Som ething fromªN othingº - Eight
W eak Estrogenic Chem icals
Com binedat Concentrations below
N O ECs ProduceSignificant M ixture
E L I S A B E T E
N I S S A N K A R A J A P A K S E , A N D
A N D R E A S K O R T E N K A M P *
Centre for Toxicology, Department of Pharmacology,
The School of Pharmacy, University of London,
29-39 Brunswick Square, London WC1N 1AX
S I L V A ,
We tested whether multicomponent mixtures of xe-
noestrogens would produce significant effects when each
component was combined at concentrations below its
individualNOECorEC01level.Theestrogenic effects ofeight
chemicals of environmental relevance, including hydroxyl-
ated PCBs, benzophenones, parabenes, bisphenol A, and
screen (YES). To ensure that no chemical contributed
disproportionately to the overall combination effect, a
mixture was prepared at a mixture ratio proportional to the
potency of each individual component. The performance
effects (concentration addition, toxicity equivalency
factors, effect summation, and independent action) was
compared. Experimental testing of the predictions revealed
that concentration addition and its application, the
toxicity equivalency factor approach, were valid methods
for the calculation of additive mixture effects. There
In contrast, independent action and effect summation
responses. Crucially, there were substantial mixture
effects even though each chemical was present at levels
well below its NOEC and EC01. We conclude that
estrogenic agents are able to act together to produce
significant effects whencombinedat concentrations below
their NOECs. Our results highlight the limitations of the
traditional focus on the effects of single agents. Hazard
assessments that ignore the possibility of joint action of
underestimations of risk.
The discrepancies between the high concentrations of
estrogenic chemicals that are needed to elicit effects in
laboratory assays and their low levels in the environment
have lent credence to the belief that risks to human health
into account that humans and wildlife are exposed not to
single agents but to mixtures of multiple estrogenic agents,
so-called xenoestrogens. The low concentrations of the
estrogenic chemicalsneed to beinvoked to explain possible
health risks. This has motivated a systematic search for
synergistic combination effects (reviewed in ref 2). In 1996,
a report claiming spectacular synergisms between binary
thepaperhadtobewithdrawn (4) becausetheexperimental
results could not be reproduced by other laboratories (5, 6).
justifiably always heighten concerns about health risks, the
combination effects have not received adequate attention.
Furthermore, it is crucial to explore whether xenoestrogens
can act together to yield measurable responses when
combined at concentrations which individually produce
undetectable effects. Here, we present work that addresses
theseproblemsexperimentally.Westudied theactivation of
thehuman estrogen receptor(alpha) in arecombinantyeast
system, the Yeast Estrogen Screen (YES). The assay utilizes
yeast cells genetically modified to harbor DNA coding for
the alpha human estrogen receptor. Estrogen receptor
plasmids that carry estrogen response elements (ERE) in
tandem with the reporter gene lac-Z. Upon binding of the
receptor-ligand complex to ERE, beta-galactosidase is
expressed and secreted into the culture medium where it
reacts with its substrate chlorophenol red beta-D-galacto-
pyranoside (CPRG) to cause a color change from yellow to
considerable theoretical challenges and is fraught with
conceptual controversies. Synergisms or antagonisms are
generally identified asoutcomesthat exceed, orfall short of,
responses expected from additive interactions of the indi-
additive combination effects should be calculated (8-10).
All too often, mixture effects have been assessed without
explicit reference to additivity expectations (3, 11-14).
A widely used and intuitively appealingway ofpredicting
additiveeffects is based on theexpectation that theeffect of
a mixture should be the arithmetic sum of the effects of its
individual components.It isnot alwaysappreciated that the
generalized use of this method, termed effect summation,
leads to logical inconsistencies with agents that exhibit
sigmoidal dose-effect curves (discussed in ref 2). For this
reason, there are serious doubts as to whether effect sum-
mation can be regarded as a reliable method (8, 9).
