&EPA
United Stales
Environmental Pnotecnoti
Agency
Analysis Plan for Human Health and Ecological
Risk Assessment
For the Review of the Lead
National Ambient Air Quality Standards
Draft
May 31, 2006
U.S. EPA
Office of Air Quality Planning and Standards
Research Triangle Park, NC
-------
This draft scope and methods plan has been prepared by staff from the Ambient
Standards Group, Office of Air Quality Planning and Standards, U.S. Environmental Protection
Agency, in conjunction with ICF International (through Contract No. 68-DO1-052). Any
opinions, findings, conclusions, or recommendations are those of the authors and do not
necessarily reflect the reviews of the EPA or ICF International. This document is being
circulated to obtain review and comment from the Clean Air Scientific Advisory Committee
(CASAC) and the general public. Comments on this document should be addressed to Ginger
Tennant, U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards,
C539-01, Research Triangle Park, North Carolina 27711 (email: tennant.ginger@epa.gov).
-------
Table of Contents
1.0 INTRODUCTION 1
1.1 Purpose and Scope of the Assessments 2
1.1.1 Air Quality Scenarios 3
1.1.2 Exposure and Human Health Risk Assessment 3
1.1.3 Ecological Exposure and Effects Analysis 4
1.2 Background Regarding the Previous Review 4
1.3 Organization of this Document 5
2.0 CASE STUDY APPROACH 6
2.1 Primary Lead Smelter 7
2.2 Other Significant Stationary Sources 8
2.3 Near-Roadway re-entrainment 8
HUMAN EXPOSURE AND HEALTH RISK ASSESSMENT
3.0 OVERVIEW OF ANALYSIS PLAN 10
3.1 Conceptual Model for Pb Human Health Risk Assessment 10
3.2 General Approach 13
3.3 Spatial Scale of Analysis 17
3.4 Uncertainty and Variability Characterization 18
4.0 EXPOSURE ASSESSMENT 20
4.1 Estimating Media Concentrations 20
4.1.1 Ambient Air 20
4.1.2. Indoor Air 22
4.1.3 Outdoor Soil/Dust 23
4.1.4 Indoor Soil/Dust 23
4.1.5 Estimating Inhalation Exposure Concentrations 24
4.1.6 Background (Non-Air-Related) Exposure 24
4.2 Study Population(s) and Potential Stratification Based on Socioeconomic Factors 25
4.3 Estimating Blood Lead 26
4.3.1 Children 27
4.3.2 Adults 28
4.3.3 Blood Lead Metrics 28
4.3.4 Probabilistic Analysis of Exposure Parameters 29
4.3.5 Preparation of Models and Inputs, and Evaluation of Performance 30
5.0 EFFECTS ASSESSMENT 32
5.1 Endpoints to be Evaluated 32
5.2 Models for Estimating Adverse Effects 32
5.2.1 IQ Reduction in Children 32
5.2.2 Blood Pressure Effects in Adults 33
5.2.3 Renal Effects in Adults 34
DRAFT-May 31,2006
-------
6.0 RISK ASSESSMENT 36
6.1 Defining Modeled Populations in (Assigning Exposure Concentrations) 36
6.2 Risk Metrics for Children 37
6.3 Risk Metrics for Adults 38
7.0 UNCERTAINTY AND VARIABILITY ASSESSMENT 39
7.1 Integrated Approach for Considering Uncertainty and Variability 40
7.2 Risk and Exposure Metrics Generated with the Integrated Approach 42
ECOLOGICAL RISK ASSESSMENT
8.0 OVERVIEW OF ANALYSIS PLAN 44
8.1 Conceptual Model for Lead Ecological Risk Assessment 44
8.2 General Overview of Analysis 46
9.0 ECOLOGICAL RISK ANALYSIS PLAN 50
9.1 Tier 1: Screening Level Analysis 50
9.1.1 Overview 50
9.1.2 Data Sources for Determining Media Concentrations 50
9.1.3 Case Study Selection 53
9.1.4 TRIM.FaTE Model 54
9.1.5 Determination of Potential Adverse Effects 54
9.2 Tier 2: Analysis of Sensitive Receptors 57
9.2.1 Overview 57
9.2.2 Estimation of Exposure for Sensitive Receptor(s) 57
10.0 REFERENCES 58
10.1 Human Health Risk Assessment 58
10.2 Ecological Risk Assessment 63
DRAFT-May 31,2006
-------
1.0 INTRODUCTION
The U.S. Environmental Protection Agency (EPA) is presently conducting a review of the
national ambient air quality standards (NAAQS) for lead (Pb). Sections 108 and 109 of the
Clean Air Act (Act) govern the establishment and periodic review of the NAAQS. These
standards are established for pollutants that may reasonably be anticipated to endanger public
health and welfare, and whose presence in the ambient air results from numerous or diverse
mobile or stationary sources. The NAAQS are to be based on air quality criteria, which are to
accurately reflect the latest scientific knowledge useful in indicating the kind and extent of
identifiable effects on public health or welfare which may be expected from the presence of the
pollutant in ambient air. The EPA Administrator is to promulgate and periodically review, at
five-year intervals, "primary" (health-based) and "secondary" (welfare-based)1 NAAQS for such
pollutants.2 Based on periodic reviews of the air quality criteria and standards, the Administrator
is to make revisions in the criteria and standards, and promulgate any new standards, as may be
appropriate. The Act also requires that an independent scientific review committee advise the
Administrator as part of this NAAQS review process, a function now performed by the Clean Air
Scientific Advisory Committee (CASAC).
EPA's overall plan and schedule for this Pb NAAQS review is presented in the Plan for Review
of the National Ambient Air Quality Standards for Lead (EPA, 2006a), which is available at:
http://www.epa.gov/ttn/naaqs/standards/pb/data/pbreviewplan_feb2006.pdf. That plan discusses
the preparation of two key documents in the NAAQS review process: an Air Quality Criteria
Document (AQCD) and a Staff Paper. The AQCD provides a critical assessment of the latest
available scientific information upon which the NAAQS are to be based, and the Staff Paper
evaluates the policy implications of the information contained in the AQCD and presents staff
conclusions and identifies standard-setting options for the Administrator to consider in reaching
decisions on the NAAQS.3 In conjunction with preparation of the Staff Paper, staff in EPA's
Welfare effects, as defined in section 302(h) of the Act include, but are not limited to, "effects on soils, water,
crops, vegetation, man-made materials, animals, wildlife, weather, visibility and climate, damage to and
deterioration of property, and hazards to transportation, as well as effects on economic values and on personal
comfort and well-being."
Section 109(b)(l) [42 U.S.C. 7409] of the Act defines a primary standard as one "the attainment and maintenance
of which in the judgment of the Administrator, based on such criteria and allowing an adequate margin of safety, are
requisite to protect the public health." Section 109(b)(2) of the Act directs that a secondary standard is to "specify a
level of air quality the attainment and maintenance of which, in the judgment of the Administrator, based on such
criteria, is requisite to protect the public welfare from any known or anticipated adverse effects associated with the
presence of [the] pollutant in the ambient air."
3 NAAQS decisions involve consideration of the four basic elements of a standard: indicator, averaging time, form,
and level. The indicator defines the pollutant to be measured in the ambient air for the purpose of determining
compliance with the standard. The averaging time defines the time period over which air quality measurements are
to be obtained and averaged, considering evidence of effects associated with various time periods of exposure. The
form of a standard defines the air quality statistic that is to be compared to the level of the standard (i.e., an ambient
concentration of the indicator pollutant) in determining whether an area attains the standard. The form of the
standard specifies the air quality measurements that are to be used for compliance purposes (e.g., the mean average
over 90 days), the monitors from which the measurements are to be obtained (e.g., one or more population-oriented
monitors in an area), and whether the statistic is to be averaged across multiple years.
DRAFT-May 31,2006 1
-------
Office of Air Quality Planning and Standards (OAQPS) conduct various policy-relevant
assessments, including for this Pb NAAQS review, a quantitative human exposure analysis and a
human health risk assessment, as well as an ecological exposure and effects analysis.
The purpose of this analysis plan is to outline the scope, approaches, and methods that staff is
planning to use for the human health and ecological risk assessments. This plan also highlights
key issues in the estimation of human health and ecological exposure and risks posed by Pb
under existing air quality levels ("as is" exposures and health risks), upon attainment of the
current Pb NAAQS, and upon meeting various alternative standards in selected sample areas.
This analysis plan is intended to facilitate consultation with the CASAC, as well as public
review, for the purpose of obtaining advice on the overall scope, approaches, and key issues in
advance of the completion of such analyses and presentation of results in the first draft of the Pb
Staff Paper. The assessment approaches described in this plan are intended to build upon and
extend the exposure and risk assessment approaches employed for the last review,4 and on
Agency experience with Pb exposure and risk assessment since that time, while also drawing
from information presented in the December 2005 and May 2006 drafts of the AQCD for the
current review (EPA 2005a, 2006b). The final assessments will reflect information in the final
AQCD, which will be completed later this year.
1.1 Purpose and Scope of the Assessments
The focus of these assessments is the estimation of risk resulting from exposure to lead released
into ambient air. To help inform future policy considerations, this assessment will further attempt
to distinguish between Pb exposures associated with currently emitting sources (e.g., primary
and secondary Pb smelters) and historical (no longer operative) sources (e.g., emissions
associated with leaded gasoline).
This plan recognizes several distinctions with regard to the scope of the risk assessments for Pb.
First, because exposure to atmospheric Pb particles occurs not only via direct inhalation, but also
via ingestion of deposited particles, the problem being assessed is multimedia in nature, with
exposure occurring via both the inhalation and ingestion routes. And, for human health exposure
assessment of Pb, in contrast to ozone or particulate matter, the dose metric or biomarker most
commonly used and associated with health effects information is internal (i.e., blood Pb).
Additionally, the exposure duration of interest (i.e., that influencing internal dose pertinent to
health effects of interest) may span months to potentially years, as does the time scale of the
environmental processes influencing Pb deposition and fate.
There are a variety of situations involving exposure to Pb in ambient air (and Pb deposited from
air to non-air media). These include populations living in the vicinity of a primary Pb smelter,5
those in areas where multiple significant point sources are concentrated, and those in areas near
roadways where anthropogenic Pb contained in roadway dust may become suspended in ambient
air and undergo deposition and transport to non-air media. To the extent that the internal
biomarker, blood Pb, is used in exposure assessment, the separation of contributions related to
4 The last Pb NAAQS review was completed in 1991 with no revisions made to the standards at that time.
5 Currently only one primary Pb smelter (in Herculaneum, MO) is operating in the U.S..
DRAFT-May 31,2006
-------
new Pb emissions to ambient air from those reflecting inhalation and ingestion of historically
deposited ambient Pb poses a particular challenge.
1.1.1 Air Quality Scenarios
We intend to assess human exposure and health risks, and ecological risks for the following air
quality scenarios.
• Current (baseline) air quality
• Attainment of the current NAAQS (i.e., any reduced ambient air concentrations necessary
to meet the current NAAQS)
• Attainment of alternate NAAQS
As exposures associated with deposited Pb are of interest to this review, the assessment of each
of these air quality scenarios will include estimates of future Pb concentrations in other media
(e.g., outdoor soil/dust) associated with deposition.
In modeling scenarios for alternate NAAQSs, the 1990 Staff Paper, with concurrence from
CASAC, relied on the assumption that the soil Pb level would remain constant for at least six
years after a new standard would be implemented, while they projected changes in indoor air and
dust concentrations with the changing air concentration. The plan presented here includes
consideration of the use of current methods for estimating the relationship of soil Pb levels to
changes in air concentrations (i.e., projection of outdoor soil/dust concentrations for future points
in time resulting from changes in ambient air Pb reflecting specific NAAQS scenarios).
As feasible, we are considering utilizing 2005 air quality data as the basis for the baseline
scenario. We are investigating the available information on Pb emissions estimates and Pb
concentrations in other media to identify that most appropriate for use with 2005 air quality data.
1.1.2 Exposure and Human Health Risk Assessment
The planned Pb exposure analysis and health risk assessment address exposures (of multiple
years) to Pb and associated health effects. The risk assessment will characterize health risks
associated with Pb exposure pathways that include ambient air (i.e., "NAAQS-relevant") and put
those risks into context with other Pb exposure pathways that do not include ambient air, such as
indoor paint and drinking water (i.e., "background").
This assessment will focus on health endpoints associated with the range of exposures expected
to most closely reflect current levels and for which there is adequate information to develop
quantitative risk assessments. There are some health endpoints, however, for which there
currently is insufficient information to develop quantitative risk estimates. We plan to discuss
these additional health endpoints qualitatively in the Pb Staff Paper.
DRAFT-May 31,2006
-------
The risk assessment is intended as a tool that, together with other information on these health
endpoints and other health effects evaluated in the Pb AQCD and Pb Staff Paper, can aid the
Administrator in judging whether the current primary standard is adequate to protect public
health with an adequate margin of safety, or whether revisions to the standard are appropriate.
The goals of the Pb risk assessment are: (1) to characterize the risks, including estimation of the
potential magnitude of risk, associated with current Pb levels and with attaining the current Pb
NAAQS (as well as the reduced effects associated with attaining alternative Pb standards), and
(2) to develop a better understanding of the influence of various inputs and assumptions on the
risk estimates.
1.1.3 Ecological Exposure and Effects Analysis
The planned Pb ecological exposure and risk analyses address the risk to sensitive receptors from
exposure to media concentrations resulting from ambient air deposition of Pb over time. In the
first tier of this analysis, we will attempt to describe the current media Pb concentrations in the
environment and model what those concentrations are likely to be in the future. These future
concentrations can then be compared to screening values for effect in soil, water, and sediment to
focus the analysis further on those receptors that may be most sensitive to Pb. The second tier of
the analysis will focus more detailed modeling on those receptors which seem most likely to be
affected to determine intake values or body/tissue concentrations which can then be compared to
available concentration effects data.
1.2 Background Regarding the Previous Review
The current primary and secondary standard for Pb is 1.5 micrograms per cubic meter (|ig/m3) in
total suspended particulate matter (TSP), as the maximum arithmetic mean averaged over a
calendar quarter.
As part of its last review, EPA's Office of Air Quality Planning and Standards (OAQPS)
performed an exposure assessment to estimate blood Pb levels among populations exposed under
the current NAAQS, as well as alternate Pb regulations in the future (EPA 1989). For the
different regulatory scenarios, the focus of the quantitative risk assessment was primarily on
estimating the percentages of children with blood Pb levels above 10 and above 15 |ig/dL. The
available data at that time indicated impaired neurobehavioral function and development
associated with levels of 10-15 ug/d. The available data did not indicate a clear threshold at this
range, rather suggesting a continuum of health risks down to the lowest levels measured,
although effects below this range were increasingly difficult to detect and their significance more
difficult to determine (EPA 1990). The last risk assessment also developed estimates of the
percentages of middle-aged men with blood Pb levels above 10 and 12 |ig/dL. The dose-response
information on blood pressure changes available at the time of the last review was less clear than
the information on children, however, the same approximate range of blood Pb levels as for
children were also considered for assessing risks among adult men.
The consideration of environmental effects during the last Pb NAAQS review did not include a
quantitative risk assessment (EPA 1990). Rather, critical aspects of the evidence regarding Pb
DRAFT-May 31,2006 4
-------
effects on terrestrial and aquatic ecosystems that was presented in the AQCD (EPA 1986) were
summarized to provide a basis for staff conclusions and recommendations. Based on this
information, EPA concluded that, under ambient Pb exposure conditions existing at the time, the
primary standard provided adequate protection to the environment.
1.3 Organization of this Document
The remainder of this document is generally organized into four parts. The first includes a
discussion of the case study approach (Section 2), which, like the discussion of air quality
scenarios in Section 1.1.1, pertains to both the human health and ecological risk assessments.
Next is the plan for the human exposure and health risk assessment (Sections 3 through 7),
followed by the plan for the ecological risk assessment (Sections 8 through 9). References for
both plans are provided in Section 10. While separate plans are described for the ecological and
human health risk assessments, the assessments will draw, to a large extent, from common
environmental characterizations (e.g., air quality scenarios with associated media concentrations,
and some case studies).
DRAFT-May 31,2006
-------
2.0 CASE STUDY APPROACH
A case study approach will be used for both the human exposure and risk assessment (described
in Sections 3 through 7) and the ecological risk assessment (described in Sections 8 and 9).
Some case studies will be "shared" by the two assessments (e.g., with one characterization of
environmental conditions serving the purposes of both assessments), while we plan to also have
some cases studies that are used only by either the human health or ecological risk assessment.
A description of such a case study that is particular to ecological risk assessment is discussed in
Chapter 9. The discussion provided here is intended to address case studies common to both
types of assessment as well as those that may serve only the purposes of the human health
assessment.
The general strategy for assessment of each case study will consider characteristics of the
source(s) of interest, as well as the availability of measurement and source characterization data.
Selection of case study locations will reflect the following factors:
• Variety of ambient air sources ofPb emissions: Pb emissions sources ranging from the
primary Pb smelter, to other significant stationary sources (e.g., metal refinishing and
ceramics manufacturing) and emitters of historically deposited Pb (e.g., urban roadways)
will be considered. In addition to the magnitude of Pb emissions, other attributes related
to emissions that influence exposure (e.g., effective stack height, particle size profiles,
etc.) will be considered.
• Available monitoring data for ambient air: As discussed below, characterization of
ambient air concentrations and soil deposition will rely on (a) monitoring data (if a case
study location has sufficient monitor coverage) and/or (b) dispersion modeling,
supported by monitoring data, particularly those that are appropriate for source-
apportionment. Consequently, selection of case studies may favor locations which have
local Pb monitoring data that can be source-apportioned. In addition to national Pb
monitoring networks (see Chapter 3 of the AQCD), state and local sources will be
considered.
• Available soil monitoring data and, for purposes of the human health assessment,
biomonitoring (bloodPb) data: When available, measurements of Pb soil concentrations
for a study area together with blood Pb levels for the residential population can
considerably increase confidence levels associated with characterizing human health risk
associated with baseline conditions. Therefore, locations where relevant biomonitoring
and/or collection of Pb soil concentrations has been conducted will be favored in the
selection of case study locations. Note, that such data heavily influenced by sources
unrelated to air (e.g., Pb paint) will be less useful for our purposes.
