Appendix E. Analysis of chronic toxicity data and acute chronic
ratios (ACRs) in support of deriving chronic HC5s:
Acetylcholinesterase inhibitors
David R. Mount
National Health and Environmental Research Laboratory
Mid-Continent Ecology Division
Duluth, MN
December 20, 201

-------
1.0  BACKGROUND

The analysis described in this appendix is one of several conducted in support of developing a
common methodology for assessing chemical effects on aquatic animals for use by the USEPA
Office of Pesticide Programs (OPP) and the Office of Water (O W). Other appendices describe
methods for estimating/predicting the acute toxicity values and/or the acute HC5, the 5th
percentile in a species sensitivity distribution for acute toxicity. However, effects assessments
by both  OPP and OW also require estimation of chronic toxicity thresholds.  This could include
predicting chronic thresholds for specific organisms,  or the estimation of the chronic HC5, the 5th
percentile of a chronic species sensitivity distribution.

A common approach for extrapolating from acute to chronic toxicity is the Acute-Chronic Ratio
(ACR),  which is calculated as the ratio of the acute effect concentration to a chronic effect
concentration, generally for the same chemical and organism.  Both OPP and OW currently use
ACRs in some contexts, generally as a "within chemical" extrapolation tool, meaning that an
ACR developed for a chemical tested with one or more species is used to estimate chronic effect
thresholds for other organisms exposed to the same chemical.  The underlying premise of the
ACR is  that the combination of toxicokinetics and toxicodynamics that determine the separation
of acute and chronic effect concentrations has some degree of commonality among organisms (or
sub-groups thereof) and can therefore serve as a basis for extrapolating chronic thresholds for
organisms having only acute toxicity data.

As an example, EPA's guidelines for deriving aquatic life criteria (ALC)(USEPA  1985; referred
to as the "1985 Guidelines") describe ACRs as one method for deriving the Final Chronic Value
(FCV), which is intended to represent a theoretical 5th percentile of chronic toxicity for a
distribution of aquatic species — essentially, the chronic HC5.  The method in the 1985
Guidelines requires that acute and chronic toxicity data are available to calculate ACRs for a
minimum of one fish, one invertebrate, and one acutely sensitive freshwater organism. A Final
Acute-Chronic Ratio (FACR) is then calculated as the geometric mean of the available ACRs, or
by other means as warranted by the data.  For example, if the ACRs tend to decrease as acute
sensitivity decreases, then the FACR may be calculated as the geometric mean of ACRs only for
acutely sensitivity species.  The Final Acute Value (FAV; the theoretical 5th percentile of the
acute species sensitivity distribution) is then divided by the FACR to derive the  FCV.

The FAV and FCV under the 1985 guidelines represent estimates of the acute HC5 and chronic
HC5, and therefore present a case example of using ACRs to estimate the chronic HC5. To
generalize this case, we can represent the approach as:

      chronic HC5 = (acute HC5) / ACRHcs

where ACRncs refers specifically to an ACR derived for the purpose of relating an acute HC5 to
a chronic HC5.  In the case of the 1985 guidelines, the FACR = ACRncs- As outlined above,
calculation of the FACR under the 1985 Guidelines requires the availability of paired acute and
chronic  toxicity data for a minimum group of organisms exposed to the same chemical.
However, cases  exist where OPP or OW may wish to estimate chronic effect thresholds for

-------
chemicals with limited or no chronic toxicity data available; in such cases, an appropriate ACR
must be developed from other data.

As part of the Great Lake Water Quality Initiative (GLI), a method was proposed to use
"default" values in the calculation of an ACR.HC5 (referred to in the methodology as a "Secondary
Acute-Chronic Ratio" or SACR). The basis for this default value was an analysis by Host et al.
(1995), who developed a distribution of FACR values from all ALC documents available at the
time.  The 80th percentile of this distribution (18), was then adopted under the GLI as a default
ACR value to be used in lieu of measured ACRs. Calculation of the SACR uses a combination
of available measured ACRs and the default value of 18.  If two measured ACRs are available,
the SACR is the geometric mean of 18 and the two measured ACRs; if one measured ACR is
available, the SACR is the geometric mean of 18, 18, and the measured ACR. If no chronic data
are available, then the SACR is simply 18.

The FACR distribution compiled by Host et al. (1995) combined FACRs from freshwater ALC
across all chemical types. TenBrook et al. (2010) took a similar approach, but used FACR data
only for pesticides (combining pesticides of different types).  The 80th percentile of that
distribution was 12.4, and TenBrook et al. (2010) recommended application of this value in a
manner parallel to the SACR calculation under the GLI.

In the ACRncs estimation methods described above, default ACRs were derived from
distributions of chemicals with multiple mechanisms of action. In the context of the current
effort, it was hypothesized that the ACRncs approach could be refined by analyzing ACRs in the
context of MOAs. The working assumption was that the distributional characteristics of ACRs
would be related to the MOA for a chemical, meaning that particular MOAs might lead to
consistently higher or lower ACRs compared to averages obtained across multiple MOAs.
MOA-specific ACRs would logically extend also to the possibility of particular taxa having
higher or lower ACRs because of the intersection of their physiology and the way in which the
chemical  elicits adverse effects.

This appendix analyzes ACR data for pesticides whose primary MOA is the inhibition of
acetylcholinesterase (AChE); this includes two primary pesticide groups, organophosphates and
carbamates. While focusing specifically on AChE-inhibiting chemicals, this analysis  is also
intended as demonstrative of analyses that might be used for developing MOA-specific ACRs
for other groups of chemicals in the future.

