature analysis of low-dose synergy 379

Mice were exposed to 200 nmol mirex plus varying doses
of TPA (2, 5, and 8 nmol) or 200 nmol mirex alone or TPA
alone at doses of 2, 5, 8, 12, and 16 nmol. The number
of tumors per mouse after 20 weeks of exposure to the
TPA/mirex mixture, versus TPA or mirex alone, was the
response measure. The measure of interaction was the
ratio of observed tumors per mouse from the mixture
to that predicted from response addition (sum of the
tumors per mouse from each chemical given separately).
A response interaction magnitude of 1.9 was calculated
based on the ratio of total skin tumors/mouse. The apparent 1.9-fold interaction magnitude in the mice treated
with 200 nmol mirex and 2 nmol TPA produced 37% mortality over the 20-week treatment period. This compares
to no deaths observed among mice treated with acetone
(control) and then with 200 nmol mirex combined with 2
nmol TPA. Thus, the apparent 1.9-fold greater than additive response was associated with significant treatmentrelated mortality. At the higher TPA dosages (5 and 8
nmol) in combination with 200 nmol mirex, a reduction
in tumor load was observed. The authors attributed the
reduced tumor development with the higher TPA concentrations to dermal cytotoxicity.

4. Discussion
4.1. Critical evaluation of reported magnitude of
synergy

additivity, i.e., synergy in mixtures, and reported quantified values for the magnitude of synergy. Most of the
studies used doses that caused overt toxic effects from
acute exposures and thus were at exposure levels likely
to be higher than those associated with the PODs for
chronic effects of the components of the mixtures. The
overall finding of the current work is that the magnitude
of synergy at low doses in these studies ranged from 1.5 to
3.5. Within the studies, there was appreciable variability
with respect to how the definition, measurement, and
calculation of synergy was reported. As discussed above,
with the exception of Korsak etal. (1988), no attempt was
made to recalculate magnitudes of synergy. However, the
methods used in the publication were evaluated and, if
justified, respective entries were removed from the final
table of values of synergy magnitude.
As shown in Table 1, the 11 studies that reported synergy magnitudes were critically evaluated and a determination was made as to whether or not they should be
included in a final summary of magnitudes of synergy. A
brief description of the conclusion of this assessment is
provided in the far right-hand column of the table; where
a study was excluded upon critical evaluation, more
detailed discussion including the basis for exclusion is
included in Appendix C (Supplementary Information).

4.2. Overall conclusions/future directions

This literature review identified a limited number of studies that measured responses other than lethality, were
purposely designed to examine the question of supra-

Understanding the role of interactions amongst chemicals that might increase or decrease overall joint toxicity
of mixtures continues to be subject to debate by scientists and within the risk assessment community. Several

4
3.5

Interaction magnitude

3
2.5
2
1.5
1
0.5

duration
route
species
endpoint

Moser et al Moser et al Korsak et al

(2005)
(2006)
(1988)

Crofton et al
(2005)

Meyer et al
(1994)

4hr
oral
rat
ED20
motor
activity, gait
score, ChE
levels

4 days
oral
rat
ED30
serum T4
level

20 wks
dermal
mouse
# skin
tumors

4hr
oral
rat
ED20
motor
activity, gait
score, ChE
levels

4hr
inhalation
rat
ED50
rotatod
performance

Chen et al
(2004)
epi study
multi
human
# lung cancer
cases

Figure 3. Reported magnitudes of synergy from literature meeting inclusion criteria. Where shown, the bars indicate the range of values
obtained in the study, most often reflecting different interaction magnitudes at different doses.
2011 Informa Healthcare USA, Inc.

