External Letter Peer Review of EPA's Draft Report,
Inhibition of the Sodium-Iodide Symporter by Perchlorate:
An Evaluation of Lifestage Sensitivity Using Physiologically-based
Pharmacokinetic (PBPK) Modeling
Contract EP-C-07-024
Task Order 54
Submitted to:
U.S. Environmental Protection Agency
Office of Research and Development
National Center for Environmental Assessment
Research Triangle Park, NC 27711
Submitted by:
Eastern Research Group, Inc.
110 Hartwell Avenue
Lexington, MA 02421-3136
November 12, 2008
Printed on Recycled Paper
-------
QUALITY NARRATIVE STATEMENT
ERG selected reviewers according to selection criteria provided by EPA. EPA confirmed that the
scientific credentials of the reviewers proposed by ERG fulfilled EPA's selection criteria. Reviewers
conducted the review according to a charge prepared by EPA and instructions prepared by ERG. ERG
checked the reviewers' written comments to ensure that each reviewer had provided a substantial response
to each charge question (or that the reviewer had indicated that any question[s] not responded to was
outside the reviewer's area of expertise). ERG organized reviewer comments by charge question,
however, since this is an independent external review, ERG did not edit the reviewers' comments in any
way, but rather transmitted them unaltered to EPA.
-------
Contents
Responses to General Charge Questions
(Gl) Is EPA's analysis logical, clear and appropriate in depth and length? Has EPA accurately,
clearly and objectively represented and synthesized the scientific evidence for the
changes made to or specification of the model code and input parameters? 3
(G2) Please identify any additional studies or other data sources that should be considered in
the assessment of the specific parameters addressed below, including urinary clearance
(of iodide and/or perchlorate) and ingestion rates (breast milk, formula and water),
especially in neonates 7
(G3) Besides the specific parameters identified below, please comment on any other parameter
or model choices described in the document that you think are incorrect or require further
explanation or where other available information could provide better estimates 11
(G4) Please discuss research that you think would be likely to increase confidence in these
models and their use in predicting RAIU inhibition by perchlorate in different life stages
(for an average individual within each life-stage) 14
(G5) Does EPA accurately characterize the strengths and limitations of the analysis? Please
comment on any particular strengths that may not be mentioned or adequately
characterized for the estimates of RAIU for different life stages. Please also comment on
any specific sources of uncertainty that you believe have been overlooked in EPA's
analysis or that require further discussion, and which might be significant to EPA's
estimates of RAIU for different life stages 16
(G6) As recommended in EPA's 2000 Risk Characterization Handbook, is the analysis
transparent in terms of the steps, logic, key assumptions, limitations, and decisions?
Specifically, does characterization of the results of EPA's work fully explain: a) the
analysis approach employed; b) the use of assumptions and their impact on the analysis;
c) the use of extrapolations and their impact on the analysis; d) plausible alternatives and
the choices made among those alternatives; e) the impacts of one choice vs. another on
the analysis; f) significant data gaps and their implications for the analysis; g) the
scientific conclusions identified separately from default assumptions and policy
decisions, if any; and h) the major conclusions, and the discussions of EPA's confidence
and uncertainties in the conclusions? 18
Responses to Parameter-Specific Charge Questions
(A) Urinary Clearance 23
(Al) Please comment on the appropriateness of the input values selected for maternal urinary
clearance during pregnancy and lactation in EPA's analysis. Are the available data and
rationale for the values selected transparently and objectively described? Are you aware
of other publications or data that could be used to guide these choices or which provide
alternative input values that are equally valid or more appropriate? Likewise, are the
values selected for urinary clearance for the infant and older child the best estimates,
given the available science and data? Are there other data that would provide better (or
equally valid alternative) guidance or estimates? Is there any reason to believe that
urinary clearance might not be a limiting factor in the elimination of perchlorate for
infants? 23
(B) Breast-milk ingestion 29
(B1) Infants are generally known to consume very little milk or formula on the first day of life,
with ingestion quickly increasing over time. However the first time-point for which
i
-------
ingestion data are available for infants is at 7 days, which EPA extrapolated back to 0
ingestion attime=0 in its analysis (birth; Figure 1). The breast-milk ingestion rate was
estimated based on mean values reported in Arcus-Arth et al. (2005), while the water
ingestion used for the breast-feeding mother was the 90th percentile value from U.S.
EPA (2004) 29
(C) Water ingestion 32
(CI) For pregnancy, EPA used a normalized (90th percentile) water ingestion rate of 33
mL/kg-day (U.S.EPA 2004), which was multiplied by the maternal BW as described by
the PBPK model growth-functions during pregnancy to obtain total water ingestion for
the mother 32
(C2) For lactation, EPA used a fixed total (90th percentile) water ingestion rate of 2959 mL/day
(U.S. EPA 2004), with the rationale that while the woman's weight and self-water
demand are expected to drop after pregnancy (as described by the PBPK model time-
dependent weight equations), the demand for milk production would be increasing, and
the reported value was not for a specific age-range of child 36
(C3) For bottle-fed infants, EPA made extrapolations of the 90th percentile water-ingestion in
early life based on measurements made for the age ranges: 0-30 days, 1-3 months, and 6-
12 months (points in Figure 2, upper panel). For the purposes of this analysis, bottle-fed
implies feeding with formula requiring the addition of water 40
(D) Perchlorate concentrations in formula 43
(Dl) EPA used 1.42 (ig/L as the concentration of perchlorate in formula for bottle-fed infants.
This estimate was based on information from FDA's Total Diet Study, supported by
Pearce et al.'s (2007) findings 43
(E) Radioiodide excretion into breast-milk by NIS 46
(El) In the model, EPA included perchlorate inhibition of NIS radioiodide excretion into
breast-milk, as well as inhibition of radioiodide transport by perchlorate for all NIS-
containing tissues, thereby making the code consistent with the model description in
Clewell et al. (2007). Is this inclusion appropriate? Are the impacts of this inclusion
transparently and objectively described? 46
Additional Reviewer Comments 49
Appendix A. Individual Reviewer Comments A-l
Appendix B. Additional References Submitted by Reviewers B-l
ii
-------
Responses to General Charge Questions
i
-------
-------
(Gl) Is EPA's analysis logical, clear and appropriate in depth and length? Has EPA accurately, clearly
and objectively represented and synthesized the scientific evidence for the changes made to or
specification of the model code and input parameters?
Janusz The reviewed EPA draft document is logical, clear and concise. In general, reasons for
Byczkowski changes in code and input parameters and most of their their consequences are objectively
described. However, given a large number of abbreviations and acronyms, a glossary
listing these abbreviations and acronyms would be helpful.
Brian EPA's analysis is logical and clear. The length is appropriate and the appendix aid in
Cummings understanding their approach while adding to depth. Care is taken to explain what
changes were made to the model, why there were made, the scientific evidence and
literature used, the implications of these changes. In many cases EPA has simulated these
changes and found little, to no effect on the model.
Panos The document under review, "Inhibition of the Sodium-Iodide Symporter (NIS) by
Georgopolous Perchlorate: an Evaluation of Lifestage Sensitivity Using Physiologically-Based
Pharmacokinetic (PBPK) Modeling", has a very specific, and deliberately narrow,
objective. Indeed it defines (p. 21) sensitivity "as the predicted response in percent RAIU
(radioactive iodide uptake) inhibition 24 hours after iodide intravenous injection for an
average individual within a specific subgroup (e.g., bottle-fed infants) relative to the
predicted response in percent RAIU inhibition for an average, non-pregnant adult, where
response is the percent RAIU inhibition 24 hours after iodide IV injection." Though this
is rather constrained as a sensitivity metric, it can be reasonably argued that it addresses
adequately the biological issue of concern here.
The emphasis of the analysis is on sensitivity with respect to "lifestage". Parameters and
processes, related to different lifestages, were modeled based on assumptions that are
discussed in rather extensive detail in the document under review. The lifestages
evaluated in the document correspond to "average" adult, non-pregnant woman of child-
bearing age, pregnant woman, lactating woman, fetus, breast-fed infant, bottle-fed infant,
1 year old child, and and 2 year old child. The tools employed for the analysis were the
PBPK models of Clewell et al. (2007) for the pregnant woman/fetus and for the lactating
woman/breastfed infant. Results for the "bottle-fed" neonate were obtained by altering the
dose specification in the model for the breast-fed infant. The PBPK model for the average
adult was that of Merrill et al. (2005), while the model for the non-pregnant woman of
childbearing age was a direct modification of the model for the pregnant woman, obtained
by removing the placental and fetal compartments, but retaining the mammary
compartment.
The above PBPK models, with various corrections and adjustments (that are discussed in
detail in the appendices of the document) were used to estimate the predicted percent
RAIU inhibition for the average adult and different specific ("average") individuals
representing potentially sensitive subgroups. It should be mentioned here that the actual
text of the document under review states that the calculations were made for "subgroups,
including potentially sensitive subgroups"; however population-based modeling (with
considerations of inter-individual and intra-individual variability) was not actually
3
-------
pursued.
"Base" calculations were made assuming a dose equal to the point of departure (POD) of
7 (ig/kg-day, (consistent with the recommendations of the National Research Council -
NRC, 2005) and were summarized in Table 3 of the document under review. The relative
sensitivity of different subgroups was determined by comparing the percent RAIU
inhibition of each subgroup to the percent RAIU inhibition for an average adult at a dose
equal to the POD.
The document states that the "model predictions may generally be considered central
estimates for each subgroup (at the consumption levels modeled) that account for PK
(pharmacokinetic) differences, and do not take into account within-group variability in
pharmacokinetics, uncertainty in model parameters and predictions, or population
differences in PD." It should be noted that fetal simulations were reported for only the end
of gestation (Gestation Week 40).
The analysis presented in the document concluded that urinary clearance was a "key"
parameter (i.e., model predictions were highly sensitive to the values of this variable).
Though for modeling pregnancy and early infancy a conservative parameterization was
adopted, the document emphasizes that "a full population analysis of urinary clearance
was not conducted, and given that variability in other PK parameters was not addressed,
these estimates should not be considered a true upper confidence bound on RAIU
inhibition" (p. 23). The document also identified the fetus as the most sensitive subgroup
with respect to percent RAIU inhibition at a dose equal to the POD, in general agreement
with earlier PBPK modeling (Clewell et al., 2007) and estimating approximately 5-fold
higher percent RAIU inhibition for the fetus at gestational week 40 than for the average
adult. In fact it is also stated that "simulations at earlier gestation weeks indicate that the
fetus is more sensitive than the adult throughout pregnancy, but are considered too
quantitatively uncertain to assign exact relative sensitivities" (p. 23).
Overall it can be stated that EPA's analysis is clear and with sufficient discussion of
assumptions involving model and parameter specification (including adjustments and
corrections to the original models and their codes). It should be noted, however, that in
multiple instances (discussed further in the answers to the following questions) the
rationale behind specific assumptions and parameterizations relates more to
"convenience" rather than to scientific defensibility. Although this may not necessarily
affect the general conclusions, it is nevertheless a weakness of the analysis presented in
the document under review.
References Cited in Answer to Question G-l:
Clewell, R.A., Merrill, E.A., Gearhart, J.M., Robinson, P.J., Sterner, T.R., Mattie, D.R.,
and Clewell, H.J., 3rd. 2007. Perchlorate and radioiodide kinetics across life stages in the
human: using PBPK models to predict dosimetry and thyroid inhibition and sensitive
subpopulations based on developmental stage. J Toxicol Environ Health A 70 (5):408-28.
Merrill, E.A., Clewell, R.A., Robinson, P.J., Jarabek, A.M., Gearhart, J.M., Sterner, T.R.,
and Fisher, J.W. 2005. PBPK model for radioactive iodide and perchlorate kinetics and
perchlorate-induced inhibition of iodide uptake in humans. Toxicol Sci 83 (l):25-43.
NRC. 2005. Health Implications of Perchlorate Ingestion. National Research Council of
4
-------
the National Academies, National Academies Press. Washington, D.C.
http ://www.nap .edu/catalog/11202 .html
Sean Hays Yes
Frederick Kaskel Physiologically based pharmacokinetic (PBPK) models were modified to predict
inhibition of the sodium-iodide symporter (NIS)5 for pregnant and lactating women,
nursing infants, and for the subsequent stages of childhood. The published models were
modified by EPA to fix errors and incorporate new data, particularly data on lifestage
variability in the urinary clearance of pechlorate, to which NIS inhibition is sensitive. The
models are suitable to provide quantitative predictions to the Agency on the lifestyle
variability of perchlorate NIS inhibition of thyroidal iodide uptake. EPA's analysis is
logical, clear and appropriate in depth and length and EPA has accurately, clearly and
objectively represented and synthesized the scientific evidence for the changes made to or
specification of the model code and input parameters.
Kannan • The EPA analysis of perchlorate-mediated inhibition of the NIS in humans is
Krishnan based on Merrill et al. (2005) and Clewell et al. (2007) PBPK models, and
specifically addresses the variability of NIS inhibition as a function of lifestage.
The document is clear and concise. The depth and legnth of presentation are
appropriate, given the objective.
• The structure of the PBPK models published by Merrill/Clewell has not been
altered; rather some of the input parameters as well as equations have been
modified either to correct an error or to reflect current state of knowledge more
appropriately.
Chensheng Lu The Reviewer is convinced that EPA has performed an outstanding job in improving the
PBPK model so the codes written in the model are consistent to the physiology of iodide
uptake in the thyroid glands and the uptake inhibition by perchlorate. It is also clear that
EPA has tried to perfect the input parameters to increase the predictability of the model.
Lauren Zeise In general EPA's analysis is clear and logical, and succinct, although there are several
suggestions for improvement in the comments below. With regard to depth, the limited
treatment of variability and uncertainty is problematic. Pharmacokinetic models can
provide a structure for exploring and integrating variability, but this was not done in this
analysis. This major limitation is recognized by EPA (page 25). EPA points out the
model predictions apply to "a subgroup average for typical, healthy individuals, and
effectively describe the RAIU inhibition relative to that same individual as his/her own
control." EPA further points out that "These models were not designed to account for
whether the pregnant women are hypothyroid or iodine deficient." Analysis of such
large, susceptible populations is a critical aspect of understanding the potential health
impact of perchlorate drinking water exposure. A more rigorous and explicit treatment of
variability is needed to get a better handle on intra-human variability in response to
perchlorate exposure. The analysis would also be improved by more rigorous statistical
and quantitative treatment of uncertainty. The degree to which the analysis for the GW
5
-------
40 fetus may or may not represent the first and second trimester fetus needs explicit and
careful treatment.
On a smaller point, it would help if greater motivation was provided for some of the
statistical fits to data. Some statistical fits provided an expedient and practical way
forward in the analysis but appeared to introduce logical inconsistency. It would be
preferable for a more expanded discussion to provide a context for the approach taken.
6
-------
(G2) Please identify any additional studies or other data sources that should be considered in the
assessment of the specific parameters addressed below, including urinary clearance (of iodide
and/or perchlorate) and ingestion rates (breast milk, formula and water), especially in neonates.
Janusz It seems, that most of the published relevant data have been already reviewed and/or
Byczkowski included in the modeling and analysis.
However, this Reviewer could not identify the reference posted in the EPA Perchlorate
human lactation model code (pp. 74 and 76, lines 67 and 131, respectively): Gentry et al.
(2001). It is suggested, that for the VMk parameter ("Residual milk volume", see the
answer to G-3 below), another data source could be used: Dewey K.G., Heinig M.J.,
Nommsen L.A., and Lonnerdal B.: Maternal versus infant factors related to breast milk
intake and residual milk volume: the DARLING study. Pediatrics 1991: 87: 829-837.
Zuckier et. al., Journal of Nuclear Medicine, 45(3), 500-507, 2004. This study mainly
assesses perrhenate, but also studies the interaction of iodide and perchlorate in NIS
tissues, both in vivo and in vitro. It specifically studies biodistributions of these
compounds in the presence and absence of each other. It takes into account the effect of
NIS on this distribution. It may prove helpful.
A search of pubmed did not reveal any references to iodide or perchlorate clearance not
already mentioned by the authors. The most recent article I could find on either subject
was DeWoskin and Thompson, 2008, which the authors use.
Why was a compartment analysis figure not shown for this revised model? Such figures
are useful to readers in conceptualizing the model. These were included in the literature
on which the current model was based.
The scientific literature relevant to NIS inhibition by perchlorate is currently growing fast;
the same holds true for related literature areas covering fields such as demographics and
exposure informatics and modeling, Physiologically-Based Pharmacokinetic and
Pharmacodynamic modeling methods, etc. Though the document under review is not
expected to provide a thorough literature review of the subject of perchlorate inhibition of
NIS and of related exposure and risk issues, it could certainly provide a more complete
picture to its readers, by incorporating some of the references suggested below.
These suggestions are grouped in three categories: (a) "general references," that cover
various aspects of perchlorate exposure and effect, (b) references that focus on studies of
human exposure to perchlorate, and (c) references that focus on biological (physiological
and biochemical) issues, either directly specific to perchlorate and NIS inhibition or
indirectly related, such as e.g. references on information for urinary clearance related
parameters or on information for PBPK modeling specific to infants.
It should be noted in particular that USFDA (The US Food and Drug Administration) has
developed PBPK modeling recommendations, as well as computer software that
implements them, for early life stages (Luecke et al., 2007, 2008); at a minimum, it would
be useful to examine how these parameterizations compare to the ones adopted in the
analysis presented in the document under review. (Similarly, it would be useful to
Brian
Cummings
Panos
Georgopolous
7
-------
compare exposure-related parameter selections used in the reviewed work to
corresponding relevant recommendations in USEPA's Child-Specific Exposure Factors
Handbook).
General:
ATSDR. 2008. Toxicological Profile for Perclorates. Agency for Toxic Substances and
Disease Registry. Atlanta, GA. http://www.atsdr.cdc.gov/toxprofiles/tpl62.pdf
Charnley, G. 2008. Perchlorate: Overview of risks and regulation. Food and Chemical
Toxicology 46 (7):2307-2315.
De Groef, B., Decallonne, B.R., Van der Geyten, S., Darras, V.M., and Bouillon, R. 2006.
Perchlorate versus other environmental sodium/iodide symporter inhibitors: potential
thyroid-related health effects. Eur J Endocrinol 155 (1): 17-25.
Gu, B., and Coates, J.D. 2006. Perchlorate: Environmental Occurrence, Interactions and
Treatment. New York: Springer.
Kirk, A.B. 2006. Environmental perchlorate: why it matters. Anal Chim Acta 567 (1):4-
12.
Kirk, A.B., Dyke, J.V., Martin, C.F., and Dasgupta, P.K. 2007. Temporal patterns in
perchlorate, thiocyanate, and iodide excretion in human milk. Environ Health Perspect
115 (2): 182-6.
Kirk, A.B., Martinelango, P.K., Tian, K., Dutta, A., Smith, E.E., and Dasgupta, P.K.
2005. Perchlorate and iodide in dairy and breast milk. Environ Sci Technol 39 (7):2011-7.
Wang, R.Y., and Needham, L.L. 2007. Environmental chemicals: from the environment
to food, to breast milk, to the infant. J Toxicol Environ Health B Crit Rev 10 (8):597-609.
Exposure:
Baier-Anderson, C., Blount, B.C., Lakind, J.S., Naiman, D.Q., Wilbur, S.B., and Tan, S.
2006. Estimates of exposures to perchlorate from consumption of human milk, dairy milk,
and water, and comparison to current reference dose. J Toxicol Environ Health A 69 (3-
4):319-30.
Blount, B.C., Valentin-Blasini, L., Osterloh, J.D., Mauldin, J.P., and Pirkle, J.L. 2007.
Perchlorate exposure of the US Population, 2001-2002. J Expo Sci Environ Epidemiol 17
(4):400-7.
Ginsberg, G.L., Hattis, D.B., Zoeller, R.T., and Rice, D.C. 2007. Evaluation of the U.S.
EPA/OSWER preliminary remediation goal for perchlorate in groundwater: focus on
exposure to nursing infants. Environ Health Perspect 115 (3):361-9.
Zender, R., Bachand, A.M., and Reif, J.S. 2001. Exposure to tap water during pregnancy.
J Expo Anal Environ Epidemiol 11 (3): 224-30.
8
-------
Physiological/Biochemical:
Brandt, J.R., Wong, C.S., Hanrahan, J.D., Quails, C., McAfee, N., and Watkins, S.L.
2006. Estimating absolute glomerular filtration rate in children. Pediatr Nephrol 21
(12): 1865-72.
Clewell, R.A., and Gearhart, J.M. 2002. Pharmacokinetics of toxic chemicals in breast
milk: use of PBPK models to predict infant exposure. Environ Health Perspect 110
(6):A333-7.
Dohan, O., De la Vieja, A., Paroder, V., Riedel, C., Artani, M., Reed, M., Ginter, C.S.,
and Carrasco, N. 2003. The Sodium/Iodide Symporter (NIS): Characterization,
Regulation, and Medical Significance. Endocrine Reviews 24 (l):48-77.
Hawcutt, D.B., and Smyth, R.L. 2008. One size does not fit all: getting drug doses right
for children. Archives of Disease in Childhood 93 (3): 190-191.
Ito, S., and Alcorn, J. 2003. Xenobiotic transporter expression and function in the human
mammary gland. Adv Drug Deliv Rev 55 (5):653-65.
Johnson, T.N. 2008. The problems in scaling adult drug doses to children. Arch Dis Child
93 (3):207-l 1.
Kurz, H., Sandau, K., Dawson, T.H., Brown, J.H., Enquist, B.J., and West, G.B. 1998.
Allometric ccaling in biology. Science 281 (5378):751a-.
Lewandowski, T.A., Seeley, M.R., and Beck, B.D. 2004. Interspecies differences in
susceptibility to perturbation of thyroid homeostasis: a case study with perchlorate. Regul
Toxicol Pharmacol 39 (3):348-62.
Luecke, R.H., Pearce, B.A., Wosilait, W.D., Slikker, W., Jr., and Young, J.F. 2007.
Postnatal growth considerations for PBPK modeling. J Toxicol Environ Health A 70
(12): 1027-37.
Luecke, R.H., Pearce, B.A., Wosilait, W.D., Doerge, D.R., Slikker, W., Jr., and Young,
J.F. 2008. Windows based general PBPK/PD modeling software. Comput Biol Med 38
(9):962-78.
McManaman, J.L., and Neville, M.C. 2003. Mammary physiology and milk secretion.
Adv Drug Deliv Rev 55 (5):629-41.
Packard, G.C., and Birchard, G.F. 2008. Traditional allometric analysis fails to provide a
valid predictive model for mammalian metabolic rates. J Exp Biol 211 (Pt 22):3581-7.
Spitzweg, C., Dutton, C.M., Castro, M.R., Bergert, E.R., Goellner, J.R., Heufelder, A.E.,
and Morris, J.C. 2001. Expression of the sodium iodide symporter in human kidney.
Kidney Int 59 (3): 1013-23.
Strawson, J., Zhao, Q., and Dourson, M. 2004. Reference dose for perchlorate based on
thyroid hormone change in pregnant women as the critical effect. Regulatory Toxicology
9
-------
and Pharmacology 39 (l):44-65.
West, G.B., Brown, J.H., and Enquist, B.J. 1997. A general model for the origin of
allometric scaling laws in biology. Science 276 (5309): 122-126.
Sean Hays N/A
Frederick Kaskel Additional studies that should be considered in the assessment of the specific parameters
such as urinary clearance of iodide and/or perchlorate and ingestion rates (breast milk,
formula and water) in neonates and these include data on maturation of tubular transport
rates. On page 39 of the report, the issue of the role of pendrin transporter for iodide
during development is addressed. One cannot assume that perchlorate and iodide are
handled similarly by the developing kidney based on their similar charge and diameter;
more data is needed in the investigation of tubular maturation of the transporters that
regulate the clearance of iodide and perchlorate. This is also addressed on page 40,
second paragraph in the report where it is stated that one cannot assume that the relative
clearance for iodide and perchlorate should be constant across all ages and life stages.
Additionally on page 42 the EPA states that there is no data on renal transporters during
infancy to suggest the level and pattern of expression changes required to change
clearance/GFR. Thus, the report used DeWoskin and Thompson's published data for
scaling of renal excretion for infants by body weight and on page 44 the EPA extended its
extrapolation to a 60-day-old, 5 kg child is sound. These assumptions are reasonable but
indicate the importance of additional investigations in newborn models and in humans.
Kannan This reviewer is not aware of any studies in neonates that would provide better estimates
Krishnan of urinary clearance of perchlorate and iodide. Even though isolated studies reporting
ingestion rates (breast milk, formula and water) in infants in other parts of the world could
be obtained from the literature, such studies probably would only introduce further
uncertainty. However, the study of Kirk et al. (2005). Perchlorate and iodine in dairy and
breast milk. Environ Sci technol 39: 2011-17 may used to corroborate the findings of the
present study - as it relates to the relationships between drinking water concentration and
breast milk concentration.
Chensheng Lu Considering the elevated inhibition of RAIU in bottle-fed infants, EPA should seek for
additional data to re-affirm the water ingestion rates that are used by EPA, particularly the
use of the 90th percentile values in which the situations exceed the expectation of the
fundamental knowledge. The Reviewer has no knowledge of whether there are studies or
data sources that EPA could use.
Lauren Zeise Additional possible studies and data sources are identified in response to specific charge
questions below.
10
-------
(G3) Besides the specific parameters identified below, please comment on any other parameter or model
choices described in the document that you think are incorrect or require further explanation or
where other available information could provide better estimates.
The numerical value of the VMk parameter, identified in the EPA Perchlorate human
lactation model code (p. 76), line 131: "CONSTANT VMk = 0.6320 ! Residual milk
volume (L) (Gentry et al 2001)" seems to be unrealistically high. It is closer to the low
daily breast milk intake by infant rather than to the residual milk volume (< 650 g/day vs
109 g/day; Dewey et al., 1991).
Since the mammary glands respond to feeding stimuli by secreting breast milk on
demand, the "residual milk volume" usually refers only to the small volume of
unconsumed milk. Without suckling stimulation, even lower void volume of milk remains
in alveoli, lactiferous ducts and sinuses between the feeding sessions, and it stays in
equilibrium with blood under near steady-state conditions (Byczkowski, J.Z. in U.S. EPA
(2002): Final Report "EXPLORATION OF PERINATAL PHARMACOKINETIC
ISSUES", EPA/63O/R-01/004, May 10, 2001. On-line:
http://oaspub.epa.gov/eims/eimscomm.getfile?p_download_id=120867). While milk
intakes vary with caloric demand of the infant (represented by KTrans in the EPA model),
as reported by Dewey et al. (1991), infants with low intakes left as much milk
unconsumed as those with higher intakes, which justifies the residual milk volume to
remain constant (109 g/day).
On the other hand, the EPA PBPK modeling algorithm suggests that the VMk parameter
corresponds rather to the initial volume of milk (632 g), further linked to the growth
function of infant's body weight by KTrans, which infant receives in each "pulse". Even
though this approach may adequately describe the volume of breast milk actually ingested
by growing infant, it is not physiologically accurate and does not allow for any
interspecies extrapolation. Since VMk affects concentrations of both iodide and
perchlorate in breast milk, it is difficult to predict if and how the suggested change in
VMk description would change the RAIU inhibition in the breast-fed neonate. Before any
change, a sensitivity analysis should be performed with varied VMk value, to evaluate
how VMk parameter affects the PBPK model output and to decide if and how this
potential problem should be addressed.
Brian The methods used for scaling of clearance to body weight, age and surface area are
Cummings appropriate; however, such scaling is most accurate for clearance when the substance in
question is not reabsorbed or secreted. Given that fact that both pendrin and NIS are
reported to act on perchlorate, and given reports that NIS expression does not scale to
bodyweight in some tissues (see below), do the authors feel that their approach is still
valid? Do alterations in NIS expression need to be included in this model? If they are,
would this increase the risk for children age 10-14, when NIS expression is believed to
altered?
GFR in children is typically scaled according to muscle mass, which scales well with the
cube of height in boys and girls from 6 months to adult (see Check, DB et al., Am. J. Clin.
Nutr, 30:851, 1977). Scaling formulas have even been derived for children based on
creatinine levels (See Diseases in the Kidney, Chapter 80, Seventh Edition, Editor =
Janusz
Byczkowski
11
-------
Schrier, Page 2355). Could these formulas be used to more accurately reflect GFR in
children when calculating perchlorate and iodide clearance?
Panos As discussed in more detail in the answer to the questions regarding the characterization
Georgopolous °f urinary clearance processes, there is a need to develop and thoroughly test a consistent
framework for modeling these processes for different lifestages. "Correcting" the
inconsistencies, that are in fact identified in Appendix B of the document under review,
would be a first step towards the implementation of such a framework.
Sean Hays N/A
Frederick Kaskel There are no other parameters or model choices described in the document that are
incorrect or require further explanation or provide better estimates.
Kannan The parameters of this model consist of:
Krishnan
• Physiological parameters
• Intake/contact rates
• Partition coefficients
• Permeability-area cross product
• Urinary clearance
• Binding parameters
• Maximal velocity and affinity constants
The parameter values found in the original reports and refined following EPA's
evaluation would appear to be supported by available literature. However, focused data
collection might facilitate the improvement of the partition coefficient values used in the
model as well as the urinary clearance values for perchlorate and iodide in the various
lifestages.
Chensheng Lu EPA should explain the rationale of using the 90th percentile values in the analysis. It
seems to the Reviewer that such choice is deemed to create an upper bound limit,
however, throughout the document, EPA has stated that this is not the purpose due to the
uncertainties involved in the model simulation and other reasons.
12
-------
Lauren Zeise
_ Co-exposures, _
no biologic ?ri susceptibility
No co-exposures or
biolog ic a I susceptibility
5
1
2s
I
¦S
Value of phytisSufiea! jrarranter
Rp» 1. Distribution of a typical physiological parameter within tha population and how that tnav vary
depending on the influence of chemical and biologic background,
The above figure, taken from Woodruff et al. (2008; EHP 116:1568), illustrates the main
limitation in the analysis and approach to modeling. The question in evaluating the
potential risks from perchlorate in drinking water is about the extent to which the
incremental exposure from water results in adverse effects to the mom, her baby, her
developing fetus or others. In the above figure it is above the extent to which the
perchlorate drinking water exposure, in the presence of coexposure and biological
sensitivity, is creating adverse outcomes in the population. The fidelity of the analysis
depends on whether individuals with biological susceptibility have been adequately
addressed and also whether coexposures that affect iodide inhibition have been
adequately considered.
