UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
November 13, 2006
EPA-SAB-07-002
Honorable Stephen L. Johnson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, NW
Washington, DC 20460
Subject: EPA Science Advisory Board (SAB) Ad Hoc All-Ages Lead Model Review
Panel's Peer Review of the "All-Ages Lead Model (AALM) Version 1.05
(External Review Draft)"
Dear Administrator Johnson:
EPA's Office of Research and Development (ORD) has developed the All-Ages Lead
Model (AALM), which is designed to predict lead concentrations in body tissues and organs for
a hypothetical individual, based on a simulated lifetime of lead exposure. The precursor to the
AALM was the Integrated Exposure Uptake Biokinetic (IEUBK) Model for Lead in Children,
which underwent peer review by the EPA Science Advisory Board (SAB) in 1991. In response
to ORD's request, the SAB convened an ad hoc expert panel to conduct a peer review of the
model (Version 1.05) and the Guidance Manual on October 27-28, 2005 in Washington, D.C.
The SAB panel members were generally supportive of progress in developing the model.
However, in the judgment of this Panel, the current version of the model is not ready for
deployment due to a number of deficiencies, as detailed in the report.
Regarding features and operation of the model, the SAB Panel recommended that the
AALM be made more transparent and easier to understand by diverse users. The Panel noted
that the predictive accuracy of the model could be improved by incorporating new biokinetic
data that has been available since 1993. These data fall generally into three areas (that is,
absorption, skeletal turnover, and blood/plasma components). Both of the existing Leggett and
O'Flaherty models are incomplete and do not include current understanding in these areas. EPA
should sponsor experimental and computational research to improve the AALM parameterization
in these three areas. Panel members also suggested a more rigorous examination of all lead
models, including a summary of each model's advantages and limitations, as well as differences
in their conceptual structures, and use these as a basis for justifying the structure of the AALM.
Three different components of the model need to be addressed: dust exposure, gastrointestinal
absorption of lead, and soil exposure. Bioavailability, particularly with respect to soil, is not
-------
addressed in the model and should be one of its key parameters. Differences in bioavailability
among lead in soils of different origins and character, is likely to be a major factor in model
predictions. Real-world data should be used to evaluate the predictive accuracy and reliability of
the model (i.e., environmental lead values compared with blood urine and bone lead for children
and adults). Improvement is also needed in the predictive accuracy and reliability of the model.
The model needs to predict a distribution rather than predicting a single value. Furthermore, the
model needs to incorporate uncertainty more directly. In particular, a high degree of uncertainty
is introduced in the modeling effort by specifying so many parameters.
The user interface of the model was generally deemed to be quite good. The menu-
driven interface is intuitive and the learning curve is not steep; however, the SAB panel
suggested additional features. The guidance manual was useable but still in need of
improvement. The manual should provide both a theoretical framework to understand the
structure of the model and its scientific basis, and a step-by-step procedure from data input to
evaluation of the predicted outcomes. The parameters dictionary was judged to be an extremely
important component of both the guidance and the help feature. The Panel also identified
problems in quality control. The model often did not perform correctly, at times yielding strange
results, with coding errors suspected. Lastly, the SAB Panel recommended that the model be run
with the same datasets as the Leggett model, that plausibility checks also be run, and that other
human lead pharmacokinetic data sets be examined.
Detailed suggestions for improvement of the draft AALM are presented in the report,
organized by four sub-group areas: (1) conceptual construct of the model; (2) predictive accuracy
and reliability of the model; (3) computer coding and quality assurance; and (4) AALM
documentation (e.g., guidance manual, parameters dictionary, etc.)
In conclusion, the SAB encourages the Agency to continue its development of the
AALM. The SAB stands ready to offer additional advice and recommendations to assist EPA in
this effort, and wishes the Agency staff well in this important endeavor.
Sincerely,
/Signed/ /Signed/
Dr. Granger Morgan Dr. Meryl Karol
Chair Chair
EPA Science Advisory Board SAB Ad Hoc AALM Review Panel
-------
NOTICE
This report has been written as part of the activities of the U.S. Environmental
Protection Agency's (EPA) Science Advisory Board (SAB), a Federal advisory
committee administratively located under the EPA Science Advisory Board (SAB) Staff
Office that is chartered to provide extramural scientific information and advice to the
Administrator and other officials of the EPA. The SAB is structured to provide balanced,
expert assessment of scientific matters related to issue and problems facing the Agency.
This report has not been reviewed for approval by the Agency and, hence, the contents of
this report do not necessarily represent the views and policies of the EPA, nor of other
agencies in the Executive Branch of the Federal government, nor does mention of trade
names or commercial products constitute a recommendation for use. SAB reports are
posted on the SAB Web site at: http://www.epa.gov/sab.
-------
U.S. Environmental Protection Agency
Science Advisory Board (SAB) Staff Office
SAB Ad Hoc All-Ages Lead Model Review Panel
CHAIR
Dr. Meryl Karol*, Associate Dean for Academic Affairs, University of Pittsburgh, Pittsburgh,
PA
PANEL MEMBERS
Dr. Mary Jean Brown, Chief, Lead Poisoning Prevention Branch, U.S. Centers for Disease
Control and Prevention (CDC), Atlanta, GA
Dr. Deborah Cory-Slechta*, Director, University of Medicine and Dentistry of New Jersey and
Rutgers State University, Piscataway, NJ
Dr. Bruce Fowler, Assistant Director for Science, Division of Toxicology and Environmental
Medicine, Office of the Director, Agency for Toxic Substances and Disease Registry, U.S.
Centers for Disease Control and Prevention (ATSDR/CDC), Chamblee, GA
Dr. Philip Goodrum, Senior Scientist/Manager, Blasland, Bouck & Lee, Inc., Syracuse, NY
Dr. Roberto Gwiazda, Assistant Researcher, Environmental Toxicology, University of
California - Santa Cruz, Santa Cruz, CA
Mr. Sean Hays, President, Summit Toxicology, Allenspark, CO
Dr. Marlin Mickle, Professor, Electrical and Computer Engineering, University of Pittsburgh,
Pittsburgh, PA
Dr. Paul Mushak, Principal, PB Associates, and Visiting Professor, Albert Einstein College of
Medicine (New York, NY), Durham, NC
Dr. Joel Pounds, Scientist, Cell Biology & Biochemistry, Biological Sciences Division, Battelle
- Pacific Northwest National Laboratory (PNNL), Richland, WA
Dr. Michael Rabinowitz, Geochemist, Marine Biological Laboratory, Woods Hole, MA
Dr. Joel Schwartz, Professor, Environmental Health, Harvard University School of Public
Health, Boston, MA
Dr. Alan Stern, Section Chief-Risk Assessment/Adjunct Associate Professor, Division of
Science, Research & Technology/Dept. of Environmental & Occupational Health, New Jersey
Dept. of Environmental Protection (NJDEP)/University of Medicine & Dentistry of NJ-School of
Public Health, Trenton, NJ
11
-------
Dr. Ian von Lindern, Senior Scientist, TerraGraphics Environmental Engineering, Inc.,
Moscow, ID
SCIENCE ADVISORY BOARD STAFF
Mr. Fred Butterfield, CASAC Designated Federal Officer, 1200 Pennsylvania Avenue, N.W.,
Washington, DC, 20460, Phone: 202-343-9994, Fax: 202-233-0643 (butterfield.fred@epa.gov)
(Physical/Courier/FedEx Address: Fred A. Butterfield, III, EPA Science Advisory Board Staff
Office (Mail Code 1400F), Woodies Building, 1025 F Street, N.W., Room 3604, Washington,
DC 20004, Telephone: 202-343-9994)
* Members of the statutory EPA Science Advisory Board (SAB) appointed by the EPA Administrator
in
-------
U.S. Environmental Protection Agency
Science Advisory Board (SAB)
CHAIR
Dr. M. Granger Morgan, Professor and Head, Department of Engineering and Public Policy,
Carnegie Mellon University, Pittsburgh, PA
VICE CHAIR
SAB MEMBERS
Dr. Gregory Biddinger, Environmental Programs Coordinator, ExxonMobil Biomedical
Sciences, Inc, Houston, TX
Dr. James Bus, Director of External Technology, Toxicology and Environmental Research and
Consulting, The Dow Chemical Company, Midland, MI
Dr. Trudy Ann Cameron, Raymond F. Mikesell Professor of Environmental and Resource
Economics, Department of Economics, University of Oregon, Eugene, OR
Also Member: COUNCIL
Dr. Deborah Cory-Slechta, Director, Environmental and Occupational Health Sciences
Institute, Robert Wood Johnson Medical School, University of Medicine and Dentistry of New
Jersey and Rutgers State University, Piscataway, NJ
Dr. Maureen L. Cropper, Professor, Department of Economics, University of Maryland,
College Park, MD
Dr. Virginia Dale, Corporate Fellow, Environmental Sciences Division, Oak Ridge National
Laboratory, Oak Ridge, TN
Dr. Kenneth Dickson, Professor, Institute of Applied Sciences, University of North Texas,
Denton, TX
Dr. Baruch Fischhoff, Howard Heinz University Professor, Department of Social and Decision
Sciences, Department of Engineering and Public Policy, Carnegie Mellon University, Pittsburgh,
PA
Dr. A. Myrick Freeman, William D. Shipman Professor of Economics Emeritus, Department of
Economics, Bowdoin College, Brunswick, ME
Dr. James Galloway, Professor of Environmental Sciences, Environmental Sciences
Department, University of Virginia, Charlottesville, VA
Dr. Lawrence Goulder, Shuzo Nishihara Professor of Environmental and Resource Economics,
Department of Economics, Stanford University, Stanford, CA
IV
-------
Dr. Rogene Henderson, Scientist Emeritus, Lovelace Respiratory Research Institute,
Albuquerque, NM
Also Member: CASAC
Dr. Philip Hopke, Bayard D. Clarkson Distinguished Professor, Department of Chemical
Engineering, Clarkson University, Potsdam, NY
Dr. James H. Johnson, Dean, College of Engineering, Architecture & Computer Sciences,
Howard University, Washington, DC
Dr. Meryl Karol, Associate Dean for Academic Affairs, Graduate School of Public Health,
University of Pittsburgh, Pittsburgh, PA
Dr. Catherine Kling, Professor, Department of Economics, Iowa State University, Ames, IA
Dr. George Lambert, Associate Professor and Director, Center for Child and Reproductive
Environmental Health & Pediatric Clinical Research Center, Department of Pediatrics, UMDNJ-
Robert Wood Johnson Medical School/ University of Medicine and Dentistry of New Jersey,
New Brunswick, NJ
Dr. Jill Lipoti, Director, Division of Environmental Safety and Health, New Jersey Department
of Environmental Protection, Trenton, NJ
Dr. Genevieve Matanoski, Professor, Department of Epidemiology, Johns Hopkins University,
Baltimore, MD
Dr. Michael J. McFarland, Associate Professor, Department of Civil and Environmental
Engineering, Utah State University, Logan, UT
Dr. Jana Milford, Associate Professor, Department of Mechanical Engineering, University of
Colorado, Boulder, CO
Dr. Rebecca Parkin, Professor and Associate Dean, Environmental and Occupational Health,
School of Public Health and Health Services, The George Washington University, Washington,
DC
Mr. David Rejeski, Foresight and Governance Project Director, Woodrow Wilson International
Center for Scholars, Washington, DC
Dr. Joan B. Rose, Professor and Homer Nowlin Chair for Water Research, Department of
Fisheries and Wildlife, Michigan State University, E. Lansing, MI
Dr. Kathleen Segerson, Professor, Department of Economics, University of Connecticut, Storrs,
CT
-------
Dr. Kristin Shrader-Frechette, O'Neil Professor of Philosophy- Concurrent Professor of
Biological Sciences-and Director of the Center for Environmental Justice and Children's Health,
Department of Biological Sciences and Philosophy Department, University of Notre Dame,
Notre Dame, IN
Dr. Robert Stavins, Albert Pratt Professor of Business and Government, Environment and
Natural Resources Program, John F. Kennedy School of Government, Harvard University,
Cambridge, MA
Dr. Deborah Swackhamer, Professor, Division of Environmental Health Sciences, School of
Public Health, University of Minnesota, Minneapolis, MN
Dr. Thomas L. Theis, Professor and Director, Institute for Environmental Science and Policy,
University of Illinois at Chicago, Chicago, IL
Dr. Valerie Thomas, Anderson Interface Associate Professor of Natural Systems, School of
Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, GA
Dr. Barton H. (Buzz) Thompson, Jr., Robert E. Paradise Professor of Natural Resources Law,
Stanford Law School, and Director, Woods Institute for the Environment, Stanford University,
Stanford, CA
Dr. Robert Twiss, Professor, University of California-Berkeley, Ross, CA
Dr. Terry F. Young, Consultant, Environmental Defense, Oakland, CA
Dr. Lauren Zeise, Chief, Reproductive and Cancer Hazard Assessment Section, California
Environmental Protection Agency, Oakland, CA
SCIENCE ADVISORY BOARD STAFF
Mr. Thomas Miller, Designated Federal Officer, 1200 Pennsylvania Avenue, NW
1400F, Washington, DC, Phone: 202-343-9982, Fax: 202-233-0643, (miller.tom@epa.gov)
VI
-------
U.S. Environmental Protection Agency
Science Advisory Board
PEER REVIEW
of the
ALL-AGES LEAD MODEL (AALM)
Version 1.05 (External Review Draft)
by the
SAB Ad Hoc AALM Review Panel
October xx, 2006
-------
Table of Contents
Executive Summary 3
Background and Introduction 9
Compiled Responses to Agency Charge Questions from Panel Sub-Groups 9
I. Conceptual Construct of the Model 9
II. Predictive Accuracy and Reliability of the Model 17
III. Computer Coding and Quality Assurance 35
IV. AALM Documentation 37
Appendix A - Charge to the SAB Ad Hoc All-Ages Lead Model Review Panel A-1
Appendix B-References B-l
-------
Executive Summary
The AALM Panel strongly supports the Agency's development of the model, but offers
extensive suggestions for its improvement. Importantly, however, the current version of the
AALM does not model a population response and therefore does not meet the goals of EPA.
