ENVIRONMENTAL EFFECTS BRANCH (7403)

    HEALTH AND ENVIRONMENT REVIEW DIVISION

                FEBRUARY, 1984
ESTIMATING "CONCERN LEVELS" FOR CONCENTRATIONS
  OF  CHEMICAL SUBSTANCES  IN THE  ENVIRONMENT
  OFFICE OF  POLLUTION  PREVENTION  AND TOXICS
     U.S.  ENVIRONMENTAL PROTECTION AGENCY
          WASHINGTON,  DC  20460-0001

-------
           The Environmental Effects Branch Position on
        Estimating "Concern Levels"  for Concentrations of
          Chemical Substances in the Aquatic Environment


Background and Purpose

    In evaluating PMNs, the Environmental Effects Branch attempts
to estimate levels of a given compound which, if met or exceeded
in the environment, could cause adverse effects.  The data set
from which such concentrations are estimated may run the gamut
from fairly complete packages, containing both acute and chronic
test data, to only the structure of the chemical itself.

    The purpose of this position document is to describe the
procedure and rationale followed to estimate environmental
concentrations of concern for a particular chemical.  The
methodology described should identify (at least 95% of the time)
concentration concern levels for TSCA chemicals which may cause
adverse environmental effects.

    This procedure is based upon generally accepted scientific
rationale and currently available data.   The attached Support
Document provides a detailed discussion of these data.  As
additional data become available and are analyzed,  any revisions
necessary to maintain this procedure at the state-of-the-art will
be made.

-------
Definitions,  Rationale,  and  Assumptions  Employed

    The  number  used  to  adjust  effective  concentrations
(laboratory LC50, MATCs, etc.)  to  arrive at environmental
concentrations  of concern  is defined  as  an assessment factor.
Before proceeding further, we  will attempt' to describe what a
concern  level for concentrations of a specific chemical would be
in the natural  environment.

    An environmental concentration of concern is that
concentration of a chemical  at which  populations of organisms are
adversely affected as found  in a field study conducted under
simulated or actual conditions of production, use, and disposal
(PUD) of a particular TSCA chemical.  Thus, a field study would
measure what would be expected to occur normally in the natural
environment.  Changes in PUD or site  specificity could alter
exposure concentrations but  should not significantly change the
actual concentrations at which effects occur.

-------
    The  assessment  factors  described in this paper take into
                  •

account  the  uncertainties due  to such variables  as test species


sensitivity  with  regard  to  acute and chronic toxicity,  laboratory


test conditions,  and  age-group susceptibility.   For the purposes


of ease  and  simplicity,  each of the  assessment factors  used  in


this.document has been rounded off to the  nearest  whole number.





    The  rationale for using assessment factors is  similar  to  that


of any other government  agency charged with  protecting  the


environment or public health.   Simply stated, the  more  the
                                             *
toxicological endpoints  of given  chemicals are characterized, the


less there is a need  for an assessment factor in estimating the


concentration of chemical which would  be expected  to cause


adverse environmental effects  in  nature.  A  chemical with a no-


effect level established by both  acute and chronic  testing would


require a relatively  smaller assessment factor to  establish the


concern level.   But a chemical with a  more scanty data base


(perhaps only a QSAR estimate  of  acute toxicity)  would require a


relatively larger assessment factor to establish  the concern


level.

-------
    The diagram below shows how each assessment factor decreases
as the data improves.  Note that the assessment factor for each
given data set relates only back to natural environment
conditions.  The assessment factors are not intended to relate
among each data set, even though the magnitude of the assessment
factors are partially derived from relationships between the
different types of data.
                             One Acute Concentration or QSAR
           Natural
         Environment
                             Lowest Effect Based On Many
                             Acute Concentration
                                      Lowest Chronic Effect
                                       Concentration

-------
     The following assumptions are used in selecting the most
 appropriate assessment factor to be employed in establishing the
 concentration of concern.

     1.     The actual test data or QSAR employed in applying
 assessment factors will be valid.  In the absence of suitable
 information or data, assessment factors should not be used.

     2.     The assessment factor is used to predict a
 concentration of a given chemical which would be expected  to
 cause adverse effects  on a population,  rather than individual
 organisms.

     3.     For purpose  of applying these assessment factors,  and
 only in  the  absence  of  data on  the chemical  under  review,  data on
 an appropriate analog  chemical  will  be  treated  as  if  it were  data
 on the chemical  itself.   For  example,  if  there  are  no ecotoxicity
 data on  a  PMN chemical  but there  are data on  an analog  (single or
 multiple LC50, MATC), the  assessment  factor appropriate to  the
 kind of  (analog) data available would be  applied.

    Finally,  it  should be  emphasized that  the Assessment Factors,
 as described  here  and in  the  attached Support Document, are not
 the same as  "S«f«ty Factors"  <«lso c«ll«d  "Margin of S«f«ty").
 Traditionally, safety factor means a factor (usually 100 or 1000)
 applied to a no-observed-effect-level (NOEL) for an adverse
 (human health or environmental) effect  to give  a figure below
which exposures are presumed to be "safe."  Additionally,  when

-------
safety factors are used,  there  is  usually  a good  set of  hazard
data available.  BEB's assessment  factor,  however, means a  factor
applied to an MATC or to  effect  levels  (e.g., an  LC-50)  to  give a
figure which if met or exceeded  by exposures could cause adverse
effects.  Also, assessment  factors are  used when  there is only
limited hazard data available.   Note also  that use of assessment
factors does not identify a "safe" exposure level.  EEB  cannot
make any statement about  the safety or  risk or exposures which
fall below concern levels derived  from  use of assessment
factors.  Assessment factors are intended  to be used in  the PMN
process to identify those chemicals which  should  be tested  under
TSCA Section 5 to more fully characterize  their ecotoxicity
hazards.

The Assessment Factors
    One of three assessment factors is used in estimating the
chemical concentration of concern based on the following types of
data:  1) Lowest chronic effect concentration; 2) Lowest LC-50
concentration of many acute tests; or 3) One LC5Q from an acute
test, or Quantitative structure-activity relationships.

    1.    Lowest Chronic Effect Concentration

          If available information on a chemical includes chronic
effect and no-effect concentrations, an assessment factor of 10
is proposed to estimate concern levels for concentrations of the

-------
chemical in the natural environment.  For example, if the lowest
available no-observable-effect concentration  (NOEC) were for
daphnid reproduction, and it was 0.1 mg/L, a concentration of
0.01 mg/L (i.e., 0.1/10 = 0.01) would be used in preliminary
hazard assessment.

          In addition to the data discussed in the Support
Document (attached), the following issues were considered in
arriving at this particular assessment factor.

          a)     Other species from the same environment  may  be
                more sensitive than the species tested.

          b)     Test conditions may not fully represent  the
                natural  environment.   Typically,  laboratory  test
                conditions  are more optimal for survival  of
                organisms  exposed to  the chemical than those  in
                the natural  environment.

          c)     Other adverse effects may occur  at  levels  lower
                than the effect measured.

-------
          d)     The  presence  of  other toxic  chemicals  in  the
                 natural environment may  amplify  the  effects of
                 the  tested  chemical.   For  example,  fish are more
                 sensitive to  aryl phosphates  in  the  presence of
                 certain pesticides.

