Review of toxicity of chemical mixtures: Theory, policy and regulatory practice

University of Florida, Gainesville, Florida, United States
Regulatory Toxicology and Pharmacology (Impact Factor: 2.03). 08/2006; 45(2):119-43. DOI: 10.1016/j.yrtph.2006.03.004
Source: PubMed
An analysis of current mixture theory, policy, and practice was conducted by examining standard reference texts, regulatory guidance documents, and journal articles. Although this literature contains useful theoretical concepts, clear definitions of most terminology, and well developed protocols for study design and statistical analysis, no general theoretical basis for the mechanisms and interactions of mixture toxicity could be discerned. There is also a poor understanding of the relationship between exposure-based and internal received dose metrics. This confounds data interpretation and limits reliable determinations of the nature and extent of additivity. The absence of any generally accepted classification scheme for either modes/mechanisms of toxic action or of mechanisms of toxicity interactions is problematic as it produces a cycle in which research and policy are interdependent and mutually limiting. Current regulatory guidance depends heavily on determination of toxicological similarity concluded from the presence of a few prominent constituents, assumed from a common toxicological effect, or presumed from an alleged similar toxic mode/mechanism. Additivity, or the lack of it, is largely based on extrapolation of existing knowledge for single chemicals in this context. Thus, regulatory risk assessment protocols lack authoritative theoretical underpinnings, creating substantial uncertainty. Development of comprehensive classification schemes for modes/mechanisms of toxic action and mechanisms of interaction is needed to ensure a sound theoretical foundation for mixture-related regulatory activity and provide a firm basis for iterative hypothesis development and experimental testing.


Available from: Christopher J Borgert, Jun 30, 2015
Regulatory Toxicology and Pharmacology 45 (2006) 119–143
0273-2300/$ - see front matter © 2006 Elsevier Inc. All rights reserved.
Review of the toxicity of chemical mixtures: Theory, policy,
and regulatory practice
L.S. McCarty
, C.J. Borgert
L.S. McCarty ScientiWc Research & Consulting, 94 Oakhaven Drive, Markham, Ont., Canada L6C 1X8
Applied Pharmacology & Toxicology Inc., 2250 NW 24th Street, Gainesville, FL, USA
Center for Environmental and Human Toxicology, College of Veterinary Medicine, University of Florida, Gainesville, FL, USA
Received 18 October 2005
Available online 15 May 2006
An analysis of current mixture theory, policy, and practice was conducted by examining standard reference texts, regulatory guidance
documents, and journal articles. Although this literature contains useful theoretical concepts, clear deWnitions of most terminology, and
well developed protocols for study design and statistical analysis, no general theoretical basis for the mechanisms and interactions of mix-
ture toxicity could be discerned. There is also a poor understanding of the relationship between exposure-based and internal received dose
metrics. This confounds data interpretation and limits reliable determinations of the nature and extent of additivity. The absence of any
generally accepted classiWcation scheme for either modes/mechanisms of toxic action or of mechanisms of toxicity interactions is prob-
lematic as it produces a cycle in which research and policy are interdependent and mutually limiting. Current regulatory guidance
depends heavily on determination of toxicological similarity concluded from the presence of a few prominent constituents, assumed from
a common toxicological eVect, or presumed from an alleged similar toxic mode/mechanism. Additivity, or the lack of it, is largely based
on extrapolation of existing knowledge for single chemicals in this context. Thus, regulatory risk assessment protocols lack authoritative
theoretical underpinnings, creating substantial uncertainty. Development of comprehensive classiWcation schemes for modes/mechanisms
of toxic action and mechanisms of interaction is needed to ensure a sound theoretical foundation for mixture-related regulatory activity
and provide a Wrm basis for iterative hypothesis development and experimental testing.
© 2006 Elsevier Inc. All rights reserved.
Keywords: Mixture; Toxicity; Mode; Mechanism; Interaction; Risk; Regulations; Exposure; Organic chemicals
1. Introduction
The companion paper (McCarty and Borgert, this issue)
examined published research on complex mixture toxicity
where at least one component was a chlorinated organic
chemical. This second paper focuses on the theoretical
aspects of and interpretive approaches to mixture toxicity,
with emphasis on applications of this knowledge for regula-
tory policy and practices.
1.1. Goals and tasks
This review has two goals: examination of current views
of and approaches to mixture toxicity, including identiWca-
tion and analysis of key scientiWc literature pertaining to
the theoretical aspects of mixture toxicology, and a review
of how mixture toxicity theory and data are employed in
current mixture regulations and associated science-based
policy. For these goals, a representative, but not exhaustive,
sample of general toxicological reference works, mode/
mechanism of toxic action classiWcation publications, liter-
ature on toxicity statistics and models, and regulatory doc-
uments were examined.
Corresponding author.
E-mail address: (L.S. McCarty).
Page 1
120 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
As was the case in the companion paper, attention was
focused on several speciWc areas: exposure-dose relation-
ships and how mixture concentrations aVect dose metrics,
chemical characteristics of toxicologically similar mixtures
and whether toxicological similarity appears to be predict-
able on the basis of the chlorinated compounds in the mix-
ture, relationships between toxicological thresholds and the
concentration of the most toxic constituents of the mix-
tures, and relationships between mixture concentration and
mode of action.
2. Review of general approaches to and views of mixture
Important views and opinions regarding mixture toxicity
and interaction classiWcation have been summarized from
major toxicological reference books, regulatory guidance
from various government agencies, and nationally promi-
nent scientiWc organizations in the USA and Europe. Fol-
lowing this is an examination of views from the general
scientiWc literature subdivided into three topic areas: gen-
eral methods/review/commentary, mode/mechanism classi-
Wcation, and statistics and models.
International agencies such as the Organization for Eco-
nomic Co-operation and Development (OECD) and the
International Agency for Research on Cancer (IARC) also
consider mixture toxicity. As explained in more detail later,
the OECD provides a focus for coordinating mixture classi-
Wcation schemes from various national governments, but
this approach is hazard-based rather than risk-based, and is
primarily regulatory rather than toxicological in focus.
IARC recognizes mixtures but does not employ a mixture
classiWcation scheme; rather it examines speciWc deWned
mixtures or products for carcinogenicity (IARC, 1987).
2.1. Toxicological reference books
Hayes (2001) in the book “Principles and Methods of
Toxicology”, fourth edition, provides a brief overview of
the interactions that occur with exposure to multiple chem-
icals. Although noting that more complicated schemes have
been proposed, it recommends the use of a basic mixture
interaction classiWcation scheme with categories of addi-
tion, antagonism and synergism. The importance of
absorption, distribution, metabolism and excretion in
determining interactions is stressed because one chemical
can aVect either the delivery of another chemical to the site
of toxic action or interfere with reactions at the site. It also
cautions that the timing of exposures and the nature of the
biological eVect elicited can both act and interact diVerently
in mixtures depending on the doses of the components. This
text concludes that with the exception of some pesticides
“ƒ little information is available on toxic mechanisms.”
Based upon an examination of a number of large mixture
studies, the authors concluded that synergism is likely to
occur less frequently at typical environmental levels than at
the higher exposure levels often encountered in therapeutic
or occupational situations. The book also includes a short
overview of the typical regulatory approaches—hazard
index, toxicity equivalency factors, and treating some com-
plex mixtures (e.g. PCBs) as a single substance—used to
facilitate decision-making when the mechanistic toxicologi-
cal knowledge is poor.
In the book “Casarett & Doull’s Toxicology: The Basic
Science of Poisons,” sixth edition (Klaassen, 2001), the sec-
tion on mixtures includes a brief description and review of
the types of interactions that can occur: addition, syner-
gism, potentiation, and functional, chemical and disposi-
tional antagonism. As noted in McCarty (2002), the fourth
edition (Amdur et al., 1991) provided a mixture classiWca-
tion scheme based on eVects on various biochemical pro-
cesses and pathways while the Wfth edition (Klaassen et al.,
1996) used a mixture classiWcation scheme based on a com-
bination of broad eVect types (e.g. cancer, immunotoxicity)
and organ systems.
Krieger (2001) in the book “The Handbook of Pesticide
Toxicology,” second edition, an update to comprehensive
text on pesticides by
Hayes and Laws (1990), includes a
thorough review of the basics of toxicology, primarily from
a mammalian toxicology perspective. The general descrip-
tions of mode/mechanisms of toxic action are at a relatively
gross level involving only broad groupings of types of
changes (general lethal toxicity, neurotoxicity, carcinoge-
nicity, mutagenicity, teratogenicity, immune reactions, etc.).
Mixture toxicity is discussed more in an indirect way with
reference to mode/mechanism and interactions. This text
describes compound interactions in conventional terms
(simple addition, antagonism, potentiation, etc.) and
includes a discussion of the means by which such interac-
tions occur (enzyme induction, kinetics, dynamics, etc.) and
various types of toxicity-modifying factors. Reference to
mode/mechanism of action is made for various pesticides
but this is not done consistently and there is no general uni-
fying discussion of mode/mechanism of action. Pesticides
are grouped for discussion primarily according to chemical
structure, with some groups formed from closely related
chemicals that produce toxicity in a similar manner (e.g.
organophosphates, pyrethroids, etc.).
“Loomis’s Essentials of Toxicology,” fourth edition
(Loomis and Hayes, 1996) provides a good coverage of the
broad scope of modern toxicology, including the inXuence
of various exposure routes, various modifying factors, and
a classiWcation scheme for mechanisms of harmful eVects of
substances. However, it does not address the issue of mix-
tures in any comprehensive manner other than a substan-
tial thread that focuses on known issues and examples with
Lu (1996) in “Basic Toxicology: Fundamentals, Target
Organs, and Risk Assessment,” third edition, examines tox-
icology using four broad main topic areas: principles, test-
ing, target organs/systems, and groups of toxic substances.
Mixture toxicity does not receive a detailed examination,
although the modifying eVect caused by interactions due to
exposure to multiple chemicals is covered brieXy. Limited
Page 2
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 121
information is presented on classifying carcinogenic and
non-carcinogenic modes of action.
In Chapter 1 of “Fundamentals of Aquatic Toxicology
II” (Rand et al., 1995), the basics of toxicology are
reviewed, primarily from an aquatic toxicology perspective,
and a discussion of the inXuence of various modifying fac-
tors, such as metabolism, body size, etc. is included. A
quantitative structure–activity relationship (QSAR)-
derived approach with a simple physiologically based phar-
macokinetic/pharmacodynamic (PBPK) model is used to
interpret standard LC50 and chronic aquatic toxicity test
data and estimate a received whole-body dose or body resi-
due based on McCarty and Mackay (1993). Modes of
action are classiWed using the Wsh acute toxicity syndrome
(FATS) scheme proposed by researchers at the US EPA-
Duluth laboratory (narcosis, polar narcosis, respiratory
uncoupler, acetylcholinesterase inhibitor, membrane irri-
tant, CNS seizure agent, respiratory blocker; see Russom
et al., 1997) plus a TCDD category. This scheme illustrates
that there are characteristic body residues dose ranges for
acute and chronic toxicity in the various categories for var-
ious aquatic organisms. The scheme is limited to organic
chemicals, although residue-based info for some metals is
mentioned. Aquatic mixture toxicity information is
reviewed and it is demonstrated how this scheme can
explain existing mixture observations. Essentially, it
appears that when organic chemicals are combined in con-
centrations below their individual thresholds for observable
toxicity, they contribute additively toward the narcotic tox-
icity of the mixture proportional to their molar ratios in the
mixture. This holds regardless of the type or mechanism of
observable toxicity produced by the mixture components
when administered individually in higher concentrations.
Thus, dose-dependent diVerences in the modes of toxicity
may be a general theme that needs more attention as
researchers strive to improve the predictive accuracy of tox-
icity models for mixtures.
In his book “Multiple Chemical Interactions,” Calabrese
(1991) focuses on interactions among chemicals in mix-
tures. This comprehensive treatise includes a detailed
review of interaction classiWcation systems and statistical
analysis as well as various pharmacokinetic and pharmaco-
dynamics aspects of interactions. Carcinogenesis, terato-
genesis and other response endpoints are examined.
Mixture toxicity is reviewed both by and within broad
chemical classes: inorganics (including metals, elements,
gases), organics (including PAHs, pesticides, organochlo-
rines), and drugs. The text reviews various regulatory
approaches to mixture toxicity prior to 1991 and presents
two fundamentally important observations on mixture tox-
“Since most toxic substances have multiple toxic eVects,
the nature of any chemical interaction may vary depending
on the response that one measures.” (p. 95)
“While it may be a losing proposition, it is argued here
that only three basic classes of interaction can be recog-
nized as the most fundamental foundations for the present
study of joint interactions (Berenbaum, 1981): additivity,
synergy, antagonism.” (pp. 13–14)
“Quantitative Toxicology: Selected Topics” by Filov
et al. (1979) is a translation with revisions and update anno-
tations of the 1973 Russian textbook. The perspective is
diVerent from many texts written in English with consider-
able referencing of Russian scientiWc literature not well
known in the West. Despite an excellent review of various
dose–response issues, modifying factors, and structure–
activity relationships, no comprehensive classiWcation of
modes of toxic action is suggested. After reviewing mixture
interaction nomenclature schemes the authors concluded:
“For this purpose it seems advisable to use only one term,
‘additivity’ (‘additive’). Accordingly, a combined eVect
should be referred to as ‘additive’, ‘more than additive’, or
‘less than additive’, depending on whether it is equal to,
more than, or less than the sum of the individual eVects
produced by the substances concerned.” (p. 273).
Another important related point was made:
“As regards the nomenclature used to describe joint
actions, it appears that a distinction should be drawn
between terms describing the mechanisms of action and
those used to describe the toxic eVect itself.” (p. 272)
2.2. General methods, reviews and commentary
2.2.1. Mammalian-based publications
Subsequent to their exhaustive review of binary mixture
toxicity (
Krishnan and Brodeur, 1991), Krishnan and co-
workers published several additional papers on mixture
toxicity, some of which considered chlorinated chemicals.
Haddad et al. (2001) published a substantial improvement
of their earlier PBPK modeling approach to mixture toxic-
ity (Haddad and Krishnan, 1998; Krishnan et al., 1994) by
incorporating a measure of received dose; speciWcally, a tis-
sue dose metric based on the area under the curve of the
time-tissue concentration relationship. Haddad et al.
(2000a,b) also validated a rat PBPK model with a meta-
bolic degradation component, through prediction of
observed venous blood levels of various mixtures of sol-
vents (benzene, ethyl-benzene, toluene xylene, dichloro-
methane, and trichloroethylene). Krishnan and Brodeur
(1994) reviewed mixture toxicity interaction data for gases,
pesticides, metals/metalloids, and solvents. Chlorinated
organic chemicals were present in the pesticides and sol-
vents data and both interactions that were greater and less
than dose additivity have been reported in animal experi-
ments within these two groups. Although there are a num-
ber of examples of toxicokinetic mechanisms through
which such interactions can occur, there is only very limited
evidence, primarily from high dose occupational and acci-
dental exposures, for such mixture eVects in humans.
Finally, in a of review of drinking water contaminants and
Canadian regulatory risk assessment methods of such
Page 3
122 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
mixtures, Krishnan et al. (1997) concluded that the compo-
nent-based methodology for mixture risk assessment, with
appropriate use of dose addition, response addition, or
toxic equivalency factor approaches, is the current state of
the art. However, improvements to mixture toxicity meth-
ods, such as explicit consideration of interaction eVects and
improved mechanistic understanding that can be obtained
by techniques such as PBPK modeling, were recommended.
In a series of papers Cassee, Feron, Groten and various
co-workers in the Netherlands (Cassee et al., 1998; Groten,
2000; Feron et al., 1995a,b, 1998a,b; Henschler et al., 1996)
reviewed the state of the art in mixture toxicity theory
focusing on terminology, experimental methods, statistics,
interpretation, and risk assessment. Under terminology,
they recommended three basic combined toxic action cate-
gories: simple similar action (dose addition), simple dissimi-
lar action (eVect or response addition), and interactions
(non-additivity via synergism, potentiation, or antago-
nism). Mixtures were classiWed as either simple mixtures,
with typically 10 or fewer components with fully estab-
lished composition, and complex mixtures, typically with
tens to thousands of components where there was an
incomplete knowledge of the mixture composition. Experi-
mental methods could be divided into those employing a
bottom-up approach of component interaction analysis
versus those with a top-down approach where mixtures are
Wrst studied in their entirety whole mixtures, then in frac-
tions of the original mixture.
Based largely on mammalian liver and kidney toxicity
studies carried out in their laboratories, the Dutch investi-
gators identiWed four diVerent types of mixtures that
appeared to yield consistent results in multi-component
testing. Despite some minor interactive eVects, mixture tox-
icity appeared to approximate the toxicity of the most toxic
component of the mixture when: (1) components of the
mixture aVected diVerent target organs and/or operated via
diVerent modes of toxic action, (2) components of the mix-
ture aVected the same target organ but through diVerent
modes of toxic action. For such mixtures, exposure limits
based on the NOAEL of the most toxic component appear
to provide adequate protection against the toxicity of the
whole mixture. For mixtures where the components act on
the same target organ but diVerent target sites (3), there
were no signs of interaction and the NOAEL of the most
risky chemical in the mixture was a good approximation
for the whole mixture. For mixtures where components
aVected the same target organ via the same mode of action
(4), toxicity of the components was best approximated by
simple dose addition of each component. This method is
likely to overestimate mixture toxicity due to generalized
additivity assumptions based on similar target organs and
presumed mechanisms, however, it is unlikely to underesti-
mate mixture toxicity.
