United States
Environmental Protection
Agency
Office of Pollution
Prevention and Toxics
Washington. DC 20460
Peer Review Draft
August. 2000
Risk Analysis to Support Standards for
Lead in Paint, Dust, and Soil
Supplemental Report
-------
DRAFT FOR PEER REVIEW - DO NOT CITE OR QUOTE
August 28, 2000
RISK ANALYSIS TO SUPPORT STANDARDS
FOR LEAD IN PAINT, DUST, AND SOIL
SUPPLEMENTAL REPORT
National Program Chemicals Division (7404)
Office of Pollution Prevention and Toxics
U.S. Environmental Protection Agency
Washington, DC 20460
EPA Contract Number 68-W-99-033
-------
DISCLAIMER
The material in this document has not been subject to Agency technical and
policy review. Views expressed by the authors are their own and do not
necessarily reflect those of the U.S. Environmental Protection Agency.
Mention of trade names, products, or services does not convey, and should not
be interpreted as conveying, official EPA approval, endorsement, or
recommendation. Do not quote or cite this document.
This report is copied on recycled paper.
-------
CONTRIBUTING ORGANIZATIONS
This report is a supplement to EPA 747-R-97-006 ("Risk Analysis to Support Standards
for Lead in Paint, Dust, and Soil"). Efforts to produce this report were funded and managed by
the U.S. Environmental Protection Agency. The risk analysis was conducted by Battelle
Memorial Institute under contract to the U.S. Environmental Protection Agency. Each
organization's responsibilities are listed below.
Battelle Memorial Institute (Battelle)
Battelle was responsible for performing the additional data analyses, literature reviews,
and documentation presented in this report. Battelle was also responsible for preparing this
report.
U.S. Environmental Protection Agency (EPA)
The Environmental Protection Agency was responsible for providing direction on the
technical issues to be presented in this report, providing relevant information for the report,
reviewing the report, contributing to the development of conclusions, and managing the peer
review and publication of the report. The EPA Work Assignment Manager was Mr. Ronald
Morony. The Deputy Work Assignment Managers were Mr. Brad Schultz and Mr. Dave
Topping. The EPA Project Officer was Ms. Sineta Wooten.
DRAFT--DO NOT CITE OR QUOTE ill August 28,2000
-------
This page intentionally blank.
DRAFT - DO NOT CITE OR QUOTE iv August 28,2000
-------
TABLE OF CONTENTS
Page
EXECUTIVE SUMMARY xxi
1.0 INTRODUCTION 1
2.0 HAZARD IDENTIFICATION 3
2.1 Review of the Adverse Health Effects of Lead Exposure, With
a Focus on Neurological Effects, As Observed in Animal Studies 4
2.1.1 Approach to Reporting on the Findings of Animal Studies 4
2.1.2 Neurological, Behavioral, and Developmental Health Effects 5
2.1.2.1 Physiological Effects of Lead on the
Neurological System 6
2.1.2.2 Behavioral and Developmental Effects of Lead 10
2.1.2.3 Findings from Specific Studies 13
2.1.3 General Health Effects 17
2.1.3.1 Death 17
2.1.3.2 Systemic Effects 17
2.1.3.3 Immunological Effects 18
2.1.3.4 Reproductinve and Genotoxic Effects 19
2.1.3.5 Carcinogenicity 19
2.1.4 Conclusions from Animal Studies Investigation 19
2.2 Support for the Causality of Adverse Health Effects Due to
Lead Exposure 21
2.2.1 Principles of Causality 21
2.2.2 Causality as Addressed in Longitudinal Studies 24
2.2.3 Conclusions on Causality 27
2.3 The Association Between Blood-Lead Concentration and
IQ Score 28
2.3.1 Linearity and Slope Assumptions 28
2.3.2 Threshold Assumption 32
2.3.3 Verifying the Results of Schwartz (1994) 38
2.4 Impact of Certain Residential Dust Characteristics
on Dust-Lead Exposure 38
2.4.1 Review of Literature: Effects of Chemical Composition
on Lead Bioavailability in Dust 39
2.4.1.1 Research on Lead Bioavailability in
Controlled Animal Studies 40
2.4.1.2 Research on Lead Bioavailability in Soils 40
2.4.2 Review of Literature: Effects of Particle Size
on Lead Bioavailability in Dust 41
2.4.3 Information Gaps, Issues and Conclusions 41
3.0 EXPOSURE ASSESSMENT 43
3.1 The National Survey of Lead and Allergens in Housing 44
3.2 Comparison of Environmental-Lead Levels in the HUD
National Survey with Those of Other Key Studies 53
DRAFT-DO NOT CITE OR QUOTE V August 28,2000
-------
TABLE OF CONTENTS
(Continued)
3.2.1 Characterizing Dust-Lead Loadings on Floors and
Window Sills 55
3.2.1.1 Data Summaries for the §403 Risk Analysis
Versus the Interim NSLAH 56
3.2.1.2 Data Summaries for the §403 Risk Analysis
Versus Three Other Studies 67
3.2.1.3 Calculating National Exceedance Percentages
For Household Average Floor Dust-Lead Loading 81
3.2.1.4 Interpreting the Observed Differences with
Other Studies 88
3.2.1.5 Conclusions of the Dust-Lead Data Comparisons 89
3.2.2 Characterizing Soil-Lead Concentrations 90
3.2.2.1 Data Summaries for the §403 Risk Analysis
Versus the Interim NSLAH 94
3.2.2.2 Data Summaries for the §403 Risk Analysis
Versus Other Studies 102
3.2.2.3 Calculating National Exceedance Percentages
For Yardwide Average Soil-Lead Concentration 120
3.2.2.4 Interpreting the Observed Differences with
Other Studies 134
3.2.2.5 Conclusions of the Soil-Lead Data Comparisons 137
3.3 Evaluation of Soil Pica in Children 138
3.3.1 What is Soil Pica? 138
3.3.2 How Does the §403 Risk Analysis Account for Soil Pica? 143
3.3.3 Prevalence of Soil Pica Behavior 144
3.3.3.1 Literature Review 144
3.3.3.2 Prevalence of Soil Pica Separate from
Paint Pica 145
3.3.4 Estimating the Frequency of Ingestion and Amount
of Soil Ingested by Children Who Exhibit Soil Pica 147
3.3.5 Conclusions on Soil Pica 150
3.4 Characterizing the Population of Children in the Nation's
Housing Stock 150
3.5 Summaries of Dust-Lead Levels on Surfaces Other than Uncarpeted
Floors and Window Sills 152
3.5.1 Distribution of Dust-Lead on Surfaces Other than Floors
And Window Sills 153
3.5.2 Evidence of a Relationship Between Children's Blood-Lead
Concentrations and Dust-Lead on Surfaces Other Than
Floors and Window Sills 165
3.5.3 Implications of the Available Information for
Regulatory Standards 168
3.6 Distribution of Childhood Blood-Lead 170
3.6.1 Evaluation of the HUD Lead-Based Paint Hazard Control
Grant Program ("HUD Grantees") 171
DRAFT-DO NOT CITE OR QUOTE vi August 28,2000
-------
TABLE OF CONTENTS
(Continued)
3.6.2 Evidence of the Impact of Housing Age/Condition
on Blood-Lead Concentration 177
4.0 DOSE-RESPONSE ASSESSMENT 179
4.1 HUD Model 18°
4.1.1 Form of the HUD Model 180
4.1.2 Development of the HUD Model 181
4.1.3 Interpreting Results of Fitting the HUD Model 186
4.1.4 Conclusions 188
4.2 Alternative Multimedia Models for Predicting a Geometric
Mean Blood-Lead Concentration Based on Environmental-
Lead Levels 188
4.3 Supplemental Information on Model-Based Approaches in
the §403 Risk Analysis 191
4.3.1 The "Scaling" Algorithm Used to Determine a
Post-Intervention Blood-Lead Concentration Distribution 191
4.3.2 Adjusting the Empirical Model Parameter Estimates
To Reflect Measurement Error 194
5.0 RISK CHARACTERIZATION 197
5.1 Risk Characterization Sensitivity and Uncertainty Analysis 198
5.1.1 Estimates of Individual Risks from Applying the HUD Model 198
5.1.2 Estimates of Individual Risks from Applying the
Alternative Rochester Multimedia Model 207
5.1.3 Considering Potential Declines in Blood-Lead Concentration
from NHANES III Phase 2 Measures 210
5.1.4 Considering How Baseline Environmental-Lead Levels May
Have Changed Since the HUD National Survey 212
5.1.5 Impact on the Estimated Incidence of IQ Point Decrement
Assuming Certain Thresholds on the IQ/Blood-Lead
Relationship 215
6.0 ANALYSIS OF EXAMPLE OPTIONS FOR THE §403 STANDARDS 219
6.1 Performance Characteristics Analyses 220
6.1.1 Data Used in the Performance Characteristics Analysis 220
6.1.2 Analysis Approach 223
6.1.3 Results Cited in the §403 Proposed Rule 227
6.1.4 Results of Analysis on Specified Sets of Standards 230
6.1.4.1 Analyses performed on 41 Combinations of
Candidate Standards, in Three Iterations 230
6.1.4.2 Considering Only Soil and Dust Standards 245
6.1.4.3 Analysis Involving Only Dust-Lead Standards and
a Standard on the Amount of Deteriorated Paint 253
6.2 Investigating Incidence of Elevated Blood-Lead Concentration
in Housing Units Meeting All Example Options for Standards 260
6.2.1 The Model-Based Approach 260
DRAFT ~ DO NOT CITE OR QUOTE vii August 28.2000
-------
TABLE OF CONTENTS
(Continued)
6.2.2 Example of Applying the Model-Based Approach 261
6.3 Review of Published Information on Post-Intervention
Dust-Lead Loadings 263
6.3.1 Post-Intervention Floor Dust-Lead Loadings 263
6.3.2 Post-Intervention Window Sill Dust-Lead Loadings 265
6.4 Sensitivity and Uncertainty Analyses for Risk Management
Analyses 267
6.4.1 Considering How Baseline Environmental-Lead Levels
May Have Changed Since the HUD National Survey 267
6.4.2 Impact on the Estimated Incidence of IQ Point
Decrement Assuming Certain Thresholds on the
IQ/Blood-Lead Relationship 269
6.4.3 Considering Alternative Assumptions on Post-Intervention
Dust-Lead Loadings 271
6.4.4 Characterizing the Post-Intervention Blood-Lead Distribution
Based on Relative Change from Baseline in the Geometric
Mean and the Probability of a Child's Blood-Lead
Concentration Exceeding 10 //g/dL 274
7.0 REFERENCES 277
LIST OF APPENDICES
Appendix A. Glossary to Section 2.1 A-1
Appendix B. Calculating Average IQ Decrement Assuming a Non-zero
Threshold on the IQ/Blood-Lead Concentration Relationship B-1
Appendix C. Method to Imputing Household Average Environmental-Lead
Levels for Housing Units in the National Survey of Lead and
Allergens in Housing (NSLAH) C-1
Appendix D1. Summaries of Interim Dust-Lead Loading Data from the
National Survey of Lead and Allergens in Housing (NSLAH),
Where Imputed Data are Excluded D1-1
Appendix D2. Summaries of Interim Yard-Wide Average Soil-Lead Concentration
Data from the National Survey of Lead and Allergens in Housing
(NSLAH), Where Imputed Data are Excluded D2-1
Appendix E. Method to Estimating Total Soil-Lead Concentration from
Analytical Results for Fine and Coarse Soil Fractions E-1
Appendix F. Comparison and Contrast of Risk Estimates from the HUD
Model and the Rochester Multimedia Model Developed in
the §403 Risk Analysis F-1
Appendix G. Performance Characteristics Analysis Cited in the
§403 Proposed Rule G-1
Appendix H. Review of Published Information on Post-Intervention Wipe
Dust-Lead Loadings on Floors and Window Sills H-1
DRAFT - DO NOT CITE OR QUOTE
VIII
August 28, 2000
-------
TABLE OF CONTENTS
(Continued)
LIST OF TABLES
Table 2-1. Summary of Lead Exposure Levels and Key Findings for Selected
Animal Studies 14
Table 2-2. Summary of Key Findings from Studies that Investigate the
Relationship Between Blood-Lead Concentration and IQ Score 30
Table 3-1. Differences in Approaches and Outcomes Between the HUD
National Survey of Lead-Based Paint in Housing and the
HUD National Survey of Lead and Allergens in Housing 46
Table 3-2. Estimated Number of Occupied Housing Units in the U.S.
Housing Stock Within Year-Built Categories, According to
Four Recent Surveys and/or Analyses 51
Table 3-3. Estimated Number of Occupied Housing Units in the U.S.
Housing Stock Within Each Census Region, According to
Four Recent Surveys and/or Analyses 52
Table 3-4. Descriptive Statistics of Area-Weighted Average Floor Wipe
Dust-Lead Loadings for Households, As Reported in the
§403 Risk Analysis Versus the Interim NSLAH Data 57
Table 3-5. Descriptive Statistics of Area-Weighted Average Window Sill
Wipe Dust-Lead Loadings for Households, As Reported in the
§403 Risk Analysis Versus the Interim NSLAH Data 57
Table 3-6. Descriptive Statistics of Area-Weighted Average Floor Wipe
Dust-Lead Loadings for Households, Presented bv Housing Age
Category. As Reported in the §403 Risk Analysis Versus
the Interim NSLAH Data 61
Table 3-7. Descriptive Statistics of Area-Weighted Average Window Sill
Wipe Dust-Lead Loadings for Households, Presented by
Housing Age Category. As Reported in the §403 Risk Analysis
Versus the Interim NSLAH Data 62
Table 3-8. Descriptive Statistics of Area-Weighted Average Floor Wipe
Dust-Lead Loadings for Households, Presented bv Census
Region. As Reported in the §403 Risk Analysis Versus the
Interim NSLAH Data 65
DRAFT •- DO NOT CITE OR QUOTE
IX
August 28, 2000
-------
TABLE OF CONTENTS
(Continued)
Table 3-9.
Table 3-1 Oa.
Table 3-1 Ob.
Table 3-11 a.
Table 3-11b.
Table 3-12.
Table 3-13.
Table 3-14.
Descriptive Statistics of Area-Weighted Average Window Sill
Wipe Dust-Lead Loadings for Households, Presented bv Census
Region. As Reported in the §403 Risk Analysis Versus the
Interim NSLAH Data
66
Descriptive Statistics of Area-Weighted Average Floor Wipe
Dust-Lead Loadings for Households, Presented bv Housing Age
and Census Region. As Reported in the §403 Risk Analysis
Versus the Interim NSLAH Data Where No Adjustments Were
Made to Not-Detected Results
70
Descriptive Statistics of Area-Weighted Average Floor Wipe
Dust-Lead Loadings for Households, Presented bv Housing Age
and Census Region. As Reported in the §403 Risk Analysis
Versus the Interim NSLAH Data Where Not-Detected Results
Were Replaced bv LOD/2
71
Descriptive Statistics of Area-Weighted Average Window Sill
Wipe Dust-Lead Loadings for Households, Presented by
Housing Age and Census Region. As Reported in the §403
Risk Analysis Versus the Interim NSLAH Data Where No
Adjustments Were Made to Not-Detected Results
72
Descriptive Statistics of Area-Weighted Average Window Sill
Wipe Dust-Lead Loadings for Households, Presented by
Housing Age and Census Region. As Reported in the §403
Risk Analysis Versus the Interim NSLAH Data Where Not-
Detected Results Were Replaced bv LOD/2
73
Descriptive Statistics of Area-Weighted Average Pre-
Intervention Floor Wipe Dust-Lead Loadings for Households,
As Reported in the §403 Risk Analysis, the HUD National
Survey, and Other Studies
77
Descriptive Statistics of Area-Weighted Average Pre-
Intervention Window Sill Wipe Dust-Lead Loadings for
Households, As Reported in the §403 Risk Analysis, the
HUD National Survey, and Other Studies
78
Descriptive Statistics of Area-Weighted Average Pre-
Intervention Floor Wipe Dust-Lead Loadings for Households,
Presented bv Housing Aoe Category. As Reported in the §403
Risk Analysis, the HUD National Survey, and Other Studies . .
82
DRAFT - DO NOT CITE OR QUOTE
August 28, 2000
-------
Table 3-15.
Table 3-16.
Table 3-17.
Table 3-18.
Table 3-19.
Table 3-20.
Table 3-21 a.
Table 3-21b.
TABLE OF CONTENTS
(Continued)
Descriptive Statistics of Area-Weighted Average Pre-
Intervention Window Sill Wipe Dust-Lead Loadings for
Households, Presented bv Housing Aae Category. As Reported
in the §403 Risk Analysis, the HUD National Survey, and
Other Studies
Page
Estimated Percentages of 1997 U.S. Housing Exceeding
Specified Thresholds of Household Average Dust-Lead Loading
Information on Soil Sampling and Analysis Protocols for
Studies Whose Soil-Lead Data Were Compared to Results
from the §403 Risk Analysis and the HUD National Survey . .
Table 3-22a.
Table 3-22b.
Descriptive Statistics of Yard-Wide Average Soil-Lead
Concentrations for Households, As Reported in the
§403 Risk Analysis Versus the Interim NSLAH Data .
84
88
91
95
Descriptive Statistics of Yard-Wide Average Soil-Lead
Concentrations for Households, Presented bv Housing Age
Category. As Reported in the §403 Risk Analysis Versus
the Interim NSLAH Data
98
Descriptive Statistics of Yard-Wide Average Soil-Lead
Concentrations for Households, Presented bv Census
Region. As Reported in the §403 Risk Analysis Versus the
Interim NSLAH Data
100
Descriptive Statistics of Yard-Wide Average Soil-Lead
Concentrations for Households, Presented bv Housing Age
and Census Region. As Reported in the §403 Risk Analysis
Versus the Interim NSLAH Data Where No Adiustments Were
Made to Not-Detected Results
103
Descriptive Statistics of Yard-Wide Average Soil-Lead
Concentrations for Households, Presented by
Housing Age and Census Region. As Reported in the §403
Risk Analysis Versus the Interim NSLAH Data Where Not;
Detected Results Were Replaced bv LOD/2
Descriptive Statistics of Yard-Wide Average Soil-Lead
Concentrations. According to Study and Within Specific
Subsets of the Sampled Housing Within a Study
104
109
Descriptive Statistics of Average Soil-Lead Concentrations
in Specific Yard Areas and/or for Certain Subsets of the
Sampled Housing Within a Study
DRAFT - DO NOT CITE OR QUOTE
XI
113
August 28, 2000
-------
TABLE OF CONTENTS
(Continued)
Table 3-23a. Descriptive Statistics of Yard-Wide Average Soil-Lead
Concentrations. According to Study and Housing Aoe
Category and Within Specific Subsets of the Sampled
Housing Within a Study 121
Table 3-23b. Descriptive Statistics of Average Soil-Lead Concentrations
in Specific Yard Areas and/or for Certain Subsets of the
Sampled Housing Within a Study, Presented by Housing
Aae Category 126
Table 3-24. Estimated Percentages of 1997 U.S. Housing Exceeding
Specified Thresholds of Yardwide Average Soil-Lead
Concentration 133
Table 3-25. Results of Literature Review on Children's Exposure to
Lead Through Soil Pica 140
Table 3-26. Estimated Rates of Paint and Soil Pica Behavior Reported in the
USLADP Studies, the Rochester Lead-in-Dust Study, and the
Baltimore R&M Study 147
Table 3-27. Alternative Estimates of the Average Number of Children Per
Unit in the 1997 National Housing Stock, by Age of Child 151
Table 3-28. Alternative Estimates of the Average Number of Children in
the 1997 National Housing Stock, by Age of Child and
Year-Built Category, Based on Data Obtained Since the
§403 Risk Analysis 152
Table 3-29. Studies for Which Dust Samples Have Been Collected from
Exterior Areas, Air Ducts, Window Troughs, and Upholstery
for Lead Analysis 154
Table 3-30. Summary of Data from Studies Where Exterior Dust Samples
Were Collected for Lead Analysis 155
Table 3-31. Summary of Data from Studies Where Air Duct Dust Samples
Were Collected for Lead Analysis 159
Table 3-32. Summary of Data from Studies Where Window Trough Dust Samples
Were Collected for Lead Analysis 160
Table 3-33. Summary of Data from Studies Where Upholstery Dust Samples
Were Collected for Lead Analysis 162
DRAFT - DO NOT CITE OR QUOTE
XII
August 28, 2000
-------
Table 3-34.
Table 3-35.
Table 4-1.
Table 5-1 a.
Table 5-1 b.
Table 5-2.
Table 5-3.
Table 5-4a.
TABLE OF CONTENTS
(Continued)
Summary of Children's Pre-lntervention Blood-Lead Concentration
in the HUD Grantees Evaluation According to Blood Collection
Method. Child Age Category, and Grantee (ages 1-2 years only) .
172
Percentage of Children with Elevated Blood-Lead Concentration
(at Pre-lntervention) in the HUD Grantees Evaluation According
to Blood Collection Method, Child Age Category, and Grantee
(ages 1 -2 years only)
Parameter Estimates and Associated Standard Errors for the
Three Alternative Multimedia Models Fitted to Rochester Study
Data to Predict Log-Transformed Blood-Lead Concentration . . .
174
190
Yard-Wide Average Soil-Lead Concentration at Which the
Percentage of Children Aged 1 -2 Years With Blood-Lead
Concentration At or Above 10 //g/dL is Estimated by the
IEUBK Model at 1, 5, or 10%, for Three Assumed Dust-Lead
Concentrations (Table 5-5 in §403 risk analysis report)
201
Yard-Wide Average Soil-Lead Concentration at Which the
Percentage of Children Aged 1 -2 Years With Blood-Lead
Concentration At or Above 10 //g/dL is Estimated by the
HUD Model at 1, 5, or 10%, for Eight Assumed Dust-Lead
Loadings and Two Assumed Geometric Standard Deviations .
Floor Dust-Lead Loadings at Which the Percentage of Children
Aged 1 -2 Years With Blood-Lead Concentration At or Above
10 A/g/dL is Estimated by the HUD Model at 1, 5, or 10%, for
Five Assumed Soil-Lead Concentrations and Two Assumed
Geometric Standard Deviations
201
205
Floor Dust-Lead Loadings at Which the Percentage of Children
Aged 1 -2 Years With Blood-Lead Concentration At or Above
10 //g/dL is Estimated by the Rochester Multimedia Model at
1, 5, or 10%, for Five Assumed Soil-Lead Concentrations, Two
Assumed Window Sill Dust-Lead Loadings, and Two Assumed
Geometric Standard Deviations (expanded version of Table 5-6
in §403 risk analysis report)
206
Uncarpeted Floor Dust-Lead Loadings at Which the Percentage
of Children Aged 1 -2 Years With Blood-Lead Concentration At
or Above 10 //g/dL is Estimated by the Alternative Rochester
Multimedia Model (A) at 1, 5, or 10%, for Fixed Levels of
Yardwide Average Soil-Lead Concentration and Window Sill
Dust-Lead Loading
209
DRAFT » DO NOT CITE OR QUOTE
XIII
August 28,2000
-------
TABLE OF CONTENTS
(Continued)
Table 5-4b.
Table 5-4c.
Table 5-5.
Table 5-6.
Table 5-7.
Table 6-1.
Table 6-2.
Table 6-3.
Table 6-4.
Window Sill Dust-Lead Loadings at Which the Percentage
of Children Aged 1-2 Years With Blood-Lead Concentration At
or Above 10/yg/dL is Estimated by the Alternative Rochester
Multimedia Model (A) at 1, 5, or 10%, for Fixed Levels of
Yardwide Average Soil-Lead Concentration and Uncarpeted
Floor Dust-Lead Loading
209
Yardwide Average Soil-Lead Concentration at Which the
Percentage of Children Aged 1 -2 Years With Blood-Lead
Concentration At or Above 10 //g/dL is Estimated by the
Alternative Rochester Multimedia Model (A) at 1, 5, or 10%,
for Fixed Levels of Dust-Lead Loadings for Uncarpeted Floors
and Window Sills
210
Sensitivity Analysis for the Estimated Baseline Number and Percentage
of Children Aged 1 -2 Years Having Specific Health Effect and
Blood-Lead Concentration Endpoints, Assuming Various Percentage
Declines in Blood-Lead Concentration Since Phase 2 of NHANES III . . .
211
Sensitivity Analysis on How Changes in Household Average
Baseline Dust-Lead Loadings/Concentrations and Soil-Lead
Concentration Impact Pre-1403 Estimates of Health Effect and
Blood-Lead Concentration Endpoints for Children Aged 1 -2 Years .
Sensitivity Analysis on the Assumed Blood-Lead Concentration
Threshold on IQ Decrement and Its Impact on the Pre-§403
Estimates of IQ Decrement Endpoints for Children Aged 1 -2 Years
Definitions of Performance Characteristics Used to Evaluate
How Various Combinations of Environmental-Lead Standards
Classify Housing Units in the Rochester Lead-in-Dust Study
Set of 21 Housing Units in the Rochester Study in Which No
Standard Was Exceeded in at Least One of the 864 Combinations
of Candidate Standards
Results of Performance Characteristics Analysis Performed on
Data for 177 Units in the Rochester Lead-in-Dust Study for
Specified Sets of Standards
Results of Performance Characteristics Analysis Performed on
Data for 184 Units in the Rochester Lead-in-Dust Study for
Specified Sets of Standards
214
217
224
229
232
236
DRAFT -- DO NOT CITE OR QUOTE
XIV
August 28, 2000
-------
Table 6-5.
Table 6-6.
Table 6-7.
Table 6-8.
Table 6-9.
Table 6-10.
Table 6-11.
Table 6-12.
Table 6-13.
Table 6-14.
TABLE OF CONTENTS
(Continued)
Numbers of Housing Units with Missing Data for Four Endpoints
and the Imputed Data Values Assigned to These Units in
This Analysis
Page
Results of Performance Characteristics Analysis Performed on
Data for 205 Units in the Rochester Lead-in-Dust Study for
Specified Sets of Standards
Estimates of Sensitivity and Negative Predictive Value Presented
in Tables 6-3, 6-4, and 6-6
240
241
244
Results of Performance Characteristics Analysis Performed on
Data for Housing Units in the Rochester Lead-in-Dust Study, for
Specified Sets of Candidate Standards for Lead in Dust and
Soil Only
Results of Performance Characteristics Analysis Performed on
Data for Housing Units in the Rochester Lead-in-Dust Study, for
Specified Sets of Candidate Standards for Dust-Lead Loadings
and Observed Amount of Damaged Paint on a Tested Surface .
247
254
Results of Performance Characteristics Analysis Performed on
Data for Housing Units in the Rochester Lead-in-Dust Study, for
Specified Sets of Candidate Standards for Dust-Lead Loadings
and Observed Amount of Damaged Lead-Based Paint on a
Tested Surface
Summaries of Pre- and Post-Intervention Floor Wipe Dust-Lead
Loadings for Housing Groups Within Seven Studies
Summaries of Pre- and Post-Intervention Window Sill Wipe
Dust-Lead Loadings for Housing Groups Within Seven Studies
257
264
266
Sensitivity Analysis on How Changes in Household Average
Baseline Dust-Lead Loadings/Concentrations and Soil-Lead
Concentration Impact Post-§403 Estimates of Health Effect and
Blood-Lead Concentration Endpoints for Children Aged 1 -2 Years
Under a Specified Set of Example Standards
268
Sensitivity Analysis on the Assumed Blood-Lead Concentration
Threshold on IQ Decrement and Its Impact on the Post-§403
Estimates of IQ Decrement Endpoints for Children Aged 1-2 Years,
Under a Specified Set of Example Standards
270
DRAFT - DO NOT CITE OR QUOTE
XV
August 28, 2000
-------
TABLE OF CONTENTS
(Continued)
Table 6-15.
Table 6-16.
Sensitivity Analysis on How Changing the Assumption on the
Post-Intervention Household Average (Wipe) Dust-Lead Loadings
on Floors and Window Sills Impact Post-§403 Estimates (Based
on the Empirical Model) of the Health Effect and Blood-Lead
Concentration Endpoints for Children Aged 1 -2 Years Under a
Specified Set of Example Standards
Estimated Post-§403 Health and Blood-Lead Concentration
Endpoints Under the Original and Alternative Scaling Algorithms
for Characterizing the Post-§403 Blood-Lead Distribution ....
273
276
Figure 3-1.
Figure 3-2.
Figure 3-3.
Figure 3-4.
Figure 3-5.
Figure 3-6.
LIST OF FIGURES
Boxplots of Area-Weighted Average Floor Wipe Dust-Lead Loadings
(/ug/ft2), As Observed in the §403 Risk Analysis (Using HUD National
Survey Data) and in the Interim NSLAH (under 2 approaches to
handling not-detected values)
59
Boxplots of Area-Weighted Average Floor Window Sill Wipe
Dust-Lead Loadings Oug/ft2), As Observed in the §403 Risk
Analysis (Using HUD National Survey Data) and in the Interim
NSLAH (under 2 approaches to handling not-detected values) .
60
Boxplots of Area-Weighted Average Floor Wipe Dust-Lead Loadings
Oug/ft2), by Housing Age Category, As Observed in the §403 Risk
Analysis (Using HUD National Survey Data) and in the Interim
NSLAH (under 2 approaches to handling not-detected values)
63
Boxplots of Area-Weighted Average Floor Window Sill Wipe
Dust-Lead Loadings U/g/ft2), by Housing Age Category,
As Observed in the §403 Risk Analysis (Using HUD National
Survey Data) and in the Interim NSLAH (under 2 approaches
to handling not-detected values)
64
Boxplots of Area-Weighted Average Floor Wipe Dust-Lead Loadings
Ovg/ft2), by Census Region, As Observed in the §403 Risk
Analysis (Using HUD National Survey Data) and in the Interim
NSLAH (under 2 approaches to handling not-detected values)
68
Boxplots of Area-Weighted Average Floor Window Sill Wipe
Dust-Lead Loadings (//g/ft2), by Census Region, As Observed
in the §403 Risk Analysis (Using HUD National Survey Data)
and in the Interim NSLAH (under 2 approaches to handling
not-detected values)
69
DRAFT « DO NOT CITE OR QUOTE
XVI
August 28, 2000
-------
TABLE OF CONTENTS
(Continued)
Figure 3-7. Boxplots of Area-Weighted Average Pre-lntervention Floor Wipe
Dust-Lead Loadings (//g/ft2) for Houses in the HUD National Survey,
Baltimore R&M Study, Rochester Lead-in-Dust Study, and Grantees
Within the HUD Grantees Evaluation 75
Figure 3-8. Boxplots of Area-Weighted Average Pre-lntervention Window Sill
Wipe Dust-Lead Loadings (//g/ft2) for Houses in the HUD National
Survey, Baltimore R&M Study, Rochester Lead-in-Dust Study, and
Grantees Within the HUD Grantees Evaluation 76
Figure 3-9. Boxplots of Area-Weighted Average Pre-lntervention Floor Wipe
Dust-Lead Loadings (//g/ft2) for Houses in the HUD National Survey,
Baltimore R&M Study, Rochester Lead-in-Dust Study, and HUD
Grantees Evaluation, by Age of House Category (pre-1979 only) 79
Figure 3-10. Boxplots of Area-Weighted Average Pre-lntervention Window Sill
Wipe Dust-Lead Loadings (//g/ft2) for Houses in the HUD National
Survey, Baltimore R&M Study, Rochester Lead-in-Dust Study, and
HUD Grantees Evaluation, by Age of House Category (pre-1979 only) 80
Figure 3-11. Estimated Distribution of Household Average Floor Dust-Lead
Loading in the Nation's Housing Stock, and Corresponding Estimates
of the Percentage of Homes Exceeding Specified Thresholds (with
95% Confidence Intervals on the Corresponding Number of Homes,
in Millions), Based on Data from the HUD National Survey (top plot)
and the Interim NSLAH (bottom plot) 87
Figure 3-12. Boxplots of Yard-Wide Average Soil-Lead Concentration (//g/g),
As Observed in the §403 Risk Analysis (Using HUD National
Survey Data) and in the Interim NSLAH (under 2 approaches
to handling not-detected values) 96
Figure 3-13. Boxplots of Yard-Wide Average Soil-Lead Concentration (//g/g),
by Housing Age Category, As Observed in the §403 Risk
Analysis (Using HUD National Survey Data) and in the Interim
NSLAH (under 2 approaches to handling not-detected values) 99
Figure 3-14. Boxplots of Yard-Wide Average Soil-Lead Concentration (//g/g),
by Census Region, As Observed in the §403 Risk Analysis
(Using HUD National Survey Data) and in the Interim NSLAH
(under 2 approaches to handling not-detected values) 101
Figure 3-15. Boxplots of Household Average Soil-Lead Concentrations (//g/g)
for Houses in the HUD National Survey, Baltimore R&M Study,
Rochester Lead-in-Dust Study, and Grantees Within the HUD
Grantees Evaluation 105
DRAFT - DO NOT CITE OR QUOTE XVJi August 28.2000
-------
TABLE OF CONTENTS
(Continued)
Page
Figure 3-16. Summary Statistics of Average Household Soil-Lead Concentrations
U/g/g) for Selected Studies as Compared to Summaries Based
on Data from the HUD National Survey 107
Figure 3-17. Boxplots of Household Average Soil-Lead Concentrations (//g/g)
for Houses in the HUD National Survey, Baltimore R&M Study,
Rochester Lead-in-Dust Study, and HUD Grantees Evaluation,
by Housing Age Category (pre-1979 only) 119
Figure 3-18. Estimated Distribution of Yardwide Average Soil-Lead Concentration
in the Nation's Housing Stock, and Corresponding Estimates
of the Percentage of Homes Exceeding Specified Thresholds (with
95% Confidence Intervals on the Corresponding Number of Homes,
in Millions), Based on Data from the HUD National Survey (top plot)
and the Interim NSLAH (bottom plot) 132
Figure 3-19. Estimated Distribution of Yardwide Average Soil-Lead Concentration
Among Urban Housing in the HUD National Survey, and Corresponding
Estimates of the Percentage of Homes Exceeding Specified Thresholds
(with 95% Confidence Intervals on the Corresponding Number of Urban
Homes in the Nation, in Millions) 135
Figure 3-20. Fitted Regression Models Predicting Children's Blood-Lead
Concentration as a Function of Area-Weighted Arithmetic Average
Floor Dust-Lead Loading (Wipe Collection Method), for the
Various Grantees in the HUD Grantees Evaluation and for
the Rochester Lead-in-Dust Study 175
Figure 3-21. Fitted Regression Models Predicting Children's Blood-Lead
Concentration as a Function of Area-Weighted Arithmetic Average
Window Sill Dust-Lead Loading (Wipe Collection Method), for
the Various Grantees in the HUD Grantees Evaluation and for
the Rochester Lead-in-Dust Study 176
Figure 3-22. Geometric Mean Blood-Lead Concentration Versus Child Age,
As Reported Within the Cincinnati Prospective Lead Study and
Presented According to Housing Age and Condition 178
Figure 5-1. Percentage of Children's Blood-Lead Concentrations, as Predicted
by the IEUBK and HUD Models, That Will Exceed or Equal 10/jg/dL
as a Function of Yard-Wide Average Soil-Lead Concentration and
at Fixed Levels of Dust-Lead Concentrations or Loadings 200
DRAFT - DO NOT CITE OR QUOTE XVJii August 28.2000
-------
TABLE OF CONTENTS
(Continued)
Page
Figure 5-2. Percentage of Children's Blood-Lead Concentrations, as Predicted
by the HUD Model and the Rochester Multimedia Model, That Will
Exceed or Equal 10 yug/dL as a Function of Floor Dust-Lead Loading
for Five Soil-Lead Concentrations and Two Window Sill Dust-Lead
Loadings (Geometric Standard Deviation = 1.6) 203
Figure 5-3. Percentage of Children's Blood-Lead Concentrations, as Predicted
by the HUD Model and the Rochester Multimedia Model, That Will
Exceed or Equal 10 pg/dL as a Function of Floor Dust-Lead Loading
for Five Soil-Lead Concentrations and Two Window Sill Dust-Lead
Loadings (Geometric Standard Deviation = 1.72) 204
Figure 6-1. Example of an Ideal Situation for Establishing Potential Dripline
Soil-Lead Standards 225
Figure 6-2. Example of a Situation Where the Negative Predictive Value and
Sensitivity Equal 100%, but the Positive Predictive Value and
Specificity are Less than 100% 226
DRAFT - DO NOT CITE OR QUOTE XJX August 28,2000
-------
This page intentionally blank.
DRAFT -- DO NOT CITE OR QUOTE XX August 28,2000
-------
1 EXECUTIVE SUMMARY
2 This report is a supplement to the EPA report "Risk Analysis to Support Standards for
3 Lead in Paint, Dust, and Soil" (USEPA, 1998a), which presented the methods and findings of a
4 risk analysis that supported efforts by the U.S. Environmental Protection Agency (EPA) to set
5 regulatory standards for lead levels in dust and soil and to control lead-based paint hazards within
6 most pre-1978 housing and child-occupied facilities. These regulatory standards were mandated
7 through §403 of the Toxic Substances Control Act (TSCA), as specified within Title X, the
8 Residential Lead-Based Paint Hazard Reduction Act of 1992 (42 U.S.C. 4851). The §403 risk
9 analysis provided EPA with a scientific foundation for establishing the regulatory standards. On
10 June 3, 1998, EPA proposed a regulation to establish these standards (40 CFR Part 745); this
11 regulation is referred to as the "§403 proposed rule."
12 In 1998, the Environmental Health Committee of EPA's Science Advisory Board (SAB)
13 conducted a technical review of the §403 risk analysis. In response to this review, EPA deemed
14 that a supplement to USEPA (1998a), referred to as the "§403 risk analysis report," was
15 necessary to provide additional technical analyses and to clarify certain key analyses and findings
16 presented within the report. EPA used the summaries and analyses of data considered in the
17 original §403 risk analysis to prepare responses to the public comments on the §403 proposed
18 rule and to prepare the final rule. Although this supplement also presents summaries and
19 analyses of data made available to EPA since the §403 proposed rule was released, such as
20 interim data from the National Survey of Lead and Allergens in Housing (NSLAH), this type of
21 information is meant only to provide an alternative to the findings presented in the §403 risk
22 analysis report. EPA has not used these new data in efforts to select the hazard standards or
23 levels of concern that are included within the final rule. Furthermore, this supplement does not
24 replace any parts of the original §403 risk analysis, but rather supplements this risk analysis with
25 more detailed analyses on selected topics.
26 The format of this supplement follows that of the §403 risk analysis report, with separate
27 chapters for each phase of the risk assessment (hazard identification, exposure assessment, dose-
28 response assessment, and risk characterization) and a final chapter on risk characterization under
29 example options for standards. The remainder of this executive summary summarizes the key
30 issues and findings presented within each chapter of this supplement.
31 Hazard Identification (Chapter 2)
32 Additional evidence that lead-based paint hazards pose a health risk to children and the
33 magnitude of such risk was provided in the following four areas within this supplement:
34 • Adverse health effects of lead exposure, with a focus on neurological effects, as
35 observed in animal studies (to address the point raised in the SAB review that
36 animal data can support causality effects over and above the available data on
37 humans).
DRAFT-DO NOT CITE OR QUOTE XXi August 28.2000
-------
1 • Evidence supporting causality between lead exposure and adverse health effects.
2 • The association between blood-lead concentration and reduction in intelligence
3 quotient (IQ) score.
4 • The role that dust particle size and the chemical composition of lead compounds
5 in lead-contaminated dust may play in determining the extent to which lead in
6 residential dust is bioavailable to humans.
7 Adverse health effects, especially neurotoxicitv. observed in animal studies. Lead has
8 been observed to have widespread neurotoxic effects, as well as behavioral and cognitive
9 symptoms, in humans. These observations are largely consistent with the findings that have been
10 demonstrated in controlled, dose-response studies on animals (e.g., rodents, dogs, non-human
11 primates). Animal studies are also congruent with observations of lead exposure in humans,
12 suggesting an increased susceptibility of the young brain to lead poisoning (Banks et al., 1997).
13 For example, lead-induced alteration of protein kinase C (PKC) activity in the brain has been
14 shown to correlate with poor performance in several learning tasks in animal studies, and
15 researchers have suggested that some of the learning and memory deficits observed in children
16 are likely to be causally related to the types of PKC activity alterations exhibited in these studies
17 (Chen et al., 1998). In addition, animal studies have provided physiological evidence that many
18 of the effects of lead on the differentiation of the developing nervous system, such as synaptic
19 and dendritic development, and myelin and other nerve structure formation, have the potential to
20 be long-lived (USEPA, 1986). Although animal models do not duplicate the human response to
21 lead exposure, they do serve to provide strong support for expecting certain health effects to
22 occur when humans are exposed.
23 Cause-effect relationship of lead exposure and adverse health effects. The combined
24 weight of human and animal studies provides evidence that lead may be assumed to cause
25 adverse neurological effects in young children. For example, longitudinal studies in humans
26 (e.g., Boston, Port Pirie) have shown that disturbances occur in neurobehavioral development
27 early in life even at low lead exposure levels. These studies have observed effects of lead
28 exposure even after accounting for other demographic factors (e.g., socioeconomic status,
29 maternal IQ) that could affect neurological development. While these other demographic factors
30 tend to be highly correlated with blood-lead levels early in life, the influence that lead exposure
31 has on blood-lead tends to increase with a child's age. This is because over time, measures of a
32 child's lead exposure tend to change at a faster rate than the child's demographic measures, and
33 more recent lead exposures continue to be predictive of a child's current health consequences.
34 The §403 risk analysis assessed these adverse neurological effects in children through measuring
35 average IQ score decrements in the population due to lead exposure.
36 Association between blood-lead concentration and IP score decrement. This supplement
37 provides additional technical justification regarding the assumptions made in the §403 risk
38 analysis on the association between blood-lead concentration and decrement in IQ score.
DRAFT-DO NOT CITE OR QUOTE XXJi August 28,2000
-------
1 While prior investigations into this relationship have considered both linear and log-linear
2 associations, a linear relationship (with positive slope) was used in the §403 risk analysis based
3 on the evidence taken from these investigations and the desire not to unduly over-estimate the IQ
4 decrement associated with children having low blood-lead concentrations.
5 Recent studies and meta-analyses investigating the presence of a threshold in the blood-
6 lead/IQ relationship (e.g., Schwartz, 1994) have concluded either that no non-zero threshold
7 exists, or if one does exist, if may be very low (e.g., less than 1.0 ug/dL; Schwartz, 1993).
8 Researchers who disagree with this conclusion have not reached a clear consensus on a value for
9 this threshold. Several older studies that have suggested high thresholds (e.g., over 10 ug/dL)
10 involved few, if any, children at low blood-lead levels, thereby preventing their ability to provide
11 information on potential health effects at low levels. Other researchers have used visual
12 inspection of data summaries to conclude that thresholds exist at relatively high levels, rather
13 than using statistical inference techniques that would yield more scientifically defensible
14 conclusions. As it would have been necessary to have clear, scientifically defensible evidence of
15 a particular non-zero threshold to justify its adoption within the §403 risk analysis, and based on
16 the findings of recent meta-analyses, the approach taken in the §403 risk analysis was to assume
17 that no threshold exists. Nevertheless, sensitivity analyses documented in this supplement have
18 investigated the impact of assuming a positive threshold on the risk analysis estimates.
19 Effect of residential dust characteristics on the bioavailabilitv of lead in dust. It has been
20 suggested that lead speciation and particle size may affect the bioavailability of lead in dust
21 through their influence on solubility. Therefore, this supplement included an investigation of
22 current knowledge about bioavailability of lead in dust and how this knowledge may impact the
23 rulemaking process.
24 There is relatively little in the literature which examines the relationship between
25 bioavailability and chemical composition specifically for household dust. Animal studies have
26 concluded that different lead compounds are associated with different rates of absorption (e.g.,
27 metallic-lead was associated with low absorption). Otherwise, most of the investigations into the
28 effect of lead speciation on bioavailability primarily address lead in soil. However, soil can have
29 a considerable influence on dust-lead levels. Evidence exists that correlation between blood-lead
30 and soil-lead levels can be influenced by both particle size and chemical composition. Generally,
31 the smaller the particle size, the greater the absorption of lead due to more rapid dissolution in
32 the gastrointestinal tract. Some researchers have also hypothesized that smaller particle sizes can
33 contain higher lead concentrations.
34 While evidence does suggest that particle size and chemical composition can influence
35 the level of bioavailability of lead in dust, the current information base may be inadequate to
36 determine how such factors can reasonably be incorporated into the rulemaking effort.
