UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
August 30, 2007
EPA-CASAC-07-006
Honorable Stephen L. Johnson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, NW
Washington, DC 20460
Subject: Clean Air Scientific Advisory Committee's (CASAC) Review of EPA-OPPT's
Draft Approach for Estimating IQ Change from Lead Renovation, Repair, and
Painting (LRRP) Activities and the OPPT Dust Study
Dear Administrator Johnson:
The Clean Air Scientific Advisory Committee (CASAC or Committee) Panel for Review
of EPA's Lead Renovation, Repair, and Painting (LRRP) Activities (CASAC Panel) convened
on July 9-10, 2007 in Durham, NC to conduct a peer review of the Agency's Draft Approach for
Estimating Changes in Children's IQfrom Lead Dust Generated During Renovation, Repair,
and Painting in Residences and Child-Occupied Facilities (Draft LRRP Activity IQ-Change
Methodology, June 2007), and the Draft Final Report on Characterization of Dust Lead Levels
After Renovation, Repair, and Painting Activities (OPPT Dust Study, January 2007); and met
again on August 7, 2007 via teleconference to hold further deliberative discussions concerning
this letter. The roster of CASAC members is attached as Appendix A of this letter, and the Panel
roster is found in Appendix B. EPA's charge to the Panel is contained in Appendix C to this
letter. Panel members' responses to those Agency charge questions are provided in Appendix D,
and Panelists' individual written comments are attached as Appendix E.
EPA's Office of Pollution Prevention and Toxics (OPPT) had requested that the CASAC
conduct a peer review of these two Agency documents — and an earlier consultation on OPPT's
Draft Assessment to Support the LRRP Rule (1st Draft LRRP Assessment, January 2007) — in
support of the EPA's LRRP rule-making activity. In a notice published in the Federal Register
on January 10, 2006 (71 FR 1587-1636), the Agency proposed new requirements to reduce
exposure to lead hazards created by renovation, repair, and painting (RRP) activities that disturb
lead-based paint. This action supports the attainment of the Federal government's goal of
eliminating childhood lead poisoning by 2010. The rule is intended to address EPA's concern
that RRP work conducted by untrained and uncertified contractors may create new lead hazards,
thus increasing the risk of lead exposure to the residents of homes containing lead-based paint.
-------
The peer review of the Agency's Draft LRRP Activity IQ-Change Methodology, June
2007), and the OPPT Dust Study was a follow-up to the CASAC's consultation on 1st Draft
LRRP Assessment, January 2007) conducted on February 5, 2007, as documented in our letter to
you dated April 3, 2007 (EPA-CASAC-07-004). Some of the key points that the CAS AC Panel
suggested to the Agency have yet to be addressed; for example, a separate consideration of
uncertainty and variability; use of the most recent epidemiology studies indicating that children
are more sensitive to lead poisoning than previously thought; and, in particular, greater emphasis
on the use of empirical data rather than on model estimates alone.
Overall Evaluation
The CASAC Panel is pleased to review and provide advice to the Agency on this topic.
This letter provides a summary of the major findings with respect to the CASAC's peer review
of the OPPT Dust Study and the Draft Approach for Estimating IQ Change from LRRP
Activities, which OPPT developed to support the benefits assessment in the economic analysis
required for "significant regulatory action[s]" such as the LRRP rule-making. The Agency
authors are to be commended for the significant effort that went into the document to define and
describe the conceptual elements linking Rule and non-Rule LRRP protocols to established
biomarkers of lead exposure and established toxic endpoints from these exposures. In addition,
the Panel found that the OPPT Dust Study was reasonably well-designed, considering the
complexity of the problem, and that the report provided information not available from any other
source. Nevertheless, as detailed below, the Panel's overall conclusion was that the available
experimental or empirical data are limited and that the modeling procedures and analyses are
inadequate to support the proposed modeling approach for estimating the IQ changes in
children exposed during renovation procedures. Therefore, the modeling approach in its present
form would not adequately support a rigorous cost/benefit analysis. Accordingly, while the
Panel believes the overall concepts in the proposed methodological approach are reasonable,
the CASAC Panel cannot endorse the specific steps, procedures, and data analyses contained in
the draft Agency methodology document.
Furthermore, the Panel had two other overarching concerns beyond the review of these
documents:
• Although the topic is not strictly related to the review of the draft methodology, the Panel
questions whether it is appropriate to conduct economic analyses on each increment of
exposure to a multimedia pollutant such as lead (e.g., during a restoration) without
considering the degree to which that increment builds on other pathways of exposure.
Evaluating each source of lead as an increment to all other sources could, in the extreme,
lead to the conclusion that no individual source is significant while it is clear that the
combined, accumulated effects of multimedia lead exposure is harmful.
• EPA should give much greater priority to this effort of developing improved processes
for lead RRP activities so as to decrease childhood lead exposures in homes and other
child-occupied facilities (COFs) in various parts of our country. Since 2000, when the
Agency joined with other Federal agencies in establishing the worthy and, indeed, critical
goal of eliminating childhood lead poisoning within 10 years, the evidence suggests that
EPA has to date given only limited emphasis to developing a reliable means to decrease
-------
exposure of children to lead used as pigments in many of the interior and exterior paints
applied during the original construction and renovation of homes, schools, day-care
centers, and other buildings that are frequently occupied by children. In particular, the
CASAC agrees that there is ample evidence that exposure of children to lead dust poses a
major health risk. The data are sufficiently compelling as to require both prevention of
new, and control of existing, dust lead exposures. Repair and renovation of homes and
COFs where lead-based paint surfaces are present require practices that minimize dust
lead exposures to children. The Panel recommends that young children be removed from
such premises where feasible. If this is not feasible, it is important for the Agency to
provide guidance to reduce such exposures to children.
With respect to its review of the two Agency documents, the Panel was concerned about
both regulatory and methodological issues, as follows:
Regulatory Concerns
• OPPT's draft methodology is likely to underestimate IQ loss because blood lead levels of
concern for adverse health effects are based on the exposure-response information under-
lying the current lead national ambient air quality standards (NAAQS) that are presently
under Agency review. The CASAC has recently recommended that the Lead NAAQS be
lowered substantially (see the CASAC's letter to you [EPA-CASAC-07-003] dated
March 27, 2007). In the CASAC's opinion, these standards need to be strengthened in
view of recent epidemiological data indicating that children are more susceptible to
effects from lead exposure than was previously thought.
• Outdated residual surface contamination standards (i.e., dust lead cleanup levels of 40
|ig/ft2 for floors and 250 |ig/ft2 for window sills) are being used that are insufficiently
protective of children's health, as indicated by recent epidemiological studies. In
addition, while setting maximum allowable cleanup concentrations is a reasonable
approach, benefit estimates must take into account benefits that accrue from decreasing
environmental exposures to lead below those concentrations, which any reasonable
cleanup protocol will do much of the time (see the individual written comments of Panel
member Dr. Joel Schwartz on page E-48).
Methodological Concerns
• The cleaning procedures employed are inadequate, such that post-cleaning lead levels do
not even meet the existing EPA standards. Moreover, the qualitative and simplistic
method used to verify the effectiveness of these cleaning procedures — i.e., the "white
cloth verification tests" — does not yield consistently reliable results, leading to an
inaccurate assessment of cleaning efficiency after repair and renovation activities.
• The limited data from residential housing units and COFs included in the Dust Study,
which are used as input into the biokinetic models, most likely do not represent a
statistically-valid sample of housing at the national level. On the one hand, the estimated
risks may overstate what might be encountered on average nationally because the
-------
examples studied likely represent high-hazard scenarios reflective of the upper tails of the
distribution. On the other hand, several modifications in the study design falsely
diminish the estimated risk of lead hazards due to repair and renovation activities, e.g.,
excluding eight out of 35 housing units because they were in poor condition; excluding
housing units with floors in poor condition; use of sample trays in place of window sills
because of inability to achieve clearance standards; and the use of plastic sheeting on
some tool and observation rooms. In order to use the OPPT data, EPA must quantify the
extent to which the factors just mentioned, as well as others, might lead to either an
overestimate or an underestimate of risk, and hence an overestimate or underestimate of
the regulation's benefits.
• In both the draft IQ-Change Methodology and the Dust Study, there is no indication of
the variability of the findings (e.g.., standard deviation, range, etc.)., which should be
given in the main body of each document.
• High uncertainties are associated with the use of the biokinetic lead models, especially
for episodic exposures. The Integrated Exposure Uptake Biokinetic (IEUBK) model for
lead in children is unsuitable in this exposure scenario because it does not allow use of
episodic exposure data input and is only concerned with steady-state conditions. As a
result, unless it is decided that the exposure-response function can use a time-weighted
average exposure rate over some significant period of time that includes pre- and post-
cleanup, and that shorter-term temporal patterns of exposure are not important, the
IEUBK model will not be suitable. The Leggett model does allow use of episodic
exposure data, but predicted blood lead concentrations appear to be biased high when
compared to the IEUBK and O'Flaherty models applied to the same steady-state
exposure scenarios. Therefore, it is unclear if Leggett model is biased high when applied
to an acute or episodic sub-chronic exposure scenario. Moreover, the use of biokinetic
models for episodic exposure creates additional uncertainties because of short-term
variations in behavioral, dietary, and biokinetic parameters that broaden the distribution
of expected blood lead concentrations. Accordingly, the Panel recommends use of the
Leggett model with an uncertainty analysis that identifies and quantifies sources of
potentially high uncertainty associated with the model estimates. The CASAC Panel also
strongly favors greater use of empirical data for estimating blood lead levels following
exposure to lead dust during renovation activities, to aid in the evaluation of the
usefulness of the biokinetic component of the Leggett model.
• There is a lack of consideration for the activity patterns of children as a sensitive
subpopulation. The Panel felt that changing the geometric standard deviation (GSD)
from 1.2 to 1.6 would, in part, alleviate this problem. However, a GSD of 2.0 or 2.1 is
probably most appropriate given all of the uncertainties in the analyses to be performed.
A more scientifically defensible approach, and one used routinely in other regulatory
analyses, would be to develop separate distributions of activity patterns for children, with
their own central tendency and GSD values.
The Panel's specific comments and recommendations on these two Agency documents
are as follows:
-------
Comments on OPPT Dust Study
The Dust Study, which was conducted by an outside contractor (Battelle), was designed
to compare environmental lead levels at appropriate stages after various types of RRP activities
were conducted on the interior and exterior of residential housing units and COFs. All jobs
disturbed more than two square feet (2 ft2) of lead-based paint, which is the de minimus amount
of disturbed area to which the proposed rule applies. Of particular interest was the impact of
using specific work practices that repair and renovation contractors would be required to follow
under the proposed rule. The RRP procedures undertaken represented the range of activities that
are permitted under the proposed rule. Importantly, the Dust Study also provided input for the
type of exposure data needed for the draft LRRP document.
The Panel found that the OPPT Dust Study was reasonably well-designed, considering
the complexity of the problem, and that the report provided information not available from any
other source. The study indicated that the rule cleaning procedures reduced the residual lead
(Pb) remaining after a renovation more than did the baseline cleaning procedures. Another
positive aspect of the Dust Study was that it described deviations from the protocol when they
occurred.
Despite these positive aspects of the OPPT Dust Study, the CAS AC Panel had several
areas of significant concern relative to both regulatory issues and study methodology:
Regulatory Concerns
• The lead dust loading values of 40 |ig/ft2 for floors and 250 |ig/ft2 for window sills are
presented as adequately protective of children against lead poisoning, i.e., to guard
against blood lead levels of greater than ten (>10) |ig/dL. However, the Panel notes that
these residual surface contamination standards are obsolete on the basis of recent
epidemiology findings that indicate that adverse health effects are found in children with
blood lead levels less than five (<5) pg/dL (Lanphear et al., 2005; Lanphear et al., 2002,
Lanphear et al., 1998, Lanphear et al., 1996, and Malcoe et al., 2002). Unless EPA's
new LRRP regulation reflects the underlying exposure-response information reported in
the forthcoming Lead NAAQS documents that are presently undergoing Agency review,
public health will not be adequately protected.
Methodological Concerns
• It is problematic that the Dust Study appears to ignore measured lead values indicating
that post-cleaning Pb levels do not meet even the current EPA standards. Such non-
compliant measurement data strongly suggest that a modification of the cleaning
procedures is required. As an example of an effective cleaning method, there are data
available from the ongoing Cincinnati Children's Hospital Medical Center HOME
(Health Outcomes and Measures of the Environment) Study, funded in part by EPA,
demonstrating that over 99% of housing units can achieve dust lead loading values of 10
Hg/ft2 and 50 |J,g/ft2 on floors and windows sills, respectively, after the implementation of
interim lead hazard controls (which are similar to repair and renovation activities).
-------
• The method used to assess the effectiveness of the cleaning procedures does not yield
results that are consistently reliable. The use of white cloth (wet or dry) swipes to verify
the degree of cleanup by discoloration was employed as an alternative to lead
measurements because the white cloth test was "quick, inexpensive, reliable and easy to
perform." However, quality assurance tests of the method showed that the results were
not reliable. As an example, the report states, "Overall, only three window sills failed the
first wet cloth verification despite the fact that nineteen window sills had post-cleaning
levels >250 |ig/ft2." The study also points out that the simple coloration test had several
other problems associated with it as well, including the fact that some forms of lead are
white and that the cleanup solution (Simple Green®) sometimes discolored the white
cloths. One member of the CAS AC Panel suggested that the white cloth test might be
used as a screening method, but that a measured value should be obtained for final
verification of the effectiveness of the clean-up. Indeed, for as important a responsibility
as protecting children against lead poisoning, the Panel strongly feels that it is imprudent
to substitute a simplistic and qualitative white cloth test for highly-specific, analytical
measures of lead in house dust.
• The limited data from residential housing units and COFs included in the Dust Study,
which are used as input into the biokinetic models, most likely do not represent a
statistically-valid sample of housing at the national level. On the one hand, the estimated
risks may overstate what might be encountered on average nationally because the
examples studied likely represent high-hazard scenarios reflective of the upper tails of the
distribution. On the other hand, several modifications in the study design falsely
diminish the risk of lead hazards due to repair and renovation activities, e.g., excluding
eight out of 35 housing units because they were in poor condition; excluding housing
units with floors in poor condition; use of sample trays in place of window sills because
of inability to achieve clearance standards; and the use of plastic sheeting on some tool
and observation rooms. In order to use the OPPT data, EPA should quantify the extent to
which the factors just mentioned, as well as others, might lead to either an overestimate
or an underestimate of risk, and hence an overestimate or underestimate of the benefits of
the regulation.
In addition to these major concerns, several additional minor or editorial changes are
recommended:
• The conclusions of the report should be linked to the Dust Study's objectives.
• Dust loadings, as well as dust lead loading, should be reported, if possible.
• The tables and figures are not complete, and some figures do not have properly
labeled axes. In some tables, only the p-values are given; the magnitude of changes
should also be stated.
• The Panel thought that the Dust Study should have documented blood lead levels in
the workers before and after their work activities. This important information could
have been reported in a form that could not be linked to the individual workers.
-------
Additional remarks on the OPPT Dust Study are found in Panel members' individual
written comments provided in Appendix E.
Comments of Draft Approach for Estimating IQ Change from Lead RRP Activities
(Draft LRRP Activity IQ-Change Methodology)
OPPT developed the Draft LRRP Activity IQ-Change Methodology to support the
benefits assessment of the economic analysis required for "significant regulatory action[s]" such
as the LRRP rule-making. As stated in the introduction to the draft Approach document, "The
quantified benefits analysis will be based primarily on changes in neurocognitive function in
children (as measured by IQ) due to lead exposure from specific renovation, repair and painting
(RRP) activities. OPPT is using data from a variety of sources ... to determine the specific types
and frequencies of RRP activities that occur in residences and child-occupied facilities." To
support this economic analysis, OPPT developed a proposed approach for estimating lead
exposures and resulting changes in IQ for children under age six that could result from various
RRP activities conducted in all of the residential houses and child-occupied facilities required for
the economic analysis. As noted above, OPPT's Lead Dust Study provided the input exposure
data for use in the estimation of IQ changes in children exposed to lead during renovation
activities.
The Agency authors are to be commended for the significant effort that went into the
document to define and describe the conceptual elements linking Rule and non-Rule LRRP
protocols to established biomarkers of lead exposure and established toxic endpoints from these
exposures. In addition, the general, three-step approach described in this draft methodology
document — (1) estimating the dust lead generated from specific renovation activities and
converting the dust lead loadings to dust lead concentrations; (2) estimating blood lead levels
from exposure to the dust lead concentrations; and (3) from those values, estimating IQ changes
in exposed children — is both logical and reasonable.
However, among the CAS AC Panel's major concerns is that the methodology, in its
current form, is not adequate for the main objective of the approach., i.e., cost/benefit analyses.
These concerns are delineated as follows by the three steps of the approach:
• Step 1: Significantly, for the first step in this approach, there are insufficient data to
determine if the results of the limited dust studies and measurements conducted by
Battelle are representative of a national sampling of renovations of either homes or child-
occupied facilities. As mentioned previously, this limitation is a critical issue that EPA
should address either by quantifying the extent to which this factor and others may result
in either the overestimation or underestimation of risk. In addition, the data on the
effectiveness of lead cleanup in the Dust Study is also limited, and yet this information is
essential for determining lead exposures associated with the renovation procedures.
Thus, while the document does a good job of noting its limitations, there are no
suggestions of how to improve on or eliminate those limitations.
• Step 2: For the step in the approach that converts exposure to dust lead to blood lead
concentrations, two biokinetic models were proposed. The Panel discussed at length the
advantages and limitations of the IEUBK model and the Leggett model for use in
-------
predicting the blood levels of children exposed during renovation activities (see, in
particular, the response of Panel member Mr. Sean Hays to Charge Question #3 found on
pages D-22 & D-23). The IEUBK model was considered inadequate because it is based
on steady-state conditions for exposure to lead — not episodic, peak exposures as would
be encountered in renovation activities, and so would be inappropriate if used in
conjunction with exposure-response data that employ short-term exposure measures. The
Leggett model, which predicts higher blood lead levels than the IEUBK model, allows
for input of data from such acute, peak exposures. The fact that the IEUBK model
produces values closer to the Centers for Disease Control and Prevention (CDC) National
Health and Nutrition Examination Survey (NHANES) data is probably because both the
NHANES data set and the data input for the IEUBK model include homes (and children
living in homes) with no lead paint. It is likely that the upper tail of the NHANES data
would be a better basis for comparison.
The Panel recommends that the Leggett model be used as the biokinetic model of choice
for modeling of childhood lead exposures occurring during LRRP activities, because: (1)
the IEUBK model is inappropriate for acute, short-term lead exposures of children; and
(2) time does not allow substitute use of the potentially-better O'Flaherty model. The
Leggett model has a positive bias to its output (i.e., blood lead) when compared to blood
lead outputs of the IEUBK and O'Flaherty models applied to the same steady-state
exposure scenarios, but the size of that bias in absolute terms is not precisely known,
especially for episodic exposures. Therefore, the uncertainty associated with use of the
Leggett model must be described. The use of the Leggett model also has the advantage
that it is less likely to underestimate blood lead levels and subsequent health risks.
As also mentioned during the CASAC's February 2007 consultation with OPPT, the
Panel strongly favors greater use of empirical data for estimating blood lead levels
following exposure to lead dust, which will aid in evaluating the usefulness of the
biokinetic model predictions. At least one set of published data was suggested for this:
Rabinowitz et a/., Amer. J. Public Hlth, 1985. (In addition, see the individual written
comments of Panel member Dr. Paul Mushak found on pages E-35 to E-41). This dataset
provides quantitative data relating dust lead loadings in houses with children's blood lead
levels and shows that repair and renovation activities result in elevations of these blood
lead levels. The article does not, however, relate dust lead loading during renovation/
construction with changes in blood lead levels. Therefore, this dataset will primarily be
useful for estimating children's steady-state blood lead levels associated with post-
renovation dust lead loading clean-up standards and as a "reality check" for the Leggett
model predictions of the respective blood lead levels. Unfortunately, no quantitative data
are available that relate changes in dust or air lead loadings from repair and renovation
activities with acute changes in children's blood lead levels. Therefore, the Leggett
model will need to be relied upon for predicting transient changes in children's blood
lead levels as a function of dust levels during renovations.
Furthermore, one member of the Panel outlined how such an empirical approach might be
undertaken: (1) Begin with population blood lead levels for a cohort of one- to two-year-
old children from NHANES using the upper tail of the distribution; (2) estimate the
increase in blood lead concentration due to renovation and repair activities using
-------
empirical published values in the range of a 12.5% to 30% increase (Rabinowitz et a/.,
Amer. J. Public Hlth, 1985; Lanphear et a/., unpublished data); (3) estimate the reduction
in IQ associated with a increase in population mean blood lead concentration in this range
using the -2.94 IQ decrement per 1 |j,g/dL from the piecewise linear analysis for children
with blood lead concentration <7.5 ng/dL; (4) determine the range of estimated IQ
benefits of the proposed rule; and (5) use these empirical estimates to calculate the cost-
benefit of the proposed rule for a U. S. birth cohort of one- to two-year-old children. Such
an empirical model could be used in conjunction with the Leggett model. (For additional
detail, see the individual written comments of Panel member Dr. Bruce Lanphear found
on pages E-20&E-21).
• Step 3: The next step described in the draft Approach document was to estimate IQ loss
in children based on their estimated blood levels. The approach was, understandably,
based on the exposure-response characteristics underlying current standards for lead
toxicity, but these characteristics are out-of-date due to recent epidemiological data.
These data extend the dose-response relationship between blood Pb and IQ loss to even
lower Pb lead levels than those underlying the current standards, and, importantly,
demonstrate that children are more sensitive to lead toxicity than previously thought. If
the EPA goal of eliminating lead poisoning in children by the year 2010 is to be
achieved, the toxicity estimates must be based on the most recent data.
The Panel discussed the log-linear model with a cutpoint of 1 |ig/dL versus the piecewise
linear model for estimating IQ change based on blood lead. Arguments can be made for
each of the models (see the responses of Panel members Dr. Robert Goyer and Mr. Sean
Hays to Charge Question #4 found on pages D-20 & D-21 and D-23 & D-24, respective-
ly). Both models fit the existing data well, with a slight edge to the log-linear model for
the entire range of blood Pb levels. However, compared to the piecewise linear model,
the log-linear model underestimates the magnitude of effects on IQ for those children
currently having low blood Pb values (i.e., <7.5 jig Pb/dL) because the slope of the log-
linear model is not as steep as that of the piecewise linear model in this lower range.
Also, the piecewise analysis may be more appropriate because the majority of children
have maximal baseline blood levels below 7.5 |ig/dL, and the mean increase in blood
lead concentration, on a population level, would generally be up to, but not exceeding, a
blood lead concentration of 7.5 |ig/dL. Thus, the majority of the Panel recommends
using the piecewise linear model for estimating potential effects on IQ of household lead
arising from renovation activities. The newly-recognized steep slope for the exposure
response curve below 7.5 |ig/dL (i.e., -2.94 IQ decrements per 1 |ig/dL blood lead)
should be used to estimate the effect of the exposures. The approach should also take
into consideration sensitive subpopulations of children such as those with amino-
levulinate dehydratase (ALAD) polymorphisms or children with either particularly-low
initial blood lead levels or children of low socio-economic status (SES).
Finally, the following significant technical concerns were also noted:
• The sparseness of data is illustrated in the Monte Carlo simulations for uncertainty in
indoor dust lead concentrations. In Appendix D of the draft LRRP Activity IQ-Change
Methodology document, the authors state, with no defense, that low and high values were
-------
assumed to represent two standard deviations above and below the mean. While this
approach is used commonly in risk assessments, we can see no scientific basis for the use
of such assumptions in the current case (or at least no basis is provided in the document),
and so the Monte Carlo simulations are suspect.
• Another assumption — that children occupy the entire house or yard equally during a
renovation project — is simply not realistic. Activity patterns for children as a sensitive
subpopulation should have been included in the analyses. Changing the GSD in the
biokinetic model from 1.2 to 1.6 should, in part, alleviate this problem. However, a GSD
of 2.0 or 2.1 is probably most appropriate given all of the uncertainties in the analyses to
be performed. In addition, a sensitivity analysis can be performed using a GSD in the
range of 1.6 to 2.1 (see the response of Panel member Dr. Fred Miller to Charge Question
#2 found on pages D-27 & D-28). In any event, the best approach would be to develop a
child-specific distribution with central tendency and GSD determined from child-specific
data. There is also a need to consider newer data for deposition of airborne lead dust both
in the respiratory tract and in the head.
• As discussed in the earlier consultation, variability needs to be quantified separately from
uncertainty. In the judgment of the Panel, the sensitivity analyses were performed using
too small an alteration in input variables (i.e., 10%); larger alterations should therefore be
probed. Additionally, there is no indication given as to how the 10% alteration relates to
variability in the data (that is, it is not clear what fraction of a standard deviation is
represented by a 10% change in the input variables).
As with the OPPT Dust Study, additional remarks on the Draft LRRP Activity IQ-
Change Methodology are provided in Panelists' individual written comments (Appendix E).
The CAS AC Panel was pleased to be of service to EPA in its review of these two
documents related to the avoidance of lead exposures during restoration of residential
properties and COFs contaminated with lead. If the Agency wishes additional advice or
recommendations from the CASAC on EPA documents related to the forthcoming LRRP
rule, the Panel would be happy to assist you again in the future. As always, we wish the
Agency well in this important task.
