United States
Environment Protection
Agency
Office of Pollution
Prevention and Toxics
EPA 747-R-97-001
August 1997
Lead-Based Paint
Abatement and Repair
and Maintenance Study
in Baltimore:
Findings Based on the
First Year of Follow-up
-------
EPA 747-R-97-001
August 1997
LEAD-BASED PAINT ABATEMENT AND
REPAIR AND MAINTENANCE STUDY
IN BALTIMORE:
FINDINGS BASED ON THE FIRST YEAR OF FOLLOW-UP
Technical Branch
National Program Chemicals Division
Office of Pollution Prevention and Toxics
Office of Prevention, Pesticides, and Toxic Substances
U.S. Environmental Protection Agency
401 M Street, SW
Washington, DC 20460
Recycled/Recyclable • Printed with Vegetable Based Inks on Recycled Paper (20% Postconsumer)
-------
DISCLAIMER
The material in this document has been subject to Agency technical and policy review and
approved for publication as an EPA report. Mention of trade names, products, or services,
does not convey, and should not be interpreted as conveying, official EPA approval,
endorsement, or recommendation.
11
-------
CONTRIBUTING ORGANIZATIONS AND ACKNOWLEDGMENTS
The study described in this report was funded and managed by the U.S. Environmental
Protection Agency (EPA) and conducted collaboratively as described below.
Kennedy Krieger Research Institute (XKRI)
KKRI was responsible for the overall design and conduct of this study, including the field,
laboratory and data analysis activities, and the preparation of this report. The KKRI investigators
were Mark R. Farfel, Sc.D., Project Director and J. Julian Chisolm, Jr., M.D. The Johns
Hopkins University co-investigators were Peter S.J. Lees, Ph.D., Department of Environmental
Health Sciences, and Charles Rohde, Ph.D., Department of Biostatistics. Study staff included
William Derbyshire, Project Manager; Brian Rooney, Data Analyst; Desmond Bannon, Trace
Metals Laboratory Supervisor; Pat Tracey, Outreach Coordinator; and Ken Watts, R&M QC
Officer. Field staff were Eula Kemmer, Earnestine Powell, Tammy Smith, and Marc Talley.
Laboratory staff included Mike Burns, Mavis Harby, Lori Losh, Catherine Murashchik, and
Becky Zapf.
Special acknowledgment is given to the numerous collaborating organizations and individuals
including Battelle Memorial Institute and Midwest Research Institute for technical and
administrative support during the planning and pilot phases of the study, Maryland Department
of the Environment, Baltimore City Departments of Health and Housing and Community
Development, City Homes, Inc., Paul Constant, Gary Dewalt, Jack Hirsch, Susan Guyaux,
Michael and Susan Kleinhammer, Patrick Connor, Barry Mankowitz, Clark McNutt, Ron
Menton,Vance Morris, Charlotte Pinning, Ruth Quinn, Marge Sheehan, Mary Snyder-Vogel,
Jennifer Steciak, Amy Spanier, and participating property owners.
Maryland Department of Housing and Community Development
This agency reserved and administered loan funds from a special residential lead paint abatement
loan program to finance the Repair & Maintenance interventions performed in this study.
US Environmental Protection Agency
The U.S. Environmental Protection Agency (EPA) was responsible for managing the study, for
providing technical oversight, guidance and direction, and for overseeing the peer review and
finalization of the report. The EPA Project Leader was Benjamin S. Lim. The EPA Work
Assignment Managers were Benjamin S. Lim and Brad Schultz. The EPA Project Officers were
Phil Robinson and Jill Hacker. Cindy Stroup was the Branch Chief of the Technical Programs
Branch (TPB) who initiated this study and provided valuable input. Special Acknowledgment is
given to Darlene Watford, the Acting Methods Section Chief, for her careful review and input.
111
-------
CONTENTS
EXECUTIVE SUMMARY vii
1.0 INTRODUCTION 1
1.1 Report Objectives 4
1.2 Purpose of the R&M Study 4
1.3 Peer Review 5
2.0 SUMMARY AND DISCUSSION OF FINDINGS 7
3.0 QUALITY ASSURANCE 15
3.1 System Audit 15
3.2 Data Audit 15
3.3 Performance Audit 16
4.0 STUDY DESIGN AND SAMPLE COLLECTION PROCEDURES 21
4.1 Overview Of Study Design 21
4.2 Repair & Maintenance Interventions and Comprehensive Abatement 23
4.3 Recruitment And Enrollment 24
4.4 Selection Criteria For Houses And Children 25
4.5 Characteristics Of Study Houses And Participants 27
4.6 Sample And Data Collection Procedures 28
5.0 LABORATORY ANALYSIS PROCEDURES 30
6.0 DATA PROCESSING AND STATISTICAL ANALYSIS PROCEDURES 31
6.1 Data Processing 31
6.2 Data Summary 31
6.3 Statistical Analysis 34
7.0 RESULTS 47
7.1 Descriptive Statistics For The First Year Of Follow-Up 47
7.2 Descriptive Statistics At The 12-Month Campaign 48
7.3 Longitudinal Data Analysis 81
8.0 REFERENCES 92
APPENDIX A: Descriptive Statistics for Dust Lead and Dust Loadings
APPENDIX B: Descriptive Statistics for Baseline Blood Lead Concentrations
IV
-------
TABLES
Table 1: Comparison of Elements of Repair & Maintenance Levels I - III 2
Table 2: Descriptive Statistics And Tolerance Limits For Percent Recovery For SRM
And Spiked Samples And Percent Differences Between Spike And Spike
Duplicate Samples 18
Table 3: Descriptive Statistics And Tolerance Limits For Percent Recovery For ICV
And CCV 19
Table 4: Descriptive Statistics For Field Blanks And Method Blanks 20
Table 5: Data Collection Plan 22
Table 6: Types Of Field Samples 29
Table 7: Summary Of Laboratory Procedures 30
Table 8: Types And Numbers Of Samples Collected And Analyzed For Lead
(Excluding QC Samples) As A Part Of The 12-Month Campaign 32
Table 9: Types And Numbers Of Samples Collected By Group As A Part Of The
12-Month Campaign 33
Table 10: Family Moves, Reoccupancies, And New Subjects Enrolled Between The
Initial Campaign And The 12-Month Campaign 35
Table 11: Variability Accounted for by Factor Loadings Across Campaigns 40
Table 12: Factor Patterns For The Five Study Groups Across Campaigns 41
Table 13: Factor Patterns For R&M Groups Across Campaigns 42
Table 14: Definitions of Variables 44
Table 15: Numbers of Children With Initial Blood Lead <20/*g/dL and >20jtg/dL ... 46
Table 16: Descriptive Statistics For Blood Lead Concentrations By Group At The
12-Month Campaign 49
Table 17: Descriptive Statistics For Soil Lead Concentrations By Study Group At
The Six-Month Campaign 75
Table 18: Descriptive Statistics For Water Lead Concentrations By Study Group At
The Six-Month Campaign 75
Table 19: Correlations Between Dust Lead Concentrations At The 12-Month Campaign
76
Table 20: Correlations Between Dust Lead Loadings At The 12-Month Campaign ... 77
Table 21: Correlations Between Dust Loadings At The 12-Month Campaign 78
Table 22: Correlations Between Blood Lead and Dust Measures Using The Youngest
Child Per Household In Continuing Houses At The 12-Month Campaign ... 79
Table 23: Correlations Between Blood Lead and Dust Measures Using All Children
Per Household In Continuing Houses At The 12-Month Campaign 80
Table 24: Predicted Blood Lead Concentration (PbB, /xg/dL) By Group And By
Campaign In Children With Initial PbB < 20 jwg/dL 84
Table A-l: Descriptive Statistics For Dust Lead Concentrations By Surface Type And
Study Group At The 12-Month Campaign 97
Table A-2: Descriptive Statistics For Dust Lead Loadings By Surface Type And Study
Group At The 12-Month Campaign 98
Table A-3: Descriptive Statistics For Dust Loadings By Surface Type And Study Group
At The 12-Month Campaign 99
-------
FIGURES
Boxplots:
Figure 1: Dust Lead Loadings Across Campaigns By Group For Floors 50
Figure 2: Dust Lead Loadings Across Campaigns By Group For Window Sills 51
Figure 3: Dust Lead Loadings Across Campaigns By Group For Window Wells .... 52
Figure 4: Dust Lead Loadings Across Campaigns By Group For Interior Entryways . . 53
Figure 5: Dust Lead Concentrations Across Campaigns By Group For Floors 54
Figure 6: Dust Lead Concentrations Across Campaigns By Group For Window Sills . 55
Figure 7: Dust Lead Concentrations Across Campaigns By Group For Window Wells . 56
Figure 8: Dust Lead Concentrations Across Campaigns By Group For Interior
Entryways 57
Figure 9: Dust Loadings Across Campaigns By Group For Floors 58
Figure 10: Dust Loadings Across Campaigns By Group For Window Sills 59
Figure 11: Dust Loadings Across Campaigns By Group For Window Wells 60
Figure 12: Dust Loadings Across Campaigns By Group For Interior Entryways 61
Figure 13: Blood Lead Across Campaigns By Group For Children With Initial Blood
Lead Concentrations < 20 ug/dL 62
Plots:
Figure 14: Children's Blood Lead Concentrations Across Time - R&M I 63
Figure 15: Children's Blood Lead Concentrations Across Time ~ R&M II 64
Figure 16: Children's Blood Lead Concentrations Across Time - R&M III 65
Figure 17: Children's Blood Lead Concentrations Across Time — Modern Urban .... 66
Figure 18: Children's Blood Lead Concentrations Across Time - Previously Abated . . 67
Bar Graphs:
Figure 19: Dust Lead Loadings At 12 Months By Surface Type And By Group 68
Figure 20: Dust Lead Concentrations At 12 Months By Surface Type And Group .... 69
Figure 21: Dust Loadings At 12 Months By Surface Type And By Group 70
Figure 22: Overall Lead Levels And Dust Loadings By Group At 12 Months 73
Plots based on Longitudinal Data Analysis
Figure 23: Environmental Model Least Square Means — R&M Groups 87
Figure 24: Environmental Model Least Square Means - All Five Study Groups 88
Figure 25: Comparison Model Predicted Blood Lead Levels (Initial PbB < 20/^g/dL) . . 89
Figure 26: Comparison Model Predicted Blood Lead Levels (Initial PbB;>20yUg/dL) . . 90
Figure 27: Exposure Model Adjusted Residual Plot Of Factor 1 Dust Lead
Versus Blood Lead 91
VI
-------
EXECUTIVE SUMMARY
In recent years, there has been growing interest in the use of interim measures to
temporarily control the problem of extensive residential lead-based paint hazards in U.S. housing
in a cost-effective manner. Title X of the Housing and Community Development Act of 1992 (P.L.
102-550) defined interim controls as " a set of measures designed to reduce temporarily human
exposure or likely exposure to lead-based paint hazards, including specialized cleaning, repairs,
maintenance, painting, temporary containment, ongoing monitoring of lead-based paint hazards
or potential hazards and the establishment of management and resident education programs." The
1995 HUD Guidelines for the Evaluation and Control of Lead-Based Paint Hazards in Housing
provide detailed information on interim control practices. However, little is known about the
short-and long-term effectiveness of these approaches in terms of reducing lead in dust and in
children's blood.
This report presents the first year of follow-up of the Lead-Based Paint Abatement and
Repair & Maintenance (R&M) Study in Baltimore. The study is designed to characterize and
compare the short-term (two to six months) and longer-term (12-24 months) effectiveness of three
levels of interim control interventions (R&M I-III) in low-income housing where children are at
high risk of exposure to lead in dust and paint. The study has two control groups — urban houses
built after 1979, and presumably free of lead-based paint, and previously abated houses that
received comprehensive abatement in the past. The study population consists of non-Hispanic
black households with at least one participating child. At the outset, mean ages of study children
ranged from 25 months to 33 months across groups, and their geometric mean blood lead
concentrations were 10 /*g/dL in R&M I, 14 /-tg/dL R&M II, 14 pig/dL in R&M m, and 13 /xg/dL
in the previously abated houses. The geometric mean blood lead concentration in children in the
modern urban houses was 5 pig/dL, a value slightly above the geometric mean of 2.7 ug/dL for all
U.S. children aged 12 to 60 months.1
During the first year of follow-up, objectives related to enrollment, laboratory
performance, data quality and data completeness were met. Furthermore, families were informed
by letter of the results of dust lead and blood lead tests from each campaign. For this reason, the
study intervention was a combination of R&M work and the provision of information to families
on a periodic basis. The main findings based on dust lead and blood lead data from the five study
groups collected during the pre- and post-intervention campaigns, as well as during the two, six
and 12 months post-intervention data collection campaigns are summarized below.
• All three levels of R&M intervention were associated with statistically significant
reductions in house dust lead loadings and in total dust loadings. These loadings were
sustained below pre-intervention levels during the first year of follow-up. Dust lead
concentrations were significantly reduced following intervention in the middle (R&M H)
and high (R&M IE) intervention houses, but not in the low intervention houses (R&M I).
• The dust lead loadings, lead concentrations, and dust loadings during the first year of
follow-up were related to the intensity of the intervention. Immediately following
intervention and at two months, six months, and 12 months post-intervention, dust lead
loadings, lead concentrations and dust loadings were lowest in R&M HI houses,
Vll
-------
intermediate in R&M II houses, and highest in R&M I houses. Statistically significant
differences were found between R&M groups on these dust measures over time. For
example, at 12 months post-intervention, weighted average dust lead loading estimates
were 21-fold higher in R&M I houses than in R&M III houses, and five-fold higher in
R&M I houses than hi R&M II houses.
• The modern urban control group had significantly lower dust lead loadings and
concentrations across time than the other four groups. These houses, located in clusters
of urban houses built after 1979, were expected to reflect the lowest residential and
ambient lead levels in the urban environment. Low dust lead concentrations (geometric
mean <310 uglg, equivalent to ^0.03 percent) and soil lead concentrations (geometric
mean <,15 /ig/g) support the presumption the these houses were free of lead-based paints.
Dust lead levels in the previously abated control houses three years to five years post-
abatement were generally intermediate between those in R&M II and R&M ni houses at
the end of the first year of follow-up.
• At the end of the first year, the unadjusted geometric mean blood lead concentrations were
lower for each group - 8 /ig/dL in R&M I, 11 /ig/dL in R&M H, and 12 ^ig/dL in R&M
ffl, 12 /xg/dL in previously abated, and 3 /-ig/dL in modern urban. Children in the modern
urban group had significantly lower blood lead concentrations over tune, compared with
each of the other four groups; their blood lead concentrations were < 10 jUg/dL, the
Center for Disease Control's level of concern.
• Using all five study groups in the longitudinal data analysis, a statistically significant
relationship was found between a composite measure of house dust lead in an entire house
(both concentration and loading) and children's blood lead concentration, controlling for
covariates including age and season.
• Children with pre-intervention blood lead concentrations >20 /ig/dL had statistically
significant reductions in blood lead concentration during follow-up, controlling for age and
season. Statistically significant blood lead changes were not found in children hi the three
R&M groups with pre-intervention blood lead concentrations < 20 /*g/dL, again controlling
for age and season. Cumulative body lead burden and neighborhood housing characteristics
are discussed as two factors that may have mediated children's blood lead responses to the
R&M interventions and contributed to the differences in blood lead concentrations
observed between children in the modern urban group and those in the other four groups.
The next report will investigate changes in blood lead and dust lead during the second year of
follow-up. It should be emphasized that the R&M interventions under investigation are interim
control or partial abatement approaches to reducing lead-based paint hazards. As such, they are
not expected to be as long-lasting as comprehensive abatement. During the first year of follow-up,
none of the interventions in individual houses failed, that is, all or most of the dust samples
showed lead loadings at, or below, pre-intervention levels. Thus, a major study objective with
important policy implications remains the documentation of the longevity of the R&M
interventions. It is also important to note that the costs of the interventions in this project may not
be generalizable to other settings and time periods.
VlILl
-------
1.0 INTRODUCTION
This report presents the results of the first year of follow-up in the Lead-Based Paint
Abatement and Repair & Maintenance (R&M) study in Baltimore, conducted by the Kennedy
Krieger Research Institute. The study is a longitudinal trial of housing interventions designed to
reduce children's exposure to lead in paint and settled dust in their homes.2 Baseline demographic,
environmental, and biological data were reported previously for the five groups of houses and
residents studied, which included houses designated for R&M intervention levels I through HI,
modern urban control houses built after 1979, and previously abated control houses that had
received comprehensive abatement.3 This document represents the first report on changes in lead
levels in settled house dust and children's blood associated with the three levels of interim control
interventions under investigation (R&M I-in, Table 1). These interventions and the comprehensive
abatements are described in section 4.2. The next report will include findings from the second
year of follow-up.
At baseline, the study population consisted of non-Hispanic black households (140 children
in 107 houses) with low-to-moderate monthly rents or mortgages who resided in city rowhouses.
Mean ages of children studied ranged from 25 months to 33 months across the groups. Initial
geometric mean blood lead concentrations were 10 jttg/dL in the R&M I group, and 14 ptg/dL hi
both the R&M n and R&M HI groups, and 5 j*g/dL hi the modern urban group and 13 /ig/dL in
the previously abated group. Baseline blood lead concentrations in the modern urban group were
statistically lower than baseline levels hi the other four groups. Further, children's blood lead
concentrations were correlated significantly (r=.28 to .64) with measures of lead in dust from six
types of ulterior house surfaces and exterior entry ways.
Houses in all study groups were generally similar hi terms of characteristics that might be
expected to influence patterns of dust movement into and within a house, including overall size,
number of windows, house type and design, condition, degree of setback from the street, and the
presence of porches and yards. Statistically significant differences were not found hi demographic
characteristics, children's blood lead concentrations, and dust lead concentrations between R&M
groups at baseline. However, baseline dust lead loadings tended to be highest hi R&M HI houses
(vacant at baseline), lowest hi R&M I houses (occupied at baseline), and intermediate in R&M n
houses (a mix of vacant and occupied houses). Baseline weighted average lead loadings within
an entire house were 16,600 jig/ft2 hi R&M I houses, 24,000 pig/ft2 hi R&M H houses, and
47,500 jig/ft2 in R&M in houses, compared to 83 jug/ft2 in the modern urban houses. Similar
weighted average measures of baseline dust lead concentrations were nearly two orders of
magnitude higher hi R&M houses (19,000 /ig/g hi level I; 14,400 /*g/g hi level II; and 17,500
/xg/g in level ffl) than in modern urban houses (235 /^g/g). Previously abated houses had
intermediate dust lead concentrations of 2,400 ptg/g and lead loadings of 900 /xg/ff. The baseline
data collection campaign in the previously abated houses represents a point in time two years to
four years post-abatement. In these houses, the geometric mean lead loadings for floors and
window sills, but not window wells, were found to be at or below the interim clearance standards
set by the U.S. Department of Housing and Urban Development (HUD)* and the clearance
standards guidance published by the U.S. Environmental Protection Agency (EPA/.
-------
Table 1: Comparison of Elements of Repair & Maintenance Levels I - III
ELEMENT OF
INTERVENTION
R&MLEVELI
R & M LEVEL II
R & M LEVEL HI
TESTING
Test for the presence of lead-based paint
(LBP) on interior and exterior surfaces.
Use results to develop the R&M Plan.
Test for the presence of lead-based paint (LBP)
on interior and exterior surfaces.
Use results to develop the R&M Plan.
Test for the presence of lead- based paint
(LBP) on interior and exterior surfaces.
Use results to develop the R&M Plan.
FLOOR TREATMENTS
Pace textured walk-off mat at main
entry way.
If LBP, provide floor covering. If not LBP,
seal floors to make smooth and cleanable
surfaces. Place textured walk-off mat at main
entry way. In occupied units, treat floors to
extent possible.
If LBP, provide floor covering. If not LBP,
make floors smooth and cleanable with
combination of coverings and sealants. Place
textured walk-off mats at main entry way.
TRIM COMPONENT
TREATMENTS
Remove loose and peeling LBP on interior
surfaces, and on exterior surfaces to limit of
budget. Repaint treated components.
Remove loose and peeling LBP on interior
surfaces, and on exterior surfaces to limit of
budget. Repaint treated components. If not
LBP, make interior surfaces smooth and
cleanable.
Seal, encapsulate, or enclose LBP on interior
and exterior surfaces. If not LBP, make
interior surfaces smooth and cleanable.
STAIRWAY TREATMENTS
None
If LBP present, encapsulate treads and risers, at
minimum. If not LBP, make smooth and
cleanable.
If LBP present, enclose treads and risers using
durable materials. If not LBP, make smooth
and cleanable.
WINDOW TREATMENTS
Install aluminum cap on window wells.
Prepare and repaint all exterior window
trim. Repaint interior stool with non-flat
paint.
If LBP present, treat in-place to reduce friction.
Stabilize paint on exterior trim. Install
aluminum caps on wells. Repaint interior sill
with non-flat paint. If not LBP, make smooth
and cleanable.
If LBP present, replace window and abate
exterior window trim by enclosing with
aluminum coverings. If not LBP, make
smooth and cleanable.
DOOR TREATMENTS
Same as TRIM COMPONENT
TREATMENTS.
If LBP, rework interior and exterior doors to
reduce friction. Remove peeling LBP paint and
stabilize exterior door trim. Repaint treated
surfaces. If not LBP, make smooth and
cleanable.