EPA analysis enables biological susceptibility and coexposures to be partially addressed
in the assessment, but it needs to move further to enable a fuller treatment. With regard to
biological susceptibility EPA considers susceptible subgroups - the infant, fetus, mom -
and an important factor that increases susceptibility in these groups - low renal clearance.
But the analysis does not enable the agency to consider the extent of impact on other
sensitive subgroups in these populations, such as those with clinical and subclinical
hypothyrodism, those that may be genetically predisposed (see e.g., Scinicariello, EHP
113(11): 1479-84), and those that are iodine deficient. The EPA analysis also considers an
important coexposure - perchlorate intake via food. However, the analysis does not
consider the combined impact with thiocyanate, which also affects iodide uptake at the
NIS. Thiocyanate is also found in breast milk (see e.g., Kirk et al. 2007, EHP, 115:182-
186), cigarette smoke, and common foods. The recent finding in women who smoked,
that those with low urinary iodine levels had decreasing T4 with increasing perchlorate
(Steinmaus et al. 2007, EHP, 115:1333-1338) as well as reduced content of iodine in
breast milk and the urine of breast feeding infants of smokers (Laurberg et al. 2004, J Clin
Endo Met 89:181-187) indicates the importance of considering coexposures to
thiocyanate. Nitrate, ubiquitous though far less potent than perchlorate, should also be
considered (see e.g., DeGroef et al., 2006, Eur J Endocrin 155: 17-25).
13
-------
(G4) Please discuss research that you think would be likely to increase confidence in these models and
their use in predicting RAIU inhibition by perchlorate in different life stages (for an average
individual within each life-stage).
Janusz
Byczkowski
If the overall model output is sensitive to VMk parameter, the confidence in the PBPK
modeling of"Breast-fed neonate" could be increased by a better description of "residual
milk volume" (see answers to G-2 and G-3, above).
Brian
Cummings
Studies directly assessing the effect of perchlorate on the clearance of NIS substrates are
needed. Further, data on the mechanisms of perchlorate inhibition of NIS is lacking.
Research investigating the toxicity of substrates of NIS, in the presence and absence of
perchlorate, is needed. Finally, urinary clearances of environmental pollutants in infants
and neonates are needed.
Panos The main challenge related to this question is "defining" an "average individual within
Georgopolous each lifestage." Equally challenging would be the identification of an "average exposure"
within each lifestage. EPA should consider the merits of a probabilistic
sensitivity/uncertainty analysis and eventually the feasibility of population-based PBPK
modeling (with explicitly defined sensitive subpopulations). Such an approach will
provide a more realistic assessment of the actual ranges of the outcomes considered in the
analysis, will eventually improve risk characterization efforts, and help explain both inter-
individual and intra-individual variability within a (sub) population.
Sean Hays The results of this modeling effort should be classified as theoretical since no validation
exercises have been performed and EPA should be clear to state this. There are existing
data which could help to validate the model predictions, especially related to the most
sensitive scenario (e.g., the nursing infant of the exposed mother). In particular, Pearce et
al. (2007) provides matched data on perchlorate and iodine in breast milk and urine
samples from nursing mothers. EPA should obtain this data from the study authors. This
will greatly help to test the model predictions. Furthermore, the authors found no
correlation (either positive or negative) between perchlorate and iodine in breast milk
samples. The authors of this study indicate this is consistent with other researchers. This
may raise questions about the results of EPA's modeling efforts. Since no results for the
concentrations of perchlorate and iodine in milk as a function of perchlorate dose are
provided in EPA's report, it is impossible to determine the validity of this issue.
Frederick Kaskel Newer estimates of renal function have been provided by Schwartz which should be
evaluated
Kannan • In the iodide/perchlorate models, the chemical concentration entering the tissues
Krishnan corresponds to the arterial PLASMA concentration whereas the flow rate to
tissues corresponds to BLOOD (RBC + Plasma) FLOW rates. Either the influx
in all mass balance equations should correspond to whole blood concentration or
the flow rate should correspond to plasma flows - since the RBC:plasma partition
14
-------
coefficient (PRBC_p) is not always equal to 1 (see for example lines 112 on page
38, or line 115 on page 4 of the EPA model code file), and the chemical
movement between plasma and RBC is diffusion-limited and no flow-limited.
The consequence of this modeling assumption may be verified to ensure
confidence in the use of these models. For example, if the simulations indicate
that the concentration profile of perchlorate is identical in RBC and plasma
compartments, qualitatively and quantitatively, then the above observation has no
consequence.
• Further, consideration should be given to the possibility of being able to simulate
iodine-deficient (or hypothyroid) situation in pregnant women by modulating
specific parameters of the model.
• Both the response to the previous question and the comments under general
overview are all applicable here.
Chensheng Lu What will significantly increase the confidence in these PBPK model is to use the real-
world data (such as perchlorate in drinking water and the level of iodide in blood or
thyroidal functions in population) liking perchlorate exposure and iodide inhibition. The
article published by Blount et al. (EHP 2006 114(12) 1865-1871) would be an ideal
application for these PBPK models. Unfortunately, data used in Blount et al. study
(NHANES) do not include children ages 6 and below.
Lauren Zeise Research to get a better handle on renal clearance of iodine and perchlorate during
pregnancy and postpartum; biomonitoring of perchlorate, iodide, thiocyanate and thyroid
hormone during and after pregnancy during lactation in smoking and non-smoking
women. Measurements of perchlorate in baby formula - in non-composited samples.
15
-------
(G5) Does EPA accurately characterize the strengths and limitations of the analysis? Please comment on
any particular strengths that may not be mentioned or adequately characterized for the estimates
of RAIU for different life stages. Please also comment on any specific sources of uncertainty that
you believe have been overlooked in EPA's analysis or that require further discussion, and which
might be significant to EPA's estimates of RAIU for different life stages
Janusz The reviewed EPA draft document adequately describes strengths and limitations of the
Byczkowski analysis. However, the overall strength of any PBPK modeling used for risk assessment is
its ability to simulate adequately the experimental data. While a partial validation of the
unmodified iodide PBPK model was performed and reported by Clewell et al., in the EPA
document, only limited comparisons of some variables to the actual data have been
presented (in Appendix C). The strength of the EPA modeling analysis could be better
presented by including comparisons of the PBPK model simulations to all significant real-
life experimental data.
Brian EPA does a good job of characterizing the strengths and limitations of the model, but
Cummings needs a separate paragraph at the end directly addressing these points.
Did EPA take into account any hormone effects on NIS or thyroid function? The
previous literature, on which this model is based, devotes some discussion to this subject.
This is particularly important when discussing susceptibility during puberty.
Panos In this reviewer's opinion the main strength of the analysis is the explicit listing of
Georgopolous unresolved issues (and inconsistencies) in the modeling described in the document under
review.
The main limitations involve:
• the emphasis on point estimates rather than on pursuing a distributional
(probabilistic) approach for characterizing exposures (with explicit variability and
uncertainty of activities for each individual and across a sub-population,
• the emphasis on "average" individuals for each lifestage rather than on pursuing
population-based modeling with explicit characterization of inter-individual and
intra-individual pharmacokinetic (physiological and biochemical) variabilities,
and
• the consideration of iodide and perchlorate exposure "in isolation" and not in a
context of "total" exposure that would consider other NIS inhibitors (thiocyanate,
nitrates).
Sean Hays In places, EPA adequately highlights the strengths and weaknesses of their analysis.
However, there are other areas where the limitations have not been adequately addressed
(e.g., lack of validation).
16
-------
Frederick Kaskel EPA accurately characterized the strengths and limitations of the analysis. However, as
indicated in the report, additional information on the possible effects of maturation of
glomerular filtration, tubular reabsorption and secretion, and changes in body composition
during the neonatal period is indicated in order to more confidently apply EPA's
estimates of RAIU for different life stages.
Kannan • The strength relates to the use of PBPK model to assess the lifestage sensitivity to
Krishnan inhibition of the sodium-iodide symporter by perchlorate; the use of fetus as a
subgroup to evaluate the relative sensitivity to adults; consideration of relevant
route/source of exposure (drinking water);
• The weaknesses are related to the fact that the analysis did not include certain
subgroups (e.g., elderly, foetus during early gestation periods, iodine-deficient or
hypothyroid status during pregnancy) and did not address variability of parameter
values within subgroups in the simulations (i.e., with the use of Bayesian or
Monte Carlo type methods).
Chensheng Lu This document is well written and has highlighted what EPA has accomplished in
assessing RAIU inhibition at different life stages resulting from perchlorate exposure. The
Reviewer thought EPA has thoroughly discussed the strengths and limitation of this
analysis, including the uncertainty analysis.
One uncertainty, however, has not been addressed by EPA is the use of direct IV dose of
radioiodide to the bottle-fed infants in order to determining iodide uptake inhibition
caused by perchlorate in formula. Although this approach seems intuitive, it may not
reflect the real-world scenario in which iodide intake is usually taking place by oral
ingestion. Pharmacokinetically speaking, the absorption of chemicals in humans could
vary significantly between oral ingestion and bolus iv injection. EPA needs to conduct an
uncertainty analysis to assure that such approach would not impact the outcomes
significantly.
Lauren Zeise EPA does not sufficiently elaborate on the limitation of focusing on "healthy" individuals,
and the lack of consideration of the large susceptible populations.
Some parts of the analysis are scenario based, using 90th percentile values, while other
parts use mean values. With over 4 million infants born in the US each year, scenario
analyses should be added. These would be directed at ascertaining the inhibition levels for
the some plausible higher susceptibility cases, such as infant and fetus exposures
associated with a mom with relatively high thiocyanate exposure (e.g., from broccoli
consumption or smoking), low renal clearance, who got all her fluids directly or indirectly
from tap water.
17
-------
(G6) As recommended in EPA's 2000 Risk Characterization Handbook, is the analysis transparent in
terms of the steps, logic, key assumptions, limitations, and decisions? Specifically, does
characterization of the results of EPA's work fully explain: a) the analysis approach employed; b)
the use of assumptions and their impact on the analysis; c) the use of extrapolations and their
impact on the analysis; d) plausible alternatives and the choices made among those alternatives; e)
the impacts of one choice vs. another on the analysis; f) significant data gaps and their implications
for the analysis; g) the scientific conclusions identified separately from default assumptions and
policy decisions, if any; and h) the major conclusions, and the discussions of EPA's confidence and
uncertainties in the conclusions?
It seems that the reviewed EPA draft document complies with guidelines and
recommendations of the EPA 200 Risk Characterization Handbook. As stated in the
answer to G-l (above), it is logical, clear and concise. The alternatives are presented and
the decision points are adequately explained.
Perhaps, the contribution of specific PBPK uncertainties to the overall uncertainty of
modeling RAIU inhibition by perchlorate could be summarily discussed and presented in
a separate section.
Brian The analysis conforms to EPA guidelines on transparency with regards to steps and logic.
Cummings The key assumptions of the model are clearly listed, as well as the decisions involved in
parameters changes and impact of these changes. The limitations of the model could be
more clearly listed (see above).
The analysis approach employed is adequately explained as are the basis for assumption
used in this model. EPA goes to great lengths to test the impact of these assumptions by
determining the sensitivity for each parameter changed.
The extrapolations used are clearly outlined and well as their rationale for them; however,
for the most part, only the theoretical impact of these extrapolations are discussed. Some
validation is presented, but this mostly uses previously validated data. Was the model
applied to any data sets in the literature not previously studied? Are such data available at
this time?
The impacts of choice for most of the critical parameters are discussed. This is particular
true when discussing choices for BW, clearance and scaling these values.
Plausible alternatives for some of the parameters could be more clearly listed. This is
particularly true of the impact on altering the level of water ingestion to 90%. What were
the alternatives to this value and how did they affect them model?
Alternative choices for scaling to BW and clearance are clearly listed as well as their
impacts.
The major and scientific conclusions for this work are clearly stated and appear to be
separated from any grand statements on policy decisions.
The authors do discuss data gaps throughout the manuscript, but a specific section is
needed, towards the end, listing these gaps in itemized, or table form. This should be
Janusz
Byczkowski
18
-------
followed by paragraph that list major perceived weaknesses and uncertainty of this model,
which need to be more, clearly stated.
The document and - in particular the appendices - are quite explicit ("transparent") in
listing and discussing all the assumptions and approximations involved in the analysis.
This takes place at a level of detail that exceeds what is typically expected in the peer
reviewed literature and the authors of the document under review should be commended
for this. However, the justifications of what can be called "emergency solutions" to
various problems discussed in the document, especially those related to inconsistencies in
modeling urinary clearance processes (including inconsistencies in scaling factors as well
as "ad hoc" adjustments to achieve agreement for predicted values) are often weak. It can,
probably reasonably, be argued that, for the ranges of concentrations and exposures
considered, the effect of correcting the above inconsistencies will not have a substantive
impact on calculated outcomes; however, it is still very important to develop a "fully
defensible" model that incorporates up-to-date scientific information and assumptions that
are consistent "across lifestages". Clearly, resolving the inconsistencies that have been
identified through the efforts presented in the document under review will be useful in
numerous other applications involving exposures to chemicals in utero and during
infancy.
Sean Hays Yes, the analysis is transparent.
Frederick Kaskel The analysis is transparent in terms of the steps, logic, key assumptions, limitations, and
decision. The characterization of the results of EPA's work fully explains: a) the analysis
approach employed; b) the use of assumptions and their impact on the analysis; c) the use
of extrapolations and their impact on the analysis; d) plausible alternatives and the
choices made among those alternatives; e) the impacts of one choice vs. another on the
analysis; f) significant data gaps and their implications for the analysis; g) the scientific
conclusions identified separately from default assumptions and policy decisions, and h)
the major conclusions and the discussions of EPA's confidence and uncertainties in the
conclusions.
Kannan The analysis is transparent and the assumptions as well as alternative approaches are
Krishnan generally described in sufficient detail. The conclusions are essentially scientific in
nature, based on data obtained from PBPK model simulations. The following
improvements are suggested:
1. The reason for limiting the present analysis to eight sub-groups (i.e., pregnant
woman,fetus, lactating woman, breast-fed infant, bottle-fed infant, 1 year old and 2 year
old child,"average" adult, and non-pregnant woman of child-bearing age) may be
specified at the outset. In this regard, it may be useful to clarify as to why the elderly and
teens were not part of the sub-groups analysed in this study.
2. Clarify as to why the results of this analysis are also applicable to chronic exposure
exposure situations (compared to typically acute (short-term) simulations)
Panos
Georgopolous
19
-------
3. The justification of the choice of 24-hr RAIU as the endpoint should be included. Was
24-hr AUC considered as an alternative measure ? What was the scientific basis for
basing the analysis on a single RAIU value in infants and adults obtained at one specific
time point (i.e,. 24 hr). Some consideration/discussion of the sensitivity of that time point
to the key input parameters as a function of age might be useful.
Chensheng Lu EPA has clearly explained their approaches employed including the assumptions,
alternatives, and the use of extrapolations and their impacts on the analyses. Apparently,
there are significant data gaps, particularly for newborn infants that lead to some
limitations of using this revised PBPK model. However, it is rather common for many
PBPK modeling work, and therefore should NOT be considered a major limitation of this
analysis.
It is apparent that the outcome of the PBPK model prediction is dictated by the use of
urinary clearance of perchlorate and iodide. Other parameters have somewhat less
impacts on the results. EPA has taken the right approach focusing on the parameters
related to perchlorate exposure and iodide intakes. The revised PBPK model that EPA
modified has demonstrated the importance of those parameters, and the Reviewer agrees
with the EPA's scientific conclusion in which the modified Clewell et al. model is
acceptable to calculate the lifestage differences in the degree of NIS inhibition of
thyroidal radioiodide uptake at a given level of perchlorate exposure.
Lauren Zeise EPA does a reasonably good job laying out the logic, key assumptions, limitations and
decisions. But for the most part, it is done in a manner that will be understandable to
someone with a modeling background. It will be difficult to follow and very accessible to
a more general reader. More motivation of the forms for the statistical fits is needed, and
a more quantitative and rigorous treatment of uncertainty. EPA reasoning for using 90th
percentile values for some parameters and mean values for others is not explained well.
Failure to address certain large susceptible populations and the possible sizes of these
populations should be discussed. The degree to which the analysis for the GW 40 fetus
may or may not represent the first and second trimester fetus needs explicit and careful
treatment.
20
-------
Responses to Parameter-Specific Charge Questions
21
-------
-------
(A) Urinary Clearance
(Al) Please comment on the appropriateness of the input values selected for maternal urinary clearance
during pregnancy and lactation in EPA's analysis. Are the available data and rationale for the
values selected transparently and objectively described? Are you aware of other publications or
data that could be used to guide these choices or which provide alternative input values that are
equally valid or more appropriate? Likewise, are the values selected for urinary clearance for the
infant and older child the best estimates, given the available science and data? Are there other data
that would provide better (or equally valid alternative) guidance or estimates? Is there any reason
to believe that urinary clearance might not be a limiting factor in the elimination of perchlorate for
infants?
Janusz The urinary clearance has been addressed appropriately and discussed as well as
Byczkowski presented in a sufficient detail (in Appendix B). Given the limited data, particularly for
perchlorate, it seems tat the approach and values presented by EPA are reasonable and
this Reviewer is not aware of any data that could contradict this approach.
Brian The authors do a good job discussing the choice and values used maternal urinary
Cummings clearance during and after pregnancy and lactation. This includes are significant
discussion of the impact of changes made to these parameters and alternatives.
The available data and analysis, for the most part, are rationale, transparently and
objectively described, with one exception (see C-2) below. I am not aware of any further
publication that could be used for input values that are more appropriate. One question
that I did have is what was the rationale for choosing the lower clearance value from
Clewell et al., (2007) as opposed to the others (Page 12, 2nd paragraph).
Please see my comments above concerning other data that provide alternative guidance
with regards to urinary clearance in neonates, infant and children. Several studies support
the hypothesis that urinary clearance is a limiting factor of perchlorate elimination.
While it's possible that it may not be the only factor involved, it is clearly a major one. In
absence of any data to the contrary, which could not be found, the authors are correct in
this assumption.
The analysis presented in the document under review concluded that urinary clearance
was a "key" process/parameter; i.e. model predictions were very sensitive with respect to
the magnitude of urinary clearance. However, there are a number of issues concerning the
parameters employed in modeling urinary clearance that need substantial clarification.
Appendix B (pp 39-50) of the document under review provides an extensive discussion of
the assumptions and approximations involved in selecting and estimating these
parameters. Various inconsistencies in the selection/estimation procedures are in fact
recognized explicitly in Appendix B, but in general these inconsistencies are "accepted"
on the basis of either a minimal anticipated effect on the calculations of the model, or as a
means for avoiding a more complex analysis. For example, the last paragraphs of p. 40
states that because "... renal clearance is largely controlled by glomerular filtration and
non-specific fluid resorption, the expectation is that the relative clearance for iodide and
perchlorate [....] should be constant across ages, body weights, and lifestages. In EPA's
evaluation for the child and "average" (non-pregnant, non-lactating) adult, this
Panos
Georgopolous
23
-------
proportionality has been maintained." However, in the model of Clewell et al. (2007) "the
maternal urinary clearance value [....] was set at 60% of the value in the non-pregnant
human based on observed difference in the pregnant and male rat models These
maternal lactation values go against the argument given just above that the proportionality
should be maintained, but EPA chose to use the maternal values as so set. It is likely
worthwhile to evaluate these maternal values in light of the generally higher urinary
excretion seen in pregnant/lactating women, but alteration of these clearance constants
would require refitting of other parameters, and so EPA chose not to conduct that specific
evaluation." Clearly this is an issue that requires further consideration. It should also be
mentioned that this discussion is preceded by the following rather puzzling statement that
"The tables in the papers identify the units of [urinary clearance] as L/h/kg, but clearly
this should be L/h/kg°75 to be consistent with this mathematical formulation, which is
how the CLU values are calculated in the computer code." Such a selection
units/dimensions contradict the physics of the problems and in fact it appears that the
tables in the original articles (Clewell et al., 2007 and Merrill et al., 2005) state the correct
units. (The first paragraph on page 41 of the document under review also employs correct
units/dimensions.) The issue of consistent allometric scaling is an important one and there
exist various publications that can be helpful in clarifying issues such as the above (e.g.
Johnson, 2008; West et al., 1997; Kurz et al., 1998). Issues of inconsistent scaling in fact
appear across the entire description of urinary clearance parameters (Appendix B).
Various other inconsistencies are discussed and "accepted" in relation to the calculations
of urinary clearance in the neonate (pages 42-43) and in the pregnant/lactating woman.
For example, on page 47, it is stated:
"Keeping with the assumed proportionality between perchlorate and iodide, based
on these data the same relationship would be expected to hold: higher clearance
rather than reduced. A dilemma occurs in considering the data of Aboul-Khair et
al. (1964); however, in that the control iodine clearance as measured by them is
31.05 ± 3.66 ml./min (mean ± SE), while the value determined by Merrill et al.
(2005) for non-pregnant adults is 44.3 mL/min. Likewise Aboul-Khair et al.
(1964) report thyroid iodide uptake at 2.5 hr postinjection as 21.4 ± 1.4 % of the
administered dose, but the amount predicted by the Merrill et al. (Merrill et al.,
2005) model (in the absence of perchlorate) is 7.78%. Therefore, the data of
Aboul-Khair et al. (1964) was normalized to their own controls for both urinary
clearance and iodide uptake, and then use that relative change as a model input
(for clearance, multiplying the non-pregnant clearance rate constant by the
pregnant:control ratio from Aboul-Khair et al. (1964) or in estimating changes in
thyroid NIS (to fit relative increases in thyroid uptake)."
Clearly, the inconsistency in absolute values reported in the above paragraph should be
the focus of further study; while the normalization employed by EPA offers a way of
circumventing the issue, this "solution" could only be considered qualitative in nature.
In this reviewer's opinion, a consistent treatment of the urinary clearance process for
various life stages emerges clearly as a research need, based on the outcomes of the
sensitivity testing and the issues presented in Appendix B of the document under review.
References Cited in Answer to Question A-l:
Clewell, R.A., Merrill, E.A., Gearhart, J.M., Robinson, P.J., Sterner, T.R., Mattie, D.R.,
24
-------
and Clewell, H.J., 3rd. 2007. Perchlorate and radioiodide kinetics across life stages in the
human: using PBPK models to predict dosimetry and thyroid inhibition and sensitive
subpopulations based on developmental stage. J Toxicol Environ Health A 70 (5):408-28.
Johnson, T.N. 2008. The problems in scaling adult drug doses to children. Arch Dis Child
93 (3):207-l 1.
Kurz, H., Sandau, K., Dawson, T.H., Brown, J.H., Enquist, B.J., and West, G.B. 1998.
Allometric ccaling in biology. Science 281 (5378):751.
Merrill, E.A., Clewell, R.A., Robinson, P.J., Jarabek, A.M., Gearhart, J.M., Sterner, T.R.,
and Fisher, J.W. 2005. PBPK model for radioactive iodide and perchlorate kinetics and
perchlorate-induced inhibition of iodide uptake in humans. Toxicol Sci 83 (l):25-43.
West, G.B., Brown, J.H., and Enquist, B.J. 1997. A general model for the origin of
allometric scaling laws in biology. Science 276 (5309): 122-126.
Sean Hays The available data and rationale are transparently described. However, I disagree with the
rationale for the choice of maternal urinary clearance values. This is probably the most
sensitive parameter for the most sensitive scenario/receptor and the EPA has chosen to
use rodent data over human data. This is inadequate. EPA should choose to use the
available human data which indicates there is no measurable or consistent difference in
urinary clearance during pregnancy as compared to the non-pregnant state. The EPA
chose a reasonable urinary clearance for the lactating mother scenario. I agree with EPA's
choice for the urinary clearance among infants and older children.
Frederick Kaskel The input values selected for maternal urinary clearance during pregnancy and lactation
used in EPA's analysis are appropriate, transparent and objectively described. The choice
of three alternatives for pregnancy is a rational compromise in lieu of the lack of
additional human data. The use of a lower clearance is a safe assumption. On page 10
urinary clearance values for perchlorate and iodide across all lifestages were determined
to be sensitive parameters for prediction ofNIS thyroidal iodide uptake inhibition by
perchlorate. EPA determined that urinary clearance of perchlorate and iodide in neonates
is slower than is indicated by scaling based on body weight. Urinary elimination of a
number of compounds including drugs and drug metabolites also indicate that renal
clearance is slower per unit of body weight in neonates. Modification of the PBPK
models to describe slower clearance of perchlorate and iodide in neonates resulted in an
increase in predicted levels ofNIS inhibition in infants.
The values selected for urinary clearance for infants and older children are the best
estimates for the available data. The interpretation of the data that suggested an increase
in predicted levels ofNIS inhibition in infants at a perchlorate dose-rate of 7 ug/kg-day is
a safe assumption. The indices of renal function are based on the literature which
indicated that the GFR increases steadily postnatally but does not reach adult values until
approximately 2 years of age. I know of no other data that would provide better guidance
or estimates and it is unlikely that there are other factors than the urinary clearance in the
elimination of perchlorate for infants. However, one should consider that tubular function
in this age group is not fully matured and possible developmental changes in transport
activity might be important but no data is available for perchlorate elimination during
25
-------
development. On page 11 EPA chose to estimate perchlorate induced inhibition using
scaling of urinary cleareance proportional to body weight for children at 1 year of age and
older which results in somewhat higher estimates of iodide uptake inhibition than reported
by Clewell although still slightly less than predicted for the average adult exposed at the
same dose. EPA's estimates of urinary clearance in infants and children are lower than
those used in Clewell but reflects published GFR values.
The inadequacy of use of urinary clearance values in various human lifestages based on
(i) the pregnant:nonpregnant values in rats, and
(ii) the scaling of renal function for neonates on the basis of BW°75
- are well justified by EPA. The outcome is consistent with available experimental and/or
physiological data. The selection of lower clearance value for pregnancy as well as the
option 2 for lactating women, though not the optimal (given the interindivudal
variability), would appear to be pragmatic and consistent with the rationale provided by
EPA. However this reviewer has the following additional observations:
• The R2 value for the fit described in Figure B-6 is poor raising concern about the
adequacy of the equation
• Did EPA analyze the data in Figure B-5 on the basis of body surface data for the
various age groups (of pregnant women)?
• On page 42, para 3, Figure B-l should read Figure B-2?
• What does GFR-based scaling mean in Figure B2? Is it body surface scaled?
Chensheng Lu Considering that perchlorate, as well as iodide, does not further metabolize in human
body, the urinary clearance should be a limiting factor in removing perchlorate and iodide
from humans at all lifestages, and an important parameter in the perchlorate PBPK model.
Unfortunately, data for urinary clearance of perchlorate and iodide by the mother during
pregnancy and lactation are not consistent among three sources (Clewell, Aboul-Khair,
and Delange) cited by EPA. The choices that EPA made for selection clearance for
pregnancy and lactation are quite arbitrary, and the reasoning, if any, are not found. If
GFR is corresponding to the cardiac output (meaning higher blood flow rate equal to
higher GFR), urinary clearance of any given compound during the pregnancy should be
higher than non-pregnancy. Urinary clearance during lactation period might be the
opposite to the pregnancy due to the difference of cardiac output. EPA should seek for
differences of urinary clearance (mainly via GFR) of compounds during pregnancy and
lactation outside the iodide and perchlorate literatures.
EPA has clearly documented how they determined the alternative scaling of urinary
clearance of perchlorate and iodide by body weight and has provided a thorough
explanation of why EPA chose to use (BW)1, instead of commonly used (BW)0 75, in
neonates. The justification is sound and supported by the data published in the literature.
Similar justification of using (BW)1 scaling for perchlorate clearance in older children
Kannan
Krishnan
26
-------
(ages 2-12) is also provided, however, the sentence of "EPA's estimates of urinary
clearance in infants and children are lower than those used in Clewell et al. (2007), but are
values EPA judges to be scientific estimate, not bounds." (on page 11, 1st paragraph) is
not clear to the Reviewer. The information to support this sentence may come from
Appendix B (pages 44-46), particularly from Figure B-4. However, Figure B-4 itself is
difficult to understand (for instance, how is Lower 95% related to the yellow diamonds,
and how the line of Data average is constructed?), and therefore renders less convincing
remark of using (BW)1 scaling for older children. EPA may want to review this and
provide a clearer explanation on the data presented in Figure B-4.
Lauren Zeise The discussion of maternal urinary clearance values could be somewhat improved. In
regard to Figure B-6 motivation is not given for the fitting of the quadratic function to the
data for iodide clearance vs gestation week, and it is unclear where the postpartum data
set - greater than week 39 data set - on the plot appeared from and why it is included in
the modeling of clearance during pregnancy. The highest mean value was measured by
Adoul-Khair at the latest pregnancy time point. Inclusion of the extra data set weighs the
function down late in pregnancy when the highest value was measured by Aboul-Khair.
Further, including a gestation week of 45 on the plot axis is confusing to the reader.
There is a large extrapolation to clearance during the early pregnancy time point and renal
clearance can be increased fairly early in gestation. The quadratic fit may underpredict
clearance during this period. However, given that EPA is declining to estimate early fetal
effects, this portion of the extrapolation is not critical. It is unclear why the fit is being
presented however. Finally variability among individuals is an important consideration
and it would therefore be of interest to see on the plot or otherwise reported an indication
of variability in the individuals studied. Some indication of this is given in Table 2 of
Aboul-Khair et al., where renal clearance values for iodide have been serially averaged
for each pregnant individual studied. In addition, individual measurements for controls
are given. For Figure B-7 it would be good to show error bars or confidence bounds.
Minor error, on page 42, data from Guignard et al. are plotted in Figure B-2 not B-l.
Are you aware of other publications or data that could be used to guide these choices or
which provide alternative input values that are equally valid or more appropriate?