37 Furthermore, needing to characterize dust in this manner within a risk assessment would likely
38 add to the expense of dust analyses, and dust standards that distinguish between these various
39 characterizations could add considerable complexity to the rule.
DRAFT - DO NOT CITE OR QUOTE XXJii August 28.2000
-------
1 Exposure Assessment (Chapter 3)
2 When using model-based procedures to estimate health risks associated with lead-based
3 paint hazards, the §403 risk analysis used data from the 1989-1990 HUD National Survey of
4 Lead-Based Paint in Housing to characterize lead levels in dust and soil within the nation's
5 housing stock. As mentioned above, more recent data have been made available to EPA since
6 the §403 risk analysis report and the proposed rule were published, and some commenters on the
7 proposed rule suggested that EPA should use these data when available. These data include
8 interim data (collected in 1998 and 1999) for 706 housing units from the NSLAH, an on-going
9 national survey of lead levels in dust and soil in the nation's housing. HUD assigned interim
10 sampling weights to these 706 surveyed units in such a way as to allow interim results that
11 properly incorporate these sampling weights to be nationally representative of occupied housing
12 in which children can possibly reside. Other recently-acquired data that are summarized in
13 Chapter 3 of this supplement include additional data from the Evaluation of the HUD Lead-
14 Based Paint Hazard Control Grant Program ("HUD Grantees") and data from the 1997 American
15 Housing Survey.
16 Comparing dust-lead loadings in the HUD National Survey with those of other studies.
17 Before dust-lead loadings measured in the HUD National Survey could be compared to dust-lead
18 loadings from other studies, the loadings needed to be adjusted to reflect the fact that the HUD
19 National Survey used a vacuum technique rather than a wipe technique to collect the dust
20 samples. The §403 risk analysis included a procedure for performing this adjustment. The
21 resulting dust-lead loadings, adjusted to represent dust samples collected using wipe techniques,
22 tended to be lower than the wipe dust-lead loadings reported in other recent lead exposure studies
23 (e.g., Rochester study, HUD Grantees evaluation), even when taking into account housing age
24 category and Census region. One major exception, however, were data from the interim
25 NSLAH, whose distribution of dust-lead loadings was lower than in the HUD National Survey
26 for both floors and window sills. For example, household average floor dust-lead loadings based
27 on the interim NSLAH data had a median of less than 2 ug/ft2, compared to approximately 5.3
28 ug/ft2 as estimated in the §403 risk analysis based on the HUD National Survey data. The
29 estimated percentage of housing units with average floor dust-lead loadings that exceed 50 ng/ft2
30 (i.e., the proposed floor dust-lead standard) was 6.4% based on HUD National Survey data used
31 in the §403 risk analysis, and 0.9% based on interim data from the NSLAH.
32 Comparing soil-lead concentrations in the HUD National Survey with those of other
33 studies. Data on yardwide average soil-lead concentrations from the HUD National Survey were
34 compared with interim NSLAH data, as well as data for 22 other studies that characterized soil-
35 lead concentrations in urban areas prior to any lead abatement. While geometric mean yardwide
36 average soil-lead concentration was lower in the HUD National Survey relative to most of the
37 other studies reviewed, the distributions of yardwide average soil-lead concentration were similar
38 between the HUD National Survey and the interim NSLAH (although the estimated median was
39 nearly 50% lower in the HUD National Survey versus the interim NSLAH for homes built prior
40 to 1940). Regardless of which national survey is considered, the risk analysis supplement
41 estimates that the yardwide average soil-lead concentration exceeds 2000 ppm in approximately
DRAFT -- DO NOT CITE OR QUOTE XXJV August 28,2000
-------
1 1.7% of housing, approximately 3.3% of housing exceeds an average soil-lead concentration of
2 1200 ppm, and between 11% and 12% of housing exceeds an average soil-lead concentration of
3 400 ppm.
4 Soil pica in children. Because the impact of paint pica (i.e., the purposeful ingestion of
5 paint chips) on blood-lead concentration in the presence of deteriorated lead-based paint is not
6 represented within other environmental exposures to lead, the §403 risk analysis accounted for
7 paint pica as a separate factor when estimating risks. While the analysis did not consider the
8 independent impact of soil pica (i.e., the purposeful ingestion of soil) over and above paint pica,
9 it considered the impact of soil pica as part of the relation between soil-lead concentration and
10 blood-lead concentration. While this supplement does not change the approach taken in the
11 original risk analysis, it documents information obtained on the component of soil-lead exposure
12 that may be attributable to soil pica.
13 Based on what has been found in the literature on studies in which paint pica and soil pica
14 behaviors were characterized and could be separated, approximately 10% to 20% of study
15 children appeared to exhibit soil pica behavior in the absence of paint pica. The frequency of soil
16 pica episodes depends on many factors, including climate, access to bare soil, socioeconomic
17 standing, age of child, and parental supervision. While estimates of the amount of soil ingested
18 during pica episodes can vary widely among mass balance studies (i.e., from 500 to 13,000
19 mg/day), average daily ingestion over a year may be much lower. Because soil pica behavior
20 tends to be episodic in nature, it is currently uncertain whether the amount of lead in soil ingested
21 on an infrequent basis would be sufficient to elevate blood-lead concentration to unsafe levels.
22 Again, it should be noted that the original §403 risk analysis did include an effect of soil pica in
23 the relationship between soil-lead and blood-lead levels; the additional information presented in
24 this supplement addresses how the effect of soil pica is separated from the effect of paint pica.
25 Dust-lead levels on surfaces other than floors and window sills. Comments were received
26 on the §403 proposed rule recommending that EPA establish dust standards for surfaces other
27 than floors and window sills. In response, EPA conducted further analysis on whether surfaces
28 other than floors and window sills might provide significant additional information for risk
29 assessments. EPA reviewed what has been published regarding dust-lead loadings from exterior
30 locations, air ducts, window troughs (also known as window wells), and upholstery, with special
31 attention given to studies that investigated the correlation between such dust-lead levels and
32 blood-lead concentrations.
33 It is difficult to use data from lead exposure studies to characterize the effect of lead
34 exposure from exterior dust on children's blood-lead levels, primarily because studies differ
35 considerably in how they collect exterior dust and how they report the results of lead analysis.
36 For example, some studies have collected both exterior dust and soil for lead analysis but report
37 results that are combined across both media. Also, studies often differ in their sampling
38 approaches (e.g., surface scrapings, vacuum sampling) and in the locations at which the samples
39 are taken. Furthermore, studies frequently have not investigated the specific influence of exterior
40 dust-lead on blood-lead level in the presence of lead in other environmental media. For these
DRAFT -- DO NOT CITE OR QUOTE XXV August 28.2000
-------
1 reasons, along with the fact that lead in exterior dust can be highly correlated with soil-lead, a
2 scientifically defensible standard for lead in exterior dust was not identified, nor was such a
3 standard deemed essential over and above the planned standards for lead in soil and interior dust.
4 This latter conclusion becomes more acceptable if risk assessors are made aware that lead in
5 exterior dust can be associated with adverse health effects in children, and as a result,
6 recommend corrective action in cases where lead is present in exterior dust and is suspected to be
7 an important pathway of exposure (e.g., when children spend a considerable amount of time on
8 hard surfaces immediately outside of the residence).
9 For air ducts and upholstery, there is insufficient data upon which to characterize the
10 role that lead in dust from these surfaces plays in a child's total lead exposure, and therefore, to
11 develop a hazard standard for lead on these surfaces.
12 Because dust from window troughs (i.e., window wells) has historically been sampled in
13 lead exposure studies (along with floors and window sills), and because risk assessors sample
14 dust from window troughs in determining clearance following a lead-based paint intervention,
15 reports of window trough dust-lead levels are generally prevalent in the literature. However,
16 several studies have reported that the association between window trough dust-lead and blood-
17 lead is not statistically significant after taking into account the effects of dust-lead on floors and
18 window sills (for which standards have been proposed under §403). For this reason, along with
19 the likelihood that exceeding a window sill dust-lead standard will prompt a cleaning of leaded-
20 dust from window troughs as well as sills, it does not appear that an additional standard for
21 window troughs is necessary either to identify a home with a hazard or to guide corrective
22 actions. Also, given the correlation in lead levels between window troughs and window sills, it
23 is likely that if more sampling is to be done beyond a minimal risk assessment, more benefit
24 would be obtained from sampling more windows at only the sill rather than fewer windows at
25 both the sill and trough.
26 Distribution of childhood blood-lead. Information on the distribution of blood-lead
27 concentrations in children based on data from the HUD Grantees evaluation was updated to
28 reflect additional data collected through January, 1999. These data provided a means by which
29 estimates based on data from Phase 2 of the Third National Health and Nutrition Examination
30 Survey (NHANES ffl) (i.e., the data used in the §403 risk analysis to characterize blood-lead
31 levels in the nation) could be evaluated. The HUD Grantees evaluation data (under venipuncture
32 blood collection) had a geometric mean of 9.3 ug/dL for children aged 1-2 years and 8.0 ug/dL
33 for children aged 3-5 years. In contrast, the geometric means based on data from Phase 2 of
34 NHANES m were 3.1 ug/dL for children aged 1-2 years and 2.5 ug/dL for children aged 3-5
35 years. Also, 51 percent of children aged 1-2 years sampled via venipuncture methods had blood-
36 lead concentrations at or above 10 ug/dL, compared to the estimates of 5.9% for Phase 2 of
37 NHANES m, 53.8% for the Baltimore R&M study (pre-intervention), and 23.4% for the
38 Rochester Lead-in-Dust study. The trend toward high blood-lead levels in the HUD Grantees
39 evaluation reflects, among other factors, the HUD Grantees program's procedure of selecting
40 high-risk children for monitoring. While NHANES IQ provides the most nationally
DRAFT -- DO NOT CITE OR QUOTE XXVJ August 28,2000
-------
1 representative data on children's blood-lead concentration, it does not provide environmental-
2 lead data that could be used to investigate the effect of environmental-lead exposure on blood-
3 lead concentration. Therefore, other data sources such as the HUD Grantees evaluation must
4 provide this information.
5 Regression modeling of the blood-lead concentration data suggests that the relationships
6 between blood-lead concentration and household average dust-lead loading were relatively
7 consistent across grantees. In particular, these relationships were similar to that observed for data
8 from the Rochester study (i.e., the data used to develop the empirical model presented in Chapter
9 4 of the §403 risk analysis). This conclusion is important in that the data from the HUD
10 Grantees evaluation reflect a much larger geographical area than the Rochester study and
11 represent several types of exposure conditions.
12 This risk analysis supplement includes evidence that housing age and condition play
13 important roles in the likelihood of a resident child having an elevated blood-lead concentration.
14 The association between older housing and the prevalence of lead hazards has been well-
15 documented and is accepted by many experts in residential lead exposure. The level of
16 deterioration is an important variable in the accessibility of lead-based paint hazards to children.
17 Dose-Response Assessment (Chapter 4)
18 The objective of the dose-response assessment in the §403 risk analysis was to develop a
19 statistical procedure to characterize the relationship between environmental-lead exposure and
20 the resulting adverse health effects in young children. This characterization would then be used
21 to estimate health risks at specified environmental-lead levels or over the entire population. The
22 modeling tools used in this characterization were EPA's Integrated Exposure, Uptake, and
23 Biokinetic (IEUBK) model and an empirical model developed especially for the §403 risk
24 analysis from data collected in the Rochester Lead-in-Dust study. In this supplement, additional
25 models were considered to quantify this characterization: a new model developed from
26 epidemiological data collected from 12 lead exposure studies and made available to EPA after
27 the §403 risk analysis was completed, and revisions to the multimedia model developed for the
28 §403 risk analysis using the Rochester study data ("Rochester multimedia model"). In addition,
29 this supplement provides additional detail on specific aspects of the model-based analysis
30 employed within the §403 risk analysis, such as how post-intervention blood-lead concentration
31 distributions are characterized and how measurement error was handled when fitting the
32 empirical model.
33 HUD Model. An additional model for predicting blood-lead concentration as a function
34 of environmental-lead levels became available after the §403 risk analysis report was published.
35 Some commenters on the §403 proposed rule suggested that EPA use this new model. This new
36 model is a log-linear regression model developed on behalf of the U.S. Department of Housing
37 and Urban Development (HUD) from epidemiological data collected from 12 studies. Thus this
38 model is referred to in this supplement as the "HUD Model." The goal of this model was to
DRAFT - DO NOT CITE OR QUOTE XXVJi August 28.2000
-------
1 "estimate the contribution of lead-contaminated house dust and soil to children's blood-lead
2 levels" (Lanphear et al., 1998).
3 When using the HUD model to predict blood-lead concentration as a function of lead
4 levels in various environmental media, this risk analysis supplement has noted several caveats
5 associated with interpreting the predicted blood-lead concentration:
6 • Risks associated with exposure to specific environmental-lead levels (as estimated
7 from the HUD model) are generally not comparable to population-based risks (as
8 estimated by the IEUBK and empirical models in the §403 risk analysis).
9 • The prediction parameters in the HUD model are not independent. Therefore, it is
10 not appropriate to interpret the parameter estimates in the HUD model (or in the
11 models developed for the §403 risk analysis) in isolation.
12 • The HUD model has adjusted for measurement error in certain environmental-
13 lead measures used as input. Therefore, the model assumes that these input values
14 represent "actual" exposure levels. In contrast, the models developed for the §403
15 risk analysis use measured levels as input that would be reported from a risk
16 assessment. Because the §403 standards will be compared to lead exposure
17 measures that are subject to being measured with error, this latter approach is
18 more relevant for rulemaking purposes.
19 Further discussion of the HUD model is provided in Chapters 4 and 5 of this supplement.
20 Risk Characterization (Chapter 5)
21 Health risks associated with current (i.e., baseline) lead exposures for children aged 1 to 2
22 years were characterized in the §403 risk analysis. Both individual risk estimates (i.e., risks
23 associated with specific environmental-lead levels) and population-based risk estimates (i.e.,
24 average risks over the entire nation) were presented. In this supplement, additional sensitivity
25 and uncertainty analysis associated with the baseline risk characterization was performed, where
26 possible alternatives to various approaches taken and assumptions made in the risk
27 characterization were identified and incorporated into the analysis, and the resulting impact on
28 the risk estimates was evaluated.
29 When predictions under the Rochester multimedia model (developed in the §403 risk
30 analysis to characterize individual risks) were compared to predictions under the HUD model,
31 the following general findings were observed:
32 • At very low floor dust-lead loadings (i.e., 1 -5 ng/ft2), the HUD model and the
33 Rochester multimedia model yield similar predictions for the geometric mean
34 blood-lead concentration, which also results in similar predictions for the health-
35 effect endpoints that are calculated directly from this geometric mean (e.g.,
DRAFT -- DO NOT CITE OR QUOTE XXViii August 28,2000
-------
1 percentage of children with blood-lead concentration at or above a specified
2 threshold; average IQ decrement resulting from lead exposure).
3 • The predicted geometric mean blood-lead concentration under the HUD model
4 ranges from 20% to nearly 60% higher than the prediction under the Rochester
5 multimedia model as floor dust-lead loadings increase from 15 to 100 pg/ft2 and
6 as soil-lead concentrations decrease from 2000 ppm to 10 ppm (assuming, for the
7 Rochester multimedia model, that window sill dust-lead loadings are at their
8 estimated national median level). Note that for a fixed value of the geometric
9 standard deviation (GSD) for the blood-lead distribution, the average IQ
10 decrement in the population that is associated with lead exposure is a multiple of
11 the geometric mean (as calculated in the §403 risk analysis). Therefore, similar
12 differences in predictions between the two models would occur for average IP
13 decrement.
14 • If the geometric standard deviation (GSD) associated with the blood-lead
15 distribution is fixed, then as floor dust-lead loadings increase beyond 10 ug/ft2,
16 the predicted percentage of children with blood-lead levels at or above 10 ue/dL
17 increases at a much faster rate under the HUD model (at a constant soil-lead
18 level). For example, if window sill dust-lead loading is at its estimated national
19 median and soil-lead concentration is below 2000 ppm, the predicted percentage
20 under the HUD model is at a minimum twice as large as the prediction under the
21 Rochester multimedia model. This difference in predictions gets even greater as
22 the assumed soil-lead concentration gets lower. For example, at a GSD of 1.6, a
23 floor dust-lead loading of 100 ug/ft2, and a soil-lead concentration of 10 ppm, the
24 prediction is over 7 times higher for the HUD model compared to the Rochester
25 multimedia model (13.1% versus 1.76%).
26 Other findings within the additional sensitivity analyses performed to support the baseline
27 risk characterization were as follows:
28 • If it is assumed that a 50% across-the-board decline in blood-lead concentration
29 has occurred relative to the distribution portrayed by data from Phase 2 of
30 NH ANES m, the estimated number of children whose blood-lead concentration
31 was at or above 20 ng/dL declined by 95% (from 46,800 to 2,130), while the
32 estimated number at or above 10 ug/dL was reduced by nearly 90% (from 458,000
33 to 46,800). The estimated average IQ decrement in the population due to lead
34 exposure is cut in half under this assumption (from 1.06 to 0.53 points).
35 • Model-based baseline risk estimates were calculated for various alternative
36 assumptions on the percentage decline in dust-lead and soil-lead levels that may
37 have occurred in the housing stock since the HUD National Survey was
38 conducted. Risk estimates under the empirical model seemed to be more sensitive
39 to these changes than the estimates under the BEUBK model, and reductions in
DRAFT -- DO NOT CITE OR QUOTE XXJX August 28,2000
-------
1 soil-lead concentration seemed to have more of an impact on reducing these risk
2 estimates than reductions in dust-lead levels. Baseline estimates for the
3 percentage of children with blood-lead concentrations at or above 10 pg/dL were
4 45% lower under the empirical model when both dust-lead and soil-lead levels
5 were decreased by 50%, with smaller declines occurring for less drastic total
6 reductions in the environmental-lead levels.
7 • In an effort to determine whether an assumption of no threshold made in the §403
8 risk analysis was particularly sensitive to the risk estimates, baseline estimates of
9 the IQ-related health effect endpoints were calculated under the assumption that
10 specified non-zero thresholds exist in the relationship between blood-lead
11 concentration and IQ score decrement. While the §403 risk analysis estimated an
12 average IQ decrement of 1.06 points occurs due to lead exposure across the
13 population of children aged 1-2 years, this average declines by approximately 44%
14 under a assumed threshold of 2 ug/dL (0.588 points) and by 90% under a
15 threshold of 8 ug/dL (0.103 points).
16 Analysis of Example Options for the §403 Standards (Chapter 6)
17 Prompted by public comments on the §403 proposed rule and risk analysis, this
18 supplement included the following on methods used in the §403 rulemaking process to evaluate
19 candidate hazard standards and levels of concern:
20 • Detailed information on performance characteristics analyses (also known as
21 sensitivity/specificity analysis), used by EPA to help establish levels of concern
22 within the §403 rule.
23 • Characterizing the extent to which children with elevated blood-lead
24 concentrations reside in homes where no candidate standard is met or exceeded
25 (i.e., children who would be "missed" by a specified set of candidate standards),
26 as part of the candidate standards evaluation process.
27 • An additional investigation into the assumptions made in the risk management
28 study on post-intervention dust-lead loading (40 |jg/ft2 on floors, 100 ug/ft2 on
29 window sills).
30 • Additional sensitivity and uncertainty analyses for the analyses performed and
31 documented within Chapter 6 of the §403 risk analysis, including alternative
32 assumptions on baseline and post-intervention environmental-lead levels.
33 Performance characteristics analysis. Performance characteristics analysis is a non-
34 modeling approach (based on calculating conditional probabilities) to assessing how often a
35 specified set of candidate hazard standards would "trigger" interventions in housing units within
36 the studies in question and the extent to which these units contained a child with an elevated
DRAFT -- DO NOT CITE OR QUOTE XXX August 28.2000
-------
1 blood-lead concentration (^ 10 ug/dL). Data from the Rochester Lead-in-Dust study were used in
2 these analyses. This supplement contains a detailed discussion of how to interpret performance
3 characteristics analysis and provides additional information on analysis results that were cited in
4 the §403 proposed rule. Furthermore, this supplement presents the results of follow-on
5 performance characteristics analyses which EPA considered when responding to public
6 comments and in preparing the final rule. While the analysis presented in the proposed rule was
7 based on data for 77 housing units in the Rochester study, additional assumptions made to the
8 soil-lead data permitted up to 184 units to be represented among the data analyzed in the follow-
9 on performance characteristics analyses. One goal of these analyses was to identify those sets of
10 candidate dust-lead loading standards for which the analysis estimated that no more than 5% of
11 children living in housing units with environmental-lead levels below the standards would have
12 elevated blood-lead concentrations.
13 Investigating incidence of elevated blood-lead concentrations in homes where no
14 candidate standard is met or exceeded. As an alternative to the performance characteristics
15 analysis, a model-based approach was developed to determine the likelihood of a child with
16 elevated blood-lead concentration residing in a housing unit that exceeds none of a given set of
17 candidate standards. This approach was designed to use data from the Rochester study and to
18 yield results that would be directly comparable to those from the performance characteristics
19 analysis.
20 Review of published information on post-intervention dust-lead loadings. To evaluate
21 the performance of a given set of candidate standards in reducing population-based health risks
22 to lead exposure, the §403 risk analysis needed to make assumptions on lead levels in dust and
23 soil that would occur after performing interventions that would be prompted by exceeding the
24 example standards. Assumptions made on post-intervention dust-lead loadings (40 ug/ft2 for
25 floors, 100 ug/ft2 for window sills) within the §403 risk analysis were evaluated in this
26 supplement in a detailed review of results from studies that evaluated abatement effectiveness.
27 In the reviewed studies, geometric mean or median floor dust-lead loadings were
28 generally at or below 41 ug/ft2 over periods ranging from 6 months to 6 years post-intervention,
29 with several studies reporting levels below 21 ug/ft2 at follow-up periods ranging from
30 12 months to 2 years. Of the eight grantees participating in the HUD Grantees evaluation that
31 had post-intervention floor dust-lead loadings available at 12 months post-intervention, four had
32 median values for these loadings that were at or below 10 ug/ft2. Median pre-intervention floor
33 dust-lead loadings for these four grantees ranged from 9 to 26 fig/ft2. For post-intervention
34 window sill dust-lead loadings, geometric means or medians ranged from 24 Mg/ft2 to 958 ug/ft2
35 in the reviewed studies. Most of the study groups had geometric mean or median post-
36 intervention window sill dust-lead loadings below 100 ug/ft2, while a few were at or below 51
37 ug/ft2.
38 As a result of the post-intervention dust-lead loadings review, a sensitivity analysis
39 documented in this supplement applied the methods developed in the risk analysis under
40 alternative post-intervention floor dust-lead loadings of 10 and 25 ug/ft2 and post-intervention
DRAFT -- DO NOT CITE OR QUOTE XXXJ August 28.2000
-------
1 window sill dust-lead loadings of SO and 75 ug/ft2. Results of this sensitivity analyses indicated
2 that while more housing units may be assumed to achieve reductions in average dust-lead loading
3 as a result of lowering the post-intervention dust-lead loading assumptions, the corresponding
4 reduction in the estimated blood-lead concentration and health effect endpoints appeared to be
5 modest, especially compared to the reduction that occurs from pre-intervention conditions.
DRAFT -- DO NOT CITE OR QUOTE XXXJi August 28,2000
-------
1 1.0 INTRODUCTION
2 On June 3, 1998, the U.S. Environmental Protection Agency (EPA) proposed regulation
3 to establish standards for lead-based paint hazards in most pre-1978 housing and child-occupied
4 facilities (40 CFR Part 745, "Lead; Identification of Dangerous Levels of Lead; Proposed Rule").
5 EPA proposed these standards in accordance with Section 403 of the Toxic Substances Control
6 Act (TSCA), as amended by Title X, the Residential Lead-Based Paint Hazard Reduction Act of
7 1992. The proposed standards are as follows:
8 • Dust-lead hazards: Household average dust-lead loadings equal to or exceeding 50
9 Mg/ft2 °n uncarpeted floors and 250 ug/ft2 on window sills, assuming wipe
10 collection techniques for dust;
11 • Soil-lead hazards: Total lead levels equal to or exceeding 2,000 ppm based on a
12 yard-wide average soil-lead concentration
13 • Hazardous lead-based paint: Lead-based paint in poor condition, defined as
14 follows:
15 • More than 10 ft2 of deteriorated paint on exterior components with large
16 surf ace areas
17 • More than 2 ft2 of deteriorated paint on interior components with large
18 surface areas
19 • Deteriorated paint consisting of more than 10% of the total surface area of
20 exterior or interior components with small surface areas.
21 These standards, a focal point of the Federal lead program, identify the presence of lead-based
22 paint hazards, defined within TSCA Section 401 as the condition of lead-based paint and the
23 levels of lead-contaminated dust and soil that "would result" in adverse human health conditions.
24 To provide a scientific basis for selecting the §403 standards, EPA conducted a risk
25 analysis to assess the health risks to young children (aged 1-2 years) from exposures to lead-
26 based paint hazards, lead-contaminated dust, and lead-contaminated soil in the nation's housing.
27 This risk analysis also documented EPA's approach to estimate the reduction in these risks
28 following promulgation of the §403 standards and applied this methodology to evaluate example
29 options for the §403 standards. Finally, the risk analysis provided estimates of the numbers of
30 homes and children that would be affected by various example standards. EPA published this
31 risk analysis in June, 1998, in a document hereby referred to as the "§403 risk analysis report"
32 (USEPA, 1998a).
33 A period of public comment followed publication of the §403 proposed rule, extending to
34 March, 1999. Several comments received during this period requested additional analyses and
35 investigation. In addition, the Environmental Health Committee of EPA's Science Advisory
36 Board (SAB) performed a review of the technical aspects of the §403 risk analysis, the §403
37 economic analysis, and the proposed rule. While the SAB concluded that many approaches taken
DRAFT - DO NOT CITE OR QUOTE 1 August 28, 2000
-------
1 in the §403 risk analysis were technically sound and scientifically defensible, their final report
2 provided detailed comments and recommendations for additional investigation and analysis to be
3 considered when preparing a final rule (USEPA, 1998b).
4 This report is a supplement, or addendum, to the §403 risk analysis report. It contains
5 additional information obtained since the report was published that further supports the findings
6 and conclusions made in that report. It also contains the results of additional analyses requested
7 by the SAB and by key public comments. This supplement does not replace any parts of the
8 original risk analysis, but rather supplements the original risk analysis with more detailed
9 analyses on selected topics.
10 Reflecting its close ties to the original §403 risk analysis report, this supplement contains
11 the same chapters as those found in the §403 risk analysis report. These chapters represent the
12 different components of the risk analysis: Hazard Identification, Exposure Assessment, Dose-
13 Response Assessment, Risk Characterization, and Analysis of Example Options for the §403
14 Standards. The additional analyses and investigations presented in this supplement are found
15 within the specific chapters to which their findings contribute. Each analysis or investigation is
16 presented as an independent module within each chapter. See the chapter introductions for the
17 contents of each chapter and the motivation for each analysis being presented. Furthermore, the
18 reader is referred to the §403 risk analysis report for details on the risk analysis approach and
19 findings.
DRAFT -- DO NOT CITE OR QUOTE 2 August 28, 2000
-------
1 2.0 HAZARD IDENTIFICATION
2 Chapter 2 of the §403 risk analysis report presented information on the toxicity of lead,
3 through a discussion of how body-lead burden is measured, how lead works in the body, the
4 resulting adverse health effects, and populations at risk. This chapter introduced the endpoints
5 used in the risk analysis to represent the adverse health effects resulting from lead exposure and '
6 to estimate the benefits of the §403 rule. These endpoints included the likelihood of exceeding a
7 specified blood-lead concentration threshold (10 or 20 ug/dL), the likelihood of achieving a
8 specified IQ score decrement as a result of lead exposure (1,2, or 3 points), the likelihood of
9 achieving an IQ score less than 70 points due to lead exposure, and the average IQ score
10 decrement in the population that results from lead exposure. The blood-lead concentration
11 thresholds were among those established by the Centers for Disease Control and Prevention
12 (CDC) as levels of concern. The IQ-related endpoints represented measures of the neurological
13 effects of lead exposure. The representative population upon which the §403 risk analysis
14 focused was children aged 1-2 years, as it was considered the most appropriate age range for the
15 estimation of health effects.
16 In this chapter of the supplemental report, the results of four additional investigations into
17 hazard identification are presented:
18 • Section 2.1: A review of the adverse health effects of lead exposure, with a focus
19 on neurological effects, as observed in animal studies.
20 • Section 2.2: Support for the causality of adverse health effects due to lead
21 exposure.
22 • Section 2.3: Characterizing the relationship between blood-lead concentration
23 and IQ score.
24 • Section 2.4: Documenting what is known about the role that dust particle size and
25 the chemical composition of lead compounds in lead-contaminated dust may play
26 in determining the extent to which lead in residential dust is bioavailable to
27 humans.
28 The motivation for including each of these sections into this chapter is presented within the
29 introduction to each section.
DRAFT -- DO NOT CITE OR QUOTE 3 August 28,2000
-------
1 2.1 REVIEW OF THE ADVERSE HEALTH EFFECTS OF LEAD EXPOSURE.
2 WITH A FOCUS ON NEUROLOGICAL EFFECTS.
3 AS OBSERVED IN ANIMAL STUDIES
4 The §403 risk analysis used data from human exposure studies to characterize the
5 relationship between environmental-lead levels and measures of the various blood-lead
6 concentration and health effect endpoints. However, as the SAB's review of the §403 risk
7 analysis points out (USEPA, 1998b), causality is difficult to establish using human studies alone,
8 due to the potential for confounding factors being present. For example, while average IQ score
9 may differ significantly between one group of children with low blood-lead concentration and
10 another group with elevated blood-lead concentration, the reason for this difference may not be
11 solely due to lead exposure, but to demographic and other factors that cannot be controlled
12 completely by the researcher. The factors left uncontrolled in the analyses of data from human
13 exposure studies contribute to large uncertainties associated with these analyses. While ethical
14 considerations preclude the use of humans in controlled lead-exposure experiments, a substantial
15 amount of published literature is available on controlled lead-exposure experiments involving
16 animals. Such studies use animals that are chosen from a homogeneous population, reared under
17 identical conditions and randomly assigned to groups where the mode, duration, and amount of
18 lead exposure is controlled within each group (these groups can include one or more control
19 groups). Therefore, in a well-controlled animal study, the presence of significant differences
20 between dose groups can be inferred to be the result of lead exposure at certain doses.
21 To address the SAB's recommendation to consider the findings of "... animal data, since
22 they support human data by establishing causality, due to the absence of confounding variables,
23 and potential mechanisms for adverse health effects" (USEPA, 1998b), this section presents the
24 key findings of animal studies that have investigated the impact of lead exposure on adverse
25 health effects. The major lead-induced adverse health effects noted in humans, including
26 neurological, neurodevelopmental, immunological, and systemic (e.g., cardiovascular,
27 hematological, and renal effects) have been demonstrated in controlled, dose-response studies in
28 rodents, dogs, and/or non-human primates. While this section recognizes the variety of adverse
29 health effects, its primary focus is on the neurological effects of lead, as the §403 risk analysis
30 has recognized that young children are most susceptible to neurological effects due to their
31 developing central nervous systems. In particular, this section also documents those animal
32 studies who have observed adverse health effects at low doses of lead.
33 A glossary of the terms used in this section to discuss health effects can be found in
34 Appendix A.
35 2.1.1 Approach to Reporting the Findings of Animal Studies
36 This subsection reviews the incidence of adverse health effects associated with lead
37 exposure, as reported in publications on animal studies. The emphasis of this review is on the
38 neurological, developmental, and neurobehavioral effects of lead. The findings and conclusions
39 from the two review publications which the SAB's review (USEPA, 1998b) cites as "important
40 references" on animal studies data (Rice, 1996; Cory-Slechta et al., 1997), as well as other key
DRAFT -- DO NOT CITE OR QUOTE 4 August 28, 2000
-------
1 review references (e.g., USEPA, 1986; USDHHS, 1999) are included in this chapter. In addition,
2 certain relevant publications referenced within these reviews and the §403 risk analysis report
3 (USEPA, 1998a) were also identified and reviewed for this chapter. Finally, to help identify any
4 key articles that may have been published since these review references were prepared and
5 published, a review of the scientific literature over the last five years was conducted. The search
6 strategy for this literature search was on the key words "lead," "effect," "exposure,"
7 "neurologic," "behavior," "development," "teratology," and "animal" as whole or root words.
8 Upon review of abstracts identified from this literature search, articles found to be relevant to the
9 objectives of this report were obtained and reviewed, with high priority placed on study reviews.
10 Section 2.1.2 focuses on the findings of animal studies on the neurological,
11 developmental, and neurobehavioral effects of lead. This section contains an overview of the
12 physiological consequences of these effects, along with a review of general findings of studies
13 investigating these types of health effects. In addition, certain animal studies are discussed in
14 greater detail with summaries of their design and conclusions.
IS Lead has been documented to have considerably more effects on the health of humans
16 and animals than just neurological effects. Therefore, Section 2.1.3 presents general health effect
17 information and how it relates to lead exposure, as observed in animal studies. Information in
18 this section is organized according to the type of health effect.
19 2.1.2 Neurological, Behavioral, and Developmental
20 Health Effects
21 Lead has been observed to have widespread neurotoxic effects, as well as to cause
22 behavioral and cognitive symptoms, in humans. These effects are largely consistent with results
23 of morphological, electrophysiological, biochemical, and behavioral studies on animals.
24 Although lead toxicity research has isolated many of the specific neurological effects of lead, it is
25 generally considered to be a relatively indiscriminate toxin within the neurological system. This
26 consideration is largely due to the ability of lead to disturb several fundamental biotic processes
27 such as cellular metabolism and energy production, ion transport across membranes, and protein
28 function. In addition, the neurological effects of lead frequently have been observed to occur as a
29 series of interrelated events. Thus, lead poisoning is likely to cause simultaneous and interrelated
30 disturbances in a number of processes within the nervous system. As an additional consequence,
31 the ability to separate the direct and indirect effects of lead on the neurological system is often
32 difficult (Banks et al. 1997).
33 Research, based on both human observation and animal studies, indicates that relatively
34 low doses of lead can adversely affect both the peripheral and central nervous systems, while
35 high lead exposures can result in acute lead encephalopathy, and may ultimately lead to death.
36 Lead induced damage to the brain and nervous system may be manifested as various and diverse
37 developmental symptoms, including both behavioral and cognitive impairments.
DRAFT -- DO NOT CITE OR QUOTE 5 August 28, 2000
-------
1 In the following sections, the physiological effects of lead on the brain and nervous
2 system, and subsequent effects on development and behavior, will be discussed. As these effects
3 will be presented as evidenced by experimental animal research, caution must be taken when
4 extrapolating these findings to humans, as well as to other species. For example, lead neuropathy
5 in rats is primarily characterized by demyelination of nerves, while cats, rabbits, and humans
6 generally show damage to the axons of nerves (Davis et al., 1990). Developmental age of the
7 brain also varies between animal species and humans. For example, at birth, the rat brain is
8 relatively less developed and is roughly equivalent to the human brain at 5-6 months of gestation
9 (Winneke et al., 1996). Furthermore, although rats and possibly monkeys tend to have higher
10 tolerances for the general toxic effects of lead exposure relative to humans (i.e., they require a
11 higher exposure level to reach an equivalent blood toxicity level), the lowest blood-lead levels at
12 which lead-induced developmental / neurobehavioral effects have been observed in animals and
13 in humans are reported to be similar in magnitude (Banks et al., 1997; Davis et al., 1990;
14 USDHHS, 1999).
15
16 2.1.2.1 Physiological Effects of Lead on the Neurological System.
17 Overview. As lead is known to adversely affect several universal processes within
18 biological systems, the symptoms of lead poisoning have been observed in many cell types,
19 tissues, and organs within the neurological system.
20 On the cellular level, lead has been observed to cause disruption of mitochondria!
21 function (i.e., cellular metabolism), damage to cell structural components (e.g., microtubules),
22 and damage to glial cells and the myelin sheaths and axons of nerve cells (USEPA, 1986).
23 Electrophysiological and biochemical processes at the cellular level may also be disrupted by
24 lead exposure. Specific alterations of these processes reported in animal studies include
25 impairment of synaptic events and neuron function, interference with neurotransmitter function,
26 and protein activity inhibition (e.g., enzymes and hormones) (USEPA, 1986). Much of lead's
27 role in cellular level dysfunctions is suggested by many researchers to be attributed to its
28 interference with calcium-mediated processes (Banks et al., 1997). Calcium is an important ion
29 in many biological systems, and is specifically involved in neurological phenomena such as
30 enzyme-protein activation, secondary messenger regulation of metabolic pathways, membrane
31 potential/ion channel regulation, and neurotransmitter release. Lead ions, as they are similar to
32 calcium ions in both size and charge, can substitute for calcium and thus competitively interfere
33 with these types of calcium-mediated cell processes.
34 Disruption of cellular processes can eventually result in damage to tissues and organ
35 systems within the neurological system. Reported effects of lead in animals at the organ and
36 system level within the central nervous system include compromise of the blood-brain barrier,
37 disruption of the limbic system and cerebral cortex, and damage to the cerebellum (Banks et al.,
38 1997; USEPA, 1986). Animals exposed to lead in early post-natal life have also exhibited
39 reductions and delays in development of various brain regions, including the hippocampus and
40 cerebral cortex (Banks et al., 1997; USDHHS, 1999). The limbic system (e.g., hippocampus) is
41 of particular interest in lead toxicity studies as it is key in many of the processes that appear to be
42 affected by lead poisoning, including emotion, motivation, behavior, memory, and various
DRAFT--DO NOT CITE OR QUOTE 6 August 28,2000
-------
1 autonomic functions. Rice (1996), for example, observed that lead exposed animals showed
2 learning and memory impairment on a variety of tasks. Some researchers have even suggested
3 that symptomatic similarities between lead toxicity and other experimental limbic system
4 disruption indicate that the limbic system is a target site for lead toxicity in the brain (Walsh and
5 Tilson, 1984, as cited in Banks et al., 1997).
6 In the peripheral nervous system, visual and auditory system functioning has also been
7 reported to be impaired by lead exposure in animal studies. A review study by USEPA (1986)
8 reported demyelination of peripheral nerves in adult rats exposed to relatively high levels of lead.
9
10 Blood-brain barrier. Banks et al. (1997) reviewed animal studies that examined the
11 effect of lead exposures on the blood-brain barrier that regulates the movement of chemical
12 substances in and out of the brain. In studies of rats exposed to relatively high lead levels, higher
13 concentrations of lead were observed in the barrier capillaries than were observed in the brain as
14 a whole. USEPA (1986) reviewed studies which provided evidence that lead transport within the
15 brain is by the same mechanisms as calcium transport. Thus, as the capacity for calcium
16 transport (specifically, into neural cell mitochondria) is known to be much higher in the brain
17 than other body tissues, a lead accumulation in brain capillaries is not unexpected (USEPA,
18 1986). Acute lead toxicity as lead accumulations in brain capillaries may disrupt barrier
19 permeability, allow greater influxes of water, ions, and other substances, and result in swelling of
20 the brain (encephalitis) (USEPA, 1986). Although lower lead exposure levels (< 40//g/dL) have
21 not been reported to result in specific damage to the barrier or in disproportionate accumulation
22 of lead in the capillaries of the blood-brain barrier (Banks et al., 1997), developing brains have
23 been suggested to be particularly susceptible to the transfer of even very low levels of lead from
24 the blood into the brain since the blood-brain barrier is not yet fully functional (Altmann et al.,
25 1993, as cited in Winneke et al., 1996; Cory-Slechta, 1997).
26 General cellular processes. Several in vitro studies with animal cells have provided
27 evidence that a major site of lead interference is with cellular metabolism and energy transfer via
28 disruption of the normal mitochondrial ion gradient (Banks et al., 1997; USEPA, 1986). In
29 addition, mitochondria in the cerebellum and in developing brains of animals have been observed
30 to show a greater sensitivity to lead disruption than mitochondria in other body tissues or at other
31 ages (USEPA, 1986). This has been suggested to provide a possible explanation for one of the
32 root causes of the greater sensitivity of the neurological system, and of the young in particular, to
33 lead poisoning (USEPA, 1986). Impairment of mitochondrial energy production subsequently
34 can affect many other energy-requiring cellular processes, such as protein synthesis, lipid
35 synthesis, and membrane integrity. In animal studies, the normal development of proteins in
36 neurons was altered in rats exposed to lead perinatally (USDHHS, 1999). Other studies reviewed
37 by Banks et al. (1997) indicate that moderate levels of lead can interfere with microtubule
38 formation (in vitro animal cell studies) and the formation of the myelin sheath in neurons in both
39 the central and peripheral nervous systems of lead exposed rodents.
40 Lead can also interfere directly with calcium-mediated cell processes. This may include
41 disruption of protein function (e.g., enzyme regulation of cell growth and differentiation), ion
DRAFT - DO NOT CITE OR QUOTE 7 August 28, 2000
-------
1 transport systems across membranes, and membrane potentials (Banks et al, 1997; USEPA,
2 1986). For example, hippocampal protein kinase C (PKC) activity, which has been correlated
3 with performance in several learning tasks, was observed to be altered by relatively low internal
4 lead exposures (<30 (J.g/dL) in postnatal rats (Chen et al., 1998). This study is discussed in
5 further detail in Section 2.1.2.3.
6 Neuron and neurotransmitter function. Lead-induced disruption of cellular processes
7 may result in neuron dysfunction. According to a body of experimental animal research
8 reviewed by Banks et al. (1997), evidence exists that moderate to low levels of lead exposure can
9 impair synapse formation in the hippocampus during postnatal development of rats, in the visual
10 cortex of primates, and in the frontoparietal cortex of guinea pigs, and also appears to interfere
11 with synaptic transmission in fetal rat hippocampal neurons by blocking postsynaptic receptors.
12 USEPA (1986) reviewed numerous studies in which rats exhibited morphological effects such as
13 decreased glial cell and synaptic density, and delayed maturity of synapses in neurons of the
14 cerebral cortex with lead exposure. Several of the reviewed studies also reported abnormal
15 development of neuron dendrites and persistent impairment of electrical activity (i.e., reduced
16 firing rates) in the cerebellum of cats and rats exposed perinatally to low levels of lead (Banks et
17 al., 1997). The cerebellum is responsible for the regulation and coordination of complex
18 voluntary muscular movement, and is also implicated in cognitive attention-switching activity.
19 The cerebral cortex is largely responsible for higher brain functions, including sensation,
20 voluntary muscle movement, thought, reasoning, and memory. Thus, damage to neurons in these
21 areas may explain some attention deficits, learning and memory impairments, and disturbances in
22 motor coordination that have been observed in behavioral studies of lead exposure (Banks et al.,
23 1997).
24 Lead has also been observed to interfere with the release and uptake of neurotransmitters
25 in various in vitro and in vivo animal studies, including the inhibition of neurotransmitter release
26 from calcium sensitive voltage channels in snails, and alteration of synthesis and turnover rates
27 of various neurotransmitters in different regions of the brain (hippocampus, cerebellum,
28 hypothalamus, brainstem) in lead-exposed rats (Banks et al., 1997; USEPA, 1986; Nagymajtenyi
29 et al., 1998). Altered activity of various neurotransmitters has been observed in studies involving
30 rats exposed to lead prenatally and postnatally (USDHHS, 1999). In the reviewed literature, lead
31 has generally been reported to have highly variable and non-specific effects on the various
32 neurotransmitter systems, possibly due to its more general effect on metabolic processes (Banks
33 et al., 1997).
34 Cory-Slechta (1997) conducted a review of studies linking disruptions in neurotransmitter
35 systems (i.e., dopaminergic (DA) and glutamatergic (GLU) systems) with behavioral and
36 cognitive impairments in lead exposed animals. The author cites previous research which
37 indicates that the DA and GLU neurotransmitter systems are critical to various cognitive
38 functions, and also are sensitive to lead-induced disruptions. In addition, the author reviews
39 studies reporting non-lead related disruptions of these systems and the concurrent appearance of
40 behavioral symptoms that are very similar to symptoms of lead toxicity. However, the author
41 also concedes that much of the historical research linking behavioral impairments to biochemical
DRAFT - DO NOT CITE OR QUOTE 8 August 28, 2000
-------
1 effects is based solely on correlations and that relationships are complicated by the fact that any
2 given behavioral symptoms may have multiple physiological mechanisms. Cory-Slechta (1997)
3 summarized several studies conducted in her own lab on rats exposed, postweaning (i.e., at 21
4 days of age), to lead (0,50 and 250 ppm lead as lead acetate) in drinking water for varying
5 durations of time. Results of these studies indicated that the DA neurotransmitter system is
6 vulnerable to lead-induced modifications, such as impaired regulation of DA synthesis and
7 release. These modifications to DA system function were suspected to contribute to alterations
8 in response control (fixed interval schedule-controlled behavior) that were observed in lead-
9 exposed rats in the reviewed studies. Results indicated that the GLU system was also involved in
10 lead-induced impaired learning, primarily as manifested by an increase in perseverative errors in
11 lead-exposed rats, relative to controls. There was no evidence that the DA system was involved
12 in learning accuracy. The author suggests that confirmation of the involvement of the GLU and
13 DA systems in lead-induced effects, could also have implications extending beyond cognitive
14 concerns. For example, similar patterns of neurotransmitter disruptions have been associated
15 with schizophrenia, drug addiction, and psychosis (Cory-Slechta, 1997).