Toxicity information could be classed into several
approaches for use in risk assessment of chemical mixtures.
The toxicity equivalency factor (TEF) approach, as
employed for chemicals with TCDD-like toxicity, appears
to be a viable approach where groups of chemicals express
diVering potencies within a similar mode of toxic action.
Safe et al. (1998) have discussed criteria and limitations for
using the TEF approach. A fractionation and recombina-
tion approach to evaluating whole mixtures shows promise
and is being further developed. The combination of QSAR,
lumping analysis, and PBPK modeling is another promis-
ing strategy for interpretation and risk assessment (Verhaar
et al., 1997). Using the toxicity of the ten riskiest chemicals
in a complex mixture as a surrogate measure and the weight
of evidence (WOE) approach are additional approaches
that are currently being evaluated.
Several papers have reviewed the issue of damage–repair
mechanisms and kinetics using primarily data on pheno-
barbital and chlordecone potentiation of carbon tetrachlo-
ride liver toxicity (Calabrese and Mehendale, 1996;
Mehendale, 1995; Soni and Mehendale, 1998). When pres-
ent in mixtures at doses that produce low toxicity sepa-
rately, chlordecone (Kepone) and carbon tetrachloride
produce about 67 times the toxicity to rat liver as the car-
bon tetrachloride alone. Phenobarbital potentiation of car-
bon tetrachloride liver toxicity is even greater; however, the
animals exposed to the chlordecone–CCl
mixture tend to
die while those exposed to the phenobarbital–CCl
tend to recover and survive. The explanation oVered is a 2-
stage, dose-dependent damage–repair model. In phase 1,
which is thought to occur above an initial threshold level,
the parallel operations of damage and damage–repair
mechanisms allow recovery from damage. In phase 2, which
occurs beyond a second threshold, the repair mechanism is
inhibited and the damage accumulates unrepaired, often
causing all death. DiVerent chemicals have varying poten-
cies for phase 2 and the mortality versus recovery response
noted above has also been observed for some other combi-
nations of chemicals. Until it is better understood, the pres-
ence of dose-dependent damage–repair processes with such
signiWcant inXuence on mixture toxicity is likely to be a sig-
niWcant confounding factor in attempts to explain and pre-
dict the toxicity of mixtures.
Yang and co-workers have prepared a number of papers
on the theory and practice of mixture toxicity that have
provided an important focus for advancing the theory and
practice of mixture toxicity (el-Masri et al., 1995, 1997;
Yang, 1996; Yang et al., 1995a,b, 1998). Several important
points are made in this series of papers. Given the magni-
tude of the challenge of mixture toxicity, it is clear that sub-
stantial enhancements to experimental and risk assessment
methods are needed. In general terms, they suggest that
improvements can be achieved by using organism-based
uptake, distribution, and elimination modeling coupled
with data from well-deWned, model-based in vivo and
in vitro experiments analyzed with improved statistical and
mathematical protocols. Combining the PBPK with data
on tissue doses is suggested for modeling the fate of the
components of chemical mixtures in organisms and extrap-
olating experimental results for risk assessment purposes.
QSARs, exposure/eVect biomarkers, and Monte-Carlo
Page 4
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 123
simulations are suggested as means for improving toxicity
test protocols and improving the interpretation of test data
to facilitate enhanced conceptual toxicity models. An addi-
tional improvement suggested for mixture toxicity testing is
routine establishment of the interaction threshold in mix-
ture testing; that is, the point where biochemical interac-
tions between components begin to aVect the response of
the organism to the mixture. Factorial experimental designs
and improved isobolographic and response surfaces analy-
sis are both methods that show promise for improving sta-
tistical/mathematical aspects of mixture testing. Additional
important factors identiWed for improving the eVectiveness
and eYciency of mixture toxicity investigations include
minimizing animal usage, shortening experimental expo-
sures, and employing environmentally realistic exposure
Chambers and Carr (1995) reviewed published LD50
and LC50 levels for a variety of insecticides in several ver-
tebrate species and concluded that a wide range of toxic-
ity levels exist, and these cannot be easily predicted within
either a chemical group or within a species. Also, there is a
relatively limited database documenting interactions
between insecticides and other chemicals, either agricul-
tural or non-agricultural. They note a clear potential for
toxicity interactions due to the fact that most major insec-
ticide groups perturb nervous system function as their pri-
mary mechanism of acute toxicity; speciWcally, (1)
nervous system hyperexcitability via opening neuronal
sodium channels, e.g., chlorinated diphenylethanes such
as DDT or pyrethroids; (2) nervous system hyperexcit-
ability via antagonism at the inhibitory GABA receptor
chlorine ionophore, e.g. chlorinated cyclodienes; and (3)
nervous system hyperexcitability via anticholinesterases,
e.g. organophosphates and carbamates. Finally, metabo-
lism of insecticides appears to be far more inXuential
in some species than others in determining the toxicity
Seed et al. (1995) reviewed mixture terminology and
mixture risk assessment methods within the three tier US
EPA approach: actual toxicity data for the whole mixture
in question, data for a mixture with a similar composition,
and a component-based assessment using either a hazard
index or a toxicity equivalency factor approach. For carcin-
ogens, the various methods are typically variations on
response/eVect addition, although some mechanistic models
are available. They also reviewed publications that
addressed mixture toxicity at low doses and made three
general observations: adverse toxic eVects are unlikely to
occur for non-cancer endpoints when all components of a
mixture are below their respective eVect threshold, unpre-
dictable non-additive eVects may occur when some or all
mixture components are at or above their respective thresh-
olds for adverse eVects, and greater than additive carcino-
genic eVects have been observed in whole animal studies at
relatively low doses. The issue of mixture toxicity at low
doses remains controversial and is discussed further in
Section 4.
Sexton et al. (1995) reviewed the major issues and litera-
ture on mixture toxicity and two points vital for future
research and public health decision-making. The Wrst is that
a clear, consistent taxonomy for mixture-related phenom-
ena is needed. The second is that what critical features con-
stitute a common mechanism remains unclear.
Simmons (1994) noted the paucity of data on nephrotox-
icity for multiple exposures and mixtures. Existing data
indicates the presence of non-additive responses but a num-
ber of issues (6 are noted) confound interpretation of the
extent and signiWcance of this. The lack of sensitive, easily
measured indicators of nephrotoxicity maybe a contribut-
ing factor. Simmons (1995) indicates that the challenge
areas for mixture toxicology and risk assessment include
increasing the peer-reviewed publication of human studies,
improving access to peer-reviewed data and examining
multiple target organs. Two important aspects are the
development of a common, consistent terminology and the
use of appropriate experimental designs and analyses. She
also proposes that establishing mechanism(s) of toxic
action oVers a rational basis for extrapolation across dose
levels, exposure durations and exposure routes as well as to
other species and to other similar chemicals. Finally,
research should aim to reduce uncertainty in risk assess-
ments for chemical mixtures and focus on speciWc mixtures
where there is greater potential to reduce impact on human
Chambers and Dorough (1994) carried out a review of
pesticide mixtures and impurities of organophosphorus and
organochlorine pesticides. They concluded that literature
indicates synergistic, additive, and antagonistic eVects can
occur in mixtures but that predicting the nature of the mix-
ture interaction is diYcult and uncertain.
Woo et al. (1994) describes the development of an inno-
vative computerized system for ranking and predicting
potential cancer risks of chemical mixtures. The system is
said to consider both the additive risk of individual carcin-
ogens and the projected overall interaction eVect of the
mixture based on the possible interaction eVects of all
binary pairs of individual constituents of the mixture.
Using this system, the authors predict that some PAH mix-
tures should have a carcinogenic risk lower than that calcu-
lated by the simple additivity model, whereas the reverse is
true for a number of other mixtures.
2.2.2. Aquatic toxicity-based publications
Escher and Hermens (2002) present the clearest recent
review on various issues associated with the use of toxico-
logical information in environmental risk assessment. They
emphasized the theoretical and practical advantages of
organism-based rather than exposure-based dose metrics
for a number of areas, including mixture toxicity and
modes of toxic action. Despite the signiWcant conceptual
advantages, and the thorough investigation of baseline
(narcosis) toxicants, the lack of mixture toxicity test data
for various speciWcally acting compounds is currently limit-
ing. Although a number of mode/mechanisms of toxic
Page 5
124 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
action were discussed, no generally accepted classiWcation
scheme for modes of toxic action or mechanisms of toxicity
interactions were reported.
In their review of aquatic toxicity mixture interaction
terminology and statistics, Konemann and Pieters (1996)
noted that studies on mixtures often contain statements
about the type of combined action (e.g., additive, synergistic
or antagonistic). For a number of reasons, however, the
results and interpretations from such studies are often not
comparable. First, the terminology for indicating combined
action is far from consistent. Second, depending on the
model, diVerent conclusions may be drawn from the same
results. It is therefore important to provide clear deWnitions
of the null hypothesis. Third, adequate statistical methods
should be used for testing the null hypothesis. In the past,
many mixtures studies either used no statistics or used
statistics incorrectly. Konemann and Pieters (1996) recom-
mend that environmental toxicologists focus on the low-
dose region of the dose–eVect curve and conclude that
although interactions are less plausible at low doses, dose
additivity cannot be excluded.
Vighi and Calamari (1996) conducted a general review
of the issues and approaches for assessing the aquatic toxic-
ity of chemical mixtures. They concluded that water quality
guidelines for aquatic mixtures could be based on concen-
tration addition for toxicants with the same mode of action.
Based on an examination of published aquatic mixture tox-
icity data, Warne and Hawker (1995), proposed the funnel
hypothesis, which predicts that there is less deviation from
dose additivity as the number of components in a mixture
increases. It also predicts that the toxicity of mixtures mea-
sured using biological endpoints that require high toxicant
concentrations will deviate more from toxic additivity than
endpoints that require low concentrations. They use essen-
tially the same explanation as McCarty and Mackay
(1993), that when organic chemicals in complex mixtures
fall below the level where speciWc toxicity is expressed,
they then contribute to narcotic eVects in a dose additive
Shirazi and Linder (1991) developed a response surface
model for acute Wsh LC50 test data using a Weibull distri-
bution. The single chemical mixture data set of Broderius
and Kahl (1985) for 26 organics was used to develop the
model. Validation of the model with some binary and com-
plex mixtures test data was mostly additive with some posi-
tive and negative deviation from additivity.
2.3. Mode/mechanism of toxic action classiWcation
There are primarily three schools of thought on this
issue: one based largely on mammalian toxicity, a second
based on hazard classiWcation, and the third based largely
on aquatic toxicity.
2.3.1. Schemes based on mammalian toxicity Broad classiWcation schemes. Rosenkranz and Cunn-
ingham (2001) evaluated at high production volume (HPV)
chemicals via a QSAR approach and attempted to deWne
mode of toxic action categories for organic chemicals. Their
conclusion was: “We have demonstrated that traditional
organic chemical categories do not encompass groups of
chemicals that are predominately either toxic or nontoxic
across a number of toxicological endpoints or even for spe-
ciWc toxic activities. At this juncture, the use of arbitrary
categories will require further investigation to be useful in
predicting the toxicity of chemicals for speciWc toxicities or
to identify broadly toxic chemicals.”
Feron, Groten, Jonker and colleagues have conduced a
number of mammalian mixture toxicity studies where they
used a mammalian target organ classiWcation scheme. A
recent paper (Groten et al., 2000) used this scheme to exam-
ine possible mixture toxicity in food additives. Their
scheme was developed in earlier work, which uses some
organochlorines in the mixture experimentation. Feron
et al. (1995a) reported on two 4-week studies where rats
were exposed to 8 or 9 widely diVering chemicals (pesti-
cides, metals, inorganics, organics, organochlorines, drugs)
combined at the respective LOAEL and NOAEL. The
organochlorines employed were dichloromethane and
mirex. They deWned major toxicological targets at LOAEL
of adrenals, body weight, hemoglobin, red blood cell, stom-
ach, liver, kidneys, and thymus. The target organs for these
chemicals were identiWed as liver, stomach, blood, red
blood cell, heart, and nose. Two modes of action were
deWned and studied in more detail. The Wrst was nephrotox-
icants, which had four subclasses: metal ion chelator,
mitochondrial dysfunction, -lyase mediated activation
-globulin accumulation. These were studied using
four-week exposures to mixtures of chemicals at their
respective LOAELs and below. The second mode of action
examined was nasal irritants and 1 and 3 day inhalation
which employed a similar mixture regime. They concluded
that mixtures of chemicals with diVerent target organs or
the same target organ but with diVerent sites or diVerent
modes of action did not appear to be more hazardous than
individual chemicals in 4 week exposures, providing the
dose of each chemical in the mixture did not exceed its
respective NOAEL. Additivity did not hold but was judged
applicable when mixture components had the same target
organ and same mechanism of action or receptor.
Groten et al. (1997) carried out an extension of the 9
mixture study (dichloromethane plus 8 non-chlorinated
chemicals) using 4 week exposures at the minimum
observed adverse eVect level (MOAEL), NOAEL and 1/3
NOAEL plus a satellite experiment using a two level facto-
rial design with nine dose combinations. Although there
were a couple of modest but statistically detectable changes
in some exposure regimes, their conclusion was that “in
general, combined exposure to the single chemicals does
not constitute an evidently increased hazard, provided the
exposure level of each chemical is similar or lower than its
own NOAEL”.
Craig et al. (1997) developed a methodology for address-
ing occupational exposure to mixtures of non-radiological
Page 6
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 125
materials in facilities subject to the authority of the US
Department of Energy. For non-carcinogenic endpoints, a
standard hazard index summation process is employed. For
each substance the HI numerator is the TWA concentra-
tion at the receptor (1, 15, or 60 min averaging speciWed for
various circumstances) and the denominator is the appro-
priate Emergency Response Planning Guideline (ERPGs)
or Temporary Emergency Exposure Limit (TEEL). A HI
sum less than or equal to 1.0 was considered acceptable.
For carcinogens, the sum of the incremental cancer risks
was calculated and acceptability judged against applicable
facility guidelines. In addition, matrices were generated that
crossed the toxicological classiWcation for each chemical
with its target organ. Information for the matrices was
obtained from health code numbers found in standard
occupational references such as Patty’s (Cralley and Cral-
ley, 1985) and Sax’s (Lewis, 1996). This information was
used to decide which chemical-speciWc hazard indices
should be added to estimate mixture risk and which should
be treated independently. An example was evaluated using
14 organic chemicals of which 5 were chlorinated organics.
Jonker et al. (1996) compared a 4-week rat nephrotoxi-
cant study with 4 chemicals with diVerent modes of action
(hexachloro,1,3-butadiene, mercuric chloride, lysinoala-
nine, and d-limonene), to a 32-day rat nephrotoxicant study
with 4 chemicals with a similar mode of action (trichloro-
ethylene, tetrachloroethylene, hexachloro,1,3-butadiene,
and, 1,1,2-trichloro-3,3,3-triXuoro-1-propene). Rats were
exposed at various doses at or below the lowest-observed-
nephrotoxic-eVect level (LONEL) and the no-observed-
nephrotoxic-eVect level (NONEL) for each chemical, both
singly and in mixtures. For the diVerent modes of action
study, mixture exposure at the LONEL resulted in
increased growth depression and increased renal toxicity in
male but not in female rats. Mixtures at the NONEL pro-
duced slightly retarded growth and increased renal weight,
while mixture exposure at the NONEL/4 did not produce
any treatment-related changes. This suggests that neither
synergism nor additivity occurred in this experiment. In the
study with the similarly acting nephrotoxicants, relative
kidney weight was increased on exposure to the individual
compounds at their LONEL and, to about the same extent,
on combined exposure at the NONEL or the LONEL/3.
These results suggest that in those dose ranges, nephrotoxi-
cants with the same mode of action are additive in mixture.
McInnes and Brodie (1988) reviewed drug interactions
under the main categories of interactions aVecting either
pharmacokinetics or pharmacodynamics. Pharmacokinet-
ics may aVect absorption, distribution, metabolism, enzyme
induction/inhibition, and/or excretion and, due to substan-
tial inter-individual variability in humans, are diYcult to
predict accurately. On the other hand, pharmacodynamic
interactions are less easily classiWed, but are considered to
be generally more predictable. The most common mecha-
nism of drug interaction was identiWed as simple summa-
tion of similar or opposing independent eVects of multiple
drugs on the same system, organ, cell, or enzyme via direct
or indirect action on receptors. Although a number of spe-
ciWc examples of both classes of interactions were reviewed,
the mechanisms of action involved in pharmacodynamic
interactions were discussed only in very broad terms and no
attempt was made at a classiWcation scheme for either
modes or mechanisms. McInnes and Brody also point out
that many interactions have been identiWed on a biochemi-
cal or cellular basis that never manifest clinically signiWcant
changes. They call for a more reasoned and productive
approach to the study of drug interactions that focuses on
interactions that actually matter from a clinical standpoint.
Such insights from clinical pharmacology may be instruc-
tive as environmental toxicologists focus more intensely on
potential interactions in environmental chemical mixtures. Narrow or partial classiWcation schemes. Krieger (2001)
and Hayes and Laws (1990)
examined various aspects of pes-
ticide toxicity and discussed them using several diVerent
approaches: speciWc chemicals (e.g., DDT, pentachlorophe-
nol), chemical type (e.g., inorganics, organometals), chemi-
cal structure (e.g., pyrethoids, phenoxy herbicides), and
biochemical mechanism of action (e.g., protoporphyrino-
gen oxidase inhibitors, inhibitors/uncouplers of oxidative
phosphorylation). Although various modes/mechanisms of
action were reviewed and discussed, and a basic classiWca-
tion scheme of non-speciWc (e.g., corrosion and Ferguso-
nian narcosis) and speciWc (various types of biochemical
reactions) toxic actions was noted, no detailed comprehen-
sive framework of types of mode/mechanisms of toxic
action was discussed or proposed.