Increased urinary clearance of iodide during pregnancy is a widely recognized
phenomenon and data on the magnitude besides that reported by Aboul-Khair would be
useful and important to locate. Further, the inconsistency of PK outcomes and the Aboul-
Khair measured values in controls for IV iodide dose uptake 2.5 hours post injection is
quite troublesome and calls into question the PK modeling. The approach on page 49
described to deal with the inconsistency is not entirely satisfactory. EPA should look
hard for additional data sets to cross check assumptions regarding iodide uptake and renal
clearance during pregnancy and early postpartum.
Likewise, are the values selected for urinary clearance for the infant and older child the
best estimates, given the available science and data? Are there other data that would
provide better (or equally valid alternative) guidance or estimates? Is there any reason
to believe that urinary clearance might not be a limiting factor in the elimination of
perchlorate for infants?
I am not aware of better values for infants and the older child. The EPA laid out a
27
-------
reasonable analysis and approach for developing estimates for the infant and older child.
There is interindividual variability in clearance and it would be preferable if this were
more emphasized and acknowledged in the discussion, and attempts to better describe it
quantitatively, for example in terms of varying glomerular filtration rates normalized by
body size.
28
-------
(B) Breast-milk ingestion
(Bl) Infants are generally known to consume very little milk or formula on the first day of life, with
ingestion quickly increasing over time. However the first time-point for which ingestion data are
available for infants is at 7 days, which EPA extrapolated back to 0 ingestion at time=0 in its
analysis (birth; Figure 1). The breast-milk ingestion rate was estimated based on mean values
reported in Arcus-Arth et al. (2005), while the water ingestion used for the breast-feeding mother
was the 90th percentile value from U.S. EPA (2004).
Is EPA's extrapolation and rationale transparently and objectively described? Does this function
appropriately characterize the available data and information? Are there other data that could be
used to obtain a better or equally valid alternative estimate of mean breast-milk ingestion rate for
infants in the first few days of life? Should an estimate other than the mean be used to determine
the breast-milk ingestion rate?
Janusz The Arcus-Arth et al. (2005) data for breast milk intake are, perhaps, the best currently
Byczkowski available, and thus the smooth function that fits these data is the right approach. It seems
that the exponentially growing intake for the first week postpartum may be an
overestimate, but given the inadequate information about intake during the first weeks
postpartum, the applied growth function appears to be a reasonable, practical solution,
even though seemingly quite a health-protective. The idea of adding on the top of this
overestimation the upper bound (e.g., the estimate of 90th percentile intake) would lead to
a marked exaggeration of milk consumption, unrealistically high for a typical infant.
While this Reviewer cannot suggest any better approach to the milk intake, please see the
answers to G-2 and G-3 for a closely related potential problem with the estimated residual
breast milk volume.
Brian EPA's rationale for ingestion rates for infants at time 0 follows rationale assumptions and
Cummings the approach appears objective. While its true that infants consume little in the first day
of life, the extrapolation of 7 day data to day 1 allows for a margin of safety. I am not
aware of any data that EPA has not presented. While other estimates for breast-mile
ingestion may exist, the ones used by the EPA are clear, rationale and tractable. Thus,
they will be easily validated in future models.
Panos EPA's selection of point values for the testing analysis appears appropriate and
Georgopolous adequately justified as a reasonable conservative assumption. However, in this reviewer's
opinion, the uncertainties and variability inherent in the problem at hand would be better
addressed by a distributional (probabilistic) rather than point calculation. The large
population above the 90th percentile and the potential "spread" of exposure factors above
that percentile, would further justify such an analysis.
Sean Hays The approach and methods are objectively described. However, this is one of the weakest
portions of EPA's analysis. The modeling of perchlorate and iodine kinetics in the
neonate is highly uncertain. EPA needs to recognize this and make this clear to the reader.
If the purpose of this analysis is to determine the relative difference in inhibition of iodide
uptake by the thyroid for the various scenarios (e.g., normal adult, child, nursing infant,
fetus, and bottle-fed infant), then the mean on all exposure and pharmacokinetic
29
-------
parameters should be used and used consistently throughout the analysis. Otherwise, the
current approach exhibits bias for one scenario over the other.
Frederick Kaskel EPA's extrapolation and rationale are transparent and objectively described to assess the
breast milk ingestion rate.
Kannan • The fitting of the available data from Kahn and Stralka, with a mathematical
Krishnan function is adequate. However, the extrapolation from day 7 towards day 0 (or at
birth) is not warrented given that the newborn is not a sub-group used in the
assessment of relative sensitivity of lifestages (Table 3, page 22).
• The motivation for choosing 90th pctle for consumption rates needs to be clearly
presented, since the expectation is a calculation either based on mean values in
the various groups or 95th pctle values. Therefore, the rationale and scientific
basis for the choice and use of the 90th pctle values in these calculations should be
more clearly presented.
Chensheng Lu Based on the data presented in Figure 1, the milk ingestion rates, or suckling rate, are
quite different between Clewell et al. and Arcus-Arth et al., however, EPA's decision to
use Arcus-Arth's data requires further clarification. EPA claimed that Clewell et al.'s
data is inadequate to describe the suckling rates in the first couple weeks of life, however,
based on the Reviewer's examination on Figure 1, the abrupt increase of milk ingestion
during Day 1, and between Day 1 and 7 as presented by Arcus-Arth et al. seems unlikely.
The difficulty of collecting breast-milk ingestion rate for infants in the first few days of
life is understandable, and the deviation of the mean breast-milk ingestion from the true
value might not be as large as we thought. Therefore, the mean breast-milk ingestion rate
might be robust enough for use.
Lauren Zeise EPA's approach is objectively and transparently described, and the Agency was correct
that the Clewell et al. description is inconsistent with the currently available peer
reviewed literature. It is unclear why a mean value is used for infants and an upper 90th
percentile is used for the breast feeding mother. This is not adequately explained.
Does this function appropriately characterize the available data and information?
The function does not characterize the available data and information. It will be quite
confusing to anyone but a modeler.
The equation on page 15 has milk describes milk ingestion rate as
Milk ingestion rate (mL/hr) = KTRANS = 28.3*(BW-3.375)0.175
It then plots milk ingestion as a function of bodyweight and shows values for days 1,3,5
and 7 of life as on the bw vs milk ingestion plot. This formulation was used as a
convenient way of giving values to KTRANS but is problematic because it works only for
the specific circumstances using the mean values for breast milk intake in Arcus-Arth et
30
-------
al. data and will be confusing for anyone but a modeler.
In using bodyweight as a surrogate for age (3.375 kg as the zero age bodyweight) it
builds in an illogical structure that will be hard for the general public to understand and
limits the usefulness of the model for using data beyond the mean values in Arth-Arcus et
al. For example, there is zero milk ingestion for a bodyweight of 3.375 and milk ingestion
rates below that value cannot be defined. Furthermore, the expression has milk ingestion
increasing with increasing bodyweight indefinitely. This also is contrary to what occurs -
as infants age solid foods and other liquids are introduced and breast feeding reduces.
Arth-Arcus et al. show that for the available data sets milk consumption - in terms of
volume per bodyweight per day - decreases with age in a linear fashion. Thus there is
another inconsistency introduced by the way the model is formulated. At different ages
the mean milk ingestion at a given bodyweight will differ.
A more logical approach would be to develop an expression for milk ingestion in terms of
volume per bodyweight per day could be expressed as a function of age. A separate
expression could then be used to convert this to KTRANS. To deal with the early low
consumption rate on days 1-3 the measured values could be used.
Figure 1 notes that the data are from Arcus-Arth et al. but in fact it is entirely inconsistent
with Arcus-Arth et al. for the above discussed reasons.
Are there other data that could be used to obtain a better or equally valid alternative
estimate of mean breast-milk ingestion rate for infants in the first few days of life?
It is not correct that the first time point for which ingestion data are available is 7 days.
There are values in the literature for intake on days 1, 2, 3, 4 and 5. Indeed at days 4 and 5
the intake is quite high and consistent with the linear relationship for volume consumed
per kg per day vs age reported in Arth-Arcus et al. See table 8 in that paper.
Should an estimate other than the mean be used to determine the breast-milk ingestion
rate?
Whether or not the mean is used depends on later steps in the process, and ways that
variability will be taken into account. There are over 4 million births in the US annually.
The overall procedure for characterizing intra-species variability and central tendency
needs to be designed to be able to address the large number of infants "in the tails" of the
distribution. It would be preferable to build a PK approach that would enable fuller
description of variability in iodide uptake inhibition. The use of mean values and the
formulation used to compute KTRANS precludes this.
31
-------
(C) Water ingestion
(Cl) For pregnancy, EPA used a normalized (90th percentile) water ingestion rate of 33 mL/kg-day
(U.S.EPA 2004), which was multiplied by the maternal BW as described by the PBPK model
growth-functions during pregnancy to obtain total water ingestion for the mother.
Is EPA's approach and rationale transparently and objectively described? Is this approach
appropriate for characterizing the upper-bound for ingestion of pregnant women? Are there other
approaches or data that could be used to obtain a better or equally valid alternative estimate for
this parameter? (Note that the self-reported body weight values in U.S. EPA (2004) indicate the
same average body weight for pregnant women as for non-pregnant; data appears inaccurate, so
the PBPK model body-weight description was used instead.)
Janusz
Byczkowski
It seems that linking the upper bound estimate of water ingestion to maternal body weight
growth function is an appropriate approach, evidently, more realistic than the self-
reported data. The water ingestion issue has been mentioned in the reviewed EPA draft
document but not extensively discussed. It is not clear what (if any) modification to the
water ingestion variable has been made by EPA in comparison to the original pregnancy
model reported by Clewell el al. (the algorithm for "rdose p" is already marked in the
original CSL file as "modified by PMS").
Brian
Cummings
EPA's approach is rationale, transparent and objectively described. I agree with using
90% water ingestion for pregnant woman for a upper bound rate, as it adds a safety factor;
however clarification is needed as to how this value was derived. A review of the
documents used by EPA to determine this value reports 90% bootstrapping levels as
apposed to overall ingestions levels? Was the 90% bootstrap value used, or did EPA
calculate 90% ingestion from these data. How exactly was the value of 33 mL/kg-day
determined?
Panos
Georgopolous
The approach taken by EPA appears reasonable. As in the answer to the previous
question, this reviewer's opinion is that a distributed zonal analysis (Monte Carlo) can
provide more substantial insight on patterns of potential exposure rather than the point
calculations presented here.
Sean Hays
Throughout this analysis, EPA was inconsistent in choosing upper bounds or means for
various parameters. As such, there is no clear understanding of the objectives of EPA's
analysis. If the purpose of this analysis is to determine the relative difference in inhibition
of iodide uptake by the thyroid for the various scenarios (e.g., normal adult, child, nursing
infant, fetus, and bottle-fed infant), then the mean on all exposure and pharmacokinetic
parameters should be used and used consistently throughout the analysis. Otherwise, the
current approach exhibits bias for one scenario over the other.
Frederick Kaskel
EPA's approach and rationale for pregnancy water ingestion rate is transparent and
objectively described, and is based on the available literature. I know of no additional data
that could be used to obtain a better estimate of mean breast-milk ingestion rate for
32
-------
infants in the first few days of life. The mean estimate is fine.
• The cited value of 33 ml/kg-day corresponds to the 90th percentile value for
pregnant women ("consumers only") of the direct and indirect community water
ingestion (chapter 6, page 16, U.S. EPA 2004). However, the 90th percentile value
of total water ingestion for the same group was 39 ml/kg-day. It is unclear then
as to why EPA specifes the use of total water ingestion rate but actually uses the
value corresponding to another group (i.e,. community water ingestion).
Furthermore, the 90th percentile value for pregnant women, reported in US EPA
(2004), was associated with a small sample size (n=65, which does not meet the
minimum reporting requirements described in the "Third report on Nutrition
Monitoring in United States"). This raises the question of why not use (or justify
the non-use of) the value from Ershow et al (1991) based on much larger sample
size (n=188). These authors reported 90th pctle values for tap and total water
ingestion of 34.5 and 48.9 ml/kg-day respectively.
• It is also unclear to this reviewer as to why 90th pctle value is chosen for the
computations and not either the median or the 95th percentile value.
• This reviewer is not concerned about the use of subject-specific or group-specific
body weight in the PBPK model to facilitate the calculations for pregnant women,
as long as the ingestion rate is expressed in units of ml/kg-day, as done here.
Chensheng Lu EPA's has objectively described its approach in using water ingestion rate of 33 mL/kg-
day. However, the rationale of using the 90th percentile value was not provided by the
EPA in this analysis. As for the BW estimates, it is unclear of how accurate it is to use
the PBPK model growth-functions during pregnancy for estimating BW of pregnant
women. Will NHANES data provide some sort of national average of the water ingestion
rates stratified by lifestages and the BW of pregnant women? Or could EPA validate the
PBPK model growth-functions for weight estimates using the NHANES data?
Lauren Zeise The approach is transparently and objectively described, but the rationale is somewhat
unclear. Some parameters are based on mean values, others on midpoints and still others
on upper 90% bounds. It would be of interest to understand parameter distributions and
how this translates to distributions for iodide uptake inhibition. This may be beyond what
EPA has resources and time to do, but failing that, it would be desirable to have a clear
presentation of the approach. EPA appears to be taking a plausible scenarios approach.
But a clearer explanation is needed.
The table below is taken from EPA (2004). It shows the 95th percentile upper bound for
community water as 43 mL/kg/day, a reasonably higher level than the 90th percentile. In
the perchlorate document, the reason for choosing the 90th percentile and not some other
value needs to be justified. It is also worth noting that the number of pregnant women
captured in the survey is quite small, and raises some concern that the upper bound values
may be under estimates. For example, the upper bound estimate on the 90th percentile for
pregnant women was 46 mL/kg-day.
Kannan
Krishnan
33
-------
Table 63.B2. Per Capita Water CMramptioo—Pregnant Womea (la^kg/day)
Estimate
Total "Water
lP"i NFCS
'Hrdiovr et a3.
1001
Taj) Water
lf"8 NTC"
i Ei div.- et i>i.
!!>¦¦>: i
Total Wafer
Pi.
CljHII
Vv'aicr
fS CSm
Sample Siac
1KS-
!S«»
®*
05#
Mean
32.1
IS 3
21*
14*
SO" %
30.5
lft.4
10*
pt
mf m>
48.0
345
30*
33*
.as* m>
53.5
:-p S
44*
43*
* Women aged 15 so 40 years: # Wcasea aged 15 to 4- years.
Is this approach appropriate for characterizing the upper-bound for ingestion ofpregnant
women?
With 4.3 million births in the US each year, above the 90th percentile will be 430,000
women. Thus a very large number of women may consume water above this level, and
one is left wondering about the importance of the assumption and how sensitive the
results are to it. Following EPA (2004), the upper 95 percentile is 44 mL/kg/day, still
representing a rather large number of women - 215,000.
Are there other approaches or data that could be used to obtain a better or equally valid
alternative estimate for this parameter? (Note that the self-reported body weight values
in U.S. EPA (2004) indicate the same average body weight for pregnant women as for
non-pregnant; data appears inaccurate, so the PBPK model body-weight description was
used instead.)
There are other approaches that could be used to obtain a better or equally valid
alternative estimate. One consideration is the extent to which we may be confident that a
pregnant woman may use drinking water in cooking and for her fluid intake without
having to be concerned about harming her fetus. For this analysis one might consider the
basic water requirments for women living in hot climates. For this one might select a
value somewhat above the value of 3.0 L/day considered an "adequate intake" by the
Institute of Medicine (2004; Dietary Reference Intakes for Water, Potassium, Sodium
Chloride and Sulfate, IOM Food and Nutrition Board).
Another would be to pick a plausible upper bound value from the cumulative distribution
observed. For example, from the figure below, taken from EPA (2004), it can be seen
that a reasonable plausible upper bound may fall between 3.5 and 4 liters per day.
34
-------
Figure fi.2.Cla. Cumulative Distrihutinns of Par Capita Direct and Indirect Water Ingestion
Pregnant Consumers Only
ml/person/day
T
O
\V.*1 or I-1H. —H-- >11 : i V I n 11 ir«>r-^TMir-.f>i yr1„t v)
35
-------
(C2) For lactation, EPA used a fixed total (90th percentile) water ingestion rate of 2959 mL/day (U.S.
EPA 2004), with the rationale that while the woman's weight and self-water demand are expected
to drop after pregnancy (as described by the PBPK model time-dependent weight equations), the
demand for milk production would be increasing, and the reported value was not for a specific age-
range of child.
Is EPA's approach and rationale transparently and objectively described? Is this an appropriate
value to use for the ingestion rate of lactating women? Are there other better or equally valid
alternative approaches or values that could be used?
Janusz The use of 90th percentile of water ingestion by lactating woman apparently covers the
Byczkowski extra demand due to breast-feeding, and perhaps, it is health protective. The approach is
adequately described in the reviewed EPA draft document. This Reviewer is not aware of
any better approach.
Brian EPA approach is rational as it assumes that lactating women will still have increased
Cummings water needs. While the approach is transparently and objectively described I am have
some questions about the actual level, which results in -45.5 ml/kg-day, which is
substantially higher than the 90% ingestion rate reported above (assuming that the value
was not a bootstrap). Doesn't it seem more likely that water ingestion would equalize?
While the overall demand would decrease after pregnancy, this decrease would be
countered by lactation? Are the water demands for lactation higher than pregnancy?
What data exist on this subject other than models?
Panos The approach taken by EPA appears reasonable, though a probabilistic (Monte Carlo)
Georgopolous analysis would provide additional insight regarding the range of potential exposures.
Furthermore, since the simulations for lactating women produce estimates of perchlorate
concentration in breast milk, a population/distribution-level analysis with appropriate
parameterizations could be used to provide valuable testing of the model in relation to the
available data presented in Pearce et al. (2007) as well as Kirk el al. (2005, 2007)
References Cited in Answer to Question B-2:
Kirk, A.B., Martinelango, P.K., Tian, K., Dutta, A., Smith, E.E., and Dasgupta, P.K.
2005. Perchlorate and iodide in dairy and breast milk. Environ Sci Technol 39 (7):2011-7.
Kirk, A.B., Dyke, J.V., Martin, C.F., and Dasgupta, P.K. 2007. Temporal patterns in
perchlorate, thiocyanate, and iodide excretion in human milk. Environ Health Perspect
115 (2): 182-6.
Pearce, E.N., Leung, A.M., Blount, B.C., Bazrafshan, H.R., He, X., Pino, S., Valentin-
Blasini, L., and Braverman, L.E. 2007. Breast milk iodine and perchlorate concentrations
in lactating Boston-area women. J Clin EndocrinolMetab 92 (5): 1673-7.
Sean Hays Throughout this analysis, EPA was inconsistent in choosing upper bounds or means for
various parameters. As such, there is no clear understanding of the objectives of EPA's
analysis. If the purpose of this analysis is to determine the relative difference in inhibition
36
-------
of iodide uptake by the thyroid for the various scenarios (e.g., normal adult, child, nursing
infant, fetus, and bottle-fed infant), then the mean on all exposure and pharmacokinetic
parameters should be used and used consistently throughout the analysis. Otherwise, the
current approach exhibits bias for one scenario over the other.
Frederick Kaskel EPA's approach and rationale for lactation water ingestion rate is appropriate and
transparent and objectively described. I know of no other approaches.
Kannan • The ingestion rate of 2959 ml/day, used by EPA, corresponds to the 90th
Krishnan percentile value of "consumers only" lactating women for direct and indirect
community water ingestion. In comparison, the 90th percentile value of total
water ingestion in consumers only lactating women is reported to be 3021 ml/ day
(chapter 6 page 17). The EPA report (page 24, para 2) states that the intent was
to use the "total" consumers-only water intake in the calculations. The source and
consequence of this discrepancy should be addressed.
• Further, U.S. EPA (2004) indicated that the 90th pctl value (2959 ml/day) is
associated with a small sample size (n=41, which does not meet the minimum
reporting requirements described in the "Third report on Nutrition Monitoring in
United States"), raising a concern of its use rather than the value from Ershow et
al. (1991). Additionally, it is unclear as to why the 90th pctl rather than 95th pctl
of the water ingestion is used in these calculations.
• In light of the fact that the water ingestion rate in lactating women is significantly
greater (see chapter 6 pages 16-17, U.S. EPA 2004), on a ml/kg-dav basis, than
in pregnant women, the rationale used for using a fixed ingestion rate needs to be
more fully articulated.
Chensheng Lu The rationale of using a fixed total water ingestion rate is justifiable and transparently and
objectively described. However, the reasoning of selection of 2,959 mL/day at the 90th
percentile is missing in this analysis. It would be assuring if EPA could provide the
complete distribution of the estimates of total water ingestion.
Lauren Zeise The approach is transparently and objectively described, but the rationale is somewhat
unclear. As noted in response to C-l, some parameters are based on mean values, others
on midpoints and still others on upper 90% bounds. A clearer explanation is needed on
why the 90th percentile is chosen here, and not some other higher bound given the number
of women-infant pairs affected.
The tables below are taken from EPA (2004). They show the 90th percentile upper bound
for community water is not substantially smaller than the 95th percentile when expressed
as mL/kg/day, but appears more different when expressed as mL/person/day (2959 vs
3588), suggesting the difference may be driven by bodyweight differences at the 90th and
95th percentile. Still the reason for choosing the 90th percentile requires further
explanation.
37
-------
Mb <*.'• CI PerCapita Water ¦rocsoni^ttML—L«:1 atiuj it ui-y ui ilit Csi/Si?- Z '/.tz
Iff
Il:Iio7- ;l.
i-f f Ii
'.fi trr:c
.1(01.
ff ::n]
isf-i-sc. ir^rz
raciplt D'hz
=
jA=
kfcaa
i :c
imm
^
2.1M
l ; :
IJMt
14*0*
vtPm
1
-..-.v.
i Cll
us®- m
T
:
; ~r i
;r-ii
• WtaKnapsd IS w ¦geH
Is this an appropriate value to use for the ingestion rate oflactating women?
Similar to the response given to charge question C-l, use of the 90th percentile raises
concerns that a substantial number of mother infant pairs are not sufficiently considered.
The majority of newborn infants breast feed, and substantial numbers of infants do so
through age 6 months, and there are still large numbers above the 90th percentile.
With regard to arguments on water needs of lactating women, there is a paucity of data.
One could add the argument that IOM (2004) made that the intake of non-pregnant
women added to the fluid output in breastfeeding provides a reality check on water
ingestion rate.
Are there other better or equally valid alternative approaches or values that could be
used?
38
-------
There are other approaches that could be used to obtain a better or equally valid
alternative estimate. One consideration is the extent to which we may be confident that a
lactating woman may use drinking water in cooking and for her fluid intake without
having to be concerned about harming her baby. The basic water requirements for
women living in hot climates might be considered. For this one might select a value
somewhat above the value of 3.8 L/day considered an "adequate intake" for lactating
women by the Institute of Medicine (2004).
Looking at the cumulative distribution observed for lactating women, in the figure below
taken from EPA (2004), it can be seen that a reasonable plausible upper bound may fall
somewhere around 4 liters per day.
Figure 6.2,Clb- Uir'HilrfhVi' Jtivtii".ni1iiuiN nl Dili f anil Indirect Water Ingestion
1 i"rnu I j s.nmrs * >nl
i I pin• ¦ u^iuv
s.SCrtt
39
-------
(C3) For bottle-fed infants, EPA made extrapolations of the 90th percentile water-ingestion in early life
based on measurements made for the age ranges: 0-30 days, 1-3 months, and 6-12 months (points in
Figure 2, upper panel). For the purposes of this analysis, bottle-fed implies feeding with formula
requiring the addition of water.
Is EPA's extrapolation and rationale transparently and objectively described? Does the overall
function used to represent the changes in ingestion with age (and body weight) appropriately
characterize the available data and information? Are there other data that could be used to obtain
a better or equally valid alternative estimate of water ingestion rate for infants in the first few days
of life (e.g., 7-day old)? The water ingestion rates used by EPA are based on 90th percentile
ingestion data and thus are likely to exceed (minimal) physiological needs of infants as defined in
nutritional guidelines. Are the water ingestion rates used by EPA reasonable in comparison to the
physiological needs of infants at these various life stages? Are there other approaches or data (e.g.,
see the FDA memo) that could be used to obtain a better or equally valid alternative estimate for
this parameter?
Janusz For bottle-fed infant, analogously to maternal water ingestion, linking water intake to
Byczkowski body weight growth function is an appropriate approach (see answer to C-l, above). This
issue has been adequately described in the reviewed EPA draft document. The water
ingestion rates used by EPA appear to be realistic, and thus, reasonable. This Reviewer is
not aware of any better approach.
Brian EPA approach is rationale and transparently described. It is standard practice to scale
Cummings water ingestion to body weight, which does increase with age. Weight gain in the new
born scales more rapidly than almost any other time period and it sequestered into specific
groups; however, why was the calculation not scaled to 2 and 3 years as it done by WHO?
Another point of interest is in regards to the 1st seven days of birth. Most infants either
maintain birth weight or slightly lose 10% of their birth weight. This, as pointed out by
EPA is directly related to water ingestion. Should this than represent another group or be
removed (i.e. 0-7 days, or 7-30 days?).
Panos
Georgopolous
EPA's extrapolation and rationale are adequately described. However, it is doubtful that
the use of any single-point estimate would provide adequate understanding of the
potential range of exposures, and corresponding doses, for bottle-fed infants.
Sean Hays
Same response as last question.
Frederick Kaskel
EPA's extrapolation and rationale for bottle fed infant's water ingestion in early life is
transparent and objectively described. The overall function used to represent the changes
in ingestion with age (and body weight) appropriately characterize the available data. The
water ingestion rate for infants used by the EPA are reasonable in comparison to the
physiological needs of infants. The estimated 90% water intake rate used by EPA in
PBPK model stimulations is appropriate.
40
-------
Kannan
Krishnan
Even though the EPA rationale is satisfactory, it is unclear as to why the
emphasis is placed on the section of the curve (i.e., first few days after birth)
which is neither used in the lifestage analysis nor supported by any data.
• A report published by a Public Health Agency in Quebec contains data on water
consumption of 393 infants of 8 weeks of age. For bottle-fed only infants (n =
278), mean (IC95%) value for total water ingestion was 122±43 (117-127)
ml/kg-day, or 655 ± 233 (627-682) ml/day. The corresponding 90th percentile
values were 179 ml/kg-day and 981 ml/day. For more details the following
source may be consulted:
• http://www.inspq.qc.ca/publications/default.asp?NumPublication=334
• A copy of the above report in PDF is also attached herewith.
Chensheng Lu EPA has not informed the rationale of using the 90th percentile total water ingestion rate
in early life stage, as well as during the lactaton (as stated earlier in the review), and
therefore, the possibility of the estimated numbers are likely exceeding minimal
physiological needs of infants raises a concern. If this is the case in which the 90th
percentile total water ingestion rate exceeds the norms, this approach of using the 90th
percentile is problematic. Since this sub-analysis focuses on bottle-fed infants, EPA could
follow the nutritional guidelines to estimate the total water-ingestion rate (such as the
frequency of feeding per 24 hours and the quantify of formula and water mixing per
feeding).
Lauren Zeise Yes, although the reason for using a quadratic relationship was not described. The
approach of expressing water ingestion in units mL/kg/day and modeling it as a function
of age is much preferred over the approach used for breast milk consumption (e.g., in
Figure 1).
Does the overall function used to represent the changes in ingestion with age (and body
weight) appropriately characterize the available data and information?
This approach is a reasonable way of describing the upper 90th bound given in the Kahn
and Stralka (2008) paper.
Are there other data that could be used to obtain a better or equally valid alternative
estimate of water ingestion rate for infants in the first few days of life (e.g., 7-day old)?
An alternative for estimating ingestion rate for the first few days of life would be to rely
on data sets for breast milk consumption during the first 7 days (e.g., Casey et al. 1986,
Am J Dis Child 140:933; Neubauer et al. 1993, Am J Clin Nutr, 58:54), since breast fed
infants do not require supplemental water and the results may be more indicative than the
assumed relationship used, although sample sizes are relatively small. It is noteworthy
that intake in mL/kg/d during this period is not a smooth function of bodyweight. It is
quite low during the first two days of life but by age four or five days the intake is
essentially the same as at age 7 days. It is possible that the function 1 -e"day does a
reasonably good job of describing this. EPA could compare the values predicted by this
41
-------
function at days 1-7 to those seen in the literature for breast milk consumption on those
days.
The water ingestion rates used by EPA are based on 90th percentile ingestion data and
thus are likely to exceed (minimal) physiological needs of infants as defined in nutritional
guidelines. Are the water ingestion rates used by EPA reasonable in comparison to the
physiological needs of infants at these various life stages? Are there other approaches or
data (e.g., see the FDA memo) that could be used to obtain a better or equally valid
alternative estimate for this parameter?
The available data used by EPA indicate that infant formula consumption varies by
individuals, and correspondingly water consumption does as well. It is reasonable to
consider the minimal physiological needs of infants, as defined in nutritional guidelines,
although a precise understanding energy needs and use in infancy still appears to be a
matter of discussion (Reilly et al. 2005, Br J Nutr 94: 56-63). At any particular age
bodyweights, growth rate and degree of activity varies, and so consumption can not be
precisely calculated based on formula energy content and recipes for making up bottle fed
formula. Further, some infants are overfed and others are underfed. Thus although it
would be useful to compare water consumption with what one would expect given
nutritional guidelines and typical formula recipes, the nutritional guideline would not lead
to a reliable upper bound value for water consumption. Assumptions would be needed to
go from the water consumption based on the nutritional guideline level to an upper bound
estimate. Though as the FDA memo notes, "there is a relationship between the volume
of water an infant needs, and his/her caloric requirements for healthy growth" the exact
relationship to assume and the interindividual variability in that relationship has not been
provided and it is unclear that it would provide a more reliable estimate of water
consumption than is given in EPA's perchlorate report.
42
-------
(D) Perchlorate concentrations in formula
(Dl) EPA used 1.42 jig/L as the concentration of perchlorate in formula for bottle-fed infants. This
estimate was based on information from FDA's Total Diet Study, supported by Pearce et a/.'s
(2007) findings.
Is EPA's approach and rationale transparently and objectively described? Is 1.42 jig/L an
appropriate value to use for the concentration of perchlorate in infant formula? Are there other
better or equally valid alternative approaches or values that could be used?
