United States           Office of Water        EPA-822-R-00-009
          Environmental Protection       Office of Science and      August 2000
          Agency	Technology (4304)	

-------
                                  CONTENTS

CANCER/NONCANCER WORKGROUP RESPONSES                           1

1.  Response to the Workgroup's Technical Charge Comments and
   Recommendations on Cancer (5.2.1)	  1
   Issue 1.  New approaches to dose-response assessment and modeling  	  1

2.  Response to the Workgroup's Technical Charge Comments and
   Recommendations on Inconsistencies Between Cancer and Noncancer
   Methodologies (5.2.2)                                                        1
   1. Application of an RfD range rather than default point-estimate 	  1
   2. Separate methods for noncarcinogens versus carcinogens based on a nonlinear low-
dose extrapolation	  2
   3. Proposed effects on AWQC	  2
   4. Consideration of noningestive exposures 	  3
   5. Human interindividual variability	  3
   6. Methods for noncarcinogens and carcinogens based on a nonlinear low-dose
extrapolation do not estimate risk	  3

3.  Response to the Workgroup's Technical Charge Comments and
   Recommendations on Noncancer (5.2.3)	  4
   Issue 1. Application of an RfD range rather than default point-estimate  	  4
   Issue 2	  4
       Severity of effect 	  4
       Less-than-90-day studies	  4
       PBPK modeling	  5
   Issue 3.  Reproductive/developmental, immunotoxicity, and neurotoxicity data	  5
   Issue 4.  Nonthreshold mode of action	  5
   Issue 5.  Guidance on noncancer methods	  5

EXPOSURE ASSESSMENT WORKGROUP RESPONSES                        7

1.  Response to the Workgroup's General Comments (6.1)  	  7

2.  Response to the Workgroup's Technical Charge Comments and
   Recommendations (6.2) 	  7
   Issue 1. Inhalation ane dermal exposures	  7
       Relative source contribution (RSC) recommendations	  8
   Issue 2	  9
       Use of the USDA data	  10
       Consumption among minority populations	  11
       Species designation	  12
       Use of cooked versus uncooked data	  13
   Issue 3. Separate intake and body weight assumptions 	  14

-------
3.  Response to Other Issues Addressed by the Workgroup (6.3)	  15
    1. Monte Carlo and other statistical techniques	  15
    2. Policy regarding incremental risk	  16
    3. Federal Register guidance document equations	  17

4.  Response to Issues Considered by All Peer Reviewers (6.4)	  18
    1. Procedures for States and Tribes	  18
    2. Method to aggregate exposure from various sources  	  18
    3. Population (subgroup or percentile) being protected	  20

5.  Response to Three Additional Related Issues (6.5)                              23
    1. Use of reliable/empirical/adequate data  	  23
    2. Encouraging State/Tribal risk assessments	  24
    3. Risks to individuals and populations	  24

References 	  25

BIOACCUMULATION WORKGROUP RESPONSES                             26

1.   Response to the Workgroup's Technical Charge Comments and
    Recommendations (7.1) 	  26
    General issue	  26
       General comments	  26
       Document readability	  28
       Scale of application	  29
    Issue 1	  29
       Default lipidvalue	  30
       Freely dissolved fraction	  34
       Food chain multipliers	  36
    Issue 2. Metabolism  	  39
    Issue 3. Alternative models 	  41
    Issue 4. 1980 versus new methodology	  42

2.   Response to Issues for Public Comment Listed in the Federal Register (7.2)  ...  44
    Issue 1. Tiered hierarchy	  44
    Issue 2. National default lipid value 	  47
    Issue 3. Freely dissolved fraction equation	  48
    Issue 4. Default DOC and POC values  	  48
    Issue 5. Metabolism  	  49
    Issue 6. Models and FCMs   	  49
    Issue 7. Selecting Kow values	  49
    Issue 8. Field-based FCMs	  50

References 	  50
                                         111

-------
               CANCER/NONCANCER WORKGROUP RESPONSES

1.  Response to the Workgroup's Technical Charge Comments and Recommendations
    on Cancer (5.2.1)

    Issue 1: EPA has presented a detailed discussion concerning scientific issues
    associated with the cancer risk assessment methodology and its intentions for
    incorporating the Agency's new Proposed Guidelines for Carcinogen Risk Assessment
    (1996). Specifically EPA requests comment on applying the new approaches to dose-
    response assessment and modeling to its water quality criteria program.

The workgroup supported the EPA proposal that risks in the range of 10"6 are appropriate for
the average person (general population) at risk and 10"4 for a person who is highly exposed
because of specific exposure circumstances.  One workgroup member also wanted EPA to
present ambient water quality criteria (AWQC) at risks of 10"4, 10"5, and 1CT6 instead of a
single value based on  1CT6 risk. EPA's response to this issue is discussed in the section on
Exposure Assessment Workgroup Responses (in subsection 4.3, beginning near the bottom of
page 21).

The workgroup recommended that EPA use the same method for carcinogens based on a
nonlinear low-dose extrapolation and noncarcinogens and combine Equations ID-1 and ID-2 for
the reference dose (RfD) and point of departure/safety factor (PdP/SF) on p. 50 in the Federal
Register notice. EPA agrees that Equations ID-1 and ID-2 are operationally similar equations.
However, it is better to keep the two operations separate during analysis because PdP/SF may
differ from RfD.  The  RfD, by definition, should be from the most sensitive noncancer endpoint.

The workgroup wanted to see more criteria (for using the linear versus the nonlinear equations)
adequately spelled out in the Federal Register notice and the Technical Support Document
(TSD).  EPA's response to this comment is stated above in the first paragraph under Issue 1.

2.  Response to the Workgroup's Technical Charge Comments and Recommendations
    on Inconsistencies Between Cancer and Noncancer Methodologies (5.2.2)

    1.  Application of an RfD range rather than default point-estimate

The workgroup discussed the EPA proposal to allow the use of a point within an RfD range as
the basis for deriving water quality criteria rather than the single point default estimate of the RfD.
The workgroup thought that the mechanism suggested for selecting the range from which an
alternate to the default RfD could be selected was not scientifically justified.  They were also not
certain that the flexibility offered by this option would be useful to the risk assessor.

EPA agrees that the log-based apportionment for the range from which an RfD other than the
calculated RfD can be chosen is not based on specific data. It is simply a partitioning of an
order of magnitude into equal segments on either side of the calculated RfD.  It is important to
note that the uncertainty range about the calculated RfD establishes a domain from which a risk

                                          1

-------
assessor can select a single point to use as the alternate RED under defined circumstances.  One
example of a situation where a point other than the calculated RfD might be applied would be
that where there is a difference in the bioavailability of the contaminant in the water component
of the AWQC as opposed to the fish component. In such an instance, the decreased
bioavailability from fish tissues could be used to support selection of an RfD value greater than
the calculated value if the critical study was one that administered the contaminant in drinking
water.

Because the methodology says that a point within the range is selected when the uncertainty
factor (UF) is 100 or greater and the range is either a quarter or half log unit to either side of the
calculated RfD, it offers some flexibility for site-specific or contaminant-specific situations but
remains protective of public health.

    2.  Separate (yet effectively identical) methods for noncarcinogem versus carcinogens
       based on a nonlinear low-dose extrapolation

The workgroup made a number of comments supporting harmonization of the cancer and
noncancer methodologies, using benchmark modeling for all noncancer endpoints, and
expressing RfD values in terms of risk. This cluster of workgroup comments regarding EPA
methodologies for noncancer endpoints addressed issues about previously published, EPA-wide
guidelines for health risk assessment or current, active projects of the agency Risk Assessment
Forum to revise carcinogen risk assessment guidelines and to harmonize assessment methods for
noncancer and cancer endpoints.  The comments of the workgroup are noted and will be
considered in the context of revisions to our assessment guidelines. The human health
methodology is an application of broader EPA-wide guidelines. Revisions to fundamental
Agency guidelines are beyond the scope of this document.

    3.  Proposed effects on A WQC of incorporating physiologically based
       pharmacokinetic (PBPK) modeling for noncarcinogens vs. (linear or nonlinear
       low-dose extrapolated) carcinogens

As part of a their request that EPA harmonize the cancer and noncancer approaches, the
workgroup supported viewing the interspecies UFs as being made up of a pharmacokinetic and
pharmacodynamic component.

We agree that both toxicokinetics and toxicodynamic differences contribute to the difference in
response between animals and humans. We also accept that there are differences in opinion as
to the magnitude of the UF for toxicokinetics and that for toxicodynamics.

The Peer Review Report states the Health Risk Assessment Committee (HRAC) conclusion that
the toxicodynamic factor should be independent of the toxicokinetic factor and "be about the
same size as the correction factor used to adjust for interspecies toxicokinetic differences."

On the default UF for extrapolation from animal  dose to human equivalent dose, the workgroup
recommended a "unified approach" for both carcinogens (linear and nonlinear approaches) and

-------
noncarcinogens.  The suggested approach uses two factors: a scaling factor, adjusting for
toxicokinetics (UFA.PK) and a UFA.PD, adjusting for toxicodynamics. Thus, the UF becomes a
multiplication of these two factors [i.e., UF = (UFA.PK)(UFA.PD)]. For carcinogens, the
workgroup agreed with the use of 0.75-power of body weight to scale for toxicokinetics
between species in the absence of adequate PBPK data/models. For noncarcinogens, assigning
a 3 to both factors as proposed in the noncancer section of the human health methodology
preserves the default interspecies UF of 10 and gives each factor equal weight. However, the
workgroup recommended that UFA.PD (3) also be applied to carcinogens based on linear low-
dose extrapolation.  This will be specifically addressed in the final carcinogen assessment
guidelines. EPA has a separate ongoing effort to harmonize assessment approaches for differing
endpoints, and the workgroup comments will be considered in that effort. The human health
methodology is an application of broader EPA-wide guidelines. Revisions to fundamental
Agency guidelines are beyond the scope of this  document.

   4.  Consideration ofnon-ingestive exposures for noncarcinogens and carcinogens
       based on nonlinear low-dose extrapolation versus linear low-dose extrapolation.

The workgroup recommended using relative source contribution (RSC) for both nonlinear and
linear low-dose extrapolated carcinogens.  The Agency does not consider it necessary to apply
another factor such as RSC for carcinogens based on linear low-dose extrapolation because of
the conservatism built in a linear nonthreshold model. For carcinogens based on linear low-dose
extrapolation, the method does not assume any  threshold at low doses, whereas for
noncarcinogens or carcinogens based on nonlinear low-dose extrapolation, a threshold is
assumed below which there is no risk.

   5.  In contrast to methods for noncarcinogens and carcinogens based on nonlinear
       low-dose extrapolation, the methods proposed for carcinogens based on linear
       low-dose extrapolation do not consider human interindividual variability.

The workgroup would like EPA to consider another UF for interindividual variability for
carcinogens based on linear low-dose extrapolation.  EPA agrees with the National Research
Council's recommendation (NRC, 1994) that "the conservatism inherent in a linear-no-threshold
model obviates the need for any explicit consideration of interindividual variability in human
susceptibility to environmentally induced cancer." This comment by NRC was with respect to
general population exposure.  The NRC recommended possible consideration of an extra factor
for interindividual variability when assessing risk to a special population such as one exposed to
a fenceline risk from a dispersive source of air toxics. The present methodology is more
applicable to the general population., Even though an additional factor is not considered
necessary generally, specific data on sensitive subpopulations such as children, who may be
particularly sensitive to a specific chemical, will be considered in risk decisions.

   6.  In contrast to methods for carcinogens based on linear low-dose extrapolation, the
       methods proposed for noncarcinogens and carcinogens based on nonlinear low-
       dose extrapolation do not estimate risk.

-------
The workgroup stated that the method proposed for noncarcinogens and carcinogens based on
nonlinear low-dose extrapolation does not estimate risk. They suggested applying a modified
benchmark procedure for all carcinogens based on nonlinear low-dose extrapolation and for
noncarcinogens.  The benchmark dose would be associated with a specific risk of an explicitly
defined adverse response.  Thus, a corresponding distribution of risk could always be estimated
for any specified distribution of actual or potential environmental exposures to noncarcinogens
and/or carcinogens based on nonlinear low-dose extrapolation. The issue is related to Agency
policy on linearity/nonlinearity and dose response, which will be addressed in the final guidelines
for carcinogen risk assessment. See our discussion above in the first paragraph under Section 1,
Issue 1.

3.  Response to the Workgroup's Technical Charge Comments and Recommendations
    on Noncancer (5.2.3)

    Issue 1:  The use of a point within the RfD range for deriving water quality criteria,
    rather than a single point default estimate,  and the factors for determining its
   justification.
This charge to the workgroup related to the option for using a point within a range about the
calculated RfD as an alternative for risk assessment when there was adequate justification for the
alternate RfD. The workgroup's sentiments on regarding this issue and the EPA response are
presented above under Section 2, item 1.

    Issue 2: Incorporating information on severity of effect, less-than-90-day studies, and
    Physiologically-Based Pharmacokinetic data into derivation ofRfDs.

    Severity of Effect

The workgroup thought that the state of the science does not support a quantitative adjustment
for severity of effect in the development of an RfD. They stated that it is not simple to determine
whether adverse effects are mild or moderate and that it is often not possible to determine
whether effects are reversible or irreversible from a less-than-90-day study. EPA agrees that it
is difficult to quantify severity of effect in risk assessment. However, when the mode of action is
known and a sequence of precursor events is well established, it may be possible to establish a
quantitative relationship between a dose for a precursor event and the adverse effects and, thus,
quantify the severity of the precursor event when it is used as the point of departure.

    Less-Than-90-Day Studies

The workgroup did not support using less than 90-day studies for derivation of an RfD except
under unusual circumstances.  In a special case where a less-than-90-day study was used, the
workgroup stated that an additional UF of 10 should be added to the RfD calculation. The
Agency agrees with the workgroup's suggestion that studies of less-than-90-day duration be
used in the derivation of the RfD only if the reason for doing so it is carefully explained. We

-------
disagree with the suggestion that an additional UF of 10 always be used when an RED is based
on a less-than-90-day study. If the RfD is based on an acute effect and is simultaneously
protective against chronic effects, it is appropriate to use an acute study and not apply an added
10-fold UF.  If the contaminant is a nutrient, a 10-fold UF applied to a NOAEL from a short-
term human or animal study solely because of the duration of the study is often not appropriate.
    PBPK Modeling

The workgroup supported the use of physiologically based pharmacokinetic modeling but
pointed out that PBPK modeling does not account for differences in pharmacodynamics
between species.  The human health methodology supports the use of PBPK modeling and is in
agreement with the workgroup that pharmacodynamics as well as pharmacokinetics must be
considered in calculating the RfD.

    Issue 3: The use of reproductive/developmental,  immunotoxicity, and neurotoxicity
    data as the basis for deriving RfDs.

EPA asked the peer reviewers if they believed it was appropriate to use reproductive/
developmental, immunotoxicity and/or neurotoxicity data as a basis for deriving an RfD.  The
workgroup responded that all relevant lexicological data should be considered in the RfD

derivation process. They agreed with EPA's concern that some immunological data are difficult
to utilize in RfD development.

We agree with the workgroup's recommendation that it can be appropriate to use
reproductive/developmental, immunotoxicity, and neurotoxicity data as the basis for deriving
RfDs and concur with their suggestions regarding the vagaries of using certain immunotoxicity
endpoints in assessment.

    Issue 4: Case-by-case consideration of a nonthreshold mode of action for certain
    chemicals that cause noncancer effects when deriving RfDs.

The workgroup agreed that in some cases a nonthreshold mode of action is appropriate for a
noncarcinogen. However, they said that the example of nickel used in the human health
methodology was not appropriate, even for a sensitized person. They thought lead would be a
better example. We accept this recommendation that lead is a better example than nickel of a
noncarcinogen for which a nonthreshold approach risk assessment may be appropriate.

    Issue 5: Whether EPA should develop guidance for when to use each noncancer
    method (i.e., NOAEL, Benchmark Dose,  Categorical Regression).

The workgroup supported the development of guidance on when to select the
NOAEL/LOAEL, benchmark or categorical  regression methodologies for RfD development.
They stated that the guidance document should address the following questions:

-------
•   When are the data sufficient for modeling?
•   Where can the data be found?
•   When do the available data not support the development of an RfD?

Peer reviewers also expressed a strong preference for the benchmark methodology over the
categorical regression methodology.

Workgroup support for developing guidance for users of the AWQHH methodology regarding
the selection of an approach for derivation of a RfD for a noncarcinogen is appreciated.
Experience with the benchmark dose approach is growing, and the Agency is preparing a
guidance document for the methodology including a discussion on the data sets that are best
suited to such an analysis.  The guidance document is currently being reviewed within EPA.  It is
hoped that a similar document will be developed for the categorical regression methodology.

-------
              EXPOSURE ASSESSMENT WORKGROUP RESPONSES

1.  Response to the Workgroup's General Comments (6.1)

EPA acknowledges the workgroup's approval of certain features of the methodology revisions.
Specifically, we acknowledge the following:  endorsement of the Agency specifically indicating
where decisions or guidance is based in science, science policy, risk management, or, perhaps,
some combination of these; the flexibility offered to States and Tribes for deriving more site-
specific criteria with their water quality standards programs; and the use of examples in the TSD.
Our intention is to make both the final methodology guidance and the TSD as clear and useful as
possible for State and Tribal programs.  To this end, we will continue to identify areas where
discussion of science/policy/management issues, appropriate flexibility, and the inclusion of more
examples will enable better understanding and greater utilization by all States and Tribes.

Regarding the points made in the first bullet of section 6.1 (p. 6-1) of the Workshop Summary
Report, we also intend to expand on the discussion of inhalation and dermal exposures. We will
cross-reference existing Agency guidelines and known State guidance documents on assessing
exposures from inhalation and dermal exposures. Additionally, we plan on refining the
methodology, at least to incorporate information summarizing EPA's own guidance and example
assessments in the TSD that account for inhalation and dermal exposures. We also wish to
respond to the following workgroup comment:

    EPA [has]  finally abandoned the idea that a number can be developed from a small
    data set and used throughout different regions. EPA now has much more data.