Butterworth and BogdanVy (1999) reviewed genotoxic
versus non-genotoxic modes of action for cancer risk
assessment. They provide good deWnitions for and discuss
the distinctions between the terms mode and mechanism of
action. Examples of diVerent modes of action based on the
scheme proposed by Butterworth et al. (1995) include:
genotoxic, non-genotoxic cytotoxic, and non-genotoxic
mitogenic carcinogens. The scheme is focused on carcinoge-
nicity and does not have broad application to other eVects
of organic chemicals.
Mileson et al. (1998) reported the results of an ILSI
working group commenced to evaluate EPA’s determina-
tion that organophosphate pesticides act via a common
mechanism of action on classiWcation of organophosphate
pesticide toxicity to mammals. They noted “Compounds
that act by a common mechanism of toxicity may cause the
same critical eVect, act on the same molecular target at the
same target tissue, act by the same biochemical mechanisms
of action, and share a common intermediate” (p. 18). After
considering several alternative hypotheses regarding vari-
ous mechanistic subgroups, the working group concluded
that OP pesticides should be considered to act via a com-
mon mechanism of toxicity if they inhibit AChE by phos-
phorylation and elicit any spectrum of cholinergic eVects.”
Pope (1999) reviewing the same issue cautioned that,
despite a common general mechanism, some organophos-
phate pesticides alter the common mechanism and/or
Page 7
126 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
include other mechanisms/pathways that produce qualita-
tive and/or quantitative diVerences in toxicity. It is sug-
gested that subclasses within the acetylcholinesterase
inhibition mechanism of action may be appropriate.
Soni and Mehendale (1998) noted in their review of tis-
sue damage and repair in mammalian liver toxicity that
four well-established mechanisms of toxicity were evident:
free radical formation, reactive intermediate formation,
lipid peroxidation, and blockade of endogenous pathways.
Feron et al. (1995a) in their studies of various complex mix-
tures identiWed four modes of action of nephrotoxicants:
metal ion chelator, mitochondrial dysfunction, -lyase
mediated activation and
-globulin activation. Schemes used in regulatory policy and enforcement.
The OECD (2001a,b) reports on mixture toxicity classiW-
cation are part of a series of reports whose objective is
international harmonization of various classiWcation
schemes, such as those for transportation, industrial chemi-
cals, consumer products, pesticides, and occupational expo-
sure. This report reviews various classiWcation schemes
from Japan, Canada, USA, Europe, and South America.
These comprehensive classiWcation schemes are based on
various eVect or response endpoints (irritation, cancer,
acute and chronic toxicity, reproductive eVects, aquatic tox-
icity, etc.), some critical chemical characteristics (Xamma-
bility, explosive, etc.) for both single substances and
mixtures. Although useful and practical for many purposes,
such types of hazard-based classiWcation schemes do not
represent a useful framework for a toxicologically based
classiWcation of modes/mechanisms of action for either sin-
gle chemicals or mixtures.
Paustenbach (1994) summarized the OSHA substances
toxicity classiWcation system, which employs 20 health code
numbers (from 1 to 20) to categorize chemicals by their pri-
mary adverse health eVects. This system is a mixture of
types of eVects (cancer, reproduction, acute, chronic, etc.),
target systems (nervous system, respiratory system, blood,
eye, skin, etc.) and other groupings (explosive, nuisance
particulates, odor). There is no explicit consideration of
mixtures in this scheme. Craig et al. (1997) used this OSHA
system to formulate an approach to dealing with mixtures.
They used the health code numbers as a basis for grouping
chemicals into similar modes of toxic action and adverse
toxic endpoints categories. They then prepared mode/end-
point-speciWc mixture hazard indexes (HI) for each health
code number by summing individual HI values calculated
as the respective peak exposure concentration divided by
the respective guideline value for each chemical found in a
given health code grouping. The authors judged that the
methodology produced a more realistic estimate of the risk
of chemical mixtures than either considering each compo-
nent chemical independently or summing the HI values for
all chemicals.
Sowinski and Cavender (1993) summarized various
national and international schemes and criteria for classify-
ing toxicity. They cover the same basic material as OECD
(2001a) with an emphasis on human occupational exposure
regulations. Although many of the schemes could be used
to classify deWned mixtures by essentially treating them as a
unique substance, no scheme was reported to be speciWcally
designed to address mixtures. Schemes based on aquatic toxicity. Researchers at
the US EPA lab in Duluth Minnesota have long been
involved in the attempt to prepare a classiWcation system
for modes of toxic action in aquatic organisms. Drummond
et al. (1986) examined four well-known modes of toxicity in
Wsh using fathead minnows: narcosis, uncouplers of oxida-
tive phosphorylation, acetylcholinesterase inhibitors and
skin irritants. They concluded that behavioral changes
alone are inadequate to identify modes of action and
stressed that gross morphological changes are an equally
important consideration. The US EPA-Duluth lab has
since reWned their mode of action scheme for acute toxic
action in the fathead minnow providing a detailed descrip-
tion of chemical groups and/or substructural fragments
associated with each mode of action category in the scheme
(Russom et al., 1997). Using this reWned scheme, which
includes QSARs, joint toxic action studies, toxicodynamic
les, and behavioral and dose–response interpretation
of 96-h LC50 tests, they successfully classiWed approxi-
mately 600 organics chemicals into the following six cate-
1. Narcotics (baseline or nonpolar, polar, ester narcosis)
2. Oxidative phosphorylation uncouplers
3. Respiratory inhibitors
4. Electrophiles/proelectrophiles
5. Acetylcholinesterase inhibitors
6. Central nervous system seizure agents
Despite this success in classiWcation, there is an ongoing
controversy about the existence of the polar narcosis cate-
gory used by the US EPA and in other schemes noted below.
It appears to be a composite of narcotics and uncouplers of
varying polarity rather than a separate mode of action (see
Escher and Schwarzenbach, 2002; Vaes et al., 1998).
Dutch researchers (Hermens, Verhaar, Deneer and their
colleagues) have also conducted extensive investigations
over the past decade on modes of action and mixtures.
Their detailed scheme for a mode of toxic action classiWca-
tion scheme consists of Wve basic classes:
1. Inert (narcosis)
2. Less inert (polar narcosis)
3. Reactive
4. SpeciWcally acting
5. Not possible to classify according to the scheme.
A decision tree using the chemical groups and/or substruc-
tural fragments associated with each mode category in the
scheme is provided by Verhaar et al. (1992). Van Loon et al.
(1997) support the approach of McCarty and Mackay (1993)
Page 8
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 127
using tissue-based dose metrics in conjunction with a deWned
mode of action scheme. In a recent paper (Vaal et al., 2000)
Hermens and co-workers commented on the importance of
this approach:
“We recommend the use of toxicologically based classiW-
cation schemes at an early stage of the risk assessment
procedure. Screening programs are most eYciently run
when only one species per compound is tested to prioritize
substances. The toxicity of compounds belonging to the
class of nonpolar narcotics is highly predictable and shows
little interspecies variation. For these compounds quanti-
tative structure–activity relationships (QSARs) can be
used to estimate eVect levels. Most eVort should be put
into testing reactive compounds and compounds with a
speciWc mode of action as toxicity to some species can be
105–106 times higher compared with less sensitive
Recently, an enhanced version of this mode of action
scheme was presented (Escher and Hermens, 2002). Ten
modes of ecotoxicological action were identiWed using a
classiWcation scheme that considered target sites, interac-
tions with toxicological targets, and molecular mechanisms
of action, speciWcally:
1. Baseline toxicity (narcosis)
2. Degradation of membrane lipids and membrane pro-
3. Inhibition of the electron transport chain
4. Uncoupling
5. Inhibition of ATP synthesis/depletion of ATP
6. Photosynthesis inhibition
7. Damage and depletion of biomolecules
8. Inhibition of competition (e.g., acetylcholine esterase,
estrogen receptor, etc.)
9. Indirect mutagenicity (DNA repair, combination,
10. Direct mutagenicity (frameshift, cross-links, strand
breaks, deletion, etc.)
McCarty and Mackay (1993, also in Rand et al., 1995))
expanded on the QSAR-based approach of their earlier
papers and combined an early version of the US EPA-
Duluth Wsh toxicity mode classiWcation plus dioxin with a
whole body residue eVect ranges for acute and chronic tox-
icity response endpoints in Wsh and aquatic invertebrates
exposed to organic chemicals. This categorization scheme
1. Narcosis
2. Polar narcosis
3. Respiratory uncoupler
4. Acetylcholinesterase inhibitor
5. Membrane irritant
6. Central nervous system seizure agent
7. Respiratory blocker
8. Dioxin (TCDD)-like
For organic chemicals, the least toxic mode of action
appears to cause acute toxicity in the 2–8 mmol/kg range
while for TCDD, the most toxic mode of action, acute tox-
icity is in the 0.000003–0.00004 mmol/kg range. Chronic
endpoints resulted from body residues about 0.1 times the
respective acute residue range. For other modes of toxic
action, acute and chronic eVects were observed in various
bands within this million fold range. Although there are a
number of factors e.g., lipid content and distribution,
metabolism, ionization, nature of toxic mechanism, etc.,
that confound some applications of residue-based interpre-
tation, the approach does provide a much-needed improve-
ment in toxicity test interpretation. The residue-based
approach also provides a valuable mechanism for explain-
ing the largely additive mixture toxicity observed in com-
plex mixtures of organic chemicals: when organic chemicals
with speciWcally toxic modes of action are present in a mix-
ture below about 1/3 to 1/10 of their acutely toxic level, they
drop below the range where their speciWc activity is
expressed and they contribute to narcotic toxicity.
Lipnick (1991) proposed a mechanism of toxic action
scheme based on QSAR analysis of rat oral LD50 and Wsh
LC50 data. Five basic classes, (modes of action) with some
subclasses are described. The scheme is consistent with
organic chemistry principles and does not address metals or
1. Baseline narcosis
2. Electrophile nonelectrolytes
a. Nucleophilic substitution: allylic and proparglyic
b. Nucleophilic substitution: benzylic activated
c. Nucleophilic substitution: -halo-(CBX, CCX)
d. Acid anhydrides
e. Strained three-membered heterocyclic ring
f. Michael-type addition
g. SchiV base formation
3. Proelectrophile non-electrolytes
a. Alcohol dehydrogenase activation
b. Monooxygenase activation
c. Glutathione transferase activation
4. Cyanogenic non-electrolytes
a. Cyanide release via cyanohydrin-like toxicant
b. Cyanide release via monoxygenase activation
5. Multiple step/mechanism
2.4. Statistics and models
Over decades, considerable resources have been devoted
to devising methods for predicting the dose–response char-
acteristics of mixtures based on information about their
individual chemical constituents. Intuitively, this might
seem to be a simple computational problem that requires
knowing only the concentration of each chemical constitu-
ent of the mixture and the endpoint response to each con-
centration. The mixture response would then be the sum of
the responses to the individual chemicals in the mixture.
Page 9
128 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
Such a function is known as independence (Bliss, 1939) or
response additivity (Finney, 1971) and is applicable only if
the organism responds to each chemical as if the other
chemicals in the mixture are not present. As such, indepen-
dence (response additivity) is an appropriate null hypothe-
sis for interaction studies, but two complications diminish
the applicability of such a simple function for predicting
mixture eVects.
The Wrst complication is that responses may not increase
linearly with dose. Mathematically, E
may not be equiv-
alent to E
+ E
. This can occur regardless of whether A
and B represent doses of the same chemical or diVerent
chemicals. Thus, some situations exist whereby response
addition leads to the paradoxical conclusion that two doses
of a single chemical are synergistic. An alternative function
for predicting combined action is based on a consistent
proportionality of response between chemicals, i.e., that the
dose–response curves are parallel (Loewe and Muischnek,
1926). Mathematically a/A + b/B D 1, where A and B are
doses that produce an equivalent response when adminis-
tered alone and a and b are doses that produce the equiva-
lent response when given in combination. This function,
called dose addition, is based on the premise that doses of
the same agent cannot interact and it may also be used as
the null hypothesis for interaction experiments. Only when
the dose response curves of individual chemicals are linear
and parallel would dose addition and response addition
functions be expected to yield equivalent predictions; this
expectation does not hold for other forms of the dose
response curves (Berenbaum, 1989; Borgert et al., 2001;
Greco et al., 1995).
The second complication that diminishes the applicabil-
ity of these functions for predicting mixture eVects is simply
that responses to mixtures may not conform to either func-
tion. When this occurs, an interaction is inferred. Because
of their potential utility in medicine and insect control,
identifying biologically signiWcant chemical interactions is
an active area of research. Despite considerable investiga-
tion, few interactions of real pharmacological or toxicologi-
cal consequence have been identiWed. Thus, one might
assume a model of non-interaction as the general case, and
then predicting mixture eVects becomes an exercise in iden-
tifying combinations of chemicals that interact.
Dose addition seems to have gained favor as a non-inter-
action model for interaction studies, probably because it
avoids the aforementioned paradoxical inference that can
result from a generalized application of response addition.
Binary mixtures (two chemicals) can be tested against the
null hypothesis of dose addition using a simple graphical
technique called an isobologram (Loewe and Muischnek,
1926). Numerous publications have described useful modi-
Wcations of the isobolographic technique and study designs
for testing interactions between binary combinations of
chemicals. These modiWcations include various statistical
methods, study designs and computational methods (exam-
ples include Carter and Carchman, 1988; Carter and Gen-
nings, 1994; Chen and Pounds, 1998; Christensen and
Chen, 1985; Drescher and Boedeker, 1995; Gennings et al.,
2005; Gennings and Carter, 1995; Gennings, 1995; Gessner,
1995; Haas, 1992; Price et al., 2002; Suhnel, 1996). Other
publications review and compare various models, review
applications of the isobolographic technique, or review sta-
tistical approaches for isobolographic analysis (examples
include Altenburger et al., 1990; Berenbaum, 1981, 1989;
Eide and Johnsen, 1998; Gerig et al., 1989; Groten et al.,
1996; Haas et al., 1996, 1997; Kodell et al., 1995; Kortenk-
amp and Altenburger, 1999, 1998; Krewski et al., 1989;
Simmons, 1996). When a third dimension is added to repre-
sent data from diVerent levels of response, an isobologram
becomes a response surface (for example, see Carter and
Gennings, 1994; Greco et al., 1995).
2.5. Government and related regulatory guidance
The following sections review various government and
regulatory guidance, with particular attention to apparent
similarities and diVerences in their underlying assumptions
and conceptual foundation.
2.5.1. US Department of Human and Health Services
The Agency for Toxic Substances and Disease Registry
(ATSDR) has recently undertaken the development of a
series of “Interaction Pro
Wles” to complement its series of
Toxicological ProWles mandated under the 1986 Superfund
Amendments and Reauthorization (SARA) of CERCLA
(Comprehensive Environmental Response, Compensation
and Liability Act). SARA stipulates that toxicological pro-
Wles be developed for hazardous substances most com-
monly found at facilities on the National Priorities List
(NPL). ATSDR derives its authorization to undertake the
development of interaction proWles from language in CER-
CLA that directs ATSDR to develop methods for deter-
mining the health eVects of substances in combination with
other substances with which they are commonly found.
Toward this end, ATSDR’s Division of Toxicology has
developed and coordinated a program to identify the mix-
tures most often found in environmental media and to
assess their toxicological eVects through trend analysis,
in vivo and in vitro toxicity testing, quantitative modeling
of joint action, and methodological development. Hansen
et al. (1998) summarized the general features of the
ATSDR approach and the purpose for the research pro-
gram on mixtures that has been funded by the Agency.
Two guidance documents relating to the toxicological
assessment of chemical mixtures have been published by
ATSDR, including a general guidance document for the
assessment of joint toxic action for chemicals in mixtures
(US DHHS, 2004a,b,c,d) and a guidance document for pre-
paring an interaction proWle (US DHHS, 2001). The guid-
ance document for assessing joint toxic action expresses a
clear preference for using data on the mixture of concern
for public health assessments and follows a tiered approach
to prioritizing available data quite similar to the scheme
originally recommended by the US EPA (1986). When data
Page 10
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 129
for the mixture itself are unavailable, the Agency prefers
data on a suitable “similar” mixture, deWned as:
a mixture having the same constituent chemicals but in
diVerent proportions to those found in the mixture of
concern, or
a mixture having most of the same constituents chemi-
cals that are present in highly similar proportions as
found in the mixture of concern.
The rationale for these deWnitions appears to be derived
from general concepts about structure activity relationships
rather than empirical evidence. No studies are cited by
ATSDR, nor were any found by these authors, that vali-
date the ATSDR criteria (or any other criteria) for predict-
ing the toxicological similarity of chemical mixtures.
In the absence of data on the mixture itself or a suY-
ciently similar mixture, ATSDR calls for using information
on the mixture components. The ATSDR approach begins
with the traditional hazard index and cancer risk estimates
using EPA’s reference doses (RfDs) and Cancer slope fac-
tors. ATSDR’s approach departs from standard practice at
that point, with intent to account for potential additive tox-
icity or toxicity interactions. Only when fewer than two
chemicals in a mixture exceed a hazard quotient of 0.1 for
non-cancer eVects or a 10
risk level for cancer does
ATSDR consider the mixture unlikely to constitute a sig-
niWcant health concern. In essence, ATSDR’s procedure for
mixtures adds additional safety factors of 10 for non-cancer
and 100 for cancer above the levels typically considered to
be of insigniWcant toxicological concern (e.g., Hazard Index
of unity and cancer risk of 10
) (US EPA, 1989).