Accordingly, in the judgment of this Panel, the AALM is not ready for use. The Panel offers the
following comments and suggestions for improvement of the model.
I. Conceptual Construct of the Model
The model should be more transparent and easier to understand by diverse users. The
AALM Guidance Manual should define each of the various lead pharmacokinetic models,
including their advantages and limitations, and differences in their conceptual structures. This
review, with appropriate literature references, should be used to help justify the structure of the
model. Another general recommendation is the need for standard units of measure to be used
throughout the model and reported in all outputs, including graphics; this feature is currently
missing and makes it difficult to understand the outputs. The AALM Panel further recommends
that the descriptions of the biokinetic parameters in the AALM be made more consistent with the
Leggett model's descriptors. Bioavailability, particularly with respect to soil, is not addressed in
the model and should be one of its key parameters. In addition, pica should not be used in the
model as a surrogate for soil ingestion.
In terms of model performance, it is critical to compare its "outcomes" (model results)
with the empirical predictions from existing, high-quality databases that relate measured lead
concentrations in environmental media to blood lead and bone lead concentrations in exposed
populations, i.e., the AALM should be shown to be capable of providing as accurate a reflection
as possible of the empirical outcomes from these databases. For purposes of protecting the most
highly exposed, it is also vital that the model yield predictions of blood lead and bone lead that
can be compared with both the mean and the upper percentiles of the distribution of measured
concentrations. The model should allow users to incorporate information on variability in
exposure, uptake, and biokinetics. In addition, it would be useful to allow users to separately
characterize variability and parameter uncertainty in order to compare confidence intervals in
both the model output and the empirical measurements.
The model should be modified to assure that all biokinetic parameters remain internally
consistent, since a change in one biokinetic parameter without corresponding adjustments in all
the transfer and tissue/organ deposition streams will affect the reliability of the outputs.
Likewise, exposure parameters are also linked and should be synchronized. Caution should be
provided to users about the consequences of implementing parametric changes. Second, EPA
should consider allowing adjustments to the intake and uptake parameters, since such
adjustments are required for site-specific or circumstance-specific lead scenarios of direct
interest to users, but it should restrict alterations in the biokinetic parameters, as with the IEUBK
model. If the Agency does allow variation in biokinetic parameters, the user should be warned if
a particular parameter value was estimated by calibrating the model to match empirical data on
blood lead concentrations or other variables. In these cases, changing the parameter values may
result in model predictions that are no longer supported by previous calibration exercises. Third,
-------
the overall complexity of the model gives it aspects of a "black box." This might be partly
addressed by allowing users to evaluate intermediate outputs from individual modules. Fourth,
checks for mass balance errors need to be included in the model. If the AALM uses fractions of
rate constants, how is the sum of fractions maintained to have a sum of 1? A warning should be
provided the user when a mass balance has not been achieved.
Other specific recommended changes include: (a) the need to differentiate and explicitly
address three different components: dust exposure, gastrointestinal absorption of lead, and soil
exposure; (b) implementation of more realistic age range breakouts for the youngest age bands.
The peak in oral exploratory activity and hand-mouth activity occurs at 12 to 30 months of age.
The current age interval for toddlers too broad and unlikely to capture the heightened exposures
of 12 to 30 months of age toddlers; (c) breast milk exposures should be accounted for; (d) age-
specific intake exposure factors (e.g., breathing rate, drinking water intake rate, etc.) need to be
consistent with EPA's "Child-Specific Exposure Factors Handbook" (EPA-600-P-00-002B;
2002); and (e) bioavailability and bio-accessibility differences should be developed outside of
the exposure module in a manner consistent with how this lead will be treated in the Absorption
module.
II. Predictive Accuracy and Reliability of the Model
Several suitable data sets were identified that could be used to examine the models
predictive veracity, and to calibrate the model. These existing data sets include environmental
lead values paired with blood, urine and/or bone lead values for children and adults. Suitable
data sets for validating the model include the National Health and Nutrition Examination Survey
(NHANES) data set and the Lanphear compilation of multiple studies.
Regardless of the actual values predicted by the model, several issues of internal
consistency were noted. For example, blood lead values changed abruptly with age, which was
troublesome. These results appeared to be sensitive to the step size selected. Also, the
integration algorithms need to be verified.
The model should predict a distribution of blood lead values. For the AALM to be used
to characterize variability and uncertainty in blood lead and other output variables, a probabilistic
approach is needed. The AALM Panel generally recommended the use of Monte Carlo analysis
in which exposure and/or biokinetic parameters are characterized by probability distributions.
Both variability and uncertainty in an output variable can be characterized depending on the
choice of input distributions and the choice of Monte Carlo simulation methods. A second
probabilistic approach may also be desirable as an alternative to Monte Carlo analysis, and to
facilitate the transition from the IEUBK model to the AALM. Currently, users of IEUBK are
familiar with the use of a lognormal distribution assumption applied directly to the output
variable (i.e., the geometric standard deviation [GSD] term). While the process for selecting
site-specific GSDs has been a source of considerable debate among the risk assessment
community, it is a simpler method of characterizing distributions and can be informed by
empirical data. One shortcoming of this approach is that it does not allow for quantitative
uncertainty analysis, since plausible bounds or confidence intervals on model predictions cannot
be determined.
-------
The predictive accuracy of the model could be improved by including newer information
about absorption and internal distribution of lead, RBC-plasma partitioning, and air-dust
relationships. The default values should be reexamined, and the ability to change selected
biokinetic parameters should be added. In addition, a high degree of uncertainty is introduced in
the modeling effort by specifying so many biokinetic model parameters for which there is limited
information about their values.
III. Computer Coding and Quality Assurance
The user interface of the AALM is quite good. The menu-driven interface is intuitive, and
the learning curve is not very steep. However, there are additional features that would enable the
model to be more useful for either hypothetical or real-world risk assessment problems.
Limitations of the AALM include the following:
A batch mode is needed similar to the current functionality of the IEUBK model to
facilitate an evaluation of the proportion of the population that exceeds a target risk-based
concentration in an exposure medium. The user should be able to specify an input file
with a set of site-specific factors (e.g., paired concentrations in soil, dust, and water at a
residence). The AALM would benefit greatly by allowing either point estimates or
probability distributions to be calculated for each exposure unit.
The AALM needs to incorporate variability and uncertainty more directly. It would be
useful to be able to specify expected distributions of parameters, and get out a probability
distribution of blood lead for a population. Default distributions, rather than default point
estimates, for these parameters would be preferred so that variability and uncertainty are
more properly accounted for in the risk assessment without the requirement of tedious
repetitions by the risk assessor.
Even for the research community, caution should be given about changing the biokinetic
parameters, since they were not derived independently, and changing one often implies
changes to others. The results would also no longer correspond to the Leggett model.
Interest was expressed in adding the physiological (O'Flaherty) model option.
Specific suggestions for the AALM include: eliminating the need to set the gender option in
three locations; and increasing flexibility in the graphic display.
Regarding the Quality Assurance/Quality Control (QA/QC) concerns of the program, the
AALM does not perform correctly. For example, Manton and Cook's data indicate that plasma
lead should be about 0.2% of blood lead when blood lead is less than 25 mg/dL. The Leggett
model, on which the AALM is based, predicts this successfully. However, the AALM not only
does not meet this design criteria, it produces non-single-valued functions. Small errors in the
parameterization of the kinetics of this compartment can propagate very rapidly to errors in the
amounts of lead in all other compartments. Since the AALM derives from the Leggett model, it
is assumed that this is a result of coding errors.
In addition, plausibility checks should be run such as making sure results behave as
expected as the number of years is lengthened, that different intakes for different periods behave
-------
as one would expect, given cumulative dose, etc. It would also be worthwhile to examine other
data sets. Finally, several of the input assumptions are unreasonable, and should be changed.
This includes assuming the same gut absorption rate for food and water lead, the default values
for water lead concentration, etc.
IV. AALM Documentation
The AALM Panel deems that the present Guidance Manual is useable, but should be
made more user-friendly. It is incomplete in several areas and contains many errors and
confusing wording. The manual should provide both a theoretical framework for the uninitiated
user to understand the structure of the model and its scientific basis, and a step-by-step procedure
that would walk the user from the data input to the evaluation of the predicted outcomes. EPA
should also consider developing and releasing a companion Technical Support Document to
augment the Guidance Manual that includes verification and validation exercises, utilizing real
world demonstrations, and appropriate cautions that would aid the user in understanding,
interpreting and utilizing the model.
In addition, the AALM Panel notes that the output options provided in Version 1.05 are
interim choices and will therefore need to be developed more fully in subsequent releases. The
data files, if possible, should be exportable to other traditional software programs. More
explanation of the structure, underlying nature and accessibility of these data sets should be
provided. There should also be explanation regarding the type of environmental information to
be entered, so that it is standardized to the type for which the model was calibrated. The
Parameters Dictionary was an extremely important component of both the guidance and the help
feature. The AALM Panel suggests that more specific information be provided in the guidance
and support documents with regard to each individual parameter, including its origin, source of
support data, possible range of values, any information regarding central tendencies and
variance, uncertainty, the rationale for the default setting, relationship to other parameters, and
appropriate cautions as needed for modifications.
To the extent practicable, the AALM approach, guidance and application should be
consistent, and evolve concurrently with similar models and guidance presently endorsed by
EPA. The Agency should consider issuing guidance regarding the required (or recommended)
use of the default or prescribed bio-kinetic parameters in regulatory applications. Finally, the
AALM should be evaluated relative to the Agency's current Draft Guidance on the
Development, Evaluation, and Application of Regulatory Environmental Models.
Background and Introduction
The EPA Science Advisory Board was established by 42 U.S.C. § 4365 to provide
independent scientific and technical advice, consultation, and recommendations to the EPA
Administrator on the technical basis for Agency positions and regulations. The SAB is a Federal
advisory committee chartered under the Federal Advisory Committee Act (FACA), as amended,
5 U.S.C., App. The AALM Panel consists of 14 members, two of whom are also members of the
chartered SAB appointed by the EPA Administrator. The AALM Panel provides its advice
-------
through the SAB, and complies with the provisions of FACA and all appropriate SAB Staff
Office procedural policies.
EPA's Office of Research and Development (ORD), National Center for Environmental
Assessment (NCEA), requested that the SAB Staff Office form the AALM Panel to provide
advice and recommendations to the Agency on EPA's recently-developed AALM. The AALM
is designed to predict lead concentrations in body tissues and organs for a hypothetical
individual, based on a simulated lifetime of lead exposure. Statistical methods can be used to
extrapolate to a population of similarly-exposed individuals. The precursor to the AALM was
the Integrated Exposure Uptake Biokinetic (IEUBK) Model for Lead in Children. The IEUBK
Model underwent peer review by the SAB in 1991 and was subsequently revised in response to
that review, leading to release of Version 0.99d of the IEUBK Model in March 1994. Since
then, the IEUBK Model has been widely accepted and used in the risk assessment community as
a tool for implementing the site-specific risk assessment process when the issue is childhood lead
exposure. Based on further refinement of the IEUBK Model and its expansion for use with
additional age groups beyond pediatric populations six years old or younger, the AALM has
recently been developed to cover older childhood and adult lead exposure. The anticipated
outcome of this model is reduced uncertainty in lead exposure assessments for children and
adults.
