xvEPA
EPA
OSWER #9285.7-46
August 2001
REVIEW OF ADULT LEAD MODELS
EVALUATION OF MODELS FOR
ASSESSING HUMAN HEALTH RISKS
ASSOCIATED WITH LEAD EXPOSURES AT
NON-RESIDENTIAL AREAS OF SUPERFUND
AND OTHER HAZARDOUS WASTE SITES
Office of Solid Waste and Emergency Response
U.S. Environmental Protection Agency
Washington, DC 20460
-------
NOTICE
This document provides guidance to EPA staff. It also provides guidance to the public and to the
regulated community on how EPA intends to exercise its discretion in implementing the National
Contingency Plan. The guidance is designed to implement national policy on these issues. The
document does not, however, substitute for EPA's statutes or regulations, nor is it a regulation
itself. Thus, it cannot impose legally-binding requirements on EPA, States, or the regulated
community, and may not apply to a particular situation based upon the circumstances. EPA may
change this guidance in the future, as appropriate.
-------
CONTRIBUTING MEMBERS OF ADULT PB SUB-COMMITTEE OF
U.S. ENVIRONMENTAL PROTECTION AGENCY
TECHNICAL REVIEW WORKGROUP FOR LEAD
The Technical Review Workgroup for Lead (TRW) is an interoffice workgroup convened by the
U.S. EPA Office of Solid Waste and Emergency Response/Office of Emergency and Remedial
Response (OSWER/OERR).
Region 2
Mark Maddaloni
New York, NY
CHAIRPERSON
MEMBERS
Region 1 Region 5
Mary Ballew Patricia VanLeeuwen
Boston, MA Chicago, IL
Region 4 Region 6
Kevin Koporec Ghassan Khoury
Atlanta, GA Dallas, TX
Region 5 OERR Mentor
Mark Johnson Larry Zaragoza
Chicago, IL Office of Emergency and Remedial Response
Washington, DC
-------
Review of Adult Lead Models
Evaluation of Models for Assessing
Human Health Risks Associated with
Lead Exposures at Non-Residential
Areas of Superfund and Other
Hazardous Waste Sites
Final Draft: August 2001
Prepared by the
Adult Lead Risk Assessment Committee of the
Technical Review Workgroup for Lead (TRW)
NOTICE
This document has been reviewed in accordance with U.S. EPA
policy and is approved for publication. Mention of trade names or
commercial products does not constitute endorsement of
recommendation.
-------
-REVIEW OF ADULT LEAD MODELS-
Table of Contents
Table of Contents ii
List of Tables iv
List of Figures vi
List of Equations vii
List of Acronyms and Abbreviations viii
Contributing Members ix
Executive Summary x
1.0 Introduction 1
1.1 Purpose of Review 1
1.2 EPA Interim Adult Lead Methodology 3
1.3 Selection of Models 5
1.4 Model Review Criteria and Process 7
2.0 Model Review Summaries 10
2.1 Rabinowitz 10
2.1.1 Introduction 10
2.1.2 Evaluation Criteria 14
2.1.3 Summary 15
2.2 California 16
2.2.1 Introduction 16
2.2.2 Evaluation Criteria 18
2.2.3 Summary 20
2.3 Bert 21
2.3.1 Introduction 21
2.3.2 Evaluation Criteria 24
2.3.3 Summary 25
2.4 Stern (1994) and Stern (1996) 26
2.4.1 Introduction 26
2.4.2 Evaluation Criteria 31
2.4.3 Summary 34
2.5 Leggett 35
2.5.1 Introduction 35
2.5.2 Evaluation Criteria 44
2.5.3 Summary 46
2.6 O'Flaherty 48
2.6.1 Introduction 48
2.6.2 Evaluation Criteria 61
2.6.3 Summary 64
3.0 Summary and Recommendations 66
3.1 Summary of Model Reviews 66
3.2 Recommendations 68
-------
-REVIEW OF ADULT LEAD MODELS-
4.0 References 70
Appendix A. Comparison of Gastrointestinal Absorption Factors in Leggett and O'Flaherty
Models A-l
Appendix B. Comparison of Biokinetic Slope Factors (BKSF) for Children Predicted by the
IEUBK Model and O'Flaherty Model B-l
Appendix C. Biokinetic Slope Factors (BKSF) for Adults Predicted by the O'Flaherty Model ... C-l
Appendix D. Comparison of Biokinetic Slope Factors (BKSF) for Children Predicted by the
IEUBK Model and Leggett Models D-l
Appendix E. Biokinetic Slope Factor (BKSF) for Adults Predicted by the Leggett Model E-l
Appendix F. Modeling of Bone Lead and Bone Metabolism During Adolescence - O'Flaherty
Model F-l
Appendix G. Miscellaneous Issues Related to the O'Flaherty Model ACSL Code G-l
Appendix H. Miscellaneous Issues Related to the O'Flaherty Model C++ Code H-l
Appendix I. Mass Balance in O'Flaherty Model 1-1
Appendix J. Comparisons of Blood Lead Concentrations Predicted by the Leggett Model and
the EPA ALM J-l
Appendix K. Comparisons of Blood Lead Concentrations Predicted by the O'Flaherty Model and
the EPA ALM K-l
Appendix L. Comparisons of Blood Lead Concentrations Predicted by the Bert Model and the
EPA ALM L-l
-------
-REVIEW OF ADULT LEAD MODELS-
List of Tables
Table 1.1. Summary of Default Values for Variables in the EPA Adult Lead Methodology 5
Table 1.2. Models Considered by the TRW for Lead Risk Assessment 6
Table 1.3. Evaluation Categories for Adult Models 8
Table 2.1.1. Summary of Experimental Studies with Humans to Assess Clearance Rates of Lead
from Blood and Extracellular Fluid 11
Table 2.1.2. Inputs for Comparison Simulations of the Rabinowitz Model and ALM Shown in
Figure 2.1.2 13
Table 2.2.1. Media and Pathway-Specific Slope Factors used in the California Model 17
Table 2.2.2. Comparison of ALM With the California Model Industrial Scenario 19
Table 2.3.1. Inputs for Comparison Simulations of the Bert Model and ALM Shown in
Figure 2.3.2 23
Table 2.4.1. Simulations Executed to Evaluate the Hybrid Stern Model 28
Table 2.4.2. PRGs Associated with Default Point Estimate Exposure Assumptions Applied to the
Stern (1994) de minimis PbB Methodology 29
Table 2.4.3. PRGs Associated with Default Probability Density Functions for Exposure
Assumptions Applied to Stern (1994) de minimis PbB Methodology 30
Table 2.5.1. Inputs for Comparison Simulations of the Leggett Model and
ALM Shown in Figure 2.5.3 38
Table 2.5.2. Comparison of Lead Uptake Slope Factors Based on the IEUBK, ALM, and
Leggett Models 43
Table 2.6.1. Inputs for Comparison Simulations of the O'Flaherty Model and ALM Shown in
Figure 2.6.6 56
Table 2.6.2. Comparison of Lead Uptake Slope Factors (BKSF) for Adults Based on the ALM and
O'Flaherty Models 58
Table 2.6.3. Comparison of BKSFs Predicted from the IEUBK and
O'Flaherty Models 60
Table 3.1. Summary of Model Scores by Evaluation Category 66
-------
-REVIEW OF ADULT LEAD MODELS-
List of Figures
Figure 1.1. Decision Tree for Evaluating Alternative Models in Terms of Options for Retaining,
Replacing, or Modifying the Adult Methodology 2
Figure 2.1.1. Rabinowitz et al. (1976) BIOKINETIC MODEL FOR LEAD 11
Figure 2.1.2. Comparison of Blood Lead Concentrations Predicted by the Rabinowitz Model and
ALM 14
Figure 2.2.1. Conceptual Model of Lead Exposure and
Biokinetics in the California Model 16
Figure 2.3.1. Lead Body Burden Associated with Intakes of Lead to the Gastrointestinal and
Respiratory Tracts for a Typical Adult Male 21
Figure 2.3.2. Comparison of Blood Lead Concentration Predicted
by the Bert Model and the ALM 24
Figure 2.4.1. Cumulative Probability Distribution (CDF) of Output from Hybrid Stern Model,
Using Stern (1994) de minimis Blood Lead Concentration Model with Stern (1996)
Adult Inputs for PDF 32
Figure 2.4.2. Comparison of Soil Lead Concentrations (PRGs) Using EPA (1996) ALM and
Probabilistic Stern Model (See Table 2.4.3) as a Function of Baseline Blood Lead
Concentrations (PbB0) 33
Figure 2.5.1. Schematic of the Biokinetic Model of Lead Metabolism Proposed by Leggett (1993) 36
Figure 2.5.2. Default Values in the Leggett Model for Lead Absorption in the Gastrointestinal
Tract (Percent) as a Function of Age 38
Figure 2.5.3. Comparison of Blood Lead Concentrations Predicted by the Leggett Model and
ALM 40
Figure 2.5 Aa. Leggett Model Simulation of Baseline Blood Lead Concentration and Cortical and
Trabecular Bone Lead 41
Figure 2.5 Ab. Leggett Model Simulation of Baseline Blood Lead Concentration 42
Figure 2.5.5. Biokinetic Slope Factors (BKSFs) for Adults Predicted by the Leggett Model 43
Figure 2.6.1. Compartments and Pathways of Lead Exchange in the O'Flaherty Model 51
Figure 2.6.2. Bone Volume-Age Relationships for Human Juvenile and
Mature Bone as Represented in the O'Flaherty Model 53
Figure 2.6.3. Bone Volume-Age Relationships for Human Juvenile Cortical and Trabecular Bone
as Represented in the O'Flaherty Model 54
Figure 2.6.4. Default Values for Gastrointestinal Lead Absorption as a Function of Age as
Simulated in the O'Flaherty Model 55
Figure 2.6.5. O'Flaherty Model Simulation of Baseline Blood Lead (PbB) Concentration 57
Figure 2.6.6. Comparison of Blood Lead Concentration Predicted by the O'Flaherty Model and
ALM 59
Figure 2.6.7. Biokinetic Slope Factors for Adults Predicted by the O'Flaherty Model 61
-------
-REVIEW OF ADULT LEAD MODELS-
List of Equations
Equation 1.2.1 3
Equation 1.2.2 3
Equation 1.3.1 6
Equation 1.3.2 6
Equation 1.3.3 6
Equation 2.4.1 26
Equation 2.4.2 26
Equation 2.4.3 27
Equation 2.4.4 31
Equation 2.6.1 48
Equation 2.6.2 48
Equation 2.6.3 48
Equation 2.6.4 48
Equation 2.6.5 49
Equation 2.6.6 53
-------
-REVIEW OF ADULT LEAD MODELS-
List of Acronyms and Abbreviations
ACSL Advanced Continuous Simulation Language
AF absorption fraction
ALM Adult Lead Methodology
AM arithmetic mean
BKSF biokinetic slope factor
ED exposure duration
EF exposure frequency
EPA Environmental Protection Agency
GM geometric mean
GSD geometric standard deviation
IEUBK Integrated Exposure Uptake Biokinetic Model for Lead in Children
IR ingestion rate
NHANES National Health and Nutrition Examination Survey
ORD Office of Research and Development
Pb lead
PbB blood lead
PbB0 baseline blood lead
PbS soil lead
PDF probability distribution function
PRO preliminary remediation goal
RAF relative absorption fraction
SF slope factor
TRW Technical Review Workgroup for Lead
% percent
A delta
|j,g microgram
d day
dL deciliter
g gram
L liter
m3 cubic meters
mm millimeters
mo month
ppm parts per million
wk week
yr year
-------
-REVIEW OF ADULT LEAD MODELS-
Contributing Members of Adult Pb Sub-Committee of
U.S. Environmental Protection Agency
Technical Review Workgroup for Lead
CHAIR
Mark Maddaloni
Region 2
New York, NY
MEMBERS
Mary Ballew Kevin Koporec
Region 1 Region 4
Boston, MA Atlanta, GA
Mark Johnson Patricia Van Leeuwen
Region 5 Region 5
Chicago, IL Chicago, IL
Ghassan Khoury Larry Zaragoza
Region 6 Office of Emergency and Remedial
Dallas, TX Response
Washington, DC
-------
-REVIEW OF ADULT LEAD MODELS-
Executive Summary
In 1996, in response to the need for a scientifically defensible approach for assessing human health lead
risks at non-residential hazardous waste sites, the U.S. Environmental Protection Agency's (EPA's)
Technical Review Workgroup for Lead (TRW) developed the Adult Lead Methodology (ALM). The
ALM was released by the EPA as an interim report entitled Recommendations of the Technical Review
Workgroup for Lead (TR W) for an Interim Approach to Assessing Risks Associated with Adult Exposures
to Lead in Soil. The effort to provide interim guidance in a timely manner limited the scope of the
approach to adult workers and specific exposure media (i.e., soil/dust). Therefore, as a follow-up to the
ALM, a more exhaustive effort was undertaken to evaluate other currently available modeling
approaches and their potential applicability to assessing non-residential lead exposures and risks.
This report evaluates six lead biokinetic models that have been published in the peer-reviewed scientific
literature. These models have been used to assess the relationship between environmental lead exposures
and blood lead (PbB) concentration in adults. Each model was evaluated and compared to the ALM
using the following general evaluation criteria:
• Completeness of exposure module
• Kinetic performance
• Utility of model output
• Ease of use and flexibility
This report presents summaries of the model reviews and provides recommendations for whether:
• a superior model is currently available that could replace the existing ALM;
• a new model should be developed;
• the existing model should be modified;
• the ALM should be retained.
The models reviewed were slope factor and multi-compartmental models and included the California
Carlisle and Wade (1992), Stern (1994,1996), Rabinowitz (1976), Bert (1989), Leggett (1993a,b, 1996),
and O'Flaherty (1993, 1995) models. All models reviewed showed strengths and weaknesses. The
models are discussed in greater detail in subsequent sections of this report. Although no single model
reviewed by the TRW was judged to be a significant improvement over the ALM, various components
from the different models were determined to offer refinements in adult lead modeling. These
components could be integrated into a hybrid model; however, such modifications would require a long-
term effort (i.e., months to years). The decision not to proceed with development of a hybrid model is
discussed below. However, in lieu of such an effort, it is worth noting that the kinetic performance of all
the models produced similar estimates of quasi-steady-state PbB concentrations when exposure
parameters were normalized across models (i.e., all were set to approximate ALM inputs).
While model evaluations were being conducted by the TRW, the EPA Office of Research and
Development (ORD) initiated an effort to develop an All Ages Lead Model capable of simulating
multimedia exposures and biokinetics over the entire human lifespan. The TRW has long contemplated
the expansion of the Integrated Exposure Uptake Biokinetic Model for Lead in Children (IEUBK model)
to accommodate a wider age range of exposure. Conceptually, the goals of the All Ages Lead Model are
similar to the IEUBK model in that it could have an integrated multimedia exposure module and a
relatively complex (e.g., multi-compartment) biokinetic module.
-------
-REVIEW OF ADULT LEAD MODELS-
To support the All Ages Lead Model effort and to minimize overlapping of EPA's goals for developing
lead risk assessment models, the TRW elected to direct recommendations for model improvements based
on the adult model reviews to the All Ages Lead Model initiative.
Based on the TRW evaluations described in this report and recognition that the All Ages Lead Model is
currently under development, the TRW has chosen to retain the ALM as an interim tool for assessing
soil-borne lead risks for non-residential exposure scenarios. In addition, to assist users with the
application of the ALM, the TRW has prepared a fact sheet providing answers to frequently asked
questions and may incorporate modifications to model input parameters as new information becomes
available. The TRW recognizes that in specific situations such as: (1) short-term exposures or
intermittent exposure scenarios; (2) varied age range (e.g., trespasser, recreational); and (3) and/or
multimedia contamination, the risks of lead exposures may be more amenable to assessment by
alternative models, such as those described in this report. As previously mentioned, the TRW will
provide support to the All Ages Lead Model effort.
-------
-REVIEW OF ADULT LEAD MODELS-
1.0 Introduction
1.1 PURPOSE OF REVIEW
In December 1996, the Technical Review Workgroup for Lead released the report,
Recommendations of the Technical Review Workgroup for Lead (TRW) for an Interim Approach
to Assessing Risks Associated with Adult Exposures to Lead in Soil. The report filled a direct
need for a scientifically defensible approach for assessing soil-borne lead hazards at
non-residential hazardous waste sites. Though the report provided readily available guidance, it
limited the scope of the approach to a narrowly defined receptor population (i.e., an adult female
worker of child-bearing age at a non-residential site or with non-residential exposure scenarios)
and specific media (i.e., soil/dust). The Adult Lead Methodology resulting from this study was
offered as an interim approach pending a more detailed effort to identify the most scientifically
defensible approach for modeling non-residential lead exposure over a broader age range. In
performing a more exhaustive evaluation, the TRW initiated a review of currently existing, peer-
reviewed lead biokinetic models. Specifically, as the first objective, the TRW focused on
addressing the following options:
• Is there a single existing model that should replace the ALM as the recommended
approach for non-residential risk assessment?
• Is there a hybrid model that could be constructed with minimal effort that would
represent an improvement?
• Should the ALM be retained as the primary model for assessing risks to adults
associated with non-residential lead exposures or should the TRW recommend
development of a new model?
The options above formed the basis of a decision tree used to determine whether to retain,
replace, or modify the ALM (Figure 1.1).
-------
-REVIEW OF ADULT LEAD MODELS-
FIGURE 1.1. DECISION TREE FOR EVALUATING ALTERNATIVE MODELS IN TERMS OF
OPTIONS FOR RETAINING, REPLACING, OR MODIFYING THE ADULT METHODOLOGY
Should an existing
modelreplace ALM ?
Fes
Replace ALM with
single b est m o d el
No
Should a hybrid modelbe
developed to replace ALM ?
Fes
Develop new model
No
Fes
Should the ALM be modified?
M o d ify A L M
No
R eta in A L M w itho u t
mod ific a tie ns
While model evaluations were being conducted by the TRW, the ORD initiated an effort to develop an
All Ages Lead Model capable of simulating multimedia exposures and biokinetics over the entire human
lifespan. Conceptually, the goals of the All Ages Lead Model are similar to the IEUBK model in that it
could have an integrated multimedia exposure module and a relatively complex (e.g., multi-compartment)
biokinetic module. To support the All Ages Lead Model effort and to minimize overlapping of EPA's
goals for developing lead risk assessment models, the TRW elected to provide recommendations for
model improvements to the All Ages Lead Model initiative.
Specifically, the TRW defined the following additional objectives for the model reviews:
• Identify the research and data needs relevant to achieving an optimal model.
• Identify potentially useful features of other models that might be relevant to the development
of the All Ages Lead Model.
In order to provide the detailed information necessary to support the development of the All Ages Lead
Model and to provide technical users with specific information regarding how the models were evaluated
(e.g., how inputs to the Leggett model were derived in order to create a baseline blood lead
concentration), the TRW is preparing several technical appendices to this report. The appendices are
internal working documents that are available from the TRW upon request. In addition, the TRW
recognizes that environmental models are constantly evolving. To the extent possible, the TRW will
review these revised models and address their relative strengths and weaknesses in the technical
appendices to this report.
-------
-REVIEW OF ADULT LEAD MODELS
1.2 EPA INTERIM ADULT LEAD METHODOLOGY
The ALM is a modified version of a similar approach, described by Bowers et al. (1994), which was used
by EPA Region 8 in the assessment of the California Gulch Superfund site in Colorado (Weston 1995).
The ALM uses a biokinetic slope factor (BKSF) to represent lead biokinetics and a relatively simple
exposure model in which all exposure pathways, other than soil ingestion, are represented by a
background PbB concentration. The model is implemented with the following algorithms:
RBC = PbS - (PbBadultMgoal - PbBadult:J - AT
*U"adult,central,goal ~
(BKSF • IRS • AFS •
fetal.Q35.goal
Equation (1.2.1)
Equation (1.2.2)
i.adult ' ^-fetal/maternal
where:
RBC
PbBadult,0
PbS
AT
BKSF
AF,
Phft =
^ adult, central, goal
* ""fetal, 0.95, goal
risk-based concentration (RBC); soil lead concentration (PbS) that
would be estimated to result in a specified central tendency PbB
concentrations in adults (i.e., women of child-bearing age) at the
site (PbBadultcentralgoal) and corresponding 95th percentile fetal PbB
concentration (PbB fetlliOSSigoJ)
typical PbB concentration (|ig/dL) in women of child-bearing age
at the site in the absence of exposures to the site that is being
assessed
soil lead (|ig/g)
averaging time; the total period during which soil contact may
occur (365 days/year for continuing long term exposures)
ratio of (quasi-steady state) increase in typical adult PbB
concentration to average daily lead uptake (|_ig/dL PbB increase
per |j,g/day lead uptake)
intake rate of soil, includes both outdoor soil and indoor
soil-derived dust (g/day)
absolute gastrointestinal absorption fraction for ingested lead in
soil and lead in dust derived from soil (dimensionless)
exposure frequency for contact with assessed soils and/or dust
derived in part from these soils (days of exposure during the
averaging period); exposure frequency can be considered days per
year for continuing, long-term exposure
goal for central estimate of blood lead concentration (|ig/dL) in
adults (i.e., women of child-bearing age) that have site exposures;
the goal is intended to ensure that PbB fetal 0 95 goal does not exceed
10ng/dL
goal for the 95th percentile blood lead concentration (|ig/dL)
among fetuses of women having exposures to the specified site
soil concentration; this is interpreted to mean that there is a
95 percent likelihood that a fetus, in a woman who experiences
such exposures, would have a blood lead concentration no greater
than PbB fetal 095 goal (i.e., the likelihood of a blood lead con-
-------
-REVIEW OF ADULT LEAD MODELS-
centration greater than 10 |-ig/dL would be less than 5 percent for
the approach described in this report)
GSDt adult = estimated value of the individual geometric standard deviation; the
GSD among adults (i.e., women of child-bearing age) that have
exposures to similar on-site lead concentrations, but that have
non-uniform response (intake, biokinetics) to site lead and
non-uniform off-site lead exposures
^fetai/matemai = constant of proportionality between fetal PbB concentration at
birth and maternal PbB concentration.