If LBP, rework interior and exterior doors to
reduce friction or replace. Remove peeling
paint. If not LBP, make smooth and cleanable.
Enclose LBP on exterior door trim with
aluminum coverings.
-------
Table 1: Comparison of Elements of Repair and Maintenance Levels I - III
(Continued)
ELEMENT OF yfe ;•<- ;
INTERVENTION
WALL TREATMENTS
FINAL CLEAN-UP
CLEANING KITS
EDUCATION
R & M LEVEL I
Same as TRIM COMPONENT
TREATMENTS.
HEPA vacuum all horizontal surfaces and
window components (ceilings excluded). Then
wet clean horizontal surfaces.
Provide cleaning kits to occupants for use after
R&M work is completed.
Provide educational materials about lead
poisoning to occupants.
R&M LEVEL O
If LBP and < 25 % of component is damaged,
repair damaged area and seal component, at a
minimum. If LBP and
>25% of component is damaged, repair
damaged area and treat by use of flexible
encapsulant or rigid enclosure.
HEPA vacuum all surfaces excluding ceilings.
Then wet clean horizontal surfaces.
Provide cleaning kits to occupants for use after
R&M work is completed.
Provide educational materials about lead
poisoning to occupants.
R&M LEVEL III
If LBP and < 25% of component is
damaged, repair damaged area and
encapsulate, at a minimum. If LBP and
>25% of component is damaged, then treat
by use of flexible encapsulant or rigid
enclosure.
HEPA vacuum all surfaces excluding
ceilings. Then wet clean horizontal surfaces.
Provide cleaning kits to occupants for use
after R&M work is completed.
Provide educational materials about lead
poisoning to occupants.
-------
1.1 Report Objectives
The primary objectives of this report are to:
• Describe lead loadings and concentrations in settled house dust for the three levels of R&M
intervention at baseline and across the four data collection campaigns conducted during the
first year of follow-up, i.e., immediate post-intervention, and two months, six months, and
12 months post-intervention.
• Describe changes in lead loadings and concentrations in settled dust between baseline and
the 12-month campaign for the control houses which consist of modern urban houses built
after 1979 and houses that received comprehensive abatement in the past.
• Apply the study's statistical models for longitudinal data analysis to the dust lead and blood
lead data.
• Report on compliance with laboratory data quality objectives.
1.2 Purpose of the R&M Study
Past studies have documented the short-term (2 months to 6 months) and longer-term (12
months or longer) effectiveness of comprehensive approaches to residential lead paint abatement
intended to attain long-term control of lead paint hazards.6'7 In recent years, there has been
growing interest in the concept of interim measures to temporarily control the extensive problem
of lead-based paint hazards in housing in a cost-effective manner. Title X of the Housing and
Community Development Act of 1992 (P.L. 102-550) defined interim controls as " a set of
measures designed to reduce temporarily human exposure or likely exposure to lead-based paint
hazards, including specialized cleaning, repairs, maintenance, painting, temporary containment,
ongoing monitoring of lead-based paint hazards or potential hazards and the establishment of
management and resident education programs." More recently, the June 1995 HUD Guidelines
for the Evaluation and Control of Lead-Based Paint Hazards in Housing operationalized the
concept by compiling information on interim control practices.4 Many believe these measures
will benefit large numbers of current and future occupants of housing with lead-based paint
hazards. However, little is known about the short- and long-term effectiveness of this approach.8
The R&M study is designed to document the short- and long-term effectiveness of a range
of housing interventions, including interim control measures, designed to reduce children's
exposure to lead in paint and in settled dust. This research is important because house dust and
residential paints containing lead have been identified as major sources of exposure in U.S.
children,9"15 Primarily via the hand-to-mouth route of ingestion3:15"18 Families with children
under seven years of age occupy approximately 10 million of the 57 million privately owned and
occupied U.S. housing units that are estimated to contain some lead-based paint.19 Children living
in the nearly 4 million houses with deteriorating paint and elevated dust lead levels are at highest
-------
risk of exposure.19 Given the extent of the problem and the high costs of remediation by
comprehensive abatement practices, the preventive R&M approach may provide a practical means
of reducing exposure for future generations of U.S. children who will continue to occupy housing
that contains lead-based paint. This study represents the first systematic examination of the R&M
approach.
The goal of the study is to contribute to the existing scientific bases needed to develop a
standard of care for lead-painted houses through the analysis of environmental and biological data
from a longitudinal intervention study. Specific study objectives are as follows:
• Measure the short- and longer-term changes in the lead concentration and lead loading of
settled house dust and changes in children's blood lead concentrations associated with the
three levels of R&M intervention (I-III), as compared to houses that had undergone
previous comprehensive abatement and to a group of modern urban houses presumed free
of lead-based paint based on their post-1979 year of construction.
• Characterize the nature of the relationship between lead in children's blood and settled
house dust.
• Evaluate and compare methodologies for the collection and analysis of lead in residential
dusts, including wipe and cyclone methods. This objective has been addressed in previous
reports.20-22
1.3 Peer Review
All four of the independent external reviewers recommended publishing the report after
minor revisions. A number of the reviewers' comments were related to the study design and the
interpretation and generalizability of study findings, potential confounders, and policy
implications. In light of this, the report places additional emphasis on the nature of the
intervention (R&M work plus feedback of information to families), its relation to generalizability
of the findings, and the limited generalizability of study data on R&M costs and lead
concentrations of soil and tap water. Soil and water were tested for lead to account for these
sources in the analysis of the longitudinal dust lead and blood lead data; the study was not intended
to answer other scientific questions about these sources. Further, we plan to request EPA funding
to test for lead-based paint in the modern urban control houses to validate our assumption that they
are free of lead-based paints. If testing is done, paint lead data will be included in subsequent
reports. Otherwise, a discussion of this assumption will be added to the next study report. Lastly,
we agree that additional data on residential dust lead loadings and dust lead concentrations in
communities across the country would increase the utility of study findings and better inform the
policy making process.
-------
To aid the reader in the interpretation of study data, a description of the comprehensive
abatements performed in the previously abated control houses was added to Section 4.2, and data
on the distributions of baseline blood lead concentrations were included in Appendix B. Further,
it is noted that the variances of baseline blood lead concentrations across the three R&M groups
were essentially the same and that housing characteristics such as degree of setback from the street
and the presence of a porch were not significant additions to the statistical models for dust lead
loadings and concentrations in the presence of season, group and campaign. Section 2.0 now
explains that the difference in average dust lead loadings between the R&M in and the previously
abated houses may be due to differences in time since intervention. The report also clarifies the
following points in response to reviewers' comments: (a) the study can not determine effect of
R&M I on the blood lead concentrations of children with initial blood lead concentrations >20
/ug/dL due to limited data, and (b) demographic data are not available to compare study households
to the small proportion of households which did not express an interest in participating. More than
90 percent of households identified as potentially eligible for the study indicated an interest in
participating.
A reviewer questioned whether the study could determine the effectiveness of R&M
interventions in preventing childhood lead exposure, as measured by blood lead concentrations,
given the ages of the children and the extent of their lead-exposure at baseline. The study can
determine in several ways whether the R&M interventions are effective in preventing lead
exposure, as measured by children's blood lead concentrations. First, it can show whether or not
their blood lead concentrations reach levels that trigger medical management (^ 15-20 /ug/dL
according to the CDC guidelines) during the post-intervention period of follow-up. Second, it can
show whether R&M interventions are associated with acute increases in blood lead concentration
during the immediate post-intervention phase; this is important because past studies have
documented acute increases in children's blood lead following improper lead-paint abatement
work. Third, to assess the potential for primary prevention of lead poisoning, the study design
includes the enrollment of newborns during the follow-up phase, once they reach the age of six
months. Children born into study households were not included in this report due to small
numbers; however, a separate analysis of their blood lead data is planned for the report on the
second year of follow-up. Also with regard to primary prevention, this study shows that the
modern urban control houses are associated with children's blood lead concentrations below the
CDC's level of concern (10
One reviewer commented that the performance of R&M interventions in vacant and
occupied houses was likely to have an effect on whether there were significant reductions in dust
lead loading and concentrations in R&M I houses (occupied at time of intervention) as compared
with R&M IE houses (vacant at the time of intervention). This point would be of particular
importance if this were a study of changes in dust lead loadings and concentrations immediately
following R&M interventions. However, in this report data were analyzed in terms of dust lead
loadings and concentrations during the first year of follow-up, both within and between groups,
and not in terms of absolute change in dust lead levels immediately following intervention.
Moreover, in the report on pre-intervention findings, the baseline data were analyzed by treatment
-------
group rather than by occupancy status because our primary interest was to assess the comparability
of groups on measures of dust lead and blood lead at baseline and then to compare dust lead
loadings and concentrations and blood lead concentrations over time within and between groups.
Further, the group variable accounted for more of the variability in the dust lead data than the
status variable (occupied/vacant) in our statistical analysis.
On a related point, a reviewer pointed out that R&M II interventions were performed in
both vacant and occupied units and asked whether vacancy influenced the amount, quality, or cost-
effectiveness of the R&M activities. Vacancy did influence the extent to which floors were treated
in R&M II houses and the need for precautions to protect furnishings and other belongings. As
noted in the report, in the case of occupied houses, family members were out of the house while
work was in progress and floors were treated to the extent feasible, given the presence of
furnishings and the drying times of the sealants selected for use. In vacant R&M n houses, all
floors were available for treatment and there was no need to take precautions to protect furnishings
and personal belongings.
One reviewer noted that the finding of some sporadic reaccumulation of lead dust loading
to pre-interventions levels at some sites in R&M houses highlights the importance of counseling
families to conduct regular, targeted lead cleaning in their homes, even after deleading. The
potential for reaccumulation of lead in dust was one reason why this study was designed to include
feedback of information on household dust lead loadings to participating families. Additionally,
HUD guidelines recognize the need for ongoing inspection and maintenance of houses that have
received interim control interventions. Further, it is important to emphasize that the longevity of
the three R&M interventions under investigation is unknown; sustained reductions were found in
dust lead loadings during the first year of follow-up in each of the three intervention groups as
explained below.
2.0 SUMMARY AND DISCUSSION OF FINDINGS
During the first year of follow-up, this study met objectives related to enrollment,
laboratory performance, data quality and data completeness (section 3.0). The latter is attributable
to the study families' willingness to cooperate with the blood lead testing and the environmental
sampling components of the study. This, in turn, is a reflection of the good rapport established
between study staff and participating households. During the first year of follow-up, 21 (20
percent) of the 107 original families moved from study houses. In all but three cases, the house
was subsequently reoccupied and the new family was enrolled in the study. This assured that, at
a minimum, the house remained in the study. Most of the new families also had eligible children
who were enrolled in the blood lead testing component of the study.
All families were informed by letter of the results of the dust lead and blood lead tests from
each campaign. Results of dust tests were provided on a qualitative basis with recommendations
for housekeeping priorities to address areas with high dust lead loadings. For this reason, it is
-------
important to emphasize that the study intervention consisted of a combination of R&M work and
the provision of information to families on a periodic basis. The nature of the intervention limits
the degree to which study findings can be generalized to houses which will receive R&M
interventions, but no periodic monitoring of dust lead levels and/or feedback of results to families.
It is also important to note that the costs of the interventions in this project may not be
generalizable to other settings and time periods due to potential differences in labor and material
costs and overhead rates.
This section summarizes and discusses the main findings, including those based on the
fitting of the study's statistical models for longitudinal data analysis (section 6.3) to the dust lead
and blood lead data. The longitudinal models enabled investigation of lead levels in house dust and
in children's blood across time within study groups and comparisons between groups during the
first year of follow-up, accounting for age, season, and other potential covariates. These models
also address statistical issues associated with having multiple measurements per house and repeated
measures over time.
Dust Lead In R&M Houses
All three levels of R&M interventions under investigation (section 4.2) were associated
with statistically significant reductions in both interior dust lead loadings and dust loadings that
were sustained below pre-intervention levels during the first year of follow-up. Moreover, none
of the interventions in individual houses failed, that is, all or most of the dust samples showed lead
loadings at, or below, pre-intervention levels during the first year of follow-up. Reaccumulation
of dust and dust lead loadings in all three R&M groups was the greatest during the first two month
post-intervention, while there was relatively little reaccumulation between two months and 12
months post-intervention (Figures 23 and 24). This early reaccumulation was most evident in
R&M n and R&M HI houses and may be due in part to the possible importation of dust and lead
into the house during move-in by study families. Half of the R&M II houses, all of the R&M El
houses, and none of the R&M I houses were vacant at the time of intervention. Vacancy is also
believed to account for the finding of highest baseline dust lead loadings in R&M in houses
(vacant at baseline) and lowest baseline lead loadings hi R&M I houses (occupied at baseline).
As expected, the dust lead loadings, lead concentrations, and dust loadings during the post-
intervention period of follow-up were related to the intensity of the intervention. Environmental
samples collected immediately following intervention and at two, six and 12 months post-
intervention consistently showed dust lead loadings, lead concentrations, and dust loadings to be
lowest in R&M III houses, intermediate hi R&M n houses, and highest in R&M I houses.
Statistically significant differences were found between R&M groups on these three dust measures
throughout the first year of follow-up. Weighted average measures of dust lead levels on floors,
window sills, and window wells hi an entire house indicated that the relative differences hi
exposure between groups were large. For example, at twelve months post-intervention, weighted
average dust lead loading estimates were 21-fold higher in R&M I houses than hi R&M ffl houses,
and 4-fold higher in R&M II houses than in R&M III houses. In R&M I houses, the 12-month
-------
geometric mean dust lead loading for floors in rooms with windows was 94 jug/ft2, for window
sills it was 470 /xg/ft2, and for window wells it was 16,698 jig/ft2. In R&M II houses, the 12-
month geometric mean dust lead loading for floors in rooms with windows was 76 jug/ft2, for
window sills it was 237 fj-g/f?, and for window wells it was 2,587 /zg/ff. Finally, in R&M III
houses, the 12-month geometric mean dust lead loading for floors in rooms with windows was 50
/^g/ft2, for window sills it was 29 pig/ft2, and for window wells it was 220 /xg/f? .a Differences
in lead loadings between groups are attributable mainly to differences in lead concentrations
between groups and secondarily attributable to differences in dust loadings (Figure 22).
Dust lead concentrations were found to be statistically significantly reduced following
intervention in R&M II and R&M III houses, but not in R&M I houses. Significant differences
in dust lead concentrations between R&M groups were anticipated based on differences between
the three levels of intervention. By design, R&M III interventions, and to a lesser extent R&M
II interventions, directly addressed lead-based paint, a source of high lead concentrations in house
dust. For example, R&M III interventions typically involved the replacement of lead-painted
windows and the use of durable aluminum coverings to enclose lead paint on exterior components
of windows and doorways. In R&M II interventions, window friction surfaces were treated to
reduce the abrasion of lead paint, but windows generally were not replaced. In contrast, R&M I
interventions directly addressed paint sources only to the extent that deteriorating paint on interior
and exterior surfaces was stabilized and window wells were capped with aluminum coverings.
Sustained reductions in lead concentrations in R&M II and R&M III houses, and less frequent
observations of paint chips on sampled window surfaces during follow-up, indicate that these
interventions contributed to the control of paint as a source of high lead concentrations in house
dust for a one-year period. A reduced rate of lead input into dust from paint may explain in part
the downward trend in dust lead concentrations in R&M II houses two months following
intervention even in the presence of rising dust loadings. A sharp rise in dust lead concentrations
in future data collection campaigns would likely signal the presence of lead paint hazards and the
need for further remediation.
The patterns observed in dust loadings and dust lead concentrations between R&M groups
also may be related to the degree to which the surfaces in the three levels of intervention were
made smooth and easily cleanable. For example, in R&M in houses, floors were covered or sealed
to make them smooth and easily cleanable. Floors in R&M II houses were sealed (to the degree
possible in the subset of houses occupied at the time of intervention), while floors in R&M I
houses were neither sealed nor covered. Also, the window wells in the houses in all three R&M
groups were covered in some manner. The provision of smooth and easily cleanable surfaces has
been shown to be an important element of effective residential lead paint abatement.6'7 In this
study, surface conditions would have influenced the effectiveness of the post-R&M cleanup by
a It should be noted that the cyclone device used to collect dust in this study has been shown to
produce higher estimates of dust lead loadings compared to wipes across a range of surface types and
conditions. However, the cyclone device tends to yield lower estimates of dust lead loadings than wipes
on smooth surfaces with low lead loadings (< -100 ^g/ft2). 21
-------
contractors and housekeeping by study families. Based on field data recorded at the time of
sampling, window well surfaces in the three groups of intervention houses were generally
smoother and less deteriorated one year post-intervention compared to the pre-intervention
baseline.
Dust Lead In Control Houses
The modern urban and previously abated control houses were characterized by a relative
stability of dust lead loadings, lead concentrations, and dust loadings over time (Figure 24). No
statistically significant differences in the three dust measures were found within in these two
groups during the first year of follow-up. Downward but nonstatistically significant trends were
noted in lead loadings and dust loadings across time for both groups of houses. These trends may
be related, in part, to families becoming more aware of the importance of lead dust control as a
result of study participation and to the fact that dust was repeatedly removed from household
surfaces by the sampling process.
The modern urban control houses are rowhouses located hi clusters of houses built after
1979 and presumably free of lead-based paint because of the year of construction.48 It is expected
that this type of housing reflects the lowest residential and ambient lead levels in the urban
environment. The consistently low overall interior dust lead concentrations (geometric mean ^310
/-ig/g (ppm), equivalent to ^0.03%) and low soil lead concentrations (geometric mean ^75 jug/g)
support the presumption that these houses are free of lead-based paints. As noted previously, the
investigators are seeking EPA funding to test the paint hi the modern urban control houses to
determine directly if the paint contains lead additives. This group of houses had significantly lower
dust lead loadings and lead concentrations compared to each of the other study groups at baseline
and throughout the first year of follow-up. At one year, weighted average lead loadings in modern
urban houses were three tunes lower than in R&M III houses. The geometric mean dust lead
loading for floors in these houses was 8 pig/fl2, for window sills it was 9 /xg/ft2. and for window
wells it was 208 /xg/ft2 compared to previously abated houses where the geometric mean dust lead
loading for floors was 77 /^g/ft2, for window sills it was 75 /xg/ff, and for window wells it was
1,164/ng/ft2.
The previously abated control houses had geometric mean lead loadings at the six-month^
and 12-month campaigns that tended to be between the levels found in R&M II and R&M ffl
houses. These findings may be related to differences in time since intervention between R&M
groups and this control group. The 12-month campaign occurred three years to five years post-
abatement hi the previously abated control houses. Further, average dust lead concentrations hi
R&M II and R&M III houses were not significantly different from those in previously abated
houses at six and 12 months. This is consistent with the fact that none of these interventions,
including the comprehensive abatements, involved the complete removal of all lead-based paint
from a home. As illustrated by the case in which a child's blood lead concentration rose to 53
/xg/dL during follow-up, the previously abated control houses were comprehensively, but not fully,
abated of lead paint. In these houses, some ulterior (in this case basement) surfaces that had not
10
-------
been treated due to resource limitations, and some painted exterior surfaces that had been
stabilized as part of the original abatement were found to be in deteriorated condition during the
R&M study sampling campaigns. These problems combined with deteriorating exterior paint on
neighboring houses, were likely sources of this child's exposure. This case points to the need for
ongoing inspection and maintenance of houses, particularly those houses that receive less intensive
interventions.
It should be emphasized that the three R&M interventions being investigated in this study
are interim control or partial abatement approaches to reducing lead-based paint hazards in
housing. As such, they are not expected to be as long-lasting as comprehensive abatement, and
documentation of the longevity of the R&M interventions is, therefore, a major study objective.
To date, dust lead loadings at sporadic sites in individual study houses (particularly in R&M I
houses) have reaccumulated to levels close to pre-intervention levels. However, during the first
year of follow-up, none of the R&M interventions in houses exhibited widespread failure. All or
most of the interior dust lead loading measurements in R&M houses were at or below pre-
intervention levels during the first year of follow-up. If failures do occur, contingency funds will
be used to perform additional remediation work.
Lead In Drip-Line Soil And Tap Water
Soil and water samples were tested in order to take these sources into account in the
analysis of the longitudinal dust lead and blood lead data. Soil lead data were limited due to the
absence of drip-line soil at most study houses, except for at modern urban houses. Geometric mean
soil lead concentrations at baseline and six months ranged from 700-730 /xg/g in R&M I and R&M
n houses and were an order of magnitude higher than the soil lead levels measured at the modern
urban houses over time (geometric means of 63-73 /jg/g). The low soil lead concentrations found
at the modern urban houses are consistent with the possible use of replacement sod or soil at the
time of construction of these houses (Table 17). Tap water was found to have low concentrations
of lead. Geometric mean water lead concentrations across groups were <,4 /-ig/L (ppb) at the initial
and six month campaigns, and only a small number of readings exceeded the EPA drinking water
standard of 15 /tg/L (Table 18). The combination of low water lead concentrations and the absence
of a significant correlation between children's blood lead concentrations and water lead
concentrations indicates that water was not likely to have been an important source of lead
exposure in study children. Beyond this, no major conclusions were drawn with regard to these
sources due to the limited generalizability of these water and soil data.
Blood Lead
The majority of U.S. children with elevated blood lead concentrations defined by the U.S.