Compiled Responses to Agency Charge Questions from Panel Sub-Groups
I. Conceptual Construct of the Model
* Charge Questions
(1) In general, to what extent are the parameters and relationships represented by various
AALM features adequately supported by available research findings in published peer-
reviewed literature or by reasonable extrapolations from such findings? That is, are the
specifications of key components of the AALM model scientifically supportable in
characterizing particular parameters or relationships of the types noted above?
Standard units of measure should be used throughout the model and should be reported in
all outputs, including graphics. This is critical for both comprehension of AALM outputs and for
ease of application in the regulatory world. For example, the output levels in Window Figure 8
are total lead content values that require different metrics for routine use, e.g., |ig/dL for whole
blood Pb, and ppm for wet weight of soft tissues. Use of routine measurements and
specifications would eliminate much confusion.
All the assumptions utilized in the model, both on the computational and the biological
side, should be specified and made as transparent as possible. Users of the model may not be
using a common vocabulary and nomenclature, nor have the requisite background to
comprehend discipline-specific issues in the same way.
-------
The committee recommends that the Guidance Manual define all of the models more
lucidly, including delineation of the differences in the conceptual structures of the models.
Users of the models who are not toxicokineticists or pharmacokineticists need to understand the
distinctions among the models, including their advantages and limitations. Currently, there are
abstract definitions in the literature of what comprises a "PB-PK" model but no universally
accepted features of existing models for simulation of human lead exposures that define them
and functionally distinguish one model from another. Indeed, there seems to be actual
disagreement regarding labeling among those who have introduced various models. O'Flaherty
(1998) states (p. 1501, col. 1) that "The IEUBK [Integrated Exposure Uptake Biokinetic] model
developed by the U.S. EPA is not physiologically based in the sense in which either the
Leggett/ICRP model or the O'Flaherty model is..." Pounds and Leggett (1998) state (p. 1507,
col. 1) that "The IEUBK model is the most commonly used physiologically-based
pharmacokinetic (PBPK) model for lead in children." The Draft AALM Manual states (p. 40,
Bottom) that: "The Leggett method is generally considered to be anatomically-based...The
O'Flaherty method is physiologically based..." NCEA/EPA's 10/20/05 draft models
background document describes the IEUBK model (p. 10, last par.) as a "multicompartmental
pharmacokinetics" model. A more in-depth presentation of these models' limitations is critical
to understanding the rationale for the current model.
One issue that requires additional consideration is the need to modify the model to assure
that all biokinetic parameters remain internally consistent. That is, a change in one biokinetic
parameter without corresponding adjustments in others will affect the reliability of the outputs.
The EPA AALM designers should highlight the consequences of such scenarios for the
edification of, e.g.., risk assessors, using concrete examples. One possible consideration is, when
programming, to link the parts of the computational stream in the model that are affected by
isolated parameter changes. In this way, if a non-modeler arbitrarily makes changes in isolation,
appropriate changes are automatically made in other parameters to produce mass balance.
Alternatively, such changes could trigger a dialogue about the consequences and a directive with
respect to other parameters that would have to be changed. This would address the situation in
which a change in one parameter would potentially require changes in numerous other biokinetic
parameter combinations, not simply one or several.
Some Panel members suggested that EPA should consider the desirability of allowing
adjustments to the intake and uptake parameters, but restricting alterations in the biokinetic
parameters. This is the situation with the IEUBK model in which the intake and uptake
parameters can be edited, but few of the biokinetic parameters are accessible. It is difficult to
foresee circumstances where typical users would be more interested in changes in the biokinetic
parameters than lead intake and uptake. The latter are much more driven by site-specific or
circumstance-specific lead scenarios of direct interest to users than are parameters in the
Biokinetic module. The logic for making the IEUBK Biokinetic module inaccessible to users
applies as well to users of the AALM. That logic (stated on Page 4-58 of the 1994 IEUBK
guidance manual) is:
"The IEUBK model has a very detailed biokinetic modeling component. This
component of the model is not accessible to the user because, in our judgment, most users
will neither wish to change the biokinetic parameters nor have the need to change any of
-------
the biokinetic parameters. The biokinetic parameters are used to define intrinsic
biological variables that do not change from one exposure scenario to another once a
child's age is specified. "
Similar to the above, once an age interval is specified in the AALM, the biokinetic parameters
should not require changes by the user.
There is significant complexity in the individual modules and even greater complexity
when the interaction of the modules is considered. This complexity makes it very difficult to
evaluate the specification of the parameters within and across the modules in the abstract. This
might be partly addressed by allowing users to evaluate intermediate outputs from individual
modules, although such an option will not address the inter-module complexity.
The model fails to specify parameters relating to exposure and uptake of Pb in soil.
Although the model addresses gastrointestinal absorption of "dust", and "pica" ingestion, it does
not explicitly address soil. This is more than a semantic problem. Clearly, there is exposure via
ingestion of soil by both children and, to a lesser extent, adults. This exposure pathway is one of
the major drivers for hazardous site cleanup decisions. "Dust" is generally considered to be large
diameter indoor particulates, but indoor dust contains both soil-derived particles as well as
particles derived exclusively from indoor activities. Some members of the Panel also suggested
that it appears that the model envisions dust to include the top, easily accessed layer of soil, but
this does not necessarily correspond to the way soil is accessed particularly by children, and
by some adults, including gardeners. Additionally, it also appears that the model envisions
"pica" to be the intentional ingestion of bolus quantities of soil. However, pica is more properly
viewed as the persistent eating of non-nutritive substances for at least one month in a manner
inappropriate to the developmental level and without cultural sanction. (Citation: Diagnostic and
Statistical Manual of Mental Disorders, 4th Ed., Text Revision. Washington DC: American
Psychiatric Association, 2000). While there are, indeed, children who ingest bolus quantities of
soil, in general soil ingestion occurs along a continuum, with some children occasionally
consuming bolus doses, and others continually ingesting small quantities of material on their
hands over the course of the day. "Pica" should not be used in the model as a surrogate for soil
ingestion. Some members of the Panel further suggested that it is necessary that the model
consider exposure to soil as a separate, clearly defined component including that part of indoor
dust that is soil-derived. Without such an approach the model cannot be used to define Pb-
contaminated site cleanup goals. Finally, some AALM Panel members also suggested that
bioavailability, particularly with respect to soil, is not addressed in the model and should be one
of its key parameters. Differences in bioavailability among Pb in soils of different origins and
character, is likely to be a major factor in model predictions.
Historical exposures were not addressed in any quantitative way by the Panel. This
leaves unresolved how well one can evaluate or calibrate the AALM output. Arguably, in model
testing, users of the model would be confined to those data sets where one or the other of the
pairs of data may have measurement problems peculiar to this type of site testing. For example,
PbB testing at such sites is typically done only once. Mining, milling and smelting sites have
produced such pairs of measurement data but the statistical handling of the measurements can be
problematic.
-------
Simulations in which EPA is dealing with a contaminated site that has a likelihood of
producing lifelong exposures for affected communities, starting with current newborns, entails a
number of assumptions about the relative stability of the environmental Pb levels that serve as
exposure inputs. For example, the galena form of lead ore in milling wastes weathers to more
bioavailable lead carbonate (cerussite). Bioavailabilities estimated for current chemical species
of lead might well be underestimates for future decades.
Another important issue regarding specification of parameters is the use of point
estimates for default parameter values in the AALM. Given the variability in exposure and
biokinetics in populations, it is critical that predictions of lead exposure be capable of addressing
not only the mean exposure but also the upper percentiles of the exposure distribution. This is
important since highly exposed individuals may differ from the mean by several standard
deviations. Many of the parameters are too poorly characterized to be adequately described as
distributions. Clarification is needed on a case-by-case basis, particularly since full distributional
descriptions are not necessarily required to allow a reasonable estimate of the distribution of the
outcome parameter (i.e.., blood or bone lead). Estimated or screening distributions such as
triangular distributions can often be derived from limited data, and can provide adequate input to
an overall distributional analysis. The ability to fully describe all parameters with distributions
notwithstanding, useful information can be generated even by limiting the distributional
descriptions to a single module in the AALM. In particular, exposure parameters are generally
well characterized, and many, if not most, of the relevant parameters have been described by
distributions in the published literature.
Once obvious errors and deficiencies in the model have been identified and addressed
(see below), the appropriate question about the specification of the model and its components at
this point should not relate to a parameter-by-parameter assessment of the science underlying the
specific values and model structures. Rather, the performance of the model should be compared
with the empirical predictions from those existing good quality databases that relate measured
lead concentrations in environmental media to blood lead and bone lead concentrations in
exposed populations. As is clear from existing databases, there are differences among
individuals in biokinetics of lead. There will also be errors of measurement in characterization
of exposures, both human and environmental. Nevertheless, these are the data that often serve as
the basis for decisions regarding public health policies and interventions. Therefore, the model
should be shown to be capable of providing a reasonable reflection of the empirical outcomes
from these databases. If significant differences are found, comparisons can be made on a more
detailed level, including sensitivity analyses.
More specifically, what are the AALM Review Panel's views with regard to:
(a) The adequacy of the values specified for the exposure parameters for different media and
how well the model interprets exposure throughout the various age groups;
A detailed review of these parameters should be undertaken as a specific and focused
effort on a module-by-module basis. There was some concern that the AALM does not appear to
have taken full advantage of the extensive development of exposure parameters in the IEUBK
10
-------
model exposure modules. This is true with respect to age breakdowns, ingestion rates, route-
specific bioavailability and bio-accessibility and historical exposure default values. These
parameters have proven to be highly useful in regulatory risk assessment and risk management
activities and should not be "left behind" in advancing lead health modeling efforts by the
Agency.
More realistic age range breakouts should be considered for the youngest age bands. The
current 6 to 48 months age interval for toddlers is too broad and unlikely to capture the
heightened exposures of 12-24 or 30 months of age toddlers due to increased hand-mouth
activity. This adjustment will require revision within the exposure module, specifically the
parameters seen in Window Figure 21. Figure 21 shows a daily dust intake of 85 mg/d for
toddlers (42 months age interval total) and 135 mg/d for preschoolers (24 months interval). The
IEUBK model exposure module more correctly shows dust/soil intakes of 135 mg/d in children
12 - 48 months of age.
It is not clear where the values in Figure 21 are from. The Leggett (1993) paper says
nothing about this set of parameters. For example, are they 40% lower for these infants and
toddlers than those in the IEUBK model because of removal of the discrete soil component? If
so, this is all the more reason not to put soil in a subsidiary role under "pica" but to place it
within the main media tabs. Also, where does the value 135 mg/d for preschoolers, age 49 - 72
months, come from? It is not plausible that the older a child, the more dust he/she will ingest.
Also, the use of the 135 mg value suggests that an altered role of soil in these intake amounts is
not the reason for the change from the IEUBK.
EPA/NCEA's current "Child-Specific Exposure Factors Handbook" (EPA-600-P-00-
002B; 2002) reviewed the totality of the soil ingestion literature and concluded (Chapter 5, p. 5-
21, Table 5-19) that the best estimate of the soil ingestion portion of the soil + dust pair is 100
mg/day. Addition of interior dust to that figure makes 135 mg/d as it appears in the IEUBK
exposure module. This is much more plausible for infants and toddlers than 85 mg/d.
Restoration of the two lead-containing media as in the IEUBK model, while
simultaneously refining the age band for the youngest childhood subsets, will have the net effect
of making the PbB outputs consistent with a childhood lead exposure literature showing a broad
peak in PbB around 24 months, see, e.g., Figure 9, O'Flaherty, 1998; Clark et al., 1985; Billick et
al., 1979.
Additional specific recommended changes include the following:
The uptake value for Pb in drinking water should be close to 100% as the Pb is already in
a soluble form.
Breast milk exposures should be accounted for.
The default media lead concentrations need to be more scientifically justified.
Age-specific intake exposure factors (e.g., breathing rate, drinking water intake rate, etc.)
need to be consistent with existing EPA exposure factor guidance.
11
-------
(b) The adequacy of the uptake/absorption parameters or any need for modification of the
methodology for determining absorption for various routes of exposure;
Several considerations, some related to those raised in (a) above should be considered:
Bioavailability should be a key component of the uptake module.
The absorption of Pb in drinking water should be reconsidered.