The default values for the variables in Equations 1.2.1 and 1.2.2 are presented in Table 1.1.
-------
-REVIEW OF ADU LT LEAD MODELS-
TABLE 1.1. SUMMARY OF DEFAULT VALUES FOR VARIABLES IN THE
EPA ADULT LEAD METHODOLOGY"
Variable
PbBfetalj0095igoal
GSDijadl]]t
n
fetal/maternal
PbBadllltj0
BKSF
IRS
EFS
AFS
Value
10
1.8-2.1
0.9
1.7-2.2
0.4
0.05
219
0.12
Unit
M-g/dL
-
-
|ig/dL
Lig/dL per
fig/day
g/day
days/year
-
Comment
Used to estimate RBCs based on risk to the developing
fetus.
Value of 1.8 is recommended for a homogeneous
population, while 2.1 is recommended for a more
heterogeneous population.
Based on Goyer (1990) and Graziano et al, (1990).
Plausible range based on NHANES III phase 1 for
Mexican-American, non-Hispanic black, and white women
of child bearing age (Brody et al., 1994). Point estimate
should be selected based on site-specific demographics.
Based on analysis of Pocock et al., (1983) and Sherlock et
al., (1984) data.
Predominantly non-residential exposures to indoor soil-
derived dust rather than outdoor soil
(0.05 g/day=50 mg/day).
Based on EPA (1993) guidance for average time spent at
work by both full-time and part-time workers.
Based on an absorption factor for soluble lead of 0.20 and
a relative bioavialability of 0.6 (soil/soluble).
"Variables refer to Equations 1 and 2 in text, as described in EPA (1996).
1.3 SELECTION OF MODELS
Seven models were identified from scientific literature searches or brought to the TRW's attention as the
result of reviews of lead risk assessments conducted in the various EPA regions (see Table 1.2). Six
models were selected for review by the TRW. The Bowers et al. (1994) model was not included in this
review effort, because of the similarity between the Bowers and ALM models. The basic algorithms for
the Bowers model were used for the California Gulch site and form the basis for the current ALM model.
As indicated in Table 1.2, the models can be broadly organized into two categories based on the
conceptual approach used to represent lead biokinetics in each model.
-------
-REVIEW OF ADU LT LEAD MODELS-
TABLE 1.2. MODELS CONSIDERED BY THE TRW FOR LEAD RISK ASSESSMENT
Model
ALMa
California13
Stern0
Rabinowitzd
Berte
Leggettf
O'Flaherty6
Biokinetics Modeling Approach
Slope Factor
X
X
X
Multi-compartmental
X
X
X
X
aALM modified from Bowers et al, 1994
"Carlisle and Wade, 1992
"Stern 1994, 1996
"Rabinowitz et al., 1976
"Bert etal, 1989
fLeggettetal., 1993a,b, 1996
8O'Flahertyl993, 1995
Slope Factor Models
For the slope factor models (i.e., Bowers, California, Stern), PbB concentration is represented as a simple
linear relationship between PbB concentration and lead uptake or intake in Equation 1.3.1 and 1.3.2:
where:
APbB = APbUptake • BKSF
= kPblntake • SFi
APbB
APbUptake
APblntake
SF,
BKSF
Equation (1.3.1)
Equation (1.3.2)
increase in blood lead concentration (|j,g/dL)
increase in the rate of lead absorption (|j,g/day)
increase in the rate of lead intake (|j,g/day)
slope factor; an empirically-based estimate of the slope of the
linear relationship between PbB concentration and lead intake or
uptake (|-ig/dL per |j,g/day). Slope factors may be intake (SF:) or
uptake (BKSF)
biokinetic slope factor, an empirically-based estimate of the slope
of the linear relationship between PbB concentration and lead
uptake (|-ig/dL per |j,g/day)
In this report, the slope factor in Equation 1.3.1 is referred to as an uptake or biokinetic slope factor
(BKSF) because it reflects the biokinetics of absorbed, rather than ingested, lead. The BKSF is used in
combination with a separate parameter for the lead absorption fraction (AF) to calculate PbB
concentration in Equation 1.3.3:
APbB = hPblntake • AF • BKSF
Equation (1.3.3)
-------
-REVIEW OF ADULT LEAD MODELS-
The slope factor in Equation 1.3.2 is referred to as an intake slope factor because it is based on ingested
rather than absorbed lead and, thus, reflects a combination of lead absorption and biokinetics. The ALM
is a slope factor model which uses a BKSF with a separate AF to calculate PbB concentrations
(Equation 1.3.3). This approach allows for explicit adjustment of the AF.
Multi-compartmental Models
Multi-compartmental models, such as the Rabinowitz, Bert, Leggett, and O'Flaherty models, simulate
lead biokinetics as one or several interconnected tissue compartments that exchange lead via a central
blood or plasma compartment. Two approaches have been used to model the exchanges of lead between
tissues and the central compartment. In the Rabinowitz, Bert, and Leggett models, exchanges are
represented as first-order rate constants for the transfer of lead across compartment boundaries. Thus,
these models are also referred to as transport-limited or diffusion-limited models because the rates of
change of lead masses in the various compartments are assumed to be governed by rates of transfer
across compartment boundaries. Compartment lead concentrations are determined from the lead masses
and compartment volumes.
An alternative to simulating inter-compartmental exchanges as transport-limited processes is to simulate
flow-limited exchanges. In a typical flow-limited model, the central compartment (usually plasma) is
represented as a dynamic process that is characterized by volume and flow (sometimes other char-
acteristics, such as binding interactions and subcompartments) rather than as a static volume. The flows
reflect actual rates of flow to the tissues represented in the model. Lead is assumed to instantaneously
partition between plasma and soft tissues and to archive an equilibrium (i.e., partition coefficient).
Therefore, the rates of change of lead masses in soft tissues are limited by the rates of delivery of lead to
the tissues, given by the product of the plasma concentration of lead and the rate of plasma flow to the
tissue, rather than by limiting steps in the transfer of lead across tissue boundaries. In the O'Flaherty
model, exchanges between plasma and soft tissues are simulated as flow-limited processes, whereas
exchanges between plasma and bone are represented as a combination of flow-limited and diffusion-
limited processes.
1.4 MODEL REVIEW CRITERIA AND PROCESS
The model reviews focused on four evaluation categories: Category 1, Exposure; Category 2, Kinetic
performance; Category 3, Output; and Category 4, Ease of use and flexibility of the model for application
at hazardous waste sites. Examples of considerations for each category are presented in Table 1.3.
The model review process consisted of the following phases:
• Model procurement and implementation. Documentation on each model was obtained
and reviewed, and computation platforms for simulations were established. The latter
was achieved by procuring computer code when available (e.g., Leggett, O'Flaherty), or
by developing a computer code when not available (e.g., Rabinowitz, Bert).
• Preliminary summary and comparisons. The main features of each model were
summarized, compared, and contrasted to the ALM. Standard exposure scenarios were
defined for initial comparison simulations. The purpose of these simulations was to
compare the output of each model when similar inputs were used. The scenarios were
intended to represent adult exposures to soil at either 1,000 or 10,000 ppm, with an
exposure frequency of either 1 or 5 days/week. Additional simulations were run as
needed to explore specific issues identified in workgroup discussions.
-------
-REVIEW OF ADU LT LEAD MODELS-
Narrative discussion. The summary narrative for each model was discussed in
workgroup sessions. The discussions focused on describing the conceptual and
computational structure, results of simulations that demonstrated important aspects of
model performance, strengths and weaknesses, and applicability of each model to the All
Ages Lead Model.
Semi-quantitative ranking. For each evaluation criterion, each model was relatively
ranked and each received a (+) for outperforming or a (-) for underperforming the ALM.
A score of 0 meant that a model performed similarly to the ALM. A positive score
indicated that the model provided an improvement over the ALM in a given category;
whereas, a negative score indicated that the model was more limiting than the ALM in a
given category.
TABLE 1.3. EVALUATION CATEGORIES FOR ADULT MODELS
Category 1: Exposure
Source Identification
Evaluated comprehensiveness of exposure to accommodate all
environmental media, age groups, and exposure duration.
Category 2: Kinetic performance
Nature of fit to available data
Plausibility
Saturation
Predictability
Kinetic performance was compared to and with available data.
There is a biological basis for the model.
The model accommodates saturation kinetics.
When tested with different combinations of exposure, frequency, intensity,
and duration, the model performs as expected.
Category 3: Output
Completeness
Decision making
The output provides an adequate and full summary of how the model was
run and the specific endpoints.
The model provides sufficient information for decision making at
hazardous waste sites.
The model can run both to predict risk and calculate preliminary
remediation goal (PRG) values.
Category 4: Ease of use/flexibility
Ease of use
Clarity
Flexibility
The model assesses site-specific exposures as specified by the Superfund
program.
The model variables/inputs can easily be changed.
The model application is easy to understand. For example, the help screens
aid the user in applying the model.
The model code could be easily changed or updated.
The model components could be easily changed or updated.
-------
-REVIEW OF ADU LT LEAD MODELS-
This page is intentionally left blank.
-------
10
-REVIEW OF ADULT LEAD MODELS
2.0 Model Review Summaries
2.1 RABINOWITZ
2.1.1 Introduction
Description of Biokinetics
The Rabinowitz et al. (1976) model simulates changes in PbB concentrations in adult males in response
to lead uptakes. The model is based on data collected from five healthy subjects who received oral doses
of stable lead isotopes for various periods of time (see Calibration and Evaluation sub-section). The
model includes three compartments representing kinetically distinct lead pools in the body. The central
compartment (Pool 1) represents whole blood and other extracellular fluids that rapidly equilibrate with
whole blood. The apparent volume of the central compartment is approximately 1.5-2.2 times the whole
blood volume, and the compartment contains approximately 1 percent of the total lead body burden; it
has the shortest half-life (36±5 days) of the three pools. Lead in the central compartment exchanges
directly with Pools 2 and 3. Pool 2 includes primarily those soft tissues, which equilibrate more slowly
with whole blood and possibly parts of the skeleton where more active (and relatively rapid) exchanges
occur with the central compartment (i.e., rapid relative to exchanges with Pool 3). Pool 2 represents less
than 0.5 percent of the lead body burden, and relatively little of the lead in this pool is returned to blood.
Pool 3 represents approximately 98-99 percent of the lead body burden and is assumed to include bone
and other slowly exchanging tissues. Approximately 54-78 percent of the lead which leaves the body
each day in urine is assumed to come from the central compartment, while other excretion pathways
(e.g., bile, hair, sweat, and nails) are assumed to originate from Pool 2.
Rabinowitz et al. (1976) reported values for the lead pool masses and rates of movement of lead between
pools (|j,g/day); however, the actual rate constants (k) for exchanges between the various pools that were
used in the model were not reported. First order rate constants for lead exchanges were derived from
isotopic composition measurements and from mass balance data obtained from the study subjects. The
values shown in Figure 2.1.1 were calculated for the purpose of implementing the model, and are the
ratios of the rates of movement between pools and the pool lead masses. The rates of lead movement
between pools were assumed by Rabinowitz et al. (1976) to be independent of the pool lead masses in
the subjects studied. This assumption may only apply to the relatively narrow range of PbB
concentrations observed in the subjects, 17-25 |-ig/dL (range of means). The sizes of the pools were also
assumed to be constant over the period of study (i.e., there is mass equilibrium), due to the slow turnover
of Pool 3. These assumptions would imply that rates of lead transfer between compartments and lead
excretion vary linearly with lead uptake over the range of intakes and lead body burdens examined in the
study. The relationship between lead intake and uptake is not given explicitly in the Rabinowitz et al.
(1976) model, although Rabinowitz estimated gastrointestinal absorption in the 1976 study (see Uptake
sub-section). The relationship between intake, PbB concentration, and excretion predicted by the
Rabinowitz model based on the adult male subjects may not apply to females during periods of
physiological change, such as adolescence, pregnancy, and menopause, in which the blood pool size or
bone turnover rates may be very different from adult males.
-------
3
Bone
, = 3.7 x 10-3 d-1
k=1.9x10-2d-1
1e
1
Blood
1*1,
Uhne
2
Soft
Tissue
Bile, Hair
Sweat, Nails
FIGURE 2.1.1. RABINOWITZ ef a/. (1976) BIOKINETIC MODEL FOR LEAD
Although not reported in Rabinowitz et al. (1976), the reported estimate for lead residence times and
sizes of Pool 1 allow calculation of a clearance (dL/day) of lead from Pool 1 (see Table 2.1.1). These
values are consistent with an estimated BKSF of 0.4 |-ig/dL per |j,g/day absorbed lead.
TABLE 2.1.1. SUMMARY OF EXPERIMENTAL STUDIES WITH HUMANS TO ASSESS CLEARANCE
RATES OF LEAD FROM BLOOD AND EXTRACELLULAR FLUID
Subject
A
B
C
D
E
Mean±SD
Vx'
(dL)
74
100
101
99
113
97±14
T-> b
(day)
34
40
37
40
27
36±5
TC
1/2
(day)
24
28
26
28
19
25±4
C/
(dL/day)
2.2
2.5
2.7
2.5
4.2
2.8±0.8
BKSF
(d/dL)
0.46
0.40
0.37
0.40
0.24
0.37±0.8
aThe volume of Pool 1, which refers to blood and a rapidly exchangeable extracellular fluid compartment. Mass units for the
compartment, reported in Rabinowitz et al., 1976, were converted to units of volume, (assuming that one kg of blood has an
approximate volume of 10 dL).
"The reported residence time for lead in Pool 1.
"The half-time of lead in Pool 1; Tlfl=(T1)xIn(2).
"Clearance of lead from Pool 1; Q = V^.
-------
12
-REVIEW OF ADULT LEAD MODELS
Description of the Exposure Component
The model does not allow for the consideration of separate intakes of lead from environmental exposure
concentrations. Input to the model is total lead uptake (e.g., |j,g/day) from dietary and atmospheric
sources.
Description of Uptake
Uptake of lead into the central compartment from dietary or atmospheric sources is not modeled
separately; the combined contributions from both sources are considered to be a single input variable.
Absorption of lead tracer from the gastrointestinal tract was estimated in the five subjects from the
difference between ingested lead tracer, total fecal excretion, and endogenous fecal excretion.
Endogenous fecal excretion was estimated by measuring the amounts of 204Pb tracer and isotopic
composition of tracers in the excreta (salivary, gastric, biliary, and pancreatic secretions). Estimates of
gastrointestinal absorption of ingested lead in the five fed subjects ranged from 6.5 to 13.7 percent
(mean=9.7 percent). Endogenous fecal excretion was estimated to be approximately 0.5 percent of
intake. In a subsequent stable isotope tracer study, gastrointestinal absorption of lead was estimated in
fed and fasted subjects (Rabinowitz et al, 1980). A tracer dose of lead nitrate, cysteine or sulfide (form
did not affect absorption) along with carrier lead (144-221 |j,g) was given at mid-point of meals or on the
9th hour of a 16-hour fast; absorption estimates were based on modeling of the kinetics of PbB
concentration and fecal lead excretion. Absorption in fed subjects was 8.2±2.8 percent and in fasted
subjects was 35±13 (based on nine estimates on five subjects).
Calibration and Evaluation
The model is based on data on PbB isotopic concentrations and excretion kinetics in five healthy males
(ages 25-53 years). Subjects were fed diets of constant lead content for periods of 10-210 days and
received doses of 204Pb or 207Pb as lead nitrate in water with each meal to restore their total dietary intake
of lead to the pre-study levels. The men were maintained (not confined) in a metabolic unit and had
additional exposure to ambient lead concentrations in air. Three of the subjects were moved to units in
which the air was filtered to remove the lead contribution from atmospheric air; this contribution to the
total lead intake was supplemented by the administration of a second oral tracer, 207Pb as lead nitrate.
The contribution to the total lead input from atmospheric lead was estimated from isotopic composition
measurements.
The model was calibrated to achieve agreement between observed and predicted PbB concentrations in
the five subjects in the tracer study. The subjects had average PbB concentrations for the time (early
1970s) that ranged from 17 to 25 |_ig/dL. The model has not been validated against an independent set of
observations, nor has it been validated for predicting PbB concentrations associated with lower exposure
levels (e.g., exposures corresponding to PbB concentrations less than 10 |-ig/dL) or higher PbB
concentrations that might be encountered with non-residential exposures.
Simulation
The model is represented mathematically as a series of coupled first order differential equations
(Rabinowitz et al., 1976). These equations were implemented either in an Excel spreadsheet using a
forward Euler function for numerical integration, or in Advanced Continuous Simulation Language
(ACSL) using a 4th order Runge-Kutta function for numerical integration, with a 1-day time step; both
approaches yielded nearly identical results.
-------
13
-REVIEW OF ADULT LEAD MODELS-
The model yields estimates of quasi-steady state maternal PbB concentrations that are very close to those
predicted by the ALM if similar values for the soil lead absorption fraction (AFS) and soil ingestion rate
(IRS) variables are used to estimate lead uptakes, and if the same values for the baseline PbB
concentration (PbB0) are assumed. Table 2.1.2 provides the inputs that were used in a typical set of
simulations in which the outputs of the ALM and Rabinowitz models were compared. The soil lead
exposure was assumed to be to 1,000 |ig/g, 5 days/week for 260 days/year. An IRS of 0.05 g/day and a
gastrointestinal lead AF of 0.12 were assumed. Note, the latter two parameters are not components of the
Rabinowitz model; however, they were used to calculate lead uptakes that would be equivalent to those
simulated in the ALM for the same soil lead exposure. PbB0 was introduced into the Rabinowitz model
simulations as a constant daily uptake (4 |j,g/day) that yielded a quasi-steady state PbB concentration of
2 |-ig/dL in the absence of exposure to the designated soil lead level. After the PbB0 concentration was
achieved, the simulation was continued with a daily lead uptake equal to the sum of the baseline uptake
(4 |j,g/day) and the soil lead uptake (6 |j,g/day) (i.e., the product of the soil lead concentration, 1,000 |ig/g;
IR,,, 0.05 mg/day; and AF, 0.12). Figure 2.1.2 shows the adult PbB concentrations predicted by the ALM
together with the PbB concentrations predicted by the Rabinowitz (1996) model. The ALM predicted a
quasi-steady state PbB concentration of 3.7 |-ig/dL, whereas the Rabinowitz (1996) model predicted a
slightly higher value of 4.1 |j,g/dL. If an AF of 0.1 was assumed in the Rabinowitz (1996) model, as
estimated for the five subjects on which the model was based (Rabinowitz et al, 1976), the model yields
a quasi-steady state PbB concentration of 3.7 |-ig/dL, which agrees nearly exactly with the prediction
from the ALM. The ALM also calculates a 95th percentile fetal PbB concentration, which is not shown in
the figure. However, if the same fetal/maternal ratios and variability model (i.e., the GSD) were applied
to the central tendency estimate of PbB concentration, the models would yield similar values for the
95th percentile fetal PbB concentration.
TABLE 2.1.2. INPUTS FOR COMPARISON SIMULATIONS OF THE RABINOWITZ MODEL AND
ALM SHOWN IN FIGURE 2.1.2
Parameters
PbS
IRS
AFS
PbB0
EF
ED (exposure duration)
Output
ALM
l,OOOLlg/g
0.05 g/day
0.12
2 Llg/dL
5 days/week (260 days/year; model
default is 219 days/year)a
Not a parameter in the model;
duration sufficient to achieve quasi-
steady state is assumed
Adult PbB concentration
Rabinowitz
1,000 |-lg/g (not a parameter in the model)
0.05 g/day (not a parameter in the model)
0.12 (not a parameter in the model)
Model was iterated with a daily uptake that
yielded a quasi-steady state PbB concentration
of 2 |-lg/dL (4 fig/day), after which, the model
was iterated with a daily uptake equal to the
sum of 4 fig/day and the product PbS*
IRS*AFS.
5 days/week (260 days/year)
1 year beginning on day 365 of simulation
Adult PbB concentration
aThe default exposure frequency for the ALM is 219 days/year; however, the assumption of 260 days/year in the
simulations would not change the outcome of the model comparisons.
-------
14
-REVIEW OF ADULT LEAD MODELS-
FIGURE 2.1.2. COMPARISON OF BLOOD LEAD CONCENTRATIONS PREDICTED BY THE
RABINOWITZ MODEL AND ALM
5.0
D.D
720
8 10
Note: See text and Table 2.1.2 for details on the inputs used in the model simulations.
