&EPA
United States
EnviroimnU Protection
Agency
Lead: Human Exposure and Health Risk
Assessments for Selected Case Studies
Volume I. Human Exposure and Health Risk
Assessments - Full-scale
-------
-------
EPA-452/R-07-014a
October 2007
Lead: Human Exposure and Health Risk Assessments for
Selected Case Studies
Volume I. Human Exposure and Health Risk Assessments - Full-scale
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina
-------
DISCLAIMER
This document has been reviewed by the Office of Air Quality Planning and Standards,
U.S. Environmental Protection Agency (EPA), and approved for publication. This document has
been prepared by staff from the Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency. Any opinions, findings, conclusions, or recommendations are
those of the authors and do not necessarily reflect the views of the EPA Mention of trade names
or commercial products is not intended to constitute endorsement or recommendation for use.
Any questions or comments concerning this document should be addressed to Zachary Pekar,
U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards, C504-06,
Research Triangle Park, North Carolina 27711 (email: pekar.zachary@epa.gov).
-------
TABLE OF CONTENTS
List of Tables iv
List of Figures v
1 INTRODUCTION 1-1
1.1 MULTIMEDIA ASPECT OF THE RISK ASSESSMENT 1-1
1.2 RISK ASSESSMENT FROM LAST REVIEW 1-4
1.3 PILOT PHASE ASSESSMENT FOR THE CURRENT REVIEW 1-5
1.4 CASAC ADVICE 1-6
1.5 ORGANIZATION OF THE DOCUMENT 1-8
REFERENCES 1-9
2 DESIGN OF EXPOSURE AND RISK ASSESSMENTS 2-1
2.1 BACKGROUND INFORMATION ON LEAD EXPOSURE AND RISK .... 2-1
2.1.1 Sources, Pathways and Routes 2-3
2.1.2 At-risk Populations 2-4
2.1.3 Internal Disposition 2-6
2.1.4 Health Endpoints 2-7
2.1.4.1 Developing Nervous System 2-8
2.1.4.2 Adult Nervous System 2-10
2.1.4.3 Cardiovascular System 2-11
2.1.4.4 Renal System 2-12
2.1.4.5 Heme Synthesis 2-12
2.1.4.6 Immune System 2-13
2.1.5 Metric and Model for Risk Quantitation 2-14
2.2 USE OF CASE STUDIES AND LOCATION SELECTIONS 2-19
2.2.1 General Urban Case Study 2-20
2.2.2 Point Source Case Studies 2-20
2.2.2.1 Primary Pb Smelter Case Study 2-21
2.2.2.2 Secondary Pb Smelter Case Study 2-24
2.3 ASSESSMENT SCENARIOS 2-25
2.3.1 Air Concentrations 2-25
2.3.2 Policy-relevant Background 2-28
2.3.3 Outdoor Soil/Dust 2-29
2.4 ANALYTICAL APPROACH 2-31
2.4.1 Temporal Aspects 2-31
2.4.2 Spatial Scale and Resolution 2-31
-------
2.4.3 Categorization of Policy-relevant Exposure Pathways 2-32
2.4.4 Overview of Analytical Steps 2-34
2.4.4.1 Exposure Assessment 2-37
2.4.4.2 Risk Characterization 2-39
2.4.5 Variability Characterization 2-39
2.4.6 Uncertainty Characterization and Sensitivity Analysis 2-40
2.4.6.1 Performance Evaluations 2-41
2.4.6.2 Generating Multiple Sets of Results 2-41
2.4.6.3 Sensitivity Analysis 2-42
2.4.6.4 Qualitative Discussion of Sources of Uncertainty 2-43
REFERENCES 2-44
3 EXPOSURE ASSESSMENT 3-1
3.1 METHODS FOR ESTIMATING MEDIA CONCENTRATIONS 3-1
3.1.1 Ambient Air Concentrations 3-3
3.1.1.1 General Urban Case Study 3-3
3.1.1.2 Primary Pb Smelter Case Study 3-4
3.1.1.3 Secondary Pb Smelter Case Study 3-5
3.1.2 Inhalation Exposure Concentrations 3-5
3.1.3 Outdoor Surface Soil/Dust Concentrations 3-6
3.1.3.1 General Urban Case Study 3-6
3.1.3.2 Primary Pb Smelter Case Study 3-6
3.1.3.3 Secondary Pb Smelter Case Study 3-7
3.1.4 Indoor Dust Concentrations 3-8
3.1.4.1 General Urban Case Study 3-10
3.1.4.2 Primary Pb Smelter Case Study 3-12
3.1.4.3 Secondary Pb Smelter Case Study 3-14
3.2 METHODS FOR ESTIMATING BLOOD PB LEVELS 3-15
3.2.1 Blood Pb Modeling 3-15
3.2.1.1 Primary Analysis 3-16
3.2.1.2 Sensitivity Analysis 3-17
3.2.2 Exposure Pathway Apportionment and Probabilistic Population
Modeling 3-18
3.2.2.1 General Urban Case Study 3-20
3.2.2.2 Point Source Case Studies 3-20
3.2.3 GSD for Population Blood Pb Modeling Procedure 3-23
3.2.3.1 General Urban Case Study 3-25
3.2.3.2 Point Source Case Studies 3-26
3.3 ESTIMATED MEDIA CONCENTRATIONS 3-26
3.4 ESTIMATED BLOOD PB LEVELS 3-29
3.5 UNCERTAINTY CHARACTERIZATION AND SENSITIVITY
ANALYSIS 3-33
11
-------
3.5.1 Performance Evaluation Related to Exposure Media Modeling 3-33
3.5.1.1 Evaluation of Modeled Ambient Air Pb Concentrations 3-33
3.5.1.2 Evaluation of Modeled Outdoor Soil/Dust Pb
Concentrations 3-34
3.5.1.3 Evaluation of Modeled Indoor Dust Pb Concentrations 3-35
3.5.2 Performance Evaluation Related to Blood Pb Modeling 3-38
3.5.2.1 Evaluation of Candidate Blood Pb Models 3-38
3.5.2.2 Evaluation of model-derived outdoor air Pb-to-blood Pb
ratios 3-39
3.5.2.3 Comparison of modeled blood Pb levels to nationally
representative data 3-42
REFERENCES 3-46
4 RISK ASSESSMENT 4-1
4.1 METHODS FOR DERIVING RISK ESTIMATES 4-1
4.1.1 Concentration-Response Functions 4-1
4.1.1.1 Log-Linear Function with Cutpoint 4-3
4.1.1.2 Log-Linear Function with Low-Exposure Linearization 4-3
4.1.1.3 Two-piece Linear Function 4-4
4.1.2 Projection of Population Risk 4-5
4.2 RISK ESTIMATES 4-6
4.3 UNCERTAINTY CHARACTERIZATION AND SENSITIVITY
ANALYSIS 4-13
4.3.1 Qualitative Discussion of Key Sources of Uncertainty 4-13
4.3.2 Sensitivity Analysis 4-17
4.3.3 Performance Analyses 4-19
4.3.4 Uncertainty in Modeling Approaches - Multiple Sets of Results 4-21
REFERENCES 4-23
5 ADDITIONAL ANALYSES 5-1
5.1 DESIGN OF EXPOSURE AND RISK ASSESSMENTS 5-1
5.1.1 Assessment Scenarios 5-1
5.1.2 Analytical Approach 5-1
5.1.3 Location-Specific Urban Case Studies 5-2
5.2 EXPOSURE ASSESSMENT 5-4
5.2.1 Methods for General Urban and Primary Pb Smelter Case Studies 5-4
5.2.2 Methods for Location-specific Urban Case Studies 5-5
5.2.2.1 Ambient Air and Inhalation Exposure Concentrations 5-5
5.2.2.2 Other media Concentrations 5-6
5.2.2.3 Blood Pb levels 5-7
5.2.3 Media concentrations 5-7
5.2.3.1 Location-specific Urban Case Studies 5-7
in
-------
5.2.3.2 General Urban and Primary Pb Smelter Case Studies 5-8
5.2.4 Blood Pb levels 5-13
5.2.5 Uncertainty Characterization 5-16
5.2.5.1 Performance Evaluation of Modeled MediaConcentrations 5-16
5.2.5.2 Performance Evaluation of Modeled Blood Pb Levels 5-17
5.3 RISK ASSESSMENT 5-20
5.3.1 Methods for Deriving Risk Estimates 5-20
5.3.1.1 Concentration-response Functions 5-20
5.3.1.2 Project! on of Population Risk 5-24
5.3.2 Risk Estimates 5-24
5.3.2.1 Population Risk Distribution Estimates 5-25
5.3.2.2 IQ Loss Incidence Estimates 5-28
5.3.3 Uncertainty Characterization and Sensitivity Analysis 5-31
5.3.3.1 Qualitative Discussion of Key Sources of Uncertainty 5-31
5.3.3.2 Performance Analyses 5-33
5.3.3.3 Uncertainty in Modeling Approaches - Multiples Sets
of Results 5-34
5.3.3.4 Sensitivity Analysis - Indoor Dust Pb Modeling 5-34
REFERENCES 5-37
List of Tables
Table 2-1. Key aspects of primary Pb smelter case study 2-23
Table 2-2. Key aspects of secondary Pb smelter case study 2-24
Table 2-3. Air quality scenarios assessed for the general urban case study 2-28
Table 3-1. Case study approaches for estimating media Pb concentrations 3-2
Table 3-2. Hybrid model for indoor dust Pb in general urban case study 3-12
Table 3-3. Estimated annual ambient air concentrations 3-27
Table 3-4. Estimated inhalation exposure concentrations 3-28
Table 3-5. Estimated outdoor soil/dust concentrations 3-28
Table 3-6. Estimated indoor dust concentrations 3-29
Table 3-7. Summary of blood Pb estimates for medians in total-exposure blood Pb
distributions 3-31
Table 3-8. Summary of blood Pb estimates for 95th percentiles in total-exposure blood Pb
distributions 3-32
Table 3-9. Evaluation of model-predicted indoor dust Pb levels against empirical data obtained
from the literature 3-37
Table 3-10. Air-to-blood Pb ratios for "recent air" contribution to concurrent blood Pb level. 3-40
Table 3-11. Blood Pb levels for 7 year olds in the U.S. (interpolated from NHANES IV,
1999-2002) 3-44
Table 4-1. Comparison of total and incremental IQ loss estimates below 10 |ig/dL for the three
concentration-response functions 4-2
Table 4-2. Summary of risk estimates for medians of total-exposure risk distributions 4-11
IV
-------
Table 4-3. Summary of risk estimates for 95th percentiles of total exposure risk
distributions 4-12
Table 4-4. Impact of multiple sources of uncertainty on risk results 4-22
Table 5-1. Estimated annual average ambient air concentrations 5-9
Table 5-2. Estimated inhalation exposure concentrations 5-10
Table 5-3. Estimated outdoor soil/dust concentrations 5-11
Table 5-4. Estimated indoor dust concentrations 5-12
Table 5-5. Summary of blood Pb level estimates for median total blood Pb 5-14
Table 5-6. Summary of blood Pb level estimates for high-end total blood Pb 5-15
Table 5-7. Air-to-blood ratios derived by comparing air quality scenario air and blood Pb
estimates 5-19
Table 5-8. Comparison of total and incremental IQ loss estimates below 10 |ig/dL for the four
concentration-response functions 5-23
Table 5-9. Summary of risk estimates for medians of total-exposure risk distributions 5-26
Table 5-10. Summary of risk estimates for 95th percentiles of total exposure risk
distributions 5-27
Table 5-11. Incidence of children with >1 point IQ loss 5-29
Table 5-12. Incidence of children with >7 points IQ loss 5-30
Table 5-13. Comparison of hybrid indoor dust model with a modified form of the model 5-36
List of Figures
Figure 1-1. Principal pathways of human and ecological exposure toPb 1-3
Figure 2-1. Conceptual model for Pb human health risk assessment 2-2
Figure 2-2. Overview of analysis approach 2-36
Figure 2-3. Modeling approaches for case study analyses presented in chapters 3 and 4 2-42
Figure 3-1. Procedure for generating population blood Pb distributions for point source case
studies 3-22
Figure 3-2. Comparison of NHANES IV blood Pb levels with modeled estimates 3-44
Figure 4-1. Comparison of three concentration-response functions for concurrent
blood Pb levels< 10 |ig/dL 4-2
Figure 5-1. Core modeling approach for each case study 5-2
Figure 5-2. Comparison of four concentration-response functions for concurrent blood
Pb levels < 10 |ig/dL 5-23
-------
1 INTRODUCTION
This document is the first volume of the report Lead: Human Exposure and Health Risk
Assessments for Selected Areas. This volume describes the quantitative human exposure and
health risk assessments1 conducted to inform the U.S. Environmental Protection Agency's
(EPA's) current review of the National Ambient Air Quality Standards (NAAQS) for lead (Pb).
As with the last review of the Pb NAAQS (see Section 1.2), the human exposure and
health risk assessments (the risk assessment)2 for this review reflect multimedia exposure
pathways, and their influence on blood Pb levels as an internal index of exposure or dose
(Section 1.1). The assessment for this review, as with that for the last review, utilizes a case
study approach wherein a set of specific locations or case studies associated with policy-relevant
Pb exposures are evaluated in detail. The case studies have been selected to provide a
perspective on the nature and magnitude of air source Pb exposures and risk in the United States.
There are two phases to the risk assessment for the current review: pilot and full-scale.
The first phase (the pilot assessment, described in Section 1.3) was presented in the first draft
Staff Paper and accompanying technical report (USEPA, 2006a; ICF, 2006), and was the subject
of a review by the Clean Air Scientific Advisory Committee (CASAC) on February 6 and 7,
2007 described in Section 1.4 (Henderson, 2007a). The initial full-scale analyses were presented
in the July 2007 draft report (USEPA, 2007b) and were the subject of a CASAC review at a
public meeting on August 28 and 29, 2007. In response to CASAC recommendations described
in Section 1.4 (Henderson, 2007b), additional analyses, using a core modeling approach were
conducted to complete the full-scale assessment.
All analyses for the full-scale assessment are described in this document, with the initial
analyses being the focus in Chapters 3 and 4 and the additional analyses in Chapter 5. Further,
given the significant time constraints of this review, risk results are provided in this document
without substantial interpretation. Rather, interpretative discussion of these results is provided in
the Staff Paper.
1 As described in the Preface to this document, the ecological risk analysis performed for this review, which
will be considered in the policy assessment for the secondary standard, is presented in the draft technical report of
the pilot phase risk assessments (ICF, 2006) and described in the Staff Paper (USEPA, 2007a).
2 Throughout the remainder of this document, the term "risk assessment" will be used to refer to both the
human exposure and health risk assessments collectively, unless specific reference to either the human exposure or
health risk assessment is required.
1-1
-------
1.1 MULTIMEDIA ASPECT OF THE RISK ASSESSMENT
The focus for this Pb NAAQS risk assessment is on Pb derived from those sources
emitting Pb to ambient air. In designing and implementing this assessment, we have been faced
with significant limitations and complexity that go far beyond the situation for similar
assessments typically performed for other criteria pollutants. In addition to the constraints of the
timeframe allowed for this review, we are also constrained by significant limitations with regard
to data and tools needed for the assessment. The multimedia and persistent nature of Pb and the
role of multiple exposure pathways contribute significant additional complexity to the
assessment as compared to other assessments that focus only on the inhalation pathway.
First, exposures to Pb emitted into the air occur via multiple pathways. As described in
the Air Quality Criteria for Lead (USEPA, 2006b; henceforth referred to as the CD), "The
multimedia aspects of Pb exposure can be seen in that Pb emissions to the air contribute to Pb
concentrations in water, soil and dusts; Pb in soil and dust also can make important contributions
to Pb concentrations in ambient air" (CD, p. 3-1).
Inhalation exposures can result from Pb emitted to the ambient air recently or from Pb
emitted in the past that has deposited from air to soil or dust and then become resuspended in the
ambient air. Further, Pb emitted into the ambient air can contribute to ingestion exposures
(associated with indoor dust, outdoor soil/dust, agricultural products and surface water) of
recently deposited Pb and of Pb that was deposited in the past. Consequently, this is a
multipathway risk assessment in which we are considering both airborne Pb, as it contributes to
human exposures through direct inhalation of particles containing Pb, and also Pb that has
deposited from air to dusts, soil and other environmental media and that contributes to human
exposures through ingestion. Further, we are considering that Pb, once deposited, may be
resuspended in the air, contributing to human inhalation exposures or, upon redeposition, to
human ingestion exposures. Thus, as illustrated in Figure 1-1, pathways that are directly relevant
to a review of the NAAQS include both newly emitted Pb from currently operating sources, and
Pb emitted in the past, either from currently operating sources or historic sources, which are
collectively referred to as "policy-relevant sources".
1-2
-------
Policy-relevant
Background*
Policy-relevant Sources
Policy-relevant
Background*
Historically emitted Pb
Non-air Pb
releases
Outdoor Soil, Dusts
Ecological Exposures
Human Exposures
-Inhalation & ingestion
*Policy-relevant background sources and pathways are indicated by dashed lines.
"""Dietary exposure should not be considered to be limited to policy-relevant background, as it reflects a combination
of Pb introduced into food items during processing (policy-relevant background), as well as Pb associated with
atmospheric deposition (policy-relevant sources).
Figure 1-1. Principal pathways of human and ecological exposure to Pb. Among the
policy-relevant pathways, heavy arrows indicate the predominant human
exposures.
Due to limited data, models, and time available, however, we are not able to fully and
completely characterize in our risk assessment all of the various complexities associated with Pb.
Consequently, in our efforts to focus on and characterize risk associated with the ambient air-
related3 sources and exposures, we have made a number of simplifying assumptions in a number
of areas. For example, Figure 1-1 illustrates that people are also exposed to Pb that originates
from nonair sources, including leaded paint or drinking water distribution systems. For purposes
of this assessment, the Pb from these nonair sources is collectively referred to as "policy-relevant
background" 4. Although Pb in diet and drinking water sources may derive from Pb emitted into
Ambient air related sources are those emitting Pb into the ambient air (including resuspension of
previously emitted Pb), and ambient air related exposures include inhalation of ambient air Pb as well as ingestion of
Pb deposited out of the air (e.g., onto outdoor soil/dust or indoor dust).
4 This categorization of policy-relevant sources and background exposures is not intended to convey any
particular policy decision at this stage regarding the Pb standard. Rather, it is simply intended to convey an area of
interest to this review.
1-3
-------
the ambient air, the contribution from air pathways to these exposure pathways is not explicitly
recognized, such that these exposures are treated as policy-relevant background.
1.2 RISK ASSESSMENT FROM LAST REVIEW
In the risk assessment conducted in support of the last review, air quality scenarios were
compared in terms of their impact on the percentage of modeled populations that exceeded
specific blood Pb levels chosen with consideration of the health effects evidence at that time
(USEPA, 1986a, 1986b, 1990a). The 1990 analysis focused on both children (birth through 7
years of age) and middle-aged men residing in three case study locations, two near secondary Pb
smelters and one near a primary Pb smelter (USEPA, 1990b). The analysis also introduced the
use of pharmacokinetic blood Pb modeling for children, although it used empirically derived
slope models for adult men to relate changes in air Pb to changes in blood Pb.
In the 1990 Staff Paper, staff concluded that at levels of 10-15 micrograms per deciliter
(ug/dL) of blood Pb, there appeared to be "a convergence of evidence of lead-induced
interference with diverse set of physiological functions and processes, particularly evident in
several independent studies showing impaired neurobehavioral function and development"
(USEPA, 1990).5 Accordingly, the staff used blood Pb levels of 10 and 15 ug/dL to evaluate
effects of alternate NAAQS on children in the 1990 analysis (USEPA, 1990). These values were
chosen with consideration of the full body of health effects evidence at that time. Staff then used
dispersion modeling (the Industrial Source Complex (ISC) model) combined with source
characterization data to generate Pb air concentrations for each case study area. Statistically
derived relationships based on data from other industrial locations, including Pb smelters, that
linked concentrations of Pb in air to Pb in indoor dust and outdoor soil were then used to predict
Pb in these media for the three case study locations, based on the modeled air Pb concentrations.
An uptake/biokinetic model was also developed to predict child blood Pb levels. This model was
used in place of a statistical regression slope model to allow consideration of the dynamic nature
of Pb exposure in children. EPA combined model-derived central tendency blood Pb levels with
an estimated geometric standard deviation (GSD) reflecting interindividual variability in blood
Pb levels, to generate population distributions of blood Pb levels. These distributions were then
used to estimate the percentage of children at each case study location that exceeded the blood
Pb levels 10 and of 15 ug/dL, respectively.
For adult men, the 1990 assessment used blood Pb levels of 10 and 12 ug/dL to compare
relative effects of alternate NAAQS (USEPA, 1990). The same approach was used for
5 As a result of a parallel activity, the U.S. Centers for Disease Control and Prevention in 1991 reduced the
children's blood Pb level warranting individual intervention to 15 ug/dL and identified a level of 10 ug/dL for
implementing community-wide prevention activities (CDC, 1991; CDC, 2005).
1-4
-------
generating media concentrations for the adult analysis as was used for the child assessment.
However, rather than using a biokinetic model, as for the children's assessment, the 1990
analysis for adults used statistically derived slope models to relate air Pb to blood Pb levels with
two versions of the slope models being employed: (a) the aggregate model which predicts blood
Pb in adults based solely on air Pb levels (here a single slope factor captures both the direct
inhalation pathway as well as the more complex pathway of Pb deposition to soil and dust
followed by incidental ingestion) and (b) the disaggregate model which uses media-specific
slopes to predict blood Pb based on Pb concentrations in soil, dust and air. Since the projected
blood Pb levels were mean population levels, a GSD term was included to develop population-
level blood Pb distributions. The GSD estimates for adults and children were derived from
information on observed blood Pb levels in these subgroups. These population-level
distributions were then queried to identify the percentage of adult men at each case study
location with modeled blood Pb levels exceeding the levels of interest for adults (10 and 12
Hg/dL).
The primary difference between the risk assessment approach used in the current pilot
analysis and the assessment completed in 1990 involves the risk metric employed, which reflects
the quantitative and qualitative health effects evidence available today that was not available in
1990 (CD). Rather than estimating the percentage of study populations with exposures above
blood Pb levels of interest as was done in the last review (i.e., 10, 12 and 15 ug/dL), the current
pilot analysis estimates the degree of health decrement in study populations exposed to Pb.
Specifically, the pilot analysis estimates the distribution of IQ loss associated with Pb exposure
for child populations at each of the case study locations with that IQ loss further differentiated
between background Pb exposure and policy-relevant exposures.
1.3 PILOT PHASE ASSESSMENT FOR THE CURRENT REVIEW
The pilot phase of the risk assessment for the current review is described in the first draft
Staff Paper and accompanying technical report (USEPA 2006a; ICF 2006). The pilot assessment
was intended primarily as a demonstration of the risk assessment methodology being developed
for the current review. Consequently, exposure and risk results from the pilot assessment are
considered preliminary.
The pilot assessment included three case studies: (a) a primary Pb smelter (in
Herculaneum, Missouri), (b) a secondary Pb smelter (in Troy, Alabama), and, (c) a near roadway
(urban) location along a short road segment in Houston, Texas.6 The case studies modeled for
6 This case study was intended to provide perspective on the near roadway exposure scenario but was not
intended to estimate total population risk for a full urban or metropolitan area.
1-5
-------
the pilot were selected to provide a preliminary perspective on the nature and magnitude of air
sourced Pb exposures and risk. In addition, they provided a range of exposure scenarios in
which to test the risk assessment methodology developed for the current review. Because of
differences in the exposure scenarios and available data at each of the case study locations, the
approach used for modeling exposure and risk differed among the case studies. Results from the
pilot assessment, as well as comments received from the public and CASAC (see Section 1.4)
have informed decisions on the design for the full-scale assessment, including the types of case
studies included.
1.4 CASAC ADVICE
The staff consulted with the CASAC on the draft analysis plan for the risk assessment
(USEPA, 2006c) in June 2006 (Henderson, 2006), and subsequently developed the pilot
assessment, summarized in Section 1.3, and described in the first draft Staff Paper and
accompanying technical report (USEPA, 2006a; ICF 2006). On February 6-7, 2007, the CASAC
Pb panel met to discuss these documents and CASAC's written comments and recommendations
were provided in March 2007 (Henderson, 2007a).
Consistent with their mandate under the Clean Air Act, CASAC provided comments on
both scientific aspects of the risk assessment and aspects related to the standards themselves
(Henderson, 2007a).7 With regard to the risk assessment, they recommended that the case study
approach implemented for the pilot risk assessment be supplemented with a "population-based"
analysis, and, in discussion at the public meeting, the panel raised the general occurrence of
lower Pb levels in urban areas beyond more point source impacted areas as being an important
focus for the risk assessment. As described below, consideration of comments in this area led to
a significant difference in the design of the full-scale assessment as compared to the pilot
assessment.
CASAC also recommended that uncertainty be characterized with regard to the
relationship between a change in the NAAQS and the distribution of population blood Pb
concentrations, and with regard to the relationship between blood Pb concentrations and the risk
of adverse health effects (Henderson, 2007a). With respect to alternate NAAQS for
consideration by EPA, CASAC recommended consideration of levels less than or about 0.2
|ig/m3 (micrograms per cubic meter) and of a monthly averaging time. Additionally, they
indicated that they consider a population loss of 1-2 points in intelligence quotient (IQ) to be
highly significant from a public health perspective and that the primary Pb standard should be set
7 Consistent with the focus of this document on the human exposure and health risk assessment, CASAC
comments regarding the ecological risk assessment and secondary standard considerations are not discussed here.
1-6
-------
to protect 99.5 percent of the population from exceeding that loss (Henderson, 2007a, p. 6).8
CASAC also recommended conducting future Pb monitoring with samplers for particulate matter
less than ten microns in size (PMio) rather than with samplers for total suspended particulate
matter (TSP) (Henderson, 2007a, p. 7).
In consideration of CAS AC's comments on the pilot phase assessment (Henderson,
2007a), we considered a number of alternate approaches for the full-scale assessment. As a
result, several additions and modifications to the assessment design were implemented for the
initial analyses of the full-scale assessment presented in the July 2007 draft risk assessment
report (USEPA, 2007a) and chapters 1 through 4 of this document. The most significant of these
modifications was the replacement of the near roadway case study with a general urban case
study. This case study was designed to provide estimates of risk in urban areas associated with
broad population level exposures to different ambient air levels of Pb. The general urban case
study was assessed in addition to the two point source case studies, and differs from those case
studies in basing the estimate of air quality on monitoring data (rather than on results from air
quality modeling). The alternate NAAQS levels were selected to overlap with the range of levels
suggested by CASAC. The target population and endpoint for the assessment remains young
children and risk of IQ decrements associated with Pb exposure. To address CASAC comments
on the cutpoint employed in the pilot assessment (Henderson, 2007a), the blood Pb concentration
response function was re-examined, and three alternatives were included in the assessment.
Additionally, in consideration of CASAC recommendations regarding the geometric standard
deviation used in the blood Pb modeling, two values were included in the initial analyses for the
general urban case study.
In their review of the July draft risk assessment report, in addition to reiterating their
previous recommendations on aspects of the standards and concurring with several aspects of the
draft assessment, the CASAC Pb Panel made several recommendations for additional exposure
and health risk analyses (Henderson, 2007b), of which several notable ones are summarized here.
First, they recommended that the general urban case study be augmented by the inclusion of risk
analyses in specific urban areas of the U.S. Further, they recommended use of the hybrid dust Pb
model in preference to the alternate that had additionally been employed in the initial analyses
for the general urban case study. The Panel also recommended limiting presentation of
population risk results to the 5th through 95th percentile of the distribution.9 In response to
CASAC's reference to 99.5 percent of the population in this statement is consistent with their recognition
earlier in the same letter of the 99.5 percent figure being the EPA's risk management choice when they established
the Pb NAAQS in 1978 (Henderson, 2007a, pp. 5 and 6).
9 This recommendation was in recognition by Panel members of uncertainties in quantitative estimates
from the assessment for this point on the distribution (Henderson, 2007b, Appendix D).
1-7
-------
discussion by the CASAC Pb Panel associated with these recommendations at the August public
meeting, staff has developed additional analyses that include three location-specific urban case
studies, and focus on a core modeling approach (with regard to dust Pb model, blood metric and
concentration-response functions). With regard to the blood Pb concentration-response function,
CASAC recommended use of a two-piece or dual linear function that recognizes a change in
slope (to a notably higher value) at blood Pb levels of 7.5 ug/dL and indicated less favor for the
two-piece linear function with hinge at 10.82 ug/dL derived for the initial analyses of the full-
scale assessment (see Section 4.1.1.3) (Henderson, 2007b). Accordingly, a different set of
concentration-response functions was employed in the additional analyses (see Section 5.3.1.1),
and risk estimates for the two-piece linear function with hinge at 10.82 ug/dL are not included in
the range of risk estimates presented for the initial analyses in Chapter 4.
1.5 ORGANIZATION OF THE DOCUMENT
The remainder of this document is organized as follows. Chapter 2 describes the design
of the exposure and risk assessments, covering such topics as the conceptual model used in
designing the analysis (Section 2.1), the case studies included in the assessment (Section 2.2), the
air quality scenarios simulated in the assessment (Section 2.3), and an overview of the analytical
approach (Section 2.4). Chapter 3 describes the methods and results for the exposure
assessment, as well as the performance evaluation. Chapter 4 describes the methods for deriving
risk estimates, the resultant estimates, sensitivity analyses and a characterization of uncertainty.
Chapter 5 presents additional analyses using a core modeling approach completed in response to
the August 2007 CASAC review of initial full-scale analyses.
1-8
-------
REFERENCES
Centers for Disease Control (1991) Preventing lead poisoning in young children: a statement by the Centres for
Disease Control. Atlanta, GA: U.S. Department of health and Human Services, Public Health Service;
October 1. http://wonder.cdc.gov/wonder/prevguid/p0000029/p0000029.asp
Centers for Disease Control and Prevention (2005) Preventing lead poisoning in young children: a statement by the
Centers for Disease Control and Prevention. Atlanta, GA: U.S. Department of Health and Human Services,
Public Health Service. August.
Henderson, R. (2006) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee, to
Administrator Stephen L. Johnson. Re: Clean Air Scientific Advisory Committee (CASAC) Lead Review
Panel's Consultation on EPA's draft Analysis Plan for Human Health and Ecological Risk Assessment for
the Review of the Lead National Ambient Air Quality Standards. July 26, 2006.
Henderson, R. (2007a) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee, to
Administrator Stephen L. Johnson. Re: Clean Air Scientific Advisory Committee's (CASAC) Review of
the 1st Draft Lead Staff Paper and Draft Lead Exposure and Risk Assessments. March 27, 2007.
Henderson, R. (2007b) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee, to
Administrator Stephen L. Johnson. Re: Clean Air Scientific Advisory Committee's (CASAC) Review of
the 2nd Draft Lead Human Exposure and Health Risk Assessments Document. September 27, 2007.
ICF International. (2006) Lead Human Exposure and Health Risk Assessments and Ecological Risk Assessment for
Selected Areas. Pilot Phase. Draft Technical Report. Prepared for the U. S. EPA's Office of Air Quality
Planning and Standards, Research Triangle Park, NC. December.
U.S. Environmental Protection Agency. (1986a) Air Quality Criteria for Lead. Environmental Criteria and
Assessment Office, Office of Research and Development, Research Triangle Park, NC. EPA/600/8-83-028
a-d. June 1986.
U.S. Environmental Protection Agency. (1986b) Lead effects on cardiovascular function, early development, and
stature: an addendum to U.S. EPA Air Quality Criteria for Lead (1986) Environmental Criteria and
Assessment Office, Office of Research and Development, Research Triangle Park, NC. September, 1986.
U.S. Environmental Protection Agency. (1990a) Air Quality Criteria for Lead: Supplement to the 1986 Addendum.
Research Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and
Assessment Office; report no. EPA/600/8-89/049F. Available from: NTIS, Springfield, VA; PB91-138420.
U.S. Environmental Protection Agency. (1990b) 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; report no. EPA-450/2-89/022. Available from: NTIS,
Springfield, VA; PB91-206185. Available on the web:
http://www.epa.gov/ttn/naaqs/standards/pb/data/rnaaqsl asti.pdf
U.S. Environmental Protection Agency. (2006a). Review of the National Ambient Air Quality Standards for Lead:
Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper - First Draft. Office of Air
Quality Planning and Standards, Research Triangle Park, NC. EPA-452/P-06-002. Available online at:
U.S. Environmental Protection Agency. (2006b) Air Quality Criteria for Lead. Washington, DC, EPA/600/R-
5/144aF. Available online at: www.epa.gov/ncea/
U.S. Environmental Protection Agency. (2006c) Analysis Plan for Human Health and Ecological Risk Assessment
for the Review of the Lead National Ambient Air Quality Standards. Office of Air Quality Planning and
1-9
-------
Standards, Research Triangle Park, NC. Available online at:
http://www.epa.gov/ttn/naaqs/standards/pb/sjb cr_pd.html
U.S. Environmental Protection Agency. (2007a). Review of the National Ambient Air Quality Standards for Lead:
Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper. Office of Air Quality
Planning and Standards, Research Triangle Park, NC. EPA-452/R-07-013. Available online at:
http://www.epa.gov/ttn/naaqs/standards/pb/s_pb_cr_sp.html
U.S. Environmental Protection Agency. (2007b). Lead: Human Exposure and Health Risk Assessments for Selected
Case Studies (Draft Report). Volume I. Human Exposure and Health Risk Assessments - Full-scale, and
Volume II. Appendices. Office of Air Quality Planning and Standards, Research Triangle Park, NC. EPA-
452/D-07-001a. andEPA-452/D-07-001b. Available online at:
htto://www.erja.£ov/ttn/naaas/standards/rjb/s t)b cr td.html
1-10
-------
2 DESIGN OF EXPOSURE AND RISK ASSESSMENTS
The risk assessment design relies on the use of case studies. The types of case studies
included, as well as the analytical aspects of the assessment of each, reflect consideration of the
evidence presented in the CD, air quality analyses, and findings of the pilot assessment (Section
1.3), as well as comments received from CASAC (Section 1.4) and the public.
Drawing primarily from the CD, Section 2.1 provides background for the risk
assessment, with regard to key elements of Pb exposure and effects. The assessment scenarios
evaluated in the assessment are described in Section 2.2. Background information on the three
case studies is described in Section 2.3. Section 2.4 describes the analytical approach, with
attention to key analytical steps, and discussion of temporal and spatial aspects of the
assessment, as well as the categorization of policy-relevant exposure pathways, and the
uncertainty characterization.
2.1 BACKGROUND INFORMATION ON LEAD EXPOSURE AND RISK
As recognized in Section 1.1, there are a variety of complexities associated with the
assessment of air-related Pb exposure and risk. In this risk assessment, we have attempted to
focus effort on those aspects that are most important and feasible to address within our scope and
given the constraints of time, pertinent data, models, etc. With regard to some aspects,
simplifying assumptions have been implemented. The following subsections describe elements
of Pb exposure and effects pertinent to evaluating public health risks associated with Pb from
ambient air, and specify those that are explicitly addressed in this quantitative risk assessment.
This is summarized in Figure 2-1, with boxes outlined in bold indicating items included in the
quantitative risk assessment and sources and pathways for which ambient air has played a role
identified in bold text.
2-1
-------
Figure 2-1. Conceptual model for Pb human health risk assessment.
Newly Emitted
Pb
Soil/Oust - Outdoors
from Past Pb
Emissions
RE-ENT
J
Diet
i L
3AINMENT
t
si
D£ (0
111 O
I- 0.
Ambient Air
i
DEPOSITION
4
Air * Indoor
t
Inhataton
REENTRA1NM
-Nff
J >
SofifDust *
Outdoors
Dust -
Indoors
Lead Paint -
indoors
1
Dust - indoors
Note: Boxes outlined in bold are included in the quantitative risk assessment. Sources and pathways for
which ambient air has played a role are in bold text.
2-2
-------
2.1.1 Sources, Pathways and Routes
As described in Section 1.1, for the purposes of this assessment, policy-relevant sources
(in Figure 2-1 in bold type) include both sources of new Pb emissions (e.g., from active
stationary and mobile sources) and re-emission or resuspension of historically deposited Pb (e.g.,
near roadways or associated with now inactive, or now lower emitting stationary sources, as
discussed in Appendix A, Section A. 1.1.3).
There are more than 13,000 individual sources in the U.S. for which we have estimated
Pb emissions to the air (Appendix A, Section A. 1.2). Cumulatively, those sources, in addition to
mobile sources and other sources not individually quantified, emit some 1600 tons per year (tpy)
of Pb in the U.S. (Appendix A). The largest categories (in terms of aggregate national
emissions) include mobile sources (specifically combustion of leaded general aviation fuel),
boilers and process heaters, and metals processes, such as primary and secondary Pb smelting.
Of these, metals processing industries are among the largest emitters of Pb, in terms of emissions
from individual facilities (Appendix A, Section A.I). Another potentially large category of Pb
emissions, for which we do not have quantitative estimates in our national emissions inventory,
is resuspension of recent and historically deposited Pb (Appendix A, Section A.I.1.3; CD,
Section 2.3.3). Studies of emissions in southern California suggest that Pb in resuspended road
dust may represent up to 40%-90% of Pb emissions in some areas (Appendix A, Section
A. 1.1.3). Resuspension is represented to differing degrees in the three case studies included in
the risk assessment (Sections 2.2.1 and 2.2.2)
Lead in outdoor dust and soil may be derived from a range of sources including current
and historical air emissions sources, as well as miscellaneous nonair sources (e.g., land disposal
of wastes and subsequent weathering). Outdoor dust and soil may play a substantial role in
human exposures, particularly for children (CD, Section 3.2). Additionally, Pb in house dust,
which may be derived from Pb in outdoor dust and soil as well as from ambient air Pb (including
previously deposited Pb resuspended into ambient air), is another source of children's exposure
(CD, Sections 3.2 and 4.4). For example, blood Pb levels in children have been shown to be
particularly influenced by exposures to Pb in dust (e.g., Lanphear and Roghmann 1997;
Lanphear et al., 1998). Such findings and "other studies of populations near active sources of air
emissions (e.g., smelters), substantiate the effect of airborne Pb and resuspended soil Pb on
interior dust and blood Pb" (CD, p. 8-22).
In addition to airborne emissions (recent or those in the past), sources of Pb to the
environment or to human exposure included old leaded paint, Pb in drinking water and Pb in the
diet (Figure 1-1). As mentioned in Section 1.1, Pb in diet and drinking water may have air
pathway related (i.e., policy-relevant) contributions as well as contributions from policy-relevant
2-3
-------
background (e.g., Pb solder on water distribution pipes and Pb in materials used in food
processing). Limitations in our data and modeling tools have handicapped our ability to separate
these contributions in the risk assessment, such that we have labeled diet and drinking water
policy-relevant background. Consequently, these sources of Pb exposure are depicted in Figure
2-1 as policy-relevant background (in nonbold type), although this is not intended to convey any
particular policy decision at this stage regarding the Pb standard. Policy-relevant pathways (bold
text in Figure 2-1) include inhalation of newly or previously emitted Pb, ingestion of outdoor
soil/dust containing previously deposited Pb, and ingestion of indoor dust containing newly or
previously emitted Pb.
Human exposure to environmental Pb occurs predominantly via ingestion and inhalation
routes, with ingestion (including incidental ingestion of dust and soil) recognized as generally
playing a larger role for the general human population (CD, Section 4.5). The dermal route is
relatively less well characterized but is not considered to play a large role in total Pb exposure
(CD, Section 4.5), and is not included in this assessment (Figure 2-1).
2.1.2 At-risk Populations
In considering populations for inclusion in the risk assessment, we considered evidence
regarding those with increased susceptibility (i.e., physiological factors contributing to a greater
response for the same exposure), and those with increased exposure (including that resulting
from behavior leading to increased contact with contaminated media). A behavioral factor of
great impact on Pb exposure is the incidence of hand-to-mouth activity that is prevalent in very
young children (CD, Section 4.4.3). Physiological factors include both conditions contributing
to a subgroup's increased risk of effects at a given blood Pb level, and those that contribute to
blood Pb levels higher than those otherwise associated with a given Pb exposure (CD, Section
8.5.3). An additional population characterization for which evidence was considered was
vulnerability to pollution-related effects which additionally encompasses situations of elevated
exposure, such as residing in old housing with Pb-containing paint or near sources of ambient
Pb, as well as socioeconomic factors, such as reduced access to health care or low
socioeconomic status (SES) (USEPA, 2003, 2005) that can contribute to increased risk of
adverse health effects from Pb.
Three particular physiological factors contributing to increased risk of Pb effects at a
given blood Pb level are recognized in the CD (e.g., CD, Section 8.5.3): age, health status, and
genetic composition (or genotype). With regard to age, the susceptibility of young children to
the neurodevelopmental effects of Pb is well recognized (e.g., CD, Sections 5.3, 6.2, 8.4, 8.5,
8.6.2), although the specific ages of vulnerability have not been established (CD, pp 6-60 to 6-
64). Early childhood may also be a time of increased susceptibility for Pb immunotoxicity (CD,
2-4
-------
Sections 5.9.10, 6.8.3 and 8.4.6), and childhood exposures have been associated with increased
risk of cardiovascular and neurodegenerative effects in adulthood (CD, p. 8-74). Health status is
another physiological factor in that subpopulations with pre-existing health conditions may be
more susceptible (as compared to the general population) for particular Pb-associated effects,
with this being most clear for renal and cardiovascular outcomes. For example, African
Americans as a group, have a higher frequency of hypertension than the general population or
other ethnic groups (NCHS, 2005), and as a result may face a greater risk of adverse health
impact from Pb-associated cardiovascular effects. A third physiological factor relates to genetic
polymorphisms. That is, subpopulations defined by particular genetic polymorphisms (e.g.,
presence of the 5-aminolevulinic acid dehydratase-2 [ALAD-2] allele) have also been recognized
as sensitive to Pb toxicity, which may be due to increased susceptibility to the same internal dose
and/or to increased internal dose associated with same exposure (CD, p. 8-71, Sections 6.3.5,
6.4.7.3 and 6.3.6).
Several physiological factors pertain to susceptibility by contributing to increased blood
Pb levels (i.e., increased internal dose levels) over those otherwise associated with a given Pb
exposure (CD, Section 8.5.3). These include nutritional status, which plays a role in Pb
absorption from the GI tract (CD, Section 5.10.2.5); polymorphism for the vitamin D receptor,
which studies suggest may contribute to increased Pb absorption from the GI tract (CD, Section
8.4.2.7); presence of the ALAD-2 allele, which studies suggest contribute to increased blood Pb
levels (Section 8.5.3); and bone demineralization, such as occurs during pregnancy, lactation,
and aging, which appears to influence Pb release from bone into the blood (CD, Section 4.3.2).
In summary, there are a variety of ways in which Pb exposed populations might be
characterized and stratified for the purposes of health risk assessment. Age or lifestage is used to
distinguish potential groups on which to focus the risk assessment (see Figure 2-1) in recognition
of its influence on exposure and susceptibility. In consideration of the health effects evidence
regarding endpoints of greatest public health concern and a recognition of effects on the
developing nervous system as a sentinel endpoint for public health impacts of Pb (see Section
2.1.4), young children have been selected as the priority population for this risk assessment (see
Figure 2-1). We recognize, however, other population subgroups as described above may also
be at risk of Pb-related health effects of public health concern at similar or higher exposures. As
currently available data do not generally support quantitative modeling that differentiates blood
Pb levels and associated health risk within a particular population group such as young children
on the basis of enhanced or reduced susceptibility to Pb effects (e.g., concentration response
functions for IQ loss that differentiate between populations that are calcium deficient and those
that are not), the assessment does not develop separate risk estimates for such subpopulations of
2-5
-------
young children. In the risk assessment, however, interindividual variability in blood Pb is
quantitatively considered (Section 3.2.3).
2.1.3 Internal Disposition
Once inhaled or ingested and absorbed into the blood stream, Pb is distributed throughout
the body via the blood, with bone being the predominant site of Pb accumulation and storage in
the body (CD, Sections 4.2 and 4.3). Additionally, the epidemiologic evidence indicates that Pb
freely crosses the placenta resulting in continued fetal exposure throughout pregnancy, and that
exposure increases during the later half of pregnancy (CD, Section 6.6.2).
During childhood development, bone represents approximately 70% of a child's body
burden of Pb, and this accumulation continues through adulthood, when more than 90% of the
total Pb body burden is stored in the bone (CD, Section 4.2.2). Accordingly, levels of Pb in bone
are indicative of a person's long-term, cumulative exposure to Pb. In contrast, blood Pb levels
are usually indicative of recent exposures. Depending on exposure dynamics, however, blood Pb
may - through its interaction with bone - be indicative of past exposure or of cumulative body
burden (CD, Section 4.3.1.5).
Throughout life, Pb in the body is exchanged between blood and bone, and between
blood and soft tissues (CD, Section 4.3.2), with variation in these exchanges reflecting "duration
and intensity of the exposure, age and various physiological variables" (CD, p. 4-1). For
example, resorption of bone (e.g., in pregnant or nursing women, and associated with
osteoporosis in postmenopausal women or, to a lesser magnitude, in older men) results in a
mobilization of Pb from bone into circulation (CD, Sections 4.3.2.4 and 4.3.2.5). Past exposures
that contribute Pb to the bone, consequently, may influence current levels of Pb in blood. Where
past exposures were elevated in comparison to recent exposures, this influence may complicate
interpretations with regard to recent exposure (CD, Sections 4.3.1.4 to 4.3.1.6). That is, higher
blood Pb concentrations may be indicative of higher cumulative exposures or of a recent
elevation in exposure (CD, pp. 4-34 and 4-133).
Bone measurements, as a result of the generally slower Pb turnover in bone, are
recognized as providing a better measure of cumulative Pb exposure (CD, Section 8.3.2). The
bone pool of Pb in children, however, is thought to be much more labile than that in adults due to
the more rapid turnover of bone mineral as a result of growth (CD, p. 4-27). As a result, changes
in blood Pb concentration in children more closely parallel changes in total body burden (CD,
pp. 4-20 and 4-27). This is in contrast to adults, whose bone has accumulated decades of Pb
exposures (with past exposures often greater than current ones), and for whom the bone may be a
significant source long after exposure has ended (CD, Section 4.3.2.5).
2-6
-------
Given the association with exposure, particularly recent exposure, and the relative ease of
collection, blood Pb levels are extensively used as an index or biomarker of exposure by national
and international health agencies (CD, Section 4.3.1.5). Although recent methods are making
bone Pb measurements easier to collect (CD, Section 4.3.2.2) and consequently, their use more
widespread, epidemiological and toxicological studies of Pb health effects and dose-response
relationships tend to be dominated by blood Pb as the exposure metric (CD, Sections 4.3.1.3,
8.3.2 and Chapter 5).
Accordingly, blood Pb level is the index of exposure or exposure metric in this risk
assessment. The use of concentration-response functions that rely on blood Pb (e.g., rather than
ambient Pb concentration) as the exposure metric reduces uncertainty in the causality aspects of
Pb risk estimates. The relationship between specific sources and pathways of exposure and
blood Pb level is needed, however, in order to identify the specific risk contributions associated
with those sources and pathways of greatest interest to this assessment (i.e., those related to Pb
emitted into the air). For example, the blood Pb-response relationships developed in
epidemiological studies of Pb-exposed populations do not distinguish among different sources or
pathways of Pb exposure (e.g., inhalation, ingestion of indoor dust, ingestion of dust containing
leaded paint). In the exposure assessment for this review, models that estimate blood Pb levels
associated with Pb exposure (e.g., CD, Section 4.4) are used to inform estimates of contributions
to blood Pb arising from ambient air related Pb versus contributions from other sources.
2.1.4 Health Endpoints
Lead has been demonstrated to exert "a broad array of deleterious effects on multiple
organ systems via widely diverse mechanisms of action" (CD, p. 8-24 and Section 8.4.1). This
array of health effects, the evidence for which is comprehensively described in the CD, includes
• Heme biosynthesis and related functions;
• Neurological development and function;
• Reproduction and physical development;
• Kidney function;
• Cardiovascular function; and,
• Immune function.
There is also some evidence of Pb carcinogenicity, primarily from animal studies, with limited
human evidence of suggestive associations (CD, Sections 5.6.2, 6.7, and 8.4.10).l
1 Lead has been classified as a probable human carcinogen by the International Agency for Research on
Cancer, based mainly on sufficient animal evidence, and as reasonably anticipated to be a human carcinogen by the
2-7
-------
This review is focused on those effects most pertinent to ambient exposures. Given the
reductions in ambient Pb levels over the past 30 years, these effects are generally those
associated with the lowest Pb levels of exposure. These are neurological, hematological and
immune effects for children, and neurological, hematological, cardiovascular and renal effects
for adults (CD, Tables 8-5 and 8-6), with neurological effects in children and cardiovascular
effects in adults appearing to be of greatest public health concern (CD, p. 8-60). The
toxicological and epidemiological information available since the time of the last review
"includes assessment of new evidence substantiating risks of deleterious effects on certain health
endpoints being induced by distinctly lower than previously demonstrated Pb exposures indexed
by blood Pb levels extending well below 10 ug/dL in children and/or adults" (CD, p. 8-25). The
CD indicates some health effects associated with blood Pb levels that extend below 5 |ig/dL,
with some studies observing these effects at the lowest blood levels considered (i.e., threshold
levels for these effects cannot be discerned from the currently available studies).
The endpoints identified above are important considerations for this review and are
described briefly in sections below, with detailed discussion of the evidence presented in the CD.