• For purposes of the human health assessment, socioeconomic status (SES) and other
demographic attributes related to Pb exposure and risk: If specific demographic and/or
SES factors are linked quantitatively to increased exposure or concentration-response for
Pb effects being modeled, then those factors may also be considered in selecting case
study locations. Potential factors currently under consideration include: (a) general
DRAFT-May 31,2006 6
-------
population density (would favor locations with higher population densities since will
have larger modeled population and more fully characterized population risk
distributions), (b) housing age (for background exposure levels to Pb paint), and (c) SES
(race and income) to inform estimation of risk characteristics (e.g., baseline blood
pressure levels, which could result in greater blood pressure impacts following Pb
exposure).
• Available water quality and sediment monitoring data, for purposes of ecological risk
assessment: Locations which include the presence of a surface water body for which
measurements of Pb are available for water and sediment will be favored in selection of
case study locations.
We also plan to select case studies that reflect particular types of ambient Pb emissions and
exposure situations, such as the following:
• Primary lead smelter. This case study would be expected to reflect Pb contamination of
the surrounding study area which is dominated by emissions from this facility, with other
sources being of relatively lesser importance.
• Other significant stationary sources: This category of case study would reflect a location
impacted by Pb emissions from multiple point and area sources, with emissions from any
particular source being significantly lower than those for the primary Pb smelter, but
where combined impacts from these multiple sources (including re-entrainment of
historically deposited near-roadway Pb) may be significant.
• Near roadway re-entrainment: This category of case study reflects environmental Pb
contamination dominated by re-emission of historically deposited Pb near roadways (i.e.,
re-entrainment of near-roadway Pb). This may include consideration of a general urban
scenario with heavy road traffic, historic dust (e.g., reflecting old housing sources) and
proximal residential areas, but without large local sources of industrial Pb emissions.
It is anticipated that one or more locations will be modeled for each of the above categories of
case studies described in the following section, the final number for each category will reflect
consideration of a number of factors including a desire to characterize population risk variability
associated with these situations within the US. The following subsection focuses on distinctions
among these categories with regard to the approaches we intend to employ to characterize
exposure and risk.
2.1 Primary Lead Smelter
There is presently only one primary Pb smelter operating in the U.S. Source characterization
information is available from past regulatory analyses, EPA Regional Office permit files, and
state/local offices. The Staff anticipates that there will be more available site-specific ambient
air, soil, and blood Pb level measurement data for areas near this facility than for any other
facility in the U.S. As a result, it is expected that measurement data (e.g., ambient air monitoring
DRAFT-May 31,2006
-------
data and soil sampling data) will play a major role in characterizing exposures and risks
associated with this facility.
The primary Pb smelter case study is anticipated to have the most comprehensive site-specific
blood Pb monitoring data among the three categories of case studies, a feature relevant to the
human health assessment. This may support the site-specific characterization of blood Pb levels
for key target populations. These data will also be useful in evaluating performance of the blood
Pb modeling approach applied to this case study location. For example, we may be able to
compare model predicted blood Pb distributions for the study area to site-specific blood Pb data
in order to (a) evaluate the overall representativeness of the blood Pb modeling approach and (b)
guide refinement of that modeling approach as appropriate.
2.2 Other Significant Stationary Sources
As described in the Pb AQCD (EPA 2006b), there are numerous stationary sources of Pb to air,
including secondary Pb smelters, metal finishing facilities, and ceramic manufacturers.
Assessments for this category of case study will rely on locations and emissions characterization
information for these facilities from EPA's National Emissions Inventory (NEI), supplemented
by available data from Regional Office permit files, emissions standards (MACT and residual
risk) files, and state/local offices. This case study category will address co-location of these
types of facilities to the extent feasible. We anticipate that air, soil and blood Pb measurements
in locations for this category of case studies may be limited relative to site-specific data available
for the primary Pb smelter, necessitating a greater reliance on modeling to estimate
environmental concentrations (as described in sections 4.1.2 and 4.1.3).
Because this case study category is expected to have limited (or no) site-specific blood Pb
monitoring data, the degree to which blood Pb modeling performance can be evaluated will be
limited. Therefore, more general regional or national-scale blood Pb data obtained form
NHANES (possibly differentiated by relevant SES factors) may be used for performance
evaluation. Note, however, that locations for which local blood Pb data are available will be
given preference in selection.
2.3 Near-Roadway re-entrainment
The use of leaded gasoline in the U.S. resulted in accumulation of Pb in soils near roadways and
the resuspension of this Pb may have the potential to contribute significantly to concentrations of
airborne Pb (Harris et al. 2005). Besides historic combustion, there are other auto-related
releases of Pb near roadways (see EPA 2006b). Measurements of concentrations near some
roadways have been reported (Filippelli et al. 2005; Teichman et al. 1993). This case study
category is intended to provide some characterization of associated exposures and risks for this
type of air emissions source. Consequently, selection of locations will favor areas that are not
significantly influenced by other types of air sources.
To develop estimates of associated media concentrations for this case study category, we intend
to consider the available literature, as well as potential use of modeling tools, ambient
measurements and source apportionment methods (discussed in Section 4.1.1 and 4.1.2). The
DRAFT-May 31,2006 8
-------
use of multiple methods may inform characterization of uncertainty in estimated concentrations.
As with the last case study category discussed, locations for which blood Pb measurements are
available will be favored. As available such data will be used, in preference to NHANES data,
and in conjunction with socioeconomic and demographics data, to evaluate blood Pb model
performance.
DRAFT-May 31,2006
-------
HUMAN EXPOSURE AND HEALTH RISK ASSESSMENT
3.0 OVERVIEW OF ANALYSIS PLAN
This risk assessment is designed to estimate various health effects associated with current, and
some alternate, exposures to Pb emitted into ambient air in the U.S. This includes new emissions
of Pb into ambient air as well as emissions of historically deposited Pb. As recognized in Section
1.1, the exposure assessment will not only include inhalation but also the ingestion route (e.g.,
incidental ingestion of ambient Pb deposited to residential soils and that transported indoors). In
addition to ambient air associated Pb, exposure to Pb from other (i.e., "background") sources will
be assessed. "Background" for the purposes of this assessment refers to sources, pathways and
exposures that do not involve ambient air (e.g., Pb in drinking water and Pb paint-related dust).
As described in Section 2.0, we will use a case study approach wherein a small set of locations
intended to illustrate the variety of exposure conditions associated with ambient air associated Pb
that occur in the U.S. This assessment will focus on health endpoints, associated with the range
of exposures expected to most closely reflect current levels, and for which there is adequate
information to develop quantitative risk assessments. For example, we intend to assess
neurological effects in young children, and we are also considering other endpoints for adults. In
the risk characterization, we will characterize the magnitude and distribution of Pb-related risk
within the case study populations.
The risk assessment will be completed in two stages. The first stage, referred to as the Pilot, will
be completed by the end of this year and will be described in the first draft of the exposure/risk
report and summarized in the first draft of the Staff Paper, both of which will undergo CAS AC
review. We anticipate including 3-4 case studies for the Pilot, with a primary focus being on
testing and refining the assessment approaches. The methodology for the Full-Scale assessment
will build on our experience with and CASAC comments on the Pilot assessment. The Full-
Scale risk assessment will be described in the second draft exposure/risk report, which we intend
to complete in mid-2007. We anticipate expanding the number of case studies for the Full-Scale
analysis (e.g., 5-10).
The following subsections provide an overview of the analysis plan. Section 3.1 presents the
conceptual model for Pb human health risk assessment, describing both the elements generally
pertinent to the assessment of health risk associated with Pb, and identifying those elements to be
included and a focus of this assessment. Section 3.2 expands on the elements of this assessment,
although with a grouping of conceptual model elements that reflects basic steps of the general
approach for this assessment. Section 3.3 describes issues related to spatial scale and the
definition of the GIS-based spatial template used as the basis for modeling in the analysis.
Lastly, Section 3.4 generally describes how we will characterize uncertainty and variability
associated with those steps.
3.1 Conceptual Model for Pb Human Health Risk Assessment
This section presents the conceptual model (Figure 3-1) intended to illustrate the elements
pertinent to assessment of public health risks associated with environmental Pb exposures.
DRAFT-May 31,2006 10
-------
For purposes of this risk assessment, "background" is intended to refer to sources of, exposures
to, etc, Pb associated with pathways that do not involve ambient air. Included among these would
be pathways associated with indoor Pb paint, Pb in drinking water, Pb in food, etc. Background
elements are shown in Figure 3-1 in non-bold (regular) type. As shown in Figure 3-1, the
assessment will include contributions from all sources. We intend, however, to characterize risks
associated with ambient air associated sources separately from background sources.
Sources: The focus of this review is on sources to ambient air (e.g., current sources of new Pb
emissions and re-emission of historically deposited Pb). Other ("background") sources of
environmental Pb that are significant contributors to population Pb exposures (e.g., indoor Pb
paint) will also be considered in estimating blood Pb levels associated with all sources.
Pathways: Figure 3-1 is intended to generally illustrate the many pathways by which Pb emitted
into the environment becomes available for human exposure. Those not passing through ambient
air are considered "background" for the purposes of this assessment. For those case studies in
which modeling is employed for exposure assessment (in lieu of adequate measurements), those
pathways shown with white background in Figure 3-1 will not be explicitly modeled.
Routes: The ingestion and inhalation routes are considered the primary routes of human
exposures to environmental Pb, and the ingestion route (e.g., inclusion of incidental ingestion) is
expected to be the more significant. Both routes will be included in this assessment.
Exposed Populations: The Pb exposed populations can be characterized and subset based on a
variety of characteristics. Figure 3-1 identifies groups based primarily on age or lifestage, which
has an influence on behaviors that can influence exposure or susceptibility. It is recognized that
more specific factors (e.g., calcium deficiency, to name one) also influence susceptibility. Such
characteristics will be addressed as feasible in the assessment, given limitations of the currently
available information (e.g., differing exposure/dose - response functions). Where quantitative
assessment of particular subgroups (e.g., children with calcium deficiencies) may not be feasible,
qualitative characterization will be important. In this assessment, we intend to quantify risks to
young children, and are considering also quantifying risks for adult populations.
Internal Disposition: While Pb is distributed throughout the body, bone is an established site of
internal accumulation of Pb, while blood is an established internal dose metric for purposes of
both exposure and risk assessment. This risk assessment will rely primarily on blood Pb with
corresponding dose-response functions. The tools employed will recognize the role of bone as a
reservoir with the potential to act as source and storage site.
Endpoints: Figure 3-1 generally identifies the wide variety of health endpoints recognized in
the draft AQCD (EPA, 2006b) as associated with Pb exposures. As mentioned previously, the
endpoints of interest for this assessment are those associated with the range of exposures
expected to most closely reflect current levels, and for which there is adequate information to
develop quantitative risk assessments. Recognizing that, the primary endpoint we currently
intend to include in the assessment (indicated in Figure 3-1 via bold outlined box) is neurological
effects in children, while we are considering the additional endpoints of cardiovascular and renal
effects in adults.
DRAFT-May 31,2006 11
-------
Figure 3-1. Conceptual Model for Pb Human Health Risk Assessment
LU
o:
o
Q_
X
LU
O
LU
LU
{£.
UJ O
I- Q.
Z CO
Q
Stationary and
Mobile Sources
1
Soil/Dust -
Outdoors2
REENTR
\
MNM
\
Otr
Miscellc
Sou
\
ENT
Ambient Air
1
DEPO,
1 Air - Indoor 1
1 r 1
. ^
3ITION DEPO,
, ,
Soil/Dust -
Outdoors
|
Soil/Dust -
Indoors
'
Air - 1
Inhalation
'
ndoor
3ITION DEPCX
, ,
3ITION
k ,
1 Soil/Dust- Soil/Dust- 1
Outdoors Outdoors 1
i
i
r
Air - Indoor 1
DEPO
J
Soil/Dust- 1
Indoors 1
' i
3ITION
L
Soil/Dust- 1
Indoors 1
' i
r
1
Diet
Drinking
Water
Lead Paint -
Indoors
Soil/Dust - Indoors
ngestion
Young
Children
Adult
Men
Cardiovascular
System
(for adults)
Kidney
(for adults)
Other
(genotoxic effects/cancers,
reproductive/developmental effects,
blood effects, immune system effects,
bone and teeth effects)
Blood Pressure
Change
Hypertension
Incidence
Creatinine
Clearance
Decrement
NOTES
1 Components with gray background and solid borders will be included in this assessment. Components depicted in gray
but without borders are being considered for inclusion in the assessment. Further, a distinction is made between
components linked to exposure pathways involving ambient air (shown in bold) and components involving other pathways
(i.e., background).
2 Includes contributions of historical sources, including (but not limited to) emissions from the use of leaded gasoline,
historical emissions from stationary sources, and exterior leaded paints.
DRAFT-May 31,2006
12
-------
Metrics: Figure 3-1 generally recognizes that there are many metrics that might be considered
for risk assessment. Recognizing the need for the metrics used in this assessment to have
sufficient support for use in quantifying population health risk, we currently intend to use IQ
decrement in children, and are considering the inclusion of blood pressure and kidney effects in
adults. We will use these metrics to characterize the magnitude and distribution of Pb-related
risk within the exposed populations.
3.2 General Approach
This section presents the general approach for the exposure and risk assessment methodology for
generation of population-level risk estimates.
The specific approach used to estimate exposure and risk for individual case studies is expected
to differ depending on the amount of site-specific monitoring data available for characterizing Pb
concentrations in ambient air and outdoor soil/dust. For example, where there is adequate
monitoring coverage, we will rely primarily on measurements to establish baseline conditions for
both ambient air and soil Pb concentrations. By contrast, for those case study sites lacking in
ambient monitoring data, we will depend primarily on modeling to characterize baseline soil and
ambient air Pb conditions. In all cases, future projections of soil Pb concentrations (and related
media concentrations such as indoor dust), designed to reflect conditions under various NAAQS
scenarios, will rely on modeling.
The general approach for a case study location at which modeling is used to characterize baseline
ambient air Pb concentrations or soil Pb concentrations is illustrated in Figure 3-2. This
approach is divided into four components, the first three of which comprise steps in the exposure
assessment (intended to cover the components indicated in bold in the conceptual model, Figure
3-1) and the last of which represents the effects and risk assessment. The four components are:
(a) estimating ambient air concentrations and deposition to soil, (b) estimating soil and indoor
dust concentrations, (c) estimating blood Pb levels and (d) health effects incidence estimation
(risk characterization). Each of these components is discussed below. This discussion also
highlights ways in which the fully-modeled approach presented in Figure 3-2 would be modified
for those case studies where there are sufficient ambient air and/or soil measurements to support
characterization of baseline conditions without or with lesser use of modeling. The approach for
characterizing uncertainty and variability associated with the steps in this approach, while not
illustrated in Figure 3-2, is described in section 3.4.
Estimating Ambient Air Concentrations and Deposition to Soil
We intend to rely on air monitoring data, where available, in preference to modeling, to
characterize baseline conditions across the study area. For those case studies lacking sufficient
ambient air monitoring data to characterize baseline conditions across the study area, we intend
to model the dispersion of air emissions in order to characterize the spatial and temporal
distribution of ambient air concentrations across the study area, and also to estimate Pb
deposition to soil. If a case study area has some ambient monitors (but not enough to fully
characterize baseline conditions), then we will consider source-apportioning those monitoring
data and using the results to performance-evaluate the ambient air concentration surface
DRAFT-May 31,2006 13
-------
Exhibit 3-2. Overview of Analysis Approach relying on Modeling. Where
measurement data are available and sufficient, they will be used in preference to, or in
combination with, modeling steps shown here to characterize baseline conditions. The
lower two boxes are to be implemented for each population and endpoint assessed.
CO
•z.
o
p
o
o
o
o
LLJ
S
O
\
LLJ
1
LLJ
CO
CO
<
LLJ
K
§
Q_
X
LLJ
N
£
LLJ
O
O
CO
Characterizing Ambient Air Levels
and Deposition to Soil
Estimating
re-entrained
lead (near
road way)
Tj
Emissions from
current sources
f-1
oersion modeling for
i emissions sources
lent air
ring data
|
1
Soil
profiles
Source
apportionment
modeling
(of monitored levels)
1
Deposition rates
across study area
Ambient air
concentrations for
study area
Estimating Soil and
Indoor Dust
Concentrations
Estimating
residential soil
concentrations
Used to
evaluate
performance of
source
modeling and
characterize
uncertainty
E PA'S 1999 NATA
National Scale
Assessment
. concentrations
Estimating indoor
residential lead
dust levels
Indoor residential dust j
;| concentrations j
ZZZZZCZZZZ!
Estimating Blood Lead Levels
Background levels for
additional lead
exposures
(drinking water, diet)
~L
L
Pathway-specific
exposure rates
(where relevant)
Blood lead level modeling
(children and adults)
Ambient air
concentrations
(see above)
Calibration of
blood lead level
models using
NHANES data
and additional
data sources
Health Effects
Incidence
Estimation (Risk
Characterization)
Demographics
Health effects
incidence estimation
Health effects incidence estimates
for each endpoint/study population
combination
Concentration-response
functions for key endpoints:
• IQ (child)
• Blood pressure (adults)
Key:
Input dataset
Modeling step
| Modeled output
DRAFT-May 31,2006
14
-------
projected by modeling. If the modeled ambient air surface differs significantly from the source-
apportioned monitor values (in the vicinity of the monitor) then key elements of the dispersion
modeling, including emissions profiles for modeled sources, would be reevaluated in order to
reduce uncertainty and improve model performance.
For those case studies with sufficient monitor coverage for the study area, we are considering
using those monitor data to characterize the spatial and temporal distribution of ambient air Pb
concentrations across the study area without relying on dispersion modeling. We are
additionally considering the use of source-apportionment to determine which fraction of each
monitored concentration is associated with a particular source grouping in order to obtain
separate estimates of ambient air concentration for different source groupings of interest. In
these case study locations, modeling will still be required to project deposition to soil for
characterization of future conditions.
Estimating Soil and Indoor Dust Concentrations
We intend to rely on soil measurements, where available, in preference to modeling, to
characterize baseline conditions across the study area. For those case study locations in areas
impacted by both emissions of new Pb as well as historically deposited Pb arising from
combustion of leaded fuels, but lacking sufficient soil concentration data to characterize baseline
conditions, a hybrid modeling approach which uses background soil Pb data together with multi-
media fate and transport modeling is being considered to characterize baseline soil Pb
concentrations. This hybrid approach has two steps:
• Estimation of soil concentrations associated with historically deposited Pb associated
with historical leaded automobile emissions: Soil Pb monitoring data from locations
similar to the case study area but not impacted by significant ongoing Pb emissions will
be used to represent Pb concentrations from older (primarily auto emissions) deposition.
