Summary of Public Comments on Materials for the Peer Review of EPA's Draft Report: "Proposed Modeling Approaches for a Health-Based Benchmark for Lead in Drinking Water"
                Prepared by Eastern Research Group, Inc. (ERG)
                                April 20, 2017
                                       
 Introduction
This report summarizes public comments on the U.S. Environmental Protection Agency's (EPA's) draft report, "Proposed Modeling Approaches for a Health-Based Benchmark for Lead in Drinking Water", and associated charge questions. Eastern Research Group, Inc. (ERG), a contractor to EPA, is organizing a peer review of this report, and will provide this comment summary and a complete set of public comments and attachments to the reviewers as they prepare their preliminary individual written comments in response to EPA's charge questions. The charge questions seek reviewer comment on the scientific aspects of three potential modeling approaches to associate lead in drinking water with blood lead levels (BLL) for sensitive life stages, including formula-fed infants and children.
In the draft report, EPA's Office of Water (OW) documented three modeling approaches, input parameters, and exposure scenarios that will inform development of a health-based benchmark: the household action level (HAL) for lead in drinking water. OW is considering revising the National Primary Drinking Water Regulations for Lead and Copper (the "Lead and Copper Rule" or LCR) to improve public health by adding this action level. OW will consider the specific role for this value when proposing revisions to the LCR after completion of the peer review. The peer review will focus on scientific aspects of the modeling approaches. Reviewers will not be asked to comment on the specific role for a health-based value. 
EPA provided the draft report and charge questions for public comment on January 18, 2017, with an initial deadline of March 6, 2017, which EPA extended to April 5, 2017, based on an early comment requesting an extension. Twelve commenters (see Table 1) submitted written comments, which are summarized in five sections: 
 Section 2: General context and purpose of the peer review. 
 Section 3: Public comment opportunity. 
 Section 4: EPA's draft charge questions.
 Section 5: The three modeling approaches described in EPA's draft report. This section is organized into four areas: model scenarios, model inputs, modeling approaches, and model evaluation/sensitivity analysis.
 Section 6: Interpretation and use of a modeled health-based benchmark for lead in drinking water. 
In this summary, organizations commenting are identified by the acronym shown in the first column of Table 1.  Three organizations submitted attachments with their comments, as documented in Table 1. 
                          Table 1: Public Commenters
                                 Organization
                                   Signator
                                  Submission
American Water Works Association (AWWA)
G. Tracy Mehan III, Executive Director of Government Affairs
2-page letter requesting extension to comment deadline.
American Water Works Association (AWWA)
G. Tracy Mehan III, Executive Director of Government Affairs
6-page letter and one attachment consisting of a 21-page analysis describing an AWWA consultant's (Dr. Douglas Crawford-Brown's) review of the draft materials.
Anonymous #1
Anonymous 
Two-sentence statement.
Association of State Drinking Water Administrators (ASDWA)
J. Alan Roberson, Executive Director
1-page letter.
Anonymous #2
Anonymous 
Brief (one-paragraph) statement.
Academy of Nutrition and Dietetics (AND)
Jeanne Blankenship, Vice President for Policy Initiatives and Advocacy; and Pepin Andrew Tuma, Senior Director for Government & Regulatory Affairs
3-page letter.
American Water (AW)
Matthew Corson, Manager for Environmental Compliance and Stewardship
1-page letter.
Environmental Defense Fund (EDF)
Tom Neltner, Chemicals Policy Director; Ananya Roy, Health Scientist; and Lindsay McCormick, Project Manager
4-page letter and an attachment consisting of three recent EDF blog posts: (1) "With draft report, EPA takes major steps to help communities access risks from lead in drinking water"; (2) "When it comes to lead, formula-fed infants get most from water and toddlers from food, but for highest exposed children the main source of lead is soil and dust"; and (3) "EDF's assessment of a health-based benchmark for lead in drinking water."