EPA is required under the Clean Water Act (CWA), § 304(a), to develop national default
criteria that States and Tribes may use as guidance to establish water quality standards. As
such, the numerical criteria values that we develop and revise remain potentially applicable to the
nation. We encourage States and Tribes to use the methodology to develop criteria based on
local/regional information and believe that criteria reflecting such local conditions are desirable.

2.  Response to the Workgroup's Technical Charge Comments and Recommendations
    (6.2)

    Issue 1: The appropriateness of including inhalation and dermal exposures when
    deriving criteria and how they should be estimated

As indicated above, EPA intends to address inhalation and dermal exposures in greater detail.
We acknowledge that the potential for these exposures exists and that an approach to
accounting for them in the context of developing individual water quality criteria is appropriate.
In the short term, we will cross-reference existing Agency guidance and methods for inhalation
and dermal exposures. We will also consider the workgroup's recommendation for providing
more specific guidance on the relevance of these exposure routes to the ambient water quality
criteria in the form of future refinements of the methodology guidance.

-------
EPA acknowledges the workgroup's comments regarding State inhalation/dermal guidelines.
References to such documents (e.g., California) will be considered for inclusion in the final TSD.
We also appreciate the workgroup's concept of the TSD as a "living document" and, as stated
in the published draft methodology revisions (i.e., the Federal Register Notice), we anticipate
that our future role in the program includes refinement of the revised methodology.  Specifically,
we anticipate that as more current data and methods become available, these would be
incorporated into the methodology to reflect the latest science.

   Relative Source Contribution (RSQ Recommendations

The workgroup also provided specific recommendations on the concept of RSC, which they
linked to their discussion of inhalation and dermal exposures.  Although the workgroup  concurs
that the RSC concept is an important part of the AWQC-setting process, they recommend more
clarity and additional examples.  EPA acknowledges that additional discussion is needed to
clarify what constitutes the RSC, that is, clarifying language on what sources, routes, and
pathways of exposure will be specifically considered when setting a CWA 304(a) criterion, and
what data sources are appropriate. Regarding the workgroup's specific recommendations, we
offer the following responses.

•  EPA will add information either to the Decision Tree figure or to the text to be more explicit.

•  EPA will develop a more detailed TSD that offers additional examples in order to provide
   further clarification on how the RSC method works; this would include addressing the Box
    15 allocations that the workgroup identified. This will likely take the form of an addendum
   (or followup document), given the current timeline for final publication of the Federal
   Register guidance.  We are committed to providing methodology guidance that will enable
   States and Tribes to derive site-specific criteria values, if they so choose, and will pursue this
   endeavor.

•  EPA is limited in its ability to coordinate the RSC process with other agencies. We have
   discussed our RSC policy with appropriate staff from USDA and FDA, and representatives
   from the FDA participated in the EPA workgroup that developed the Decision Tree
   approach. However, the specific requirements of the CWA and EPA's  particular
   approaches to conducting risk assessments and deriving protective water quality criteria may
   vary substantially from the legal requirements, science policies, and risk management
   decisions made by various other agencies for vastly different program goals.  Therefore, the
   RSC process is likely not the same as that used by other agencies. EPA has coordinated
   with USDA and FDA regarding the use of data relevant to the exposure assessments (e.g.,
   food consumption data, contaminant monitoring data), and will  continue  to do so in  the
   future.

•  It is not clear what the workgroup meant by recommending that "the RSC for non-cancer is
   too vague" (see p.  6-2 of the report). If this refers to the discussion of its application and
   what constitutes the RSC (as indicated above), EPA will work to improve the clarity of this
   discussion.  If this  statement is related to the distinction between carcinogens based  on linear

-------
    low-dose extrapolation and noncarcinogens (as is the statement that follows it), we reiterate
    our policy on the distinction here.  Specifically, different approaches for addressing nonwater
    exposure pathways are used in setting AWQC for the protection of human health depending
    on the toxicological endpoint of concern. For those that are considered carcinogens based
    on linear low-dose extrapolation, only drinking water consumption and fish ingestion are
    accounted for in the derivation of the AWQC. The RSC is not applied to nonwater sources
    because, for these chemicals, the AWQC are being determined with respect to acceptable
    incremental lifetime risk posed by a substance's presence in water, given that the estimates
    are considered upper-bound on potential risk, and are not being set with regard to an
    individual's overall cancer risk from all sources of exposure. For carcinogens with a mode of
    action indicating nonlinearity or for a noncancer endpoint where a threshold is assumed to
    exist, nonwater exposures are accounted for when deriving the AWQC. The rationale for
    this approach has been that for pollutants exhibiting threshold effects, the objective of the
    AWQC is to ensure that an individual's total exposure does not exceed that threshold level.

    Furthermore, health-based, medium-specific criteria values based on linear low-dose
    extrapolation typically vary from other medium-specific values in terms of the concentration
    value, and often the associated risk level. Therefore, the RSC concept could not even
    theoretically apply unless all risk assessments for a particular carcinogen based on linear
    low-dose extrapolation resulted in the same concentration value and same risk level; that is,
    an apportionment would need to be based on a single concentration value and risk level.

The workgroup expressed curiosity about RSC and other EPA programs (the Safe Drinking
Water Act [SDWA] and the Food Quality Protection Act [FQPA] were specifically
mentioned).  EPA explicitly stated in the Federal Register Notice on these draft revisions that it
believes, for a given pollutant, the drinking water component of an AWQC should be consistent
with the Maximum Contaminant Level Goal (MCLG) established under SDWA. We therefore
propose to use similar assessment methodologies for deriving AWQC and MCLGs.  The EPA
Office of Water (OW) has been working with the Office of Pesticide Programs regarding their
implementation of the FQPA, in order to share information on  how the two offices approach
addressing multiple exposure sources as part of their assessment programs (i.e., tolerance-
setting, health criteria). Additionally, OW has recently been working with the Office of Air
Quality Planning and Standards on issues related to aggregate exposure and cumulative risk.
With each of these efforts, the EPA offices are attempting to identify areas where common
policies and approaches may be appropriate.

Finally, the workgroup indicates support for use of an 80 percent ceiling with the RSC. EPA
acknowledges this support. In addition to the workgroup's understanding of the possibility that
"new exposures and situations will arise," we reiterate here that the ceiling also is intended to
provide adequate protection for those who experience exposures (from any or several sources)
higher than the available data indicate. For many of the chemical contaminants that we evaluate,
the data available are not extensive.

    Issue 2:  The use of the USD A survey data to choose estimates offish consumption
    among different population groups, in addition to decisions made on species

-------
    designations, cooked weight values, and potential cooking-related changes to the
    toxicants.
    Use of the USD A Data

EPA acknowledges the workgroup's support for the four-preference hierarchy. Regarding the
workgroup's suggestion to provide guidance on how to conduct a consumption survey (in
addition to guidance on analyzing the results), we have already done this. Specifically, the draft
methodology revisions, in the discussion of the first preference for using local data, reference
EPA's Guidance for Conducting Fish and Wildlife Consumption Surveys (EPA Report No.
EPA/823/B-98/007).

The workgroup questioned the use of short-term data for long-term fish consumption estimates.
Specifically, the workgroup stated that short-term data do not "capture 'chronic' usual intakes"
and are "not appropriate to use when estimating long term exposures."  The workgroup instead
recommended use of the Tuna Research Institute (TRI) data, from the EPA/ORD Exposure
Factors Handbook and estimates made in the Mercury Study Report to Congress (MSRC)
using food frequency data from the National Health and Nutrition Examination Survey
(NHANES HI). The TRI data the workgroup refers to is actually the National Purchase Diary
(NPD) study conducted more than 25 years ago.  The NPD is the basis of the 6.5 g/day default
value that EPA has historically used for fresh/estuarine fish consumption. At the  1992 national
workshop that EPA conducted, one of the initial components identified for revision was the fish
intake default rate.  At that time, many participants considered the 6.5 g/day value to be
inadequate and advocated the use of much more recent data.  Dietary information suggests that
consumption offish has increased since that time because of nutritional, cultural, and other
preferential choices, and EPA has endeavored to identify more recent survey data. We have
received consistently strong input from many of our stakeholders (including EPA Regions,
States, and Tribes) to this effect, urging an update. The workgroup's recommendation of the
NPD data somewhat contradicts their statement (see p. 6-3) that "estimates are poor when the
data are derived from older national surveys conducted for other purposes, but then adjusted to
derive . . . AWQC."

The MSRC states that it is "rarely possible to measure a large number of days of dietary intake
for individual subjects; consequently, a sample of one or several days is used to represent the
true intake (Willett, as cited in USEPA 1997)." The report emphasizes that these samples are
typically 24-hour recalls, 3-day recalls or records, or 7-day recalls or records. The MSRC
indicates that data from such studies provide reasonable (unbiased) estimates of mean intake,  but
that standard deviations can be greatly overestimated.  We reiterate here that the CSFII mean
values are not biased; specifically, the intra-individual variation does not bias estimates of the
mean intake of a population (Hegsted 1972).  The estimates of the upper percentiles of per
capita fish consumption based on 3 days of data may be biased upward, thereby  resulting in a
conservative estimate of risk.  However, the extent to which this is overestimated is not known.
We note that we did not exclusively analyze the CSFn data; rather the data were compared with

                                          10

-------
those of other studies (especially for recreational fisher and subsistence fisher estimates) that
support our decision. The MSRC inevitably relies on the CSFn data from USD A, along with
the NHANES HI estimates offish consumption patterns (from the early 1990s) for making
estimates on fish consumption in the general population.  The NPD data are also presented, for
comparison.

EPA believes that the CSFn data are adequately representative offish intake rates among the
general U.S. population for purposes of national criteria. Although the MSRC indicates the
potential for underestimating the extent offish consumption due to the 3-consecutive-day
limitation of the assessment, it states that the dietary recall/record assessment provides "more
precise estimates of the quantities offish consumed that [sic] would be obtained with a food
frequency record." As part  of the CSFII analysis, sampling weights were adjusted to account
for nonresponse and were subsequently reweighted using regression techniques that calibrated
the sample to match characteristics correlated with eating behavior.  States and Tribes are
encouraged to use local data on dietary preferences to establish criteria when national estimates
are not suitable.

The Exposure Factors Handbook indicates the advantages of the NPD in terms of its high
response rate, national representativeness, and consumption record  over a 30-day period.
However, according to the Handbook, the upper percentiles from the NPD data are (as is the
CSFn) likely to overestimate the corresponding upper percentiles of long-term intake (the same
is indicated for the standard deviation). According to the MSRC, there were other limitations in
the NPD. For example, the survey did not include data on the quantity offish represented by a
serving (or information to calculate actual consumption offish from numerous entries, e.g.,
breaded fish, fish mixed with other ingredients), and there may have  been underreporting over
time because of the survey diary completion requirements. Also, several studies indicate that the
quantities and types offish consumed have changed over the past 25 years.  Further,
comparisons between these  data and newer studies are not possible because of the unavailability
of the survey sample weights and participants' body weights.

Advantages of the CSFn, according to the Handbook, include its large sample size,
representativeness, and relative currency. The Handbook describes it as the "key study" for
estimating mean fish intake.  The Handbook does recommend the NPD data for use in
estimating long-term distributions; however, it actually recommends  adjustments to the data to
account for age of the data,  and it presents values from a study that did exactly that.  The CSFH
study, however, suggests even higher increases in fish consumption than the adjusted values
made on the NPD data.  EPA also believes that the 3-day CSFII data are superior to the
NHANES 1-day recall for characterizing fish consumption. Furthermore, the NHANES food
frequency information is not useful because it does not break out the data by habitat and species
(it is only divided into categories ofjinfish and shellfish)., which are needed to estimate
fresh/estuarine species intake. Given that the data are much more recent, the fact that the CSFII
describes a nationally representative sample of individuals, and the strong support to revise the
NPD-based default, EPA believes that the CSFH is the best source of current data available.
The current draft TSD identifies the NPD as the basis of the 6.5 g/day assumption. We will
consider including additional information on the NPD in the final TSD, as the workgroup

                                           11

-------
recommends.

    Consumption Among Minority Populations

The workgroup reports that EPA's analysis of the CSFII (1994-95 data), as it appears in the
MSRC, indicates that Asian populations do not consume greater amounts offish than other
minorities or whites. The workgroup specifically states that "although Asians consumed fish
more frequently, the amount eaten per serving was less than in other groups, thus the intake in
g/kg BW/day was less than that for other ethnic groups." Although this is true in terms of the
data described in the MSRC as "per user" (i.e., similar data that are described in the
methodology as "acute" consumption) and useful for an indication of meal size, the "per capita"
data indicate greater consumption for Asian and Pacific Islander groups at both median and
upper percentile values (see MSRC,  Volume IV, p. 4-82, Table 4-67). We believe that the per
capita rates are more appropriate to use for protection of human health from chronic exposures
(i.e., for chemical toxicants that are of chronic health concern).  Similarly, the MSRC analysis of
"month-long estimates" for both fish/shellfish consumption (presented in grams/day) and mercury
exposures (presented in • g/kg/day) based on NHANES data, indicate higher intakes of both for
the "other" ethnic/racial category than for either the "white/nonhispanic" or "black/nonhispanic"
categories (see MSRC, Volume IV,  p. 4-83, Tables 4-68a and b).

A point made by EPA in the draft methodology revisions was that local and regional studies exist
that indicate that Native American, Pacific Asian American, and subsistence population groups
may consume greater amounts than the general U.S. population. EPA recommended—and
continues to recommend—the use  of such studies where appropriate, as indicated by EPA's first
two preferences in the hierarchy. This idea was strongly supported by the experts from the
1992 national workshop.

    Species Designation

The workgroup stated that EPA's explanation of the species habitat designation for shrimp is not
correct. However, the workgroup  simply states that shrimp should be referred to as
"anadromous."  The term anadromous generally refers to a species that spawns in fresh water or
near-fresh water and then migrates into the ocean to grow to maturity, or to an ocean species
that similarly spawns in fresh or near-fresh waters.  The life cycles of anadromous  species vary in
terms of whether they remain in fresh or near-fresh waters until they die or whether they return to
ocean waters after spawning. As such, the description provided by EPA in the draft
methodology revisions is correct and  does not conflict with the term anadromous. EPA can add
the term to the discussion when finalizing the documents. However, regardless of their
anadromous status, shrimp  have been included in the default value (i.e., designated as a
fresh/estuarine species) because of their life cycle, as described in the draft TSD. The amount of
time that shrimp spend in near-shore and estuarine waters is substantial  enough to include them in
the default assumption, thereby accounting for their potential to contribute to health risks if
contaminated and, more importantly,  ensuring the AWQC are protective regarding their
consumption.
                                          12

-------
The workgroup also stated their disbelief that the "99th percentile of the salmon consumed is
marine." EPA clarifies here the context of its statement in the draft methodology revisions.  The
USDA food codes containing salmon do not indicate the source of the salmon (e.g., landlocked
freshwater, farm-raised, or wild).  We based our allocation of salmon between freshwater and
marine habitats on commercial landings data provided by the National Marine Fisheries Service
for the period 1989 to 1991. All landings of Pacific salmon, including chum, coho, king, pink, or
sockeye were assigned to the marine habitat. All landlocked Great Lakes salmon and farmed
salmon received the classification of freshwater.  The resulting apportionment for salmon was
1.18 percent to the freshwater habitat and 98.82 percent to the marine habitat.

Regarding the other species identified for designation by the workgroup (p. 6-4), EPA
acknowledges that a limited number of freshwater fish are listed.  The species listed directly
reflect the consumption of the CSFII survey respondents.  Therefore, the absence of striped
bass or crayfish is due to the fact that neither were reported consumed. However, we intend to
incorporate the CSFII data from the years 1994 through 1996, which will result in inclusion of
additional species.  We believe we have correctly apportioned all clam and oyster species to the
appropriate habitat categories (i.e., estuarine/marine and estuarine-only, respectively). The
workgroup believed that clams should be in freshwater and marine categories,  and that oysters
should be added to the marine category.  EPA not only believes that the estuarine/marine
allocation for clams is most accurate, but we also note that the non-marine designated species
are included in the default intake rate regardless of whether being called estuarine or freshwater.
For oysters, we are not aware of open-ocean harvesting and the designation of all oysters to the
estuarine habitat is a more protective exposure assumption. Oysters may be present in waters
outside of estuaries which are considered marine in terms of salinity, but these  are near-shore
waters to which water quality standards apply.

    Use of Cooked Versus Uncooked Data

The workgroup advocated using data on uncooked fish weights "as recommended in the
Exposure Factors Handbook''  Separately, the workgroup recommended the uncooked
weights "because of the bioaccumulation factor in the AWQC equations presented in the TSD
and the Federal Register.  Furthermore, chemical residue data are typically available for
uncooked fish." EPA understands that chemical residue data and field-measured BAFs are
usually described for uncooked fish and, thus, the uncooked fish weight is consistent with the fish
tissue bioaccumulation value.

EPA has considered the pros and  cons of using uncooked versus as consumed weights on
several levels. First, the intake parameters of the criteria derivation equation are intended to
capture ingestion—that is, what people actually consume and are exposed to. By and large,
people consume cooked fish, and  where raw shellfish or sushi were consumed by the CSFII
respondents, those intakes were included in the as consumed weights. This assumption is also
consistent with the dietary estimates based on prepared foods (not raw commodities) that are
made by both the EPA pesticide program and the FDA Total Diet Study program. We also
considered the "consistency"  issue in the context of the fact that the CSFII survey respondents
estimated the weight offish they had consumed. Similarly, the basis of EPA's  Great Lakes

                                           13

-------
Water Quality Guidance was a consumption survey of as consumed fish intakes.