16 Altered bioelectric activity in the brain and nervous system may also occur as a result of
17 lead-induced disruption of neuron and neurotransmitter function. Nagymajtenyi et al. (1998)
18 observed disruptions, including changes of EEG and slowed nerve conduction velocity relative to
19 controls, in the somatosensory area of the cerebral cortex in prenatal and postnatal rats exposed
20 by gavage to relatively low levels of lead (80, 160, and 320 mg/kg body weight as lead acetate
21 dissolved in water). The observed changes in electrophysiological functioning depended on the
22 dose and timing (i.e., age of animal at exposure) of lead administration. The authors suggested
23 that non-invasive monitoring of electrical disturbances in the nervous system may provide a
24 valuable and early indicator of low-level lead poisoning.
25 The limbic system and hippocampus. As the hippocampus plays an important role in
26 many of the processes that appear to be affected by lead poisoning, such as cognition and
27 memory formation, it has been the focus of much research on lead neurotoxicity. Altman et al.
28 (1993), for example, found a correlation between hippocampal disruptions and active avoidance
29 learning deficit in lead treated animals (as cited in Winneke et al., 1996). Other reported
30 behavioral changes that may be attributable to hippocampal damage include increased
31 aggressiveness, seizures, inappropriate responsiveness, reversal problems, visual discrimination
32 deficits, impaired motor coordination, and other types of learning deficiencies (Petit et al., 1983).
33 Animal studies reviewed by Banks et al. (1997) and Petit et al. (1983) showed effects
34 such as significant reductions in size and weight of the hippocampus and reductions in
35 hippocampal cell layer thickness with relatively high perinatal lead exposure in rats. Damage
36 and stunting of hippocampal structural cells (glial cells/astrocytes) was observed in rats and
37 monkeys exposed to lead prenatally and postnatally. As discussed earlier, Chen et al. (1998)
38 found developmental lead exposure to alter protein kinase C (PKC) activity in the hippocampus
39 of postnatal rats. Perinatally exposed rats also exhibited reductions in development of
40 hippocampal neuronal dendrites and mossy fiber pathways, which are both involved in the
41 transmission of nerve impulses (Petit et al., 1983).
DRAFT - DO NOT CITE OR QUOTE 9 August 28, 2000
-------
1 In addition, relative to other regions of the brain, greater impairment of neuron function
2 has been observed in hippocampal cells of rats exposed to lead perinatally (Banks et al., 1997).
3 Studies reviewed by Petit et al. (1983) indicated that higher levels of lead tend to accumulate in
4 hippocampal cells of rat brains. Although the review by Petit et al. (1983) suggests that the
5 hippocampus may be a target organ for lead, other studies reviewed by Banks et al. (1997) did
6 not report a preferential accumulation of lead in the hippocampus of young lead-exposed rats.
7 Therefore, some researchers have hypothesized that the particular vulnerability of the
8 hippocampus to lead poisoning may be due more to sensitivity rather than increased lead
9 accumulation in this region (Banks et al., 1997). In addition, as the most rapid phase of
10 development of hippocampus is known to occur postnatally (late compared to other brain
11 regions), a particular sensitivity of the hippocampus to early-life lead exposure is plausible (Petit
12 et al, 1983).
13 The visual system. Lead exposed rats have exhibited persistent decreases in visual acuity
14 and spatial resolution (USDHHS, 1999). Davis et al. (1990) reviewed several studies in which
15 lead exposed rats and monkeys exhibited decreased responsiveness of neurons to visual stimuli,
16 as assessed by parameters (i.e., visual evoked potentials) which measure nerve conduction.
17 Nagymajtenyi et al. (1998) also observed disruptions in the functioning of optical nerves,
18 including changes of EEG and slowed nerve conduction velocity, in prenatally and postnatally
19 lead-exposed rats (80, 160, and 320 mg lead/kg body weight as lead acetate, dissolved in water,
20 and administered by gavage). Reviews of animal studies have also reported that rats exposed to
21 moderate levels of lead have sustained direct damage to the eye, including decreases in rod cells,
22 thinning of retinal layers, reductions in the number of axons in the optic nerve, and necrosis of
23 photoreceptors and cells in the inner retinal layer (Banks et al., 1997; USEPA, 1986). Such
24 damage was not reported in rats at lower levels of lead exposure.
25 The auditory system. Slowing of nerve conduction in the auditory pathway, as assessed
26 by brainstem auditory evoked potentials has been observed in lead exposed monkeys (Lilienthal
27 et al., 1990, as cited in Banks et al., 1997). Nagymajtenyi et al. (1998) also observed disruptions
28 in the auditory pathway, including changes of EEG and slowed nerve conduction velocity, in
29 prenatally and postnatally exposed rats.
30 2.1.2.2 Behavioral and Developmental Effects of Lead. Several recent review papers
31 have summarized previous research and advances in the area of neurotoxicological effects of
32 lead. Key studies as reviewed by Banks et al. (1997), Cory-Slechta (1997), Davis et al. (1990),
33 and Rice (1996), among others, are presented below. When possible, attention was focused on
34 health effects at low levels of lead exposure or at low blood-lead levels. Many animal studies
35 have investigated the neurobehavioral effects of lead exposure, that is, effects upon learning and
36 performance activities. Some studies also addressed related issues such as: (1) what blood-lead
37 levels were associated with learning and performance deficiencies in test animals; (2) whether
38 other signs of toxicity were present; (3) what biochemical mechanisms produce toxicity; and (4)
39 how blood levels in exposed animals correlate with specific blood-lead levels in humans at
40 which analogous effects upon learning/IQ are observed. Many experiments utilize rats or
DRAFT •- DO NOT CITE OR QUOTE 10 August 28, 2000
-------
1 monkeys as subjects since they respond to motivational factors, usually food rewards, and can be
2 induced to learn certain behaviors of interest to researchers. Animals were exposed to lead either
3 in utero, during infancy (through mother's milk/formula), during maturation or adulthood, or a
4 combination of these life phases.
5 Rice (1996) conducted an extensive review of animal studies, interpreted in conjunction
6 with learning disabilities in children as reported in human epidemiological studies. In
7 experiments with rats or monkeys, a general learning deficiency was frequently observed at high
8 lead levels. However, for monkeys exposed to low or moderate lead levels, the majority of the
9 impairment was evident only with the performance of more complex tasks. This phenomenon
10 was observed in (a) monkeys exposed to lead from birth with preweaning blood-lead levels of 50
11 ug/dL and adult blood-lead levels of 30 ug/dL (Rice, 1984; Rice, 1988); (b) monkeys with
12 blood-lead levels of 30-35 ug/dL from infant formula and postweaning levels of 19-22 ug/dL
13 (Rice and Gilbert, 1990a); and (c) monkeys with blood-lead levels of 15-25 ug/dL during infancy
14 and steady-state levels of 11-13 ug/dL during adulthood (Rice, 1985). In these studies, lead
15 exposure had ceased by the time performance tests were conducted.
16 Rice (1996) describes experiments that were conducted with visual discrimination
17 problems with the addition of "reverse performance" requirements, and/or the addition of
18 irrelevant distracting signals. In the operant discrimination reversal task, the researcher changes
19 the pattern of rewards so that previously-learned correct and incorrect responses become
20 switched. Lead-exposed animals sometimes perform as well as controls in the original learning
21 acquisition, but perform poorly when the change of rules requires learning a change in strategy.
22 In one study (Bushnell and Bowman, 1979, as cited in Rice, 1996), some adult monkeys exposed
23 to lead in infancy performed poorly on spatial discrimination reversal tasks, even though blood-
24 lead levels had returned to normal by the time of testing. Also, lead-exposed animals tend to be
25 distracted by irrelevant details more than do controls (although in some cases they may perform
26 similar to controls in the absence of such distraction). Gilbert and Rice (1987) report an
27 experiment in which the group exposed to low lead levels performed poorly on a task with
28 unfamiliar distracting cues but adequately on a task with familiar distractions, whereas the group
29 exposed to high levels performed poorly on both tasks.
30 Rice and Gilbert (1990a) report a study in which three groups of monkeys (13 monkeys
31 per group) received lead exposure either continuously from birth, during infancy only, or after
32 infancy only. Lead was administered orally as lead acetate in gelatin capsules, with the dose
33 equivalent of 1.5 mg/kg/day. An equivalent control group which received the vehicle only was
34 included in the study. All lead-exposed groups showed learning impairment, although those
35 dosed during infancy-only were less impaired than those exposed later. All exposed groups
36 tended to exhibit "perseverative behavior" and an inability to suppress inappropriate responses
37 during test delay intervals. Rice summarizes results from both human epidemiology and animal
38 studies as follows: "Increased distractability, inability to inhibit inappropriate responses,
39 perseveration, and inability to change response strategy are common themes that may be
40 extracted from both literatures."
DRAFT--DO NOT CITE OR QUOTE 11 August 28,2000
-------
1 Similar impairments in learning and performance have been noted in experiments with
2 lead-exposed rats. Cory-Slechta (1997) found that selective learning deficits were present after
3 lead exposures of 50 and 250 ppm (as lead acetate in drinking water) with resulting blood-lead
4 levels as low as 20-25 ug/dL. Lead-exposed rats performed as well as controls during the
5 performance component of the experiment (i.e., the correct sequence of responses remained
6 constant across trials), but less accurately during the repeat acquisition component (the correct
7 sequence changed in an unpredictable way with each new set of trials). Rats also displayed
8 perseverative behavior, pressing the same lever repeatedly, even though the experiment
9 precluded this pattern of response from generating a food reward. The author discusses the
10 research attempting to elucidate some of the biochemical mechanisms underlying these results,
11 including evidence which strongly suggests the existence of a link between learning impairments
12 and lead-induced disruptions of neurotransmitters, particularly those of the glutamine system.
13 Cory-Slechta (1997) also reviewed the reported effects of lead on fixed interval (FI) test
14 performance (delayed response operant schedule of food reinforcement). The FI test requires the
15 animal to bar-press only at specific minimum time intervals before a food reward can be
16 provided. Studies conducted by Cory-Slechta showed that rats exposed to low doses of lead (25-
17 300 ppm with postweaning exposures) tended to respond more rapidly than did controls, even
18 though this behavior resulted in withheld rewards. In contrast, animals exposed to higher doses
19 (500 ppm and above) showed decreased rates of FI responding, at least initially. It is
20 hypothesized that the increased response rates may actually be a form of perseverative behavior.
21 The author notes that, in other studies, lead-induced interference with the dopamine system has
22 been suggested as a possible mechanism for perseverative behavior. Also of significance, Cory-
23 Slechta reports that dose-dependent patterns in FI performance, like those in her own study, have
24 consistently been observed across a wide range studies and methodological conditions, including
25 species (i.e., reported effects described in rats, monkeys, sheep, pigeons, and mice) and
26 developmental period (i.e., prenatal, postnatal, postweaning, adult, old adult) during which lead
27 exposure occurs. Thus, evidence in the literature strongly suggests that changes in FI schedule-
28 controlled behavior seem to be one of the most reliable parameters, relative to other measures,
29 for assessing the behavioral effects of experimental lead exposures (Cory-Slechta, 1997).
30 USEPA (1986) contains an extensive review of experimental studies using rats exposed
31 to lead. A variety of behavioral and physiological responses affected by neurotoxicity, as well as
32 some social interactions, were examined. Many studies (e.g., Angell and Weiss, 1982; Cory-
33 Slechta and Thompson, 1979; Geist et al., 1985) have observed greater effects in rats exposed
34 after weaning or during maturation than in those exposed prenatally or during infancy. Rats were
35 exposed to lead in varying doses, using different modes of dose administration, and during
36 various life stages, and it was not possible to standardize the results across all studies. However,
37 many adverse effects were observed in rats with blood-lead levels around 30 Mg/dL, and some
38 effects upon learning were detected even when maximum blood-lead levels were below 20
39 ug/dL. In many studies, lead-exposed rats were different from controls according to some but
40 not all criteria evaluated. As with monkeys, lead-exposed rats acquired performance skills on
41 discrimination tests more slowly than controls, and they committed more errors.
DRAFT - DO NOT CITE OR QUOTE 12 August 28, 2000
-------
1 USEPA (1986) also reviewed studies investigating the impact of lead upon social
2 behavior in rats. While early studies reviewed suggested that lead exposure produced increased
3 aggressiveness, more recent studies indicated reduced levels of aggressive behavior. For
4 example, in a study testing electric shock-elicited aggression (Hastings et al., 1977, as cited in
5 USEPA, 1986), lead-exposed animals with blood-lead levels of 5-9 pg/dL showed significantly
6 less aggressive behavior than did controls. One of the few studies (Cutler, 1977, as cited in
7 USEPA, 1986) to investigate gender differences with respect to rodent behavior found that lead-
8 exposed male mice showed significantly reduced levels of aggressive behavior (even though
9 body weights and overall activity levels were not affected) compared to controls; however, lead-
10 exposed females showed significantly increased social/sexual investigative behavior compared to
11 controls. Social interactions were also affected in maternal behavior toward offspring where
12 both had been lead-exposed. Weight gain by pups was delayed, mothers retrieved pups to the
13 nest for longer times, and pups showed reduced exploration outside the nest, when exposed to
14 lead.
15 A review of animal studies which evaluated the persistence of lead-induced effects on
16 cognitive development was conducted by long (1998). Neurobehavioral toxicity was reported to
17 persist for up to 10 years of age in monkeys exposed to low levels of lead and was strongly
18 suggested in some studies to be an irreversible neurotoxin. The physiological evidence of lead
19 disruption of the developing brain (e.g., neuron and synapse formation), and the fact that
20 intracellular lead may not be removable from neural cells, also supports the plausibility of
21 enduring, and possibly irreversible, deficits in neurobehavioral function with lead exposure.
22 To synthesize the neurobehavioral effects of lead across species, Davis et al. (1990)
23 conducted a review of rodent, primate, and human studies. The lowest levels of internal lead
24 exposure during early development at which neurobehavioral effects have been observed were
25 reported to be <20 ug/dL for rodents (Cory-Slechta et al., 1985), 15-25 ug/dL for primates (Rice,
26 1985), and 10-15 Mg/dL for children (USEPA, 1989).
27 2.1.2.3 Findings from Specific Studies. This section presents detailed findings from
28 several recent experimental animal studies which have advanced current understanding of the
29 neurotoxicological effects of lead. An overview, including lead exposure levels and key
30 findings, of selected studies reviewed in this section is presented in Table 2-1. These studies
31 varied considerably with respect to the amount of detail used in describing the experimental
32 design, randomized treatment group allocation, procedures employed to control sources of
33 variability, and statistical methods used to analyze results. In general, experiments using smaller
34 animals (e.g., rodents) can employ larger numbers of subjects and can control more detailed
35 features of the experimental design than experiments using larger animals (e.g., primates).
DRAFT -- DO NOT CITE OR QUOTE 13 August 28, 2000
-------
1 Table 2-1. Summary of Lead Exposure Levels and Key Findings for Selected Animal
2 Studies
Authors
Subject
Species
Lead Exposure
Key Findings /
Effects of Lead Exposure
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Kuhlman et
al. (1997)
Rat
750 or 1000 ppm lead acetate, as feed;
exposed in utero, in utero to adulthood,
or postweaning to adulthood; controls
received no lead exposure
performance impairment in water
maze for all rats exposed in utero; no
impairment for rats exposed post-
weaning
Nagymajtenyi
et al. (1998)
Rat
80, 160, 320 mg/kg/day lead acetate in
distilled water, by gavage; exposed pre-
or post-natally; controls received same
volume of distilled water by gavage
dose dependent increase in
behavioral and bioelectric aberrations
(e.g., hyperactivity)
Rice and
Gilbert
(1990b)
Monkey
1.5 mg/kg/day lead acetate, in
capsules; exposed continuously after
birth, postnatally until 400 days, or
after 300 days from birth; controls
received vehicle only
all exposed groups showed spatial
delayed alternation task impairment
Mello et al.
(1998)
Rat
1.0 mM lead acetate as drinking water;
exposed in utero and during nursing via
dams; control dams received deionized
water only
selective motor skill impairment
(accelerated fist eye opening, startle
reflex, and free-fall righting; impaired
spontaneous alternation performance
in maze)
Burger et al.
(1998)
Turtle
0.25, 1.0, or 2.5 mg/g lead acetate in
deionized water, by injection; one-time
exposure; controls received an isotonic
saline solution injection
death with high dose; dose-
dependent righting response
impairment with low and moderate
doses
Chen et al.
(1998)
Rat
0.2% lead acetate as drinking water;
exposed in utero and during nursing via
dams, and postweaning directly;
controls received 0.145% sodium
acetate in drinking water
altered protein kinase C (PKC)
distribution in the hippocampus
Cory-Slechta
(1997)
(review)
Rat
50 or 250 ppm lead acetate as drinking
water; exposed postweaning; control
groups received no lead exposure
disruption of neurotransmitter
systems (i.e., dopamine and
glutamine systems); selective
learning deficits (e.g., impaired repeat
acquisition performance); dose-
dependent alteration of fixed interval
schedule-controlled response rates
20 Kuhlman et al. (1997) conducted an experiment with 10 rats reared in each of five groups.
21 The control group received no lead exposure. The "maternal exposure" group was exposed to
22 lead in utero and during lactation (750 ppm lead acetate in feed via dams), but was moved to a
23 control diet after weaning. The permanent group was exposed to lead both in utero and
24 continuously afterward into adulthood (750 ppm lead acetate in feed). Finally, there were two
25 post-weaning groups, in which no exposure occurred until after weaning, and then pups were fed
DRAFT - DO NOT CITE OR QUOTE
14
August 28, 2000
-------
1 diets containing two different lead concentrations (750 or 1000 ppm lead acetate in feed). While
2 animals were randomized among groups, littermates were places in different groups to prevent
3 genetic variability/similarity from becoming a confounding factor in the experimental design.
4 Rats were performance-tested using a water maze at 100 days of age, and their blood-lead levels
5 were measured at that time. A highly significant impairment in performance was observed in
6 both the maternal and permanent exposure groups. The results seen in the maternal group
7 demonstrated the impact of lead upon early development, even though average blood levels in
8 that group (1.8 Mg/dL) had returned to control levels by the time of the test. The post-weaning
9 exposure groups did not show any significant performance impairment, even though average
10 blood-lead levels exceeded 20 //g/dL at the time of testing.
11 Nagymajtenyi et al. (1998) also observed rats exposed to lead during gestation and found
12 an increase in behavioral and bioelectric aberrations compared to control rats. Lead was
13 administered by gavage (to 120 female and 60 male rats in the parental generation, and 120 male
14 offspring) at either 80.0,160.0, or 320.0 mg/kg body weight in the form of lead acetate dissolved
15 in water. The control rats received vehicle only. There were three variations on the treatment
16 schedule: (1) pregnant females were dosed only during the 5th -15th day of pregnancy; (2) dosing
17 occurred during pregnancy and through lactation; or (3) dosing occurred during pregnancy,
18 lactation, and directly to pups, postweaning, for 8 weeks. Behavioral observations of male
19 offspring were made at 12 weeks. Results showed that lead dosing during pregnancy was related
20 to a significant, dose-dependent increase in hyperactive behavior in the offspring. Nagymajtenyi
21 et al. (1998) also observed electrophysiological (electrocorticogram (ECoG), cortical evoked
22 potentials, etc.) parameters in male offspring at 12 weeks. Electrophysiological functions
23 showed both dose- and treatment-dependent changes, including decreased mean amplitude and
24 increased frequency of the EEC, and lengthened latency and duration of the evoked potentials.
25 The authors conclude that low-level lead exposure during prenatal and postnatal development
26 can interfere with normal development and bioelectric functioning within the nervous system and
27 is associated with behavioral changes. In addition, the authors suggest that the functional and
28 behavioral effects of lead, which occur without other overt signs of lead toxicity, are much more
29 harmful than has been previously supposed.
30 In Mello et al. (1998), rat pups were exposed to lead in utero and during nursing, using
31 1.0 mM lead acetate administered to dams as drinking water. Eleven litters from control females
32 and nine litters from lead-exposed females were used, for a total of 160 pups. The pups were
33 observed for physical development and tested for reflexes/behavior at aged 17-19 days by a
34 single researcher who was blind to the group status of individual pups. While lead exposure
35 appeared to significantly accelerate the appearance of first eye opening, startle reflex, and free-
36 fall righting, it significantly impaired spontaneous alternation performance in a maze. No
37 explanation was given for the seemingly contradictory effects in this study, though the authors
38 suggested that any lead-induced alterations in animal development or behavior, regardless of
39 direction, must be considered deleterious.
40 In a study performed by Rice and Gilbert (1990b), 52 infant monkeys were assigned
41 shortly after birth to one of four feeding groups: (1) a control diet; (2) a lead-containing diet
DRAFT - DO NOT CITE OR QUOTE 15 August 28, 2000
-------
1 continuously after birth; (3) a lead-containing diet from birth until age 400 days, followed by a
2 control diet; and (4) a control diet from birth until age 300 days, followed by a lead-containing
3 diet. Lead was administered orally in gelatin capsules as lead acetate in 0.05 M sodium
4 carbonate, equivalent to 1.5 mg/kg/day. Feeding regimes were maintained up to and through the
5 time of spatial delayed alternation task testing at age 6-7 years. All three exposure groups
6 showed performance impairment relative to controls and were impaired to an approximately
7 equal degree. These results suggested that lead exposure only during infancy (i.e., < 400 days)
8 results in impairment comparable to exposure that continues beyond infancy.
9 Burger et al. (1998) conducted an experiment on 48 slider turtle hatchlings which were
10 randomly assigned in equal numbers to a control group and three lead exposure groups. The
11 exposed groups were injected intramuscularly at one of three doses (0.25, 1.0, or 2.5 mg/g as lead
12 acetate, respectively). After lead exposure, turtles were maintained and observed by technicians
13 who were blind to the group status of individual animals. Behavioral observations, growth, and
14 survival data were taken both prior to and at various times after lead injection (i.e., weekly during
15 the first month, and at 4 weeks, 4 months, and 6 months post-exposure). No survivors remained
16 in the high dose group by 120 days of age. In the low and medium dose groups, one of the
17 behaviors tested was the time required by an animal to right itself when turned upside down on
18 its shell. At the age of 6 months, righting response adjusted for body weight was significantly
19 impaired by lead dosing. Both the low and medium dose groups performed significantly worse
20 than did controls, and the medium dose group performed significantly worse than did the low
21 dose group, indicating a dose-response effect. The range of lead doses used in the study had such
22 a marked impact on survival that "the threshold for behavioral effects is on the same order of
23 magnitude as the LD50" (i.e., the lowest lead dose at which only 50% of animals would survive).
24 Also, the authors note some of the limitations of testing this species, as few of its behaviors were
25 suitable endpoints for detecting differences between controls and dosed animals. For example,
26 the experimenters attempted to measure time in seeking cover as a potential response endpoint,
27 but even control animals did not always exhibit consistent behavior.
28 In an attempt to elucidate some of the physiological mechanisms of lead-induced learning
29 deficits, Chen et al. (1998) investigated the effects of developmental lead exposure on protein
30 kinase C (PKC) activity in the hippocampus of rats at various postnatal ages. Lead was
31 administered orally as 0.2% lead acetate in drinking water to pregnant and lactating female rats
32 and then directly to their weanling pups (weaned at postnatal day 21) in drinking water. Controls
33 received 0.145% sodium acetate in drinking water. Four to six rat pups were randomly selected
34 from different dams at postnatal days 7, 14, 28, and 56, necropsied, and PKC activity was
35 measured in both the membrane and cystolic fractions of the hippocampi. Results showed that
36 lead exposure increased PKC activity in the cystolic fraction at postnatal day 56, and decreased
37 PKC activity in the membrane fraction at postnatal day 7. The ratio of membrane to cystolic
38 PKC activity, which is indicative of PKC distribution, decreased at postnatal days 28 and 56.
39 A review of studies in Chen et al. (1998) indicated that PKC activity has been associated
40 with various brain functions (e.g., ion channel function, receptor function, and neurotransmitter
41 release) and that alteration of hippocampal PKC, in particular, has been correlated with poor
DRAFT--DO NOT CITE OR QUOTE 16 August 28, 2000
-------
1 performance in several learning tasks. Therefore the authors hypothesize that the lead-induced
2 alterations of PKC activity and distribution observed in the current study may have caused
3 functional changes in the animal brain, including modulation of ion channels, desensitization of
4 receptors, and enhancement of neurotransmitter release. Chen et al. (1998) also suggested that
5 some of the learning and memory deficits observed in children are likely to be causally related to
6 the types of PKC activity alterations exhibited in this study.
7 2.1.3 General Health Effects
8 Although the health effects of lead are diverse, and in general depend on the duration and
9 degree of lead exposure, they are all thought to originate from lead's ability to interfere with
10 fundamental biochemical processes (i.e., mitochondrial energy production, calcium-mediated
11 processes, and protein function). All of the major types of health effects of lead that have been
12 observed in humans, including hematological, neurodevelopmental, immunological,
13 cardiovascular, and renal effects, have been demonstrated in controlled, dose-response studies in
14 rodents, dogs, and/or non-human primates. Although conclusive evidence for carcinogenicity is
15 still lacking in human studies, animal studies have also indicated that lead has the potential to be
16 carcinogenic.
17 While the focus of this report is on the neurobehavioral effects of lead exposure, any type
18 of health effect that is demonstrated to have resulted from lead exposure in animal studies could
19 be of interest, especially at low lead doses. Therefore, an overview of experimental animal
20 studies which have investigated the general physical health effects that result from lead exposure
21 is presented below. This overview will provide only brief statements of findings across animal
22 studies. Much of the information presented here has previously been cited in USDHHS (1999)'
23 and USEPA (1986).
24 2.1.3.1 Death. At high levels of exposure, lead is known to cause death in humans
25 following severe lead encephalopathy, and has been suggested to be a causative agent in Sudden
26 Infant Death Syndrome (USDHHS, 1999). In children, blood-lead levels associated with
27 encephalopathy-related death range from 125-750 //g/dL (NAS, 1972, as cited in USDHHS,
28 1999). The data are minimal, however, regarding high exposures to lead and death in animals.
29 Increased mortality has been observed in studies with rats and mice given lead in food or
30 drinking water, although in some cases mortality did not occur in a dose-related manner
31 (USDHHS, 1999).
32 2.1.3.2 Systemic Effects.
33 Hematological Effects. There have been numerous studies in animals demonstrating the
34 adverse effects of lead on heme (hematin) biosynthesis, which can in turn affect many organ
35 systems. In acute and intermediate-duration studies, the activities of several enzymes involved in
1 A 1998 draft of this document was used in obtaining information for this section that is cited as originating from
USDHHS, 1999.
DRAFT » DO NOT CITE OR QUOTE 17 August 28, 2000
-------
1 the heme biosynthetic pathway have been observed to be altered by administration of lead to rats.
2 Adverse hematological effects have also been observed in rats and dogs in longer term, lower
3 dose studies (USDHHS, 1999).
4 Renal Effects. The effects of lead on the renal system have been documented in animal
5 studies involving rats, dogs, monkeys, and rabbits. Symptoms of renal insufficiency, including
6 both transient and irreversible kidney lesions, tubular dysfunction, and increased excretion of
7 amino-acids and nitrogen compounds in urine, have been observed at high lead exposures in
8 these studies (USDHHS, 1999; USEPA, 1986).
9 Cardiovascular Effects. Exposure to lead has been associated with adverse
10 cardiovascular effects in studies with laboratory animals. While older animal studies concluded
11 that hypertension was clearly associated with extremely high doses of lead, the direct effects of
12 lead on blood pressure and the secondary effects (i.e., hypertension as a result of renal damage)
13 of lead were difficult to separate (USEPA, 1986). However, USDHHS (1999) includes a review
14 of several more recent chronic-duration experiments in rats and found that lower lead exposures
15 (i.e., levels that were otherwise non-toxic) were associated with sustained increases in blood
16 pressure as compared to controls. Other adverse cardiovascular effects, such as structural and
17 functional changes relative to controls (e.g., degeneration of the myocardium and aorta), have
18 also been observed in rats following lead ingestion (USDHHS, 1999).
19 Hepatic Effects. Indications of possible lead toxicity to the hepatic system, as suggested
20 by increased liver weight, morphological changes, and changes in liver enzyme activity, have
21 been observed in animal studies involving mice, rats, dogs, and baboons (USDHHS, 1999;
22 USEPA, 1986). In general, adverse effects of lead on the liver in animal studies were observed at
23 high exposure levels.
24 Respiratory Effects. There is limited evidence from inhalation studies that prolonged
25 exposure to lead may cause respiratory system irritation in mice. Data suggest that lung
26 irritation, if significant, is largely dependent on the duration of exposure (USDHHS, 1999).
27 Other Systemic Effects. Lead, at relatively high exposure levels, has also been associated
28 in experimental animal studies with other various systemic effects. These include general
29 impairment of cellular function (via interference with membranes, calcium ion transport, and
30 mitochondria! respiration), long-term visual system deficits in animals exposed during post-natal
31 development, weight loss, impairment of vitamin D metabolism, and endocrine disruption
32 leading to a decrease in thyroid function (USDHHS, 1999; USEPA, 1986).
33 2.1.3.3 Immunoloqical Effects. The literature provides good evidence that lead can
34 have an immunosuppressive effect on animals at relatively low doses. Although additional
35 research is required to elucidate the exact mechanisms of action, the macrophage is postulated to
36 be the primary immune system target cell for lead (USEPA, 1986). Studies reviewed in USEPA
37 (1986) and USDHHS (1999) showed that in rats, mice, and rabbits, both oral and airborne lead
38 exposures contributed to increased susceptibility to bacterial and viral infections, decreased
DRAFT -- DO NOT CITE OR QUOTE 18 August 28, 2000
-------
1 antibody counts, decreased cell-mediated immunity, depressed lymphocyte function, and
2 suppressed macrophage-dependent immune response, as compared to control animals. In many
3 of the studies cited, lead-induced immunosuppression occurred at low exposure levels that
4 otherwise induced no overt symptoms of systemic lead toxicity, such as increased blood pressure
5 or weight loss.
6
7 2.1.3.4 Reproductive and Genotoxic Effects. The observed effects of lead on the
8 reproductive system are mixed across animal studies. Adverse effects observed included
9 decreased pregnancy rates, decreased male fertility, damage to ovaries and testes, and altered
10 pubertal progression, all presumably by interference with hormone production. High doses of
11 lead have also been associated with fetal stunting and fetotoxicity. Lead was not observed to be
12 teratogenic in limited rodent studies, except when lead was administered via injection
13 (USDHHS, 1999). In various animal studies on the genotoxic effects of lead, chronic exposures
14 to lead produced either slight or no significant increases in chromosomal aberrations in mice or
15 monkeys, except in one study (Deknudt et al., 1977, as cited in USDHHS, 1999) in which
16 monkeys were given a calcium-deficient diet (USDHHS, 1999).
17 2.1.3.5 Carcinoqenicitv. Data obtained in laboratory animal studies indicate that high
18 levels of oral exposure (approximately 25-100 mg/kg/day) to lead acetate and lead phosphate
19 increases the incidence of renal tumors in rats and mice (USDHHS, 1999). However, due to the
20 extremely high cumulative doses of lead used in these studies, and the uncertainties regarding
21 mechanisms by which lead induces tumors in the rat kidney (i.e., by acting on species-specific or
22 trans-species proteins), extrapolation to low-level exposure in humans is difficult. The
23 carcinogenicity of lead at lower doses in animals is yet to be determined.
24 2.1.4 Conclusions from Animal Studies Investigation
25 Lead has been observed to have widespread neurotoxic effects, as well as behavioral and
26 cognitive symptoms, in humans. These observations are largely consistent with the findings of
27 morphological, electrophysiological, and biochemical studies on animals. Animal studies are
28 also congruent with observations of lead exposure in humans, suggesting an increased
29 susceptibility of the young brain to lead poisoning (Banks et al., 1997). In addition, animal
30 studies have provided physiological evidence that many of the effects of lead on the
31 differentiation of the developing nervous system, such as synaptic and dendritic development,
32 and myelin and other nerve structure formation, have the potential to be long-lived effects
33 (USEPA, 1986). Although animal models do not duplicate the human response to lead exposure,
34 they do serve to provide strong support for expecting certain health effects to occur when humans
35 are exposed. In addition, animal studies allow for more specific determination of dose-response
36 relationships. Single-route exposure and dose-response data are generally not available for
37 humans. Animal studies provide an opportunity to elucidate some of the physiological
38 mechanisms of lead toxicity, and thus they are a valuable tool for assessing the potential risks of
39 lead exposure for human health.
DRAFT -- DO NOT CITE OR QUOTE 19 August 28, 2000
-------
1 A large resource of published literature also exists in which animal experiments were
2 conducted to investigate the neurobehavioral effects caused by lead dosing. Most of these used
3 monkeys or rats in experiments to assess the effects of lead dosing upon learning and
4 performance activities and were conducted using low to moderate doses of lead. In many studies
5 using monkeys, no other overt signs of toxicity (e.g., weight loss) were observed even when
6 learning impairments were observed.
7 Several experiments have demonstrated that lead-dosed animals perform similarly to
8 controls on simple tasks, but perform worse than controls on more complex tasks. In particular,
9 both rats and monkeys perform poorly in reverse discrimination tasks (where they have to change
10 a previously-learned strategy), or in fixed-interval/delayed response tests. Perseverative behavior
11 is a recurring phenomenon in the latter type of test, whereby the animal is unable to inhibit
12 inappropriate, repetitious movements despite having rewards withheld as a consequence.
13 Experiments to compare the severity of impairments in animals dosed only during early
14 development (i.e., prior to birth or during infancy) versus those dosed only during maturation did
15 not yield consistent results. Generalizations about dosing schedules could not be made across all
16 studies or across species, as to which groups of animals faired worse.
17 There are general similarities in studies and across species, however, with regards to the
18 lowest levels of internal lead exposure during early development at which neurobehavioral
19 effects are commonly observed. For example, internal lead exposures associated with
20 neurobehavioral effects were reported to be as low as the following: <20 ug/dL for rodents
21 (Cory-Slechta et al., 1985, as cited in Davis et ah, 1990), <15 ug/dL for primates (Rice, 1985, as
22 cited in Davis et al., 1990), and 10-15 ug/dL for children (USEPA, 1989, as cited in Davis et al.,
23 1990). Although direct comparisons of blood-lead levels are generally not considered to be
24 accurate across species, evidence that rodents and primates may actually tolerate higher
25 exposures to lead before reaching a given blood-lead level, suggests that exposures and blood-
26 lead levels of concern would even lower in humans than in animals (Davis et al., 1990).
27 The published literature varied in the amount of detail given to explaining experimental
28 designs used. For example, it was often implied but not explicitly stated whether subjects were
29 randomized to exposure groups according to some statistical criteria, and whether the
30 experimenters observing the performance tests were blind to the dosing status of subjects.
31 Articles usually provided detail about experimental procedures involving diets and weight gain,
32 as lead exposure was often given through food or water. Blood-lead levels at the time of
33 performance testing were usually measured and reported.
34 The choice of response endpoint was clearly of particular importance in the design of
35 performance tests. Tests that were either too simple or too complex would not enable the
36 researcher to detect significant differences between dosed and control animals. Also, lead
37 exposure may delay or accelerate the physical development of some reflexes, so peak
38 performance may occur at different times in different exposure groups. Therefore, the time at
39 which statistical comparisons are made may produce seemingly paradoxical results. In
DRAFT - DO NOT CITE OR QUOTE 20 August 28,2000
-------
1 particular, any apparent non-monotonic dose-response relationships must be interpreted with
2 caution.
3 Despite the aforementioned caveats, animal studies do provide an opportunity to elucidate
4 some of the physiological mechanisms of lead toxicity, as well as generate single-route exposure
5 and dose data that are generally not available in humans. Although animal models cannot claim
6 to duplicate the human response to lead exposure, the resulting ability to control confounding
7 factors permits these models to succeed in providing substantial evidence that supports the
8 existence of a causal relationship between low level lead exposure and neurological impairment,
9 especially in the young.
10 2.2 SUPPORT FOR THE CAUSALITY OF ADVERSE HEALTH
11 EFFECTS DUE TO LEAD EXPOSURE
12 Chapter 2 of the §403 risk analysis report documented previously-published information
13 on adverse health effects associated with lead exposure to humans. Subsequent chapters
14 characterized how environmental-lead exposure impacts selected blood-lead concentration and
15 health effect endpoints in children. Specifically, IQ-based health endpoints were selected to
16 represent the neurological effects associated with lead exposure.
17 A concern, particularly with the IQ-based health endpoints, is whether lead can be
18 assumed to cause the adverse health effects. For ethical reasons, the controlled lead exposure
19 studies necessary to test this hypothesis can not be performed on humans. However, animal
20 studies can be used to supplement the evidence of human studies in this area. To this end,
21 Section 2.1 of this document presented key findings of animal studies that investigated the
22 impact of lead exposure on adverse health effects. This section examines whether the combined
23 evidence of human and animal studies suggests that lead exposure causes neurological damage
24 that can be measured through intelligence testing.
25 2.2.1 Principles of Causality
26 On the issue of causality, Needleman & Gatsonis (1990) make the following assertions
27 and present the following principles of causality (citing Kenny, 1979, as a reference):
28 "Epidemiologic studies cannot, by themselves, establish causal relationships.
29 Causality is not subject to empirical proof, whether in the field or in the
30 laboratory. Given that direct demonstration of proof of a low-dose lead effect in
31 a naturalistic setting is not achievable, epidemiologists rely on canons that, if
32 satisfied, permit the conservative drawing of causal inferences. They are (1) time
33 precedence of the putative cause, (2) biologic plausibility, (3) nonspuriousness,
34 and (4) consistency."
35 Needleman & Gatsonis (1990) also introduce a fifth principle, biologic gradient, which is part of
36 biologic plausibility. Needleman (1998) indicates that these principles originate from
DRAFT -- DO NOT CITE OR QUOTE 21 August 28,2000
-------
1 investigations of whether tobacco use causes cancer and attributes them to the British statistician
2 Sir Austin Bradford Hill. Together, these principles can be applied to the findings of existing
3 studies (both human and animal studies) to imply that lead exposure causes neurological damage
4 that may be measured through IQ score decrements.
5 The following paragraphs discuss the five principles of causality relative to the evidence
6 in human and animal studies.
7 Time precedence. Time precedence means that the proposed cause must exist before the
8 proposed effect occurs. That is, a child must become exposed to lead before neurological deficit
9 is observed.
10 In cross-sectional studies in humans, it is impossible to establish the time precedence of
11 lead exposure and neurological deficit. Longitudinal studies in humans, however, have shown
12 that disturbances in early neurobehavioral development occur even at low lead exposure levels in
13 early life. For example, in Boston, 4 to 8 point differences in performance on the Bayley Mental
14 Development Index were reported at 6, 12, 18, and 24 months, after adjusting for other
15 covariates, when children with low prenatal blood-lead levels (mean of 1.9 ug/dL) Were
16 compared to children with modestly elevated prenatal blood-lead levels (mean of 14.6 ug/dL)
17 (Bellinger et al., 1985a, 1985b, 1986a, 1986b, 1987a; 1991; 1992). Additional detail on this and
18 other longitudinal studies is presented in Section 2.2.2 below and Section 2.3.1 of the §403 risk
19 analysis report.
20 Controlled animal studies further support the time precedence of lead exposure. For
21 example, groups of monkeys exposed to lead either continuously from birth or only after infancy
22 showed learning impairment on a series of non-spatial discrimination reversal tasks, relative to a
23 control group that were not exposed to lead (Rice and Gilbert 1990a). These lead-exposed
24 monkeys, along with a group exposed only during infancy, also showed performance impairment
25 relative to controls in a spatial-delayed alternation task at age 6-7 years. Rats exposed to lead in
26 utero and during lactation displayed a highly significant impairment in performance on a water
27 maze test at 100 days of age, relative to controls, even though the average blood-lead level had
28 declined to 1.8 ug/dL by this time (Kuhlman et al., 1997). Additional animal research described
29 in Section 2.1 also supports the conclusion that lead exposure precedes the neurological deficit.
30 Biologic plausibility. Biologic plausibility means that the causal relationship between
31 exposure and adverse health effects must be consistent with known biological function.
32 While investigation of the mechanisms of lead toxicity remains the subject of active
33 research, it is known that lead can interfere with cell function by competing with essential
34 minerals, such as calcium and zinc, for binding sites on membranes and proteins. Lead binding
35 to membranes or transport proteins can inhibit or alter ion transport across the membrane or
36 within the cell. In the brain, lead can substitute for calcium and zinc in ion transport events at the
37 synapse (i.e., the junction where the axon of one neuron terminates with the dendrite of another
38 neuron, through which nerve impulses must travel to move from one nerve cell to another). The
DRAFT - DO NOT CITE OR QUOTE 22 August 28,2000
-------
1 normally-developing brain appears to delete synapses that are unused and to keep and strengthen
2 synapses that are used. Goldstein (1990, 1992) suggested that lead may disrupt, or delay, the
3 development of synapses and that, perhaps, the resulting connections in the brain are "poorly
4 chosen," leading to functional impairment. Silbergeld (1991) found that exposure of fetal
5 animals to lead affects both regional growth and synaptogenesis, with synaptogenesis being the
6 more sensitive. Although some of these conclusions are speculative, it is biologically plausible
7 that exposure to lead causes neurological damage.
8 Biologic gradient. Biologic gradient means that a dose-response relationship must be
9 present (i.e., increased doses of lead should cause increased impairment).
10 It is well known that high-level exposure to lead produces encephalopathy in children,
11 starting at blood-lead levels of 80 to 100 ug/dL. At lower exposure levels, IQ decrements, fine
12 motor dysfunction, and disturbances in neurobehavioral development have been related to
13 varying levels of lead exposure. These effects are summarized in Section 2.3.1 of the §403 risk
14 analysis report. As summarized in Section 4.4 and Appendix D2 of the §403 risk analysis report,
15 many studies have focused on estimating the magnitude of this dose-response relationship.
16 Nonspuriousness. Nonspuriousness means that confounding factors associated with
17 adverse health effects must be ruled out.
18 Studies that investigate the relationship between children's IQ and blood-lead
19 concentration have adjusted for other demographic factors that may affect neurological
20 development, such as Home Observation for Measurement of the Environment (HOME) score,
21 maternal IQ, and socioeconomic status. These studies are summarized in Appendix D2 (Tables
22 D2-1 and D2-2) of the §403 risk analysis report. In many of these studies, the level of lead
23 exposure remains a highly-significant factor after adjusting for these potential confounding
24 factors. Because the method of adjusting for confounding factors in human exposure studies
25 does not necessarily remove all confounding, however, animal research is required to supplement
26 human subject research. Animal studies minimize confounding by administering lead in
27 controlled doses to randomly-selected subjects that are genetically similar and are otherwise
28 treated similarly. Controlled animal studies described in Section 2.1 of this document support
29 the conclusion that lead exposure precedes neurological deficit. In addition, these studies
30 demonstrate that the effects are nonspurious by establishing study designs that include control
31 groups and that attempt to control for confounding factors.
32 Consistency. Consistency, also known as coherence, requires that the phenomenon be
33 demonstrated in different studies under similar, but not identical conditions.
34 Human studies investigating the relationship between blood-lead concentration and IQ
35 scores have generally associated IQ decrements with increases in blood-lead concentration in
36 various populations around the world, as summarized in Appendix D2 (Tables D2-1 and D2-2)
37 of the §403 risk analysis report. The magnitude of the estimated IQ decrement varied from study
38 to study, as may be expected, and the relationship was not always statistically significant.
DRAFT-DO NOT CITE OR QUOTE 23 August 28.2000
-------
1 However, the relationship was consistently negative (i.e., increased blood-lead concentration was
2 associated with IQ decrement in more than ten studies). Furthermore, a loss of approximately 2-
3 3 IQ points was associated with an increase in blood-lead concentration from 10 to 20 ug/dL in
4 several of these studies.