380 A. Boobis etal.

methods for assessing the joint toxicity of mixtures have
been proposed, developed, and published, and, in some
cases, have been incorporated in the official risk assessment guidelines of various organizations and government agencies. However, agreement on the impact of
interactions following exposures to chemical mixtures is
largely incomplete, and concern has been raised about
the possibility that exposure to low levels of multiple
chemicals could lead to enhanced toxicity as a result of
synergistic interactions.
This work represents an in-depth analysis of documented examples of synergy in the published literature.
A key objective of this analysis was to identify the greatest magnitudes of synergy reported in these low-dose
studies to evaluate the potential impact on screeninglevel approaches such as the TTC for the assessment of
mixtures.
Considerable effort was devoted to the search for
publications on synergy over the last 20 years. The search
included reports in journals, the gray literature, and government publications. Ninety peer-reviewed published
studies were identified that reported results on combinations of 204 different chemicals but many of these studies
did not contain sufficient information to enable any definite conclusions regarding synergy, its characteristics,
or magnitude to be drawn. After excluding questionable
studies, only 43 publications were found to contain novel
data on synergy in mammals. These reported synergy in
studies of 126 chemicals using conventional in vivo toxicological methodologies and endpoints. These studies
varied widely in exposure route, duration of exposure,
and the number of component chemicals in the mixture
as well as in the approaches used to identify interactions,
express interaction potential, and calculate magnitudes
of interaction.
The literature search was not able to provide quantitative information on the frequency of the occurrence
of synergy for several reasons. First, the search terms
were based on the occurrence of, or a consideration, of
synergy. Studies of the toxicity of mixtures that do not
mention synergy (or related terms) would not have been
identified by this strategy. As a result, there is no way of
determining the proportion of studies of the toxicity of
mixtures where synergy did not occur, because unless
this was explicitly reported, it would have been captured
by the search strategy. A second problem concerns the
possibility of publication bias, where a clear finding
of synergy would be expected to greatly increase the
probability of publication over results showing additivity or independent action. It is not known whether the
significant but opposite interaction, antagonism, would
similarly increase the chance of publication. Such uncertainties about underreporting of negative findings, common in the health literature, complicates any attempt
to estimate the likelihood of synergy in environmental
mixtures. Third, in many instances where synergy was
discussed, it is not clear whether true synergy had actually occurred. This uncertainty resulted from a variety of


factors, including insufficient reporting of details and/or

limitations in design of the study or the survey. Lastly, it
should be noted that the breadth of the survey was somewhat limited. The search included only publications from
1990 to the present and because of the specific limitations of the search (see Section 2.1) resulted in studies on
a small number of chemical classes.
The studies in the literature used a variety of methods
for assessing the occurrence and magnitude of synergy.
In some cases, synergy was observed only for endpoints
where the toxicological relevance was unclear (e.g., concentrations in blood or hair). In other cases, the authors
used ratios of doses at overt toxic effect levels (e.g., LD50
values). It is difficult to evaluate such findings in a risk
assessment context because of issues related to extrapolation from high doses and severe endpoints to the more
relevant low doses and less severe endpoints that are
generally of concern in an assessment of the risks to
human health. Consistent approaches should be applied
in the quantification of synergy, including that synergy
be defined in terms of departure from dose additivity;
uniform procedures be developed for assessing synergy
at low exposures; and the method for determining the
POD (e.g., benchmark response for equitoxic dose) for
calculating synergy be standardized.
After a thorough and careful evaluation of the six studies that provided useful quantitative estimates, the magnitude of synergy at low dose did not exceed the levels
predicted by additive models (Figure 3) by more than a
factor of 4. This finding is significant because it gives the
first empirical insight into the magnitude of interactions
that have been found to occur in experimental studies.
Specifically, this literature search for low-dose synergy
found evidence for the existence of at most a modest
(<4-fold) increase in toxicity, or decrease in equitoxic
dose, from six toxicological studies of defined mixtures
ranging from 2 to 18 chemicals. No evidence was found
for the existence of large, say >10-fold, increases in toxicity
from chemical interactions. This finding is strengthened
by the level of effort performed in identifying positive
studies. However, the small number of studies in which
synergy was quantified also suggests that the application
of statistical and other mathematical methods is infrequent in mixture studies.
In considering the implications of the findings of this
review, a number of factors need to be considered. First,
in most of the positive studies, mixtures were composed
of chemicals expected, or known, to affect a common target or to operate by a common mechanism; therefore, it
is not clear how these findings would apply to mixtures of
chemicals operating by a variety of mechanisms. Second,
in most of the positive studies, the doses of the individual
mixture components exceeded their respective PODs,
requiring at least a qualitative extrapolation of findings
of mixture interaction down to the range of the POD.
Third, in a number of the positive studies, the occurrence
of synergy was dose dependent and observed only at the
higher doses used in the study.
Critical Reviews in Toxicology