Second, EPA considered the differences as discussed in the Exposure Factors
Handbook—that is, the possibility of overestimating consumption but underestimating dose if the
cooking process results in an increased concentration in the cooked fish (there is typically a
weight loss in cooking). However, the Handbook fails to consider the opposite, where chemical
concentration loss exceeds the loss offish weight when cooking. The latter has been shown with
chemicals that accumulate in fat tissue, as we discussed in the draft methodology revisions. As
we previously stated, there are comparatively few chemicals for which measurements are
available and the process is complicated further by the variability in parts of a fish where the
chemical may accumulate, the method of preparation, and how the cooking process may
transform the chemical. What is certain is that the mass of the contaminant will either remain
constant or be reduced. The resulting concentration is harder to predict. The Handbook stated
that it is "more conservative and appropriate to use uncooked fish intake rates." However, the
Handbook also stated that "if concentration data can be adjusted to account for changes after
cooking, then the  'as consumed' intake rates are appropriate." The Handbook presents both as
consumed and uncooked values "so that the assessor can choose the intake data that best
matches the concentration data that is being used."  [We recommended the use of as consumed
weights in the draft methodology revisions and an adjustment  of the bioaccumulation factor for
cooking loss, if information was available.  Otherwise, we recommended using the as consumed
weight along with the full bioaccumulation factor (unadjusted for cooking loss), which would
produce a slightly more stringent AWQC.]

Third, EPA has received input from its stakeholders regarding potential confusion over the fact
that uncooked weights are used in the Agency's fish advisory program and that having two sets
of values may prove confusing to States and Tribes, as well as the general public.  Furthermore,
the measures of a contaminant in fish tissue samples that would be applicable to either
compliance monitoring or the permitting program are related to the uncooked fish weights.

Therefore, EPA has reconsidered its position based on these facts in contrast to the fact that as
consumed values more accurately represent actual  intake. The approach of using an uncooked
weight in the calculation will result in a somewhat more stringent AWQC (studies indicate that,
typically, the weight loss in cooking is about 20 percent).  EPA will derive its national default
criteria on the uncooked weight fish intakes based on the input received, especially that from the
States over the potential for confusion with the fish advisory program. In addition, EPA will
provide guidance on site-specific modifications in its TSD volume on exposure assessment.
Specifically, EPA will  describe  an alternate approach, by  calculating the AWQC with the as
consumed weight —again, more directly associated with exposure and risk—and then adjusting
the value by the approximate 20 percent loss to an uncooked equivalent. Thus, the AWQC
conversion to an uncooked equivalent can be consistently used with State/Tribal standards
programs and still represent the same relative risk as the as consumed value. It is important to
understand that the two approaches will not result in the same AWQC value. Whereas the
second is more scientifically rigorous and, again, represents a  more direct translation of the as
consumed risk to the uncooked equivalent, it may be too intensive a process to expect of State
and Tribal organizations whose resources are  already constrained.

                                           14

-------
    Issue 3:  The use of separate intake and body weight assumptions (e.g., 17.80 g/day of
   fish and 70 kg body weight) versus assumptions that combine intake and body weight
    (e.g., 254.3 mgfish/kg body weight).

The workgroup recommended combining the intake and body weight parameters, "especially if
children are being evaluated," and modifying the equation to reflect this. The workgroup
provided no additional rationale or advice. Presumably, they believe that combining the two will
provide a more accurate estimate. When we presented the issue for review by the Agency's
Science Advisory Board (SAB), the board provided the following advice:

    In theory  it would be better to develop standards on a per kilogram body
    weight basis.  However, in practice the results are not different enough to
    make much difference in the magnitude of AWQC.  In particular, data should
    not be rejected because individual body weights are not available, and funds
    should not be allocated for collecting such data since no conceivable benefit
    would accrue.

EPA has also received input from its State stakeholders regarding potential  confusion over
combining the two parameters.  Most believe that the difference in accuracy is negligible but that
the difficulty in associating the units of mg/kg-BW/day with a meal size, especially for public
communication and understanding, is great and, therefore,  not particularly useful. Several
stakeholders believed that if the data were combined as part of a study, or if a strong,
demonstrated correlation between intake and body weight exists, the combined parameter
should be used. We are evaluating recent information on both drinking water intake and fish
intake from the 1994-1996 CSFn data and are assessing the differences between the two units
of measure—including an emphasis of the differences with finer age categories for children when
mg/kg-BW/day are used. [Note: SAB's comment on the unavailability of individual body
weights is not an issue with the CSFII; that is, this information is available.]

EPA intends to provide tables in the final exposure assessment TSD of all fish/population
categories for both grams/day and mg/kg-BW/day.  EPA also intends to derive its national
default criteria using grams/day (for fish) and L/day (for drinking water), along with a body
weight assumption, as recommended by the States. However, EPA will refine the exposure
assessment TSD to provide examples on how to derive criteria using either, including identifying
situations where the latter estimate would provide substantively more accurate estimates.

3.  Response to Other Issues Addressed by the Workgroup (6.3)

    1.  Monte Carlo and other statistical techniques should be used only if data support
       their  use.

EPA generally agrees with the workgroup's statements on the potential for use of statistical
methods in assessing exposure when deriving AWQC.  We intend to expand the discussion in
the TSD to provide additional guidance on the complexity and limitations of using Monte Carlo
and other techniques, and on the need for clear, scientifically defensible, and reproducible

                                          15

-------
analyses. Existing EPA documents will be relied upon and cited for the user's reference.  In the
context of exposure distributions, we will try to clarify how population segments can be
protected at desired levels (see discussion on p. 6-6 of the peer review report, for complete
comment).

We clarify here our position on two points from the workgroup's comments on this issue.  First,
the workgroup referred to using Monte Carlo to:

    give a clearer representation of the relationship of the conservative deterministic
    AWQC to the range of possible criteria that would be protective of various segments
    of the population.

Our inclusion of the discussion in the TSD addresses our potential use of probabilistic techniques
to estimate exposures when deriving EPA national default criteria. However, in terms of the risk
assessments, we derive criteria for the population most relevant to the lexicological basis of the
RfD or cancer assessment. By basing default criteria on this population group, we are confident
of protecting the overall population, especially given the conservative manner in which the
RfD/cancer assessment is derived. However, we will continue to rely on values approved by the
Agency (as published on IRIS) for the AWQC risk assessments and will not be publishing a
"range of possible criteria that would be protective of various segments of the population."

Second, the CWA requirements and the goals of the water quality criteria program do not make
the specific development of reasonable  maximum exposure (RME) or maximally exposed
individual (MET) descriptors useful, as the workgroup mentioned in their report (see discussion
on p. 6-8 of the peer review report, for complete comment).

    2.  The policy regarding incremental risk needs to be expanded (FR -pg. 163).

It is not clear what the workgroup meant by their comments on considering background risk.
Specifically, the workgroup stated (referring to the Federal Register Notice, p.  163, not the
TSD):

    In the  context of the TSD, background risk is not considered (i.e., only incremental
    risk is considered). However, background risk is considered in other documents
    when discussing drinking water. An explanation addressing why background risk
    is not considered should be provided.

The page cited from the Federal Register discussion refers to consideration of nonwater
sources of exposure (e.g., diet, air) when setting AWQC—that is, background
exposures—which the workgroup may be describing as background risk. The distinction we
made was between chemical substances where the toxic  endpoint was carcinogenicity based on
linear low-dose extrapolation versus a nonlinear-extrapolated endpoint. The distinction is  as
follows: (1) For chemical substances where the toxicity basis is that of carcinogenicity based on
nonlinear low-dose extrapolation or a noncancer endpoint and a threshold is assumed to exist,
the resulting numerical value is thought to be a level below which the adverse effect (i.e., the

                                           16

-------
effect the dose is based on) will not occur. Therefore, EPA will account for other common
sources for the population being targeted in order to ensure that an individual's total exposure
does not exceed that threshold level. (2) For chemicals that do exhibit carcinogenicity based on
linear low-dose extrapolation, the AWQC are being set on the basis of the chemical substance's
presence in the water. Nonwater sources are not considered because the criteria are protecting
only the incremental lifetime risk posed by the chemical from this specific source, and are not
being set with regard to an individual's total risk (of the chemical's linear-extrapolated
carcinogenicity) from all exposure sources.

Contrary to the workgroup's statement in the workshop report, the drinking water program at
EPA has followed the same approach—accounting for other exposure sources (by applying a
relative source contribution factor) has not been done with carcinogens in the past, whereas
accounting for nonwater exposures for noncarcinogens (most  often a default value) has routinely
been done. It is not clear to what "other documents" the workgroup is referring.

Regarding the workgroup's comments on using other sources of lexicological data, EPA has
primarily relied in the past and continues to rely on the consensus values in the IRIS database for
its risk assessment information. We believe it is acceptable for States and Tribes to use
lexicological data and risk assessments outside of the IRIS database as long as the information
and/or assessment has been externally peer reviewed and is either published or otherwise
available to the public.  As stated in the Agency's Peer Review Handbook, EPA policy is to
peer review scientifically and technically based products  that  are used to support EPA decisions
(U.S. EPA 1998).  Therefore, we recommend that States and Tribes follow this same approach
when using lexicological data outside of IRIS to ensure that the resulting risk assessments are
scientifically defensible.

    3.  Federal Register Guidance Document Equations [Note:  List of toxicants and
       populations protected also discussed.]

EPA acknowledges the workgroup's suggestion to provide the AWQC equations in their most
complex forms (i.e., "the level of detail provided in Equation  7.1.1"). As stated in the Federal
Register, the "generalized" equations were presented to simplify understanding for the reader,
with a footnote explaining the trophic level breakouts and where they appear in the documents.
EPA will revise the methodology to explain the more complex forms at the first point where the
criteria equations discussion appears.

The workgroup's comments on presentation of intakes based  on mg/kg and modifying the
equations to address different population groups (e.g., pregnant women and children) are
addressed in the response to combining intake assumptions in Section 2, Issue 3,  and the
discussion on "various segments of the population" in Section 3.1, respectively. Apparently,  one
panelist described doing "an analysis in two different ways (e.g., benzene) and then choose the
most appropriate one."  This  panelist presumably refers to conducting various exposure
scenarios for different target populations and basing each criterion on the population at greatest
risk. We agree with this, in principle, and are open to developing multiple estimates, where
appropriate. However, as stated in  Section 3.1, above, we derive criteria for the population

                                           17

-------
most relevant to the most sensitive lexicological endpoint and are, therefore, generally confident
that the criteria are protective of the overall population.  The workgroup also commented on
specifying the population to protect and referred to EPA's assessment for lead in which EPA
"wanted 95 percent of the population to have lead levels below a specific level." For a response
to this issue, refer to 4.3, below.

The workgroup stated, "it would be helpful to have the list of toxins that occur in accumulated
fish tissues and information was solicited for such an open-ended list." We have not currently
compiled such a specific list. We listed 29 chemicals that we ranked highest priority for AWQC
revisions from a larger list (that, by and large, also comprised chemicals with existing criteria) in
terms of toxicity, occurrence data in fish tissue and sediments, and BAF values from the Great
Lakes Initiative. Additionally, we stated that the Agency welcomed suggestions from the public
at any time.  If the workgroup was simply expressing their desire to obtain a compiled list as a
reference, we will consider developing a list, available to the public, after the methodology is final
and the overall state of the science predicting bioaccumulation merits such a list.

4. Response to Issues Considered by All Peer Reviewers (6.4)

    1.  EPA needs to provide procedures for the States and Tribes to create water quality
       standards that do not require Federal resources or are not impeded by Federal
       constraints.

States and Tribes  are not impeded by any  constraints in Federal resources. EPA encourages
States and Tribes to develop their own AWQC to reflect local and regional conditions.  This is
reiterated in the draft methodology revisions (Appendix n (C)) and other policy and guidance
documents related to the development of water quality standards. (See the Water Quality
Standards Handbook, Advance Notice of Proposed Rulemaking at 63 FR 36741.)

States and Tribes are also encouraged to use their own data in the development or refinement of
their criteria, whether through the EPA methodology or through other scientifically defensible
methods as specified in 40 CFR 131.1 l(b). If a State or Tribe does not have an alternative
methodology it wishes to use,  components within the draft methodology may be refined based
on site-specific information,  such as lifetime cancer risk or fish consumption values. Where the
State or Tribe chooses not to refine AWQC based on local or regional  conditions, EPA
publishes 304(a) criteria as recommendations for States and Tribes to use when adopting water
quality criteria and for use when it becomes necessary for us to promulgate replacement Federal
standards under CWA  §303(c).

The revised human health methodology establishes a scientifically defensible approach to
deriving §304(a) criteria for  the protection of human health. This methodology may also be used
by States and Tribes in the development of their own criteria based on their own data. A State
or Tribe is not required  to use this methodology or to adopt EPA's recommended criteria if they
are able to develop alternative criteria based on scientifically defensible  methods. EPA is not
required to develop additional methodologies and believes that this methodology provides
sufficient detail for States and Tribes to use in the development of criteria for local or regional

                                           18

-------
conditions that may occur for waters under their jurisdiction.

    2.  Method to aggregate exposure from various sources (FR -pg. 180) — The RSC
       approach vs. route-specific margin of exposure approaches.

EPA acknowledges the workgroup's general support of the RSC approach and addressing
aggregate exposures.  EPA also acknowledges the workgroup's comments on the fact that
cumulative risks are not assessed in the derivation of AWQC.  The workgroup specifically
expressed concern that

    The TSD does not address risks from multiple chemicals (or other threats for
    that matter). This situation arises when the population being protected has risks
    from exposure to chemicals not addressed in the AWQC criteria and standards,
    pathogens, air emissions, etc.  EPA could address the way in which these other
    risks are taken into account, or explain that it cannot factor those risks in at present.

Assuming that all multiple exposures from multiple chemicals are additive is scientifically sound if
they exhibit the same toxic endpoints and modes of action. We are very much aware of the
complex issues and implications of cumulative risk and are developing an overall approach at the
Agencywide level. Numerous publications relevant to cumulative risk can assist States and
Tribes in understanding the complex issues associated with cumulative risk.  These include the
following:

Durkin PR, Hertzberg RC, Stiteler W, Mumtaz M. 1995. The identification and testing of
interaction patterns. Toxicol Lett 79:251-264.

Hertzberg RC, Rice G, Teuschler LK. 1999. Methods for health risk assessment of combustion
mixtures. In: Roberts S, Teaf C, Bean J, eds. Hazardous waste incineration: evaluating the
human health and environmental risks. Boca Raton: CRC Press LLC, pp. 105-148.

Rice G, Swartout J, Brady-Roberts E, Reisman D, Mahaffey K, Lyon B. 1999.
Characterization of risks posed by combustor emissions. Drug Chem Toxicol 22(1)221-240.

U.S. Environmental Protection Agency. 1999. Guidance for conducting health risk assessment of
chemical mixtures.  Final draft. Risk Assessment Forum Technical Workgroup. September.
NCEA-C-0148. www.epa.gov/ncea/raf/rafpub.html.

U.S. Environmental Protection  Agency. 1998. Methodology for assessing health risks associated
with multiple pathways of exposure to combustor emissions. EPA/600/R-98/137. (Update to
EPA/600/6-90/003, Methodology for assessing health risks associated with indirect exposure to
combustor emissions), http://www.epa.gov/ncea/combust.html.

U.S. Environmental Protection Agency.  1996. PCBs: cancer dose-response assessment and
application to environmental mixtures. National Center for Environmental Assessment,
Washington, DC.  EPA/600/P-96/001F.

                                          19

-------
U.S. Environmental Protection Agency. 1993. Review draft addendum to the methodology for
assessing health risks associated with indirect exposure to combustor emissions. Office of Health
and Environmental Assessment. Office of Research and Development, Washington, DC.
November 10. EPA/600/AP-93/003.

U.S. Environmental Protection Agency. 1993. Provisional guidance for quantitative risk
assessment of poly cyclic aromatic hydrocarbons. Office of Research and Development,
Washington, DC. July. EPA/600/R-93/089.

U.S. Environmental Protection Agency. 1990. Technical support document on health risk
assessment of chemical mixtures. Office of Research and Development, Washington, DC.
August. EPA/600/8-90/064.

U.S. Environmental Protection Agency. 1989a. Risk assessment guidance for Superfund, vol.
1.  Human health evaluation manual (part A). EPA/540/1-89/002.

U.S. Environmental Protection Agency. 1989b. Interim procedures for estimating risks
associated with exposures to mixtures of chlorinated dibenzo-p-dioxins and -dibenzofurans
(CDDs and CDFs) and 1989 update. Risk Assessment Forum. March.  EPA/625/3-89/016.

U.S. Environmental Protection Agency. 1986. Guidelines for the health risk assessment of
chemical mixtures. Risk Assessment Forum, Office of Research and Development, Washington,
DC. September.  EPA/630/R-98/002.

The Agency's program offices (including OW and OPP) are also engaged in ongoing discussions
on how to sort through the great complexities, methodological challenges,  data adequacy needs
and other information gaps, as well as the science policy and risk management decisions that will
need to be made, as they pursue developing a sound strategy and, eventually, specific guidance
for addressing cumulative risks. Additionally, OPP has factored cumulative risk into a recent
assessment of several pesticides determined to have the same mode of action (i.e., triazine
pesticides). Unfortunately, the workgroup stated that they had no specific  suggestions on how
cumulative risk should be factored into the derivation of AWQC. EPA can add a discussion
about the concept of cumulative risk and the inadequate state of the science when finalizing the
methodology documents. As a matter of internal policy, EPA is committed to refining the
methodology as advances in relevant aspects of the science improve, as has been previously
indicated.

The workgroup commented that EPA's RSC approach does not account for "effects  that are
specific to the route of exposure." The draft Federal Register language (p. 180) discussed
inclusion of inhalation and ingestion exposures, and accounting for them either as part  of the RSC
or by using the RED along with the RfC in determining an acceptable hazard index. EPA also
discussed differences in bioavailability and absorption rates, including recommendations for
situations where data exist and where they do not.  We will expand this discussion and our
position on route-specific differences in exposure, in a future addendum to the TSD.  We
acknowledge the workgroup's comment on the simplicity  of the RSC approach versus the more

                                         20

-------
complicated but accurate approach of using both reference values (RfD and RfC). We also
intend to expand the discussion on these alternatives.  OW and OPP, as previously stated, are
discussing issues and methods of aggregating multiple pathway exposures.

The workgroup also stated the following:

    The currently proposed Margin of Exposure (MOE) approach may be of great
    applicability to Ambient Water Quality Criteria Standards.

EPA believes that the MOE approach has merit as an alternative way of expressing risk and the
Agency has used it for quite a while.  OPP has used the MOE approach for residential exposure
analyses and  is considering using it for their aggregate evaluations. However, OPP continues to
utilize Hazard Index and Aggregate Risk Index approaches also.  Furthermore, EPA is
considering using the MOE approach for assessing chemical carcinogens based on nonlinear
low-dose  extrapolation.