    2.    Multiple Acute Toxicity Data

          When  there are acute toxicity  data  for several  species,
an assessment factor of 100 is used to estimate  a concern level
for enviromental concentrations.  The assessment factor will be
applied to data  from the most sensitive  species.  For example, if
the EC5o values  for  at least  5 species of  fish and  invertebrates
range from 20 mg/L to 1 mg/L, the 1 mg/L would be adjusted to
0.01 mg/L (i.e., 1/100 =0.01).  This value  (0.01 mg/L) would be
used in any preliminary hazard assessment.

          For purposes of applying this  assessment factor, the
term "multiple acute toxicity data" refers to either of the
following two cases:  1) at least one acute test  for each of
three taxonomic groups (fish, invertebrates, and algae); or 2) at
least five acute tests divided among two of the  three taxonomic
groups listed above.  Note that two or more acute tests in only
one taxonomic group  are not ^multiple acutes* un^er this
definition;  in such a case the lowest LC5Q and the assessment
factor for a single LC50 would be used (see below).   A detailed
discussion of the "species cluster" concept, upon which the
multiple-acute assessment factor is based, is presented in the
attached Support Document.

                                8

-------
          In addition to the data discussed  in  the attached
Support Document and the four uncertainty factors a - d
enumerated under Lowest Chronic Effect Concentration, the
following additional issues were considered  in arriving at the
above figure:

          e)    Acute toxicity tests are short-term tests, while
                chronic tests are conducted under long-term
                exposures.   Total exposure in a chronic test is
                greater than in an acute test at higher
                concentrations because of the longer duration of
                exposure.   Long-term exposures generally result
                in lower concentrations causing  effects.
                Therefore,  acute LC/EC5Qs are adjusted  downward
                by assessment factors to account for longer
                exposure.   Even  in the absence of prolonged
                exposure due to  persistence,  repeated or
                intermittent exposure may result  in  delayed
                effects  or  effects at concentrations  lower than
                that  indicated by  acute  tests.

          f)     Acute tests generally measure only lethality  as
                an endpoint,  while chronic tests  generally
                measure  sublethal  effects which  normally occur  at
                lower concentrations  than those which cause
                lethality.   Also an LC50  is based on median
                response.   Chronic tests  often seek effects on

-------
       less  than  50%  of  the  population  (e.g.,  an  avian
       reproduction test has the goal of detecting
       effects on 25% or less of a population; a  fish
       early  life  stage may  get significant results'at
       an even lower  percentage response).

g)     EEB has compiled data on the acute-to-chronic
       ratios (ACR) of selected toxicants on fathead
       minnows.  Additional  information on ACRs of
       various fish species  for organic chemials  is
       undergoing analysis.  Although there is
       substantial variation in the ACRs for various
       chemicals, these data support an assessment
       factor of 100 when going from the lowest of
      several LC50 values  to natural environments.

h)    Acute toxicity  tests generally use juvenile and
      adult organisms.   Early-life stages  (eggs  and
      embryos)  are generally more  sensitive to toxic
      chemicals  than  are other  life  stages (juveniles
      and  adults).
                     10

-------
    3.    Acute Toxicity  Data

          An assessment factor  of  1000  is used  for estimating a
concern level  for environmental concentrations  of a chemical when
only one acute test  is available.  For  example,  if the acute LC
or EC5Q for a single species of fish or invertebrate is 8 ppm,
then the concern level for environmental concentrations is 0.008
ppm or 8 ppb (8/1000 = 0.008).

          In addition to  the data discussed in  the attached
Support Document and the  uncertainty factors a - g enumerated
earlier, the following additional issue was considered in
arriving at this figure of 1000.

          h)    There are abundant data showing that different
                species have different sensitivities  to a
                chemical,  but no one or two species  are
                consistently more sensitive  to a broad  array  of
                chemicals.  This natural variation in
                sensitivity,  along with  previously noted
                uncertainty,  supports an assessment  factor of
                1000  when  going  from  a single  LC5Q value  to
                natural  environments.
                               11

-------
    4.
QSAR
          An  assessment  factor of  1,000  is  used  to  predict  a
concern  level  for environmental concentrations when
extrapolations  are  made  from  quantitative structural  activity
relationships  (QSAR).

          In  addition  to  the  data  discussed  in the  attached
Support  Document, the  rationale for  the  above figure  is as
follows:

          i)    QSAR estimates  an  acute  toxicity  level based upon
                certain structural correlates.  Within a given
                class  of  organic compounds,  the actual versus
                estimated acute toxicity (LC^) may vary 10-
                fold.  Considered  in conjunction  with previous
                areas  of  uncertainty, such variation supports an
                assessment  factor  of 1,0OOX when  going from an
                acute  toxicity  value derived from QSAR to natural
                environments.
                         SUPPORT DOCUMENT
Introduction
    EEB has devised a simple but reliable system to estimate
concern levels for environmental effects of chemicals when
minimal data are available.  This approach is intended for use in
                                12

-------
 the PMN review process when the available ecotoxicity data
 consist of either a single LC5Q, multiple LC5() values, QSAR data
 or an acceptable MATC value.   The result is that a determination
 is made relative to exposure  levels as to whether some effects
 testing is appropriate.

     The basis for the system  is the use of "assessment factors"
 to apply  to the  available  aquatic toxicity data  to arrive  at  a
 level which is likely to cause adverse environmental  effects  in
 the  field.   An adverse field  effect is defined as any effect  upon
 growth,  reproduction,  or survival which would  either  temporarily
 or permanently alter  the population level  of wild plants or
 animals.   The laws  supporting  this concept are summarized  in
 "Environmental Effects of  Regulatory Concern under TSCA" (EEB,
 1983).

     If  one  has in situ evidence  of  population  effects  such as
might be derived by biological monitoring or full  scale field
studies, the  results  are taken at face  value and  thus  an
assessment  factor of one is applied to  determine  the significant
field effect  level.

     If  the  best evidence of what  the field effect  level would  be
were  limited  to a valid MATC*, an assessment factor of 10 is
*  To be valid, an MATC must (among other things) be preceded by
acute toxicity tests on several species; the "most sensitive"
species would then be used in the chronic test.
                                13

-------
applied  to  account  for  the various  laboratory-to-field
variables.  Similarly,  if  the  best  evidence of  what  the  field
effect level would  be were limited  to  a  cluster of acceptable
acute LC50  values,  a factor  of 100  is  applied to the  lowest  LC5Q
to project  the  concern  level for  the  field.  Finally,  a  factor  of
1000 is  appropriate when only  a single acute aquatic  toxicity
value or QSAR-based LC50 are known.

    These assessment factors are  derived  to account  for  the  many
individual  experimental and  environmental  factors which  act  to
alter acute laboratory  toxicity test values away from  field
effect levels.  For convenience,  however,  we have avoided great
complexity  by grouping  these numerous  influences under only  three
categories.  They are comparative toxicology, chronicity, and
laboratory-to-field differences.

    Readily available evidence for the values used in each
assessment  factor will constitute the bulk of the remainder of
this technical  support document.
                               14

-------
    Figure #1 illustrates  the  ralationships between the data
which are intended by the  given assessment  factors.
              1000
                   SINGLE ACUTE or QSAR
                    100
                       LOWEST ACUTE
                           MATC
                         10
                                   FIELD EFFECT LEVEL
                               15

-------
    Figure #2 illustrates these  linkages with  a more
sophisticated model which pictures how the  idealized
relationships between the various forms of  data are viewed
statistically.  The figure makes use of the log dose-probit
response model which so often  fits the data.