CDC as £ 10 /ig/dL have lead concentrations in the range of 10-20 /ig/dL.23 Little is known,
however, about blood lead changes associated with lead paint hazard reduction interventions in
the homes of children with low-to-moderate blood lead concentrations.8-24 In this study, the
unadjusted geometric mean blood lead concentrations (PbB) at baseline were 10 /tg/dL for R&M
I children, 14 /ig/dL for R&M II children, and 14 /ig/dL for R&M m children, and 13 /ig/dL for
11
-------
children in the previously abated houses and 5/*g/dL for children in the modern urban houses. At
the end of the first year of follow-up, the unadjusted geometric mean blood lead concentrations
were lower for each group: 8 /xg/dL for R&M I children, 11 jug/dL for R&M II children, and 12
/xg/dL for R&M III children, and 12 /xg/dL for children in the previously abated houses and 3
jug/dL for children in the modern urban houses (Table 16). Mean ages at the end of the first year
of follow-up ranged from 39 months to 44 months across groups .b
One of the study's longitudinal data analysis models allowed for comparisons of blood lead
concentrations within and between groups, and for control of age, season and other potential
covariates. This comparison model was fit separately for children with blood lead concentrations
<20/*g/dL and >20/*g/dL. According to CDC guidelines, children with blood lead concentrations
>20/*g/dL should be referred for medical evaluation and management.10 Children born into study
houses were not included in this report due to the small number of children involved.
In children with initial blood lead concentrations (PbB) <20 pg/dL, no statistically
significant changes in blood lead concentration were found within any of the five study groups
during the first year of follow-up, controlling for age and season. Further, no significant
differences in blood lead concentrations were found between R&M groups during the first year
of follow-up, again controlling for covariates. Children in the modern urban control group had
statistically significantly lower blood lead concentrations than children hi the other four groups.
This was the only statistically significant blood lead finding among study children with initial
blood lead concentrations < 20 jug/dL. The blood lead concentrations of children hi the modern
urban group were all less than or equal to the CDC's blood lead level of concern (10 /xg/dL)
during the first year of follow-up.
The absence of an increase in blood lead concentration at two months post-intervention is
noteworthy because past studies have attributed short-term rises in children's blood lead
concentrations to improper abatement practices.24"26 Precautions taken in R&M houses included
having children out of the house while R&M work was in progress and the use of work practices
to minimize, contain, and remove lead-contaminated dust. Further, one could hypothesize that,
accounting for age, the R&M interventions prevented increases in blood lead concentrations during
the entire first year of follow-up that study children might have experienced otherwise hi the
absence of the R&M interventions. For ethical reasons, the study design did not include a non-
b The geometric mean blood lead concentration (PbB) for children in the modern urban
group was slightly above the geometric mean of 2.7 |u.g/dL reported for U.S. children aged 12
months to 60 months but very similar to that estimated for all non-Hispanic black children in this
age range, 4.3 ng/dL.1 The unadjusted geometric mean PbB in each of the other four study
groups was similar to, or higher than, the estimated geometric mean PbB value of 9.7 ug/dL
previously reported for U.S. non-Hispanic black children for low-income families living in
central cities (populations >1 million, 1988-1991).23
12
-------
intervention control group to test this hypothesis.
As anticipated, nearly all children with baseline blood lead concentrations >20/*g/dL were
in the R&M II and R&M III groups because the policy of one of the collaborating housing
organizations was to rent its improved properties to families with lead-poisoned children. Children
across all groups with initial blood lead concentration >20pig/dL (n=19) had a statistically
significant reduction in blood lead concentration (in most cases to levels < 20/ig/dL) during the
first year of follow-up, controlling for age and season (Figures 14-18). Due to small numbers,
the effect of R&M I intervention on the blood lead concentrations of children with baseline blood
lead concentrations >20jug/dL cannot be assessed in this study.
Relationship Between Blood Lead And Dust Lead
At the end of the first year of follow-up, statistically significant correlations, ranging from
.28 to .44, were found between children's blood lead concentrations and dust lead loadings (both
on the log scale) for various surface types (Tables 22 and 23). These low-to-moderate magnitude
correlations are consistent with those reported in the literature.14'15 A statistical model was used
to assess the relationship between blood lead concentration and dust lead loadings and
concentrations, controlling for covariates. Using data from all five study groups in the longitudinal
data analysis, blood lead concentration was found to be significantly related to a linear
combination of floor, window sill, and window well dust lead loadings and to a similar composite
measure of dust lead concentrations, after controlling for age and season. Gender and hand-to-
mouth activity were not found to be consistently significant contributors to the model in this study.
The latter may be attributed to the more-or-less truncated blood lead concentration distribution and
the aging of study children. Further, a statistically significant relationship was not found between
dust lead loadings and concentrations and blood lead when the statistical model was fitted to blood
lead concentration data from just the three R&M groups. This was likely due to the narrower
range of post-intervention dust lead loadings and concentrations, compared with pre-intervention
dust lead loadings and concentrations, exacerbated by the absence of the low-lead modern urban
houses and children living in these types of houses from the analysis. Other studies, including a
recent study in Rochester,14 have found a statistically significant relationship between children's
blood lead concentrations and lead in settled dust in their homes.
Seasonal change in children's blood lead concentration was estimated to be +1.2
in summer relative to the other seasons, controlling for age, campaign and dust lead loading and
concentration. Other studies have reported seasonal trends in children's blood lead concentrations
for different years and populations that vary in the estimated magnitude of the seasonal difference
but generally were higher than that reported here.27"29
Considerations In The Interpretation Of Blood Lead Findings
Multiple factors can theoretically mediate a child's blood lead concentration response to
an intervention. These factors may include cumulative body lead burden, age, degree of hand-to-
13
-------
mouth activity, ambient lead levels, and neighborhood housing characteristics.
Housing history data, combined with the baseline blood lead concentration data, suggest
that children in the modern urban houses had lower body lead burdens at the time of enrollment
than did children in the other four study groups. Most children in the modern urban group had
lived in the same low-lead house since birth, and all of them had baseline blood lead
concentrations less than or equal to the CDC's blood level of concern (10 /ig/dL). By contrast, it
is likely that the children in the R&M and previously abated houses had spent most or all of their
lives prior to enrollment in low-income rental housing and thus were at risk of high exposure to
lead in dust and paint due to poor housing conditions. On average, baseline blood lead
concentrations in these four groups of children were two to three times higher than those of
children in the modern urban group. Body lead burdens could have mediated children's blood lead
concentration responses to the R&M interventions because blood lead reflects a mixture of recent
exposure and lead that the body has stored.
Most (-70 percent) of the lead hi children is stored in their bones,30 and the half-life of
lead in human adult cortical bone is estimated to be 20 years.31 This skeletal lead can be an
ongoing internal source of lead measured in blood even after external exposure and children's lead
ingestion are reduced following lead remediation interventions. This was the case hi an earlier
study of children with much higher blood lead concentrations (geometric mean=63 pig/dL) who
received inpatient chelation therapy and were monitored for several years following discharge to
"lead-free" public housing and abated houses.32 Because the bone lead concentrations of R&M
study children are unknown and the kinetics of lead mobilization from children's bones is not well
understood, it is not possible to estimate the magnitude and duration of bone lead's contribution
to children's blood lead concentrations measured hi the post-intervention phase of this study. For
this reason, additional time beyond 12 months post-intervention may be needed to measure
significant blood lead changes in R&M children. The newborns who are being added to the study
over tune during the period of follow-up are of particular interest because they are likely to have
had minimal exposure to lead prior to enrollment (age six months) and therefore can be followed
to assess the potential for primary prevention of lead poisoning.
Additionally, ambient lead levels hi study neighborhoods may have mediated the children's
blood lead responses to intervention and contributed to blood lead differences between the modern
urban group and the other four groups. By design, the modern urban houses were all located in
housing clusters built after 1979 and are presumably free of lead-based paint. The low lead
concentrations found hi ulterior dust, exterior dust, and soil support the notion that these control
houses were associated with low ambient lead levels. The children hi this group were, therefore,
at low risk of exposure to lead in paint and in the general environment, compared to children
living hi the R&M houses and previously abated houses which are located hi low-income lead-
contaminated neighborhoods. Such neighborhoods often have housing in poor condition and in
close proximity to abandoned and boarded houses.
Because hand-to-mouth activity is recognized as a major entry route for lead into pre-
14
-------
school children,16"18 age and frequency of hand-to-mouth activity are other potential factors
mediating children's blood lead response to an intervention. At the 12-month campaign, most
study children were 36 months to 48 months of age, a range in which the frequency of mouthing
behavior is likely to be less than in infants and young toddlers. This potential reduction in hand-
to-mouth activity could account, in part, for the lack of statistically significant changes in blood
lead concentration within and between R&M groups in children with baseline blood lead
concentration <20 pg/dL, despite the differences in dust lead exposure between and within groups
over time.
The children with blood lead concentrations >20 ptg/dL may have had higher blood lead
concentrations due to more frequent hand-to-mouth activity. It also is possible they may have had
a relatively greater contribution to their blood lead from current exposure rather than from bone
lead, compared to children with blood lead concentrations < 20 /xg/dL. Therefore, their blood lead
concentrations may have been more responsive to the reduction in lead exposure associated with
the R&M interventions than children with lower baseline blood lead concentrations.
The reader is referred to section 7.0 for a more detailed presentation of these and other
findings during the first year of follow-up.
3.0 QUALITY ASSURANCE
3.1 System Audit
Laboratory and field activities have been subjected to regular review to assure conformance
with procedures proscribed in the Quality Assurance Project Plan.2 This ongoing audit has
focused on the sampling and analytical procedures used, their documentation, the training of field
and laboratory personnel, and the adequacy of related facilities and equipment. Reports have been
generated and forwarded to the project officer annually. Inadequacies noted hi early reports have
been subsequently corrected. Only minor problems, not directly related to data quality, were
noted during the first year of follow-up.
3.2 Data Audit
To verify the accuracy of the data used in this report, the quality control officer conducted
a stratified random audit of 5 percent of the field and laboratory data generated during the first
two years of this study. Prior to the audits, laboratory and data staff had completed independent
checks of the data. The audit procedure involved the verification of information in the final data
base against the original field and laboratory data. Samples to be audited were selected by
computer using random number sequences. Sampling was stratified to ensure that samples were
randomly selected to represent every analytical batch. Probably as a result of the extensive quality
control effort prior to the audits by the quality control officer, the audits did not identify any
errors.
15
-------
3.3 Performance Audit
In order to assure that the sampling and analytical protocols employed in the R&M study
yielded data of sufficient quality, a number of different types of quality control samples were
included in the study design. These samples were designed to control and assess data quality in
each phase of the data collection and analysis process, which were potentially subject to random
and/or systematic error. Blank samples, including field blanks and method blanks, were included
to assess procedural contamination by lead. Recovery samples, including standard reference
materials, spiked samples, and calibration verification samples, were included to indicate the
accuracy of analyses. Duplicate samples were used to indicate precision of analyses. Standard
control charts were generated quarterly showing percent recovery of a standard reference material,
percent recovery of spiked samples, spike/spike duplicate precision, initial calibration values,
continuing calibration values, percent recovery of continuing calibration values, and drift of
continuing calibration values within a run. Separate control charts were generated for each
combination of sample matrix and analytical instrument used. Of the more than 6,000 quality
control samples included in these analyses, the control limit (+30 percent) was rarely exceeded
for any quality control parameter. Data on field and method blanks also have been reviewed on
a periodic basis as part of the performance audit.
In addition to these internal quality control efforts, the Kennedy Krieger Research Institute
(KKRI) Trace Metals Laboratory has participated in external quality control programs for
environmental lead samples and blood lead concentrations as a part of the R&M study. Beginning
in September 1993, the laboratory participated in the Environmental Lead Proficiency Analytical
Testing (ELPAT) program for environmental samples. This program is administered through the
National Lead Laboratory Accreditation Program and is sponsored in part by EPA Office of
Pollution Prevention and Toxics. Blind samples are analyzed quarterly; the KKRI Trace Metals
Laboratory has been rated as "proficient" for the evaluation of lead in paint chips, soil, and dust
wipes since joining the program. The Trace Metals Laboratory also participates in the Health
Resources and Services Administration/Wisconsin Blood Lead Proficiency Testing Program.
Three blind blood samples are analyzed every month as a part of this program. Since beginning
this analysis in 1993 the KKRI laboratory has achieved a 100 percent accuracy rating for Graphite
Furnace Atomic Absorption Spectroscopy (GFAA) analysis of blood lead for all rounds in which
the laboratory participated.
Statistical Analyses of OC Data
Because of the overlapping nature of the sampling campaigns in this longitudinal study,
samples from several campaigns are generated and analyzed concurrently. Consequently, there
is no unique set of quality control data that can be attributed to any particular sampling campaign
or set of campaigns. As a result, the quality control data reported here represent all data submitted
as a part of quarterly reports through Oct. 25, 1996. These data include all of the samples from
the initial sampling campaign through the 12-month campaign, plus varying numbers of samples
16
-------
from subsequent campaigns. Statistical analyses of the quality control samples are included in
Tables 2 through 4. With the exception of soil samples, the percent recovery of standard reference
material and the percent recovery of spike and spike duplicates all fell within a tolerance interval
of 70 percent to 130 percent. Precision was very high, with generally less than a one percent
difference between spike and spike duplicate samples. Percent recovery of initial and continuing
calibration samples fell within a tolerance interval of 90 percent to 110 percent. Drift was limited
to an average of less than two percent over a run. Field and method blanks showed extraneous
lead contamination of the samples to be, on average, trivial. No evidence of systematic
contamination was observed.
Additional quality control analyses were conducted on the environmental sampling data to
assess potential bias resulting from sampling conducted by different field personnel. No
statistically significant differences were found between the estimates of dust lead loadings, dust
lead concentrations, and dust loadings based on samples collected by the various members of the
field staff, after controlling for surface type and study group.
17
-------
Table 2: Descriptive Statistics And Tolerance Limits For Percent Recovery For SRM And Spiked Samples And Percent
Differences Between Spike And Spike Duplicate Samples
Sample TV***1
Standard
Reference
Material
(SRM)
Spike/Spike
Duplicate
Type of Analysis
ICP-DV3
GFAA-DV
GFAA-S3
GFAA-W3
ICP-DV
SPIKE
ICP-DV
SPIKE DUPLICATE
ICP-DV
PERCENT DIFFERENCE
GFAA-DV
SPIKE
GFAA-DV
SPIKE DUPLICATE
GFAA-DV
PERCENT DIFFERENCE
GFAA-S
SPIKE
GFAA-S
SPIKE DUPLICATE
GFAA-S
PERCENT DIFFERENCE
GFAA-W
SPIKE
GFAA-W
SPIKE DUPLICATE
GFAA-W
PERCENT DIFFERENCE
Number of
Samples
505
425
20
73
505
505
505
427
427
427
20
20
20
73
73
73
Minimum
(%)
76.27
79.34
43.14
50.99
82.33
77.09
-20.99
80.00
79.00
-36.09
-263.00
35.00
-25.89
72.80
40.80
-7.41
Maximum
<%)
153.64
119.59
108.39
129.18
119.92
121.03
13.29
118.00
139.00
29.31
289.00
142.00
47.01
117.80
120.60
64.87
Mean.
(%>
93.38
92.96
91.47
98.16
97.05
96.87
0.20
98.12
98.04
0.12
82.18
92.16
-0.03
97.45
97.19
0.54
Standard
Error
0.43
0.32
3.23
1.85
0.21
0.22
0.13
0.31
0.34
0.24
21.44
5.78
3.06
0.98
1.31
0.97
Lower Limit
£5&Tofera««!
Interval
I <%>
73.27
79.13
51.66
61.98
87.18
86.54
-0.05
84.71
83.46
-0.35
-181.8
20.94
-6.44
78.19
71.57
-1.40
Upper Limit
95% Tolerance
Interval
(%)
113.49
106.78
131.28
134.33
106.91
107.20
0.44
111.52
112.62
0.59
346.11
163.37
6.37
116.72
122.80
2.48
DV = cyclone dust, S = soil, W = water
18
-------
Table 3: Descriptive Statistics And Tolerance Limits For Percent Recovery For ICV And CCV
, Sffittjj>te%»e
Initial
Calibration
Verification
(ICV)
Continuing
Calibration
Verification
(CCV)
Tj-pe of Analysis
ICP-DV3
GFAA-DV
GFAA-S3
GFAA-W*
ICP-DV
% TRUE VALUE
ICP-DV
% DRIFT
GFAA-DV
% TRUE VALUE
GFAA-DV
% DRIFT
GFAA-S
% TRUE VALUE
GFAA-S
% DRIFT
GFAA-W
% TRUE VALUE
GFAA-W
% DRIFT
Notaberdf
Sampler
287
120
34
63
1937
1937
476
476
77
77
173
173
MiBimtaa
m
93.08
92.50
93.50
96.00
90.02
-13.95
90.50
-12.15
89.00
-13.88
90.50
-12.80
Maximum '•
m
109.98
110.00
109.00
110.00
112.70
14.53
112.50
11.46
109.00
9.23
110.00
11.86
Meatt
m
100.50
103.41
102.57
103.40
98.82
-1.68
102.80
-0.83
101.14
-1.09
102.77
-0.51
Standard
JError
0.18
0.34
0.60
0.41
0.09
0.10
0.20
0.20
0.58
0.54
0.34
0.33
Lower Limit
£5% T$9KHHje
Interval \
m
93.97
95.28
93.89
95.77
90.94
-1.88
93.82
-1.22
89.47
-2.17
93.21
-1.17
Upper Limit
9$% Tolerance
Interval
m
107.04
111.54
111.25
111.04
106.70
-1.49
111.77
-0.43
112.81
-0.01
112.33
0.14
DV = cyclone dust, S = soil, W = water
19
-------
Table 4: Descriptive Statistics For Field Blanks And Method Blanks
$a*t)$ie1^3£
Field Blank
Method Blank
Type of Sample
Dust3
Soil
Water
Dust
Soil
Water
Naukttt? of
Samples
796
107
364
507
20
73
MittittlWttt
Cmg/k)
0.06
0.01
0.15
-152.00
-0.40
-0.80
Maximuw
(mg/L)
621.00
2.59
92.00
549.00
14.00
8.90
Mean
(n»g/t)
8.67
0.17
1.42
7.81
2.55
0.62
Standard Irror
1.39
0.03
0.27
1.83
0.86
0.14
Field blanks are analyzed by ICP or GFAA
20
-------
4.0 STUDY DESIGN AND SAMPLE COLLECTION PROCEDURES
The R&M study targeted houses in low-income neighborhoods where children are at
highest risk of lead-poisoning due to exposure to lead in dust and in deteriorating paint. It is
important to emphasize that the R&M study was not designed as an intervention study in the
homes of lead-poisoned children per se, although some study children did have blood lead
elevations at baseline. Instead, the study started by identifying eligible intervention and control
houses with eligible children. The eligibility criteria for children were based on age and other
parameters, but not blood lead concentration (see section 4.4). It is also important to recognize
that the study was not designed to assess the specific effects of the various elements of the
interventions (e.g., provision of information to families) on the study outcomes. Instead, the study
investigated the effectiveness of the R&M interventions as a whole. To assess the potential for
primary prevention of lead poisoning, the study design included the enrollment of newborns once
they reached the age of six months. The sections below provide an overview of the study design
followed by descriptions of the R&M interventions, recruitment and enrollment procedures,
selection criteria for houses and children, selected characteristics of the study houses, and sample
collection procedures.
4.1 Overview Of Study Design
The R&M study has two main components and five groups of study houses. The first
component is designed to obtain serial measurements of lead in the venous blood of children
between the ages of six months and 60 months at enrollment. The study also obtained serial
measurements of lead in house dust, exterior soil, and drinking water in three groups of 25 houses
(a total of 75 houses), each being subjected to one of three levels of R&M intervention. The
second component was designed to collect a comparable set of measurements in two groups of
control houses. Table 5 summarizes the types of data planned for collection by study group and
by campaign. Blood lead and dust lead measurements were planned in all R&M study houses at
each campaign, except blood lead was not collected at the immediate post-intervention campaign.
Measurements of lead in exterior soil and drinking water were made as part of a subset of all
campaigns. The study questionnaire, designed to obtain information on demographics and
covariates that could influence lead exposure in the home (e.g., hobbies and child behavior), was
administered at six month intervals starting at enrollment.
R&M intervention houses (vacant and occupied) were identified in collaboration with
owners and operators of low-income rental properties as explained in section 4.3. Occupied
houses that were eligible for R&M intervention were randomly assigned to receive either R&M
I (low level intervention) or R&M II (intermediate level intervention). Vacant houses that were
eligible for R&M intervention were randomly assigned to receive R&M II or R&M HI (high level
intervention). The R&M II intervention was designed to be performed in both occupied and
vacant houses, and the randomization scheme was designed to ensure that equal numbers of houses
(n=25) were assigned to each R&M intervention level. To allow for a better estimation of the
post-intervention rate of re-accumulation of lead in dust and for periodic assessments of the need
for further cleanups/repairs during the follow-up period, more frequent sampling campaigns were
planned in the R&M groups during the first year of follow-up (Table 5).