5 2.2.2 Causality As Addressed in Longitudinal Studies
6 The consistency principle of causality implies that causality can not be concluded from
7 the findings of a single human monitoring study. In turn, the time precedence principle indicates
8 that repeated data collection over time for study subjects within longitudinal studies provides an
9 important component of an investigation into the specific role of lead exposure as the cause of
10 certain adverse health effects. Therefore, the findings of multiple longitudinal studies on the
11 association between lead exposure and diminished performance on cognitive function or
12 intelligence testing can be an important contribution to the argument of causality (while the
13 ability to conclude causality exclusively from such findings remains very limited). The level of
14 importance of the contribution increases when consistent findings are observed across studies
15 (and across different cohorts having different demographic characteristics and lead exposure
16 potential).
17 This section provides some key findings from two longitudinal studies (Boston and Port
18 Pirie) that can be used in evaluating the hypothesis of causality. This is not meant to be an
19 exhaustive presentation of all results (significant or otherwise) across all longitudinal studies, but
20 instead is a presentation of only key findings from selected studies. For example, results from
21 other longitudinal studies that monitored lead exposure (e.g., Cleveland, Cincinnati, Sydney,
22 Yugoslavia) have not been reviewed. In addition, the potential for reverse causality (i.e., children
23 with lower levels of intelligence are more prone to elevated blood-lead concentrations) is not
24 addressed.
25 The Boston Prospective Study (Bellinger et al., 1985a, 1985b, 1986a, 1986b, 1987a; 1991; 1992)
26 The Boston prospective study considered infants bom at the Brigham and Women's
27 Hospital in Boston, MA, from August, 1979, to April, 1981. Cord blood-lead levels were
28 measured for 9489 infants (97% of all available infants), and those whose cord blood-lead levels
29 were within one of the following three categories were considered for the study: < 3 ug/dL
30 (low), 6-7 ug/dL (mid), and ;> 10 ug/dL (high). These categories represented the 10lh, 50th, and
31 90lh percentiles of the cord blood-lead distribution observed in the first three months of the study.
32 A total of 249 infants in these categories were enrolled in the study (85 in the low category
33 having mean 1.5 ug/dL, 88 in the mid category having mean 6.5 ug/dL, and 76 in the high
34 category having mean 14.6 ug/dL).
35 The cohort was considered to be of high socioeconomic standing (e.g., 87% of the
36 enrolled infants were white; 92% were considered to come from intact families). While such a
37 cohort tends to have a lower likelihood of lead exposure than more disadvantaged children (e.g.,
38 those in poor, inner-city neighborhoods), thereby restricting the ability to generalize results to a
DRAFT -• DO NOT CITE OR QUOTE 24 August 28,2000
-------
1 wide population, the low occurrence of certain demographic conditions that are typically highly
2 correlated with lead exposure gives this study a greater opportunity to isolate the effect of lead
3 exposure (especially at low levels) on cognitive function.
4 Post-natal blood-lead concentrations were measured on the study cohort at ages 6,12,18,
5 24, and 57 months, and at age 10 years. Capillary blood was obtained through 24 months of age,
6 and venous blood was obtained at ages 57 months and 10 years. Cognitive function was
7 measured by the Mental Development Index (MDI) of the Bayley Scales of Infant Development
8 at 6,12, 18, and 24 months of age, by the McCarthy Scales of Children's Abilities at age 57
9 months, and by the Wechsler Intelligence Scale for Children-Revised (WISC-R) and Battery
10 Composite scores on the Kaufman Test of Educational Achievement -Brief Form (K-TEA) at age
11 10 years.
12 Investigations on the association between lead exposure (as measured by cord-lead or
13 blood-lead concentration) and intellectual functioning within the study cohort were performed at
14 various time points during the study. The extent of association was measured by multiple
15 regression modeling, adjusting for parameters that represent potential confounding factors. This
16 resulted in the following key findings:
17 1. At ages 6, 12, 18, and 24 months, children in the high cord-blood group had
1 g significantly lower MDI scores relative to each of the other two groups (4.8 points
19 lower than the low group and 3.8 points lower than the mid group) at the 0.05
20 level. Furthermore, the MDI scores at these ages were not significantly associated
21 with post-natal blood-lead concentrations measured up to that time (Bellinger et
22 al., 1987). The level of association increased upon adjusting for potential
23 confounders within the multiple regression equation.2
/
24 2. By age 57 months, the association between cord-blood grouping and cognitive
25 function diminished considerably from what was observed at earlier ages and was
26 no longer statistically significant, except in the instance where only children with
27 blood-lead concentrations at or above 10 ug/dL at age 57 months was considered.
28 However, a significant (inverse) association was observed between blood-lead
29 concentration at 24 months of age and score on the McCarthy scales at 57 months,
30 upon adjusting for potential confounders3 (Bellinger et al., 1991). The perceptual-
31 performance subscale of the McCarthy scales, which measures visual-spatial and
2 Mother's age, mother's race (white vs nonwhite), mother's IQ (as measured by the Peabody Picture vocabulary test),
mother's education level, # of years mother smoked, # alcoholic dnnks per week by mother in the third trimester of pregnancy,
Hollmgshead Four-Factor Index measure of family social class, quality of care-giving environment, child's sex, child's birth
weight, child's gestational age, child's birth order
3 Family social class, maternal IQ, marital status, preschool attendance, HOME score, # hours per week of "out-of-
home" care, # family residence changes, recent medication use, # adults in household, gender, race, birth weight, birth order.
DRAFT-DO NOT CITE OR QUOTE 25 August 28.2000
-------
1 visual-motor integration skills, was especially sensitive to post-natal lead
2 exposure.
3 3. The association between increased blood-lead concentration at 24 months of age
4 and observed deficits in full-scale and verbal IQ scores was statistically significant
5 at age 10 years, even after adjusting for confounding variables4 (p=0.007 for full-
6 scale IQ, 0.004 for verbal IQ, but only 0.091 for performance IQ). Blood-lead
7 concentrations at other ages, however, were not significantly associated with such
8 deficits at the 0.05 level.
9 Finding #1 shows that despite a child's lead exposure during the first two years, cognitive
10 function during this period is more significantly associated with pre-natal exposure, which is
11 generally less prone to behavioral and environmental confounding than is direct, post-natal
12 exposure. As a child ages, as seen in Findings #2 and #3, post-natal lead exposure (especially its
13 peak at approximately two years of age) becomes more dominant than pre-natal exposure in its
14 association with the child's performance on intelligence tests. This suggests that the effect of
15 cumulative post-natal exposure through approximately age 2 years eventually outweighs pre-
16 natal exposure. In particular, children can eventually see reduced performance on intelligence
17 testing if their post-natal lead exposure becomes significant, despite their pre-natal lead exposure.
18 Furthermore, a child's post-natal lead exposure tends to change at a faster rate than other
19 demographic variables that are highly correlated with blood-lead levels early in life, with the
20 more recent lead exposures being predictive of a child's current health consequences. These
21 findings support the hypothesis of causality, especially in the relatively homogeneous and highly-
22 privileged cohort considered in this study.
23 Typically, in environments having high lead levels (e.g., inner-cities, smelter/mining
24 communities), correlations between blood-lead concentrations measured at different ages are so
25 highly correlated that it is difficult to separate out the age effects. However, the lower age-to-age
26 correlations observed in the Boston study (resulting from relatively low lead exposure potential)
27 allowed for investigating age-specific vulnerabilities within this study.
28 The Port Pirie Cohort Study (Baghurst et al., 1992; Tong et al., 1996; Bums et al., 1999)
29 This study consisted of 723 subjects born in the lead smelting community of Port Pirie,
30 Australia (and surrounding rural communities) from May, 1979, to May, 1982. Cord-blood was
31 obtained and analyzed for lead (geometric mean = 8.3 ug/dL). In addition, capillary blood
32 samples were collected at ages 6 and 15 months and annually from ages 2 through 7 years. A
33 venous blood sample was collected at age 11-13 years. From the measured blood-lead levels,
34 average lifetime blood-lead concentration was calculated for each child using trapezoidal
35 integration. Geometric mean blood-lead concentrations increased to 21.2 ug/dL by age 2, then
4 HOME score at 10 years, total HOME score al 57 months, child stress, maternal age, race, maternal IQ,
socioeconomic status, sex, birth order, maternal marital status, and # family residence changes prior to 57 months.
DRAFT -- DO NOT CITE OR QUOTE 26 August 28,2000
-------
1 declined to 7.9 ug/dL by age 11-13, when 375 children remained in the cohort. Children from
2 more advantaged backgrounds were more likely to remain in the cohort through age 11-13 years
3 than children from disadvantaged backgrounds.
4 Measures of developmental status were made at ages 2 years (Bayley scales), 4 years
5 (McCarthy scales), and at 7 and 11-13 years (Wechsler intelligence scale). In addition, emotional
6 and behavioral problems were assessed by their mothers using a Child Behavior Checklist, and
7 other demographic parameters were measured via questionnaire.
8 The key findings within this longitudinal study on how lead exposure is associated with
9 decreased developmental status measures were as follows:
10 1. The inverse relationships between IQ at age 7 years and average lifetime blood-
11 lead concentration measured at 15 months and at 2, 3, and 4 years were
12 statistically significant at the 0.05 level after adjusting for potential confounding
13 factors5. This result held for both full-scale and verbal IQ measures, but not
14 performance IQ.
15 2. Despite blood-lead concentration declining after age 2 or 3 years, the association
16 with cognitive development continued into later childhood. Furthermore, lifetime
17 blood-lead concentration was significantly associated with childhood emotional
18 and behavioral problems (after adjusting for such confounding factors as HOME
19 score, maternal psychopathology, and child's IQ) at ages 11-13 years.
20 Finding #1 was consistent with the Boston prospective study in that no significant relationship
21 was observed between IQ and pre-natal blood-lead concentration at approximately age 7 years,
22 while the relationship was significant when considering blood-lead concentration at 2 years of
23 age.
24 2.2.3 Conclusions on Causality
25 The combined weight of human and animal studies provide evidence, consistent with the
26 principles of causality presented in Section 2.2.1, that lead may be assumed to cause adverse
27 neurological effects in young children. In particular, longitudinal studies in humans have shown
28 that disturbances occur in neurobehavioral development early in life even at low lead exposure
29 levels. These studies have observed effects of lead exposure even after accounting for other
30 demographic factors (e.g., socioeconomic status, parents' IQ) that could affect neurological
31 development. While these other demographic factors tend to be highly correlated with blood-
32 lead levels early in life, the influence that post-natal lead exposure has on blood-lead tends to
5 Sex, parents' level of education, maternal age at delivery, parents' smoking habits, socioeconomic status, quality of
home environment, maternal IQ, birth weight, birth order, feeding method (bottle, breast, both), duration of breast feeding,
whether the child's natural parents were living together
DRAFT - DO NOT CITE OR QUOTE 27 August 28,2000
-------
1 increase with a child's age. This is because over time, measures of a child's lead exposure tend
2 to change at a faster rate than the child's demographic measures, and more recent lead exposures
3 continue to be predictive of a child's current health consequences. For the §403 risk analysis,
4 adverse neurological effects are assessed through IQ score decrements. Reasons for selecting IQ
5 score decrement as a health effect endpoint are presented in Section 2.5.2 of the §403 risk
6 analysis report.
7 2.3 THE ASSOCIATION BETWEEN BLOOD-LEAD CONCENTRATION
8 AND IQ SCORE
9 In its risk characterization, the §403 risk analysis used IQ score decrement associated
10 with lead exposure as the basis for measures of neurological effects. As the risk analysis used
11 blood-lead concentration as its primary measure of body lead burden to quantify environmental-
12 lead exposure, it was necessary to determine IQ score as a function of blood-lead concentration
13 and to characterize the extent to which a change in IQ score occurs when blood-lead
14 concentration changes within a child. The following assumptions were made in this analysis on
15 the association between blood-lead concentration and IQ score for the representative population
16 of 1-2 year old children:
17 • The relationship between blood-lead concentration and IQ score decrement was
18 assumed to be linear.
19 • The risk characterization assumed a loss of 0.257 IQ points per 1 ng/dL increase
20 in blood-lead concentration (with alternatives of 0.185 and 0.323 considered in
21 sensitivity analyses).
22 • No threshold was assumed in this relationship (i.e., no blood-lead concentration
23 exists below which a relationship between blood-lead concentration and IQ score
24 is not apparent), although selected non-zero thresholds have been assumed in
25 sensitivity analyses presented in Sections 5.1.5 and 6.4.2 of this report.
26 While the §403 risk analysis report discusses the basis for making these assumptions (e.g., see
27 Section 4.4 and Appendix D2), this section presents additional information that is necessary to
28 judge the correctness and accuracy of the assumptions. Section 2.3.1 addresses the linearity and
29 slope assumptions, while Section 2.3.2 addresses the threshold assumption.
30 2.3.1 Linearity and Slope Assumptions
31 How was such an assumption made ? As researchers have used primarily linear and log-
32 linear models to characterize the relationship between blood-lead concentration and IQ scores ,
33 these two types of models were considered for use in the risk analysis. The log-linear model
34 predicts IQ score as a linear function of log-transformed blood-lead concentration (plus other
35 important confounding variables, such as maternal IQ and HOME score), while the linear model
36 does not take a log transformation of the blood-lead concentration. The scientific community
DRAFT -- DO NOT CITE OR QUOTE 28 August 28.2000
-------
1 does not appear to have reached a consensus on which form is more appropriate. For example,
2 the meta-analysis in Schwartz (1994) included three studies that employed log-linear models and
3 four studies that employed linear models.
4 To obtain a single measure of the relationship that would be comparable across studies,
5 despite the different model forms used, Schwartz (1994) used the change in IQ score associated
6 with a doubling of blood-lead concentrations from 10 to 20 ug/dL. The meta-analysis (Schwartz,
7 1994) yielded an estimated decrease of 2.57 IQ points for an increase in blood-lead concentration
8 from 10 to 20 ug/dL. This was the slope estimate used in the §403 risk analysis.
9 Using the measure discussed in the previous paragraph, Schwartz (1994) provides some
10 evidence that the log-linear relationship may be more appropriate than the linear relationship. In
11 an analysis to investigate the presence of a threshold (see Section 2.3.2), Schwartz (1994)
12 estimated an IQ point decrement of 3.23 IQ points for the three studies with mean blood-lead
13 concentrations below 15 ug/dL, compared to a 2.32 IQ decrement for the four studies with mean
14 blood-lead concentrations at or above 15 ug/dL. Thus, if anything, a trend toward greater IQ loss
15 associated with lower blood-lead concentrations was observed. This result is consistent with a
16 log-linear relationship.
17 Despite this evidence, a linear relationship was applied in the §403 risk analysis. The
18 assumption of a linear model reduces the likelihood of overestimating the number of children
19 with low blood-lead concentrations at risk, or who may benefit from actions taken in response to
20 the §403 standards. See Section 4.2.1 and Appendix D2 of the §403 risk analysis report for
21 additional information.
22 Additional information: Tables D2-1 and D2-2 in Appendix D2 of the §403 risk analysis
23 report summarize a total of 18 studies that report the relationship between children's blood-lead
24 concentration and IQ. Each of these studies was used in at least one of the meta-analysis studies
25 reviewed in Appendix D2 of the §403 risk analysis report. Table 2-2 provides a subset of the
26 information previously reported in Tables D2-1 and D2-2 of the §403 risk analysis report for
27 these studies and also reports the type of model (linear or log-linear) which each study used to
28 predict IQ as a function of blood-lead concentration.
29 A few of the studies included in Table 2-2 used a log-linear model rather than a linear
30 model to characterize the effect of blood-lead concentration on IQ. However, the overall
31 evidence that these studies provide regarding a log-linear relationship was more limited than the
32 existing evidence on a linear relationship. Furthermore, if EPA had adopted a log-linear model
33 approach, the risk analysis would have estimated that blood-lead concentration had a greater
34 impact on IQ at lower levels than at higher levels. This would have resulted in a greater
35 possibility that the risk analysis would have overestimated benefits at lower levels, compared to
36 underestimating benefits. For these reasons, EPA felt that a linear model was the better approach
37 over a log-linear model. However, the §403 risk analysis did include a sensitivity analysis which
38 considered the effects of a steeper slope in the linear model, in order to evaluate the possibility of
39 underestimating the relationship between blood-lead concentration and IQ.
DRAFT -- DO NOT CITE OR QUOTE 29 August 28,2000
-------
1
2
Table 2-2. Summary of Key Findings from Studies that Investigate the Relationship
Between Blood-Lead Concentration and IQ Score
Study
Hatzakis et al. (1987)
Hatzakis et al. (1989)
Bellinger et al. (1991)
Bellinger et al. (1992)
Baghurstetal (1992)
Ernhartetal. (1989)
Cooney et al. (1991)
Schroeder et al (1985)
Hawketal. (1986)
Dietrich et al (1993)
Yuleetal (1981)
Lansdown et al (1986)
Wmnekeetal (1990)
Silva (1988)
Harvey et al (1988)
Wangetal (1989)
Wmneke et al. (1985a)
Fulton etal (1987)
Type of
Study
Prospective
Prospective
Prospective
Prospective
Prospective
Prospective
Prospective
Prospective
Replication
of Schroeder
Study
Prospective
Pilot Study
Replication
of Yule
Study
Multi-Center.
Cross -
Sectional
Study
Cross -
Sectional
Cross -
Sectional
Cross -
Sectional
Cross -
Sectional
Cross -
Sectional
Location
Lavrion, Greece
Lavrion. Greece
Boston, MA
Boston. MA
Port Pine,
Australia
Cleveland, OH
Sidney. Australia
Wake County.
NC
Lenoir & New
Hanover
counties. NC
Cincinnati, OH
London, England
London.
England
Bucharest
Budapest
Moden
Sofia
Dusseldorf
Dusseldorf
Dunedin, New
Zealand
Birmingham,
England
Shanghai, China
Nordenham,
Germany
Edinburgh.
Scotland
N
509
509
150
147
494
212
175
104
75
231
166
166
301
254
216
142
109
109
579
177
157
122
501
PbB
Mean (SD)
U/g/dL)
23.7 (9.2)
23.7 (9 2)
64(4 1)
6 5 (4.9)
20
167 (645)
142
20.9 (9 7)
15 2 (11 3)
1352 (4 13)
12 75 (307)
GM = 189 (1 3)
GM = 18 2 (1 7)
GM=11 0 (1 3)
GM=182(1.6I
GM = 83 (1 4)
74(1 3)
11 1 (491)
12.3 (02)
21 1 (10 11)
8.2 (1 4)
GM = 11 5
IQ Score
Mean (SD)
87 7 (14.8)
1155(145)
119 1 (148)
1047
87 5(166)
Range =
45-140
Range =
59-118
86.9(11 3)
98 21 (1344)
105 24 (14 2)
116
1089(15 12)
1059(106)
89
1202(103)
112(134)
Association Between IQ and Blood-
Lead Levels
Change in IQ
as PbB
increases
*__-_
iTOm
10-20 i/g/dL
-2 7
-27
-1 6
-58
-33
-1 1
+ 04
-20
-2 6
-1 3
-56
+ 1 5
-1 5
-9
-26
P-Value
<0001
<0001
023
0007
004
<001
<001
<005
<0 10
0084
063
<0 1
<0 1
<0 1
<0 1
<0 1
<0 1
<0 1
0003
Model
Form
linear
linear
log-linear
linear
log-linear
linear
linear
linear
linear
linear
log-linear
log-linear
linear
linear
linear
linear
linear
linear
log-linear
linear
linear
linear
log-linear
3
4
5
6
1
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
PbB = blood-lead concentration (pg/dL), SD = standard deviation, GM = geometnc mean
DRAFT - DO NOT CITE OR QUOTE
30
August 28, 2000
-------
1 A recent article by Marais and Wecker (1998) has suggested that researchers who have
2 characterized IQ as a function of blood-lead concentration using linear regression techniques
3 have often reported biased estimates for the effect of blood-lead concentration on IQ for one or
4 both of the following reasons:
5 • by not having all four of the following predictor variables in the model: blood-
6 lead concentration, mother's intelligence, father's intelligence, and socioeconomic
7 status.
8 • by not taking into account measurement error in these predictor variables.
9 By analyzing data from four case studies, the authors imply that the bias overestimates the effect,
10 thereby making it likely that the researcher would declare that blood-lead concentration
11 (especially at low levels) has a significant effect on IQ, when in reality, such an effect is
12 insignificant. The authors show how to arrive at an estimate of the blood-lead effect that is not
13 subject to this bias; this estimate is a function of the correlations among the four predictor
14 variables and the measurement variability associated with these variables. The article prompted
15 two responses that were published simultaneously with the article, both of which challenged the
16 article' s conclusions.
17 Despite their claims, the findings in Marais and Wecker (1998) have not resulted in any
18 change to the approach taken by the §403 risk analysis to characterize the relationship between
19 blood-lead concentration and children's IQ for the following reasons:
20 • Sensitivity analyses performed by other researchers6 and documented in a
21 response published with the article by Marais and Wecker (1998) have shown that
22 the approach to obtaining an "unbiased" estimate for the effect of blood-lead on
23 IQ is highly sensitive to the values of the estimates for the correlations and
24 variability among the four predictor variables that are input to the calculation,
25 thereby implying that input values that do not represent the target population can
26 lead to a highly inaccurate estimate for this effect under this approach.
27 • The authors have not shown that any overestimation associated with this bias is
28 always significantly large enough to warrant concern or will result in an incorrect
29 declaration that blood-lead has a significant effect on IQ.
30 • The meta-analysis documented in Schwartz (1994), which the §403 risk analysis
31 used to characterize the relationship between blood-lead concentration and IQ,
32 utilizes the estimated blood-lead effects on IQ that were reported by seven studies,
33 six of which estimated these effects after taking into account both parental IQ and
6 Watemaux, C, Petkova, E., and DuMouchel, W. "Comment: Problems with Using Auxiliary Information to Correct
for Omitted Variables When Estimating the Effect of Lead on IQ." Journal of the American Statistical Association. 93:505-513
DRAFT -- DO NOT CITE OR QUOTE 31 August 28,2000
-------
1 socioeconomic status (i.e., the predictor variables of most concern to Marais and
2 Wecker). Furthermore, the §403 risk analysis considered not only the outcome of
3 this meta-analysis, but also the sensitivity analyses associated with this analysis,
4 when investigating the effect of deviation from the meta-analysis outcome on the
5 risk analysis results.
6 • The need to adjust for measurement error in the predictor variables is not relevant
7 to the §403 risk analysis, as the goal is to predict how a measured blood-lead
8 concentration (after adjusting for the measured values of other potentially
9 important variables) is associated with IQ.
10 The issue of whether a linear model is appropriate over the entire range of blood-lead
11 concentration must address the presence of a threshold in the relationship. This is discussed
12 further in the next section.
13 2.3.2 Threshold Assumption
14 Despite the claims of some researchers on the presence of a threshold in the blood-
15 lead/IQ relationship, the majority of findings across studies and in meta-analyses have failed to
16 find sufficient evidence of a non-zero threshold. Furthermore, when claims of a non-zero
17 threshold were made, the value of this threshold (when suggested) differed considerably across
18 these claims. Therefore, the approach taken in the §403 risk analysis was to assume that no
19 threshold exists (although risk calculations assuming certain non-zero threshold values have been
20 included in sensitivity analyses found in Sections 5.1.5 and 6.4.2 of this document).
21 In Section 3.3 of USEPA (1998b), the SAB concurred that "available data have not
22 identified a clear threshold," and, therefore, "the assumption of no threshold for lead effects on
23 IQ score is both defensible and appropriate statistically." However, it was desired to document
24 the technical justifications for this assumption more thoroughly. Furthermore, the investigation
25 into the presence of a threshold could be addressed by evaluating whether the dose-response
26 function is linear across the entire range of blood-lead concentration.
27 How was such an assumption made? The assumption of no threshold made in the §403
28 risk analysis was based on the findings of Schwartz (1994), who noted that the presence of a
29 threshold would result in a decline in the estimated slope associated with blood-lead
30 concentration as the range of blood-lead concentrations declined across studies. However, as
31 mentioned in Section 2.3.1 above, a larger effect size was observed in the four studies with mean
32 blood-lead levels of 15 ug/dL or lower (-0.323 ± 0.126) compared to the other three studies
33 (-0.232 ± 0.040). This observed trend toward higher slopes at lower concentrations discounted
34 the likelihood of a threshold.
35 Also, Schwartz (1994) examined data from the Boston prospective lead study (discussed
36 in Section 2.2.2 above) specifically to investigate the presence of a threshold. This study was
37 selected as it had the lowest mean blood-lead concentration at two years of age (6.5 ug/dL;
DRAFT -• DO NOT CITE OR QUOTE 32 August 28,2000
-------
1 n=133) of the studies considered in the meta-analysis, thereby allowing thresholds at low blood-
2 lead levels to be identified if present. In addition, the study cohort's high socioeconomic (SES)
3 standing may have limited the likelihood of certain confounding, and the Boston study
4 coordinators found relatively weak association between blood-lead concentration at two years of
5 age and various sociodemographic characteristics and psychosocial environment parameters
6 (Bellinger et al., 1986; Bellinger et al., 1992).
7 Schwartz's examination of the Boston study data involved fitting two separate regression
8 curves to the same set of covariates: one using IQ score (at 10 years) as the dependent variable,
9 and the other using blood-lead concentration (at 2 years) as the dependent variable. The
10 covariates included age, race, stress and HOME scores, maternal IQ, educational level and
11 occupational status for each parent, mother's time working out of the house, marital status,
12 gestational age, birth weight, mother's use of alcohol during pregnancy, otitis media history, birth
13 order, and SES. Then, a nonparametric smoothed curve (LOESS) was used to characterize the
14 residuals from the IQ score regression as a function of the residuals from the blood-lead
15 regression. The residuals were used in this curve-fitting exercise as they represent blood-lead
16 and IQ score measures after any effects of the above covariates have been removed. The LOESS
17 technique allowed for nonlinear curve fits, such as those that would result if a threshold was
18 present. The curve fit suggested that IQ score decrement was associated with declines in blood-
19 lead concentration even when blood-lead levels were below 5 ug/dL, supporting the hypothesis
20 that a blood-lead threshold on IQ score decrement was essentially not present.
21 In Schwartz (1993), this nonparametric smoothing approach was performed on McCarthy
22 index data collected at age 57 months and blood-lead concentration data collected at 24 months,
23 as recorded in the Boston prospective lead study (Bellinger et al., 1991). Again, after adjusting
24 for potential confounding variables, a definite relationship was observed even at levels below 10
25 ug/dL, with no evidence of a threshold (Schwartz, 1993). To allow any potential threshold to be
26 identified, a piecewise-linear regression model was fitted to these data which allowed the
27 relationship to resemble a "hockey-stick" (i.e., the fit resembled two lines of different slopes that
28 meet at some point representing the potential threshold, with the line below the threshold having
29 nearly a zero slope, and the line above the threshold having a larger, positive slope). This model
30 fit suggested that any potential threshold would be less than 0.0001 ug/dL (Schwartz, 1993).
31 The meta-analysis by Pocock et al. (1994) involving 26 studies concluded that no single
32 study has collected a sufficient amount of information to make definitive statements on the
33 presence of a threshold, and contradictory results (due to chance) on the presence of a threshold
34 can be observed for different studies. Thus, the analysis did not have enough evidence to reject
35 the hypothesis that no threshold exists.
36 Identifying a threshold. If a statistical hypothesis is used to determine the presence of a
37 threshold, the test should take the following form:
38 Null hypothesis: No threshold exists (i.e., the "threshold" is at 0 Mg/dL)
39 Alternative hypothesis: A non-zero threshold exists.
DRAFT--DO NOT CITE OR QUOTE 33 August 28,2000
-------
1 If a statistical test is used as a scientific basis for making a decision, one either rejects the null
2 hypothesis or fails to reject it. One never says that the null hypothesis is "true." Therefore, given
3 a set of data and the statistical methods being applied, one either rejects the hypothesis that no
4 threshold exists or cannot reject it.
5 The statistical method used to test the above hypotheses can also vary from study to
6 study. In general, the method is applied as part of an investigation into a dose-response
7 relationship between blood-lead concentration and IQ. Two examples of statistical approaches
8 are as follows:
9 • Several studies (e.g., Dietrich et al., 1993; Hatzakis et al., 1989) investigated
10 dose-response by placing the study cohort into from 5 to 10 groups according to
11 blood-lead concentration, determining the predicted IQ score associated with the
12 mean blood-lead concentration in each group (using some pre-determined
13 regression model), calculating confidence intervals associated with the prediction,
14 and determining how the groups differ in their predictions (as well as any patterns
IS among the groups).
16 • Another approach focuses on attempting to fit the piecewise-linear "hockey-stick"
17 regression model discussed above that predicts IQ score as a function of blood-
18 lead concentration (and other confounding variables), where the fitted line has a
19 different (larger, positive) slope once blood-lead concentration achieves a certain
20 level, which is interpreted as a threshold value.
21 Problems associated with suggesting that a threshold exists: Determining whether a
22 threshold exists in the relationship between blood-lead concentration and IQ score is problematic
23 due to the difficulties in accurately characterizing the blood-lead/IQ relationship and the inability
24 to generalize findings across studies and to the nation as a whole. Major sources of these
25 difficulties include the following:
26 • Different protocols for measuring IQ and different IQ measures (e.g., performance
27 IQ, verbal IQ, full-scale IQ measures are all associated with the Wechsler
28 protocol) are used in different studies.
29 • Study designs differ, as do the methods used to make inferences from the data.
30 • Children's IQ can be difficult to measure and can be more variable than adult IQ.
31 • Outcomes are often highly dependent on the given set of confounding variables
32 being considered. This set differs from one study to the next. Furthermore, when
33 multiple studies consider the same confounding variables, these variables are
34 often measured differently, using different protocols, from study to study.
DRAFT -- DO NOT CITE OR QUOTE 34 August 28,2000
-------
1 • Different ages of children and different ranges of blood-lead concentration are
2 found across studies.
3 A non-zero threshold would result in reduced estimates of the likelihood of adverse
4 health effects, as children with blood-lead concentrations below the threshold would no longer be
5 labeled as experiencing an exposure-related IQ decrement. The level of reduction would depend
6 on how large the value of the threshold is. Thus, if a decision on a non-zero threshold was made
7 in error, the incidences of adverse health effects would be underestimated. The impact that a
8 non-zero threshold has on reducing the risk estimates calculated in this risk analysis is addressed
9 in sensitivity analyses presented in Sections 5.1.5 and 6.4.2 of this document.
10 Examples of Possible Non-zero Thresholds Concluded from Study Findings. Relatively
11 high thresholds (e.g., 10 ug/dL or above) have been suggested in some older studies conducted
12 10 or 15 years ago. However, some of the higher suggested thresholds appear to have lost their
13 legitimacy as they are higher than the levels for which more recent studies have observed some
14 type of health effect. For those older studies that did not report a possible threshold, many
15 involved children with a range of blood-lead concentrations that would be considered high by
16 today's standards. Thus, the findings from these studies cannot be used to determine whether
17 thresholds exist at lower lead levels (i.e., levels below the observed ranges). A study's design
18 must allow for a sufficiently large range of blood-lead concentrations, and in particular, cover a
19 sufficient range of lower-lead levels (i.e., below 5 or 10 ug/dL), to ensure that any threshold
20 value would occur within the observed range.
21 Some researchers attempting to prove the existence of a threshold have reviewed results
22 of applying the first statistical approach above (i.e., making predictions within groups of the
23 cohort), but have made conclusions based upon simple plots of the results rather than by citing
24 the outcome of statistical comparisons. For example, Kaufman (1996) has concluded that
25 threshold effects may exist at about 20 ug/dL from data presented in Dietrich et al. (1993), at
26 from 10-15 ug/dL from data presented in Bellinger et al. (1992), and at from 25-35 ug/dL in
27 Hatzakis et al. (1989). However, each conclusion was based on visually interpreting selected
28 figures within these articles rather than on the results of controlled statistical hypothesis tests.
29 Furthermore, the following must also be considered when interpreting these conclusions:
30 • Dietrich et al. (1993): From this prospective study conducted in Cincinnati, OH,
31 Kaufman (1996) cites the authors' presentation of predicted Wechsler Scale
32 Performance IQ (PIQ) for four groups of children approximately 6.5 years of age,
33 where the groups are determined by lifetime mean blood-lead concentration (i.e.,
34 average blood-lead concentration measured at 3-month intervals from age 3 to 60
35 months, and at ages 66 and 72 months). The predicted PIQ score was lower in the
36 group with the highest lifetime mean blood-lead concentration (>20 ug/dL)
37 compared to the other groups. However, the authors caution against interpreting
38 this finding as evidence of a threshold effect, as most children in this group had
39 one or more individual measurements above 30 ug/dL. Furthermore, in his meta-
40 analysis, Schwartz (1994) considered a different relationship cited by the authors:
DRAFT - DO NOT CITE OR QUOTE 35 August 28,2000
-------
1 full-scale IQ as a function of average blood-lead concentration through age 3 years
2 (and other covariates, including HOME score and material IQ). This relationship,
3 cited in Appendix D2 of the §403 risk analysis report, is more relevant to the
4 representative population in the risk analysis.
5 • Bellinger et al. (1992): From this prospective study conducted in Boston, MA,
6 Kaufman (1996) cites the authors' presentation of predicted WISC-R full-scale IQ
7 and K-TEA Battery Composite scores for four groups of 10-year old children,
8 where the groups are determined by blood-lead concentration at 24 months of age
9 (PbB24). Kaufman (1996) indicates that differences were apparent only between
10 the two lowest and the two highest groups, suggesting an apparent threshold
11 between them (i.e., 10-15 Mg/dL). However, little difference among any of the
12 four groups would have been identified if the assertion was based on confidence
13 intervals associated with the predictions, rather than standard errors for the
14 individual groups. Meanwhile, the regression model used to predict IQ from
15 PbB24 indicated a highly significant linear trend (p=0.007) across the entire range
16 of observed values of PbB24. This trend was present even among the three groups
17 having the lowest blood-lead concentrations, suggesting that the trend is in fact
18 present at the lower range of the observed concentrations (i.e., below 15 Mg/dL).
19 • Hatzakis et al. (1989): From this study conducted in Greece within a city in which
20 lead mining and smelting occurred, Kaufman (1996) cites how the authors present
21 predicted full-scale IQ for primary school-aged children grouped by blood-lead
22 concentration. Blood-lead concentrations in this study were high: the average
23 blood-lead concentration in this study was 23.7 ug/dL, no child had blood-lead
24 concentration below 7 ug/dL, and more than 90% of the children exceeded 10
25 Mg/dL. The mean predicted IQ in the first two groups (s 14.9 Mg/dL, 15-24.9
26 Mg/dL) appeared to be statistically equivalent, then steadily declined for the
27 remaining three groups, suggesting the presence of a threshold around 25 Mg/dL.
28 However, many other more recent studies (including animal studies) have
29 observed neurological and developmental effects at lower blood-lead
30 concentrations, making the concept of a threshold at 25 Mg/dL highly unlikely. In
31 fact, the linear regression model developed in this study (which included 17
32 covariates) had a highly significant slope for blood-lead concentration (p < 0.001)
33 across the entire range of data in this study, even though more than 50% of the
34 data occurred from 7-25 Mg/dL. Furthermore, it is unclear if different conclusions
35 would have been made if the groups of children were defined differently.
36 These examples illustrate the complexities associated with characterizing the dose-response
37 relationship and the ability to conclude that a threshold exists in this relationship.
38 The findings of a study by Fulton et al. (1987), which was included in the Schwartz
39 (1994) meta-analysis, appear to discount the high threshold level suggested by the findings of
40 Hatzakis et al. (1989). This study was conducted on 501 children aged 6-9 years in Scotland.
DRAFT - DO NOT CITE OR QUOTE 36 August 28,2000
-------
1 The predicted B ASC score was calculated for ten groups of children determined by log-
2 transformed blood-lead concentration. No evidence of a threshold was found among these data,
3 and the estimated slope associated with log-transformed blood-lead concentration was significant
4 (p=0.003). These findings were observed despite having only 10 study children with blood-lead
5 concentrations exceeding 25 ug/dL.
6 Concluding Possible Thresholds for Tooth-Lead Concentration: Some studies (e.g.,
7 Bellinger and Needleman, 1983; Rabinowitz et al., 1992) have observed the potential for
8 thresholds in tooth-lead concentration when relating tooth-lead concentration to IQ score. Based
9 on their investigation of the relation between tooth-lead concentration and IQ score, Rabinowitz
10 et al. (1992) suggested that a threshold for blood-lead concentration exists at approximately 8
11 ug/dL. Their investigation was centered around their 1989-1990 study of 764 children in grades
12 1-3 in Taiwan (an average age of 6.7 years). In this study, teeth shed by these children were
13 analyzed for lead. In addition, the children were administered Raven's Colored Progressive
14 Matrices (CPM) test, the score of which is considered a measure of IQ (average=25, SD=5.7). A
15 model was developed which found CPM test score to be highly correlated with selected non-lead
16 predictors (parental education level, sex, grade level, and whether or not the child is
17 ambidextrous). The difference between the model-predicted and observed test scores for a child
18 (the "CPM score deficit") was interpreted as a measure of the change in the test score that results
19 from lead exposure.
20 Each of the 380 children for which CPM score deficit could be calculated was placed into
21 one of two groups according to whether or not their tooth-lead level (|ag/g) exceeded a specified
22 value. Then, a Mann-Whitney test was performed to determine whether the mean CPM score
23 deficit differed significantly between the two groups. This was done for a series of grouping
24 values for tooth-lead, from 2 to 6 ug/g. Significant differences between the two groups (p <
25 0.05) were seen at grouping levels of tooth-lead at 3.5 ng/g or above, but not at 3 ug/g or below.
26 Therefore, the authors concluded that a tooth-lead threshold for intelligence deficit existed at
27 approximately 3.25 ug/g. Finally, the authors relate this tooth-lead threshold value to blood-lead
28 by applying a modeled relationship between tooth-lead and blood-lead levels (formulated from
29 data for 88 Boston children aged 57 months), and concluding that a tooth-lead threshold of 3.25
30 pg/g corresponded roughly to a blood-lead threshold of 8 ug/dL (± 2 ug/dL).
31 While the cohort was considered to have low tooth-lead concentrations, the following
32 must be considered when interpreting the above conclusion on the presence of a threshold and its
33 relevance to the §403 risk analysis:
34 • A non-zero threshold existing for tooth-lead concentration does not necessarily
35 imply that one exists for blood-lead concentration, as tooth-lead may impact
36 children's health in a different way from blood-lead.
37 • When noting the lack of significant difference between two groups defined by
38 whether or not tooth-lead exceeds a given threshold when this threshold gets low
39 enough, it is uncertain of the extent to which the lack of significance is actually
DRAFT -- DO NOT CITE OR QUOTE 37 August 28,2000
-------
1 due to reduced power to detect differences as the sample size in the non-
2 exceedance group declines with the threshold being considered.
3 • It is uncertain whether the model to predict blood-lead based on tooth-lead from
4 the Boston study, which was used to obtain the blood-lead threshold estimate of 8
5 ug/dL, can be applied directly to the findings of the Taiwan study without needing
6 to consider certain statistical issues. For example, measurement error associated
7 with tooth-lead levels may differ between the Boston and Taiwan studies.
8 • Rabinowitz et al. (1992) state that, when attempting to fit a "hockey-stick"
9 regression model to the data, "this data shows no change in the slope (or intercept)
10 of the lines across any trial threshold." While this statement appears to support
11 the hypothesis that no tooth-lead threshold exists, the authors do not provide any
12 further information on the outcome of this model-based analysis.
13 2.3.3 Verifying the Results of Schwartz (1994)
14 Schwartz (1994) applied a random effects modeling approach suggested by DerSimonian
15 and Laird (1986), a highly-regarded reference on meta-analysis. As a result, it was considered
16 among the best encountered by the §403 risk analysis, and therefore, was an important
17 contributor to how the §403 risk analysis characterized the relationship between blood-lead
18 concentration and IQ score in children. For this reason, the meta-analysis findings were verified
19 as part of the §403 risk analysis. Using either the weighted noniterative method or the weighted
20 maximum likelihood method suggested by DerSimonian and Laird (1986), the §403 risk analysis
21 obtained the same finding as Schwartz (1994): that a decrease of 0.257 (± 0.041) IQ points was
22 associated with an increase in blood-lead concentration of 1.0 ug/dL within the range of 10-20
23 ug/dL. In addition, it was noted that heterogeneity of variance among the seven studies
24 considered by Schwartz (1994) was not significant, and the random effects model gave the same
25 results as a fixed effects model.
26 2.4 IMPACT OF CERTAIN RESIDENTIAL DUST
27 CHARACTERISTICS ON DUST-LEAD EXPOSURE
28 The bioavailability of lead can be an important factor in determining the toxic effects of
29 lead exposure to children within a specific environment. Because lead is found in a variety of
30 chemical and physical forms depending on its source, the bioavailability of lead has been studied
31 as a function of chemical make-up (i.e., the particular form of lead present) and particle size in
32 various environmental matrices (e.g., dusts and soils, mining wastes). Generally, the literature
33 concludes that the bioavailability of lead can depend on, among other things, the particular lead
34 species present (which varies depending on the source of lead), the size of the lead-containing
35 particles, the matrix incorporating the lead species, and the types of nutrients or other compounds
36 ingested with the lead (Freeman et al., 1992; USEPA, 1994). It has been suggested that lead
37 speciation and particle size may affect the bioavailability of lead through their influence on
38 solubility (USEPA, 1994). For example, lead bioavailability appears to be lower in mining areas
DRAFT - DO NOT CITE OR QUOTE 38 August 28,2000
-------
1 relative to urban and smelter areas. Some authors (e.g., Rieuwerts and Farago, 1995; Davis et al.,
2 1995; Freeman et al., 1992) have suggested that this difference may be, in part, explained by
3 variations in chemical form (dissolutions rates) and particle size. Studies have also shown that
4 correlations between soil-lead and blood-lead levels are influenced by particle size and
5 composition of the lead compounds (USDHHS, 1992).
6 The purposes of this section are:
7 1. To present a brief review of the some of the available literature which specifically
8 examines the bioavailability of lead in dust, as a function of particle size and
9 chemical composition of the dust.
10 2. To determine if there is evidence which warrants the consideration of particle size
11 and lead speciation. as related to lead bioavailability in dust, in the §403
12 rulemaking; and, if warranted, determine if there is sufficient information
13 available in the literature to allow for a thorough consideration.
14 3. If relevant, to identify significant information gaps and potential issues that may
15 warrant further research.
16 2.4.1 Review of Literature: Effects of Chemical Composition on
17 Lead Bioavailability in Dust
18 There is substantial evidence in the scientific literature that the particular chemical
19 species, as well as the matrix (e.g., mineralogy, organic matter content) within which the lead
20 compound is found, are important in determining lead bioavailability (USDHHS, 1999; USEPA,
21 1994). Many of the studies in the literature have been based on comparisons of relatively simple
22 lead compounds in controlled animal feeding studies or have focused on lead bioavailability in
23 urban and mining-associated soils. With respect to household dust in particular, there is
24 relatively little in the literature which specifically examines the relationship between
25 bioavailability and chemical composition.
26 The literature does recognize, however, that the composition of interior dust is
27 substantially influenced by soil and exterior dust (Diemel et al., 1981; USEPA, 1994). For
28 example, USEPA (1994) characterizes total lead in household dust as being comprised of soil-
29 lead, air-lead, lead from outside sources (e.g., workplace, school), and lead from household paint.
30 As a default value to the IEUBK exposure model, EPA has set the ratio of household dust-lead
31 concentration to soil-lead concentration at 0.70, which was considered appropriate for
32 neighborhoods or residences where loose particles of surface soil are readily transported into the
33 house (USEPA, 1994). Thus, soil particles have the potential to be a significant contributor to
34 lead levels in household dust.
35 The following sections will provide a brief review of scientific findings related to the
36 general physical and chemical principles of lead and how they relate to bioavailability differences
DRAFT - DO NOT CITE OR QUOTE 39 August 28.2000
-------
1 in controlled environments and in soil studies. Because the literature recognizes that the
2 composition of interior dust is influenced by soil and exterior dust, discussion of these general
3 bioavailability factors will, in the absence of more dust-specific data, serve as a starting point in
4 understanding the relationship between lead bioavailability in household dust and chemical
5 composition and particle size.