Earth Justice, Louisiana State University Health Sciences Center, Childhood Lead Action Project, Flint Rising, Freshwater for Life Action Coalition, Parents for Nontoxic Alternatives, Water Alliance, and Women for a Healthy Environment (EJ et al.)
Jennifer C. Chavez, Staff Attorney, et. al.
5-page letter.
Natural Resources Defense Council (NRDC)
Mae Wu, Senior Attorney; and 
Miriam Rotkin-Ellman, Senior Scientist
4-page letter and two attachments: (1) State of California. Public Health Goals For Chemicals In Drinking Water Lead (April 2009); (2) California Environmental Protection Agency. Office of Environmental Health Hazard Assessment. Development Of Health Criteria For School Site Risk Assessment Pursuant To Health And Safety Code Section 901(G): Child-Specific Benchmark Change In Blood Lead Concentration For School Site Risk Assessment (April 2007).
Texas Commission on Environmental Quality (TCEQ)
Richard A. Hyde, Executive Director
Cover letter and 5-page comment statement.
Michigan Department of Environmental Quality (MDEQ)
Christine Flaga, Chair of Toxics Steering Group
3-page letter.

 Review Context and Purpose
Several commenters (AND, Anonymous #2, ASDWA, AW, EDF, and MDEQ) acknowledged the importance of a model that defines the relationship between lead in drinking water and blood, and expressed support of EPA's principal goal of developing a health-based benchmark for lead in drinking water. In general, these commenters highlighted the need to develop the proposed models to improve public health protections of the LCR. One commenter (AND) described the cost of inaction, as quantified by childhood lead poisonings in the U.S., as another reason to support the proposed regulatory revisions. Two commenters (AW and MDEQ), however, encouraged EPA to provide further context for the eventual use of modeled outputs. For example, MDEQ found the proposed modeling approaches to be scientifically robust but expressed difficulty evaluating the appropriateness of the different approaches in the absence of a defined adverse health effect association and a specifically defined role for the modeled health-based benchmark. AW expressed similar concerns and suggested that EPA start the formal peer review meeting with a clear, robust discussion of what the modeled numbers represent.

Two commenters did not support development of a health-based benchmark with the proposed modeling approaches (NRDC, and TCEQ). NRDC suggested that EPA base a public notification on the Maximum Contaminant Level Goal (MCLG) or a Maximum Contaminant Level (MCL) that is as close to the MCLG as feasibly possible. This commenter also suggested that EPA consider the approach used by the California Office of Environmental Health Hazard Assessment (OEHHA). TCEQ) argued that the draft report and January 19, 2017, notice in the Federal Register should be withdrawn entirely until the new leadership of EPA has an opportunity to study the materials. 

Another commenter (EJ et al.) was cautiously optimistic about EPA's efforts to better characterize the health effects of lead in water on BLLs, but expressed concern with the benchmark concept that EPA is contemplating.
Several commenters (ASDWA, AW, EDF, and EJ et al.) acknowledged general support of EPA's decision to seek expert peer review of the potential modeling approaches that may be used to derive a health-based benchmark for the LCR. For example, one commenter (EJ et al.) stated, "rigorous peer review is needed, at a minimum, to determine whether the models are capable of deriving a conservative value that reliably reflects both potential exposure and effects of exposure on blood lead levels (BLLs) in the most vulnerable population."
 Public Comment Opportunity
Most commenters expressed appreciation for the opportunity to comment on the draft report (AND, ASDWA, AW, AWWA, EDF, MDEQ, NRDC, and TCEQ). One commenter (TCEQ) stressed that EPA should carefully consider and respond to public comments, in a thoughtful, scientific, and common-sense way.
Two commenters (AWWA and TCEQ) did not think the original public comment period was sufficient. The AWWA requested that the public comment period be extended 60 days beyond the original March 6, 2017 deadline. This organization argued that the importance of the modeling efforts warranted additional time for the public to evaluate the materials and provide meaningful comments to the peer review panel. This concern was echoed by TCEQ, who stated that the 45-day public comment period was inadequate to complete an in-depth review of the materials. TCEQ further argued that the 30-day extension, offered two weeks prior to the original deadline, did not help organizations conduct a more thorough analysis. To meet the original due date, organizations would have already shifted their focus to a few key issues, rather than critically reviewing the draft report and proposed modeling approaches, including input parameters and exposure scenarios.