-------
2.0  DATA COMPILATION AND ANALYSIS

Chronic toxicity data were extracted from a database developed by USEPA called
"AquaChronTox". This database is a compilation of chronic toxicity data for aquatic animals
obtained from studies that would meet the definition of an acceptable chronic test under the 1985
Guidelines; in general terms, this means life-cycle tests (including reproduction) with
invertebrates, and life-cycle, partial life-cycle, or early life stage (ELS) tests with fish.  Sources
of data included a previous chronic toxicity data base  compiled by Glen Thursby (USEPA-
Narragansett); an existing database of chronic toxicity data compiled and largely generated by
the USEPA laboratory in Duluth; chronic toxicity data submitted to USEPA under F1FRA
pesticide registration requirements; and other data from the open literature  (compiled in the
ECOTOX data base and/or evaluated by USEPA as part of ecological risk assessments
supporting pesticide registration).  While the development of AquaChronTox is ongoing, priority
was given to entry of data specifically for AChE inhibitors to support the analysis described in
this appendix. Data coverage for AChE inhibitors within AquaChronTox is thought to be robust,
though not necessarily fully comprehensive.

The AquaChronTox database contains a wide variety  of extracted data, including both general
experimental parameters and treatment level results for individual endpoints related to survival,
growth, or reproduction. Also captured were statistically determined NOEC and LOEC values
(where available) from the study authors or later re-analysis (e.g., data evaluation records
completed as part of USEPA risk assessments).  Both freshwater and marine/estuarine organisms
were included in the data compilation.  Both measured (where available) and nominal exposure
concentrations are contained  in the database, but measured concentrations were used for data
analysis whenever available.

Initially, a total of 119 chronic studies were identified for organophosphate or carbamate
insecticides.  These data were then subject to additional screening wherein studies were
eliminated if there were test conditions that would disqualify a test from consideration using
evaluation criteria similar to that used in the development of water quality criteria (e.g., unusual
dilution water characteristics, use of surfactants in combination with toxicants, poor performance
of control  organisms). Where NOEC and LOEC values were reported for individual endpoints,  a
single NOEC and LOEC value  for the study was established based on the most sensitive
endpoint related to survival, growth, or reproduction (i.e., that endpoint showing significant
effects at the  lowest exposure concentration). As part of this review, exposure response curves
for those endpoints were evaluated qualitatively to insure that the designation of the NOEC and
LOEC values was not notably ambiguous and seemed biologically reasonable. Through this
review, a small number of tests were discarded because of a highly irregular exposure response
curves. While it is possible that such tests represent true response profiles,  they occurred in less
than two percent of tests, making it seem less likely that the "true" response curve is actually
non-monotonic. Another characteristic of a small number of tests was that the lowest
NOEC/LOEC was reported for organism length, but there was no indication of a concomitant
decrease in organism weight. In general, one expects organism weight to vary with the cube of
organism  length (assuming similar body morphology), so even very small changes in length
would be expected to cause substantial changes in weight. In cases where small differences in
length were defining the overall NOEC/LOEC without a corresponding indication of a response

-------
in weight (or reproduction), the NOEC/LOEC for the test were increased to match the
NOEC/LOEC for the next most sensitive endpoint. Finally there were a few tests for which the
spacing of exposure concentrations was large (e.g., greater than 3-fold concentration intervals).
These tests were discarded because of the associated uncertainties in their NOEC/LOEC values.

Those tests passing the previous filters were then coded into two groups according to whether the
data yielded  "defined" or "unbounded" NOEC or LOEC values.  In this context, the term
"unbounded" means that there were either significant effects at the lowest tested concentration,
or no effects at the highest tested concentration;  in such cases, the NOEC and LOEC can only be
expressed as inequalities. For most analyses, only those tests with defined NOEC and LOEC
values were used, because of the innate difficulty in conducting distributional analyses with a
mixture of defined and unbounded values. However, for some analyses unbounded values were
included as check to insure that limiting analyses to defined values did not skew the analysis
(e.g., if eliminating tests with effects noted at the lowest tested concentration had the effect of
biasing data  away from  tests indicating high chronic sensitivity).  From the original set of 119
chronic studies, a subset of 73 tests meeting the screening criteria and with defined NOEC and
LOEC values was established. For these tests, the geometric mean of the NOEC and LOEC was
also calculated, denoted as "GM" in this document. GM is the equivalent of the "Maximum
Acceptable Toxicant Concentration" (MATC), a term that appears frequently in the literature.
MATC is not used here  because its name implies a risk management judgement (what is
"acceptable") that is not intrinsic to the calculated  value itself.

In addition to NOEC, LOEC, and GM, regression analysis was used to estimate EC 10, EC20,
and EC50 concentrations for invertebrate chronic tests meeting all the previous qualification
criteria. The Toxicity Response Analysis Program (TRAP, Version  1.02; U.S. EPA, Mid-
Continent Ecology Division, Duluth, MM) was used to derive these values. A sigmoid model
with finite tails was used to fit the exposure response data for individual endpoints, except in a
very small number of cases where this model did not seem to reflect the shape of the underlying
data, in which case a piecewise linear model was used.  The exposure concentration was
expressed on a logic scale while the response variable was expressed on a linear scale. Because a
log exposure scale prevents the control concentration from being entered as 0, the control
exposure was assigned a concentration equal to one-tenth of the lowest exposure concentration.
The test endpoint with the lowest EC20 value was  identified from among the reported endpoints,
and the EC 10, EC20, and EC50 values for that endpoint were selected to represent that test. The
EC20 was used because of its precedent as the chronic effect concentration currently used in the
derivation of ALC.  However, because of differences in the slope of the concentration-response
curve, it was possible for an endpoint that did not have the lowest EC20 value to have the lowest
EC10 or EC50 value. In the infrequent cases where this occurred, the differences were small.