2.1.2 Evaluation Criteria
Biokinetics
The Rabinowitz (1996) model is the least complicated of the compartmental models examined, in that
lead biokinetics are described in terms of three lead pools and two excretory pathways. While these
pools do not explicitly represent specific tissues, the slow pool is assumed to include bone and other
tissues that exchange lead very slowly with the central compartment. Sub-compartments within bone are
not represented. The model was developed to predict PbB concentrations associated with long-term lead
intakes, and it appears to do so reasonably well for the calibration data set. However, the model also will
calculate PbB concentrations associated with intermittent or non-steady state exposure conditions,
although the validity of any predictions would need to be further evaluated.
The model is completely linear; all lead transfer coefficients are constants. This approach appeared to
adequately predict PbB concentrations in the subjects that were used to calibrate the model over a
limited, but relatively high, exposure range (i.e., PbB concentrations of 17-25 |_ig/dL). This assumption
may not hold for lower or higher level exposures. Other models have assumed a limited capacity of red
blood cells to accumulate lead; this results in a curvature of the lead uptake-PbB concentration
relationship as PbB concentration approaches and exceeds approximately 25-30 |-ig/dL (e.g., Leggett and
O 'Flaherty models).
-------
15
-REVIEW OF ADULT LEAD MODELS
The model was calibrated to represent the biokinetics of adult males and may not adequately represent
the biokinetics of adolescents or females (e.g., different body masses); or biokinetic changes associated
with physiological status (e.g., adolescent growth, pregnancy).
Exposure
The exposure component is daily lead uptake. The model does not allow calculation of intakes from
environmental exposure levels (e.g., soil or dust lead concentrations). However, it would be relatively
easy to link an exposure model to the Rabinowitz model.
Although the model was calibrated with data on PbB concentrations, the model calculates lead masses in
a slow (e.g., bone) and fast exchange pool and in urine and other excreta combined.
Ease of Use/Flexibility
The model is relatively easy to understand and can be readily implemented on a variety of computational
platforms (e.g., spreadsheets). An exposure module that calculates lead intakes from environmental
exposure levels could be easily linked to the model. The model could be expanded to include additional
compartments (see Bert model, Section 2.3).
2.1.3 Summary
The Rabinowitz model is the least complicated of the compartmental models examined and has provided
a basis for other more complex models (e.g., Bert model). The Rabinowitz model was designed and
calibrated to predict quasi-steady state PbB concentrations corresponding to long-term exposures. Blood
lead concentrations corresponding to intermittent exposures can be easily calculated with the model;
however, the validity of these predictions would need to be determined. The quasi-steady state PbB
concentrations predicted by the model compare well with the ALM, if the same assumptions are made
about soil ingestion and maternal lead transfer.
Limitations of the Rabinowitz model include the following:
• Parameter values are for adult males and are not age-specific; therefore historical
exposures to lead during infancy, childhood, or adolescence cannot be simulated.
• Changes in lead biokinetics that may occur during pregnancy are not simulated.
• Exposure and uptake are not modeled; however, external models could be linked to the
biokinetic model.
• Variability is not modeled; however, any external model for variability could be linked to
the biokinetic model as in the ALM.
-------
16
-REVIEW OF ADULT LEAD MODELS-
2.2 CALIFORNIA
2.2.1 Introduction
The California model predicts adult and child PbB concentrations for a residential exposure scenario and
adult PbB concentration for an industrial exposure scenario. Specific exposure media and pathways are
evaluated independently using intake estimates from the ingestion of lead from dietary sources, drinking
water, soil, dust ingestion, inhalation of air-borne lead, and direct dermal contact with lead in soil. Lead
absorption and biokinetics are represented as intake-PbB concentration slope factors for each pathway.
The conceptual model is shown in Figure 2.2.1.
FIGURE 2.2.1. CONCEPTUAL MODEL OF LEAD EXPOSURE AND
BIOKINETICS IN THE CALIFORNIA MODEL
Blood
C ompartment
Lead absorption and biokinetics are represented as medium- and pathway-specific intake-PbB
concentration slope factors which relate the incremental change in lead intake to an incremental change
in the quasi-steady state PbB concentration. These slope factors are referred to in this report as intake
slope factors to distinguish them from the biokinetic slope factor (BKSF) which applies to the lead
uptake (absorption)-PbB concentration relationship.
The model provides estimates of percentiles of PbB concentration (50th to the 99th). The relative
contribution of each medium and exposure pathway is calculated and displayed as a percentage of the
total PbB concentration. Using a specific target PbB concentration, the preliminary remediation goal
(PRO) is back-calculated from specified or default intake estimates for other lead sources.
-------
17
-REVIEW OF ADU LT LEAD MODELS-
Description of Biokinetics
The California model calculates PbB concentrations using pathway-specific intake-PbB concentration
slope factors (see Table 2.2.1). PbB concentrations corresponding to each exposure pathway are summed
to estimate a multi-pathway geometric mean of a lognormal distribution of PbB concentrations having a
GSD of 1.42. The resulting probability distribution (i.e., mean and standard deviation) is used to
calculate percentiles.
TABLE 2.2.1. MEDIA AND PATHWAY-SPECIFIC SLOPE FACTORS USED
IN THE CALIFORNIA MODEL
Intake Pathway
Intake Slope Factor
(|ig Pb/dL blood) (|ig Pb/day)
Reference
Dietary
Child: 0.16
Adult: 0.04
(Pb concentration in plants is
0.045 percent soil Pb
concentration)
Chaney et al., 1982; plant uptake
study
Drinking water
Child: 0.16
Adult: 0.04
EPA, 1986
Soil and dust ingestion
Child: 0.07
Adult: 0.018
Chaney et al., 1990; 0.44 ratio of
soil Pb to Pb acetate uptake from
diet in rats
Inhalation
Child: 1.92 (|J,g/dL blood)
(|Ig/m3 air)
Adult: 1.64 (|J,g/dL blood)
(|Ig/m3 air)
EPA, 1986
Dermal contact
0.0001 (|ig Pb/dL blood)
(|Ig dermal Pb/day)
Adjustment of ingestion slope
factor by ratio of dermal absorption
(0.06 percent; Moore et al., 1980)
to oral absorption (11 percent;
ATSDR, 1990)
Description of the Exposure Component
The California model simulates inhalation of lead in airborne dust; ingestion of lead in soil; dermal
contact with lead in soil; ingestion of lead in the diet, including lead incorporated into plants from the
soil; and ingestion of lead in drinking water. Each exposure pathway is represented as the product of an
exposure concentration and a contact rate. Default exposure variables are provided for three scenarios:
adult residential exposures, child residential exposures, and adult industrial exposure.
Description of Uptake
Lead absorption and the biokinetics of absorbed lead are represented in the intake-PbB concentration
slope factors assigned to each medium and pathway. The amounts (or rates) of lead absorbed from the
oral, dermal, and inhalation pathways are not explicitly calculated in the model.
-------
18
-REVIEW OF ADULT LEAD MODELS
Calibration and Evaluation
The output from the California model does not appear to have been compared to specific set of empirical
data. Carlisle and Wade (1992) compared the output of the California model with the output of the
IEUBK model and concluded that the two models responded similarly to varying levels of lead in diet,
soil, and drinking water, and did not differ markedly in their predictions of PbB concentrations for the
inputs compared.
Simulation
The California model can be implemented in a spreadsheet. The predicted output of the California model
as presented in Carlisle and Wade (1992) is significantly different from the estimated PbB concentrations
from the ALM. However, this model can be adjusted to achieve agreement with the quasi-steady state
PbB concentrations predicted by the ALM (see Table 2.2.2). In Simulation 2, the California model
predicted 50th and 95th percentile PbB concentrations of 3.6 |-ig/dL and 10.9 |-ig/dL, respectively, which
were within the ranges predicted from the ALM (reflecting the ranges for the baseline PbB concentration
variable in the ALM). This was achieved by making the following adjustments to the California model:
• The exposure concentration for drinking water was assumed to be 4 |-ig/L, the IEUBK
model default value.
• Soil ingestion rate was assumed to be 0.05 g/day, the ALM default value.
• The intake-PbB concentration slope factor for drinking water was assumed to be 0.08,
which is equivalent to the assumptions in the ALM for soluble lead (the product of the AF
for soluble lead and BKSF is [0.08, 0.2x0.4=0.08]), (EPA, 1996).
• The intake-PbB concentration slope factor for soil and dietary lead was assumed to be
0.048 to correspond to the ALM (0.12x0.4=0.048).
• A GSD of 1.95 was assumed for the lognormal PbB concentration probability distribution,
the ALM average for the default range.
The resulting simulation predicted a PbB concentration of approximately 2.0 |-ig/dL, in the absence of the
soil ingestion pathway, corresponding to the 2.0 |-ig/dL baseline PbB concentration assumed in the ALM.
2.2.2 Evaluation Criteria
Biokinetics
As is true for other slope factor models, the California model does not include pharmacokinetic
parameters for estimating lead concentrations in physiologic compartments other than blood. It also
assumes linearity in the relationship of intake parameters with PbB concentrations. This assumption of
linearity could result in either an overestimate or underestimate of PbB concentration, depending on the
specific parameter and the soil lead concentration relative to the linear range.
The default value of the intake SF for soil lead in the California model is 0.018. The product of the ALM
values for the BKSF (0.4) and the AF for soil lead (0.12) is 0.048. Thus, relative to the ALM, the
California model assumes a lower value for the soil lead BKSF or soil lead AF, or lower values for both
variables. These differences presumably reflect the basis for the values in the models; the default value
for the soil lead intake SF in the California model was based on the results of rat studies; the soil lead
-------
19
-REVIEW OF ADU LT LEAD MODELS-
TABLE 2.2.2. COMPARISON OF ALM WITH THE CALIFORNIA MODEL INDUSTRIAL
SCENARIO
Parameters
PbS
IRS
Dust
Air cone.
Water cone.
AF
BKSF
PbB0
EF-ED
Maternal PbB
(Llg/dL)
GSD
Industrial
PRO (95th)
ALM
l,OOOLlg/g
0.05 g/day
-
-
-
AFS=0.12
0.4
2 Llg/dL
5 days/week-1 year
perc entile PbB f|J.g/dL)
50 3.1-3.6
95 9.4-10.9
1.951
926 ppm
California model
Simulation 1
(initial defaults)
l,OOOLlg/g
0.025 g/day
50 Llg/m3
0.1 Llg/m3
1 5 Llg/L (MCL)
AF x BKSF=route-specific
constant
0.04-water
0.082-air
0.018-soil and diet
0.000 11 -dermal
5 days/week-1 year
perc entile PbB (Lia/dL)
50 2.6
95 4.7
1.42
6,406 ppm
Simulation 2
(change IRS, BKSF,
[water], GSD)
1,000 Llg/g
AF x BKSF=route-specific
constant
0.08-water
0.082-air
0.048-soil and diet
0.00011-dermal
5 days/week-1 year
perc entile PbB (Llg/dL)
50 3.6
95 10.9
1.95
865 ppm
Notes:
AF = absorption fraction; BKSF = Biokinetic slope factor; GSD = Geometric standard deviation; IRS = Soil ingestion rate;
PbB0 = baseline blood lead concentration; PbS = Soil lead; PRO = preliminary remediation goal
Bold italics text indicate a change from default faults.
Source: Based on EPA Drinking Water Standard or Maximum Contaminant Level
-------
20
-REVIEW OF ADULT LEAD MODELS
BKSF in the ALM was based on an analysis of the data from Pocock et al. (1983), as supported by soil
lead ingestion studies in swine and humans (Casteel et al., 1996; Maddaloni et al., 1998).
The intake SF for dermal exposure to soil lead was estimated as the product of the SF for the soil
ingestion pathway and the relative absorption ratio for dermal/oral. This approach assumes that the
BKSFs for lead that is absorbed through the skin and gastrointestinal tract are similar, which may not be
accurate. The approach also requires an estimate of the relative absorption ratio for dermal/oral AF, for
which empirical support is lacking.
The AF and BKSF are represented as a single variable, the intake SF. This approach reflects the
available data on relationships between PbB concentrations and exposure concentrations or lead intakes
and the relative lack of data on the BKSF. Nevertheless, the inclusion of separate variables for the AF
and BKSF, as in the ALM, allows the two variables to be independently evaluated and adjusted to reflect
new data on each variable.
Exposure
Unlike the ALM, the California model does not allow for the inclusion of information about population
baseline PbB concentrations; however, this can be simulated by adjusting variables in the non-soil
pathways. The separation of the intake parameters for each exposure pathway and media in the
California model has the advantage that it allows for the inclusion of site-specific data about other
sources of lead in the risk estimate. It also allows a quantitative analysis of the relative contributions of
the various pathways to risk. The dermal soil pathway in the California model is unique in that the model
was the only one evaluated that considered a dermal absorption pathway. The empirical support for any
given value for the AF from soil adhered to the skin is weak.
The default value for combined soil and dust IR in adults is 0.025 g/day, compared to 0.05 g/day in the
ALM.
The California model includes several unique output features that would be useful in site risk assessment:
(1) the percent contribution of each exposure pathway to the predicted quasi-steady state PbB
concentration; (2) the estimated percentiles for PbB concentration; and (3) the estimated percentiles for
the risk-based soil lead concentration (e.g., PRO). The model lacks a graphics output; however, graphics
can be easily added to the existing spreadsheet.
Ease of Use/Flexibility
The model is easily implemented in a spreadsheet, and is easy to understand.
2.2.3 Summary
The exposure and variability features in the California model are similar to the IEUBK model. Although
it lacks the graphics display capability of the IEUBK model, an entire profile, including exposure and
intake parameters, PbB output, and estimated PRGs, is displayed in a single page of the spreadsheet. It
calculates the PbB concentration for both children and adults for residential scenarios and for adults in an
industrial scenario. Although this model does not simulate the distribution of lead between specific
tissue compartments, it represents a reasonably simple screening tool for evaluating the contribution of
soil lead exposure to PbB concentrations and for estimating PRGs.
-------
21
-REVIEW OF ADULT LEAD MODELS-
2.3 BERT
Reference: Environmental Research 48: 117-127 (1989)
2.3.1 Introduction
Description of Biokinetics
The Bert et al. (1989) model calculates the lead body burden associated with intakes of lead to the
gastrointestinal and respiratory tracts for a typical adult male (Figure 2.3.1). The central compartment
represents whole blood and other spaces that rapidly equilibrate with lead in whole blood. The apparent
volume of the central compartment is assumed to be approximately 1.5 times the blood volume; this
value is attributed to Rabinowitz et al. (1976), although the Rabinowitz model assumes a volume of
distribution for the central compartment of 1.7 times the blood volume. Lead in the central compartment
exchanges directly with cortical bone, trabecular bone, and other tissues.
FIGURE 2.3.1. LEAD BODY BURDEN ASSOCIATED WITH INTAKES OF LEAD TO THE
GASTROINTESTINAL AND RESPIRATORY TRACTS FOR A TYPICAL ADULT MALE
Bert et at. (1989) Biokinetic Model for Lead
Bile, Saliva,
Gastric Secretions
Be.
The cortical and trabecular bone compartments are distinguished by a more rapid exchange between
blood and trabecular bone compared to blood and cortical bone. External inputs to the central
compartment include the gastrointestinal tract, lung, and extraneous sources. Excretion in urine is
assumed to occur from the central compartment, while other excretion pathways (e.g., bile, saliva, gastric
-------
22
-REVIEW OF ADULT LEAD MODELS
secretions, and other pathways) are assumed to emanate from the other tissue compartment. Lead inputs
to the central compartment include digestive and respiratory tracts. Exchanges between tissue
compartments and transfers to excreta are represented as first order rate constants and were estimated for
the typical adult male based on average values estimated for four individuals from the Rabinowitz et al.
(1976) study (see Section 2.1). Because transfer coefficients are assumed to be constant, results in
calculated rates of lead transfer between compartments and lead excretion vary linearly with lead intake.
Description of the Exposure Component
The model does not have a complete multi-pathway exposure model. Exposure inputs to the model
include dietary lead intakes (|j,g/day) and air lead concentration (|j,g/m3). Intake of air-borne lead is
represented as the product of the air lead concentration and an inhalation day-volume; 15 mVday was
assumed to be typical for a sedentary lifestyle.
Three sources of lead uptake into the central compartment are represented in the model: gastrointestinal
tract, respiratory tract (referred to as lung in Beriet al., 1989), and extraneous sources. Lead uptake from
the respiratory tract is calculated as the product of a constant AF and inhalation intake; a value of
0.35 for the AF is attributed to Batschelet et al. (1979). Lead uptake from the gastrointestinal tract is
calculated as the product of a constant AF and the dietary intake; the value of 0.08 for the AF is assumed,
based on Marcus (1985), Batschelet et al. (1979), and Bernard (1977). Uptakes from other exposure
pathways are represented as extraneous uptake (|j,g/day).
Calibration and Evaluation
Model predictions of PbB concentrations compared well in magnitude and trends with the experimental
measurements from Rabinowitz et al. (1976) on which the transfer coefficients used in the model are
based. However, the model predicted substantially lower PbB concentrations when compared to an
independent data set from Griffin et al. (1975). In the Griffin study, subjects were exposed to high levels
of airborne lead and to dietary lead; blood and urinary lead concentrations were then measured at various
times. The divergence of predicted and observed PbB concentrations could be rectified by adjusting the
extraneous lead uptake to achieve a quasi-steady state PbB concentration (and corresponding body
burden) equal to the concentration at the start of exposure for each subject. This is, in effect, a
calibration of the model to each subject. The initial lead burden in the cortical bone was based on data
from Barry (1975). After calibration, agreement between observations and model predictions was greatly
improved; however, the model predicted PbB concentrations greater than those observed: a steeper
increase in PbB concentration and a higher maximum PbB concentration at the end of the high exposure
period.
The performance of the Bert model was also compared with the Bernard model (1977) for simulating
relatively long exposure periods (up to approximately 14 years) as reported in Ashford et al. (1977). The
Bert model predicts a more rapid increase in PbB concentrations in the early phase of increased lead
intakes; however, the two models predict similar quasi-steady state PbB concentrations.
Simulation
The model is represented mathematically as a series of coupled first order differential equations (Bert et
al., 1989). These equations were implemented in an Excel spreadsheet using a forward Euler function
for numerical integration with a 1-day time step.
-------
23
-REVIEW OF ADULT LEAD MODELS-
The model yields estimates of quasi-steady state maternal PbB concentrations that are very close to those
predicted by the ALM, if similar values for the AFS and IRS variables are used to estimate lead uptakes
and if the same values for the PbB0 are assumed. Table 2.3.1 provides the inputs that were used in a
typical set of simulations in which the outputs of the ALM and Bert model were compared. The soil lead
exposure was assumed to be to 1,000 |ig/g, 5 days/week for 260 days/year. An IRS of 0.05 g/day and a
gastrointestinal lead AF of 0.12 were assumed. Note, the latter two parameters are not components of the
Bert model; however, they were used to calculate lead uptakes that would be equivalent to those
simulated in the ALM for the same soil lead exposure. Baseline PbB concentration was introduced into
the Bert model simulations as a constant daily uptake (4.6 |j,g/day) that yielded a quasi-steady state PbB
concentration of 2 |_ig/dL in the absence of exposure to the designated soil lead level. After the baseline
PbB concentration was achieved, the simulation was continued with a daily lead uptake equal to the sum
of the baseline uptake (4 [i g/day) and the soil lead uptake (6 |_i g/day) (i.e., the product of the soil lead
concentration, 1,000 |ig/g; soil ingestion rate, 0.05 mg/day; and absorption fraction, 0.12).
Figure 2.3.2 shows the adult PbB concentrations predicted by the ALM together with the PbB
concentrations predicted by the Bert model. The ALM predicted a quasi-steady state PbB concentration
of 3.7 |-ig/dL, whereas the Bert model predicted a slightly higher value of 4.0 |_ig/dL. If an AF of
0.08 was assumed in the Bert model, which is the model default, the model yields a quasi-steady state
PbB concentration of 3.4 |-ig/dL, which is slightly lower than the prediction from the ALM. The ALM
also calculates a 95th percentile fetal PbB concentration, which is not shown in the figure. However, if
the same fetal/maternal ratios and variability mo del (i.e., the GSD) were applied to the central tendency
estimate of PbB concentration, the models would yield similar values for the 95th percentile fetal PbB
concentration.
TABLE 2.3.1. INPUTS FOR COMPARISON SIMULATIONS OF THE BERT MODEL AND
ALM SHOWN IN FIGURE 2.3.2
Parameters
PbS
IRS
AF
PbB0
EF
ED
Output
ALM
l,OOOLlg/g
0.05 g/day
0.12
2 Llg/dL
5 days/week (260 days/year; model
default is 219 days/year)3
Not a parameter in the model;
duration sufficient to achieve
quasi-steady state is assumed
Maternal PbB concentration
Bert
1,000 |-lg/g (not a parameter in the model)
0.05 g/day (not a parameter in the model)
0.12 (default value is 0.08)
Model was iterated with a daily uptake that
yielded a quasi-steady state PbB concentration
of 2 |-lg/dL (4.6 fig/day), after which, the
model was iterated with a daily uptake equal
to the sum of 4.6 fig/day and the product
PbS*IRs*AFs
5 days/week (260 days/year)
1 year beginning on day 365 of simulation
Adult male PbB concentration
aThe default exposure frequency for the ALM is 219 days/year; however, the assumption of 260 days/year in the
simulations would not change the outcome of the model comparisons.