Two competing pharmacological concepts for the cal-
culation of expected additive mixture effects have gained
broad acceptance, concentration addition, and independent
action. The origins of concentration addition can be traced
to Fraser (15) and Loewe (16). The concept rests on the
way, such that one can be replaced by an equal fraction of
an equieffective concentration of another, without diminish-
ing the overall mixture effect. By implication, this means
that every mixture component contributes to the overall
combination effect in proportion to its concentration, even
below zero effect levels. There is a consensus that concentra-
*Corresponding author phone/fax: 0044-207-753 5908; e-mail:
10.1021/es0101227 CCC: $22.00
Published on Web 00/00/0000
xxxx American Chemical Society VOL. xx, NO. xx, xxxx / ENVIRON. SCI. & TECHNOL.9A
PAGE EST: 5.9
of action is the topic of a long-standing controversy (8, 10).
of isoboles (16), toxic unit summation (19), and in the toxic
equivalency factor approach (20).
Originally under the name ªindependent joint actionº,
the concept of independent action was developed by Bliss
(21) on the basis of stochastic considerations. It assumes
mixture constituents with different subsystems of an organ-
ism. It is often applied to mixtures composed of agents with
zero effect levels are not expected to contribute to the total
of the concept of independent action (9). Applications of
independent action (23).
The above approaches can be utilized to predict the
ofindividual mixturecomponents, but it isunclearwhich of
the methods should be applied to xenoestrogens in experi-
ments with the YES. In previous work with xenoestrogen
mixtures of up to four components we have found that
concentration addition and independent action produced
almost identical additivity expectations, both with mixtures
of estrogen receptor agonists in the YES (similarly acting
agents) (24) and with combinations of mitogenic agents
showing diverse modes of action in the E-SCREEN assay
(25). In both cases, observed and expected combination
effects agreed well. Since the YES responds to agents that
activate the estrogen receptor (alpha) and is blind to any
othereffects, itmay beexpected thatconcentration addition
should produce valid calculations of additive combination
effects with estrogen receptor agonists. The toxicity equiva-
lency factor approach should also produce accurate predic-
tions, provided the equivalency factors are accurately
components are parallel (26). Conversely, the concepts of
effectsummation andindependentaction shouldprovetobe
However, decisive evidence to support these notions is not
Thus, before addressing the main topic of this papers
combination effects of xenoestrogens at very low effect
a mixture of xenoestrogens where the differences in com-
bination effect predictions derived from concentration ad-
dition and independent action were large enough to be
discriminated experimentally. Drescher and Bo Èdeker (27)
have shown that concentration addition may give higher or
lower (and sometimes identical) mixture effect predictions
than those derived from independent action. This is de-
pendent on the number of mixture components, their
concentration ratio, the steepness of their concentration-
response functions, and the biometrical model used to
describetherelationship between concentration andeffects
ofsingleagents. To avoid thedisproportionatecontribution
of a single xenoestrogen to the overall mixture effect, it was
imperative to choose concentration ratios that reflected the
individual potency of mixture components. Assuming that
the steepness of concentration-response functions of xe-
noestrogens as well as a suitable biometric model are not
open to experimental manipulation, large discriminations
between mixture effect predictions could only be achieved
by increasing the number of mixture components. We
therefore selected eight xenoestrogens for in-depth mixture
The main motivation of our work, i.e., to assess the
at levels that elicit effects indistinguishable from those seen
with untreated controls, made it necessary to estimate as
accurately as possiblelow effect concentrations of all tested
agents. To achieve this, concentration-response relation-
the basis for predictions of entire concentration-response
curves for mixtures of defined composition, assuming
additive combination effects. The predictions were made
using concentration addition, the toxic equivalency factor
use and appeal, effect summation was also included here,
although the approach is theoretically ill-founded. The
predicted (additive) combination effects were then tested
The eight estrogenic chemicals (28, 29) listed in Table 1
such as bisphenol A, a number of 4-hydroxylated polychlo-
rinated biphenyls, parabene, benzophenone, and related
Experim ental Section
purchased from Sigma (Poole, Dorset, U.K.), resorcinol
purity), and 2,4-dihydroxybenzophenone(99% purity) from
Aldrich Chemicals (Poole, Dorset, U.K.), bisphenol A (97%
purity) from Acros Organics (Geel, Belgium), 4′-chlorobi-
from UltraScientific (North Kingston, RI). All agents were
used as supplied and prepared in HPLC-grade ethanol as 1
mM stock solutions. A mixture of test agents was made by
combining appropriatevolumesofstock solutions.All stock
TABLE1. Sum m ary of Param eters for Test Agents inthe YES
7. bisphenol A
8. resorcinol monobenzoate
aDefined in Materials and Methods. Absorbance units for max.bNumbered as in Figure 2.cRatio of the concentration of each component to
total mixture concentration.dConcentration producing effect 0.017 absorbance units, i.e., EC01of maximal response in YES.eCalculated from the
total mixture concentration (1.43 µM) by the multiplication with the fraction of each component in the mixture, see Figure 4.