• Deposition modeling and application of a soil reservoir model to estimate soil Pbfor
current/ongoing emissions sources: The contribution of ongoing local point source and
mobile emissions to soil concentrations will be modeled by using the deposition rates
from ongoing (current) air emissions, to predict loading to soil and translating those to
soil concentrations with a simple reservoir model.
The combination of these two estimates is intended to produce an estimate of baseline soil Pb
concentrations that reflects both older (e.g., primarily automobile-related deposition) and more
current loading related to local sources of interest (e.g., secondary smelters). As with ambient air,
future soil concentrations will be projected using modeling to reflect any changes in soil Pb
concentrations resulting from deposition of Pb under alternate NAAQS scenarios.
We are considering estimating indoor dust concentrations attributable to outdoor air and soil/dust
based on empirical relationships of Pb concentrations in outdoor soil and ambient air to indoor
dust concentrations. For residences impacted by Pb paint, this approach would estimate total
indoor dust concentrations by combining these modeled concentrations with existing estimates of
DRAFT-May 31,2006 15
-------
Pb concentrations in indoor dust impacted by Pb-based paint, taking into account the variability
in indoor dust concentrations based on the presence of Pb paint.
Characterization of baseline conditions for locations with sufficient soil monitoring data will be
simpler than the modeled approach described above, since these monitor data reflect the
aggregate impact of older and ongoing (current) deposition, and can be used directly. For
projection of future soil concentrations for these locations, however, we will still rely on
modeling to project changes in soil concentrations associated with reductions in ambient air
concentrations and deposition of Pb.
Estimating Blood Pb Levels
Exposures to media containing Pb for the populations of interest will be quantified using blood
Pb concentration as the exposure metric. For example, we plan to use the IEUBK model to
estimate blood levels for children less than 7 years of age. Other models being considered are
described in Section 4.3. Exposure to air, outdoor soil, and indoor dust will be estimated from
the concentrations described above. These exposures, along with estimates of exposures from
other pathways (e.g., drinking water, diet), will be provided as inputs to the blood Pb models.
Any and all models used will be subjected to performance evaluation to ensure that: (1) they are
being applied appropriately; (2) assumptions related to Pb intake and uptake (where they are
calculated) are consistent across models; and (3) the models, as applied, produce estimated blood
Pb levels that are reasonably consistent with their underlying basis and previous evaluations, and
with available data sets (e.g., NHANES).
In modeling exposure for children, we are planning to use IEUBK as typically applied in the
regulatory context, including the use of a GSD to provide coverage for inter-individual
variability in both exposure and biokinetics. However, we are also considering an alternative
approach of using probabilistic simulation to model inter-individual variability in exposure
outside of IEUBK and then running those simulated individuals through IEUBK in batch-mode.
Inter-individual variability in biokinetics might then be addressed using a purpose-defined GSD.
A similar approach is being considered for the adult age group, should risk for that group be
modeled.
Estimating Health Effects Incidence (Risk Characterization)
From the estimated blood Pb levels, we will estimate distributions of health effects for the
populations of interest using concentration-response functions for the chosen endpoints. Risk
estimates will be generated by applying the blood-Pb adverse effect relationships to the blood Pb
distributions derived as described above. For the Pilot analysis, we are planning to model IQ
loss for children and are considering the option of modeling of blood pressure changes and renal
effects for adults. In the case of children (IQ loss), we are planning to use concentration-
response functions derived from the Lanphear et al. (2005) study, including nonlinear models
that predict higher slopes at lower blood Pb levels. While Lanphear (2005) represents our
preferred study, we are also considering including a range of studies to provide coverage for
uncertainty in the modeling of this endpoint. For adults, we are considering the use of the
Nawrot et al. (2002) meta-analysis as the basis for a concentration-response function for blood-
DRAFT-May 31,2006 16
-------
pressure effects, while for renal effects we have not identified a single preferred study and
instead, are considering a range of studies.
The focus of this analysis is on characterizing population-level risk for modeled study
populations (i.e., characterization the distribution of risk levels across a particular study
population). Once these distributions have been developed, we then plan to generate several
types of population-level risk metrics from these distributions including: (a) the degree of risk
(e.g., IQ loss) experienced by individuals at various percentiles on the risk distribution and (b)
the number of simulated individuals falling within specific risk ranges (e.g., number of children
predicted to have 1-2 IQ point loss). We are planning to generate risk metrics that will specify
the degree of risk attributable to NAAQS-relevant emissions sources at a particular case study
location, relative to risk associated with background exposures. As noted earlier, for the Pilot
analysis, we are also planning to generate risk metrics for both current conditions and any
reductions in air concentrations associated with attainment of the current NAAQS standard.
Estimation of risk reductions associated with attaining alternative standards would be reserved
for the Full Scale risk analysis.
3.3 Spatial Scale of Analysis
To capture the spatial variability in exposure media concentrations, a "GIS-based spatial
template" will be created for each case study location that defines the outer boundary of the
study area and determines the level of refinement to be incorporated in subdividing the area into
discrete spatial units.
Defining the Study Area Boundary
The overall size of the area to be modeled will depend on the specific attributes of each case
study (see Section 2.0). For case studies where transport of Pb through the air beyond the
immediate area surrounding the facility of interest, the study area size may extend out to 50 km
from the primary source. For a case study that includes multiple contributing sources, the study
area size may extend up to 50 km from the source with the highest emissions. A maximum
radius of 50 km is being considered because application of Gaussian plume dispersion models
(e.g., AERMOD, ISC) are not recommended beyond this distance. This radius is expected to
capture the area over which the impact of the source is distinguishable from background
concentrations. A smaller radius may be selected for case studies with lower emissions levels
where estimated media concentrations are close to background.
Refinement within the Study Area
The level of spatial refinement within the modeling region will depend on the available data and
modeling methods that are employed. As appropriate for the case study type (see Section 2.0),
media concentrations will be estimated for spatial units that adequately represent the observed or
expected variability in media concentrations. Consequently, the level of spatial refinement may
vary by media.
DRAFT-May 31,2006 17
-------
For case studies employing air dispersion modeling, modeled air concentration and deposition
estimates will be generated at grid points designed to capture variability in media concentrations
across the study area (taking into account the locations of any monitoring stations present in the
case study area). The resolution of the grid may vary depending on the case study and within a
case study (e.g., dense near the source and less dense with greater distance from the source).
The grid for each case study will also be designed taking into account locations of measured air
concentrations and soil concentrations in order to optimize comparisons with and utilization of
monitoring data. At a minimum, the grid will be designed to allow for the calculation of media
concentrations at the resolution of the demographic data being used.
To select the appropriate resolution for demographic data, the staff considered the expected
spatial variability in media concentrations and the availability of demographic data at different
resolutions from the U.S. Census. Based on these considerations, U.S. Census block groups are
being considered as the demographic spatial units because they provide the required
demographic information and sufficient resolution to capture spatial variability in media
concentrations (including "hot spots"). In many locations, Census tracts would be too coarse to
capture spatial variability, and Census blocks, while providing even more resolution than block
groups, offer only limited demographic data and would require increased computational
resources (see Section 4.2 for additional detail on study populations).
3.4 Uncertainty and Variability Characterization
For the Pilot analysis, we are considering the use of an integrated approach for addressing both
uncertainty and variability, which will combine probabilistic simulation (for addressing
exposure-related variability) with sensitivity analysis techniques (for addressing parameter and
model uncertainty). At this stage, it is not possible to address uncertainly using probabilistic
simulation due to data limitations, necessitating the need for a sensitivity-based approach that
characterizes the range of risk results reflecting specific sources of model and parameter
uncertainty.
The integrated approach will use a modeling options "tree" to represent model uncertainty, with
each branch on the tree representing a distinct combination of modeling options for the analysis
(a number of the modeling steps in the analysis are subject to model uncertainty, leading to
multiple modeling options at those nodes). For the Pilot analysis, we are considering the option
of identifying the high-bound (Max), low-bound (Min) and central tendency modeling branches,
these reflecting, respectively, the combination of model options that produce the highest risk,
lowest risk and central-tendency risk for a given age-group/endpoint combination. We would
then examine those three modeling branches (Max, Min and central tendency) in greater detail,
by considering (a) parameter uncertainty through the application of various sensitivity analysis
techniques and (b) exposure parameter-related variability using probabilistic simulation.
This integrated approach, will generate risk metrics that reflect the distribution of risk across
modeled age group/endpoint combinations resulting from exposure-related variability. With
regard to uncertainty, this approach will show the range of potential risk results resulting from
uncertainty in key model steps and input parameters. However, as noted above, we will not be
DRAFT-May 31,2006 18
-------
producing confidence intervals for population risk distributions since uncertainty is being
addressed with sensitivity analysis techniques and not through probabilistic simulation.
An expanded discussion of methods used in the Pilot Analysis to address both variability and
uncertainty is included in Section 7.0.
DRAFT-May 31,2006 19
-------
4.0 EXPOSURE ASSESSMENT
4.1 Estimating Media Concentrations
This section describes the approaches we intend to use and others we are considering for
estimating Pb concentrations in the exposure media of interest for this analysis. In order to
model the exposure pathways outlined in Figure 3-1, we need to estimate Pb concentrations in
the following media:
• Ambient air;
• Indoor air;
• Soil/dust - outdoors; and
• Soil/dust - indoors.
Our needs for characterizing media concentrations for this analysis can be divided broadly into
(a) estimating baseline conditions (for all relevant media) and (b) estimating future conditions
(for all media) reflecting a particular NAAQS scenario. The extent to which empirical data will
be used relative to modeling in establishing baseline conditions for a particular case study
location will depend on the degree of coverage provided by available monitoring data. In some
instances, it may be possible to establish baseline conditions largely using measured data for
ambient air, soil and even indoor dust, while other case study locations may require modeling to
characterize baseline conditions, with measured data being used largely for performance
assessment of those modeled concentrations. And there may be locations for which a
combination is employed. By contrast, future conditions for ambient air will be estimated either
by (a) adjusting air concentration surfaces to reduce air concentrations in areas exceeding a
particular NAAQS standard or (b) conducting modeling with reduced emissions estimates in
order to achieve a modeled surface without exceedances of a particular standard. Future
conditions for other media besides ambient air (e.g., outdoor soil, indoor dust) will be estimated
through modeling.
The remainder of this section begins with a brief discussion of the general approach for
characterizing spatial variability in demographics and media concentrations using a GIS-based
spatial template, followed by descriptions of the approaches being considered for estimating
media concentrations in each medium of interest.
4.1.1 Ambient Air
As described in Section 3.2, the approach used to characterize ambient air Pb concentrations for
a particular case study will depend on the sufficiency of the monitoring data available. For
current conditions, we will rely upon air monitoring data, as feasible, to estimate ambient air
concentrations. As described in the draft AQCD (EPA, 2006b) there are a number of monitoring
networks that may be useful for this purpose, and we will also investigate the availability of
additional data in particular locations (e.g., near sources of interest). We will augment this
approach with modeling as needed for current and for future conditions.
DRAFT-May 31,2006 20
-------
We have identified a number of modeling tools which may be combined with ambient
monitoring data in order to characterize air concentrations resulting from different emissions
sources (i.e., current direct releases and reentrainment). These tools include:
• Air quality modeling: We intend to rely on air quality models (e.g., AERMOD, ISC, and
CALPUFF)6 for modeling ambient air concentrations associated with sources of new Pb
emissions (e.g., ongoing stationary and mobile sources), as well as re-emission of
historically deposited Pb (i.e., re-entrainment of Pb associated with soil/dust).
• Source apportionment modeling: We are considering the use of source apportionment
models for two purposes: 1) to identify contributions of different source types to
measured air concentrations (e.g., to apportion ambient monitor Pb concentrations
between specific industrial point sources and reentrainment), and 2) to evaluate model
performance and guide refinement of air quality modeling (i.e., compare modeled
surfaces in the vicinity of a monitor to the relevant source-apportioned component of that
monitor to determine whether modeling seems representative and if not, then reexamine
elements of modeling including emissions profiles). Specific source apportionment
methods being considered include: (a) Chemical Mass Balance (CMB) and (b) Positive
Matrix Factorization (PMF). In addition, we will consider using source-apportioned
results conducted for PM (from the literature) and then applying soil/dust Pb
concentration values to convert those PM-related sourced-apportioned data into Pb-
specific estimates.
• Reentrainment modeling: To estimate re-emissions of historically deposited Pb (for use
in the air quality modeling), we are considering the following tools: (a) particulate
emission factors (PEFs) as described in the Superfund Soil Screening Guidance (EPA,
1996a) and EPA's AP-42 emission factor documentation and (b) Wind Erosion
Prediction System (WEPS) which is a dust resuspension model developed by the USDA
Agricultural Research Service.
• Use of National Air Toxics Assessment-national scale assessment (NATA-nsa): Another
approach for developing estimates of re-emissions of historically deposited Pb that we
are considering involves comparison of the ambient concentrations predicted in the most
recent NATA-nsa7 to ambient monitoring data. As the NATA-nsa estimates are
developed from dispersion modeling of emissions from the National Emissions
Inventory (NEI), which does not include reentrainment of road and soil dust, the delta,
which may result form several factors including reentrainment, may provide a
perspective on the contribution of that source type to ambient air concentrations.
6 AERMOD and ISC (Industrial Source Complex) are Gaussian plume dispersion models, with AERMOD being the
more current model. CalPuff, another refined model, is a puff dispersion model.
http://www.epa.gov/scram001/dispersionjrefrec.htnrfaermod
7 EPA's NATA-nsa is a national scale assessment of air toxics emissions, ambient concentrations, inhalation
exposures, and human health risk performed on a triennial basis, http://www.epa.gov/ttn/atw/natal999/ The most
recently available assessment is based on the 1999 NEI.
DRAFT-May 31,2006 21
-------
• Literature-based re-entrainment estimates: We intend to consider, as appropriate,
information in the literature in developing estimates of emissions of historically
deposited Pb , e.g., that described and cited in the draft AQCD (EPA 2006b).
Figure 3-3 illustrates an application of tools described above to characterize baseline ambient air
concentrations (e.g., a "fully modeled approach"). In this example, both point/mobile source
emissions and re-entrainment of historically deposited Pb are modeled explicitly. As feasible,
performance evaluation of this type of approach could involve source-apportioned monitor data.
Specifically, modeled air concentrations resulting from point and mobile sources, as well as
reentrainment could be compared with the relevant source-apportioned signal from ambient
monitors within the study area.
Figure 3-3 Overview of Modeling Approach for Ambient Air (where measurement data
are available and sufficient, they will be used in preference to, or in combination with
modeling steps shown here to characterize baseline conditions)
1
Base
Ambie
Concen
(Cur
Emiss
Point/Mobile
Source Emissions
Data
i
V
Re-entrainment
(WEPS/PEF)
i
"Current"
Reentrainment
Emissions
......
AERMOD
1 j
^
! Wet/Dry "
1 Deposition [
' 7
Simple Soil I
Reservoir
Model 1
L^
Baseline Pb Soil
Concentrations
1 ' Attributable to
Point/Mobile
Sources
Soil Measurement Data From Different Location Representing
Older Auto-Related Emissions (location similar to case study
but not impacted by local point sources)
j-l
i Baseline Pb Soil Concentrations [
[ Attributable Historic (auto) Impacts i
V1
Re-entrainment
modeling
(WEPS/PEF)
j 1
Ambient Air Emissions
(reentrainment of
historical auto-sourced
soil/dust lead)
' ~~~l
AERMOD W '
H
line Base ine Ambient Air B
ntAlr Concentrations
rations (Reentrainment of Current (reen
re™ emissions) a
ions) „
[
i
jseline Ambient Air
Concentrations
trainment of historical lnPut Dataset
1 Modeling
Step
] Output i
^i Base ine Ambient Air '
^' Concentrations ,
4.1.2. Indoor Air
To estimate indoor inhalation exposures indoors associated with ambient (outdoor) air
concentrations, we intend to use either ambient-to-exposure concentration ratios or inhalation
exposure modeling. The inhalation exposure modeling approach, described in Section 4.1.5,
DRAFT-May 31,2006
22
-------
takes into account differences in indoor and ambient air concentrations, time spent indoors vs.
outdoors and other activity patterns, and variability in concentrations across a variety of different
microenvironments (e.g., cars, offices, outside).
4.1.3 Outdoor Soil/Dust
The approach used to characterize outdoor soil/dust concentrations for a particular study area
will depend on the amount of site-specific soil measurement data available. However, it is
anticipated that for all case study locations, potential future changes in soil Pb concentrations
under different NAAQS scenarios will need to be modeled to reflect the potential impact of
decreases in deposition of Pb across the study area, reflecting decreases in ambient air-related
Pb.
To predict future soil Pb concentrations (or baseline conditions for those case studies lacking
sufficient monitoring data), we are considering using deposition predicted from the air quality
modeling combined with a simple soil reservoir model, such as EPA's Multipathway Exposure
(MPE) methodology (EPA, 1998a). MPE is a set of algorithms developed to estimate cumulative
soil concentration as a function of dry and wet particle deposition, soil mixing depth and bulk
density, and a soil loss constants for processes of interest that impact concentrations in surface
soil. As an alternative to MPE, we are considering the option of using EPA's Total Risk
Integrated Methodology Fate, Transport and Ecological Exposure model (TRIM.FaTE) which
employs a fully coupled mass balanced approach to simulating distribution of pollutants of
interest among media of the simulated ecosystem (http://www.epa.gov/ttn/fera/trim_fate.html).
TRIM.FaTE includes the capability for dynamic as well as steady-state modeling and has
sensitivity analysis and MonteCarlo features.
4.1.4 Indoor Soil/Dust
The approach used to characterize indoor dust concentrations for a particular study area will
depend on the amount of site-specific measurement data available. For example at the primary
smelter case study location, some site-specific residential dust data are available. For other case
study locations, we may consider the use of literature-based concentrations for comparable
residences or a modeling approach (e.g., below). Additionally, for all locations under future air
quality scenarios, we intend to use a modeling approach.
We are considering the modeling approach employed by the IEUBK model (EPA 2005a, EPA
2006b, EPA 1994) which presumes a linear regression relationship between air Pb (PbA),
outdoor soil Pb (PbS) and indoor dust Pb (PbD), and accommodates user-specified inputs.
PbD = p(other sources) + PS PbS + PA PbA
The IEUBK recommended default values for PS and PA, respectively, are 0.70 and 100 ug/g per
ug/m3, with the former being based on site-specific data where soil was a major contribution to
household dust. Prior to adoption, we intend to evaluate the application of this equation and
these defaults in light of alternates currently available in the literature, including information
discussed in the draft AQCD (EPA, 2006b). Note, that Pother sources) covers background exposure
DRAFT-May 31,2006 23
-------
sources, including Pb paint in older houses (the characterization of background exposures
including Pb paint levels in indoor dust is discussed in Section 4.1.6). We also plan to evaluate
the performance of the modeling approach in light of the current literature.