-------
24
-REVIEW OF ADULT LEAD MODELS-
FIGURE 2.3.2. COMPARISON OF BLOOD LEAD CONCENTRATION PREDICTED
BY THE BERT MODEL AND THE ALM
5 0
4 0 -
TJ
230
3 ^-u
CO
°- 20
3
T3
< 10 -
0 0
c
r
I
} 1C
-r-- •"
)0 2C
)0 3t
/
X
)0 4C
Da
/"
"^dult!
)0 5C
ys
Bert
j/^wmHW*
EPA
>oil Exp<
)0 6C
n^Bwntfc
»sure B
)0 7C
jgins
)0 8C
)0
Note: See text and Table 2.1.1 for details on the inputs used in the model simulations.
2.3.2 Evaluation Criteria
Biokinetics
The Bert model can be viewed as an expansion of the Rabinowitz model, in which the bone compartment
is subdivided into separate compartments representing cortical bone and trabecular bone. The model is
based on a mass balance of lead in the major compartments where lead is distributed. The constant
transfer coefficients used in the model are average values estimated for four individuals from the
Rabinowitz et al. (1976) study. Although all tissue compartments are not individually represented, the
model fit experimental data reasonably well when it was calibrated to each subject by including an
extraneous lead uptake to match the initial experimental PbB concentrations. The extraneous lead values
ranged from 22 to 38 percent of the highest lead uptake. The volume of the central compartment
(denoted blood in the model) is assumed to be 1.5 times the volume of whole blood, with the whole blood
volume varying in direct proportion with body weight; a default value of whole blood volume of 52 dL
was assumed.
-------
25
-REVIEW OF ADULT LEAD MODELS
A useful feature of this model is the calculation of the initial average lead value in cortical bone for
different average ages of a group of individuals studied by Barry (1975). The cortical bone, which
contains the largest mass of lead in the body, is a "sink" for lead accumulation in this model. Thus, lead
in the cortical bone compartment never truly comes to equilibrium with the central compartment. The
assumption of equilibrium between compartments at the start of simulations actually introduces a
significant error in model simulations. The model achieves a quasi-steady state in blood after 3 months
of continuous exposure, consistent with experimental data. The model predicts urinary excretion of lead;
however, it was found to predict higher rates of lead excretion than were observed in experimental
studies.
The Bert model assumes a gastrointestinal AF of 0.08 compared to 0.12 used in ALM; however, when
the same values for the AF are assumed, the two models yield nearly identical estimates of quasi-steady
state PbB concentrations. The Bert model uses route-specific intake exposures instead of media-specific
concentration inputs.
Lead transfer coefficients are constants; the model was calibrated to represent biokinetics of adult males
and may not adequately represent the biokinetics of adolescents or females (e.g., different body masses),
or changes in biokinetics associated with physiological status (e.g., adolescent growth, pregnancy).
Exposure
The Bert model does not include a complete lead exposure model. Route-specific intakes (e.g., ingestion
of dietary lead and ambient air inhalation) are inputs to the model rather than media-specific
concentrations (e.g., soil, dust, drinking water). An exposure module could be linked to the Bert
biokinetic module.
The model is designed to predict lead in major body compartments. It predicts the mass of lead in blood,
cortical bone, trabecular bone, and other tissues combined (referred to in the model as the "tissue"
compartment). The model also predicts the amount of lead excreted through the urine. The model
assumes a blood distribution volume to report PbB concentration in |j,g/dL.
Ease of use/flexibility
The model is easy to use and can be readily implemented in a spreadsheet or ACSL. Currently,
environmental media concentrations are not inputs to the model; however, an exposure module could be
linked to the biokinetic module. The model can be used for exposures greater than one year.
2.3.3 Summary
In general, the model was found to fit experimental data reasonably well and to yield estimates of the
quasi-steady state PbB concentrations that are similar to those estimated by the ALM. The model could
be enhanced by adding an exposure module. The Bert model may not accurately simulate the kinetics of
lead in adolescence or during pregnancy.
-------
26
-REVIEW OF ADULT LEAD MODELS
2.4 STERN (1994) AND STERN (1996)
References: Risk Anal 14: 1049-156 (1994)
Risk Anal 16: 201-210 (1996)
U.S. EPA 1996
2.4.1 Introduction
Stern has developed two models to assess risks from exposures to lead in soil using different scenarios,
receptors, and toxicological endpoints. The Stern (1994) model was developed for residential land use
scenarios; the exposed population of concern is children ages 1-7 years; and the measured endpoint is the
incremental increase in PbB concentration, which correlates with adverse effects on the developing
central nervous system in the young child. The Stern (1996) model was developed for non-residential
land use scenarios; the exposed population of concern is adult males; and the selective critical effect is
elevated blood pressure. Both models represent lead absorption and biokinetics as an intake-PbB
concentration SF, which relates the incremental change in soil lead intake to an incremental change in the
quasi-steady state PbB concentration. The intake SF is distinguished in this report from the term BKSF,
which applies to the lead uptake (absorption)-PbB concentration relationship, such as that used in the
ALM.
The Stern (1994) model evaluates the residential child exposure as follows:
Csoil = APbB Equation (2.4.1)
1,-R- [(A, • T, • (I,/!)) + (S-A2-F-T2- (I3/I))]
where:
Csoil = concentration of lead in outdoor soil (|ig/g)
PbB = de minimis increase PbB concentration; total change in PbB
concentration resulting from ingestion of soil and soil-derived dust
(Hg/dL)
It = overall rate of daily soil ingestion integrating the contribution from
outdoor soil and indoor soil-derived dust (g/day)
R = slope of the relationship between PbB concentration and lead intake
from diet (|j,g/dL per |ig/day)
A, 2 = ratio of lead absorption from (1) soil and (2) soil-derived dust to lead
absorption from food (unitless)
Tj 2 = fraction of waking day spent (1) outdoors and (2) indoors (unitless)
I, = rate of outdoor soil ingestion (g/day)
I3 = rate of total indoor dust ingestion (g/day)
S = enrichment factor; ratio of lead concentration in soil-derived dust to lead
concentration in soil (unitless)
F = fraction of indoor dust that is soil-derived (unitless)
For adult nonresidential exposure to lead, the Stern (1996) model describes the relationship between soil
lead and change in blood pressure with the following equation:
P2 = [ln((I 'CS'R + PbBJ/PbBJ • S] + P, Equation (2.4.2)
-------
27
-REVIEW OF ADULT LEAD MODELS
where:
P2 = the resultant systolic blood pressure due to the increase in PbB
concentration (mm Hg)
/ = soil ingestion rate (mg/day)
Cs = concentration of lead in soil (|ig/g)
R = slope of the relationship between intake from soil lead ingestion and
PbB concentration (|j,g/dL per |ig/day)
PbB, = baseline PbB concentration for the adult male population (|j,g/dL)
S = slope of the relationship between PbB concentration and systolic blood
pressure (mm Hg/|j,g/dL)
P, = baseline systolic blood pressure in the adult male population (mm Hg)
Application of the Stern models utilizes the concept of a lexicologically de minimis PbB concentration,
defined by the Stern (1994) model as an incremental increase in PbB concentration (from a single
medium) that represents a suitably small increase relative to the generally accepted PbB concentration
(e.g., 10 ng/dL).
Variables in Equations 2.4.1 and 2.4.2 are represented as probability density functions (PDFs), and
output is a distribution of Csoil or P2. PDFs are defined for selected input variables to reflect variability
and/or uncertainty. Monte Carlo simulations are used to randomly sample from the set of input PDFs in
order to estimate a PDF for soil lead concentrations that corresponds to a de minimis concentration of
2.0 |-ig/dL (see Equation 2.4.1). The Stern (1994) model provides both default point estimates and
default parameter values for PDFs for selected input variables; the Stern (1996) model does not include
point estimates with default PDFs.
Description of Biokinetics
Lead absorption and biokinetics are combined in an intake SF, which relates the incremental change in
soil lead intake to an incremental change in the quasi-steady state PbB concentration.
Description of the Exposure Component
Environmental lead levels are not explicitly represented as input variables in the residential child model
Stern (1994). The output for this model is soil lead concentration. In both Stern (1994, 1996) models,
exposure pathways are limited to soil ingestion (both outdoors and as a component of indoor dust).
Alternative source contributions to indoor dust (e.g., paint) are not explicitly represented. Chronic lead
intake is estimated as a single long-term average (ages 1-7 years for Stern, 1994; ages 18-65 years for
Stern, 1996), and intermittent changes in dose levels cannot be simulated. Exposure to other sources of
lead cannot be assessed using these models.
Soil and Dust Ingestion Rate. For the residential model Stern (1994), daily soil and dust IRS are
calculated separately (Equation 2.4.3). This approach is similar to the "alternative method" described by
EPA (1996; pp. A19-A22).
7, = (r, • /,) + (r2- 72) = (r, • /,) + r2- (/3 • F) Equation (2.4.3)
-------
28
-REVIEW OF ADU LT LEAD MODELS-
where:
F
Description of Uptake
overall rate of daily soil ingestion integrating the contribution from
outdoor soil and indoor soil-derived dust; parameter estimated in fecal
tracer studies (g/day)
rate of ingestion of outdoor soil (g/day)
fraction of the waking day spent outdoors (unitless, point estimate=0.1)
rate of ingestion of soil-derived dust (g/day)
rate of ingestion of total indoor dust (g/day)
fraction of indoor dust that is soil-derived (unitless, point estimate=0.3)
fraction of the waking day spent indoors (unitless, point
estimate=l-0.1=0.9)
As previously noted, the Stern (1994) model uses an intake SF that relates an incremental increase in
quasi-steady state PbB concentration to dietary lead intake; thus, an AF for dietary lead is not a separate
variable, to estimate lead uptake. However, since the intake SF is based on experimental data with lead
in drinking water and milk, intakes are adjusted with a linear multiplier, which represents the relative
bioavailability of lead in soil (Al) and in dust (A2) compared to soluble lead. The latter model does not
require an additional factor since the model calculates PbB from soil lead, not from dietary lead.
The Stern (1996) model uses an intake SF that also relates an incremental increase in quasi-steady state
PbB concentration to soil lead intake, based on empirical data relating lead in soil to PbB concentration.
Calibration and Evaluation
Calibration and evaluation of the Stern models were not described in the documentation.
Simulation
In order to directly compare the Stern model with the ALM, ideally risks and PRGs would be estimated
using the same exposure scenario (non-residential), receptors (adults), and endpoints of concern (PbB).
Neither of Stern's two models (Stern, 1994, 1996) provide this combination of approaches. Therefore, a
hybrid of the two models was developed by combining the input assumptions for the non-residential
exposure scenarios in which adults are the receptor of concern (Stern, 1996) with a toxicological
endpoint defined by a de minimis change in PbB concentration assuming a baseline PbB of 2 |-ig/dL
(Stern, 1994). The hybrid model was executed with both point estimate and probabilistic inputs. Monte
Carlo simulations were implemented using @Risk as an add-in software to Microsoft Excel. The four
simulations using the hybrid model are summarized in Table 2.4.1.
TABLE 2.4.1. SIMULATIONS EXECUTED TO EVALUATE THE HYBRID STERN MODEL
Simulation
l
2
3
4
Reference for
Inputs
Stern (1994)
EPA (1996)
Stern (1994)
Stern (1996)
Point
Estimates
X
X
PDFs
X
X
Endpoint
Children
Adults
Children
Adults
Results
Reported
Table 2. 4.2
Table 2. 4.2
Table 2. 4. 3
Table 2. 4. 3
-------
29
-REVIEW OF ADU LT LEAD MODELS-
The population of concern for the Stern (1994) model is children ages 1-7 years. Soil lead
concentrations are calculated under two different scenarios: soil ingestion occurs both outdoors and
indoors (as soil-derived dust); and soil ingestion occurs outdoors only. Table 2.4.2 summarizes the point
estimate input assumptions and corresponding PRGs for these two scenarios. For purposes of
comparison, the PRGs associated with the Stern model are compared with the PRGs calculated using the
default inputs recommended for the ALM (EPA, 1996). Table 2.4.3 summarizes the probabilistic input
assumptions recommended for use in residential exposure scenarios (i.e., children ages 1-7 years) and
non-residential exposure scenarios (i.e., adults). Results from these simulations yield probability
distributions for PRGs; Table 2.4.3 gives the geometric mean and arithmetic mean. Each of the
simulations is described in further detail below.
TABLE 2.4.2. PRGs ASSOCIATED WITH DEFAULT POINT ESTIMATE EXPOSURE
ASSUMPTIONS APPLIED TO THE STERN (1994) DEMINIMIS PbB METHODOLOGY
Symbol
APbB
PbBfeta/PbBmat
I,
R
BKSF
At
A2
A3
T,
T2
IA
Vi,
LA
s
F
Csoll (PRO)
csoll,mtdoors
Description
Change in blood lead concentration
Ratio of fetal PbB to maternal PbB
Ingestion rate of soil and soil-derived dust
Slope factor relating blood lead to lead
INTAKE
Slope factor relating blood lead to lead
UPTAKE
Relative absorption factor (AF1011/AFsohlble)
Relative absorption factor (AFdus/AFloluble)
Absolute absorption factor (AFlohlble x
AFsoll/AFsohlble
Fraction of day outdoors
Fraction of day indoors
Ratio of IRsoll to IR all dust
Ratio of IRsoll to IR^,, + dust
Ratio of IRsoll
Ratio of [Pb](DUST)/[Pb](SOIL
Fraction of indoor dust that is soil
Cone, of Pb in soil
Cone, of Pb in soil, outdoor exposure only
Units
Hg/dL
unities s
g/day
Hg/dL per
Hg/day
Hg/dL per
Hg/day
unities s
unities s
unities s
unities s
unities s
unities s
unities s
unities s
unities s
unities s
^g/g
^g/g
Point Estimate
Stern"
2.00
NA
0.10
0.16
NA
0.63
0.76
NA
0.10
0.90
1.00
2.70
2.70
1.20
0.30
150
198
EPA"
2.00
0.90
0.10
NA
0.40
NA
NA
0.12
NA
NA
NA
NA
NA
NA
NA
375
NA
"Stern, 1994.
"Revision of Stern's default inputs based on EPA, 1996
NA = variable not available in the default model; PRG = preliminary remediation goal
-------
-REVIEW OF ADU LT LEAD MODELS-
TABLE 2.4.3. PRGs ASSOCIATED WITH DEFAULT PROBABILITY DENSITY FUNCTIONS FOR
EXPOSURE ASSUMPTIONS APPLIED TO STERN (1994) DEMINIMIS PbB METHODOLOGY
Symbol
APbB
I,
R
A:
A2
T,
T2
IA
VI,
yi,
s
F
Csoll (PRO)
c
^soil, outdoors
Description
Change in blood lead concentration
Ingestion rate of soil and soil-derived dust
Slope factor relating blood lead to lead
INTAKE
Relative absorption factor (AFdlls/AFsolllble)
Relative absorption factor (AFdlls/AFsohlble)
Fraction of day outdoors
Fraction of day indoors
RatioofIRsolltoIRalldust
Ratio of IRsoll to IRSOiL + dUSt
Ratio of IRall iast to IRsoll+ dus,
Ratio of [Pb](DUST)/[Pb](SOIL)
Fraction of indoor dust that is soil
Cone, of Pb in soil
Cone, of Pb in soil, outdoor exposure only
Units
Hg/dL
g/day
Hg/dL per
Hg/day
unitless
unitless
unitless
unitless
unitless
unitless
unitless
unitless
unitless
^g/g
^g/g
PDF
Residential"
2.00
triang (0.05,
0.10,0.20)
triang (0.056,
0.16,0.18)
triang (0.15,
0.63,0.71)
triang (1.0, 1.2,
1,5) xAl
triang (0.05,
0.10,0.20)
1.00-T1
uniform (0.50,
0.20)
l/[Tl+((FxT2)/
(11/13))]
(Il/(Il/I3))/Ii
triang (1.0, 1.2,
3.0)
triang (0.20,
0.30,0.50)
GM=160
AM=190
GM=280
AM=320
Non-Residential1"
2.00
lognormal (0.0547,
0.0328)
uniform (0.014,
0.034)
NA
NA
triang (0.05, 0.10,
0.20)
1.00-T1
uniform (0.50, 0.20)
l/[Tl+((FxT2)/
(11/13))]
(Il/(Il/I3))Ii
triang (1.0, 1.2, 3.0)
triang (0.20, 0.30,
0.50)
GM=1,270
AM=2,230
GM=1,850
AM=2,230
'Stern, 1994.
"Stern, 1996. PDFs for li and R are for adult populations. No absorption fraction is assumed by Stern. Values for Tl, T2, II, 13, S, and F are from
the residential model.
NA = variable not available in the default model; PRG = preliminary remediation goal
Simulation 1. This approach calculates soil PRGs using point estimates for exposure assumptions given
by Stern (1994). If ingestion is assumed to occur both outdoors and indoors, the resulting PRG is
150 |-ig/g. If ingestion is assumed to occur outdoors only, the resulting PRG is 198 |-ig/g. PRG is lower
when dust ingestion is excluded because of the enrichment factor for soil in dust (S), and the higher
indoor dust ingestion is included because of the enrichment factor (AFdust/AFsoluble) compared with soil
(AFsoil/AFsolllble).
-------
31
-REVIEW OF ADU LT LEAD MODELS
Simulation 2. The soil PRO using the ALM defaults applied to Stern's de minimis PbB concentration
methodology (see Equation 2.4.4) is 375 |ig/g.
Csoil = &PbB • PbBfetal Equation (2.4.4)
I, • BKSF • AFsd
where:
Csoil = concentration of lead in outdoor soil (|ig/g)
PbB = de minimis increase in PbB concentration (|j,g/dL)
PbBfetal/PbBmatemal = ratio of fetal to maternal PbB concentration (unitless)
/, = overall rate of daily soil ingestion integrating the contribution
from outdoor soil and indoor soil-derived dust (g/day)
BKSF = biokinetic slope factor relating PbB concentration and lead
uptake from soil ingestion (jig /dL per |ig/day)
AFsd = absolute absorption fraction of lead from soil and dust (unitless)
Simulation 3. This approach estimates the distribution of soil PRGs that corresponds to a de minimis
PbB of 2 |-ig/dL combined with probability distribution functions (PDFs) for selected exposure variables.
For these scenarios, the PDFs are intended to characterize interindividual variability. Using the default
PDFs recommended for residential exposure scenarios (Stern, 1994) yields a geometric mean PRO of
160 |ig/g and arithmetic mean of 190 |_ig/g when outdoor and indoor exposures are combined, compared
with a geometric mean of 280 |_ig/g and arithmetic mean of 320 |_ig/g for outdoor exposures alone.
Simulation 4. The hybrid Stern model uses adult-specific PDFs for the following variables: soil and dust
ingestion rate (I;) and slope factor (R). Note that the slope factor is based on empirical data on lead in
soil; therefore, a relative absorption factor is not used in the adult male model (Stern, 1996).
Assumptions in the child model (Stern, 1994) used to differentiate between indoor and outdoor exposures
were applied to the hybrid model. There is uncertainty in this approach given that the relationship
between indoor and outdoor exposures were presumably developed to simulate activity patterns of young
children, rather than that of adult workers. The cumulative PDF of the model output is shown in
Figure 2.4.1. The PRGs for Simulation 4 are approximately an order of magnitude greater than
Simulation 3; this difference is due mainly to the lower slope factor relating blood lead concentration to
lead intake for adults compared with children.
As shown in Figure 2.4.2, the output from the hybrid Stern model (GM=1,270 |ig/g) and the EPA (1996)
ALM (1,080 |ig/g) yield approximately the same soil PRG when the baseline PbB concentration is
assumed to be 2.0 |_ig/dL. In contrast to the ALM, Stern's model is insensitive to the absolute magnitude
of PbB concentration (influenced by the PbB0).
2.4.2 Evaluation Criteria
Biokinetics
Lead absorption and biokinetics are combined in an intake SF. In the residential child model, the intake
SF for dietary lead is represented as a triangular distribution having a mode of 0.16 and mean of 0.13 (see
Table 2.4.3). This value is lower than the intake SF of 0.17 predicted from the IEUBK model; the latter
can be estimated as the product of the BKSF predicted from the IEUBK model (0.34) and from the
default lead AF for dietary lead at low lead intakes (0.5; i.e., 0.34x0.5=0.17). The product of the
assumed BKSF ([0.4] and the AF for soil lead 90.12) results in an intake SF of 0.048. Thus, relative to
-------
-REVIEW OF ADULT LEAD MODELS-
the IEUBK model, the Stern (1994) residential child model assumes a higher value for the soil lead
BKSF or soil lead AF, or higher values for both variables.
A similar comparison can be made between the intake SF in the Stern (1996) non-residential adult model
and the SF in the ALM. The Stern (1996) model represents the intake SF for soil lead as a uniform
distribution of 0.014-0.034 and mean of 0.024 (see Table 2.4.2). The product of the ALM values for the
BKSF (0.4) and the AF for soil lead (0.12) is 0.048. Thus, relative to the ALM, the Stern (1996) model
assumes either a higher value for the soil lead BKSF or soil lead AF, or higher values for both variables.