B9ENVIRON. SCI. & TECHNOL. / VOL. xx, NO. xx, xxxx
stored at -20 °C. Chlorophenol red-beta-D-galactopyrano-
side (CPRG) was obtained from Boehringer (Mannheim,
The yeast estrogen screen was carried out exactly as
described previously (30), following the protocol developed
by Routledge and Sumpter (7). Single agent samples were
times. Nominal concentrations were used.
Dosimetry. The colorimetric readings in the YES assay
werecorrected for absorbances seen with untreated control
cultures (for details see ref 30). Scatterplots of corrected
absorbance readings (ªeffectº) vs log concentration were
constructed. Nonlinear regression analysis was carried out
using the asymmetric Hill function
whereMin and Max aretheminimal and maximal observed
effects,respectively,c istheconcentration oftestagent,EC50
is the concentration of test agent yielding half-maximal
for readings in untreated controls, Min equaled zero. The
were also calculated. Nonlinear regression analysis was
carried out by using SigmaPlot software (version 5.0, SPSS
Inc., Chicago, U.S.A.).
Successful completion of our studies required the prepa-
ration of mixtures with each component present at levels
that did not produce statistically significant effects. One
approach torealizing such ªineffectiveº concentrationsisto
choose the no-observed-effect-concentrations (NOECs) of
testing methods used in estimating NOECs are increasingly
recognized. It has been shown (31) that such procedures all
too easily overlook low-dose effects. It has been argued that
more reliable estimations of low dose effects can be made
by interpolation on the basis of complete dose-response
curves (ªbenchmark concentrationsº) (31). In light of these
considerations we have not only estimated NOECs by using
Dunnett's test (32) but also based our assessments on effect
concentrations yielding 1% of the maximal observed effect
monobenzoate (EC01). EC01values were estimated by inter-
polation using the Hill regression model.
Experimental Approach to Mixture Testing. The estro-
genic activity of the eight-component mixture was deter-
mined and assessed using the fixed mixture ratio design
described by Altenburger et al. (33) and Backhauset al. (34).
was made, with the concentration ratios given in Table 1.
These reflect equieffective concentrations of each chemical
which were established in prescreening runs. The total
concentration of the mixture was varied, and complete
concentration-response relationships were recorded.
Calculation of Predicted Mixture Effects. On the basis
of the Hill regression models for each mixture component,
the joint effects of a mixture with known composition were
calculated by using concentration addition, the toxicity
equivalency factor approach, effect summation, and inde-
their mathematical underpinnings can be found in refs 25
compoundsinduced concentration-dependent increasesin
the activity of beta-galactosidase, indicative of activation of
data showed little experimental variation, as exemplified by
the scatter plot and the regression model for 2′,3′,4′,5′-
chemicals gave plots of similar quality, and the data proved
to be reproducible. Figure 2 shows regression models of all
tested chemicals, including the steroid hormone 17 beta-
estradiol. Important parameters of theregression models of
in Table1.Theregression modelswereusedtoestimateEC01
values of thesechemicals (Table1); NOEC weredetermined
using Dunnett's test. In many cases, the NOEC of the
individual chemicals were lower than their EC01values. The
potencies of the eight estrogenic chemicals varied consider-
ably. By far the most potent of the tested chemicals were
yl-4-ol, and the least potent was resorcinol monobenzoate.
compounds spanned almost 3 orders of magnitude. Slope
parameters and maximal effects of all chemicals were very
et al. (28) for benzyl-4-hydroxyparabene, bisphenol A, and
resorcinol monobenzoate. In our hands, however, dihy-
droxybenzophenone was more potent than communicated
Effect ) Min + (Max - Min)/[1 + (c/EC50) ∧ (-p)]
FIGURE 1. Concentration-response relationship for 2′,3′,4′,5′-
tetrachlorobiphenyl-4-ol intheYES.Experimental effectdata(solid
circles) fromthree independent experiments, withthe regression
model (black solid sigmoidal line) and the upper and lower
interval of the population mean of untreated controls (n ) 20).
