(Do not Cite or Quote) EPA 600/AP-93/001 a
July 1993
Urban Soil Lead Abatement
Demonstration Project
Volume I:
Integrated Report
NOTICE
This document is a preliminary draft. It has not been formally
released by EPA and should not at this stage be construed to
represent Agency policy. It is being circulated for comment
on its technical accuracy and policy implications.
Environmental Criteria and Assessment Office
Office of Health and Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Printed an Recycled Paper
-------
DISCLAIMER
This document is an external draft for review purposes only and does not constitute
U.S. Environmental Protection Agency policy. Mention of trade names or commercial
products does not constitute endorsement or recommendation for use.
July 15, 1993 ii DRAFT-DO NOT QUOTE OR CITE
-------
TABLE OF CONTENTS
Page
LIST OF TABLES vii
LIST OF FIGURES ix
LIST OF PARTICIPANTS xiii
LIST OF REVIEWERS xvii
LIST OF ABBREVIATIONS AND ACRONYMS xix
1. EXECUTIVE SUMMARY 1-1
1.1 SCOPE OF REPORT 1-1
1.2 BACKGROUND AND OVERVIEW 1-2
1.2.1 Stody Designs . 1-3
1.3 BRIEF DESCRIPTION OF HOW INTERVENTION
WAS PERFORMED 1-9
1.4 BRIEF SUMMARY OF INDIVIDUAL STUDY REPORTS .... 1-11
1.4.1 Summary of the Boston Study 1-11
1.4.2 Summary of the Baltimore Study 1-12
1.4.3 Summary of the Cincinnati Study 1-14
1.4.4 Loose Paint Stabilization Approaches 1-14
1.5 SUMMARY OF RESULTS AND STATISTICAL
INFERENCES 1-17
1.5.1 Quality of the Data 1-17
1.5.2 Effectiveness and Persistency of Intervention 1-18
1.5.3 Summary of Statistical Inferences 1-19
1.6 SUMMARY STATEMENT OF PROJECT CONCLUSIONS
AND THEIR IMPLICATIONS 1-20
1.6.1 Project Conclusions 1-20
1.6.2 Implications . 1-21
2. BACKGROUND AND OVERVIEW OF PROJECT 2-1
2.1 PURPOSE OF THIS DOCUMENT 2-1
2.2 PROJECT BACKGROUND 2-3
2.2.1 Historical Perspective 2-3
2.2.2 Site Selection 2-5
2.3 PROJECT OVERVIEW . 2-8
2.3.1 Project Terminology 2-9
2.3.2 Study Groups 2-11
2.3.3 Intervention Strategies 2-11
2.3.4 Measurements of Exposure 2-17
2.3.4.1 Blood Lead 2-18
2.3.4.2 Hand Lead 2-18
2.3.4.3 House Dust 2-18
July 15, 1993 iii DRAFT-DO NOT QUOTE OR CITE
-------
TABLE OF CONTENTS (cont'd)
Page
2.3.5 Treatment Approaches 2-19
2.3.5.1 Soil Abatement Approaches , 2-19
2,3.5.2 Exterior Dust Abatement Approach . 2-21
2,3,5.3 Interior Dust Abatement Approaches , 2-21
2.3.5.4 Loose Paint Stabilization Approaches 2-22
2,3.6 Project Activity Schedule 2-23
2.3.7 Quality Assurance/Quality Control Plan 2-23
2.3.7.1 Quality Assurance/Quality Control
for Soils and Dusts , 2-24
2.3,7.2 Quality Control and Quality Assurance
for Hand Dust 2-27
2,3.7,3 Quality Control and Quality Assurance
for Blood Lead 2-27
2.3.8 External Factors That Could Influence the
Outcome of the Project 2-28
2.3.8.1 Cycles and Trends in Environmental
Lead Concentrations 2-28
2.3.8.2 Unexplained and Unexpected Sources
of Lead 2-31
2.3.8.3 Movement of Lead in Soil and Dust , . 2-32
2.4 AHD EVALUATION OF INDIVIDUAL
STUDY REPORTS 2-33
2.4.1 Summary of the Boston Study . 2-33
2.4.2 Summary of the Baltimore Study ............... 2-34
2.4.3 Summary of the Cincinnati Study 2-36
2.5 CONCLUSION 2-38
2.5,1 Summary of Project Description 2-38
2.5.2 Conclusions ............._..,......,... 2-39
3. PROJECT RESULTS , 3-1
3.1 DATA QUALITY 3-1
3.1.1 Round Robin I: Common Standards and
Analytical Methods 3-1
3.1.2 Double Blind Audit Program for Soil
and Dust 3-6
3.1.3 Round Robin El: Biweight Distribution
and Final Laboratory Calibration 3-11
3.1.4 Disposition of Audit Data ...,..,,,.....,.... 3-14
3.1.5 Database Quality 3-14
July 15, 1993 iv DRAFT-DO NOT QUOTE OR CITE
-------
TABLE OF CONTENTS (cont'd)
Page
3,2 OVERVIEW OF PROJECT DATA 3-16
3.2.1 Description of the Data , 3-16
3.2.1,1 Types of Data 3-16
3.2.1,2 Data Collection Patterns 3-16
3.2.1.3 Data Linkages 3-16
3.2.1,4 Data Transformations 3-17
3.2.L5 Adjustments and Corrections to the Data .... 3-20
3.3 PRESENTATION OF THE DATA ................... 3-21
3,3.1 Effectiveness and Persistency of Soil Abatement 3-23
3.3.2 Effectiveness and Persistency of Exterior
Dust Abatement ..,..,.....,... 3-31
3.3.3 Effectiveness and Persistency of Interior
Dust Abatement 3-34
3.3.4 Hand Dust Results 3-34
3.3.5 Blood Lead Results ....................... 3-46
3.3.5,1 Blood Lead Correction Factors 3-48
3.3.5.2 Reanalysis of Boston Study Blood
Lead Data , 3-49
3.3.5.3 Reanalysis of Cincinnati Study
Blood Lead Data 3-49
3.4 SUMMARY OF RESULTS 3-52
4. STATISTICAL INTERPRETATION OF "HIE RESULTS ......... 4-1
4.1 PREPARATION OF PROIFCT DATA SF^S FOR
STATISTICAL ANALYSIS 4-!
4.1.1 Description of Data Sets 4-2
4.1.2 Validation of the Data Sets .................. 4-3
4.1.3 Modifications in the Project Data Sets 4-3
4,1.3.3 Restmcted Data Sets 4-3
4.1.3,2 Missing Information Procedures . . . . d-4
4.1.4 Statistical Methods . 4-5
4.1,4.1 Repealed Measures Analysis ............ 4-5
4.1.4.2 Autoregressive Regression Models 4-7
4.3.4.3 Structural Equations Modeling ........... &-9
4.1.5 Limitations of the Statistical Methods . 4-42
4.2 IMPACT OF INTERVHK7ION 4-14
4.2.1 Impact of Soil Abatement on Exterior and
Interior DUST 4-14
4.2.2 Impact of Soil and Dust Abatement on Hand
Lead Loading 4-15
42.3 Impact of Soil and Dust Abatement on Blood
Lead Concentrations 4-16
4.3 RESULTS OF STATISTICAL ANALYSES ............. 4-17
4.3.1 Baltimore Study 4-17
[5.1993 v PRA^T-DQ NOT QUOTE OR f"T
-------
TABLE OF CONTENTS (cont'd)
Page
4.3.2 Boston Study 4-24
4,3.3 Cincinnati Study . 4-29
4.4 DISCUSSION AND CONCLUSIONS . 4-36
4.4.1 Comparison Across the T^ree Studies 4-36
4.5 SUMMARY OF STATISTICAL INFERENCES 440
5. CONCLUSIONS 5-!
5.1 SUMMARY OF PROJECT 5-]
5.2 SUMMARY OF RESULTS 5-2
5.3 SUMMARY OF STATISTICAL INFF.RHNCES 5-3
5.3.1 Baltimore Study 5-3
5.3,2 Boston Study 5-3
5.3.3 Cincinnati Study 5-4
5.4 IN" -.GRAMF} PROJECT CONCLUSIONS ............. 5-4
5.4.) Findings 5-5
5-4.2 Implications 5-6
5.4 3 Recommendations
6. REFERENCES
Julv 15. 1993 vi DRAFT-DO NOT QUOTE OR CITE
-------
LIST OF TABLES
Number Page
1-1 Description of Study Groups and Types of Intervention 1-5
1-2 Number of Project Participants by Round 1-6
1-3 Soil Abatement Statistics for the Three Studies 1-9
2-1 Treatment Group Nomenclature with Cross-Reference to
Individual Reports 2-10
2-2 Number of Project Participants by Study Group and Round ..... 2-12
2-3 Soil Abatement Statistics for the Three Studies . 2-20
2-4 Wet Chemistry and Instrumental Methods Used for the First
TtitercaJjbration Study 2-25
3-1 Wet Chemistry and Instrumental Methods Used for the First
Intercalibration Study .....,......,., 3-2
3-2 Analytical Results of the First Intercalibration Study:
Lead Concentrations in the Total and Fine Fractions of
1.0 Soils from Each Study .,......,...,,..,.....,.. 3-3
3-3 Soil and Dust Audit Program Results 3-9
3-4 Preliminary and Final Biweight Distributions for Soil and
Dust Audit Program 3-10
3-5 Results of the Final Intercalibration Study 3-12
3-6 Consensus Values and Correction Factors from the
Final Intercalibration Program 3-13
3-7 Summary of Boston Study Data ....,,.,,,,.,.,.,..... 3-22
3-8 Summary of Baltimore Study Data 3-23
3-9 Summary of Cincinnati Study Data .................... 3-24
4-1 Repeated Measures Analysis of Variance for Log (Blood Lead)
for Baltimore Study. Analysis of Variance Tables 4-18
4-2 Repeated Measures Analysis of Covariance for Log
(Blood Lead) for Baltimore Study. Ancova Tables . . , , 4-18
July 15, 1993 vii DRAFT-DO NOT QUO Tit OR CITE
-------
LIST OF TABLES (cont'd)
Number Page
4-3 Autoregressive Regression Model for Blood Lead on Hand Lead,
Fitted in Log Form, for Baltimore Study 4-19
4-4 Autoregressive Regression Model for Blood Lead on
Environmental Lead, Fitted in Log Form, for Baltimore Study,
Preabatement 4-20
4-5 Autoregressive Regression Model for Blood Lead on
Environmental Lead, Fitted in Log Form, for Baltimore Study,
Postabatement 4-21
4-6 Regression Coefficients for Baltimore Structural Equations
Model, Using the GLS Method 4-23
4-7 Repeated Measures Analysis of Variance for Log (Blood Lead)
for Boston Study, Analysis of Variance Tables 4-24
4-8 Repeated Measures Analysis of Covariance for Log
(Blood Lead) for Boston Study, Analysis of Covariance Tables . . . 4-25
4-9 Autoregressive Regression Model for Blood Lead on Hand Lead,
Fitted in Log Form, for Boston Study . 4-26
4-10 Autoregressive Regression Model for Blood Lead on
Environmental Lead, Fitted in Log Form, for Boston Study 4-26
4-11 Regression Coefficients for Boston Structural Equations Model,
Using the AGLS Method 4-28
4-12 Repeated Measures Analysis of Variance for Log (Blood Lead)
for Cincinnati Study, Analysis of Variance Tables 4-30
4-13 Repeated Measures Analysis of Covariance for Log (Blood Lead)
for Cincinnati Study, Analysis of Covariance Tables 4-30
4-14 Autoregressive Regression Model by Neighborhood for
Blood Lead on Environmental Lead, Fitted in Log Form,
for Cincinnati Study 4-32
4-15 Autoregressive Regression Model for Blood Lead on
Environmental Lead, Fitted in Log Form, for Cincinnati Study . . . 4-34
4-16 Regression Coefficients for Boston Structural Equation Model,
Using the AGLS Method 4-39
July 15, 1993 viii DRAFT-DO NOT QUOTE OR CITE
-------
LIST OF FIGURES
Number Page
1-1 Generalized concept of the sources and pathways for lead
exposure in humans 1-7
1-2 Typical pathways of childhood exposure to lead in dust 1-7
2-1 Generalized concept of the sources and pathways of lead
exposure in humans 2-13
2-2 Typical pathways of childhood exposure to lead in dust 2-14
2-3 Pathway intervention scheme for dust exposure (Boston Soil
Abatement Study) 2-15
2-4 Pathway intervention scheme for dust exposure (Baltimore
Soil Abatement Study) 2-16
2-5 Pathway intervention scheme for dust exposure (Cincinnati
Soil Abatement Study) 2-17
2-6 Project activity schedule showing the tunes of sampling
and interviewing and soil abatement 2-24
2-7 Literature values for seasonal patterns for childhood
blood lead (age 25 to 36 months) , 2-29
2-8 Literature values for seasonal patterns for blood lead
in children and adults (NHANES n, age 6 months to
74 years) 2-29
2-9 Expected changes in blood lead during early childhood . . 2-30
3-1 Comparison of uncorrected data for two wet chemistry methods of
soil analysis showing the comparability of hot and cold nitric
acid for the Cincinnati laboratory . 3-4
3-2 Comparison of uncorrected data for atomic absorption spectroscopic
analysis by two laboratories (Baltimore and Cincinnati) using the
hot nitric acid method of soil analysis .- 3-5
3-3 Interlaboratory comparison of uncorrected data for the X-ray
fluorescence method of soil analysis showing the comparability
of the Boston and Georgia Institute of Technology laboratories . . . 3-5
July 15, 1993 ix DRAFT-DO NOT QUOTE OR CITE
-------
LIST OF FIGURES (cont'd)
Number Page
3-4 Interlaboratory comparison of uncorrected data for soil analysis
showing the comparability of inductively coupled plasma emission
spectroscopy and atomic absorption spectroscopy for the Baltimore
and Cincinnati laboratories 3-7
3-5 Comparison of uncorrected data for soil analysis showing the
comparability of inductively coupled plasma emission spectroscopy
and atomic absorption spectroscopy within the Baltimore
laboratory , 3-7
3-6 Interlaboratory comparison of uncorrected data for soil analysis
showing the comparability of X-ray fluorescence and atomic
absorption spectroscopy for the Cincinnati and Boston
laboratories 3-8
3-7 Interlaboratory comparison of uncorrected data for soil analysis
showing the comparability of X-ray fluorescence and atomic
absorption spectroscopy for the Baltimore and Boston
laboratories 3-8
3-8 Departures from consensus dust values for each of the
three studies 3-15
3-9 Departures from consensus soil values for each of the
three studies 3-15
3-10 Hypothetical representation of intervention impact on soil and
dust concentrations 3-27
3-11 The arithmetic means of Boston soil lead concentrations by
study group show the effectiveness and persistency of soil
abatement 3-27
3-12 Cincinnati soil lead concentrations 3-28
3-13 Reconstruction of the expected effectiveness of soil abatement
in the Baltimore study 3-28
3-14 Boston soil lead concentrations above 2,500 micrograms
per gram 3-30
3-15 Cincinnati, exterior dust load measurements 3-32
July 15, 1993 x DRAFT-DO NOT QUOTE OR CITE
-------
LIST OF FIGURES (cont'd)
Number
3-16 Cincinnati exterior dust lead load measurements , 3-32
3-17 Cincinnati exterior dust lead concentrations 3-33
3-18 Boston floor dust lead concentration 3-35
3-19 Boston floor dust load . 3-35
3-20 Boston floor dust lead load 3-36
3-21 Boston window dust lead concentrations 3-36
3-22 Boston window dust load 3-37
3-23 Boston window dust lead load 3-37
3-24 Cincinnati floor dust lead concentrations 3-38
3-25 Cincinnati floor dust load . 3-38
3-26 Cincinnati floor dust lead load 3-39
3-27 Cincinnati window dust lead concentration 3-39
3-28 Cincinnati window dust load 3-40
3-29 Cincinnati window dust lead load 3-40
3-30 Cincinnati mat dust lead concentration 3-41
3-31 Cincinnati mat dust load 3-41
3-32 Cincinnati mat dust lead load ' 3-42
3-33 Cincinnati entry dust lead concentration 3-42
3-34 Cincinnati entry dust load . , . 3-43
3-35 Cincinnati entry dust lead load 3-43
3-36 Boston hand lead load 3-44
3-37 Baltimore hand lead load 3-45
My 15, 1993 xi DRAFT-DO NOT QUOTE OR CITE
-------
LIST OF FIGURES (cont'd)
Number Page
3-38 Cincinnati hand lead load 3-45
3-39 Baltimore uncorrected blood lead concentrations 3-46
3-40 Baltimore blood lead concentrations corrected for
seasonal cycle and long-term time trends 3-47
3-41 Seasonally adjusted correction factor for blood
lead concentrations 3-47
3-42 Boston unconnected blood lead concentrations 3-50
3-43 Boston blood lead concentrations corrected for seasonal
cycles and long-term time trends and normalized to BOS P 3-50
3-44 Cincinnati uncorrected blood lead concentrations 3-51
3-45 Cincinnati blood lead concentrations corrected for seasonal
cycles and long-term time trends and normalized to CIN NT to
show possible impact of soil and dust abatement 3-52
4-1 Structural equation diagram for the Baltimore study 4-22
4-2 Structural equation diagram for the Boston study 4-27
4-3 Structural equation diagram for the Cincinnati study 4-33
July 15, 1993 xii DRAFT-DO NOT QUOTE OR CITE
-------
LIST OF PARTICIPANTS
Urban Soil Lead Abatement Demonstration Projects
Dr. Ann Aschengrau
Boston University School of Public Health
80 East Concord Street, T-355
Boston, MA 02118
Dr. David Bellinger
Children's Hospital
Gardner House Room 455
300 Longwood Avenue
Boston, MA 02115
Dr. Robert Bornschein
University of Cincinnati
Department of Environmental Health
3223 Eden Avenue #56
Cincinnati, OH 45267-0056
Ms. Dawn Boyer
Inorganic Chemistry Department
Lockheed ESC
1050 East Flamingo
Las Vegas, NV 89119
Ms. Merrill Brophy
MDE/Lead & Soil Project
Maryland Department of the Environment
2500 Broening Highway
Baltimore, MD 21224
Dr. Richard Brunker
U.S. EPA - Region m
Site Support Section MD-3HW26
841 Chestnut Street
Philadelphia, PA 19107
Mr. Barry Chambers
Maryland Department of the Environment
2500 Broening Highway
Baltimore, MD 21224
Dr. Rufus Chancy
U.S. Department of Agriculture
ARC, Building 318 BARC-East
BeltsviUe, MD 20705
Dr. Julian Chisolm
Kennedy Institute
707 N, Broadway
Baltimore, MD 21205
Dr. Scott Clark
University of Cincinnati
Department of Environmental Health
Mail Stop 56
3223 Eden Avenue
Cincinnati, OH 45267-0056
Ms. Linda Conway-Mundew
University of Cincinnati
Department of Environmental Health
1142 Main Street
Cincinnati, OH 45210
Dr. Robert Elias
U.S. EPA
Environmental Criteria Assessment Office
Mail Drop 52
Research Triangle Park, NC 27711
Dr. Eatherine Farrell
Arnie Arundell County Health Department
3 Harry S. Truman Parkway
Annapolis, MD 21401
Ms. Beverly Fletcher
U.S. EPA - Region I
Environmental Services Division
60 Westview Street
Lexington, MA 02173
My 15, 1993
xui
DRAFT-DO NOT QUOTE OR CITE
-------
LIST OF PARTICIPANTS (cont'd)
Urban Soil Lead Abatement Demonstration Projects
Mrs. Barbara Gordon
Cincinnati Health Department
3101 Burnet Avenue, Room 309
Cincinnati, OH 45229
Ms. Jo Ann Grote
University of Cincinnati
Department of Environmental Health
1142 Main Street
Cincinnati, OH 45210
Mr. Bill Hanson
Cincinnati Health Department
UC Soil Project
1142 Main Street
Cincinnati, OH 45210
Mr. Reginald Harris
U.S. EPA - Region HI
Site Support Section MD-3HW15
841 Chestnut Street
Philadelphia, PA 19107
Mr. Ronald Jones
Cleveland Department of Public Health
1925 East St. Claire Ave.
Cleveland, OH 44114
Dr. Boon Lim
Environmental health Program
Maryland Department of the Environment
2500 Broening Highway
Baltimore, MD 21224
Dr. Allan Marcus
Battelle Applied Statistics
100 Park Drive
P.O. Box 13758
Research Triangle Park, NC 27709
Dr. Tom Matte
Public Health Service - Region n
26 Federal Plaza, Room 3337
New York, NY 10278
Ms. Lisa Matthews
U.S. EPA - OS 230
401 M Street, SW
Washington, DC 20460
Mr. J. Todd Scott
The Cadmus Group
3580 Cinderbed Road
Suite 2400
Newington, VA 22122
Mr. Dave Mclntyre
U.S. EPA - Region I
Environmental Services Division
60 Westview Street
Lexington, MA 02173
Mr. William Menrath
University of Cincinnati
Department of Environmental Health
1142 Main Street
Cincinnati, OH 45210
Dr. Winkey Pan
University of Cincinnati
Department of Environmental Health
3223 Eden Avenue
Cincinnati, OH 45267-0056
Dr. Dan Paschal
Centers for Disease Control
1600 Clifton Road, NE
Mail Stop F-18
Atlanta, GA 30333
July 15, 1993
xiv
DRAFT-DO NOT QUOTE OR CITE
-------
LIST OF PARTICIPANTS (cont'd)
Urban Soil Lead Abatement Demonstration Projects
Ms. Sandy Roda
University of Cincinnati
Department of Environmental Health
3223 Eden Avenue
Cincinnati, OH 45267-0056
Dr. Charles Rohde
Biostatistics
Johns Hopkins University
615 N. Wolfe Street
Baltimore, MD 21205
Ms. Penny Schmitgen
University of Cincinnati
Department of Environmental Health
3223 Eden Avenue
Cincinnati, OH 45210
Dr. James Simpson
Centers for Disease Control
CEHIC/EHHE, Mail Stop F28
1600 Clifton Road
Atlanta, GA 30333
Dr. Tom Spittler
U.S. EPA - Region I
Environmental Services Division
60 Westview Street
Lexington, MA 02173
Mr. Warren Strauss
MDE/Lead & Soil Project
Maryland Department of the Environment
2500 Broening Highway
Baltimore, MD 21224
Dr. Paul A. Succop
University of Cincinnati
Department of Environmental Health
3223 Eden Avenue
Cincinnati, OH 45267-0056
Dr. Pat VanLeeuwen
U.S. EPA - Region V
Technical Support Unit 5HR-11
230 S. Dearborn Street
Chicago, IL 60404
Dr. Harold A. Vincent
U.S. EPA
Quality Assurance Division
Environmental Monitoring Systems
Laboratory - Las Vegas
P.O. Box 93478
Las Vegas, NV 89193-3478
Dr. Michael Weitzman
Chief of Pediatrics
Rochester General Hospital
1425 Portland Avenue
Rochester, NY 14621
July 15, 1993
xv
DRAFT-DO NOT QUOTE OR CITE
-------
Preceeding Page Blank
LIST OF REVIEWERS
COPY IS STILL TO COME.
July 15, 1993 xvii DRAFT-DO NOT QUOTE OR CITE
-------
Page Intentionally Left Blank
-------
LIST OF ABBREVIATIONS AND ACRONYMS
AAS
ANCOVA
ANOVA
Atomic Absorption Spectroscopy
Analysis of Covariance
Analysis of Variance
Autoregressive Regression Model Statistical procedure for multiple linear regression where
one variable is regressed on preceding values of the
variable as well as other related variables
BALP
BALSP
BOSP
BOS PI
BOS SPI
CDC
GIN I-SE
CINNT
CINSEI
CORR
dL
Double Blind
Dust Loading
ECAO/RTP
Baltimore Study Group with Paint Intervention
Baltimore Study Group with Soil and Paint Intervention
Boston Study Group with Paint Intervention
Boston Study Group with Paint and Interior Dust
Intervention
Boston Study Group with Soil, Paint, and Interior Dust
Intervention
Centers for Disease Control
Cincinnati Study Group with Interior Dust Intervention,
followed by Soil and Exterior Dust Intervention (second
year)
Cincinnati Study Group with No Treatment
Cincinnati Study Group with Soil, Exterior Dust, and
Interior Dust Intervention
Systat Procedure for pair-wise correlations
Deciliter, Used here as a measure of blood lead in jug/dL
Analytical audit sample where analyst knows neither that
the sample is an audit sample nor the concentration
Mass of dust per unit area
Environmental Criteria and Assessment Office/Research
Triangle Park
July 15, 1993
xix
DRAFT-DO NOT QUOTE OR CITE
-------
LIST OF ABBREVIATIONS AND ACRONYMS (cont'd)
EPA/ORD
EPA/OSWER
FAM
GIS
GLIM
GLM
GSD
Hand Dust
HEPA
ICP
K3D
MEANS
MGLH
NBHD
NHANESH
NONLIN
P-value
Pb
Pb Concentration
July 15, 1993
EPA/Office of Research and Development
EPA/Office of Solid Waste and Emergency Response
Datafile indexed by Family or Living Unit
Geographic Information Systems
Numerical Algorithms Group software package for a
general linear model
SAS procedure for general linear models approximately
equivalent to Systat MGLH
Geometric Standard Deviation
Sample taken by wiping the child's hand thoroughly; a
measure estimating the ingestion of lead
High Efficiency Particle Accumulator
Inductively Coupled Plasma Emission Spectroscopy
Datafile indexed by child
SAS Procedure for calculating means
Systat procedure for general linear models approximately
equivalent to SAS GLM
Datafile indexed by Neighborhood
National Health Assessment and Nutrition Examination
Survey n
Systat program for single response nonlinear regression
models
Statistical term for the likelihood that an observed effect
differs from zero
Lead
Mass of Pb per mass of medium (soil, dust, water)
xx DRAFT-DO NOT QUOTE OR CITE
-------
LIST OF ABBREVIATIONS AND ACRONYMS (cont'd)
Pb Loading
PC
Pearson Correlation Coefficient
Project
PROP
QA/QC
Repeated Measures Analysis
RFP
Round
SARA
SAS
SES
Single Blind
Study
SYSTAT
U.S. EPA
USLADP
XRF
July 15, 1993
Mass of Pb per unit area
Personal Computer
Statistical term for one measure of correlation between
two variables
In this report, "project" refers collectively to the three
individual studies that compose the Urban Soil Abatement
Demonstration Project.
Datafile indexed by Property
Quality Assurance/Quality Control
Statistical procedure for analyzing normally distributed
responses collected longitudinally
Request for Proposal
Period of sampling and data collection during study
Superfund Amendments and Reauthorization Act
Statistical Software Package
Socio-economic Status
Analytical audit sample where analyst knows sample is an
audit sample but doesn't know concentration (see Double
Blind)
In this report, "study" refers to one of the three
individual soil abatement studies that compose the Urban
Soil Abatement Demonstration Project.
Statistical Software Package
U.S. Environmental Protection Agency
Urban Soil Lead Abatement Demonstration Project
X-Ray Fluorescence
xxi DRAFT-DO NOT QUOTE OR CITE
-------
LIST OP ABBREVIATIONS AND ACRONYMS (cont'd)
XRFE Exterior measurements of Lead-based paint using portable
XRF instruments
XRH Interior measurements of Lead-based paint using portable
XRF instruments
July 15, 1993 xxii DRAFT-DO NOT QUOTE OR CITE
-------
i 1. EXECUTIVE SUMMARY
2
3
4 1.1 SCOPE OF REPORT
5 This document describes the results of the Urban Soil Lead Abatement Demonstration
6 Project (USLADP) from the perspective of the reanalysis of data from the three studies that
7 participated in the project. Taken individually, the reports of the three studies could support
8 three different conclusions concerning the feasibility of reducing lead exposure by abating
9 soil. Collectively, a common picture emerges that places a significant role for soil abatement
10 in the total scheme of lead exposure reduction. Because of the common design of the three
11 studies and their focus on key experimental parameters, it is possible to treat certain key
12 parameters as an integrated data set and analyze the results of each study separately but
13 identically. This report presents the results and conclusions drawn from a detailed statistical
14 reanalysis of the integrated data set.
15 The Urban Soil Lead Abatement Demonstration Project, known also as the Three City
16 Study, was authorized in 1986 under Section lll(b)(6) of the Superfund Amendments .and
17 Reauthorization Act (SARA). The purpose of the project was to determine whether
18 abatement of lead in soil could reduce the lead in blood of inner city children.
19 Until 1986, the concept of soil abatement had been applied mainly to residential
20 (usually nonurban) areas located near Superfund sites where lead was a major contaminant.
21 The decision to abate soil was usually based in part on the distribution of blood lead
22 concentrations within the population of children. There were few, if any, attempts to
23 measure the effects of this abatement and little or no opportunity to follow up with further
24 studies of the results.
25 This project is three coordinated longitudinal studies of urban children where
26 intervention into the pathway of lead exposure was expected to reduce the children's blood
27 lead. There have been many cross-sectional studies of childhood lead exposure under a wide
28 range of exposure conditions. These studies showed that differences in lead exposure
29 produced differences in blood lead concentrations. They did not show that changes in
30 exposure produce changes in blood lead. This would require a longitudinal study. Before
31 now, there were few longitudinal studies; none involved extensive intervention, and some
July 15, 1993 1-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 showed that Increasing lead exposure results in increasing blood lead concentrations. The
2 unique aspect of this project is that it measures a response to intervention, not contamination.
3 There is a difference between children who are suddenly placed into a "clean"
4 environment and those who have lived continuously in a clean environment. Because of the
5 physiology of lead mobilization in body tissues, there is a difference between the rate of
6 change in a population with increasing lead exposure and in one with decreasing exposure.
7 In other words, the decrease in blood lead concentrations in response to intervention was not
8 expected to be at the same rate as the increase in blood lead concentrations with increasing
9 exposure.
10 The project began in December 1986 with the appointment of a Steering Committee to
11 develop recommendations for implementing the SARA, lead-in-soil demonstration project.
12 A panel of experts was formed in March 1987 to set the criteria for selection of sites and the
13 minimum requirements for a. study at each site. An early decision was that the options for
14 soil abatement methods were limited, because only excavation and removal had been used in
15 similar programs. Therefore, there would be no attempt in the project to evaluate alternative
16 methods of abatement because of limited time and resources. The panel met again in April
17 1987 to discuss technical issues and study designs and evaluate technical criteria for selection
IS of urban areas as potential soil-lead abatement demonstration project sites. They established
19 site selection criteria that in December, 1987 led to the selection of Boston, Baltimore, and
20 Cincinnati as the participating sites.
21
22
23 1.2 BACKGROtJND AND OVERVIEW
24 In the mid 1980s, concern for childhood lead exposure increased with mounting
25 evidence that urban environments were exposed to lead in soil to a degree that might be
26 related to potential health effects. Evidence for this concern came from the apparent
27 correlation between the incidence of high blood lead concentrations and high concentrations
28 of lead in residential soils. At that time, there were several other sources of exposure that
29 could potentially account for unusually high blood lead in a population of urban children.
30 Among these were lead in the air (primarily from automobile emissions), lead in food
31 (primarily from canned foods with lead soldered side seams), lead in drinking water
July 15, 1993 1-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 (primarily from lead pipes or newly soldered copper pipes), and lead in paint. The lead in
2 the soil was believed to be a mixture of lead from the atmosphere and lead from exterior
3 paint. At that time, regulations were in place that would remove nearly all tetraethyl lead
4 from gasoline by the end of 1986, and there was a voluntary program among food processors
5 to phase out cans with lead soldered side seams and use only cans without lead solder. The
6 relationship between soil lead and blood lead is an indirect relationship in the sense that
7 children most commonly do not eat soil directly but ingest small amounts of dust derived, in
8 part, from this soil. In the child's environment, soil is only one of several sources of lead.
9 Likewise, the lead in blood reflects not only exposure from these sources but also the
10 biokinetic processes that distribute and redistribute lead between blood and other body
11 tissues, especially bone tissue.
12
13 1.2.1 Study Designs
14 Each study was designed around the concept of participating families within a definable
15 neighborhood. There were two or three study groups in each study, with one to three
16 neighborhoods in a study group. Each study group was evaluated during three phases:
17 preabatement, abatement and postabatement. This means that prior to and after abatement,
18 the environment of the child was extensively evaluated through measurements of lead in soil,
19 dust, drinking water, and paint, and through questionnaires about activity patterns, eating
20 habits, family activities, and socioeconomic status. The objective of the preabatement phase
21 was to determine the exposure history and status (stability of the blood lead and
22 environmental measures) prior to abatement. During the abatement phase, all possible
23 measures were taken to prevent exposure that might result from the abatement activities.
24 During the postabatement phase, samples were taken to measure the duration of the effect of
25 soil abatement and to detect possible ^contamination.
26 Because of the complex nature of this exposure assessment, intermediate exposure
27 indices, such as street dust, house dust, and hand dust were measured wherever possible.
28 This required the development of new sampling and analysis protocols that were not
29 generally available hi the scientific literature. These protocols were developed through a
30 Scientific Coordinating Committee composed of representatives from each study, the three
31 Regional offices, the CDC, EPA/Office of Solid Waste and Emergency Response,
July 15, 1993 1-3 DRAFT-DO NOT QUOTE OR CITE
-------
I EPA/Office of Research and Development, and Battelle/Research Triangle Park, NC. They
2 were largely complete by the start of the preabatement phase in January 1989.
3 Table 1-1 describes the study groups and the form of intervention. The Cincinnati
4 study design used intervention on the neighborhood scale, where the soil in parks, play areas
5 and other common grounds were abated, and paved surfaces in the neighborhood were
6 cleaned of exterior dust lead. In Boston and Baltimore, only the soil on the individual
7 properties was abated. Table 1-2 describes the study design characteristics for each of the
8 three studies and their respective participant groups. The general characteristics are that soil
9 lead concentrations are typically high in Boston, and it is common to find lead in drinking
10 water and in both exterior and interior paint. In the Boston areas studied, housing is
11 typically single family with relatively large lot sizes. In the Baltimore neighborhoods, nearly
12 every house had lead-based paint, the houses were mixed single and multifamily, and the lots
13 were smaller than Boston lots, with typical yards less than 100 square meters. Residential
14 units in Cincinnati were mostly multifamily with little or no soil on the residential parcel of
15 land.
16 Figure 1-1 illustrates the generalized concept of human exposure to lead, showing the
17 pathways of lead from the several sources in the human environment to four compartments
18 immediately proximal to the individual. One of these proximal sources, dust, is the route of
19 concern in this project. Figure 1-2 expands this dust route to show the complexity of the
20 many routes of dust exposure for the typical child. The strategies for intervention used in
21 this project were designed to interrupt the movement of lead along one or more of these
22 pathways.
23 Intervention is defined here as the interruption of the flow of lead along an exposure
24 pathway. There were three forms of intervention in this project: soil abatement, dust
25 removal, and paint stabilization. Soil abatement was by excavation and removal. If done
26 correctly, this abatement should establish an effective and persistent barrier to the dust
27 movement. Dust intervention was by vacuuming, wet mopping, and, in some cases,
28 replacement of rugs and upholstered furniture. Cincinnati and Boston performed interior dust
29 abatement, and Cincinnati removed neighborhood dust with mechanical sweepers and hand
30 tools. Dust intervention was not expected to be permanent, because dust continually flows
31 through the human environment. Instead, the removal of dust with elevated lead
July 15, 1993 1-4 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 1-1. DESCRIPTION OF STUDY GROUPS AND TYPES OF INTERVENTION
Cross-Reference to
Treatment Group Individual Study
Namea Report . Description of Treatment
BOSTON
BOS SPI Study Group Soil and interior dust abatement, and
exterior and interior paint stabilization at
beginning of first year, no further treatment
BOS PI Control Group A Interior dust abatement and exterior and
interior paint stabilization at beginning of
first year
BOS P Control Group B Exterior and interior paint stabilization at
beginning of first year
BALTIMORE
BAL SP Study Area Soil abatement and exterior paint
stabilization at beginning of first year, no
further treatment
BAL P-l Control Area Exterior paint stabilization at beginning of
first year, no further treatment
BAL P-2 Study Area Exterior paint stabilization at beginning of
first year, no further treatment
CINCINNATI
CIN SEI Area A Soil, exterior dust, and interior dust
abatement at beginning of first year, no
further treatment
CIN I-SE-10 Area B, Back Street Interior dust abatement at beginning of first
and Findlay year, soil and exterior dust abatement at
neighborhoods beginning of second year, no further
treatment
CIN I-SE-2C Area B, Dandridge Interior dust abatement at beginning of first
neighborhood year, soil and exterior dust abatement at
beginning of second year, no further
treatment
CIN NT Area C No treatment, soil abatement at end of
study
aThe treatment group designation indicates the location of the study (BOS = Boston, BAL = Baltimore,
CIN = Cincinnati), the type of treatment (S = soil abatement, E = exterior dust abatement, I = interior dust
abatement, P = loose paint stabilization, NT = no treatment).
Treated as one group in the Baltimore report, analyzed separately in this report.
°Treated as one group in the Cincinnati report, analyzed separately in this report.
July 15, 1993 1-5 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 1-2. NUMBER OF PROJECT PARTICIPANTS BY ROUND3
Study
BOSTON
Middatc
Children1"
Round 1
10/17/89
150
Round3
4/9/90
146
Round 4
9/12/90
147
Famlics
125
121
122
Properties
BALTIMORE
Middatc
Children1"
Families0
Properties
CINCINNATI
Middatc
Children
Families'5
Properties
100
Round 1
10/25/88
408
200
193
Round 1
7/6/89
201
71
141
96
Round 2
4/1/89
322
203
201
Rounds
11/14/89
185
67
129
97
Round 3
2/17/90
269
157
156
Round 5
7/1/90
219
66
124
Round 4
1/27/91
200
116
114
Round?
11/17/90
198
94
124
Round 5 Round 6
6/7/91 9/3/91
196 187
110 109
108 108
Round 9
6/16/91
169
82
124
"Number shown is based on samples taken and does not include individuals enrolled but not sampled.
Based on number of children sampled for blood.
eased on number of households sampled for dust.
Based on number of properties (Boston, Baltimore) or soil parcels (Cincinnati) sampled.
July 15, 1993
1-6
DRAFT-DO NOT QUOTE OR CITE
-------
Auto
Emissions
Paint,
Industrial
Dusts
Figure 1-1. Generalized concept of the sources and pathways of lead exposure in
humans.
Atmospheric
Particles
Local
Fugitive
Dust
Soil
Exterior Paint
Dust
Interior Paint
Dust
Exterior
Dust
Interior
Dust
Secondary^
| Occupational
Dust j
Hand
Dust
Child
Figure 1-2. Typical pathways of childhood exposure to lead hi dust.
July 15, 1993 1-7 DRAFT-DO NOT QUOTE OR CITE
-------
1 concentrations was expected to enhance the impact of soil abatement on the child's
2 environment.
3 In the home, dust is a mixture of exterior dust and soil, interior weathering products
4 including paint flakes and chips, workplace dust carried home by adults, and dust generated
5 from human activities within the household. It is believed that most of the mass of the
6 interior dust originates from soil immediately exterior to the home, but this can vary greatly
7 by the types of family activities. Nevertheless, in the absence of lead-based paint inside the
8 home, it would seem reasonable to assume that most of the lead in household dust comes
9 from soil.
10 Many of the Boston and Baltimore households selected for the project had chipping and
11 peeling lead-based paint, both interior and exterior. In order to reduce the impact of this
12 paint, much of which was lead-based paint, the walls and other surfaces were scraped and
13 smoothed, and repainted to reduce the impact of lead-based paint on the pathways .of lead
14 exposure. It is important to note that this approach in not a full scale paint abatement and
15 was not designed to place a permanent barrier between the paint and the child. Paint
16 stabilization was used on exterior and interior surfaces in Boston, and on exterior surfaces in
17 Baltimore. Paint stabilization was not used in Cincinnati because the lead-based paint had
18 been removed from these homes hi the early 1970s.