Calculating ACRs also requires an acute value for  the same species and chemical. Acute values
were acute LC50 (or immobilization EC50) values (48-h for cladocerans; 96-h for other species)
selected according to a sequence of priorities as follows:

    •   An LC50 provided by the study author as being paired with the chronic study
    •   An LC50 identified in the USEPA data evaluation record  as being paired with the chronic
       study

-------
   •   An LC50 from an acute test conducted by the same laboratory as conducted the chronic
       study, but potentially at a different point in time.
   •   An LC50 (or geometric mean of multiple LC50 values) derived from the data
       compilation underlying Web-ICE (Version 3.1; Raimondo et al.  2010). Values from the
       Web-ICE data were selected based on factors including:
          o  Must meet minimum data  requirements for inclusion in Web-ICE models
          o  Preference given to values with measured chemistry
          o  Preference given to values using flow-through methods
          o  Preference given to values generated using USEPA/OPPTS methods

In the context of developing ALC, ACRs are calculated as the ratio of the acute LC50 (or EC50)
to the chronic EC20 or GM. For ecological risk assessments for pesticide registration, ACRs are
calculated as the ratio of the acute LC50 (or EC50) to the chronic NOEC.  In this appendix,
ACR is used when the endpoint is not relevant, such as when speaking conceptually, or to a point
that applies to ACRs of all types. In cases where the ACR is specific to a particular  chronic
endpoint, subscripts are used to  indicate the chronic effect measure used to calculate the ACR.
For example, ACRoM refers to the ratio of the LC50 to the GM, and the  ACRNOEC is the ratio of
the LC50 to the NOEC. As defined in the introduction, an ACRncs is an ACR intended to relate
the acute HC5 to the chronic HC5.

In some cases, ACR values of less than one were calculated, meaning that the LC50 was actually
lower than the NOEC, GM or EC20. While it is counterintuitive that this could be the case, one
must remember that the acute value is generally not derived from the same study as the chronic
endpoints (because of differences in methodology), and since sources of variation exist for both
the acute and chronic tests that define the ACR, it is possible for ACRs  less than one to occur as
a result of experimental variation.  Another possible cause of ACRs less than one is where the
food added during chronic testing has an  effect on the bioavailability or toxicity of the chemical
causing it to be less toxic than under the acute test conditions (which generally preclude
feeding); this has been observed with daphnid tests and some metals. For this analysis, ACRs
less than 1 were retained at their calculated value for purposes of evaluating the distribution of
ACRs. In a previous analysis of ACRs, Raimondo et al. elected to exclude tests where ACRs
were less than 1. This was not done here because of concern that it would bias the ACR
distribution to higher values.  This is not  to suggest that ACRs less than one would be
appropriate as an ACRncs, only that they should be retained during evaluation of the distribution
of ACRs.

Much of the analysis that follows focuses on the ACRcM, though parallel calculations were done
using ACRNOEC and ACREC20 (for invertebrates) values. The conclusions reached using the
ACRNOEC are essentially the same as those obtained from the ACRcMJ this is because data were
only used from tests with similar and relatively close spacing of exposure concentrations. This
makes the ratio of the ACRcM to the ACRNOEC fairly consistent; for the  73 chronic tests
evaluated, the median ratio of the ACRGM to the ACRNoEC was  1.40, with 10th and 90th
percentiles of 1.28 and 1.55, and minimum and maximum of 1.11  and 2.00.  ACRec20 values
were only calculated for the 26 chronic tests conducted with invertebrates.  Within those data,
the median ratio of the ACRncio to the ACRNOEC was  1.41, the 10th and  90th  percentiles were

-------
1.07 and 3.11 and the minimum and maximum were 0.89 and 6.91. Comparing the ACREc2o to
the ACRcM gave a median ratio of 1.01, the 10th and 90th percentiles were 0.75 and 2.41, and the
minimum and maximum were 0.63 and 5.12.

Few tests of statistical significance were used to compare groups of ACRs.  This is because data
availability was highly irregular across species and chemicals, so data sets that compare, for
example, two taxa across the same set of chemicals, are generally limited. Because of the high
variability in ACRs, statistical comparisons of ACRs with low sample size lack power.
Accordingly, it was decided that direct inspection of the data was sufficient for most evaluations
presented.

-------
3.0  RESULTS AND DISCUSSION

Tables 1, 2, and 3 provide a summary of the ACRcM, ACRiMOEC, and ACREC2o (respectively) data
for specific combinations of species and chemicals.  Figure 1 shows the cumulative distribution
of ACRcM values across all taxa and AChE inhibitors, showing that overall ACRs varied widely,
spanning more that four orders of magnitude. Figure 2 separates the distributions for
invertebrates and fish and shows that while ACRs for both groups vary considerably, the
distribution of invertebrate ACRs was consistently lower, and somewhat less variable than that
for fish.

Figure 3, shows the same distribution as Figure 2, except that data are coded as to whether they
represent OP or carbamate chemicals.  While there are fewer data for carbamate chemicals, their
ACRs show a fairly wide dispersion across the overall distributions, giving little reason to
believe that combining data for OP and carbamate chemicals is inappropriate.
Figure 4, shows the same distribution with data coded as to whether the test species was a
freshwater or marine/estuarine organism.  For the fish data, the overlap of freshwater and
saltwater species is fairly broad. For invertebrates, however, 8 of the 9 saltwater ACRs lie at or
above the 50th percentile. Figure 5 shows the distributions of invertebrate ACRs by species; as
indicated in the figure, mysid shrimp (Americamysis bahid) is the only marine invertebrate
organism, and freshwater data are all for cladocerans, predominately Daphnia magna.  To
provide a more focused evaluation of the comparative sensitivity of mysids and D. magna, a
subset of the overall data was extracted containing data for only those chemicals that have data
for both species (Table 4).  The ratio of the mysid ACRcM to the daphnid ACRoM was calculated
for each chemical, then the logs of those ratios were evaluated with a t-test to determine if the
ratios were significantly different (greater) than 1 (which would indicate that mysids were
significantly more sensitive).  In 6 of 8 direct comparisons, the mysid ACRoM was higher than
the daphnid ACRoM, with an average ratio of 1.8. However, a one-tailed test (p=0.05)  of
whether this was significantly different from 1 was close to, but not quite, significant (p=0.057).