    Experimental evidence exists linking each  of these tests
statistically.  The frequency  of one test result (say the single
LC^Q) being a certain amount higher than another test result (say
the lowest of several LC50 values) can be derived from available
data on a wide array of test chemicals.

    Before proceeding with experimental evidence of these
relationships, an historical review of regulatory use of
assessment factors to represent  these relationships is in
order.  Such perspective should provide an  indication of the
degree to which regulatory use has been made of this concept.

B.  Historical Review of Assessment Factors

    Previous use of factors in the absence of  specific chemical
data is well established.   In the human health field,  for
example, Dourson and Stara (in press)  have reviewed the extensive
history of the use of the  numbers 10,  100, and 1000 in regulatory
                                16

-------
Figure 2.  Relative Effect Level Model
                                                               most likely
                                                               single LC-50
                       field effects
                       concern
                       leve
                                    10-fold
                                  V
                                  1
   10-fold    10-fold
Y
10
• f
 100
1000
                                        Relative Concentrations
                                   16a

-------
 history.   One example,  that of the 100-fold uncertainty factor,
 is  attributed variously to Lehman and Fitzhugh (1954),  Bigwood
 (1973),  and  Vettorazzi  (1976 and 1980).
   •

     Although the  specific  areas  of uncertainty described  by  these
 authors  differ somewhat,  they can be  viewed as due  to  intra- or
 interspecies variability.   It has also been suggested  that two
 10-fold  uncertainty  factors,  one for  each  type of variability, be
 used  to  describe  the  100-fold factor  in  some  instances  (Bigwood,
 1973; Klassen and Doull, 1980; Food Safety  Council,  1982).


    The  FDA  expanded  their  approach when chronic data were
 unavailable.   In  cases  where  only subchronic  animal  NOELs (no
 observed effect levels) or  NOAEL  (no observed  adverse effect
 levels) were available  for  only  two species,  the FDA recommended
 a factor of  1000  instead of 100.   The additional 10-fold was due
 ostensibly to  the added uncertainty when estimating an ADI
 (acceptable  daily intake) from adequate short-term toxicity data
 (Kokoski, 1976).  If  subchronic data were available for only one
 species,  a 2000-fold  uncertainty  factor was recommended (Shibko,
 1981).  The  National Academy of Sciences (NAS, 1977) has
 recommended a similar approach for drinking water.   Finally,  the
USEPA (1980)  has
                               17

-------
previously used  the MAS  reasoning  for  pollutants  in  ambient
waters, under the consent decree.   Figure  3,  from  Dourson  and
Stara (1983),  illustrates a  human  health  use of uncertainty
factors.

    The record of  use of uncertainty factors does  not  stop with
human health.  The Research  and Development  Department of the
Procter and Gamble Company (Duthie, 1979) proposed that "Testing
for aquatic effects is indicated for any  chemical  that is
expected to occur  regularly  or continuously  in surface waters at
concentrations greater than  0.001 mg/1..."   This 0.001 mg/1 "is
calculated as resulting  from the disposal of  a material that is
uniformly distributed across the U.S.  in quantities of 1 million
pounds/year."  The proposal  continues  that "the same
concentration could be obtained by the release of much smaller
quantities in a narrowly bounded locality."   The proposal further
states that the LCSOs of  several species of  fish should be
                               18

-------
Figure 3  Guidelines for the Use of Uncertainty (Safety) Factors

(1) Use a 10-fold factor when extrapolating from valid
    experimental results from studies on prolonged ingestion by
    man.  This 10-fold factor protects the sensitive members of
    human population estimated from data garnered on average
    healthy individuals

(2) Use a 100-fold factor when extrapolating from valid results
    of long-term feeding studies on experimental animals with
    results of studies of human ingestion not available or scanty
    (e.g., acute exposure only).  This represents an additional
    10-fold uncertainty factor in extrapolating data from the
    average animal to the average man.

(3) Use a 1000-fold factor when extrapolating from less than
    chronic results on experimental animals with no useful long
    term or acute human data.  This represents an additional 10-
    fold uncertainty factor in extrapolating from less than
    chronic to chronic exposure.

(4) Use an additional uncertainty factor of between 1 and 10
    depending  on the sensitivity of the adverse effect when
    deriving an ADI from LOAEL.   This uncertainty factor drops
    the LOAEL  into the range of  a NOAEL.

    FROM:  Dourson and Stara (in  press).
                               19

-------
 greater than 100 times the concentration to exist in surface
 waters  if  concerns  for hazard are to be  dropped.   Beck  et  al.
 (1980)  and Beck  et  al. (1981) proposed essentially the  same
 concepts but laid them out in greater detail.

    Kimerle et al.  (1978)  of  Monsanto Industrial  Chemicals
 Company  presented hazard evaluation  criteria as follows:   If the
 screening  LC50 is between  1 and  l/1000th  the estimated
 environmental concentration (EEC), continue testing.  If the
 lowest of  additional acutes are  more than l/500th the EEC,
 continue testing.   If  the  short-term chronic (estimated MATC) is
 more than  l/50th  the EEC or the  long-term chronic  (measured MATC)
 is more  than l/20th the EEC,  continue testing.  If the impact on
 ecosystems  in field studies is minor, continue testing.   If the
 ecosystem  is impaired, no  use or restricted use of the chemical
 was proposed.

    The American  Institute of  Biological  Sciences (unpublished)
reported, to EPA, their Task Group recommendation that
 "...experimentation has shown, however that there sometimes
exists a direct relationship between acute and  chronic [aquatic]
t ox i c i ty
                               20

-------
 of  chemicals  to aquatic organisms and that the acute LC50 can  be
 used,  within  certain  limits  to estimate  the MATC."   AIBS
 described  the ratio as  generally between 2 and .10,000,  but felt
 that 100 was  a useable  figure.  '

    The European standard-setting organization EIFAC and  the FAO
 suggested  that concentrations  which  are  at least  four orders of
 magnitude  lower than  the 96-hr LC50  from a standard  acute
 toxicity test are surely harmless to fish  populations.  Calamari
 et  al. (198-)  in quoting this  figure (i.e.,  10,000)  calls  it too
                      *
 restrictive.

    The Organization  for Economic Cooperation  and Development
 proposed that  further confirmatory testing  take place if  the NOEC
 (similar to our  MATC) for Daphnia, fish, or  algae is  not greater
 than 10 times  the environmental concentration  (OECD,  1980).

    Finally,  OTS contracted with  Life System,  Inc. to bring
 together the  top environmental  toxicology  scientists and chemists
 frcn the United States  to a series of testing  trigger workshops
 in  1982 and 1983.  The  formal report of this workshop to EPA
 contained the  recommendation that  if the EC50 or LC50 from a
 single aquatic species,  divided by 1000,  equalled or exceeded the
 environmental exposure,  acute tests with three or more species
 should be initiated.  Also proposed by the workshop  for use by
 EPA was that  "chronic"  tests (fish early-life stage, invertebrate
 life cycle) be initiated if the environmental exposure would be
equal to or greater than the acute effect level for  the most
sensitive of three or  more [aquatic]  animal species.