19 The frequency and timing of sampling relative to abatement and seasonal cycles is an
20 important aspect of this project. The original design focussed on sampling blood lead during
21 the late summer, as it was known that the seasonal cycle is highest at that time. Where this
22 schedule could not be adhered to, an effort was made to schedule the followup blood lead
23 sampling to characterize this cycle and possibly permit extrapolation to the summer peaks.
24 In order to accurately measure the effectiveness and persistency achieved by soil
25 abatement, and the impact of this abatement on reducing lead exposure for children, the
26 sampling and analysis plans for soil and dust required robust quality control and quality
27 assurance objectives. Protocols were developed to (1) define sampling schemes that
28 characterize the expected exposure to soil for children; (2) collect, transfer, and store
29 samples without contamination; and (3) analyze soil, dust, handwipe, and blood samples in
30 a manner that would maximize interlaboratory comparison.
31
July 15, 1993 1-8 DRAFT-DO NOT QUOTE OR CITE
-------
1 1.3 BRIEF DESCRIPTION OF HOW INTERVENTION WAS
2 PERFORMED
3 A total of 93 Boston properties were abated. The information on area treated and
4 volume of soil removed from these properties appears in Table 1-3.
5
TABLE 1-3. SOIL ABATEMENT STATISTICS FOR THE THREE STUDIES
Number of properties*
2
Surface area (m )
Volume soil removed (m3)
Surface area/property (m )
Volume soil/property (m )
Boston
36
7,198
1,212
200
34
Baltimore
56
4,100b
690
73
llb
Cincinnati
171
12,089
1,813
71
11
alncludes only properties abated during the study. Properties abated at the end of the study, where no further
sampling was reported, are not included in this analysis, but are included in the individual study reports.
In Cincinnati, a property is the location of the soil abatement, not the location of the child's residence.
Surface area not provided by Baltimore report. Calculated using Boston vol/surf ratio, which is equivalent to
an average removal depth of 17 cm.
1 In Baltimore, 63 properties in the BAL SP treatment group (see Table 1-1) were abated
2 between August and November 1990. An additional seven properties that did not meet the
3 requirements for abatement were transferred to the control group (BAL P). Unpaved
4 surfaces were divided into areas on each property, usually front, back, and one side; any
5 area with soil lead concentrations above 500 jug/g was abated entirely.
6 Within each neighborhood, the Cincinnati study identified all sites with soil cover as
7 discrete study sites. The decision to abate was based on soil lead concentrations for each
8 parcel of land, and for the depth to which the lead had penetrated. Lead was measured in
9 the top 2 cm, and at a depth of 13 to 15 cm. If the concentration in the top sample was
10 greater than 500 jug/g, the soil was abated. Additional parcels were abated if there was
11 evidence that lead had penetrated into the soil profile.
12 Exterior dust abatement was performed in the Cincinnati study only. The approach to
13 this abatement was to identify all types of paved surfaces where dust might collect, obtain
14 permission to sample and abate these areas and to clean them once with vacuum equipment,
July 15, 1993 1-9 DRAFT-DO NOT QUOTE OR CITE
-------
1 suitable for the area, that had previously been tested and shown to remove about 95 % of the
2 available dust on the area. The groups of surfaces selected were streets, alleys, sidewalks,
3 parking lots, steps, and porches. For data analysis, these were grouped as (1) targeted (steps
4 and porches); (2) streets, sidewalks, and alleys; and (3) parking lots and other locations.
5 The exterior dust measurements in the Cincinnati study (and the interior dust
6 measurements of all three studies) were made in a manner that determined the lead
7 concentration (p.g Pb/g dust), the dust loading (mg dust/in), and the lead loading (/xg Pb/m )
8 for the surface measured. This required that a dry vacuum sample be taken over a
r\
9 prescribed area, usually 0.25 to 0.5 m . It is important to note that dust abatement is not
10 expected to cause an immediate change in the lead concentration on dust surfaces, only in the
11 dust and lead loading.
12 Household dust was abated in the Boston and Cincinnati studies, but not in Baltimore.
13 The BOS SPI and CEST SEE groups (see Table 1-1) received interior dust abatement at the
14 same time as soil abatement, the BOS PI group received interior dust abatement without soil
15 abatement, and the CINI-SE-1 and CDSTI-SE-2 groups received interior dust abatement in
16 the first year, followed by soil and exterior dust abatement in the second year.
17 In Boston, interior dust abatement was performed after loose paint stabilization.
18 Families were moved off site during interior dust abatement. Hard surfaces (floors,
19 woodwork, window wells, and some furniture) were vacuumed, as were soft surfaces such as
20 rugs and upholstered furniture. Hard surfaces were also wiped following vacuuming.
21 Common entries and stairways outside the apartment were not abated.
22 The Cincinnati group performed interior dust abatement after exterior dust abatement
23 and also moved the families off site during this activity. Vacuuming, was followed by wet
24 wiping with a detergent. They vacuumed hard surfaces and replaced one to three carpets and
25 two items of upholstered furniture per housing unit. Their previous studies had shown that
26 these soft items could not be cleaned effectively with vacuuming alone.
27 Most homes in the Cincinnati group had received paint abatement 20 years prior to the
28 project, but in Boston and Baltimore lead-based paint occurred in nearly every home.
29 Because Ml paint abatement was not within the scope of this project, the alternative was to
30 retard the rate of movement of paint from the walls to household dust to the extent possible.
31 The interior and exterior surfaces of all Boston homes and the exterior surfaces of all
July 15, 1993 .1-10 DRAFT-DO NOT QUOTE OR CITE
-------
1 Baltimore homes received loose paint stabilization approximately one week before soil
2 abatement.
3 In Boston, loose paint stabilization consisted of removing chipping and peeling paint
4 and washing the surfaces. Window wells were painted with a fresh coat of primer.
5 Baltimore homes were wet scraped over the chipping and peeling surfaces, followed by
6 vacuuming. The entire surface was primed and painted with two coats of latex paint
7
8
9 1.4 BRIEF SUMMARY OF INDIVIDUAL STUDY REPORTS
10 1.4.1 Summary of the Boston Study
11 The Boston study retained 149 of the original 152 children enrolled, although
12 22 children moved to a new location but were retained in the study. Children with blood
13 lead concentrations below 7 /xg/dL or above 24 /xg/dL had been excluded from the study and
14 two children were dropped from the data analysis when they developed lead poisoning,
15 probably due to exposure to lead-based paint at another location.
16 Baseline characteristics (age, SES, Soil lead, dust lead, drinking water lead, and paint
17 lead) were similar for the three study groups (BOS P, BOS SP, BOS SPI). The pre-
18 abatement blood lead concentration was higher for BOS P. The proportion of Hispanics was
19 higher in BOS P the BOS SP or BOS SPI, and the proportion of Blacks was lower. There
20 was a larger proportion of male children in BOS P.
21 Data were analyzed by analysis of covariance (ANCOVA), which showed a significant
22 effect of group assignment (intervention) for both the BOS SP and BOS SPI groups. These
23 results did not change with age, sex, socioeconomic status, or any other variable except race
24 and paint. When the paint variable was added, the effect was diminished; when the race
25 variable was added, the effect became insignificant.
26 Although designed and conducted to produce rigorous results, the study has several
27 limitations. Participants were chosen to be representative of the population of urban
28 preschool children who are at risk of lead exposure by using the Boston Childhood Lead
29 Poisoning Prevention Program to identify potential participants from neighborhoods with the
30 highest rates of lead poisoning and by using as wide a range of blood lead levels as was
31 practical. Since no study subjects had blood lead levels below 7 jwg/dL or in excess of
July 15, 1993 1-11 DRAFT-DO NOT QUOTE OR CITE
-------
I 24 jtg/dL at baseline, the study provides no information about the effect of lead contaminated
2 soil abatement for children with these lead levels. Similarly, a different effect might have
3 been found for children who had a greater blood lead contribution from soil, such as in
4 communities with smelters or other stationary sources where soil lead levels are substantially
5 higher than those seen in this study, or where differences in particle size result in differences
6 in bioavailability.
7 It is possible that the intervention would have been associated with a greater reduction
8 in children's blood lead levels had they been followed for a longer period of time.
9 In addition, all children in the study were exposed to lead contaminated soil prior to
10 enrollment and so we are unable to investigate whether exposure to lead contaminated soil in
11 the first year of life is associated with higher blood lead levels. Lastly, the unit of abatement
12 was the single premises rather than clusters of premises. It is possible that the effect of lead
13 contaminated soil abatement on children's blood lead levels would have been greater had we
14 also removed lead contaminated soil from properties that surrounded Study Group children's
15 premises.
16 In conclusion, this intervention study suggests that an average 1,856 ppm reduction in
17 soil lead levels results in a 0.8 to 1.6 /tg/dL reduction in the blood lead levels of urban
18 children with multiple potential sources of exposure to lead.
19 This study provides information about soil abatement as a secondary prevention
20 strategy, that is the benefit to children already exposed to lead derived, in part, from
21 contaminated soil. It can not be used to estimate the primary prevention effect of soil
22 abatement. Since children's postabatement blood lead levels reflect both recent exposure and
23 body burdens from past exposure, the benefit observed is probably less than the primary
24 prevention benefit, that is the benefit of abating lead contaminated soil before children are
25 exposed to it so as to prevent increases in blood levels and body stores.
26
27 1.4.2 Summary of the Baltimore Study
28 The Baltimore study recruited 472 children, of whom 185 completed the study.
29 Of those that completed the study, none were excluded from analysis. The recruited children
30 were from two neighborhoods, originally intended to be a study and a control group.
31 Because soil concentrations were lower than expected, some properties in the study group did
July 15, 1993 1-12 DRAFT-DO NOT QUOTE OR CITE
-------
1 not receive soil abatement. The Baltimore report transferred these properties to the control
2 group. In this report, the low soil properties in the study group are treater as a separate
3 group.
4 Because of logistical problems, there was an extended delay between recruitment and
5 soil abatement that accounted for most of the loss of the participating families from the
6 project. In their report, the Baltimore group applied several statistical models to the two
7 populations to evaluate the potential bias from loss of participating children. These analyses
8 showed the two populations remained virtually identical in demographic, biological and
9 environmental properties.
10 The Baltimore study was not designed to focus on measurements of the movement of
11 lead through the child's environment, Repeat measurements of soil were on abated
12 properties only, to confirm abatement. There were no measurements of exterior dust, no
13 interior paint stabilization, and no followup measurements of house dust. Rather, the study
14 design focused on changes in biological parameters, hand dust and blood lead over an
15 extended period of time,
16 Including the pre-study screening measurements of hand dust and blood lead in the
17 original cohort of participants, the Baltimore study made six rounds of biological
18 measurements that spanned twenty months. It is unusual to have a data set of this
19 composition and quality. In this integrated report, the baltimore blood lead measurements
20 were the basis for determining the key parameters in the seasonal cycle conversion factor
21 equation discussed in Section 3.3.5.1,
22 Soil was abated between the third and fourth rounds of biological measurements. The
23 mean soil decrease was 550 ^ig/g. At Round 4, the blood lead concentrations were about
24 0.5 ^ig/dL lower in the study group than in the control, or 1 /ug/dL per decrease of
25 1,000 yiig/g in soil, which is comparable to the response observed in the Boston study.
26 By Rounds 5 and 6, the study group blood lead concentrations had returned to their
27 preabatement levels and were in fact higher than the control group,
28 From the perspective of the Baltimore study alone, it is reasonable to conclude, as the
29 Baltimore report did, that soil abatement has no effect on children's blood lead. But from
30 the perspective of the Boston study, where a blood lead reduction of the same magnitude was
31 found to be persistent when house dust abatement was performed, and from the perspective
July 15, 1993 1-13 DRAFT-DO NOT QUOTE OR CITE
-------
1 of the Cincinnati study, where blood lead concentrations were shown to rise and fall in
2 tandem with house dust concentrations, the results of the Baltimore study are consistent with
3 the observation that sou abatement, in conjunction with other environmental interventions,
4 can permanently reduce exposure to lead.
5
6 1.4.3 Summary of the Cincinnati Study
7 The Cincinnati study recruited 307 children, including 16 children born to participating
8 families during the study, and an additional 50 children who were recruited after the
9 beginning of the study. In their final report, the Cincinnati group excluded these children
10 who were recruited after the start of the study, plus 31 children who were living in
11 nonrehabilitated housing suspected of having lead-based paint, and four children (in two
12 families) who had become lead-poisoned from other causes. Thus, data for 210 children
13 were analyzed in the Cincinnati report and these same children were included in this
14 integrated report.
15 The Cincinnati study achieved effective and persistent abatement of soE on the
16 140 parcels of land scattered throughout the neighborhoods. In CIN SEI, where soil
17 abatement was performed in the first year, the arithmetic mean concentration dropped from
18 680 /tg/g down to 134 /*g/g. In the two groups where soil abatement occurred in the second
19 year, CIN I-SE-1 and CIN I-SE-2, the soil lead concentration dropped from 262 jig/g to
20 125 ^g/g and 724 jtg/g to 233 /zg/g, respectively.
21 If soil were the only source of lead in the neighborhoods, exterior and interior dust
22 should have responded to the reduction in soil lead concentrations. Exterior dust lead
23 loading decreased following both soil and dust abatement, but returned to preabatement levels
24 within one year. In their report, the Cincinnati group concluded that recontamination of
25 exterior dust began soon after abatement. They observed corresponding changes in house
26 dust, hand lead, and blood lead that paralleled changes in exterior dust. Because blood lead
27 concentrations also decreased in the control area, the Cincinnati group concluded that there is
28 no evidence for the impact of soil and dust abatement on blood lead concentrations. This
29 integrated report concludes, through a more detailed structural equation analysis, that there is
30 a strong relationship between exterior dust and interior dust in this subset of the Cincinnati
31 study where the impact of lead-based paint was minimized. From the perspective of all three
v
July 15, 1993 1-14 DRAFT-DO NOT QUOTE OR CITE
-------
1 studies, this means that when neighborhood and living unit sources of lead are removed,
2 exposure is reduced.
3 The Scientific Coordinating Committee attempted to establish uniformity among the
4 three studies for several aspects of the project. Although there were differences in form and
5 content, each study plan contained:
6 1. a statement of the objectives of the study;
7 I . '
8 2, a testable hypothesis that provided direction and focus to the study;
9 |
10 3. protocols for collecting and analyzing the data;
11 j
12 4. an array of treatment groups that addressed all features of the hypothesis;
13
14 5, measures to be taken to ensure that all phases of the study would be conducted as
15 planned; and
16 I
17 6. procedures by which the results of the study would be processed, analyzed, and
18 interpreted.
19 [
i ...
20 The objectives, protocols for sampling and analysis, QA/QC plans, and data processing
21 procedures were nearly identical for all three studies. Elements that differed slightly among
j
22 the three studies were the hypotheses and the array of treatment groups. The hypotheses
23 differed only slightly, as seen from the following statements.
24 The central hypothesis of the Urban Soil Lead Abatement Demonstration Project is:
25 !
26 A reduction of lead in residential soil accessible to children will
21 result in a decrease in their blood lead levels.
28 | ,
29 The formal statement of the Boston hypothesis is:
30
31 A significant reduction (equal to or greater than 1,000 pg/g) of lead
32 in soil accessible to children will result in a mean decrease of at
33 least 3 pg/dL in the blood lead levels of children living in areas with
34 multiple possible sources of lead exposure and a high incidence of
35 lead poisoning.
36 1
37 The Baltimore hypothesis, stated in the null form, is:
38 j
39 A significant reduction of lead (> 1,000 pg/g) in residential soil
40 accessible to children will not result in a significant decrease
41 (3 to 6 ng/dL) in their blood lead levels.
42 |
July 15, 1993 j 1-15 DRAFT-DO NOT QUOTE OR CITE
-------
1 The Cincinnati hypothesis was separated into two parts: a
2
3 (1) A reduction of lead in residential soil accessible to children will result
4 i?i a decrease in their blood lead levels.
5
6 (2) Interior dust abatement, when carried out in conjunction with exterior
7 dust and soil abatement, would result in a greater reduction in blood
8 lead tJian would be obtained with interior dust abatement alone, or
9 exterior dust and soil abatement alone.
10
11 Secondary hypotheses in the Cincinnati study are:
12
13 (3) A reduction of lead in residential soil accessible to children will result
14 in a decrease in their hand lead levels.
15
16 (4) Interior dust abatement, when carried out in conjunction with exterior
17 dust and soil abatement, would result in a greater reduction in hand
18 lead than would be obtained with interior dust abatement alone, or
19 exterior dust and soil abatement alone.
20
21 The array of treatment groups differed considerably among the three studies
22 (Table 1-1). Each treatment group, however, had several features in common. All groups
23 were taken from one to three demographically similar neighborhoods. All groups had some
24 prior evidence of elevated lead exposure, usually a greater than average number of reports of
25 lead poisoning. Each group received the same pattern of treatment: baseline phase for 3 to
26 18 mo, intervention (except for controls), and followup for 12 to 24 mo.
27 In each treatment group, even the controls, there was an attempt to minimize the impact
28 of lead-based paint. In Boston, this was done by paint stabilization of both interior and
29 exterior paint. In Baltimore, only exterior paint was stabilized. Therefore, in these two
30 studies, the effects of soil abatement should be evaluated in the context of some intervention
31 for lead-based paint. In Cincinnati, most of the living units had been abated of lead-based
32 paint more than 20 years before the start of the study. Those that had not been abated were
33 measured but not treated prior to the study and were in included in the final analysis.
34 Another difference between the studies was the parallel intervention scheme used in
35 Boston and Baltimore, compared to the staggered scheme used in Cincinnati. In other
36 words, intervention in Boston (and Baltimore) took place at the same time for all treatment
37 groups, and the followup period was of the same duration. But in Cincinnati, the
July 15, 1993 1-16 DRAFT-DO NOT QUOTE OR CITE
-------
1 intervention was delayed for one group, CINI-SE, such that followup varied between
2 12 and 24 mo.
3
4
5 1.5 SUMMARY OF RESULTS AND STATISTICAL INFERENCES
6 From the perspective of the child's environment, changes in the soil lead concentration
7 are expected to bring about changes in the house dust concentration, the hand dust, and the
8 blood lead concentration. In each of the three studies, the soil lead concentrations were
9 reduced to approximately 50 jitg/g in the study area, and for most children, there was a
10 measurable reduction of blood lead, although not always statistically significant. When
11 corrected for seasonal and age related cyclic variations on blood lead, the impact was even
12 greater, and the effect was maximized when street dust and house dust were also removed
13 from the environment.
14
15 1.5.1 Quality of the Data
16 In the absence of certified standards for soil and dust, it was necessary to put into
17 place a program that would ensure that analyses performed by the three participating
18 laboratories would be internally accurate and externally consistent with similar analyses by
19 other researchers. This program consisted of identifying acceptable analytical and
20 instrumental methods, establishing a set of soil and dust standards, and monitoring the
21 performance of the participating laboratories through an external audit program.
22 Because chemical extraction of 75,000 soil and dust samples presented a costly burden
23 on the project both in terms of time and expense, and because of the advantage of
24 nondestructive analysis for a project of this nature, the Scientific Coordinating Panel
25 recommended the use of XRF for soil analysis on the condition that a suitable set of common
26 standards could be prepared for a broader concentration range and that a rigorous audit
27 program be established to ensure continued analytical accuracy. Two groups, Boston and
28 Baltimore, elected to use XRF for dust analysis also, whereas Cincinnati opted for hot nitric
29 extraction with AAS. During the study, the Baltimore group recognized problems with
30 analyzing dust by XRF when the sample size was small, less than 100 mg. They reanalyzed
31 the dust samples by AAS and reported both measurements. In Boston, this problem was
July 15, 1993 1-17 DRAFT-DO NOT QUOTE OR CITE
-------
1 solved by compositing the floor dust samples for XRF analysis, reporting one floor dust
2 sample per housing unit.
3 During the project, there were two rounds of soil and dust intercalibration
4 measurements, one near the beginning and one at the completion of the soil and dust
5 analyses. These exercises involved the three participating laboratories and two additional
6 laboratories for each exercise. These exercises provided the basis for the conversion factors
7 used to compare soil and dust data between laboratories, and for the evaluation of the
8 performance of each laboratory in the audit sample program.
9 For soil and dust, multiple measurements must be reduced to a single representative
10 data point for each property or living unit for each round of measurement. This measure of
11 central tendency was reported differently for each of the three studies. Boston used the
12 arithmetic mean, giving equal weight to all values. Cincinnati used the geometric mean,
13 which gives lesser weight to the extremes and is always lower than the arithmetic mean
14 except when the distribution is perfectly normal. Baltimore used a tri-mean approach that
15 gives lesser weight to the extremes while not underestimating exposure for right-skewed
16 distributions.
17 Each study maintained rigorous standards for database quality. These included double
18 entry, 100% visual confirmation, and standard procedures for detecting outliers. Additional
19 errors were found during the preparation of this report and corrected prior to use in this
20 report. None of these errors would have impacted the conclusions drawn by the individual
21 study. To minimize further errors that might impact this report, statistical procedures were
22 repeated to replicate the results of each report and confirm the exactness of data.
23
24 1.5.2 Effectiveness and Persistency of Intervention
25 Soil abatement was found to be effective in all three studies and persistent in both
26 Boston and Cincinnati. There was no measure of soil abatement persistency in Baltimore.
27 Evidence for exterior dust recontamination in Cincinnati suggests lack of effectiveness or
28 persistency of abatement. Further analysis of the data may resolve the issue of the source of
29 this recontamination.
30 Interior dust abatement was effective and persistent in both Boston and Cincinnati,
31 even though some recontamination occurred in Cincinnati in response to the exterior dust
July 15, 1993 1-18 DRAFT-DO NOT QUOTE OR CITE
-------
1 recontamination. Paint stabilization appeared to have some impact on exposure, but there
2 were no measures of effectiveness or persistency.
3
4 1.5.3 Summary of Statistical Inferences
5 From the Boston study, a statistical analysis of the data shows that when dust lead and
6 soil lead levels show a persistent decrease as a result of effective abatement, blood lead
7 levels also show a persistent decline. The postabatement blood lead levels are lower when
8 postabatement dust lead levels are persistently lower over a long time.
9 From the Baltimore study, the analyses show that soil lead abatement had little effect
10 on the primary factors responsible for elevating child blood lead levels, which appear to be
11 interior lead-based paint and interior dust lead.
12 In Cincinnati, there appear to be additional sources of environmental lead exposure
13 that had different effects on the neighborhoods during the course of the study and were not
14 related to the abatement methods used in the Cincinnati study.
15 The analysis of the data from the three studies showed evidence that blood lead
16 responds to changes in environmental lead. This suggests that abatement of any type and to
17 any degree will cause a reduction in the blood lead of children. All three studies and all
18 groups within each study produced data supporting this conclusion, although not statistically
19 significant in two of the groups in the Baltimore study.
20 All three studies also showed evidence for a quantifiable impact of intervention. This
21 may have been intervention from soil abatement, dust abatement, or paint stabilization.
22 In Baltimore, this impact was temporary at best and was marginally significant.
23 In Cincinnati, the impact was quickly swamped by other sources of environmental lead.
24 In Boston, the impact was persistent. The best estimate for this effect is 1 /ig/dL per
25 1,000 /Ag/g decrease in soil. Similar decreases in exterior dust would be expected to have a
26 similar effect.
27 There is evidence from all three studies that lead moves throughout the child's
28 environment. This means that lead in soil becomes lead in street or playground dust, lead in
29 paint becomes lead in soil, and lead in street dust becomes lead in house dust, A more
30 detailed analysis of the data may show the relative contribution from two or more sources,
31 but the present analyses confirm that this transfer takes place. In the Baltimore study, there
July 15, 1993 1-19 DRAFT-DO NOT QUOTE OR CITE
-------
1 was statistical evidence for implied causal pathways, such as paint to exterior dust, but in the
2 Boston and Cincinnati studies, the pathways were explicit.
3 Finally, there is evidence for the continued impact of nonabated sources following
4 abatement. This means that abatement of soil probably does not reduce the contribution of
5 paint lead to the child's exposure.
6
7
8 1.6 SUMMARY STATEMENT OF PROJECT CONCLUSIONS AND
9 TKQEIR IMPLICATIONS
10 1.6.1 Project Conclusions
11 This report concurs with the results reported by the individual studies. The reanalysis
12 of the individual study data sets, where performed in the same manner as the report, revealed
13 no evident errors in statistical analysis. The statistical analyses revealed that the relationship
14 between environmental lead and blood lead was more or less uniform across all three studies.
15 When the environmental lead increased, the blood lead increased, and when environmental
16 lead decreased, blood lead decreased.
17 The results of these three studies demonstrated a clear relationship between
18 environmental lead and blood lead. There were few instances where changes in the blood
19 lead concentrations could not be attributed to changes in environmental lead. Unexplained
20 changes appeared to cause an increase in blood lead concentrations. The fact that both hand
21 dust loads and blood lead concentrations responded accordingly gives credence to the strong
22 link between environmental lead and blood lead.
23 Finally, the project results shed additional light on the well-known phenomenon of
24 seasonal cycles in blood lead concentrations. The rare opportunity to evaluate three
25 independent longitudinal studies with similar sampling and analysis protocols led to the
26 conclusion that the amplitude of the cycle is roughly 15 % in all three cities, that the peak
27 occurs about August 15-20, and that these factors appear to be independent of environmental
28 lead.
29 In terms of changes attributed to intervention, all three studies observed a quantifiable
30 change in response to intervention. The analyses in Chapter 4 show that, although not
31 always statistically significant, this quantifiable response to intervention is consistent even at
July 15, 1993 1-20 DRAFT-DO NOT QUOTE OR CITE
-------
1 low levels of environmental lead. Normalized to a decrease in soil of 1,000 j«g/g, the
2 response appears to be in the range of 0.7 to 1.5 jug/dL. This suggests that there is no
3 plateau, within the ranges measured in this project, where the removal of environmental lead
4 will not produce a corresponding reduction in blood lead concentrations.
5 It is expected that the data of this project will be analyzed and reanalyzed by many
6 qualified scientists to extract further information from this massive database. As the need
7 arises and time permits, further analyses may shed light, through additional structural
8 equation modeling and meta analysis, on the intricate pattern of environmental lead exposure
9 in urban neighborhoods. Much of the data are amenable to Geographic Information System
10 (GIS) studies and can be useful to state and local public health officials in assessing the
11 extent of lead exposure in their own domain. The efforts of the many investigators in this
12 project have produced much useful information and, as usual in a well-planned and executed
13 study, raised many additional questions.
14
15 1.6.2 Implications
16 In spite of the recent successes in reducing exposure to lead by removing lead from
17 gasoline and canned food, lead exposure remains a complex issue. This integrated report
18 attempts to assess exposure to lead in soil and house dust. It is only one component of the
19 risk assessment process and cannot by itself be the sole basis for a risk management decision,
20 However, the observations and conclusions are based on sound scientific measurements and
21 reasonable interpretations of these measurements. A thorough understanding of the results of
22 this project can provide guidance for regulatory decisions and public health policies.
23 This report concludes that a reduction in environmental lead corresponding to a
24 decrease of 1,000 jug/g in soil will result hi a reduction of about 1 /^g/dL in blood lead.
25 Although this modest decrease suggests that soil abatement as a form of environmental
26 intervention would not be particularly effective in clinical treatment of a lead poisoned child,
27 in an environmental intervention program where the goal is to reduce the incidence of blood
28 lead concentrations above 10 pcg/dL, this small change could reduce this incidence by 10 to
29 15%.
30 Lead in soil and lead-based paint are closely linked in the child's environment.
31 If there is exterior lead-based paint, then soil lead is likely to be elevated. If there is interior
July 15, 1993 1-21 DRAFT-DO NOT QUOTE OR CITE
-------
1 lead-based paint, then soil abatement measures to reduce the impact of soil lead on house
2 dust will be ineffective. Public health programs designed to reduce lead exposure will not
3 achieve that objective unless both paint and soil abatement are implemented.
4 From a regulatory standpoint, where the goal is to determine a safe level of lead in
5 soil, this report concludes that abatement of soil above 500 /*g/g will measurably reduce
6 blood lead concentrations. It does not say that this reduction in blood lead would be
7 permanent or cost-effective.
8 From another perspective, decisions about soil abatement are likely to be made on an
9 individual home basis or on a neighborhood basis. For an individual home, the owner or
10 renter may require only peace of mind in knowing that the property is safe-for children if the
11 soil lead concentrations are below an acceptable level, or, if not, that soil abatement would
12 be a cost effective way to reduce or eliminate the problem.
13 This project shows that, on an individual house basis, soil abatement reduces the flow
*
14 of lead into the home and its incorporation into house dust. The magnitude of this reduction
15 depends on the concentration of lead in the soil, the amount of soil-derived dust that moves
16 into the home, and the frequency of cleaning in the home. The number and ages of children
17 and the presence of indoor/outdoor pets are factors known to increase this rate, whereas the
18 frequency of cleaning with an effective vacuum cleaner and removing shoes at the door serve
19 to reduce the impact of soil lead on house dust.
20 On a neighborhood basis, the focus of concern is usually directed at the local public
21 health officer, who faces a risk management decision for which an exposure assessment
22 based on the results of this project is only one element. Guidance in this case should provide
23 general exposure scenario information that would assist the officer in predicting blood lead
24 concentrations should soil be abated. There are many alternatives to soil abatement, such as
25 control of pets, frequency of cleaning, or preservation of ground cover, available to the
26 individual home owner or renter, that may not be practical for the public health officer
27 making a decision on the neighborhood level.
July 15, 1993 1-22 DRAFT-DO NOT QUOTE OR CITE
-------
i 2. BACKGROUND AND OVERVIEW OF PROJECT
2
3
4 2.1 PURPOSE OF THIS DOCUMENT
5 The purpose of this document is to describe the results of the Urban Soil Lead
6 Abatement Demonstration Project (USLADP) from the perspective of all three studies that
7 participated in the project. Taken individually, the reports of the three studies could support
8 three different conclusions concerning the feasibility of reducing lead exposure by abating
9 soil. Collectively, a common picture emerges that places a significant role for soil abatement
10 in the total scheme of lead exposure reduction.
11 The purpose of USLADP was to determine if intervention in the form of soil abatement
12 would reduce childhood exposure to lead. The project has taken nearly eight years from
13 conception to completion. Each of the three studies in the project is a longitudinal study of
14 the impact of an altered environment on the lead exposure of children. There are few other
15 longitudinal studies of this type, and none of this scope or duration. Furthermore, the three
16 studies were conducted using common protocols where possible, so that integrated analyses
17 can broaden the base of information beyond the limits of a single study or location.
18 The project provides information for policy makers who need to recommend acceptable
19 cleanup levels of lead in soil and dust that apply broadly to entire cities, states or the entire
20 country, for public health officials who must provide guidance on site specific abatement
21 decisions for individuals and families about the potential hazard on a single property, and for
22 the scientific community whose job it is to challenge and refute incorrect information and to
23 correct that information by further research.
24 For the policy maker, the Executive Summary (Chapter 1), provides a brief overview
25 of the project, a short synopsis of the results, and a discussion of the recommendations and
26 conclusions of the authors. For the Public Health Official who needs more detail in order to
27 relate site specific situations to the generalized findings of this report, Chapter 2 provides a
28 detailed description of the project and Chapter 5 gives an extended discussion of the
29 conclusions and recommendations. A graphical presentation of the results may be found in
30 Chapter 3 for the reader who seeks to understand in full detail the movement of lead in the
July 15, 1993 2-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 human environment. These graphs, which systematically demonstrate the nature of this
2 process, are supported by an array of statistical tests that are described in Chapter 4.
3 This document will reach its final form only after an extensive review process, First,
4 the reports of the individual studies were reviewed by a panel of experts, revised, and
5 presented to the U.S. EPA in their final form, along with the data sets that were used as the
6 basis for the individual reports. These data sets were reanalyzed by EPA using rigorous
7 statistical techniques to extract information not easily accessible with a single data set, and
8 the integrated report, this document, was written based on these analyses. Following internal
9 review and revision, the integrated report will be released in draft form for public comment
10 and external review. The report will become final after the comments from-the external
11 panel of experts have been addressed and its release has been approved by senior EPA
12 officials. At that time, members of the scientific community who have a legitimate research
13 interest in the analysis of the data can obtain a copy of the data set for continued review and
14 analysis.
15 Although the three studies were conducted independently, an effort was made to
16 coordinate the critical scientific aspects of each study hi order to provide comparable data at
17 their completion. This effort included several workshops where the study designs, sampling
18 procedures, analytical protocols, and QA/QC requirements of each study were discussed with
19 a goal toward reaching a common agreement. In most cases, a consensus was reached on the
20 resolution of specific issues, but the individual studies were not bound to conform to that
21 consensus or to adhere to it throughout the study. Therefore, some attention will be given in
22 this section to the differences in study design and experimental procedures among the three
23 individual studies.
24 The results of these projects were presented at a symposium hi August 1992. These
25 presentations included the data analysis and conclusions of the three individual studies,
26 Following this open discussion with the scientific community, the three groups submitted
27 their respective reports to the designated U.S. Environmental Protection Agency (EPA)
28 Regional Offices (Boston, Region I; Baltimore, Region HI; and Cincinnati, Region V).
29 These reports and their associated data sets were passed on to EPA/Office of Research and
30 Development and EPA/Office of Solid Waste and Emergency Response (OSWER) for the
31 preparation of this integrated report. Although it is unlikely that major findings have been
July 15, 1993 2-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 overlooked in these first two phases, it is not at all unreasonable that more detailed
2 information will be retrieved and reported by the extended investigations made possible by
3 this open policy for data release.
4 This report presents the results and conclusions of the three studies in a manner in
5 which they can be compared, both through a broad overview of the general findings and
6 through the benefit of a detailed statistical reanalysis of the data. Because of the common
7 design of.the three studies and their focus on key experimental parameters, it is possible to
8 combine the data for certain key parameters into a single data set for meta-analysis, and the
9 results of this meta-analysis are also presented in this report.
10
11
12 2.2 PROJECT BACKGROUND
13 The Urban Soil Lead Abatement Demonstration Project, known also as the Three City
14 Study, was authorized in 1986 under Section lll(b)(6) of the Superfund Amendments and
15 Reauthorization Act (SARA). The scientific evidence for a correlation between soil lead and
16 blood lead was sufficient to warrant concern for the health of children, but not strong enough
17 to support a large scale program for soil lead abatement. Consequently, SARA (1986) called
18 for EPA to conduct a "pilot program for the removal, decontamination, or other actions with
19 respect to lead-contaminated soil in one to three different metropolitan areas."
20 To fulfill this mandate, it was necessary to design a project that would measure the
21 change in exposure from a single source (soil) amid continual changes in exposure to other
22 sources (air and food) and demographically irregular exposure from still more sources
23 (drinking water and paint). Furthermore, the range of exposure for a 6-year-old child is
24 much more diverse than for a 1-year-old child. Consequently, it was necessary to monitor
25 all sources of lead and to include the abatement of entire neighborhoods as well as single
26 residences.
27
28 2.2.1 Historical Perspective
29 In the past 25 years, concern for children with lead poisoning has steadily increased
30 with mounting evidence for the subtle but serious metabolic and developmental effects of lead
31 exposure levels previously thought to be safe. Childhood lead poisoning was formerly
July 15, 1993 2-3 DRAFT-DO NOT QUOTE OR CITE
-------
I considered a severe medical problem usually traced to swallowed chips of peeling lead-based
2 paint. Scientific evidence has systematically revealed deleterious effects of lead at lower
3 levels of exposure, and regulatory agencies such as EPA and the Centers for Disease Control
4 (CDC), have repeatedly lowered the level of concern for children's lead burden that requires
5 environmental or clinical intervention. Whereas this concern was initially directed toward
6 symptomatic children with blood lead levels of 60 /xg/dL and above, since November 1991,
7 lead poisoning has been defined by CDC as a blood lead level of 10 /Ag/dL or greater.
8 Children are exposed to lead through complex pathways from multiple sources. In the
9 mid 1980s, attention to sources of childhood lead exposure turned to urban environments
10 with high concentrations of lead in soil. Evidence for this concern came from the apparent
11 correlation between the incidence of high blood lead concentrations and high concentrations
12 of lead in residential soils. At that time, there were several other sources of exposure that
13 could potentially account for unusually high blood lead in a population of urban children.
14 Among these were lead in the air (primarily from automobile emissions), lead in food
15 (primarily from canned foods with lead soldered side seams), lead in drinking water
16 (primarily from lead pipes or newly soldered copper pipes), and lead in paint. The lead in
17 the soil was believed to be a mixture of lead from the atmosphere and lead from exterior
18 paint. Regulations were in place that would remove lead from gasoline by the end of 1986,
19 and there was a voluntary program among food processors to phase out cans with lead
20 soldered side seams and use only cans without lead solder.
21 The concept of soil abatement was not new in 1986. Many residential (usually
22 nonurban) areas were located near Superfund sites where lead was a major contaminant, and
23 the decision to abate soil was usually based in part on the distribution of blood lead within
24 the population of children. There was, however, limited experience on the effects of this
25 abatement and little or no opportunity to followup with studies of the results.
26 These complexities, and the overwhelming magnitude of the implications of this
27 project, made it necessary to plan a project that was as broad in scope as possible within the
28 resources available. One important concern was that there is a difference between a
29 population of children that is suddenly placed into a "clean" environment and one that has
30 lived continuously in a clean environment. Because of the physiology of lead mobilization in
31 body tissues, there is a difference between the rate of change in a population with increasing
July 15, 1993 2-4 DRAFT-DO NOT QUOTE OR CITE
-------
1 lead exposure and in one with decreasing exposure. In other words, the decrease in blood
2 lead anticipated in this project was not expected to be at the same rate as the increase
3 observed in other studies.
4
5 2.2.2 Site Selection
6 The site selection process began in December 1986 with the appointment of a steering
7 committee to develop recommendations for implementing the SARA lead-in-soil
8 demonstration project. A panel of experts was formed in March 1987 to set the criteria for
9 selection of sites and the minimum requirements for a study at each site. It would be
10 necessary to design three concurrent studies that were identical except for key parameters
11 such as the scope of the abatement (neighborhood versus single residence) and amount of soil
12 contamination. An early decision was that the options for soil abatement methods were
13 limited because only excavation and removal had been used hi similar programs. Therefore,
14 there would be no attempt in the project to evaluate alternative methods of abatement because
15 of limited tune and resources. The panel met again in April 1987 to discuss technical issues
16 and study designs and evaluate technical criteria for selection of urban areas as potential soil-
17 lead abatement demonstration project sites. They established the following site selection
18 criteria.
19
20 A. To be considered for selection, a metropolitan area must have:
21
22 1. Agreement by the appropriate EPA regional office to provide general project
23 oversight, and to disburse the funds.
24
25 2. An established entity, preferably the state, documented as willing to be responsible
26 . for removing and disposing of lead contaminated soil. This includes identification
27 of an appropriate facility within the' state for disposal of the soil, facilitation of
28 permits, community relations and education, and any other activities necessary to
29 expeditiously provide for disposal.
30
31 3. The administrative infrastructure to carry out a large scale project. This includes a
32 key government department with appropriate authority to coordinate the project,
33 and generally includes active participation by the state, by community groups, and
34 by all the different metropolitan departments with some responsibility for the
35 project.
36
July 15, 1993 2-5 DRAFT-DO NOT QUOTE OR CITE
-------
1 4. Access to scientific and medical expertise is needed. Scientific advice will be
2 needed to ensure that sampling and analysis are properly conducted; medical care
3 will be needed for children found to have lead toxicity,
4
5 5. Evidence that there are both children with elevated blood lead levels as defined by
6 the CDC in its childhood lead screening guidelines, and soil in residential areas to
7 be abated with lead levels of 1,500 jitg/g or greater. It would be desirable for
8 leaded paint to be established as a major contributor to the soil lead levels.