FIGURE 2. Regression models for the effects of all eight mixture
components andof 17beta-estradiol (E2). The horizontal solidand
of untreatedcontrols (n) 20). The numbering of the chemicals is
as follows: 1: 2′,3′,4′,5′-tetrachlorobiphenyl-4-ol, 2: 2′,5′-dichlo-
robiphenyl-4-ol, 3: 4-chlorobiphenyl-4-ol, 4: genistein, 5: dihy-
droxybenzophenone, 6: benzyl-4-hydroxyparabene, 7: bisphenol
A, 8: resorcinol monobenzoate.
VOL. xx, NO. xx, xxxx / ENVIRON. SCI. & TECHNOL. 9 C
by these authors (median effect concentrations of 0.7µM vs
Estimation of Combination Effects. Weused theregression
models shown in Figure 2 to predict the mixture effects that
areexpected to occur from additiveinteractionsof theeight
estrogenic chemicals.Thecombined effectswerecalculated
the toxicity equivalency factor approach, effect summation,
and independent action.
The predicted combination effects were tested experi-
produced a complete concentration-response curve, with
experimental variation highest in the linear portion of the
curve. For the entire range of effect levels, the observed
combination effects confirmed decisively the responses
predicted by concentration addition. The regression model
the concentration addition expectation.
The TEFs shown in Table 2 were used to estimate
concentrations of 17beta-estradiol expected to be equief-
fective with the eight-component mixture. The regression
model for the steroid hormone was employed to calculate
a concentration-response curve that could be compared
with the prediction curves of the remaining concepts, this
Confirming that the conditions of the TEF approach were
met, the resulting ªTEF-curveº was congruent with the one
obtained by using concentration addition (Figure 3).
In contrast, the concepts of effect summation and
independent action led to significant underestimations of
the observed combination effects. The predicted median
of 5% of the maximal effect, in reality produced significant
of the independent action prediction curve was comparable
to that of concentration addition. The maximal effect
predicted on the basis of independent action was slightly
higher than experimentally observed. As expected from its
the leveling off of responses normally seen at higher
concentrations. Instead, this concept predicted a steep rise
in response, even beyond the upper limits of maximally
possible effects observable in the YES.
Combination Effects at Concentrations Producing Ef-
fectsIndistinguishablefrom Thosein Untreated Controls.
that would be observed if they acted alone at levels
corresponding to 50% their EC01values. Although none of
the eight xenoestrogens on their own would have produced
combined at these concentrations. The observed mixture
effect, corrected for untreated controls, was 0.39 ( 0.02
absorbance units (mean ( 95% CI of mean response), the
effect predicted by concentration addition was 0.35 absor-
bance units. Simple summation of the individual effects of
all the mixture components underestimated the observed
response by a factor of 20 (0.02 predicted vs 0.39 ( 0.02
In view of the excellent agreement between prediction and
experimental observation, there can be little doubt about
the usefulness of concentration addition for the prediction
and assessment of the joint action of xenoestrogens at the
level of estrogen receptor activation. The predictive power
of concentration addition was apparent for the entire range
FIGURE3. Predictedandobservedeffectsofamixtureofall tested
estrogenic chemicals 1-8(see legendof Figure 2fornumbering).
Experimental effect data (open circles) from three independent
andthe upper andlower confidence limits of the best estimate of
meanresponses (shading).The redsolidline shows the predicted
producedusing thetoxicityequivalencyfactorapproachis almost
here.The lines labeledIA andES are the predictions derivedfrom
independent actionandeffect summation, respectively. The solid
and dashed horizontal lines are the mean and 95% confidence
interval of the mean of untreated controls (n ) 20).
TABLE2. Toxicity Equivalency Factors Derivedfor the Eight
TestedEstrogenic Chem icalsa
compoundtoxicity equivalency factor
7. bisphenol A
8. resorcinol monobenzoate
2.86 × 10-3
3.00 × 10-3
0.281 × 10-3
0.288 × 10-3
0.253 × 10-3
0.224 × 10-3
6.16 × 10-5
1.35 × 10-5
aEstradiol was used as the reference compound for these calcu-
FIGURE 4. Effects of individual mixture components 1-8 at the
concentrations present in 1.43 µM of the mixture. ES: effect
summation, i.e., expected mixture effect obtained by calculating
the arithmetic sum of individual effects of agents 1-8. CA:
concentration addition prediction. MIX: observed mixture effect.