FIGURE 2.4.1. CUMULATIVE PROBABILITY DISTRIBUTION (CDF) OF OUTPUT FROM HYBRID
STERN MODEL, USING STERN (1994) DEMINIMIS BLOOD LEAD CONCENTRATION MODEL
WITH STERN (1996) ADULT INPUTS FOR PDF
Inn
.uu
£ 07^-
ig u. o
1
a
I-H
PH o eft
a, U.3U
1 :
i— 1
3
£ n i^
^ U.io
Onn
^
_y
/
Hj^brid Stern Model :
y
j_
i
i
.uu
0 500 1000
-"
X
EPA, 1996
1,080 jig/g
^•^
_-o^ "
mean —
5th %ile =
10th %ile -
50th %ile =
75th %ile -
90th %ile -
.^-0
1,590
1,113
453
567
1,270
1,922
2,933
1500 2000 2500 3000 3500 4000
Soil Pb Concentration (ilg/g)
Note: EPA (1996) soil PRO is based on default inputs.
GSD = 2.0; PbB0 = 2.0
-------
Figure 2.4.2. Comparison of Soil Lead Concentrations (PRGs) Using EPA (1996) ALM and
Probabilistic Stern Model (See Table 2.4.3) as a Function of
Baseline Blood Lead Concentrations (PbB0)
1 nnn
4,UUU
T •rnn -
j,jiJU
3000 -
a
to
^t "i ^nn
£ J,JUU •
5
s
h i nnn -
+; J,UUU
8
u
i
J~ 1 •rnn
JJ l,jUU -
£
•3 1 nnn
5 l.UUU
«!
^nn
jUU
X
EPA(1W6)
N
OJO
X
i X
N
^
\
X
Sfcrn, 9 0th 'i .life
= 2,930 ng/g
Stem, 50th % die
=1,270 Hg/g
X
1JO 2.0 3JO
Baseline PbB (|i^dL)
4.0
5.0
The intake SF used in the Stern (1994) child model is based on experimental data with lead in drinking
water and lead in milk; therefore, soil and dust lead intakes are adjusted with a relative absorption factors
(RAF) of lead in soil (AFsoil/AFdiet, Al) and lead in dust (AFdust/AFdiet, A2) as compared to soluble lead
(Equation 2.4.1). The RAF for soil lead is represented in the Stern (1994) model as a triangular
distribution having a mode of 0.63 and mean of 0.50 (see Table 2.4.2). These values are similar to the
IEUBK model default value of RAF for soil lead of 0.6, as reflected in the ratio of the AFs for soil lead
(0.3) and dietary lead (0.5) at low lead intakes (0.3/0.5=0.6). The Stern (1994) model assumes that the
RAF for dust varies dependency on the RAF for soil lead. This dependence is represented as the product
of the distribution for RAFsoil and a triangular distribution having a mode of 1.2 and a mean of 1.23 (see
Table 2.4.3); the resulting combined distribution, reflecting RAFdust, has a mean of 0.61, slightly higher
than RAFsoil. The IEUBK model default values for the dust lead and soil lead AFs and corresponding
RAFs are identical.
Exposure
The exposure algorithms in the Stern (1994, 1996) models and ALM are very similar. The treatment of
soil and dust as separate pathways is described in EPA (1996) as an alternative method for calculating
soil and dust ingestion as separate pathways. One difference between the two exposure models is the
concept in the Stern (1994) child model of an enrichment factor for the concentration of lead in dust as
-------
34
-REVIEW OF ADULT LEAD MODELS
compared to lead in soil. The source of the lead is important in determining the relative degree of
enrichment; lead from mining sites would tend to be less enriched in the dust than the lead from smelting
sites due to differences in size distribution of lead-associated particles. As Stern (1994) points out, there
is considerable uncertainty about what value to use as a default for this enrichment factor.
Computations for additional exposure pathways could be readily added to the Stern (1994) model.
However, the model is limited to assessing constant exposures of sufficient duration to achieve a quasi-
steady state PbB concentration. Intermittent changes in exposure cannot be simulated with this model.
The output is a probability distribution reflecting combined variability and uncertainty associated with
input variables. This is calculated using a one-dimensional Monte Carlo approach in which the
combined variability and uncertainty associated with selected variables are represented with PDF. An
alternative approach would be to segregate the representations of variability and uncertainty as separate
dimensions in a two-dimensional Monte Carlo approach.
Ease of use/flexibility
The de minimi approach to lead risk assessment is currently not consistent with EPA policy; thus, the
model as currently configured might not be applicable to Superfund assessments.
The model can be implemented with point estimates for input variables in a relatively simple
spreadsheet; however, additional software is required to implement the model with PDF for input
variables (e.g., @Risk, Crystal Ball).
2.4.3 Summary
The Stern (1996) adult exposure model uses change in blood pressure in the adult male (and resultant
increase in incidence of hypertension in the exposed population) as the sensitive endpoint. This outcome
differs substantively from the ALM, which uses fetal effects as the sensitive endpoint. The bottom line
of the Stern (1996) model is that exposure by adult males to 1,000 |ig/g soil lead concentration results in
an increase of approximately 1 mm Hg in systolic blood pressure, an increase of approximately 1 percent
in the incidence of systolic hypertension (pressure >140 mm Hg), and an increase in PbB concentration
of 1-3 |-ig/dL (these effects are defined by Stern as de minimi). Since the simulation runs derived
relatively similar soil lead PRGs for the hybrid Stern model and ALM, this would indicate that the EPA
approach is protective for the male hypertension endpoint (assuming that Stern's quantitative cause and
effect relationship between soil lead and blood pressure is correct).
Stern's (1994) child model focuses on the de minimi increase in PbB concentration; therefore, the model
does not account for a baseline PbB concentration. Use of this approach could result in cleanups that are
over- or under-protective, depending on the specific baseline lead exposure of the receptor. Stern's
(1996) adult model does have the baseline PbB concentration as an input parameter. The conclusions,
however, focus on the ability to correlate blood pressure to a calculated increase in PbB concentration,
downplaying the importance of the baseline PbB concentration value.
-------
35
-REVIEW OF ADULT LEAD MODELS
2.5 LEGGETT
References: Environmental Health Perspectives 101: 598-616 (1993a)
Health Physics 64: 260-71 (1993b)
Environmental Health Perspectives 106:1505-1511 (1998)
2.5.1 Introduction
The Leggett (1993) model was developed in the last 10 years by Oak Ridge National Laboratories. The
model is currently used by the International Commission on Radiological Protection to predict internal
radiation doses of a variety of radionuclides that have biokmetics similar to those of calcium. The
descriptions and discussions that follow are based on the FORTRAN version (5/15/97), provided by Joel
Pounds (Wayne State University) and described in Pounds and Leggett (1998). The FORTRAN code
was modified as needed to develop simulations and capture model output for analysis and display.
Description of Biokinetics
The Leggett model simulates the movement and deposition of lead in the body as exchanges between
various tissue compartments (and sub-compartments) and a central diffusible plasma compartment
(Figure 2.5.1). Tissue compartments represented in the model include bound plasma (e.g., lead bound to
plasma proteins), brain, extravascular fluid, gastrointestinal tract, kidney, liver, lung, other soft tissues,
red blood cells, and skeletal tissues (cortical and trabecular bone). Excretory routes represented in the
model include feces, sweat, urine, and other routes (e.g., hair, nails, skin). Exchanges are described by
age-specific transfer coefficients (analogous to first-order rate constants). In the default model, transfer
coefficients are specified for ten ages: 0, 0.274, 1,5, 10, 15, 25, 30, 40, and 60 years of age (although the
ages can be changed by the user); coefficients for intermediate ages are computed by linear interpolation.
The Leggett model can represent the saturation of lead uptake by red blood cells as a non-linear process
(user option).
The bone compartment in the Leggett model includes the following features:
• The model simulates slow, intermediate, and rapid kinetics of bone lead as distinct lead
compartments and pools.
• Bone is divided into several compartments: cortical volume, trabecular volume, cortical
surface, and trabecular surface. The two bone volume compartments are divided into
exchangeable and non-exchangeable lead pools.
• The slow kinetic component is attributed to the non-exchangeable pools of the bone
volume compartments. Lead enters the non-exchangeable pools from the exchangeable
pools and exits only when bone is resorbed.
• The intermediate kinetic component is attributed to the exchangeable pools of the bone
volume compartments, from which lead exchanges with the bone surface compartments.
• The fast kinetic component is attributed to the surface bone compartments from which
lead exchanges with plasma.
Description of the Exposure Component
Environmental lead levels are not explicitly represented. The model includes three routes of lead intake
(|-ig/day): ingestion, inhalation, and direct input to blood (injection). Exposure may occur at any age, and
-------
36
-REVIEW OF ADULT LEAD MODELS-
there is an option to include exposure prior to birth (fetal exposure). The latter calculates a PbB
concentration at birth as a user-defined ratio of the maternal PbB concentration and distributes lead to
FIGURE 2.5.1. SCHEMATIC OF THE BIOKINETIC MODEL OF LEAD METABOLISM
PROPOSED BY LEGGETT (1993)
Compartments and Pathways of Lead Exchange in the Leggett Model
Losses in
H=iir N=iilc 4
Skin
Other SoftTissues
Intermediate
Turnover
Rapid Tens
Turnover Turr
Skeleton
rinrtjrqj \/ri|um°
Nnn- i « * ^
U U
1 fc
ical
j t.
F^h.nnR! t!.
r*-
Kidneys
Other
Kidney
Tissue
Urinary
Path
^
,4
^
™-t-
Bladder
Contents
4
Urine
Uittu^ible -4
Plasma
! Extra- • "-» Br
•
; Vascular : .
! RBC ; «-»• Liv
1 Bound : ++ Llv
i Plasma :
^
,
Sv^eat
cious
lover
.
1
b LUn j] c
A
a In
^J— —
ver /*^ ~~~~-^.^
er 1 ( Lead Intake
V /
^^~T^~^"^
er 2
^
b, r-.i Ti-art
1
Feces
Source: Derived from Leggett 1993
Note: Transfer rates for exchanges between com pa rtrrie nfe are age-specific. See Leggett (1993) for the age-specific transfer rates
used in the model.
-------
37
-REVIEW OF ADULT LEAD MODELS
tissues at birth according to user-defined fractions of body burden. The model allows for evaluation of
acute or chronic exposure. Chronic exposure may occur at up to 50 different dose levels over the course
of a lifetime.
For exposures that begin at other times, lead masses in all compartments are null at the start of exposure,
although simple programming can be used to set other values, (i.e., to simulate a baseline condition).
Alternatively, baseline conditions can be simulated as a baseline lead intake scenario.
Description of Uptake
Uptake is represented in the Leggett (1993) model with age-specific AFs. The AF is a linear multiplier
of lead intake and represents the fraction of the lead mass delivered into the small intestine that enters the
plasma. Default values for gastrointestinal AFs are as follows: 0.45 at birth and 100 days of age; 0.3 at 1,
5, 10, and 15 years of age; and 0.15 at 25, 30, 40, and 60 years of age. Absorption fractions at
intermediate ages are calculated by linear interpolation, as shown in Figure 2.5.2.
Calibration and Evaluation
Leggett (1993) has compared model predictions to urinary or fecal excretion data obtained from ten
separate studies on human subjects who were exposed via injection, inhalation, or ingestion and found
that the predictions fit the observations reasonably well (Leggett, 1993b; Pounds and Leggett, 1998).
Model predictions provided good agreement with postmortem data on lead concentrations in bone, blood,
liver, kidneys, and brain in humans in a variety of age groups. The model predicted the acute changes in
PbB concentrations in individual children following chelation therapy reasonably well; for these
comparisons, pre-chelation exposure was simulated to achieve observed pre-chelation PbB
concentrations (Pounds and Leggett, 1998).
Simulation
The model is represented mathematically as a series of coupled first order differential equations (Leggett,
1993a). These equations were implemented in FORTRAN (5/15/97). The user can vary the time step
(variable integration) used in the integration, and the data logging rate (e.g., every 100 time steps). The
program reports the concentration in the compartments once every 100 time steps (the communication
interval) (if the time step is too large, the Leggett model, as with most other models, will not yield
accurate simulations).
The performance of the Leggett model was investigated by running a variety of simulations. These were
not exactly the same as those explored in the other models because the models differed so much in
structure that it was not always possible to set up the same simulations. Also, unique features of this
model were of interest and were further understood through different types of simulations. In this section
six main points are discussed: (1) how the simulations using the Leggett model compare to the adult lead
slope factor approach (ALM); (2) how the baseline PbB concentration in a population was simulated;
(3) how exposure to a given concentration of lead in soil was simulated; (4) how the nonlinear
component of the Leggett model works; (5) how the biokinetic slope factor used in the IEUBK compares
to a crude estimate of the slope factor from the Leggett model; and (6) the effects of brief changes in
childhood blood lead upon adult blood lead.
Comparisons of output from the Leggett model and ALM. When simulations use similar values for
soil intake (IRS) and absorption (AFS), and baseline blood lead (PbB0) without site exposures, the Leggett
model yields estimates of blood lead very similar to the EPA approach (ALM). Table 2.5.1 provides the
inputs used in a typical set of simulations that compare the Leggett model and the ALM. In order to
generate comparable simulations, some of the inputs (e.g, IRS) had to be approximated for the Leggett
model. When similar input parameters were used in both models and a constant exposure to lead
-------
-REVIEW OF ADU LT LEAD MODELS-
FIGURE 2.5.2. DEFAULT VALUES IN THE LEGGETT MODEL FOR LEAD ABSORPTION IN THE
GASTROINTESTINAL TRACT (PERCENT) AS A FUNCTION OF AGE
40 -•
i 304.--*
10 -•
60
120 180
Age (months)
240
300
TABLE 2.5.1. INPUTS FOR COMPARISON SIMULATIONS OF THE LEGGETT MODEL AND
ALM SHOWN IN FIGURE 2.5.3
Parameters
PbS
IRS
AF
PbB0
EF
ED
Output
ALM
l,000|lg/g
0.05 g/day
0.12
2 [ig/dL
5 days/week (260 days/year; model default is
219 days/year)3
Not a parameter in the model; duration sufficient
to achieve quasi-steady state is assumed
Adult PbB concentration
Leggett
1,000 |-lg/g (not a parameter in the model)
0.05 g/day (not a parameter in the model)
0.12 (default varies with age from 0.3 at
age 15 to 0.15 for ages 25-50 years)
Model was iterated with a daily intake that
yielded a quasi-steady state PbB
concentration of 2 flg/dL (see
Figure 2.5.4a.), after which, an additional
increment in lead intake was simulated that
was equal to the sum of the product of *
IRS
5 days/week (260 days/year)
Ages 17-45 years
Adult male PbB concentration
aThe default exposure frequency for the ALM is 219 days/year; however, the assumption of 260 days/year in the
simulations would not change the outcome of the model comparisons.
-------
-REVIEW OF ADULT LEAD MODELS-
in soil (i.e., 1,000 |ig/g) was used, the quasi-steady state PbB concentrations predicted by the two models
were similar: 4.1 |-ig/dL for Leggett and 3.7 |-ig/dL for the ALM (Figure 2.5.3).
Simulated of baseline blood lead. The ALM has a parameter for baseline blood lead in adults (i.e.,
PbB0) which has a default value of approximately 2 |j,g/dL. In order to reproduce this value and
associated tissue lead levels in the Leggett model, pre-adult lead intakes were simulated to approximate
the present day population exposures to lead that results in a typical population blood lead of 2 |j,g/dL.
The baseline simulation was constructed by altering the lead intake to produce blood leads at various
ages that were similar to those of U.S. females in Phase I of the National Health and Nutrition Survey III
study (NHANES III; Brody et al, 1994) (Figure 2.5.4a.). Shown in Figure 2.5Aa. are the minimum and
maximum values of geometric means for the race/ethnicity categories in NHANES HI (Table 2 of Brody
et al., 1994). Simulated lead concentrations in cortical and trabecular bone are also shown in
Figure 2.5 Aa. Figure 2.5 Ab. shows the lead exposure (jig/day) that produced the profile for the baseline
blood lead graphed in Figure 2.5 Aa. The intakes ranged from 10-16 |j,g/day in children and increased up
to 26 |j,g/day in adults.
Simulation of exposure to a given concentration of lead in soil. The Leggett model can simulate a
wide variety of lead exposure conditions, however, input to the model must be in units of lead intake
(|ig/day). Numerous simulations were run to compare the model with the ALM. A typical simulation
consisted of a baseline intake plus an additional increment in lead intake equivalent to a given exposure
to soil lead. For example, to simulate an adult exposure to 1,000 |_ig/g lead in soil, a baseline simulation
was run (dotted line in Figure 2.5.3), then an increment in lead intake representing the soil lead exposure
(1,000 |ig/g x 0.05 g soil/day ingestion rate) was introduced into the simulation at age 17 years (solid line
Figure 2.5.3; inputs in Table 2.5.1). For comparison purposes, the absorption fraction in the Leggett
model for ages 15 years and older was changed to the default value of 0.12 used in the ALM. The
simulated soil lead exposure (1,000 |_ig/g lead in soil) resulted in similar quasi-steady state blood leads
for both the Leggett (4.1 |-ig/dL) and the ALM (3.7 |_ig/dL). However, if the absorption fractions
(decreasing from 0.3 at age 15 to 0.15 at age 25 years) from the Leggett model were used instead, the
difference in between the Leggett model and ALM was more substantial; the Leggett predicted a higher
PbB concentration, 4.9 |_ig/dL. Figure (2.5.3) also displays estimates of the concentrations of lead in
cortical and trabecular bone from the Leggett model.
Nonlinear component of the Leggett model. The Leggett model gives the user the option of simulating
the lead uptake by red blood cells as a linear or nonlinear process. The nonlinear process could
correspond to the saturation of binding sites (lead substituting for calcium) on the membrane of the red
blood cells and the subsequent increase of lead concentration in the plasma and the urine. Since
available data suggests that the relationship between blood lead concentration and lead intake is non-
linear (Leggett, 1993), the non-linear option was selected for all simulations run in this evaluation. At
low lead concentrations, the kinetics are linear; nonlinear kinetics start when the lead concentration in
red blood cells reaches 60 |-ig/dL which corresponds to a blood lead of about 25 |_ig/dL. For simulations
of blood lead below 25 |-ig/dL, the linear and nonlinear options of the model produce similar results.
Over 25 |-ig/dL, the linear option predicts higher blood lead than the non-linear option and the disparity in
the predictions increases as blood lead increases (see the appendices). On the other hand, use of the
linear option had no effect on the simulated concentrations of lead in bone (see the appendices).
-------
40
-REVIEW OF ADULT LEAD MODELS-
FIGURE 2.5.3. COMPARISON OF BLOOD LEAD CONCENTRATIONS PREDICTED BY THE
LEGGETT MODEL AND ALM
T 4 -
—1 H
TJ
In
&
T3 o _
« 3
Of
_i
•o
o
2 -
CO
„„ Baseline Blood
(Leggett)
/
^— Baseline + Soil Blood
(Leggett)
'\ / _j
"•"•'! X"
: ^^^^^^^^m i
\
_.J
X ,f~~
Cortical Bone
_i— — - — -^
d^ ~... — — ~~
0 10
Begin Soil Exposure
i,
Baseline + Soil Blood \
(EPA ALM) ^--^
^— —
^-^-"__
^—
30 30
Age (years)
A
Trabecular Bone
40
^
9
r 6
s
4
m
o
3
to
3 |-
a>
&»
o.
2 CQ
&
1
0
0
Note: The Leggett model simulation includes a baseline exposure followed by additional exposure to 1,000 ppm soil lead, 5 days/
week, beginning at age 17 years (see Table 3.5.1 for details on model inputs). A gastrointestinal AF of 0.12 forages) 15 years
and older was used in the Leggett model to correspond to the default value in the ALM. The dotted line shows the adult PbB0
concentration predicted by the ALM. Cortical and trabecular bone lead concentrations from the Leggett model are shown on the
right vertical axis.
-------
41
-REVIEW OF ADULT LEAD MODELS-
FIGURE 2.5.4a. LEGGETT MODEL SIMULATION OF BASELINE BLOOD LEAD CONCENTRATION
AND CORTICAL AND TRABECULAR BONE LEAD
6 T
; I Leaflet! - Cort. Bone| iLeqqett-Trab.Bone
CD
a
03
CD
EL
10
20 30
Age (years)
40
90
Note: Figure 2.5 An shows the blood and bone lead output for the baseline simulation. The dashed line shows the range of
geometric mean PbB concentrations for U.S. females of different race/ethnicity categories from Phase I of NHANES III (Table 2
of Brody et al, 1994). The Leggett model default values for the gastrointestinal absorption fraction (Figure 2.5.2) were used in
these simulations.
Comparison of BKSFs derived from the IEUBK and Leggett models with ALM. The ALM uses a
linear BKSF derived from epidemiologic studies to predict blood lead concentrations resulting from
different soil lead exposures (EPA, 1994). Although biokinetic models, such as Leggett and IEUBK,
have nonlinear behavior at some blood lead concentrations, one can crudely approximate a linear
biokinetic slope factor relating lead uptake from exposure and blood lead from the model results. The
Leggett model predicts a BKSF of 0.43 in adults, which is similar to the BKSF of 0.4 chosen for the
ALM (Table 2.5.2.)