9
10 B. To be considered for selection, a metropolitan area should have:
11
12 6. A documented high incidence of children with elevated blood lead levels in the
13 proposed study areas. This means that the municipality supports an activd
14 childhood lead screening program.
15
16 7. A pattern of high density population in study areas. The number of children
17 available for evaluation as part of the project is important to the statistical validity
18 of the study.
19
20 8. Availability of other sources of funding for portions of the project not funded by
21 SARA (1986). Such items might include de-leading the outside of houses, or
22 intensive interior vacuuming to remove residual leaded dust.
23
24 The Steering Committee approved Boston as the initial study site and issued a Request
25 for Proposal (RHP) in the summer of 1987 for two additional sites. Additional proposals
26 were received from five other metropolitan areas: Baltimore, Cincinnati, Minneapolis,
27 Detroit, and East St. Louis. These were reviewed on December 3 and 4, 1987, by the
28 Steering Committee with additional experts. Baltimore and Cincinnati were selected to
29 participate with Boston in the project. The following points were the basis for this decision.
30 1. Cincinnati proposed a neighborhood level abatement study where housing units had
31 been previously gutted and rehabilitated approximately 20 years ago, and were
32 considered free of lead-based paint.
33
34 2. The Cincinnati sites contained soil lead from 220 to 900 jtg/g, exterior surface dust
35 (primarily from paved areas) from 2,000 to 5,000 jig/g, and a number of children
36 with blood lead concentrations above 25 ^ig/dL.
37
38 3. The Baltimore project proposed individual housing units with soil lead
39 concentrations in excess of 10,000 /tg/g. Lead-based paint had been abated in
40 some, but not all houses.
41
42 4. There were few children in Minneapolis with blood-lead concentrations above
43 25 /ig/dL, and most of the soil lead concentrations were below 1,000 /wg/g.
July 15, 1993 2-6 DRAFT-DO NOT QUOTE OR CITE
-------
1 5. Cincinnati demonstrated a high degree of organizational infrastructure, with
2 commitments from the City of Cincinnati and the University of Cincinnati. There
3 was an established structure for neighborhood associations that was perceived to be
4 a plus for the project.
5
6 6. The Baltimore proposal was prepared by the State of Maryland and showed a
7 satisfactory level of organizational infrastructure and local scientific expertise.
8 There were problems noted with the statistical approach.
9
10 7. Although Detroit and East St. Louis both demonstrated significant lead problems
11 through the numbers of children with elevated blood lead and the incidence of high
12 soil lead, the organizational infrastructure and local scientific expertise in Detroit
13 and East St. Louis were not perceived to be strong enough to support a project of
14 this magnitude.
15
16
17 The selection panel was unanimous in their opinion that Baltimore and Cincinnati were
18 the best choices to complement the Boston study, although they clearly recognized that
19 Detroit and East St. Louis had a significant problem with urban soil lead. The panel
20 believed that, if the study were conducted in Minneapolis where the soil lead concentrations
21 were much lower than the other cities, the results would be too subtle to provide the level of
22 statistical significance required for the study.
23 With the selection of Boston, Cincinnati, and Baltimore, a Scientific Coordinating
24 Committee, with representatives from the three studies, the three EPA Regional Offices,
25 OSWER, the Environment Criteria and Assessment Office/Research Triangle Park, NC
26 (ECAO/RTP), and the Centers for Disease Control^ was established to provide scientific and
27 technical support for the three studies and to coordinate the exchange of scientific
28 information. At an early stage, it was decided that the individual reports of the three studies,
29 although published independently, would be combined into a comprehensive report that
30 would seek to extract significant information that might not be otherwise available from the
31 individual studies. The task of organizing the Scientific Coordinating Committee and writing
32 this integrated report was assigned to ECAO/RTP. Major policy decisions remained with the
33 Steering Committee.
34 Following the selection of Boston, Baltimore, and Cincinnati as the three sites for the
35 project, the funding mechanisms were set into place individually through the respective EPA
36 Regional Offices (Regions I, HI, and V). Each of these regional offices set up an
37 independent funding mechanism and oversight plan. The regional project officer became the
July 15, 1993 2-7 DRAFT-DO NOT QUOTE OR CITE
-------
1 liaison to the Steering Committee and to the Scientific Coordinating Committee. Each city
2 submitted a work plan, which included the project description, organization, operation plan,
3 and reporting mechanisms, and the Q/A plan. These work plans required more than one
4 year to complete and pass Regional approval. In the meantime, the projects were staffed and
5 became operational. Community relations programs were initiated that began the process of
6 recruiting the study participants. Coordination between the three cities was accomplished
7 through a series of workshops, approximately three per year, that continued through the
8 preparation of the final reports.
9
10
11 2.3 PROJECT OVERVIEW
12 To place this project in perspective, it is helpful to look at the similarities and
13 differences among the three studies. They are similar in that their hypotheses and study
14 designs were drawn from the same general hypothesis, that removing lead from soil will
15 reduce lead exposure. In a general sense, each study was designed around the concept of
16 participating families within a definable neighborhood. There were one to three
17 neighborhoods in each study group, and two or three study groups in each study, For each
18 study group, there was a preabatement, abatement, and postabatement phase. This means
19 that the environment of the child was extensively evaluated prior to and after abatement,
20 through measurements of lead in soil, dust, drinking water, and paint, and through
21 questionnaires about activity patterns, eating habits, family activities, and socioeconomic
22 status. The objective of the preabatement phase was to achieve a clear understanding of the
23 exposure history and status (stability of the blood lead and environmental measures) prior to
24 abatement. During the abatement phase, attention was given to preventing any possible
25 exposure that might result from the abatement activities. During the postabatement phase,
26 the project was designed to determine the duration of the effect of soil abatement and to
27 detect possible recontamination.
28 The project objective was to measure the relationship between soil lead and blood lead.
29 This is an indirect relationship in the sense that children most commonly do not eat soil
30 directly but usually ingest small amounts of dust derived, in part, from this soil. Likewise,
31 the lead in. blood reflects not only exposure from all sources, but a host of physiological
July 15, 1993 2-8 DRAFT-DO NOT QUOTE OR CITE
-------
1 processes that include distribution and redistribution of lead to other body tissues, especially
2 bone tissue.
3 Throughout all phases, the timing of the blood lead measurements was in the context of
4 a seasonal cycle of blood lead levels that peaks in the late summer and the age-related pattern
5 that peaks at 18 to 24 mo. Because of the complex nature of this exposure assessment, it
6 was deemed advisable to measure intermediate exposure indices, such as street dust, house
7 dust, and hand dust, wherever possible. This required the development of new sampling and
8 analysis protocols that were not generally available in the scientific literature, and thus
9 became a major topic during the early coordinating workshops.
10 The studies differ in several respects. The pathways of soil to exterior dust and paint
11 to house dust differ slightly among the studies, as do the intervention strategies to interrupt
12 the flow of lead along these pathways. Collectively, these differences in study design
13 broaden the scope of the project to cover aspects of lead exposure intervention not possible
14 through the study of a single neighborhood or even a single city.
15
16 2.3.1 Project Terminology
17 The reader will more easily understand the discussions in this report if some
18 consistency is given to certain descriptive terms that are found in the reports of the individual
19 studies. These terms are described in the glossary of this document. An obvious example is
20 the use of the terms "study" and "project". In order to avoid confusion, the term "study"
21 refers to one of the three individual studies, and the term "project" is used in reference to the
22 three studies collectively.
23 The names that were used by each study to identify the treatment groups have been
24 modified in this report to assist the reader in remembering the type of intervention performed
25 on each group. Table 2-1 lists these names, with a brief description and the corresponding
26 term in the report of the individual study. The revised group names are linked to the
27 location of the study and the nature of the intervention. For example, BOS SPI refers to the
28 Boston group that received Soil, Paint, and Interior dust intervention. A hyphen is used to
29 show that the intervention was separated by a period of time, as in CINI-SE, where interior
30 dust abatement took place about 1 year before soil and exterior dust abatement. The reader
31 may want to become familiar with this nomenclature for the eight groups of participants in
July 15, 1993 2-9 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 2-1. TREATMENT GROUP NOMENCLATURE WITH
CROSS-REFERENCE TO INDIVIDUAL REPORTS
Treatment
Group Name8
Cross-Reference to
Individual Study
Report
Description of Treatment
BOS SPI
BOS PI
BOSP
BALSP
BAL P-l*
BAL P-21
ONSET
CINI-SE-1C
CIN I-SE-20
CIN NT
Study Group
Control Group A
Control Group B
Study Area
Control Area
Study Area
Area A
Area B, Back
Street and Findlay
neighborhoods
Area B, Dandridge
neighborhoods
Area C
BOSTON
Soil and interior dust abatement, and exterior and
interior paint stabilization at beginning of first year,
no further treatment
Interior dust abatement and exterior and interior
paint stabilization at beginning of first year
Exterior and interior paint stabilization at beginning
of first year
BALTIMORE
Soil abatement and exterior paint stabilization at
beginning of first year, no further treatment
Exterior paint stabilization at beginning of first
year, no further treatment
Exterior paint stabilization at beginning of first
year, soil not above cutoff lead, no further
treatment
CINCINNATI
Soil, exterior dust, and interior dust abatement at
beginning of first year, no further treatment
Interior dust abatement at beginning of first year,
soil and exterior dust abatement at beginning of
second year, no further treatment
Interior dust abatement at beginning of first year,
soil and exterior dust abatement at beginning of
second year, no further treatment
No treatment, soil and interior dust abatement at
end of study
*The treatment group designation indicates the location of the study (BOS = Boston, BAL = Baltimore,
CIN *» Cincinnati), the type of treatment (S = soil abatement, E = exterior dust abatement, I = interior dust
abatement, P = loose paint stabilization, NT = no treatment).
Treated as one group in the Baltimore report, analyzed separately in this report.
"Treated as one group in the Cincinnati report, analyzed separately in this report.
July 15, 1993
2-10
DRAFT-DO NOT QUOTE OR CITE
-------
1 the project, as the data and results will be presented using these designations without further
2 explanation. One further note: The BOS PI and BOS P groups each received soil abatement
3 at the end of the study. Because no data were reported following this intervention, the
4 designation "-S" was not used in order to avoid confusion.
5 Other departures from the terminology of the respective study reports are conversion to
6 a common system of units (metric where possible) and standard terms for phases, stages, or
7 rounds of the project. In the latter case, the term "Round" is used for the study phase, and
8 there is no consistent pattern for when abatement occurs (i.e., after Round 1, Round 3, etc.).
9
10 2.3.2 Study Groups
11 Variations in the nature and form of intervention were included in the study designs,
12 such that specific types of information could be retrieved from individual studies based on
13 the unique characteristics of the cities and their neighborhoods. For example, neighborhoods
14 in Cincinnati were selected because they were relatively free of lead-based paint, a known
15 confounder in the relationship between soil lead and blood lead. As it happened, these
16 neighborhoods were mostly multifamily housing with little or no soil on the residential parcel
17 of land. The study design used intervention on the neighborhood scale, where the soil in
18 parks, play areas, and other common grounds could be abated, and paved surfaces in the
19 neighborhood could be abated of exterior dust lead. Table 2-2 describes the study design
20 characteristic for each of the three studies and their respective neighborhood groups. The
21 general characteristics are that soil lead concentrations are typically high in Boston, where it
22 is also common to find lead in drinking water and in both exterior and interior paint. In the
23 areas studied, housing is typically single family with relatively large soil cover. In the
24 Baltimore neighborhoods, nearly every house had lead-based paint, the houses were mixed
25 single and multifamily, and the soil areas were smaller, typical less than one hundred square
26 meters.
27
28 2.3.3 Intervention Strategies
29 Intervention is defined here as the interruption of the flow of lead along an exposure
30 pathway. Soil abatement is one form of intervention. If done correctly, this abatement
31 should establish an effective and persistent barrier to the dust movement. Other forms of
July 15, 1993 2-11 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 2-2. NUMBER OF PROJECT PARTICIPANTS BY STUDY
GROUP AND ROUND3
Study Group
BOSTON
Middatc
Children*
Famlies0
Properties'5
BALTIMORE
Middatc
Children13
», ... b
Families
Properties
CINCINNATI
Mid date
Children11
Families0
Properties
BOS SPI
BOS PI
BOSP
BOS SPI
BOS PI
BOSP
BOS SPI
BOS PI
BOSP
BALSP
BALP
BALSP
BALP
BALSP
BALP
CINSEI
GIN I-SE
CIN NT
CINSEI
CIN I-SE
CIN NT
CINSEI
CIN I-SE
CIN NT
Rl
10/17/89
52
51
47
43
43
39
34
36
30
Rl
10/25/88
75
333
53
210
48
213
Rl
7/6/89
54
86
61
31
58
40
55
74
86
R3
4/9/90
52
48
46
43
40
38
34
33
29
R2
4/1/89
70
252
43
160
44
157
R3
11/14/89
52
81
52
30
56
37
39
121
85
R4
9/12/90
52
49
46
43
41
38
34
34
29
R3
2/17/90
95
174
32
125
51
105
R5
7/1/90
46
92
81
31
56
35
39
121
85
R4
1/27/91
89
111
23
93
49
65
R7
11/17/90
37
87
74
31
74
63
40
119
84
R5
6/7/91
85
111
22
88
45
63
R9
6/16/91
31
77
61
30
60
52
40
121
84
R6
9/3/91
81
106
21
88
45
63
t
"Round designations (Rl, R2, etc) are the same as used in the individual study reports. Some rounds are
omitted from this table because no participant data were collected. Intervention occurred between Rl and R3
in Boston, R3 and R4 in Baltimore, Rl and R3 in the first year of the Cincinnati study, and R5 and R7 in the
second year.
Based on number of children sampled for blood.
''Based on number of households sampled for dust.
Based on number of soil areas sampled.
My 15, 1993
2-12
DRAFT-DO NOT QUOTE OR CITE
-------
1 intervention used in this project were exterior dust abatement, interior dust abatement, and
2 paint stabilization. Because dust is a very mobile constituent of the human environment,
3 exterior and interior dust abatement would not be expected to form a permanent barrier to
4 lead unless other sources of lead, such as soil, were also abated.
5 Figure 2-1 illustrates the generalized concept of the pathway and sources of human
6 exposure to lead, showing the routes of lead from the several sources in the human
7 environment to four compartments (inhaled air, dusts, food, drinking water) immediately
8 proximal to the individual. One of these proximal sources, dust, is the primary route of
9 concern in this project. Figure 2-2 expands this dust route to show the complexity of the
10 many routes of dust exposure for the typical child.
11
Auto
Emissions
Industrial
Emissions
Crustal
Weathering
Surface and
Ground Water
Paint,
Industrial
Dusts
Solder
Lead Glazes
Drinking
Water
Figure 2-1. Generalized concept of the sources and pathways of lead exposure in
humans.
July 15, 1993
2-13
DRAFT-DO NOT QUOTE OR CITE
-------
I Atmospheric I
Particles
v.
Soli
Exterior Paint
Dust
Interior Paint
Dust
Local
Fugitive
Dust
Secondary
Occupational |
Dust
figure 2-2. Typical pathways of childhood exposure to lead in dust.
The strategies for intervention used in this project were designed to interrupt the
movement of lead along one or more of these pathways. The pattern for these interventions
for each study appears on Figures 2-3, 2-4, and 2-5.
The strategy for soil abatement was to remove all soil with concentrations above a
specific level (determined for each study), and replace this soil with clean soil below a
specified lead concentration. This method, called excavation and removal, was used in all
three studies. In some cases, repair and maintenance of ground cover was used where the
soil concentrations did not warrant excavation and removal.
To further interrupt the flow of lead along the exposure pathways, entire neighborhoods
were cleaned of exterior dust using vacuum equipment and hand tools. This intervention
method was used only in Cincinnati.
Interior dust abatement was expected to have an immediate effect on the blood lead of
children because household dust is believed to be their primary source of lead. Because
household dust is a mixture of several sources of lead, abating house dust temporarily
removes these sources, and their return and consequent impact on the child's environment
July 15, 1993 , 2-14 DRAFT-DO NOT QUOTE OR CITE
-------
"\
Atmospheric
Particles
y
/
Soil
Exterior Paint
Dust
Interior Paint!
Dust
Local ^
Fugitive
Dust y
^^-^
Exterior
Dust
Interior
Dust
f Secondary""
<«r Qccu national
I Dust y
X*. ^r
Hand
Dust
Child
- Full Abatement
= Stabilization
Figure 2-3. Pathway intervention scheme for dust exposure (Boston Soil Abatement
Study).
1 can be evaluated by careful measurements of the household dust. It is essential that
2 information about the change in lead concentration, lead loading, and dust loading be a part
3 of these measurements. Following dust abatement, there should be an immediate decrease in
4 the dust loading, with no change in the lead concentration for those groups that did not
5 receive soil, exterior dust, or paint intervention. The rate at which this dust loading returns
6 to preabatement levels reflects the rate of movement of dust from other sources into the
7 home and the general housekeeping effectiveness of the home. Both of these factors are
8 believed to have some influence on the amount of dust that a child ingests.
9 The effectiveness of both paint stabilization and soil and dust abatement can be
10 observed by changes in the lead concentrations of house dust. In the presence of lead-based
11 paint, the concentration of lead in house dust is expected to be greater than 1,500 to
12 2,000 /wg/g, whereas without the influence of lead-based paint, the house dust is expected to
13 be comparable to external dust and soil (U.S. Environmental Protection Agency, 1986).
July 15, 1993 2-15 DRAFT-DO NOT QUOTE OR CITE
-------
[ Atmospheric
I Particles
\.
Soil
Exterior Paint
Dust
Interior Paint
Dust
Local
Fugitive
Dust
Secondary
Occupational j
Dust
= Full Abatement
Stabilization
Figure 2-4. Pathway intervention scheme for dust exposure (Baltimore Soil Abatement
Study).
I House dust is a mixture of dusts from many sources within and outside the home.
2 In the absence of lead-based paint inside the home, it would seem reasonable to assume that
3 most of the lead in household dust conies from soil and other sources immediately external to
4 the home. Therefore, to enhance the impact of soil abatement, interior dust abatement was
5 carried out for some study groups in Boston and Cincinnati, but not in Baltimore.
6 Many of the Boston and Baltimore households selected for the project had chipping and
7 peeling paint, both interior and exterior. In order to reduce the impact of this paint, much of
8 which was lead-based paint, the walls and other surfaces were scraped and smoothed and
9 repainted to reduce the impact of lead-based paint on the pathways of lead exposure. It is
10 important to note that this approach in not a full scale paint abatement and was not designed
11 to place a permanent barrier between the paint and the child. Paint stabilization was used
12 on exterior and interior surfaces in Boston and on exterior surfaces in Baltimore. Paint
July 15, 1993
2-16
DRAFT-DO NOT QUOTE OR CITE
-------
Atmospheric
Particles
V J
/
Soil
Exterior Paint
Dust
Interior Paint
Dust
Local ^ \.
Fugitive >
Dust J
Exterior
Dust
KX >
>CX *"
Interior
Dust
C Secondary A
< Occupational
^ Dust ^
Full Abatement
Lead Based Paint Previously Removed
Figure 2-5. Pathway intervention scheme for dust exposure (Cincinnati Soil Abatement
Study).
1 stabilization was not used in Cincinnati because the lead-based paint had been removed from
2 these homes in the early 1970s.
3
4 2.3.4 Measurements of Exposure
5 Exposure is the amount of a substance that comes into contact with an absorbing
6 surface over a specific period of time. In the case of lead, the absorbing surface can be the
7 gastrointestinal tract or the lungs. Exposure is measured in pg/day. Thus, an exposure of
8 10 /^g/day represents a total ingestion and inhalation of 10 ^g lead from all sources; a
9 fraction of this 10 /*g would be absorbed into the body. In this project, blood lead was used
10 as an indicator of exposure, and reductions in blood lead were expected as a result of any
11 combination of the interventions described above. The units are /ig Pb/dL blood, and they
12 are not compatible with the normal units of exposure, ^g Pb/day. This illustrates that lead in
13 1 dL of blood reflects the accumulation of an unknown number of days. Other measures of
My 15, 1993
2-17
DRAFT-DO NOT QUOTE OR CITE
-------
exposure are hand lead and house dust. The amount of lead on the child's hands is believed
to be closely related to the blood lead. House dust can be an effective estimator of exposure
when then child spends most of the day playing inside, or when the house dust is composed
largely of exterior dust.
2.3.4.1 Blood Lead
The amount of ingested lead that is actually absorbed depends on the bioavailability of
the particular form of lead. The amount of absorbed lead that reaches specific body tissues
depends on the biokinetics of lead in the human body. Because there is also a relationship
between blood lead and the onset of health effects of lead, it is convenient for this measure
of blood lead to be used both as an indicator of exposure and a measure of the potential
health risk to the child. Blood tissue is in dynamic equilibrium with all other body tissues,
including bone tissue, where the lead is stored for longer periods of time. This situation
becomes important when measuring the rate at which blood lead concentrations might decline
following abatement. For a child with lead stored in bone tissue following a long history of
lead exposure, the decline in blood lead might be expected to be slower than a child without
previous exposure.
2.3.4.2 Hand Lead
Because blood lead reflects exposure to lead from all environmental sources, a second
measure, hand lead, was used to focus directly on the immediate pathway of dust into the
child. The units of measure are pg Pb/pair of hands, and like blood lead, this measure does
not reflect the rate at which lead moves to the body in the form of ^g Pb/day. Instead, this
hand dust is a measure of lead loading. It is a measure of the "dirtiness" of the hand in the
same sense that dust loading is a measure of the dirtiness of the floor, as discussed in the
next section. Hand dust loading could be converted to jtg/day if there were a measure of the
number of "hands" (or hand area) mouthed by the child during one day.
2.3.4.3 House Dust
House dust is a mixture of lead from many sources, including soil, street dust, interior
paint, and several biological sources such as insects, pets, and humans. The units of
July 15, 1993 2-18 DRAFT-DO NOT QUOTE OR CITE
-------
7 2.
1 measurement are jttg Pb/g (lead concentration), jitg Pb/m (lead loading), and mg dust/m
2 (dust loading). When expressed as /wg Pb/g, the measurement can be converted to an
3 exposure measurement by assuming a specific amount of dust ingested per day, usually about
4 100 mg/day for preschool children. Exposure to household dust then becomes /ig/day:
5
Pb Concentration X Ingestion = Exposure
(2-1)
gdust day day
6
7 To understand the importance of this measurement, compare the exposure from
8 household dust to other sources. Food, drinking water, and inhaled air normally account for
9 about 5 to 15 /ig Pb/day. If the lead concentration in household dust is 200 /^g/g and dust
10 ingestion is 0.1 g/day, the exposure is 20 jwg/day or more than the other sources combined.
11 In this project, the maximum lead concentration household dust was 107,000 jwg/g.
12 House dust can be an important measure of potential lead poisoning, and abatement
13 efforts, whether soil or paint abatement, will not be successful unless there is a reduction in
14 exposure to lead in house dust.
15
16 2.3.5 Treatment Approaches
17 2.3.5.1 Soil Abatement Approaches
18 The approach to soil abatement used in Boston was to remove the top 15 cm of soil,
19 apply a synthetic fabric, and cover with a layer of about 20 cm of clean topsoil. The new
20 soil was covered with sod and watered through dry months. Areas not resodded were
21 covered with a bark mulch. Some driveways and walkways were covered with 5 cm soil and
22 15 cm gravel or crushed bank (stone with dust). On four properties, the driveway was
23 capped with 7.5 cm asphalt without soil removal, at the owner's request. A total of
24 93 Boston properties were abated in this manner. The information on area treated and
25 volume of soil removed from these properties appears in Table 2-3. The method of
26 excavation was by small mechanical loader (Bobcat) and hand labor, for the most part.
27 Initially, six properties were abated with a large vacuum device mounted on a truck, but this
28 proved unsatisfactory due to the size and lack of maneuverability. During one extreme cold
July 15, 1993 2-19 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 2-3. SOIL ABATEMENT STATISTICS FOR THE THREE STUDIES
Number of properties*1
Surface area (in)
Volume soil removed (m3)
f\
Surface area/property (m )
Volume soil/property (m3)
Boston
36
7,198
1,212
200
33.7
Baltimore
63
4,100b
690
73
10.9
Cincinnati
171,
12,089
1,813
70.7
11
"includes only properties abated during study. Properties abated at the end of the study, where no further
sampling was reported, are not included in this analysis, but are included in the individual study reports.
In Cincinnati, a property is the location of the soil abatement, not the location of"the child^s lesidence.
Surface area not provided by Baltimore report. Calculated using Boston vol/surf ratio, which is equivalent to
an average removed depth of 17 cm.
1 spell, it was necessary to remove large blocks of frozen soil, often greater than 15 cm thick,
2 by loosening with a jackbammer.
3 In Baltimore, 63 properties in BAL SP were abated between August and November
4 1990. An additional seven properties that did not meet the requirements for abatement were
5 transferred to the control group (BAL P). TJnpaved surfaces were divided into areas on each
6 property, usually front, back, and one side; and any area with soil lead concentrations above
7 500 jtg/g was abated entirely. Soil and ground cover were removed down to 15 cm and
8 replaced to the original level with soil lead concentrations less than 50 j«g/g. These areas
9 were sodded or reseeded as appropriate. Bare areas were prepped and reseeded even if soil
10 lead concentrations did not warrant excavation. Additional information appears in Table 2-3.
11 Within each neighborhood, the Cincinnati study identified all sites with soil cover as
12 discrete study sites. The decision to abate was based on soil lead concentrations for each
13 parcel of land, and for the depth to which the lead had penetrated. Lead was measured in
14 the top 2 cm and at a depth of 13 to 15 cm. If the concentration in the top sample was
15 greater than 500 /*g/g, the soil was abated. For areas where the top concentration was less
16 than 500 jig/g but the profile (mean of top and bottom) was between 300 and 500 jug/g, the
17 soil was resodded if bare. Initially, there was an option to cultivate by rototilling, but this
18 approach was abandoned as not feasible in this study. Excavation was by front end loader,
July 15, 1993 2-20 DRAFT-DO NOT QUOTE OR CITE
-------
1 backhoe and hand tools down to 15 cm, and the replacement soil lead concentration was less
2 than 50 j«g/g. Further information can be found in Table 2-3.
3
4 2.3.5.2 Exterior Dust Abatement Approach
5 Exterior dust abatement was performed in the Cincinnati study only. The approach to
6 this abatement was to identify all types of paved surfaces where dust might collect, obtain
7 permission to sample and abate these areas and to clean them once with vacuum equipment,
8 suitable for the area, that had previously been tested and shown to remove about 95 % of the
9 available dust on the area. The groups of surfaces selected were streets, alleys, sidewalks,
10 parking lots, steps, and porches. For data analysis, these were grouped as (1) targeted (steps
11 and porches); (2) streets, sidewalks, and alleys; and (3) parking lots and other locations.
12 If it were true that soil is the only source of lead in the urban neighborhood, then
13 analysis of external dust would provide a measure of the movement of lead. In the case
14 where the soil was abated, then external dust abatement would speed up the rate at which the
15 effectiveness of this abatement could be seen on the interior dust of homes. Where the soil
16 was not the only source of lead, the recontamination of exterior dust might shed some light
17 on the movement of lead from other sources onto the hard surfaces that characterize child
18 playtime activities.
19 The exterior dust measurements in the Cincinnati study (and the interior dust
20 measurements of all three studies) were made in a manner that determined the lead
21 concentration (p,g Pb/g dust), the dust loading (mg dust/m ), and the lead loading (/ng Pb/m )
22 for the surface measured. This required that a dry vacuum sample be taken over a
2
23 prescribed area, usually 0.25 to 0.5 m . It is important to note that dust abatement is not
24 expected to cause an immediate change in the lead concentration on dust surfaces, only the
25 dust and lead loading.
26
27 2.3.5,3 Interior Dust Abatement Approaches
28 Household dust was abated in the Boston and Cincinnati studies, but not in Baltimore.
29 The BOS SPI and CIN SEI groups received Interior dust abatement at the same time as soil
30 abatement, the BOS PI received interior dust abatement without soil abatement, and the CIN
July 15, 1993 2-21 DRAFT-DO NOT QUOTE OR CITE
-------
1 I-SE received interior dust abatement in the first year followed by soil abatement in the
2 second year.
3 In Boston, interior dust abatement was performed after loose paint stabilization.
4 Families were moved offsite during interior dust abatement. Hard surfaces (floors,
5 woodwork, window wells, and some furniture) were vacuumed with a High Efficiency
6 Particle Accumulator (HEPA) vacuum, as were soft surfaces such as rugs and upholstered
7 furniture. Hard surfaces were also wiped with a wet cloth (an oil treated rag was used on
8 furniture) following vacuuming. Common entries and stairways outside the apartment were
9 not abated.
10 The Cincinnati group performed interior dust abatement after exterior dust abatement
11 and also moved the families offsite during this activity. Vacuuming of non-carpeted areas,
o
12 which was done two times, each at a prescribed rate of 1 m /min, was followed by wet
13 wiping with a detergent. They replaced one to three carpets and two items of upholstered
14 furniture per housing unit. Their previous studies had shown that these soft items could not
15 be cleaned effectively with vacuuming alone. Where carpets were left in the home, they
fj
16 were vacuum cleaned three times, each at a rate of 1 m /min.
17 Although it is clear that interior dust abatement had a positive effect on reducing the
18 lead in the child's environment in the Boston study, the influence of paint stabilization
19 (discussed below) on household dust must also be considered in conjunction with the impact
20 of soil abatement.
21
22 2.3.5.4 Loose Faint Stabilization Approaches
23 It was the intent of the project to maximize the impact of soil abatement by minimizing
24 the influence of lead-based paint in all three studies. Most homes in the Cincinnati group
25 had received paint abatement 20 years prior to the project, but in Boston and Baltimore,
26 lead-based paint occurred in nearly every home. Because full paint abatement was not within
27 the scope of this project, the alternative was to retard the rate of movement of paint from the
28 walls to household dust to the extent possible. The interior and exterior surfaces of all
29 Boston homes and the exterior surfaces of all Baltimore homes received loose paint
30 stabilization at the beginning of the project, prior to any other intervention.
July 15, 1993 2-22 DRAFT-DO NOT QUOTE OR CITE
-------
1 In Boston, loose paint stabilization consisted of removing chipping and peeling paint
2 with a HEPA vacuum and washing the surfaces with a trisodium phosphate and water
3 solution. Window wells were painted with a fresh coat of primer. The exterior painted
4 surfaces of Baltimore homes were wet scraped over the chipping and peeling surfaces,
5 followed by HEPA vacuuming. The entire surface was primed and painted with two coats of
6 latex paint.
7 Although there were subsequent measurements made of the presence of lead-based
8 paint, there were no measurements made of the movement of lead from paint to house dust
9 that would reflect the effectiveness or persistency of paint stabilization. It was believed that
10 any contamination from lead-based paint would be readily apparent in the dust samples, and
11 this appears to be the case (Chapter 3).
12
13 2.3.6 Project Activity Schedule
14 The project activity schedule, shown in Figure 2-6, illustrates the major intervention
15 and measurement activities of the individual studies and the sequence and duration of these
16 activities. The frequency and timing of sampling relative to abatement and seasonal cycles
17 are important issues in the study design. These time lines are the actual occurrence of these
18 events and they differ somewhat from the planned schedule. The original design focused on
19 sampling blood lead during the late summer, as it was known that the seasonal cycle for
20 blood lead reaches a peak during this period. Where this schedule could not be adhered to,
21 an effort was made to schedule the followup sampling to characterize this cycle and possibly
22 permit extrapolation to the summer peaks.
23
24 2.3,7 Quality Assurance/Quality Control Plan
25 Each study established a sampling and analysis plan that included rigorous quality
26 control and quality assurance (QA/QC) objectives. To achieve these objectives, protocols
27 were developed to: (1) define sampling schemes that characterize the expected exposure to
28 soil for children; (2) collect, transfer, and store samples without contamination; and
29 (3) analyze samples with the maximum degree of accuracy and precision. Several
30 intercalibration exercises were performed to guarantee that the analytical results for
July 15, 1993 2-23 DRAFT-DO NOT QUOTE OR CITE
-------
Project Activity Schedule
J F f, A M J Jl A SO N D J F
JFJ/AMJJASCri:
Figure 2-6. Project activity schedule showing the times of sampling and interviewing
(shaded bars) and soil abatement (solid bars).
1 measurements of soil, dust, handwipes, and blood were accurate and that the data would be
2 intercomparable.
3
4 2.3.7.1 Quality Assurance/Quality Control for Soils and Dusts
5 A major objective of the QA/QC program was to ensure that the three studies could
6 achieve a comparable level of expertise in the analysis of soil samples. One measure of this
7 expertise is whether the laboratories of the three studies would each get the same results
8 when analyzing the same soil sample. Two round robin calibration exercises were
9 conducted, one at the beginning and one near the end of the project. In each exercise, two
10 additional laboratories were included in order to determine some measure of comparability
11 with other studies reported in the scientific literature. All laboratories reported their results
July 15, 1993 2-24 DRAFT-DO NOT QUOTE OR CITE
-------
1 independently. In the interval between these two calibration exercises, double blind audit
2 samples were inserted into the sample stream of each study to measure the persistency of
3 analytical precision throughout the study.
4
5 Round Robin Calibration Exercise I
6 In May 1988, prior to the begraning of each study, each of the three laboratories
7 collected ten soil samples from areas similar to those that would be included in their study.
8 One of the samples from Cincinnati was a street dust sample of very high lead concentration.
9 The other 29 samples were selected from soils with lead concentrations expected to range
10 from 250 to 8,000 /ig/g. The samples were dried and sieved according to the study
11 protocols. Approximately 200 g of each sample were sent to the other two laboratories and
12 to an outside lab at Georgia Tech Research Institute. Table 2-4 shows the instrumentation
13 and method of analysis used by each laboratory. The results are presented in Chapter 3.
14 In making these analyses, each laboratory used its own internal standards for instrumental
15 calibration and shared a common set of five standards provided by Dr. Rufus Chaney at the
16 U.S. Department of Agriculture. The intercalibration exercise successfully established a
17 baseline for cross study comparison of soil and dust results.
18
TABLE 2-4. WET CHEMISTRY AND INSTRUMENTAL METHODS USED FOR
THE FIRST INTERCALIBRATION STUDY
Participating Laboratories
Methoda Boston
Hot HNO3/AAS
Cold HNO3/AAS
HotHN03/ICP
XRF X
Baltimore
X
X
Cincinnati
X
X
GTRIb USDAC
X
X
HNOg = Nitric acid; AAS = Atomic absorption spectroscopy; ICP = Inductively coupled plasma emission
spectroscopy; XRF = x-ray fluorescence,
bGTRI = Georgia Tech Research Institute.
°USDA = U.S. Department of Agriculture.
July 15, 1993 2-25 DRAFT-DO NOT QUOTE OR CITE
-------
1 Intercalibration Standards and Audits
2 The first intercalibration exercise also revealed a need for a set of common standards
3 that could be used to hitercalibrate the laboratories and to monitor the performance of each
4 laboratory throughout the project. The three studies each collected three soil samples in bulk
5 (about 30 kg) in a range thought to be high, medium, and low for their area. The samples
6 were dried, sieved, and analyzed at the EPA Environmental Monitoring Systems Laboratory
7 in Las Vegas, NV (EMSL/LV). Following homogenization, approximately fifty samples of
8 each of the soils were analyzed by laboratory scale X-ray fluorescence (XRF) at the
9 EMSL/LV laboratory. Three of the nine soils were distributed to the participating cities for
10 use as common external reference standards. The remaining six were used as double blind
11 external audits. These were aliquoted into approximately 20-g samples, distributed to the
12 QA/QC officer of each study, and inserted into the soil sample stream fully disguised as field
13 soil samples.
14 Each city appointed a QA/QC officer who was not directly involved with the analysis
15 of the soil samples, but who had access to the soil sample preparation stream on a daily
16 basis. This person mailed prelabeled soil sample containers with typical sample numbers to
17 the EMSL/LV laboratory. Soil aliquots typical for each city were placed in the sample
18 containers and returned to the QA/QC officer in lots of 20 to 30. The identification numbers
19 and soil concentration values were sent to the project QA/QC officer at ECAO/RTP. Each
20 city's QA/QC officer inserted the double blind samples into the sample stream on a random
21 basis at a frequency that would ensure about four QA/QC samples per analytical day. These
22 were occasionally placed as duplicates in the same batch to provide Information about
23 replication within the batch.
24 The preliminary acceptance range for the double blind audit samples was established
25 using only the 50 XRF analyses by the Las Vegas laboratory. As the analytical results were
26 reviewed by the study QA/QC officer, the audit sample results were sent to the project
27 QA/QC officer. If the audit samples were outside the acceptable range, the study QA/QC
28 officer was informed and could recommend either reanalysis or flagging the data for that
29 entire batch. In most cases the data were flagged, because the range for the six audit
30 samples was very narrow, having been established based on analyses by a single laboratory
31 (EMSL/LV). There was no allowance for interlaboratory variation. Final decisions on the
July 15, 1993 2-26 DRAFT-DO NOT QUOTE OR CITE
-------
1 disposition of the audit sample anomalies were deferred until the completion of the second
2 ultercalibration exercise near the end of the study, which provided a basis for determining
3 the means and ranges for these audit samples.
4
5 Round Robin Intercatibration Exercise H
6 Near the end of the project, aliquots of the nine soil and six dust audit samples used
7 during the project were redistributed to the three study laboratories for single blind analysis.
8 The analyst was aware that the samples were audit samples, but did not know their
9 concentrations. These measurements were the basis for establishing the final range of
10 acceptability for the audit samples, and for correcting the soil and dust measurements to
11 values common to the project.
12
13 2.3.7.2 Quality Control and Quality Assurance for Hand Dust
14 The collection and analysis of hand wipes is a new procedure developed just prior to
15 the beginning of the project. There were few published reports of the measurement
16 techniques, no certified standards, no internal standards, and little to base decisions on
17 acceptable analytical precision. Double blind audit samples were provided the study QA/QC
18 officer as an external control for hand wipe analysis. These were prepared as simulated
19 samples by placing a known amount of an appropriate solution of lead nitrate solution onto
20 the blank hand wipe at the EMSL/LV laboratory, wrapping and labeling according to the
21 field protocol and returning to the participating laboratory for insertion into the sample
22 scheme. There was no attempt to determine interlaboratory variance or to calculate
23 correction factors. The study QA/QC officer was responsible for reporting problems to the
24 laboratory director.
25
26 2.3.7.3 Quality Control and Quality Assurance for Blood Lead
27 The QA/QC program for blood analysis was directed by Dr. Dan Pascal of the Centers
28 for Disease Control using the protocols developed'for the CDC blood lead certification
29 program. Each laboratory received double blind bovine blood samples with one of four lead
30 concentrations. The range of acceptable measured concentrations was the same as for the
July 15, 1993 2-27 DRAFT-DO NOT QUOTE OR CITE
-------
1 certification program, and all three participating laboratories reported satisfactorily in this
2 program.
3
4 2.3.8 External Factors That Could Influence the Outcome of the Project
5 The Scientific Coordinating Panel recognized several factors that might influence the
6 outcome of the project and that were generally beyond the control of the investigators.
7 Among these are seasonal cycles and time trends of childhood blood lead concentrations,
8 unexplained or unexpected sources of lead in the homes or neighborhoods, changes in the
9 public perception of the hazards of lead exposure and awareness of exposure reduction
10 efforts, and movement of lead in soil either down the soil column or laterally as drainage or
11 fugitive dust.