    3.   What population (subgroup or percent! le) is EPA trying to protect?  What level of
       protection is EPA shooting for??

The workgroup suggested that the methodology have clear policy and implementation goals on
the population protected (issue also discussed separately on p. 6-10 of the Peer Review Report,
where the workgroup referred to EPA's assessment for lead and the goal to have "95 percent of
the population below a specific level").

EPA described in its Federal Register Notice issues regarding identifying the population
subgroup  that the AWQC are designed to protect (see Appendix in.C. 1 .(a) of the draft
Federal Register Notice, p. 154). Nevertheless, we can provide greater clarity is characterizing
the intake parameters used to derive the criteria in the context of the population subgroup(s),
specifically describing the population segment as the target population or the criteria basis
population, estimating the exposures, and discussing why we believe the criteria are protective
of that segment of the population.

However, associating the derived criteria with a specific percentile is far more difficult, and such
a quantitative descriptor typically requires detailed distributional exposure and dose information.
EPA's Guidelines For Exposure Assessment (57 FR 22901, May 29, 1992) describes the
extreme difficulty in making accurate estimates of exposures and indicates that uncertainties at
the more extreme ends of the distribution increase greatly. On quantifying population
exposures/risks, the Guidelines specifically state:

    In practice, it is difficult even to establish  an accurate mean health effect risk
    for a population.  This is due to many complications, including uncertainties in
    using  animal data for human dose-response relationships, nonlinearities in the
    dose-response curve, projecting incidence data from one group to another dissimilar
    group, etc.  Although it has been common practice to estimate the number of cases
    of disease, especially cancer, for populations exposed to chemicals, it should be

                                           21

-------
    understood that these estimates are not meant to be accurate estimates of real (or
    actuarial) cases of disease.  The estimate's value lies in framing hypothetical risk
    in an understandable way rather than in any literal interpretation of the term "cases."

EPA also recommended that States and Tribes consider developing more stringent criteria to
protect highly exposed populations if they determined that criteria based on the general
population would not be adequately protective.  We will expand the discussion regarding our
recommendations for States and Tribes and their flexibility in deriving their own criteria and/or
adopting water quality standards.

Regarding the four conclusions described on p. 6-13, we offer the following responses.

•   EPA agrees with the workgroup majority that the values for cancer  effects (using the 90th
    percentile offish consumption and a cancer risk level of 10"6) are protective of public health.
    We believe the use of our fish intake assumption and drinking water intake assumption (2
    L/day), along with other conservative assumptions of the risk assessment, provide an
    adequate level of protection for the vast majority of the population and are appropriate for
    use in deriving national default criteria.  However, we  are also aware that exposure patterns
    in general and fish consumption in particular vary substantially. We  strongly emphasize our
    preference that States and Tribes use fish intake levels derived from local data, when
    available, instead of the default values when deriving AWQC to ensure that the level chosen
    will be protective of highly exposed subgroups in the population. [Note: The same idea also
    applies to the other exposure parameters, although available data indicate that the fish intake
    parameter is the most variable and, thus, the most subject to local/regional differences.] We
    recognized in the draft methodology revisions that risk  management decisions involved with
    the derivation of AWQC are, in many cases, better made at the State and Tribal  level. If, as
    the one workgroup member cautioned, a State or Tribe does not believe our default criteria
    would adequately protect populations that face a high risk, they have the flexibility to
    develop more stringent criteria for use in their standards programs.

•   For subsistence fishers, EPA has not prescribed the combinations offish intake levels and
    cancer risk levels that the workgroup indicates in this conclusion.  We have recommended
    default intake rates for various higher fish-consuming populations for State and Tribal use.
    Again, States and Tribes have the flexibility to use any of these intake level/cancer risk
    combinations or use their own fish consumption  data, as long as they can demonstrate that
    the most highly exposed population subgroup would not exceed a 10"4  cancer risk level. We
    also emphasized that approval of a Statewide 10"4 cancer risk level would be unlikely
    because of the need to ensure, and substantiate with data, that this level would not be
    exceeded. EPA notes that  special circumstances and assessment of natural contaminants
    may lead to numbers outside the 10"6 to 10"5 risk range. Based on the  support received
    from States and Tribes, we intend to finalize the methodology using the 99th percentile fish
    consumption rate from the  CSFII survey. However, it must be emphasized that we also
    intend to derive our CWA  Section 304(a) national  default criteria based on the general
    population (while using the 90th percentile fish intake rate from the CSFII in an effort to
    protect most consumers of fresh/estuarine fish) and based on a cancer risk level of 10"6.

                                           22

-------
(See our response for the fourth conclusion below, on the relation of cancer risk levels.)

EPA believes that the criteria developed for noncancer effects will also be protective of most
consumers of fresh/estuarine fish.  We acknowledge that our water quality criteria do not
account for cumulative exposures from multiple noncarcinogenic compounds (see response
to 4.2, above). We appreciate the idea that cumulative exposure from other compounds
with the same lexicological endpoint could make even the 50 percent ceiling on the RSC
(i.e., the workgroup's reference to Table 2.3.27 in the TSD where a 50 percent
apportionment of the RfD was used) not protective enough.  As previously stated, we are
not currently able to account quantitatively for specific cumulative chemical risks when
deriving our national default AWQC.  However, we continue to work Agency wide to
develop policies on cumulative risk. We will consider further the workgroup's idea of using
the RSC policy to address cumulative risk by possibly applying more conservative ceilings,
where appropriate, as implied by the workgroup's comment.

EPA's CWA Section 304(a) national default criteria serve as guidance to States and Tribes
who must, in turn, adopt legally  enforceable numerical criteria into water quality standards.
States and Tribes have the option of developing their own criteria and the flexibility to base
those criteria on population groups that they determine to be at potentially greater risk from
higher exposures, if they so choose—although many States have adopted EPA's Section
304(a) default criteria directly into their standards. We believe that basing our 304(a)
criteria on general U.S. population exposures is most appropriate, given their use as a default
value for the nation as a whole.  Furthermore, we cannot oblige the States to set their
standards  on a particular "sensitive population" because these criteria are guidance to the
States, not enforceable regulations, and do not impose legally binding requirements.
Nevertheless, in our methodology  guidance, we recommended that States and Tribes give
priority to identifying and adequately protecting the most highly exposed population by
adopting more stringent criteria,  if the State or Tribe determines that the highly exposed
population would not be adequately protected by criteria based on the  general population.

Also, we are not recommending a cancer risk level of 10"4, as the workgroup suggests.
States and Tribes have the option of deriving their criteria on a 10"6 risk level, as we propose
to do with our default criteria, combined with fish consumption rates for highly exposed
population groups. What we have stated in our methodology is that we consider
establishment of criteria that will be protective of the general population at an upper-bound
cancer risk in the range of 10"5 to 10"6 to be an appropriate risk management goal.
However, consistent with the Agency's risk management policy in other programs, we now
explicitly urge States and Tribes to ensure that the most highly exposed populations do not
exceed a risk level of 10"4.  In this respect, we have for the first time in our water quality
criteria program established a ceiling above which incremental cancer risk levels are not
considered acceptable.  We would disapprove any State or Tribal standard in which
information indicated that greater risk levels may be experienced by such highly exposed
groups.

It should be clarified that the incremental cancer risk levels are relative,  meaning that any

                                       23

-------
    given criterion associated with a particular cancer risk level is also associated with specific
    exposure parameter assumptions (i.e., intake rates, body weights).  When these exposure
    values change, so does the risk. Therefore, the workgroup's recommendation that we
    "protect these communities at the same level as the general population" is not conceptually
    accurate.  Given a criterion derived on the basis of a cancer risk level of 10"6, individuals
    consuming up to 10 times the assumed fish consumption rate would be protected at a 10"5
    risk level.  Similarly, individuals consuming up to 100 times the assumed rate would still be
    protected at a 10"4 risk level. Therefore, with a criterion based on EPA's default fish intake
    rate (17.8 g/day) and a risk level of 10"6, those consuming a pound per day would be
    protected at a 10"5 to 10"4 risk level (closer to 10"5). If a criterion were based on a "95%
    percentile level of exposure  and ... at 10"6" (as the workgroup suggests on p. 6-13), then it
    is likely that an average fish  consumer would be protected at a cancer risk level of
    approximately 10"8. The point here is that the risks for different population groups are not
    the same.

5. Response to Three Additional Related  Issues (6.5)

    1.  Use of Reliable/Empirical/Adequate Data

EPA agrees with the workgroup's comments regarding the use of reliable, empirical data for
inputs to the AWQC equation, including  the need to address data adequacy.  We will encourage
the use of such data by  States and Tribes and the generation of new data where resources allow
and where the collection of new data would improve the assessment. We acknowledge the
workgroup's approval of our minimum data requirements discussion and will emphasize that
States and Tribes need to characterize their assessments as completely as possible, especially
when the assessments are based  on combinations of data that are older/newer, national/regional,
and so on.

    2.   Encouraging State/Tribal Risk Assessments

As stated in comment-response 4.1, States and Tribes are not constrained by Federal limitations
in risk assessment evaluation schedules, resources,  or other factors;  States and Tribes always
have the option of undertaking their own  evaluations to develop water quality criteria, as long as
the criteria are consistent with CWA requirements.  Indeed many States have derived chemical
criteria values in the absence of EPA guidance  for those criteria and will continue to be able to
do so. We are well aware that the resources and expertise within States and Tribal authorities
vary greatly and, although we encourage  them  to pursue their own criteria and standards
development programs, we anticipate that many will continue to rely on our expertise and default
criteria.  We also acknowledge the workgroup's idea that some chemicals do not necessarily
require intensive risk assessments, whereas other chemicals are of great importance and require
greater accuracy.  In this respect, we intend to  devote our efforts to the development or revision
of criteria for chemicals of high priority and national importance, as proposed in the draft
Federal Register Notice.

    3.  Risks to Individuals and Populations

                                           24

-------
EPA agrees with the workgroup's statement that the AWQC approach should be one of "public
health protection," and we have developed our methodology with the protection of human health
in mind.  The language contained in both the original 1980 methodology and the draft
methodology revisions refers to risks, exposures, consumption rates, etc. tor populations, and
not specific individuals. The workgroup is correct that we do not explicitly assume that
"protecting the individual also protects the population."  However, the cancer risk estimates
developed for AWQC are derived for targeting specific incremental cancer incidence and, as
such, can be thought of as representing both individuals and populations—that is, a 10"6 cancer
risk represents one additional cancer case (individual) in one million (population)—and is clearly
a "defined risk-based goal" as the workgroup recommends.
In addition to deriving our default criteria to protect the general population, we have encouraged
States and Tribes to identify and protect more highly exposed subpopulation groups based, in
particular, on their water and fish consumption patterns. We have specifically referred to the
following groups: adults in the general population; sport (recreational) fishers; subsistence fishers;
women of childbearing age; and children. We also consider sensitive subgroups in calculating
dose-response estimates and in hazard identifications, where data warrant. In this sense, we are
concerned with risks to individuals (as represented by these population subgroups) and to the
overall population. However, as the workgroup suggests, our approach preferentially minimizes
risks to populations. We have also acknowledged that choosing intake rates for protection of a
certain percentage of the general population is a risk management decision and have emphasized
that in choosing a 90th percentile fish consumption value from the USDA national survey as a
default (a survey of 11,912 individuals), we are intending to protect a majority of the population
offish  consumers.  We also believe that our default rates for the sportfisher/sport angler and
subsistence fisher are protective of a majority of the individuals in those groups.

References

Hegsted DM.  1972. Problems in the use and interpretation of the recommended dietary
allowances. Ecol Food Nutr 1:255-265.

U.S. Environmental Protection Agency. 1998.  Science policy council handbook: peer review.
Prepared by the Office of Science Policy, Office of Research and Development, Washington,
DC. EPA/100/B-98/001.

WillettW.  1990. Nature of variation in diet. In: Willett W, ed. Nutritional epidemiology.
Monographs in epidemiology and biostatisties, vol. 15. New York/Oxford: Oxford University
Press, pp. 34-51. [Cited in the Mercury Study Report to Congress, 1997. EPA /452/R-
97/006.]
                                           25

-------
               BIOACCUMULATION WORKGROUP RESPONSES

This section contains EPA's response to all comments of the Bioaccumulation Workgroup on the
bioaccumulation factor (BAF) portion of the human health AWQC methodology as contained in
Section 7 of the September 1999 Peer Review Summary Report. Section 7 contained the
combined comments from the workgroup on the BAF portion of the proposed AWQC
methodology.

1.  Response to the Workgroup's Technical Charge Comments and Recommendations
    (7.1)

    General Issue: EPA requests comment on the recommended methodology guidance
   for estimating BAFs using a tiered approach that depends on the availability of data
    and resources, and the choice of the default parameter values provided.

    General Comments

The workgroup stated  that they were in  general agreement that using BAFs can result in better
predictions of bioaccumulation than BCFs for some nonpolar (nonionic) organic chemicals and
that the choice of most of the default parameters appears generally to reflect the state of the
science.  However, the workgroup expressed concerns that the draft BAF methodology is much
more complex and includes more assumptions than the previous BCF methodology.  The
workgroup also stated  that many model  parameters were highly uncertain and some assumptions
have a tenuous scientific basis.  In addition, the workgroup said that, as written, the draft
methodology has only had limited testing and could not be applied to ionizable compounds (e.g.,
pentachlorophenol).

We agree with the workgroup that BAFs are better predictors of chemical accumulation than
bioconcentration factors (BCFs) for certain types of compounds such as highly persistent, highly
hydrophobic chemicals. Numerous studies have confirmed the finding that for some chemicals,
BAFs exceed BCFs because of food web biomagnification (e.g., Russell et al. 1999; Fisk et al.
1998;  Oliver and Niimi 1983, 1985, 1988; Niimi 1985; Swackhamer and Kites 1988).  We
further agree with the reviewers that for some compounds (e.g., nonionic organic chemicals that
exhibit relatively low hydrophobicity), the BAF is expected to be similar to the BCF. To
address this issue, we  have revised the 1998 draft BAF methodology so that BAFs and BCFs
for minimally hydrophobic organic chemicals are considered equally in determining the National
BAF for an aquatic species, all else being equal.

We appreciate the need to balance complexity versus simplicity in developing guidance for
assessing bioaccumulation for deriving AWQC.  The 1998 draft bioaccumulation methodology is
more complex than the 1980 methodology, which emphasized the use of measured BCFs or
BCFs predicted from  Kow values. However, significant scientific advancements have occurred
over the past 20 years  that have greatly expanded our understanding of the bioaccumulation
process.  Therefore, the added complexity of the draft methodology is required to increase the
scientific soundness and accuracy of AWQC through incorporation  of these scientific advances.

                                        26

-------
For example, methods to directly address the effect of organic carbon on the bioavailability of
nonionic organic chemicals are absent from the 1980 methodology but are explicitly
incorporated in the revised bioaccumulation methodology.  Furthermore, the use of model-
derived food chain multipliers in combination with measured or estimated BCFs has been shown
to provide more accurate estimates of bioaccumulation for persistent, highly hydrophobic
chemicals than the use of BCFs alone (Burkhard et al. 1997; U.S. EPA 1995, 1998).  These
methods are also absent in the 1980 methodology. In addition, trophic-level dependence of
bioaccumulation, which can be important for some types of chemicals, is not explicitly addressed
in the 1980 methodology but is addressed in the new methodology. It should be noted that the
added complexity of the revised 2000 methodology also provides greater opportunity to
stakeholders to modify national BAFs to address site- or region-specific attributes, which again
was lacking in the 1980 methodology.

We agree that for some types of chemicals, the procedures for deriving BAFs can be simplified
from those presented in the 1998 draft methodology. Accordingly, we have revised the 1998
draft methodology so that the derivation of BAFs is tailored to specific  categories of chemicals,
some of which require less complex procedures. For example, we revised the draft
methodology to limit the use of food chain multipliers (FCMs) to groups of chemicals where they
are most likely to impact the BAF (e.g., highly hydrophobic organic chemicals that have
reasonable likelihood of persisting in aquatic biota). We do not recommend use of FCMs for
other types of chemicals (e.g., organic chemicals that have been shown  to metabolize
substantially in biota and those with low hydrophobicity). We have also limited the derivation of
separate BAFs for each trophic level to groups of chemicals where such distinctions are most
meaningful (e.g., highly hydrophobic chemicals).

Responses are provided below for the various comments pertaining to model parameters and
assumptions.

•  Evaluation of the draft bioaccumulation methodology focused on persistent, hydrophobic
   chemicals in selected locations (e.g., Lake Ontario, Green Bay, Bayou d'Inde, Louisiana)
   because of a general lack of appropriate data for other types of chemicals in other
   geographic areas. The workgroup raised concerns about the applicability of certain portions
   of the  methodology to certain classes of chemicals. In response to  this, we have developed
   additional guidance that restricts some aspects of the methodology  to certain types of
   chemicals. For example, we have removed the use of Kow-based BAF estimates and
   model-derived FCMs for chemicals that have been consistently shown to be metabolized
   substantially in aquatic biota (e.g., benzo[a]pyrene in vertebrates).

•  Regarding the locations in which certain  aspects of the methodology have been tested (FCM
   and biota-sediment accumulation factor (BSAF) approach), we agree that these sites are
   few in number, largely because the availability  of appropriate field data is so limited.
   Although few in number, these sites do provide a range of ecosystem types from which to
   evaluate the BAF methodology.  Specifically, they include the hydrodynamically complex
   and tidally influenced area of Bayou d'Inde (Lake Charles, LA) and the more stable,
   oligotrophic system of Lake Ontario. A limited evaluation of the BSAF methodology was

                                          27

-------
    also performed with the more shallow, eutrophic system of Green Bay, Lake Michigan. To
    obtain an assessment of the performance of the BAF methodology in lotic systems, we
    evaluated two other data sets (PCBs in the Hudson River and Fox River/Green Bay).  We
    believe that placing additional limitations on the use of predicted BAFs, as noted above, and
    the further evaluations we conducted to compare predicted BAFs in other systems (e.g.,
    Hudson River and/or Fox River/Green Bay) to field-measured BAFs, give the revised
    bioaccumulation methodology a better scientific foundation and supports using it to derive
    national BAFs.