21
-------
Table 5: Data Collection Plan For Lead Paint Abatement And Repair & Maintenance Study
R&MI
Blood
Dust
Soil
Water
Questionnaire
V
V
R&MII
Blood
Dust
Soil
Water
Questionnaire
V
V
V
/*
/*
V
V
R&M III
Blood
Dust
Soil
Water
Questionnaire
/
V
V*
V
V
Control Houses:
Previously
Abated and
Modern Urban
Blood
Dust
Soil
Water
Questionnaire
V
V
Shading indicates data covered in this report
Blood, questionnaire, and water samples were not collected in vacant houses until the family moved in following intervention.
The need for additional cleanups/repairs during the entire follow-up period will be determined by a comparison of the follow-up
22
-------
dust lead loadings and blood lead concentrations with their corresponding pre-intervention levels.
Further cleanups/repairs will be performed when dust lead loadings at most interior sites in a
house re-accumulate to levels that exceed pre-intervention levels. This assessment will exclude
interior sites with low baseline dust lead loadings (e.g., < 100 jug/ft2) that remain low at follow-
up, despite small increases in their lead loadings. In contrast, clean-up/repair will be considered
for sites with high levels at baseline and at follow-up (e.g., >25,000 /xg/ff) where the follow-up
level approaches, but does not exceed, the corresponding baseline value.
The second component of the study is to obtain serial measurements of lead in venous
blood of children six months through 60 months of age at enrollment, and in house dust, soil, and
drinking water in two groups of control houses. The first control group consisted of 16 houses
drawn from a group of houses that received comprehensive lead-paint abatement in demonstration
projects in Baltimore between May 1988 and February 1991.6'7 The second control group
consisted of 16 modern urban houses built after 1979, which were presumably free of lead-based
paint. The types and frequencies of measurement were the same in the two control groups (Table
5). The two years of follow-up planned for the previously abated control group will provide an
opportunity to measure the effectiveness of comprehensive abatement four years to six years after
abatement.
It should be noted that the sample sizes of the control groups were reduced from 25 to 16
houses each, due to reductions in the scope and funding of the project. The number of control
houses, rather than the number of R&M houses, was reduced because the former (and in particular
the modern urban houses) were expected to have less inter-house variability with respect to both
blood lead and dust lead. This was borne out in the study findings.3 Furthermore, two types of
houses were originally planned for inclusion in the modern urban control group: houses in
clusters of urban houses built after 1979, and houses in scattered sites, that had been extensively
rehabilitated after 1979. When the sample size of modern urban houses was reduced to 16 houses,
only the former were included as the negative (no lead paint) control group (see section 4.5 for
additional descriptive information). It was expected that this type of cluster housing would reflect
the lowest residential and ambient lead levels in the urban environment.
4.2 Repair & Maintenance Interventions and Comprehensive Abatement
R&M Levels I-IH
The R&M interventions were financed by the Maryland Department of Housing and
Community Development (DHCD) through a special loan program open to low-income owner-
occupants and private property owners who rent their properties to low-income tenants. To meet
DHCD loan eligibility requirements and the pre-requisites for R&M-type interventions imposed
by the study, the three levels of R&M interventions were planned for study in lead-painted houses
that had no structural defects and that were maintained according to the eligibility criteria listed
in section 4.4. The R&M intervention costs were capped by DHCD as follows: R&M I, $1,650;
R&M H, $3,500; and R&M HI, $6,000 to $7,000. The last range is due to program criteria and
pre-existing program agreements. It is important to note that the costs of the interventions in this
23
-------
project may not be generalizable to other settings and time periods due to differences in labor and
material costs and overhead rates.
The three levels of intervention, described in detail elsewhere,2 are described briefly below
and in Table 1. R&M I included the following elements: wet scraping of peeling and flaking
lead-based paint on interior surfaces; limited repainting of scraped surfaces; wet cleaning with a
tri-sodium phosphate detergent (TSP) and vacuuming with a high efficiency paniculate air (HEPA)
vacuum to the extent possible in an occupied house; the provision of an entryway mat; the
provision of information to occupants; and stabilization of exterior surfaces to the extent possible,
given the project's budget cap. The R&M II interventions included two key additional elements:
floor treatments to make them smooth and more easily cleanable and in-place window and door
treatments to reduce abrasion of lead-painted surfaces. In addition to all of this, R&M III
intervention included window replacement and encapsulation of exterior window trim with
aluminum coverings as the primary window treatment, encapsulation of exterior door trim with
aluminum, and more durable floor and stairway treatments. R&M households received cleaning
kits for their own cleaning efforts. The kits each included a bucket, sponge mop, sponges, a
replacement sponge mop head, a TSP cleaning agent, and an EPA brochure entitled "Lead
Poisoning and Your Children."
Elements of Comprehensive Lead-Paint Abatement
The previously abated control houses received a comprehensive form of lead-paint
abatement in demonstration projects in Baltimore between May 1988 and February 1991,6'7 These
comprehensive abatements included the following elements:
• Treatment of all lead painted (^0.7 mg/cm2 or ^0.5% lead by weight) surfaces primarily
using replacement and enclosure methods;
• Minimal use of on-site paint removal methods;
• Fixing water leaks and other pre-existing conditions that would impediment effective
abatement;
• Installation of vinyl replacement windows and enclosing of the exterior window trim with
aluminum coverings;
• Making floors smooth and easily cleanable by the use of vinyl tile and sealant coatings;
• Door and stairway treatments, including replacement of lead-painted components;
• Cleaning by wet washing and the use of HEPA vacuum cleaners.
4.3 Recruitment And Enrollment
R&M study houses were identified from lists of addresses provided by owners of private
rental properties in low-income neighborhoods of Baltimore and by City Homes, Inc., a non-profit
housing organization, which owns and operates low-income rental properties to demonstrate
methods of managing and maintaining such properties. The small number of owner-occupant
properties in the R&M intervention groups (n=4) were identified through the KKRI's Lead
Poisoning Prevention Program and outside sources. The previously abated houses were identified
24
-------
from lists of houses abated in past years as part of lead paint abatement demonstration projects
conducted by Baltimore and KKRI. The modern urban houses built after 1979 were identified by
house-to-house visits conducted in multiple clusters of such housing in Baltimore.
The enrollment process was done in two stages: pre-enrollment and formal enrollment.
These activities were undertaken by study field workers who conducted extensive home visits
(1,100 visits to more than 650 modern urban, previously abated, and candidate R&M houses)
during the spring and summer of 1992. More than 90 percent of households identified as
potentially eligible for the study indicated an interest in participating. Unfortunately, demographic
data are not available to compare those households to households which did not express interest
in participating. This pre-enrollment activity yielded 100 interested and eligible households for
formal enrollment. Formal enrollment entailed obtaining signed informed consent statements for
study participation from parents or legal guardians for both environmental and biological
sampling. Separate consent statements were obtained for each child enrolled in the study using
forms approved by the Joint Committee on Clinical Investigation of the Johns Hopkins Medical
Institutions.
Between the time of formal enrollment and the commencement of the initial data collection
campaign in January 1993, some enrolled households became ineligible, primarily due to the
children growing too old to participate and the families moving to other dwellings. In some cases,
the losses re-initiated pre-enrollment activity to identify an additional pool of potential study
participants. The initial environmental sampling campaign in the modern urban and previously
abated control houses was performed between January 1993 and July 1993. The baseline
environmental sampling in R&M houses was conducted between March 1993 and November 1994.
4.4 Selection Criteria For Houses And Children
Houses and children were selected for participation in the study based on a rigid set of
criteria. The first set of selection criteria listed below was applied to all five study groups.
Additional selection criteria were applied to the three R&M groups and to the previously abated
control group.
Selection criteria applied to all five study groups:
• House size was approximately 800 to 1,200 ft2.
• The house was structurally sound without pre-existing conditions that could impede
or adversely affect the R&M treatments and the safety of the workers and field staff
(e.g., roof leaks or unsafe floor structures). This criterion eliminated substandard
housing in need of major renovation and, therefore, not suitable for R&M-type
interventions. It also allowed a house to qualify for the special state loans that
financed the R&M interventions.
25
-------
• Utilities (heat, electric, and water) were available to facilitate interventions and
field sampling.
• Each household included at least one child who was six months through 60 months
of age at enrollment and was not mentally retarded or physically handicapped or
had restricted movement. The house also had to be the child's primary residence
(i.e., the child was reported to spend at least 75 percent of time at the address).
Also, the child's family had no definite or immediate plans to move at the tune of
enrollment.
• The house did not contain a large amount of furniture. This criterion allowed dust
collection in all houses, as well as the intervention and cleanup efforts in occupied
R&M houses.
Additional selection criteria applied to R&M houses:
• The house contained lead-based paint (defined in Maryland as sO.7 mg Pb/cm2 or
^0.5 percent lead by weight, as determined by wet chemical analysis) on at least
one surface in a minimum of two rooms or, in the absence of testing, was
constructed prior to 1941 (when lead-based paints were commonly used19).
• Interior house dust lead loadings, prior to intervention, exceeded Maryland's
interim post-abatement clearance levels (i.e., 200 pig/fl2 for floors, 500 pig/ft2 for
window sills, and 800 /ig/ft2 for window wells) at a minimum of three
locations.33iC
• The house had 12 or fewer windows needing R&M work. This was to allow for the
implementation of the R&M interventions, given limited resources.
Additional selection criterion applied to previously abated houses:
• At least two pairs of pre-abatement and immediate post-abatement dust-wipe lead
measurements from the same floor, window sill, and window well surfaces were
available from previously collected data. This ensured that data were available to
the R&M study on pre- and post-abatement baseline dust lead levels hi these
control houses.
c In 1990, these clearance levels were adopted as interim post-abatement clearance levels by
the U.S. Department of Housing and Urban Development (HUD). In 1995, HUD revised its interim
clearance standard for floors to be 100 /ug/ft2.4
26
-------
4.5 Characteristics Of Study Houses And Participants
The R&M houses and the previously abated houses are all scattered-site houses located in
older residential neighborhoods in Baltimore. These study houses were built prior to 1941. More
than 98 percent of the R&M houses and 100 percent of previously abated houses were rowhouses,
which constitute the predominant type of housing in inner-city Baltimore neighborhoods. As
mentioned previously, the 16 modern urban houses are rowhouses located in clusters built after
1979. The clusters of modern urban houses, which served as the sampling frames for this study,
were all located in, or are adjacent to, urban housing neighborhoods constructed prior to 1941.
Each cluster had multiple rows of housing built after 1979 and the rows generally extended the
length of a city block. The characteristics of the study houses were typical of housing hi low-
income neighborhoods in Baltimore. Unfortunately, data do not exist to allow a comparison of
dust lead levels in study homes to those in city homes in general.
Study houses generally were similar in terms of characteristics that might influence patterns
of dust movement into and within a house (i.e., overall size, number of windows, house type and
design, condition, degree of setback from the street, and the presence of porches and yards).3 The
selection criteria ensured that the study houses would be similar in terms of size, number of
windows, and, to some degree, overall condition. With regard to housing type, all five groups
of houses consisted primarily of two-story rowhouses (not located at the end of the row) with two
or three rooms on each level. Floor plans were produced for each study house to facilitate the
sample collection activities. The proportion of carpet samples in composites was, on average, very
low - essentially zero - in R&M I, R&M n, R&M HI, and previously abated houses. On average,
the proportion of carpets making up floor dust composites hi modern urban houses was very high,
averaging close to 100 percent. In all groups, differences were noted hi the distribution of carpets
between first and second stories.
Further, most study houses did not have porches (84 percent), were not located on narrow
alleys (77 percent), and were not set back far from the street (77 percent). Houses with minimal
set-backs had no front yards and entryways leading directly from the sidewalk, or from stairs
ascending directly from the sidewalk. The other 23 percent of study houses were more than
minimally set-back from the street, primarily due to the presence of porches or small front yards.
Only four houses (3 percent) were classified as being set-back from the street by more than a
modest amount as described above. Unlike the other four groups of houses, most of the modern
urban control houses had yards hi the front or back of the house. For this reason, exterior soil
was available for collection at baseline from 69 percent of the modern urban houses, as opposed
to only 15 percent of the R&M houses and 19 percent of the previously abated houses.
As reported previously,3 a comparison of the 75 R&M houses to 27 R&M candidate
houses that were sampled but not included hi the study revealed no evidence of selection bias
based on environmental lead concentrations, lead loadings, dust loadings or the blood lead
27
-------
concentrations of resident children.
4.6 Sample And Data Collection Procedures
Venous blood was collected from study children at the Kennedy Krieger Institute's Lead
Poisoning Clinic by a pediatric phlebotomist into 3 mL Vacutainers* with EDTA added as an
anticoagulant. Information on the study children and their households was collected using a
structured interview questionnaire. Trained field teams administered the questionnaires and
collected all environmental samples, including field quality control (QC) samples.
Settled house dust was collected using a modified high-volume cyclone sampler originally
developed for EPA for the evaluation of pesticide residues in house dust.34 The modified device,
referred to as the R&M cyclone, is described in detail and characterized elsewhere.20-21 The device
consists of a Teflon'-coated cast aluminum cyclone attached to hand-held Dirt Devil * vacuum as
the air mover for the system. A 100 mL Teflon* microwave digestion liner was used as the sample
collection container to eliminate a sample transfer step in the laboratory, thereby reducing the risk
of sample loss.
The sampling plan for settled dust included the collection of three composite floor dust
samples in each of the houses at each campaign: one composite in rooms with windows on the first
story, one composite in rooms with windows on the second story, and one composite in first and
second story rooms without windows. Each composite was composed of samples collected from
two randomly selected 1 ft2 (929 cm2) perimeter floor locations in each appropriate room. If a
randomly selected location were carpeted or covered with an area rug, this information was
recorded on the sample collection form and the carpet or rug was sampled using the R&M
cyclone. Settled dust also was collected in two composite window sill samples and two composite
window well samples in each house at each sampling campaign. Samples were composited by
story from all windows available for sampling. Examples of windows not available for sampling
were those with window air conditioners and those blocked by furniture. Settled dust also was
collected as individual (i.e., not composite) samples from horizontal portions of air ducts, from
interior and exterior entry ways, and from the main item of upholstered furnishing in each house.
Three individual soil core samples were collected from the top 0.5 inch (1.3 cm) of soil
from three randomly selected locations at the drip-line and then combined as one composite
sample. Each soil core was collected into a polystyrene liner using a six-inch (15.2 cm) stainless
steel recovery probe.
Drinking water samples were collected as two-hour fixed-time stagnation samples from the
kitchen faucet. This procedure involved running the cold water for at least two minutes to flush
the pipes and, after a two-hour interval, collecting the first flush of water in a 500 mL
polyethylene bottle. A list of field sample types is provided in Table 6.
28
-------
Table 6: Types Of Field Samples
Sample Type
Perimeter Floor Composite Settled
Dust
Window Sill Composite Settled
Dust
Window Well Composite Settled
Dust
Air Duct/Upholstery Settled Dust
Interior Entryway Settled Dust
Exterior Entryway Settled Dust
Soil Core
Drinking Water
Field QC
Sampling Locations/Specifics
First story and second story rooms with windows;
rooms without windows
First and second story
First and second story
Upholstery was sampled if air ducts were unavailable
Not directly on entryway mat
Not directly on entryway mat
Drip-line composite
Kitchen faucet
Blanks and duplicates for all field sample types
Families were informed by letter of the results of dust lead and blood lead tests from each
campaign. Results of dust tests were provided on a qualitative basis with recommendations for
housekeeping priorities to address areas with high lead loadings. Additionally, letters were sent
to the parents/guardians of the study children with the results of the blood lead tests to be shared
with the child's primary care provider. All blood lead test results were reported to the Maryland
Blood Lead Registry, as required by Maryland law.
29
-------
5.0 LABORATORY ANALYSIS PROCEDURES
Interior and exterior settled dust, exterior soil, water and venous blood samples were
analyzed at the Kennedy Krieger Research Institute's Trace Metal Laboratory using established
analytical methods. Closed vessel microwave digestion was used for dust, soil, and water samples,
according to modified SW 846 Methods 3015 and 3051. Analysis of dust digestates was performed
using Inductively Coupled Plasma-Atomic Emission Spectrometry (ICP), according to SW 846
Method 6010 and/or Graphite Furnace Atomic Absorption Spectrometry (GFAA), according to
SW 846 Method 7421. Soil and drinking water were analyzed by GFAA according to SW 846
Method 7421. Venous blood was analyzed by GFAA and by anodic stripping voltammetry
(ASV).35 Table 7 summarizes these procedures.
Table 7: Summary Of Laboratory Procedures
Sample Type
Dust
Soil
Drinking Water
Blood
Fre-PreparatioR i
Drying and
gravimetrics
Drying, sieving and
homogenization
Acidified
Stabilized in EDTA
after collection
Preparation ;
Microwave digestion
using 1:1 HNO3: H2O
Microwave digestion
using 1:1HNO,:H2O
Microwave digestion
using 1:1 HNO3: H2O
Addition of matrix
modifier/Triton X-100
solution
Analysis
ICP/GFAA3
GFAA
GFAA
GFAA/ASVb
Samples with lead concentrations below the limit of quantitation of the ICP instrument were
analyzed by GFAA.
ASV used in addition to GFAA for rapid reporting of blood lead
30
-------
6.0 DATA PROCESSING AND STATISTICAL ANALYSIS PROCEDURES
6.1 Data Processing
Data analyzed as a part of this study were derived from field collection forms, laboratory
instruments, and questionnaires. Raw data of all types were transferred to the data manager who
uploaded the data to a VAXStation 3100 computer for later analysis. A summary of the data
processing steps employed for the three sources of data is presented below.
• The field data consist of all data recorded on the field collection forms for settled dust,
soil, and drinking water samples, as well as room and window inventory data and general
study data. Data were entered twice for verification from the field forms into ASCII data
files by a commercial data entry firm. These raw data files were transferred to the data
management team for management, storage, and later analysis. Field data forms were
checked for completeness and accuracy by the outreach coordinator and data manager prior
to data entry. Data were verified again by laboratory staff from final SAS* file printouts.
• Laboratory data were electronically stored by each laboratory instrument. Gravimetric
data (tared and loaded weights for dust and soil samples) were generated and stored by the
Mettler Balance. Lead concentration measurements for dust samples were made and
recorded by the ICP. Lead content in drinking water, soil, and blood, as well as dust
samples with low lead concentrations, were measured by GFAA. Electronically stored
laboratory data from the Mettler, ICP, and GFAA instruments were imported to Paradox*
(v.4.0) by laboratory staff for tracking of samples. Paradox* data were then converted to
ASCII files by the data management team for uploading to the VAXStation. A SAS*
program read in the laboratory data for environmental and blood samples and created SAS*
data sets for data analysis. The data were verified again by laboratory staff from final
SAS* file printouts.
• Questionnaire data forms were entered twice for verification by a data entry firm into
ASCII data files. These raw data files were verified in-house and transferred to the data
manager. A SAS* program read in the raw data and created SAS* data sets for analysis.
6.2 Data Summary
Environmental dust data from four surface types (perimeter floor, window sill, window
well, and interior entryway) included ine^ch of the first five data collection campaigns (pre-R&M,
post-R&M, two months post R&M, six months post R&M, and twelve months post-R&M) are
included in this report as well as data collected less frequently (i.e., airduct dust, upholstery dust,
soil, and water). Tables 8 and 9 display the types and numbers of 12-month campaign samples
planned, collected, and analyzed for lead, by study groups, for the 104 active houses included in
this report.
31
-------
Table 8: Types And Numbers Of Samples Collected And Analyzed For Lead
(Excluding QC Samples) As A Part Of The 12-Month Campaign
Sample Type
Perimeter Floor
Dust Composite in
Rooms with
Windows
Perimeter Floor
Dust Composite in
Rooms without
Windows
Window Sill Dust
Composite
Window Well Dust
Composite
Interior Entry way
Dust
Exterior Entryway
Dust
Air Duct Dust
Upholstery Dust
TOTAL DUST
Soil Core - drip line
Drinking Water
Venous Blood
GRAND TOTAL
Planned
per
House
2a
1
2a
2a
1
-
!b
b
9
_
I/child
^10
Collected
inJt04
Active
Houses
212
55
207
203
104
-
57
46
884
.
126
1010
Collected :
and i
Analyzed ;
for Lead :
212
55
207
203
104
-
57
46 j
884
-
126
1010
Unavailable
Samples HI
the 184
Active
Houses
0
49C
ld
5e
0
lf
0
56
_
-
-
56
a One composite sample was obtained jper story. Some houses had samples in basements used as living spaces.
b Upholstery samples were collected if air duct samples could not be obtained.
c 49 houses did not have rooms without windows.
d Sills on one story were inaccessible in one R&M I house.
e Wells on one story were inaccessible hi five instances hi R&M I houses.
f Ah- duct & upholstery were inaccessible/not present in one R&M II house (see footnote b).
32
-------
Table 9: Types And Numbers Of Samples Collected By Group (Excluding QC Samples)
As A Part Of The 12-Month Campaign3
Sample Type
Perimeter Floor Dust
Composite in Rooms
with Windows
Perimeter Floor Dust
Composite in rooms
without windows
Window Sill Dust
Composite
Window Well Dust
Composite
Interior Entryway
Dust
Exterior Entryway
Dust
Air Duct Dust
Upholstery Dust
TOTAL DUST
Soil Core - drip line
Drinking Water
Venous Blood
TOTAL
CoBficted
iwiS
Modem
Urbaa
Houses
31
4
30
30
15
11
4
125
_
14
139
Collected
to 14
Previously
Abated
Houses
28
6
28
28
14
5
9
118
_
24
142
Collected
in 25
! R&M I
\ Houses
53b
17
49
45
25
-
10
15
214
.