For mixtures that require an evaluation of additivity and
interactions (i.e., those with two or more constituents with
hazard quotient greater than 0.1 or cancer risk exceeding
), ATSDR recommends application of PBPK models
to evaluate potential additivity and interactions. When
such models are unavailable, modiWcations of the hazard
index approach are recommended, speciWcally the Binary
Weight-of-Evidence (WOE) approach of Mumtaz and
Durkin (1992) and Mumtaz et al. (1994) for chemicals with
the same critical eVect, or the Target Toxicity Dose method
of Mumtaz et al. (1997) for chemicals with several diVerent
critical eVects. Cancer is assessed as if it were the same criti-
cal eVect regardless of the tumor type or location. Both
approaches rely heavily on the ability to discern the mecha-
nism of toxicity suYciently well to assume which chemicals
will be additive and which will not. To date, the level of
mechanistic detail required to make such a determination
has not been deWned or validated by ATSDR or others.
Despite this, the binary WOE forms the basis for the
ATSDR approach and is used in several of ATSDR’s Inter-
action ProWles, including the proWle for breast milk con-
taminants. Although not deWning the criteria necessary to
make a toxicity similarity determination ATSDR’s docu-
ments do provide examples of how WOE methods might be
applied in diVerent situations.
Mumtaz et al. (1998) evaluated the accuracy of their
WOE methodology by testing it against empirical data
from studies speciWcally designed to evaluate interactive
eVects of similarly and dissimilarly acting nephrotoxi-
cants and hepatotoxicants. The WOE method was devel-
oped for use as a quantitative modiWer to the hazard
index in risk assessment involving multiple chemicals.
They concluded that for groups of similarly acting chemi-
cals, the WOE method modiWed the hazard index in the
direction of the observed data. In other words, for such
chemicals, the WOE method was more accurate than
strict dose additivity. On the other hand, the WOE
method was less accurate than dose additivity when the
mechanism for the eVect was unknown and it did not cor-
rectly predict interactive eVects for dissimilarly acting
chemicals. For dissimilarly acting chemicals, response
additive models appeared to be more predictive than the
WOE method.
Durkin et al. (1995) presented a systematic approach
for assessing data on chemical interactions for three types
of binary interaction patterns: chemical–chemical interac-
tions, chemical–class interactions, and class interactions.
Chemical–chemical interactions were deWned simply as
interactions that occur between two speciWc chemicals.
Chemical–class interactions were deWned as those that
occur between a chemical and all members of another
class of chemicals. Class interactions were deWned as those
that occur among all members of a chemical class. The
authors proposed calculating summary vector for the
chemical–chemical and chemical–class interactions. This
calculation involves tabulating published reports of inter-
actions according to the direction of the interaction, i.e.,
synergistic (+1), additive (0) and antagonistic (¡1). An
index of variance is then calculated that is mathematically
analogous to calculation of a population variance. The
authors demonstrated the utility of the method by
describing its application to a set of chemicals for which a
robust data set exists (carbon tetrachloride and alcohol).
They also listed the interaction patterns they calculated
for 21 chemicals pairs and discussed the limitations of
applying the method to chemicals for which interaction
data are sparse.
Mumtaz and Hertzberg (1993) reviewed the status of
interaction data with respect to its use in mixture risk
assessment, noting problems with the application of proper
statistical methods in many interaction studies as pointed
out previously by US EPA (1990). Mumtaz et al. (1997)
reviewed various modiWcations of the hazard index
approach that can be applied to the risk assessment of
chemical mixtures, such as the target toxicity dose and
binary weight of evidence approaches developed previously
by these researchers.
In addition to its guidance for assessing joint toxic
action, ATSDR also released a guidance document
designed to assist contractors with preparing interaction
proWles for the Agency (US DHHS, 2001). Interaction pro-
Wles are intended to assist public health assessors who need
Page 11
130 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
to know whether interactions are expected to occur among
components of chemical mixtures and how to incorporate
concerns regarding interactions into the public health
assessment of a site. The proWles provide interpretations of
the interaction data for a given set of chemicals. These
interpretations are said to involve judgments and assump-
tions that are matters of policy rather than objective sci-
ence. The proWles are intended to reXect ATSDR’s
evaluations of the validity of particular studies and the
inferences that can be drawn from them, but do not provide
all the information necessary to support these evaluations
or to allow users to weigh the evidence for themselves (US
DHHS, 2001). ATSDR has developed several interaction
proWles, including proWles for: Persistent chemicals found
in breast milk and Wsh (chlorinated dibenze-p-dioxins,
hexachlorobenzene, p,p-DDE, methylmercury and poly-
chlorinated biphenyls) (US DHHS, 2004b,c) and some
chlorinated ethanes and ethylenes (US DHHS, 2004d), as
well as some other nonchlorinated substances.
2.5.2. US Environmental Protection Agency
Chen et al. (2001) proposed a formal statistical proce-
dure to estimate the cumulative risk posed by mixtures of
chemicals that share a common mode of action. The proce-
dure Wts dose response data for the individual chemicals to
a dose–response function for the mixture under the
assumption of dose addition. The assumption of dose addi-
tivity is said to be consistent with EPA policy for pesticides
that cause related pharmacological eVects. The authors
illustrate the procedure using data for four analgesics pre-
sented previously by Finney (1971). Fenner-Crisp (2000)
reviewed EPA methods for cumulative risk estimation and
discussed EPA policy approaches to pesticide mixtures.
Putzrath (2000) reviewed EPA methodologies for mixture
risk estimation and identiWed inconsistencies that arise
from application of the various methods. She proposed
methods aimed at reducing the uncertainty in these risk
estimates regardless of the underlying accuracy of the
EPA’s most recent guidance document for mixtures (US
EPA, 2000) focuses on procedures for dose–response
assessment and risk characterization for human health risk
assessment from multi-chemical exposures. Many of the
procedures described therein might also be adaptable to
ecological risk assessments. The document is said to con-
tain some procedures that have had no previous applica-
tion in actual health risk assessments. This supplementary
guidance does not recommend or prioritize any single
approach for mixture risk assessment (i.e., use of data on
the mixture of concern, similar mixture or components) and
as such, represents a departure from the Agency’s 1986
guideline. The guidance intends to provide risk assessors a
number of options for choosing approaches that Wt the type
of mixture, the available data and site characteristics. As
speciWed in the 1986 guideline, the 2000 guidance also
begins the assessment process with an evaluation of data
EPA’s 2000 Supplementary Guidance document is con-
siderably more detailed than its 1986 guideline (US EPA,
1986), and it expands and updates many of the subjects dis-
cussed previously in the 1988 Technical Support document
(US EPA, 1988). The document provides an explanation of
many diVerent approaches for assessing combined eVects of
mixtures and gives several examples of their use. Future
directions for mixture assessment are considered and the
overall direction of mixtures assessment is said to proceed
along the same lines as risk assessment for single chemicals.
For both, there is an increasing emphasis on information
about speciWc modes of action and expanded use of mathe-
matical and statistical models. EPA diVerentiates mixture
assessments from those for single chemicals by its more
extensive use of quantitative inference from tested chemi-
cals to untested chemicals.
EPA suggests two approaches that might be taken to
evaluate the inWnite variety of chemical mixtures possible in
the environment. The Wrst approach would be to directly
investigate a few high-priority mixtures and use data on
these mixtures to extrapolate to other environmental mix-
tures based on degrees of similarity. The challenges posed
by this approach are in determining the priority of mixtures
to test. Considerations for priority setting are said to
include the documentation of public health problems in
populations exposed to speciWc mixtures and the number of
pairs of chemicals in the mixtures that have been reported
to produce toxicological interactions. The second approach
for addressing the remaining mixtures would be to focus on
developing methods to extrapolate exposure and toxicity
estimates from tested to untested mixtures. Exposure issues
and methods are especially highlighted. Issues of changing
mixture composition and toxicologic interactions between
components challenge the development of such methods
and are said to require integrating information from several
disciplines. EPA speciWcally mentions dose–response
extrapolation, basic toxicology, epidemiology and occupa-
tional studies and mathematical modeling.
Overall, the document relies heavily on concepts related
to modes of toxic action and toxicological similarity
between diVerent mixtures. Data on interactions appears to
play an increasingly important role in mixtures assess-
ments, yet criticisms of the interactions literature, such as
those contained in the 1988 Technical Support document,
are not highlighted. Little speciWc guidance is provided for
how to determine when two mixtures should be considered
toxicologically similar. Considerable professional judgment
and numerous assumptions regarding methodological
details will be required to implement this guidance, and
therefore it seems likely that the traditional approaches of
calculating hazard indices and summing cancer risk esti-
mates is likely to remain the predominant form of mixture
risk assessment.
Wilkinson et al. (2000) reviewed and analyzed methods
of adding toxicity values for use in cumulative risk assess-
ment as speciWed under the 1996 Food Quality Protection
Act. The paper identiWes several conceptual problems in the
Page 12
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 131
various approaches and concludes that approaches based
on margins of exposure or points of departure appear most
defensible, whereas approaches based on reference doses
and other estimates of acceptable exposure are less useful
due to the diVerent uncertainty factors used to derive the
Several authors provided reviews of testing and interpre-
tation and/or recommended updates for EPA’s 1986 mix-
ture guidelines, including analysis of the hazard index
approach for mixtures risk assessment (Houk and Waters,
1996; Svendsgaard and Hertzberg, 1994; Teuschler et al.,
1999; Teuschler and Hertzberg, 1995).
EPA’s 1988 technical support document (US EPA, 1990)
was developed in response to a recommendation of the Sci-
ence Advisory Board pursuant to its review of EPA’s 1986
Guidelines (US EPA, 1986). It discusses the available (to
1988) toxicity and interaction information useful in assess-
ing human health risks from mixtures. Mathematical mod-
els and statistical techniques are reviewed and research
needs identiWed in the document. Noteworthy in this docu-
ment is the report of an evaluation of the statistical meth-
ods used to test interactions in more than 400 studies.
Among a subset of 32 studies evaluated in detail for the
adequacy of the statistical tests for the experimental design
and the interpretations drawn from the data, only one was
deemed acceptable. The document also reviewed the evi-
dence for diVerent mechanisms of interaction and made
suggestions for future research. EPA considered the most
signiWcant conclusions of the document to be: (1) that the
available literature is extremely inadequate for use in
quantifying the extent of synergism expected from environ-
mental exposures, and (2) that validation of in vitro and
short-term in vivo studies seemed to oVer the most promise
for improving risk assessment of complex mixtures. The
document also makes other salient points regarding inter-
action studies in general. The Agency states: “The diYcul-
ties in obtaining quantitative measures on toxicant
interactions are exacerbated by the fact that many of the
studies on binary mixtures that purport to quantify toxi-
cant interactions are improperly designed and the reported
results are either uninterpretable or are diYcult to compare
among diVerent studies.”
This document also recognizes the importance of consid-
ering the various ways in which chemicals are known to
interact, a recognition that appears to have been set aside in
subsequent guidance from the Agency. Section 3 of this
document constitutes a 16-page discussion of various
mechanisms of interaction, and a summary table provides
examples arranged according to a modiWcation of a scheme
proposed by Veldstra (1956). Regarding modes of interac-
tion, EPA quotes an extensive review by WHO (1981),
“ƒthe available evidence from in vitro animal experiments
and from human observations has shown that a limited
number of mechanisms seem to account for the majority of
the important biological interactions.” EPA’s interpreta-
tion is that “ƒthe basic mechanisms by which toxicants
interact as detailed by Veldstra (1956) are based on classic
pharmacologic principles that have not changed substan-
tially over the past 30 years.”
EPA’s initial concise guideline for risk assessment of
chemical mixtures (US EPA, 1986) provides a stepwise
decision tree for assessing the toxicity of mixtures accord-
ing to the type of data available for the particular mixture
of interest. The decision tree begins with an assessment of
data quality, which involves utilizing a speciWc classiWcation
scheme for categorizing the quality of the available data.
From highest to lowest quality, the document lists data on
the mixture itself, data on a similar mixture, data on the
individual components of the mixture and their interac-
tions, and combined toxicity based on additivity assump-
tions as the quality categories into which mixture risk
assessments may be grouped.
In their 1986 Guideline, the Agency cites studies by Poz-
zani et al. (1959), Smyth et al. (1969, 1970) and Murphy
(1980) as justiWcation for assuming additivity between mix-
ture components. These same studies are reviewed in more
detail and again cited as the supporting data for additivity
assumptions in EPA’s 2000 Supplemental Guidance docu-
ment (US EPA, 2000). In their 1986 Guideline, EPA
acknowledges that it is highly questionable whether inter-
action data from acute toxicity studies can be used to infer
interactions from chronic low-level exposure, unless the
mechanism by which the interaction occurs is known to
operate at both high and low doses. Saturation phenomena
are said to explain the reason that interactions occurring at
high doses are unlikely at have a signiWcant impact on tox-
icity at low doses.
2.5.3. Other American contributions
The National Research Council was commissioned to
develop strategies and experimental approaches for assess-
ing the toxicity of complex chemical mixtures on the basis
of a selective review of the available literature. Its report
(NRC, 1988) reviews the general concepts pertinent to the
study of mixtures, including mixture characterization,
exposure, dose–response characterization and predictive
dose–toxicity response models for multiple chemicals. The
report oVers no speciWc recommendations for either models
or experimental designs, instead stressing the general need
for continued development of more robust models veriWed
by experimental data. Although informative in a general
way and useful in some contexts, many of the approaches
described for testing mixtures and evaluating mixture
eVects have been superseded by newer modiWcations, but
none have gained general acceptance in the Weld.
It is important to recognize that the NRC report pre-
sents one side of a polarized debate that is at variance with
the thinking of prominent authorities in the mixture Weld.
For example, NRC relies heavily on mechanistic inference
to select appropriate null hypotheses for testing chemical
interactions, stating that the validity of the non-interaction
model depends upon the extent to which it is based on an
understanding of biologic mechanisms. In contrast, leading
researchers in the Weld have pointed out that presumptions
Page 13
132 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
about mechanisms are a poor basis for selecting the null
hypothesis for testing interactions (Berenbaum, 1989;
Greco, 2001). Berenbaum (1989) has explained that reliance
on mechanistic understanding to choose a non-interaction
model creates a conundrum whereby the inference of inter-
action cannot be made from an interaction study. Rather
than supporting an inference of interaction, departure from
the non-interaction model would call into question the
researcher’s mechanistic understanding. Greco’s commen-
tary on Borgert et al. (2001) provides several arguments
against formulating the null hypothesis based on mechanis-
tic assumptions, including the fact that most agents have
more than one action and our current understanding about
most mechanisms is probably wrong. Consequently, the
practice adds a subjective step that contributes confusion
rather than clariWcation to the analysis of mixture eVects.
Under its cooperative agreement with EPA’s OYce of
Water, the ILSI Risk Science Institute (1998a) conducted a
comprehensive review of the literature on the toxicity of
complex mixtures in drinking water and the methods used
to assess human health risks associated with exposure to
those mixtures. Numerous toxicological studies were identi-
Wed and reviewed. Based on the review, the methods used to
assess disinfection mixtures were grouped into three general
categories: assessment of the toxicity of concentrates of
drinking water; assessment of the toxicity of humic acid
solutions treated with various disinfectants; and, assess-
ment of the toxicity of deWned mixture of chemicals.
Numerous diYculties in assessing disinfection byprod-
uct mixtures were identiWed and discussed. Most of these
diYculties are instructive for all types of chemicals,
although they have been most extensively analyzed for dis-
infection byproducts. The diYculties include: (1) that
attempts to concentrate drinking water inevitably change
the original composition of the water both qualitatively
and quantitatively; (2) the composition of water varies
from source to source and season to season; and (3) use of
surrogate mixtures raises the question how similar deWned
mixtures are to drinking water.
ILSI concluded that despite the diYculties, it is possible
to make some general statements about the signiWcance of
the database. Although the database provides clear evi-
dence of the in vitro mutagenic potential of disinfection
byproduct mixtures, none of the in vivo studies show evi-
dence of carcinogenicity and little evidence of systemic tox-
icity, even at levels far exceeding estimates of human
exposure. Although it is theoretically possible to delineate
mechanisms that would lead to additive, synergistic or
antagonistic toxicity of constituents of disinfection byprod-
uct mixtures, it is doubtful whether such mechanisms mani-
fest adverse health outcomes, or are even operative in the
concentration ranges pertinent to human consumption.
ILSI identiWed two sets of studies as particularly impor-
tant: those of the TNO Nutrition and Food Research insti-
tute in the Netherlands (Groten et al., 1994; Jonker et al.,
1990, 1993, 1994) and the studies of deWned mixtures con-
ducted by the National Toxicology Program (Heindel et al.,
1994, 1995; US NTP, 1993). The TNO studies indicated no
discernible toxic eVects until the concentrations of mixture
components approached or exceeded their NOAELs. How-
ever, regardless of the mechanisms of action of the individ-
ual constituents, interactions occurred when levels were
equivalent to the individual LOAELs and thus, above their
individual toxic thresholds. Although not designed speciW-
cally to investigate the relationship between levels of indi-
vidual chemicals and their thresholds, results of the NTP
studies were interpreted to support the conclusion that
adverse eVects are unlikely when the mixture components
are present at levels well below their individual thresholds
(Seed et al., 1995). ILSI concluded that inadequacies in the
similar mixture approach for assessing risks of disinfection
byproducts as well as inadequacies of the component-based
approach to identify signi
Wcant interactions at low doses
makes an approach that focuses on the health eVects of a
few critical compounds most promising.
