-------
I
(5
I
(3
•i
10
i
If) .O
^ I
a '>
I
oJ a
00
o>
II
c
0>
9-1
tO V
09 C
O
b
o
15
4M
tt
I?
2-*
C CJ
« r»
o«
CJ C
2
o
e
o
o
CD
a.
•
" a =
S >g
^ •"* «l *fc
n] •— 0) A
5 g z 1,
: i
•••
' .S .9> o "5
1 Q..C S 5
•o «
150
-------
1998). Most quantitative estimates of the amount of soil ingested that were reviewed in this
report were obtained using the mass-balance/tracer element approach.
Incidence rates (i.e., prevalence) of pica for soil in young children may be estimated from
parental questionnaires. This approach can yield biased results, however, as it relies on the
observation and accurate reporting of pica by the adult. In addition, the response depends on the
wording of the question. Various surveys have asked whether the child eats dirt (Annest and
Mahaffey, 1984; Bates et al., 1995; USEPA, 1996a - Boston and Baltimore portions), whether
the child puts dirt or sand in mouth while playing outside (USEPA, 1996a - Cincinnati portion;
USEPA, 1996c), or whether the child puts fingers or toys in mouth while playing outside
(Barltrop et al., 1974). Clearly, these questions would elicit different responses from the same
caregiver. In addition, response choices may be simply yes or no, may specify a timeframe
(e.g., in the past month), or may be open-ended. These choices, too, would result in differing
responses. Thus, care must be taken in comparing soil pica prevalence rates originating from
parental questionnaire data.
While mass-balance studies provide soil ingestion rates to support prevalence data, these
studies are also subject to error and have disadvantages. For example, Calabrese et al., (1989)
acknowledge that analyzing chemical tracers without the use of a mass-balance approach
(i.e., not correcting for intake) can result in soil ingestion estimates that are increased by factors
of 2 to 6. In addition, the particular tracer used, the duration of the study, and the frequency of
sampling may also influence reported results (Calabrese et al., 1989; Calabrese et al., 1997). For
example, the short duration of most mass-balance studies makes it difficult to determine a
"normal" rate of soil ingestion for a child. Approaches using chemical tracers also have
disadvantages in that they are more expensive and generally have small sample sizes.
3.3.2 How Does the §403 Risk Analysis Account for Soil Pica?
Within the exposure assessment (Chapter 3) portion of the §403 risk analysis report
(USEPA, 1998a), soil was considered an indirect source of lead exposure, although summary
information on soil pica frequency from two lead exposure studies was presented in Table 3-3b
of the §403 risk analysis report.
An indicator of soil pica was considered as a candidate predictor variable in the
development of the empirical model for the §403 risk analysis. The soil pica variable was based
on the parental questionnaire administered hi the Rochester Lead-in-Dust study. This variable
measured the child's tendency to put dirt or sand in the mouth using a scale of 0 (never) to
4 (always). The soil pica variable was borderline significant in single media models, which
assessed the relationship between blood-lead concentration and each predictor variable under
consideration. These single media models were the first step in developing the empirical model.
Variable selection for the multimedia exposure model was based on several properties, including
the strength of the relationship with blood-lead concentration as estimated using the bivariate
statistical models, predictive power of each variable when included into a model with competing
sources of lead exposure, and interpretability of parameter estimates. The soil pica variable was
151
-------
dropped during this phase of the empirical model development. Additional information on the
development of the empirical model can be found in Appendix G of the §403 risk analysis report.
Age-dependant soil and dust ingestion rates for the Integrated Exposure, Uptake, and
Biokinetic (BEUBK) model were taken from the IEUBK guidance manual (USEPA, 1994) and
represent central tendencies within the range of values seen in different studies. Combined soil
and dust ingestion amounts ranged from 85 to 135 mg/day, as shown in Table 4-1 of the
§403 risk analysis report, of which 45 percent was assumed to be from soil. Thus, soil ingestion
was assumed to be between 38 and 61 mg/day for children aged 0 to 7 years, with the maximal
ingestion estimated for children aged 1 to 3 years. These ingestion rates are consistent with
estimates of inadvertent soil ingestion presented in this report, but are not representative of pica
episodes. While IEUBK model predicted blood-lead levels were adjusted in the §403 risk
analysis to allow consideration of paint pica in homes with damaged lead-based paint, as
described in Section 4.1 and Appendix Dl of the §403 risk analysis report, no such adjustment
was made for the effect of soil pica.
It should be noted that while neither model used in the §403 risk analysis had explicitly
accounted for soil pica as a separate factor independent of paint pica, the impact of soil pica was
included in the analysis as part of the relation between soil-lead concentration and blood-lead
concentration which the analysis characterized.
3.3.3 Prevalence of Soil Pica Behavior
Estimates reported in the scientific literature of the percentage of children who ingest soil
are summarized in this section. From the literature, it was not possible to estimate the percentage
of children who exhibit pica for soil but not paint The §403 risk analysis did account for the
effect of paint pica on blood-lead concentration estimates. For children who ingest both paint
chips and soil, it is reasonable to assume that the effect of soil pica is insignificant compared to
that of paint pica. Thus, in estimating the percentage of children who ingest soil, it is important
to exclude those who also ingest paint chips. It was possible to estimate the percentage of
children who exhibit soil pica, but not paint pica, using information from parental questionnaires
administered in the USLADP study (USEPA, 1996a), Baltimore R&M study (USEPA, 1996c),
and Rochester Lead-in-Dust study (USHUD, 1995a; Lanphear et al., 1996a). This information is
also summarized in this section.
3.3.3.1 Literature Review. Most sources in the literature reported prevalence rates for
general pica behavior (mouthing or eating non-food items) or for soil pica (eating dirt). One
source (Stanek et al., 1998) reported pica rates for a variety of specific non-food items (soil, paint
chips, paper, toys, etc), but did not cross-tabulate [e.g., 18 percent of children ages 1-6 years, as
assessed by parent interview, were reported to ingest/mouth dirt at least monthly and 3 percent to
ingest/mouth paint chips, but information was not provided on how many children eat both soil
and paint chips (Stanek et al., 1998)]. An overview of selected studies estimating pica
prevalence is shown in Table 3-25 above. It is important to note that in the cited studies, various
definitions of "pica" were used in reporting the prevalence of pica behavior. For example,
152
-------
Greene et al., (1992) defined soil pica only as the occurrence of soil ingestion and reported the
percentage of children who ingest soil based on caretaker interview. In comparison, Calabrese et
al., (1997) defined soil pica as consumption exceeding 0.5 grams per day and reported the
prevalence of pica behavior as assessed quantitatively by mass-balance methods. It should be
noted that only primary research findings are reported in Table 3-25, with the exception of
Calabrese and Stanek (1993). Several review articles were also obtained. These were excluded
from Table 3-25, as insufficient details of the source studies were provided. General findings of
the review articles are cited in the text.
Table 3-25 shows that the estimated percentage of children ingesting soil ranged from
1.6 to 62 percent and varied with definition/criteria for soil pica used, age group of children, and
socioeconomic status. In general, 12 of 16 observations in the table report a prevalence of soil
ingestion in children of 13 percent or lower (where, at a minimum, limiting criteria are defined as
ingesting soil at least once). "Normal" mouthing behavior, however, is typically exhibited more
commonly, particularly in the younger age groups. For example, Barltrop et al., (1974) reported
that 43 percent of children exhibited pica defined to include mouthing behavior, but that
9 percent were estimated to swallow soil. Stanek et al., (1998) assessed non-food ingestion and
mouthing behaviors in 533 children, ages 1 to 6, by parental interview. Results of the survey
indicated that 38 percent of 1 year old children and 21 percent of 2-year old children
ingest/mouth soil at least monthly. In contrast, at ages 3 to 6 years, less than 10 percent of
children were observed to ingest/mouth soil at least monthly. At age 1 year, 11 percent of
children were observed to ingest/mouth soil daily compared to one percent or less among
children aged 2 to 6 years.
Of the studies that used mass-balance methodology, prevalence of soil pica ranged from
1.6 to 20.8 percent. For studies that employed parent or caretaker interview methodology, soil
pica prevalence ranged from 3.2 to 19 percent, although one study reported a rate of 62 percent,
which appears to be more consistent with studies that monitored general mouthing behavior. For
both methodologies, the prevalence of soil ingestion tended to be higher in the younger age
groups and for children in families with lower socioeconomic status. Although, as mentioned in
Section 3.3.1 above, techniques utilizing parental interview are generally considered less reliable
than quantitative methodologies, the issue of consistently defining "pica" when reporting study
results is an issue not only in studies using questionnaire methodology, but also in the
mass-balance/chemical tracer studies. Differences in values reported, both between and within
the various assessment techniques, may largely be due to differences in how pica behavior is
being defined in the study. Thus, many studies estimating soil ingestion prevalence may not
consistently monitor the actual overall risk associated with soil pica behavior in children.
3.3.3.2 Prevalence of Soil Pica Separate from Paint Pica. Although the literature
review provided several estimates of the prevalence of soil pica behavior, none of the cited
sources provided information about concurrent paint pica behavior. For the purpose of this
report, the prevalence of soil pica behavior in the absence of paint pica is of interest, as the §403
risk analysis did account for paint pica behavior. For children who ingest both paint chips and
soil, it is reasonable to assume that the effect of soil pica is insignificant compared to that of
153
-------
paint pica. Unless there is industrial contamination, or the home is in an area with heavy traffic,
where residual leaded gasoline emissions are present, lead in residential soil is usually derived
primarily from lead-based paint. Thus, soil pica can be considered an indirect pathway of
exposure to leaded paint, whereas paint pica is a direct exposure pathway.
Information on pica behavior for paint, soil, and other objects was collected in the three
USLADP studies (USEPA, 1996a), the Rochester Lead-in-Dust study (USHUD, 1995a;
Lanphear et al., 1996a), and the Baltimore R&M study (USEPA, 1996c), through parental
reporting of observed behaviors. Therefore, it is possible to use these data to estimate the
prevalence of soil pica separately from paint pica. This information is summarized hi Table
3-26. As can be seen hi this table, rates of soil pica only range from 9.1 to 40.9 percent, while
rates of both soil and paint pica range from 1.4 to 7.4 percent.
Some of the disparity hi the rates reported in Table 3-26 can be explained by the survey
questions and other factors associated with the study. For example, in the Rochester Lead-in-
Dust, Baltimore R&M, and Cincinnati USLADP studies, parents were asked how frequently the
child put dirt or sand in his or her mouth. In contrast, parents in the Boston and Baltimore
portions of the USLADP were asked how frequently the child ate dirt or sand. The paint pica
questions were more consistent across studies, querying how frequently the child put paint chips
in his or her mouth. In the Cincinnati USLADP study, the time-period of observation for both
soil and paint was limited to the previous month, whereas the other studies used open-ended time
periods. Response rates in Rochester were consistent with literature estimates of soil pica that
included mouthing behavior, while the Baltimore R&M and Cincinnati USLADP studies
provided substantially lower estimates. Because most homes in the Baltimore R&M study had
small or no yards, the low estimates of soil mouthing behavior are not unexpected. The lower
response in Cincinnati is probably due to the limited period of observation.
Since inadvertent soil ingestion due to mouthing behavior was included in the IEUBK
model analysis for the §403 risk analysis, the prevalence of soil ingestion, rather than mouthing
behavior, is of interest hi the context of this report. Thus, the Boston and Baltimore portions of
the USLADP study provide the best estimates of soil pica behavior in the absence of paint pica.
These estimates are 14.4 and 16.3 percent hi Boston and Baltimore, respectively. These
estimates are greater than those derived from the mass-balance studies, but consistent with other
studies that rely on parental reporting methods. The prevalence of pica for both paint and soil
was low in Boston (1.4 %), but somewhat higher in Baltimore (6.0 %). Adding these rates to the
reported rates for soil pica alone does not substantially increase the estimates, however, which
remain hi the range of other studies that rely on parental reporting.
154
-------
Table 3-26. Estimated Rates of Paint and Soil Pica Behavior Reported in the
USLADP Studies, the Rochester Lead-in-Dust Study, and the
Baltimore R&M Study
Study
Boston
USLADP
(146 children)
Baltimore
USLADP
(400 children)
Cincinnati
USLADP
(220 children)
Baltimore R&M
Pre-intervention
(165 children)
Rochester
Lead-in-Dust
(203 children)
Type of Pica
Behavior
Soil only
Paint only
Soil and Paint
neither
Soil only
Paint only
Soil and Paint
neither
Soil onty
Paint only
Soil and Paint
neither
Soil only
Paint only
Soil and Paint
neither
Soil only
Paint only
Soil and Paint
neither
StudyiChOdren Exhibiting Such Pica Behavior.
Percent (#) of Y
Study Children1
14.4% (21)
9.6% (14)
1 .4% (2)
74.7% (109)
16.3% (65)
10.5% (42)
6.0% (24)
67.3% (269)
23.2% (51)
2.7% (6)
2.3% (5)
71.8% (158)
9.1% (15)
7.3% (12)
7.3% (12)
76.4% (126)
40.9% (83)
2.5% (5)
7.4% (15)
49. 3% (100)
F Average Age
. (months)
NA
NA
NA
NA
35.7
33.1
26.5
41.7
35.9
19.7
28.0
28.1
27.5
26.7
24.2
31.1
21.0
23.0
20.0
21.4
Geometric Mean,
Blood-Lead Cone. (pg/dL).
12.5
12.7
13.4
11.7
12.0
11.7
15.2
10.3
12.8
15.3
14.7
8.9
10.5
15.3
20.7
9.4
6.1
11.5
8.5
6.2
' A response of "Unknown* was treated as missing and was not included in the calculation of these percentages.
NA = Not applicable
3.3.4 Estimating the Frequency of Ingestion and Amount
of Soil Ingested by Children Who Exhibit Soil Pica
As discussed in Section 3.3.3.1 above, studies reporting soil ingestion prevalence have
the potential to misrepresent the extent of soil pica behavior in children due to differences in
methodology and criteria for defining pica. Therefore, estimates of ingestion quantity and
frequency may also be employed to assess the severity of soil pica behavior.
Mass-balance studies provide data on the frequency of ingestion and amount of soil
ingested by children who exhibit soil pica. These studies estimate the typical amounts of soil
inadvertently ingested by normal children as ranging from 9 to 246 mg/day. The estimated
155
-------
quantities ingested in actual "pica" episodes are between 500 and 13,000 mg/day (Table 3-25).
The literature generally reports pica behavior to be episodic hi nature, varying both amongst
different children and within individual children. In addition, the occurrence (including both
frequency and quantity) of soil ingestion was observed to be influenced by the age of the child
(Stanek et al., 1998; Annest and Mahaffey, 1984), as well as by variety of factors that may alter
the child's access to soil, including seasonal variation and/or climate/vegetation differences
(Simon, 1998; Calabrese and Stanek, 1993), socioeconomic status (Annest and Mahaffey, 1984;
Bhatia, 1988), and parental supervision (Calabrese and Stanek, 1993; Bhatia, 1988). However,
Davis et al., (1990) found that although there was considerable variability in soil ingestion
estimates among children, there was no consistent demographic or behavioral factor that was
predictive of soil ingestion.
Calabrese et al., (1989) estimated median soil ingestion rates, including those involved in
non-pica behavior, between 9 and 40 mg/day (n = 64). Calabrese et al., (1997) also observed
median soil ingestion rates under 40 mg/day in 12 children (selected from the population
described in Stanek et al., 1998) identified by their parents as likely to ingest soil at a high rate.
These levels of soil ingestion typically would not be considered pica behavior. Each of these
studies, however, did report observations of a child exhibiting extreme soil pica behavior, with
one child ingesting from 5 to 8 grams of soil per day (Calabrese et al., 1989) and another child
ingesting between 0.5 and 3.0 grams of soil per day on 4 of 7 days (Calabrese et al., 1997).
Calabrese et al., (1991) found that the soil pica behavior for the former child occurred only on
two days during the two weeks of observation with an ingestion rate ranging from 10-13 grains
of soil per day, suggesting that the issue of variability in soil pica behavior may be very
important, meriting further research. Implications of these patterns were demonstrated by
Calabrese et al., (1993), who observed that on the two days when the child displayed soil pica
behavior, she also displayed striking increases in fecal lead excretory values. In contrast, the pica
child reported in Calabrese et al., (1997) consistently ingested large quantities of soil
(0.5-3 g soil/day on 4 of 7 days).
Calabrese and Stanek (1993) presented results of a 4-month mass-balance/chemical tracer
study performed by M.S. Wong of 52 Jamaican children of generally normal intelligence hi an
institutional setting. The children were partitioned into a younger (0.3-7.5 years) group and an
older (1.8-14 years) group. One of the children in the older group exhibited mental retardation.
This was the only child in the older group (of 28 children) that exhibited soil pica exceeding one
gram of soil per day. This child had an average ingestion rate of 41 g soil/day over 4 months
(observations on 1 day per month), hi the younger group, 10.5 percent of total observations (n =
84) included soil pica, and five of the 24 children exhibited soil pica on at least one occasion.
The Wong study showed that soil pica occurred more frequently in younger children, and there
was a fairly high degree of daily variation in soil ingestion among the children exhibiting soil
pica. For example, 3 of 6 children displayed pica on only 1 of 4 days. Furthermore, even for the
children who consumed soil more consistently with regards to frequency, the rates were still
variable (e.g., 1.0-10.3 g/day). Calabrese and Stanek (1993) suggest that although this study
confirms that soil pica, strictly defined as ingestion greater than 1.0 g/day, is likely to be rare in
156
-------
older children, the Wong study is important in that it challenges the idea that pica is a rare event
in younger children.
Using daily soil ingestion data from their 1989 study, Stanek and Calabrese (1995)
developed annual soil ingestion distribution estimates as follows. First, the mean and variance of
daily soil ingestion were estimated for each of the 64 children in the 1989 study, based on 4 to
8 daily estimates for each child. Then 365 daily soil ingestion amounts for each child were
calculated as percentiles of a log-normal distribution with the estimated mean and variance, in
increments of 1/365. Based on these distributions, Stanek and Calabrese conclude that 33
percent of children are expected to ingest more than 10 grams of soil on 1-2 days per year and
that 16 percent of children are expected to ingest more than 1 gram of soil on 35-40 days per
year. These ingestion levels are consistent with amounts estimated for soil pica episodes. The
median and 95th percentile for average daily soil ingestion resulting from this method were
75 mg/day and 1,751 mg/day, respectively. While the median estimate is similar to previous
estimates, the estimated 95th percentile is substantially greater than most other estimates.
Assumptions and limitations of this approach include:
1. The assumption of a log-normal distribution for daily soil ingestion. Insufficient
data were available to determine whether this assumption is reasonable.
2. The estimation of the mean and variance for each child based on very small
sample sizes. The annual estimates were strongly affected by the tails of the
distribution, which are imprecise due to large variability in the estimates of the
mean and variance.
3. The extrapolation of daily soil ingestion estimates from a 2-week period in
autumn to the remainder of the year, without regard to possible seasonal effects.
In addition, the children studied were a nonrandom sample residing in or near an
academic community in western Massachusetts. Thus, the soil ingestion behavior
of these children may not be representative of those living in other climates,
geographic regions, or in inner-city or rural areas.
4. The presence of trace elements in fecal matter was assumed to be entirely due to
soil consumption, after correcting for food consumption, with no contribution
from indoor dust.
Many of these assumptions and limitations serve to introduce positive bias to the daily soil
ingestion estimate, while the effect of others is unclear. Nonetheless, this analysis is at present,
the only available source of both frequency of soil pica episodes and amount ingested during soil
pica episodes.
157
-------
3.3.5 Conclusions on Soil Pica
The following conclusions can be made from the findings presented in this section:
• The prevalence of soil pica, exclusive of paint pica, is most likely between 10 and
20 percent in young children. For the purpose of this report, the Boston and
Baltimore portions of the USLADP provide the best estimates of soil pica
behavior in the absence of paint pica (14.4 and 16.3 percent, respectively).
• Soil pica behavior is episodic in nature. The frequency of soil pica episodes
depends on many factors, including climate, access to bare soil, socioeconomic
standing, age of child, and parental supervision. In one study of 12 children
identified by their parents to be predisposed to pica for soil, only one child
displayed soil pica during the two week observation period (Calabrese et al.,
1997). Only one study estimated annual rates for pica episodes (Stanek and
Calabrese, 1995). This study suggested that 33 percent of children would ingest
more than 10 grams of soil on 1-2 days per year, and that 16 percent of children
are expected to ingest more than 1 gram of soil on 35-40 days per year.
* Estimates of the amount of soil ingested during pica episodes vary widely among
the mass balance studies, from 500 to 13,000 mg/day. The average daily ingestion
over a year, however, may be much lower. Assuming the frequencies estimated
by Stanek and Calabrese (1995), children who ingest 15 grams of soil on 1-2 days
per year and 50 mg/day on remaining days would have an average daily soil intake
of 132 mg/day over the course of a year. Children who ingest 1.5 grams of soil on
40 days per year and 50 mg/day on remaining days would have an average daily
soil intake of 209 mg/day. A question, however, is whether the amount of lead in
soil ingested on the small number of days where pica episodes occurred would be
sufficient to elevate the blood-lead concentration to unsafe levels.
3.4 CHARACTERIZING THE POPULATION OF CHILDREN
IN THE NATION'S HOUSING STOCK
For the §403 risk analysis, it was necessary to estimate numbers of children of specific
age groups who reside within the 1997 national housing stock in order to characterize the extent
to which various environmental-lead levels provide exposures to children and to characterize the
benefits associated with performing interventions under §403 rules. These estimates were based
on numbers of housing units determined by sampling weights within the HUD National Survey
(conducted in 1989-1990), revised to represent the 1997 national occupied housing stock and on
average numbers of children per housing unit determined from the 1993 American Housing
Survey (AHS). The estimates used in the §403 risk analysis were presented in Section 3.3.2 and
Appendix C of the §403 risk analysis report. This section provides alternative estimates using
158
-------
more recent data (i.e., interim data from the NSLAH and data from the 1997 American Housing
Survey).
The method to calculating the alternative estimates involved determining the numbers of
children in a given age group for each of the 706 housing units surveyed within the NSLAH
whose interim data were made available to this effort. Methods used to obtain these estimated
numbers of children were similar to those presented in Section 1.2 of Appendix Cl of the §403
risk analysis report. For a given age group of children, the estimated number of children
associated with a given NSLAH-surveyed unit was determined by the following formula:
# children = (1997weight)*(Average#residentsperunit)*(#chUdrenperperson)
(1)
The "1997 weight" factor in equation (1) was the interim sampling weight from the NSLAH for
the unit. The factor "average # residents per unit" in equation (1) was calculated for the housing
group based on information obtained from the 1997 AHS. The 1997 AHS database provided
information on up to 18 residents within each housing unit in the AHS. Once units surveyed in
the 1997 AHS were placed within the four year-built categories (pre-1940,1940-1959,1960-
1979, post-1979), the average number of people residing in a unit (regardless of their ages) was
calculated for each group. This average ranged from 2.5 to 2.7 across the four year-built
categories. Therefore, a common average of 2.6 residents per unit was used for the entire
national housing stock. The third factor in equation (1), "# children per person," represented the
average number of resident children (of the given age group) in a housing unit. This factor was
determined by dividing the total number of residents in the housing stock of a given age group by
the total number of residents regardless of age, where both totals were calculated from data in the
1997 AHS. The method for calculating this third factor differed from the approach used in the
§403 risk analysis, where forecasted birth rate and population estimates from the Bureau of the
Census were used.
Table 3-27 contains estimates of average number of children per unit in the 1997 national
housing stock, according to age group. These number are the product of the final two factors in
equation (1). Therefore, these number are multiplied by the sampling weights for each housing
unit in the interim NSLAH to obtain a revised number of children per housing unit. For children
aged 12-35 months, the estimated average of 0.073 children per unit is about 9% lower than the
estimate of 0.080 used in the §403 risk analysis.
Table 3-27. Alternative Estimates of the Average Number of Children Per Unit in the
1997 National Housing Stock, by Age of Child
Age Group
12-35 months
12-71 months
Estimated Average Number of
Children Per Unit
2.6*0.0281 = 0.073
2.6-0.0732 = 0.190
159
-------
By summing the estimates across surveyed units in the interim NSLAH, the updated
number of children aged 12-35 months and 12-71 months residing within the 1997 national
housing stock is obtained by year-built category and for the nation. Table 3-28 provides these
alternative estimates on the number of children residing in the 1997 housing stock according to
age of housing unit and age of child. The overall estimate of approximately 6.51 million children
aged 1-2 years is approximately 18% lower than the estimate of 7.96 million made in the §403
risk analysis. The lower estimates are due to the lower per-unit estimate from Table 3-27 and on
the lower sample weight total in the interim NSLAH data compared to the HUD National Survey
(Table 3-2). They are also likely to be underestimates of the numbers of children of the given
age category, based upon population projections previously published by the U.S. Bureau of the
Census (e.g., Day, 1993).
Table 3-28. Alternative Estimates of the Average Number of Children in the 1997
National Housing Stock, by Age of Child and Year-Built Category, Based on
Data Obtained Since the §403 Risk Analysis
"Years in Which Housing
Units Were Built
Prior to 1940
1940-1959
1960-1977
After 1977
Unknown1
All Housing2
Age of Child Within These Housing Units
1-2 Year$::-':P '•' :
1,053,000
1,234,000
1,877,000
1,759,000
591,000
6,513,000
1-S Years
2,743,000
3,214,000
4,889,000
4,582,000
1 ,540,000
16,967,000
1 There are 66 units in the interim NSLAH which have missing age of house data.
2 Values in this row may differ from sum of previous rows due to rounding.
3.5 SUMMARIES OF DUST-LEAD LEVELS ON SURFACES OTHER
THAN UNCARPETEP FLOORS AND WINDOW SILLS
The exposure assessment in Chapter 3 of the §403 risk analysis report concluded that
even at low to moderate lead levels, lead-contaminated dust can affect children's blood-lead
concentration. The assessment focused on dust-lead found on floor and window sill surfaces, for
which §403 regulatory standards were proposed. However, dust found on other surfaces, such as
exterior dust and dust hi air ducts, carpeted floors, window troughs (also known as window
wells), and upholstery, may also potentially present a lead exposure hazard. Many issues
concerning potential exposure to dust-lead on these other surfaces were raised throughout the
§403 Dialogue Process as well as in comments received on preliminary drafts of the §403 risk
analysis and on the proposed rule. For example, there was extensive discussion during the
Dialogue Process concerning whether standards were necessary for window troughs (i.e.,
window wells) as long as there are standards for window sills and the window troughs are
160
-------
thoroughly cleaned. Concern was also expressed about sampling on carpeted and upholstered
surfaces.
The purpose of this section is to supplement information in the original exposure
assessment by assessing the potential exposure to dust-lead found on surfaces other than floors
and window sills. In particular this assessment seeks to answer the following questions:
1. What information is available to assess residential lead exposure resulting from dust
on surfaces other than floors and window sills?
2. What does this information say about the distribution of environmental lead levels
for dust on these other types of surfaces?
3. Is there evidence of a relationship between these exposures and children's blood-
lead concentrations?
4. Is the information sufficient to set a regulatory standard and is a standard necessary?
The exposure assessment is based on a review of the literature to identify studies with
potentially useful data for assessing lead hazards due to dust-lead on these other surfaces. It
should be noted that this section deals only with several specific surfaces other than uncarpeted
floors and window sills. These surfaces include exterior dust, air ducts, window troughs, and
upholstery. Hazards associated with lead-contaminated dust from carpeted floors are addressed
in Section 6.5 and Appendix I of this report.
The literature review for this assessment drew upon previous literature reviews conducted
for the §403 risk analysis and reviews conducted for other EPA published reports (e.g., USEPA,
1997b). Most of the studies that were found that addressed the various surfaces are included
here, regardless of whether there is any information specifically relating dust-lead levels and
blood-lead levels. For those studies where blood samples were collected from resident children,
those results are also presented. Table 3-29 provides a summary of the studies that were
examined. The table indicates the surfaces from which dust samples were collected and in the
case of exterior dust samples, where those samples were collected.
Table 3-29 indicates that the review of the literature found fourteen studies that have
examined exterior dust as a source of lead exposure and seven studies that similarly assessed
window troughs. Only four studies were located with significant information on dust-lead levels
in air ducts, and four with similar information on upholstery.
3.5.1 Distribution of Dust-Lead on Surfaces Other than Floors
and Window Sills
Tables 3-30 through 3-33 present summary information from the studies related to
exterior dust samples, air duct samples, window trough samples, and upholstery samples,
161
-------
Table 3-29. Studies for Which Dust Samples Have Been Collected from Exterior Areas,
Air Ducts, Window Troughs, and Upholstery for Lead Analysis
= medium was sampled in the given study
I.-'..' •:.•'''•'• '. '" . Study:'; , . ,:v. . ./\-
Lead-Based Paint Abatement and Repair & Maintenance
-------
01
M
•5
§
£
"5
O
10
M
o.
i
**
(0
S
O> 00 OJ 00 09
o> co •* 01 •»
i^ in ui
85*
*
I
C9
If
III
&.&.SL
SO TO 00
ss
at
0)
O
1
*•«
CO
2
n»
i
O
I
(0
6
CO
a
il
EC
m
Hi
I
55
163
-------
ep a>» ev i
CO * CD CO CO '
go n co to o> co •** co ,, -^
a • en oo • i~ co 03 'uico • r^ p-
•r- CM « •» in
-------
CD o> to I-; r-^
S.J2.H2.HLS2.
'v 2 S3 § I
Su> CM •- r«- r^ r» m
„ v *_ oj »- t- ~
i < oj o a uj u. O
eo" u£ ^* %? &? &? 35" ^* TO^ 3% £% *
s.si^.s2.siaiaj.a!.s2.aia^3
?s-sr
;s,s>.
.
o
m
a
if
:1
{]
&•
I
CO
li
a. sL IL £2, si al al si.
;
i o a>
o ed id co
£ u. u_ u.
m o o uj u. O
g g g
• w en en
•g-i-i-i
-S S S^
ffl CD OQ CD
CO
CD UJ I C,
165
-------
S2.Ei.QSi
to cd ^ cd
Q i
(O.C
.
9cj . in
—si a. 21
to .
*'
a.
166
-------
CO
I
<
"8
5
CD
t»
£
"3
U
e
(0
t!
3
ts
a
i_
<
s
I
•O
w
I
"S
>.
0)
a
Si
5-r«
*?
CO CO *T CM V
25i"
l5Sp
iiill
I
o
1
O
E
co 7
833
cjcjf
fi
Fl
!l
k
.2
£1
2-
5.
s-
o
|i|
asE
E
II
M 25'
w «1.
n SS
zo<:
CO II
JQcr
S u
gs
167
-------
*
I
<
•o
a
I
•o
o
o
S
0
S
a
CO
Q
f
2
o
•o
i
S
o
1
*<
0)
S
«0
O
•5
a
CO
ci
n
n
o
S
.2
e>o>c\i cq oq
o» CT * cu •*
S .SB .a
i
lifll
g
iiiii
.
in in
CO CO S S
—.ass
8SS2E:
O) CO
CM
.
*? *f *? S
^ ce i- T r
cn o
i
m
i!
CD I
J
CD
II
II
lill
nil
Illl
llsl!*
iilii!
1
168
13
-------
CM
CO
n
9
5
a
3
coco
-------
1
a
8
1
§
o
S
o
a
_«
a
S
tl
8'
*=*w . *
*^ <0 (O 05 CVI 05»
O> CT •* CM •*
333S?
i-=|||
i
0
_
>« rj-ejl
§ So
.£--- ~r-e
ent
Ab
' S 3^.
ip
111
1
-oil
0 " «S
^ 52 n u) t
O
e
O
B/6
^
«?
s
gf
I
1
*
PA Supetfund
(1989-1 990)
5 ° < ll S it ll eS
n II -g "22 M
z(3<5 wm a ir
170
-------
respectively. Included in the tables are sampling location, dust collection methods used, number
of dust samples taken, and distribution statistics available from the literature on dust-lead
concentrations and loadings. Also presented in the tables are simple summaries of blood-lead
concentrations for those studies which also sampled blood.
Exterior Dust
Table 3-30 summarizes the data from the fourteen studies in which exterior dust samples
were collected and analyzed for lead content. In addition, there are also results for several
studies in which dust samples were collected inside the home in the entryway. These can
occasionally be good representations of exterior dust samples. The summary consists of
descriptive statistics for the dust-lead measures and also for blood-lead measures, when
available. Summary statistics are presented separately for different study groups and different
sampling periods, where appropriate. As indicated in Table 3-29, exterior dust samples were not
collected at a consistent location across all studies. Sampled locations included the house
entrance (including doormats and porches), street, and the child's play area. Likewise, samples
were collected by a variety of methods including surface scraping, vacuum sampling, and wipe
sampling. The sampling location and method have a significant impact on the lead loading and
concentration estimates.
Overall, even with the variability in sampling location and method, Table 3-30 indicates
the potential for significant amounts of lead in exterior dust, with concentrations often exceeding
400 ug/g and loadings often exceeding 100 ug/ft2.
Air Ducts
As indicated in Table 3-31, only four studies were identified that contained information
on lead levels in dust within air ducts in residential housing. While only limited information was
encountered, a consensus across studies was that air ducts can contain high amounts of dust and
lead. This was due partially to the general lack of cleaning of ah* ducts over time and the ability
of lead particles to enter air ducts from outside of the unit via ventilation filters.
In units with a potential for containing lead hazards, dust-lead loadings in air ducts
typically exceeded 100 ug/ft2, with individual samples often exceeding 1,000 ug/ft2. Lead levels
can vary considerably among dust samples within the same unit and in different units. Older
ductwork and HVAC systems, as well as vacant units in which no cleaning is performed and
HVAC systems may not be used, tend to have high dust loadings, and therefore, higher dust-lead
loadings when a lead source is present. Several methods were used across studies to collect dust
in air ducts. As air ducts often have metal surfaces, issues concerning static electricity must be
considered when sampling dust from air ducts.
The EPA Comprehensive Abatement Performance (CAP) study involving occupied
homes assumed to be free of lead-based paint for at least two years, provided the greatest amount
of information on lead in dust within air ducts; levels were relatively low in this study compared
171
-------
to the others. Nevertheless, in a typical housing unit in the CAP study, average dust-lead
loadings from air ducts exceeded all other sampled surfaces except for window troughs and
entryways. In general, air duct dust-lead levels in the Baltimore R&M study and the Renovation
and Remodeling study were considerably higher than in the CAP study, as these studies included
older, vacant units in need of repair and maintenance.
The relative sparsity of published information indicates that many open questions exist on
the nature of lead contamination of dust within residential air ducts and whether the lead in this
dust is available for exposure to residents (especially children). Nevertheless, evidence exists
that air ducts can contain some of the highest levels of lead in dust within a housing unit.
Window Troughs
Table 3-32 presents summary information on seven studies that sampled dust-lead levels
in window troughs (also known as window wells). The HUD Grantees evaluation is a significant
data source on window trough dust-lead levels in high risk housing that is not included in Table
3-32, as these data are still being collected and reported.
In general, partially because of the published data summarized in Table 3-32, and
partially because a standard for window troughs has been historically used in risk assessments
and to determine clearance (following EPA's Interim Guidance for §403 standards and the HUD
Guidelines), window troughs are widely recognized as a major reservoir of dust-lead in
residences. As shown in Table 3-32, levels often exceed 800 ug/ft2 and it is not uncommon to
see levels above 10,000 ug/ft2 in high risk housing. However, unlike the other surfaces discussed
in this report, national estimates of the distribution of dust-lead in window troughs are available
from the HUD National Survey. The estimated national geometric mean dust-lead loading in
window troughs from the HUD National Survey (as modified in the §403 risk analysis to reflect
the 1997 housing stock and wipe techniques) was 460 ug/ft2 (Table 3-32), with 30% of homes
estimated to have average window trough dust-lead loadings at or above 800 ug/ft2.
Upholstery
Table 3-33 summarizes the data from the four studies which collected dust-lead samples
from upholstery. In general, dust-lead loadings for these surfaces averaged below 100 ug/ft2.
As the sample sizes in all of these studies were small, and sampling techniques, sampling
locations, and study goals varied considerably from study to study, more information would be
necessary to fully characterize potential lead hazards associated with upholstery.
172
-------
3.5.2 Evidence of a Relationship Between Children's Blood-Lead
Concentrations and Dust-Lead on Surfaces Other Than
Floors and Window Sills
The information available to assess children's exposure to dust-lead on surfaces covered
in this section is discussed in detail for each surface type below. It should be noted, however,
that in general it is difficult to establish the causal link between these surfaces and children's
blood-lead concentrations. This is true for many reasons. Often other important sources of lead
exposure are not well characterized in the studies that provide data on these special surfaces.
Correlations are often estimated based on small sample sizes and without adjusting for other
exposure variables such as lead in floor-dust and soil. Moreover, there is often correlation
between lead levels on these surfaces and lead levels on floors, window sills, and in soil. For all
of these reasons, it must be noted that even significant correlation coefficients should not be
interpreted as the degree to which dust-lead on these surfaces causes a change in blood-lead
concentration. In almost all cases, in order to characterize the pathway of lead from these
surfaces to children's blood, additional data collection or analyses are needed.
Exterior Dust
Table 3-30 (in the previous subsection) contains a summary of the dust-lead data and
blood-lead data separately for studies which collected exterior dust. However, it contains no
results providing information about the relationship between exterior dust-lead levels and blood-
lead levels. The reports describing these analyses were examined to assess the relationship
between exterior dust-lead levels and blood-lead levels. In the Repair and Maintenance Study
(USEPA, 1996c, 1997c), exterior dust samples were collected at five separate times. The
(Pearson) correlation coefficients between blood-lead concentrations and entryway dust-lead
concentrations ranged from 0.23 to 0.49, and the correlation coefficients between blood-lead
concentrations and entryway dust-lead loadings ranged from 0.10 to 0.46. In most cases, those
correlation coefficients were statistically significant at the o=0.01 level. In the Rochester Study
(USHUD, 1995a), the University of Cincinnati Dust Vacuum Method (DVM) and Baltimore
Repair and Maintenance vacuum method (BRM) were used to collect external dust samples. The
correlation coefficients for the BRM and DVM techniques were 0.21 and 0.27 for the correlation
between blood-lead concentrations and exterior dust-lead concentrations and 0.34 and 0.18 for
the correlation between blood-lead concentrations and exterior dust-lead loadings, respectively.
The correlations were all statistically significant at the o=0.05 level, with the correlation for the
DVM measured loading significant at the a=0.01 level. On the other hand, the Mexico City
Study (Romieu et. al., 1995), Arnhem Study (Brunekreef et. al., 1981) and the Midvale Study
(Bomschein et. al., 1990) reported the correlations between external dust measurements and
blood-lead levels to be statistically insignificant.
Multivariate regression and structural equation modeling was used in some of the studies
to examine how multiple sources of environmental lead exposure and other factors affect blood-
lead levels. Regression analyses were carried out in most of the studies where both external dust-
lead and blood-lead measurements were collected, but the external dust-lead measurements were
173
-------
not included as an explanatory variable in any of the reported regression models. Reasons for
excluding external dust-lead were not clearly stated. Speculatively, such reasons may include a
lack of interest in the relationship, poor data quality in the external dust-lead measurements,
colinearity of external with internal dust-lead measurements and omission of the variable through
step-wise regression. Structural equation modeling was carried out in the (Three Cities) Urban
Soil Lead Abatement Demonstration Project (USEPA, 1996a), Butte-Silver Bow Study (Butte-
Silver Bow DoH et. al., 1991) and Midvale Study (Bornschein et. al., 1990). In the Three Cities
Study, exterior dust was lead considered as a component of the lead exposure pathway in the
general structural equation model, but the component was excluded in the actual implementation
of the model. In the Butte-Silver Bow Study, external dust-lead was also excluded from the
structural equation model, but external dust-lead was observed to be correlated (Pearson
correlation r=0.64) with soil lead, which was included as a component of the lead exposure
pathway in the model. The Midvale Study was the only one to include external dust-lead hi the
actual implementation of the structural equation model, but the dust-lead to blood-lead
relationship was reported as being statistically insignificant.
hi summary, there is much difficulty in distinguishing between direct and indirect
exposure in cases where external dust-lead levels is closely related to levels in other sources of
environmental lead. Correlation and univariate regressions with external dust-lead and blood-
lead fail to account for the possibility that external dust-lead by itself may only play a small part
in aggregate lead exposure when other sources of lead and exposure pathways are considered.
Multivariate regressions using lead measurements from multiple sources do not solve this
problem due to problems with colinearity. The preferred approach would be to use structural
equation models, which allow multiple source and exposure pathways to be modeled in a
reasonable way, but this approach requires more effort in terms of implementation and
interpretation of the model, and is not well-reported in the literature. Therefore, quantitative
estimates of the effect of external dust-lead on children's blood-lead concentrations have not
been well established in the literature.
Air Ducts
Most of the encountered articles provided only preliminary information on lead exposures
associated with ah* ducts. It is unclear to what extent dust-lead in air ducts is accessible to
children. Children would not typically be expected to encounter the dust lodged in air ducts
directly. One case study found that dust-lead levels in living areas outside of contaminated air
ducts can be orders of magnitude lower than what is found in the air ducts. However, if dust in
air ducts is disturbed, it is more likely to be introduced to the air and to nearby surfaces with
which children can come into direct contact. In particular, HVAC ductwork removal can yield
extensive contamination of surfaces in the general area of the ductwork.