 Draft Charge Questions
Three commenters (ASDWA, AWWA, and MDEQ) commented on EPA's draft charge questions to peer reviewers. MDEQ and ASDWA approved of the questions and found them sufficient to evaluate the proposed modeling approaches. AWWA suggested that the panel charge be expanded to request a recommended minimum appropriate biological averaging period for the models and advice on the minimum duration of exposure that can be appropriately evaluated using the available modeling. AWWA stressed that the peer review panel will need to know the specific risk management strategies to be evaluated with model outputs and the analytical strategies into which the model output will be integrated. AWWA further recommended that EPA provide the peer review panel with the National Drinking Water Advisory Council (NDWAC) recommendation concerning developing a household action level for lead and stressed that the panel be explicitly charged with evaluating modeling approaches with the NDWAC use in mind. Finally, AWWA encouraged the panel to focus on the specific needs of the drinking water program, and not those of the air, waste, and toxic substance control programs.

 Review Materials
5.1	Model Scenarios
The proposed modeling approaches consider lead exposures for three sensitive age groups (i.e., 0- to 6-month-old infants fed formula reconstituted with tap water, 1- to 2-year-old children, and 0- to 7-year-old children). Two commenters (MDEQ and EJ et al.) expressed concern that the models do not evaluate fetal exposures associated with maternal exposures to lead in drinking water. MDEQ encouraged EPA to prioritize their research efforts toward developing a model that would assess exposure risk to the developing fetus while reminding EPA that the existing Adult Lead Model (ALM) only accounts for maternal lead exposures from soil on the fetus. EJ et al. highlighted this as a serious concern and suggested that EPA include, at the very least, a disclaimer concerning the population groups that are not accounted for in model outputs. 
One commenter (EDF) was confused by why some sections of the analysis only considered formula-fed infants, while others dealt with all infants. This commenter recommended a comparison of modeling results between formula-fed infants and those not fed formula. 
5.2	Model Inputs 
Five commenters (EDF, EJ et al., MDEQ, TCEQ, and AWWA) suggested changes to the existing input parameters for the various modeling approaches, and three commenters (Anonymous #1, EDF, and EJ et al.) proposed other parameters. 
      Changes to Existing Input Parameters 
Two commenters (EDF and EJ et al.) identified drinking water intake calculations as a critical aspect of the proposed modeling approaches. EDF was concerned that the drinking water ingestion rate used for 0- to 6-month-old infants may underestimate true ingestion for infants fed formula reconstituted with tap water (bottle-fed infants). EJ et al. expressed similar concerns regarding the static intake assumptions for formula-fed infants, stressing this population's high water intake on a body weight basis and high absorption rate for lead. Both commenters encouraged EPA to consider the World Health Organization's (WHO's) default value for water intake by bottle-fed infants (0.75 liters/day) and EPA's own exposure guidelines (ranging from 0 liters/day to 1.6 liters/day, with a 90[th] percentile of 0.88 liters/day and 95[th] percentile of 1.0 liter/day). EDF also questioned why the daily water consumption rates for 0- to 7-year-old children drop off from 0.410 liters/day to 0.151 liters/day when a child reaches 1 year of age, and sought clarification on whether this shift was due to consumption of other beverages after that age.
One commenter (MDEQ) identified an internal inconsistency with the geometric mean maternal BLL used in the IEUBK model, referencing the EPA Office of Solid Waste and Emergency Response's OLEM Directive #9285.6-55 released on August 2, 2016. This directive recommended a BLL of 0.7 ug/dL, while the proposed modeling approaches apply a BLL of 0.61 ug/dL.