Under its cooperative agreement with EPA’s OYce of
Water, the ILSI Risk Science Institute (1998b) convened a
panel of experts to develop and prioritize research recom-
mendations to investigate the potential toxicity associated
with disinfection byproducts in drinking water. Overall, the
panel concluded that the evidence is insuYcient to believe
that disinfection byproducts pose a potential health risk,
but could not quantify the level of risk without additional
information. Germane to the subject of mixtures is the
panel’s determination that the risks posed by disinfection
byproducts cannot be quantiWed using traditional
approaches based on single compound toxicity testing. The
panel recommended that integration of the latest advances
in computational technology, such as PBPK/PD modeling,
biologically based dose–response (BBDR) modeling, and
QSAR modeling, as well as molecular biology should be
considered for attacking toxicological questions regarding
disinfection byproduct mixtures.
The panel stressed the important concern for interac-
tions between disinfection byproducts that might increase
toxicity, and stressed that the critical issue is not whether
interactions have been identiWed per se, but whether they
are likely to occur at the levels found in drinking water.
Examples were given for interactions that occur by mecha-
nisms that either would or would not be expected to oper-
ate at low doses. Due to the sheer number of combinations
that would have to be tested for interactions, the panel rec-
ommended a tiered testing approach using whole mixture
studies, with the Wrst tier consisting of QSAR analysis and
in vitro tests for mutagenicity, cellular toxicity, and hor-
mone receptor binding. Mixtures testing negative would
not be further evaluated, but those testing positive would
move to the next tier, consisting of short-term in vivo
assays. Tier 3 would consist of mutigenerational studies
and chronic toxicity/oncogenicity studies. Mechanistic
studies to predict interactions were also discussed, with rec-
ommendations made for studies to address the dose-depen-
dence of those mechanisms. The panel predicted that only
interactions that occur in the range where pharmacokinetic
Page 14
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 133
and metabolic variables are linear with dose would be
important at the levels at which disinfection byproducts are
found in drinking water. Molecular and epidemiological
approaches were also discussed.
SpeciWc recommendations were made to use epidemio-
logical approaches to identify sensitive subpopulations and
to elucidate potential associations between disinfection
byproducts and speciWc diseases. Molecular approaches
were recommended to identify markers of exposures to dis-
infection byproducts, biological markers of toxic eVects,
and to identify sensitive individuals.
Craig et al. (1997) developed a methodology for address-
ing carcinogenic and noncarcinogenic occupational expo-
sure to mixtures of nonradiological materials in facilities
subject to the authority of the US Department of Energy.
For noncarcinogenic endpoints, a standard hazard index
summation process is used, while for carcinogens the sum
of the incremental cancer risks is employed. Furthermore,
matrices of toxicological classiWcations developed from
health code numbers (HCN) obtained from standard occu-
pational references were used to determine when chemical-
speciWc risk should be added together and when they
should be treated individually. An example was examined
using 14 organic chemicals of which 5 were chlorinated
organics: carbon tetrachloride, chlorobenzene, methylene
chloride, tetrachloroethylene, and 1,1,1-trichloroethane.
2.5.4. European contributions
Binderup et al. (2003) produced an extensive and com-
prehensive review of the theory and practice of the toxic-
ity of mixtures of industrial and environmental
chemicals. They also examined the risk assessment and
regulatory approaches to mixture toxicity by various
agencies around the world. Although they examined
many of the issues about modes of toxic action, they did
not discuss mode of action classiWcation in any detail.
Some mode of action data was reviewed using some
experimental work with simple, well-deWned mixtures. As
well, they presented their detailed review of mixture tox-
icity by diVerent toxicological eVect areas: local irrita-
tion, genotoxicity, carcinogenicity, reproductive toxicity,
endocrine disrupting chemicals, neurotoxicity, and
immunotoxicity. In their summary, Binderup et al. (2003)
make several points including: the presence or absence of
interactions between components of a mixture is vital in
any risk assessment; most risk assessment guidelines for
mixtures do not adequately address mode of action issues
in their simplistic recommendations regarding dose or
response addition methodologies; the use of dose addi-
tion approaches to mixture toxicity must be carefully
applied because risks may be substantially overestimated
where the requisite assumptions are not valid; toxicoki-
netics are vital in understanding and interpreting expo-
sure-based toxicity results as the amounts of substances
received in the body of exposed organisms (i.e., internal
or target site dose) may varying widely in diVerent cir-
cumstances, yet is seldom known, and Wnally; no single
approach for risk assessment of mixtures is yet suitable
for universal application.
ReVstrup (2002) reviewed available risk assessment pro-
cedures, and the underlying toxicological understanding
and available experimental data, with the objective of eval-
uating the risk to humans due to consumption of food con-
taining mixtures of pesticides. The component-based rather
than the whole mixture approach was judged to be appro-
priate for use in this case. Various risk assessment protocols
were recommended for pesticide contamination based on a
case-speciWc examination of available toxicological data.
Toxicological similarity was determined by a combination
of structural groupings, biochemical pathways, adverse
eVects, and target organs.
The UK Committee on Toxicity of Chemicals in Food,
Consumer Products and the Environment (2002) carried
out an extensive and comprehensive review of the presence
of and exposure to mixtures of pesticides and veterinary
medicines in human food. Current regulations in the UK
and USA were examined, evidence for dietary exposure
evaluated, biomonitoring and biological eVect monitoring
assessed, basic concepts and models for mixture toxicity
summarized, and the state of probabilistic risk assessment
reviewed. The experimental toxicity data examined was
organized, not by mode of toxic action, but by various
types of target system eVects e.g., respiratory systems, skin,
neurotoxicity, nephrotoxicity, carcinogenicity, genotoxicity,
Four items from the conclusions and recommendations
are noteworthy for the purpose of this manuscript. First,
current regulations for pesticides and veterinary medicines
focus on individual chemicals/products and do not address
mixtures. Second, dose additivity has been observed at high
and low exposure/eVect levels for some substances with the
same target organ and mode of action, however, for sub-
stances with diVerent target organs and/or modes of action,
no additivity or potentiating interactions are generally
found (i.e., response addition) when mixture component
exposure levels are near or below the range of the respective
NOAELs. Third, the presence and nature of mixture toxic-
ity at clearly toxic eVect levels is not necessarily predictive
of lower exposure levels near or below the lowest observed
eVects levels. Fourth, additional work was recommended to
develop common mechanism groups (essentially a mode of
toxic action classiWcation scheme).
In an update of a 1985 report on the same issue, the
Health Council of the Netherlands (2002) examined the
human health risk assessment processes for simultaneous
or near simultaneous multiple exposures to substances via
one or more exposure pathway, along with associated toxi-
cological knowledge and data. Rather than a prescriptive
approach with detailed protocols, it represents overall guid-
ance on a structured approach to various issues concerning
toxicology and risk. Although the report recommends fur-
ther work in clariWcation and collection of additional data
on working mechanisms of toxicity, this refers to reWne-
ment of a basic Plackett and Hewlett (1967) classiWcation of
Page 15
134 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
four working mechanisms of mixture toxicity (dose or
response additivity with or without interaction) rather than
a speciWc mode of toxic action classiWcation scheme.
The European Centre for Ecotoxicology and Toxicology
of Chemicals produced a monograph reviewing the aquatic
toxicity of mixtures including the theoretical background,
acute and chronic toxicity testing in the laboratory and
Weld, and risk assessment protocols. (ECETOC, 2001). They
concluded that acutely toxic mixture eVects appear to be
generally additive in laboratory testing, particularly for
large numbers of chemicals at approximately equitoxic con-
centrations. Chronic eVects of mixtures were also judged to
be largely additive, but more variability was often observed
in mixtures of metals. Recommendations included more
study of existing risk assessment processes (summation of
environmental/no eVect concentrations, additional mixture-
related correction factors for risk assessments, and tailoring
mixture risk assessment reWnement to case-speciWc environ-
mental status), evaluation of additional risk assessment
processes, and consensus on terminology, statistical
approaches, and interpretation of experimental designs for
ILSI-Europe examined 350 food additives registered for
use in the European Union for possible joint action or
interaction (Groten et al., 2000). The mode of action classi-
Wcation was primarily by grouping according to target
organ toxicity using the ADI. A initial screening using crite-
ria including clear organ toxicity above the NOAEL and
production of toxic eVects in the same organ, indicated that
65 chemicals should be further examined for joint toxicity/
interaction and a common target organ. It was concluded
that, for approved food additives, there are only a few cases
where joint toxic action or interactions cannot be excluded
using current toxicological principles. For various reasons,
the few non-excluded cases did not appear to represent an
increased risk for joint action or interaction. Overall, it was
judged that joint toxic actions or interactions between
approved food additives do not represent a signiWcant
health concern.
3. Summary
3.1. Toxicological reference books
Overall, there is wide variability in the nature and extent
of coverage about mixtures in standard toxicological refer-
ence material. It can range from very brief to good coverage
of both the theory and practice of mixture toxicology with
good general deWnitions of key concepts plus information
on risk assessment and regulatory applications. Despite the
diVerences in detail, the nature and thrust of the presenta-
tions is generally consistent. Several general key points were
Little information is available about modes/mechanisms
of toxic action. Although some attempts at preparing
mode/mechanisms classiWcation schemes have been
attempted, none are widely accepted or generally appli-
Various toxicity interaction schemes have been proposed
but the simplest scheme—less than additive (synergism),
additive, more than additive (synergism), with a clear
mathematical deWnition of “additive”—is currently con-
sidered the best as more sophisticated schemes require
knowledge of the mechanisms of toxic eVects and toxic
interactions that is usually not readily available.
A number of modifying factors can aVect toxicokinetics
and/or toxicodynamics of components in mixtures but
speciWc quantitative information is largely unavailable.
Limitations associated with converting between the dose
metrics in the two basic categories of exposure doses and
received doses are a major restriction to more complete
and thorough interpretation of mixture toxicity.
3.2. General methods/review/commentary
The material covered in this section is largely integrative
in nature, but covers more speciWc issues in greater detail
than the reference material covered in Section 3.1
. These
publications reviewed various aspects of mixture toxicity
theory and practice and discussed improvements to existing
methodology, as well as examining new approaches. PBPK
modeling, incorporation of received dose metrics rather
than exclusively exposure-dose metrics, and consideration
of damage–repair theories to interpretation of experimental
results were considered promising approaches. Various reg-
ulatory risk methodologies, such as dose addition, response
addition, and toxic equivalency factors, were also reviewed.
Some key issues noted were:
a clear, generally accepted taxonomy for mixture-related
phenomena is needed;
a critical question for mixture toxicity remains unan-
swered: What constitutes a common mechanism of
3.3. Mode/mechanism classiWcation
It has been conWrmed that, at least for organic chemi-
cals, standard grouping based on similarity in chemical
structure is not suitable for toxicological classiWcation.
Although structure and substructure are important deter-
minants of toxicological characteristics of a chemical, they
are not necessarily the same structure/substructure aspects
used in classiWcation schemes devised by chemists. The use
of classiWcation by target system/organ eVects has demon-
strated some utility, but has not been developed to the
point where it is broadly applicable. ClassiWcation eVorts
with narrow scopes (selected types of pesticides, cancer,
liver/kidney toxicants) have demonstrated some utility in
addressing important aspects of mixture toxicity. Overall,
work on development of classiWcation schemes for both
modes of toxic action and mechanisms of interaction are
recognized as being important, but those working in the
Page 16
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 135
Weld of mammalian toxicology have proposed no widely
applicable schemes for consideration and reWnement.
Although various schemes for classifying chemicals for
regulatory purposes have been developed, they are often
hazard based and usually consider a variety of factors in
addition to toxicity information. When mixtures are con-
sidered, it is usually as a deWned mixture which is treated
essentially as a unique chemical rather than a mixture.
Perhaps the most development on broad mode/mecha-
nisms of toxic action classiWcation schemes has been carried
out by those working in the Weld of aquatic toxicity. QSAR
and general mammalian mode/mechanism information has
been combined to generate schemes for organic chemicals
that have been demonstrated to have signiWcant utility in
both classifying individual chemicals on the basis of similar
mode/mechanism of toxic action and explaining toxicity
test data from aquatic organisms exposed to speciWc mix-
tures in various circumstances. Although incomplete, espe-
cially because inorganics and metals are not included in the
existing schemes, the extant work illustrates that broad gen-
eral mode of action classiWcation schemes are feasible, espe-
cially when advanced concepts such as kinetics and received
dose metrics are included.
3.4. Statistics and models
The statistical challenges of mixtures study design have
been covered extensively and the issues fully vetted in the
literature. This rich discussion and debate is quite special-
ized and appears to have been, unfortunately, largely unap-
preciated within the toxicological community. A thorough
discussion of statistical issues is beyond the scope of this
review. The reader is referred to excellent discussions found
in Berenbaum (1989), Carter and Carchman (1988), Carter
and Gennings (1994), Gennings and Carter (1995), Gessner
(1995), Greco et al. (1995) and US EPA (1990). More exten-
sive interaction between statisticians and toxicologists will
surely be required to advance the Weld of environmental
mixture assessment. Not only is there an increasing need for
statistically sound interaction studies to be conducted, but
innovative approaches for addressing the eVects of low lev-
els of chemicals present in mixtures will need to be devel-
oped. Approaches that incorporate sound statistical
approaches for interaction assessment into biologically
based modeling (e.g., PBPK) will likely deWne future
research on chemical mixtures.
3.5. Government and related regulatory publications
Over the years, regulatory methods for assessing mix-
tures have continued to increase in complexity and sophis-
tication. These enhancements focus primarily on the
extrapolation of data from single chemicals, or simple mix-
tures of only a few chemicals, to predict the toxicity of more
complex mixtures. In addition, more attention has been
given to the use of data on chemical interactions in mixture
risk assessments. SpeciWc methodologies have been pro-
posed for modifying standard methods, such as the hazard
index approach, to address mixture toxicity more speciW-
cally. In addition, the use of mechanistic information is a
particular feature that increases through successive guid-
ance documents, for example, those from EPA (US EPA,
1986, 1990, 2000). Mechanistic information is used to pre-
dict the type of combined action for chemicals (additivity,
synergy, antagonism) and to judge the toxicological simi-
larity of diVerent mixtures. The rationale behind this is that
a more extensive use of information on toxicological mech-
anisms will enable toxicity information to be extrapolated
from a mixture on which toxicity data are available to one
for which data are lacking.
At present, several underlying assumptions must be
made to utilize mechanistic approaches. A fundamental
assumption is that chemicals with similar modes of action
produce dose additive toxicity and that chemicals with dis-
similar modes of action produce response additive (inde-
pendent) toxicity. Although there is a theoretical basis for
this assumption and some empirical data to support it,
many exceptions have been observed, i.e., interactions. The
second assumption is that chemicals can be readily grouped
based on their modes or mechanisms of action. The meth-
odology for making mechanistic groupings is currently
vague and criteria for determining mechanistic similarity
are only beginning to be developed. No rigorous testing has
been conducted that can be used to validate proposed crite-
ria or methodologies for grouping chemicals according to
mechanisms of action. A third assumption implicit in this
discussion is that clear mechanistic data can be developed
for most chemicals. An example that calls into question this
assumption is that the mechanism of action has remained
elusive even for some of the most widely used and exten-
sively studied medications.
Despite potential diYculties, the focus on mechanistic
information to predict mixture toxicity represents a poten-
tial advancement in the sophistication of mixtures assess-
ment if the underlying assumptions can be further tested
and either validated or reW
ned further. In like fashion, using
mixture similarity to extrapolate data from a well-charac-
terized mixture to other, less well characterized mixtures
requires assumptions about structure–activity relation-
ships, dose–response characteristics and interactions in
mixtures. Methodologies and clear criteria for making such
assessment appear to be in the formative stages of develop-
4. Discussion
Based on the above summaries some additional com-
ments on theory, policy, and regulatory practice for mix-
tures of organic chemicals can be made. There are good
general deWnitions of several key mixture-related toxicity
concepts. A mechanism of toxicity is deWned as a molecular
sequence of events from absorption of an eVective dose to
production of a speciWc biological response. A mode of
toxic action is a set of physiological and behavioral signs
Page 17
136 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
characterizing an adverse biological response. The deWni-
tion of mechanism includes mode but the more general deW-
nition of mode is not necessarily restricted to a single
mechanism. An “interaction” greater than additivity (some-
times referred to as synergism) or less than additivity
(sometimes referred to as antagonism) is inferred if the level
of response produced by any combination of diVerent
agents diVers from the response expected on the basis of a
theoretical model of non-interaction.
Although more philosophical than technical, the above
toxicity deWnitions represent a solid starting point. One
major roadblock to more thorough assessment of mixture
toxicity remains the lack of a comprehensive, generally
accepted mechanism/mode of toxic action classiWcation sys-
tem. Until single chemicals can be grouped into common
types of mechanisms/modes with linked eVect relationships
that span the entire range of levels of biological organiza-
tion, little progress can be made toward accurately predict-
ing mixture toxicity or assessing risks. Some current
examples of mechanism/mode/eVect groupings are in very
early stages of development, as can be seen below in Table 1
(from McCarty, 2002).
Neither the mammalian nor the aquatic system pre-
sented in Table 1 is adequate, although the aquatic classiW-
cation schemes represent examples of practical approaches
which have seen some use. A detailed discussion of the
issues associated with technical deWnitions for mode and
mechanism of toxicity, and current inconsistencies and lim-
itations is presented in Borgert et al. (2004).