_
24
238
Collected
In 23
R&M JI
Houses
46
15
46
46
23
15
7
198
.
_
30
228
coBeeted
mil
RAM HI
Houses
54
13
54
54
27
-
16
11
229
.
_
34
263
a Two R&M II houses were reclassified to R&M III on the basis of the actual work done
in the house at the time of intervention.
b Includes two samples collected in basements used as living spaces.
33
-------
Some of the original study households moved or voluntarily withdrew from the study
between the initial and the 12-month data collection campaigns. Table 10 reports the frequency
of moves for each study group and the numbers of replacement households enrolled among
families moving into vacated study houses. Approximately 20 percent (21 out of 107) of the
original families in the 107 original study houses moved prior to the 12-month campaign. By the
end of the 12- month campaign, all but three of these study families were replaced by the next
family that moved into the house. Despite our success in gaining the participation of these new
families, they had fewer eligible children than the original families. By the end of the 12-month
campaign, the study also had gained nine children who were newborns that became of age (^6
months) for blood lead testing and two other eligible children who joined study households. It is
hoped that future campaigns will provide sufficient longitudinal data on children born into study
houses to allow for a separate analysis of this subgroup, which would help to increase our
understanding of the role of R&M interventions in the primary prevention of lead poisoning.
One R&M n house was vacant at the time of the 12-month sampling. None of the houses
included in this report are known to have had any major renovations or repairs during the first
year of follow-up. One R&M I house had its front and back doors replaced due to break-ins that
damaged the original doors, and in another house, the wallpaper was removed by the occupants
from the first floor rooms using a steam process.
6.3 Statistical Analysis
This section describes the statistical methods employed in the analysis of data from the first
year of follow-up. The first section describes the methods used to generate descriptive statistics
and graphical displays of the data. The second section provides an overview of the statistical
method used for the analysis of longitudinal data. The last section describes the use of factor
analysis as a method for combining individual sample readings in a house and specifies the
longitudinal models fitted to the dust and blood data.
SAS36 PROC MIXED software (version 6.09E) was used for longitudinal data analysis.
Interpretation of the estimates obtained by the mixed model obey the usual rules of interpretation
of regression coefficients, i.e., the coefficient of a covariate is the expected change in the response
variable associated with a unit change in the covariate in the presence of the other covariates.
When the covariate is a dummy variable, a unit change in the covariate corresponds to the
expected difference between the response at the level of the covariate compared to the omitted
level.
For data analysis purposes, lead values less than the instrument detection limit (DDL) were
coded as the IDL/\/2.37 For lead values less than the limit of quantitation (LOQ), but greater than
the IDL, the observed value was used in the data analysis. Also, one child in a previously abated
house had a blood lead increase to a concentration of 53 /wg/dL at the 12-month campaign and was
provided with chelation therapy. This child is an outlier in this study and was excluded from the
statistical data analysis relating blood lead to dust lead.
34
-------
Table 10: Family Moves, Reoccupancies, And New Subjects Enrolled Between The Initial
Campaign And The 12-Month Campaign.
Study Group
R&MI
(25 houses)
R&MII
(23 houses)
R&MIII
(27 houses)
Modern Urban
(16 houses)
Previously Abated
(16 houses)
Total
Moved
'
Households
6
6
6
1
2
21b
Chadrea*
16
10
9
4
6
45
Replaced
Households
6
5
6
1
0
18
OnTdren
2
8
8
0
0
18
Other
Children
ftnr ",
member of
bouseboM
0
0
0
0
2
2
ISew&Bnis
3
2
1
0
3
9
Includes children/families who moved although other members of household remained.
This number represents 20 percent of the original study households.
35
-------
Descriptive Statistics
The study outcome variables were dust lead concentration Cug/g), dust lead loading
(yug/ft2), dust loading (rng/ff) and blood lead concentration Oug/dL). The main study variables
included study group, data collection campaign, type of environmental sample (e.g., dust, water),
and surface type (e.g., floor, window sill, window well, entryway, upholstery). A Shapiro-Wilk
Test indicated that the distributions of the dust and blood lead data were skewed.38 As expected,
use of the log transformation reduced the amount of skewness and produced histograms and
boxplots that were approximately normal (Figures 1-13). Descriptive statistics on blood and dust
were produced after transforming the data using the natural logarithm (In).
A further characteristic of the data set is the repeated measures from a house, which violate
the assumption of independence invoked for most analyses. To overcome this problem, a mixed-
effects model was used to account for the correlation of samples within a house. These calculations
result in a better estimate of the mean and confidence interval for the settled dust from floors in
rooms with windows, window sills, window wells, and children's blood. These calculations were
done by study group and surface type.
Descriptive statistics for all dust sample types are presented in Appendix A. Tables 16-18
display descriptive statistics for blood, soil and water. Since multiple observations were available
in each of the houses for settled dust from window sills, and window wells, floors in rooms with
windows, as well as for children's blood, additional analysis was performed using SAS® PROC
MDCED with house as a random effect to address the issue of clustering (i.e. multiple observations
per house). Geometric mean values, standard errors, and 95 percent confidence intervals were
obtained using the intercept models fitted separately for each study group, surface type (floors in
rooms with windows, window sills, window wells), and matrix (dust, blood).
Side-By-Side Boxplots
Side-by-side boxplot figures with median traces are presented in this report as a means of
displaying lead levels across campaigns within and between study groups. In a boxplot display,
50 percent of the data is contained in the box shown in the figure; the bottom of the box is the
lower quartile and the top of the box is the third quartile, the horizontal line inside the box
represents the sample median. The vertical lines extending from the box represent the expected
lower and upper range of the data, based on the variability of the central portion of the data. The
fences are 1.5 interquartile ranges from the upper and lower edges of the box. Extreme values
are indicated by an asterisk.39 The widths of the boxes in any given side-by-side boxplot are
proportional to the number of observations. The descriptive statistics presented hi this report
include "extreme values" that are indicated by the symbol '*' hi the boxplot displays.
36
-------
Statistical Method for Analysis of Longitudinal Data
Statistical methods for the analysis of longitudinal data have developed rapidly over the last
decade.4M6 These methods, which are natural extensions of multiple regression and analysis of
variance, are extremely flexible. Current longitudinal methods allow for the inclusion of random
and fixed effects, longitudinal (time-dependent) covariates and constant covariates, as well as for
discrete and continuous covariates, all in a multiple regression context. In this study, for example,
we have the following types of covariates:
• study group - fixed effect, discrete
• house - random effect, discrete
• dust lead - fixed time dependent continuous covariate
• child - random effect, discrete
• campaign fixed time dependent covariate, discrete
• age of child fixed time dependent covariate
• season - fixed discrete covariate
The response variable modeled was dust lead reading or blood lead concentration (log-
transformed). These response variables, as well as their associated covariates, have been and will
be observed at times described in Table 5.
For the dust lead measurements let Y; denote the vector of responses over time for the 7-th
house, i.e., Yj is an n( x vector of the form Yj = (yji,yj2,---,yin)T where yy is the response for the
7-th house at time tj and "T" stands for the transpose operation. Then, the general form of the
model is:
Yj = Xjp + Z& + 6i (Eq.l)
where Xj is an n. xp matrix of covariate values for the fixed effects, P is a pxl vector of
parameters for the fixed effects, Zj is an r^xq matrix of covariate values for the random effects,
b; is a qxl vector of random effect parameters and ^ is an i^xl vector representing random error.
Estimates of the parameters in the overall model are obtained using the methods outlined
in published papers."^ The essential feature of these methods is the use of weighted least squares
with a "working" estimate of the covariance matrix followed by iteration with an updated estimate
of the covariance matrix until convergence. The estimate of the variance-covariance matrix of the
fixed effects is robust, in the sense that it is consistent, regardless of the form of the "working"
estimate of the covariance matrix. The model for blood lead is similar to the above model,
specified for each child.
Our primary interest in this study is in the parameters of the model that represent the effect
of R&M interventions on dust lead and blood lead. The fact that this model allows estimation of
these parameters in the presence of heterogeneity between houses and temporal correlation, and
produces variance estimates that are robust, is extremely important.
37
-------
The general nature of the model makes it ideal for a study of this type where there is the
potential for unbalance. Since the model is house-specific or child-specific, depending on whether
dust lead or blood lead is being modeled, we do not require that the number of observations
through time be equal. Thus, should a child move or otherwise be eliminated from the study, the
house data can be analyzed while the data for that child can be included up to the point of
departure. Should another child be entered into the study at that house, his or her blood lead
readings can be included hi the blood lead analysis for the remainder of the study, thus providing
partial information for that child. The common residence of the children is included hi the house
covariate, which allows for correlation structure between these observations.
Age-related effects hi the analysis of blood lead concentration responses need to take into
account the fact that blood lead is not linearly related to age, since it tends to increase between six
months and two years and decrease slowly among children over two years of age. This is done
by the use of linear and quadratic terms for age in the model. The presence of several children
in a house, which introduces another source of correlation, (i.e. between children hi the same
house) is accounted for by using the house as a random effect, which introduces the required
correlation.
Specifications of Longitudinal Models for Dust
In the analysis of the data from the first year of follow-up, we have fit the statistical models
proposed in the Quality Assurance Project Plan.2 The results of the compositing self study
indicated that an overall measure of lead exposure could be considered with little loss of
information.47 Factor analysis confirms this as described below. This was true for both dust lead
concentrations, lead loadings and dust loadings. These results suggest that the readings from
multiple sample sites in a house can be combined to produce an overall measure to use as a
covariate hi the model relating environmental lead to blood lead. Consequently, we have explored
the use of factor analysis as a method for combining individual sample results. The use of the
results of exploratory factor analysis to guide the construction of variables for analysis is a
standard approach used in data analysis. Our general approach is outlined below:
• Data for floors in rooms with windows, window sills and window wells were used hi the
analysis. These data were composited across stories in a house in the calculation of
weighted averages for each of the three dust endpoints, for each house, and for each
campaign.
• The weighted averages were transformed using natural logarithms.
• Factor analysis was first performed for each dust endpoint by campaign and then again not
broken out by campaign. The latter results were then used in the longitudinal analysis.
These steps were repeated anew for each analysis because of the different combination of
study groups and campaigns for intervention and for control houses.
38
-------
Occasionally, a composite was incomplete because a sill or well was not accessible. On
a very few occasions, all sills or wells in a single story were inaccessible and, thus, no composite
value was available. If both first and second story composites were missing, no attempt was made
to estimate missing data.
The results indicate that:
• The first factor (factorl) accounts for 64 percent to 82 percent of the variability of
environmental dust lead across campaigns, when all five groups are analyzed together, and
54 percent to 65 percent of the variability, when the three R&M groups are analyzed
separately (Table 11).
• The second factor (factor2) characterizes the difference between the floor lead
measurements and the window sill and window well lead measurements and accounts for
12 percent to 26 percent of the variability, when all five groups are analyzed together, and
22 percent to 31 percent of the variability, when the three R&M groups are analyzed
separately (Table 11).
Thus far, the percentages of the variability of the dust readings accounted for by the factor
loadings have remained relatively stable over study groups and campaigns (Table 11). The factor
patterns for all five groups also were stable over time (Table 12). The factor patterns for the three
R&M groups by surface type across campaigns also were consistent over time, except for factor2
at the initial campaign (Table 13). The latter may be different due to the fact that half of the R&M
houses were vacant at the time of the initial campaign and/or to an intervention effect on factor
patterns. Table 13 also shows that the factor patterns are consistent within campaigns for the three
types of dust measurements. Both factorl and factor2 are normally distributed.
Given the stability of the factors over time, we used them as the variable to measure
environmental lead levels. The first factor was used as the dependent variable in the longitudinal
data analysis of the three dust endpoints. This factor reflects the campaigns up to, and including,
the 12-month campaign. We found that the use of the first factor in the data analysis explains
more of the variability in the dust endpoints, as compared to raw average or to weighted average
measures.
39
-------
Table 11: Variability Accounted for by Factor Loadings Across Campaigns
Five Study Groups Combined:
\ Dust Measure
Lead Loading
Lead Concentration
Dust Loading
Initial
i Campaign
factorl factor2
.81 .14
.82 .12
.65 .24
Six-Month
Cainpaign
factorl factor2
.68 .23
.73 .20
.60 .24
12-Month
Campaign
factorl factor2
.64 .26
.65 .22
.55 .32
Three R&M Groups:
i Dust Measure
Lead Loading
Lead Concentration
Dust Loading
i Campaign :
factorl factor2
.56 .29
.57 .28
.58 .26
Campaign
factorl factor2
.65 .22
.59 .29
.55 .26
Campaign
factorl factor2
.59 .31
.59 .28
.57 .31
SMfop*
K' -6
factorl factor2
.65 .26
.64 .27
.62 .25
12-Montfe
Campaign
factorl factor2
.58 .31
.54 .31
.56 .32
40
-------
Table 12: Factor Patterns For The Five Study Groups Across Campaigns
Dust Measure
+
Lead Loading
Lead
Concentration
Dust Loading
Surface Type
Floor
Sill
Well
Floor
Sill
Well
Floor
Sill
Well
Campaign
Initial
factor 1 factor!
0.87 0.48
0.95 -0.09
0.90 -0.37
0.88 0.45
0.93 -0.09
0.94 -0.35
0.76 0.62
0.90 -0.06
0.80 -0.52
Six-Month
factor 1 factor2
0.71 0.70
0.90 -0.23
0.87 -0.34
0.73 0.67
0.91 -0.25
0.90 -0.30
0.71 0.70
0.81 -0.24
0.79 -0.38
la-Montfti
f actor Ifactor2
0.68 0.72
0.89 -0.17
0.84 -0.40
0.74 0.67
0.85 -0.30
0.85 -0.28
0.49 0.85
0.88 -0.08
0.80 -0.44
41
-------
Table 13: Factor Patterns For R&M Groups Across Campaigns
Dust Measure
Lead Loading
Lead
Concentration
Dust Loading
Surface
Type
Floor
Sill
Well
Floor
Sill
Well
Floor
Sill
Well
Campaign
Initiat
factorl factor2
0.82 -0.29
0.82 -0.27
0.55 0.83
0.76 -0.47
0.82 -0.14
0.63 0.76
0.82 -0.28
0.81 -0.33
0.65 0.76
Post-Intervention
factorl factor2
0.76 0.61
0.87 -0.07
0.79 -0.51
0.51 0.86
0.88 -0.19
0.85 -0.32
0.73 -0.55
0.80 -0.08
0.68 0.69
TworMoirfh
factorl factor2
0.49 0.87
0.90 -0.13
0.86 -0.36
0.57 0.82
0.86 -0.30
0.87 -0.24
0.49 0.85
0.88 -0.09
0.80 -0.43
Six-Month
factorl factor2
0.57 0.82
0.88 -0.30
0.90 -0.23
0.53 0.84
0.90 -0.22
0.89 -0.29
0.66 0.74
0.86 -0.18
0.81 -0.41
12-Month
f actor Ifactor2
0.40 0.91
0.90 -0.12
0.87 -0.29
0.40 0.91
0.86 -0.22
0.86 -0.21
0.38 0.92
0.89 -0.09
0.85 -0.32
42
-------
Consequently, the following models were fit to the dust data (see Table 14 for definitions of
variables)*1
Environmental Model:
//7/"*/"/! f* 7 — rt -4- /i ^c^/rc/iw -4- /? ^O'f/iy/n
jULLuri^i — p0 T p1 zcuburiij -r f>2 8rtJUPik
+ @3* campaign + f34 group %* campaign
+ b-*houset + eijkl (Eq.2)
where,
"i" refers to house, "j" to season, "k" to study group, "I" to campaign, group*campaign to the
interaction of group and campaign. Following standard practice, regression coefficients
corresponding to "fixed effects" are denoted by Greek letters, while regression coefficients
corresponding to "random effects" are denoted by non-Greek letters (e.g. b).
This model was fit to the lead concentration, lead loading and the dust loading data. The
models were run using all five study groups and then again using just the three R&M groups in
order to include the post-intervention and two-month campaign data in the analysis.
Specifications of longitudinal models for blood lead
To address the study objectives with regard to blood lead, we fit two main types of models
to the data. The first model, referred to as the exposure model, was used to characterize the
relationship between blood lead and dust lead (both dust lead concentrations and lead loadings).
In this model, the two dust lead factors were included as dependent variables, along with
demographic and behavioral variables. The second model, referred to as the comparison model,
was used to investigate blood lead concentrations across groups and within groups over tune.
d Our exploratory analysis indicated that the covariance structure varied little over time.
Therefore, when fitting the longitudinal models using SAS Proc Mixed, we used the random statement
that built in the necessary covariance structure.
43
-------
Table 14: Definitions of Variables
¥ariatjJs
f actor 1
factor!
age
mouthing
season
Definition
Linear combination of floor, window sill and window well data
(composite measure of exposure in a house).
Linear combination of floor, window sill and window well data
(represents the difference between floor and window values).
Child's age in months
The sum of four questionnaire variables dichotomized into a
low/high variable
Fall: September 21 through December 20
Winter: December 21 through March 20
Spring: March 21 through June 20
Summer: June 21 through September 20
44
-------
The two models are as follows:
Exposure Model
ln(PbB)iklm = P0 + p^factorliklm + p*factor2iUn
+ P3*ageiklm + j34*age2Mm + p5*summerik[m
+ P^campaig^
+ bfhousei + bm(I)*childm(I) + eiklm (Eq.3)
where,
"i" refers to house, "k" to group, "1" to campaign, "m" to child within house, group *campaign
to the interaction of group and campaign. Regression coefficients corresponding to "fixed effects"
are denoted by Greek letters, while regression coefficients corresponding to "random effects" are
denoted by ordinary letters (e.g. b).
The initial campaign blood and dust lead values for children who moved into the vacant
R&M II and R&M HI houses after intervention were excluded from the exposure model. Their
initial blood lead values at the time they moved in reflect body burdens associated with exposures
in their past living environments, not in their new home environments.
Study group was left out of the exposure model because of its association with our
exposure variables. This model was run using all five study groups and then again using the three
R&M groups. Due to the consistency of the factor patterns noted above across campaigns, the
interaction between f actor 1 and campaign and between factor2 and campaign were not found to
be statistically significant and were dropped from later applications of the model. Other variables
such as gender and mouthing variables were added to this basic model.
Comparison Model
ln(PbB)iklm = P0+ P^ageiklm + P2*age2iklm + P3* summer iklm +
+ P5*groupk + p^campaigrii
+ b^housei + bm(I*childm(I) + eiklm (Eq.4)
(Refer to the exposure model above for an explanation of the notation used in Eq.4).
The comparison model was fit separately for children with blood lead concentrations
<20/ng/dL and ^20/xg/dL. According to CDC guidelines, children with blood lead concentrations
^20/ig/dL should be referred for medical evaluation and management.10 Table 15 displays the
numbers of children included in these models by initial blood lead concentration and by group.
Although most children with baseline blood lead concentrations ^20/xg/dL were in R&M II and
R&M HI, the variances of baseline blood lead concentrations across the three groups were
45
-------
essentially the same. Descriptive statistics and box plots of baseline blood lead concentrations
by study group are displayed in Appendix B.
Table 15: Numbers of Children With Initial Blood Lead <2<>Mg/dL and *20jig/dL
Study group
R&MI
R&MH
R&Mm
Previously Abated
Modern Urban
Initial Blood Lead
<20#g/dL
(u)
34
27
29
20
20
Initial Blood Lead
s2fl/tg/dL
(a)
1
7
8
3
0
The group*campaign interaction term and the gender and mouthing variables were not
statistically significant. It should be noted that although the model includes a term for child within
house, there were in actuality small numbers of households that had more than one child per
house.
Measurement Error
A number of researchers have raised the issue of measurement error in environmental
variables. Measurement errors in the covariates or explanatory variables can affect the magnitude
of the estimated regression coefficients in linear models. This effect is called attenuation and
implies that observed effects are underestimated by an amount related to the magnitude of the
errors hi the covariates. The modeling approach used in our analysis uses factor analysis to derive
environmental measures from the basic environmental samples. The use of latent variables implicit
in the measurement error models is thus present in our approach where these variables are
explicitly treated as part of the model. While measurement error is present in the environmental
samples, we believe that the approach using factor analysis adequately accounts for the presence
of measurement error.
46
-------
7.0 RESULTS
This section is divided into three parts. The first part provides descriptive statistics on
environmental data and blood data from the first year of follow-up, including a series of side-by-
side boxplot figures with median traces to graphically display trends across time. The second part
presents descriptive statistics on data derived from the 12-month campaign and an analysis of the
correlations between children's blood lead concentrations and their dust lead exposure (section
7.2). These descriptive statistics do not take into account season or any other potential covariates.
Part three presents findings of the longitudinal data analysis and includes a summary of the
statistical significance of trends in dust lead and blood lead over tune within and across groups
(section 7.3).