-------
 Data  In  Support  of  Assessment Factors



    With all  these  authors  and respected authoritative groups

 proposing  and  using factors of 10,  100,  and 1000,  two questions

 arise.   First, why  have  all these people come  to  the same

 conclusion?  Second,  why are the  numbers the magnitude that  they

 are?  Let  us  next  investigate the available experimental evidence

 to see what justification exists  for these  numbers.



    What are these  variables which  can so greatly  influence  the

 outcome  of ecotoxicity studies?   They must  be  taken  into account

 if test  results  are used to project possible field population

 effects.  Crosby et  al.  (1966) reported  on  the  importance of the

 variables which  they  noticed  to significantly  alter  ecotoxicity

 figures  derived  in  aquatic  bioassays with Daphnia magna.  One

 might expect these  variables*  to  affect  other  aquatic  test

 results  as well.  The list  included age  and instar,  temperature

 (400), light intensity,  light  quality and periodicity,  food

quantity and type, water quality  (100),  pH  (61) (Huey et al.,

 1982), oxygen-carbon dioxide content of  the test water, test

animal population density,  test duration, length of pre-test

 fasting  period, criteria  for detection of toxicity (6),

 randomization and selection of individual test organisms, purity

of the toxicant,  and physical  form of the toxicant (2).  If these
*  Factors in parentheses after each variable given above are
extreme ratios found.  Though these extremes seldom occur,
knowledge of them should give one caution when decisions to
forego further, or any, testing  are made.


                                22

-------
and other  variables  can influence the outcome of  an ecotoxicity
test  in the  laboratory,  it  can  be expected  that the same
variables  can  likewise  influence  the  outcome  of chemical
exposures  in the  field.   A  thorough discussion of  this  can  be
found  in Tucker and  Leitzke (1979).

    The fact that many  modern-day aquatic testing  procedures have
been  standardized to some extent  should work  to reduce  the
likelihood of  seeing extreme  variability in the toxicity
values.  But it is not  to be  concluded that nature has  been
standardized.  These variables  will not disappear  under field
exposure conditions.  Nor are the data available for evaluating
PMN chemicals  always derived  from state-of-the-art  testing
protcols and procedures

    Since there are  entirely  too  many factors which may alter
effect levels  to take all of  them  individually into careful
account everytime a  testing decision  is made, it is proposed to
group all the  variables under just three headings  for
convenience.   They are species-to-species sensitivity related
variables,  chronicity related variables,  and laboratory-to-field
related variables.
                               23

-------
 Species-To-Species  Variation

    This section deals  with readily available  information on
 acute  toxicity  to aquatic  organisms which  shows  that  an
 assessment  factor of  about 10  can be used  to account  for  species
 to species  variation, on average.   That  is  to  say, given  any
 single  LC-50  for an aquatic species,  dividing  by an assessment
 factor  of 10  will provide  a figure  which is likely to encompass
 the most sensitive  species about  50%  of  the time.  This
 assessment  factor,  therefore,  is  applied when  there is only  a
 single  LC50 for  aquatic species.
    a.    The "Cluster Concept"

    Don  Mount of  the EPA Duluth Environmental  Research
 Laboratory* has  said, "We  can  see that [aquatic] species
 sensitivity distributes itself in a  rather consistent way for
 most chemicals.... The distribution resembles a  log normal one.
We can then take  the approach of sampling this distribution  in
order to predict  the range  about the mean."  He goes on to say,
 "This is in reality our objective.  Each species we test then is
 not representating any other species but is one estimate of
 sensitivity and with several such estimates then of the overall
 range of sensitivity for all species."  Dr. Mount concludes:  with
 •{the data in Table  I of th« consent 
-------
 species  -  rainbow trout,  fathead,  bluegill,  D.  magna and Gammarus
 (included  2  species)  -  is  a  large  part  of  the range  of  all
 species  tested  for most chemicals."   This  idea  has  been called
 the  "cluster concept."

     Specifically,  Dr. Mount  is arguing  that  a limited set of
 aquatic  toxicity  data for  several  fish  and invertebrates  is
 representative  enough to provide a good indication of the most
 sensitive  aquatic  species.   This concept is  useful since  we
 cannot reasonably  require  testing  on  all aquatic  species  which
 might be exposed  to a chemical.  Please note also that  Dr.
 Mount's  use  of  the  cluster concept includes  only  fish and
 invertebrates;  EEB has  included algae in the species cluster to
 be considered,  although we do  not  have  hard  data  to  support this
 approach.

 b.  Analysis  of Data

    Work by  comparative ecotoxicologists initially focused on
 identifying  relationships between  specific species reactions to
 chemicals.   This relationships was typically expressed  as a range
 of toxicity  values, from a numerically  low value (most
 toxic/sensitive'species) to  a numerically high value (least
 toxic/insensitive specie*).  For example, although Tiot backed fey
any presented data, Kenaga (1978)  reported that a Daphnia LC50
would lie within ± 3.26  orders of magnitude (i.e., the  Daphnia
LC50 divided or multiplied by about 200) of the LC50 of the same
                                25

-------
 chemical for the rainbow trout, 95% of the time for all chemical
 groups together.  He similarly reported that the various species
 of saltwater fish LC50 values lie within ± 0.83 orders of
 magnitude (i.e., a single saltwater fish LC50 multiplied or
 divided by about 7)  95% of the time for chlorinated hydrocarbon
 insecticides.   Most  other aquatic species correlations he made
 also tended to fall  between ± 1 and 3  orders of magnitude,  95%  of
 the  time (i.e.,  a single LC50 multiplied  or  divided by 10 or
 10UO).   This approach   is not useful  for  regulatory purposes
 because  we  are not interested in  how  insensitive other species
 may  be  compared  to a given LC-50,  but  in  how sensitive other
 species  may  be.

     Statistical  Consultants  Inc.*  under contract to OTS,  reported
 results  of a similar nature  in  a way more useful for regulatory
 purposes.  They  based  their  results upon  large  numbers  of fish
 toxicity figures  of a  type  (tests, species,  chemicals,  etc.)
 expected to  be used under TSCA.  They determined the factors by
 which one would have to divide any specified fish LC50  in order
 to arrive at the  lowest of the LC50 values for any of eleven
other fish species for which they had toxicity data.  The result
was  that a divisor of 48.9 would
   Personal communication

                               26

-------
 include  the  most  sensitive  of  these  eleven  species  95%  of  the
 time.  Likewise,  19.6 would include  the  most  sensitive  of  the
 fish  90% of  the time, and 2.4  would  include the  most  sensitive
 50% of the time.  A  total of 920  LC50  values  were used  in  these
 computations.  However,  these  data were  solely on one taxonomic
 group of aquatic  organisms,  i.e., fishes.   In reality,  we  must
 protect  a far more diverse  group  of  organisms than  just bony
 fish.