Janusz
Byczkowski
The estimate of perchlorate concentration in formula for bottle-fed infants seems to be
appropriate and well substantiated, as it was based on the two independent sets of data
(FDA TDS and Pearce et al., 2007), which by themselves do not differ significantly from
each other. The derivation of average value is clearly described in the reviewed EPA draft
document.
Brian
Cummings
EPA's approach and rationale are transparent and well described. The approach is
logical and based on the most current information. A search for other levels for this
values results in similar results. This is obvious an area for which more data is needed
(see above).
Panos
Georgopolous
The rationale for the selection of 1.42 |_ig/L as a preset value for the concentration of
perchlorate for bottle-fed infants is not adequately discussed. This value is the average of
12 samples (8 of them above detection limit) presented in Murray et al. (2008); it is also
close to the average value (1.45 ppb) of the 17 samples analyzed by Pearce et al. (2007).
It should be noted that the values of perchlorate concentrations in the samples of Pearce et
al. range from 0.2 to 4.1 ppb. It would be useful to examine the sensitivity of uptake for a
reasonable concentration range rather than only the average value.
References Cited in Answer to Question D-l:
Murray, C.W., Egan, S.K., Kim, H., Bern, N., and Bolger, P.M. 2008. US Food and Drug
Administration's Total Diet Study: Dietary intake of perchlorate and iodine. J Expos Sci
Environ Epidemiol 18 (6):571-580.
Pearce, E.N., Leung, A.M., Blount, B.C., Bazrafshan, H.R., He, X., Pino, S., Valentin-
Blasini, L., and Braverman, L.E. 2007. Breast milk iodine and perchlorate concentrations
in lactating Boston-area women. J Clin EndocrinolMetab 92 (5): 1673-7.
Sean Hays
Based on the stated objectives of this analysis, EPA should adjust the intake of
perchlorate from infant formula to result in a daily exposure consistent with the point of
departure. This will yield consistent results across all scenarios to assure a fair and
impartial comparison of relative differences in inhibition of thyroid iodine uptake can be
made.
43
-------
Frederick Kaskel EPA's approach and rationale for the concentration of perchlorate in formula for bottle
fed infants is appropriated and transparency
Kannan The EPA,s approach is clearly described and appears to be consistent with the current
Krishnan state of knowledge. However, it would be better to clearly identify the basis for the
choice of the mean value rather than median, 90th or 95th pctl value (presumably the
limited, available data did not permit such a determination).
Chensheng Lu The Reviewer believes that perchlorate level in formula used by EPA is the best available
data; especially this level is consistent to the public numbers from an independent
research. The Reviewer is not aware other better or equally valid alternative approaches
or values that could be used.
Lauren Zeise The approach is not entirely transparent and the description could be improved. A sample
calculation for Table 4 describing how perchlorate intake for bottle fed infants is
estimated would be helpful.
Is 1.42 fig/L an appropriate value to use for the concentration ofperchlorate in infant
formula? Are there other better or equally valid alternative approaches or values that
could be used?
According to the Pearce et al. methodology:
"Seventeen brands of infant formulae were also assessed for iodine and perchlorate
levels. A single sample of each different type of liquid formula available at a local
supermarket was purchased for testing. Nine brands were sold in concentrated form
and designed to be diluted by half before use. Iodine and perchlorate levels were
measured directly in these samples, and the results were divided by half to reflect the
concentration intended for infant use. The other eight brands were sold ready for
use."
Thus, for the nine formula that were designed to be diluted, Pearce et al. assumed that
there was no perchlorate in the diluting water. EPA reports the correct average of 1.45
(ig/L calculated from 17 Pearce et al. samples. But for the young bottle fed infant the
calculation should reflect the intake of perchlorate from neat formula plus the intake from
the water used to dilute it. It is reasonable to assume that the only perchlorate intake in
the seven and 60 day infant would be water and formula. Thus the undiluted values for
formula perchlorate should be used.
The undiluted average from Pearce is 1.97 j^ig/L, but that includes formula that is ready to
use undiluted as well as formula that requires dilution. For use in Table 4, the focus
should be on concentrations of formula that would require dilution. The young seven and
60 day infant population drinking ready to use formula with no other consumption is more
a concern of the FDA than the EPA; they would not be receiving perchlorate
contaminated tap water. In the Pearce et al. study, the perchlorate concentration in the 9
samples of formula that would be diluted was 1.96 j^ig/L. The two highest of the nine
values reported would require dilution correspond to 3 ng/L and 3.2 ng/L, double the
44
-------
value reported in Pearce et al. Table 1.
The problem with the FDA data is that they represent composite samples, prepared as
they would be expected to be consumed. Also, the detection limit used by FDA is 1 j^ig/L.
The composite would be averaged across different formula brands and certain types.
Thus they do not provide an indication of what higher end exposures might be. The
composite sample results, in units (.ig/L. are:
202 Infant formula, milk, hi-Fe: ND, 2.5, 2.0, 2.0
203 Infant formula, milk, lo-Fe: 1.2, ND, 3.6, 2.1
309 BF, infant formula, soy: ND, ND, 0.8 *, 0.8 *
* indicates above the limit of detection but below the limit of quantitation and ND
indicates not detected.
Each value represents a composite from three cities in a given region. Thus a
concentration in particular product may be three times as high as the value reported.
Because of consumer loyalty and habit it is far more likely that a consumer will use the
same product over an extended period of time. From the values tabulated, value of 1.42
(ig/L will be an underestimate of perchlorate concentration in contaminated infant
formula. Further, the concentration of perchlorate in water used by FDA to prepare the
formula in to-be-eaten form has not been reported, but is likely to be low or not present,
given the several NDs in the table. Because FDA uses composite samples, it would be
preferable to use the high end value from FDA (3.6 j^ig/L) or a value of say 3 j^ig/L from
Pearce et al. Clearly better and more extensive measurement of perchlorate in infant
formula is desirable.
45
-------
(E) Radioiodide excretion into breast-milk by NIS
(El) In the model, EPA included perchlorate inhibition of NIS radioiodide excretion into breast-milk, as
well as inhibition of radioiodide transport by perchlorate for all NIS-containing tissues, thereby
making the code consistent with the model description in Clewell et al. (2007). Is this inclusion
appropriate? Are the impacts of this inclusion transparently and objectively described?
Janusz
Byczkowski
Because the mechanism of iodide transfer from blood into the thyroid gland and other
tissues, including mammary, is mediated by the sodium-iodide symporter (NIS), it was
logical and appropriate for EPA to include the algorithm for competitive inhibition of NIS
by perchlorate in these extra-thyroid tissues too. The inclusion of such a mechanistic
approach has been briefly mentioned in main text of the EPA draft document (in Section
2), and then, extensively discussed in the Appendix A. The rationale and the consequent
improvement in PBPK model predictions have been described objectively and presented
clearly (in Appendix A).
Brian
Cummings
This inclusion is appropriate because it's unlikely that a situation exists where perchlorate
or iodide is absent from the diet. This represents a logical and important refinement in the
model. However, a section is needed, towards the end, which clearly discusses the
impacts of perchlorate inhibition. Further, the rationale for scaling NIS to body weight
and other tissues is not clear. Studies suggest that NIS expression changes over
development (see below). At least one study suggests that the expression of NIS is
higher per g of tissue in young children (<12 years) compared to adults. Further, another
study suggests that this difference accounts for higher levels of iodide uptake in children
than in adults. Thus, scaling NIS levels to body weight (i.e. age) may not be appropriate.
Faggino et al., Journal of Nuclear Medicine, 45(2), 232-237.
Panos
Georgopolous
The inclusion of perchlorate inhibition of NIS radioiodide excretion into breast-milk, as
well as of inhibition of radioiodide transport by perchlorate for all NIS-containing tissues,
is appropriate. The impacts of this inclusion are adequately ("transparently and
objectively") described in the document.
Sean Hays
Based on the data from Pearce et al. (2007), one would expect there to be no effect of
perchlorate on the excretion of iodine in breast milk. As such, this feature of the model
that EPA has included may not be accurate with perchlorate kinetics. EPA should
investigate the data of Pearce et al. more fully and explore other data sets to see what
evidence is available to include such a feature in the model.
Frederick Kaskel
The inclusion of perchlorate inhibition of NIS radioiodide excretion into breast-milk, as
well as inhibition of radioiodide transport by perchlorate for all NIS-containing tissues, is
appropriate, transparent and objectively described. The EPA added inhibition of
radioiodide transport by perchlorate for radioiodide excretion into breast milk by NIS
markedly increased the predicted percent inhibition of thyroidal radioiodide uptake in the
breast fed infant.
46
-------
Kannan
Krishnan
Yes. The EPA's approach is logical and internally-consistent. The impact of this
inclusion is described in sufficient detail.
Chensheng Lu Yes, this inclusion is not only appropriate but also needed, and the impacts of this
inclusion are transparently and objectively described in this analysis. This work reflects
EPA's efforts in reviewing the model established by Clewell et al. and seeking for
improvement of the PBPK model.
Lauren Zeise It is reasonable and appropriate to assume that perchlorate inhibits the transport of iodide
in NIS containing tissues and iodide excretion into breast-milk. The impact of its
inclusion is transparently and objectively described. There is a straightforward layout in
Appendix A of changes in model assumptions and their impacts. Further, the effect of
decreased iodide levels in breast milk from smoking - with potential inhibition caused by
thiocyanate - has also been observed (Laurberg et al. 2004, J Clin Endo Met 89:181-187),
consistent with the finding that this should be taken into account in the modeling.
47
-------
-------
Additional Reviewer Comments
Miscellanous Comments;
Brian
Cummings ] An extra period is present in the first bullet point on page 59.
2. The third paragraph on page 42 refers to Figure B-l. Should this be Figure B-2?
3. Please list the years for the references for Merrill et al. listed on page 39.
Sean Hays Additional comments:
The approach of using PBPK modeling is admirable and the EPA should be commended.
However, the EPA should also have considered easier and more straightforward
approaches. The one-compartment PK model developed by EPA (Lorber, 2008), paired
with measured perhclorate and iodine levels in breast milk and infant formula would have
provided simpler and equally valid approaches for answering the question of the relative
difference in steady-state perchlorate levels (this is ultimately the endpoint of interest) in
the various receptors/scenarios. While I agree with using PBPK modeling, I also think
EPA should think about simpler approaches that are equally or more valid, and sometimes
much simpler and more easily embraced by the regulatory and risk assessment
community.
Appendix A: I agree with the model modifications made by EPA.
Appendix B: Appendix B is well written and easier to follow than the corresponding text
in the main report relating to urinary clearance values (section 3.1).
Overview:
This EPA report summarizes work conducted to evaluate the PBPK models for
perchlorate and radioiodide for quantitating relative sensitivity of different subgroups
(lifestages). The two-stage model evaluation process involved verification of model
codes and examination of the parameterization approaches. Following the revision of the
PBPK models by EPA, they were checked by a contractor who also verified the output of
the model by reproducing various figures from original publications. Despite the
thoroughness of the work, this life-stage variability analysis (either due to lack of data or
due to uncertainty associated with available data) did not account for certain subgroups
(e.g., elderly, foetus during early gestation periods, iodine-deficient or hypothyroid status
during pregnancy) and did not account for variability of parameter values within
subgroups in the simulations (i.e., with the use of Bayesian or Monte Carlo type
methods).
ADDITIONAL COMMENTS
In section 4.3. of the report, it is indicated as follows:
"For the 6- to 12-month and 1- to 2-year-old children, the water intake rates of 0.971
L/kg-day and 0.674 L/kg-day, respectively, were set based on 90th percentile values for
Kannan
Krishnan
49
-------
direct and indirect water consumers-only intake (Kahn and Stralka, 2008). Additionally,
to calculate L/day for these age groups, the corresponding age group mean body weights
obtained from NHANES1999-2006 were used: 9.2 kg for 6- to 12-month and 11.4 kg for
1-to 2-year-old children. "
The above statements indicate that the water intake by a 6 - 12 months old child would be
about 9 L/day whereas the body weight itself is only about 9 kg. So the above numbers
should be verified with the original report of Kahn and Stralka (2008). This reviewer's
verification with the original report would indicate 90th percentile values of 120 ml/kg/d
and 64 ml/kg/d, respectively, for the 6-12 month and 1-2 yr old groups. This is in
contrast to EPA's numbers of 971 ml/kg/d and 674 ml/kg/d as indicated above.
50
-------
Appendix A
Individual Reviewer Comments
A-l
-------
-------
PEER REVIEW COMMENTS FROM
Janusz Byczkowski, Ph.D, DABT
Consultant in Pharmacology and Toxicology
Fairborn, OH 45324
937-878-55321
Email: ibvczkowski@,netscape.com
A-3
-------
-------
Januzs Byczkowski
Contract No. EP-C-07-024
Task Order No. 54
October 22, 2008
INHIBITION OF THE SODIUM-IODIDE SYMPORTER BY PERCHLORATE:
AN EVALUATION OF LIFESTAGE SENSITIVITY USING PHYSIOLOGICALLY-BASED
PHARMACOKINETIC (PBPK) MODELING
(DUE DATE: NO LATER THAN MONDAY, NOVEMBER 10, 2008)
Background and Purpose
The U.S. Environmental Protection Agency (EPA) has prepared a draft document analyzing the inhibition
of thyroidal radioiodide uptake (RAIU) by perchlorate (an anion in CASRNs 7790-98-9; 7791-03-9;
7778-74-7 and 7601-89-0) for multiple life-stages, including pregnancy, fetal development, lactation,
infancy, childhood, and adulthood, using physiologically based pharmacokinetic modeling (PBPK). The
EPA has asked to review the changes applied to modeling parameters and specific changes in the PBPK
code, that were intended to make the code consistent with the published description.
Answers to Charge Questions
General Charge Questions:
G-l. Is EPA 's analysis logical, clear and appropriate in depth and length? Has EPA accurately, clearly
and objectively represented and synthesized the scientific evidence for the changes made to or
specification of the model code and input parameters?
The reviewed EPA draft document is logical, clear and concise. In general, reasons for changes in
code and input parameters and most of their their consequences are objectively described. However, given
a large number of abbreviations and acronyms, a glossary listing these abbreviations and acronyms would
be helpful.
G-2. Please identify any additional studies or other data sources that should be considered in the
assessment of the specific parameters addressed below, including urinary clearance (of iodide and/or
perchlorate) and ingestion rates (breast milk, formula and water), especially in neonates.
It seems, that most of the published relevant data have been already reviewed and/or included in
the modeling and analysis.
However, this Reviewer could not identify the reference posted in the EPA Perchlorate human
lactation model code (pp. 74 and 76, lines 67 and 131, respectively): Gentry et al. (2001). It is suggested,
that for the VMk parameter ("Residual milk volume", see the answer to G-3 below), another data source
could be used: Dewey K.G., Heinig M.J., Nommsen L.A., and Lonnerdal B.: Maternal versus infant
factors related to breast milk intake and residual milk volume: the DARLING study. Pediatrics 1991: 87:
829-837.
A-5
-------
Januzs Byczkowski
G-3. Besides the specific parameters identified below, please comment on any other parameter or model
choices described in the document that you think are incorrect or require further explanation or where
other available information could provide better estimates.
The numerical value of the VMk parameter, identified in the EPA Perchlorate human lactation
model code (p. 76), line 131: "CONSTANT VMk = 0.6320 ! Residual milk volume (L) (Gentry et al 2001)"
seems to be unrealistically high. It is closer to the low daily breast milk intake by infant rather than to the
residual milk volume (< 650 g/day vs 109 g/day; Dewey et al., 1991).
Since the mammary glands respond to feeding stimuli by secreting breast milk on demand, the
"residual milk volume" usually refers only to the small volume of unconsumed milk. Without suckling
stimulation, even lower void volume of milk remains in alveoli, lactiferous ducts and sinuses between the
feeding sessions, and it stays in equilibrium with blood under near steady-state conditions (Byczkowski,
J.Z. in U.S. EPA (2002): Final Report "EXPLORATION OF PERINATAL PHARMACOKINETIC
ISSUES", EPA/630/R-01/004, May 10, 2001. On-line:
http://oaspub.epa.gov/eims/eimscomm.getfile?p_download_id=120867). While milk intakes vary with
caloric demand of the infant (represented by KTrans in the EPA model), as reported by Dewey et al.
(1991), infants with low intakes left as much milk unconsumed as those with higher intakes, which
justifies the residual milk volume to remain constant (109 g/day).
On the other hand, the EPA PBPK modeling algorithm suggests that the VMk parameter
corresponds rather to the initial volume of milk (632 g), further linked to the growth function of infant's
body weight by KTrans, which infant receives in each "pulse". Even though this approach may
adequately describe the volume of breast milk actually ingested by growing infant, it is not
physiologically accurate and does not allow for any interspecies extrapolation. Since VMk affects
concentrations of both iodide and perchlorate in breast milk, it is difficult to predict if and how the
suggested change in VMk description would change the RAIU inhibition in the breast-fed neonate. Before
any change, a sensitivity analysis should be performed with varied VMk value, to evaluate how VMk
parameter affects the PBPK model output and to decide if and how this potential problem should be
addressed.
G-4. Please discuss research that you think would be likely to increase confidence in these models and
their use in predicting RAIU inhibition by perchlorate in different life stages (for an average individual
within each life-stage).
If the overall model output is sensitive to VMk parameter, the confidence in the PBPK modeling
of "Breast-fed neonate" could be increased by a better description of "residual milk volume" (see
answers to G-2 and G-3, above).
B-6
-------
Januzs Byczkowski
G-5. Does EPA accurately characterize the strengths and limitations of the analysis? Please comment on
any particular strengths that may not be mentioned or adequately characterized for the estimates of RAIU
for different life stages. Please also comment on any specific sources of uncertainty that you believe have
been overlooked in EPA's analysis or that require further discussion, and which might be significant to
EPA 's estimates of RAIUfor different life stages.
The reviewed EPA draft document adequately describes strengths and limitations of the analysis.
However, the overall strength of any PBPK modeling used for risk assessment is its ability to simulate
adequately the experimental data. While a partial validation of the unmodified iodide PBPK model was
performed and reported by Clewell el ah, in the EPA document, only limited comparisons of some
variables to the actual data have been presented (in Appendix C). The strength of the EPA modeling
analysis could be better presented by including comparisons of the PBPK model simulations to all
significant real-life experimental data.
G-6. As recommended in EPA 's 2000 Risk Characterization Handbook, is the analysis transparent in
terms of the steps, logic, key assumptions, limitations, and decisions? Specifically, does characterization
of the results of EPA's work fully explain: a) the analysis approach employed; b) the use of assumptions
and their impact on the analysis; c) the use of extrapolations and their impact on the analysis; d)
plausible alternatives and the choices made among those alternatives; e) the impacts of one choice vs.
another on the analysis; f) significant data gaps and their implications for the analysis; g) the scientific
conclusions identified separately from default assumptions and policy decisions, if any; and h) the major
conclusions, and the discussions of EPA 's confidence and uncertainties in the conclusions?
It seems that the reviewed EPA draft document complies with guidelines and recommendations
of the EPA 200 Risk Characterization Handbook. As stated in the answer to G-1 (above), it is logical,
clear and concise. The alternatives are presented and the decision points are adequately explained.
Perhaps, the contribution of specific PBPK uncertainties to the overall uncertainty of modeling
RAIU inhibition by perchlorate could be summarily discussed and presented in a separate section.
Parameter-Specific Charge Questions:
(A) Urinary clearance
A-l. Please comment on the appropriateness of the input values selected for maternal urinary clearance
during pregnancy and lactation in EPA 's analysis. Are the available data and rationale for the values
selected transparently and objectively described? Are you aware of other publications or data that could
be used to guide these choices or which provide alternative input values that are equally valid or more
appropriate? Likewise, are the values selected for urinary clearance for the infant and older child the
best estimates, given the available science and data? Are there other data that would provide better (or
equally valid alternative) guidance or estimates? Is there any reason to believe that urinary clearance
might not be a limiting factor in the elimination ofperchlorate for infants?
A-7
-------
Januzs Byczkowski
The urinary clearance has been addressed appropriately and discussed as well as presented in a
sufficient detail (in Appendix B). Given the limited data, particularly for perchlorate, it seems tat the
approach and values presented by EPA are reasonable and this Reviewer is not aware of any data that
could contradict this approach.
(B) Breast-milk ingestion
B-l. Infants are generally known to consume very little milk or formula on the first day of life, with
ingestion quickly increasing over time. However the first time-point for which ingestion data are
available for infants is at 7 days, which EPA extrapolated back to 0 ingestion at time=0 in its analysis
(birth; Figure I). The breast-milk ingestion rate was estimated based on mean values reported in Arcus-
Arth et al. (2005), while the water ingestion used for the breast-feeding mother was the 90th percentile
value from U.S. EPA (2004).
Is EPA's extrapolation and rationale transparently and objectively described? Does this function
appropriately characterize the available data and information? Are there other data that could be used
to obtain a better or equally valid alternative estimate of mean breast-milk ingestion rate for infants in
the first few days of life? Should an estimate other than the mean be used to determine the breast-milk
ingestion rate?
The Arcus-Arth et al. (2005) data for breast milk intake are, perhaps, the best currently available,
and thus the smooth function that fits these data is the right approach. It seems that the exponentially
growing intake for the first week postpartum may be an overestimate, but given the inadequate
information about intake during the first weeks postpartum, the applied growth function appears to be a
reasonable, practical solution, even though seemingly quite a health-protective. The idea of adding on the
top of this overestimation the upper bound (e.g., the estimate of 90th percentile intake) would lead to a
marked exaggeration of milk consumption, unrealistically high for a typical infant.
While this Reviewer cannot suggest any better approach to the milk intake, please see the answers
to G-2 and G-3 for a closely related potential problem with the estimated residual breast milk volume.
(C) Water ingestion
C-l. For pregnancy, EPA used a normalized (90th percentile) water ingestion rate of 33 mL/kg-day
(U.S.EPA 2004), which was multiplied by the maternal BWas described by the PBPK model growth-
functions during pregnancy to obtain total water ingestion for the mother.
Is EPA 's approach and rationale transparently and objectively described? Is this approach appropriate
for characterizing the upper-bound for ingestion ofpregnant women? Are there other approaches or
data that could be used to obtain a better or equally valid alternative estimate for this parameter? (Note
that the self-reported body weight values in U.S. EPA (2004) indicate the same average body weight for
B-8
-------
Januzs Byczkowski
pregnant women as for non-pregnant; data appears inaccurate, so the PBPK model body-weight
description was used instead.)
It seems that linking the upper bound estimate of water ingestion to maternal body weight growth
function is an appropriate approach, evidently, more realistic than the self-reported data. The water
ingestion issue has been mentioned in the reviewed EPA draft document but not extensively discussed. It
is not clear what (if any) modification to the water ingestion variable has been made by EPA in
comparison to the original pregnancy model reported by Clewell el al. (the algorithm for "rdose p" is
already marked in the original CSL file as "modified by PMS").
C-2. For lactation, EPA used a fixed total (90th percentile) water ingestion rate of2959 mL/day (U.S.
EPA 2004), with the rationale that while the woman's weight and self-water demand are expected to drop
after pregnancy (as described by the PBPK model time-dependent weight equations), the demand for milk
production would be increasing, and the reported value was not for a specific age-range of child.
Is EPA 's approach and rationale transparently and objectively described? Is this an appropriate value to
use for the ingestion rate oflactating women? Are there other better or equally valid alternative
approaches or values that could be used?
The use of 90th percentile of water ingestion by lactating woman apparently covers the extra
demand due to breast-feeding, and perhaps, it is health protective. The approach is adequately described
in the reviewed EPA draft document. This Reviewer is not aware of any better approach.
C-3. For bottle-fed infants, EPA made extrapolations of the 90th percentile water-ingestion in early life
based on measurements made for the age ranges: 0-30 days, 1-3 months, and 6-12 months (points in
Figure 2, upper panel). For the purposes of this analysis, bottle-fed implies feeding with formula
requiring the addition of water.
Is EPA's extrapolation and rationale transparently and objectively described? Does the overall function
used to represent the changes in ingestion with age (and body weight) appropriately characterize the
available data and information? Are there other data that could be used to obtain a better or equally
valid alternative estimate of water ingestion rate for infants in the first few days of life (e.g., 7-day old)?
The water ingestion rates used by EPA are based on 90th percentile ingestion data and thus are likely to
exceed (minimal) physiological needs of infants as defined in nutritional guidelines. Are the water
ingestion rates used by EPA reasonable in comparison to the physiological needs of infants at these
various life stages? Are there other approaches or data (e.g., see the FDA memo) that could be used to
obtain a better or equally valid alternative estimate for this parameter?
For bottle-fed infant, analogously to maternal water ingestion, linking water intake to body
weight growth function is an appropriate approach (see answer to C-l, above). This issue has been
adequately described in the reviewed EPA draft document. The water ingestion rates used by EPA appear
A-9
-------
Januzs Byczkowski
to be realistic, and thus, reasonable. This Reviewer is not aware of any better approach.
(D) Perchlorate concentration in formula
D-l. EPA used 1.42 fig/L as the concentration ofperchlorate in formula for bottle-fed infants. This
estimate was based on information from FDA 's Total Diet Study, supported by Pearce etal. 's (2007)
findings.
Is EPA 's approach and rationale transparently and objectively described? Is 1.42 fig/L an appropriate
value to use for the concentration ofperchlorate in infant formula? Are there other better or equally
valid alternative approaches or values that could be used?
The estimate of perchlorate concentration in formula for bottle-fed infants seems to be
appropriate and well substantiated, as it was based on the two independent sets of data (FDA TDS and
Pearce et al., 2007), which by themselves do not differ significantly from each other. The derivation of
average value is clearly described in the reviewed EPA draft document.
(E) Radioiodide excretion into breast-milk by NIS
E-l. In the model, EPA included perchlorate inhibition of NIS radioiodide excretion into breast-milk, as
well as inhibition of radioiodide transport by perchlorate for all NIS-containing tissues, thereby making
the code consistent with the model description in Clewell et al. (2007). Is this inclusion appropriate? Are
the impacts of this inclusion transparently and objectively described?
Because the mechanism of iodide transfer from blood into the thyroid gland and other tissues,
including mammary, is mediated by the sodium-iodide symporter (NIS), it was logical and appropriate for
EPA to include the algorithm for competitive inhibition of NIS by perchlorate in these extra-thyroid
tissues too. The inclusion of such a mechanistic approach has been briefly mentioned in main text of the
EPA draft document (in Section 2), and then, extensively discussed in the Appendix A. The rationale and
the consequent improvement in PBPK model predictions have been described objectively and presented
clearly (in Appendix A).
B-10
-------
PEER REVIEW COMMENTS FROM
Brian Cummings, Ph.D
Assistant Professor
University of Georgia
College of Pharmacy
Athens, GA 30602
706-542-3792
Email: bsc@rx.uga.edu
A-ll
-------
-------
Brian Cummings
Technical Charge to Peer Reviewers
Contract No. EP-C-07-024
Task Order No. 54
October 22, 2008
INHIBITION OF THE SODIUM-IODIDE SYMPORTER BY PERCHLORATE:
AN EVALUATION OF LIFESTAGE SENSITIVITY USING
PHYSIOLOGICALLY-BASED PHARMACOKINETIC (PBPK) MODELING
WRITTEN COMMENTS ARE DUE NO LATER THAN MONDAY, NOVEMBER 10, 2008
CHARGE QUESTIONS: My responses to the charge questions are in italics
General Charge Questions:
G-l. Is EPA's analysis logical, clear and appropriate in depth and length? Has EPA accurately, clearly
and objectively represented and synthesized the scientific evidence for the changes made to or
specification of the model code and input parameters?
EPA 's analysis is logical and clear. The length is appropriate and the appendix aid in understanding
their approach while adding to depth. Care is taken to explain what changes were made to the model,
why there were made, the scientific evidence and literature used, the implications of these changes. In
many cases EPA has simulated these changes and found little, to no effect on the model.
G-2. Please identify any additional studies or other data sources that should be considered in the
assessment of the specific parameters addressed below, including urinary clearance (of iodide and/or
perchlorate) and ingestion rates (breast milk, formula and water), especially in neonates.
Zuckier et. al., Journal of Nuclear Medicine, 45(3), 500-507, 2004. This study mainly assesses
perrhenate, but also studies the interaction of iodide and perchlorate in NIS tissues, both in vivo and in
vitro. It specifically studies biodistributions of these compounds in the presence and absence of each
other. It takes into account the effect of NIS on this distribution. It may prove helpful.
A search of pubmed did not reveal any references to iodide or perchlorate clearance not already
A-13
-------
Brian Cummings
mentioned by the authors. The most recent article I could find on either subject was DeWoskin and
Thompson, 2008, which the authors use.
Why was a compartment analysis figure not shown for this revised model? Such figures are useful to
readers in conceptualizing the model. These were included in the literature on which the current model
was based.
G-3. Besides the specific parameters identified below, please comment on any other parameter or model
choices described in the document that you think are incorrect or require further explanation or
where other available information could provide better estimates.
The methods used for scaling of clearance to body weight, age and surface area are appropriate;
however, such scaling is most accurate for clearance when the substance in question is not reabsorbed or
secreted. Given that fact that both pendrin and NIS are reported to act on perchlorate, and given reports
that NIS expression does not scale to bodyweight in some tissues (see below), do the authors feel that
their approach is still valid? Do alterations in NIS expression need to be included in this model? If they
are, would this increase the risk for children age 10-14, when NIS expression is believed to altered?
GFR in children is typically scaled according to muscle mass, which scales well with the cube of height in
boys and girls from 6 months to adult (see Check, DB etal., Am. J. Clin. Nutr, 30:851, 1977). Scaling
formulas have even been derived for children based on creatinine levels (See Diseases in the Kidney,
Chapter 80, Seventh Edition, Editor = Schrier, Page 2355). Could these formulas be used to more
accurately reflect GFR in children when calculating perchlorate and iodide clearance?
G-4. Please discuss research that you think would be likely to increase confidence in these models and
their use in predicting RAIU inhibition by perchlorate in different life stages (for an average
individual within each life-stage).
Studies directly assessing the effect ofperchlorate on the clearance of NIS substrates are needed.