Literature analysis of low-dose synergy 381

Despite these caveats, there is probably merit in the
default regulatory approaches that assume toxicological interactions are not likely to occur at the low doses
permitted under existing exposure standards. However,
it is too early to draw firm conclusions, particularly for
cumulative and low level chronic exposures.
Measuring synergy presents a conundrum for toxicologists. On the one hand, the goal is to determine the
impact of synergy on the PODs for chronic exposures to
mixtures in animal models; it is changes in these lowdose findings that are most relevant to setting regulatory standards. However, the costs of long-term studies
of mixtures in animals are so great that such research is
rarely feasible except for the highest priority mixtures.
Studies of mixtures using short-term dose regimens,
however appealing, may make extrapolation of results
to more realistic long-term exposures difficult. A second problem is that accurate measurements of deviations from additivity are difficult to perform for doses
close to the NOAEL where the majority of the animals
may show no dose related effects. As a result, synergy
is often defined in terms of changes in the ED20 or ED50
and so, it is not surprising that the doses used in such
studies are greater, sometimes much greater, than the
POD.
The solution to this problem will not be found by continuing the current practice of testing mixtures at high
doses. Rather, new techniques and short-term assays are
required to detect and evaluate the low-level molecular
and cellular changes that are precursors to chronic toxicity (Teuschler etal., 2002; Yang etal., 2004). Techniques,
such as toxicogenomics, proteomics, and signaling pathway assessments could provide insights into changes
that take place in the cell following chronic exposures.
Predictive in silico modeling studies, such as those
addressing pharmacokinetic interactions, also show
promise for characterizing interaction potential as long
as tissue concentrations of parent chemicals and relevant
metabolites can be linked to toxic endpoints. In order to
take advantage of this potential for understanding the
role, nature and relevance of interactions, future studies will need to be designed specifically for the assessment of interactions that might occur following chronic
exposure.
This review has highlighted a number of issues that
should be considered when designing mixtures studies
for use in hazard assessment. First and foremost, dose
ranges of the mixture that bracket the PODs of the mixture components should be used whenever possible.
This may help alleviate possible confounding due to
saturation of kinetic and dynamic processes that occur
only at high doses. Dose- or concentration-response
curves for the individual chemicals should contain sufficient data points above and below the no-effect level to
establish a practical threshold (if applicable to the measured endpoint) and sufficient data points to enable the
statistical fit of a dose-response curve. Second, for highpriority mixtures of regulatory concern, the proportion
2011 Informa Healthcare USA, Inc.

of chemicals within the mixture should be based on the

fractional amounts of constituent chemicals in the environmental exposures being regulated. For example, coexposure to TCCD and PCB 153 almost always occurs,
but the amounts are very different with PCB 153 concentrations routinely being orders of magnitude higher
than TCDD (Crofton et al., 2005). Such constraints on
the experiment, however, can substantially limit the
extent to which test results can be generalized to any
mixture (arbitrary fractions) of those component chemicals. Third, use of rigorous statistical models are critical
for testing hypotheses of effect- or dose-addition, and
determining whether antagonism or synergism exists
(Feron and Groten, 2002; Casey et al., 2004; Gennings
et al., 2004; Christiansen et al., 2009). Finally, use of
mechanistically based mathematical models is strongly
encouraged in order to understand why the observed
interaction occurs and how it might change under different exposure conditions.
In summary, an intensive literature search for positive findings of low-dose synergy located only a small
number of studies. After a thorough and careful evaluation of the six studies that provided useful quantitative
estimates of synergy, it was found that synergy at low
doses resulted in only modest increases in toxicity, with
a magnitude that did not exceed the levels predicted by
additive models by more than a factor of 4. A recent study
by Christiansen et al. (2009) examining the effects of a
mixture of four antiandrogens on male sexual development in the rat demonstrated a magnitude of synergy
(<4-fold) consistent with the range reported from our literature search. A more detailed description of this study
is reported in Appendix C. In addition, a recent Opinion
of the European Food Safety Authority (EFSA) Panel
on Plant Protection products and their Residues (PPR)
concluded from a limited review that significant toxic
interactions between chemicals are much less likely to
occur at doses below the effect levels for individual component compounds than at higher doses. (EFSA, 2008).
The examples identified in this review are not sufficient
to support robust conclusions regarding the frequency
with which synergy is likely to occur following low-dose
chronic exposures; however, these results might indicate
an upper bound estimate when screening of untested
mixtures for synergistic potential. A different approach is
to investigate an interaction threshold, defined as the
boundary that separates the dose response space into
regions of interaction and additivity (Hamm etal., 1995).
The actual experimentally determined interaction
threshold requires detailed data not readily available
for chemical mixtures. Hence, modeling techniques have
been proposed to estimate the interaction thresholds of
mixtures (El-Masri etal., 2004).
Overall, the results of this analysis provide support for
the feasibility of developing the TTC as a screening-level
risk assessment in a tiered approach for the assessment
of combined exposures to chemicals. The magnitude of
any synergy appears to be low, but more work is required