    In one exercise, we  have used data from the  Handbook of Acute
 Toxicity of  Chemicals to Fish  and Aquatic Invertebrates.
 (Johnson & Finley, 1980).   The Handbook  contains 1587 acute tests
 with 271 chemicals on 28 species of  fish and  30  species of
 aquatic  invertebrates.   However, there  were  only 67  chemicals
 for which data was presented for 6 or more species.    For these
 chemicals the median overall range of sensitivity of  the tested
organisms was 450.  Now if one takes the view of Dr. Mount quoted
earlier, that the distribution of species sensivity is  log
normal,  Figure #4 would well represent the distribution of LC50
values.
                               27

-------
Figure 4
                              28

-------
    Thus  if the  lowest LC50 value were  set at 1 ppb,  the highest

LC50 value would be projected  to be 450 ppb.  Owing to the log

normal nature of this curve, the highest frequency LC50 would be

calculated to lie at 21.12 ppb.  If one assumes therefore that

the most  likely LC50 to be reviewed by EPA is 21 ppb  and the

lowest is 1 ppb, a divisor of  21 is indicated as the  50th

percentile when projecting lowest LC50  from a "typical" LC50.
    To summarize, we have presented data (See Table 1) to show

that an assessment factor of about 10 can be used to estimate,

from a single LC50, a figure which encompasses the LC50 for the

most sensitive aquatic species.  As can be seen from the data in

Table 1, our assessment factor of 10 is a ballpark figure;

assessment factors derived from actual data can vary depending on

the number and kinds of both species and chemicals tested.
Table 1.  Assessment Factors for Species to Species Variation
                             Assessment
Species         Chemicals    Factor  Confidence      Ref.
Fish            all groups    10      50%            this
invertebrates                            paper
algae
Fish only
(11 species)
Fish/
invertebrates
(6 or more
species)
TSCA-type    48.9    95% statistical
chemicals    19.6    90% consultants
              2.4    50% (pers. comm.)
67           21      50% EEB analysis
chemicals                of data in
                         Johnson &
                         Finley (1980)
                               29

-------
 Acute - To - Chronic Toxicity Ratio

     Data which are readily available have been reviewed in the section
 that follows to evaluate the level of support for an additional factor
 of 10 to account for chronicity and related variables.

     Any assessment factors based upon acute toxicity data alone must
 take into account the fact that longer exposures or longer response
 times may greatly reduce the dosage level at which toxic effects
 occur.   The degree to which this happens is called "chronicity."  A
 measure of chronicity which we  commonly use is that of  the LC50
 divided by the MATC.   This acute-to-chronic ratio (ACR)  contains other
 elements  in addition  to  pure chronicity.   For example,  a 50%  effect is
                              ^
 denoted by the LC50 while  a "just significant effect"  level  is  denoted
 by the  MATC.   Also, the  MATC can  be based upon effects  such as
 reproduction which can be  more  sensitive  to a chemical  than  is
 mortality.

    The assessment factors  used by  EEB  are  based,  among  other things,
 upon an MATC which is derived from  less than  full  length
 chronic tests.  We utilize  early  life stage  tests  to derive MATC
 values, for  example.  While  several  life  stages are implied in this
MATC, it  is  not based on full-length chronic  testing.  It would  need
 to be divided by about two, on an average, based on 34 chemicals to
project the  "real" chronic MATC values  from  full life-cycle studies
 (McKim, 1977).
                                30

-------
     Goodman (in press)  found chronicity ratios arrayed between 2-and
 3300-fold  for  estuarine fish.   The most sensitive effects were
 reported  as reduced survival,  growth,  or fecundity.   Call et al.
 (1983)  reported that the acute LC50 was 17,551 times higher than  the
 MATC for  propanil  on fathead minnow.   We have seen few ratios as  high
 as  this one.

     Although we have, on occasions seen very large acute-to chronic
 (LC50/MATC  = ACR)  ratios,  this does not mean they are always large.
 EEB  staff have  from time to  time  performed  analyses  upon readily
 available data  in  limited  efforts  to determine the statistical
 relationships  between the  array of acute-to-chronic  (ACR)  ratios  for
 various species.

                                                          •
    One series  of  computations (Ells,  personal communication)  resulted
 in our  determination of  the  approximate  aquatic chronicity  ratio  which
 is exceeded in  tests with  50%  of  the chemicals.   More data  would  be
 required to reliably establish the ACR  which  is exceeded  just  5%  of
 the  time.   Using a  set  of  29 available ACRs  on fathead minnows for
organic chemicals having acute LC50 values in  excess  of  1 ppm, the
median  chronicity ratio  was  found  to be  7.   This  means if LC50 values
 for  fathead minnows  for  these  chemicals  were divided  by  7, half of the
MATCs would be  found below the dividend  and  half  would be found
above.  On this same data set,  a divisor of  28 would have reduced the
LC50 to a value below which  about  5% of  the MATC  values would lie.
Stated another way, one could  have 95% confidence  that dividing an
LC50 by 28 in this data  set would  produce a concentration below which
the true MATC would not  be found in more than  5%  of the cases.
                                31

-------
     When additional data on fathead minnows was added  to  this data
 set, the median chronicity ratio dropped very slightly to 6.4, but the
 95th percentile rose dramatically to 300!  When acute  to chronic
 ratios for many other species were added to this data set to make a
 total of 219 cases, the median chronicity ratio was 6.4,  but the 95th
 percentile was 500.  This data set included much the same mix of
 species found in OTS's ecotoxicity guidelines:  fathead minnows,
 sheepshead minnows, rainbow trout, brook trout,  Daphnia maqna,  and
 mysid shrimp.

     Further data analyses on the same  data set indicated  to  us  that
 restricting the chemicals to just organics (omitting metals  and
 inorganics)  would bring  the 50th and 95th percentile ratios  down
 somewhat,  in this  instance  to  5.2 and  180,  respectively.

     Contract statistical  studies  are underway  by Statistical
 Consultants  Incorporated  (SCI)  to refine  the acute  to chronic ratio
 estimates  by use  of  additional  data sets.   We  have  received  interim
 phone reports as  follows:   An analysis has  been done on a  data set  for
 fish  containing ACRs for  95 chemicals  including metals, pesticides,
 and  industrial  chemicals.   The  median ACR was  reported  as  8.46.  This
means that half of  the tested chemicals produced MATC values below the
 acute LC50 divided by 8.46.  Five percent of these chemicals produced
MATC values  below l/875th of the  acute LC50.   This ratio comes to
1/522 if only chemicals having  LC50 values below 1 ppm are considered.
                                32

-------
    In summary, a factor of about  10 'seems  likely  to  be  a  useful
factor to project "typical" chronicity  (See Table  2).  No  doubt many
more special relationships will arise as we and SCI continue  to obtain
and evaluate increasing amounts of data.  These will  improve  our
analyses and could refine the statistical relationships  between
various toxicity test results.  We will continue to test certain
surrogate species in the laboratory and apply  the  results  to  untested
or untestable species.  The real proof of these methodologies will be
the degree to which these assessment factors will  flag the appropriate
chemicals for adequate testing/ upon which results sound risk
assessments can be made.  A specific available test of these
assessment factors, then, is to see what would have happened  had we
applied these factors to chemicals for which we already  know  the field
effect levels.   The next section is an attempt to  review these
laboratory to field ratios.
                                33

-------
 Table  2.
Assessment Factors for Acute-to Chronic Ratio
                   Assessment
      Chemicals    Factor  Confidence      Ref.
      all groups
 Speci
es
 fish
 Invertebrates
 algae

 Fathead
 Minnow

 Fathead
 mi nnow
 Fathead Minn.
 Sheep she ad
 Minnow
 Rainbow &
 brook trout
 D. magna,
 mys ids

 Same as
 above
Fish
                        10
      29 chemicals  7
                   28
      unkn
      219
      unk.  #:
      same  as
      above excl.
      metals &
      inorganics

      95
      (includes
      metals,
      pesticides,
      industrial
      chems.)
                        6.4
                        300

                        6.4
                        500
                        5.2
                        180
                        8.46
                        875
 50%  this paper
 50% Ells
 95% (pers. comm.)