•   Regarding ionizable chemicals, we agree that the draft methodology did not clearly
    differentiate between nonionizable and ionizable chemicals and have revised the draft
    methodology to include separate procedures specific to determining BAFs for ionizable
    chemicals.

    Document Readability

The workgroup stated that the bioaccumulation methodology needed revision before it could be
applied on a national scale. Specifically, the methodology needed better direction and improved
readability, including a more precise description of what to do and when to do it.  The
workgroup recommended that a more prescriptive approach be developed that retains the
intended flexibility and site-specific alternatives. In one instance the workgroup also commented
that once revised, the methodology might be implemented on a more limited State or site-specific
scale.

EPA has made substantial revisions to the 1998 draft bioaccumulation methodology as a result
of workgroup comments. To improve readability and clarity of the methodology we separated
the  guidance for developing national BAFs from the guidance for developing site- or region-
specific BAFs. The revised national BAF methodology is written in a more prescriptive manner
so that it is clear how EPA plans to derive national BAFs. In the guidance for site- or region-
specific BAFs, we have expanded the guidance to better enable such adjustments to be made by
States, Territories, and authorized Tribes. For example, the databases used to develop national
default values for lipid content in aquatic biota and organic carbon content in water were
updated and expanded to make data more accessible so that States  and authorized Tribes can
more readily develop site- or region-specific values. After publication of the revised
methodology, we will also develop detailed guidance to stakeholders for designing and
conducting field studies to measure site-specific BAFs and BSAFs. This guidance will specify
our recommendations for how, when, where, and how often one should sample water, biota,
and sediment for producing reliable measurements of BAFs and BSAFs. We expect to
complete this guidance within a year following publication of the revised AWQC methodology.

In addition to improved clarity and expanded guidance, we have revised the draft
bioaccumulation methodology to address and reduce uncertainty in various aspects of the
methodology, as recommended by the workgroup. For example, to reduce uncertainty in
national BAFs as a result of improper application of the methodology to a certain chemical
group, and to simplify procedures, we developed separate procedures for deriving BAFs for

                                          28

-------
different chemical classes (e.g., high versus low hydrophobicity, high versus low metabolism in
biota, ionic versus nonionic organics).  We also revised the guidance to recommend that KQW-
based estimates of BAFs and FCMs not be used for nonionic organics that are known to be
metabolized substantially in targeted biota. Restrictions have been put on the use of the BSAF
methodology, such that it is applied only to highly hydrophobic organic chemicals.
    Scale of Application

Although we recognize that even with the revisions to the BAF methodology, significant
uncertainty might exist in the derivation and application of national BAFs at some sites
throughout the United States because of the influence of site-specific factors, we do not agree
that the methodology should be limited to State or site-specific use. We believe the revised
methodology is applicable on a national basis and will result in broadly applicable national BAFs
for several reasons.  First, for the predictive methods that incorporate factors affecting
bioavailability and bioaccumulation (i.e.,DOC/POC, lipid) we use default values for the factors
based on average values derived using large nationally representative data sets. Second, we
obtained bioaccumulation field data for a representative range of ecosystems (e.g., Lake
Ontario, Green Bay/Fox River, Hudson River, Bayou d'Inde), chemicals (PCBs, dioxins,
chlorinated benzenes, pesticides), species, and trophic levels and shown through comparisons of
field-measured and predicted BAFs, that when used appropriately, the different predictive
methods result in BAFs that agree very well to field-measured BAFs with few exceptions.
Third, by improving the readability and direction of the bioaccumulation methodology and by
limiting the use of the different BAF methods to certain groups of chemicals for which they are
most appropriate, we have also reduced the potential uncertainty that might occur from
inappropriately applying the methodology to certain groups of chemicals.

We believe that deriving national  304(a) water quality criteria using national BAFs is a sound
scientific approach and results in criteria that can be implemented effectively throughout the
United States.  For more than two decades, EPA has developed and implemented its national
304(a) water quality criteria (aquatic life and human health) through State and, on  occasion,
Federal water quality standards programs. Implementation of this program has relied on the use
of national  304(a) criteria as a cornerstone, and has evolved to allow the use of procedures to
modify national criteria by States, Territories, authorized Tribes, and other stakeholders where
appropriate. The revised national bioaccumulation methodology is consistent with this
programmatic practice, by enabling States, Territories, and authorized Tribes to readily adopt
national 304(a) water quality criteria into standards (based on national BAFs) that achieve the
Clean Water Act goals of protecting public health while also allowing site- or State-specific
adjustments to be made in situations where national AWQC may be considered to be
overprotective or in some cases, underprotective. In contrast to the workgroup
recommendation, we believe that restricting the bioaccumulation methodology only to the
development of State or site-specific BAFs would greatly hinder implementation of water quality
criteria throughout the United States.  This would be the case because many States, Territories,
and authorized Tribes lack the resources to develop State- or site-specific BAFs for all of the
numerous pollutants of concern and thus, subsequent adoption of AWQC would be delayed

                                           29

-------
substantially.

    Issue 1:  The appropriateness of the recommended procedures for estimating the
    consumption-weighted default lipid value, the equation to derive the freely dissolved
   fraction of a chemical (including estimates ofKDOC andKPOC), and the choice of food
    web structures used to calculate food chain multipliers.
    Default Lipid Value

The workgroup stated that the general approach for deriving the default lipid level is appropriate
but had several concerns that led them to question the representativeness of the trophic-level
mean lipid values. One concern related to the low or unknown sample sizes of lipid values
supporting many of the species-mean lipid content values.  Although the workgroup was
unaware of any other compilation that EPA could use to augment its existing database, they
indicated that individual studies that report lipid content could be used to provide a more robust
database.

EPA agrees that several data sets supporting the species-specific lipid values are of low or
unknown sample size because of limitations in the available data. Most of these data sets pertain
to the estuarine species, which are not widely represented in available databases, such EPA's
environmental monitoring database called STORET (data STOrage and RETrival). STORET is
a repository for water quality, biological, and physical data and is used by state environmental
agencies, EPA and other federal agencies, universities, private citizens, and many others for
environmental management purposes.  Generally, the sources used for estimating lipid content for
estuarine species report data in summary format and do not reveal the underlying sample size.  In
order to increase the certainty of species-mean lipid values, we have conducted additional data
searches that specifically target lipid data for species where the sample size is low or unknown.
Where appropriate, we have expanded the data sets to include additional data for these species.

Another workgroup concern regarding the representativeness of the recommended national
default lipid fraction values related to the aggregation of lipid data to the trophic level category,
including both freshwater and saltwater species, given the variability in lipid content that can
occur within and across species in the same trophic level.  The workgroup recommended that
additional guidance be developed on how site-specific data could be combined with the national
default data.

As discussed in the 1998 draft bioaccumulation methodology TSD, lipid  content can vary
significantly not only across aquatic species but also within a species because of a variety of
factors, including age, size, sex,  and diet of the fish; sampling season; and environmental
conditions (pp. 185, 239). Furthermore, the representativeness of the national default values
may vary for different sites.  As a result, we recommended (and will continue to recommend in
the  revised national bioaccumulation methodology) that wherever possible, States, Territories,
and authorized Tribes use site-specific or region-specific data to determine the identity  and lipid

                                          30

-------
content of consumed aquatic species. To enable States, Territories, and authorized Tribes to
develop their own lipid content estimates, we have revised the lipid database used in the 1998
draft methodology. The new lipid database includes more data for aquatic species having low
sample sizes (see previous response to comment) and to include additional aquatic species that
may be commonly consumed but were not part of the original database. Such species (e.g.,
walleye) were usually omitted from the database in the 1998 draft methodology because they did
not reflect the types of aquatic biota that were being consumed by humans as reported by the
USDA's Continuing Survey of Food Intake by Individuals-CFSII, 1989-1991.  We expect the
revised and expanded lipid database to be much more useful to States, Territories, and
authorized Tribes when they are modifying the national default lipid content values to better
reflect their situation.

States, Territories, and authorized Tribes will not always have the resources or data to develop
site- or region-specific lipid values. In these situations, the national values provide States,
Territories, and authorized Tribes with reasonable default values of lipid content in commonly
consumed aquatic organisms based on the best available data.  Regarding aggregation of lipid
content data to the trophic level, it should be noted that the mean lipid fraction value determined
for each CFSn consumption category in the 1998 draft methodology was weighted
appropriately by the corresponding consumption rate determined from the survey. For example,
variation in mean lipid content between the CSFn category of "perch" and "estuarine salmon"
(both  assigned to trophic level four) was accounted for in the national default lipid value
calculation by weighting by their individual consumption rates.  In recognition that lipid content
can vary appreciably across species comprising each of the CSFn consumption categories (e.g.,
lake trout versus brook trout in the "trout" category), we derived "average," "low," or "high"
estimates of the recommended national default values in the draft 1998 methodology, based on
differing assumptions of the representativeness of different species for a given trophic level.
Although these estimates were originally done as a sensitivity analysis, we have added additional
guidance to States, Territories, and authorized Tribes in the TSD on how to adapt the national
default lipid values to reflect State  and local consumption patterns where such data are available.
To enable such modifications to be made, we will make the raw data available to States,
Territories, and authorized Tribes for the purposes of selecting species-specific lipid content
different from the default values used by EPA in derivation of national 304(a) criteria.

The workgroup had several comments concerning the method of lipid extraction and analysis.
These concerns include (1) not specifying the lipid extraction method used for data constituting
the national default lipid values, (2) the need to recommend method(s) for measuring lipid
composition, including what tissues to analyze, and (3) a recommendation that alternative (but
unspecified) lipid extraction methods used for fish residue analysis would be more appropriate
than the Bligh-Dyer method, due to its greater affinity for polar lipids.

As discussed on page 185 of the 1998 TSD, various lipid extraction methods can extract
differing quantities of lipid from the same tissue of aquatic  organisms. In one study (Randall et al.
1991), lipid fraction varied by nearly fourfold among four extraction methods, but varied by
twofold or less among two of the more common extraction methods (chloroform-methanol and
acetone-hexane). Additionally,  the relative importance of lipid extraction method might vary

                                           31

-------
depending on the lipid content of the tissue, with lean tissues containing proportionally more
polar lipids (and greater potential difference due to the type of extraction solvent used)
compared with tissues with more adipose (nonpolar) lipids. Other attributes (e.g., high
temperature, pH, lipid decomposition due to exposure to light and oxygen) also can affect lipid
extractions.

Although it is desirable to have one standardized method for extracting and analyzing lipids for
the purposes of normalizing residues of nonionic organic chemicals, a clear consensus has not
emerged on which method is most appropriate for all tissues, species, and nonionic chemicals.
Furthermore, it might be true that no single method is equally appropriate for all chemicals and
tissues because different tissues have different lipid compositions (e.g., polar versus nonpolar
lipids), which in turn may alter the partitioning of various nonionic organic chemicals to varying
degrees. The science is not presently clear on which lipid fractions (e.g., phospholipids, free
fatty acids, mono-, di- and triglycerides) are most lexicologically relevant with respect to
different organic chemicals. For example, DDT has been reported to bind to more polar
membrane-associated lipids, which might render them lexicologically relevant (Chefurka and
Gnidec 1987, as cited by Randall et al. 1991).  In a followup study, Randall et al. (1998)
reported that 27% of extractable PCBs were analytically associated with the more polar,
membrane-bound lipid pool (i.e., extractable with chloroform/methanol) whereas 73% were
associated with the neutral lipid pool (i.e., extractable with hexane).  This finding further suggests
that membrane-bound lipids should not be ignored with lipid extraction techniques, at least for
some pollutants.

Although there is practical appeal to using the same solvent system to extract both lipids and
nonionic organic chemicals (i.e., separate chemical and lipid extraction methods would not be
necessary, as is required with the Bligh and Dyer method), different analytical methods can vary
in their extraction methods even for the same pollutant. For example, EPA method 1613 for
chlorinated dioxins and dibenzofurans uses a 50:50 mixture of hexane:methylene chloride to
extract fish tissues (U.S. EPA 1994). Another EPA method for analyzing PCBs and TCDD for
fish tissue uses acetonitrile (U.S. EPA 1980).  Thus, coextraction of lipids with the target
analytes may still result in different lipid fractions being measured for the same tissue, depending
on the analytical method used.  Furthermore, as noted above, it is not clear that the more polar
lipids (e.g., membrane-bound phospholipids) are lexicologically irrelevant.

For the sake of consistency in measuring BAFs and BSAFs using field studies, we continue to
recommend the use of the Bligh and Dyer (1959) chloroform/methanol extraction method (or the
less toxic solvent system of Kara and Radin (1978) which uses hexane/isopropanol) in
combination with gravimetric analysis for lipid measurement (p.  185 of TSD).  We recommend
the Bligh-Dyer method because it is widely used for lipid measurements and has been well
characterized in terms of the types of lipids extracted.  The Bligh-Dyer method also extracts both
polar and nonpolar lipids, both of which might be lexicologically important. These and other
considerations led Randall el al. (1998) lo recommend the Bligh-Dyer melhod as a slandard
technique for tola! lipid extinction pending more research lo identify the complex neulral
pollulanl and lipid relationships and subsequenl developmenl of a final slandard melhod.  Randall
el al. (1998) further recommended lhal if other lipid extinction melhods are used, comparisons

                                           32

-------
should be made to the Bligh-Dyer method to allow conversion of the lipid results to Bligh-Dyer
equivalents.  EPA has added similar guidance in the revised bioaccumulation methodology.

Regarding the tissues to be extracted, we have added guidance recommending that the percent
lipid be measured on the tissue used to derive the BCF or BAF study, which should be the
edible tissue of the organisms (e.g., fillet, whole body, soft tissue, etc., depending on the
species).  Guidance was provided on the preferred tissue type on pages 175  and 185 of the
draft 1998 TSD, and we have added clarity to this language.

Finally, where data were available, we have summarized which lipid extraction methods were
used to develop the database that supports the recommended national default lipid values. We
reviewed the lipid data and removed data derived using methods that were considered to be
suspect. It should be noted, however, that we weighed the added uncertainty of basing national
default lipid values on substantially fewer lipid data (because of incomplete information on
extraction method) against the uncertainty that might result from including data with different or
unknown lipid extraction methods.  In some cases, lipid records contained little or no information
on the extraction method, yet they were retained (appropriately flagged) in the database used to
derive the national default lipid values.

The workgroup recommended that the tissue type (edible, fillet, whole-body) be specified on
Table 2.4.8 and 2.4.10 in the TSD.  We agree and have made this change.

The workgroup stated that use of the consumption-weighted default lipid value in the AWQC
estimation process assumes that each trophic level is contaminated at the highest allowable
concentration. The workgroup considered this assumption not realistic and recommended that
some method to provide a distribution of contamination be used, or at least a differential source-
based contamination scheme for separating fish from waters of concern versus fish from  other
sources.

As authorized by Section 304(a) of the Clean Water Act (CWA), EPA is charged with
developing water quality criteria that reflect the latest scientific knowledge of the effects of
pollutants on human health and welfare.  EPA's 304(a) water quality criteria are often used by
States, Territories, authorized Tribes, and EPA to set enforceable water quality standards that
are designed to meet the designated uses of a water body (e.g., fishing,  swimming, propagation
of aquatic life, recreation). In developing the methodology for deriving human health criteria, we
made estimates about exposure to contamination from eating fish taken from surface waters.
The purpose of the estimates was to ensure that if criteria were met in a water body designated
for fishing, most people could safely eat fish from that water body.  In addition to the estimate
that 17.8 grams offish are consumed per day (a value reflecting the 90th percentile of the
general population), we also estimated that fish and shellfish are taken from water with pollutant
concentration at the criterion level. It is our view that to ensure that people can safely eat fish
from waters designated for fishing, it is necessary to assume that all of the consumed fish are
taken from water bodies with chemical concentrations present at the criteria level (i.e.,
contaminated to the maximum safe level). Fishing patterns (i.e., extent and location of fishing),
and the degree to which fish and shellfish bioaccumulate contaminants from waters across the

                                          33

-------
United States, may differ from the exposure assumptions used to calculate national 304(a)
criteria. The national criteria (which States, Territories, and authorized Tribes may modify) are
designed to be protective for the general population.

The data do not exist to enable 304(a) criteria to reliably account for the myriad of spatial and
temporal differences in fishing patterns, bioaccumulation, and subsequent differences in exposure
to fish contaminants at the national level.  For example, a particular water body might not be of
concern to individuals of one subpopulation because they do not use it for fishing. However, this
waterbody might be of concern to individuals of another subpopulation because it serves as a
significant resource for their diet. Data at the national level that would enable such fine
distinctions to be made are not available. It should also be noted that, once adopted into State
or Tribal standards, AWQC must protect the designated use of the water body (e.g., fishable,
swimmable) regardless of the extent to which that designated use is actually being exploited. For
these reasons, we believe that the exposure assumptions use to derive national 304(a) criteria
are  necessary to achieve adequate protection of humans from  exposure to waterborne
pollutants. Where States, Territories, and authorized Tribes have concerns regarding the level of
protection afforded by EPA's national criteria, EPA encourages States, Territories, and
authorized Tribes to make appropriate adjustments to reflect local conditions affecting fish
consumption and bioaccumulation.  Guidance for making such modifications is provided in the
revised methodology.

    Freely Dissolved Fraction

The workgroup commented that the equation to estimate the freely dissolved fraction of nonionic
organic chemicals generally reflects the current state of knowledge, but they were concerned that
the  equation did not allow for ionization despite unspecified methods being available to  account
for this phenomenon.

The workgroup concurred with the use of the three-phase  partitioning model to estimate the
freely dissolved concentrations of nonionic organic chemicals in ambient waters.  They suggested
that the three-phase partitioning model can be extended so that ionizable chemicals such as
pentachlorophenol, silvex, • -naphthylamine,  and aniline can be addressed.

