United States
Environmental Protection
Agency
Environmental
Research Laboratory
Gulf Breeze, FL 32561
Research and Development
EPA/600/SR-92/091 August 1992
EPA Project Summary
Statistical Approach to
Predicting Chronic Toxicity of
Chemicals to Fishes from Acute
Toxicity Test Data
F.L. Mayer, G.F. Krause, M.R. Ellersieck, and G. Lee
A methodology and a computer pro-
gram were developed cooperatively by
the U.S. Environmental Protection
Agency (U.S. EPA) Ecological Risk As-
sessment Research Program and the
University of Missouri-Columbia to pre-
dict chronic toxicity of chemicals from
acute toxicity test data. A comprehen-
sive approach to predicting chronic tox-
icity from acute toxicity data was de-
rived in which simultaneous consider-
ation was given to concentration, de-
gree of response, and time course of
effect. A consistent endpoint (lethality)
and degree of response (0%) were used
to compare acute and chronic tests.
The software, Multifactor Probit
Analysis (MPA), calculates the lethal
concentration of a chemical for ex-
pected effect, P (probability of re-
sponse), for extended periods of expo-
sure time. The MPA software is versa-
tile, and the user can choose from sev-
eral probit models and seven different
transformation combinations of the in-
dependent variables. This software is
entirely menu driven, allows the user
to predict concentration of a toxicant
at any time and any percent effect, and
calculates a point estimate and a mea-
sure of dispersion (95% approximate
confidence limits).
Predicted no-effect concentrations
were highly accurate 92% of the time
(within a factor of 2.0 of the limits of
the maximum acceptable toxicant con-
centrations for lethality) when the tech-
nique was applied to a data base of 18
chemicals and 7 fish species. Predic-
tions were also quite accurate for a
pond study and two quail tests. Growth
effects can be estimated from predicted
chronic lethality, but reproductive ef-
fects should not be.
This Project Summary was developed
by EPA's Environmental Research
Laboratory, Gulf Breeze, FL, to an-
nounce key findings of the research
project that is fully documented in a
separate report of the same title (see
Project Report ordering information at
back).
Introduction
Using acute lethality data to estimate
chronic toxicity to fishes customarily in-
volves deriving an application factor or an
acute-chronic ratio, both of which require
acute and chronic toxicity testing. The ap-
plication factor is derived by dividing
the MATC for a compound, as determined
in a chronic toxicity test with a given fish
species, by the acute flow-through LC50
for the same compound tested with the
same species. The acute-chronic ratio
(ACR) is the inverse of AF. The AF or
ACR is then used to estimate chronic no-
effect concentrations for other species for
which only acute toxicity data exist. Both
approaches have limitations in using these
ratios to estimate chronic toxicity.
One limitation is that biological endpoints
and degrees of response are often not
comparable between acute and chronic
toxicity data. When one uses either the
AF or ACR, the acute median lethal con-
centration (LC50) is compared with the
MATC, often derived from an endpoint
other than lethality. Even though the mode
of action for lethality is often assumed to
be the same under acute and chronic
exposures, the mode of action may not be
Printed on Recycled Paper
-------
the same for different endpoints (e.g.,
growth or reproduction compared with le-
thality). Although different degrees of re-
sponse (acute 50% versus chronic no-
affect or 0%) could be used when re-
sponse slopes are similar, the slopes may
be different. Additionally, the use of the
AF or ACR method does not take into
consideration the progression of lethality
through time that is observed in acute
toxicity tests. The concentration-time-re-
sponse interaction has been addressed
previously, but it has been directed to-
ward better defining the LC50. The acute
toxicity value represents only one point in
time (96-h LC50), and the progression of
degree of response with duration of expo-
sure should be essential when one pre-
dicts chronic toxicity from acute toxicity
data.
A more comprehensive, alternative ap-
proach is proposed here in which simulta-
neous consideration is given to concen-
tration, degree of response, and time
course of effect, all of which are usually
included in the results of an acute test,
but seldom used. A consistent endpoint
(lethality) and degree of response (0%)
are used to predict chronic lethality from
acute toxicity tests. Two assumptions may
be required: (1) concentration-response is
a continuum in time, and (2) the mode of
action for lethality is similar under acute
and chronic exposures.
Methods
Simple linear regression (Y - a+bX)
was used to derive lethal concentrations
of no effect (LCO = 0.01%) for each ob-
servation time in an acute toxicity test and
to predict the chronic no-effect concentra-
tion for lethality from those LCDs.