Of these health endpoints, the focus of the quantitative health risk assessment is developmental
neurotoxicity in children, with IQ decrement as the risk metric (Figure 2-1), described in Section
2.1.5
2.1.4.1 Developing Nervous System
The nervous system has long been recognized as a target of Pb toxicity, with the
developing nervous system affected at lower exposures than the mature system (CD, Sections
5.3, 6.2.1, 6.2.2, and 8.4). While blood Pb levels in U.S. children ages one to five years have
decreased notable since the late 1970s, newer studies have investigated and reported associations
of effects on the neurodevelopment of children with these more recent blood Pb levels (CD,
Chapter 6). Functional manifestations of Pb neurotoxicity during childhood include sensory,
motor, cognitive and behavioral impacts. Numerous epidemiological studies have reported
neurocognitive, neurobehavioral, sensory, and motor function effects in children at blood Pb
levels below 10 ug/dL (CD, Section 6.2). Further, "extensive experimental laboratory animal
evidence has been generated that (a) substantiates well the plausibility of the epidemiologic
findings observed in human children and adults and (b) expands our understanding of likely
mechanisms underlying the neurotoxic effects" (CD, p. 8-25; Section 5.3).
U.S. National Toxicology Program (CD, Section 6.7.2). U.S. EPA classified it in the past as a probable carcinogen
(http://www.epa.gov/iris/subst/0277.htmX
2-8
-------
Cognitive effects associated with Pb exposures that have been observed in
epidemiological studies have included decrements in intelligence test results, such as the widely
used IQ score, and in academic achievement as assessed by various standardized tests as well as
by class ranking and graduation rates (CD, Section 6.2.16 and pp 8-29 to 8-30). As noted in the
CD with regard to the latter, "Associations between Pb exposure and academic achievement
observed in the above-noted studies were significant even after adjusting for IQ, suggesting that
Pb-sensitive neuropsychological processing and learning factors not reflected by global
intelligence indices might contribute to reduced performance on academic tasks" (CD, pp 8-29 to
8-30).
Other cognitive effects observed in studies of children have included effects on attention,
executive functions, language, memory, learning and visuospatial processing (CD, Sections
5.3.5, 6.2.5 and 8.4.2.1), with attention and executive function effects associated with Pb
exposures indexed by blood Pb levels below 10 ug/dL (CD, Section 6.2.5 and pp. 8-30 to 8-31).
The evidence for the role of Pb in this suite of effects includes experimental animal findings
(discussed in CD, Section 8.4.2.1; p. 8-31), which provide strong biological plausibility of Pb
effects on learning ability, memory and attention (CD, Section 5.3.5), as well as associated
mechanistic findings. Further, Pb-induced deficits observed in animal and epidemiological
studies, for the most part, have been found to be persistent in the absence of markedly reduced
environmental exposures (CD, Sections 5.3.5, 6.2.11, and 8.5.2). It is additionally important to
note that there may be long-term consequences of such deficits over a lifetime. Poor academic
skills and achievement can have "enduring and important effects on objective parameters of
success" later in life, as well as increased risk of antisocial and delinquent behavior (CD,
Section 6.2.16).
Other neurological effects associated with Pb exposures indexed by blood Pb levels near
or below 10 |ig/dL include behavioral effects, such as delinquent behavior (CD, Sections 6.2.6
and 8.4.2.2), sensory effects, such as those related to hearing and vision (CD, Sections 6.2.7,
7.4.2.3 and 8.4.2.3), and deficits in neuromotor function (CD, p. 8-36).
Neurocognitive impact, specifically decrement in IQ in young children, is a focus of the
quantitative risk assessment due to the strength of evidence for association with blood Pb levels
below 10 ug/dL, and the strength of the dose-response information at these exposure levels.
As discussed in the CD (Section 8.4.2) and by Rice (1996), while there is no direct
animal test parallel to human IQ tests, "in animals a wide variety of tests that assess attention,
learning, and memory suggest that Pb exposure (of animals} results in a global deficit in
functioning, just as it is indicated by decrements in IQ scores in children" (CD, p. 8-27). The
animal and epidemiological evidence for this endpoint are consistent and complementary (CD, p.
8-44). As stated in the CD (p. 8-44):
2-9
-------
Findings from numerous experimental studies of rats and of nonhuman primates, as
discussed in Chapter 5, parallel the observed human neurocognitive deficits and the
processes responsible for them. Learning and other higher order cognitive processes
show the greatest similarities in Pb-induced deficits between humans and experimental
animals. Deficits in cognition are due to the combined and overlapping effects of 'Pb-
inducedperseveration, inability to inhibit responding, inability to adapt to changing
behavioral requirements, aversion to delays, and distractibility. Higher level
neurocognitive functions are affected in both animals and humans at very low exposure
levels (<1Q /ug/dL), more so than simple cognitive functions.
Further, "epidemiologic studies of Pb and child development have demonstrated inverse
associations between blood Pb concentrations and children's IQ and other outcomes at
successively lower Pb exposure levels" over the past 30 years (CD, p. 6-64). This is supported
by multiple studies performed over the past 15 years (see CD, Section 6.2.13), with particularly
compelling evidence for decrements in IQ at blood Pb levels below 10 ug/dL provided by a
recent international pooled analysis of seven prospective studies (Lanphear et al., 2005; CD,
Section 6.2.13). For example, this pooled analysis estimated a decline of 6.2 points (with a 95%
confidence interval bounded by 3.8 and 8.6) in full scale IQ occurring between approximately 1
and 10 ug/dL blood Pb level, measured concurrent with the IQ test (CD, p. 6-76). This analysis
(Lanphear et al., 2005) is relied upon in the quantitative risk assessment for this endpoint.
2.1.4.2 Adult Nervous System
The nervous system has long been recognized as a target of Pb toxicity (CD Sections
5.3.1, 8.4.2). For example, those chronically exposed in the workplace are at risk for various
neurological effects including peripheral sensory nerve impairment, visuomotor and memory
impairment, and postural sway abnormalities, with a blood Pb concentration >14 ug/dL being a
possible threshold (CD, p. 6-87). Past occupational exposure also increases the risk of
developing amyotrophic lateral sclerosis and motor neuron disease (CD, Section 6.3.5 and p. 6-
87). Essential tremor is also associated with Pb exposures, particularly for those with genetic
susceptibility (CD, Sections 6.3.5 and 6.3.6 and p. 6-86).
In elderly populations, significant associations have been reported between bone Pb
levels and impaired cognitive performance or dysfunction (CD, Section 6.3.3 and 6.3.3.1), but
not with blood Pb levels, perhaps indicating a role of cumulative and/or past Pb exposures (CD,
p. 6-83). During demineralization of bone in the elderly, Pb may be released into the blood, thus
augmenting blood Pb associated with current ambient exposures (CD, Section 4.3.2.4). An
increased susceptibility among the elderly to Pb effects on cognitive function is supported by
animal evidence (Section 5.3.7). With lifetime exposure, senescent animals have exhibited an
2-10
-------
increased susceptibility to Pb, due to the increased exposure from bone resorption, and an
apparently greater sensitivity to the biochemical effects of Pb (CD, Section 5.3.7). Laboratory
animal research in rats and monkeys also indicates a potential for cognitive function effects in
the elderly to be related to physiological effects (regulation of protein thought to play a role in
Alzheimer's disease) of Pb exposures in early childhood (CD, p. 5-67; Basha et al., 2005). Thus,
early life exposure to Pb may contribute to neurocognitive effects later in life due to the
redistribution of Pb body burden from bone to brain and by enhanced susceptibility caused by
age-related degenerative changes in various organs, including brain (CD, p. 8-40).
2.1.4.3 Cardiovascular System
Epidemiologic and experimental toxicology studies provide strong support for the
relationship between Pb exposure and increased adverse cardiovascular outcomes, including
increased blood pressure, increased incidence of hypertension, and cardiovascular morbidity and
mortality (CD, Sections 5.5, 6.5 and 8.4.3). The cardiovascular effect most frequently examined
in epidemiological studies is increased systolic blood pressure in adults, which has been
repeatedly associated with Pb exposure (CD, Sections 8.4.3, 8.6.3, 6.5.2.3, and 6.5.7). The
association has been observed with Pb levels in bone and also, in some cohorts, with Pb in blood
(including blood Pb levels below 10 ug/dL). A recent meta-analysis by Nawrot and others
(2005), that included a range of blood Pb levels from 2.3 to 63.8 ug/dL, reported an association
of increased systolic blood pressure and decreased diastolic pressure with increased blood Pb
level, including levels below 10 ug/dL. The magnitude of change observed has considerable
significance at the population level (CD, p. 8-45, Section 8.6.3). The epidemiological evidence
is supported by evidence in numerous animal studies of arterial hypertension with low Pb
exposures, an effect that persists in animals long after cessation of exposure (CD, Sections 5.5
and 8.4.3).
Multiple studies reporting positive associations of blood pressure and hypertension with
bone Pb levels highlight the important role of cumulative past Pb exposure in development of
cardiovascular health effects (Sections 6.5.2.3 and 6.5.7). A study of young adults who lived as
children in an area of high Pb exposures also indicates the potential role of childhood exposure.
In this study, higher bone Pb levels were associated with higher systolic and diastolic blood
pressure (CD, p. 6-138), while current blood Pb levels (mean of 2.2 ug/dL) were not (CD, p. 6-
124).
Systolic blood pressure exerts a strong influence on more serious cardiovascular events
by its role in hypertension and its adverse cardiovascular sequelae (CD, p. 8-83). Several
analyses of National Health and Nutrition Examination Survey (NHANES) cohorts, including
some recently released, have collectively suggested a "significant effect of Pb on cardiovascular
2-11
-------
mortality in the general U.S. population" (CD, p. 8-88, Sections 6.5.3.2 and 8.6.3). For example
recent analyses of NHANES blood Pb data from 1976 to 1980 and 1988 to 1994 provide
supportive evidence for an increased risk of cardiovascular mortality, consistent with projected
likely increases in serious cardiovascular events (stroke, heart attack) resulting from Pb-induced
increases in blood pressure (CD, Section 8.6.3).
2.1.4.4 Renal System
Lead nephrotoxicity is mediated by alterations in the glomerular filtration rate (CD,
Sections 5.7.3 and 8.4.5). The interaction of Pb with the kidney, including occurrences and
mechanisms of Pb uptake by and accumulation in the kidney, and associated cellular alterations,
is well described in animal research (CD, Section 5.7). A set of screening tests involving
markers of nephrotoxic effects have been established for screening individuals exposed to Pb
occupationally or environmentally (CD, Section 5.7.1). In the epidemiological literature,
associations between blood Pb and indicators of renal function impairment (e.g., measures of
glomerular integrity, such as creatinine levels in urine) have been found at blood Pb levels
extending below 10 ug/dL, to as low as ~2 to 4 ug/dL (CD, Sections 6.4.4.1.5 and 8.4.5).
Associations are also observed with cumulative Pb dose, assessed via bone Pb, and longitudinal
renal function decline (CD, p. 6-94), indicating the potential role of earlier exposures.
The findings for non-occupational populations since the last review provide "strong
evidence that renal effects occur at much lower blood Pb levels than previously recognized"
(CD, p. 6-113). These findings of lower Pb renal effects thresholds in environmental compared
to occupational research may be a result of potentially larger proportions of susceptible
individuals in the general population as compared to occupational cohorts (CD, p. 6-107). The
data available are not sufficient to determine whether these effects are related more to current
blood-Pb levels, higher levels from past exposure, or both (CD, p. 8-49).
The findings regarding Pb exposures and renal effects are of particular concern with
regard to certain susceptible subpopulations. At levels of exposure in the general U.S.
population overall, Pb combined with other risk factors, such as diabetes, hypertension, or
chronic renal insufficiency from causes unrelated to Pb, can result in clinically relevant effects.
Notably, the size of such susceptible populations is increasing in the United States due to obesity
(CD, p. 6-113). That is, Pb is recognized as acting cumulatively with other renal risk factors to
cause early onset of renal insufficiency and/or a steeper rate of renal function decline in
individuals already at risk for renal disease (CD, p. 6-107).
2.1.4.5 Heme Synthesis
It has long been recognized that Pb exposure is associated with disruption of heme
synthesis in both children and adults. The evidence regarding effects on heme synthesis and
2-12
-------
other hematological parameters in animal and humans is strong, and includes documented
quantitative relationships between exposure and effects in children and adults. Interference with
heme synthesis was identified as one of the targets of low-level Pb toxicity in children during the
time of the last NAAQS review (USEPA, 1990), and was the primary focus for the initial setting
of the Pb NAAQS in 1978 (USEPA, 1978).
Mechanisms associated with Pb interference with heme synthesis include inhibition of
the enzymes ALAD and ferrochelatase (Table 3-1; CD Sections 5.2.1, 6.9.1, 6.9.2; USEPA
1986). Inhibition of ALAD has been associated with increased blood Pb concentrations at and
somewhat below 10 ug/dL, in children and adults (Tables 3-1 and 3-2; CD, Table 6-7). Blood
Pb concentrations at and above approximately 15 ug/dL, in children, and 15-30 ug/dL, in adults,
are associated with elevation of erythrocyte protoporphyrin (EP), and notable reductions in
hemoglobin synthesis (Tables 3-1 and 3-1; CD, p. 8-47; USEPA, 1986). In the setting of the Pb
NAAQS in 1978, the Agency concluded that "the state of elevated EP must be regarded as
potentially adverse to the health of young children" (USEPA, 1978). Blood Pb concentrations at
and above 40 ug/dL are associated with frank anemia, a clinical sign of severe Pb poisoning (CD,
p. 8-47). The evidence regarding Pb disruption of heme synthesis and associated mechanisms is
presented in detail in past CDs (USEPA 1986, 1977), with more recent findings, including the
role of genetic polymorphisms, discussed in the current CD (Sections 8.4.4, 5.2.1, 6.9.1 and
6.9.2).
2.1.4.6 Immune System
Since the time of the last review, there has been substantial research on the
immunotoxicity of Pb. As summarized in the CD, "studies across humans and a variety of
animal models are in general agreement concerning both the nature of the immunotoxicity
induced by Pb as well as the exposure conditions that are required to produce
immunomodulation" (CD, p. 5-244, Section 5.9). Lead is distinguished from other
immunotoxicants, however, by the fact that the most sensitive biomarkers of its immunotoxicity
are associated with specific functional capacities that influence risk of disease, as opposed to
being associated with changes in immune cell numbers or pathological changes of lymphatic
system organs (CD, Section 5.9.1). The main immune system targets of Pb are macrophages
and T lymphocytes, leading to a potential for increased tissue inflammation, reduced cell-
mediated immunity, and increased risk of autoimmunity (See CD, Figure 5-18, Section 5.9.11).
Additionally, Pb exposures in both animal and human studies are associated with increased
production of IgE, an immunoglobulin involved in allergic responses and asthma (CD, Section
5.9.3.2). These effects have been reported in epidemiologic studies of children, and supported
2-13
-------
by evidence in neonatal and juvenile animals, at blood Pb levels extending below 10 ug/dL (CD,
p. 6-197 and Sections 5.9.10 and 8.4.6).
2.1.5 Metric and Model for Risk Quantitation
The health endpoint on which we focused in the quantitative health risk assessment for
this review is developmental neurotoxicity in children (Section 2.1.4.1), with IQ decrement as
the risk metric. Among the wide variety of health endpoints associated with Pb exposures, there
is general consensus that the developing nervous system in young children is the most sensitive
and that neurobehavioral effects (specifically neurocognitive deficits), including IQ decrements,
appear to occur at lower blood levels than previously believed (i.e., at levels <10 ug/dL). For
example, the overall weight of the available evidence, described in the CD, provides clear
substantiation of neurocognitive decrements being associated in young children with blood Pb
levels in the range of 5 to 10 ug/dL, and some analyses indicate Pb effects on intellectual
attainment of young children ranging from 2 to 8 ug/dL (CD, Sections 6.2, 8.4.2 and 8.4.2.6).
That is, while blood Pb levels in U.S. children ages one to five years have decreased notably
since the late 1970s, newer studies have investigated and reported associations of effects on the
neurodevelopment of children with these more recent blood Pb levels (CD, Chapter 6).
The evidence for neurotoxic effects in children is a robust combination of
epidemiological and toxicological evidence (CD, Sections 5.3, 6.2 and 8.5). The
epidemiological evidence is strongly supported by animal studies that substantiate the biological
plausibility of the associations, and provides an understanding of mechanisms of action for the
effects (CD, Section 8.4.2). The selection of children's IQ for the quantitative risk assessment
reflects consideration of the evidence presented in the CD as well as advice received from
CASAC (Henderson, 2006, 2007a).
The epidemiological studies that have investigated blood Pb effects on IQ (see CD,
Section 6.2.3) have considered a variety of specific blood Pb metrics, including: 1) blood
concentration "concurrent" with the response assessment (e.g., at the time of IQ testing), 2)
average blood concentration over the "lifetime" of the child at the time of response assessment
(e.g., average of measurements taken over child's first 6 or 7 years), 3) peak blood concentration
during a particular age range and 4) early childhood blood concentration (e.g., the mean of
measurements between 6 and 24 months age). All four specific blood Pb metrics have been
correlated with IQ (see CD, p. 6-62; Lanphear et al., 2005). In the international pooled analysis
by Lanphear and others (2005), however, the concurrent and lifetime averaged measurements
were considered "stronger predictors of lead-associated intellectual deficits than was maximal
measured (peak) or early childhood blood lead concentrations," with the concurrent blood Pb
level exhibiting the strongest relationship (CD, p. 6-29). It is not clear in this case or for similar
2-14
-------
findings in other studies, whether the cognitive deficits observed were due to Pb exposure that
occurred during early childhood or were a function of concurrent exposure. Nevertheless,
concurrent blood Pb levels likely reflected both ongoing exposure and preexisting body burden
(CD, p. 6-32).
Using concurrent blood Pb level as the exposure metric and IQ as the response from the
pooled dataset of seven international studies, Lanphear and others (2005) employed
mathematical models of various forms, including linear, cubic spline, log-linear, and piece-wise
linear, in their investigation of the blood Pb concentration-response relationship (CD, p. 6-29;
Lanphear et al., 2005). They observed that the shape of the concentration-response relationship
is nonlinear and the log-linear model provides a better fit over the full range of blood Pb
measurements than a linear one (CD, p. 6-29 and pp. 6-67 to 6-70; Lanphear et al., 2005). In
addition, they found that no individual study among the seven was responsible for the estimated
nonlinear relationship between Pb and deficits in IQ (CD p. 6-30). Others have also analyzed the
same dataset and similarly concluded that, across the range of the dataset's blood Pb levels, a
log-linear relationship was a significantly better fit than the linear relationship (p=0.009) with
little evidence of residual confounding from included model variables (CD, Section 6.2.13;
Rothenberg and Rothenberg, 2005).
A nonlinear blood Pb concentration-response relationship is also suggested by several
other studies that have observed that each ug/dL increase in blood Pb may have a greater effect
on IQ at blood Pb levels below 10 ug/dL than at higher levels (CD, pp. 8-63 to 8-64). While this
may at first seem at odds with certain fundamental toxicological concepts, a number of examples
of non- or supralinear dose-response relationships exist in toxicology (CD, pp. 6-76 and 8-83 to
8-89).2 With regard to the effects of Pb on neurodevelopmental outcome such as IQ, the CD
suggests that initial neurodevelopmental effects at lower Pb levels may be disrupting very
different biological mechanisms (e.g., early developmental processes in the central nervous
system) than more severe effects of high exposures that result in symptomatic Pb poisoning and
frank mental retardation (CD, p. 6-76). In comparing across the individual studies and the
pooled analysis, it is observed that at higher blood Pb levels, the slopes derived for log-linear and
linear models are almost identical, and for studies with lower blood Pb levels, the slopes appear
to be steeper than those observed in studies involving higher blood Pb levels (CD, p. 8-78,
Figure 8-7).
Given the evidence summarized here and described in detail in the CD (Chapters 6 and
8), and in consideration of CASAC recommendations (Henderson, 2006, 2007a, 2007b), the risk
2 Similarly, a nonlinear concentration-response relationship was observed for the relationship between
blood Pb levels and blood pressure in adults (CD, pp. 8-83 to 8-89).
2-15
-------
assessment for this review relies on the functions presented by Lanphear and others (2005) that
relate absolute IQ as a function of concurrent blood Pb or of the log of concurrent blood Pb, and
lifetime average blood Pb, respectively. As discussed above, the slope of the concentration-
response relationship described by these functions is greater at the lower blood Pb levels (e.g.,
less than 10 ug/dL). The impact of the nonlinear slope is illustrated by the estimates of IQ
decrements associated with increases in blood IQ for different ranges of blood Pb level reported
for the log-linear model (Lanphear et al., 2005). These estimates were IQ decrements of 3.9
(with 95% confidence interval, CI, of 2.4-5.3), 1.9 (95% CI, 1.2-2.6) and 1.1 (95% CI, 0.7-1.5),
for increases in concurrent blood Pb from 2.4 to 10 ug/dL, 10 to 20 ug/dL, and 20 to 30 ug/dL,
respectively (Lanphear et al., 2005). For an increase in concurrent blood Pb levels from <1 to 10
ug/dL, the log-linear model estimates a decline of 6.2 points in full scale IQ which is comparable
to the 7.4 point decrement in IQ for an increase in lifetime mean blood Pb levels up to 10 ug/dL
observed in the Rochester study (CD, pp 6-30 to 6-31).
Several studies have examined the relationship of IQ decrement with blood Pb, quantified
by a variety of metrics, at lower blood Pb levels. On a change in IQ per ug/dL basis, estimates of
IQ decrement associated with blood Pb levels (concurrent, 24-month, peak, lifetime average or
lifetime cumulative) below 10 ug/dL range from -0.4 to -1.8 (CD, Table 8.7). At the upper end
of this range are the slopes derived for the subsets of children in the Rochester and Boston
cohorts for which peak blood Pb levels were <10 ug/dL; these slopes are -1.8 (for concurrent
blood Pb influence on IQ) and -1.6 (for 24-month blood Pb influence on IQ), respectively. The
number of children in these low blood Pb subsets of the Rochester and Boston cohorts are 101
and 48, respectively. A similar stratification of the pooled dataset by Lanphear and others (2005)
yielded a slope for the linear function of IQ change associated with concurrent blood Pb of-0.8
for the subset of the children in the pooled data set for which maximal or peak blood Pb levels
were below 10 ug/dL. Of the 1333 children in the full pooled dataset, there were 244 in this
subset. When the full dataset was restricted to a still smaller subset of 103 children for which
peak blood Pb levels were below 7.5 ug/dL the slope of concurrent blood Pb and IQ was -2.94
(Lanphear et al., 2005). The analysis of this latter subset supported the authors' conclusions that
"for a given increase in blood lead, the lead-associated intellectual decrement for children with a
maximal blood lead level <7.5 ug/dL was significantly greater than that observed for those with
a maximal blood lead level > 7.5 ug/dL (p-0.015)" and that "environmental lead exposure in
children who have maximal blood lead levels < 7.5 ug/dL is associated with intellectual
deficits". This subset was composed primarily of children from the Rochester cohort (69
children), with smaller numbers of children from five of the other seven cohorts (Lanphear et al.,
2005). The Rochester data included IQ test and concurrent blood Pb measurements taken at age
six (Lanphear et al., 2005). The linear slope observed for this subset of the pooled dataset,
2-16
-------
however, was notably greater than that previously reported for the low blood Pb subset of the
Rochester cohort at age five described above, and greater than those slopes from other studies for
blood Pb < 10 ug/dL summarized in the CD (e.g., CD, Table 8-7), providing some uncertainty
with regard to the precise magnitude of slope for the full range of blood Pb below 7.5 ug/dL.
As discussed in the CD, threshold blood Pb levels for these effects cannot be discerned
from the currently available epidemiological studies, and the evidence in the animal Pb
neurotoxicity literature does not define a threshold for any of the toxic mechanisms of Pb (CD,
Sections 5.3.7 and 6.2).
In applying relationships observed with the pooled analysis (Lanphear et al., 2005) to the
risk assessment, which includes blood Pb levels below the range represented by the pooled
analysis, several alternative blood Pb concentration-response models were considered in
recognition of a reduced confidence in our ability to characterize the quantitative blood Pb
concentration-response relationship at the lowest blood Pb levels represented in the recent
epidemiological studies. The functions considered and employed in the initial risk analyses for
this review include the following.
• Log-linear function with low-exposure linearization, for both concurrent and lifetime
average blood metrics, applies the nonlinear relationship down to the blood Pb
concentration representing the lower bound of blood Pb levels for that blood metric in
the pooled analysis and applies the slope of the tangent at that point to blood Pb
concentrations estimated in the risk assessment to fall below that level.
• Log-linear function with cutpoint, for both concurrent and lifetime average blood
metrics, also applies the nonlinear relationship at blood Pb concentrations above the
lower bound of blood Pb concentrations in the pooled analysis dataset for that blood
metric, but then applies zero risk to all lower blood Pb concentrations estimated in the
risk assessment.
• Two-piece linear function, for both concurrent and lifetime average blood metrics,
applies a two-piece linear model derived from the log-linear function to all blood Pb
concentrations estimated in the risk assessment.
In the additional risk analyses performed subsequent to the August 2007 CASAC public
meeting (Section 1.4) using the core modeling approach, the first two functions listed above and
the following two functions were employed (see Section 5.3.1 of the Risk Assessment Report for
details on the forms of these functions as applied in this risk assessment).
• Population stratified dual linear function for concurrent blood Pb, derived from the
pooled dataset stratified at peak blood Pb of 10 |ig/dL and
• Population stratified dual linear function for concurrent blood Pb, derived from the
pooled dataset stratified at 7.5 |ig/dL peak blood Pb.
2-17
-------
In interpreting risk estimates derived using the various functions, consideration should be
given to the uncertainties with regard to the precision of the coefficients used for each analysis.
The coefficients for the log-linear model from Lanphear et al. (2005) had undergone a careful
development process, including sensitivity analyses, using all available data from 1,333 children.
The shape of the exposure-response relationship was first assessed through tests of linearity, then
by evaluating the restricted cubic spline model. After determining that the log-linear model
provided a good fit to the data, covariates to adjust for potential confounding were included in
the log-linear model with careful consideration of the stability of the parameter estimates. After
the multiple regression models were developed, regression diagnostics were employed to
ascertain whether the lead coefficients were affected by collinearity or influential observations.
To further investigate the stability of the model, a random-effects model (with sites random) was
applied to evaluate the results and also the effect of omitting one of the seven cohorts on the lead
coefficient was assessed. In the various sensitivity analyses performed, the coefficient from the
log-linear model was found to be robust and stable. The log-linear model, however, is not
biologically plausible at very low blood Pb concentrations as they approach zero; therefore, in
the first two functions the log-linear model is applied down to a cutpoint, selected based on the
low end of the blood Pb levels in the pooled dataset, followed by a linearization or an assumption
of zero risk at levels below that point.
In contrast, the coefficients from the two analyses using the population stratified dual
linear function with cutpoints at 7.5 ug/dL and 10 ug/dL, peak blood Pb, have not undergone
such careful development. These analyses were primarily done to compare the lead-associated
decrement at lower blood lead concentrations and higher blood lead concentrations. For these
analyses, the study population was stratified at the specified cutpoint and separate linear models
were fitted to the data above and below the cutpoint. The fit of the model or sensitivity analyses
were not conducted (or reported) on these coefficients. While these analyses are quite suitable
for the purpose of investigating whether the slope at lower concentration levels are greater
compared to higher concentration levels, use of such coefficients in a risk analysis to assess
public health impact may be inappropriate. Further, only 103 children had maximal blood lead
levels less than 7.5 ug/dL and 244 children had maximal blood lead levels less than 10 ug/dL.
While these children may better represent current blood lead levels, not fitting a single model
using all available data may lead to bias. Slob et al. (2005) noted that the usual argument for not
considering data from the high dose range is that different biological mechanisms may play a
role at higher doses compared to lower doses. However, this does not mean a single curve across
the entire exposure range cannot describe the relationship. The fitted curve merely assumes that
the underlying dose-response follows a smooth curve over the whole dose range. If biological
mechanisms change when going from lower to higher doses, this change will result in a
2-18
-------
gradually changing slope of the dose-response. The major strength of the Lanphear et al. (2005)
study was the large sample size and the pooled analysis of data from seven different cohorts. In
the case of the 7.5 ug/dL cutpoint, less than 10% of the available data is used in the analysis,
with more than half of the data coming from one cohort (Rochester) and the six other cohorts
contributing zero to 13 children to the analysis. Such an analysis dissipates the strength of the
Lanphear et al. study.
In consideration of the preceding discussion, we place greater confidence in the log-linear
model form compared to the dual-linear stratified models for our purposes in this risk
assessment. Further, in considering risk estimates derived from the four core functions (log-
linear function with low-exposure linearization, log-linear function with cutpoint, dual linear
function, stratified at 7.5 |ig/dL peak blood Pb, and dual linear function, stratified at 10 |ig/dL
peak blood Pb), we have assigned greatest confidence to risk estimates derived using the log-
linear function with low-exposure linearization since this function (a) is a nonlinear function that
describes greater response per unit blood Pb at lower blood Pb levels consistent with multiple
studies identified in the discussion above, (b) is based on fitting a function to the entire pooled
dataset (and hence uses all of the data in describing response across the range of exposures), (c)
is supported by sensitivity analyses showing the model coefficients to be robust, and (d) provides
an approach for predicting IQ loss at the lowest exposures simulated in the assessment
(consistent with the lack of evidence for a threshold). Note, however, that risk estimates
generated using the other three concentration-response functions are also presented to provide
perspective on the impact of uncertainty in this key modeling step.
2.2 USE OF CASE STUDIES AND LOCATION SELECTIONS
Consistent with the risk assessment performed during the prior Pb NAAQS review and
for the pilot phase assessment, the full-scale risk assessment for the current review relies on a
case study approach. This approach is intended to provide a framework for considering the
nature and magnitude of Pb exposures associated with air sources and associated risk to human
health, and for comparing the impact of alternate NAAQS on those exposures and risks.
With consideration of CAS AC comments on the pilot phase assessment (see Section 1.4),
the design for the full-scale assessment expanded from the pilot assessment by replacing the
urban near roadway case study with a general, non-location-specific, case study focused on
population exposures in urban areas (Section 2.2.1), while retaining the two stationary source
case studies (Sections 2.2.2.1 and 2.2.2.2). The point source case studies were intended to
illustrate risks associated with Pb near point sources, with the primary Pb smelter case study
representing one of significant impact on an individual basis, while the general urban case study
is intended to assess more widespread Pb exposures, due to the large populations in urban areas.
2-19
-------
Background information for all three case studies is presented in appendices A and B, and briefly
summarized in Sections 2.2.1 and 2.2.2, below.
Consideration of CASAC comments on the draft full-scale assessment (see Section 1.4)
led to the development of additional case studies which are presented in Chapter 5.
2.2.1 General Urban Case Study
In consideration of CASAC comments on the pilot assessment (Henderson, 2007a), a
general urban case study is included in the assessment. This case study is designed to provide
estimates of risk associated with the current and alternate NAAQS in urban areas, as well as with
current conditions. This case study differs from the point source case studies in several ways.
First it is not based on a specific location. Rather, it is designed to generally represent large
urban areas in the U.S.3 Second, as discussed further in subsequent sections (e.g., Sections 2.4.2
and 3.1.1), the media concentrations are assumed to be spatially uniform throughout the case
study area (i.e., spatial variation within the area is not considered). Third, the ambient air quality
for this case study is specified, based on analyses of Pb levels in large urban areas of the U.S.
(Appendix A and Section 2.3.1), rather than derived from air quality modeling of particular air
Pb sources. As a result of this third difference, this case study includes different types of
uncertainties as compared to the case studies employing air quality models, and it does not
distinguish among the different air Pb sources influencing the air concentrations, be they
currently active stationary or mobile sources, or resuspension of previously deposited Pb.
Fourth, the case study does not rely on any specific demographic values; that is, a theoretical
population of unspecified size is assumed to be uniformly distributed across the study area. All
of these distinctions of this case study from the others have produced a platform that is a
simplified representation of urban areas intended to inform our assessment of the impact of
changes in ambient Pb concentrations on risk.
2.2.2 Point Source Case Studies
Based on the Analysis Plan (USEPA, 2006a) and conclusions from the pilot phase risk
assessment (USEPA, 2006b; ICF, 2006), the point source case studies for this full-scale
assessment include a study area near a primary Pb smelter, and a second near another, smaller,
stationary Pb source. The locations for these two case studies were chosen with consideration of
3 While the air monitoring data used to characterize the current conditions scenario for this case study area
are from large urban areas, other empirical datasets used in developing this case study, such as those for outdoor
residential soil Pb and indoor dust Pb, are generally representative of U.S. residential properties (see Sections
3.1.1.1, 3.1.3.1 and 3.1.4.1). As most U.S. residential properties are in large urban areas because that is where a
significant share of the U.S. population resides, these datasets will include greater representation by urban areas,
particularly large ones, than non-urban areas. These datasets, however, are not limited to urban locations.
2-20
-------
factors described in the Analysis Plan: (a) availability of site-specific monitoring data for
ambient air Pb, (b) availability of measurement data for other environmental media (soil and
indoor dust) and for Pb exposure (i.e., blood Pb levels), and (c) demographic and socioeconomic
factors related to Pb exposure and risk (USEPA 2006a).
Both of these case studies rely on air quality modeling to estimate air Pb concentrations
used in the risk assessment and as a result they include different types of uncertainties as
compared to those associated with the air concentrations for the general urban case study. One
of these is related to the representation of resuspension of previously deposited Pb as a source of
Pb to ambient air. To the extent that emissions estimates are available for this source, this is
represented in these case studies. However, as recognized in Section 2.1.1 and Appendix A,
Section A. 1.1.3, such emissions estimates are often uncertain or unavailable. Some resuspension
sources are included in the primary Pb smelter case study, however no such sources are explicitly
modeled in the secondary Pb smelter case study.
Background information for these case studies is briefly summarized below with regard
to: (a) population characteristics, (b) reported emissions, (c) ambient air Pb levels, and (d) the
availability of site-specific data characterizing levels of Pb in key media and Pb exposures (e.g.,
soil, blood Pb level data).
2.2.2.1 Primary Pb Smelter Case Study
The primary Pb smelter case study is focused on the only currently operating facility in
the U.S., which is located in Herculaneum, Missouri. At primary Pb smelters, Pb-bearing ore
concentrates are smelted to produce Pb metal. This smelter is one of the largest individual
sources of Pb metal in the U.S., has been active for over a century and there exist a large amount
of site-specific data in the surrounding area characterizing both media concentrations (soil,
indoor dust, outdoor air) and population blood Pb levels (Appendix B, Section B.I).
The facility's century of operation has contributed to Pb contamination of the area
surrounding the facility4, which has triggered various remediation activities (e.g., removal of Pb-
contaminated residential soil) as well as enlargement of the facility property to encompass many
of the most heavily impacted private properties. The remediation activity introduces a
complication to the risk modeling, especially aspects involving characterization of the
relationship of ambient air Pb and residential soil Pb to indoor dust Pb (see Section 3.1.4).
4 Portions of this study area comprise an active Superfund site and are subject to ongoing evaluation under
the Superfund program administered by the Office of Solid Waste and Emergency Response. Methods used in
conducting the human health exposure and risk assessment for the pilot analysis have been selected to address policy
questions relevant to the Pb NAAQS review and consequently may differ from those used by the Superfund
program.
2-21
-------
Some key aspects of the background information for this case study (Appendix B) are
summarized briefly in Table 2-1.
2-22
-------
Table 2-1. Key aspects of primary Pb smelter case study.
Population
As of the 2000 U.S. Census, approximately 38,000 people lived within 10 kilometers
(km) of the facility, 10% of which were children less than 8 years of age. Since 2000,
actions associated with reducing facility-related Pb exposures have reduced the
population size within 2 km of the facility such that six previously occupied census
blocks near the facility were unoccupied in 2004. With those counts subtracted from
the 2000 counts, the numbers of children (less than 8 years old) residing within 2 km,
between 2 and 5 km and between 5 and 10 km were 171, 1,545 and 2,164,
respectively (Appendix B).
Emissions
Lead is emitted from a wide variety of activities associated with the primary Pb
smelter, including the transport of materials into and within the facility. The facility is
estimated to be the largest Pb emitter in the U.S. (Appendix B, Section B.1 and
Appendix A, Section A.1).
Air Quality
In 2005, annual average concentrations of Pb in total suspended particulate matter
(Pb-TSP) at the nine monitors in this town, for which data are reported in the U.S. Air
Quality System (AQS), ranged from 0.046 to1.56 ug/m3. All of these nine monitors,
fall within the top 30% of the 2005 annual average levels at AQS monitors nationally,
with four of the nine monitors falling in the top 10% (Appendix B). Maximum quarterly
average Pb-TSP concentrations at one of these monitors exceeded the current
NAAQS in 2005, 2006 and 2007 (Appendix A; AQS).
The area within the city limits of Herculaneum is designated nonattainment for the Pb
NAAQS and the existing State Implementation Plan (SIP) was approved in 2002 (67 FR 18497).
EPA determined the existing SIP to be inadequate to attain and maintain the current NAAQS in
2006 (71 FR 19432), and consequently the state of Missouri developed a revised SIP for the area.
U.S. EPA, Region 7 received Missouri's proposed SIP revision on May 31, 2007 (MDNR,
2007). The air dispersion modeling performed for the risk assessment described in this
document built on the information and modeling developed for the revised SIP.5
The significant amount of site-specific data available for Herculaneum, paired with air
dispersion modeling for the facility conducted in support of SIP development for Pb, provides a
substantial data set for this study area which enhances the modeling of exposure and risk. For
example, the Herculaneum facility has more site-specific monitoring data available to support
risk assessment than the second point source case study location, including residential yard soil,
indoor dust and road dust Pb measurements collected in areas potentially impacted by the
facility. In addition, the Agency for Toxic Substances and Disease Registry (ATSDR) has
conducted a number of health consultations that involved the collection of blood Pb
measurements for children (Appendix B, Section B. 1.3). Extensive air Pb monitoring data are
also available and were considered in the performance evaluation of the modeling setup. The
Herculaneum case study location also has a number of attributes that add complexity to the
modeling of Pb exposure and risk including (a) complex terrain and meteorology which
' The 2007 draft SIP revision (MDNR, 2007), including the modeling, is currently under review by EPA.
2-23
-------
complicate the modeling of Pb transport in ambient air, (b) a large and complex facility with a
long history of operation and significant opportunity for fugitive emissions, making source
characterization challenging, and (c) a history of remediation activities which has contributed to
widely varying residential soil Pb concentrations across the town.
2.2.2.2 Secondary Pb Smelter Case Study
The secondary Pb smelter case study, in Troy, Alabama, involves a smaller point source
than the primary Pb smelter case study, with relatively less site-specific data characterizing
media concentrations and exposure levels. Secondary Pb smelters produce Pb from scrap and
provide the primary means for recycling Pb acid automotive batteries, and are among the larger
source categories of Pb emitters (see Appendix A, Section A. 1). The Troy facility was one of 15
secondary Pb smelters operating within the U.S. in 2002 (see Appendix B, Section B.2). Some
key aspects of the background information for this case study (Appendix B) are summarized
briefly in Table 2-2.
Table 2-2. Key aspects of secondary Pb smelter case study.
Population
As of the 2000 U.S. Census, approximately 18,000 people lived within 10 km of the facility,
10 percent of which were children less than 8 years of age. Specifically, 187 children of
that age group lived within 2 km of the facility, 896 lived between 2 and 5 km and 589 lived
are between 5 and 10 km from the facility (Appendix B).
Emissions
Lead is emitted from the facility operations, from materials storage and handling, and from
facility roadway dust. Similar to most secondary Pb smelters, emissions from this facility
are estimated to fall between 1 and 5 tpy (see Appendix B, Section B.2.2 and Appendix A,
Attachment A-1).
Air Quality
Annual average concentrations of Pb-TSP for 2005 at the two monitors located within 1 km
of the Troy facility (300 and 800 m from the facility), for which data are reported in the U.S.
Air Quality System, are approximately 0.4 and 0.1 ug/m3, respectively. These values fall
within the top 15% of Pb-TSP annual average values for 2005 (see Appendix B, Section
B.2.5.1). Maximum quarterly average Pb-TSP concentrations at one of the monitors
exceeded the current NAAQS in 2003 (Appendix A).
In contrast to the Herculaneum facility, we have not identified soil or indoor dust Pb
measurements for this case study location. Additionally, although there are blood Pb
measurements in children available at the county level, they are not available at a more refined
scale that might relate more directly to this case study. The relative sparseness of site-specific
Pb measurements means that the exposure assessment conducted for the secondary Pb smelter
case study is more dependent on model projections, and consideration of measurements available
for similar locations, and that there is less opportunity for rigorous performance evaluation of the
modeling steps. The available air Pb monitoring data, however, are used in the performance
evaluation of the air quality modeling.
2-24
-------
2.3 ASSESSMENT SCENARIOS
The design of the scenarios assessed for each case study includes aspects particular to air
concentrations of Pb (Section 2.3.1), surface soil/dust concentrations of Pb (Section 2.3.3) and
background (Section 2.3.2). As the scenarios are primarily distinguished by the differences in air
concentrations, we generally refer to the different assessment scenarios as air quality scenarios in
this document (including the appendices). The different air quality scenarios include current
conditions, meeting the current NAAQS of 1.5 |ig/m3 (maximum quarterly average) and meeting
several alternate, lower NAAQS.
2.3.1 Air Concentrations
The air concentrations assessed in the different air quality scenarios include current
conditions, meeting the current NAAQS of 1.5 |ig/m3 (maximum quarterly average) and meeting
several alternate, lower NAAQS. In consideration of the range of levels suggested by CASAC
(Henderson, 2007a), the alternate NAAQS scenarios included in the assessment are: 0.5 |ig/m3,
0.2 |ig/m3 and 0.05 |ig/m3 as maximum monthly averages, and 0.2 |ig/m3 as a maximum
quarterly average. In response to discussion at the August 2007 meeting of the CASAC Pb Panel
to extend lower the range of alternate levels considered, an alternate NAAQS scenario of 0.02
|ig/m3, as a maximum monthly average, is also included in the additional analyses presented in
Chapter 5. While the current and alternate NAAQS scenarios are characterized by quarterly or
monthly averaging times, it is the associated annual average ambient air concentrations that are
then used in the risk assessment6.
The current conditions scenario, performed for the general urban and secondary Pb
smelter case studies, is intended to generally reflect recent conditions for these case studies based
on data available for the characterization. For example, for the urban case study, air Pb levels for
current conditions are based on 2003-2005 air quality data (Appendix A). For the secondary Pb
smelter case study, for which we used air quality modeling, air Pb levels for current conditions
are based on emissions characterizations drawn from currently available emissions information
and recent meteorological data (see Appendix E).
The current NAAQS attainment air quality scenario was performed for the primary Pb
smelter case study, for which current monitoring data indicate exceedance of the current Pb
6 Use of the annual average concentration is consistent with the temporal period of this input to the primary
blood Pb model (Appendix H) and also the generally longer term resolution of the blood Pb metrics associated with
the concentration-response functions ("concurrent" and "lifetime average") (Lanphear et al., 2005).
2-25
-------
NAAQS (Appendix B, Section B.I.5.1).7 Additionally, although we consider it extremely
unlikely that air concentrations in urban areas across the U.S. that are well below the current
standard would increase to just meeting the standard, we recognize the potential for air Pb
concentrations in some areas currently well below the standard to increase to just meet standard
by way of, for example, expansion of existing sources (e.g., facilities operating as secondary
smelters exercising previously used capabilities as primary smelters) or by the congregation of
multiple Pb sources in adjacent locations. Accordingly, we have simulated this type of scenario
(increased Pb concentrations to just meet the current standard) in the general urban case study
and in the additional analyses for three location-specific urban case studies described in Chapter
5.
In developing the reduced air concentrations for the alternate NAAQS scenarios for the
point source case studies (for which air quality models are employed), the maximum monthly or
quarterly average (depending on averaging time for the alternate NAAQS) for each modeled
receptor point is compared to the NAAQS level to identify the factor by which the highest
average exceeds the NAAQS level. All monthly or quarterly averages are then reduced by this
factor (i.e., a proportional roll back is implemented) and the associated annual average
recalculated for each receptor point.
Two different current conditions scenarios are assessed for the general urban case study
based on air Pb concentrations for the period 2003-2005 at monitors in U.S. urban areas with
population greater than one million (Appendix A).8 One of these two scenarios is based on the
mean maximum calendar quarter average for these monitors, and the second is based on the 95th
percentile of maximum calendar quarter averages for these monitors. Additionally, although the
mean and 95th percentile maximum quarterly average of the large urban area monitors nationally
do not exceed the current NAAQS level, an increased air concentration scenario (i.e., to the level
of the current standard) has been included for this case study.
As the air Pb concentrations for this case study do not vary spatially (see Section 3.1.1.1),
the air Pb concentration is simply set to the level specified for each scenario as either a monthly
or quarterly average. The annual average air concentration (the metric used in the dust and blood
7 Given the status of this area with regard to nonattainment and SIP revision, as well as the use of the
modeling set-up developed for the SIP attainment demonstration, a current conditions scenario was not developed
for the primary Pb smelter case study.
8 In designing the general urban case study, we recognized CAS AC's interest in our development of risk
estimates that were more informative to the larger national population and exposures related to ambient air reflecting
the aggregate contributions from multiple sources of Pb than might be discrete populations near particular point
sources of Pb (Henderson, 2007a). Urban areas of greater than one million population were identified as
informative for large populations in established cities (as compared to more newly-developed urban areas).
2-26
-------
Pb modeling) is derived from the current conditions or NAAQS level (and averaging time) using
relationships based on current Pb-TSP monitoring data for monitors in large U.S. urban areas
(Appendix A). Lead in TSP monitoring data for the time period 2003-2005 from monitors in
U.S. urban areas of population size greater than one million were analyzed to derive estimates of
maximum quarterly average, maximum monthly average and annual average for each monitor.
From these estimates, the ratios of the maximum quarterly and maximum monthly averages to
the annual average were derived for each monitor and the arithmetic mean and 95th percentile
value of the monitor-specific ratios were derived. To derive the annual average air concentration
used in the dust and blood Pb modeling for each scenario, one of these ratios was applied to the
air quality scenario level. For example, the alternate NAAQS level of 0.5 (for a maximum
monthly averaging time) was divided by the mean monitor ratio of maximum monthly average to
annual average to derive the annual average air concentration estimate for that alternate NAAQS
scenario. The air quality values associated with the different scenarios assessed for the urban
case study are summarized in Table 2-3.
2-27
-------
1 Table 2-3. Air quality scenarios assessed for the general urban case study.
Air Quality Scenario
95tn Percentile Current Conditions
Mean Current Conditions
Alternate NAAQS
Alternate NAAQS
Alternate NAAQS
Alternate NAAQS
Current NAAQS
Level
(ug/m3)
0.87
0.14
0.5
0.2
0.2
0.05
1.5
Averaging Time
(Form)
calendar quarter
(maximum)
calendar quarter
(maximum)
Month
(maximum)
Month
(maximum)
calendar quarter
(maximum)
Month
(maximum)
calendar quarter
(maximum)
Ratio
7.6a
2.5°
4.0C
4.0C
2.5°
4.0C
2.5°
Associated Annual
Average Concentration
(ug/m3)
0.11
0.056
0.13
0.05
0.08
0.013
0.60
3 This is the 95m percentile of the ratios of maximum quarterly average to annual average for monitors at sites in
urban areas with population of one million or more people.
" This is the mean of the ratios of maximum quarterly average to annual average for monitors at sites in urban
areas with population of one million or more people.
cThis is the mean of the ratios of maximum monthly average to annual average for monitors at sites in urban areas
with population of one million or more people.
3 2.3.2 Policy-relevant Background
4 Given the multimedia, multipathway nature of Pb, levels of Pb in all exposure media
5 (including those other than air) are essential aspects of the scenarios assessed for each case study.
6 As discussed in Section 1.1, some of the Pb in other media may be derived from policy-relevant
7 sources, while, for our purposes here, we have categorized others as policy-relevant background.
8 Some amount of Pb in the air also derives from background sources, such as volcanoes, sea salt,
9 and windborne soil particles from areas free of anthropogenic activity (CD, Section 2.2.1). The
10 impact of these sources on current air concentrations is expected to be quite low and has been
11 estimated to fall within the range from 0.00002 |ig/m3 and 0.00007 |ig/m3 based on mass balance
12 calculations (CD, Section 3.1 and USEPA 1986, Section 7.2.1.1.3). The midpoint in this range,
13 0.00005 |ig/m3, has been used in the past to represent the contribution of naturally occurring air
14 Pb to total human exposure (USEPA 1986, Section 7.2.1.1.3). It is noted that the data available
15 to derive such an estimate are limited and that such a value might be expected to vary
16 geographically with the natural distribution of Pb. Comparing this to reported air Pb
17 measurements is complicated by limitations of the common analytical methods and by
18 inconsistent reporting practices. This value is one half the lowest reported nonzero value in
19 AQS. For the purposes of this assessment, however, the value of 0.00005 |ig/m3 was selected as
20 representative of policy-relevant background Pb in air. Unlike for other criteria pollutants, the
2-28
-------
1 role of this value for Pb is limited. In considering risk contributions from policy-relevant
2 background, the contributions from exposures to nonair media are such that any credible estimate
3 of policy-relevant background in air is likely insignificant in comparison. In developing the air
4 Pb concentrations associated with the alternate NAAQS scenarios, the estimate of policy-
5 relevant background in air was the floor below which concentrations were not lowered.