382 A. Boobis etal.

to determine how frequently synergy is likely to occur at
real world exposures.

Acknowledgements
The authors would like to thank Drs. Roger Meyerhoff
and Joel Bercu (Eli Lilly) for valuable review, input, and
edits and Drs. Jason Lambert (US EPA) and Ed Carney
(Dow Chemical) for their thoughtful reviews.

Declaration of interest
The employment affiliations of the authors are shown on
the cover page. These individuals had the sole responsibility for the writing and content of the paper. The
individual authors worked as professionals in preparing
the article and not as agents of their employers. The literature review used as the basis of this article was performed by three of the authors (R.H., S.C., and D.K.) and
funded by the HESI Mixtures Committee, which collects
funding from member companies to support the project. Four of the authors (A.B., D.K., K.S., and R.Y.) are
affiliated with universities, two authors (R.H. and S.C.)
are independent consultants providing services to public and private organizations, three of the authors* are
affiliated with government agencies, one author (M.E.)
is affiliated with a nonprofit organization and six of the
authors (R.B., S.F., G.M., P.P., and R.Z.) are employed by
private corporations. Government and academic committee participants were reimbursed for travel expenses
to attend committee meetings and did not receive any
other compensation. (*The views expressed in this paper
are those of the authors and do not necessarily reflect
the opinions or policy of the US EPA or the Centers for
Disease Control, ATSDR.)

References
Agency for Toxic Substances and Disease Registry (ATSDR). (2004a).
Assessment of joint toxic action of chemical mixtures. Available at:
http://www.atsdr.cdc.gov/interactionprofiles/ipga.html.
Agency for Toxic Substances and Disease Registry (ATSDR). (2004b).
Interaction profile for: benzene, toluene, ethylbenzene, and xylenes.
Available
at:
http://www.atsdr.cdc.gov/interactionprofiles/
IP-btex/ip05.pdf
Barlow S. (2005). Threshold of Toxicological Concern (TTC)A Tool
for Assessing Substances of Unknown Toxicity Present at Low
Levels in the Diet. Washington, DC: International Life Sciences
Institute.
Boobis A, et al. (2006). IPCS framework for analysing the relevance
of a cancer mode of action for humans. Toxicol Lett 164(Suppl
1):S254S255.
Carpy SA, Kobel W, et al. (2000). Health risk of low-dose pesticides
mixtures: A review of the 19851998 literature on combination
toxicology and health risk assessment. J Toxicol Environ Health B
3:125.
Casey M, Gennings C, Carter JWH, Moser VC, Simmons JE. (2004).
Detecting interaction(s) and assessing the impact of component
subsets in a chemical mixture using fixed-ratio mixture ray designs.
J Agricult Biol Environ Stat 9:339361.
Chen C-L Hsu, L-I, Chiou H-Y Hsueh, Y-M, Chen S-Y Wu, M-M, etal.
(2004). Ingested arsenic, cigarette smoking, and lung cancer risk:


A follow-up study in arseniasis-endemic areas in Taiwan. JAMA

292:29842990.
Chen H, Eastmond DA. (1995). Synergistic increase in chromosomal
breakage within the euchromatin induced by an interaction of
the benzene metabolites phenol and hydroquinone in mice.
Carcinogenesis 16:19631969.
Christiansen S, Scholze M, Dalgaard M, Vinggaard AM, Axelstad M,
Kortenkamp A, Hass U. (2009). Synergistic disruption of external
male sex organ development by a mixture of four antiandrogens.
Environ Health Perspect 117:18391846.
Cramer GM, Ford RA, Hall RL. (1978). Estimation of toxic
hazardA decision tree approach. Food Cosmet Toxicol 16:
255276.
Crofton KM, Craft ES, Hedge JM, Gennings C, Simmons JE, Carchman
RA, et al. (2005). Thyroid-hormone-disrupting chemicals:
Evidence for dose-dependent additivity or synergism. Environ
Health Perspect 113:15491554. Available online at: http://www.
ehponline.org/members/2005/8195/8195.pdf.
Dennison JE, Bigelow PL, Mumtaz MM, Andersen ME, Dobrev ID, Yang
RSH. (2005). Evaluation of potential toxicity from co-exposure to
three CNS depressants (toluene, ethylbenzene, and xylenes) under
resting and working conditions using PBPK modeling. J Occup
Environ Hyg 2:127135.
El-Masri HA, Mumtaz MM, Yushak ML. (2004) Application of
physiologically based pharmacokinetic modeling to investigate
the toxicological interaction between chlorpyrifos and parathion
in the rat. Environ Toxicol Pharmacol 16:5771.
European Food Safety Authority. (2008). Opinion of the Scientific
Panel on Plant Protection Products and their Residues to evaluate
the suitability of existing methodologies and, if appropriate,
the identification of new approaches to assess cumulative and
synergistic risks from pesticides to human health with a view
to set MRLs for those pesticides in the frame of Regulation (EC)
396/20051. EFSA J 704:1284.
Feron VJ, Groten JP. (2002). Toxicological evaluation of chemical
mixtures. Food Chem Toxicol 40:825839.
Gennings C, Carter WH Jr, Carney EW, Charles GD, Gollapudi BB,
Carchman RA. (2004). A novel flexible approach for evaluating
fixed ratio mixtures of full and partial agonists. Toxicol Sci 80:
134150.
Goldsworthy TL, Hanigan MH, Pitot HC. (1986). Models of
hepatocarcinogenesis in the ratContrasts and comparisons.
CRC Crit Rev Toxicol 17:6189.
Haddad S, Tardif R, Charest-Tardif G, Krishnan K. (1999). Physiological
modeling of the toxicokinetic interactions in a quaternary
mixture of aromatic hydrocarbons. Toxicol Appl Pharmacol 161:
249257.
Hamm AK, Carter H Jr, Gennings C. (1995). Analysis of an interaction
threshold in a mixture of drugs and/or chemicals. Stat Med
24:24932507.
Hasegawa R, Miyata E, Futakuchi M, Hagiwara A, Nagao M, Sugimura
T, Ito N. (1994). Synergistic enhancement of hepatic foci
development by combined treatment of rats with 10 heterocyclic
amines at low doses. Carcinogen 15:10371041.
Hertzberg RC, MacDonell MM. (2002). Synergy and other ineffective
mixture risk definitions. Sci Total Environ 288:3142.
Hertzberg RC, Teuschler LK. (2002). Evaluating quantitative formulas
for dose-response assessment of chemical mixtures. Environ
Health Perspect 110(Suppl 6):965970.
Hirose M, Takesada Y, Tanaka H, Tamano S, Kato T, Shirai T. (1997).
Carcinogenicity of antioxidants BHA, caffeic acid, sesamol,
4-methoxyphneol and catechol at low doses, either alone or in
combination, and modulation of their effects in a rat mediumterm multi-organ carcinogenesis model. Carcinogen 19:207212.
Ito N, Hasegawa R, Shirai T, Fukushima S, Hakoi K, Takaba K, et al.
(1991). Enhancement of GST-P positive liver cell foci development
by combined treatment of rats with five heterocyclic amines at low
doses. Carcinogenesis 12:767772.
Ito N, Imaida K, Hirose M, Shirai T. (1998). Medium-term bioassays
for carcinogenicity of chemical mixtures. Environ Health Perspect
Critical Reviews in Toxicology