Only one study (the Baltimore R&M Study) estimated (in a quantitative manner) the
association between blood-lead concentrations in children and dust-lead levels found in air ducts.
This relationship was expressed as a simple correlation coefficient. Unlike correlations between
blood-lead concentrations and dust-lead levels on other surfaces, the correlation coefficient
174
-------
involving dust-lead levels from air ducts was not significant at the 0.05 level. However, this
analysis was based on a small sample size and did not adjust for the effects of other exposure
variables such as lead in floor-dust and soil. Moreover, as evidence of a significant correlation
was observed between air duct dust-lead levels and lead levels on other surfaces, such as floors,
even significant correlation coefficients should not be interpreted as the degree to which air duct
dust causes a change in blood-lead concentration. In order to characterize the pathway of lead
from air ducts to children's blood, additional data collection and analyses are needed.
Window Troughs
Of the seven studies listed in Table 3-32 above that collected information on dust-lead in
window troughs (also known as window wells), three also collected blood-lead data from
resident children. Correlation coefficients between blood-lead levels and window-trough dust-
lead concentrations in the R&M study ranged from 0.20 to 0.39, and correlation coefficients
between blood-lead levels and window-trough dust-lead loadings ranged from 0.06 to 0.44. The
correlations between dust-lead concentrations and blood-lead concentrations were statistically
significant in 4 of the 5 sampling campaigns, and the correlation between dust-lead loading and
blood-lead concentration was statistically significant only in the pre-maintenance sampling. In
the Rochester Study, correlation coefficients between blood-lead concentrations and dust-lead
loadings were 0.35 for the BRM samples, 0.31 for the DVM samples, and 0.29 for wipe samples,
while correlation coefficients between blood-lead concentrations and dust-lead concentrations
were 0.23 for both BRM and DVM samples.
Previous analyses (Battelle, 1996a; Battelle, 1996b) have examined whether the
predictive ability of a model improves when adding window trough lead levels to a model which
already accounts for dust-lead on floors and window sills. Results of these analyses on the
Rochester Lead-in-Dust Study and Baltimore R&M study data indicated that the estimated effect
of window trough dust-lead on blood-lead was either not statistically significant or only
marginally significant after adjusting for the effects of floor lead, sill lead, and temporal
.variation. A pathways analysis (USEPA, 1998c) using structural equations modeling concluded
that window troughs were a significant pathway for lead exposure, both as a direct pathway of
lead to children's blood-lead concentration (seen when Rochester Lead-in-Dust study data were
analyzed) and as an indirect pathway through window sills and floors to blood-lead
concentrations (seen when both the Rochester and Baltimore R&M study data were analyzed).
In summary, the association between blood-lead concentrations and window trough dust-
lead has been well established in the literature. The more difficult question of the degree to
which window troughs contribute directly or indirectly to children's lead exposure is not well
established.
175
-------
Upholstery
Table 3-33 in the previous subsection included results for two studies which measured
both children's blood-lead concentrations and dust-lead on upholstery. In the Baltimore R&M
study, correlation coefficients were calculated between blood-lead concentrations in children and
both the loading and concentration of lead in upholstery dust. These correlation coefficients
ranged from 0.19 to 0.61 for dust-lead concentrations and from 0.06 to 0.47 for dust-lead
loading. These correlations were statistically significant in the pre-intervention sampling. As
with most of the other surfaces discussed in this section, upholstery dust-lead levels were not
included in any analyses to determine which lead sources were most significantly related to
blood-lead levels. In the Mexico City Study, the correlation between upholstery dust-lead levels
was not statistically significant, resulting in the absence of upholstery dust-lead levels from
models linking blood-lead levels and environmental lead levels.
The results of the two studies assessing the importance of upholstery dust as a source for
lead exposure in children differ. In one case, the relationship between blood-lead and upholstery-
dust-lead is significant, while in the other it is not. Moreover, as upholstery dust-lead is often
correlated with other lead exposure variables, such as floor dust-lead and soil-lead, as cautioned
earlier, the positive correlation coefficient should not be interpreted as the degree to which
upholstery dust causes a change in blood-lead concentration. In order to characterize the
pathway of lead from upholstery to children's blood (and perhaps hands), additional data
collection and analyses are needed.
3.5.3 Implications of the Available Information For Regulatory Standards
Two primary questions related to the need and feasibility of regulatory standards for dust-
lead on surfaces other than floors and window sills are:
1. Is there sufficient information available on which to base a standard?
2. Is the standard necessary to either identify a lead hazard at a residence or to
characterize the risk to determine appropriate corrective actions?
The answers to these questions are discussed for each surface type below.
Exterior Dust
In general, there is a fair amount of data on exterior dust, including studies where exterior
dust has been measured along with other lead exposure variables and blood-lead concentrations.
The amount of data implies that analyses could be conducted to provide a quantitative basis for
an exterior dust standard. However, implementation and interpretation of such analyses for
exterior dust will face many difficulties. For example, in many of the studies it is difficult to
distinguish between exterior dust and soil samples because of aggregation of the samples or of
the measurements. Some external sampling for lead was carried out using surface scrapings
176
-------
which measures lead levels from a mix of both soil and dust-lead and some analyses averaged the
external soil and dust measurements and recorded the value as a single external lead
measurement. (Hence only studies for which a clear distinction between external soil and dust-
lead levels is possible were included in this summary.) It is also difficult to determine what
locations for exterior dust should be included. Should the focus be on enclosed spaces or also
include unenclosed areas such as sidewalks, stoops, and unenclosed porches? One primary
reason to focus only on enclosed areas is because exposures to unenclosed areas are not under the
direct control of property owner. Exposure and cleaning scenarios for enclosed versus
unenclosed areas are likely to be very different as well. In conclusion, decisions on the specific
focus of a standard for exterior dust would impact the feasibility of establishing a good
quantitative basis on which to set the standard.
The question of whether a standard for exterior dust is necessary is also a difficult one,
for which the literature does not have a clear answer. While it is reasonable to assume that
measurements of lead in interior dust and exterior soil might capture a lead hazard if one exists,
there is not a strong body of information on which to base this conclusion. A separate standard
may not be necessary if risk assessors are aware of the potential hazard from exterior dust, and
include testing or corrective actions in cases where it is suspected to be an important pathway of
exposure (for example, in the case where a child spends a considerable amount of time on a
paved surface, such as a driveway or patio).
Air Ducts
There is insufficient data upon which to develop a hazard standard for lead in air duct
dust, or upon which to draw conclusions about the necessity of a standard to either identify a
hazard or determine corrective actions.
Window Troughs
The fact that regulatory standards have been proposed for dust-lead on floors and window
sills based on data sets (most notably the Rochester Lead-in-Dust Study and the HUD National
Survey) that also include window troughs implies that sufficient data exists on which to base a
standard for window troughs.
However, while there is sufficient information on which to base a standard, analyses
conducted to assess the necessity of a window trough standard given the existence of a floor and
window sill standard suggest that a window trough standard may not be necessary to identify a
residence with a lead-based paint hazard. These analyses include the sensitivity/specificity
analyses included in a companion §403 report as well as the analyses that examine the effect of
adding window troughs to a statistical model that already includes floors and window sills
(Battelle, 1996a; Battelle, 1996b). Given the correlation between window trough and window
sill lead levels, it is likely that if more sampling is to be done beyond a minimal risk assessment,
more benefit will be obtained from sampling more windows at the sill rather than sampling fewer
windows but at both the sill and trough. Moreover, cleaning of window troughs is
177
-------
recommended for all homes that require a dust intervention, and clearance standards have been
proposed to guide assessment of the effectiveness of the cleaning. For these reasons, it does not
appear that an additional standard for window troughs is necessary either to identify a home with
a hazard or to guide corrective actions.
Upholstery
There is insufficient data upon which to develop a hazard standard for lead in upholstery
dust, or upon which to draw conclusions about the necessity of a standard to either identify a
hazard or determine corrective actions.
3.6 DISTRIBUTION OF CHILDHOOD BLOOD-LEAD
This section updates the information presented in Section 3.4 of the §403 risk analysis
report on the distribution of childhood blood-lead concentration hi the United States, with a focus
on the 1-2 year (12-35 month) age range as the population of interest. In addition to a national
characterization based on data from Phase 2 of the Third National Health and Nutrition
Examination Survey (NHANES ffl), Section 3.4 of the §403 risk analysis report summarized data
from other studies (e.g., the Baltimore R&M study, the Rochester Lead-in-Dust study, and the
HUD Grantees evaluation) to provide supporting information on the prevalence of elevated
blood-lead concentrations in children living in urban locations and in older housing or housing
likely to contain lead-based paint. Blood-lead data from these other studies were also considered
because the NHANES DI did not collect environmental-lead data, despite having the most
nationally representative data on blood-lead levels.
Section 3.6.1 below is an update of Section 3,4.4 of the §403 risk analysis report. It
contains revised data summaries of pre-intervention blood-lead concentrations hi children
monitored within the HUD Grantees evaluation and revised regression model fits to predict
blood-lead concentration as a function of dust-lead loading for each individual grantee, as well as
for the Rochester Lead-in-Dust study (i.e., the study that provided the data used to developed the
empirical model developed within the §403 risk analysis. These revisions were possible as
additional pre-intervention data from the HUD Grantees evaluation (through 1/99) have been
made available to the risk analysis since the report was released.
Section 3.6.2 provides information from the Cincinnati Prospective Lead study (Clark et
al., 1985) and summarized by the Centers for Disease Control and Prevention (CDC) on the
relationship between children's blood-lead concentration and housing age/condition, and how
this relationship may change with the age of the child (CDC, 1991). CDC used this information
in their recommendations for blood-lead screenings of young children.
178
-------
3.6.1 Evaluation of the HUD Lead-Based Paint Hazard Control
Grant Program ("HUD Grantees")
Blood-lead concentrations of children residing in households participating in the
evaluation phase of the HUD Grantees evaluation (Section 3.2.2.3 of the §403 risk analysis
report) were measured, along with environmental-lead levels in various media. The population
of children targeted for participation in the program differed among the fourteen grantee
recipients, due to the different enrollment criteria among the grantees (see Table 3-4 of the §403
risk analysis report). These criteria included targeting high-risk neighborhoods, enrolling only
homes with a lead-poisoned child, and considering unsolicited applications. Pre-intervention
data collected through January 1999 are presented in this section; these data provide some of the
most recent information on the relationship between children's blood-lead concentration and
environmental-lead levels.
Across all grantees, pre-intervention blood-lead concentration data through 1/99 were
available for 526 children aged 1-2 years and for 764 children aged 3-5 years. For these children,
Table 3-34 summarizes measured blood-lead concentration for each combination of blood
collection type (venipuncture, fingerstick) and age of child (1-2 years, 3-5 years, and 1-5 years).
Table 3-34 also summarizes measured blood-lead concentration for children aged 1-2 years for
each combination of blood collection type and grantee. Note that fingerstick methods were
predominant for Wisconsin, Milwaukee, and Vermont, while Rhode Island used both methods
for similar numbers of children. The remaining nine grantees (excluding New Jersey) used the
venipuncture either exclusively or predominantly.
According to Table 3-34, the geometric mean blood-lead concentration via the
venipuncture collection method was 9.3 (Jg/dL for children aged 1-2 years and 8.0 ng/dL for
children aged 3-5 years. In contrast, the geometric means based on data from Phase 2 of
NHANES UI were 3.1 ug/dL for children aged 1-2 years and 2.5 ng/dL for children aged 3-5
years (Table 3-36 of the §403 risk analysis report). The larger values in the HUD Grantees
evaluation reflect the HUD Grantees program's procedure of selecting high-risk children for
monitoring. The differing enrollment criteria across grantees also contributed to considerable
differences in the geometric mean blood-lead concentration among the grantees.
Under venipuncture, the geometric means of children aged 1-2 years for Individual
grantees reporting more than three blood-lead results ranged from 4.2 ug/dL (California, which
only targeted older units) to 15.9 ug/dL (Cleveland, which targeted units with lead-poisoned
children).
The geometric mean blood-lead concentration via the fingerstick collection method was
9.4 ug/dL for children aged 1-2 years and 8.9 ug/dL for children aged 3-5 years. When data were
available for more than one child under fingerstick collection methods, the geometric means for
children aged 1-2 years ranged from 5.9 ug/dL (Wisconsin) to 13.5 ug/dL (Milwaukee).
179
-------
Table 3-34. Summary of Children's Pre-lntervention Blood-Lead Concentration in the
HUD Grantees Evaluation According to Blood Collection Method, Child Age
Category, and Grantee (ages 1 -2 years only)
-'..;' '
: Age Category
1-2 Years
3-5 Years
1-5 Years
'"-I" Grantee
Alameda County
Baltimore
Boston
California
Cleveland
Massachusetts
Minnesota
New Jersey
Rhode Island
Wisconsin
Milwaukee
Chicago
New York City
Vermont
Age Range
1-2 Years
3-5 Years
1-5 Years
: Grantee
Cleveland
Massachusetts
Minnesota
Rhode Island
Wisconsin
Milwaukee
Vermont
Number
of
ChBdren
Hood-Lead Concentration fe/g/dU
Arithmetic
Mean •
; ' '_.
.Mean
Geometric
Standara •
Deviation
Minimum
Rood Collection Method -
361
536
897
12.5
10.6
11.4
9.3
8.0
8.5
2.3
2.2
2.2
0.7
0.0
0.0
25th
Pefcentife
Median
;••:•>,•-• ' :'".
-•ITWir-/
rarcanuie
Maximum
= Vehipuncture ' .
5.4
4.5
5.0
10.0
8.6
9.0
17.0
15.0
16.0
53.0
48.0
53.0
Blood Collection Method = Venipuncture {Children Aged 1-2 Years only}
27
25
20
21
64
43
75
1
14
9
3
28
23
8
6.5
9.3
12.7
5.3
19.3
11.2
14.5
3.0
10.0
10.2
26.0
13.9
5.2
13.5
4.7
7.7
10.4
4.2
15.9
9.1
10.7
3.0
8.1
8.7
25.1
11.7
4.7
12.4
2.2
1.9
2.0
2.0
1.9
1.9
2.4
-
2.0
1.8
1.4
2.0
1.6
1.6
1.4
2.0
3.0
1.4
4.0
3.0
0.7
3.0
2.0
4.0
18.0
1.0
2.0
6.0
: hi Blood I Collection Method
164
232
396
11.7
11.5
11.6
9.4
8.9
9.1
1.9
2.0
2.0
2.0
2.0
2.0
3.0
6.0
6.0
3.2
11.5
6.0
6.0
3.0
6.0
6.0
18.0
9.5
4.0
8.5
4.7
7.0
14.5
3.8
17.0
9.0
11.0
3.0
8.5
8.0
25.0
12.0
5.0
14.5
6.6
10.0
19.0
6.0
28.0
16.0
22.0
3.0
14.0
12.0
35.0
19.0
7.0
17.0
24.8
26.0
27.0
16.9
53.0
40.0
43.0
3.0
21.0
24.0
35.0
35.0
12.0
22.0
= Fingerstick '•. •• ,';:< - ' •" ' <„•'
6.0
5.0
5.0
9.0
9.0
9.0
15.0
15.0
15.0
48.O
62.0
62.0
Blood Collection Method = Fingerstick (Children Aged 1-2 Years only)
1
4
1
9
43
82
24
13.0
7.8
33.0
8.8
6.2
16.0
7.9
13.0
7.3
33.0
8.2
5.8
13.2
7.0
-
1.5
-
1.5
1.4
1.9
1.6
13.0
4.0
33.0
5.0
3.5
2.0
3.5
13.0
6.0
33.0
7.0
4.0
9.0
5.0
13.0
8.5
33.0
7.0
6.0
14.5
6.5
13.0
9.5
33.0
11.0
8.0
20.0
11.0
13.0
10.0
33.0
15.0
14.0
48.0
16.0
Note: All pre-intervention blood-lead concentration data available and collected through 1/99 are included in the above
summaries.
180
-------
The percentages of children with elevated blood-lead concentrations (Le., concentrations
at or above 10,15,20 or 25 ug/dL) at pre-intervention are summarized hi Table 3-35. According
to this table, 51 percent of children aged 1-2 years sampled via venipuncture methods had blood-
lead concentrations at or above 10 ug/dL, compared to the estimates of 5.88% for Phase 2 of
NHANES m, 53.8% for the Baltimore R&M study (pre-intervention), and 23.4% for the
Rochester Lead-in-Dust study (Tables 3-37, 3-41, and 3-42, respectively, of the §403 risk
analysis report). For individual grantees having more than three children with a measured blood-
lead concentration, the percentage of children aged 1-2 years with blood-lead concentrations
(venipuncture) at or above 10 ug/dL varied from 4% (New York City, which targeted housing
and neighborhoods rather than lead-poisoned children) to 80% (Cleveland). The range of
percentages under the fingerstick method were similar to that under the venipuncture method, but
less data were available to estimate them.
Figures 3-20 and 3-21 illustrate the nature of the linear relationship observed in the HUD
Grantees evaluation between a child's (log-transformed) blood-lead concentration and the
household's (log-transformed) area-weighted arithmetic average wipe dust-lead loading for floors
and window sills, respectively. The figures portray fitted linear regression models for each
grantee, as well as for the Rochester Lead-in-Dust study and, in Figure 3-21, the Baltimore R&M
study (for comparison purposes). The regression model used only the log-transformed average
dust-lead loading as a predictor variable; the impact of other potentially important predictor
variables on blood-lead concentration was not considered in the model fittings. The regression
lines span the ranges of the observed area-weighted average dust-lead loadings, except data for
five HUD Grantees households (three from Cleveland and one each from Baltimore and Rhode
Island) were omitted from Figure 3-21 as their average window sill dust-lead loadings were
extremely low (less than 0.05 ug/ft2) compared to the other households and were considered too
influential to the model fittings.
When fitting the regression models in Figures 3-20 and 3-21 to the HUD Grantees data, it
was desired to have each household having blood-lead and dust-lead data be represented by only
a single data point. This was possible only if blood-lead data were considered for a single child
in that household. In situations where data for multiple children were available for a single
household, only data for the youngest child older than 12 months of age were considered. This
approach resulted in a single blood-lead result for each household with blood-lead data. In
addition, only data for children meeting the following criteria were included in the regression
modeling:
• Children who lived in the sampled housing unit for at least three months and
before dust and soil samples were collected;
• Children whose blood samples were taken within four months of dust and soil
sample collection;
• Children not having medical treatment for lead poisoning.
181
-------
Table 3-35. Percentage of Children with Elevated Blood-Lead Concentration (at Pre-
Intervention) in the HUD Grantees Evaluation According to Blood Collection
Method, Child Age Category, and Grantee (ages 1-2 years only)
Age Range
1-2 Years
3-5 Years
1-5 Years
*"...•'• Grantee ' . •;::!.*
Alameda County
Baltimore
Boston
California
Cleveland
Massachusetts
Minnesota
New Jersey
Rhode Island
Wisconsin
Milwaukee
Chicago
New York City
Vermont
Age; Range
1-2 Years
3-5 Years
1-5 Years
Grantee
Cleveland
Massachusetts
Minnesota
Rhode Island
Wisconsin
Milwaukee
Vermont
Number of
• Chadren
Percentage of C
ilOpg/dL
hQdren mUi Elevate
"I i 15#g/dL
id Blood-Lead Concentration (%) ,-,
220pg/dL:
•» 25jig(dL
Rood Collection Method = Venipunctura
361
536
897
51
43
46
^ Hood Conection Method =
27
25
20
21
64
43
75
1
14
g
3
28
23
8
22
36
55
14
80
47
61
0
36
44
100
75
4
63
35
27
30
18
12
14
12
7
9
= Venipuncture (Ctiiidren Aged 1-2 Years only) -;/-
11
20
50
5
59
30
44
0
21
22
100
39
0
50
4
8
15
0
38
9
31
0
7
11
67
14
0
13
0
4
5
0
30
7
21
0
0
0
67 .
7
0
0
Blood Collection Method « Fingerstick -.
164
232
396
46
44
45
28
26
27
13
15
14
9
8
9
Bood; Collection Method — Rngerstick (Children Aged 1-2 Years onlyl •
1
4
1
9
43
82
24
100
25
100
33
9
70
29
0
0
100
11
0
50
13
0
0
100
0
0
26
O
0
0
100 .
0
O
17
0
Note: All pre-intervention blood-lead concentration data available and collected through 1/99 are included in the above
summaries.
182
-------
100-
10
0.1 1.0 10.0 100.0 1000.0 10000.0
Carpeted and Uncarpeted Floor Wipe Area-Weighted Arithmetic Average Dust-Lead Loading (
A A A AtamodB Canty
C E E Ctevetand
».i IA..I. rw
IWW flNII Mqr
Figure 3*20. Fitted Regression Models Predicting Children's Blood-Lead Concentration as
a Function of Area-Weighted Arithmetic Average Floor Dust-Lead Loading
(Wipe Collection Method), for the Various Grantees in the HUD Grantees
Evaluation and for the Rochester Lead-in-Dust Study
(Note: Venipuncture blood-lead data were exclusively used in each fitting except for Wisconsin. Milwaukee, and Vermont,
where fingerprick blood-lead data were exclusively used.)
183
-------
100
10-
1-
0.1
1.0 10.0 100.0 1000.0 10000.0
Window Sill Area-Wteighted Arithmetic Average Dust-Lead Loading (
100000.0
A A A Afcmdi County
U O H BflDlNMI
f? r* r- ii.,..,,ii,.,.,di.
I I r MRBaBGnUSBm
Vialf—i M |«WHj(m
O O O BflSfenore RAM
B B B CMRMnb
I I I Rhodotd
aj itj ij >.« vhA
ifi m in now IUK
Figure 3-21. Fitted Regression Models Predicting Children's Blood-Lead Concentration as
a Function of Area-Weighted Arithmetic Average Window Sill Dust-Lead
Loading (Wipe Collection Method), for the Various Grantees in the HUD
Grantees Evaluation and for the Rochester Lead-in-Dust Study
(Note: Venipuncture blood-lead data were exclusively used in each fitting except for Wisconsin, Milwaukee, and Vermont,
where fingerprick blood-lead data were exclusively used.)
The regression models in Figures 3-20 and 3-21 were fitted to blood-lead concentration
data only under the venipuncture method for all but the three grantees (Wisconsin, Milwaukee
and Vermont) for which fingerstick sample results were predominant. For these three grantees,
only fingerstick blood-lead concentration data were used in the regressions.
Note that the slopes of the fitted regression lines in Figures 3-20 and 3-21 are generally
similar in sign and magnitude (given expected ranges of variability) across the grantees and the
two other studies. This suggests that the relationships between blood-lead concentration and
household average dust-lead loading were relatively consistent across grantees. In particular,
184
-------
these relationships were similar to that observed for data from the Rochester study (i.e., the data
used to develop the empirical model presented in Chapter 4 of the §403 risk analysis). This
conclusion is important in that the data from the HUD Grantees evaluation reflect a much larger
geographical area than the Rochester study and represent several types of exposure conditions.
3.6.2 Evidence of the Impact of Housing Age/Condition
on Blood-Lead Concentration
The role that housing age plays in the increased likelihood of a resident child having an
elevated blood-lead concentration has been well-documented and is accepted by many experts in
residential lead exposure. Older housing is more likely to contain lead-based paint hi a
deteriorated condition, which contributes to lead in other environmental media within the
residence, especially those media that is most likely to come into direct contact with children. In
particular, the importance that the level of deterioration plays in the accessibility of lead-based
paint hazards implies that housing condition is an additional key factor in predicting blood-lead
concentration.
Table 3-39 of the §403 risk analysis report summarized data from Phase 2 of NHANES
m to illustrate how geometric mean blood-lead concentration and the percentage of elevated
blood-lead concentrations (i.e., percentage exceeding a given threshold) for children are related
to housing age category. For example, the percentage of children aged 1-5 years with blood-lead
concentration of at least 10 ug/dL increases from 1.6% for children living in post-1973 housing
to 8.6% for children living in pre-1946 housing, with a corresponding geometric mean increase
from 2.0 to 3.8 ug/dL. The Centers for Disease Control and Prevention (CDC) cited these same
results in their 1997 document, Screening Young Children for Lead Poisoning, to support their
conclusion that older housing (i.e., housing built prior to 1950) contained the greatest risk for
lead-based paint hazards.
Figure 6-1 of the CDC's 1991 document, Preventing Lead Poisoning in Young Children
-A Statement by the Centers for Disease Control, presents results from the Cincinnati
Prospective Lead Study (Clark et al., 1985) to illustrate how the combination of housing age and
condition is related to children's blood-lead concentration and how this relationship changes with
the age of the child. This figure is duplicated in Figure 3-22. This figure shows that children's
blood-lead concentration tends to peak at 18-24 months, with the most rapid increase occurring
between 6-12 months. The highest blood-lead levels are associated with housing built prior to
World War n, as well as older housing (predominantly 19th century) that once contained
considerable lead-based paint but which later underwent rehabilitation. Within these groups of
.housing/children living in units in a deteriorated or dilapidated condition had consistently higher
geometric mean blood-lead concentrations through their first three years, with this geometric
mean exceeding 20 ug/dL from about 12 to 24 months of age. CDC used the information
presented in Figure 3-22 to prepare a recommended screening schedule for testing children's
blood-lead levels.
185
-------
Deteriorated
Sit- $»fofact«T
t»i#Wli
fob,» NMic
0369
12 13 1$ 21 24 27 30 33 36 39 42
Child's Age (in Months)
Figure 3-22. Geometric Mean Blood-Lead Concentration Versus Child Age, As Reported
Within the Cincinnati Prospective Lead Study and Presented According to
Housing Age and Condition
(Note: Duplicated from Figure 6-1 of CDC, 1991. Blood-lead concentrations for the same cohort of children
were measured over time. Numbers in parentheses indicate numbers of children with blood-lead information at
18 months of age.)
186
-------
4.0 DOSE-RESPONSE ASSESSMENT
The objective of dose-response assessment was to characterize the relationship between
environmental-lead exposure and the resulting adverse health effects in young children. The
foundation of this characterization was the relationship between environmental-lead levels and
blood-lead concentration. EPA's Integrated Exposure, Uptake, and Biokinetic (IEUBK) model
and an empirical model developed for this risk assessment (Sections 4.1 and 4.2 of the §403 risk
analysis report) were employed to make this characterization.
Section 4.1 of this chapter documents an additional tool, obtained since the §403 risk
analysis report was published, for predicting blood-lead concentration as a function of
environmental-lead levels. This tool is a regression model developed from epidemiological data
collected from 12 studies and was suggested for use in the §403 risk analysis by some
commenters on the §403 proposed rule. As the U.S. Department of Housing and Urban
Development (HUD) sponsored the development of this model, it is referred to in this report as
the "HUD Model." The goal of this model was to "estimate the contribution of lead-
contaminated house dust and soil to children's blood-lead levels" (Lanphear et al., 1998). This
goal is consistent with the objectives of the §403 risk assessment, and so the model merits
consideration for this analysis. Section 4.1 includes documentation on the HUD model, key steps
that were taken in its development, and issues that are necessary to consider when interpreting
the results of the model fits.
Section 4.2 of this chapter contains a revision of the Rochester Multimedia model,
introduced in Section 4.2.3 of the §403 risk analysis report, to allow the model to predict results
that are more comparable to the results of the performance characteristics analysis presented in
the preamble to the §403 proposed rule. The Rochester Multimedia model predicted a geometric
mean blood-lead concentration as a function of average dust-lead loadings in floors and window
sills, dripline soil-lead concentration, and a variable which indicates the presence of deteriorated
lead-based paint and a child with paint pica tendencies. In contrast, the performance
characteristics analysis in the preamble estimated risks associated with dust-lead loadings on
uncarpeted floors, dust-lead loadings on window sills, yardwide average soil-lead concentration,
and the percentage of painted components with deteriorated lead-based paint. Because the
definitions of the data inputs were not always consistent between these two statistical
approaches, their findings were not comparable. Thus, the revised model presented in Section
4.2 uses the same types of data inputs as those used for the performance characteristics analysis.
Section 4.2 also documents multimedia models that omit one or more of the dust, soil, and paint
input variables to obtain predicted blood-lead concentration in instances where data for one or
more of these media were not available.
Section 4.3 provides additional information regarding key assumptions made in the risk
characterization process: the "scaling" algorithm used to determine a post-intervention blood-
lead concentration distribution that is comparable to the baseline distribution, and the issue of
187
-------
adjusting for measurement error when deriving the empirical model used in the §403 risk
analysis.
4.1 HUD MODEL
The HUD model (Lanphear et al., 199S) was developed by a team of researchers and
sponsored by HDD's Office of Lead Hazard Control. This modeling effort used data from 12
epidemiologic studies (hence its frequent reference as a "pooled analysis" model) to make
statistical inferences on the contribution of lead-contaminated house dust and residential soil to
children's blood-lead concentration.
The HUD model predicts a geometric mean blood-lead concentration for children aged 6-
36 months (i.e., the age range of data considered from the 12 studies) as a function of exposure to
specific lead levels in dust, soil, paint, and water, and as a function of other important
demographic variables. Therefore, the model is used to estimate individual risks, or the risks
associated with a class of children determined by specified environmental-lead levels to which
they are exposed. EPA addressed minimizing individual risks when establishing "levels of
concern" for lead in dust and soil within the §403 proposed rule. However, EPA was obliged to
consider population-based risks within a cost-benefit analysis to establish lead hazard standards
within the proposed rule.
4.1.1 Form of the HUD Model
The HUD model takes the following form:
(1) LnfPbB) = 1.496 + 0.183-Ln(DustLead) + 0.01398-Ln(WaterLead) + 0.02116-Ln(ExtLead) +
0,005787-Ln(ExtLead)-ExtType +• 0.4802-Ln(ExtLead)'ExtLoc - 0.1336-ExtType •*• 0.5858-ExtLoc
- 0.02199-Ln(MaxXRF) + 0.0381l-Ln(MaxXRF)-PaintCond - 0.0808-PaintCond + 0.02126-Age -
0.001399-Age2 + 0.00007854-Age3 - 0.3932-Boston - 0.01167-Butte + 0.2027-Bcreek +
0.2392-Cpgm + 0.5383-Csoil + 0.05717-Leadville + 0.1761-Magna - 0.04209-RochLong +
0.07257-RochUD - 0.3712-Sandy + 0.1777-Midvale + 0.123-Race + 0.3175-SES1 + 0.2138-SES2
+ 0.1799-SES3 + 0.1691-SES4 - 0.03233-MouthQften - 0.2454-MouthRare - 0.1397-MouthSome
+ 0.002649-Ln(DustLead)-Age - 0.0003381-Ln(DustLead)-Age2 • 0.00001281-Ln(DustLead)-Age3
+ 0.2212-Ln(ExtLead)-MouthOften + 0.07892-Ln(ExtLead)-MouthRare +
0.1663-Ln(ExtLead)-MouthSome + 0.5305-Ln(WaterLead)-SESl - 0.0136-Ln(WaterLead)-SES2 +
0.1033-Ln(WaterUad)-SES3 - 0.09098-Ln(WaterLead)-SES4 + 0.01192-Age-Race -
0.01023-Age-SESl + 0.003849-Age-SES2 + 0.00008468-Age-SES3 - 0.01679-Age-SES4 + error
where
Ln(PbB) - log-transformed blood-lead concentration (ug/dL)
Ln(DustLead) = log-transformed interior (wipe) floor dust-lead loading (ug/ft2), minus
the mean of the log-transformed data used to develop the model (2.605 ug/ft2)
Ln(WaterLead) = log-transformed water-lead concentration (ppb), minus the mean of the
log-transformed data used to develop the model (0.785 ppb)
Ln(ExtLead) = log-transformed exterior-lead concentration (ppm), minus the mean of the
log-transformed data used to develop the model (6.232 ppm), where the exterior
188
-------
sample is either soil collected at the perimeter of the foundation, soil from the
child's play area, or exterior dust
ExtType = indicator of the type of exterior sample (1 = dust, 0 = soil)
ExtLoc — indicator of whether exterior sample is represented by soil at the perimeter of
the house's foundation (1 = exterior sample is not from perimeter soil, 0 =
exterior sample is from perimeter soil)
Ln(MaxXRF) = log-transformed maximum lead-paint measurement on interior surfaces
(mg/cm2, as measured by XRF), minus the mean of the log-transformed data used
to develop the model (0.921 mg/cm2)
PaintCond = indicator of paint condition (1 = damaged, 0 = undamaged)
Age = age of child (months) minus the mean age of children whose data were used to
develop the model (16.3 months)
Age2 - Age2 - (85.5 + 4.82-Age) (quadratic orthogonal polynomial)
Age3 = Age3 - (-490.71 + 10.32-Age2 + 122.3-Age) (cubic orthogonal polynomial)
Boston, Butte, Bcreek, Cpgm, Csoil, Leadville, Magna, RochLong, RochUD, Sandy, and
Midvale are indicators that the data come from the particular study being represented (1 =
data comes from the particular study, 0 = otherwise)
Race = Race indicator (0 = white, 1 = other)
SES1 = indicator of whether the pseudo-Hollingshead measure of socioeconomic status is
equal to 1 (1 = yes, 0 = no)
SES2 = indicator of whether the pseudo-Hollingshead measure of socioeconomic status is
equal to 2 (1 = yes, 0 = no)
SES3 = indicator of whether the pseudo-Hollingshead measure of socioeconomic status is
equal to 3 (1 = yes, 0 = no)
SES4 = indicator of whether the pseudo-Hollingshead measure of socioeconomic status is
equal to 4 (1 = yes, 0 = no)
MouthOften = indicator of whether mouthing behavior occurs often in the child (1 =
often, 0 = otherwise)
MouthRare = indicator of whether mouthing behavior occurs rarely in the child (1 =
rarely, 0 = otherwise)
MouthSome = indicator of whether mouthing behavior occurs sometimes in the child (1 =
sometimes, 0 = otherwise)
error = random error between the observed log-transformed blood-lead concentration and
what is predicted by the model.
4.1.2 Development of the HUD Model
This section presents several issues on how the HUD model was developed that have a
direct impact on the predicted blood-lead concentration and how this prediction should be
interpreted. These issues include how studies were selected, how study effects were represented
in the model, and how data were handled or adjusted prior to or during the model development
exercise.
189
-------
Study Selection and Potential Selection Bias
The HUD model was developed from environmental-lead, blood-lead, and demographic
data from 12 studies performed over a 15-year time frame (1982-1997). These studies
investigated the relationship between environmental-lead levels and children's blood-lead levels
in various locations and subpopulations. Five of the studies (representing 62% of the data used
to fit the model) were conducted in urban environments:
• Boston Longitudinal Study (Rabinowitz et al., 1985)
• Cincinnati Longitudinal Study (Bornschein et al., 1985b)
Cincinnati Soil Study (Clark et al., 1991)
• Rochester Longitudinal Study (Lanphear et al., unpublished)
• Rochester Lead-in-Dust Study (Lanphear et al., 1996a,b)
The remaining seven studies (representing 38% of the data used to fit the model) were conducted
in milling, mining, or smelter environments:
Bingham Creek, Utah (1993)
• Butte, Montana (1990)
Leadville, Colorado (1991)
Magna, Utah (1994)
Midvale, Utah (1989)
• Palmerton, Pennsylvania (1994)
Sandy, Utah (1994)
According to Lanphear et al. (1998), these 12 studies were selected based on the following
criteria:
• The studies had well-defined sampling protocols for blood and environmental
media (particularly dust, soil, and paint).
• The studies took measures of dust-lead levels, soil-lead levels, paint-lead content
(via XRF), and paint condition.
• The original data were available and could be reanalyzed.
• Dust samples were collected via wipe techniques or by the Dust Vacuum Method
(DVM)
* Dust samples were taken within three months of collecting blood samples from
the resident child(ren) (to address seasonal variation in blood-lead concentration).
• Children were not selected on the basis of having a high blood-lead concentration.
In addition, only cross-sectional data were considered (i.e., data were not considered that could
reflect changes in environmental-lead levels over time).
190
-------
As a result of the inclusion criteria, data for at least seven other studies considered by
HUD were excluded from the model development process. These studies and the primary
reasons for their exclusion (when specified within Lanphear et al., 1998) were
• UK Study (Davies et al., 1990) - dust collection method not wipe or DVM
• Boston and Baltimore segments of the EPA Urban Soil Lead Abatement
Demonstration Study (USEPA, 1996a; Weitzman et al., 1993) - dust collection
method not wipe or DVM
• Australian National Survey (1996) - lack of XRF paint-lead levels
Baltimore R&M Study (USEPA, 1996c; USEPA, 1997c) - method for selecting
children did not meet the criteria, and data were not available to the analysis
• Telluride, CO (1987) - reason for exclusion not given
• Trail, BC (1992) - reason for exclusion not given.
Lanphear et al. (1998) indicates that the 12 studies were not chosen to represent the entire
nation or even communities like those in which the studies were conducted. In fact, these studies
were conducted in communities with a recognized environmental-lead hazard, and any abatement
efforts within each study targeted those hazards. Furthermore, the study effects included in the
model were treated as fixed effects (i.e., they are the only studies of interest in the model-
building process) rather than random effects (i.e., they are assumed to be a random sample of a
larger population of studies). Thus, if the model is used to estimate risks to a broader population
of children than simply those within the 12 studies, additional information is needed to determine
the extent to which the pooled data used to develop the model are representative of the U.S.
housing stock.
Fixed vs. Random Study Effects and Interaction with the Study Effects
As stated in the previous paragraph, the study effect in the HUD model is a series of fixed
effects. If the study effect was assumed to be random instead of fixed (i.e., the studies can be
considered a random selection of all such residential-lead exposure studies), then study-to-study
variation would become a contributor to total variation in the prediction. Based on work with
previous models that incorporate a random study effect, the study-to-study component is typically
a major portion of total variability in the prediction. Thus, the variability associated with
predictions by the HUD model is likely underestimated.
Additional underestimation in variability may result from the absence of interaction terms
in the model between study effects and other environmental exposure factors. This can
underestimate variability associated with inferences involving the environmental exposure
factors, including the principal inferences which involve dust-lead and exterior-lead levels.
Adjusting for Measurement Error in Environmental-Lead Predictor Variables
The HUD model parameter estimates were determined from Simulation Extrapolation
(SIMEX) methods, which attempted to quantify the theoretical relationship between blood-lead
191
-------
and "error-free" measures of environmental-lead. As a result, the parameter estimates were
adjusted to reflect measurement error present among the environmental-lead variables. However,
the goal is to predict children's blood-lead concentrations as a function of wipe dust-lead
loadings and soil-lead concentrations as they would be measured in a risk assessment, not their
true, "error-free" (but unobservable) values. Carroll et al. (1995) states that for prediction
problems, adjusting for the effects of measurement error in predictor variables is rarely
necessary. Adjusting for measurement error in these predictor variables tends to increase the
values of the slope parameters associated with these variables, which in turn can inflate predicted
blood-lead concentrations. Thus any predictions from the HUD model fits should be properly
labeled that values of the predictor variables are assumed to be "error-free" rather than measures
of environmental-lead levels taken from activities such as a risk assessment (as the predictor
variables hi the models used in the §403 risk analysis were assumed to represent).
Making Survey Variable Definitions Consistent Across Studies
Because different survey designs in different studies can result in different definitions for
a common survey measure (e.g., SES, mouthing behavior, paint condition), which in nun can
introduce considerable complication when interpreting model predictions, certain survey
measures were redefined to make them more consistent across studies. Each redefinition
transforms the original data values to values generated from a domain that is consistent across the
studies. However, such a transformation does not remove all study-to-study differences in these
values. In particular, it does not consider factors that impact how the specific study
measurements were obtained and which differ from study to study, such as the use of different
survey instruments and different approaches to administering the instruments.
Converting DVM Dust-Lead Loadings to Wipe-Equivalents
While it was desired to have floor dust-lead loading assuming wipe dust collection as a
predictor variable in the HUD model, some of the 12 studies used DVM methods to collect dust
samples. Rather than exclude data from these studies from consideration in the model
development effort, a procedure was derived to convert DVM dust-lead loadings reported in
these studies to wipe-equivalent loadings. This procedure used data from the Butte study to
develop the following conversion equations:
for carpeted floors,
for hard floors.
= 0.7727 + 0.9821-Log^DVM)
Log^Wipe) = 0.7762 + 0.4839-Log,JDVM)
where "Wipe" and "DVM" indicate wipe dust-lead loadings and DVM dust-lead loadings,
respectively (Westat, 1998). Note that these same two equations were used to convert DVM
dust-lead loadings in each study, regardless of whether the relationship between DVM and wipe
dust-lead loadings differed among the studies.
The method for deriving these conversion equations included a procedure to adjust for
measurement error in the DVM dust-lead loading measurements. However, the purpose of the
192
-------
conversion was to predict a measured wipe dust-lead loading based on a measured DVM dust-
lead loading, not the (unobservable) "true" DVM dust-lead loading. Thus, some bias may have
been introduced in this conversion process.