6 2.4.1.1 Research on Lead Bioavailabilitv in Controlled Animal Studies. The relative
7 bioavailabilities of simple lead compounds have been studied under controlled conditions in
8 animal studies. For example, Barltrop and Meek (1975) (as cited in USDHHS, 1992) compared
9 the absorptions of 12 different lead compounds in rats by measuring the kidney contents
10 following oral exposure. They found that the absorption of metallic-lead (particle size 180-250
11 //m) was the lowest of the lead compounds tested. Data also suggested that the absorption of
12 lead sulfide (particle size <50 //m) was significantly less than the oral bioavailability of other
13 lead salts (oxide, acetate). Lead carbonate had the highest absorption, which was suggested to be
14 due to its high solubility in gastric juice.
15 Dieter et al. (1993) also found differing blood-lead, bone-lead and kidney-lead levels in
16 rats fed different lead compounds, indicating variability in bioavailability. For example,
17 maximum blood-lead levels were higher (80 /ug/dl) in rats fed lead acetate and lead oxide, in
18 comparison to rats fed lead sulfide and a lead ore concentrate. Similar differences were observed
19 in bone-lead and kidney-lead levels between the rats receiving the more soluble (e.g., lead acetate
20 and oxide) and less soluble (e.g., lead sulfide and ore) lead compounds.
21 2.4.1.2 Research on Lead Bioavailabilitv in Soils. The literature also suggests that the
22 soil matrix itself is a major factor in determining the bioavailability of lead (Maddaloni et al.).
23 For example, Freeman et al. (1992) found that tissue-lead concentrations were lower in rats fed
24 lead-contaminated mining waste soils from Butte, Montana, as compared to rats fed comparable
25 doses of soluble lead acetate. It was suggested that the inherent chemical properties of soil-
26 adsorption sites and the alteration of lead-bearing solids (e.g., encapsulation processes which
27 inhibit dissolution) may reduce the bioavailability of soil-lead, as compared to lead ingested
28 without soil. In general, the fate and bioavailability of lead in soils are affected by the species of
29 lead incorporated into the soil, the degree of absorption at mineral interfaces, precipitation of
30 solid phases, and the formation of relatively stable complexes/chelates with organic matter, as
31 well as other complex soil matrix factors such as pH (USDHHS, 1999; McKinney, 1993;
32 Freeman et al., 1992; Maddaloni et al.). This has been suggested to be largely due to the
33 influence of these factors on solubility, although it is important to note that solubility is but one
34 factor in the bioavailability of lead to humans or animals (USEPA, 1994).
35 In a study of lead bioavailability in soil, Laperche et al. (1997) found that apatite (calcium
36 fluoride phosphate) amendments to a lead-contaminated soil lowered the bioavailability of soil
37 lead (as determined by plant uptake) by inducing the formation of geochemically-stable lead
38 phosphate compounds. Similarly, in a study of an old mining village with elevated lead levels in
39 both garden soils and house dust, Cotter-Howells (1994) identified the predominance of lead
40 phosphate compounds (of limited bioavailability) as probable explanation of why blood-lead
DRAFT -- DO NOT CITE OR QUOTE 40 August 28,2000
-------
1 levels in the village were not elevated in an otherwise contaminated area. In a study of mining-
2 associated soils in Butte, Montana, Davis et al. (1995) suggested that the predominance of lead
3 sulfide/sulfate and oxide/phosphates in soil and mine waste samples might provide an
4 explanation for the limited lead bioavailability that was observed when the Butte soils were fed
5 to rats in a previous study (Freeman et al., 1992).
6 Urban area soils are typically contaminated with alkyl lead species originating from
7 combustion of leaded gasoline; lead halides (chlorides and bromides) from auto exhaust
8 participates; or lead carbonate, chromate, and octoate (as chips, flakes, and dusts) from exterior
9 and interior lead-based paint (USEPA, 1994). Lead halides in soils are quickly transformed to
10 (or associated with) oxides or sulfates (USEPA, 1986 as cited in USEPA, 1994). In many lead-
11 mining districts, the predominant form of lead is galena or lead sulfide (USDHHS, 1992).
12 2.4.2 Review of Literature: Effects of Particle Size on Lead Bioavailability in Dust
13 Data in the literature are limited with specific regards to how particle size of lead-
14 contaminated house dust influences the bioavailability of lead in the dust. However, studies have
15 been conducted to examine the general relationship in soils between panicle size and
16 bioavailability, as well as particle size and lead concentration. For example, Barltrop and Meek
17 (1979) found that the bioavailability of lead in the intestinal tract of rats fed metallic-lead of
18 various particle sizes increased fivefold as particle size decreased from 197 microns to 6
19 microns. Particle size, due to kinetic limitations that control dissolution rates in the
20 gastrointestinal tract, has also been hypothesized to contribute to the lower bioavailability of lead
21 observed in mining waste soils relative to urban and smelter soils (Davis et al., 1995). In general,
22 the smaller the particle size, the greater the absorption of lead due to more rapid dissolution
23 (small particles have higher surface area to mass) in the gastrointestinal tract (Freeman et al., -
24 1992).
25 Que Hee et al. (1985) found that when lead concentrations were measured in dust samples
26 categorized by size fraction, lead concentration was generally independent of the particle size.
27 However, most of the dust particle mass (about 75%), and thus most of the lead (about 77%),
28 was present in the <149 //m size fraction. Lead concentration in smaller particle size ranges may
29 possibly maximize intestinal absorption, and thus increase bioavailability (USDHHS, 1992).
30 Duggan and Inskip (1985) performed an extensive literature review on the variation of lead
31 concentration with particle size and reported that higher lead concentrations are usually found in
32 the smaller-sized fractions of soil and dust. As reported by the Agency for Toxic Substances and
33 Disease Registry, numerous studies have also observed the lead content of soil, street dust, city
34 dust, and house dust to increase with decreasing particle size (USDHHS, 1992).
35 2.4.3 Information Gaps, Issues and Conclusions
36 Although the literature is generally lacking in data which directly address the
37 bioavailability of lead in household dust as a function of chemical composition and particle size,
38 by recognizing that interior dust composition can be greatly influenced by outside soil, it can
DRAFT - DO NOT CITE OR QUOTE 41 August 28,2000
-------
1 reasonably be expected that the factors which affect lead bioavailability in soils will also
2 influence the bioavailability of lead in household dust. Therefore, based on general knowledge of
3 the bioavailability of simple lead compounds and studies of lead compounds in soil matrices,
4 evidence suggests that particle size and chemical composition have the potential to significantly
5 affect lead bioavailability in dust. Nonetheless, the current information base which specifically
6 addresses particle size and chemical composition of dust as factors in lead bioavailability may be
7 inadequate to determine how such factors can reasonably be incorporated into the rulemaking
8 effort. Furthermore, needing to characterize dust by particle size and lead by chemical speciation
9 within a risk assessment will likely add to the expense of dust analyses, and dust standards that
10 distinguish between these various characterizations could add considerable complexity to the
11 rule.
12 Specific uncertainties that remain concerning bioavailability of lead in household dust
13 include the following: physical and chemical properties that may be unique to dust versus soil;
14 whether the effect of lead speciation in dust is significant enough to affect dust standards for
15 lead; distributions of lead across panicle sizes found in household dust (e.g., whether dust is
16 enriched with the smaller size fraction relative to outside soil) and whether particle size
17 differences are significant enough to affect standards; and possibly variances in exposure
18 mechanisms that may occur across particle sizes.
19 The need for further research in these and related areas has been supported by several
20 authors. For example, Freeman et al. (1992), based on comparisons of mining waste soils and
21 other soil types in reviewed studies, emphasized the importance of evaluating the soil mineralogy
22 and lead species present when predicting bioavailability values for lead in soils. In addition,
23 USEPA (1994) in the Guidance Manual for the Integrated Exposure Uptake Biokinetic Model
24 for Lead in Children, notes that adequate characterization of lead contaminated media, for the
25 purpose of estimating bioavailability, should include assessment of physical and chemical
26 parameters, such as particle size and media solubility.
DRAFT -- DO NOT CITE OR QUOTE 42 August 28,2000
-------
1 3.0 EXPOSURE ASSESSMENT
2 According to Chapter 3 of the §403 risk analysis report, the goal of the exposure
3 assessment was to document the important sources of lead in the environment, to document the
4 major pathways by which children are exposed to lead, to characterize the current (baseline)
5 distribution of environmental-lead levels in the nation's housing stock, and to characterize the
6 current distribution of average blood-lead concentration among the nation's children.
7 In particular, Chapter 3 introduced those data sources used to characterize environmental-
8 lead levels in the nation's housing stock and presented summaries of household average lead
9 levels in dust and soil as reported in these studies. The U.S. Department of Housing and Urban
10 Development (HUD)'s National Survey of Lead-Based Paint in Housing ("HUD National
11 Survey", Section 3.3.1.1 of the §403 risk analysis report) was selected as the data source for
12 characterizing baseline environmental-lead levels in the nation's housing stock. Pre-intervention
13 data from other selected studies, such as the Rochester Lead-in-Dust study and the ongoing
14 Evaluation of HUD Lead-Based Paint Hazard Control Grant Program ("HUD Grantees") were
15 also summarized in Section 3.3.1 of the §403 risk analysis report to provide supporting
16 information on environmental-lead levels and to obtain information on the relationship between
17 these levels and blood-lead concentration in children.
18 Since the §403 risk analysis report was published, additional data on environmental-lead
19 levels in the nation's housing stock have been made available to EPA. These data include interim
20 data from the National Survey of Lead and Allergens in Housing, and additional data from the
21 HUD Grantees evaluation. In addition, updated data from the U.S. Census Bureau are available
22 on numbers of young children associated with the various types of lead exposures found in the
23 national housing stock. Some comments on the §403 proposed rule suggested that EPA use these
24 additional data when available. Therefore, EPA has investigated these new data to document
25 additional, more recent information on lead levels in the nation's housing stock and, when
26 available, blood-lead levels in children exposed to these lead levels. For example, it was of
27 interest to document more recent information on the distribution of lead levels in dust deposited
28 on interior uncarpeted floors and window sills (i.e., the surfaces included in the proposed §403
29 standards), as well as on other types of surfaces (e.g., exterior surfaces, window troughs) to help
30 evaluate their potential contribution to overall lead exposure at a residence. It was also of
31 interest to characterize the national distribution of residential soil-lead levels and percentages of
32 the housing stock whose soil-lead levels exceed specified thresholds. Therefore, this chapter
33 provides additional information on lead exposure within the following sections:
34 • Section 3.1: Information on the National Survey of Lead and Allergens in
35 Housing (NSLAH), a national survey begun in 1997 of lead levels in dust and soil
36 in U.S. residential housing.
37 • Section 3.2: Comparison of the HUD National Survey data summaries for dust-
38 lead loading and soil-lead concentration with summaries from other lead exposure
39 studies, including interim data (for 706 households) from the NSLAH and pre-
DRAFT - DO NOT CITE OR QUOTE 43 August 28.2000
-------
1 intervention data from the HUD Grantees evaluation that have been revised and
2 augmented since the §403 risk analysis report was published).
3 • Section 3.3: Information on the prevalence of soil pica tendencies in young
4 children and how such tendencies may occur over and above paint pica
5 tendencies.
6 • Section 3.4: Updated information on numbers of children in the nation's housing
7 stock, using interim data from the NSLAH.
8 • Section 3.5: Distribution of dust-lead levels on surfaces other than uncarpeted
9 floors and window sills.
10 • Section 3.6: Revised summaries of pre-intervention blood-lead concentration
11 based on updated data from the HUD Grantees evaluation.
12 3.1 THE NATIONAL SURVEY OF LEAD AND ALLERGENS IN HOUSING
13 The National Survey of Lead and Allergens in Housing (NSLAH) is a currently-ongoing
14 survey sponsored by the U.S. Department of Housing and Urban Development (HUD) and the
15 National Institute of Environmental Health Sciences (NEEHS) to assess the lead and allergen
16 burden in that portion of the regularly-occupied U.S. housing stock that can potentially include
17 young children among its residents. In particular, the survey is assessing lead burden by
18 characterizing levels of lead-contaminated dust, lead-based paint, and lead-contaminated soil in
19 housing and residential areas. HUD initiated this survey in 1997 and has been approved by the
20 Office of Management and Budget to collect information through April 2001 for up to 1000
21 housing units.
22 The NSLAH provides a more recent nationally-representative characterization of
23 environmental-lead levels in the U.S. housing stock than the 1989-1990 HUD National Survey
24 and involves sampling in considerably more housing units. In addition, dust samples in the
25 NSLAH are collected using wipe techniques (i.e., the technique assumed in the §403 rule) rather
26 than the Blue Nozzle vacuum method used in the older survey, and the NSLAH did not restrict
27 the sampling frame to only housing built prior to 1980. Therefore, the information collected in
28 the NSLAH is very important for the §403 risk analysis to consider. However, the survey's
29 scheduled completion date and the expected date for finalizing the survey's database do not fall
30 within the time frame necessary to complete the risk analysis. Therefore, in order to utilize data
31 from the NSLAH, the risk analysis could only consider data collected up to an interim point in
32 the survey.
33 Interim NSLAH data for 706 housing units, collected from 1998-1999, were made
34 available to the §403 risk analysis in August, 1999. This is a preliminary subset of the survey's
35 final database that will represent an expected 825 housing units. To allow the data for these 706
DRAFT -- DO NOT CITE OR QUOTE 44 August 28,2000
-------
1 units to be considered a nationally-representative characterization of lead levels in the housing
2 stock, the interim database included sampling weights assigned to each unit based on its set of
3 selection probabilities within each stage of the multi-staged sampling design and adjusted for
4 nonresponse. These are interim sampling weights as they were generated by only considering the
5 706 units represented in the interim database. As the final sampling weights to be assigned at the
6 end of the survey will reflect all housing units in the survey, and as there is a potential for
7 additional correction of the existing data before the survey database is finalized, any analysis
8 results based on the interim database of 706 housing units will likely differ from those to be
9 based on the final database.
10 Table 3-1 contains key design specifications and approaches of the NSLAH, such as the
11 types of rooms in which dust samples were collected and paint-lead levels were measured, the
12 approach to taking soil samples, and laboratory analytical methods. Also included for
13 comparison purposes in Table 3-1 are the design specifications and approaches taken in the older
14 HUD National Survey. Note that in both surveys, dust samples were taken from the same types
15 of surfaces (floors, window sills, and window troughs, also known as window wells) and
16 analyzed under similar methods, and soil sampling occurred in the same areas of the yard. The
17 method for analyzing soil samples was changed from ICP-AES in the older survey to FAA in the
18 NSLAH due to the need to reduce detection limits associated with the method. Specific focus
19 was made in the NSLAH to ensure that rooms in which children frequently reside are more
20 dominantly represented in the sampling design.
21 Various types of data are being collected from housing units participating in the NSLAH.
22 Household questionnaire data are collected at two time points: at the initial contact with the
23 household during recruitment (to screen for eligibility and to perform an inventory on interior
24 rooms) and during an interview with residents during the study (to obtain information on the
25 building, household, and residents). Allergen dust levels are measured by collecting and
26 analyzing vacuum dust samples. Lead levels in the unit are characterized through the following
27 types of measures:
28 • Dust samples: Dust-lead loadings (jig/ft2, assuming wipe collection techniques)
29 for floors, window sills, and window troughs (also known as window wells)
30 • Soil samples: Soil-lead concentrations (pg/g) at entry way, dripline, and mid-yard
31 • Lead on painted surfaces: X-rav fluorescence (XRF) measurements (mg/cm2)
32 To determine the numbers of housing units represented by the interim NSLAH sampling
33 weights within certain housing categories and how these numbers compare with estimates made
34 in the §403 risk analysis and by the U.S. Census Bureau, Tables 3-2 and 3-3 provide estimated
35 numbers of occupied housing within specified housing age categories and the four Census
36 regions, respectively. These totals are presented based on data from the NSLAH as well as from
37 the following additional surveys/analyses:
DRAFT - DO NOT CITE OR QUOTE 45 August 28,2000
-------
1
2
8
Table 3-1. Differences in Approaches and Outcomes Between the HUD National Survey of Lead-Based Paint in Housing
and the HUD National Survey of Lead and Allergens in Housing
3
4 m
c O
6 o
7 c
3
10
11
12
13
14
o>
Area
HUD National Survey of Lead-Based Paint in Housing1
HUD National Survey of Lead and Allergens in Housing1
Types/numbers of
housing units selected
for the survey and
whose data were
available to the §403
risk analysis
284 housing units selected from privately-owned, year-
round occupied housing in the 48 conterminous states
built prior to 1980 and having the potential for containing
children. Institutional and group (i.e., housing units with at
least 10 unrelated persons) housing were excluded from
consideration for the survey.
Interim data for 706 housing units selected from year-round
occupied housing in the 50 states and the District of Columbia
having the potential for containing children were provided to
EPA on August 13, 1999 (out of an expected 825 housing
units in the survey). The sample represents 67 of the planned
75 primary sampling units (PSUs). Institutional and group (i.e.,
housing units with at least 10 unrelated persons) housing were
excluded from consideration for the survey.
Breakdown of selected
units by year built
Pre-1940: 27%
1940-1959: 31%
1960-1979:42%
Post-1979:0%
Pre-1940: 18%
1940-1959: 23%
1960-1977:31%
Post-1977: 28%
(Percentages are relative to the 640 units with housing age
information from either the recruitment or resident
questionnaire.)
Dates of
environmental
sampling
November 1 989 to March 1 990
August 1998 to February 1999 (according to dates specified in
the survey's interim database - sampling in a small number of
units may have occurred earlier in 1998)
(O
ro
oo
-------
Table 3-1. (cont.)
o
o
o
o
3J
o
Area
HUD National Survey of Lead-Based Paint in Housing1
HUD National Survey of Lead and Allergens in Housing1
Selecting rooms for
environmental
sampling
Telephone household interview provided information on
rooms. One room was selected for sampling in each of the
following strata:
• Wet room -- rooms containing plumbing (e.g., kitchen,
bathroom, laundry room, utility room)
• Dry room -- all rooms not classified as wet rooms
• Main entrvwav (floor dust samples only)
Room Inventory Form from the Screening/Recruiting
Questionnaire was used to obtain information on rooms.
room was randomly selected for sampling in each of the
following four strata:
One
• Kitchen
• Common living area (e.g., living room, den, family room)
• Bedroom in which one or more children aged 17 years or
younger regularly slept, or any regularly-occupied bedroom
if no such children lived in the unit (occasionally, two such
bedrooms were selected)
• Other random room among the remaining rooms in the
housing unit. (Note: Two rooms were randomly selected
from this stratum if the stratum contained at least six
rooms. Adult bedrooms were included if a child's bedroom
was available for selection in the bedroom stratum.)
In addition to the selected rooms, floor dust samples from the
main entryway were collected.
Method to assigning
sampling weights
Weights reflect the various stages of sampling and were
designed to sum to the approximately 77 million pre-1980
homes then in the occupied housing stock. The weights
were stratified to control for the number of housing units
with children (13.9 million) and without children. Total of
the sampling weights within a given census region equaled
the estimated number of units with children under age 7
years in the census region.
Interim weights reflect the various stages of sampling and were
designed to sum to the estimated 89 million housing units in
the occupied housing stock that do not exclude children.
Method for taking
dust samples for lead
analysis
Blue Nozzle vacuum (a few wipe samples were also
collected)
Wipes, collected in accordance with ASTM E1728-95, Practice
for the field determination of settled dust samples using wipe
sampling methods for lead determination by atomic absorption
spectrometry techniques.
»
Number and location
of floor-dust samples
per room
One sample from each selected room (location not dictated
in the protocol)
One sample from each selected room, generally taken from the
largest open area.
-------
Table 3-1. (cont.)
. I
\ I
m
O
4 S
\°
Area
HUD National Survey of Lead-Based Paint in Housing1
HUD National Survey of Lead and Allergens in Housing1
Window sill/trough
dust sampling
approach
A window was selected within each selected room
according to some ranking scheme. Sampling was
performed from both the sill and trough of the selected
window until "enough" dust was collected or until the
entire sill or trough was vacuumed.
Entire sill and trough sampled from a random window in the
selected room.
m
Number and location
of sill and trough dust
samples per room
One sample from each of the sill and trough of the
selected window in the selected wet room and dry room
One sample from the sill and one sample from the trough of
the selected window in each selected room
7
8
9
10
Method of analyzing
dust samples
GFAA (with SW-846 digestion method)
FAA (Digestion method: modification of SW-846 Method 3050
or ASTM ES 36-94 - hot-plate digestions utilizing
nitric/perchloric acid and H2O2)
Method must be that used in proficiency testing within the
Environmental Lead Laboratory Accreditation Program (ELLAP)
Soil sampling
approach
One composite sample of 3 core samples (the latter two
taken within 20 inches of the first), each taken at a depth
of 10 cm, was collected at each of the following locations:
entrywav. drip-line, and remote area (i.e., an area halfway
between the unit and its property boundary, or within 25
feet of the unit).
a
ro
CD
8
o
o
Samples were collected from bare soil when possible. If no
bare soil existed, soil samples were collected from covered
surfaces if possible. Two sides of the unit were selected for
soil sampling: the side containing the major entryway (Wall 1)
and a second, randomly-selected side (Wall 2). Samples were
collected from the top 0.5 inches of soil at the following three
locations:
• Main entry - a single sample from Wall 1
• Foundation/drip-line - one sample from each of Walls 1
and 2, each sample being a composite of 3 core samples
taken within 3 feet of the foundation
• Mid-yard area - one sample from each of Walls 1 and 2,
each sample being a composite of up to 4 core samples
taken midway between the drip-line and the closer of the
boundary line or another building on the property.
Soil samples were collected in accordance with core sampling
procedures based on ASTM E1727-95 (described in the HUD
Guidelines and in EPA's Residential Sampling for Lead:
Protocols for Leaded Dust and Soil Sampling)
-------
Table 3-1. (cont.)
D
O
1
O
m
O
33
i
Area
HUD National Survey of Lead-Based Paint in Housing1
HUD National Survey of Lead and Allergens in Housing1
1
2
Method of analyzing
soil samples
ICP-AES (with SW-846 digestion method)
ICP-AES
Digestion method: modification of SW-846 Method 3050 or
ASTM ES 36-94 (hot-plate digestions utilizing nitric acid and/or
HCI/H202). or SW-846 Method 3051 (microwave nitric acid
digestion)
Method must be that used in proficiency testing within the
Environmental Lead Laboratory Accreditation Program (ELLAP)
3
4
5
6
7
8
9
10
11
12
13
Handling dust-lead and
soil-lead
measurements below
the detection limit
As log-transformed lead amounts are reported in the
database, only positive measurements are represented.
No indication is given as to when data may have been
truncated due to being below detection limits.
The final results as reported by the instrument are recorded in
the database (i.e., not-detected results are not censored), along
with detection limits.
Method for taking
paint-lead
measurements
MAP-3 XRF instrument (single 60-second "spectrum
reading" measurement using a 40 millicune cobalt source).
Measurements were adjusted to statistically correct for
measurement bias.
XRF (Niton XL-309 running software version 5.1)
Approach to selecting
interior painted
components for paint-
lead measurements
Painted surfaces were categorized into the following four
strata:
• Walls/ceilings/floors
• Metal substrate
• Non-metal substrate
• Other surfaces
Five painted components were selected randomly for
testing in each of the selected wet and dry rooms, one
from each stratum along with a fifth selected randomly
from among all strata. In addition, up to two purposive
measurements were taken from paint anywhere in the unit
that may be suspected to contain lead.
A list of 25 possible interior components was developed and
included:
All four major walls
Ceiling
Floor
Window system components
Doors and doorways
Trim
Porches
All components present in a given room were tested.
ro
oo
-------
Table 3-1. (cont.)
o
o
o
a
o
3J
O
I
m
Area
HUD National Survey of Lead-Based Paint in Housing1
HUD National Survey of Lead and Allergens in Housing1
1
2
3
4
Approach to selecting
exterior painted
components for paint-
lead measurements
Painted surfaces were categorized into the following four
strata:
• Wall (randomly-selected)
• Metal substrate within the selected wall
• Non-metal substrate within the selected wall
• Other surfaces within the selected wall
Five painted components were selected randomly for
testing from the side of the unit containing the selected
wall, one from each stratum along with a fifth selected
randomly from among all strata. In addition, up to two
purposive measurements were taken from paint anywhere
on the exterior of the unit that may be suspected to
contain lead.
A list of 25 possible exterior components was developed and
included:
Siding
Window system components
Doors and doorways
Trim
Porches
All components present on the sampled wall were tested.
01
o
Information reflects only that part of the survey whose data and information were used in the §403 risk analysis.
(Q
a
ro
CD
8
§
-------
1
2
3
4
5
Table 3-2. Estimated Number of Occupied Housing Units in the U.S. Housing Stock
Within Year-Built Categories, According to Four Recent Surveys and/or
Analyses
Year in Which the
Unit Was Built
Prior to 1 940
1940-1959
1960-1979
( 1960-1 977 f or NSLAH)
After 1979
(After 1977 for NSLAH)
Not specified5
Total
Number of Units in the National Housing Stock (and Percentage
of Total Units), as Estimated by the ...
1995
American
Housing
Survey1
19,308,000
(20%)
19,885,000
(20%)
35,300,000
(36%)
23,201,000
(24%)
~
97,693,000
§403 Risk
Analysis2
19,676,000
(20%)
19,718,000
(20%)
34,985,000
(35%)
24,893,000
(25%)
-
99,272,000
1997
American
Housing
Survey3
19,441,000
(20%)
19,797,000
(20%)
34,884,000
(35%)
25,367,000
(25%)
--
99,487,000
National
Survey of
Lead and
Allergens in
Housing
(NSLAH)4
14,412,000
(18%)
16,886,000
(21%)
25,688,000
(32%)
24,076,000
(30%)
8,089,000
89,151,000
# Units Surveyed
1989-90
HUD
National
Survey
77
87
120
--
-
284
NSLAH
(interim)
114
145
201
180
66
706
8
9
10
11
12
13
14 ' Estimates represent only year-round occupied housing in the 1995 national housing stock and were obtained from
15 information within Table 1A-1 of 'American Housing Survey for the United States in 1995" (Current Housing Reports
16 H150/95RV, published by the Bureau of the Census and HDD's Office of Policy Development and Research). This national
17 survey was conducted on about 55,000 surveyed units from August 1995 through February 1996. An updated report
18 reflecting the 1997 American Housing Survey data has not yet been published.
19 * Estimates were obtained from Table 3-5 of the §403 risk analysis report Estimates are based on data from the 1989-90
20 HUD National Survey of Lead-Based Paint in Housing, augmented by other Census information in order to represent the 1997
21 housing stock (see the §403 risk analysis report for details).
22 3 Estimates were obtained from 1997 American Housing Survey data summaries prepared by HUD (in particular. Table 2-1).
23 These HUD data summaries are available at http //www.huduser.org/datasets/ahs/ahsdata97.html (It does not appear that
24 these summaries have been publically released outside of this Internet location.) The estimates represent total occupied
25 units in the 1997 national housing stock (which may also include occupied seasonal housing) and were obtained from data
26 that appear to have been corrected by HUD for possible errors.
27 * This survey, conducted from 1998-1999, characterized only occupied housing in which a young child could reside.
28 Information in this table is based on an interim dataset for 706 surveyed housing units. Year-built information was
29 determined from responses given in the survey's resident questionnaire. If no year-built information was available from the
30 resident questionnaire, any year-built information provided from the recruitment questionnaire (when available) was used.
31 Note differences in how year-built categories were defined in this survey. The specified percentages are relative to the total
32 minus the number of units represented by surveyed units with no year-built information specified (i.e., 89,151,000 -
33 8,089,000 - 81,062,000).
34 s Total sampling weights for surveyed units where either no housing age information was provided, or responses of "Don't
35 Know* or "Not Ascertained" were given, on both the resident and recruitment surveys.
DRAFT - DO NOT CITE OR QUOTE
51
August 28, 2000
-------
1
2
3
Table 3-3. Estimated Number of Occupied Housing Units in the U.S. Housing Stock
Within Each Census Region, According to Four Recent Surveys and/or
Analyses
Census Region
Northeast
Midwest
South
West
Total
Number of Units in the National Housing Stock (and Percentage
of Total Units), as Estimated by the ...
1995
American
Housing
Survey1
19,200,000
(20%)
23,662,000
(24%)
34,236,000
(35%)
20,596,000
(21%)
97,693,000
§403 Risk
Analysis2
1 5,878,000
(16%)
22,313,000
(22%)
41,733,000
(42%)
1 9,348,000
(19%)
99,272,000
1997
American
Housing
Survey3
19,484,000
(20%)
23,951,000
(24%)
34,808,000
(35%)
21,245,000
(21%)
99,487,000
National
Survey of
Lead and
Allergens in
Housing
(NSLAH)4
14,977,000
(17%)
22,202,000
(25%)
32,519,000
(36%)
19,453,000
(22%)
89,151,000
# Units Surveyed
1989-90
HUD
National
Survey
53
69
116
46
284
NSLAH
(interim)
109
150
265
182
706
5
6
7
8
9
10 ' Estimates represent only year-round occupied housing in the 1995 national housing stock and were obtained from
11 information within Table 1A-1 of "American Housing Survey for the United States in 1995" (Current Housing Reports
12 H150/95RV, published by the Bureau of the Census and HUD's Office of Policy Development and Research). This national
13 survey was conducted on about 55,000 surveyed units from August 1995 through February 1996. An updated report
14 reflecting the 1997 American Housing Survey data has not yet been published.
15 2 Estimates were obtained from Table 3-5 of the §403 risk analysis report. Estimates are based on data from the 1989-90
16 HUD National Survey of Lead-Based Paint in Housing, augmented by other Census information in order to represent the 1997
17 housing stock (see the §403 risk analysis report for details).
18 3 Estimates were obtained from 1997 American Housing Survey data summaries prepared by HUD (in particular. Table 2-1).
19 These HUD data summaries are available at http //www.huduser.org/datasets/ahs/ahsdata97 html (It does not appear that
20 these summaries have been publically released outside of this Internet location.) The estimates represent total occupied
21 units in the 1997 national housing stock (which may also include occupied seasonal housing) and were obtained from data
22 that appear to have been corrected by HUD for possible errors.
23 * This survey, conducted from 1998-1999, characterized only occupied housing in which a young child could reside.
24 Information in this table is based on an interim dataset for 706 surveyed housing units.
DRAFT -- DO NOT CITE OR QUOTE
52
August 28, 2000
-------
1 • the 1995 American Housing Survey (i.e., the last survey in which estimates of
2 these totals were published in documents issued by the Census Bureau)
3 • the §403 risk analysis (which characterized the 1997 housing stock by revising
4 the sampling weights from the 1989-90 HUD National Survey)
5 • the 1997 American Housing Survey (based on information obtained from the
6 HUD web-site)
7 As noted in these tables, the sum of the interim sampling weights in the NSLAH (89,151,000) is
8 over ten million units lower than the corresponding sums from the §403 risk analysis and the
9 1997 American Housing Survey. It is possible that this difference is due to the NSLAH's
10 exclusion of housing that forbids resident children (i.e., adult-only housing), while the §403 risk
11 analysis and the 1997 American Housing Survey results reflect the entire regularly-occupied
12 housing stock.
13 When reviewing the sum of sampling weights by housing age category (Table 3-2), the
14 interim NSLAH data represent a slightly smaller percentage of pre-1940 housing compared to the
15 other surveys. The housing age categories for the NSLAH differ slightly from the categories in
16 which the other surveys are represented and what was used in the 1989-90 HUD National
17 Survey. Approximately eight million housing units in the U.S. housing stock are represented by
18 66 units in the interim NSLAH dataset that do not have a housing age specified. The final two
19 columns of Table 3-2 present numbers of surveyed units by housing age category in both HUD
20 surveys.
21 The percentages of housing units within Census regions (Table 3-3) are similar between
22 the interim NSLAH and the 1997 American Housing Survey except for the Northeast, where the
23 percentage in the interim NSLAH was lower than in the 1997 American Housing Survey.
24 Differences relative to the 1997 American Housing Survey were even greater for the §403 risk
25 analysis, where the adjustments made to the sampling weights in the 1989-90 HUD National
26 Survey to represent the 1997 housing stock did not take into account Census region.
27 Summaries of the interim NSLAH data and comparison to the 1989-90 HUD National
28 Survey data summaries (cited in the §403 risk analysis report) are provided in the next section.
29 3.2 COMPARISON OF ENVIRONMENTAL-LEAD LEVELS IN THE HUD NATIONAL
30 SURVEY WITH THOSE OF OTHER KEY STUDIES
31 As discussed in Sections 3.2 and 3.3 of the §403 risk analysis report, the risk analysis
32 used data from the HUD National Survey to represent baseline (pre-§403) environmental-lead
33 levels (paint, dust, soil) in the nation's housing stock. To help evaluate how accurate this
34 representation may be and how environmental-lead levels may have changed since the HUD
35 National Survey was conducted (1989-1990), the survey data were compared with data from
36 other environmental field studies that were conducted more recently and that measured
DRAFT -- DO NOT CITE OR QUOTE 53 August 28,2000
-------
1 environmental-lead levels in a large number of housing units. This section also summarizes how
2 housing selection, sample collection techniques, laboratory testing practices, and the distribution
3 of environmental-lead levels reported in the HUD National Survey differ from those in these
4 other studies.
5 The studies whose dust-lead and soil-lead data were used in the comparisons in this
6 section included the ongoing National Survey of Lead and Allergens in Housing (NSLAH,
7 introduced in Section 3.1 above), the Baltimore Repair & Maintenance (R&M) Study, the
8 Rochester Lead-in-Dust Study, and the various portions of the ongoing HUD Grantees evaluation
9 (design information and data for the latter three studies were summarized in Section 3.2.2 of the
10 §403 risk analysis report). These studies were conducted since 1993 in locations within the
11 United States where a specific point source of lead was not necessarily present. The latter three
12 studies provided the §403 risk analysis with the most useful and available data on the
13 relationship between environmental-lead levels (paint, dust, and soil) and childhood blood-lead
14 concentration. In particular, dust samples in these studies were collected from floors and
IS window sills using either a wipe technique or a method whose resulting dust-lead loadings could
16 be converted to wipe-equivalent loadings using methods developed in the §403 risk analysis.
17 Data summaries for the HUD Grantees evaluation were updated from the §403 risk analysis
18 report summaries to reflect data collected through February, 1999.
19 The risk associated with elevated soil-lead concentrations and intervention practices
20 designed to alleviate that risk are more frequently debated in the scientific literature than are the
21 risk from and the intervention practices targeting elevated dust-lead loadings. As a result, this
22 section supplements the comparison of the HUD National Survey's characterization of soil-lead
23 concentrations to the interim NSLAH and the aforementioned three recent studies with the results
24 of other relevant studies.
25 Boxplots were used in this section to summarize household average dust-lead and soil-
26 lead levels graphically. A boxplot, also known as a box-whisker plot, portrays the distribution
27 visually by using a box to represent data falling within the 25th and 75th percentiles and using
28 different graphical symbols for the remaining data values according to their distance from the
29 box. The following features are included within the boxplots presented in this section:
30 • A horizontal line within the box corresponds to the median.
31 • A dot within the box corresponds to the geometric mean.
32 • The bottom and top edges of the box correspond to the 25th and 75th percentiles,
33 respectively.
34 • Central vertical lines ("whiskers") extend to 1.5 interquartile ranges (IQR, equal
35 to the difference between the 75* and 25* percentiles on a log scale) of the box.
36 However, if the data extend to less than 1.5 IQRs of the box, the whiskers extend
37 only as far as the data exist.
38 • Open circles represent data values that exceed 1.5 IQRs but no more than 3 IQRs
39 from the box.
40 • Asterisks represent data values that exceed 3.0 IQRs from the box.
DRAFT -- DO NOT CITE OR QUOTE 54 August 28,2000
-------
1 The boxplots were plotted on a logarithmic scale to improve the readability of the data
2 distributions, due to the tendency of the data to be skewed toward the lower end of these
3 distributions. Selected information portrayed within the boxplots have also been included within
4 tables of descriptive statistics presented throughout this section.
5 Dust-lead loading data comparisons are provided in Section 3.2.1, while soil-lead
6 concentration data are addressed in Section 3.2.2.
7 3.2.1 Characterizing Dust-Lead Loadings on Floors and Window Sills
8 Household area-weighted average dust-lead loadings (assuming wipe techniques) as
9 calculated in the §403 risk analysis were the basis for the comparisons made in this section. This
10 average, calculated for each building component sampled for dust (i.e., floor, window sill),
11 represented a single dust-lead measure for the component within a housing unit and was
12 calculated by weighting each dust sample's result by the area that was sampled.
13 While the household average dust-lead loadings assumed wipe collection techniques, the
14 dust collection device differed among the studies:
15 • HUD National Survey: Blue Nozzle vacuum
16 • Baltimore R&M study: BRM vacuum
17 • NSLAH. Rochester study, and the HUD Grantees evaluation: wipes.
18 To obtain wipe-equivalent dust-lead loadings for samples taken in the HUD National Survey and
19 the Baltimore R&M study, the reported loadings were entered into the conversion equations
20 presented in Sections 4.3.1 (Blue Nozzle vacuum to wipe) and 4.3.2 (BRM vacuum to wipe) of
21 the §403 risk analysis report. Note that dust-lead loadings for samples collected by other
22 collection methods in these studies were not included in determining the area-weighted
23 averages.7
24 In the §403 risk analysis, the household averages were calculated on wipe-equivalent
25 sample loadings associated with the 284 units in the HUD National Survey, with imputed
26 averages assigned to those units having no available data (Section 3.3.1.1 of the §403 risk
27 analysis report). When characterizing the distribution of these averages across units, the §403
28 risk analysis weighted each unit by its 1997 sample weight as calculated for the §403 risk
29 analysis (Appendix Cl of the §403 risk analysis report), and each unit built between 1960 and
30 1979 and without lead-based paint also represented post-1979 housing (Section 3.3.1.5 of the
31 §403 risk analysis report). The resulting data distribution was used in the §403 risk analysis to
32 characterize the distribution of average dust-lead loadings in the nation's housing stock.
7 The HUD National Survey database included a few wipe dust-lead loadings that were used as reported in
determining household area-weighted averages.
DRAFT - DO NOT CITE OR QUOTE 55 August 28,2000
-------
1 3.2.1.1. Data Summaries for the §403 Risk Analysis Versus the Interim NSLAH.
2 Descriptive statistics of household average dust-lead loadings for floors and window sills as
3 calculated in the §403 risk analysis using the HUD National Survey data are presented in this
4 subsection as they compare with the same statistics calculated on interim data for 706 housing
5 units in the NSLAH. Note that these statistics reflect the sampling weights used in the §403 risk
6 analysis and the interim NSLAH sample weights, thereby allowing these summaries to be
7 nationally representative of the 1997 housing stock.
8 The interim NSLAH summaries include imputed average dust-lead loading data values
9 which are assigned to households when no such data are available for a given surface (floors,
10 window sills). Assigning an imputed dust-lead loading average to a household that has no dust-
11 lead loading data ensures that it (and its corresponding sampling weight representing a given
12 portion of the national housing stock) is represented in the risk analysis. The method used to
13 impute data closely follows the method used in the §403 risk analysis for housing units in the
14 HUD National Survey; this method is detailed in Appendix C. This appendix also gives the
15 imputed data values and how they were assigned to housing units. Summaries of the interim
16 NSLAH dust-lead loading data with imputed data excluded are found in Appendix D1.
17 When using the interim NSLAH data to calculate a household's average dust-lead loading
18 for floors or window sills, five different approaches were considered for handling individual
19 sample results that fell below the instrument's detection limit. These five approaches, which
20 include censoring the not-detected results, are presented in Appendix Dl. The data summaries
21 that exclude imputed data values, found in Appendix D1, were performed and presented for each
22 of these five approaches. Of these five approaches, two were specifically identified as most
23 likely to be applied in the supplemental risk analysis involving the interim NSLAH data:
24 • making no adjustment to not-detected data values, and
25 • replacing not-detected data values with one-half of the detection limit.
26 The first approach eliminates potential bias that can be introduced when an adjustment is made to
27 a reported data value, but it also permits a household's average to be zero or below, preventing
28 the data from being used as input to the empirical model within the §403 risk analysis. (As the
29 survey's analytical method adjusted for potential analytical bias by subtracting a specified
30 amount from a given sample result, reported results of less than zero were possible. Such results
31 were included in the survey database used in this analysis.) The second approach prevents this
32 problem from occurring and represents the best estimate of a sample's actual lead amount value
33 when the analytical result is only known to fall somewhere between zero and the instrument's
34 detection limit. Interim NSLAH data summaries under both approaches are presented in this
35 section to illustrate the impact that any one approach has on the characterized distribution.
36 National comparisons
37 Tables 3-4 and 3-5 present descriptive statistics of average household dust-lead loadings
38 for floors and window sills, respectively, for the 1997 national housing stock. These summaries
DRAFT - DO NOT CITE OR QUOTE 56 August 28,2000
-------
1
2
3
Table 3-4. Descriptive Statistics of Area-Weighted Average Floor Wipe Dust-Lead
Loadings for Households, As Reported in the §403 Risk Analysis Versus the
Interim NSLAH Data
Study
How Not-
Detected
and
Negative
Data were
Handled
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced
by LOD/2
Area-Weighted Average Floor Dust-Lead Loading (/ig/ft2)1
#
Surveyed
Units with
Positive
Averages
284
633
706
Arith-
metic
Mean
16.5
10.4
10.8
Geo-
metric
Mean2
6.27
1.22
1.82
Geo-
metric
Std.
Dev.2
3.49
4.57
2.78
Minimum
0.508
-1.23
0.750
25*
Percen-
tile
2.65
0.300
0.950
Median
5.32
1.05
1.31
75"1
Percen-
tile
12.2
2.30
2.46
Maximum
375
5940
5950
5
6
7
8
9
10
11
12
13
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with floor dust-lead data are
used to calculate the remaining statistics).
3 Summaries include imputed data for households having no floor wipe dust-lead loading data. The method for imputation is
presented in Appendix C.
14
15
16
Table 3-5. Descriptive Statistics of Area-Weighted Average Window Sill Wipe Dust-
Lead Loadings for Households, As Reported in the §403 Risk Analysis
Versus the Interim NSLAH Data
Study
How Not-
Detected
and
Negative
Data were
Handled
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced
by LOD/2
Area-Weighted Average Window Sill Dust-Lead Loading (fig/ft2)1
#
Surveyed
Units with
Positive
Averages
284
690
706
Arith-
metic
Mean
550
137
137
Geo-
metric
Mean2
23.0
14.5
15.8
Geo-
metric
Std.
Dev.2
15.8
7.83
6.57
Minimum
0.0118
-9.43
0.445
25*
PercBn-
tile
4.35
2.90
3.35
Median
19.5
12.8
13.6
75*
Percen-
tile
198
51.3
51.0
Maximum
43700
11100
11100
17
18
19
20
21
22
23
24
25
26
1 All statistics are i
• All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with window sill dust-lead
data are used to calculate the remaining statistics).
3 Summaries include imputed data f<
imputation is presented in Appendix
' Only household averages greater than zero are usi
data are used to calculate the remaining statistics).
3 Summaries include imputed data for households having no window sill wipe dust-lead loading data The method for
imputation is presented in Appendix C.
DRAFT - DO NOT CITE OR QUOTE
57
August 28, 2000
-------
1 imply that the average dust-lead loadings for both floors and window sills based on the interim
2 NSLAH data are considerably lower than that reported in the §403 risk analysis (based on the
3 HUD National Survey after converting to wipe-equivalent loadings). For example, the median
4 floor dust-lead loading is less than 2 ug/ft2 based on the interim NSLAH data compared to 5.3
5 MS/ft2 from the §403 risk analysis, and the median window sill dust-lead loading is less than 12
6 Mg/ft2 based on the interim NSLAH data compared to nearly 20 ug/ft2 from the §403 risk
7 analysis.
8 Median detection limits for dust-lead loadings in the interim NSLAH were 1.5 ug/ft2 for
9 floors and 3.6 ug/ft2 for window sills. When considering all dust samples in the interim NSLAH
10 that had lead amounts reported, approximately two-thirds of the floor dust-lead samples and one-
11 third of the window sill dust-lead samples had results below the detection limit.
12 Boxplots of the data distributions presented in Tables 3-4 and 3-5 are found in Figures
13 3-1 and 3-2, respectively. Appendix Dl contains these tabular summaries and boxplots after
14 excluding imputed data values.