Sincerely,
/Signed/
Dr. Rogene Henderson, Chair
Clean Air Scientific Advisory Committee
Appendices (A-E)
cc: Marcus Peacock, Deputy Administrator
10
-------
NOTICE
This report has been written as part of the activities of the U.S. Environmental
Protection Agency's (EPA) Clean Air Scientific Advisory Committee (CASAC), 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
CASAC 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. CASAC reports are posted on the SAB
Web site at: http://www.epa.gov/sab.
11
-------
Appendix A - Roster of the Clean Air Scientific Advisory Committee
U.S. Environmental Protection Agency
Science Advisory Board (SAB) Staff Office
Clean Air Scientific Advisory Committee (CASAC)
CHAIR
Dr. Rogene Henderson, Scientist Emeritus, Lovelace Respiratory Research Institute,
Albuquerque, NM
MEMBERS
Dr. Ellis Cowling, University Distinguished Professor At-Large, North Carolina State
University, Colleges of Natural Resources and Agriculture and Life Sciences, North Carolina
State University, Raleigh, NC
Dr. James D. Crapo, Professor, Department of Medicine, National Jewish Medical and
Research Center, Denver, CO
Dr. Douglas Crawford-Brown, Director, Carolina Environmental Program; Professor,
Environmental Sciences and Engineering; and Professor, Public Policy, Department of
Environmental Sciences and Engineering, University of North Carolina at Chapel Hill, Chapel
Hill, NC
Mr. Richard L. Poirot, Environmental Analyst, Air Pollution Control Division, Department of
Environmental Conservation, Vermont Agency of Natural Resources, Waterbury, VT
Dr. Armistead (Ted) Russell, Georgia Power Distinguished Professor of Environmental
Engineering, Environmental Engineering Group, School of Civil and Environmental
Engineering, Georgia Institute of Technology, Atlanta, GA
Dr. Frank Speizer, Edward Kass Professor of Medicine, Channing Laboratory, Harvard
Medical School, Boston, MA
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)
A-l
-------
Appendix B - Roster of the CASAC Panel for Review of EPA's LRRP Activities
U.S. Environmental Protection Agency
Science Advisory Board (SAB) Staff Office
Clean Air Scientific Advisory Committee (CASAC)
CASAC Panel for Review of EPA's
Lead Renovation, Repair, and Painting (LRRP) Activities
CHAIR
Dr. Rogene Henderson*, Scientist Emeritus, Lovelace Respiratory Research Institute, Albuquerque, NM
MEMBERS
Dr. Joshua Cohen**, Research Associate Professor of Medicine, Tufts University School of Medicine,
Institute for Clinical Research and Health Policy Studies, Center for the Evaluation of Value and Risk,
Tufts New England Medical Center, Boston, MA
Dr. Deborah Cory-Slechta**, J. Lowell Orbison Distinguished Alumni Professor of Environmental
Medicine, Department of Environmental Medicine, University of Rochester School of Medicine and
Dentistry, Rochester, NY
Dr. Ellis Cowling*, University Distinguished Professor-at-Large, North Carolina State University,
Colleges of Natural Resources and Agriculture and Life Sciences, North Carolina State University,
Raleigh, NC
Dr. James D. Crapo [M.D.]*, Professor, Department of Medicine, National Jewish Medical and
Research Center, Denver, CO
Dr. Douglas Crawford-Brown*, Director, Carolina Environmental Program; Professor, Environmental
Sciences and Engineering; and Professor, Public Policy, Department of Environmental Sciences and
Engineering, University of North Carolina at Chapel Hill, Chapel Hill, NC
Dr. Richard Fenskef, Professor, Department of Environmental and Occupational Health Sciences,
School of Public Health and Community Medicine, University of Washington, Seattle, WA
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 Goodrumf, Senior Scientist I/Manager, ARCADIS BBL, ARCADIS of New York, Inc.,
Syracuse, NY
Dr. Robert Goyer [M.D.]**, Emeritus Professor of Pathology, Faculty of Medicine, University of
Western Ontario (Canada), Chapel Hill, NC
Mr. Sean Hays**, President, Summit Toxicology, Allenspark, CO
B-l
-------
Dr. Bruce Lanphear [M.D.]**, Sloan Professor of Children's Environmental Health, and the Director of
the Cincinnati Children's Environmental Health Center at Cincinnati Children's Hospital Medical Center
and the University of Cincinnati, Cincinnati, OH
Dr. Frederick J. Miller**, Consultant, Gary, NC
Dr. Maria Morandif, Assistant Professor of Environmental Science & Occupational Health, Department
of Environmental Sciences, School of Public Health, University of Texas - Houston Health Science
Center, Houston, TX
Dr. Paul Mushak**, Principal, PB Associates, and Visiting Professor, Albert Einstein College of
Medicine (New York, NY), Durham, NC
Mr. Richard L. Poirot*, Environmental Analyst, Air Pollution Control Division, Department of
Environmental Conservation, Vermont Agency of Natural Resources, Waterbury, VT
Dr. Michael Rabinowitz**, Geochemist, Marine Biological Laboratory, Woods Hole, MA
Dr. Armistead (Ted) Russell*, Georgia Power Distinguished Professor of Environmental Engineering,
Environmental Engineering Group, School of Civil and Environmental Engineering, Georgia Institute of
Technology, Atlanta, GA
Dr. Joel Schwartz**, Professor, Environmental Health, Harvard University School of Public Health,
Boston, MA
Dr. Frank Speizer [M.D.]*, Edward Kass Professor of Medicine, Channing Laboratory, Harvard
Medical School, Boston, MA
Dr. Ian von Lindern**, Senior Scientist, TerraGraphics Environmental Engineering, Inc., Moscow, ID
Dr. Barbara Zielinska**, Research Professor, Division of Atmospheric Science, Desert Research
Institute, Reno, NV
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)
* Members of the statutory Clean Air Scientific Advisory Committee (CASAC) appointed by the EPA
Administrator
** Members of the CASAC Lead Review Panel
•f Members of the Science Advisory Board (SAB) or SAB panel
B-2
-------
Appendix C - Agency Charge to the CASAC Panel
Charge to the CASAC Panel for the Review of EPA-OPPT's Draft Approach for Estimating
Changes in Children's IQ from Lead Dust Generated During Renovation, Repair, and
Painting in Residences and Child-Occupied Facilities (Draft LRRP Activity IQ-Change
Methodology, June 2007)
1. Overall Approach.
Please comment overall on the Approach and its utility for "building" all of the houses and
child-occupied facilities (COFs) required for the economic analysis. Please comment on the
clarity and transparency of the document.
2. Sensitivity and Monte Carlo Analyses
The approach of this document assumes that variable reduction (reduction of the number of
potentially influential factors carried through the analysis) is carried out following a
sensitivity analysis and that Monte Carlo analyses permit the estimates to account for
magnitude of uncertainty as well as variability.
a. The document describes a sensitivity analysis for each of the two examples. They
suggest which factors are important to describe the features of lead (Pb) exposure. The
examples, however, provide only a sense of the impact on that particular example and not
necessarily for the whole. Please comment on the strengths and weaknesses of the
sensitivity analyses. Please comment on whether the sensitivity analysis using the two
examples is sufficient to characterize the factors that are most important for determining
Pb exposure or should a separate sensitivity analysis be conducted for each of the houses
and COFs that will be "built" for the economic analysis.
b. The document describes Monte Carlo analyses for each of the two examples. Please
comment on the strengths and weaknesses of the Monte Carlo analyses. Please comment
on whether the Monte Carlo analyses using the two examples is sufficient to characterize
the variability in Pb exposures or should a separate Monte Carlo analysis be conducted
for each of the houses and COFs that will be "built" for the economic analysis.
c. Dust study results that are observed to be non-monotonic across increasing Control
Options will likely translate into similar patterns following application of the approach to
estimate IQ changes. IQ change models only use geometric means from the Dust Study.
Please comment on the usefulness of an additional Monte Carlo step as the way to
account for the variances in the Dust Study.
d. The blood Pb models assume that variability in the population around any mean blood Pb
is approximately that displayed in the general population. The assumption that the
estimated mean blood lead values are accompanied by geometric standard deviations of
1.2 is made explicit in the IEUBK model documentation and is extended implicitly in
these analyses for the Leggett model. Nonetheless, the assumption currently is not
C-l
-------
discussed in the description of the Approach given to the CAS AC nor would it be
displayed numerically in any results from its application; the approach shown carries out
all simulations during one phase of analysis. A Monte Carlo step between the application
of the blood lead model and a model for estimating changes in IQ would expand the
characterization of differences between similarly aged children experiencing the same
renovation, repair and painting (RRP) activities. Please comment on the usefulness of an
additional Monte Carlo step between the application of the blood lead model and the IQ
change model as the way to display differences.
e. In addition to the aspects addressed by 2b-2d, the document mentions several ways in
which assumptions have been incorporated into the approach in a deterministic fashion.
Please comment on the strengths, weaknesses, and necessity of introducing additional
Monte Carlo analyses or markedly changing these assumptions, and whether these would
be applied to each of the houses and COFs that will be "built" for the economic analysis.
3. Blood Lead Modeling
The document describes use of the Leggett and IEUBK models for each of the two examples.
Both models are used because exposures to Pb from RRP activities are anticipated to be of
short duration, and fluctuate frequently. In this context, applying the IEUBK to estimate the
impacts of short-term fluctuations in Pb exposure (weekly in this approach) may stretch the
IEUBK to the limits of its temporal resolution. Both models are used in this document to
display the impact of model uncertainty. The two examples presented in this document show
that predictions by the Leggett model are about three times those predicted by the IEUBK.
This is consistent with the findings of Pounds and Leggett (1998) who compared predictions
from the Leggett model with the deterministic predictions of blood Pb levels generated by the
IEUBK model, using the IEUBK default inputs. In addition, the relative difference between
the two models seems to be similar for single and multiple RRP activities. Please comment
on whether both the IEUBK and Leggett models should be used to estimate blood Pb levels
for all of the houses and COFs that will be "built" for the economic analysis.
4. Estimates of IQ Change
This document describes the use of two strategies to address the limitations and uncertainties
associated with the log-linear IQ model. Please comment on the strategies EPA has used to
address limitations and uncertainties. Have these limitations and uncertainties been
accurately and transparently described? These include the use of a log-linear model with a
"cutpoint" of 1 |ig/dL blood Pb and the use of a piecewise linear model. Both models are
drawn from Lanphear et al. (2005). The coefficient for the piecewise linear is derived from
concurrent blood Pb levels. Both models, however, are being used with lifetime average
blood Pb values in the context of this document. Please comment on the strengths and
weaknesses of the models. Please comment on whether both the log-linear IQ model and the
piecewise linear model should be used for all of the houses and COFs that will be "built" for
the economic analysis.
C-2
-------
5. Adaptation of Approach for Child-Occupied Facilities
With the range of potential COF configurations, the fact that children may spend most of
their time in a limited part of the COF, and the fact that there may be multiple children under
age six (6) in different rooms of the same COF, there is no simple way to develop a COF-
wide loading estimate. The proposed approach would estimate the Pb loadings in three
different types of rooms in a COF (workspace, adjacent, and rest of COF) by assuming that
all RRP activities take place in the same workspace. It is proposed that loadings in each
room would be estimated for each type of activity individually and then composite loadings
would be estimated for each multiple activity scenario by summing the relevant activity-
specific loadings for each type of room. The estimated loadings for the workspace would
therefore represent the high-end exposure scenario, the rest of COF would represent the low-
end exposure scenario, and the adjacent room would represent the mid exposure scenario.
Please comment on the strengths and weaknesses of the overall approach for COFs.
6. Adaptation of Approach using Age of Housing
The HUD surveys of lead-based paint in housing indicate that the level of lead in paint will
vary by the age of the housing and housing component. The OPPT Dust Study included
houses dating from around 1920 and a school built in 1967. The lead levels in the lead-based
paint in the OPPT Dust Study varied considerably. The Approach uses lead loadings from
the OPPT Dust Study as a proxy for lead loadings in newer houses. Please comment on: (1)
whether it is appropriate to adjust the lead loadings from the OPPT Dust Study downward
based on the age (i.e., vintage) of the building for newer buildings; (2) a suggested approach
for making the adjustment, if recommended; and (3) the application of such an adjustment
for COFs in public or commercial buildings, as well as for residential buildings.
7. Adaptation of Approach for Exterior Renovation, Repair, and Painting
The examples provided in the Approach are for interior renovation jobs. The proposed rule
also addresses exterior renovation, repair, and painting. When the Approach is used to build
the houses for the economics analysis, exterior jobs will be represented. Modifications or
enhancements may be needed to the approach to account for lead exposure from exterior
jobs. In particular, lead dust created by exterior jobs may be tracked into a housing unit or
COF or otherwise enter the unit or COF, and contribute to the indoor dust loading. Please
comment on: (1) the extent to which the approach should consider this "tracked in" dust
contribution to the indoor dust loading of a single property, and provide suggestions for
incorporating it, if recommended; and (2) how to estimate potential lead exposures to
occupants of neighboring dwellings from exterior renovations and for occupants of
neighboring units in multi-family housing from interior renovations.
8. Adaptation of Approach for Other Contributions
The Approach was developed to consider the range of permutations and combinations of
exposure scenarios and houses/COFs that would need to be built for this rulemaking. Please
comment on whether any potential exposure scenarios and/or housing/COF considerations
have been overlooked and should be considered when building the houses for this rule-
C-2
-------
making. Please comment on any additional issues with building houses in which many low
or high dust generating activities are used (e.g., small repairs or power sanding).
Charge to the CAS AC Panel for the Review of EPA-OPPT's Draft Final Report on
Characterization of Dust Lead Levels After Renovation, Repair, and Painting Activities
(OPPT Dust Study, January 2007).
Organization of the Report
Chapter 1 covers background and study objectives. Chapter 2 includes a brief summary of study
conclusions, and addresses the peer review of the study design and the human subjects review.
Chapter 3 summarizes the study design. Chapter 4 summarizes the field work. Chapter 5
contains the statistical analysis plan. Chapters 6 and 7 present the analysis of the data. Chapter 8
summarizes study quality assurance. Chapter 9 presents the study conclusions in detail.
Appendix A provides details on the individual jobs in the study. Appendices B to H contain
plots of study data. Appendices I to O provide more detail on the data analysis in Chapter 7.
Issue 1. Study Objectives
The study was designed to meet several objectives. The study objectives were determined
through consultations with the ultimate users of the data, who would conduct the risk approach
and the economic analysis. The study objectives are listed on pages 1-2 and 1-3 in Chapter 1 of
the report.
Question 1. Are each of the study objectives objectively and transparently addressed in the data
analyses and conclusions in the report?
Issue 2. Study Conclusions
The study conclusions are presented briefly at the beginning of Chapter 2, and in detail in
Chapter 9. The conclusions are based on the analyses of the data that was collected in the study.
Question 2. Is each of the study conclusions in the report supported by the data analyses and
other information in the report? If you do not agree that the conclusions are supported by the data
and analyses, please discuss your concerns and if possible, provide specific language to describe
the conclusions.
Issue 3. Range of Data
Data collected from field studies tend to be variable due to a number of factors. That was the
case in this study.
C-4
-------
Question 3. Do the tables, graphs, figures and other information in the report objectively and
transparently convey the range of the data in the study?
Issue 4. Report Organization and Clarity
The report was written to support rule development, and is likely to be read by persons who do
not necessarily have a technical background, but who are interested in the final rule.
Question 4. Is the report logically laid out, consistent, and easy to follow?
Issue 5. Data Collection and Descriptive Analysis
The study design has been peer reviewed previously. There has been no other external peer
review of the data analyses in the report.
Question 5. Are the descriptive analyses in Chapter 6 for interior and exterior jobs appropriate
for the study objectives and the collected data? Have the collection and the descriptive analyses
of the data been objectively and transparently described in Chapters 3, 4 and 6?
Issue 6. Statistical Modeling Results
In addition to the descriptive analysis in Chapter 6, the report includes statistical modeling in
Chapter 7.
Question 6. Please provide any specific comments on the modeling analyses in Chapter 7. Are
the statistical methods appropriately applied to the data? Are the methods objectively and
transparently described?
C-5
-------
Appendix D - Panel Members' Responses to Agency Charge Questions
This appendix contains the preliminary and/or final individual responses to EPA
charge questions on the Agency's Draft Lead Renovation, Repair, and Painting (LRRP)
Activity IQ-Change Methodology and the OPPT Dust Study from those members of the
Clean Air Scientific Advisory Committee Panel for Review of EPA's LRRP Activities
who submitted such comments. These written responses do not represent the views of
the CAS AC, the EPA Science Advisory Board, or the EPA itself. The names of the
panelists who provided written responses to Agency charge questions comments are
listed on the next page, and their individual responses follow.
D-l
-------
Panelist Page #
Dr. James Crapo D-3
Dr. Douglas Crawford-Brown D-5
Dr. Richard Fenske D-13
Dr. Bruce Fowler D-16
Dr. Philip Goodrum D-17
Dr. Robert Goyer D-20
Mr. Sean Hays D-22
Dr. Bruce Lanphear D-25
Dr. Frederick J. Miller D-27
Dr. Ian von Lindern D-29
Dr. Barbara Zielinska D-38
D-2
-------
Dr. James Crapo
James D. Crapo, M.D.
Comments: An Approach for Estimating Children's Health Risk (IQ) Changes Associated
with Dust Lead Generated by RRP Activities
Charge Question #1: Overall approach and utility
The approach used by the EPA for estimating IQ change from LRRP activity is appropriate, and
the work done to date is important in identifying how these models could and should be used as
well as the inputs that are necessary to support these models. Unfortunately, the draft approach
for estimating changes in children's IQ from LRRP activities has sufficient problems with lack
of adequate data for model input and a lack of model validation that it cannot be reliably used at
the present time to estimate health or economic cost-benefits and should not be used to determine
national regulatory standards.
There is insufficient data for input into the biokinetic blood lead level models, and work
validating the models is completely absent. This approach cannot be reliably carried out and
completed at the current time. The OPPT Dust Study provides important data to address
questions regarding the extent of lead dust generated from LRRP activities and provides some of
the important activity-required inputs for the IEUBK and LEGGETT models for estimating
blood lead levels. Unfortunately, substantial additional work is necessary to provide the
necessary inputs to these models and to validate the model results. This work needs to be
completed before either of these models are used for cost-benefit analyses and the results
considered in the development of national regulatory standards.
A substantial amount of data is currently available to guide regulatory decision making and
which can provide an adequate base for regulatory decision making. What is known is that:
• Lead exposure results in elevated blood lead levels with children being at primary risk
because of their activity patterns.
• Elevated blood levels effect neurocognitive development including IQ.
• The effects of blood lead levels on neurocognitive development are particularly strong
below levels of 10 ug/dL.
• All lead exposures contribute to the blood lead level.
• Room floor dust is a substantial factor in determining blood lead levels in children with the
key factor being lead loading or ug Pb/ft2.
• Renovation, repair and painting activities in older homes can lead to substantial increases
in house (floor and window sill) lead levels post renovation and can mobilize lead into
various media including air, flood, house dust and soil.
• Standard cleaning of older residences post RRP activities can result in high residual lead
levels in the home (elevated 5- to 6-fold according to the OPPT Dust Study).
D-3
-------
• Enhanced protection and cleaning as studied in the OPPT Dust Study can substantially
reduce the risk for lead exposures in homes post activities. (Floor lead levels are decreased
up to 80% following cleaning under the proposed rule change compared to standard
cleaning.)
The above data and the supporting documents in the OPPT Dust Study provide adequate support
for rule making to reduce lead exposure post renovation, repair and painting activities in older
homes. I would recommend that the EPA support additional research activities to enhance and
validate biokinetic models of blood lead levels that can be used for future analysis of the
effectiveness of interventions to reduce lead exposures to children. Critical studies that would
enable the application and validation of these models include:
1. Measurement of blood lead levels before and after RRP activities. This should be done in
children and adults living in the home and should include assessment of the impact of lead
exposure on workers carrying out RRP activities in older homes. Both short- and long-term
follow-up should be included.
2. Measurement of home (floor and window sill) lead levels before and after RRP activities
in older homes with follow-up for multiple years to determine the duration of potential
lead exposures following RRP activities. These measures should be done with and without
the proposed rule change in lead prevention and cleaning standards.
3. The impact of the proposed rule change on lead prevention and cleaning in homes
undergoing RRP activities should be assessed on homes of differing age ranging from pre-
1920 construction to post-1970 construction.
The IEUBK and LEGGETT models should be validated based on their ability to accurately
predict measured blood lead levels in relationship to the RRP activities described above. Only
after validation should these models be generalized for cost-benefit analyses.
D-4
-------
Dr. Douglas Crawford-Brown
Doug Crawford-Brown (6-23-07)
Addressing Question 6: Concerning the potential correction for the fact that homes sampled
were all 1920s vintage.
The issue arises here because (1) the lead content of paint has changed historically, with
reductions in 1950 and then again in 1978, (2) all homes in the OPPT study were of 1920 vintage
and, hence, were first painted when the controls on lead in house paint were not in place and (3)
the EPA wishes to extend the results of the Dust Study to a national sample of homes, which
would include homes built between the 1920s and today.
The proposed solution in the IQ Change document is to use the results of the OPPT study (on
1920s vintage homes) for all homes in the national sample. The justification given is first that
data necessary to make a correction for newer homes are not available. The second justification
is that the distribution of lead in the paint chip samples from the OPPT study homes was wide
(although it was only a GSD of 1.5 to 2, which is not a particularly large dispersion). I suppose
the argument is that this wide distribution would in some way "cover" the post-1920s homes,
including post-1950 homes.
Let me begin by saying that I don't know the actual distribution of lead content in paint sold
prior to 1950, between 1950 and 1978, or after 1978.1 assume after 1978 the lead content is
negligible, and in any event these homes don't seem to be the focus of the proposed rule anyway.
But it is safe to say that paint produced between 1950 and 1978 would contain a lower lead
content than paint produced before 1950, and certainly the paint first used in the OPPT homes.
So, we should have a curve (Figure 1 below) that looks something like this (and I am not
drawing this to scale, since I don't know the values):
D-5
-------
So, if this curve is correct in some sense, we might imagine that homes painted entirely with pre-
19508 paint and homes painted entirely with 1950-1978 paint might have some sort of
distributions of lead levels in paint as follows (Figure 2 below):
PDF
for
Lead
level in
paint
Lead content in paint
But homes in the OPPT sample would have multiple layers of paint that accumulated over the
years, including paint from all three periods of time (pre-1950; 1950-1978; post 1978). So in a
sense, the paint scraped from the OPPT homes would have an average lead content represented
by the average of the layers of paint that had accumulated. The only effect this would have
would be to "push" the Pre-1950 curve to the left. So the argument of the EPA (the second
justification I mentioned) is that this existence of multiple layers of paint would produce
something like this (Figure 3 below):
PDF
for
Lead
level in
paint
Average lead content in paint layers
D-6
-------
This does seem qualitatively to make some sense. Whether it does quantitatively depends on
three factors:
1. How far apart the two distributions in Figure 2 lie (their degree of overlap).
2. How many paint layers have been applied to the OPPT homes
3. The timing of those paint layers
We should be able to state with confidence that use of the OPPT study for all homes will tend to
overestimate risk from the activities, which I suppose could count as a kind of conservatism
(even if I am not a believer that conservatism necessarily leads to optimal protection of public
health). But it also will then overstate (for the same reasons) the benefits of the rule practices in
post-1950 homes (i.e. overstate the benefit-cost ratio).
We can be a bit more mathematical about this (which doesn't necessarily mean more quantitative
unless we can get the needed parameter values), in the following way. Let's assume there is
some time, L years, between paint layers, and that L does not depend on when the home was
built. Let's further divide the time at which a home was built, TB, into three periods:
TBlispre 1950
TB2is 1950 to 1978
TBS is post 1978
TB1, TB2 and TBS are the date, in years (e.g. 1925)
During these three periods, based on Figure 1 above, the lead content is:
Cl inTBl
C2 in TB2
C3 in TBS
Finally, let TR be the time of the renovation (e.g. 2007).
Now, any given home will have some number of layers of paint on the walls. We will assume
each layer of paint uses the same amount of paint. The fraction of the layers at Cl, C2 and C3
will be proportional to the fraction of the total time period between TB and TR represented by
each time period. Since L is assumed constant through the three periods, it will drop out of the
math here.
Let's let Caverage be the average lead content in the wall paint, averaged across the layers. There
are now three cases:
1. If TB falls in TBS, then
Caverage = C3
D-7
-------
2. IfTBfallsinTB2, then
Caverage = [C3 x (TR-1978) + C2 x (1978-TB)] / (TR-TB)
2. If TB falls in TB1 (as in the OPPT study) then
Caverage = [C3 x (TR-1978) + C2 x (1978-1950) + Cl x (1950-TB)] / (TR-TB)
I made a simple EXCEL sheet to calculate these three values. I will assume that C3 is negligible,
so it is 1 (that can be corrected in the attached EXCEL sheet). The figure below shows how
Caverage changes with time if C2 is 10 and Cl is 100.
Caverage (Y axis) and TB (x axis) for values of 1, 10 and 100
40.0
35.0
1900 1920 1940 1960 1980
Year built
2000
2020
The next figure below shows how Caverage changes with time if C2 is 5 and Cl is 20.
Caverage (Y axis) and TB (x axis) for values of 1, 5 and 20
10.0
8.0
2.0
0.0
1900 1920 1940 1960 1980
Year built
2000
2020
From these two figures, it seems clear to me that there can be a rather large overestimate of paint
lead levels using only the OPPT values for all ages of homes. From this analysis, I would guess
that the PDFs for the different periods of homes are closer to those of Figure 1 than Figure 3:
D-8
-------
PDF
for
Lead
level in
paint
Lead content in paint
I recommend that the EPA find reasonable values of Cl, C2 and C3; place these into the attached
spreadsheet (or equivalent); and apply the resulting correction factors. Absent this, some mention
must be made of the likely magnitude of the over-estimate of the risk reduction benefits for
1950-1978 homes and post 1978 homes.