Another commenter (TCEQ) highlighted an inconsistency with EPA's statement on page 19 of the draft that states "the input parameters used in this analysis do not represent high-end exposure." This commenter provided details on why the proposed outdoor air, soil/dust lead concentration, and soil ingestion rate inputs do include higher-end exposures. Their main points were:
 Outdoor air  -  TCEQ argued that the outdoor air concentration used in the models is based on the average of the highest rolling 3-month average air monitoring data from 50 urban areas and therefore represents "worst-case, high-end, relatively short­term urban exposure" and not "typical childhood exposure."
 Soil/dust concentrations  -  TCEQ argued that the soil and dust geometric mean lead concentrations used in the models are based on soil sample results from the American Healthy Homes Survey for homes built prior to 1950 and therefore represent high-end lead exposures at homes with lead-based paint. To the contrary, MDEQ was concerned that statistics based on these data may not accurately reflect the disproportionate lead-contaminated housing stock present in older cities. MDEQ cited the number of homes constructed before 1950 from the draft report (approximately 20%), and compared this to Detroit, where approximately 50% of homes were built before 1950. Based on these differences, MDEQ concluded that the soil and dust lead level inputs may underestimate lead levels in urban areas with older housing stock. TCEQ also mentioned that mean concentration inputs likely represent high-end exposure in Texas and other southern states.
 Soil/dust ingestion rates  -  TCEQ argued that central tendency soil and dust ingestion rate inputs may overestimate mean exposure. Based on a meta-analysis of soil ingestion rates, they estimated that the mean soil ingestion rate may overestimate exposure by an order of magnitude for children aged 0 to 2 years and 2-fold for children aged 2 and 3 years.
      Other Input Parameters Proposed
Three commenters (Anonymous #1, EDF, and EJ et al.) mentioned other inputs that EPA should consider in their proposed modeling approaches. One commenter (Anonymous #1) noted that the models do not account for exposure to lead from tobacco smoke and cited an estimate from the European Food Safety Authority that up to 20% of overall lead exposure in infants could be due to environmental tobacco smoke. Another commenter (EJ et al.) questioned why the models do not consider exposure to lead from drinking water through the uptake of cooked food, and suggested that this weakness could greatly underestimate the effects of lead in water on BLLs. A similar concern was expressed by a third commenter (EDF), who questioned whether the water consumption rates included in the model accounted for water retained in food from home cooking.
One commenter (EDF) was concerned that the proposed modeling approaches do not account for the strong seasonal variability of BLLs associated with changes in dust, soil, and paint chip exposures to children, as well as higher lead solubility in water and increased water ingestion during the summer. They argued that lower baseline lead levels from background exposures (e.g., dust, soil, paint chips) coupled with underestimated water consumption during the summer months could result in higher modeled estimates of the water concentration required to reach a given BLL in children.

5.3	Modeling Approaches	
EPA proposed three modeling approaches in the draft report, all of which use the Exposure Uptake Biokinetic Model (IEUBK) Model for Children. Two of the approaches are individual-based, while the third approach is population-based and incorporates the Stochastic Human Exposure and Dose Simulation (SHEDS) multimedia model. Several commenters (EDF, NRDC, and TCEQ) discussed concerns that apply to all three of the proposed modeling approaches, as briefly summarized below.
 It was not clear to EDF whether EPA updated the IEUBK model based on 2011 recommendations from the Agency's Science Advisory Board. 
 NRDC did not support development of a household action level or health-based benchmark, but suggested EPA consider the California OEHHA approach, if EPA decides to proceed with modeling. NRDC recommended that EPA use the OEHHA-derived level of concern (1 ug/dL) to translate a drinking water level using health protective values.