Some evidence that runs counter to a presumption that mysids are more sensitive than daphnids
lies in three mysid chronic tests reported by Thursby et al (1990a,b; 1991). These three tests
(carbaryl, dichlorvos, and propoxur) were excluded because of reduced control survival;
however, analysis of these data suggest fairly low ACRs (ACRoM = 0.85, 1.21, 5.29,
respectively).  These values would fall in the low end of the overall ACRcM distribution for
invertebrates, though they cannot be directly compared because ACRcM  values for D. magna are
not available for these same chemicals.

While fish showed high ACRs for AChE inhibitors, this does not translate to high chronic
sensitivity offish relative to invertebrates.  Because the ACRcM is a ratio of LC50/GM, high
ACRs can be produced not only by having low chronic effect concentrations, but also by having
very high LC50 values. Relative to invertebrates, the high ACRs for fish exposed to AChE
inhibitors appears to be driven largely by a very low acute sensitivity (high LC50 values), rather
than by having high chronic sensitivity. This low level of acute sensitivity is affirmed by SSD
analyses for AChE inhibitors  conducted by Erickson (see Appendix D).

-------
To help put the chronic sensitivity offish in perspective, the NOECs from chronic fish tests were
plotted against the acute LC50 for D. magna for the same chemical (Figure 6).  In addition to
tests with defined NOEC values, this plot also includes tests with unbounded NOECs, and values
are coded with respect to whether they were derived from ELS tests or life-cycle (or partial life-
cycle) tests.  As points of reference, three lines were drawn representing:

   •  the D. magna LC50
   •  an estimate of a median acute HC5, estimated by dividing the D. magna LC50 by 2.89,
       the median acute AChE inhibitor EF derived by Erickson (Appendix D) for data sets
       including D. magna, rainbow trout, and bluegill sunfish;
   •  an estimate of the 90th percentile HC5, estimated  by dividing the D. magna LC50 by 16,
       the 90th percentile acute AChE inhibitor EF derived by Erickson (Appendix  D) for data
       sets including D. magna,  rainbow trout, and bluegill sunfish.

Several inferences may be drawn for Figure 6.  First, the vast majority offish NOEC values are
at or above the D. magna LC50, meaning that in most cases, the  chronic sensitivity offish to
AChE inhibitors is less than the acute sensitivity of D. magna to the same chemicals.  However
even in cases where the fish NOEC was lower (more sensitive) than the D. magna LC50, almost
all were at or above the estimates of acute HC5 values based on EFs calculated by Erickson
(Appendix D). The conclusion here is that the ACR for fish may not be very important for
estimating the chronic HC5 for AChE inhibitors, because of the greater sensitivity of
invertebrates.

Another aspect of Figure 6 is the comparison of the general sensitivity of chronic tests with fish
tested in life-cycle (including reproduction) versus early  life stage (ELS) exposures.  Based on
inspection, there is some tendency for life-cycle tests to be toward the lower end of the scatter,
suggesting they may be more  sensitive indicators of chronic toxicity to fish. Using a different
method of comparison, Raimondo et al., (2007) found that ACRs from fish life-cycle tests were
significantly higher than those for ELS tests.  However, this tendency is not dramatic, as the
distribution of ELS values encompasses the range of life-cycle values.  In regard to estimating a
chronic HC5, the same conclusion is reached from both tests types, that fish are comparatively
insensitive to AChE inhibitors.

As outlined in the introduction, the 1985 Guidelines, calculation of the FACR requires ACRs
(ACRcM or ACREC2o) for at least three species, encompassing at least one fish species, at least
one invertebrate species, and at least one acutely sensitive freshwater species.  Depending on the
variability and distribution of the ACRcM values, the FACR is determined in different ways. If
the ACRcM values are smaller for organisms with high acute sensitivity (and vice versa), then the
FACR is determined by the ACRoM for species whose acute sensitivity is near the FAV, rather
than by the mean of all ACRs. The rationale for this approach is that if organisms with high
acute sensitivity have smaller ACRs, then applying larger ACRs associated with acutely
insensitive species would result in a chronic HC5 that was too low.  For AChE  inhibitors, ACR
values are related to acute sensitivity, with invertebrates  having both higher acute sensitivity and
lower ACRs. Therefore, using the same logic, it seems appropriate to base the  ACRncs  for
AChE inhibitors on invertebrate ACRs.

-------
As shown in Figure 5, the invertebrate ACRs for AChE inhibitors are dominated by two species,
D. magna and mysids.  If the ACR.HC5 is based on an invertebrate ACR, one must decide whether
to segregate ACRs by species (e.g., use D. magna ACRs for freshwater ACRncs and mysids for
marine ACRncs), or to use the joint distribution of all invertebrates. This decision hinges in part
on the previously discussed issue of whether mysids do in fact have larger ACRs than do
daphnids and, even if they do, might there be acutely sensitive freshwater organisms that have
ACRs more similar to mysids than daphnids. Neither of these questions appear easily answered
from the data in hand. Again drawing from the 1985 Guidelines, that methodology requires a
minimum of three ACRs to include at least one fish species, one invertebrate species, and one
acutely sensitive freshwater species.  Given the species for which there are chronic data for
AChE chemicals, the two ways these requirements would be met would be either: 1) D. magna,
and two fish species, or 2) D. magna, mysids, and a fish. In the former case, the D. magna
ACRoM would be the FACR; in the latter case, the FACR would likely be the mean of the ACR
values for D. magna and mysids, since both are acutely sensitive species.

Given that:

    1)  it is unclear whether mysids actually have higher ACR values than D. magna;
    2)  even if they do, they may represent ACRs for other relevant invertebrates; and
    3)   a FACR under the 1985 Guidelines would be calculated as the geometric mean of the
       ACRs for the two  species

 it seems  reasonable to base an ACRncs on an ACR selected from the joint distribution of ACRs
for all invertebrate species. Figure 7 shows the distribution of invertebrate ACREC20> ACRcM,
and ACRNOEC combined across all AChE inhibitors and freshwater/saltwater species. Figure 8
directly compares EC20 and NOEC values, while Figure 9 compares EC20 and GM values. In
all but one case, the NOEC was lower than the EC20, indicating that the level of effect at the
NOEC was generally lower than 20%. The EC20 values showed greater similarity to the GM
concentrations, with only a few GM values falling outside a range of 0.7 to 1.5 of the EC20.