We can discuss these issues further at the July meeting.
D-9
-------
Spreadsheet for the lead content of in of different vintage.
TR =
C1 =
C2 =
C3 =
2W7
in
JB is the build year
The units on C are arbitrary here, since we are interested in relative values
For hom.es where 1B > 1978, Gavetage =
For homes where 1850 < TB < 197B, Caverage =
1978
1877
»75
1870
1961
1958
1657
1956
1954
IQS3
1952
1.0D
1.13
1.28
1.58
1.78
1.96
2.03
2.10
2.17
2.24
2.3S
2.42
2.4S
2.53
2.58
2.63
2 73
2.81
2.89
Composite:
TB
23 37
23 y.
IDDf
1304
20D3
2332
233-
2330
D-10
-------
For homes where TB < 1860, Caverage =
TB
use
-94;
•£44
'£43
•£4C
•933
"532
•63C
•?2S
'£25
'£25
'524
'£22
•620
Coverage
2,?e
3.2e
3.54
3.. 32
-t.uc
•1.34
-.59
•1.53
?,oe
?.2£
5.5'
£.72
f.M
8 '3
f.32
e.5:
e.7D
e.c£
7. Of
7.22
7.3&
7.5f
7 7^
7.ie
9.0*
a.:e
S.3C
8.44
8.5E
3.71
s.a«
•33*
•33=
"35"
:&?«:
*«55
•&M
-.332
'992
*SS'
'335
*di="
•3S5
*3i7
'555
IJ35
•954
•63C-
•31
"55'
•S35
•S"?
'.9~:
"», •
:fl"5
•9"5
•374
*9 ';
•§";
'3; *
•970
^3:*
'35:
•§?7
•&36
•33?
•534
•»33
•331
Ifl
U
1.0
1.0
1.0
10
1.0
1,0
1,0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.1
1.3
1.4
1.5
1.8
1.7
1,1
1,8
1.0
2.0
2.1
2,2
2,2
2,3
2,4
2.4
D-ll
-------
10.0
9.0
Caverage- (Y axis) and TB (x axis) for values, of 1, 5 and ',
1800
1920
1940
1980
2000
2020
Year built
1B61
*550
'B5?
"35!
•354
'353
'651
'95'
•35C
•94:
34E
•347
T45
•345
•344
•34C-
•94;
•34-
'S4Q
'93=
'3:4
2.5
2.5
2.8
2.8
2.7
2.7
2.8
2.8
2.B
2.B
2.B
3.D
3.0
3.3
3.5
3.8
4.1
4.3
4.8
4.8
5.1
5.3
5.5
5.7
5.8
6.1
6.3
6.5
7.1
7.2
7.4
7.B
7,7
925
•324
32?
8.0
8.2
8.3
8.4
8.6
8.7
D-12
-------
Dr. Richard Fenske
"Draft Final Report on Characterization of Dust Lead Levels after Renovation,
Repair and Painting Activities"
Response to Charge Questions
July 9, 2007
Richard Fenske
University of Washington
Question #1. Are each of the study objectives objectively and transparently addressed in the data
analyses and conclusions in the report?
The report provides six objectives, as well as a seventh objective ("Objective X"). The objectives
are presented in an unusual format, as they are posed as interrogatives rather than as specific
aims. It would be helpful if the objectives were stated as declaratives rather than interrogatives.
Several of the objectives contain multiple questions (Objectives 2, 3 and 4). Each question
should be restated as a specific aim.
Objective "X" is a bit mysterious. The report states that there are six study objectives, but it
seems that an additional objective has been added. Please make objective X a numbered specific
aim, or omit.
The first objective is focused on the effect of low-, medium-, and high-level RRP on post-work
and post-verification dust-lead levels. This objective is not addressed directly in Section 9 of the
report. The conclusions presented in Section 9.1 appear to address the mysterious objective X.
Please revise the report so that the conclusions address the objectives clearly and sequentially.
The second objective is focused on the ability of heavy-duty polyethelyne sheeting to reduce
lead levels. This objective is addressed directly in section 9.2. In that section it is stated that
"Figure 6-3 indicates that use of plastic did not consistently result in lower geometric mean work
room floor lead levels across job types at the post-cleaning phase." While this statement is
generally true, no statistical results are included in Figure 6-3, so it is hard to understand the
extent to which plastic sheeting did or did not reduce lead levels. The geometric means are
presented, but the geometric standard deviations are omitted. It might be more helpful to plot the
results as distributions, using percentiles; e.g., lead levels would be the x-axis and number of
samples would be the y-axis.
The third objective focuses on a comparison between the proposed rule cleaning method (HEPA
vacuum and wet mopping with cleaning solution) and the baseline cleaning methods. Section 9.3
of the report addresses this issue directly. The report states that the "differences in cleaning
method is [sic] focused on the Work room, as little impact of cleaning method was observed in
either of the two non-work rooms." The reader is referred to a table in Appendix C (Table
C2.7a). However, this table does not provide a statistical test of geometric mean differences. The
D-13
-------
reader is then referred to Figure 6-4 and 6-9, but no statistical analysis is provided for either of
these figures. Jumping from the main report to appendices and back again is quite difficult for
the reader. It is recommended that that final report present all of the key findings within the
main body of the report.
The fourth objective focuses on lead dust migration from the Work room to adjacent rooms.
This objective raises three separate issues related to dust migration: the effect of different levels
of RRP work, the use and non-use of plastic, and the use and non-use of proposed cleaning
methods. This objective is not addressed directly in Section 9.
The fifth objective focuses on the use of plastic ground coverings to reduce the amount of lead
falling on the ground during exterior work. Section 9 does not provide a succinct response to this
issue.
The sixth objective focuses on the lead levels remaining after the two steps of the cleaning
verification process; i.e., the wet cloth step and the dry cloth step. This objective is not addressed
directly in Section 9.
Question #2. Are each of the study conclusions in the report supported by the data analyses and
other information in the report? If you do not agree that the conclusions are supported by the
data and analyses, please discuss your concerns and if possible, provide specific language to
describe the conclusions.
Yes, but the data could be presented in a more reader-friendly format. For example, Tables 9-1
and 9-2 provide no results of statistical analyses. It should not be necessary for the reader to
consult the appendices to review key information.
Question #3. Do the tables, graphs, figures and other information in the report objectively and
transparently convey the range of data in the study?
No. The graphs present mean values, but do not include variability; e.g. standard deviations,
geometric standard deviations, or 95% confidence intervals. It is recommended that the key
tables be presented within the report, rather than in the appendices.
Question #4. Is the report logically laid out, consistent and easy to follow?
In general, yes. But there is a disconnect between the objectives and the conclusions. It is
recommended that the objectives and conclusions be linked in a direct manner, and that key
information currently relegated to the appendices be included in the main body of the report.
This recommendation holds true for all sections of the report.
Question #5. Are the descriptive analyses in Chapter 6 for interior and exterior jobs
appropriate for the study objectives and the collected data? Have the data collection and
the descriptive analyses of the data been objectively and transparently described in
Chapters 3, 4, and 6?
D-14
-------
Figure 6-1 is a scatter plot of disturbed area and average post-work workroom floor lead loading.
It seems that this figure should be revised such that disturbed area is the predictor variable (x-
axis) and lead loading is the outcome variable (y-axis). The explanation for the change in pre-
work goals for lead levels (Section 6.3) is not satisfactory. The 4-fold increase in pre-work lead
dust level goals needs further justification.
Again we have the situation in which the reader is asked to consult an appendix (appendix G) in
order to verify assertions that are presented in the main body of the report. Key findings should
be presented in the main report.
Figure 6-2 does a poor job of illustrating differences across jobs. The use of a log-scale for the y-
axis is not discussed. The use of this scale makes it hard to see differences. It would probably be
more effective to present these data in a table. Statistical differences for the different jobs are not
provided.
There is a lack of statistical analysis in Section 6. In most cases the reader is referred to the
appendices. The key findings should be documented in the main body of the report.
Question #6. Please provide any specific comments on the modeling analyses in Chapter 7. Are
the statistical methods appropriately applied to the data? Are the methods objectively and
transparently described?
It would be more effective to consign the modeling analyses to an appendix, and to bring more of
the empirical findings and statistical analyses into the body of the main report.
D-15
-------
Dr. Bruce Fowler
Bruce Fowler
Pre-meeting Comments on Charge Question #3
The primary focus of Charge Question #3 is whether to use both the IEUBK and Leggett
models for predicting blood leads in children living in or around housing stock that is undergoing
lead renovation, repair and / or painting. It seems to me that this would be the pragmatic thing to
do because if EPA chose to use one model and not the other there would likely be many
questions and criticisms. I also think it best to use both models so that future decisions regarding
relationships between lead in dust and soil and predicted blood lead values will have the benefit
of a complete perspective.
In addition, I would like to again point out that blood lead values represent an established
surrogate for a target dose of lead to sensitive organ systems such as the brain, kidneys and the
hematopoietic organ system. It is also known that there are subgroups within the general
population (on the basis of ALAD polymorphisms for example) who may react in a more
sensitive manner to a given blood lead value than others. It is my hope that as the agency evolves
its LRRP Rule that it will take into account these sensitive subpopulations and consider the issue
of predicted blood lead values by whatever models in this context. It is my view that such an
approach will help the EPA in the long -term to provide more precise risk assessment values in
this important area by taking advantage of the most modern available science.
D-16
-------
Dr. Philip Goodrum
CAS AC Consultation on "Draft Final Report on Characterization of Dust Lead Levels after
Renovation, Repair and Painting Activities"
Philip Goodrum, Ph.D.
ARCADIS BBL, Syracuse, NY
Philip.Goodrum@arcadis-us.com
General Comments
Overall the report is well written and provides enough detail to support the study conclusions.
My comments focus on areas in which additional information could be added to improve the
readability. Also, some shortcuts were taken in the statistical analysis; a more rigorous data
exploration and analysis would strengthen the overall presentation, but I suspect the overall
conclusions would not change.
This study focuses on the relationship between lead content on interior surfaces prior to RRP
activity, and lead content on floor and sill surfaces within 1 hour of the activity as well as after
cleaning. The study does not address the difference between lead loading and dust loading. In
terms of compliance with standards, the choice to measure dust lead loading makes sense.
Likewise, exposure models will generally require information on contact rates (square feet per
time) and lead content per square feet. However, to better understand and explain differences in
dust lead loadings associated with different experimental conditions, as well as to generalize the
results to other jobs, it would be informative to present information on dust loading. A
moderate correlation was observed between lead concentrations in paint and post-cleaning lead
loadings on floor and sill surfaces. To what extent is the variability in lead loadings due to
variability in dust loading? This component of the variability cannot be quantified without
including dust loading in the statistical models. Presumably, containment and cleaning practices
that are most effective at reducing dust will, in turn, be most effective at reducing dust lead
loading.
Recommend including a definition for dust lead loading in the glossary.
The analysis focuses on data collected at discrete points throughout numerous RRP experiments.
A figure should be presented early in the introduction of the study design to more clearly
delineate a generic timeline of sampling activities relative to stages of the experiment. Points on
the timeline would include stages of the experiment: cleaning, RRP activity, cleaning, and
verification. Sampling events would include: pre-experiment test, air sampling during RRP
activity, air sampling after RRP activity, dust sampling after RRP activity, dust sampling after
cleaning, dust sampling for verification.
D-17
-------
RESPONSES TO CHARGE QUESTIONS
Issue 1: Study Objectives
Q. 1. Are each of the study objectives presented objectively and transparently?
Yes. Although I would recommend adding a paragraph to explain that the study is not designed
to investigate the relationship between experimental conditions and dust loading (see general
comment above).
Issue 2: Study Conclusions
Q. 2. Are each of the study conclusions supported by data?
Yes, although the Chapter 2 summary is reduced to one paragraph each for interior and exterior
jobs. It may be helpful to expand the summary to include bullets of the main points from
Chapters 6 and 9. Specifically address factors that have a primary influence on dust lead loading
- is it the conditions of units prior to RRP activity (e.g., paint lead concentration, surface area
remediated), the level of dust generated by a job, or the containment and cleaning practices?
Issue 3. Range of Data
Q. 3. Do the tables, graphs...in the report objectively and transparently convey the range of
data...?
No. Most comparisons of subsets of data focus on differences in the geometric mean (GM). For
example, almost all figures in Chapter 6 present side-by-side bar charts of GMs. But there are no
tables in main text or supporting appendices that convey the corresponding variability. For
example, Figure 6-3 compares post-cleaning work room floor lead levels by job with and without
plastic use. None of the box-and-whisker plots in Appendix C illustrate the variability for the
same subsets of data. Same observation for Figure 6-4 (post-cleaning work room floor lead
levels by job and cleaning method (proposed rule vs. baseline). Information should be presented
in summary tables in Chapter 6 (or an appendix) to include both the GM and measures of
variability (e.g., standard deviation, interquartile range, min/max).
Issue 4. Report Organization and Clarity
Q. 4. Is the report logically laid out, consistent and easy to follow?
Yes. There is a lot of information to present, and good choices were made to move many of the
supporting graphs and tables to the appendices. Summary bullets are helpful. Table of contents
should be proofread - Section 9 TOC does not match titles of sections/sub-sections in text.
D-18
-------
Issue 5. Data Collection and Descriptive Analysis
Q. 5. Are the descriptive analyses in Chapter 6...appropriate? Have the data collection and the
descriptive analyses...been adequately described...in Chapters 3, 4, and 6?
Yes, although implications of some simplifying assumptions need to be investigated.
1. Goodness-of-Fit (GOF) testing. Section 5.2 indicates that data were evaluated for both
normal & lognormal distribution using the Shapiro-Wilks test. Appendix B (p. B-l)
indicates the statistical significance of GOF is attributed to p-value < 0.001. Text should
introduce how the GOF testing was used (to support log-transformation; to evaluate
underlying assumptions of subsequent statistical analysis, etc). Should note that p-value
of 0.001 is much lower than standard p < 0.05, effectively relaxing the test (will less
often reject the null hypothesis that data are normal/1 ognormally distributed).
2. Units. All tables and graphics in appendices should include units; graphics should
include x-y axis titles. Particularly confusing was Table Cl.l which appears to provide
information on loadings (|ig/ft2) when in fact it is sample sizes.
Issue 6. Statistical Modeling Results
Please provide any specific comments on the modeling analyses.
The R2 of regressions can be strongly influenced by including non-detects. Recommend
excluding NDs, or conducting a separate analysis on detects only, especially for regressions that
involve comparisons to clearance levels (e.g., Figures C3.2a - C.3.2c; Figure F3.2a; Figures
G1.12andG3.12)
D-19
-------
Dr. Robert Goyer
July 3, 2007
From: Robert A. Goyer, M.D.
Response to Charge Question 4. Estimate oflQ change, draft document "An Approach for
Estimating changes in Children's IQ from Lead Dust During Renovation, Repair, and
Painting (LRRP) Rule —.
Chapter 4 provides an overview of the uncertainties and limitations associated with the log-linear
IQ model. The discussion is logical and transparent.
Choice of Blood Pb metric
With regard to the choice of blood-Pb metric, reports in the literature do suggest that blood -Pb
concurrent to IQ testing may be a better indicator of future IQ than lifetime average but I do
agree with the decision for the purpose of the LRRP rule that lifetime, thru 6years, takes into
account the renovation exposure activity of children of all ages avoiding problem of high
concurrent levels of children with renovation activity ages 5 or 6 years old.
Input data and Assumptions
The limitations of small sample size variation in cleaning efficiencies will contribute to data
uncertainty as discussed. However, the most difficult uncertainty to account for will be the
amount of time children spend in different locations in buildings and time out of doors and
variation in activities. Both of these uncertainties are likely to be significant and vary
considerably from child to child. Maybe the variations will cancel out reducing the uncertainty.
Choice of biokinetic models
As discussed there are a number of uncertainties common to both the IEUBK and the Leggett
models but the fact that the IEUBK model is more consistent than the Leggett model, I tend to
favor selection of the IEUBK model. But I really do not know which to choose. A problem for
me is that the results of the two models sometimes differ greatly.
Choice of Strategies to Overcome Limitations of the Log-linear Model:
The two strategies proposed to overcome limitations of the log-linear model are the application
of a 1 ng/dl cutpoint or the use of a piecewise linear model. It was helpful to include the
examples comparing results of the two models. The 1 |ig/dl cutoff model not allowing estimates
of IQ loss with blood Pb levels below 1 |ig/dl seems most appropriate because there were few B-
Pb measures below the 1 jig cutpoint and the log-linear model goes to infinity at zero blood Pb
which is not realistic.
D-20
-------
Which model is most realistic?
The 1 |ig/dl cutoff model seems most realistic.
The demonstration exhibits show that the piecewise model results generally in smaller IQ loss
per ng/dl exposure than the log cutoff model. Exhibit 6-10 shows that the piece-wise model
results in lower IQ deficits in newer vintage homes which seems realistic to me.
Should both models be used for all homes to be built for economic analysis?
No, that doesn't seem practical. Any later decisions based on the economic analysis would have
the problem of deciding which result to use.
D-21
-------
Mr. Sean Hays
Comments on "An Approach for Estimating Changes in Children's IQ from Lead Dust
Generated during Renovation, Repair, and Painting in Residences and Child-Occupied
Facilities"
Prepared by:
Sean Hays
Summit Toxicology
165 Valley Rd.
Lyons, CO 80540
July 10, 2007
Charge Question # 3: Blood Lead Modeling
1) The IEUBK model is not scientifically valid for the application intended for this LRRP
rule.
a. IEUBK is only valid for steady-state exposures (see White et al., 1998). "The
[IEUBK] model was designed for applications where there are long periods of
relatively steady exposure, not to acute or relatively rapid subchronic exposures"
quoted from: The Conceptual Structure of the Integrated Exposure Uptake
Biokinetic Model for Lead in Children (1998) Paul D. White, Patricia Van
Leeuwen, Barbara D. Davis, Mark Maddaloni, Karen A. Hogan, Allan H. Marcus,
and Robert W. Elias, p. 1513 (http://www.ehponline.org/docs/1998/Suppl-6/1513-
1530white/abstract.html).
2) The Leggett model is the only model of the two that is valid for predicting changes in
blood lead levels associated with short-term/acute exposures.
a. Leggett model was not validated against environmental exposure scenarios for
validation of predicting absolute blood lead levels (hence the high estimates for
background)
b. However, the model should be valid for incremental increases in blood lead levels
associated with incremental increases in environmental exposures.
c. I applaud EPA for conducting an analysis of incremental increase in blood lead
levels for the various scenarios. This incremental increase analysis is more valid
and reliable.
3) The O'Flaherty PBPK model should be used along with the Leggett model
a. This will help to make modeling uncertainty more transparent
4) Dose Metrics: Concurrent versus lifetime average (0-6 years of age)
a. Must be paired with appropriate IQ decrement model
b. Currently, it appears EPA has used lifetime (0-6 years) average blood lead
modeling and paired it with Lanphear IQ analysis which relied on concurrent
D-22
-------
blood lead levels. EPA should clarify this? The dose metrics from the lead
modeling and the IQ versus blood lead concentration (PbB) relationship provided
by Lanphear et al. (2005) must match to be scientifically valid.
c. It would provide additional context for risk management purposes if EPA would
provide a plot similar to Exhibit 5-11 but showing the PbB profiles expected for
each of the example control measures.
d. The EPA should consider using incremental increase peak PbB as the dose metric
of concern for the risk assessment rather than lifetime average (0-6 years of age).
Charge Question # 4: Estimates of IQ Change
1) Reliance on the recommended dose metric, incremental increase in peak PbB, requires
the use of a single linear slope describing the relationship between IQ decrement and
PbB. A re-analysis of Lanphear et al. (2005) data should allow the use of a single linear
relationship.
The following examples show the type of analysis which could provide this relationship.
2)
99 -
98 -
97 -
96 -
O 95 -
94 -
93 -
92 -
91 -
Lanphear - Figure 3
-date+ 97,95
R* = Q 88
0 5 10 15 20 25
Concurrent Blood Lead (ug/dL)
30
35
Constant slope allows for a constant
IQ decrement per incremental
increase in blood lead levels. If
blood lead levels < 30 |ig/dl is the
region of relevance,
IQ decrement = Loss of 0.2 IQ
points per 1 ug/dl increase in blood
lead levels
Lanphear- Figure 3 (eliminating >30 ug/dL datapoint)
5 10 15
Concurrent Blood Lead (ug/dL)
20
Constant slope allows for a constant
IQ decrement per incremental
increase in blood lead levels. If
blood lead levels < 20 |ig/dl is the
region of relevance,
IQ decrement = Loss of 0.3 IQ
points per 1 ug/dl increase in
blood lead levels
3) Some additional benefits of relying on incremental increase in blood lead and IQ
decrement
a. Less uncertain for blood lead modeling results
b. Alleviates issues/uncertainties about knowing what region of the blood lead vs. IQ
curve one is at.
c. Alleviates the issues with "cutpoints" for blood lead levels and IQ
d. Likely to substantially simplify cost-benefit analysis
D-23
-------
4) It needs to be considered whether the Lanphear analysis, which relies on observing IQ
and blood lead levels among children mostly at steady-state, is valid for use in predicting
IQ loss following acute/transient increases in blood lead levels. While it may not be
scientifically valid to make this extrapolation, it may be the only available approach.
This should be recognized in the text.
5) There is some consideration of using empirical relationships between dust loading and
blood lead levels in children. However, these types of relationships should only be
considered applicable for the LRRP Rule is the empirical relationships have been
developed in homes undergoing renovations/remodels and paired children's blood lead
levels. These empirical relationships are NOT applicable if they were derived for homes
not undergoing renovations (i.e., both the home's dust lead loadings and children's blood
lead levels were at steady-state).
Charge Question # 8: Adaptation of approach for other contributions
1) I recommend EPA consult with the building/construction/remodeling trade associations
for advice on this issue. They are in a better position to advise on issue related to the
types of construction activities that are most likely to result in dust generation.
D-24
-------
Dr. Bruce Lanphear
Comments on "Draft Final Report on Characterization of Dust Lead Levels after Renovation,
Repair and Painting Activities"
Bruce Lanphear
RESPONSES TO CHARGE QUESTIONS
Issue 1: Study Objectives
Q. 1. Are each of the study objectives presented objectively and transparently?
With two important exceptions, the objectives were presented objectively and
transparently. The first exception was the validation step, which was presumably a stealth
approach to using the white glove technique that was soundly criticized at the February CASAC
meeting. This technique or "verification" is a non-scientific and foolhardy attempt to minimize
the costs of the LRRP. This criticism is consistent with the understated conclusions in the
Report (see p. 9-6, part 9.4), which states:
"The cleaning verification process as stated in the proposed rule resulted in decreases in
lead levels, but under the conditions of the study was not always accurate in identifying the
presence of levels above EPA standards for floors and sills. Factors such as floor condition,
contractor performance, job type, and dust particle characteristics impacted the cleaning
verification process in the study."
The second exception is the decision by US EPA staff to ignore scientific evidence
indicating that it is necessary to achieve settled dust lead loading values considerably less than
40 |J,g/ft2 to protect children from lead hazards. If the EPA staff chooses to ignore or deny this
evidence, they should boldly declare this in the introduction of the document and take full
responsibility for denying its existence.
Issue 2: Study Conclusions
Q. 2. Are each of the study conclusions supported by data?
Yes, except for the conclusions about the verification method and the reliance on dust lead
standards described above. Still, the report fails to make several important and directly relevant
conclusions. For example, the study confirms that regulations are needed to protect children
from lead hazards. This conclusion should be prominently described. This conclusion is
justified because the lead hazards generated by repair and renovation using baseline practices
consistently produced dust lead levels in excess of levels shown to be associated with children
having blood lead levels > 10 ng/dL; produced significantly higher dust lead loading values than
the rule practices in the Work Room; led to marginally significantly lower dust lead levels in the
D-25
-------
Observation and Tool Rooms; and they often exceeded US EPA residential dust lead standards.
The proposed rule's procedures consistently - if not always statistically significant - led to
significantly lower post-renovation dust lead loading values.
Issue 3. Range of Data
Q. 3. Do the tables, graphs...in the report objectively and transparently convey the range of
data...?
No. The tables and graphs are overwhelming and overly complex. The titles and footnotes do
not fully or accurately describe what is being presented. It would have been easier to understand
if the report focused on the primary analyses and results, preserving the extensive output for the
appendices. Instead, the reader must wade through the document numerous times to read and
then re-read to verify the information presented.
Issue 4. Report Organization and Clarity
Q. 4. Is the report logically laid out, consistent and easy to follow?
No. The report is overly complex. It would have been easier to understand if the report focused
on the primary analyses and results, preserving the extensive output for the appendices.
Issue 5. Data Collection and Descriptive Analysis
Q. 5. Are the descriptive analyses in Chapter 6...appropriate? Have the data collection and the
descriptive analyses...been adequately described...in Chapters 3, 4, and 6?
Yes.
Issue 6. Statistical Modeling Results
Please provide any specific comments on the modeling analyses...
3. The results of the statistical modeling should be presented more simply. The results
should also explore the proportion of various dust lead loading values achieved by use of plastic,
proposed rule cleaning, and baseline cleaning (e.g., < 10 ng/ft2, < 20 ng/ft2, < 30 |J,g/ft2, and so
on). Achieving floor dust lead loading values of 40 ng/ft2 and sill dust lead loading values of
250 |J,g/ft2 only provides an illusion of safety.