In response to the workgroups comments regarding ionization  of organic chemicals in water, we
revised human health methodology by dividing the chemical universe into three general classes:
nonionic organics, ionic organics, and inorganics including organometallics. Ionic organics
include chemicals containing functional groups with exchangeable protons such as hydroxyl,
carboxylic and sulfonic groups,  and functional groups that readily accept protons such as amino
and aromatic heterocyclic nitrogen (pyridine) groups.  In the revised methodology, the users are
directed to the section on ionic organics when the chemical of interest is of this class, and this
section also provides methodologies for deriving AWQC for this class of chemicals.

As part of the revisions, we reviewed  the literature describing ionization of organic chemicals in
water.  In general, most organic acids, (e.g., pentachlorophenol and silvex) exist mostly in the
ionized form in ambient waters because their pKa's (4.75  and  3.07) are much smaller than the

                                           34

-------
pH of the ambient waters.  Conversely, most organic bases (e.g., aniline) exist mostly in the un-
ionized form in ambient waters because their pKb's (4.63) are much smaller than the pH of the
ambient waters. When the species of the chemical is predominately in the un-ionized form, the
chemical can be treated as if it were a nonionic organic chemical. Significant ionization (more
than 99% ionized) occurs for organic acids and bases when the pH > pKa +  2 and pH < pKb +
2, respectively.

During the revisions we also reviewed available models for predicting the partitioning and
bioavailability of ionized forms of organic chemicals (e.g., for review, see Spacie, 1994; Suffet et
al., 1994). Although the neutral species of ionic organic chemicals are thought to behave in a
similar manner as nonionic organic compounds (e.g., partitioning to lipids and organic carbon as
a function of hydrophobicity), the ionized (cationic, anionic) species exhibit a considerably more
complex behavior involving multiple environmental partitioning mechanisms (e.g., ion exchange,
electrostatic, and hydrophobic interactions) and a dependency on pH and other factors including
ionic strength and ionic composition (Jafvert et al., 1990; Jafvert 1990; Schwarzenbach, et al.,
1993). As a consequence, methods to predict the environmental partitioning of organic cations
and anions are less developed and validated  compared to nonionic organic chemicals (Spacie,
1994; Suffet et al.,  1994).  Given the current limitations in the state of the science for predicting
the partitioning and bioaccumulation of the ionized species of ionic organic chemicals, EPA has
decided not to extend the freely dissolved equation to include ionic organic chemicals. Rather,
EPA has developed separate procedures for addressing bioaccumulation of ionic organic
chemicals which depend on the extent to which the fraction of the total chemical is likely to be
represented by the ionized (cationic, anionic) species in U.S. surface waters. When a significant
fraction of the total chemical concentration is expected to be present as the ionized species in
water,  procedures for deriving the national BAF rely on empirical (measured)  methods (i.e., field
BAF or laboratory BCF). When an insignificant fraction of the total chemical  is expected to be
present as the ionized species (i.e., the chemical exists essentially in the neutral form),
procedures for deriving the national BAF follow those established for nonionic organic
chemicals, which address the freely dissolved form. As the science improves on predicting the
partitioning and bioavailability of ionic organic chemicals, EPA plans to consider the  use of
partitioning and bioavailability models on  a case-by-case basis.
Additional information on partitioning of ionic organic chemicals is presented in the revised
methodology for ionic organics.

The workgroup commented that implicit decisions and assumptions are used with the three-
phase partitioning model for estimating the freely dissolved concentration for a nonionic organic
chemical in the ambient water. The workgroup recommended that the guidance document needs
explicitly to identify these assumptions, and to provide discussion and information "at a level that
will allow the user to gain a sense of the uncertainty of this approach."  The peer reviewers
recommended that this detailed discussion be placed in an appropriate appendix.

Three implicit assumptions in the methodology were highlighted by the workgroup:

•  The values for the paniculate and dissolved organic carbon partition coefficients are set to
   default values depending on the type of aquatic environment.

                                           35

-------
•   The assumption that the freely dissolved chemical in the water is in equilibrium with
    particulate organic carbon (POC), dissolved organic carbon (DOC), phytoplankton, and
    zooplankton.

•   Chemical bioavailability to aquatic organisms is reduced because of sorption of the chemical
    to DOC and POC.

EPA has provided more detailed discussions and information highlighting the uncertainties
associated with the implicit assumptions used with the three-phase partitioning model in the
bioaccumulation factor part of the TSD.  As suggested by the workgroup, we conducted an up-
to-date literature review and subsequent evaluation of the default values for KDOC and KPOC
values. In the methodology section of the guidance document, we more clearly identified the
implicit assumptions and refer the reader to the appropriate appendix for additional details.
Where default values were used in the methodology for individual parameters, these
selections/decisions have also been more clearly identified.

    Food Chain Multipliers

Clearer guidance on use ofFCMs.  The workgroup commented that although EPA provided
three different examples ofFCMs that varied  depending on the mix of pelagic and benthic
components, no clear guidance was provided on which one to use.

The lack of clarity in the draft methodology was caused in part by mixing of the national and site-
specific methodologies in the same guidance. To address this issue, in the revised methodology
we have divided the national and site-specific BAF methodologies into separate documents.
The detailed technical basis for the methodologies appears where appropriate in the revised
national and site-specific BAF guidance documents.

For derivation of national BAFs, we chose to use a mixed benthic/pelagic food web because we
believe that this food web is the most broadly applicable and typical food web encountered in
nature. The use of a mixed benthic/pelagic food web also results in FCMs that are midway
between a pure pelagic and pure benthic food web structure. Discussions have been provided in
the revised methodology that allow the user to gain a sense of the uncertainties associated with
using the mixed benthic/pelagic food web as the default food web.

For determination of site-specific BAFs, the document provides additional guidance on which of
EPA's recommended FCMs to use depending on the situation. In addition, EPA also strongly
recommends that site-specific FCMs be determined whenever possible using site-specific food
web parameters (e.g., diet, lipid content, and weight for each organism and sediment-water
disequilibrium). The revised document provides guidance on how one might assess the diet, lipid
content, and weight for organisms at their field site.

Applicability ofFCMs.  For several reasons, the workgroup commented that the proposed
FCMs are not broadly applicable to all chemicals and all aquatic ecosystems.  Specifically, the

                                          36

-------
workgroup said that further validation of FCMs was needed for additional chemicals and diverse
systems before FCMs could be used on a national scale. In addition, concerns were raised that
the FCMs should not be applied to chemicals that are readily metabolized or those that do not
reach steady state in the food chain during the same time as the environmental half-life.

EPA has revised the guidance to limit the use of FCMs, which account for biomagnification
processes in aquatic food webs, to high Kow nonionic organic chemicals that have been shown
to persist (or have a reasonable likelihood of persisting) in aquatic biota of concern. The
workgroup suggested that FCMs should not be used on a national scale.  However, the
workgroup did not provide alternative recommendations on how to account for biomagnification
processes on either a national scale or a site-specific basis, and EPA does not know of any
other sound approaches to account for biomagnification processes. (Note that FCMs and
BMFs [biomagnification factors] are the same approach because BMFs are equal to ratio of
FCMs.  Also, it should be noted that FCMs were derived using food web models, and thus
food web models and FCMs are one and the same approach as well.)

The workgroup suggested further field verification of the use of FCMs because a wide diversity
of ecosystems have not been included.  EPA has performed successful verification studies in two
different ecosystems: (1) Lake Ontario,  an oligotrophic freshwater ecosystem,  and (2) Bayou
d'Inde, Lake Charles, LA, an estuarine ecosystem with variable salinities. We have also
evaluated additional data from the Hudson River and the Fox River/Green Bay ecosystems to
further field verify the predicted BAF methods, as was suggested by the workgroup.

EPA agrees with the workgroup that FCMs that assume no metabolism in the food web could
be inappropriate for chemicals that are metabolized. However, as stated by the workgroup,
"there is no reliable, broadly applicable (universal) approach to predicting the metabolic
breakdown of organic chemicals by biota."  In view of the lack of methodologies for predicting
metabolism, EPA has made a science policy decision to assume no metabolism when deriving
FCMs.  For chemicals and species wherein metabolism has been shown to be  important, EPA
has revised the guidance to recommend not using model-derived FCMs.  EPA has added text
and guidance caveating the limitations of the FCMs because of their inability to account for
metabolism processes.  In the methodology, B AFs derived using the product of a measured
BCF and FCM do include metabolic processes of the organisms used in the BCF measurement.
The revised BAF methodology allows modification of the national BAFs to account for site-
specific applications, which includes procedures to account for metabolism in the derivation of
site-specific FCMs when appropriate metabolic rate data exist.

The workgroup suggested that "chemicals that do not reach steady-state in the food chain during
the same time frame as the environmental half-life should not be included here." We do not
agree with this because the  suggestion does not recognize that time to steady-state and
environmental half-life are not necessarily related. In addition, loadings of the chemical to the
ecosystem are an important and controlling factor in establishing the concentrations of the
chemical in the ecosystem (e.g., ambient water). Consider the following example:  Assume that
one unit of a chemical is added to a lake per day, the lake is well mixed, all chemical is retained
in the lake, and the  chemical is lost using a first-order rate loss.  Given enough time, the total

                                          37

-------
amount of chemical in the lake will plateau at 43.8 units assuming a 30-day environmental half-
life for the chemical.  Because the total amount of chemical in the lake plateaus, the time to
steady-state in the food web has no relevance to whether the residue is formed or not formed in
the organisms composing the food web.  As one decreases the environmental half-life of the
chemical, the total amount of chemical in the lake becomes less; for example, for environmental
half-lives of 10, 5, and 1 days, the total amounts of chemical in the lake are 14.9, 7.7, and 2.0
units. When the environmental half-life becomes very small, less than 1 day in this example, the
total  amount of chemical in the lake becomes very small, e.g., «1 unit.  Even when the
environmental concentrations become very low, bioaccumulation processes still occur. In
effluent-dominated systems, bioaccumulation of chemicals with relatively short environmental
half-lives might be important because of the continuous loading of the pollutant to the system.

The workgroup's suggestion about considering time to steady-state and environmental half-lives,
although interesting, does not seem to resolve the problem  of metabolism in food webs. The
workgroup's suggestion seems reasonable because chemicals with small environmental half-lives
rarely produce measurable residues in aquatic organisms.  However, even though a chemical is
rarely detected in aquatic organisms, this does not necessarily mean that bioaccumulation
processes do not occur.  It could be that the concentrations of the chemical in the ambient water
are so low that the residues in aquatic organisms are not detectable even with the
bioaccumulation processes.

Finally, we note that the  issue of environmental persistence is most appropriately addressed
during the permitting process with the use of dynamic water quality models.  Such models can
account for degradation processes (e.g., hydrolysis, volatilization, photolysis) in the wasteload
allocation and subsequent derivation of the permit limit (U.S. EPA 1991).

Sediment interaction.  The workgroup noted that sediment interaction (benthic-pelagic
coupling) can be a dominant driver of bioaccumulation and that it should be incorporated in
some manner.  The FCM methodology does include the benthic-pelagic coupling, as noted by
the workgroup. The three sets of FCMs provided in the 1998 draft of the methodology
presented FCMs for a purely benthic food web, a mixed benthic-pelagic food web, and a purely
pelagic food web.  Benthic-pelagic coupling is incorporated via the diet of the consumers in the
food web.

Other FCM Comments. The workgroup pointed out that FCMs have some potential
problems. First, no one food web can realistically represent the entire United  States. However,
the workgroup did indicate that a  default food web could be used in many cases to provide
acceptable estimates. Second, the workgroup suggested that for broad, general application (of
the default food web structure), there  remain a number of unvalidated assumptions that might be
the source of considerable uncertainty (when using the default food web). Uncertainties include:
(1) variability in diet, physiology, and ambient conditions; (2) dietary lipid content and its
relationship to bioaccumulation and toxicity; and (3) all chemicals and organisms act ideally (i.e.,
various physical, chemical, and biological modifying factors are identical).  The workgroup
recommended that additional guidance and limitations be provided on the use of FCMs in
criteria development.

                                          38

-------
The workgroup agreed with EPA that the scientific understanding of aquatic food web processes
is known well enough to develop a default food web structure that provides a realistic
representation of the processes occurring in aquatic food webs for many ecosystems.

EPA is using state-of-the-art food web models for deriving FCMs, which incorporate the latest
thinking and knowledge on the processes occurring in aquatic food webs.  The workgroup
suggests that the assumptions used in constructing these models are largely unvalidated. We
recognize that any modeling formulation of contaminant behavior in aquatic food webs requires
simplification of a very complex biological system in order to assemble a tractable model. These
simplifications do not imply  or mean that our scientific understanding of all processes occurring in
food webs is complete.  As documented in the scientific literature by Gobas and coworkers,
MacKay and coworkers, and Thomann and coworkers (all model-building research groups),
these simplifications provide reasonable model formulations with good predictive power.

EPA has performed an analysis of the importance and sensitivities of individual input parameters
for food web models and of the overall uncertainties associated with predictions from food web
models (Burkhard 1998). We have provided additional discussion in the TSD outlining the
results from these analyses and their implications for deriving FCMs. Comparisons between
measured and predicted BAFs for the Lake Ontario food web using the Gobas and Thomann
food web models resulted in average ratios of 1.2 and 2.5, respectively, for PCBs and
chlorinated pesticides. The  overall uncertainties (expressed as the ratio of the 90th to 10th
percentile values in the distribution of predicted BAFs) associated with the Gobas and Thomann
models were a factor of 3.6  and 4.0, respectively, for a chemical with a log Kow of 6.5 for the
Lake Ontario food web.  The small ratios (of predicted to measured BAFs) and small
uncertainties associated with both the Gobas and Thomann food web models strongly suggest
that the assumptions used do not introduce large uncertainty into the model predictions as
suggested by the workgroup.

We have fully considered the workgroup's comment regarding applicability of the proposed
FCMs, and, consistent with  our responses to that comment and to the previous one requesting
clarity on which FCM to use, we have limited  the use of FCMs to nonionic organic chemicals
with log KQWS • 4.0 in both the national default methodology and the site-specific methodology.
In addition, EPA has restricted the use of model-derived FCMs in situations where metabolism
has been shown to be important.  EPA appreciates the workgroup's  concern that not all
chemicals have identical behavior.

The workgroup suggested that additional guidance be provided on the uncertainties associated
with input parameters such as diet, organism physiology (e.g., weight, temperature preferences,
and lipid content), sediment water disequilibrium, lipid content of the diet, and with processes
modifying bioaccumulation potential. Because EPA understands that the default food web
structure might not be appropriate in some site-specific conditions, the methodology includes
procedures for making site-specific modifications to BAFs derived using the national
methodology. These procedures allow the use of site-specific parameters in the generation of
FCMs. As  suggested by the workgroup, EPA has provided additional guidance, information,
and clarification on the uncertainties associated with the use of the food web models to

                                           39

-------
determine FCMs.

    Issue 2:  Available approaches and data to account for metabolism in the
    determination of a BAF value, and to predict food chain multipliers.

    Metabolism

The workgroup confirmed EPA's assertion that no reliable, broadly applicable approach exists
to predict the metabolic breakdown of organic chemicals by biota, and that this is a significant
limitation in the state of the science and affects the proposed methodology. Rather than assume
no metabolism for all organic chemicals, the workgroup recommended that EPA develop a
chemical grouping scheme and guidance for circumstances in which an assumption of no
metabolism is reasonable (e.g., nonplanar PCBs, several chlorinated pesticides) and in which
complete loss of a chemical via biotransformation could be assumed (e.g., many aromatic
hydrocarbons).

This suggestion is a reasonable idea in theory, but the general lack of metabolism data prevents
implementation of such an approach. Data do not exist for either individual chemicals or
chemical classes, or for the metabolic abilities of individual organisms.  Generally, invertebrate
species (e.g., muscles, clams, benthic invertebrates, lobster, shrimp, and crabs) tend to have
much lower metabolic abilities than vertebrate  species (e.g., fish in aquatic food webs).
Although many of the users of these guidance documents are focused on fish as their target
species, it is important to note that on average invertebrate aquatic species compose a large
portion of the human diet.  Such organisms include shrimp, crabs, lobster, scallops, and clams,
and most of these organisms  have substantially lower metabolic abilities than vertebrate species
like fish. Given that these methods are for the protection of human health, EPA cannot ignore
the invertebrate species in the determination of bioaccumulation potential for the chemical of
interest. In some cases, bioaccumulation potential might be fairly small in fish because of
metabolism processes for a chemical, whereas, in contrast, bioaccumulation potential might be
very large in invertebrate species because these organisms do not possess the metabolic
pathways or have  substantially lower metabolic abilities for metabolizing the chemical.

EPA has developed a table to be put in the TSD for chemicals that are not substantially
metabolized or are very slowly metabolized.  This table in all likelihood contains no false
positives (i.e., chemicals that  are on the list but are easily metabolized) is not all-inclusive
because there are numerous chemicals (e.g., hundreds of thousands in use commercially today)
for which few or no metabolism data exist.

We disagree with the workgroup  that a table of completely metabolized chemicals can be
developed. This belief is based on the lack of whole-organism metabolic rate data for fish and
other aquatic species, the lack of metabolic rate databases of any type for any species, and the
general inability to extrapolate from in vitro studies using liver microsomes, cells, or organ slices
to whole organism rate constants.  In addition, predictions from QSAR relationships based on  in
vitro data have extremely large uncertainties.  The following example for PAHs, which are
suggested by the reviewers as being completely metabolized in vertebrate species, illustrates the

                                           40

-------
difficulties in developing a table of completely metabolized chemicals. Burkhard and
Lukasewycz (2000) reported log of field-derived BAFs on a freely dissolved and lipid
normalized basis versus the log of the octanol-water partition coefficient (Kow) for
phenanthrene, fluoranthene, pyrene, benz[a]anthracene, and chrysene/triphenylene in lake trout.
Their data showed that bioaccumulation did occur for all five PAHs because all of the log BAFs
were greater than zero, even though the concentrations in the fish were very low (i.e., 0.06 to
2.9 ppb). If no metabolism  occurred, the log BAF should be equal to the chemical's log Kow
(ignoring biomagnification processes).  The ratios of the field-measured log BAFs to log KQWS
ranged from 0.2% to 34%, suggesting that some metabolism did occur.  Pyrene's log BAF was
34% of its log KQW, suggesting that partial metabolism but not complete metabolism occurred
for this chemical. This example highlights the difficulties faced in constructing a table for
completely metabolized chemicals, because even for a class of chemicals believed to be
completely metabolized, field data suggest otherwise.  If the completely metabolized table were
limited to only those chemicals with known metabolic data, this table would in all likelihood
contain very few entries. If we were  to use expert opinion, advice, or best scientific judgment,
we believe that numerous false positives would be present because of the lack of data; that is,
chemicals that are not completely metabolized and would likely not be defensible from a public
health protection perspective. In view of these difficulties, EPA does not agree with the
workgroup's suggestion of developing a table of completely metabolized chemicals.