-------
42
-REVIEW OF ADULT LEAD MODELS-
FIGURE 2.5.4b. LEGGETT MODEL SIMULATION OF BASELINE BLOOD LEAD CONCENTRATION
30
25
20
15
O)
Q)
C3
J:
1 10
10
20 30
Age (years)
40
50
To obtain the BKSF for the Leggett model, a series of simulations was run assuming different lead
intakes that produced a range of lead uptakes and corresponding PbB concentrations. In the simulations,
exposure via only the ingestion pathway occurred over the age period 25-31 years. The lead intakes
simulated started at 10 |j,g/day and increased by steps of 40 |j,g/day until they reached a maximum of
810 |j,g/day. Next, the intake was multiplied by the absorption fraction to obtain the uptake; this
approach produced lead uptakes ranging from 1.5 to 121.5 |_ig/day in a total of 21 runs. The monthly
output after age 25 from the Leggett model was used to calculate 7-year averages for blood lead
concentration and lead uptake. The mean lead uptake was plotted against the mean PbB concentration
and the BKSF was obtained by simple linear regression (e.g., Figure 2.5.5). The BKSF from the Leggett
model exhibits a strong nonlinearity with increasing lead concentration. Below a mean lead uptake of
about 60 |j,g/day, the BKSF appeared approximately linear. Estimates of BKSFs from the Leggett model
were tabulated for various age ranges (see Table 2.5.2). The adult BKSFs predicted by the Leggett
model are lower than those predicted for children. The differences presumably reflect age-related
differences in the kinetics of tissue distribution or excretion of lead. This trend is distinctly different
from that predicted by the O'Flaherty model, which predicts a higher BKSF in adults than in children
(see Section 2.6). The simulations used to estimate the BKSF from the Leggett model are described in
greater detail in appendices.
-------
43
-REVIEW OF ADU LT LEAD MODELS-
FIGURE 2.5.5. BIOKINETIC SLOPE FACTORS (BKSFs) FOR ADULTS PREDICTED BY THE
LEGGETT MODEL
50
40
J 30
CJ
CD
£ 20
c
10
Linear Portion
y=0.43 »+• 0.002
BKSF = 0.43
Full Range
y=0.34x+0.2.73
BKSF = 0.34
20
40 60 BO 100
Mean Pb Uptake (|ig/day)
120
140
TABLE 2.5.2. COMPARISON OF LEAD UPTAKE SLOPE FACTORS BASED ON THE IEUBK,
ALM, AND LEGGETT MODELS
Age Group
1-84 months
25-48 months
25-32 years
Lead Uptake Slope Factor
EPA
0.34 (IEUBK)
0.33 (IEUBK)
0.4 (ALM)
Leggett
0.57
0.63
0.43
Percent Difference"
68
91
8
al OOx(Leggett-EPA)/EPA
-------
44
-REVIEW OF ADULT LEAD MODELS
Effect of child blood lead concentration on blood lead concentration during adulthood. A variety of
other simulations of the Leggett model (see the appendices) were run to explore the effects of early
intense exposures to lead on blood lead levels in the adult. Two scenarios were explored: (1) an infant
born with a high blood lead (perhaps from a high maternal exposure); and (2) a brief high exposure to
lead in early childhood (ages 2-3 years). The first scenario was simulated by raising the blood lead
concentration at birth from 2 |-ig/dL to 50 |-ig/dL and following this with a baseline simulation (NHANES
III). The second scenario was simulated by increasing the lead intake to 50 |j,g/day between ages two and
three (baseline simulation for all other ages) which resulted in a PbB concentration of approximately
30 |-ig/dL. Neither scenario had a substantial effect upon adult blood lead after age 17 years. Possibly,
the increases in blood volume with age diluted the lead concentrations in the blood and increased the
influence of current rather than historic lead exposures. Rapid turnover of bone and bone lead during
childhood may also contribute the predicted lack of effect of historic child exposures on adult blood lead
concentrations. However, the intense early childhood exposures represented in these simulations could
have an effect upon the developing nervous system; however, blood measurements in older teens or
adults are a poor indicator for early childhood exposures. This model, like the other biokinetic models
for lead, does not have a way to reflect, irreversible tissue damage tied to a given concentration in blood.
2.5.2 Evaluation Criteria
Biokinetics
The Leggett model was the most complex of the compartmental models that were evaluated, simulating
the largest number of compartments. The model includes several important features that are unique in
comparison to the other compartmental models that were evaluated:
• The model simulates lead kinetics from birth through adulthood.
• The central exchange compartment is blood plasma rather than whole blood.
• The model simulates the partitioning of lead between bound and unbound pools of lead
in plasma.
• Saturation of lead uptake by red blood cells is simulated.
• Trabecular and cortical bone are treated as kinetically heterogeneous compartments to
simulate fast, intermediate, and slow components of the uptake and release of lead from
bone.
• The model includes a compartment representing the brain, and a very simple fetal
compartment.
The ages and transfer coefficients that define intercompartmental lead exchanges can be easily modified
to accommodate more recent data. Compartments can be added or deleted without major reconstruction
of the entire source code; however, this process may require a recalibration (and validation) of the model.
The number of ages with explicit transfer coefficients can also be modified.
The model simulates the uptake of lead into red blood cells as either a saturable or non-saturable process.
The user can select either of two options (linear or nonlinear) and adjust the maximum concentration of
lead allowed in the red blood cells. This feature is useful because some uncertainty exists about the
impact of saturation on PbB concentration. The nonlinear option largely performed as expected in the
simulations. Substantial curvature in the PbB concentration-uptake relationship becomes apparent at a
PbB concentration of about 25 |_ig/dL. This results in a decrease in the BKSF when the estimated PbB
concentration exceeds 25 |-ig/dL; for example, the estimated BKSF for adults decreases from 0.43 at PbB
-------
45
-REVIEW OF ADULT LEAD MODELS
concentrations below 25 |-ig/dL to 0.24 at PbB concentrations exceeding 25 |_ig/dL. When simulations of
two different soil lead concentrations (i.e., intakes) were compared, the nonlinear option produced a
greater effect on PbB concentration at the higher lead exposure. However, although the linear and
nonlinear options produced noticeably different effects on PbB concentration, there was no noticeable
difference in their effects on cortical and trabecular bone lead. One would expect nonlinearities in PbB
concentration to have some effect on bone lead concentrations; however, the direction and magnitude of
the effect may be complex and not easy to predict. Red blood cell saturation would be expected to affect
plasma lead concentrations and, thereby, lead excretion and exchanges with other tissue compartments.
It appears from the simulations that saturation effects are not important to include in a biokinetic model
for site work unless concentrations of lead in soil are fairly high. Using the Leggett model, soil
concentrations would have to exceed 8,000 ppm before they would affect the linearity of the BKSF. If
the absorption fraction of the Leggett model is decreased to match that of the ALM (0.15 to 0.12), then
the soil concentration would have to exceed 11,000 ppm before it would affect the linearity of the BKSF.
The model does not currently simulate nonlinear kinetics of absorption of lead from the gastrointestinal
tract; however, this could be included in the model. The default gastrointestinal AF for adults is
0.15 which applies to all intake sources (e.g., diet, water, soil). This value is slightly higher than the
ALM default value, which is 0.12 for soil. The age-specific default values for the gastrointestinal AF are
easily modified in the Leggett model. When various simulations were run using the Leggett default
value, the predicted quasi-steady state PbB concentrations in adults were higher than those predicted by
the ALM; the two models converged to more similar predictions when a value of 0.12 was assumed in
both models.
Leggett (1993a,b) developed the initial ideas about the transfer coefficients from injection studies, which
confirmed the biological plausibility of the transfer coefficients from numerous ingestion and inhalation
studies (it is easier to derive transfer coefficients from injection studies). It has been suggested that
injection data should not be used as a basis for transfer coefficients in a model because injected lead may
not be in the same ionic form as lead that has been ingested or inhaled (e.g., discussions at the 1998 EPA
Lead Model Validation Workshop, Chapel Hill, NC). However, the significance of this to the
performance and utility of the Leggett model has not been evaluated.
Exposure
The model does not currently include an exposure component to calculate intakes from environmental
concentrations. The model includes three routes of intake: ingestion, inhalation, and injection (directly
to the blood compartment). Inputs to these pathways can be adjusted overtime to simulate exposures of
varying duration or intensity. Thus, exposure can begin in utero, at birth, or at any other age. For fetal
exposures, PbB concentrations at birth are calculated as a ratio of the maternal lead concentration, and
tissue lead concentrations at birth are set according to user-defined tissue-specific fractions of the body
burden. For exposures that begin at other times, lead masses in all compartments are null at the start of
exposure. A slight modification of FORTRAN code allows users to set the starting tissue lead masses at
times other than at birth to simulate a baseline starting body burden and PbB concentration. However,
the user cannot access these variables from the current input file. Therefore, baseline conditions can be
simulated by modifying lead intake.
The Leggett model can track and record tissue lead accumulation over any selected age range. The
model can also provide estimates of blood and tissue lead over a wide time frame that can vary from
minutes to decades.
-------
46
-REVIEW OF ADULT LEAD MODELS
The model does not currently have a graphics or statistics output. Data files are generated that can be
imported into spreadsheets for further tabular or graphical processing.
Ease of use/flexibility
The source code for implementing the model can be easily executed with a FORTRAN compiler on any
PC. User input involves editing an input text file that is read by the FORTRAN compiler. A more user-
friendly, menu-driven interface would need to be considered for wide-spread use as a regulatory model.
By contrast with the C++ code used in the O'Flaherty model, the FORTRAN code used in the model is
very flexible in terms of the integration time steps and communication intervals that can be selected. The
user has access to all of the biokinetic variables and constants. For example, the integration time steps
can range from 0.001 days for the first day to every day for longer term exposures. The model can
capture the results of every 100th calculation time step and present it in an output file. This provides a
good balance between computation time and output file size, as well as temporal resolution of the model
output. The model allows a variety of exposure situations; lead levels can change as much as 50 times
within one simulation. This feature is useful for simulating complex exposures, including intermittent
recreational or non-residential exposures or baseline exposures in combination with site-related
exposures.
2.5.3 Summary
The Leggett model is currently used widely within the radiation health risk assessment community. The
model has been tested against a wide variety of data sets, which include a number of radioactive elements
in addition to lead. The biokinetic component performed well for adult lead exposure in a wide variety
of exposure situations. The model also performed well, followed anticipated patterns, and gave
reasonable output for a wide variety of simulations. When the AF in the ALM and Leggett model were
made equivalent, the two models produced similar results. The Leggett model also produced similar,
although higher, results compared to the IEUBK model.
The model predictions fit observations for relatively high exposures in children and adults. The ability of
the model to predict lead distributions in children whose blood lead concentrations are more similar to
those experienced by the U.S. population (2-5 |_ig/dL) has not been tested (Pounds and Leggett, 1998.
A strength of the Leggett model is that it is an all-ages model that can simulate lead accumulation in a
variety of tissues over any selected age range for a wide range of lead intake patterns, including intakes
that vary in intensity over time. The Leggett model works well over a wide variety of conditions as
assessed by comparison between predicted and observed PbB concentrations and urinary or fecal
excretion in adults and comparisons of predicted and observed post-mortem tissue lead concentrations.
The flexibility of the model makes it relatively easily to simulate complex lead scenarios, including those
that might represent baseline exposures.
An additional strength of the model is that the source code for implementing the model can be easily
executed on any PC with a FORTRAN compiler. The user has access to all variables and constants in
the model. The model is very flexible in terms of the integration and communication intervals that can
be selected by the user to accommodate complex time-varying exposures.
A most significant limitation of the Leggett model is the lack of an exposure model. In order to simulate
an exposure scenario, the user must calculate and input the lead intakes (|j,g/day) that correspond to the
exposure scenario. This approach maybe too flexible for a regulatory model; guidance on calculating
intakes from exposure concentrations would be needed.
-------
47
-REVIEW OF ADULT LEAD MODELS
The lack of a menu-driven input interface may also be a challenge for the novice user. Other limitations
to the model include the lack of statistical and graphics modules and the inability to start simulations
from a baseline lead burden. The latter limitation requires the user to construct an intake scenario to
simulate the baseline condition (this is readily accomplished with the flexible lead intake module).
-------
48
-REVIEW OF ADULT LEAD MODELS
2.6 O'FLAHERTY
References: Toxicology and Applied Pharmacology 118: 16-29 (1993)
Toxicology and Applied Pharmacology 131: 297-308 (1995)
2.6.1 Introduction
The O'Flaherty model is an extension and modification of an earlier model of lead biokinetics in the rat
(O'Flaherty, 1991a,b). A distinguishing feature of this model is that it simulates the delivery of lead to
tissues as a function of blood flow to the tissues. The central plasma compartment has dimensions of
volume and flow, and transfers of lead from the central plasma compartment are simulated as entirely, or
partially, perfusion-limited processes. This general approach has been referred to as "physiological-
based" in some pharmacokinetics texts (e.g., Gibaldi and Perrier, 1982) to distinguish it from the more
traditional compartmental approach, such as that used in the Bert, Leggett, and Rabinowitz models. This
designation derives in part from the idea that transfer constants in perfusion-limited processes are given
by the tissue perfusion rates (Equation 2.6.1) (O'Flaherty, 1987):
dM} dC,
= V] = -Q(Cj - Cy ) Equation (2.6.1)
dt dt
where:
M! = mass of lead in the plasma (|j,g)
G! = concentrations in plasma and tissue water (|ig/L)
C2 = concentration in plasma and tissue water
Y! = the volume of plasma (L)
Q = the transfer constant and is equal to the tissue plasma flow rate
(L/min)
In compartmental models, the term Q would be replaced with a term reflecting the rates constants for
transfer between the tissue and plasma compartments (Equations 2.6.2-4):
~~~~ = Vl ' —- = ~kJ2 ' Vl ' Cl + k21 ' V2 ' C2 Equation (2.6.2)
At steady state,
dCl
v
V } '
-/ r- ^- /T £ ->\
dt Equation (2.6.3)
and
kt = k!2 ' V! = k2! ' V2 Equation (2.6.4)
-------
49
-REVIEW OF ADULT LEAD MODELS
where:
Iq = compartment clearance
k12 = first order rate constants for transfer from plasma to tissue and
tissue to plasma
k21 = first order rate constants for transfer from plasma to tissue and
tissue plasma
Y! = the apparent volumes of distribution to which the rate constants
apply (which may not be the actual plasma and tissue volumes)
V2 = the apparent volumes of distribution to which the rate constants
apply (which may not be the actual plasma and tissue volumes)
The transfer constant (Q) in Equation 2.6.2 is scaled across species or within species by accounting for
differences in perfusion rates, which are readily measurable entities. On the other hand, rate constants
and compartment volumes in Equation 2.6.3, are derived from modeling (e.g., curve fitting) of empirical
data on tissue and plasma concentrations, and different assumptions must be made in scaling these
constants when empirical data are not available, which is often the case. Conceptually, the approach
represented by Equation 2.6.2 is not necessarily "more physiological" than Equation 2.6.3. Both
approaches are simplifications of physiology, and in practice usually require numerous assumptions that
may not be completely accurate, and both approaches may yield similar approximations of the
measurable kinetics of blood lead for similar exposure scenarios. This appears to be the case for the
simulations of steady state PbB concentrations developed from the O 'Flaherty and Leggett models
described in this report.
At the time this analysis was conducted, the O'Flaherty model computer code was available in either
ACSL or C++ language. ACSL is a commercial software package that can implement a wide variety of
numerical integration algorithms for solving the simultaneous differential equations that are the core of
the O'Flaherty model. The major difference between the execution of the ACSL and C++ versions of the
model relates to user access to model variables and constants. All variables and constants can be
accessed in the ACSL code, whereas all variables and constants are not accessible in the C++ code.
Where such limitations are noted in this report, they refer to the C++ code which was provided by Dr.
O'Flaherty (PBKM, 5/26/97). Simulations that support the evaluations described in this report were
conducted, for the most part, with the ACSL (6/96), also provided by Dr. O'Flaherty, and no attempt was
made to compare results between the two programs. Dr. O'Flaherty provided the TRW a draft User's
Guide for the C++ program (O'Flaherty, 1997), which provided the basis for describing the structure and
algorithms in the O'Flaherty model.
Description of Biokinetics
The O'Flaherty model estimates age-specific exchanges of lead between blood plasma and specific tissue
compartments to simulate the disposition of lead in the human body. The current version simulates eight
tissue compartments: the lungs, gastrointestinal tract, liver, kidney, bone (trabecular and cortical), blood
plasma, well perfused tissues, and poorly perfused tissues (Figure 2.6.1). Exchanges between plasma and
soft tissues are simulated as flow-limited processes, whereas exchanges between plasma and bone are
represented as a combination of flow-limited and diffusion-limited processes. In flow-limited exchanges,
lead is assumed to instantaneously partition between tissue water and any bound or sequestered forms in
the tissue according to a partition coefficient; thus, the general form of Equation 2.6.1 becomes
(O'Flaherty, 1987):
dt dt IT Equation (2. 6. 5)
-------
50
-REVIEW OF ADULT LEAD MODELS
where:
C 2 = concentration of lead in whole tissue (including free and bound
forms)
II = partition coefficient; partition coefficient is a measure of the
affinity of a molecule for an environment
C 2/n = concentration of lead in tissue water, which is assumed to be the
same as venous plasma leaving the tissue
All lead in plasma is assumed to be available to exchange with tissues. Lead bound to plasma proteins is
assumed to be in rapid equilibrium with diffusible forms. Therefore, binding has no affect on rates of
exchanges with tissues. Blood flow to tissues, tissue volumes, and respiration rate are scaled to body
mass (lean body mass), which is represented by a continuous gender-specific growth function.
Therefore, exchanges of lead between plasma and tissues and between tissue lead masses and
concentrations are age-dependent as a consequence of the linkage between flows and body mass.
-------
51
-REVIEW OF ADULT LEAD MODELS-
FIGURE 2.6.1. COMPARTMENTS AND PATHWAYS OF LEAD EXCHANGE
IN THE O'FLAHERTY MODEL
Well-perfused Tissues
Poorly-perfused Tissues
Cortical Bone
Trabecular Bone
o
o
Elimination pools of the body
Intake from environmental
Media (ug lead/day)
Body compartment
Li/er
Kidney
-------
52
-REVIEW OF ADULT LEAD MODELS
O'Flaherty (1993) modeled the kinetics of exchange of lead between plasma and bone as three
interdependent processes that occur in series: (1) flow-limited exchange with surface bone; (2) exchange
between plasma leaving the bone surface and the metabolically active region of bone, limited by age-
dependent rates of bone formation and resorption; and (3) exchange between plasma leaving the
metabolically active region of bone and mature bone (where rates of bone formation and resorption are
low), limited by diffusion (representing ionic exchange of lead and calcium).
The 1993 version of the model was subsequently modified (O'Flaherty, 1995, 1997) as follows to include
separate compartments for cortical and trabecular bone, based on results of pharmacokinetic studies
conducted in non-human primates: (1) the separate module representing exchange with surface bone was
eliminated; (2) exchanges between plasma and metabolically active trabecular and cortical bone were
modeled as separate parallel processes; and (3) cortical bone was represented as containing both
metabolically active regions of bone formation and resorption as well as a region of lead-calcium
exchange, which receives blood leaving the metabolically active region of cortical bone. Exchanges
between plasma and bone are represented as age-dependent processes by their linkage to variables that
represent age-dependent rates of bone formation and resorption.
O'Flaherty (1995) derived expressions relating age and bone formation and resorption rates (and total
bone weight), based on various types of empirical observations made in humans, including studies of
bone histomorphology, bone turnover biomarkers, and calcium uptake. Important features of the bone
model derived from these data are as follows:
• Bone is classified as either juvenile or mature. Juvenile bone has a high rate of formation and
resorption (turnover); mature bone has a low rate of turnover. The relative portion of bone
mass that is juvenile transitions from 100 percent at birth to 0 percent at age 25 years. The
bone volume-age relationships for juvenile and mature bone as represented in the O'Flaherty
model are shown in Figure 2.6.2. Osteoporosis is not modeled.
• Bone is assumed to consist of 80 percent cortical and 20 percent trabecular bone. The relative
distribution of bone turnover that occurs in mature trabecular and cortical bone is
approximated by observations made in Cynomolgus monkeys (C.P. Jerome, unpublished).
Turnover rate of mature bone is assumed to be 10 percent/year; 65 percent of bone turnover is
assumed to occur in trabecular bone and 35 percent in cortical bone. (Note: It is not clear
whether this same distribution was applied to juvenile bone.) Thus, trabecular bone turnover
rate is 32.5 (0.65x0.10/0.2) percent/year, and cortical bone turnover is approximately
4.4 percent (0.35x0.1/0.8)/year. The bone volume-age relationships for cortical and trabecular
bone as represented in the O'Flaherty model are shown in Figure 2.6.3.
• The relationship between bone formation rate and age is approximated by an empirical
relationship between age and bone calcium uptake in humans.
• The relationship between bone resorption rate and age is given by the bone formation rate and
an empirically-based allometric expression relating bone mass and body mass.
Description of the Exposure Component
The model includes two routes of lead intake: ingestion and inhalation. Lead concentrations in dust, soil,
drinking water, infant formula or milk, and air (both ambient and workplace) are inputs to each
simulation. Lead intake rates from these media are computed using medium-specific ingestion or
respiration rates as a function of age and gender. Soil and dust intake are explicitly modeled as age-
specific ingestion only in children (0.3-6.5 years). The ingestion of dust and soil lead by adults may be
simulated using a model term for miscellaneous adult lead ingestion. Food lead intakes for young
children and adults also may be specified. Intakes from lead in ambient air and food vary with date of
-------
53
-REVIEW OF ADULT LEAD MODELS-
birth; this relationship simulates the declines of lead in food and air since 1970 and 1975, respectively.