D-26
-------
Dr. Frederick J. Miller
Input on Charge Question 2 for the CASAC Panel letter to the EPA Administrator
2a - As a strength, a sufficient number of appropriate variables were chosen for sensitivity
analyses. However, there are a number of weaknesses in the implementation of the sensitivity
analyses for the two examples studied. Among these were: (1) studying only a 10% change in the
value of input variables, (2) not providing how this 10% change relates to the standard deviations
of the variables, (3) computing elasticity but providing no explanation of the importance in the
range of computed outcomes, and (4) not putting the sensitivity scores into perspective so the
reader could adequately judge which input variables provide the most variability in predicted
outcomes. Given that the two examples studied reflect probably the minimum and the maximum
amount of renovation that will be done, there is no need to conduct separate sensitivity analyses
for each of the houses and COFs that will be "built" for the economic analysis.
2b - The framework for the Monte Carlo analyses that were performed was fine. However, the
outcome of these analyses is of limited value because of deficiencies such as those noted above.
The Panel does not feel that separate Monte Carlo analyses need to be conducted for each of the
houses and COFs that will be "built" for the economic analysis.
2c - There probably is not a need for an additional Monte Carlo analysis as a way to account for
the variances in the dust study. The non-monotonic results that were seen across increasing
Control Options is probably an artifact due to the way non-detectible sample values were
handled and the failure to use statistical methods for left censored distributions. Clearly, the dust
levels should decrease across the increasingly rigorous Cleaning Options. A failure to do so
reflects poor or inadequate data.
2d - The use of a geometric standard deviation (GSD) of 1.2 in both the IEUBK and Leggett
models is inappropriate and has contributed to the sensitivity analyses showing that most input
variables are not important for dust loadings (and subsequently dust concentrations). A GSD
between 1.6 and 2.1 should be used as a value in this range incorporates some of the
uncertainties in estimating mean blood Pb levels. An additional Monte Carlo step between the
application of the blood Pb model and the IQ change model would be quite useful.
2e - Some assumptions bases on deterministic data are imparting variability that is not really of
interest for the project objectives. For example, the dietary Pb intake values for ages 2 to 6 vary
up and down by an amount that is not likely to be biologically significant. Setting the dietary
level to the average across these ages would eliminate this factor, which is not of interest here.
The deterministic value used for lung deposition fraction is outdated and needs to be replaced by
age-specific lung deposition fractions. The current analyses ignore head deposition of particles,
which is going to be a major source of Pb input to the G.I. tract. The use of a GSD of 1.2 in the
blood Pb modeling is not appropriate and needs to be increased. If the type of corrections
mentioned above were made, the assumptions that were incorporated in a deterministic fashion
would probably be fine. As noted earlier, except for the possible use of additional Monte Carlo
analyses for the blood Pb and IQ change modeling, other Monte Carlo modeling beyond the
D-27
-------
types already being done is not necessary nor is it necessary to apply them to each of the houses
and COFs that will be "built" for the economic analysis.
D-28
-------
Dr. Ian von Lindern
Clean Air Scientific Advisory Committee
OPPT RRP Lead Rule Consultation July 9-10, 2007
Post-meeting Response to Charge Questions
Ian H. von Lindern
Charge Question 1 Clarity and Transparency of the Document.
Overview: The problem is well presented from the immediate perspective of the EPA regulatory
need to produce a document to support cost-benefit evaluations. The analyses, strategies, and
systematic approach to the problem are well laid out and supported by appropriate graphs and
Tables. However, the ultimate questions to be answered, and how the Agency and OMB will be
informed by these analyses, are not well defined. It is unclear exactly how the outputs will be
used in decision-making.
The discussions are brief and well edited for the reader looking for findings and conclusions at
the various stages, and the document flows well. However, the details are difficult to follow into
the appendices. Many procedural and data use implementation decisions have to be accepted at
face value, although the appropriate caveats and rationale are usually there, the numbers are not
so easy to access and assess. There is no doubt, however, that it will always be a difficult task to
either present or follow the numbers through a document containing so many mathematical
models in different systems.
The data discussion is straight-forward, concluding that there are few sources available to
support these analyses and none specifically designed to its needs. This perennial complaint of,
at least, some critics of every study, seems to be true in this case. There are few studies and data
to rely on, and as a result, considerable uncertainty in the outcomes.
The model structures are sound and have been used successfully in past regulatory and scientific
activities. The outcomes of each step are, for the most part, appropriately conveyed to the next.
However, the limitations imposed in each step may constrain the analyses in succeeding steps,
and at times, lead to erroneous interpretations. The sensitivity analysis is well conceived, but it
inadequate to the task as presented. The Monte Carlo application add some useful information,
must be presented in manner that does not imply it increases the quantifiable nature of this sparse
data set.
The parameter selection process tends toward simplicity, that is proper, given the reach that must
be extended to develop the estimates necessary to support the models, and even more so, the
Monte Carlo analyses. However, the range of parameter values that should be accommodated in
the blood lead modeling and sensitivity analyses is inadequate.
Overall, the significance of these short-comings is not sufficiently illuminated nor well-enough
echoed in the late stage uncertainty discussions. It is unlikely that an unqualified conclusion and
D-29
-------
quantitative answers can be provided to the economists, given the overall degree of uncertainties.
The committee was obviously uncomfortable with endorsing the scientific validity of these
analyses unless some corrective measures are undertaken, and the results presented in the context
of the significant uncertainties and how those unknowns must be treated in extended analyses.
Risk Assessment Concerns: Despite all the caveats, the greatest problem with this analysis is
that it could lead to an erroneous conclusion and ultimately subject children to significant risk
that might otherwise be avoided. This is in contrast to the current flavor of the document that
seems to suggest these are transient exposures, will result in short-term spikes in blood lead
levels, with little or no significant IQ decrement.
Much of the frustration indicated by the committee, I believe, comes from the committee's
collective understanding that:
1) the levels of lead generated in the RRP activities can cause substantial increases in
household dust levels,
2) exposure to these dusts could result in dangerous spikes in blood lead levels, that put
children at risk of significant IQ loss,
3) extraordinary measures (relative to current state-of-the-art practices) must be
undertaken to minimize these exposures,
4) even the most stringent controls proposed will result in increased exposures,
5) as a result, the rule should advise and promote every reasonable pre-caution,
6) reasonable pre-cautions likely translate to containment of dust and isolation of children
during construction, extensive (frequent and aggressive) and repeated cleaning afterward, and a
reliable clearance test.
The difficulty is that, after all the modeling, this document could be leading to the conclusion
that only small IQ decrements result, and many committee members fear that an economist may
find that the costs associated with implementing these precautions are excessive in relation to the
small IQ decrement. Much of this confusion, in my opinion, is that the presentation of resultant
blood lead and IQ, respectively, as an increment and decrement to baseline can be misleading.
The problem is complex, and the complexities multiply through the series of non-linear systems.
This is difficult to grasp conceptually and describe mathematically. There are problems with lack
of data and uncertainty associated with parameter selections. However, the modeling effort is
comprehensive and most of the components have been peer-reviewed and utilized successfully in
past scientific and regulatory activities. The process and the results should be informative in
assessing the risk and outcomes. Doubtless, some significant aspects and factors are left out, but
the analyses should aid in making a better decision.
D-30
-------
Lead Health Policy Concerns: The analyses of the modeling results, however, neglect two of
EPA's critical mandates and much of what we have learned about lead poisoning over the years.
EPA's mission is to:
1) eliminate lead poisoning by 2010,
2) protect the most vulnerable members of the population, and
Because lead is a potent multimedia toxin, eliminating lead poisoning will only be accomplished
by addressing all sources of lead. Evaluating each source of lead as an increment to all the other
sources could (in the extreme) lead to the conclusion that no individual source is significant
while it is clear that the combination is harmful.
Purposefully, allowing increases in children's exposures is counter to the federal strategy to
eliminate lead poisoning and promote lead-safe housing.
I believe it is because the modeling approach masks the most significant effects by the manner in
which the increments are evaluated.
As I understand the approach, the model assumes a "background" or baseline dust concentration
for each vintage home and then estimates outcome blood lead levels for this "background"
situation. Then an incremental dust exposure associated with the RRP is added to the mean
background dust exposure and a second outcome mean blood lead level is estimated. IQ deficits
are then projected for both these total blood lead levels and the difference is attributed to the
RRP. A GSD of 1.2 is applied to the blood lead means to estimate 5th and 95th percentile blood
leads and the same increment/decrement calculation is performed.
In effect, the model assigns the same baseline blood lead level (and hence IQ deficit) to every
child that lives in a housing strata. The procedure then calculates various distributed blood lead
increments and calculates IQ deficits for children. This effectively is operating only in the upper
portion of the distribution, where it is assumed the most vulnerable children are found, consistent
with the theory that those children with the highest blood lead levels are at greatest risk. As a
result, the largest overall IQ deficits identified are associated with the children with the highest
combined background and RRP increment. However, when the IQ deficit attributed to
"background or baseline" is subtracted out, a smaller decrement is noted.
This process misses another vulnerable segment of the population for these analyses. That is the
child with a low baseline or background attributable blood, (left side of the background
distribution) that is then exposed to high RRP sources (right side of the RRP dust lead
distribution), responds with a high response rate (right side of the individual blood lead dose-
response distribution) at the low end (highest slope) of the IQ/blood lead dose-response curve.
This is the child will that will see large IQ deficits, that otherwise may have been avoided by
controlling the RRP exposure. This child could be teased out with modifications to the sensitivity
analyses.
D-31
-------
Summary: Overall, I believe the analyses are well structured and can, with modifications inform
the decision-makers in their respective activities related to characterizing risk. I do, however,
have reservations regarding the quantitative reliability and the exact nature of the deliverable to
the economists. From a risk assessment perspective there are three critical elements missing from
this document. Those are:
1) presentation of the number of children whose blood lead levels will be elevated above the 10
|ig/dl threshold currently considered unacceptable in all other agency programs (also
currently under serious review as being inadequate to protect children). This critical
consideration seems to be lost in the rush to provide economists with numbers to crunch.
Allied to this is the suspicion that the rule is insufficient to protect children without lowering
the residual floor dust loading to, at least, half the current acceptable value.
2) Identification of, and descriptions of the potential effects on, the susceptible populations.
Which segments of the national population will suffer the greatest harm from not
implementing the rule, and what is the effect at the individual child level?
3) Qualitative discussion of the uncertainties in terms of the error effect on outcomes and
determinations need to substantially augmented. Do these analyses, the critical parameter
selections, assumptions, and model forms lead to false negative or false positive
determinations. At the end of the day are we more likely to make false positive (impose
unnecessary precautions in the rule and waste resources) or false negative (leave children at
risk) conclusions.
Charge Question 2 Sensitivity and Monte Carlo Analyses
Sensitivity Analyses: The sensitivity analysis is well developed, but is applied to a limited
portion of the overall modeling effort and is incomplete in some key areas. These analyses, in my
opinion, are among the most important in the document. There are, as noted through out the
document and the Committee's findings, large uncertainties that result from the limited database.
These uncertainties are compounded through this chain of calculations performed largely with
non-linear models. The sensitivity analyses should be accomplished with both negative and
positive increments in key variables. The increments employed should be larger and in line with
the known or anticipated range in expected values. Runs, at least, should be made at double and
half the typical values; or with larger differences if one and two standard deviations are
substantially greater than the double and half scenarios. It would not be necessary to display all
these results, but it should be discussed whether the model sensitivity breaks down or implodes
with values in the extremes of the distributions.
The current analysis provides some good information with respect to the dust lead concentration
estimations. However, the parameter selections must be extended toward 1 to 2 SDs from the
typical values in the model and the elasticity measurements evaluated accordingly across this
range. The elasticity and sensitivity indices provide insight about the model's behavior and
performance, but it is even more important to use these sensitivity analyses to convey
information regarding the effects of the uncertainties on outcome. For example, what are the
consequences in blood lead and IQ increments and decrements if the dust lead prediction
D-32
-------
equation is off by a factor of two? This is critical information to those who intend to use this
output. This should be followed with a discussion concerning the probability of this occurrence
and whether it leads to overstating or understating the true outcome.
Sensitivity analyses should be conducted on dust, blood lead, and IQ outcomes.
Non-monotonic Results: The lack of monotonic results in the outcomes could trace back to non-
monotonic input from the RRP studies. This well could be due to the data points for different
strata not reflecting mean or median observations within these distributions. That is, not enough
data points were collected in each strata to obtain a consistent typical or central tendency values
for modeling analyses. This is not a fatal flaw, but it must be recognized in the uncertainty
discussions that neither the variance nor an effective mean has been identified.
Monte Carlo Analyses: This is especially true for the Monte Carlo analyses. The reader should
be cautioned that these analyses don't lend to better quantification of the outcomes, but can be
useful in qualitatively assessing the uncertainty in results derived from the limited data base.
There has long been caution offered regarding adopting probabilistic front ends to the IEUBK to
avoid using an empirical overall GSD in evaluating tail effects. Much of this caution relates to
the inability to specify the variance in the input factors with more informed guesses. Multiple
guesses on the front end of the model won't improve on an aggregate guess based on observed
data in the backend.
Similarly there are several instances where no-detects and background, baseline or fixed
variables tend to "freeze" or truncate the left-hand portions of the distributions. These fixed or
informed guesses could have important artifactual effects on subsequent analyses, particularly
with random sampling techniques. These analyses must pay attention to both tails of the
distributions, as the most vulnerable children will be affected by both the low and high end
distributions. For example, the traditional use of IEUBK is to evaluate effects and relationships
at the mean and project the percentage of outcomes in exceeding critical levels in the upper tail.
When considering incremental analyses, the greatest affects may be observed in children moved
from the left tail to the right side of the distribution due to the incremental exposure. This doubly
true when considering the delta IQ, as this child will be responding in the steepest portion of the
IQ/blood lead relationship.
Additional Monte Carlo analyses will not be helpful if it is disengaged from model results. All of
these types of models must appropriately describe how they inform the treatment of uncertainty.
It must be pointed out that without an effective description of variance in the input parameters
the Monte Carlo results are qualitative or, at best, semi-quantitative. The imputed variance must
be discussed is in context of other unknowns and uncertainties and the discussion must be
informative of variance and parameter, model or decision rule uncertainty.
Charge Question 3 Blood Lead Modeling
Choice of Models: The blood lead estimates are well-presented and the contrast in model results
between the two primary models is demonstrative of a lack of consensus regarding the capability
to project children's blood lead levels following short-term exposures. These differences are
D-33
-------
probably not as glaring as this document would leave a regulator to believe. The Leggett model,
as pointed out in the document, is known to predict higher blood lead levels than the IEUBK.
Some of the over-prediction may be to absorption factors for soil and dust. For the most part,
however, the Leggett model is likely showing spikes that are collapsed in the longer-tern
averages captured in the IEUBK. In all likelihood, the appropriate value for use in these analyses
lies between the IEUBK and Leggett estimate. Bracketing the analyses by both these estimates,
likely captures this value.
The best blood lead level to utilize in the IQ model, however, is not clear because little is known
regarding the effect of short-term excursions in blood lead levels. Both concurrent and lifetime
average values have been shown to be related to IQ deficit, but there are questions regarding
which would be better applied in these analyses. In either case, the effect is likely related to some
cumulative measure of elevated blood lead. In this document that cumulative blood lead
excursion is best illustrated in the time-series curves shown in Exhibits 5-10 and 5-11. The area
under these curves shows the combined attenuation in exposure and predicted blood lead level
that leaves small children at risk for significant periods of time, even considering the "spike" in
dust concentration. The greatest concern evident in those figures is the area under the curve for
1-3 year old children, who would be the most vulnerable due to their habits, hygiene and
physiological predisposition to adverse effects. This could lead to the suggestion that the IEUBK
2-year old blood lead estimate be used. The best course of action is probably to conduct the
analyses using the highest and lowest estimates and convey the outcomes and associated
uncertainties to the decision-makers in informed discussion.
Selection of GSD: There are difficulties with the manner in which variance in blood lead
estimates were handled in these analyses. There is a fundamental error in the use of the 1.2 GSD.
This value fails to account for several sources of variance in blood lead distributions. Typically,
most blood lead populations show GSDs in the range of 1.5 to 2.0 and greater. EPA has
recommends 1.6 as a typical value when evaluating lead contaminated hazardous waste sites.
NHANES national strata data indicate values near 2.0. These GSDs are empirical and reflect
several sources of variance in blood lead response, ranging from differences in media
concentrations, accessibility, exposure, intake, absorption, individual blood lead response, bio-
kinetics and SES factors. When large national populations such as NHANES are assessed, all of
these factors show large variation and resultant overall GSDs are generally 2.0, or greater.
Conversely, when some of the exposure and population factors are limited, such as at an
individual hazardous waste site where the primary sources, pathways, media concentrations,
bioavailability, SES factors, etc. are known or constrained, lower GSDs are appropriate.
In this report, the selection of the GSD is not a simple determination. Evaluation of a single
constructed home would suggest a GSD near 1.6. However, to apply the results across the
national population the aggregate GSD should reflect the NHANES findings of near 2.0. Most
probably the 1.6 should be applied and the results of all homes when aggregated should reflect
the increased variance in response noted in NHANES.
Because the GSD is driven to a large extent by the population-wide variance in exposures and
SES, the GSD for this case should be selected to describe the variation in the "background"
population. That population(s) should be the appropriate NHANES stratification relative to the
D-34
-------
housing vintage under consideration. This GSD will likely be much larger than the 1.2 value
used and will result in significantly higher 95th percentile estimates. Because this is also the
population to which the increment must be applied, the same GSD should be used in assessing
the increment.
Consideration of the Most-vulnerable Children: In these analyses, the ultimate need is to
assess an IQ increment associated with RRP exposure. The current approach does calculate a
mean and distribution for the "background" case and then adds a dust exposure increment to
intake, resulting in a blood lead increment. This is accomplished at the mean in both models and
then the extremes of the response distribution are assessed by applying the GSD. However, this
method 1) underestimates the extreme blood lead levels by not using the appropriate GSD, and
2) implicitly estimates the increment by subtracting each point in the RRP distribution from its
respective point in the background distribution. The latter is discussed in response to Charge
Questions 1 and 2. As a result, using a larger GSD and investigating the potential impact for the
most vulnerable child will likely show a much larger potential IQ deficit for those individuals.
There are two distinct populations that could be considered most-vulnerable in these populations.
Those are 1) the children who (under current guidance) would see an increase in blood lead
levels to above 10 |ig/dl, that would not otherwise be at risk of exceeding this criteria, and 2)
those that would experience the greatest IQ deficit, that would have otherwise not have suffered
this result (even if their blood lead level did not exceed the 10 |ig/dl threshold). The former
should be evaluated in this analysis as these are the children that conventional lead health risk
assessment would identify. If large numbers of children are identified as experiencing excess
absorption (>5% under current criteria) by these analyses, then the efficacy of the rule should be
re-examined in light of the findings. Current empirical evidence is suggestive that the rule will be
ineffective in protecting these children, unless the post-activity loading criteria are significantly
reduced. In the case of those children who would suffer significant unintended IQ deficit but
remain below 10 |ig/dl, these should be explored by the sensitivity analysis suggested in
Response to Charge Question 2.
IEUBK Sensitivity Analyses: The IEUBK model clearly has the disadvantage that it is not
designed to estimate short-term blood lead levels and the resultant inapplicability to this problem
must be pointed out. However, this weakness is clearly biased toward under-estimation of risk
and the IEUBK can be used as a reality check for the overall analyses and to assess the potential
of other uncertainties to over- or under-estimate risk. The IEUBK can inform these analyses by
performing "sensitivity" runs to:
1) Aiding in selecting a GSD that describes the population variance as described above.
2) Estimating the percentage of children put a risk of exceeding thelO |ig/dl threshold.
3) Estimating the blood lead increment seen by children at risk for the greatest IQ
decrement (i.e. by performing a "sensitivity " run for children from a low background soil and
dust home that sees a maximum dust lead increase due to RRP.
D-35
-------
4) Examining the blood lead effect of repeated RRP activities over a number of months
or years.
5) Examining the effect of different "background" or other source contributions to blood
lead levels and the "area under the curve."
Charge Question 4 Estimates oflQ Change
Please see the response to Charge Questions 1 and 3 with regard to where in dose-response curve
the IQ decrement is calculated, particularly with respect to the overall decrement that is
attributed to background blood lead levels. See also the discussion in response to Charge
Question 3 under Choice of Models as it pertains to the appropriate blood lead level for
estimating IQ decrement and the "area under the curve" in time series blood lead exhibits.
Charge Question 5 Adaptation of Approach for Child-Occupied Facilities
I believe the application is sufficient provided the approach is consistent with the rule as it will
be applied.
Charge Question 6 Adaptation of Approach using Housing Age
My understanding of this question relates to the observation that the OPPT (Battelle) Study dust
lead loadings (that provides the data to estimate dust concentration estimates for use in blood
lead modeling) were observed from renovation work performed only on older vintage homes.
The concern expressed is that these loadings, and hence the concentrations developed may be
overestimated when used as surrogates for newer vintage homes, This could lead to
overestimating the blood lead increment, IQ decrement, benefit, and cost. Initially, this seems to
be a straight-forward question with a straight-forward answer. There is more lead in older homes,
and likely more lead in dusts generated from RRP activities that is likely reflected in lead
loadings from RRP debris. This would lead to the conclusion that using the older results for
newer homes likely over-estimates risk and should be viewed as a "conservative" or protective
uncertainty factor consideration in decision-making. At the end of the process this would
translate to there being a greater probability of making a false positive (protect children not
needing protection) rather than a false negative (waste money and resources) determination. The
policy risk managers would be so informed. This is my understanding of how the question would
be handled in the current version, although I don't specifically see that discussion in the text. The
issue thus far, was raised only in the charge questions.
However, there is reason to suspect that risk may be overestimated for newer homes because
there is more lead in older homes and background dust loadings and concentrations are higher in
older homes (HUD and others). This is known from XRF surveys and dust lead measurements.
One alternative would be to develop an "adjustment factor" to reduce this uncertainty and try to
incorporate the belief that lower lead loadings result in lowered dust concentrations and down-
the-line outcomes. The uncertainty discussions would then need to reflect the level of knowledge
and evidence supporting the parameter selections, models and assumptions incorporated in the
"adjustment".
D-36
-------
The first question to ask is whether the surrogate levels are inappropriate. There are considerable
uncertainties in these data, even for the older homes. This is a single data base, relying on few
observations, with no corroborating evidence, and little to build meaningful variance estimates.
It's unlikely that the uncertainty can be quantitatively improved by developing additional
adjustments, as high, low and mid estimates need to be developed to support the probabilistic
analyses. However, the understanding of the level of protectiveness in the outcome could be
improved, depending on the reliability of the "adjustment".
In the absence of RRP data generated from newer vintage homes, any adjustment parameters
would likely be developed from the XRF and loading / concentration observational database. It
seems that none of the available data are RRP activity specific, or even related to RRP activities.
As a result, the adjustment would likely be a "scaling factor" related to relative "background" or
XRF loading factors. Whether these are more or less appropriate for each RRP activity is
unknown. I suspect appropriate, but not constant values could be developed for each vintage.
The use of lead paint and the lead content of the paint varied by vintage and component. Such
factors, as demolition of lathe and plaster verse types of wallboard or textured surfaces etc. might
have some affect on bulk dust, and as such, concentration levels.
Most important, however, is recognizing that the variable being constructed for use in predicting
outcomes, is dust concentration. Corrections for vintage are made in the conversions applied in
this document. Moreover, the conversions use the same variables that would likely be used in an
adjustment factor. Statistical analyses of the HUD database showed that the dust concentration
was related to house age, XRF loading and outdoor soil. However, the correction selected was
univariate using only house age. This form subsumes the XRF and soil variables and accounts
for these in the intercept. As such, the adjusted correction may have already been accommodated
in the dust conversion equation.
Additionally, the second Battelle report provided shortly before the meeting dated June 29, 2007
suggests that there is little significant difference in XRF lead loading in paints in various interior
housing components by vintage (for homes that show readings >1.0 mg/cm2). Should these
analyses be accepted in peer review, the adjustment for vintage is likely not needed, provide the
homes exceed the 1.0 mg/cm2 criteria.
Charge Question 7 Adaptation of Approach for Exterior RRP.
Outdoor soils are implicated in both outcome blood lead levels and as contributors to house dust
lead concentrations. As such soils have the potential for longer term residual impact on blood
lead, IQ and residual dust lead levels. The significance of exterior RRP on these soils should be
included in the analyses.
Charge Question 8 Adaptation of Approach for Other Contributions
Please see the discussion regarding children at-risk and incorporate the findings of the sensitivity
analyses in the houses to be built so as to appropriately include these children in the risk
assessment and cost-benefit studies.
D-37
-------
Dr. Barbara Zielinska
Barbara Zielinska
Comments: An Approach for Estimating Children's Health Risk (IQ) Changes Associated
with Dust Lead Generated by RRP Activities
Charge Question #8: Adaptation of approach for other contributions.
I understand that the question is if the methodology developed in this document could be used in
any other exposure scenarios and houses/child occupied facilities (COFs) that it may be
necessary to examine these in a subsequent economic analysis. In my opinion, this document is
very comprehensive and well written and could be adapted to other combinations of exposure
scenarios, providing that certain changes and corrections, as discussed at the CAS AC meeting on
July 9-10, 2007, are made. The goal of this document is not to show the efficiency (or lack of
efficiency) of the proposed LRRP rule, but rather to develop an approach that would allow for
building multiple exposure scenarios in different residences or COFs in different geographical
areas and to estimate changes in children's IQ from lead exposure due to these scenarios. In this
respect, I have only a few general comments:
1. As discussed in Section 7.1, the approach developed in this document relies heavily on
the experimental data derived from the OPPT Battelle Dust Study, which was rather
limited in scope. The small sample size, limited geographical area, lack of evaluation of
different control options for the same activity in the same building, limited list of possible
RRP activities makes the model input data rather uncertain. A more comprehensive data
set would be required to make the approach developed in this document universally
applicable.