Table 5 provides the 50th, 80th, and 90th percentile values for each of the ACR calculation
methods along with related values from other ACR  analyses. The Host et al. (1995) and
TenBrook et al. (2010) analyses were discussed in the introduction. The Raimondo et al. (2007)
was a compilation of 456 ACRcM values from a broad range of chemicals, the distributions of
which were analyzed by a  variety of sub-groupings. Interestingly, Raimondo et al. (2007) did
not find a significant difference between ACRs for AChE inhibitors compared to other MOA.
However, their comparisons across MOA pooled fish and invertebrate ACRs within a MOA, so
the resulting distribution would have been very large (similar to Figure 1),  and would not have
reflected  differences among taxa within an MOA, as has been described here for AChE
inhibitors. Likewise, the Raimondo et al. comparison offish and invertebrate ACRs did not
show the pronounced differences shown in the current analysis, presumably because the
Raimondo et al. comparison offish v. invertebrates included all chemicals, and the large
differences shown for AChE inhibitors likely doesn't exist for all chemical groups. While
median ACRoM values are similar between Raimondo et al. and the current analysis, limiting
selection of the ACRncs for AChE inhibitors to invertebrates greatly reduces the ACRs in the
upper ends of the distributions.

-------
Another difference is that Raimondo et al. discarded all ACRcM values that were less than 1
whereas these values were retained in the current analysis.  As explained earlier, this has the
potential to bias the ACR distribution toward higher values; the same sources of variability that
create artificially low ACRs presumably create artificially high ACRs also, but these are not
distinguishable from other values and are therefore left in the analysis. However, the bias
created by data censoring is probably not high, because Raimondo et al. used similar data
sources to the current analysis, and the total number of ACRs for AChE inhibitors reported by
Raimondo et al. (78) is actually slightly higher than the number included in this analysis (73). A
preliminary comparison of the invertebrate ACRs for AChE inhibitors from both analyses (data
not shown) showed some differences in the low end of the distribution (which is expected
because that is where values were excluded), but close agreement in the 50th to 90th percentiles.

The Host et al. (1995) and TenBrook et al. (2010) analyses differ from Raimondo et al. (2007) in
that they are based on distributions of FACR values from ALC documents. Although these
FACRs are based on ACRoM (or ACRecio) values,  they are not directly comparable to raw
ACRcM values, because FACRs reflect a variety of different derivations - for example, some
may represent the ACRcM for a single acutely sensitive species, while others may be the average
of multiple ACRcM values. That difference aside, because the current analysis suggests that
invertebrate ACR values represent a logical basis for an ACRHcs for AChE inhibitors, the
invertebrate ACRcM and/or ACREC20 values from the current analysis are functionally parallel to
the Host et al. FACR distribution. Invertebrate ACRoM and ACRnczo distributions from this
analysis range from roughly equal to, to about 2-fold lower than, comparable values from the
Host et al. (1995) and TenBrook et al. (2010) distributions.

When making this comparison, it is important to note that the degree of similarity between the
Host et al. (1995) and TenBrook et al. (2010) FACR distributions and the proposed ACRHcs
values for AChE inhibitors is created in part by the previous finding that fish ACRs were not
important, because the invertebrates have greater chronic sensitivity. If the fish were more
chronically sensitive to AChE inhibitors, then the conclusions would likely change.  As such, it
is highly uncertain whether the same conclusion would be reached for another mode of action
with different taxonomic sensitivity.

Because invertebrate ACRs for AChE chemicals show a substantial range, estimation of a
chronic HC5 using an ACRncs will require a risk management decision as to where in the
distribution the ACRncs will be selected. One important consideration is that because ACRs are
ratios and both the numerator and denominator have uncertainty, variability in the ACRs from
chance alone will be larger than the variability in either of the component values.  In a Monte
Carlo simulation, a theoretical distribution of ACRoM values was created by assuming true
values of5.91 for the LC50 and 1 for the GM-this creates a "true" ACR of 5.91, which was the
50th percentile for the invertebrate ACRcM for AChE inhibitors. Both the LC50 and GM were
assumed to vary within 2-fold, meaning that 95% of values would fall between 0.5x and 2x of
the true value, with a log-normal distribution. This factor of 2 was based on a common "rule of
thumb" regarding variability in toxicity test results; it is simply illustrative and is not based on
any specific evaluation of test variability. This simulation gave 50th, 80th, and  90th percentile
        of 5.91  (the true mean), 11.0, and 15.6; the corresponding ACRcM values from the

-------
current analysis were 5.91, 13.6 and 18.8 (Table 5). Thus,  if the true variability in LC50 and
GM values were on the order of 2-fold, the majority of the observed range in ACRcM values
could be attributable to random variability alone, even if the "true" ACR were the same for all
AChE inhibitors.

Less easily evaluated is the possibility that sources of error beyond random variability exist,
causing extreme values.  For example, in the current analysis the highest invertebrate ACRcM is
255 for D. magna and disulfoton. The reported measured LC50 used to calculate this ACR is 13
ug/L. However, this value seems inconsistent with the much greater mortality response observed
in the chronic test (37.5% survival at 0.64 ug/L), and is also inconsistent with a much lower (less
than 1 ug/L) LC50 predicted by the ECOSAR system. While not certain, these latter
observations suggest that the measured LC50 and resulting ACR may be too high.  Although the
calculated value of 255 was retained in the analysis, the  potential for aberrantly high or low
values emphasizes the potential inaccuracies that might be introduced by selecting values that are
in the extreme tails of the distribution.