The status of a comprehensive classiWcation scheme for
modes/mechanisms of toxicity interactions is similarly
incomplete and represents another major roadblock. It is
generally accepted that there are two basic categories of
combined action: non-interaction and interaction. Non-
interaction has been deWned as additivity, for which there
are two fundamentally diVerent models. One predicts that
the response to a combination of agents can be modeled
like dilutions of the same agent given simultaneously by the
summation of doses of the individual mixture components
(Berenbaum, 1981; Loewe and Muischnek, 1926). This is
usually called “dose addition,” and is the basis for toxic
equivalency factor approaches (Safe, 1990; Safe, 1998). The
second model of non-interaction, often called response
addition, is based on statistical probabilistic independence
(Bliss, 1939) such that the toxicity (and dose–response char-
acter) of each agent is unaVected by other agents in the mix-
ture. The terms “Loewe additivity” and “Bliss
independence,” should be used to distinguish these con-
cepts in their historical context, but the terms “dose addi-
tion and “response addition” (or independence) are more
commonly used in the general literature and regulatory
guidance documents. Recently, a method for unifying these
concepts has been proposed based on detecting diVerences
in the slopes of dose–response curves between mixtures and
their components (Gennings et al., 2005).
Conceptually, dose addition has been associated with
chemicals that have the same mode/mechanism of action
and response addition (independence) with chemicals that
have diVerent modes/mechanisms of action, however, a
strong empirical basis for these distinctions is lacking and
needs more rigorous assessment. The terms synergism and
antagonism, both on a dose and/or response basis, fall into
the interaction category. A number of complicated schemes
for classifying types of toxicity interactions exist, but these
schemes are unwarranted complications given the current
knowledge base and can be readily reduced to the simple
classiWcation scheme of more than additive, additive, and
less than additive (
Borgert et al., 2001; Calabrese, 1991;
Filov et al., 1979). An integrated interaction addition model
has recently been proposed (Rider and Leblanc, 2005).
Besides interactions themselves, there is the issue of
dose-dependence. Both the mode/mechanism of toxic
action and the mode/mechanism of toxicity interactions
can be aVected by the dose, with diVerent eVects and/or
interactions possible at diVerent dose ranges. A chemical
may produce diVerent toxic eVects by diVerent modes/
mechanisms, but these diVerent pathways often become
operative in diVerent dose ranges. Thus, dose inXuences
toxicity because mechanism is a function of dose. This
dose-dependence of toxic action leads to a dose-dependence
Table 1
Examples of classiWcation schemes for grouping toxicological modes and eVects
Mammalian systems Aquatic systems
By type (Amdur et al., 1991) By system (Klaassen et al., 1996) By type (Verhaar et al., 1992) By system (McCarty and Mackay, 1993)
Receptor–ligand binding Carcinogenesis Inert (narcosis) Narcosis
Excitable membranes Genetic Less inert Polar narcosis
Nonlethal mutations Respiratory Polar narcosis Oxidative uncoupler
Binding to biomolecules Immune SpeciWcally acting ACHase inhibitor
Cellular energy production Nervous Reactive Membrane irritant
Selective cell death Reproductive CNS seizure
Calcium homeostasis Developmental Respiratory blocker
Skin Dioxin-like (TCDD)
Heart, vascular, blood
Page 18
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 137
for toxicity interactions because the interaction of two or
more chemicals is determined in part by which mode/mech-
anism of toxic action is operative. This is a critical consider-
ation as much of the available toxicity knowledge base was
obtained with controlled testing where relatively high levels
of exposure were employed. In contrast, many environmen-
tal concerns center on the toxicity of mixtures of multiple
chemicals where each are present at low levels, often below
levels where substantial overt toxicity is attributable to any
single chemical acting alone. As well, the issue of external
versus internal dose metrics, and the typical lack of test-
speciWc information on toxicity modifying factors, contin-
ues to confound interpretation.
In addition to dose magnitude, the inXuence of temporal
dose factors must be considered. In the simplest case of
multiple exposures to the same chemical, the duration of
the exposure period as well as the spacing between expo-
sure episodes can aVect the nature and degree of any
response. Multiple exposures can span a range from a
continuous, constant exposure to continuous, variable
exposure to distinct but additive pulses to sequential inde-
pendent exposures. In mixtures, the complexity of potential
toxicity interactions rises with the number of components,
plus, there is the additional complication of the order of
exposure. For example, the response produced by an expo-
sure of chemical A then B may be diVerent from B then A,
as when a metabolic degradation pathway for A is induced
by an initial exposure to B. Examination of the temporal
inXuence of dose-dependence on mixture toxicity represents
an extremely complex issue that might be better
approached by building upon a viable framework for toxic
action and toxicity interactions.
As well as the fundamental components of toxicity,
interaction, and dose-dependence, two additional aspects
are vital to any mixture toxicity classiWcation scheme that
aims to be of general and widespread utility. The Wrst addi-
tion is consideration of multiple levels of biological organi-
zation. There are scientiWc and regulatory interests in
mixture eVects from biochemistry, through whole organ-
isms, to ecosystems. To ensure the utility any mixture
framework, it should be constructed with information for
diVerent levels of biological organization so that it can be
used broadly. The second addition is consideration of a
wide range of types of living organisms. Because not all
modes/mechanisms of toxic action will be present in each
type of living organism, toxicity interactions will not be
found universally. Thus, dose-dependence will vary among
organisms due both to diVering sensitivity to the same
mode/mechanism and to the lack of some types of modes/
mechanisms in some types of organisms.
Although there are a number of other modifying and
inXuencing factors, any comprehensive framework that
seeks to predict and explain the eVects of chemical mixtures
must take into account the following: the mechanisms of
toxicity of the component chemicals, the potential points at
which these mechanisms interact, the dose-dependence of
both the mechanisms of toxicity and the mechanisms of
interaction, be designed to be used at various levels of bio-
logical organization, and account for species-speciWc diVer-
ences in both toxicity and interaction. We are not aware
that any such comprehensive framework has been
These signiWcant knowledge gaps represent a major
complication thwarting both academic investigations and
regulatory approaches addressing the toxicity of chemical
mixtures. For example, recent guidance documents pub-
lished by ATSDR and EPA recommend using toxicological
data on the mixture of concern, on similar mixtures as sur-
rogates for the mixture of concern, or on toxicity and inter-
actions between individual components of the mixture
depending upon data availability and quality (US DHHS,
2001, 2004a,b,c,d; US EPA, 2000). However, despite the
promulgation of guidance for assessing mixtures, clear cri-
teria have yet to be developed for determining when two
mixtures are suYciently similar to use one as a toxicological
surrogate for the other.
For many mixtures, it may be the chemical similarity of
only a few prominent constituents that is used to conclude
similarity. For other mixtures, similarity might be assumed
from a single common toxicological eVect or a common
mode of toxic action. However, there is no generally
accepted classiWcation scheme for categorizing toxicologi-
cal eV
ects or modes/mechanisms of toxic action (Borgert
et al., 2004; McCarty, 2002; Sexton et al., 1995). Therefore,
no reliable method exists to judge whether two mixtures
exert eVects via a single common toxicological pathway or
by one or more independent toxicological pathways.
As well, because few mixtures have been adequately
characterized and because of technical limitations to such
studies, many risk assessments for mixtures utilize informa-
tion derived from chemical interaction testing. However,
much of the published data on chemical interactions lack
the necessary rigor to be used eVectively in mixture risk
assessment (Borgert et al., 2001; US EPA, 1988). This pre-
sents considerable diYculty for risk assessors who must use
the substantial body of literature on interactions to comply
with methods outlined in recent mixtures guidance docu-
ments. ATSDR has recently developed interaction proWles
for some chlorinated compounds (US DHHS, 2004b,c,d),
but their methodology does not require a rigorous evalua-
tion of interaction data and interpretations.
Without a rigorous understanding of the toxicology of
chemical mixtures, conservatism will dictate that the toxic-
ity of an untested mixture be considered similar to the most
toxic surrogate mixture. Reported interactions between
chemicals will likely be used without regard for the quality
of the interaction data or interpretations, or their relevance
to human and/or environmental risk. Furthermore, toxico-
logical data on whole mixtures and interactions, i.e., data
from the dose range that produces observable toxicity, will
likely be used without regard for the inXuence of dose mag-
nitude on mechanisms of toxicity and mechanisms of inter-
action. In many instances, the resulting overestimates of
risk could lead to risk management decisions that do not
Page 19
138 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
protect public health and/or the environment. For example,
the known health beneWts of breast feeding (Borgert et al.,
2003; LaKind et al., 2002) and drinking aseptic water
(Anderson, 1991) might be foregone to avoid theoretical
risks based on assumed mixture eVects. Yet, the mixture
eVects assumed might be plausible only at doses far above
those that could be obtained from typically low-level envi-
ronmental exposures.
Without a more cohesive framework for understanding
mixture toxicity, mixtures of concern will continue to be
deWned by premises such as source, location, or regulatory
authority, which ignores the simple fact that living organ-
isms eat, drink, breathe, live in, and indeed, are themselves
chemical mixtures. Purporting to identify toxicologically
important subsets of the chemical exposure, i.e., mixtures,
on such arbitrary premises makes for convenient policy,
but is scientiWcally unsupportable. ScientiWcally, one must
acknowledge that toxic (or therapeutic) eVects of single
chemicals are never observed in isolation, but always in the
context of biological and environmental mixtures. Only by
exaggerating dose can we presume to test eVects of single
agents or speciWc mixtures. Dose exaggeration, however,
confounds any generalization of results beyond the experi-
mental conditions, a problem that underscores the need for
developing a comprehensive framework for assessing mix-
tures at environmentally relevant concentrations.
Of necessity, various governmental agencies have
adopted policies and promulgated guidance documents for
the risk assessment of chemical mixtures. Without excep-
tion, these incorporate a blend of convention and empiri-
cism reXecting the amorphous scientiWc literature on
mixtures toxicology noted above. For policy and guidance
to be improved, a more cohesive body of toxicity data on
chemical mixtures must be developed, but such are unlikely
to emerge spontaneously without a more cohesive and
comprehensive conceptual framework to guide research in
mixtures toxicology. This is not meant as a criticism of
either researchers or policy makers, but as recognition of
the status of a very complex issue and to encourage the
breaking of a cycle in which research and policy are inter-
dependent and mutually limiting.
5. Conclusions and recommendations
Chemical mixtures present conceptual challenges not yet
satisfactorily resolved in toxicology. Past eVorts suVer less
from an absence of empirical data and more from a lack of
coherence among those data as there is no generally
accepted theorem for mixture toxicity. Because there is no
widely accepted classiWcation scheme for either mode/
mechanism of toxic action or mechanism of toxicity inter-
actions, current approaches rely on unsubstantiated
notions about similarity or dissimilarity between modes of
toxic action and about relationships between modes of
action and the toxicity of mixtures. These approaches fail
to account for how dose inXuences mechanism of action
and thereby the mechanistic similarity or dissimilarity
between chemicals. They also fail to account for dose-
dependence of mechanisms of toxicity interaction.
Carrying out experimental mixture research without a
well-deWned, working classiWcation scheme for modes/
mechanisms of toxic action and interaction is futile in terms
of advancing the theoretical foundation of toxicology as
the results are not interpretable in a consistent manner.
This is abundantly clear from this examination of the mix-
ture literature. Other than toxicity by narcosis, there has
been only limited progress toward understanding and esti-
mating mixture toxicity, developed using only narrowly
deWned conditions and endpoints with small subsets of
chemicals. As the regulatory thrust for addressing mixtures
is ultimately broad, the only scientiWcally sound approach
is to develop a comprehensive joint scheme for mode/mech-
anisms of toxic action and mechanisms of interaction.
Previous recommendations regarding modes/mecha-
nisms of toxicity and mechanisms of toxicity interaction
related to the issue of mixtures have been made (Borgert
et al., 2004; Borgert et al., 2001; McCarty, 2002) and repre-
sented a starting point for the recommendations that fol-
low. As noted previously there is much experimental data
for both single chemicals and various types of mixtures
which has provided the basis for considerable development
in many technical issues such as experimental design, statis-
tical interpretation, inXuence of modifying factors, etc.
which are important in improving the scientiWc understand-
ing of mixture toxicity.
However, despite the development in these areas,
advances in fundamental toxicological concepts and theo-
ries critical to mixture toxicity has not kept pace. Two vital
theoretical issues are crucial to the advancement of knowl-
edge and practice of mixture toxicity. These two key inter-
related issues are:
development of a mode/mechanism of toxicity classiWca-
tion scheme
development of a mechanism of toxicity interaction clas-
siWcation scheme
These two schemes should be coordinated such that clas-
siWcation designs and associated technical deWnitions
address the issue of translation/conversion across the full
range of biological levels of organization, from biochemical
through whole organisms to ecosystems. To fulWll this
requirement it will be necessary to develop two additional
broad classiWcation schemes; speciWcally:
a dose classiWcation and interconversion scheme
a response/adverse eVect classiWcation and interconver-
sion scheme
Thus, doses should be deWned with multiple dose metrics
(both external or exposure doses and internal or received
doses) appropriate for various levels of organization and
interconversion advice should be speciWed. Similarly,
responses/adverse eVects should be linked so that some
Page 20
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 139
understanding of how various eVects are expressed at various
levels of biological organization and interconversion advice is
also speciWed. Although the relative simplicity which might
be achieved by focusing only an a single level of biological
organization, such as the whole organism, has considerable
attraction, the reality of both available data and toxicological
concerns across the spectrum of the levels of biological orga-
nization dictates that multiple levels must be considered.
Developing broadly applicable, generally accepted
schemes addressing modes/mechanisms of toxic action
and mechanisms of toxicity interaction to facilitate
improvements in understanding and predicting mixture
toxicity will be neither simple nor quick. It requires initi-
ating a long-term development plan. A multi-disciplinary
eVort directed at the issues outlined here will ensure that
mixture-related regulatory activity will be able to draw
upon a sound and continuously improving scientiWc foun-
dation. Adopting the proven, time-honored method of
iterative hypothesis development, testing, and improve-
ment still seems the best approach to examining mixture
toxicity, and although arduous, may well be the most
expedient approach.
The authors thank the Chlorine Chemistry Council of
the American Chemistry Council for Wnancial support that
facilitated this project and Jennifer Muller of Applied Phar-
macology & Toxicology Inc. for technical revision and edit-
ing of the manuscript.
Note: over 100 additional publications were identiWed in the literature
search. For various reasons (no original data or data already cited in
another paper, material addressed in another paper, etc.) these documents
were not cited.
Altenburger, R., Boedeker, W., Faust, M., Grimme, L.H., 1990. Evaluation
of the isobologram method for the assessment of mixtures of chemi-
cals. Ecotox. Environ. Safety 20, 98–114.
Amdur, M.O., Doull, J., Klaassen, C.D. (Eds.), 1991. Casarett & Doull’s
Toxicology: The Basic Science of Poisons, fourth ed. McGraw-Hill,
New York.
Anderson, C., 1991. Cholera epidemic traced to risk miscalculation.
Nature 354, 255.
Berenbaum, M.C., 1981. Criteria for analyzing interactions between bio-
logically active agents. Adv. Cancer Res. 35, 269–335 (As cited in Cala-
brese, 1991).
Berenbaum, M.C., 1989. What is synergy? Pharmacol. Rev. 41, 93–141.
Binderup, M.L, Dalgaard, M., Dragsted, L.O., Hossaini, A., Ladefoged, O.,
Lam, H.R., Larsen, J.C., Madsen, C., Meyer, O., Rasmussen, E.S., ReVst-
rup, T.K., Soborg, I., Vinggaard, A.M., Ostergard, G., 2003. Combined
actions and interactions of chemicals in mixtures: the toxicological eVects
to exposures of mixtures of industrial and environmental chemicals.
Report 2003:12. Danish Veterinary and Food Administration, Soborg.
Bliss, C.I., 1939. The toxicity of poisons applied jointly. Ann. Appl. Biol.
26, 585–615.
Borgert, C.J., Price, B., Wells, C.S., Simon, G.S., 2001. Evaluating chemical
interaction studies for mixture risk assessment. Hum. Ecol. Risk
Assess. 7, 259–305.
Borgert, C.J., LaKind, J.S., Witorsch, R.J., 2003. A critical review of meth-
ods for comparing estrogenic activity of endogenous and exogenous
chemicals in human milk and infant formula. Environ. Health Per-
spect. 111, 1020–1036.
Borgert, C.J., Quill, T.F., McCarty, L.S., Mason, A.M., 2004. Can mode of
action predict mixtures toxicity for risk assessment? Toxicol. Appl.
Pharmacol. 201, 85–96.
Broderius, S., Kahl, M., 1985. Acute toxicity of organic chemical mixtures
to the fathead minnow. Aquat. Toxicol. 6, 307–322.
Butterworth, B.E., BogdanVy, M.S., 1999. A comprehensive approach for
integration of toxicity and cancer risk assessments. Regul. Toxicol.
Pharmacol. 29, 23–36.
Butterworth, B.E., Conolly, R.B., Morgan, K.T., 1995. A strategy for estab-
lishing mode of action of chemical carcinogens as a guide for
approaches to risk assessments. Cancer Let. 93, 129–146.
Calabrese, E.J., 1991. Multiple Chemical Interactions. Lewis Publishers,
Chelsea, MI.
Calabrese, E.J., Mehendale, H.M., 1996. A review of the role of tissue
repair as an adaptive strategy: why low doses are often non-toxic and
why high doses can be fatal. Food Chem. Toxicol. 34, 301–311.
Carter Jr., W.H., Carchman, R.A., 1988. Mathematical and biostatistical
methods for designing and analyzing complex chemical interactions.
Fund. Appl. Toxicol. 10, 590–595.