Fetal exposure is not addressed in the model, although the ACSL code may be modified to set initial
integration conditions of any parameter (e.g., PbB concentration) at any age, including at birth.
Description of Uptake
AFs are linear multipliers in the calculation of lead uptake by various routes of exposure. The
gastrointestinal AF represents the fraction of bioavailable ingested lead that is absorbed from the
gastrointestinal tract and transported to the liver. The AF is not medium-specific in the O'Flaherty
model; it is simulated as a function of age, ranging from 0.58 at birth to 0.08 for ages eight to adult
(Figure 2.6.4):
AF = 0.6-
0.52
30-s
Age
Equation (2.6.6)
where:
AF
Age
absorption fraction
age in months
FIGURE 2.6.2. BONE VOLUME-AGE RELATIONSHIPS FOR HUMAN JUVENILE AND
MATURE BONE AS REPRESENTED IN THE O'FLAHERTY MODEL
O'Flaherty Model: Bone Volume (Females)
-------
54
-REVIEW OF ADULT LEAD MODELS-
FIGURE 2.6.3. BONE VOLUME-AGE RELATIONSHIPS FOR HUMAN JUVENILE CORTICAL AND
TRABECULAR BONE AS REPRESENTED IN THE O'FLAHERTY MODEL
0.45
0.40
0.35
_0-30
£ 0.25
"3
0) 0.20
o
CD
0.15
0.10
0.05
0.00
Juvenile
Trabecular Bone
10 15
Age (years)
Juvenile Cortical
Bone
20
25
-------
55
-REVIEW OF ADULT LEAD MODELS-
FIGURE 2.6.4. DEFAULT VALUES FOR GASTROINTESTINAL LEAD ABSORPTION AS A FUNCTION
OF AGE AS SIMULATED IN THE O'FLAHERTY MODEL
60
o
"S.
o
til
T3
as
CD
SO-
40-
30-
20-
10-
0.52
60
120 180
Age (rnonth^
240
:300
Note: The function relating age in months and the absorption fraction (AF) is given in this figure.
Relative bioavailability factors are used in the O'Flaherty model only in conjunction with childhood
(0.3 to 6.5 years) dust and soil lead exposures. These bioavailability factors refer to the fractions of
ingested dust or soil lead that are available for absorption through the lining of the gastrointestinal tract
(O'Flaherty, 1997). Bioavailability values in the O'Flaherty model are equal to 1 for exposures other
than dust or soil lead ingestion. The lung AF represents the fraction of inhaled lead that is absorbed into
arterial blood. The default AF value is -0.3 for all ages, although this value may be adjusted by the user
to simulate changes in the lung AF over time.
Calibration and Evaluation
The O'Flaherty model has been developed from data on observed blood and skeletal lead concentrations
associated with short-term and long-term lead exposures. These exposures include epidemiological
observations of human children and adults, adult human experimental exposures, and non-human primate
exposures (O'Flaherty, 1993, 1995, 1998; O'Flaherty et al., 1998).
Simulations
The model is represented mathematically as a series of coupled differential equations (O'Flaherty, 1997).
These equations were implemented in ACSL which is a commercial software package that can implement
a wide variety of numerical integration algorithms for solving the simultaneous differential equations.
The ACSL version of the O'Flaherty model uses a Gear integration algorithm which integrates over a
time step that varies according to the rates of change of the various variables being computed. Where the
-------
56
-REVIEW OF ADULT LEAD MODELS-
rate of change is low (e.g., PbB concentration is changing slowly), the integration time step is increased.
This approach serves to decrease computation time.
The O'Flaherty model yields estimates of quasi-steady state maternal PbB concentrations that are very
close to those predicted by the ALM, if similar values for the absorption fraction (AF) and soil ingestion
rate (IRS) are used to estimate lead uptakes, and if the same values for the baseline blood lead
concentration (PbB0) are assumed. Table 2.6.1 provides the inputs that were used in a typical set of
simulations in which the outputs of the ALM and O'Flaherty model were compared. The soil lead
exposure was assumed to be to 1,000 |ig/g, 5 days/week for 260 days/year. A soil ingestion rate of
0.05 g/day was assumed. Note, the soil ingestion rate parameter is not a component of the O'Flaherty
model; however, it was used to calculate lead intakes that would be equivalent to those simulated in the
ALM for the same soil lead exposure. The gastrointestinal absorption fraction in the O'Flaherty is
calculated with Equation 2.6.6 which yields a value of 0.08 for adults (see Figure 2.6.4). This was not
modified for comparisons with the ALM, which uses a value of 0.12 for the AF in adults.
TABLE 2.6.1. INPUTS FOR COMPARISON SIMULATIONS OF THE O'FLAHERTY MODEL AND
ALM SHOWN IN FIGURE 2.6.6
Parameters
IRS
AF
PbB0
EF
ED
Output
EPA (1996)
l,OOOLlg/g
0.05 g/day
0.12
2 Llg/dL
5 days/week (260 days/year;
model default is
219 days/year)a
23 years
Maternal PbB concentration
O'Flaherty (1996)
1000 |-lg/g (not a parameter in the model)
0.05 g/day (not a parameter in the model)
0.08 (default)
To simulate female PbB0, inhalation exposures and
child ingestion were minimized to yield a PbB
concentration approximating 2 flg/dL at >17 years of
age. The model was then run using an additional adult
ingestion intake equal to * IRS
5 days/week (260 days/year) (O'Flaherty model
modified to include this parameter)
40 year (not a parameter in the model) (childhood PbB
simulation is obligatory)
Adult female PbB concentration
aThe default exposure frequency for the ALM is 219 days/year; however, the assumption of 260 days/year in the
simulations would not change the outcome of the model comparisons.
The O'Flaherty model does not include an input for the PbB0 term. Therefore, in order to compare the
O'Flaherty model with the ALM, the PbB0 was simulated. The baseline was constructed by altering the
Pb intake (|j,g/day) at various ages to produce a PbB concentration profile that matched as closely as
possible the age-specific PbB concentrations reported for U.S. females in Phase I of NHANES III study
(Brody et al, 1994) and approximated 2 |-ig/dL (the default in the ALM) in adults. The resulting
PbB0 profile is shown in Figure 2.6.5. The upper and lower bound geometric mean PbB concentration
for U.S. females of different race/ethnicity from Phase I of NHANES III study (from Table 2 of Brody et
al., 1994) are shown for comparison. Also shown in Figure 2.6.5, are the bone lead concentrations (total
mass of lead in bone, cortical and trabecular, divided by the bone volume) predicted by the model (right
-------
57
-REVIEW OF ADULT LEAD MODELS-
vertical axis). Note that the bone lead concentrations are somewhat higher than those predicted by the
Leggett model for the baseline simulation (Figure 2.5.4a.).
FIGURE 2.6.5. O'FLAHERTY MODEL SIMULATION OF BASELINE BLOOD LEAD (PbB)
CONCENTRATION
35
40
Figure 2.6.5. shows the blood and bone lead output for the baseline simulation. The dashed line shows the range of geometric
mean PbB concentrations for U.S. females of different race/ethnicity categories from Phase I of NHANES III (Table 2 of Brody
et al., 1994). The O'Flaherty model default values for the gastrointestinal absorption fraction (Figure 2.6.4) were used in these
simulations.
To simulate an adult exposure to 1,000 |_ig/g soil lead, a baseline simulation was run with an additional
intake from soil included in the simulation beginning at age 17 years (the product of 1,000 |_ig/g soil and
an IR of 0.05 g soil/day). Figure 2.6.6 compares the PbB concentrations predicted by the ALM and the
O'Flaherty model (using inputs presented in Table 2.6.1). The ALM predicted a steady state PbB
concentration of 3.7 |-ig/dL, whereas the O'Flaherty model predicted slightly higher values which ranged
from 3.6 |-ig/dL at age 18 years to 4.7 |-ig/dL at age 40 years. The average for the last 10 years of the
simulation (ages 30-40 years) was 4.6 |_ig/dL. Note, these blood lead concentrations were obtained with
the O'Flaherty model simulations when the default values for the absorption fraction in the O'Flaherty
model was used (i.e., 0.08 for adults). If the ALM default AF value of 0.12 was used in both models, the
difference between the two models would have been greater. Rather than attempting to modify the AF
algorithm in the O'Flaherty model (Equation 2.6.6), the effect of the difference in the AF used in the two
models can be demonstrated by replacing the default value of 0.12 in the ALM with the O'Flaherty adult
value of 0.08. This change results in a predicted PbB concentration of 3.1 |-ig/dL from the ALM,
-------
58
-REVIEW OF ADU LT LEAD MODELS-
compared to the average value of 4.6 for the O'Flaherty model. Also shown in Figure 2.6.6 are the bone
lead concentrations predicted by the model (right vertical axis). Note that the bone lead concentrations
are somewhat higher than those predicted by the Leggett model for the same adult exposure scenario
(Figure 2.5.3). The ALM also calculates a 95th percentile fetal PbB concentration, which is not shown in
the figure.
The BKSF predicted by the O'Flaherty model is approximately 0.65 in adults, which is higher than the
value of 0.4 assumed in the ALM (see Figure 2.6.7). To estimate slope factors from the O'Flaherty
model, soil lead exposure was simulated in adults as either an intermittent "non-residential" exposure,
5 days/week with a soil ingestion rate of 50 |j,g/day), or continuous exposure, 7 days/week with a time-
weighted average soil lead intake equivalent to the latter intermittent exposure. Adult simulations were
run assuming either default exposures from birth to 17 years, followed by the incremental exposure to
soil, or lead exposure was initiated at age 17 years. Slope factors derived from the O'Flaherty model
exhibit a strong nonlinearity with increasing PbB concentration. Therefore, separate SFs were estimated
for PbB concentration ranges: <10 |-ig/dL and >10 |_ig/dL. Estimates of BKSFs for adult ages for various
exposure scenarios and age ranges are summarized in Table 2.6.2. Table 2.6.2 shows that the adult
BKSFs predicted by the O'Flaherty model are consistently higher than BKSFs predicted for children.
All things being equal, a model with a slope factor will tend to predict a greater increase in blood lead for
a given intake of lead.
TABLE 2.6.2. COMPARISON OF LEAD UPTAKE SLOPE FACTORS (BKSF) FOR ADULTS BASED
ON THE ALM AND O'FLAHERTY MODELS
Exposure scenario
"Intermittent non-
residential",
5 days/week,
1 00-3,000 ppm
"Continuous non-
residential" time-
weighed average
of 5 days/ week,
100-3,000 ppm
Initial Pb exposure
(age-year)
0
0
PbB and uptake
averaging time
(age-years)
17-45
25-32
17-45
25-32
O'Flaherty BKSF
(|J.g/dL per
|ig/day)
0.64
0.62
0.65
0.64
Percent difference
fromALM"
60
55
62
60
'100x(O'Flaherty-EPA)/EPA
-------
59
-REVIEW OF ADULT LEAD MODELS-
FIGURE 2.6.6. COMPARISON OF BLOOD LEAD CONCENTRATION PREDICTED BY THE
O'FLAHERTY MODEL AND ALM
Baseline + Soil Blood
(O'Flaherty)
Baseline Blood
(O'Flaherty)
10 15 f
Begin Soil Exposure
20 25
Age (yrs)
30
35
40
Note: The O'Flaherty model simulation includes a baseline exposure followed by additional exposure to 1,000 ppm soil lead,
5 days/week, beginning at age 17 years (see Table 2.6.1 for details on model inputs). A gastrointestinal AF of 0.12 for ages)
15 years and older was used in the Leggett model to correspond to the default value in the ALM. The dotted line shows the adult
PbB0 concentration predicted by the ALM. Cortical and trabecular bone lead concentrations from the Leggett model are shown
on the right vertical axis.
-------
60
-REVIEW OF ADU LT LEAD MODELS-
The differences presumably reflect age-related differences in the kinetics of tissue distribution or lead
excretion. This was examined further in the O'Flaherty model by tracking the blood-to-urine (C^^) and
blood-to-bone (C^.bo) clearances of lead in children and adults, which is easily accommodated in the
ACSL version of the O'Flaherty model. Both Cb,.ur and Cbl.bo were higher in young children relative to
adults. Thus, under the simulated exposure conditions, the O'Flaherty model predicts that a higher lead
uptake rate to bone per unit of blood volume (or body mass) and higher kidney clearance of lead from
blood in young children, resulting in a smaller increase in PbB concentration for a unit change in uptake
rate (uptake slope factor) in children as compared to adults. A more detailed description of these
simulations are provided in appendices.
A second possible explanation for higher BKSFs being predicted for adults than children by the
O'Flaherty model is the release of sequestered bone lead in adults. This was examined by comparing
BKSFs from simulations that initiated all lead exposures at age 17 with slope factors from simulations
that initiated lead exposures at birth (see Table 2.6.2). In the simulations that initiated lead exposures at
age 17, there was no lead in blood or bone at the beginning of adulthood (this unrealistic scenario was
adopted for the purpose of revealing an effect of prior body burden on the adult SF). The resulting adult
SFs from adult-only lead exposures were the same, 0.65 |-ig/dL per |j,g/day for age groups 17-45 and
25-32-years-old, respectively, and were nearly identical to adult SFs predicted when lead exposures
began at birth (0.64 for age groups 17-45 and 25-32-year-olds, respectively). This data would suggest
that accumulation of a lead burden in bone does not appear to greatly affect the BKSF predicted by the
O'Flaherty model.
The BKSFs predicted by the O'Flaherty and IEUBK models are compared in Table 2.6.3. Slope factors
derived from the O'Flaherty model exhibit a strong nonlinearity with increasing PbB concentration;
therefore, separate SFs were estimated for PbB concentration ranges less than or greater than 10 |_ig/dL.
The BKSFs predicted by the O'Flaherty model were approximately 12-25 percent lower than those
predicted by the IEUBK model, depending on the age averaging range used in the estimation of the slope
factors.
TABLE 2.6.3. COMPARISON OF BKSFs PREDICTED FROM THE IEUBK AND
O'FLAHERTY MODELS
Age group
(months)
1-84
1-84
25-48
25-48
PbB range
(M-g/dL)
<10
>10
<10
>10
Uptake slope factor (|j,g/dL per
Hg/day)
IEUBK
0.36
0.35
0.35
0.34
O'Flaherty
0.32
0.28
0.26
0.23
Percent
Difference"
ll.l
20.0
25.7
32.4
a 1 OOx(O Tlaherty-IEUBK)/IEUBK.
-------
61
-REVIEW OF ADULT LEAD MODELS-
FlGURE 2.6.7. BlOKINETIC SLOPE FACTORS FOR ADULTS PREDICTED BY THE O'FLAHERTY
MODEL
10
O)
CQ
(C
y=0.646x-i- 0.227
R2= 1.000
BKSF = 0.65
5 10
Mean Pb Uptake (jig/d)
15
2.6.2 Evaluation Criteria
Biokinetics
The O'Flaherty model is an example of a modeling approach known as physiologically-based
pharmacokinetic (PBPK) modeling. The PBPK modeling approach used in the O'Flaherty model is
similar to the approach used in the compartment lead models (e.g., Leggett model) in that lead
biokinetics are simulated as a series of tissue compartments that exchange lead with plasma or blood.
However, the O'Flaherty model has several important features that distinguish it from the compartmental
models. Unlike the compartment models that represent the central compartment as a static volume, the
central compartment of the O'Flaherty model (i.e., blood plasma) is represented as dynamic process that
is characterized by volume and flow, and flows reflect actual rates of blood flow to the discrete tissues
represented in the model. Lead is assumed to partition instantaneously between plasma and tissues
according to an assumed equilibrium (i.e., partition coefficient). Therefore, the rates of change of lead
masses in tissues are limited by the rates of delivery of lead to the tissues (i.e., flow-limited exchange),
rather than by rate-limiting steps in the transfer of lead across tissue boundaries (i.e., transport-limited
exchange) as in compartment models. The O'Flaherty model is, for the most part, a flow-limited model,
in that exchanges between plasma and soft tissues are simulated as flow-limited processes. Exchanges
between plasma and bone are represented as a combination of flow-limited and diffusion-limited
processes. The model achieves a mass balance with respect to lead uptake and elimination (lead
absorbed minus lead excreted), flow (cardiac output minus sum of blood flows to all tissues) and lead
mass (lead absorbed minus lead excreted minus body burden).
-------
62
-REVIEW OF ADULT LEAD MODELS
In calibration studies, the O'Flaherty model predicted PbB concentrations that agreed reasonably well
with observed concentrations associated with short-term and long-term lead exposures. These include
epidemiological observations of human children, adult human experimental exposures, and non-human
primate exposures (O'Flaherty, 1993, 1995, 1998; O'Flaherty et al, 1998). As these were calibration
studies, the model was adjusted to optimize the fit to the empirical data. In the large epidemiologic study
on adults presented by O'Flaherty, the model fit was reasonable for the data in the center of the
distribution, but it was not possible from the presented data to evaluate the fit in the tails of the
distribution.
The O'Flaherty model predicts a gradual upward trend in the adult PbB concentration over time at a
constant lead intake until ages 60-80 years. This prediction is distinctly different from the assumptions
in the ALM and the predictions from the Rabinowitz and Bert models, which predict a more rapid
approach to a quasi-steady state in blood. The Leggett model also predicts an upward trend in PbB
concentration with age. There is no overwhelming empirical support for this trend or for the more rapid
approach to steady-state predicted by the other models.
The O'Flaherty model represents gastrointestinal absorption of lead as a constant fraction of intake and
does not simulate a saturable component to absorption. This approach is similar to the ALM, but distinct
from the uptake module in the IEUBK model.
The O'Flaherty model predicts a higher BKSF in adults than in children, which is, in part, the result of
higher clearances of lead from blood to bone and urine in children. The model also predicts a SF for the
age range 24-36 months that is lower for the age range 0-84 months; peak PbB concentrations are
predicted to occur in children ages 4-7 years. Empirical data to support a lower BKSF is lacking.
Epidemiologic data suggest that peak PbB concentrations occur in children over the age range
12-36 months (Brody et al., 1994; Pirkle et al., 1998); this may result from a combination of age-related
behavioral and physiological changes and, thus, is not necessarily inconsistent with the O'Flaherty model
predictions.
The predicted bone lead mass and concentrations are highly dependent on the assumed skeletal growth
curves for which empirical support is relatively weak. It is unclear whether the model adjusts
appropriately for the rapidly changing bone volumes in young children and adolescents when calculating
the lead concentrations in bone. Changes in bone volume with age would be expected to produce greater
interindividual variation in bone lead concentrations during these age ranges.
Exposure
The O'Flaherty model has an exposure component as well as a biokinetic component. The model
simulates lead intakes from air, food, water, or soil. Exposure from all these media is of interest to the
EPA. The model simulates age-specific dietary lead intakes by interpolating between pre-1970 intakes
(200 |j,g/day) and current dietary intakes (10 |j,g/day); thus, the dietary intakes in any given simulation are
linked to a user-specific birth date and cannot be modified easily (PBKM 5/26/97). A more flexible
exposure model that would allow the user to assign age-specific exposures would be needed for site risk
assessments. An option for setting a baseline PbB concentration and corresponding tissue lead burdens
would be useful as well as user specification of AFs for each environmental medium.
The O'Flaherty model should be flexible enough to consider gender, racial, or ethnic differences in
exposure or physiology, if quantitative information is available. It is possible that racial differences in
body weight, muscle to bone mass ratios, or blood mass may affect the model output. The model also
lacks the probabilistic component that would be needed for looking at population risks. The exposure
assumptions in the current version of the model overestimate present day population blood lead
concentrations, estimated from NHANES data.
-------
63
-REVIEW OF ADULT LEAD MODELS
The C++ program currently used by the O'Flaherty model (PBKM, 5/26/97) reports only blood and bone
lead concentrations. The output is easy to read and understand. It would be relatively easy to change the
model to report other tissue concentrations; however, while this functionality maybe useful for model
research, it may not be useful for risk assessment purposes. Validation of the model would be required
for making predictions about lead concentrations in other tissues; validation may not be possible given
the limited data available. The ACSL code allows output of the values of all variables in the model.
Ease of use/flexibility
The model is fast and easy to use as long as the user needs fairly simple and standard exposure
assumptions. The User's Guide is easy to understand for routine applications. The input demands would
be sufficient for most routine sites. Modifications to the computer code to simulate more complex
exposure situations is difficult (PBKM, 5/26/97). Simple changes in medium-specific lead levels and
intake rates over time may be simulated readily; however, the simulation of complex intake scenarios
may require extensive manipulation of either data input files (Windows), a batch command file (ACSL),
or the source code (ACSL). Due to the computation difficulties mentioned previously, it might be
advisable to re-program the model using the same initial equations and concepts if the model becomes
widely used.
The O'Flaherty model ACSL code uses an integration algorithm that automatically adjusts the integration
time step according to the rate of change in calculated variables, while the communication interval (i.e.,
the frequency with which variable values are reported) is constant. As a result, there can be a mismatch
between the time of communication and time of the last integration. For example, rapid changes in PbB
concentrations over time will automatically be captured in the model by downward adjustments of the
integration time step; however, these may not be reported if the communication interval is too large
relative to the rates of change of the PbB concentration.