12
13 2.3.8.1 Cycles and Trends in Environmental Lead Concentrations
14 Investigators have known for some time that there is a seasonal pattern to the blood
15 lead measurements taken for a population of children. Most epidemiological studies are
16 planned so that measurements can be taken at the peak of this cycle, generally during the late
17 summer. Prior studies of large numbers of children show a clear sinusoidal pattern, even
18 when the measurements do not include sequential measurements for the same child. During
19 the development of the study designs, it was apparent that an understanding of the seasonal
20 cycles and temporal trends in blood lead would play an important part in the interpretation of
21 data collected over several years. Many earlier studies had shown a clear seasonal cycle,
22 with a peak in late summer, for the blood lead concentrations of urban children. Figure 2-7
23 illustrates this pattern for Chicago during the 1970s, at the same time showing a downward
24 trend throughout the decade. The National Health Assessment and Nutrition Examination
25 Survey n (NHANES IT) data for the entire country and all age groups (Figure 2-8) show a
26 similar seasonal cycle and downward trend during the last half of that decade. Although this
27 project was designed to maximize the measurements of blood lead during the late summer for
28 eacli of the three studies, many measurements were made during other times of the year in
29 order to observe changes immediately after abatement. Consequently, there are sequential
30 measurements for a large number of children that should be adjusted for seasonal effects in
31 order to interpret the response to intervention.
July 15, 1993 2-28 DRAFT-DO NOT QUOTE OR CITE
-------
"a
1
3 30
m
| ,0
o
10
i—i—r
T 1 i T
1970 1971 1972 1973 1974 197S 1976 1977 1978 1979 1980
Year
Figure 2-7. Literature values for seasonal patterns for childhood blood lead (age 25 to
36 mo).
Source: U.S. Environmental Protection Agency (1986).
25
IT
I20
!
I
15
T3
O. 10
m
-------
1
2
3
4
5
6
Two other patterns, long-term time trends and early childhood age dependent patterns,
are applicable to this project. Little is known about age related patterns, but one study in
Cincinnati, prior to the project, showed a pattern of blood lead changes during early
childhood growth patterns (Figure 2-9).
10
I6
01 23456
Age (years)
Figure 2-9. Expected changes in blood lead during early childhood.
Source: Adapted from Hasselblad et al. (1980).
1 Long-term downward trends were documented for child blood lead concentrations
2 during the 1970s and 1980s and have been attributed to decreasing concentrations of lead in
3 food and air. There is little evidence for decreasing concentrations of lead in soil or dust,
4 but sequential measures in these media are rarely made. Data for this project were analyzed
5 for similar trends, and the results are reported in Chapter 3. The QA/QC measures
6 described above and reported in detail in Chapter 3 rule out any possibility of this trend
7 being caused by a measurement artifact such as analytical drift.
July 15, 1993
2-30
DRAFT-DO NOT QUOTE OR CITE
-------
1 There is a question whether that the seasonal cycle is caused by fluctuations in
2 exposure, as opposed to physiological processes that affect the biokinetic distribution of lead
3 within the body. Other studies attribute fluctuations in exposure to changing environmental
4 lead concentrations or changing consumption. During the late summer months, the child
5 may eat food or dust with high lead concentrations or ingest more dust during outdoor play.
6 This project was designed to observe changes in lead concentrations in soil and dust, but not
7 changes in consumption patterns. The observations made on these fluctuations and the
8 interpretation of these observations are reported in Section 3.3.5.
9
10 2.3.8.2 Unexplained and Unexpected Sources of Lead
11 Occasionally, measurements of environmental lead are higher than expected. This
12 section discusses the possibility that such phenomena can be attributed to changes in air
13 concentration alone. Because this study began after the phasedown of lead in gasoline, the
14 air concentrations of lead in these cities had decreased to less than 0,1 jig/m3 during the
15 study. The following is a theoretical calculation of the amount of lead that could be
16 transferred to soil or dust from this source alone.
17 Atmospheric deposition during the study was assumed to be typical for air
^ f\ ^
18 concentrations that averaged 0.1 /Ag/m (0.1 x 10" /^g/cm ). At a deposition rate of
19 0.2 cm/s, this would accumulate 0.6 /jg/cm -year at the soil surface. Assuming that this lead
2 3
20 would be retained in the upper 1 cm of soil surface, 1 cm of soil surface equals 1 cm of
21 soil, and the annual increment would be 0.6 /ig/cm . Because 1 cm of soil weighs about
22 2 g, the annual incremental increase in lead concentration would be 0.3 /*g Pb/g soil, an
23 insignificant amount in soils that average several hundred |tig/g. The calculation for annual
24 deposition to a surface is
25
1 x 10-7 KJ* x 0.2 EL x 3.15 x 107 _Ł_ = 0.6 ^ Pb (2-2)
cm3 s year cm2 year
26
27 For the accumulation of dust on hard surfaces, however, the same calculation indicates
28 a potentially greater influence of atmospheric lead. Converting to units of lead loading, and
29 assuming that changes occur over a shorter time period, the 0.6 /ig/cmz*year becomes
? 9 1
30 6,000 ^g/m -year, or 16 /tg/m -day. Therefore, a change of 0.1 /j,g/m in air concentration
July 15, 1993 2-31 DRAFT-DO NOT QUOTE OR CITE
-------
I could account for a change of 16 pig Pb/m2 per day in the dust lead loading of an up-facing
fj
2 surface, A change of 160 /tg/m over 10 days is in the range of the observed changes in
3 surface dust loading in this project. The impact of atmospheric deposition on exterior dust is
4 discussed in greater detail in Section 3.3.2.
5
6 2.3.8.3 Movement of Lead in Soil and Dust
7 In this study, measurements of lead in soil were believed to represent the total amount
8 of lead present in soil, including lead in the crystalline matrix of the parent rock material, in
9 the surfaces of soil particles, in soil moisture, and attached to organic matter. Changes
10 independent of intervention would be expected to occur that would increase 4ead
11 concentrations as the result of atmospheric deposition, exterior paint chipping and chalking,
12 and human activity such as household waste dumping. Decreasing lead concentrations might
13 occur as a result of leaching downward into the lower soil horizon, or by the reentrainment
14 of surface dust. The downward leaching of lead through the soil profile occurs at a very
15 slow rate. Estimates of a few millimeters per decade are generally considered the most
16 reasonable (Grant et al., 1990). The reentrainment of dust at the soil surface was considered
17 to be in equilibrium with the local environment, such that inputs would equal outputs by this
18 pathway. This is reasonable if there is no flaking or peeling paint within the neighborhood,
19 and no industrial source of fugitive dust in the vicinity of the neighborhood. A limited effort
20 was made to monitor and control the impact of lead-based paint on soil concentrations.
21 Buildings with exterior lead-based paint were stabilized by removal of the chipping and
22 peeling paint, done in a manner to avoid contaminating the soil. There were no attempts to
23 control the introduction of lead to the soil by human activity such as household waste
24 dumping.
25 In summary, the concentration of lead in soil was expected to be stable throughout the
26 study, but there are a number of reasons why localized soil lead fluctuations might occur.
27 Lead in household dust is a mixture of that brought into the house from outside and that
28 generated from within the home. Studies have shown that about 85% of the mass of dust
29 comes from outside the home and much of this is apparently brought in on the feet of
30 children and pets (Roberts et al., 1991). In the case where there are no internal sources of
31 lead, such as lead-based paint, the household dust lead concentration is largely a function of
July 15, 1993 2-32 DRAFT-DO NOT QUOTE OR CITE
-------
1 the soil concentration in the immediate vicinity of the house. The time delay is unknown,
2 but is believed to be on the order of weeks or months. Thus, changes in soil concentrations
3 would probably appear as changes in household dust concentrations within a few days, and
4 would probably reach equilibrium by 90 days.
5
6
7 2.4 REVIEW AND EVALUATION OF INDIVIDUAL STUDY REPORTS
8 2.4.1 Summary of the Boston Study
9 The Boston study retained 149 of the original 152 children enrolled, although
10 22 children moved to a new .location but were retained in the study. Children with blood
11 lead concentrations below 7 >tg/dL or above 24 /ig/dL had been excluded from the study and
12 two of the 149 children were dropped from the data analysis when they developed lead
13 poisoning, probably due to exposure to lead-based paint at another location.
14 Baseline characteristics (age, SES, soil lead, dust lead, drinking water lead, and paint
15 lead) were similar for the three study groups (BOS P, BOS SP, BOS SPI). The
16 preabatement blood lead concentration was higher for BOS P. The proportion of Hispanics
17 was higher in BOS P the BOS SP or BOS SPI, and the proportion of Blacks was lower.
18 There was a larger proportion of male children in BOS P.
19 Data were analyzed by analysis of covariance (ANCOVA), which showed a significant
20 effect of group assignment (intervention) for both the BOS SP and BOS SPI groups. These
21 results did not change with age, sex, socioeconomic status, or any other variable except race
22 and paint. When, the paint variable was added, the effect was diminished; when the race
23 variable was added, the effect became insignificant.
24 Although designed and conducted to produce rigorous results, the study has several
25 limitations. Participants were chosen to be representative of the population of urban
26 preschool children who are at risk of lead exposure by using the Boston Childhood Lead
27 Poisoning Prevention Program to identify potential participants from neighborhoods with the
28 highest rates of lead poisoning and by using as wide a range of blood lead levels as was
29 practical. Since no study subjects had blood lead levels below 7 ^g/dL or in excess of
30 24 /jg/dL at baseline, the study provides no information about the effect of lead contaminated
31 soil abatement for children with these lead levels. Similarly, a different effect might have
July 15, 1993 2-33 DRAFT-DO NOT QUOTE OR CITE
-------
1 been found for children who had a greater blood lead contribution from soil, such as in
2 communities with smelters or other stationary sources where soil lead levels are substantially
3 higher than those seen in this study, or where differences in particle size result in differences
4 in bioavailability.
5 It is possible that the intervention would have been associated with a greater reduction
6 in children's blood lead levels had they been followed for a longer period of time.
7 In addition, all children in the study were exposed to lead contaminated soil prior to
8 enrollment and so we are unable to investigate whether exposure to lead contaminated soil in
9 the first year of life is associated with higher blood lead levels. Lastly, the unit of abatement
10 was the single premises rather than clusters of premises. It is possible that the effect of lead
11 contaminated soil abatement on children's blood lead levels would have been greater had we
12 also removed lead contaminated soil from properties that surrounded Study Group children's
13 premises.
14 In conclusion, this intervention study suggests that an average 1,856 ^tg/g reduction in
15 soil lead levels results in a 0.8 to 1.6 ^g/dL reduction in the blood lead levels of urban
16 children with multiple potential sources of exposure to lead.
17 This study provides information about soil abatement as a secondary prevention
18 strategy, that is the benefit to children already exposed to lead derived, in part, from
19 contaminated soil. It can not be used to estimate the primary prevention effect of soil
20 abatement. Since children's postabatement blood lead levels reflect both recent exposure and
21 body burdens from past exposure, the benefit observed is probably less than the primary
22 prevention benefit, that is the benefit of abating lead contaminated soil before children are
23 exposed to it so as to prevent increases in blood levels and body stores.
24
25 2.4.2 Summary of the Baltimore Study
26 The Baltimore study recruited 472 children, of whom 185 completed the study.
27 Of those that completed the study, none were excluded from analysis. The recruited children
28 were from two neighborhoods, originally intended to be a study and a control group.
29 Because soil concentrations were lower than expected, some properties in the study group did
30 not receive soil abatement. The Baltimore report transferred these properties to the control
July 15, 1993 2-34 DRAFT-DO NOT QUOTE OR CITE
-------
1 group. In this report, the low soil properties in the study group are treater as a separate
2 group,
3 Because of logistical problems, there was an extended delay between recruitment and
4 soil abatement that accounted for most of the loss of the participating families from the
5 project. In their report, the Baltimore group applied several statistical models to the two
6 populations to evaluate the potential bias from loss of participating children. These analyses
7 showed the two populations remained virtually identical in demographic, biological and
8 environmental properties.
9 The Baltimore study was not designed to focus on measurements of the movement of
10 lead through the child's environment. Repeat measurements of soil were on abated
11 properties only, to confirm abatement. There were no measurements of exterior dust, no
12 interior paint stabilization, and no follow-up measurements of house dust. Rather, the study
13 design focused on changes in biological parameters, hand dust and blood lead over an
14 extended period of time.
15 Including the prestudy screening measurements of hand dust and blood lead in the
16 original cohort of participants, the Baltimore study made six rounds of biological
17 measurements that spanned twenty months. It is unusual to have a data set of this
18 composition and quality. In this integrated report, the baltimore blood lead measurements
19 were the basis for determining the key parameters in the seasonal cycle conversion factor
20 equation discussed in Section 3.3.5.1.
21 Soil was abated between the third and fourth rounds of biological measurements. The
22 mean soil decrease was 550 /jg/g. At Round 4, the blood lead concentrations were about
*
23 0.5 pg/dL lower in the study group than in the control, or 1 /jg/dL per decrease of
24 1000 /xg/g in soil, which is comparable to the response observed in the Boston study.
25 By Rounds 5 and 6, the study group blood lead concentrations had returned to their
26 preabatement levels and were in fact higher than the control group.
27 From the perspective of the Baltimore study alone, it is reasonable to conclude, as the
28 Baltimore report did, that soil abatement has no effect on children's blood lead. But from
29 the perspective of the Boston study, where a blood lead reduction of the same magnitude was
30 found to be persistent when house dust abatement was performed, and from the perspective
31 of the Cincinnati study, where blood lead concentrations were shown to rise and fall in
July 15, 1993 2-35 DRAFT-DO NOT QUOTE OR CITE
-------
I tandem with house dust concentrations, the results of the Baltimore study are consistent with
2 the observation that soil abatement, in conjunction with other environmental interventions,
3 can permanently reduce exposure to lead.
4
5 2.4.3 Summary of the Cincinnati Study
6 The Cincinnati study recruited 307 children, including 16 children born to participating
7 families during the study, and an additional 50 children who were recruited after the
8 beginning of the study. In their final report, the Cincinnati group excluded these children
9 who were recruited after the start of the study, plus 31 children who were living in
10 nonreliabilitated housing suspected of having lead-based paint, and four children (in two
11 families) who had become lead-poisoned from other causes. Thus, data for 210 children
12 were analyzed in the Cincinnati report and these same children were included in this
13 integrated report.
14 The Cincinnati study achieved effective and persistent abatement of soil on the
15 140 parcels of land scattered throughout the neighborhoods. In CEST SEI, where soil
16 abatement was performed in the first year, the arithmetic mean concentration dropped from
17 680 ftg/g down to 134 jig/g. In the two groups where soil abatement occurred in the second
18 year, GIN I-SE-1 and CINI-SE-2, the soil lead concentration dropped from 262 jig/g to
19 125 fig/g and 724 ptg/g to 233 /*g/g, respectively.
20 If soil were the only source of lead in the neighborhoods, exterior and ulterior dust
21 should have responded to the reduction in soil lead concentrations. Exterior dust lead
22 loading decreased following both soil and dust abatement, but returned to preabatement levels
23 within one year. In their report, the Cincinnati group concluded that recontamination of
24 exterior dust began soon after abatement. They observed corresponding changes in house
25 dust, hand lead, and blood lead that paralleled changes in exterior dust. Because blood lead
26 concentrations also decreased in the control area, the Cincinnati group concluded that there is
27 no evidence for the impact of soil and dust abatement on blood lead concentrations. This
28 integrated report concludes, through a more detailed structural equation analysis, that there is
29 a strong relationship between exterior dust and interior dust hi this subset of the Cincinnati
30 study where the impact of lead-based paint was minimized. From the perspective of all three
July 15, 1993 2-36 DRAFT-DO NOT QUOTE OR CITE
-------
1 studies, this means that when neighborhood and living unit sources of lead are removed,
2 exposure is reduced.
3 The central hypothesis of the Urban Soil Lead Abatement Demonstration Project is
4
5 A reduction of lead in residential soil accessible to children will
6 result in a decrease in their blood lead levels.
1
8 The formal statement of the Boston hypothesis is
9
10 A significant reduction (equal to or greater than 1,000 y,g/g) of lead
11 in soil accessible to children will result in a mean decrease of at
12 least 3 pg/dL in the blood lead levels of children living in areas with
13 multiple possible sources of lead exposure and a high incidence of
14 lead poisoning.
15
16 The Baltimore hypothesis, stated in the null form, is
17
18 A significant reduction of lead (> 1,000 fj-g/g) in residential soil
19 accessible to children will not result in a significant decrease (3 to
20 6 tig/dL) in their blood lead levels.
21
22 The Cincinnati hypothesis, separated into two parts, is
23
24 (1) A reduction of lead in residential soil accessible to children will
25 result in a decrease in their blood lead levels.
26
-27 (2) Interior dust abatement, when carried out in conjunction with exterior
28 dust and soil abatement, would result in a greater reduction in blood
29 lead than would be obtained with interior dust abatement alone, or ,
30 exterior dust and soil abatement alone.
31
32 Secondary hypotheses in the Cincinnati study are
33
34 (3) A reduction of lead in residential soil accessible to children will
35 result in a decrease in their hand lead levels.
36
37 (4) Interior dust abatement, when carried out in conjunction with exterior
38 dust and soil abatement, would result in a greater reduction in liand
39 lead than would be obtained with interior dust abatement alone, or
40 exterior dust and soil abatement alone.
41
42 The array of treatment groups differed considerably among the three studies
43 (Table 2-1). Each treatment group, however, had several features in common. All groups
44 were taken from one to three demographicaUy similar neighborhoods. All groups had some
July 15, 1993 2-37 DRAFT-DO NOT QUOTE OR CITE
-------
1 prior evidence of elevated lead exposure, usually a greater than average number of reports of
2 lead poisoning. Each group received the same pattern of treatment: baseline phase for 3 to
3 18 mo, intervention (except for controls), and followup for 12 to 24 mo.
4 In each treatment group, even the controls, there was an attempt to minimize the impact
5 of lead-based paint. In Boston, this was done by paint stabilization of both interior and
6 exterior paint. In Baltimore, only exterior paint was stabilized. Therefore, in these two
7 studies, the effects of soil abatement should be evaluated in the context of some intervention
8 for lead-based paint. In Cincinnati, most of the living units had been abated of lead-based
9 paint more than 20 years before the start of the study. Those that had not been abated were
10 measured but not treated prior to the study.
11 Another difference between the studies was the parallel intervention scheme used in
12 Boston and Baltimore, compared to the staggered scheme used in Cincinnati. In other
13 words, intervention in Boston (and Baltimore) took place at the same time for all treatment
14 groups, and the followup period was of the same duration. But in Cincinnati, the
15 intervention was delayed for one group, CINI-SE, such that followup varied between 12 and
16 24 mo.
17
18
19 2.5 SUMMARY AND CONCLUSION
20 2.5.1 Summary of Project Description
21 This project focuses on the exposure environment of the individual child. One measure
22 of short term exposure is the child's blood lead. Two other indicators of exposure are house
23 dust and hand dust. From the perspective of the child's environment, changes in the soil
24 concentration are expected to bring about changes in the house dust concentration, the hand
25 dust concentration, and the blood lead concentration. In each of the three studies, the soil
26 lead concentrations were reduced to approximately 50 j«g/g in the study area, and for most
27 children, there was a measurable reduction of blood lead, although not always statistically
28 significant. When corrected for seasonal and age related cyclic variations on blood lead, the
29 impact was even greater, and the effect was maximized when the rate of movement of dust
30 through the human environment was taken into account. That is, when street dust and house
July 15, 1993 2-38 DRAFT-DO NOT QUOTE OR CITE
-------
1 dust were also removed from the environment so that the clean soil represented the major
2 source of lead to the child's environment, the impact of abatement was the greatest.
3
4 2.5.2 Conclusions
5 This review of the study designs, analytical procedures and data quality measures has
6 shown no major flaws that would cast doubt on the conclusions of the individual reports.
7 We are now prepared to evaluate the data in a systematic, analytical manner in order to
8 answer the following question: If residential soil is abated will blood lead concentrations
9 decline?
10 To confirm or reject this soil lead/blood lead hypothesis the reader must pass stepwise
11 through the series of logical arguments described below. Although these statements seem a
12 bit pedantic, each step of the pathway from soil to blood must be scrutinized closely with
13 every detail examined and every possible relationship evaluated. In biogeochemical terms,
14 substances move from one source to another along real, definable pathways. This means that
15 if soil lead is not ingested, either directly or after passing through other sources, then blood
16 lead concentrations cannot respond to changes in soil lead concentrations.
17 The statements attached to each of these steps give a hint at the conclusions of this
18 report. Data are presented in Chapter 3 that support these statements, followed by statistical
19 inferences in Chapter 4.
20 1. There is lead in soil.
21
22 Lead was measured in soil in the range of less than 50 pig/g to more that 18,000 for
23 the combined studies. Each measurement of soil was treated as representing an
24 • equal area of soil surface for a given property or soil parcel. If a parcel of 100 m2
25 had four samples, each with 500 ^g Pb/g soil, then the upper 2 cm of soil on this
26 parcel (about 4,000,000 g) would contain 2,000,000,000 /tg or two kilograms of
27 lead.
28
29 2. Lead hi soil can move directly onto the child's hand.
30
31 Conceptually, this is difficult to measure, and there is no part of these studies that
32 would confirm this statement. Except for a pica situation, the child is not likely to
33 ingest the same fraction of soil that would be sampled with a 2 cm core. During
34 normal playtime activity, the child would probably get the part of soil that
35 corresponds roughly to playground dust, which is similar to the measurements made
36 of exterior dusts.
37
July 15, 1993 2-39 DRAFT-DO NOT QUOTE OR CITE
-------
I 3. Lead in soil can also move to other compartments of the child's environment, such
2 as exterior dust.
3
4 Evidence for this statement was shown in the Cincinnati study. When lead in soil
5 decreased through abatement, lead in exterior dust also decreased. In the
6 Cincinnati study, however, the relationship between soil and exterior dust was
7 found to be very weak, giving rise to the next question.
8
9 4. There are sources of lead other than soil that contribute to exterior dust.
10
11 Because the changes in lead in soil do not account for all of the changes in exterior
12 dust, it is reasonable to conclude from the Cincinnati study that there are other
13 sources for lead in exterior dust. In Cincinnati, the soil parcels were not on the
14 individual properties of the participating families, as was the case in Boston and
15 Baltimore. There are no measurements of exterior dust in the Boston or Baltimore
16 studies to confirm or reject the conjecture that exterior dust on the residential study
17 site is more closely linked to lead in soil.
18
19 5. Lead in exterior dust can move directly onto the child's hand.
20
21 There is no portion of these studies that directly measures this effect. Baltimore
22 reported that the lead loading on hands increased during the summer months, by
23 inference due to the increased playtime outside. During the interviews with the
24 family, questions were asked in all three studies about the activity patterns of the
25 children, including the amount of time spent outside, but none of the studies
26 attempted to determine the play activities immediately before the hand wipe sample
27 was taken.
28
29 6. Lead in exterior dust can also move into other components of the child's
30 environment, such as interior dust.
31
32 In the Cincinnati study, when exterior dust lead concentrations changed, interior
33 dust lead concentrations also changed. This was especially obvious when the
34 exterior dust sample closest to the residence was compared to the interior floor dust
35 sample taken just inside the entryway door.
36
37 A living unit with 130 m2 of floor space (1,400 ft2) and 1,000 /tg Pb/m2
38 (a relatively high value from tables in Section 3.3) would have 130,000 /*g of lead,
39 or less than 1 % of the lead available from soil in paragraph 1 above. Additional
40 lead would be in rugs and upholstered furniture.
41
42 7, There are sources of lead other than exterior dust that contribute to interior dust.
43
44 Taken individually, none of the studies decisively demonstrated this effect. The
45 most obvious source of lead inside the home is lead-based paint, which was
46 common in the Boston and Baltimore studies but excluded from the Cincinnati
47 study. Because neither Boston nor Baltimore measured exterior dust, measurements
July 15, 1993 2-40 DRAFT-DO NOT QUOTE OR CITE
-------
1 of interior dust in these studies cannot easily be broken down into contributions
2 from lead-based paint and from exterior dust. However, structural equation
3 analyses on the Boston study showed a string influence of both interior and exterior
4 lead-based paint on interior dust.
5
6 8. Lead in interior dust can move directly onto the child's hand.
7
8 In most cased, when interior dust changed, hand dust changed. Because hand dust
9 lead is only a measure of the amount of lead on the hand, not the concentration nor
10 the amount of dust, it is difficult to make a quantitative estimate of this pathway.
11 It is not likely that the amount of dust on the hand is strictly a function of the
12 amount of dust on the playing surface, as there is probably an equilibrium effect
13 where some dust falls off after time. There is no aspect of these studies that could
14 measure this interesting problem.
15
16 9. Lead in interior dust can also move into other components of the child's
17 environment, such as food,
18
19 This pathway was not investigated by any of the three studies. Measurements of
20 lead in food before and after kitchen preparation would be required. Conceptually,
21 this lead and other routes such as the direct mouthing activities on toys, furniture,
22 and window sills is included in the measurement of interior dust when the
23 assumption is made that a child ingests about 100 mg of dust per day by all routes
24 and through all activity patterns.
25
26 10. There are sources of lead other than dust that contribute to the child's lead
27 exposure.
28
29 In this project, lead was measured in drinking water once or twice during each
30 study. Ambient levels of lead in air were assumed, as were national averages of
31 lead in food. Ethnic food preferences and individual use of cosmetics or other lead
32 containing products were not investigated,
33
July 15, 1993 2-41 DRAFT-DO NOT QUOTE OR CITE
-------
i 3. PROJECT RESULTS
2
3
4 3.1 DATA QUALITY
5 The participating cities recognized the need for standardizing the sampling and
6 analytical protocols so that data from each study could be compared at the end of the project,
7 This was ultimately accomplished for soil and dust by measuring the analytical difference
8 between each the three labs. But before this was possible, common standards needed to be
9 prepared and a program for assuring data quality had to be put into place. A three step
10 program was agreed to that involved: (1) a round robin calibration study of soil samples to
11 measure differences between laboratories and differences between analytical methods and
12 instrumentation, (2) a double blind audit system for soil and dust to monitor the performance
13 of each laboratory during the project, and (3) a second round robin calibration study to
14 determine the arithmetic correction factor that would allow the conversion of dust and soil
15 data to a common project basis. This program would ensure that analyses performed by each
16 of the three participating laboratories would be internally accurate and externally consistent
17 with similar analyses by other research laboratories.
18
19 3.1.1 Round Robin I: Common Standards and Analytical Methods
20 At the beginning of this project, the proposed methods for soil and dust analysis were
21 reviewed by the Scientific Coordinating Panel. The preferred method, hot nitric acid
22 digestion followed by atomic absorption spectroscopy (AAS), was time consuming and
23 expensive). The number of samples was expected to exceed 75,000 per project, so more
24 rapid and less expensive methods were evaluated. Laboratory scale X-ray fluorescence
25 (XRF) spectroscopy and inductively coupled plasma (ICP) emission spectroscopy were
26 proposed, and a cold nitric acid extraction method was also considered. The first round
27 robin intercalibration study was organized, as described in Chapter 2. In summary, the test
28 conditions were that each laboratory would be provided with instructions for preparing the
29 samples (drying, sieving, and chemical extraction) but would use their own internal standards
30 and instrumental settings. They would have access to a set of external standards (from U.S.
31 Department of Agriculture) with known values from which they could make corrections if necessary,
July 15, 1993 3-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 Each of the three study laboratories sent aliquots of 10 samples to the other two
2 participating laboratories and to two external laboratories. One of the samples from
3 Cincinnati was a street dust sample with a lead concentration in excess of 15,000 jug/g, The
4 other 29 samples were soils. The samples were subdivided by sieving during preparation to
5 a "total" and "fine" fraction. Thus there were 30 samples, each with two size factions
6 analyzed by each of five laboratories using either one or two analytical methods. The
7 analytical and wet chemistry methods used by are shown in Table 3-1, and the results of the
8 analyses appear in Table 3-2.
9
TABLE 3-1. WET CHEMISTRY AND INSTRUMENTAL METHODS USED FOR
THE FIRST INTERCALIBRATION STUDY
Methodfl
Hot HN03/AAS
Cold HNO3/AAS
Hot HNO3/ICP
XRF
Participating
Laboratories
Boston Baltimore Cincinnati GTRIb
X
X
X
X
X
X
USDAC
X
aHN03 = Nitric acid; AAS = Atomic absorption spectroscopy; ICP = Inductively coupled plasma emission
speelroscopy; XRF = X-ray fluorescence.
bGTRI = Georgia Tech Research. Institute.
"TJSDA = U.S. Department of Agriculture.
1 The cold nitric acid extraction method was found to be essentially equivalent to the hot
2 nitric acid extraction method for soils with lead concentrations up to 8,000 ,ug/g (Figure 3-1)
3 for the samples analyzed in this study. The AAS method used by Cincinnati and Baltimore
4 was also equivalent (Figure 3-2), showing a high degree of comparability between these two
5 laboratories under these test conditions.
6 The interlaboratory comparison of X-ray fluorescence (XRF) between the Boston and
7 Georgia Tech Research Institute (GTRI) Laboratories showed the method was acceptable,
8 although not fully linear above 5,000 pg/g. There were no soil standards available above
9 2,000 jig/g so the analysts had some difficulty calibrating their XRF instruments above this
10 level. The data of Figure 3-3 suggest a systematic difference between the two laboratories
July 15, 1993 3-2 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 3-2. ANALYTICAL RESULTS OF THE FIRST
INTERCALIBRATION STUDY: LEAD CONCENTRATION (jtg/g)
IN THE TOTAL AND FRACTIONS OF 10 SOILS FROM EACH STUDY
Sample
Fraction0
IT
2T
3T
4T
5T
6T
7T
8T
9T
10T
11T
12T
1ST
14T
15T
16T
17T
1ST
19T
20T
21T
22T
23T
24T
26T
27T
28T
29T
30T
IF
2F
3F
4F
5F
6F
7F
8F
9F
10F
IIP
12F
13F
14F
15F
16F
17F
18F
19F
20F
21F
22F
Boston
XRF
1,200
1,750
400
550
1,100
1,450
1,000
500
550
1,450
250
800
100
700
550
220
220
75
50
4,800
500
950
1,700
2,400
2,800
3,800
5,200
4,000
6,500
1,500
2,650
500
1,600
1,700
2,400
1,200
600
650
2,200
220
1,800
100
800
620
300
100
100
50
5,100
550
1,100
Baltimore
Hot HN03
AAS
1,418
2,893
492
619
1,058
2,323
1,359
683
608
1,649
484
1,069
2,200
1,754
264
126
106
9
15,792
496
850
1,559
2,260
2,484
3,846
5,092
5,097
7,995
1,545
3,540
625
1,814
1,793
3,137
1,344
723
686
2,398
356
2,707
96
100
796
3,200
118
142
7,866
606
1,118
Hot HN03
ICP
1,324
2,544
389
462
882
1,955
1,098
535
485
1,330
365
878
53
1,701
1,410
200
62
48
7
12,030
372
698
1,298
1,880
2,119
3,440
4,667
4,510
6,560
1,421
2,921
507
1,554
1,475
2,387
1,105
598
558
1,946
244
2,220
68
779
616
236
73
85
10
6,000
506
916
Cincinnati
HotHN03 ColdHNO3
AAS AAS
1,552
2,868
387
423
964
1,876
1,383
491
455
1,679
316
1,850
63
2,068
747
253
59
74
2
14,593
387
837
1,567
2,284
2,754
4,337
5,454
5,586
8,467
1,560
3,335
478
1,678
1,689
2,835
1,306
595
593
1,808
267
2,683
68
926
635
237
73
91
3
8,109
480
1,069
1,215
2,211
466
415
854
1,722
990
725
417
1,228
348
1,103
45
1,713
785
295
58
61
3
8,147
378
739
1,368
2,003
2,401
3,835
4,747
4,700
7,502
1,404
3,127
508
1,595
1,971
2,009
1,184
298
601
1,116
. 277
2,683
64
818
642
239
66
87
2
7,432
467
944
GTRf
XRF
1,174
1,912
400
500
980
1,524
651
400
261
1,660
180
900
100
652
505
187
30
100
20
4,817
383
111
1,390
2,021
2,331
3,500
4,460
3,280
4,704
1,223
2,263
440
1234
1,290
2,134
815
490
375
1,980
180
1,680
100
693
600
236
100
100
30
4,780
505
980
USDA"
Cold HNO3
AAS
1,338
2,695
417
464
988
1,808
1,473
726
605
1,764
304
1,944
73
1,710
825
286
83
111
13
14,733
1,120
1,761
2,561
2,472
4,983
3,184
6,473
10,042
1,569
3,273
515
1,824
1,683
2,682
1,297
672
630
280
2,610
89
895
664
242
80
92
20
8,451
470
904
July 15, 1993
3-3
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 3-2 (cont'd). ANALYTICAL RESULTS OF THE FIRST
INTERCALIBRATION STUDY: LEAD CONCENTRATION (/ig/g)
IN THE TOTAL AND MNE FRACTIONS OF 10 SOILS FROM EACH STUDY
Sample
Fraction
23F
24F
25F
26F
27F
28F
29F
30F
Boston
3CRF
1,700
2,200
2,200
2,800
4,000
3,100
4,500
8,000
Baltimore
Hot HN03
AAS
1,679
2,331
2,372
2,899
4,833
3,087
5,896
8,555
HotHN03
TCP
1,424
2,014
2,000
2,402
3,969
2,616
4,717
7,443
Cincinnati
HotHNOg ColdHNOg
AAS AAS
1,710
2,328
1,665
2,946
4,531
3,073
5,606
8,679
1,431
2,010
2,089
2,568
4,130
2,720
4,869
7,789
GTRIa
XRF
1,320
1,940
2,005
2,249
3,739
2,445
4,240
6,015
USDAb
ColdHNOg
AAS
1,640
2,492
3,156
4,979
6,194
6,680
9,754
* Georgia Tech Research Institute.
"USDA = U.S. Department of Agriculture.
"T *» Total fraction, F = Fine fraction.
Cincinnati Hot HNO3 (pg/g)
Thousands
Figure 3-1. Comparison of uncorrected data for two wet chemistry methods of soil
analysis showing the comparability of hot and cold nitric acid for the
Cincinnati laboratory. The straight line indicates a slope of 1.
July 15, 1993
3-4
DRAFT-DO NOT QUOTE OR CITE
-------
5 10
Cincinnati AAS Hot HNOa
Thousands
15
20
Figure 3-2. Comparison of unconnected data for atomic absorption spectroscopic
analysis by two laboratories (Baltimore and Cincinnati) using the hot nitric
acid method of soil analysis. The straight line indicates a slope of 1.
DC
45
Boston XRF
Thousands
Figure 3-3. Interlaboratory comparison of unconnected data for the X-ray fluorescence
method of soil analysis showing the comparability of the Boston and
Georgia Institute of Technology laboratories. The straight line indicates a
slope of I.
July 15, 1993
3-5
DRAFT-DO NOT QUOTE OR CITE
-------
1 that could be corrected with a more uniform calibration. Both interlaboratory (Cincinnati
2 and Baltimore in Figure 3-4) and intralaboratory (Baltimore in Figure 3-5) comparisons of
3 AAS versus ICP demonstrated equivalency between these two instrumental methods. These
4 comparisons showed that there is likewise a systematic difference that can be corrected
5 arithmetically.
6 Finally, the interlaboratory comparison of XRF versus AAS (Boston and Cincinnati in
7 Figure 3-6, and Boston and Baltimore hi Figure 3-7) led to the conclusion that of suitable
8 soil standards at higher concentration could be made available, XRF is an acceptable
9 alternative method to AAS for soil analysis.
10 Based on this study and the awareness that chemical extraction of 75,000 soil samples
11 presented a costly burden on the project both in terms of time and expense, and the value in
12 that nondestructive analysis would preserve the samples for reanalysis, the Scientific
13 Coordinating Panel recommended the use of XRF for soil analysis on the condition that a
14 suitable set of common standards could be prepared for a broader concentration range and
15 that a rigorous audit program be established to ensure continued analytical accuracy.
16 The Round Robin I calibration exercise also raised the need for a broader scale
17 calibration exercise to determine the arithmetic correction factor for converting the data to a
18 common basis. For routine analyses, two groups, Boston and Baltimore, elected to use XRF
19 for dust analysis also, whereas Cincinnati opted for hot nitric extraction with AAS. During
20 the study, Baltimore recognized problems with analyzing dust by XRF when the sample size
21 was small, less than 100 mg. They reanalyzed the dust samples by AAS and reported both
22 measurements. In Boston, this problem was solved by compositing the floor dust samples
23 for XRF analysis, reporting one floor dust sample per housing unit.
24
25 3.1.2 Double Blind Audit Program for Soil and Dust
26 The procedures for the audit program are discussed in Section 2.3.6.1. The results of
27 that program are given in Table 3-3 based on the find biweight distributions in Table 3-4.
28 The preliminary biweight distributions, shown also in Table 3-4, contained no measure of
29 interlaboratory variability because the preliminary analyses were performed by only the
30 EMSL-LV laboratory. Thus they could only be used in the audit program for identifying and
July 15, 1993 3-6 DRAFT-DO NOT QUOTE OR CITE
-------
10 15
Cincinnati AAS (ng/g)
Thousands
20
Figure 3-4. Intel-laboratory comparison of uncorrected data for soil analysis showing
the comparability of inductively coupled plasma emission spectroscopy and
atomic absorption spectroscopy for the Baltimore and Cincinnati
laboratories. The straight line indicates a slope of 1.
20
I „
o"
Z w
il
.•_• C0
Ł1
10-
O
s
co
CO
0
10
Baltimore AAS Hot HNO3
Thousands
15
20
Figure 3-5. Comparison of uncorrected data for soil analysis showing the comparability
of inductively coupled plasma emission spectroscopy and atomic absorption
spectroscopy within the Baltimore laboratory. The straight line indicates a
slope of 1.
July 15, 1993
3-7
DRAFT-DO NOT QUOTE OR CITE
-------
^ 15
I"
0
5 10
Cincinnati Hot HNQ
15
.3 MS (ug/g)
Thousands
20
Figure 3-6. Interlaboratory comparison of uncorrected data for soU analysis showing
the comparability of X-ray fluorescence and atomic absorption spectroscopy
for the Cincinnati and Boston laboratories. The straight line indicates a
slope of I.
Baltimore Hot HNO, AAS
o
Thousands
Figure 3-7. Interlaboratory comparison of uncorrected data for soil analysis showing
the comparability of X-ray fluorescence and atomic absorption spectroscopy
for the Baltimore and Boston laboratories. The straight line indicates a
slope of 1.
July 15, 1993
3-8
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 3-3. SOIL AND DUST AUDIT PROGRAM RESULTS
Study/ Audit Sample
BOSTON DUST
BAL03
CIN01
CIN02
BOSTON SOIL
BOS M
BALH
CINL
CINH
BALTIMORE DUST
BAL 02
CIN01
BOS 01
BALTIMORE SOIL
BOS M
BALH
CINL
CINH
Study
CINCINNATI DUST
BAL 03
BOS 01
CIN01
CIN02
CINCINNATI SOIL
BOS M
BALH
CIN L
CINH
Number of
Samples
N/Aa
N/A
N/A .
N/A
N/A
N/A
N/A
(XRF)
8
10
10
15
15
15
15
Number of
Samples
(AAS)
34
35
38
26
32
49
130
31
Mean
1,232
2,671
331
6,786
1,044
399
14,074
218
3,280
14,444
5,046
838
286
11,290
Mean
1,727
24,104
2,683
259
6,654
1,016
301
14,890
Range
980-1,441
2,075-3,228
115-461
6,015-7,549
747-1,244
207-570
11,407-16,592
159-281
800-3,660
14,080-14,920
4,800-4,200
433-916
266-307
10,100-12,500
STDDEV
275
2,337
225
44
268
40
11
635
Percent Within
Final Biweight
Distribution
92
100
65
100
73
61
50
100
90
100
100
60
100
53
Percent Within
Final Biweight
Distribution
N/A
N/A
100
100
100
100
100
N/A
aN/A = Not available.