Carter Jr., W.H., Gennings, C., 1994. Analysis of chemical combinations:
relating isobolograms to response surfaces. In: Yang, R.S.H. (Ed.),
Toxicology of Chemical Mixtures. Case Studies, Mechanisms and
Novel Approaches Academic Press, San Diego, pp. 643–663.
Cassee, F.R., Groten, J.P., van Bladeren, P.J., Feron, V.J., 1998. Toxicolog-
ical evaluation and risk assessment of chemical mixtures. Crit. Rev.
Toxicol. 28, 73–101.
Chambers, J.E., Carr, R.L., 1995. Biochemical mechanisms contributing to
species diVerences in insecticidal toxicity. Toxicology 105, 291–304.
Chambers, J.E., Dorough, G.D., 1994. Toxicologic problems associated
with pesticide mixtures and pesticide impurities. In: Yang, R.S.H. (Ed.),
Toxicology of Chemical Mixtures. Case Studies, Mechanisms and
Novel Approaches Academic Press, San Diego, pp. 135–155.
Chen, J.J., Chen, Y-J., Rice, G., Teuschler, L.K., Hamernik, K., Protzei, A.,
Dodel, R.L., 2001. Using dose addition to estimate cumulative risks from
exposures to multiple chemicals. Regul. Toxicol. Pharmacol. 31, 35–41.
Chen, D.G., Pounds, J.G., 1998. A nonlinear isobologram model with Box-
Cox transformation to both sides for chemical mixtures. Environ.
Health Perspect. 106, 1367–1371.
Christensen, E.R., Chen, C.Y., 1985. A general noninteractive multiple tox-
icity model including probit, logit, and Weibull transformations. Bio-
metrics 41, 711–725.
Craig, D.K., Baskett, R.L., Davis, J.S., Dukes, L.L., Hansen, D.J., Powell,
T.J., Petrocchi, A.J., Sutherland, P.J., 1997. Comprehensive default
methodology for the analysis of exposures to mixtures of chemicals
accidentally released to the atmosphere. Contract Number: AC09-
96SR18500. US Department of Energy, Washington, DC.
Cralley, L.J., Cralley L.V. (Eds.), 1985. Patty’s Industrial Hygiene and Tox-
icology. Volume 3. Theory and Rationale of Industrial Hygiene Prac-
tice. The Work Environment, second ed. John Wiley and Sons, New
York, NY (Chapter 3A).
Drescher, K., Boedeker, W., 1995. Assessment of the combined eVects of
substances: the relationship between concentration addition and inde-
pendent action. Biometrics 51, 716–730.
Drummond, R.A., Russom, C.L., Geiger, D.L., DeFoe, D.L., 1986. Behav-
ioral and morphological changes in fathead minnows (Pimephales
promelas) as diagnostic endpoints for screening chemicals according to
mode of action. In: Poston, T.M., Purdy, R., (Eds.), Aquatic Toxicology
and Environmental Fate: Ninth Symposium. ASTM STP 921. Wiley,
New York, pp. 415–435.
Durkin, P., Hertzberg, R., Stiteler, W., Mumtaz, M., 1995. The identiWca-
tion and testing of interaction patterns. Toxicol. Lett. 79, 251–264.
ECETOC, 2001. Aquatic toxicity of mixtures. Technical Report No. 80.
European Centre for Ecotoxicology and Toxicology of Chemicals,
Brussels. 64 p.
Page 21
140 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
Eide, I., Johnsen, H.G., 1998. Mixture design and multivariate analysis in
mixture research. Environ. Health Perspect. 106, 1373–1376.
el-Masri, H.A., Thomas, R.S., Benjamin, S.A., Yang, R.S.H., 1995. Physio-
logically based pharmacokinetic/pharmacodynamic modeling of chem-
ical mixtures and possible applications in risk assessment. Toxicology
105, 275–282.
el-Masri, H.A., Reardon, K.F., Yang, Y.S., 1997. Integrated approaches for
the analysis of toxicologic interactions of chemical mixtures. Crit. Rev.
Toxicol. 27, 175–197.
Escher, B.I., Hermens, J.L.M., 2002. Modes of action in ecotoxicology:
their role in body burdens, species sensitivity, QSARs and mixture
eVects. Environ. Sci. Technol. 36, 4201–4217.
Escher, B.I., Schwarzenbach, R.P., 2002. Mechanistic studies on baseline
toxicity and uncoupling of organic compounds as a basis for modeling
eVective membrane concentrations in aquatic organisms. Aquat. Sci.
64, 20–35.
Fenner-Crisp, P.A., 2000. FQPA Science issues: common mechanism of
toxicity and cumulative risk assessment. Regul. Toxicol. Pharmacol. 31,
Feron, V.J., Groten, J.P., van Zorge, J.A., Casse, F.R., Jonker, D., van Blad-
eren, P., 1995a. Toxicity studies in rats of simple mixtures of chemicals
with the same or diVerent target organs. Toxicol. Lett.(82/83), 505–512.
Feron, V.J., Groten, J.P., Jonker, D., Cassee, F.R., van Bladeren, P.J.,
1995b. Toxicology of chemical mixtures: challenges for today and the
future. Toxicology 105, 415–427.
Feron, V.J., Groten, J.P., van Bladeren, P.J., 1998a. Exposure of humans to
complex chemical mixtures: hazard identiWcation and risk assessment.
Arch. Toxicol. Suppl. 20, 363–373.
Feron, V.J., Cassee, F.R., Groten, J.P., 1998b. Toxicology of chemical mix-
tures: international perspective. Environ. Health Perspect. 106, 1281–
Filov, V., Golubev, A., Liublina, E., Tolokontsev, N., 1979. Quantitative
Toxicology: Selected Topics. John Wiley and Sons, New York, NY.
Finney, D.J., 1971. Probit Analysis. Cambridge University Press.
Gennings, C., Carter Jr., W.H., Carchman, R.A., Teuschler, L.K., Simmons,
J.E., Carney, E.W., 2005. A unifying concept for assessing toxicological
interactions: changes in slope. Toxicol. Sci. 88, 287–297.
Gennings, C., 1995. An eYcient experimental design for detecting depar-
ture from additivity in mixtures of many chemicals. Toxicology 105,
Gennings, C., Carter Jr., W.H., 1995. Using concentration-response data
from individual components to detect statistically signiWcant depar-
tures from additivity in chemical mixtures. Biometrics 51, 1264–1277.
Gerig, T.M., Blum, U., Meier, K., 1989. Statistical analysis of the joint
inhibitory action of similar compounds. J. Chem. Ecol. 15, 2403–2412.
Gessner, P.K., 1995. Isobolographic analysis of interactions: an update on
applications and utility. Toxicology 105, 161–179.
Greco, W.R., 2001. Evaluating chemical interaction studies for mixture
risk assessment—comments. Hum. Ecol. Risk Assess. 7, 306.
Greco, W.R., Bravo, G., Parsons, J.C., 1995. The search for synergy: a crit-
ical review from a response surface perspective. Pharmacol. Rev. 47,
Groten, J.P., 2000. Mixtures and interactions. Food Chem. Toxicol. 38,
Groten, J.P., Kuper, F.C., Schoen, E.D., Van Balderen, P.J., Feron, V.J.,
1994. Subacute toxicity studies of a combination of nine chemicals in
rats. In: Poster Presentation at the Second Annual HERL Symposium,
Chemical Mixtures and Quantitative Risk Assessment, Raleigh, NC
(As cited in ILSI Risk Science Institute, 1998a).
Groten, J.P., Schoen, E.D., Feron, V.J., 1996. Use of factorial designs in
combination toxicity studies. Food Chem. Toxicol. 34, 1083–1089.
Groten, J.P., Schoen, E.D., Van Bladeren, P.J., Kuper, C.F., Van Zorge,
J.A., Feron, V.J., 1997. Simultaneous exposure to nine chemicals in
rats—eVects on NOAELs. Fund. Appl. Toxicol. 36, 15–29.
Groten, J.P., Butler, W., Feron, V.J., Kozianowski, G., Renwick, A.G.,
Walker, R., 2000. An analysis of the possibility for health implications
of joint actions and interactions between food additives. Regul. Toxi-
col. Pharmacol. 31, 77–91.
Haas, C.N., 1992. A new approach for the analysis of mixture toxicity
data. Water Sci. Technol. 26, 2345–2348.
Haas, C.N., Cidambi, K., Kersten, S., Wright, K., 1996. Quantitative
description of mixture toxicity: eVect of level of response on interac-
tions. Environ. Toxicol. Chem. 15, 1429–1437.
Haas, C.N., Kersten, S.P., Wright, K., Frank, M.J., Cidambi, K., 1997. Gen-
eralization of independent response model for toxic mixtures. Chemo-
sphere 34, 699–710.
Haddad, S., Charest-Tardif, G., Krishnan, K., 2000a. Physiologically based
modeling of the maximal eVect of metabolic interactions on the kinet-
ics of components of complex chemical mixtures. J. Toxicol. Environ.
Health 61A, 209–223.
Haddad, S., Charest-Tardif, G., Tardif, R., Krishnan, K., 2000b. Validation
of physiological modeling framework for simulating the toxicokinetics
of chemicals in mixtures. Tox. Appl. Pharmacol. 167, 199–209.
Haddad, S., Beliveau, M., Tardif, R., Krishnan, K., 2001. A PBPK model-
ing-based approach to account for interactions in the health risk
assessment of chemical mixtures. Toxicol. Sci. 63, 125–131.
Haddad, S., Krishnan, K., 1998. Physiological modeling of toxicokinetic
interactions: implications for mixture risk assessment. Environ. Health
Perspect. 106 (Suppl. 6), 1377–1384.
Hansen, H., DeRosa, C.T., Pohl, H., Fay, M., Mumtaz, M.M., 1998. Public
health challenges posed by chemical mixtures. Environ. Health Per-
spect. 106, 1271–1280.
Hayes, W.J. Jr., Laws, E.R. Jr. (Eds.), 1990. Handbook of Pesticide Toxi-
cology. Volume 1, General Principles; Volumes 2 and 3, Classes of Pes-
ticides. Academic Press, San Diego, CA.
Hayes, A.W., 2001. Principles and Methods of Toxicology, fourth ed. Tay-
lor and Francis, Philadelphia, PA.
Health Council of the Netherlands, 2002. Exposure to combinations of
substances: a system for assessing health risks. Health Council of the
Netherlands, The Hague, Netherlands.
Heindel, J.J., Chapin, R.E., Gulati, D.K., George, J.D., Price, C.J., Marr,
M.C., Myers, C.B., Barnes, L.H., Fail, P.A., Grizzle, T.B., Schwartz,
B.A., Yang, R.S.H., 1994. Assessment of the reproductive and develop-
mental toxicity of pesticide/fertilizer mixtures based on conWrmed pes-
ticide contamination in California and Iowa groundwater. Fund. Appl.
Toxicol. 22, 605–621.
Heindel, J.J., Chapin, R.E., George, J., Gulati, D.K., Fail, P.A., Barnes,
L.H., Yang, R.S., 1995. Assessment of the reproductive toxicity of a
complex mixture of 25 groundwater contaminants in mice and rats.
Fund. Appl. Toxicol. 25, 9–19.
Henschler, D., Bolt, H.M., Jonker, D., Pieters, M.N., Groten, J.P., 1996.
Experimental designs and risk assessment in combination toxicology:
panel discussion. Food Chem. Toxicol. 34, 1183–1185.
Houk, V.S., Waters, M.D., 1996. Genetic toxicology and risk assessment of
complex environmental mixtures. Drug Chem. Toxicol. 19, 187–219.
IARC (International Agency for Research on Cancer), 1987. IARC
Monographs on the Evaluation of Carcinogenic Risks to Humans.
Preamble, Supplement 6.
ILSI Risk Science Institute, 1998a. The toxicity and risk assessment of
complex mixtures in drinking water. ILSI Risk Science Institute, Wash-
ington, DC.
ILSI Risk Science Institute, 1998b. Assessing the toxicity of exposure to
mixtures of disinfection by-products: research recommendations.
Expert Working Group on Mixtures of Disinfection By-products, ILSI
Risk Science Institute, Washington, DC.
Jonker, D., Woutersen, R.A., Van Bladeren, P.J., Til, H.P., Feron, V.J., 1990.
4-week oral toxicity study of a combination of eight chemicals in rats:
comparison with the toxicity of the individual compounds. Food Chem.
Toxicol. 28, 623–631 (As cited in ILSI Risk Science Institute, 1998a).
Jonker, D., Woutersen, R.A., Van Bladeren, P.J., Til, H.P., Feron, V.J.,
1993. Subacute (4-wk) oral toxicity study of a combination of four
nephrotoxins in rats compared with the toxicity of the individual com-
pounds. Food Chem. Toxicol. 31, 45–52 (As cited in ILSI Risk Science
Institute, 1998a).
Jonker, D., Woutersen, R.A., Feron, V.J., 1994. The additivity assumption
tested for combinations of similarly acting nephrotoxicants. In: Poster
Page 22
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 141
Presentation at the Second Annual HERL Symposium, Chemical Mix-
tures and Quantitative Risk Assessment, Raleigh, NC (As cited in ILSI
Risk Science Institute, 1998a).
Jonker, D., Woutersen, R.A., Feron, V.J., 1996. Toxicity of mixtures of
nephrotoxicants with similar or dissimilar mode of action. Food Chem.
Toxicol. 34, 1075–1082.
Klaassen, C.D. (Ed.), 2001. Casarett & Doull’s Toxicology: The Basic Sci-
ence of Poisons, sixth ed. McGraw-Hill, Medical Publications Division,
New York, NY.
Klaassen, C.D., Amdur, M.O., Doull, J. (Eds.), 1996. Casarett & Doull’s
Toxicology: The Basic Science of Poisons, Wfth ed. McGraw-Hill, New
York, NY.
Kodell, R.L., Ahn, H., Chen, J.J., Springer, J.A., Barton, C.N., Hertzberg,
R.C., 1995. Upper bound risk estimates for mixtures of carcinogens.
Toxicology 105, 199–206.
Konemann, W.H., Pieters, M.N., 1996. Confusion of concepts in mixture
toxicology. Food Chem. Toxicol. 34, 1025–1031.
Kortenkamp, A., Altenburger, R., 1998. Synergisms with mixtures of
xenoestrogens: a reevaluation using the method of isoboles. Sci. Total
Environ. 221, 59–73.
Kortenkamp, A., Altenburger, R., 1999. Approaches to assessing combina-
tion eVects of oestrogenic environmental pollutants. Sci. Total Environ.
233, 131–140.
Krewski, D., Thorslund, T., Withey, J., 1989. Carcinogenic risk assessment
of complex mixtures. Toxicol. Ind. Health 5, 851–867.
Krieger, R. (Ed.), 2001. Handbook of Pesticide Toxicology, second ed.
Academic Press, San Diego, CA.
Krishnan, K., Andersen, M.E., Clewell III, H.J., Yang, R.S.H., 1994. Physi-
ologically based pharmacokinetic modeling of chemical mixtures. In:
Yang, R.S.H. (Ed.), Toxicology of Chemical Mixtures. Case Studies,
Mechanisms and Novel Approaches Academic Press, San Diego, CA,
pp. 399–437.
Krishnan, K., Paterson, J., Williams, D.T., 1997. Health risk assessment of
drinking water contaminants in Canada: the applicability of mixture
risk assessment methods. Regul. Toxicol. Pharmacol. 26, 179–187.
Krishnan, K., Brodeur, J., 1991. Toxicological consequences of combined
exposure to environmental pollutants. Arch. Complex Environ. Studies
3, 1–106.
Krishnan, K., Brodeur, J., 1994. Toxic interactions among environmental
pollutants: corroborating laboratory observations with human experi-
ence. Environ. Health Perspect. 102, 11–17.
LaKind, J.S., Birnbach, N., Borgert, C.J., Sonawane, B.R., Tully, M.R.,
Friedman, L., 2002. Human milk surveillance and research of environ-
mental chemicals: concepts for consideration in interpreting and pre-
senting study results. J. Toxicol. Environ. Health Part 65A, 1909–1928.
Lewis, Sr., R.J. (Ed.), 1996. Sax’s Dangerous Properties of Industrial Mate-
rials, ninth ed. Van Nostrand Reinhold, New York, NY.
Lipnick, R.L., 1991. Outliers: their origin and use in the classiWcation of
molecular mechanisms of toxicity. Sci. Total Envir. 109, 131–153.
Loewe, S., Muischnek, H., 1926. EVect of combinations: mathematical
basis of problem. Arch. Exp. Pathol. Pharmakol. 114, 313–326.
Loomis, T.A., Hayes, A.W., 1996. Loomis’s Essentials of Toxicology,
fourth ed. Academic Press, San Diego, CA.
Lu, F.C., 1996. Basic Toxicology: Fundamentals, Target Organs, and Risk
Assessment, third ed. Taylor & Francis, Washington, DC.
McCarty, L.S., 2002. Issues at the interface between ecology and toxicol-
ogy. Toxicology 181–182, 497–503.
McCarty, L.S., Mackay, D., 1993. Enhancing ecotoxicological modeling
and assessment: body residues and modes of toxic action. Environ. Sci.
Technol. 27, 1719–1728.
McInnes, G.T., Brodie, M.J., 1988. Drug interactions that matter. A critical
reappraisal. Drugs 36, 83–110.