Further, data on the mechanisms ofperchlorate inhibition of NIS is lacking. Research investigating the
toxicity of substrates of NIS, in the presence and absence of perchlorate, is needed. Finally, urinary
clearances of environmental pollutants in infants and neonates are needed.
B-14
-------
Brian Cummings
G-5. Does EPA accurately characterize the strengths and limitations of the analysis? Please comment on
any particular strengths that may not be mentioned or adequately characterized for the estimates of
RAIU for different life stages. Please also comment on any specific sources of uncertainty that you
believe have been overlooked in EPA's analysis or that require further discussion, and which might
be significant to EPA's estimates of RAIU for different life stages.
EPA does a good job of characterizing the strengths and limitations of the model, but needs a separate
paragraph at the end directly addressing these points.
Did EPA take into account any hormone effects on NIS or thyroid function? The previous literature, on
which this model is based, devotes some discussion to this subject. This is particularly important
when discussing susceptibility during puberty.
G-6. As recommended in EPA's 2000 Risk Characterization Handbook, is the analysis transparent in
terms of the steps, logic, key assumptions, limitations, and decisions? Specifically, does
characterization of the results of EPA's work fully explain: a) the analysis approach employed; b)
the use of assumptions and their impact on the analysis; c) the use of extrapolations and their impact
on the analysis; d) plausible alternatives and the choices made among those alternatives; e) the
impacts of one choice v.v. another on the analysis; f) significant data gaps and their implications for
the analysis; g) the scientific conclusions identified separately from default assumptions and policy
decisions, if any; and h) the major conclusions, and the discussions of EPA's confidence and
uncertainties in the conclusions?
The analysis conforms to EPA guidelines on transparency with regards to steps and logic. The key
assumptions of the model are clearly listed, as well as the decisions involved in parameters changes and
impact of these changes. The limitations of the model could be more clearly listed (see above).
The analysis approach employed is adequately explained as are the basis for assumption used in this
model. EPA goes to great lengths to test the impact of these assumptions by determining the sensitivity
for each parameter changed.
The extrapolations used are clearly outlined and well as their rationale for them; however, for the most
part, only the theoretical impact of these extrapolations are discussed. Some validation is presented, but
this mostly uses previously validated data. Was the model applied to any data sets in the literature not
previously studied? Are such data available at this time?
The impacts of choice for most of the critical parameters are discussed. This is particular true when
discussing choices for BW, clearance and scaling these values.
A-15
-------
Brian Cummings
Plausible alternatives for some of the parameters could be more clearly listed. This is particularly true of
the impact on altering the level of water ingestion to 90%. What were the alternatives to this value and
how did they affect them model?
Alternative choices for scaling to BW and clearance are clearly listed as well as their impacts.
The major and scientific conclusions for this work are clearly stated and appear to be separated from any
grand statements on policy decisions.
The authors do discuss data gaps throughout the manuscript, but a specific section is needed, towards the
end, listing these gaps in itemized, or table form. This should be followed by paragraph that list major
perceived weaknesses and uncertainty of this model, which need to be more clearly stated.
Parameter-Specific Charge Questions:
(A) Urinary clearance
A-l. Please comment on the appropriateness of the input values selected for maternal urinary
clearance during pregnancy and lactation in EPA's analysis. Are the available data and rationale for
the values selected transparently and objectively described? Are you aware of other publications or
data that could be used to guide these choices or which provide alternative input values that are
equally valid or more appropriate? Likewise, are the values selected for urinary clearance for the
infant and older child the best estimates, given the available science and data? Are there other data
that would provide better (or equally valid alternative) guidance or estimates? Is there any reason to
believe that urinary clearance might not be a limiting factor in the elimination of perchlorate for
infants?
The authors do a good job discussing the choice and values used maternal urinary clearance during and
after pregnancy and lactation. This includes are significant discussion of the impact of changes made to
these parameters and alternatives.
The available data and analysis, for the most part, are rationale, transparently and objectively described,
with one exception (see C-2) below. I am not aware of any further publication that could be used for
input values that are more appropriate. One question that I did have is what was the rationale for
choosing the lower clearance value from Clew ell etal., (2007) as opposed to the others (Page 12, 2nd
B-16
-------
Brian Cummings
paragraph).
Please see my comments above concerning other data that provide alternative guidance with regards to
urinary clearance in neonates, infant and children. Several studies support the hypothesis that urinary
clearance is a limiting factor ofperchlorate elimination. While it's possible that it may not be the only
factor involved, it is clearly a major one. In absence of any data to the contrary, which could not be
found, the authors are correct in this assumption.
(B) Breast-milk ingestion
B-l. Infants are generally known to consume very little milk or formula on the first day of life, with
ingestion quickly increasing over time. However the first time-point for which ingestion data are
available for infants is at 7 days, which EPA extrapolated back to 0 ingestion at time=0 in its
analysis (birth; Figure 1). The breast-milk ingestion rate was estimated based on mean values
reported in Arcus-Arth et al. (2005), while the water ingestion used for the breast-feeding mother
was the 90th percentile value from U.S. EPA (2004).
Is EPA's extrapolation and rationale transparently and objectively described? Does this function
appropriately characterize the available data and information? Are there other data that could be
used to obtain a better or equally valid alternative estimate of mean breast-milk ingestion rate for
infants in the first few days of life? Should an estimate other than the mean be used to determine the
breast-milk ingestion rate?
EPA's rationale for ingestion rates for infants at time 0 follows rationale assumptions and the approach
appears objective. While its true that infants consume little in the first day of life, the extrapolation of 7
day data to day 1 allows for a margin of safety. I am not aware of any data that EPA has not presented.
While other estimates for breast-mile ingestion may exist, the ones used by the EPA are clear, rationale
and tractable. Thus, they will be easily validated in future models.
A-17
-------
Brian Cummings
(C) Water ingestion
C-l. For pregnancy, EPA used a normalized (90th percentile) water ingestion rate of 33 mL/kg-day
(U.S.EPA 2004), which was multiplied by the maternal BW as described by the PBPK model
growth-functions during pregnancy to obtain total water ingestion for the mother.
Is EPA's approach and rationale transparently and objectively described? Is this approach
appropriate for characterizing the upper-bound for ingestion of pregnant women? Are there other
approaches or data that could be used to obtain a better or equally valid alternative estimate for this
parameter? (Note that the self-reported body weight values in U.S. EPA (2004) indicate the same
average body weight for pregnant women as for non-pregnant; data appears inaccurate, so the PBPK
model body-weight description was used instead.)
EPA 's approach is rationale, transparent and objectively described. I agree with using 90% water
ingestion for pregnant woman for a upper bound rate, as it adds a safety factor; however clarification is
needed as to how this value was derived. A review of the documents used by EPA to determine this value
reports 90% bootstrapping levels as apposed to overall ingestions levels? Was the 90% bootstrap value
used, or did EPA calculate 90% ingestion from these data. How exactly was the value of 33 mL/kg-day
determined?
C-2. For lactation, EPA used a fixed total (90th percentile) water ingestion rate of 2959 mL/day (U.S. EPA
2004), with the rationale that while the woman's weight and self-water demand are expected to drop
after pregnancy (as described by the PBPK model time-dependent weight equations), the demand
for milk production would be increasing, and the reported value was not for a specific age-range of
child.
Is EPA's approach and rationale transparently and objectively described? Is this an appropriate
value to use for the ingestion rate of lactating women? Are there other better or equally valid
alternative approaches or values that could be used?
EPA approach is rational as it assumes that lactating women will still have increased water needs. While
the approach is transparently and objectively described I am have some questions about the actual level,
which results in -45.5 ml/kg-day, which is substantially higher than the 90% ingestion rate reported
above (assuming that the value was not a bootstrap). Doesn 't it seem more likely that water ingestion
would equalize? While the overall demand would decrease after pregnancy, this decrease would be
countered by lactation? Are the water demands for lactation higher than pregnancy? What data exist on
this subject other than models?
C-3. For bottle-fed infants, EPA made extrapolations of the 90th percentile water-ingestion in early life
B-18
-------
Brian Cummings
based on measurements made for the age ranges: 0-30 days, 1-3 months, and 6-12 months (points in
Figure 2, upper panel). For the purposes of this analysis, bottle-fed implies feeding with formula
requiring the addition of water.
Is EPA's extrapolation and rationale transparently and objectively described? Does the overall
function used to represent the changes in ingestion with age (and body weight) appropriately
characterize the available data and information? Are there other data that could be used to obtain a
better or equally valid alternative estimate of water ingestion rate for infants in the first few days of
life (e.g., 7-day old)? The water ingestion rates used by EPA are based on 90th percentile ingestion
data and thus are likely to exceed (minimal) physiological needs of infants as defined in nutritional
guidelines. Are the water ingestion rates used by EPA reasonable in comparison to the
physiological needs of infants at these various life stages? Are there other approaches or data (e.g.,
see the FDA memo) that could be used to obtain a better or equally valid alternative estimate for
this parameter?
EPA approach is rationale and transparently described. It is standard practice to scale water ingestion
to body weight, which does increase with age. Weight gain in the new born scales more rapidly than
almost any other time period and it sequestered into specific groups; however, why was the calculation
not scaled to 2 and 3 years as it done by WHO?
Another point of interest is in regards to the 1st seven days of birth. Most infants either maintain birth
weight or slightly lose 10% of their birth weight. This, as pointed out by EPA is directly related to water
ingestion. Should this than represent another group or be removed (i.e. 0-7 days, or 7-30 days?).
(D) Perchlorate concentration in formula
D-l. EPA used 1.42 j^ig/L as the concentration of perchlorate in formula for bottle-fed infants. This
estimate was based on information from FDA's Total Diet Study, supported by Pearce el al.'s
(2007) findings.
Is EPA's approach and rationale transparently and objectively described? Is 1.42 (ig/L an
appropriate value to use for the concentration of perchlorate in infant formula? Are there other
better or equally valid alternative approaches or values that could be used?
EPA's approach and rationale are transparent and well described. The approach is logical and based
on the most current information. A search for other levels for this values results in similar results. This
is obvious an area for which more data is needed (see above).
A-19
-------
Brian Cummings
(E) Radioiodide excretion into breast-milk by NIS
E-l. In the model, EPA included perchlorate inhibition of NIS radioiodide excretion into breast-milk, as
well as inhibition of radioiodide transport by perchlorate for all NIS-containing tissues, thereby
making the code consistent with the model description in Clewell et al. (2007). Is this inclusion
appropriate? Are the impacts of this inclusion transparently and objectively described?
This inclusion is appropriate because it's unlikely that a situation exists where perchlorate or iodide is
absent from the diet. This represents a logical and important refinement in the model. However, a
section is needed, towards the end, which clearly discusses the impacts of perchlorate inhibition.
Further, the rationale for scaling NIS to body weight and other tissues is not clear. Studies suggest that
NIS expression changes over development (see below). At least one study suggests that the expression of
NIS is higher per g of tissue in young children (<12 years) compared to adults. Further, another study
suggests that this difference accounts for higher levels of iodide uptake in children than in adults. Thus,
scaling NIS levels to body weight (i.e. age) may not be appropriate.
Faggino et al., Journal of Nuclear Medicine, 45(2), 232-237.
Miscellanous Comments;
1. An extra period is present in the first bullet point on page 59.
2. The third paragraph on page 42 refers to Figure B-l. Should this be Figure B-2?
3. Please list the years for the references for Merrill et al. listed on page 39.
B-20
-------
PEER REVIEW COMMENTS FROM
Panos Georgopolous, Ph.D.
Professor, Department of Environmental
& Occupational Medicine
Rutgers University
170 Frelinghuysen Road
Piscataway, NJ 08854
732-445-0159
Email: panosg@fidelio.rutgers.edu
A-21
-------
-------
Panos Georgopolous
CHARGE QUESTIONS
General Charge Questions:
G-l. Is EPA's analysis logical, clear and appropriate in depth and length? Has EPA accurately,
clearly and objectively represented and synthesized the scientific evidence for the changes
made to or specification of the model code and input parameters?
The document under review, "Inhibition of the Sodium-Iodide Symporter (NIS) by Perchlorate: an
Evaluation of Lifestage Sensitivity Using Physiologically-Based Pharmacokinetic (PBPK) Modeling",
has a very specific, and deliberately narrow, objective. Indeed it defines (p. 21) sensitivity "as the
predicted response in percent RAIU (radioactive iodide uptake) inhibition 24 hours after iodide
intravenous injection for an average individual within a specific subgroup (e.g., bottle-fed infants) relative
to the predicted response in percent RAIU inhibition for an average, non-pregnant adult, where response
is the percent RAIU inhibition 24 hours after iodide IV injection." Though this is rather constrained as a
sensitivity metric, it can be reasonably argued that it addresses adequately the biological issue of concern
here.
The emphasis of the analysis is on sensitivity with respect to "lifestage". Parameters and processes,
related to different lifestages, were modeled based on assumptions that are discussed in rather extensive
detail in the document under review. The lifestages evaluated in the document correspond to "average"
adult, non-pregnant woman of child-bearing age, pregnant woman, lactating woman, fetus, breast-fed
infant, bottle-fed infant, 1 year old child, and and 2 year old child. The tools employed for the analysis
were the PBPK models of Clewell etal. (2007) for the pregnant woman/fetus and for the lactating
woman/breastfed infant. Results for the "bottle-fed" neonate were obtained by altering the dose
specification in the model for the breast-fed infant. The PBPK model for the average adult was that of
Merrill et al. (2005), while the model for the non-pregnant woman of childbearing age was a direct
modification of the model for the pregnant woman, obtained by removing the placental and fetal
compartments, but retaining the mammary compartment.
The above PBPK models, with various corrections and adjustments (that are discussed in detail in the
appendices of the document) were used to estimate the predicted percent RAIU inhibition for the average
adult and different specific ("average") individuals representing potentially sensitive subgroups. It should
be mentioned here that the actual text of the document under review states that the calculations were
made for "subgroups, including potentially sensitive subgroups"; however population-based modeling
(with considerations of inter-individual and intra-individual variability) was not actually pursued.
A-23
-------
Panos Georgopolous
"Base" calculations were made assuming a dose equal to the point of departure (POD) of 7 (ig/kg-day,
(consistent with the recommendations of the National Research Council - NRC, 2005) and were
summarized in Table 3 of the document under review. The relative sensitivity of different subgroups was
determined by comparing the percent RAIU inhibition of each subgroup to the percent RAIU inhibition
for an average adult at a dose equal to the POD.
The document states that the "model predictions may generally be considered central estimates for each
subgroup (at the consumption levels modeled) that account for PK (pharmacokinetic) differences, and do
not take into account within-group variability in pharmacokinetics, uncertainty in model parameters and
predictions, or population differences in PD." It should be noted that fetal simulations were reported for
only the end of gestation (Gestation Week 40).
The analysis presented in the document concluded that urinary clearance was a "key" parameter (i.e.,
model predictions were highly sensitive to the values of this variable). Though for modeling pregnancy
and early infancy a conservative parameterization was adopted, the document emphasizes that "a full
population analysis of urinary clearance was not conducted, and given that variability in other PK
parameters was not addressed, these estimates should not be considered a true upper confidence bound on
RAIU inhibition" (p. 23). The document also identified the fetus as the most sensitive subgroup with
respect to percent RAIU inhibition at a dose equal to the POD, in general agreement with earlier PBPK
modeling (Clewell el al.. 2007) and estimating approximately 5-fold higher percent RAIU inhibition for
the fetus at gestational week 40 than for the average adult. In fact it is also stated that "simulations at
earlier gestation weeks indicate that the fetus is more sensitive than the adult throughout pregnancy, but
are considered too quantitatively uncertain to assign exact relative sensitivities" (p. 23).
Overall it can be stated that EPA's analysis is clear and with sufficient discussion of assumptions
involving model and parameter specification (including adjustments and corrections to the original
models and their codes). It should be noted, however, that in multiple instances (discussed further in the
answers to the following questions) the rationale behind specific assumptions and parameterizations
relates more to "convenience" rather than to scientific defensibility. Although this may not necessarily
affect the general conclusions, it is nevertheless a weakness of the analysis presented in the document
under review.
References Cited in Answer to Question G-l:
Clewell, R.A., Merrill, E.A., Gearhart, J.M., Robinson, P.J., Sterner, T.R., Mattie, D.R., and Clewell,
H.J., 3rd. 2007. Perchlorate and radioiodide kinetics across life stages in the human: using PBPK
B-24
-------
Panos Georgopolous
models to predict dosimetry and thyroid inhibition and sensitive subpopulations based on
developmental stage. J Toxicol Environ Health A 70 (5):408-28.
Merrill, E.A., Clewell, R.A., Robinson, P.J., Jarabek, A.M., Gearhart, J.M., Sterner, T.R., and Fisher,
J.W. 2005. PBPK model for radioactive iodide and perchlorate kinetics and perchlorate-induced
inhibition of iodide uptake in humans. Toxicol Sci 83 (l):25-43.
NRC. 2005. Health Implications of Perchlorate Ingestion. National Research Council of the National
Academies, National Academies Press. Washington, D.C. http://www.nap.edu/catalog/11202.html
G-2. Please identify any additional studies or other data sources that should be considered in the
assessment of the specific parameters addressed below, including urinary clearance (of iodide
and/or perchlorate) and ingestion rates (breast milk, formula and water), especially in
neonates.
The scientific literature relevant to NIS inhibition by perchlorate is currently growing fast; the same holds
true for related literature areas covering fields such as demographics and exposure informatics and
modeling, Physiologically-Based Pharmacokinetic and Pharmacodynamic modeling methods, etc.
Though the document under review is not expected to provide a thorough literature review of the subject
of perchlorate inhibition of NIS and of related exposure and risk issues, it could certainly provide a more
complete picture to its readers, by incorporating some of the references suggested below.
These suggestions are grouped in three categories: (a) "general references," that cover various aspects of
perchlorate exposure and effect, (b) references that focus on studies of human exposure to perchlorate,
and (c) references that focus on biological (physiological and biochemical) issues, either directly specific
to perchlorate and NIS inhibition or indirectly related, such as e.g. references on information for urinary
clearance related parameters or on information for PBPK modeling specific to infants.
It should be noted in particular that USFDA (The US Food and Drug Administration) has developed
PBPK modeling recommendations, as well as computer software that implements them, for early life
stages (Luecke et al., 2007, 2008); at a minimum, it would be useful to examine how these
parameterizations compare to the ones adopted in the analysis presented in the document under review.
(Similarly, it would be useful to compare exposure-related parameter selections used in the reviewed
work to corresponding relevant recommendations in USEPA's Child-Specific Exposure Factors
Handbook).
A-25
-------
Panos Georgopolous
General:
ATSDR. 2008. Toxicological Profile for Perclorates. Agency for Toxic Substances and Disease Registry.
Atlanta, GA. http://www.atsdr.cdc.gov/toxprofiles/tpl62.pdf
Charnley, G. 2008. Perchlorate: Overview of risks and regulation. Food and Chemical Toxicology 46
(7):2307-2315.
De Groef, B., Decallonne, B.R., Van der Geyten, S., Darras, V.M., and Bouillon, R. 2006. Perchlorate
versus other environmental sodium/iodide symporter inhibitors: potential thyroid-related health
effects. Eur J Endocrinol 155 (1): 17-25.
Gu, B., and Coates, J.D. 2006. Perchlorate: Environmental Occurrence, Interactions and Treatment. New
York: Springer.
Kirk, A.B. 2006. Environmental perchlorate: why it matters. Anal ChimActa 567 (1):4-12.
Kirk, A.B., Dyke, J.V., Martin, C.F., and Dasgupta, P.K. 2007. Temporal patterns in perchlorate,
thiocyanate, and iodide excretion in human milk. Environ Health Perspect 115 (2): 182-6.
Kirk, A.B., Martinelango, P.K., Tian, K., Dutta, A., Smith, E.E., and Dasgupta, P.K. 2005. Perchlorate
and iodide in dairy and breast milk. Environ Sci Technol 39 (7):2011-7.
Wang, R.Y., and Needham, L.L. 2007. Environmental chemicals: from the environment to food, to breast
milk, to the infant. J Toxicol Environ Health B Crit Rev 10 (8):597-609.
Exposure:
Baier-Anderson, C., Blount, B.C., Lakind, J.S., Naiman, D.Q., Wilbur, S.B., and Tan, S. 2006. Estimates
of exposures to perchlorate from consumption of human milk, dairy milk, and water, and
comparison to current reference dose. J Toxicol Environ Health A 69 (3-4):319-30.
Blount, B.C., Valentin-Blasini, L., Osterloh, J.D., Mauldin, J.P., and Pirkle, J.L. 2007. Perchlorate
exposure of the US Population, 2001-2002. J Expo Sci Environ Epidemiol 17 (4):400-7.
Ginsberg, G.L., Hattis, D.B., Zoeller, R.T., and Rice, D.C. 2007. Evaluation of the U.S. EPA/OSWER
preliminary remediation goal for perchlorate in groundwater: focus on exposure to nursing infants.
Environ Health Perspect 115 (3):361-9.
Zender, R., Bachand, A.M., and Reif, J.S. 2001. Exposure to tap water during pregnancy. J Expo Anal
Environ Epidemiol 11 (3):224-30.
Physiological/Biochemical:
Brandt, J.R., Wong, C.S., Hanrahan, J.D., Quails, C., McAfee, N., and Watkins, S.L. 2006. Estimating
absolute glomerular filtration rate in children. Pediatr Nephrol 21 (12): 1865-72.
Clewell, R.A., and Gearhart, J.M. 2002. Pharmacokinetics of toxic chemicals in breast milk: use of PBPK
models to predict infant exposure. Environ Health Perspect 110 (6):A333-7.
B-26
-------
Panos Georgopolous
Dohan, O., De la Vieja, A., Paroder, V., Riedel, C., Artani, M., Reed, M., Ginter, C.S., and Carrasco, N.
2003. The Sodium/Iodide Symporter (NIS): Characterization, Regulation, and Medical
Significance. Endocrine Reviews 24 (l):48-77.
Hawcutt, D.B., and Smyth, R.L. 2008. One size does not fit all: getting drug doses right for children.
Archives of Disease in Childhood 93 (3): 190-191.
Ito, S., and Alcorn, J. 2003. Xenobiotic transporter expression and function in the human mammary
gland. Adv Drug Deliv Rev 55 (5):653-65.
Johnson, T.N. 2008. The problems in scaling adult drug doses to children. Arch Dis Child 93 (3):207-l 1.
Kurz, H., Sandau, K., Dawson, T.H., Brown, J.H., Enquist, B.J., and West, G.B. 1998. Allometric ccaling
in biology. Science 281 (5378):751a-.
Lewandowski, T.A., Seeley, M.R., and Beck, B.D. 2004. Interspecies differences in susceptibility to
perturbation of thyroid homeostasis: a case study with perchlorate. Regul Toxicol Pharmacol 39
(3):348-62.
Luecke, R.H., Pearce, B.A., Wosilait, W.D., Slikker, W., Jr., and Young, J.F. 2007. Postnatal growth
considerations for PBPK modeling. J Toxicol Environ Health A 70 (12): 1027-37.
Luecke, R.H., Pearce, B.A., Wosilait, W.D., Doerge, D.R., Slikker, W., Jr., and Young, J.F. 2008.
Windows based general PBPK/PD modeling software. Comput Biol Med 38 (9):962-78.
McManaman, J.L., and Neville, M.C. 2003. Mammary physiology and milk secretion. Adv Drug Deliv
Rev 55 (5):629-41.
Packard, G.C., and Birchard, G.F. 2008. Traditional allometric analysis fails to provide a valid predictive
model for mammalian metabolic rates. J Exp Biol 211 (Pt 22):3581-7.
Spitzweg, C., Dutton, C.M., Castro, M.R., Bergert, E.R., Goellner, J.R., Heufelder, A.E., and Morris, J.C.
2001. Expression of the sodium iodide symporter in human kidney. Kidney Int 59 (3): 1013-23.
Strawson, J., Zhao, Q., and Dourson, M. 2004. Reference dose for perchlorate based on thyroid hormone
change in pregnant women as the critical effect. Regulatory Toxicology and Pharmacology 39
(l):44-65.
West, G.B., Brown, J.H., and Enquist, B.J. 1997. A general model for the origin of allometric scaling
laws in biology. Science 276 (5309): 122-126.
G-3. Besides the specific parameters identified below, please comment on any other parameter or
model choices described in the document that you think are incorrect or require further
explanation or where other available information could provide better estimates.
As discussed in more detail in the answer to the questions regarding the characterization of urinary
clearance processes, there is a need to develop and thoroughly test a consistent framework for modeling
A-27
-------
Panos Georgopolous
these processes for different lifestages. "Correcting" the inconsistencies, that are in fact identified in
Appendix B of the document under review, would be a first step towards the implementation of such a
framework.
G-4. Please discuss research that you think would be likely to increase confidence in these models
and their use in predicting RAIU inhibition by perchlorate in different life stages (for an
average individual within each life-stage).
The main challenge related to this question is "defining" an "average individual within each lifestage."
Equally challenging would be the identification of an "average exposure" within each lifestage. EPA
should consider the merits of a probabilistic sensitivity/uncertainty analysis and eventually the feasibility
of population-based PBPK modeling (with explicitly defined sensitive subpopulations). Such an approach
will provide a more realistic assessment of the actual ranges of the outcomes considered in the analysis,
will eventually improve risk characterization efforts, and help explain both inter-individual and intra-
individual variability within a (sub) population.
G-5. Does EPA accurately characterize the strengths and limitations of the analysis? Please
comment on any particular strengths that may not be mentioned or adequately characterized
for the estimates of RAIU for different life stages. Please also comment on any specific
sources of uncertainty that you believe have been overlooked in EPA's analysis or that
require further discussion, and which might be significant to EPA's estimates of RAIU for
different life stages.
In this reviewer's opinion the main strength of the analysis is the explicit listing of unresolved issues (and
inconsistencies) in the modeling described in the document under review.
The main limitations involve:
• the emphasis on point estimates rather than on pursuing a distributional (probabilistic) approach
for characterizing exposures (with explicit variability and uncertainty of activities for each
individual and across a sub-population,
• the emphasis on "average" individuals for each lifestage rather than on pursuing population-based
modeling with explicit characterization of inter-individual and intra-individual pharmacokinetic
(physiological and biochemical) variabilities, and
• the consideration of iodide and perchlorate exposure "in isolation" and not in a context of "total"
exposure that would consider other NIS inhibitors (thiocyanate, nitrates).
B-28
-------
Panos Georgopolous
G-6. As recommended in EPA's 2000 Risk Characterization Handbook, is the analysis transparent
in terms of the steps, logic, key assumptions, limitations, and decisions? Specifically, does
characterization of the results of EPA's work fully explain: a) the analysis approach
employed; b) the use of assumptions and their impact on the analysis; c) the use of
extrapolations and their impact on the analysis; d) plausible alternatives and the choices
made among those alternatives; e) the impacts of one choice vs. another on the analysis; f)
significant data gaps and their implications for the analysis; g) the scientific conclusions
identified separately from default assumptions and policy decisions, if any; and h) the major
conclusions, and the discussions of EPA's confidence and uncertainties in the conclusions?
The document and - in particular the appendices - are quite explicit ("transparent") in listing and
discussing all the assumptions and approximations involved in the analysis. This takes place at a level of
detail that exceeds what is typically expected in the peer reviewed literature and the authors of the
document under review should be commended for this. However, the justifications of what can be called
"emergency solutions" to various problems discussed in the document, especially those related to
inconsistencies in modeling urinary clearance processes (including inconsistencies in scaling factors as
well as "ad hoc" adjustments to achieve agreement for predicted values) are often weak. It can, probably
reasonably, be argued that, for the ranges of concentrations and exposures considered, the effect of
correcting the above inconsistencies will not have a substantive impact on calculated outcomes; however,
it is still very important to develop a "fully defensible" model that incorporates up-to-date scientific
information and assumptions that are consistent "across lifestages". Clearly, resolving the inconsistencies
that have been identified through the efforts presented in the document under review will be useful in
numerous other applications involving exposures to chemicals in utero and during infancy.
Parameter-Specific Charge Questions:
(A) Urinary clearance
A-l. Please comment on the appropriateness of the input values selected for maternal urinary
clearance during pregnancy and lactation in EPA's analysis. Are the available data and
rationale for the values selected transparently and objectively described? Are you aware of
other publications or data that could be used to guide these choices or which provide
alternative input values that are equally valid or more appropriate? Likewise, are the values
selected for urinary clearance for the infant and older child the best estimates, given the
available science and data? Are there other data that would provide better (or equally valid
alternative) guidance or estimates? Is there any reason to believe that urinary clearance
might not be a limiting factor in the elimination of perchlorate for infants?
The analysis presented in the document under review concluded that urinary clearance was a "key"
process/parameter; i.e. model predictions were very sensitive with respect to the magnitude of urinary
clearance. However, there are a number of issues concerning the parameters employed in modeling
urinary clearance that need substantial clarification. Appendix B (pp 39-50) of the document under review
A-29
-------
Panos Georgopolous
provides an extensive discussion of the assumptions and approximations involved in selecting and
estimating these parameters. Various inconsistencies in the selection/estimation procedures are in fact
recognized explicitly in Appendix B, but in general these inconsistencies are "accepted" on the basis of
either a minimal anticipated effect on the calculations of the model, or as a means for avoiding a more
complex analysis. For example, the last paragraphs of p. 40 states that because "... renal clearance is
largely controlled by glomerular filtration and non-specific fluid resorption, the expectation is that the
relative clearance for iodide and perchlorate [....] should be constant across ages, body weights, and
lifestages. In EPA's evaluation for the child and "average" (non-pregnant, non-lactating) adult, this
proportionality has been maintained." However, in the model of Clewell et al. (2007) "the maternal
urinary clearance value [....] was set at 60% of the value in the non-pregnant human based on observed
difference in the pregnant and male rat models [....].These maternal lactation values go against the
argument given just above that the proportionality should be maintained, but EPA chose to use the
maternal values as so set. It is likely worthwhile to evaluate these maternal values in light of the generally
higher urinary excretion seen in pregnant/lactating women, but alteration of these clearance constants
would require refitting of other parameters, and so EPA chose not to conduct that specific evaluation."