15 In addition to these data summaries that are based solely on the observed data and the
16 sampling weights, it was desired to characterize the national distribution of household average
17 floor dust-lead loading in such a way that the percentage of housing where this average exceeds a
18 specified threshold could be estimated. This was done for both the HUD National Survey and
19 interim NSLAH data by assuming that these data originate from a lognormal distribution. Then,
20 the fitted distributions and corresponding estimated exceedance percentages were compared
21 between the two surveys. These results are presented in Section 3.2.1.3 below.
22
23 Comparisons by housing age category
24 While the summaries in Tables 3-4 and 3-5 represent the entire nation, Tables 3-6 and 3-7
25 present descriptive statistics according to the housing age category scheme defined in Table 3-2
26 above. Considerable declines in the geometric means and medians from the §403 risk analysis to
27 the interim NSLAH data were observed in all four age categories.
28 Boxplots of the data distributions presented in Tables 3-6 and 3-7 are found in Figures
29 3-3 and 3-4, respectively. Appendix Dl contains these tabular summaries and boxplots after
30 excluding imputed data values.
31 Comparisons bv Census region
32 Tables 3-8 and 3-9 present descriptive statistics according to Census region. Declines in
33 the geometric means and medians were observed from the §403 risk analysis to the interim
34 NSLAH data for all regions but the West region, where very slight increases in these estimates
35 were observed. The greatest declines were observed in the Northeast and Midwest.
DRAFT - DO NOT CITE OR QUOTE 58 August 28,2000
-------
o
o
o
o
m
O
3J
O
c
o
m
en
CO
8
cr
HUONS
(••OS)
NSLAH
(LOO/Z)
(Q
ro
oo
Figure 3-1.
Boxplots of Area-Weighted Average Floor Wipe Dust-Lead Loadings (ug/ft2) As Observed in the §403 Risk
Analysis (Using HUD National Survey Data) and in the Interim NSLAH (under 2 approaches to handling not-
detected values)
g (Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using the methods documented in the
§ §403 risk analysis report. Boxplots include imputed household averages but not negative averages. See text for definitions of labels along the horizontal axis.)
-------
O
o
O
m
O
30
O
§
m
en
o
I
I
HUONJ
(403)
NSLAH
(LOO/2)
(Q
C
cn
po
Figure 3-2.
Boxptots of Area-Weighted Average Window Sill Wipe Dust-Lead Loadings (ug/ft2) As Observed in the §403
Risk Analysis (Using HUD National Survey Data) and in the Interim NSLAH (under 2 approaches to handling
not-detected values)
^ (Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using the methods documented in the
§ §403 risk analysis report. Boxplots include imputed household averages but not negative averages. See text for definitions of labels along the horizontal axis.)
-------
1 Table 3-6. Descriptive Statistics of Area-Weighted Average Floor Wipe Dust-Lead
2 Loadings for Households, Presented bv Housing Aae Category. As Reported
3 in the §403 Risk Analysis Versus the Interim NSLAH Data
Study
How Not-
Detected
and
Negative
Data were
Handled
Area-Weighted Average Floor Dust-Lead Loading (/fg/ft2)1
#
Surveyed
Units with
Positive
Averages
Arith-
metic
Mean
Gee-
rriBtric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
25th
Percen-
tile
Median
75"1
Percen-
tile
Maximum
Units Built Prior to 1940
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
77
111
114
47.9
36.9
37.0
22.6
3.74
4.00
3.63
4.53
3.97
0.991
-0.600
0.750
8.84
1 30
1.45
17.7
2.42
2.71
79.7
9.50
9.50
375
5940
5950
Units Built from 1940 - 1959
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
87
134
145
18.1
4.11
4.38
8.74
1.90
2.31
3.34
3.57
2.64
0.508
-0.720
0750
407
0.719
1.05
7.81
1.80
1.99
224
4.00
4.00
171
71.0
71.0
Units Built from 1960-1977 (1960 - 1979 for the §403 risk analysis)
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
120
176
201
6.74
1.50
1.96
4.14
0.912
1.46
2.45
3.47
1.92
0.657
-0733
0750
2.25
0.236
0.900
3.62
0900
1.20
7.59
1.68
1 92
106
28.5
288
Units Built After 1977 (after 1979 for the §403 risk analysis)
§403 Risk Analysis
(HUD Natl Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
28
151
180
4.16
1.20
1.71
3 14
0.545
1.14
2.06
3.35
1 72
1.06
-1.05
0750
1.76
0.146
0.750
2.84
0.400
1.00
566
1.08
1.35
12.9
265
265
NSLAH Units with Unspecified Year-Built Indicator
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
61
66
31 7
32.1
1.37
220
6.64
392
-1.23
0.750
0300
1.00
1.24
1.40
2.72
2.56
1040
1040
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with floor dust-lead data are
used to calculate the remaining statistics).
3 Summaries include imputed data for households having no floor wipe dust-lead loading data. The method for imputation is
presented in Appendix C.
DRAFT - DO NOT CITE OR QUOTE
61
August 28, 2000
-------
1 Table 3-7. Descriptive Statistics of Area-Weighted Average Window Sill Wipe Dust-
2 Lead Loadings for Households, Presented by Housing Age Category. As
3 Reported in the §403 Risk Analysis Versus the Interim NSLAH Data
Study
How Not-
DotGCtGO
and
Negative
Data were
Handled
Area-Weighted Average Window Sill Dust-Lead Loading Urg/ft1)1
#
Surveyed
Units with
Positive
Averages
Arith-
metic
Mean
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
25"1
Percen-
tile
Median
75*
Percen-
tile
Maximum
Units Built Prior to 1940
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
77
113
114
2060
400
400
168
77.5
76.8
16.7
6.59
644
0.0155
-0.152
1.03
35.6
21 2
21.2
198
79.8
79.8
1220
294
294
43700
11100
11100
Units Built from 1940 - 1959
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
87
144
145
285
129
129
22.0
245
26.1
107
6.80
5.97
00118
-1.73
0.923
6.47
635
658
19 1
23.0
22.0
107
886
88.6
16100
3630
3630
Units Built from 1960-1977 (1960 - 1979 for the §403 risk analysis)
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
120
195
201
Units
28
174
180
184
36.6
36.9
Built After
83.0
15.6
16.0
16.2
10.7
11.3
14.6
4.71
4 18
0.0164
-232
1.02
2.05
2.89
3.17
16.6
9.40
9.54
217
290
293
5790
1390
1390
1977 (after 1979 for the §403 risk analysis)
8.17
3.56
4.57
9.94
527
3.79
0.0164
-943
0.445
2.58
0916
1.72
8.11
3.19
3.67
578
10.3
999
1590
426
427
NSLAH Units with Unspecified Year-Built Indicator
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
64
66
367
367
39.8
40.2
7.32
6.72
-0.629
0.720
18.6
18.8
36.4
36.4
118
118
9030
9030
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
1 AM statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with window sill dust-lead
data are used to calculate the remaining statistics).
3 Summaries include imputed data for households having no window sill wipe dust-lead loading data. The method for
imputation is presented in Appendix C.
DRAFT -- DO NOT CITE OR QUOTE
62
August 28, 2000
-------
a
o
•z.
o
o
m
O
ID
D
c
O
en
oo
(Q
oo
100000 -
10000 -
l£-
I
^c
TJ
1
TJ
I I
I
Prt.rto19«0
1940-1159
1 960 - 1977(79)
Aft.r 1977(79)
Figure 3-3.
Boxplots of Area-Weighted Average Floor Wipe Dust-Lead Loadings (M9/ft2), by Housing Age Category, As
Observed in the §403 Risk Analysis (Using HUD National Survey Data) and in the Interim NSLAH (under 2
approaches to handling not-detected values)
(Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using the methods documented in the
§403 risk analysis report. Boxplots include imputed household averages but not negative averages. See text for definitions of labels along the horizontal axis.)
-------
o
D
o
o
z
o
o
H
m
O
1000000 -
O
m
o>
.p..
o> tooo -
o»
^c
TO
g too -
•fc
i
I 10"
1 -
O.I -
O.O1 -
-
I
I
1
-
j
I
1
«
1 (
-
|
t i
<
I
1
,
!
i
[
«
i
: i
i
!
•
•
i
i
c
1 I 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
HUONS NSLAH NSLAH HUONS NSLAH NSLAH HUONS NSLAH NSLAH HUONS NSLAH NSLAH NSLAH NSLAH
(403) (LOO/2) (4O3) (LOO/2) (4O3) (LOO/2) (403) (LOO/2) (LOO/2)
Prior to 194O 194O-19S9 1980-1977(79) Altar 1 977(79) Unknown
ca
S2.
to
oo
Figure 3-4. Boxplots of Area-Weighted Average Window Sill Wipe Dust-Lead Loadings (ug/ft2), by Housing Age Category,
As Observed in the §403 Risk Analysis (Using HUD National Survey Data) and in the Interim NSLAH (under 2
approaches to handling not-detected values)
o (Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using
8 §403 risk analysis report. Boxplots include imputed household averages but not negative averages. See text for definitions of lab
the methods documented in the
labels along the horizontal axis.)
-------
1 Table 3-8. Descriptive Statistics of Area-Weighted Average Floor Wipe Dust-Lead
2 Loadings for Households, Presented bv Census Region. As Reported in the
3 §403 Risk Analysis Versus the Interim NSLAH Data
Study
How Not-
Detected
and
Negative
Data were
Handled
Area-Weighted Average Floor Dust-Lead Loading (/ig/ft2)1
#
Surveyed
Units with
Positive
Averages
Arith-
metic .
Mean
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
25*
Percen-
tile
Median
75"1
Percen-
tile
Maximum
Northeast
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
53
103
109
35.6
10.0
10.3
14.9
2.28
2.90
3.95
4.42
3.15
0.632
-0.620
0.750
479
0.800
1.20
11.0
1.90
2.13
76.3
6.00
6.00
375
617
617
Midwest
§403 Risk Analysis
(HUD Natl Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
§403 Risk Analysis
(HUD Natl Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
73
136
150
134
235
265
52
159
182
14.7
14.7
15.0
13.3
265
3.07
9.81
187
19.1
6.32
1.34
2.04
5.01
0.981
1.55
4.97
0.949
1.46
3.26
5.81
3.39
South
3.28
394
2.25
West
2.75
3.66
2.31
0.508
-0.733
0750
0.735
-1.05
0.750
1.06
-1.23
0.750
2.83
0.283
0.760
2.00
0.254
0.970
2.65
0.255
0.800
6.32
1.20
1.29
3.89
0.940
1.21
401
0.800
1.20
11.0
2.48
3.25
10.0
1.76
1.94
8.43'
1.67
1.88
173
1040
1040
236
265
265
197
5940
5950
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with floor dust-lead data are
used to calculate the remaining statistics).
3 Summaries include imputed data for households having no floor wipe dust-lead loading data. The method for imputation is
presented in Appendix C.
DRAFT -- DO NOT CITE OR QUOTE
65
August 28, 2000
-------
1 Table 3-9. Descriptive Statistics of Area-Weighted Average Window Sill Wipe Dust-
2 Lead Loadings for Households, Presented bv Census Region. As Reported in
3 the §403 Risk Analysis Versus the Interim NSLAH Data
Study
How Not-
Detected
and
Negative
Data were
Handled
Area-Weighted Average Window Sill Dust-Lead Loading (ug/ft2)1
#
Surveyed
Units with
Positive
Averages
Arith-
metic
Mean
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
25"1
Percen-
tile
Median
75*
Percen-
tile
Maximum
Northeast
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
53
107
109
1440
172
172
92.2
21.5
226
16.1
8.01
7.06
0.0155
-1.89
0.578
15.3
5.94
5.94
173
16.0
160
335
89.5
90.0
14600
5530
5530
Midwest
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
73
145
150
134
259
265
52
179
182
564
218
218
432
115
116
62.2
54.3
54.4
48.5
21.0
21.6
19.6
13.8
156
4.45
7.73
8.72
13.2
7.25
6.49
South
12.4
8.11
642
West
12.7
6.65
5.59
0.0706
-2.32
1.12
0.118
-9.43
0.646
0.0118
-0.115
0.445
776
4.00
4.75
4.60
2.88
3.06
1.68
2.07
2.30
83.0
16.6
164
150
128
13.9
5.40
7.54
7.76
309
60.1
60.1
127
53.8
53.8
28.0
29.0
29.3
43700
9630
9630
28400
11100
11100
1400
3630
3630
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with window sill dust-lead
data are used to calculate the remaining statistics).
3 Summaries include imputed data for households having no window sill wipe dust-lead loading data. The method for
imputation is presented in Appendix C.
DRAFT » DO NOT CITE OR QUOTE
66
August 28, 2000
-------
1 Boxplots of the data distributions presented in Tables 3-8 and 3-9 are found in Figures
2 3-5 and 3-6, respectively. Appendix Dl contains these tabular summaries and boxplots after
3 excluding imputed data values.
4 Comparisons bv combination of housing age and Census region
5 Tables 3-10a and 3-1 Ob present descriptive statistics for household average floor dust-
6 lead loadings according to the 16 combinations of Census region and housing age category.
7 Table 3-10a considers no adjustment to the interim NSLAH data when not-detected results were
8 observed, while Table 3-10b summarizes data where not-detected data were replaced by one-half
9 of the detection limit. Tables 3-1 la and 3-1 Ib present the same descriptive statistics for
10 household average window sill dust-lead loadings. As the central tendency of the dust-lead
11 loading data was of primary interest to compare across the different combinations, these tables
12 only contain estimates of the arithmetic and geometric means, geometric standard deviation
13 (GSD), and median. Appendix Dl contains these tabular summaries after excluding imputed
14 data values.
15 Due to the small number of housing units within certain combinations, caution is
16 warranted when making inferences based on the numbers in these tables.
17 3.2.1.2. Data Summaries for the §403 Risk Analysis Versus Three Other Studies.
18 This subsection provides descriptive statistics of household average dust-lead loadings for floors
19 and window sills for the HUD National Survey (both as collected and as used in the §403 risk
20 analysis), comparing these summaries to those for the three studies identified in the introduction
21 to this section that provided the most useful and available information to the §403 risk analysis
22 on the relationship between environmental-lead levels and childhood blood-lead concentration:
23 the Baltimore R&M study, the Rochester Lead-in-Dust study, and the ongoing HUD Grantees
24 evaluation (data collected through February 1999).
25 Summaries of the reported dust-lead loadings in the HUD National Survey and the
26 Baltimore R&M study were performed on wipe-equivalent dust-lead loadings using conversion
27 methods presented in the §403 risk analysis report. In addition, the household averages based on
28 HUD National Survey data were summarized in two different ways: by ignoring the sample
29 weights assigned to the surveyed housing units and any imputed data for households with
30 missing data, and by handling the data as used in the §403 risk analysis (described earlier in this
31 section).
32 Because the HUD Grantees program emphasizes local control of the individual programs,
33 each grantee participating in the HUD Grantee evaluation is responsible for designing and
34 implementing lead-hazard reduction approaches applicable to its specific needs and objectives.
35 These responsibilities include the recruitment methods, enrollment criteria, and intervention
36 strategies. However, to enable comparison of results from the various approaches, grantees
37 participating in the evaluation follow the same sampling protocols and use standard data
38 collection forms developed specifically for this evaluation. Table 3-4 of the §403 risk analysis
DRAFT -- DO NOT CITE OR QUOTE 67 August 28,2000
-------
o
O 100000 -
O
O
m 1000°-
O
0
c
0 ,000
m
e
i
? 100 -
I J
O) o
" i
O.I -
0.01 -
. .
.
I
!
1 1
i •
1
,
•
•
j
.
.
* • • "
• S . .
1
1
II
j
1 .
' ! !
t! n n
1 J I • •
ft 1 !l
8 IS I j j
HUDHS
HUONS
(40J)
NSLAH
(LOO/Z)
NSLAH
(LOO/2)
C
CO
c
ro
CD
Figure 3-5. Boxplots of Area-Weighted Average Floor Wipe Dust-Lead Loadings (ug/ft2), by Census Region, As Observed
in the §403 Risk Analysis (Using HUD National Survey Data) and in the Interim NSLAH (under 2 approaches to
handling not-detected values)
(Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using the methods documented in the
§403 risk analysis report. Boxplots include imputed household averages but not negative averages. See text for definitions of labels along the horizontal axis.)
-------
o
o
2
O
o
H
m
O
ID
O
§
m
CO
1 1
1 1 :
•
i
I |
.
1 11
I
it ...
II •
• • •
'
i
•
i
,
i
1
•
1 1 1 1 1 1 1 I 1 1 1 1 1 1 1 1
HUONS NSLAH NSLAH HUONS NSLAH NSLAH HUONS NSLAH NSLAH HUONS NSLAH NSLAH
(403) (LOO/2) (4OJ) (LOD/2) (*O3) (LOD/2) <«03) (LOO/2)
North.omt Mldwval South w«t
co
Figure 3-6. Boxplots of Area-Weighted Average Window Sill Wipe Dust-Lead Loadings (ug/ft2), by Census Region, As
Observed in the §403 Risk Analysis (Using HUD National Survey Data) and in the Interim NSLAH (under 2
approaches to handling not-detected values)
^ (Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using the methods documented in the
§ §403 risk analysis report. Boxplots include imputed household averages but not negative averages. See text for definitions of labels along the horizontal axis.)
-------
1 Table 3-1 Oa. Descriptive Statistics of Area-Weighted Average Floor Wipe Dust-Lead
2 Loadings for Households, Presented bv Housing Age and Census Region. As
3 Reported in the §403 Risk Analysis Versus the Interim NSLAH Data Where
4 No Adjustments Were Made to Not-Detected Results
Census
Region
Northeast
Midwest
South
West
Study
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal
Interim NSLAH
§403 Risk Anal
Interim NSLAH
§403 Risk Anal
Interim NSLAH
§403 Risk Anal
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Housing Age
Category
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1940
1940- 1959
1960-1977
( 1960-79 for §403)
After 1977
(1979 for §403)
Area-Weighted Average Floor Dust-Lead Loading 0/g/ft2)
#
Surveyed
Units
26
41
17
21
10
19
15
19
33
21
35
29
32
4
25
19
26
33
42
64
71
18
72
13
11
16
36
17
54
6
39
Arithmetic
Mean
63.5
23.7
13.2
3.75
7.00
3.34
1.12
31.3
849
158
5.48
6.33
1.52
332
0.913
507
11.0
254
3.66
8.06
1.16
4.19
1.04
34.9
264
14.6
2.86
4.50
1.16
4.60
1.75
Geometric
Mean
36.5
5.02
8.84
2.37
4.73
1.72
0.714
14.7
2.62
6.69
205
4.58
0.737
2.77
0.545
20.8
3.66
10.3
1.63
4.13
0.825
316
0.549
16.2
384
9.04
1.70
3.53
0.949
3.36
0.454
Geometric
Std. Dev.
3.39
4.31
2.54
3.36
2.23
3.76
2.78
3.01
4.47
3.95
4.16
2.35
477
1.83
3.86
4.01
3.93
3.91
3.40
2.74
3.04
2.05
3.12
3.51
6.17
2.46
2.92
2.03
2.42
2.21
3.67
Median
76.3
4.20
7.81
2.38
4.76
1.46
0.867
8.94
2.16
5.79
1.59
4.44
1.12
2.80
0.320
19.0
2.74
10.0
1.77
3.39
0.880
2.84
0.480
17.2
2.30
7.47
1.36
3.35
0.990
3.00
0.270
5
6
8
10
11
12
Note: Summaries include imputed data for households having no floor wipe dust-lead loading data. The method for
imputation is presented in Appendix C.
DRAFT -- DO NOT CITE OR QUOTE
70
August 28, 2000
-------
1
2
3
4
5
6
Table 3-1 Ob. Descriptive Statistics of Area-Weighted Average Floor Wipe Dust-Lead
Loadings for Households, Presented bv Housing Age and Census Region. As
Reported in the §403 Risk Analysis Versus the Interim NSLAH Data Where
Not-Detected Results Were Replaced bv LOD/2
Census
Region
Northeast
Midwest
South
West
Study
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Housing Age
Category
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Area-Weighted Average Floor Dust-Lead Loading (//g/ft2)
#
Surveyed
Units
26
41
17
23
10
21
16
19
36
21
36
29
37
4
30
19
26
33
48
64
81
18
84
13
11
16
38
17
62
6
50
Arithmetic
Mean
63.5
23.8
13.2
4.03
7.00
3.58
1.68
31.3
8.79
158
5.80
6.33
200
3.32
1.31
50.7
11.1
25.4
394
8.06
1.67
4.19
1.54
34.9
264
14.6
3.07
4.50
1.62
4.60
2.34
Geometric
Mean
36.5
5.47
8.84
2.86
4.73
2.16
1.43
14.7
2.88
6.69
2.57
4.58
1.50
2.77
1.09
20.8
3.87
10.3
1.99
4.13
1.31
3.16
1.13
16.2
403
904
1.99
3.53
1.40
3.36
1.07
Geometric
Std. Dev.
3.39
3.91
2.54
2.23
2.23
2.60
1.72
3.01
3.41
3.95
3.20
2.35
2.03
1.83
1.67
4.01
3.76
3.91
2.35
2.74
1.73
2.05
1.57
3.51
5.91
2.46
2.34
2.03
1 65
2.21
1.95
Median
76.3
4.35
7.81
2.40
4.76
1.68
1.29
8.94
2.19
5.79
1 53
4.44
1.20
280
0.938
19.0
2.70
10.0
1.54
3.39
1.18
2.84
1.06
17.2
2.19
7.47
1.52
3.35
1.38
3.00
0.900
9
10
11
12
Note: Summaries include imputed data for households having no floor wipe dust-lead loading data. The method for
imputation is presented in Appendix C.
DRAFT - DO NOT CITE OR QUOTE
71
August 28, 2000
-------
1
2
3
4
5
6
Table 3-11 a. Descriptive Statistics of Area-Weighted Average Window Sill Wipe Dust-
Lead Loadings for Households, Presented bv Housing Age and Census
Region. As Reported in the §403 Risk Analysis Versus the Interim NSLAH
Data Where No Adjustments Were Made to Not-Detected Results
Census
Region
Northeast
Midwest
South
West
Study
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Housing Age
Category
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
( 1979 for 5403)
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for 5403)
Area-Weighted Average Window Sill Dust-Lead Loading (pg/ft2)
ft
Surveyed
Units
26
4O
17
23
10
2O
16
19
36
21
35
29
33
4
30
19
26
33
48
64
80
18
80
13
11
16
38
17
62
6
48
Arithmetic
Mean
2700
396
98.5
62.7
499
13.9
18.3
1660
361
982
103
223
279
62.5
21.0
2450
600
657
160
149
554
112
18.2
125
47.6
107
186
58.7
261
9.66
5.64
Geometric
Mean
265
99.4
32.6
20.1
38.9
7.88
3.28
435
72.5
17.7
20.0
20.9
9.94
27.5
6.57
64.0
112
38.9
30.7
24.0
14.3
9.09
3.93
11.5
14.2
7.35
29.0
3.83
834
2.65
1.99
Geometric
Std. Dev.
15.8
6.33
5.55
4.31
20.8
2.67
5.69
5.79
6 15
11.6
6.33
11.6
475
6.78
3.64
23.1
5.87
993
8.58
12.6
5.44
8.60
6.00
14.7
5 17
132
7.21
11 5
4.19
11.6
4.08
Median
176
91.7
50.7
18.5
217
6.49
2.06
542
673
17.4
17.1
48.3
9.54
83.0
5.86
24.4
115
26.2
32.0
32.0
15.4
7.58
389
7.05
17.1
6.96
33.8
4.35
7.51
5.94
1.63
10
11
12
Note: Summaries include imputed data for households having no window sill wipe dust-lead loading data. The method for
imputation is presented in Appendix C.
DRAFT -- DO NOT CITE OR QUOTE
72
August 28, 2000
-------
1
2
3
4
5
6
Table 3-11 b. Descriptive Statistics of Area-Weighted Average Window Sill Wipe Dust-
Lead Loadings for Households, Presented by Housing Age and Census
Region. As Reported in the §403 Risk Analysis Versus the Interim NSLAH
Data Where Not-Detected Results Were Replaced bv LOD/2
Census
Region
Northeast
Midwest
South
West
Study
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal
Interim NSLAH
Housing Age
Category
Prior to 1940
1940-1959
1960-1977
1 1960-79 for §403)
After 1977
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1 977
(1979 for §403)
Prior to 1 940
1940- 1959
1960-1977
1 1960-79 for §403)
After 1977
(1979 for §403)
Area-Weighted Average Window Sill Dust-Lead Loading (//g/ft2)
«
Surveyed
Units
26
41
17
23
10
21
16
19
36
21
36
29
37
4
30
19
26
33
48
64
81
18
84
13
11
16
38
17
62
6
50
Arithmetic
Mean
2700
396
98.5
62.7
499
14.7
18.6
1660
361
98.2
103
223
284
62.5
21.4
2450
600
657
160
149
55.7
112
18.8
125
47.8
107
186
58.7
26.0
9.66
5.77
Geometric
Mean
265
90.1
32.6
19.6
38.9
8.39
4.80
435
75.7
17.7
20.9
20.9
10.3
27.5
7.01
64.0
112
389
35.5
24.0
15.3
9.09
5.21
11.5
15.9
7.35
30.6
3.83
8.77
2.65
2.57
Geometric
Std. Dev.
15.8
6.91
5.55
4.49
20.8
2.55
3.80
5.79
565
11.6
549
11.6
3.81
6.78
3.54
23.1
5.86
9.93
6.78
12.6
4.88
8.60
386
14.7
4.23
13.2
651
11.5
3.88
11.6
3 14
Median
176
91.7
50.7
18.9
217
7.37
3.73
542
67.3
17.4
17.6
48.3
9.54
83.0
6.20
24.4
115
26.2
32.0
32.0
15.8
7.58
4.00
7.05
17.2
6.96
33.8
4.35
7.51
5.94
1.85
10
11
12
Note: Summaries include imputed data for households having no window sill wipe dust-lead loading data. The method for
imputation is presented in Appendix C.
DRAFT - DO NOT CITE OR QUOTE
73
August 28, 2000
-------
1 report documented the differences between grantees in their enrollment/recruitment criteria. As a
2 result, the HUD Grantees data summaries in this subsection are presented by grantee.
3 Overall data summaries
4 Figures 3-7 and 3-8 present boxplots of the area-weighted household average dust-lead
5 loadings for floors and window sills, respectively. Each of these two figures contains a boxplot
6 for each study, along with separate boxplots for each grantee in the HUD Grantees evaluation8.
7 Each figure also includes three boxplots associated with the HUD National Survey data:
8 • "HUDNS (U)" summarizes the data without regard to sampling weights
9 • "HUDNS (403)" summarizes the data as used in the §403 risk analysis (e.g., using
10 sampling weights reflecting the 1997 housing stock; incorporating imputed data
11 assigned to housing units with missing data)
12 • "HUDNS (OW)" summarizes the data weighted according to the original weights
13 assigned in the survey.
14 Tables 3-12 and 3-13 present values of the statistics presented in the boxplots (geometric
15 mean, minimum, median, maximum, 25th and 75lh percentiles), along with other important
16 information not explicitly observable from the boxplots (number of houses whose data enter into
17 these statistics, geometric standard deviation) that is necessary when comparing distributions
18 across studies. The GSD reported for the overall HUD Grantees evaluation is the exponentiation
19 of the square root of the weighted average of log-transformed variances for the different grantees,
20 where the weights correspond to the numbers of units with data.
21 Comparisons bv housing age category
22 Figures 3-9 and 3-10 contain boxplots on pre-1980 housing data (floors and window sills,
23 respectively) from the HUD National Survey, Baltimore R&M, and Rochester studies, and pre-
24 1978 data from the HUD Grantees evaluation (data combined across grantees) according to three
25 housing age categories (pre-1940, 1940-1959, 1960-1977/79). As in the overall summaries
26 above, the HUD National Survey data are presented within three boxplots for each age category.
27 Caution is warranted when interpreting results in these figures for the Rochester study, as the
28 actual age of certain houses may be older than what was specified in the Rochester study
29 database (see Section 3.3.1.3 of the §403 risk analysis report). Also for this reason, and since the
30 other studies surveyed few, if any, post-1979 homes, boxplots were not created for homes built
31 after 1979. Boxplots for non-control houses in the Baltimore R&M study, all of which were built
32 prior to 1941, are also included in these figures and are displayed in the "pre-1940" category.
8 "Alam"=Alameda County, "Balt"=Baltimore, "Bos"=Boston, "CA"=Califomia, "Cle"=Cleveland,
"MA"=Massachusetts; "MN"=Minnesoia; "NJ"=New Jersey; "Rl"=Rhode Island, "WI"=Wisconsm, "Milw"=Milwaukee,
"Chic"=Chicago; "NYC"=New York City; "VT"=Vermont
DRAFT -- DO NOT CITE OR QUOTE 74 August 28,2000
-------
0
u
O
O 100000 -
0
H
O
m 10000 -
O
33
O
S^ 1OOO -
m
c
g> 100 -
T> 10 "
i -
0.1 -
0.01 -
•
•
• <
. .
1
i
1
•
•
•
; •
• i
. i
1
HUONS MUONS HUDNS Itoatw
(U) <40J> (OW)
1
mtmr »ottl
n
•
:
9
1
*
•
: i
• 9
• | •
• • • |
! * ' : 1
: i :
: 1 • i ! i
i •
i
L
I
; i
i
i
[
1
-
]
! i
i i
i
i 1 1 1 i i i i i i i r
nor> »lom maa «,. CA Cl. M» UN HJ Kl Wl Mil. CMo HTC VT Oov
1M 1 HUQ Gr«««^-« 1
> Figure 3-7. Boxplots of Area-Weighted Average Pre-lntervention Floor Wipe Dust-Lead Loadings (ugffi2) for Houses in the
HUD National Survey, Baltimore R&M Study, Rochester Lead-in-Dust Study, and Grantees Within the HUD
Grantees Evaluation
c
(Q
C
cn
ro
oo
(Note: Dust-lead loadings from the HUD National Survey and Baltimore R&M study have been converted from vacuum to wipe-equivalents using the methods
documented in the §403 risk analysis report. See text for definitions of labels along the horizontal axis.)
-------
o
ID
1000OOO -
z
0
H
O
1^ 100000 -
rn
O
3)
D
C 10000 -
3
m C^
•3- 1000 -
J"
1
g 100 -
fi
o
» 5
1
C
1 -
O.I -
0.01 -
• I
• 1
1 1
1 1
• I
1 1
-
..
1
1
1
1
1
.
<
;
i
.
1
]
<
I
i
I
' 4
1
li
1
1 '
1
j I
.
1
1
4
:
i
t I
I
1
.
,
1
.
i
i
]
•
i
4
4
|
4
|
!
I
.
;
i
1 I
T T
-•
i
i
i
i
i
••
HUD Grantees .
> Figure 3-8.
-------
1
2
3
Table 3-12. Descriptive Statistics of Area-Weighted Average Pre-lntervention Floor Wipe
Dust-Lead Loadings for Households, As Reported in the §403 Risk Analysis,
the HUD National Survey, and Other Studies
Study
HUD
National
Survey
Approach/
Grantee
unweighted1
orig. weights2
§403 RA3
Rochester Lead-in-Dust
Baltimore R&M4
HUD
Grantees
Alameda Co.
Baltimore
Boston
California
Cleveland
Massachusetts
Minnesota
New Jersey
Rhode Island
Wisconsin
Milwaukee
Chicago
New York City
Vermont
All Grantees
Area-Weighted Average Pre-lntervention Floor Dust-Lead Loading (fig/ft2)
# Units
with
Data
281
281
284
205
90
168
402
114
90
190
229
212
45
203
236
291
158
399
354
3091
Arith-
metic
Mean
21.0
21.0
16.5
110
54.3
127
642
205
130
232
408
202
308
530
172
247
234
462
515
366
Geo-
metric
Mean
8.19
7.97
6.27
17.7
40.9
31.4
149
61.3
246
39.4
64.4
27.3
682
60.6
24.8
32.2
29.4
52.6
679
50.1
Geo-
metric
Std. Dev.
3.66
3.70
3.49
3.20
2.27
5.78
5.48
4.79
4.89
6.51
6.47
5 14
6.71
5.85
692
4.61
4.40
590
6.70
5.76
Minimum
0.508
0.508
0.508
1.21
4.48
0.250
0.250
1.00
2.75
1.00
0.521
0.333
1.75
0.250
0400
1.50
0.200
0.0880
0.750
00880
25th
Percen-
tile
3.23
3.17
2.65
10.4
29 1
8.59
53.2
18.8
7.95
10.3
17.0
10.9
105
17.7
5.99
11 0
11.5
18.5
158
14.3
Median
7.27
694
5.32
16.1
45.2
31.0
167
55.3
15.6
364
59.8
192
93.4
54.0
16.9
27.5
28.2
32.9
49.9
40.2
75"1
Percen-
tile
17.3
17.0
12.2
26.6
70.4
98.0
456
170
59.3
134
234
62.4
298
187
79.1
76.3
69.2
94.4
219
165
Maximum
375
375
375
8660
266
3730
89100
2490
2650
8800
16600
13800
4250
59200
2780
31900
26400
22200
15600
89100
5
6
7
8
9
10
11
12
13
14
15
16
17
18
1 Area-weighted average dust-lead loadings, as reported in the HUD National Survey but converted to wipe-equivalent
loadings, are summarized without weighting by sample weights.
2 Area-weighted average dust-lead loadings, as reported in the HUD National Survey but converted to wipe-equivalent
loadings, are summarized by weighting each average by the original sample weights assigned in the survey.
3 Area-weighted average dust-lead loadings, as calculated in Chapter 3 of the §403 risk analysis, are summarized by
weighting each average to reflect the 1997 U.S. housing stock and imputing averages for units with missing data.
* BRM dust-lead loadings are converted to wipe-equivalent loadings prior to summary in this table
DRAFT - DO NOT CITE OR QUOTE
77
August 28, 2000
-------
1 Table 3-13. Descriptive Statistics of Area-Weighted Average Pre-lntervention Window
2 Sill Wipe Dust-Lead Loadings for Households, As Reported in the §403 Risk
3 Analysis, the HUD National Survey, and Other Studies
Study
HUD
National
Survey
Approach/
Grantee
unweighted1
orig. weights2
§403 RA3
Rochester Lead-in-Oust
Baltimore R&M4
HUD
Grantees
Alameda Co.
Baltimore
Boston
California
Cleveland
Massachusetts
Minnesota
New Jersey
Rhode Island
Wisconsin
Milwaukee
Chicago
New York City
Vermont
All Grantees
Area-Weighted Average Pre-lntervention Window Sill Dust-Lead Loading (fig/ft1)
# Units
with
Data
245
245
312
196
90
178
402
95
81
185
206
193
51
192
234
271
146
382
318
2934
Arith-
metic
Mean
678
721
550
558
627
677
6690
4090
909
4050
2990
3160
758
4930
2790
3520
1600
1580
3740
3360
Geo-
metric
Mean
21.7
24.9
23.0
196
356
118
1560
452
316
259
425
308
93.7
659
279
536
260
267
246
380
Geo-
metric
Std. Dev.
15.4
17.9
158
3.96
3.55
9.14
7.39
9.87
4.60
16.2
7.13
6 17
27.8
11.9
844
6.89
571
5.58
14.6
868
Minimum
0.0118
0.0118
0.0118
2.83
20.6
0.0016
< 00001
0.0053
11.0
< 0.0001
2.15
566
< 0.0001
< 0.0001
0.0008
1.00
3.02
0.320
<00001
< 0.0001
25"1
Percen-
tile
3.57
5.22
4.35
80.6
112
37.7
444
135
94.2
72.8
108
726
32.8
186
80.7
127
867
97.0
45.0
102
Median
17.6
36.3
19.5
183
576
134
1690
385
293
288
369
262
104
666
256
424
267
183
227
343
75*
Percen-
tile
149
217
198
416
960
464
5800
2040
1030
949
1420
1030
435
2450
845
2110
877
670
1260
1490
Maximum
43700
43700
43700
14900
2330
19700
220000
106000
9630
241000
76100
300000
8450
132000
142000
88000
50500
57100
98100
300000
5
6
7
8
9
10
11
12
13
14
15
16
17
18
1 Area-weighted average dust-lead loadings, as reported in the HUD National Survey but converted to wipe-equivalent
loadings, are summarized without weighting by sample weights.
2 Area-weighted average dust-lead loadings, as reported in the HUD National Survey but converted to wipe-equivalent
loadings, are summarized by weighting each average by the original sample weights assigned in the survey.
3 Area-weighted average dust-lead loadings, as calculated in Chapter 3 of the §403 risk analysis, are summarized by
weighting each average to reflect the 1997 U S. housing stock and imputing averages for units with missing data.
* BRM dust-lead loadings are converted to wipe-equivalent loadings prior to summary in this table.
DRAFT ~ DO NOT CITE OR QUOTE
78
August 28, 2000
-------
o
n
o
o
I
o
m
O
O
§
m
o
o
J
S
•o
i
o
§
I
o
o
o
o
o
o
o
o
o
888
§ 8
CD
HUDNV
(U)
HUONm
(401)
MOON1 HUON« He
(«0») (OW)
It BO - l»77(7») .
-------
0
o
o
o
H
m
O
D
c
O
m
oo
o
a o <
) O O
) O O
I I
>
[
•
o o o
o o o
o <
1 O
I t I
? n
s
tl
I
-•
1 i 11 i i i i 1 1 1 1 1 1 1 1 r
ft1 Tss "tss? "~~'" -ssr o^^. H?g?m TSS? TS? — — * o^ss- "vsr* TSS? TcS? —• "* *ri'
1 FrtortolMO 1 1 1.4O - 1««t 1 1 ,((o ttrT(r*) 1
Figure 3-10. Boxplots of Area-Weighted Average Pre-lntervention Window Sill Wipe Dust-Lead Loadings (ug/ft2) for
Houses in the HUD National Survey, Baltimore R&M Study, Rochester Lead-in-Dust Study, and HUD Grantees
Evaluation, by Age of House Category (pre-1979 only)
-°° (Note: Dust-lead loadings from the HUD National Survey and Baltimore R&M study have been converted from vacuum to wipe-equivalents using the methods
o documented in the §403 Risk Analysis report. See text for definitions of labels along the horizontal axis. Caution must be taken when categorizing houses in the
S Rochester studv bv ane of housp }
o Rochester study by age of house.)
-------
1 Values of the statistics entering into the boxplots in Figures 3-9 and 3-10 are included
2 within Tables 3-14 and 3-15. While not included in the figures, these tables include summary
3 statistics for homes labeled as post-1979 (although the Rochester study units may not have
4 actually been built in this time period, as mentioned in the previous paragraph). The post-1979
5 results labeled as "HUD National Survey (§403 RA)" represent surveyed homes built from 1960-
6 1979 that contain no lead-based paint (Section 3.3.1.5 of the §403 risk analysis report).
7 3.2.1.3 Calculating National Exceedance Percentages for Household Average Floor
8 Dust-Lead Loading. With respect to the national summaries of household average floor dust-
9 lead loading presented in Section 3.2.1.1 above, it was desired to estimate the percentage of
10 housing with average floor dust-lead loadings at or above specified thresholds (i.e., "exceedance
11 percentage"), with separate estimates originating from data for each of the two national surveys
12 (i.e., HUD National Survey and the interim NSLAH). This was done by fitting a lognormal
13 distribution to the household average floor dust-lead loadings summarized in Section 3.2.1.1 and
14 calculating the exceedance percentages based on this distribution.9 If the household averages
15 from the two surveys could each be considered a sample from their respective fitted lognormal
16 distributions, with the probability of selection for the sample determined by the sampling
17 weights, then the estimates based on these fitted distributions would be considered representative
18 of actual percentages for the nation. The fitted lognormal distributions and the resulting
19 exceedance percentage estimates are now presented for both surveys.
20 For both surveys, normal probability plots prepared on the log-transformed average floor
21 dust-lead loadings indicated that a lognormal distribution did not adequately represent data in the
22 upper tails of the distribution (i.e., typically the upper quartile). This was because the fitted
23 distribution was heavily influenced by the considerable amount of data at the lower end of the
24 distribution. Because it was necessary in this exercise to characterize the upper tail of the
25 distribution as accurately as possible (due to calculating exceedance percentages from the
26 distribution), the actual values of the data at the lower end of the distribution did not need to
27 influence the fitted distribution to the extent that they were. Under these considerations, the
28 procedure to fit a lognormal distribution was as follows:
29 • For values of P from 5 to 50 (in multiples of 5), the value of the log-transformed
30 average floor dust-lead loading (call this value X) was identified for which P% of
31 the (weighted) data fell below.
32 • For each value of P, log-transformed data values falling below the value X were
33 considered to be censored at X. That is, rather than using these actual log-
34 transformed data values, the procedure assumed that each of these values was
35 somewhere at or below X.
Q
For the interim NSLAH, household averages calculated from data where no adjustment was made when below
detection limits were used in this exercise
DRAFT -- DO NOT CITE OR QUOTE 81 August 28,2000
-------
1 Table 3-14. Descriptive Statistics of Area-Weighted Average Pre-lntervention Floor Wipe
2 Dust-Lead Loadings for Households, Presented bv Housing Aae Category.
3 As Reported in the §403 Risk Analysis, the HUD National Survey, and Other
4 Studies
Study
Approach/
Grantee
Area-Weighted Household Average Pre-lntervention Floor Dust-Lead Loading 0/g/ft2)
# Units
with
Data
Arith-
metic
Mean
Geo-
metric
Mean
Geo-
metric
Std. Dev.
Houses Built Prior to
HUD
National
Survey
unweighted1
orig. weights2
§403 RA3
Baltimore R&M4
Rochester Lead-in-Dust
HUD
Grantees
Alameda Co.
Baltimore
Boston
California
Cleveland
Massachusetts
Minnesota
New Jersey
Rhode Island
Wisconsin
Milwaukee
Chicago
New York City
Vermont
All Grantees
76
76
77
74
172
138
345
71
35
173
146
182
26
123
214
262
144
375
288
2522
43.9
47.9
47.9
63.6
127
118
672
222
269
209
147
171
511
197
183
254
60.7
470
478
328
19.5
22.4
22.6
55.5
19.8
306
153
55.4
48.4
347
31 7
21.3
215
44.2
28.4
30.7
25.6
50.0
63.7
459
3.68
3.65
3.63
1.65
3 18
5.40
5.10
5.11
6 75
6.30
5.03
4.72
4.19
4.68
6 77
4.42
3.92
593
6.81
5.45
Minimum
1940
0.991
0.991
0.991
22.0
1.66
0.250
1 00
1.00
2.75
1.00
0.521
0333
10.5
200
0400
1 50
0200
00880
0.750
00880
25"1
Percen-
tile
8.45
8.84
8.84
38.7
11 3
103
546
16.0
8.38
9 50
11.9
10.0
134
16.4
726
11.0
10.7
18 1
15.8
13.7
Median
17.1
17 7
17.7
54.3
16.9
31.0
164
35.0
350
31.0
26.5
168
239
38.7
18 5
26.3
254
31.4
49.0
36.1
75*
Percen-
tile
47.1
797
797
76.0
30.0
97.7
456
151
250
121
83.1
40.0
513
106
99.5
71 0
62.4
844
197
145
Maximum
375
375
375
266
8660
3730
89100
2490
2650
8800
4540
13800
4250
6050
2780
31900
668
22200
15500
89100
Houses Built From 1940 - 1959
HUD
National
Survey
unweighted1
orig. weights2
§403 RA3
Rochester Lead-in-Dust5
HUD
Grantees
Alameda Co.
Baltimore
Boston
California
Massachusetts
Minnesota
87
87
87
19
19
43
4
51
5
1
198
18.1
18.1
11.8
153
494
57.3
41.7
55.5
149
9.20
8.74
8.74
8.36
32.1
120
26.6
15.4
46.5
149
3.53
3.34
3 34
2.61
7.15
9.13
4.46
3.29
1.93
--
0508
0.508
0.508
1 21
2.00
0.250
500
2.75
22.5
149
4.20
4.07
4.07
3.54
5.75
39.5
100
6.25
30.0
149
8.32
7.81
7.81
11.1
170
197
27.0
10 1
39.8
149
22.5
22.4
22.4
19.2
157
648
105
33.3
70.3
149
171
171
171
26.9
909
4170
170
825
115
149
DRAFT - DO NOT CITE OR QUOTE
82
August 28, 2000
-------
Table 3-14. (cont.)
Area-Weighted Household Average Pre-lntervention Floor Dust-Lead Loading (/ig/ft1)
Study
Approach/
Grantee
# Units
with
Data
Arith-
metic
Mean
Geo-
m6tfic
Mean
Geo-
metric
Std. Dev.
Minimum
25-
Percen-
tile
Median
75*
Percen-
tile
Maximum
Houses Built From 1940 - 1959 (cont.)