 50% EEB
 95% analysis

 50% EEB
 95% analysis
50% EEB
95% analysis
50% Statist.
95% consultant
    (pers. comm. )
Laboratory - To - Field Comparisons



    This section deals with some of the readily available field effect

levels and compares these to the levels projected by the EEB

assessment factors.  Due to limited field data these comparisons can

at present only be anecdotal rather than statistical in nature.
                               34

-------
     Some may question the need to divide an acute LC50 for, say, a
 fish by a number as great as 1000 to arrive at a level above which EEB
 would assume concern for adverse environmental effects in the field.
 The  technical staff of EEB,  using their best scientific judgement,
 felt that levels above 1000th of any .single LC50 value represented too
 great a risk of  field effects.   This was based upon  experience  with
 laboratory and field effects levels, plus a sense that while no
 particular level of sureness has been  made into a policy,  any less
 than 95%  sureness would not  be  acceptable.  Even at  95%,  one in twenty
 environmentally  hazardous PMN chemicals would  accidently  be  allowed
 through the review process without  testing or  regulation.

     How well  did the  staff do?   Available  data are reviewed  below;
 much  more  data exist  and,  with  adequate resources, could  be  included.

     In  1969,  standard  application of malathion used  in  mosquito
 control resulted  in an immediate concentration of 2.0-3.2  ug/1  in  a
 saline  marsh  with  rapid reduction thereafter.   If one had  the goldfish
 acute LC50  of  10,700 ug/1  from Johnson  and  Finley (1980),  it would
 have been easy to  assume that this level  (2-3  ug/1) was safe since
 this LC50  is  about  3300 to 5300  times higher than the highest field
 exposure.   It would have been easy,  but  it would have been wrong,
 because 14  to 80%  of the valuable brown  shrimp  and white shrimp were
 killed  in the marsh (Conte & Parker, 1975).  But some would say
goldfish are  insensitive.  Had the available LC50 been  for carp it
would have been 6590 ug/1.  Had the  available  LC50 been for an OTS
guideline recommended species, such  as the fathead minnow, the LC50
                                35

-------
 would be 8650 ug/1.  This is a good example of how the EEB assessment
 factor of 1000, large as it may seem still would not be sufficiently
 protective.  Other examples follow.

     Pimental (1971) discusses data by Oliver et al.  (1966) showing how
 algal productivity was reduced in aquatic habitats by as little as 2
 Ib/A of cacodylic acid.   This translates to less than 2 mg/1,  and
 likely less than 1 mg/1.   The LC50 for mosquito fish  and taillight
 shiners is about 1000 mg/1.

     Wellborn (1969)  reported an LC50  of 315 mg/1  for  diquat  on striped
 bass.   But Tatum and Blackburn (1962)  had already  reported that  0.5
 mg/1 adversely  affected  plankton in outdoor ponds.

     Bottom-dwelling  organisms  of  a great  variety were reduced  by  50%
 or more following  applications  of  0.5  to  10.0 mg/1, of  simazine
 (Walker,  1962).  On  the other  hand, the  laboratory LC50  of simazine
 for  bluegills (a test  species  recommended  by OTS)  is 118 mg/1
 (Bohmont,  1967).

     The LC50 of copper sulfate  to  bluegills is 45 mg/1 at 8 grains
 water hardness  (McKee & Wolf,  1963).  In contrast 0.05-0.08 mg/1
 applied to ponds had demonstrated  effects upon a great variety of pond
animals and plants.  In this instance an assessment factor of about
900 would have been required to protect the aquatic environment.
Since the LC5Q for bluegill drops  to 0.6 mg/1 when water hardness is
only 1 grain, an assessment factor of  only 12 might have sufficed.
                               36

-------
 But which LC50 would have been the first or only one submitted if this
 were a PMN?

     Farm ponds with 2.8  mg/1  silvex (K + salt)  showed disruption of
 the benthic community and rapid  decreases in several species.   The
 bluegill LC50 for the same material is 100 mg/1  (Hughes  &  Davis,
 1965).

     Treatment of  outdoor ponds with 2.39  ug/1 Dursban produced  a  46%
 kill  of  bass  and  a  55% kill of bluegills.   A treatment of  0.97  ug/1
 produced  10%  kill of  bass in  the ponds  (Macek et  al.,  1972).   By  way
 of  comparison,  the  LC5Q  of Dursban  for  channel catfish, a  common  test
 species,  is 280 ug/1  (Johnson & Finley, 1980).

     In the previous paragraphs we have  noted several  examples where
 assessment factors of at  least 236, 289,  630, 900, 1000, or even 5300
 would have been needed (See Table 3).   To be sure, lower factors might
 sometimes suffice, but we have no means as yet of telling when this
would be the case for any PMN chemical when we have but a single
aquatic LC50 as our basis.
                               37

-------
Table 3.
Species
Fish
Inverte-
brates
algae


Chemicals
Malathion


Cacodylic
acid


Diquat




Simazine



Copper
sulfate

Assessment Factors for LC-50 to Field Effect Levels
Chemicals Assessment Factor Confidence Ref.
all groups

Test
species
Goldfish
Carp
Fathead
Minnow
Mosquito
fish &
tail-light
shiners
Striped
Bass



Bluegill



Bluegill

1000 95%

Lab Field
LC50 Species
10,700 ug/L Brown &
6,590 ug/L White
8,650 ug/L Shrimp
this
EEB's
Effects &
Effect Level
14-80%
mortality
at
paper

Calculated
Assess, factor
3343-5350
2059-3295
2703-4325
2.0-3.2 ug/L
both algae
§1000 reduced
mg/1 at

-------
(Chart continue)

Silvex          Bluegill
Dursban
             100 mg/L    Fann pond
                     benth ic
                     species
                         2.8 mg/L
                        disruption or
                        rapid  popul.
                        decrease  at
Channel
Catfish
280 ug/L    Bass and    46%  bass  kill
        Bluegill        55%  bluegill
            kill at  2.39
            ug/1;
            10%  bass kill
            at 0.77  ug/L
35.7
                                                                     117  &
                                                                     288

                                                                     Finley
Hughes &
Davis (1965)
                Macek et
                (1977) &
                Johnson &

                (1980)
         al

-------
     Based  upon expert judgement and experience, the use of an
 assessment factor of 1000 seems reasonable to project concern
 levels  from any single LC50.   If additional resources become
 available,  we  should focus additional study on the degree to
 which  laboratory results can  be predictive of field effect
 levels.  More  collection and  analysis of  past field data (such  as
 that of  the US Fish and Wildlife Service),  more actual field
 studies, and more terrestrial studies are  needed to refine our
 projections from laboratory to field.   But we do now have a
 working model.

 QSAR-Calculated  LC-SOs

     In the  absence  of  other data,  LC-50s  for aquatic species  can
 be calculated  from  the  log  P  of  certain compounds  (alchols,
 ke tones, ethers,  alkyl  halides,  and  benzenes).   Veith et  al.
 (1983). examined  54  industrial  chemicals from the TSCA Inventory
 which fall  into  the  above  chemical classes.   QSAR  -  based  LCSOs
 were calculated  for  these  chemicals  and compared to  actual
 (measured)  LCSOs  for the  same  compounds.  They  concluded,  "All of
 the predicted  values were within one order of magnitude of the
measured values,  with a  range  for observed/predicted  ratios from
 0.12 for 3,3-dimethyl-2-butanone to  4.22 for  2-propanol."