Clearly this is an issue that requires further consideration. It should also be mentioned that this discussion
is preceded by the following rather puzzling statement that "The tables in the papers identify the units of
[urinary clearance] as L/h/kg, but clearly this should be L/h/kg°75 to be consistent with this mathematical
formulation, which is how the CLU values are calculated in the computer code." Such a selection
units/dimensions contradict the physics of the problems and in fact it appears that the tables in the original
articles (Clewell et al., 2007 and Merrill et al., 2005) state the correct units. (The first paragraph on page
41 of the document under review also employs correct units/dimensions.) The issue of consistent
allometric scaling is an important one and there exist various publications that can be helpful in clarifying
issues such as the above (e.g. Johnson, 2008; West et al., 1997; Kurz et al., 1998). Issues of inconsistent
scaling in fact appear across the entire description of urinary clearance parameters (Appendix B).
Various other inconsistencies are discussed and "accepted" in relation to the calculations of urinary
clearance in the neonate (pages 42-43) and in the pregnant/lactating woman. For example, on page 47, it
is stated:
"Keeping with the assumed proportionality between perchlorate and iodide, based on these data
the same relationship would be expected to hold: higher clearance rather than reduced. A
dilemma occurs in considering the data of Aboul-Khair et al. (1964); however, in that the control
iodine clearance as measured by them is 31.05 ± 3.66 mL/min (mean ± SE), while the value
determined by Merrill et al. (2005) for non-pregnant adults is 44.3 mL/min. Likewise Aboul-
B-30
-------
Panos Georgopolous
Khair et al. (1964) report thyroid iodide uptake at 2.5 hr postinjection as 21.4 ± 1.4 % of the
administered dose, but the amount predicted by the Merrill et al. (Merrill et al., 2005) model (in
the absence of perchlorate) is 7.78%. Therefore, the data of Aboul-Khair et al. (1964) was
normalized to their own controls for both urinary clearance and iodide uptake, and then use that
relative change as a model input (for clearance, multiplying the non-pregnant clearance rate
constant by the pregnant:control ratio from Aboul-Khair et al. (1964) or in estimating changes in
thyroid NIS (to fit relative increases in thyroid uptake)."
Clearly, the inconsistency in absolute values reported in the above paragraph should be the focus of
further study; while the normalization employed by EPA offers a way of circumventing the issue, this
"solution" could only be considered qualitative in nature.
In this reviewer's opinion, a consistent treatment of the urinary clearance process for various life stages
emerges clearly as a research need, based on the outcomes of the sensitivity testing and the issues
presented in Appendix B of the document under review.
References Cited in Answer to Question A-l:
Clewell, R.A., Merrill, E.A., Gearhart, J.M., Robinson, P.J., Sterner, T.R., Mattie, D.R., and Clewell,
H.J., 3rd. 2007. Perchlorate and radioiodide kinetics across life stages in the human: using PBPK
models to predict dosimetry and thyroid inhibition and sensitive subpopulations based on
developmental stage. J Toxicol Environ Health A 70 (5):408-28.
Johnson, T.N. 2008. The problems in scaling adult drug doses to children. Arch Dis Child 93 (3):207-l 1.
Kurz, H., Sandau, K., Dawson, T.H., Brown, J.H., Enquist, B.J., and West, G.B. 1998. Allometric ccaling
in biology. Science 281 (5378):751.
Merrill, E.A., Clewell, R.A., Robinson, P.J., Jarabek, A.M., Gearhart, J.M., Sterner, T.R., and Fisher,
J.W. 2005. PBPK model for radioactive iodide and perchlorate kinetics and perchlorate-induced
inhibition of iodide uptake in humans. Toxicol Sci 83 (l):25-43.
West, G.B., Brown, J.H., and Enquist, B.J. 1997. A general model for the origin of allometric scaling
laws in biology. Science 276 (5309): 122-126.
A-31
-------
Panos Georgopolous
(B) Breast-milk ingestion
B-l. Infants are generally known to consume very little milk or formula on the first day of life,
with ingestion quickly increasing over time. However the first time-point for which ingestion
data are available for infants is at 7 days, which EPA extrapolated back to 0 ingestion at
time=0 in its analysis (birth; Figure 1). The breast-milk ingestion rate was estimated based on
mean values reported in Arcus-Arth et al. (2005), while the water ingestion used for the
breast-feeding mother was the 90th percentile value from U.S. EPA (2004).
Is EPA's extrapolation and rationale transparently and objectively described? Does this
function appropriately characterize the available data and information? Are there other data
that could be used to obtain a better or equally valid alternative estimate of mean breast-milk
ingestion rate for infants in the first few days of life? Should an estimate other than the mean
be used to determine the breast-milk ingestion rate?
EPA's selection of point values for the testing analysis appears appropriate and adequately justified as a
reasonable conservative assumption. However, in this reviewer's opinion, the uncertainties and variability
inherent in the problem at hand would be better addressed by a distributional (probabilistic) rather than
point calculation. The large population above the 90th percentile and the potential "spread" of exposure
factors above that percentile, would further justify such an analysis.
(C) Water ingestion
C-l. For pregnancy, EPA used a normalized (90th percentile) water ingestion rate of 33 mL/kg-
day (U.S.EPA 2004), which was multiplied by the maternal BW as described by the PBPK
model growth-functions during pregnancy to obtain total water ingestion for the mother.
Is EPA's approach and rationale transparently and objectively described? Is this approach
appropriate for characterizing the upper-bound for ingestion of pregnant women? Are there
other approaches or data that could be used to obtain a better or equally valid alternative
estimate for this parameter? (Note that the self-reported body weight values in U.S. EPA
(2004) indicate the same average body weight for pregnant women as for non-pregnant; data
appears inaccurate, so the PBPK model body-weight description was used instead.)
The approach taken by EPA appears reasonable. As in the answer to the previous question, this reviewer's
opinion is that a distributed zonal analysis (Monte Carlo) can provide more substantial insight on patterns
of potential exposure rather than the point calculations presented here.
B-32
-------
Panos Georgopolous
C-2. For lactation, EPA used a fixed total (90th percentile) water ingestion rate of 2959 mL/day
(U.S. EPA 2004), with the rationale that while the woman's weight and self-water demand are
expected to drop after pregnancy (as described by the PBPK model time-dependent weight
equations), the demand for milk production would be increasing, and the reported value was
not for a specific age-range of child.
Is EPA's approach and rationale transparently and objectively described? Is this an
appropriate value to use for the ingestion rate of lactating women? Are there other better or
equally valid alternative approaches or values that could be used?
The approach taken by EPA appears reasonable, though a probabilistic (Monte Carlo) analysis would
provide additional insight regarding the range of potential exposures. Furthermore, since the simulations
for lactating women produce estimates of perchlorate concentration in breast milk, a
population/distribution-level analysis with appropriate parameterizations could be used to provide
valuable testing of the model in relation to the available data presented in Pearce el al. (2007) as well as
Kirk etal. (2005,2007)
References Cited in Answer to Question B-2:
Kirk, A.B., Martinelango, P.K., Tian, K., Dutta, A., Smith, E.E., and Dasgupta, P.K. 2005. Perchlorate
and iodide in dairy and breast milk. Environ Sci Technol 39 (7):2011-7.
Kirk, A.B., Dyke, J.V., Martin, C.F., and Dasgupta, P.K. 2007. Temporal patterns in perchlorate,
thiocyanate, and iodide excretion in human milk. Environ Health Perspect 115 (2): 182-6.
Pearce, E.N., Leung, A.M., Blount, B.C., Bazrafshan, H.R., He, X., Pino, S., Valentin-Blasini, L., and
Braverman, L.E. 2007. Breast milk iodine and perchlorate concentrations in lactating Boston-area
women. J Clin EndocrinolMetab 92 (5): 1673-7.
A-33
-------
Panos Georgopolous
C-3. For bottle-fed infants, EPA made extrapolations of the 90th percentile water-ingestion in
early life based on measurements made for the age ranges: 0-30 days, 1-3 months, and 6-12
months (points in Figure 2, upper panel). For the purposes of this analysis, bottle-fed implies
feeding with formula requiring the addition of water.
Is EPA's extrapolation and rationale transparently and objectively described? Does the
overall function used to represent the changes in ingestion with age (and body weight)
appropriately characterize the available data and information? Are there other data that
could be used to obtain a better or equally valid alternative estimate of water ingestion rate
for infants in the first few days of life (e.g., 7-day old)? The water ingestion rates used by
EPA are based on 90th percentile ingestion data and thus are likely to exceed (minimal)
physiological needs of infants as defined in nutritional guidelines. Are the water ingestion
rates used by EPA reasonable in comparison to the physiological needs of infants at these
various life stages? Are there other approaches or data (e.g., see the FDA memo) that could be
used to obtain a better or equally valid alternative estimate for this parameter?
EPA's extrapolation and rationale are adequately described. However, it is doubtful that the use of any
single-point estimate would provide adequate understanding of the potential range of exposures, and
corresponding doses, for bottle-fed infants.
(D) Perchlorate concentration in formula
D-l. PA used 1.42 jig/L as the concentration of perchlorate in formula for bottle-fed infants. This
estimate was based on information from FDA's Total Diet Study, supported by Pearce et al.'s
(2007) findings.
Is EPA's approach and rationale transparently and objectively described? Is 1.42 jig/L an
appropriate value to use for the concentration of perchlorate in infant formula? Are there
other better or equally valid alternative approaches or values that could be used?
The rationale for the selection of 1.42 (ig/L as a preset value for the concentration of perchlorate for
bottle-fed infants is not adequately discussed. This value is the average of 12 samples (8 of them above
detection limit) presented in Murray et al. (2008); it is also close to the average value (1.45 ppb) of the 17
samples analyzed by Pearce et al. (2007). It should be noted that the values of perchlorate concentrations
in the samples of Pearce et al. range from 0.2 to 4.1 ppb. It would be useful to examine the sensitivity of
uptake for a reasonable concentration range rather than only the average value.
References Cited in Answer to Question D-l:
Murray, C.W., Egan, S.K., Kim, H., Bern, N., and Bolger, P.M. 2008. US Food and Drug
Administration's Total Diet Study: Dietary intake of perchlorate and iodine. J Expos Sci Environ
Epidemiol 18 (6):571-580.
B-34
-------
Panos Georgopolous
Pearce, E.N., Leung, A.M., Blount, B.C., Bazrafshan, H.R., He, X., Pino, S., Valentin-Blasini, L., and
Braverman, L.E. 2007. Breast milk iodine and perchlorate concentrations in lactating Boston-area
women. J Clin EndocrinolMetab 92 (5): 1673-7.
(E) Radioiodide excretion into breast-milk by NIS
E-l. In the model, EPA included perchlorate inhibition of NIS radioiodide excretion into breast-
milk, as well as inhibition of radioiodide transport by perchlorate for all NIS-containing
tissues, thereby making the code consistent with the model description in Clewell et al. (2007).
Is this inclusion appropriate? Are the impacts of this inclusion transparently and objectively
described?
The inclusion of perchlorate inhibition of NIS radioiodide excretion into breast-milk, as well as of
inhibition of radioiodide transport by perchlorate for all NIS-containing tissues, is appropriate. The
impacts of this inclusion are adequately ("transparently and objectively") described in the document.
A-35
-------
-------
PEER REVIEW COMMENTS FROM
Sean Hays, M.S.
President, Summit Toxicology, L.L.P.
Allenspark, CO 80510
303-747-0722
Email: shavs@,summittoxicology.com
A-37
-------
-------
General Charge Questions:
G-l. Is EPA's analysis logical, clear and appropriate in depth and length? Has EPA accurately, clearly
and objectively represented and synthesized the scientific evidence for the changes made to or
specification of the model code and input parameters?
YES.
G-2. Please identify any additional studies or other data sources that should be considered in the
assessment of the specific parameters addressed below, including urinary clearance (of iodide and/or
perchlorate) and ingestion rates (breast milk, formula and water), especially in neonates.
NA
G-3. Besides the specific parameters identified below, please comment on any other parameter or model
choices described in the document that you think are incorrect or require further explanation or
where other available information could provide better estimates.
NA
G-4. Please discuss research that you think would be likely to increase confidence in these models and
their use in predicting RAIU inhibition by perchlorate in different life stages (for an average
individual within each life-stage).
The results of this modeling effort should be classified as theoretical since no validation
exercises have been performed and EPA should be clear to state this. There are existing data
which could help to validate the model predictions, especially related to the most sensitive
scenario (e.g., the nursing infant of the exposed mother). In particular, Pearce et al. (2007)
provides matched data on perchlorate and iodine in breast milk and urine samples from
nursing mothers. EPA should obtain this data from the study authors. This will greatly help to
test the model predictions. Furthermore, the authors found no correlation (either positive or
negative) between perchlorate and iodine in breast milk samples. The authors of this study
indicate this is consistent with other researchers. This may raise questions about the results of
EPA's modeling efforts. Since no results for the concentrations of perchlorate and iodine in
milk as a function of perchlorate dose are provided in EPA's report, it is impossible to
determine the validity of this issue.
A-39
-------
Sean Hays
G-5. Does EPA accurately characterize the strengths and limitations of the analysis? Please comment on
any particular strengths that may not be mentioned or adequately characterized for the estimates of
RAIU for different life stages. Please also comment on any specific sources of uncertainty that you
believe have been overlooked in EPA's analysis or that require further discussion, and which might
be significant to EPA's estimates of RAIU for different life stages.
In places, EPA adequately highlights the strengths and weaknesses of their analysis. However,
there are other areas where the limitations have not been adequately addressed (e.g., lack of
validation).
G-6. As recommended in EPA's 2000 Risk Characterization Handbook, is the analysis transparent in
terms of the steps, logic, key assumptions, limitations, and decisions? Specifically, does
characterization of the results of EPA's work fully explain: a) the analysis approach employed; b)
the use of assumptions and their impact on the analysis; c) the use of extrapolations and their impact
on the analysis; d) plausible alternatives and the choices made among those alternatives; e) the
impacts of one choice v.v. another on the analysis; f) significant data gaps and their implications for
the analysis; g) the scientific conclusions identified separately from default assumptions and policy
decisions, if any; and h) the major conclusions, and the discussions of EPA's confidence and
uncertainties in the conclusions?
Yes, the analysis is transparent.
Parameter-Specific Charge Questions:
(A) Urinary clearance
A-l. Please comment on the appropriateness of the input values selected for maternal urinary
clearance during pregnancy and lactation in EPA's analysis. Are the available data and rationale for
the values selected transparently and objectively described? Are you aware of other publications or
data that could be used to guide these choices or which provide alternative input values that are
equally valid or more appropriate? Likewise, are the values selected for urinary clearance for the
infant and older child the best estimates, given the available science and data? Are there other data
that would provide better (or equally valid alternative) guidance or estimates? Is there any reason to
believe that urinary clearance might not be a limiting factor in the elimination of perchlorate for
infants?
The available data and rationale are transparently described. However, I disagree with the
rationale for the choice of maternal urinary clearance values. This is probably the most
sensitive parameter for the most sensitive scenario/receptor and the EPA has chosen to use
rodent data over human data. This is inadequate. EPA should choose to use the available
human data which indicates there is no measurable or consistent difference in urinary
clearance during pregnancy as compared to the non-pregnant state. The EPA chose a
B-40
-------
reasonable urinary clearance for the lactating mother scenario. I agree with EPA's choice for
the urinary clearance among infants and older children.
(B) Breast-milk ingestion
B-l. Infants are generally known to consume very little milk or formula on the first day of life, with
ingestion quickly increasing over time. However the first time-point for which ingestion data are
available for infants is at 7 days, which EPA extrapolated back to 0 ingestion at time=0 in its
analysis (birth; Figure 1). The breast-milk ingestion rate was estimated based on mean values
reported in Arcus-Arth et al. (2005), while the water ingestion used for the breast-feeding mother
was the 90th percentile value from U.S. EPA (2004).
Is EPA's extrapolation and rationale transparently and objectively described? Does this function
appropriately characterize the available data and information? Are there other data that could be
used to obtain a better or equally valid alternative estimate of mean breast-milk ingestion rate for
infants in the first few days of life? Should an estimate other than the mean be used to determine the
breast-milk ingestion rate?
The approach and methods are objectively described. However, this is one of the weakest
portions of EPA's analysis. The modeling of perchlorate and iodine kinetics in the neonate is
highly uncertain. EPA needs to recognize this and make this clear to the reader. If the
purpose of this analysis is to determine the relative difference in inhibition of iodide uptake by
the thyroid for the various scenarios (e.g., normal adult, child, nursing infant, fetus, and
bottle-fed infant), then the mean on all exposure and pharmacokinetic parameters should be
used and used consistently throughout the analysis. Otherwise, the current approach exhibits
bias for one scenario over the other.
(C) Water ingestion
C-l. For pregnancy, EPA used a normalized (90th percentile) water ingestion rate of 33 mL/kg-day
(U.S.EPA 2004), which was multiplied by the maternal BW as described by the PBPK model
growth-functions during pregnancy to obtain total water ingestion for the mother.
Is EPA's approach and rationale transparently and objectively described? Is this approach
appropriate for characterizing the upper-bound for ingestion of pregnant women? Are there other
approaches or data that could be used to obtain a better or equally valid alternative estimate for this
parameter? (Note that the self-reported body weight values in U.S. EPA (2004) indicate the same
average body weight for pregnant women as for non-pregnant; data appears inaccurate, so the PBPK
model body-weight description was used instead.)
Throughout this analysis, EPA was inconsistent in choosing upper bounds or means for
various parameters. As such, there is no clear understanding of the objectives of EPA's
A-41
-------
Sean Hays
analysis. If the purpose of this analysis is to determine the relative difference in inhibition of
iodide uptake by the thyroid for the various scenarios (e.g., normal adult, child, nursing
infant, fetus, and bottle-fed infant), then the mean on all exposure and pharmacokinetic
parameters should be used and used consistently throughout the analysis. Otherwise, the
current approach exhibits bias for one scenario over the other.
C-2. For lactation, EPA used a fixed total (90th percentile) water ingestion rate of 2959 mL/day (U.S. EPA
2004), with the rationale that while the woman's weight and self-water demand are expected to drop
after pregnancy (as described by the PBPK model time-dependent weight equations), the demand
for milk production would be increasing, and the reported value was not for a specific age-range of
child.
Is EPA's approach and rationale transparently and objectively described? Is this an appropriate
value to use for the ingestion rate of lactating women? Are there other better or equally valid
alternative approaches or values that could be used?
Throughout this analysis, EPA was inconsistent in choosing upper bounds or means for
various parameters. As such, there is no clear understanding of the objectives of EPA's
analysis. If the purpose of this analysis is to determine the relative difference in inhibition of
iodide uptake by the thyroid for the various scenarios (e.g., normal adult, child, nursing
infant, fetus, and bottle-fed infant), then the mean on all exposure and pharmacokinetic
parameters should be used and used consistently throughout the analysis. Otherwise, the
current approach exhibits bias for one scenario over the other.
C-3. For bottle-fed infants, EPA made extrapolations of the 90th percentile water-ingestion in early life
based on measurements made for the age ranges: 0-30 days, 1-3 months, and 6-12 months (points in
Figure 2, upper panel). For the purposes of this analysis, bottle-fed implies feeding with formula
requiring the addition of water.
Is EPA's extrapolation and rationale transparently and objectively described? Does the overall
function used to represent the changes in ingestion with age (and body weight) appropriately
characterize the available data and information? Are there other data that could be used to obtain a
better or equally valid alternative estimate of water ingestion rate for infants in the first few days of
life (e.g., 7-day old)? The water ingestion rates used by EPA are based on 90th percentile ingestion
data and thus are likely to exceed (minimal) physiological needs of infants as defined in nutritional
guidelines. Are the water ingestion rates used by EPA reasonable in comparison to the
physiological needs of infants at these various life stages? Are there other approaches or data (e.g.,
see the FDA memo) that could be used to obtain a better or equally valid alternative estimate for
this parameter?
Same response as last question.
B-42
-------
(D) Perchlorate concentration in formula
D-l. EPA used 1.42 j^ig/L as the concentration of perchlorate in formula for bottle-fed infants. This
estimate was based on information from FDA's Total Diet Study, supported by Pearce el al.'s
(2007) findings.
Is EPA's approach and rationale transparently and objectively described? Is 1.42 (ig/L an
appropriate value to use for the concentration of perchlorate in infant formula? Are there other
better or equally valid alternative approaches or values that could be used?
Based on the stated objectives of this analysis, EPA should adjust the intake of perchlorate
from infant formula to result in a daily exposure consistent with the point of departure. This
will yield consistent results across all scenarios to assure a fair and impartial comparison of
relative differences in inhibition of thyroid iodine uptake can be made.
(E) Radioiodide excretion into breast-milk by NIS
E-l. In the model, EPA included perchlorate inhibition of NIS radioiodide excretion into breast-milk, as
well as inhibition of radioiodide transport by perchlorate for all NIS-containing tissues, thereby
making the code consistent with the model description in Clewell et al. (2007). Is this inclusion
appropriate? Are the impacts of this inclusion transparently and objectively described?
Based on the data from Pearce et al. (2007), one would expect there to be no effect of
perchlorate on the excretion of iodine in breast milk. As such, this feature of the model that
EPA has included may not be accurate with perchlorate kinetics. EPA should investigate the
data of Pearce et al. more fully and explore other data sets to see what evidence is available to
include such a feature in the model.
Additional comments:
The approach of using PBPK modeling is admirable and the EPA should be commended. However,
the EPA should also have considered easier and more straightforward approaches. The one-
compartment PK model developed by EPA (Lorber, 2008), paired with measured perhclorate and
iodine levels in breast milk and infant formula would have provided simpler and equally valid
approaches for answering the question of the relative difference in steady-state perchlorate levels
(this is ultimately the endpoint of interest) in the various receptors/scenarios. While I agree with
using PBPK modeling, I also think EPA should think about simpler approaches that are equally or
more valid, and sometimes much simpler and more easily embraced by the regulatory and risk
assessment community.
A-43
-------
Sean Hays
Appendix A: I agree with the model modifications made by EPA.
Appendix B: Appendix B is well written and easier to follow than the corresponding text in the
main report relating to urinary clearance values (section 3.1).
B-44
-------
PEER REVIEW COMMENTS FROM
Frederick Kaskel, M.D., Ph.D.
Assistant Professor
Albert Einstein College of Medicine ofYeshiva University
Section of Pediatric Nephrology
111 East 21 Oth Street
Bronx, New York 10467
718 655-1120
Email: fkaskel@montefiore.org
A-45
-------
-------
Frederick Kaskel
General Charge Questions:
G-l. Physiologically based pharmacokinetic (PBPK) models were modified to predict inhibition of the
sodium-iodide symporter (NIS)47 for pregnant and lactating women, nursing infants, and for the
subsequent stages of childhood. The published models were modified by EPA to fix errors and
incorporate new data, particularly data on lifestage variability in the urinary clearance of pechlorate, to
which NIS inhibition is sensitive. The models are suitable to provide quantitative predictions to the
Agency on the lifestyle variability of perchlorate NIS inhibition of thyroidal iodide uptake. EPA's
analysis is logical, clear and appropriate in depth and length and EPA has accurately, clearly and
objectively represented and synthesized the scientific evidence for the changes made to or specification of
the model code and input parameters.
G-2. Additional studies that should be considered in the assessment of the specific parameters such as
urinary clearance of iodide and/or perchlorate and ingestion rates (breast milk, formula and water) in
neonates and these include data on maturation of tubular transport rates. On page 39 of the report, the
issue of the role of pendrin transporter for iodide during development is addressed. One cannot assume
that perchlorate and iodide are handled similarly by the developing kidney based on their similar charge
and diameter; more data is needed in the investigation of tubular maturation of the transporters that
regulate the clearance of iodide and perchlorate. This is also addressed on page 40, second paragraph in
the report where it is stated that one cannot assume that the relative clearance for iodide and perchlorate
should be constant across all ages and life stages. Additionally on page 42 the EPA states that there is no
data on renal transporters during infancy to suggest the level and pattern of expression changes required
to change clearance/GFR. Thus, the report used DeWoskin and Thompson's published data for scaling of
renal excretion for infants by body weight and on page 44 the EPA extended its extrapolation to a 60-day-
old, 5 kg child is sound. These assumptions are reasonable but indicate the importance of additional
investigations in newborn models and in humans.
G-3. There are no other parameters or model choices described in the document that are incorrect or
require further explanation or provide better estimates.
G-4. Newer estimates of renal function have been provided by Schwartz which should be evaluated
G-5. EPA accurately characterized the strengths and limitations of the analysis. However, as indicated in
the report, additional information on the possible effects of maturation of glomerular filtration, tubular
A-47
-------
Frederick Kaskel
reabsorption and secretion, and changes in body composition during the neonatal period is indicated in
order to more confidently apply EPA's estimates of RAIU for different life stages.
G-6. The analysis is transparent in terms of the steps, logic, key assumptions, limitations, and decision.
The characterization of the results of EPA's work fully explains: a) the analysis approach employed; b)
the use of assumptions and their impact on the analysis; c) the use of extrapolations and their impact on
the analysis; d) plausible alternatives and the choices made among those alternatives; e) the impacts of
one choice vs. another on the analysis; f) significant data gaps and their implications for the analysis; g)
the scientific conclusions identified separately from default assumptions and policy decisions, and h) the
major conclusions and the discussions of EPA's confidence and uncertainties in the conclusions.
Parameter-Specific Charge Questions:
(A) Urinary clearance
A-l. The input values selected for maternal urinary clearance during pregnancy and lactation used in
EPA's analysis are appropriate, transparent and objectively described. The choice of three alternatives for
pregnancy is a rational compromise in lieu of the lack of additional human data. The use of a lower
clearance is a safe assumption. On page 10 urinary clearance values for perchlorate and iodide across all
lifestages were determined to be sensitive parameters for prediction of NIS thyroidal iodide uptake
inhibition by perchlorate. EPA determined that urinary clearance of perchlorate and iodide in neonates is
slower than is indicated by scaling based on body weight. Urinary elimination of a number of compounds
including drugs and drug metabolites also indicate that renal clearance is slower per unit of body weight
in neonates. Modification of the PBPK models to describe slower clearance of perchlorate and iodide in
neonates resulted in an increase in predicted levels of NIS inhibition in infants.
The values selected for urinary clearance for infants and older children are the best estimates for the
available data. The interpretation of the data that suggested an increase in predicted levels of NIS
inhibition in infants at a perchlorate dose-rate of 7 ug/kg-day is a safe assumption. The indices of renal
function are based on the literature which indicated that the GFR increases steadily postnatally but does
not reach adult values until approximately 2 years of age. I know of no other data that would provide
better guidance or estimates and it is unlikely that there are other factors than the urinary clearance in the
elimination of perchlorate for infants. However, one should consider that tubular function in this age
group is not fully matured and possible developmental changes in transport activity might be important
but no data is available for perchlorate elimination during development. On page 11 EPA chose to
B-48
-------
Frederick Kaskel
estimate perchlorate induced inhibition using scaling of urinary cleareance proportional to body weight
for children at 1 year of age and older which results in somewhat higher estimates of iodide uptake
inhibition than reported by Clewell although still slightly less than predicted for the average adult exposed
at the same dose. EPA's estimates of urinary clearance in infants and children are lower than those used
in Clewell but reflects published GFR values.
(B) Breast-milk ingestion
B-l. EPA's extrapolation and rationale are transparent and objectively described to assess the breast milk
ingestion rate.
(C) Water ingestion
C-l EPA's approach and rationale for pregnancy water ingestion rate is transparent and objectively
described, and is based on the available literature. I know of no additional data that could be used to
obtain a better estimate of mean breast-milk ingestion rate for infants in the first few days of life. The
mean estimate is fine.
C-2 EPA's approach and rationale for lactation water ingestion rate is appropriate and transparent and
objectively described. I know of no other approaches.
C-3 EPA's extrapolation and rationale for bottle fed infant's water ingestion in early life is transparent
and objectively described. The overall function used to represent the changes in ingestion with age (and
body weight) appropriately characterize the available data. The water ingestion rate for infants used by
the EPA are reasonable in comparison to the physiological needs of infants. The estimated 90% water
intake rate used by EPA in PBPK model stimulations is appropriate.
(D) Perchlorate concentrations in formula
D-l. EPA's approach and rationale for the concentration of perchlorate in formula for bottle fed infants is
appropriated and transparency.
(E) Radioiodide excretion in breast milk by NIS
(E-l) The inclusion of perchlorate inhibition of NIS radioiodide excretion into breast-milk, as well as
inhibition of radioiodide transport by perchlorate for all NIS-containing tissues, is appropriate, transparent
and objectively described. The EPA added inhibition of radioiodide transport by perchlorate for
A-49
-------
Frederick Kaskel
radioiodide excretion into breast milk by NIS markedly increased the predicted percent inhibition of
thyroidal radioiodide uptake in the breast fed infant.
B-50
-------
Kannan Krishnan
PEER REVIEW COMMENTS FROM
Kannan Krishnan, Ph.D.
Associate Professor
Department of Occupational and Environmental Health
Faculty of Medicine, Universite de Montreal
2375 chemin de la Cote Ste-Catherine
Montreal, QC, H3T 1A8 Canada
514-343-6581
Email: kannan.krishnan@umontreal.ca
A-51
-------
-------
Kannan Krishnan
Dr. Kannan Krishnan's Review of the Document entitled
"INHIBITION OF THE SODIUM-IODIDE SYMPORTER BY PERCHLORATE:
AN EVALUATION OF LIFESTAGE SENSITIVITY USING PHYSIOLOGICALLY-BASED
PHARMACOKINETIC (PBPK) MODELING"
Overview:
This EPA report summarizes work conducted to evaluate the PBPK models for perchlorate and
radioiodide for quantitating relative sensitivity of different subgroups (lifestages). The two-stage model
evaluation process involved verification of model codes and examination of the parameterization
approaches. Following the revision of the PBPK models by EPA, they were checked by a contractor who
also verified the output of the model by reproducing various figures from original publications. Despite
the thoroughness of the work, this life-stage variability analysis (either due to lack of data or due to
uncertainty associated with available data) did not account for certain subgroups (e.g., elderly, foetus
during early gestation periods, iodine-deficient or hypothyroid status during pregnancy) and did not
account for variability of parameter values within subgroups in the simulations (i.e., with the use of
Bayesian or Monte Carlo type methods).