HUD
Grantees
Rhode Island
Wisconsin
Milwaukee
Chicago
Vermont
All Grantees
34
15
5
5
31
213
81.3
87.0
14.0
5300
38.4
276
27.3
7.22
6.78
102
26.4
30.1
5.47
6.99
4.14
23.8
2.23
5.39
0.250
0.800
1.50
16.4
8.00
0.250
7.60
1.60
2.25
17.8
15.0
8.00
36.6
5.72
4.88
19.2
17.8
24.3
77.3
17.1
225
75.8
45.5
893
617
1050
38.8
26400
219
26400
Houses Built From 1960 - 1979 (1960 - 1977 for HUD Grantees)
HUD
National
Survey
unweighted1
orig. weights2
§403 RA3
Rochester Lead-in-Dust5
HUD
Grantees
Boston
Cleveland
New Jersey
Wisconsin
All Grantees
118
118
120
4
1
1
16
6
24
7.14
6.74
6.74
9.65
18.8
9.25
32.6
4.42
240
4.30
4.11
4.14
7.84
18.8
9.25
13.6
4.01
10.0
2.50
2.46
2.45
2.40
-
-
3.70
1.61
3.14
0.657
0.657
0.657
2.13
18.8
9.25
1.75
2.40
1 75
2.26
2.25
2.25
6.38
18.8
9.25
6.58
2.50
4.45
3.85
3.62
3.62
11.6
18.8
9.25
10.0
3.84
8.88
7.59
7.59
7.59
12.9
18.8
9.25
34.6
5.93
24.6
106
106
106
13.2
18.8
9.25
245
8.02
245
Houses Built After 1979 (After 1977 for HUD Grantees)
HUD National Survey
(§403 RA)3
Baltimore R&M4
Rochester Lead-in-Dust5
HUD
Grantees
Minnesota
Rhode Island
All Grantees
28
16
10
1
3
4
4.16
10.9
37.2
32.4
984
746
3.14
9.97
15.0
32.4
838
372
2.06
1.55
334
-
2.00
2.00
1.06
4.48
3.48
32.4
440
32.4
1.76
7.13
5.57
32.4
440
236
2.84
10.5
16.8
32.4
763
602
5.66
147
21.2
32.4
1750
1260
12.9
17.4
250
32.4
1750
1750
2
3
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
1 Area-weighted average dust-lead loadings, as reported in the HUD National Survey but converted to wipe-equivalent
loadings, are summarized without weighting by sample weights.
1 Area-weighted average dust-lead loadings, as reported in the HUD National Survey but converted to wipe-equivalent
loadings, are summarized by weighting each average by the original sample weights assigned in the survey.
3 Area-weighted average dust-lead loadings, as calculated in Chapter 3 of the §403 risk analysis, are summarized by
weighting each average to reflect the 1997 U.S. housing stock and imputing averages for units with missing data.
4 BRM dust-lead loadings are converted to wipe-equivalent loadings prior to summary in this table.
5 Some houses in this housing age category may belong to an earlier age category, as some houses may have actually been
built earlier than the year specified within the study's database.
DRAFT - DO NOT CITE OR QUOTE
83
August 28, 2000
-------
1 Table 3-15. Descriptive Statistics of Area-Weighted Average Pre-lntervention Window
2 Sill Wipe Dust-Lead Loadings for Households, Presented bv Housing Age
3 Category. As Reported in the §403 Risk Analysis, the HUD National Survey,
4 and Other Studies
Study
Approach/
Grantee
Area-Weighted Household Average Pre-lntervention Window Sill Dust-Lead Loading (fig/ft2)
# Units
with
Data
Arith-
metic
Mean
Geo-
metric
Mean
Geo-
metric
Std. Dev.
Houses Built Prior to
HUD
National
Survey
unweighted1
orig. weights2
§403 RA3
Baltimore R&M4
Rochester Lead-in-Dust
HUD
Grantees
Alameda Co.
Baltimore
Boston
California
Cleveland
Massachusetts
Minnesota
New Jersey
Rhode Island
Wisconsin
Milwaukee
Chicago
New York City
Vermont
All Grantees
71
71
77
74
164
148
347
71
35
172
146
177
26
123
211
261
140
368
269
2494
1610
2060
2060
751
613
767
7070
5150
1530
4120
1770
3320
1080
5780
3020
3610
1630
1530
3860
3480
54.7
146
168
555
234
138
1600
577
506
233
322
282
328
816
294
543
259
258
272
391
19.6
16.8
16.7
2.41
3.67
9.34
7.69
7.89
5.70
16.9
5.81
6.30
5.08
6.49
8.91
6.89
5.78
5 51
15.9
8.22
Minimum
1940
0.0155
0.0155
0.0155
44.0
2.85
0.0016
< 0.0001
14.0
28.0
< 0.0001
2.60
5.66
21.3
120
0.0008
1 00
3.02
0.320
< 00001
< 0.0001
25"1
Percen-
tile
7.05
35.6
35.6
399
95.3
44.7
451
135
159
63.8
93.8
71 0
99.6
192
81.4
127
85.0
95.6
72.0
106
Median
67.1
198
198
628
223
164
1690
425
524
270
296
190
276
709
258
413
267
175
275
351
75*
Percen-
tile
442
1220
1220
989
475
566
6140
2410
2440
876
1090
945
1170
2500
1090
2110
852
543
1340
1470
Maximum
43700
43700
43700
2330
14900
19700
220000
106000
9630
241000
63400
300000
8450
132000
142000
88000
50500
57100
98100
300000
Houses Built From 1940 - 1959
HUD
National
Survey
unweighted1
orig. weights2
§403 RA3
Rochester Lead-in-Dust5
HUD
Grantees
Alameda Co.
Baltimore
Boston
California
Massachusetts
Minnesota
79
79
87
18
20
43
4
42
4
1
430
285
285
399
152
4310
382
395
142
289
23.1
17.9
22.0
72.0
47.7
1330
150
203
59.7
289
11.4
10.5
10.7
6.16
8.04
5.39
5.20
3.41
8.20
~
0.0118
0.0118
0.0118
2.83
0.140
33.0
39.4
11.0
2.79
289
6.47
6.47
6.47
23.0
14.5
256
39.6
89.9
47.9
289
21.7
19.1
19.1
56.0
71.1
1600
160
190
123
289
107
107
107
194
260
4820
724
565
237
289
16100
16100
16100
4390
580
29400
1170
1850
321
289
DRAFT -- DO NOT CITE OR QUOTE
84
August 28, 2000
-------
Table 3-15. (cont.)
Study
Approach/
Grantee
Area-Weighted Household Average Pro-Intervention Window Sill Dust-Lead Loading (fig/ft2)
# Units
with
Data
Arith-
metic
Mean
Geo-
metric
Mean
Geo-
metric
Std. Dev.
Minimum
25th
Percen-
tile
Median
75*
Percen-
tile
Maximum
Houses Built From 1940 - 1959 (cont.)
HUD
Grantees
Rhode Island
Wisconsin
Milwaukee
Chicago
Vermont
All Grantees
34
16
6
5
30
205
1520
497
552
835
52.4
1350
416
148
140
449
40.4
222
7.53
4.24
7.09
3.84
2.08
4.94
0.500
24.0
18.0
111
7.00
0.140
144
47.4
28.8
120
31.0
45.0
617
105
123
521
45.0
205
1120
338
797
1170
45.0
814
9970
4750
2220
2250
212
29400
Houses Built From 1960 - 1979 (1960 - 1977 for HUD Grantees)
HUD
National
Survey
unweighted1
ong. weights2
§403 RA3
Rochester Lead-in-Dust5
HUD
Grantees
Boston
Cleveland
New Jersey
Wisconsin
All Grantees
95
95
120
4
1
1
20
6
28
190
184
184
54.4
289
409
59.8
209
112
Houses Built
HUD National Survey
(§403 RA)3
Baltimore R&M4
Rochester Lead-in-Dust5
HUD
Grantees
Minnesota
Rhode Island
All Grantees
28
16
10
1
1
2
83.0
50.8
134
2350
816
1580
10.3
9.10
16.2
523
289
409
129
153
27.8
13.3
14.5
14.6
1 38
-
-
63.1
2.90
39.3
0.0164
0.0164
0.0164
36.2
289
409
<0.0001
21.0
< 0.0001
1.68
2.05
2.05
40.0
289
409
17.8
105
20.9
8.69
16.6
16.6
55.2
289
409
29.6
240
44.4
51.3
217
217
68.7
289
409
72.7
289
179
5790
5790
5790
70.7
289
409
333
359
409
After 1979 (After 1977 for HUD Grantees)
8.17
45.6
113
2350
816
1390
9.94
1.65
1.95
-
--
-
0.0164
20.6
26.9
2350
816
816
2.58
27.1
75.7
2350
816
816
8.11
52.6
125
2350
816
1580
57.8
66.5
159
2350
816
2350
1590r
85.9
320
2350
816
2350
2
3
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
1 Area-weighted average dust-lead loadings, as reported in the HUD National Survey but converted to wipe-equivalent
loadings, are summarized without weighting by sample weights.
2 Area-weighted average dust-lead loadings, as reported in the HUD National Survey but converted to wipe-equivalent
loadings, are summarized by weighting each average by the original sample weights assigned in the survey.
3 Area-weighted average dust-lead loadings, as calculated in Chapter 3 of the §403 risk analysis, are summarized by
weighting each average to reflect the 1997 U.S. housing stock and imputing averages for units with missing data.
4 BRM dust-lead loadings are converted to wipe-equivalent loadings prior to summary in this table.
5 Some houses in this housing age category may belong to an earlier age category, as some houses may have actually been
built earlier than the year specified within the study's database.
DRAFT - DO NOT CITE OR QUOTE
85
August 28, 2000
-------
1 • For each value of P, a normal distribution was fitted to the log-transformed data,
2 taking into account the censoring of the lower P% of the data and the sample
3 weights, using the LIFEREG procedure in the SAS® System.
4 • The value of P (and its corresponding cut-off X) was identified that resulted in the
5 best fit for normality in the upper tail of the distribution (based on review of
6 normal probability plots). The exceedance percentages were estimated based on
7 this final distribution, using normal probability theory.
8 This procedure was applied separately to HUD National Survey data and interim data from the
9 NSLAH. Exceedance percentages were estimated for each of the following floor dust-lead
10 loading thresholds: 5, 10, 20, 30,40, and 50 ug/ft2.
11 Figure 3-11 contains the fitted distributions based on the HUD National Survey data (top
12 plot) and the interim NSLAH data (bottom plot). (The top plot is labeled "Section 403 risk
13 analysis" as it reflects sample weights adjusted for the 1997 housing stock and dust-lead loadings
14 converted to wipe-equivalents, both done within the §403 risk analysis.) Each plot contains a bar
IS chart of the observed data, onto which the fitted lognormal distribution curve is superimposed.
16 Note that the same floor dust-lead loading (horizontal) axis is used for both plots, so that the two
17 plots can be directly compared. As can be noted in this figure (and which was seen in the
18 summaries in Section 3.2.1.1), the distribution based on the interim NSLAH data covers a
19 considerably lower range compared to the distribution based on the HUD National Survey data
20 used in the §403 risk analysis. Thus, the estimated exceedance percentages for each of the six
21 thresholds, also annotated within each plot, are considerably lower based on the interim NSLAH
22 data, especially as the threshold increases.
23 Each estimated exceedance percentage within Figure 3-11 is accompanied by an
24 approximate 95% confidence interval on the number of homes in the U.S. housing stock that
25 exceeds the threshold. These intervals were calculated based on the estimated total number of
26 housing units in the housing stock, as determined by the sum of the sampling weights for the
27 given survey (which is specified within each plot).
28 In Figure 3-11, the distribution based on the HUD National Survey data used in the §403
29 risk analysis was determined by censoring data values below 3.81 ug/ft2 (i.e., the bottom 40
30 percent of the data, taking into account the sample weights). The distribution based on the
31 interim NSLAH data was determined by censoring data values below 0.2025 ug/ft2, which
32 corresponds to the bottom 20 percent of the observed weighted distribution, including negative
33 values.
34 For both surveys, the estimated exceedance percentages specified within Figure 3-11 for
35 household average floor dust-lead loading, based on the fitted lognormal distribution, are also
36 included within Table 3-16 (columns 2 and 4) for the same six thresholds. Also included in
37 Table 3-16 (columns 3 and 5) are estimated exceedance percentages that were determined solely
38 by the proportion of total sampling weights in the survey that corresponded to surveyed units
DRAFT - DO NOT CITE OR QUOTE 86 August 28,2000
-------
o
3)
O
O
O
g
H
m
O
3J
o
c
o
m
SECTION 403 RISK ANALYSIS
Homes Above 10 n
Homes Above 20 ng/ft2:
18.5% (14.9 M to 21.6 M) Homes Above
33.6% (29.1 M to 37.5 M )
Homes Above 5 ug/ft2:
51 .9% (46.7 M to 56.3 M )
(Total Number of Homes: 99.3 M)
1 2% (9.1 M to 1 5 M )
Homes Above 40 |ig/ft2:
8.5% (6.1 M to 1 1 .1 M)
Homes Above 50 iig/ft2:
6.4% (4.3 M to 8.6 M )
001
0.1
1
10
100
Floor D ust-Lead Loading (ng/ft )
1000
oo
INTERIM
NSLAH
q
d
Homes Above 1 0 ug/ft2:
7.9% (5.7 M to 8.3 M
Homes Above 5 ng/ft2:
1 5.8% (1 2.1 M to 15.9
(Total Number of
(Rep
Homes Above 20 iig/ft2:
3.4% (2.2 M to 3.8 M )
Homes Above 30 ug/ft2:
2% (1 .2 M to 2.4 M )
Homes Above 40
ia/(t
2.
1 .3% (0.8 M to 1 .6 M )
Homes Above 50 ng/ft
0.9% (0.5 M to 1 .2 M)
0.01
0.1
TT-T 1 1 1—i i i i i i 1 1 1—i—
1 10
Floor D ust-Lead Loading (ug/ft2
100
1000
> Figure 3-11. Estimated Distribution of Household Average Floor Dust-Lead Loading in the Nation's Housing Stock, and
| Corresponding Estimates of the Percentage of Homes Exceeding Specified Thresholds (with 95% Confidence
£ Intervals on the Corresponding Number of Homes, in Millions), Based on Data from the HUD National Survey
p° (top plot) and the Interim NSLAH (bottom plot)
§ Note: The estimated exceedance percentages are calculated based on the fitted distribution (solid curve).
-------
1 Table 3-16. Estimated Percentages of 1997 U.S. Housing Exceeding Specified
2 Thresholds of Household Average Dust-Lead Loading
Dust-
Lead
Loading
Threshold
U/g/ft2)
5
10
20
30
40
50
§403 Risk Analysis - Based on Data from the
HUD National Survey (n = 284)
Based on the Fitted
Lognormal Distribution
(i.e., the curve in
Figure 3-11)
51.9%
33.6%
18.5%
12.0%
8.5%
6.4%
Based on the
Weighted Observed
Data (i.e.. the bar
chart in Figure 3-11)
51.1%
30.6%
15.6%
11.9%
9.6%
8.3%
Data from the Interim NSLAH (n = 706)
Based on the Fitted
Lognormal
Distribution (i.e., the
curve in Figure 3-11)
15.8%
7.9%
3.4%
2.0%
1.3%
0.9%
Based on the
Weighted Observed
Data (i.e.. the bar
chart in Figure 3-11)
13.2%
7.2%
4.0%
2.0%
1.4%
1.2%
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Note: Data are imputed for those surveyed units with missing data prior to calculating the above statistics (3 observations in
the HUD National Survey and 9 observations in the interim NSLAH had imputed data). The estimates based on the weighted
observed data are simple weighted percentiles that do not originate from a fitted distribution
17 whose household average floor dust-lead loadings exceeded the given threshold (i.e., information
18 from the bar charts within Figure 3-11). These results are included to evaluate the similarity
19 between the lognormal-based estimates and those generated from an approach that uses only the
20 observed data without an underlying distribution assumption. As Table 3-16 shows, the
21 lognormal-based estimates are slightly lower for the lower thresholds and slightly higher for the
22 higher thresholds, while the two approaches yield nearly equivalent estimates at the threshold of
23 30 ug/ft2. It should be noted that the lognormal-based estimates for the exceedance percentages
24 (which were also portrayed in Figure 3-11) should be used when making inferences on the
25 nation's housing stock.
26 3.2.1.4 Interpreting the Observed Differences with Other Studies. In order to make
27 proper interpretations from the results portrayed in this subsection, in particular why differences
28 exist between the studies, one must be aware of how the housing selection procedure and sample
29 collection and analysis procedures differ between the studies and can contribute to the
30 differences observed in the boxplots and tables. For the studies highlighted in the §403 risk
31 analysis report, this information was summarized in Tables 3-3a through 3-3f of that report.
32 Some of the differences among these studies that may contribute to differences in the reported
33 data are as follows:
34 • All non-control housing units in the Baltimore R&M study, approximately 88
35 percent of units selected in the HUD Grantees evaluation, and at least 84 percent
36 of the Rochester study units were built prior to 1941. In contrast, only 27 percent
37 of the housing units in the HUD National Survey were built prior to 1940.
DRAFT « DO NOT CITE OR QUOTE
88
August 28. 2000
-------
1 • The neighborhoods surveyed within the Baltimore R&M study and HUD Grantees
2 evaluation had a high prevalence of homes with lead-based paint hazards, along
3 with a history of children with elevated blood-lead concentrations and/or
4 considered at high-risk for lead poisoning.
5 • The HUD National Survey targeted occupied permanent housing throughout the
6 48 contiguous states. The units were selected via a statistically-based sampling
7 design to represent the national housing stock built prior to 1980. Excluding two
8 grantees from California, the HUD Grantees (as well as the Rochester and
9 Baltimore R&M studies) sampled housing from the Northeast and Midwest
10 census regions. Approximately two out of every five homes sampled in the HUD
11 National Survey were from the South.
12 • For the HUD Grantees evaluation, 28 percent of the homes were single-family
13 buildings (16 percent were single-family detached, and 12 percent were single-
14 family attached, or rowhouses). All homes in the R&M intervention group within
IS the Baltimore R&M study were urban rowhouses (single-family attached). Eighty
16 percent of the homes in the HUD National Survey were single-family dwellings.
17 • From 8 percent to 67 percent of the dwelling units for any one Grantee were
18 vacant prior to sampling. Out of the 5,265 dwelling units in total that were
19 enrolled as of January 1999,1524 units were vacant prior to pre-intervention
20 sampling. Overall vacancy rate was 29 percent for the Evaluation. On the other
21 hand, the HUD National Survey contained dwelling units which were permanent
22 and occupied, with the potential for containing children.
23 • The dates of environmental sampling were 11/89-3/90 for the HUD National
24 Survey, 12/93-1/99 for the HUD Grantees evaluation (pre-intervention), 8/93-
25 11/93 for the Rochester study, and 3/93-11/94 for the Baltimore R&M study.
26 Therefore, the HUD National Survey performed sampling roughly three years
27 before each of the other studies and during the late fall and winter months.
28 Section 3.1 discussed differences in approaches and methods between the HUD National Survey
29 and the NSLAH that could impact observed differences in the reported data.
30 3.2.1.5 Conclusions of the Dust-Lead Data Comparisons. The following
31 conclusions could be made upon review of the dust-lead loading summaries within Tables 3-4
32 through 3-16 and Figures 3-1 through 3-11:
33 • For both floors and window sills, the interim NSLAH data are considerably lower
34 than that reported in the §403 risk analysis (and based on the HUD National
35 Survey data), as well as for all other sources of data available to the risk analysis.
36 Household average floor dust-lead loadings had a median of less than 2.0 ug/ft2
DRAFT--DO NOT CITE OR QUOTE 89 August 28,2000
-------
1 across the interim NSLAH data, while household average window sill dust-lead
2 loadings had a median of approximately 12.0 ug/ft2. Approximately two-thirds of
3 the floor-dust samples and one-third of the window sill-dust samples had lead
4 measurements below the detection limit in the interim NSLAH. Further
5 investigation is necessary to determine the reasons for such low dust-lead loadings
6 in the interim NSLAH.
7 • Compared to the other lead exposure studies whose data were considered in the
8 §403 risk analysis (e.g., Rochester study, Baltimore R&M study, HUD Grantees
9 evaluation), geometric mean dust-lead loadings tended to be lower in the HUD
10 National Survey. However, all of these studies had similar ranges of observed
11 dust-lead loading data. This suggests that 1) the conversions to wipe-equivalent
12 dust-lead loadings performed on the HUD National Survey data in the §403 risk
13 analysis did not lead to extreme adjustments overall, and 2) there is not sufficient
14 evidence that data from the HUD National Survey are higher than what is
15 representative of the 1997 housing stock simply because it was performed some
16 years earlier.
17 • The importance of housing age is evident in the summaries within the four
18 housing age categories. Older housing is more likely to contain higher average
19 dust-lead loadings compared to newer housing. However, within an age category,
20 the summaries were quite consistent across studies (with the exception of the
21 interim NSLAH).
22 • The percentage of housing units with average floor dust-lead loadings that exceed
23 SO ug/ft2 (i.e., the proposed floor dust-lead standard) was 6.4% based on data used
24 in the §403 risk analysis, and 0.9% based on interim data from the NSLAH.
25 3.2.2 Characterizing Soil-Lead Concentrations
26 This subsection summarizes observed soil-lead concentrations in the HUD National
27 Survey and how these data were used to characterize soil-lead levels in the §403 risk analysis,
28 and compares these summaries with summaries of the interim NSLAH data (Section 3.1), as well
29 as data for 22 other studies that characterized soil-lead concentrations in urban areas prior to any
30 lead abatement. These 22 studies include the three recent studies included in the dust-lead data
31 summaries of the previous section (Baltimore R&M study, Rochester study, and HUD Grantees
32 evaluation) and other studies dating to the early 1970s (e.g., Omaha, Charleston). Sampling and
33 laboratory protocols for the 22 additional studies are summarized in Table 3-17. The soil-lead
34 data summaries for these 22 studies were either calculated directly from the available data set or
35 culled from the published scientific literature.
36 Household mass-weighted average soil-lead concentration for a specific portion of the
37 yard was the basis for the comparisons made in this section. This average was calculated by
38 weighting the result for each soil sample taken at that location by the sample's mass. If this
DRAFT -- DO NOT CITE OR QUOTE 90 August 28.2000
-------
1 Table 3-17. Information on Soil Sampling and Analysis Protocols for Studies Whose Soil-
2 Lead Data Were Compared to Results from the §403 Risk Analysis and the
3 HUD National Survey
Study
[Reference]
Soil Sampling and Analysis Details
Soil-Lead Parameter(s) Used in This Section
for Comparison to HUD National Survey
Baltimore
R&M
(USEPA.
1996el
1993-94. Three 0.5" core samples per composite, taken
from randomly determined areas along the dnpline using a 6"
stainless steel recovery probe and collected into a polysterene
liner. Samples were sieved and homogenized and digested
using SW 846-3015 and SW 846-3051. GFAA (SW
846-7421) laboratory analysis method. Only data for
occupied units were used.
Soil-lead concentration for each composite
sample (one composite sample per housing
unit, taken from the dnpline).
Baltimore
Urban Garden
Soil (Mielke et
al., 1983)
1982. Samples were from garden soil in random locations
within a 30-mile radius of downtown Baltimore. Samples
were air-dried and sieved with a 2mm stainless steel mesh
screen and digested in nitric acid. Extracts were filtered and
analyzed using a Varian atomic absorption spectrophotometer
with deuterium background correction.
Soil-lead concentration for each collected
sample.
3-City
(Baltimore,
Boston,
Cincinnati)
(USEPA,
1996a)
(Also known
as the Urban
Soil Lead
Abatement
Demonstra-
tion Project)
Only round 1 (pre-abatement) measurements were used.
Baltimore/Boston: Only results for the top 2 cm of a 15 cm
core sample were considered.
Cincinnati: Soil samples were collected within neighborhoods.
as well as within the yards of surveyed housing units.
Baltimore: Yard-wide average for a unit,
equal to the unweighted arithmetic average
of the unit's average dnpline, average mid-
yard and average boundary soil-lead
concentrations within a property (set to
missing if any of these measurements were
missing). The location averages were also
summarized.
Boston- Yard-wide average for a unit, equal
to the average across all samples
associated with that unit
Cincinnati: Yard-wide average for a unit,
equal to the unweighted arithmetic average
of average building and average play area
soil-lead concentrations for the unit (set to
missing if either of these measurements are
missing). These and other location
averages were also summarized.
Boston
Brigham and
Women
(Rabmowitz et
al., 1985)
3 samples were collected one meter apart and at least 3
meters from any road structure (preference given to obvious
play areas). These samples were composited prior to
analysis. Soil sampling occurred twice: when the resident
child of interest was 18 and 24 months of age. Laboratory
analysis method was atomic absorption spectrophotometry
(AAS).
Soil-lead concentration for each unit, equal
to the unweighted arithmetic average of
soil-lead concentrations for composite
samples taken at the 18 and 24 month
visits.
CAP Study
(USEPA,
1996d)
1990. Soil samples taken from Denver units that were
abated in 1989 during the HUD Abatement Demonstration
Study. Samples were collected from the dnpline, entryway,
and remote areas of the yard with a soil recovery probe (1*
diameter liner and 12" core sampler). At each location, a
composite sample consisted of 3 cores, each 0.5" in depth.
The sample preparation method was EPA SW846 Method
3050 (included use of nitric acid and hydrogen peroxide for
digestion). The laboratory analysis method was ICP-AES.
Soil-lead concentration for the dnpline,
entryway, and remote areas of the yard.
(One composite sample per location per
unit.)
DRAFT -- DO NOT CITE OR QUOTE
91
August 28, 2000
-------
Table 3-17 (cont.)
Study
[Reference]
Soil Sampling and Analysis Details
Soil-Lead Parameter(s) Used in This Section
for Comparison to HUD National Survey
California
(Simon et al.,
1995)
1987-91. Older units in Oakland, Los Angeles, and
Sacramento. Composite soil samples (of 4 subsamples) were
collected at each of the front, side, and rear yards. In
addition, units in Oakland and Los Angeles had a composite
soil sample collected from a secondary structure (e.g.,
garage) and a single sample collected from rain drains. All but
ain drain samples were composites. Samples were < 1" in
depth and were collected using a trowel (visible paint chips
removed first). The laboratory analysis method was AAS.
Soil-lead concentration for each collected
sample (from 3 to 5 per unit).
Cincinnati
Longitudinal
(Bornschem et
al., 1985a;
1986; Que
Hee et al.,
1985)
1980-87. Surface scrapings rather than soil cores were
taken. Laboratory analysis method was AAS. Enrolled
expectant mothers residing in areas with a history of child
residents with elevated blood-lead concentrations.
Not determined1
Cincinnati
Roadside
-------
Table 3-17 (cont.)
Study
[Reference]
Soil samples collected from perimeter and play areas at each
housing unit.
Soil Sampling and Analysis Details
Soil-Lead Parameter(s) Used in This Section
for Comparison to HUD National Survey
Milwaukee
(Pendleton)
Yard-wide average for a unit, equal to the
unweighted arithmetic average of perimeter
and play area soil-lead concentrations (set
to missing if either of these measurements
are missing).
Minneapolis
Clean-Up
(Mielke et al.,
1992)
Only pre-cleanup data were considered. Deep scrape samples
were taken at a depth of 2.5 cm, air-dried and sieved with a
2 mm stainless steel mesh screen, and digested in nitric acid.
Extracts were filtered and analyzed using a Varian atomic
absorption spectrophotometer with deuterium background
correction.
Not determined1.
Minnesota
(Schmitt et
al., 1988;
Mielke et al.,
1989)
1986-87. Only results for St. Paul and Minneapolis were
considered (except results labeled "Whole Study" also
included Duluth, Rochester, St. Cloud and rural areas).
Foundation samples were taken within 1.5 m of building.
Yard samples (front, side, and back) were taken at the
midpoint of the yard and at least 1.5 m from the foundation.
Street samples were taken within 1 5 m of a curb. Samples
were from the top 2 cm of soil. The laboratory analysis
method was ICP-AES.
Soil-lead concentration for each collected
sample.
New Orleans
(Mielke,
1995; 1993)
1983. Samples taken from residential neighborhoods within
283 census tracts in the New Orleans metropolitan area.
Foundation samples were taken within 1 m of a house.
Streetside samples were taken from within 1 m of a street.
Open area samples were from vacant lots or parks. The
laboratory analysis method was AAS with deuterium
background correction.
Soil-lead concentration for each collected
sample.
New Haven,
Connecticut
(Stark et al.,
1982)
1974-77. Samples (5-10 g) collected from homes of children
who lived at the same address for at least one year. Only the
top 0.5" of soil was analyzed.
Not determined1
Omaha
(Angle et al.,
1979)
1971-77 Soil core samples (2" depth) self-selected from
halfway between the building and lot line on four sides of the
selected units.
Yard-wide average for a unit, equal to the
arithmetic average soil-lead concentration
across all collected samples at the unit.
Rochester
Lead-in-Dust
(USHUD,
1995a;
Lanphear et
al., 1996a)
1993. Two composite samples, one from the dnpline (12
samples per composite) and one from play areas (8-10
samples per composite). Core samples were taken at a depth
of 0.5". Composites were mixed and sieved into fine and
coarse fractions and analyzed separately. Digestion method
was SW 846-3050, and the laboratory analysis method was
FAA (method 239.1).
Total soil-lead concentrations were computed as 0.25*Fine
Soil Fraction + 0.75*Coarse Soil Fraction (see Appendix E).
Yard-wide average for a unit (for both total
soil and fine soil only), equal to the
unweighted arithmetic average of the unit's
dnpline and play area soil-lead
concentrations (set to missing if either of
these measurements was missing). The
soil-lead concentration for the dnpline
sample at each unit was also summarized
(for both total soil and fine soil only).
Washington,
DC
(Elhelu et al.,
1995)
Housing units were randomly selected from each of the 8
wards of Washington, DC. Soil samples were collected from
unpaved front yards approximately 1 m from the unit and at a
depth of 15cm. Average dwelling distance from the road was
4.5 m. Fine soil samples were analyzed with a Perkin Elmer
2100 Atomic Absorption Spectrophotometer, with one result
associated with each surveyed unit.
Soil-lead concentration for each collected
sample (i.e., each housing unit).
32
Most likely soil-lead concentration for each collected sample.
DRAFT -- DO NOT CITE OR QUOTE
93
August 28, 2000
-------
1 average could not be calculated for a given study due to insufficient data, then alternative
2 statistics were calculated. For example, if mass weights were not available, the arithmetic
3 average soil-lead concentration was instead calculated.
4 When possible, a yard-wide average soil-lead concentration was calculated in a manner
5 that attempted to be consistent with the §403 risk analysis. This involved taking a weighted
6 arithmetic average of the soil-lead concentrations reported at the dripline, unit entryway, and
7 remote areas of the yard, with remote concentrations weighted twice as much as the dripline and
8 entryway concentrations. (When only one of the dripline or entryway concentrations was
9 available at a housing unit, the yard-wide average was the unweighted arithmetic average of that
10 one concentration and the remote soil-lead concentration.) Thus, the yard-wide average was
11 essentially an arithmetic average of two measures: the average soil-lead level at the dripline and
12 unit entryway (i.e., "near" the housing unit) and the soil-lead level at a remote area of the yard
13 (i.e., "far" from the housing unit). It was assumed that "play areas" represented remote areas of
14 the yard. Imputed data values replaced missing values for a housing unit in the §403 risk
15 analysis summaries, where imputation methods discussed in Section 3.3.1.1 of the §403 risk
16 analysis report were used.
17 3.2.2.1 Data Summaries for the §403 Risk Analysis Versus the Interim NSLAH.
18 Descriptive statistics of yard-wide average soil-lead concentrations as calculated in the §403 risk
19 analysis using the HUD National Survey data are presented in this subsection as they compare
20 with the same statistics calculated on interim data for 706 housing units in the NSLAH. Note
21 that these statistics reflect the sampling weights used in the §403 risk analysis and the interim
22 NSLAH sample weights, thereby allowing these summaries to be nationally representative of the
23 1997 housing stock. In addition, the interim NSLAH summaries do not include any data that
24 may have been imputed within the revised §403 risk analysis when missing data for key
25 parameters were encountered for a housing unit.
26 As in the dust-lead loading summaries (Section 3.2.1.1), the interim NSLAH summaries
27 include imputed values of yard-wide average soil-lead concentration for those housing units
28 having no reported soil-lead concentration data. As discussed in Appendix C, the imputation
29 method involved imputing values for average dripline/entryway soil-lead concentration and for
30 average mid-yard soil-lead concentration, then averaging these two imputed values together. If
31 data existed for one of the two locations but not the other, the yard-wide average for that unit
32 equaled the average soil-lead concentration at the location represented by the available data.
33 Appendix C also gives the imputed data values and how they were assigned to housing units.
34 Summaries of the interim yard-wide average soil-lead concentration data from the interim
35 NSLAH excluding any imputed data can be found in Appendix D2.
36 Also, in the same manner as the dust-lead loading summaries (Section 3.2.1.1), Appendix
37 D2 presents soil-lead concentration summaries for the interim NSLAH under five different
38 approaches (including data censoring) to handling sample results that were below the detection
39 limit. The summaries in this subsection were calculated under two of these approaches:
DRAFT -- DO NOT CITE OR QUOTE 94 August 28,2000
-------
1 • making no adjustment to not-detected data values
2 • replacing not-detected data values with one-half of the detection limit.
3 These two approaches, the same two used in the dust-lead loading data summaries in Section
4 3.2.1.1, were included together in the summary tables to illustrate the impact that any one
5 approach has on the characterized distribution of yard-wide average soil-lead concentration.
6 National comparisons
7 Table 3-18 presents descriptive statistics of yard-wide average soil-lead concentrations
8 for the 1997 national housing stock. These results indicate that only a slight downward shift in
9 the distribution of soil-lead concentrations was observed from the §403 risk analysis to the
10 interim NSLAH data, (e.g., a decline in the geometric mean from 62 ug/g to approximately 53
11 ug/g)- This decline was much smaller than that observed for dust-lead loadings.
12 Boxplots of the data distributions presented in Table 3-18 are found in Figure 3-12.
13 When not-detected data in the NSLAH were replaced by one-half of the detection limit, the
14 observed distribution of yard-wide average soil-lead concentration appears similar to what was
15 characterized in the §403 risk analysis. Appendix D2 contains tabular summaries and boxplots
16 after excluding imputed data values.
17 Table 3-18. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentrations for
18 Households, As Reported in the §403 Risk Analysis Versus the Interim
19 NSLAH Data
Study
How Not-
Detected
and
Negative
Data were
Handled
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced
by LOD/2
Yard-Wide Average Soil-Lead Concentration U/g/g)1
#
Surveyed
Units with
Positive
Averages
284
689
706
Arith-
metic
Mean
235
200
200
Geo-
metric
Mean2
61.9
53.0
52.6
Geo-
metric
Std.
Dev.2
4.46
5.09
4.73
Minimum
463
0.00
4.62
25th
Percen-
tile
21.3
16.6
16.8
Median
49.2
41.8
41.4
75th
Percen-
tile
142
158
158
Maximum
7030
9270
9270
20
21
22
23
24
25
26
27
28
29
30
31
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with soil-lead data are used to
calculate the remaining statistics).
3 Summaries include imputed data for households having no soil-lead concentration data. The method for imputation is
presented in Appendix C.
Note: The yard-wide average for a household is the average of the following two statistics: 1) the average of the mid-yard
sample results, and 2) the average of results for the driplme and entryway samples
DRAFT - DO NOT CITE OR QUOTE
95
August 28, 2000
-------
o
33
a
O
O
o
-i
m
O
O
c
o
m
100000 4
10000 -
a.
100 -
co
CD
S
5
•o
2
HUONS
(403)
NSLAH
NSLAH
(LOO/2)
IQ
C
00
M
Figure 3-12. Boxplots of Yard-Wide Average Soil-Lead Concentrations (ug/g) As Observed in the §403 Risk Analysis
(Using HUD National Survey Data) and in the Interim NSLAH (under 2 approaches to handling not-detected
values)
(Note: Boxplots include imputed data but not negative or zero values.)
-------
1 The detection limit for soil-lead concentrations in the interim NSLAH ranged from 7.2 to
2 12.4 ug/g, with a mean (and median) of 9.9 ug/g. Of those soil samples in the interim NSLAH
3 with soil-lead concentrations reported, approximately 22% (covering approximately 38% of
4 housing units reporting soil-lead concentrations) had soil results below the detection limit.
5 In addition to these data summaries that are based solely on the observed data and the
6 sampling weights, it was desired to characterize the national distribution of yardwide average
7 soil-lead concentration in such a way that the percentage of housing where this average exceeds a
8 specified threshold could be estimated. Like what was done in Section 3.2.1.3 above for floor
9 dust-lead loading, this was done for both the HUD National Survey and interim NSLAH data by
10 assuming that these data originate from a lognormal distribution. Then, the fitted distributions
11 and corresponding estimated exceedance percentages were compared between the two surveys.
12 These results are presented in Section 3.2.2.4 below.
13 Comparisons bv housing age category
14 The distribution of yard-wide average soil-lead concentrations is portrayed for each study
15 according to housing age category in Table 3-19. The importance of housing age on yard-wide
16 average soil-lead concentration is seen in both surveys, as the geometric mean and median
17 concentrations tend to increase with the age of house. The method to handling not-detected
18 values in the interim NSLAH dataset affected the data summaries only slightly, if at all.
19 Boxplots associated with the data distributions portrayed in Table 3-19 are found in
20 Figure 3-13. Appendix D2 contains tabular summaries and boxplots after excluding imputed
21 data values.
22 Comparisons bv Census region
23 The distribution of yard-wide average soil-lead concentrations is portrayed for each study
24 according to Census region in Table 3-20. Geometric mean estimates declined from the §403
25 risk analysis to the interim NSLAH data for each Census region, but the magnitude of the
26 declines were typically small. Observed median values increased from the §403 risk analysis to
27 the interim NSLAH data for the Midwest and West, but these increases were likely due to
28 random chance. No changes from the §403 risk analysis in the pattern of the yard-wide soil-lead
29 concentration distributions across Census regions were observed, with the Northeast continuing
30 to be associated with somewhat higher concentrations compared to the others (although the
31 ranges of observed soil-lead concentrations are comparable across all Census regions).
32 Boxplots associated with the data portrayed in Table 3-20 are found in Figure 3-14.
33 Appendix D2 contains tabular summaries and boxplots after excluding imputed data values.
DRAFT -- DO NOT CITE OR QUOTE 97 August 28.2000
-------
1 Table 3-19. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentration for
2 Households, Presented bv Housing Age Category. As Reported in the §403
3 Risk Analysis Versus the Interim NSLAH Data
Study
How Not-
Detected
and
Negative
Data were
Handled
Yard-Wide Average Soil-Lead Concentration 0/g/g)1
#
Surveyed
Units with
Positive
Averages
Arith-
metic
Mean
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
25*
Percen-
tile
Median
75th
Percen-
tile
Maximum
Units Built Prior to 1940
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
77
114
114
761
646
646
463
297
297
3.09
3.56
3.56
17.4
12.8
108
259
135
135
569
294
294
1030
711
711
4620
9270
9270
Units Built from 1940 - 1959
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
87
145
145
287
264
264
92.6
112
114
3.15
3.43
3.33
5.40
1 65
4.62
44.3
45.2
45.2
77.3
110
110
162
273
273
7030
4340
4340
Units Built from 1960-1977 (1960 - 1979 for the §403 risk analysis)
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
120
198
201
55.0
76.7
77.2
32.8
31.8
333
2.56
3.65
324
4.63
0.00
4.83
19.7
14.0
14.7
29.7
29.4
29.4
61.6
583
58.3
996
1120
1120
Units Built After 1977 (after 1979 for the §403 risk analysis)
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
28
168
180
31.3
27.4
28.2
22.4
15.7
16.2
2.31
3.19
2.65
5.35
0.00
4.65
136
6.07
6.34
21 2
16.0
14.9
45.0
28.7
28.7
97.4
474
475
NSLAH Units with Unspecified Year-Built Indicator
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
64
66
175
175
72.9
68.9
4.15
4.13
0.00
4.74
22.3
22.4
63.8
64.4
211
211
2290
2290
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with soil-lead data are used to
calculate the remaining statistics).
3 Summaries include imputed data for households having no soil-lead concentration data. The method for imputation is
presented in Appendix C.
Note: The yard-wide average for a household is the average of the following two statistics 1) the average of the mid-yard
sample results, and 2) the average of results for the dnpline and entryway samples
DRAFT - DO NOT CITE OR QUOTE
98
August 28, 2000
-------
D
D
0
o
Z
O
o
o
n
D
§
m
CD
CD
ro
CD
s
10000 -
1000 -
100 -
1 »<
I
! !
i i
i
i
I
T
~T
~T
T
HUONS NSLAH NSLAH
(403) (LOO/2)
Prior 10 1940
HUONS NSLAH NSL»H
(403) (LOO/2)
194O-1S39
HUONS NSLAH NSLAH
(4O3) (LOO/2)
1 »60 - 1 977(79)
HUONS NSLAH NSLAH
(*O3) (LOO/2)
Alt.r 1 977(79)
LAH NSLAH
(LOO/2)
Unknown
I" Figure 3-13. Boxplots of Yard-Wide Average Soil-Lead Concentration (ug/g), by Housing Age Category, As Observed in the
| §403 Risk Analysis (Using HUD National Survey Data) and in the Interim NSLAH (under 2 approaches to
handling not-detected values)
§ (Note: Boxplots include imputed data but not negative or zero values.)
-------
1 Table 3-20. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentration for
2 Households, Presented bv Census Region. As Reported in the §403 Risk
3 Analysis Versus the Interim NSLAH Data
Study
How Not-
Detected
and
Negative
Data were
Handled
Yard-Wide Average Soil-Lead Concentration (jig/g)1
#
Surveyed
Units with
Positive
Averages
Arith-
metic
Mean
Geo-
metric
Mean*
Geo-
metric
Std.
Dev.1
Minimum
25m
Percen-
tile
Median
75"1
Percen-
tile
Maximum
Northeast
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
53
109
109
437
423
423
206
160
162
3.58
4.24
4.16
148
3.92
6.24
60.1
52.3
52.9
279
176
176
569
396
396
4320
3460
3460
Midwest
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
§4O3 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH3
No
adjustment
Replaced by
LOD/2
73
149
150
134
258
265
52
173
182
404
220
220
125
162
163
112
682
69.0
81.4
65.5
65.8
44.5
37.3
36.4
34.4
30.5
31.7
6.33
4.97
4.71
South
2.94
4.62
4.38
West
3.92
4.36
3.55
4.63
0.00
4.90
5.22
000
4.65
479
0.00
4.62
19.7
22.1
22.1
22.6
11.9
13.1
142
12.5
12.8
51.6
63.2
63.2
40.8
276
27.9
27.2
29.4
29.4
264
206
206
79.3
79.2
79.2
61.6
775
79.3
2750
7070
7070
7030
9270
9270
2020
776
776
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with soil-lead data are used to
calculate the remaining statistics).
3 Summaries include imputed data for households having no soil-lead concentration data. The method for imputation is
presented in Appendix C.
Note: The yard-wide average for a household is the average of the following two statistics: 1) the average of the mid-yard
sample results, and 2) the average of results for the dripline and entryway samples.
DRAFT -- DO NOT CITE OR QUOTE
100
August 28, 2000
-------
o
33
3 100000 -
1
0
H ,0000-
0
3D
D
c
0 ,000-
m
0.
S
.§ 100 •
8
|
0 J
1
i -
O.I
O.Ol
i 1
: :
i
••
:
i
i
«
• •
i
•
i
• „
•
i
i
I
•
•
:
i
•
i
>
•
•
i
•
I
i 1 i
>
HUONS
(403)
NSLAH
(LOD/2)
HUONS NSLAH NSLAH
(403) (LOO/2)
uldw.it
HUONS NSLAH NSLAH
(*03) (LOO/2)
South
HUONS NSLAH NSLAH
(403) (LOO/2)
W«t
Figure 3-14. Boxplots of Yard-Wide Average Soil-Lead Concentration (pg/g), by Census Region, As Observed in the §403
Risk Analysis (Using HUD National Survey Data) and in the Interim NSLAH (under 2 approaches to handling
not-detected values)
§ (Note: Boxplots include imputed data but not negative or zero values.)
-------
1 Comparisons bv combination of housing age and Census region
2 Tables 3-21 a and 3-2 Ib present descriptive statistics for yard-wide average soil-lead
3 concentration according to the 16 combinations of Census region and housing age category.