July 15, 1993
3-9
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 3-4. PRELIMINARY AND FINAL BIWEIGHT DISTRIBUTIONS FOR SOIL
AND DUST AUDIT PROGRAM
... Preliminary Values
Clranln A tiHit *
Final Values
Type Sample Mean Low High Mean Low High
1
2
3
4
5
6
7
Dust BAL01 78 58 99
Dust BAL02 331 288 374
Dust BAL03 1,480 1,346 1,613
Dust CIN01 2,851 2,660 3,042
Dust CIN02 252 216 288
Soil BOSL 3,131 2,858 3,405
Soil BOSM 6,090 5,748 6,431
Soil BOSH 14,483 13,071 15,895
Soil BALL 639 555 724
Soil BALH 923 850 997
Soil CINL 303 284 322
SoU CINH 13,585 12,872 14,297
Soil REF5
Soil REF6
Soil REF7
Soil REF8
Soil REF9
Soil REF10
84 4 16.3
309 138 480
1,438 1,091 1,786
2,617 1,422 3,812.
233 93 372
3,101 2,283 3,919
6,219 4,742, 7,696
13,369 11,980 14,754
626 468 783:
1,017 847 1,187
315 204 426
12,729 11,361 14,096
413 258 568
936 738 1,134
1,042 758 1,326
2,354 1,950 2,759
3,913 2,943 4,888
735 615 854
flagging batches of soil samples that might need to be reanalyzed pending the determination
of the final biweight distributions.
As the audit program progressed, two patterns
were systematically low or high, and this was not of
could be resolved by a more detailed intercalibration
emerged. In some cases, laboratories
major concern, as these discrepancies
exercise and arithmetic correction. The
Cincinnati group elected to make a midcourse change in instrumental parameters that reduced
this difference, as described in the Cincinnati report.
In other cases, the measured audit
July 15, 1993
3-10
DRAFT-DO NOT QUOTE OR CITE
-------
1 sample was sporadically high or low, in which case the laboratories investigated the problem
2 and resolved it. Most of these discrepancies occurred for dust samples where the sample size
3 for XKF analysis was below 200 mg. Here, the Baltimore group elected to analyze by AAS
4 and the Boston group composited the floor samples from several rooms in a single residence
5 to obtain a larger sample size. The Boston group also found, but did not report in detail,
6 that a calibration curve for XRF analysis using standards that were also less than 200 mg
7 would provide a suitable correction to the original data.
8 Batch analyses with audit discrepancies that could not be resolved by one of these two
9 means were retained in the data set but flagged. This was necessary during the study
10 because the final values for the upper and lower acceptance limits could not be determined
11 until the end of the study.
12
13 3.1.3 Round Robin II: Biweight Distribution and Final Interlaboratory
14 Calibration
15 The nine soil and five dust samples that were used for external standards and audit
16 samples were reanalyzed in a more detailed round robin exercise near the end of the project,
17 as described in Section 2.3.6.1. The purpose of this exercise was to determine the correction
18 factor for mathematically converting the soil and dust data from each study to a common
19 basis and to revise the biweight distribution values for the audit samples to reflect the
20 multilaboratory variance and systematic differences between laboratories. Additional
21 analyses by AAS were performed by Baltimore and Cincinnati for soil and dust, even though
22 only dust was analyzed by AAS during the study. Boston and Las Vegas analyzed the
23 samples by ICP for the purposes of obtaining a broader perspective on the application of this
24 method. The data from this exercise are in Table 3-5. They are the basis for determining
25. the consensus values and correction factors that appear in Table 3-6.
26 The data evaluation subcommittee of the Scientific Coordinating Panel was appointed to
27 determine the consensus values and methods of statistical interpretation of the intercalibration
28 results. Several methods were discussed in great detail. Tests were made for outliers using
29 the method of Barnett and Lewis (1984), and none were found. The data were of good
30 quality and were highly linear. The r2 values ranged from 0.997 to 0.999 using a consensus
31 based on the simple arithmetic means of the reported values. The subcommittee chose to
July 15, 1993 3-11 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 3-5. RESULTS OF THE FINAL INTERCALEBRATION STUDY
1
2
3
4
5
6
7
Sample
DUST1
DUST2
DUSTS
DUST4
DUSTS
SOIL1
SOIL2
SOIL3
SOIL4
SOILS
S01L6
SOIL?
SOILS
SOIL9
SOILIO
SOIL11
SOIL12
SOILI3
SOIL14
SOILI5
BOSK
120
320
1,430
2,000
280
450
900
1,050
2,200
3,800
710
650
950
2,800
5,600
12,500
310
12,000
810
1,450
explore alternatives
BOSX
510
910
1,100
2,300
4,000
770
930
930
2,900
5,300
13,000
290
12,000
850
1,600
XRF
BAL
121
482
1,686
3,771
267
388
808
961
2,100
3,486
640
559
896
2,514
5,200
11,000
283
10,500
793
1,400
to the arithmetic
AAS
CIN
92
329
1,307
2,924
233
441
1,033
1,080
2,555
4,227
789
675
1,036
3,126
6,493
15,963
305
14,156
929
1,705
LV
78
288
1,288
2,456
212
310
833
923
2,264
3,974
611
532
798
2,972
5,956
15,984
286
13,530
763
1,509
BAL
15
201
1,363
2,335
150
383
1,001
1,100
2,468
4,044
741
567
1,032
3,401
6,861
13,175
321
13,000
875
1,731
mean and eventually settled
weighted for within-laboratory variance. The
Version 3
Table 3-6.
regression
CIN
66
236
1,581
2,451
273
452
1,013
1,120
2,502
4,251
798-
650
1T067
3,263
6,937
13,955
379
13,195
986
1,766
ICP
BOS LV
1
2
2
3
3
5
12
13
1
94
284
,428
,109
244
401
850
972
,230
,748
699
597
944
,148
.932
,6-52,
300
,167
907
,631
on a multiplicative
model was run with GLIM
.77, Update 2, and gave consensus values and
. Although great
care was
taken to
, the consensus values produced by
correction factors
evaluate several alternatives
the GLIM
72
307
1,346
2,296
191
379
912
1,006
2,286
3,843
660
626
998
3;,158
6,360
12,608
294
11,440
900
1,650
model
statistical software,
shown in
to simple
procedure differed only slightly
from those of a simple linear regression. The correction factors
the three studies to
convert
their soil
and dust
data to a common
on Table
3-6 were
project basis.
used by
A plot of the
July 15, 1993 3-12 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 3-6. CONSENSUS VALUES AND CORRECTION FACTORS FROM
THE FINAL INTERCALIBRATION PROGRAM
XRF
AAS
Interlaboratory Consensus Values
Sample
DUST1
DUST2
DUSTS
DUST4
DUSTS
92.8
342.7
1,319.0
, 2,943.4
228.3
54.2
221.9
1,492.2
2,378.1
23.2.4
ICP
for Dust
81.7
283.4
1,362.3
2,133.4
206.2
Interlaboratory Correction Factors
Study
BOS
BAL
CIN
Sample
SOIL1
SOIL2
SOILS
SOIL4
SOILS
SOIL6
SOIL?
SOILS
SOIL9
SOIL10
SOIL11
SOIL12
SOIL13
SOIL14
SOIL15
Study
BOS
BAL
CIN
1.1527
0.7803
1.0074
Interlaboratory Consensus
460.2
960.7
1,140.5
2,493.5
4,139.3
761.0
664.1
1,062.3
2,987.8
6,175.2
13,120.7
335.3
12,498.5
941.3
1,663.2
Interlaboratory Correction
1.0370
1.1909
0.8698
1.0416
0.9616
Values for Soil
430.5
1,002.1
1,106.2
2,474.2
4,164.1
776.9
623.3
1,049.4
3,272.6
6,863.2
13,645.4
361.5
13,041.6
949.5
1,744.1
Factors for Soil
1.0166
0.9839
1.0707
426.6
909.6
1,018.8
2,342.1
3,706.1
736.1
656.0
1,005.4
3,274.9
6,411.5
13,224.7
323.6
13,080.0
923.3
1,716.8
1.0166
July 15, 1993
3-13
DRAFT-DO NOT QUOTE OR CITE
-------
1 dust (Figure 3-8) and soil (Figure 3-9) reported values versus the consensus means derived
2 from the GLIM analysis illustrates the reliability of this method.
3
4 3.1.4 Disposition of Audit Data
5 Based on the results of the second intercalibration exercise, a consensus value was
6 determined for each dust and soil sample, biweight distributions were determined for those
7 that had been used in the audit program. This new distribution incorporated interlaboratory
8 variation. When the correction factor is applied to the reported audit samples results, the
9 revised number should lie between the upper and lower boundaries of the biweight
10 distribution. Table 3-3 lists the number and percentage of these audit sample values that fell
11 within these new boundaries. Most of the discrepancies were resolved by the corrective
12 measured taken by the laboratories as described in Section 3.1.2.
13 When the audit sample values fell outside the boundaries of the final biweight
14 distribution the batches were flagged but not rejected. This decision was made for two
15 reasons. The quality of soil and dust analysis in this project was a step or two above the .
16 acceptable standards for research studies involving soil and dust analyses. Furthermore, to
17 attempt to raise this level one more step would have been costly and would have produce
18 little more in terms of scientific information. The simple fact is that if these data had been
19 rejected because of discrepancies with the audit samples, then all previous and subsequent
20 research studies in which a double blind audit program was not used would also have to be
21 rejected. By flagging the batches of data associated with these outlying audit samples the
22 groups and other users of the data could attempt to determine, on a one-to-one basis, the best
23 explanation for the apparent discrepancy. The options could be to exclude these data from
24 their statistical analysis, reanalyze the samples, or use the original data based on the
25 assumption that the data are correct.
26
27 3.1.5 Database Quality
28 Each study maintained rigorous standards for database quality. These included double
29 entry, 100% visual confirmation, and standard procedures for detecting outliers. In spite of
30 these measures, some errors were found, confirmed, and corrected prior to use in this report.
31 None of these errors would have impacted the conclusions drawn by the individual study.
July 15, 1993 3-14 DRAFT-DO NOT QUOTE OR CITE
-------
3000
*•— •>,
•g
,5 2500!
8
o 2000
3
.8
.3 1500
"§ 1000
•c
o
I" 500
500 1000 1500 2000 2500
Consensus XRF (ng/g)
3000
D Boston o Baltimore > Cincinnati x EMSL-LV
Figure 3-8. Departures from consensus dust values for each of the three studies.
I
20
15
10
T3
i
O
Q.
Cincinnati x EMSL-LV
Figure 3-9. Departures from consensus soil values for each of the three studies.
July 15, 1993
3-15 DRAFT-DO NOT QUOTE OR CITE
-------
1 3.2 OVERVIEW OF PROJECT DATA
2 3.2.1 Description of the Data
3 3.2.1.1 Types of Data
4 The analytical data used in this report consist of measurements of soil, exterior dust
5 (sometimes referred to as street dust), interior dust (house dust), hand dust, blood lead,
6 exterior paint, interior paint, and drinking water. The age and sex of the child and the date
7 of collection are also included in these analyses.
8
9 3.2.1.2 Data Collection Patterns
10 Each study produced the same or similar information about the occurrence of lead in
11 the environment. The data sets among the studies are not identical, however, in that they
12 differed in the timing of the collection relative to intervention, the spatial distribution of the
13 sampling points relative to the expected exposure to the child, and the manner in which the
14 data were reduced to a central tendency.
15 Data were collected, in rounds. That is, during a specific period of time, samples were
16 taken of soil, dust, etc., for a specific objective, such as establishing the concentration of
17 lead prior to intervention. Usually a round lasted for several weeks, perhaps 3 to 4 mo.
18 Rounds are not contiguous, in that there were gaps during which no samples were taken.
19 It is easy to see that it may be important to know when a sample was taken during a round,
20 especially following intervention, in order to evaluate the impact on exposure. Consider the
21 pathway from soil =* street dust => house dust => hand lead =* blood lead. One would expect,
22 if soil alone (not house dust) were abated and the exposure were mainly through house dust,
23 there would be a lag between abatement and response, and the impact of intervention might
24 become greater with increasing time. Conversely, the impact of intervention might be
25 reduced with time if there were recontamination, as would be expected if house dust were
26 abated but soil were not.
27
28 3.2.1.3 Data Linkages
29 Data linkages are important to the interpretation of the results. Soil data alone would
30 show only the average lead concentration at various stages of each study. Soil data by study
31 group would show apparent differences between groups, which by statistical inference might
July 15, 1993 3-16 DRAFT-DO NOT QUOTE OR CITE
-------
1 indicate the impact of intervention. When the soil data are linked to external dust or house
2 dust, the impact of soil on dust, with or without intervention, becomes discemable. Through
3 these data linkages, it is ultimately possible to construct a crude exposure scenario for the
4 individual child. These scenarios begin with the simplest case, but can become complex in
5 short order. For example, a young child may spent most of the time indoors, whereupon the
6 exposure scenario becomes the lead that is available to the child through food, drinking
7 water, air, and dust (see Figure 2-1). Each of these is influenced by one or more other
8 sources of lead.
9 Most of the rest of this chapter consists of a discussion of specific data linkages that are
10 arithmetic means of a specific compartment, such as soil, dust, hand dust, or blood lead,
11 within study groups The reader is encouraged to become familiar with these graphical
12 presentations of the data, because they form the basis of the statistical inferences of
13 Chapter 4, and the conclusions of Chapter 5.
14 Data are also linked by a primary identifier or index. Some data are linked to the
15 individual child, such as blood lead and hand lead. Some are specific for the living unit or
16 family, and some are specific for the property. In Cincinnati, soil and exterior dust data are
17 linked at the neighborhood level. It is important to be aware of this distinction because of
18 the duplication effect that can occur when there are several siblings in a family and several
19 families in a dwelling. This means that a single soil value could be heavily weighted if there
20 were, for example, five children living on the same property.
21
22 3.2.1.4 Data Transformations
23 Processing Original Data Sets
24 The data were received in their original form as personal computer compatible data sets
25 constructed according to the data management procedures of the individual study. From the
26 original data set, certain identifiers were removed to protect the privacy of the individuals
27 and their families, and apparent discrepancies (missing or unused data) were resolved with
28 the database managers for each study.
29 From this set of corrected original data, merged databases were formed in spreadsheet
30 format, according to the primary index. There is one KID, FAM, and PROP file for each
31 study, and a NBHD file for Cincinnati. Each of these intermediate files is large and
July 15, 1993 3-17 DRAFT-DO NOT QUOTE OR CITE
-------
1 cumbersome, but is amenable to extraction of information for specific statistical and data
2 analysis purposes. The final condensed files used for this report consisted of selected fields
3 from the intermediate files. Certain new fields were added for the convenience of data
4 analysis. These were:
5 (1) a numeric value for the number of days since the start of the project (January 1,
6 1989),
7
8 (2) an arithmetic mean for soil and dust (discussed in Section 3.2.1.4), and
9
10 (3) the blood lead concentrations corrected for seasonal cycles and long-term trends
11 (discussed in Section 3.3.5.1).
12
13 With these new fields, the condensed files were converted to SYSTAT format for the
14 statistical analysis of the data as described in Chapter 4.
15
16 Measures of Central Tendency for Soil and Dust
17 For soil and dust, there is a need to reduce multiple measurements to a single
18 representative data point for each property or living unit within a round. In order to
19 determine the appropriate central tendency for this measurement, the participating groups
20 discussed several alternatives at great length without reaching a consensus. Therefore,
21 different measures of central tendency were reported in each of the three studies. The
22 following is an extended discussion of each of these measures, followed by an argument for
23 the use of the arithmetic mean as the best measure in these circumstances.
24 The procedures for selecting a representative soil sample were based on the statistical
25 distribution of data in each study. The Boston study used the arithmetic mean, giving equal
26 weight to all values. The Cincinnati study used the geometric mean, a method that is often
27 used when the measured values are lognormally distributed because it gives lesser weight to
28 extremes. The geometric mean is always lower than the arithmetic mean (except in the case
29 of a perfectly normal distribution) and therefore may be an underestimate of the exposure to
30 the child.
31 The distribution problem was approached differently in Baltimore, where the tri-mean
32 was calculated as the weighted average of the first, second, and third quartiles:
33
34
My 15, 1993 3-18 DRAFT-DO NOT QUOTE OR CITE
-------
X = * (3-1)
4
1 where X = tri-mean, and
2 Qn = upper cutoff for the nth quartile
3
4 The tri-mean approach gives some consideration to the uneven distribution of values
5 without unduly weighting the extremes. As with the geometric mean, the tri-mean is
6 equivalent to the arithmetic mean if the distribution is perfectly normal. For distributions
7 skewed to the left, the estimate is less than the arithmetic mean, and the estimate is greater
8 than the arithmetic mean for right skewed distributions.
9 The ideal measurement of central tendency is one that perfectly represents exposure to
10 the child. In this case a sample would be taken at each location where the child played and
11 this sample would be weighted according to the time spent playing there and the frequency of
12 hand-to-mouth activity during that time. Because this information is not available, a
13 simplification assumption is that weight should be given to location rather that concentration
14 because location, not lead concentration, is the basis of choice for the child's play
15 environment.
16 All three approaches assume that the sampling pattern is random and that exposure to
17 soil is spatially random. Neither condition is strictly true in all three studies. One-third to
18 one-half of the soil samples were taken 1 m from the foundation of the home, where
19 concentrations are known to be higher than elsewhere. Because of playtime interests,
20 parental instructions, or other influences, the child tends to play in specific areas that may
21 represent less than 25% of the total soil area.
22 It would seem reasonable that the ideal method for selecting a representative value
23 should focus on the relationship between the soil and the child. In this respect, an exposure
24 weighted mean of the soil samples would seem to be the most direct approach. This would
25 be an arithmetic mean of soil values corrected for the degree of exposure to the child. For
26 example, a sample taken from bare soil in an area observed to be a play area would be given
27 a high weighting factor for exposure. Grass covered areas with limited accessibility would
28 be weighted on the low end of exposure. Although cumbersome, this method is feasible
29 because such information was collected at the time of sampling in each study. The drawback
July 15, 1993 349 DRAFT-DO NOT QUOTE OR CITE
-------
1 is that the method emphasizes the direct, outdoor playtime contact between the child and the
2 soil, and does not consider other routes of dust exposure, such as soil => household dust,
3 An alternative solution is to consider that the child has equal exposure to the top 2 cm
4 of the entire surface of the soil. In this case, the perfect sample would be to scrape up this
5 upper 2 cm of soil, homogenize it and take a sample. Theoretically, this is equivalent to
6 sampling in a random pattern and taking the arithmetic mean of these samples. In this
7 project, random locations were taken along lines specifically selected to represent the
8 expected high- and low-concentration areas of the plot of soil. In this sense, the arithmetic
9 mean is the best measure of the central tendency of soil data, and is the statistic used in this
10 report.
11
12 3.2.1.5 Adjustments and Corrections to the Data
13 Subjects Dropped from Study
14 During the analysis of their data, the Boston group discovered that two children of the
15 same family had apparently become exposed to lead-based paint while staying at a house
16 outside their neighborhood during a time when it was being remodeled. Both siblings had
17 blood lead concentrations that had tripled in less than 5 mo, between Rounds 2 and 3, from
18 10 to 35 and 17 to 43 jig/dL. The Boston group analyzed their data with and without these
19 children, eventually excluding these data from the analyses used to confirm their hypothesis.
20 This report accepts the conclusion that the data are outliers and dropped them from further
21 analysis. There were no other individuals, families, or properties excluded as outliers from
22 any of the three studies.
23
24 Unit Conversion
25 All data were converted to common units, usually metric. No further corrections were
26 made for analytical blanks or similar analytical adjustments, other than as reported by each
27 study.
28
29 Missing Data
30 The Baltimore data set for children's blood lead contained a large number of missing
31 data. This was because of the complexity of scheduling clinic visits for large numbers of
July 15, 1993 3-20 DRAFT-DO NOT QUOTE OR CITE
-------
1 children over an extended period of time. Statistical procedures for imputing values for these
2 data were used and the results of the reanalysis are given in Chapter 4. Although these
3 imputed data would have changed the conclusions slightly, this procedure for imputing
4 missing data was not used in reaching the primary conclusions of this report. Such
5 procedures for imputing data are becoming more acceptable, and the reader should be aware
6 of their possible impact.
7
8 Seasonal Cycles
9 In the review of the individual reports, the participants were encouraged to explore
10 possible indicators of seasonal and long-term effects on children's blood lead. None were
11 discovered in the individual study reports. However, from the vantage point of the three
12 studies viewed side by side, it is apparent that there is a persistent seasonal pattern for blood
13 lead concentrations that is consistent among all three studies and that can be normalized to
14 reduce the impact of blood measurements taken over a span of weeks during each round.
15 The normalizing parameters appear to be independent of obvious environmental factors.
16 Furthermore, there appears also to be a long-term downward trend in blood lead similar
17 to that observed in other studies. The slope of this downward trend was not the same for all
18 three studies. The discussion of the mathematical methods for making these two corrections
19 to the blood lead concentrations is presented in Section 3.3.5.1.
20
21
22 3.3 PRESENTATION OF THE DATA
23 The purpose of this section is to visually present the data in a manner that illustrates
24 apparent relationships between lead and the child's environment. For each set of figures
25 there is a conclusion drawn in Section 3.4 and supporting statistical analyses in Chapter 4.
26 This approach is intended to give the reader of Chapter 3 an opportunity to digest the data
27 and preliminary hypotheses, then accept or reject these hypotheses based on the statistical
28 analyses and inferences in Chapter 4.
29 The summarized data for all three studies used in this reanalysis appear in Tables 3-7,
30 3-8, and 3-9. For the most part, these data are the basis for the statistical analyses hi
31 Chapter 4.
July 15, 1993 3-21 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 3-7. SUMMARY OF BOSTON STUDY DATA
Mean Soil Pb Cone, f/ig/g)
BOS SPI
BOS PI
BOSP
Mean Soil Pb Cone. > 2,500 Qng/g)
BOS SPI
BOS PI
BOSP
Mean Floor Dust Pb Cone. Oig/g)
BOS SPI
BOS PI
BOSP
Mean Floor Dust Load (mg/nr)
BOS SPI
BOS PI
BOSP
Mean Floor Dust Pb Load Gtg/nf)
BOS SPI
BOS PI
BOSP
Mean Window Dust Pb Cone. 0*g/g)
BOS SPI
BOS PI
BOSP
Mean Window Dust Load (mg/m2)
BOS SPI
BOS PI
BOSP
Mean Window Dust Pb Load (jug/nf)
BOS SPI
BOS PI
BOSP
Mean Hand Pb Load (/tg/pair)
BOS SPI
BOS PI
BOSP
Mean Blood Pb Cone. Oig/dL)
BOS SPI
BOS PI
BOSP
GM Corrected Blood Pb Cone. G*g/dL)
BOS SPI
BOS PI
BOSP
ROUND 1
2,605
2,822
2,780
3,167
3,471
3,518
6,761
4,202
5,231
51
39
47
352
117
291
10,343
10,393
13,030
0.140
0.182
0.176
3.11
6.30
10.97
14.88
13.97
14.88
13.10
12.37
12.02
11.52
11.52
11.52
ROUND 2
2,780
2,780
3,518
3,518
2,445
1,763
3,337
53
31
47
148
50
185
5,452
2,304
11,414
0.079
0,033
0.165
1.22
0.22
7.04
14.88
14.88
12.02
12.02
11.52
11.52
ROUND 3
141
2,662
2,729
142
3,193
3,440
3,239
1,476
1,443
•
54
39
47
245
51
79
6,906
6,350
9,799
0.176
0.171
0.155
2.09
2.22
3.11
14.59
14.44
16.18
10.19
8.85
9.83
11.28
10,30
11,52
ROUND 4
232
2,502
2,679
282
2,914
3,362
1,311
1,337
1,863
27
30
30
39
39
67
11,529
12,184
13,171
0.190
0,226
0.231
2.87
4.29
3,92
18.08
18.10
21.99
10.65
11.49
11.35
10.22
11.07
11.52
July 15, 1993
3-22
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 3-8. SUMMARY OF BALTIMORE STUDY DATA
Round !
1 Round 2
Round 3
Round 4
Round 5
Round 6
Mean Soil Pb Cone. Og/g)
BALSP
BALP-1
BALP-2
Mean Hand Pb Load
BALSP
BALP-1
BALP-2
Mean Blood Pb Cone
BALSP
BALP-1
BALP-2
GM Corrected Blood
BALSP
BALP-1
BAL P-2
501
552
402
Og/pair)
10.7
13.6
9.9
• Gtg/dL)
12.0
12.7
12.5
Pb Cone.
10.0
10.5
10.1
N.D.a
N.D.
N.D.
12.9
14.8
13.7
11.0
11.8
11.3
(ftg/dL)
10.8
11.4
11.2
N.D.
N.D.
N.D.
7.4
9.5
9.0
10.5
10.1
12.0
10.7
9.5
11.4
36
N.D.
N.D.
8.5
6.0
6.5
9.5
8.9
10.1
10.3
9.5
11.4
N.D.
N.D.
N.D.
12.6
17.3
15.5
11.1
9.3
10.8
9.8
8.6
10.1
N.D.
N.D.
N.D.
14.9
13.0
13.0
10.6
9.7
10.6
9.0
8.0
9.3
aN.D. = Not determined.
1 3.3.1 Effectiveness and Persistency of Soil Abatement
2 Soil abatement took place between two rounds of soil measurements. Figure 3-10
3 visually illustrates the concept that preabatement soil concentrations remained constant until
4 abatement, then decreased abruptly during abatement to the level measured at the next round.
5 The reader should understand that these extrapolations may not represent the actual rate at
6 which concentrations changed, but likewise, it would be inappropriate to assume gradual
7 change suggested by the dashed line between the preabatement and first postabatement
8 measurements and to attach some significance to the slope. This issue becomes important
9 again in Sections 3.3.4 and 3.3.5 when the impact of intervention on hand and blood lead is
10 discussed.
July 15, 1993 3-23 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 3-9. SUMMARY OF CINCINNATI STUDY DATA
Mean Soil Pb Cone.
ONSET
QN I-SE-1
ON I-SE-2
CINNT
Mean Street Dust Pb
CINSEI
ON I-SE-1
GIN I-SE-2
CINNT
Round 1
Otg/g)
680
237
N.D.
339
Cone, (/tg/g)
3,937
3,665
3,644
1,583
Round 2
134
247
1,012
346
3,398
3,416
1,990
1,156
Rounds
142
240
555
330
2,118
3,411
1,920
891
RoundS
103
262
724
256
2,559
2,275
1,680
968
Round 6 Round 7
122 166
125 182
233 251
331 267
3,231
3,040
2,905
1,086
Round 9
132
138
442
266
Mean Street Dust Load (mg/m2)
CINSEI
aN I-SE-1
CIN I-SE-2
ON NT
Mean Street Dust Pb
CINSEI
CIN I-SE-1
ON I-SE-2
ON NT
Mean Floor Dust Pb
CINSEI
ON I-SE-1
CIN I-SE-2
ON NT
454
649
760
624
Load (jig/nf)
1,162
2,364
2,440
1,005
Cone. (#g/g)
438
417
447
295
242
561
726
755
789
1,618
973
957
439
512
513
290
363
326
533
481
641
1127
739
498
468
478
431
241
452
420
508
477
968
943
648
587
2,530
2,924
694
1,984
310
126
134
654
808
371
302
442
2,247
2,040
1,117
2,369
Mean Floor Dust Load (g/m2)
CINSH
CIN I-SE-1
ON I-SE-2
CINNT
1.67
0.82
0.82
0.44
0.77
0.09
0.21
0.47
0.28
0.16
0.20
0.20
0.82
0.67
3.28
0.64
0,51
0.54
0.33
0.29
Mean Floor Dust Pb Load (/ig/rn2)
CINSEI
ON I-SE-1
CIN I-SE-2
CINNT
802
415
288
125
350
55
37
69
176
76
88
51
2,456
1,356
1,528
1,085
1,467
1,059
482
1,675
My 15, 1993
3-24
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 3-9 (cont'd). SUMMARY OF CINCINNATI STUDY DATA
Round 1 Round 2
Rounds
Round 5
Round 6 Round 7 Round 9
Mean Window Dust Pb Cone, (jig/g)
ONSET
CIN I-SE-1
ON I-SE-2
CIN NT
1,807 1
4,282 2
2,622 1
1,975 1
,493
,312
,900
,443
1,210
1,611
1,532
992
5,080
3,458
5,940
2,982
4,033
4,014
4,390
,2,412
Mean Window Dust Load (g/m2)
CINSEI
CIN I-SE-1
CIN I-SE-2
CIN NT
10.3
2.64
5.57
13
1.86
3.3
2.71
13.8
0.52
0.87
0.71
0.73
17
12.6
11.0
14.5
6.56
4.62
6.0
5.3
Mean Window Dust Pb Load (jig/m2)
CINSEI
CIN I-SE-1
CIN I-SE-2
CIN NT
Mean Mat Dust
CINSEI
CIN I-SE-1
CIN I-SE-2
CIN NT
Mean Mat Dust
CIN SEI
CIN I-SE-1
CIN I-SE-2
CIN NT
Mean Mat Dust
CINSEI
CIN I-SE-1
CIN I-SE-2
CIN NT
33,700 135
11,400 5
47,400 4
22,700 18
Pb Cone, (jig/g)
150
215 1
204 1
174
Load (mg/m2/day)
*
Pb Load Oig/m2/day)
,000
,260
,600
,500
913
,950
,163
435
6.5
18.7
8.5
-1.8
6.54
7.65
5.83
3.30
1,413
1,560
1,220
726
827
769
997
391
7.7
4.7
7.9
2.0
7.62
5.14
8.04
4.67
60,990
50,400
198,000
112,000
4,284
2,271
1,121
3,692
4.4
4.9
6.2
2,7
2^38
3.20
6.24
0.99
20,500
26,600
34,500
24,500
3,340
2,423
1,017
2,619
28.2
16.6
3.6
12.2
9.80
8.02
2.06
5.29
Mean Entry Dust Pb Cone, (/ig/g)
CINSEI
CIN I-SE-1
CIN I-SE-2
CIN NT
456
452
665
- 348
789
812
681
475
569
456
833
369
1,057
819
1,127
1,804
1,767
922
1,224
1,534
676
1,507
862
1,419
2,148
1,789
675
2,061
July 15, 1993
3-25
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 3-9 (cont'd). SUMMARY OF CINCINNATI STUDY DATA
Round 1 Round 2
Round 3
RoundS
Round 6 Round 7
Round 9
Mean Entry Dust Load (g/m2)
CINSEI
CIN I-SE-1
CIN I-SE-2
ON NT
Mean Entry Dust Pb
CINSEI
ON I-SE-1
GIN I-SE-2
cm NT
Mean Hand Pb Load
CINSEI
CIN I-SE-1
C3N I-SE-2
ON NT
36.4 6.1
1.0 0,11
1.24 0.15
12.0 2.45
Load Otg/m2)
16,800 5,670
540 82
828 144
6,715 513
(fig/pair)
8.6 7.1
12.3 8.3
14.1 11.2
5.3 4.4
4.8
0.50
0.66
0.32
2,856
261
496
113
5.9
6.5
7.1
2.7
6.6
10.9
6.12
22.5
9,517
8,280
9,678
23,300
17.7
9.0
16.6
5.9
2.6 0.34
1.07 0.64
0.86 0.89
1.46 2.26
3,634 238
967 657
1,186 589
4,800 1,270
16.0 8,9
10.4 7.5
11.3 , 14.1
11.3 4.7
1.8
2.24
1.31
17.8
3,042
2,029
1,695
12,900
18.2
16.3
39.6
14.2
Mean Blood Pb Cone. ftig/dL)
CINSEI
GIN I-SE-1
CTN I-SE-2
ON NT
GM Corrected Blood
CINSEI
CIN I-SE-1
CIN I-SE-2
ON NT
10.4
10.3
13.7
9.2
Pb Cone. 0*g/dL)
8.3
8.5
9.8
7.3
8.3
8.8
11.8
6.6
8.9
9.3
11.9
7.4
10.2
8.1
10.7
7.6
9.6
8.5
10.0
7.6
9.3
7.1
9.8
7.2
11.8
10.3
12.1
9,9
9.9
9,5
12.1
7,9
12.0
11.3
13.6
10.7
1 In order to form an effective, permanent barrier between the source of lead and the
2 human environment, soil abatement must reduce the concentration of lead in the soil in a
3 manner that is persistent for a period of years. In each of the three studies, measurements
4 were made prior to abatement and immediately after abatement (within 3 mo). Followup
5 measurements were made periodically until the end of the study in Cincinnati and Boston.
6 The results of these soil analyses, corrected for small errors, are graphically illustrated in
7 Figures 3-11, 3-12, and 3-13. These data show, for all three studies, a substantial reduction
July 15, 1993 3-26 DRAFT-DO NOT QUOTE OR CITE
-------
o5 1500
O)
a.
c
o
"S
~ 1000
I
o
O
;Q
^ 500
3
O
O
to n
. . ___* :,„- *
* » •» ; — ~-^
\ i
\ 1 -
\ 1
\ I
\ 1
\ 1
\i
/*!4
l\_
Abate
i i i i i i i i i
0 100 200 300 400 500 600 700 800 900 1000
Days from Start of Project
. • Abated Soil or Dust —0— Unabated Soil or Dust
Figure 3-10. Hypothetical representation of intervention impact (solid lines, shaded
areas) on soil and dust concentrations. When intervention occurs between
rounds of measurements, the dashed line misrepresents the transition from
the first to the second measurement hi the abated group.
4,000
3,500
•§?
5 3,000
•S 2,500
§ 2,000
o
0 1,500
.0
Q_
1 1'000
500
JU|
oct-89
May-90
200 300 400
Days from Start of Project
-•-BOS SPi -*- BOS PI -»— BOS P
500
Figure 3-11. The arithmetic means of Boston soil lead concentrations by study group
show the effectiveness and persistency of soil abatement. The date shown
is the arithmetic mean of the individual site sampling dates. Figure 3-14 is
a similar plot of soil lead concentrations above 2,500 jig/g.
July 15, 1993
3-27
DRAFT-DO NOT QUOTE OR CITE
-------
1,100
1,000
900
•55
"Si 800
100
Mar-89
Sep-39
Abate I
Jul-91
Abaton
i
100 200 300 400 500 600 700 800 900 1,000
Days From Start of Project
CIN SEI
CIN l-SE-1
CIN I-SE-2
CIN NT
figure 3-12. Cincinnati soil lead concentrations. The arithmetic means by study group
of the soil lead concentrations that show the effectiveness and persistency
of soil abatement.
ouu
700
seoo
OS
g 500
IS
g 400
Ł300
=§
W 200
100
ft
-
- **»
Abate
ir
j
i
j
$
-I Janr9l
0 200 300 400 500 600 700 80
Days from Start of Project
BALSP —*— BALP-1 *
BALP-2
Figure 3-13. Reconstruction of the expected effectiveness of soil abatement in the
Baltimore study. Measurements were made for both study groups before
abatement, but only for the BAL SP group after abatement. The lines for
BAL P-l and BAL-P-2 are shown only for comparison purposes.
July 15, 1993
3-28
DRAFT-DO NOT QUOTE OR CITE
-------
1 in the amount of lead in soil measured immediately after abatement, and that in Boston and
2 Cincinnati, where followup soil measurements were taken, this reduction persisted for the
3 duration of the study. For this calculation, the arithmetic mean was used in order to give
4 equal weight to all soil samples taken. It should be noted that in Baltimore, the
5 postabatement measurements were made only in the locations where soil had been excavated
6 and removed.
7 Each study was able to achieve the targeted concentration for abated soil. The mean
8 soil concentrations following abatement are not substantially higher than the specifications for
9 clean soil. The amount of soil lead reduction actually achieved directly because influences
10 the expected changes in dust lead and blood lead. In Section 3.3.5, an attempt will be made
11 to evaluate the treatment/response relationship for each step of the pathway of lead in the
12 human environment.
13 To determine the effectiveness and persistency of soil abatement, the arithmetic mean
14 for each parcel of land was taken for each round where soil measurements were made. The
15 parcel means for the Boston and Cincinnati studies show that abated soil concentrations
16 (BOS SPI and GIN SEI) dropped significantly after abatement (Figures 3-11, 3-12, and
17 3-14), whereas unabated soil (BOS PI, BOS P, and CIN NT) appear to decrease only
18 slightly, if at all. This reduction in the abated soil was persistent through the end of the
19 study. The Cincinnati groups CIN SEI-1 and CIN SEI-2, which received soil and exterior
20 dust abatement during the second year, showed a decrease in the range expected, but there
21 were no followup measurements to demonstrate persistency. Because there were no
22 measurements, in the Baltimore study, of abated soil beyond the first measurement
23 immediately after abatement, and there were no measurements of unabated soil after the
24 beginning of the study, the data show only that abatement was effective, with no
25 measurement of persistency.
26 Because it is known that blood lead concentrations follow a seasonal cycle and a
27 downward trend, every effort was made in this document to note and discuss possible trends
28 or cycles in the environmental sources of lead that might explain this blood lead pattern.
29 There appears to be some indication of a general downward trend of lead concentrations
30 in the unabated soil. Although not statistically significant for any individual group, the fact
31 that all groups where the soil remained unabated show this phenomenon lends some credence
July 15, 1993 3-29 DRAFT-DO NOT QUOTE OR CITE
-------
4,000
3,500
Is 3|00°
Ł.
g" 2,500
I
G 2,000
g 1,500
i 1,000
p
500
0
Jul-89
May-90
200
Abate
300
400
500
Days from Start of Project
BOSSPI —4— BOS PI -*— BOSP
Figure 3-14. Boston soil lead concentrations above 2,500 /ig/g. Similar to Figure 3-11,
these data show the effectiveness and persistency in the Boston study for
soils with lead concentrations greater than 2,500 /ig/g.
1 to this observation. Analysis of quality assurance/quality control audit samples shows this
2 trend cannot be attributed to analytical drift (see Section 3.1). There are several obstacles to
3 demonstrating a temporal trend or cycle in soils. Soil lead concentrations vary widely over a
4 relatively small distances, even less than 1 m. Because it was not feasible to return to the
5 exact spot for sequential soil sample, it is not reasonable to expect two sequential samples to
6 have the same value. It is reasonable to expect, however, that the differences with time to
7 be random in the absence of a seasonal cycle or trend.
8 On the issue of recontamination following abatement, it is curious that such an increase
9 did not occur in the comparable sites where no abatement occurred. This suggests that soil
10 lead might reach some equilibrium with other components of the environment. Although this
July 15, 1993
3-30
DRAFT-DO NOT QUOTE OR CITE
-------
1 is not an unreasonable conclusion, there is no complete explanation for the mechanisms that
2 regulate this equilibrium.
3
4 3.3.2 Effectiveness and Persistency of Exterior Dust Abatement
5 Dust is measured in both concentration and surface loading. Concentration is measured
6 in micrograms of lead per gram of dust, whereas loading is measured in milligrams of lead
7 per square meter. When dust abatement is performed, the amount of dust changes, but the
8 concentration of lead in the dust does not. Therefore, there should be no change in dust lead
9 concentration unless the source of the dust changes. Where soil abatement has been
10 performed in connection with dust abatement, the dust lead concentration should also
11 decrease abruptly. If there is a mixture of dust sources and only one has been abated, the
12 lead concentration would change less abruptly, according to the contribution from each
13 source.