Errorbars are upper95% confidence limits of the bestestimate of
mean responses. Concentrations of test agents in 1.43 µM of the
mixture are depicted in Table 1.
D 9ENVIRON. SCI. & TECHNOL. / VOL. xx, NO. xx, xxxx
the prediction curve and the 95% confidence interval of the
regression model of observed mixture effectssa criterion
which we used to assess our results statistically.
Our data also show that the toxicity equivalency factor
of xenoestrogens. We attribute this to the fact that the
assumptions(26) underlyingtheTEF approachwerefulfilled
in our case, namely that all mixture components acted
through the same pathway (estrogen receptor activation)
were essentially parallel. The TEF approach may, therefore,
be regarded as a convenient and easy to use alternative to
concentration addition. Against this however, we have to
balance the resources and efforts needed to ascertain that
all mixture components indeed produce parallel concentra-
tion-response curves. This is an essential requirement,
because toxicity equivalency factors for individual mixture
components will inevitably vary with the effect level chosen
for analysis, if the curves show different slopes. A dilemma
may arise with mixture components that elicit similar but
not exactly parallel curves. How much deviation from the
requirement of parallelism can be tolerated to justify ap-
plication of the TEF approach?As far as we are aware, there
are no clear-cut, rational criteria to decide this question.
Considering these complications, we favor use of the
concentration addition concept. Unlike the TEF approach,
it copes well with agents showing response curves with
differing slopes (8, 33).
are clearly unsuitable for the assessment of the joint effects
ofxenoestrogensin thisassay.Thisishighlighted by thefact
that both concepts predicted effects of only 0.1 (corrected
absorbanceunits) forthemedian effectconcentration ofthe
mixture(2.3µM). This represents 11% of theexperimentally
observed value of 0.85.
The inappropriateness of independent action may be
theactivation oftheestrogen receptorprotein by binding to
its ligand binding domain. It is blind to all other possible
through yeast cell killing. On theoretical grounds therefore,
independent action could have been ruled out a priori as a
concept suitable for dealing with xenoestrogens in receptor
activation assays, because only similarly acting agents will
beactivein such systems. However, in previous studies (24)
we have observed that concentration addition and indepen-
combination effects in the yeast estrogen screen, which
Hence, evidencewasrequired thatconcentration addition is
indeed appropriate for use with estrogen receptor agonists
in this assay. The present study provides proof of this
hypothesis. This could only be achieved because the dif-
ferences between the additivity expectations derived from
concentration addition and independent action were suf-
ficiently large as not to encounter problems with discrimi-
nating between prediction and observation.
oftheeffect summation concept, ourdemonstration oflarge
underestimations of mixture effects with this approach is of
considerable practical importance. We show that the com-
bined effect of eight agents that individually produce
estrogenic effects (corrected absorbance units) of e.g. 0.5,
0.4, 0.1, 0.15, 0.14, 0.17, 0.21, and 0.45 is not 2.12. The fatal
flaw in simply adding up the individual effects of mixture
components is underlined by the fact that this method
produces combination effect predictions in wild excess of
the biologically possible maximal effect of 1.765! Inevitably
therefore, the effect summation curve in Figure 2 shoots up
steeply to unattainably high responses, beyond all known
biological reality. Crucially, the uncritical use of effect
summation would have led us to the erroneous conclusion
that the joint effect of the eight xenoestrogens studied here
is synergistic. This is because the observed combination
effects far exceeded those predicted by effect summation.
Concentration addition implies that every mixture com-
ponent contributes to the total mixture effect in proportion
to its concentration, even when present at concentrations
below zero effect levels (ªno effect concentrationsº, NEC).
Therefore, does the excellent agreement between experi-
addition prediction mean that not even NECsaresafewhen
Asiswell established, it isimpossibleto determineNECs
with confidence, because true ªzero effectsº cannot be
is estimated as NOEC depends to a large degree on the
biological variability of the test system and the number and
spacing of tested concentrations (31). Often, effects below
reliably from effects seen in untreated controls (31) and
(31, 35, 36). NOECs define a range of concentrations where
low effects can neither be quantified nor ruled out with
So-called benchmark concentrations (ECx) are increas-
ingly viewed as alternatives to NOEC determinations (38,
39), and we became interested in comparing NOECs with
benchmark concentrations corresponding to low effects.