Another consideration in selecting an ACR from the distribution is that a chronic HC5 derived
by dividing the acute  HC5 by the ACRncs contains the uncertainties from both the acute HC5
estimation and the ACR distribution. If the acute HC5 derivation contains substantial
conservatism (e.g., an EF selected from a high percentile of the distribution) and the same is
done for selecting an  ACRncs, the estimated chronic HC5 may have an undesirably high
aggregate conservatism.

Based on the experience  of developing this analysis, suggestions for future study include the
following:

   1.  Determine if additional screening of data to more carefully "match"  (e.g., same lab, same
       water, same time frame) the acute and chronic data would decrease variability in ACRs.
   2.  Consider additional data collection to evaluate specific toxicity values that are
       contributing to extreme values within ACR distributions.
   3.  Conduct similar analyses on chemicals in other MOAs to explore more fully whether
       consideration  of MOA in developing ACRHcss provides a substantive decrease in
       uncertainty in deriving chronic HC5s as compared to more generalized ACRHC5
       distributions.
   4.  Consider approaches that could be used to identify chemicals or chemical groups for
       which ACROSS are likely to exceed default values (e.g., "structural alerts" similar to
       Ahlers et al. 2006), and thereby result in overestimating the chronic  HC5.

-------
4.0 CONCLUSIONS

    1)  When compared across their distributions, ACRs for AChE inhibitors are much higher
       for fish than for invertebrates.
    2)  Despite having higher ACRs, fish generally have low acute sensitivity to AChE
       inhibitors, such that their chronic effect concentrations are typically higher than the acute
       effect concentrations for D. magna exposed to the same chemical, and almost certainly
       higher than an acute HCs for vertebrates and invertebrates combined; for this reason,
       estimation offish ACR values is not important to estimating a chronic HC5 for AChE
       inhibitors for all species combined.
    3)  If the chronic HC5 for AChE inhibitors is estimated by applying an ACRncs to the acute
       HC5, it is logical to select the ACRncs from the distribution of invertebrate ACR values;
       doing so should produce a chronic HC5 for AChE inhibitors that is protective of both fish
       and invertebrate species.
    4)  The AChE inhibitor-specific distribution of invertebrate ACRs derived here is  generally
       lower than those from literature analyses using datasets less specific to the AChE
       inhibition MOA. Accordingly, the same intended level of protection may be achieved
       with a higher chronic HC5 if the MOA-specific ACRpcs is used.

-------
References

Ahlers, J., C. Riedhammer, M. Vogliano, R. Ebert, R. Kuhne, and G. Schuurman. 2006, Acute to
chronic ratios in aquatic toxicity ~ variation across trophic levels and relationship with chemical
structure. Environ. Toxicol. Chem. 25:2937-2945.

Host, G.E., R.R. Regal, and C.E. Stephan. 1995. Analyses of acute and chronic data for aquatic
life. Draft report. United States Environmental Protection Agency. Washington, D.C. March 16,
1995.

Raimondo, S., B.J. Montague and M.G. Barron. 2007. Determinants of variability in acute to
chronic ratios for aquatic invertebrates and fish. Environ. Toxicol. Chem. 26:2019-2023.

TenBrook, P.L., A.J. Palumbo, and R.S. Tjeerdema. 2009. Methodology for derivation of
pesticide water quality criteria for the protection of aquatic life. Phase II:  Methodology
development and derivation of chlorpyrifos criteria. University of California Davis, Davis, CA,
USA.

Thursby, G.B. 1990a. Memo to D.J. Hansen: Flow-through acute and chronic toxicity of
dichlorvos to Mysidopsis bahia. USEPA, Office of Research and Development, Narragansett,
RI.

Thursby, G.B. 1990b. Memo to D.J. Hansen: Flow-through acute and chronic toxicity of
propoxur to Mysidopsis bahia. USEPA, Office of Research and Development, Narragansett, RI.

Thursby, G.B. 1991. Memo to D.J. Hansen: Flow-through acute and chronic toxicity of
carbaryl to Mysidopsis bahia. USEPA, Office of Research and Development, Narragansett, RI.

USEPA.  1985. Guidelines for Deriving Numerical National Water Quality Criteria  for the
Protection of Aquatic Organisms and their Uses. United States Environmental Protection
Agency. Stephan, C.E., D.I. Mount, D.J. Hansen, J.H. Gentile, G.A. Chapman and W.A. Brungs.
PB85-227049. National Technical Information Service, Springfield, VA.

USEPA. 2010. Exploration of Methods for Characterizing Effects of Chemical Stressors to
Aquatic Animals. Draft Report. United States Environmental Protection Agency. November 2,
2010.

-------
Figure 1:  Distribution of ACRoM values, combined across all AChE inhibitors and all species.
  C
  
 Q_

  0)

 '•M
 J2
  3

  E
  2
 O
99.8


 99
 98

 95

 90

 80

 70


 50 -


 30

 20

 10

  5

  2 -
  1 -
 0.5 -

 0.2 -

   0
          1    0.3    1
10   30
100  300   1000  3000 10000
                   Acute Chronic Ratio (LC50/GM)

-------
Figure 2:  Distribution offish and invertebrate ACRcM values for all ACHE inhibitors.
     99.8

      99
      98
      95
      90
      80
      70
 c
 0)
 o
 0)
a.
 0)
5   50 H
 re
"5   so
 E   20
O   10
      5
      2
      1
     0.5
                                                   A Invertebrates
                                                   T Fish
        0.1    0.3     1     3     10    30    100   300   1000  3000  10000
                   Acute Chronic Ratio (LC50/GM)

-------
Figure 3:  Distribution offish and invertebrate ACRoM values for all ACHE inhibitors, coded to
indicate data for carbamate and organophosphate chemicals.
     99.8

      99
      98
 -   95

 O   90
 QJ   80
 0-   70
 0)
 £   50
 3
 E
 3
 o
30
20

10
 5
 2 -
 1 -
0.5
        0.1
A  Invert Carbamate
A  Invert OP
V  Fish Carbamate
V  Fish OP
               1           10         100        1000