4.1.5 Estimating Inhalation Exposure Concentrations
Estimates of inhalation exposure concentrations for the population subgroups of interest will be
derived in the Pilot assessment using ambient air concentration estimates (see Section 4.1.1). A
potentially important factor to consider in assessing inhalation exposure is an individual's daily
activity patterns including their commuting activity (e.g., home to work, home to school). In
order to incorporate consideration for daily mobility into our modeling of inhalation exposure,
we are considering the use of data generated as part of the recent NATA-nsa. Specifically, we
are considering also using ratios of modeled exposure Pb concentrations to ambient air Pb
concentrations (the former reflecting daily mobility and generated at the US Census tract-level)
to adjust our ambient air concentration values to reflect consideration for daily mobility. For the
NATA-nsa, annual average ambient air concentrations were estimated for all U.S. census tracts
and 30 replicates of annual average exposure concentrations were estimated for five different
population subgroups (i.e., five age groups) in each census tract reflecting daily activity patterns
including commuting trips. We may use these data to develop 30 replicate exposure-to-ambient
concentration ratios per census tract and population group in a study area and these ratios in turn,
used to derive 30 replicate estimates of annual average inhalation exposure concentrations for
each population subgroup in each geographic unit (block group) in the study area. The 30
replicate estimates of annual average inhalation could then be used to establish central tendency
inhalation estimates at the US Census block group-level and the full distribution could be used in
Monte Carlo simulation of exposure-related variability (see Section 4.3.4).
If the Pilot analysis suggests that inhalation exposure to Pb is a significant contributor to overall
Pb exposure, then we may consider a further refinement in our consideration of daily mobility by
running HAPEM or APEX for each study area and population subgroup of interest rather than
relying on the ratios generated from the NATA-nsa.
4.1.6 Background (Non-Air-Related) Exposure
In addition to estimating media concentrations for air-related exposure pathways, we will also
estimate Pb exposures for media pathways that are not directly linked to air concentrations (i.e.,
background exposure levels). Consideration of these background exposures allows us to model
total Pb exposure for our study populations. Potential sources of data and a brief overview of
how media concentrations will be estimated are presented here by medium:
• Indoor and Outdoor Soil/Dust Associated with Lead Paint. EPA, HUD, FDA, and other
Agencies have developed extensive monitoring programs to address non-air related Pb
exposure concentrations. For children living in older housing or in urban areas, Pb levels
in outdoor soils and house dust are likely to contribute the bulk of total Pb intake. For the
portion of these exposures not due to re-entrainment, we are considering a number of
surveys of urban soil and house dust Pb levels (Lanphear et al. 1998; EPA 1998b, 2000,
Jacobson et al., 2002) as potential sources of exposure concentration values spanning the
DRAFT-May 31,2006 24
-------
range of housing stock and urban settings. The National Research Council (2005) has
also reviewed a number of soil Pb data sets for communities near Pb mining and smelting
operations.
• Dietary Exposures. Similar to air Pb levels, Pb concentrations in food have also been
decreasing over the last two decades. In a recent project for the Office of Water (OW),
Agency contractors reviewed available data sources related to dietary Pb concentrations
(ICF 2005) and evaluated the trends in Pb concentrations in major food groups reported
in FDA's Total Diet Survey data from 1992-2002 (FDA 2004). These and similar data
are being considered, along with information related to food intake patterns, to estimate
Pb intake from foods. We do not expect to identify any site-specific food consumption
data for the case study sites.
• Drinking Water. We plan to gather data on the distribution of Pb concentrations in
drinking water to monitor compliance with the "Lead and Copper Rule" (EPA 2004b).
These data document decreases in water concentrations since the promulgation of that
rule in 1991. However, the summarized data provided in this data set are not suitable for
detailed exposure estimation because (1) the sampling program is targeted rather than
random and (2) data are reported as 90th percentile values rather than means, percentiles,
or individual values. As part of ongoing work for OW, we will pursue additional data,
either individual sample data from EPA's SDWIS system and/or data from a small
number of individual systems in an effort to characterize exposure concentrations related
to drinking water.
4.2 Study Population(s) and Potential Stratification Based on
Socioeconomic Factors
The selection of subpopulations to be included in this assessment focused on those for which
there are endpoints associated with the range of exposures expected to most closely reflect
current levels and for which there is adequate information to support quantitative risk estimates.
We are currently planning on modeling risk metrics related to IQ loss for children under seven
years of age and are considering various options for modeling blood pressure and renal endpoints
for adults. Each of these demographic groups is discussed below, including the degree to which
they can be differentiated in modeling based on gender and race. In addition, several additional
subpopulations of potential interest are discussed (e.g., pregnant women, post-menopausal
women), although we do not expect to quantitatively assess risks to them.
• Children from birth through age 6: We will consider the option of differentiating blood
Pb level modeling for different socioeconomic subpopulations to the extent that there are
data that (a) specify different behavior translating into differential exposure levels for key
media (e.g., dietary ingestion rates), or (b) specify different Pb concentration levels (e.g.,
paint-related dust contamination based on age of housing). Consideration of
socioeconomic-related differences in exposure may be treated as part of sensitivity
analyses rather than explicitly in population-level exposure modeling.
DRAFT-May 31,2006 25
-------
• Adults: We are considering modeling both working-age (18-64 year olds) and retired (65
to 80 year olds) adult populations. In addition, we may consider modeling adult
subpopulations known to have high occupational exposures to Pb, or subpopulations
expected to experience high Pb exposure due to a confluence of Pb exposure-related
factors (e.g., housing age, dietary exposure, proximity to ambient air sources). As with
the child study population, we are considering the option of differentiating blood Pb
modeling for adults based on socioeconomic status (e.g., differences in behavior that
translates into increased exposure levels, differences in Pb concentrations in key media
such as indoor dust related to housing age).
• Pregnant women: While this subgroup will not be quantitatively assessed here (see EPA
2006b for summary of currently available studies), fetal exposure from the mother will be
include in estimating blood Pb levels for children.
• African Americans (and blood pressure modeling): Racial and ethnic differences in blood
Pb concentrations are well-demonstrated and age-specific average blood pressure is
known to be substantially higher in African Americans than in other groups (Pirkle et al.
1998, Vupputuri et al. 2003). However, we are not aware of an approach to differentiate
exposure-response relationships for the African American sub-population in particular.
However, should we undertake an assessment of the potential shift in population-level
blood pressure distributions for the case studies as part of the Pilot, we may consider
African Americans separately through the use of socioeconomically-differentiated blood
pressure distributions and the impact that Pb-related exposure has on those distributions.
4.3 Estimating Blood Lead
Once exposure levels in the form of either modeled intake rates (e.g., for dietary items and
indoor dust) or exposure concentrations (e.g., for ambient air) have been generated for study
populations of concern, the next step, as outlined in Figure 3-2, is to model blood Pb levels for
those populations. The concentration of Pb in whole blood is the most commonly used measure,
or "biomarker," primarily because it is most convenient and easily measured, but also because
blood Pb tends to be a good indicator of recent exposures. Pb in long-term body stores
(primarily bone) may also contribute to blood Pb concentrations and to the risk of adverse
effects. Thus, most approaches for estimating adverse effects take into account the "biokinetics"
(i.e., uptake, deposition, mobilization, and excretion) of Pb in the body. Simplified forms of
biokinetic models (e.g. "slope factor models") involve a linear projection of blood Pb from either
intake or uptake estimates developed outside the model. "Empirical" approaches bypass the
explicit modeling of biokinetics and predict blood Pb levels directly based on concentrations in
exposure media. This section discusses the approaches we are considering for modeling blood
Pb levels in children and adults, which include all three types of models.
Although the various models involve different ways of accepting "inputs" (e.g., media
concentration with ingestion rate vs intake rate), we intend to ensure that we provide the models
used with the same inputs. For example, the Model B input for intake rate will match the
combination of Model A's inputs for media concentration and ingestion rate. This will ensure
that differences in model outputs will be reflective of differences in the models themselves.
DRAFT-May 31,2006 26
-------
The section begins by discussing the models selected for children, followed by options identified
for modeling adults. We note that in our use of any model, all inputs will be considered, with
values chosen based on the best available information as replacement for defaults as supported.
The section then discusses blood Pb metrics under consideration for both children and adults
(e.g., concurrent or lifetime-averaged values). The section then discusses a probabilistic Monte
Carlo-based approach being considered for exposure variables, with subsequent "batch mode"
modeling of blood Pb levels through the various models (Note, this probabilistic simulation of
exposure is an option that applies both to children and adults and consequently is discussed in its
own subsection). The section concludes with a discussion of model preparation and evaluation.
4.3.1 Children
We intend to use the IEUBK model (EPA, 1994, 2002a, 2002b) for modeling blood Pb levels in
children. We are also considering employing an additional biokinetic model (Leggett et al.,
1993) and an empirical model (Lanphear et al., 1998) in order to consider model uncertainty in
this key step in the risk assessment. As discussed below, the Leggett model is being considered
for predicting both children's and adult (if included) blood Pb levels. Chapter 4 of the draft
AQCD (2006b) discusses these models.
We are also considering the pharmacokinetic model of O'Flaherty et al. (1993) as another option
for use in this analysis. While its approach to blood Pb modeling is also quite sophisticated, the
O'Flaherty model has been subject to considerably less independent validation against human
population data and other models than the Leggett model. Also, the precise model structure and
parameter values are less well-documented than those of the other models.
As noted earlier, we are also considering an "empirical" model (the Lanphear model) for
estimating blood Pb levels in children, however, its use poses some challenges. The model
includes no component for inhalation exposures, so it is necessary to assume that the main blood
Pb impact of changes in ambient air levels occurs through changes in soil and indoor dust Pb
levels. In addition, the full Lanphear et al. model includes parameter values for many variables
for which data may not be available for some or all of the study populations. In addition, the
study identifier variable (specifying the different populations combined in the analysis) is highly
significant and has a substantial impact on estimated blood Pb levels. This variable presumably
captures unidentified site-specific covariates that strongly effect blood Pb levels, that may or
may not apply to a specific case study location.
One way to apply the model that is being considered involves adapting the Lanphear et al. (1998)
output tables as "response surfaces" for evaluating changes in exposure concentrations. The
tables give predicted geometric mean blood Pb levels for various combinations of soil Pb
concentration and house dust Pb loading, with all other covariates held at mean or median
values. Changes in ambient and indoor air concentrations associated with the different case
studies/exposure scenarios would enter into the model through their effect on soil concentrations
and indoor dust Pb loading, estimated through deposition modeling. We recognize, however,
that this approach would ignore the significant role of covariates in predicting blood Pb levels.
DRAFT-May 31,2006 27
-------
4.3.2 Adults
For adults, we are considering the use of both a biokinetic model (Leggett 1993) and a "slope
factor" model (the Adult Lead Methodology, ALM, EPA, 1996b, Maddaloni 2005 or Bowers et
al 1994). The Leggett model was discussed in the previous section and will not be covered
further here. However, it is important to note that if employed, the Leggett model will provide a
"bridge" between the children's and adult's blood estimates. The remainder of this section will
focus on the ALM, derived from the Bowers et al (1994) model.
The ALM was originally developed by EPA (1996b) for the purpose of estimating changes in
long-term (months or years) adult blood Pb levels arising from non-occupational (i.e.,
residential) exposures to contaminated soil near "Superfund sites." The model has two parts.
The first, which is implicit, is the calculation of a long-term average Pb uptake from soil Pb
concentration data. The parameters in the uptake calculation are the soil Pb concentration, a
combined ingestion rate of soil, including both outdoor soil and indoor soil-derived dust, an
absolute gastrointestinal absorption fraction (for soil and dust from soil), and exposure frequency
and averaging times.
The estimated Pb uptake is multiplied by a "biokinetic slope factor", derived from data from
human exposures (Pocock et al 1983, Sherlock et al 1982, 1984), which is intended to capture
the relationship between long-term Pb intake and the estimated increment in blood Pb levels.
The increment is added to a population-specific "background" central tendency blood Pb level to
generate an estimate of the central tendency blood Pb concentration associated with soil
exposures at the specified concentration. In hazardous waste site applications, this background
blood level has been derived from a national databases such as the NHANES.
The output of the ALM is a central estimate of the "quasi-steady state blood Pb concentration for
the population exposed to the specified Pb concentration in soil. Maddaloni et al (2005)
presented an approach to estimating the distribution of estimated blood Pb levels based on the
assumption of lognormality, again using data from NHANES. The ALM model can be
expressed as a set of simple equations, facilitating adapting of the model for Monte Carlo
analysis, should that be undertaken.
4.3.3 Blood Lead Metrics
The "raw" outputs from the blood Pb models will be time profiles of blood Pb levels across the
ages of interest, which will be converted to metrics that can serve as inputs to the adverse effects
models discussed in the following section. Metrics under consideration for children and adults
include:
• Child blood lead metrics: The metrics will be those particular to the adverse effects
models chosen to quantify IQ risk, such as "concurrent" (i.e., blood level at a particular
age typically associated with testing for the endpoint of interest such as IQ) or lifetime
average blood Pb levels (birth through particular age). The models being considered are
presented in Table 5-1.
DRAFT-May 31,2006 28
-------
• Adult blood lead metrics: If adults are included, estimated adult blood Pb concentrations
will be used with the adverse effects models chosen to quantify blood pressure or renal
effects. Blood Pb metrics for adults will focus on the average over the adult-aged
exposure period (e.g., the average modeled blood Pb levels for ages 18-64 for the
"working aged" adult population).
4.3.4 Probabilistic Analysis of Exposure Parameters
As described in Section 3.2, the IEUBK model is typically applied with the use of a GSD to
estimate the distribution of blood Pb concentrations reflecting inter-individual variability in both
exposure and biokinetics. In their recent evaluation of a risk assessment conducted for soil Pb
contamination at large Superfund Sites, the NRC (2005) suggested that future use of the IEUBK
and similar blood Pb models, in that context, should employ probabilistic simulation of
exposure-related variables. We are considering employing such an approach8 in this assessment
for all of the selected models. This approach would involve implementation of external Monte
Carlo-based modeling of population exposure, followed by "batch-mode" blood Pb modeling for
the set of Monte Carlo generated exposure simulations.
If this is pursued, the probabilistic simulation of exposure will be conducted using a one-
dimensional Monte Carlo model which uses probability distributions for exposure concentrations
and intake/uptake factors in the case of IEUBK, Leggett and ALM models. The probabilistic
exposure model used to feed the Lanphear model would include distributions of estimated
changes in soil Pb and house dust concentrations, since that model does not include any exposure
factor values.
While the focus here is on sources of variability (with sources of uncertainty reserved for the
sensitivity analysis - see Section 7.0), as is often the case with stochasticity in exposure, some
sources mix variability and uncertainty. For example, distributions of exposure concentrations
estimated for specific portions of a study area using a particular modeling approach can reflect
both spatial variability in Pb levels across that area, as well as uncertainty in capturing that
variability. In this case, it is difficult to separate these two sources of variability in Pb levels and
it is likely that Monte Carlo simulation of exposure variation will reflect both variability and
uncertainty in the case of this factor.
Exposure factor distributions used in probabilistic modeling (reflecting variations in receptor
behavior) will be developed based on available literature data. For some variables (drinking
water and food consumption), substantial data are available from surveys and other studies to
support the derivation of probability distributions reflecting the expected wide range of
variability across the exposed populations (EPA 1997, 2002a, 2002b) Similarly, the databases
used in the inhalation exposure models include probabilistic representations of time activity
patterns. Children's soil and house dust ingestion have been subject to extensive study, while
for other exposure factors (adult soil ingestion, for example) there are far fewer data to support
estimation of probability distributions (Maddaloni et al. 2005). Probabilistic representations of
DRAFT-May 31,2006 29
-------
gastrointestinal absorption fraction will be chosen to reflect the literature related to GI absorption
from various Pb species (EPA 2005b, Maddaloni et al. 2005).
An initial assumption used in developing the probabilistic simulation would be that the input
distributions of the exposure concentrations and exposure factors are independent (not
correlated) except to the extent that they co-vary in space and are age-specific. This assumption
could be examined more closely in an effort to evaluate uncertainty related to correlation for the
Full Scale analysis, should probabilistic simulation be undertaken.
Evaluation of this modeling analysis will include comparison of the resultant modeled
distributions of blood Pb levels with empirical measures of variability in blood Pb levels
obtained either from site-specific data or NHANES. Such a comparison may also inform our
understanding of the contribution to population blood Pb variability (e.g., as characterized by
GSD) from biokinetic variability for modeled populations.
If this option is pursued, the blood Pb distributions generated will be carried through the
assessment of adverse health effects in order to generate distributions of IQ loss estimates and
blood pressure effects for modeled populations.
4.3.5 Preparation of Models and Inputs, and Evaluation of Performance
Briefly, the preparation of models for application and their evaluation will involve the following
steps.
• Develop and document consistent sets of intake/uptake assumptions (exposure factor
values, absorption fractions, etc.) for the three biokinetic models9 derived from the
outputs of the case study exposure concentration models.
• Run the IEUBK, and Leggett, and Lanphear et al. models (assuming these other models
are employed) for a small number of children's exposure scenarios covering a credible
range of exposure conditions and parameter values. If adults are included, run Leggett
and the ALM for a small number of adult exposure scenarios.
• Compare results of models to age-specific population blood Pb data from the most recent
NHANES, and data from Superfund sites and/or HUD housing studies.
• Compare model results to each other within age ranges.
• Conduct sensitivity analysis to identify parameters/assumptions contributing the most to
differences/uncertainties in results. The first step in this analysis will be the comparison
of model performance in the selected test scenarios; additional model runs will be
conducted if necessary.
• Fully document the process.
• Evaluate overall performance of the models for each age group.
In these steps, we will draw extensively on the previous published modeling comparisons and
studies, and may confer with the model authors and other experts during the process. The
9 A preliminary review of the available literature suggests that previous comparisons did not always involve
consistent exposure, intake, and uptake models and parameter values across the models being compared. Once
consistent intake/uptake assumptions have been defined, the results from the various models can be compared.
DRAFT-May 31,2006 30
-------
sensitivity of the models to changes in specific parameters will be evaluated. Systematic
differences among the model predictions will also be noted; that is, where one or more model
predictions differ consistently relative to another under specified sets of exposure conditions.