2. The important limitation of this approach is that it does not account for the child's
activity pattern in a building and an adjacent yard that undergoes renovation. Children
may spend more time in different areas or be removed from the building for the duration
of renovation and in these cases their exposures would be significantly different. This is
especially important in the case of multiple activities and COFs- it is rather unlikely that
children would be left in the building during the major renovation activities. The post-
renovation Pb level would be the most important in this case.
3. As noted in the document, the assessment of the cleaning efficiency after renovation is
also very limited. The supplemental material "Cleaning Verification Extract" that
describes the methods used for assessment of cleaning efficiency makes it clear that these
methods do not provide real Pb-loading data. Since the buildings may not be even
occupied during the renovation activities, the cleaning efficiency may be very important
in terms of the Pb exposure.
4. Since the two biokinetic models used in this document give very different blood Pb
concentrations, it would be necessary to resolve the question which of these two models
D-38
-------
is appropriate to use for LRRP analysis before the more general application of the
developed methodology. Although the Leggett model seems to be more appropriate for
the RRP activities since it can accommodate the shorter exposure span, it generates
values that are much higher than the observed blood levels. However, the higher
estimates of Pb blood levels may be more consistent with the protection of children's
health. I think that the empirical data should be used for model verification. Perhaps
IEUBK model could be used for estimating the background concentrations and Leggett
model for incremental changes resulted from RRP activities.
Since the estimated incremental IQ changes due to the major renovation activities were
rather small (Chapter 6), I don't believe that the approach developed in this document
could be applicable to low dust generating activities, such as small repairs. Due to the
high overall uncertainties of this analysis, low exposure scenarios would generate rather
unreliable results.
D-39
-------
Appendix E - Review Comments from Individual CASAC Panel Members
This appendix contains the preliminary and/or final individual written review
comments on the Agency's Draft LRRP Activity IQ-Change Methodology and the OPPT
Dust Study from those members of the CASAC Panel for Review of EPA's LRRP
Activities who submitted such comments. The comments are included here to provide
both a full perspective and a range of individual views expressed by Panel members
during the review process. These comments do not represent the views of the CASAC,
the EPA Science Advisory Board, or the EPA itself. The names of the Panelists who
provided review comments are listed on the next page, and their individual comments
follow.
E-l
-------
Panelist Page #
Dr. Joshua Cohen E-3
Dr. Ellis Cowling E-8
Dr. Douglas Crawford-Brown E-10
Dr. Bruce Lanphear E-20
Dr. Frederick J. Miller E-26
Dr. Maria Morandi E-32
Dr. Paul Mushak E-35
Dr. Michael Rabinowitz E-42
Dr. Joel Schwartz E-48
Dr. Frank Speizer E-49
E-2
-------
Dr. Joshua Cohen
Comments on the sensitivity and Monte Carlo analyses
The description of the sensitivity analysis and Monte Carlo analysis should be revised to
help better convey several important issues.
1 Quantify Impact of Uncertainty and Variability Separately
EPA's report does not describe the purpose of the sensitivity and Monte Carlo analyses.
Is their purpose to characterize variability or uncertainty? The first sentence of Section 4.6.1 (the
overview of the sensitivity and Monte Carlo section) reads in part, "/Y is important to
characterize the impacts of this variability and uncertainty on the estimated dust
concentrations."
The analyses include a large number of parameters, but the results do not parcel out how much of
the estimated variation is due to uncertainty, and how much is due to variability.
Variability represents differences between different members of the population (or, in this
case, different dwellings) that cannot be reduced by the collection of additional information.
That information can help determine the extent to which the proposed intervention might better
be targeted at those dwellings where they yield the largest benefit. There are conceivably several
important sources of variability that were not, as far as I could tell, explicitly addressed. These
sources include differences in: 1) building condition; 2) lead paint coverage on surfaces that will
be disturbed; 3) type of renovation; 4) number of children living in the dwelling, among other
factors. It is possible that the net benefit of the proposed rule will vary substantially. In some
cases, it could conceivably be negative, while in other cases, more intensive measures might be
justified.
Uncertainty, on the other hand, indicates that there is a range of plausible results because
of a lack of knowledge. Quantification of uncertainty helps decision makers understand if the
knowledge base is sufficiently robust to support rule making, and if it is not sufficiently robust, it
helps identify those sources of uncertainty that are most important to resolve. The sensitivity and
Monte Carlo analyses should quantify the extent to which an outcome of interest varies
depending on the assumptions used. EPA reports the extent to which various outputs change in
response to an arbitrary change in the input quantities (e.g., a 10% change), yielding a response
elasticity. This information does not, however, account for how uncertain each input quantity is.
Two assumptions can have the same elasticity, but one may be more uncertain than the other and
therefore make a larger contribution to the overall uncertainty of the output.
E-2
-------
2 Use of Dust Lead Concentration
Use of dust lead concentration as a measure of exposure is problematic because it
introduces substantial uncertainty, and the analysis does not sufficiently quantify its impact on
the results. Tables A-5, A-6, and A-7 in Appendix A do a reasonable job laying out the
assumptions used in the sensitivity analysis and Monte Carlo analysis. The treatment of the
load-concentration slope coefficient is not clear, however. The central value in Table A-6 is 0.5
and no hi and low values are listed. The footnotes read:
• " The Load-concentration Intercept and Load-concentration Slope were
determined through ICF analysis. Please see Appendix Cfor a more detailed
discussion of this analysis."
and
• "There were no low and high values for the Load-concentration Slope, however a
CVof2 was deemed too high. Therefore, the Load-concentration Slope CVwas
estimated by setting the CVequal to the Load-concentration Intercept CV."
Appendix C (referred to in the first footnote passage above) describes the regression analysis,
which indicates that the estimated dust lead concentration is very uncertain, spanning a bit more
than e3, or a factor of about 20.
The second footnote passage does not help. Section Dl explains that "If no low and high
values were available, then the variable was assumed to be normally distributed and was
assigned a default coefficient of variation of two, representing a conservative estimate of the
variability.'" Why was CV = 2 deemed too high in this case? If it was deemed to be too high,
then there must be an alternate value that makes more sense. Using the CV from the intercept
does not make sense and probably understates uncertainty because the intercept is probably more
certain than the slope coefficient. In any case, there is no reason to believe that the CV for the
intercept is a good proxy for the slope's CV.
It appears that EPA used concentration (ug lead/g dust), rather than loading (ug lead/ft2
floor space) because the biokinetic models require a concentration value. This conversion
introduces uncertainty by using an uncertain relationship to replace an empirical measurement
(loading) with a proxy (concentration). Comments from Bruce Lanphear indicate, moreover,
that even without consideration of this additional uncertainty, concentration is an inferior
predictor of blood lead than the original loading measurement. EPA would be better off if it
could use loading as an input to its biokinetic model, rather than concentration. I briefly describe
several approaches.
The simplest approach would be to compute background blood lead and the increment
associated with exposure to RRP-generated dust lead separately, and then to add the two
together. The background could be computed using NHANES data or the IEUBK model, as
E-4
-------
discussed in the July 9-10 meeting, while the increment could be taken directly from the
relationship between dust lead loading and blood lead described in Bruce Lanphear's 2005 paper
(I do not have the cite, but he showed me the paper during our meeting). The shortcoming of
this approach is that the relationship described in Lanphear et al. (2005) is based on a chronic
exposure to lead. A second approach uses a biokinetic model to address this limitation. Let us
assume that the Leggett model is used for this purpose because it can model the impact on blood
lead of short time scale changes in lead exposure. EPA should determine the chronic change in
dust lead concentration that increases blood lead levels by the same amount that the Lanphear et
al. model predicts a given change in dust lead loading will produce. This mapping should be
computed for a range of incremental dust lead loadings. Note that Lanphear et al. (2005) use a
log-linear relationship to predict blood lead as a function of dust lead loading, so the impact on
blood lead depends on both the beginning and end dust lead loading. The numbers in the table
below are fabricated but illustrate the approach.
Start Loading
0
0
0
10
10
End
Loading
10
20
30
20
30
Change in PbB
(GM) Predicted
by Lanphear
(2005)
2
O
3.5
1
1.5
Corresponding Dust Lead
Concentration Increment Producing
Same Increase in PbB in the Leggett
Model
200
300
350
100
150
Note that I am assuming that in the Leggett model, the relationship between dust lead
concentration and blood lead is roughly linear.
The next step is to produce the appropriate dust lead concentration time series that will be
input into the Leggett model. EPA should assume that the same dust lead concentration values
that reproduce the Lanphear relationship for chronic exposure will also work for short term
exposure. For example, suppose that the assumed background dust lead loading is 10 ug/ft2, that
the immediate post-RRP level is 30 ug/ft2, and that after 1 week, the level is 20 ug/ft2. The dust
lead concentration time series to be entered into the Leggett model would then be:
E-5
-------
Time Point
Before RRP
Post-RRP
0
Assumed
Baseline Dust
Lead Loading
10 ug/ft2
10
10
Measured
Incremental Loading
0 ug/ft2 (by definition)
20
10
Corresponding Dust Lead
Concentration Increment
Producing Same Increase in
PbB in the Leggett Model
0
150
300
3 Results Display
Exhibit 4-14 in the main report illustrates key percentile values for week 0 dust lead
concentrations for each of the four control options evaluated in this study. This quantity is not,
however, what is important for two reasons. First, what we are interested in is the difference
between the impact of the control options. Because all of these sources of uncertainty affect the
impact of each control option to a similar degree, the uncertainty across control options is
correlated. As the figure suggests, the differences between control options are probably not as
the uncertainty within each control option. It is not clear exactly how correlated the different
control options are, however. It is also likely that the quantities that influence individual control
option results contribute more to uncertainty of the differences between the control option
cleanup levels than do assumptions that influence all of the control options. Second, we are
interested in the uncertainty in the estimate of exposure aggregated over the entire remediation
process. This quantity serves as a good indicator of the impact of remediation on blood lead
levels, which in turn serves as a good indicator of the impact of remediation on IQ. The results
in Exhibit E are likewise uninformative because they report week 0 and week 10 dust lead
concentrations, and the number of weeks it takes for dust lead concentrations to return to
background.
4 Use of "Conservative" Assumptions in Monte Carlo Analysis
The Monte Carlo analysis involves a mix of distributions and, in some cases,
conservative estimates for parameters. "Conservative" estimates are used the decay constant for
air lead (p. 56), and assumptions related to child access to the work area (p. 87). Because the
analysis is designed to evaluate an intervention, rather than the burden associated with the
presence of a chemical, it is not clear what "conservative" means in this context. If we
overestimate the size of the risk (the burden), we end up overestimating the benefit of the
intervention. It can be argued, though, that a "conservative" evaluation of an intervention would
underestimate its net benefit.
E-6
-------
5. Addressing Report Limitations
It is not clear that the sensitivity analyses and Monte Carlo analysis addressed the most
important sources of uncertainty. Both sections 3.5 and 4.7 of this report list important
limitations to the analysis. These sections are very useful. The limitations can be addressed in a
number of ways. It may be possible to dismiss some limitations as not having an important
impact on the results of the analysis - i.e., they introduce unimportant sources of uncertainty. If
that is not the case, it would be best to attempt to quantitatively account for these limitations in
the sensitivity and Monte Carlo analyses. Short of that, the report can explain how these
limitations qualify the results. Perhaps the findings cannot be generalized to a certain class of
dwellings. If even that is not possible, the report should explain what sort of data should be
collected to overcome the limitation.
While listing limitations is useful, it is insufficient. Without somehow addressing the
limitations, they become reasons for dismissing the usefulness of the analysis entirely. From my
perspective, the most important limitations are those that relate to the quality of the data. In
particular, it is not clear to me how a convenience sample of 12 residential dwellings and 3 child
occupied facilities can be used to represent the range of building types and conditions that will be
encountered in the U.S. At the very least, the uncertainty inherent in such an extrapolation
should be addressed. Without doing so, it is not clear how meaningful the results the results of
this analysis are.
E-7
-------
Dr. Ellis Cowling
Dr. Ellis Cowling
North Carolina State University
July 23, 2007
Post-Review-Meeting Comments on the EPA Report titled:
"An Approach for Estimating Children's Health Risk (IO) Changes Associated with Dust
Lead Generated by Renovation, Repair, and Painting Activities"
Several of my colleagues on CASAC are much better prepared by experience and scientific
expertise to provide very constructive comments on specific details of the approach described in
this EPA report on "Estimating Children's Health Risk (IQ) Changes Associated with Dust Lead
Generated by Renovation, Repair, and Painting Activities". My natural predilection as a member
of C ASAC, and also my assigned task in our recent review of this report, was to focus on the
most general aspects (Charge Question 1 - Overall approach and utility) rather than more
specific aspects of this report and its proposed methodologies for estimating IQ changes related
to lead exposure in residential housing and other "child occupied facilities."
Charge Question 1: Overall approach and utility
Various agencies of our Federal government have established a very worthy goal for
environmental protection — "eliminating childhood lead poisoning [in the United States] by the
year 2010" — [now only 3 years away!].
In this connection, EPA has proposed new requirements aimed at decreasing the exposure of
children to the lead used mostly during various decades of the 20th Century as pigments in many
of the interior and exterior paints applied during the original construction and renovation of
homes, school buildings, day-care centers, churches, and other buildings that frequently will be
occupied by children in the early years of the 21st Century. These requirements will affect the
manner in which renovation and repair of already existing child occupied facilities will be
achieved during the next few years of the present Century. Thus it is important that EPA's
Office of Pollution Prevention and Toxics do its job both very well and in a very timely way!
Many aspects of the general approach outlined in this EPA report seem very reasonable.
Unfortunately, however, the report contains few, if any, references to indicate how the approach
and resulting estimates proposed by the USEPA will be related to and pursued in conjunction
with the activities and responsibilities of other federal agencies whose leadership and resulting
actions also will be necessary to achieve the noble goal of "eliminating childhood lead poisoning
by the year 2010."
E-S
-------
As we look forward to the time it will take for the USEPA to implement the approach
outlined in this report, I could not help but wonder why CASAC was asked to review this
approach when the goal of our government is "eliminating childhood lead poisoning by 2010" -
is now only 3 years away!
During CASAC's recent review of this report, we asked and learned that:
1) This interagency goal of our federal government was established in 1992,
2) EPA promulgated its first regulations with regard to renovation, repair, and painting of child-
occupied facilities in 1996,
3) EPA provided its first guidance for implementation of these rules and regulations in 1997, and
4) Until very recently, relative little further actions were taken by the USEPA with regard to
reconsideration and implementation of these rules and regulations.
This very brief and no doubt inadequate chronology of events within the USEPA led to
further inquiry into the long-term history of both governmental and private sector actions and
activities with regard to lead pollution in the United States and other developed countries around
the world.
The attached references indicate that the currently proposed actions and reports regarding
lead pollution and resulting lead poisoning in the United States are long overdue. We commend
the present administration of the USEPA for undertaking their presently renewed interest and
actions with regard to lead pollution and poisoning and hope that the USEPA will do its
important and appropriate part — together with other federal agencies - and thus help our country
make further progress toward our national goal of "eliminating childhood lead poisoning -
[hopefully] by 2010," if not by then "as soon thereafter as may be possible. Better late than
never!
References:
1) Childhood lead poisoning prevention. Too little, too late. B P Lanphear. J Amer Med Assn.
2005 May, 293(18):2274-2276.
2) "Cater to the children": The role of the lead industry in a public health tragedy, 1900-1955. G
Markowitz and D Rosner. Am J Public Health. 2000 January; 90(l):36-46.
3) The secret history of lead: How General Motors, Standard Oil and Du Pont colluded to make
and market gasoline containing lead—a deadly poison—although there were safe
alternatives. Abetted by the US government, they suppressed scientific knowledge that
lead kills. Still sold in countries all over the world, leaded gasoline continues to poison
the planet. J L Kitma. The Nation. March 20, 2000: 11-44.
4) Warnings unheeded: A history of child lead poisoning. R Rabin. Am J Public Health. 1989
December; 79(12): 1668-1674.
5) "A 'Gift of God'?: The public health controversy over leaded gasoline during the 1920s." D
Rosner and G Markowitz. Am J Public Health. 1985 April; 75(4): 344-352.
E-9
-------
Dr. Douglas Crawford-Brown
Review of An Approach for Estimating Changes in Children's IQ..., Draft for CAS AC Review
Doug Crawford-Brown
General Comments
I begin by noting that this is a mature document that lays out quite clearly (in most instances)
how the assessment will proceed. I found the organization of the document, including the use of
the flow diagrams for computations (and the sub-diagrams for specific subsections) particularly
useful in keeping track of the discussion. The authors are to be congratulated for their
organization.
I found throughout that (for the parts where I feel I have the necessary expertise) the authors
have chosen the appropriate studies and models to use, including the appropriate parameter
values for the models.
Before getting into specific comments, and then answering the Charge Questions, I do have one
quibble. It relates not to the way in which the assessment will be done for a specific activity or
set of activities in a home, but rather to how the EPA intends to use any results. The proposed
methodology is no different in kind from the methodologies used for any number of national
assessments, such as that of incinerators. In the latter case, however, there is a distinct set of
sources with a defined (however well or poorly) emissions inventory. In the current case, I don't
think we can specify the distribution of activities that will be conducted, unless we assume that
the future distribution is the same as the distribution of activities that have taken place so far.
And I doubt this will be a valid assumption.
So it seems problematic to me to produce anything like a national variability distribution of IQ
loss over the potentially exposed population. But then this may not be the intent in the first place.
The intent might be to produce inter-subject variability distributions of IQ loss for hypothetical
populations exposed to each specific activity and some representative subset of complexes of
activities. The goal might then be to say something like: For the following activities and
complexes of activities, the inter-subject variability distribution oflQ loss is acceptable if the
contractor uses the Rule Procedures, and for the remaining activities and complexes of activities
even the Rule Procedure is insufficient to provide adequate protection of public health. There
could even be some subset in which the Rule Procedures are not needed to protect public health
(although the data in the OPPT Dust Study suggests this is unlikely).
E-10
-------
The concern I raise here doesn't affect the methodology proposed. That methodology remains
sound. I simply, as a reader, was not completely clear how the assessment results would relate to
specific public health decisions that needed to be made.
Some Specific Comments
1. The flow diagram on Page 10 has a step termed ... convert dust lead loadings to dust lead
concentrations. It doesn't say concentrations in what. I presume air (indoors) and soil (outdoors),
but this could be made more clear. Also, there are hints throughout the document that at least
some of the OPPT Dust Study results contained air sampling, and that these air samples might be
used directly. This would remove the need to estimate air dust from dust loadings. However,
there also are places in the document where I received the distinct impression that all indoor air
concentrations would be estimated from loadings. This needs to made more clear for the reader.
2. In the section on the epidemiological studies, it at first worried me that the studies do not
adequately characterize prenatal exposure to Pb. There are legitimate arguments to be made for
prenatal exposure being quite important. But then as I thought about it further, I became less
worried. That is because one might assume that prenatal exposure is in part (perhaps even in
large part) correlated with post-natal blood lead levels. If the infant has high blood lead levels
that are maintained throughout the first 6 years, this probably indicates exposure in the home,
and the mother might be expected to have received similar exposures during pregnancy. So
perhaps this is not such as major weakness of the epidemiological studies. Just a word or two to
this effect might alleviate concerns.
3. Again in the epidemiological study section, it would be an improvement to include the
Lanphear et al data in some sort of dose-response graphical form. I find it unsatisfactory to just
note (as on Page 17) that there was a ".. .decline of 6.2 points in full scale IQ for an increase in
concurrent blood lead levels from less than 1 to 10...:" Since essentially all of the "action" in
this document is applied at blood lead levels between these 2 values, there is a pressing need to
see what the D-R curve looks like in there. Absent this, the later assumption of a possible 1 |ig/L
threshold appears arbitrary, as does the assumption of linearity, piece-wise linearity, etc.
4. The authors note that their use of the word "Phase" (see Page 22) differs from that in the
OPPT Dust Study report. It would be better if this difference could be resolved. It has no
implication for the results of the current report, since the authors make the translations correctly.
But it struck me as odd at first having just reviewed the OPPT Dust Study report.
5.1 noted in Figure 3-7 that there is no mention here of people ingesting dust from surfaces. The
last box ends up being concentrations. Later in the document, the authors note correctly that
there will be dust ingestion (or at least they mention ingestion of dust lead and suggest some
E-ll
-------
fraction is due to ingestion outdoors and some fraction to ingestion indoors). This discrepancy
needs to be cleaned up.
6. On Page 33, the issue of background levels of indoor air is raised. There seems to be the
suggestion that this will be estimated from outdoor, ambient, air. Perhaps this is the best that can
be done, but I find it hard to believe the penetration factor for the indoor microenvironment is 1,
and that indoor sources don't elevate the indoor air above outdoor air concentrations in the kinds
of homes being considered here. Are there no data from the OPPT Dust Study or the HUD study
(mentioned on Page 34) that can be used to improve on this assumption?
7. Page 36, Section 3.3.2.2 has a statement that I don't believe can be correct. It says that
".. .values below 0.375 |ig/ft2 set equal to 0.375 |ig/ft2, which is equal to one quarter of the
detection limit". But if 0.375 is one quarter of the detection limit, that limit must be 0.375 x 4 =
1.5 |ig/ft2. So how can there be ANY measurements down below 1.5 |ig/ft2 if that is the
detection limit? Or is this an artifact of the measurement method, which produces results down to
0 (and even below 0 when background is subtracted) even if ones below 1.5 |ig/ft2 are not to be
considered statistically significant detects? But if the latter is the case, then the procedure should
be ... values below 1.5 |ig/ft2 set equal to 0.375 |ig/ft2, which is equal to one quarter of the
detection limit. I am at a loss here to understand this paragraph.
8. On Page 36, near the top, the authors state that".. .the increased ventilation in the outdoor
environment as compared to indoor ventilation would likely reduce the effect". But "reduce" and
"make insignificant" are not the same things.
9. On Page 43, on about the 11th line, the authors state that"... geometric means.. .correct for
positively skewed data...." I don't understand what they mean by "correct for" here. There is no
"correction" being made. The median is simply more stable with respect to extreme values in the
tails. This doesn't make it a better estimate of, or correction for, the mean.
10. Also on Page 43, the authors introduce the idea of using 10% of the window sill loading to
add to the floor loading. This value comes out of no where. I can see no justification for it. And
surely it would depend on the ratio of window sill area to total floor area.
11. On Page 43, the last paragraph begins by saying that Exhibit 3-13 contains the loadings. No
such loading values are provided in that table. I assume the authors mean that Exhibit states
WHICH loadings will be considered representative in specific cases.
12. On Page 45,1 was not clear whether it will be assumed that children play on the outdoor
plastic sheet during the Rule Procedure.
13. Between Pages 45 and 46, the bridging sentence appears to suggest that if the yard is divided
into the three areas, the soil value used will be the median of the three medians. Perhaps I have
E-12
-------
misread this. This doesn't strike me as an areal-weighted average unless the three kinds of areas
have the same surface area (which they don't).
14. On Page 54, the assumption is that a person spends an equal fraction of time in each square
foot of the house. This may be the best one can do for now, but if there are data on fraction of
time spent in each room, that might be more appropriate. But again, such data may not be
available and the suggested approach is at least an approximation.
15. On Page 55,1 cannot understand why Equation 4-2 contains the denominator of n. I would
assume the air concentrations contributed by separate activities would be additive. Imagine a
home with just one activity that yields a concentration of 10 (choose your units). Are the authors
saying that if the contractor chooses to do a second activity that releases NO lead, the new air
concentration will be (10 + 0) / 2 = 5? That makes no sense. I am missing something here.
16. On Page 76,1 like the areal weightings applied to a yard, but I don't see any information on
the national sample that will be used to establish these weightings. And it needs to be a sample
tailored to the kinds of homes being remediated, since I suspect these have smaller than average
yards. Or is there really no areal weighting, but rather the procedure I mention in Comment 13?
17. On Page 77, Exhibit 4-11 contains values that are not described as medians, means, 95th
percentiles, etc. Which are they? And it will be important to ensure that whatever statistical value
is used, it meshes with the assumptions employed from the OPPT study and other aspects of the
methodology in the current report.
18. Beginning on Page 91,1 began to have some questions about the Monte Carlo analysis. The
authors propose a reasonable methodology, but I was not clear as to how an individual would be
followed over multiple years. Once some parameter value is selected (e.g. air inhalation rate) for
an individual, is that same value used for that individual throughout that year? And when the
next year occurs (and the age has changed), will a completely random value from the underlying
inter-subject variability distribution be selected, or will it be chosen at the same Z-value as for
the first year? For example, if an individual is at the high end of inhalation rates for the first year,
will they also be at the high end for subsequent years? The decision here will influence the
variance of the final inter-subject variability distribution for IQ loss.
19. In Exhibit 5-3, Page 94, the authors refer to "inhalation absorption fraction". I presume this
includes both the deposition fraction and the fraction transferred to the bloodstream?
20. In the same Exhibit, the "water consumption rate" is given. Is this solely tap water? And does
it include only direct ingestion or also use in food and drink preparation? I assume the latter.