 TCEQ recommended an approach based on 5% increase in the probability of exceeding 5 ug/dL and justified the approach by claiming it is: "(1) consistent with that used historically by EPA for assessing acceptable lead exposures (i.e., no more than a 5% chance of exceeding 10 ug/dL for children's surface soil impacted by historic Superfund sites); (2) more meaningful in terms of actual risk; and (3) consistent with a more stringent (yet reasonable) probability criterion that may be used by EPA for more serious effects in benchmark dose modeling (i.e., a benchmark response of 5% as opposed to 10%)." TCEQ further recommended the health-based benchmark be derived using geometric mean background soil and dust concentrations of 37 ug/dL and 72 ug/dL, respectively. These conditions result in a drinking water concentration of 15.4 ug/L, which is nearly identical to the current lead action level (15 ug/L). TCEQ concludes that "based on the options currently in the draft report, it appears the current action level that is based on the technical feasibility of corrosion control is also health protective for the formula-fed, average infant scenario recommended by EPA's own NDWAC (as well as all infants combined)."
Comments specific to each approach are summarized below.
 Approach 1: This approach estimates the concentration of lead in drinking water that would result in a 1 to 5 percent increase in the probability of a child having an elevated BLL. One commenter (EDF) thought this approach provided the most useful information because it is grounded in the Reference Level established by the Centers for Disease Control and Prevention (CDC), which is updated every four years to reflect progress made in reducing children's exposure to lead. EDF also emphasized that the approach is conceptually consistent with EPA's method to derive its lead-based paint hazards standard. 
 Approach 2: The second modeling approach estimates the concentration of lead in drinking water that would result in a 0.5 ug/dL or 1 ug/dL increase in a child's geometric mean BLL. One commenter (EDF) found this to be a reasonable approach due to its focus on the incremental contribution of lead in drinking water to BLLs. This commenter did, however, question why EPA chose the specified geometric mean BLL increases and stated that those values are not linked to the overall risk to children.
 Approach 3: The third proposed modeling approach estimates the amount of lead in drinking water that would result in a predicted distribution of BLLs for a target population of children with a 95[th] and 97.5[th] percentile of 3.5 ug/dL and 5 ug/dL, respectively. EDF found the description of the proposed approach to be confusing, but thought that the model provided useful insight into the relative contribution of the main sources of lead exposure. This commenter also saw the value in comparing the models to blood lead data recorded from the National Health and Nutrition Examination Survey (NHANES). AWWA (which commented only on third of the three approaches) recommended that EPA consider modifying this approach to first consider all routes of exposure and then examine what happens to the distribution of BLLs when lead in water is reduced below a series of target levels. MDEQ expressed concern regarding whether this modeling approach, which simulates exposure over a 30-day period, violates any underlying assumptions of the models (e.g., the reliability of the IEUBK and ALM has not been assessed for durations less than 3 months). They also asked for justification on how the 5 ug/dL reference BLL from the CDC and the additional modeled value of 3.5 ug/dL are appropriate for the 0- to 6-month scenario since this age group falls outside of CDC's lead reference age range. In general, MDEQ expressed a preference for this approach since the other two approaches rely entirely on the IEUBK model, which assumes a constant concentration of lead in drinking water and thus, may underestimate BLLs. TCEQ cautioned EPA against tying an action level to an ever-moving target such as the 97.5[th] percentile of the U.S. childhood BLL distribution since that value is neither health-based or static.
5.4	Model Evaluation and Multimedia Exposure Pathway/Sensitivity Analyses
The proposed modeling approaches evaluate different lead exposure scenarios including (1) drinking water as the lone source of lead, and (2) drinking water exposures in conjunction with other lead exposure pathways (e.g., air, food, soil, dust, maternal contributions). One commenter (EDF) commended EPA on the inclusion of all likely sources of exposure. However, AWWA commented that the modeling approaches allowed other lead exposures "to occupy the bulk of the lead risk cup and then assume that the small incremental changes in lead exposure through water are determining increases in BLLs." This commenter suggested that the panel evaluate an aggregate lead exposure analysis, where the concentration of lead in water is fixed at candidate pre-defined levels. AWWA also recommended that EPA provide the panel with a sensitivity analysis that tests the model predictions against input data and assumptions. This would offer the panel information to "(1) determine which elements of the model are most influential, (2) evaluate confidence in the model predictions given the primary drivers within the model, (3) inform where the inputs to the model need to be refined or outputs constrained, and (4) judge the appropriate application of specific modeling approaches given the embedded biases within the model algorithms." 