7.1 Descriptive Statistics For The First Year Of Follow-Up
Side-by-Side Boxplots With Dust Data
Figures 1-12 show the distributions of dust lead loadings, dust lead concentrations, and
dust loadings by study group across campaigns for each of four main surface types. The boxplots
are displayed on the log scale, due to the wide ranges of dust values between groups and within
groups across time (see section 6.3 for an explanation of the components of a boxplot). These
figures reveal the following trends:
• Median traces for dust lead loadings across surface types show a pattern of maximally
reduced levels at post-intervention. This pattern is most pronounced for R&M III houses,
intermediate for R&M II houses, and smallest for R&M I houses. At two months, lead
loadings were increased over post-intervention levels, but they were below pre-intervention
levels. They remained below pre-intervention levels through six months and 12 months of
follow-up. At six months and 12 months, median lead loadings were relatively stable, or
moderately increased, in R&M I and II houses across surface types, while in R&M III
houses, median lead loadings tended to be relatively stable or moderately decreased
(Figures 1-4). Deviations from this pattern were evident for floors and entry ways of R&M
I and R&M II houses in which lead loadings did not increase at two months.
• Median traces for dust lead concentrations reveal a downward trend at post-intervention
and at two months across sample types. This trend was most pronounced in R&M III
houses, intermediate hi R&M n houses, and least pronounced in R&M I houses. At six and
12 months, lead concentrations remained relatively stable or slightly increased hi R&M I
houses and relatively stable or moderately decreased in R&M II and R&M III houses
(Figures 5-8).
• The median traces for dust loadings show a pattern of reductions at post-intervention that
was greatest in R&M HI houses, intermediate in R&M II houses, and smallest in R&M I
houses (Figures 9-12). At two months, dust loadings tended to reaccumulate over the post-
47
-------
intervention loadings, but median loadings generally remained below pre-intervention
levels throughout the first year of follow-up.
• The modern urban and previously abated control houses show a pattern of relatively stable
median lead loadings, lead concentrations, and dust loadings. There is a slight downward
trend at six months and 12 months in lead loadings and dust loadings (Figures 1-12).
Side-By-Side Boxplots Of Blood Lead Concentrations
Figure 13 provides boxplot displays of unadjusted blood lead concentrations by study group
for children with initial blood lead concentrations <20 /xg/dL. The child with a blood lead
concentration of 53 jag/dL in the previously abated group at 12 months does not appear on the
figure. The median traces for all five study groups, unadjusted for covariates, indicate little change
over time.
"Hair Clip" Line Plots With Blood Lead Concentrations for Individuals
Figures 14-18 are "hair clip" line plots of blood lead concentrations for individual children
in each of the five study groups. These figures display each study child's unadjusted blood lead
concentrations during the first year of follow-up. As seen in these plots, most of the children with
baseline blood lead concentrations >20 /ig/dL were in the R&M II and R&M El study groups.
Children with baseline blood lead concentrations >20 /ig/dL experienced reductions in their blood
lead concentrations over time, while those with baseline blood lead concentration < 20 /xg/dL
tended to remain <20 /ig/dL during the first year of follow-up.
7.2 Descriptive Statistics At The 12-Month Campaign
Blood Lead Concentrations At 12 Months
Table 16 provides descriptive statistics for children's blood lead concentrations by group
at the 12-month campaign. The unadjusted geometric mean blood lead concentrations were 8
/xg/dL for children in R&M I houses, 11 pig/dL for children in R&M n houses, and 12 /xg/dL for
children hi R&M HI houses, 12 /ig/dL for children in previously abated houses and 3 /ig/dL for
children in modem urban houses. The mean age of children across the five groups at the 12-month
campaign ranged from 39 months to 44 months.
Dust Lead Loadings. Lead Concentrations And Dust Loadings At The 12-Month Campaign
Descriptive statistics for settled dust at the 12-month campaign are graphically displayed
as bar graphs showing geometric mean dust lead loadings (jttg/tf), dust lead concentrations (/*g/g),
and dust loadings (mg/ft2) by group and by surface type in Figures 19 to 21. Tables with
descriptive statistics (geometric mean, n, minimum, maximum, standard deviation) for lead
loadings, lead concentrations and dust loadings by group and by surface type are in Appendix A.
48
-------
Table 16: Descriptive Statistics For Blood Lead Concentrations By Group At The 12-Month Campaign
Study Group
R&MI
R&MII
R&M III
Previously Abated
Modem Urban
, ft
24
30
34
24
14
Minimum
WdL)
2
4
4
1
2
Maximum
>g&L)
20
31
30
53"
6
Geometric
m*m
fcBW
8
11
12
12
3
S>1>, on
fog scale
0.538
0.422
0.480
0.731
0.371
tower £5% CI
Itoefifcf
, <«6ftH3
6
10
10
8
3
VHprftwa
ferSM"
(pg/«t4
10
13
14
18
4
a
b
GM values and confidence intervals were obtained from SAS* PROC MIXED
This outlier was excluded from the longitudinal data analysis (see section 6.3)
49
-------
R&M Study: 12 Month Final Report
Figure 1 Dust Lead Loadings (PbD in ug/ftA2) across Campaigns for Floor Surfaces
Q
Xi
o"
§
7
6
5
4
3
2
1
0
-1
RM-I
7
6
5 ^
4
3
2 ^
1
0 -
-1 -
RM-II
fl-
-a
7
6
5
4 -
3 -
2 -
1 -
0 -
-1 -
RM-III
-a
IN PI 2M
6M
12M
IN PI 2M 6M
12M
IN PI 2M
6M
12M
Q
£
o
8s
7
6
5
4
3
2
1
0
-1
Previously-Abated
IN
6M
7 •
6 •
5 -
4 •
3 -
2 -
1 •
0 •
-1 -
Modern
12M
IN
6M
12M
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
6M = 6 Months
12M = 12 Months
50
-------
R&M Study: 12 Month Final Report
Figure 2 Dust Lead Loadings (PbD in ug/ftA2) across Campaigns for Window Sill Surfaces
Q
£
o
§>
7
6
5
4
3
2
1
0
-1
RM-I
RM-II
RM-III
7
6
5
4 -
3 •
2
1
0 ^
-1
7 -
6 -
5 -
4 -
3 •
2 -
1
0
-1
B
IN PI 2M 6M
12M
IN PI 2M 6M
12M
IN PI 2M
6M
12M
8"
7
6
5
4
3
2
1
0
-1
Previously-Abated
Modern
IN
6M
a
12M
7 -
6 -
5 -
4 -
3 -
2 -
1
0 -
-1 -
IN
6M
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
6M = 6 Months
12M = 12 Months
12M
51
-------
R&M Study: 12 Month Final Report
Figure 3 Dust Lead Loadings (PbD in ug/ftA2) across Campaigns for Window Well Surfaces
Q
.Q
a^
o
^3>
o
7
6
5
4
3
2
1
0
-1
RM-I
-6-
7 ^
6
5
4
3
2 -
1 -
0
RM-II
RM-III
-fr
7
6
5
4
3
2
1
0 -
-1 -
IN PI 2M 6M
12M
IN PI 2M 6M
12M
IN PI 2M
6M
12M
Q
_Q
7
6
5
4
3
2
1
0
-1
Previously-Abated
7 -
6
5
4
3 ^
2
1
0
-1 H
Modern
9-
IN
6M
12M
IN
6M
•a
12M
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
6M = 6 Months
12M = 12 Months
52
-------
R&M Study: 12 Month Final Report
Figure 4 Dust Lead Loadings (PbD in ug/ftA2) across Campaigns for Interior Entryway Surfaces
RM-I
RM-II
RM-III
Q
J3
Q.,
O
7
6
5 -
4 -
3 -
2 -
1
0
-1 H
fl
7 -
6 -
5
4 -
3 -
2 -
1 -
0 -
-1
f—fi Q
7 -
6
5 -
4 -
3 -
2 -
1 -
0 -
-1 -
IN PI 2M 6M
12M
IN PI 2M 6M
12M
JN PI 2M 6M
12M
Q
.a
a.
Previously-Abated
7 -
6 -
5 -
4 •
3 •
2 -
1 -
0 -
-1 -
rt
u
I_J— »
T^
(I
U
*
*
Br—
,
l_*J
Modern
IN
6M
12M
7 •
6 •
5 -
4 -
3 •
2 •
1 -
0 -
-1 -
, I-T-I
n -fl
rp FT
I .j.,
*
4-
^-.
*
IN
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
6M = 6 Months
12M = 12 Months
6M
12M
53
-------
R&M Study: 12 Month Final Report
Figure 5 Dust Lead Concentrations (PbD-C in ug/g) across Campaigns for Floor Surfaces
RM-I
IN PI 2M
6M
12M
RM-I
IN PI 2M
6M
12M
RM-III
6 •
5 •
? 4-
f 3.
0
3 2'
1 -
o -
! *
T1 pi r— , T< *y
S~I?H^ n n
-J-- i S ~5~ bd
1r
* *
6 •
5 •
4 •
3 •
2 -
1 -
0 -
EH:k6 h T
^ u^ : y q
6
5 •
4 •
3 •
2 •
1 •
0 -
1-71
• i *
pi : T1 * *
i&-0— -o— _S
|J-1 "-I-1 LJ UJ ' -| 1
IN PI 2M
6M
12M
O
Q
£
o
Previously-Abated
6 •
5 •
4 •
3 •
2 -
1 •
0 -
.
n
u
^*-i
*
n
u
*
a1
i_i_j
IN
6M
12M
6 -
5 -
4
3
2
1 -
0 -
e-
Modern
IN
6M
12M
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
6M = 6 Months
12M = 12 Months
54
-------
R&M Study: 12 Month Final Report
Figure 6 Dust Lead Concentrations (PbD-C in ug/g) across Campaigns for Window Sill Surfaces
RM-I
IN PI 2M
6M
12M
RM-II
IN PI 2M
6M
12M
RM-III
6
5 •
§ «•
n
t 3 •
o
s 2-
1 •
o •
*
6~BHi n -3
"-^ UJ LJ 1 l UJ
.-L*
6 •
5 •
4 •
3 -
2 -
1 -
0 •
'-T ry, ry, "T"
9--C1 r1! h rj
nnj — g — g
6 -
5 •
4 -
3 •
2 •
1 -
0 -
»~T~*
IN PI 2M
6M
12M
Previously-Abated
IN
6M
12M
IN
Modern
oglO(PbD-C)
6 i
5 •
4 •
3 -
2 •
1 •
0 -
*
D h -—5
s-1 u M
•-• * j i— «— *
• * •
6 •
5 -
4 •
3 •
2 •
1 -
0 -
~" n n
£3~ ^ M
*
6M
12M
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
6M = 6 Months
12M = 12 Months
55
-------
R&M Study: 12 Month Final Report
Figure 7 Dust Lead Concentrations (PbD-C in ug/g) across Campaigns for Window Well Surfaces
RM-I
IN PI 2M 6M
12M
RM-I
IN PI 2M 6M
12M
RM-I
6 •
5 •
2 4-
£ 3-
0
3 *•
1 -
o -
T I"r' •— i ""' *"•"
Bl J rh I-** rh
. uT H & W
*""* bJ_i I_LJ *-*-•
*
6
5 •
4 •
3 •
2 -
1 -
0 -
m m * *
9-4] fS PI ri
•j- u u y g
i_t_i i * * I
6 •
5
4
3 •
2
1 -
0 -
,-r-l
|
fl T 4- T
>J-M iTl rS rn _i_
IH — 1 r^
* LJ trJ UJ fcp
* - * • * • •-1-'
*
IN PI 2M 6M
12M
O
Q
§"
6
5
4
3
2 -
1 •
0 -
Previously-Abated
IN
6M
12M
6
5
4
3
2
1 •
0 -
e-
Modern
-O-
IN
6M
12M
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
6M = 6 Months
12M = 12 Months
56
-------
R&M Study: 12 Month Final Report
Figure 8 Dust Lead Concentrations (PbD-C in ug/g) across Campaigns for Interior Entryway Surfaces
9
Q
£
o
s?
co in Tt co CM T- o
RM-I
-Q
6 •
5
4
3 -
2 -
1 -
0 -
RM-II
IN PI 2M
6M
12M
IN PI 2M 6M
12M
6
5
4
3
2
1
0
RM-III
B-
IN PI 2M 6M
12M
.0
0-,
o
1?
6
5
4
3
2
1
0
Previously-Abated
IN
6M
12M
6
5
4
3 -
2 -
1 -
0 •
Modern
IN
6M
12M
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
6M = 6 Months
12M = 12 Months
57
-------
3-
2-
1 -
Q
o 0-
-1-
-2-
-3-
R&M Study: 12 Month Final Report
Figure 9 Dust Loadings (DL in mg/ftA2) across Campaigns for Floor Surfaces
RM-I
-6-
-9
2-
1 -
0-
-1-
-2-
-3-
RM-II
3-
2-
1 -
0-
-1 -
-2-
-3-
RM-III
IN PI 2M
6M
12M
IN PI 2M
6M
12M
IN PI 2M
6M
12M
IN
Previously-Abated
6M
12M
Modern
3-
2-
1 -
0-
-1-
-2-
-3-
T"1 r-^i
n n n
u u td
. ! . 1 1 1 «. ni !
3-
2-
1 -
0-
-1-
-2-
-3-
•
§ $- 1
t_>_l
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
6M = 6 Months
12M = 12 Months
IN
6M
12M
58
-------
3-
2-
1 -
o 0 -
-1-
-2-
-3-
R&M Study: 12 Month Final Report
Figure 10 Dust Loadings (DL in mg/ftA2) across Campaigns for Window Sill Surfaces
RM-I
-B-
3-
2-
1 -
0-
-1-
-2-
-3-
RM-I
-9-
-s
3 -
2-
1 -
0-
-1 -
-2-
-3-
RM-III
IN PI 2M
6M
12M
IN PI 2M
6M
12M
IN PI 2M
6M
12M
3-
2-
1 -
Q
o 0-
-1 -
-2-
-3-
Previously-Abated
Modern
3i
2-
1 -
0-
-1-
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
6M = 6 Months
12M = 12 Months
IN
6M
12M
IN
6M
12M
-------
8
3-
2-
1 -
0-
-1-
-2-
-3-
R&M Study: 12 Month Final Report
Figure 11 Dust Loadings (DL in mg/ftA2) across Campaigns for Window Well Surfaces
RM-I
3-
-H
-2H
-3H
RM-II
1
0
-1 -
-2-
-3-
RM-III
IN PI 2M
6M
12M
IN PI 2M
6M
12M
IN PI 2M
6M
12M
3-
2-
1 -
0
o OH
8
-H
-2H
-3^
Previously-Abated
3-
2-
1 -
0-
-1-
-2-
-3-
Modern
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
6M = 6 Months
12M = 12 Months
IN
6M
12M
IN
6M
12M
60
-------
3-
2-
1 -
o 0-
8
-1-
-2-
-3-
R&M Study: 12 Month Final Report
Figure 12 Dust Loadings (DL in mg/ftA2) across Campaigns for Interior Entryway Surfaces
RM-I
3-
2-
1 -
0-
-1-
-2-
-3-
RM-I
3 -
2-
1 -
-2-
-3-
RM-III
IN PI 2M
6M
12M
IN PI 2M
6M
12M
IN PI 2M
6M
12M
Previously-Abated
Modern
3-
2-
1 -
^7
O
o 0-
f
-1 -
-2-
-3-
i
•
-^
i --•—*
—
.
r-i-i
!
-i-
__
1 - •
I
h
_
" M
i_^j
3-
2-
1 -
0-
-1-
-2-
-3-
""•"* ^
i^i n n
|_| M' H
•-^ "^^
•
Campaigns
IN = Initial
PI = Post Intervention
2M = 2 Months
AM - ft K/lnnthc
OIV1 *~ D IVIOIHMo
12M = 12 Months
IN
6M
12M
IN
6M
12M
61
-------
R&M Study: 12 Month Final Report
Figure 13 Blood Lead Concentrations (PbB in ug/dL) for Children with Initial Blood Pb < 20 ug/dL
RM-I
IN 2M
6M
12M
RM-II
IN 2M 6M
12M
RM-III
2.0-
1.5-
CQ
5- 1.0-
0.5-
o.o-
-_.i.
Br~"~i I~~^ i
: 2 P— fl
* * *
*
2.0-
1.5-
1.0-
0.5-
o.o-
t—l— « l—r— I t— — 1
— PI — n pi
U LJ M
*
*
2.0-
1.5-
1.0-
0.5-
0.0-
^-1-1
' ' ' l~'
R—fi — H R
LjJ "-pJ
* * '
*
IN 2M 6M
12M
Previously-Abated
IN
6M
12M
Modern Urban
2.0-
1.5-
s
| 1.0-
0.5-
o.o-
*
*-~i *
Br-2-!
LJ p
*
2.0-
1.5-
1.0-
0.5-
o.o-
*
IN
6M
12M
Campaigns
IN = Initial
2M = 2 Months
6M = 6 Months
12M = 12 Months
62
-------
U)
40
35.
30.
20 _
§
Q
-o
§
s
10J
5.
0,
Figure 14: Repair & Maintenance Study - 12 Month Report
Children's Blood Lead Levels Across Time - R&M I Houses
Initial, 02, 06 and 12 month campaigns
Note: Each Bne represents an individual chfld.
Second family move-ins and single blood lead values excluded.
Dec-1992 Jun-1993 Dec-1993 Jun-1994 Dec-1994
Sample Collection Date
Jun-1995
Dec-1995
Jun-1996
-------
Figure 15: Repair & Maintenance Study - 12 Month Report
Children's Blood Lead Levels Across Time - R&M II Houses
Initial, 02, 06 and 12 month Campaigns
40.
35 _
30 _
1
^
«J
20_J
s
•g
o
3
10 _
5_
0.
B
Note: Each line represents an individual child.
Second family move-ins and single blood lead values excluded.
Dec-1992
Jun-1993 Dec-1993
Jun-1994
Dec-1994
Jim-1995
Dec-1995
Jun-1996
Sample Collection Date
-------
40.
Figure 16: Repair & Maintenancestudy - 12 Month Report
Children's Blood Lead Levels Across Time - R&M III Houses
Initial, 02, 06 and 12 month Campaigns
35 _
30 _
25 _
c
o
20 _
J
T3
O
O
15 _j
5_
Note: Each line represents an individual child
Second family move-ins and single blood lead values excluded
The blood lead value that appears as 40 is a value of W ug/dL.
-€>
Deo-1992 Jun-1993 Dec-1993 Jun-1994 Dec-1994
Blood Collection Date
Jun-1995
Dec-1995
Jun-1996
-------
40.
Figure 17: Repair & Maintenance Study -12 Month Report
Children's Blood Lead Levels Across Time - Modern Urban Houses
Initial, 06 and 12 month Campaigns
35 _
Note: Each line represents an individual chfld.
Second family move-ins and single blood lead values excluded.
30 _
•a
ON
20 _
§
U
o
o
CQ
5_
Dec-1992
Jun-1993
Dec-1993
Sample Collection Date
Jun-1994
Dec-] 994
-------
Figure 18: Repair & Maintenance Study - 12 Month Report
Children's Blood Lead Levels Across Time - Previously Abated Houses
Initial, 06 and 12 month Campaigns
40.
35 _
30 _
J>25_
I
20 _
§
u
•g
o
ffl
5_
0.
Note: Each fine represents an individual child.
A high blood lead value (53 ug/dL, 12 month) was excluded to facilitate plotting
Second family move-ins and single blood lead values excluded.
Dec-1992
Jun-1993
Dec-1993
Sample Collection Date
Jun-1994
Dec-1994
-------
Figure 19: Geometric Mean Dust Lead Loadings By Surface Type And Study Group
At The 12 Month Campaign
100000
CD
CO
(D
.2
is
o
-------
Figure 20: Geometric Mean Dust Lead Concentrations By Sample Type and Study
Group At The 12 Month Campaign
100000
0)10000
1000
100
10
CO
CD
o
CD
O
0.1
(ED R&M Level I EH R&M Level II • R&M Level III [^ Previously Abated D Modern Urban)
Air Duct
Floor Floor, Interior Upholstery Window Window
No Windows Entrance Sill Well
Surface Type
69
-------
Figure 21: Geometric Mean Dust Loading By Sample Type and Study
Group At The 12 Month Campaign
100000
R&M Level I E3 R&M Level II • R&M Level HI M Previously Abated D Modern Urban]
Air Duct
Floor Floor, Interior Upholstery Window Window
No Windows Entrance Sill Well
Surface Type
70
-------
Figures 19-21 show that air ducts and window wells had the highest lead loadings and dust
loadings among the various surfaces types across study groups. Lead concentrations tended to be
highest for window wells and window sills.
Geometric mean lead loadings were <470 jug/fl2 across groups and surface types at the 12-
month campaign, except for air ducts in all groups and window wells in R&M I, R&M II and
previously abated houses. Geometric mean air duct lead loadings ranged from 856 ^g/ft2 in
modern urban houses to 15,237 /tg/ft2 in R&M II houses. For window wells, geometric mean lead
loadings ranged from 208 jig/ft2 in modern urban houses to 16,698 ^g/ft2 in R&M I houses. In
R&M I houses, the geometric mean lead loadings were 94 jig/ft2, 76 /ng/ft2 in R&M II houses,
and 50 ^g/ft2 in R&M III houses for floors in rooms with windows. When measuring window
sills, the geometric mean dust lead loadings were 470 jig/ft2 for R&M I houses, 237 pig/ft for
R&M II houses, and 29 /xg/ft2 for R&M III houses. Geometric mean lead loadings for window
wells were 16,698 /ig/ft2 in R&M I houses, 2,587 /ig/fi in R&M II houses, and 220 jig/ft in
R&M III houses.