             Acute Chronic Ratio (LC50/GM)
                 10000

-------
Figure 4: Distribution offish and invertebrate ACRcM values for all ACHE inhibitors, coded to
indicate data for freshwater and saltwater species.
     99.8
  c
  0)
  u
  0)
 0-
  0>
 .5
 3
 E
 3
 o
99 -
98 -
95
90

80
70
 £   50-
30
20

10
 5

 2 -
 1 -
0.5
                      SW Invertebrates
                      FW Invertebrates
                      SW Fish
                      FW Fish
         0.1
        0.3
10
30
100   300    1000  3000  10000
                    Acute Chronic Ratio (LC50/GM)

-------
Figure 5: Distributions of ACRFc2o values for individual invertebrate species, combined across
all ACHE inhibitors.


^_/
c
0)
0
1_
0)
Q.

o



E
0


so -
95 -
90 -
80
70 -



50

30
20 -
10 -
5 -
•j
V
A
T
T A
T
T A
T A
T A

•^ A
A
7 A
T
YA
A Americamysis bahia
^ Daphnia magna
• Ceriodaphnia dubia

       0.1  0.2   0.5  1   2     5   10   20    50  100200  5001000

                Acute Chronic Ratio (LC50/EC20)

-------
Figure 6: NOEC values from fish chronic tests with AChE inhibitors plotted as a function of the
Daphnia magna acute LC50 and estimates of acute HC5s for the same chemical.  See text for
detailed explanation of the lines.
 ro
 3
 *->
 (A
 .2
 o
 'E
 o
 .c
 o
 £.
 V)
 L
 E
 o
 *«-
 u
 UJ
 O
     1000 -
100 -
              Fish chronic
              less sensitive
                                                Fish chronic
                                                more sensitive
         0.1
                        1                  10
                 Daphnia magna 48 -h LC50 (ug/L)
100
                              V   ELS Fish NOEC
                              V   ELS Unbounded NOEC
                              A   Life Cycle Fish NOEC
                              A   Life Cycle Unbounded NOEC
                             	 NOEC = LC50
                             	NOEC = Median acute HC5
                                   NOEC = 90th% acute HC5

-------
Figure 7:  Distribution of ACRNQEC- ACRoM- and ACREC20 values for invertebrates, combined
across all  ACHE inhibitors.
 c
 0)
 o
 0)
 Q.
 0)
 3

 3
 O
99
98 -

95 -
90 -

80
70 -

50 -

30 -
20

10 -
 5

 2
 1
             A
ACRchv
ACREC2o
       0.1  0.2    0.5   1
                    10   20
                                            50  100  200   500
                                  ACR

-------
Figure 8: Comparison of EC20 and NOEC values for invertebrate species combined across all
AChE inhibitors. Solid line designates unity, dashed line shows 2x deviation from unity, and
dotted line shows 4x deviation from unity.
      100
 CD
 3
       10 -
        1 -
 o
 O    0.1 1
 LU
     0.01 -
    0.001
 ,'/''    \
*/''        \/_
        0.001
     0.01
0.1
10
100
                              NOEC (ug/L)

-------
Figure 9: Comparison of EC20 and GM values for invertebrate species combined across all
AChE inhibitors. Solid line designates unity, dashed line shows 2x deviation from unity, and
dotted line shows 4x deviation from unity.
      100
       10 -
 0)
 o
 3    0.1
 LU
     0.01 -
    0.001
      x
-------

Aminocarb
Carbaryl (Sevin)
Carbofuran
Methomyl
Propoxur (Baygon)
Azinphos-methyl
Carbophenothion
Chlorpyrifos (Dursban)
Diazinon
Dichlorvos(DDVP)
Disulfoton
Ethoprop
Fensulfothion
Fenthion
Fonofos
Malathion
Methidathion
Methyl parathion
Naled
Parathion
Phosmet
Profunofos
Terbufos
Geomean Carbamate
Geomean OP
Geomean All
Americamysis
bahia



5.55



1.75
13.51


956


6.50

19.11

13.89
29.89

8.65

5.55
10.02
9.38
Ceriodaphnia
d ubia

















2.59






2.59
2.59
Daphnia
magna


1.65
16.14



3.11
4.12

255.44
1.94


4.49
6.71
8.15
0.41
2.26

5.01
6.27

5.15
4.97
4.99
Cyprinodon
variegatus



3.25

9.43

42.25
250.63
5.53
43.32
12.02


4.23
5.50







3.25
16.91
14.08
Menidia
beryllina







3.61
















3.61
3.61
Menidia
peninsulae







L38
















1.38
1.38
Jordanella
floridae















36.05








36.05
36.05
Oncorhynchus
mykiss





14.08



13.87
9.87


44.52
7.48
3.13




52.06



13.98
13.98
Opsanus
beta







2988
















29.88
29.88
Pimephales
promelas
31.96
23.82
8.61
25.84
9.12

28.05
151.50
793.56
21.37
28.93

6.44




15.62
324.37



28.38
17.29
49.94
34.19
Geomean
Invert


1.65
9.46



2.33
7.46

255.44
4.33


5.40
6.71
12.48
1.03
5.60
29.89
5.01
7.36

5.28
6.37
6.21
Geomean
Fish
31.96
23.82
8.61
9.16
9.12
11.52
28.05
15.71
445.97
11.79
23.13
12.02
6.44
44.52
5.63
8.53

15.62
324.37

52.06

28.38
13.09
20.72
19.11
Geomean
all
31.96
23.82
3.77
9.31
'-) u
11.52
28.05
9.11
57.67
11.79
42.16
6.09
6.44
44.52
5.51
8.03
12.48
2.55
21.67
29.89
16.15


9.67
12.67
10.52
o   H
-+,  v?

3   S
n>   >
<   n
n
t/i
     ££_

     B
     n
     •s.
       .
     r.