There is concern that the AALM has combined soil and dust under a single route that
seems to have a fixed absorption rate in the gut identical to that for water and food. The
absorption rate requires that bioavailability and bio-accessibility differences be developed
outside of the exposure module in a manner cognizant of exactly how this lead will be
treated in the absorption module.
The fraction absorbed as a function of age needs to be better validated and made
consistent with the scientific literature.
Scientific justification of transdermal absorption of lead is needed before this parameter
is made part of the AALM.
(c) Whether there are any errors in AALM methods for determining biokinetic distribution or
errors in assigning values to biokinetic parameters;
Little information is available to answer this question without relying on the original
Leggett publication. The biokinetic parameters of the AALM are not all consistent with the
parameters used by Leggett. EPA appears to have used a different approach by employing
fractional rate constants (deposition fraction) that scale off the overall rate of elimination from
the diffusible plasma compartment. While this is the approach that Leggett used, Leggett only
reported the individual rate constants for transfer between compartments (total transfer rate times
the deposition fraction). Although a semantic issue, the user may refer to Leggett and see that
the AALM has used a different approach. In addition, the "Transfer Rate from RBC" entry in
the AALM corresponds with the parameter that Leggett calls "red blood cells (RBCs) to Plasma-
D". This is a subtle, but important difference. EPA's entry seems to imply this is the overall
elimination from RBCs to all compartments. Leggett's is more specific. The Panel recommends
that the descriptions of the biokinetic parameters in the AALM be more consistent with Leggett's
descriptors and that differences between the two approaches be made more explicit. EPA should
provide specific information as to whether Leggett's original values have been adopted or
modified, why, and whether and how additional information accumulated over the past twelve
years influenced that determination.
Table 1, p. 40, and Figure 42, p. 41, appear to have the wrong "RBC Threshold
Concentration." The figure of 60 |ig/dl is more correct for the value of whole blood Pb, PbB,
above which the equilibrium ratio of plasma/serum Pb to PbB becomes curvilinear upward
(Bergdahl et al., 1997; Manton and Cook, 1984; de Silva, 1981; Marcus, 1985; discussions in
U.S. EPA's lead criteria document, 1986, Ch. 10; and NAS/NRC, 1993, p. 159). That is, the
level of lead in plasma plotted against PbB remains stable (linear) until ca. 40-60 |ig/dl PbB,
when the relative plasma level increases. A PbB value of 50 |ig/dl corresponds to an
approximate Pb-RBC value of 125 |ig/dl (PbB/0.4 = Pb-RBC); for 60 |ig/dl, the erythrocyte Pb
12
-------
level is 150 |ig/dl. This is based on a hematocrit of 40%. The relationship is depicted in Figures
10-2 and 10-5 in Ch. 10 of the EPA 1986 lead criteria document. This curvilinear relationship
with increasing PbB has been suggested as one biokinetic explanation for the Chamberlain
(1983) observation that the relative lead excretion rate in adults increases with increasing PbB.
The Leggett model is a multicompartment model in which the model parameter values
and their relationships to one another (because Leggett used fractions of rate constants to achieve
an overall rate constant) are valid in their current state, only with the data sets for which it was
parameterized. Allowing the user to modify the biokinetic parameters will most likely make the
model inconsistent with the literature with which the Leggett model was parameterized.
Therefore, if EPA would allow the user to vary these parameters, the user must be warned
explicitly that changing the values for the biokinetic parameters will most likely make the model
invalid. Currently, the user is allowed to change the age cutoffs for the biokinetic parameters in
the AALM. However, if this is done, the AALM will be inconsistent with the Leggett model
The specific values used for some of the tissue-specific rate constants in the AALM could
not be reproduced and do not match those Leggett reported. For instance, the parameter
"Deposition fraction of lead in the brain by age range" reports a value for the first age range of
(Age Range, Deposition Fraction in Brain): (0.000, 0.00045). Leggett reports a value of
(0.557/2000=0.000279). Why the difference? These values should be QA/QC'd. If the AALM
uses fractions of rate constants, how is the sum of fractions maintained to have a sum of 1? Will
this cause a mass balance error? A warning should be provided to the user when a mass balance
has not been achieved.
(d) Does the AALM model correctly account for elimination of lead via various pathways?
Breast milk should be added if possible. Non-absorbed lead could be summarized in the
output to provide a confirmation that all lead entering the body is accounted for in the AALM. If
breast milk is incorporated, it will have to be accounted for as a route of elimination for the
mother. The urinary excretion should also be reported as elimination (jig/day). This is the way
that urinary excretion data are reported in many of the scientific papers on lead excretion. For
validation purposes, the model should provide this as an option.
It is not clear if elimination via the dermal pathway is tied to a "skin" compartment that
has feedback with the transdermal absorption pathway. If transdermal absorption is deemed
scientifically justified, is the percent absorbed dependent on blood lead concentrations? If not,
then the transdermal absorption factor can be described as an independent compartment. If the
percent absorbed is found to be dependent on blood lead concentrations, then the skin
compartment will have to be incorporated into the model in a way that the blood lead
concentrations can have a "feedback" type control over transdermal absorption.
II. Predictive Accuracy and Reliability of the Model
Several data sets were identified that could be used to examine the models predictive
veracity. These existing data sets include environmental lead values paired with blood, urine
13
-------
and/or bone lead values for children and adults. The AALM could be calibrated with these data
sets. Regardless of the actual values predicted by the model, several issues of internal
consistency were noted. For example, blood lead values changed abruptly with age, apparently a
result of the step size selected. Also, the integration algorithms need to be verified.
The model should predict a distribution of blood lead values. For the AALM to be used
to characterize variability and uncertainty in blood lead and other output variables, a probabilistic
approach is needed. The AALM Panel generally recommended the use of Monte Carlo analysis
in which exposure and/or biokinetic parameters are characterized by probability distributions.
Both variability and uncertainty in an output variable can be characterized depending on the
choice of input distributions and the choice of Monte Carlo simulation methods. A second
probabilistic approach may also be desirable as an alternative to Monte Carlo analysis, and to
facilitate the transition from the IEUBK model to the AALM. Currently, users of IEUBK are
familiar with the use of a lognormal distribution assumption applied directly to the output
variable (i.e., the geometric standard deviation [GSD] term). While the process for selecting
site-specific GSDs has been a source of considerable debate among the risk assessment
community, it is a simpler method of characterizing distributions and can be informed by
empirical data. One shortcoming of this approach is that it does not allow for quantitative
uncertainty analysis, in that plausible bounds or confidence intervals on model predictions
cannot be determined.
The predictive accuracy of the model could be improved by including considerably newer
information about absorption and internal distribution of lead. For example, much has been
learned about age-dependent bone kinetics. Additionally, improvements to the modeling could
be made in RBC-plasma partitioning, and air-dust relationships. Introducing an "Injection term"
is suggested to isolate the Biokinetic module from the Exposure and Absorption modules. The
default values, particularly for water lead and for the indoor/outdoor lead ratios, should be
reexamined, and the ability to change selected biokinetic parameters should be added. In
addition, a high degree of uncertainty is introduced in the modeling effort by specifying so many
biokinetic model parameters for which there is limited information about their values. Finally,
suitable data sets for validating the model include the National Health and Nutritional Evaluation
Survey (NHANES) data set and the Lanphear compilation of multiple studies.
* Charge Questions 2 & 3: Predictive Accuracy and Reliability of the Model
(2) Based on EPA 's demonstration of the model, what can be stated with regard to the
predictive accuracy and reliability of the AALM regarding comparisons of: (a) model-
generated outputs of projected blood lead distributions derived from real-world lead
exposure data inputs with (b) actual distributions of blood lead (or bone lead)
concentrations for individuals experiencing such lead exposures? In addition, have SAB Ad
Hoc AALM Review Panel members made any "test runs " to apply the current draft version
(1.05) of the AALM to "real-world" datasets that may be available to them; and, if so, what
were the outcomes of such efforts?
COMPARISONS WITH REAL WORLD DATA
14
-------
In that no outputs of projected blood lead were presented to the panel at this time, the
panel considered various existing data sets that might be suitable. The following characteristics
of such suitable data sets were suggested:
Paired concentration of blood lead (PbB) and perhaps also bone lead data with multi-
media lead exposure measurements.
Dust data were collected in ways compatible with the methods used by the model as
inputs. Different dust sampling methods generate both concentration (|ig/g) and lead
loading (jig/area) values.
Data sets that had been examined by structural equation modeling.
To include paint lead observation, but again with a caution about the units being
consistent with the model's need for area loading or lead concentration.
Age in months available, particularly if the time spent in different environments was also
available.
Suitable data sets include the National Health and Nutritional Evaluation Survey
(NHANES) data set and the Lanphear compilation of multiple studies. The data should have
been generated with sufficient concerns for quality assurance, not simply screening data. The
blood and the environmental data need to be paired for each subject.
Additional key lead pharmacokinetic datasets exist that could be simulated and would be
helpful to the scientific community to determine how the AALM performs. These include (but
may not be limited to):
Manton & Malley (1983): urinary lead excretion (jig/day) versus blood lead (|ig/dl)
Van de Vyver et al. (1988): bone lead versus blood lead for workers and the general
population
Other real-world datasets include those from areas where there is extensive lead
contamination and areas where the ambient contamination is much less. These include
Hu and Hernandez-Avila in Mexico, Guilson in Australia and several of the central
European studies. In addition, NHANES (1999-2002) collected both blood and dust lead
from a representative sample of the U.S. population that could be used for this purpose.
The researchers might agree to provide unidentifiable data for this effort that would
streamline 1KB or OMB procedures. HIPAA should not apply in this case.
Dust and other environmental data need to be in the same format as that used by the
model for calibration. For example, dust values (either concentration or surface loading) must be
the same for the data set and the model, as must the location of the sample within the residence
(floor, windowsill, furniture). Similarly, paint lead values need to be surface loading or
concentration.
Calibration of the model should be done with data sets that have both concentration and
loading of dust values. It is important that the way the sample was physically collected and
assayed corresponds with the way the variable is specified to be entered into the model. The
manual must address this point. For example, if the model asks for floor dust lead loading (jig/
unit area), because it was calibrated with that, then the measured dust samples, which will be
15
-------
applied in the future to the model, must be in the same units and from the same location in the
residence. Similar concerns apply to air, water, and other environmental inputs. This would be a
necessary part of the benchmark calibration of the model.
It will not be possible to further refine the various internal biokinetic parameters even if
additional precise environmental lead values and matching tissue lead levels were generated.
Because of the many degrees of freedom, fitting the tissue data will not yield unique solutions.
1. INTERNAL CONSISTENCY
Regardless of the actual values predicted by the model, which could be changed with
calibration efforts, several issues of internal consistency were noted. Results appear to be
sensitive to the step size selected the integration algorithms needs to be verified. Different
opinions were expressed on whether users should be given access to modify the time step. The
step size could be hard-wired, i.e., use a variable time step that is not accessible to the user. In
any case, the manual should caution the user to not change the value.
The model needs to predict blood lead and bone lead trajectories that vary smoothly and
reasonably with time, without abrupt changes at some ages or age boundaries. The abrupt
changes in predicted lead at certain ages, for example, with the onset of middle age, are striking
(see figure below). The current model has this numeric instability, which may be caused by
integration step sizes being too long, or caused by abrupt changes with age at arbitrary age
boundaries. Both of these causes of numeric instability need to be addressed.
Abrupt Changes in Blood Lead with Age
2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 18,000 20,000 22,000 24,000 26,000 28,000
Age(Days)
The discontinuity in the various parameters (only blood concentration is shown above) between
15 years of age and 25 years of age is not justified in the manual and only barely justified in the
1993 Leggett paper.
16
-------
Perhaps related to this were the observed whole blood-plasma irregularities. That the
PbB vs. plasma curves do not overlap, suggests a coding issue. The predicted relationship
between plasma and whole blood lead levels was neither constant nor smooth, nor consistent
with the Leggett model. This requires investigation. It may simply be a matter of choosing more
appropriate time step sizes.
In its current version, the model does not provide numerically correct solutions, i.e.,
either there are coding errors or numerical integration errors. This may be occurring because the
time step is independent of the transfer rates. Plasma lead turnover rate is in the order of a few
minutes, yet the default integration time step is one day and the minimum possible time step is 1
hour. In addition, point changes in biokinetic parameters at specific ages should be smoothed
over time. For example:
0.8 -.
0.7
1 0.6
8 0.5
^^
f) 4 ~~
.Q n ^
55 0.2
0 1
0
v - ...
?
J
--- -1 day time step
1 hour time step
I I I I I I
0 2 4 6 8 10 12
Blood Pb
14
The two curves in this figure show the fraction of blood lead in the plasma for children
from birth to 6 years versus blood lead calculated using 1 day time steps and 1 hour time steps.