Degree of Response
In chronic toxicity tests, we are most
often interested in the no-effect concen-
tration (e.g., that concentration causing
0% effect), whereas in acute tests, the
degree of response usually used is 50%.
Although a prob'rt value does not exist for
0% or 100%, an approximate value can
be derived. In the use of probit analysis of
acute toxicity data, the probit value used
for 100% mortality is actually the probit
value for 99.99%. An approximate value
for LCO can thus be derived by subtract-
ing the probit value for 99.99% (8.7190)
from 10 to provide a probit value of 1.2810
for 0.01% mortality.
Time Course of Effect
Predicting chronic toxicity from acute
toxicity data requires a means of estimat-
ing the LCO for an indefinite period of time
(chronic) from an acute toxicity test con-
ducted over a finite period of time (96-h
LC50). Approaches to the problem of esti-
mating tolerance over an indefinite time
period have b^en developed by other re-
searchers, although it was with the LD50
or LC50. They noted that as the time of
exposure becomes sufficiently long, the
LD50 or LC50 approaches an asymptotic
value. A hyperbola describes this relation-
ship and can £>e expressed as a straight
line by using the reciprocal of time (t) as
the independent variable. The equation
becomes LD50 = a+b(1/t). Since 1/t ap-
proaches zero as t approaches infinity,
the intercept (a) represents the LD50 over
an indefinite time of exposure.
Technique
The acute toxicity test must be con-
ducted with strict adherence to standard
test methods to obtain estimates of LCO
over time. The times of 24, 48, 72, and 96
h were selected because observations in
standard acute toxicity tests are usually
made at these time periods. Less than
24-h observations were used when avail-
able. Inclusion of these observations is
very important when most toxicity occurs
during the early part of a 96-h test. The
greatest concentration that causes no mor-
tality and the least concentration that
causes complete mortality were used for
0% and 100% responses. All concentra-
tions causing mortality (0%^x<100%) were
also included in our calculations. When
regression analysis could not be conducted
8.719
(less than 3 observations), the highest
nonlethal concentration was used as the
estimate of LCO for that observation time.
Having a range of mortalities for all time
periods is best; although observation times
with only 0 and 100% mortalities are ac-
ceptable if a concentration-response is
evident in time.
Linear regression analysis was used to
calculate the estimated LCO at all obser-
vation times from acute flow-through tests
(Figure 1) as probit % mortality = a+b(log
concentration). The LCD's at each time
period were then regressed against the
reciprocal of time (Figure 2) where LCO =
a+b(1/t). The intercept (a) of this regres-
sion is the predicted no-effect concentra-
tion for chronic lethality. Log transforma-
tions, log LCO = a+b (l/t) or log LCO =
a+b log (l/t), were required for ten tests
because of negative intercepts and/or cur-
vilinear nature of the data.
When test data permits, response-sur-
face models (multiple regression) for ana-
lyzing all data from an acute toxicity test
simultaneously (Figure 3) are preferable
to the two-step simple linear regression
approach described above. We therefore
developed a probit surface methodology
and a user-friendly software program for
simple linear and multiple regression mod-
els to predict chronic toxicity based on
acute time-exposure-effect data. This
method is called Multifactor Probit Analy-
sis (MPA) and uses the iterative reweighed
least squares method to estimate the pa-
i
5.000
1.281
1.2
1.4
0.6 0.8 1.0
Log Concentration
Figure 1. Dose-response curves used to derive the LCO (0.01%) for various observation times in acute
toxicity tests (1.281 = a+bX). Probit % mortality: 1.281 = 0.01%, 5.000 = 50%, and8 719 =
99.99%.
-------
8
10
8.0
6.0
4.0
2.0
E.
F.
centration of toxicant 1, exposure
time, and a third variable which
could be a second toxicant).
Simple probit analysis, using con-
centration as the independent
variable, at each level of expo-
sure time.
, a = 2.9 \ag/L, NOEC = 1.2-3.1 \ig/L
0.010
0.021
1/time (h)
0.042
' Figure 2. Prediction of the chronic no-effect value for lethality from acute toxicity test data with Kepone
andfatheadminnows(LCO=a+b[1/t]).Theintercept(a)representstheLCO(2.9\ig/L)over
an indefinite exposure time, and the maximum acceptable toxicant concentration (MATC)
for chronic lethality was between 1.2 and 3.1 \ig/L
rameters of the probit surface. The inde-
pendent variables consist of time of expo-
sure and concentration of the toxicant.