14 Measurement of dust loading on smooth interior surfaces has in the past provided
15 reliable information of the contribution of dust from multiple sources. The attempt to use
16 this procedure for exterior dust in Cincinnati presented some problems. The surfaces such as
17 asphalt and concrete were not as smooth as hardwood floors, and the sources were not as
18 easily identified. These factors should be taken into consideration when interpreting the
19 Cincinnati exterior dust data (Figures 3-15, 3-16, and 3-17).
20 In the Cincinnati study, the exterior dust load data (Figure 3-15) suggest a trend or
21 cycle with a period of about 1 year for the unabated group (CIN NT). When this pattern is
22 considered in the abated areas, it was not clear whether abatement of exterior dust was
4
23 effective or whether the abated areas were recontaminated before the postabatement
24 measurements were taken (1 week after abatement).
25 , Exterior dust was measured and abated only in the Cincinnati study. In the CIN SFJ
26 group, exterior dust concentrations decreased after abatement, then increased steadily through
27 the end of the study. In the CIN I-SE-1 group, exterior dust concentrations dropped similar
28 to CIN SEI without abatement in 1989, then increased through the end of the study even
29 following abatement in 1990. This indicates that even though the relative contribution of
30 lead from other sources changed over time, exterior dust abatement did not seem to impact
31 the contribution from these sources.
July 15, 1993 3-31 DRAFT-DO NOT QUOTE OR CITE
-------
300
400 500
Days From Start of Project
600
700
CIN SB
GIN l-SE-1
CIN l-SE-2
CIN NT
Figure 3-15. Cincinnati exterior dust load measurements. The data indicate that
exterior dust abatement was effective but not persistent for more than
150 to 200 days.
3,000.
I
I
2 1,000-
o
1
500
2,500 -Aus-89
2,000
1,500
Oct-90
Abate I
Abate n
§00
300
400 500
Days From Start of Project
600
700
CIN SEl
CiN I-SE-1
CIN l-SE-2
CIN NT
figure 3-16. Cincinnati exterior dust lead load measurements. The data indicate small
changes in the lead load that may have been persistent for several months,
as indicated by the recovery time for CIN SEI of about 1 year.
July 15, 1993
3-32
DRAFT-DO NOT QUOTE OR CITE
-------
Q
.g
0)
s
4,500
I 4,000^
*c 3,500
S 3,000
-------
1 studies, little evidence is available to accept or reject this hypothesis. However, in the
2 context of exposure pathways, the parcels of soil in Boston and Baltimore were on the
3 individual properties, whereas in Cincinnati, they were in areas separated spatially from the
4 living units, such as parks and vacant lots.
5
6 3.3.3 Effectiveness and Persistency of Interior Dust Abatement
7 The data for the Boston study interior dust are shown in Figures 3-18 through 3-23.
8 In both BOS SPI and BOS PI, there was a general decrease in the floor dust lead loading
9 following interior dust abatement, as shown in Figure 3-20, and further decreases were
10 observed at 7 to 12 mo after abatement. In the window wells, however, the lead loading
11 decreased immediately after dust abatement (Figure 3-23), persisted for a few months, then
12 returned to original levels about 12 mo after abatement. The high concentrations of lead in
13 the window well dust (5,000 to 22,000 jwg/g) may indicate lead-based paint was present
14 (Figure 3-22).
15 The Cincinnati study (Figures 3-24 through 3-35) found an immediate reduction in floor
16 dust lead loading that persisted for at least 5 mo, followed by an increase at 12 mo to 70%
17 of the preabatement level in CIN SHE, where soil abatement had taken place, and to nearly
18 twice the preabatement level in CIN I-SE-1 and CIN I-SE-2, where soil had not yet been
19 abated. Similar patterns were observed in the window wells (Figures 3-27 through 3-29) and
20 entry ways (Figures 3-30 through 3-35). The window well concentrations were lower in
21 Cincinnati (1,000 to 2,300 pg/g) than in Boston, suggesting a minimum influence of lead-
22 based paint,
23
24 3.3.4 Hand Dust Results
25 Becatise hand-to-mouth activity is one route by which lead may be ingested, the amount
26 of lead on the child's hand is an indicator of exposure. Only lead loading information is
27 available because it was necessary to take the sample with wet wipes. The units of
28 measurement are micrograms per pair of hands rather than micrograms per square meter.
29 Nevertheless, these data are important intermediates between soil/dust and blood lead.
30 In Boston, there was a general increase in hand lead throughout the study
31 (Figure 3-36). Although there is no explanation for this increase, there appears to be less of
July 15, 1993 - 3-34 DRAFT-DO NOT QUOTE OR CITE
-------
o>
8,000
7,000
6,000
.o
Ł 5,000
O
4,000
3,000
2,000
1
1,000
Oct-89
Jul-90
Abate
250 300 350 400 450 500
Days from Start of Project
-•- BOS SPI —*— BOS PI ->— BOSP
550
600
Figure 3-18. Boston floor dust lead concentration. The data appear to support the
observation that two sources of lead are present in household dust: soil
and paint. Dust abatement alone is not expected to change the dust lead
concentration.
60
55
50
I
^ 45
es
o
« 40
Q
fc
o
E
35
30
25
Jan-90
Apr-90
250 300 350 400 450 500
Days from Start of Project
-•- BOS SPI —*— BOS PI ->— BOS P
550
600
Figure 3-19. Boston floor dust load. The dust load was not reduced in the BOS SPI
group, indicating an immediate recovery. A longer recovery period is
indicated for the BOS PI group.
July 15, 1993
3-35
DRAFT-DO NOT QUOTE OR CITE
-------
"oS
"S
400
350
300
250
200
150
100
50
Ocl-89
Apr-90
250 300 350 400 450 500
Days from Start of Project
-m- BOSSPI—*- BOS PI -*— BOSP
550
600
Figure 3-20. Boston floor dust lead load. Even though the dust load in Figure 3-19
indicates a quick recovery, the lead load did not recover immediately,
indicating that the source of the lead was cut off, at least temporarily.
o>
o
O
14,000
12,000
10,000
8,000
6,000
Q 4,000
1 2,000
Ocl-89
JUl-90
250
300
350 400 450 500
Days from Start of Project
BOSSPI-4— BOS PI ->- BOSP
550
600
Figure 3-21. Boston window dust lead concentrations. Faint stabilization and soil
abatement appear to have been effective and persistent for several
hundred days, similar to floor dust. The recovery observed between April
and July 1990 was not observed for the floor dust load data.
July 15, 1993
3-36
DRAFT-DO NOT QUOTE OR CITE
-------
0.25
ff- 0.2
"oi
•§ 0.15
0.1
0,05
Oct-89
Abate
Jul-90
250 300 350 400 450 500
Days from Start of Project
-m- BOSSPI-*- BOS PI ->— BOSP
550
600
Figure 3-22. Boston window dust load. These data show the effectiveness of window
dust abatement, which appears to recover after about 150 days, similar to
floor dust loads observed in Figure 3-19.
0
250
300 Abate 350 400 450 500
Days from Start of Project
BOS SPI -~4~ BOS PI -+— BOS P
550
600
Figure 3-23. Boston window dust lead load. As with floor dust lead loads, the window
data indicate that both paint and soil sources of lead were interrupted, at
least temporarily. The data appear to be consistent with Figure 3-20.
July 15, 1993
3-37
DRAFT-DO NOT QUOTE OR CITE
-------
3,000
•§2,500
2,000
g 1,500
Ł
tj 1.000
B 500
in
J_
Oct-90
t
00 200 300 400 500 600 700
«-
Days From Start of Project
—•— CINSEI —•— ClNI-SE-1 +• CINI-SE-2 —>— GIN NT
Figure 3-24. Cincinnati floor dust lead concentrations. The small changes in lead
concentrations suggest that the sources of lead did not change as a result
of the abatement activities. The abrupt increase after November 1989
indicates one or more sources changed markedly during that tune.
3.5
3
2
1.5
1
0.5
Jul-89 Abate
Jul-90
00 200 300 400 500
Days From Start of Project
• CINSEI • CINI-SE-1 + CINI-SE-2
Oct-90
600
700
CINNT
Figure 3-25. Cmcmnati floor dust load. These data confirm the effectiveness of the
household dust abatement and show that this reduction was persistent for
more than 100 days.
July 15, 1993
3-38
DRAFT-DO NOT QUOTE OR CITE
-------
3,000
200
CIN SEI
300 4OO 500
Days From Start of Project
CIN l-SE-1
CIN l-SE-2
600
CIN NT
700
Figure 3-26. Cincinnati floor dust lead load. The data suggest that the sources of lead
were interrupted by the abatement activities, but that at least one source
recovered after November 1989.
6,000
5,000
.2
4,000
Ł
%
Q
3,000
2,000
1,000
j
?00
Pct-80
—
200 300 40O 500 600
Days From Start of Project
CIN SEI —•— CIN l-SE-1 —*— CIN l-SE-2 —^— CIN NT
700
Figure 3-27. Cincinnati window dust lead concentration. The response to abatement
appears to be consistent with the observations of the floor dust in
Figure 3-24.
My 15, 1993
3-39
DRAFT-DO NOT QUOTE OR CITE
-------
20
1= 15
D
I
C
Jul-90
JuI-89 S®
ict-90
00 200 300 400 500
Days From Start of Project
—m— CIN SEI • CIN I-SE-1 » CIN l-SE-2
600
700
CIN NT
Figure 3-28. Cincinnati window dust load. The impact of abatement and the change in
the CIN NT group are consistent between floor dust load (Figure 3-25) and
window dust load.
20o,ooq
150,000-
100,000-
•g 5Q,COC -
IOO
200
CIN SEI
300 400 500
Days From Start of Project
CINI-SE-1 —*— CIN l-SE-2
600
CIN NT
700
Figure 3-2P. Cincinnati window dust lead load. These data indicate one of the few
instances where exposure to lead, as indicated by the household dust lead
load, increased during abatement for one of the study groups (CIN SEI).
July 15, 1993
3-40
DRAFT-DO NOT QUOTE OR CITE
-------
2 Months
GIN SEI
3 Months 10 Months 2 Months
Mat Sequence Number
CIN1-SE-1 • CINI-SE-2 Fl GIN NT
Figure 3-30. Cincinnati mat dust lead concentration. Clean mats were placed at the
entry to the housing unit. It was expected that the lead concentrations
would reach equilibrium with the dust entering the home and remain
constant for the duration of the study.
30
f
f 25
"5s
§ 20
1
3
15
3 10
.o
o_
ts
& 5
to
Jul - Aug '89
Jun-Jul'90
2 Months
GIN SEI
3 Months 10 Months
Mat Sequence Number
2 Months
GIN l-SE-1
GIN l-SE-2
GIN NT
Figure 3-31. Cincinnati mat dust load. La a general sense, the dust load should
increase as long as the mats are in place, which appears to be the case
from the data shown here.
July 15, 1993
3-41
DRAFT-DO NOT QUOTE OR CITE
-------
2 Months
GIN SEI
3 Months 10 Months
Mat Sequence Number
2 Months
CIN l-SE-1
CINI-SE-2
CIN NT
Figure 3-32. Cincinnati mat dust lead load. These data appear to be consistent with the
observation that there was an increase in exposure to house dust that
occurred between November 1989 and October 1990.
3,000,
2,500
2,000-
I
I
.0 1,000-
I
3 500
Jun-91
Jul-90 0ct,QO
Nov-90
Abate
100 200 300 400 500 600 700
Days From Start of Project
CIN SEI —•— CINI-SE-1 —•— CIN l-SE-2
800 900 1,000
CIN NT
Figure 3-33. Cincinnati entry dust lead concentration. The entry way subset of the
floor dust shows a pattern different from the complete floor dust data of
Figure 3-24 and the mat dust data of Figure 3-30. Note the two additional
rounds, November 1990 and June 1991.
July 15, 1993
3-42
DRAFT-DO NOT QUOTE OR CITE
-------
100 200
CIN SB
300 400 500 600 700
Days From Start of Project
CIN l-SE-1
CIN l-SE-2
800 900 1,000
CIN NT
Figure 3-34. Cincinnati entry dust load. Similar to Figure 3-25, dust abatement at the
entry appears to have been effective and persistent through November
1989.
100 200 300 400 500 600 700 800 900 1,000
Figure 3-35. Cincinnati entry dust lead load. There were a few housing units with
entry dust lead loads three or more orders of magnitude higher than floor
dust lead loads, which obscured the interpretation of the entry dust lead
load.
July 15, 1993
3-43
DRAFT-DO NOT QUOTE OR CITE
-------
23
22
21
20
19
I"
= 16
15
14
Sep-90
Oct-89
1l50 300 350 400 450 600 550~
Days From Start Of Project
-m- BOSSP! —*— BOS PI -*— BOSP
600
650
Figure 3-36, Boston hand lead load. The amount of lead on the hands increased in the
same manner as the floor dust load (Figure 3-19) and similar to the floor
dust lead load (Figure 3-20).
1 an effect for the groups that received soil and dust intervention, and this reduction is greatest
2 for the group that received soil, dust, and paint intervention.
3 Baltimore hand lead values did not follow a discernable pattern (Figure 3-37) and there
4 appears to be no difference between the two intervention groups.
5 In Cincinnati, the hand dust lead load (Figure 3-38) appears to follow the pattern of
6 change observed in the floor dust lead load (Figure 3-26). This is an important link to
7 establish in the exposure pathway. The measurement of lead on children's hands is a new
8 exposure assessment tool that was introduced in this project as an alternative to blood lead
9 measurements, which were expected to respond more slowly to environmental changes.
10 Of the three studies, only the Cincinnati group had previous experience with hand lead
11 analysis. The discussion below of the relationship of hand lead to blood lead will shed
12 further light on this critical pathway.
July 15, 1993
3-44
DRAFT-DO NOT QUOTE OR CITE
-------
Sep-!!1
200 400 600
Days From Start of Project
800
1000
BALSP
BALP-1
BAL P-2
Figure 3-37. Baltimore hand lead load. There were no sequential measurements of
Baltimore house dust to compare the hand lead load.
100
200
GIN SB
300 400 500 600 700
Days From Start of Project
—•— CINI-SE-1 —A— CINI-SE-2
900 1,000
CINNT
Figure 3-38. Cincinnati hand lead load. The pattern of hand lead load change, both
increases and decreases, appears to follow the pattern of floor dust lead
load hi Figure 3-26. Arrows indicate time of abatement.
July 15, 1993
3-45
DRAFT-DO NOT QUOTE OR CITE
-------
1 3.3.5 Blood Lead Results
2 The seasonal cyclic patterns and long-term trends observed in many studies of
3 children's blood lead were discussed in Chapter 2. The importance of understanding these
4 patterns is illustrated in the next series of graphs of the Baltimore data. The first,
5 Figure 3-39, shows the uncorrected blood lead data that suggest the data are fairly smooth
6 through most of the study, and essentially identical between groups. But there is a
7 problematic increase at the end, especially for the BAL SP group. Notice that none of the
8 rounds' measurements are taken at exactly the same time of year. Figure 3-40 fits a sinusoid
9 to these data. The sinusoid has an amplitude of approximately 15 % of the blood lead, a
10 period of I year, and a maximum at August 18.
11
15
I
o
I
Q
O
•a
8
DQ
Ocl-88
Jun-91
•S>00
Feb-90
Sep-91
200 400 600 800
Days From Start of Project
- BALSP S BALP-1 + BALP-2
1000
1200
Figure 3-39. Baltimore uncorrected blood lead concentrations. There appears to be
little difference between study groups.
July 15, 1993
3-46
DRAFT-DO NOT QUOTE OR CITE
-------
15
i
o
O
Ł
T3
I 5
CO
8
Oot-88
Apr-89
Feb-90
Jun-91
Sep-91
200 400 600 800
Days From Start of Project
1000
1200
BALSP
BALP-1
BAL P-2
Figure 3-40. Baltimore blood lead concentrations corrected for seasonal cycle and
long-term time trends. The manner of correction for seasonal cycles is
described in Figure 3-41.
1.2
100 200 300
Calander Date Of Blood Collection
400
Figure 3-41. Seasonally adjusted correction factor for blood lead concentrations. Blood
lead concentrations typically vary by about 30% seasonally, and peak in
the late summer. This correction appears is the same for all three studies,
both in the offset ($) and the amplitude (b2).
July 15, 1993
3-47
DRAFT-DO NOT QUOTE OR CITE
-------
1 3.3.5.1 Blood Lead Correction Factors
2 By minimizing the impact of these seasonal effects, the impact of intervention becomes
3 more systematic and more amenable to statistical evaluation and interpretation. The fact that
4 the corrections made to the data were minimal, in the sense that the same peak date
5 (August 18) and relative magnitude (15%) was applied to all children in all three studies
6 argues strongly for the validity of this correction.
7 To correct for seasonal variations and apparent long-term trends in children's blood
8 lead concentrations, the following equation was used:
9
10 CPbB{ - (PbBi + b^ [1 + b2cos(2-jrdi/365 + $)] • (3-2)
11
12 where CpbEi = corrected blood Pb for child i, in ^.g/dL;
13 PbBf = measured blood Pb for child /, in j«g/dL;
14 b± = time trend constant, in jitg/dL/day;
15 di = date of blood measurement for child i, in days;
16 b2 = seasonal cycle constant, unitless;
17 = offset for calendar year, in radians.
18
19 The first segment of this equation, (PbBj + b^, adjusts the measured blood lead for
20 the long-term downward trend believed to exist for the general population. This constant is
21 different for each city and can be measured in the child populations prior to intervention, or
22 where no intervention occurred. The values for this constant were -0.0025 ^g/dL/day for
23 Boston and —0.00025 jug/dL/day for Baltimore and Cincinnati. The significance of the
24 10-fold difference between Boston and the other two studies will be discussed later.
25 The second segment, (1 + b2cos(2irdi/365 + $)), adjusts the blood lead for the
26 seasonal cycle. This determination is based on a sinusoidal cycle and is largely independent
27 of the observed blood lead concentrations. The two coefficients that impact this component
28 of the equation are b2 and 3>, both of which are the same for all three studies. The seasonal
29 cycle constant, b2, determines the magnitude of the cycle, and the offset, <Ł, is based on the
30 date of the maximum. The availability of three simultaneous longitudinal studies provided a
31 rare opportunity to evaluate a large database for seasonal cyclic patterns. Analysis of the
July 15, 1993 3-48 DRAFT-DO NOT QUOTE OR CITE
-------
1 three studies led to the conclusion that the blood lead varied by ±15% throughout the year.
2 Hence, the constant, b2, is 0,15.
3 The peak occurred about August 18 for each study. The offset to the date midway
4 between the maximum and minimum would be 91 days, or 0,82 radians in terms of the
5 sinusoid. Figure 3-41 illustrates the manner in which the seasonal cycle correction was
6 made.
7 '
8 3.3.5.2 Reaualysis of Boston Study Blood Lead Data
9 The uncorrected blood lead concentrations for the Boston study are shown in
10 Figure 3-42, and the corrected values are shown in Figure 3-43. The corrected blood lead
11 concentrations graphically illustrate the conclusions of the Boston report, that intervention
12 probably accounted for a decrease of 0.8 to 1,5 ^g/dL in the blood lead. The observation
13 that all three Boston study groups experienced an increase in uncorrected blood lead
14 concentrations between Round 3 (April 1990) and Round 4 (September 1990) is consistent
15 with similar observations in the hand dust lead load and, to a lesser degree, the window dust
16 lead load. The apparent absence of a comparable increase in floor dust lead load runs
17 counter to the expected pattern of the floor dust lead load being the primary route for dust
18 exposure in children.
19
20 3.3.5.3 Reanalysis of Cincinnati Study Blood Lead Data
21 The wealth of information from the more detailed measurements of household dust in
22 the Cincinnati study presents a proportionally greater challenge to the paradigm of dust
23 exposure pathways. The uncorrected blood lead concentrations shown in Figure 3-44
24 correspond roughly to the changes observed in the hand dust lead loads of Figure 3-38. And
25 there are several points where the blood lead concentrations are consistent with the observed
26 changes in the various forms of house dust. The floor and window dust lead loads are
27 especially indicative to the exposure route, and the mat dust lead load seems to account for
28 the increase in blood lead concentrations after November 1990.
July 15, 1993 3.49 DRAFT-DO NOT QUOTE OR CITE
-------
14
13
Ł
10
Oct-89
Sop-90
250
300
350
400
450
500
550
600
~eter
700
Days From Start Of Project
BOS SPi —*- BOS PI -»— BOS P
Figure 3-42. Boston uncorrected blood lead concentrations. The approximate time of
soil and dust abatement is shown by arrows.
12.S
•3- 12
|
xT 11.s
1
I
2
11
10'5
10
Sep-90
250 300 350 400 450 500 550
Days From Start Of Project
-m- BOS SPi ->— BOS P! ->— BOS P
600
650
700
Figure 3-43. Boston blood lead concentrations corrected for seasonal cycles and
long-term time trends and normalized to BOS P. The approximate time of
soU and dust abatement is shown by arrows.
July 15, 1993
3-50
DRAFT-DO NOT QUOTE OR CITE
-------
15
14
O* 13
1
•3 12
11
o
o
S 10
0
8 8
3 7
CO
6
5
Jul-89
Jun-91
100 200 300 400 500 600 700 800 900 1,000
Days From Start Of Project
-•-GIN SE! -*-C!N l-SE-1 —*-CIN l-SE-2 -fr—CIN NT
Figure 3-44. Cincinnati uncorrected blood lead concentrations. The approximate time
of soil and dust abatement is shown by small arrows. Compare to hand
lead load patterns in Figure 3-38.
1 The corrected blood lead concentrations in the Cincinnati study (Figure 3-45) suggest
2 an impact of changes in environmental lead. On the same time scale as the Boston and
3 Baltimore projects, the observed decrease at July 1990 (between 0.6 and 1.2 jig/dL) is
4 similar to that seen in the Boston study. The group that received soil abatement in the first
5 year, CIN SET, continued to show increasing blood lead concentrations through the following
6 year, and the CIN I-SE-1 and CIN I-SE-2 groups responded negatively following soil and
7 exterior dust abatement in the second year. These uncertainties require that final conclusions
8 on the impact of abatement be postponed until more detailed analyses, as described hi
9 Chapter 4, can be made.
July 15, 1993
3-51
DRAFT-DO NOT QUOTE OR CITE
-------
10
9.5-
8.5
6.5
6
Nov-89
JuI-90
Nov-90
Jui-sa
Jun-91
100 200 300
CIN sa
400 500 600 700
Days From Start Of Project
CINI-SE-1 » CINI-SE-2
800 900 1,000
CIN NT
Figure 3-45. Cincinnati blood lead concentrations corrected for seasonal cycles and
long-term time trends and normalized to CIN NT to show possible impact
of soil and dust abatement. The approximate tune of soil and dust
abatement is shown by small arrows.
1
2
3
4
5
6
7
8
9
10
11
12
13
3.4 SUMMARY OF RESULTS
The data presented in this section lead to the following conclusions:
(1) Soil abatement in each study effectively reduced the concentration of lead
in the soil in the areas where soil abatement was performed.
(2) In the Boston and Cincinnati studies, the effectiveness of soil abatement
was persistent through the end of the study. There were no followup
measurements of soil in Baltimore to determine persistency.
(3) Many of the study groups showed a slight downward trend in soil lead
concentration independent of abatement.
July 15, 1993
3-52
DRAFT-DO NOT QUOTE OR CITE
-------
1 (4) Exterior dust abatement, performed only in Cincinnati, was not persistent,
2 indicating a source of lead other than soil at the neighborhood level.
3
4 (5) Interior dust following abatement as performed in Cincinnati and Boston,
5 responded to subsequent changes in exterior dust and soil lead. Entry was
6 measurements of lead concentration and lead load were a good indicator
7 of the movement of environmental lead.
8
9 (6) Hand lead measurements reflected general trends in blood lead
10 measurements and may be a reasonable estimate of recent exposure.
11 However, the results do not shown that hand lead measurements, as
12 performed in these studies, can be an adequate surrogate for blood lead
13 measurements.
14
15 (7) Because paint stabilization was performed on all homes with lead-based
16 paint in Boston (exterior and interior) and Baltimore (interior only), there
17 is no measure of the effectiveness or persistency of this form of
18 intervention.
19
20 From the standpoint of changes in environmental lead independent of intervention,
21 there were few instances where changes in the blood lead concentrations could not be
22 attributed to these changes. Some of these changes were expected, although uncontrolled.
23 An example is the abrupt change in floor dust load between April and July 1990, shown on
24 Figure 3-19. It appears that all homes were thoroughly cleaned during this period, but there
25 was no corresponding decrease in the window dust load in Figure 3-22. Some changes were
26 unexpected and equally difficult to explain. The dramatic increase in dust lead
27 concentrations in the Cincinnati study between November 1989 and July 1990, observed in
28 nearly every instance (floor, window, mat, and entry), remains unexplained. Although this
29 apparent contamination appears to have overwhelmed the intervention efforts, the fact that
30 both hand dust loads and blood lead concentrations responded to environmental lead gives
31 credence to the strong link between environmental lead and blood lead.
32 In terms of changes attributed to intervention, it is appropriate to note that all three
33 studies observed a quantifiable blood lead concentration change in response to soil, dust and
34 paint intervention. The analyses in Chapter 4 will show that, although not always
35 statistically significant, this quantifiable response to intervention is consistent even at low
36 levels of environmental lead. Normalized to a decrease in soil lead concentration of
37 1,000 jug/g, the response in blood lead concentration appears to be about 1 /wg/dL. This
July 15, 1993 3.53 DRAFT-DO NOT QUOTE OR CITE
-------
1 suggests that there is no plateau, within the ranges measured in this project, where the
2 removal of environmental lead will not produce a corresponding reduction in blood lead
3 concentrations,
4
July 15, 1993 3.54 DRAFT-DO NOT QUOTE OR CITE
-------
i 4. STATISTICAL INTERPRETATION
2 OF THE RESULTS
3
4
5 4.1 PREPARATION OF PROJECT DATA SETS FOR STATISTICAL
6 ANALYSIS
7 Blood lead is a measure of the recent history of lead exposure and may respond to
8 decreases environmental changes in lead within a 2- to 4-mo time frame. Reductions in
9 exposure might be somewhat attenuated by the remobilization of lead in bone tissue. There
10 is little information on these biokinetic translocations of lead when the total body burden is
11 decreasing. If the total lead exposure of the child decreases, there seems to be no doubt that
12 the blood lead concentrations would decrease, but there are no other comparable studies of
13 the rate at which blood lead concentrations respond to decreasing lead exposure.
14 Although the objective of the project was to reduce blood lead concentrations in
15 children by reducing the concentrations of lead in soil, other measures of the impact of soil
16 abatement are possible and, indeed, perhaps more realistic. These are reductions in house
17 dust lead concentrations and reductions in hand lead. The rationale for evaluating all three
18 measures is that changes in house dust lead concentrations, especially in entry areas where
19 the influence of exterior dust and soil is greatest, should be more responsive than blood lead
20 to changes in soil lead and should not be influenced by nondust sources of lead, such as food
21 and drinking water.
22 Hand dust reflects a mixture of dust lead sources throughout the child's environment,
23 both indoors and outdoors. It may be a representative measure of total dust exposure, but
24 the impact of soil abatement is diminished when there are sources of dust lead other than soil
25 in the child's environment.
26 In this project, changes in blood lead must be interpreted in the context of four time-
27 dependent effects that are independent of each other. These are
28
29 (1) the typical seasonal changes in children's blood lead concentrations, found
30 in virtually every longitudinal study, that usually indicate a peak in
31 concentration during the late summer months;
32 -
July 15, 1993 4-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 (2) the changes that occur with age during early childhood that usually peak
2 between 18 and 27 mo;
3
4 (3) long-term changes in national baseline levels of exposure, believed to be
5 mostly from reductions of lead in gasoline and in food, that are reflected
6 in a downward trend for childhood blood lead levels observed since 1978;
7 and
8
9 (4) changes that can be attributed to interventions of this project,
10
11
12 Evidence for all four of these trends and cycles can be extrapolated from the data, as
13 discussed in Section 3.3.5, and this evidence is an important part of the statistical inferences
14 drawn from the data.
15
16 4.1.1 Description of Data Sets
17 The data sets were provided by the principal investigators of the three studies. These
18 data sets were edited to remove information that would identify the participants. Obvious
19 data errors were detected and either corrected or eliminated, and data discrepancies were
20 resolved by the Environmental Criteria and Assessment Office at Research Triangle Park,
21 NC (ECAO/RTP). For each of the three studies, three files were created, as LOTUS (Lotus
22 1-2-3 v3.2, 1992) spreadsheets and then were exported into SYSTAT (Wilkinson, 1991) data
23 files. The SYSTAT statistical system was used for additional data editing, including the
24 creation of new variables in each data set, such as natural logarithms of lead variables and
25 ages of the children when blood, hand, environmental, or interview samples were collected.
26 Subsets of SYSTAT data files were used as input for structural equation modeling using EQS
27 software (Rentier, 1989).
28 The three data files for each study were intended to be the child (KID), family (FAM),
29 and properly (PROP) files. The KID file was the unit that identified response to lead
30 abatement. The FAM file was the unit that defined the interior dust and paint exposures and
31 socio-demograpnic influences on lead exposure. The PROP file was the unit defined by the
32 soil exposure and abatement data. In Cincinnati, the PROP file was subdivided into PROP
33 and NBHD because most of the soil data could not be identified with a specific property,
34
July 15, 1993 4-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 The project data files were set up as
2
3 (1) a KID file with each child listed as a single record in the data set;
4 ' . ''.'.•
5 (2) a FAM file in which each family was listed as a single record in the data
6 set, including those families with two or more siblings enrolled in the
7 study;
8
9 (3) a PROP file in which each dwelling unit was listed as a single record in
10 the data set; and
11
12 (4) a NHBD file (Cincinnati only) in which data from each soil parcel or
13 exterior dust sampling location were listed as a single record.
14
15
16 4.1.2 Validation of the Data Sets
17 The data sets were validated by reproducing the statistical results reported by the three
18 studies. Only minor modifications were made hi the original data sets for this exercise.
19 In aU instances, the validation exercise reproduced identical numbers of participants.
20 Validations of the six models examined in the Baltimore report found no major
21 discrepancies. The reported and validation results agreed for the regression coefficients and
22 their standard errors.
23 Results for the appropriate hypotheses tests coincide with those in the Boston report.
24 Analyses showed the average declines in blood lead levels between Round 1 and Round 2
25 (before and after abatement) to be significantly different from zero for all three groups. This
26 pattern held throughout the validation of the Boston analyses.
27
28 4.1.3 Modifications in the Project Data Sets
29 4.1.3.1 Restructured Data Sets
30 Certain modifications in the structure of the project data sets were necessary to facilitate
31 the statistical analysis described below. Some households consisted of children from at least
32 two distinct families, and children from each family were in the study, so that there were
33 duplicates of dwelling unit data in some of the family file cases. There were also several
34 cases in which there were two or more apartments in the same building. These were
35 identified as separate premises in the Boston and Cincinnati studies, but could not be
36 distinguished in the data set provided to ECAO/RTP by the Baltimore study group, and so
July 15, 1993 4.3 DRAFT-DO NOT QUOTE OR CITE
-------
1 were condensed to single premises. Most of the analyses reported here were performed
2 using the KID data sets.
3 In Baltimore, soU lead was collected once at each dwelling unit near the beginning of
4 the study, but only at the soil-abated dwelling units in BAL SP after abatement, which
5 occurred between blood lead Rounds 3 and 4. Dust lead was collected only once, before
6 abatement. In Boston, soil lead was collected once near the beginning of the study for all
7 dwelling units and twice after soil abatement to study recontamination. Soil was collected
8 immediately after abatement at Round 2 only for dwelling units in BOS SPI, where soil
9 abatement had been performed. Dust was collected at Rounds 1,3, and 4 for all dwelling
10 units, but at Round 2 only for the dwelling units in BOS SPI and BOS PI, and not for
11 BOS P. This introduced some built-in missing value patterns in the longitudinal design of
12 the study.
13
14 4.1.3.2 Missing Information Procedures
15 The Baltimore study design called for six rounds of blood lead measurements. Because
16 only those children with all six measurements were used in the data analysis, a problem arose
17 with the large number of children who were excluded from the analysis. To explore
18 solutions to this problem, a method, called the Missing Information Principle, was used to
19 fill in the missing data and analyze the entire data set. The method was described by
20 Woodbury and Hasselblad (1970), and further refinements were made by Orchard and
21 Woodbury (1972) and Dempster et al. (1977). The method is now called the E-M algorithm.
22 It starts with any reasonable first estimate of the parameters followed by two steps, called the
23 E and M steps. The E step consists of estimating the sufficient statistics, in this case the
24 sums and sums of squares of cross products. The M step consists of recomputing the
25 estimates of the mean and covariance from the completed sums and sums of squares and
26 cross products.
27 For this analysis, the data set was limited to individuals present for the first round of
28 measurements, but the procedure could be expanded with additional effort. The two
29 assumptions for this analysis were that the blood lead data were missing at random and that
30 the blood lead values were multivariate (log)normal. All observations were transformed
31 logarithmically. Using the E-M algorithm, 20 children were returned to the data set.
July 15, 1993 4-4 DRAFT-DO NOT QUOTE OR CITE
-------
1 A reanalysis of the data showed that Round 1 data are less correlated with the other five
2 rounds than the other five are with themselves. For logistical reasons, this reanalysis was
3 performed on blood lead data not corrected for seasonal cyclic patterns or long-term time
4 trends. When Rounds 2 and 3 were used as the preabaternent blood lead concentration and
5 Rounds 4, 5, and 6 were used as the postabatement blood lead concentration, several
6 observations emerged. First, the change in the group of children in BAL SP that did not
7 receive abatement was not zero. This is consistent with the observation that there were
8 seasonal cyclic patterns and long-term time trends in the blood lead data from all three
9 studies. Second, the estimated change in the blood lead for BAL SP where abatement did
10 occur was 1.1 ^ig/dL per decrease of 1,000 /ig/g lead in soil. Although not statistically
11 significant, this result is consistent with the results of the Boston study.
12 The imputing of missing values in the Baltimore data set was an exercise to see if
13 additional information could be extracted using this technique. Time constraints prevented
14 further statistical analyses that might refine the conclusions. For now, this procedure is
15 treated only as evidence that the impact of intervention is continuous down to low levels of
16 soil lead concentrations, although this impact may be too small to measure below a decrease
17 of 1,000 jig/g.
18
19 4.1.4 Statistical Methods
20 4.1.4.1 Repeated Measures Analysis
21 The common form of the statistical analysis that can be most easily applied to data from
22 all three studies is the form of multivariate analysis of variance or analysis of covariance
23 called repeated measures analysis. In this method, observations on each subject or unit of
24 the study are supposed to occur in an ordered sequence, normally a time series. The subjects
25 are in separate treatment groups or categories that do not change with time. The response
26 variable Y (e.g., log blood Pb) for subject i in group j at time t is denoted by Yyt. The
27 model is then
28
Kf - M + S, + G. + R, + Ajt + E9 (4-1)
29
July 15, 1993 4-5 DRAFT-DO NOT QUOTE OR CITE
-------
1 where
2
3 M is the grand mean over all groups and times,
4 5 is the variability specific to subject i in group j for all times,
5 G is the effect of being in group 7 for all times,
6 R is the effect of being observed at time or round t,
7 A is the effect of being in group j at time t over and above the effect of being
8 in groupy (e.g., abated versus not abated), and
9 E is the random measurement error.
10
11 The effect S is assumed to be random, and the effects G, R, and A are "fixed" effects
12 in that they are not expected to change over time. They are characteristics of the
13 experimental design (i.e., neighborhood and treatment). The average fixed effect is zero.
14 Using an asterisk (*) to denote an appropriate weighted average over the subscript, the
15 assumptions are
16
17 G* = 0 (average across all groups),
18 .R* = 0 (average across all rounds), and
19 Aj* = /4*t = 0 for the null hypothesis (average across groups, average across
20 rounds).
21
22
23 The expected subject effect and the expected measurement error are zero. The sum of
24 squares for testing the hypothesis that G = 0 (no treatment group effect) is that of the
25 between-subject variability 5, whereas the sum of squares for testing that R •= 0 (no time
26 effect, including significant linear trends or seasonal differences) or that A = 0 (there is no
27 interaction between treatment category and time) is that of the within-subject or repeated
28 measurement variability E, Parameter estimation of effects and tests of hypotheses were
29 performed using the repeated measures option in the procedure Multivariate General Linear
30 Hypothesis (MGLH) in the SYSTAT package (Wilkinson, 1991).
31 We were most interested in testing the null hypothesis that the interaction term A = 0,
32 because the treatment or abatement occurs during the course of the study. Even if the
33 treatment groups were equal before abatement, they are not expected, to be equal after
34 abatement.
35 A similar model can be used when there are numeric covariates available (here denoted
36 generically Xy in Equation 4-2). In the standard repeated measures model, the covariates
July 15, 1993 4-6 DRAFT-DO NOT QUOTE OR CITE
-------
1 may be different for each subject, but pertain to the whole study and should not change with
2 time. An alternative specification for use with time-dependent covariates is shown in the
3 next section. The general form of the repeated measures analysis of covariance model is
4
7f-M+S9 + GJ+Kt+Af+ b( Xy + E9 (4-2)
5
6 where
7
8 bt is the response coefficient (slope), and
9
10 Xy is the coviariate (e.g., log soil Pb).
11
12 The hypothesis tests for regression coefficients are similar to the tests for between-subject
13 and within-subject variation.
14 A problem arises if the response variable Y must be transformed, say by a logarithmic
15 transformation for blood lead or for hand lead, in order to reduce skewness and to stabilize
16 variances across treatment groups. The implied model for the original untransformed
17 variable is then multiplicative in treatment effects and random variation. This is probably
18 acceptable for the analysis of variance (ANOVA) model in Equation 4-1, but is likely to
19 produce a physically or biologically meaningless specification for the covariate model in
20 Equation 4-2 when the covariates are indicators of distinct and additive sources of lead, such
21 as soil lead and interior lead-based paint. For this reason, we also evaluated an
22 autoregressive regression model specification as an alternative approach to dealing with the
23 longitudinal design of the study.
24
25 4.1,4,2 Autoregressive Regression Models
26 Biokinetic Basis for an Autoregressive Blood Lead Model
27 Lead in blood is stored in various body tissues and organs for varying amounts of time.
28 The red blood cells retain lead for a few days, the soft tissues such as the kidney retain lead
29 for some weeks or months, and the skeletal tissues may retain lead for years or even
30 decades. We will not consider multicompartmental models for blood lead in this report,
31 except to note that the apparent blood lead half-life or the whole-body lead retention time
32 may appear to depend on the interval between blood lead samples. The blood lead
July 15, 1993 4-7 DRAFT-DO NOT QUOTE OR CITE
-------
1 concentration (denoted PbB, measured in micrograms of lead per deciliter of whole blood)
2 will depend dynamically on the absorbed lead uptake rate (denoted U, measured in
3 micrograms of lead per deciliter of whole blood per day) and on the mean blood lead
4 residence time (denoted T, measured in days). A first-order differential equation
5 specification may be assumed as a reasonable approximation for low to moderate levels of
6 exposure:
(4-3)
dt T
7
8 Equation 4-3 may be solved explicitly for constant intake rate, U:
PbB(f) = PbB(ty e-'IT + UT [l - e~"T] (4~4)
9
10 Equation 4-4 has an important implication for models of blood lead observed over time. The
11 present blood lead, PbB(f), depends on the preceding blood lead, PbB(G), but the
12 proportionality is not constant. In fact, the relationship is decreasing with increasing time
13 interval t. Thus, it is not appropriate to use the difference in blood lead levels between
14 successive sampling times as an index of effectiveness of abatement. In fact, Equation 4-4
15 implies that
PbB(f) - PbB(0) = \TJT- PbB(0)] [1 - exp(-r/7)] (4-5)
16
17
18 Thus, the change in blood lead over time, for short periods of time, must depend on the
19 starting blood lead PfoB(0)» The actual dependence is more complicated because lead has
20 multicompartmental biokinetics instead of the one-compartment kinetics of Equation 4-3 .