These are estimated by interpolation using concentration-
response relationships (31). Unlike NOEC determinations,
which are derived from comparisons of only two concentra-
tions (ªtreatedº versus ªcontrolsº), the benchmark concen-
ofourstudy, wechoseEC01.Asshown in Table1, theNOECs
determinedforeachsingleagentwereoften lowerthan their
corresponding EC01. We attribute this to the low variability
associated with YES and to the low tested concentrations of
The results of our experiments show convincingly that
xenoestrogens act together to produceeffects when present
at concentrations that individually yield responses indis-
tinguishable from those of untreated controls. There were
joint effects when the test agents were combined at con-
concentrationswerebelow theNOECsin all cases(Table1).
These results put into sharp relief the limitations of the
traditional focus on single agent effects during hazard and
risk assessments of endocrine disrupting chemicals. The
assertion thatindividualestrogenic chemicalsposenoharm
becausethey arepresentatlow, ineffectivelevelsin humans
or in wildlife may be irrelevant when dealing with mixed
mixtureeffect in an additivefashion, well below NOECsand
EC01, the task of identifying the potential risks associated
with estrogenic exposures becomes first and foremost a
question of establishing the sheer number of estrogenic
chemicals in the environment.
focus on searching for synergistic mixture effects with
estrogenic agents has been unnecessary. Clearly, additive
combination effects are of importance and require urgent
may pose for humans and wildlife.
VOL. xx, NO. xx, xxxx / ENVIRON. SCI. & TECHNOL. 9 E
Despite our success in predicting the joint effect of Download full-text
estrogenic chemicalswewould liketohighlight anumberof
factors that are likely to complicate the assessment of
exposure to real existing mixtures of xenoestrogens:
agreement between prediction and observation was due to
the high reproducibility of the YES. It remains to be seen
whether the methodology employed here can be applied
successfully to bioassays that yield less reproducible data.
We have initiated such studies using the E-SCREEN, with
encouraging results (25).
wehad to select xenoestrogens that yielded maximal effects
very similartothoseofsteroidal estrogens.Thisisduetothe
fact that for mathematical reasons, the concept of concen-
tration addition cannot predict mixture effects that exceed
those of the mixture component with the lowest maximal
effect. It is now necessary to explore the predictability of
combination effects with mixtures composed of xenoestro-
gens that produce differing maximal effects (30).
3. Endpoints. In the present study we have explored
receptors. An important task will be to investigate whether
our findings hold true for biological effects at higher levels
of cellular organization, e.g. estrogen-mediated cell prolif-
eration, or at the organism level. A study of in vivo effects
of mixtures of estrogenic chemicals has been recently
Endogenous steroidal estrogens exert quite strong effects at
the levels normally found in body fluids and tissues.
Therefore, it is crucial to explore whether relatively weak
xenoestrogens are able to create an impact on the actions
of endogenous estrogens, when combined at low levels that
individually produce undetectable effects.
Studies addressing these points are underway in our
Uxbridge, U.K. for providing the recombinant yeast cells.
Elisabete Silva is grateful for financial support by Programa
PRAXISXXI-Fundac ¸a ÄoparaaCie ÃnciaeTecnologia,Portugal
and Nissanka Rajapakse for a School of Pharmacy Student-
(1) Safe, S. H. Environ. Health Perspect. 1995, 103, 346-351.
(2) Kortenkamp, A.; Altenburger, R. Sci. Total. Environ. 1998, 221,
(3) Arnold, S. F.; Klotz, D. M.; Collins, B. M.; Vonier, P.; Gillette, L.
J.; McLachlan, J. A. Science 1996, 272, 1489-1492.
(4) McLachlan, J. A. Science 1997, 277, 462-463.
(5) Ramamoorthy, K.; Wang, F.; Chen, I.; Norris, J. D.; McDonnel,
D. P.; Leonard, L. S.; Gaido, K. W.; Bocchinfuso, W. P.; Korach,
K. S.; Safe, S. Endocrinology 1997, 138, 1520-1527.
(6) Ashby, J.; Lefevre, P. A.; Odum, J.; Harris, C. A.; Routledge, E.
J.; Sumpter, J. P. Nature 1997, 385, 494.