Interpreting the "Exterior-Lead" Predictor
Certain households whose data were used hi the HUD model development effort did not
have soil-lead data for various reasons (e.g., no bare soil). In these instances, exterior dust-lead
concentration was generally measured instead. As a result, among the predictor variables in the
HUD model was an indicator variable that identified whether or not an exterior-lead
measurement was from dust or soil. This indicator allowed both the model intercept and the
slope factor associated with exterior-lead concentration to change according to its value.
However, no other consideration was made for differences in sampling and analysis methods
between soil and exterior dust and the impact such differences can have on the reported lead
levels. In addition, no indication was given that differences in bioavailability between soil-lead
and exterior dust-lead were considered in developing the model.
When a household had soil-lead data available, data from foundation perimeter (i.e.,
dripline) soil were used when available; otherwise, play-area soil-lead data were used. Like the
soil vs. exterior-dust issue in the previous paragraph, the HUD model includes an indicator
variable that identified whether or not soil-lead levels represent dripline soil. This indicator
allowed both the model intercept and the slope factor associated with exterior-lead concentration
to change according to its value. However, certain study-to-study differences in collecting soil
samples or obtaining a soil-lead measurement were not considered in model development, such
as the depth of soil sampling, soil surface type (e.g., covered vs. bare), chemical methods for the
digestion and analysis of soil samples, and soil compositing.
Handling Missing Water-Lead Measurements
While the HUD model included water-lead concentration as a predictor variable, it was
necessary to impute values for this measurement during model development when data for a
given household were not available. Water-lead data were unavailable for all households in two
studies and up to 12% of study households in the other ten studies. In such instances, imputed
measurements were randomly generated from a lognormal distribution with geometric mean
equal to that observed from data for other study households (if data for other households were
available) or to the community-wide average (if data for other households were not available).
Handling Data Reported Below a Detection Limit
When data values reflected measurements at or below some detection limit, the HUD
model development effort replaced them with probability-based values between zero and the
detection limit, where probabilities were determined from a lognormal distribution associated
with data above the detection limit. According to Table 2.15 of Westat (1998), the incidence of
not-detected results was high with XRF paint-lead level, and to a lesser extent, with water-lead
193
-------
concentration. In half of the studies, the percentage of not-detected paint-lead results ranged
from 20% to 86% (with the detection Limit being reported at either 0.1 or 0.7 mg/cm2). The
percentage of not-detected water-lead concentrations (i.e., results somewhere below 5 ppb)
exceeded 80% in two studies. In contrast, none of the lead levels in blood, dust, or soil samples
were reported below a detection limit in 10 of the 12 studies, and lead levels in no more than 9%
of these samples were reported below a detection limit in the other two studies.
Selecting from Multiple Observations Within a Household
When blood-lead concentration data were available for multiple children within a
household (as occurred in nine of the studies), only data for one child selected at random from
the household were considered in the HUD model development. If blood-lead data existed at
multiple time points for the same child (such as at 6,18, and 24 months of age in the Boston
Longitudinal study), those time points whose data met the initial inclusion criteria were identified
(e.g., lead interventions did not occur between the time points), and data for the time point
having dust-lead measurements taken more closely in time were selected for the model
development. Similarly, when environmental-lead measurements were repeatedly taken over
time for a given household, data for the time point closest to a blood-lead measurement were
used.
Handling Seasonalitv Effects
The effect of seasonality on blood-lead concentrations was given some consideration in
the modeling effort (e.g., blood and dust samples must have been collected within three months
of each other for their data to be included). However, there is no effect of seasonality included hi
the final model. It is unclear whether seasonality was determined not to be a significant effect
among the pooled data, or whether a seasonality term was intentionally left out of the model.
4.1.3 Interpreting Results of Fitting the HUD Model
The previous section discussed issues concerning the pooled study data and development
of the HUD model that should be understood when using the model to estimate risks, as is done
in Section 5.1.1 and Appendix F. This section addresses the interpretation of results from fitting
the HUD model and, in particular, caveats associated with certain interpretations.
Individual risks vs. population-based risks
As mentioned earlier, the HUD model estimates individual risks associated with lead
exposure to children aged 6-36 months. While the §403 risk analysis included individual risks
analyses, which EPA used in efforts to establish levels of concern for lead in environmental
media, EPA was required to employ cost-benefit analysis to select §403 hazard standards. The
cost-benefit analyses used population-based risks (i.e., risks posed by childhood lead exposure to
the nation as a whole) to estimate the benefit and cost associated with performing interventions
and other activities in response to §403 rules.
194
-------
Individual risks and population-based risks are generally not comparable. This must be
understood when attempting to compare the individual risks estimated by fitting the HUD model
at specified environmental-lead levels with the population-based risk estimates found in the §403
risk analysis.
Interpreting Model Parameter Estimates
The prediction parameters in the HUD model are not independent. For example, it is
known that soil-lead and dust-lead measures are correlated. Therefore, it is not appropriate to
interpret the parameter estimates in the HUD model (or in the models developed for the §403 risk
analysis) in isolation. Using the parameter estimates to characterize a cause-and-effect
relationship that is attributable to a single parameter alone, such as measuring the extent of an
increase in blood-lead concentration associated with a given increase in dust-lead loading, is very
problematic.
One example of how correlation among the predictor variables can influence the model
parameter estimates is seen with maximum XRF paint-lead measurement. One would expect a
positive correlation between maximum XRF paint-lead measurement and blood-lead
concentration, and as a result, a positive slope parameter. However, the estimated slope
parameter is negative (-0.022), although not significantly different from zero. The negative
estimate is likely due to confounding between paint-lead measurements and other predictor
variables. The likelihood of confounding increases with the number of parameters in the model.
Problems with interpreting model parameter estimates in isolation emphasizes the need to
consider total exposure (i.e., prediction based on considering the joint effect of all model
parameters) rather than exposure associated with a single environmental medium. In the §403
situation, protectiveness needs to be judged by recognizing that hazard standards exist for dust,
soil, and paint, and that resulting actions from these multiple standards will determine the level
of protection, not just the actions associated with a single standard. For example, the level of
protection associated with a dust-lead loading standard of 5 ng/ft2, without consideration of other
standards, may equal that associated with a joint set of standards that involve a higher dust-lead
loading standard.
Interpreting Results at Low Environmental-Lead Exposures
The HUD model and the models developed for the EPA risk analysis are "log-log"
models. That is, they predict log-transformed blood-lead concentration as a linear function of
log-transformed environmental-lead levels. As the log transformation "stretches out" the lower
portion of the scale and contracts the upper portion of the scale, very low environmental-lead
levels and blood-lead concentrations have undue influence on inferences made from the models.
For example, the effect of increasing dust-lead loading from 1 to 10 jig/ft2 is equal to the effect of
increasing dust-lead loading from 10 to 100 ug/ft2. Therefore, inferences at such low levels can
be overestimated and misleading. Thus, any inferences at very low dust-lead loadings, such as 1
or 5 ug/ft2, should be made with caution.
195
-------
4.1.4 Conclusions
The following conclusions can be made on the HUD model and the comparison of risk
estimates originating from this model versus those originating from models used in the §403 risk
analysis:
• As the HUD model parameters associated with environmental-lead measurements
in specific media have been adjusted for measurement error, the input parameters
to this model are assumed "true" lead levels in these media. This can provide
biased results when the model is used to predict blood-lead concentration
associated with lead levels measured in current risk assessments. The Rochester
multimedia model and the empirical model did not have such an adjustment
incorporated.
• While the HUD model contains study effects, they are considered fixed effects
and therefore allow the model to make predictions for only the group of children
represented by the 12 studies. Furthermore, the study effects impact only the
intercept of the model, and any study-to-study differences that may be present in
other model terms (such as in environmental-lead measurements) are not
represented.
• The HUD model handles "exterior-lead measurements" (e.g., soil) differently than
the §403 models; the impact of such difference has not been determined.
4-2 ALTERNATIVE MULTIMEDIA MODELS FOR PREDICTING
BASED ON ENVIRONMENTAL-LEAD LEVELS
As discussed in the introduction to this chapter, the Rochester multimedia model,
presented in Section 4.2.3 of the §403 risk analysis report, was developed using data from the
Rochester Lead-in-Dust study to explain children's blood-lead concentration as a function of
dust-lead loadings from floors (carpeted and uncarpeted) and window sills, dripline soil-lead
concentration, and an indicator variable on the presence of deteriorated lead-based paint and a
child with paint pica tendencies. This model was used in the risk characterization (Section 5.3 of
the §403 risk analysis report) to determine the probability that a child exposed to specific levels
of lead in paint, dust, and soil would have a blood-lead concentration at or above 10 ug/dL. EPA
used these estimates of individual risk, as well as the findings of performance characteristics
analyses detailed in Section 6.1 of this report, in proposing levels of concern for lead in dust
(page 30318 in the §403 proposed rule).
The §403 proposed rule considered uncarpeted floors and a yard-wide average soil-lead
concentration when proposing dust and soil standards and levels of concern. The performance
characteristics analysis cited in the proposed rule considered these types of dust-lead and soil-
196
-------
lead measures. However, in the Rochester multimedia model, the floor dust-lead loading
measure did not limit the type of floor surface to uncarpeted floors, the soil-lead measure
represented only dripline soil, and the paint/pica indicator variable was different from the paint
measure used in the performance characteristics analysis (the percentage of tested components in
the home with deteriorated lead-based paint). For these reasons, it was difficult to compare
estimates of individual risks based on this model to results obtained from the performance
characteristics analyses. Thus, it was desired to fit an alternative multimedia model (cited as
"Model A" in this section) that replaced the floor dust-lead and soil-lead predictor variables used
in the Rochester multimedia model with uncarpeted floor dust-lead loading and yard-wide
average soil-lead concentration, respectively, and replaced the paint/pica indicator variable with a
measure of the percentage of tested components containing lead-based paint.
While a household risk assessment for lead-based paint hazards is expected, at a
minimum, to characterize lead levels in floor dust and to identify the extent of deteriorated lead-
based paint, it is possible that some risk assessments may not measure lead levels in soil or
window sill dust. Therefore, to investigate how individual risks would be characterized in these
types of risk assessments, two alternative multimedia models were fitted that were reduced
versions of Model A. One model excluded soil-lead concentration as a predictor variable
("Model B"), and the other model excluded both soil-lead concentration and window sill dust-
lead loading as predictor variables ("Model C").
The three alternative multimedia models were fitted using data from the Rochester Lead-
in-Dust study, using the same approach used to fit the Rochester multimedia model in the §403
risk analysis. The models were log-linear in nature, where the dust-lead and soil-lead measures
were log-transformed, and the models predicted a log-transformed blood-lead concentration. For
example, Model A took the following form:
log(PbB) = po + p,*log(PbF) + p2*log(PbW) + p3*log(PbS) + p4*PbP
where PbB represents blood-lead concentration (ug/dL), PbF represents household average dust-
lead loading for uncarpeted floors (fig/ft2), PbW represents household average dust-lead loading
for window sills (ug/ft2), PbS represents yard-wide average soil-lead concentration (ug/g), and
PbP represents the larger (between the interior and exterior of the housing unit) of the
percentages of tested components containing deteriorated lead-based paint. As with the
Rochester multimedia model, ordinary least squares regression methods were used to fit the
models to the Rochester data.
Table 4-1 presents the estimates of the model parameters for each of the three alternative
multimedia models. Note that the model fits were based on different numbers of housing units,
as eliminating certain predictor variables from the above model resulted in more housing units
that had all necessary data available for fitting the model. An investigation of model diagnostics
showed that the extent of collinearity among the predictor variables in these models was low.
Generally, the slope estimates associated with the paint variable were very low, and except for
197
U.S. EPA Headquarters Library
Mail code 3201
1200 Pennsylvania Avenue NW
Washington DC 20460
-------
Table 4-1. Parameter Estimates and Associated Standard Errors for the Three
Alternative Multimedia Models Fitted to Rochester Study Data to Predict
Log-Transformed Blood-Lead Concentration1
Parameter
Po
p,
>
P3
p*
R2
0&«r
n
Predictor Variable
Intercept
log (PbF): Area-Weighted Arithmetic Mean
(Wipe) Dust-Lead Loading from Uncarpeted
Floors
log (PbW): Area-Weighted Arithmetic Mean
(Wipe) Dust-Lead Loading from Window
Sills
log (PbS): Yardwide Average Soil-Lead
Concentration2 (fine soil fraction)
PbP: The larger of the following two
percentages: % of interior tested surfaces
that contain deteriorated LBP, and % of
exterior tested surfaces that contain
deteriorated LBP3
Coefficient of Determination
Error
# data points included in model fitting
Parameter Estimate (Standard Error)
Model A
0.331
(0.263)
0.114
(0.049)
0.082
(0.037)
0.115
(0.040)
0.001
(0.002)
20.03%
0.561
177
Model B
0.899
(0.183)
0.130
(0.048)
0.101
(0.035)
—
0.002
(0.002)
15.62%
0.572
188
Model C
1.337
(0.122)
0.140
(0.043)
—
—
0.004
(0.002)
11.38%
0.580
196
"-* indicates that the variable is not included in the model as a predictor. The models are log-linear in nature.
' One housing unit (cid = 01689) had an uncarpeted floor dust-lead loading measurement of 18130.0/yg/ft2 (only one
uncarpeted floor wipe sample was collected in this unit). This data value was omitted when fitting the above models as it
was highly influential and led to a noticeable reduction in the estimate of p,.
2 Yardwide soil-lead concentration at a given housing unit was calculated as the unweighted arithmetic average of dripline
and play area soil-lead concentrations. If one or the other is missing (but both are not missing), yardwide concentration was
set equal to the non-missing value. If both are missing, yardwide concentration is missing.
3 If one or the other of these two percentages is missing (but both are not missing), the value of this variable is set equal to
the non-missing value. If both are missing, the value is missing.
198
-------
Model C, were not significantly different from zero at the 0.05 level. See footnotes to this table
for additional details on the model fits.
Other alternative multimedia models were considered when initially developing the
Rochester multimedia model. These models, and information used to evaluate these models in
selecting the final version used in the §403 risk analysis, were presented in Appendix G of the
§403 risk analysis report.
4.3 SUPPLEMENTAL INFORMATION ON MODEL-BASED
APPROACHES IN THE S403 RISK ANALYSIS
Some comments on the §403 proposed rule addressed issues concerning approaches taken
in the §403 risk analysis in which statistical models were used to predict a post-intervention
distribution of blood-lead concentration, and therefore, a means of assessing how health risks
associated with lead-based paint hazards would change as a result of implementing the §403 rule.
The types of models used in this analysis and the approach to characterize post-intervention
health risks were presented in Chapters 4 and 6 of the §403 risk analysis report. In addition,
technical details on these models and approaches were provided in appendices to the report.
However, certain comments on the §403 proposed rule indicated that this information may not
have been provided in sufficient detail or would have benefitted from additional clarity. In
particular, two issues in question involved how the §403 risk analysis estimated a post-
intervention blood-lead concentration distribution that was comparable to the baseline
distribution that was characterized by data from Phase 2 of NHANES m, and how the empirical
model (Section 4.2 of the §403 risk analysis report) account for measurement error issues
associated with the predictor variables. The following subsections provide additional
information on these two issues, specifically geared toward addressing the specific areas raised
by selected public comments.
4.3.1 The "Scaling" Algorithm Used to Determine a Post-
Intervention Blood-Lead Concentration Distribution
In the §403 risk analysis, EPA used data from Phase 2 of NHANES m (collected from
1991-1994) as the basis for the baseline ("pre-§403 rule") characterization of children's blood-
lead concentration in the U.S. housing stock. As discussed in Section 3.5 of the §403 risk
analysis report, EPA took this approach because these data were considered the best available
data (as well as the most recent data) on blood-lead measures, the data consisted of actual blood-
lead measurements from a nationally-representative survey, and it was preferred (and considered
more defensible) to use such data to characterize the baseline distribution rather than data
generated from statistical prediction models. However, because the "post-§403 rule" time period
has not yet occurred, and it was desired to compare the blood-lead distribution in this time period
to the baseline blood-lead distribution, it was necessary to use statistical prediction models to
generate how the baseline distribution would change following interventions performed as a
result of the §403 rule.
199
-------
When estimating the blood-lead concentration and health effect endpoints used in the
§403 risk analysis, it was assumed that the national distribution of blood-lead concentration was
lognormally distributed. Initial investigations into the weighted NHANES ffi blood-lead
concentration data used in the §403 risk analysis (Figure 5-3 of the §403 risk analysis report)
suggested that this was a satisfactory assumption. The lognormal distribution is characterized by
the geometric mean and geometric standard deviation (GSD) of the data (or, equivalently, the
exponentiated mean and exponentiated standard deviation of the log-transformed data). Once the
geometric mean and GSD were calculated from the weighted NHANES ffl data, the additional
assumption of lognormality was used to obtain baseline estimates of the health effect and blood-
lead concentration endpoints considered in the §403 risk analysis (e.g., probability that a child's
blood-lead level was at or above 10 ug/dL). These estimates were presented in Table 5-1 of the
§403 risk analysis report.
To estimate how the blood-lead distribution changed between the "pre-§403 rule" and
"post-§403 rule" environments (based on a given set of candidate §403 standards and on
assumed changes in environmental-lead levels resulting from implementing the §403 rule),
model-based estimates of the blood-lead distribution were made for both environments. For
reasons explained in the §403 risk analysis report, EPA chose to characterize both the pre-§403
and post-§403 blood-lead distributions twice, with the first characterization using the IEUBK
model and the second characterization using the empirical model. Each characterization
involved fitting the given model to environmental-lead data separately for each home in the HUD
National Survey and weighting each prediction appropriately to represent a given proportion of
the nation's children. Therefore, although the HUD National Survey homes do not represent a
random sample of homes in the national housing stock, and therefore, the set of predicted blood-
lead concentrations do not themselves represent a random sample of blood-lead levels in the
nation's children, the fact that each prediction is weighted appropriately to represent a given
proportion of the nation's children allows the total set of predictions to be a good estimate of the
national blood-lead distribution, in either a pre-§403 or post-§403 environment.
Appendix E2 of the §403 risk analysis report discusses how model-predicted blood-lead
concentrations generated for the HUD National Survey homes are used to estimate the geometric
mean and GSD associated with the national blood-lead distribution. Recall that for a given HUD
National Survey home, the model-predicted blood-lead level represents a geometric mean of
children whose exposure is characterized by the environmental-lead levels in that home.
Therefore, the estimated GSD of the national blood-lead distribution is characterized not only by
the variability among the predicted blood-lead levels, but also by the assumed variability in
blood-lead levels among individual children exposed to the same environmental-lead levels.
^.uux.. Under a given model (i.e., either the IEUBK or empirical model), the "scaling" algorithm
(Appendix Fl of the risk analysis report) involves calculating the proportional change in the
geometric mean and GSD of the model-predicted blood-lead distribution from the "pre-§403
rule" to the "post-§403 rule" environment. Then, the same proportional change in both statistics
was applied to the geometric mean and GSD of the baseline distribution determined from the
NHANES ffl data:
200
-------
The resulting geometric mean and GSD, along with an assumption that lognormality still holds,
characterized a blood-lead concentration that represented the post-§403 environment and that
was considered comparable to the baseline distribution.
This type of algorithm was necessary due to the difficulties associated with comparing a
model-based post-§403 blood-lead distribution directly with a baseline distribution that was
characterized from observed (NHANES HI) data. While the empirical model was calibrated so
that its estimate for the baseline national geometric mean blood-lead concentration for children
aged 1-2 years (as obtained within the approach taken in the §403 risk analysis) equaled the
NHANES DI Phase 2 estimate (although the predicted GSD was not similarly calibrated), such a
calibration was not possible for the BEUBK model. As a result, using the HUD National Survey
data as input, these models could not both predict the same national estimate of the geometric
mean blood-lead concentration as Phase 2 of NHANES HI. In addition, the empirical model
could not be developed based on data from a national survey that measured both blood-lead and
environmental-lead levels, which would have facilitated direct comparisons of the predicted
blood-lead distribution between pre-§403 and post-§403 environments. Therefore, the "post-
1403 rale" blood-lead distributions predicted by the two models had some inconsistency with the
baseline distribution estimated from the NHANES HI data, making direct comparisons
problematic. For example, if a model underestimates the geometric mean, the benefits associated
with the §403 rale could be overestimated, while if a model overestimates the geometric mean,
this could result in estimates of negative benefits. Therefore, the "scaling" algorithm used in the
risk analysis used the models to predict the change in the geometric mean and GSD that occurs
from a pre- to post-§403 rule environment, then applied this same change to the baseline
distribution.
Note that the scaling algorithm does not require that the two model-based blood-lead
distributions (pre-§403 and post-§403) be independent of each other. In fact, the two
distributions are dependent, because the post-§403 environmental-lead levels, used to predict
post-§403 blood-lead levels, are dependent on the pre-§403 levels. Only the geometric mean and
GSD of these two distributions are necessary to characterize, and they are used simply to
estimate the proportional change in the geometric mean and GSD of the national blood-lead
distribution between pre-§403 and post-§403 conditions.
While the geometric mean and GSD are scaled separately, one change is not necessarily
independent of the other. For example, if the pre-§403 geometric mean and GSD both have high
values, they are both likely to be reduced at a greater rate than at lower values. The approach
was kept as simple as possible while retaining scientific defensibility, in order that it be easily
applied during risk characterization and in the economic analysis.
In the peer review of the §403 risk analysis report, EPA specifically asked the peer
reviewers to comment on whether the scaling procedure was scientifically defensible in general,
201
-------
and in particular, whether it was relevant in the situation where the environmental-lead data
(from the HUD National Survey) and the blood-lead data (from NHANES ID) were collected at
different periods in time. None of the peer reviewers specifically criticized the scaling algorithm.
Furthermore, the Science Advisory Board reviewing the §403 risk analysis stated that, in general,
the approach was scientifically defensible as presented, and specifically, the multi-step approach
was warranted due to the need to use various datasets of differing sources and representing
different time periods to make the characterization.
See Section 6.4.4 below for an alternative approach to applying this scaling algorithm
where the probability of a child's blood-lead concentration exceeding 10 ug/dL is scaled rather
than the GSD.
4.3.2 Adjusting the Empirical Model Parameter
Estimates to Reflect Measurement Error
The approach to the measurement error adjustment, discussed in Section 4.2.4 of the §403
risk analysis report, attempted to correct for the fact that while the empirical model was
developed using data from the Rochester Lead-in-Dust study, it was used to predict a (pre-403)
geometric mean blood-lead concentration assuming that the data input to the model originated
from the HUD National Survey. Because the Rochester study and the HUD National Survey
used different sampling schemes involving different collection devices and instruments, as well
as different analytical methods, and because the ranges of observed environmental-lead levels
differed between the two studies, it was necessary to adjust the model parameter estimates to
reflect these differences prior to allowing the model to accept HUD National Survey
environmental-lead data as input to the prediction.
Note that the measurement error adjustment made to the empirical model was not to
address the more-standard "errors in variables" issue (Carroll et al., 1995) which attempts to take
into account that a value input to the model represents a measurement subject to error, rather than
a "true" value. As discussed in Section 4.2.4 and Section G4.2 (Appendix G) of the §403 risk
analysis report, the empirical model was not intended to be used in the risk analysis as a dose-
response model, which would have required the predictor variables to reflect actual exposures.
Instead, the model assumed that its input environmental-lead information reflected measurements
that would have been made as a result of a risk assessment within a home. Therefore, adjusting
the model for the fact that its inputs reflect measured rather than actual lead levels was
considered inappropriate for this analysis. This decision in the type of application that is
represented by the §403 risk analysis has been concurred upon in the published literature (e.g.,
Carroll and Galindo, 1999).
When using the empirical model to predict a post-403 geometric mean blood-lead
concentration, some of the HUD National Survey dust-lead and soil-lead data (i.e., those data for
homes that exceed the candidate 403 standards) were modified to reflect the impact of
performing interventions in response to the 403 rule on these measured data values (see Table
6-2 of the §403 risk analysis report), then the model is fitted to the modified data. These
202
-------
modified data are still considered to be measured lead levels, rather than actual (or "true") lead
levels. However, the modified data values must first be transformed to represent measurements
that would have made under the methods used in the HUD National Survey. For example, the
assumed post-intervention floor wipe dust-lead loading of 40 ug/ft2 must be converted to a Blue
Nozzle vacuum-equivalent loading prior to using it as input to the empirical model.
203
-------
This page intentionally blank.
204
-------
5.0 RISK CHARACTERIZATION
Chapter 5 of the §403 risk analysis report documented the final portion of the risk
assessment phase of the §403 risk analysis, in which the methods introduced in the earlier
chapters of the §403 risk analysis report were applied to characterize risks associated with current
(i.e., baseline) lead exposures for children aged 1-2 years. The baseline distribution of blood-
lead concentration in this population was characterized using data from Phase 2 of the Third
National Health and Nutritional Examination Survey (NHANES ID), conducted from 1991 to
1994. Alternative pre-§403 risk estimates were also calculated as a function of environmental-
lead levels by using data from the HUD National Survey as input to the EUBK and empirical
models. Both individual risk estimates (i.e., risks associated with specific environmental-lead
levels) and population-based risk estimates (i.e., average risks over the entire nation) were
presented. As mentioned in Section 2.0, the specific blood-lead concentration and health effect
endpoints used to measure the risks of lead exposure to children aged 1-2 years were
• Incidence of blood-lead concentration greater than or equal to 10 ug/dL
• Incidence of blood-lead concentration greater than or equal to 20 ug/dL
• Incidence of IQ score less than 70 in the population of U.S. children, which results
from lead exposure
• Incidence of IQ score decrement (in the population of U.S. children) greater than
or equal to 1 resulting from lead exposure
• Incidence of IQ score decrement (in the population of U.S. children) greater than
or equal to 2 resulting from lead exposure
• Incidence of IQ score decrement (in the population of U.S. children) greater than
or equal to 3 resulting from lead exposure
• Average IQ decrement within the population of U.S. children that results from
lead exposure.
The risk characterization included a sensitivity and uncertainty analysis where possible
alternatives to various approaches taken and assumptions made in the risk characterization were
identified and incorporated into the analysis, and the resulting impact on the risk estimates was
evaluated. This analysis resulted in a measure of the uncertainty associated with the risk
estimates due to methodological assumptions, thereby producing a range of estimates within
which the true risk may reasonably be expected to fall. Section S.I of this chapter contains the
following additional sensitivity and uncertainty analyses that were performed and documented
since the §403 risk analysis report was published:
• Calculate individual risks associated with specified lead levels in floor-dust and
soil, as predicted by the HUD model introduced in Section 4.1 and the alternative
multimedia models presented in Section 4.2.
• Calculate estimates assuming a 50% decline in the estimated geometric mean
blood-lead concentration of children aged 1-2 years from the estimate generated
205
-------
from Phase 2 of NHANES m (in addition to the estimates associated with 10%,
20%, and 30% declines that were presented in Section 5.4.3 of the §403 risk
analysis report).
* Calculate model-based estimates of the pre-§403 blood-lead distribution under
revised environmental-lead levels (from the HUD National Survey) input to the
models, with the revisions representing the potential change in these levels that
may have occurred since the survey was performed.
• Calculate baseline estimates of the IQ-related health effect endpoints assuming
that specified non-zero thresholds exist in the relationship between blood-lead
concentration and IQ.
5.1 RISK CHARACTERIZATION SENSITIVITY AND UNCERTAINTY ANALYSIS
The following subsections present the results of additional sensitivity and uncertainty
analyses performed to gauge the level of uncertainty in baseline risk estimates associated with
methodological assumptions. These results should be considered with those presented in the
sensitivity and uncertainty analyses in Section 5.4 of the §403 risk analysis report to characterize
overall uncertainty associated with the methods and assumptions taken in the risk assessment.
5.1.1 Estimates of Individual Risks from Applying the HUD Model
In Section 5.3 of the §403 risk analysis report, the concept of individual risks was
introduced, and estimates of individual risks associated with lead exposure in children were
generated by fitting the IEUBK and Rochester multimedia models to specified environmental-
lead levels. Briefly, within the context of the §403 risk analysis, individual risks refer to the risks
associated with a young child's exposure to specified levels of environmental-lead. Once
environmental-lead levels were specified for each medium, the model-predicted blood-lead
concentration at these levels, along with the assumption that blood-lead concentrations have a
lognormal distribution with a specified variability, were used to estimate the percentage of
children exposed to the specified set of environmental-lead levels that would have elevated
blood-lead concentrations (i.e., at or above 10 |ag/dL). Then, those sets of environmental-lead
levels associated with estimated elevated blood-lead percentages of 1%, 5%, and 10% were
identified and presented in Tables 5-5 through 5-7 and Figures 5-7 and 5-8 of Section 5.3 of the
§403 risk analysis report. The IEUBK model was used to identify soil-lead concentrations
associated with these elevated blood-lead percentages (at specified dust-lead loadings), while the
Rochester multimedia model was used to identify (wipe) dust-lead loadings associated with these
elevated blood-lead percentages (at specified soil-lead concentrations). These results contributed
to the information which EPA used in proposing dust and soil levels of concern in the §403
proposed rule. See Section 5.3 of the §403 risk analysis report for additional details.
As discussed in Section 4.1 of this report, the HUD model was published shortly after the
§403 .risk analysis report was finalized, and some commenters to the §403 proposed rule
206
-------
suggested that it be considered within the §403 risk analysis. The HUD model can be used in the
same manner as the §403 risk analysis models to predict individual risks associated with
exposure to a specified set of environmental-lead levels. Therefore, this section presents
estimates of individual risks based on fitting the HUD model to specified environmental-lead
levels and compares these risk estimates with those based on the EEUBK model (for soil) or the
Rochester multimedia model (for dust). Supporting summaries and discussion for comparing
HUD model results with those of the multimedia models developed for the §403 risk analysis are
presented in Appendix F.
Soil-Lead Concentrations
When fitting the HUD model to evaluate individual risks associated with yard-wide
average soil-lead concentration, the model's soil-lead parameters were set to the following
values:
• ExtType = 0 (indicating that soil was sampled rather than exterior dust)
• ExtLoc = 0.5 (indicating that the total soil sampled was a rough average of drip-
line and non-drip-line soil)
Figure 5-1 plots estimates of the percentage of children's blood-lead concentrations at or
above 10 ug/dL as a function of soil-lead concentration, as predicted by the IEUBK model and
the HUD model. The left-most panel in Figure 5-1 corresponds to the results of IEUBK model
fits at specified dust-lead concentrations (100,200, and 500 ug/g) and is identical to Figure 5-7
in the §403 risk analysis report. The middle and right-most panels of Figure 5-1 correspond to
fits of the HUD model at specified dust-lead loadings (5,10,20,25,40,50,100 and 200 ug/ft2)
and differ according to the assumed geometric standard deviation (GSD) associated with the
blood-lead concentration distribution (GSD=1.6 for the middle panel; GSD=1.72 for the right-
most panel). A GSD of 1.72 was estimated within the HUD model publication (Lanphear et al.,
1998).
The IEUBK and HUD model fits portrayed in Figure 5-1 are not directly comparable as
the IEUBK model controls for dust-lead concentration while the HUD model controls for dust-
lead loading. However, the plots do suggest that the predicted patterns of change in blood-lead
concentration with soil-lead concentration differ considerably for the two models.
Table 5-la, identical to Table 5-5 of the §403 risk analysis report, presents the soil-lead
concentrations for which the IEUBK model-predicted percentages of children having blood-lead
concentrations of at least 10 ug/dL equal 1%, 5%, and 10% assuming dust-lead concentrations of
100,200, and 500 ug/g. Table 5-lb presents the soil-lead concentrations for which the HUD
model-predicted percentages of children having blood-lead concentrations of at least 10 ug/dL
equal 1%, 5%, and 10% assuming the dust-lead loadings considered in the HUD model panels of
Figure 5-1. Table 5-lb (HUD model) indicates that, for all dust-lead loadings considered, the
soil-lead concentrations estimated to maintain risks at 1% and 5% are less than 400 ug/g; even a
soil-lead concentration of less than 6 ug/g would not achieve these levels of protection if children
207
-------
88
ggssssssss
-IP/BTT oicaqd »« 't4iii«il»«i=>-'d
r§
0*
"S O
I
in
2
3
208
-------
Table 5-1 a. Yard-Wide Average Soil-Lead Concentrations at Which the Percentage of
Children Aged 1-2 Years With Blood-Lead Concentration At or Above 10
A/g/dL is Estimated by the 1EUBK Model at 1, 5, or 10%, for Three Assumed
Dust-Lead Concentrations (Table 5-5 in §403 risk analysis report).
Blgg^^
HHHiHiiiHI
100
200
500
155
35
Not achievable
'.;•:, . -
365
245
Not achievable
l&^MgSSSSiSSSSIiSS^BBS^SSH
515
395
25
Table 5-1 b. Yard-Wide Average Soil-Lead Concentrations at Which the Percentage of
Children Aged 1-2 Years With Blood-Lead Concentration At or Above 10
//g/dL is Estimated by the HUD Model at 1, 5, or 10%, for Eight Assumed
Dust-Lead Loadings and Two Assumed Geometric Standard Deviations
i^iwai^gtt^H
5
10
20
25
40
50
100
200
.'.. .. : • '. .."•
5
10
20
25
40
50
100
200
BBHflBBBiHHBBHBilBffff'''*
GSD
74.4
49.4
32.9
28.8
21.8
19.1
12.7
8.5
;'-;•: -:;h;;-: •'- GSD
45.9
30.5
20.3
17.8
13.5
11.8
7.9
5.2
186.2
123.7
82.3
72.1
54.7
47.9
31.9
21.2
_ t;72 1 ' ",::. • •
132.4
88.0
58.5
51.3
38.9
34.1
22.7
15.1
opi^usni
»M»W!PP«BHSralHB«HHKiSKPlBWS»
303.7
201.8
134.2
117,6
89.2
78.2
52.0
34.5
232.8
154.8
102.9
90.2
68.4
60.0
39.9
26.5
209
-------
were exposed to dust-lead loadings of 40 ug/ft2 or more. Based on Table 5-lb, if dust-lead
loadings are equal to 5 jig/ft2, a soil-lead concentration of approximately 300 ug/g (100 ug/g)
maintains the percentage of blood-lead concentrations greater than or equal to 10 ug/dL at 5% for
a GSD of 1.60 (1.72).
Floor Dust-Lead Loadings
Figures 5-2 and 5-3 plot the estimated percentages of children having blood-lead
concentrations at or above 10 ug/dL as a function of floor dust-lead loadings as predicted by the
HUD model and the Rochester multimedia model assuming GSDs of 1.60 and 1.72, respectively,
on the blood-lead distribution. Soil-lead concentrations are assumed to be fixed at 100,400,
1200,2000 and 5000 ug/g and, for the Rochester multimedia model, window sill dust-lead
loadings are assumed to be fixed at 200 and 500 ug/ft2. In each figure, the left-most panel
contains estimates based on fitting the HUD model. The Rochester multimedia model panels for
GSD equal to 1.60 and soil-lead concentrations of 100 and 400 ug/g were presented in Figure 5-8
in the §403 risk analysis report.
Tables 5-2 and 5-3 present the floor dust-lead loadings that are predicted by the HUD
model and Rochester multimedia model, respectively, to mamtaim the percentage of children
having blood-lead concentrations above or equal to 10 ug/dL at 1%, 5%, and 10% for specified
levels of soil-lead concentration, window sill dust-lead loading and GSD. Approximate 95%
upper confidence bounds, which account for the variability of parameter estimates from the
Rochester multimedia model, are also provided in Table 5-3. See Appendix C2, Section 5.0 of
the §403 risk analysis report for the methodology used to compute these confidence bounds. The
rows of Table 5-3 corresponding to a GSD of 1.60 and soil-lead concentrations of 100 and 400
ug/g are identical to Table 5-6 in the §403 risk analysis report (after correcting an error in the
computation of the confidence bounds).
Some key findings noted when comparing the individual risk estimates presented in
Appendix F between the HUD model and Rochester multimedia model include the following:
• " At very low floor dust-lead loadings (i.e., 1-5 ug/ft2), the HUD model and the
Rochester multimedia model yield similar predictions for the geometric mean
blood-lead concentration, which also results in similar predictions for the health-
effect endpoints that are calculated directly from this geometric mean (e.g.,
percentage of children with blood-lead concentration at or above a specified
threshold; average IQ decrement resulting from lead exposure). However, due to
the forms of these models and concerns involving the accuracy of very low dust-
lead measurements, any conclusions made at such low dust-lead loadings must be
made with caution.
• The predicted geometric mean blood-lead concentration under the HUD model
ranges from 20% to nearly 60% higher than the prediction under the Rochester
multimedia model as floor dust-lead loadings increase from 15 to 100 Mg/ft2 and
210
-------
r§
§ •!
f Mi
S S S S S
S 2
! I 1 i
1 ! i
i i
III!
ill
l. ' i
I ! i
11
I I
i i '' :
1 ' !
1 1 ! i
\ \\ \
\ I 1
'• \\ i
\
\ I \ '•
\ \ • \
* \ \ !
\ \ \ i
\ \ > '•
\ \N
» x \
^•>^-.
• s
o o
i n
' 1 «^
1 1 ^
K- ! 1 <.
.§S§ii il
KS W* » |4
i i i 3jr
^ 1 ?
.g 1 | S|
" 1 1 ]}
g S
* § ^ ^
-7
^
sssssass"
s s s s s s R
-IP/Or* Ol*8
-------
s
§ §1
§
s
s s
ass
§ §1
• i
fl"
a
1
-I §1
•1
i
lili
1111
co o> o
= c a
l
«
il
O O 3 =
Q.S 11.5
in
212
-------
Table 5-2. Floor Dust-Lead Loadings at Which the Percentage of Children Aged 1-2
Years With Blood-Lead Concentration At or Above 10//g/dL is Estimated by
the HUD Model at 1, 5, or 10%, for Five Assumed Soil-Lead Concentrations
and Two Assumed Geometric Standard Deviations
t^
100
400
1200
2000
5000
. - . . - ..T^j, ::.:;;;
100
400
1200
2000
5000
-------
Table 5-3. Floor Dust-Lead Loadings at Which the Percentage of Children Aged 1-2
Years With Blood-Lead Concentration At or Above 10 figldL is Estimated by
the Rochester Multimedia Model at 1, 5, or 10%, for Five Assumed Soil-
Lead Concentrations, Two Assumed Window Sill Dust-Lead Loadings, and
Two Assumed Geometric Standard Deviations (expanded version of Table 5-
6 in §403 risk analysis report).
« - „ uSP = 1 .BU
100
400
1200
2000
5000
100
400
1200
2000
5000
200
500
200
500
200
500
200
500
200
500
200
500
200
500
200
500
200
500
200
500
0.05
0.02
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
i':' '.",- '...
0.00
o.oo
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.37
0.14
0.04
0.02
0.01
0.00
0.00
0.00
0.00
0.00
GSQ!t=1.7J
0.04
0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
6.70
2.00
0.61
0.18
0.09
0.03
0.04
0.01
0.01
0.00
8;J"-f:- '• : '
1.11
0.33
0.10
0.03
0.02
0.00
0.01
0.00
0.00
0.00
22.00
8.93
2.81
1.12
0.59
0.22
0.28
0.10
0.07
0.02
4.90
1.95
0.64
0.24
0.12
0.04
0.06
0.02
0.01
0.00
89.08
26.62
8.13
2.43
1.22
0.36
0.50
0.15
0.10
0.03
327.73
92.01
20.32
9.24
4.86
1.99
2.47
0.96
0.69
0.25
21.87
6.54
2.00
0.60
0.30
0.09
0.12
0.04
0.03
0.01
67.96
24.54
7.04
3.00
1.59
0.61
0.77
0.29
0.20
0.07
1 The 95% upper confidence bounds here differ from those in Table 5-6 of the 5403 risk analysis report. The values here
are corrected for a mistake in the original computations.
214
-------
as soil-lead concentrations decrease from 2000 ppm to 10 ppm (assuming, for the
Rochester multimedia model, that window sill dust-lead loadings are at their
estimated national median level; Tables F-l and F-2 of Appendix F). Note that
for a fixed value of the geometric standard deviation (GSD) for the blood-lead
distribution, the average IQ decrement in the population that is associated with
lead exposure is a multiple of the geometric mean (as calculated hi the §403 risk
analysis). Therefore, similar differences in predictions between the two models
would occur for average IP decrement.
If the geometric standard deviation (GSD) associated with the blood-lead
distribution is fixed, then as floor dust-lead loadings increase beyond 10 ug/ft2,
the predicted percentage of children with blood-lead levels at or above 10 ug/dL
increases at a much faster rate under the HUD model (at a constant soil-lead
level). For example, if window sill dust-lead loading is at its estimated national
median and soil-lead concentration is below 2000 ppm, the predicted percentage
under the HUD model is at a minimum twice as large as the prediction under the
Rochester multimedia model. This difference in predictions gets even greater as
the assumed soil-lead concentration gets lower. For example, at a GSD of 1.6, a
floor dust-lead loading of 100 ug/ft2, and a soil-lead concentration of 10 ppm, the
prediction is over 7 times higher for the HUD model compared to the Rochester
multimedia model (13.1% versus 1.76%; Tables F-3 and F-4 of Appendix F).