6 2.3.3 Outdoor Soil/Dust
7 With regard to surface soil9 Pb concentrations for the alternate NAAQS scenarios, the
8 presence of historically deposited Pb, associated with past periods of higher air concentrations
9 and associated atmospheric deposition, affects the soil Pb dynamics. Under the alternate
10 NAAQS scenarios, atmospheric deposition of Pb will continue to occur, albeit it will be reduced
11 from the current rate which is reduced from historic rates. At any location, the type of response
12 of the surface soil concentrations to the changed deposition rate will depend on the relationship
13 of current surface soil concentrations to the surface soil concentrations associated with a steady
14 state condition (i.e., when the rate of Pb addition to the surface soil equals the rate of Pb loss
15 from the surface soil) at that location. If current surface soil concentrations are below their
16 steady state levels for the current conditions (i.e., the rate of Pb addition is greater than the rate
17 of Pb loss) and air concentrations are reduced (i.e., the rate of Pb addition via deposition is
18 reduced), surface soil concentrations might be expected to continue an increasing trend, although
19 at a reduced rate from the current rate. Alternatively, if current surface soil concentrations are
20 above their steady state levels (i.e., the rate of Pb loss is greater than the rate of Pb addition) and
21 air concentrations and associated deposition are reduced, surface soil concentrations might be
22 expected to continue a decreasing trend and do so at a greater rate than the current one.
23 Information regarding the current dynamics of Pb concentrations in surface soils is
24 limited with the predominant focus for such studies being somewhat remote forested areas (e.g.,
25 CD, Section 3.2 and 3.2.2 and pp. AX7-33 to AX7-34). Findings to date indicate that systems
26 with little influence from local point sources are still responding to reduced Pb deposition rates
27 associated with reduced atmospheric emissions of Pb, including those associated with the phase-
28 out of leaded gasoline. For example, studies of forest soils have concluded that surface
29 concentrations of Pb are decreasing in response to the reduced Pb deposition rates since the
30 phase-out of leaded gasoline (Miller and Friedland, 1994; Kaste and Friedland, 2003). Studies in
9 In the risk assessment, outdoor surface soil or dust is an important exposure pathway. Use of the term
surface soil here is intended to include the terms outdoor soil and outdoor dust, with there being some overlap
between those two terms in that the surface layer of outdoor soil might be referred to as outdoor dust. Specifically,
the phrase "outdoor dust" refers to particles deposited on any outdoor surface, including, for example, soil,
sidewalks, roadways, etc.
2-29
-------
1 urban areas of southern California, where Pb has accumulated from past sources, suggest an
2 environment in which Pb may remain at the soil surface (and other surfaces), contributing to air
3 concentrations via resuspension in the near term (CD, pp. 2-65 to 2-67 and 3-18 to 3-19).
4 Accordingly, the temporal trend in surface soil concentrations in this environment is considered
5 to be influenced by the rate of resuspension, such that little to no reduction in soil Pb
6 concentration in southern California is expected over the next few hundred years (CD, pp. 2-65
7 to 2-67 and 3-18 to 3-20; Harris and Davidson, 2005). Temporal trends in surface soils near
8 established point sources are not well characterized. Available information for a few areas
9 surrounding smelters after implementation of pollution controls shows a decline in Pb
10 concentrations in outdoor dustfall, street dust and indoor dustfall, but has not indicated a
11 noticeable decline in soil Pb concentrations (CD, pp. 3-23 to 3-24).
12 The above discussion suggests that a reduced air concentration in the three case studies
13 would not be expected to yield a changed surface soil concentration over the near term, yet may
14 yield a reduced surface concentration over a much longer term. An exception to this may be
15 some areas of the primary Pb smelter case study where contaminated soil has been removed and
16 replaced with "clean" soil. Measurements taken of Pb concentrations in such "clean" soil placed
17 within % mile of the facility exhibit small increasing temporal trends over a period of a few
18 year10 (USEPA, 2006c). In lieu of additional data or a multimedia modeling analysis, however,
19 the surface soil concentrations for the current and alternate NAAQS scenarios in all case studies
20 have been set equal to those used for the current conditions scenarios. This is generally believed
21 to be a reasonable representation of soil Pb response to alternate NAAQS for at least six years,
22 and likely much longer, after a new standard might be implemented.11 A potential exception is
23 the area of the primary Pb smelter case study within 3/4 mile of the facility, where it may be that
24 surface Pb concentrations in remediated soil may increase to higher levels under the current and
25 some of the alternate NAAQS. This remains an area of uncertainty with potential implications
26 for areas in which a Pb source or combination of Pb sources may locate where ones of
27 comparable size had not been previously.
28 Additionally, we recognize that implementation of some alternate NAAQS could in some
29 areas (e.g., areas of substantial past atmospheric deposition) involve control of surface soil/dust
30 to reduce surface soil/dust Pb levels. That is, in places where surface soil/dust Pb concentrations
31 contribute substantially to air concentrations, controls implemented to attain various alternate
32 lower NAAQS might include reducing soil Pb concentrations. Such specific control actions have
10 An increasing trend was not seen with soil at the two locations just beyond 3/4 miles away.
11 This was also the approach used in the risk assessment performed in the last Pb NAAQS review
(USEPA, 1990).
2-30
-------
1 not been addressed in this assessment, and as stated above, outdoor soil/dust concentrations in all
2 air quality scenarios have been set equal to the values for the current conditions scenarios.
3 2.4 ANALYTICAL APPROACH
4 This section provides an overview of the analytical approach, describing key elements
5 including: (a) temporal aspects, (b) spatial scale of the analysis and the type of spatial template
6 used in modeling, (c) overview of the analytical steps of predicting media concentrations,
7 modeling exposure, and modeling risk, (d) performance evaluation completed in support of the
8 analysis and (e) the approach used to characterize uncertainty and evaluate model sensitivity.
9 The approach described here pertains to the initial analyses of the full-scale assessment
10 for which results are described in Chapters 3 and 4, and to some aspects of the core modeling
11 approach employed for the additional analyses described in Chapter 5.
12 2.4.1 Temporal Aspects
13 The risk assessment conducted for each case study uses a simulated child population for
14 which exposure begins at birth and continues for 7 years. That is, the study population is
15 assumed to be a single group, for which exposure begins at birth and continues until the group
16 reaches 7 years of age.12 Furthermore, it is assumed that no migration or immigration of these
17 children occurs during this simulation period; that is, none of the children move out of the study
18 area and no children move in.
19 For the point source case studies, the use of modeling (with a constant emissions rate and
20 temporally varying meteorology) provides temporally varying air concentrations. However, the
21 primary blood Pb model for this assessment is limited in the temporal resolution of its inputs (see
22 Section 3.2), because the finest temporal resolution of inputs to the blood Pb model is a year.
23 Consequently, in characterizing exposure media concentrations, annual averages are used.
24 With regard to temporal variation across the seven-year exposure period, several
25 exposure factors and physiological parameters are varied on an annual basis within the blood Pb
26 modeling step (see Section 3.2). Once set for the air quality scenario, however, the media
27 concentrations of Pb are held constant throughout the seven year period (see Section 3.1).
28 2.4.2 Spatial Scale and Resolution
29 The size and resolution of the study area differed among the three case studies. The
30 general urban case study is not set in a specific location, and involves non-spatially-varying
31 media concentrations and population density. There is no spatially differentiated template per se
12 Modeling of blood Pb levels for the child population includes contributions representative of prenatal Pb
exposure.
2-31
-------
1 and instead, a single generic urban study area is assumed with uniform population density and
2 exposure concentrations. For the exposure modeling for the point source case studies, however,
3 spatial templates were developed. The templates subdivide the study area into subunits,
4 composed of U.S. Census blocks or block groups, across which media concentrations differ.13
5 Media concentration estimates (e.g., for outdoor air, soil and indoor dust) are developed for each
6 block or block group, and from these a central tendency estimate is developed of concurrent and
7 lifetime average blood Pb levels for the resident children. Interindividual variability of blood Pb
8 levels for children within a block or block group is considered through the use of a statistically
9 derived GSD. The specific spatial templates used for each of the point source case studies are
10 presented in Appendices D (Section D. 1) and E (Section E. 1).
11 For the two point source case studies, population risk estimates were derived separately
12 for a portion of the full study area in addition to the estimates derived for the full study area. The
13 subareas extended approximately 1.5 km out from the point source locations (see Appendix P).
14 2.4.3 Categorization of Policy-relevant Exposure Pathways
15 To inform policy aspects of the Pb NAAQS review, we have attempted to parse the
16 assessment estimates for indoor dust Pb, blood Pb and IQ loss into the fraction associated with
17 policy-relevant background (e.g., diet and drinking water) versus that associated with policy-
18 relevant pathways, which include inhalation, outdoor soil/dust ingestion and indoor dust
19 ingestion (Section 2.1.1). We have further categorized the policy-relevant pathways into one of
20 two categories, "recent air" or "past air". Conceptually, the recent air category includes those
21 pathways involving Pb that is or has recently been in the air, whether or not it was also in the air
22 in the past, and the past air category includes those pathways involving Pb that was in the air in
23 the past and was not in the air recently. As discussed below, this conceptual distinction could
24 not be entirely reflected in the risk assessment due to technical limitations. Thus, risk estimates
25 associated with "past air" reflect some Pb that was recently in the air. Similarly, diet and
26 drinking water are treated entirely as policy-relevant background, despite also reflecting
27 contributions from policy-relevant pathways such as atmospheric deposition.
28 Conceptually, recent air refers to exposure contributions associated with inhalation of
29 ambient air Pb and ingestion of the fraction of indoor dust Pb derived from recent ambient air Pb.
30 To the extent that ambient air Pb includes contributions from resuspension of previously
13 US Census block groups vary in size from several city blocks in densely populated urban areas to many
square miles in less populated rural areas. Their population count varies from 600 to 3000 people per block group
with the typical block group in the U.S. containing 1,500 people. US Census blocks are more refined than block
groups and typically contain several hundred people or less. Their size can vary from a single city block in urban
areas to multiple square miles in less populated rural locations.
2-32
-------
1 deposited Pb, that source is represented in the recent air category. Thus, a "recent air" exposure
2 may involve previously deposited Pb that is (1) resuspended into the air and inhaled or (2)
3 resuspended into the air, transported into a building, deposited into the dust, contacted and
4 ingested.
5 Conceptually, past air includes exposure contributions from ingestion of outdoor soil/dust
6 that is contacted on surfaces outdoors, and ingestion of indoor dust Pb that is derived from past
7 air sources. Although Pb that is currently in outdoor soil/dust may have been in the air recently
8 or some time ago, we have assigned ingestion of outdoor soil/dust Pb contacted outdoors to the
9 past air category in recognition of our inability to maintain a dynamically changing
10 categorization of recent versus past air.14 The past air category also includes the ingestion of
11 indoor dust Pb that was in the ambient air in the past but not recently. These sources to indoor
12 dust Pb include any residual legacy of historical air Pb in the indoor dust of older homes, as well
13 as Pb occurring in indoor dust that is derived from outdoor soil/dust Pb that was not transported
14 indoors by an air pathway. This latter pathway includes outdoor soil/dust Pb that is carried
15 indoors by human contact (e.g., "tracking in").
16 To implement this categorization in the assessment, we developed estimates of the recent
17 air portion of indoor dust Pb (i.e., contributions associated with recent ambient air Pb levels), and
18 assigned the remainder of indoor dust Pb to other sources, which include those relevant to past
19 air. That is, the "other sources" component of indoor dust Pb refers to contributions from indoor
20 paint, outdoor soil/dust and additional sources. Among the additional sources is any residual
21 legacy of historical Pb in the indoor dust of older homes.
22 The indoor dust Pb subdivision reflects and is limited by the models and inputs used to
23 estimate indoor dust Pb levels for the different scenarios. All of them predict dust Pb
24 concentration as a function of, among other factors, ambient air Pb concentration, and all of the
25 models include a constant (e.g., the intercept in the regression-based models) that captures
26 "other" sources, as well as uncertainty associated with the relationship with ambient air Pb.15
27 One of the models (used in the secondary Pb smelter case study and in part of the primary Pb
28 smelter case study) also includes a dependency on outdoor soil/dust Pb concentration. This
29 difference among the models leads to an inconsistency across the case studies in the ability to
30 separate the contribution to indoor dust Pb from outdoor soil/dust. Consequently, we have
14 In concept, this assignment appears to inherently contribute to an underestimate of the recent air
category. However, the reality of much higher air emissions in the past that have contributed to Pb concentrations in
other media that are higher than those that would be associated with more recent lower emissions, complicates
consideration of this assumption (see Section 2.3.3).
15 The extent to which this intercept captures uncertainty about the relationship with ambient air will
indicate presence of a recent air pathway in the past air category.
2-33
-------
1 limited to two categories the subdivision of indoor dust contributions, with that from outdoor
2 soil/dust included in "other" whether it is estimated by a model that includes a soil concentration
3 coefficient or is accommodated by the constant or intercept term. Additionally, we have not
4 subdivided indoor dust Pb estimates for the primary Pb smelter due to uncertainty regarding this
5 categorization with the model used for that case study (Section 3.1.4.2). In presenting risk
6 estimates associated with policy-relevant pathways (Chapter 4), we have included risk associated
7 with the "other" component of indoor dust in the past air category. We recognize that the
8 potential for this "other" component to include ambient Pb unrelated to air emissions contributes
9 to a potential for the risk estimates associated with past air to be overestimates. Additionally, as
10 the recent air portion of indoor dust Pb, which derives from transport of airborne Pb into a house,
11 depends on the estimate of ambient air Pb concentration, selection of an ambient air Pb estimate
12 that is not appropriate to this relationship (e.g., one that does not represent ambient air Pb levels
13 immediately outside the house) may contribute to an under- or overestimate of recent air indoor
14 dustPb.
15 There is inherent uncertainty associated with the approaches used to divide indoor dust-
16 related Pb exposures and risk into contributions from "recent ambient air" and from "other"
17 sources. Further, the uncertainty may differ among the three case studies due to the different
18 approaches used in modeling indoor dust Pb. For example, uncertainty associated with the
19 hybrid mechanistic-empirical model used in the general urban case study includes that which
20 arises from model inputs and model performance, while the empirical, regression-based,
21 statistical models used in the point source case studies entail uncertainty regarding similarity of
22 the conditions from which the model was derived to those to which it is applied, as well as
23 uncertainty regarding variables that may be correlated with those explicitly represented in the
24 model. Thus, given the various limitations of our modeling tools, blood Pb levels associated
25 with air-related exposure pathways and current levels of Pb emitted to the air (including via
26 resuspension) are likely to fall between the estimates derived for "recent air" pathways and those
27 for "recent" plus "past air" pathways.
28 2.4.4 Overview of Analytical Steps
29 As illustrated in Figure 2-2, the risk assessment completed for the two point source case
30 studies generally includes four analytical steps: (a) fate and transport of Pb released into outdoor
31 air, including the dispersion of Pb away from the point of release and the deposition of Pb onto
32 surfaces, (b) prediction of the resulting concentration of Pb in media of concern including
33 outdoor air, outdoor surface soil/dust and indoor dust, (c) use of these Pb concentrations together
34 with estimates of Pb in background exposure pathways, including diet, to estimate associated
35 blood Pb levels in children using biokinetic modeling and (d) use of concentration-response
2-34
-------
1 functions derived from epidemiology studies to estimate IQ loss associated with the estimated
2 blood Pb levels. The modeling approach for the general urban case study is somewhat simpler,
3 since it does not involve fate and transport modeling for air concentration estimates and instead,
4 uses ambient monitor levels combined with an assumption of uniform ambient air Pb levels
5 across the study area. Subsequent steps in the general urban case study analysis are fairly similar
6 to what is described above for the point source case studies, with the generation of population
7 blood Pb levels being somewhat simplified for the general urban case study. Figure 2-2
8 identifies the key input data sets, modeling steps and intermediate model output in each of the
9 four analytical steps. The first three steps are employed in the exposure assessment (Section
10 2.4.4.1), while the fourth is the risk assessment step (Section 2.4.4.2).
11 Prior to focusing on a single modeling approach (see Chapter 5), and because of the
12 quantitative influence of certain analytical steps on the results, we employed multiple approaches
13 to perform some of the analytical steps. The multiple sets of results generated in this way for a
14 case study and air quality scenario are intended to span a range indicative of the model and
15 parameter uncertainty associated with those analytical steps of the risk analysis (see Section
16 2.4.6).
2-35
-------
1 Figure 2-2. Overview of analysis approach.
GO
GO
w
GO
GO
Characterizing ambient air levels and
deposition to soil
Ambient monitoring data (mean and
high-end values) (urban)
Air dispersion modeling
(primary, secondary Pb smelters)
Ambient air
concentrations
for study areas
(
Lead deposition rates
across study area
Characterizing soil and indoor
dust concentrations
Nationally-
representative
residential soil Pb
value selected from
literature
(urban)
Multi-media fate & transport
model-derived soil
Combination of (a) statistical concentrations scaled using
(regression) and/or (b) mechanistic measured data from surrogate
(compartmental) indoor dust prediction <— location
.. ,v (all3^ud1es) I (secondaryPb smelter) j Site-specific soil
W I V J \ monitoring data
combined with
statistical
Indoor residential dust I t I Outdoor residential extrapolation
concentrations j j soil concentrations (primary Pb
"\ " | '' smelter)
Characterizing blood Pb levels 4 f '*•;
( . ~~~ 1 Ambient air
1 Biokmetic blood Pb .
„ , , _, , , . , ,. | concentrations
Background Pb exposure levels i w modeling ^ ' , ,
,. , (see above)
• diet (all 3 case studies)
• drinking water ' ^ 1 '
• indoor paint (actually reflected in |
dust modeling) I GSD reflecting inter-
1 s ' ^ individual variability
• 1 Probabilistic modeling of blood Pb I in behavior related to
Demographics (distribution of levels for children within each case | Pb exposure and
children within study areas) —' study location biokinetics
(two point source case studies) (all 3 case studies) '"' •*'
: ' ' V.
§ :
H ! Characterizing risk (IQ loss)
Concentration-response function
• log-normal distribution (Lanphear pooled
Risk estimation - IQ
Distribution of IQ loss for change estimation 4 ' " breakpoint reflecting confidence in CR
study populations , ,, , , ,. -. function
i (all j case stucti&s)
(partitioned between policy- •^-J l_ J ' effects estimation based on concurrent and
relevant exposures and lifetime average blood Pb metric
policy-relevant background)
GO I ''••• /
2
2-36
-------
2 2.4.4.1 Exposure Assessment
3 Concentrations of Pb are estimated in ambient media and indoor dust using a
4 combination of empirical data and modeling projections. The use of empirical data brings with it
5 uncertainty related to the potential inclusion of background source signals in these measurements
6 (e.g., house paint contributions to indoor dust and outdoor soil Pb). Conversely, the use of
7 modeling tools introduces other uncertainties (e.g., model and parameter uncertainties). Both of
8 these uncertainties are recognized in Section 4.3. Specific approaches used at the three case
9 study locations are briefly described below.
10 Characterization of Pb in ambient air relies on (a) dispersion modeling of facility-related
11 (including fugitive) Pb emissions for the primary and secondary Pb smelter case studies and (b)
12 the use of ambient monitor data for the general urban case study. For the general urban case
13 study, monitoring data for U.S. urban areas of more than a million in population were used to
14 identify two current conditions scenarios, one "typical", and one higher end, and to relate
15 alternate NAAQS (of other forms) to annual average levels needed for blood Pb modeling. A
16 key aspect of the general urban case study is that ambient air lead levels do not vary spatially
17 within the study area. The approach is monitor-based rather than source-based as compared to
18 two point source case studies. This means that we did not explicitly model specific source
19 contributions for the general urban case study (e.g., resuspension of roadside dust, "fresh"
20 industrial emissions) and instead, relied on empirical data to define ambient air Pb levels for this
21 general case study, with these levels reflecting contributions from all contributing sources, be
22 they currently active stationary or mobile sources, resuspension of previously deposited Pb or
23 other.
24 Characterization of Pb concentrations in outdoor surface soil/dust, resulting from
25 deposition of airborne Pb is based on the use of (a) existing site-specific measurements (primary
26 Pb smelter case study), (b) nationally representative residential soil measurements obtained from
27 the literature (general urban case )study and (c) fate and transport modeling (secondary Pb
28 smelter case study). In the case of the primary Pb smelter case study, soil Pb concentration data
29 were available for a zone close to the facility and statistical extrapolation from the available
30 empirical data was used to predict soil levels for portions of the study area beyond this zone.
31 To predict concentrations of ambient Pb in indoor dust, we have relied on a combination
32 of (a) regression-based models that relate indoor dust to outdoor air Pb and/or outdoor soil Pb
33 and (b) mechanistic models that predict indoor dust Pb based on key mechanisms (e.g., exchange
34 of outdoor air with indoor air, deposition rates of Pb to indoor surfaces, house cleaning rates).
35 For both point source case studies, a combination of regression-based models obtained from the
2-37
-------
1 literature and developed based on site-specific data were used, and a customized hybrid
2 empirical-mechanistic model was developed for the general urban case study. This reflected the
3 fact that available regression-based models had been developed largely based on residential
4 exposures near large point sources and were not considered representative of more general urban
5 exposures. Consequently, a mechanistic model, augmented with empirical data, was developed
6 for the general urban case study. Additional detail on methods used to characterize media Pb
7 concentrations for each case study can be found in Section 3.1.
8 Blood Pb levels are predicted from estimates of Pb contained in various media (e.g.,
9 ambient air, diet, water, indoor dust) using the Integrated Exposure and Uptake Biokinetic
10 (TEUBK) model (Section 3.2.1.1). A second biokinetic model, the International Commission for
11 Radiation Protection model (hereafter referred to as the "Leggett model"), is included in the
12 sensitivity analysis (Section 3.2.1.2).16 The same fundamental approach was used to estimate
13 population distributions of blood Pb levels for each of the two point source case studies, and a
14 somewhat simpler approach was used for the general urban case study. The approach used for
15 the two point source case studies involved two main steps:
16 1) Use biokinetic model to predict central tendency blood Pb levels for children within
17 each exposure zone: The model outputs are then aggregated into the "concurrent"
18 and "lifetime average" blood Pb metrics used in the concentration-response functions.
19 2) Implement probabilistic exposure model to generate a population distribution of
20 blood Pb levels for children in each case study location: The probabilistic model
21 generates a distribution of simulated blood Pb levels for the children in each study
22 area based on consideration of three key factors: (a) the central tendency blood Pb
23 levels generated for each exposure zone in the preceding step, (b) demographic data
24 (distribution of children 0-7 years of age) across the zones comprising a given study
25 area and (c) use of a GSD characterizing interindividual variability in blood Pb (e.g.,
26 reflecting differences in behavior and biokinetics related to Pb).
27 The step involving modeling population-level exposures for the general urban case study is
28 somewhat simpler than that used for the two point source case studies in that demographic data
29 for a specific location is not considered. As discussed in Section 3.2.2, this avoids the need for
30 implementing a population-weighted probabilistic sampling procedure.
16 The Leggett model was included along with IEUBK in the pilot analyses. Findings for the model in the
pilot analyses and in subsequent performance analyses (Appendix J and Section 3.5) contributed to the decision to
use the IEUBK model as the primary model in this full-scale assessment (Section 3.2), and the Leggett model in the
sensitivity analysis (Section 3.5.2).
2-38
-------
1 2.4.4.2 Risk Characterization
2 The risk characterization step involves generating a distribution of IQ loss estimates for
3 the set of children simulated in the exposure assessment. Specifically, estimated blood Pb levels
4 (for the two blood Pb metrics) are combined with three blood Pb concentration-response
5 functions for IQ loss (see Section 4.1). Three differing concentration-response functions
6 (described in Section 4.1.1) have been selected to provide three different characterizations of
7 behavior at low exposures. The decision to use three different functions is in recognition of
8 uncertainty related to modeling this endpoint, particularly at lower exposure levels (e.g., blood
9 Pb levels < 5 |ig/dL). These three functions are all based on the lognormal concentration-
10 response function described in the Lanphear et al, (2005) pooled analysis of epidemiology
11 studies focusing on IQ loss in children. As these three functions were developed for each of the
12 blood Pb metrics included in the analysis, concurrent and lifetime average, six separate functions
13 were used in the analysis.
14 For each of the two point source case studies, we have produced two categories of risk
15 metrics:
16 • Population risk percentiles: The IQ loss associated with policy-relevant exposure
17 pathways for specific percentiles of the child population (e.g., the 50th, 90th, and 95th
18 percentile modeled child). This category of metric provides perspective on the
19 distribution of IQ loss resulting from policy-relevant exposure pathways, ranging
20 from the typical or average child (50th percentile, mean) to children experiencing
21 higher exposures (90th, 95th percentiles).
22 • Child frequency counts associated with specific risk percentiles: Number of children
23 associated with each of the population percentiles (e.g., the number of children
24 predicted to have risk levels at or above the 95th percentile). This risk metrics
25 provides a perspective on the number of children associated with various levels of IQ
26 loss for a particular case study.
27 For the general urban case study, because it is not location-specific, only the first type of
28 risk metric, population risk percentiles, was developed because this case study is not location-
29 specific. Child frequency counts are not applicable, since a specific location with associated
30 demographic data was not modeled.
31 Additional detail on the risk characterization is presented in Sections 4.1 and 4.2.
32 2.4.5 Variability Characterization
33 There are a variety of sources of variability associated with the results of this assessment
34 which are presented in terms of risk estimates for specific population percentiles:
2-39
-------
1 • variability in the concentration of lead in key media (e.g., diet, drinking water, ambient
2 air, indoor dust),
3 • variability in behaviors which effect Pb exposure (e.g., dust ingestion, soil ingestion),
4 • variability in physiological response to Pb exposure leading to variations in blood Pb
5 levels, and
6 • variability in the toxic response to Pb, resulting in differing degrees of IQ loss for the
7 same degree of Pb exposure.
8
9 A variety of methods have been used to incorporate, to a limited extent, the above
10 sources of variability in the risk assessment such that they are reflected in the results for the three
11 case studies.
12 • For the two point source case studies, use of spatial templates developed using a
13 geographic information system (GIS) to reflect the distribution of children across a
14 study area (in relation to ambient air Pb and related media concentrations) in
15 projecting population exposures.
16 • For the general urban case study, inclusion of two current conditions scenarios (mean
17 and 95th percentile air Pb scenarios) which together, provide a degree of coverage for
18 variation in ambient air Pb levels seen across urban areas in the U.S.
19 • Use of empirically derived GSDs reflecting interindividual variability in blood Pb
20 levels, to provide coverage for multiple sources of variability associated with Pb
21 exposure and biokinetics. Note, that the application of these GSDs provides the
22 primary means of reflecting interindividual variability in blood Pb levels in this
23 analysis. These GSDs also reflect uncertainty associated with measuring blood Pb
24 levels and characterizing population-level distributions of those levels.
25
26 There is significant uncertainty associated with reflecting the sources of variability
27 identified above in population-level exposure and risk. For example there is uncertainty
28 associated with the GSD selected to reflect interindividual variability in blood Pb levels for a
29 particular case study. This is considered in the uncertainty characterization (Section 2.4.6).
30 2.4.6 Uncertainty Characterization and Sensitivity Analysis
31 Several methods have been used to examine uncertainty in our modeling approach and its
32 potential impact on exposure and risk estimates (Section 4.3). These include: (a) development of
33 multiple sets of exposure and risk estimates for each case study and air quality scenario that
34 illustrate the combined impact of different key models and parameters on risk results and the
35 associated uncertainty, (b) evaluation of model performance (e.g., by comparison with empirical
2-40
-------
1 data) to provide confidence in individual modeling steps and (c) qualitative discussion of key
2 sources of uncertainty and their potential impact on exposure and risk estimates. Each of these
3 elements of the uncertainty characterization is briefly summarized below.
4 In addition, we have completed a sensitivity analysis, intended to characterize the
5 potential impact of individual modeling elements on risk results (see Section 2.4.63).
6 2.4.6.1 Performance Evaluations
7 Performance evaluation for the exposure assessment (Section 3.5) focused on evaluation
8 of estimates of Pb in ambient air, outdoor soil, and indoor dust (discussed in Section 3.5.1) and
9 estimates of Pb in blood (covered in Section 3.5.2). This evaluation focused on those estimates
10 based on modeling.
11 2.4.6.2 Generating Multiple Sets of Results
12 There are multiple models or inputs that could be implemented for each of the analytical
13 steps of the assessment. For those more highly influential analytical steps for which it is not
14 clear which model or input would generate "best estimate" results, we have implemented
15 multiple modeling approaches in the initial analyses presented in Chapters 3 and 4. Risk results
16 considered across these multiple modeling approaches provide perspective on the range of
17 potential risk, given key sources of uncertainty in the analysis. The multiple modeling
18 approaches for each case study were developed by the following stepwise strategy:
19 • Identification of those modeling elements believed to contribute significant
20 uncertainty to risk results,
21 • Identification of a set of plausible options for each of these key modeling elements
22 (e-g-, alternative models or input parameters), and
23 • Development of alternative modeling approaches by combining individual options
24 from the previous step.
25 Identification of the modeling elements believed to contribute significant uncertainty (step 1)
26 involved consideration of a number of factors including the results of the sensitivity analysis
27 completed for the pilot analysis, and comments provided by CASAC and the public on the pilot
28 analysis and analysis plan.
29 Because each of the case studies uses different modeling approaches for some of the
30 analytical steps (e.g., different indoor dust models are used for each case study, and these are
31 associated with differing uncertainty), the identify and size of the areas of uncertainty associated
32 with each case study differs. The specific modeling approaches for each case study and their
33 elements are presented in Figure 2-3. For the general urban case study, two different dust
34 models and two GSDs were used, compared to one model and GSD for these analytical steps in
2-41
-------
1 the two point source case studies. However, the same number of blood Pb metrics and IQ loss
2 functions are used for all three case studies.
3 Figure 2-3. Modeling approaches for case study analyses presented in chapters 3 and 4.
Case Study
General Urban
Case Study
Each Point
Source Case
Study
Elements of modeling approaches
Indoor dust
modeling
2 models:
(a) hybrid
mechanistic-
empirical
(b) statistical
(regression)
1 model:
statistical
(regression)
approach b
Blood Pb metric
2 metrics:
(a) concurrent
(b) lifetime
average
GSD
2 sets of GSDs,
representing:
(a) smaller
scale
(b) larger,
regional scale
1 set of GSDs
Concentration-
response function
3 functions:
(a) log-linear with
linearization,
(b) log-linear with
cutpoint, and
(c) two-piece linear
Number of
sets of results
2 * 2 * 2 * 3 = 24
1 * 2 * 1 * 3 =6
b Different models used for each point source case study.
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
The set of exposure and risk results generated for each case study and air quality scenario
using the alternative modeling approaches indicates to some extent the magnitude of uncertainty
surrounding the risk results. However, as discussed in Section 4.3, these sets of risk results do
not represent an uncertainty distribution, since confidence levels are not specified for each
modeling approach. In presenting the multiple sets of results generated for each case study, we
have selected the highest and lowest sets of risk results from those generated, and used them to
represent, respectively, upper and lower bounds on risk. The degree to which these actually
represent upper and lower bounds depends on the whether the various modeling approaches
evaluated in this analysis capture the largest sources of uncertainty. For example, if an important
source of uncertainty was excluded in designing the alternative modeling approaches for a given
case study, than the bounds represented by the set of risk results generated for that case study
might not be wide enough.
2.4.6.3 Sensitivity Analysis
Sensitivity analysis techniques were used to examine the uncertainty for individual
modeling elements and its impact on exposure and risk estimates. We used a "one element at a
time elasticity analysis" approach, running the full risk model with one of the selected modeling
elements adjusted to reflect an alternate input value or modeling choice. The results of that run
with the modified modeling element would then be compared to those for the "baseline risk" run
to determine the magnitude of the impact on risk results of selections for that one modeling
element.
2-42
-------
1 While the sensitivity analysis for the pilot was based on the primary Pb smelter case
2 study, for the full-scale analysis we have focused the sensitivity analysis on the general urban
3 case study. This reflects a desire to more fully understand the sensitivity of the modeling
4 approach used for the general urban case study to key sources of uncertainty, recognizing that the
5 results of the pilot sensitivity analysis are informative with regard to the point source case
6 studies. The sensitivity analysis completed for the full-scale analysis focused on those modeling
7 elements (including input datasets and modeling steps) believed to have a significant potential
8 for impacting exposure and risk results. Those modeling elements include oral uptake factor,
9 interindividual blood Pb variability GSD, biokinetic model, concentration-response function for
10 IQ loss. This type of sensitivity analysis indicates which of the modeling elements included in
11 the sensitivity analysis has the greatest impact on risk results, and can be used to guide future
12 efforts to refine the overall risk model.
13 2.4.6.4 Qualitative Discussion of Sources of Uncertainty
14 In addition to the quantitative analyses described above, we have also included a
15 qualitative discussion of key sources of uncertainty in the analysis. This includes sources not
16 explicitly included in any of the above quantitative analyses due to a lack of necessary data
17 (Section 4.3.1). This discussion also attempts to describe the nature of the impact of these
18 sources of uncertainty on risk results, e.g., would a particular source of uncertainty likely result
19 in an over- or underestimation of risk. To the extent possible, the likely magnitude in qualitative
20 terms of a particular source of uncertainty is also discussed.
2-43
-------
1 REFERENCES
2
3 Basha, M.R., Wei, W., Bakheet, S.A., Benitez, N., Sidiqi, H.K., Ge, Y.-W., Lahiri, O.K., Zawia, N.H. (2005) The
4 fetal basis of amyloidogenesis: exposure to lead and latent overexpression of amyloid precursor protein and
5 (3-amyloid in the aging brain. J Neurosci. 25: 823-829.
6 Henderson, R. (2006) Clean Air Scientific Advisory Committee (CASAC) Lead Review Panel's Consultation on
7 EPA's Draft Analysis Plan for Human Health and Ecological Risk Assessment for the Review of the Lead
8 National Ambient Air Quality Standards. Memorandum to Stephan Johnson, EPA Administrator, from Dr.
9 Rogene Henderson. July. Available at http://www.epa.gov/sab/pdf/casac-con-06-006.pdf
10 Henderson, R. (2007a) Clean Air Scientific Advisory Committee (CASAC) Review of the 1st Draft lead Staff Paper
11 and Draft Lead Exposure and Risk assessments. Memorandum to Stephan Johnson, EPA Administrator,
12 from Dr. Rogene Henderson. July. Available at http://www.epa.gov/sab/pdf/casac-07-003.pdf
13 Henderson, R. (2007b) Clean Air Scientific Advisory Committee (CASAC) Review of the 2nd Draft Lead Human
14 Exposure and Health Risk Assessments Document. Memorandum to Stephen Johnson, EPA Administrator,
15 from Dr. Rogene Henderson. September. Available at http://www.epa.gov/sab/pdf/casac-07-007.pdf
16 ICF International. (2006). Lead Human Exposure and Health Risk Assessments and Ecological Risk Assessment for
17 Selected Areas. Pilot Phase. Draft Technical Report. Prepared for the U.S. EPA's Office of Air Quality
18 Planning and Standards, Research Triangle Park, NC. December. Available at:
19 http://www.epa.gov/ttn/naaqs/standards/pb/sjb crtd.html
20 Lanphear, B. P.; Roghmann, K. J. (1997) Pathways of lead exposure in urban children. Environ. Res. 74: 67-73.
21 Lanphear, B. P.; Burgoon, D. A.; Rust, S. W.; Eberly, S.; Galke, W. (1998) Environmental exposures to lead and
22 urban children's blood lead levels. Environ. Res. 76: 120-130.
23 Lanphear, B.P., Hornung, R., Khoury, J., Yolton, K., Baghurst, P., Bellinger, D.C., Canfield, R.L., Dietrich, K.N.,
24 Bornschein, R., Greene, T., Rothenberg, S.J., Needleman, H.L., Schnaas, L., Wasserman, G., Graziano, J.,
25 and Robe, R. (2005) Low-level environmental Pb exposure and children's intellectual function: An
26 international pooled analysis. Environmental Health Perspectives. 113(7):894-899.
27 Missouri Department of Natural Resources (MDNR). (2007) 2007 Revision of the State Implementation Plan for the
28 Herculaneum Lead Nonattainment Area, as adopted by the Missouri Air Conservation Commission April
29 26,2007.
30 National Center for Health Statistics. (2005) Health, United States, 2005. With Chartbook on Trends in the Health
31 of Americans. Hyattsville, Maryland.
32 Rothenberg, S.J.; Rothenberg, J.C. (2005) Testing the dose-response specification in epidemiology: public health
33 and policy consequences for lead. Environ. Health Perspect. 113: 1190-1195.
34 Slob, W.; Moerbeek, M.; Rauniomaa, E.; Piersma, A. H. (2005) A statistical evaluation of toxicity study designs
3 5 for the estimation of the benchmark dose in continuous endpoints. Toxicol. Sci. 84: 167-185.
36 U.S. Environmental Protection Agency. (1977) Air Quality Criteria for Lead.: Office of Research and Development,
37 Washington, DC. EPA report no. EPA-600/8-77/017.
38 U.S. Environmental Protection Agency. (1978) National Primary and Secondary Ambient Air Quality Standards for
39 Lead. Federal Register 43(194): 46246-46263. Oct 5, 1978. Available at:
40 http://www.epa.gOv/ttn/naaqs/standards/pb/s pb pr fr.html
2-44
-------
1 U.S. Environmental Protection Agency. (1986) Air Quality Criteria for Lead. Office of Health and Environmental
2 Assessment, Environmental Criteria and Assessment Office Research Triangle Park, NC: EPA report no.
3 EPA-600/8-83/028cF. Available from: NTIS, Springfield, VA; PB87-142378.
4 U.S. Environmental Protection Agency. (1990) Review of the National Ambient Air Quality Standards for Lead:
5 Assessment of Scientific And Technical Information: OAQPS Staff Paper. Research Triangle Park, NC:
6 Office Of Air Quality Planning and Standards; report no. EPA-450/2-89/022. Available from: NTIS,
7 Springfield, VA; PB91-206185. http://www.epa.gov/ttn/naaqs/standards/pb/s_pb_pr_sp.html.
8 U.S. Environmental Protection Agency. (2003) Framework for Cumulative Risk Assessment. Risk Assessment
9 Forum, Washington, DC, EPA/630/P-02/00IF. May
10 U.S. Environmental Protection Agency. (2005) Review of the National Ambient Air Quality Standards for
11 Paniculate Matter: Policy Assessment of Scientific And Technical Information: OAQPS Staff Paper.
12 Research Triangle Park, NC: Office Of Air Quality Planning and Standards. EPA-452/R-05-005a.
13 December.
14 U.S. Environmental Protection Agency. (2006a). Draft Analysis Plan for Human Health and Ecological Risk
15 Assessment for the Review of the Pb National Ambient Air Quality Standards. Office of Air Quality
16 Planning and Standards, Research Triangle Park, NC. May 31, 2006. Available at:
17 http://www.epa.gov/ttn/naaqs/standards/pb/s_pb cr_pd.html
18 U.S. Environmental Protection Agency. (2006b). Review of the National Ambient Air Quality Standards for Lead:
19 Policy Assessment of Scientific and Technical Information. OAQPS Staff Paper - First Draft. Office of Air
20 Quality Planning and Standards. Research Triangle Park, NC. December 2006. Available at:
21 http://www.epa.gov/ttn/naaqs/standards/pb/s_pb cr sp.html
22 U.S. Environmental Protection Agency. (2006c) Lead soil trend analysis through May, 2006. Evaluation by
23 individual quadrant. Herculaneum lead smelter site, Herculaneum, Missouri. Prepared by TetraTech for
24 U.S. EPA, Region 7. Available on the web, at:
25 http://www.epa.gov/region7/cleanup/superfund/herculaneum pbtrend thru mav2006.pdf
2-45
-------
3 EXPOSURE ASSESSMENT
This chapter describes the methods and results of the exposure assessment and
performance evaluation for the initial analyses of the full-scale assessment. Additional analyses
are described in Chapter 5.
3.1 METHODS FOR ESTIMATING MEDIA CONCENTRATIONS
To estimate media Pb concentrations for the three case studies, we used a combination of
empirical data and fate and transport modeling, reflecting the different availability of Pb
measurements for the two point source case studies and the non-location-specific nature of the
general urban case study (Table 3-1). For all three case studies, media concentrations were
estimated for multiple air quality scenarios including a range of alternative NAAQS (see Section
2.3). However, outdoor dust/soil concentrations for each of the three case studies were
established for current conditions or the current NAAQS scenarios and those values were used
for the current NAAQS and each of the alternate NAAQS scenarios evaluated (see Sections 2.3.3
and 3.1.3). Further, as described in Section 2.4.1, media concentrations, once defined, are held
constant for the full exposure period simulated with the blood Pb modeling. For ambient air,
outdoor soil/dust and indoor dust, estimates of annual average concentration were used for this
purpose for all three case studies. Additionally, from the ambient air concentrations, annual
average inhalation exposure concentrations were estimated with consideration for daily activity
patterns by children and differences in outdoor (ambient) versus indoor air Pb levels (see Section
3.1.2).
3-1
-------
Table 3-1. Case study approaches for estimating media Pb concentrations.
Modeling Step
General Urban Case Study
Primary Pb Smelter Case
Study
Secondary Pb
Smelter Case Study
Spatial
template
Single generic study area (with
spatially uniform media
concentrations and population
density)
Combination of U.S. Census
blocks and block groups out
to a 10 km radius around the
facility (with media
concentrations and
demographics uniform within
blocks^lock groups)
U.S. Census blocks
out to a 10km radius
around the facility
(with media
concentrations and
demographics
uniform within
blocks)
Ambient air
concentrations
for current
conditions
and/or current
NAAQS
Monitor data (mean and high-
end urban monitors selected to
represent two current conditions
scenarios)
Dispersion modeling
Performance
evaluation
[Monitoring data used as basis
characterizing ambient air
levels]
Comparison to Pb-TSP monitor data from study area
Inhalation
exposure
concentrations
Estimates for all three case studies are based on ambient air concentrations and reflect the
application of location-specific adjustment factors that account for (a) the time spent by
children at different locations and at various activity levels and (b) differences between
indoor and outdoor ambient air Pb levels
Outdoor soil
concentrations
Nationally representative
residential soil value selected
from literature (same value used
for entire study area and for all
air quality scenarios)
Near facility (remediation
zone) relied on sampling
data, and regression model
used for outer portions of
study area
Multiple Pathways of
Exposure (MPE)
model used to predict
spatial distribution of
surface soil Pb levels,
then scaled using
empirical data from
surrogate location
Performance
evaluation
[Empirical data used in
characterizing soil levels]
[Estimates based on surrogate
data]
Modeled estimates
compared to surrogate
data for other
industrial (point
source) locations
Indoor dust
concentrations
Two approaches used: (a)
hybrid mechanistic
(compartmental) model
augmented with empirical data
developed specifically for this
analysis and (b) empirical (air-
only regression) model obtained
form the literature
Near facility (remediation
zone) relied on site-specific
regression model (based on
air) and pooled analysis
regression model (based on
air plus soil) for remainder of
study area
Statistical (air-only
regression) model
obtained from the
literature
Performance
evaluation
Subcomponents of hybrid
model evaluated, and case study
estimates compared to literature
estimates and national-scale
survey
Site-specific sampling data
used in deriving regression
model for remediation zone;
case study estimates
compared to literature
estimates and national-scale
survey
Case study estimates
compared to literature
estimates and
national-scale survey
3-2
-------
3.1.1 Ambient Air Concentrations
Different methods were used for estimating annual average ambient air concentrations for
the three case studies. For the primary and secondary Pb smelter case studies, air dispersion
modeling of Pb emissions was performed, while Pb-TSP measurement data from the years 2003-
2005 for urban areas with greater than one million residents were used for the general urban case
study.
3.1.1.1 General Urban Case Study
The general urban case study is not site-specific and has been designed to represent
general Pb exposures experienced by children residing in urban areas within the United States.
The case study involves spatially uniform Pb concentrations in ambient air, outdoor soil/dust and
indoor dust with a matching uniformly distributed child resident population. This is in contrast
to the point source case studies which each have a relatively large number of exposure zones to
track potentially significant spatial gradients in their concentrations of Pb in environmental
media and the spatial distribution of children.
Two current conditions scenarios are included in the general urban case study based on
Pb-TSP monitoring data from urban areas in the United States. Specifically, these two scenarios
include: (a) a central tendency current conditions scenario based on the mean maximum
quarterly average Pb measurement (0.14 |ig/m3) seen, in 2003-05 time period, at Pb-TSP
monitors in urban areas with greater than a million people, and (b) a high end current conditions
scenario based on the 95th percentile maximum quarterly average Pb measurement (0.87 |ig/m3)
obtained from the same urban Pb-TSP dataset. The data analysis associated with these values is
described in Appendix A, Section A.2.2.2.
In addition to these current conditions scenarios, the current NAAQS scenario (1.5
|ig/m3, as a maximum quarterly average) and four alternate NAAQS scenarios have also been
evaluated (see Section 2.3 for additional details on the NAAQS scenarios). In each of these
instances, the specific NAAQS of interest has been evaluated by assuming that air Pb
concentrations in the entire study area are reduced to meet that specific ambient air Pb level and
form.
Because the blood Pb modeling (Section 3.2) is based on annual average media
concentrations, the maximum quarterly and maximum monthly average values used in defining
the air quality scenarios were translated into equivalent annual average ambient air
concentrations for the blood Pb modeling (Section 2.4.1). This is accomplished using ratios
obtained from the 2003-2005 Pb-TSP monitoring dataset that relates maximum quarterly or
maximum monthly averages to associated annual average values for each monitor located in an
urban area of population more than a million (Appendix A, Section A.2.2). Ratios were selected
3-3
-------
for each of the air quality scenarios according to the averaging time (calendar quarter versus
month) and the percentile represented by the air quality scenario (e.g., mean versus high end for
current conditions). Consequently, the annual average Pb concentration for the central tendency
current conditions scenario was derived from the mean of monitor-specific maximum quarterly
average concentrations using the mean of the monitor-specific maximum quarterly-to-annual
average ratios, while the high end current conditions scenario was derived from the 95th
percentile of monitor-specific maximum quarterly average concentrations using the 95th
percentile of the monitor-specific maximum quarterly-to-annual average ratios. The mean of the
maximum quarterly-to-annual average ratios was also used to derive the annual average
concentration estimates for the alternative NAAQS air quality scenario with a quarterly
averaging time. The mean of the maximum monthly-to-annual average ratios was used to derive
the annual average concentration estimates for the alternative NAAQS scenarios with monthly
averaging times. The derivation of these ratios is described in Appendix A, Section A.2.2.
3.1.1.2 Primary Pb Smelter Case Study
The air quality scenarios included for the primary Pb smelter case study were the current
NAAQS and four alternate NAAQS (see Section 2.3.1). As the study area of the primary Pb
smelter is currently in nonattainment for the current NAAQS and emissions profiles from the
facility are being modified as additional controls are put in place, ambient air Pb concentrations
for the current NAAQS scenario were estimated using the model, emissions and source
parameters used in developing the 2007 proposed revision to the State Implementation Plan for
the area (MDNR, 2007a, 2007b). Annual average Pb concentration estimates for the current
NAAQS scenario were derived for each census block or block group from model outputs.
Additionally, for the purposes of developing the alternative NAAQS scenarios, hourly estimates
from the model were used to generate quarterly average and monthly average concentrations for
each census block or block group. A proportional rollback procedure was then used to adjust the
set of maximum monthly or quarterly averages to represent meeting a particular NAAQS
scenario. That is, the block or block group with the greatest exceedance was reduced to meet the
particular NAAQS and all locations were then reduced by that same fraction. After the
proportional rollback procedure had been applied to the set of location-specific monthly or
quarterly averages to meet a particular NAAQS, these adjusted quarterly or monthly average
values were then used to derive annual averages which, in turn, were used in the exposure
analysis.
The development of air Pb concentration estimates for this case study is described more
fully in Appendix D, Section D.2.
3-4
-------
3.1.1.3 Secondary Pb Smelter Case Study
Outdoor air concentration and deposition rates for Pb were estimated for the secondary
Pb smelter case study current conditions scenario using the AERMOD dispersion model1 and
source characterization and emissions information for the facility (See Appendix E, Section E.2
for details). The Pb emissions modeled reflected processes at the facility (e.g., stack emissions)
and fugitive dust emissions from materials storage and handling and roadway dust.
Annual average Pb concentrations for the current conditions scenario were derived for
each census block from model outputs. As with the primary Pb smelter case study, alternative
NAAQS scenarios were modeled using the proportional rollback procedure (see Section 3.1.1.2).
The development of air Pb concentration estimates for this case study is described more
fully in Appendix E, Section E.2.
3.1.2 Inhalation Exposure Concentrations
Inhalation exposure concentrations for Pb were estimated for young children, the
population of interest, from the estimated ambient air concentrations using age group- and
location-specific relationships for Pb developed using the exposure modeling component of
EPA's 1999 national-scale air toxics assessment (USEPA 2006a), one of the U.S. EPA's
National Air Toxics Assessment (NATA) activities. These relationships account for air
concentration differences indoors and outdoors and mobility or time spent in different locations
(e.g., outdoors at home, inside at home etc.) for the population of interest.