(7) Routledge, E.; Sumpter, J. Environ. Toxicol. Chem. 1996, 15,
(8) Berenbaum, M. Pharmacol. Rev. 1985, 41, 93-141.
(9) Greco, W. R.; Bravo, G.; Parsons, J. C. Pharmacol. Rev. 1995, 47,
(10) Po Èch, G. Combined effects of drugs and toxic agents: Modern
evaluation in theory and practice; Springer-Verlag: New York,
(11) Arnold, S.F.;Vonier, P.M.;Collins, B.M.;Klotz, D.M.;Guilette,
L. J., Jr.; McLachlan, J. A. Environ. Health Perspect. 1997, 105,
(12) Soto, A. M.; Sonnenschein, C.; Chung, K. L.; Fernandez, M. F.;
Olea, N.; Olea Serrano, F. Environ. Health Perspect. 1995, 103,
(14) Gray, L.E., Jr;Ostby, J.;Furr, J.;Wolf, C.J.;Lambright, C.;Parks,
L.; Veeramachaneni, D. N.; Wilson, V.; Price, M.; Hotchkiss, A.;
Orlando, E.; Guilette, L. Hormone Reproduction Update 2001,
(15) Fraser, T. R. BMJ 1872, 2, 485-487.
(16) Loewe, S.; Muischnek, H. Arch. Exp. Pathol. Pharmakol. 1926,
(17) EIFAC (European Inland Fisheries Advisory Commission,
fish); 1987. Report on combined effects on freshwater and fish
and other aquatic life of mixtures of toxicants in water. EIFAC
Technical Paper 37; FAO, Rome.
(18) EPA (U.S. Environmental Protection Agency); 1986, Federal
Register 51, 34014-34025.
(19) Sprague, J. B. Water Res. 1970, 4, 3-32.
(20) Safe, S. Crit. Rev. Toxicol. 1990, 21, 51-88.
(21) Bliss, C. I. Ann. Appl. Biol. 1939, 26, 585-615.
(22) Colby, S. R. Weeds 1967, 15, 20-22.
(23) Plackett, R. L.; Hewlett, P. S. Ann. Appl. Biol. 1948, 35, 347-
(24) Payne, J.; Rajapakse, N.; Wilkins, M.; Kortenkamp, A. Environ.
Health Perspect. 2000, 108, 983-987.
2001, 109, 391-397.
(26) Safe, S. Environ. Health Perspect. 1998, 106, 1051-1058.
(27) Drescher, K.; Boedeker, W. Biometrics 1995, 51, 716-730.
(28) Miller, D.; Wheals, B. B.; Beresford, N.; Sumpter, J. P. Environ.
Health Perspect. 2001, 109, 113-138.
(29) Blair, R. M.; Fang, H.; Branham, W. S.; Hass, B. S.; Dial, S. L.;
Moland, C. L.; Tong, W. D.; Shi, L. M.; Perkins, R.; Sheehan, D.
M. Toxicol. Sci. 2000, 54, 138-153.
(30) Rajapakse, N.; Ong, D.; Kortenkamp, A. Toxicol. Sci. 2001, 60,
(31) Moore, D. R. J.; Caux, P. Y. Environ. Toxicol. Chem. 1997, 16,
(32) Dunnett, C. W. Biometrics 1964, 20, 482-491.
M.; Grimme, L. H. Environ. Toxicol. Chem. 2000, 19, 2341-
M.; Grimme, L. H. Environ. Toxicol. Chem. 2000, 19, 2348-
(35) Bruce, R. D.; Versteeg, D. J. Environ. Toxicol. Chem. 1992, 11,
(36) Hoekstra, J. A.; Van Ewijk, P. H. Environ. Toxicol. Chem. 1993,
(37) Crump, K. Risk Analysis 1995, 15, 79-89.
Chem. 1996, 15, 77-79.
(39) Van der Hoeven, N.; Noppert, F.; Leopold, A. Environmetrics
1997, 8, 241-248.
(40) Thorpe, K.L.;Hutchinson, T.H.;Hetheridge, M.J.;Scholze, M.;
Received for review May 2, 2001. Revised manuscript re-
ceived November 20, 2001. Accepted January 17, 2002.
F9ENVIRON. SCI. & TECHNOL. / VOL. xx, NO. xx, xxxx PAGE EST: 5.9