The workgroup suggested that an alternate methodology that relied on chemical-specific or
species/trophic level-specific elimination rates might be used for addressing metabolism.
However, it was acknowledged that this approach would be resource-intensive and that much of
the information would be direct measures of organism-level metabolic rates (e.g., enzyme
induction, metabolite structures).  The workgroup suggested that such an approach might be
used to screen or prioritize chemicals most appropriate for AWQC derivation using the current
methodology.

The workgroup suggested that research initiatives on the extrapolation of organism-level
metabolic rates to whole organism rates and species to species extrapolations (e.g., from rats to
fish) be initiated. In the future, once a sufficient body of knowledge from such efforts becomes
available, EPA will consider revising  these guidance documents to include the results of the
investigations.

We disagree with the workgroup's suggestion that using direct measures of organisms' metabolic
rate to evaluate bioaccumulation potential for screening or prioritizing chemicals for AQWC
derivation is practical now or advisable. EPA will continue to use the risk-based approach for
selecting chemicals to derive AWQC, rather than just relying on bioaccumulation potential,
because other factors can contribute to potentially significant health risks for individual chemicals.

   Issue 3: Any other available models that EPA should consider for inclusion in the
   revised methodology for estimating bioaccumulation.

   Alternative Models
                                          41

-------
The alternative models available for assessing bioaccumulation were judged by the workgroup to
all have similar structure, assumptions, and limitations. The differences were not thought to be
important on a national scale, but might be for a specific site.  The workgroup thought that
additional modifications could be made to address some of the model assumptions (e.g.,
sediment disequilibrium, metabolism) in addition to the use of isotopes for empirically modeling
residues.

We agree with the workgroup that there is a general lack of alternative models that differ in their
structure and assumptions for estimating bioaccumulation. In the revised guidance, we have
allowed appropriately validated alternative models to be used  on a site-specific basis. EPA
notes that the proposed model and other available models include explicit consideration of
sediment-water disequilibrium and include the capability to account for metabolism of the
chemical by the inclusion of the first-order, whole organism metabolism rate constants, km.
When no metabolism is assumed, k,,, is set equal to zero.

The workgroup suggested that use of stable isotopes for nitrogen and carbon (e.g., • 15N or
• 13C) might be useful in modeling chemical residues.  EPA does not believe that • 15N or • 13C
signatures can be used to predict chemical residues in  a given ecosystem for a given organism,
because chemical residues in fish and other aquatic organisms are a function of the chemical
loading to the ecosystem from past and current practices. However, as demonstrated by
Cabana and Rasmussen (1994) using mercury and Canadian shield lakes, it may be feasible to
perform this prediction with a reasonable degree of accuracy for ecosystems of similar nature
and loading patterns, assuming  a predictive relationship can be developed for other chemicals.

EPA believes that • 15N or • 13C signatures are extremely useful in establishing food chain length,
trophic levels status of individual organisms, and food web structure or function for specific
ecosystems. Traditional methodologies for determining these characteristics in  aquatic food
webs such as visual observation, gut analysis, and professional interpretation of expected feeding
interactions are all very difficult and have high uncertainties. In addition, • 15N or • 13C signatures
provide a time course integration of dietary consumption patterns, whereas traditional
methodologies (e.g., gut analysis) represent dietary consumption patterns for single moments in
the organism's life.  In the past, if food web structure  was not known, a linear food web
structure was assumed in modeling effects (e.g., Thomann 1989).  However, in more recent
modeling efforts, Morrison et al. (1997) used food web structures that represent actual dietary
consumption patterns. • 15N or • 13C signatures allow one to determine trophic level status of
organisms on a continuous scale rather than the lumping of organisms into general categories as
previously done, that is, trophic levels 1, 2, 3, or 4.  Numerous investigators, including Bromann
et al. (1992), Cabanna and Rasmussen (1994), Kiriluk et al. (1995), Kucklick et al. (1996), and
Kidd et al. (1998), have demonstrated that • 15N signatures are well correlated with
biomagnification of PCBs, DDE, PCDD/Fs, and mercury in specific aquatic food webs. Their
results suggest that stable isotope data of nitrogen may be useful in estimating biomagnification
factors for nonmetabolizable hydrophobic chemicals such as PCBs and DDE.

Minigawa and Wada (1984) have reported that 3.4%o enrichment of • 15N should be expected
on average  for a predator consuming a prey with a constant • 15N signature.  Vander Zanden

                                           42

-------
and Rasmussen (1996) reported that omnivory feeding behavior often causes differences in the
• 15N signatures to be smaller than 3.4%o between lake trout and their primary prey, forage fish.
Vander Zanden and Rasmussen (1996), when calculating BMFs for a full trophic level (i.e.,
+3.4%o enrichment of • 15N) observed much larger BMP estimates than those ignoring omnivory
behavior. EPA believes that • 15N or • 13C signatures will, in all likelihood, be a very useful tool
for estimating BMFs for nonmetabolizable hydrophobic chemicals in the future.  The scientific
underpinnings of this tool are under active research and until further developments are made, we
believe that using this tool for determining BMFs for use in the derivation of national AWQC
would be premature.

   Issue 4:  Whether the draft BAF methodology is an improvement over the 1980
   methodology and, in particular, whether it is likely to be more predictive of
   bioaccumulation.

    1980 Versus 1998 draft Methodology

The workgroup stated that the 1998 draft methodology is a theoretical improvement over the
1980 methodology and is more predictive for the chemicals and sites referenced in the TSD.
However, the workgroup said that it is not known whether the 1998 draft methodology is more
predictive than the 1980 methodology for other chemicals and sites and that further verification
and comparisons are needed to address this issue. The workgroup also commented that the
1998 draft methodology requires many assumptions and measurements and, as proposed in its
current form (1998), contains an aggregate uncertainty that is too high for broad regulatory
application.

We agree with the reviewers that the draft 1998 methodology represents a theoretical
improvement over the 1980 methodology because it emphasizes a more explicit and systematic
assessment of bioaccumulation (i.e., chemical accumulation from water, diet, sediment)
compared with the 1980 methodology, which emphasizes the assessment of bioconcentration
(i.e., uptake from water only). [It should be noted that the 1980 AWQC Guidelines do allow for
the use of "field-BCFs" (currently termed field-BAFs) when such values are substantially higher
or lower than laboratory-measured or Kow-estimated BCFs. Guidance on the use of "field
BCFs" is very limited in the 1980 guidelines.]  EPA notes that consideration of dietary and other
sources in addition to water has been shown to be very important for many persistent, highly
bioaccumulative pollutants of concern (e.g., Russell et al. 1999; Burkhard et al. 1997; Oliver and
Niimi 1983,  1988; Niimi 1985; Swackhamer and Kites 1988;  Watras and Bloom 1992; U.S.
EPA 1997).   We  recognize that evaluation of various aspects of the draft 1998 methodology
(such as predicted BAFs using Kow, food chain multipliers, and BSAFs) has focused on
persistent, hydrophobic organic chemicals including PCBs, chlorinated pesticides, PCDDs, and
PCDFs in relatively few (but diverse) ecosystems. These systems include the hydrodynamically
complex and tidally influenced area of Bayou d'Inde, Louisiana; the more stable, oligotrophic
system of Lake Ontario; and by comparison to Lake Ontario, the more shallow, eutrophic
system of Green Bay, Lake Mchigan.  The primary reason for this focus is the limited availability
of high-quality field data on bioaccumulation from which to draw such comparisons. Such high-
quality field data include studies that measure contaminants in water, sediment, and the food web

                                          43

-------
over appropriate time scales in addition to other measurements such as organic carbon and lipid
fraction.  We also note that other aspects of the methodology (e.g., Kow-based estimates of
bioconcentration factors) have been tested extensively in the scientific literature for numerous
organic chemicals (Vieth et al. 1979; Oliver andNiimi 1983; Mackay 1982; Chiou 1985; and
others) and have been used in developing AWQC for nearly two decades.  Nevertheless, we
agree that additional testing of various aspects of the methodology is desirable and have
provided further evaluation of the national BAF methodology using two data sets (Fox River and
Hudson River). We believe that these sites are appropriate for additional comparisons because
they represent lotic systems that would likely differ in their ecological and hydrological
characteristics compared with the sites already examined.

Regarding the uncertainty in the draft 1998 bioaccumulation methodology, we have made
extensive revisions to the methodology that we believe address many of the workgroup's
concerns and reduce overall uncertainty in BAF estimates.  These changes include the following:
(1) development of separate procedures for deriving BAFs for different chemical classes (e.g.,
high versus low hydrophobicity, high versus low metabolism in biota, ionic versus nonionic
organics); (2) simplification of procedures for organic chemicals with low hydrophobicity; (3)
recommending that KoW-based estimates of BAFs and food chain multipliers not be used for
nonionic organics that are known to metabolize substantially in targeted biota (e.g.,
benz[a]pyrene in fish); and (4) restricting the use of the BSAF methodology. It should be noted
that many of the uncertainties raised by the workgroup are also present in the  1980
methodology, which is oriented toward assessment of bioconcentration factors using lab-BCFs
and Kow-based predictions.  Furthermore, in the  1980 guidance no explicit guidance exists on
modifying the national default BCFs for site- or region-specific concerns.  Therefore, we believe
that the revised 1998 bioaccumulation methodology represents a substantial improvement over
the 1980 methodology for assessing bioaccumulation for deriving AWQC.

The workgroup commented that the uncertainties in the 1998 draft methodology would likely
overestimate the BAF but were not certain by how much.  We agree with the  workgroup that
some of the procedures of the 1998 draft methodology (e.g., Kow and FCM-predicted BAFs)
might lead to overestimates of BAFs for certain types  of pollutants, such as those that are
metabolized substantially to chemical forms not addressed by the AWQC. However, we
disagree with the comment that implies that, in general, EPA's draft methodology would lead to
across-the-board overprediction of BAFs.  Field BAFs, the first tier in the data preference
hierarchy, represent direct measures of bioaccumulation.  We are not aware of any reason why
the treatment of uncertainty or variability in such field BAF estimates would consistently lead to
across-the-board overestimates of bioaccumulation, regardless of the pollutant type, since such
values are based on central tendency estimates within and across species of a given trophic level.
Similarly, the calculation of freely dissolved concentration for nonionic organic  chemicals using
the three-phased partitioning model  is based on central tendency estimates of input parameters.
We also know of no reason why the treatment of uncertainty and variability in laboratory-
measured BCFs would result  in consistent overestimates of BAFs.  For highly hydrophobic
contaminants that do not metabolize substantially  in tissues, use of the BSAF-predicted and Kow
x FCM-predicted BAFs also does not appear to be biased toward overestimating field-
measured BAFs. This is demonstrated by  comparisons made in the TSD (Exhibits 2.4.1, 2.4.3,

                                          44

-------
and 2.4.6) for BSAFs and those made by Burkhard et al. (1997) for the Kow x FCM method
(Figure 2).

The workgroup questioned how AWQC will (or could) be linked to sediment quality criteria
(SQC) and inquired whether there will be SQC for the protection of human health. EPA agrees
that sediments can serve as important sinks and sources for waterborne contaminants, which if
not considered can lead to unacceptable ecological and human health risks. EPA is nearing
completion of the first equilibrium-partitioning sediment guidelines (ESGs) for PAH mixtures,
cationic metals, and dieldrin/endrin and expects to publish these guidelines in late spring of 2000.
EPA notes that the partitioning theory of nonionic organic chemicals used to calculate ESGs is
the same as that used in EPA's draft bioaccumulation methodology, which helps to ensure
consistency in the two approaches. It should be noted that to the extent that contaminated
sediments are contributing to bioaccumulation in aquatic food webs, the draft (and subsequently
revised) procedures for measuring BAFs and BSAFs in the field do account for this exposure at
the sites from which the BAFs and BSAFs are measured. Finally, although we are not currently
planning to develop human health AWQC that are solely based on sediment concentrations, we
are in the early stages of conceptualizing environmental criteria that integrate  exposure from
multiple sources, including sediments.  These integrated criteria would also evaluate risks to
multiple receptors (aquatic life, wildlife, human health) simultaneously.  Thus, EPA's desire is to
develop future criteria that are truly comprehensive in their exposure and receptor evaluations
and rely on a consistent set of methodologies that reflect the current state of the science.

2. Response to Issues for Public Comment Listed in the Federal Register (7.2)

   Issue 1: Is the suggested hierarchy for developing BAFs appropriate? Are  there any
   alternatives to the four methods that could be used to derive AWQC?

   Tiered Hierarchy

The workgroup considered the hierarchy to be acceptable for a site-specific analysis (given
more explicit guidance), but not for use on a national level as was originally proposed.  They
further recommended that EPA make clear that although the national BAF can be derived based
on a reliable field-measured BAFs with limited regional coverage, site-specific studies are
allowed if the national BAF appears to not be representative of that site.

As we discussed in our responses in Section 1 above, we have made substantial changes to the
1998 draft methodology in regard to general concerns about using BAFs, document readability
and scale of application, and the 1980 versus new methodology, which we believe addresses
many of the peer reviewers' concerns and resulted in an improved methodology for assessing
bioaccumulation that can be implemented effectively throughout the United States. The revisions
include improvements in the readability and clarity of the guidance, such as separating EPA's
guidelines for deriving national BAFs from its guidelines for deriving site- or region-specific
BAFs. We have also simplified parts of the draft methodology for certain types of pollutants
where the benefits of the added complexity (i.e., improved accuracy) are not likely be realized.
For example, for nonionic organic chemicals of low hydrophobicity, the methodology has been

                                          45

-------
revised to give equal consideration to the use of laboratory-measured BCFs and field-measured
BAFs since the benefits of field-measured BAFs over laboratory-measured BCFs would likely
be marginal. For the same reason, the requirement to derive separate, trophic level-specific
BAFs for low Kow chemicals has been relaxed.  Changes were also made to address the
workgroup's concerns about uncertainty in certain key areas, such as placing restrictions on the
use of Kow- and B SAP -predicted BAFs for certain types of pollutants.

Regarding the use of site-specific studies to modify national BAFs, EPA emphasized in the 1998
draft methodology, and will continue to encourage, that such studies are appropriate and
recommended in situations where there are concerns about the representativeness of a national
BAF. We are developing detailed procedures for designing and conducting field BAF studies,
which are scheduled for completion within the year following the publication of the revised
AWQC methodology.  The guidance for conducting field BAF studies addresses  differing
chemical properties of the pollutant (e.g., KQW) and site characteristics that can impact
uncertainty in BAF measurements. We recognize (and have made clear in the revised
methodology) that because of current limitations in availability of high-quality field studies, some
national BAF values might be based on results from a few field studies, which might represent to
varying degrees different sites around the United States. However, EPA notes that although
such national BAFs may be directly  supported by a few studies, indirect support  (and greater
confidence) can be derived by comparing the results of BAF estimates using other tiers of the
methodology (e.g., BSAF, laboratory-measured BCFs, Kow x FCM-predicted BAFs).  Thus,
in some cases, a field-measured BAF may be supported by multiple lines of evidence. In other
cases, uncertainties in a field-measured BAF may outweigh its preference to BAFs derived from
the lower tiers. We have provided additional text to better emphasize the assessment of
uncertainty in field-measured BAFs (and BAFs derived using other tiers) when deriving national
BAFs.

Finally, regarding limiting the guidance to site-specific application, EPA notes that the current
304(a) AWQC have been implemented on a national scale for nearly two decades. As
discussed under Section 1, Issue 1, in our response to the workgroup's comments on default
lipid values, we believe that restricting the bioaccumulation methodology only to the development
of site-specific BAFs would greatly hinder the implementation of water quality criteria throughout
the United States.  We believe this would be the case because many States, Territories, and
authorized Tribes lack the resources to develop site-specific BAFs for all of their pollutants of
concern and subsequent adoption of AWQC would be delayed substantially.

Although the BSAF approach may be reasonable for the chemicals examined in the TSD, the
workgroup thought that it was not reliable for all  chemicals and that it should be dropped from
the proposed hierarchy.  Specifically, the workgroup made the following comments: (1) the
BSAF approach ignores differences in gut assimilation efficiency, metabolism, and bioavailability
from sediment; (2) BSAFs have not been widely validated for general use with organic
chemicals; and (3) the relative contribution of food and water routes to the BSAF vary with
EPA believes that the BSAF method for determination of BAFs is valid and is needed for

                                          46

-------
chemicals with nondetectable or difficult-to-predict concentrations in water.  We agree that the
BSAF method should not be used for all organic chemicals that may be addressed through the
human health methodology.  Although proper choice of reference chemicals to match properties
of less hydrophobic target chemicals should allow the BSAF method to work for a wider range
of chemicals, EPA has restricted application of the method to the following: (1) chemicals that,
because of their chemical properties, cannot be measured or are very difficult to measure in
water; and (2) chemicals that perhaps could be measured but have not been, yet need an
assessment of bioaccumulation. We have also provided more specific guidance on selection of
reference chemicals and use of multiple reference chemicals to secure the most accurate estimate
of a chemical's BAF.

We do not agree that the BSAF method ignores differences between chemicals in their gut
assimilation efficiency, metabolism, and bioavailability from sediment. The ability to measure
these differences through BSAFs for chemicals without directly measurable BAFs is precisely
why the BSAF method was proposed.  Gut assimilation efficiencies for nonionic organic
chemicals with log KQWS • 4.0 are uniformly above 80% (e.g., Nichols et al, 1998).  The
cumulative effects of metabolism of the chemical in the food chain on the chemical concentration
in the organism are incorporated in the BSAF in the same manner as in a measured BAF (same
numerator). Bioavailability differences between organic chemicals are a function of their log
Kows and are most important for log KQWS •  4.0. Based on the critical condition that
• socw/K0w for both chemicals are similar, the BSAF method has been shown to accurately
predict BAFs in two different Great Lakes ecosystems. The factor • socw is  a ratio that
represents the disequilibrium between the concentration of a chemical in sediment (normalized
for organic carbon content) and water.  However, EPA agrees that the BSAF method could be
bolstered with further validation.  Thus, we have added more validation in the revised TSD using
new data sets that meet the water, surface sediment, and biota sampling and  analysis
requirements.