The exposure modeling component of the NATA national-scale assessment generated
inhalation exposure concentrations for sets of modeled children for each U.S. Census tract
(USEPA, 2006a). For the two point source case studies, we used the ratio of these NATA
national-scale assessment inhalation exposure concentrations to that assessment's corresponding
estimates of ambient air Pb concentration matched by U.S. Census tract to develop adjustment
factors that could be used to derive inhalation exposure concentrations from our estimates of
ambient air Pb concentration (see Attachment D, Section D.2.3, and Attachment E, Section E.2.3
for additional detail on this procedure). Although analyses of the ambient air concentrations
predicted in the NATA national-scale assessment indicate potential underpredictions, and there
are particular uncertainties in the assessment predictions at small scales, the relationship between
ambient air concentrations and exposure concentrations (i.e., the comparison used here) is not
expected to be affected by these factors. For the general urban case study, we used the median
adjustment factor from the full NATA national-scale assessment. The 0-4 years old age group is
1 AERMOD is the current preferred Gaussian plume dispersion model for assessing stationary sources
under the Clean Air Act (70FR(216): 68217-68261).
3-5
-------
the closest age group to the age group of interest for this assessment for which outputs are
available. For this age group, the adjustment factors or ratios between NATA national-scale
assessment Pb inhalation exposure concentrations and ambient air concentrations ranged from
0.37 to 0.46 for the Census tracts within the two point source case study areas. A value of 0.43
was used for the general urban case study (see Appendix C, Section C. 1.2). Use of these ratios
for the 0 to 4 year old age group to represent the 0 to 7 year old age group modeled in this
assessment contributes some uncertainty in the estimate of inhalation exposure concentrations.
3.1.3 Outdoor Surface Soil/Dust Concentrations
Pb concentrations in outdoor surface soil/dust were characterized for the current
conditions or current NAAQS scenarios for the three case studies using a combination of
modeling and empirical data. These estimates were also used for the alternate NAAQS
scenarios. That is, it was assumed that reductions in ambient air concentrations associated with
the alternate NAAQS scenarios did not have a significant impact on soil concentrations2.
3.1.3.1 General Urban Case Study
The outdoor surface soil concentration for the general urban case study was derived after
considering empirical data collected both at urban and other residential areas across the United
States (Appendix C, Section C.2). A single outdoor soil Pb level representative of general
residential yards in the U.S. was then selected for use with this case study. Specifically, a
nationally representative arithmetic mean of soil Pb levels in residential yards (198 |ig/g) was
obtained from the National Study of Lead and Allergen in Housing (NSLAH) (USEPA, 2000).
The NSLAH survey, which was conducted by the Department of Housing and Urban
Development (HUD) between 1998 and 1999, was intended to generate a nationally
representative sample of residential housing, including both private and public residences
constructed between 1940 and 1998 (but excluding institutional and group housing).
3.1.3.2 Primary Pb Smelter Case Study
In the primary Pb smelter case study, a different approach was used to estimate outdoor
surface soil/dust Pb concentrations near the facility than that used for more distant locations.
This difference is in recognition of the remediation activities that have included the removal of
contaminated soil at many of the residential yards closest to the facility, and replacement with
"clean" soil. Consequently, soil Pb concentrations are estimated using a combination of
2 As mentioned in Section 2.3.3, this also presumes that implementation methods for any of the alternate
NAAQS do not involve taking action to separately change soil Pb concentrations.
3-6
-------
measurement data for blocks within the remediation zone and statistically based predictions
beyond the remediation zone (see Appendix D, Section D.3 for more detail).
Surface soil/dust Pb levels near the facility (within the remediation zone) are based on the
most recent postremediation measurements available for a given block (obtained between 2001
and 2004 - see Appendix B Section, B.I.5.2). Preremediation soil levels are not used in
estimating soil levels within the remediation zone. Although analyses of sequential (over time)
postremediation measurements have indicated recontamination of remediated yards (USEPA,
2006b), the postremediation measurement based estimates were used for the current NAAQS
scenario, and for the alternate NAAQS scenarios. It is recognized that this approach may
produce an underestimate of Pb exposure for the ingestion of surface soil/dust for the current
NAAQS scenario, and perhaps some alternate NAAQS scenarios. However, an analysis of the
impact of a reduced NAAQS on the remediation zone temporal trend in surface soil/dust Pb
levels was not available to inform an alternate approach. Further, the impact of such a potential
bias is limited to the surface soil/dust pathway and does not affect the indoor dust pathway
because the indoor dust Pb concentrations for this region of the primary Pb smelter case study
were derived using an approach that did not rely on outdoor soil/dust Pb concentrations.
Characterization of soil levels for blocks and block groups beyond the remediation zone
are based on a regression model predicting soil Pb as a function of distance from the facility,
which was fitted to preremediation soil measurement data (available closer to the facility). The
use of preremediation soil data in deriving the regression equation reflects the fact that little
remediation has occurred in these more distant locations and consequently, spatial trends seen in
the preremediation soil levels are more likely to be representative for these outer portions of the
study area. The regression model used in these estimates has an r2 of 0.92 which suggests a good
fit and increases overall confidence in these statistical estimates. However, it should be noted
that this increased confidence holds for areas of interpolation (areas with sampling data used to
fit the model - out to about 2.3 km from the facility) more than for areas of extrapolation (areas
without sampling data - beyond 2.3 km from the facility).
The development of surface soil/dust Pb concentration estimates for this case study is
described more fully in Appendix D, Section D.3.
3.1.3.3 Secondary Pb Smelter Case Study
As noted in Section 2.2.2.2, soil sampling data for Pb were not identified for this case
study. Consequently, a hybrid mechanistic-empirical modeling approach was used to
characterize soil Pb levels for this case study, with fate and transport modeling employed to
derive a soil Pb concentration surface for the study area and sampling data obtained from a
surrogate secondary Pb smelter study area employed to adjust (calibrate) that surface.
3-7
-------
In lieu of historical estimates of emissions, the fate and transport modeling was
performed using current emissions estimates over a period consistent with the operating period
of the facility. As the emissions estimates used did not reflect levels of historical emissions,
generally believed to have been much higher than current estimates, the resultant Pb
concentrations did not reflect current conditions for the location, and, as expected, were lower
than concentrations reported for areas near other secondary smelters (see comparisons in
Appendix E, Section E.3). Accordingly, the soil concentration surface was scaled up based on a
set of factors (differing with distance from the facility) derived from empirical data for a
surrogate secondary Pb smelter location. Specifically, Pb concentrations across the entire
modeled surface were increased by a factor ranging froml to 13 (depending on distance from the
facility) to obtain surface soil/dust Pb concentrations consistent with those reported in the
literature for areas near other secondary Pb smelters. These estimates for the current conditions
scenario were also used for the alternate NAAQS scenarios. That is, it was assumed that
reductions in ambient air concentrations associated with the alternate NAAQS scenarios did not
have a significant impact on soil concentrations.
The development of surface soil/dust Pb concentration estimates for this case study is
described more fully in Appendix E, Section E.3.
3.1.4 Indoor Dust Concentrations
Pb in indoor dust can originate from a variety of sources including (a) outdoor soil which
is tracked into the house, (b) Pb in outdoor soil which is resuspended into the air and
subsequently transported indoors (c) Pb released directly into outdoor air through ongoing
anthropogenic activity (e.g., industrial point emissions) which is transported indoors and (d)
interior sources of Pb (e.g., paint, hobbies) (Adgate et al., 1998, Von Lindern, 2003). In the
exposure assessment conducted for the 1990 Staff Paper, indoor dust Pb concentrations were
predicted based on Pb concentrations in outdoor soil and ambient air (USEPA, 1989). This is
also the case for the default approach in the exposure component of the IEUBK model (USEPA,
1994a).
The importance of outdoor soil relative to outdoor air in influencing indoor dust Pb levels
appears to depend on the nature of the Pb sources involved. Investigations in urban areas and
near contaminated waste sites with elevated soil Pb levels without a currently active industrial
point source emitter of Pb have indicated a greater association of measurements of dust Pb
concentration with measurements of soil Pb concentration than with measurements of ambient
air concentration (e.g., Adgate, 1998 and Von Lindern, 2003). By contrast, investigations in
areas with currently operating large point sources of Pb (e.g., active Pb smelters) have implicated
ambient air Pb as an important source of Pb to indoor dust (Hilts, 2003). Contributions of
3-8
-------
ambient air Pb to indoor dust Pb levels have also been illustrated by a deposition study
conducted in New York City (Caravanos et al., 2006). Caravanos and others described Pb
deposition indoors resulting primarily from exterior environmental sources and not from interior
Pb sources. For additional discussion of the relationship between indoor dust Pb, outdoor
ambient air Pb and other related factors, refer to Appendix G, Section G.2
The prediction of indoor dust Pb based on Pb concentrations in outdoor ambient air and
other media (e.g., outdoor soil/dust Pb, indoor paint) can be conducted using empirical models,
mechanistic models, or by a combination of both techniques. Empirical models (typically
implemented as regression models) have the advantage of being able to specify a relationship
between indoor dust Pb and predictor variables (e.g., outdoor soil/dust, outdoor air) even when
these relationships are complex and uncertain. However, statistical models have the
disadvantage of requiring a significant amount of site-specific data for their derivation and not
being well suited to extrapolation to scenarios with conditions different from those underlying
their development. Mechanistic models, by contrast, can be developed in the absence of
extensive site-specific data and are not as limited in the types of scenarios to which they can be
applied as long as those scenarios are conceptually consistent with the scenario underlying their
development. However, the development of mechanistic models can be quite challenging and
subject to uncertainty if the system or process being predicted is complex and uncertain.
To date, efforts to predict indoor dust Pb have focused primarily on the development of
empirically-based regression models (EPA, 1989). Furthermore, most of these regression
models have been based on data associated with large industrial point sources (smelters) and
their impacts on surrounding residential populations. Little progress has been made in
developing comparable models for areas where ambient Pb levels are not so greatly influenced
by a large industrial point source, such as in more general urban residential areas, including
either regression models specific to these more general urban scenarios, or mechanistic models
which could be applied more readily to different exposure scenarios including urban residential
populations.
In this assessment, a combination of (a) statistical (regression) models obtained from the
literature, (b) statistical regression model developed specifically for individual case studies and
(c) mechanistic models (developed specifically for the general urban case study) was used in
predicting indoor dust Pb. This reflects the fact that varying amounts of site-specific data were
available across the three case studies for characterizing indoor dust Pb and related factors. In
addition, the absence of statistical regression models in the literature specifically focused on
urban residential locations necessitated the development of a hybrid mechanistic-empirical
model for the general urban case study (see below). The approaches used to predict indoor dust
3-9
-------
Pb for each of the case studies are presented below and described in greater detail in Appendix
G.
3.1.4.1 General Urban Case Study
Two models are used to estimate indoor dust Pb in the general urban case study: a hybrid
mechanistic-empirical model (the hybrid model), and a regression-based model. Application of
these models with the ambient air Pb concentrations for this case study produced two sets of
indoor dust Pb estimates for each air quality scenario (see Appendix C, Section C.3).
The hybrid model uses a mechanistic model to predict indoor dust Pb resulting from the
infiltration of outdoor air containing Pb into indoor residential air with subsequent contribution
to indoor dust Pb. This portion of indoor dust Pb derived with this model is subsequently
referred to in this document as the recent air or recent ambient air related component or
contribution to indoor dust Pb. To the extent that outdoor air Pb includes contributions from
resuspension of historically deposited Pb, it is represented here. Other contributions to indoor
dust Pb (e.g., tracking of outdoor soil/dust indoors, indoor paint flaking, etc.) are addressed using
an empirically based estimate derived from a national-scale dataset characterizing indoor dust Pb
loadings in U.S. residences (the U.S. Housing and Urban Development [HUD] National Survey
of Lead-based Paint in Housing - USEPA, 1995). When combined, the recent air component
(derived using the mechanistic model) and the other contributions component (derived using the
HUD dataset) provide an estimate of total indoor dust Pb for the general urban case study
(Appendix C, Exhibit C-7). Note that while the indoor dust Pb concentration contributed by the
hybrid model's recent air Pb component will vary, depending on the ambient air Pb level, the
concentration contributed from other sources will remain constant across the different air quality
scenarios. The two components of the hybrid model (the mechanistic and empirical) are
described in greater detail below.
The mechanistic model linking ambient air Pb to indoor dust Pb was developed by
obtaining the steady state solution to a dynamic mass balance equation that predicts Pb in both
indoor air and indoor floor dust as a function of outdoor air Pb. This dynamic mass balance
equation was developed specifically for this assessment. In recognition of the complexity of the
larger task of simulating contributions to indoor dust Pb from all sources, the mechanistic model
development activity was limited to the area considered most essential to the needs of this
assessment, i.e., the contribution to indoor dust Pb from recent ambient air. See Appendix G,
Section G.3.2 for additional detail on the derivation of the mechanistic ambient air-to-indoor dust
Pb model.
As noted above, contributions of other sources to Pb in indoor dust were addressed in the
hybrid model using empirical data (rather than trying to model them mechanistically). The HUD
3-10
-------
dataset, described above, that characterized the national distribution of residential indoor dust Pb
(USEPA, 1995) was selected as the basis for this other sources component of indoor dust Pb.
Specifically, a median value from the HUD dataset characterizing residential indoor dust Pb
loadings (USEPA, 1995) was identified, and the ambient air related component of this indoor
dust value estimated and subtracted from the HUD median value, leaving only the other sources
component. This provided a central tendency estimate of the component of indoor dust Pb
loadings associated with sources other than recent air Pb, for typical residences in the United
States (see Appendix G, Section G.3.3 for additional detail). As noted earlier, this single
estimate of other sources indoor dust Pb is used for modeling all air quality scenarios, with no
change associated with the ambient air Pb level being evaluated.
The hybrid model generates estimates of indoor dust Pb in terms of loading, rather than
concentration, while concentration is the form required by the blood Pb models used in this
analysis. Conversion from loading to concentration was accomplished using a log-log regression
equation derived from the HUD dataset described above. This dataset has matched data for
sampled residences for both indoor dust loadings from vacuum samples, and indoor dust
concentrations. Use of the HUD dataset based loading to concentration conversion required an
additional conversion between loading estimates based on wipe samples (the form used by the
hybrid model) and those based on vacuum samples (the form of the HUD data). This was
accomplished using an equation developed by EPA (USEPA, 1997). Details on both
conversions are provided in Appendix G, Sections G.3.4 and G.3.4.1.
The individual elements of the hybrid model, including both the mechanistic and
empirical components as well as the loading-to-concentration conversion equations are presented
in Table 3-2. The final hybrid equation (including all components) is presented last.
3-11
-------
Table 3-2. Hybrid model for indoor dust Pb in general urban case study.
Model component
Hybrid mechanistic equation for "recent" air
Hybrid empirical component for other sources
Converting wipe loadings to vacuum loadings
Converting vacuum loadings to concentration
Final combined hybrid equation
Formula
FLOOR LOADING (PbWipe) = 104.2 * PbAIR
Where, PbAIR is Pb in outdoor ambient air, and FLOOR
in terms of wipe sample loading.
LOADING is
1.15ug/m2
PbVAC = 0.185 * PbWIPE °921
Ln(PbCONC) = 4.92 + 0.52 * In(PbVAC)
PbDust = EXP [4.92 + 0.52 * Ln (0.185 * (104.2 * PbAir H
M.15)0931)]
Because the hybrid model has not been subject to extensive review and application
outside of this analysis and given the influence of indoor dust Pb on Pb exposure and risk, dust
Pb concentrations for the general urban case study are also estimated using an additional dust Pb
model (see Appendix G, Section G.3.5). The use of two models is intended to inform our
characterization of model uncertainty in this key portion of the analysis. Specifically, we have
included the air-only regression model (EPA, 1989) as a second, parallel approach in predicting
indoor dust Pb. That model estimates indoor dust Pb based on (a) outdoor ambient air Pb
(multiplied by an air-related factor) and (b) an intercept which captures other impacts besides air
(e.g., indoor paint). The air factor used in this equation is expected to capture longer-term
impacts of outdoor air Pb on indoor dust, including the indirect effect of air Pb on outdoor
soil/dust Pb with subsequent impacts of that outdoor soil/dust Pb on indoor dust Pb through other
mechanisms (EPA, 1989). The air-only regression model is presented below:
PbDUST(mg/kg or ppm) = 60 + 844 * PbAIR(ug/m3)
3.1.4.2 Primary Pb Smelter Case Study
We used different regression models for predicting Pb concentrations in indoor dust in
areas near the primary Pb smelter facility where soil has been remediated and more distant areas.
For the remediation zone near the facility, a regression equation was developed using dust Pb
measurements which had been collected from some of the houses within this area. These data,
while adequate for development of a site-specific regression model, did not have sufficient
spatial coverage to be used alone to represent indoor dust Pb levels for that portion of the study
area. For the remainder of the study area, we employed a regression equation developed for the
3-12
-------
last review (USEPA, 1989). Because of the presumed impact of the remediation activity on dust
Pb, the site-specific dust Pb model developed for the remediation zone was not considered
appropriate for use in areas beyond that area.
The dataset used to develop the model for the remediation zone was based on indoor dust
samples collected in 17 houses within the remediation zone. Independent variables included in
the analysis were: (a) estimated annual average Pb concentrations in air at census block centroids
located within 200 meters of each of the 17 houses, (b) road dust Pb measurements for locations
within 300 meters of each house and (c) postremediation residential soil Pb measurements for the
yard of each house. Preremediation soil Pb concentrations were not included in the regression
analysis since they were not expected to represent current conditions at the site. Multiple
samples for each medium associated with a specific house within the dataset (e.g., reflecting
multiple samples collected over time) were averaged to produce a "temporally averaged" value.
A number of regression models were evaluated, (see Appendix G, Section G.4), and the "H5"
model was ultimately selected based on goodness of fit and other considerations. This model
relates the natural log of indoor house dust to the natural log of ambient air Pb (r2=0.625):
ln(house dust, mg/kg or ppm) = 7.7892 + 0.7200*ln(air Pb, ug/m3)
In applying the H5 model to the remediation zone portion of the study area, a constraint
was applied in recognition of the our application of this regressions to scenarios with much lower
air Pb than that for the dataset on which it is based, and findings for similar towns of residual
levels of Pb in house dust that reflect historical contributions (e.g., von Lindern et al., 2003;
Hilts, 2003). A floor value for the predicted dust Pb for alternative NAAQS scenarios was
assigned to any model predictions falling below it. This value was 60 ppm, which is the
intercept for the air only regression model derived at the time of the last review, see previous
section. While this value may be an underestimate of the historical component of Pb in house
dust of existing houses or houses existing prior to remediation, its use is intended to balance the
need for a floor for such houses and the potential for the population to also have newer houses.
The use of and need for this floor recognizes the uncertainty associated with this model.
Additionally, the regression model used here does not lend itself to partitioning the recent air
related Pb from other contributions (discussed in Section 2.4.3); accordingly, this partitioning is
not done for the primary Pb smelter case study.
Several points regarding the other variables considered for the remediation zone
regression are noted here. For example, road dust Pb concentration was not found to have
significant predictive power for indoor dust Pb. This may reflect the fact that the road dust Pb
dataset does not provide significant coverage for homes located near to the truck haul routes.
Additionally, yard soil Pb concentration was found to be slightly, and statistically significantly,
negatively correlated with indoor dust Pb levels. This counterintuitive finding may be a result of
3-13
-------
the existence within the remediation zone of a patchwork of remediated yards, such that the
remediation activity may have interfered with any correlation between yard soil Pb levels,
ambient air Pb levels and indoor dust Pb levels that might have existed previously. The resulting
slight negative correlation of dust Pb levels with soil Pb levels led us to exclude soil Pb in
predicting indoor dust Pb, leading to the selection of the H5 model, which only considers
ambient air Pb in predicting indoor dust Pb. The y-intercept for the selected model may reflect a
number of factors not correlated with ambient air or distance from the facility, such as a general
level of soil Pb contamination in the area and indoor Pb paint.
For areas beyond the remediation zone, we used a regression equation developed during
the last review from data collected at a number of operational primary Pb smelters, including this
case study location (USEPA, 1989, Appendix B). This model, the "AGG" or "aggregate" model,
predicts indoor dust Pb concentration from both outdoor soil and ambient air Pb concentrations.
We have selected the AGG model for the nonremediation portion of the primary Pb smelter case
study area since this area has not been subjected to extensive remediation and is therefore likely
to resemble the locations included in the pooled dataset used in deriving this model in terms of
relationships among air, outdoor surface soil/dust and indoor dust. The AGG air and soil
regression model (USEPA, 1989), selected for areas beyond the remediation zone is the
following:
House dust (mg/kg or ppm) = 31.3 + 638*air Pb (ug/m3) + 0.364*soil Pb (mg/kg)
3.1.4.3 Secondary Pb Smelter Case Study
A version of the empirical regression model used for the primary Pb smelter (USEPA,
1989) was also used for the secondary Pb smelter case study. In this case study, an air-only
version of the model (USEPA, 1989) was employed reflecting the reduced overall confidence
associated with soil characterization at this case study (see Section 3.1.3.3). The AGG model for
estimating indoor dust (USEPA, 1989) was derived in two forms including an air-only model
that based indoor dust concentrations on outdoor ambient air Pb (without explicitly considering
outdoor soil Pb levels) and an air+soil model which based estimates on both outdoor soil and
ambient air Pb data. It is important to note, however, that the air-only model implicitly reflects
some consideration for the air-to-soil-to-indoor dust mechanism in the air signal. Specifically,
the larger air factor for the air-only model relative to the air+plus dust model's air factor, reflects
contribution of air Pb both directly to dust through penetration indoors and subsequent
deposition to surfaces and indirectly to dust through deposition to outdoor soil which impacts
indoor dust (USEPA, 1989). This air-only regression model (USEPA, 1989) is as follows:
House dust (mg/kg or ppm) = 60 + 844*air Pb (ug/m3)
3-14
-------
The model used for this case study was based on a number of studies focusing mainly on
primary Pb smelters (a number of primary Pb smelters were operational at the time of model
development). This may introduce uncertainty into indoor dust Pb predictions generated for this
case study to the extent that factors related to the dependency of indoor dust Pb on ambient air
Pb, such as particle size profiles and the nature of the airborne Pb compounds, differ for primary
versus secondary Pb smelters.
3.2 METHODS FOR ESTIMATING BLOOD PB LEVELS
This section presents the methodology used to estimate blood Pb levels in the child study
populations. The section begins with an overview of the primary biokinetic model used in this
full-scale risk analysis, the Integrated Exposure and Uptake Biokinetic model (USEPA, 1994a).
Findings associated with the use of both the IEUBK and Leggett models in the pilot analysis, in
addition to subsequent performance evaluation analyses (see Appendix J) contributed to the
decision to use IEUBK as the primary blood Pb model and reserve the Leggett model for use in
the sensitivity analyses. In addition, we have considered an empirical slope model (the Lanphear
model) in conducting our performance evaluation (Appendix J, Section J.2).
Following the overview of the IEUBK biokinetic model, the probabilistic approach used
to generate population-level distributions of blood Pb levels for each study population is
described. The section ends with a discussion of the GSDs used within each case study to reflect
interindividual variability in behavior related to Pb exposure and Pb biokinetics (a key
component in modeling population-level blood Pb).
3.2.1 Blood Pb Modeling
Blood Pb models are used in the assessment in order to (a) apportion exposure, the metric
for which is blood Pb, between policy-relevant Pb exposures and policy-relevant background,
and (b) estimate potential changes in blood Pb level distributions that would result from alternate
ambient air Pb levels.
As discussed in Section 4.4.1 of the CD, there are two broad categories of blood Pb
models available to support exposure and risk assessment:
• Statistical (regression) models, which attempt to apportion variance in measured blood Pb
levels for a study population to a range of determinants or control variables (e.g., surface
dust Pb concentrations, air Pb concentrations). The development of these models
requires paired predictor-outcome data which restricts these empirical models to the
domain of their observations (i.e., to applications involving the study population(s) and
exposure scenarios used in their derivation or at least to scenarios very similar to the
original study conditions).
3-15
-------
• Mechanistic models, which attempt to model the process of transfer of Pb from the
environment to human tissues. While these models are considerably more complex
compared with the regression models (in terms of both the number of variables and their
computational structure), by incorporating variables that vary temporally and spatially, or
across individuals or populations, mechanistic models can be extrapolated to a wide
range of scenarios, including those outside of the original populations and exposure
scenarios used to develop/parameterize the models.
Given concerns over applying regression models to populations and exposure scenarios
other than those used in their derivation, and consistent with recommendations from CASAC
(see Section 1.4), we have relied primarily on mechanistic models in conducting the exposure
analysis for this assessment. Additionally, a regression model developed by Lanphear et al.
(1998) and described in Appendix H, Section H.2.3.was included in the blood Pb modeling
performance evaluation (Appendix J, Section J.3.2). The CD (Section 4.4) describes three
mechanistic (biokinetic) models developed over the past several decades including IEUBK for
modeling child Pb exposure (CD, Section 4.4.5 and Appendix H, Section H.2.1) and two models
designed to simulate Pb biokinetics and blood Pb levels from birth through adulthood, the
Leggett model (CD, Section 4.4.6) and the model developed by O'Flaherty (CD, Sections 4.4.7).
The Leggett and O'Flaherty models simulate smaller time steps and consequently demonstrate
blood Pb responses on much shorter time scales than the IEUBK model, the outputs for which
are described as indicative of quasi steady-state conditions (CD, Sections 4.4.5-4.4.7). All three
models have the potential for application in Pb risk assessment and have been evaluated to
varying degrees using empirical datasets (CD, Section 8.3.4). Based on results of the pilot
analysis, subsequent performance evaluations, and advice from CASAC, we selected the IEUBK
model as the primary blood Pb model for the full-scale analysis. The Leggett model has been
included as part of the sensitivity analysis intended to assess the potential impact of uncertainty
in blood Pb modeling on the overall analysis (See Section 4.3.2). A brief overview of the
IEUBK and Leggett models is presented below.
3.2.1.1 Primary Analysis
The IEUBK model was selected for use in generating the primary set of exposure and
risk results. IEUBK is a multicompartment pharmacokinetics model for children 0 through 84
months of age (the first 7 years of life), which predicts average quasi-steady state blood Pb
concentrations corresponding to daily average exposures, averaged over periods of a year of
more. The exposure module provides average daily intakes of Pb (averaged over a 1 year time
increment) for inhalation (air, including consideration for both outdoor and indoor) and ingestion
(soil, indoor dust, diet and water) (Section 4.4.5.1 of the CD). The model is intended to be
applied to groups of children experiencing similar levels of Pb exposure and to generate a central
3-16
-------
tendency blood Pb estimate for that group (USEPA, 1994a). In applications of IEUBK,
interindividual variability in biokinetics and behavior (e.g., varying rates of dietary Pb ingestion)
of the study population is typically characterized through the incorporation of a GSD which,
together with the lEUBK-generated blood Pb level, provides a distribution of blood Pb levels for
a group of modeled children.
For each exposure zone in each case study air quality scenario, estimates of the
concurrent and lifetime average blood Pb level are derived from the outputs of the IEUBK model
as described in Appendix H. Briefly, the concurrent metric is derived as the average over ages
73 to 84 months (approximately 6 to 7 years of age)3, and the lifetime average metric is the
average of blood Pb levels between ages 6 and 84 months.
Additional detail on the IEUBK model is described in Appendix H (Section H.2.1) and in
Section 4.4.5 of the CD. Application of the IEUBK model is described in Appendix H, Section
H.3.1, and input parameter values, and their basis, are described in Appendix H, Section H.4.
3.2.1.2 Sensitivity Analysis
The Leggett model has been included as part of the sensitivity analysis (Appendix L) and
in model performance evaluations (Appendix J), but has not been used in generating primary
exposure and risk results for this assessment. Originally developed from a model designed to
simulate radiation doses for bone-seeking radionuclides, the Leggett biokinetic model has a
temporal resolution of one day and can model exposure from infancy through adulthood (CD,
Section 4.4.6). The daily resolution in Leggett allows more comprehensive treatment of the
temporal pattern of exposure and its shorter-term impact on blood Pb levels than IEUBK,
although for this assessment, which focuses on longer-term trends in Pb exposure, this
functionality is less relevant. The Leggett model does not include a detailed pathway-level
exposure module as does IEUBK. Rather the Leggett model takes total ingestion and inhalation
exposure estimates as inputs. However, it is possible to link the Leggett model to a more
detailed pathway-level exposure model, thereby allowing a more detailed treatment of Pb
exposure pathways and their impact on blood Pb. The use of this type of external exposure
model including pathway-specific modeling of exposure levels was implemented for the pilot
analysis. As with IEUBK, Leggett can be used to derive central tendency blood Pb levels for
groups of similarly exposed children. The same GSD used for IEUBK is then used to produce
3 As described in Appendix I (Section 1.2.1), the concurrent metric is calculated by averaging blood Pb
estimates generated for ages 75 and 81 months with these estimates representing the first and second halves of the
7th year of life, respectively. Use of the 7th year of life is consistent with or similar to the average age of the IQ
testing in the Lanphear et al. (2005) study on which the concentration-response function is based (Sections 2.1.5 and
4.1).
3-17
-------
estimates of the distribution of blood Pb levels within study populations. For additional details
on the Leggett model see Section 4.4.6 of the CD and Appendix H.
3.2.2 Exposure Pathway Apportionment and Probabilistic Population Modeling
This section describes the method used to estimate contributions to blood Pb and IQ from
exposure pathways of interest (see Section 2.4.3), and the method used for probabilistic
population modeling. Both are applied to the central tendency estimates of blood Pb developed
for the two blood Pb metrics (concurrent and lifetime average) from the IEUBK model outputs.
To the extent feasible with the modeling tools and assessment design, estimates of the
contribution to blood Pb (central tendency estimate) are developed for policy-relevant
background versus policy-relevant exposures (Section 2.4.3). This is done by considering blood
Pb model estimates for an exposure zone derived using only the pathways of interest (Appendix
I, Section I.I). As discussed in Section 2.4.3, there are limitations on the resolution to which
policy-relevant exposures can be distinguished which results in some simplifying assumptions.
We developed estimates of contribution to blood Pb estimates (and IQ estimates) for the
following pathways (or pathway combinations):
• Inhalation of ambient air Pb (i.e., "recent air" Pb): This is derived using the blood Pb
estimate resulting from Pb exposure limited to the inhalation pathway.
• Ingestion of "recent air" indoor dust Pb: This is derived using the blood Pb estimate
resulting from Pb exposure limited to ingestion of the Pb in indoor dust that is
predicted to be associated with ambient air concentrations (i.e., via the air
concentration coefficient in the regression-based dust models or via the mechanistic
component of the hybrid blood Pb model (Section 3.1.4). For the primary Pb smelter
case study, estimates for this pathway are not separated from estimates for the pathway
described in the subsequent bullet due to uncertainty regarding this categorization with
the model used for this case study (Section 3.1.4.2).
• Ingestion of "other" indoor dust Pb: This is derived using the blood Pb estimate
resulting from Pb exposure limited to ingestion of the Pb in indoor dust that is not
predicted to be associated with ambient air concentrations (i.e., that predicted by the
intercept in the dust models plus that predicted by the outdoor soil concentration
coefficient, for models that include one (Section 3.1.4)). This is interpreted to
represent indoor paint, outdoor soil/dust, and additional sources of Pb to indoor dust
including historical air (see Section 2.4.3). As the intercept in regression dust models
will be inclusive of error associated with the model coefficients, this category also
includes some representation of dust Pb associated with current ambient air
concentrations (described in previous bullet). For the primary Pb smelter case study,
estimates for this pathway are not separated from estimates for the pathway described
above due to uncertainty regarding this categorization with the model used for this case
study (Section 3.1.4.2).
3-18
-------
• Ingestion of outdoor soil/dust Pb: This is derived using the blood Pb estimate resulting
from Pb exposure limited to ingestion of outdoor soil/dust Pb.
• Ingestion of drinking water Pb: This is derived using the blood Pb estimate resulting
from Pb exposure limited to ingestion of drinking water Pb.
• Ingestion of dietary Pb: This is derived using the blood Pb estimate resulting from Pb
exposure limited to ingestion of dietary Pb.
The goal of this probabilistic exposure modeling is to generate population-level
distributions of blood Pb levels that allow (a) specific percentiles of exposure (e.g., 50th, 90th,
and 95th) within a study population to be identified. These are presented along with the
differentiation by exposure pathway (e.g., policy-relevant background versus policy-relevant
exposures, with the latter further differentiated as to ambient air inhalation, indoor dust ingestion
and outdoor surface soil/dust ingestion, as indicated above). Therefore, for example, we may
have an estimate of exposure for the 95th percentile child in the primary Pb smelter case study,
with that blood Pb level differentiated as to the fraction coming from (a) diet and drinking water,
(b) recent air Pb, including ambient air inhalation and ingestion of the recent air component of
indoor dust Pb, with this fraction potentially including resuspended, previously deposited Pb (see
Section 2.4.3), and (c) other sources including outdoor soil/dust and historical air Pb, as well as
indoor paint. The effort to differentiate the recent air and other sources of Pb to indoor dust is
subject to different degrees of uncertainty for the various case studies, reflecting the different
approaches used in modeling indoor dust (see Sections 2.4.3 and 3.1.4). Furthermore, it is noted
that given the various limitations of our modeling tools (Sections 2.4.3), blood Pb levels
associated with air-related exposure pathways and current levels of Pb emitted to the air
(including via resuspension) are likely to fall between the estimates for "recent air" and those for
"recent" plus "past air".
Probabilistic exposure modeling differs for the two point source case studies and the
general urban case study. Because the two point source case studies are location-specific,
demographic data for each study area are used to evaluate the interaction between child
populations in those study areas and the distribution of Pb concentrations in each media (e.g.,
outdoor ambient air, outdoor soil/dust and indoor dust). By contrast, because the general urban
case study is not location-specific, probabilistic exposure modeling does not involve location-
specific demographic data and instead, is based on the assumption of a uniformly distributed
child receptor population contacting media with uniformly distributed Pb.
3-19
-------
3.2.2.1 General Urban Case Study
The approach used to generate exposure distributions for the general urban case study is
simpler than that used for the two point source case studies (Section 3.2.2.2), due to the use of a
uniformly distributed population and spatially uniform media concentrations of Pb in the design
of this case study. This design negates the need for population-weighted sampling in generating
the exposure distribution. Instead, the central tendency blood Pb level generated for a specific
air quality scenario using IEUBK, is combined deterministically with the GSD reflecting
interindividual variability in Pb exposure and biokinetics to produce a population-level
distribution of blood Pb levels.4 Specific percentile exposure levels are then identified from that
distribution, with these levels interpreted as representing modeled individual children for the
general urban case study (i.e., estimates conceptually equivalent to those generated for the two
point source case studies). However, because the analysis is not location-specific, population
incidence estimates are not generated.
The study area for the general urban case study can be considered a single exposure zone
with a single lEUBK-derived central tendency blood Pb level that is combined with the GSD to
produce the population-level exposure distribution for that study area. As is the case with
individual exposure zones modeled for the point source case studies, the entire child population
in the single exposure zone modeled here is given the same pathway apportionment. As with the
point source case studies, this approach introduces uncertainty into the analysis, especially for
simulated individuals with high-end blood Pb levels, since pathway apportionment would likely
not be the same for all individuals in an exposure zone, even if all were exposed to the same Pb
concentrations in each medium.
3.2.2.2 Point Source Case Studies
Probabilistic exposure modeling for the two point source case studies relied on
information in three areas as summarized below:
• Central tendency blood Pb levels for each exposure zone: Outputs from the biokinetic
blood Pb modeling (considered to be central tendency estimates) are used to produce
central tendency estimates of concurrent and lifetime average blood Pb levels for each
exposure zone (i.e., census block or block group) in each case study area.
4 Note that as discussed in Section 3.2.3.1, the range of GSDs used in the general urban case study to reflect
interindividual variability in lead exposure and biokinetics has been selected to include some degree of coverage for
small-scale variation in media concentrations. Therefore, while the overall scenario modeled in the general urban
case study is based on the assumption of generally uniform media concentrations (e.g., an ambient air concentration
level which is generally constant across the study area), the use of larger GSDs provides some coverage for
residence-to-residence variation in these media concentrations.
3-20
-------
• Demographics (child distribution within study areas): The distribution of children (<7
yrs old) within each case study area is used to insure that the generation of population-
level blood Pb level distributions for each case study reflects where children are located.
• GSD reflecting interindividual variability in blood Pb levels: A GSD is used to reflect
interindividual variability for blood Pb levels in groups of similarly exposed children
(i.e., within each exposure zone of a case study area). The GSD is combined with the
central tendency blood Pb level estimates to generate a distribution of blood Pb levels for
the group of children located in each exposure zone.
The stepwise procedure used to generate population-level blood Pb distributions for each
of the point source case studies is illustrated in Figure 3-1, with the information described in the
bullets above recognized as input data.
Several points related to implementation of this procedure are noted. For this assessment,
50,000 simulated individuals were generated for each point source case study in order to insure
that the population-level blood Pb distributions generated met target stability goals (see
Appendix M, Section M.2.2).5 This simulation count represents a higher total child count than
actually is associated with either study area. Using a higher number of simulated individuals was
necessary, however, to generate blood Pb distributions with "stable" higher-end exposure
estimates. If simulations matching the actual population count at each case study had been
conducted, the distributions that would have resulted would have been "unstable" at higher
percentiles. It is important to note, however, that child population count estimates for individual
percentiles (Section 3.4) have been scaled to reflect the actual child count associated with each
study area.
5 The analysis of simulation stability focused on the general urban case study and demonstrated that even
the highest population percentiles generated for this analysis (95th percentile) were relatively stable with coefficients
of variation of less than 1% for total IQ loss estimates (see Appendix M, Section M.2.2).
3-21
-------
Figure 3-1. Procedure for generating population blood Pb distributions for point source
case studies.
Input distributions and datasets used in the Monte Carlo-based population-exposure modeling:
• Central-tendency blood Pb level dataset: Blood Pb modeling (described in Section 3.2.1)
generates a central-tendency blood Pb level for each exposure zone (e.g., block, or blockgroup)
within each study area (results dimensioned on blood metric and blood Pb model).
• Demographic dataset: The child count (children <7 yrs of age) for each exposure zone is used
to support population-weighted sampling of exposure zones.
• Adjustment factor distribution representing inter-individual variability in behavior and
biokinetics related to Pb exposure: These adjustment factor distributions are derived by
centering the GSDs (representing inter-individual variability in PbB levels) on a median value of
1.0. This results in distributions that can be used to obtain adjustment factors reflecting the
behavior and biokinetics (related to Pb exposure) for a single simulated child.
Monte Carlo-Based Population Blood Pb Level Modeling Procedure:
Step 1 - select a central-tendency blood Pb level for a specific
exposure zone based on population-weighted random sampling:
The central-tendency blood Pb level dataset is combined with the
demographic dataset to conduct population-weighted sampling of a
central-tendency blood Pb level for a specific exposure zone (i.e.,
sampling proportional to child count within each exposure zone is
used to select a specific zone and the central-tendency blood Pb level
for that zone is chosen as the output form Step 1).
Step 2 - select an adjustment factor reflecting inter-individual
variability in behavior and biokinetics related to Pb exposure for
a single simulated child: A value is sampled randomly from the
adjustment factor distribution described above. This single value
represents the behavior and biokinetics (related to Pb exposure) for a
single simulated child.
Step 3 - generate a blood Pb level for a single simulated
individual: The central-tendency blood Pb level selected in Step 1 is
multiplied by the adjustment factor in Step 2 to produce an "adjusted"
blood Pb level for a single simulated child within the exposure zone
selected in Step 1 (this value reflects both the central-tendency PbB
for that exposure zone as well as consideration for inter-individual
variability in PbB levels).
Step 4 - place simulated child blood Pb level (output of Step 3) in
pool of modeled blood Pb levels for that study area: This pool of
simulated child blood Pb levels represents the distribution of blood Pb
levels across the study area and reflects (a) the demographic
distribution of children across that study area (and their relation to Pb
levels in contact media) and (b) inter-individual variability in behavior
and biokinetics related to Pb exposure.
Repeat procedure
50,000 times to
generate set of 50.000
simulated individuals
(Note, this set can be
weighted down to
reflect the actual
population of children
(0-7 years) within
each study area in
generating
population-count
related risk metrics).
3-22
-------
An additional point of clarification relates to the process used in differentiating specific
percentile total blood Pb levels into pathway-specific fractions for the two point source case
studies. All simulated individuals associated with a given exposure zone (i.e., U.S. Census block
or block group), were assigned the same pathway-specific apportionment, which was derived
from the blood Pb modeling performed for the exposure zone. The set of simulated individuals
with a range of total blood Pb levels, produced subsequently via application of the GSD, are
assumed to all have the same pathway-specific apportionment of those blood Pb levels (i.e., the
same apportionment as that generated for the central tendency blood Pb level modeled for that
exposure zone). In reality, it is likely that pathway apportionment would vary across children
with different blood Pb levels located in the same exposure zone (e.g., the contribution of indoor
dust exposure to total blood Pb might differ for kids living near each other who demonstrate
different total blood Pb levels). The modeling approach used, however, does not provide
pathway apportionment within an exposure zone, only across exposure zones (i.e., each exposure
zone has a different pattern of pathway apportionment for its simulated children).
The modeling approach presented in Figure 3-1 and described above, generates a
population-level distribution of total blood Pb levels with pathway apportionment as described
above. These distributions are used to generate several types of exposure metrics including:
• Population-weighted exposure percentiles: total blood Pb levels (with pathway
apportionment) for simulated individuals representing specific points along the population
blood Pb distribution (e.g., 50, 90, and 95th percentile).
• Incidence counts: number of children within a given study area projected to experience a
specific degree of Pb exposure (total blood Pb level).
3.2.3 GSD for Population Blood Pb Modeling Procedure
A key aspect of the population-level blood Pb modeling for all three case studies is the
application of the GSD reflecting interindividual variability in blood Pb levels. This GSD
reflects a number of factors which operate together to produce interindividual variability in blood
Pb levels, including: (a) biokinetic variability (differences in the uptake, distribution or clearance
of Pb), (b) differences in behavior related to Pb exposure (e.g., varying hand-to-mouth activity,
tap water ingestion rates, and time spent playing indoors) and (c) differences in environmental
Pb exposure concentrations (e.g., spatial gradients in ambient Pb levels of a resolution beyond
that simulated in each case study, differences in cleaning/vacuuming rates and air exchange
rates).
GSDs will tend to be larger for more diverse populations and/or larger study areas,
reflecting the potential for greater variability in the factors listed above. Specifically, more
diverse populations will tend to demonstrate greater diversity in behavioral and biokinetic factors
3-23
-------
related to Pb exposure, thereby producing greater variation in blood Pb levels. Larger study area
will tend to produce greater GSDs due both to more diverse populations, with their greater
behavioral and biokinetic variability, as well as greater variation in Pb media concentrations
across the study area.
The dramatic reduction in blood Pb levels among children in the United States that was
observed between the NHANES surveys of the late 1970s and those of the 1980s and later was
accompanied by an increase in GSD (see Appendix H, Exhibit H-7). Possible reasons for this
include the likelihood that, as blood Pb levels decrease, a wider variety of exposure pathways
begin to play a role in determining overall blood Pb levels (at higher blood Pb levels, it likely
that one, or a few related pathways dominate exposure). As more pathways come into play, the
potential for interindividual variability in behavior and biokinetics related to these pathways
increase, thereby producing greater variability in blood Pb levels (see Section 4.2.2 of EPA,
1994b). Another possible explanation for the increase in GSDs is that, while overall Pb exposure
levels have decreased, some fraction of children nationwide continue to be exposed to Pb paint
and Pb in drinking water (associated with Pb solder used in older plumbing). These higher
nonair-related exposures can produce elevated blood Pb level, especially when compared to
average blood Pb levels in the current general population. Therefore, while the geometric mean
blood Pb level may have decreased, the tail of the distribution may have remained anchored (for
these paint and drinking water exposed children) resulting in a larger GSD.
A number of studies have been conducted over the past three decades which provide
insights into interindividual variability in Pb levels under various exposure conditions. Many of
the studies from the 1970's and 1980's focused on populations living near smelters with fairly
elevated blood Pb levels compared with levels modeled for our three case studies. For example,
as seen in Appendix H, Exhibit H-7, geometric mean blood Pb levels for populations near active
smelters in the past tended to exceed 8 |ig/dL with GSDs on the order of 1.7. Note, that these
earlier studies do not have readily available summaries of blood Pb levels for the metrics of
interest in this analysis (concurrent and lifetime average metrics) and instead, provide more
generalized summaries based on the individual measurements collected.
Beginning in the 1980's and extending through the 1990's, a number of studies were
conducted focusing on urban populations in specific U.S. cities (e.g., Lanphear et al 2005). These
studies, which were intended to examine the link between Pb exposure and neurological effects
in children, have readily available blood Pb summary data for both concurrent and lifetime
average blood metrics. As can be seen in Appendix H, Exhibit H-9, geometric mean blood Pb
levels vary significantly across these studies, with concurrent values ranging from 5.5 |ig/dL to
14.5 |ig/dL and lifetime average values ranging from 4 |ig/dL to 14.2 |ig/dL. GSDs also vary
3-24
-------
across the studies with GSDs of 1.4 to 1.7 reported for the lifetime average metric and 1.5 to 1.9
for the concurrent metric.
In addition to these smelter-related and city-specific studies, the CDC has also conducted
several iterations of its NHANES national-scale survey over the past three decades which track
changes in the national distribution of blood Pb levels among children in the U.S. (see CD,
Section 4.3.1.3 and Appendix H). As mentioned above, between the earliest NHANES surveys
and the later (post 1980) surveys average blood Pb levels in children in the U.S. decreased
dramatically (following initiatives to remove Pb from gasoline and other products in the late
1970's and 1980's), and the GSD increased significantly. This is most pronounced in the
geometric mean blood Pb levels and associated GSDs reported for the first NHANES survey
(1967-1980) (CDC, 2005) and the second NHANES survey (1988-1991) (CDC, 2005) with
blood Pb levels decreasing from 14.9 to 3.6 and GSDs increasing from 1.4 to 2.1.
GSDs were selected for each case study based on consideration for the study data
summarized above. The selection of these GSDs reflected consideration for a number of factors
including: (a) the type of study area and underlying population involved (e.g., point source
versus more general urban area), (b) the fact that all three study areas use exposure zones with
fairly uniform media concentrations and that the GSD selected will be used to represent blood Pb
variability within each of those zones (c) age of the underlying survey population (the goal was
to match the survey population to our study population to the extent possible), and (d) date of the
survey (generally there is a desire, when possible, to use studies that are more contemporary to
capture any underlying downward trends in blood Pb levels which have occurred). The GSDs
for each of the case studies, as well as the rationale for their selection, is presented below.
3.2.3.1 General Urban Case Study
For this case study, two sets of population blood Pb estimates are developed for each
blood Pb metric. This was done using GSDs intended to reflect: (a) a more uniform population
living in a smaller urban area (represented by a smaller GSD for each metric) and (b) a more
diverse urban population living in a larger urban area (represented by a larger GSD for each
metric). Together, these two sets of estimates provide a range of results for each blood Pb metric
intended to reflect uncertainty associated with this key component of the analysis (i.e.,
interindividual variability in blood Pb levels). The lower-bound GSDs (representing the more
uniform, smaller urban population) were obtained from the Boston study (Bellinger, 1992) (i.e.,
concurrent GSD of 1.7 and a lifetime average value of 1.6). This study represents one of the
more contemporary of the urban studies focusing on a smaller population. By contrast, the
upper-bound GSDs (representing the larger, more diverse population) were obtained using
NHANES IV (CDC, 2005). Here, the GSD from NHANES IV for 1-5 yr olds (2.1) is used for
3-25
-------
the concurrent metric, and the GSD for the lifetime average metric (2.0) was derived by scaling
the value for the concurrent metric using the relationship between concurrent and lifetime seen in
the smaller urban-scale Boston study (Bellinger, 1992).
3.2.3.2 Point Source Case Studies
A critical consideration in identifying the GSD for use in the point source case studies, is
the fact that each is modeled using a spatial template that divides the study area into exposure
zones with relatively uniform media concentrations. As noted earlier, the GSDs are used to
reflect interindividual variability in blood Pb levels for the group of modeled children located
within each of those zones. Therefore, the GSD does not need to provide full coverage for media
concentration variability and its impact on exposure, since this is covered to some extent by the
spatial template. Therefore, a larger GSD such as that suggested by NHANES would likely
overstate variability and a smaller GSD, possibly in line with values from smaller-scale studies
such as the smelter or city-specific studies would seem to be more appropriate. Following this
logic, a GSD of 1.7 was selected for the concurrent blood metric, based on consideration for the
range of values from these smaller-scale studies. A matching lifetime average metric GSD value
of 1.6 was selected, also based on the city-specific studies.
3.3 ESTIMATED MEDIA CONCENTRATIONS
This section summarizes the media concentration estimates for all air quality scenarios at
all three case studies (Tables 3-3 to 3-6). The complete set of media concentration estimates for
each air quality scenario is presented in Appendix C for the general urban case study, in
Appendix D (Attachments D-7 through D-l 1) for the primary Pb smelter case study, and in
Appendix E (Attachments E-3 through E-7) for the secondary Pb smelter case study. Estimates
presented in this section are presented to three (for air) or zero (for dust and soil) decimal places,
which results in various numbers of implied significant figures. This is not intended to convey
greater precision for some estimates than others; it is simply an expedient and initial result of the
software used for the calculation.