21 After a long period of time, the information in starting blood lead PbB(G) is negligible and
22 the current blood lead PbB(t) reflects only lead uptake U, which is approximately a linear
23 function of the environmental lead levels.
24
25 Fitting an Alternative Autoregressive Regression Model
26 The model implied by Equation 4-4 can be estimated using nonlinear regression
27 techniques. If we let PbB, PbH, PbS, PhD, and PbP, denote levels of lead in blood, hands,
July 15, 1993 4-8 0RAFT-DO NOT QUOTE OR CITE
-------
1 soil, dust, and paint, respectively, then a generic form of an autoregressive regression model
2 is
3
4 PbB(f) = PbB(t-\) r(t) + (b^t) + PbH(f) + b2(t} PbS(t)
5 + b3(t) PbD(f) + b4(f)PbP(t) + ... (4-6)
6
7 As noted above, log-transforming blood lead to reduce skewness and stabilize variances
8 requires another specification for model parameter estimation:
9
10 log[PbB(t)] = log[P2>fi(M) r(f) + b^t) PbH(f) + b2(f) PbS(f)
11 + b3(t)PbD(f) + b4(t)PbP(f) + ...] (4-7)
12
13 This was fitted using the procedure NON1IN in the SYSTAT package (Wilkinson, 1991).
"14 The models fitted here show that there is a dose-response of blood lead to environmental
15 lead. The effectiveness of abatement is thus reduced to testing the effectiveness of the
16 abatement in reducing environmental lead. If the postabatement regression coefficients, b(t),
17 are not significant, then there is no persistent effect of abatement that is not characterized by
18 the autoregressive coefficient, r(f). Even if b(t~) is not significant, but r(f) for the abatement
19 group is significantly smaller than r(r) for the nonabatement group, then there is a persistent
20 long-term effect of abatement.
21
22 4.1.4.3 Structural Equations Modeling
23 The most complete and technically correct evaluation of these studies requires a
24 simultaneous assessment of changes in blood lead levels and changes in environmental lead
25 pathways following soil lead and/or dust lead abatement. Underlying any analysis of time-
26 dependent relationships are the following assumptions:
27 - .
28 (1) Both preabatement and postabatement blood lead levels reflect, in part,
29 contemporary environmental lead exposures that can be characterized by
30 measurements of lead levels in soil, dust, paint and other media;
31
My 15, 1993 4.9 DRAFT-DO NOT QUOTE OR CITE
-------
1 (2) Postabateraent blood lead levels may also reflect, in part, preabatement blood
2 lead levels due to the contribution of preabatement body burdens of lead
3 (principally in the skeleton) from earlier exposures;
4
5 (3) Postabatement dust lead levels may also reflect, in part, preabatement dust lead
6 levels due to mixing of incompletely abated or unidentified sources of lead in
7 dust for which preabatement dust lead levels are a surrogate indicator;
8
9 (4) Postabatement soil lead levels may also reflect, in part, preabatement soil lead
10 levels due to mixing of incompletely abated or unidentified sources of lead in soil
11 for which preabatement soil lead levels are a surrogate indicator;
12
13 (5) Even when lead-based paint has been stabilized, lead paint levels measured by
14 XRF may also help to predict postabatement soil and dust lead levels from
15 incompletely abated or unidentified sources of lead in soil and dust for which
16 lead-based paint levels are a surrogate indicator.
17
18 Based on these assumptions, there are several testable hypotheses that can be
19 formulated:
20
21 HYPOTHESIS 1. Postabatement blood lead levels in the abatement group(s) are
22 relatively lower than blood lead levels in nonabatement or low-impact abatement
23 group(s), after adjustment for current environmental lead exposures, due to a persistent
24 effect of the abatement in reducing body burden;
25
26 HYPOTHESIS 2. Postabatement blood lead levels in the abatement group(s) are not
27 necessarily lower than blood lead levels in nonabatement or low-impact abatement
28 group(s), after adjustment for current environmental lead exposures, but are lower in
29 the abatement groups for which there has been a persistent reduction in soil and dust
30 lead exposure levels.
31
32 An explicit parametric model for these hypotheses is based on the following relationship
33 between preabatement and postabatement blood lead levels for each subject:
34
35 PbBlood(Post) = R * PbBlood(Pre) + B0 + Bt * PbSoil(Post) + ... (4-1)
36
37 The autoregression coefficient R in Equation A includes such factors as the exponential
38 washout or biologic elimination of lead from the body over time. R also includes other time
39 effects such as seasonal variations and a general reduction in lead exposure from diet and
July 15, 1993 4-10 DRAFT-DO NOT QUOTE OR CITE
-------
1 airborne sources. The valid use of Equation A requires that pre- and postabatement
2 measurements be as nearly synchronous as possible. All three studies have preabatement and
3 postabatement blood lead measurements taken during the summer or early fall, about 10 to
4 12 mo apart. These can be used as a basis for comparison among the three studies, without
5 requiring substantial adjustments for seasonality and time trends.
6 A more complete analysis of these data would also use the measurements made
7 immediately after abatement, and longer-term measurements assessing the extent and
8 response to environmental recontamination, and aging of the subjects. While tune series
9 models are usually expressed in terms of autoregression of residuals, the lack of synchrony
10 between blood lead samples and environmental samples makes such analyses less useful in
11 the present situation.
12 The models shown here represent the best models we obtained by backward elimination
13 of nonsignificant predictor variables from the most complete model, including those predictor
14 variables for blood lead that were predicted pathway variables. Data were analyzed using the
15 EQS program (Bender, 1991) with GLS (generalized least squares) or AGLS (asymptotically
16 distribution-free generalized least squares) methods. Cases with missing values of model
17 variables were eliminated, and no effort was made at this time to impute any of the missing
18 values. We tested model specifications with combinations of dust lead concentrations and
19 loadings. All of the lead pathway equations may be developed by writing down linear
20 models implied by the pathway diagrams. Hand lead was not used at this time, due to the
21 complexity of the longitudinal models.
22 Negative coefficients in the model are physically or biologically uninterpretable, and
23 coefficients were constrained to be nonnegative. Most of the zero-constrained coefficients
24 were associated with variables eliminated from the model, and the remaining zero-
25 constrained coefficients should not seriously affect the predictions relative to a
26 nonconstrained set of coefficients. However, since all of the coefficients were constrained to
27 be nonnegative, one-tailed statistical tests should be used in interpreting the results.
28
29
30
July 15, 1993 4-11 DRAFT-DO NOT QUOTE OR CITE
-------
1 4.1.5 Limitations of the Statistical Methods
2 The, statistical methods we used here were reasonably appropriate and could be used by
3 many other investigators with access to standard statistical software packages. However, the
4 methods have certain limitations that should be understood. The repeated measures analyses
5 assume only that the response variables are correlated with each other, with no implication of
6 temporal causality. The goodness of Fit of the models was significantly improved by use of
7 covariate analyses. The usual repeated measures analyses require that the covariates have no
8 time dependence. This is appropriate for exposure covariates such as lead-based paint levels,
9 which were not changed during the course of the study. It could also be used for dust lead
10 levels in the Baltimore study that were only measured at one time, although there is a high
11 likelihood that the dust lead levels were changing during the course of the study. The
12 availability of environmental data to characterize time-varying lead exposures in the Boston
13 and Cincinnati studies suggests that more powerful statistical methods, such as structural
14 equation models, would be more appropriate.
15 It is also important to understand that the repeated measures model applied to log(Blood
16 lead) implies that blood lead levels are multiplicative functions of their main factors.
17 By using a response variable,
18
19 Y = log(Blood lead)
20
21 and exponentiating Equation 4-1 or Equation 4-2, we obtain
22
Blood leadijt = eu x es< x e°' x eR> x eA> x eE» C4'8)
23
24 In this model, e is the geometric mean blood lead over all groups, treatments, rounds, and
25 subjects. For the covariate model in Equation 4-2, if we used
26
27 Xy = log(Environmental lead)^
28
29 as we did in Chapter 3, and exponentiate Equation 4-2, we obtain
30
July 15, 1993 4-12 DRAFT-DO NOT QUOTE OR CITE
-------
Blood leadyt = eM x es x eR> x eAf x (EnvironmentalPb)ijt e** (4^9)
1
2 This model does not reproduce the additive model we believe is more appropriate.
3 The autoregressive regression model is closer to the basic biokinetics of lead exposure.
4 However, the model implied by Equation 4-4 is only partially correct since the exposure
5 variables embraced in the U term are also varying over time. There are secular trends, such
6 as steadily decreasing exposures from air lead and from lead in food cans. There are
7 seasonal variations that may be multiplicative in nature. It would be preferable to fit a
8 parametric form of the model. For example, to include a multiplicative seasonal variation
9 with relative amplitude denoted a and phase angle denoted F, we would use
10
11 log[PZ>B(r)] = log[PZ>B(M) r(f) + b^t) PbH(f) +
12 b2(t) PbS(f) + b3(f) PbD(f) + b4(f) PbP(f) '+ ...] +
13 log[l + a x cos(r/58.131 + F)] (4-10)
14 ,
15 We have used 58.131 as the number of days in a year of 365.25 days that are equivalent to a
16 radian of angle. When tune-varying covariates are not available for important predictors,
17 such as the lack of hand lead levels in the data set for Cincinnati that was provided to us,
18 then it may be more informative to use a model in which the relative reduction in blood leads
19 postabatement could be used in the form
20
21 log[PWJy(r)] = log[P&By(f - 1) rj(f) + b2(f) PbDy(i) + ...] (4-11)
22
23 where the autoregression coefficients rfi) may depend on both group identifier j and time
24 (round) t. The model in Equation 4-11 provided a better fit than the covariate-adjusted group
25 means autoregressive model,
26
27 log[P%(0] = log[PbBy(t - 1) r(t) + gj(t) + b^(f) PbD^f) + ...] (4-12)
28
29 where the intercept terms denoted Cj(f) may depend on both group identifier j and time
30 . (round) r, but the autoregression term r(f) is the same for all groups.
July 15, 1993 4-13 DRAFT-DO NOT QUOTE OR CITE
-------
1 Extension of repeated measures analyses to covariates such as environmental lead levels
2 that change with time, and extension of the autoregressive regression models to simultaneous
3 estimation of parameters at multiple time points where the output of one regression equation
4 is used as input for the next, can be done using a single technique, structural equation
5 modeling. These analyses are more powerful and general diagnostic tools.
6
7
8 4.2 IMPACT OF INTERVENTION
9 4.2.1 Impact of Soil Abatement on Exterior and Interior Dust
10 Exterior dust was measured and abated in Cincinnati only. In this study, the results
11 suggest a recontamination rate for exterior dust of less than 2 weeks, and that the source of
12 this recontamination is not the soil. Additional measurements have been made to identify the
13 source and rate of recontamination, but no data are available at this time. The source may
14 be a single episode (possibly external), a seasonal effect such as dry weather, or a continuous
15 process possibly related to traffic. With a neighborhood level perturbance of this type, it is
16 not possible to measure the impact of soil abatement on house dust directly. However, if
17 abatement is considered on the broader scope, where neighborhood cleanup would include
18 soil, external dust, and any other sources of lead external to the home, then the house dust
19 measurements made immediately inside the homes can be used as a measure of this "total
20 neighborhood abatement". For those cases in the Cincinnati study where there was no
21 immediate recontamination of this entryway dust, this measurement can be used as a
22 surrogate for soil abatement. To make this determination, it is also necessary to evaluate the
23 fraction of exposure that would derive directly from soil or from playground dust, which
24 would not be included in the interpretation of house dust alone.
25 The key to understanding the impact of soil (and external dust) abatement on interior
26 dust is to observe changes in the three components of the interior dust measurement: lead
27 concentration (micrograms of lead per gram of dust), lead loading (micrograms of lead per
28 square meter), and dust loading (milligrams of dust per square meter). Where there was no
29 interior dust abatement, the lead concentration in interior dust should decrease gradually over
30 time, provided that the influence of lead-based paint has been minimized. Also, the lead
31 loading should decrease if the dust loading remains constant or the lead loading is normalized
July 15, 1993 4-14 DRAFT-DO NOT QUOTE OR CITE
-------
1 to dust loading. This normalization is believed to correct for differences in housekeeping
2 efficiency. If interior dust abatement has occurred, the lead concentration should decrease
3 markedly and remain low where the influence of lead-based paint is minimal, and the lead
4 loading and dust loading should decrease and then increase in tandem.
5 The impact of lead-based paint can be minimized in three ways: (1) observe only cases
6 where there is no lead-based paint; (2) stabilize the paint so that the rate of incorporation to
7 house dust is minimized; and (3) compare measurements where the influence of lead-based
8 paint is probably high, such as window wells, to areas where the influence of soil is high,
9 such as entryways. A crude measure of the rate of recontamination of house dust from lead-
10 based paint can be observed from the changes in window well dust lead concentrations
11 following interior dust abatement, for units with and without lead-based paint.
12 The analysis of three types of internal dust measurements, (1) entry, (2) floor, and
13 (3) window well, can provide additional information about the impact of soil abatement. The
14 entry measurement probably shows the greatest influence of exterior lead from soil and dust.
15 If the entryway to the housing unit is somewhat removed from the building entrance, such as
16 an apartment on the second or third floor, then a comparison of these two measurements
17 should demonstrate the effect of soil lead on multifamily houses. Likewise, where interior
18 dust abatement has taken place, the rate of recontamination of interior dust should be
19 entry > floor > window well.
20 f>.
21 4.2.2 Impact of Soil and Dust Abatement on Hand Lead Loading
22 It was expected that hand dust would serve as an surrogate measure of changes in
23 exposure following abatement to augment information about blood lead changes. Hand dust
24 reflects the child's recent exposure (since the latest hand washing), but is a measure only of
25 lead loading, not lead concentration or dust loading, because the total amount of dust is not
26 measured. Consequently, it is not possible to estimate the source of lead (soil or paint) by
27 differences in concentration, nor is it possible to correct for housekeeping effectiveness by
28 observing changes in dust loading, as with house dust. It is believed that the amount of dust
29 (not mud or dirt) on the hand reached equilibrium after a short period of time, perhaps
30 30 min to 2 h. The dustiness of the house would affect only the rate at which this
31 equilibrium is reached, riot the total amount of dust at equilibrium.
My 15, 1993 : 445 DRAFT-DO NOT QUOTE OR CITE
-------
1 The hand dust measurements in this report should be viewed with caution because of
2 the analytical difficulties discussed earlier and in the individual study reports. Nevertheless,
3 both the Boston and Cincinnati studies showed a reduction in lead loading on the hands
4 following interior dust abatement, but very little response to soil abatement.
5
6 4.2.3 Impact of Soil and Dust Abatement on Blood Lead Concentrations
7 Blood lead concentrations should respond to soil and dust abatement through the impact
8 of abatement on two routes of exposure: (1) hand-to-mouth activity, reflecting the impact of
9 interior house dust and exterior play area dust on exposure; and (2) food contamination,
10 reflecting the incorporation of house dust in food during kitchen preparation. There was no
11 measure of the incorporation of house dust into food during this project. Intuitively, the
12 impact of interior dust abatement should be the same, or at least comparable., for food and
13 hand dust. Li some homes, however, lead-based paint is more common in kitchens and
14 bathrooms, and the rate of return of lead-based paint following stabilization would have a
15 greater impact on food than hand dust. There is a limited amount of data, not yet analyzed,
16 where kitchen floor dust can be compared to bedrooms and other living areas, and likewise
17 for window wells. Most of these data, however, are from the Cincinnati study, where there
18 was a minimum influence of lead-based paint.
19 The Boston study showed a small but statistically significant effect of soil abatement on
20 blood lead. This is expressed as a 1 to 1.5 jtg/dL decrease in blood lead per 1,000 jtg/g
21 decrease of lead in soil. Although a greater effect was expected, these findings are not
22 surprising for two reasons. First, the earlier studies that predicted a greater effect were not
23 based on a reduction in exposure but on extrapolations from data of cross-sectional studies of
24 children with different ranges of exposure to soil lead. Second, the measurement of soil lead
25 in these earlier studies was of a distinctly lower quality than in this project. Fewer samples
26 were taken, little effort was made to take representative samples, and no reference materials
27 were available to standardize analytical procedures. In these earlier studies and the Boston
28 Study, there was no attempt to measure the impact of neighborhood-level exposure.
29 The Baltimore study showed no influence of soil abatement on blood lead
30 concentrations. The Baltimore study did not measure the impact of soil abatement in the
31 absence of interior lead-based paint, and it is possible that soil abatement would be swamped
July 15, 1993 4-16 DRAFT-DO NOT QUOTE OR CITE
-------
1 by the presence of paint lead in the house dust. This negative result is an important finding
2 of this study and the integrated project that suggests, in the absence of interior dust
3 abatement and interior paint stabilization (or abatement), soil, exterior dust, and exterior
4 paint abatement will have little impact on childhood lead exposure.
5 The Cincinnati study showed no effect of soil abatement alone on the blood lead
6 concentrations, but showed a positive effect of interior dust abatement and a marginal effect
7 of total abatement when the interior-entry dust immediately inside the home was used as a
8 surrogate of neighborhood lead abatement. The importance of these findings is that when the
9 sources of lead that recontaminate exterior dust can be identified and abated, the impact of
10 neighborhood-level abatement will be greater than single dwelling abatement alone,
11
12 . .
13 4.3 RESULTS OF STATISTICAL ANALYSES
14 4.3.1 Baltimore Study
15 The repeated measures analysis of variance is shown in Table 4-1. The main effects of
16 category (Control in Area 1, Control in Area 2, Abatement in Area 2) are nonsignificant in
17 Table 4-1 (p = 0.18), due to the large between-subjects variance MS = 1.17. There is a
18 large difference within subjects among Rounds 1 through 6, with p < 0.000001. The round
19 by category interaction (soil abatement effects) are very small, none exceeding 0.039 (4%
20 difference in blood lead) and show no consistent pattern of pre-Round 3 versus post-Round 3
21 direction. The soil abatement effects appear to be random, and this is borne out by the
22 results in Table 4-1, where p = 0.8886 » 0.05.
23 Attempts to refine the analysis by use of covariates is shown in Table 4-2. This
24 reduces the category effects overall, with a corresponding nonsignificant p = 0.86 in
25 Table 4-2. The initial interior dust lead loading is highly predictive of blood lead levels
26 (p = 0.0079), and the interior lead-based paint level is still marginally significant even after
27 interior dust loading is considered (p = 0.0598). However, the round by category
28 interaction is still nonsignificant (p = 0.59), so that soil abatement apparently had little effect
29 on blood lead levels.
July 15, 1993 4-17 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 4-1. REPEATED MEASURES ANALYSIS OF VARIANCE FOR LOG
(BLOOD LEAD) FOR BALTIMORE STUDY, ANALYSIS OF VARIANCE TABLES
Effect
MS
df
P
Between Subjects
Category
Variability
2.0305
1.1729
2
139
0.1809
Within Subjects
Round
*
Round Category
Variability
2.0587
0.0253
0.0504
TABLE 4-2. REPEATED MEASURES ANALYSIS
(BLOOD LEAD)
Effect
5
105
695
<10"6
0.8886
OF COVARJANCE FOR LOG
FOR BALTIMORE STUDY, ANCOVA
MS
df
TABLES
P
Between Subjects
Category
Log(XREB)
Log (XRM)
Log (PbLD-AA)
Variability
0.1579
1.5108
3.8463,
7.7800
1.0629
2
1
1
1
108
0.8621
0,2358
0.0598
0.0079
Within Subjects
Round
Round X Category
Round X Log (XRFE)
Round X Log (XRM)
Round X Log (PbDL-AA)
Variability
0.2487
0.0396
0.0246
0.0831
0.0222
0.0473
5
10
5
5
5
540
0.0001
0.5926
0.7613
0.1201
0.7999
July 15, 1993
4-18
DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
8
9
10
11
This hypothesis is further developed in Tables 4-3 and 4-4. In Table 4-3, there is a
statistically significant relationship between hand lead and blood lead (one-tailed p < 0.05)
for Rounds 1, 2, 4, and 6, and a marginally significant relationship (one-tailed p < 0.10) for
Round 3. A more detailed examination in Table 4-4 shows that there is a significant
relationship between blood lead and at least one of the variables (hand lead, dust lead, or
interior paint lead) preabatement at Rounds 1,2, and 3. Table 4-5 shows that this also
occurs in Round 4 for both control and soil abatement groups, and in Round 6 for the soil
abatement group, but not in Round 5 or in the Round 6 control group. Soil lead is a
significant predictor of blood lead in Round 2, and in the soil abatement group at Rounds 4
and 6.
TABLE 4-3. AUTOREGRESSIVE REGRESSIO
ON HAND LEAD, FlTTJttU IN LOG FORM,
>N MODEL FOR BLOOD LEAD
FOR BALTIMORE STUDY3
Round
Variable 1 2 3
Intercept 8.84b 2,50b 1.95b
(/ig/dL) (0.45) (0,31) (0.45)
Autoregression NE 0,662b 0.685b
(0,029 (0.048)
Hand Lead 202.0b 13.7b 38.7
(Mg/pair x 103) (41.1) (8,3) (25.8)
N 408 307 212
Residual GSD 1.666 1.323 1.380
456
1.12b 0.44 0.85b
(0.36) (0.27) (0.27)
0.693b 1.024b 0.884b
(0.044) (0.042) (0.037)
61.7b -1.3 29.8b
(27.5) (12.7) (13.6)
193 192 184
1.336 1.256 1.235
Note: Soil abatement occurred between Rounds 3 and 4.
a-
'NE = not estimated; values not in parentheses are model estimate parameter, standard error is shown in
parentheses, GSD = geometric standard deviation.
Statistically significant positive effect, one-tailed p < 0.05.
1 It appears then that children with higher exposures to interior dust or interior lead-
2 based paint will more consistently have higher blood leads hi these two neighborhoods than
3 will children with exposure to exposure to elevated soil lead levels, and that it may be
4 necessary to directly intervene to reduce dust and paint lead levels in these houses. Although
5 exterior lead-based paint was correlated with soil lead, the addition of an exterior X-ray
July 15, 1993
4-19
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 4-4. AUTOREGRESSIVE REGRESSION MODEL FOR BLOOD LEAD
ON ENVIRONMENTAL LEAD, FITTED IN LOG FORM, FOR
BALTIMORE STUDY, PREABATEMENT3
Variable
Intercept
frgfdL)
Autoregression
Hand Lead
(jtg/pair x 103)
Soil Lead
(/ig/g x 103)
Dust Lead
(/ig/m2 X 103)
Exterior Paint
(mg/cm2)
Interior Paint
(mg/cm2)
Round 1
7.77b
(1.19)
NE
45.8
(53.9)
1.48
(1.83)
4.04b
(2.38)
0.066
(0.068)
0.442
(0.276)
Round 2
0.69
(0.72)
0.720b
(0.040)
0.0
(NE)
1.84b
(0.82)
1.04
(0.89)
0.0
(NE)
0.196
(0.120)
Round 3
1.38b
(0.53)
0.675b
(0.060)
67. 8b
(31.8)
0.0
(NE)
0.0
(NE)
0.0
(NE)
0.147
(0.157)
*NE «= not estimated, values not in parentheses are the model estimate parameter, the standard error is shown
in parentheses.
One-tailed p < 0.05; statistically significant positive effect.
1 fluorescence term to the model did not significantly improved the ability to estimate blood
2 lead levels. This does not imply that there is no benefit to abatement of soil or exterior lead-
3 based paint, however, because there clearly was a relationship between postabatement soil
4 lead levels and blood lead that may indicate that reducing soil lead has reduced the soil
5 component of interior dust lead loading.
6 The model adopted for Baltimore is shown in Figure 4-1 and results are shown in
7 Table 6. The BAL SP group needed to be split into separate components. The nonabated
8 units in BAL SP (now denoted BAL P-2) were similar in response, and these differed from
9 the abated units in BAL SP (still denoted BAL SP). We therefore show three groups in our
10 analyses. The models included blood lead levels at Rounds 3 and 4, environmental lead
11 levels at Round 1, and postabatement soil lead levels in group BAL SP after Round 4
12 (February 1990 and January 1991, respectively).
July 15, 1993 4-20 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 4-5. AUTOREGRESSIVE REGRESSION MODEL FOR BLOOD LEAD
ON ENVIRONMENTAL LEAD, FITTED IN LOG FORM, FOR BALTIMORE
STUDY, POSTABATEMENT3
Round
Variable
Intercept
(pg/dL)
Autoregression
Hand Lead
Gig/pair x 103)
Soil Lead
(pg/g X 103)
Dust Lead
G*g/m2 X 103)
Exterior Paint
(mg/cm2)
Interior Paint
o
(mg/cm )
N
Residual GSD
4C
1.18b
(0.50)
0.646b
(0:069
112.2b
(59.3)
0.
(NE)
0.88
(1.26)
0.
(NE)
0.035
(0.144)
82
1.308
4S
0.40
(0.83)
0.693b
(0.059)
45.4
(33.6)
13.18b
(6.06
2.55b
(1.33)
0.
(NE)
0.
(NE)
83
1.318
5C
-0.49
(0.38)
1.063b
(0.048)
0.6
(1.3)
0.56
(0.45)
0.
(NE)
0.
(NE)
0.
(NE)
96
1.191
5S
-0.05
(0.56)
1.071b
(0.071)
0.
(NE)
0.
(NE)
1.11
(1.21)
0.064
(0.060)
0.
(NE)
81
1.265
6C
0.34
(0.44)
0.912b
(0.063)
25.4
(20.4)
0.
(NE)
0.
(NE)
0.
(NE)
0.190
(0.124)
85
1.277
6S
0.66
(0.39)
0.822b
(.046)
27.9
(18.3)
8.16b
(4.15)
0.
(NE)
0.
(NE)
0.295b
(.103)
74
1.167
aNE = not estimated, values not in parentheses are the model estimate parameter, the standard error is shown
in parentheses, GSD = geometric standard deviation.
One-tailed p < 0.05; statistically significant positive effect.
1 Floor lead concentrations and floor lead loadings were not as consistently associated
2 with blood lead levels as in the Boston study, whether measured by AAS or XRF. We
3 developed a model that used preabatement dust lead concentration as an indicator of lead
4 indoor exposure. Interior and exterior XRF levels were as indicators of lead-based paint, for
5 direct exposure and as source terms for soil and dust lead.
6 Unlike Boston, the abatement or nonabatement groups corresponded to spatially distinct
7 neighborhoods that also differed in some important preabatement characteristics. We must
8 therefore allow for the possibility that the primary environmental pathways, and the
July 15, 1993
4-21
DRAFT-DO NOT QUOTE OR CITE
-------
XRF
Interior
0
0.46
NE
0.12
0.0
0.39
0
0.46
NE
XRF
Exterior
12.9
4.2
0.0
Pb Blood 3
2.5
0.0
3.8
0.04
0.03
0.74
Pb Dust 1
2.5
8.1
0
Pb Soil 1
Pb Blood 4
10
BALSP
Abatement
Only
Pb Soil 4
figure 4-1. Structural equation diagram for the Balitmore study. Numbers next to
arrow, from Table 4-6, are the regression coefficients (top to bottom) for
BAL SP, BAL P-l, and BAL P-2.
1 effectiveness of the abatements in interdicting those pathways, may differ among different
2 neighborhoods.
3 We found strong relationships between preabatement blood lead level and floor dust
4 lead concentration in BAL P-2, and postabatement between blood lead level and soil lead
5 concentration in BAL P-2. Exterior lead-based paint contributed to soil lead levels both
6 preabatement in BAL SP (significant) and BAL P (positive but nonsignificant), but not in
7 BAL P-2. Interior lead-based paint contributed significantly to floor dust lead preabatement
8 in BAL SP, but not in BAL P.
9 There was at least some reasonable consistency in estimates of the blood lead
10 autoregression coefficient R. The estimates decreased from 0.850 in BAL P to 0,720 in
July 15, 1993
4-22
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 4-6. REGRESSION COEFFICIENTS FOR BALTIMORE STRUCTURAL
EQUATIONS MODEL, USING THE GLS METHOD
Response Variable (Output)
Preabatement
Predictor Variable (Input)
Preabatement Blood
Preabatement Dust Cone.
Pre/Postabatement Soil Cone.
Mean XRF Exterior
Pb-Based Paint
Mean XRF Interior
Pb-Based Paint
Group
BALSP
BALP-1
BALP-2
BALSP
BALP-1
BALP-2
BALSP
BAL P-l
BALP-2
BALSP
BALP-1
BALP-2
BALSP
BAL P-l
BALP-2
Blood
0.04
0.03
0.74a
2.53
0.0
3.79
0.12
0.0
0.39
0.0
0.46
ME
Dust Cone. Soil Cone.
2.50
8.11
0.0
12.86C
4.19
0.0
628C "
0.0
NE
Postabatement
Blood
0.72a
0.85a
0.57 la
1
10.00b
0.0
, 5.01a
Significance level < 0.001.
Postabatement soil lead used in this group.
cSignificance level 0.01 to 0.05 one-tailed (0.02 to 0.10 two-tailed).
1
2
3
4
5
6
7
,8
1
2
BAL SP, but to 0.571 in BAL P-2. The difference in R between BAL SP and BAL P-l was
only marginally significant in testing Hypothesis 1. (Hypothesis 1 asserts that R for
nonabatement is greater than for effective abatement, so a one-tailed test is appropriate).
In summary, the relative decrease in blood lead levels between pre- and postabatement
samples was not significantly larger in the soil abatement area BAL SP than in the
designated nonabatement area BAL P, after adjusting for current exposure. Effects of the
abatements could only be detected after adjusting for changes in postabatement lead
exposure.
July 15, 1993
4-23
DRAFT-DO NOT QUOTE OR CITE
-------
1 4.3.2 Boston Study
2 The results for the Boston study are shown in Tables 4-7 and 4-8. The analysis of
3 variance shown in Table 4-7 shows no significant difference among categories. However,
4 there is a very significant difference among Rounds 1,3, and 4 overall. The round by
5 category interaction effects are statistically significant (p = 0.0020) as shown in Table 4-7.
6
TABLE 4-7, REPEATED MEASURES ANALYSIS OF VARIANCE FOR LOG
(BLOOD LEAD) FOR BOSTON STUDY, ANALYSIS OF VARIANCE TABLES
Effect
MS
df
P
Between Subjects
Category
Variability
0.0493
0.3626
2
143
0,8729
Within-Subjects
Round
Round X Category
Variability
3.3324
0.2196
0.0505
2
4
286
<10"6
0.0020
1 The effectiveness of abatement is explored in more detail in the analysis of covariance
2 in Table 4-8. Initial soil lead and dust lead levels in Table 4-8 show that the effects are not
3 statistically significant. The lead paint variables showed even less predictiveness and are not
4 shown here. The abatement interaction terms are about equally significant as shown in
5 I Table 4-8 (p = 0.0024). There is some marginal significance for the initial dust lead term
6 (p = 0.0787) and for an interaction between initial soil lead and abatement effect as
^ characterized by the round by soil lead interaction (p = 0.0664). Therefore, there is
8 evidence for an effect of soil lead abatement on blood lead concentrations that depends—not
9 surprisingly—on the initial soil and dust lead levels. Because soil lead and interior dust lead
10 are highly correlated, we may infer that, in these Boston houses, exposure to dust lead (as a
11 primary vector) that was derived from soil lead can be reduced over time by a combined
12 removal of soil and dust lead, but not by dust lead abatement alone.
July 15, 1993 4-24 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 4-8. REPEATED MEASURES ANALYSIS OF COVARIANCE FOR LOG
(BLOOD LEAD) FOR BOSTON STUDY, ANALYSIS OF COVARIANCE TABLES
Effect
MS
df
P
Between Subjects
Category
Log [PbS(l)]
Log [PbLDF(l)]
Variability
0.1494
0.5899
1.0640
0.3387
2
1
1
130
0.6442
, 0.1893
0.0787
Within Subjects
Round
Round X Category
Round X Log [PbS(l)]
Round X Log [PbLDF(l)]
Variability
0.2255
0.2144
0.1381
0.0510
0.0504
2
4
2
2
260
0.0123
0.0024
0.0664
0.3647
1 Additional analyses are shown in Table 4-9. An autoregressive regression model using
2 hand lead as a surrogate exposure variable performs very well, with residual geometric
3 standard deviations (GSD) of 1.33 to 1.38 that are nearly as small as any seen in regression
4 models of cross-sectional studies. There is a large and highly significant autoregression
5 coefficient for Round 4 blood leads on Round 3 blood leads taken 2 to 3 mo earlier (adjusted
6 for hand lead) of 0.670, which captures the longitudinal course during the recontamination
7 phase. This is very similar to the autoregression coefficient of 0.683 for postabatement
8 blood lead at Round 3 on preabatement blood lead at Round 1.
9 Assessment of the predictiveness of other environmental lead variables is shown in
10 Table 4-10. Hand lead is again predictive of blood lead, with coefficients that are not much
11 different than those given in Table 4-9. Soil lead levels, including postabatement values,
12 never made a significant contribution to increasing blood levels when hand lead and dust lead
13 loadings were included in the model. Postabatement dust lead loadings were highly
14 significant predictors of postabatement blood lead. There were significant reductions in
15 blood lead between Rounds 1 and 3, with decreases of geometric mean blood lead relative to
July 15, 1993 4-25 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 4-9. AUTOREGRESSIVE REGRESSION MODEL FOR BLOOD LEAD
ON HAND LEAD, PITTED IN LOG FORM, FOR BOSTON STUDY3
Variable
Intercept
fcg/dL)
Autoregression
Hand Lead
(jig/pair x 103)
N
Residual GSD
Round 1
9.74b
(0.73)
NE
148. 8b
(49.3)
150
1.377
Round 3
-0.89
(0.72)
0.683b
(0.061)
107.4b
(35.0)
146
1.362
Round 4
4.26b
(0.61)
0.670b
(0.065)
' 1.56b
(18.4)
143
1.333
*NE = not estimated, values not in parentheses ate the model estimate parameter, the standard error is shown
in parentheses, GSD = geometric standard deviation.
One-tailed p < 0.05; statistically significant positive effect.
Note: Abatement occurred between Rounds 1 and 3.
TABLE 4-10. AUTOREGRESSIVE REGRESSION MODEL FOR BLOOD LEAD ON
ENVIRONMENTAL LEAD, FITTED IN LOG FORM, FOR BOSTON STUDY3
Variable
Intercept
(Mg/dL)
Autoregression
Hand Lead
(jig/pair x 103)
Soil Lead
(Mg/g X 103)
Dust Lead
Gig/m2 x 103)
Dust Abatement
Area Int. 0*g/dL)
Soil Abatement
Area Int. Qig/dL)
N
Residual GSD
Round 1
8.86b
(0.89)
NE
192.2b
(52.8)
0.
(NE)
0.50
(0.39)
-0.21
(0.39)
0.74
(0.78)
139
1.364
Round 3
0.26
(0.84)
0.666b
(0.062)
78. 2b
(33.4)
0.
(NE)
1.96b
(0.83)
-1.33
(0.53)
-0.70
(0.54)
125
1.343
Round 4
3.82b
(0.97)
0.646b
(0,082)
24.3
(27.8)
0.
(NE)
9.12b
(4.78)
-0.75
(0.68)
-0.75
(0.68)
106
1.329
aNE = not estimated, values not in parentheses are the model estimate parameter, the standard error is shown
in parentheses, GSD = geometric standard deviation.
One-tailed p < 0.05; statistically significant positive effect.
Note: Abatement occurred between Rounds 1 and 3.
July 15, 1993
4-26
DRAFT-DO NOT QUOTE OR CITE
-------
1 control of 0.74 - (-0.70) = 1.44 j^g/dL in the soil abatement group, and 0.74 - (-0.75)
2 =1.49 jLig/dL between Rounds 1 and 4. For the group that received only dust abatement,
3 there was an initial improvement in geometric mean blood lead between Rounds 1 and 3 of
4 —0,21 — (-1.33) = 1.12 /xg/dL, but after Round 3 there was a postabatement
5 recontamination that increased blood lead to control levels. These differences are not the
6 same as the raw blood lead mean differences because they have been adjusted for blood lead
7 autoregression, hand lead, and dust lead.
8 The adopted model is shown in Figure 4-2. Regression coefficients are shown in
9 Table 4-11. The model includes blood lead levels at Rounds 1 and 4 (October 1989 and
10 September 1990, respectively). Dust lead levels at Rounds 1 and 4, and soil lead levels
11 preabatement and postrecontamination were used. The mean XRF lead paint measurement
12 was a better predictor than the maximum XRF or the total area of chipped and peeling paint,
13 even though these were stabilized during the study. Analyses were run for each group BOS
14 SPI, BOS PI, and BOS P.
15
1
/
XRF Exterior
Lead-based Paint
K/
0
0
1363
440\
96 \
218 \
S
0
0
0
X
Preabatement
Blood Pb
i
k0.8
1.5
0.3
Preabatement
Floor Dust Cone.
t
°
0.44
0
Preabatement
Soil Cone.
0.6
0.5
.0.03
0.19
0.43
0.02
0.68
0.75
Postabatement
Blood Pb
' i
k 1.3
1.2
0.00
Postabatement
Floor Dust Cone.
i
0.96
0.01
0
Postabatement
Soil Cone.
^
0
0.13
0.49
Figure 4-2. Structural equation diagram for the Boston study. Numbers next to arrow,
from Table 4-11, are the regression coefficients (top to bottom) for BOS SP,
BOS PI, and BOS P.
July 15, 1993
4-27
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 4-11. REGRESSION COEFFICIENTS FOR BOSTON STRUCTURAL
EQUATIONS MODEL, USING THE AGLS METHOD
Response Variable (Output)
Predictor Variable (Input)
Preabatement Blood Lead
Preabatement Floor Dust
Concentration
Postabatement Floor Dust
Concentration
Prcabatement Soil
Concentration
Postabatement Soil
Concentration
Mean XRF Measure of
Pb-Based Paint
Group
BOS SPI
BOS PI
BOSP
BOS SPI
BOS PI
BOSP
BOS SPI
BOS PI
BOSP
BOS SPI
BOS PI
BOSP
BOS SPI
BOS PI
BOSP
BOS SPI
BOS PI
BOSP
Preabatement
Floor
Dust
Blood Cone. Soil
0.76a
1,45s
0.25°
0,0 0.0
0.0 0.44e
0.0 0.0
0.0 0.0 440a
0.36 0.0 96
0.0 l,363a 218d
Postabatement
Blood
0.60a
0.60a
0.51b
1.33e
1.18*
0.06
0.0
0.13
0.49e
Floor
Dust
Cone.
0.03
0.20
0.43d
0.96
0.01
0.0
0.96e
0.011
0.0
Soil
0.02
0.68a
0.75a
"One-tailed p less than 0.0005.
bOne-tailed p between 0.0005 and 0.005.
cOne-tailed p between 0.005 and 0.025.
dOne-tailed p between 0.025 and 0.05.
"One-tailed p between 0.05 and 0.10.
1 Preabatement blood lead was significantly related to preabatement dust lead, but not
2 directly related to soil or paint lead. The indirect effects are very strong, however, from the
3 paint -> soil -» dust -> blood pathway. The component of soil lead not accounted for by lead-
4 based paint also makes a large contribution to preabatement blood lead.
5 Postabatement blood lead levels are highly correlated with postabatement dust lead
6 levels in the abatement groups BOS SPI and BOS PI, but not in the nonabatement group
July 15, 1993
4-28
DRAFT-DO NOT QUOTE OR CITE
-------
1 BOS P. The coefficient (about 1.2 fig/dL blood lead per 1,000 ppm dust lead) is similar to
2 that found in many other studies, and may be an underestimate since it does not represent
3 blood lead that is fully equilibrated with the new dust lead levels. There is a small additional
4 effect from postabatement soil lead that is not statistically significant.