Analysis of the patterns of differences will inform consideration of specific model intake/uptake
parameters. In evaluating the reasonableness of model predictions, greater weight would be
given to deviations from observational data (when available) than to differences among models.
DRAFT-May 31,2006
-------
5.0 EFFECTS ASSESSMENT
Lead exposure has been shown to affect a wide range of organ systems and physiological
processes at differing doses in adults and children. The most recent epidemiological studies have
been reviewed by EPA in the latest draft AQCD (EPA 2006b). Because extensive de novo
biostatistical analyses are beyond the scope of the assessment, we intend to draw from the
evaluation and identification of adverse effects models contained in the draft AQCD (EPA,
2006b). The staffs basic approach will be to (1) identify a small number of the most appropriate
studies and associated models and (2) apply them to each endpoint of concern to derive a range
of model-dependent risk estimates.
5.1 Endpoints to be Evaluated
For the Pilot analysis, we are planning to model IQ effects in children and will consider options
for modeling blood pressure and renal effects using creatinine clearance (a commonly used
measure of glomerular filtration rate (GFR), an index of kidney function) in adults. As feasible,
the full assessment may include all three endpoints. As described in the current draft of the draft
AQCD (EPA 2006b), there are a number of high-quality studies reporting these effects that are
available and a consensus that these effects are "adverse" and occur in the range of exposure
anticipated to be relevant to this analysis and the case studies included. The draft AQCD (EPA,
2006b) presents findings regarding susceptible subpopulations (e.g., renal effects in hypertensive
subpopulations) that will be considered in characterizing risk associated with the endpoints
included in the assessment.
5.2 Models for Estimating Adverse Effects
5.2.1 IQ Reduction in Children
We intend to focus on the Lanphear et al (2005) study as our primary source of exposure-
response functions for use in quantifying the relationship between children's blood Pb and IQ
decrements. In our use of Lanphear et al (2005) in this assessment, we intend to rely on a
nonlinear model, such as the log-linear or piece-wise linear fits depicted in Figures 3 and 4,
respectively, of that study, to estimate IQ decrement associated with estimated blood Pb
concentrations, and to capture the higher slope indicated by this analysis at lower blood Pb levels
(note, the current approach does not call for conducting adjustments for covariates for individual
case study populations).
We are also considering the other studies identified in the draft AQCD (Table 6-2.2 of EPA
2006b) reporting quantitative relationships of IQ and blood Pb for populations with blood Pb
levels less than 10 |ig/dL. 10 These studies are listed in Table 5-1. Although included in this set
are several studies on non-U.S. populations, the responses of these populations do not appear to
differ substantially from the responses in the U.S. populations. For example, the estimated
slopes for blood Pb under 10 jig/dL provided in Table 6-2.2 of the draft AQCD (EPA 2006b) for
the two U.S. studies range from -0.8 to -1.6 IQ points per |ig/dL blood Pb, while the other
10 We have not included Al-Saleh et al (2001) which studied children of notably older ages (6-12 years) than the
other studies being considered.
DRAFT-May 31,2006 32
-------
studies, which include non-US populations, range from -0.8 to -1.1 IQ points per |ig/dL blood
Pb. In selecting a set of studies for use in characterizing the range of uncertainty associated with
modeling the IQ endpoint, preference will be given to studies conducted in the US reflecting
concerns over study population and exposure condition relevance.
Table 5-1. Studies with Quantitative Relationships of IQ and Blood Lead for Blood Lead
Levels Less than 10 ug/dL (drawn from draft AQCD Table 6-2.2, EPA, 2006b).
Reference
Tellez-Rojo et al. (2006)
Bellinger etal. (1992)
Kordas et al. (2006)
Canfield et al. (2003)
Lanphear et al. (2005)
Study Location
Mexico City, Mexico
Boston, Massachusetts
Torreon, Mexico
Rochester, New York
International Pooled
Analysis
Exposure/Dose Metric having
strongest association with IQ
Pb-B @ 24 months
Pb-B @ 24 months
Concurrent Pb-B , 6-8 yr
Concurrent @ 5yrs
Concurrent Pb-B
4.8 - 10 yrs
5.2.2 Blood Pressure Effects in Adults
For quantifying blood pressure changes associated with blood Pb levels in this assessment, if
included, we are focusing on the Nawrot et al (2002) meta-analysis and its log-linear slope
estimate as our primary blood Pb-blood pressure model for adult populations. The study also
provides confidence limits that can be considered for use as an index of the uncertainty in the
assessment associated with sample sizes considerations, and for sensitivity analyses.
We are also considering the other studies identified in the draft AQCD (Table 6-5.1 of EPA
2006b) reports quantitative relationships between systolic blood pressure and blood Pb for mean
blood Pb levels below 10 |ig/dl. These studies (listed in Table 5-2) analyzed blood pressure
effects in particular populations differentiated by gender and race. If this endpoint is included,
these studies may be relied on in the full assessment to provide risk estimates based on gender
and race specific response functions. Additionally, the current draft of the AQCD includes a
meta analysis inclusive of studies published since the Nawrot et al (2002) analysis (EPA 2006b).
This analysis also is being considered for use in quantifying this effect.
DRAFT-May 31,2006
33
-------
Table 5-2. Studies under Consideration for Quantification of Blood Pressure Change
Associated with Blood Lead Levels (drawn from draft AQCD Table 6-5.1, EPA, 2006b).
Reference
Vupputuri et al. (2003)
Den Hondetal. (2002)
Nash et al. (2003)
Cheng etal. (2001)
Procter etal. (1996)
Nawrot et al. (2002)
Study Location - and gender
NHANESIII* - White Males
NHANES III - White females
NHANES III - Black Males
NHANES III - Black females
NHANES III 1988-94 - White males
NHANES III 1988-94 - White females
NHANES III 1988-94 - Black males
NHANES III 1988-94 - Black females
NHANES II - Females
Boston normative aging- Males (about 97% white)
US Boston Normative aging study -Males
3 1 US and European Studies
(includes occupationally exposed)
* NHANES - United States population sample.
5.2.3 Renal Effects in Adults
For quantifying renal effects associated with blood Pb levels in this assessment, if included, we
intend to use changes in urinary creatinine clearance as the specific endpoint. Creatinine
clearance is a commonly used measure of glomerular filtration rate (GFR), an index of kidney
function. The draft AQCD describes multiple studies in which a relationship of this effect with
blood Pb (and bone Pb) has been observed for both the general population and specific
subgroups such as hypertensives (EPA 2006b). To quantify this endpoint in the risk assessment,
we are considering the studies for which estimates of slope for creatinine clearance11 per blood
Pb are presented in Figure 6-4.1 of the draft AQCD (EPA 2006b). These studies, which were for
general populations and generally had mean blood Pb levels less than 10 |ig/dL, are listed in
Table 5-3.
11 For studies where creatinine clearance was not measured, it was estimated using the relationship based on
creatinine levels, age and weight from Cockcroft and Gault (1976).
DRAFT-May 31,2006
34
-------
Table 5-3. Studies under Consideration for Quantification of Creatinine Clearance Change
Associated with Blood Lead Levels (drawn from draft AQCD Figure 6-4.1 and Table 6-4.1,
EPA, 2006b).
Reference
Paytonetal. (1994)
Kim etal. (1996)
Tsaih et al. (2004)
Staessen etal. (1992)
Akesson et al. (2005)
Study location - Study population
Boston, MA - Normative Aging Study, 1988-1991
Boston, MA - Normative Aging Study, 1979-1994
Boston, MA- Normative Aging Study- 1991—2001
Belgium - Cadmibel Study
Women's Health in the Lund Area Study, Sweden - 1999-2000
We are also considering potential use of Muntner et al (2003), which included a focused analysis
on creatinine clearance changes in hypertensives.
DRAFT-May 31,2006
35
-------
6.0 RISK ASSESSMENT
Once blood Pb metrics have been generated for the study populations of interest (for each case
study location), the next step, as outlined in Figure 3-2, is to combine these exposure estimates
with concentration responses functions developed as part of Effects Assessment to generate risk
estimates. This analysis is primarily focused on generating population risk metrics (i.e.,
distributions of risk levels across modeled populations). Note, however, that we may consider
generating more hypothetical individual or small-group risk estimates in order to cover higher-
level exposures which may not have been captured in the population-level risk estimates
generated for the analysis. We are planning currently, to generate risk estimates for IQ loss for
children and are considering the option of generating risk estimates covering blood pressure and
renal function effects for adults. This section begins by describing options for defining modeled
populations from the stand point of risk characterization, including options for assigning
exposure concentrations (e.g., modeling a cohort as beginning exposure all at the same age, or
modeling a cohort with distributed ages that experiences varying exposure windows). The
section then describes risk metrics being considered for both the child and adult age groups.
Note, that risk metrics generated as part of the integrated strategy for addressing uncertainty and
variability described in Section 7.0 are discussed separately in Section 7.2.
6.1 Defining Modeled Populations in (Assigning Exposure
Concentrations)
A key issue in generating population risk estimates is whether study populations are defined as
cohorts which begin exposure all at the same age and then are tracked forward in terms of
exposure and risk, or as cohorts with a distribution of ages experiencing a range of exposure
windows across the simulation period. An additional factor to consider in modeling population-
level risk, is how to treat the period before the exposure of interest begins (e.g., assign all
modeled individuals baseline exposure levels prior to implementation of the alternate NAAQS
standard). There are several options under consideration for defining study populations for
purposes of this analysis. As this issue has policy-related aspects, we do not expect to be
selecting a specific approach, and instead, may run multiple options in order to characterize the
sensitivity of risk results to this factor. Specific options available for defining the modeled
populations in terms of assigning exposure concentrations include:
• Uniform age cohort: This option defines the population as a group of individuals at the
age defining the lower end of the age range being modeled (e.g., for children, beginning
all modeled individuals at age 0 yrs). Therefore, exposure to the levels of interest (e.g.,
those associated with a particular alternate NAAQS standard), would begin when they are
at the lower-bound age. In the case of adults (and the mothers of children for purposes of
predicting prenatal Pb contributions where appropriate), we would assume that exposures
prior to the start age is at background levels.
• Distributed (demographically representative) cohort: This option defines the population
as distributed across the age range of interest, as reflected in available demographic data
(e.g., 10% of children at 0-1 yrs, 15% at 1-2 yrs and so on). With this option, modeled
individuals begin the exposure of interest (e.g., an alternative NAAQS level) at different
DRAFT-May 31,2006 36
-------
ages across the age range. As with the uniform age cohort option, the assumption would
be made that exposures at ages before initiation of the NAAQS level of interest, would
occur at background levels.
In modeling adult blood Pb levels, depending on the approach used, it may be necessary to track
blood Pb levels across consecutive age ranges. For the childhood period, biokinetic and/or
empirical models could be used to estimate central tendency blood Pb profiles as a function of
age for the adult population. For adults of working age, it is likely adult average blood Pb levels
would be used across the ages of 18 to 64 to estimate health impacts. For retired adults, the
working age exposure scenarios would be run, followed by the exposures scenarios for
retirement, and time-average blood levels for ages 65 to 80 would be used to estimate health
impacts.
6.2 Risk Metrics for Children
In previous analyses of Pb risks, EPA has adopted a number of metrics for expressing population
risks from Pb exposure in children. These have included average changes in blood Pb levels,
changes in the numbers of children with blood Pb levels above 10 |ig/dL, and population average
and aggregate changes in children's IQ.
For the Pilot, we are planning to generate risk metrics characterizing IQ changes in children,
including degrees of IQ reductions for specific percentiles of the population as well as
population-level estimates of the number of kids experiencing specific ranges of IQ reductions.
In addition to population-level risk metrics, we may consider changes in population-level blood
Pb distributions.
We are also planning to differentiate all risk metrics as to background versus NAAQS-relevant
exposures (e.g., Pb exposure-related IQ loss for a particular percentile of the children at a case
study location would be further differentiated as to magnitude of IQ loss from exposures to
"NAAQS-relevant exposures" versus exposures to "background".
It is important to note that the degree to which we can generate refined population-level risk
metrics that capture risks at the tails of the distribution (i.e., for more highly exposed
individuals), or identify the number of individuals with blood Pb levels exceeding science-policy
thresholds of interest, will depend critically on the degree of uncertainty associated with our
characterization of population variability in blood Pb levels. This, in turn, will depend on the
overall degree of uncertainty associated with the two methods described above for modeling
population variability in blood Pb levels (i.e., application of the GSD approach and use of
probabilistic simulation to cover exposure-related variability).
It is also important to make clear that the population-level risk metric planned for the Pilot
assessment is the distribution of IQ loss across the case study population, not the change in
absolute IQ levels in the exposed populations. The process of modeling shifts in absolute IQ for
a population requires that (a) the underlying IQ distribution for that population be established
and (b) that a reasonable understanding of the correlation between absolute IQ and the level of
Pb-related IQ loss exist (e.g., is greater IQ loss correlated with lower IQ for a particular
DRAFT-May 31,2006 37
-------
population or are they uncorrelated?). The process of establishing the absolute IQ distributions
for specific populations associated with a given case study location would be challenging and
subject to considerable uncertainty, which in turn, would mean that an analysis of a shift in
absolute IQ resulting from Pb-related exposure would be subject to as much, if not more
uncertainty (since the potential correlation between Pb-related IQ loss and absolute IQ could
have an significant impact on this risk metric). Because of the significant uncertainty associated
with projecting reductions in absolute IQ distributions for specific populations (uncertainty that
can not be reduced without significant research to identify representative data for a particular
population), we are not planning to project shifts in absolute IQ as part of the Pilot analysis.
6.3 Risk Metrics for Adults
As was the case for children, if undertaken, risk calculations for adults will be based on estimates
of blood Pb changes for central tendency adults derived from the Leggett or ALM models.
Corresponding log-normal distributions of blood Pb levels will be derived using GSD estimates
derived from large population data (NHANES) for groups with similar demographic and
socioeconomic characteristics as the exposed populations or for the general population where
population-specific data are not available. Average increases in central tendency blood Pb levels
will serve as simple summary values for comparing the general magnitude of effects of different
exposure scenarios. Percentile values and proportions of the exposed populations with estimated
blood Pb levels in specified ranges will also be provided.
As with children, the staff will also consider using a Monte Carlo-based simulation of blood Pb
levels for adults, reflecting exposure factor variability. This probabilistic analysis will generate
population-level blood Pb results paralleling those generated using the GSD approach described
above, with the exception that they will not provide coverage for biokinetic variability (we will
look into options for developing a GSD just to cover this specific factor). The primary outputs
from the adult risk assessment will be tabular summaries of the changes in the distribution of
blood pressure in the exposed populations.
In the Pilot assessment, we will explore the estimation of incidence of clinical hypertension by
adding the estimated increase in blood pressure to national gender- and age-specific absolute
blood pressure distributions relevant to the case study locations.
As noted earlier, there will be no attempt to predict the changes in incidence of cardiovascular
disease resulting from Pb exposure in the risk assessment.
DRAFT-May 31,2006 38
-------
7.0 UNCERTAINTY AND VARIABILITY ASSESSMENT
Modeling of Pb-related exposure and risk is subject to a variety of sources of variability (e.g.,
residential location, daily activity patterns, dietary ingestion rates, Pb uptake rates) as well as
sources of uncertainty (e.g., different blood Pb models, different health endpoint concentration-
response functions). A comprehensive analysis of uncertainty and variability typically involves
two-dimensional probabilistic simulation with one dimension designed to represent variability
and the other uncertainty. Probabilistic simulation of uncertainty requires development of
confidence distributions for uncertain input variables (parameter uncertainty) and the
establishing of degrees of confidence for multiple competing models identified for the same
modeling step (model uncertainty). Because of data limitations and constraints, it is not feasible
to develop confidence distributions for many of the sources of parameter and model uncertainty
identified for this risk analysis. Therefore, for the Pilot analysis, the staff will use sensitivity
analysis techniques to examine the impact of sources of uncertainty on exposure and risk results.
These techniques involve establishing plausible ranges for input parameters and identifying
multiple modeling options reflecting model uncertainty for specific modeling steps. Then each
uncertainty parameter (or possibly pairings of parameters expected to exhibit correlation) is
varied across its plausible range as the remaining parameters are held at their central tendency
(or expected) values. The impact of each input parameter on model outputs can then be
ascertained. The impact of individual parameters can then be compared to identify those having
the greatest impact on model results and the general degree of that impact can be evaluated to
gain perspectives on the magnitude of potential uncertainties. Similarly with model uncertainty,
alternate models can be substituted one at a time into the analytical framework to determine the
consequent impact on model results. As with sensitivity analysis results for parameter
uncertainty, these results can be evaluated to compare the impact from different sources of model
uncertainty and the overall degree of the impact from model uncertainty on model results can be
determined.
Regarding variability, as noted earlier in Section 7.2, we are considering the use of probabilistic
(Monte Carlo-based) simulation to characterize the impact of exposure variability on blood Pb
levels and ultimately, risk metrics. It is also important to note that the modeling framework
developed for this analysis reflects coverage for a range of other sources of variability besides
the exposure related factors covered in the probabilistic analysis. For example, the GIS-based
spatial modeling framework explicitly reflects variability resulting from the demographic spatial
profiles associated with modeled populations within a given study area, as well as the
intersection of those populations with the Pb media concentration fields of interest (e.g., ambient
air, outdoor soil/dust, indoor dust). Spatial and temporal variability in media concentration fields
will also be reflected in both the air monitor data and dispersion modeling used to establish those
fields.
Table 7-1 presents examples of: (a) competing models for specific steps in the modeling
framework (representing model uncertainty), (b) sources of input parameter uncertainty and (c)
specific sources of exposure-related variability associated with the risk modeling framework. It
is these sources of model and parameter uncertainty as well as exposure-related variability that
are addressed in the integrated approach discussed below.