The curves are very different, indicating the computer model has errors.
Predicted values should be continuous. One of the discontinuities, for example at ~ 3
years of age, does not occur at a point where there is a dramatic change in biokinetic parameter
values. Regardless of the time step, the fraction of Pb in plasma is too high. It should be around
0.2%. (Leggett [1993], fig. 15, at PbB values below 20 |ig/dl). Independent of the time step
used, there should be only one function, relating Pb in plasma and Pb in whole blood.
2. OTHER CONCERNS FROM RUNNING TESTS OF VERSION 1.05
A. Using environmental values from Boston, applied to children, the model appears to
overestimate PbB by more than five fold. Some calibration is likely necessary.
17
-------
B. Examining the slopes of relationships should be encouraged. In evaluating the model,
it would be useful to look not only for predicted lead levels, but, also to compare the predicted
and observed slopes for the blood/dust, blood/water, or blood/air relationships, to see if they are
consistent with published values from epidemiological surveys.
C. Attention should be given to achieving an appropriate variance in blood lead. To
properly describe the variance of blood lead levels expected in a population, one suggested
method would be to put a distribution of gut absorption rates into the model, rather than a fixed
value. This would generate a distribution in blood lead values. Alternatively, with specified
biokinetic and environmental values, a point estimate in predicted blood lead could be calculated
and then transformed into a suitable (log-normal, perhaps) distribution. By either method, a
reasonable range for blood lead distributions needs to be generated, rather than just a "point
estimate."
3. NEED TO PREDICT A DISTRIBUTION
In its memo to the SAB panel, the EPA states that the goal in developing this model is to
address lead-related regulatory or remedial action decisions. These decisions involve the
estimation of the impact of lead in different media on body burdens of lead in a subpopulation. A
model that would assist the decision maker in estimating the effect of such regulatory or
remedial action would have to predict the impact of exposure on the particular population, not in
a specific individual. The present version of the model is not capable of modeling a population
impact and thus it does not meet the goals of the EPA. It is not clear what "actual distributions
of blood lead" refers to in question (b). The model does not predict blood lead distributions,
rather it provides single estimate versus time. Varying the biokinetic and exposure parameters
can yield the desired distributions.
It would be possible to generate a distribution by varying the biokinetic parameters.
There is a physiological basis for this approach. Indeed, the gut absorption rate is not a constant
among people. Even in carefully controlled metabolic ward settings, with a constant diet, gut
absorption rates vary in the same person from week to week, and vary even more from person to
person (Rabinowitz et al. [1976]). By allowing the gut absorption rate to vary, the model would
create a distribution of blood lead levels.
(3) What advice can the Panel offer with regard to identification of specific features of the
AALM that should be further refined in order to improve its predictive accuracy or to make it
more user friendly? For example, what comments can be offered with respect to default
values assigned for various parameters in the current version of the AALM software?
Which, if any, of those default values may need to be changed and why?
1. IMPROVING THE PREDICTIVE ACCURACY OF THE MODEL
1.1 Include Newer Information
New biokinetic data, that have been available since 1993, should be incorporated into the
model. The data fall generally into three areas (Absorption, Skeletal turnover, Blood/plasma
18
-------
components). Both the Leggett and O'Flaherty models are incomplete and do not include current
understanding in these areas. The EPA should sponsor experimental and computational research
to improve the AALM parameterization in these three areas. Consideration should be given to
the EPA STAR grant program, working within the Superfund Basic Research Program, or via
research contracts.
Regarding absorption, experimentally measured values for gastrointestinal (GI)
absorption range from near zero to almost complete absorption. Since publication of the Leggett
and O'Flaherty models, many relevant studies have described GI absorption or bioavailability
using stable isotope dilution, in humans, swine, etc. EPA should promote retrospective analysis
of existing data and support new research to better define the multimedia bioavailability and age-
dependent absorption of ingested lead. Such effort may require revising and reparameterizing
the Absorption module, for example, to be more like the IEUBK with saturable uptake. Much
new data about the more important metabolic rates can be obtained by stable isotope methods
with fairly brief experimentation times. For example, with the ingestion of a single bolus of lead
tracer, several days of fecal and urine collection, and a few blood samples, most of the basic,
essential rates can be determined, for example, gut absorption, blood pool size, blood turnover
rates, blood to urine rate, and a rate for the movement from blood to deeper pools.
The Absorption module needs to be improved to utilize current data. Gut absorption is
such a driving variable of major importance, that more data about this rate is needed. Research
on better understanding of gut absorption rates should be encouraged.
It should be emphasized in the AALM User's Guide that dust absorption needs to
consider that dust can be re-suspended in the air, and this represents a pathway of exposure to
dust lead. Personal PM2 5 exposure studies suggest that the "personal cloud" is a non-trivial
source of airborne exposure, and re-suspension is an important part of this cloud. The dust model
appears to assume exposure only via ingestion.
Regarding absorption through the lungs, the absorption module does not appear to allow
the deposition rate, or the transfer rate out of the lung, to vary with either the size of the particle,
or the speciation, or at least some surrogate for bioavailability. Size and speciation matter in
absorption. Partially-complexed divalent cations on the surface of a particle deposited in the
lung are easily mobilized and detectable in the blood within 10 minutes of instillation. Stable
oxides will behave quite differently. In addition, since 1993 there has been a great increase in
knowledge of particle deposition in the lung. The information concerning PM2.5 in the Agency's
Final Air Quality Criteria Document (October 2004) is a good source of recent information
regarding absorption in the lung, and deposition parameters in the model.
The Skeletal module needs to be improved to utilize data among age groups that were not
previously available. Much improvement can be made, for example, from neurotrophins (NTs)
and other biomarkers of bone turnover, much has been learned that could be incorporated.
The description of the age-dependent skeletal growth, bone turnover, and lead
accumulation and loss by the Leggett and the O'Flaherty models are incomplete. There is much
room for improvement in the parameterization of skeletal growth, and cortical and trabecular
19
-------
bone formation and turnover. The literature describing bone density and mass using DEXA and
bone formation and resorption using biochemical markers such as circulating osteocalcin, cross-
linked collagen peptides, radioisotopes, etc, as well as skeletal lead (stable isotope, XRF studies)
offers a rich source of information to better parameterize the skeletal lead compartments.
There are data on repeated measurements of tibia and patella lead over time in the
Normative Aging Study, and NTX measurements (a surrogate of bone turnover) at one time. It
would be useful to test the bone model against these data to see how well it predicts bone lead
decline. The default turnover rate for cortical bone may be too small for older adults. A half-life
of-20 years is seen in the Normative Aging Study.
In the blood compartment, the speciation and partitioning of Pb, particularly in the small
and rapidly exchanging plasma pool, needs to be re-considered. Several recent studies describe
the partitioning of Pb in RBCs and plasma. The parameterization and structure of the Central
blood/Plasma compartment should be revisited by reviewing the literature with emphasis on
partitioning and speciation in plasma, including chelating agents. Generation of new
experimental data may be warranted. The EPA should consider using stability constants to
describe the speciation and equilibrium of Pb-small molecule complexes in plasma.
1.2 Improving the Modeling
Given the uncertainty in the biokinetic parameters, one appealing feature of the
O'Flaherty model is that it is a simpler model than Leggett. Although bone as modeled in
O'Flaherty model is not ideal because it doesn't lend to comparisons with XRF data, it could be
useful.
Regarding hair and skin, in section 5.2.16, it is stated that lead removal through hair/skin
and nails is a major route of lead excretion, two orders of magnitude larger than sweat because
its deposition fraction (0.4) is two orders of magnitude larger than that of sweat (0.0035). This
statement is incorrect. The 0.4 indicates that 40% of lead in the intermediate soft tissue pool is
eliminated via hair/nails/skin but 0.0035 means that 0.35% of lead in diffusible plasma is
eliminated via sweat. Since plasma lead turnover rate (<1 min =2000 per day) is much faster than
intermediate soft tissue turnover rate (144 days= 0.007 per day or half life of 110 days) the
amount of lead eliminated in sweat can be much higher than that eliminated via hair/skin/nails.
etal.
Improvements in the biokinetic model settings screen are suggested. Several items in the
editing menu are potentially confusing. Referring to the figure below:
20
-------
1d1. Edi
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
""
i Wuda! jbtujji;a
Parameters
Exposure Age
Last Day
Maximum Cycles
Fetal Exposure On/Off
Fixed Length Delta Option
Delta Step Lengths
Output Step Interval
Acute/Chronic
Mode of Intake
Linear/Nonlinear Model
RBC Threshold Concentration
Nonlinear Parameter 2
Power
Chelation OnjOff
Fixed Delta
Body Size Curve (on/off)
Bone Computation?
Values
0.002740
4745
4745
0
0
Edit
100
2
3
1
60
350
1.5
0
1.0
0
Use Leggett
Default
Units Descriptions
years Age at acute exposure or beginning of chronic exposure
days Maximum number of days
Maximum number of cycles
Fetal Exposure Switch (1 =On 0=0ff) INTERNAL
Use fix length delta (0=Variable >Q=Fixed/TimeStep)
Delta step lengths by age range
days Write output to file only on these steps
Switch for acute or chronic (1 =Acute 2=Chronic)
Selection for mode of intake (0=lnjection 1 =lnhalation 2=lngestion 3=Com
(0= Linear Model 1 =Nonlinear Model)
ug/dl Lead concentration on RBC above whcih a nonlinear model is used
Nonlinear Parameter 2
| Power
Chelation Switch (0=Off 1 =0n)
days Length for Fixed Delta
Computation will use growth curve (1 =on, 0=off)
Bone computation to be used (i.e. Leggett or O'Flaherty)
Save
m
Keys
expage
endday
ncycle
fetal
deltQ
delta
iskip
iacute
inmode
irbc
rbcnl
satrat
power
ichel
deltfix
bUseBodySiie
sBoneComputi
[ Cancel i|
Line 1. Why is this value expressed in units of years when all other inputs are in days? It
would be less confusing to express age in post-natal days (e.g., 0).
Line 4. Change description to (0=off, l=on) to be consistent order with other switch
descriptions.
Line 5. Should read 0=variable, 1= fixed...
Line 7. Is the unit for output step variable in days or in cycles? The original publication and
many programs use cycles.
1.3 An Injection Term
"Injection" (exposure/uptake independent dosing) is an important feature to isolate the
Biokinetic module from the Exposure and Absorption modules. This feature provides
opportunity to compare various Biokinetic and PBPK models of lead and may be useful for
simulating stable isotope studies. It would be appropriate and convenient to include a row for
"injection" in the Exposure module. This approach would work well for situations where
injections are chronic to circumvent the Exposure and Absorption module. For example, an
"injection" of 5 iig Pb/d would facilitate understanding the role of age-dependent changes in
biokinetic parameters in predicted PbB without the complication of age-dependent changes in
person activities and absorption.
Unfortunately, "injection" simulations with only a single injection may be more
complicated to organize through the Exposure module. For example, simulating an "injection" of
21
-------
a stable isotope may be difficult through the exposure page. Thus accommodating "injection"
may require some thought.
2. CHANGES IN DEFAULT VALUES
The water default values need to be re-examined. The flushed water value should be re-
examined, and the daily water intake varying across ages, may not be suitable.
The estimates of the ratio of indoor-outdoor air can be improved based on recent
literature. For the airborne exposure route, the indoor/outdoor ratio of 0.3 chosen for the default
seems low. Many studies have looked at penetration rates and indoor-outdoor ratios in
development of the particulate matter (PM) National Ambient Air Quality Standards (NAAQS).
This literature should be incorporated here, since airborne lead is a particle. In general, readings
of 0.3 are typical of winter-time studies, with summer ratios more like 0.6-0.7, and spring/fall
ratios are closer to summer than winter values. The principal determinant of these ratios is
ventilation rate, which varies geographically. In addition to choosing a more reasonable default,
it would be useful to point the user to this literature, and raise the issue of regional variation.
In assigning the default values to biokinetic parameters, it should be noted that there are
no biokinetic distributions in the model, just point estimates. The lead compartments and lead
flows between compartments represented in the biokinetic component of the AALM are in
reasonable agreement with the proposed kinetic behavior of lead and its disposition in tissues.
However, all values assigned to the model are those presented in the Leggett paper. It should be
noted that some of the compartments are model constructs without necessarily an anatomical
correlate. For example, compartment liver 2 is added to account for a fraction of lead in liver
with very low turnover rate. It does not mean that lead in liver is compartmentalized in two
different physical reservoirs. The kinetic constant for this second compartment is a mathematical
derivation necessary for a fully accounting of the kinetics of lead in liver as constructed in this
model. In other words, other models of lead may choose to parameterize lead in liver differently
and may not need a second compartment and a second rate, or may choose to have three rates.