For each air quality scenario for the two point source case studies, a range of percentile
estimates derived from a population-weighted distribution of these media concentrations are
presented for each exposure medium. For the general urban case study, however, only a single
value is presented for each exposure medium. This reflects the fact that, while the point source
case studies are modeled using spatial templates that include a large number of US Census
blocks and/or block groups (allowing percentile media concentrations to be identified), the
general urban case study is modeled using a single study area with uniform media
3-26
-------
concentrations. Consequently, there is only one value presented for the general urban case study
for each medium in each air quality scenario.
As discussed in Section 2.3.3, Pb concentration in outdoor soil/dust is not changed with
the alternate air quality scenarios. Rather, outdoor soil/dust concentration is held constant at the
current conditions or current NAAQS level. This reflects the judgement that in most cases, a
reduced air concentration would not yield a changed soil concentration over the near term (e.g.,
years to decades). In the case of an area such as the remediation zone of the primary Pb smelter
case study, however, where soil dynamics have been changed by the substitution of
contaminated soil with clean soil, or in areas where local sources may pose a more significant
source to outdoor soil/dust than historic sources - and where there may be a currently increasing
trend in surface Pb concentration - this may underestimate soil concentrations under some
alternate NAAQS.
As expected, the highest media concentrations for each case study are associated with the
current NAAQS scenario. The relatively lower media concentrations for the alternate NAAQS
scenarios vary in the following order of decreasing estimates: 1) the 0.5 |J,g/m3 maximum
monthly average, 2) the 0.2 |J,g/m3 maximum quarterly, 3) 0.2 |J,g/m3 maximum monthly and 4)
the 0.05 ng/m3 maximum monthly average. In the case of the general urban case study, both of
the current conditions scenarios (mean and high-end) generate media concentrations below the
current NAAQS, since both monitor-based values are below the current NAAQS.
Table 3-3. Estimated annual ambient air concentrations.
Statistic
Average Annual Air Pb Concentration (jig/m3)
Current
Conditions
Alternative NAAQS
Current 0.2 |ig/m3,
NAAQS max quarterly
0.5 |ig/m3
max monthly
0.2 |ig/m3
max monthly
0.05 ng/m3
max monthly
General urban case study
NA- single
study area
High-end: 0.11
Mean: 0.056
0.600
0.080
0.130
0.050
0.013
Primary Pb smelter case study
Maximum
95m percentile
Median
5tn percentile
Minimum
NA
0.740
0.153
0.042
0.015
0.006
0.161
0.033
0.009
0.003
0.001
0.326
0.067
0.019
0.007
0.003
0.130
0.027
0.007
0.003
0.001
0.033
0.007
0.002
0.001
< 0.001
Secondary Pb smelter case study
Maximum
95m percentile
Median
5tn percentile
Minimum
0.126
0.015
0.003
0.001
< 0.001
NAa
0.034
0.004
0.001
< 0.001
< 0.001
0.071
0.008
0.002
< 0.001
< 0.001
0.028
0.003
0.001
< 0.001
< 0.001
0.007
0.001
< 0.001
< 0.001
< 0.001
aThe current conditions scenario for secondary Pb smelter case study met the current NAAQS.
3-27
-------
Table 3-4. Estimated inhalation exposure concentrations.
Statistic
Average Annual Inhalation Ex
Current
Conditions
Current
NAAQS
posure Concentration of Pb (ng/m3)
Alternative NAAQS
0.2 (ig/m3 max
quarterly
0.5 u.g/m3
max monthly
0.2 ng/m3
max monthly
0.05 jig/m3
max monthly
General urban case study
NA - single
study area
High-end: 0.049
Mean: 0.024
0.026
0.034
0.054
0.021
0.005
Primary Pb smelter case study
Maximum
95m percentile
Median
5tn percentile
Minimum
NA
Secondary Pb smelter case studi
Maximum
95tn percentile
Median
5tn percentile
Minimum
0.056
0.007
0.001
< 0.001
< 0.001
0.310
0.064
0.017
0.006
0.002
0.067
0.014
0.004
0.001
< 0.001
0.136
0.028
0.007
0.003
0.001
0.055
0.011
0.003
0.001
< 0.001
0.014
0.003
0.001
< 0.001
< 0.001
j
NAa
0.015
0.002
< 0.001
< 0.001
< 0.001
0.031
0.004
0.001
< 0.001
< 0.001
0.013
0.002
< 0.001
< 0.001
< 0.001
0.003
< 0.001
< 0.001
< 0.001
< 0.001
aThe current conditions scenario for secondary Pb smelter case study met the current NAAQS.
Table 3-5. Estimated outdoor soil/dust concentrations.
Statistic
Projected Average Outdoor Soil/Dust Pb
Concentration (mg/kg)
(Same for all air quality scenarios)3
General urban case study
NA - single
study area
198
Primary Pb smelter case study
Maximum
95m percentile
Median
5tn percentile
Minimum
958
245
85
30
17
Secondary Pb smelter case study
Maximum
95tn percentile
Median
5m percentile
Minimum
315
66
12
1
<1
a Estimates developed for current conditions (or current
NAAQS) scenario were used for all alternate NAAQS
scenarios.
3-28
-------
Table 3-6. Estimated indoor dust concentrations.
Statistic
Projected Average Indoor Dust Pb Concentration (mg/kg or ppm)
Current
Conditions
Alternative NAAQS
Current 0.2 (ug/m3)
NAAQS max quarterly
0.5 (ng/m3)
max monthly
0.2 (u.g/m3)
max monthly
0.05 (n.g/m3)
max monthly
General urban case study
NA- single
study area
High-end: 157-198
Mean: 107-146
426-566
128-169
166-206
102-140
71-88
Primary Pb smelter case study
Maximum
95tn percentile
Median
5tn percentile
Minimum
NA
1,944
219
84
53
41
648
152
68
45
38
1,077
172
73
47
39
557
149
67
44
38
383
138
63
43
38
Secondary Pb smelter case study
Maximum
95tn percentile
Median
5m percentile
Minimum
166
73
63
60
60
NAa
89
63
61
60
60
120
67
61
60
60
84
63
61
60
60
66
61
60
60
60
a The current conditions scenario for secondary Pb smelter case study met the current NAAQS.
3.4 ESTIMATED BLOOD PB LEVELS
Estimates of concurrent and average lifetime blood Pb level derived from outputs of the
IEUBK model (Section 3.2) have been developed for each air quality scenario in each case study
(see Appendix I). Further, multiple sets of blood Pb estimates were generated for each air
quality scenario of each case study, reflecting an effort to consider key sources of uncertainty
(e.g., indoor dust model, blood metric, GSD) and their impact on blood Pb estimates (see Section
2.4.6.2). That is, eight separate blood Pb distributions were generated for each air quality
scenario of the general urban case study (four for each of the two blood Pb metrics) and two
distributions were generated for each air quality scenario of the point source case studies (one for
each of the two blood Pb metrics) (see Table 2-3). The greater number of blood Pb distributions
for the general urban case study reflects the larger number of modeling approaches implemented
for this more conceptual case study, which differ by indoor dust model and GSD.
Because general trends in blood Pb levels across both population percentiles and air
quality scenarios (for a given case study) are similar for both the concurrent and lifetime average
blood metrics, we have only presented estimates for the concurrent metric here. Full results are
presented in Appendix I. Concurrent blood Pb estimates reflecting all pathways (total), the
"recent air" pathways (see Section 3.2.2), and the recent plus past air pathways are presented in
Table 3-7. The total estimates presented in Table 3-7 are those for the median in the distribution
for each case study, and the estimates for the other two categories are the values for those
3-29
-------
categories associated with the median for the total blood Pb estimate. The corresponding
estimates for the 95th percentile in the distribution of total blood Pb estimates for each case study
are presented in Table 3-8. In these tables, all values are rounded to one decimal place, and we
have presented all values below 0.05 as <0.1.
It is noted that given the various limitations of our modeling tools (Sections 2.4.3 and
3.2.2), blood Pb levels associated with air-related exposure pathways and current levels of Pb
emitted to the air (including via resuspension) are likely to fall between the estimates for "recent
air" and those for "recent" plus "past air". Additionally, with regard to the urban case study
"recent air" and "recent" plus "past air" categories, an artifact of the hybrid dust Pb model tends
to mask trends in the two components of dust Pb (recent air and other), which contribute to
"recent air" and "past air" estimates, respectively (see Section 4.3.1 discussion of this
uncertainty). For the primary Pb smelter case study, uncertainty in parsing out the "recent air"
and "other" components of indoor dust (specifically for the site-specific regression model used in
the remediation zone) have lead us to conclude that only "recent plus past air" exposures should
be presented and "recent air" should not be separately presented, as is done for the other case
studies (see Section 3.1.4.2).
With regard to total blood Pb, estimates for the general urban and primary Pb smelter
case studies indicate higher values for the current NAAQS scenario compared to any of the other
scenarios. The difference is 1-2 |ig/dL for the median estimates and up to 6 |ig/dL difference for
the 95th percentile estimates. Although no difference in median total blood Pb estimates
(rounded to whole numbers) is observed between current conditions or alternative NAAQS
scenarios for the general urban or primary Pb smelter case studies, differences among alternate
NAAQS scenarios are observed for the primary Pb smelter case study subarea (Appendix P).
Additionally, a 1 |ig/dL difference in total blood Pb is observed at the 95th percentile between
current conditions and lower alternative NAAQS scenarios for the general urban case study.
3-30
-------
Table 3-7. Summary of blood Pb estimates for medians in total-exposure blood Pb
distributions.
Case Study and Air Quality Scenario
General urban case study
Current NAAQS (1.5 ug/rrr3, max quarterly)
Current conditions - 95 percentile (0.87 ug/nT3, max
quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Current conditions - mean (0.14 ug/rrr3, max quarterly)
Alternative NAAQS (0.2 ug/nT3, max monthly)
Alternative NAAQS (0.05 ug/nr3, max monthly)
Concurrent Blood Pb Level3
Recent airb
Low
1.9
0.4
0.5
0.3
0.2
0.2
0.1
High
2.2
0.8
0.9
0.7
0.6
0.5
0.2
Recent plus
past airb
Low High
2.6 3.2
1.4 1.6
1.5 1.6
1.3 1.5
1.2 1.4
1.2 1.3
1.1 1.1
Total Pb
exposure
Low High
3.1 3.7
2.0 2.1
2.0 2.2
1.9 2.0
1.8 1.9
1.7 1.9
1.6 1.7
Primary Pb smelter - full study area
Current NAAQS (1.5 ug/nT3, max quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/nr3, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
NAC
0.8
1.1
0.7
0.5
0.7
1.5
1.4
1.4
1.4
1.4
Primary Pb smelter - 1.5 km subarea
Current NAAQS (1.5 ug/rrr3, max quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/nT3, max monthly)
Alternative NAAQS (0.05 ug/mj, max monthly)
NAC
4.0
2.6
1.8
1.8
1.1
4.6
3.2
2.5
2.3
1.7
Secondary Pb smelter - full study area
Current conditions
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/nT3, max monthly)
Alternative NAAQS (0.05 ug/mj, max monthly)
<0.1
<0.1
<0.1
<0.1
<0.1
0.4
0.1
0.4
0.4
0.4
1.0
1.0
1.0
1.0
1.0
Secondary Pb smelter -1.5 km subarea
Current conditions
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/mj, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
0
1
<0.1
0
0
1
1
<0.1
0.7
0.7
0.8
0.8
0.6
1.3
1.3
1.3
1.3
1.2
a - Estimates are rounded to one decimal place.
b -The term "past air" includes contributions from the outdoor soil/dust contribution to indoor dust, historical air
contribution to indoor dust, and outdoor soil/dust pathways, while "recent air" refers to contributions from inhalation
of ambient air Pb or ingestion of indoor dust Pb predicted to be associated with outdoor ambient air Pb levels, with
outdoor ambient air also potentially including resuspended, previously deposited Pb (see Section 2.4.3).
c- "Recent air" estimates were not developed for the primary Pb smelter case study (see Section 3.1.4.2).
3-31
-------
Table 3-8. Summary of blood Pb estimates for 95
distributions.
th
percentiles in total-exposure blood Pb
Air Quality Scenario
(and case study)
General urban case study
Current NAAQS (1.5 ug/nT3, max quarterly)
Current conditions - 95th% (0.87 ug/rrr3, max quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Current conditions - mean (0.14 ug/rrr3, max quarterly)
Alternative NAAQS (0.2 ug/nT3, max monthly)
Alternative NAAQS (0.05 ug/mj, max monthly)
Concurrent Blood Pb Level3
Recent airb
Low
4.6
1.1
1.2
0.8
0.5
0.5
0.1
High
7.5
2.8
2.6
2.3
1.9
1.8
0.7
Recent plus
past airb
Low High
6.3 9.9
3.4 5.4
3.5 5.1
3.1 5.1
2.9 4.7
2.9 4.6
2.6 3.9
Total Pb
exposure
Low High
7.6 11.5
5.1 7.2
4.8 6.7
4.4 6.9
4.2 6.5
4.2 6.4
3.9 5.7
Primary Pb smelter - full study area
Current NAAQS (1.5 ug/nr3, max quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/mj, max monthly)
Alternative NAAQS (0.05 ug/mj, max monthly)
NAC
2.8
3.1
2.4
2.5
1.9
4.6
4.2
4.0
4.0
3.8
Primary Pb smelter - 1.5 km subarea
Current NAAQS (1.5 ug/rrr3, max quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/mj, max quarterly)
Alternative NAAQS (0.2 ug/nT3, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
NAC
10.3
7.6
5.9
4.9
3.5
12.3
8.5
6.6
6.1
4.5
Secondary Pb smelter - full study area
Current conditions
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/nT3, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
<0.1
0
1
<0.1
<0.1
<0.1
0.8
1.1
0.8
1.1
0.8
2.4
2.4
2.3
2.4
2.4
Secondary Pb smelter - 1.5 km subarea
Current conditions
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/nT3, max monthly)
Alternative NAAQS (0.05 ug/mj, max monthly)
0
0
0
5
4
2
<0.1
<0.1
2.2
2.5
2.3
1.4
1.6
3.3
3.3
3.1
3.1
3.1
a - Estimates are rounded to one decimal place.
b -The term "past air" includes contributions from the outdoor soil/dust contribution to indoor dust, historical air
contribution to indoor dust, and outdoor soil/dust pathways, while "recent air" refers to contributions from
inhalation of ambient air Pb or ingestion of indoor dust Pb predicted to be associated with outdoor ambient air
Pb levels, with outdoor ambient air also potentially including resuspended, previously deposited Pb (see
Section 2.4.3).
c- "Recent air" estimates were not developed for the primary Pb smelter case study (see Section 3.1.4.2).
3-32
-------
3.5 UNCERTAINTY CHARACTERIZATION AND SENSITIVITY ANALYSIS
The characterization of uncertainty associated with exposure assessment included
performance evaluation which is discussed in this section. In addition, uncertainty in exposure
assessment is also considered as part of characterizing uncertainty in the risk estimates and is
reflected in inclusions of multiple modeling approaches and the qualitative discussion of key
sources of uncertainty which are discussed in Section 4.3. Elements of exposure assessment are
also included in the sensitivity analysis described in Section 4.3.2.
Performance evaluation for the exposure assessment focused on evaluation of projections
of Pb in exposure media (i.e., ambient air, outdoor soil, and indoor dust) (discussed in Section
3.5.1) and projections of Pb in blood (covered in Section 3.5.2). Those case studies (or case
study elements) for which media concentrations were estimated using empirical data as the basis
are not considered here; only those estimates based directly on modeling were included.
Comparing model output to monitoring data is often considered the most desirable form
of performance evaluation. In addition to monitoring data, or in the absence of such data,
outputs from other models and expert opinion about how outputs should look can be used as
comparison benchmarks in performance evaluation. The performance evaluation activities
reported here have focused on the comparison of model outputs to measurements. Favorable
comparisons, including a lack of systematic trends in either over- or underestimation of modeled
results relative to empirical data, provide confidence in the modeling estimates.
3.5.1 Performance Evaluation Related to Exposure Media Modeling
This section discusses performance evaluation conducted for modeling of Pb
concentrations in exposure media including ambient air, outdoor soil and indoor dust.
3.5.1.1 Evaluation of Modeled Ambient Air Pb Concentrations
Performance evaluation for ambient air Pb predictions focused on the two point source
case studies, in which air dispersion modeling was used.
For the primary Pb smelter case study current NAAQS scenario, the ISCST3-Prime
model was used with input files (e.g., source characterization, meteorological data) used in
developing the 2007 proposed revision to the SIP for that location (Appendix D). The
submission to EPA for the proposed SIP revision included a model performance evaluation,
focused on the "actual value" modeling scenario (MDNR 2007a). The actual value modeling
included three separate comparisons based on relating model predictions to measured Pb
concentrations at five monitor sites in the study area. These comparisons included:
3-33
-------
• Day-to-day evaluation of modeling output compared to monitor values: The review
of the model performance evaluation conducted by the state of Missouri concluded
that all sites demonstrated a pattern of overall accuracy for directional prediction (i.e.,
high modeled days being high monitored days and low modeled days being low
monitored days), suggesting that the model was performing well in relating wind
direction to Pb transport (MDNR, 2007a).
• Comparison of source contribution analysis using chemical mass balance (CMB) with
dispersion model predicted relative contributions: Source contribution analysis using
CMB of monitor filter residue to identify significant sources of Pb for each monitor
(e.g., roads inside facility boundaries and yard dust, blast furnace) were compared
with relative contributions predicted by the dispersion model for individual modeled
sources. The review of the model performance evaluation concluded that there was
generally good agreement between the CMB results and the air dispersion results in
terms of major sources contributing Pb at each monitor (MDNR, 2007a).
• Comparison of overall average modeled results with monitored levels: This
performance evaluation involved comparing modeled results (for 247 days simulated
for 2005) at six monitor locations with actual measured values for that same period at
those locations. Results of this evaluation suggested a slight overprediction bias
(<10%) for those sites likely to have the greatest impacts from the facility.
This evaluation of model performance for the actual value modeling scenario increases
confidence in estimates developed for the current NAAQS scenario with this modeling setup and
inputs particular to the 2007 proposed SIP revision.
Performance evaluation of air dispersion modeling for the secondary Pb smelter is
discussed in Appendix E (Section E.2.4), and involved comparing modeled ambient air Pb levels
to measured levels at two monitoring locations within the study area. Results of that evaluation
suggest that the model might be slightly underpredicting levels at the closest monitor, and
slightly more underpredictive of levels at the more distant monitors. The use of meteorological
data that is not site-specific and may not be fully representative of actual wind patterns in the
area may be contributing to this. When modeled concentrations at distances matching those of
the monitors, and for all directional points around the facility, are compared with monitor values,
modeled values are identified which match or exceed the measured values.
3.5.1.2 Evaluation of Modeled Outdoor Soil/Dust Pb Concentrations
Modeling was used in estimating the spatial pattern of outdoor soil/dust Pb
concentrations for the secondary Pb smelter case study (see Section 3.1.3.3). As the modeling
did not use estimates of historical Pb emissions (presumed to be higher than current emissions),
absolute values of the model predictions were not expected to be representative of current Pb
concentrations in soil/dust in this case study. Although measurements were not available for the
3-34
-------
case study location, comparison of model predicted soil Pb levels with Pb concentrations
reported for a surrogate secondary Pb smelter location confirmed this expectation (see Appendix
E, Section E.3). Model predictions were - on average - a factor of 3 lower than Pb
concentrations at the surrogate location. Accordingly, we used a hybrid approach for
characterizing outdoor soil Pb levels for this case study, scaling the model estimates of soil Pb
concentration across the study area by factors derived from studies at other secondary Pb smelter
locations (see Appendix E, Section E.3).
3.5.1.3 Evaluation of Modeled Indoor Dust Pb Concentrations
Performance evaluation completed in support of indoor dust Pb prediction involved two
components. First, the hybrid model developed for the general urban case study was evaluated
using available data from the literature (see Appendix G, Section G.3.6). Results of this
evaluation help us assess the overall reasonableness of this model in supporting indoor dust
prediction for the general urban case study in particular. Additionally, predicted indoor dust Pb
levels for all three case studies were compared with data identified in the literature, including
measurements from individual studies focusing on smaller areas and data from national-scale
surveys. The general urban case study model and model projections are discussed here first
The hybrid indoor dust model developed for the general urban case study (Section 3.1.4.1
and Appendix G) is a combination of mechanistic model (to relate outdoor ambient air Pb to
indoor dust Pb) and empirical data (to characterize the nonair related fraction of indoor dust Pb
in the residential urban setting). Components of the mechanistic portion of the model were
subjected to a range of evaluations based on available data in the literature. In addition, the
conversion of model-generated indoor dust Pb loadings to concentrations (a key step required
prior to blood Pb modeling) was evaluated. Finally, model predictions of indoor dust Pb levels
were evaluated using data from several studies for specific cities.
Evaluation of the mechanistic component of the hybrid dust model focused on (a)
predicted deposition fluxes of Pb to indoor surfaces and (b) prediction of relationship between
indoor air Pb and ambient outdoor air Pb (Appendix G, Section G.3.6). Generally, modeled
indoor deposition fluxes for Pb appear to be in line with values reported for a location in
Manhattan with closed windows (Caravanos et al., 2006). The model predicted ratio of indoor
air Pb to outdoor ambient air Pb are lower (-50%) than ratios based on data collected from
residences in the Midwest (Roy et al., 2003) but similar to ratios generated by a different hybrid
model which combines empirical data with a mass balance modeling approach in predicting
indoor ambient air concentrations (Riley et al., 2002). The underestimate of this ratio, when
compared with the empirical data from Roy et al. (2003), may result from the fact that the hybrid
model does not consider resuspension of indoor dust Pb. Generally, these findings with regard to
3-35
-------
substeps of the mechanistic (and air-focused) component of the hybrid model (Appendix G,
Section G.3.6) suggest a potential to underpredict the influence of outdoor ambient air Pb on
indoor dust Pb. The conversion for predicted indoor dust loadings to concentrations, however,
may overpredict concentrations based on a comparison data collected in residences in the
Midwest (Roy et al., 2003; Appendix G, Section G.3.6). This finding may counterbalance any
potential for the mechanistic component of the model to underestimate the influence of ambient
air Pb on indoor dust Pb.
Indoor dust Pb concentrations (in terms of mass per mass) were identified from a few
studies in the literature (e.g., CD, Table 3-8; Tang et al., 2004), as well as the HUD National
Survey of Lead-based Paint in Housing (USEPA, 1995) described in Section 3.1.4.1. These
studies are identified in Table 3-9. Although the HUD survey was used for the empirical
component of the hybrid model, it is considered here with regard to the total dust Pb
concentrations predicted by both the hybrid and air-only models used in the general urban case
study (use of these models is described in Section 3.1.4.1).6 Comparison of the model-predicted
indoor dust concentrations for the current conditions scenarios (146 and 198 ppm for the hybrid
model and 107-157 ppm for the air-only model) to empirical data collected in Jersey City, NJ,
and Ottawa Canada (Table 3-9) suggest that the model may be underpredicting indoor dust Pb,
although the housing stock in these studies was much older than housing generally in U.S.,
which may mean that the measurements are impacted to a greater extent by indoor Pb paint than
would be the general case in the U.S. Thus, it would be expected that those reported values
would be higher than the model predictions. Additionally, the model-predicted values fall
between the medians for the youngest and oldest houses sampled in the HUD national survey
(Table 3-9). Given that indoor dust Pb modeling completed for the general urban case study was
aimed at capturing central tendency indoor dust Pb levels (and is not expected to, for example,
capture variability related to cleaning rates or indoor paint Pb levels), this finding provides
confidence in the estimates.
6 The use of HUD indoor dust Pb data for performance evaluation is considered reasonable, even in light of
its use in deriving the air-related portion of the hybrid model, since total indoor dust Pb levels generated by the
hybrid model (reflecting both the air-related and non air-related components of the model) are being examined in the
performance evaluation. In this context, nationally representative indoor dust Pb concentrations (obtained from the
HUD dataset) are considered a useful empirical dataset for the evaluation.
3-36
-------
Table 3-9. Evaluation of model-predicted indoor dust Pb levels against empirical data
obtained from the literature.
Case study
General urban
case study
Primary Pb smelter
case study
Secondary Pb
smelter
case study
Modeled indoor dust Pb levels (ppm)
Air quality
scenario
Current
conditions
(air-only
regression
and hybrid)
Current
NAAQS
Current
conditions
Median
5tn to 95tn
Percentile
(min-max)
Mean current conditions:
(107 to 146)
95th percentile current
conditions: (157-198)3
84
63
53-219
(41 -1,944)
60-73
(60-166)
Indoor Dust Pb Observations reported in
the literature
- Residences near smelters: 1283-4140 ppm
(CD, Table 3-8)
- Jersey City, NJ housing (floor): 857 ppm
(CD, Table 3-8)
- Residences in the Midwest (windowsill):
954ppm (CD, Table 3-8)
- Ottawa Canada housing (floor): 222 ppm
(median), 406 ppm (mean) (Tang et al., 2004)
- HUD survey of US housing: 87 ppm (median
for newest houses, built 1960-1979), 406
ppm (median for oldest housing, build <1940)
(USEPA, 1995).
a - because a single study area is used for the general urban case study, a single dust estimate was generated for each
combination of current conditions air quality scenario and indoor dust model (i.e., four separate estimates were generated for the
general urban case study).
For the two point source case studies, the model-predicted indoor dust Pb concentrations
were compared to observations from the literature (see Table 3-9). An important factor to keep
in mind when reviewing the modeled results presented in Table 3-10 is that the current NAAQS
scenario modeled for the primary Pb smelter involved regions of the study area with ambient air
Pb levels significantly higher than either of the other case studies.
In consideration of indoor dust Pb levels predicted for the primary Pb smelter, we have
focused more on the central tendency values. This reflects the fact that the higher-end model-
generated values for this case study likely reflect significant ambient air impacts which are not
captured in any of the empirical data identified in the literature, thereby reducing the utility of
performance evaluations for these higher-end predictions. The median indoor dust Pb
concentration generated for the primary Pb smelter case study (84 ppm) falls near the lower end
of the range in the HUD dataset (i.e., near the median of 87 ppm reported for newer housing in
that study). This seems reasonable since a significant fraction of the study area for this case study
has ambient air Pb levels not significantly different from ambient air Pb levels seen across the
U.S. (see Table 3-4).
Regarding the secondary Pb smelter case study, the median value for this case study (63
ppm) is below the range of values reported in the past for smelters (CD, Table 3-8), and also just
3-37
-------
below median value reported for the youngest housing in the HUD national data set (USEPA,
1995), suggesting that indoor dust Pb levels may be underpredicted for this case study.
3.5.2 Performance Evaluation Related to Blood Pb Modeling
Performance evaluation completed in support of blood Pb modeling involved three steps:
(a) evaluation of candidate blood Pb models for the analysis, (b) comparison of ambient outdoor
air Pb-to-bloodPb ratios generated for the three case studies against ratios obtained from the
literature and (c) comparison of modeled blood Pb levels for these three case studies against
NHANES IV data. Each of these evaluation steps is discussed below.
3.5.2.1 Evaluation of Candidate Blood Pb Models
Evaluation of candidate blood Pb models (IEUBK and Leggett) involved three separate
stages: (a) application of the candidate models to three hypothetical individual child exposure
scenarios used previously by EPA and others in evaluations of blood Pb models, (b) comparison
of candidate model predictions for a general U.S. childhood exposure scenario (using typical Pb
exposures for key pathways) to NHANES IV empirical data, and (c) evaluation of candidate
model performance in replicating measurements of urban child blood Pb levels obtained in
Rochester. Detailed results of the evaluation of candidate blood Pb models are presented in
Appendix J (and summarized in Section J.4).
The first stage (focusing on reproducing results of previous performance evaluations)
demonstrated that we were applying the candidate models correctly. Tests of the models against
specific individual exposure scenarios (Section J.I.3) reproduced, to a high degree, the results of
previous model comparison.
The second stage of the model evaluation (focusing on reproducing general US child
blood Pb levels presented in NHANES IV) demonstrated that, depending on assumptions
regarding typical outdoor soil/dust and indoor dust Pb concentrations, the IEUBK model either
moderately overpredicted GM blood Pb levels (by two-fold or less) or generated predictions
close to NHANES summary statistics (see Section J.3.1). By contrast, predictions from the
Leggett model were more than three to six times higher than the age-specific NHANES IV GM
values.
The third stage of the model evaluation focused on evaluating the candidate models in
predicting blood Pb levels for an urban child cohort (Appendix J, Section J.3.2). The dataset,
which included matched Pb media concentrations (outdoor soil/dust and indoor dust levels) and
blood Pb levels, was collected as part of an epidemiological study focusing on the effects of Pb
exposure in children living in Rochester, NY (Lanphear et al., 1995; Lanphear and Roghmann,
1997). Blood Pb levels for each child sampled in the study were predicted using IEUBK and
Leggett, with the measured media Pb levels collected as part of the study as inputs. These
3-38
-------
predicted blood Pb levels were then compared to the measured blood Pb levels for each child.
The results of this stage of the model evaluation were similar to those from the NHANES
evaluation in that IEUBK results suggested a moderate overprediction (-70%) while Leggett
results indicated a much greater overprediction (a factor of 2 to 6).
Results of the model evaluation for IEUBK and Leggett suggest that IEUBK generates
more representative blood Pb estimates relative to Leggett in the context of the evaluations
conducted here. Based on the results of the model evaluation, we decided to use IEUBK in
generating the primary set of exposure and risk results for this analysis, and include Leggett as
part of the sensitivity analysis but not in the primary analysis (see Section 4.3.2).
3.5.2.2 Evaluation of model-derived outdoor air Pb-to-blood Pb ratios
In deriving the current NAAQS in 1978 (43 FR 46246), USEPA used an estimate of the
relationship between ambient air Pb concentration and associated blood Pb concentration (i.e.,
1:2, |ig/m3 to |ig/dL). In this assessment, we rely on several distinct modeling steps which, when
taken together, translate ambient air Pb into blood Pb. As part of the blood Pb model
performance evaluation we have extracted air-to-blood Pb ratios from the modeling completed
for the three case studies for comparison to estimates reported in the literature.
Ratios were developed that relate ambient air Pb to blood Pb contributed from the
following different exposure pathways or pathway combinations: (a) inhalation of ambient air,
(b) inhalation of ambient air plus ingestion of the Pb in indoor dust that is predicted to be
associated with ambient air Pb levels (i.e., "recent air" per Section 2.4.3), and (c) inhalation of
ambient air plus ingestion of indoor dust plus ingestion of outdoor soil/dust (i.e., not including
the diet and drinking water ingestion pathways). The limitations of our modeling tools precluded
us from parsing air-related blood contributions any more finely. The ratios (actually, the blood
Pb side of the ratio) will be larger as you move from (a) to (c) because there is a progressively
larger fraction of overall blood Pb (exposure) being associated with air. With regard to the
potential impact of ambient air Pb on blood Pb, the first ratio (inhalation pathway) is an
underestimate because it excludes the important ingestion pathways to which ambient air Pb can
contribute. Conversely, the third ratio, although not including any impact of air Pb on diet (and
blood), potentially includes some contributions to blood Pb that are not influenced by air (e.g.,
indoor paint). For the purposes of this model performance discussion, we have focused on the
second type of ratio (those for "recent air") derived using the concurrent blood Pb metric. The
full set of ratios are presented for each case study in Appendix I
In this evaluation, we have considered the ratios derived from the blood Pb estimate prior
to application of the GSD reflecting interindividual variability in Blood Pb levels (i.e., the central
tendency estimate of blood Pb derived for each US Census block or blockgroup for the point
3-39
-------
source case studies and for the entire study area for the general urban case study). Although air-
to-blood Pb ratios can be derived for individual simulated children after application of the GSD,
these would be less relevant to empirical ratios reported in the literature, which tend to capture
central tendency or typical ratios for a study population (through statistical regression modeling,
for example). While this interindividual variability in exposure parameters is not reflected in the
ratios evaluated here, variation in exposure concentration is reflected by the range of ratios
derived for the different exposure zones of the two point source case studies. For the general
urban case study, ratios are presented for both current conditions scenarios (reflecting a mean
and high-end estimate of air Pb concentration).
Table 3-10. Air-to-blood Pb ratios for "recent air" contribution to concurrent blood Pb
level.
Case study
Annual average
ambient air
Pb concentrations
(ug/m3)
Air-to-Blood
Pb
Ratio
General urban - current conditions
Current conditions (mean)
Current conditions (95th percentile)
0.056
0.114
1:4&1:10a
1:4&1:7a
Primary Pb smelter - current NAAQS
Median air concentration
95th percentile air concentration
0.093
0.458
1:3
1:7
Secondary Pb smelter- current conditions
Median air concentration
95th percentile air concentration
0.005
0.011
1:5
1:4
Air-to-Blood Pb Ratios
Identified from the
Literature
- Review of studies published
before 1984 reports air-to-
blood Pb ratios for children
generally ranging from 1 :3 to
1:5(Brunekreef, 1984).
- Pooled analysis of air-to-
blood Pb relationship (log-log
regression) based on above
studies yielded ratios of 1:3 to
1:6(Brunekreef, 1984).
- An air-to-blood ratio
developed based on more
recent study data (Hilts, 2003)
is 1 :7 (for children in the
vicinity of an operating
smelter which experienced a
modification in facility
operations, leading to a
marked decrease in air Pb
emissions).
a The two ratios for the general urban case study correspond to results obtained from the two indoor dust models.
Several observations can be made regarding higher ratios generated for two of the case
studies presented in Table 3-10:
• For the general urban case study, the higher ratios (i.e., 1:10 and 1:7) result from
application of the hybrid dust model which produces a higher indoor dust
3-40
-------
concentration per unit ambient outdoor air Pb relative to the air-only regression
model.
• For the primary Pb smelter case study, the higher ratio (1:7) is also the result of a
more potent dust model (the site-specific regression model developed for the
remediation zone of this study area). That indoor dust model was used for the higher-
impact (95th percentile) block, while the regression (air plus soil) model was used for
the majority of the study area, including the median block (producing the 1:3 ratio).
The air-to-blood Pb ratios from the literature presented in Table 3-10 all come from a
combination of older studies summarized in a single review from 1984 (Brunekreef, 1984) and
more recent data (Hilts, 2003). The Brunekreef, 1984 review presents both (a) the ratios
identified from the reviewed studies, which included surveys focused on smelter and urban study
areas (generally range from 1:3 to 1:5) and (b) the results of a pooled analysis where a log-log
regression model was developed relating ambient air Pb directly to blood Pb levels based on the
underlying study data (this yields the range of 1:3 to 1:6 presented in Table 3-10).
Data presented in Hilts, 2003, track Pb exposure levels (e.g., outdoor ambient air Pb,
outdoor soil Pb and indoor dust Pb) and measured blood Pb levels (annual geometric mean blood
Pb levels for children 6 to 60 months of age), for a population of children living in the vicinity of
smelting operations in Trail, British Columbia for the period 1996 to 2001. The facility
experienced a change in operations in 1997 which resulted in a significant decrease in Pb
emissions from the facility and a resulting decrease in ambient air Pb levels in the vicinity of the
facility. In deriving an air-to-blood Pb ratio from these data, we compared ambient air Pb levels
for 1996 and 1999 to geometric mean blood Pb levels for the same years (all data were obtained
from Table 3 in Hilts, 2003). An air-to-blood Pb ratio of 1:7 was generated based on this
comparison. Note, also, that a second air-to-blood Pb ratio could have been generated by
comparing data for 1996 and 2001 (which would have tracked the continued decrease in both
ambient air Pb levels and geometric blood Pb levels). However, blood Pb measurements in 2001
focused on children 6-36 months of age, while blood Pb measurements for 1996 and 1999,
involved children 6-60 months of age. This disconnect in the age groups would have prevented a
direct comparison of the ratios generated using the two pairings of years. Therefore, we focused
on the ratio developed by comparing 1996 and 1999 since the blood Pb measurements for these
two years focused on an age group (6-60 month old children) closer to that used in our modeling
of risks for the case studies (children <7 years of age).
The air-to-blood Pb ratio of 1:7 generated using the Hilts, 2003 data is slightly higher
than the range presented in Brunekreef, 1984 (1:3 to 1:6). This might reflect the fact that overall
exposures associated with the Hilts, 2003 study are lower than those reflected in the Brunekreef,
3-41
-------
1984 study and, as noted in Brunekreef, 1984, there appears to be greater efficiency in translating
air Pb to blood Pb at lower exposure levels, compared with higher levels.
Although the blood Pb levels associated with Hilts, 2003 (geometric mean blood Pb
levels of 4 to >10 |ig/dL) are closer to those associated with our modeled case studies than are
the levels in Brunekreef (1984), the levels in Hilts are still higher. In the case of the blood Pb
levels reflected in the studies summarized in Brunekreef (1984), these measured blood Pb levels
(typically > lOug/dL) are significantly higher than those modeled for our case studies. This
disconnect between the studies in the literature and the modeled case studies (regarding blood Pb
ranges) introduces uncertainty into performance evaluation of air-to-blood Pb ratios completed
here. That is, it is possible that ratios seen at blood Pb levels significantly above 5 to 10 |ig/dL
(levels seen in the literature) are different from those below 5 |ig/dL (those associated with the
modeled case studies), although this can not be verified without updated characterizations of air-
to-blood Pb ratios based on contemporary trends in Pb exposure and blood Pb levels. Despite
this uncertainty, however, comparison of the air-to-blood Pb ratios developed for the three case
studies against those found in the literature is considered useful.
The lower air-to-blood ratios generated for the three case studies (1:4 for the general
urban case study, 1:3 for the primary Pb smelter case study and 1:4 to 1:5 for the secondary Pb
smelter case study) fall within the range of ratios identified in the literature (1:3 to 1:7).
However, the higher ratios, ranging from 1:7 to 1:10, are higher than the bulk of Brunekreff
(1984) reported ratios and just include the 1:7 ratio developed from the Hilts (2003) dataset,
although a subset of the individual studies reviewed by Brunekreef (1984) presented air-to-blood
Pb ratios that were 8.5 and higher. In several instances, such higher ratios are associated with
lower blood Pb levels and lower ambient air Pb levels (both factors that would seem to be more
relevant to exposure conditions found in the three case studies modeled for this analysis).
However, the studies reporting these higher ratios are complicated by a number of factors (e.g.,
involving older children, use of ambient air measurement techniques which may underestimate
results, thereby inflating ratios, relatively small sample sizes, etc.).
3.5.2.3 Comparison of modeled blood Pb levels to nationally representative data
Evaluation of modeled blood Pb levels using empirical data depends on the availability
of empirical datasets for populations with exposure similar to those in the case studies. While
there are many blood Pb studies reported in the literature, they are largely composed of
populations experiencing exposures from all pathways that are higher than exposures reflected in
our case studies. No datasets were identified for specific subpopulations with exposures
corresponding to current exposures that would be appropriate for performance evaluation here.
3-42
-------
In the case of the general urban case study, while it is not location-specific, we have not
identified any contemporary datasets for urban locations with ambient air Pb levels matching
those modeled in the current scenario that are not heavily influenced by Pb-based paint. For the
primary Pb smelter case study location, the most generally available blood Pb monitoring study
was completed in 2001 and 2002, and given the changes in exposures since that time and the use
of a future (current NAAQS) scenario these older blood Pb surveys do not correspond to the
exposures modeled for this case study. In the case of the secondary Pb smelter, the blood Pb
data available for the county containing the secondary Pb smelter does not indicate residence
location handicapping efforts to consider blood Pb levels for children in the secondary Pb
smelter case study location.
While blood Pb datasets were not identified for populations representative of those
modeled for the three case studies, we have compared the modeled blood Pb levels against
general national-scale blood Pb levels (for a child age group matching that modeled for the case
studies). While each of the case studies involves a population that is different than the national
population (reflecting the nature of the exposure scenario being considered), the background
exposures for the case studies and the national population are similar. Accordingly, the
relationship between the modeled blood Pb levels for each study area and the national
distribution are evaluated. For this evaluation, we have used measured values interpolated from
the NHANES IV dataset. Specifically, we have interpolated a series of percentile estimates for 7
yr olds based on summary data presented for NHANES IV (CDC, 2005). The process of
interpolating these values involved the following steps: (a) use summary statistics (GM and
associated confidence intervals) for 1-5 yr olds and 6-19 yr olds (for years 1999-2002) to
establish lognormal distributions for each age range, (b) identify population percentiles of
interest (e.g., median, 90th, 95th percentiles) for each age group using these lognormal
distributions, and (c) interpolate a series of percentiles for a 7yr age cohort using the percentiles
for the two age ranges (the 1-5 yr old and 6-19 yr old). This interpolation procedure resulted in
the population percentile estimates for a 7 yr old cohort (for the years 1999-2002) presented in
Table 3-11.
3-43
-------
Table 3-11. Blood Pb levels for 7 year olds in the U.S. (interpolated from NHANES IV,
1999-2002).
Population
percentile
median
75th
90th
95th
Blood Pb level
(ug/dL)
1.5
2.55
4.1
5.25
Figure 3-2 compares modeled blood Pb levels for the three case studies with the
interpolated NHANES IV blood Pb levels presented in Table 3-11. Modeled blood Pb levels
presented in Figure 3-2 include (a) the (mean) current conditions scenario for the general urban
case study, including both low-end and high-end estimates reflecting application of different
indoor dust models, and GSDs for this case study, and (b) the current NAAQS scenarios for both
the primary and secondary Pb smelter case studies.
Figure 3-2. Comparison of NHANES IV blood Pb levels with modeled estimates.
median 75th 90th 95th
Population Percentile
-NHANES-IV (interpolated 1999-
2002, 7yr old)
-General urban case study (mean
current conditions - low er)
-General urban case study (mean
current conditions - higher)
- Primary Pb smelter (current
NAAQS)
- Primary PB smelter (smaller
1.5km study area - current
NAAQS)
-Secondary Pb smelter (current
conditions)
-Secondary Pb smelter (smaller
1.5km study area - current
conditions)
The following observations are made regarding the blood Pb modeling results for the
three case studies (Figure 3-2):
• General urban case study: The low- and high-end estimates for the mean current
conditions scenario bracket the NHANES IV data. Of the three case studies, it is
reasonable to assume that results for the current conditions (mean) scenario for this
3-44
-------
case study might be the closest to matching the NHANES. Interestingly, of the
results for the low- and high-end modeling approaches for this scenario, the high-end
results appear to be closer to the NHANES data, which would indicate that modeling
for this case study may not be biased high with regard to total blood Pb levels.
Additionally, unlike the point source case studies, the general urban case study
included consideration of variability in background exposure pathways through
application of the larger GSD in the high-end modeling approach for that case study.
• Primary Pb smelter case study: It would be expected that higher percentile exposure
levels for this case study might exceed the NHANES data, however, that is not seen
either at the 90th or 95th percentiles in Figure 3-2. There are three factors which
may account for this observation. First, we have not generated higher population
percentile estimates (i.e., greater than 95th percentile) and it is possible that
exposures for those higher percentiles at the primary Pb smelter might begin to
exceed levels seen with the NHANES-interpolated data. Second, it is important to
reiterate that modeling for this case study did not account fully for exposure
variability in background Pb exposures. It is likely that if modeling for this case
study had included consideration of background exposure variability (e.g., relatively
high paint Pb and drinking water exposures), there would be a greater difference
between high-end blood Pb levels modeled for this case study and those interpolated
form NHANES IV, with the potential for the primary Pb smelter percentiles to
exceed those derived from the NHANES IV data. Note, however, this lack of full
accounting for blood Pb variability from background sources is not considered a
major limitation in this analysis because the focus is on ambient air-related
exposures. Finally, as discussed in Section 4.3.1, modeling for the point source case
studies did not consider other sources of ambient air Pb besides the smelter. This
could produce low bias in predictions of exposure, especially for portions of the
study area further from the facility.
• Secondary Pb smelter case study: Modeled blood Pb levels for this case study are
significantly lower than NHANES IV levels, especially for higher population
percentiles. As with the primary Pb smelter case study, this likely reflects three
factors: (a) absence of predictions (due to technical limitations) of higher population
percentiles above the 95th percentile, where exposure related to facility emissions is
likely more dominant, (b) exclusion of consideration for higher background
exposures to Pb in paint and drinking water, and (c) failure to consider ambient air Pb
sources impacting the study area, other than the smelter facility. Note, also that, for
the secondary Pb smelter case study, overall uncertainty associated with modeling of
exposure and risk is considered sufficiently high to significantly reduce overall
confidence in results for this case study (see Section 4.3.1).
3-45
-------
REFERENCES
Adgate, J. L.; Willis, R. D.; Buckley, T. I; Chow, J. C.; Watson, J. G.; Rhoads, G. G.; Lioy, P. J. (1998) Chemical
mass balance source apportionment of Pb in house dust. Environ. Sci. Technol. 32: 108-114.
Bellinger, D.C., Stiles, K.M., Needleman, H.L. (1992) Low-level lead exposure, intelligence and academic
achievement: A long-term follow-up study. Pediatrics 90(6): 855-861.
Brunekreef, B. (1984) The relationship between air lead and blood lead in children: a critical review. Science of the
total environment, 38: 79-123.
Caravanos, J.; Weiss, A. L.; Jaeger, R. J. (2006) An exterior and interior Pbed dust deposition survey in New York
City: results of a 2-year study. Environ. Res. 100: 159-164.
Centers for Disease Control and Prevention (CDC). (2005) Blood Lead Levels - United States, 1999-2002.
Available online at: http://www.cdc.gov/mmwr/preview/mmwrhtml/mm5420a5.htm.
Gustavsson, N., Bolviken, B., Smith, D.B., and Severson, R.C. (2001) Geochemical Landscapes of the
Conterminous United States - New Map Presentations of 22 Elements. U.S. Department of the Interior,
U.S. Geological Survey. Professional Paper 1648. Available at http://pubs.usgs.gov/pp/pl648.
Hattis, D. (2005) Preliminary Analysis of Blood Pb Distributional Data from Various Years of the National Health
and Nutrition Surveys (NHANES). Prepared for ICF International. October.
Hilts, S. R. (2003) Effect of smelter emission reductions on children's blood Pb levels. Sci. Total Environ. 303: 51-
58.
Lanphear, B. P.; Emond, M; Jacobs, D. E.; Weitzman, M; Tanner, M; Winter, N. L.; Yakir, B.; Eberly, S. (1995)
A Side-by-Side Comparison of Dust Collection Methods for Sampling Lead-Contaminated House Dust.
Environ. Res. 68(2): 114-123.
Lanphear, B.P. and Roghmann, K. J. (1997) Pathways of Lead Exposure in Urban Children. Environ. Res. 74(67):
73.
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.
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 Robe, R. (2005) Low-level environmental Pb exposure and children's intellectual function: An
international pooled analysis. Environmental Health Perspectives. 113(7):894-899.
Missouri Department of Natural Resources (MDNR). (2007a) Doe Run - Herculaneum State Implementation Plan
(SIP) Dispersion Modeling Review. Memorandum from Jeffry D. Bennett to John Rustige. February 12,
2007. Available at: http://www.dnr.mo.gov/env/apcp/herculaneumsip.htm
Missouri Department of Natural Resources (MDNR). (2007b) 2007 Revision of the State Implementation Plan for
the Herculaneum Lead Nonattainment Area, as adopted by the Missouri Air Conservation Commission
April 26, 2007.
Riley, W. J., McKone, T. E., Lai, A. C., Nazaroff, W. W. (2002) Indoor Paniculate Matter of Outdoor Origin:
Importance of Size-Dependent Removal Mechanisms. Environ. Sci. Technol. 36(2): 200-207.
3-46
-------
Roy, A., Georgopoulos, P. G., Ouyang, M, Freeman, N., Lioy, P. J. (2003) Environmental, Dietary, Demographic,
and Activity Variables Associated With Biomarkers of Exposure for Benzene and Lead. J. Expo. Anal.
Environ. Epidemiol. 13(6): 417-426.
Tang, K. M., Nace, C. G., Jr., Lynes, C. L., Maddaloni, M. A., LaPosta, D., Callahan, K. C. (2004) Characterization
of Background Concentrations in Upper Manhattan, New York Apartments for Select Contaminants
Identified in World Trade Center Dust. Environ. Sci. Technol. 38(24): 6482-6490.
U.S. Environmental Protection Agency. (1989) Review of National Ambient Air Quality Standard for Lead:
Exposure Analysis Methodology and Validation. Office of Air Quality Planning and Standards, Research
Triangle Park, NC. EPA-450/2-89-011. June.
U.S. Environmental Protection Agency. (1994a) Technical Support Document: Parameters and Equations Used in
the Integrated Exposure Uptake Biokinetic Model for Pb in Children (v.099d). EPA 540/R-94/040.
U.S. Environmental Protection Agency. (1994b) Guidance Manual for the IEUBK Model for Lead in Children. EPA
PB93-963510 (OSWER #9285.7-15-1).
U.S. Environmental Protection Agency. (1995) Report on the National Survey of Lead-Based Paint in Housing:
Appendix I: Design and Methodology. EPA 747-R95-004. Office of Pollution Prevention and Toxics.
U.S. Environmental Protection Agency. (1997) Conversion Equations for Use in Section 403 Rulemaking. EPA
747-R-96-012. Office of Pollution, Prevention, and Toxics.
U.S. Environmental Protection Agency. (2000) Hazard Standard Risk Analysis Supplement - TSCA Section 403.
Available online at: http://www.epa.gov/lead/pubs/403risksupp.htm.
U.S. Environmental Protection Agency. (2006a) 1999 National-Scale Air Toxics Assessment. Available at
http://www.epa.gov/ttn/atw/natal999/nsata99.html.