AWWA also encouraged EPA and the peer review panel to explore the truncation and correlation of the underlying distribution of the input parameters. This commenter was concerned with the impact of input parameters at the extreme ends of the distribution and does not believe that the model outputs are adequately robust beyond the 90[th] percentile of BLL predictions. The commenter suggested that model predictions be limited accordingly. 
Another commenter (EJ et al.) discussed limitations of the proposed modeling approaches and suggested that EPA review the empirical epidemiological and biomonitoring exposure studies, especially those which consider cumulative impacts of chronic exposures to low-dose waterborne lead. This commenter specifically discussed the lack of consideration of absorption factors and variability of lead leaching and lead-scale release, as well as uncertainty surrounding kinetics of lead transfer across the placenta and in breast milk. Citing EPA's draft report, the commenter concluded by stating that "deriving a household‐level benchmark solely on the basis of models with their inherent uncertainties is not only inadequate but dangerous."
 Policy Implications	
Several commenters expressed concern, summarized below, about how modeled output will be interpreted and used in practice. 
6.1	Interpretation of Modeling Output
Some commenters expressed concern with how the public will interpret a modeled health-based benchmark (EJ et al., NRDC, TCEQ, and AW). These concerns largely stem from the vague description of how model outputs might be used in risk management, and use of the term "health-based benchmark". EJ et al. acknowledged support for EPA's attempt to differentiate the concept of a benchmark derived from health effects (i.e., increase in BLLS) from the existing lead action level. However, the commenter was concerned that a modeled benchmark may create a false sense of security regarding water that tests below the benchmark in any one-time sampling. NRDC noted the same concerns and stated that "any number set above 0 parts per billion (ppb) will give the public the false sense of security that their water is safe, even if it contains lead." Two of these commenters (EJ et al. and TCEQ) encouraged EPA to remove the word "health" or revise the terminology. TCEQ stressed that the term "health-based benchmark" may unnecessarily confuse the public, and recommended that EPA "revise the terminology to accurately depict the derivation and intended use of the value, or provide the necessary documentation to justify a health-based benchmark in a separate document made available for public and peer review." EJ et al. mentioned that the term "health-based" is inherently misleading and ripe for misinterpretation and misapplication. One additional commenter (AW) stated that water systems will lack clear instruction on how to respond to questions raised by customers and the media regarding the health-based benchmark.
6.2	Use of Modeling Output
Numerous commenters cautioned EPA that any modeled numbers will likely be used as point of comparison to existing compliance data (AW, AWWA, and EJ et al.). AW found this troublesome since the current compliance sampling scheme is designed to capture worst-case scenarios, whereas comparing against a typical exposure would be a more appropriate comparison when using a health-based benchmark. AWWA noted that all available data in the U.S. have been obtained by deliberately sampling in a manner that represents higher than typical lead levels and stressed that EPA should therefore caution against direct comparison to any modeled benchmark. EJ et al. questioned how associations between lead-in-water measurements and health effects can be derived from existing lead-in-water data, since they do not accurately capture the concentrations of lead flowing from consumer taps, and requested that EPA defend the scientific validity of such extrapolations.
Related to these concerns, several commenters remarked on the variability inherent with lead exposures from drinking water and questioned how this would affect the use of a modeled benchmark (EJ et al., AWWA, and MDEQ). EJ et al. suggested that EPA consider the variability in lead release and water use patterns, and uncertainties about the presence and nature of lead‐bearing plumbing in homes, daycares, and schools. AWWA echoed these concerns and stressed that the peer review panel be fully informed about the variability of lead exposure through drinking water and the variability in both the form and level of lead present in water as a function of water age, timing of water use, water flow, and other variables.  MDEQ mentioned that the outputs presented in the draft report must reflect specifically defined point-of-use drinking water sampling methodology, and that applicable sampling methodology needs to be standardized before the appropriateness of any downstream drinking water-based health-protective benchmark can be effectively evaluated. 