21. The comment 18 above also applies to the Monte Carlo approach to blood lead modeling. I
was not clear as to whether for a given individual, one value of a parameter is chosen and applied
E-13
-------
throughout multiple days of simulation, or new samplings were done many times for a given
individual. I assume the former. But then it strikes me that NO Monte Carlo analysis will be
done using the model. Instead, the national distribution of blood lead levels will be used in a
post-processing mode. I support this latter approach.
22. Throughout the comparison between Leggett and IEUBK, there is an unstated issue about the
difference between what the two central compartments mean in these models. It is evident that
the IEUBK model compares better to the NHANES data, but this MAY be because the Leggett
central compartment is not the same biological medium as is sampled in the NHANES study. It
could be that the Leggett model is predicting concentration only in some sub-compartment of
what the NHANES study uses as the sampled medium, in which case the Leggett values would
go down if they were averaged over the entire medium being sampled in NHANES. I don't know
the answer here, but someone should look at that issue.
23. On Page 109, the authors state that "This methodology is intended to give a conservative
estimate of IQ changes..." I cannot understand why it would produce conservative estimates.
This requires some explanation.
24.1 am not sure how I feel about Exhibits 6-2 and 6-3. They purport to show IQ losses due to
background exposures to Pb. The problem I have with this is that everyone in the
epidemiological studies was exposed to background. I don't see how these studies can measure
anything other than the incremental loss of IQ due to incremental exposure above background. I
suppose these two Exhibits simply assume that the dose-response curve above background
extends down below background, and these Exhibits are predicting what the effect would be if
even background exposures to Pb were removed from the environment. But I am not convinced
this is a valid use of the epidemiological studies. I am ready to be convinced otherwise by our
epidemiological colleagues.
Specific Charge Questions
1. My comments on the overall methodology are provided in the previous section. Applying
them to the economic analysis will be more difficult since one must hypothesize some
distribution of the kinds of renovations that will be done. But this problem is avoided by doing a
benefits analysis only on specific renovation scenarios, rather than for the entire nation. As I
mentioned earlier, I am not clear as to which is to be the application here.
2a. I believe the sensitivity analysis done so far indicates the most important factors to consider,
and will be applicable over a large range of modeled scenarios.
2b. I think doing a separate Monte Carlo analysis for each house is both a waste of resources and
likely to be inaccurate anyway. The parameter distributions used are difficult to relate to a
specific house (and many are difficult to relate to a specific scenario), and so they are better
E-14
-------
applied to broad categories of houses and scenarios. I think a different Monte Carlo analysis for
each house would be over-interpreting the data.
2c. I don't think there is any way to account for uncertainty and variability properly when
making the transition from the Dust Study to IQ change results. The temporal pattern of
exposures in these case studies is simply too different from those in the epidemiological studies
to make any reasonable or defensible corrections. You are already doing the best that can be
done.
2d. I do not recommend a separate Monte Carlo step. The epidemiological results already
INCLUDE the effect of differences in the biokinetics of individuals in the population (i.e. the
slope factors already reflect this variability), and it would be inappropriate (in my opinion) to
apply these differences again in the proposed study through a Monte Carlo biokinetics model.
2e. I am comfortable with the way the probabilistic and deterministic aspects are melded in the
current approach.
3.1 would not use both until you can be sure that the central compartment model in which blood
lead levels are calculated in the Leggett model are the same as those used in NHANES and in the
epidemiological studies. See my earlier comment on this.
4. It is essential that the blood lead indicator (concurrent or lifetime average) be the same in the
epidemiological study and the application. Given that lifetime average will be calculated for the
case studies, it is appropriate to use the results of the epidemiological studies based on lifetime
average.
5.1 think the approach recommended is the best that can be done at present. I would not
recommend changes.
6. This is outside my area of expertise.
7. Again, this is outside my area of expertise.
8.1 believe you have chosen the correct ones. The one caveat I would make here is that I did not
see any explicit mention of kids gnawing on window ledges. I had thought that was an issue.
E-15
-------
Review of Draft Final report on Characterization of Dust Lead Levels After Renovation, Repair
and Painting Activities
Doug Crawford-Brown
My first, and most general, comment is that this is quite an extensive study. Given the vast
variability that exists between structures, and the variability between the way different
individuals conduct any given task and even in the way the same individual will conduct the
same task at different locations, study design was a particular challenge. The result is a lot of
variability in results, not just in magnitude of change with and without rule practices, but even at
times in the direction of change. Having said this, there really isn't anything the researchers
could have done to reduce this variability, and one would expect this kind of variability to exist
in practice anyway. As a result, I am comfortable with the overall design and conduct of the
study.
My second general comment concerns a statement (and it is a key one) on page 2-2: "The study
results may underestimate the levels of dust that would result from a renovation job due to the
absence of these "real world" factors, but the study will achieve its goal of providing
comparative data on the differences in dust lead levels when lead-based paint is disturbed under
proposed versus baseline work practices". There is an implicit assumption here, often made in
both experimental and modeling studies, that missing factors, or uncontrolled factors, might
change the magnitude of results for a given part of the study, but not the RELATIVE magnitude
of different parts. In this case, the assumption is that these missing factors won't change the
relative magnitude, and hence the ratio, of dust lead levels with and without the rule procedures.
I don't buy this idea in general, and it becomes particularly less tenable as the ratios get closer to
1. I'm not saying the research design was flawed, because I believe the best approach to the
study WAS to simplify matters by getting furniture, etc out of the building. I would go further to
say that any responsible remediator of a home should clear the entire room before doing
anything. And by having bare rooms, the study in effect removed a major source of variability
that would have complicated the picture even further. But the authors need to do a slightly better
job of explaining why the particular conditions of the study leave the results valid, rather than
having this implicit assumption that their inclusion would not have changed the ratios of results
for rule and baseline practices.
Homes were selected in part on the basis of the lead content of paint, which is appropriate. I even
agree with the criterion, or cut-point, used for selection. But it would also have been good to
show where the particular homes selected, with their specific lead concentrations (or fractions by
weight), fall within a national distribution. I doubt there are any implications for study
interpretation, but as a reader I just kept wondering how representative the levels are of homes
E-16
-------
that might need remediation in the future. I don't think this would have affected the ratios of
floor, sill and/or soil lead levels (rule over non-rule practices), however.
At the beginning of the report, the authors are quite clear that there was a lot of variability in the
background lead loadings on floors and sills in the various homes, even after the cleaning that
took place prior to work. But then I can't see any place in the report in which these differences
are reflected in the rule to non-rule comparisons. I would expect, for example, that the ratio of
rule to non-rule lead loadings on floors might get closer to 1 as the pre-work loadings increase,
since then the loading due to the work would become a smaller fraction of the overall loading.
Perhaps I missed it, but I have looked several times for some sort of way in which this was
accounted for. It could be accounted for by using the difference between pre and post loadings as
the measure of impact, or it could be done by stratifying the results according to pre-work levels.
At least some mention of this issue would be helpful.
I never fully understood the chosen length of time between the end of work and the taking of
floor samples. It was about an hour, and this surely was selected based on the settling velocity of
the lead in the air, but this wasn't made clear quantitatively. A simple graph showing the
expected RATE of settling (grams per minute) in an originally contaminated air sample would
have been helpful. That would show the decrease in airborne lead over time, and show that most
of the airborne lead was gone by 1 hour, 2 hours, or whatever time was used (and presumably,
therefore, present on the floor and sills). I also note that on Page 4-7, it mentions that there were
times when this was as short as 30 to 40 minutes. Again, a curve would help clarify whether this
is or is not a large issue.
A minor quibble: On page 4-3, the authors list the four Phases. Phase III says "No plastic
covering and rule cleaning after work completion". It is not clear to the reader whether this
means there was (i) no plastic covering but (ii) there was rule cleaning after work completion; or
that there was (i) no plastic covering and (ii) also no rule cleaning after work completion. The
same problem holds for the Phase IV description.
On Page 4-8 (and at a few other places, but 4-8 will illustrate the point), the authors state that at
times, pre and post-work samples were not necessarily from the same spatial location. This
occurred because a worker couldn't identify the precise location of the pre-work sample. This
isn't troubling if there is not much spatial variability (on the scale of the difference between the
two sample locations), but can be a large problem if there IS significant spatial variability.
Something more needs to be said about this issue of spatial variability and its impact.
In a few places (Figure 6-1 is a good example), one of the axes in a figure has no units. In the
case of Figure 6-1, the x-axis is unit-less. It is presumably the loading on the floor, but the reader
doesn't know that.
E-17
-------
I don't understand why Figure 6-2 shows only the floor loadings by job, rather than comparing
with and without some sort of rule practice. Figure 6-2 just seems to aggregate results over too
many dimensions for a given activity.
I realize I should be able to keep track of this myself, but I kept having to refer back to the text to
understand the difference, in figures, between whether rule practices included BOTH plastic and
cleanings. I kept feeling the headings should consistently describe what rule practices meant in
all cases.
The bullet that straddles pages 6-13 and 6-14 confused me. I didn't think the job type or the
amount of area remediated depended on the floor type. Are the authors suggesting that some
types of jobs may have been disproportionately represented in homes that had wood floors, and
especially in homes that had wood floors in poor condition? This would, of course, be
problematic if true.
Figure 6-6 was the first figure where it struck me that error bars, or confidence intervals, really
are needed in many of the figures. The two sets of bars are so close to each other in height that I
began to wonder whether they had any significant differences. I realize this is dealt with later in
the pair-wise statistical analyses, but the bars or intervals on these figures would have provided a
better visual cue for me.
The third bullet on Page 6-23 mentions that air levels were higher when plastic is used on the
floor. This may be due either to differences in the jobs conducted or the pre-work levels of lead
(differences between rooms with and without plastic) or due to easier re-entrainment of lead dust
when plastic is on the floor. Some comment seems warranted here.
In Table 6-10 on Page 6-27, the authors report results of dust loadings in trays. Presumably, this
loading depends on the length of time the trays are exposed to the air. Is there a time associated
with this sampling? Is it at some form of equilibrium?
The statistical analysis was well designed in my opinion, although there are so many analyses, on
so few sample sites, I worry about the strength of any conclusions. But at least the results,
summarized primarily in Figures 9-1 and 9-2 are consistent and suggest there is a beneficial
effect of the rule practices. As mentioned previously, these figures would be improved by error
bars on all bars of the graphs. It also would be good to show a horizontal line with the EPA
target level (e.g., 40 |ig/ft2 for floors).
In tables such as 9-1, there is a value of the number of samples above a target value (e.g., the last
column in that table). It would be useful to show the total samples collected for each row so the
reader knows the fraction of samples above the target. The report doesn't explain whether
compliance means NO sample is above the target, whether each part of a floor needs to be below
the target, whether the target applies to the average of a floor area, etc. I realize that would be
E-18
-------
explained elsewhere in the actual rule, but it would be helpful for the reader to understand that
here.
I wasn't clear whether Table 9-3 combines results across all categories of activities. I presume
from Table 9-4 that this is the case, but the reader should be told this. And if this is true, is the
conclusion therefore that in activities other than door planning and high heat gun, the rule
practices are effective in reducing floor lead levels below the target? That seems to be a valid
conclusion.
I now address the specific questions in the memorandum, although my answers are all contained
at one point or another in the above:
Question 1. Yes, I understood the study objectives from the writing.
Question 2. Yes, I agree with the conclusions, although the discussion section could do a better
job of explaining how the degree of inter-site variability in results produces some specific levels
of uncertainty in the conclusions.
Question 3. Yes, subject to the comments I have above about axis labels and error bars and
slightly improved figure headings.
Question 4. Yes, it was very well laid out. It was difficult to track through all of the various
bodies of data, but that was because there were a lot of data and a lot of analyses of those data,
rather than being due to any deficiencies in presentation.
Question 5. Yes, I found it easy to understand how the data were collected and how they related
to the specific study objectives. I had to keep referring back to earlier sections to be sure I
understood why a particular table or figure was important, but this is only natural in a report of
this complexity.
Question 6. Statistical methodology is not my forte, so I can't comment on this.
E-19
-------
Dr. Bruce Lanphear
Comments on "An Approach for Estimating Changes in Children's IQ from Lead Dust
Generated during Renovation, Repair and Painting"
Bruce Lanphear
Most of my comments are to present an epidemiologic perspective to provide a reality check on
the numerous assumptions underlying the two biokinetic models.
1. Page 92: The definition of hypothetical children and the time in their life when they are
when they experienced the RRP activity is one reasonable approach, but many children
experience more than one or even ongoing renovation or repair activities (see attached
slide). It would be important to consider the impact on blood lead concentration and IQ
when there is ongoing or > 1 renovation activities during a child's lifetime exposure.
2. Page 100: The estimated "Lifetime Average Blood Lead Levels" in Exhibit 5-6 are quite
small across the Control Options. Since we know that much greater reductions in dust
lead loading values can be achieved - and that the dust lead loading values can be quite
high after renovation using baseline renovation and repair activities and cleaning
procedures - the estimated reduction and subsequent benefits of the proposed rule dust is
dramatically underestimated. Given these incorrect assumptions, it isn't surprising that
there are small estimated differences in IQ scores by the various Control Options.
3. Page 107: The Report states, "Regrettably, there are little data available concerning the
patterns of lead exposure of the NHANES participants." It is also regrettable that the US
EPA didn't take advantage of the NHANES data set and insert relevant questions in
NHANES because they collected blood lead levels and floor dust lead levels for housing
units with children > 6 years of age. It if further regrettable that the EPA didn't plan
ahead and fund the necessary epidemiologic studies in the mid-1990s, when Congress
mandated the promulgation of this rule. Too often, there are delays in rule making until
there is a lawsuit and then the EPA staff are given too little time and have too little
empirical data to set scientifically-based health standards. This is an extreme example of
a failure of leadership at EPA that has tremendous adverse consequences for the health of
the US public.
4. The effects of IQ were apparently made only after accounting for background exposures.
If so, you will underestimate the effects because renovation activities occur more
frequently in owner-occupied housing in which children have lower mean baseline blood
lead levels and because the greatest decrements in IQ occur at the lowest increments in
E-20
-------
blood lead concentration and because children who live in owner-occupied housing are
more likely to be exposed to renovation and repair activities and have lower baseline
blood lead concentrations.
5. I would argue that the piece-wise analysis is more appropriate because the majority of
children have maximal baseline blood lead levels below 7.5 ng/dL, and the mean increase
in blood lead concentration, on a population level, would generally be up to but not
exceeding a blood lead concentration of 7.5 |j,g/dL. Thus the estimated -2.94 IQ
decrement per 1 |j,g/dL should be used to estimate the effect of the proposed LRRP rule.
6. In addition to the current approaches, EPA staff should explore calculations to test
whether the biokinetic models are valid or are in direct conflict with empirical data. One
simple estimate of the impact of renovation and repair practices and the proposed rule on
blood lead levels and IQ using empirical data is:
A. Begin with population mean blood lead levels for a cohort of 1 or 2 year old children
from NHANES;
B. Estimate the increase in blood lead concentration due to renovation and repair activities
using three values: 12.5% (Lanphear et al.) from unpublished research and by -28-30%
(from M. Rabinowitz, et al. AJPH 1985), and by about 20% (the midpoint of these two
studies (i.e., a range of estimated increases in blood lead concentrations due to
renovation).
C. Estimate the reduction in IQ associated with a increased in population mean blood lead
concentration for the three scenarios (12.5%, 20% and 30%) using the -2.94 IQ
decrement per 1 |j,g/dL from the piece-wise linear analysis for children with blood lead
concentration < 7.5
D. Estimate the range of estimated IQ benefits of the proposed rule (i.e., if the proposed rule
reduces lead exposure and blood lead concentration by 25%, 50%, 100% or 150%)
relative to the baseline practices. The estimated reductions in blood lead levels could use
published reductions as well as the OPPT Dust Study. They should also include
estimates of achieving dust lead loading values <10 ng/ft2, as indicated by the HOME
Study (Lanphear, et al. unpublished data). It should, as Rogene pointed out, result in a
protective effect (i.e., an increase in IQ) for some children;
E. Use these empirical estimates to calculate the cost benefit of the proposed rule for a US
birth cohort of 1 to 2 year old children.
E-21
-------
Comments on "Draft Final Report on Characterization of Dust Lead Levels after Renovation,
Repair and Painting Activities"
Bruce Lanphear
There are many aspects of this study that provide valuable data for the promulgation of a rule
to protect children from lead poisoning after repair and renovation activities. Overall, for
example, the results showed quite conclusively that the use of the proposed rule practices and
plastic led to significantly lower dust lead loading values compared with baseline cleaning and
baseline practice. I will focus my comments on the problems and limitations of the study.
Several decisions or exclusions described in this document will minimize or underestimate
lead hazards created by repair and renovation. Similarly, they will underestimate the benefits of
the proposed rule. This should be considered in the final rule. These include:
4. The draft report blindly accepts as truth that achieving floor dust lead loading values of
40 ng/ft2 and sill dust lead loading values of 250 ng/ft2 are sufficient to protect children,
despite considerable evidence to the contrary. If the LRRP relies on clearance levels on
floors above 10 |J,g/ft2 or 15 ng/ft2, it will inevitably fail to protect children who live in
housing units that undergo repair and renovation. Any clearance levels in excess of 15
Hg/ft2 on floors and 50 |J,g/ft2 on window sills should be justified in relation to existing
epidemiologic data (see attachments for Lanphear, 1996, Lanphear 1998, Malcoe 2002,
Lanphear 2002, Lanphear 2005). This is a problem throughout the document, but
relevant examples can be found on 2-1, 1st paragraph and 3-4, 4th paragraph. Failure to
account for the risk associated with floor dust lead loading values of 40 ng/ft2 and sill
dust lead loading values of 250 ng/ft2 will result in a dramatic underestimate of the
generation of lead hazards by renovation and repair activities.
5. Page 4-2, 3rd paragraph: Excluding 8 out of 35 housing units because they were in poor
condition would tend to minimize or underestimate the generation of lead hazards from
repair and renovation. Similarly, by excluding housing units with floors in poor
condition, the differences in dust lead loading values between rule practices and baseline
practices would be underestimated.
6. Page 4-7, 5th paragraph: Use of sample trays in place of window sills occurred because of
inability to achieve clearance and insufficient sill surface area. Use of sample trays
because of inability to achieve clearance would tend to underestimate lead hazards after
repair and renovation.
E-22
-------
7. Page 4-8, 3rd paragraph: The study design called for sampling to occur on actual floor
surfaces, but plastic sheeting was used on some tool and observation room floors
"because floors that were in poor condition were difficult to clean".
8. On the other hand, the use of procedures and practices that are prohibited (e.g., use of a
heat gun > 1100°) and the inclusion of vacant housing units, may have exaggerated the
lead hazards from renovation and repair of residential dwellings. Nevertheless, we can
anticipate that many renovations and repairs will not be done using lead-safe work
practice. Thus, I agree with the peer-reviewers of the OPPT Dust Study that adding these
prohibited practices actually makes the study more realistic. Moreover, the cleaning done
prior to implementing the study reduces the problem of including vacant housing.
Other comments:
9. As noted in the previous comments, there should be a review of the epidemiologic
research linking renovation and remodeling with lead poisoning. This is particularly
important because the limited sample size and questionable generalizability of the
sample. There should also be a review of the epidemiologic evidence supporting the
selection of floor dust lead loading values of 40 |J,g/ft2 and sill dust lead loading values of
250 ng/ft2.
10. Page 1.1, first paragraph: It would be useful if the report described how the paint lead
levels and dust lead loading values in the LRRP studies compared with those in the
National Survey of Lead and Allergens.
11. Page 4-8, paragraphs 4 and 5: Verification was not adequately defined nor was the
description of post-verification floor lead sampling. Verification, which refers to a
version of the "white-glove" test, is a misnomer. It is described by EPA as:
Disposable Cleaning Cloth/White Glove Study. EPA began looking for an
alternative to dust clearance sampling that would be quick, inexpensive,
reliable, and easy to perform. EPA conducted a series of studies using
commercially available disposable cleaning cloths to determine whether
variations of a "white glove" test could serve as an effective alternative to
dust clearance sampling. White disposable cleaning cloths were used to wipe
windowsills and wipe floors, then examined to determine whether dust was
visible on the cloth. This determination was made by visually comparing the
cloth to a photographic standard that EPA developed to correlate to a level of
contamination that is below the dust lead hazard standard in 40 CFR 745.65(b).
Cloths that matched the standard were considered to have achieved "white
glove."
E-23
-------
12. The problems identified in this Report highlight how ridiculous it is to try to replace a
highly specific measure of lead in house dust (a dust sampling wipe test that quantifies
the amount in lead in a uniform area and that has been extensively validated for its ability
to predict children's blood lead concentration) with a simplistic and non-specific measure
of discoloration. For example, during the study, the contractors found that using "Simple
Green" could result in discoloration and that "some RRP contractors appeared more
likely than others to consider discolored wet verification cloths to be clean, attributing
discolorations to factors other than residual paint dust (with no apparent basis for such
conclusions)". Finally, as noted in the Report, "Overall, only 3 window sills failed the
first wet cloth verification despite the fact that nineteen window sills had post-cleaning
levels > 250 jug/ft2'."
Despite these obvious problems and lack of firm evidence, the US EPA states: "EPA
believes that adherence to this post-renovation cleaning verification protocol, in
combination with the proposed training, containment, and cleaning requirements is a safe,
reliable and effective system of ensuring that renovation activities do not result in an
increased risk of exposure to lead-based paint hazards. In the great majority of cases,
windowsills and floors that achieve post-renovation cleaning verification will also pass
dust clearance sampling."
13. Page 6-1, Table 6-1 and page 6-5, 1st paragraph: Interior paint lead levels ranging from
0.8 to 13 percent (average = 4%) by weight seemed low. How do these values compare
with the National Survey XRF values?
14. Overall, the results showed quite conclusively that the use of the proposed rule practices
and plastic led to significantly lower dust lead loading values compared with baseline
cleaning and baseline renovation and repair practices. Although this wasn't statistically
significant across all rooms (e.g., observation rooms), the lack of a difference were likely
due to small sample sizes.
15. Page 9-5, paragraph 3: The differences in dust lead loading values with and without the
use of plastic sheeting (41, 43 and 39 ng/ft2 for post-work, post-cleaning and post-
verification compared with 51,61 and 79 ng/ft2) in the tool room, for example, were
described as "small". I disagree. These differences are quite substantial on a population
level, especially in view of the high dust lead loading values. The mean values will also
mask exposures of some children to considerably higher dust lead loading values.
16. Page 9-6, 3rd paragraph; Of particular concern, "nearly half of the experiments ended
with average work room floor lead levels above 40 ng/ft2" - a level of settled lead-
contaminated house dust associated with about 20% of children having a blood lead
concentration > 10 |j,g/dL. Not surprisingly, the majority (20 out of 29) of floors that
exceeded 40 ng/ft2 were in housing units that had floors in poor condition. This finding
E-24
-------
indicates that further research and more extensive cleaning will oftentimes be necessary
to make housing units safe after renovation and repair.
E-25
-------
Dr. Frederick J. Miller
An Approach for Estimating Changes in Children's IQ from Lead Dust Generated during
Renovation, Repair, and Painting in Residences and Child-Occupied Facilities
General Comments:
The 2nd draft of this document is much improved relative to the presentation of the methodology.
Staff have laid out the most important aspects of the factors to be considered in support of the
analyses of the impact of renovation, repair, and painting activities in residences with lead-based
paint or in child-occupied facilities.
For the types of analyses needed in support of the proposed rule, the Leggett model would
appear to be more appropriate than the IEUBK model to use. Staff repeatedly phrase sentences
towards use of the IEUBK model because the predicted blood Pb levels are lower and tend to
agree more with values from NHANES. However, using NHANES as the baseline for
comparison is equivalent to comparing apples to oranges as the NHANES survey had a large
number of homes that did not have lead-based paint. Moreover, the stratification according to
various demographics in the NHANES study argues that the overall population value not be
used. Hence, it is quite likely that the upper tail NHANES data are the values that one should be
comparing the blood Pb modeling results with. Discussions during the July 9-10 meeting with
CAS AC brought out the point that the IEUBK model did pretty well at handling the trends in the
blood Pb data and most likely provides a lower bound on expected blood Pb values. It is also
clear that the Leggett model results probably provide an upper bound on expected blood Pb
values. I endorse using both models and letting comparisons with experimental data provide the
decision point on the final answer as to which model provides the most relevant estimates for the
exercise being undertaken here.
The document still has some technical errors that need to be corrected. One of the most
egregious of these is a failure to use newer data on the deposition in the respiratory tract of
children of various sizes of particles that will be present in indoor dust after Pb renovation
activities. The result is an under prediction or an over prediction of lung mass absorbed
depending upon particle size and age of the child. In addition, the models and methods do not
appear to account for deposition in the head following inhalation of indoor dust, and yet this
source likely accounts for more than 220 jig of Pb (assuming an indoor air level of 10 |ig/m )
that is transferred to the G.I. tract on a daily basis, which would be a value about 80-fold higher
than what is input from the diet on a daily basis. Some way needs to be developed to incorporate
head deposition in the inputs to the blood Pb models.
The authors of this document seem to have combined the concepts of uncertainty and variability
into one lumped category. Variability stems from differences in the population being studied
with respect to any number of factors such as type of housing, inhalation rate, ability to clear Pb
E-26
-------
from various organs, etc for which no additional data can be obtained to delineate among the
individuals. Uncertainty, on the other hand, reflects the fact that the choice of variables or the use
of methods, models, etc can lead to plausible different answers about risk, some attribute like
blood Pb, etc. As such, the sensitivity analyses need to focus on whether a variable or factor
imparting uncertainty can make a difference in the "bottom line" outcome.