    The authors  also noted, "Predicting the mode-of-action of
chemicals remains a  difficult  task and the errors associated with
selecting the wrong  structure-toxicity mechanism may be greater
                               40

-------
 than  the  standard  error of  the estimate associated  with any
                       •
 individual  QSAR.   We  recommend that  the equation  presented  herein

 be  used only  for  [alcohols,  ketones,  ethers,  alkyl  halides,  and

 benzenes] or  for chemicals  for which  there  is  evidence  that- they

 or  analogous  structures are  narcotic."



    Since the QSAR-calculated  LCSOs  for these  classes of

 chemicals is  correct  within  one order of magnitude  when compared

 to  measured LCSOs, EEB  assumes  that QSAR-calculated LCSOs for

 chemicals in  the above  classes  may be treated  for practical
                                                 *
 purposes as if they were measured LCSOs.  Therefore, in using a

 QSAR-calculated LC50  to determine concern levels  for field

 concentrations, we will apply  the assessment  factor for single,

 measured LCSOs (i.e., 1000).   It should also  be noted that QSAR-
                                              \
 calculated LCSOs are  minimum values,  and that  actual toxicity may

 be  greater  if other,  or additional, modes of action than narcosis

 are operating.



 Summary




    EEB has developed a series  of assessment factors to estimate

concern levels for field concentrations of a chemical when only

 limited data on that  chemical are available.  These assessment

 factors are  intended  to account for the different levels of

certainty associated with variously-derived toxicity figures (See

Table 4).
                                41

-------
Table 4.  EEB Assessment Factors
Toxicity Derivation          Assessment
method                       Factor
QSAR or Acute LC50           1000
with single species

Acute LC50 with several      100
species (use value for most
sensitive species)

Chronic toxicity value       10
 (MATC)

Field test value             1
    We have presented anecdotal data to support an assessment
factor of 1000 for estimating concern levels for field
concentrations from a single LC50.  We have also presented more
quantitative data to support a factor of 10 between a single
"typical" LC50 and the LC50 for the most sensitive species (i.e.,
the single LC50 divided by 10).  Since the assessment factor for
single LC50 to field level is 1000, and since the factor between
single and multiple LCSOs is 10,  it follows that the assessment
factor to estimate field concern  levels from multiple LCSOs is
100.
                               42

-------
    We  have  also  presented  quantitative data to support an
 additional factor of  10  to  account  for  variables between the  most
 sensitive of multiple LCSOs and  the MATC (chronic toxicity
 value).  Since  the assessment  factor for multiple LCSOs to field
 concern level  is  100,  and since  the factor  between multiple LCSOs
 and the  MATC is 10,- it follows that the assessment factor  to
 estimate field  concern levels  from  the  MATC  is  10.

    We  have also  presented  data  (cited  in Veith  et  al.,  1983) to
 show that QSAR-calculated LCSOs  for specified classes of
 chemicals are correct within an order of magnitude, compared to
 the measured LCSOs.   Therefore we assume  that,  for  practical
purposes, QSAR-calcluated LCSOs may  be  treated as  if they were
measured LCSOs for a  single species, and that an  assessment
 factor of 1000 should be used to estimate field concern levels.
                               43

-------
 Literature Cited

 American Institute of Biological Sciences,   Aquatic Hazards of Pesticides Task
      Group.  Unpublished.   Report to the Environmental Protection Agency:
      Criteria and  rationale for decision making in aquatic hazard evaluation.
      U.S. E.P.A, Off. Pest. Progr.,  Contract  No.  68-01-2457.

 Beck, L. W.,  A.  W. Maki, N. R.  Artman,  and  E.  R. Wilson.  1980.  Outline and
      criteria for  evaluating the safety of  new detergent  chemicals, p. VII-
      1-12.  In:  Environmental risk assessment  — An integrated  approach.  First
      Ann. Meet.  Soc.  Environ. Toxicol.  Chem.,  November 23-25,  1980.

 Beck, L. W.,  A.  W. Maki, N. R.  Artman,  and  E.  R. Wilson.  1981.  Outline and
      criteria for  evaluating the safety of  new chemicals.   Reg. Tox. Pharm.
      1:19-58.

 Bigwcod,  E. J. 1973.   The acceptable daily  intake  of  food  additives.  C.R.C.
      Crit.  Rev.  Toxicol. June,  41-93.

 Bohmont,  B. L. 1967.   Toxicity  of herbicides to livestock,  fish,  honeybees, and
      wildlife.   20th  West.  Weed Cont. Conf.  ,  Proc. 21:25-27.

 Butler,  P. A. 1964.   Coimercial fishery inestigations: Effects of DDT in a
      tidal marsh,  p.  14-15.   In:  Pesticide-wildlife studies, 1963: A review
      of  Fish  and Wildlife Service investigations during the calender year.
      U.S.D.I., Fish Wildl.  Serv., Circ.  199.

 Cairns, J., Jr., K.   L. Dickson, and A. W. Maki (eds.). 1978.  Estimating the
      hazard of chemical substances to aquatic  life.  Amer. Soc. Test. Mat.,
      ASTM Spec. Tech.  Publ. 657.  278 p.

 Calamari, D., R. Da Gasso, S. Galassi, A. Provini, and M. Vighi. 198 .
      Estimating the hazard of eight amines on aquatic life, p. 329-341.
      Source is an  unidentified  book.

Call, D. J.,  L. T. Brooks, R. J. Kent, M. L. Knuth, C. Anderson, and C.
      Moriarity. 1983.  Toxicity, bioconcentration, and metabolism of the
      herbicide propanil (31, 4'-dichloropropionanilide) in freshwater fish.
     Arch. Environ. Contain. Toxicol. 12(2)sl75-182.

Conte, F. S. and J. C. Parker.  1975.  Effect of aerially-applied malathion
     on juvenile brown and white shrimp Penaeus aztecus and P. setiferus.
     Trans. Amer. Fish. Soc. 104(4):793-799.                ~

Crosby, D. G., R. K. Tucker, and N. Aharonson.  1966.  The detection of acute
      toxicity with Daphnia magna.  Food Cosmet. Toxicol. 4:503-514.

Dourson, M. L. and J. F. Stara.  In press.  Regulatory history and  experimental
     support of uncertainty (safety)  factors.  Reg. Toxicol. Pharmacol.  15 p.

Duthie, J. R. Prepublication. Outline and criteria for evaluating the safety
     of new chemicals.  Proctor  and Gamble Company, Res. Dev. Dep.,  Cincinati,
     Ohio.  37 p.

-------
Food Safety Council. 1982.  A proposed food safety evaluation process.  The
     Nutrition Foundation, Inc. Washington, D. C.

Goodman, L. R. In press.  Chronic toxicity of organophosphorus pesticides to
     estuarine fish  (Abstract).  Presented at the Symposium on organophosphorus
     pesticides in the marine environment, June 7-9, 1981, Duke Univ. Mar. Lab.

Harp, G. L. and R. S. Campbell. 1964.  Effects of the herbicide silvex on
     benthos of a farm pond.  J. Wildl. Managem. 28(2):308-317.