U.S. Environmental Protection Agency. (2006b) Lead soil trend analysis through May, 2006. Evaluation by
individual quadrant. Herculaneum lead smelter site, Herculaneum, Missouri. Prepared by TetraTech for
U.S. EPA, Region 7.
VonLindern, I. H.; Spalinger, S. M.; Bero, B. N.; Petrosyan, V; VonBraun, M. C. (2003) The influence of soil
remediation on lead in house dust. Sci. Total Environ. 303(1-2): 59-78.
3-47
-------
4 RISK ASSESSMENT
This chapter describes the approach used to characterize risk for the initial analyses of the
full-scale assessment, including discussion of the methodology (Section 4.1), presentation of risk
estimates (Section 4.2), and uncertainty characterization (Section 4.3). Additional analyses
completed after the August 2007 CASAC meeting (Henderson, 2007a), involving a core
modeling approach, are presented in Chapter 5.
4.1 METHODS FOR DERIVING RISK ESTIMATES
Risk characterization for this assessment focuses on IQ loss in children. IQ loss is
derived using a set of concentration-response functions developed based on results from a pooled
analysis of epidemiology studies (Lanphear et al., 2005). These concentration-response
functions are combined with the population-level blood Pb distributions generated for the case
studies to produce distributions of IQ loss estimates for each study population. IQ loss is also
apportioned among different exposure pathways using the pathway apportionment information
generated as part of the exposure analysis.
Two key elements of the risk methodology are described in greater detail below: (a) the
concentration-response functions used in the analysis (Section 4.1.1) and (b) the stepwise
analytical procedure used to generate the IQ loss (risk) distributions (Section 4.1.2).
4.1.1 Concentration-Response Functions
As discussed in Section 2.1.5, log-linear concentration-response functions for IQ loss for
the concurrent and lifetime average blood Pb metrics were obtained from a large pooled study
(Lanphear et al., 2005) and used as the basis for estimating IQ changes in children in this
analysis. Three types of concentration-response functions were drawn from this study, and two
equations of each type are used in this assessment: one for the concurrent blood Pb metric, and a
second for the lifetime average blood Pb metric. The three types include: (a) a log-linear
function with cutpoint (Section 4.1.1.1), (b) a log-linear function with low-exposure linearization
(Section 4.1.1.2) and (c) a two-piece linear function (Section 4.1.1.3). Plots of the three types of
IQ loss functions developed for this assessment are presented in Figure 4-1 for the concurrent
blood Pb metric. Table 4-1 presents a comparison of total IQ loss and incremental IQ loss (IQ
loss/|ig/dL) for the three functions across a range of concurrent blood Pb levels. As can be seen
by comparing the plots of the three functions together with the results presented in Table 4-1, the
log-linear function with low-exposure linearization will generate the greatest IQ change across
the exposure range 0 to 10 |ig/dL followed by the log-linear function with cutpoint and then the
two-piece linear function.
4-1
-------
Inclusion of these three types of concentration-response functions in this assessment
produces a range of risk estimates that is indicative of the model uncertainty related to predicting
IQ loss based on modeled blood Pb levels.
Figure 4-1. Comparison of three concentration-response functions for concurrent blood
Pb levels < 10 ug/dL.
log-linear with outpoint
log-linearwith low-exposure
linearization
-two-piece linear
Concurrent PbB (ug/dL)
Table 4-1. Comparison of total and incremental IQ loss estimates below 10 ug/dL for the
three concentration-response functions.
Performance Metric
Total IQ loss
Incremental IQ loss
(points per ug/dL)
at 2 ug/dL
at 5 ug/dL
at 7.5 ug/dL
at 10 ug/dL
<2 ug/dL
<5 ug/dL
<7.5 ug/dL
<10ug/dl_
Concentration-Response Function
Log-linear
with
outpoint
Log-linear
with low-
exposure
linearization
Two-piece
linear
Points, IQ loss
1.9
4.3
5.4
6.2
0.94
0.87
0.73
0.62
4.6
7.0
8.1
8.9
2.29
1.41
1.09
0.89
0.9
2.3
3.4
4.5
0.45
0.45
0.45
0.45
4-2
-------
4.1.1.1 Log-Linear Function with Cutpoint
This concentration-response function is the log-linear function for IQ change from
Lanphear et al. (2005), with incorporation of a cutpoint at the blood Pb level corresponding to
the lowest levels represented by measurements in the underlying pooled analysis. The values
used for cutpoints, for both the concurrent and lifetime average blood Pb metrics, are based on
the corresponding low end of the range of values for these two indices in the pooled analysis
(Hornung, 2007a, b). The values are 1.0 |ig/dL and 1.47 |ig/dL, for the concurrent and lifetime
average blood Pb metrics, respectively. Slopes for the two log-linear functions were obtained
directly from the study (Lanphear et al., 2005). Parameterization for these two log-linear
functions with cutpoints is presented below, along with the mathematical form of the function:
• Form of function: IQ loss = beta * In (concurrent blood Pb/cutpoint)
• Log-linear function with cutpoint (concurrent metric):
- beta (slope): -2.70
- cutpoint: 1.0 |jg/dL
• Log-linear function with cutpoint (lifetime average metric):
- beta (slope): -3.04
- cutpoint: 1.47 |jg/dL
4.1.1.2 Log-Linear Function with Low-Exposure Linearization
We also developed risk estimates that reflect the possibility that IQ loss is associated with
the entire range of Pb exposure all the way down to "zero" exposure, as recommended by
CASAC (Henderson, 2007b). The risk assessment included IQ loss prediction based on a log-
linear IQ change function with a linearization of the log-linear function taking over at lower
exposure levels. The transition point from the log-linear function to the linearized slope was
selected with the same basis as the cutpoint described in Section 4.1.1.1, the lower end of the
range of the pertinent blood Pb index values in the pooled analysis (Lanphear et al., 2005). The
linearized slope is obtained by taking the tangent to the log-linear function at the point of
departure, with different slopes being identified for the concurrent and lifetime average blood Pb
metrics (per Lanphear et al, 2005). Parameterization for the two log-linear functions with low-
exposure linearization is presented below, with the mathematical form of the function:
• F orm of functi on:
- For blood Pb level > cutpoint:
IQ loss = beta * In (concurrent blood Pb/cutpoint) + linear slope * cutpoint
- For blood Pb level < cutpoint:
IQ loss = linear slope * concurrent blood Pb
4-3
-------
• Log-linear function with low-exposure linearization (concurrent metric):
- beta (slope): -2.70
- linear slope to be applied below 1.0 |jg/dL: -2.70
• Log-linear function with low-exposure linearization (lifetime average metric):
- beta (slope): -3.04
- linear slope to be applied below 1.47 |jg/dL: -2.1
4.1.1.3 Two-piece Linear Function
This category of concentration-response function was developed by fitting a two-piece
linear model to the log-linear IQ loss function obtained from the pooled analysis (Lanphear et al.,
2005). Inclusion of this function in the risk assessment is intended to address the practical
problem of the shape associated with application of the log-linear model at the low end of the
blood Pb value range. In this case, we consider an alternate functional form that allows ready
prediction of IQ loss across the full range of modeled Pb exposure. The procedure involved first
generating blood Pb values for each of the two blood Pb metrics, concurrent and lifetime
average, for a set of N = 1,333 simulated children representative of those included in the pooled
analysis (Lanphear et al., 2005). This was accomplished by sampling from a blood Pb
distribution constructed from the median and 95th percentile of the concurrent and lifetime
average blood Pb indices, respectively, reported in Lanphear et al. (2005). IQ values for each of
the 1,333 simulated children were then estimated using the reported log-linear models that relate
blood Pb to absolute IQ (Lanphear, et al., 2005). Nonlinear regression techniques were then
used to fit two piece linear models to these two sets of simulated children with their matched
pairs of blood Pb and IQ values. The regressions provided parameter estimates (slopes and
"hinge" point) of the best fitting two piece linear segment function for each blood Pb metric,
concurrent and lifetime average. As discussed in Section 4.3.1, the use of a constructed dataset
introduces uncertainty related to fitting of a model to a model. Parameterization for the two two-
piece linear functions is presented below, with the mathematical form of the function:
• F orm of functi on:
- For blood Pb level > hinge:
IQ loss = beta 2 * concurrent blood Pb
- For blood Pb level < hinge:
IQ loss = beta 1 * concurrent blood Pb
• Two-piece linear function (concurrent metric):
- "hinge" linking two segments: 10.82 |jg/dL
- beta 1 (slope at < 10.82 |ig/dL): -0.4539
- beta 2 (slope at >10.82 |jg/dL: -0.1130
4-4
-------
• Two-piece linear function (lifetime average metric):
- "hinge" linking two segments: 13.39 |jg/dL
- beta 1 (slope at < 13.39 pg/dL): -0.3790
- beta 2 (slope at >13.39 |jg/dL: -0.1187
4.1.2 Projection of Population Risk
Risk characterization completed for this assessment involved converting the population-
level blood Pb distributions generated for the three case studies into population-level
distributions of IQ loss using the three types of concentration-response functions described in the
last section. This procedure is described below for each of the functions.
• Log-linear function with outpoint: Each modeled blood Pb level is compared against
the cutpoint. If the blood Pb level is lower than the cutpoint, then no IQ loss is
estimated because the simulated individual's blood Pb level is below the level for
predicting IQ loss with this function. If the blood Pb level is greater than the
cutpoint, then the log-linear function is used to predict IQ loss for the portion of the
estimated blood Pb level extending above the cutpoint.
• Log-linear function with low-exposure linearization: Each modeled blood Pb level is
compared against the point of linearization. If the blood Pb level is below the point
where the function becomes linear, than the linear slope is used to predict IQ loss. If
the modeled Pb level is above the point where the function becomes linear, than IQ
loss is calculated as the sum of IQ loss across the linear portion of the curve plus the
additional contribution from the log-linear portion of the function extending up to the
total blood Pb level.
• Two-piece linear function: Similar to the last function, the modeled blood Pb level is
compared against the blood Pb level of the "hinge". If the blood Pb level falls below
the hinge, as is the case for most simulated individuals at the three case studies, then
the steeper, low-exposure slope is used to estimate IQ loss. If the simulated blood Pb
level falls above the hinge, then IQ loss associated with the low-exposure (steeper
slope) piece of the function (for the portion of the blood Pb level up to the hinge) is
combined with IQ loss estimated using the shallower and higher exposure slope, for
that portion of the blood Pb level extending above the hinge.
The IQ loss estimates generated using this approach are pooled to form a population-level
distribution of IQ loss for a given study area. Each of these IQ loss estimates are pathway
apportioned among policy-relevant pathways and policy-relevant background based on pathway
contribution to the underlying blood Pb levels (see Sections 2.4.3 and 3.2.2). As with pathway
apportionment for blood Pb levels (Section 3.2.2), pathway apportionment of risk estimates is
also at the level of exposure zone. All simulated individuals from a given zone are assigned the
4-5
-------
same pathway apportionment. As noted with regard to blood Pb levels (Section 3.2.2), there is
increased uncertainty associated with pathway apportioned IQ loss estimates for higher-end risk
percentiles, since these reflect an assumption that relative pathway contributions at central
tendency exposure levels hold at the high-end percentiles of the blood Pb level distribution.
Just as with the population-level exposure estimates discussed in Section 3.2.2, risk
estimates generated using the approach outlined above are used to generate several types of risk
metrics, depending on the case study. For the two point source case studies, because they are
location-specific and include a defined and enumerated receptor population, two categories of
risk metrics are generated:
• Population-weighted risk (IQ loss) percentiles: IQ loss (with pathway apportionment)
for simulated individuals representing specific points along the population risk
distribution (e.g., 50, 90, 95 percentile simulated individuals).
• Incidence counts: Number of children within a given study area projected to experience a
magnitude of IQ loss.
As for the exposure estimates, the general urban case study risk metrics are restricted to
population-weighted risk (IQ loss) percentiles.
4.2 RISK ESTIMATES
Estimates of IQ loss resulting from Pb exposure have been developed for each air quality
scenario in each case study (see Appendix K). In addition, IQ loss results have been developed
for subareas associated with each of the point-source case studies (see Appendix P). Further,
multiple sets of risk results were generated for each combination of case study and air quality
scenario, in an effort to consider key sources of uncertainty and their impact on blood Pb
estimates (see Section 2.4.6.2). That is, twenty four separate risk distributions were generated
for each air quality scenario of the general urban case study and six distributions were generated
for each air quality scenario of the point source case studies (see Table 2-3).
As discussed in Section 2.4.6.2, generating multiple sets of risk results for each
combination of case study and air quality scenario provides a range of results reflecting the
impact of key sources of uncertainty on risk results. However, because we could not assign
specific confidence levels to each modeling approach, these multiple sets of results are not
translated into single uncertainty distributions of risk for each air quality scenario in each case
study. Therefore, we consider the multiple sets of risk results to span the best estimate risk
distribution. In response to CASAC comments on the July 2007 draft risk assessment report
(Henderson, 2007a), however, we have omitted results from the two-piece linear (with hinge at
4-6
-------
10.82 |ig/dL for concurrent blood metric) from the summary of results presented in this chapter
(these results are available in Appendix K).
Three types of risk estimates are summarized in Tables 4-2 and 4-3:
• Risk estimates for recent air: This estimate is the portion of total IQ loss derived
from inhalation of ambient air Pb and ingestion of Pb in indoor dust that is predicted
to be associated with ambient air Pb concentrations based on the indoor dust model
(i.e., "recent air" risk per Sections 2.4.3 and 3.2.2) Given the modeling approach
used in this analysis and its limitations, this set of risk estimates is expected to be
most responsive to alternative NAAQS.
• Risk estimates for recent air plus past air (other indoor dust and outdoor soil
contribution): This estimate is the portion of total IQ loss derived from inhalation of
ambient air Pb, ingestion of indoor dust and ingestion of outdoor soil/dust (Sections
2.4.3 and 3.2.2). That is, this estimate includes the recent air pathways in addition to
other indoor dust and outdoor soil pathways. This estimate include some contribution
from indoor paint because the indoor dust models handicapped our ability to parse out
this contribution. Otherwise, these estimates reflect contributions from all except the
policy-relevant background pathways of diet and drinking water.
• Risk estimates for all pathway contributions: This estimate is the IQ loss associated
with the total blood Pb concentration.
As noted previously (Sections 2.4.3 and 3.2.2), policy-relevant pathways are represented
in the first two bullets above, with the true values for the policy-relevant pathways considered to
fall between the estimates for "recent air" and those for "recent" plus "past air". Additionally,
with regard to the urban case study "recent air" and "recent" plus "past air" categories, an artifact
of the hybrid dust Pb model tends to mask trends in the two components of dust Pb (recent air
and other), which contribute to "recent air" and "past air" estimates, respectively (see Section
4.3.1 discussion of this uncertainty). For the primary Pb smelter case study, uncertainty in
parsing out the "recent air" and "other" components of indoor dust (specifically for the site-
specific regression model used in the remediation zone) have led us to conclude that only "recent
plus past air" IQ loss estimates should be presented and "recent air" results should not be
separately presented for the primary Pb smelter, as is done for the other case studies (see Section
3.1.4.2).
IQ loss estimates reflecting all exposure pathways (total), the "recent air" pathways (see
Section 3.2.2), and the recent plus past air pathways are presented in Tables 4-2 and 4-3. The
total estimates presented in Table 4-2 are those for the median in the distribution of total risk,
and the estimates for the other two categories in that table (i.e., recent air and recent plus past air)
are the values for those categories associated with the median for the total exposure pathway
4-7
-------
estimate. The corresponding estimates for the 95th percentile in the distribution of IQ loss
estimates for total blood Pb for each case study are presented in Table 4-3. This presentation
may lead to some seeming inconsistencies in trends for recent air and recent air plus past air risk
estimates across air quality scenarios. This is because the recent air or recent plus past air
estimates associated with the median and 95th percentile total blood Pb estimate may not
necessarily be the median and 95th percentiles of the distribution of estimates for those specific
categories. This is because the blood Pb level (and associated total IQ loss) for a simulated child
reflects not only the total Pb uptake (from both background and ambient air-related pathways),
but also the GSD-based adjustment factor. This means that two simulated children could have
identical blood Pb levels (and IQ loss estimates), but one child could have a higher recent air
exposure, with the other child having their lower recent air exposure compensated for by a higher
GSD-based adjustment factor, thereby resulting in both children having the same total blood Pb
level.
In presenting IQ loss estimates, all values are rounded to one decimal place. The
complete set of risk estimates are presented in Appendix K. The risk estimates in Tables 4-2 and
4-3 indicate the following:
• IQ loss estimates were uniformly higher for the current NAAQS air quality scenario
compared with any of the alternative NAAQS or current conditions air quality
scenarios for the primary Pb smelter and general urban case studies. This trend held
for both the median and 95th percentile risk results. Risk estimates for the secondary
Pb smelter case study (both the full study area and sub-area) differed little across any
of the air quality scenarios.
• IQ loss estimates decreased with increasingly lower alternative NAAQS, as a general
observation across case studies. However, this trend varied in consistency and
magnitude. For example, the trend was more pronounced with the recent air estimates
(both median and 95th percentile) for the general urban case study, and was less
obvious with the risk estimates for total blood Pb (in this case, it is likely that
background exposures overwhelmed the recent air related exposures). The trend was
also stronger for the subarea of the primary Pb smelter, compared with the full study
area, reflecting the fact that the subarea has higher ambient air Pb impacts and
therefore, is likely to demonstrate greater variation in IQ impacts across alternative
NAAQS.
• As expected, risks associated with recent plus past air contributions are larger, across
all population percentiles for all case studies and air quality scenarios evaluated,
compared with the recent air risk estimates. This reflects that the "past air" category
includes contributions from the outdoor soil/dust contribution to indoor dust, historical
air contribution to indoor dust, and outdoor soil/dust pathways, while "recent air"
refers to contributions associated with outdoor ambient air Pb levels, either by
inhalation of ambient air Pb or ingestion of indoor dust Pb predicted to be associated
with outdoor ambient air Pb levels, including resuspended, previously deposited Pb
4-8
-------
(see Section 2.4.3). As noted previously (Sections 2.4.3 and 3.2.2), policy-relevant
pathways are represented in both of these categories, with the true values for the
policy-relevant pathways considered to fall between the estimates for "recent air" and
those for "recent" plus "past air".
Recent air and recent plus past air IQ loss estimates were higher for the subareas of the
point source case studies. This reflects the fact, mentioned above, that these subareas
experience significantly greater ambient air-related impacts compared with the full
study areas. While the full study areas include a large number of children in areas with
ambient air Pb levels well below the alternative NAAQS considered, whose exposures
consequently are less affected under alternative NAAQS, these areas are omitted from
the subarea, thus increasing the percentage of simulated children whose exposures and
associated risks are notably reduced under alternate NAAQS.
Risk estimates generated for both of the point source case studies (full study areas) are
lower across all population percentiles (for the same air quality scenario) compared
with the general urban case study. This reflects two factors. First, the point source
case studies include spatial gradients, such that areas near the facility may be at or near
a given ambient air Pb level (associated with an air quality scenario), while the
majority of the study area (and consequently study population) experiences ambient air
Pb levels considerably lower than the level associated with the air quality scenario. By
contrast, the general urban case study assumed a uniform ambient air Pb concentration
at the level of the air quality scenario being evaluated. Consequently, the entire
hypothetical study population would experience that uniform air concentration. The
second reason for the point source case studies having risk estimates lower than those
for the general urban case study, is that they only consider Pb emitted from the
industrial facility of interest and do not consider emissions from any other sources
within the study area, or near the study area that could impact populations within the
study area. By contrast, current conditions for the general urban case study is based on
monitoring data for urban locations, which will by default reflect the actual mix of Pb
sources influencing air Pb concentrations. Consequently, the general urban case study
reflects the combined impact of different Pb sources, while the point source case
studies only consider emissions from the smelter under consideration and the influence
of these emissions lessens (and consequently air Pb concentrations decline) with
distance from the facilities.
For the general urban case study, the large difference seen in the recent air risk
estimates between the three higher and the lowest alternate NAAQS scenario for the
95th percentile is not seen for the risk estimates for recent plus past air contributions.
This reflects an artifact of the approach used in the hybrid indoor dust model to
characterize "other" (non-"recent air"-associated) contributions to indoor dust (see
Appendix G). Because of the method used to convert dust loadings to dust
concentrations within the hybrid dust model and the approach used to apportion total
dust concentrations between other and recent air components, the "other" indoor dust
concentration predicted by this model varies with air quality scenario, with that value
increasing as the ambient air Pb level decreases (see Appendix C). This means that, as
the recent air contribution to exposure through indoor dust ingestion decreases (as the
lower alternative NAAQS levels are considered), the estimate of contribution of
4-9
-------
"other" indoor dust actually increases. This partially accounts for the trend noted
above. That is, even though recent air risk estimates drop, they are partially offset by
risk resulting from an increase in exposure to the "other" Pb contributions to indoor
dust. The impact of this aspect of the hybrid dust model is identified in Section 4.3.1
and evaluated in a sensitivity analysis presented in Chapter 5.
Risk estimates for the secondary Pb smelter case study are considerably lower than risk
estimates for the other case studies (for all air quality scenarios considered).
Limitations in the design for the secondary Pb smelter case study (described in Section
4.3.1) contribute significant uncertainty to these estimates. Accordingly these results
should not be used to draw conclusion regarding similar facilities risks experienced by
populations living in the vicinity of other similar point sources.
4-10
-------
Table 4-2. Summary of risk estimates for medians of total-exposure risk distributions.
Case Study and Air Quality Scenario
Points IQ loss3
Recent airb
Low
General urban case study
Current NAAQS (1.5 ug/nT1, max quarterly)
Current conditions - 95tn percentile
(0.87 ug/m3, max quarterly)
Alternative NAAQS (0.5 ug/nT1, max monthly)
Alternative NAAQS (0.2 ug/nT1, max quarterly)
Current conditions - mean
(0.14 ug/m3, max quarterly)
Alternative NAAQS (0.2 ug/nT1, max monthly)
Alternative NAAQS (0.05 ug/nf', max monthly)
2.1
0.4
0.5
0.3
0.2
0.2
High
Recent plus
past airb
Low
4.3
2.0
2.1
1.7
1.4
1.4
<0.1 0.6
2.9
1.3
1.4
1.2
1.1
1.0
0.9
High
Total Pb
exposure
Low
6.0
4.0
4.1
3.8
3.6
3.6
3.1
3.5
1.8
1.9
1.7
1.5
1.5
1.3
High
7.0
5.4
5.4
5.2
5.0
5.0
4.6
Primary Pb smelter - full study area
Current NAAQS (1.5 ug/nT1, max quarterly)
Alternative NAAQS (0.5 ug/nT1, max monthly)
Alternative NAAQS (0.2 ug/nv1, max quarterly)
Alternative NAAQS (0.2 ug/nT1, max monthly)
Alternative NAAQS (0.05 ug/nf1, max monthly)
NAC
0.6
0.8
0.6
0.4
0.4
2.5
2.2
2.7
2.1
2.4
1.1
1.0
0.9
0.9
0.9
3.8
3.7
3.6
3.6
3.5
Primary Pb smelter - 1.5 km subarea
Current NAAQS (1.5 ug/nT1, max quarterly)
Alternative NAAQS (0.5 ug/nT1, max monthly)
Alternative NAAQS (0.2 ug/nv1, max quarterly)
Alternative NAAQS (0.2 ug/nT1, max monthly)
Alternative NAAQS (0.05 ug/nf1, max monthly)
NAC
3.2
2.1
1.5
1.2
0.9
6.3
4.9
4.3
4.0
2.9
3.7
2.6
2.0
1.9
1.4
6.8
5.8
5.2
2.0
4.6
Secondary Pb smelter - full study area
Current conditions
Alternative NAAQS (0.5 ug/nT1, max monthly)
Alternative NAAQS (0.2 ug/nv1, max quarterly)
Alternative NAAQS (0.2 ug/nT1, max monthly)
Alternative NAAQS (0.05 ug/nf1, max monthly)
0° <0.1
0 <0.1
0 <0.1
0 <0.1
0 <0.1
0°
0
0
0
0
1.0
1.0
1.0
0.9
1.0
0°
0
0
0
0
2.8
2.7
2.7
2.7
2.7
Secondary Pb smelter - 1.5 km subarea
Current conditions
Alternative NAAQS (0.5 ug/mj, max monthly)
Alternative NAAQS (0.2 ug/nv1, max quarterly)
Alternative NAAQS (0.2 ug/nT1, max monthly)
Alternative NAAQS (0.05 ug/mj, max monthly)
<0.1 0.4
<0.1 0.1
<0.1 <0.1
<0.1 0.1
<0.1 <0.1
0.4
0.4
0.3
0.4
0.3
2.7
1.9
1.7
1.9
2.1
0.8
0.7
0.6
0.6
0.6
3.8
3.7
3.6
3.6
3.6
a - With the exception of "true zero values" (see note d below), all estimates are rounded to one decimal place.
b -The term "past air" includes contributions from the outdoor soil/dust contribution to indoor dust, historical air
contribution to indoor dust, and outdoor soil/dust pathways, while "recent air" refers to contributions from
inhalation of ambient air Pb or ingestion of indoor dust Pb predicted to be associated with outdoor ambient air
Pb levels, with outdoor ambient air also potentially including resuspended, previously deposited Pb (see
Section 2.4.3).
c- "Recent air" estimates were not developed for the primary Pb smelter case study (see Section 3.1.4.2).
d - Table entries of "0" (in the "low" column) are truly zero values and reflect application of log-linear
concentration-response function with cutpointto blood Pb estimates below 1 ug/dL.
4-11
-------
rth
Table 4-3. Summary of risk estimates for 95m percentiles of total exposure risk
distributions.
Case Study and Air Quality Scenario
Points IQ loss3
Recent airb
Low
General urban case study
Current NAAQS (1.5 ug/m3, max quarterly)
Current conditions - 95tn percentile
(0.87 ug/m3, max quarterly)
Alternative NAAQS (0.5 ug/m3, max monthly)
Alternative NAAQS (0.2 ug/m3, max quarterly)
Current conditions - mean
(0.14 ug/m3, max quarterly)
Alternative NAAQS (0.2 ug/m3, max monthly)
Alternative NAAQS (0.05 ug/m3, max monthly)
3.3
1.0
1.0
0.7
0.5
0.4
0.1
High
Recent plus
past airb
Low
6.0
3.4
3.6
2.9
2.5
2.3
1.0
4.6
3.1
3.1
2.8
2.7
2.7
2.4
High
Total Pb
exposure
Low
8.3
6.6
6.9
6.4
6.1
6.0
5.5
5.5
4.2
4.2
4.0
3.9
3.9
3.7
High
10.5
8.8
9.1
8.7
8.5
8.4
8.1
Primary Pb smelter - full study area
Current NAAQS (1.5 ug/m3, max quarterly)
Alternative NAAQS (0.5 ug/m3, max monthly)
Alternative NAAQS (0.2 ug/m3, max quarterly)
Alternative NAAQS (0.2 ug/m3, max monthly)
Alternative NAAQS (0.05 ug/m3, max monthly)
NAC
2.3
2.5
1.9
2.0
1.5
4.2
4.5
5.2
4.0
4.9
3.7
3.4
3.2
3.2
3.1
6.8
6.6
6.5
6.4
6.3
Primary Pb smelter - 1.5 km subarea
Current NAAQS (1.5 ug/m3, max quarterly)
Alternative NAAQS (0.5 ug/m3, max monthly)
Alternative NAAQS (0.2 ug/m3, max quarterly)
Alternative NAAQS (0.2 ug/m3, max monthly)
Alternative NAAQS (0.05 ug/m3, max monthly)
NAC
5.7
5.1
4.7
3.9
2.8
8.8
7.5
6.9
6.7
4.7
6.8
5.8
5.3
4.9
3.6
9.5
8.5
7.8
7.6
6.5
Secondary Pb smelter - full study area
Current conditions
Alternative NAAQS (0.5 ug/m3, max monthly)
Alternative NAAQS (0.2 ug/m3, max quarterly)
Alternative NAAQS (0.2 ug/m3, max monthly)
Alternative NAAQS (0.05 ug/m3, max monthly)
<0.1 <0.1
0.1
0.1
<0.1 <0.1
<0.1 <0.1
<0.1 <0.1
0.8
1.0
0.8
1.1
0.8
1.8
2.0
2.1
1.8
1.7
2.4
2.3
2.3
2.3
2.3
5.2
5.1
5.1
5.1
5.1
Secondary Pb smelter - 1.5 km subarea
Current conditions
Alternative NAAQS (0.5 ug/m3, max monthly)
Alternative NAAQS (0.2 ug/m3, max quarterly)
Alternative NAAQS (0.2 ug/m3, max monthly)
Alternative NAAQS (0.05 ug/m3, max monthly)
0.5
0.4
0.2
1.4
0.5
0.1
<0.1 0.2
<0.1 <0.1
2.1
2.4
2.3
1.4
1.6
5.0
3.9
3.2
4.3
4.2
3.2
3.2
3.1
3.1
3.0
6.3
6.3
6.1
6.1
6.1
a - Estimates are rounded to one decimal place.
b -The term "past air" includes contributions from the outdoor soil/dust contribution to indoor dust, historical air
contribution to indoor dust, and outdoor soil/dust pathways, while "recent air" refers to contributions from
inhalation of ambient air Pb or ingestion of indoor dust Pb predicted to be associated with outdoor ambient air
Pb levels, with outdoor ambient air also potentially including resuspended, previously deposited Pb (see
Section 2.4.3).
c- "Recent air" estimates were not developed for the primary Pb smelter case study (see Section 3.1.4.2).
4-12
-------
4.3 UNCERTAINTY CHARACTERIZATION AND SENSITIVITY ANALYSIS
This section discusses uncertainty related to exposure and risk estimates generated for
this analysis. Several methods have been used to examine uncertainty in our modeling approach
and its potential impact on exposure and risk estimates. These include:
• qualitative discussion of key sources of uncertainty and their potential impact on
exposure and risk estimates (Section 4.3.1),
• evaluation of model performance, including comparison with empirical data (Section
4.3.3), and,
• development of multiple sets of exposure and risk estimates for each assessment
scenario that illustrate the combined impact of different models and input data on risk
results and the associated uncertainly (Section 4.3.4).
In addition to these methods for considering uncertainty, a sensitivity analysis was also
conducted to characterize the potential impact of individual modeling elements on risk results
(see Section 4.3.2). Each of these elements of the uncertainty characterization, along with the
sensitivity analysis, is briefly summarized below.
4.3.1 Qualitative Discussion of Key Sources of Uncertainty
Given the complexity of this assessment and the range of models and input data used in
completing it, there is a wide variety of sources of uncertainty potentially impacting the exposure
and risk results generated (Appendix M, Table M-l). This section identifies those sources of
uncertainty with the potential to have a significant impact on risk results (see Appendix M, Table
M-l bold text). When it was feasible with the available methods and data, these key sources of
uncertainty have been quantitatively assessed for their potential impact on risk results either as
part of the multiple modeling approaches implemented (see Section 4.3.4). In addition, some of
these modeling elements have been included in the sensitivity analysis (see Section 4.3.2). Key
sources of uncertainty include:
• Temporal aspects: As described in Section 2.4.1, exposure for the simulated child
population begins at birth and continues for 7 years, with Pb concentrations in all
exposure media remaining constant throughout the period, and children residing in the
same exposure zone throughout the period. In characterizing exposure media
concentrations, annual averages are derived and held constant through the seven year
period. Exposure factors and physiological parameters vary with age of the cohort
through the seven year exposure period, several exposure factors and physiological
parameters are varied on an annual basis within the blood Pb modeling step (see
Section 3.2). These aspects are a simplification of population exposures that
contributes uncertainty to our exposure and risk estimates.
4-13
-------
• General urban case study: This case study differs from the others in several ways
(described in more detail in Section 2.2.1). It is by definition a general case study and
not based on a specific location. There is a single exposure zone for the case study
within which all media concentrations of Pb are assumed to be spatially uniform; that
is, no spatial variation within the area is simulated (see Sections 2.4.2, and 3.1.1).
Additionally, the case study does not rely on any specific demographic values. Within
the single exposure zone a theoretical population of unspecified size is assumed to be
uniformly distributed. Thus this case study is a simplified representation of urban
areas intended to inform our assessment of the impact of changes in ambient Pb
concentrations on risk, but which carries with it attendant uncertainties in our
interpretation of the associated exposure and risk estimates. For example, the risk
estimates for this case study, while generally representative of an urban residential
population exposed to the specified ambient air Pb levels, cannot be readily related to a
specific urban population. Specific urban populations are spatially distributed in a
nonuniform pattern and experience ambient air Pb levels that vary through time and
space. Consequently, interpretations of the associated blood Pb and risk estimates with
regard to their relevance to specific urban residential exposures carry uncertainty and
presumably an upward bias in risk, particularly for large areas, across which air
concentrations may vary substantially.
• Point source case studies: Dispersion modeling was used to characterize ambient air
Pb levels in the point source case studies. This approach simulates spatial gradients
related to dispersion and deposition of Pb from emitting sources. In the case of the
point sources modeled, sources were limited to those associated with the smelter
operations, and did not include other sources such as resuspension of roadside Pb not
immediately related to facility operations, and other stationary sources of Pb within or
near the study area. This means that, with distance from the facility, there is likely
underestimation of ambient air-related Pb exposure because with increased distance
from the facility there would be increasing influence of other sources relative to that of
the facility. We believe this limitation to have more significant impact on risk
estimates associated with the full study than on those for the subareas, and to perhaps
have a more significant impact on risk estimates associated with the smaller secondary
Pb smelter (see below).
• Secondary Pb smelter case study: Air Pb concentration estimates derived from the air
dispersion modeling completed for the secondary Pb smelter case study are subject to
appreciably greater uncertainty than that for those for the primary Pb smelter case
study due to a number of factors, including: (a) a more limited and less detailed
accounting of emissions and emissions sources associated with the facility (particularly
fugitives), (b) a lack of prior air quality modeling analyses and performance analyses,
and (c) a substantially smaller number of Pb-TSP monitors in the area that could be
used to evaluate and provide confidence in model performance1. Further, as mentioned
in the previous bullet, no air sources of Pb other than those associated with the facility
were accounted for in the modeling. Given the relatively smaller magnitude of
1 The information supporting the air dispersion modeling for the primary Pb smelter case study (see Section
3.5.1.1) provides substantially greater confidence in estimates for that case study.
4-14
-------
emissions from the secondary Pb smelter, the underestimating potential of this
limitation with regard to air concentrations with distance from the facility has a greater
relative impact on risk estimates for this case study than for the primary Pb smelter
case study. The aggregate uncertainty of all of these factors has left us with low
confidence in estimates for this case study. We note that exposure and risk estimates
are lower than those for the other case studies. Although we had initially intended to
use this case study as an example of areas near stationary sources of intermediate size
(smaller than the primary Pb smelter), our experience with this analysis indicates that
substantially more data and multiple case studies differing in several aspects would be
needed to broadly characterize risks for such a category of Pb exposure scenarios.
• Indoor dust Pb modeling for the general urban case study: The hybrid indoor dust Pb
model was developed for the general urban case study due to a lack of an existing
urban-focused dust model, and this hybrid model is subject to particular uncertainties.
Key among these uncertainties is failure to consider house-to-house variability in
factors related to the infiltration of ambient air Pb indoors and subsequent buildup of
Pb on indoor surfaces. This handicaps our ability to predict variation in indoor dust Pb
levels for non-typical residential conditions. In addition, a lack of comprehensive data
in the literature on rates and efficiency of indoor cleaning, introduces uncertainty into
the model. The method used to convert Pb loadings generated by the hybrid model to
Pb concentrations is also subject to uncertainty, partially due to the age of the
underlying dataset. Because the underlying dataset is an older dataset (see Section
3.1.4.1), there is potential for bias toward greater representation of housing with Pb-
based paint. Additionally, the component of the hybrid model that converts dust Pb
loadings to concentrations (Appendix G, Section G.3.4) introduces an uncertainty into
the estimates of percent contribution from recent air compared to other pathways. This
is related to a nonlinearity in this conversion and results in the "other" indoor dust
concentration predicted by this model varying with air quality scenario, with that value
increasing as the ambient air Pb level decreases (see Appendix C). This means that, as
the recent air contribution to exposure through indoor dust ingestion decreases for the
lower alternative NAAQS levels, the estimate of contribution of "other" indoor dust
actually increases. This issue of the uncertainty in estimating the "other" component of
indoor dust (and relating it to the "recent air" component) has been examined further as
a sensitivity analysis (see Section 5.3.3.4).
• Estimates of indoor dust Pb concentrations for the primary Pb smelter case study
(application of the site-specific regression model): There is uncertainty associated
with the site-specific regression model applied in the remediation zone (see Section
3.1.4.2), and relatively greater uncertainty associated with its application to air quality
scenarios that simulate notably lower air Pb levels. Limitations in the dataset from
which the model was derived limited its form to that of a simple regression that
predicts dust Pb concentration as a function of air Pb concentration plus a constant
(intercept). We recognize, however, that there may be variables in addition to air that
influence dust Pb concentrations and their absence in the regression contributes
uncertainty to the resulting estimates. To the extent that these unaccounted for
variables are spatially related to the smelter facility Pb sources, our estimates could be
biased, not with regard to the absolute dust Pb concentration, but with regard to
4-15
-------
differences in dust Pb concentration estimate between different air quality scenarios.
Those differences may be overestimated because of potential overestimation of the air
coefficient and underestimation of the intercept in the regression model. Examples of
such not-accounted-for variables are roadside dust Pb and historically contributions to
current levels of indoor dust Pb (e.g., dust Pb contributed to a house in the past that
continues to contribute to current dust Pb levels). See Sections 3.1.4.2 and 3.5.1.3 for
additional discussion.
• Estimates of outdoor soil/dust Pb concentrations: Outdoor soil/dust Pb concentrations
in all air quality scenarios have been set equal to the values for the current conditions
scenarios. That is, we are not simulating an impact of changes in air Pb concentrations
on soil concentrations, or the associated impact on dust concentrations, blood Pb and
risk estimates. In areas where air concentrations have been greater in the past,
however, implementation of a reduced NAAQS might be expected to yield reduced soil
Pb levels over the long term. As described in Section 2.3.3, however, there is
potentially uncertainty associated with this specification, particularly with regard to
implications for areas in which a Pb source may locate where one of comparable size
had not been previously. Additionally, we note that control measures implemented to
meet alternative NAAQS may result in changes to soil Pb concentrations; these are not
reflected in the assessment
• Recontamination of residential yard soil near the primary Pb smelter: Although data
collected in residential yards within % mile of the primary Pb smelter indicate a trend
of increasing surface soil Pb concentrations, soil Pb estimates in this area of the
primary Pb smelter case study were not adjusted using this information. Therefore,
these estimates likely were underestimates of soil Pb levels associated with the current
ambient air Pb levels. Because the indoor dust Pb model used in this part of this case
study did not rely on outdoor soil/dust concentrations this did not affect indoor dust Pb
estimates, although as described in a previous bullet the lack of soil data to consider in
developing the indoor dust Pb model increased uncertainty in that model's results.
Additionally, the outdoor soil/dust ingestion pathway is a small contributor to blood Pb
levels, which also reduced the impact of this factor on blood Pb estimates .
• Interindividual variability GSD: There is uncertainty associated with the GSD
specified for each case study. In the case of the general urban case study, the use of a
uniform ambient air Pb concentration across the hypothetical study area complicates
selection of a GSD. This is because any urban area for which GSDs can be developed
would be subject to spatial variation in ambient air Pb levels. To the extent that these
ambient air Pb levels influence children's blood Pb levels, this influence would be
reflected in the associated GSD. In the case of the two point source case studies, the
use of the spatial templates complicates selection of a GSD for a different reason.
Because the spatial templates are intended to contribute variability in risk results
related to spatial gradients in air-related Pb media concentrations, the GSDs used for
these case studies need to reflect the remaining sources of variability in blood Pb levels
(e.g., interindividual variability in behavior and biokinetics related to Pb exposure as
well as variability in non-air related Pb exposures). The extent to which the specified
GSDs reflect the sources of variability at play in each of these case studies is unknown.
4-16
-------
• Exposure pathway apportionment for higher percentile blood Pb level and IQ loss
estimates: As discussed in Section 3.2.2, pathway apportionment of blood Pb levels
for higher population percentiles is specified to be the same as that estimated using the
central tendency estimate of blood Pb in an exposure zone. This introduces uncertainty
into projections of pathway apportionment for higher population percentiles of blood
Pb and IQ loss. In reality, pathway apportionment may shift as you consider higher
exposure percentiles. For example, paint and/or drinking water exposures may
increase in importance, with air-related contributions decreasing as an overall
percentage of blood Pb levels and associated risk.
• Projection oflQ loss at lower exposure levels: Because available epidemiological data
are limited in their description of the relationship of IQ decrement to blood Pb at lower
blood Pb levels (e.g., < 5 |ig/dL), there is uncertainty associated with projecting IQ loss
at these blood Pb levels which are particularly relevant to the current population of
children in the U.S. (see discussion in Section 2.1.5). For additional discussion of
uncertainty associated with predicting IQ loss resulting from Pb exposure, see Sections
5.3.3.1.
4.3.2 Sensitivity Analysis
Sensitivity analysis techniques were used to evaluate the impact of individual modeling
elements on exposure and risk estimates. Specifically, we used a "one element at a time
elasticity analysis" approach, in which the full risk model was run with one of the selected
modeling elements adjusted to reflect an alternate input value or modeling choice. The results of
that run with the modified modeling element was then compared to those for the "baseline risk"
run to determine the magnitude of the impact on risk results of selections for that one modeling
element.
The sensitivity analysis described in Appendix L focused on the general urban case study,
reflecting the fact that this case study has potential relevance for a larger number of Pb-exposed
children compared with the two point source case studies. Additionally, we recognized that
exposure patterns for urban children can be highly variable compared with exposures near Pb
smelters and the availability of data sets for specific urban areas that reflect current blood Pb
levels is limited. The modeling elements examined in the sensitivity analysis included those
inputs and modeling steps believed to have a significant potential for impacting exposure and
risk results (e.g., oral uptake factor, interindividual blood Pb variability GSD, biokinetic model,
concentration-response function for IQ loss). The one-element-at-a-time sensitivity analysis
indicates which of the modeling elements included in the sensitivity analysis has the greatest
impact on risk results, which can be used to guide future efforts to refine the overall risk model.
Note, that an additional sensitivity analysis involving the hybrid indoor dust model and
specifically the derivation of the "other" component of indoor dust was conducted as part of the
core analysis (see Section 5.3.3.4).
4-17
-------
The results of the sensitivity analysis (Appendix L) can be considered both in terms of
the impact of individual modeling element choices on (a) overall risk results and (b) recent air
risk results, i.e., IQ loss estimates for inhalation plus ingestion of indoor dust Pb predicted to be
associated with ambient outdoor air Pb levels, where ambient outdoor air potentially includes
resuspended, previously deposited Pb (see Section 2.4.3). Given the relevance of the recent air
exposure pathways to the NAAQS review, the results of the sensitivity analysis are summarized
here in terms of their impact on recent air risk results, rather than total risk results. Results of the
sensitivity analysis for total risk estimates, however, are similar (Appendix L).
Results of the sensitivity analysis showed the following modeling elements to have the
greatest impact on recent air risk estimates for the general urban case study (i.e., >40% impact on
the 95th percentile recent air risk estimates and an even greater impact, in terms of percent
change, on median estimates). These elements are presented in order of decreasing magnitude of
impact on risk estimates (Note, results presented here reflect the impact on the 95th percentile
risk results - see Appendix L for sensitivity analysis results reflecting impact on the median
population percentile):
• IQ loss function: Use of the log-linear with linearization IQ loss model (the model
producing the greatest IQ loss) resulted in 226% increase in recent air risk results over
results for baseline which used the 2-piece linear model. These results suggest that
characterization of the relationship between Pb exposure and IQ loss has the greatest
impact on high-end risk results.
• Indoor dust Pb modeling: Use of a combination of high-end modeling parameters for
the hybrid dust model results in a 139% increase in recent air risk estimates over results
using the hybrid model with those input values used in the primary analysis. In
addition, use of the air-only regression model in place of the hybrid model results in a
60% decrease in the recent air risk results. These results together suggest that
predicting the relationship between outdoor ambient air Pb and indoor dust Pb is an
important factor impacting risk results.
• Blood Pb modeling: Use of the Leggett model as compared to the IEUBK model,
results in a 170% increase in recent air risk estimates. This indicates the importance of
the blood Pb modeling step in the analysis. However, these results need to be
considered in the context of the performance evaluation for blood Pb which suggested
that the Leggett model may be significantly overpredicting blood Pb levels, and the
fact that the Legget model is designed to provide results over shorter time frames (e.g.,
with daily time step), while the IEUBK model provides longer-term quasi steady-state
estimates.
• Monitor-based ratio of maximum quarterly to annual average Pb-TSP concentrations:
A ratio of monitored maximum quarterly average to annual average air Pb
concentrations is used in the general urban case study to translate maximum quarterly
average air concentrations into the annual average values used in the risk assessment.
4-18
-------
Use of the 95th percentile of urban Pb-TSP monitor ratios as compared to the mean
value, results in a 50% reduction in recent air risk estimates.
• GSD reflecting interindividual variability in bloodPb levels: Use of the larger GSD
(2.1) in place of the smaller value (1.7) resulted in a 40% increase in the recent air risk
estimates.
Alternate selections for the other modeling elements included in the sensitivity analysis
yielded changes to the 95th percentile risk results of less than 40% from baseline.
4.3.3 Performance Analyses
Performance evaluation for the exposure assessment (Section 3.5) focused on evaluation
of estimates of Pb in ambient air, outdoor soil, and indoor dust (discussed in Section 3.5.1) and
estimates of Pb in blood (covered in Section 3.5.2). Consideration of the results of performance
evaluation can provide insights into potential sources of uncertainty in an analysis, by identifying
those elements of the analysis that appear inconsistent with available empirical data. This can, in
turn, point to underlying bias or other errors associated with that particular modeling step,
reflecting either parameter or model uncertainty. This section identifies key findings of the
performance analysis describing the nature of associated uncertainty, including results which
either supported modeling elements or suggested increased uncertainty.
• Modeled ambient air Pb levels: The evaluation of the air model performance for the
primary Pb smelter case study (Sect 3.5.1.1) indicated performance generally
consistent with empirical data, increasing our confidence in air-related results
generated for this case study. Evaluation of air dispersion model performance for the
secondary Pb smelter suggested the potential for low bias in predictions of ambient air
Pb concentrations.
• Estimates of outdoor soil/dust Pb concentration for secondary Pb smelter case study:
The use of a combination of dispersion and soil mixing models to generate a spatial
pattern of concentrations combined with a scaling factor based on a surrogate location
contributes significant uncertainty to the soil Pb characterization for this case study.
Specifically, while the approach used is believed unlikely to significantly
underestimate soil Pb levels at this type of facility, the exact extent to which it is
representative of conditions at this specific location is not known.
• Modeled indoor dust Pb concentrations: Evaluation of the hybrid indoor dust model
used in the general urban case study suggested the potential for both under- and
overestimation. The mechanistic ambient air-related portion of the model may
underestimate that component of dust Pb, while the indoor dust loading-to-
concentration conversion algorithm may contribute to overestimate of dust Pb. It is not
know to what extent these two biases cancel out each other. Overall comparison of
indoor dust Pb concentrations generated for the three case studies against available
empirical data suggest that: (a) for the general urban case study, estimates fall within
the range of measured values from a national-scale study, adding confidence to the
estimates, (b) central tendency estimates for the primary Pb smelter are close to the
4-19
-------
lower end of the range for the national-scale dataset referenced above for the general
urban case study and high-end estimates seem to fit with available data near smelters,
and (c) comparison of estimates for the secondary Pb smelter against empirical data
suggest that these estimates may be biased low.
• Evaluation of candidate blood Pb models: A number of performance evaluations were
completed on the two candidate blood Pb models considered for this analysis (IEUBK
and Leggett), as well as the Lanphear empirical model. The results of these
performance evaluations, which included application of both models in replicating
national-scale child blood Pb levels (NHANES IV results) and blood Pb levels for an
urban child cohort, suggested that the Leggett model consistently overpredicted blood
Pb levels by a factor of 3 to 6, while IEUBK estimates were usually within a factor of
2. These findings, in addition to CASAC recommendations, resulted in our selecting
the IEUBK model as primary blood Pb model for this assessment, with the Leggett
model being reserved for application in the sensitivity analysis.
• Outdoor air Pb-to-bloodPb ratios: Three sets of outdoor air Pb-to-blood Pb ratios
were derived. These related outdoor ambient air Pb to blood Pb resulting from: (1) the
inhalation pathway only; (2) all recent air pathways (inhalation plus ingestion of indoor
dust Pb predicted to be associated with ambient air Pb levels, with ambient air
potentially including resuspended, previously deposited Pb); and (3) all recent and past
air pathways (see Section 2.4.3). All ratios were derived prior to application of the
GSD reflecting interindividual variability in blood Pb levels and therefore reflect
central tendency blood Pb levels and not high-end population percentiles. The
modeled ratios were compared to both empirical data and statistically derived ratios
based on a pooled analysis (Section 3.5.2.2). With the exception of the primary Pb
smelter case study recent air ratio using 95th percentile air concentration and the
general urban case study recent air ratios for the hybrid dust model, the ratios for recent
air contribution to concurrent blood Pb level (Table 3-10) generated for the three case
studies were reasonable and supported by available empirical data. As described in
Section 3.5.2.2, the exceptions relate to use in those case studies of the dust models
predicting the greatest influence of air concentration on indoor dust Pb among the set
of dust models used. These were the general urban case study hybrid dust model and
the site-specific regression dust model used in the remediation zone for the primary Pb
smelter case study. In both cases, the ratios based on these models were higher than
empirically derived ratios obtained from the literature. However, the literature
indicated a potential for higher ratios for locations with either lower ambient air Pb
levels or lower blood Pb levels, both conditions of which are present at the two case
studies. Therefore, the higher modeled ratios obtained using the site-specific indoor
dust models for the general urban and primary Pb smelter case studies do not
necessarily point to potential high bias in predicting blood Pb levels for the study
populations.