Geometric mean dust lead concentrations across all groups and surface types at 12 months
were < 1,500 /xg/g, except for window sills in R&M I houses, which were 6,964 /*g/g, and 3,165
/*g/g in R&M II houses, and window wells in R&M I houses (20,921 jug/g), R&M n houses
(4,989 /ig/g) and previously abated houses (3,031 /ig/g). At 12 months, geometric mean dus_t
loadings by group and by surface type were all < 800 mg/fl2, except for air ducts, which ranged
from 8,474 to 19,000 mg/ft2.
Modern urban houses continued to have the lowest lead loadings at the 12-month campaign.
Geometric mean lead loadings were <30 /xg/ft2 across surface types, except for window wells (208
fjig/ft2) and air ducts (856 ^ig/ft2). At 12 months, R&M I houses had statistically significantly
higher geometric mean lead loadings for floors in room with windows (94 /-tg/ft2), for window
sills (470 Mg/ft2), and for window wells (16,698 /zg/ft2), compared to R&M m houses (50 /xg/ft2
for floors in rooms with windows, 29 /-tg/ft2 for window sills, and 220 jug/ft2 for window wells).
Geometric mean lead loadings in R&M II houses were intermediate (76, 237, and 2,587 /ig/fi2,
respectively).
At 12 months, modern urban houses continued to have the lowest geometric mean lead
concentrations across all surface types (<440 /^g/g). The geometric mean lead concentrations for
interior entryways and interior floors across the other four study groups were higher and were not
statistically different from each other. R&M I houses had statistically higher geometric mean lead
concentrations for window sills (6,964 /-ig/g) and for window wells (20,921 /xg/g) compared to
R&M III houses which had readings of 881 /*g/g for window sills and 1,071 jug/g for window
wells, and compared to R&M II houses which had intermediate lead concentrations of 3,165
for window sills and 4,989 /xg/g for window wells.
71
-------
The five groups of houses were most similar to each other in terms of dust loadings. As
with the other measures, however, dust loadings tended to be highest in R&M I houses, lowest
in R&M in houses, and intermediate in R&M II houses. For windows wells, R&M I houses had
a statistically higher geometric mean dust loading (777 mg/ff) than R&M HI houses (205 mg/ft2).
R&M II houses had intermediate dust loadings (519 mg/fl?).
Summary Measures Of Dust Data For A House
Summary measures of dust data for each house were calculated based on the weighted
average of the measurements from surface types common to all campaigns. Lead loadings and dust
loadings were weighted by the surface area sampled. Lead concentrations were weighted by the
sample mass. The "weighted average loading" within each house was calculated as the total mass
of lead collected divided by the total area sampled (or total dust mass, depending on the dust
endpoint). These weighted averages were computed based on samples collected from floors in
rooms with windows, window sills, and window wells. The results were then transformed using
the natural logarithm.
Figure 22 displays the geometric mean of the weighted averages for each dust endpoint at
the 12-month campaign for each group. Based on these summary measures, geometric mean lead
loadings at the 12-month campaign were approximately 21 times higher in R&M I houses than hi
R&M HI houses. This difference is due to the order of magnitude higher lead concentrations and
approximately two-fold higher dust loadings in R&M I houses, relative to R&M in houses. Lead
concentrations and lead loadings in the modern urban and previously abated houses were one to
two orders of magnitude lower than corresponding levels in the intervention groups. The five
study groups were most similar hi terms of weighted average dust loadings.
Paint Chips On Sampled Window Surfaces And Window Surface Conditions
For each sub-area included in a composite dust sample from window sills and window
wells, field staff noted the presence or absence of paint chips and rated the surface condition
(smooth and intact to rough and deteriorated). At 12 months, observations of the presence of paint
chips on window sills and window wells were reduced for all three R&M groups relative to pre-
intervention. The decline was greatest in R&M III houses, intermediate hi R&M H houses, and
lowest hi R&M I houses. Similarly, 12-month observations of surface conditions for window sills
and window wells showed improvement over pre-intervention observations for all R&M groups.
The improvement was greatest hi R&M ffl houses and intermediate in R&M H houses for window
sills, and similar in all three groups for window wells.
72
-------
Figure 22: Overall Weighted Geometric Mean Lead Concentrations, Lead Loading,
And Dust Loading By Study Group At The Twelve Month Campaign
100000
10000
CO
CD
8
CD
1000
100
10
0.1
R&M Level I O R&M Level II • R&M Level Ilia Previously Abated a Modern Urban)
Lead Concentration
(M9/9)
Lead Loading
Dust Loading
(mg/ft2)
Summary measures are based on weighted averages of all dust sample types within houses
73
-------
Lead In Drip-Line Soil
Drip-line soil samples were not collected at the 12-month campaign. Therefore, this report
provides data on soil lead concentrations at the six-month campaign for each study group (Table
17). These data are limited due to the lack of soil for most study houses. Soil lead concentrations
in the modern urban houses at six months (geometric mean 73 /zg/g, range 34 to 229 /ig/g)
remained similar to their corresponding levels at the initial campaign (geometric mean 63 |Kg/g,
range 29 to 154 jtig/g). Across previously abated and R&M houses, individual soil sample values
ranged from 182 to 7,845 /ig/g at six months. A similar range of readings were obtained at the
initial campaign (range of 233 to 15,968 ptg/g).
Lead In Drinking Water
Drinking water samples were not collected at the 12-month campaign. Water lead
concentrations at six months were unchanged from their geometric mean baseline level of ^4 yug/L
(ppb) across all groups. The range of values also remained the same across time (less than the
instrumental limit of detection (
-------
Table 17: Descriptive Statistics For Soil Lead Concentrations By Study Group At The Six-Month Campaign
X ^
-------
Table 19: Correlations Between Dust Lead Concentrations At The 12-Month Campaign
Pearson Correlation Coefficients / Number of Observations
Air Duct
Interior
Entryway
Floors in
Rooms with
Windows
Window Sill
Upholstery
Window
Well
Floors in
Rooms
without
Windows
r
n
r
n
r
n
r
n
r
n
r
n
r
n
Air Duct \
-
-
-
-
-
-
-
Interior
Btotryway
0.41**
57
-
-
-
-
-
-
Floors in
Rooms
with
Wtodws
0.62**
57
0.53**
104
-
-
-
-
-
Window Sill
0.47**
57
0.42**
104
0.43**
104
-
-
-
-
Upholstery
0
0.32*
46
0.23
46
0.28
46
-
-
-
Window
WeB
0.34**
56
0.49**
103
0.44**
103
0.61**
103
0.33*
46
-
-
Floors in
Ko»)t»$
without
Wjtaftamt
0.48**
30
0.52**
55
0.53**
55
0.39**
55
0.42*
25
0.36**
54
-
p-value is < .05 ** p-value is < .01
76
-------
Table 20: Correlations Between Dust Lead Loadings At The 12-Month Campaign
Pearson Correlation Coefficients / Number of Observations
s,.
Air Duct
Interior
Entryway
Floors in
Rooms with
Windows
Window Sill
Upholstery
Window
Well
Floors in
Rooms
without
Windows
r
n
r
n
r
n
r
n
r
n
r
n
r
n
Air Buct
-
-
-
-
-
-
-
Interior
latr.jwj'
0.24
57
-
-
-
-
-
-
Moors in
KOJHttS
with
Windows
0.27*
57
0.37**
104
-
-
-
-
-
Window Silt
0.34**
57
0.37**
104
0.43**
104
-
-
-
-
Upholstery
0
0.10
46
0.23
46
0.03
46
-
-
-
Window
Weft
0.30*
56
0.37**
103
0.33**
103
0.67**
103
0.03
46
-
-
Floors in
Rooms
without
Wfrt<*ow$
-0.01
30
0.36**
55
0.39**
55
0.24
55
0.15
25
0.18
54
-
* p-value is < .05 ** p-value is < .01
77
-------
Table 21: Correlations Between Dust Loadings At The 12-Month Campaign
Pearson Correlation Coefficients / Number of Observations
Cv 4 \
Air Duct
Interior
Entryway
Floors in
Rooms with
Windows
Window Sill
Upholstery
Window
Well
Floors in
Rooms
without
Windows
r
n
r
n
r
n
r
n
r
n
r
n
r
n
Air Duct !
-
-
-
-
-
-
-
Interior
'Btttryway
-0.02
57
-
-
-
-
-
-
Moors an
K09BIS
with
Wfcjjfow?
-0.09
57
0.07
104
-
-
-
-
-
Window Sill
0.03
57
0.50**
104
0.06
104
-
-
-
-
tJphotstery
0
-0.11
46
0.06
46
-0.10
46
-
-
-
Window
Weft
-0.10
56
0.26**
103
0.08
103
0.70**
103
-0.15
46
-
-
Etoorsiii
Kooms
without
Wf»«tow$
0.34
30
-0.01
55
0.28*
55
-0.04
55
-0.14
25
0.05
54
-
* p-value is < .05 ** p-value is < .01
78
-------
Table 22:
Correlations Between Blood Lead and Dust Measures Using The Youngest Child Per Household In Continuing
Houses At The 12-Month Campaign
Pearson Correlation Coefficients / Number of Observations
Dusl Measure
Correlated with
Blood Lead
log of dust
lead concentration
0»g/g)
log of dust
lead loading
(Mg/ft2)
log of
dust loading
(mg/ft2)
SIMM TVPI
Inferior Eatrywsy
r
n
r
n
r
n
0.28**
92
0.14
92
-0.01
92
Floors to
Rooms with
Wiftdws
0.44**
92
0.35**
92
-0.01
92
Floors in
Rooms
Without
Windows
0.32*
50
0.31*
50
0.02
50
Upholstery
0.40*
40
0.42**
40
0.22
40
Window
I Sill
0.18
92
0.16
92
<0.01
92
Window
Well
0.19
91
0.09
91
-0.10
91
AirDtoct
0.38**
52
0.35*
52
0.13
52
* = p-value is <05
** = p-value is
79
-------
Table 23:
Correlations Between Blood Lead and Dust Measures Using All Children Per Household In Continuing
Houses At The 12-Month Campaign
Pearson Correlation Coefficients / Number of Observations
Dust Measure
Correlated, wftfe
Blood I*ad
^
log of dust
lead concentration
(Mg/ft2)
log of dust
lead loading
(Mg/ft2)
log of
dust loading
(rig/ft2)
GEAH/T'ID-T t* •'WBU1
CTfjylyM jr fjf?_i 1 JC Jfjj
Interior Entry way
x
r
n
r
n
r
n
0.29**
125
0.14
125
<-001
125
floors in
Rooms wjtfo
Windows
0.44**
125
0.35**
125
<-001
125
iloors in
Boonts
witliout
-Wiadow$
0.29*
66
0.33**
66
0.05
66
tfpholsfery
:
-
0.28*
59
0.33*
59
0.17
59
Window
S.H
0.20*
125
0.16
125
<-001
125
Window
Weil
0.17
123
0.08
123
-0.10
123
; Air Buct
-
0.47**
66
0.42**
66
0.15
66
* = p-value is <. .05
** = p-value is £.01
80
-------
7.3 Longitudinal Data Analysis
Environmental Dust Model
The environmental dust model (described in section 6.3) was developed for the data for
lead loadings, lead concentrations, and dust loadings. The dependent variable for the
environmental model, called factorl, was obtained from a factor analysis and accounted for most
of the variability of environmental dust lead.
Figures 23(a-c) are plots of the least square mean estimates for each of the three dust
endpoints (lead loadings and concentrations, and dust loadings ) derived from the environmental
model when fit to data from the three R&M groups only. Figures 24(a-c) are plots of the least
square mean estimates derived from the same model fit to data from all five groups. Note that
solid lines are used to connect the points in these plots. This is done for ease of display. These
lines should not be taken to indicate that trends in the intervals between campaigns are known.
Study group, campaign and the interaction of study group and campaign were found to be
statistically significant in all six applications of the environmental model, after controlling for
season. Season was found to have a significant fixed effect only in the models fit to the dust
loading data. The significant interaction term indicates that the relationship between group and
campaign for the three dust endpoints is not the same across study groups. The main findings of
the applications of the environmental model are listed below.
Environmental Dust Model - Comparison Of Groups At Specific Campaigns
• Pre-intervention dust lead loadings were significantly higher in R&M HI houses than in
R&M I and R&M II houses. Statistically significant differences were found between the
three R&M groups at each post-intervention campaign, except for between R&M I houses
and R&M II houses at two months. During follow-up, R&M III houses consistently had
the lowest lead loadings, R&M I the highest lead loadings, and R&M II had intermediate
lead loadings. Modern urban houses had statistically significantly lower lead loadings than
each of the other four study groups at each campaign.
• Pre-intervention dust lead concentrations were not significantly different across the three
R&M groups. Lead concentrations were significantly lower (generally p < .01) hi R&M
in houses than in R&M I and R&M II houses at immediate post-intervention, two months,
six months, and at 12 months. During follow-up, dust lead concentration was lowest in
R&M HI houses, highest in R&M I houses, and intermediate in R&M II houses. Only at
12 months, were lead concentrations in R&M II houses significantly lower than those in
R&M I houses. R&M I-HI houses and previously abated house all had significantly higher
dust lead concentrations during follow-up than modern urban houses. Lead concentrations
in R&M HI houses were not significantly different from those in previously abated houses
at six months and 12 months.
81
-------
• At pre-intervention, dust loadings were significantly higher in R&M III houses than in
R&M I and R&M n houses. Except for at two months post-intervention, dust loadings in
R&M in houses were significantly less than those in R&M I houses. R&M II houses had
intermediate dust loadings at each follow-up campaign; they were statistically significantly
less than those in R&M I houses at post-intervention and at 12 months. Dust loadings in
the modern urban houses were not statistically significantly different from those in the
other four groups at six months and 12 months.
Environmental Dust Model — Changes Over Time Within Groups
• For all three R&M groups, lead loadings during follow-up were statistically significantly
lower than the corresponding pre-intervention lead loadings. Lead loadings at two months,
six months, and 12 months were significantly higher than the corresponding immediately
post-intervention lead loadings, except for R&M I houses at two months. Further, no
statistically significant changes in dust lead loadings were found within any of the R&M
groups between two months and 12 months post-intervention.
• R&M I intervention was not associated with a statistically significant reduction in dust lead
concentration. In R&M II and R&M m houses, lead concentrations were significantly
lower at all post-intervention campaigns compared to baseline, except for R&M II houses
at immediately post-intervention. R&M HI was the only R&M group to have a significant
reduction in lead concentration immediately after the intervention.
• Dust loadings were reduced significantly immediately post-intervention and remained
significantly below pre-intervention levels during the first year in all there groups of R&M
houses, despite significant increases in dust loadings at two months in R&M II and R&M
HI houses.
• Statistically significant changes were not found for dust lead loadings, lead concentrations
and dust loadings in modern urban and previously abated houses during the first year of
follow-up, despite downward trends in lead loadings and dust loadings in both groups.
Blood Lead Comparison Model
The main findings of the comparison model (see section 6.3) for investigating blood lead
changes within and between groups are listed below. The model was fit separately for children
with initial blood lead concentrations < 20 /ig/dL and for those with blood lead concentrations ^20
/ig/dL. Figures 25(a,b) and 26(a,b) are plots of the predicted blood lead concentrations based on
the longitudinal data analysis of children with baseline blood lead concentrations <20 /ig/dL in
the three R&M groups and in all five study groups. Table 24 displays the predicted blood lead
concentrations with 95 percent confidence intervals for children with initial blood lead
concentrations <20 /ig/dL, by study group.
82
-------
Children With Baseline Blood Lead Concentration < 20 jug/dL
The interaction between group and campaign was not statistically significant and the
models were refitted without the interaction term. Age and season, but not gender, were
found to be statistically significant in all applications of the comparison model.
For children with baseline blood lead concentrations <20 ^ig/dL, no statistically
significant differences in blood lead concentration were found between and within R&M
groups during the first year of follow-up, controlling for age, gender and season. Group
and campaign were not found to be significant in this analysis. R&M I children tended to
have lower blood lead concentrations at each campaign, including baseline, compared to
R&M in children. The group variable was statistically significant in the five group model
when controlling for age, gender, and season.
Controlling for age and season, children in modern urban houses had blood lead
concentrations that were statistically lower than those of children in each of the other four
study groups at the initial, six-month, and 12-month campaigns. Children in the modern
urban houses had a small, but statistically nonsignificant, increase in blood lead
concentration over baseline at the six-month campaign.
Children with initial blood lead concentrations < 20 fig/dL in the previously abated control
houses had no statistically significant blood lead changes at the six-month and 12-month
campaigns, controlling for age, gender, and season.
Children With Baseline Blood Lead Concentrations >20 /ig/dL
None of the children in the modern urban group had blood lead concentrations^ 11 jig/dL.
For the 19 children in the other four groups with initial blood lead concentration .2 20
Mg/dL, a statistically significant downward trend hi blood lead concentration was found
during the first year of follow-up, when controlling for age, season, and group. (Only one
child in the R&M I group had an initial blood lead concentration _^20 /xg/dL).
83
-------
Table 24:
Predicted Blood Lead Concentration (PbB, /tg/dL) By Group And By
Campaign In Children With Initial PbB < 20 jtg/dL*
;:lb«Jy<*roB|*
R&MI
R&MII
R&M III
Previously Abated
Modern
Urban
Initial Campaign
Predicted PbB
05ft €9
8.8
(7.6 to 10.3)
10.5
(9.2 to 11. 9)
11.3
(9.9 to 13.0)
11.7
(10.7 to 12.8)
3.3
(2.9 to 3.8)
Two Monfb
•y*x-£ •fiff5^&^
•• •• x-.^\ ^v? •
I'^i^^pJI'bB
8.8
(7.3 to 10.4)
11.2
(9.7 to 12.9)
12.4
(10.4 to 14.8)
not applicable
not applicable
;. -Six Month
" Campaign
Predicted PbB
#S%O)
9.0
(7.6 to 10.7)
11.6
(10.2 to 13.2)
11.9
(10.1 to 14.0)
13.7
(12.2 to 15.5)
3.8
(3.2 to 4.5)
*)Cwelve Month
Predict* PfcB
' $s& CD
7.8
(6.1 to 9.8)
10.3
(9.0 to 11. 9)
10.7
(9.2 to 12.5)
12.2
(10.9 to 13.5)
3.6
(3.1 to 4.1)
* Based on the application of the comparison model for longitudinal data analysis described in
section 6.3)
84
-------
Exposure Model Fitted To Blood Lead Concentration Data
The main findings of the exposure models (see section 6.3) used to investigate the relationship
between blood lead concentration and dust lead (loading and concentration) are below:
• Age, age3, and season (summer vs nonsummer) were significant contributors to the model
for the three R&M groups and for all five groups. Gender and hand-to-mouth activity
(high vs low) were not found to be consistently significant contributors to the model .e
• Controlling for age, campaign, dust factorl, and factor!, the seasonal change in children's
blood lead concentration was estimated to be 4-1.2 jug/dL in summer, relative to the other
seasons.
• Using all five study groups in the model, dust lead loadings and concentrations (factorl
and factorl) were significantly related to children's blood lead concentration after adjusting
for age, season, campaign and the inclusion of random effects for houses and multiple
children in each house. Factorl and factorl were not found to be significant contributors
to the model for the three R&M groups.
• The interactions of factorl and factorl with campaign were not statistically significant for
lead concentration factors and lead loading factors. For this reason, the exposure models
do not include these interaction terms.
Figures 27a and 27b are partial-residual plots of blood lead concentration versus factorl
dust lead loading and factorl dust lead concentration, derived from the exposure model for all five
study groups. These types of plots reflect the relationship between the dependent variable (blood
lead concentration) and a specific independent variable (factorl dust lead) after both variables are
adjusted for all of the other independent variables in the model. The slope of the regression line
in the figure is non-zero and positive, indicating a statistically significant relationship between
blood lead concentration and dust lead loading, and between blood lead concentration and dust lead
concentration. The positive slope indicates that blood lead concentration increases as exposure
increases. Factorl is a composite measure of lead exposure in a house based on a linear
combination of floor, window sill, and window well lead loadings or lead concentration data. Due
to the nature of factorl, it is not possible to interpret the model findings in terms of a unit change
in blood lead concentration predicted for a unit change in factorl. These partial-residual plots also
indicate that the model assumptions, with respect to the normal distribution of residuals, is not
violated.
e One measure of hand-to-mouth activity had borderline statistical significance using data
from all five study groups through the 12-month campaign. Within some groups, one of the various
measures of hand-to-mouth activity reached statistical significance (.05), or borderline significance.
85
-------
Carpet Dust Data
Although this study was not designed to study carpets, longitudinal data analysis was
performed to determine whether dust lead loadings and concentrations and dust loadings varied
by the amount of carpet included in the composite dust samples from floors. Dust loadings and
dust lead loadings tended to increase as the amount of carpet area included in composite samples
increased, when accounting for group, campaign, the interaction of group and campaign, and story
(1st vs 2nd floor). Dust lead concentrations, however, decreased slightly. This pattern of findings
suggests that carpets are dust traps or sinks. The significance of this pattern is not clear; other
analyses indicated that the amount of carpet included in composite samples was not a predictor of
children's blood lead concentrations.