5 Postabatement dust lead levels are significantly lower relative to preabatement dust lead
6 concentrations in BOS PI than in BOS P, and much lower yet in BOS SPI. This is by far the
7 most important contribution to reduced blood lead levels. Soil lead levels in the non-soil-
8 abatement areas BOS PI and BOS P were also relatively lower after abatement, by about
9 25 to 30%. .
10 A formal statistical test of the equality of the blood lead autoregression coefficient R
11 was also carried out by constraining the coefficients to be equal in the three groups. There is
12 no reason to believe that R is significantly different in the three groups.
13 In summary, blood lead reductions in the Boston SPI group 8 mo after abatement
14 appear to be associated mth a persistent long-term reduction in the rate of transport of
15 exterior lead (largely from soil) into household dust. This amounted to about 1.2 jug Pb per
16 1,000 ppm dust lead reduction, or with a dust to soil lead ratio of about 0.7, about 0.8 to
17 0.9 /ig/dL per 1,000 ppm soil lead. In the Boston PI group, a similar blood lead reduction
18 could have occurred if there were no recontamination of household dust, but dust lead levels
19 and blood lead levels did rebound to high preabatement levels. Stabilized lead-based paint
20 did not contribute significantly to recontamination in the first few months after the
21 abatement. Lead-based paint was associated with a significant part of the elevated lead levels
22 in soil and dust found at the beginning of the study, contributing roughly 400 ppm to soil
23 lead per mg Pb/cm2 as a mean XRF level.
24
25 4.3.3 Cincinnati Study
26 The results for the Cincinnati study are shown in Tables 4-12 and 4-13. We found that
27 the six neighborhoods in the study showed some puzzling differences, even within the same
28 abatement category (Glencoe and Mohawk for controls and Findlay, Back, and Dandridge for
29 dust abatement only in Year 1). We therefore carried all six neighborhoods in the analyses.
30 The analysis of variance in Table 4-12 shows that there are significant differences among
31 categories. However, there is a very significant difference among Rounds 1,3, and
5;
July 15, 1993 4-29 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 4-12, REPEATED MEASURES ANALYSIS OF VARIANCE FOR LOG
(BLOOD LEAD) FOR CINCINNATI STUDY, ANALYSIS OF VARIANCE TABLES
Effect
MS
df
P
Between Subjects
Category
Variability
2.1850
0.8752
5
169
0.0328
Within Subjects
Round
Round X Category
Variability
TABLE 4-13. REPEATED
1.9427
0.2790
0,1316
MEASURES ANALYSIS
2
10
338
OF
(BLOOD LEAD) FOR CINCINNATI STUDY, ANALYSIS
Effect
MS
df
< 10"6
0.0225
COVARIANCE FOR LOG
OF COVARIANCE TABLES
P
Between Subjects
Category
Log (PbDF-Rl)
Variability
0.8589
5.1040
0.6457
5
1
135
0.2516
0.0057
Within Subjects
Round
Round X Category
Round X Log (PbDF-Rl)
Variability
0.3836
0.3353
0.1903
0.1043
2
10
2
270
0.0265
0.0006
0.1632
1 5 overall. The round by category interaction effects are statistically significant (p = 0.0225)
2 as shown in Table 4-12. However, there is no obvious similarity in pattern between the two
3 control neighborhoods. Two of the dust abatement neighborhoods (Findlay and Back) appear
4 to have similar preabatement levels and response to dust abatement, but the Dandridge
?
July 15, 1993 4-30 DRAFT-DO NOT QUOTE OR CITE
-------
1 neighborhood is different initially (Round 1) and very different postabatement (Round 5).
2 Analyses were not extended to Round 9 due to unavailability of Round 9 environmental data.
3 The initial Round 1 lead concentration in floor dust was the most predictive covariate and
4 was used in the analysis of covariance in Table 4-13. Consideration of initial floor dust lead
5 levels greatly increased the estimated effectiveness of the soil abatement between Rounds 1
6 and 5, amounting to a difference in interaction effects of 0.2132 - (-0.1233) = 0.3365;
7 that is, a reduction of e ' — 1 = 0.40, or 40% reduction in blood lead, everything else
8 being equal. However, this conclusion should not be over-interpreted because the baseline of
9 the group average includes unexplained variation in the "control" groups. As has been often
10 noted, the different Cincinnati neighborhoods appear to have been exposed before Round 5 to
11 variable external lead sources that were not related to this study.
12 The autoregressive regression model results shown in Table 4-14 do not include hand
13 lead levels, which were not available to us at the time of this analysis. It became evident
14 that there were exposure-related differences among the groups that could not be adequately
15 described by assigning different geometric mean blood lead levels to each neighborhood,
16 except for Round 1. Thus, the first column hi Table 4-14 shows blood lead levels or
17 intercepts for a model that also includes dependence on floor dust lead concentration. The
18 autoregressive models for Rounds 3 and 5 differ from the analyses in Tables 4-9 and 4-10,
19 but are similar to the postabatement analyses in Tables 4-4 and 4-5, in that a separate
20 autoregressive slope is used for each neighborhood. There are statistically significant
21 differences in these coefficients even within abatement groups: between control
22 neighborhoods Glencoe and Mohawk for Rounds 3 and 5 and between dust abatement
23 neighborhoods Findlay and Dandridge at Round 3. Floor dust lead levels are highly
24 predictive of within-group blood lead differences at Rounds 1 and 3, with a magnitude of
25 about 2 /xg/dL per milligram of lead per gram of dust, but not at Round 5. However, the
26 residual GSDs are much larger than for the categorical repeated measures models hi
27 Tables 4-12 and 4-13, so this model is much less adequate than the autoregressive regression
28 models for Baltimore and Boston.
29 The model adopted for Cincinnati is shown in Figure 4-3 and results are shown in
30 Table 4-15. After careful evaluation of the preliminary results, it became evident that the
31 CIN I-SE group needed to be split into separate components. The Back Street and Findlay
My 15, 1993 4-31 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 4-14. AUTOREGRESSIVE REGRESSION MODEL BY NEIGHBORHOOD
FOR BLOOD LEAD ON ENVIRONMENTAL LEAD, FITTED IN
LOG FORM, FOR CINCINNATI STUDY3
Auto Regression
Variable
Intercept Qxg/dL)
Autoregiession Round
Autoregression
Glencoe (NT)
Autoregression
Mohawk (NT)
Autoregression
Findlay (I-SE)
Autoregression
Back St. (I-SE)
Autoregression
Dandridge (T-SE)
Autoregression
Pendleton (SET)
Hand Lead (/tg/pair x
Baseline Round 3 Round
—
—
7.64°
(0.68)
6.30°
(1.20)
9.42C
(1.09)
11.32C
(1.67)
11.15°
(1.15)
8.66C
(0.82)
103) 53.0C
(30.2)
2.37°
(0.39)
1
0.347°
(0.058)
0.510°
(0.132)
0.546°
(0.070)
0.598°
(0.099)
0.577°
(0.067)
0.429°
(0.063)
81.1°
(53.3)
1.84°
(0.58)
1
0.517C
(0.097)
0.959C
(0.252)
0.661C
(0.105)
0.428C
(0.151)
0.673°
(0.102)
0.667°
(0.112)
-12.6
(13.0)
5 Round 5
1.46°
(0.44)
3
0.851°
(0.177)
1.484°
(0.262)
0.870°
(0.104)
0.567°
(0.150)
0.855°
(0.095)
0.894C
(0.104)
-26.1
(10.7)
b Round 7
3.04°
(0.24)
5
0.496°
(0.055)
0.535°
(0.056)
0.600°
(0.050)
0.431°
(0.098)
0.618°
(0,052)
0.638°
(0.058)
-2.6
(14.3)
Round 9
2.43°
(0.41)
7
0.605°
(0.082)
0.897°
(0.121)
0.820°
(0.087)
0.586°
(0.223)
0.861°
(0.083)
0.676C
(0.078)
14.4
(10.8)
*NE = not estimated, values not in parentheses are the model estimate parameter, the standard error is shown
in parentheses.
Alternate analyses.
cOne-tailed p < 0.05; statistically significant positive effect, :
1 neighborhoods (now denoted CINI-SE1) were similar in response, and these differed
2 substantially from the Dandridge neighborhood (now denoted CIN I-SE2). We therefore ,
3 show four groups in our analyses. The models included blood lead and environmental lead
4 levels at Rounds 1 and 5 (July 1989 and July 1990, respectively).
5 Floor lead concentrations and floor lead loadings were not as consistently associated
6 with blood lead levels as in the Boston study. Dust lead concentrations or loadings at the
7 entrance of the residence unit were better predictors than floor lead levels. We developed a
July 15, 1993
4-32
DRAFT-DO NOT QUOTE OR CITE
-------
Exl
a
Lead-
0
0
409
0
Preabatement
g-57 y Blood Pb
o'aa/^ *
/ 0
enorTnm *? Preaba
ii iu wctii > Floor Di
based Paint
0,
\ 104
\. 0
0
0
2.6
0.12
tement
ist Load
10.9
6.1
0
0
\J Preabatement
| Entry Dust Cone.
0.1 B
0.38
0.56
0.65 ^
u.
0.83
0.01
0
0 >.
3.1
o.a
0.3
0 1.6
0.3
1.4
181
0
0
78
1
1.3
0.7
0
r°
Postabatement
Blood Pb
1
k 0
0.7
0
0.45
Postabatement
Floor Dust Load
j
' 1.1
0.3
1.3
0
Postabatement
Entry Dust Cone.
/
L
3.1
0.7
0.2
0
Figure 4-3. Structural equation diagram for the Cincinnati study. Numbers next to
arrow, from Table 4-15, are the regression coefficient (top to bottom) for
CIN SEI, CIN I-SE-1, CIN I-SE-2, and CIN NT.
1 model that used entrance dust lead concentration as an indicator of external lead exposure,
2 and floor dust lead loading as a proximate indicator of indoor exposure.
3 Unlike Boston, the abatement groups corresponded to spatially distinct neighborhoods
4 that also differed in some important preabatement characteristics. We must therefore allow
5 for the possibility that the primary environmental pathways, and the effectiveness of the
6 abatements in interdicting those pathways, do differ among different neighborhoods.
7 The lead-based paint measurements used in our models included interior wall and trim
8 (denoted XRWL1 and XRTRM1 respectively) and exterior wall and trim (denoted XREWL
9 and XRETEM, respectively). There were significant XRF levels ( greater than 1 mg
10 Pb/cm2) at least somewhere in many of these units, even the majority were described as
11 rehabilitated ("gut rehab"). Only 7 of 157 children lived in nonrehabilitated units. The
12 analyses were done with and without these children, all of whom Eved in I-SE1.
July 15, 1993
4-33
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 4-15. REGRESSION COEFFICIENTS FOR CINCINNATI STRUCTURAL
EQUATIONS MODEL, USING THE AGLS METHOD
Response Variable (Output)
Preabatement
Predictor Variable (Input)
Preabatement Blood Lead
Preabatcment Floor Dust
Loading
Postabatement Floor
Loading
Preabatement Entry Dust
Cone.
Postabatement
Entry Cone.
XRF Exterior Trim
Pb-Based Paint
XRF Exterior Wall
Pb-Based Paint
Group
CIN SEI
CIN I-SE-1
CIN I-SE-2
CIN NT
CIN SEI
CIN I-SE-1
CIN I-SE-2
CIN NT
CIN SEI
CIN I-SE-1
CIN I-SE-2
CIN NT
CIN SEI
CIN I-SE-1
CIN I-SE-2
CIN NT
CIN SEI
CIN I-SE-1
CIN I-SE-2
CIN NT
CIN SEI
CIN I-SE-1
CIN I-SE-2
CIN NT
CIN SEI
CIN I-SE-1
CIN I-SE-2
CIN NT
Blood
0.0
0.0
2.61
0.12
3.13e
0.79
0.29
0.0
0.57
0.0
2.49d
0.22
0.0
1.38
0.24
0.0
Floor Entry
Loading Cone.
0.94d
6.11a
0.0
0.0
0.0 0.0
362d 104d
0.0 0.0
0.0 0.0
506 0.0
1,810 774
0.0 0.0
' 0.0 166
Postabatement
Floor Entry
Blood Loading Cone,
0.162
0,38b
0,564a
0.649a
0.829a
0.012
0.0
0.0
0,0
0,69e
0.0
0.45
1.56a
0.30d
1.42a
0.38
1.12 0.0
0.28 0.048
1.32d 0.035
0,0 0.199
1.28° 0.0 181a
0.18 0.0 0.0
0.0 409a 0.0
0.0 0.0 78
0.0 230
1.24 0.0
1.21 260
0.0 131
"One-tailed p less than 0.0005.
bOne-tsiled p between 0.0005 and 0.005.
cOne-tailed p between 0.005 and 0.025.
dOne-tailed p between 0.025 and 0.05.
"One-tailed p between 0.05 and 0.10.
July 15, 1993
4-34
DRAFT-DO NOT QUOTE OR CITE
-------
1 We found strong relationships between blood lead level and floor dust lead loading in
2 CIN I-SE2 and in CIN SEI, both pre- and postabatement, and between blood lead level and
3 entrance dust lead concentration postabatement in both groups, but only in CIN SEI
4 preabatement. Exterior lead-based paint contributed to blood lead levels both pre- and
5 postabatement in CIN I-SE1 and CIN I-SE2, but not in CIN SEI or CIN NT. Exterior lead-
6 based paint contributed to floor dust lead loadings preabatement in CIN I-SE1 and CIN SEI,
7 but not in CIN I-SE2 or CIN NT. Exterior lead-based paint contributed to entrance dust lead
8 concentrations postabatement in CIN I-SE2, CIN NT, and CIN SEI, but not in CIN I-SE1.
9 This may be accounted for to some extent because the autoregression of entrance lead
10 postabatement to entrance lead concentration preabatement was 0.30 to 0.38 for CIN I-SE1
11 and CIN NT, but 1.42 to 1.56 for CIN I-SE2 and CIN SEI.
12 There was at least some reasonable consistency in estimates of the blood lead
13 autoregression coefficient R. The estimates decreased from 0.649 in CIN NT to 0.564 in
14 CIN I-SE2 to 0.380 in CIN I-SE1 to 0.162 in CIN SEI, as per Hypothesis 1. The difference
15 between R for CIN NT and CIN SEI was statistically significant, p = 0.0255 one-tailed
16 (Hypothesis 1 asserts that R for nonabatement is greater than for effective abatement, so a
17 one-tailed test is appropriate). In summary, in spite of the large changes in dust lead
18 loadings and concentrations, increasing in some neighborhoods and decreasing in other
19 neighborhoods, the relative decrease in blood lead levels between pre-. and postabatement
20 samples was significantly larger in the soil abatement area than in the nonabatement areas,
21 after adjusting for current exposure. The relative blood lead reduction was also larger in the
22 areas with interior dust abatement only (albeit not significantly). It appears that, whatever
23 transient effectiveness the soil and dust abatements may have had in the first few months
24 „ after abatement, these effects were being overtaken by increases in lead exposure that
25 presumably could not be controlled by the study. Effects of the abatements could only be
26 detected after adjusting for changes in postabatement lead exposure.
27 . • .
28
29
July 15, 1993 4.35 DRAFT-DO NOT QUOTE OR CITE
-------
1 4.4 DISCUSSION AND CONCLUSIONS
2 4.4.1 Comparison Across the Three Studies
3 The effectiveness of soil lead abatement in reducing blood lead varied greatly among
4 the three cities. The variability in abatement effects is probably due to substantial differences
5 in lead sources and pathways among the neighborhoods in these studies. We will discuss
6 these differences for each study.
7 The Baltimore study had two neighborhoods, Upper Park Heights and Walbrook
8 Junction. The area to which abatement was assigned (Park Heights) had enrolled families
9 whose residences did not have soil lead levels that were high enough to justify abatement.
10 The soil lead levels in the nonabatement premises in Park Heights that were measured in the
11 preabatement phase were not significantly smaller than those of the abated houses or of the
12 control premises in Walbrook Junction. We therefore used the nonabatement houses in Park
13 Heights as an additional control group. Unlike the other studies, the soil abatement was not
14 accompanied by interior dust abatement. There was essentially no significant effect of soil
15 abatement in the abated houses, compared to the control group. Statistical covariate
16 adjustment hi both repeated measures analyses and autoregressive regression analyses showed
17 that the differences in blood lead levels both before and after abatement were significantly
18 dose-related to interior lead-based paint and (nonabated) interior dust. It is likely that
19 interior paint contributed to child lead exposure, either directly by ingestion of paint chips,
20 or indirectly by the hand-to-mouth exposure pathway: ,
21
22 interior paint => interior dust => hands => blood.
23
24 Cross-sectional and longitudinal structural equation analyses could be used to explore this
25 hypothesis. However, because there were no repeated measurements of household dust lead,
26 it will be very difficult to assess changes in exposure over time except by use of hand lead
27 data.
28
29 // is likely that soil lead abatement had little effect on the primary factors
30 responsible for elevated child blood lead levels in these two neighborhoods,
31 wJuch appear to be interior lead-based paint and interior dust lead.
32
July 15, 1993 4-36 DRAFT-DO NOT QUOTE OR CITE
-------
1 The Boston study was conducted with blood and hand leads measured at one
2 preabatement round and at postabatement rounds about 2 and 8 mo after abatement. Soil and
3 dust lead measurements were available for all three rounds at about the same time. There
4 were also environmental data on soil and dust immediately after abatement for the soil
5 abatement premises, and there were dust lead measurements immediately after abatement for
6 the dust-only abatement premises. These data allowed a very complete analysis of blood lead
7 responses to changes in dust and soil lead over time. The results showed clearly that there
8 was a persistent 10% reduction in blood lead levels (1.4 or 1.5 ^g/dL) in the soil lead
9 abatement children, and that, on average, the postabatement blood leads were lowest in
10 premises that had the lowest postabatement soil lead and dust lead loadings. Interior and
11 exterior lead paint were not significant predictors of blood lead for Boston children.
12
13
14 When dust lead and soil lead levels show a persistent decrease as a result of
15 effective abatement, blood lead levels also show a persistent decline. The
16 postabatement blood lead levels are lower when postabatement dust lead levels
17 are persistently lower over a long time.
18
19 The Cincinnati study had collected blood and lead and environmental samples in six
20 Cincinnati neighborhoods. We were able to generate analyses comparable to those reported
21 for the Baltimore and Boston studies. After some analyses using models similar to those for
22 Baltimore and Boston, it became evident that the neighborhoods within each treatment group
23 (two neighborhoods as controls, three as interior dust abatement only) were not identical in
24 some ways, so the analyses were run with six neighborhood-treatment groups instead of
25 three. Although the statistical tests showed that there were strong interactions between group
26 and .round, there was no clear pattern of effect within any group except for a modest
27 decrease in blood leads in the surface soil abatement neighborhood and in one of the dust
28 abatement neighborhoods in Round 5. This reduction was hard to interpret because there
29 were changes in blood lead in the control neighborhoods during this same time interval.
30 Although there was a strong dependence of blood lead on environmental lead, particularly on
31 hand lead and on current or previous dust lead loadings on floors, there was no clear pattern
32 of change or response of interior dust lead levels after abatement.
July 15, 1993 4.37 DRAFT-DO NOT QUOTE OR CITE
-------
1 Some of the analyses in Tables 4-14 and 4-16 require further discussion. The intercept
2 corresponds roughly to the component of new lead exposure associated with "background"
3 sources such as diet. We have assumed that the background sources were the same in all
4 neighborhoods. The autoregression coefficient is a composite of time-dependent factors.
5 If everything else were constant (stationary over time), then the autoregression coefficient
6 would be equal to an exponential function of the pharmacokinetic mean residence time and
7 the time between successive blood lead measurements. However, the relationship of blood
8 lead measurements to a previously measured blood lead concentration is also a function of
9 child age, seasonal variations, and other changes in lead exposure over time. We have not
10 adjusted the autoregressions for these factors.
11 Note that the autoregression coefficient for Round 3 versus Round 1 blood lead is much
12 lower in the soil abatement neighborhood (Pendleton) than in the three neighborhoods that
13 received only interior dust lead abatement. For example, in Table 4-14, the difference
14 between Dandridge (dust abatement only) and Pendleton (soil and dust abatement) for
IS Round 3 versus Round 1 is 0.577 — 0.429 = 0.146. In other words, the nonbackground
16 (presumably soil and dust) steady-state component of blood lead was about 14% smaller in
17 the soil abatement neighborhood than in a nearby neighborhood with dust abatement only.
18 This suggests that there is at least some initial effect of soil lead abatement over and above
19 that of dust lead abatement. The control neighborhoods, Glencoe and Mohawk, appear to
20 have very different autocorrelation coefficients and other statistics. The control
21 neighborhoods were at some distance from the soil abatement neighborhood, and the three
22 indoor dust abatement neighborhoods were between them. The clustering of neighborhood
23 locations and blood lead response variables suggests that some simpler comparisons may be
24 needed in order to compare the effectiveness of soil lead treatment versus nontreatment.
25 The autocorrelations for the soil abatement and dust abatement neighborhoods were
26 similar (differences were not statistically significant) after Round 3. This suggests that the
27 environmental abatements did not affect hand and dust lead pathways, but may have reduced
28 the overall soil and dust lead exposure. These preliminary findings and interpretations
29 should be verified by more thorough longitudinal statistical analyses using structural equation
30 models.
July 15, 1993 4-38 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 4-16. AUTOREGRESSIVE REGRESSION MODEL FOR BLOOD LEAD ON
ENVIRONMENTAL LEAD, FITTED IN LOG FORM, FOR CINCINNATI STUDY3
Variable
Intercept 0*g/dL)
Autoregression Round
Autoregression
Glencoe (NT)
Autoregression
Mohawk (NT)
Autoregression
Findlay (I-SE)
Autoregression
Back St. (I-SE)
Autoregression
Dandridge (I-SE)
Autoregression
Pendleton (SET)
Hand Lead
Og/pair x 103)
Floor Dust Lead Loading
(fig/m2 x 103)
Dust Round
Floor Dust Lead Loading
Og/m2 x 103)
Dust Round
Round
—
1 Round
Intercept 1
7.
(0.
5.
(1.
9.
(1.
11.
(1.
11.
(1.
70C
67)
82C
13)
40C
06)
24C
56)
61C
16)
9.06C
(0.88)
35.
(22.
T
2)
0.086
(0.111)
1
—
0.357C
(0.060)
0.491C
(0)
0.511C
(0.071)
0.612C
(0.102)
0.586C
(0.070)
0.415°
(0.067)
73.2
(56.8)
0.089
(0.100)
1
4.58
(2.93)
3
3 Round5
3
<°
1
0
(0
0
(0
0
(0
0
(0
0
(0
0
(0
64
(33
2
(1
4
.46C
.64)
.296C
.087)
.581C
.195)
.467°
.106)
.498C
.192)
.451°
.098)
.400C
.106)
.lc
.9)
.49
.97)
(2.93)
Round 5
2.54°
(0.55)
3
0.585C
(0.110)
L160C
(0.237)
0.709C
(0.105)
0.650°
(0.185)
0.692C
(0.098)
0.716°
(0.120)
34.2
(29.8)
0.38
(1.66)
4
(2.93)
b Round 7 Round 9
3
(0
5
0
(0
.64°
.61)
7
.467C
.111)
0.559°
(0.161)
0
(0
0
(0
0
(0
0
(0
8
(38
1
(1
4
0
(0
6
.557°
.097)
.283°
.127)
.490C
.090)
.663°
.122)
.6
.5)
.17
•91)
.80C
.40)
aValues not ifl parentheses are the model estimate parameter, the standard error is shown
in parentheses.
Alternate analyses.
cOne-tailed p < 0.05; significant positive effect.
1 There are some differences between Tables 4-14 and 4-16. The estimated regression
2 coefficients between blood lead and hand lead for Rounds 3 through 9 are nonsignificant or
3 negative in Table 4-14. When floor dust lead loadings were also included as predictors of
4 blood lead, the estimated relation of hand lead to blood lead was always positive, although
5 significant only for Round 5. When the hand lead regression coefficient was not significant,
July 15, 1993
4-39
DRAFT-DO NOT QUOTE OR CITE
-------
I then the dust lead loading regression coefficient was significant or nearly so. This suggests
2 that there are some interactions among blood lead, hand lead, and dust lead variables,
3 possibly as causal pathways.
4 We are inclined to accept the conclusion of the Cincinnati investigators that blood and
5 dust lead levels were affected differently at different times and places by exogenous sources,
6 possibly related to repainting of nearby highway bridges or other events not under their
7 control. However, the dose-dependence exhibited in the models suggests that reducing
8 interior dust lead levels did reduce blood lead levels, at least for a while. The problem is
9 that the abatements did not always persistently reduce dust lead levels.
10 ,
11 We conclude that there were additional sources of environmental lead exposure
12 that had different effects on the neighborhoods during the course of the study
13 and were not related to the abatement methods used in the Cincinnati study.
14 It will be necessary to use other analysis methods, such as structural equations
15 modeling, in order to assign changes in Gncinnati child blood lead levels to
16 changes in lead exposure.
17
18
19 4.5 SUMMARY OF STATISTICAL INFERENCES
20 This report concurs with the results reported by the individual studies. The reanalysis
21 of the individual study data sets, where performed in the same manner as the report, revealed
22 no evident errors in statistical analysis. Reconstruction of the exact statistical procedures and
23 selection of records to be included proved to be a formidable task.
24 Certain procedures were not performed by one or more of the three studies, but were
25 included in the reanalysis of the data. The repeated measures analysis and autoregressive
26 regression model revealed that the relationship between environmental lead and blood lead
27 was more or less uniform across all three studies. When the environmental lead increased,
28 the blood lead increased, and when environmental lead decreased, blood lead decreased. The
29 removal of variance caused by seasonal cycles and long-term time trends appeared to resolve
30 most of the nonenvironmental variance in the blood lead measurements. The imputing of
31 values for missing data in the Baltimore study restored enough observations to the data set to
32 suggest a relationship between, environmental lead and blood lead similar to that found in
33 Boston, although the difference in Baltimore was not statistically significant.
July 15, 1993 4-40 DRAFT-DO NOT QUOTE OR CITE
-------
1 Several of the statistical analyses confirm, or at least support, the observations in
2 Chapter 3 that there is a strong link between environmental lead and blood lead. The most
3 crucial analysis requires the development of structural equation models, a time-consuming
4 process that will not be completed within the tune limits of this report.
July 15, 1993
4-41
DRAFT-DO NOT QUOTE OR CITE
-------
i 5. CONCLUSIONS
2 .
3
4 5.1 SUMMARY OF PROJECT
5 This project focuses on the exposure environment of the individual child. One measure
6 of short-term exposure is the child's blood lead. Two other indicators of exposure are house
7 dust lead concentrations and hand dust lead loading. From the perspective of the child's
8 environment, changes in the soil lead concentration are expected to bring about changes in
9 the house dust and blood lead concentrations, and hand dust loading. In each of the three
10 studies, the soil lead concentrations were reduced to approximately 50 /xg/g in the study area.
11 For most children, there was a measurable, although not always statistically significant,
12 reduction of blood lead. When corrected for seasonal and age-related cyclic variations on
13 blood lead, the impact was even greater, and the effect was maximized when the rate of
14 movement of dust through the human environment was taken into account. That is, when
15 street dust and house dust were also removed from the environment so that the clean soil
16 represented the major source of lead to the child's environment, the impact of abatement was
17 the greatest.
18 The earlier sections of this document evaluated the following statements:
19
20 (1) that the abatement of the soil resulted in a reduction of soil lead
21 concentration on a case-by-case basis, and that this reduction persisted
22 throughout the study (Chapter 3);
23
24 (2) that the intermediate exposure elements (street dust, house dust, and hand
25 dust) responded to this soil abatement (Chapter 4); and
26
27 (3) that there was a measurable decrease in blood lead that could be directly
28 related to the abatement of soil (Chapter 4).
29
30
31 Chapter 3 discussed the measurement and abatement of lead in soil. It was apparent
32 that where soil abatement was conducted, this abatement was effective and persisted for the
33 duration of the study. Chapter 4 provided the statistical analysis required to support the
34 conclusions in Chapter 5. ,
July 15, 1993 5-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 In the Cincinnati study, the persistency of external dust abatement was very short,
2 leading to the conclusion that sources of lead other than soil were predominant in the
3 neighborhood. At this time, there is no further evidence for the nature or size of this source,
4 It is clear, however, that until this source is identified and controlled, soil abatement at the
5 neighborhood level under conditions similar to those in Cincinnati would have little influence
6 on the house dust of individual homes and would only influence the blood lead concentrations
7 for those children playing in the areas with the soil abatement.
8
9
10 5.2 SUMMARY OF RESULTS
11 The results of these three studies demonstrated a clear relationship between
12 environmental lead and blood lead. There were few instances where changes in the blood
13 lead concentrations could not be attributed to changes in environmental lead. Unexpected
14 changes, such as the dramatic increase in the Cincinnati dust lead concentrations between
15 November 1989 and July 1990, were observed in nearly every instance (floor, window, mat,
16 and entry) of every study group, but the sources remain unexplained. Although this apparent
17 contamination appears to have overwhelmed the intervention efforts, the fact that both hand
18 dust loads and blood lead concentrations responded accordingly gives credence to the strong
19 link between environmental lead and blood lead. Other, less dramatic changes also produced
20 corresponding changes in blood lead concentrations,
21 In terms of changes attributed to intervention, it is appropriate to note that all three
22 studies observed a quantifiable change in response to intervention. The analyses in Chapter 4
23 show that, although not always statistically significant, this quantifiable response to
24 intervention is consistent even at low levels of environmental lead. Normalized to a decrease
25 in soil lead concentration of 1,000 /*g/g, the aggregate response in blood lead concentrations
26 appears to be about 1 /*g/dL. This suggests that there is no plateau, within the ranges
27 measured in this project, where the removal of environmental lead will not produce a
28 corresponding reduction in blood lead concentrations.
29 Finally, the project results shed additional light on the well-known phenomenon of
30 seasonal cycles in blood lead concentrations. The rare opportunity to evaluate three
31 independent longitudinal studies with similar sampling and analysis protocols led to the
July 15, 1993 5-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 conclusion that the amplitude of the cycle is roughly 15% in all three cities, that the peak
2 occurs about August 15-20, and that these factors appear to be independent of environmental
3 lead.
4
5
6 5.3 SUMMARY OF STATISTICAL INFERENCES
7 5.3.1 Baltimore Study
8 It is likely that soil lead abatement had little effect on the primary factors responsible
9 for elevated child blood lead levels in these two neighborhoods, which appear to be interior
10 lead-based paint and interior dust lead.
11 It appears then that children with higher exposures to interior dust or interior lead-based
12 paint will more consistently have higher blood lead concentrations in these two
13 neighborhoods than will children with elevated soil lead levels, and that it may be necessary
14 to directly intervene to reduce dust and paint lead levels in these houses. Although exterior
15 lead-based paint was correlated with soil lead, the addition of an exterior paint term to the
16 model did not significantly improve the ability to estimate blood lead levels. This does not
17 imply that there is no benefit to abatement of soil or exterior lead-based paint, however,
18 because there was clearly a relationship between postabatement soil lead levels and blood
19 lead that may indicate that reducing soil lead has reduced the soil component of interior dust
20 lead loading.
21
22 5.3.2 Boston Study
23 There is evidence for an effect of soil lead abatement on blood lead concentrations that
24 depends—not surprisingly—on the initial soil and dust lead levels. Because soil lead and
25 interior dust lead are highly correlated, we may infer that in the Boston houses studied,
26 exposure to dust lead (as a primary vector) that was derived from soil lead can be reduced
27 over time by a combined removal of soil and dust lead, but not by dust lead abatement alone.
28 Soil lead levels, including postabatement values, never made a significant contribution
29 to increasing blood levels when hand lead and dust lead loadings were included in the model.
30 Postabatement dust lead loadings were highly significant predictors of postabatement blood
31 lead. There were significant reductions in blood lead between Rounds 1 and 3, with
July 15, 1993 5-3 DRAFT-DO NOT QUOTE OR CITE
-------
I decreases of geometric mean blood lead relative to control of 0.74 — (—0.70) = 1.44 ,wg/dL
2 in the soil abatement group, and 0.74 — (—0.75) = 1.49 jwg/dL between Rounds 1 and 4.
3 For the group that received only dust abatement, there was an initial improvement in
4 geometric mean blood lead between Rounds 1 and 3 of —0.21 — (—1.33) = 1.12 j«g/dL,
5 but after Round 3 there was a postabatement recontamination that increased blood lead to
6 control levels. These differences are not the same as the raw blood lead mean differences,
7 because they have been adjusted for blood lead autoregression, hand lead, and dust lead.
8 However, they point to a very clear conclusion from the Boston study: When dust lead and
9 soil lead levels show a persistent decrease as a result of effective abatement, blood lead
10 levels also show a persistent decline. The postabatement blood lead levels are lower when
11 postabatement dust lead levels are persistently lower over a long tune.
12
13 5.3.3 Cincinnati Study
14 We conclude that there were unexplained sources of environmental lead exposure that
15 had different effects on the neighborhoods during the course of the study and were not
16 related to the abatement methods used in the Cincinnati study. It will be necessary to use
17 other analysis methods such as structural equations modeling in order to assign changes hi
18 Cincinnati child blood lead levels to changes in lead exposure. Lead paint was absent and
19 was not a confounding factor.
20
21
22 5.4 INTEGRATED PROJECT CONCLUSIONS
23 This report concurs with the conclusions reported by the individual studies. The
24 reanalysis of the individual study data sets, where performed in the same manner as the
25 report, revealed no evident errors in statistical analysis. The repeated measures analysis and
26 autoregressive regression model revealed that the relationship between environmental lead
27 and blood lead was more or less uniform across all three studies. When the environmental
28 lead increased, the blood lead increased, and when the environmental lead decreases, the
29 blood lead decreases.
30 The removal of variance caused by seasonal cycles and long-term time trends appeared
31 to resolve most of the nonenvironmental variance in the blood lead measurements. The
July 15, 1993 5-4 DRAFT-DO NOT QUOTE OR CITE
-------
1 imputing of values for missing data in the Baltimore study restored enough observations to
2 the data set to suggest a relationship between environmental lead and blood lead similar to
3 that found in Boston, although the difference in Baltimore was not statistically significant.
4 Several of the statistical analyses confirm, or at least support, the observations in
5 Chapter 3 that there is a strong link between environmental lead and blood lead. The most
6 crucial analysis requires the development of structural equation models, which were beyond
7 the scope of this reanalysis.
8
9 5.4.1 Findings
10 The analysis of the data from the three studies showed evidence that blood lead
11 responds to changes in environmental lead. This suggests that abatement of any type and to
12 any degree will cause a reduction in the blood lead of children. All three studies and all
13 groups within each study produced data supporting this conclusion, although not statistically
14 significant hi two of the groups in the Baltimore study.
15 All three studies also showed evidence for a quantifiable impact of intervention. This
16 may have been intervention from soil abatement, dust abatement, or paint stabilization.
17 In Baltimore, this impact was temporary at best and was marginally significant.
18 In Cincinnati, the impact was quickly swamped by other sources of environmental lead.
19 In Boston, the impact was persistent. The best estimate for this effect is 1 jug/dL per
20 1,000 ,«g/g decrease in soil. Similar decreases in exterior dust would be expected to have a
21 similar effect.
22 There is evidence from all three studies that lead moves throughout the child's
23 environment. This means that lead in soil becomes lead in street or playground dust, lead in
24 paint becomes lead in soil, and lead in street dust becomes lead in house dust. A more
25 detailed analysis of the data may show the relative contribution from two or more sources,
26 but the present analyses confirm that this transfer takes place. In the Baltimore study, there
27 was statistical evidence for implied causal pathways^ such as paint to exterior dust, but in the
28 Boston and Cincinnati studies, the pathways were explicit.
29 Finally, there is evidence for the continued impact of nonabated sources following
30 abatement. This means that abatement of soil probably does not reduce the contribution of
31 paint lead to the child's exposure.
July 15, 1993 5-5 DRAFT-DO NOT QUOTE OR CITE
-------
I 2. New techniques to measure dust in urban neighborhoods should be
2 developed. The mat placement experiment in the Cincinnati study shows
3 promise of being a simpler approach to measuring house dust than other
4 methods available. If developed further to determine the optimum time of
5 placement and recovery, a public health officer could place these mats and
6 recover them with minimum intrusion into the home and reasonable
7 convenience in recovering the sample in the laboratory.
8
9 Other techniques, such as hand wipe analysis and methods for measuring
10 dust loading as well as dust lead concentration should be further developed
11 and become routine is studies of this type.
12
13 3. Judgements on whether or not to conduct soil abatement should be based
14 on more than the expected decrease in blood lead concentrations. This
15 study shows that children living in urban environments are exposed to
16 many sources of lead. Abatement of one source should be considered only
17 in the context of other sources, including lead-based paint. This decision
18 should also take into the consideration the impact on children moving into
19 the neighborhood from areas with lower exposure. Effective abatement
20 prior to this move would probably have a greater impact on these children
21 than on those residing in the neighborhood before abatement,
22
July 15, 1993 5-8 DRAFT-DO NOT QUOTE OR CITE
-------
6. REFERENCES
4 Annest, J.L.; Pirkle, J.L.; Makuc, D.; Neese, J.W.; Bayse, D.D.; Kovar, M.G. (1983) Chronological trend in
5 blood lead levels between 1976 and 1980. N. Engl. J. Med. 308:1373-1377.
6
7 Bamett and Lewis (1984) Outliers in Statistical Data. John Wiley and Sons, NY.
8
9 Bentler, P.M. (1989) EQS Structural Equations Program Manual. BMDP Statistical Software, Los Angeles, CA.
10
11 Dempster, A.P.; Laird, N.M,; Rubin, D.B. (1977) Maximum likelihood from incomplete data via the EM
12 algorithm. J. Royal Statistical Soc. Series B, Vol 38: 1-11.
13
14 Grant, L.D.; Elias, R.W.; Gayer, R.; Nicholson, W.; Olem, H. (1990) Indirect health effects associated with
15 acidic deposition. National Acid Precipitation Assessment Program. SOS/T Report 23. U.S.
16 Government Printing Office, Washington, DC.
17
18 Hagselblad, V.; Stead, A.G.; Galke, W. (1980) Analysis of coarsely grouped data from the lognormal
19 distribution. J. Am. Stat. Assoc. 75:771-778.
20
21 Lotus Development Corporation (1990) Lotus 1-2-3 Release 3.1. Lotus Development Corporation, Cambridge,
22 MA.
23
24 Orchard, T.; Woodbury, M.A. (1972) A missing information principle: Theory and applications. Proc 6th
25 Berkeley Symp on Mathematical Statistics and Probability. Vol 1, pp 697-715.
26
27 Roberts, J.W.; Camaan, D.E.; Spittler, T.M. (1991) Reducing lead exposure from remodeling and soil track-in
28 in older homes. Air and Waste Management Association Paper 91-134.2, 84th Annual Meeting and
29 Exhibition, Vancouver, British Columbia, June 16-21, 1991.
30
31 U.S. Environmental Protection Agency (1986) Air quality criteria for lead. Research Triangle Park, NC: Office
32 of Health and Environmental Assessment, Environmental Criteria and Assessment Office; EPA report no
33 EPA-600/8-83/028aF-dF. 4v. Available from: NTIS, Springfield, VA; PB87-142378.
34
35 Wilkinson, L. (1990) SYSTAT: The system fot statistics. Version 5.03 (1991). Evanston, IL: SYSTAT, Inc.
36
37 Woodbury, M.; Hasselblad, V. (1970) Maximum likelihood estimates of the variance covariance matrix from the
38 multivariate normal. Proc of SHARE XXXIV, Vol 3, pp 1550-1561.
39
July 15, 1993 6-1 DRAFT-DO NOT QUOTE OR CITE
*U.S. GOVERNMENT PRINTING OFFICE: 1993 - 750-068/60014
------- |