Modifications to the model
Functions that this model lacks or performs poorly could be designed into the All Ages Lead Model, or a
refined version of this model could be built. Some of these functions are mentioned above. A model
could also include capabilities for sensitivity analysis, so that one could easily identify the most
important parameters in a complex exposure and biokinetic situation. A valuable addition to the
O'Flaherty model would be a fetal component, or a component to simulate lead changes during
pregnancy, that could predict whether or when the fetal blood reaches the target concentration during
pregnancy. The O'Flaherty model (PBKM, 5/26/97) lacks a graphical interface; such an interface is
particularly important when running quick simulations to determine whether the simulation is reasonable.
A graphical interface also facilitates simple bookkeeping when running multiple simulations. Some
improvements in the user interface are necessary to model complex exposure situations. The ability to
simulate changes in PbB concentrations starting from a baseline PbB concentration, rather than making a
lot of long-term assumptions about lead exposures during childhood and young adulthood, would be
helpful. This model, like most of the biokinetic models, lacks a way to account for the variability in
environmental lead concentrations and in the population response to lead exposure.
-------
64
-REVIEW OF ADULT LEAD MODELS
2.6.3 Summary
The major attributes of the O'Flaherty model are: (1) the model simulates PbB concentrations for ages
from birth through adulthood; (2) the model simulates both short-term and long-term exposures; (3) the
model can simulate biokinetics of females or males; (4) the model calculates tissue lead accumulation in
any compartment for any age range; (5) the model simulates nonlinear kinetics of lead in blood; (6) the
model can be calibrated to achieve a reasonable fit to epidemiologic and experimental data and; (7) the
model output includes graphics (ACSL version only).
The major limitations of the model include relatively weak empirical support for some of the model
components; the model is not designed to simulate maternal biokinetics during pregnancy; the exposure
module for adults is limited to age-specific intakes; and a variety of limitations in the C++ code,
including limited graphics capability, cumbersome user interface, and limited access to certain variables
and constants. Moreover, the program crashes when certain combinations of simulation durations and
communication intervals are selected.
The O'Flaherty model fits a variety of field and laboratory data reasonably well. The model appears to
predict higher PbB concentrations than the ALM given the same exposure assumption; thus, if used in
site risk assessment, the model is likely to support lower risk-based soil lead concentrations than the
ALM. The program that implements the model could be redesigned to reduce the potential for user
errors.
Another issue pertaining to the model is that some of the model runs, in the hands of an inexperienced
user, can produce results that are incorrect because the users may not know when they have chosen an
incorrect combination of communication intervals and integration time. Both the lead excretion and
absorption rates for children and the uptake and exposure parameters for lifetime exposure may need re-
working. For example, the default exposure assumptions in the model produces population blood lead
concentrations much higher than those found in NHANES IE. And because this model is based on a
balance of flows rather than amass balance of lead, any parameter changes must be checked carefully to
make sure it is consistent with the physiological requirements.
Overall, although the output of the model is plausible, the empirical support for the model is incomplete.
Uncertainties derive from the limited empirical support for model components such as kinetics of lead in
plasma; discussion of lead in bone; age-related changes in lead absorption; age-related changes in bone
formation and resorption; and variability in body and organ growth, particularly in children.
-------
65
-REVIEW OF ADU LT LEAD MODELS
This page is intentionally left blank.
-------
66
-REVIEW OF ADU LT LEAD MODELS-
3.0 Summary and Recommendations
The TRW compared six models for assessing soil lead at hazardous waste sites to the ALM (as detailed
in Recommendations of the Technical Review Workgroup for Lead for an Interim Approach to Assessing
Risks Associated with Adult Exposures to Lead in Soil, EPA, 1996). This evaluation focused on
application of the models to Superfund sites or RCRA Corrective Action facilities where adult exposure
to lead in hazardous waste is an issue.
3.1 SUMMARY OF MODEL REVIEWS
As part of the review, the TRW analyzed the performance of the models both qualitatively and
quantitatively. Simulations were run under widely varying exposure assumptions to assess the sensitivity
and performance of each model and to also determine under which set of conditions the exposure
assumptions break down. Each model was also compared using the same set of exposure conditions.
Additionally, the TRW reviewed features of each model, critiqued the scientific basis for each model,
and reviewed and interpreted the simulation model runs.
The six models that were compared to the ALM included slope factor approaches (California, Stern) and
biokinetic models of lesser (Rabinowitz, Bert) and greater (Leggett, O'Flaherty) complexity. Given the
disparity in the forms and function of the models relative to the ALM, the TRW used a semi-quantitative
system to compare them to the ALM. The strengths and weakness of each model are shown in Table 3.1.
Based on how the models performed in each evaluation category, they were assigned a (+), (-), or a (0).
A score of (+) indicated that the model performed better or offered some advantage over the ALM in a
given category, a score of (-) indicated that the model did not perform as well as the ALM in a given
category, and a score of (0) indicated that the model showed no significant difference in performance
from the ALM in a given category. A score of (0/+) indicated that the model showed marginal
improvement over the ALM in a given category.
TABLE 3.1. SUMMARY OF MODEL SCORES BY EVALUATION CATEGORY"
Evaluation Categories
Category 1:
Exposure
Category 2:
Kinetic
performance
Category 3:
Output
Category 4:
Ease of use/flexibility
Models Reviewed
Rabinowitz
0
+
0
0/+
California
+
0
0
+
Bert
0
+
0
0/+
Stern
0/+
0
0
-
Leggett
-
+
-
0/+
O'Flaherty
+
+
0
-
'Values were adjusted from earlier numerical rankings.
In Category 1 (Exposure), the California and O'Flaherty models were determined to offer some
advantage over the ALM because they allow consideration of input from media other than soil and dust.
By contrast, other media are included only as part of the background exposure in the ALM. The Leggett
model was determined to be less attractive than the ALM, since it does not contain an exposure
-------
67
-REVIEW OF ADULT LEAD MODELS
component. Furthermore, none of the methods presented have exposure components that fully satisfy the
TRW's requirements. The Stern model received a (0/+) score in Category 1 because the use of
probability distribution functions in the exposure component showed some benefit over the approach
used in the ALM.
In Category 2 (Kinetic Performance), the O'Flaherty, Bert, Rabinowitz, and Leggett models all offered
some advantage over the ALM. These four biokinetic models are all based on empirical data. In
addition, the O'Flaherty and Leggett models incorporate saturation kinetics, have been tested against
large data sets, and both are biologically plausible. Only the Leggett model contained a simple
fetal/maternal module.
In Category 3 (Output), the Leggett model, which does not contain an output component for summarizing
and displaying the model simulations, was determined to be less useful than the ALM and thus, rated
lower in this category. However, the outputs of all other models tested were generally inconsistent with
the TRW's reporting requirements (i.e., regulatory application and supporting risk management
decisions). Therefore, none of these models offered any significant improvement over the ALM.
In Category 4 (Ease of Use/Flexibility), the TRW determined that the California model was the easiest to
use and offered the most flexibility with regard to modifying model default inputs. The Leggett,
Rabinowitz, and Bert models were all easy to use, modify, and understand mathematically, but only if the
user understands Fortran, simulation, or pharmacokinetic software. The O'Flaherty model, which is
more data intensive, presented a disadvantage for site-specific adult exposure assessment, since it
requires the user to estimate cumulative lifetime exposure from infancy. If earlier estimates of chronic
lead exposure from media, such as diet and air, are overestimated, then the resulting PbB concentration
of the adult will be overestimated. The O'Flaherty model is easy to use when run with default settings;
however, the model and its programming language are not easy to manipulate for other purposes (e.g.,
determining baseline PbB concentration). The Stern model received a (-) score in Category 4 because the
model's de minimis approach to evaluating risk does not fit into the EPA paradigm of relative risk, and
also because of the complexities the TRW identified with the PDFs.
Overall, the majority of the models gave very similar results to the ALM when identical exposure inputs
were used and when the exposures were relatively low (i.e., exposures to soil concentrations between
1,000 and 10,000 ppm). The TRW determined that, when each model was run as intended by the author,
differences in results between each model could be attributed mostly to variations in exposure
assumptions.
The TRW compared the performance of the models that incorporated nonlinear saturation of the carrying
capacity for lead in blood (O'Flaherty, Leggett) to the models without such functionality (California,
Bert, Stern, Rabinowitz, and the ALM). It was observed that saturation effects started to appear at
exposures ranging from 8,000 to 11,000 ppm soil lead concentration. These saturation effects were
shown to increase slowly with increasing soil lead concentration. All the models performed similarly at
the lower soil average concentrations typical of many smaller Superfund and RCRA sites
(1,000-10,000 ppm); however, when average soil lead concentrations increased above 100,000 ppm, a
biokinetic model with a non-linear saturation component was better suited to estimate risks.
Selected components of individual methods, if incorporated into the ALM, could improve or make the
ALM more flexible. Some models had features that would improve performance under more widely
varied exposure conditions than is typically seen at Superfund and RCRA Corrective Action sites.
However, none of the models evaluated, as written by their authors, were great improvements upon the
ALM.
-------
-REVIEW OF ADULT LEAD MODELS-
3.2 RECOMMENDATIONS
The TRW established a decision tree (see Figure 1.1) with the following possible outcomes from the
model reviews:
• Replace the existing ALM with a superior model.
• Create a hybrid model from available components.
• Retain the ALM.
As a result of this evaluation, no single model was judged to be a significant improvement over the ALM,
although various components from the different models were determined to be refinements in adult lead
modeling. Collectively, these components could be integrated into a hybrid model; however, such
modifications would require a long-term effort. Therefore, the TRW recommends that the ALM should
be retained as an interim methodology. At the present time, EPA's ORD is developing an all ages model
that is expected to offer improvements over the ALM.
All of the models evaluated will perform adequately if exposures to lead in soil are relatively low
(1,000-10,000 |ig/g in soil if other environmental exposures are at typical background levels) and
chronic. If exposures are high enough such that the PbB concentration approaches saturation (25 |-ig/dL
or greater), or the exposures are highly variable or acute in nature, then the O'Flaherty or Leggett
biokinetic models maybe more useful. The Bert model is the easiest biokinetic model to understand;
however, the model does not contain a saturation component and should not be used for exposures where
PbB concentrations are likely to approach saturation. The Bert model can provide information on the
combination of exposures that will result in a PbB concentration close to 25 |_ig/dL. The Bert model can
be used for both short- and long-term exposure conditions, but has not been thoroughly tested on a
variety of data sets.
The TRW does not endorse many of the exposure assumptions made by the models. In some early
biokinetic models, such as Rabinowitz, the exposure assumptions do not reflect current conditions. In
the O'Flaherty model, exposure inputs from childhood to adult are required. Consequently, if the user
over-estimates the chronic lead exposure received during childhood then the predicted adult PbB
concentration will be too high. Therefore, the TRW recommends that the exposure assumptions for any
of these models be reviewed prior to using for site-specific purposes.
The TRW recommends continued use of the ALM for most site work, with the caveats listed in the
methodology. The TRW has and will continue to issue fact sheets to provide answers to frequently asked
questions regarding use of the interim ALM and may incorporate modifications to model input
parameters as new information becomes available. In situations where (1) multi media lead exposure is
being modeled; (2) exposure is acute or highly variable in nature; (3) receptor populations span a wide
age range; or (4) saturation kinetics are anticipated, one of the alternative models reviewed in this report
may be appropriate. However, given the current development stage of each of these models, it may be
difficult to modify the models appropriately for a given set of site conditions. Moreover, such
modifications often require knowledge of pharmacokinetics, programming, and/or the ability to use
pharmacokinetic modeling or dynamic simulation packages.
At the present time, EPA's ORD is developing a research-oriented biokinetic model, the All Ages Lead
Model, which may become a regulatory model in the future. The model evaluation described in this
report revealed several features of existing models that would be useful to consider in the development of
new, expanded models for adult lead risk assessment, including the All Ages Lead Model. These
features include the following:
-------
-REVIEW OF ADULT LEAD MODELS-
A flexible multimedia exposure module that includes variables to model different age groups,
genders, racial, and demographic groups.
Exposure and biokinetic models that will reliably simulate acute and intermittent exposures, and
the corresponding biokinetics, including PbB concentration dynamics as well as the quasi-steady
state.
An option to represent soil and dust lead ingestion and absorption as separate variables.
A variable(s) to represent enrichment of lead in dust.
Options for translating dust lead loading into dust lead concentrations and/or intakes.
A biokinetic module that predicts lead level and/or tissue concentrations for which biomonitoring
data are available, including blood, bone, teeth, plasma.
An option to simulate linear and nonlinear kinetics of gastrointestinal absorption and red blood
cell uptake of lead.
A complete maternal-fetal biokinetic component.
An option for a default or site-specific simulation of baseline PbB concentration, (e.g., the
baseline PbB concentration achieved prior to an adult exposure at a site).
1 A user-friendly, menu-driven interface.
Output that includes summary statistics (central tendency, percentiles), graphics capability, PRO,
and relative contribution of specific environmental sources and exposure pathways to the
predicted PbB concentratioa
Options for representing the inter-individual variability in PbB concentrations, including
empirically based GSDs and a Monte Carlo approach using PDFs for input variables.
A capability for automated sensitivity analysis that allows multiple variables to vary
simultaneously across defined ranges.
-------
70
-REVIEW OF ADULT LEAD MODELS
4.0 References
Barry, P.S.I. 1975. A comparison of concentrations of lead inhuman tissues. Br. J. Ind. Med. 32:119-
139.
Batschelet, E., L. Brand and A. Steiner. 1979. On the kinetics of lead in the human body. J. Math. Biol.
8: 15-23.
Bernard, S.R. 1977. Dosimetric data and metabolic model for lead. Health Phys. 32:44-46.
Bert, J.L., LJ. Van Dusen and J.R. Grace. 1989. A generalized model for the prediction of lead body
burdens. Environ. Res. 48:117-127.
Bowers, T.S., B.D. Beck and H.S. Karam. 1994. Assessing the relationship between environmental lead
concentrations and adult blood lead levels. Risk Analysis. 14(2): 183-189.
Brody, D.J., J.L. Pirkle, R.A Kramer, K.M. Flegal, T.D. Matte, E.W. Gunter and D.C. Paschal. 1994.
Blood lead levels in the U.S. population. Phase 1 of the third National Health and Nutrition Examination
Survey (NHANES HI, 1988 to 1991). JAMA. 272(4): 277-283.
Carlisle, J.C. and MJ. Wade. 1992. Predicting blood lead concentrations from environmental
concentrations. Regul. Toxicol. Pharmacol. 16: 280-289.
Casteel, S.W., L.D. Brown, M.E. Dunsmore, et al. 1996a. Bio availability of lead in soil samples from
the New Jersey Zinc NPL Site Palmerton, Pennsylvania. U.S. EPA, Region 8, Denver. .
Casteel, S.W., L.D. Brown, M.E. Dunsmore, et al. 1996b. Bioavailability of lead in slag and soil
samples from the Murray Smelter Superfund Site. U.S. EPA, Region 8, Denver.
Gibaldi, M. and D. Perrier. 1982. Pharmacokinetics. 2nd edition. Marcel Dekker, Inc.
Goyer, R.A. 1990. Transplacental transport of lead. Environ. Health Perspect. 89: 101-105.
Graziano, J.H., D. Popovac, P. Factor-Litvak, P. Shrout, J. Kline, M.J. Murphy, Y. Zhao, A. Mehmeti, X.
Ahmedi, B. Rajovic, Z. Zvicer, D. Nenezic, N. Lolacono and Z. Stein. 1990. Determinants of elevated
blood lead during pregnancy in a population surrounding a lead smelter in Kosovo, Yugoslavia. Environ.
Health Perspect. 89:95-100.
Griffin, T.B., F. Coulston and H. Wills. 1975. Biological and clinical effects of continuous exposure to
airborne particulate lead. Arh. Hig. Toksikol. 26(Suppl.): 191-208.
Griffin, T.B., F. Coulston, H. Wills and J.C. Russell. 1975. Clinical studies on men continuously
exposed to airborne particulate lead. Environ. Qual. Saf. Suppl. 2: 221-240.
Leggett, R.W. 1993. An age-specific kinetic model of lead metabolism in humans. Environ. Health
Perspect. 101: 598-616.
Maddaloni, M., N. Lolacono, W. Manton, C. Blum, J. Drexler and J. Graziano. 1998. Bioavailability of
soilborne lead in adults, by stable isotope dilution. Environ. Health Perspect. 106: 1-9.
-------
71
-REVIEW OF ADULT LEAD MODELS
Marcus, A.H. 1985. Multicompartment kinetic model for lead. I. Bone diffusion models for long-term
retention. Environ. Res. 36: 441-458.
Moore, M.R., P.A. Meredith, W.S. Watson, DJ. Sumner, M.K. Taylor and A. Goldberg. 1980. The
percutaneous absorption of lead-203 in humans from cosmetic preparations containing lead acetate, as
assessed by whole-body counting and other techniques. Food Chem. Toxicol. 18: 399-405.
O'Flaherty, EJ. 1991. Physiologically based models for bone-seeking elements. I. Rat skeletal and bone
growth. Toxicol. Appl. Pharmacol. Ill: 299-312.
O'Flaherty, EJ. 1991. Physiologically based models for bone-seeking elements. II. Kinetics of lead
disposition in rats. Toxicol. Appl. Pharmacol. Ill: 313-331.
O'Flaherty, EJ. 1991. Physiologically based models for bone-seeking elements. IE. Human skeletal and
bone growths. Toxicol. Appl. Pharmacol. 111:332-341.
O'Flaherty, E J. 1993. Physiologically based models for bone-seeking elements. IV. Kinetics of lead
disposition in humans. Toxicol. Appl. Pharmacol. 118:16-29.
O'Flaherty, E J. 1995. Physiologically based models for bone-seeking elements. V. Lead absorption and
disposition in childhood. Toxicol. Appl. Pharmacol. 131: 297-308.
O'Flaherty, EJ. 1997. User manual for lead model (draft). Attached to May 6 1997 letter from Ellen
O'Flaherty to Gary Diamond.
O'Flaherty, EJ. et al. 1998. Evaluation and modification of a physiologically based model of lead
kinetics using data from a sequential isotope study in cynomolgus monkeys. 149:1-16.
O'Flaherty, EJ., MJ. Inskip, A.P. Yagminas and CA. Franklin. 1996. Plasma and blood lead
concentrations, lead absorption, and lead excretion in nonhuman primates. Toxicol. Appl. Pharmacol.
138: 121-130.
Pirkle, J.L., R.B. Kaufmann, DJ. Brody, T. Hickman, E.W. Gunter and D.C. Paschal. 1998. Exposure
of the U.S. population to lead, 1991-1994. Environ. Health Perspect. 106: 745-750.
Pocock, S J., AG. Shaper, M. Walker, C J. Wale, B. Clayton, T. Delves, R.F. Lacey, R.F. Packham and
P. Powell. 1983. Effects of tap water lead, water hardness, alcohol, and cigarettes on blood lead
concentrations. J. Epidemiol. Commun. Health. 37: 1-7.
Rabinowitz, M.B., G.W. Wetherill and J.D. Koppel. 1976. Kinetic analysis of lead metabolism in health
humans. J. Clin. Invest. 58: 260-270.
Rabinowitz, M.B., J.D. Koppel and G.W. Wetherill. 1980. Effect of food intake on fasting
gastrointestinal lead absorption in humans. Am. J. Clia Nutr. 33: 1784-1788.
Sherlock, J.C., D. Ashby, H.T. Delves, G.I. Forbes, M.R Moore, WJ. Patterson, S J. Pocock, M J.
Quinn, W.N. Richards and T.S. Wilson. 1984. Reduction in exposure to lead from drinking water and
its effect on blood lead concentrations. Human Toxicol. 3: 383-392.
Stern, A.H.. 1994. Derivation of a target concentration of Pb in soil based on elevation of adult blood
pressure. Risk Analysis. 16:201-210.
-------
72
-REVIEW OF ADULT LEAD MODELS
U.S. EPA. 1986. Air Quality Criteria for Lead Volumes I - IV. Environmental Criteria and Assessment
Office, Office of Research and Development, RTF, NC. EPA 600/8-83-028 a-d.
U.S. EPA. 1993. Superfund's Standard Default Exposure Factors for the Central Tendency and RME-
Draft. Working Draft, November 1993.
U.S. EPA. 1994a. Revised Interim Soil Lead Guidance for CERCLA Sites and RCRA Corrective Action
Facilities. OSWER Directive No. 9355.4-12. Office of Emergency and Remedial Response,
Washington, D.C. EPA/540/F-94/043, PB94-963282.
U.S. EPA. 1994b. Technical Support Document: Parameters and Equations Used in the Integrated
Exposure Uptake Biokinetic Model for Lead in Children (v. 0.99d). Office of Emergency and Remedial
Response, Washington, D.C. EPA/540/R-94/040, PB94-963505.
U.S. EPA. 1994c. Guidance Manual for the Integrated Exposure Uptake Biokinetic Model for Lead in
Children. Office of Emergency and Remedial Response, Washington, D.C. EPA/540/R-93/081, PB93-
963510.
U.S. EPA. 1996. Recommendations of the Technical Review Workgroup for Lead for an Interim
Approach to Assessing Risks Associated with Adult Exposures to Lead in Soil. U.S. Environmental
Protection Agency Technical Review Workgroup for Lead.
------- |