Literature analysis of low-dose synergy 383

106(Suppl 6):13311336. Available online at: http://www.
ehponline.org/members/1998/Suppl-6/1331-1336ito/ito-full.
html.
Kodell R, Pounds J. (1991). Assessing the toxicity of mixtures of
chemicals. In: Krewski D, Franklin C, eds. Statistics in Toxicology.
New York: Gordon and Breach, 559591.
Knemann WH, Pieters MN. (1996). Confusion of concepts in mixture
toxicology. Food Chem Toxicol 34:10251031.
Korsak Z, Sokal J, Dedyk A, Tomas T, Jedrychowski R. (1988).
Toxic effects of combined exposure to toluene and xylene
in animals. I. Acute inhalation study. P J Occup Med 1:
4550.
Korsak Z, Sokal J, Gorny R. (1992). Toxic effects of combined exposure
to toluene and m-xylene in animals. IIII. Subchronic inhalation
study. P J Occup Med 5:2733.
Kroes R, Renwick AG, Cheeseman M, Kleiner J, Mangelsdorf I, Piersma
A, Schilter B, Schlatter J, van Schothorts F, Vos JG, Wurtzen G.
(2004) Structure-based Thresholds of Toxicological Concern (TTC):
Guidance for application to substances present at low levels in the
diet. Food Chem Toxicol 42:6583.
Lester RR, Green LC, Linkov I. (2007). Site-specific applications of
probabilistic health risk assessment: Review of the literature since
2000. Risk Anal 27:635658.
Meyer SA, Kim TW, Moser GJ, Monteiroriviere NA, Smart RC. (1994).
Synergistic interaction between the nonphorbol ester-type
promoter mirex and 12-o-tetradecanoylphorbol-13-acetate in
mouse skin tumor promotion. Carcinogenesis 15:4752.
Moser VC, Casey M, Hamm A, Carter WH Jr, Simmons JE, Gennings
C. (2005). Neurotoxicological and statistical analyses of a mixture
of five organophosphorus pesticides using a ray design. Toxicol Sci
86:101115.
Moser VC, Simmons JE, Gennings C. (2006). Neurotoxicological
interactions of a five-pesticide mixture in preweanling rats. Toxicol
Sci 92:235245.

2011 Informa Healthcare USA, Inc.

Mumtaz MM, Durkin PR. (1992). A weight of evidence scheme for

assessing interactions in chemical mixtures. Toxicol Ind Health
8:377 406.
Munro I, Ford RA, Kennepohl E, Sprenger JG. (1996). Correlation
of a structural class with no-observed-effect levels: A proposal
for establishing a threshold of concern. Food Chem Toxicol
34:829867.
Sonich-Mullin C, et al. (2001). IPCS conceptual framework for
evaluating a mode of action for chemical carcinogenesis. Regul
Toxicol Pharmacol 34:146152.
Tardif R, Charest-Tardif G, Brodeur J. (1996). Comparison of the
influence of binary mixtures versus a ternary mixture of inhaled
aromatic hydrocarbons on their blood kinetics in the rat. Arch
Toxicol 70:405413.
Tardif R, Charest-Tardif G, Brodeur J, Krishnan K. (1997).
Physiologically based pharmacokinetic modeling of a ternary
mixture of alkyl benzenes in rats and humans. Toxicol Appl
Pharmacol 144:120134.
Teuschler L, Klaunig J, Carney E, Chambers J, Conolly R, Gennings C,
Giesy J, Hertzberg R, Klaassen C, Kodell R, Paustenbach D, Yang
R. (2002). Support of science-based decisions concerning the
evaluation of the toxicology of mixtures: A new beginning. Regul
Toxicol Pharmacol 36:3439.
UK Committee on Carcinogenicity of Chemicals In Food, Consumer
Products, and the Environment. (2008). Meeting minutes. Available
at: http://www.iacoc.org.uk/meetings/documents/201108.pdf.
US EPA. (2000). Supplementary Guidance for Conducting Health Risk
Assessment of Chemical Mixtures. EPA/630/R-00/002. Washington,
DC: Risk Assessment Forum.
Yang RSH, El-Masri HA, Thomas RS, Dobrev ID, Dennison JE, Bae
D-S Campain, JA, Liao KH, Reisfeld B, Andersen ME, Mumtaz
M. (2004). Chemical mixture toxicology: From descriptive to
mechanistic, and going on to in silico toxicology. Environ Toxicol
Pharmacol 18:6581.

Copyright of Critical Reviews in Toxicology is the property of Taylor & Francis Ltd and its content may not be
copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.