While on the right track, the sensitivity analyses that were performed are not sufficient to provide
a sense of the importance of the input variables that were allowed to vary relative to estimating
dust concentrations. Specifically, all input variables were allowed to vary only by 10% and the
resulting elasticity and sensitivity statistics computed. Without knowledge of the standard
deviations about the mean for all of these input variables, one would have to suspect that not
enough variation in the input variables was studied. For each input variable, does a 10% change
correspond to a standard deviation, two standard deviations, l/4th of a standard deviation or
what? This review suspects that for most input variables, the 10% change is far less than a single
standard deviation from the mean and therefore, is not sufficient to really establish whether the
input variable contributes significantly to estimations of dust concentration. Discussions at the
July 9-10 meeting with CAS AC clearly reinforced the above comments.
It would be helpful to the reader if the authors would include a discussion of how far the
elasticity and sensitivity statistics need to depart from zero before one should conclude that the
calculated deviation is indeed significant, if not statistically, then from a data interpretation
viewpoint. Without such a context, one can not really identify which input variables are the most
important drivers of changes in the Pb dust concentration changes.
The Monte Carlo simulations that were done for uncertainty in indoor dust Pb concentrations are
described briefly in the text and more explicitly in Appendix D. The text description of how the
Monte Carlo simulations were done glosses over the major distributions assumption that is
explained in Appendix D. The authors state in this appendix how the low and high values for the
variables listed in Exhibit 4-13 were used to determine whether a normal or a lognormal
distribution was more appropriate to use in the simulations if a standard deviation was not given
for one of the input variables. The authors simply state the assumption without any defense of it.
The assumption was that the low observed value represents two standard deviations below the
mean and the high value represents two standard deviations above the mean. There is no
scientific basis to defend this assumption, which makes the Monte Carlo simulations suspect as
to what extent uncertainty in Pb dust concentrations has indeed been captured.
For the Monte Carlo simulations, Appendix D states that a coefficient of variation (CV) of 2%
was used if no low and high values were available and also that the variable was assumed to be
normally distributed. Discussions at the July 9-10 meeting brought out that the 2% was really
supposed to be 200%, which is more than satisfactory as a default CV. Other assumptions, such
as assuming children occupy the entire house or the entire yard equally during the renovation
project, are not justified and weaken this reviewer's confidence in the Monte Carlo modeling
E-27
-------
results. Also, the use of 1.2 as the GSD in the Pb lead models is a gross underestimation that
minimizes variability in the predicted outcome variable, and as noted during the July 9-10
meeting needs to be between 1.6 and 2.1 to capture uncertainties in this part of the analysis.
Furthermore, since the real issue is the extent to which the indoor air dust concentrations of Pb
impact blood Pb levels and thus IQ, more attention should have been given to Monte Carlo
modeling that extended to the final endpoint - estimated impact on IQ changes due to the various
types of renovation projects.
Specific Comments:
Page, Line #
Comment
Exhibit 3-6
The use of black and grey does not come through very well in this exhibit
and in others. Perhaps the steps that are discussed in Chapter 4 should be
italicized instead.
Sec. 3.3.1.3
The assumption that the background indoor air concentration equals the
background ambient air Pb concentration is not reasonable. From the PM
CD, we know that not all outdoor particles penetrate indoors and that the
percent that do penetrate is a function of particle size. Thus, the 0.025 jig
per m is too high and should be adjusted downward.
Sec. 3.3.2.2
The practice of setting values below the detection limit to a level of 1/41
of the detection limit is not really defensible. Staff should be using the
statistical methods developed by EPA (Dr. John Creason) for left
censored distributions. The authors should indicate what percentage of
data comprising Exhibit 3-9 is reflected by values below the detection
limit. The fact that 5/9ths of the cells in Exhibit 3-9 are values below the
detection limit compounds the error being made by not using the
statistical methods for left censored distributions.
Sec. 3.3.3.2
What was the method used to determine outdoor soil values and how
could the concentrations be < 0? No defense is provided for setting
values to 5 jig/g when the measured value is < 0. The authors should
describe how frequently this type of adjustment had to be made.
Sec. 3.4.3.1
In the first paragraph of this section, the authors refer to the geometric
mean, minimum and maximum concentrations as being labeled mid, min
and max, and they state that Exhibit A-4 gives these indoor air
concentrations. Only the mid value appears in Exhibit A-4. What is the
range of sample size associated with the various combinations for which
geometric means were generated? If the Ns are small, this presents a
major problem for the subsequent analyses and models that use the
indoor air concentrations of Pb.
p. 43, line 6
A geometric mean of geometric means is described here, but the
calculated mean should be a weighted geometric mean unless the sample
sizes are essentially the same for all of the combinations being averaged.
E-28
-------
p. 44
Sec. 3.4.4
p. 48, line 15
Eq. 4-3
Exhibit 4-3
Exhibit 4-4
Eq. 4-5
Exhibit 4-9
Exhibit 4- 11
Exhibit 4-14
p. 85, line 21
Sec. 5.1
It is noted that loading values for the Rest of Building increased between
2
the post work and post cleaning measurements, from 1 jig/ft to 1.5
|ig/ft . This is probably an artifact of the arbitrary decision to make
values equal to l/4th of the detection limit.
How did staff come up with the example scenario of multiple
renovations? Some discussion of this selection is needed.
The paragraph states that it is unclear whether the assumption of indoor
air concentrations being equal to outdoor air concentrations produces
either a positive or negative bias in estimated blood Pb levels. Indeed,
since not all particles penetrate inside a home, a positive bias should be
expected.
Shouldn't the term in the exponent of e have a negative sign in front of
it? Otherwise the decay constant must be negative because the equation is
relating to estimating concentrations during the Settling phase and the air
concentration should decrease.
If I understand this table correctly, the base control option, which
includes no special cleaning or plastic placement measures, produces the
lowest Pb air concentrations. What am I missing?
Same comments as for Exhibit 4-3 .
What is the adjusted R for this equation? This is needed so the reader
can judge just how good the model fits the data.
This reviewer does not see where Option 3 is plotted.
This exhibit gives non-water dietary Pb intake estimates. Beyond the first
year of life, the values go up and down by year and only vary between
2.6 and 2.99 |ig/day. The question arises as to whether these values are
really different from each other. In addition, their use in the blood Pb
models imparts variability that is not of interest and that is probably not
biologically significant. Consider using the overall average for years 1 to
6, which is 2.76 |ig/day
The reader can not currently tell which is Control Option 2 and which is
Control Option 3. One would assume that the last bar in each set of 4 is
Control Option 3, but the reader should not have to assume anything.
The statement is made that the approach (no effort made to ensure that
children remain outside the work area during renovation) gives a
conservative estimate of exposure. To the contrary, this ensures that a
liberal exposure scenario is used. As an aside, given our societal
knowledge about the harmful effects of Pb, it is unthinkable that a parent
would not ensure that a child is removed from the area while major
renovations are being implemented.
Why isn't the O'Flaherty model discussed here? It certainly has more
positives that the IEUBK model for this particular application.
E-29
-------
Exhibit 5-2
This exhibit is not needed. The text adequately describes the exposure
patterns for the 6 children example.
Exhibit 5-3
The absorption factor for the lung given in the exhibit is 0.42 based upon
a 1989 USEPA publication. This absorption fraction only holds for total
deposition of particles that are 0.5 jim in size. Lung deposition for
children is typically much less than this. There are more recent estimates
of the respiratory tract absorption of particles in children that should be
used (e.g., the MPPD model available from the Hamner Institutes for
Health Science). This reviewer pointed this out for the Pb CD and is
including a table at the end of these comments appropriate for a 3-year
old child. There is no excuse whatsoever to not use the correct absorption
fractions as a function of age and of the particle size distribution that the
indoor air will contain when renovations occur. Moreover, the renovation
activities will lead to a great amount of larger particles that will deposit
in the head. I see no place where URT deposition is taken into account.
The Leggett model should be able to be adjusted to incorporate URT
deposition.
p. 99, line 3
The NHANES data are not appropriate for use as a benchmark to
determine which model (Leggett or IEUBK) provides the most
appropriate modeling results unless the data relate to specific strata from
NHANES. One is comparing apples and oranges here because probably
most of the NHANES houses did not contain lead-based paint.
Sec. 5.5
The discussion about dust ingestion rates and GI absorption likely being
important contributors and sources of uncertainty, while appropriate,
begs the question on a major source of uncertainty that is not being
addressed anywhere - namely deposition in the head of most of the dust
from the renovation projects and the quick translocation of this material
to the G.I. tract.
p. 109, line 8
While it is correct that the log-linear model produces unrealistic results at
a blood Pb of zero, the model should never be applied there. Given the
uncertainties in the analyses being conducted here, the models should not
be used at low Pb blood levels (i.e., < 0.5 to 1 |ig/dL.
E-30
-------
Particle Size
(um)
0.5
1
2
2.5
3
5
7.5
10
12.5
15
20
Regional Deposition Fraction for a 3 Yr-Old Child3
Head
0.208
0.243
0.264
0.27
0.243
0.332
0.443
0.562
0.647
0.686
0.649
TB
0.047
0.047
0.064
0.077
0.047
0.178
0.317
0.324
0.245
0.167
0.074
P
0.164
0.167
0.269
0.312
0.167
0.313
0.133
0.027
0.0027
1.5E-4
7.0 E-7
Total
0.419
0.457
0.597
0.659
0.457
0.824
0.894
0.912
0.895
0.853
0.723
The deposition values are from the MPPD model, which is available from the Hamner Institutes for Health
Science, 6 Davis Drive, Research Triangle Park, NC
E-31
-------
Dr. Maria Morandi
Characterization of Dust Lead Levels after Renovation, Repair, and Painting Activities (OPPT
Dust Study)
Preliminary comments for July 9 meeting.
Maria Morandi
General Comments:
The report presents a detailed description of the objectives, rationale, practical (operational)
limitations due to field constraints over initial protocols, sample collection, and data analysis.
The description of actual field conditions and deviations from original protocol presented in
Chapter 3 and the field experiments in Chapter 4 is detailed and a strength of the Report.
Appendix A is very helpful in providing further detail, as are the rest of the appendices.
However, there is a sense of too much information so that sometimes it is difficult to keep in
mind the essential objectives of the study.
Given the nature and amount of the data collected, it would be useful to have summary of results
and conclusion beyond the very brief Summary presented on page 2-1. My suggestion is to move
the May 24 - June 11, 2007, peer review from the main body of the Report and perhaps place it
in an Appendix together with any subsequent peer reviews. The Summary section should be
expanded in a manner that provides the findings addressing each of the objectives listed on pages
1-2 and 1-3, followed by a paragraph presenting limitations and caveats.
While the conclusions of the study are generally correct in that the rule package did result in
overall lower post-job levels than the baseline conditions as determined from analysis of the log-
transformed data, but there was variability among contractors, workers and jobs, with the
differences not been uniform. These results raise the question of how (and if) the Dust Study
differences may underestimate variability upon rule implementation given the broader universe
of contractors and owners that will not be knowingly part of an experimental study and subject to
the typical time pressures "to get done with the job" Obviously, this could have implications for
the estimation of blood lead levels and IQ decrements.
There is a concern that when the post-job or post-cleaning Pb values do not achieve the EPA
guideline, the values tend to be somewhat dismissed either because of some alteration of the
protocol or housing condition. This is problematic because it is likely that these are highly likely
to occur during actual abatement implementation. The main objective of the rule is to assure
appropriate and safe abetment for all Pb contaminated housing, so if the proposed rule is not
effective for certain activities or residential/building condition, it would be advisable to
E-32
-------
determine what additional modifications to the rule or further modifications need to be
undertaken.
The response to the prior reviewers of the report accurately indicated that analyses of workers Pb
exposures as part of the report was not done because it was not contemplated in any of the study
objectives. However, this is a weakness in the report. As we all know, there were serious
concerns about worker exposures in the asbestos abatement program and it would seem
appropriate to address the potential for excessive occupational exposures, bearing in mind that
PPE principles were probably more strongly adhered to than it would likely occur once
implementation of the rule takes effect, as the Report indicates on page 3-8. It is not quite
convincing that these data cannot be used because of 1KB stipulations since the data analyses and
presentation can be done using personal de-identifiers which is a standard approach for reporting
such findings. It seems a waste of potentially important information from an occupational health
stand point not to use the exposure data, obviously a new 1KB application would be required for
secondary use of the personal data.
Charge questions:
Issue 1) Study Objectives
Question 1) Are each of the study objective objectively and transparently addressed in the data
analysis and conclusion of the report?
The study objectives are addressed in the report. However, because of the profusion of data and
their analysis it would be advisable to provide an objective by objective conclusion in the
Summary. I addition, it would help the reader if the pertinent sections of the report are cited
under each of the objectives (following the same format as in the last paragraph of page 5-3, or
Chapter 7, for example).
Issue 2) Study Conclusions
Question 2. Is each of the study conclusions in the report supported by the data analyses and
other information in the report? If you do not agree that the conclusions are supported by the data
and analyses, please discuss your concerns and if possible, provide specific language to describe
the conclusions.
In general yes, except when the post cleaning levels do not achieve the target levels, then the
data are dismissed.
E-33
-------
Issue 3. Range of Data
Question 3. Do the tables, graphs, figures and other information in the report objectively and
transparently convey the range of the data in the study?
Yes. I found the graphs especially useful.
Issue 4. Report Organization and Clarity.
Question 4. Is the report logically laid out, consistent, and easy to follow?
The layout of the report is logical and consistent. Given the amount of data and analyses
presented, it cannot be described as an "easy read", but it is clear and can be followed and
probably clearer that most reports of this type.
Issue 5. Data Collection and Descriptive Analysis
Question 5. Are the descriptive analyses in Chapter 6 for interior and exterior jobs appropriate
for the study objectives and the collected data? Have the data collection and the descriptive
analyses of the data been objectively and transparently described in Chapters 3, 4 and 6?
As indicated earlier under General Comments, the description of the data collection and field
activities, especially the deviations from the original protocols, are strength of the report and the
authors should be commended for that.
As much as possible, descriptive statistics on paint Pb content should include some indicator of
distribution, as is done in a limited way for other information such as the duration of activities or
cleaning times. The scatter plots are especially illustrative.
Issue 6. Statistical Modeling Results
Question 6. Please provide any specific comments on the modeling analysis in chapter 7. Are the
statistical methods appropriately applied to the data? Are the methods objectively and
transparently described?
The methods are described in a manner that is understandable. What are less clear are the
assumptions used for the choice of statistical analysis or models? /**-*-y concern is that given
the variability in the post-data, modeling based on geometric measures of central tendency may
not be the best approach to evaluate if adoption of the rules will meet the intended objectives.
E-34
-------
Dr. Paul Mushak
POST-MEETING COMMENTS: REVIEWS OF THE EPA-OPPT DUST STUDY AND
DRAFT DOCUMENT ON AN APPROACH FOR IQ CHANGES DUE TO LRRP
ACTIVITIES
Reviewer: Paul Mushak, Ph.D.
There appear to be two Panel review positions, judging from the mix of submitted
comments and views expressed at the 7/9-10/07 meeting. Some members favor setting aside or
greatly adding to this draft. Others favor making enough interim repairs to the draft to make it at
least minimally acceptable for cost-benefit analyses. The repair camp appears to focus on just
getting the draft materials to a certain point, albeit with flaws. The trade-in group focuses on
whether the approach will be adequate for reliable use throughout the nation.
I am more or less in the trade-in camp, but I'm quite prepared to be convinced otherwise.
I still see high barriers to acceptance of the draft approach. Furthermore:
• I admit I still don't quite follow what we're being asked to do for this draft
approach review.
• We first have the general and underlying problem of communication. OPPT
seems to be talking to the Panel in "reg-speak" and "law-speak" while the Panel is
talking to OPPT in "tech-speak." This gets into dialogue problems of non-aligned
vocabularies that complicate effective peer review.
• OPPT appears mainly interested, and has repeatedly reminded the panel of that
interest, in getting the Panel's take on the conceptual and qualitative virtues of its
approach. OPPT is seemingly less interested in the specific outputs of the
methodologies in that approach. OPPT presenters appeared frustrated in the
meeting by the Panel's continuous framing of the dialogue through a battery of
quantitative criteria it considered necessary for review.
• Is the principal need for review of the OPPT's draft approach simply that (1)
OPPT needs the Panel to say whether the approach is feasible or not, rather than
(2) the Panel needs to conclude that the approach can only be judged feasible if it
is judged to be case-accurate or even case-reasonable?
• I believe the Panel generally accepts the notion that adherence of the OPPT
methodologies to quantitative criteria is the critical prerequisite to review sign-
off How rigid the applications of those quantitative criteria have to be appears to
E-35
-------
define the bifurcated views of the Panel. Nonetheless, I don't see how either
OPPT or the Panel can escape the link between conceptual validity of some
approach and quantitative criteria for validating the concepts.
• If the above indivisibility for validation does not govern, then how would one
advise OPPT on any revised approach strategy, not just the one at issue? How
could one tell OPPT that if the end results of the OPPT formulas and potions in
the review draft or revised version do not make quantitative scientific sense, there
is still conceptual validity to the approach?
• I suspect that the Panel would have a more unitary view as to advising OPPT if, at
the end of the day, it concluded (i) one could harvest enough RRP field data
relevant to national scenarios, (ii) one could generate plausible and reliable
environmental media lead data sets for biokinetic modeling, (Hi) one had access to
reliable biokinetic, mechanistic models into which these reliable data sets could
be input, and (iv) one could generate quantitatively reliable health risk dose-
response relationships as IQ changes. One can question whether the first three of
these elements are even present. That is, it is not a matter of the Agency and the
authors merely doing a better conceptual structuring and presentation of existing
data.
Differing views and perspectives can perhaps be better represented to the various
consumers of the review by a simpler illustration. Consider a parcel shipping approach instead of
a health risk causal chain approach. If one asks in broadest terms whether one can have delivery
of a parcel to Seattle by sending it from New York the answer is obviously yes. One also
answers yes to questions that address a set of specifics. Can one get a parcel delivered from NY
to Seattle by shipping via FedEx or UPS Air? UPS/FedEx Ground? U.S. Postal Service?
Someone in NY actually taking it to Seattle by flying there? Driving there? Someone going by
ship with the parcel from NY through the Panama Canal to the Port of Los Angeles and taking a
train to Seattle? And so forth.
• All the above approaches are qualitatively feasible but obviously differ in such
quantitative characteristics as overall cost, effort and temporal uncertainty as to date of
delivery or knowing all of the variables that operate and have to mesh to assure delivery.
• For purposes of cost-benefit analysis, any one of the above scenarios within the overall
goal can be monetized. So, the real question for the shippers is which scenario is
governing, if any?
• One can then elevate the role of uncertainty in the process by saying that (1) the parcels
contain valuable but perishable contents, and (2) if the delivery is not timely, the contents
spoil and the parcel is worthless. We have specified the nature of the parcel at issue,
E-36
-------
regardless of how it's shipped. We then have to stratify the level of uncertainty, i.e., get
quantitative, because the risk of contents becoming worthless increases with the increase
in uncertainty about delivery being prompt enough to prevent the contents becoming
worthless.
• We readily conclude that the odds for preserved value of contents are greatest using an
air express overnight shipment. That is, this choice will result in the least likelihood of a
spoiled, worthless shipment because it is the most temporally certain of the options.
Inasmuch as the other alternatives for delivery of the valuable but perishable contents
have increasingly greater uncertainty, their usefulness for the purpose at hand C assuring
delivery of contents that have not perished and retain their values declines significantly.
• A panel of parcel shipping experts advising on various shipping options would be able to
address qualitative feasibility straight away but certainly would not see simple questions
of feasibility as those that are ultimately governing. They would focus on the best way to
preserve value and utility of perishable shipments. They would never argue that the
method of shipment is irrelevant because the level of temporal uncertainty is irrelevant to
preservation of a parcel content's value. They would particularly never argue that
shipments whose stability and utilitarian value are time-sensitive do not require any more
certainty about beating the clock than those that are not time-sensitive.
In the case of this Panel review, another broad aspect the Panel needs to wrestle with is
the atypical nature of the RRP activity to be regulated vis-a-vis the proposed approach. This was
not communicated very well to the Panel. The proposed RRP Rule, as I understand the Rule's
regulatory purpose, addresses a situation that is the flip side of typical TSCA or even other
regulatory rule making on lead (CERCLA, RCRA, CAA NAAQS for Pb, the Lead-Copper
Water Rule, etc). Specifically, much of lead hazard control rule-making in the Federal
government confronts situations where a lead exposure/health hazard already exists and the point
of the rule is to oversee discernible reduction of the hazard, e.g., lowered 95th percentile or
median Pb-B levels.
In cases of Superfund site evaluations, one of the areas I've been involved with, one
includes a human health risk assessment for the no-action option ("Baseline" risk assessments).
That is, one quantifies hazards to risk populations if an already-existing hazardous waste site is
not cleaned up. For lead, the relevant cohort of exposure children at the site will typically have a
log-normal distribution of Pb-B values and one is especially focused on the most exposed
children. That is, the focus is on those in the upper tail of the exposure (Pb-B) distribution.
The RRP Rule has to address the atypical scenario where an activity, RRP jobs involving
lead-painted surfaces, will produce or potentially significantly increase lead hazards and some
level of associated lead exposures of children rather than reducing them. As Ian von Lindern
correctly noted, the RRP scenario is one where children in older housing with relatively low Pb-
E-37
-------
B levels are at risk to have more exposure, not less, with lead dust-generating RRP work. Taking
Dr. von Lindern's point further, the purpose of the Rule becomes preventing increased blood
lead for those kids in the lower tail of the distribution as much as it is preventing increased Pb-B
levels of children in the upper tail of the distribution.
Toxicokinetically, moving children in the lower tail who have low Pb-B values to higher
Pb-B concentrations not only produces higher health risks but induces higher risks along a
portion of the dose-response curve for IQ loss that is the steepest, i.e., most robust for blood lead
maximal impact. For example, if RRP activities increase a child's Pb-B from 2 to 6 |ig/dl the
impact on incremental IQ point loss is comparatively more than IQ change in the upper but
shallower portion of the Pb-B versus IQ loss curve.
Other serious flaws in the draft approach exist. Collectively, virtually all assumptions
used in the approach have had the net impact on the methodology outputs of understating health
hazards, specifically reduction in IQ, with RRP activity intensity and multiplicity.
The inclusion of sensitivity and similar testings, e.g., Monte Carlo analysis, are useful in
direct proportion to the amount of empirical data available for crunching. The size of the data
sets used here versus the needs of the approach are so disparate that it's not clear how much
value accrues to such analyses. A number of the Panel members also made it clear in writing and
meeting comments that incorrectly used constraints on parameter specifications for doing the
sensitivity and related analyses in the draft report created a host of artifactually predetermined
responses to parameter tweaking.
The two biokinetic models selected for simulation of lead exposure of the target children
are themselves limited for the purposes to which they were put. The IEUBK model, on balance,
is ill suited kinetically and mathematically to the type of body lead changes arising from
transitory changes in lead intakes from RRP activities and would not be recommended for that
purpose.
Note that the absolute and incremental values for Leggett versus IEUBK models
tabulated across the various Chapter 5 Tables track in about a 3:1 Leggett/IEUBK ratio. That
does not validate the utility of IEUBK for very short-term Pb intakes from RRP activities. That
ratio is simply an artifact of the IEUBK model not differentiating between acute/sub-chronic
intakes/inputs to the models and stable, steady-state inputs. That is, with dust lead, inputting a
daily dust Pb intake of X jig to the IEUBK for a three-day Pb intake at that level is perceived by
the model as being the same as a steady-state 6-months of daily dust Pb inputs. It will produce
Pb-B simulations that assume input stability. One fools the model by ignoring the model's
assumption and requirement of stable, steady state lead inputs for proper use. The model's
exposure module does not filter or censor for temporal stability of environmental inputs.
E-38
-------
The IEUBK model is still valid for background Pb-B from preexisting (pre-RRP or
stable) dust lead based on its close match to the relevant NHANES III, Phase 2 Pb-B descriptive
statistics for Pb-Bs of U.S. children 1-5 years of age in variably-aged housing.
The Leggett Model also presents problems for this particular OPPT draft approach. One
problem is that the Leggett model starts with the background, pre-RRP Pb-B distributions at too
high a value, based on the above relevant strata of NHANES data. Leggett also shifts
incremental changes in Pb-B with various RRP activities to the shallow higher portion of the Pb-
B vs. IQ curve, i.e., > 10 |ig/dl. We then have an artifactually created lower IQ vs. Pb-B
relationship compared to the curve segment below 7.5 |ig/dl Pb-B. Sean Hays is correct that the
increments from Leggett are not the absolutes from Leggett, but the increments from Leggett are
numerically higher for some Pb intake change than the increments from the O'Flaherty and
IEUBK models when applied to stable, steady-state conditions. The comparison is evident from
the text material evaluated by the EPA All Ages Lead Model evaluation Panel where all three
models are compared and contrasted with empirical data.
On balance, we cannot say that the actual Pb-B is bounded by the two model estimates.
There are no reliable bookends and, even if there were, the models would be bracketing outputs
as Pb-Bs that are derived from modeled, not measured, data. Also, the bracketed modeled Pb-B
values would have a range that is arguably very small when superimposed on the overall huge
uncertainty to Pb-B values propagated from start to finish in the derivation of both Pb-Bs and the
IQ changes from those Pb-B changes.