5.1.2 Estimates of Individual Risks from Applying the
Alternative Rochester Multimedia Model
The approach to estimating individual risks that was discussed and applied (using the
IEUBK model and Rochester multimedia model) in Section 5.3 of the §403 risk analysis report
and was applied using the HUD model in Section 5.1.1 above was also applied using the
alternative Rochester multimedia model ("Model A") that was documented in Section 4.2 of this
report. Recall that this model uses 1) average dust-lead loadings on uncarpeted floors, 2) average
dust-lead loadings on window sills, 3) yardwide average dust-lead concentration, and 4) the
percentage of painted components containing deteriorated lead-based paint as predictor variables.
The blood-lead concentration predicted by this model is an estimate of the geometric mean of the
distribution of blood-lead concentration, which is assumed to be lognormal with geometric
standard deviation 1.6. The resulting distribution is then used to estimate the percentage of
children with blood-lead concentration at or above 10 fig/dL. This subsection documents the
results of estimating individual risks based on the alternative Rochester multimedia model.
In this analysis, it is assumed that the given residential environment for which individual
risks will be estimated has no deteriorated lead-based paint (as was done in Section 5.3 of the
§403 risk analysis report). Then, levels of two of the remaining three predictor variables are
fixed, and the level of the third variable is determined so that either 1%, 5%, or 10% of children
215
-------
exposed to the combined lead levels given by the three predictor variables would be predicted to
have a blood-lead concentration at or above 10 ug/dL.
Table 5-4a contains the estimated floor dust-lead loading at which the predicted
percentage of children with blood-lead concentration at or above 10 ug/dL is either 1%, 5%, or
10%, given that window sill dust-lead loading is at either 200 or 500 ug/ft2, and soil-lead
concentration is at either 100 or 400 ug/g. Table 5-4b contains the estimated window sill dust-
lead loading at which the predicted percentage of children with blood-lead concentration at or
above 10 ug/dL is either 1%, 5%, or 10%, given that floor dust-lead loading is at either 25 or 100
ug/ft2, and soil-lead concentration is at either 100 or 400 ug/g. These two tables correspond to
Tables 5-6 and 5-7, respectively, in Section 5.3 of the §403 risk analysis report, where the
estimates were based on the original Rochester multimedia model. Finally, Table 5-4c contains
the estimated yard-wide average soil-lead concentration at which the predicted percentage of
children with blood-lead concentration at or above 10 ug/dL is either 1%, 5%, or 10%, given that
floor dust-lead loading is at either 25 or 100 ug/ft2, and window sill dust-lead loading is at either
200 or 500 ug/ft2. (A corresponding table for the soil-lead concentration does not exist in
Section 5.3 of the §403 risk analysis report as the soil-lead concentration predictor variable in the
original Rochester multimedia model assumed only dripline soil rather than a yard-wide average,
and the IEUBK model did not accept dust-lead loadings as input.)
Effect on risk analysis: If the target percentage of children with elevated blood-lead
concentration is 5%, the estimates in Table 5-4a (fourth column) are only slightly higher than the
corresponding estimates in Table 5-6 of the §403 risk analysis report, suggesting that at the fixed
levels of soil-lead concentration and window sill dust-lead loading, the corresponding floor dust-
lead loading is nearly the same for the two methods determined by the two forms of the
multimedia model. However, at a soil-lead concentration of 100 ug/g, the estimated floor dust-
lead loadings are considerably smaller (and below 50 ug/ft2) than in Table.5-6 of the §403 risk
analysis report under 10% risk and considerably larger (but still below 1 ug/ft2) under a 1% risk.
For window sills (Table 5-4b), estimated dust-lead loadings are reduced under the
alternative Rochester multimedia model compared to the estimates hi Table 5-7 of the §403 risk
analysis report at the 5% and 10% risk levels. At the 5% risk level, the estimated window sill
dust-lead loadings were at 40 ug/ft2 or lower at the specified soil-lead and uncarpeted floor dust-
lead levels (Table 5-4b, fourth column).
Yardwide average soil-lead concentration needed to be below 150 ug/g at each of the
specified levels of floor and window sill dust-lead loadings to achieve risks of 10% or lower, and
at 32 ug/g or lower to achieve risks of 5% or lower (Table 5-4c). The estimated soil-lead
concentration increases as the two specified dust-lead loadings decrease, illustrating how the soil-
lead standard could become less stringent as the dust-lead loading standards became more
stringent.
216
-------
Table 5-4a. Uncaroeted Floor Dust-Lead Loadings at Which the Percentage of Children
Aged 1-2 Years With a Blood-Lead Concentration At or Above 10/ig/dL is
Estimated by the Alternative Rochester Multimedia Model (A) at 1 %, 5%, or
10%, Under Fixed Levels of Yardwide Average Soil-Lead Concentration and
Window Sill Dust-Lead Loading
Yardwide Average
Soil-Lead Cone.
. U'g/gJ
100
400
Window SHI Dust-
Lead Loading
teg/ft2)
200
500
200
500
Uncarpeted Floor Dust-Lead Loading (//g/ft2) at Which the
Estimated % of Children With Blood-Lead Concentration
At or Above 10 f/gfdL is ...
1%
0.48
0.25
0.12
0.06
5%
7.9
4.1
1.9
1.0
10%
35
18
8.7
4.5
Note: The percentages of 1%, 5%, and 10% were determined assuming that the blood-lead distribution is
lognormal with geometric mean as predicted by the alternative Rochester multimedia model (A) and a
geometric standard deviation of 1.6.
Table 5-4b. Window SHI Dust-Lead Loadings at Which the Percentage of Children Aged
1-2 Years With a Blood-Lead Concentration At or Above 10/ig/dL is
Estimated by the Alternative Rochester Multimedia Model (Al at 1 %, 5%, or
10%, Under Fixed Levels of Yardwide Average Soil-Lead Concentration and
Uncarpeted Floor Dust-Lead Loading
Yardwide Average
, Soil-Lead Cone.
(P9/3)
100
400
Uncarpeted Floor
1 Dust-Lead
Loading (//g/ft2)
25
100
25
100
Window Sill Dust-Lead Loading (pg/ft2) at Which the
Estimated % of Children With Blood-Lead Concentration
At or Above 10/sg/dl is ...
1%
0.78
0.11
0.11
0.02
5%
40
5.7
5.6
0.80
10%
320
46
45
6.5
Note: The percentages of 1%, 5%, and 10% were determined assuming that the blood-tead distribution is
lognormal with geometric mean as predicted by the alternative Rochester multimedia model (A) and a
geometric standard deviation of 1.6.
217
-------
Table 5-4c. Yardwide Average Soil-Lead Concentration at Which the Percentage of
Children Aged 1-2 Years With a Blood-Lead Concentration At or Above 10
//g/dL is Estimated by the Alternative Rochester Multimedia Model (A) at
1%, 5%, or 10%, Under Fixed Levels of Dust-Lead Loadings for Uncarpeted
Floors and Window Sills
Uncarpeted Floor
- Dust-Lead.
Loading (pg/ft2)
25
100
Window Sill Dust-
Lead Loading
dig/ft2)
"•"'•* 1" j 1^ * . J
*i-* •& *• •<
• > *
200
500
200
500
Yardwide Average Soil-Lead Concentration (frg/g) at
. Which the Estimated % of Children With Blood-Lead
Concentration At or Above 10 //g/dL is ...
'' -":i% :s ';-»'.
2.0
1.0
0.50
0.26
L >B%"~
32
17
8.0
4.2
"' fcjf . *
"- *10% ? - '
140
73
35
18
Note: The percentages of 1%, 5%, and 10% were determined assuming that the blood-lead distribution is
lognormal with geometric mean as predicted by the alternative Rochester multimedia model (A) and a
geometric standard deviation of 1.6.
5.1.3 Considering Potential Declines in Biood-Lead Concentration
from NHANES III Phase 2 Measures
The results of this subsection are an extension of the analysis in Section 5.4.3 of the §403
risk analysis report. In that subsection, the geometric mean blood-lead concentration of 3.14
ug/dL for children aged 1-2 years, estimated from data collected in Phase 2 of NHANES ffl
(1991-1994), was assumed to be either 10%, 20%, or 30% lower, and the resulting impact on the
baseline risk estimates was investigated. This analysis was performed due to the likelihood of
continued decline in blood-lead concentrations in the U.S. population that has occurred in recent
years. It was desired to augment this analysis by considering an additional assumption on the
percentage decline since Phase 2 of NHANES m 50%.
Table 5-5 presents the baseline estimates of the blood-lead concentration and health effect
endpoints for children aged 1-2 years, where each blood-lead concentration measurement in
Phase 2 of NHANES ffl was reduced by the same amount: 10%, 20%, 30%, or 50%. Thus, the
analysis assumed a constant percentage decline for the entire blood-lead concentration
distribution as characterized by Phase 2 of NHANES ffl. This table is an extension of the results
presented in Table 5-11 of the §403 risk analysis report and includes the baseline estimates
reported in the risk analysis for comparison purposes (i.e., where no reduction is assumed).
Note that within NHANES ffi, the estimated geometric mean blood-lead concentration
for children aged 1-2 years declined from 4.05 ug/dL in Phase 1 to 3.14 (ag/dL in Phase 2,
representing a 22.5% decline. This is within the range of declines being considered in the
218
-------
Table 5-5. Sensitivity Analysis for the Estimated Baseline Number and Percentage of
Children Aged 1-2 Years Having Specific Health Effect and Blood-Lead
Concentration Endpoints, Assuming Various Percentage Declines in Blood-
Lead Concentration Since Phase 2 of NHANES III
Health Effect and Blood-Lead
Concentration Endpoints
PbB* 20^g/dL
PbBi 10//g/dL
IQ score less than 70
IQ score decrement x 1
IQ score decrement s 2
IQ score decrement * 3
Average IQ score decrement
Geometric Mean (pg/dL)
: : Numbers (%) of ChOdren Aged 1-2 Years
Risk Analysis
Estimate Oablft 5-1
of the §403 risk
:- analysts report)
46,800
(0.588%)
458,000
(5.75%)
9,130
(0.115%)
3,060,000
(38.5%)
863,000
(10.8%)
294,000
(3.70%)
1.06
3.14
-;- ~DlAWAAM+.hMA nAjJma in tt
: - , • Percentage Decline m at
; Since NHAN1
10%
30,900
(0.388%)
340,000
(4.27%)
8,610
(0.108%)
2,640,000
(33.2%)
669,000
(8.40%)
213,000
(2.68%)
0.951
2.82
20%
18,900
(0.238%)
239.000
(3.00%)
8,160
(0.102%)
2,190,000
(27.6%)
493,000
(6.19%)
146,000
(1.83%)
0.845
2.51
ood-Lead Concentration
ES III Phase 2
30%
10,600
(0.133%)
156,000
(1.96%)
7,760
(0.098%)
1,740,000
(21.8%)
340,000
(4.27%)
91,900
(1.15%)
0.740
2.20
:- :"50%:fe
2,130
(0.0268%)
46,800
(0.588%)
7,140
(0.0897%)
863,000
(10.8%)
117,000
(1.47%)
25,200
(0.317%)
0.528
1.57
sensitivity analysis within Table 5-5. However, due to the NHANES HI survey design and how
this survey was performed, caution must be taken when interpreting observed differences in
results between the two phases of this survey.
Effect on risk analysis: According to Table 5-5, if it were assumed that a 50% across-
the-board decline in blood-lead concentration (resulting in a national geometric mean blood-lead
concentration of 1.57 ug/dL), this would reduce, the estimated number of children whose blood-
lead concentration was at or above 20 ug/dL from 46,800 to 2,130, a decline of 95%, while the
estimated number at or above 10 ug/dL would be reduced by nearly 90%, to 46,800 children.
The 50% decline resulted in percentage declines of 12%, 86%, and 91% for numbers of children
with IQ score decrements of 1,2, or 3, respectively, as a result of lead exposure. The estimated
average IQ decrement in the population due to lead exposure is cut in half under this assumption
(from 1.06 to 0.53 points), matching the assumed 50% decline in blood-lead concentration
because the IQ/blood-lead concentration relationship is assumed to be linear across the entire
range of blood-lead concentration. The effects of the lower assumed percentage declines (10%-
30%) were discussed in Section 5.4.3 of the §403 risk analysis report.
219
-------
The results in Table 5-5 are based on the assumption that the blood-lead concentrations
for each child in the population have been reduced by the same percentage since Phase 2 of
NHANES ffl. In reality, different subgroups have achieved different rates of change over this
time. However, considering different percentage declines for different subgroups would be very
difficult, and the resulting estimates of the health effect and blood-lead concentration endpoints
would likely differ only slightly from that observed in Table 5-5.
5.1.4 Considering How Baseline Environmental-Lead Levels May Have
Changed Since the HUD National Survey
Although interim data from the NSLAH (Section 3.1) have recently been made available
to this risk analysis and have been summarized throughout this report, the fact that the public
could not have reviewed these summaries during the public comment period limits the extent to
which these data could be considered in the rulemaking. Therefore, for purposes of the
rulemaking, data from the HUD National Survey continue to be the only nationally-
representative data source on baseline environmental-lead levels in the nation's housing stock.
Nevertheless, it was desired to estimate how changes in these environmental-lead levels that have
occurred since the HUD National Survey was conducted would affect the baseline (i.e., pre-
1403) risk characterization. Therefore, a sensitivity analysis was performed where the HUD
National Survey data was adjusted to reflect possible change in the distribution over time. The
adjusted data would yield a surrogate distribution of baseline environmental-lead levels. Several
alternative adjustments would be considered, and risk estimates based on each set of adjusted
data would be calculated.
To help in determining appropriate adjustments to the HUD National Survey data, the
summaries presented in Section 3.2 of this document compared the distribution of dust-lead and
soil-lead data reported in the HUD National Survey with distributions from other studies
performed more recently, but typically in specific locations that may not necessarily be
nationally-representative. These summaries showed that the distributions were quite consistent
across studies, suggesting that the distributions based from the HUD National Survey data, even
after converting from Blue Nozzle dust-lead loadings to wipe-equivalent loadings, are likely
adequate for characterizing environmental-lead levels even up to ten years after the survey. In
fact, the HUD National Survey data distributions were often centered at lower lead levels than in
the other studies. (A primary exception was household average dust-lead loadings, where data
from the interim NSLAH were considerably lower than in other studies,'including the HUD
National Survey. It is currently uncertain of the degree to which the observed distribution from
the interim NSLAH reflects actual declines in dust-lead levels since the HUD National Survey.)
Furthermore, the nature of the distribution appears to be affected by housing age, with higher
lead levels associated with older housing.
For this sensitivity analysis, the following five alternatives for adjusting the HUD
National Survey data were made, in an effort to reflect more current environmental-lead levels in
households:
220
-------
• Average dust-lead loading and dust-lead concentration reduced by 20%, (yard-
wide) average soil-lead concentration reduced by 20%
• Average dust-lead loading/concentration reduced by 50%, average soil-lead
concentration reduced by 50%
• Average dust-lead loading/concentration reduced by 50%, average soil-lead
concentration reduced by 0%
• Average dust-lead loading/concentration reduced by 0%, average soil-lead
concentration reduced by 50%
• Average dust-lead loading/concentration increased by 25%, average soil-lead
concentration increased by 25%.
The dust-lead loading assumptions are assumed to be for both floors and window sills and are
made to the reported Blue Nozzle loadings (i.e., those estimates used as input to the empirical
model). The same changes are assumed for Blue Nozzle floor dust-lead concentration, which is
used as input to the IEUBK model.
Each of the above five sets of alternatives implies that the same percentage change would
be applied to data from each housing unit in the HUD National Survey. Thus, the resulting
national distribution of baseline environmental-lead levels would be a simple shift in the current
distribution used in the §403 risk analysis, with no change in the variability associated with the
distribution. Insufficient data exist to determine how a distribution's variability may have
changed, so it is assumed to remain unchanged. Within each set, different percentage changes
are considered for dust-lead loading and soil-lead concentration to allow for added flexibility in
how lead levels may have changed in different media. Four of the five sets represent declines
over time, which are expected due to the increased prevalence of homes with no lead-based paint
hi the housing stock and the reduced likelihood of residual contamination associated with leaded
gasoline emissions. Nevertheless, one set representing an increase is considered, due to the
potential for the new survey to include housing with generally higher levels than those houses
included in the original survey and the continued potential for deteriorated lead-based paint and
other lead sources to contaminate dust and soil.
Note that in all five sets of adjustments, the assumed within-house geometric standard
deviation (GSD) remains equal to 1.6. Alternative values of this GSD assumption were
considered in the sensitivity analysis presented in Section 5.4.6 of the §403 risk analysis report.
Table 5-6 presents the pre-§403 model-based estimates for the health effect and blood-
lead concentration endpoints, under each of the above five sets of data adjustments, as calculated
based on blood-lead distribution generated from IEUBK and empirical model fits. For
comparison purposes, the table includes the estimates assuming no adjustments (i.e., as the data
were used in the §403 risk analysis) as reported in Table 5-2 of the §403 risk analysis report.
221
-------
Table 5-6. Sensitivity Analysis on How Changes in Household Average Baseline Dust-
Lead Loadings/Concentrations and Soil-Lead Concentration Impact Pre-§403
Estimates of Health Effect and Blood-Lead Concentration Endpoints for
Children Aged 1-2 Years
Assumed Percentage Change in Average Dust-Lead-Loadings and Concentrations ,
(Both Floor and Window SOI) and in Yard-wide Average Soil-Lead Concentration s - •
Dust:
Soil:
. No change
No change
20%
decrease
20%
decrease
50%.
decrease
50%
decrease
50%
decrease
No change
•J— .J.-. nnt*
no cnange
50%
decrease
25%
increase
25%
- increase
Predicted Health Effect And Blood-Lead Concentration Endpoints {Based on Empirical Model)
PbB 220 {%)
PbB2lO(%)
IQ < 70 <%)
IQ decrement 2 1 (%)
IQ decrement 22 (%>
IQ decrement 23 (%)
Avg. IQ decrement
0.0278
1.54
0.0997
34.5
4.53
0.718
0.932
0.0212
1.28
0.0983
31.8
3.87
0.584
0.896
0.0117
0.849
0.0957
26.5
2.74
0.373
0.825
0.0187
1.17
0.0977
30.6
3.61
0.532
0.880
0.0176
1.13
0.0974
30.1
3.49
0.509
0.873
0.0364
1.85
0.101
37.2
5.27
0.877
0.969
Predicted Health Effect And Blood-Lead Concentration Endpoints (Based on IEUBK Model)
PbB 220 (%)
PbB2lO{%)
IQ < 70 (%)
IQ decrement 2! (%)
IQ decrement 22 (%)
IQ decrement 28 (%)
Avg. IQ decrement
2.24 .
12.4
0.146
50.4
19.9
8.95
1.40
1.39
9.33
0.131
45.1
15.8
6.46
1.24
0.427 .
4.60
0.110
34.6
8.97
2.90
0.978
0.957
7.28
0.121
40.3
12.8
4.92
1.12
1.44
9.70
0.132
46.4
16.4
6.72 ^
1.26
3.06
15.3
0.160
55.4
23.8
11.3
1.56
Effect on risk analysis: The greatest total decline in baseline environmental-lead levels
being considered is the set containing 50% declines in both dust- and soil-lead levels (i.e., the
fourth column of Table 5-6). Under the empirical model, Table 5-6 indicates that the most
sensitive endpoints to the 50% decline in both dust-lead and soil-lead are the incidence of IQ
decrement of at least 3 and the incidence of blood-lead concentration of at least 10 ug/dL, where
declines of 48% and 45%, respectively, were observed in these estimates relative to no decline in
environmental-lead levels. The empirical model-based estimates appear to be more sensitive to
changes in soil-lead concentration than in dust-lead concentration, as lower estimates were
observed when soil-lead concentrations declined by 50% (and no change was made to dust-lead
loadings) than when dust-lead loadings declined by 50% (and no change was made to soil-lead
concentrations). This is explained by the empirical model's larger slope estimate for soil-lead
222
-------
concentration than for dust-lead loading in either floors or window sills (Table 4-3 of the §403
risk analysis report).
When the IEUBK model is used to estimate the distribution of blood-lead concentration,
corresponding declines of 68% and 63% were observed for incidence of IQ score decrement at or
above 3 and for incidence of blood-lead concentration at or above 10 ug/dL, respectively, when
50% declines are assumed for both dust-lead and soil-lead concentration (Table 5-6). However,
hi this same scenario, the greatest decline (81%) among the endpoints is observed with the
incidence of blood-lead concentration at or above 20 ug/dL. This is a considerable decline
compared to the 24% decline for this endpoint observed under the empirical model. Contrary to
the type of finding observed under the empirical model, the IEUBK model-based estimates
appear to be more sensitive to changes in dust-lead concentration than in soil-lead concentration,
as lower estimates were observed when dust-lead concentrations declined by 50% (and no change
was made to soil-lead concentrations) than when soil-lead concentrations declined by 50% (and
no change was made to dust-lead concentrations).
Under both models, the last column in Table 5-6 shows that only modest increases in the
risk estimates were observed under the one adjustment assumption involving increases in
environmental-lead levels (i.e., 25% increases in both dust-lead and soil-lead levels).
5.1.5 Impact on the Estimated Incidence of IQ Point Decrement
Assuming Certain Thresholds on the IQ/Blood-Lead Relationship
As discussed in Chapter 4 of the §403 risk analysis report, results of the meta-analysis
documented hi Schwartz (1994) indicate that an average IQ point loss of 0.257 is predicted for
every 1.0 ug/dL increase in blood-lead concentration, with no evidence of a threshold in this
relationship (i.e., non-zero blood-lead concentration below which the predicted IQ point loss is
zero). These results were used in the §403 risk analysis to characterize the IQ/blood-lead
relationship. Section 2.3 of this report provides additional justification for making these
assumptions.
As discussed in Section 2.3, some researchers have suggested that a non-zero threshold
exists in the IQ/blood-lead relationship. While no consensus on a single threshold has been
adopted among those making this conclusion, and such conclusions are occasionally made by
visual inspection of data rather than on statistical criteria, this sensitivity analysis considers the
impact of assuming non-zero blood-lead concentration thresholds on the baseline and model-
based pre-§403 risk estimates. A non-zero threshold will result in reduced estimates for health
effects measured by IQ decrement, as children with blood-lead concentrations below the
threshold will have an estimated IQ decrement of zero due to lead exposure.
Estimates of the IQ decrement parameters under the following thresholds are presented in
this subsection: 1,2, 3, 5,8, and 10 ug/dL. hi addition, the estimates under an assumed
"threshold" of 0 ug/dL (i.e., those measured in the §403 risk analysis and meaning that any
blood-lead level, regardless of how small, would have an adverse effect on a child's IQ score) are
223
-------
presented for comparison purposes. The candidate threshold of 8 ug/dL has been suggested by
Rabinowitz et al. (1992), as discussed in Section 2.3. The candidate threshold of 10 (ig/dL was
selected as it represents the action level reported by the Centers for Disease Control and
Prevention (Section 2.5.1 of the §403 risk analysis report). It also is representative of the higher-
level thresholds reported by some early studies; thresholds any higher than 10 ug/dL would result
in extremely low (and likely very underestimated) risk estimates and have been discounted by
more recent studies. Levels of 1,2, 3, and 5 ug/dL represents possible candidates of a very low,
but positive, threshold. Such thresholds would not be detectable by many studies in the literature
as they tend to fall below the range of observed data or the detection limit of the blood-lead
measurement procedure.
The assumption of a positive threshold requires a minor modification to the method used
to predict IQ score decrement based on blood-lead concentration. An average IQ point loss of
0.257 continues to be predicted for every 1.0 ug/dL increase in blood-lead concentration, but
only above the assumed blood-lead concentration threshold value. Thus, if T represents the
threshold, then the predicted IQ score decrement at a blood-lead concentration of C would equal
0.257*(C-T) if C is greater than T, or zero if C is less than or equal to T. While the
methodology used to obtain risk estimates remains the same as that documented in Appendices
El and £2 of the §403 risk analysis report, slight differences were required for calculating the
average and standard deviation of IQ decrement, as this measure was no longer assumed to be
lognormally distributed. See Appendix B for how these statistics are calculated assuming a non-
zero threshold.
Table 5-7 presents the estimated percentages of children with IQ score decrements greater
than or equal to 1,2, or 3, and the average and standard deviation IQ point decrement under
assumptions of an IQ score decline of 0.257 points for every 1.0 ug/dL increase in blood-lead
concentration above the specified threshold. These estimates are presented assuming the baseline
blood-lead distribution (top section of the table), the pre-§403 distribution as generated by
BEUBK model fits (middle section of the table), and the pre-§403 distribution as generated by the
empirical model fits (bottom section of the table) for children aged 1-2 years.
Effect on risk analysis: The magnitude of the assumed blood-lead concentration
threshold has a considerable impact on the percentage of children affected by decrements in IQ
score. As seen in Table 5-7, while the §403 risk analysis estimated an average IQ decrement of
1.06 points occurs due to lead exposure across the population of children aged 1-2 years, this
average declines by approximately 44% under a assumed threshold of 2 ug/dL (0.588 points) and
by 90% under a threshold of 8 ug/dL (0.103 points). An estimated 38.5% of children aged 1-2
years were expected to experience an IQ score decrement of at least 1 if a threshold was not
assumed. This percentage is decreased by approximately 50% under a threshold of 2 ug/dL
(19.6%) and by 90% under a threshold of 8 ug/dL (3.5%). The percentage decline is decreased
in magnitude as the lower limit of IQ score decrement increases to 3, but it remains at least a
39% decline for a threshold of 2 ug/dL and 83% for a threshold of 8 ug/dL.
224
-------
Table 5-7. Sensitivity Analysis on the Assumed Blood-Lead Concentration Threshold on
IQ Decrement and Its Impact on the Pre-§403 Estimates of IQ Decrement
Endpoints for Children Aged 1-2 Years
Assumed
Threshold
: (pg/dU
0
1
2
3
5
8
10
% of Children Aged 1-2 Years with a Specified IQ
Decrement Due to Lead Exposure1
IQ Decrement 2
'"i:;;,U'L;;
10. Decrement 2
•';•" ":':';2 •..'
Baseline Estimates (Section 5.1
38.5
27.3
19.6
14.2
7.83
3.50
2.15
10.8
8.08
6.10
4.66
2.80
1.40
0.915
IQ Decrement 2 ;
•.' ' '-3 •
Average IQ
Decrement
(# points)2
. Standard
Deviation of IQ
Decrement?
.1 of §403 risk analysis report)
3.70
2.88
2.26
1.80
1.16
0.627
0.429
1.06
0.804
0.588
0.428
0.233
0.103
0.0638
0.895
0.891
0.860
0.802
0.666
0.494
0.408
Pre-§403 Estimates Based on IEUBK Model-Generated PbB Distribution
(Section 5.1.2 of 1403 risk analysis report)
0
1
2
3
5
8
10
50.4
39.3
30.8
24.4
15.7
8.58
5.96
19.9
16.0
13.0
10.6
7.27
4.31
3.13
8.95
7.42
6.19
5.20
3.73
2.35
1.76
1.40
1.15
0.921
0.738
0..483
0.273
0.194
1.35
1.35
1.33
1.28
1.15
0.964
0.854
Pre-1403 Estimates Based on Empirical Model-Generated PbB Distribution
(Section 5.1.2 of §403 risk analysis report)
0
1
2
3
5
8
10
34.5
20.4
12.0
7.14
2.61
0.652
0.278
4.53
2.76
1.71
1.07
0.443
0.130
0.0613
0.718
0.464
0.330
0.202
0.0926
0.0312
0.0158
0.932
0.675
0.442
0.271
0.0972
0.0224
0.00912
0.538
0.537
0.514
0.453
0.309
0.162
0.108
' A 0.257 IQ decrement is assumed for each 1.0//g/dL increase in PbB above the assumed threshold (see Section 4.4.1 of
the §403 risk analysis report). Thus, the fallowing hold:
• P[IQ i 11 « P[PbB 2 (threshold + 3.9 jig/dL]
• P[IQ i 2] = PIPbB i (threshold + 7.
• PHQ * 31 = RPbB * (threshold +11.
2 Average and standard deviation of tQ decrement are calculated assuming no IQ decrement occurs below the assumed
threshold, and a 0.257 IQ decrement is assumed for each 1.0 (igldL increase in PbB above the threshold.
-------
Similar patterns of decline were seen for the pre-§403 estimates generated under the
EUBK and empirical model-based blood-lead distributions, with the empirical model predicting
greater reductions for the larger thresholds. These model-based estimates were used in the
procedure to characterize changes from baseline that occur in a post-§403 environment, which is
addressed in Chapter 6.
226
-------
6.0 ANALYSIS OF EXAMPLE OPTIONS FOR THE §403 STANDARDS
Chapter 6 of the §403 risk analysis presented the methodology used to characterize
reductions to childhood health effect and blood-lead concentration endpoints expected to result
after interventions are conducted in response to the proposed §403 rule and applied this
methodology to a broad range of example options for standards. Assumptions were made on
post-intervention environmental-lead levels, which were applied to those HUD National Survey
housing units where a particular intervention was triggered as a result of having environmental-
lead levels that exceeded an example standard. Then, the EEUBK and empirical models were
used to generate the post-§403 blood-lead concentration distribution given post-§403
environmental-lead levels. These results, combined with similar model-based estimates in the
pre-§403 environment presented in Chapter 5, were used to obtain a final post-§403 blood-lead
distribution which was comparable to the baseline distribution generated by data from Phase 2 of
NHANES m. This procedure was detailed in Chapter 6 and Appendix Fl of the §403 risk
analysis report. This was the distribution upon which the health effects and blood-lead
concentration endpoints were estimated in the post-§403 environment.
The risk management procedure in Chapter 6 of the §403 risk analysis report considered
example standards for the following risk assessment measures:
• Average floor dust-lead loading
• Average window sill dust-lead loading
• Average soil-lead concentration
• Amount of deteriorated lead-based paint requiring paint maintenance
• Amount of deteriorated lead-based paint requiring paint abatement
Note that the lead-based paint standards considered in the risk management procedure differed
somewhat from the standards proposed in the §403 rule (see Chapter 1 of this report), as the rule
considered only a single tier rather than a two-tiered standard.
Section 6.1 presents additional detail and results on the performance characteristics
analyses, a non-modeling data analysis procedure used by EPA to help establish levels of concern
within the §403 rule. Performance characteristics analyses cited in the §403 proposed rule are
detailed, and additional performance characteristics analyses performed after the proposed rule to
address public comments and to finalize the rule are presented.
Section 6.2 investigates the incidence of children with elevated blood-lead concentrations
in homes where no candidate standard is met or exceeded (i.e., children who would be "missed"
by a specified set of candidate standards).
Since the §403 risk analysis report was published, public comment resulted in an
additional investigation into the assumptions made in the risk management on average dust-lead
loading following an intervention involving dust cleaning (40 ug/ft2 on floors, 100 Mg/ft2 on
227
-------
window sills). The results of this investigation are presented in Section 6.3. Based on this
investigation, the impact of alternative assumptions on post-intervention dust-lead loadings on
characterizing the reduction hi risk as a result of implementing §403 rules was evaluated through
a sensitivity analysis presented in Section 6.4. Also included in Section 6.4 are sensitivity
analyses applied to baseline (pre-§403) data within Section 5.1 of this report to evaluate the
impact of potential changes to the HUD National Survey data and assumptions on non-zero
thresholds for the IQ/blood-lead relationship, where the analyses are implemented on data
representing the post-§403 environment.
6.1 PERFORMANCE CHARACTERISTICS ANALYSES
The procedures defined and discussed in the §403 risk analysis report used statistical
modeling techniques to characterize risks of lead exposure to children in the nation's housing
stock and how these risks may be reduced as a result of interventions performed to reduce lead-
based paint hazards in the housing stock under the §403 rule. While using the findings of this
risk analysis to evaluate options for the standards specified in the §403 rule, EPA also wished to
base its evaluation partially on a non-modeling approach using data from field studies that
measured lead levels in both children's blood and in the same environmental media targeted by
the §403 rule. In particular, given the data reported in these studies, EPA was interested in
observing how often a specified set of candidate standards would "trigger" interventions hi
housing units within these studies and the extent to which these units contained a child with an
elevated blood-lead concentration (s 10 ng/dL). Such an investigation provided useful
information on the performance of a specified set of candidate standards without some of the
complexities associated with making conclusions from statistical modeling analyses.
EPA employed performance characteristics analysis, sometimes referred to as
sensitivity/specificity analysis, as a non-modeling approach to evaluating candidate §403
standards. The underlying statistical principle of this approach involves conditional probabilities
and has been documented hi references such as Fleiss (1981, Section 1.2). This chapter presents
the findings of performance characteristics analyses applied to data from the Rochester Lead-in-
Dust study. Applying data from this study was highly appropriate under the objective to evaluate
candidate lead standards in the §403 rulemaking. The form of the study data used in this analysis
is discussed hi detail within Section 6.1.1. The methods used to perform this performance
characteristics analysis are presented in Section 6.1.2. Section 6.1.3 presents the results of
performance characteristics analysis presented in the preamble and which were used in the §403
rulemaking. Finally, Section 6.1.4 presents additional performance characteristics analyses
performed after the §403 proposed rule was published, where these analyses considered other
sets of standards (including the standards specified in the §403 proposed rule) and other means of
handling data on amount of deteriorated paint within a household.
6.1.1 Data Used in The Performance Characteristics Analysis
The performance characteristics analysis was applied to data from the recently-conducted
Rochester Lead-in-Dust study. A summary of objectives and design information for this study is
228
-------
found in Section 3.2.2.2 of the §403 risk analysis report. The Rochester study data were selected
for this analysis for the following reasons:
• The study reported information for all media for which §403 standards were
proposed (e.g., dust-lead on floors and window sills, soil-lead, condition of lead-
based paint).
• The study measured blood-lead concentration in 205 children aged 12-31 months
who resided in the selected homes.
* The dust sampling methods used in this study included the wipe technique, from
which dust-lead loadings were measured.
• For some homes, soil was sampled from multiple locations (i.e., dripline and play
areas), allowing for yardwide average soil-lead concentration to be estimated.
• While homes and children were targeted for selection in this study, the selection
process was more random and more representative of a general population than is
the case with other lead exposure studies.
The primary concern with using data from the Rochester study in this analysis is the degree to
which the study may be considered representative of the nation as a whole. The study selected a
targeted sample which was limited to a single geographic area. The sample consisted of children
who had moderate exposure to lead in their home environment and did not necessarily include
children with very high or very low exposure to lead. In particular,
• 22.9% of the children in this study (47 children total) had blood-lead
concentrations at or above 10 ng/dL, compared to the national estimate of 5.9%
for children aged 1-2 years according to Phase 2 of the Third National Health and
Nutrition Examination Survey (NHANES HI) (CDC, 1997).
• The geometric mean blood-lead concentration for the study children was 6.38
|ng/dL with a geometric standard deviation (GSD) of 1.85. This compares with a
geometric mean of 3.1 (ig/dL and GSD of 2.09 estimated for U.S. children aged 1-
2 years according to Phase 2 of NHANES m (CDC, 1997).
• At least 84% of the housing units included in this study were built prior to 1940,
compared to the estimated 20% of the entire U.S. housing stock made within the
§403 risk analysis (Table 3-5 of USEPA, 1998). There is a well-documented
relationship between age of housing and presence of lead-based paint hazards.
• While geometric mean floor dust-lead loadings were comparable between the
Rochester study and the HUD National Survey (after converting the data to wipe-
equivalent loadings), whose results are considered nationally-representative,
229
-------
geometric mean estimates of window sill dust-lead loading and soil-lead
concentration were higher for the Rochester study relative to HUD National
Survey estimates (Section 3.2).
Despite these limitations, the Rochester study is considered one of the best resources of data for
characterizing the relationship between children's blood-lead concentration and residential
environmental-lead levels, and therefore, for evaluating national standards for lead in the nation's
housing stock.
While data were available for 205 units in the Rochester study, somewhat fewer of these
units had values for all required data endpoints for this analysis. In particular, 177 units had data
reported on the amount of deteriorated lead-based paint, plus lead measurements for floor dust
(wipe), window sill dust (wipe), and soil (dripline and/or play area). Of these units, 77 had soil-
lead data for both dripline and play areas, thereby allowing an average concentration across these
two areas to be calculated.
For the analysis presented in the §403 proposed rule, the following five data endpoints
were calculated for each Rochester study housing unit:
• Area-weighted average uncarpeted floor wipe dust-lead loading (i.e., the measured
loading for each sample was weighted by the area of the sample when averaged)
• Area-weighted average window sill wipe dust-lead loading
• Average of dripline and play area soil-lead concentrations
• The percentage of interior painted components tested in the study that contained
lead-based paint (measurements at or above 1.0 mg/cm2) and some level of
deterioration (paint condition listed as fair or poor)
• The percentage of exterior painted components tested in the study that contained
lead-based paint and some level of deterioration.
Note that these endpoints are comparable to the standards included in the §403 proposed rule,
with the exception of the latter two paint-lead measurements. While the proposed §403 standard
for the paint component is expressed as a square footage of deteriorated lead-based paint for
components with large surface areas (2 ft2 for interior surfaces, 10 ft2 for exterior surfaces) or as
the percentage of total painted surface area that is deteriorated for components with small surface
areas (10%), no indication on the amount of deteriorated lead-based paint on a given component
(either in square feet or as a percentage of the total surface area) was recorded in the Rochester
study. Instead, each paint-lead measurement was associated with an indicator of the paint's
condition (good, fair, poor). Therefore, for this analysis, the amount of deteriorated lead-based
paint in a housing unit was taken to be the percentage of tested components in the housing unit
that contained lead-based paint along with some level of deterioration (i.e., condition of paint
230
-------
either fair or poor). This result was assumed to he a good estimate of the total amount of lead-
based paint in the unit that was deteriorated.
6.1.2 Analysis Approach
The performance characteristics analysis classified each housing unit in the Rochester
study according to two different criteria:
1. Whether or not the unit exceeded anv of the candidate standards for the various
media being controlled.
2. Whether or not the unit contained a child with elevated blood-lead concentration
(alOug/dL).
The first criterion represented whether a housing unit was "triggered" for any intervention by
exceeding at least one candidate standard, while the second represented whether the unit
contained a child requiring attention as a result of having an elevated blood-lead concentration.
The first criterion was determined by noting whether the value for at least one of the five
endpoints mentioned at the end of the previous section exceeded the standard associated with the
type of measurement represented by that endpoint.
For a given set of candidate standards, the set of housing units in the Rochester study was
identified that had data for all of the above five endpoints. These units were classified according
to whether or not they achieved the above two criteria. These results are summarized in the
manner illustrated within the 2x2 frequency table in Table 6-1. From this information, the four
performance characteristics defined in Table 6-1 were then calculated: sensitivity, specificity,
positive predictive value (PPV), and negative predictive value (NPV). These characteristics
provide the necessary information for evaluating the sets of standards on their ability to target the
proper set of units for intervention.
In this analysis, a "false positive" corresponds to triggering a housing unit for intervention
when it does not contain a child with an elevated blood-lead concentration, while a "false
negative" corresponds to not triggering a housing unit containing a child with an elevated blood-
lead concentration. Note that the proportion of false positives is equal to one minus the
specificity, while the proportion of false negatives is equal to one minus the sensitivity.
While information from all four performance characteristics are important for evaluating
the performance of a given set of standards, typically one or two characteristics are given more
weight than the others in the performance evaluation process. For example, in the preamble,
EPA evaluated candidate standards for dust-lead loading on uncarpeted floors and window sills
according to whether the performance characteristics analysis yielded a value of NPV from 95 to
99 percent under the given set of standards. This implied that no more than 5% of children living
in housing units with environmental-lead levels below the standards would have elevated blood-
lead concentrations (i.e., at or above 10 ug/dL). More recent Agency inquiries have focused on
231
-------
Table 6-1. Definitions of Performance Characteristics Used to Evaluate How Various
Combinations of Environmental-Lead Standards Classify Housing Units in the
Rochester Lead-in-Dust Study
Blood-Lead Concentration
At or Above 10/ig/dL?
In the above table, the lette
or above 10^/g/dL who live
specified standards. Letters
equals a + b+c+d. From th
Performance
Characteristic
Sensitivity
(or True Positive Rate, or
1 - False Negative Rate)
Specificity
(or True Negative Rate, or
1 - False Positive Rate)
Positive Predictive Value
(PPV)
Negative Predictive Value
(NPV)
Yes
No
Any of the Standards Exceeded?
No
a
c
Yes
b
d
r 'b' represents the number of children which have a blood-lead concentration at
in a residence with environmental-lead levels that exceed at least one of the
> 'a', 'c', and 'd' represent similar counts. The total number of housing units
sse counts, the following performance characteristics are calculated:
Definition ;
Probability of a housing unit exceeding at least one
standard given that there is a resident child with an
elevated blood concentration (2 1 0 //g/dU
Probability of a housing unit not exceeding at least
one standard given that a resident child has a low
blood-lead concentration (< 10//g/dL).
Probability of a resident child having an elevated
blood-lead concentration <210 j/g/dL) given that the
housing unit exceeds at least one standard.
Probability of a resident child having a low blood-
lead concentration « 10 //g/dL) given that the
housing unit does not exceed at least one standard.