Because of this, it would be difficult to validate some of these rate constants against literature
values because they are mathematical constructs of this particular model.
3. OTHER IMPROVEMENTS TO THE MODEL ACCURACY
An important part of the AALM validation process is to compare AALM performance to
output of the IEUBK, Leggett, and O'Flaherty Models. The inputs and outputs of each module
should be provided so the modules can be evaluated in isolation. Thus it is critical that the
AALM provide outputs and inputs of each model to facilitate comparisons of inputs and outputs
amongst these models (for example, the output of the Exposure module (jig Pb/d) in the .mod
file). The data contained in this file can be reformatted as "intake" to the Leggett or O'Flaherty
models making it possible to use identical "intakes" to the AALM, Leggett, and O'Flaherty
models.
As discussed elsewhere in this document, the inadequate mathematical description of Pb
22
-------
absorption and bioavailability of Pb in the gastrointestinal tract (GIT) is a central limitation of
the current AALM version. The AALM program should develop an approach to manage
variability in Pb INTAKE between individuals in a population, and in an individual in different
physiological states, e.g.., fed vs. fasted. The future AALM should accommodate user control of
INTAKE and UPTAKE parameters in time domains appropriate to simulate the existing and
future UPTAKE and bioavailability data from studies utilizing stable isotopes. These features are
necessary to support evaluation and simulation of media-dependent Pb bioavailabilities. The
usefulness of a PBPK module to simulate Pb as part of the AALM or to compare and contrast
with the Biokinetic module is limited in the absence of a more accurate and experimentally
validated Absorption module. The time and resources required to incorporate a parallel PBPK
module do not address or solve deficiencies of the Absorption module and thus, the improvement
of the Absorption module should be the higher priority.
The historical dietary exposure data should be revisited to allow reanalysis with
consistent assumptions on non-detects. The historical food data should be re-examined. The
FDA, which generated these data, changed their method for inputting data when lead levels were
below detection limits in their laboratories. Since many values were below detection limits, and
the input methods changed, the values should be recalculated to get a consistent set of historical
data.
Regarding historical air lead data, historical air lead concentrations can be well predicted
from historical gasoline lead usage, which is available. The regression models to do this have
been published (see: Schwartz J, Pitcher H. The relationship between gasoline lead and blood
lead in the United States. J Off Stat 1990;5:421-431; Rabinowitz, M and Needleman H (1983)
Gasoline lead sales and umbilical cord blood lead levels in Boston, Massachusetts. The Lancet
8314: 63; and Rabinowitz M, Needleman H, Burly M, Finch H and Rees J (1984). Lead in
umbilical blood, indoor air, tap water, and gasoline in Boston. Archives of Environmental
Health 39: 299-301.)
If breast milk is incorporated, it will have to be accounted for as a route of elimination for
the mother. In terms of excretion, not only should breast milk be considered an excretory
mechanism for the mother, but perhaps more importantly the fetus is an excretory mechanism.
4. OTHER MODELING CONCERNS
The observed nonlinearities provoke some unease. This is illustrated in the following
observation from an AALM Panel member regarding the relationship of the model equations.
"The model includes a very important non-linearity. It relates to the adjustment of all
deposition fractions (and, thereby, rate constants) out from plasma, based on the deposition
fraction from plasma to RBC (TOORBC). TOORBC is adjusted downward when RBC
concentration (RBCONC) exceeds a certain limit (RBCNL). This results in an upward
adjustment of all other deposition fractions from plasma (i.e.., see variable CF).
Conceptually, what is being simulated is capacity-limited transfer of lead from plasma to
RBC, with lead transfer out of plasma being diverted from RBC to other tissues, when
transfer to RBC approaches capacity."
23
-------
"... the Gear (in ACSL) runs slower (shorter cycle length) when the TOORBC
adjustment is allowed than when it is not (the latter simulating capacity-unlimited transfer
to RBC), presumably because it forces a shorter integration cycle to achieve the specific
error limits on the integration."
It is quite possible that this non-linearity may cause the model to hunt while hitting the
various limits imposed by the non-linearity. For example, when a person is being weighed on a
true balance scale in the doctor's office, if the adjustments that are performed are too coarse, the
balance simply bangs against one mechanical limit and then the other (limit cycles). In the
actual body, although the process may be nonlinear, it is highly unlikely that the body goes
through the same limit cycles.
In the IEUBK model, it was found that there was a very sensitive point involving two
variables, CONRBC and TPLRBC. It is not suggested that a similar problem exists in the
AALM, but it is a coincidence that the question involves RBC.
Stability of the model is also a concern. In general, for the linear portion of the model, the
stability can be determined by the eigenvalues from the stated variable formulation of the
AALM. For stability, these eigenvalues should be real and negative. It is unclear if the model
has been formulated as such. However, in the IEUBK model, it was noticed that the eigenvalues
started off as negative real, as the model evolved in time, these values were reduced in value
approaching zero (0) towards the end of the model time (84 months). A major concern in the
AALM is that the time period is much longer thus emphasizing the direction and magnitude of
the eigenvalues. One test that should be run is the integration (simulation) over the entire time
period with zero (0) input.
5. FRIENDLINESS
5.1 Guidance Manual
To be friendly, the manual needs to be free of textual errors. The current Guidance
manual incurs many errors both in its paraphrasing of the Leggett paper and in mixing flows in
and out of a compartment in the same section and even in the same sentence (sees examples
below). Often these two flows are treated as dependent on each other when in fact they are not.
For example: In 5.2.15 Transfer from fast soft tissue: the reader is directed to figures 68 and 69
which in fact describe parameters controlling the reverse transfer, from plasma to fast soft tissue.
When this reverse flow is addressed in 5.2.23 Deposition fraction from diffusible fraction to fast
soft tissue, figures 68 and 69 are repeated as figures 84 and 83. The same duplication occurs
with the descriptions of plasma lead transfers to and from intermediate soft tissues, slow soft
tissue, and brain.
In addition, several figures in the manual are never cited in the text (figures 35 through
40, and figure 43; figure 43 is a repeat of figure 35). The manual needs a thorough editing to
identify these mistakes and insure that the conveyed information is accurate.
24
-------
Below, are listed the instances where incorrect statements and confusing information
were identified:
In page 36: bottom paragraph: "This..." It is never stated what "this" refers to.
The paragraph's last sentence is missing the contribution of plasma lead directly
to the small intestine.
Figures 37 and 38 are inconsistent for the age 0.274 (i.e., .274 years, or roughly
100 days). Figure 37 lists 0.45 GI absorption and in the graph (figure 38) it is
0.66.
In page 39 model settings: Two parameters referred as having drop down
windows. There are in fact seven parameters with drop down windows.
Page 43: ".. .biokinetic model settings, line 6 units are percent fraction per
day...." Units are not percent, but fraction per day.
Page 51: "... Transfer from the kidney has two components: Kidney 1 from the
kidney back to diffusible plasma and kidney 2 to kidney 1 from the bladder..." It
should be "... two components: kidney 2 from the kidney back to diffusible
plasma and kidney 1 from the kidney to bladder...."
Page 51, section 5.2.12: "...Urinary excretion...The model includes two routes:
diffusible plasma to urinary path and diffusible plasma to Other kidney Tissue.
This is incorrect: It should read: diffusible plasma to urinary path and diffusible
plasma to Urinary Bladder contents....
Page 52 Section 5.2.13 is titled transfer from liver 2 (to plasma, should be added),
but the text deals mostly with the outflow of lead from liver 1 to plasma and to
other compartments.
Page 53:"... this transfer from kidney 2 represents the amount of lead that passes
to the bladder..." In fact, it represents the amount of lead that goes back to
diffusible plasma
Page 58: ".. .The deposition fraction for lead in feces is 0.006.. .and represents the
lead entering the digestive tract from the mucocilliary...." This is incorrect. It is
the lead entering the small intestine from diffusible plasma.
Page 61:".. .fraction of lead deposited in Liver 1 does not vary with age..." this is
incorrect; it varies in Leggett's paper. This reference should be excluded from the
section dealing with deposition fraction that does not vary with age. Furthermore
the section makes reference to deposition fractions in general when in Leggett's
paper the deposition fraction term is used to address the percentage of lead
flowing to different compartment from diffusible plasma, and not the division of
flow out of any other compartments. This needs clarification. The section
continues: ".. .This Liver 1 fraction receives 4% of the lead released by diffusible
plasma, giving a transfer rate of 80/day and a removal half time often days..."
This section is misleading since it implies that the removal half time of 10 days
for Liver 1 is a consequence of the input from diffusible plasma, when in fact it is
due to the flows from Liver 1 to diffusible plasma, to liver 2 and to the small
intestine, with a combined rate of 0.0693/day for a half life of ln(2)/0.0693/day =
10 days. The end of the passage states:"... Forty-five percent (0.45) of the liver 1
25
-------
fraction is deposited in the Small intestine though the bile duct. Most of this is
eliminated with feces; a small amount may be reabsorbed into the diffusible
plasma..." This latter amount is not necessarily small. This lead is reabsorbed
into the plasma at the same rate as ingested lead (page 605, Leggett's paper), i.e.,
the fraction of lead from Liver 1 entering the small intestine that is reabsorbed is
determined by the GI fractional absorption which starts at 45% in early childhood
and decreases to 15% by middle adulthood. In the next paragraph:".. .Most of the
lead...." It is not most: it is 45%.
In page 63: Section 5.2.24 needs thorough rewriting: ".. .This is the fraction
deposited in intermediate Soft tissue from diffusible plasma, with a turnover rate
of 25 to 300 days..." That is incorrect; the turnover rate of the intermediate soft
tissue compartment is not a result of the deposition fraction of plasma-D but is
dictated by its outflow to hair, skins and nails and back to plasma with a
combined transfer rate of (0.00416/day+0.00217/day), resulting in an age
invariable half life of ln(2)/ 0.013/day = 110 days. It continues ".. .this
compartment has a deposition fraction from 0.005, giving a transfer rate of
0.00277.. .etc". The deposition fraction of plasma to this compartment and the
outflow from this compartment are not related to each other, contrary to what the
above implies.
In page 64: "... A small amount of lead is transferred from diffusible plasma to
slow soft tissue with a turnover rate from 1500 to 10000 days..." Again, the
turnover rate of lead in the slow soft tissue compartment is not dictated by the
incoming flow from plasma, which is very fast, but by the slow transfer rate of the
compartment. These two rates are independent. Further, the removal half life of
this compartment in age invariant ln(2)/ 0.00038/day= 1824 days and not 1500 to
10000 days.
In page 66, section 5.2.27: ".. .The fraction of lead in bound plasma that is
transferred to red blood cells is the deposition fraction..."; "in RBC" should be
added. It continues:"... The fraction of lead that is deposited in Extra Vascular
Fluids from diffusible plasma and red blood cells is 0.5..."; "and red blood cells"
should be excluded since it is the fraction coming exclusively from diffusible
plasma.
Some additional, minor editorial suggestions for the Guidance Manual are as follows:
Page 3. Use sentence case for improved legibility
Page 14. Change |ig/g to jig Pb/g
Page 18. Figure 13. Include unambiguous units in output name.
Page 30. Figure 29. What are the units TS?
Page 34. Paragraph 4. The statement that calcium, iron, and phosphorus are
similar to lead is overly simplified.
Page 34. Paragraph 5. Again, this paragraph is overly simplistic and incomplete.
Lead speciation, gastrointestinal tract pH and contents are probably at least as
important as digestive tract calcium. Moreover, this discussion ignores dietary
26
-------
influences on lead uptake that may be mediated by hormones such as vitamin D
(that is, there is a need for more details about factors influencing gut absorption).
Page 35. Figure 34.
Change "Losses in hair..." to "loss to hair...."
Change "In Bone compartments, exchange..." to ".. .exchangeable...."
Change "RT Tract" to "Resp. Tract."
The naming of subcompartments in "Other Soft tissues" is a little confusing; also,
"tenacious turnover" is really "tenacious retention" or "slow turnover" and should
be so described.
Page 36. Figure 35. Line 1. Is the key "kdermal" correct?
Page 36. Last paragraph. First sentence is not clear. "This" is a dangling
participle; ... and liver to the gastrointestinal tract?... "It" is a dangling participle.
Delete "(slower)"
Page 37. Figure 37 (and many other figures). Line 1. "Age Range" is actually the
start age in years? "Decimal percentage" is confusing. Shouldn't this read
"decimal fraction"?