General Charge Questions:
G-l. Is EPA's analysis logical, clear and appropriate in depth and length? Has EPA accurately, clearly
and objectively represented and synthesized the scientific evidence for the changes made to or
specification of the model code and input parameters?
RE:
• The EPA analysis of perchlorate-mediated inhibition of the NIS in humans is based on Merrill et
al. (2005) and Clewell et al. (2007) PBPK models, and specifically addresses the variability of
NIS inhibition as a function of lifestage. The document is clear and concise. The depth and
legnth of presentation are appropriate, given the objective.
• The structure of the PBPK models published by Merrill/Clewell has not been altered; rather some
of the input parameters as well as equations have been modified either to correct an error or to
reflect current state of knowledge more appropriately.
A-53
-------
Kannan Krishnan
G-2. Please identify any additional studies or other data sources that should be considered in the
assessment of the specific parameters addressed below, including urinary clearance (of iodide and/or
perchlorate) and ingestion rates (breast milk, formula and water), especially in neonates.
RE:
This reviewer is not aware of any studies in neonates that would provide better estimates of urinary
clearance of perchlorate and iodide. Even though isolated studies reporting ingestion rates (breast
milk, formula and water) in infants in other parts of the world could be obtained from the literature,
such studies probably would only introduce further uncertainty. However, the study of Kirk et al.
(2005). Perchlorate and iodine in dairy and breast milk. Environ Sci technol 39: 2011-17 may used
to corroborate the findings of the present study - as it relates to the relationships between drinking
water concentration and breast milk concentration.
G-3. Besides the specific parameters identified below, please comment on any other parameter or model
choices described in the document that you think are incorrect or require further explanation or
where other available information could provide better estimates.
RE:
The parameters of this model consist of:
• Physiological parameters
• Intake/contact rates
• Partition coefficients
• Permeability-area cross product
• Urinary clearance
• Binding parameters
• Maximal velocity and affinity constants
The parameter values found in the original reports and refined following EPA's evaluation would
appear to be supported by available literature. However, focused data collection might facilitate the
improvement of the partition coefficient values used in the model as well as the urinary clearance
values for perchlorate and iodide in the various lifestages.
B-54
-------
Kannan Krishnan
G-4. Please discuss research that you think would be likely to increase confidence in these models and
their use in predicting RAIU inhibition by perchlorate in different life stages (for an average
individual within each life-stage).
RE:
• In the iodide/perchlorate models, the chemical concentration entering the tissues corresponds to
the arterial PLASMA concentration whereas the flow rate to tissues corresponds to BLOOD
(RBC + Plasma) FLOW rates. Either the influx in all mass balance equations should correspond
to whole blood concentration or the flow rate should correspond to plasma flows - since the
RBC:plasma partition coefficient (PRBC_p) is not always equal to 1 (see for example lines 112
on page 38, or line 115 on page 4 of the EPA model code file), and the chemical movement
between plasma and RBC is diffusion-limited and no flow-limited. The consequence of this
modeling assumption may be verified to ensure confidence in the use of these models. For
example, if the simulations indicate that the concentration profile of perchlorate is identical in
RBC and plasma compartments, qualitatively and quantitatively, then the above observation has
no consequence.
• Further, consideration should be given to the possibility of being able to simulate iodine-deficient
(or hypothyroid) situation in pregnant women by modulating specific parameters of the model.
• Both the response to the previous question and the comments under general overview are all
applicable here.
G-5. Does EPA accurately characterize the strengths and limitations of the analysis? Please comment on
any particular strengths that may not be mentioned or adequately characterized for the estimates of
RAIU for different life stages. Please also comment on any specific sources of uncertainty that you
believe have been overlooked in EPA's analysis or that require further discussion, and which might
be significant to EPA's estimates of RAIU for different life stages.
RE:
• The strength relates to the use of PBPK model to assess the lifestage sensitivity to inhibition of
the sodium-iodide symporter by perchlorate; the use of fetus as a subgroup to evaluate the
relative sensitivity to adults; consideration of relevant route/source of exposure (drinking water);
• The weaknesses are related to the fact that the analysis did not include certain subgroups (e.g.,
elderly, foetus during early gestation periods, iodine-deficient or hypothyroid status during
pregnancy) and did not address variability of parameter values within subgroups in the
A-55
-------
Kannan Krishnan
simulations (i.e., with the use of Bayesian or Monte Carlo type methods).
G-6. As recommended in EPA's 2000 Risk Characterization Handbook, is the analysis transparent in
terms of the steps, logic, key assumptions, limitations, and decisions? Specifically, does
characterization of the results of EPA's work fully explain: a) the analysis approach employed; b)
the use of assumptions and their impact on the analysis; c) the use of extrapolations and their impact
on the analysis; d) plausible alternatives and the choices made among those alternatives; e) the
impacts of one choice v.v. another on the analysis; f) significant data gaps and their implications for
the analysis; g) the scientific conclusions identified separately from default assumptions and policy
decisions, if any; and h) the major conclusions, and the discussions of EPA's confidence and
uncertainties in the conclusions?
The analysis is transparent and the assumptions as well as alternative approaches are generally described
in sufficient detail. The conclusions are essentially scientific in nature, based on data obtained from
PBPK model simulations. The following improvements are suggested:
1. The reason for limiting the present analysis to eight sub-groups (i.e., pregnant woman, fetus, lactating
woman, breast-fed infant, bottle-fed infant, 1 year old and 2 year old child, "average" adult, and non-
pregnant woman of child-bearing age) may be specified at the outset. In this regard, it may be useful to
clarify as to why the elderly and teens were not part of the sub-groups analysed in this study.
2. Clarify as to why the results of this analysis are also applicable to chronic exposure exposure situations
(compared to typically acute (short-term) simulations)
3. The justification of the choice of 24-hr RAIU as the endpoint should be included. Was 24-hr AUC
considered as an alternative measure ? What was the scientific basis for basing the analysis on a single
RAIU value in infants and adults obtained at one specific time point (i.e,. 24 hr). Some
consideration/discussion of the sensitivity of that time point to the key input parameters as a function of
age might be useful.
B-56
-------
Kannan Krishnan
Parameter-Specific Charge Questions:
(A) Urinary clearance
A-l. Please comment on the appropriateness of the input values selected for maternal urinary
clearance during pregnancy and lactation in EPA's analysis. Are the available data and rationale for
the values selected transparently and objectively described? Are you aware of other publications or
data that could be used to guide these choices or which provide alternative input values that are
equally valid or more appropriate? Likewise, are the values selected for urinary clearance for the
infant and older child the best estimates, given the available science and data? Are there other data
that would provide better (or equally valid alternative) guidance or estimates? Is there any reason to
believe that urinary clearance might not be a limiting factor in the elimination of perchlorate for
infants?
RE:
The inadequacy of use of urinary clearance values in various human life stages based on
(i) the pregnant:nonpregnant values in rats, and
(ii) the scaling of renal function for neonates on the basis of BW°75
- are well justified by EPA. The outcome is consistent with available experimental and/or physiological
data. The selection of lower clearance value for pregnancy as well as the option 2 for lactating women,
though not the optimal (given the interindivudal variability), would appear to be pragmatic and consistent
with the rationale provided by EPA. However this reviewer has the following additional observations:
• The R2 value for the fit described in Figure B-6 is poor raising concern about the adequacy of the
equation
• Did EPA analyze the data in Figure B-5 on the basis of body surface data for the various age
groups (of pregnant women)?
• On page 42, para 3, Figure B-l should read Figure B-2?
• What does GFR-based scaling mean in Figure B2? Is it body surface scaled?
A-57
-------
Kannan Krishnan
(B) Breast-milk ingestion
B-l. Infants are generally known to consume very little milk or formula on the first day of life, with
ingestion quickly increasing over time. However the first time-point for which ingestion data are
available for infants is at 7 days, which EPA extrapolated back to 0 ingestion at time=0 in its
analysis (birth; Figure 1). The breast-milk ingestion rate was estimated based on mean values
reported in Arcus-Arth et al. (2005), while the water ingestion used for the breast-feeding mother
was the 90th percentile value from U.S. EPA (2004).
Is EPA's extrapolation and rationale transparently and objectively described? Does this function
appropriately characterize the available data and information? Are there other data that could be
used to obtain a better or equally valid alternative estimate of mean breast-milk ingestion rate for
infants in the first few days of life? Should an estimate other than the mean be used to determine the
breast-milk ingestion rate?
RE:
• The fitting of the available data from Kahn and Stralka, with a mathematical function is adequate.
However, the extrapolation from day 7 towards day 0 (or at birth) is not warrented given that the
newborn is not a sub-group used in the assessment of relative sensitivity of lifestages (Table 3,
page 22).
• The motivation for choosing 90th pctle for consumption rates needs to be clearly presented, since
the expectation is a calculation either based on mean values in the various groups or 95th pctle
values. Therefore, the rationale and scientific basis for the choice and use of the 90th pctle values
in these calculations should be more clearly presented.
(C) Water ingestion
C-l. For pregnancy, EPA used a normalized (90th percentile) water ingestion rate of 33 mL/kg-day
(U.S.EPA 2004), which was multiplied by the maternal BW as described by the PBPK model
growth-functions during pregnancy to obtain total water ingestion for the mother.
Is EPA's approach and rationale transparently and objectively described? Is this approach
appropriate for characterizing the upper-bound for ingestion of pregnant women? Are there other
approaches or data that could be used to obtain a better or equally valid alternative estimate for this
parameter? (Note that the self-reported body weight values in U.S. EPA (2004) indicate the same
average body weight for pregnant women as for non-pregnant; data appears inaccurate, so the PBPK
model body-weight description was used instead.)
RE:
• The cited value of 33 ml/kg-day corresponds to the 90th percentile value for pregnant women
("consumers only") of the direct and indirect community water ingestion (chapter 6, page 16,
B-58
-------
Kannan Krishnan
U.S. EPA 2004). However, the 90th percentile value of total water ingestion for the same group
was 39 ml/kg-day. It is unclear then as to why EPA specifes the use of total water ingestion rate
but actually uses the value corresponding to another group (i.e,. community water ingestion).
Furthermore, the 90th percentile value for pregnant women, reported in US EPA (2004), was
associated with a small sample size (n=65, which does not meet the minimum reporting
requirements described in the "Third report on Nutrition Monitoring in United States"). This
raises the question of why not use (or justify the non-use of) the value from Ershow et al (1991)
based on much larger sample size (n=188). These authors reported 90th pctle values for tap and
total water ingestion of 34.5 and 48.9 ml/kg-day respectively.
• It is also unclear to this reviewer as to why 90th pctle value is chosen for the computations and not
either the median or the 95th percentile value.
• This reviewer is not concerned about the use of subject-specific or group-specific body weight in
the PBPK model to facilitate the calculations for pregnant women, as long as the ingestion rate is
expressed in units of ml/kg-day, as done here.
C-2. For lactation, EPA used a fixed total (90th percentile) water ingestion rate of 2959 mL/day (U.S. EPA
2004), with the rationale that while the woman's weight and self-water demand are expected to drop
after pregnancy (as described by the PBPK model time-dependent weight equations), the demand
for milk production would be increasing, and the reported value was not for a specific age-range of
child.
Is EPA's approach and rationale transparently and objectively described? Is this an appropriate
value to use for the ingestion rate of lactating women? Are there other better or equally valid
alternative approaches or values that could be used?
RE:
• The ingestion rate of 2959 ml/day, used by EPA, corresponds to the 90th percentile value of
"consumers only" lactating women for direct and indirect community water ingestion. In
comparison, the 90th percentile value of total water ingestion in consumers only lactating women
is reported to be 3021 ml/ day (chapter 6 page 17). The EPA report (page 24, para 2) states that
the intent was to use the "total" consumers-only water intake in the calculations. The source and
consequence of this discrepancy should be addressed.
• Further, U.S. EPA (2004) indicated that the 90th pctl value (2959 ml/day) is associated with a
small sample size (n=41, which does not meet the minimum reporting requirements described in
A-59
-------
Kannan Krishnan
the "Third report on Nutrition Monitoring in United States"), raising a concern of its use rather
than the value from Ershow et al. (1991). Additionally, it is unclear as to why the 90th pctl rather
than 95th pctl of the water ingestion is used in these calculations.
• In light of the fact that the water ingestion rate in lactating women is significantly greater (see
chapter 6 pages 16-17, U.S. EPA 2004), on a ml/kg-day basis, than in pregnant women, the
rationale used for using a fixed ingestion rate needs to be more fully articulated.
C-3. For bottle-fed infants, EPA made extrapolations of the 90th percentile water-ingestion in early life
based on measurements made for the age ranges: 0-30 days, 1-3 months, and 6-12 months (points in
Figure 2, upper panel). For the purposes of this analysis, bottle-fed implies feeding with formula
requiring the addition of water.
Is EPA's extrapolation and rationale transparently and objectively described? Does the overall
function used to represent the changes in ingestion with age (and body weight) appropriately
characterize the available data and information? Are there other data that could be used to obtain a
better or equally valid alternative estimate of water ingestion rate for infants in the first few days of
life (e.g., 7-day old)? The water ingestion rates used by EPA are based on 90th percentile ingestion
data and thus are likely to exceed (minimal) physiological needs of infants as defined in nutritional
guidelines. Are the water ingestion rates used by EPA reasonable in comparison to the
physiological needs of infants at these various life stages? Are there other approaches or data (e.g.,
see the FDA memo) that could be used to obtain a better or equally valid alternative estimate for
this parameter?
• Even though the EPA rationale is satisfactory, it is unclear as to why the emphasis is placed on
the section of the curve (i.e., first few days after birth) which is neither used in the lifestage
analysis nor supported by any data.
• A report published by a Public Health Agency in Quebec contains data on water consumption of
393 infants of 8 weeks of age. For bottle-fed only infants (n = 278), mean (IC95%) value for total
water ingestion was 122 ± 43 (117-127) ml/kg-day, or 655 ± 233 (627-682) ml/day. The
corresponding 90th percentile values were 179 ml/kg-day and 981 ml/day. For more details the
following source may be consulted:
• http://www.inspq.qc. ca/publications/default.asp?NumPublication=334
• A copy of the above report in PDF is also attached herewith.
B-60
-------
Kannan Krishnan
(D) Perchlorate concentration in formula
D-l. EPA used 1.42 (ig/L as the concentration of perchlorate in formula for bottle-fed infants. This
estimate was based on information from FDA's Total Diet Study, supported by Pearce el al.'s
(2007) findings.
Is EPA's approach and rationale transparently and objectively described? Is 1.42 (ig/L an
appropriate value to use for the concentration of perchlorate in infant formula? Are there other
better or equally valid alternative approaches or values that could be used?
RE:
The EPA,s approach is clearly described and appears to be consistent with the current state of knowledge.
However, it would be better to clearly identify the basis for the choice of the mean value rather than
median, 90th or 95th pctl value (presumably the limited, available data did not permit such a
determination).
(E) Radioiodide excretion into breast-milk by NIS
E-l. In the model, EPA included perchlorate inhibition of NIS radioiodide excretion into breast-milk, as
well as inhibition of radioiodide transport by perchlorate for all NIS-containing tissues, thereby
making the code consistent with the model description in Clewell et al. (2007). Is this inclusion
appropriate? Are the impacts of this inclusion transparently and objectively described?
RE:
Yes. The EPA's approach is logical and internally-consistent. The impact of this inclusion is described
in sufficient detail.
A-61
-------
-------
PEER REVIEW COMMENTS FROM
Chensheng Lu, Ph.D.
Assistant Professor
Department of Environmental Protection
School of Public Health
Harvard University
665 Huntington Avenue
Landmark Center Room 404N
Boston, MA 02115
617-988-8811
Email: cslu@hsph.harvard.edu
A-63
-------
-------
Chensheng Lu
Technical Charge to External Peer Reviewers
Contract No. EP-C-07-024
Task Order No. 54
October 22, 2008
INHIBITION OF THE SODIUM-IODIDE SYMPORTER BY PERCHLORATE:
AN EVALUATION OF LIFESTAGE SENSITIVITY USING
PHYSIOLOGICALLY-BASED PHARMACOKINETIC (PBPK) MODELING
WRITTEN COMMENTS ARE DUE NO LATER THAN MONDAY, NOVEMBER 10, 2008
BACKGROUND
The U.S. Environmental Protection Agency (EPA) is seeking an external peer review of its analysis of the
inhibition of thyroidal radioiodide uptake (RAIU) by perchlorate for multiple life-stages including
pregnancy, fetal development, lactation, infancy, childhood, and adulthood. This analysis involves
modifications to the computational implementation (code) for a set of existing PBPK models that describe
the kinetics of radioiodide and perchlorate in humans during these different life-stages. Since the models
themselves are published in the peer-reviewed literature, EPA is not seeking review of the models per se,
but of changes in a small number of parameters, and of specific changes in the code intended to make the
code consistent with the published description. EPA's analysis and modifications, including identification
of model code errors and examination of data and assumptions used for specific input parameters, are
described in the attached document.
REVIEW MATERIALS AND INSTRUCTIONS
Review Document
EPA's Draft Report, Inhibition of the Sodium-Iodide Symporter by Perchlorate: An Evaluation of
Lifestage Sensitivity Using Physiologically-based Pharmacokinetic (PBPK) Modeling (Including a report
of a contractor-led detailed review and quality assurance check of the model code in Appendix C)
Background Materials Provided on CD
1. Model Code in PDF and zipped acslXtreme workspace files
2. FDA memo on "Volume of Feeds for Infants" (for Parameter-Specific Charge Question C-3)
3. EPA's 2000 Risk Characterization Handbook (for General Charge Question G-6)
4. Supporting References
Peer review of this analysis is being sought to ensure that EPA's analysis, modifications to existing PBPK
models, and data inputs and assumptions are clearly and transparently described and are scientifically
sound and supported by the available data. Please provide detailed explanations of your responses to the
charge questions.
A-65
-------
Chensheng Lu
CHARGE QUESTIONS
General Charge Questions:
G-l. Is EPA's analysis logical, clear and appropriate in depth and length? Has EPA accurately, clearly
and objectively represented and synthesized the scientific evidence for the changes made to or
specification of the model code and input parameters?
The Reviewer is convinced that EPA has performed an outstanding job in improving the PBPK model so
the codes written in the model are consistent to the physiology of iodide uptake in the thyroid glands
and the uptake inhibition by perchlorate. It is also clear that EPA has tried to perfect the input
parameters to increase the predictability of the model.
G-2. Please identify any additional studies or other data sources that should be considered in the
assessment of the specific parameters addressed below, including urinary clearance (of iodide and/or
perchlorate) and ingestion rates (breast milk, formula and water), especially in neonates.
Considering the elevated inhibition of RAIU in bottle-fed infants, EPA should seek for additional data to
re-affirm the water ingestion rates that are used by EPA, particularly the use of the 90th percentile
values in which the situations exceed the expectation of the fundamental knowledge. The Reviewer
has no knowledge of whether there are studies or data sources that EPA could use.
G-3. Besides the specific parameters identified below, please comment on any other parameter or model
choices described in the document that you think are incorrect or require further explanation or
where other available information could provide better estimates.
EPA should explain the rationale of using the 90th percentile values in the analysis. It seems to the
Reviewer that such choice is deemed to create an upper bound limit, however, throughout the
document, EPA has stated that this is not the purpose due to the uncertainties involved in the model
simulation and other reasons.
G-4. Please discuss research that you think would be likely to increase confidence in these models and
their use in predicting RAIU inhibition by perchlorate in different life stages (for an average
individual within each life-stage).
What will significantly increase the confidence in these PBPK model is to use the real-world data (such as
perchlorate in drinking water and the level of iodide in blood or thyroidal functions in population)
B-66
-------
Chensheng Lu
liking perchlorate exposure and iodide inhibition. The article published by Blount et al. (EHP 2006
114(12) 1865-1871) would be an ideal application for these PBPK models. Unfortunately, data used
in Blount et al. study (NHANES) do not include children ages 6 and below.
G-5. Does EPA accurately characterize the strengths and limitations of the analysis? Please comment on
any particular strengths that may not be mentioned or adequately characterized for the estimates of
RAIU for different life stages. Please also comment on any specific sources of uncertainty that you
believe have been overlooked in EPA's analysis or that require further discussion, and which might
be significant to EPA's estimates of RAIU for different life stages.
This document is well written and has highlighted what EPA has accomplished in assessing RAIU
inhibition at different life stages resulting from perchlorate exposure. The Reviewer thought EPA
has thoroughly discussed the strengths and limitation of this analysis, including the uncertainty
analysis.
One uncertainty, however, has not been addressed by EPA is the use of direct IV dose of radioiodide to
the bottle-fed infants in order to determining iodide uptake inhibition caused by perchlorate in
formula. Although this approach seems intuitive, it may not reflect the real-world scenario in which
iodide intake is usually taking place by oral ingestion. Pharmacokinetically speaking, the absorption
of chemicals in humans could vary significantly between oral ingestion and bolus iv injection. EPA
needs to conduct an uncertainty analysis to assure that such approach would not impact the
outcomes significantly.
G-6. As recommended in EPA's 2000 Risk Characterization Handbook, is the analysis transparent in
terms of the steps, logic, key assumptions, limitations, and decisions? Specifically, does
characterization of the results of EPA's work fully explain: a) the analysis approach employed; b)
the use of assumptions and their impact on the analysis; c) the use of extrapolations and their impact
on the analysis; d) plausible alternatives and the choices made among those alternatives; e) the
impacts of one choice v.v. another on the analysis; f) significant data gaps and their implications for
the analysis; g) the scientific conclusions identified separately from default assumptions and policy
decisions, if any; and h) the major conclusions, and the discussions of EPA's confidence and
uncertainties in the conclusions?
EPA has clearly explained their approaches employed including the assumptions, alternatives, and the use
of extrapolations and their impacts on the analyses. Apparently, there are significant data gaps,
particularly for newborn infants that lead to some limitations of using this revised PBPK model.
However, it is rather common for many PBPK modeling work, and therefore should NOT be
considered a major limitation of this analysis.
It is apparent that the outcome of the PBPK model prediction is dictated by the use of urinary clearance of
A-67
-------
Chensheng Lu
perchlorate and iodide. Other parameters have somewhat less impacts on the results. EPA has taken
the right approach focusing on the parameters related to perchlorate exposure and iodide intakes.
The revised PBPK model that EPA modified has demonstrated the importance of those parameters,
and the Reviewer agrees with the EPA's scientific conclusion in which the modified Clewell et al.
model is acceptable to calculate the lifestage differences in the degree of NIS inhibition of thyroidal
radioiodide uptake at a given level of perchlorate exposure.
Parameter-Specific Charge Questions:
(A) Urinary clearance
A-l. Please comment on the appropriateness of the input values selected for maternal urinary
clearance during pregnancy and lactation in EPA's analysis. Are the available data and rationale for
the values selected transparently and objectively described? Are you aware of other publications or
data that could be used to guide these choices or which provide alternative input values that are
equally valid or more appropriate? Likewise, are the values selected for urinary clearance for the
infant and older child the best estimates, given the available science and data? Are there other data
that would provide better (or equally valid alternative) guidance or estimates? Is there any reason to
believe that urinary clearance might not be a limiting factor in the elimination of perchlorate for
infants?
Considering that perchlorate, as well as iodide, does not further metabolize in human body, the urinary
clearance should be a limiting factor in removing perchlorate and iodide from humans at all
lifestages, and an important parameter in the perchlorate PBPK model.
Unfortunately, data for urinary clearance of perchlorate and iodide by the mother during pregnancy and
lactation are not consistent among three sources (Clewell, Aboul-Khair, and Delange) cited by EPA.
The choices that EPA made for selection clearance for pregnancy and lactation are quite arbitrary,
and the reasoning, if any, are not found. If GFR is corresponding to the cardiac output (meaning
higher blood flow rate equal to higher GFR), urinary clearance of any given compound during the
pregnancy should be higher than non-pregnancy. Urinary clearance during lactation period might be
the opposite to the pregnancy due to the difference of cardiac output. EPA should seek for
differences of urinary clearance (mainly via GFR) of compounds during pregnancy and lactation
outside the iodide and perchlorate literatures.
EPA has clearly documented how they determined the alternative scaling of urinary clearance of
B-68
-------
Chensheng Lu
perchlorate and iodide by body weight and has provided a thorough explanation of why EPA chose to
use (BW)1, instead of commonly used (BW)0 75, in neonates. The justification is sound and supported
by the data published in the literature. Similar justification of using (BW)1 scaling for perchlorate
clearance in older children (ages 2-12) is also provided, however, the sentence of "EPA's estimates of
urinary clearance in infants and children are lower than those used in Clewell et al. (2007), but are
values EPA judges to be scientific estimate, not bounds." (on page 11, 1st paragraph) is not clear to
the Reviewer. The information to support this sentence may come from Appendix B (pages 44-46),
particularly from Figure B-4. However, Figure B-4 itself is difficult to understand (for instance, how
is Lower 95% related to the yellow diamonds, and how the line of Data average is constructed?), and
therefore renders less convincing remark of using (BW)1 scaling for older children. EPA may want to
review this and provide a clearer explanation on the data presented in Figure B-4.
(B) Breast-milk ingestion
B-l. Infants are generally known to consume very little milk or formula on the first day of life, with
ingestion quickly increasing over time. However the first time-point for which ingestion data are
available for infants is at 7 days, which EPA extrapolated back to 0 ingestion at time=0 in its
analysis (birth; Figure 1). The breast-milk ingestion rate was estimated based on mean values
reported in Arcus-Arth et al. (2005), while the water ingestion used for the breast-feeding mother
was the 90th percentile value from U.S. EPA (2004).
Is EPA's extrapolation and rationale transparently and objectively described? Does this function
appropriately characterize the available data and information? Are there other data that could be
used to obtain a better or equally valid alternative estimate of mean breast-milk ingestion rate for
infants in the first few days of life? Should an estimate other than the mean be used to determine the
breast-milk ingestion rate?
Based on the data presented in Figure 1, the milk ingestion rates, or suckling rate, are quite different
between Clewell et al. and Arcus-Arth et al., however, EPA's decision to use Arcus-Arth's data
requires further clarification. EPA claimed that Clewell et al.'s data is inadequate to describe the
suckling rates in the first couple weeks of life, however, based on the Reviewer's examination on
Figure 1, the abrupt increase of milk ingestion during Day 1, and between Day 1 and 7 as presented by
Arcus-Arth et al. seems unlikely. The difficulty of collecting breast-milk ingestion rate for infants in
the first few days of life is understandable, and the deviation of the mean breast-milk ingestion from
the true value might not be as large as we thought. Therefore, the mean breast-milk ingestion rate
might be robust enough for use.
A-69
-------
Chensheng Lu
(C) Water ingestion
C-l. For pregnancy, EPA used a normalized (90th percentile) water ingestion rate of 33 mL/kg-day
(U.S.EPA 2004), which was multiplied by the maternal BW as described by the PBPK model
growth-functions during pregnancy to obtain total water ingestion for the mother.
Is EPA's approach and rationale transparently and objectively described? Is this approach
appropriate for characterizing the upper-bound for ingestion of pregnant women? Are there other
approaches or data that could be used to obtain a better or equally valid alternative estimate for this
parameter? (Note that the self-reported body weight values in U.S. EPA (2004) indicate the same
average body weight for pregnant women as for non-pregnant; data appears inaccurate, so the PBPK
model body-weight description was used instead.)
EPA's has objectively described its approach in using water ingestion rate of 33 mL/kg-day. However,
the rationale of using the 90th percentile value was not provided by the EPA in this analysis. As for
the BW estimates, it is unclear of how accurate it is to use the PBPK model growth-functions during
pregnancy for estimating BW of pregnant women. Will NHANES data provide some sort of
national average of the water ingestion rates stratified by lifestages and the BW of pregnant women?
Or could EPA validate the PBPK model growth-functions for weight estimates using the NHANES
data?
C-2. For lactation, EPA used a fixed total (90th percentile) water ingestion rate of 2959 mL/day (U.S. EPA
2004), with the rationale that while the woman's weight and self-water demand are expected to drop
after pregnancy (as described by the PBPK model time-dependent weight equations), the demand
for milk production would be increasing, and the reported value was not for a specific age-range of
child.
Is EPA's approach and rationale transparently and objectively described? Is this an appropriate
value to use for the ingestion rate of lactating women? Are there other better or equally valid
alternative approaches or values that could be used?
The rationale of using a fixed total water ingestion rate is justifiable and transparently and objectively
described. However, the reasoning of selection of 2,959 mL/day at the 90th percentile is missing in
this analysis. It would be assuring if EPA could provide the complete distribution of the estimates
of total water ingestion.
B-70
-------
Chensheng Lu
C-3. For bottle-fed infants, EPA made extrapolations of the 90th percentile water-ingestion in early life
based on measurements made for the age ranges: 0-30 days, 1-3 months, and 6-12 months (points in
Figure 2, upper panel). For the purposes of this analysis, bottle-fed implies feeding with formula
requiring the addition of water.
Is EPA's extrapolation and rationale transparently and objectively described? Does the overall
function used to represent the changes in ingestion with age (and body weight) appropriately
characterize the available data and information? Are there other data that could be used to obtain a
better or equally valid alternative estimate of water ingestion rate for infants in the first few days of
life (e.g., 7-day old)? The water ingestion rates used by EPA are based on 90th percentile ingestion
data and thus are likely to exceed (minimal) physiological needs of infants as defined in nutritional
guidelines. Are the water ingestion rates used by EPA reasonable in comparison to the
physiological needs of infants at these various life stages? Are there other approaches or data (e.g.,
see the FDA memo) that could be used to obtain a better or equally valid alternative estimate for
this parameter?
EPA has not informed the rationale of using the 90th percentile total water ingestion rate in early life
stage, as well as during the lactaton (as stated earlier in the review), and therefore, the possibility of
the estimated numbers are likely exceeding minimal physiological needs of infants raises a concern.
If this is the case in which the 90th percentile total water ingestion rate exceeds the norms, this
approach of using the 90th percentile is problematic. Since this sub-analysis focuses on bottle-fed
infants, EPA could follow the nutritional guidelines to estimate the total water-ingestion rate (such
as the frequency of feeding per 24 hours and the quantify of formula and water mixing per feeding).