We agree that relative contributions of food, water, and sediment routes of exposure to BSAFs
(and BAFs) vary with Kow.  Although a BSAF indexes the concentration of a chemical in fish to
concentration in sediment, rather than water as for the BAF, both have a common numerator
that measures the sum of all routes of exposure. Thus, both BSAFs and BAFs vary with Kow,
depending on the relative contribution of food, water, and sediment, and the BSAF method
accounts for this variation in the same way that measured BAFs do.

The workgroup noted that "for the chemicals examined (persistent and bioaccumulative),
extrapolation to other circumstances may be reasonable." We believe that restricting the use of
the BSAF method to highly hydrophobic chemicals difficult to measure in water, clarifying the
use of reference chemicals, elaborating on the primacy of the sediment-water fugacity
equivalence condition for use of the method, and validation with additional data sets have
alleviated the workgroup's concerns about use of this new method.  Finally, it should be noted
that use of the BSAF method, by incorporating the chemical-specific effects of bioavailability
and metabolism into the BAF estimate, will allow measurement of BAFs that are significantly less
than those predicted from a BCF, a Kow with a food chain multiplier, or a food web model that
assumes no metabolism.

                                         47

-------
The workgroup commented that it is unclear whether it is EPA's intent to take the geometric
mean of the geometric mean species BAFs for calculating the trophic-level BAF or if this
calculation is correct. The goal of EPA's national 304(a) AWQC is to be protective of public
health generally.  Accordingly, EPA's national 304(a) AWQC values are derived with
parameter estimates for dose-response assessment that are upper bounds in exposure
assessment parameter estimates using a combination of mean values (e.g., body weights, non-
fish dietary intakes) and upper percentile values (e.g., drinking water and fish consumption rates)
to provide an overall high-end (conservative) public health risk estimate. EPA determines its
estimates of the BAF on central tendency estimates. When variability in BAFs occurs within or
across species of a trophic level, it is EPA's intent to express the species-mean BAF as the
geometric mean of acceptable species BAF values.  Likewise, it is EPA's intent to express the
trophic-level mean BAF as the geometric mean of the species-mean BAF values. Given the
limited data typically available for field-measured BAFs and laboratory-measured BCFs, we
believe that this procedure is appropriate. We recognize that ideally, one would want to weight
each species-mean BAF by the extent to which that species represents the likely dietary
exposure of the target population.  However, as described on pages 239-256 of the  1998
TSD, consumption rate information is not available on a national scale at the individual species
level. Therefore, each species-mean BAF value is weighted equally within a trophic level for the
purposes of deriving a trophic level mean BAF.

After reviewing this issue further, we believe that States, Territories, and authorized Tribes may
wish to weigh the contribution of one species' BAF to a greater or lesser extent than a BAF for
another species based on State or site-specific data.  Therefore, in the guidelines for developing
site-  or region-specific BAFs, we have provided additional guidance on this issue.

   Issue 2: Is the procedure for estimating the consumption-weighted lipid value of 2
   percent for aquatic species eaten by humans and the data used for  deriving the value
   appropriate? Are there other data available that could be used to calculate the  default
   lipid value?

   National Default Lipid Value

The workgroup considered the procedure used to derive the national default value as  reasonable
but cited concerns expressed earlier (see Section 1, Issue 1, "Default Lipid  Value").  EPA's
response to these concerns appears under Section 1, Issue 1, "Default Lipid Value."

   Issue 3: Are there alternatives to the equation used to derive the freely dissolved
   fraction of a chemical appropriate? If yes, what data support and alternative
   approach? Are there scientifically defensible alternatives to EPA's  Kow-based estimate
   ofKDOC andKPOC?

   Freely Dissolved Fraction Equation

The general approach chosen by EPA to estimate the freely dissolved fraction of a chemical was
considered by the workgroup to be the most appropriate one. However, the workgroup stated

                                          48

-------
that EPA needs to make modifications to these equations and provide a more explicit description
of the underlying assumptions and potential magnitude of associated error as discussed above in
Section 1, Issue 1, "Freely Dissolved Fraction." EPA's response to this comment appears
above in Section 1, Issue 1, under "Freely Dissolved Fraction."

   Issue 4: Are the default POC value ofO. 48 mg/L and the default DOC value of 2.9
   mg/L used in deriving BAFs appropriate as national defaults? Are the water body- and
   State-specific POC and DOC values provided in the TSD appropriate? Are there
   additional data that could be used to derive these  values?

   Default DOC and POC Values

The workgroup considered the national defaults as appearing to provide valid mean values of
dissolved organic carbon (DOC) and particulate organic  carbon (POC).  However, the
workgroup made several suggestions to expand on these values: (1) provide a measure of
variability around the means (e.g., confidence limits); (2) provide more detailed data in an
appendix or separate document (including State-specific values); (3) provide additional guidance
on the potential for high spatial and temporal variability in DOC and POC; and (4) provide
appropriate analytical methodologies for measuring DOC and POC.

We agree with the workgroup that the national default values for DOC and POC described in
the 1998 draft methodology provide reasonable estimates of mean values. Although in the 1998
draft TSD we provided some measure of variability around these mean estimates (e.g., standard
deviation) both across and within water-body types, we agree with the reviewers that a more
thorough characterization of variability in organic carbon values would be desirable. In the
revised 1998 draft methodology, the analysis of DOC and POC  data to has been expanded to
include other measures of variability (e.g., percentiles, confidence limits) and provide a
breakdown of estimates by State and where data allow, by water-body type within a State.
EPA has also decided to made the DOC and POC database available to States, Territories, and
authorized Tribes for use in modifying national  default estimates on a State/Tribal or local basis.
A more complete characterization of the DOC and POC  database has been provided in BAF
TSD.  Additional guidance will also be provided in the forthcoming bioaccumulation field plan
document on measuring DOC and POC, including the selection of analytical methodologies and
addressing temporal and spatial variability.
   Issue 5: What approaches could be used to account for metabolism in the
   determination of a BAF and what data are available to support these approaches?

   Metabolism

See EPA's response to this issue above under Section 1, Issue 2, "Metabolism."

   Issue 6. What other models are available that could be used to predict FCMs? What
   are the data that support these models? Is EPA's choice of food web structures used to

                                         49

-------
    calculate FCMs appropriate?

    Models and FCMs

The workgroup did not offer other models for EPA's consideration but did comment that
flexibility is necessary in the selection of model parameters to enable site-specific adjustments to
be made. They further noted that guidance on doing this is lacking in the proposed methodology
and lacking in general, thus making the development of science-based guidance on FCMs
difficult and their application on a national scale potentially inappropriate. The workgroup
recommended the generation of additional data to refine and validate food web models, including
the use of isotopic studies for analyzing food web structures.

As we discussed earlier, in Section 1, Issue 1, under "Food Chain Multipliers," we strongly
recommend wherever possible that site-specific FCMs be determined using site-specific food
web parameters.  The revised site-specific guidance TSD provides guidance on how one should
assess the diets, lipid contents, and weights of organisms composing the food web, and
important environmental parameters (e.g., sediment water column disequilibrium) used in
calculating FCMs.  Guidance has also been provided for calculating the FCMs in the revised
site-specific portion of the TSD.

As we discussed earlier, in Section 1, Issue 3, "Alternative Models," we believe nitrogen and
carbon isotopic signatures (i.e., • 15N or • 13C) are extremely useful for establishing food web
structures and trophic levels of individual organisms. However, EPA believes the science
supporting the use of • 15N or • 13C signatures is still developing and not ready for use in a
regulatory program.  Isotopic signatures may be a very useful tool for estimating BMFs for
nonmetabolizable hydrophobic chemicals in the future when the scientific underpinnings of this
tool are further clarified.

Consideration of the Electric Power Research Institute (EPRI) mercury food chain model was
recommended by the workgroup.  EPA is currently  in the process of revising its ambient human
health criteria for methylmercury. As part of this process, EPA may consider the use of the
EPRI methylmercury bioaccumulation model and other models for use in estimating mercury
bioaccumulation.  Currently, these models have the greatest appeal for use on a site-specific
basis.

    Issue 7: Is EPA's Guidance on selecting reproducible Kow values appropriate? Which
    of the two options for selecting reproducible Kow values do you consider the most
    appropriate?

    Selecting Kr.w Values

The second (more detailed) option for selecting the Kow value was preferred by the workgroup,
although they cautioned that the molecular fragment method should be used as a last resort.
Additionally, the workgroup thought that a general consensus on Kow values already exists in the
scientific community (e.g., Mackay et al. 1999) and that EPA should simply publish a list of

                                           50

-------
recommended values.

The second method for selecting the Kow values has been retained in the revised methodology
because EPA believes that it provides more complete guidance on selecting KQW values for use
in AWQC derivation. EPA further agrees that it is desirable to publish a list of KQW values that
have been selected using the Kow selection methodology. We plan to publish a list of such
chemicals following issuance of our revised AWQC methodology.

We agree with the cautionary note by the workgroup about the molecular fragment method. The
Kow selection protocol uses a weight-of-evidence approach, where "assigning a Kow ... will
necessarily involve scientific judgment in evaluating not only the reliability of all data inputs but
also the accretion/concretion of evidence in support of the recommended KQW value." The
molecular fragment method is included in the KQW  selection protocol for circumstances where
disagreement exists among Kow estimation methods (e.g., ClogP, LOGKOW, and SPARC
Kow  estimation computer programs), when Kow  measurements differ substantially from
predictions using the Kow  estimation methods, and when an absence or scarcity of reliable data
exists.  One good example of possible use of the molecular fragment method is for the chemical
photomirex, for which KQW estimation methods disagree and no measurements have been
made.  Using the molecular fragment method with mirex, an estimate can be derived for
photomirex.  This estimate then provides additional information to be used in the selection of the
recommended Kow value.

   Issue 8: Should properly derived field-measured FCMs take precedence [over] FCMs
   derived using the Gobas (1993) model?

   Field-Based FCMs

In general, the workgroup preferred the use of field-measured FCMs over model-derived
FCMs. However, they emphasized that proper evaluation of these two approaches would
require a  more complete comparison and encouraged EPA to perform such an analysis.  EPA
agrees that in theory, use of field-measured FCMs would be preferred to the use of model-
derived FCMs (which assume no metabolism), particularly for contaminants where metabolism is
of concern. Field-derived FCMs reflect any metabolism that occurred in the food web.
However, EPA believes that effective use of field-derived FCMs requires knowledge of the
food web structure at the site(s) from which they are obtained in order to derive valid FCMs.
More detailed guidance has been added to the revised methodology on the use of field-derived
FCMs.

References

Bligh EG, Dyer WJ.  1959. A rapid method of total lipid extraction and purification. Can J
Biochem Physiol 37:911-917.

Broman D, Naf C, Rolff C, Zebuher Y, Fry B, Hobbies J.  1992. Using ratios of stable nitrogen
isotopes to estimate bioaccumulation and flux of poly chlorinated dibenzo-p-dioxins (PCDDs)

                                         51

-------
and dibenzofurans (PCDFs) in two food chains from the northern Baltic. Environ Toxicol Chem
11:331-345.

Burkhard LP. 1998. Comparison of two models for predicting bioaccumulation of hydrophobic
organic chemicals in a Great Lakes food web. Environ Toxicol Chem 17(3):383-393.

Burkhard LP, Sheedy BR, McCauley DJ, DeGraeve GM.  1997.  Bioaccumulation factors for
chlorinated benzene, chlorinated butadienes and hexachloroethane.  Environ Toxicol Chem
16(8):1677-1686.

Cabana G, Rasmussen JB.  1994. Modeling food chain structure and contaminant
bioaccumulation using stable nitrogen isotopes.  Nature 372:255-257.

Chefurka W, Gnidec EPP. 1987. Binding of [14C]DDT by submitochondrial particles. Comp
Biochem Physiol 88C:213-217.

Chiou CT. 1985. Partition coefficients of organic compounds in lipid-water systems and
correlations with fish bioconcentration factors. Environ Sci Technol 19:57-62.

Fisk AT, Norstrom RJ, Cymbalisty CC,  Muir DCB. 1998. Dietary accumulation and depuration
of hydrophobic organochlorines: bioaccumulation parameters and their relationship with the
octanol/water partition coefficient. Environ Toxicol Chem  17(5):951-961.

Gobas FAPC.  1993. A model for predicting the bioaccumulation of hydrophobic organic
chemicals in aquatic food-webs: application to Lake Ontario. Ecol Modeling 69:1-17.

Kara A,  Radin NS. 1978. Lipid extraction of tissues with a low-toxicity solvent. Anal Biochem
90:420-426.

Kidd KA, Hesslein RH, Ross BJ, Koczanski K, Stephens GR, Muir DCG.  1998.
Bioaccumulation of organochlorines through a remote freshwater food web in the Canadian
Arctic.  Environ Pollut 102:91-103.

Kiriluk RM, Servos MR, Whittle M, Cabana G, Rasmussen JB.  1995. Using ratios of stable
nitrogen and carbon isotopes to characterize the biomagnification of DDE, mirex, and PCB in a
Lake Ontario pelagic food web.  Can J Fish Aquat Sci 52:2660-2674.

Kucklick JR, Harvey HR, Ostrom PH, Ostrom NE, Baker JE. 1996.  Organochlorine
dynamics in the pelagic food web of Lake Baikal.  Environ Toxicol Chem 15(8): 1388-1400.

Mackay D. 1982. Correlation of bioconcentration factors. Environ Sci Technol 16:274-278.

Minigawa M, Wada E.   1984.  Step wise enrichment of 15N along food chains: further evidence
and the relation between 15N and animal age. Geochim Cosmochim Acta 48:1135-1140.
                                         52

-------
Morrison HA, Gobas FAPC, Lazar R, Whittle DM, Haffner GD.  1997. Development and
verification of a benthic/pelagic food web bioaccumulation model for PCB congeners in western
Lake Erie.  Environ Sci Technol 31(11):3267-3273.

Nichols JW, Jensen KM, Tietge JE, Johnson RD.  1998.  Physiologically based toxicokinetic
model for maternal transfer of 2,3,7,8-tetrachlorodibenzo-p-dioxin in brook trout (Salvelinus
fontinalis). Environ Toxicol Chem 17:2422-2434.

Niimi AJ.  1985.  Use of laboratory studies in assessing the behavior of contaminants in fish
inhabiting natural ecosystems.  Wat Poll Res J Can 20:79-88.

Oliver BG, Niimi AJ. 1988. Trophodynamic analysis of poly chlorinated biphenyl congeners
and other chlorinated hydrocarbons in the Lake Ontario ecosystem.  Environ Sci Technol
22:388-397.

Oliver BG, Niimi AJ. 1985. Bioconcentration factors of some halogenated organics for
rainbow trout: limitations in their use for prediction of environmental residues.  Environ  Sci
Technol 19:842-849.

Oliver BG, Niimi AJ. 1983. Bioconcentration of chlorobenzenes from water by rainbow trout:
correlations with partition coefficients and environmental residues.  Environ Sci Technol 17:287-
291.

Randall RC, Young DR, Lee JJ, Echols SF.  1998.  Lipid methodology and pollutant
normalization relationships for neutral nonpolar organic pollutants.  Environ Toxicol Chem
17(5):788-791.

Randall RC, Lee H JJ, Ozretich RJ, Lake JL, Pruell RJ.  1991.  Evaluation of selected lipid
methods for normalizing pollutant bioaccumulation. Environ Toxicol Chem 10: 1431-1436.

Russell RW, Gobas FAPC, Haffner GD.  1999. Role of chemical and ecological factors in
trophic transfer of organic chemicals in aquatic food webs. Environ Toxicol Chem 18(6): 1250-
1257.

Swackhamer DL, Kites  RA. 1988. Occurrence and bioaccumulation of organochlorine
compounds in fishes from Siskiwit Lake,  Isle Royale, Lake Superior.  Environ Sci Technol
22:543-548.

Thomann RV. 1989.  Bioaccumulation model of organic chemical distribution in aquatic food
chains. Environ Sci Technol 23:699-707.

U.S. Environmental Protection Agency.  1998.  Water quality criteria methodology: human
health.  Technical Support Document.  Office of Water, Washington, DC.  July.  EPA-822-B-
98-005.
                                          53

-------
U.S. Environmental Protection Agency.  1997.  Mercury study report to Congress, vol. HI: fate
and transport of mercury in the environment. Office of Air Quality Planning and Standards,
Office of Research and Development, Washington, DC. December. EPA 452/R-97-005.

U.S. Environmental Protection Agency.  1995. Great Lakes water quality initiative technical
support document for the procedure to determine bioaccumulation factors.  Office of Water,
Washington, DC. March. EPA 820-B-95-005.

U.S. Environmental Protection Agency.  1994.  Method 1613. Tetra- through octa-chlorinated
dioxins and furans by isotope dilution HRGC/HRMS. Washington, DC. EPA-621-B-94-005.

U.S. Environmental Protection Agency.  1991.  Technical support document for water quality-
based toxics control. Office of Water, Washington, DC.  March. EPA- 505-2-90-001.

U.S. Environmental Protection Agency.  1980. Analysis of pesticide residues in human and
environmental samples: a compilation of methods selected for use in pesticide monitoring
programs.  Research Triangle Park, NC. EPA-600/8-80/038.

Vander Zanden MJ, Rasmussen JB.  1996. A trophic position model of pelagic food webs:
impact on contaminant bioaccumulation in lake trout. Ecol Monogr 66(4):451-477.

Vieth GD, DeFoe DL, Bergstedt BV. 1979. Measuring and estimating the bioconcentration
factor of chemicals in fish. J Fish Res Board Can 36:1040-1048.

Watras CJ, Bloom NS.  1992. Mercury and methylmercury in individual zooplankton:
implications for bioaccumulation. Limnol Oceanogr 37(6):1313-1318.
                                         54

-------