DRAFT-May 31,2006 39
-------
Table 7-1 Sources of Model/Parameter Uncertainty and Exposure-Related Variability
Potentially Impacting the Risk Assessment
Model Uncertainty
• Source apportionment
models (PMF, CMB,
NATA-based)
• Blood Pb modeling
(IEUBK, Leggett, ALM,
Lanphear)
• Health effects
(concentration-response
model) for IQ (Canfield,
Lanphear)
• OTHERS
Parameter Uncertainty
• Pb Emissions Estimation
• Blood Pb Metrics
• Pb Uptake Factors
• Soil/House Dust Pb
• Blood PbGSD
• Background blood Pb
levels
• OTHERS
Exposure-Related Variability
• Exposure/Intake Factors
• Background Exposure
Levels (e.g., diet, paint)
• NAAQS-Relevant
Exposure Levels
7.1 Integrated Approach for Considering Uncertainty and Variability
We are considering the use of an integrated approach that combines probabilistic modeling of
variability with sensitivity analysis techniques intended to examine both parameter and model
uncertainty. This approach, which is depicted in Figure 7-1, involves development of a
Figure 7-1 Integrated Approach for Considering Model Uncertainty,
Parameter Uncertainty and Exposure-Related Variability
Modeling Tree
Supporting Sensitivity
Analysis of Model
Uncertainty
(NOTE: hypothetical
"MIN" modeling branch
identified as CMB-IEUBK-
Lanphear combo)
Source Apportionment Option
Blood Lead Modeling (children)
CM
PMF
NAT
IEUB
Leggett
Concentration-response Function
r *_
""S Lanphear
Canfield
Conduct Sensitivity
Analysis for Input
Parameters for this
Modeling Branch
Evaluate impact
of key parameter
values on
variability-based
risk distribution
Conduct Probability
Simulation of Exposure-
Related Variability for this
Modeling Branch
DRAFT-May 31,2006
40
-------
modeling options tree where each branch represents a unique combination of models identified
for the analysis (i.e., a selection of specific models for each modeling step subject to model
uncertainty). This means that the risk estimates generated by a specific combination of modeling
options would be represented as a distinct set of risk results located at the end of a particular
branch of the modeling options tree. Therefore, looking across the range of risk results generated
at the ends of the branches, provides perspective on the overall impact of model uncertainty on
risk results. Note, however, that because confidence levels have not been assigned to any of the
modeling options, it is not possible to develop a confidence distribution across the branches. All
that can be said is that the spread in risk estimates across the branches reflects the range of model
uncertainty impact on risk estimates.
We may then integrate consideration of parameter uncertainty into the modeling options tree by
taking a particular modeling branch and conducting parameter sensitivity analysis on that model
combination. Specifically, a given input parameter that is subject to uncertainty can be varied
from its central tendency (or expected) value across its plausible range and the impact on risk
estimates at the end of that modeling branch recorded. This process can be repeated for other
uncertain parameters (and for pairings of input likely to exhibit correlation) to determine the
potential impact of each uncertain parameter on risk estimates. This procedure when completed
provides a parameter sensitivity analysis for that particular branch. This procedure could then be
repeated for each modeling branch in the tree, providing a set of parameter sensitivity results for
each modeling option. This then would reflect a combination of sensitivity analysis techniques
considering both model uncertainty and parameter uncertainty.
To facilitate comparison of sensitivity analysis results both within a given modeling branch and
across branches, the parameter sensitivity analysis results for each modeling branch can be
standardized by dividing the results derived using the high-end values by the all-central tendency
estimate. This will provide a rough estimate of the degree of uncertainty (roughly equivalent to
the proportion of variance) contributed to the risk estimate by each step in the analysis.
Finally, we may integrate consideration of exposure-related variability into this integrated
modeling options tree by conducting Monte Carlo simulation of exposure parameter variability
(as described in Section 4.3.4) for each modeling branch. This would produce a distribution of
risk estimates at the end of each modeling branch which represent inter-individual variability in
key exposure-related factors. In addition, we may use the parameter uncertainty sensitivity
analysis approach described above to derive multiple variability distributions for each branch,
each reflecting the impact of a specific uncertain input parameter. By looking across the risk
distributions at the end of the modeling branches, the impact of model uncertainty on the
distribution of population risk can be considered. For example, comparison of the 95th% risk
estimates at the end of the modeling branches may reflect the impact of model uncertainty on this
risk metric. Furthermore, the set of 95th% risk estimates at the end of a particular modeling
branch can be compared to provide an indication of the impact of specific uncertain input
parameters on that particular risk metric (for that particular modeling combination).
Rather than generating full probabilistic risk metrics for the full spread of modeling branches
(i.e., running the probabilistic variability simulation for each combination of modeling options),
we would likely focus the analysis on those key branches of greatest utility to decision makers.
DRAFT-May 31,2006 41
-------
These include (a) the central tendency modeling branch (i.e., the combination of model options
generating risk estimates falling in the middle of the full range of estimates generated) and (b)
the upper- and lower-bound modeling branches (i.e., the modeling options producing the highest
and lowest risk distributions of all the branches). Central tendency and upper- and lower-bound
modeling branches could be identified by first running each modeling branch without
consideration for either probabilistic variability analysis or parameter uncertainty. The point
estimate risk results generated for each modeling branch will then be compared to identify
branches with the lowest (MEN), the highest (MAX) and the central tendency risk estimates.
The staff could then conduct full comprehensive application of the integrated approach to these
three key modeling branches. This approach will allow identification of a clear central tendency
modeling approach (which can be presented as the best estimate, in the absence of additional
confidence information allowing a more explicit and quantitative treatment of model
uncertainty). Risk results for the max and min modeling branches can then be evaluated to
provide perspective on the range or spread of model uncertainty impact on modeled results.
7.2 Risk and Exposure Metrics Generated with the Integrated Approach
As mentioned above, we are considering implementing the integrated approach for model
uncertainty, parameter uncertainty and exposure-related variability for the max, min and central
tendency modeling branches. This approach will generate three sets of risk and exposure
metrics, each reflecting a particular combination of modeling options. In addition, each of these
three sets of results will involve consideration for input parameter uncertainty as described
above. Therefore, it would be possible, for example, to estimate at the number of children with
IQ loss from air pathway-related Pb sources in the range of 1-2 IQ points for the max, min and
central tendency modeling options. It would also be possible to see how this population risk-bin
metric varies given parameter uncertainty for each modeling option (and to compare the impact
of that parameter uncertainty across the three modeling branches). This hypothetical example is
presented in Figure 7-2.
DRAFT-May 31,2006 42
-------
Figure 7-2 Example Risk Metrics Generated Using the Integrated Approach
i
Modeling Tree Supporting Sensitivity Analysis of
Model Uncertainty
Source Apportionment Option used \
in fate/transport modeling ]
Blood Lead Modeling
(children
Concentration-
response Function
MIN
Branch
CENTRAL
TENDENCY
Branch
MAX
Branch
Monte Carlo-based (exposure-related)
estimate of number of children with IQ loss (22) (34) (43)
in the 1-2 point range
Sensitivity Analysis
for Input Parameters
and impact on
population risk metric
generated using Monte
Carlo simulation
Pb Emissions Estimation (20 — 44)
Blood Lead Metric (25 - 37)
Soil/House Dust Lead (31 - 36)
Baseline blood lead levels (15 - 48)
DRAFT-May 31,2006
43
-------
ECOLOGICAL RISK ASSESSMENT
8.0 OVERVIEW OF ANALYSIS PLAN
8.1 Conceptual Model for Lead Ecological Risk Assessment
Figure 8-1 represents a conceptual model of the elements pertinent to assessing ecological risks
associated with environmental Pb exposures. The freshwater and soil pathways are the primary
pathways of interest to this analysis and will be addressed quantitatively in this assessment.
Sources: The focus of the NAAQS review is on sources of Pb to ambient air (e.g., stationary and
mobile emissions sources and resuspension of anthropogenic Pb). Other non-air sources of
environmental Pb (including mining activities, contaminated landfills, etc.) may contribute to Pb
concentrations in environmental media. This analysis deals primarily with present and past
emissions to air and the resulting deposition.
Pathways: The ecologically significant pathways of exposure are through ambient air and the
accumulation of Pb in media (soil, water, sediment). Direct inhalation of Pb, while a source of
exposure, is probably the less significant one. This analysis will deal primarily with deposition
and resulting concentrations in environmental media.
Organisms: Those organisms in direct contact with Pb contaminated media whether directly or
by prey selection are most likely to be the most highly exposed organisms in the environment.
There is limited evidence for biomagnification of Pb in food chains, but sensitivities to lead do
vary widely within and among groups of organisms with similar exposures.
Endpoints: Exposure to Pb at significantly high levels can cause effects to individuals and
populations thereby altering processes and interdependences of ecosystems. Known effects of
high Pb exposure include stunted growth, decreased fecundity, and increased mortality rates
among some organisms.
Risk Metrics: Metrics could be developed to look at individual, population, and ecosystem
effects from led exposures. Individual-level toxicity data that are likely to represent thresholds
for population level effects in sensitive species were used to develop the screening methods that
are be considered for this analysis.
DRAFT-May 31,2006 44
-------
Figure 8-1. Conceptual Model of Lead in Ecosystems
LU
a:
3
W
O
Q_
X
UJ
Stationary and
Mobile Sources
Soil and Dust
Terrestrial
Organisms
Aquatic Organisms
EFFECTS
(
ENDPOINTS
Survival
,
Reproduction
Growth
to
£
ORGANISM
Toxicity for specific
intake rate
POPULATION
Growth Rates
Reproductive Rates
NOTES
Many of the processes and pathways above are circular in nature. For the clarity of the schematic
they are shown as unidirectional.
Components with gray text will not be addressed in the quantitative assessment due to uncertainty
regarding available data and modeling tools.
1 Water in this schematic represents all surface water bodies but only freshwater is addressed
in the analysis plan.
DRAFT-May 31,2006
45
-------
8.2 General Overview of Analysis
The primary route of exposure to Pb for ecological receptors is very different than the typical
inhalation exposure path of most other criteria air pollutants. In general, exposure to Pb comes
through contact with or ingestion of media that contains Pb by way of air deposition, runoff,
dispersion of man made materials containing Pb, and exposed mining waste. In addition, Pb
accumulates in the environment over time and does not dissipate readily. These two facts make
a multimedia approach to Pb analysis necessary to assess the long term effects of deposition
from ambient air.
This assessment will be performed in two steps or tiers. The first tier will use case study
locations developed in conjunction with the primary standard assessment to determine predicted
concentrations of Pb in soils and will consider water and sediment based on availability of data.
These media concentrations will then be compared to environmental screening levels developed
for ecological receptors to focus the analysis on those receptors most likely to be affected by
ambient Pb concentrations. The second tier of the analysis will then use available modeling to
determine what the intake rate is likely to be for sensitive receptors and compare these rates with
available data on concentration effects.
The sections that follow describe in more detail what is envisioned for analyzing each
component as illustrated in Figure 8-1. Section 9.1 describes how estimates of current and future
media concentrations will be generated using empirical data and modeling to provide a
comparison with several screening tools available for predicting the likelihood of adverse effect
to organisms from specific concentrations of Pb in soil, freshwater, and sediment. These
screening risk results can be used to focus further analysis on those receptors thought to be most
at risk based on the outcome of the screening step. Section 9.2 is a discussion of the proposed
detailed analysis of those receptors found to be at risk in the tier 1 analysis. A discussion of the
model and method for determining intake rates or body/tissue concentrations for the susceptible
receptor(s) is also found in this section of the plan. Figure 8-2 gives an overview of the proposed
analysis.
DRAFT-May 31,2006 46
-------
Figure 8-2. Overview of Planned Analysis
I-
Characterize current conditions in air,
soil, water, sediment for case study locations
(empirical or modeled)
Input
Load media with air concentration over time
Output
Future concentrations in media for
case study location
Compare to media-specific screening benchmarks
To determine most at risk receptor group
CM
0
I-
Detailed modeling of susceptible receptor(s)
Output
Modeled body/tissue concentration or intake rates
for susceptible receptor(s)
Compare to known concentration effects data if possible
DRAFT-May 31,2006
47
-------
8.3 Uncertainty and Variability
Modeling of Pb related exposure and risk to ecological receptors is subject to a wide array of
sources of both variability and uncertainty. Variability is associated with geographic location,
habitat types, physical and chemical characteristics of soils and water that influence Pb
bioavailability, terrestrial and aquatic community composition, Pb uptake rates by invertebrates,
fish, and plants by species and season. For wildlife, variability also is associated with food
ingestion rates by species and season, prey selection, and locations of home ranges for foraging
relative to the Pb contamination levels. Sources of uncertainty include modeling choices for
future media concentrations and assumptions used to derive the ecotoxicity screening
benchmarks. Uncertainty and variability will be discussed to the extent possible for each step of
the analysis, and where feasible, comparisons will be made between model outputs and empirical
data.
As discussed in Section 7 for the pilot human health risk analysis, we are considering the use of
an integrated approach for addressing both uncertainty and variability, which will combine
probabilistic simulation (for addressing exposure-related variability) with sensitivity analysis
techniques (for addressing both parameter and model uncertainly). The integrated approach
would include a "modeling tree" to represent model uncertainty with three branches representing
the high-bound (Max), low-bound (Min), and central tendency modeling risks for a given
receptor/medium combination. Because of data limitations and constraints, it is not feasible to
develop confidence distributions for many of the sources of parameter and model uncertainty
identified for this risk analysis. Thus, the staff will use sensitivity analysis techniques to
examine the impact of sources of uncertainty on exposure and risk results.
The sensitivity analysis techniques involve establishing plausible ranges for input parameters and
identifying multiple modeling options reflecting model uncertainty for specific modeling steps,
as described in Section 7.1. For the screening-level ecological risk assessment, the key sources
of modeling uncertainty should be those related to future predictions of concentrations of Pb in
environmental media and the assumptions or models used to develop media-specific ecotoxicity
benchmarks (e.g., EPA's sediment criteria based on the equilibrium partitioning approach).
Exposure parameters for which uncertainty is likely to be key include the emissions input
parameters and various exposure parameter values that are built into the ecotoxicity benchmarks
(e.g., for Eco-SSLs for birds and mammals, food ingestion rates, diet selection; for EPA
equilibrium-based sediment benchmarks, values for SAV, FOC, and TOC) and the applicability
of the species represented in the ecotoxicity benchmarks to a given case study site location.
Values for these parameters are both variable (if they were known with accuracy) and uncertain
(given that we don't know the true distribution of values or even have good estimates of mean
values for a parameter for given species and locations). Data from a variety of sources can be
used to bound the possible parameter values in general or for a specific case study location.
At a screening level, the impact of using more stringent compared to less stringent toxicity
reference values could be evaluated to develop both minimum and maximum estimates of risk.
For soils, these could be based on NOAEL to LOAEL-based toxicity reference values; for
sediments these could be based on TECs to PECs; and for surface waters, these could be based
on AWQC calculated for a water hardness of 50 to 200 mg/L CaCOs. Also, for soils, the impact
DRAFT-May 31,2006 48
-------
of different parameters used to calculate the Eco-SSL (e.g., for birds and mammals, soil
ingestion rate, food ingestion rate, and composition of diet) on the risk results could be
examined. Possible variability in Pb concentrations for each environmental medium could be
assessed on the basis of available data (e.g., are there enough soil or surface water/sediment
sampling sites for the case study location to characterize spatial and/or temporal variability in Pb
concentrations in these media) or discussed qualitatively.
Additional sources of uncertainty that would require qualitative discussion include uncertainty
about the presence or absence of most susceptible or sensitive species that are the basis of the
ecotoxicity benchmarks at a particular site, proportion of a population of plants or animals that
might be exposed at or above different Pb concentrations, and whether local pH levels and
organic content are much different than those associated with the experiments on which the
ecotoxicity benchmarks are based.
DRAFT-May 31,2006 49
-------
9.0 ECOLOGICAL RISK ANALYSIS PLAN
9.1 Tier 1: Screening Level Analysis
9.1.1 Overview
To establish to potential for adverse effects to ecosystems from ambient Pb, it is necessary to
first characterize current levels of Pb in environmental media (air, soil, freshwater, and
sediment). A broader analysis would include terrestrial systems, aquatic systems, marine
systems, and estuarine/wetland systems. Given the available data and the potential for effects
from Pb, this analysis focuses on current and future concentrations of Pb in terrestrial soils as
well as concentrations in ambient air. We are currently evaluating the availability of data and the
feasibility of models to characterize aquatic exposures using freshwater and sediment. Evidence
presented in the CD suggests that, in highly impacted areas at least, soil concentrations do not
change over small time intervals even if sources of Pb are removed or emissions decreased.
Given advances in modeling and larger data sets, this analysis attempts to test and expand this
assumption by applying more recent data, existing screening tools, and current models to look at
areas with high media concentrations of Pb as well as more typical areas in which Pb emissions
may be very low.
Current media concentrations for specific case study locations can then be compared to
established ecotoxicity screening benchmarks for soils (e.g., Ecological Soil Screening Levels or
Eco-SSLs), freshwater (e.g., ambient water quality criteria for the protection of aquatic life, or
AWQC), and sediments (e.g., sediment quality criteria for the protection of aquatic life) to select
locations for analysis in which there appears to be the potential for adverse effects from Pb and
to identify which media and receptor combinations, if any, are likely to be adversely impacted by
current and future Pb concentrations given current ambient air conditions. Once current Pb
concentrations are known for a specific location, future concentrations may be predicted based
on various air quality scenarios to determine what happens to Pb concentrations in media over
time given changes to Pb concentrations in ambient air.
9.1.2 Data Sources for Determining Media Concentrations
Empirical data sources for current ambient levels of Pb in air, soil, freshwater, and sediment will
be needed to establish current media conditions. Modeling will be used to estimate future media
concentrations given various N A AQS-relevant scenarios. The CAS AC review of the 1989
exposure and risk assessment recommended that a constant soil concentration should be used in
predicting future scenarios. The plan detailed here considers using current methods for
estimating the relationship between soil Pb and air concentrations for future scenarios.
Air:
Sources of data for air emissions and monitoring data for selected locations will be used to
establish current air conditions as discussed in Section 4.1.1. As described in the AQCD, there
are a number of monitoring networks that may be useful for this purpose, and we will also
investigate the availability of additional data in particular locations (e.g., near sources of
interest).
DRAFT-May 31,2006 50
-------
Soils:
Relevant literature will be scanned to determine the most reasonable source of data on current
soil concentrations for a given study location. Depending on the study location there are several
options for soil data (e.g., regional and local entities, soil monitoring protocols for primary and
secondary sources, and urban areas monitoring programs). Data sources might be needed for
"background" Pb concentrations in areas removed from historical and current anthropogenic Pb
sources as well.
In developing the Eco-SSLs, EPA compiled data on metals concentrations in soils across the
United States (EPA 2005a; Appendix A of Attachment 1-4) in areas that might be considered
"background". The data sources for Pb in Appendix A include a variety of county- and state-
wide assessments conducted using a variety of different analytic techniques and sampling
strategies. National-level assessments have been conducted by the US Geological Survey
(USGS).