The dependent variable is the probit of
.the proportion responding to an exposure
concentration. MPA allows the user to pre-
dict the concentration of a toxicant at any
time and percent mortality as well as cal-
culate a measure of variability (95% confi-
dence limits). MPA has two primary func-
tions. The first function is for entering and
editing data not only for the MPA subrou-
tine, but also for other software. Data files
already prepared in ASCII format can be
retrieved using this function. The second
function is statistical analysis. Once a data
set has been entered, a selected MPA
subroutine executes an analysis. The out-
put which is produced depends on the
analysis option chosen. Chi-square and r2
values are used for selecting the best
model. The analysis options are:
A. Simple probit analysis using one
independent variable, which is
concentration. A single exposure
time is assumed.
B. Multifactor probit analysis which
includes two independent vari-
ables; concentration and expo-
sure time. This option assumes
parallel probit regression lines at
each exposure time.
C. Multifactor probit analysis using
concentration, time, and interac-
tion as independent variables.
D.
This option is different than B in
that non-parallel probit lines over
time are assumed.
Multifactor probit analysis with
three independent variables (con-
Simple probit analysis where the
independent variable represent-
ing exposure time is the recipro-
cal of time (1/time).
G. Multifactor probit analysis where
the independent variable repre-
senting exposure time is 1/time.
The calculation of LCOs is dependent
on slope and time course of effect, both of
which are influenced by sample size (num-
ber of fish per concentration) and dose
separation (dilution factor among concen-
trations). In this study, sample sizes ranged
from 10 to 30 organisms and dilution fac-
tors ranged from 50 to 75%.
Data Base
The acute and chronic tests selected
for analyses were conducted at the Co-
lumbia National Fisheries Contaminant
Research Center (U.S. Fish and Wildlife
Service, Columbia, MO) and the U.S. EPA
Environmental Research Laboratory, Gulf
Breeze, FL, on seven fish species: rain-
bow trout, Oncorhynchus mykiss; cutthroat
100
20
20
40
Hours
10
mg/L
60
30
80
100
Figure 3. Acute flow-through toxicity test results with carbon tetrachloride and sheepshead minnows
demonstrating dose-response data in time (96-h LC50 = 19 mg/L).
-------
trout, O. clarki; brook .trout, Salvelinus
fontinalls', lake trout, S. namaycush;
fathead minnow, Pimephales promelas;
channel catfish, Ictalurus punctatus; and
sheepshead minnow, Cyprinodon
variegatus. With the exception of a few
static acute tests used, acute and chronic
tests were conducted in flow-through di-
luter systems. Each diluter delivered four
to seven concentrations of toxicant and a
control Water temperature was maintained
within ±1° C of the desired temperature,
and day length was regulated. Acute and
chronic tests were conducted in accor-
dance with standard procedures, and con-
centrations of all chemicals were mea-
sured.
Acute and chronic flow-through tests
were also conducted with carbon tetra-
chtoride and sheepshead minnows as an-
other test of the LCO method, because
carbon tetrachtoride is considered to have
different modes of action between acute
and chronfe exposures with mammals.
Two additional types of data sets hav-
ing available and acceptable acute toxic-
Hy tests were analyzed — a pond study
with bluegills, Lepomfs macrochirus, and
fluorene, a component of petroleum, and
a study with coturnix quail, Coturnix
Japonka, and mercuric or methyl mercuric
chloride. Fourteen 0.08 ha ponds were
treated with various concentrations of
fluorene (July 26, 1982). Ponds were
drained approximately 70 days after ex-
posure (early October 1982), and fish were
counted, measured, and weighed to de-
termine survival, growth, and production
of recruits. Fluorene exposures in the
ponds were based on average measured
concentrations following treatment on days
1,3, and 7. Acute toxicity tests with blue-
gills were conducted in the laboratory un-
der static conditions to simulate pond ex-
posures. Five-day acute dietary tests were
conducted with coturnix quail by present-
Ing the chemicals at various concentra-
tions in turkey starter mash for 5 days.
Daily observations for evidence of toxicity
were made until clinical signs were no
longer detectable (10 days). Chronic tox-
icity was determined by feeding the
mercurials at various concentrations in ad
libitum diets from hatching to adulthood (9
weeks).