Page 37. Figure 38. The plotted data in Figure 38 do not match the data in Figure
37.
Page 42. Figure 43. Change "Age cut-off to "end age (days)."
Page 43. First paragraph. Change "Pb decay rate" to "Pb radioactive decay rate."
Page 43. Last paragraph. "This" is a dangling participle (both of them)
Page 40. Last heading and paragraph. Change "EXCHANGE" to
"EXCHANGEABLE."
Page 51. Section 5.2.11 Change close up "kidney 1," etc. to "kidney 1" to be
consistent with program labels, liver 1, etc.
Page 51. Figure 61. What is the meaning of the label "indexes" in the Row 1?
Shouldn't this be the end age in years for the variable?
Page 53. Last paragraph. Change "... transfer for Kidney 2" to "transfer from
Kidney2."
Page 54. Last paragraphs. "Binding capacity" is mismatched to the term
"strengths." The AALM Panel suggests that the binding "capacity" should be
restated to relate to turnover. Also see comment above re: Page 35 related to
"tenacious turnover."
Page 60. Figure 78. This figure describing a "chelation" parameter should be
deleted as the chelation is not yet implemented in the AALM.
Page 61. Last paragraph. "This..." is a dangling participle.
In addition, as has been mentioned in other contexts, no guidance is given to the user as
to how to measure "dust" lead. Different methods give different values for concentrations per
gram or concentration per square centimeter (cm2). As part of the validation process, determine
which method seems to be closest to your "dust" input, and let people know.
27
-------
5.2 Guidance Manual Cautionary Notes
The user needs to be warned against changing the biokinetic model parameters by
stressing any changes in the biokinetic model parameters that can/will make the model no longer
equivalent to the Leggett model. The discussion about the differences in the freedom to change
parameters in the model for risk assessors versus researchers needs an explicit statement. The
uncertainty surrounding the numbers generated by the IEUBK model is often not explicit. In this
larger model there may be a temptation to tweak the various parameters to yield a desired
outcome. This model is highly complex and includes numerous parameters for which there is
limited information about their values. A high degree of uncertainty is introduced in the
modeling effort by specifying so many parameters. A purely physiologically-based
pharmacokinetic model, such as O'Flaherty's, has a much smaller number of lead specific
parameters thus reducing the level of uncertainty but maintaining a high degree of complexity
through the parameterization of human physiology in terms of perfusion rates and organ sizes.
Uncertainty about the values of these variables is much smaller. The O'Flaherty model has been
more thoroughly evaluated against real datasets than the Leggett model.
5.3 Making Data Entry Easier
The following are suggested to ease the process of data entry:
A. A batch mode option to simulate different distributions for different environmental
concentrations.
B. A dialogue box, or balloons, that tells the user what the model parameters are and
implications of changing them.
C. The activity patterns window could be improved by allowing the user to re-size it.
For example, resizing the window horizontally would allow as many columns as needed to be
viewed at once. This would be more convenient.
D. Consideration should be given to creating several pre-specified scenarios representing
settings expected to be of particular interest. These might include: an urban child in high risk
housing, an occupationally exposed male adult worker, an older female with osteoporosis having
body burdens from historical exposures in the 1950's, 60's, and 70's, a resident near a
Brownfield's superfund site. From entering the desired scenario, the user would have specified a
set of environmental levels, and the ages of interest. This might prove to be a time-saver.
5.4 Running the Model (including glitches to fix)
The following are suggested to improve operation of the model:
28
-------
1. What/where are the historical diet/air/dust values? Can these values be put in a
separate file (.rtf, .xls, etc)? The check "use historical air Pb concentration" box does not seem
to change the air Pb concentrations.
2. The age ranges should be edited to differentiate between infants 0-0.5 years, toddlers
0.5-3 years, preschool 3-6 years, and school age children 6-12 years. The exposure of two-year -
olds is markedly different from that of four-year-olds. The model should categorize exposures
by the ages of children. The exposure and biokinetic parameters are in some cases tied to age
category, and the AALM should apply these consistently.
3. An algorithm with a variable time step would accommodate both chronic and acute
exposures and would be a better than the current fixed time step. The time step of integration
should not be accessible to the user. The time step is dictated by the degree of numerical
stability and error tolerance in the integration, and the frequency of model output as specified by
the user. The user should be able to specify the desired output frequency and the software should
calculate the time step to be used based on the user input and the numerical needs of the
integration algorithm. The time step should not be a user input.
4. Often, the model runs for longer than the age group specified. This happens after the
software has been used for several model runs in which age groups had been deselected.
Restarting the software avoids this glitch.
5. When run for different lengths of time, the model generates inconsistent results for the
same age group even when the lead intake for the age group is the same in all simulations. This
occurs occasionally and appears to be related to the opening and closing of a new model. For
example if an age group is selected, and the model run, then the age group is expanded and the
model run again, the results of the two runs are consistent. But, if the model is closed after the
first run, and a new one is opened and run with an expanded age range and same exposure
conditions, the outputs of the models runs on overlapping age ranges are inconsistent.
5.5 Improving Reporting of Model Output
The following are suggested to improve reporting of the model output:
The output should automatically report all of the key parameters. Printed graphic output
should include a list of important variables including internal biokinetic parameters that may
have been altered, and environmental variables that may be desired in the output report. The
section explaining the output from the model is extremely poor. Instructions are needed
regarding how to modify the plots, access previous model runs, and combine outputs from
different simulations Instructions should be provided for batch runs.
The output of the AALM Absorption module (uptake to blood) should be written to a
similar file (to the .mod) so that the research modeler can evaluate the behavior of the
Absorption module in isolation. "Pb Uptakes" are not coded as output variables by Leggett
(1993). The user should have the option to customize the default plot. Output files can have
data stored as follows:
29
-------
Exposure Module Output (Intake as jig Pb/delta time)
Interval Drinking Water Dust inhaled, etc.
Total Intake GI Total intake RT Grand Total
Absorption Module Output (Uptake jig Pb/delta time)
Time interval uptake via GI uptake via RT Total uptake
The question, "How are Exposure outputs (jig Pb/d) passed to the Absorption module
(i.e., as Intake)"?, is essentially asking about the synchrony of simulation time steps as data are
passed between the three modules. It is not clear how the modeler controls the time in these time
steps.
It is assumed that the Hour/day switch on the initial AALM window defines only the
Exposure module and that the simulation time steps for the Absorption and Biokinetic modules
are controlled only by the "Edit Model Settings" menu. If this assumption is correct, how are
Uptake values passed to the Absorption module? If the Exposure module simulates Uptake (jig
Pb/hour), does the Absorption module simulate once per hour, and Biokinetic module model
simulates 1 per day? Simulation comparing 1-hr vs. 1-day time steps (selected from the initial
edit page) give appropriate values in the .mod file, but the model-predicted PbB values are very
dissimilar (data not shown). The difference in predicted PbB may be the result of asynchrony
between modules.
The AALM user should also be able to control the "default values" for plot display, not
merely the display of the current plot. This control would facilitate consistency in the axes and
other display parameters for purposes of publication, presentation, etc. The complaint is that, in
its present form, the user must change the plot display every run. The label "age range" (Figure
8) should be made more precise. It can be used to identify a particular age, such as a Start or
Stop Age. The default file name when exporting the MOD file is missing the period resulting in
an incorrect filename.
III. Computer Coding and Quality Assurance
The user interface of the AALM is quite good. The model has an easy, menu-driven
interface that is intuitive, and the learning curve is not very steep. However, there are additional
features that would enable the model to be more useful for either hypothetical or real-world risk
assessment problems. Limitations of the AALM include the following:
A batch mode is needed similar to the current functionality of the IEUBK model to
facilitate an evaluation of the proportion of the population that exceeds a target risk-based
concentration in an exposure medium. The user should be able to specify an input file
with a set of site-specific factors (e.g., paired concentrations in soil, dust, and water at a
residence). The AALM would benefit greatly by allowing either point estimates or
probability distributions to be calculated for each exposure unit.
The AALM needs to incorporate variability and uncertainty more directly. It would be
useful to be able to specify expected distributions of parameters, and get out a probability
distribution of blood lead for a population. Default distributions, rather than default point
estimates, for these parameters would be preferred so that variability and uncertainty are
30
-------
more properly accounted for in the risk assessment without the requirement of tedious
repetitions by the risk assessor.
Even for the research community, caution should be given about changing the biokinetic
parameters, given that they were not derived independently, and changing one often
implies changes to others. The results would also no longer correspond to the Leggett
model. Interest was expressed in adding the physiological (O'Flaherty) model option.
Specific suggestions for the AALM include: eliminating the need to set the gender option in
three locations; and increasing flexibility in the graphic display.
Regarding the QC of the program, the AALM does not perform correctly. For example,
Manton and Cook's data indicate that plasma lead should be about 0.2% of blood lead when
blood lead is less than 25 mg/dL. The Leggett model, on which the AALM is based, predicts
this successfully. However, the AALM not only does not meet this design criteria, it produces
non-single-valued functions. Small errors in the parameterization of the kinetics of this
compartment can propagate very rapidly to errors in the amounts of lead in all other
compartments. Since the AALM derives from the Leggett model, it is assumed that this is a
result of coding errors. It is recommended that efforts be made to fit the AALM to the same
datasets as the Leggett model.
In addition, plausibility checks should be run such as making sure results behave as
expected as the number of years is lengthened, that different intakes for different periods behave
as one would expect, given cumulative dose, etc. It would also be worthwhile to examine other
data sets. Finally, several of the input assumptions are unreasonable, and should be changed.
This includes assuming the same gut absorption rate for food and water lead, the default values
for water lead concentration, etc.
* Charge Questions 4 & 5: Computer Coding and Quality Assurance
(4) Based on any trial-run experiences of Panel members, what can be said about the "learning
curve " needed to become sufficiently-familiar with the AALM software in order to effectively
apply it? Furthermore, assuming that one had a need to apply the AALM to a hypothetical
or real-world risk assessment problem, what additional information (if any) about the AALM
might be useful for a user to have in order to correctly and efficiently apply the model and
enhance effective communication of modeling outcomes? What comments can the SAB Ad
Hoc AALM Review Panel offer concerning output features (e.g., tabular presentation of
modeling results, graphic display options, etc.)?
(5) In the judgment of the SAB Ad Hoc AALM Review Panel, to what extent has the computer
code comprising the AALM software been adequately verified and appropriate quality
assurance checks carried out and/or planned? What additional quality control/quality
assurance checks, if any, would the Panel recommend?
Question 4 relates to the user interface of the model. Overall, this subgroup found the user
interface to be quite good, but for a more limited goal than would be desirable. The model has
an easy, menu driven interface that is intuitive. The format for entry of exposure parameters was
-------
very useful and intuitive, and the results are generally presented well. Hence, the members of
the AALM Panel do not think the learning curve is very steep. However, AALM Panelists do
believe that there are additional features that would enable the model to be more useful for either
hypothetical or real-world risk assessment problems.
The ability to vary a large number of parameters, while of use to the research community,
may be confusing, and tempting, to the risk assessor. It is easy to get into trouble, and Panel
members wonder whether a risk assessor option that fixes some of the choices would be a useful
option.
Specific Suggestions/Comments:
The gender option needs to be set in three locations. This is awkward, and can lead to
errors, since the locations are not linked.
The graphic display is too inflexible. Units are not displayed, axes scales are not
flexible, and it is not clear how to save graphs. "Time" should not have a scale of
days as this makes it too hard to examine longer term results.
It is important to include a soil lead input that is separate from the dust lead input
Question 5 relates to QC of the program. Definite problems were found and quite
simply, the model does not perform correctly. For example, Manton and Cook's data indicate
that plasma lead should be about 0.2% of blood lead, when blood lead is less than 25 mg/dL.
The Leggett model, on which the AALM is based, successfully predicts this. However, the
AALM not only does not meet this design criteria, it produces non single-valued functions.
Simulations were run from birth to middle age that assumed two different levels of dust lead
intake, i.e., lead inputs of 10 ppm or 25 ppm lead in dust as the only exposure. The figure below,
with the results of those runs, can only be considered to be in error.
0.8
0.7
0.6
_
a.
c 0.4
ra 0.3
.2
55 0.2
0.1
0
10 ppm dust
25 ppm dust
0.2 0.4 0.6 0.8
blood concentration
1.2
32
-------
(Percent lead in plasma was calculated from the ratio of output variables Plasma and Blood, and
the X-axis is the output variable Blood lead concentration.)
The numbers are out of range; they differ depending on the dust lead level used to arrive
at the same blood lead concentration; and, as noted before, they are non single-valued functions.