Hays, H. W. 1969.  Uses, regulations, and registration of chemicals used in
     fish culture and management, 6 p.  Int Fish and Chemicals: A symposium
     on registration and 'clearance of chemicals for fish culture and fishery
     management.  A symposium arranged by the Bureau of Sport Fisheries and
     Wildlife for the 99th Annual Meeting of American Fisheries Society, New
     Orleans, Louisiana, September 12, 1969.

Houser, A. 1963.  Loss of weight of sunfish following aquatic vegetation
     control using the herbicide silvex.  Proc. Okla. Acad. Sci. 43:232-237.

Huey, D. W., M. C. Wcoten, L. A. Freeman, and T. L. Beitinger. 1982.  Effect of
     pH and chloride on nitrile-induced lethality in bluegill (Lepomis
     macrochirus).  Bull. Environ. Contain. Toxicol. 28(1):3-6.

Hughes, J. S. and J. T. Davis. 1965.  Comparative toxicity to bluegill sunfish
     of granular and liquid herbicides.  S. E. Assoc. Game Fish Conn., Ann.
     Conf., Proc. 16:319-323.

Johnson, W. W. and M. T. Finley. 1980.  Handbook of acute toxicity of chemicals
     to fish and aquatic invertebrates.  U.S.D.I., Fish Wildl. Serv., Res. Publ.
     137.  98 p.

Kenaga, E. E. 1978.  Test organisms and methods useful for early assessment of
     acute toxicity of chemicals.  Environ. Sci. Technol. 12(11):1322-1329.

Kimmerle, R. A., W. E. Gledhill, and G. J. Levinskas. 1978.  Environmental
     safety assessment of new materials, p. 132-146.  In; Cairns, J., Jr., K.
     L. Dickson, and A. W. Maki (eds.). Estimating the hazard of chemical
     substances to aquatic life.  Amer. See. Test. Mat., ASTM Spec. Tech. Publ.
     657.

Klassen, C. D. and J. Doull. 1980.  Evaluation of safety: Toxicological evalu-
     ation, p. 26.  In; C. D. Klassen and M. 0. Amdur. (eds.). Toxicology.
     McMillan, New York.

Kokoski, C. J. 1976.  Written testimony of Charles J. Kokoski, Oep. Heal. Educ.
     Welf., Food Drug Admin., Docket No. 76N-0070, Washington, D. C.

Lehman, A.J. and 0. G. Fitzhugh. 1954.  100-Fold margin of safety.  Assoc. Food
     Drug Off. U. S. 0. Bull. 18:33-35.
          •
Life Systems, Inc. 1982.  Surrogate species workshop. Workshop report.  Life
     Systems, Inc., Cleveland, OH,  U.S. Environmental Protection Agency
     Contract No. 68-01-6554, LSI TR-507-36B.

-------
 Macek,  K.  J.,  D.  F.  Walsh, J. W. Hogan, and D. D. Holz. 1972.  Toxicity of the
      insecticide Dursban* to fish and aquatic invertebrates in ponds.  Trans
      Amer. Fish.  Soc.  101(3):420-427.

 McKee,  J.  E.  and H.  W. Wolf. 1963.  Water quality criteria.  Calif. St. Wat.
      Qual. Contr. Board, Publ., No. 3-A.

 McKim,  J.  M.  1977.  Evaluation of tests with early life stages of fish for
      predicting long-term toxicity.  J. Fish. Res. Board Can. 34(8):1148-1154.

 Mount,  D.  Unpublished.  Aquatic surrogates.  Presented by J.  G. Eaton at the EPA,
      OTS Workshop on Surrogates Species in May,  1983 in Washington, D.C.  3 p.

 National Academy  of  Sciences. 1977.  Drinking water and health.  Nat. Acad.
      Sci., Washington, D.  C.

 O.E.C.D.,  Chemicals  Testing Programne Ecotoxicology Group.  1980.   Report on the
      assessment of potential environmental effects of chemicals:  The effects on
      organisms other than  man and on ecosystems.   Vol.  1.   O.E.C.D.,  Div.
      Technol.   Soc.  TNO Dep. Biol. Study and Infor.  Cent.  TNO Environ.  Res.,
      Delft, Netherlands.

 Oliver,  K.  H.,  G.  H. Parsons,  and C.  T.  Huffstetler.  1966.  An ecological  study
      on  the effects  of certain concentrations of  cacodylic  aid on selected  fauna
      and flora.   U.  S.  Air Force, Air Force Systems  Conmand,  Air  Proving Ground
      Center, Eglin Air Force Base, Florida.   25 p.

 Pimental,  D. 1971.   Ecological effects of pesticides  on non-target species.
      Exec. Off. President, Off.  Sci.  Technol.  220 p.

 Schmidt-Bleek,  F.  1980.  Regulation of chemicals within the European  Economic
      Community  and the  Organization for  Economic Cooperation  and  Development:
      Efforts for  international harmonization  in the control of  chemicals, p.
      56-63.  In;  Environmental risk assessment — An  integrated approach. First,
      Ann. Meet. Soc. Environ.  Toxicol. Chem., November  23-25,  1980.

 Shibko,  S.  1981.  Memorandum to M.  L. Dourson, U.S. E.P.A., Cincinati, Ohio,
      December 24,  1981.  Food  and Drug Admin., Washington, D. C.

Tatum, W. M. and R. D. Blackburn.  1962.  Preliminary study of effects of diquat
      on  the natural bottom fauna  and plankton in two subtropical ponds.  S. E.
      Assoc. Game Fish Comm. Ann.  Conf., Proc. 16:301-307.

Tucker,  R. K. and J. S. Leitzke.  1979.  Comparative toxicology of insecticides
      for vertebrate wildlife and  fish.  Pharmac. Theor. 6:167-220.

U.S.  Environmental Protection Agengy. Unpublished.  Environmental effects of
      regulatory concern under TSCA.  U. S. E.P.A., Off. Tox. Subst., Health and
     Environmental Review Division, Environmental Effects Branch.  December 2,
      1983.

U.S. Environmental Protection Agengy. Unpublished.  Testing triggers workshop:
     Workshop report.  U.S.  E.P.A., OTS, Workshop on Testing Triggers held in
     July and September, 1982 Washington, D.C.  TR-507-47C.  Project 1247 for
     Contract No. 68-01-6554, Work Assignment NO. 05.

-------
U'.S. Environmental Protection Agengy. 1980.  Guidelines and methodology used in
     the preparation of health effects assessment chapters of the consent decree
     water quality criteria.  Fed. Reg. 45:79347-79357.

Veith, G. P., D. J. Call, and L. T. Brooks. 1983.  Structure-toxicity relation-
     ships for the fathead minnow, Pimephales promelas; narcotic industrial
     chemicals.  Can. J. Fish. Aquat. Sci. 40:743-748.

Vettorazzi, G. 1976.  Safety factors and their application in the toxicological
     evaluation, p. 207-223.  In: The evaluation of toxicological data for the
     protection of public health.  Pergamon Press, Oxford.

Walker, C. R. 1962.  Toxicological effects of herbicides on the fish environ-
     ment.  Ann. Air Wat. Poll. Conf., Proc. 8:17-34.

Wellborn, T. L., Jr. 1969.  The toxicity of nine therapeutic and herbicidal
     compounds to striped bass.  Progr. Fish-Cult. 31(l):27-32.

-------