Results and Discussion
Predicted values were compared with
the observed values of chronic tests (early
life-stage and partial and full life cycle
toxfcity tests) and proved highly accurate
for a variety of chemicals and fish species
(Table 1). The predicted no-effect con-
centrations (PNEC) ware very close to or
within the limits (highest concentration w'rth-
Table 1. Comparison of Observed Maximum AcceptableToxicant Concentrations (MATC) and
Predicted No-Effect Concentrations (PNEC) for Lethality Based on Flow-Through Acute
Tests.
Chemical and species
Log
Kow
MATC
PNEC
Butyl benzyl phthalate 4.44
Fathead minnows
Carbon tetrachlpride 2.64
Sheepshead fninnows
Chlordane \ 5.80
Sheepshead,minnows
Complex efflueht
Fathead minnows
2,4-D Butyl ester 2.81
Cutthroat trout
Lake trout <
2,4-D PGBEE ; 4.88
Cutthroat trout
Lake trout i
Endosulfan ', 4.90-6.00 *
Sheepshead 'minnows
Endrin 4.56-5.30
Sheepshead minnows
EPN 4.80
Sheepshead minnows
Fluridone 1.87
Channel catfish
Heptachlor 5.44
Sheepshead minnows
Kepone ; 6.08
Fathead minnows
Methoxychlor 4.20
Rainbow trout
Sheepshead minnows
Pentachlorophenol 5.01
Fathead minnows
Phorate 3.50
Sheepshead minnows
PydraulSOE , 4.62-6.08'
Fathead minnows
t
TFM :
Brook trout
>360
4,5004X411,200
7.14X417
2.04X43.5%
244x^44
334x460
314x460
524X4100
1.14X42.5,0.924X42.1
0.124X40.31
4.14X47.9
1,0004X42,000
1.94x42.8,2.24x43.5
1.24X43.1
1.14x43.1
124x423,234x448
>142
0.244X40.41
3174X4752
4,0004X48,800
635 •
10,427
14'
5.2%
112
67'
59
74
1.4
0.12
3.9°
1,182°
2.6
2.9
0.94 "•"
12, 12, 17"
240
0.15*
592
4,311
Toxaphene 4.83
Brook trout ,
Brook trout (adult)
Fathead minnows
Channel catfish
Sheepshead minnows
0.0684X40.14
0.144X40.29
0.624X41.3
0.074X40.13
1.14x42.5
0.041
1.4°
1.7°
0.057
0.77*
Log transformation of LCD's required.
b Endosulfan I = ^.90, Endsulfan II = 6.0.
° Based on static lest.
d Acute toxicity test for rainbow trout was not available and PNEC was based on brook trout test because of similarity
In response to toxicants.
• Pydraul 50E is a hydraulic fluid consisting of three components; triphenyl phosphate = 4.62,
nonylphenyl diphenyl phosphate = 5.93, cumylphenyl diphenyl phosphate =6.08.
-------
out effect on survival and the next higher
concentration with a significant effect) of
the maximum acceptable toxicant concen-
trations (MATC) for lethality and varied by
less than a factor of two from an MATC
92% of the time. The other 18% of the
predictions (two observations) consisted
of factors of 2.5 and 4.8 of the observed
concentrations.
The technique worked very well in pre-
dicting chronic lethality of carbon tetra-
chloride to sheepshead minnows (PNEC
= 10.4 mg/L, observed = 4.532 ug/g for mer-
curic chloride and 2.0^x<8.0 u.g/g for me-
thyl mercuric chloride with PNECs of 226
and 1.3 fig/g, respectively.
The technique for deriving PNECs uses
some aspects of concepts developed pre-
viously. Acute tests have been conducted
until the toxicity curve becomes parallel to
the time axis, indicating a threshold con-
centration. An incipient LC50 is then esti-
mated by selecting an exposure time from
the asymptotic part of the toxicity curve.
The reciprocal of mean survival times
within concentrations was used as early
as 1917. Regressing the reciprocal of
mean survival time on concentration to
Tables. Regression Correlations ' of Survival and Growth No Observed Effect Concentrations
Among all Fish Species and Chemicals at Various Time Periods.
Analysis and days
of exposure
Weight vs. survival
30
60
90
n
6
10
15
Intercept
(a)
0.395
0.682
0.194
Slope "
(b)
0.920
0.901
0.993
Coefficient of
Determination
(r1)
0.931
0.901
0.916
y±95% C.I.