     CL
     gj


       -

-------

Aminocarb
Carbaryl (Sevin)
Carbofuran
Methomyl
Propoxur (Baygon)
Azinphos-methyl
Carbophenothion
Chlorpyrifos (Dursban)
Oiazinon
Oichlorvos(DDVP)
Disulfoton
Ethoprop
Fensulfothion
Fenthion
Fonofos
Malathion
Methidathion
Methyl parathion
Naled
Parathion
Phosmet
Profenofos
Terbufos
Geomean Carbamate
Geomean OP
Geomean All
Americamysis
bahia



7.90



3.50
18.26


13.41


9.39

31.82

17.81
38.71

10.91

7.90
14.47
13.53
Ceriodaphnia
dubia

















3.05






3.05
3.05
Daphnia
magna


2.51
19.29



4.40
5.65

351.35
2.70


6.45
s (,;
11.03
0.51
3.06

7 IX
X if>

6.95
6.75
6.78
Cyprinodon
variegatus



4.46

13.50

60.11
341.86
7.66
61.73
16.41


5.93
8.25







4.46
23.86
19.81
Menidia
beryllina







5.60
















5.60
5.60
Menidia
peninsulae







1.98
















1.98
1.98
Jordanella
floridae















40.58








40.58
40.58
Oncorhynchus
mykiss





17.93



19.42
13.64


64.15
10.64
4.52




71.88



19.45
19.45
Opsanus
beta







48.57
















48.57
48.57
Pimephales
promelas
50.13
4286
12.49
37.02
13.23

37.86
212.69
1095.70
28.09
44.08

8.94




17.29
478.26



38.19
26.53
6794
48.56
Geomean
Invert


2.51
12.35



3.92
10.15

351.35
6.02


7.78
8.67
18.73
1.25
7.38
i8 7 1
7.18
9.37

7.26
8.80
8.58
Geomean
Fish
50.13
42.86
12.49
12.85
13.23
15.56
37.86
23.32
612.03
16.10
33.35
16.41
8.94
64.15
, /.'M
11.48

17.29
478.26

71.88

« 1')
19.7]
28.83
26.96
Geomean
all
50.13
42.86
5.59
12.60
13.23
15.56
37.86
14.01
78.84
If. 10
HI ()')
8.41
894
tvl IS
7.86
10.70
18.73
3.00
;<> 65
38.71
22.72


14.13
17.58
14.69
  ,
-*>  05

3   Z
c   n
<   n

SL  73

n  o
vi  m
•    o
     n
     •s.


     BO
     n
     6.
     n


     g
     tfl

     i/T

-------
Table 3:  ACRecio values used in the analysis.  Highlighted values represent the geometric mean
of multiple values.

Aminocarb
Carbaryl (Sevin)
Carbofuran
Methomyl
Propoxur (Baygon)
Azinphos-methyl
Carbophenothion
Chlorpyrifos
(Dursban)
Diazinon
Dichlorvos (DDVP)
Disulfoton
Ethoprop
Fensulfothion
Fenthion
Fonofos
Malathion
Methidathion
Methyl parathion
Naled
Parathion
Phosmet
Profenofos
Terbufos
Geomean Carbamate
Geomean OP
Geomean All
Americamysis
bahia



8.84



1.75
2.64


6.13


7.33

26.71

6.73
18.92

8.42

8.84
7.03
7.21
Ceriodaphnia
dubia

















2.75






2.75
2.75
Daphnia
magna


2.04
3.67



3.84
4.97

121.12
1.56


2.75
2.79
9.19
0.35
2.38

3.92
6.45

2.73
4.09
3.84
Geomean


2.04
5.69



2.59
3.62

121.12
3.09


4.49
2.79
15.66
0.98
4.00
18.92
3.92
7.37

4.04
4.98
4.84

-------
Table 4: ACRoM values for AChE inhibitors with defined values for both Daphnia magna and
mysids.
Chemical
Chlorpyrifos (Dursban)
Diazinon
Ethoprop
Fonofos
Methidathion
Methomyl
Naled
Profenofos
Mysid
ACR
1.75
2.64
6.13
7.33
26.71
8.84
6.73
8.42
D. magna
ACR
3.84
4.97
1.56
2.75
9.19
3.67
2.38
6.45
Mysid/magna
0.456
0.532
3.941
2.664
2.907
2.411
2.832
1.306
Log
mysid/magna
-0.341
-0.274
0.596
0.426
0.463
0.382
0.452
0.116
mean = 0.228
l-tailed t p=0.05 df=7 1.895 SD= 0.357
2-tailed t p=0.05 df=7 2.365 t= 1.803
 Not significantly different under l-tailed test at p=0.05
 Not significantly different under 2-tailed test at p=0.05

-------
Table 5:  Distributional values for ACRs from the current analysis and literature sources.
ACR Distribution Source/Type
Present Analysis:
ACHE inhibitors, fish only ACRiMOEC
AChE inhibitors, fish only ACRoM
AChE inhibitors, invertebrates only ACRNOEC
AChE inhibitors, invertebrates only ACRoM
AChE inhibitors, invertebrates only ACREC20
Raimondo et al (2007):
Combined chemicals, all species ACRcM
AChE inhibitors, all species ACRoM
Combined chemicals, fish only ACRcM
Combined chemicals, invertebrates only ACRcM
Host ef a/. (1995):
Combined chemicals, freshwater FACR
Combined chemicals, saltwater FACR
TenBrook ef a/. (2010)
Combined chemicals, freshwater FACR
Percentile
50th

42.9
29.2
7.98
5.91
3.88

8.3
6.4
9.3
7.5

7.1
4.4

NR
80th

118
79.0
17.9
13.6
8.63

NR
NR
NR
NR

17.9
13.5

12.4
90th

334
247
30.6
18.8
18.0

79.5
60.2
90.0
68.3

29.1
24.2

NR
NR = Not Reported

-------