Mehendale, H.M., 1995. Toxicodynamics of low level toxicant interactions
of biological signiWcance: inhibition of tissue repair. Toxicology 105,
Mileson, B.E., Chambers, J.E., Chen, W.I., Bettbarn, W., Ehrich, M., Eldef-
rawi, A.T., Gaylor, D.W., Hamernik, K., Hodgson, E., Karczmar, A.G.,
Padilla, S., Pope, C.N., Richardson, R.J., Saunders, D.R., Sheets, L.P.,
Sultatos, L.G., Wallace, K.B., 1998. Common mechanism of toxicity: a
case study or organophosphorus pesticides. Tox. Sci. 41, 8–20.
Mumtaz, M.M., DeRosa, C.T., Durkin, P.R., 1994. Approaches and chal-
lenges in risk assessments of chemical mixtures. In: Yang, R.S.H. (Ed.),
Toxicology of Chemical Mixtures. Case Studies, Mechanisms and
Novel Approaches Academic Press, San Diego, CA, pp. 565–597.
Mumtaz, M.M., Poirier, K.A., Colman, J., 1997. Risk assessment for chem-
ical mixtures: Fine-tuning the hazard index approach. J. Clean Tech-
nol. Environ. Toxicol. Occupat. Med. 6, 189–204.
Mumtaz, M.M., DeRosa, C.T., Groten, J., Feron, V.J., Hansen, H., Durkin,
P.R., 1998. Estimation of toxicity of chemical mixtures through model-
ing of chemical interactions. Environ. Health Perspect. 106, 1353–1360.
Mumtaz, M.M., Durkin, P.R., 1992. A weight-of-evidence approach for
assessing interactions in chemical mixtures. Toxicol. Ind. Health 8,
Mumtaz, M.M., Hertzberg, R.C., 1993. The status of interactions in risk
assessment of chemicals. In: Saxena, J. (Ed.), Hazard Assessment of
Chemicals. Taylor and Francis, London, UK, pp. 47–79.
Murphy, S.D., 1980. Assessment of the potential for toxic interactions
among environmental pollutants. In: Galli, C.L., Murphy, S.D., Pao-
letti, R. (Eds.), The Principles and Methods in Modern Toxicology.
Elsevier/North Holland Biomedical Press, Amsterdam.
National Research Council (NRC), 1988. Complex Mixtures: Methods for
In Vivo Toxicity Testing. Committee on Methods for the in vivo Tox-
icity Testing of Complex Mixtures. National Academy Press, Washing-
ton, DC.
OECD, 2001a. Harmonized integrated classiWcation system for human
health and environmental hazards of chemical substances and mix-
tures. OECD Series on Testing and Assessment, Number 33. Environ-
ment Directorate, Organization for Economic Co-operation and
Development, Paris.
OECD, 2001b. Detailed review document on hazard classiWcation for mix-
tures. OECD Series on Testing and Assessment, Number 30. Environ-
ment Directorate, Organization for Economic Co-operation and
Development, Paris.
Paustenbach, D.J., 1994. Occupational exposure limits, pharmacokinetics,
and unusual work schedules. In: Harris, R.L., Cralley, L.J., Cralley,
L.V. (Eds.), Patty’s Industrial Hygiene and Toxicology, third ed., vol. 3,
Part A. John Wiley and Sons, New York.
Plackett, R.L., Hewlett, P.S., 1967. A comparison of models for quantal
responses to mixtures of drugs. Biometrics 23, 27–44.
Pope, C.N., 1999. Organophosphorus pesticides: do they all have the same
mechanism of toxicity? J. Toxicol. Environ. Health 2B, 161–181.
Pozzani, U.C., Weil, M.A., Carpenter, C.P., 1959. The Toxicological Basis
of Threshold Limit Values: 5. The experimental inhalation of vapor
mixtures by rats, with notes upon the relationship between single dose
inhalation and single dose oral data. American Industrial Hygiene
Association, Chicago, IL.
Price, B., Borgert, C.J., Wells, C., Simon, G.S., 2002. Assessing toxicity of
mixtures: the search for economical study designs. Human Ecol. Risk
Assess. 8, 305–326.
Putzrath, R.M., 2000. Reducing uncertainty of risk estimates for mixtures
of chemicals within regulatory constraints. Regul. Toxicol. Pharmacol.
31, 4–52.
Rand, G.M., Wells, P.G., McCarty, L.S., 1995. Chapter 1: Introduction to
Aquatic Toxicology. In: Rand, G.M. (Ed.), Fundamentals of Aquatic
Toxicology II: EVects, Environmental Fate, and Risk Assessment Tay-
lor and Francis, Bristol, PA, pp. 3–67.
ReVstrup, T.K., 2002. Combined actions of pesticides in food. Report
2002:19. Danish Veterinary and Food Administration, Soborg.
Rider, C.V., Leblanc, G.A., 2005. An integrated addition and interaction
model for assessing toxicity of chemical mixtures. Toxicol. Sci. 87, 520–
Rosenkranz, H.S., Cunningham, A.R., 2001. Chemical categories for
health hazard identiWcation: a feasibility study. Regul. Toxicol. Phar-
macol. 33, 313–318.
Russom, C.L., Bradbury, S.P., Broderius, S.J., Hammermeister, D.E.,
Drummond, R.A., 1997. Predicting modes of toxic action from chemi-
Page 23
142 L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143
cal structure: acute toxicity in the fathead minnow (Pimephales prom-
elas). Environ. Toxicol. Chem. 16, 948–967.
Safe, S., Wang, F., Porter, W., Duan, R., McDougal, A., 1998. Ah receptor
agonists as endocrine disruptors: antiestrogenic activity and mecha-
nisms. Toxicol. Lett. 102–103, 343–347.
Safe, S., 1990. Polychlorinated biphenyls (PCBs), dibenzo-p-dioxins
(PCDDs), dibenzofurans (PCDFs), and related compounds: envi-
ronmental and mechanistic considerations which support the devel-
opment of toxic equivalency factors (TEFs). Crit. Rev. Toxicol. 21,
Safe, S., 1998. Hazard and risk assessment of chemical mixtures using the
toxic equivalency factor approach. Environ. Health Perspect. 106
(Suppl. 4), 1051–1058.
Seed, J., Brown, R.P., Olin, S.S., Foran, J.A., 1995. Chemical mixtures: cur-
rent risk assessment methodologies and future directions. Regul. Toxi-
col. Pharmacol. 22, 76–97.
Sexton, K., Beck, B.D., Bingham, E., Brain, J.D., DeMarini, D.M., Hertz-
berg, R.C., O’Flaherty, E.J., Pounds, J.G., 1995. Chemical mixtures
from a public health perspective: the importance of research for
informed decision making. Toxicology 105, 429–441.
Shirazi, M.A., Linder, G., 1991. A model of additive eVects of mixtures of
narcotic chemicals. Arch. Environ. Contam. Toxicol. 21, 183–189.
Simmons, J.E., 1994. Nephrotoxicity resulting from multiple chemical
exposures and chemical interactions. In: Yang, R.S.H. (Ed.), Toxicol-
ogy of Chemical Mixtures. Case Studies, Mechanisms and Novel
Approaches Academic Press, San Diego, CA, pp. 335–360.
Simmons, J.E., 1995. Chemical mixtures: challenge for toxicology and risk
assessment. Toxicology 105, 111–119.
Simmons, J.E., 1996. Application of physiologically based pharmacokinet-
ics modeling to combination toxicology. Food Chem. Toxicol. 34,
Smyth, H.F., Weil, C.S., West, J.S., Carpenter, C.P., 1969. An exploration
of joint toxic action: twenty-seven industrial chemicals intubated in
rats in all possible pairs. Toxicol. Appl. Pharmacol. 14, 340–347.
Smyth, H.F., Weil, C.S., West, J.S., Carpenter, C.P., 1970. An exploration
of joint toxic action. II. Equitoxic versus equivolume mixtures. Toxicol.
Appl. Pharmacol. 17, 498–503.
Soni, M.G., Mehendale, H.M., 1998. Role of tissue repair in toxicologic
interactions among hepatotoxic organics. Environ. Health Perspect.
106, 1307–1317.
Sowinski, E.J., Cavender, F.L., 1993. Criteria for identifying and classify-
ing toxic properties of chemical substances. In: Clayton, G.D., Clayton,
F.E. (Eds.), Patty’s Industrial Hygiene and Toxicology, fourth ed., vol.
2, Part A. John Wiley and Sons, New York. pp. 77–105.
Suhnel, J., 1996. Zero interaction response surfaces for combined-action
assessment. Food Chem. Toxicol. 34, 1151–1153.
Svendsgaard, D.J., Hertzberg, R.C., 1994. Statistical methods for the toxi-
cological evaluation of the additivity assumption as used in the Envi-
ronmental Protection Agency Chemical Mixture Risk Assessment
Guidelines. In: Yang, R.S.H. (Ed.), Toxicology of Chemical Mixtures.
Case Studies, Mechanisms and Novel Approaches Academic Press,
San Diego, CA, pp. 599–642.
Teuschler, L.K., Dourson, M.L., Stiteler, W.M., McClure, P., Tully, H.,
1999. Health risk above the reference dose for multiple chemicals.
Regul. Toxicol. Pharmacol. 30, S19–S26.
Teuschler, L.K., Hertzberg, R.C., 1995. Current and future risk assessment
guidelines, policy, and methods development for chemical mixtures.
Toxicology 105, 137–144.
UK Committee on Toxicity of Chemicals in Food, Consumer Products
and the Environment, 2002. Risk Assessment of mixtures of pesticides
and similar substances. Food Standards Agency, United Kingdom
Government, London, United Kingdom.
US DHHS, 2001. Guidance manual for the preparation of an interaction
proWle. US Department of Health and Human Services, Public Health
Service, Agency for Toxic Substances and Disease Registry (ATSDR),
Washington, DC.
US DHHS, 2004a. Guidance manual for the assessment of joint toxic
action of chemical mixtures. US Department of Health and Human
Services, Public Health Service, Agency for Toxic Substances and Dis-
ease Registry (ATSDR), Washington, DC.
US DHHS, 2004b. Interaction proWle for: persistent chemicals found in
breast milk (chlorinated dibenzo-p-dioxins, hexachlorobenzene, p,p-
DDE, methylmercury, and polychlorinated biphenyls). US Department
of Health and Human Services, Public Health Service, Agency for Toxic
Substances and Disease Registry (ATSDR), Washington, DC.
US DHHS, 2004c. Interaction proWle for: persistent chemicals found in
Wsh. (chlorinated dibenzo-p-dioxins, hexachlorobenzene, p,p-DDE,
methylmercury, and polychlorinated biphenyls). US Department of
Health and Human Services, Public Health Service, Agency for Toxic
Substances and Disease Registry (ATSDR), Washington, DC.
US DHHS, 2004d. Interaction proWle for: 1,1,1-trichloroethane, 1,1-dichlo-
roethane, trichloroethylene, and tetrachloroethylene. US Department of
Health and Human Services, Public Health Service, Agency for Toxic
Substances and Disease Registry (ATSDR), Washington, DC.
US EPA, 1986. Guidelines for the health risk assessment of chemical mix-
tures. Federal Register. US Environmental Protection Agency, Wash-
ington, DC.
US EPA, 1988. Technical support document on risk assessment of chemi-
cal mixtures. EPA/600/8/-90-064. US Environmental Protection
Agency, Washington, DC.
US EPA, 1989. Risk assessment guidance for superfund. Volume 1.
Human Health Evaluation Manual (Part A). US Environmental Pro-
tection Agency, Washington, DC.
US EPA, 1990. Technical support document on risk assessment of chemi-
cal mixtures report (dated November 1988) EPA/600/8-90/064. O
of Health and Environmental Assessment, Cincinnati, OH.
US EPA, 2000. Supplementary guidance for conducting health risk assess-
ment of chemical mixtures. EPA/630/R-00/002. US Environmental
Protection Agency, Washington, DC.
US NTP, 1993. Toxicity studies of a chemical mixture of 25 groundwater
contaminants administered in drinking water to F344/N rats and
B6C3F1 mice. TOX-35 NTIS#PB94-121498. US National Toxicology
Vaal, M.A., VanLeeuwen, C.J., Hoekstra, J.A., Hermens, J.L.M., 2000. Var-
iation in sensitivity of aquatic species to toxicants: Practical conse-
quences for eVect assessment of chemical substances. Environ. Manage.
25, 415–423.
Vaes, W.H.J., Urrestarazu Ramos, E., Verhaar, H.J.M., Hermens, J.L.M.,
1998. Acute toxicity of nonpolar versus polar narcosis: is there a diVer-
ence? Environ. Toxicol. Chem. 17, 1380–1384.
Van Loon, W.M.G.M., Verwoerd, M.E., Wijnker, F.G., vanLeeuwen,
C.J., van Duyn, P., van deGuchte, C., Hermens, J.L.M., 1997. Esti-
mating total body residues and baseline toxicity of complex organic
mixtures in eZuents and surface waters. Environ. Toxicol. Chem.
16, 1358–1365.
Veldstra, H., 1956. Synergism and potentiation with special reference to
the combination of structural analogues. Pharmacol. Rev. 8, 339–387.
Verhaar, H.J.M., van Leeuwen, C.J., Hermens, J., 1992. Classifying envi-
ronmental pollutants. 1: structure–activity relationships for prediction
of aquatic toxicity. Chemosphere 25, 471–491.
Verhaar, H.J., Morroni, J.R., Reardon, K.F., Hays, S.M., Gaver, D.P.J.,
Carpenter, R.L., Yang, R.S., 1997. A proposed approach to study the
toxicology of complex mixtures of petroleum products: the integrated
use of QSAR, lumping analysis and PBPK/PD modeling. Environ.
Health Perspect. 105 (Suppl. 1), 179–195.
Vighi, M., Calamari, D., 1996. Quality objectives for aquatic life: the prob-
lem of mixtures of chemical substances. Hum. Ecol. Risk Assess. 2,
Warne, M.S.J., Hawker, D.W., 1995. The number of components in a
mixture determines whether synergistic and antagonistic or additive
toxicity predominate: the funnel hypothesis. Ecotox. Environ. Safety
31, 23–28.
WHO, 1981. Health eVects of combined exposures in the work environ-
ment. World Health Organization Technical Report Series 662.
Wilkinson, C.F., Christoph, G.R., Julien, E., Kelley, J.M., Kronenberg, J.,
McCarthy, J., Reiss, R., 2000. Assessing the risks of exposures to multi-
Page 24
L.S. McCarty, C.J. Borgert / Regulatory Toxicology and Pharmacology 45 (2006) 119–143 143
ple chemicals with a common mechanism of toxicity: how to cumulate?
Regul. Toxicol. Pharmacol. 31, 30–45.
Woo, Y.T., Di Carlo, F.J., Arcos, J.C., Argus, M.F., Polansky, G., DuBose,
J., 1994. Assessment of carcinogenic hazard of chemical mixtures
through analysis of binary chemical interaction data. Environ. Health
Perspect. 102, 113–118.
Yang, R.S.H., 1996. Some current approaches for studying combination
toxicology in chemical mixtures. Food Chem. Toxicol. 34, 1037–1044.
Yang, R.S., el-Masri, H.A., Thomas, R.S., Constan, A.A., Tessari, J.D.,
1995a. The application of physiologically based pharmacokinetic/phar-
macodynamic (PBPK/PD) modeling for exploring risk assessment
approaches of chemical mixtures. Toxicol. Lett. 79, 193–200.
Yang, R.S., el-Masri, H.A., Thomas, R.S., Constan, A.A., Tessari, J.D.,
1995b. The use of physiologically-based pharmacokinetic/pharmaco-
dynamic dosimetry models for chemical mixtures. Toxicol. Lett. 82-83,
Yang, R.S.H., Thomas, R.S., Gustafson, D.L., Campain, J., Benjamin, S.A.,
Verhaar, H.J.M., Mumtaz, M.M., 1998. Approaches to developing
alternative and predictive toxicology based on PBPK/PD and QSAR
modeling. Environ. Health Perspect. 106, 1385–1393.
Page 25
  • Source
    • "Therefore, climate change and the anthropogenic manipulation of water resources in Mediterranean rivers may lead to enhance human health risks of river water exposure. As humans are exposed to chemical mixtures rather than individual substances, new realistic approaches have been developed to assess the risks associated to combined exposure to sets of pollutants [6] [7]. Classically, two main approaches have been used to evaluate the toxicity of chemical mixtures: concentration addition (CA) and independent action (IA), which assume a similar or different mode of action (MoA), respectively. "
    [Show abstract] [Hide abstract] ABSTRACT: The hazard of chemical compounds can be prioritized according to their PBT (persistence, bioaccumulation, toxicity) properties by using Self-Organizing Maps (SOM). The objective of the present study was to develop an Integrated Risk Index of Chemical Aquatic Pollution (IRICAP), useful to evaluate the risk associated to the exposure of chemical mixtures contained in river waters. Four Spanish river basins were considered as case-studies: Llobregat, Ebro, Jucar and Guadalquivir. A SOM-based hazard index (HI) was estimated for 205 organic compounds. IRICAP was calculated as the product of the HI by the concentration of each pollutant, and the results of all substances were aggregated. Finally, Pareto distribution was applied to the ranked lists of compounds in each site to prioritize those chemicals with the most significant incidence on the IRICAP. According to the HI outcomes, perfluoroalkyl substances, as well as specific illicit drugs and UV filters, were among the most hazardous compounds. Xylazine was identified as one of the chemicals with the highest contribution to the total IRICAP value in the different river basins, together with other pharmaceutical products such as loratadine and azaperol. These organic compounds should be proposed as target chemicals in the implementation of monitoring programs by regulatory organizations.
    Full-text · Article · Jun 2013 · Journal of hazardous materials
  • Source
    • "In the first case they may act through concentration addition (CA) and independent action (IA) mechanisms also referred to as Loewe additivity and Bliss independence . CA is thought to be valid