4.35±0.41
4.63+0.31
4.33±0.22
Length vs. survival
30
60
90
16 0.284 0.968
17 0.263 0.965
18 0.275 0.971
0.945
0.949
0.941
4.64±0.18
4.60±0.15
4.41±0.17
* Log y = a+b(log x), where y is no-effect concentration for survival and x is no-effect concentration (ng/L) for growth
(length or weight).
b All slopes were significantly different from 0 (pso.01).
derive theoretical thresholds of toxicity was
further developed during 1957-67. Al-
though observing survival times in acute
tests has merits, it is laborious and is only
infrequently done.
The approach of incorporating all data
in an acute test (concentration, degree of
response, and time course of effect) to
predict chronic lethality has a technical
basis. During the last 20 years, fish chronic
toxicity tests have been shortened from
full life cycle tests to 30- to 90-day early
life stage or partial life cycle tests, and
then to 7-day subchronic tests. Reviews
of subchronic, early life stage, partial life
cycle, and full life cycle toxicity tests with
several fish species demonstrated that the
shorter tests are good estimators of chronic
toxicity and MATCs observed in the longer
life cycle tests. Although the success of
developing briefer tests to estimate chronic
toxicity is empirically based, it does sup-
port the toxicological concept of time
course of effect in using acute data to
predict chronic lethality.
Another use of acute toxicity data to
estimate chronic toxicity is the toxicity
threshold value or LC1, which is calcu-
lated for 1.0% mortality and at one point
in time. This application of acute tests
should work well for those chemicals, ef-
fluents, and others that differ little in toxic-
ity between acute and chronic effects or
where the LC1 is derived at a duration
approaching or within chronic exposure
conditions. However, the LC1 does not
take into account time course of effect,
and its use for predictive purposes is lim-
ited for a wide range of chemicals; par-
ticularly those that bioconcentrate or have
cumulative effects.
Relation to Other Endpoints
Chronic toxicity tests commonly include
the measurement of long-term effects of a
contaminant on the survival, growth,, and
reproduction of a test organism. Survival
and growth are often equally sensitive,
and growth may not be of critical impor-
tance in establishing no-effect concentra-
tions in most tests.. In tests for which
growth is the single most sensitive end-
point, survival could be used to estimate
the no-effect concentration within a factor
of 3.
Growth-related endpoints are highly pre-
dictable from survival effects with fresh-
water fishes (Table 2). Length was less
variable than weight, and although all co-
efficients of determination (r2) exceeded
0.9, they were slightly higher for length
(0.941 to 0.949) than for weight (0.901 to
0.931). Also, no alteration was noted in
the intercepts (a) for length versus sur-
vival between 30 and 90 days of expo-
sure; the intercepts of weight versus sur-
vival varied, without trends, over time. Us-
ing these equations (Table 2), estimated
no-effect concentrations for growth may
be derived from the predicted values for
chronic lethality.
No-effect concentrations were always
less for reproduction endpoints than for
survival, but attempts to relate acute le-
thality to chronic reproductive effects by
regression analysis have not been suc-
cessful. Because of the likelihood of dif-
ferent modes of action between lethal and
reproductive effects, we do not recom-
mend that reproductive effects be pre-
dicted using the proposed method. How-
ever, the proposed technique is highly ben-
eficial in the preliminary assessment of
chronic toxicity of effluents and other
chemicals and in predicting chronic no-
effect concentrations for survival and
growth with fish species that are difficult
to culture under chronic testing conditions.
•U.S. Government Printing Office: 1992— 648-080/60054
-------
-------
-------
G.F. Krause, M.R. Ellersieck, and G. Lee are with University of Missouri, Agricultural
Experiment Station, Columbia, MO 65211.
Foster L. Mayer is the EPA Project Officer (see below).
The complete report consists of paper copy and diskette, entitled "Statistical Approach to
Predicting Chronic Toxicity of Chemicals to Fishes from Acute Toxicity Test Data":
Paper Copy (Order No. PB92-169655/AS; Cost: $26.00, subject to change)
Diskette (Order No. PB92-503119/AS; Cost: $90.00, subject to change)
(Cost of diskette includes paper copy)
The above items will be available only from:
National Technical Information Service
5285 Port Royal Road
Sprlngfleld,VA22l6l
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Environmental Research Laboratory
U.S. Environmental Protection Agency
Gulf Breeze, FL 32561
United States
Environmental Protection Agency
Center for Environmental Research Information
Cincinnati, OH 45268
Official Business
Penally for Private Use
$300
BULK RATE
POSTAGE & FEES PAID
EPA PERMIT NO. G-35
EPA/6QO/SR-92/091
------- |