(D) Perchlorate concentration in formula
D-l. EPA used 1.42 j^ig/L as the concentration of perchlorate in formula for bottle-fed infants. This
estimate was based on information from FDA's Total Diet Study, supported by Pearce el al.'s
(2007) findings.
Is EPA's approach and rationale transparently and objectively described? Is 1.42 (ig/L an
appropriate value to use for the concentration of perchlorate in infant formula? Are there other
better or equally valid alternative approaches or values that could be used?
The Reviewer believes that perchlorate level in formula used by EPA is the best available data, especially
this level is consistent to the public numbers from an independent research. The Reviewer is not
aware other better or equally valid alternative approaches or values that could be used.
A-71
-------
Chensheng Lu
(E) Radioiodide excretion into breast-milk by NIS
E-l. In the model, EPA included perchlorate inhibition of NIS radioiodide excretion into breast-milk, as
well as inhibition of radioiodide transport by perchlorate for all NIS-containing tissues, thereby
making the code consistent with the model description in Clewell et al. (2007). Is this inclusion
appropriate? Are the impacts of this inclusion transparently and objectively described?
Yes, this inclusion is not only appropriate but also needed, and the impacts of this inclusion are
transparently and objectively described in this analysis. This work reflects EPA's efforts in reviewing
the model established by Clewell et al. and seeking for improvement of the PBPK model.
B-72
-------
PEER REVIEW COMMENTS FROM
Lauren Zeise, Ph.D.
Chief, Office of Environmental Health Hazard Assessment
California Environmental Protection Agency
1515 Clay Street, 16th Floor
Oakland, CA 94612
510-622-3190
Email: lzeise@oehha.ca.gov
Note: Dr. Zeise performed this review as a private consultant and not as an agent
of the California Environmental Protection Agency
A-73
-------
-------
Lauren Zeise
Technical Charge to External Peer Reviewers
Contract No. EP-C-07-024
Task Order No. 54
October 22, 2008
INHIBITION OF THE SODIUM-IODIDE SYMPORTER BY PERCHLORATE:
AN EVALUATION OF LIFESTAGE SENSITIVITY USING PHYSIOLOGICALLY-BASED
PHARMACOKINETIC (PBPK) MODELING
WRITTEN COMMENTS ARE DUE NO LATER THAN MONDAY, NOVEMBER 10, 2008
CHARGE QUESTIONS
General Charge Questions:
G-l. Is EPA 's analysis logical, clear and appropriate in depth and length? Has EPA accurately, clearly
and objectively represented and synthesized the scientific evidence for the changes made to or
specification of the model code and input parameters?
In general EPA's analysis is clear and logical, and succinct, although there are several suggestions
for improvement in the comments below. With regard to depth, the limited treatment of variability
and uncertainty is problematic. Pharmacokinetic models can provide a structure for exploring and
integrating variability, but this was not done in this analysis. This major limitation is recognized by
EPA (page 25). EPA points out the model predictions apply to "a subgroup average for typical,
healthy individuals, and effectively describe the RAIU inhibition relative to that same individual as
his/her own control." EPA further points out that "These models were not designed to account for
whether the pregnant women is hypothyroid or iodine deficient." Analysis of such large, susceptible
populations is a critical aspect of understanding the potential health impact of perchlorate drinking
water exposure. A more rigorous and explicit treatment of variability is needed to get a better
handle on intra-human variability in response to perchlorate exposure. The analysis would also be
improved by more rigorous statistical and quantitative treatment of uncertainty. The degree to
which the analysis for the GW 40 fetus may or may not represent the first and second trimester fetus
needs explicit and careful treatment.
On a smaller point, it would help if greater motivation was provided for some of the statistical fits to
data. Some statistical fits provided an expedient and practical way forward in the analysis but
appeared to introduce logical inconsistency. It would be preferable for a more expanded discussion
A-75
-------
Lauren Zeise
to provide a context for the approach taken.
G-2. Please identify any additional studies or other data sources that should be considered in the
assessment of the specific parameters addressed below, including urinary clearance (of iodide and/or
perchlorate) and ingestion rates (breast milk, formula and water), especially in neonates.
Additional possible studies and data sources are identified in response to specific charge
questions below.
G-3. Besides the specific parameters identified below, please comment on any other parameter or model
choices described in the document that you think are incorrect or require further explanation or
where other available information could provide better estimates.
Co-exposures,
no biological susceptibility
No co-exposures or
biological susceptibility
§
JE
S
n.
¦=•
a.
Co-exposures and
biological susceptibility
Q
=
a-
*
a>
a.
Value of physiological parameter
Figure 1. Distribution of atypical physiological parameter within the population and hew that may vary
depending on the influence of chemical and biologic background.
The above figure, taken from Woodruff et al. (2008; EHP 116:1568), illustrates the main limitation
in the analysis and approach to modeling. The question in evaluating the potential risks from
perchlorate in drinking water is about the extent to which the incremental exposure from water
results in adverse effects to the mom, her baby, her developing fetus or others. In the above figure it
is above the extent to which the perchlorate drinking water exposure, in the presence of coexposure
and biological sensitivity, is creating adverse outcomes in the population. The fidelity of the
analysis depends on whether individuals with biological susceptibility have been adequately
addressed and also whether coexposures that affect iodide inhibition have been adequately
considered.
EPA analysis enables biological susceptibility and coexposures to be partially addressed in the
B-76
-------
Lauren Zeise
assessment, but it needs to move further to enable a fuller treatment. With regard to biological
susceptibility EPA considers susceptible subgroups - the infant, fetus, mom - and an important
factor that increases susceptibility in these groups - low renal clearance. But the analysis does not
enable the agency to consider the extent of impact on other sensitive subgroups in these populations,
such as those with clinical and subclinical hypothyrodism, those that may be genetically predisposed
(see e.g., Scinicariello, EHP 113(11): 1479-84), and those that are iodine deficient. The EPA
analysis also considers an important coexposure - perchlorate intake via food. However, the
analysis does not consider the combined impact with thiocyanate, which also affects iodide uptake at
the NIS. Thiocyanate is also found in breast milk (see e.g., Kirk et al. 2007, EHP, 115:182-186),
cigarette smoke, and common foods. The recent finding in women who smoked, that those with low
urinary iodine levels had decreasing T4 with increasing perchlorate (Steinmaus et al. 2007, EHP,
115:1333-1338) as well as reduced content of iodine in breast milk and the urine of breast feeding
infants of smokers (Laurberg et al. 2004, J Clin Endo Met 89:181-187) indicates the importance of
considering coexposures to thiocyanate. Nitrate, ubiquitous though far less potent than perchlorate,
should also be considered (see e.g., DeGroef et al., 2006, Eur J Endocrin 155: 17-25).
G-4. Please discuss research that you think would be likely to increase confidence in these models and
their use in predicting RAIU inhibition by perchlorate in different life stages (for an average
individual within each life-stage).
Research to get a better handle on renal clearance of iodine and perchlorate during pregnancy and
postpartum; biomonitoring of perchlorate, iodide, thiocyanate and thyroid hormone during and after
pregnancy during lactation in smoking and non-smoking women. Measurements of perchlorate in
baby formula - in non-composited samples.
G-5. Does EPA accurately characterize the strengths and limitations of the analysis? Please comment on
any particular strengths that may not be mentioned or adequately characterized for the estimates of
RAIUfor different life stages. Please also comment on any specific sources of uncertainty that you
believe have been overlooked in EPA 's analysis or that require further discussion, and which might
be significant to EPA's estimates of RAIUfor different life stages.
EPA does not sufficiently elaborate on the limitation of focusing on "healthy" individuals, and the
lack of consideration of the large susceptible populations.
Some parts of the analysis are scenario based, using 90th percentile values, while other parts use
A-77
-------
Lauren Zeise
mean values. With over 4 million infants born in the US each year, scenario analyses should be
added. These would be directed at ascertaining the inhibition levels for the some plausible higher
susceptibility cases, such as infant and fetus exposures associated with a mom with relatively high
thiocyanate exposure (e.g., from broccoli consumption or smoking), low renal clearance, who got all
her fluids directly or indirectly from tap water.
G-6. As recommended in EPA 's 2000 Risk Characterization Handbook, is the analysis transparent in
terms of the steps, logic, key assumptions, limitations, and decisions? Specifically, does
characterization of the results of EPA 's work fully explain: a) the analysis approach employed; b)
the use of assumptions and their impact on the analysis; c) the use of extrapolations and their
impact on the analysis; d) plausible alternatives and the choices made among those alternatives; e)
the impacts of one choice vs. another on the analysis; f) significant data gaps and their implications
for the analysis; g) the scientific conclusions identified separately from default assumptions and
policy decisions, if any; and h) the major conclusions, and the discussions of EPA's confidence and
uncertainties in the conclusions?
EPA does a reasonably good job laying out the logic, key assumptions, limitations and decisions.
But for the most part, it is done in a manner that will be understandable to someone with a modeling
background. It will be difficult to follow and very accessible to a more general reader. More
motivation of the forms for the statistical fits is needed, and a more quantitative and rigorous
treatment of uncertainty. EPA reasoning for using 90th percentile values for some parameters and
mean values for others is not explained well. Failure to address certain large susceptible populations
and the possible sizes of these populations should be discussed. The degree to which the analysis for
the GW 40 fetus may or may not represent the first and second trimester fetus needs explicit and
careful treatment.
Parameter-Specific Charge Questions:
(A) Urinary clearance
A-l. Please comment on the appropriateness of the input values selected for maternal urinary
clearance during pregnancy and lactation in EPA 's analysis. Are the available data and rationale
for the values selected transparently and objectively described?
The discussion of maternal urinary clearance values could be somewhat improved. In regard to
Figure B-6 motivation is not given for the fitting of the quadratic function to the data for iodide
B-78
-------
Lauren Zeise
clearance vs gestation week, and it is unclear where the postpartum data set - greater than week 39
data set - on the plot appeared from and why it is included in the modeling of clearance during
pregnancy. The highest mean value was measured by Adoul-Khair at the latest pregnancy time point.
Inclusion of the extra data set weighs the function down late in pregnancy when the highest value was
measured by Aboul-Khair. Further, including a gestation week of 45 on the plot axis is confusing to
the reader. There is a large extrapolation to clearance during the early pregnancy time point and renal
clearance can be increased fairly early in gestation. The quadratic fit may underpredict clearance
during this period. However, given that EPA is declining to estimate early fetal effects, this portion of
the extrapolation is not critical. It is unclear why the fit is being presented however. Finally variability
among individuals is an important consideration and it would therefore be of interest to see on the
plot or otherwise reported an indication of variability in the individuals studied. Some indication of
this is given in Table 2 of Aboul-Khair et al., where renal clearance values for iodide have been
serially averaged for each pregnant individual studied. In addition, individual measurements for
controls are given. For Figure B-7 it would be good to show error bars or confidence bounds.
Minor error, on page 42, data from Guignard et al. are plotted in Figure B-2 not B-l.
Are you aware of other publications or data that could be used to guide these choices or which
provide alternative input values that are equally valid or more appropriate?
Increased urinary clearance of iodide during pregnancy is a widely recognized phenomenon and data
on the magnitude besides that reported by Aboul-Khair would be useful and important to locate.
Further, the inconsistency of PK outcomes and the Aboul-Khair measured values in controls for IV
iodide dose uptake 2.5 hours post injection is quite troublesome and calls into question the PK
modeling. The approach on page 49 described to deal with the inconsistency is not entirely
satisfactory. EPA should look hard for additional data sets to cross check assumptions regarding
iodide uptake and renal clearance during pregnancy and early postpartum.
A-79
-------
Lauren Zeise
Likewise, are the values selected for urinary clearance for the infant and older child the best
estimates, given the available science and data? Are there other data that would provide better (or
equally valid alternative) guidance or estimates? Is there any reason to believe that urinary
clearance might not be a limiting factor in the elimination ofperchlorate for infants?
I am not aware of better values for infants and the older child. The EPA laid out a reasonable analysis
and approach for developing estimates for the infant and older child. There is interindividual
variability in clearance and it would be preferable if this were more emphasized and acknowledged in
the discussion, and attempts to better describe it quantitatively, for example in terms of varying
glomerular filtration rates normalized by body size.
(B) Breast-milk ingestion
B-l. Infants are generally known to consume very little milk or formula on the first day of life, with
ingestion quickly increasing over time. However the first time-point for which ingestion data are
available for infants is at 7 days, which EPA extrapolated back to 0 ingestion at time =0 in its
analysis (birth; Figure I). The breast-milk ingestion rate was estimated based on mean values
reported in Arcus-Arth et al. (2005), while the water ingestion used for the breast-feeding mother
was the 90th percentile value from U.S. EPA (2004).
Is EPA's extrapolation and rationale transparently and objectively described?
EPA's approach is objectively and transparently described, and the Agency was correct that the
Clewell et al. description is inconsistent with the currently available peer reviewed literature. It is
unclear why a mean value is used for infants and an upper 90th percentile is used for the breast
feeding mother. This is not adequately explained.
Does this function appropriately characterize the available data and information?
The function does not characterize the available data and information. It will be quite confusing to
anyone but a modeler.
The equation on page 15 has milk describes milk ingestion rate as
Milk ingestion rate (mL/hr) = KTRANS = 28.3*(BW-3.375)0.175
B-80
-------
Lauren Zeise
It then plots milk ingestion as a function of body-weight and shows values for days 1,3,5 and 7 of
life as on the bw vs milk ingestion plot. This formulation was used as a convenient way of giving
values to KTRANS but is problematic because it works only for the specific circumstances using the
mean values for breast milk intake in Arcus-Arth et al. data and will be confusing for anyone but a
modeler.
In using bodyweight as a surrogate for age (3.375 kg as the zero age bodyweight) it builds in an
illogical structure that will be hard for the general public to understand and limits the usefulness of
the model for using data beyond the mean values in Arth-Arcus et al. For example, there is zero
milk ingestion for a bodyweight of 3.375 and milk ingestion rates below that value cannot be
defined. Furthermore, the expression has milk ingestion increasing with increasing bodyweight
indefinitely. This also is contrary to what occurs - as infants age solid foods and other liquids are
introduced and breast feeding reduces. Arth-Arcus et al. show that for the available data sets milk
consumption - in terms of volume per bodyweight per day - decreases with age in a linear fashion.
Thus there is another inconsistency introduced by the way the model is formulated. At different
ages the mean milk ingestion at a given bodyweight will differ.
A more logical approach would be to develop an expression for milk ingestion in terms of volume
per bodyweight per day could be expressed as a function of age. A separate expression could then
be used to convert this to KTRANS. To deal with the early low consumption rate on days 1-3 the
measured values could be used.
Figure 1 notes that the data are from Arcus-Arth et al. but in fact it is entirely inconsistent with
Arcus-Arth et al. for the above discussed reasons.
Are there other data that could be used to obtain a better or equally valid alternative estimate of mean
breast-milk ingestion rate for infants in the first few days of life?
It is not correct that the first time point for which ingestion data are available is 7 days. There are
values in the literature for intake on days 1, 2, 3, 4 and 5. Indeed at days 4 and 5 the intake is quite
high and consistent with the linear relationship for volume consumed per kg per day vs age reported
in Arth-Arcus et al. See table 8 in that paper.
A-81
-------
Lauren Zeise
Should an estimate other than the mean be used to determine the breast-milk ingestion rate?
Whether or not the mean is used depends on later steps in the process, and ways that
variability will be taken into account. There are over 4 million births in the US annually.
The overall procedure for characterizing intra-species variability and central tendency needs
to be designed to be able to address the large number of infants "in the tails" of the
distribution. It would be preferable to build a PK approach that would enable fuller
description of variability in iodide uptake inhibition. The use of mean values and the
formulation used to compute KTRANS precludes this.
(C) Water ingestion
C-l. For pregnancy, EPA used a normalized (90th percentile) water ingestion rate of 33 mL/kg-day
(U.S.EPA 2004), which was multiplied by the maternal BWas described by the PBPK model
growth-functions during pregnancy to obtain total water ingestion for the mother.
Is EPA's approach and rationale transparently and objectively described?
The approach is transparently and objectively described, but the rationale is somewhat
unclear. Some parameters are based on mean values, others on midpoints and still others on
upper 90% bounds. It would be of interest to understand parameter distributions and how
this translates to distributions for iodide uptake inhibition. This may be beyond what EPA
has resources and time to do, but failing that, it would be desirable to have a clear
presentation of the approach. EPA appears to be taking a plausible scenarios approach. But
a clearer explanation is needed.
The table below is taken from EPA (2004). It shows the 95th percentile upper bound for
community water as 43 mL/kg/day, a reasonably higher level than the 90th percentile. In the
perchlorate document, the reason for choosing the 90th percentile and not some other value
needs to be justified. It is also worth noting that the number of pregnant women captured in
the survey is quite small, and raises some concern that the upper bound values may be under
B-82
-------
Lauren Zeise
estimates. For example, the upper bound estimate on the 90th percentile for pregnant women
was 46 mL/kg-day.
Table i.JU2. Per Capita Water Gaosmnptioti—Pregnant Women fiat/kg/daty)
Estimate
Total Watei
IP'S NFCS
Erilic-M- ft. a]
l>ai.
T.ip \v\iteL1
1°"8 N1TCJ
Eic.ho"^' ft si.
iopl.
Total
QS
'-.iDBTI
C-Miujiniiity Yi'ate!1
Sample Sise
1SB*
»8«
00*
85#
Mean.
:;.i
18 3
21*
14*
§§"•
23.5
10.4
19'
Q>
90* *
4S.P
34.5
30*
33*
OS* *
53.5
44*
43*
* Wotaai aged 15 to 4P yean; f Wanes agrf 15 to 44 ysass.
Is this approach appropriate for characterizing the upper-bound for ingestion of pregnant
women?
With 4.3 million births in the US each year, above the 90th percentile will be 430,000
women. Thus a very large number of women may consume water above this level, and one
is left wondering about the importance of the assumption and how sensitive the results are
to it. Following EPA (2004), the upper 95 percentile is 44 mL/kg/day, still representing a
rather large number of women - 215,000.
Are there other approaches or data that could be used to obtain a better or equally valid
alternative estimate for this parameter? (Note that the self-reported body weight values in
U.S. EPA (2004) indicate the same average body weight for pregnant women as for non-
pregnant; data appears inaccurate, so the PBPK model body-weight description was used
instead.)
There are other approaches that could be used to obtain a better or equally valid alternative
estimate. One consideration is the extent to which we may be confident that a pregnant
woman may use drinking water in cooking and for her fluid intake without having to be
A-83
-------
Lauren Zeise
concerned about harming her fetus. For this analysis one might consider the basic water
requirments for women living in hot climates. For this one might select a value somewhat
above the value of 3.0 L/day considered an "adequate intake" by the Institute of Medicine
(2004; Dietary Reference Intakes for Water, Potassium, Sodium Chloride and Sulfate, IOM
Food and Nutrition Board).
Another would be to pick a plausible upper bound value from the cumulative distribution
observed. For example, from the figure below, taken from EPA (2004), it can be seen that a
reasonable plausible upper bound may fall between 3.5 and 4 liters per day.
Figure 8,2.Cta. CbbiuWvb Distributions of ft* Capita Direct and Indirect Water Ingestion
Pregnant Con$UTri@r5 Only
ml^ecsoaa/day
I ' ' ¦ ¦ I ¦ ' ' ' I ¦ ¦ ¦ ¦ I ¦ ¦ ¦ 1 I • ' ' ¦ I ' ' ' ' 1 ¦ ¦ ¦ ¦ I ¦ ¦ ¦ ¦ I 1 1 • ' I ¦ ' ' ' I ' ¦ ¦ ¦ I ¦ ¦ ¦ ¦ I—
O 5<'n> -©SOSi- J3u<> »u int'Hr .JOttO JijVoO HUHWJ -aoS-iK" ¦Sm.Ki
B-84
-------
Lauren Zeise
C-2. For lactation, EPA used a fixed total (90th percentile) water ingestion rate of2959 mL/day (U.S.
EPA 2004), with the rationale that while the woman's weight and self-water demand are expected to
drop after pregnancy (as described by the PBPK model time-dependent weight equations), the
demand for milk production would be increasing, and the reported value was not for a specific age-
range of child.
Is EPA 's approach and rationale transparently and objectively described?
The approach is transparently and objectively described, but the rationale is somewhat
unclear. As noted in response to C-l, some parameters are based on mean values, others on
midpoints and still others on upper 90% bounds. A clearer explanation is needed on why
the 90th percentile is chosen here, and not some other higher bound given the number of
women-infant pairs affected.
The tables below are taken from EPA (2004). They show the 90th percentile upper bound
for community water is not substantially smaller than the 95th percentile when expressed as
mL/kg/day, but appears more different when expressed as mL/person/day (2959 vs 3588),
suggesting the difference may be driven by bodyweight differences at the 90th and 95th
percentile. Still the reason for choosing the 90th percentile requires further explanation.
TsftLs Fsi Capita Wafer Ciimn wi|iliiin—ZadatiBg Von (mLolstf rt)
tm-: v-:*.
LS>"1 ET
1PP1-
Tai:
l
iKi:Lor: ^ -1
LPPi.
T::^l
i::4—it
r o iii in i ¦ Lii ^ ¦ v
i«--=: :=
77 ¦
11 •
JJ#
J >3^
J7.0
2:1.4
2»
no*
so* m
3S,t
ms
23*
20*
wr «t
53,7
311
33#
34*
PS** It
JP.2
37.4
57*
33*
¦" X-cjttx 'ft.*. ¦.; -j -v ¦ *!-: = "i "o-jiji: ? 1 j ¦; —
I TL: jungle t-x: an ¦ apimnEols as described in. Ifae 'medtepartwliwriiinii
kJ'lLU'C*. III Z
A-85
-------
Lauren Zeise
Table CI, Per Capla Water CcuraHupric»—LaetntnigWoaaai (adL^ie ::ml '-ji;
Tof;l ".V; Iti
:f-f N7CC
;l.
iLff
.•.in-
T: Vr;:
::n]
ii-'T ""jttfi-
f-l-f "i.
TaLi-pl* Jin
77'
:-i=
Men
i .r it
IMS#
:
L 4
-------
Lauren Zeise
Looking at the cumulative distribution observed for lactating women, in the figure below
taken from EPA (2004), it can be seen that a reasonable plausible upper bound may fall
somewhere around 4 liters per day.
Kigiirn fi.2.C"lh. Cumulative Distributions of Pur Capita Miroct find Inriirnrt Watnr Ingestion
[.firlHlinu, [ Onlv
ml/person/day
W,"»r t-iu. ; i -.i11111 iri.r.^m.r^u Jt 1-iir)
C-3. For bottle-fed infants, EPA made extrapolations of the 90th percentile water-ingestion in early life
based on measurements made for the age ranges: 0-30 days, 1-3 months, and 6-12 months (points in
Figure 2, upper panel). For the purposes of this analysis, bottle-fed implies feeding with formula
requiring the addition of water.
Is EPA's extrapolation and rationale transparently and objectively described?
Yes, although the reason for using a quadratic relationship was not described. The approach of
expressing water ingestion in units mL/kg/day and modeling it as a function of age is much
preferred over the approach used for breast milk consumption (e.g., in Figure 1).
A-87
-------
Lauren Zeise
Does the overall function used to represent the changes in ingestion with age (and body weight)
appropriately characterize the available data and information?
This approach is a reasonable way of describing the upper 90th bound given in the Kahn and Stralka
(2008) paper.
Are there other data that could be used to obtain a better or equally valid alternative estimate of
water ingestion rate for infants in the first few days of life (e.g., 7-day old)?
An alternative for estimating ingestion rate for the first few days of life would be to rely on data sets
for breast milk consumption during the first 7 days (e.g., Casey et al. 1986, Am J Dis Child 140:933;
Neubauer et al. 1993, Am J Clin Nutr, 58:54), since breast fed infants do not require supplemental
water and the results may be more indicative than the assumed relationship used, although sample
sizes are relatively small. It is noteworthy that intake in mL/kg/d during this period is not a smooth
function of body weight. It is quite low during the first two days of life but by age four or five days
the intake is essentially the same as at age 7 days. It is possible that the function l-e"daydoes a
reasonably good job of describing this. EPA could compare the values predicted by this function at
days 1-7 to those seen in the literature for breast milk consumption on those days.
The water ingestion rates used by EPA are based on 90th percentile ingestion data and thus are
likely to exceed (minimal) physiological needs of infants as defined in nutritional guidelines. Are
the water ingestion rates used by EPA reasonable in comparison to the physiological needs of
infants at these various life stages? Are there other approaches or data (e.g., see the FDA memo)
that could be used to obtain a better or equally valid alternative estimate for this parameter?
The available data used by EPA indicate that infant formula consumption varies by individuals, and
correspondingly water consumption does as well. It is reasonable to consider the minimal
physiological needs of infants, as defined in nutritional guidelines, although a precise understanding
energy needs and use in infancy still appears to be a matter of discussion (Reilly et al. 2005, Br J
Nutr 94: 56-63). At any particular age bodyweights, growth rate and degree of activity varies, and
so consumption can not be precisely calculated based on formula energy content and recipes for
making up bottle fed formula. Further, some infants are overfed and others are underfed. Thus
although it would be useful to compare water consumption with what one would expect given
nutritional guidelines and typical formula recipes, the nutritional guideline would not lead to a
reliable upper bound value for water consumption. Assumptions would be needed to go from the
water consumption based on the nutritional guideline level to an upper bound estimate. Though as
the FDA memo notes, "there is a relationship between the volume of water an infant needs, and
B-88
-------
Lauren Zeise
his/her caloric requirements for healthy growth" the exact relationship to assume and the
interindividual variability in that relationship has not been provided and it is unclear that it would
provide a more reliable estimate of water consumption than is given in EPA's perchlorate report.
(D) Perchlorate concentration in formula
D-l. EPA used 1.42 fig/L as the concentration ofperchlorate in formula for bottle-fed infants. This
estimate was based on information from FDA 's Total Diet Study, supported by Pearce etal. 's
(2007) findings.
Is EPA's approach and rationale transparently and objectively described?
The approach is not entirely transparent and the description could be improved. A sample
calculation for Table 4 describing how perchlorate intake for bottle fed infants is estimated would be
helpful.
Is 1.42 fig/L an appropriate value to use for the concentration ofperchlorate in infant formula? Are
there other better or equally valid alternative approaches or values that could be used?
According to the Pearce et al. methodology:
"Seventeen brands of infant formulae were also assessed for iodine and perchlorate levels. A
single sample of each different type of liquid formula available at a local supermarket was
purchased for testing. Nine brands were sold in concentrated form and designed to be diluted by
half before use. Iodine and perchlorate levels were measured directly in these samples, and the
results were divided by half to reflect the concentration intended for infant use. The other eight
brands were sold ready for use."
Thus, for the nine formula that were designed to be diluted, Pearce et al. assumed that there was no
perchlorate in the diluting water. EPA reports the correct average of 1.45 j^ig/L calculated from 17
Pearce et al. samples. But for the young bottle fed infant the calculation should reflect the intake of
perchlorate from neat formula plus the intake from the water used to dilute it. It is reasonable to
assume that the only perchlorate intake in the seven and 60 day infant would be water and formula.
Thus the undiluted values for formula perchlorate should be used.
The undiluted average from Pearce is 1.97 |_ig/L. but that includes formula that is ready to use
undiluted as well as formula that requires dilution. For use in Table 4, the focus should be on
A-89
-------
Lauren Zeise
concentrations of formula that would require dilution. The young seven and 60 day infant population
drinking ready to use formula with no other consumption is more a concern of the FDA than the
EPA; they would not be receiving perchlorate contaminated tap water. In the Pearce et al. study, the
perchlorate concentration in the 9 samples of formula that would be diluted was 1.96 (ig/L. The two
highest of the nine values reported would require dilution correspond to 3 (ig/L and 3.2 (ig/L, double
the value reported in Pearce et al. Table 1.
The problem with the FDA data is that they represent composite samples, prepared as they would be
expected to be consumed. Also, the detection limit used by FDA is 1 j^ig/L. The composite would
be averaged across different formula brands and certain types. Thus they do not provide an
indication of what higher end exposures might be. The composite sample results, in units (.ig/L. are:
202 Infant formula, milk, hi-Fe: ND, 2.5, 2.0, 2.0
203 Infant formula, milk, lo-Fe: 1.2, ND, 3.6, 2.1
309 BF, infant formula, soy: ND, ND, 0.8 *, 0.8 *
* indicates above the limit of detection but below the limit of quantitation and ND indicates not
detected.
Each value represents a composite from three cities in a given region. Thus a concentration in
particular product may be three times as high as the value reported. Because of consumer loyalty
and habit it is far more likely that a consumer will use the same product over an extended period of
time. From the values tabulated, value of 1.42 |_ig/L will be an underestimate of perchlorate
concentration in contaminated infant formula. Further, the concentration of perchlorate in water
used by FDA to prepare the formula in to-be-eaten form has not been reported, but is likely to be
low or not present, given the several NDs in the table. Because FDA uses composite samples, it
would be preferable to use the high end value from FDA (3.6 j^ig/L) or a value of say 3 (ig/L from
Pearce et al. Clearly better and more extensive measurement of perchlorate in infant formula is
desirable.
B-90
-------
Lauren Zeise
(E) Radioiodide excretion into breast-milk by NIS
E-l. In the model, EPA included perchlorate inhibition of NIS radioiodide excretion into breast-milk, as
well as inhibition of radioiodide transport by perchlorate for all NIS-containing tissues, thereby
making the code consistent with the model description in Clew ell etal. (2007). Is this inclusion
appropriate? Are the impacts of this inclusion transparently and objectively described?
It is reasonable and appropriate to assume that perchlorate inhibits the transport of iodide in
NIS containing tissues and iodide excretion into breast-milk. The impact of its inclusion is
transparently and objectively described. There is a straightforward layout in Appendix A of
changes in model assumptions and their impacts. Further, the effect of decreased iodide
levels in breast milk from smoking - with potential inhibition caused by thiocyanate - has
also been observed (Laurberg et al. 2004, J Clin Endo Met 89:181-187), consistent with the
finding that this should be taken into account in the modeling.
A-91
------- |