The USGS data primarily were collected in the 1970's and therefore were taken before the
removal of Pb in gasoline; however, collection sites were far from urban areas and other point
sources of Pb and also were at least 100 m from any road. Preliminary review of this data set
indicates that the higher soil Pb concentrations in the US (i.e., 25 to 30 ppm) tend to be localized
in certain areas, including the western mineral mining areas, but high Pb concentrations also are
present in areas of the northeast. Given that Pb is very stable in soil, these data are probably still
the best available for large areas of the country.
For screening-level ecological risk assessments for areas in the vicinity of selected case study
sites, the county- and state-wide soil assessments presented in the Eco-SSL documentation
(Appendix A of Attachment 4-1) may provide additional information, depending on the location.
Staff currently is assessing the degree to which the USGS soil sampling data, from the 1970s,
may or may not be representative of current concentrations. For this data set, only 14 percent of
samples fell below the detection limit of 10 ppm, which is very close to the Eco-SSL for birds of
11 ppm based on the woodcock (see Section 9.1.5.1). More than 99 percent of these background
soil concentrations for Pb are less than the next highest Eco-SSL for birds of 46 ppm for the dove
(see Section 9.1.5.1).
Freshwater Surface Waters
Ambient water concentrations of Pb may be found in several national datasets which are
currently being evaluated for this analysis, including both EPA and USGS datasets.
The EPA STORET database from the EPA Office of Water includes the water quality sampling
data developed by states to determine compliance with water quality standards and reported to
EPA annually, sampling data from Superfund sites, and data from other sources (e.g., U.S. Army
Corps of Engineers). EPA's EnviroMapper can be used to focus on localized geographic areas to
identify sampling locations and to download requested water quality data for those locations.
Limitations of STORET data include differences in sampling density and detection limits across
states. For Pb, a limitation is that many states do not attempt to measure dissolved Pb
DRAFT-May 31,2006 51
-------
concentrations, only total Pb concentrations, because most of the Pb in most surface waters is
sorbed to particulate matter in the water column or is in the sediments. (The AWQC for Pb is
based on dissolved (bioavailable) Pb, not total Pb). For the states that do sample for dissolved
Pb, the analytic techniques used do not detect any dissolved Pb in most samples (exceptions near
urban areas and mining sites). Moreover, the detection limit for dissolved Pb varies by state and
some states use analysis methods with detection limits at or above the chronic AWQC for Pb in
freshwater (2.5 ug/L at a water hardness of 100 mg/L as CaCOs). Preliminary review of data in
STORET indicate exceedances of the AWQC for aquatic life at only a few locations, principally
in the midwest near mining sites and at some urban or industrial locations.
The USGS maintains water quality data in two separate data sets: the National Water
Information System (NWIS) Network and the National Water Quality Assessment (NAWQA)
Data Warehouse. The NWIS system was originally developed for water flow data, and includes
data from 1.5 million sites across all 50 states, the District of Columbia, and Puerto Rico.
Chemical quality data have been added to it in recent years. The data have been compiled from a
variety of projects ranging from national level studies to small watershed projects up through
2004; therefore, the sampling methods, density of sampling sites, and detection limits are
variable across the data set. Lead has been an analyte at some of the sites; however, how well
different geographic areas are represented for Pb is not known at this time and might be difficult
to determine.
The NAWQA Data Warehouse includes the sampling data from the NAWQA program that
started systematically collecting chemical, biological, and physical water quality data from 42
study unit basins across the nation in 1991 (data compiled through 9/30/2004 at this time).
Basins from all regions of the United States are included; however, only approximately 50
percent of the land base is covered by these basins. Lead is one of the analytes in the program,
and measurements of dissolved and total Pb are available from most locations. An analysis of
the data in NAWQ by EPA, the current draft AWQCD found a total of 3,445 measurements of
dissolved Pb in surface waters, for which 86 percent were non-detects. When looking at a subset
of those data determined to be "natural" or background areas, of 430 samples for dissolved Pb in
surface waters, 88 percent were nondetects. The mean and upper 95th percentile concentrations
of dissolved Pb in the total sample were 0.66 and 1.10 ug/L, respectively, both of which are less
than the most stringent chronic AWQC of 1.2 ug/L (at a water hardness of 50 mg/L CaCOs).
The mean and upper 95th percentile concentrations of dissolved Pb in the natural samples were
slightly lower, 0.52 and 0.50 ug/L, respectively, indicating a skewed distribution of Pb
concentrations, as is expected. Thus, risks for aquatic biota in the water column, as assessed by
exceedance of the AWQC for Pb (see Section 9.1.5.2), are expected at less than 5 percent of the
NAWQA sampling stations nationwide. Maximum dissolved Pb concentrations for all sites and
for the subset of natural sites were 29.8 and 8.4 ug/L, both of which exceed the AWQC. With
the latitude and longitude of sampling locations known for both databases, it is possible to use
existing GIS data layers to identify the local land uses and other relevant river basin/watershed
data in the vicinity of the sampling locations that show exceedances of AWQC.
Of the three available surface water quality databases, STORET and NWIS include samples from
more locations in the US than does the NAWQA data set, but the sampling techniques and
sampling methods represented in STORET and NWIS are inconsistent from site to site. The
DRAFT-May 31,2006 52
-------
NAWQA data set provides representative, although not complete, coverage of the US and a
consistent approach to sampling and analysis of the elements assessed. Thus, the NAWQA data
set may be more appropriate for a national-level ecological risk screening assessment than the
other two data sets. For the assessment of ecological risks at selected case study locations, the
availability of data from NAWQA, STORET, and NWIS could be assessed, in that order.
Sediments:
Sediment data are available from the USGS NAWQA program as well as from the EPA
STORET database. The USGS program database includes Pb concentrations in sediments
measured using consistent sampling techniques from 42 river basin areas across the
conterminous United States. The STORET sediment Pb data, while providing more complete
coverage of the United States, again vary from state to state with respect to density of sampling
locations, sampling methods, and detection limits. With the latitude and longitude of sampling
locations known for both databases, it is possible to use existing GIS data layers to identify the
local land uses and other relevant river basin/watershed data.
An analysis of the data in the NAWQA Data Warehouse by EPA, the AWQCD found a total of
1,466 measurements of Pb in bulk sediments with a grain size less than 63 um. Of the total, only
0.48 percent of all samples were non-detect. Looking at a subset of 258 of those data determined
to be from "natural" areas, 1.2 percent were nondetects. The mean and upper 95th percentile
concentrations of Pb in bulk sediments for all samples were 120 and 162 ug/g dry wt,
respectively. The mean and upper 95th percentile concentrations of Pb in the natural samples
were slightly lower, 109 and 162 ug/L, respectively. Using the "Consensus-based" sediment
quality guidelines for Pb of 36 mg/kg dry weight as a concentration that is unlikely to cause
adverse effects in benthic organisms (threshold-effect concentrations or TECs) and 130 mg/kg
dry weight as a concentration that is likely to cause adverse effects on sediment-dwelling
organisms (probable-effect concentrations or PECs) (MacDonald et al. 2003) (see Section
9.1.5.3), it is apparent that a much higher proportion of the samples ofPb concentrations in
sediments in the NAWQA data set exceed one or both of the sediment quality guidelines for the
protection of benthic life than do the surface water concentrations exceed AWQC. The
consensus-based guidelines were developed by the USGS in cooperation with EPA and other
federal, state, and local agencies, as discussed in Section 9.1.5.3.
9.1.3 Case Study Selection
The challenge in selecting locations for analysis are that the ideal location would be 1)
ecologically relevant (contain significant ecological resources such as forests, water bodies, and
known populations of specific receptors used in the screening levels); 2) contain complete
empirical data sets from air monitoring, media sampling, and land use data; and 3) contain a
known air source with good data on emissions. It is also important to analyze a location that is
highly impacted by Pb emissions (exceeds ecotoxicity screening values) to be certain that the
long term effects at this atypically high level of deposition be captured as well as a site(s) that are
more typical. However, it is possible that at such sites, existing Pb concentrations in soils and
sediments are sufficiently high as to be relatively insensitive to alternative NAAQS scenarios.
Therefore, locations in areas where current Pb contamination is close to levels that might reflect
DRAFT-May 31,2006 53
-------
a threshold for adverse ecological effects might be more informative in analyzing alternative
NAAQS. To conserve time and resources, locations with higher exposures to Pb, as described in
section 2.0, will be parameterized as much as possible to include ecologically relevant criteria
and thereby used in the welfare assessment. In addition, 1-2 locations that are more typical of
low level exposures or alternate ecological scenarios might be modeled in tier 2 as well.
9.1.4 TRIM.FaTE Model
In addition to the MPE model discussed in section 4.1.3, TRIM-FaTE is being considered to
model both future media concentrations and intake rates for susceptible receptors. TRIM is a
time series modeling system with multimedia capabilities for assessing human health and
ecological risks from hazardous and criteria air pollutants developed by the US EPA. The
Environmental Fate, Transport, and Ecological Exposure module, TRIM.FaTE, accounts for
movement of a chemical through a comprehensive system of discrete compartments (e.g., media
and biota) that represent possible locations of the chemical in the physical and biological
environments of the modeled ecosystem and provides an inventory, over time, of a chemical
throughout the entire system. TRIM.FaTE is a mass-balanced based multimedia model, with
broad flexibility in spatial, temporal and simulation design complexity.
In the tier 1 analysis, this model (or MPE) may be used to load various media over time under
current ambient air conditions to estimate concentrations of Pb in soil, freshwater, and sediment
for case study locations. The model will need to be parameterized for Pb and it requires inputs
for current media concentrations and air emissions/deposition rates for each location. Data for
emissions and deposition rates will be similar to those discussed in earlier sections of this plan.
The contents of the biota compartments may be simplified to provide only media concentrations
not biotic concentrations for this tier of the analysis.
9.1.5 Determination of Potential Adverse Effects
9.1.5.1 Ecological Soil Screening Levels
Ecological Soil Screening Levels (Eco-SSLs) were developed by the EPA (2005a,b) as
concentrations of contaminants in soil that are protective of ecological receptors that commonly
come into contact with soil or ingest biota that live in or on soil (Table 9-1). They are derived
separately for four groups of ecological receptors: plants, soil invertebrates, birds and mammals
and as such, are presumed to provide protection of terrestrial ecosystems. Several species in
each receptor category were assessed to develop the Eco-SSL for that category. In general, the
same toxicity reference value (TRV) applies (e.g., the reference dose for the most sensitive of the
adverse ecological effects on birds) to all species in each receptor category; however, differences
in chemical intake associated with foraging techniques and diet generally result in different Eco-
SSL values, expressed as a soil concentration, for the different receptors (e.g., bird species) in
each receptor category. Scenarios were developed for each of the representative species, and the
most sensitive organism for which the Eco-SSL was lowest was chosen in each category to
establish the Eco-SSL. They are conservative values and are intended to be applied at the
screening stage of an ecological risk assessment.
DRAFT-May 31,2006 54
-------
Table 9-1. Ecological Soil Screening Levels for Lead for Various Receptors
(EPA 2005b)
Receptor
Plants
Soil Invertebrates
Birds
Mammals
Soil Concentration (mg Pb/kg soil
dry weight)
120
1,700
11
56
In this analysis, Eco-SSLs are intended to be used as screening values to focus further analyses
on receptors that are likely to be exposed to soils that exceed the relative Eco-SSL for that
receptor, thereby indicating the potential for adverse effects to that receptor. It should be noted
that the value given for birds (11 mg/kg dry weight in soil) is likely to be exceeded currently by
most soils in the United States. Therefore, this value is not useful in focusing the analysis or
looking at the impact of current ambient air on this receptor. As the avian value is too
conservative to reflect conditions in the real world, there are three options for the purpose of this
analysis: 1) ignore birds in the analysis and use other receptors as necessary, 2) use another bird
value from the original Eco-SSLs for Pb to develop the Eco-SSL for birds (e.g., the next highest
avian Eco-SSL of 46 mg/kg soil dry weight for a dove), or 3) recalculate value for the current
bird species, the woodcock, based on more realistic diet composition (not 100 percent
earthworms), food intake rates and incidental soil intake rates for woodcock (both are somewhat
higher in the calculation of the Eco-SSL for woodcock than provided in EPA's 1993 Wildlife
Exposure Factors Handbook; EPA 1993a,b), and a best estimate of earthworm absorption of Pb,
rather than an upper 95th percentile estimate of earthworm absorption of Pb.. The second option
may be most useful for this analysis as it would rely on peer-reviewed Eco-SSLs and woodcock
are found only in the eastern half of the United States and are not found in all habitat types there.
Recalculation of the woodcock Eco-SSL using more realistic exposure assumptions for the
woodcock would also be considered for an ecological risk assessment for areas in the eastern
half of the United States for which the conservative Eco-SSL value for woodcock is exceeded.
9.1.5.2 Ambient Water Quality Criteria
Ambient Water Quality Criteria (AWQC) were developed by the EPA to provide guidance to
states and tribes to use in adopting water quality standards. AWQC values for Pb are given for
chronic and acute exposures as well as for freshwater and marine environments (EPA 1985).
The freshwater criteria depend on water hardness. The values provided in Table 9-2 for
freshwater are based on a water hardness of 100 mg/L as CaCO3; while the values for saltwater
are not dependent on any water characteristics. The current equations/values may be reissued
based on pH at the conclusion of the current revision scheduled to be completed in 2007. These
are values based on toxicity testing in aquatic organisms.
Table 9-2. Recommended Ambient Water Quality Criteria
Freshwater (|ig Pb/L)
Acute
65
Chronic
2.5
Saltwater (|ig Pb/L)
Acute
210
Chronic
8.1
DRAFT-May 31,2006 55
-------
In this analysis AWQC will be used to evaluate the potential for adverse effect to freshwater
organisms (exposed via the water column) over time under the current ambient air levels.
9.1.5.3 Sediment Criteria
Sediment screening-level (and other) criteria or benchmarks have been established by EPA
regions and at the state level for Pb. Some work also has been done at the national level (EPA
2005c). For the EPA national sediment quality benchmarks, equations based on an equilibrium
partitioning approach have been established to relate the Pb concentration benchmark to pH (acid
volatile sulfide or AVS), fraction organic carbon (FOC), and total organic carbon (TOC). These
benchmarks also assess three levels of effect on benthic invertebrates: no-effect, threshold-effect,
and probable-effect concentrations. The dependence of these benchmarks on values for AVS,
FOC, and TOC mean that appropriate data for their calculation might or might not be available
for a given site. That dependence also means that the values are not readily useful for screening
national-level data to identify locations where sediment concentrations might exceed the
threshold- or probable-effect levels.
Another source of national-lev el sediment quality benchmarks is from the cooperative
Freshwater Sediment Quality Assessment Initiative that began in 2000 and that included the
Florida Department of Environmental Protection (FDEP), the EPA, and the USGS, with
participants from the US Fish and Wildlife Service (USFWS), National Oceanic and
Atmospheric Administration (NOAA), consultants, academics, county governments, and water
management districts (MacDonald et al. 2003). The Initiative recommended use of a
"Consensus approach" to developing threshold- and probable-effects concentrations (TEC and
PEC values, respectively) of metals in sediments. The approach basically consisted of
calculating a geometric mean sediment concentration from, in the case of Pb, sediment values
from five different approaches to estimating both the TEC and PEC. These values can be used in
the absence of data on AVS, FOC, and TOC; however, it is likely that they will be less
"accurate" for sediments for which the AVS, FOC, and TOC values are far from typical values.
Note that the screening-level sediment quality benchmarks representing a threshold for effects
(not a probable-effect level) for EPA Regions 4, 5, and 6, Environment Canada, and EPA's
Assessment and Remediation of Contaminated Sediments (ARCS) Program are all between 30
and 36 mg Pb/kg dry weight.12 Given that mean sediment concentrations measured in the
NAWQA program are well above these levels, at probable-effect level may be more appropriate
for the screening ecological risk assessment. As an initial screen for a large number of sites, the
Consensus-based benchmarks may be easier to use because additional data on AVS, FOC, and
TOC are not needed. For specific case study locations, once selected, EPA's equilibrium
partitioning approach-based values, based on site-specific values for AVS, FOC, and TOC, may
be more appropriate for comparison with current and future predicted concentrations of Pb in
sediments.
12 Available from the Risk Assessment Information System (RAIS) (see
http://risk.lsd.ornl.gov/homepage^enchmark.shtml).
DRAFT-May 31,2006 56
-------
9.2 Tier 2: Analysis of Sensitive Receptors
9.2.1 Overview
The second tier analysis consists of more detailed modeling of intake rates or body/tissue
concentrations for receptor(s) identified for potential adverse effects in the tier 1 analysis and
may include additional case study locations based on empirical data. This step will provide less
conservative estimates of the potential for effect based on published concentration effects for
specific receptors, if such data is available, and based on more realistic exposure factors (e.g.,
dietary composition, food ingestion rates) and possibly a less conservative threshold for effects.
For example, the current Eco-SSL toxicity reference values (TRVs) for birds and mammals are
based on a geometric mean of no-observed-adverse-effect levels (NOAELs) for the most
sensitive of the effects on reproduction or growth. It may be that use of a lowest-observed-
adverse-effect level (LOAEL) for such effects may be more representative of a threshold for
population-level effects in the most susceptible bird or mammal populations.
9.2.2 Estimation of Exposure for Sensitive Receptor(s)
9.2.2.IDetailed Modeling of Intake or Body/Tissue Concentration
TRIM.FaTE may be used to provide a more detailed picture of actual intake rates or body tissue
concentrations expected in sensitive receptor(s) given a specific soil (or surface water or
sediment) concentration. This step may be undertaken for those locations in the tier one
screening analysis where there is the potential for adverse effect to a specific receptor (screening
values are exceeded either currently or in the future). The model will be more fully developed to
include more biotic compartments and tailored to produce either intake rates or body/tissue
concentrations depending on the metric used by published concentrations effects studies for that
receptor.
9.2.2.2 Comparison to Known Concentration Effects
Reference toxicity values expressed as a daily dose (daily intake via ingestion) have been
developed for the Eco-SSLs for birds and mammals. Those values are based on NOAELs, rather
than LOAELs, for reproduction and growth. The corresponding LOAEL data are provided,
however, and could be analyzed for use in this context.
If other receptor groups turn out to be more at risk than birds and mammals, in support of a Tier
2 analysis, we will identify appropriate toxicity benchmarks from other sources or perform a
review of scientific literature to identify concentration effects studies relevant to these other
sensitive receptor(s) chosen in the tier one analysis. Specific known physiological effects, if any,
will be discussed based on likely intake rates or body/tissue concentrations resulting from the
future media concentrations. This step is dependent on the availability of data on specific effects
to specific receptors (or classes of receptors).