• Comparison of modeled blood Pb levels to nationally representative data: Our ability
to compare modeled blood Pb levels to empirical data was handicapped by a lack of
studies for populations of children directly comparable to those in the three case studies
(see Section 3.5.2.3). Therefore, we focused most of our evaluation of modeled blood
Pb levels on consideration of the national-level data obtained from NHANES IV. Note
4-20
-------
that while higher-end population percentiles for the three case studies might not be
reflected in a national-level dataset, blood Pb levels associated with background
exposures for all three case studies are similar to blood Pb levels in the national
dataset. Thus, we considered a comparison of modeled blood Pb levels for the three
case studies against a national-scale dataset for central tendency population percentiles
appropriate. In this comparison, we included the highest and lowest of the multiple
sets of blood Pb estimates for the general urban case study generated using the multiple
modeling approaches employed used for this case study. These two blood Pb
distributions bracketed the NHANES IV based distribution, which adds confidence to
blood Pb modeling conducted for this case study. Results for the primary Pb smelter
(full study area) were similar to the NHANES IV distribution across population
percentiles up to the 95th. This result is not surprising given the fact that exposure
modeling for the primary Pb smelter did not fully consider variability in background Pb
exposure (e.g., paint, diet, drinking water). Had background exposures been fully
considered, the degree of divergence between the case study and national-level
percentiles would likely have been much greater (with the primary Pb smelter
distribution exceeding the NHANES IV distribution at higher population percentiles).
Results for the subarea of the primary Pb smelter case study were consistently higher
than the NHANES IV distribution across all population percentiles, as is expected
given that the entire study area experiences elevated facility-related ambient air Pb
levels. Blood Pb levels for all population percentiles for the secondary Pb smelter are
lower than those for the national-scale dataset. As with the primary Pb smelter case
study, this likely reflects the fact that variability in background Pb exposures was not
fully considered in modeling this case study.
4.3.4 Uncertainty in Modeling Approaches - Multiple Sets of Results
For those more highly influential analytical steps for which it is not clear which model or
input would generate "best estimate" results, we have implemented multiple modeling
approaches (see Section 2.4.6.2). Because each of the case studies uses different modeling
approaches for some of the analytical steps such as the indoor dust modeling step, and these
approaches are associated with differing uncertainty, the identity and size of the areas of
uncertainty associated with each case study differs. The specific modeling approaches for each
case study and their elements are presented in Figure 2-3. For the general urban case study, two
different dust models and two GSDs were used as compared to one model and GSD for these
analytical steps in the two point source case studies. However, the same number of blood Pb
metrics and IQ loss functions are used for all three case studies.
Consideration of the range of risk results generated using the multiple modeling
approaches for each case study provides perspective on the combined effect of key sources of
uncertainty on risk results. The median and 95th percentile estimates associated with the
modeling approaches yielding the highest and lowest risk results for specific scenarios of the
three case studies are presented in Table 4-4.
4-21
-------
Table 4-4. Impact of multiple sources of uncertainty on risk results.
Case study
General
urban
(mean current
conditions)
Primary Pb
smelter
(current
NAAQS)
Secondary Pb
smelter
(current
conditions)
Modeling approach
Lowest risk: Dust Model (Air-only Regression-based),
GSD (1.7), blood Pb Metric (Concurrent), IQ Function
(Two-piece Linear)
Highest risk: Dust Model (Hybrid), GSD (2.0), blood Pb
Metric (Lifetime), IQ Function (Log-linear with
Linearization)
Lowest risk: blood Pb Metric (Concurrent), IQ Function
(Two-piece Linear)
Highest risk: blood Pb Metric (Lifetime), IQ Function
(Log-linear with Linearization)
Lowest risk: blood Pb Metric (Concurrent), IQ Function
(low-linear with outpoint)
Highest risk: blood Pb Metric (Lifetime), IQ Function
(Log-linear with Linearization)
IQ loss
(population
percentile)
Median
2
5
1
4
0
3
95tn
4
9
4
7
2
5
Percent difference
between highest
and lowest
approaches
Median
230%
280%
>300%
95tn
120%
80%
120%
The lack of a consistent pattern regarding the magnitude of differences between the
modeling approaches generating the lowest and highest risk results either across case studies, or
across population percentiles (Table 4-4), is due to different influences of various factors specific
to the modeling approaches for each case study. For example, the finding with regard to the 95th
percentile results that the greatest difference across approaches occurs for the secondary Pb
smelter case study (>300%) reflects the fact that the lowest predicted median blood Pb level for
this case study is below the cutpoint associated with one of the three concentration-response
functions (thereby resulting in zero predicted IQ loss). The finding for the 95th percentile results
that there is a greater difference for the general urban case study than the primary Pb smelter
case study is because of the larger range of GSDs applied in this case study (2.1 and 1.6).
In summary, results presented in Table 4-4 suggest that these key sources of uncertainty,
when acting in concert, can produce an impact on overall risk results on the order of a factor of 2
to 3.
4-22
-------
REFERENCES
Henderson, R. (2007a) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee, to
Administrator Stephen L. Johnson. Re: Clean Air Scientific Advisory Committee's (CASAC) Review of
the 2nd Draft Lead Human Exposure and Health Risk Assessments Document. September 27, 2007.
Henderson, R. (2007b) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee, to
Administrator Stephen L. Johnson. Re: Clean Air Scientific Advisory Committee's (CASAC) Review of
the 1st Draft Lead Staff Paper and Draft Lead Exposure and Risk Assessments. March 27, 2007.
Hornung, R. (2007a) Email message to David Svendsgaard, U.S. EPA. April 24, 2007.
Hornung R. (2007b) Email message to Zachary Pekar, U.S. EPA, May 1, 2007.
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 Robe, R. (2005) Low-level environmental Pb exposure and children's intellectual function: An
international pooled analysis. Environmental Health Perspectives. 113(7):894-899.
4-23
-------
5 ADDITIONAL ANALYSES
This chapter presents the additional analyses performed for the Pb NAAQS risk
assessment subsequent to the public meeting of the CASAC Pb Panel on August 28-29, 2007
(Henderson, 2007b). These analyses include reanalysis of the general urban and primary Pb
smelter case studies and new analyses for three location-specific urban case studies located in
Cleveland, Chicago and Los Angeles.1 The methods used in conducting these additional
analyses, as well as the resultant exposure and risk estimates, sensitivity analyses and uncertainty
characterization are presented in this chapter.
5.1 DESIGN OF EXPOSURE AND RISK ASSESSMENTS
This section presents an overview of the design for the additional analyses. Additional
details regarding the methods used in estimating exposure for both analyses can be found in
Section 5.2. Details on the concentration-response functions for IQ loss and methods used to
generate population risk estimates for these additional analyses can be found in Section 5.3.
5.1.1 Assessment Scenarios
The design of the assessment scenarios for the additional analyses is generally the same
as that for the initial analyses (see Section 2.3 and subsections). The main difference is the
addition of an alternate NAAQS scenario of 0.02 |ig/m3, as a maximum monthly average;
otherwise, the air quality scenarios for the additional analyses are the same as those included in
the initial analyses (Section 2.3.1). Specification of the air concentrations for the current
conditions location-specific urban case studies is summarized in Section 5.1.3 and described
more fully in Section 5.2.2.1 and Appendix O.
5.1.2 Analytical Approach
The general analytical approach employed for the initial analyses, described in Section
2.4, was also employed for the additional analyses, including the temporal aspects (Section
2.4.1), categorization of policy-relevant exposure pathways (Section 2.4.3), and the analytical
steps (Section 2.4.4). Additionally, the spatial scale and resolution aspects of the reanalyses for
the general urban and primary Pb smelter case studies were the same as for the initial analyses
(see Section 2.4.2). The spatial scale and resolution aspects of the location-specific urban case
studies are summarized in Section 5.1.3.
1 Due to limitations with the design of the secondary Pb smelter case study and the associated high level of
uncertainty noted in Sections 4.2 and 4.3.1,this case study was not part of the additional analyses described in this
chapter.
5-1
-------
A notable difference in the approach between the initial analyses and the additional
analyses is that the multiple modeling approaches described in Section 2.4.6.2 and implemented
for the initial analyses were narrowed down to a core modeling approach for the additional
analyses. This was done in consideration of CASAC comments on the initial analyses
(Henderson, 2007). Consistent with CASAC comments, the core modeling approach includes a
single indoor dust model for the general and location-specific urban case studies, focuses on the
concurrent blood Pb metric, and applies a single blood Pb GSD for the general urban case study.
Additionally, CASAC recommended use of a linear concentration-response function derived
from a subpopulation with peak blood Pb levels below 7.5 |ig/dL, and indicated less favor for the
two-piece linear function employed in the initial analyses of the full-scale assessment (Section
4.1.1.3). Accordingly, a different set of concentration-response functions was employed in the
additional analyses (see Section 5.3.1.1). Figure 5-1 identifies the core modeling approach for
each case study.
Figure 5-1. Core modeling approach for each case study.
Case Study
General Urban
Case Study
And
Location-
specific Urban
Case Studies
Primary Pb
Smelter Case
Study
Elements of modeling approaches
Indoor dust
modeling
1 model:
hybrid
mechanistic-
empirical
1 model:
statistical
(regression)
approach
Blood Pb metric
1 metric:
(a) concurrent
GSD
1 GSD,
representing
larger, regional
scale
1 GSD
Concentration-
response function
4 functions:
(a) log-linear with
low-exposure
linearization,
(b) log-linear with
cutpoint, and
(c) dual linear,
stratified at 7.5
Mg/dL
(d) dual linear,
stratified at 10 ug/dl_
Number of
sets of results
1*1*1*4=4
1 * 1 * 1 * 4 =4
5.1.3 Location-Specific Urban Case Studies
The primary design difference between the location-specific case studies and the general
urban case study, is that, unlike the spatially uniform general urban case study in which ambient
air Pb levels are uniform and fixed at the air concentration considered (see Section 2.2.1) both
the number of children and the ambient air Pb levels vary within the location-specific urban
study areas. As a result, although the data on which these location-specific urban case studies
are based are quite limited, exposure and risk estimates for these location-specific urban case
studies provide a better representation of some urban areas in the U.S. than do those for the
general urban case study.
5-2
-------
Locations were selected for these study areas based on consideration of the number and
spatial distribution of Pb-TSP monitors, the relative representation of source-oriented versus
other monitors and the relative Pb concentrations reported for those monitors (Appendix A,
Attachment A-2). Due to time constraints, the 2003-2005 dataset of Pb-TSP monitors analyzed
in Appendix A (Attachment A-2) was the focus for this selection process and for subsequent
consideration of air quality for the selected areas. Among all metropolitan areas represented in
this dataset, preference was given to urban areas with relatively more monitors having a more
uniform distribution and relatively higher Pb concentrations, and monitors that were not
considered source oriented (see Appendix A, Section A.2.2.2.3). This led to selection of areas in
Cleveland (6 monitors), Chicago (11 monitors) and Los Angeles (7 monitors).
The three study areas differ in several respects. First, they differ in Pb air levels, with the
highest reported for Cleveland and the lowest for Los Angeles. Further, while monitors in
Cleveland and Chicago included a subset classified as source oriented, this was not the case in
Los Angeles. Last, the source oriented monitors were segregated from the other monitors in
Chicago, while they were relatively more dispersed among the other monitors in Chicago.
The size of each study area is based on a polygon encompassing the area of each city in
which the monitors occur. That polygon is defined initially using those monitors as apices and
then a one mile-wide buffer is included beyond them in defining the final study area polygon
(Appendix O, Exhibits O-2 through O-4). The U.S. Census blocks whose centroids fall within
this polygon comprise the study area for each location-specific case study (Appendix O). Using
centroids as the basis for determining whether a Census block is inside or outside of the study
area provides a simple way for determining membership for those blocks that straddle the study
area boundary. The target population for each study area is those children less than 7 years of
age located within the study area blocks. The three study areas range in size from 67 to 1091
square km, and in target population size from 13,990 to 396,511 children (less than 7 years of
age).
Exposure zones within each study area were defined based on assignment of census
blocks to monitor sites, such that the number of exposure zones for a study area was equal to the
number of monitors in a study area. The total child count for each exposure zone reflects the
sum of children less than seven years old in the U.S. census blocks assigned to it, and the
ambient air Pb concentrations for that zone reflect characteristics of data for that monitor.
Consistent with the approach for all case studies, Pb concentrations in air (as well as other
media) are homogenous within each exposure zone.
Within each study area, Census blocks were assigned to exposure zones differently
depending on whether or not they were near a source oriented monitor. All blocks whose
centroids fall within a mile of a source oriented TSP monitor were assigned to a source oriented
5-3
-------
exposure zone defined by that monitor. All other blocks (i.e., those for which the centroid was
farther than a mile from a source oriented monitor) were assigned to the nearest non source
oriented monitor. This approach reflects the recognition that air Pb concentrations associated
with sizeable stationary sources exhibit sharper gradients with distance than air Pb
concentrations less influenced by such sources. Accordingly, source influenced Pb
concentrations are constrained to smaller exposure zones than non source influenced
concentrations.
Lead concentrations in indoor dust and outdoor soil/dust were estimated for the location-
specific urban case studies using specific elements of the core modeling approach applied in the
general urban case study (e.g., the indoor dust Pb model and Pb level in outdoor dust/soil).
Once the media concentrations are assigned to the exposure zones, the design of the
exposure and risk analyses for each of these location-specific urban case studies is similar to that
employed in the initial analyses to model the point source case studies (see Section 2.4.4.1 and
2.4.4.2), with a population-weighted simulation of exposure and risk conducted at the U.S.
Census block level.
5.2 EXPOSURE ASSESSMENT
5.2.1 Methods for General Urban and Primary Pb Smelter Case Studies
For estimating ambient air concentration, inhalation exposure concentration and Pb
concentration in outdoor soil/dust, the methods described in Section 3.1.1, 3.1.2 and 3.1.3 for the
general urban and primary Pb smelter case studies also apply to the additional analyses for those
case studies.
For the general urban case study, indoor dust Pb concentrations were estimated using the
hybrid model. The primary rationale for deemphasizing the air only regression model and
favoring the hybrid model in the additional analyses is that there is reduced confidence in using
the air only regression model in the urban residential context, since this statistical model was
developed with datasets that favored areas near Pb smelter sites (see Section 3.1.4.1 and
Appendix G.I.I). By contrast, the hybrid indoor dust model was developed specifically for
application in the urban residential context.
For the primary Pb smelter case study, indoor dust Pb concentrations are estimated as
described in Section 3.1.4.2.
Blood Pb concentrations for the additional analyses for the general urban and primary Pb
smelter case studies were estimated using methods previously described in Section 3.2, with two
distinctions. First, for these additional analyses, the core modeling approach employed included
only the concurrent blood Pb metric, consistent with CASAC recommendations (Henderson,
2007b). Secondly, for the general urban case study core approach we used the GSD of 2.1 which
5-4
-------
represents the GSD for the concurrent blood Pb metric identified from NHANES IV (Section
3.2.3.1). This GSD is consistent with the CASAC recommendation to use a GSD more
descriptive of the variability in blood Pb levels reported in NHANES-IV, as compared with the
smaller GSDs observed for more localized populations near specific Pb emitting industries (e.g.,
see Section 3.2.3). The rationale for selecting the larger GSD, is that larger urban areas are
likely to include more heterogeneous populations with regard to Pb exposure and blood Pb levels
that more closely resemble the national NHANES population than a smaller more homogeneous
population. For the primary Pb smelter case study, which includes a more localized study
population, we have retained the smaller GSD of 1.7 used in the initial analyses (see Section
3.2.3.2).
The remaining elements of the core analysis, related to blood Pb modeling, including the
blood Pb model, characterization of background Pb exposures, apportionment of exposure
pathway contributions and probabilistic population modeling are those described in Section 3.2.
5.2.2 Methods for Location-specific Urban Case Studies
The methods used to estimate media and blood Pb concentrations in the location-specific
urban case studies are described in the following subsections. The primary difference in
methods between the location-specific case studies and the general urban case study is the
inclusion of multiple exposure zones per study area across which numbers of children and
ambient air Pb levels vary.
5.2.2.1 Ambient Air and Inhalation Exposure Concentrations
Air quality scenarios evaluated for these case studies included: (a) a current conditions
scenario based on monitor-specific Pb-TSP statistics for the 2003-2005 dataset (Appendix A,
Attachment A-2), (b) a current NAAQS scenario based on a roll-up of Pb-TSP values to the
current NAAQS level, and (c) alternative NAAQS scenarios for levels falling below the current
conditions levels for each case study.
For each scenario, the annual average ambient air Pb concentration estimate for each
exposure zone that is the input to the blood Pb model for that exposure zone is based on the
statistics for the monitor associated with that exposure zone, using the monitor-specific statistics
presented in Appendix A, Attachment A-2. How this is derived varies with the air quality
scenario as described below.
• Current conditions scenario: Each exposure zone is assigned the Pb-TSP values
derived from the analysis of the 2003-2005 dataset in Appendix A (Attachment A-2)
for the monitor assigned to that zone. Specifically, the annual average ambient Pb
concentration estimate that is the input to the blood Pb model for a given exposure
5-5
-------
zone is the three-year annual average concentration associated with the Pb-TSP
monitor assigned to that zone.
• Current NAAQS scenario: A proportional roll-up procedure is used to derive air
concentrations for all exposure zones such that the highest maximum quarterly average
concentration among all of the study area monitors just meets the current NAAQS
(maximum quarterly average of 1.5 |ig/m3).
— For the Cleveland and Los Angeles study areas, this procedure involves adjusting
up the maximum quarterly average estimates for all monitors by the same factor as that
used to set the highest monitor equal to the current NAAQS (max quarterly value of 1.5
|ig/m3). Then those adjusted maximum quarterly values are converted into annual
averages using the ratio of annual average to maximum quarterly average specific to
each monitor (as presented in Appendix A, Attachment A-2).
— For Chicago, given the spacing of the monitors, we have defined two independent
roll-up subareas, each represented by a separate set of monitors. For this scenario, the
highest maximum quarterly average value in each of the two subareas is rolled up to
equal the current NAAQS and the other monitors in each subarea are increased by the
same proportion. Then the annual average values are calculated using monitor-specific
ratios as for the other study areas.
• Alternative NAAQS scenarios: The procedure for rolling back ambient air Pb levels to
reach alternative NAAQS is similar to the roll-up procedure used for the current
NAAQS scenario, except the values are reduced rather than increased. Only those
alternative NAAQS that are exceeded by the current conditions for a given case study
were evaluated for that case study. Consequently, several of the higher alternative
NAAQS were not evaluated at all for the location-specific urban case studies. As for
the roll-up for the current NAAQS scenario, roll-backs for the Chicago case study are
implemented separately for each of the two independent subareas, based on the highest
Pb-TSP monitor in each subarea.
Inhalation exposure air concentrations were defined for both source oriented and non
source oriented exposure zones using the same procedure employed for the general urban case
study (see Section 3.1.2).
5.2.2.2 Other media Concentrations
Characterization of Pb exposure from incidental outdoor soil/dust ingestion used the
values used for the general urban case study (Section 3.1.3.1), with the same values being used to
characterize conditions in all exposure zones (source oriented or not) in each of the three case
studies.
Indoor dust Pb concentrations are estimated for the location-specific urban case studies
using the hybrid indoor dust model (see Section 3.1.4.1). Specifically, the hybrid indoor dust
model is combined with the annual average ambient air Pb level for a particular exposure zone
and air quality scenario to generate an indoor dust Pb level for that zone. Therefore, while
background Pb exposure levels and outdoor soil/dust Pb levels are fixed across each of the urban
5-6
-------
case studies, both ambient air Pb levels and indoor dust Pb levels are allowed to vary across
source and non source exposure zones.
5.2.2.3 Blood Pb levels
Estimates of concurrent blood Pb are derived from application of the IEUBK model as
described in Appendix H, with characterization of Pb exposure from background sources (i.e.,
drinking water and diet) relying on the same values as for all case studies (Appendix H). The
apportionment of the blood Pb levels among the various exposure pathways is described in
Section 3.2.2.
Population-level exposure modeling for the location-specific urban case studies utilizes
essentially the same population-weighted sampling procedure as was used for the point source
case studies (Section 3.2.2.2, including Figure 3-1). Use of this more sophisticated, population-
weighted, multizone approach is required for the location-specific urban case studies since they
utilize multiple exposure zones with differing demographic data and ambient air Pb levels,
negating the use of the simpler, single zone approach used in the general urban case study. The
remaining elements of the blood Pb modeling approach used for the location-specific urban case
studies are identical to those used for the core approach for the general urban case study (Section
5.2.1). These include: (a) use of the concurrent blood Pb metric, (b) use of the IEUBK blood Pb
model and (c) use of the GSD of 2.1 to represent interindividual variability in blood Pb levels.
5.2.3 Media concentrations
The following subsections summarize the media concentration estimate for all case
studies and air quality scenarios. Estimates presented in these subsections are presented to three
(for air) or zero (for dust and soil) decimal places, which results in various numbers of implied
significant figures. This is not intended to convey greater precision for some estimates than
others; it is simply an expedient and initial result of the software used for the calculation.
5.2.3.1 Location-specific Urban Case Studies
This section summarizes the media concentration estimates for the core modeling
approach for each air quality scenario for the location-specific urban case studies. The complete
set of media concentration estimates for these case studies is presented in Appendix O.
As discussed in Section 2.3.3, Pb concentration in outdoor soil/dust is not changed with
the alternate air quality scenarios. Rather, outdoor soil/dust concentration is held constant at the
current conditions or current NAAQS level.
Media concentrations generated for the three location-specific urban case studies are
summarized in Tables 5-1 through 5-4, for annual ambient air Pb concentrations, inhalation
exposure concentrations, outdoor soil/dust Pb and indoor dust Pb concentrations, respectively.
5-7
-------
5.2.3.2 General Urban and Primary Pb Smelter Case Studies
This section summarizes the media concentration estimates for the core modeling
approach for each air quality scenario for the general urban and primary Pb smelter case studies.
The complete set of media concentration estimates for these case studies are presented in
Appendix N.
For each air quality scenario for the primary Pb smelter case study, a range of percentile
estimates are presented for each exposure medium. For the general urban case study, however,
only a single value is presented for each exposure medium. This reflects the fact that, while the
primary Pb smelter case study is modeled using spatial templates that include a large number of
U.S. Census blocks and/or block groups (allowing percentile media concentrations to be
identified), the general urban case study is modeled using a single study area with uniform media
concentrations. Consequently, there is only one value presented for the general urban case study
for each medium in each air quality scenario.
As mentioned above, it is assumed that outdoor soil/dust Pb concentrations are not
effected by changes in ambient air Pb levels. In the case of an area such as the remediation zone
of the primary Pb smelter case study, however, where soil dynamics have been changed by the
substitution of contaminated soil with clean soil, or in areas where local sources may pose a
more significant source to outdoor soil/dust than historic sources - and where there may be a
currently increasing trend in surface Pb concentration - this may underestimate soil
concentrations under some alternate NAAQS.
Media concentrations generated for the primary Pb smelter and general urban case studies
using the core modeling approach are summarized in Tables 5-1 through 5-4, for annual ambient
air Pb concentrations, inhalation exposure concentrations, outdoor soil/dust Pb and indoor dust
Pb concentrations, respectively. As with similar presentations in Chapter 3, the percentiles
presented here are from population weighted distributions of the media concentrations (see
description in Appendices N, O and P).
5-8
-------
Table 5-1. Estimated annual average ambient air concentrations.
Statistic*
Average Annual Air Pb Concentration (ug/m3
Current
Conditions
Current
NAAQS
Alternative NAAQS
1
0.2 ug/m3, Max
Quarterly
2
0.5 ug/m3,
Max Monthly
3
0.2 ug/m3,
Max
Monthly
4
0.05 ug/rn3,
Max
Monthly
5
0.02 ug/rn3,
Max
Monthly
Location Specific Case Study(Chicago)
Maximum
95lhpercentile
Median
5th percentile
Minimum
0.040
0.040
0.027
0.014
0.011
0.809
0.809
0.428
0.286
0.225
0.040
0.040
0.027
0.014
0.011
0.040
0.040
0.027
0.014
0.011
0.040
0.040
0.021
0.014
0.011
0.010
0.010
0.005
0.004
0.003
0.004
0.004
0.002
0.001
0.001
Location Specific Case Study(Cleveland)
Maximum
95th percentile
Median
5th percentile
Minimum
0.121
0.121
0.021
0.017
0.017
0.506
0.506
0.085
0.071
0.071
0.067
0.067
0.011
0.009
0.009
0.108
0.108
0.018
0.015
0.015
0.043
0.043
0.007
0.006
0.006
0.011
0.011
0.002
0.002
0.002
0.004
0.004
0.001
0.001
0.001
Location Specific Case Study(Los Angeles)
Maximum
95th percentile
Median
5th percentile
Minimum
0.022
0.022
0.019
0.015
0.006
0.360
0.360
0.300
0.239
0.091
0.022
0.022
0.019
0.015
0.006
0.022
0.022
0.019
0.015
0.006
0.022
0.022
0.019
0.015
0.006
0.009
0.009
0.007
0.006
0.002
0.003
0.003
0.002
0.002
0.001
General Urban Case Study
NA- Single
Study Area
High-end:
0.114
Mean:
0.056
0.600
0.080
0.125
0.050
0.013
0.005
Primary Pb SmelterCase Study- full study area
Maximum
95th percentile
Median
5th percentile
Minimum
NA
0.740
0.153
0.042
0.015
0.006
0.161
0.033
0.009
0.003
0.001
0.326
0.067
0.019
0.007
0.003
0.130
0.027
0.007
0.003
0.001
0.033
0.007
0.002
0.001
< 0.001
0.013
0.003
0.001
< 0.001
< 0.001
Primary Pb SmelterCase Study- 1.5 km sub area
Maximum
95th percentile
Median
5th percentile
Minimum
*- The percentiles
NA
presented h
0.740
0.675
0.238
0.137
0.098
sre are from p
0.161
0.147
0.052
0.030
0.021
opulation weighted
0.326
0.297
0.105
0.060
0.043
distributions of th
0.130
0.119
0.042
0.024
0.017
le media cone
0.033
0.030
0.011
0.006
0.004
entrations.
0.013
0.012
0.004
0.002
0.002
5-9
-------
Table 5-2. Estimated inhalation exposure concentrations.
Statistic*
Average Annual Inhalation Exposure Concentration of Pb (g/m3)
Current
Conditions
Current
NAAQS
Alternative NAAQS
1
0.2 ug/m3, Max
Quarterly
2
0.5 ug/m3,
Max
Monthly
3
0.2 ug/m3,
Max
Monthly
4
0.05 ug/rn3,
Max Monthly
5
0.02 ug/rn3,
Max Monthly
Location Specific (Chicago)
Maximum
95m
percentile
Median
5th percentile
Minimum
0.017
0.017
0.012
0.006
0.005
0.347
0.347
0.184
0.123
0.097
0.017
0.017
0.012
0.006
0.005
0.017
0.017
0.012
0.006
0.005
0.017
0.017
0.009
0.006
0.005
0.004
0.004
0.002
0.002
0.001
0.002
0.002
0.001
0.001
< 0.001
Location Specific (Cleveland)
Maximum
95m
percentile
Median
5th percentile
Minimum
0.052
0.052
0.009
0.007
0.007
0.217
0.217
0.037
0.030
0.030
0.029
0.029
0.005
0.004
0.004
0.047
0.047
0.008
0.006
0.006
0.019
0.019
0.003
0.003
0.003
0.005
0.005
0.001
0.001
0.001
0.002
0.002
< 0.001
< 0.001
< 0.001
Location Specific (Los Angeles)
Maximum
95m
percentile
Median
5th percentile
Minimum
0.010
0.010
0.008
0.006
0.002
0.154
0.154
0.129
0.103
0.039
0.010
0.010
0.008
0.006
0.002
0.010
0.010
0.008
0.006
0.002
0.010
0.010
0.008
0.006
0.002
0.004
0.004
0.003
0.002
0.001
0.001
0.001
0.001
0.001
< 0.001
General Urban
NA- Single
Study Area
High-end: 0.049
Mean: 0.024
0.258
0.034
0.054
0.021
0.005
0.002
Primary Pb Smelter- full study area
Maximum
95tn
percentile
Median
5th percentile
Minimum
NA
0.310
0.064
0.017
0.006
0.002
0.067
0.014
0.004
0.001
< 0.001
0.136
0.028
0.007
0.003
0.001
0.055
0.011
0.003
0.001
< 0.001
0.014
0.003
0.001
< 0.001
< 0.001
0.005
0.001
< 0.001
< 0.001
< 0.001
Primary Pb Smelter- 1.5 km subarea
Maximum
95m
percentile
Median
5th percentile
Minimum
*- The percen
NA
lies presented h
0.310
0.282
0.100
0.057
0.041
sre are from p
0.067
0.061
0.022
0.012
0.009
opulation weight
0.136
0.124
0.044
0.025
0.018
ed distribution
0.055
0.050
0.018
0.010
0.007
s of the media
0.014
0.012
0.004
0.003
0.002
concentrations.
0.005
0.005
0.002
0.001
0.001
5-10
-------
Table 5-3. Estimated outdoor soil/dust concentrations.
Statistic*
Projected Average Outdoor Soil/Dust Pb
Concentration (mg/kg)
(Same for all air quality scenarios)
Location Specific (Chicago)
NA - Full Study Area
198
Location Specific (Cleveland)
NA - Full Study Area | 198
Location Specific (Los Angeles)
NA - Full Study Area
198
General Urban
NA- Single Study Area
198
Primary Pb Smelter- full study area
Maximum
95thpercentile
Median
5th percentile
Minimum
958
245
85
30
17
Primary Pb Smelter- 1.5km sub area
Maximum
95th percentile
Median
5th percentile
Minimum
"-The percentiles presented he
media concentrations.
294
223
150
42
42
re are from population weighted distributions of the
5-11
-------
Table 5-4. Estimated indoor dust concentrations.
Statistic*
Projected Average Indoor Dust Pb Concentration (mg/kg or ppm)
Current
Conditions
Current
NAAQS
Alternative NAAQS
1
0.2 ug/m3, Max
Quarterly
2
0.5 ug/m3,
Max
Monthly
3
0.2 ug/m3,
Max
Monthly
4
0.05 ug/rn3,
Max Monthly
5
0.02 ug/rn3,
Max Monthly
Location Specific (Chicago)
Maximum
95m
percentile
Median
5th percentile
Minimum
128
128
111
91
86
491
491
363
300
269
128
128
111
91
86
128
128
111
91
86
128
128
103
91
86
84
84
74
70
68
71
71
67
65
64
Location Specific (Cleveland)
Maximum
95m
percentile
Median
5th percentile
Minimum
203
203
101
96
96
392
392
174
161
161
158
158
86
82
82
193
193
98
93
93
132
132
78
75
75
85
85
66
65
65
72
72
63
63
63
Location Specific (Los Angeles)
Maximum
95m
percentile
Median
5th percentile
Minimum
104
104
99
92
75
334
334
307
276
179
104
104
99
92
75
104
104
99
92
75
104
104
99
92
75
81
81
78
75
67
68
68
67
65
63
General Urban
NA- Single
Study Area
High-end: 198
Mean: 146
426
169
206
140
88
73
Primary Pb Smelter- full study area
Maximum
95tn
percentile
Median
5th percentile
Minimum
NA
1944
219
84
53
41
648
152
68
45
38
1077
172
73
47
39
557
149
67
44
38
383
138
63
43
38
381
120
62
42
38
Primary Pb Smelter- 1.5km subarea
Maximum
95m
percentile
Median
5th percentile
Minimum
*- The percen
NA
lies presented h
1944
1819
860
578
453
sre are from p
648
606
287
193
151
opulation weight
1077
1008
477
320
251
ed distribution
557
521
246
166
130
s of the media
205
192
91
61
60
concentrations.
106
99
60
60
60
5-12
-------
5.2.4 Blood Pb levels
Estimates of concurrent blood Pb level derived from outputs of the IEUBK model
(Section 3.2) have been developed for each air quality scenario in each case study (see
Appendices N and O). Table 5-5 presents median concurrent blood Pb estimates for the
location-specific urban case studies (Chicago, Cleveland and Los Angeles) and the core analysis
of the primary Pb smelter (and sub-area) and the general urban case study. Table 5-6 presents
95th percentile concurrent blood Pb estimates for the same set of case studies. As with blood Pb
level results presented in Section 3.4, these tables present the median and 95th percentile total Pb
levels (i.e., simulated children for each case study area ranked based on total Pb level in order to
identify population percentiles), along with the blood Pb contributions to those total blood
estimates from policy-relevant pathways.2 While some of the blood Pb levels presented in
Section 3.4 (those for the primary Pb smelter) included a range reflecting multiple modeling
options related to blood Pb estimation, the results presented in Tables 5-5 and 5-6 are single
values, reflecting the fact that the core analysis, while including multiple concentration-response
functions for IQ loss, did not include multiple modeling options related to blood Pb estimation
(see Section 5.1.2).
It is noted that given the various limitations of our modeling tools (Sections 2.4.3 and
3.2.2), blood Pb levels associated with air-related exposure pathways and current levels of Pb
emitted to the air (including via resuspension) are likely to fall between the estimates for "recent
air" and those for "recent" plus "past air". Additionally, with regard to the location-specific
urban case studies "recent air" and "recent" plus "past air" categories, an artifact of the hybrid
dust Pb model tends to mask trends in the two components of dust Pb (recent air and other),
which contribute to "recent air" and "past air" estimates, respectively (see Section 4.3.1
discussion of this uncertainty). For the primary Pb smelter case study, uncertainty in parsing out
the "recent air" and "other" components of indoor dust (specifically for the site-specific
regression model used in the remediation zone) have led us to conclude that only "recent plus
past air" exposures should be presented and "recent air" results should not be separately
presented for the primary Pb smelter, as is done for the other case studies (see Section 3.1.4.2).
2 Seemingly inconsistent trends across air quality scenarios seen in these two categories (recent air and
recent plus past air) may be a result of presenting results for these policy-relevant categories based on population-
weighted ranking of total blood Pb.
5-13
-------
Table 5-5. Summary of blood Pb level estimates for median total blood Pb.
Air Quality Scenario
(and case study)
Location-specific (Chicago)
Current NAAQS (1.5 ug/rrr3, max quarterly)
Current conditions
(0.14 ug/m3 max quarterly; 0.31 ug/m3 max monthly)
Alternative NAAQS (0.2 ug/nT3, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
Alternative NAAQS (0.02 ug/nT3, max monthly)
Concurrent Blood Pb Level
Recent
1.9
0.4
0.3
0.1
<0.1
Recent plus
Past"
2.5
1.2
1.2
1.1
1.0
Total Pb
Exposure
3.0
1.8
1.8
1.6
1.6
Location-specific (Cleveland)
Current NAAQS (1.5 ug/rrr3, max quarterly)
Current conditions
(0.36 ug/m3 max quarterly; 0.56 ug/m3 max monthly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/mj, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
Alternative NAAQS (0.02 ug/nr3, max monthly)
1.2
0.3
0.7
0.2
0.1
<0.1
<0.1
1.8
1.2
1.3
1.1
1.1
1.0
1.0
2.1
1.8
1.8
1.7
1.7
1.6
1.6
Location-specific (Los Angeles)
Current NAAQS (1.5 ug/nT3, max quarterly)
Current conditions
(0.09 ug/m3 max quarterly; 0.17 ug/m3 max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
Alternative NAAQS (0.02 ug/nT3, max monthly)
1.3
0.3
0.2
0.1
2.1
1.2
1.1
1.0
2.6
1.7
1.6
1.6
General urban
Current NAAQS (1.5 ug/rrr3, max quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Current conditions -high-end (0.87 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/m , max quarterly)
Current conditions - mean (0.14 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/mj, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
Alternative NAAQS (0.02 ug/nr3, max monthly)
1.9
0.9
0.8
0.7
0.6
0.5
0.2
0.1
2.6
1.6
1.6
1.5
1.4
1.4
1.1
1.1
3.1
2.2
2.1
2.0
1.9
1.9
1.7
1.6
Primary Pb smelter - full study area
Current NAAQS (1.5 ug/nT3, max quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/nr3, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
Alternative NAAQS (0.02 ug/mj, max monthly)
MAC
0.8
1.1
0.7
0.5
0.7
0.9
1.5
1.4
1.4
1.4
1.4
1.4
Primary Pb smelter - 1.5km study area
Current NAAQS (1.5 ug/nr3, max quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/nT3, max monthly)
Alternative NAAQS (0.05 ug/mj, max monthly)
Alternative NAAQS (0.02 ug/nT3, max monthly)
NAC
4.0
2.6
1.8
1.8
1.1
1.1
4.6
3.2
2.5
2.3
1.7
1.6
a - Estimates are rounded to one decimal place.
b -The term "past air" includes contributions from the outdoor soil/dust contribution to indoor dust, historical air
contribution to indoor dust, and outdoor soil/dust pathways, while "recent air" refers to contributions from inhalation of
ambient air Pb or ingestion of indoor dust Pb predicted to be associated with outdoor ambient air Pb levels, with
outdoor ambient air also potentially including resuspended, previously deposited Pb (see Section 2.4.3).
c- "Recent air" estimates were not developed for the primary Pb smelter case study (see Section 3.1.4.2).
5-14
-------
Table 5-6. Summary of blood Pb level estimates for high-end total blood Pb.
Air Quality Scenario
(and case study)
Concurrent Blood Pb Level
Recent
Recent plus
Past"
Total Pb
Exposure
Location-specific (Chicago)
Current NAAQS (1.5 ug/rrr3, max quarterly)
Current conditions
(0.14 ug/m3 max quarterly; 0.31 ug/m3 max monthly)
Alternative NAAQS (0.2 ug/nT3, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
Alternative NAAQS (0.02 ug/nT3, max monthly)
6.5
1.1
1.5
0.3
0.2
8.7
4.1
4.2
3.6
3.6
10.2
6.0
6.0
5.5
5.4
Location-specific (Cleveland)
Current NAAQS (1.5 ug/rrr3, max quarterly)
Current conditions
(0.36 ug/m3 max quarterly; 0.56 ug/m3 max monthly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/mj, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
Alternative NAAQS (0.02 ug/nr3, max monthly)
2.3
0.9
1.5
0.7
0.4
0.1
0.1
5.4
4.1
4.2
3.9
3.8
3.5
3.5
7.4
6.1
6.0
5.8
5.7
5.4
5.3
Location-specific (Los Angeles)
Current NAAQS (1.5 ug/nT3, max quarterly)
Current conditions
(0.09 ug/m3 max quarterly; 0.17 ug/m3 max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
Alternative NAAQS (0.02 ug/nT3, max monthly)
4.6
1.1
0.5
0.2
7.1
4.0
3.7
3.5
8.9
5.9
5.5
5.4
General urban
Current NAAQS (1.5 ug/rrr3, max quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Current conditions -high-end (0.87 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/m , max quarterly)
Current conditions - mean (0.14 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/mj, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
Alternative NAAQS (0.02 ug/nr3, max monthly)
6.4
2.9
2.7
2.3
1.9
1.7
0.7
0.4
8.8
5.6
5.4
5.0
4.7
4.6
3.9
3.7
10.6
7.4
7.2
6.8
6.5
6.4
5.7
5.5
Primary Pb smelter - full study area
Current NAAQS (1.5 ug/nT3, max quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/nr3, max monthly)
Alternative NAAQS (0.05 ug/nT3, max monthly)
Alternative NAAQS (0.02 ug/mj, max monthly)
NAC
2.8
3.1
2.4
2.5
1.9
3.2
4.6
4.2
4.0
4.0
3.8
3.8
Primary Pb smelter - 1.5km study area
Current NAAQS (1.5 ug/nr3, max quarterly)
Alternative NAAQS (0.5 ug/nT3, max monthly)
Alternative NAAQS (0.2 ug/nT3, max quarterly)
Alternative NAAQS (0.2 ug/nT3, max monthly)
Alternative NAAQS (0.05 ug/mj, max monthly)
Alternative NAAQS (0.02 ug/nT3, max monthly)
NAC
10.3
7.6
5.9
4.9
3.5
3.0
12.3
8.5
6.6
6.1
4.5
4.2
a - Estimates are rounded to one decimal place.
b -The term "past air" includes contributions from the outdoor soil/dust contribution to indoor dust, historical air
contribution to indoor dust, and outdoor soil/dust pathways, while "recent air" refers to contributions from inhalation of
ambient air Pb or ingestion of indoor dust Pb predicted to be associated with outdoor ambient air Pb levels, with
outdoor ambient air also potentially including resuspended, previously deposited Pb (see Section 2.4.3).
c- "Recent air" estimates were not developed for the primary Pb smelter case study (see Section 3.1.4.2).
5-15
-------
5.2.5 Uncertainty Characterization
As was done for the initial analyses (see Section 3.5), characterization of uncertainty
related to the exposure assessment for the additional analyses focused on performance evaluation
of both modeled media concentrations and modeled blood Pb levels. The sensitivity analysis
described in Section 4.3.2 included several modeling elements related to exposure assessment
which have bearing on blood Pb levels generated for the analyses described in this chapter for
the primary Pb smelter and general urban case study and the location-specific urban case studies
(e.g., interindividual variability GSD, blood lead model selection). Additional key sources of
uncertainty related to exposure assessment specifically for the location-specific urban case
studies (e.g., relating child study populations within the study areas to Pb-TSP monitor values in
order to assign ambient air Pb levels) have not been investigated quantitatively, but are addressed
as part of the qualitative discussion of uncertainty for the overall risk assessment (Section
5.3.3.1).
Performance evaluation completed for both the modeled media concentrations and
modeled blood Pb levels is discussed below.
5.2.5.1 Performance Evaluation of Modeled Media Concentrations
As discussed in Section 3.5.1 for the previous analyses, performance evaluation for
modeled media concentrations focuses on those estimates that were generated primarily through
modeling. Media concentrations characterized using empirical data, such as air concentrations
estimated using monitor measurements, are not subjected to performance evaluation. The
following observations can be made regarding performance evaluation of model-estimate media
concentrations for the analysis using the core modeling approach of the primary Pb smelter,
general urban and location-specific urban case studies.
• All media concentrations for the primary Pb smelter and general urban case study are
presented in Chapter 3, and the reader is referred back to Section 3.5.1 for performance
evaluation of those estimates.
• Ambient air Pb levels (and, indirectly, inhalation exposure concentrations) for the
location-specific urban case studies are based on Pb-TSP monitoring data and therefore
will not be addressed quantitatively in this performance evaluation. Note, however,
that the method for assigning Pb-TSP monitors to subsets of the child study
populations in each study area is a key step in this analysis and one which is subject to
uncertainty. However, characterizing ambient air Pb levels for the study populations at
the location-specific urban case studies is addressed as part of the qualitative
discussion of key sources of uncertainty (see Section 5.3.3.1).
• Outdoor soil/dust Pb levels for the location-specific urban case studies utilize the same
uniform value as was used for the general urban case study (see Section 3.5.1.2).
5-16
-------
• Indoor dust Pb levels for the location-specific urban case studies are generally similar
to those generated for the general urban case study. This is expected since the location-
specific urban case studies utilize the same hybrid indoor dust Pb model as was used
for the general urban case study. Because of the similarity of the indoor dust Pb levels
generated for the location-specific urban case studies and the general urban case study,
the reader is referred to the previous performance evaluation discussion for indoor dust
Pb addressing the general urban case study (see Section 3.5.1.3).
5.2.5.2 Performance Evaluation of Modeled Blood Pb Levels
In this section, we discuss air-to-blood ratios derived by comparing changes (deltas) in
median total blood Pb level (concurrent) to associated changes in annual average air Pb levels as
you step to the next lowest air quality scenario (Table 5-7). This analysis is only available for
the general urban and primary Pb smelter case studies (including both the full study area and 1.5
km subarea for the latter). Note that these air-to-blood ratios are different than those presented in
Section 3.5.2.2 for the general urban, primary Pb smelter and secondary Pb smelter case studies.
The ratios derived in Section 3.5.2.2 are based on comparing annual average ambient air Pb
levels (used in modeling these case studies) to the recent air portions of blood Pb levels.
Therefore, these ratios reflect a generalization of the relationship between air Pb and blood Pb
across the entire exposure spectrum, reflecting that relationship as an average of the two absolute
values. By contrast, the air-to-blood ratios presented in this section look at the incremental
change in blood Pb as ambient air Pb changes incrementally with lower alternate NAAQS levels
(Table 5-7).
The model-derived air-to-blood ratios presented in Table 5-7 can be compared to ratios
identified from the literature (see last column of Table 3-24). In fact, conceptually, it may be
more appropriate to compare the air-to-blood ratios presented in Table 5-7 (than those in Section
3.5.2.2) to ratios obtained from the review by Brunekreef (1984), since the individual ratios
summarized in that analysis included a large number of studies that compared blood Pb levels of
populations living in different locations with that associated different ambient air Pb levels to
derive air-to-blood ratios. Conceptually, this is fairly similar to the approach used in deriving the
air-to-blood ratios presented in this section. Similarly, Hilts (2003) considered exposure levels
(ambient air Pb) and associated blood Pb levels for the same population at two different points in
time, straddling a change in operation of a nearby smelter. This is conceptually similar to the
way the air-to-blood ratios are calculated here. Comparison of the air-to-blood ratios presented
in Table 5-7 with the ratios identified in Brunekreef (1984) and calculated based on data
presented in Hilts (2003), as summarized in Table 3-24, results in the following observations:
• Air-to-blood ratios for the general urban case study for air quality scenarios ranging
from current NAAQS down to the alternative NAAQS of 0.2 |ig/m3, maximum
monthly average, are similar to ratios reported in the literature (i.e., between 1:3 and
5-17
-------
1:5). Ratios for the two lowest alternative NAAQS are larger, ranging from 1:6 to 1:9.
However, even these larger ratios are within the range of some of the studies reported
in Brunekreef (1984) and within the range of ratios calculated based on data presented
in Hilts (2003). As noted in Section 3.5.2.2, there are a number of factors which may
produce higher air-to-blood ratios, which may pertain to urban locations, such as lower
blood Pb levels relative to those reflected in the meta analysis by Brunekreef (1984).
• Air-to-blood ratios for the primary Pb smelter case study (full study area) are similar in
magnitude to the general urban case study and the points made in the previous bullet
apply to this case study.
• Air-to-blood ratios for the 1.5 km subarea of the primary Pb smelter are significantly
higher than ratios for the other two study areas noted above, with values ranging from
1:10 to 1:19. These ratios, and particularly the higher values associated with the lower
NAAQS levels, are significantly higher than values reported in Brunekreef (1984) and
those based on data in Hilts (2003). Many of the studies cited in Brunekreef (1984) use
differences in average ambient air Pb levels and blood Pb levels between clusters of
individuals living in different areas to derive their ratios. It is possible that the indoor
dust Pb model developed for the remediation zone of the primary Pb smelter case study
captures a relatively stronger relationship between ambient air Pb and indoor dust Pb
(and hence blood Pb) which exists in the vicinity of industrial facilities. It is also
possible that reservoirs of Pb associated with past airborne Pb are still within these
houses near the facility and correlated with current ambient air Pb levels are reflected
in the site-specific regression model established between ambient air Pb levels and
indoor dust Pb. Thus, a portion of indoor dust Pb that results from contributions from
indoor reservoirs may be being attributed through our application of the model to
current ambient air Pb.
The air-to-blood ratios presented earlier in Section 3.5.2.2 for the general urban case
study and primary Pb smelter case study are also pertinent to the core modeling approach results
for these two case studies presented in this chapter and the reader is referenced to that earlier
section for discussion of these alternative ratios and implications in the context of performance
evaluation for blood Pb modeling. Other elements of performance evaluation for blood Pb
modeling presented earlier in Section 3.5.2 (e.g., evaluation of candidate blood Pb models - see
Section 3.5.2.1) also pertain to the analyses described in this section.
5-18
-------
Table 5-7. Air-to-blood ratios derived by comparing air quality scenario air and blood Pb
estimates.