Calculation
b/fa + b)
c/(c+d>
b/
-------
so
40
£ 30
|
c
-3
"g 20
o
s
10
• •. x*
\ . f .. •
400 BOO 1.200 1.600
Dripline Soil-Lead Concentration (^g/g)
2.000
2,400
Figure 6-1. Example of an Ideal Situation for Establishing Potential Dripline Soil-
Lead Standards
233
-------
50
40
30
v
o
o
u
• 30
•o
o
i
10
S . f
• 1
400 800 1.200 1.600
Dripiine Soil-Lead Concentration
2.000
2,400
Figure 6-2. Example of a Situation Where the Negative Predictive Value and
Sensitivity Equal 100%, but the Positive Predictive Value and Specificity
are Less than 100%
The performance characteristics analysis was repeated for different sets of standards. For
each analysis, the information within Table 6-1 was calculated, and those sets of standards that
maximized the desired performance criteria were identified.
The different analyses presented in the subsequent sections of this chapter were
performed on different subsets of housing units in the Rochester study. The results are purely
descriptive in that they represent combinations of candidate standards that meet the specified
performance criteria when considering the housing units in the Rochester study and are not based
on any underlying probability model. Different results are possible if this analysis were to be
applied to data from different studies. In addition, only point estimates of the performance
characteristics are presented. The uncertainty in these estimates is primarily dependent on
sample size, and to a lesser degree on measurement error.
234
-------
6.1.3 Results Cited in the §403 Proposed Rule
The analysis presented in this section were cited in section B.l.d of Part IV of the
preamble. This section of the preamble contained a brief presentation of the information
presented in Section 6.1.2 above, then cited findings of analyses documented in a memorandum
dated 9/3/97 from Battelle (Ronald Menton and Warren Strauss) to EPA (Todd Holderman).
EPA requested that Battelle perform this analysis in an action item of a meeting between Battelle
and EPA on August 27,1997. A copy of the cited memorandum is found in Appendix G.
The analyses presented in Appendix G were performed on data for the 77 housing units in
the Rochester study that had all necessary data for the analysis, including soil-lead concentrations
for both dripline and play areas. As the §403 proposed rule was to contain a yardwide average
soil-lead standard, it was desired to consider only those housing units that had soil-lead data for
both locations. The considerable reduction in the number of Rochester study housing units
whose data were considered in this analysis (from 205 to 77 units) was due primarily to the fact
that play-area soil-lead concentration was measured for less than half of the study units.
The combinations of candidate standards considered in this analysis were those requested
by EPA at the time, when EPA was actively considering candidate standards in the rulemaking.
These combinations included all 8x4x9x3=864 combinations of the following:
• uncaroeted floor dust-lead loading: 50,75, 100,125,150,175,200,400 ug/ft2
• window sill dust-lead loading: 100,300,500,800 ug/ft2
average soil-lead concentration: 200,300,400,500,600,700,900,1000,1500
• maximum of percent of interior/exterior painted surfaces with deteriorated lead-
based paint: 5,10,20%
Note that the type of endpoint that represented the paint-lead measurement in this analysis (i.e.,
the last bullet) differed from the type of paint-lead standard that EPA ultimately proposed in the
§403 proposed rule.
The purpose of this analysis was to identify those sets of candidate standards (from the
864 combinations above) which, when applying the performance characteristics analysis under
those sets of standards, resulted in values of negative predictive value (as defined in Table 6-1
above) that met one of the following three criteria:
NPV * 99%
95%*NPV<99%
90% s NPV < 95%.
The findings of this analysis are documented in Tables 1 and 2 of Appendix G.
235
-------
Twenty-one of the 77 housing units whose data were included in the analysis did not
exceed any of the candidate standards in at least one of the 864 combinations of candidate
standards. These housing units, with the values of the endpoints used to compare to the
candidate standards and children's blood-lead concentration, are listed in Table 6-2. This means
that the denominator of NPV (i.e., the number of housing units that do not exceed at least one of
the candidate standards being considered) never exceeded 21 across the 864 combinations. For
some combinations, the denominator was as small as 2. Furthermore, all but two of the 21 units
in Table 6-2 contained children with blood-lead concentrations below 10 ug/dL. As a result, the
value of NPV was no lower than 84.6% across all 864 combinations of candidate standards. At
least one of the above three criteria for NPV was met for 808 (93.5%) of the combinations. Of
these 808 combinations, NPV equaled 100% for 690 of the combinations, equaled 95% for seven
combinations, and was at least 90% but below 95% for the remaining 111 combinations.
All of the remaining 56 housing units in the analysis that are not represented in Table 6-2
exceeded either the soil-lead standard or one of the two paint standards (i.e., interior and/or
exterior) in each of the 864 combinations of candidate standards. That is, each of these houses
had at least one of the following:
• average soil-lead concentration of at least 1500 ug/g
• at least 20% of painted surfaces with deteriorated lead-based paint in the interior
and/or exterior.
Therefore, the 56 housing units not represented hi Table 6-2 were triggered in each of the 864
combinations of candidate standards, without regard to the floor or window sill standards.
The results presented in this section led to the following conclusions stated in Part IV of
the preamble:
"For uncarpeted floors, dust-lead loadings ranged from 50 ug/f? to 400 pg/f?
depending on the dust-lead loading on interior window sills and the soil-lead
concentration. For interior window sills, dust-lead loadings ranged from 100
fig/ft* to 800 fig/ft2 depending on the dust-lead loading on uncarpeted floors and
the soil-lead concentration. These ranges are significantly higher than the ranges
yielded by the multimedia approach."
"Soil-lead concentrations ranged from 200 ppm to l.SOOppm depending on dust-
lead loadings on uncarpeted floors and interior window sills and the exceedance
probability."
The ranges cited in the preamble were precisely the lower and upper ranges of the candidate
standards considered in this analysis. These findings reflect the very high values of the NPV
across the combinations of standards considered in this analysis.
236
-------
Table 6-2. Set of 21 Housing Units in the Rochester Study in Which No Standard Was
Exceeded in at Least One of the 864 Combinations of Candidate Standards
Housing ID -
00034
00132
00302
00637
00874
00974
01047
01062
01195
01228
01930
01971
01991
02290
02411
02837
03174
03360
03527
05343
05498
*.; , / Statistics Compared to, _the Candidate Standards1,, £,."-" <
Roor Dust-
Lead Loading
to/ft2}
63.60
17.30
2.55
59.00
12.90
14.90
20.83
12.40
12.25
3.37
19.35
5.10
15.50
2.65
10.48
4.29
18.60
12.43
6.08
10.30
19.15
Window Sfll
Dust-Lead
Loading -
Uig/ft2)
349.9
90.6
70.7
74.9
293.7
45.6
372.3
87.1
32.2
16.2
118.8
41.9
398.9
74.1
178.5
2.8
235.6
702.0
148.8
75.7
66.0
": rtt'j.
Average SoB-
Lead Cone.
U/g/gJ -
438.5
268.0
124.5
950.0
102.9
51.1
574.3
447.5
830.5
419.0
773.4
506.0
104.0
465.0
828.5
458.5
625.5
912.0
539.5
552.0
1150.5
% of Interior
Components
with
Deteriorated
LBP
17
0
0
0
0
0
18
0
0
0
0
11
0
11
10
0
0
0
14
13
0
% of Exterior
Components
with
Deteriorated ,
LBP
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Blood-Lead
Cone. U/g/dL)
7.1
6.0
4.8
13.3
2.1
8.9
3.9
7.4
6.9
4.6
4.6
6.1
7.5
4.9
9.0
8.9
5.8
11.3
4.5
5.6
5.8
1 See Section 6.1.1 for the definitions of these statistics.
237
-------
6.1.4 Results of Analysis on Specified Sets of Standards
The analyses presented in Section 6.1.3 were performed prior to release of the §403
proposed rule and contributed to the information presented in the preamble. Since the proposed
rule was released, EPA has requested additional performance characteristics analyses be
performed on various combinations of candidate standards, to address various issues raised
within the public comments to the proposed rule and in support of preparing the final §403 rule.
This section presents the results of these additional performance characteristics analyses.
Additional performance characteristics analysis results are presented in Appendix J.
As discussed hi Section 6.1.3, one of the limitations of the analyses presented in the
preamble was the relatively small number of housing units (77) in the Rochester study whose
data were used in the analyses. This small number was primarily due to the lack of available
soil-lead concentrations from play areas and the desire to have soil-lead data for both dripline and
play areas in order to calculate a yardwide average. Thus, the additional analyses presented in
this section re-defined how the soil-lead measure was calculated (with different approaches taken
to this re-definition), thereby increasing the number of units whose data could be included in the
analysis.
6.1.4.1 Analyses Performed on 41 Combinations of Candidate Standards, in Three
Iterations. The candidate standards that were considered in this analysis were the following:
• uncarpeted floor dust-lead loading: 5,10,20,25,40,50,100,200 jig/ft2
* window sill dust-lead loading: 250 ug/fi2
• vardwide average soil-lead concentration: 400, 1200, 2000, 5000 (ig/g
• amount of deteriorated lead-based paint: 2% of interior painted surfaces or 10% of
exterior painted surfaces.
Thus, different candidate standards for floor dust-lead loading and soil-lead concentration were
considered, while only a single candidate standard was considered for window sills (i.e., that
specified in the §403 proposed rule) and deteriorated lead-based paint. This analysis considered
a total of 41 combinations of candidate standards, corresponding to the 8x4=32 combinations of
the above candidates, as well as the additional 9 combinations:
• only the paint standards (1 additional combination)
• only the paint and soil-lead concentration standards (4 additional combinations)
• only the paint, soil-lead concentration, and window sill dust-lead loading
standards (4 additional combinations).
For each combination, the four performance characteristics were calculated and presented, as
well as the number of housing units that exceed at least one of the specified standards.
Note that the above candidate paint standard (percentage of paint that is deteriorated lead-
based paint) is not expressed in the manner that the proposed paint standard in the §403 proposed
238
-------
rule was expressed (amount of deteriorated lead-based paint, in square feet). As discussed in
Section 6.1.1 above, the Rochester study measured only lead content in paint plus an indicator of
paint condition, and therefore, did not measure the surface area containing deteriorated lead-
based paint. For the Rochester study data, the above paint standard triggered all units with
deteriorated lead-based paint present, as the lowest observed non-zero percentage of deteriorated
lead-based paint was 8% for interior surfaces and 14% for exterior surfaces.
Three iterations of this analysis was performed, with each iteration involving data for a
different number of housing units:
Iteration #1: Instead of requiring soil-lead concentrations be reported for both dripline
and play areas, as was done within the analysis cited in Section 6.1.3 above, average soil-
lead concentration was set equal to the reported concentration at one of these areas if no
concentration is reported for the other area. This approach permitted data for 177 housing
units to be used in the analysis.
Iteration #2: After taking the approach in iteration #1, any units that did not have soil-
lead concentration reported due to having no bare soil available from which to sample
were assigned a soil-lead concentration of 0 ppm. This approach was taken as Title IV of
TSCA restricts the §403 soil-lead hazard standard to bare soil and further assuming that
any covered soil at these units would not pose a soil-lead hazard. This approach
permitted data for 184 housing units to be used in the analysis.
Iteration #3: After taking the approach in iterations #1 and #2, the 21 remaining units
having missing data for at least one endpoint had an imputed value assigned to the
endpoint(s) equal to the average value across units within the same year-built category
(pre-1940,1940-1959,1960-1979, post-1979) and having the same indicator of whether
or not lead-based paint is present in the unit. This method followed the same approach
taken in the §403 risk analysis (Section 3.3.1.1 of the §403 risk analysis report) to impute
data for housing units in the HUD National Survey. This approach permitted data for 205
housing units to be used in the analysis.
The results of each iteration are now presented.
Iteration #1: Data for 177 Housing Units.
Table 6-3 presents the results of the performance characteristics analyses performed on
data for 177 housing units (#1 above) under the 41 combinations of standards listed above. Note
from this table that the fixed paint standards (which were equivalent to finding any deteriorated
lead-based paint in the unit) triggered an intervention for nearly three-fourths of the 177 units.
These paint standards considered jointly with a soil-lead concentration standard of 400 ug/g
resulted in 100% sensitivity and negative predictive value regardless of the dust standards.
Sensitivity and negative predictive values of 100% were also met at a soil-lead concentration
standard of 1200 ug/g if the floor dust-lead loading standard was at 10 ug/ft2 and the window sill
239
-------
a
£
•§
e
§
«
c
o
€
£
0)
8i>
111
§0
1
I
-5
%
e
'
l
co
CO
CM
CM
10
S
CM
co"
S
CM
CO
CM
CO
CM
CM
CO
CM
CM
CO
*
CM
CM
co
*
CM
10
CM
CO
CO
CO
CM
CO
CM
CM
CD
CO
CM
I*.
CO
CM
CO
CM
CO
CM
00
IO
CO
CO
CM
m
CM
N
CM
w
CM
O
o
N
CM
CM
CM
8
CM
CM
8
CM
8
CO
00
O>
3>
CO
CO
S
CO
CO
CO
o>
CD
CO
CO
CO
CO
CO
JO
CO
CO
CO
r*
N
CM
CO
m
CO
CM
Ifl
CO
CO
CO
CO
CO
CO
at
00
s
CO
00
CO
10
00
CO
10
CO
00
CO
10
CO
CO
CO
10
CO
00
CO
10
CO
oo
CO
10
00
CO
oo
CO
to
0>
eo
co
PS
o>
CO
CO
CO
CM
§
10
CM
10
o
CM
ID
CM
o
10
CM
o
10
CM
o
10
CM
o
10
CM
o
10
CM
to
CM
IO
CM
s
CM
o
10
CM
10
CM
to
CM
O
O
§
o
o
CM
CM
i
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
240
-------
co
q
CM
CM
CM
*t
CM
<0
CM
CO
CO
CD
CM
CM
CM
CO
CM
CM
CO
CM
CM
CO
CM
CM
CO
CM
CM
to
CO
CO
CM
IO
CD
CO
CM
CO
CM
to
CO
CM
••
n
ft
o>
IO
co
to
CO
m
CM
PO
CD
s
8
3>
CM
*
O
CM
CM
co
§
CM
CO
CM
CM
S
CM
00
CM
CM
CO
S
CO
CM
IO
CD
IO
CM
CO
SPECI
CO
CO
0>
co
to
tM
CM
CO
CD
CO
N
CM
CO
co
CN
CD
CM
O
CM
IO
CO
o
u
m
(6
(9
2
a
3
0 -i •£
CO
IO
o>
co
10
0>
CO
IO
ca
en
CM
CO
8
CO
«t
co
oo
oo
to
en
0)
co
o
o>
en
co
O
co
cn
co
g
CO
en
CO
O)
o
fs
CD
CM
CO
CO
CM
CO
CO
-------
Aiioidioads
|
n
to
a
Set of Standard
to
CO
CO
CM
CM
00
CO
ID
CM
CO
CO
8
IO
CM
CO
to
CM
CO
S
CO
8
10
CM
CM
CO
10
CM
CO
co
8
8
CM
to
en
00
CO
en
8
CM
CM
en
CM
CM
CM
CM
en
CM
00
CO
CO
CO
00
to
CM
CM
en
CO
CO
CM
CD
8
CO
en
IO
CM
CO
10
CO
IO
8
CM
tsl
~ «
«•§{;
CD
C
O
I
CO
s
o
1
5 81
m
ui
5
£
•
§
CD
f
S
5
•o
-------
dust-lead loading was at 250 ug/ft2, although the specificity declined considerably at these
standards. When the floor dust-lead loading standard was raised to 20 ug/ft2 in this situation,
both the sensitivity and negative predictive value remained above 95%. However, at soil-lead
standards of 1200 ng/g or higher, the 95% criterion for both sensitivity and negative predictive
value were no longer achieved once the floor dust-lead loading standard exceeded 20 ug/ft2.
Among the 32 combinations of standards included in Table 6-3, the sum of the four
performance characteristics (i.e., the last column of the table) was maximized at 244.6% at a
floor dust-lead loading standard of 20 ng/ft2 and a soil-lead standard of either 1200 or 2000 ug/g.
(The paint and window sill standards were fixed in each combination.)
The proposed §403 standards, assuming the different approach taken in this analysis to
interpreting the paint standards, resulted in a 93% sensitivity (40 of 43 units containing an
elevated blood-lead child are triggered) and nearly a 92% negative predictive value (see
shaded/bold row within Table 6-3). The sum of the four performance characteristics was
237.7%. Nearly 80% of the 177 units exceeded at least one of the proposed §403 standards.
Iteration #2: Data for 184 Housing Units.
Table 6-4 presents the same types of results as in Table 6-3, but it reflects analyses that
included data for seven additional housing units where soil-lead concentration was assumed to be
0 ug/g due to having no bare soil present for sampling (i.e., a total of 184 housing units). Only
one of these seven additional units contained a child with an elevated blood-lead concentration.
Slight reductions in the values of the performance characteristics were seen from Table
6-3 to Table 6-4 with the addition of these seven units. The one additional unit containing a
child with elevated blood-lead concentration did not exceed any of the paint, soil, or window sill
standards in the table and exceeded only floor dust-lead loading standards below 50 jig/ft2.
However, as in Table 2-3, sensitivity and negative predictive values of 100% (and the
considerable declines in specificity) continued to occur at a soil-lead concentration standard of
1200 ug/g if the floor dust-lead loading standard was at 10 ug/ft2 and the window sill dust-lead
loading was at 250 ug/ft2.
Despite the general declines in the values of the four performance characteristics from
Table 6-3, the largest observed value of the sum of these characteristics among the 32
combinations of standards (245.0) was slightly larger than in Table 6-3. This value was observed
for the same two combinations of standards for which the maximum occurred in Table 6-3: a
floor dust-lead loading standard of 20 ug/ft2 and a soil-lead standard of either 1200 or 2000 ug/g.
The proposed §403 standards, assuming the different approach taken in this analysis to
interpreting the paint standards, resulted in nearly a 91% sensitivity and nearly a 90% negative
predictive value, which were slight declines from Table 6-3 (see shaded/bold row within Table
6-4). The sum of the four performance characteristics was 233.2%.
243
-------
§
o
•o
ID
o
3
«
8
cc
o
ill
«•
00
w
w
:*.£i;^I-fe;
1
I
•5
s
O
CO
CO
oo
o
(O
CM
§
CO
CM
<0
O
o
IO
ao
(O
CM
CM
8
ro
CM
03
co
o
o
Tt
IO
oo
CO
CM
§
00
*t
03
CM
ex
CM
o
o
*
O>
o
o
*t
t
en
CO
CM
O
O
to
CM
O)
o
o
O)
IO
CM
CO I
ro
CM
CM
ro
co
CM
CM
CO
ro
CM
CM
CO
ro
CM
CO
CO
CO
00
CO
ro
S
n
co
CO
CO
S2
IO
ro
co
00
CO
S2
IO
ro
(O
CM
CM
CM
cq
r^
CM
10
10
to
en
IO
CM
IO
CO
IO
CM
IO
ro
#
q
id
CM
IO
ro
co
g
CO
d>
i
00
oo
oo
-------
a» m
H*
CO
CO
CD
CM
CQ
CM
10
*
CM
0>
id
CO
CM
cn
cn
CM
CM
CM
00
CM
CM
CM
00
CN
CN
CM
00
CM
CM
CM
CO
CM
CN
CN
:;.«'
$•
CO
CO
CM
CO
rx
CO
CM
IO
*
CM
cn
id
CO
CM
cn
cn
CM
CM
en
oo
CO
in
CO
o»
CO
CO
m
o>
CO
cn
CD
cn
CM
f
3
O
o
to
CO
IO
^
cn
CO
in
00
^
IO
CO
*
IO
on
IO
to
IO
00
o
to
CO
to
en
en
CO
CO
en
»—
O)
CO
en
CD
cn
CM
CO
#
8
fx^
IX
00
CM
CO
00
IO
O)
CN
in
in
cn
CO
8
o
o
CD
CO
00
eo
CO
00
00
en
CO
CD
CO
CO
CO
CO
CD
CO
go
cn
CO
.0}
S
CM
CO
o>
CN
rx
en
CO
8
CO
cn
to
O
eo
00
00
00
CD
CM
CO
CM
10
CM
CO
O
r*
CM
CD
cn
to
id
o>
m
CO
eo
CO
CD
CO
•
10
CO
CO
fx
en
en
cq
CN
00
CM
IO
CM
CO
CM
CD
cn
•s
•§
to
IO
CN
CM
to
o
o
CM
8
s
to
CM
CM
o
IO
(M
IO
CN
IO
CM
IO
CM
to
CM
O
to
CM
10
CM
to
CM
CM
g
CM
to
CM
S
CM
g
CM
IO
CM
to
CM
i
O
O
CN
CM
CM
O
O
CM
8
CM
8
CM
8
CM
CM
CM
CM
O
CM
N
CM
CM
CM
CM
CM
CM
CM
CN
CN
CN
CM
CM
CM
245
-------
i
. UI
ut >
> p ui
»5<
5*
*s*
c «"
"
= -3 §
S J-
00
01
CM
CO
m
to
CO
IO
OI
CD
OJ
00
04
01
CO
CN
01
OI
00
01
01
8
10
OI
00
CO
OI
CO
CO
CD
CO
IO
CM
8
OI
CD
OI
m
IO
01
CD
OI
IO
a>
IO
in
01
CO
g
CO
CO
10
IO
01
01
w-
8
IO
00
o>
OI
w -
8
o>
i
in
OI
246
Js?
•S s «
«•*
» c *
€.2 o
w U»
§11
S - o .2 o
0 S o ** ~K
•& at C <0 -°
S
3
C
o
S "a
il
E (D
O
U
-------
Iteration #3: Data for 205 Housing Units.
The third set of performance characteristics analyses was performed on data for all 205
housing units in the Rochester study. The previous analyses involved data for fewer housing
units as some units did not have recorded data for the key endpoints used in the analyses to
compare to the various candidate standards. Therefore, this analysis replaced incidences of
missing data with data values that were imputed from information available from other study
units. It was assumed that these imputed values were accurate estimates of what would have
been reported for these units. This estimate for a housing unit could vary considerably from what
would have been reported, however, based on actual conditions and behaviors in the household.
As all 205 housing units had reported values for child's blood-lead concentration and for
the percentage of tested ulterior components containing deteriorated lead-based paint, no
imputation was necessary for these two endpoints. The other four endpoints had at least one
housing unit with missing data. For each of these four endpoints, Table 6-5 contains the number
of housing units with missing data according to year-built category and whether or not the unit
contains lead-based paint, along with the imputed data value assigned to these units, which
equaled the average value across all units in that same category that had non-missing data. The
imputed data values depended on the year-built category and lead-based paint indicator as these
two variables are typically important predictors of these values. This same approach was used in
the §403 risk analysis to impute environmental-lead data values for HUD National Survey units
having missing data (see Section 3.3.1.1 of USEPA, 1998).
The data imputation process documented in Table 6-5 resulted in assigning imputed data
to 21 units: 19 built prior to 1940, one built from 1940-1959, and one built after 1979. A total of
eight average uncarpeted floor dust-lead loadings, nine average window sill dust-lead loadings,
six average soil-lead concentrations, and one percentage of deteriorated lead-based paint on
exterior surfaces were imputed.
Table 6-6 presents estimates of the four performance characteristics for the 41
combinations of standards, using reported and imputed data for 205 housing units in the
Rochester study. These estimates are very similar to those in Table 6-4 that were calculated from
data for 184 housing units. The same conclusions can be drawn from these results as were made
from the results in Tables 6-3 and 6-4. This implies that at the given combinations of candidate
standards considered in these analyses, the methods used in this section to estimate performance
characteristics were relatively robust across the different sets of data used in the analyses (i.e.,
177,184, or 205 units).
As sensitivity and negative predictive value are the two performance characteristics of
most interest to Agency reviewers, the results for these two characteristics from Tables 6-3,6-4,
and 6-6 are summarized in Table 6-7. This summary emphasizes the relative stability of the
estimates across the different approaches used to make the calculations.
247
-------
Table 6-5. Numbers of Housing Units with Missing Data for Four Endpoints and the
Imputed Data Values Assigned to These Units in This Analysis
Year-Built
Category
Pre-1940
1940-1959
1960-1979
Post-1979
Lead-
, Based
Paint
Present?
-" "* K & \*\ "
Yes
No
Yes
No
-
Yes
No
Area-Weighted
Average Uncarpeted
- Ftoor Dust-Lead
Loading •
# Units
with
Missing
Data
6
1
0
O
0
1
0
Imputed
Value • >
tow/ft2),1;
160.2
(157)
13.3
(8)
-
-
-
91.3
(3)
-
Area-Weighted
" -Average Window
; Sill Dust-Lead
: Loading ,
# Units
with
'Missing
'Data
5
3
1
0
0
0
0
Imputed
Value
(re/ft2)?
633.2
(158)
95.2
(6)
569.0
(12)
-
-
—
-
Average Soil-Lead
Concentration
# Units
with
Missing
Date ,
5
1
0
0
0
0
0
Imputed
Value
(migf
1258
(158)
631.7
(8)
—
-
-
-
-
% of Exterior
. ' Components
. ' Containing s-
Deteriorated Lead-
Based-Paint
# Units
with
, tiKmtSm*
Missing
Date
1
0
0
0
0
0
0
Imputed" <
Value'
{%)?
25.2%
(162)
—
—
-
-
—
-
1 Number in parentheses equals the number of values (i.e., housing units) entering into calculation of the imputed value.
which is the average of these values.
248
-------
t»
I
tn
'"£
Sol^«:
c «- S 53,
S o
•81$
CO
a
"I
.«
'JI
a>
00
CO
to
CD
CO
CM
CO
CD
CO
CN
CO
CD
CO
CM
CO
id
CO
CN
00
CN
CO
CN
IO
a>
to
CM
in
CM
CO
CM
CD
CM
CN
CO
CO
CN
CO
CO
CO
CN
CO
CO
CO
CN
CO
CO
CO
CM
CO
CO
00
IO
(ft
£
00
CM
CM
CM
CD
CD
CD
S!
IO
CM
CN
CD
CO
to
CN
CM
CD
CO
CO
CM
io
CM
IO
CN
CM
CN
CN
CN
CN
CN
CN
CN
CM
CD
CO
CO
00
CD
to
O>
CM
IO
CD
CM
IO
S
CM
CO
CO
00
CO
CO
CO
to
to
CO
10
tv
a>
w
CD
CM
CN
CD
CD
CO
CM
CD
CO
CM
00
^
CD
CM
00
CO
CN
CD
CN
CM
CO
CO
00
M
CO
CD
i
N
CO
IO
CO
N
CO
CD
2
CO
CO
CN
CO
CO
1
00
CN
to
CM
CM
CO
1ft
CM
CD
id
CN
en
to
CN
CO
id
CM
CO
id
CO
id
ct
CO
CM
CD
CM
oq
CO
CO
CO
CO
CO
00
00
CO
CO
CO
CO
CO
CO
00
00
00
CO
CO
CO
co
CO
en
CD
CO
CO
CO
co
to
to
CD
CO
CD
IO
CO
co
CD
IO
CO
CO
CD
o
o
CN
O
O
to
IO
CM
CM
CN
IA
CM
to
CM
to
CN
g
CN
CN
O
to
CN
to
CM
IO
CM
o
10
CM
10
CM
10
CN
to
CN
o
o
CM
CM
CN
O
O
CN
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CN
CN
CM
CN
249
-------
CM
in
CM
00
CM
in
en
CM
CM
CD
O
co
CM
00
CM
00
CM
CM
00
CM
CM
CM
'
CO
CO
CM
CO
CO
CM
CO
It
CM
CO
CM
in
en
CM
CM
2*
S>S
a>
en
oo
co
CD
co
I
00
§
CO
en
CO
CO
CO
#
8
^w*
in
in
*i*~
8
go
oo
co
oo
o
in
oo
92
co
CO
CO
CO
CO
CO
00
CO
00
CO
CO
CO
CM
s
CO
O)
CO
CO
CM
o>
en
CO
8
O
O
00
CO
CO
. ill
£• >
til
CN
in
(0
tn
in
CM
CO
CD
in
CM
co
i
in
CM
CO
in
CM
5
CM
en
CN_
in
in
CM
in
co
CM_
CM
CO
in
co
N
CM
CO
CO
in
CD
CM_
CM
2
CO
CO
u>
CM
CO
CD
n
in
ix1
CM
CO
in
CM
CM
|x
CD
CO
n*rl
** ia- ** . «
CO
CO
CO
CO
CM
CO
CO
en
CN
CN
CO
CO
N
CO
CO
CD
en
in
£
CO
CO
CM
CM
CM
CO
CO
CM
a
00
CO
CM
00
CO
:co
!**'
CO
CM
CM
CO
CO
en
CM
CN
CO
CO
q
CM
CO
CO
CO
en
in
in
00
I
*
CO
CD
O
JO
a
I
to
CO
CO
in
•*
oq
in
en
co
«*
CO
in
en
CD
t
en
£
o
o
CO
o
o
00
in
00
CM
CO
I
CO
CD
en
eo
co
.co
en
oo
co
co
co"
en
in
^t
oq
co
en
in
*
en
fx
o»
|x
«*
00
8
00
5-5 I 1
< 5 21
ill1"
in
i
in
co
ID
00
00
co
CO
CO
CN
0»
8
CO
en
o>
CO
in
in
in
q
CD
CM
-------
.,-5 S i -.
(D
CO
CM
CO
CM
CM
00
O
CM
CM
00
6
CM
CM
CO
CD
CM
CM
CM
CM
CM
in
oo
in
CM
in
oq
in
CM_
co
in
CO
CM
CD
CM
CO
CM
CO
CO
CO
*
co
CM
CO
CO
CO
00
CD
W
00
CO
in
in
CM
en
00
00
CM
co
CO
en
CO
09
*
CM
0)
CO
00
M
CO
CO
CM
2
09
CO
CO
CO
CM
CO
CO
CM
CO
CM
CO
CM
CO
#
in
co
to
in
CO
If)
00
in
00
CM
co
o>
00
CO
CO
CO
CO
CO
00
CO
o>
10
00
s*^
I
5
CO
o>
IO
(O
0>
10
<
O)
iq
CO
co
IO
o>
CO
co
CO
CO
CM
CM
O)
S
CO
CO
CO
8
CM
O
o
IO
!O
CM
in
m
CM
IO
CM
in
CM
10
CM
ID
CM
to
CM
CS
CM
in
CM
tn
in
O
O
O
10
§
IO
in
o
o
in
o o
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
251
II
-------
Table 6-7. Estimates of Sensitivity and Negative Predictive Value Presented in Tables
6-3, 6-4, and 6-6
_. *•
.Sot off St&nddrds
%of -
Interior
Paint tlUlt
b :
D8fIt8Q9CI
LBP ",
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2':---
2
, %of
Exterior
Pasit tfist
- IS
Damaged
IBP
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Soil-Lead;
* Cone*
(ppm)
-
400
400
400
400
400
400
400
400
400
400
1200
1200
1200
1200
1200
1200
1200
1200
1200
1200
2000
2000
2000
2000
2000
2000
Window
Sill Dust-
* Lead
Loading
frig/ft*}
-
-
250
250
250
250
250
250
250
250
250
-
250
250
250
250
250
250
250
250
250
-
250
250
250
250
250
Floor
Dust-Lead
Loading
fog/ft*).
-
-
-
200
100
50
40
25
20
10
5
»
-
200
100
50
40
25
20
10
5
-
--
200
100
50
40
SENSITIVITY*
(% of Housing Units with EBL
Children That Are At or Above
" At Least-One Standard}
Data for
177 units
(Table
6-3} x-
83.7%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
90.7%
93.0%
93.0%
93.0%
95.3%
95.3%
95.3%
97.7%
100%
100%
88.4%
90.7%
90.7%
90.7%
93.0%
93.0%
Data for
184 units
jjtaMe
:-6-4}-
81.8%
97.7%
97.7%
97.7%
97.7%
97.7%
100%
100%
100%
100%
100%
88.6%
90.9%
90.9%
90.9%
93.2%
95.5%
95.5%
97.7%
100%
100%
86.4%
88.6%
88.6%
88.6%
90.9%
93.2%
Data, for
205 units
OTabte
64)
81.3%
97.9%
97.9%
97.9%
97.9%
97.9%
100%
100%
100%
100%
100%
89.6%
91.7%
91.7%
91.7%
93.8%
95.8%
95.8%
97.9%
100%
100%
87.5%
89.6%
89.6%
89.6%
91.7%
93.8%
NEGATIVE PREDICTIVE VALUE
~ (% of Housing Units At or
Above No Standards That Do
" " Not Have EBL Children}
Data for
177 units
. (Table
/ 6-3f ,
84.4%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
90.0%
91.7%
91.7%
91.7%
94.3%
94.1%
94.1%
96.7%
100%
100%
88.1%
89.2%
89.2%
89.2%
91.7%
91.4%
Data for
184 units
fTabte
- *4)f.
83.3%
96.2%
95.8%
95.8%
95.8%
95.8%
100%
100%
100%
100%
100%
88.4%
89.7%
89.7%
89.7%
92.1%
94.4%
94.4%
96.9%
100%
100%
86.7%
87.5%
87.5%
87.5%
89.7%
91.9%
Data for
205 units
(Table
6-6):
83.3%
96.4%
96.2%
96.2%
96.2%
96.0%
100%
100%
100%
100%
100%
89.1%
90.5%
90.5%
90.5%
92.5%
94.7%
94.7%
97.1%
100%
100%
88.0%
88.4%
88.4%
88.4%
90.2%
92.3%
252
-------
Table 6-7. (cont.)
Set of Standards
. %ol
Ifitenor
Point thflt
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
2
2
2
2
%of.
Exterior
Paint that
is .
Damaged
LBP
10
10
10
10
10
1O
10
10
10
10
10
10
10
10
Soil-Lead
Cone.
Ippm)
2000
2000
2000
2000
5000
5000
5000
5000
5000
5000
5000
5000
5000
5000
Window
SiD Dust-
Lead
Losding
digits
250
250
250
250
-
250
250
250
250
250
250
250
250
250
-• Floor
Dust-Lead
Loading
(PS/ft1)
25
20
10
5
-
-
200
100
50
40
25
20
10
5
SENSITIVITY
(% of Housing Units with EBL
Chadren That Are At or Above
At Least One Standard)
Data for
177 units
(Tabte
6-3)
93.0%
97.7%
100%
100%
86.0%
88.4%
88.4%
88.4%
90.7%
90.7%
90.7%
95.3%
100%
100%
Data for:
184 units
(Table
&4)
93.2%
97.7%
100%
100%
84.1%
86.4%
86.4%
86.4%
88.6%
90.9%
90.9%
95.5%
100%
100%
Data for
205 units
(Table
fr6)
93.8%
97.9%
100%
100%
83.3%
85.4%
85.4%
85.4%
87.5%
89.6%
89.6%
93.8%
100%
100%
NEC ATTVE PREDICTIVE VALUE
{% of Housing Unto At or -
-Above No Standards That Do .
Not Have EBL ChBdren)
Data for
177 units
CTabte
6-3J_
91.4%
96.7%
100%
100%
86.4%
87.2%
87.2%
87.2%
89.5%
89.2%
89.2%
93.8%
100%
100%
*
*Wx-v f *
-Data for
184 units
fTaUe
.-*JWI
91.9%
96.9%
100%
100%
85.1%
85.7%
85.7%
85.7%
87.8%
89.7%
89.7%
94.1%
100%
100%
Data .for
205 units:
(Tabte
6-6}
92.3%
97.1%
100%
100%
84.9%
84.8%
84.8%
84.8%
86.4%
88.1%
88.1%
91.9%
100%
100%
6.1.4.2 Considering only Soil and Dust Standards. The analysis in the previous
subsection emphasized the difficulty in evaluating candidate paint standards using the Rochester
data, not only due to the fact that the Rochester study did not measure total area corresponding to
deteriorated lead-based paint, but also that most of the housing units with deteriorated lead-based
paint exceeded the candidate standards that were considered in that analysis. In the analysis
presented in this subsection, a paint standard was not considered. Instead, the performance
characteristics analysis considered only candidate standards for soil-lead, floor dust-lead, and
window sill dust-lead, and then investigated the percentage of painted surfaces that contained
deteriorated lead-based paint for those houses that did not exceed any of these three candidate
standards, in an effort to characterize the extent to which these houses would possibly exceed a
paint standard. The candidate standards for dust and soil in this analysis were the same as in the
previous subsection:
• uncarpeted floor dust-lead loading: 5,10,20,25,40,50,100,200 ng/ft2
253
-------
• window sill dust-lead loading: 250
• vardwide average soil-lead concentration: 400,1200,2000,5000 ug/g
The following 57 combinations of candidate standards were considered in this analysis:
* 8x1x4=32 combinations of the candidate floor-dust, sill-dust, and soil standards
• 4x1=4 combinations of only the candidate soil and sill-dust standards
• 1x8=8 combinations of only the candidate floor-dust and sill-dust standards
• 4 candidate soil standards without the others
* 1 sill-lead standard without the others
* 8 candidate floor-lead standards without the others.
The analysis was applied to data for housing units in the Rochester study having data that could
be compared to each of the standards included in the given combinaton. Average soil-lead
concentration for housing units equaled the average of the dripline and play area soil-lead
measures. Units having either dripline soil-lead data or play area soil-lead data, but not both, had
an average soil-lead concentration equal to the reported concentration at the area represented by
the available data. An average soil-lead concentration of 0 ppm was assigned to housing units
having no soil-lead data and no bare soil from which to sample.
Table 6-8 contains the results of the performance characteristics analysis, with each row
of the table corresponding to one of the 57 combinations of candidate standards being
considered. The following are examples of how to interpret the findings within Table 6-8:
• Consider combinations of all three standards where the candidate soil-lead
uncarpeted floor-dust standard of 50 (Jg/ft2, only one of the 44 homes containing
children with elevated blood-lead concentration did not exceed any of these three
standards and did not contain any deteriorated lead-based paint. (Two other
homes with an elevated blood-lead child also do not exceed these dust or soil
standards, but they do contain some deteriorated lead-based paint) Therefore,
under these standards, this particular unit would not be triggered for intervention,
regardless of the paint standard, despite the unit containing a child with an
elevated blood-lead concentration. However, if the uncarpeted floor-dust standard
was lowered to 40 ug/ft2, the house would exceed this lower floor standard.
• Consider the combination involving only aJEloor dust-lead standard of 20 us/ft2
and a window sill dust-lead standard of 250 us/ft2. A total of 106 of the 188
homes met or exceeded at least one of these two standards, including 36 of the 45
homes with elevated blood-lead children. Of the 82 homes that did not meet or
exceed either dust standard, 9 contained an elevated blood-lead child, of which 2
had no deteriorated lead-based paint in either the interior or exterior. This means
that if only dust and paint standards were considered, these two homes would not
be triggered for any intervention, despite containing elevated blood-lead children.
254
-------
•7
e
i
€
O
GC
O
•ft
=s
»0
•i OT
w —
3 U
o i
z •
S «
*5
c »
51
ii
&«
« £
c S
< 2
S-o
c
j«
O (A
JH ^3
_g W
O -O
O k
t o
O 3
^v5
i »
O 3
EC Q
GO
o
2
o
5
o
Al
o
o
•o
O
o
elevated
m
o>
q
A)
ID
Q.
•o
"3
Q.
.23
•s
CD
O.
•o
03
O
IT)
A O
cn (0
J £:
.11
Q} {D
O £
CD *-
CO O
fc-
•o ^~
Q} OO
».E
*- ro
CO CD
C °-
° S
c .i
« 5
CD O)
— 'C
** .—
.2 «
*? •-
w it
Q RJ
S ^
o^i
i
-
S; j j i 2
w
*TT " tO •« **
•&iSil«
Hi HI
;--•
a-.
»!•
CA •
n
n
n
CM
co 3
•^ JK
^ CM
5
5 uj
m
o>
3*
C 10
CO rsi
*~ 00
SO
00
CO
CD #
CM Z
0(6
"
LO aj
5m
0 ^
10
CM,
00
CO
CM
CO
CM
CM
CM #
0 CO
O)
CO
CM
5#
CO (0
en
-------
5l*«-
Ji-f •'
51*1
f<
*i:
CO
«D
M
CD
CM
CM
CM
CM
CO
CO
CM
CO
CM
CO
.8
A
CM
CM
CM
CN
CM
CO
SJ/Sf:
!« w
CD £
a»
2 w
2*
|q
2 «
s
Us
00 *
E*
CO CO
o> s
«- e>
m CO
0> 00
S*
5 d>
•* co
CO go
^ IS
r-. en
Is
IS
CD 3
sS
S oo
GO >£
is
is
? 10
95 CD
o> £
00
co £
o
>,
:5!
I'll
i
CO
0)
i
IO
u>
CM
h. #
to
*"" CD
CO ^t
(O
5s
§
CO
8
I-.
CO
10 jj?
5 co
10 10
CM 5
CO
g
IO
CM
00
s
256
-------
K£ i
O Z|p«
3 * *•'
3*f
S2£
8
n
N
CM
ID
CM
CM
CM
A.
§fii
Is 2?
ifi*
=> S J
t
CO
CM
CM
at
CD
S|<£o>
:|lH!