While the percent of lead in plasma is not a variable of importance from a regulatory perspective,
the amount of lead in plasma is critical in this model because it is the compartment feeding lead
to all other compartments with a very fast turnover rate. Thus, small errors in the
parameterization of the kinetics of this compartment can propagate very rapidly to errors in the
amounts of lead in all other compartments. Since the AALM derives from the Leggett model, it
is assumed that this is a result of coding errors.
Validation using the NHANES data, and in particular demonstrating that the observed
trends in U.S. population lead levels can be replicated, would be quite useful if the intent is to
use to model to examine effects of NAAQS or other regulatory changes. Furthermore, since the
aim of the "all ages" lead model is to address potentially susceptible groups beyond children,
then pregnancy, lactation, and postmenopausal bone mobilization population options should be
included.
IV. AALM Documentation
The AALM Panel deems that the present Guidance Manual is useable, but should be
made more user-friendly. It is incomplete in several areas and contains many errors and
confusing wording. The manual should provide both a theoretical framework for the uninitiated
user to understand the structure of the model and its scientific basis, and a step-by-step procedure
that would walk the user from data input to the evaluation of the predicted outcomes. EPA
should also consider developing and releasing a companion Technical Support Document to
augment the Guidance Manual that includes verification and validation exercises, utilizing real
world demonstrations, and appropriate cautions that would aid the user in understanding,
interpreting and utilizing the model.
In addition, the AALM Panel notes that the output options provided in Version 1.05 are
interim choices and will therefore need to be developed more fully in subsequent releases. The
data files, if possible, should be exportable to other traditional software programs. More
explanation of the structure, underlying nature and accessibility of these data sets should be
provided. There should also be explanation regarding the type of environmental information to
be entered, so that it is standardized to the type for which the model was calibrated. The
Parameters Dictionary was an extremely important component of both the guidance and the help
feature. The AALM Panel suggests that more specific information be provided in the guidance
and support documents with regard to each individual parameter, including its origin, source of
support data, possible range of values, any information regarding central tendencies and
variance, uncertainty, the rationale for the default setting, relationship to other parameters, and
appropriate cautions as needed for modifications.
33
-------
To the extent practicable, the AALM approach, guidance and application should be
consistent, and evolve concurrently, with similar models and guidance presently endorsed by
EPA and used by the Agency in the professional and scientific community. The Agency should
consider issuing guidance regarding the required (or recommended) use of the default or
prescribed bio-kinetic parameters in regulatory applications. Finally, the AALM should be
evaluated relative to the Agency's current Draft Guidance on the Development, Evaluation, and
Application of Regulatory Environmental Models.
»»» Charge Questions 6-9: AALM Documentation (e.g., Guidance Manual, Parameters
Dictionary, etc.)
(6) To what extent is the "AALM Guidance Manual" sufficiently clear and useful in providing
"user friendly " instructions for carrying out model runs for AALM applications? How might
the AALM user's manual be improved to help facilitate use of the model?
Implementation: The group felt that the model guidance was mechanically constructed
in a typical "point and click" format. This had the advantage of making the model easy to access
and implement. However, members thought users were able to "RUN" the model with almost no
orientation or introduction to the purpose, structure, format or construct, and suggested that the
guidance manual be significantly augmented with examples, demonstrations and appropriate
cautions that would aid the user in understanding, interpreting and utilizing the model. The
example screens provided in the guidance should be reproducible in the tutorials and
demonstrations.
Options and Features: Most of the options and features were found to be confusing,
largely because they were either unexplained in the guidance or were under development and not
available. For some of the options that were implemented, the paraphrasing and descriptions in
the text seemed inconsistent with either the information available in the Help Screens, or
information that could be deduced from applying the option.
Outputs: The group understood that the output options provided in Version 1.05 are
interim choices and will be developed more fully in subsequent releases. Accordingly, most of
the comments may be addressed with adoption of new software. Such issues included units,
rounding of values, scales, graphing inconsistent units on the same plots, etc. The panel felt that
the plotting function was especially useful and should be included and upgraded in future
releases. If possible, the data files should be exportable to other traditional software programs.
In that light, more explanation could be provided of the structure, underlying nature and
accessibility of these data sets.
Convenience: The group felt that several items would make the model more convenient
for users. Inputs could be facilitated by employing "drag and click" or "copy and paste" options
for the various age-groups, etc. Blood lead concentration should be a default output parameter
and not be plotted with other compartments with inconsistent units. The guidance should explain
the quantitative uncertainty in, and require more effort to vary, the bio-kinetic parameters. There
should be explanation regarding the type of environmental information to be entered, so that it is
standardized to the type for which the model was calibrated. Accessible interim (during model
34
-------
setup and operation) output from different modules would be a great improvement. For example,
providing a summary of route-specific inputs of lead via the dietary, soil and dust, water and air
routes following the setup of the exposure module would be most helpful. Similarly, route or
pathway specific summaries of absorbed lead would facilitate understanding of bioavailability
and bio-accessibility. A batch mode application that aggregates results for multiple individuals
in a population would be an important addition for risk assessors.
(7) To what extent are the entries in the "Parameters Dictionary "for the AALM sufficiently
clear and accurate in explaining important elements of the AALM? How might the
Parameters Dictionary be improved?
The Parameters Dictionary was considered an extremely important component of both
the guidance and the help feature. As structured, the dictionary was helpful to programmers
accessing the code. As this is expected to be an "open code" model, the information provided
should be retained. However, the overall concern was that this is a "parameter rich" model and
much more information should be provided in the guidance or available technical support
documents. The group suggested that more specific information be provided in the guidance and
support documents with regard to each individual parameter, including its origin, source of
support data, possible range of values, any information regarding central tendencies and
variance, uncertainty, the rationale for the default setting, relationship to other parameters, and
appropriate cautions when modifying them. It was suggested that this information be accessible
through "hot button" connections on-screen from the parameters dictionary or help menu.
(8) Are there any other comments or advice that the SAB AALM Review Panel wishes to provide
with regard to ways that the AALM, its software, and other associated materials can be
improved to help to facilitate its application and enhance the usefulness of its results?
The group chose to reiterate earlier concerns with regard to this charge. There were
QA/QC related to the potential interactions and inter-relationships among parameters. There
should be internal mass-balance checks, perhaps with an appropriate notice or warning that
conservation requirements are met, when parameter modifications are attempted. It is unclear
whether modifications to parameters in one screen cause (or necessitate) changes in related
screens. It was unclear if it is the user's responsibility to conserve the mass balance when
changing fractional parameters in the input screens. Special attention should be paid to units
throughout the procedures, output and feedback. There should be appropriate discussions
regarding uncertainty (both qualitatively and, to the extent practicable, quantitatively) in the
model; sensitivity to particular parameters; and limitations of the model to selected applications.
(9) Does the AALM follow the Agency's Regulatory Environmental Model Guidance found at
URL: http://cfpub.epa.gov/crem/?
To the extent practicable, the model approach, guidance and application should be
consistent, and evolve concurrently, with similar models and guidance presently endorsed by
EPA and used before the Agency in the professional and scientific community. The Agency
should consider issuing guidance regarding the required (or recommended) use of the default or
35
-------
prescribed bio-kinetic parameters in regulatory applications. Alternatively the Agency could
release two versions of the model for researchers and risk assessors. Finally, the model should
be evaluated relative to the Agency's current Draft Guidance on the Development, Evaluation,
and Application of Regulatory Environmental Models.
36
-------
Appendix A
Charge to the SAB Ad Hoc All-Ages Lead Model Review Panel
The Agency seeks the review and advice from the SAB regarding the scientific
soundness of the All-Ages Lead Model, and requests that the AALM Panel focus on the
following charge questions during its review of the AALM:
(1) In general, to what extent are the parameters and relationships represented by various
AALM features adequately supported by available research findings in published peer-reviewed
literature or by reasonable extrapolations from such findings? That is, are the specifications of
key components of the AALM model scientifically supportable in characterizing particular
parameters or relationships of the types noted above. More specifically, what are the AALM
Panel's views with regard to:
(a) The adequacy of the values specified for the exposure parameters for different media
and how well the model interprets exposure throughout the various age groups;
(b) The adequacy of the uptake/absorption parameters or any need for modification of
the methodology for determining absorption for various routes of exposure;
(c) Whether there are any errors in AALM methods for determining biokinetic
distribution or errors in assigning values to biokinetic parameters; and
(d) Does the AALM model correctly account for elimination of lead via various
pathways?
(2) Based on EPA's demonstration of the model, what can be stated with regard to the
predictive accuracy and reliability of the AALM regarding comparisons of: (a) model-generated
outputs of projected blood lead distributions derived from real-world lead exposure data inputs
with (b) actual distributions of blood lead (or bone lead) concentrations for individuals
experiencing such lead exposures? In addition, have AALM Panel members made any "test
runs" to apply the current draft version (1.05) of the AALM to "real-world" datasets that may be
available to them; and, if so, what were the outcomes of such efforts?
(3) What advice can the AALM Panel offer with regard to identification of specific
features of the AALM that should be further refined in order to improve its predictive accuracy
or to make it more user friendly? For example, what comments can be offered with respect to
default values assigned for various parameters in the current version of the AALM software?
Which, if any, of those default values may need to be changed and why?
(4) Based on any trial-run experiences of AALM Panel members, what can be said about
the "learning curve" needed to become sufficiently-familiar with the AALM software in order to
effectively apply it? Furthermore, assuming that one had a need to apply the AALM to a
hypothetical or real-world risk assessment problem, what additional information (if any) about
the AALM might be useful for a user to have in order to correctly and efficiently apply the
model and enhance effective communication of modeling outcomes? What comments can the
A-l
-------
AALM Panel offer concerning output features (e.g., tabular presentation of modeling results,
graphic display options, etc.)!
(5) In the judgment of the AALM Panel, to what extent has the computer code
comprising the AALM software been adequately verified and appropriate quality assurance
checks carried out and/or planned? What additional quality control/quality assurance checks, if
any, would the AALM Panel recommend?
(6) To what extent is the "AALM Guidance Manual" sufficiently clear and useful in
providing "user friendly" instructions for carrying out model runs for AALM applications? How
might the AALM user's manual be improved to help facilitate use of the model?
(7) To what extent are the entries in the "Parameters Dictionary" for the AALM
sufficiently clear and accurate in explaining important elements of the AALM? How might the
Parameters Dictionary be improved?
(8) Are there any other comments or advice that the AALM Panel wishes to provide with
regard to ways that the AALM, its software, and other associated materials can be improved to
help to facilitate its application and enhance the usefulness of its results?
(9) Does the AALM follow the Agency's Regulatory Environmental Model Guidance
found at URL: http://cfpub.epa.gov/crem.
A-2
-------
Appendix B
References
Bergdahl IA et al. 1991. Lead concentrations in human plasma, urine and whole blood. Scand. J.
Work Environ. Health. 23: 359-363.
Billick ffl et al. 1979. Analysis of pediatric blood lead levels in New York City for 1970-1976.
EHP31: 183-190.
Chamberlain AC. 1983. Effect of airborne lead on blood lead. Atmos. Environ. 17: 677-692.
Clark CS et al. 1985. Condition and type of housing as an indicator of potential environmental
lead exposure and pediatric blood lead levels. Environ. Res. 38: 46-53.
De Silva PE. 1981. Determination of lead in plasma and studies on its relationship to lead in
erythrocytes. Br. J. Ind. Med. 38: 209-217.
LeggettRW. 1993. An age-specific kinetic model of lead metabolism in humans. 101: 598-616.
Marcus AH. 1985. Multicompartment kinetic model for lead. III. Lead in blood plasma and
erythrocytes. Environ. Res. 36: 473-489.
Manton WI, Cook JD. 1984. High accuracy (stable isotope dilution) measurements of lead in
serum and cerebrospinal fluid. Br. J. Ind. Med. 41: 313-319.
Mushak P. 1998. The uses and limits of empirical data in measuring and modeling human lead
exposure. EHP 106 (Suppl. 6): 1467-1484).
Mushak P. 1993. New directions in the toxicokinetics of human lead exposure. NeuroToxicology
14: 29-42.
O'Flaherty EJ. 1998. A physiologically based kinetic model for lead in children and adults. EHP
106 (Suppl. 6): 1495-1503.
Pounds JG, Leggett RW. 1998. The ICRP age-specific biokinetic model for lead: Validations,
empirical comparisons, and explorations. EHP 106 (Suppl. 6)1505-1511.
Rabinowitz M, Wetherill G, Kopple J. 1976. Kinetic analysis of lead metabolism in healthy
humans. Journal of Clinical Investigations 58: 260-270.
B-l
------- |