Air Scenario
Median
Total Blood
Pb (ug/dL)
Annual
Average
Ambient Air
Concentration
(ug/m3)3
Ratio"
General urban case study
Current NAAQS (1.5 |jg/m3, max quarterly average)
Alternative NAAQS 2 (0.5 ug/m3, max monthly average)
Current Conditions (95th percentile)
Alternative NAAQS 1 (0.2 ug/m3, max quarterly average)
Current Conditions (mean)
Alternative NAAQS 3 (0.2 ug/m3, max monthly average)
Alternative NAAQS 4 (0.05 ug/m3, max monthly average)
Alternative NAAQS 5 (0.02 ug/m3, max monthly average)
3.12
2.16
2.13
2.01
1.91
1.88
1.67
1.61
0.600
0.125
0.114
0.080
0.056
0.050
0.013
0.005
1:2
1:3
1:4
1:4
1:5
1:6
1:9
Primary Pb smelter case study (full study area)
Current NAAQS (1.5 ug/m3, max quarterly average)
Alternative NAAQS 2 (0.5 ug/m3, max monthly average)
Alternative NAAQS 1 (0.2 ug/m3, max quarterly average)
Alternative NAAQS 3 (0.2 ug/m3, max monthly average)
Alternative NAAQS 4 (0.05 ug/m3, max monthly average)
Alternative NAAQS 5 (0.02 ug/m3, max monthly average)
1.50
1.44
1.40
1.40
1.38
1.37
0.042
0.019
0.009
0.007
0.002
0.001
1:3
1:3
1:4
1:4
1:7
Primary Pb smelter case study (1.5 km subarea)
Current NAAQS (1 .5 ug/m3, max quarterly average)
Alternative NAAQS 2 (0.5 ug/m3, max monthly average)
Alternative NAAQS 1 (0.2 ug/m3, max quarterly average)
Alternative NAAQS 3 (0.2 ug/m3, max monthly average)
Alternative NAAQS 4 (0.05 ug/m3, max monthly average)
Alternative NAAQS 5 (0.02 ug/m3, max monthly average)
4.58
3.20
2.50
2.35
1.75
1.63
0.238
0.105
0.052
0.042
0.011
0.004
1:10
1:13
1:15
1:19
1:19
3 For the primary Pb smelter case study entries, these are the median air concentrations in a
copulation weighted distribution of air concentration estimates for each scenario.
These are air-to-blood ratios derived by comparing changes (deltas) in median total blood Pb
levels (concurrent) to associated changes in annual average air Pb levels as you step to the next
lowest air quality scenario. Accordingly, a ratio is not presented adjacent to the current NAAQS
air quality scenario (for any of the case studies). In this case, the first ratio presented for any of
the case studies is generated by comparing median blood Pb levels at the current NAAQS level
to the median blood Pb level at the highest of the alternative NAAQS levels (the 0.05 ug/m3
maximum monthly value).
5-19
-------
5.3 RISK ASSESSMENT
This chapter describes the approach used to characterize risk using the core modeling
approach, including discussion of the methodology (Section 5.3.1), presentation of risk estimates
(Section 5.3.2), and uncertainty characterization (Section 5.3.3).
5.3.1 Methods for Deriving Risk Estimates
Risk characterization for this assessment focuses on IQ loss in children. IQ loss is
derived using a set of concentration-response functions developed based on results from a pooled
analysis of epidemiology studies (Lanphear et al., 2005). These concentration-response
functions are combined with the population-level blood Pb distributions to produce distributions
of IQ loss estimates for each case study population. IQ loss is also apportioned among different
exposure pathways using the pathway apportionment information generated as part of the
exposure analysis.
Two key elements of the risk methodology are described in greater detail below: (a) the
concentration-response functions used in the analysis (Section 5.3.1.1) and (b) the stepwise
analytical procedure used to generate the IQ loss (risk) distributions (Section 5.3.1.2).
5.3.1.1 Concentration-response Functions
For the risk analyses described in this chapter, performed subsequent to the August 2007
public meeting of the CAS AC Pb Panel, four types of concentration-response functions for IQ
loss were used. Some of the functions in this set differ from those used in the initial analyses
(Section 4.1.1), reflecting CASAC recommendations to use a two-piece or dual linear function
that recognizes a change in slope (to a notably higher value) at blood Pb levels of 7.5 ug/dL and
give less prominence to the two-piece linear function with hinge at 10.82 ug/dL derived for the
initial analyses of the full-scale assessment (Henderson, 2007b). The set of four concentration-
response functions used in the analyses presented in this chapter include the following:
• The log-linear function with a cutpoint at 1.0 [\.g/dL (for the concurrent blood Pb
metric): This function is described in Section 4.1.1.1.
• The log-linear function with low-exposure linearization (for the concurrent blood Pb
metric): This function is described in Section 4.1.1.2.
• Dual linear function, stratified at 10 [ig/dLpeak: This represents two linear functions
reported in Lanphear et al., 2005. To derive these, the full study population was
stratified into two groups - children with peak blood Pb levels above and below 10
|ig/dL - and separate linear functions were fit to the IQ and concurrent blood Pb values
for these two groups (Lanphear et al., 2005). In order to utilize this model, we
considered the relationship between peak and concurrent blood Pb levels in the pooled
dataset, as well as the relationship in the cohort comprising the bulk of the low blood
Pb subsets. For the full pooled dataset, the average peak blood Pb is 18.0 |ig/dL, while
5-20
-------
the average concurrent blood Pb is 9.7 |ig/dL, approximately a factor of two difference
(Lanphear et al., 2005). For the Rochester cohort, which comprised the majority of the
subset of children with peak blood Pb values below 10 |ig/dL, the average peak blood
Pb is 9.0 |ig/dL, while the average concurrent blood Pb is 4.0 |ig/dL, approximately a
factor of two difference (Lanphear et al., 2005). Accordingly, in specifying the
concurrent blood Pb level at which this function would change slope, we assumed that
the peak blood Pb of 10 |ig/dL corresponded to a concurrent blood Pb of 5 |ig/dL. The
form of the function as applied in this risk assessment is:
- For concurrent blood Pb level < 5 |ig/dL:
IQ loss = beta 1 * concurrent blood Pb
- For concurrent blood Pb level > 5 |ig/dL:
IQ loss = beta 2 * concurrent blood Pb
Where:
beta 1 = -0.80
beta 2 = -0.13
• Dual Linear function, stratified at 7.5 [ag/dLpeak: This represents two linear functions,
and like the previous ones, these were separately derived from subgroups of the full
study population (Lanphear et al., 2005). In this case, the population was stratified into
groups of children with peak blood Pb levels above and below 7.5 |ig/dL. For the same
considerations of the relationship between peak and concurrent blood Pb levels stated
for previously described function in specifying the concurrent blood Pb level at which
this function would change slope, we assumed that the peak blood Pb of 7.5 |ig/dL
corresponded to a concurrent blood Pb of 3.75 |ig/dL. The form of this function as
applied in this risk assessment is:
- For concurrent blood Pb level < 3.75 |ig/dL:
IQ loss = beta 1 * concurrent blood Pb
- For concurrent blood Pb level > 3.75 |ig/dL:
IQ loss = beta 2 * concurrent blood Pb
Where:
beta 1 = -2.94
beta 2= -0.16
The four functions are presented in Figure 5-2 and compared in Table 5-8 with regard to
total IQ loss and incremental IQ loss (IQ loss per |ig/dL) across a range of concurrent blood Pb
levels. A number of observations can be made by considering the plots of the four functions
together with the results presented in Table 5-8. This discussion presents IQ loss estimates
rounded to the nearest integer, although they are presented with greater precision in Table 5-8.
At blood Pb levels less than or equal to about 2 |ig/dL, with 2 |ig/dL being near the average
blood Pb level for young children in the U.S., based on NHANES-IV data and near the average
5-21
-------
blood Pb estimated in the risk assessment for the three location-specific urban case studies, the
log-linear with cutpoint model and the dual linear - stratified at 10 jUg/dLpeak generate similar
IQ loss estimates. Both models predict a total of 2 points IQ loss associated with a blood Pb
level of 2 |ig/dL. By contrast, the remaining two models (log-linear with low-exposure
linearization and dual linear - stratified at 7.5 jUg/dLpeak) generate substantially larger IQ loss
estimates, ranging from 5 to 6 points IQ loss.
At blood Pb levels above 2 |ig/dL, the pattern of predicted IQ loss shifts substantially
across the four models, with the dual linear - stratified at 7.5 ug/dL peak maintaining a constant
slope (up to 3.75 |ig/dL) which is significantly higher than the other models, all of which either
have slopes which decrease (for the log-linear models) or have a lower slope to begin with (the
dual linear - stratified at 10 jUg/dLpeak). At higher exposure levels ranging from 2 and 10
|ig/dL, the dual linear - stratified at 10 pg/dLpeak and the log-linear with cutpoint perform
similarly, with the latter predicting somewhat higher total IQ loss at 10 |ig/dL (6 points, versus 5
for the dual linear-stratified at 10 |ig/dL). Within this higher exposure range, the log-linear with
linearization model generates IQ loss estimates which are higher than the two models just
discussed, but lower than the dual linear - stratified at 7.5 peak. Specifically, at 10 |ig/dL, the
log-linear with linearization model predicts 9 points IQ loss, while the dual linear - stratified at
7.5peak model predicts 12 points IQ loss. As discussed in Section 2.1.5, the dataset from which
the lower blood Pb portion of the dual linear - stratified at 7.5 peak function is derived is much
smaller than those from which the other functions are derived and includes variable
representation of six of the seven cohorts in the pooled analysis with a dominance (-65% of
total) of one cohort.
Both of the dual linear models with stratification have lower slopes above 5 |ig/dL than
either of the log-linear models. This means that reductions in blood Pb levels for modeled
children with relatively high exposures (> 5 |ig/dL total blood Pb) will result in relatively lower
IQ benefits using the two dual-linear models, compared with the two log-linear models (even if,
in the case of the dual linear - stratified at 7.5 |ig/dL model, the overall IQ loss is higher for that
simulated child than with either of the log-linear models).
5-22
-------
Figure 5-2. Comparison of four concentration-response functions for concurrent blood
Pb levels < 10 ug/dL.
14
19
10
Q
in °
o
o 6
A
o
o
/'
/
*
/ .--
* «> -•"*""
/ ....--•"
/' .,--•••'"
f .-•'' „ , _
/'
,/•'" -->^^^~
/' _ ^^"^"^
£<^
31234567891
Concurrent blood Pb (ug/dL)
D
log-linear w ith outpoint
dual linear - stratified at
10 ug/dL, peak
— • — - - dual linear - stratified at
7.5 ug/dL, peak
log-linear w ith low -
exposure linearization
Table 5-8. Comparison of total and incremental IQ loss estimates below 10 ug/dL for the
four concentration-response functions.
Performance Metric
Total IQ loss
Incremental IQ loss
(average # points
per |jg/dl_)
at2ug/dL
at 5 ug/dL
at 7.5 ug/dL
at10|jg/dL
<2 |jg/dl_
<5 |jg/dl_
<7.5 |jg/dl_
<10|jg/dL
Concentration-Response Function
Log-
linear
with
outpoint
Log-linear
with low-
exposure
linearization
Dual linear -
stratified at
10ug/dL
peak
Dual linear -
stratified at
7.5 ug/dL
peak
Points, IQ loss
1.9
4.3
5.4
6.2
0.94
0.87
0.73
0.62
4.6
7.0
8.1
8.9
2.29
1.41
1.09
0.89
1.6
3.9
4.3
4.6
0.80
0.80
0.58
0.47
5.9
11.1
11.5
11.9
2.94
2.24
1.55
1.20
5-23
-------
5.3.1.2 Projection of Population Risk
The methods used to project population-level risk for the additional analyses are those
described in Section 4.1.2.
With regard to population-level risk metrics, risk estimates for different population
percentiles are generated for all case studies. For the location-specific urban case studies, an
additional type of population risk metric is also presented. Specifically, we have estimated the
number of children associated with different magnitudes of IQ loss resulting from total Pb
exposure under different ambient air quality scenarios (i.e., IQ loss incidence estimates). This
metric illustrates (a) the overall number of children within a given urban case study location
projected to experience various levels of IQ loss due to Pb exposure and (b) how that distribution
of incidence changes with alternate NAAQS levels.
5.3.2 Risk Estimates
Estimates of IQ loss resulting from Pb exposure have been developed for each air quality
scenario in each case study (see Appendices N and O). Further, multiple sets of risk results were
generated for each combination of case study and air quality scenario, in an effort to consider
uncertainty specifically related to specifying the relationship between Pb exposure and IQ loss
(see Section 5.1.2). That is, four separate risk distributions were generated for each air quality
scenario for each of the case studies presented here (see Table 5-1). These risk estimates include
(a) population-risk distribution estimates (in the form of estimated IQ loss for specific population
percentiles) and (b) IQ loss incidence estimates, as described below.
As noted earlier in presenting exposure results (see Section 5.2.4), IQ loss estimates
associated with air pathways and current levels of Pb emitted to the air (including via
resuspension) are likely to fall between the estimates for "recent air" and those for "recent" plus
"past air". Additionally, with regard to the location-specific urban case studies "recent air" and
"recent" plus "past air" categories, an artifact of the hybrid dust Pb model tends to mask trends in
the two components of dust Pb (recent air and other), which contribute to "recent air" and "past
air" estimates, respectively (see Section 4.3.1 discussion of this uncertainty). For the primary Pb
smelter case study, uncertainty in parsing out the "recent air" and "other" components of indoor
dust (specifically for the site-specific regression model used in the remediation zone) have led us
to conclude that only "recent plus past air" IQ loss estimates should be presented and "recent air"
results should not be separately presented for the primary Pb smelter, as is done for the other
case studies (see Section 3.1.4.2).
5-24
-------
5.3.2.1 Population Risk Distribution Estimates
Risk estimates for specific population percentiles of the risk distribution associated with
total blood Pb levels are presented in Tables 5-9 and 5-10. These include the median (Table 5-9)
and 95th percentile (Table 5-10) IQ loss estimates based on total blood Pb. Two additional types
of risk estimates are summarized in Tables 5-9 and 5-10 in addition to IQ loss based on total
blood Pb. Also presented are the contributions to IQ loss from the recent air and recent plus past
air exposure pathways (see Section 3.2.2), which as noted previously (Sections 2.4.3 and 3.2.2)
represent the policy-relevant pathways, with the true values for the policy-relevant pathways
considered to fall between the estimates for "recent air" and those for "recent" plus "past air".
More detailed pathway apportionment estimates (including percentile contributions to total IQ
loss from individual exposure pathways) are presented in the detailed risk results tables in
Appendices N, O and P. These more detailed risk results are, however, subject to the same
limitations related to presentation of pathway apportionment, as noted above for individual case
studies (e.g., reduced confidence in parsing out recent air-related risk from recent plus past air
risk for the primary Pb smelter case study).
The total Pb exposure risk estimates presented in Table 5-9 are those for the median in
the distribution of total risk, and the estimates for the other two categories in that table (i.e.,
recent air and recent plus past air) are the values for those categories associated with the median
for the total exposure pathway estimate. The corresponding estimates for the 95th percentile in
the distribution of IQ loss estimates for total blood Pb for each case study are presented in Table
5-10. This presentation may lead to some seeming inconsistencies in trends for recent air and
recent air plus past air risk estimates across air quality scenarios. This is because the recent air
or recent plus past air estimates associated with the median and 95th percentile total blood Pb
estimate may not necessarily be the median and 95th percentiles of the distribution of estimates
for those specific categories. This is because the blood Pb level (and associated total IQ loss) for
a simulated child reflects not only the total Pb uptake (from both background and ambient air-
related pathways), but also the GSD-based adjustment factor. This means that two simulated
children could have identical blood Pb levels (and IQ loss estimates), but one child could have a
higher recent air exposure, with the other child having their lower recent air exposure
compensated for by a higher GSD-based adjustment factor, thereby resulting in both children
having the same total blood Pb level. In presenting IQ loss estimates in tables 5-9 and 5-10, all
values are rounded to one decimal place.
As mentioned in Chapter 1, risk results are provided here without substantial
interpretation. Rather, interpretative discussion of these results is provided in the Staff Paper.
5-25
-------
Table 5-9. Summary of risk estimates for medians of total-exposure risk distributions.
Case Study and Air Quality Scenario
Points IQ Loss3
Recent Airb
(low, ILL and high C-R
function estimates)
Low LLL° High
Recent plus Past Airb
(low, ILL and high C-R
function estimates)
Low LLL High
Total Pb Exposure
(low, LLL and high C-R
function estimates)
Low LLL High
Location-specific (Chicago)
Current NAAQS (1 .5 |jg/m3, max quarterly)
Current conditions
(0.14 |jg/m3 max quarterly; 0.31 pg/m3 max monthly)
Alternative NAAQS (0.2 |jg/m3, max monthly)
Alternative NAAQS (0.05 |jg/m3, max monthly)
Alternative NAAQS (0.02 |jg/m3, max monthly)
1.4
0.3
0.3
0.1
<0.1
3.4
0.6
0.6
0.2
0.1
5.6
0.7
0.9
0.2
0.1
Location-specific (Cleveland
Current NAAQS (1 .5 pg/m3, max quarterly)
Current conditions
(0.36 pg/m3 max quarterly; 0.56 pg/m3 max monthly)
Alternative NAAQS (0.5 |jg/m3, max monthly)
Alternative NAAQS (0.2 |jg/m3, max quarterly)
Alternative NAAQS (0.2 |jg/m3, max monthly)
Alternative NAAQS (0.05 |jg/m3, max monthly)
Alternative NAAQS (0.02 |jg/m3, max monthly)
0.6
0.2
0.2
0.1
0.1
<0.1
<0.1
2.8
0.7
0.6
0.4
0.6
0.1
<0.1
2.1
0.9
1.9
0.5
0.4
0.1
0.1
2.0
1.0
1.0
0.9
0.8
4.7
2.9
2.9
2.6
2.6
7.4
3.5
3.6
3.2
3.1
2.4
1.4
1.4
1.3
1.3
5.6
4.2
4.2
4.0
4.0
8.8
5.2
5.2
4.8
4.7
1.3
1.0
1.0
0.9
0.9
0.8
0.8
3.9
2.9
2.9
2.8
2.8
2.6
2.6
4.6
3.6
3.9
3.3
3.2
3.1
3.0
1.7
1.4
1.4
1.4
1.3
1.3
1.2
4.7
4.2
4.2
4.1
4.1
4.0
3.9
6.3
5.2
5.2
5.0
4.9
4.7
4.6
Location-specific (Los Angeles)
Current NAAQS (1 .5 |jg/m3, max quarterly)
Current conditions
(0.09 pg/m3 max quarterly; 0.17 pg/m3 max monthly)
Alternative NAAQS (0.05 |jg/m3, max monthly)
Alternative NAAQS (0.02 |jg/m3, max monthly)
1.1
0.2
0.1
<0.1
2.7
0.7
0.3
0.1
4.0
0.9
0.4
0.2
1.7
0.9
0.9
0.8
4.2
2.9
2.7
2.6
6.2
3.5
3.2
3.1
2.1
1.4
1.3
1.3
5.3
4.2
4.0
4.0
7.7
5.1
4.8
4.7
General urban
Current NAAQS (1 .5 |jg/m3, max quarterly)
Alternative NAAQS (0.5 |jg/m3, max monthly)
Current conditions (0.87 |jg/m3 max quarterly)
Alternative NAAQS (0.2 |jg/m3, max quarterly)
Current conditions (0.14 |jg/m3 max quarterly)
Alternative NAAQS (0.2 |jg/m3, max monthly)
Alternative NAAQS (0.05 |jg/m3, max monthly)
Alternative NAAQS (0.02 |jg/m3, max monthly)
1.5
0.7
0.6
0.5
0.4
0.4
0.2
0.1
3.5
1.9
1.8
1.5
1.3
1.2
0.5
0.3
5.6
2.5
2.4
2.0
1.6
1.5
0.6
0.3
2.1
1.3
1.3
1.2
1.1
1.1
0.9
0.9
4.8
3.6
3.6
3.4
3.2
3.2
2.8
2.6
7.7
4.8
4.7
4.3
4.1
4.0
3.3
3.1
2.5
1.7
1.7
1.6
1.5
1.5
1.3
1.3
5.8
4.8
4.7
4.6
4.5
4.4
4.1
4.0
9.2
6.4
6.3
5.9
5.6
5.6
5.0
4.8
Primary Pb smelter - full study area
Current NAAQS (1 .5 |jg/m3, max quarterly)
Alternative NAAQS (0.5 |jg/m3, max monthly)
Alternative NAAQS (0.2 |jg/m3, max quarterly)
Alternative NAAQS (0.2 |jg/m3, max monthly)
Alternative NAAQS (0.05 |jg/m3, max monthly)
Alternative NAAQS (0.02 |jg/m3, max monthly)
NAd
0.6
0.8
0.6
0.4
0.4
0.6
1.9
2.9
2.3
2.7
1.7
2.6
2.3
2.7
2.6
3.0
1.9
3.0
1.2
1.0
0.9
0.9
0.9
0.9
3.8
3.7
3.6
3.6
3.6
3.6
4.4
4.2
4.2
4.1
4.0
4.1
Primary Pb smelter - 1 .5 km subarea
Current NAAQS (1 .5 |jg/m3, max quarterly)
Alternative NAAQS (0.5 |jg/m3, max monthly)
Alternative NAAQS (0.2 |jg/m3, max quarterly)
Alternative NAAQS (0.2 |jg/m3, max monthly)
Alternative NAAQS (0.05 |jg/m3, max monthly)
Alternative NAAQS (0.02 |jg/m3, max monthly)
NAd
3.2
2.1
1.5
1.2
0.9
0.9
6.0
4.5
3.8
3.7
2.8
2.9
9.4
7.7
5.6
5.1
3.4
3.3
3.7
2.6
2.0
1.9
1.4
1.3
6.8
5.8
5.2
5.0
4.2
4.0
11.2
9.4
7.4
6.9
5.1
4.8
a - Estimates are rounded to one decimal place.
b -The term "past air" includes contributions from the outdoor soil/dust contribution to indoor dust, historical air contribution to indoor dust, and outdoor soil/dust
pathways, while "recent air" refers to contributions from inhalation of ambient air Pb or ingestion of indoor dust Pb predicted to be associated with outdoor
ambient air Pb levels, with outdoor ambient air also potentially including resuspended previously deposited Pb (see Section 2.4.3).
c - Log-linear with low-exposure linearization concentration-response function.
d - "Recent air" estimates were not developed for the primary Pb smelter case study (see Section 3.1 .4.2).
5-26
-------
Table 5-10. Summary of risk estimates for 95
distributions.
th
percentiles of total exposure risk
Case Study and Air Quality Scenario
Points IQ Loss3
Recent Airb
(low, ILL and high C-R
function estimates)
Low
LLLC
High
Recent plus Past Airb
(low, ILL and high C-R
function estimates)
Low
LLL
High
Total Pb Exposure
(low, LLL and high C-R
function estimates)
Low
LLL
High
Location-specific (Chicago)
Current NAAQS (1 .5 |jg/m3, max quarterly)
Current conditions
(0.14 pg/m3 max quarterly; 0.31 pg/m3 max monthly)
Alternative NAAQS (0.2 |jg/m3, max monthly)
Alternative NAAQS (0.05 |jg/m3, max monthly)
Alternative NAAQS (0.02 |jg/m3, max monthly)
3.0
0.7
1.0
0.2
0.2
5.8
1.3
1.9
0.4
0.3
7.7
2.0
2.8
0.7
0.5
4.0
2.8
2.9
2.7
2.7
7.6
5.2
5.3
4.8
4.8
10.3
7.8
8.1
7.5
7.4
4.7
4.1
4.1
4.1
4.1
9.0
7.5
7.5
7.3
7.3
12.1
11.4
11.4
11.3
11.3
Location-specific (Cleveland)
Current NAAQS (1 .5 |jg/m3, max quarterly)
Current conditions
(0.36 pg/m3 max quarterly; 0.56 pg/m3 max monthly)
Alternative NAAQS (0.5 |jg/m3, max monthly)
Alternative NAAQS (0.2 |jg/m3, max quarterly)
Alternative NAAQS (0.2 |jg/m3, max monthly)
Alternative NAAQS (0.05 |jg/m3, max monthly)
Alternative NAAQS (0.02 |jg/m3, max monthly)
1.4
0.6
1.1
0.5
0.3
0.1
<0.1
2.6
1.2
1.9
0.9
0.6
0.2
0.1
3.7
1.8
2.9
1.4
0.9
0.3
0.1
Location-specific (Los Angeles
Current NAAQS (1 .5 |jg/m3, max quarterly)
Current conditions
(0.09 pg/m3 max quarterly; 0.17 pg/m3 max monthly)
Alternative NAAQS (0.05 |jg/m3, max monthly)
Alternative NAAQS (0.02 |jg/m3, max monthly)
2.3
0.8
0.4
0.1
4.4
1.4
0.7
0.2
6.1
2.1
1.1
0.4
3.1
2.8
2.9
2.8
2.7
2.6
2.6
5.9
5.2
5.3
5.0
4.9
4.7
4.7
8.5
7.8
8.1
7.6
7.5
7.4
7.2
4.3
4.1
4.1
4.1
4.1
4.0
4.0
8.1
7.6
7.5
7.4
7.4
7.2
7.2
11.6
11.4
11.4
11.3
11.3
11.3
11.3
3.6
2.8
2.7
2.7
6.9
5.2
4.9
4.7
9.5
7.8
7.6
7.4
4.5
4.1
4.1
4.0
8.6
7.5
7.3
7.2
11.8
11.4
11.3
11.3
General urban
Current NAAQS (1 .5 pg/m3, max quarterly)
Alternative NAAQS (0.5 |jg/m3, max monthly)
Current conditions - high-end (0.87 pg/m3 max quarterly)
Alternative NAAQS (0.2 pg/m3, max quarterly)
Current conditions - mean (0.14 pg/m3 max quarterly)
Alternative NAAQS (0.2 pg/m3, max monthly)
Alternative NAAQS (0.05 pg/m3, max monthly)
Alternative NAAQS (0.02 pg/m3, max monthly)
2.9
1.7
1.6
1.4
1.2
1.1
0.5
0.3
5.5
3.2
3.1
2.6
2.2
2.1
0.9
0.5
7.3
4.6
4.4
3.8
3.3
3.1
1.4
0.7
3.9
3.3
3.2
3.1
3.0
3.0
2.8
2.7
7.6
6.1
6.0
5.8
5.6
5.6
5.0
4.8
10.1
8.8
8.7
8.5
8.3
8.2
7.7
7.5
4.7
4.3
4.3
4.2
4.2
4.2
4.1
4.1
9.1
8.1
8.0
7.9
7.7
7.7
7.4
7.3
12.1
11.6
11.6
11.5
11.5
11.4
11.3
11.3
Primary Pb smelter - full study area
Current NAAQS (1 .5 pg/m3, max quarterly)
Alternative NAAQS (0.5 pg/m3, max monthly)
Alternative NAAQS (0.2 pg/m3, max quarterly)
Alternative NAAQS (0.2 pg/m3, max monthly)
Alternative NAAQS (0.05 pg/m3, max monthly)
Alternative NAAQS (0.02 pg/m3, max monthly)
NAd
2.3
2.5
1.9
2.0
1.5
2.5
4.2
4.9
3.9
4.0
3.1
5.3
6.8
8.3
6.6
7.0
5.4
9.2
3.7
3.4
3.2
3.2
3.1
3.1
6.8
6.6
6.5
6.4
6.3
6.3
11.2
11.1
11.1
11.1
11.0
11.0
Primary Pb smelter - 1 .5 km subarea
Current NAAQS (1 .5 pg/m3, max quarterly)
Alternative NAAQS (0.5 pg/m3, max monthly)
Alternative NAAQS (0.2 pg/m3, max quarterly)
Alternative NAAQS (0.2 pg/m3, max monthly)
Alternative NAAQS (0.05 pg/m3, max monthly)
Alternative NAAQS (0.02 pg/m3, max monthly)
NAd
4.2
4.0
3.7
3.4
2.8
2.4
8.0
7.5
6.9
6.1
5.3
4.7
10.4
10.5
10.2
9.2
8.7
7.9
5.0
4.5
4.2
4.1
3.6
3.3
9.5
8.5
7.8
7.6
6.8
6.5
12.4
11.8
11.5
11.4
11.1
11.1
a - Estimates are rounded to one decimal place.
b -The term "past air" includes contributions from the outdoor soil/dust contribution to indoor dust, historical air contribution to indoor dust, and outdoor soil/dust
pathways, while "recent air" refers to contributions from inhalation of ambient air Pb or ingestion of indoor dust Pb pred cted to be associated with outdoor
ambient air Pb levels, with outdoor ambient air also potentially including resuspended, previously deposited Pb (see Section 2.4.3).
c - Log-linear with low-exposure linearization concentration-response function.
d - "Recent air" estimates were not developed for the primary Pb smelter case study (see Section 3.1 .4.2).
5-27
-------
5.3.2.2 IQ Loss Incidence Estimates
Incidence of different levels of IQ loss has been estimated for the three location-specific
urban case studies (Appendix O). These estimates are summarized in Tables 5-11 and 5-12.
Specifically, Table 5-11 shows the number of children projected to have IQ loss greater than one
point under the current conditions scenario and the change in that metric under the current
NAAQS and the alternate NAAQS considered for each location-specific urban case study. The
value of one IQ point was selected as consistent with CASAC recommendations as to the
magnitude of IQ loss they considered to be significant from a public health perspective
(Henderson, 2007b). Similarly, Table 5-12 presents incidence estimates associated with a
greater magnitude of cognitive impact, the number of children exceeding 7 points IQ loss. The
value of 7 IQ points is the magnitude of IQ loss projected for the 95th population percentile
(using the log-linear with low-exposure linearization function) under current conditions at the
case studies evaluated in this analysis (See Table 5-10). As mentioned in Chapter 1, risk results
are provided here without substantial interpretation. Rather, interpretative discussion of these
results is provided in the Staff Paper.
5-28
-------
1 Table 5-11. Incidence of children with >1 point IQ loss.
Air Quality Scenario
(for location-specific urban case studies)
Chicago (total modeled population: 396,51 1)
Chicago Current Conditions
Current NAAQS (1 .5 ug/m3 Maximum Quarterly)
Alternative NAAQS (0.2 ug/m3 Maximum Monthly)
Alternative NAAQS (0.05 ug/m3 Maximum Monthly)
Alternative NAAQS (0.02 ug/m3 Maximum Monthly)
Cleveland (total modeled population: 13,990)
Cleveland Current Conditions
Current NAAQS (1 .5 ug/m3 Maximum Quarterly)
Alternative NAAQS (0.2 ug/m3 Maximum Quarterly)
Alternative NAAQS (0.5 ug/m3 Maximum Monthly)
Alternative NAAQS (0.2 ug/m3 Maximum Monthly)
Alternative NAAQS (0.05 ug/m3 Maximum Monthly)
Alternative NAAQS (0.02 ug/m3 Maximum Monthly)
Los Angeles (total modeled population: 372,252)
Los Angeles Current Conditions
Current NAAQS (1 .5 ug/m3 Maximum Quarterly)
Alternative NAAQS (0.05 ug/m3 Maximum Monthly)
Alternative NAAQS (0.02 ug/m3 Maximum Monthly)
dual linear - stratified at
7.5 ug/dL peak
Incidence
of
>1 point
IQ loss
391 ,602
395,797
391,158
389,572
389,176
13,809
13,893
13,770
13,789
13,759
13,729
13,720
282,216
285,272
281,112
280,740
Delta
(change in
incidence
compared to
current
conditions)
4,195
-444
-2,030
-2,427
84
-38
-20
-50
-80
-88
3,056
-1,104
-1 ,476
log -linear with
linearization
Incidence
of
>1 point,
IQ loss
389,754
395,528
389,461
387,407
386,630
13,745
13,857
13,703
13,720
13,694
13,642
13,628
280,711
284,945
279,658
279,057
Delta
(change in
incidence
compared
to current
conditions)
5,773
-293
-2,347
-3,125
112
-42
-25
-51
-103
-117
4,234
-1 ,053
-1 ,654
dual linear - stratified
at 10 ug/dL peak
Inci-
dence of
>1 point,
IQ loss
271 ,031
347,415
271 ,444
253,775
249,865
9,526
10,664
9,221
9,497
9,083
8,785
8,736
191,675
240,988
183,395
180,745
Delta
(change in
incidence
compared
to current
conditions)
76,384
412
-17,256
-21,166
1,137
-305
-29
-443
-741
-790
49,313
-8,280
-10,929
log-linear with cutpoint
Incidence
of
>1 point,
IQ loss
236,257
314,053
235,559
224,394
219,294
8,515
9,769
8,160
8,464
8,010
7,720
7,668
170,474
226,608
161,914
158,234
Delta
(change in
incidence
compared
to current
conditions)
77,795
-698
-1 1 ,864
-16,963
1,254
-354
-51
-505
-795
-846
56,134
-8,560
-12,240
5-29
-------
1 Table 5-12. Incidence of children with >7 points IQ loss.
Air Quality Scenario
(location-specific urban case studies)
Chicago (total modeled population: 396,511)
Chicago Current Conditions
Current NAAQS (1 .5 ug/m3 Maximum Quarterly)
Alternative NAAQS (0.2 ug/m3 Maximum Monthly)
Alternative NAAQS (0.05 ug/m3 Maximum Monthly)
Alternative NAAQS (0.02 ug/m3 Maximum Monthly)
Cleveland (total modeled population: 13,990)
Cleveland Current Conditions
Current NAAQS (1 .5 ug/m3 Maximum Quarterly)
Alternative NAAQS (0.2 ug/m3 Maximum Quarterly)
Alternative NAAQS (0.5 ug/m3 Maximum Monthly)
Alternative NAAQS (0.2 ug/m3 Maximum Monthly)
Alternative NAAQS (0.05 ug/m3 Maximum Monthly)
Alternative NAAQS (0.02 ug/m3 Maximum Monthly)
Los Angeles (total modeled population: 372,252)
Los Angeles Current Conditions
Current NAAQS (1 .5 ug/m3 Maximum Quarterly)
Alternative NAAQS (0.05 ug/m3 Maximum Monthly)
Alternative NAAQS (0.02 ug/m3 Maximum Monthly)
dual linear - stratified at 7.5
ug/dL peak
Incidence of
> 7 points
IQ loss
136,709
244,401
136,067
120,706
117,819
4,834
6,139
4,525
4,806
4,424
4,106
4,051
94,684
158,171
87,303
83,909
Delta
(change in
incidence
compared to
current
conditions)
107,692
-642
-16,003
-18,890
1,305
-309
-28
-410
-728
-783
63,487
-7,382
-10,775
log -linear with
linearization
Incidence
of
> 7 points
IQ loss
33,664
100,159
32,546
27,367
26,027
1,212
1,858
1,073
1,180
1,026
886
866
22,665
57,834
19,781
17,939
Delta
(change in
incidence
compared
to current
conditions)
66,495
-1,118
-6,297
-7,637
647
-139
-31
-186
-326
-345
35,168
-2884
-4726
dual linear - stratified at
10 ug/dL peak
Inci-
dence of
> 7 points
IQ loss
63
555
48
16
8
3
4
1
1
1
0
0
23
183
11
17
Delta
(change in
incidence
compared to
current
conditions)
492
-16
-48
-56
2
-2
-2
-2
-3
-3
160
-11
-6
log-linear with cutpoint
Incidence
of
> 7 points
IQ loss
1,015
5,226
1,007
864
690
46
105
40
43
43
24
27
732
3,771
624
498
Delta
(change in
incidence
compared to
current
conditions)
4,211
-8
-151
-325
59
-6
-3
-3
-22
-18
3,038
-109
-235
5-30
-------
5.3.3 Uncertainty Characterization and Sensitivity Analysis
This section discusses uncertainty related to exposure and risk estimates generated for the
results for the core modeling approach applied to the primary Pb smelter, general urban and
location-specific urban case studies. It also discusses sensitivity analyses completed to inform
application of the core modeling approach to these case studies. Several methods have been used
to examine uncertainty in our modeling approach and its potential impact on exposure risk
estimates. These methods for uncertainty evaluation include the following:
• Qualitative discussion of key sources of uncertainty and their potential impact on
exposure and risk estimates (Section 5.3.3.1). This qualitative discussion focuses on
factors particular to the new analyses presented in this section - those sources of
uncertainty that these new analyses share with the analyses presented in Chapters 3 and
4 are not repeated here (see Section 4.3.1 for that discussion).
• Evaluation of model performance, including comparison with empirical data (Section
5.3.3.2). As with the qualitative discussion of uncertainty, only those performance
evaluation elements particular to the analyses in this chapter (i.e., performance
evaluation of modeled blood Pb levels - see Section 5.2.5.2) are discussed here. The
reader is referred back to Section 4.3.3 for a more complete discussion of performance
evaluation related to modeled media concentrations and blood Pb modeling for this risk
assessment.
• Development of multiple sets of risk estimates for each assessment scenario that
illustrate the impact of different concentration-response models relating Pb exposure to
IQ loss and the associated uncertainty (see Section 5.3.3.3).
In addition to these methods for evaluating uncertainty in the modeling approach, an
additional sensitivity analysis has been completed for this portion of the analysis, specifically
focused on the hybrid indoor dust Pb model used in the general urban case study and the
location-specific urban case studies. This sensitivity analysis is discussed in Section 5.3.3.4.
The results of the sensitivity analysis described in Section 4.3.2 for the general urban case study
are broadly applicable to that case study, as well as the location-specific urban case studies.
5.3.3.1 Qualitative Discussion of Key Sources of Uncertainty
This section provides qualitative discussion of key sources of uncertainty related to the
analyses presented in this section. Specifically, it addresses sources of uncertainty specific to the
location-specific urban case studies; sources of uncertainty related to the primary Pb smelter and
general urban case studies have already been discussed (see Section 4.3.1). In addition, many of
the sources of uncertainty that impact the general urban case study also impact the location-
specific urban case studies, since both share similar modeling elements (e.g., hybrid indoor dust
Pb model, IQ loss functions, blood Pb GSD). The reader is referred back to Section 4.3.1 for
5-31
-------
discussion of those sources of uncertainty. Sources of uncertainty particular to the location-
specific urban case studies include:
• Location-specific urban case studies: As recognized in Appendix A, the Pb-TSP
monitoring network is currently quite limited. The number of monitors available to
represent air concentrations in these case studies ranged from six for Cleveland to 11
for Chicago. Accordingly, our estimates of the magnitude of and spatial variation of
air Pb concentrations are subject to uncertainty associated with the limited data. In
applying the available data to each of these case studies, exposure zones, one
corresponding to each monitor, were created and each U.S. Census block (and the
children within that demographic unit) were distributed among the exposure zones.
The details of the approach used are described in Section 5.1.3. Although this
approach provides a spatial gradient across the study area due to differences in
monitor values for each exposure zone, this approach assumes a constant
concentration within each exposure zone (i.e., no spatial gradient within a zone).
Additionally, the nearest neighbor approach to assign blocks to exposure zones
assumes that a monitor pertains to all locations that are closer to that monitor than to
any of the others in the study area. In reality, there may be different and more
variable spatial gradients in a study area than those reflected in the approach used
here. This introduces uncertainty into the characterization of risk for the urban case
studies.
• Current NAAQS air quality scenarios: For the location-specific urban case studies,
proportional roll-up procedures were used to adjust ambient air Pb concentrations up
to just meet the current NAAQS (see Sections 2.3.1 and 5.2.2.1 for detailed
discussion). Staff recognizes that it is extremely unlikely that Pb concentrations in
urban areas would rise to meet the current NAAQS and that there is uncertainty with
our simulation of such conditions. In these case studies we have simulated a
proportional roll-up, such that it is assumed that the current spatial distribution of air
concentrations (as characterized by the current data) is maintained and increased Pb
emissions contribute to increased Pb concentrations, the highest of which just meets
the current standard. There are many other types of changes within a study area that
could result in a similar result such as increases in emissions from just one specific
industrial operation that led to air concentrations in a part of the study area that just
meet the current NAAQS, while the remainder of the study area remained largely
unchanged (at current conditions). For the primary Pb smelter case study, where
current conditions exceed the current NAAQS, reaching the current NAAQS was
simulated using air quality modeling, emissions and source parameters used in
developing the 2007 proposed revision to the State Implementation Plan for the area
(see Section 3.1.1.2 for details).
• Alternative NAAQS air quality scenarios: In all case studies, proportional roll-down
procedures were used to adjust ambient air Pb concentrations downward to reach
alternative NAAQS (see Sections 2.3.1 and 5.2.2.1). We recognize that there is
uncertainty in simulating conditions associated with the implementation of emissions
reduction actions to meet a lower standard. There are a variety of changes other than
that represented by a proportional roll-down that could result in air concentrations
5-32
-------
that just meet lower alternative standards. For example, control measures might be
targeted only at the specific area exceeding standard, resulting in a reduction of air Pb
concentrations to the alternate standard while concentrations in the rest of the study
area remain unchanged (at current conditions). Consequently, there is uncertainty
associated with estimates for the alternate NAAQS scenarios.
• Relating blood Pb levels to IQ loss: Specification of the quantitative relationship
between blood Pb level (exposure) and IQ loss is subject to uncertainty, especially in
the projection of IQ loss at lower blood Pb levels (below 5 |ig/dL concurrent blood
Pb). As discussed in Section 2.1.5, this reflects limitations in the data available for
characterizing the concentration-response relationship. For example, the pooled
international dataset analyzed by Lanphear and others (2005) includes relatively few
children with blood Pb levels below 5 |ig/dL and no children with levels below 1
|ig/dL (see Section 2.1.5). As presented in Section 5.2.4, blood Pb levels in this
region are a particular focus in this review. For example, as is the case for mean
blood Pb levels nationally in the U.S. (CD, Section 4.3.1.3), concurrent blood Pb
estimates for the median of the populations simulated in this assessment fall below 5
|ig/dL. In recognition of the uncertainty in specifying a quantitative concentration-
response relationship at such levels, our core modeling approach involves the
application of four different functions to generate a range of risk estimates (see
Sections 4.2.6 and 5.3.1). The range of absolute IQ loss seen for a given case
study/air quality scenario combination when modeled using the four concentration-
response functions is typically close to a factor of 3. However, it is important to point
out that the relative (proportional) change in IQ loss across air quality scenarios (i.e.,
the pattern of risk reduction across air quality scenarios) is fairly consistent across all
four models. Note, however, that the function producing higher overall risk estimates
(the dual linear function, stratified at 7.5 |ig/dL, peak) will also produce larger
absolute reductions in IQ loss compared with the other three functions.
5.3.3.2 Performance Analyses
This section discusses the performance evaluation completed for blood Pb modeling for
the core analysis of the general urban and primary Pb smelter case studies. As mentioned above,
the performance evaluation of media concentrations for these case studies is discussed in Section
4.3.3.
The additional performance evaluation discussed in this chapter focused on a set of air-
to-blood ratios developed using a different approach than that used in Chapter 3- (see Section
5.2.5.2). The results of this analysis suggested that ratios generated for the general urban case
study and for the primary Pb smelter (full study area) were generally similar to those identified in
the literature, with the exception of ratios associated with the lowest alternative NAAQS, which
were larger than the typical range of ratios seen cited in the literature. By contrast, the set of
ratios generated for all air quality scenarios for the 1.5 km subarea of the primary Pb smelter
were noticeably larger than the values cited in the literature. A number of plausible explanations
for this discrepancy are presented in Section 5.2.5.2.
5-33
-------
5.3.3.3 Uncertainty in Modeling Approaches - Multiples Sets of Results
For the case studies included in the additional analyses presented in this chapter, four sets
of risk estimates were generated for each air quality scenario, reflecting consideration of four
concentration-response functions relating Pb exposure to IQ loss. The resulting median and 95th
percentile risk estimates for each case study and air quality scenario combination are presented in
Tables 5-9 and 5-10. As can be seen from these tables, estimates based on these four functions
result in a risk range spanning a factor of 4 for a particular case study and air quality scenario,
although the range among results for three of the four functions is much tighter.
5.3.3.4 Sensitivity Analysis - Indoor Dust Pb Modeling
One additional sensitivity analysis was completed as part of the analyses based on the
core modeling approach. Specifically, an alternate version of the hybrid indoor dust Pb model
used for the general urban and location-specific urban case studies was considered. This
alternate form of the model used a different approach for partitioning total dust Pb estimates
between recent air and "other" categories (see Section 3.1.4.1). Note, that consideration for
alternative concentration-response functions for IQ loss is not discussed here as part of the
sensitivity analysis, since all four concentration-response functions were used to evaluate the full
set of case studies and those results can be reviewed to identify differences in the performance of
these alternate models. The sensitivity analysis focusing on the hybrid indoor dust Pb model is
presented here.
Comments received from a member of the public regarding the full-scale analysis
methodology suggested a modification to the hybrid indoor dust Pb model used in evaluating the
general urban case study (see Section 3.1.4.1 for details on the hybrid indoor dust Pb model).
Specifically, rather than establishing the nonair fraction of indoor dust Pb as part of overall Pb
loading, as is currently done in the hybrid model, the recommendation was made that EPA take
the final version of the hybrid model, which predicts indoor dust Pb concentration, set the air
term to zero and solve for nonair indoor dust Pb (this providing an estimate of nonair indoor dust
Pb). This recommendation was made, in part, to address concerns raised by the commenter that
the nonair indoor dust Pb concentrations generated by the hybrid model, were not constant across
air quality scenarios. EPA recognizes that the hybrid indoor dust Pb model, as used in the full-
scale analysis, does produce nonair indoor dust Pb concentrations which vary by air quality
scenarios. This results from nonlinearity in the loading to concentration conversion algorithm
5-34
-------
within the model, which generates a greater per unit concentration at lower overall loading
levels.
To inform our understanding of model uncertainty related to predicting indoor dust Pb
levels for the urban scenarios, we have included this alternative hybrid model in this sensitivity
analysis. However, it is important to point out a key limitation in this formulation of the hybrid
dust model. By setting ambient air Pb levels to zero and then solving for indoor dust Pb, one is
using the steepest part of the loading-to-concentration curve to conduct the key step of
translating indoor dust Pb loading to equivalent concentration. In reality, we would expect to
always have a mixture of indoor dust Pb loadings from nonair and air sources, thereby resulting
in a larger total loading value, which would in turn be translated into an indoor dust Pb
concentration at a flatter portion of the curve.
Solving of the hybrid indoor dust Pb model for an ambient air Pb level of zero, yields the
fixed nonair indoor dust Pb level of 61 ppm. For the alternative hybrid model, this value has
been used as the nonair indoor dust Pb level. The ambient air-related portion of indoor dust Pb
concentration was then generated for each simulated individual by subtracting this value of 61
ppm from the total indoor dust Pb level generated by the hybrid model. If the total indoor dust
Pb concentration was lower than 61 ppm, then we fixed the concentration at 61 ppm and
assigned all of it to nonair sources.
Table 5-13 presents the median and 95th percentile IQ estimates associated with total
blood Pb estimates associated with applying the two versions of the hybrid indoor dust Pb model
to the general urban case study. The table also includes the IQ loss apportioned to recent air (i.e.,
inhalation plus ingestion of indoor dust loaded from air) for the median and 95th percentile IQ
estimates associated with total blood Pb estimates. All IQ loss estimates presented in Table 5-13
were generated using the log-linear with low-exposure linearization concentration-response
model.
5-35
-------
Table 5-13. Comparison of hybrid indoor dust model with a modified form of the model.
Air quality scenario
(General urban case study)
Current NAAQS
(1.5 |jg/m3, max quarterly)
Current conditions-high-end
(0.87 ug/m3, max quarterly)
Current conditions - mean
(0.14 ug/m3, max quarterly)
Alternative NAAQS
(0.5 ug/m3, max monthly)
Alternative NAAQS
(0.2 ug/m3, max quarterly)
Alternative NAAQS
(0.2 ug/m3, max monthly)
Alternative NAAQS
(0.05 ug/m3, max monthly)
Alternative NAAQS
(0.02 ug/m3, max monthly)
Points, IQ loss
Hybrid indoor dust model
(from full-scale analysis)
Median
Total
5.8
4.7
4.5
4.8
4.6
4.4
4.1
4.0
Recent
air
3.5
1.8
1.3
1.9
1.5
1.2
0.5
0.3
95tn percentile
Total
9.1
8.0
7.7
8.1
7.9
7.7
7.4
7.3
Recent
air
5.5
3.1
2.2
3.2
2.6
2.1
0.9
0.5
Alternate hybrid indoor dust
model (with fixed "other" dust
value)
Median
Total
5.8
4.7
4.4
4.8
4.6
4.4
4.1
4.0
Recent
air
3.1
1.4
0.9
1.5
1.1
0.8
0.3
0.1
95tn percentile
Total
9.1
8.0
7.7
8.1
7.9
7.7
7.4
7.3
Recent
air
4.8
2.3
1.6
2.4
1.9
1.5
0.5
0.3
Results presented in Table 5-13 suggest that the alternate hybrid indoor dust model (with
the fixed "other indoor dust value") does produce indoor dust Pb concentrations for recent air
that are noticeably lower than values generated using the original hybrid indoor dust model used
in the full-scale analysis. However, it is important to point out that because both versions of the
hybrid model generate the same total indoor dust Pb value (and only differ in the way that value
is apportioned between other and recent air), the trend in total IQ loss due to Pb exposure across
the air quality scenarios (moving from current NAAQS to the lowest alternative NAAQS
evaluated) is the same for both versions of the hybrid model.
5-36
-------
REFERENCES
Brunekreef, B. (1984) The relationship between air lead and blood lead in children: a critical review. Science of the
total environment, 38: 79-123.
Henderson, R. (2007a) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee, to
Administrator Stephen L. Johnson. Re: Clean Air Scientific Advisory Committee's (CASAC) Review of
the 2nd Draft Lead Human Exposure and Risk Assessments Document. September 27, 2007.
Henderson, R. (2007b) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee, to
Administrator Stephen L. Johnson. Re: Clean Air Scientific Advisory Committee's (CASAC) Review of
the 1st Draft Lead Staff Paper and Draft Lead Exposure and Risk Assessments. March 27, 2007.
Hilts, S. R. (2003) Effect of smelter emission reductions on children's blood Pb levels. Sci. Total Environ. 303: 51-
58.
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 Robe, R. (2005) Low-level environmental Pb exposure and children's intellectual function: An
international pooled analysis. Environmental Health Perspectives. 113(7):894-899.
5-37
-------
> w c
1 !.i-
I § &
5 C/2
§ £
B &
g- <»
hd
3
c±
O
& g
------- |