IIP
' !
r*» oo
co
^ to
CM pj
*~ CO
*
CO
£ •*
CO o>
>l ° * 5 5
II 2«a*
V O
n
o 3
10 *^
r- CO
^ cn
to
r- UJ
^ CO
•- ^
^ CO
•* £1
CM W
o> S
CO ^
CO IO
o> £
co
•^ ^
SCO
to
rv ID
W «
n *
CN oo
P: co
CO £l
eo
CM
co to
CO CM
If.I
%|a-ii
«1 *'
51 l;
CO
o
<5
Si-.
CO
oo
S
§
§
§*
CO ^
S
r™
CO
0)
10
S*
5 «
S
S
8
g
8
CM 10
* S
8
8
257
-------
•
CM
<*.
s
A
CM
CM
CM
(O
CO
CM
w
CO
CM 0)
2*
9J CM
CM CO
en
«0
2 *
g> oo
£ <°~
CO CO
Ii
10
"
5 jo
CO £»
ss
SD
Ii
S
8
« S
? (p
CM 00
co u>
- IO
CM 00
CO S
I?
g~~ p»
IO
!*
^ 0
(0
56
S JM
•i a
35
CO 10
s*
i ">
to en
co ™
CO CO
oo co
co 5
*
S co
CO CO
5 r-
2 "^
•* o>
S
'<* «
01
CO
co
en
CO
S
^-
8
00
ri"
o
CO
00
IO
S
1*-.
0« .I
*1-
o
IO
to
CM
ff
o
u>
(N
s
s
CM
o
to
CM
O
U)
CM
s
CM
O
§
§
258
O
O
-------
11 I :
s«l'*V
Is 21 li
sa*l5
53
i'l.
i £
1<
5s
*•
.r»
A
•*: JO
CO
(O
(O
o*
15'
W "I
"
co
CD
us
SK
CO
ss
r- CO
a> jo
» Sp_
S?
? co
en co
CM*
0> ffl
II I|l|(
CO
S t
!S t>
m n
- 9
(^ CO
co n
U) ^
»— CO
S
is!
W
00 CM
§£
00 CM
co £
ȣ
C "
co *
f- !?.
o>
§#•
? en
?s
CN IO
a £
0^ gS <
£-»^£* s •'
r* ^
0) in
S
CO 50
10 co
CM to
r- T^
Is
S« =
3
CD
S
oS
S
n
15
15-
ii
JLL
\H
!Q §.
8
8
§
8
8
S
259
-------
I
CO
(6
ii
il
*i«i-
3
"S51S.1
*||3=|
** 3
— * _J 5 «S
? I III
r*'w
il-.
^ ** Tt*
CO
g"~
2 10
CO ;>
3
0>
to
CM
O
8
IO
CM
IO
CM
1C
N
ID 3
c*
10 ri
« *
^ to
m
CO
CM
CD
10 u>
CO d
84
48
o
IO
CM
8
S
72/184
260
Il
i!
c E
.S 3
Q. »
1!
la
il
!!
|l
§ S
S*
1?
*s
II
p
Is
3-5
5
.
!l
s
S-JB
-o £
°!
IB
II
§•§
!5
* f 8
! It
i s s.
\\
I I
I I I
9 s 1
I 1 I
I - I
III
HI
I | f
« g rf
** -fc. GO
« s 2
I -I I
So?
o <*> -g
^ 1 S
€
« 1
. * I
111
! ! !
! I ~
I ;> 5
£ S §
> | .£
2 i £
•215
f I *
o *
S 5
S %
! II
o
c
S f
S 2
I 1
I I
I i
i
I I
I §
ii <11 i ! f
III i
*i? I
^
E * g
I I I
I I !
If «
i! Ill
5 g I -e
£ £ . > S
i nil
S o>0 > £•
2 &
^ o £
. g
i
S • 2
g
1
_
o "5 o o •
s s
s
-
af
i .1 7 .8 5 .1 ^ If
I te CD S »= i OK)
Sf g|if|
i« aS *S S5
-------
6.1.4.3 Analysis Involving Only Dust-Lead Standards and a Standard on the
Amount of Deteriorated Paint. In some cases, a risk assessment may involve only dust
sampling (of floors and window sills) and a visual inspection of painted surfaces for
deterioration. That is, no testing of painted surface for lead within the paint would be done, and
no soil sampling would be done. In this setting, it was of interest to investigate the extent to
which candidate dust-lead loading standards, with standards on the maximum percentage of
surfaces with deteriorated paint, performed in the absence of soil standards, within a performance
characteristics analysis. The combinations of standards considered in this analysis were the
following:
uncarpeted floor dust-lead loading: 5,10,20,25,40,50,100 ug/ft2
• window sill dust-lead loading: 125,250 ug/ft2
• maximum amount of deteriorated paint on a tested surface: >5%, >15%.
The candidate paint standards were defined to coincide with the type of paint condition
measurement made in the Rochester study. The following 63 combinations of these candidate
standards were considered in this analysis:
• 7x2x2=28 combinations of the candidate floor-dust, sill-dust, and paint standards
• 7x2= 14 combinations of only the candidate floor-dust and sill-dust standards
• 7x2=14 combinations of only the candidate floor-dust and paint standards
* 7 candidate floor-lead standards without the others.
Table 6-9 contains the results of the performance characteristics analysis, with each row
of the table corresponding to one of the 63 combinations of candidate standards being
considered. The following are examples of what can be concluded from Table 6-9:
• While, on their own, the higher candidate floor dust-lead standards trigger few
units containing elevated blood-lead children, the number of these homes that are
triggered with the addition of a deteriorated paint standard increases dramatically
(e.g., from 10.6% to 70.2% at a floor dust-lead standard of 100 ug/ft2, if the 15%
paint standard is added).
• The performance characteristics do not appear to increase substantially with an
increase in the sill standard from 125 to 250 ug/ft2.
If the risk assessment does, in fact, do paint testing for lead, then the above standard for
paint can be re-defined to represent the maximum amount of deteriorated lead-based paint on a
tested surface. Table 6-10 contains the results of the performance characteristics analysis where
the paint standard is modified in this manner.
261
-------
Table 6-9. Results of Performance Characteristics Analysis Performed on Data for
Housing Units in the Rochester Lead-in-Dust Study, for Specified Sets of
Candidate Standards for Dust-Lead Loadings and Observed Amount of
Damaged Paint on a Tested Surface
EBL - elevated blood-lead level fe10//g/dU
Sot of Csnciki&tO: Stsndvds.
f ^ < *
Uncarpeted
.Boor Dust-
Lad
* Loading
. (pgrft*)
100
50
40
25
20
10
5
100
SO
40
25
20
10
5
100
50
4O
25
20
10
5
100
50
40
25
20
10
Window
Sill Dust-
Lead
Loading r
big/ft*)
-
-
-
-
-
-
-
250
250
250
250
250
250
250
125
125
125
125
125
125
125
-
-
-
-
-
-
.Max.
Aim. Of
Damaged
Paint on a
/Tested
Surface
(%)',
••?'"' " "*•*
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
>15%
>15%
>15%
>15%
>15%
>15%
# Units
. At Or
Above
At Least
-One
Standard
/Total #
Units2 -
9/197
19/197
31/197
58/197
84/197
150/197
180/197
71/189
75/189
80/189
93/189
1O6/189
150/189
176/189
116/189
118/189
122/189
128/189
134/189
159/189
180/189
101/197
105/197
108/197
116/197
127/197
170/197
, Perfonit&nc6 C
*
Sensitivity
# {%} of Units
with EBL
Children That
Are At or'Above
At Least One
rJ^Stwidard* — :
5/47 110.6%)
9/47 (19.1%)
16/47 (34.0%)
26/47 (55.3%)
31/47 (66.0%)
44/47 (93.6%)
45/47 (95.7%)
25/45 (55.6%)
27/45 (60.0%)
30/45 (66.7%)
34/45 (75.6%)
36/45 (80.0%)
44/45 (97.8%)
44/45 (97.8%)
35/45 (77.8%)
36/45 (80.0%)
38/45 (84.4%)
39/45 (86.7%)
40/45 (88.9%)
45/45 (100%)
45/45 (100%)
33/47 (70.2%)
34/47 (72.3%)
35/47 (74.5%)
35/47 (74.5%)
38/47 (80.9%)
46/47 (97.9%)
. ,. Specificity
* ^ -
#(%) of Units
; withNoEBL
.Chadren That Are
• At or- Above No
s Standard*,
» - hi * '•< "
146/150 (97.3%)
140/150(93.3%)
135/150(90.0%)
118/150(78.7%)
97/150(64.7%)
44/1 50 (29.3%)
15/150(10.0%)
98/144(68.1%)
96/144 (66.7%)
94/144 (65.3%)
85/144 (59.0%)
74/144(51.4%)
38/144(26.4%)
12/144(8.3%)
63/144 (43.8%)
62/144(43.1%)
60/144(41.7%)
55/144 (38.2%)
50/144(34.7%)
30/144 (20.8%)
9/144 (6.3%)
82/150(54.7%)
79/150(52.7%)
77/150151.3%)
69/1 50 (46.0%)
61/150(40.7%)
26/150(17.3%)
tat JtCUH ISUCir
£PV
#{%} of Unto
At or Above At
Standard That
Have EBL
Children ^
5/9 (55.6%)
9/19 (47.4%)
16/31 (51.6%)
26/58 (44.8%)
31/84(36.9%)
44/150(29.3%)
45/180 (25.0%)
25/71 (35.2%)
27/75 (36.0%)
30/80 (37.5%)
34/93 (36.6%)
36/106(34.0%)
44/150(29.3%)
44/1 76 (25.0%)
35/116(30.2%)
36/118(30.5%)
38/122(31.1%)
39/128(30.5%)
40/134(29.9%)
45/159(28.3%)
45/180(25.0%)
33/101 (32.7%)
34/105 (32.4%)
35/108 (32.4%)
35/116(30.2%)
38/127(29.9%)
46/170(27.1%)
> V >*
i NPV
# {%) of Units At
or Above No
Standard That Do
Not Have EBL
. ChOdran*^ .
, H *^t s ?a- •
""--^-'
146/188 (77.7%)
140/178 (78.7%)
135/166(81.3%)
118/139(84.9%)
97/113(85.8%)
44/47 (93.6%)
15/17(88.2%)
98/118(83.1%)
96/114(84.2%)
94/109 (86.2%)
85/96 (88.5%)
74/83 (89.2%)
38/39 (97.4%)
12/13(92.3%)
63/73 (86.3%)
62/71 (87.3%)
60/67 (89.6%)
55/61 (90.2%)
50/55 (90.9%)
30/30 (100%)
9/9 (100%)
82/96 (85.4%)
79/92 (85.9%)
77/89 (86.5%)
69/81 (85.2%)
61/70(87.1%)
26/27 (96.3%)
262
-------
Table 6-9. (cont.)
:' Set of Candidate Standards
Uncarpetad
Floor Dust-
Lead
' Loading
(ng/ft2)
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10-
5 •
100
50
40
25
20
10
5
Window
SHI Dust-
Lead ;
1 nwii****
Loaning
<«im*>
-
-
-
-
-
-
-
-
250
250
250
250
250
250
250
125
125
125
125
125
125
125
250
250
250
250
250
250
250
Max!
Amt-Of
Damaged
Paint on a
Tested
Surface
c%r
>15%
>5%
>5%
>5%
>5%
>5%
>5%
>5%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>5%
>5%
>5%
>5%
>5%
>5%
>5%
# Units
-At Or
Above
At least
.' . One
Standard
/Totai#
Unto2
188/197
164/197
165/197
167/197
167/197
168/197
185/197
192/197
118/189
120/189
123/189
127/189
134/189
167/189
182/189
142/189
144/189
147/189
149/189
152/189
171/189
182/189
162/189
163/189
165/189
165/189
166/189
179/189
185/189
Sensitivity
#(%) of Units
whhEBL
Children That
Are At or Above
At Least One
•"•• StAIMadllP
47/47 (100%)
43/47191.5%)
44/47 (93.6%)
45/47 (95.7%)
45/47 (95.7%)
45/47 (95.7%)
47/47 (100%)
47/47 (100%)
36/45 (80.0%)
37/45 (82.2%)
38/45 (84.4%)
38/45 (84.4%)
40/45 (88.9%)
45/45(100%)
45/45(100%)
39/45 (86.7%)
40/45 (88.9%)
41/45(91.1%)
41/45(91.1%)
42/45 (93.3%)
45/45 (100%)
45/45(100%)
41/45(91.1%)
42/45 (93.3%)
43/45 (95.6%)
43/45 (95.6%)
43/45 (95.6%)
45/45 (100%)
45/45 (100%)
^•*uiii«Qii*>v \*nai€i\*tKn»tK
-------
Table 6-9. (cent.)
:- Set of Candidate Standards
:UnC0fpeted
Roar Dust-
Lead
Loading
.teg/ft*)
100
50
40
25
20
10
5
Window
SOT Dust-
Lead
Loading
{figlfft
125
125
125
125
125
125
125
Max.,
Aim.OfV
•Damaged
Paint on a
; Tested
Surface
{%>'•'
>5%
>5%
>5%
>5%
>5%
>5%
>5%
9 Units
At Or
Above
; At Least
;;One
Stsnd&rd
/Total*
Units*
166/189
167/189
169/189
169/189
170/189
179/189
185/189
- Performance C
Sensitivity ,
#(%> of Units
wfflhlBUf'V
Ctiildr6nwjnat
Are At or Above
At Least One
r'StflHuoiu „-
42/45 (93.3%)
43/45 (95.6%)
44/45 (97.8%)
44/45 (97.8%)
44/45 (97.8%)
45/45 (100%)
45/45 (100%)
Specificity
#(%) of Units
with No EBL
Chadren That Are
At or Above No
St&ndsfd^
20/144(13.9%)
20/144(13.9%)
19/144(13.2%)
19/144(13.2%)
18/144(12.5%)
10/144(6.9%)
4/144 (2.8%)
». ••mjUiu-Si-rijte
Ml CtblOl H4MA
PPV
#(%) of Units
At or Above At;
•• Least One ' "~
Standard That-
HaveffiL:
Chadren5
42/166(25.3%)
43/167(25.7%)
44/169(26.0%)
44/169(26.0%)
44/170(25.9%)
45/179(25.1%)
45/185(24.3%)
NPV
# (%) of Units At
'- '» or -Above Ma
Standard Tnat Do
_/ Not Have EBL
Children'
20/23 (87.0%)
20/22 (90.9%)
19/20(95.0%)
19/20(95.0%)
18/19(94.7%)
10/10(100%)
4/4 (100%)
1 In the Rochester study, each measurement of lead in paint had the amount of damaged paint specified as * 15%" (poor condition) of the tested surface, with no indication of total damaged surface area.
1 Total number of units having available data that could be compared to all specified candidate standards.
9 Cell entries areinumber of homes at or above at least one standard that have EBL children)/ number of homes containing EBL children),
followed by the corresponding percentage (in parentheses).
4 Cell entries are (number of homes not at or above at least one standard that do not have EBL children)/(total number of homes not
containing EBL children), followed by the corresponding percentage (in parentheses).
* Cell entries are (number of homes at or above at least one standard that have EBL chik)ren)/(total number of homes at or above at least
one standard), followed by the corresponding percentage (in parentheses).
' Cell entries are (number of homes not at or above at least one standard that do not have EBL children)/(total number of homes not at or
above any standard), followed by the corresponding percentage (in parentheses).
264
-------
Table 6-10. Results of Performance Characteristics Analysis Performed on Data for
Housing Units in the Rochester Lead-in-Dust Study, for Specified Sets of
Candidate Standards for Dust-Lead Loadings and Observed Amount of
Damaged Lead-Based Paint on a Tested Surface
EBL = elevated blood-lead level (:>10//g/dU; LBP=Lead-Based Paint
Set of Candidate Standards
Uncarpeted
Floor Dust-
Lead
Loading
tt*g/f*l
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
Window
Sin Dust-
Lead ~
Loading
frig/ft*).
-
-
-
-
-
- •
-
250
250
250
250
250
250
250
125
125
125
125
125
125
125
-
-
-
-
-
-
Max.
Aim. Of
Damaged
LBPona
Tested:
Surface"'
:<%)'
-
-
-
-
-
-
-
-
-
_
-
-
-
• -
-
-
-
-
-
-
-
>15%
>15%
>15%
>15%
>15%
>15%
# Units
At Or.
Above
At Least
One
$t&nd8rd'
/Total*
: Units2
9/197
19/197
31/197
58/197
84/197
150/197
180/197
71/189
75/189
80/189
93/189
106/189
150/189
176/189
116/189
118/189
122/189
128/189
134/189
159/189
180/189
84/197
88/197
94/197
104/197
115/197
162/197
Performance Characteristics
Sensitivity
#<%) of Units
with EBL
CrfldrenThat
Aro At or Above
At Least One
' * Standard^
5/47 (10.6%)
9/47(19.1%)
16/47(34.0%)
26/47 (55.3%)
31/47 (66.0%)
44/47 (93.6%)
45/47 (95.7%)
25/45 (55.6%)
27/45 (60.0%)
30/45 (66.7%)
34/45 (75.6%)
36/45 (80.0%)
44/45 (97.8%)
44/45 (97.8%)
35/45 (77.8%)
36/45 (80.0%)
38/45 (84.4%)
39/45 (86.7%)
40/45 (88.9%)
45/45 (100%)
45/45 (100%)
27/47 (57.4%)
28/47 (59.6%)
31/47(66.0%)
33/47 (70.2%)
36/47 (76.6%)
44/47 (93.6%)
Soeeificitv
# (%) of Units
with No &L
Children That Are
At or Above No
Standard4
146/150(97.3%)
140/150(93.3%)
135/150(90.0%)
118/150(78.7%)
97/150(64.7%)
44/150(29.3%)
15/150(10.0%)
98/144(68.1%)
96/144166.7%)
94/144(65.3%)
85/144(59.0%)
74/144(51.4%)
38/144 (26.4%)
12/144(8.3%)
63/144 (43.8%)
62/144(43.1%)
60/144(41.7%)
55/144(38.2%)
50/144(34.7%)
30/144 (20.8%)
9/144 (6.3%)
93/150(62.0%)
90/150(60.0%)
87/150(58.0%)
79/150(52.7%)
71/150(47.3%)
32/150(21.3%)
PPV
* (%) of Units,,
At or Above At
Least One-
Standard; That ;
Have EBL
Chadren*
5/9 (55.6%)
9/1 9 (47.4%)
16/31 (51.6%)
26/58 (44.8%)
31/84(36.9%)
44/150(29.3%)
45/180(25.0%)
25/71 (35.2%)
27/75 (36.0%)
30/80 (37.5%)
34/93 (36.6%)
36/106 (34.0%)
44/150(29.3%)
44/176(25.0%)
35/116(30.2%)
36/118(30.5%)
38/122(31.1%)
39/128(30.5%)
40/134(29.9%)
45/159 (28.3%)
45/180(25.0%)
27/84(32.1%)
28/88(31.8%)
31/94(33.0%)
33/104(31.7%)
36/115(31.3%)
44/162(27.2%)
NPV
#{%} of Units At
• or Above No
Standard That Do
Not Have EBL
CtiikifSfii *«,
146/188(77.7%)
140/178 (78.7%)
135/166 (81.3%)
118/139(84.9%)
97/113(85.8%)
44/47 (93.6%)
15/17(88.2%)
98/118(83.1%)
96/114(84.2%)
94/109(86.2%)
85/96 (88.5%)
74/83 (89.2%)
38/39 (97.4%)
12/13 (92.3%)
63/73 (86.3%)
62/71 (87.3%)
60/67 (89.6%)
55/61 (90.2%)
50/55 (90.9%)
30/30(100%)
9/9 (100%)
93/113(82.3%)
90/109 (82.6%)
87/103 (84.5%)
79/93 (84,9%)
71/82(86.6%)
32/35(91.4%)
265
-------
Table 6-10. (cont.)
Set of Candidate Standards _ -
Uncarpeted
Floor Dust-
Lead
Loading
-torn1)
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
Window
SiD Dust-
Lead
Lending
(ra/frt
-
-
-
-
-
-
-
-
250
250
250
250
250
250
250
125
125
125
125
125
125
125
250
250
250
250
250
250
250
Max.
Ami. Of :
Damaged
LBPona
Tested.
'Surface
(%)'
>15%
>5%
>5%
>5%
>5%
>S%
>5%
>5%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>5%
>5%
>5%
>5%
>5%
>5%
>5%
# Units
At Or
Above
At Least
One
.St&ndara
/Total*
Units2
183/197
146/197
147/197
149/197
150/197
155/197
181/197
189/197
107/189
109/189
114/189
119/189
126/189
160/189
178/189
135/189
137/189
141/189
144/189
147/189
167/189
181/189
147/189
148/189
150/189
151/189
156/189
175/189
182/189
Sensitivity
# (%) of Units
withEBL
Children That:
Are At or Above
At Least One
Standard: ~
45/47 (95.7%)
39/47 (83.0%)
40/47 (85.1%)
41/47 (87.2%)
42/47 (89.4%)
44/47 (93.6%)
47/47 (100%)
47/47 (100%)
32/45(71.1%)
33/45 (73.3%)
36/45 (80.0%)
37/45 (82.2%)
39/45 (86.7%)
44/45 (97.8%)
44/45 (97.8%)
37/45 (82.2%)
38/45 (84.4%)
40/45 (88.9%)
41/45(91.1%)
42/45 (93.3%)
45/45 (100%)
45/45 (100%)
38/45 (84.4%)
39/45 (86.7%)
40/45 (88.9%)
41/45(91.1%)
43/45 (95.6%)
45/45 (100%)
45/45 (100%)
. . ^ . ^
Specificity
# (%) of Units
with No SL
Chfldren That Are
At or Above No
Standard?
12/150(8.0%)
43/1 50 (28.7%)
43/1 50 (28.7%)
42/150(28.0%)
42/150(28.0%)
39/150(26.0%)
16/150(10.7%)
8/150(5.3%)
69/144 (47.9%)
68/144(47.2%)
66/144(45.8%)
62/144(43.1%)
57/144(39.6%)
28/144(19.4%)
10/144(6.9%)
46/144(31.9%)
45/144(31.3%)
43/144(29.9%)
41/144 (28.5%)
39/144(27.1%)
22/144(15.3%)
8/144(5.6%)
35/144 (24.3%)
35/144(24.3%)
34/144(23.6%)
34/144(23.6%)
31/144(21.5%)
14/144(9.7%)
7/144(4.9%)
P£V
# (%) of Units
At or Above At
Least One
Standard That
HaveEBL
dhadren5
45/183(24.6%)
39/146 (26.7%)
40/147 (27.2%)
41/149 (27.5%)
42/150(28.0%)
44/155(28.4%)
47/181 (26.0%)
47/189 (24.9%)
32/107.(29.9%)
33/109(30.3%)
36/114(31.6%)
37/119(31.1%)
39/126(31.0%)
44/160(27.5%)
44/178(24.7%)
37/135(27.4%)
38/137(27.7%)
40/141 (28.4%)
41/144 (28.5%)
42/147 (28.6%)
45/167(26.9%)
45/181 (24.9%)
38/147 (25.9%)
39/148(26.4%)
40/150(26.7%)
41/151 (27.2%)
43/156(27.6%)
45/175(25.7%)
45/182(24.7%)
• <•
NPV
# (%) of Units At
or Above No
Standard That Do
Not HaveEBL
Children*
12/14(85.7%)
43/51 (84.3%)
43/50 (86.0%)
42/48 (87.5%)
42/47 (89.4%)
39/42 (92.9%)
16/16(100%)
8/8 (100%)
69/82(84.1%)
68/80 (85.0%)
66/75 (88.0%)
62/70 (88.6%)
57/63 (90.5%)
28/29 (96.6%)
10/11 (90.9%)
46/54 (85.2%)
45/52 (86.5%)
43/48 (89.6%)
41/45(91.1%)
39/42 (92.9%)
22/22 (100%)
8/8(100%)
35/42 (83.3%)
35/41 (85.4%)
34/39 (87.2%)
34/38 (89.5%)
31/33 (93.9%)
14/14(100%)
7/7(100%)
266
-------
Table 6-10. (cont.)
•••: Set of Candidate Standards
»""*•- f--£il "X.
ROOT Dust-
Lead "
Loading
100
50
4O
25
20
10
5
Window
Sffl Dust-
Lead
Loading
(i*g/ft*>*
12S
125
125
125
125
125
125
Max.:;
Amt. Of
Damaged
LBPona
tested-
Surface
(%)'
>5%
>5%
>5%
>5%
>5%
>5%
>5%
# Units
At Or
Above
At Least
One
Standard
/Total #
Units -
156/189
157/189
159/189
160/189
163/189
176/189
183/189
V " . , * '••<-.. -Performance C
f "* - • -t ' ~
Sensitivity
# (%) of Units
with EBL •'
Children That
Are At or Above:
At Least One
~ Standard^.
40/45 (88.9%)
41/45(91.1%!
42/45 (93.3%)
43/45 (95.6%)
44/45 (97.8%)
45/45 (100%)
45/45 (100%)
Specificity
#(%) of Units
with No EBL
Children That Are
At or Above Mo
Standard* :
28/144(19.4%)
28/144(19.4%)
27/144(18.8%)
27/144(18.8%)
25/144(17.4%)
13/144(9.0%)
6/144 (4.2%)
haracteristics
Bfv
-V.
# (%) of Units
At or Above At
Least One "
Standard That
Have EBL
Children5
40/156(25.6%)
41/157(26.1%)
42/159(26.4%)
43/160(26.9%)
44/163(27.0%)
45/176(25.6%)
45/183(24.6%)
-
- NPV
#(%)ofUnhsAt
or Above No
Standard That Do
Not Have EBL
Children*
28/33 (84.8%)
28/32 (87.5%)
27/30 (90.0%)
27/29(93.1%)
25/26 (96.2%)
13/13(100%)
6/6 (100%)
11n the Rochester study, each measurement of lead in paint had the amount of damaged paint specified as "<5%" (good condition), *5-
15%' (fair condition), or "> 15%' (poor condition) of the tested surface, with no indication of total damaged surface area.
2 Total number of units having available data that could be compared to all specified candidate standards.
* Cell entries arelnumber of homes at or above at least one standard that have EBL children)/ number of homes containing EBL children),
followed by the corresponding percentage (in parentheses).
* Cell entries are (number of homes not at or above at least one standard that do not have EBL children)/(total number of homes not
containing EBL children), followed by the corresponding percentage (in parentheses).
* Cell entries are (number of homes at or above at least one standard that have EBL chtldren)/(total number of homes at or above at least
one standard), followed by the corresponding percentage (in parentheses).
' Call entries are (number of homes not at or above at least one standard that do not have EBL children)/(total number of homes not at or
above any standard), followed by the corresponding percentage (in parentheses).
267
-------
6.2 INVESTIGATING INCIDENCE OF ELEVATED BLOOD-LEAD
CONCENTRATION IN HOUSING UNITS MEETING ALL
EXAMPLE OPTIONS FOR STANDARDS
An alternative to the performance characteristics analysis approach (Section 6.1) to
evaluating a set of candidate standards is to use statistical modeling techniques to predict a
distribution of blood-lead concentration as a function of environmental-lead levels found in
homes which do not exceed any of the candidate standards, then estimate the percentage of
children residing in these homes that are expected to have elevated blood-lead levels (i.e., at or
above 10 ug/dL). It is desired to select a set of candidate standards so that the likelihood of
children with elevated blood-lead concentration residing in homes that do not exceed any of the
candidate standards would be very low. This section presents a modeling approach to estimate
this likelihood, using the alternative Rochester multimedia model presented in Section 4.2 of this
report ("Model A" in Table 4-1), and applies this approach to data from the Rochester study.
Recall from Section 4.2 that the reason for developing the alternative Rochester
multimedia model was to have the risk estimates from model-based analyses be more comparable
to the results of the performance characteristics analysis presented in the §403 proposed rule
(Section 6.1.3) and the results of the follow-up performance characteristics analyses (Section
6.1.4). In particular, both the performance characteristics analysis and the model-based approach
involving the alternative Rochester multimedia model use the following types of data as input
when characterizing risk:
» household average (wipe) dust-lead loading from uncarpeted floors
• household average (wipe) dust-lead loading from window sills
* yard-wide average soil-lead concentration
* the larger of the following two percentages: % of interior tested surfaces that
contain deteriorated lead-based paint (LBP), and % of exterior tested surfaces that
contain deteriorated LBP
In the model-based analysis approach presented below, the candidate standards were used to
identify a subset of homes in the Rochester study that were below all of the candidate standards,
calculate the average (across homes) of the above three measures of lead levels in dust and soil,
and fit the multimedia model to these average lead levels in order to predict a distribution of
blood-lead concentrations for children residing in these homes. For simplicity, this analysis
assumes that the homes do not contain deteriorated lead-based paint. Because the slope estimate
for the paint variable in the alternative Rochester multimedia model is nearly zero (Table 4-1 of
Section 4.2), making the assumption that no deteriorated lead-based paint exists in these homes
should have a very minor impact on the resulting risk estimates.
6.2.1 The Model-Based Approach
This model-based approach had the following four steps:
268
-------
1. For a given set of candidate standards for floor dust-lead loading, window sill
dust-lead loading, and soil-lead concentration, identify those homes in the
Rochester study that exceed none of the candidate standards in this set.
2. For each of the following three household measures, calculate the average across
the homes identified in step #1: the household average floor dust-lead loadings,
household average window sill dust-lead loading and for yard-wide average soil-
lead concentration. These three averages are assumed to represent lead levels in
housing represented by the Rochester study homes in step #1 (i.e., homes not
exceeding any of the candidate dust and soil standards).
3. Use the three averages calculated in step #2 as input to the alternative Rochester
multimedia model from Section 4.2 (assuming no .deteriorated lead-based paint
exists in the units).
4. Assume that log-transformed blood-lead concentration for children residing in the
homes identified in step #1 is normally distributed with mean equal to the
predicted log-transformed blood-lead concentration that is output from the model
fitting in step #3, and standard deviation equal to ln(1.6). (Recall that this
assumption on variability was made throughout the §403 risk analysis.) Using
normal distribution theory, determine the percentage of children represented by
this blood-lead distribution that have log-transformed blood-lead concentration or
above log(10), or equivalently, that have blood-lead concentration at or above 10
ug/dL.
6.2.2 Examples of Applying the Model-Based Approach
To illustrate how the approach in Section 6.2.1 is applied to data from the Rochester
study, the following combinations of candidate dust-lead and soil-lead standards are considered:
• (uncarpeted) floor dust-lead loading: either 40 or 50 (ig/ft2
• window sill dust-lead loading: 250 ug/ft2
• yard-wide soil-lead concentration: 400 fig/g.
When the candidate floor dust-lead loading standard is 40 ug/ft2. then the performance
characteristics analyses documented in Table 6-8 of Section 6.1 (i.e., the row of Table 6-8
corresponding to these three candidate standards) indicates that 39 of the 184 Rochester study
homes having measurements for dust-lead, soil-lead, and deteriorated lead-based paint do not
exceed any of the three candidate standards. Across these 39 homes, the following averages were
calculated from the Rochester study data:
household average (uncarpeted) floor dust-lead loading: 12.7 ug/ft2
household average window sill dust-lead loading: 87.0 ug/ft2
269
-------
* yard-wide average soil-lead concentration: 125.3 ug/g.
When fitting the alternative Rochester multimedia model to these three averages (assuming no
deteriorated lead-based paint), the model predicts a geometric mean blood-lead concentration of
4.68 ug/dL. If the standard deviation of log-transformed data is assumed to be 1.6 and normal
distribution theory is applied as described above, then the estimated percentage of children with
blood-lead concentration at or above 10 ug/dL in homes that do not exceed any of the candidate
standards is 5.30%. This matches closely with the estimate of 5.1 %, or 2 of these 39 homes in
the Rochester study dataset, which the performance characteristics analysis (Table 6-8) indicated
contained children with elevated blood-lead concentrations.
If the candidate floor dust-lead loading standard is increased to 50 ug/ft2. then the number
of Rochester study homes having measurements for dust-lead, soil-lead, and deteriorated lead-
based paint and that do not exceed any of the three candidate standards increases by one home, to
40 total homes. Across these 40 homes, the following averages were calculated from the
Rochester study data:
* household average (uncarpeted) floor dust-lead loading: 13.4 ug/ft2
• household average window sill dust-lead loading: 85.6 fig/ft2
« yard-wide average soil-lead concentration: 122.2 ug/g.
The predicted geometric mean blood-lead concentration under these assumed dust-lead and soil-
lead levels (assuming no deteriorated lead-based paint) is 4.69 ug/dL. and the estimated
percentage of children with blood-lead concentration at or above 10 ug/dL is 5.34%. This is a
very slight increase from the estimate generated under the candidate floor dust-lead loading
standard of 40 ug/ft2. The performance characteristics analysis (Table 6-8) indicated that under
these candidate standards, 7.5% of homes not exceeding any of the standards (i.e., 3 of these 40
homes in the Rochester study dataset) contained children with elevated blood-lead
concentrations.
While these examples illustrate the estimation process, they also-show that the number of
homes in the given dataset whose lead levels fall below all specified candidate standards can be
quite small, especially when at least one of the candidate standards is set at the low end of the
distribution of lead levels (i.e., most homes have data that fall above the candidate standard).
Therefore, as the set of candidate standards becomes more stringent, and as the size of the sample
from which the environmental-lead data originate becomes smaller as a result, the variability
associated with the estimated risk increases. Furthermore, as the set of candidate standards
becomes less stringent (i.e., as the standards increase), the group of homes not exceeding any of
the candidate standards is more likely to remain the same, and as a result, the estimated risk
eventually reaches a plateau. This occurs in the above examples, as increasing the candidate
floor dust-lead loading standard from 40 to 50 ug/ft2 does little, if any, to increase the estimated
risk beyond 5.3% under this approach and under the given set of data, assuming the candidate
standards for the other media (window sill dust, soil) remain fixed.
270
-------
The Rochester study data were used in this analysis as the multimedia model was fitted
based on the Rochester data. If data from other studies were used instead, it would be necessary
to verify that the model parameter estimates adequately reflect the underlying variability in these
data in the same manner that they reflect variability in the Rochester study data.
While the approach presented in this section is relatively easy to implement, it could be
modified even further in an attempt to achieve more accurate risk estimates. Such a modification
could reduce the level of simplicity associated with applying the approach. For example, rather
than calculate average environmental-lead levels across all homes and fit the model once to these
averages, a simulation approach could be applied in an attempt to more accurately represent the
entire distribution of environmental-lead levels in these homes and the resulting blood-lead
distribution associated with exposure across the entire distribution of environmental-lead levels.
6.3 REVIEW OF PUBLISHED INFORMATION ON POST-INTERVENTION
DUST-LEAD LOADINGS
This section summarizes published information on lead loadings (amount of lead per unit
surface area) in dust samples collected by wipe techniques, as reported by earlier lead
intervention studies. This information is used to evaluate assumptions made on post-intervention
dust-lead loadings (40 ug/ft2 for floors, 100 ug/ft2 for window sills) within the §403 risk analysis.
Details to supplement the summaries in this section are presented in Appendix H.
The following seven studies have been identified in which some type of paint or dust
intervention was performed, dust samples were collected using wipes or some other technique
(e.g., BRM vacuum) whose results could be converted to wipe-equivalent dust-lead loadings, and
post-intervention dust-lead loadings on floors and/or window sills were reported (references for
these studies are included in Appendix H):
• Baltimore Experimental Paint Abatement Studies
• Baltimore Follow-up Paint Abatement Study
• Baltimore Repair & Maintenance (R&M) Study
• Boston Interim Dust Intervention Study
• HUD Grantees Evaluation (data available through September 1997)
• Denver Comprehensive Abatement Performance (CAP) Study
• Jersey City Children's Lead Exposure and Reduction (CLEAR) Study
These studies employed a variety of intervention strategies, including single or repeated dust
cleanings and interim control or complete abatement of lead-based paint. Dust-lead loadings
were measured at varying intervals following intervention. Post-intervention dust-lead loadings
were summarized for 19 groups of housing units across these seven studies. These study groups
are defined hi Appendix H.
For both floors and window sills, geometric mean and median dust-lead loadings were
observed below the post-intervention assumptions established in the §403 risk analysis in a
271
-------
majority of the study groups. However, this does not preclude results for individual housing
units from being above the assumed levels. Furthermore, the extent to which results for these
studies represent the nation's housing stock has not been determined. Results are now presented
separately for floors and window sills (with more detailed presentations found in Appendix H).
6.3.1 Post-Intervention Floor Dust-Lead Loadings
Summaries of post-intervention floor (wipe) dust-lead loadings are presented in Table
6-11 according to housing group within each study. According to Table 6-11, all but two of the
19 study groups reported geometric mean or median floor dust-lead loadings at or below 41
ug/ft2 from 6 months to 6 years post-intervention. The other two study groups were from the
Baltimore Experimental Paint Abatement Study, where pre-intervention geometric mean dust-
lead loadings were much greater (556 ng/ft2 and 1261 ng/ft2) than any other study group (at most
58.6 ug/ft2). Eleven study groups reported geometric mean or median floor dust-lead loadings at
or below 21 ug/ft2 at follow-up periods ranging from 12 months to 2 years. Of these 11 groups,
four of the HUD Grantees study groups reported median floor dust-lead loadings at or below 10
ug/ft2 at 12 months post-intervention. Median pre-intervention floor dust-lead loadings in these
four groups ranged from 9 to 26 ug/ft2.
In the HUD Grantees evaluation, seven of the eight largest grantees have median floor
dust-lead loadings at or below 21 ug/ft2 at 12 months post-intervention, compared to a median of
14 ug/ft2 across all grantees. Although pre-intervention floor dust-lead loadings were lower in
the HUD Grantees evaluation compared to other studies, these preliminary results suggest that
floor dust-lead loadings can be maintained at levels below 40 ug/ft2 for at least 12 months post-
intervention.
Results from the Denver CAP study, the Baltimore Follow-up Paint Abatement study, the
Baltimore R&M study, the Boston Interim Dust Intervention study, and the Jersey City CLEAR
study suggest that geometric mean floor dust-lead loadings of below 40 ug/ft2 can be observed
even beyond 12 months post-intervention and up to six years post-intervention, under the same
conditions experienced by the housing units in these studies.
6.3.2 Post-Intervention Window Sill Dust-Lead Loadings
Summaries of post-intervention window sill wipe dust-lead loadings are presented in
Table 6-12 according to housing group. Post-intervention geometric means or medians range
from 24 jig/ft2 to 958 fig/ft2, which are considerably higher than the summaries for floors.
Eleven study groups had geometric mean or median post-intervention window sill dust-lead
loadings below 100 ug/ft2,6 groups were at or below 51 ug/ft2, and 3 groups were at or below 41
ug/ft2.
All but one of the HUD Grantees study groups (the Milwaukee grantee) had median
window sill dust-lead loadings below 100 ^g/ft2 at 12 months post-intervention. As the
intervention strategy for homes in the HUD Grantees evaluation frequently included partial or
272
-------
Table 6-11. Summaries of Pre- and Post-Intervention Floor Wipe Dust-Lead Loadings for
Housing Groups Within Seven Studies
Study
Baltimore
Experimental Paint
Abatement
Studies2
Baltimore Follow-
up Paint
Abatement Study2
Baltimore R&M
Study3
Boston Interim
Dust Intervention
Study2
HUD Grantees4
Denver CAP Study8
Jersey City CLEAR
Study
Study
Group
Study 1
Study 2
1 2-Month Follow-up
19-Month Follow-up
Previously-Abated Units
Units Slated for R&M
Intervention
Automatic Intervention
Randomized Intervention
All Grantees
Baltimore
Boston
Massachusetts
Milwaukee
Minnesota
Rhode Island
Vermont
Wisconsin
Abated Units
Intervention Group
Pro-Intervention Floor
Dust-Lead Loadings1
(pg/ft2)
1261
556
NA
NA
45.6
58,6
33.2
37.3
19
41
24 .
24
14
18
26
28
9
NA
22
Post-Intervention
Floor Dust-Lead Loadings1
Time Following
Intervention
(Months)
6-9
1.5 -3.5 Years
10-14
14-24
4 - 6 Years
24
6
6
12
12
12
12
12
12
12
12
12
2 Years
12
i Summary Value
(pg/ft*)
99
69
20
36
33.0
35.0
23.9
31.4
14
41
18
9
10
18
6
21
5
21.0
15
1 Values are geometric means except for the HUD Grantees studies, where values are medians. "NA" indicates not
available.
* Results are adjusted to reflect total dust-lead loadings by exponentiating the "bioavailable" dust-lead loadings as reported in
the study to the 1.1416 power.