DRAFT-May 31,2006 57
-------
10.0 REFERENCES
10.1 Human Health Risk Assessment
Akesson. A, Lundh, T. Vahter, M. Bjellerup, P. Lidfeldt, J. Nerbrand, C. Samsioe, G. Stromberg,
U. Skerfving. S. 2005. Tubular and Glomerular Kidney Effects in Swedish Women with Low
Environmental Cadmium Exposure. Environ Health Persp 113(11): 1627-1631.
Bellinger, D.C., Stiles, K.M. Needleman, H.L.1992. Low-level lead exposure, intelligence and
cacademic achievement: a long-term follow-up study. Pediatrics 90: 855-861.
Bowers, TS, Beck BD, Karam HS. 1994. Assessing the relationship between environmental lead
concentrations and adult blood lead levels. Risk Anal. 14: 183-189.
Canfield, R.L., Henderson, C.R., Jr., Cory-Slechta, D.A., Cox, C., Jusko, T.A., and Lanphear,
B.P. 2003. Intellectual Impairment in Children With Blood Lead Concentrations Below 10
Microg Per Deciliter. N Engl J Med 348:1517-1526.
Cheng, Y., Schwartz, J., Sparrow, D., Aro, A., Weiss, S.T., and Hu, H. 2001. Bone Lead and
Blood Lead Levels in Relation to Baseline Blood Pressure and the Prospective Development of
Hypertension: the Normative Aging Study. Am J Epidemiol 153:164-171.
Cockcroft, D.W., Gault M.H. 1976. Prediction of creatinine clearance from serum creatinine.
Nephron 16:31-41.
Den Hond, E., Nawrot, T., Staessen, J. A. 2002. The relationship between blood pressure and
blood lead in NHANES III. J. Hum. Hypertens. 16: 563-568.
EPA. 1986. Air Quality Criteria for Lead. EPA-600/8-83/028dF. Environmental Criteria and
Assessment Office, Research Triangle Park, NC.
EPA. 1989. Review of the National Ambient Air Quality Standards for Lead: Exposure Analysis
Methodology and Validation. OAQPS Staff Report. EPA-450/2-89-011. Office of Air Quality
Planning and Standards, Research Triangle Park, NC.
EPA. 1990. Review of the National Ambient Air Quality Standards for Lead: Assessment of
Scientific and Technical Information. OAQPS Staff Paper. EPA-450/2-89-022. Office of Air
Quality Planning and Standards, Research Triangle Park, NC.
EPA. 1994. Technical Support Document: Parameters and Equations Used in the Integrated
Exposure Uptake Biokinetic Model for Lead in Children (v.099d). Office of Solid Waste. EPA
540/R-94/040.
EPA. 1995. AP 42, Fifth Edition: Compilation of Air Pollutant Emission Factors, Volume 1:
Stationary Point and Area Sources: Section 13 (Miscellaneous Sources); Section 13.2.5
(Industrial Wind Erosion).
DRAFT-May 31,2006 58
-------
EPA. 1996a. Soil Screening Guidance: User's Guide. Office of Solid Waste and Emergency
Response. Publication 9355.4-23
EPA. 1996b. Recommendations of the Technical Review Work Group for Lead for an Interim
Approach to Assessing Associated with Adult Exposures to Lead in Soil. Office of Soil Waste
and Emergency Response.
EPA. 1997. Exposure Factors Handbook. Volume 2: Food Ingestion Factors. EPA/600/P-
95/002Fa. Washington, DC: Office of Research and Development, National Center for
Environmental Assessment, http://www.epa.gov/ncea/pdfs/efh/front.pdf
EPA. 1998a. Methodology for assessing health risks associated with multiple pathways of
exposure to combustor emissions. Update to EPA/600/6-90/003, EPA/NCEA (EPA 600/R-
98/137).
EPA. 1998b. Risk Analysis to Support Standards for Lead Paint, Dust and Soil. Office of
Pollution Prevention and Toxic Substances. EPA 747-R-97-006.
EPA. 2000. Risk Analysis to Support Standards for Lead in Paint, Dust, and Soil, Supplemental
Report, Volume 1. Technical Review Workgroup. EPA 747-R-00-004.
EPA. 2002a. User's Guide for the integrated Exposure Uptake Biokinetic Model for Lead in
Children (IEUBK) Windows version - 32 bit version. EPA 540-K-01-005. OSWER #9285.7-42,
Office of Solid Waste and Emergency Response, Washington, DC.
EPA. 2002b. Reference Manual: Documentation of Updates for the Integrated Exposure Uptake
Biokinetic Model for Lead in Children (IEUBK) Windows version - 32 bit version. OSWER
#9285.7-44, Office of Solid Waste and Emergency Response, Washington, DC.
EPA. 2003. Guidance for Developing Ecological Soil Screening Levels. Office of Solid Waste
and Emergency Response. OSWER Directive 9285.7-55.
EPA. 2004a. Human Exposure Measurements: The National Human Exposure Assessment
Survey. U S Environmental Protection Agency. Available from:
http ://www. epa. gov/heasd/edrb/nhexas .htm.
EPA. 2004b. Summary: Lead Action Level Exceedances for Medium (3,300-50,000) and Large
(>50,000) Public Water Systems.
EPA. 2005a. Air Quality Criteria for Lead (First External Review Draft). U S Environmental
Protection Agency. Available from:
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=141779.
EPA. 2005b. Economic Analysis for the Renovation, Repair and Painting Program Revised Rule
(Review Draft). Office of Pollution Prevention and Toxics.
DRAFT-May 31,2006 59
-------
EPA. 2006a. Plan for Review of the National Ambient Air Quality Standards for Lead. Office of
Air Quality Planning and Standards, Research Triangle Park, NC.
EPA. 2006b. Air Quality Criteria for Lead. 2nd External Review Draft. May 2006 draft.
National Center for Environmental Assessment, Research Triangle Park.
Falerios, M., Schild, K., Sheehan, P., Paustenbach, DJ. 1992. Journal of the Air and Waste
Management Association, 42:40-48.
FDA. 2004. Data on lead concentrations in food from data bases at
http://www.cfsan.fda.gov/~comm/tds-toc.html
Filippelli, G.M., Laidlaw, M.A.S., Latimer, J.C., and Raftis, R. 2005. Urban Lead Poisoning
and Medical Geology: An Unfinished Story. GSA Today 15(1):4-11.
Harris, A. and Davidson, C. 2005. The Role of Resuspended Soil in Lead Flows in the
California South Coast Air Basin. Environmental Science and Technology. 39:7410-7415.
Jacobson, DE, Clickner, RP Zhou, JY Viet, SM Marker, DA Rogers, JW Zeldin, DC Broene, P
Friedman, W. 2002. The Prevalence of Lead-based Paint Hazards in U.S. Housing.
Environmental Health Perspectives. 110(10):A599-A606.
ICF Consulting. 2005. White Paper on the Advances in Knowledge Concerning Lead
Exposures, Body Burdens, and Adverse Effects Since the 1991 Lead and Copper Rule: Options
for Reevaluation of the Drinking Water Action Level for Lead. Submitted to the Health Effects
Criteria Division, Office of Water, U.S. Environmental Protection Agency.
ICF Consulting. 2006. Memorandum: "Preliminary Findings on Non-Urban Background Lead
Concentrations in Soil and Surface Waters," to Ginger Tennant, OAQPS, from Margaret McVey
and Mercedes Bravo, ICF Consulting, April 7.
Kim, R., Rotnitsky, A. Sparrow, D. Weiss, S.T. Wager, C. Hu, H. 1996. A longitudinal study of
low-level lead exposure and impairment of renal functkon. The Normative Aging Study. J. Am.
Med. Assoc. 275:1177-1181.
Kordas, K. et al. 2006. Deficits in Cognitive Function and Achievement in Mexican First-
Graders with Low Blood Lead Concentrations. Environmental Research 100:371-386.
Lanphear, B.P., Matte, T.D., Rogers, J., Clickner, R.P., Dietz, B., Bornschein, R.L., Succop, P.,
Mahaffey, K.R., Dixon, S., Galke, W., Rabinowitz, M., Farfel, M., Rohde, C., Schwartz, J.,
Ashley, P., Jacobs, D.E. 1998. The Contribution of Lead-Contaminated House Dust and
Residential Soil to Children's Blood Lead Levels: A Pooled Analysis of 12 Epidemiologic
Studies. Environmental Research 79:51-68.
DRAFT-May 31,2006 60
-------
Lanphear, B.P., Hornung, R., Khoury, J., Yolton, K., Baghurst, P., Bellinger, D.C., Canfield,
R.L., Dietrich, K.N., Bornschein, R., Greene, T., Rothenberg, S.J., Needleman, H.L., Schnaas,
L., Wasserman, G., Graziano, J., and Roberts, R. 2005. Low-Level Environmental Lead
Exposure and Children's Intellectual Function: an International Pooled Analysis. Environ Health
Perspect 113:894-899.
Leggett,R.W. 1993. An Age-Specific Kinetic Model of Lead Metabolism in Humans. Environ
Health Perspect 101:598-616.
Maddaloni, M., Bellew, M., Diamond, G., Follansbee, M., Gefell, D., Goodrum, P., Johnson, M.,
Koporec, K., Khoury, G., Luey, J., Odin, M., Troast, R., Van, L.P., and Zaragoza, L. 2005.
Assessing Lead Risks at Non-Residential Hazardous Waste Sites. Human and Ecological Risk
Assessment 11:967-1005.
Mickle, M. 1998. Structure, Use and Validation of the IEUBK Model. Environ. Health Pers.
106(Suppl 6): 1531-1534
Muntner, P., He, J. Vupputuri, S., Coresh, J. Batuman, V. 2003. Blood lead and chronic kidney
disease in the general United States population: results from NHANES III. Kidney Int. 63: 1044-
1050.
Nash, D., Magder,L., Lustberg,M., Sherwin,R.W., Rubin,RJ., Kaufmann,R.B., and
Silbergeld,E.K. 2003. Blood Lead, Blood Pressure, and Hypertension in Perimenopausal and
Postmenopausal Women. JAMA 289:1523-1532.
National Research Council. 2005. Superfund and Mining Megasites: Lessons from the Coeur
D'Alene River Basin. Committee on Superfund Site Assessment and Remediation in the Coeur
d'Alene River Basin. Board on Environmental Studies and Toxicology. National Academies
Press. Washington, DC. http://www.nap.edu/catalog/11359.html
Nawrot,T.S., Thijs,L., Den Hond,E.M., Roels,H.A., and Staessen,J.A. 2002. An
Epidemiological Re-Appraisal of the Association Between Blood Pressure and Blood Lead: a
Meta-Analysis. J Hum Hypertens 16:123-131.
O'Flaherty, E.J. 1993. Physiologically Based Models for Bone-Seeking Elements. IV. Kinetics
of Lead Disposition in Humans. Toxicol Appl Pharmacol 118:16-29.
Payton, M., Hu, H., Sparrow, D. Weiss, S.T.. 1994. Low-level lead exposure and renal function
in the normative aging study. Am. J. Epidemiol. 140:821-829.
Pirkle,J.L., Kaufmann,R.B., Brody,D.J., Hickman,!., Gunter,E.W., and Paschal,D.C. 1998.
Exposure of the U.S. Population to Lead, 1991-1994. Environ Health Perspect 106:745-750.
Pocock SJ, Shaper AG, Walker M, et al 1983. Effects of tap water lead, water hardness, alcohol,
and cigarettes on blood lead concentrations. J Epidemiol Commun Health 37: 1-7.
DRAFT-May 31,2006 61
-------
Pounds J.G., Leggett J. 1998. The ICRP Age-Specific Biokinetic Model for Lead: Validations,
Empirical Comparisons, and Explorations Environ.. Health Pers. 106(Suppl 6):1505-1511.
Proctor, S. P., Rotnitzky, A., Sparrow, D., Weiss, S. T., Hu, H. 1996. The relationship of blood
lead and dietary calcium to blood pressure in the normative aging study. Int. J. Epidemiol.
25:528-536.
Rodes, C., Sheldon, L., Whitaker, D., Clayton, A., Fitzgerald, K., Flanagan, J., DiGenova, F.,
Hering, S., Frazier, C. 1998. Measuring Concentrations of Selected Air Pollutants Inside
California Vehicles. Final Report of the California Air Resources Board, Contract No. 95-339.
December.
Rothenberg, S. J., Kondrashov, V., Manalo, M., Jiang, J., Cuellar, R., Garcia, M., Reynoso, B.,
Reyes, S., Diaz, M., Todd, A. C. 2002. Increases in hypertension and blood pressure during
pregnancy with increased bone lead levels. Am. J. Epidemiol. 156:1079-1087.
Sherlock, JC, Smart G, Forbes GI, et al. 1982. Assessment of lead intakes and dose-response for
a population in Ayr exposed to a plumbosolvent water supply. Human Toxicol 1:115-22.
Sherlock, JC, Ashby D, Delves HT, et al. 1984. Reduction in exposure to lead from drinking
water and its effect on blood lead concentrations. Human Toxicol 3: 383-92.
Succup P., Bornschein R., Brown K., Tseng C-Y. 1998. An Empirical Comparison of Lead
Exposure Pathway Models. Environ Health Perspect 106(Suppl 6): 1577-1583.
Staessen, J. A., Lauwerys, R.R., Buchet, J.-P., Bulpitt, C.J. Rondia, D. Van Renterghem, Y.
Amery. A. 1992. Impairment of renal function with increasing blood lead concentratinons in the
general population. N. Engl. J. Med. 327:151-156.
Teichman, J., Coltrin, D., Prouty, K., and Bir, W. A. 1993. "A Survey of Lead Contamination in
Soil Along Interstate 880, Alameda County, California." American Industrial Hygiene
Association 54(9):557 - 559.
Tellez-Rojo, M.M., Bellinger, D.C., Arroyo-Quiroz, C., Lamadrid-Figueroa, H., Mercado-
Garcia, A., Schnaas-Arrieta, L., Wright, R.O., Hernandez-Avila, M., Hu, H. 2006. Longitudinal
associations between blood lead concentrations <10 |ig/dL and neurobehavioral development in
environmentally-exposed children in Mexico City. Pediatrics: in press.
Tsaih, S.-W., Korrick, S. Schwartz, J. Amarasiriwardena, C. Aro, A. Sparrow, D. Hu, H. 2004.
Lead, diabetes, hypertension, and renal function: the Normative Aging Study. Environ. Health
Perspect. 112: 1178-1182.
Vupputuri,S., He, J., Muntner, P., Bazzano, L.A., Whelton, P.K., and Batuman, V. 2003. Blood
Lead Level Is Associated With Elevated Blood Pressure in Blacks. Hypertension 41:463-468.
DRAFT-May 31,2006 62
-------
10.2 Ecological Risk Assessment
EPA. 1977. Air quality criteria for lead. Research Triangle Park, NC: Criteria and Special
Studies Office; EPA report no. EPA/600/8 77/017. Available from NTIS, Springfield, VA; PB
280411.
EPA. 1985, Ambient aquatic life water quality criteria for lead. Washington, DC: Criteria and
Standards Division; EPA report no. EPA/440/5-84-027.
EPA. 1986a. Air quality criteria for lead. Research Triangle Park, NC: Office of Health and
Environmental Assessment, Environmental Criteria and Assessment Office; EPA report nos.
EPA/600/8 83/028A-F. Available from NTIS, Springfield, VA; PB87-142949.
EPA. 1986b. Lead effects on cardiovascular function, early development, and stature: an
addendum to the EPA Air Quality Criteria for Lead (1986). Research Triangle Park, NC: Office
of Health and Environmental Assessment, Environmental Criteria and Assessment Office.
EPA. 1989. Review of the national ambient air quality standards for lead: exposure methodology
and validation. Research Triangle Park, NC: Office of Air Quality Planning and Standards,
Ambient Standards Branch; EPA report no. EPA/450/2 89/011.
EPA. 1990. Summary of selected new information on effects of lead on health and supplement to
1986 air quality criteria for lead. Research Triangle Park, NC: Office of Health and
Environmental Assessment, Environmental Criteria and Assessment Office; EPA report no.
EPA/600/8-89. Available from NTIS, Springfield, VA; PB92-235670.
EPA. 1990. Review of the National Ambient Air Quality Standards for Lead: Assessment of
Scientific and Technical Information OAQPS Staff Paper. Research Triangle Park, NC: Office of
Air Quality Planning and Standards; EPA report no. EPA/450/2-89-022.
EPA. 1993. Wildlife Exposure Factors Handbook; Volumes I and II. Prepared by ICF
Consulting (now ICF International), Fairfax, VA. Prepared for EPA's Office of Research and
Development, Washington, DC. EPA report nos. EPA/600/R-93/187a,b.
EPA. 2005a. Guidance for Developing Ecological Soil Screening Levels. Washington, DC:
Office of Solid Waste and Emergency Response. OSWER Directive No. 9285.7-55. November
2003, revised February, 2005.
EPA. 2005b. Ecological Soil Screening Levels for Lead. Washington, DC: Office of Solid
Waste and Emergency Response. OSWER Directive No. 9285.7-70.
EPA. 2005c. Procedures for the Derivation of Equilibrium Partitioning Sediment Benchmarsk
(EBSs) for the Protection of Benthic Organisms: Metal Mixtures (Cadmium, Copper, Lead,
Nickel, Silver, and Zinc). Washington, DC: Office of Research and Development; EPA report
no. EPA-600-R-02-011.
DRAFT-May 31,2006 63
-------
Federal Register. 2004. Air Quality Criteria Document for Lead: Call for Information. F. R.
(November 9) 69: 64926-64928.
Federal Register. 1979. National primary and secondary ambient air quality standards: revisions
to the National Ambient Air Quality Standards for lead, F.R. (February 8) 44:8202- 8237.
Gustavsson, N., B01viken, B., Smith, D.B., and Severson, R.C. 2001. Geochemical Landscapes
of the Conterminous United States - New Map Presentations for 22 Elements. U.S. Geological
Survey Professional Paper 1648. Washington, DC: U.S. Department of the Interior. Available
online at: http://geology.cr.usgs.gov/pub/ppapers/pl648/
MacDonald, D.D., Ingersoll, C.G., Smorong, D.E., Lindskoog, R.A., Sloane, G, and Biernacki, T.
2003. Development and Evaluation of Numerical Sediment Quality Assessment Guidelines for Florida
Inland Waters. Technical Report. Prepared for: Florida Department of Environmental Protection,
Tallahasee, Florida. Prepared by: MacDonald Environmental Sciences, Lt, British Columbia, and US
Geological Survey, Columbia Missouri. January.
DRAFT-May 31,2006 64
------- |