Use of More Empirical Data
The Panel would probably be more comfortable striking a compromise if OPPT used
more empirical data at the various steps along the way, either to augment the modeling or, more
appropriately and typically, to serve as higher priority for uses in the approach. Modeling is
typically validated and calibrated with empirical data, rather than vice-versa. This is not
inviolate. One explanation of disagreement between measured and modeled data, as I noted in a
1998 paper in EHP, is that in fact both are wrong, but each is wrong for different reasons.
The Panel discussed some empirical data options for either expanding the draft approach
or giving it higher priority. Attention was given to more use of the Lanphear data that have
within the measurements useful information about remodelings.
I believe the best choice for determining background Pb-B for variably-aged housing are
the relevant NHANES III, Phase 2 values. One cannot use NHANES national strata or
aggregated overall snapshots of national lead exposures for specific geographic and demographic
segments of the nation, e.g., for children who live in Baltimore or Atlanta, but one can use the
strata in the national NHANES picture for comparisons to OPPT approaches that are also for
national strata. GM values (Table 1) and % exceedences. 10 |ig/dl (Tables 1 and 2) in Pirkle et
E-39
-------
al., 1998 for housing in the age bands "< 1946", "1946-1973", and "After 1973" give the
corresponding NHANES GMs for 1-5 year old children of all races, income and urban status:
3.8, 2.8 and 2.0 |ig/dl respectively.
Ref:
Pirkle JL, Kaufmann RB, Brody DJ, Hickman T, Gunter EW, Paschal DC. 1998.
Exposure of the U.S. population to lead, 1991-1994. Environ. Health Perspect. 106: 745-
750.
I would suggest the OPPT authors also have a look at two papers that described lead dust
control steps as part of interim lead paint hazard controls on a national rather than localized
basis. These would be equivalent to the no plastic/Rule cleaning, i.e., Control Option 1 for the
draft approach but would reach to national results.
One focused on interior repainting of deteriorating lead paint areas and window
replacement as used in multiple programs around the nation as part of the HUD Lead Hazard
Control Program described in Galke et al., 2001. The national research projects summarized as to
hazard reduction activities and their impacts on Pb-B over time were not lead paint abatements
per se. That is, the projects in Galke et al. would have useful information about lead dust control
options for RRP activities.
Galke et al. reported on Pb-B versus dust lead changes longitudinally for 1212 dwellings around
the nation. The most common interventions in their Table 1 included repainting or repainting in
combination with window replacement. This was followed by cleaning equivalent to the Rule
cleanings. The changes in dust lead loadings with types of activities are given in their Tables 2
and 3.
Ref:
Galke W, Clark S, Wilson J, Jacobs D. et al. 2001. Evaluation of the HUD Lead Hazard
Control Grant Program: Early Overall findings. Environ. Res. 86: 149-156.
The second paper is that of Ettinger et al., 2002, where a professional interior dust lead
cleaning protocol was employed to reduce interior dust lead levels using HEPA cleaning and
two-bucket mopping with wipe cleaning afterwards. Again, it's equivalent to the no plastic/Rule
cleaning or Control Option 1 in the draft. A total of 765 homes were given dust lead clean-ups,
and 213 of these were tested for pre-and post cleaning dust-lead loadings. Cleanings were
especially effective where pre-cleaning lead loadings exceeded the EPA and HUD dust lead
standards for floors and sills (and troughs for clearance testings).
E-40
-------
Ref:
Ettinger AS, Bornschein RL, Farfel M, Campbell C, et al. 2002. Assessment of cleaning
to control lead dust in homes of children with moderate lead poisoning: Treatment of
lead-exposed children. Environ. Health Perspect. 110: A773-A779.
E-41
-------
Dr. Michael Rabinowitz
Review by Michael Rabinowitz of
'DRAFT FINAL REPORT ON CHARACTERIZATION OF DUST LEAD LEVELS AFTER
RENOVATION, REPAIR AND PAINTING ACTIVITIES" Jan 23, 2007 Draft
General Comments:
Overall, this report convincingly shows that lead contamination and exposure from old lead paint
is more severe from some activities than others, and that some appropriate work rules can
significantly reduced that exposure. The series of sites and measurements that comprise the data,
their analyses, and presentations all were executed in a way that should bring real useful clarity
to our understanding of this topic. It was clearly written, although I offer below a few
suggestions.
It might be helpful to know to what extent the specific sites studied here are representative of the
range of situations found across our nation. Climate, land use patterns (lot size and zoning, for
example), historic use of lead paint and building stock (type, materials, age, for example) all vary
even more widely than the variability studied by Battelle here. I raise this not to in any way
invalidate this study but rather to ask if we can expect similar results as these proposed rules are
applied nationwide.
I can only guess the variability and magnitude of the results if they were to be collected across
sites nationwide. How effective one method or another might be in reducing risk might depend
on humidity or how air tight or well ventilated a home might be or proximity to their neighbor.
But I would also guess that these would be relative minor effects compared to the main findings
here. The over pattern of risk reduction with rule implementations demonstrated by this Report
likely do apply nationwide.
Lead containing paint is a hazard. Over time, as it disintegrates slowly or rapidly from RRP, it
generates toxic dust. Since some of the RRP methods appear to generate much more of this dust,
one might ask if it would not be better to leave intact paint alone rather than perform the RRP in
a way that causes a short-term problem. To avoid answering this affirmatively we should be sure
the recommendations for RRP rules are sufficiently rigorous to mitigate any harm.
Some Specific Responses to the Text:
Page 1-1 line 5 For what its worth, allow me to note a further, perhaps un-necessary,
description : Because of formulation changes in American paints, not only was lead paint more
E-42
-------
common among the older homes (as is noted in the text), but the lead content of that lead paint
was much higher. For example, white ready-mix house paint in 1940 averaged 4.1 pounds of
lead per gallon compared to 2.1 pounds per gallon in 1948. (Larson, L. P. and J.H. Calbeck
Comparison of Pre-and Post- War White Outside House Paints, Paint, Oil & Chemical Review,
June 9, 1949, vol. 112, p25-50). Older paint formulations from the 1920's and 30's tended to be
even higher in lead content, and during the 1950's the fraction of lead in the lead pigments
continued to drop.
Page 1-2 This list of Objectives is admirable. I look forward to seeing it in the conclusion as
list of these along with the findings.
Page vi " Lead-based paint" are defined here as anything above /^ percent lead by weight. I
recognize that this is well accepted jargon, even in statute. However, with a view to accuracy I
feel compelled to object (likely fruitlessly) to this wording.
To me "Based" means "made of or "fundamental to". The term does apply when lead was the
major ingredient, more than half of the mass of the liquid paint, maybe four pounds per gallon.
Lead was the only or major source of the paints opacity and hiding power as well as a white base
to which colors or tints could be added. Lead also acted as a drying agent and gave the dried film
its longwearing, waterproof plastic body. However, when paint is found with less than a few
percent lead, as when some yellow or green lead based pigments or when lead napthenate drier
are used, it does not meet my criteria for being " lead-based". Lead is a minor ingredient. I
would have preferred "lead-bearing" or "lead-rich" paint, or "lead paint". But using the broad
term "lead-based" is so imbedded in our usage and recognizing it is only a matter of semantics
without substantive difference, I am close to surrender on this point. I would be pleased to see
the term "lead-based paint" replaced with "lead paint".
Page vii Specialized cleaning- Regarding the wet moping- Is it worth noting somewhere that
neither phosphate not EDTA wash waters were used. Perhaps the use of these has been
addressed elsewhere?
Page 2-1 line 2, between ... .results in lower
I would invite you to add here a descriptor of the magnitude of the reduction such as
"Somewhat", "slightly", "much "(my favorite), or "markedly "or ....
In chapter 9 where this appears again on page 9-1, a descriptor of magnitudes would be useful.
So much is said about p values and statistical significance, something more about effect size
would be welcome.
Page 3-3 Ambient Air Sampling- it looks like air volumes were so small that detect-ability
becomes a concern. Because generally air values were shown to be low, this is of less
E-43
-------
importance. I am curious what the limiting factor was. Could the air flow rate or the sampling
time be increased? Were the Samplers to noisy?
Page 3-9 Worker blood lead monitoring was required for worker protection. However, unless
the workers were generally assigned specific jobs (some always sanded, others always door
planed...) I suspect nothing would be found that might help us assign risk to each task or rule.
Still, I would be curious what the impact of this de-leading work was on the workers, how much,
if at all it went up as a group, and if any individual's stood out. I hope not.
Section 4 was especially well written. It seemed all reasonable. I particularly like the section
starting on page 4-6 about occasional deviations from the protocol. It gives the reader a much
clearer idea about what actually happened and a sense of the complexities.
Page 5-2 Fourth paragraph. Can you tell us what fraction of the environmental samples required
imputation? It may have been very low for the dusts but much higher for the airs. Could be also
addressed in Section 8?
Section 6
Is it worth making a multiple regression model of dust lead using paint lead, area of work,
remediation method (job), and phase, for example?
Page 6-6 Figure 6-1 If the independent variable is the area disturbed, why not have that on the
x, horizontal axis? The line would curve up sharply and then more to the right and become
flatter.
Page 6-10 Tables 6-5 and 6-6 I could use to illustrate one of my own personal peeves, the fear
of rounding off. Do we really need 5 or even 8 significant figures? I would not think any less
of Battelle's precision if they reported the total average lead wipe as 23.7 mg or 23,800 jig
instead of 23,770.7 jig. It is a form of clutter to me. Also, the tabulated precision exceeds by
orders of magnitude the analytical precision on which this is all based. So, I suggest rounding
off to 3 significant figures.
Figures 6-2, 6-3, 6-4 , and 6-5 present some of the most significant findings in a clear manner, so
I would assume they would be selected to be carried forward to any summaries. How can they
be improved? Is a sub-title appropriate, with information, taken from the text, that might allow
this figure to stand on its own more? Something about N's? Maybe some error bars? Reminder
that they are logs?
This repeats in Figures 9-1 and 2, page 9-1.1 want to encourage the authors to elaborate more if
they like.
E-44
-------
Page 6-23 Third bullet point: Would you speculate that the smooth plastic surface encourages
re-suspension compared to a rougher surface?
Page 9-6 section 9-3 line 2. Again I'd encourage a word about the magnitude of the effect size.
.... Yielded much (?) lower lead levels...
Responding to the Specific Questions of the May 23 Memorandum:
Question 1. This could be better, making it easier for the more casual reader. Section 1.3 clearly
listed the objectives. Could this list of these questions be duplicated with corresponding answers
to the questions in Section 9? Maybe as a Table?
Question 2. No, none came to my attention. I could have missed one.
Question 3. Yes, generally they are up to the task.
Question 4. Yes, this was one of the report's strengths.
Question 5. Yes, I think they are. Was there any more they could get out of the data? No.
Question 6. This is outside my area of expertise, but I would ask if it is worth making a multiple
regression model of dust (or soil) lead using paint lead, area of work, remediation method Gob),
and phase, for example.
After-Meeting Comments by Michael Rabinowitz
17 July 2007
The Dust Lead and RRP Study
1) Regarding the "white glove test":
Although it was an easy target of ridicule, having a simple field test to see if lead
contamination remains after clean up would be a major component of the national de-
leading effort.
E-45
-------
A variety of colorimetric field test kits for lead are available. Check your closest
hardware store. They are the size of small crayons and disposable. Many are based on
dithizone and give a color change from green to red, others using sulfide give a black
reaction to lead. As consumer products they have been tested for specificity and
sensitivity, and the sulfide method is generally disfavored. These kits are especially for
lead in paint, rather than soil lead.
My recommendation is that before deleting the whole verification step, consider using
these spot tests to visualize the lead, then mark any hot spots with bright orange spray
paint.
2) It is worth re-iterating:
Lead containing paint is a hazard. Over time, as it disintegrates slowly or rapidly from
RRP, it generates toxic dust. Since some of the RRP methods appear to generate much
more of this dust, one might ask if it would not be better to leave intact paint alone rather
than perform the RRP in a way that causes a short-term problem. To avoid answering
this affirmatively we should be sure the recommendations for RRP rules are sufficiently
rigorous to mitigate any harm.
3) Regarding air-lead:
We heard a lot about how the missing values were and should have been handled, but I
wish we heard more about why the air samples were lacking. Was the pump too noisy?
Too low a flow-rate? Too short a sampling time?
The Dust Lead from RRP and 10: Proposed Methods
1) Overall, I was heartened by the skepticism of the panel towards this undertaking.
2) Modeling Blood Lead:
a) Period of Vulnerability:
There seems to be some growing consensus that it is the contemporaneous or current
blood lead level that best correlates with these subtle neuro-cognitive deficits found at
these lower levels of exposure, aside perhaps from the perinatal period of increased
vulnerability. The matter is still murky because we must rely on longitudinal surveys of
children and most children's blood lead levels tend to track within the same exposure
category. So it is hard to resolve a specific time of vulnerability if one existed.
b) Fetal red blood cells:
The perinatal period is unusual because of both the staging of the development of the
central nervous system at that time, and the higher affinity of fetal hemoglobin for lead
compared to adult hemoglobin, which appears soon after birth. The brain competes with
red blood cells for circulating lead, so the higher affinity of fetal blood and the higher
hematocrits at birth mean that a given numerical value for whole blood lead at birth over-
estimates the brain lead level compared to the same blood value at a later age. Despite
E-46
-------
this distortion, which would give a falsely low estimate of brain lead for a given fetal
blood lead, an increased more persistent vulnerability has been reported for this age.
c) Modeling brief exposures:
RRP can cause a pulse or spike in lead exposure and lead absorption. Have we
adequately addressed the question of if the mathematical models for IQ loss, which are
based on blood lead prediction, apply well enough to the situation of short burst of lead
exposure? I ask this because with brief exposure more of the dose goes to the brain than
under more steady state conditions, because the plasma lead levels are briefly elevated?
Does our choice of bio-kinetic model take this into account?
3) Using Older Data:
Glad some older studies of blood lead in residences will be utilized to see if the proposed
modeling of RRP and blood lead perturbations gives reasonable values, according to the
lead level in the paint.
4) Assigning an Age to a House:
Whether it is based on the age of the foundation or of major structural additions, I suspect
HUD might have some general rules they used for NHANES and other surveys. You
need only do the same and follow their guidelines.
E-47
-------
Dr. Joel Schwartz
Comments of Dr. Joel Schwartz
Professor of Environmental Epidemiology and Director, Harvard Center for Risk Analysis
Harvard School of Public Health
Boston, MA
August 17, 2007
I would just like to emphasize a few points.
a) For regulatory purposes, EPA needs to set clearance levels so that there is a target for
renovation. That is quite distinct from assuming that there is a threshold at those dust
concentrations in a benefits analysis. Since children with a wide range of baseline blood lead
levels will come into contact with this dust, even if there were a threshold blood lead level below
which no benefits of further reduction in exposure occur, this would not be the case for dust lead.
Hence the benefit analysis should never assume a threshold dust lead level. This should be
clarified to EPA, and reflected in their benefit analysis for OMB. The point is that if some of the
time, the application of appropriate clean up standards produces dust lead concentrations that are
below the clearance level, that will produce additional benefits. Since clean up techniques are
presumably targeted to meeting the clearance standard almost all the time, they will in the
majority of cases produce these additional benefits, which need to be quantified and set off
against the costs.
b) We need to note that there are additional sources of variation that occur when one applies the
Leggett model to episodic exposure. There is considerable variation from day to day in many of
the parameters, both behavioral as well as physiologic, which determine lead uptake. Variations
in hand to mouth activity, iron status, febrile status, etc. These all average out when applying a
biokinetic model for a longer period. But when the exposure is episodic, there is a wider
distribution in the population of all of these factors, both related to intake as well as the
parameters of the model. Hence while the Leggett model may adequately predict the mean of a
group of children exposed, to predict the 90th percentile, a wider range of variation of all
parameters will need to be assumed. The results of the empirical studies we cite could be helpful
in determining that range, since they give distributions of blood lead levels, not just means.
c) I support the idea of Dr. Lanphear on how to use the empirical studies to do a benefit analysis.
d) I also support the use of the two linear slopes instead of the log-linear model.
e) I would like to see Monte Carlo analyses address not just uncertainty in the mean estimate of
benefit, but be used to estimate the distribution of effects in a population with varying initial
blood lead levels, varying dust ingestion, varying absorption parameters, varying lead kinetics,
etc.
E-48
-------
Dr. Frank Speizer
Draft final report on
"Characterization of Dust Lead Levels
After Renovation, Repair, and Painting Activities
January 23, 2007
Study and Report Background
May 23, 2007
Submitted by Frank E. Speizer
Date: June 21,2007
General Comment: I found the report to be an excellent summary of a series of experiments that
appear to have been well thought out and executed in manner consistent with a rather complex
protocol. The experiments follow a basic design that has provided information that will be
useful in setting up RRP guidelines and rules that should lead to protecting both contractors as
well as less formally trained self home renovators. There are a number of minor issues, which
should be discussed perhaps in more detail in the concluding chapter of the document. The only
disappointment is that no results are reported on the blood work of the workers engaged in these
experiments. I would assume the reason is that the study was deemed to not require Human
Studies Approval (this would not have passed my IRB committee) and for this reason no
personal exposure data or blood work information was retained. However, I believe that this was
a mistake and surely lessen the value of the entire enterprise, particularly if the workers involved
in this work carried out only this work during the periods of observation when blood was
obtained pre and post exposure. Such blood lead changes that were found, and I assume were
reported to the individuals as indicated page 3-9, could have been correlated with the air and
wipe measures to estimate the impact of short term accumulated exposures in well defined
settings. I would recommend that, if the data do currently exist, which I guess by IRB standards
they should not, the investigators consider ways get IRB approval and make such data available.
Detailed questions:
Issue 1: Study Objectives: The objectives are well stated and seem to all have been met.
Issue 2: Study conclusions. The section of the report described here seem to be no more than an
outline of what follows particularly in Chapters 6,7,9, with very little in the way of factual
material. If it is meant to be an executive summary more detailed material needs to be included.
The results of these experiments are impressive and important. Both here in Chapter 2 and in
Chapter 9 rather than repeating some of the finding it would be useful to provide specific
interpretive statements of what the experiment showed and how these results should direct both
rule makers and renovators, not subject to rules, to change their practices. Regarding the final
italicized comment in this section I strongly disagree with the reading of the 40 CFR rules in that
this would have been true only if blood was not drawn. By drawing blood, the situation is no
longer "usual work practices".
E-49
-------
Issue 3. Range of Data: The study protocol is well described. On page 3.4 there is a statement on
deviations. A brief table indicating frequency within each of the approximately 75 experiments
and in which portion of the experiment the deviation occurred would be informative.
Issue 4. Report organization and Clarity: The tables are well laid out. I suggest to go along with
section 4.5 on pages 4-6 and 4-7 a simple table by the 6 categories of protocol variations
indicated that the number of times or samples that deviated from protocol be indicated. If the
numbers are small the table might even have a "Comment" column giving details of the
violations. Bottom of page 4-9, back to my general comment of why the investigators should
have had Human Studies Approval.
Issue 5 and 6: Descriptive and Statistical analyses: There are quite a few tables and this seems
ok. The data are very interesting but what gets lost is the magnitude of the impact of the
potential revised rulings. Many of the graphs are log based and the differences between baseline
and experimental settings and pre and post settings are described in the text in terms of p values.
Whereas these are quite large changes in dust levels of lead and are important improvements
(most appear to be at least log different). For the lay reader or rule maker it would be more
impressive to indicate the magnitude of some of these changes rather than p values.
Pre-Meeting Comments on: "An approach for estimating changes in Children's IQ from lead
dust generation during renovation, repair, and painting (RRP) in residences and Child-Occupied
Facilities (COF)."
Submitted by: Frank E. Speizer
Date: July 3, 2007
This is a strong document that generally is well written and does, indeed, cover many of
the issues and discusses quite well most of the uncertainties contained in what I would consider a
totally flawed method. I think it is important that the document was produced, but I totally
disagree with considering it as a basis for carrying out an economic analysis as part of the rule
making procedure. Unless one finds it acceptable that "garbage in equals' garbage out" I do not
see how what has been produced can be used in any further economic analysis. What is clear is
that disturbing particularly older residences or other COF results in mobilizing lead into various
medias (air, food, dust, soil, etc.). Lead exposure results in changes in blood lead levels in
children. Elevated blood levels affect neurocognitive development. These are facts and it
appears that the later effect is linear and particularly strong below levels of 10 |ig/dL. On the
other hand with regard to the effects of RRP the uncertainties described in the methods and the
paucity of data that go along with those uncertainties described throughout this document along
with the lack of generalizability of the examples used lead me to the conclusion that to use the
data in any way other than qualitatively would be a disservice. Quite frankly, my reading of
Fehrenbacher's cover letter to the Charge Questions suggests concern that the proposed methods
may not be appropriate.
E-50
-------
As pointed out throughout the text, there are uncertainties. For example, in using the
results of the OPPT dust study at the end of chapter 3, the authors point out that the data included
Pb loading for only a limited portion of the expected sources for each job. This clearly presents
a problem in using the data. Further, in accepting the data from the selecting households for the
OPPT studies, one is restricting the possible scenarios that might apply in different setting and
different parts of the country. Houses built before 1940 (and still standing), may have undergone
significant renovations in the past producing potential higher soil levels outside with perhaps
totally eliminated indoor sources, and any combination of these scenarios. It is not clear that
further renovations of such residences or COF will result in any addition exposure or more
exposure. Thus to apply the data from the single and multiple renovations studies seems too
simplistic.
As indicated on page 122, particularly with regard to the COF, the estimation of a single
concentration for each building scenario makes unlikely assumptions about children's activities.
THE METHOD NOT DISCUSSED (for a future economic analysis). What is not contained in
this document is a discussion of the effects of removing children from buildings designated for
potential exposure during the duration of RRP and potentially for some time after completion of
work. What would be the cost of taking children from one day care facility to another while the
first is being renovated, and perhaps for a month or more after completion while dust control
abatement is going on, rather than suggesting that they be in a "non-exposed" room next door to
where the activity is going on?
BOTTOM LINE: My concern is that any use of these data for an economic analysis would be so
flawed as to be discounted from both industry and environmentalists and thus it has no part to
play in the rule making process.
I hope that the discussion of these documents at the CAS AC meeting this next week can put the
data in a more favorable light and I can be persuaded to be more positive about how the data may
be used.
Post-meeting comments for potential inclusion in Dr. Henderson's summary letter.
From: Frank E. Speizer
July 18,2007
I have reviewed Paul Mushak's Post-meeting comments and he summarizes well the
problems we all faced in reviewing this material. My specific concerns, expressed in my original
pre-meeting comments certainly are better expressed by Paul's comments. I remain essentially
unconvinced that there exist enough data to carry out the kind of analyses proposed and to have
any credibility given to those analyses. The analysis certainly can be done, but I fear it will end
up with a series of numbers that will somehow be turned into a series of policy relevant
E-51
-------
economic statements with cost/benefit ratios attached and the degrees of uncertainty attached to
those analyses will be lost.
In spite of all the work that went into gathering the data that will be used in these
analyses, the technical people "turning the cranks on the computer" seem not to understand that
the data they will be using was gathered in a convenience sample of selected homes in two cities
with very specific characteristics, yet unknown histories of prior renovations, and there is no way
that these data can be representative of what might be happening on a national sample of homes,
renovation activities, or living arrangements. To then try to apply these data both to homes being
renovated and then to other COFs (nurseries, day care centers, schools, etc.) where children less
than 6 year old might be housed for varying periods of time from hours to days, is simply beyond
the scope of the available data.
Note that as I read the text there is a general admission that there are insufficient data on
COFs and that the plan to extrapolate from the residential data where a child's time is assigned
uniformly to the whole building indoor dust and air concentrations and the whole yard outdoor
soil concentration to characterize exposure can be modified for COFs by adjusting the media
concentrations to account for where children spend their time when in a COF simply won't work.
This would be appropriate if we had a reasonable handle on where children spend their time in
the potential wide variety of COFs that exist, but I am afraid that such data simply do not exist.
From the discussion at the meeting it would appear that the marching orders given to this
analysis group was to move ahead with analyses as planned and thus meet the "legal"
obligations, so that something can move forward in the rule making process. The choice of
having CASAC be the review body for this procedure is probably one that the OPPT group now
regrets, as it is unlikely that they will be able to overcome our criticisms of the sources of data
being used in the approach, as well as specifically which models to use in the approach. I remain
open to be convinced otherwise, but thus far I am not. As Paul points out the existing models,
even if they were all used would still not bound the limits of effects, and would still be a
potential underestimate of effects.
I have great respect for the effort that went into both the gathering of the data by the
Battelle group, and the way in which the data were written up. Further the discussion of the data
and its potential use and uncertainties by the OPPT group is quite impressive and in and of itself
is an important piece of work. However, by its own admission and the discussions in chapter 7
on pages 121-123,1 believe the authors provide sufficient evidence that a quantitative use of the
data is simply inappropriate and thus they must turn to more qualitative approaches based on the
available empirical data.
Finally, the closing remarks made at the meeting left me somewhat disappointed. The
suggestion was that we had been asked to comment on the methods as presented (not to offer
alternative approaches). I am sure at that point everyone was tired and perhaps becoming
somewhat 'testy". However, I believe we are all on the same side and want a final product that
will withstand criticism both from the industrial trade groups as well as environmental activists.
To get there will require EPA to be more creative in accepting our criticisms as a good faith
effort to be constructive yet realistic as to what can and cannot be accomplished. Since there is
not legal mandated deadline for the economic analysis, as there is for the rule making, I would
E-52
-------
suggest that a less formal economic analysis be done and the rule making move forward using
more qualitative assessment techniques for establishing more qualitative risk/benefit ratios,
perhaps using "willingness to pay" techniques.
E-53
------- |