3 Results for the Baltimore R&M Study are converted from BRM dust-lead loadings to wipe-equivalent loadings.
* Data collected through September, 1997
5 Results for the Denver CAP study are converted from CAP cyclone dust-lead loadings to wipe-equivalent loadings.
273
-------
Table 6-12. Summaries of Pre- and Post-Intervention Window Sill Wipe Dust-Lead
Loadings for Housing Groups Within Seven Studies
Study
Baltimore
Experimental Paint
Abatement
Studies2
Baltimore Follow-
up Paint
Abatement Study2
Baltimore R&M
Study3
Boston Interim
Dust Intervention
Study2
HUD Grantees4
Denver CAP
Study5
Jersey City CLEAR
Study
Study
Group
Study 1
Study 2
12-Month Follow-up
19-Month Follow-up
Previously-Abated Units
Units Slated for R&M
Intervention
Automatic Intervention
Randomized Intervention
All Grantees
Baltimore
Boston
Massachusetts
Milwaukee
Minnesota
Rhode Island
Vermont
Wisconsin
Abated Units
Intervention Group
Pro-Intervention SHI
Dust-Lead Loadings1
fe/g/ft2)
15215
2784
NA
NA
163.5
778.4
787
205
258
1191
174
328
264
266
314
147
150
NA
75
Post-Intervention
Sill Dust-Lead Loadings1
Time Following
Intervention
6-9
1.5 -3.5 Years
10-14
14-24
4 - 6 Years
24
6
6
12
12
12
12
12
12
12
12
12
2 Years
12
Summary Value
(TO/*2)
958
199
41
147
97.6
204.9
210
110
90
68
49
50
217
77
85
40
51
66.4
24
1 Values are geometric means except for the HUD Grantees studies, where values are medians. "NA' indicates not
available.
2 Results are adjusted to reflect total dust-lead loadings by exponentiating the "bioavailable' dust-lead loadings as reported in
the study to the 1.1416 power.
3 Results for the Baltimore R&M Study are converted from BRM dust-tead loadings to wipe-equivalent loadings.
* Data collected through September, 1997
* Results for the Denver CAP study are converted from CAP cyclone dust-lead loadings to wipe-equivalent loadings.
274
-------
complete window replacement, these results may not be representative of the outcomes of
interventions prompted by the §403 rule.
Geometric mean window sill dust-lead loadings were below 100 ^fi2 for up to two years
post-intervention in the Baltimore Follow-up Paint Abatement study, Denver CAP study, and
Jersey City CLEAR study. However, in the Baltimore R&M study, Baltimore Experimental
Paint Abatement studies, and Boston Interim Dust Intervention study, geometric mean dust-lead
loadings remain above 100 ug/ft2 over time. In addition, the 19-month follow-up study group
within the Baltimore Follow-up Paint Abatement study and study group #2 of the Baltimore
Experimental Paint Abatement studies suggest that geometric mean dust-lead loadings can dip
, below 100 Mg/ft2 immediately after intervention, but then exceed this level after one year or so.
6.4 SENSITIVITY AND UNCERTAINTY ANALYSES FOR RISK
MANAGEMENT ANALYSES
The following subsections present the results of additional sensitivity and uncertainty
analyses performed to gauge the level of uncertainty in the post-§403 risk estimates (and the
associated decline from baseline estimates) associated with methodological assumptions. These
results should be considered with those presented in the sensitivity and uncertainty analyses in
Section 6.4 of the §403 risk analysis report to characterize overall uncertainty associated with the
methods and assumptions taken in the risk management.
6.4.1 Considering How Baseline Environmental-Lead Levels May
Have Changed Since the HUD National Survey
Section 5.1.4 of this report addressed the sensitivity of the pre-§403 model-based blood-
lead distribution and the resulting health effects and blood-lead concentration endpoint estimates
under the IEUBK and empirical models under different assumptions on how the national
distribution of baseline environmental-lead levels as estimated using HUD National Survey data
may have changed since the time of the survey (1989-1990). The same five sets of adjustments
(i.e., percentage changes) made to the average baseline dust-lead loadings, dust-lead
concentrations, and soil-lead concentrations for each housing unit in the HUD National Survey
were considered in this sensitivity analysis to observe the impact on post-§403 risk estimates
under the following set of example options for standards:
• Average floor dust-lead loading =100 ug/ft2
• Average window sill dust-lead loading = 500 ug/ft2
• Average soil-lead concentration = 2,000 \igfg
• Amount of deteriorated lead-based paint requiring paint maintenance = 5 ft2
• Amount of deteriorated lead-based paint requiring paint abatement - 20 ft2
This set of options was the primary set considered in the sensitivity analyses within Section 6.4
of the §403 risk analysis report.
275
-------
Table 6-13 presents the post-§403 estimates for the health effect and blood-lead
concentration endpoints under both the IEUBK and empirical models, for each of the five sets of
adjustments to the post-§403 environmental-lead levels hi housing units within the HUD
National Survey and under the above assumption on example standards. Also included in this
table are the percentage of homes exceeding the various example standards, which will be lower
than in the §403 risk analysis when declines in the appropriate environmental-lead levels are
considered and higher when increases are considered. The table also lists the baseline risk
estimates for comparison purposes.
Effect on risk analysis: Under the five sets of assumptions involving lower assumed
baseline environmental-lead levels, the percentage of houses that exceed at least one of the
example standards declined by at most about three percentage points (from 21.8% to 18.7%;
Table 6-13), or about three million homes. The assumption that baseline environmental-lead
levels are 25% higher than assumed in the §403 risk analysis results in an increase in the
percentage of homes exceeding at least one standard from 21.8% to 24.1%, an increase of about
2.3 million homes (Table 6-13).
As would be expected, Table 6-13 shows that all assumptions on baseline environmental-
lead levels result in post-§403 estimates of the predicted health effect and blood-lead
concentration endpoints that are lower than baseline (the last column of the table). However, as
the assumed baseline environmental-lead levels become lower in magnitude, the predicted post-
1403 risks actually increase, converging to the baseline estimates. For example, as seen in Table
6-3, baseline lead levels that are 20% below what was assumed in the §403 risk analysis resulted
in an estimated percentage of children with blood-lead concentrations at or above 10 ug/dL of
4.85%, compared to the §403 risk analysis estimate of 4.70%. When baseline lead levels are
50% below the §403 risk analysis estimates, the estimate of this percentage increases to 5.10%.
Such a finding appears counter-intuitive when first reviewing the table. However, the alternative
assumptions being considered hi this sensitivity analysis are to baseline (i.e., pre-§403)
environmental-lead levels. As assumptions on these baseline levels move lower, fewer homes
are triggered by the §403 standards, and the post-§403 distribution of environmental-lead levels
becomes less removed from the baseline distribution. As a result, post-§403 estimates of
predicted health effects and blood-lead concentration are not as different from pre-§403
estimates. In contrast, as assumed baseline environmental-lead levels increase, more homes are
triggered by the §403 standards and, therefore, have their environmental-lead levels drop as a
result of interventions, and lower post-§403 risk estimates relative to baseline are observed.
As seen in Table 6-13, the effect that different assumptions on baseline environmental-
lead levels have on the risk estimates is considerably greater under the IEUBK model than the
empirical model. The percentage of children with blood-lead concentrations at or above 20
ug/dL more than triples under the IEUBK model approach when 50% declines hi both dust-lead
and soil-lead levels were assumed (from 0.054% to 0.166%), compared to a 16% increase under
the empirical model (from 0.406% to 0.469%). Smaller percentage differences are observed for
the other endpoints for both models.
276
-------
Table 6-13. Sensitivity Analysis on How Changes in Household Average Baseline Dust-
Lead Loadings/Concentrations and Soil-Lead Concentration Impact Post-
8403 Estimates of Health Effect and Blood-Lead Concentration Endpoints for
Children Aged 1-2 Years Under a Specified Set of Example Standards1
Assumed Percentage Change In Average Dust-Lead Loadings and Concentrations
(Both Floor and Window Sill) and in Yard-wide Average Soil-Lead Concentration
Dust:
•--:,. . Soil:
.LNo
change
l*to
change
20% ;
decrease
20% ,
decrease.
50%
decrease
', w^> ,''
decrease'
50%
decrease
No.
. change
No
change
. 50%
decrease
Percentage of Homes Exceeding Example Standards/Triggers
Floor Dust
Window Sill Dust
Soil
Interior Paint Maintenance
Exterior Paint
Maintenance
Interior Paint Abatement
Exterior Paint Abatement
Any Standard/Trigger
4.04
12.5
2.49
2.92
3.49
2.43
5.77
21.8
2.34
10.8
1.52
2.92
3.49
2.43
5.77
20.6
0.694
9.10
0.746
2.92
3.49
2.43
5.77
18.7
0.694
9.10
2.49
2.92
3.49
2.43
5.77
18.9
4.04
12.5
0.746
2.92
3.49
2.43
5.77
21.6
25%
increase
. 25%
increase
Baseline
Estimate
{from
Table 5-1
of tiie
§403 risk
analysis
report)
5.68
14.3
3.27
2.92
3.49
2.43
5.77
24.1
Predicted Hearth Effect And Blood-Lead Concentration Endpoints (Based on Empirical Model)
PbB *20 (%)
PbBi10(%)
IQ < 70 (%)
IQ decrement 2 1 (%)
IQ decrement *2 (%)
IQ decrement 23 (%)
Avg. IQ decrement
0.406
4.70
0.110
36.3
9.30
2.93
1.00
0.429
4.85
0.111
36.7
9.53
3.04
1.01
0.469
5.10
0.112
37.3
9.90
3.21
1.03
0.445
4.95
0.111
36.9
9.69
3.11
1.02
0.427
4.84
0.111
36.7
9.51
3.03
1.01
0.378
4.52
0.110
35.9
9.02
2.80
0.995
0.588 •
5.75
0.115
38.5
10.8
3.70
1.06
Predicted Health Effect And Blood-Lead Concentration Endpoints (Based on IEUBK Model)
PbB 220 (%)
PbB2lO(%)
IQ < 70 <%)
IQ decrement 2! (%)
IQ decrement ±2 (%)
IQ decrement a3 (%)
Avg. IQ decrement
0.0539
1.66
0.0984
28.3
4.31
0.858
0.848
0.117
2.48
0.102
31.0
5.77
1.37
0.894
0.166
2.98
0.104
32.7
6.65
1.71
0.924
0.121
2.55
0.102
31.6
5.94
1.42
0.904
0.0681
1.86
0.0992
28.8
4.67
0.983
0.857
0.0542
1.64
0.0982
27.7
4.22
0.847
0.839
0.588
5.75
0.115
38.5
10.8
3.70
1.06
1 Example dust and soil standards were set at: 100 //g/ft2 for floor dust-lead loading, 500 //g/ft2 for window sill
dust-lead loading, and 2,000 jt/g/g for soil-lead concentration. Paint maintenance is performed if more than 5
ft2, but less than 20 ft2 of deteriorated lead-based paint exists. Paint abatement is performed if more than 20
ft2 of deteriorated lead-based paint exists.
277
-------
6.4.2 Impact on the Estimated Incidence of IQ Point Decrement
Assuming Certain Thresholds on the IQ/Blood-Lead Relationship
The sensitivity of baseline and pre-§403 model-based estimates of IQ decrements greater
than 1,2, or 3, and of the average and standard deviation of the distribution of IQ point
decrements was addressed in Section 5.1.5 of this report for various assumptions of a non-zero
threshold of blood-lead concentration on the IQ/blood-lead relationship. The following
thresholds were considered: 1,2, 3,5, 8 and 10 ug/dL. In this section, post-§403 estimates of
these health effect endpoints are estimated (under the same set of options presented in Section
6.4.1, using both the DEUBK and empirical models) under these same alternative blood-lead
concentration thresholds. These estimates are presented in Table 6-14.
Effect on risk analysis: As was also seen in Table 5-7 of this report, Table 6-14 shows
that the post-§403 risk estimates decrease as the assumed blood-lead concentration threshold
increases (i.e., smaller percentages of children experience IQ score decrements under larger
threshold assumptions). The IEUBK model is more sensitive than the empirical model to the
threshold level. For example, the probability of a child experiencing an IQ decrement of at least
1 point decreases by 63% under the IEUBK model (from 28.3% to 10.4%) when the threshold
increases from 0 to 2 ug/dL, compared to only a 52% decrease under the empirical model (from
36.3% to 17.6%). As the assumed threshold increases, the likelihood of experiencing an IQ
decrement of at least 1 point as a result of lead exposure decreases to very low values under both
models, and the average IQ score decrement in the population declines to small fractions of
points.
6.4.3 Considering Alternative Assumptions on Post-Intervention
Dust-Lead Loadings
In the risk management portion (Chapter 6) of the §403 risk analysis report, it was
necessary to make assumptions on predicted post-intervention lead levels when characterizing
the blood-lead concentration and health effect endpoints in a post-§403 environment. These
assumptions were documented in Table 6-2 of the §403 risk analysis report. Among these
assumptions were that dust cleaning activities impacted interior dust-lead loadings in the
following way:
• Post-intervention household average floor (wipe) dust-lead loadings equaled the
minimum of 40 ug/tfand the pre-intervention value.
• Post-intervention household average window sill (wipe) dust-lead loadings
equaled the minimum of 100 ug/ft* and the pre-intervention value.
A dust cleaning was assumed to be included among the interventions performed when either the
floor-dust, window sill-dust, soil, or interior paint abatement standards were exceeded within a
home. These two assumptions on post-intervention dust-lead loadings were made within the
§403 risk analysis based on data reported in EPA's Comprehensive Abatement Performance
278
-------
Table 6-14. Sensitivity Analysis on the Assumed Blood-Lead Concentration Threshold on
IQ Decrement and Its Impact on the Post-§403 Estimates of IQ Decrement
Endpoints for Children Aged 1-2 Years, Under a Specified Set of Example
Standards1
Assumed
Threshold
fc/g/dLl
% of Children Aged 1-2 Years with a Specified IQ
Decrement' Due to Lead Exposure2
IQ Decrement as-,'
T . ,;':
IQ Decrement 2
;]:,:•• 2 '..
IQ Decrement 2
3 ' '
Average IQ
Decrement
{# points)3
Standard
Deviation of IQ
n»».^. ......
-------
study and in the Baltimore Experimental Paint Abatement study (see Section 6.1.2 of the §403
risk analysis report and Section H2.0 of Appendix H of this report).
Tables 6-11 and 6-12 within Section 6.3 of this report presented additional information
on household average (wipe) dust-lead loading at pre- and post-intervention for floors and
window sills, respectively, from several recent lead intervention studies. This information, some
of which was received after the §403 risk analysis report was completed, suggests that it may be
common in some instances to observe household average post-intervention dust-lead loadings
below the assumptions made above, even from 12 months to six years post-intervention. These
findings prompted a sensitivity analysis to investigate how setting assumptions on post-
intervention household average dust-lead loadings to below the 40 ug/ft2 and 100 jug/ft2
specifications would impact the outcome of the risk management analyses.
In this sensitivity analysis, two alternative assumptions on household average post-
intervention floor dust-lead loadings were made: 10 ug/ft2 and 25 fig/ft2. As the geometric mean
(12-month) post-intervention floor dust-lead loading in the HUD Grantees evaluation was 14
Hg/ft2 (Table 6-8) and was even lower for certain grantees, an alternative of 10 ug/ft2 was
selected. The alternative of 25 ug/ft2 for floors was selected as it fell halfway between the
assumptions of 10 and 40 ug/ft2 and was within the range of expected variability in the
summaries for several of the studies in Section 6.3.1.
Similarly, two alternative assumptions on household average post-intervention window
sill dust-lead loadings were made: 50 ug/ft2 and 75 ng/ft2. Evidence from Table 6-12 indicates
that average window sill dust-lead loadings following intervention could approach 50 ug/ft2 in
some instances, especially when floor dust-lead loadings are low. The alternative of 75 ug/ft2
was selected as it fell halfway between the assumptions of 50 and 100 ug/ft2, and it was similar
to the average levels observed by grantees within the HUD Grantees evaluation (although the
HUD Grantees evaluation included window replacement, which was not among the assumed
interventions in the §403 risk analysis).
In the sensitivity analysis, if a given household's pre-intervention average floor dust-lead
loading fell below the given post-intervention assumption, its post-intervention household
average floor dust-lead loading was assumed to be equal to its pre-intervention average (as was
done in Chapter 6 of the §403 risk analysis report). Second, this sensitivity analysis considers
predictions made only by the empirical model, as the IEUBK model does not accept dust-lead
loading as input. Finally, the assumptions made in determining post-intervention soil-lead
concentrations (150 ug/g following soil removal) and amount of deteriorated lead-based paint
(none is present following paint intervention) remained the same as specified in Table 6-2 of the
§403 risk analysis report.
Table 6-15 presents the estimated post-§403 health effect and blood-lead concentration
endpoints associated with the set of example options for standards specified in Section 6.4.1
above, for the alternative assumptions on post-intervention floor and window sill dust-lead
loadings specified above. Note that each alternative assumption is evaluated on its own (i.e., it is
280
-------
•S **
IB*
»
IS-
n*
is
II
CO 10
". •' I***"
CO ID
sg
«•*
is
250
id
V*
CD
.O
CO
CO
CO
in
10
IX
ui
IO
IO
CD
CO
CO
6
CO
o
CO
cn
CO
(O
CM
CO
CO
CM
en
0)
00
CM
o
o
CO
CO
10
0)
id
CO
CO
q
oi
CO
CM
10
cn
cn
CM
00
i3
a
TJ
a
?
281
-S? t
SI
« TJ
« T> d>
C *• -C
t o •*
O 0)
I CL 5
^«
o «?.E 2
-------
the only change from the §403 risk analysis assumptions). In addition, considering the high
correlation in dust-lead loadings between floors and window sills, the two lower alternatives (10
ug/ft2 for floors and 50 ug/ft2 for window sills) and the two higher alternatives (25 fig/ft2 for
floors and 75 fig/ft2 for window sills) are evaluated together. For comparison purposes, post-
intervention estimates under the §403 risk analysis (i.e., assuming 40 fig/ft2 for floors and 100
ug/ft2 for window sills) and the estimates generated under baseline (pre-§403) conditions (both
presented in Table 6-7 of the §403 risk analysis report) are also included in Table 6-15.
Effect on risk analysis. Relative to the results reported hi the §403 risk analysis report
(column 2 of Table 6-15), the greatest deviation occurs with the most substantial change in the
assumptions, i.e., the assumptions of 10 ug/ft2for floors and 50 ug/ft2 for window sills (column 7
of Table 6-15). Under this particular set of alternative assumptions, the percentage of the
nation's children aged 1-2 years that are anticipated to have blood-lead concentration at or above
10 ug/dL following interventions conducted in response to the §403 rule (given the example
standards specified in the footnote to this table) is reduced from 4.70% to 4.53% (a 3.7% decline,
equivalent to approximately 13,700 children12). The corresponding reduction in the percentage
of children with blood-lead concentration at or above 20 ug/dL is from 0.406% to 0.380% (a
6.3% decline, equivalent to approximately 2,000 children).
Under the assumptions of 25 jig/ft2 for floors and 75 ug/ft2 for window sills (column 8 of
Table 6-15), the percentage of the nation's children aged 1-2 years that are anticipated to have
blood-lead concentration at or above 10 ug/dL is reduced from 4.70% to 4.64% (a 1.2% decline,
equivalent to approximately 4,800 children). The corresponding reduction in the percentage of
children with blood-lead concentration at or above 20 ug/dL is from 0.406% to 0.397% (a 2.3%
decline, equivalent to approximately 750 children).
Generally, even lower percentage declines occur for the IQ endpoints compared to the
blood-lead concentration endpoints. The exception occurs with the percentage of children with
IQ decline of at least 3 points, where a 4.2% decline from the §403 risk analysis assumptions was
observed under assumptions of 10 ug/fi2 for floors and 50 Mg/ft2 for window sills.
This sensitivity analysis indicates that while more housing units may achieve reductions
in average dust-lead levels on floors and window sills following a dust cleaning if the assumed
post-intervention floor dust-lead loadings are lowered from those made hi the §403 risk analysis,
the corresponding reduction in the estimated blood-lead concentration and health effect endpoints
appears to be modest, especially when compared to the reduction observed from pre- to post-
§403 conditions.
12 Assuming that 7.96 million children aged 1-2 years reside in the U.S. housing stock (Table 3-3S of the
§403 risk analysis report).
282
-------
6.4.4 Characterizing the Post-Intervention Blood-Lead Distribution
Based on Relative Change from Baseline in the Geometric
Mean and the Probability of a Child's Blood-Lead
Concentration Exceeding 10/ig/dL
As discussed in Section 4.3.1 above and in Appendix Fl of the §403 risk analysis report,
a "scaling algorithm" was used in the §403 risk analysis to characterize the distribution of blood-
lead concentration in the nation's children following interventions that would be performed as a
result of implementing the §403 rule (where the algorithm was applied under a specified set of
example options for the standards, using a specified blood-lead prediction model, and under
assumptions made on the changes in environmental-lead levels that result from the
interventions). This distribution is labeled the "post-§403" distribution. This approach
calculated the geometric mean (GM) and geometric standard deviation (GSD) of the post-§403
blood-lead distribution in the following manner:
( 1)
(2)
where the subscripts indicate the blood-lead distribution which either the GM or the GSD
represents. See Section 4.3.1 for additional information on this approach.
One comment received on the §403 risk analysis was that because the blood-lead
concentration endpoints utilized in the risk analysis were exceedance probabilities (i.e., the
likelihood of a child's blood-lead concentration exceeding a specified value), it was more
important to accurately characterize the right tail of the post-§403 distribution compared to the
remainder of the distribution, especially at blood-lead levels beyond 10 ug/dL. Therefore, a
variant of the scaling approach was considered that involved scaling the probability of a child's
blood-lead concentration exceeding 10 ug/dL rather than the GSD. If P10 was used to represent
this probability, then the alternative scaling algorithm would involve scaling the geometric mean
as in (I) above, but replacing (2) above with the following calculation:
P10post-403 ~P 1 "baseline 0" ^model-based post-403 ' ** "model-based pre-403)
(3)
The resulting value is the estimate of the probability of a child's blood-lead concentration
exceeding 10 ug/dL in a post-§403 environment. It is calculated by multiplying the probability
as calculated in the baseline distribution by the relative change in the probability from the pre-
§403 to post-§403 environment as estimated from model-based blood-lead distributions. Then,
in order to calculate the other blood-lead concentration and health effect endpoints, the GSD of
the post-§403 distribution would be calculated by assuming that this distribution is lognormal.
Therefore,
= exp{(log(10) -
(4)
283
-------
where O"1 denotes the inverse of the standard normal distribution function.
Table 6-16 presents the estimated blood-lead concentration and health effect endpoints
that result when applying this alternative scaling algorithm, under both the IEUBK and empirical
models. The example options for standards that are assumed in this analysis are the same as
those considered in Section 6.4.1 above and are specified in a footnote to Table 6-16. For
comparison purposes, this table also contains the estimates under the original version of the
scaling approach that was utilized in the §403 risk analysis.
Table 6-16. Estimated Post-§403 Health and Blood-Lead Concentration Endpoints Under
the Original and Alternative Scaling Algorithms for Characterizing the Post-
§403 Blood-Lead Distribution
Original scaling algorithm: Geometric mean and GSD are scaled.
Alternative scaling algorithm: Geometric mean and the probability of PbB exceeding 10^/g/dL are scaled.
Health Effect and Hood-Lead
: Concentration Endpoints
% of Children with PbB i 20 A/g/dL
% of Children with PbB 2 10//g/dL
%of Children with IQ < 70 due to
lead exposure
% of Children with IQ decrement 2 1
due to lead exposure
% of Children with IQ decrement 2 2
due to lead exposure
% of Children with IQ decrement 2 3
due to lead exposure
Avg. IQ decrement due to lead
exposure
Geometric Mean PbB (GSD)
Post-1403 Estimates Under the
Risk Management Analysis
(Original Scafing Algorithm)
IEUBK Model
0.0539
1.66
0.0984
28.3
4.31
0.858
0.848
2.74(1.84)
Empirical
Model
0.406
4.70
0.110
36.3
9.30
2.93
1.00
3.03 (2.04)
Post-1403 Estimates Under the
Alternative Scaling Algorithm
IEUBK Model;:
0.156
2.72
0.102
30.1
6.05
1.56
0.884
2.74 (1 .96)
- Empirical
Model 1 :
0.249
3.78
0.107
35.5
8.03
2.24
0.977
3.03(1.96)
Note: Example dust and soil standards were set at: 100 /rg/ft9 for floor dust-lead loading, 500 pg/ft* for window sill dust-
lead loading, and 2.000 ;ig/g for soil-lead concentration. Paint maintenance is performed if more than 5 ft*, but less than
20 ft1, of deteriorated lead-based paint exists. Paint abatement is performed if more than 20 ft2 of deteriorated lead-based
paint exists. GSD = geometric standard deviation. PbB = blood-lead concentration
284
-------
Effect on risk analysis. As indicated in Table 6-16, when the probability of exceeding
10 ug/dL is scaled instead of the GSD, the estimated probability is approximately 64% higher
under the IEUBK model (1.66% to 2.72%), but nearly 20% lower under the empirical model
(4.70% to 3.78%). Note that under the alternative approach, estimates based on the IEUBK and
empirical models are more similar to each other than under the original scaling algorithm. In the
alternative approach, the estimated post-§403 GSD is the same under both models: 1.96. Note
that there was no change in the manner in which the geometric mean blood-lead concentrations
were determined, and therefore, no change is noted between the two approaches.
The above results indicate that the alternative scaling approach has a more significant
impact on the IEUBK model-based estimates compared to the empirical model-based estimates.
The impact of the approach on the empirical model-based estimates is a reduction in the risk
estimates due to a 4% reduction in the estimated GSD, while the impact on IEUBK model-based
estimates is an increase in the risk estimates due to a 6.5% increase in the estimated GSD.
However, because the two approaches did not differ hi how the post-§403 geometric mean blood-
lead level was calculated, the empirical model estimates remain higher than the IEUBK model
estimates.
6.5 LEAD EXPOSURE ASSOCIATED WITH CARPETED FLOOR-DUST
While the §403 proposed rule included a proposed lead hazard standard for dust on
uncarpeted floors, EPA determined that sufficient technical data were not available to direct how
the rule should address lead-contaminated dust on carpeted floors. Based upon public comments
on the proposed rule, EPA is revisiting that determination. This section summarizes the key
findings of statistical analyses on dust-lead loading data for carpeted floors. The analysis had the
following three objectives:
1. Assess the need to have dust-lead on carpeted floors addressed by the §403 rule:
a. Characterize the relationship between floor dust-lead levels and blood-lead
concentration in young children and how this relationship differs for
carpeted and uncarpeted floors (with and without adjusting for the effects
of key demographic variables and for lead levels in other media in which
. standards have been proposed in the §403 rule).
b. Determine the added value of including a carpet dust-lead standard given
the proposed §403 standards for soil, window sills and uncarpeted floors,
or expanding the definition of floors in the rule to include carpeted as well
as uncarpeted floors.
2. Identify appropriate candidates for carpeted floor dust-lead standards and, in
particular, whether one candidate standard should correspond to 50 ug/ft2, the
uncarpeted floor dust-lead standard from the §403 proposed rule.
285
-------
3. Determine whether the wipe technique is acceptable for sampling dust from
carpeted floors for evaluating the risk of lead exposure associated with carpet-
dust, or whether alternative vacuum methods are more appropriate.
A more detailed presentation of the statistical analyses that address these three objectives is
found in Appendix I of this report.
The carpet dust-lead measurement data used in this analysis originated from two lead
exposure studies: the Rochester (NY) Lead-in-Dust study, and the pre-intervention, evaluation
phase of the HUD Lead-Based Paint Hazard Control Grant ("HUD Grantees") Program (data
collected through September, 1997). Both studies were introduced in Section 3.3.1 of the §403
risk analysis report; additional details on these studies that are relevant to this analysis is
presented in Section 13.1 of Appendix I. The results of this analysis, along with relevant findings
documented in EPA's recent literature review report on lead exposure associated with carpets,
furniture, and air ducts (USEPA, 1997b), were used to address the above objectives.
The summary of the analysis results now follows. It is formatted according to the above
three objectives. References to statistical significance are made at the 0.05 level. Unless
otherwise indicated, references to dust-lead loadings are assumed to be for samples collected
using wipe techniques. Section numbers within Appendix I are specified in parentheses where
additional information can be found.
Objective #1: Is there a need to have dust-lead on carpeted floors addressed by the §403
rule?
• Using data collected in the 1997 American Housing Survey, EPA estimates that
approximately 54 million housing units built prior to 1978 contain some wall-to-
wall carpeting. Of these units, wall-to-wall carpeting is found in a living room in
approximately 47 million units and in a bedroom in approximately 46 million
units (i.e., rooms in which children reside and play most frequently, and therefore,
would be targeted in a risk assessment).
• While the §403 proposed rule indicates that lead from floor dust is an important
exposure source for children, the proposed floor dust-lead loading standard was
only relevant for uncarpeted floors. In homes with wall-to-wall carpeting, it is
expected that floor-dust samples in certain rooms can come only from carpeted
floors. While no guidance was given in the §403 proposed rule on a standard to
which risk assessors should compare the results of lead analyses for carpet dust
samples, EPA recognizes (and many commenters on the §403 proposed rule have
noted) that some recommendation for a carpet dust-lead loading standard, based
on using wipe collection techniques, is necessary.
286
-------
• Because children come in frequent direct contact with carpeting when it is present
in their homes, any lead that may be present in carpet dust is likely to be
bioavailable to children.
Objective #1a: Is there any association between carpeted floor dust-lead loadings
and blood-lead concentration?
• For both carpeted and uncarpeted floors in the two studies, the correlation
between household average floor (wipe) dust-lead loading and children's blood-
lead concentration was positive and significantly different from zero. (Sections
14.1.1.1 and 15.1.1.1 of Appendix I)
• No evidence was found in these analyses to suggest that wipe dust-lead loadings
from uncarpeted floors are a better predictor of children's blood-lead
concentration than wipe dust-lead loadings from carpeted floors. (Sections
14.1.1.2,14.1.1.4,15.1.1.2 and 15.1.1.4 of Appendix I)
• No significant difference in the statistical relationship between average floor dust-
lead loading and blood-lead concentration was found between homes with floor
dust sampling conducted from mostly carpeted floors and homes with sampling
from mostly uncarpeted floors. (Sections 14.1.1.3 and 15.1.1.3 of Appendix I)
• Mixed results were found when investigating whether the effect of average
carpeted floor dust-lead loading on blood-lead concentration remained significant
after adjusting for the effects of lead levels in soil, window sill dust, and
uncarpeted floor dust (i.e., other environmental media addressed by the proposed
§403 standards). The carpet dust-lead loading effect was no longer statistically
significant after adjusting for these other effects when analyzing data from the
Rochester study, while the effect remained statistically significant when analyzing
data from the HUD Grantees program evaluation. (Sections 14.1.2 and 15.1.2 of
Appendix I)
' • When interpreted as a whole, these findings provide a powerful argument for
expanding the floor dust-lead standard in the §403 rule to include carpeted floors.
Objective #1 b: Is there any added benefit to adding a carpeted floor dust-lead
loading standard to the proposed §403 standards for lead in soil, window sill dust,
and dust from uncarpeted floors, or to expanding the definition of floors in the rule
to include carpeted floors? (Sections 14.1.3 and 15.1.3 of Appendix I)
• The extent of any added benefit is dependent on the value of the carpet dust-lead
loading standard and the particular criteria being considered in evaluating
performance. Adding a new standard to a set of existing standards will not reduce
sensitivity (i.e., the proportion of homes with elevated blood-lead children that are
287
-------
triggered by the set of standards), but it also will not increase specificity (i.e., the
proportion of homes with no elevated blood-lead children that are not triggered
for an intervention by the standards).
* If the uncarpeted floor dust-lead loading standard of 50 us/ft2 that was proposed
in the §403 proposed rule was extended to included carpeted floors as well, the
resulting performance of the§403 proposed standards (based on the outcome of
performance characteristics analysis) changed little, if any. If a carpeted floor
standard of 40 ug/ft2 was added to the §403 proposed standards, slight
improvements in the performance characteristics were noticed. These findings
were observed regardless of whether or not uncarpeted floors were available to
sample (i.e., whether or not the uncarpeted floor standard was considered).
• Analyses of the Rochester study data indicated that adding a carpet dust-lead
loading standard of approximately 17 fig/ft2 to the proposed §403 standards
considerably improved certain performance characteristics, particularly sensitivity,
without a large decrease in specificity.
• Analysis of the HUD Grantees evaluation data indicated that adding a carpet dust-
lead loading standard of approximately 5 ug/ft2 improved sensitivity and negative
predictive value (NPV, equal to the proportion of homes not triggered for
intervention by the standards that do not contain elevated blood-lead
concentration), but was accompanied by a considerable decrease in specificity. If
the proposed carpet dust-lead loading standard was increased to approximately J3
ug/ft2. this loss of specificity relative to the gains in sensitivity and NPV was
reduced.
• In general, these analyses concluded that expanding the proposed §403 floor dust-
lead standard (of 50 ug/ft2) to encompass both carpeted and uncarpeted floors, or
setting this standard slightly lower at 40 ug/ft2, would not lead to a large decrease
in specificity, but it would tend to result in only minor increases in sensitivity
from what was observed when carpeted floor standards were not being
considered.
Objective #2: If a carpeted floor standard is needed, what should it be? Should it be
different from the proposed uncarpeted floor dust-lead loading standard of 50 //g/ft2?
• The findings listed above for Objective #lb suggest that it may provide an
advantage to have a standard for carpeted floors that is lower than the standard for
uncarpeted floors. (Sections 14.1.1.2,14.2.1,15.1.1.2 and 15.2.1 of Appendix I)
• Having a floor dust-lead loading standard of 40 to 50 ug/ft2 that is expanded to
represent carpeted floors as well as uncarpeted floors would be at least as
protective of children (in terms of the predicted blood-lead concentration at which
288
-------
95% of children exposed at the standard level would be expected to fall below)
than if the standard represented only uncarpeted floors. (Sections 14.2.2 and 15.2.2
of Appendix I)
• When the Rochester study data was used in a performance characteristics analysis
that considered only standards for either carpeted or uncarpeted floors (Sections
14.2.3 and 15.2.3 of Appendix I), a carpeted floor dust-lead loading standard in the
range of 15 to 20 ug/ft2 maximized the total of the four performance
characteristics. In contrast, a standard of 50 fig/ft2 resulted in considerably lower
performance when the standard was for carpeted floors versus uncarpeted floors.
The level of sensitivity achieved by an uncarpeted floor dust-lead loading standard
of 50 ug/ft2 was achieved for carpeted floor dust-lead loading standards below
approximately 33 ug/ft2. However, the uncertainty associated with these estimates
may suggest that these lower levels may not actually differ from a practical
standpoint from the uncarpeted floor dust-lead loading standard in the §403 rule.
Objective #3; What dust sampling method should be used on carpeted floors? (Sections
14.3 and 15.3 of Appendix I)
• The HUD Guidelines (USHUD, 1995) support the use of wipe methods to sample
carpet dust. Participants in the §403 Dialogue Group meetings raised concerns
that requiring widespread use of vacuum techniques for collecting dust samples in
typical risk assessments would be impractical. Therefore, it would be preferable
to allow wipe sampling as an option for collecting dust samples from carpets in a
risk assessment unless wipe techniques were totally unacceptable.
• Different types of dust collection methods can collect different amounts of lead
within a dust sample, especially when sampling from carpets where surface dust is
easier to sample than dust that is deep within the carpet fibers. A laboratory study
done in conjunction with the Rochester study (Emond et al., 1997) concluded that
lead recovery from carpet dust was highest with the BRM vacuum (95.2%)
compared to the wipe (24.4%) and the DVM vacuum (31.4%). For this reason,
different dust collection methods for collecting carpet dust would require different
lead standards to which to compare the results.
• When the wipe method is used on carpets, it tends to collect only dust on the
carpet surface that can readily be removed by the method. This surface dust is
also that which is most likely to come into direct contact with children (USEPA,
1997b).
• Blood-lead concentration tends to be more highly associated with dust-lead
loading than with dust-lead concentration in carpets. (Only dust-lead loadings can
be measured under wipe techniques, while loadings or concentrations can be
measured under vacuum methods.) This contributes to the technical justification
289
-------
that a carpet dust-lead standard would be better conveyed as a loading than as a
concentration.
Each of the three dust collection methods considered in the Rochester study
(BRM vacuum, DVM vacuum, wipe) collected carpet dust samples whose dust-
lead loadings were statistically associated with blood-lead concentration, with the
level of association being similar for each method.
On both carpeted and uncarpeted floors, dust-lead loading measurements from
different dust collection methods were significantly positively correlated. This
suggests that using any of the three methods (including wipe) would portray the
extent of a carpet dust-lead hazard in a similar fashion.
As wipe sampling is currently the method of choice for uncarpeted floors and all
three methods have significant correlations with blood-lead concentration for
carpeted and uncarpeted floors, it is reasonable to develop a carpeted floor dust-
lead loading standard for the wipe sampling method. As this standard would not
apply to vacuum sampled dust-lead loadings, measurements for samples collected
using vacuum techniques could not be directly used in risk assessment via the
§403 rale without first being converted to wipe-equivalent loadings using methods
such as those documented in Section 4.3 of the §403 risk analysis report.
290
-------
7.0 REFERENCES
Abadin, H.G., Wheeler, IS., Jones, D.E., De Rosa, C.T., and MSPH (1997) "A Framework to
Guide Public Health Assessment Decisions at Lead Sites." Journal of Clean Technology,
Environmental Toxicology, and Occupational Medicine. 6(3):225-237.
Altmann, L., Weinsberg, F., Sveinsson, K., Lilienthal, Hi, Wiegand, H., and Winneke, G. (1993)
"Impairment of Long-Term Potentiation and Learning Following Chronic Lead Exposure."
Toxicol. Lett. 66:105-112.
Angell, N.F., and Weiss, B. (1982) "Operant Behavior of Rats Exposed to Lead Before or After
Weaning." Toxicology and Applied Pharmacology. 63:62-71.
Angle, CR., Manton, W.I., and Stanek, K.L. (1995) "Stable Isotope Identification of Lead
Sources in Preschool Children: the Omaha Study." Journal of Toxicology - Clinical
Toxicology. 33(6):657-662.
Angle, C. R., and Mclntire, M. S. (1979) "Environmental Lead and Children: The Omaha
Study." Journal of Toxicology and Environmental Health. 5:855-870.
Annest, J.L., and Mahaffey, K. (1984) "Blood Lead Levels for Persons Ages 6 Months - 74
Years: United States, 1976-1980." National Center for Health Statistics, Centers for
Disease Control and Prevention. Vital Health Stat Series No. 11 (233). (PHS) 84-1683.
August 1984.
Baghurst, P.A., McMichael, A.J., Wigg, N.R., Vimpani, G., Robertson, E.F., Roberts, R.J., and
Tong, S. (1992) "Environmental Exposure to Lead and Children's Intelligence at the Age of
Seven Years: The Port Pirie Cohort Study." New England Journal of Medicine. 327:1279-
1284.
Banks, E.C., Ferretti, L.E., and Shucard, D.W. (1997) "Effects of Low Level Lead Exposure on
Cognitive Function in Children: a Review of Behavioral, Neuropsychological and
Biological Evidence." Neurotoxicology. 18(l):237-282.
Barltrop, D., and Meek, F. (1979) Effect of particle size on lead absorption from the gut.
Archives of Environmental Health. 34: 280-285.
Barltrop, D., Strehlow, C.D., Thorton, I., and Webb, J.S. (1974) "Significance of High Soil Lead
Concentrations for Childhood Lead Burdens." Environmental Health Perspectives. 7:75-
82.
291
-------
Bates, M., Malcolm, M., Wyatt, R., Galloway, Y., Gairett, N., Speir, T., and Read, D. (1995)
"Lead in Children from Older Housing Areas in the Wellington Region." New Zealand
Medical Journal, 108(1009):400-404.
Battelle (1996a) "Statistical Evaluation of the Relationship Between Blood-Lead and Dust-Lead
Based on Data from the Rochester Lead-in-Dust Study." Final Report prepared by Battelle
for the Office of Prevention, Pesticides, and Toxic Substances, U.S. Environmental
Protection Agency. Contract No. 68-D2-0139. 22 February 1996.
Battelle (1996b) "Statistical Evaluation of the Relationship Between Blood-Lead and Dust-Lead
Based on Pre-Intervention Data from the R&M Study." Final Report prepared by Battelle
for the Office of Prevention, Pesticides, and Toxic Substances, U.S. Environmental
Protection Agency. Contract No. 68-D2-0139.
Bellinger, D., Stiles, K.M., and Needleman, H. (1992) "Low-Level Lead Exposure, Intelligence,
and Academic Achievement: A Long-term Follow-up Study." Pediatrics. 90(6):855-861.
Bellinger, D., Sloman, J., Leviton, A., Rabinowitz, M., Needleman H.L., and Waternaux, C.
(1991) "Low-Level Lead Exposure and Children's Cognitive Function in the Preschool
Years." Pediatrics. 87(2):219-227.
Bellinger, D.C., Leviton, A., Waternaux, C., Needleman, H., and Rabinowitz, M. (1987)
"Longitudinal Analyses of Prenatal and Postnatal Lead Exposure and Early Cognitive
Development." New England Journal of Medicine. 316:1037-1043.
Bellinger, D.C., Leviton, A., Needleman, H.L., Waternaux, C., and Rabinowitz, M. (1986a)
"Low-Level Lead Exposure and Infant Development in the First Year." Neurobehavioral
Toxicology and Teratology. 8:151-161.