86
-------
Figure 23a: Lead Concentation Least Square Mean Estimates
£
8
CO
I
re
1 2
post
intervention
567
Campaign (months)
-R&MI
-R&MII
-R&M III
10 11 12
Figure 23b: Dust Loading Least Square Estimates
1
post
intervention
45678
Campaign (months)
9 10 11
-R&MI
- R&M II
•R&M III
12
Figure 23c: Lead Loading Least Square Estimates
t
CO
1
£
1
post
intervention
567
Campaign (months)
10 11
12
87
-------
Figure 24a: Lead Concentration Least Square Means
4 6
Campaign (months)
10
12
Figure 24b: Dust Loading Least Square Mean Estimates
8
w
I
u.
468
Campaign (months)
12
Figure 24c: Lead Loading Least Square Mean Estimates
468
Campaign (months)
10
88
-------
Figure 25a: Comparison Model With Initial PbB < 20
•o
1
CO
ja
o.
T!
£
a.
8 -
6 •
4 -
2
Sample Sizes
R&M I 34
R&M II 27
R&M III 29
468
Campaign (months)
10
—•— R&M I
-•—R&M II
| -A- R&M III
12
Figure 25b: Comparison Model With Initial PbB < 20
O)
00
.o
0.
•o
8
6 -
•o
2
Q. 4
2
Sample Sizes
R&M I 34
R&M II 27
R&M III 29
Abated 20
Modem 20
468
Campaign (months)
10
-R&M I
• R&M II
-R&M III
-Abated
- Modem
12
89
-------
Figure 26a: Comparison Model With Initial PbB > 20
468
Campaign (months)
12
Figure 26b: Comparison Model With Initial PbB > 20
468
Campaign (months)
12
90
-------
Figure 27a: All Groups Lead Loading Exposure Model Adjusted Residual
Plot
1
0.8
0.6
0.4
1. 0.2 +
_c
2 o
1
H -02
«• *J-f-
&
-0.4
-0.6
-0.8
-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4
Residuals of Factor!
0.6 0.8
Figure 27b: All Groups Lead Concentration Exposure Model Adjusted
Residual Plot
-0.8 -0.6
-0.4
-0.2 0 0.2
Residuals of Factor!
0.4 0.6
91
-------
8.0 REFERENCES
1. Centers for Disease Control and Prevention. 1997. Update: Blood Lead Levels-United
States, 1991-1994. MMWR 46(7): 141-146.
2. U.S. EPA. 1992. Quality Assurance Project Plan for the Kennedy Krieger Institute Lead
Paint Abatement and Repair and Maintenance Study in Baltimore. November 1992. Office
of Pollution Prevention and Toxics, Design and Development Branch, Washington, DC.
3. U.S. EPA. Lead-Based Paint Abatement and Repair and Maintenance Study in Baltimore:
Pre-Intervention Findings. Office of Prevention, Pesticides and Toxic Substances;
Washington, DC, August 1996. EPA 747-R-95-012.
4. U.S. Department of Housing and Urban Development. Guidelines for the Evaluation and
Control of Lead-Based Paint Hazards in Housing. Washington, DC, June 1995.
5. U.S. Environmental Protection Agency. 1995. "Guidance on the Identification of Lead-
Based Paint Hazards" pursuant to Section 403 of Title IV of the Toxic Substances Control
Act (40 CFR 745). Federal Register, September 11, 1995.
6. Farfel MR, Chisolm JJ and Rohde CA. 1994. The longer-term effectiveness of residential
lead paint abatement. Environmental Research 66:217-221.
7. Farfel MR and Chisolm JJ. 1991. An evaluation of experimental practices for abatement
of residential lead-based paint: report on a pilot project. Environmental Research 55:199-
212.
8. U.S. Environmental Protection Agency. 1995. Review of Studies Assessing Lead
Abatement Effectiveness. Report No. EPA-747-R-95-006.
9. U.S. Agency for Toxic Substances and Disease Registry (ATSDR). 1988. The Nature
and Extent of Lead Poisoning in the United States: A Report to Congress. USDHHS Public
Health Service, Atlanta, Georgia.
10. U.S. Centers for Disease Control. Preventing Lead Poisoning in Children. Statement by
the Centers for Disease Control. October 1991; USDHHS PHS, Atlanta, Georgia.
11. Chisolm JJ, Mellits ED, Quaskey SA. 1986. The relationship between the level of lead
absorption in children and the age, type, and condition of housing. Environmental
Research 38:31-45.
92
-------
12. Clark CS, Bornschein RL, Grote J, et al. 1991. Urban lead exposures of children in
Cincinnati, Ohio. Journal of Chemical Speciation and Bioavailability 3:163-171.
13. Charney E, Kessler B, Farfel M, Jackson D. 1983. A controlled trial of the effect of
dust-control measures on blood lead levels. New England Journal of Medicine 309:1089-
1093.
14. Lanphear BP, Weitzman M, Tanner M, et al. 1994. The Relationship of Lead-
Contaminated House Dust and Blood Lead Levels Among Urban Children. Final Report
to the National Center for Lead Safe Housing.
15. Bornschein RL, Succop PA, Krafft KM, Clark CS, Peace B and Hammond PB. 1986.
Exterior surface dust lead, interior house dust lead and childhood exposure in an urban
environment. In: Trace Substances in Environmental Health XX, ed. D.D. Hemphill,
University of Missouri, Columbia, Missouri, 1986.
16. Charney E. 1982. Lead poisoning in children: the case against household lead dust. In:
Lead Absorption in Children: Management, Clinical and Environmental Aspects. Eds. JJ
Chisolm, Jr. and DM O'Hara. Urban and Schwarzenberg, Baltimore, Munich, pp.79-88.
17. Roels HA, Buchet J-P, Lauwerys RR, et al. 1980. Exposure to lead by the oral and
pulmonary routes of children living in the vicinity of a primary lead smelter.
Environmental Research 22:81-94.
18. Sayre JW, Charney E, Vostal J. and Pless IB. 1974. House and hand dust as a potential
source of childhood lead exposure. American Journal of Disabled Children 727:167-170.
19. U.S. Department of Housing and Urban Development. 1990. Comprehensive and
Workable Plan for the Abatement of Lead-Based Paint in Privately Owned Housing:
Report to Congress. HUD, Washington,DC.
20. Farfel MR, Bannon D, Lees PSJ, Lim BS and Rohde CA. 1994. Comparison of two
cyclone-based collection devices for the evaluation of lead-containing residential dusts.
Applied Occupational and Environmental Hygiene 9:212-217.
21. Farfel MR, Lees PSJ, Rohde CA, Lim BS and Bannon D. 1994. Comparison of wipe and
cyclone methods for the determination of lead in residential dusts. Applied Occupational
and Environmental Hygiene 9:1006-1012.
93
-------
22. Farfel MR, Bannon D, Chisolm JJ Jr., Lees PSJ, Lim BS and Rohde CA. 1994.
Comparison of a wipe and a vacuum collection method for the determination of lead in
residential dusts. Environmental Research 65:291-301.
23. Brody DJ, Pirkle JL, Kramer RA, et al. 1994. Blood lead levels in the U.S. population:
phase 1 of the third national health and nutrition examination survey (NHANES III, 1988
to 1991). Journal of the American Medical Association 272:277-283.
24. Swindell SL, Charney E, Brown MJ, Delaney J. 1994. Home Abatement and Blood Lead
Changes in Children with Class HI Lead Poisoning. Clinical Pediatrics 9/1994: 536-541.
25. Rey-Alvarez S and Menke-Hargrave T. 1987. Deleading dilemma: pitfall in the
management of childhood lead poisoning. Pediatrics 79:214-217.
26. Farfel MR and Chisolm JJ. 1990. Health and environmental outcomes of traditional and
modified practices for abatement of residential lead-based paint. Am. J. Public Health
50:1240-1245.
27. Reagan PL. 1997. Unpublished review.
28. U.S. Environmental Protection Agency. 1996. Seasonal Trends in Blood Lead Levels in
Milwaukee, 1990-1996. Report No. EPA 747-R-95-010.
29. U.S. Environmental Protection Agency. 1995. Seasonal Trends in Blood Lead Levels in
Boston, 1979-1983. Report No. EPA 747-R-94-003.
30. Barry, PSI. 1981. Concentrations of lead in tissues of children. Br. J. Industrial Med.
376:1037-1043.
31. Rabinowitz MB, Kopple JD, Whetherhill GW. 1976. Kinetic analysis of lead mobilization
in healthy humans. Journal of Clinical Investigation. 55:260.
32. Chisolm JJ, Mellits ED, Quaskey SA. 1985. The relationship between the level of lead
absorption in children and the age, type, and condition of housing. Environ. Research
38:31-45.
33. Annotated Code of Maryland. 1988. Procedures for abating lead containing substances
from buildings. COMAR26.02.07, Title 26, Maryland Department of the Environment
Regulations, Effective Date: August 8, 1988.
94
-------
34. Research Triangle Institute and Engineering Plus. "Development of a High Volume Small
Surface Sampler for Pesticides and Toxics in House Dust - Final Report" Submitted to
U.S. EPA Exposure Assessment Research Division. Research Triangle Park, NC (June 29,
1990).
35. Bannon DI, Murashchik C, Zapf CR, Farfel MR and Chisolm, JJ Jr. 1994. A graphite
furnace AAS method of blood lead measurement using matrix matched standards. Clinical
Chemistry 40:1730-1734.
36. SAS Institute Inc. 1990. SA$ Language: Reference, Version 6, First Edition. Gary, NC.
37. Hornung, RW and Reed, LD 1990. Estimation of average concentration in the presence
of nondetectable values. Applied Occupational and Environmental Hygiene 5:46-51.
38. Shapiro, SS and Wilk, MB. 1965. An analysis of variance test for normality (complete
samples), Biometrics 52:591-611.
39. Tukey, JW. 1977. Exploratory Data Analysis. Addison-Wesley, Reading, Massachusetts.
40. Laird N, Ware J. 1982. Random-Effects Model for Longitudinal Data. Biometrics 38:963-
974.
41. Zeger SL, Liang K-Y. 1986. Longitudinal Data Analysis for Discrete and Continuous
Outcomes. Biometrics 42:121-130.
42. Zeger SL, Liang K-Y, Albert PS. 1988. Models for Longitudinal Data: A Generalized
Estimating Equation Approach. Biometrics 44:1049-1060.
43. Waternaux C, Laird N, Ware J. 1989. Methods for the Analysis of Longitudinal Data:
Blood Lead Concentrations and Cognitive Development. Journal American Statistical
Association 54:33-41.
44. Liang K-Y, Zeger SL. 1986. Longitudinal Data Analysis Using Generalized Linear
Models. Biometrics 75:13-22.
45. Moulton LH, Zeger SL. 1989. Analyzing Repeated Measures on Generalized Linear
Models via the Bootstrap. Biometrics 45: 381-394.
95
-------
46. Roy all RM. 1986. Model Robust Inference using Maximum Likelihood Estimators.
International Statistical Review 54:221-226.
47. Farfel MR and Rohde CA. Determination of environmental lead, using compositing of
house dust samples. In: Lead Poisoning: Exposure, Abatement, Regulation. (Eds. JJ Breen
and CR Stroup) Lewis Publishers, Boca Raton 1995. pp. 231-235.
48. U.S. Consumer Product Safety Commission. 1977. Lead-containing paint and certain
consumer products bearing lead containing paint (16 CFR 1303). Federal Register
42:44192-44202.
96
-------
APPENDIX A
Descriptive Statistics on Dust Data
-------
Table A-l: Descriptive Statistics For Dust Lead Concentrations By Surface Type And Study Group At The 12-Month
Campaign
Swr&ceType
Air Duct
Interior
Entryway
Floors in
Rooms with
Windows
Floors in
Rooms
without
Windows
Window Sill
Upholstery
Window Well
ftM&erauv
R&M-I
R&M-H
R&M-m
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-III
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-m
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-III
Previously Abated
Modern Urban
R&M-I
R&M-H
R&M-III
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-m
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-III
Previously Abated
Modern Urban
n
10
15
16
5
11
25
23
27
14
15
53
46
54
28
31
17
15
13
6
4
49
46
54
28
30
15
7
11
9
4
45
46
54
28
30
Minimum
-------
Table A-2: Descriptive Statistics For Dust Lead Loadings By Surface Type And Study Group At The 12-Month Campaign
Sarface*Fype
Air Duct
Interior
Entryway
Floor in
Rooms with
Windows
Floors in
Rooms
without
Windows
Window Sill
Upholstery
Window Well
Study Group
R&M-I
R&M-II
R&M-III
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-IH
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-III
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-ffl
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-DI
Previously Abated
Modern Urban
R&M-I
R&M-H
R&M-IH
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-ffl
Previously Abated
Modern Urban
ft
10
15
16
5
11
25
23
27
14
15
53
46
54
28
31
17
15
13
6
4
49
46
54
28
30
15
7
11
9
4
45
46
54
28
30
Minimum
(*$$*)
245
31
144
326
112
21
7
1
8
1
6
4
2
<1
<1
8
1
7
5
<1
5
1
1
4
2
1
33
7
6
1
548
6
2
52
9
Maximum
togtoft
3,755,278
596,898
874,350
180,703
14,428
45,201
9,574
1,452
15,204
391
25,581
4,416
3,441
2,424
107
513
7,580
444
856
3
7,523
10,053
683
24,481
40
158
824
744
82
24
367,432
163,334
29,430
22,872
2,410
Geometric
Mean*
(PS/ft*)
13,239
15,237
12,040
10,020
856
365
215
94
119
30
94
76
50
77
8
56
53
44
108
1
470
237
29
75
9
36
104
64
20
7
16,698
2,587
220
1,164
208
S.D. on
tog scale
3.054
2.563
2.032
2.394
1.760
1.927
2.001
1.792
2.260
1.383
1.443
1.454
1.613
2.006
1.138
1.377
2.208
1.345
1.724
0.901
1.895
2.224
1.621
1.756
0.815
1.302
1.175
1.591
0.982
1.318
1.450
2.271
1.580
1.904
1.447
Lower 9S%d
forGM
(Ml/ft*)
1,489
3,686
4,077
513
262
165
91
46
32
14
61
43
29
32
5
28
16
19
18
<1
254
102
17
34
6
17
35
22
10
1
10,146
1,084
137
492
104
Upper 95% Cl
forGM
teift**}
117,685
62,985
35,554
195,779
2,794
808
511
191
440
64
144
134
87
187
13
114
179
99
662
4
871
550
50
168
13
74
308
187
43
61
27,479
6,173
353
2,751
416
GM values and confidence intervals for floors (rooms with windows), window sills, and window wells were obtained from SAS* PROC MIXED
-------
Table A-3: Descriptive Statistics For Dust Loadings By Surface Type And Study Group At The 12-Month Campaign
Surface Type
Air Duct
Interior
Entryway
Floors in
Rooms with
Windows
Floors in
Rooms
without
Windows
Window Sill
Upholstery
Window Well
Study Group
R&M-I
R&M-n
R&M-HI
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-III
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-III
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-III
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-III
Previously Abated
Modern Urban
R&M-I
R&M-II
R&M-III
Previously Abated
Modern Urban
R&M-I
R&M-U
R&M-III
Previously Abated
Modern Urban
rt
10
15
16
5
11
25
23
27
14
15
53
46
54
28
31
17
15
13
6
4
49
46
54
28
30
15
7
11
9
4
45
46
54
28
30
Minimum
{BJI/&2)
354
283
94
1,951
1,222
10
19
13
2
11
14
15
17
18
6
17
3
13
49
21
5
5
< 1
8
8
11
70
17
12
14
68
12
9
33
28
Mix&HtKa
feflft
155,499
103,878
167,309
34,803
176,990
8,077
2,330
1,941
2,669
760
8,924
981
617
749
675
459
509
562
156
292
1,901
1,742
264
576
354
407
1,854
2,704
214
87
10,988
10,548
1,592
5,496
3,307
Geometric
ftfeae8
11,492
18,999
12,566
8,807
8,474
244
197
136
108
250
115
113
82
135
100
87
58
84
112
62
67
75
33
66
32
72
218
167
52
36
111
519
205
384
476
S.D. on
Jog scale
2.138
1.834
1.904
1.090
1.560
1.594
1.348
1.297
1.835
1.026
1.112
1.069
0.937
1.141
1.072
1.164
1.260
1.077
0.451
1.110
1.222
1.243
1.221
1.160
0.953
1.115
1.029
1.407
0.937
0.791
0.975
1.432
1.252
1.053
1.223
Lower $$% &
for GM
2,489
6,882
4,556
2,274
2,972
126
110
81
37
141
80
76
63
82
62
48
29
44
70
11
45
50
22
41
21
39
84
65
25
10
566
306
135
250
255
Upper SS% €1
forGM
53,051
52,454
34,657
34,107
24,161
471
352
227
311
441
166
167
108
221
162
159
117
160
180
361
101
111
49
107
50
133
565
429
107
128
1,067
879
312
590
888
GM values and confidence intervals for floors (rooms with windows), window sills, and window wells were obtained from SAS* PROC MIXED
-------
APPENDIX B:
Descriptive Statistics for Baseline Blood Lead Concentrations by Group
-------
Table B-l: Descriptive Statistics For Blood Lead Concentrations By Group At Initial Campaign
! StK<|£<&aii£ i
R&M Level I
R&M Level II
R&M Level HI
Previously Abated
Modern Urban
MfoiKKKfe ;
' .''ti&t&tt I
2.0
3.5
2.0
3.5
2.0
Maximum j
<*W'
22.0
36.0
42.0
28.0
10.0
€eom«trie
Mean
;*NW
9.9
13.8
14.2
12.8
4.8
S J>. on " ;
Iflgsesfo
0.539
0.531
0.542
0.495
0.457
''Lawx$S%Cl
forGM
C^g/ril)
7.9
11.2
11.3
10.2
3.8
OpperSSI&CI
8»<3H
<^4t) %
12.3
16.9
17.9
16.1
6.1
GM values and confidence intervals were obtained from SAS* PROC MIXED
-------
Box Plots Of Blood Lead Concentrations
By Study Group At The Initial Campaign
CD
D)
O
O
•
O
*
*
MU
PA
RM-I
RM-II
RM-III
Study Group
-------
50272-101
REPORT DOCUMENT
PAGE
1. Report No.
EPA747-R-97-001
3. Recipient's Accession No.
4. Title and Subtitle
Lead-Based Paint Abatement and Repair and Maintenance Study in Baltimore:
Findings Based on the First Year of Follow-up
5. Report Date
August 1997
6.
7. Authors)
Farfel, M.R.; Rohde, C.; Lees, P.S.J.; Rooney, B.; Bannon, D.I.; Derbyshire, W.
8. Performing Organization Rept No.
9. Performing Organization Name and Address
Kennedy Krieger Research Institute (KKRI)
707 N. Broadway
Baltimore, MD 21205
10. Project/Task/Work Unit No.
11. Contract © or Grant (G) No.
Contract #68-D4-0001
12. Sponsoring Organization Name and Address
U.S. Environmental Protection Agency
Office of Pollution, Pesticide and Toxic Substances
Washington, D.C. 20460
13. Type of Report & Period Covered
Final Report; 1995-96
14.
15. Supplementary Notes
The following people were major contributors to the study: Dr. Julian Chisolm, Pat Tracy, field staff and
laboratory staff of KKRI. Special acknowledgment is given to Battelle Memorial Institute, Midwest Research
Institute for technical and administrative support during the planning and pilot phases of the study, Maryland
Department of Environment, Baltimore City Department of Health, Housing and Community Development, City
Homes Inc., Maryland Department of Housing and Community Development for financial support on the R&M
interventions and the many individuals within these organizations for their valuable contributions.
16. Abstract (Limit: 200 words)
This report presents the first year of follow-up of the Lead-Based Abatement and Repair and Maintenance
(R&M) Study in Baltimore. The R&M study is designed to characterize and compare the short (2-6 months)
and long-term (12-24 months) effectiveness of three levels of interim control interventions in low-income
housing where children are at high risk of exposure to lead in dust and paint. The study has two control
groups, i.e., urban houses built after 1979 when paint was presumably free of lead, and previously abated
houses that received comprehensive abatement in the past. The study population consists of non-Hispanic
black households with at least one participating child. At the onset, the mean ages of study children ranged
form 25 months to 33 months across groups. The main findings based on dust lead and blood lead data from
five study groups collected during the pre- and post-intervention campaigns, as well as during the two, six and
12 months post-intervention data collection campaigns are presented.
17. Document Analysis
a. Descriptors: Kennedy Craggier Research Institutes, lead, lead dust testing, blood lead testing, lead-based paint,
lead-dust loading, lead-dust concentration, dust loading, low cost repair and maintenance interventions,
children's blood lead, lead exposure reduction in children, lead hazard reduction, blood lead-dust lead
correlation.
b. identifiers/Open-Ended Terms: Lead poisoning, lead abatement, interim control, inductively coupled plasma emission
spectroscopy (ICP- AES), flame atomic absorption spectroscopy (FAAS), graphite furnace atomic absorption
spectroscopy (GFAAS), cyclone-based dust collector, Baltimore Repair and Maintenance dust collector (BRM).
c. Caseate Field/Group:
18. Availability Statement
19. Security Class (This Report)
Unclassified
20. Security Class (This Page)
Unclassified
21. No. of Pages
102
»U.S. GOVERNMENT PRINTING OFFICE: 1997-521-453/90275
------- |