Bellinger, D.C., Leviton, A., Rabinowitz, M., Needleman, H., and Waternaux, C. (1986b)
"Correlates of Low-Level Lead Exposure in Urban Children at 2 Years of Age."
Pediatrics. 77:826-833.
Bellinger, D.C., Leviton, A., Waternaux, C., et al. (1985a) "A Longitudinal Study of the
Developmental Toxicity of Low-Level Lead Exposure in the Prenatal and Early Postnatal
Periods." In: Lekkas TD, ed. International Conference on Heavy Metals in the
Environment, Athens, Greece, September, Vol 1. Edinburgh, United Kingdom: CEP
Consultants, Ltd, 32-34.
Bellinger, D.C., Leviton, A., Waternaux, C., and Allred, E. (1985b) "Methodological Issues in
Modeling the Relationship Between Low-Level Lead Exposure and Infant Development:
Examples from the Boston Lead Study." Environmental Research. 38:119-129.
292
-------
Bellinger, D.C. and Needleman, H.L. (1983) "Lead and the Relationship Between Maternal and
Child Intelligence." Journal of Pediatrics. 102:523-527.
Bhatia, MS., Rai, S., Singbhal, P.K., Nigam, V.R., Bohra, N., and Malik, S.C. (1988) "Pica:
Prevalence and Etiology." Indian Pediatr. 25:1165-1170.
Bornschein, R., Clark, S., Pan, W., and Succop, P. (1990) "Midvale Community Lead Study."
University of Cincinnati.
Bornschein, R. L., Succop, P., Krafft, K., Clark, S., Peace, B., and Hammond, P. B. (1986)
"Exterior Surface Dust Lead, Interior House Dust Lead and Childhood Lead Exposure in an
Urban Environment." Conference in Trace Metals in Environmental Health, Columbia,
MO.
Bornschein, R.L., Succop, P., Dietrich, K.N., Clark, C.S., Que Hee, S., and Hammond, P.B.
(1985a) 'The Influence of Social and Environmental Factors on Dust Lead, Hand Lead, and
Blood Lead Levels in Young Children." Environmental Research. 38:108-118.
Bornschein, R. L., Hammond, P. B., Dietrich, K. N., Succop, P., Krafft, K., Clark, S., Berger, O.,
Pearson, D., and Que Hee, S. (1985b) 'The Cincinnati Prospective Study of Low-Level
Lead Exposure and Its Effects on Child Development: Protocol and Status Report."
Environmental Research, 38:4-18.
Brunekreef, B., Veenstra, S.J., Biersteker, K., and Boleij, J.S.M. (1981) 'The Arnhem Lead
Study: Lead Uptake by 1- to 3- Year-Old Children Living in the Vicinity of a Secondary
Lead Smelter in Amhem, Netherlands." Environmental Research. 25:441-448.
Burger, J., Carruth-Hinchey, C, Ondroff, J., and McMahon M. (1998) "Effects of Lead on
Behavior, Growth, and Survival of Hatchling Slider Turtles." Journal of Toxicology and
Environmental Health. 55:495-502.
Burns, J.M., Baghurst, P.A., Sawyer, M.G., et al. (1999) "Lifetime Low-Level Exposure to
Environmental Lead and Children's Emotional and Behavioral Development: The Port
Pine Cohort Study." American Journal of Epidemiology. 149(8):740-749.
Bushnell, P.J. and Bowman, R.E. (1979a) "Persistence of Impaired Reversal Learning in Young
Monkeys Exposed to Low Levels of Dietary Lead." /. Toxicology and Environmental
Health. 5: 1015-1023.
Bushnell, P.J., and Bowman, R.E. (1979b) "Effects of Chronic Lead mgestion on Social
Development in Infant Rhesus Monkeys." Neurobehavioral Toxicology. 1:207-219.
Bushnell, P.J., Bowman, R.E., Allen, J.R., and Marlar, R.J. (1977) "Scotopic Vision Deficits in
Young Monkeys Exposed to Lead." Science 196:333-335.
293
-------
Butte-Silver Bow Department of Health, and Department of Environmental Health, University of
Cincinnati. (1991) "The Butte-Silver Bow Environmental Health Lead Study." Draft Final
Report.
Calabrese, E., Barnes, R., Stanek, EJ. m, Pastides, H., Gilbert, C.E., Veneman, P., Wang, X.,
Lasztity, A., and Kostecki, P.T. (1989) "How Much Soil Do Young Children Ingest: An
Epidemiological Study." Regulatory Toxicology and Pharmacology. 10:123-137.
Calabrese, E.J., and Stanek, EJ. (1993) "Soil Pica: Not a Rare Event." Journal of
Environmental Sci. Health, A28(2):373-384.
Calabrese, E., Stanek, EJ., and Barnes, R. (1997) "Soil Ingestion Rates in Children Identified
by Parental Observation as Likely High Soil Ingesters." Journal of Soil Contamination.
6(3):271-279.
Calabrese, E., Stanek, EJ., and Gilbert, C.E. (1991) "Evidence of Soil Pica Behavior and
Quantification of Soil Ingested." Human and Experimental Toxicology. 10:245-249.
Calabrese, E., Stanek, EJ., and Gilbert, C.E. (1993) "Lead Exposures in a Soil Pica Child."
J. Environ. Sci. Health. A28 (2):353-362.
Carroll, R.J., and Galindo, C.D. (1999) "Measurement Error, Biases, and the Validation of
Complex Models for Blood Lead Levels in Children." Environmental Health Perspectives.
106 (Supp. 6): 1535-1539.
Carroll R.J., Ruppert, D., and Stefanski. L.A. (1995) Measurement Error in Nonlinear Models.
London: Chapman & Hall.
CDC (1997) "Update: Blood-Lead Levels - United States, 1991-1994." Morbidity and Mortality
Weekly Report. Centers for Disease Control and Prevention, Public Health Service, U.S.
Department of Health and Human Services. 21 February 1997. 46(7): 141-146.
CDC (1991) "Preventing Lead Poisoning in Young Children: A Statement by the Centers for
Disease Control." Centers for Disease Control and Prevention, Public Health Service, U.S.
Department of Health and Human Services. October 1991.
Chen, H., Ma, T., Hume, A. S., and Ho, I. K. (1998) "Developmental Lead Exposure Alters the
Distribution of Protein Kinase C Activity in the Rat Hippocampus." Biomedical and
Environmental Sciences. 11:61-69.
Clark, C.S., Bomschein, R.L., Succop, P., Que Hee, S.S., Hammond, P.B., and Peace, B. (1985)
"Condition and Type of Housing as an Indicator of Potential Environmental Lead Exposure
and Pediatric Blood Lead Levels." Environmental Research. 38:46-53.
294
-------
Clark, S., Bornschein, R.L., Succop, P., Roda, S., and Peace, B. (1991) "Urban Lead Exposures
of Children in Cincinnati, Ohio." Chemical Speciation and Bioavailability. 3:163-171.
Cohen, C.J., Bowers, G.N., and Lepow, MX. (1973) "Epidemiology of Lead Poisoning: A
Comparison Between Urban and Rural Children." Journal of the American Medical
Association. 226:1430-1433.
Cory-Slechta, D.A. (1997) "Relationships Between Pb-Induced Changes in Neurotransmitter
System Function and Behavioral Toxicity." Neurotoxicology. 18(3): 673-688.
Cory-Slechta, D.A., Weiss, B., and Cox, C. (1985) "Performance and Exposure Indices of Rate
Exposed to Low Concentrations of Lead." Toxicology and Applied Pharmacology.
78:291-299.
Cory-Slechta, D.A. and Thompson, T. (1979) "Behavioral Toxicity of Chronic Post-weaning
Lead Exposure in the Rat." Toxicology and Applied Pharmacology. 47:151-159.
Cotter-Howells, J. (1994) "Lead Phosphate Formation in Soils." Environmental Geochemistry
and Health. 16(2): 86.
Cutler, M.G. (1977) "Effects of Exposure to Lead on Social Behavior in the Laboratory Mouse."
Psychopharmacology. 52:279-282.
Davies, DJ.A, Thornton, I., Watt, J.M., Culbard, E.B., Harvey, P.G., Delves, H.T., Sherlock,
J.C., Smart, G.A., Thomas, J.F.A., and Quinn, M.J. (1990) "Lead Intake and Blood Lead in
Two-Year-Old U.K. Urban Children." The Science of the Total Environment. 90:13-29.
Davis, A., Ruby, M.V. and Bergstrom, P.D. (1995) "Factors Controlling Lead Bioavailability in
the Butte Mining District, Montana, USA." Environmental Geochemistry and Health.
16(3-4): 147-157.
Davis, J.M., Otto, D.A., Weil, D.E., and Grant, L.D. (1990) "The Comparative Developmental
Neurotoxicity of Lead in Humans and Animals." Neurotoxicology and Teratology. 12:
215-229.
Davis, S., Waller, P., Buschom, R., Ballou, J., and White, P. (1990) "Quantitative Estimates of
Soil Digestion in Normal Children Between the Ages of 2 and 7 Years: Population-Based
Estimate Using Aluminum, Silicon, and Titanium as Soil Tracer Elements." Archives of
Environmental Health. 45:112-122.
Day, J.C. (1993) Population Projections of the United States, by Age, Sex, Race, and Hispanic
Origin: 2993 to 2050. U.S. Bureau of the Census, Current Population Reports P25-1104.
295
-------
Dersimonian, R. and Laird, N. (1986) "Mela-Analysis in Clinical Trials." Controlled Clinical
Trials. 7:177-188.
Diemel, J.A., Bnmekreef, B., Boleij, J.S., Biersteker, K. and Veenstra, SJ. (1981) "The Arnhem
Lead Study H Indoor Pollution, and Indoor/Outdoor Relationships." Environmental
Research. 25:449-456.
Dieter, M.P, Matthews, H.B., Jeffcoat, R.A. and Moseman, R.F. (1993) "Comparison of Lead
Bioavailability in F344 Rats Fed Lead Acetate, Lead Oxide, Lead Sulfide or Lead Ore
Concentrate from Skagway, Alaska." Journal of Toxicology and Environmental Health.
39(1): 79-93.
Dietrich, K.N., Berger, O.G., Succop, P.A., Hammond, P.B., and Bornschein, R.L.. (1993) "The
Developmental Consequences of Low To Moderate Prenatal and Postnatal Lead Exposure:
Intellectual Attainment in the Cincinnati Lead Study Cohort Following School Entry."
Neurotoxicology and Teratology. 15:37-44.
Duggan, M.J. and Inskip, M.J. (1985) "Childhood Exposure to Lead in Surface Dust and Soil: A
Community Health Problem." Public Health Reviews. 13: 1-54.
Elhelu, M. A., Caldwell, D., and Hirpassa, W. (1995) "Lead in Inner-City Soil and Its Possible
Contribution to Children's Blood Lead." Archives of Environmental Health. 50(2): 165-
169.
Emond, M.J., Lanphear, B.P., Watts, A., Eberly, S., and Members of the Rochester Lead-in-Dust
Study Group. (1997) "Measurement Error and Its Impact on the Estimated Relationship
Between Dust Lead and Children's Blood Lead." Environmental Research 72:82-92
Emhart, C.B., Morrow-Tlucak, M., Worf, A.W., Super, D., and Drotar, D. (1989) "Low Level
Lead Exposure in the Prenatal and Early Preschool Periods; Intelligence Prior to School
Entry." Neurotoxicology and Teratology. 11:161-70.
Farfel, M.R. and Chisolm, J.J., Jr. (1991) "An Evaluation of Experimental Practices for
Abatement of Residential Lead-Based Paint: Report on a Pilot Project." Environmental
Research. 55:199-212.
Farfel, M.R. and Chisolm, J.J., Jr. (1990) "Health and Environmental Outcomes of Traditional
and Modified Practices for Abatement of Residential Lead-Based Paint." American
Journal of Public Health. 80(10): 1240-1245.
Fleiss, J.L. (1981) Statistical Methods for Rates and Proportions. Second Edition. New York:
John Wiley & Sons. Section 1.2, pp. 4-9.
296
-------
Fox, D.A., Campbell, M.L., and Blocker, Y.S. (1997) "Functional Alterations and Apoptotic
Cell Death in the Retina Following Developmental or Adult Lead Exposure."
Neurotoxicology 18(3):645-664.
Francek, M.A., Makimaa, B., Pan, V., and Hanko, J.H. (1994) "Small Town Lead Levels: A Case
Study from the Homes of Pre-schoolers in Mt. Pleasant, Michigan." Environmental
Research. 84(2): 159-166.
Freeman, G.B., Johnson, J.D., Killinger, J.M., Liao, S.C., Feder, P.I., Davis, A.O., Ruby, M.V.,
Channey, R.L., Lovre, S.C., and Bergstrom, P.D. (1992) "Relative Bioavailability of Lead
from Mining Waste Soil in Rats." Fund. Appl. ToxicoL 19:399-398.
Fulton, M., Raab, G., Thomson, G., Laxen, D., Hunter, R., and Hepburn, W. (1987) "Influence of
Blood Lead on the Ability and Attainment of Children in Edinburgh." Lancet. 1:1221-
1226.
Galke, W. A., Hammer, D.I., and Lawrence, S. W. (1975) "Environmental Determinants of Lead
Burdens in Children." In: International Conference on Heavy Metals in the Environment-
Symposium Proceedings, T. C. Hutchinson, S. Epstein, A. L Page, J. VanLoon and T.
Davey (eds.), Institute for Environmental Studies, Toronto, ON, Canada, 3:53-74.
Gallacher, J.E., Elwood, P.C., Phillips, K.M., Davies, B.E., and Jones, D.T. (1984) "Relation
Between Pica and Blood Lead hi Areas of Differing Lead Exposure." Archives of Diseases
in Childhood. 59(1): 4(M4.
Geist, C.R., Balko, S.W., Morgan, M.E., and Angiak, R. (1985) "Behavioral Effects Following
Rehabilitation from Postnatal Exposure to Lead Acetate." Percept. Mot. Skills. 60:527-
536.
Gilbert, S.G., and Rice, D.C. (1987) "Low-level Lifetime Lead Exposure Produces Behavioral
Toxicity (Spatial Discrimination Reversal) in Adult Monkeys." Toxicology and Applied
Pharmacology. 91:484-490.
Goldstein, G. (1992) "Developmental Neurobiology of Lead Toxicity." In Human Lead
Exposure (Needleman, H., ed.). 125-135.
Goldstein, G. (1990) "Lead Poisoning and Brain Cell Function." Environ. Health Perspect.
89:91-94.
Greene, T., Ernhart, C.B., and Boyd, T.A. (1992) "Contributions of Risk Factors to Elevated
Blood and Dentine Lead Levels in Preschool Children." The Science of the Total
Environment. 115:239-260.
297
-------
Harrison, G. (1987) "A Survey of the Lead Distribution in the Soil of Corpus Christi, Texas."
Texas Journal of Science. 39(1): 15-22.
Harvey, P.O., Hamlin, M.W., Kumar, R., Morgan, G., Spurgeon, A., and Delves, H.T. (1988)
"Relationship Between Blood Lead, Behaviour, Psychometric and Neuropsychological Test
Performance in Young Children." British Journal of Developmental Psychology. 29:43-52.
Hastings, L., Cooper, G.P., Bomschein, R.L., and Michaelson, LA, (1977) "Behavioral Effects
of Low Lead Neonatal Lead Exposure." Pharmacology Biochemistry & Behavior. 7: 37-
42.
Hatzakis, A., Kolevi, A., Katsouyanni, K., et al. (1987) "Psychometric Intelligence and
Attentional Performance Deficits in Lead Exposed Children." International Conference on
Heavy Metals in the Environment. Edinburgh, Scotland; CEP Consultants. 16:57-67.
Hatzakis, A., Kokkeni, A., Maranelias, C., Katsouyanni, K., Salaminios, F., Kalandidi, A., et al.
(1989) "Psychometric Intelligence Deficits in Lead-exposed Children." In: Lead Exposure
and Child Development (Smith MA., Grant, LD., Sora Al, eds.). London: Kluwer,
1989:211-23.
Hawk, B.A., Schroeder, S.R., Robinson, G., et al. (1986) "Relation of Lead and Social Factors to
IQ of Low-SES Children: A Partial Replication." American Journal of Mental Deficiency.
91:178-183.
Karam, H.S., and Beck, B.D. (1990) "Current Issues in Determining Acceptable Lead
Concentrations in Soils." Comments on Toxicology. 3(6):509-529.
Kaufman, A.S. (1996) "IQ, Lead Level, and Inferences from Research Studies: Comments
Addressing the Underlying Science Forming the Basis of HUD's Regulatory Impact
Analysis." Submitted to the Environmental Health Committee of the Science Advisory
Board, U.S. Environmental Protection Agency, in their technical review of the proposed
TSCA Section 403 regulations (Identification of Dangerous Levels of Lead). August 1996.
Kenny, D.A. (1979) Correlation and Causality. New York: John Wiley & Sons, Inc. Chapter 1,
pp. 1-12.
Krueger, J. A. and Duguay, K. M. (1989) "Comparative Analysis of Lead in Maine Urban Soils."
Bulletin of Environmental Contamination and Toxicology. 42:574-581.
Kuhlmann, A.C., McGlothan, J.L., and Guilarte, T.R. (1997) "Developmental Lead Exposure
Causes Spatial Learning Deficits in Adult Rats." Neuroscience Newsletters. 233:101-104.
Lacey, E.P. (1990) "Broadening the Perspective of Pica: Literature Review." Public Health
Reports. 105(1 ):29-35.
298
-------
Lanphear, B.P., Matte, T.D., Rogers, J., Clickner, R.P., Dietz, B., Bornschein, R.L., Succop, P.,
Mahaffey, K.R., Dixon, S., Galke, W., Rabinowitz, M., Farfel, M., Rohde, C, Schwartz, J.,
Ashley, P., and Jacobs, D. (1998) 'The Contribution of Lead-Contaminated House Dust
and Residential Soil to Children's Blood Lead Levels: A Pooled Analysis of 12
Epidemiologic Studies." Environmental Research (Section A). 79:51-68.
Lanphear, B.P., Weitzman, M., and Eberly, S. (1996a) "Racial Differences in Urban Children's
Environmental Exposures to Lead." American Journal of Public Health. 86(10): 1460-1463.
Lanphear, B.P., Weitzman, M., Winter, N.L., Eberly, S., Yakir, B., Tanner, M., Emond, M., and
Matte, T.D. (1996b) "Lead-Contaminated House Dust and Urban Children's Blood Lead
Levels." American Journal of Public Health. 86(10): 1416-1421.
Lansdown, R., Yule, W., Urbanowicz, M.A., et al. (1986) "The Relationship Between Blood
Lead Concentrations, Intelligence, Attainment and Behavior hi a School Population: The
Second London Study." International Archives of Occupational and Environmental
Health. 57:225-235.
Laperche, V., Logan, T.J., Gaddam, P., and Traina, S.J. (1997) "Effect of Apatite Amendments
on Plant Uptake of Lead from Contaminated Soil." Environmental Science and
Technology. 31(10): 2745-2753.
Lasky, R.E., Maier, M.M., Snodgrass, E.B., Hecox, K.E., and Laughlin, N.K. (1995) "The
Effects of Lead on Otoacoustic Emissions and Auditory Evoked Potentials in Monkeys."
Neurotoxicology and Teratology 17(6):633-644.
Lilienthal, H., Kohler, K., Turfeld, M., and Winneke, G. (1994) "Persistent Increases in Scotopic
B-wave Amplitudes After Lead Exposure in Monkeys." Exp. Eye Res. 59:203-209.
Lilienthal, H., Winneke, G., and Ewert, T. (1990) "Effects of Lead on Neurophysiological and
Performance Measures: Animal and Human Data." Environmental Health Perspectives.
89:21-25.
Lyngbye, T., Hansen, O.N., and Grandjean, P. (1990) "Predictors of Tooth Lead Level with
Special Reference to Traffic: A Study of Lead-Exposure in Children." International
Archives of Occupational and Environmental Health. 62(6):417-422.
Marais, M.L., and Wecker, W.E. (1998) "Correcting for Omitted Variables and Measurement
Error Bias in Regression with an Application to the Effect of Lead on IQ." Journal of the
American Statistical Association. 93:494-505.
McKinney, J. (1993) "Metals Bioavailability and Disposition Kinetics Research Needs
Workshop, July 18-19, 1990." Toxicological and Environmental Chemistry. 38: 1-71.
299
-------
Mello, C.F., Kraemer, C.K., Filippin, A., Morsch, V.M., Rodrigues, A.L., Martins, A.F., and
Rubin, M.A. (1998) "Effect of Lead Acetate on Neurobehavioral Development of Rats."
Brazilian Journal of Medical and Biological Research. 31(7): 943-950.
Mielke, H. W. (1995) "Lead in New Orleans Soils: New Images of an Urban Environment."
Environmental Geochemistry and Health. 16(3/4): 123-128.
Mielke, H. W. (1993) "Lead Dust Contaminated U.S.A. Communities: Comparison of Louisiana
and Minnesota." Applied Geochemistry. Supp. Issue No. 2,257-261.
Mielke, H. W., Adams, J. E., Huff, B., Pepersack, J., Reagan, P. L., Stoppel, D., and Mielke, P.
W. (1992) "Dust Control as a Means of Reducing Inner-City Childhood Pb Exposure." In:
Trace Substances in Environmental Health - XXV. Supplement to Volume 14,
Environmental Geochemistry and Health (B. D. Beck, ed.) 121-128.
Mielke, H. W., Adams, J. L., Reagan, P. L., and Mielke, P. W. (1989) "Soil-Dust Lead and
Childhood Lead Exposure as a Function of City Size and Community Traffic Flow: The
Case for Lead Abatement in Minnesota." In: Lead in Soil: Issues and Guidelines,
Supplement to Volume 9 of Environmental Geochemistry and Health (B. E. Davis and B. G.
Wixson, eds.). 253-271.
Mielke, H. W., Anderson, J.C., Berry, K. J., Mielke, P. W., Chaney, R. L., and Leech, M. (1983)
"Lead Concentrations in Inner-City Soils as a Factor in the Child Lead Problem." American
Journal of Public Health. 73(12): 1366-1369.
Nagymajtenyi, L., Desi, L, Schulz, H., and Papp, A. (1998) "Consequences of Lead Exposure of
Rats During Pregnancy, Lactation, and Postweaning. A Combined Behavioural and
Neurotoxicological Study." International Journal of Environmental Health Research.
8(2): 121-135.
Needleman, H.L. (1998) "IQ, Lead Level, and Inferences from Research Studies." Comments to
the Environmental Health Committee of the Science Advisory Board, U.S. Environmental
Protection Agency, in their technical review of the proposed TSC A Section 403 regulations
(Identification of Dangerous Levels of Lead). 2 September 1998.
Needleman, H.L., and Gatsonis, C.A. (1990) "Low-Level Lead Exposure and the IQ of
Children." Journal of the American Medical Association. 263(5):673-678.
Otto, D.A., and Fox, D.A. (1993) "Auditory and Visual Dysfunction Following Lead Exposure."
Neurotoxicology. 14(2-3): 191-208.
Petit, T.L., Alfano, D.P., and LeBoutillier, J.C. (1983) "Early Lead Exposure and the
Hippocampus: a Review and Recent Advances." Neurotoxicology. 4(l):79-94.
300
-------
Pendleton, S. (personal communication to Brad Schultz, U.S. EPA).
Pocock, S.J., Smith, M., and Baghurst, P. (1994) "Environmental Lead and Children's
Intelligence: A Systematic Review of the Epidemiological Evidence." BMJ. 309:1189-
1197.
Que Hee, S., Peace, B., Clark, S., Boyle, J. R., Bomschein, R. L., and Hammond, P. B. (1985)
"Evolution of Efficient Methods to Sample Lead Sources, Such as House Dust and Hand
Dust, in the Homes of Children." Environmental Research. 38:77-95.
Rabinowitz, M.B., Wang, J.D., and Soong, W.T. (1992) "Apparent Threshold of Lead's Effect
on Child Intelligence." Bulletin of Environmental Contamination and Toxicology. 48:688-
695.
Rabinowitz, M., Leviton, A., Needleman, H., Bellinger, D., and Waternaux, C. (1985)
"Environmental Correlates of Infant Blood Lead Levels in Boston." Environmental
Research. 38:96-107.
Rice, D. C. (1996) "Behavioral Effects of Lead: Commonalities Between Experimental and
Epidemiologic Data." Environmental Health Perspectives. 104 (Suppl. 2): 337-351.
Rice, D.C. (1988) "Chronic Low-Lead Exposure Does Not Affect Simple Reaction Time."
Neurotoxicology. 9:105-108.
Rice, D.C. (1985) "Chronic Low-Lead Exposure from Birth Produces Deficits in Discrimination
Reversal in Monkeys." Toxicology and Applied Pharmacology. 77:201-210.
Rice, D.C. and Gilbert, S.G, (1990a) "Sensitive Periods for Lead-Induced Behavioral
Impairment (Nonspatial Discrimination Reversal) in Monkeys." Toxicology and Applied
Pharmacology. 102:101-109.
Rice, D. C. and Gilbert, S.G. (1990b) "Lack of Sensitive Period for Lead-Induced Behavioral
Impairment on a Spatial Delayed Alternation Task in Monkeys." Toxicology and Applied
Pharmacology. 103:364-373.
Rieuwerts, J.S., and Farago, M.E. (1995) "Lead Contamination in Smelting and Mining
Environments and Variations in Chemical Forms and Bioavailability." Chemical
Speciation and Bioavailability. 7(4): 113-123.
Roberts, J.W., Han, W., and Ruby, M.G. (1996) "Project Summary: Evaluation of Dust
Samplers for Bare Floors and Upholstery." National Exposure Laboratory, Office of
Research and Development, U.S. Environmental Protection Agency. EPA/600/R-96-001.
January 1996.
301
-------
Roels, H.A., Buchet, J.P., Lauwerys, R.R., Bruaux, P., Claeys-Thoreau, F., LaFontaine, A., and
Verduyn, G. (1980) "Exposure to Lead by the Oral and the Pulmonary Routes of Children
Living in the Vicinity of a Primary Lead Smelter." Environmental Research. 22:81-94.
Romieu, I., Carreon, T., Lopez, L., Palazuelos, E., Rios, C., Manuel, Y., and Hemandez-Avila,
M. (1995) "Environmental Urban Lead Exposure and Blood Lead Levels in Children of
Mexico City." Environmental Health Perspectives. 103(11): 1036-1040.
Schmitt, M.D.C., Trippler, D.J., Wachtler, J.N. and Lund, G.V. (1988) "Soil Lead
Concentrations in Residential Minnesota as Measured by ICP-AES." Water, Air, and Soil
Pollution. 39:157-168.
Schroeder, S.R., Hawk, B., Otto, D., Mushak, P., and Hicks, R.E. (1985) "Separating the Effects
of Lead and Social Factors on IQ." Environmental Research. 91:178-183.
Schwartz, J. (1994) "Low-Level Lead Exposure and Children's IQ: A Meta-Analysis and Search
for a Threshold." Environmental Research. 65:42-55.
Schwartz, J. (1993) "Beyond LOEL's, p Values, and Vote Counting: Methods for Looking at the
Shapes and Strengths of Association." Neurotoxicology and Teratology. 14(2-3):237-246.
Sedman, R.M. (1989) "The Development of Applied Action Levels for Soils Contact: A
Scenario for the Exposure of Humans to Soils in a Residential Setting." Environmental
Health Perspectives. 79:291-313.
Sedman, R.M., and Mahmood, RJ. (1994) "Soil Ingestion by Children and Adults Reconsidered
Using the Results of Recent Tracer Studies." Journal of Air & Waste Management
Association. 44:141-144.
Shellshear, I.D., Jordon, L.D., Hogan, D.J., and Shannon, F.T. (1975) "Environmental Lead
Exposure in Christchurch Children: Soil Lead as a Potential Hazard." New Zealand Med.
J. 81:382-386.
Sheppard, S.C. (1998) "Geophagy: Who Eats Soils and Where Do Possible Contaminants Go?"
Environmental Geology (Berlin). 33(2-3): 109-114.
Silbergeld, E.K. (1991) "Aneurological Perspective on Lead Toxicity." In Human Lead
Exposure (Needleman, H.L. ed.). Boca Raton, Florida: CRC Press, 89-103.
Silva, P.A., Hughes, P., Williams, S., and Faed, J.M. (1988) "Blood Lead, Intelligence, Reading
Attainment, and Behaviour in Eleven Year Old Children in Dunedin, New Zealand."
Journal of Child Psychology and Psychiatry. 29:43-52.
302
-------
Simon, S.L. (1998) "Soil Ingestion by Humans: A Review of History, Data, and Etiology with
Application to Risk Assessment of Radioactively Contaminated Soils." Health Physics,
74(6):647-672.
Stanek, E.J. m, and Calabrese, EJ. (1995) Daily Estimates of Soil Ingestion in Children.
Environmental Health Perspectives. 103:276-285.
Stanek E.J., Calabrese, E.J., Mundt, K., Pekow, P., and Yeatts, K.B. (1998) "Prevalence of Soil
Mouthing/Ingestion Among Healthy Children Aged 1 to 6." Journal of Soil
Contamination. 7(2):227-242.
Stark, A. D., Quah, R. F., Meigs, J. W., and DeLouise, E. R. (1982) "The Relationship of
Environmental Lead to Blood-Lead Levels in Children." Environmental Research. 27:372-
383.
Steuteville, W.D. (1990) "Interior Remediation of Residences: Lead Abatement (A Case
History of a Superfund Project and Public Involvement)." In: Procedures of the National
Conference of Hazardous Wastes and Hazardous Materials. Silver Springs, MD:
Hazardous Materials Research Control Institute.
Sutton, P. Athanasoulis, M., Ressel P., Guirguis, G., Haan, M., Schlag, R., and Goldman, L.
(1995) "Lead Levels in the Household Environment of Children in Three High-Risk
Communities in California." Environmental Research. 68:45-57.
Teichmean, J., Coltrin, D., Prouty, K., and Bir, W. A. (1993) "A Survey of Lead Contamination
in Soil Along Interstate 880, Alameda County, California." Journal of the American
Industrial Hygiene Association. 54(9): 557-559.
Tong, S. (1998) "Lead Exposure and Cognitive Development: Persistence and a Dynamic
Pattern." J. Pediatr. Child Health. 34:114-118.
Tong, S. T. (1990) "Roadside Dusts and Soils Contamination in Cincinnati, Ohio."
Environmental Management. 14(1): 107-114.
Tong, S., Baghurst, P., McMichael, A., Sawyer, M., and Mudge, J. (1996) "Lifetime Exposure to
Environmental Lead and Children's Intelligence at 11-13 Years: the Port Pirie Cohort
Study." British Medical Journal. 312:1569-1575.
U.S. Census Bureau (1999) Current Housing Reports, Series HI50/97, American Housing
Survey for the United States: 1997 (Table 2-1). U.S. Government Printing Office,
Washington, DC 20222.
303
-------
U.S. Department of Health and Human Services (1999) Toxicological Profile for Lead. Update,
1999. Atlanta: Agency for Toxic Substances and Disease Registry, Division of
Toxicology/Toxicology Information Branch, U.S. Department of Health and Human
Services.
U.S. Department of Health and Human Services (1992) Impact of Lead-Contaminated Soil on
Public Health. Atlanta: Agency for Toxic Substances and Disease Registry, Division of
Toxicology/Toxicology Information Branch, U.S. Department of Health and Human
Services. May 1,1992.
U.S. Department of Health and Human Services (1991) "Philadelphia Neighborhood Lead
Study." Final Report. Atlanta: Agency for Toxic Substances and Disease Registry,
Division of Toxicology/Toxicology Information Branch, U.S. Department of Health and
Human Services.
U.S. Department of Housing and Urban Development (1998) "Evaluation of the HUD Lead-
Based Paint Hazard Control Grant Program: Fifth Interim Report, Progress as of September
1,1997." Prepared by the National Center for Lead Safe Housing and the University of
Cincinnati Department of Environmental Health for the U.S. Department of Housing and
Urban Development. March 1998.
U.S. Department of Housing and Urban Development (1995a) "The Relation of Lead-
Contaminated House Dust and Blood Lead Levels Among Urban Children." Volumes I
and H. Final Report to the U.S. Department of Housing and Urban Development Grant
from the Departments of Pediatrics, Biostatistics, and Environmental Medicine, The
Rochester School of Medicine, Rochester, New York, and The National Center for Lead-
Safe Housing, Columbia Maryland.
U.S. Department of Housing and Urban Development (1995b) "Guidelines for the Evaluation
and Control of Lead-Based Paint Hazards in Housing." Office of Lead-Based Paint
Abatement and Poisoning Prevention.
U.S. Department of Housing and Urban Development (1991) "The HUD Lead-Based Paint
Abatement Demonstration (FHA)." Final Report, Office of Policy Development and
Research, U.S. Department of Housing and Urban Development.
U.S. Environmental Protection Agency (1998a) "Risk Analysis to Support Standards for Lead in
Paint, Dust, and Soil." Office of Pollution Prevention and Toxics, U.S. Environmental
Protection Agency. EPA 747-R-97-006.
U.S. Environmental Protection Agency (1998b) "An SAB Report: Technical Review of the
Proposed TSCA Section 403 Regulation (Identification of Dangerous Levels of Lead)."
Environmental Health Committee, Science Advisory Board, U.S. Environmental Protection
Agency. EPA-SAB-EHC-99-004, November 1998.
304
-------
U.S. Environmental Protection Agency (1998c) "Analysis of Pathways of Residential Lead
Exposure in Children." Office of Prevention Pesticides, and Toxic Substances. EPA 747-
R-98-007. 27 April 1998 (peer review version).
U.S. Environmental Protection Agency (1997a) "Lead Exposure Associated with Renovation
and Remodeling Activities: Environmental Field Sampling Study. Volume I: Technical
Report." Office of Pollution Prevention and Toxics, U.S. Environmental Protection
Agency. EPA 747-R-96-007.
U.S. Environmental Protection Agency (1997b) "Summary and Assessment of Published
Information on Determining Lead Exposures and Mitigating Lead Hazards Associated with
Dust and Soil in Residential Carpets, Furniture, and Forced Air Ducts." Office of
Prevention Pesticides, and Toxic Substances, U.S. Environmental Protection Agency. EPA
747-S-97-001.
U.S. Environmental Protection Agency (1997c) "Lead-Based Paint Abatement and Repair and
Maintenance Study in Baltimore: Finding Based on Two Years of Follow-Up." Office of
Prevention Pesticides, and Toxic Substances, U.S. Environmental Protection Agency. EPA
747-R-97-005.
U.S. Environmental Protection Agency (1996a) "Urban Soil Lead Abatement Demonstration
Project, Volume I: Integrated Report." Office of Research and Development, U.S.
Environmental Protection Agency. EPA 600-P-93-001aF, April 1996.
U.S. Environmental Protection Agency (1996b) "Comprehensive Abatement Performance
Study. Volume I: Summary Report." Office of Prevention, Pesticides, and Toxic
Substances, U.S. Environmental Protection Agency. EPA 230-R-94-013a.
U. S. Environmental Protection Agency (1996c) "Lead-Based Paint Abatement and Repair and
Maintenance Study in Baltimore: Pre-Intervention Findings." EPA 747-R-95-012, August
1996.
U.S. Environmental Protection Agency (1995) "Report on the National Survey of Lead-Based
Paint in Housing. Base Report." Office of Pollution Prevention and Toxics, U.S.
Environmental Protection Agency. EPA747-R95-003.
U.S. Environmental Protection Agency (1994) "Guidance Manual for the Integrated Exposure
Uptake Biokinetic Model for Lead in Children." Office of Emergency and Remedial
Response, U.S. Environmental Protection Agency. EPA/540/R-93/081, PB93-963510.
U.S. Environmental Protection Agency (1986) "Air Quality Criteria for Lead, Volume IV of IV."
Research Triangle Park NC: Environmental Criteria and Assessment Office, U.S.
Environmental Protection Agency. EPA-600/8-83/028dF. June 1986.
305
-------
Viverette, L., Mielke, H.W., Brisco, M, Dixon, A., Schaefer, J., and Pierre, K. (1996)
"Environmental Health in Minority and Other Underserved Populations: Benign Methods
for Identifying Lead Hazards at Day Care Centres of New Orleans." Environmental
Geochemistry and Health. 18(1): 41-45.
Walsh, T.J., and Tilson, H.A. (1984) "Neurobehavioral Toxicology of the Organoleads."
Neurotoxicology. 5:67-86.
Wang, L., Xu S.E., Zhang, G.D, and Wang, W.Y. (1989) "Study of Lead Absorption and its
Effect on Children's Development." Biomedical and Environmental Sciences. 2:325-330.
Weitzman, M., Aschengrau, A., Bellinger, D., Jones, R., Hamlin, J.S., and Beiser, A. (1993)
"Lead-Contaminated Soil Abatement and Urban Children's Blood Lead Levels." Journal of
the American Medical Association. 269(13): 1647-1654.
Westat (1998) "A Report on the Analysis of the Blood Lead-Environmental Lead Relationship."
Report for the U.S. Department of Housing and Urban Development. 7 August 1998.
Winneke, G., Lilienthal, H., and Kramer, U. (1996) "The Neurobehavioural Toxicology and
Teratology of Lead." Arch Toxicol Suppl. 18:57-70.
Winneke, G., Brockhaus, A., Ewers, V., Kramer, V., and Neuf, M. (1990) "Results from the
European Multicenter Study on Lead Neurotoxicity in Gulden: Implications for Risk
Assessment." Neurotoxicol Teratol 12:553-559.
Winneke, G., Beginn, U., Ewert, T., et al. (1985) "Comparing the Effects of Perinatal and Later
Childhood Lead Exposure on Neurophysiological Outcome." Environmental Research.
38:155-167.
Yule, W., Lansdown, R., Millar, L, and Urbanowiez, M. (1981) "The Relationship Between
Blood Lead Concentration, Intelligence, and Attainment in a School Population: a Pilot
Study." DevMed Child Neuro. 23:567-576.
306
-------
50272-101
REPORT DOCUMENTATION
PAGE
1. REPORT NO.
EPA 747-R-00-004
3. Recipient's Accession No.
4. Title and Subtitle
Risk Analysis to Support Standards for Lead in Paint, Dust, and Soil: Supplemental Report
5. Report Date
December 2000
6.
7. Author(s) Lordo, RA, Burgoon, DA, Ma, ZJ, Matthews, MC, Menkedick, JG, Naber, S, Nahhas,
R, Neighbor, M, Niemuth, NA, Wood, BJ, Wooton, M
8. Performing Organization
Rept. No.
9. Performing Organization Name and Address
Battelle Memorial Institute
505 King Avenue
Columbus, Ohio 43201-2693
10. ProjecVTask/Work Unit No.
0476101-20
11. Contract(C) or Grant(G) No,
(C| 68-W-99-033
12. Sponsoring Organization Name and Address
U.S. Environmental Protection Agency
Office of Pollution Prevention and Toxics
401 M Street, S.W.
Washington, D.C. 20460
13. Type of Report & Period
Covered
Technical Report
14.
15. Supplementary Notes
Support staff involved in the risk analysis and the production of this report included Haisey Boyd, Ying-Liang Chou. and Kensuke
Shirakawa.
16. Abstract (Limit 200 words)
Title X of the Housing and Community Development Act, known as the Residential Lead-Based Paint Hazard Reduction Act of 1992,
contains legislation designed to evaluate and reduce exposures to lead in paint, dust, and soil in the nation's housing. As part of Title X,
the Toxic Substances Control Act (TSCA) was amended to include Title IV, 'Lead Exposure Reduction'. Section 403 of TSCA requires
EPA to define standards for lead in paint, dust, and soil. Federal, state, and local public health agencies, as well as private property
owners and other private sector interests, will use these standards to determine in which homes actions should be taken to reduce or
prevent the threat of childhood lead poisoning.
This report is a supplement to EPA 747-R-97-006 (June 1998), which documented the outcome of a risk analysis that supported EPA's
efforts to propose standards in response to Section 403. The additional literature reviews, data summaries, and data analyses presented
in this report focus on selected topics raised in comments on the risk analysis that were made by EPA's Science Advisory Board and the
general public. EPA has cited the findings of this report when preparing responses to these comments.
17. Document Analysis
a. Descriptors
Blood-lead concentration, childhood health risks, adverse health effects, dust-lead loading, soil-lead concentration, lead hazards,
lead-based paint, risk assessment, risk characterization, risk management
b. Identifiers/Open-Ended Terms
abatement, intervention, hazard identification, exposure assessment, dose-response assessment, HUD National Survey,
Rochester Lead-in-Dust Study, Evaluation of the HUD Lead-Based Paint Hazard Control Grant Program, NHANES III,
bioavailability, modeling, performance characteristics analysis, sensitivity and uncertainty analysis. Section 403, Title X
C. COSATl Field/Group
18. Availability Statement
Release Unlimited
19. Security Class (This Report)
Unclassified
2O. Security Cfass (This Page)
Unclassified
21 . No. of Pa
340 (Vol.
ges
I); 232 (Vol. II)
22. Price
(See ANSI-239.18)
OPTIONAL FORM 272 (4-77)
(Formerly NTIS-35)
Department of Commerce
-------
-------