4 Table 3-21 a reflects the data when no adjustment to not-detected results, while not-detected
5 results are replaced by one-half of the detection limit prior to performing the summaries in Table
6 3-2 Ib. As the central tendency of the soil-lead concentrations was of primary interest to compare
7 across the different combinations, these tables only contain estimates of the arithmetic and
8 geometric means, geometric standard deviation, and median.
9 Due to the small numbers of housing units entering into each summary within Tables
10 3-2la and 3-2Ib, caution must be taken when making inferences from the results portrayed in
11 these tables. Appendix D2 contains these tabular summaries after excluding imputed data
12 values.
13 3.2.2.2 Data Summaries for the §403 Risk Analysis Versus Other Studies. This
14 subsection presents data summaries for the 22 studies in Table 3-17 that characterized soil-lead
IS concentrations in urban areas and how these summaries compare to that for the HUD National
16 Survey and to the distribution of yard-wide average soil-lead concentration characterized in the
17 §403 risk analysis. The soil-lead concentration parameters that are summarized in this
18 subsection were specified for each study in Table 3-17.
19 The 22 studies whose data are considered in this subsection include the three recent
20 studies included in the dust-lead data summaries in Section 3.2.1: Baltimore R&M study (pre-
21 intervention), Rochester Lead-in-Dust study, and HUD Grantees evaluation (pre-intervention
22 data available through 1/99). Figure 3-15 contains boxplots of household average soil-lead
23 concentration for these three studies and the HUD National Survey ("HUDNS"). These boxplots
24 represent yard-wide averages in all cases except the Baltimore R&M study, where only dripline
25 soil samples were collected. Separate boxplots are included for each grantee in the HUD
26 Grantees evaluation10.
27 As in Figures 3-7 through 3-10, the left-most three boxplots in Figure 3-15 represent
28 yard-wide average soil-lead concentration data from the HUD National Survey:
29 • "HUDNS (U)" summarizes the data without regard to sampling weights
30 • "HUDNS (403)" summarizes the data as used in the §403 risk analysis (e.g., using
31 sampling weights reflecting the 1997 housing stock; incorporating imputed data
32 assigned to housing units with missing data)
10 "Alam"=Alameda County; "Balt"=Baltimore; "Bos"=Boston; "CA"=Califomia, "Cle"=Cleveland;
"MA"=Massachusetts; "MN"=Minnesota; "NJ"=New Jersey, "RI"=Rhode Island, "WI"=Wisconsm; "Milw"=Milwaukee,
"Chic"=Chicago; "NYC"=New York City, "VT"=Vermont.
DRAFT » DO NOT CITE OR QUOTE 102 August 28, 2000
-------
1
2
3
4
6
Table 3-21 a. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentrations for
Households, Presented bv Housing Age and Census Region. As Reported in
the §403 Risk Analysis Versus the Interim NSLAH Data Where No
Adjustments Were Made to Not-Detected Results
Census
Region
Northeast
Midwest
South
West
Study2
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Housing Age
Category
Prior to 1940
1940- 1959
1960-1977
1960-79 for §403)
After 1 977
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
( 1979 for §403)
Prior to 1 940
1940 - 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Yard-Wide Average Soil-Lead Concentration1 (pg/g)
# Surveyed
Units
26
41
17
23
10
21
16
19
36
21
36
29
37
4
29
19
26
33
48
64
79
18
81
13
11
16
38
17
61
6
42
Arithmetic
Mean
542
877
573
290
79.1
132
57.8
1310
498
127
236
42.7
93.8
13.0
34.1
417
684
327
364
54.6
68.7
38.5
22.2
594
155
96.8
143
56.2
47.4
21.7
17.3
Geometric
Mean3
491
499
136
199
60.7
65.5
40.0
941
224
92.6
110
27.1
38.3
11.5
12.9
174
278
83.1
96.6
36.5
269
29.7
157
295
122
72.1
86.9
23.8
24.7
15.0
106
Geometric
Std. Dev.3
3.22
4.40
2.24
2.15
2.95
2.63
3.34
241
3.14
2.32
3.34
1.66
3.92
3.74
3.27
4.40
2.30
360
2.11
2.45
2.23
• 2.19
3.08
3.02
3.81
2.34
3.54
Median
569
60.1
273
69.7
50.9
38.8
238
123
82.0
23.4
34.6
12.4
9.36
186
81.0
77.9
34.7
26.1
25.0
15.0
158
60.4
90.3
20.0
26.9
13.6
9.53
10
11
12
13
14
15
16
17
All statistics are calculated by weighting each household by its sampling weight.
2 Summaries include imputed data for households having no soil-lead concentration data. The method for imputation is
presented in Appendix C.
3 Only household averages greater than zero are used to calculate this value (data for all units with soil-lead data are used to
calculate the remaining statistics).
Note: The yard-wide average for a household is the average of the following two statistics: 1) the average of the mid-yard
sample results, and 2) the average of results for the driplme and entryway samples.
DRAFT -- DO NOT CITE OR QUOTE
103
August 28, 2000
-------
1
2
3
4
5
6
Table 3-21 b. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentrations for
Households, Presented bv Housing Age and Census Region. As Reported in
the §403 Risk Analysis Versus the Interim NSLAH Data Where Not-Detected
Results Were Replaced bv LOD/2
Census
Region
Northeast
Midwest
South
West
Study2
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Housing Age
Category
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
Prior to 1940
194O- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Yard-Wide Average Soil-Lead Concentration1 (pg/g)
# Surveyed
Units
26
41
17
23
10
21
16
19
36
21
36
29
37
4
30
19
26
33
48
64
81
18
84
13
11
16
38
17
62
6
50
Arithmetic
Mean
542
877
573
290
79.1
132
58.1
1310
498
127
236
427
94.1
13.0
34.7
417
684
327
364
54.6
69.4
38.5
22.7
594
155
96.8
143
56.2
48.0
21.7
18.9
Geometric
Mean
491
497
136
199
60.7
65.4
42.0
941
224
92.6
111
27.1
39.0
11.5
14.0
174
278
83.1
97 7
36.5
27.8
29.7
15.4
295
122
72.1
89.8
23.8
27.8
15.0
12.1
Geometric
Std. Dev.
1.57
3.26
4.40
2.24
2 15
2.96
2.36
2.68
3.34
2.41
3.11
2.32
3.27
1.66
3.06
3.68
3.74
3.27
4 34
2.30
3.24
2 11
2.29
3.76
2.21
2.19
2.78
3.02
2.91
2.34
2.42
Median
444
569
60.1
273
69.7
50.9
388
1390
238
123
820
23.4
34.6
124
9.67
159
186
81.0
77.9
34.7
26.1
25.0
14.7
394
158
60.4
90.3
20.0
26.9
13.6
11.2
9
10
11
12
13
14
15
' All statistics are calculated by weighting each household by its sampling weight.
1 Summaries include imputed data for households having no soil-lead concentration data. The method for imputation is
presented in Appendix C.
Note: The yard-wide average for a household is the average of the following two statistics: 1) the average of the mid-yard
sample results, and 2) the average of results for the driplme and entryway samples.
DRAFT -- DO NOT CITE OR QUOTE
104
August 28, 2000
-------
o
o
I
o
m
O
O
O
m
o
en
: : : I
i ! !
7S?
CM MA MM HJ •
HUD OrontMt
at. o-k me
~ Figure 3-15. Boxplots of Household Average Soil-Lead Concentrations (//g/g) for Houses in the HUD National Survey,
-°° Baltimore R&M Study, Rochester Lead-in-Dust Study, and Grantees Within the HUD Grantees Evaluation
o (Note: Household averages represent yard-wide averages except for the Baltimore R&M study, where only dripline soil samples were collected. See text
for definitions of labels along the horizontal axis.)
-------
1 • "HUDNS (OW)" summarizes the data weighted according to the original weights
2 assigned in the survey.
3 As soil samples were sieved into fine and coarse fractions in the Rochester study, Figure 3-15
4 includes two boxplots for the Rochester soil-lead data. The boxplot labeled "Rochester (FINE)"
5 summarizes household average soil-lead concentration considering only the fine-sieved fraction
6 of the collected soil samples. The boxplot labeled "Rochester (TOTAL)" summarizes estimated
7 household average soil-lead concentration assuming the total soil sample was analyzed. Total
8 soil-lead concentration for each sample was estimated as the average of the reported
9 concentrations for the fine and coarse fractions of the sample, with the coarse fraction result
10 weighted three times that of the fine sample result (see Appendix E for the derivation of this
11 estimate using data from the Milwaukee study). Estimating soil-lead concentration in the total
12 soil sample was intended to allow soil-lead data from the Rochester study to be more comparable
13 to data from the other studies in which no sieve-fractions were calculated.
14 Figure 3-15 shows that while the ranges of average soil-lead concentrations among the
15 study households tended to overlap from study to study, the distributions based upon the HUD
16 National Survey data (including the §403 risk analysis) tended to be shifted lower than for the
17 other studies.
18 Figure 3-16 contains a graphical presentation of how the distribution of household
19 average soil-lead concentration in other selected studies listed in Table 3-17 compare with the
20 distributions based upon the HUD National Survey data (i.e., the same three distributions
21 portrayed in the boxplots labeled "HUDNS" in Figure 3-15). The studies selected for Figure
22 3-15 were among those in which an average soil-lead concentration for a particular area could be
23 determined. As only summary statistics for many of the studies in Table 3-17 were available
24 from the references or prior literature reviews, boxplots like those in Figure 3-15 could not be
25 created for these other studies. Instead, specific descriptive statistics (when cited in the
26 references) are plotted in Figure 3-16 for each study by using plotting symbols that indicate the
27 type of statistic. These statistics, with their plotting symbols following in parentheses, are the
28 minimum (MIN), 25Ih percentile (25th), median (50th), 75lh percentile (75th), maximum (MAX),
29 and geometric mean (GM) soil-lead concentrations. In studies where the arithmetic mean is
30 specified instead of the geometric mean, the arithmetic mean (AVE) was plotted. The vertical
31 dashed line in Figure 3-16 separates results based on the HUD National Survey data from the
32 results for the other studies.
33 Yard-wide average soil-lead concentration (or an average that is not specific to a given
34 location) were available or could be calculated within eight of these studies (one being the 3-
35 Cities study, which consisted of three sub-studies). Table 3-22a presents values of descriptive
36 statistics (e.g., geometric mean, minimum, maximum, selected percentiles) for yard-wide average
37 soil-lead concentration within these studies. This table also includes the estimated number of
38 averages represented in the descriptive statistics that exceed a given soil-lead concentration
39 threshold (400, 1200, 2000, and 5000 ug/g). The following features can be found within this
40 table:
DRAFT -- DO NOT CITE OR QUOTE 106 August 28, 2000
-------
O 100000 -
0 :
1
O
H
nj I0000 .
DO
s :
§
10OO -
"i"
r>
O.
I/I
C
g too -
s ;
O 0>
-v| U
O
•a 10 -
0
J_
'5
i -
0 1 -
10
C
-* 0 01 -
»
to
MAX MAX MAX
75th
75th 75lh
GM GM
SOth SOth GM SOth
25th 25th 25th
MIN MIN MIN
MAX
MAX
MAX
MAX
75th
SOth GM
MAX 25lh 7S)h
7Slh "IN 75th
75lh SOth
SOth r>u ^EIU
^ '5th emL. <«u
both GM
25th
25th 2Sm
MIN SOth
25th
MIN
MIN
WIN
HUD NS HUD NS HUD NS Baltimore Baltimore Boston Boston Charleston
(U) (403) (OW) 3-Clty Urban Garden 3-City Bngham
§ Figure 3-16. Summary Statistics of Average Household Soil-Lead Concentrations (//g/g) for Selected Studies as Compared to
Summaries Based on Data from the HUD National Survey
-------
iooooo -
o
o
o
m
O
33
O
§
m
10000 -
1000 -
E
Q.
Q.
^^
I/I
C
O
C
01
o
o
•o
o
_!_
'5
100 •
10 :
0 I
£
1
ro
oo
ro
MAX
75lh
CM
SOIh
25th
MIN
75lh
SOIh CM
25th
MIN
75th
CM
SOIh
25lh
U'N
MAX
CM
SVC
75lh
SOIh CM
25th
MIN
som GM
25lh
MIN
MIN
MIN
HUDNS
(U)
HUDNS
(403)
HUDNS
(OW)
Cincinnati
3-Cily
Cincinnati
Longitudinal
Corpus
Chnsti
HUD Abatement
Demonstration
Milwaukee
Figure 3-16. (cont.)
-------
1
2
o
o
o
m
O
3J
3
m
Table 3-22a. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentrations. According to Study and Within Specific
Subsets of the Sampled Housing Within a Study
HUD National Survey
(unweighted) '
HUD National Survey
(§403 RA) »
or Type of
All Units
Big City/ Metro
Small City/
West
All Units
Small City/
City
Yard-Wide Average Soil-Lead Concentration U/g/g)
N
250
67
60
42
24
57
109
141
31
63
111
45
284
96
73
50
28
65
146
166
53
73
134
Geometric
Mean
740
806
720
1120
65 1
535
91 5
628
144.5
750
606
750
61 9
87.5
562
779
462
46.9
833
505
2058
81.4
445
344
Geometric
Std. Dav.
40
33
36
62
36
3.8
43
37
3.7
46
35
44
45
42
38
5 5
34
4.9
4.7
42
36
63
29
3.9
Minimum
46
5.4
48
52
67
46
5.2
46
14.8
46
52
4.8
46
54
48
52
67
4.6
52
46
14.8
46
5.2
48
25th
Percentile
27.2
35.3
304
26.0
286
21 7
32.0
259
538
21.7
263
259
265
43.0
286
245
260
17 7
346
234
773
21 1
254
243
Median
59.9
70.9
606
93.9
50.8
45.2
773
538
115.2
589
52 7
60.4
61 8
945
61 1
83.3
492
359
887
51 9
2789
61.1
49.2
49.7
75th
Percentile
159.7
171 0
5856
1184
1252
2137
1359
3576
162.3
115.1
197.9
2036
313.3
161 8
391 9
929
1050
313.3
1233
6279
216.7
97.4
191.8
Maximum
1463
2019
7025
4318
2002
7025
4318
4318
2752
7025
2019
1463
2019
7025
4318
2002
7025
4318
4318
2752
7025
201 S
Soil-Lead Concentration i
400/ig/g
10.4%
100%
262%
8.3%
70%
165%
85%
226%
12.7%
72%
15.6%
17.9%
102%
194%
49%
11 0%
185%
9 5%
348%
21 2%
38%
63%
1200/ig/g
4.5%
1 7%
11.9%
42%
3.5%
73%
28%
6.5%
95%
1.8%
4.4%
2.5%
11 7%
3.8%
6.8%
64%
36%
4.4%
14.9%
08%
1.9%
2000 uglg
00%
7 1%
42%
2.8%
2.2%
00%
3.8%
34%
23%
26%
1.0%
6000 pg/g
0.0%
00%
0.0%
0.0%
O
(O
«Q
a
-------
o
3)
Table 3-22a. (cent.)
Study
HUD National Survey
(original weights)3
Baltimore Urban
Garden
Baltimore 3-City
Boston 3-City
Boston Bngham and
Women
Cincinnati 3-City
HUD Grantees
Evaluation
Subset of Units
or Type of
Soil/cover
All Units
Big City/ Metro
Big City/ Suburb
Small City/ Metro
Small City/
Suburb
Non-Metro
City
Non-City
Northeast
Midwest
South
West
Top 2 cm
Top 2 cm
Overall
Full Grass
> 1/2 Grass
< 1/2 Grass
All Bare
All Grantees
Alameda
California
Cleveland
Minnesota
Rhode Island
Wisconsin
Yard-Wide Average Soil-Lead Concentration (pg/g)
N
250
67
60
42
24
57
109
141
31
63
111
45
422
181 •
101 *
195
7»
4'
6«
4«
3"
314
58
8
99
41
40
38
Geometric
Mean
781
740
69.1
121 9
607
775
90.9
705
1502
1133
579
473
NA
4423
24309
360.8
133.1
1389
1157
1496
1036
857.5
6698
3418
16207
563.1
1146.0
3183
Geometric
Std. Dev.
4.5
33
39
65
3.6
51
46
43
3.6
60
33
40
NA
1 7
1 6
33
1.9
24
1 5
23
1.6
38
2 5
2 1
22
24
33
11 8
Minimum
46
54
4.8
52
67
46
52
46
148
4.6
52
48
1 0
1037
7443
70
558
499
71.5
568
608
00
39.5
580
3150
495
355
00
25th
Percentile
272
353
30.4
260
28.6
21 7
320
25.9
53.8
21 7
263
25.9
24 54
308.4
16780
1930
864
739
778
91 6
608
479.0
3525
3575
940.0
339.5
6085
3160
Median
599
70.9
606
939
508
45.2
773
538
1152
589
527
604
1000
4793
2380.0
3740
1129
144.9
126.3
142.1
118 1
9205
5885
4158
15450
591 5
12278
5368
75th
Percentile
145 1
1597
171.0
5856
1184
1252
213.7
135.9
357.6
162.3
115.1
1979
421.0'
6884
36000
796.0
2573
2943
151 6
3001
154.7
17300
13480
512 5
2840.0
857 5
2875.0
9170
Maximum
7025
1463
2019
7025
4318
2002
7025
4318
4318
2752
7025
2019
10900
1793
7070
13237
285
397
182
442
155
15535
12648
560
14180
4800
15535
3852
Percentage of Homes with Yard-Wide Average
Soil-Lead Concentration 2
400pg/g
15.3%
126%
11.3%
29 1%
7.3%
164%
19.4%
126%
275%
24.1%
6.6%
90%
NA
59 7%
1000%
49.2%
00%
00%
0.0%
250%
00%
828%
70.7%
625%
990%
707%
87 5%
71 1%
1200/ig/g
6.8%
2.3%
1 1%
18 2%
5.6%
10.2%
8.8%
54%
41%
19.0%
1.3%
27%
NA
2.8%
93.1%
138%
0.0%
0.0%
00%
0.0%
00%
40.1%
34 5%
0.0%
61.6%
17.1%
52.5%
158%
2000 (jglg
37%
0.0%
1.1%
10.1%
56%
5.1%
4.1%
34%
35%
8.9%
1.3%
1.4%
NA
00%
65.3%
67%
00%
0.0%
0.0%
0.0%
0.0%
21 0%
5.2%
0.0%
384%
7.3%
325%
7.9%
5000 //g/g
0.3%
0.0%
0.0%
1.9%
00%
0.0%
0.8%
0.0%
0.0%
0.0%
0.9%
0.0%
NA
0.0%
6.9%
1.0%
0.0%
0.0%
0.0%
0.0%
0.0%
4 1%
1 7%
0.0%
8.1%
00%
7.5%
0.0%
O
o
I
o
m
O
3)
O
I
m
IQ
ro
GO
s
8
-------
o
5
O
o
1
o
m
Table 3-22a. (cont.)
Omaha
Rochester
Subset of Units
or Type of
Urban
Commercial
Yard-Wide Average Soil-Lead Concentration (jig/g)
N
10
20
92
69
56
51
82
Geometric
Mean
1530.5
7079
16405
262.0
3390
81.0
8802
Geometric
Std. Dev.
20
30
2 1
NA
NA
NA
3.5
Minimum
7660
385
4490
530
200
160
467
25th
Percentile
8290
393.8
9035
NA
NA
NA
487 5
Median
11848
850.9
16055
NA
NA
NA
8078
6268
75th
Percentile
1868.8
24720
NA
NA
NA
17366
11505
Maximum
3695
1615
4792
341
10721
Soil-Lead Concentration 2
400 fig/g
750%
NA
NA
NA
793%
1200//g/9
35.0%
NA
NA
NA
2000f/g/g
150%
NA
NA
6000 //g/g
00%
NA
3.7%
1 s
o
21
5
6
10
11
NA = Not Available
Mass-weighted arithmetic average soil-lead concentration as reported in the HUD National Survey, summarized without weighting by sample weights
Mass-weighted arithmetic average soil-lead concentration as calculated in Chapter 3 of the 403 risk analysis, summarized by weighting each average to reflect the 1997 U S. housing
Mass-weighted arithmetic average soil-lead concentration as reported in the HUD National Survey, summarized by weighting with the National Survey sample weights.
20'" percentile
An initial unweighted arithmetic average of soil lead levels at the specified locations was taken prior to calculation of statistics within this table. The number in this column represents
the number of properties, not necessarily the number of houses
to
CD
-------
1 • For the HUD National Survey data, which are associated with sampling weights
2 from the original survey and revised sampling weights for the §403 risk analyses,
3 all results in Table 3-22a are portrayed three times: under each of these two sets
4 of weights as well as without regard to weights.
5 • For the HUD National Survey data, which are associated with sampling weights
6 from the original survey and revised sampling weights for the §403 risk analyses,
7 all results in Table 3-22a are portrayed three times: under each of these two sets
8 of weights as well as without regard to weights.
9 • In addition to summarizing results across all housing units or samples in a study,
10 results for selected studies are also summarized for specific subsets of housing
11 units, soil types, or soil samples. In particular, HUD National Survey results are
12 portrayed according to urbanicity and Census region, results from the HUD
13 Grantees evaluation are portrayed by grantee, and the Rochester study results are
14 portrayed for the fine soil fraction as well as for total soil.
15 Refer to Table 3-17 to verify the types of results being summarized in Table 3-22a (i.e., housing
16 unit averages versus averages for single analytical samples).
17 Table 3-22b contains the same descriptive statistics as those portrayed in Table 3-22a, but
18 they represent average soil-lead concentration for specific locations, such as dripline, play areas,
19 remote areas, geographical areas, and other locations that were considered within the individual
20 studies. As in Table 3-22a, the statistics in Table 3-22b are given over the entire study, as well as
21 for specified sets of units that are determined by urbanicity and other factors.
22 Summary statistics bv housing age category
23 As housing age category is generally regarded as an important influence on soil-lead
24 concentrations, the above summaries are also presented according to the housing age categories
25 considered in the HUD National Survey (pre-1940, 1940-1959, 1960-1979, post-1979). Figure
26 3-17 presents boxplots for pre-1980 housing data from the HUD National Survey and the
27 Rochester Lead-in-Dust study (total soil and fine soil), non-control houses in the Baltimore R&M
28 study, and pre-1978 data from the HUD Grantees evaluation (data combined across grantees).
29 As all non-control units in the Baltimore R&M study were built prior to 1941, the only boxplot
30 for this study in Figure 3-17 appears in the "pre-1940" category. Caution must be taken when
31 interpreting results in Figure 3-17 for the Rochester study, as the actual age of certain houses may
32 be older than what was specified in the Rochester study database (see Section 3.3.1.3 of the §403
33 risk analysis report).
34 Many of the other studies listed in Table 3-17 did not have information readily available
35 on housing age. Thus, no corresponding figure portraying distributions according to housing age
36 was prepared to represent these other studies.
DRAFT -- DO NOT CITE OR QUOTE 112 August 28. 2000
-------
1
2
a
o
1
o
m
O
33
O
O
Table 3-22b. Descriptive Statistics of Average Soil-Lead Concentrations in Specific Yard Areas and/or for Certain Subsets of
the Sampled Housing Within a Study
Study
HUD National
Survey
(unweighted) '
HUD National
Survey (403
RA)2
Yard Area/
Subset of Housing Units
Drlplme/
Entryway
Dnplme/
Entryway
All Units
Big City/ Metro
Big City/ Suburb
Small City/ Metro
Small City/ Suburb
Non-Metro
City
Non-City
Northeast
Midwest
South
West
All Units
Big City/ Metro
Big City/ Suburb
Small City/ Metro
Small City/ Suburb
Non-Metro
City
Non-City
Northeast
Midwest
South
West
Average Soli-Lead Concentration (pg/g)
N
263
76
62
44
24
57
120
143
38
66
113
46
312
96
73
50
28
65
146
166
53
73
134
52
Geometric
Mean
913
101 B
895
1469
835
579
1164
744
2125
952
701
818
727
1024
680
962
606
487
997
585
2510
945
518
403
Geometric
Std. Dev.
43
37
40
67
33
38
47
38
39
48
38
43
46
44
39
60
29
49
50
41
36
63
33
37
Minimum
52
79
75
52
105
56
52
56
206
58
52
75
52
79
75
52
105
56
52
56
206
58
52
75
25th
Percentile
324
402
291
318
361
260
378
294
752
317
318
266
31.8
442
289
276
354
203
373
275
853
284
294
247
Median
752
815
820
1272
71 1
455
856
674
2095
731
625
652
767
919
787
1247
700
376
961
619
3732
619
594
566
75th
Percentile
2373
2269
2633
6956
1485
131 9
2940
1788
5347
2616
1260
2633
2457
4260
2420
5892
1082
112.9
4392
1319
1007.0
2830
1233
2164
Maximum
13596
2571
1661
13596
1684
3999
13596
3999
2571
5336
13596
1149
13596
2571
1661
13596
1684
3999
13596
3999
2571
5336
13596
1149
Percentag of Homes with Soil-Lead
Concentration *
400 ug/g
164%
171%
194%
318%
83%
35%
225%
112%
368%
182%
80%
174%
155%
00%
167%
222%
49%
68%
225%
107%
485%
206%
43%
68%
1200 ug/g
4.9%
53%
16%
114%
42%
35%
75%
2.8%
132%
61%
35%
0.0%
40%
17%
05%
87%
38%
68%
46%
35%
51%
109%
17%
00%
2000 ug/g
30%
39%
00%
68%
00%
35%
50%
14%
79%
45%
18%
00%
29%
12%
00%
65%
00%
68%
34%
26%
18%
104%
08%
00%
5000 ug/g
1.1%
00%
00%
68%
00%
00%
25%
00%
00%
1.5%
1.8%
00%
1 1%
00%
00%
65%
00%
0.0%
27%
00%
00%
35%
08%
00%
w
(Q
I
ro
oo
-------
Table 3-22b. (cont.)
Study
HUD National
Survey (HUD
NS weights)3
Baltimore R&M
Baltimore 3-
City
CAP Study
California
Yard Area/
Subset of Housing Units
Dnpline/
Entryway
Driplme
Midyard
Remote
Dnpline
Entryway
Remote
All Units
Big City/ Metro
Big City/ Suburb
Small City/ Metro
Small City/ Suburb
Non-Metro
City
Non-City
Northeast
Midwest
South
West
Dnpline
Top 2 cm
Top 2 cm
Top 2 cm
All Homes
Unabated Homes
Abated Homes
All Homes
Unabated Homes
Abated Homes
All Homes
Unabated Homes
Abated Homes
Oakland
Los Angeles
Sacramento
Average Soil-Lead Concentration (ug/g)
N
263
76
62
44
24
57
120
143
38
66
113
46
28
196 '"
183 10
197 '"
117'
37'
80'
109'
37'
72'
120'
39'
81'
292
327
227
Geometric
Mean
928
902
861
1474
752
820
1099
825
1959
1320
669
532
4445
6359
2870
3370
1821
913
2506
1435
1013
1716
1204
853
1422
8970
1880
2340
Geometric
Std. Dev.
46
37
41
70
31
51
50
43
3.5
60
36
39
51
20
19
17
27
26
23
27
33
23
21
22
19
NA
NA
NA
Minimum
52
79
75
52
105
56
52
56
206
58
52
75
289
960
310
772
110
110
514
46
46
429
150
150
287
560
300
260
25th
Percentile
324
402
29.1
318
361
260
378
294
752
317
318
266
71.5
3902
1990
2300
978
558
1226
729
474
880
703
534
791
NA
NA
NA
Median
752
815
820
1272
71 1
455
856
674
2095
731
625
65.2
6869
6666
2860
3518
1904
1120
2571
148.5
1295
1528
1200
879
1502
8800
190.0
2290
75th
Percentile
237.3
2269
263.3
6956
1485
1319
2940
1788
5347
2616
1260
2633
17675
10356
4250
4656
331 8
137.0
4124
2652
215.8
3434
1976
1313
2288
NA
NA
NA
Maximum
13596
2571
1661
13596
1684
3999
13596
3999
2571
5336
13596
1149
3539
4400
2500
1850
3351
1016
3351
1068
655
1068
1073
1073
615
88176
1973
2664
Percentag of Homes with Soil-Lead
Concentration 2
400M9/9
185%
167%
225%
332%
73%
10.2%
23.3%
15.1%
42.0%
244%
72%
9.6%
607%
724%
290%
355%
20.5%
5.4%
27.5%
174%
135%
19.4%
5.0%
26%
6.2%
NA
NA
NA
1200 ug/g
57%
23%
07%
129%
56%
10.2%
66%
52%
6.5%
136%
2.8%
0.0%
429%
189%
11%
1.0%
34%
0.0%
50%
0.0%
00%
0.0%
00%
00%
0.0%
NA
NA
NA
2000 ug/g
4.3%
16%
0.0%
97%
0.0%
10.2%
49%
3.9%
22%
13.0%
1.3%
0.0%
10.7%
26%
0.5%
00%
0.9%
0.0%
13%
0.0%
0.0%
0.0%
00%
00%
0.0%
NA
NA
NA
5000 ug/g
1 6%
00%
0.0%
97%
00%
00%
39%
0.0%
00%
43%
13%
0.0%
00%
00%
0.0%
0.0%
0.0%
0.0%
00%
0.0%
00%
0.0%
0.0%
0.0%
0.0%
NA
NA
NA
O
o
1
o
m
O
3)
O
m
-------
Table 3-22b. (cont.)
Study
Cincinnati 3-
Cily
Cincinnati
Longitudinal
Cincinnati
Roadside
Yard Area/
Subset of Housing Units
Building
Bare Areas
Play Area
Other
Play Area
and/or
Entryway
Pavement
Edge
Overall
Full Grass
>1/2 Grass
All Bare
Overall
Full Grass
>1/2 Grass
<1/2 Grass
All Bare
Overall
Full Grass
>1/2 Grass
All Bare
Overall
All Bare
All Units
19th Century
<1950
>1960
Average Soil-Lead Concentration (M9/9)
N
100 ">
22 ,o
35 10
56'°
46 10
7410
312
8'°
8"
63 '"
11'°
5'°
7"
7,o
8»
g,o
3,o
2»
6"
410
BO
14
18
7
13
60"
164
Geometric
Mean
2339
2867
3035
1842
410.1
2209
3402
1067
301
2726
946
1221
863
1298
600
1078
1728
2227
959
654
13603
5720
8040
25400
26700
12560*
7520"
Nf
Geometric
Std. Dev.
46
57
43
48
33
55
171
29
44
47
19
16
1 4
1 8
23
40
1.6
1 3
51
19
47
NA
NA
NA
NA
12543'
5574'
NA
Minimum
71
169
62
71
350
54
129
423
78
54
200
693
558
620
182
58
1226
1804
58
357
760
NA
NA
NA
NA
NA
NA
9C
25th
Percentile
91 3
554
1020
576
2144
771
129
521
97
111 1
707
BO 2
683
726
321
61 3
1226
1804
597
428
NA
NA
NA
NA
NA
NA
173 C
Median
2609
2492
2909
2199
2969
2239
14913
776
210
2567
1034
1291
820
1554
757
1375
1394
2277
1044
557
NA
NA
NA
NA
NA
NA
585 C
75th
Percentile
15540
11336
5698
11307
8002
20473
1531
701
8616
1554
1643
1234
2114
982
2481
3017
2749
2756
1141
NA
NA
NA
NA
NA
Maximum
4533
4963
4897
7602
4552
2047
1128
609
4552
192
230
124
299
192
743
302
275
743
167
NA
NA
NA
NA
NA
Concentration 2
400 ug/g
455%
343%
339%
435%
392%
667%
125%
125%
413%
00%
00%
00%
00%
00%
11 1%
00%
00%
167%
00%
NA
NA
NA
NA
NA
1200 ug/g
273%
22.9%
107%
239%
176%
667%
00%
00%
175%
00%
00%
00%
00%
00%
00%
00%
00%
00%
NA
NA
NA
NA
2000 ug/g
136%
86%
130%
333%
00%
127%
00%
00%
00%
00%
NA
NA
NA
NA
5000 ug/g
00%
00%
22%
00%
00%
00%
00%
NA
NA
NA
o
o
1
Q
m
O
3)
O
§
m
cn
IV)
GO
-------
o
3J
O
O
1
O
m
O
DO
Table 3-22b. (cent.)
Study
Corpus ChnstI
1-880 (Alameda
County)
HUD
Abatement
Demonstration
Study
HUD Grantees
Evaluation
Maine Urban
Milwaukee
Minneapolis
Clean-up
Yard Area/
Subset of Housing Units
All Samples
Parks
Schools
All Others
East
West
Dnplme
Dnplme - Post-Abatement
Dnplme
All Grantees
Alameda
California
Cleveland
Minnesota
Rhode Island
Wisconsin
Milwaukee
Vermont
Homes
Parks and Playgrounds
Perimeter
Play Area
Minneapolis
St Paul
Foundation
Mid-yard
Street
Foundation
Mid-yard
Street
Average Soil-Lead Concentration (ug/g)
N
485'
94 7
12'
379'
1167
22 7
455'
455'
557
97
8
99
44
60
66
12
171
75
25
93
92
12
12
10
10
10
10
Geometric
Mean
2080'
550*
570'
2500*
5943"
2633*
7550"
8675'
11820
7763
3307
23805
5932
13924
5637
19742
15365
12750
2050
23437
6263
NA
NA
NA
NA
NA
NA
Geometric
Std. Dev.
NA
NA
NA
NA
NA
NA
NA
NA
37
27
19
23
27
33
73
25
33
NA
NA
22
23
NA
NA
NA
NA
NA
NA
Minimum
80
80
110
80
223
897
NA
NA
01
300
940
4200
450
660
01
3270
250
500
500
5870
1300
340
60
960
220
440
330
25th
Percentlle
NA
NA
NA
NA
NA
NA
NA
NA
5570
3950
2700
13500
2800
6395
4000
11665
6920
NA
NA
124BO
3780
1840
550
1380
1780
700
1060
Median
NA
NA
NA
NA
NA
NA
NA
NA
12520
7100
3600
21400
5595
15000
8590
16900
15600
NA
NA
19900
5560
7950
2720
2550
5610
1080
1530
75th
Percentlle
NA
NA
NA
NA
NA
NA
NA
NA
25800
13870
4600
45170
11505
27790
15000
34960
33800
NA
NA
36550
8600
12650
4110
2820
9800
2840
2820
Maximum
2969
318
258
2969
3187
862
NA
NA
52700
21131
780
16380
8120
26159
5733
10300
52700
10900
700
29335
9459
2240
680
373
2960
414
470
Percentag of Homes with Soil-Lead
Concentration 2
400 ug/g
NA
NA
NA
NA
NA
NA
NA
NA
837%
732%
375%
1000%
591%
883%
773%
917%
889%
50.0%'
8.0%'
1000%
73.9%
NA
NA
NA
NA
NA
NA
1200 ug/g
NA
NA
NA
NA
NA
NA
NA
NA
531%
320%
00%
808%
250%
600%
379%
750%
608%
370%*
00%G
77.2%
207%
NA
NA
NA
NA
NA
NA
2000 ug/g
NA
NA
NA
NA
NA
NA
NA
NA
321%
13.4%
00%
545%
9.1%
367%
167%
333%
41.5%
NA
0.0%
500%
98%
NA
NA
NA
NA
NA
NA
5000 ug/g
NA
NA
NA
NA
NA
NA
NA
NA
113%
4.1%
00%
20.2%
45%
150%
3.0%
167%
140%
NA
00%
18.3%
33%
NA
NA
NA
NA
NA
NA
o>
B
o
3
-------
a
3)
5
o
o
1
o
o
OJ
O
Table 3-22b. (cont.)
Study
Minnesota
New Haven,
Connecticut
New Orleans
Rochester
Study
Yard Area/
Subset of Housing Units
Entire Study. All Samples
St Paul
Minneapolis
Foundation
Garden
Backyard
Front yard
Side yard
Street side
Open
Play area
Foundation
Garden
Backyard
Front yard
Side yard
Street side
Open
Play area
Near (near (he house)
Far (near the street)
Inner City
Mid-City
Suburban
Foundation
Streetside
Open Area
Foundation
Streetside
Open Area
Foundation
Streetside
Open Area
Dnplme
Driplme ((me soil only)
Average Soli-Lead Concentration (ug/g)
N
2454 7
127'
8'
114'
108'
46'
170'
95'
164 7
199'
28'
61'
131'
170'
1197
51'
139'
260
260
201'
723'
74'
220'
765'
80'
332'
1195'
114'
185
185
Geometric
Mean
NA
4720
1740
1190
900
960
1130
660
240
6650
2530
2120
1730
1770
1860
390
220
7129
5970
NA
NA
NA
NA
NA
NA
NA
NA
NA
9926
7320
Geometric
Std. Oev.
NA
45
62
38
30
61
22
37
70
35
3.2
33
21
22
26
37
69
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
42
37
Minimum
NA
30
140
10
10
10
60
20
10
350
340
40
180
270
30
10
10
NA
NA
80
40
100
10
10
20
20
20
40
178
123
25th
Percentile
NA
2680
400
650
440
420
640
260
40
3050
109.0
110.0
1070
1060
1080
240
40
NA
NA
2490
1420
760
320
300
160
180
400
140
5458
4120
Median
NA
5760
1470
1610
1040
1330
1270
760
360
6890
2640
2470
1850
1650
2230
340
330
NA
NA
8400
3420
2120
1100
1100
400
500
860
280
11175
9590
75th
Percentile
NA
12460
11770
3000
1920
3640
204.0
1770
1040
14960
4450
5200
2890
2970
3380
730
1100
NA
NA
23700
6200
4600
4460
2460
980
1540
1710
780
23802
16480
Maximum
NA
7994
2846
1386
1377
2385
575
1466
607
20136
3858
1210
1345
1326
1876
878
788
NA
NA
69000
9450
10600
24400
6340
3960
5650
2150
540
110834
21049
Percentag of Homes with Soil-Lead
Concentration 2
400 ug/g
11 0%5
70 7% •
NA
210%'
83%'
NA
NA
5660'
NA
314%'
250%'
NA
NA
NA
NA
NA
555%'
NA
NA
21 2%'
NA
NA
92%'
NA
795%
762%
1200 ug/g
50%'
263%
NA
25%
0.0%
NA
NA
326%
NA
08%
06%
NA
NA
NA
NA
NA
91%
NA
NA
29%
NA
NA
03%
NA
47.6%
384%
2000 |ig/g
20%
NA
NA
NA
00%
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
314%
184%
5000 pg/g
NA
NA
NA
NA
00%
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
114%
38%
IQ
I
P
ro
o
3
-------
Table 3-22b. (cont.)
Study
Washington.
D.C
Yard Area/
Subset of Housing Units
Front yard
Wardl
Ward 2
Ward3
Ward 4
Wards
Ward6
Ward?
Average Soil-Lead Concentration (ug/g)
N
30
30
30
30
30
30
30
Geometric
Mean
NA
NA
NA
NA
NA
NA
NA
Geometric
Std. Dev.
NA
NA
NA
NA
NA
NA
NA
Minimum
364
483
102
327
120
138
362
25th
Percentlle
2280
3448
251
955
1013
1250
703
Median
4442
4714
537
1989
221.9
2604
144.4
75th
Percentlle
1145.0
9750
1057
294.9
3804
4279
2749
Maximum
4905
4520
815
4575
5056
1720
3740
Percenlag of Homes with Soil-Lead
Concentration t
400 ug/g
NA
NA
NA
NA
NA
NA
NA
1200 ug/g
NA
NA
NA
NA
NA
NA
NA
2000 ug/9
NA
NA
NA
NA
NA
NA
NA
5000 ug/g
NA
NA
NA
NA
NA
NA
NA
00
NA = Not Available
1 Mass-weighted anthmetic average soil-lead concentration as reported in the HUD National Survey, summarized without weighting by sample weights
2 Mass-weighted anthmetic average soil-lead concentration as calculated in Chapter 3 of the 403 risk analysis, summarized by weighting each average to reflect the 1997 U S housing stock
3 Mass-weighted anthmetic average soil-lead concentration as reported in the HUD National Survey, summanzed by weighting with the National Survey sample weights
4 Percent of samples that exceed 300 ppm
5 Percent ot samples that exceed 500 ppm
6 Percent of samples that exceed 1000 ppm
7 Number of samples (multiple samples taken at many sites)
8 Arithmetic Mean or SD
9 60 houses total, reference used did not provide number of houses by house age
10 An Initial unweighted anthmetic average of soil lead levels at the specified locations was taken prior to calculation of statistics within this table The number in this column represents the number
of properties, not necessarily the number of houses
-------
o
33
o
O
1
o
m
O
O
O
m
100000 •
10000
1000 •
J"1
J 100
1
i
1
•g '«
1 -
O.I
0.01 •
O
° P
a
Q
O O O
o o o
0 0 0 ,1| 1
J
000
0 O 0
0
0
o o o
o o o o
O O 0 n
o 9 o o o o
If!
T
i
n 1 1 1 1 1 1 1 1 1 i i i i i i i i
HUM man HUM ••» bet M nw HUM HUM HUMI ho •«* HUO HUM HUM HUM •«* too.
(u) (401) (no (rai«o (r««) u* MM (u) (401) (w) (iot«o (rut) onM (u) ( Baltimore R&M Study, Rochester Lead-in-Dust Study, and HUD Grantees Evaluation, by Housing Age Category
I (pre-1979 only)
ro
oo
ro
o
o
(Note: Data for the Baltimore R&M study are dripline results. See text for definitions of labels along the horizontal axis. Caution must be taken when categorizing
nouses in the Rochester study by age of the house.)
-------
1 The summary statistics found in Tables 3-22a and 3-22b were calculated according to
2 housing age category for relevant studies. These summaries are found in Tables 3-23a (for yard-
3 wide average soil-lead concentration) and 3-23b (for average soil-lead concentration for specific
4 locations). Note that these tables also include summary statistics for housing units built after
5 1979 (although the Rochester study units may not have actually been built in this time period, as
6 mentioned in Section 3.2.1.2). The post-1979 results labeled as "HUD National Survey (§403
7 RA)" represent surveyed homes built from 1960-1979 that contain no lead-based paint (Section
8 3.3.1.5 of the §403 risk analysis report).
9 3.2.2.3 Calculating National Exceedance Percentages for Yardwide Average Soil-
10 Lead Concentration. The soil-lead data summaries presented above suggest that the distribution
11 of measured soil-lead concentrations as reported for the HUD National Survey are reasonably
12 consistent with the distributions suggested by other studies, including the interim NSLAH data.
13 Thus, these two national surveys are expected to generate similar national distributions for
14 yardwide average soil-lead concentration, from which the estimated percentages of housing units
15 whose yardwide average soil-lead concentrations exceed specified thresholds ("exceedance
16 percentages") could be calculated. These percentages give some indication of the frequency with
17 which intervention activities might be prompted by regulations that target alleviating soil-lead
18 exposure. Soil abatement practices are often recommended both within the literature and by the
19 HUD "Guidelines for the Evaluation and Control of Lead-Based Paint Hazards in Housing"
20 (USHUD, 1995b; pages 12-47 to 12-56).
21 The methods detailed in Section 3.2.1.3, which were used to fit lognormal distributions to
22 household average floor dust-lead loadings based on data from the two national surveys, were
23 also used to fit lognormal distributions to yardwide average soil-lead concentration data from
24 these two surveys. As discussed in Section 3.2.1.3, the key objective to fitting the lognormal
25 distribution was to use the distribution to estimate exceedance percentages for specified soil-lead
26 concentration thresholds. Therefore, in order to ensure that the upper tail of the distribution was
27 as accurately portrayed as possible within the fitted distribution, this method treated a certain
28 percentage of the lowest data values as censored data when fitting the distribution. In this
29 exercise, four thresholds were of interest for yardwide average soil-lead concentration: 400,
30 1200, 2000, and 5000 ppm.
31 Figure 3-18 contains plots of the fitted lognormal distributions (superimposed on bar
32 charts of the observed data) and the estimated exceedance probabilities corresponding to these
33 distributions, for residential yard-wide average soil-lead concentrations, based on the HUD
34 National Survey data (top plot) and the interim NSLAH data (bottom plot). Recall that the
35 sampling weights corresponding to the HUD National Survey data were revised in the §403 risk
36 analysis to reflect the 1997 national housing stock. The same soil-lead concentration (horizontal)
37 axis is used for both plots, so that the two plots can be directly compared. The similarity of the
38 two distributions is noted in this plot, as the fitted distributions are nearly the same shape and
39 cover approximately the same ranges of data. Furthermore, the estimated exceedance
40 percentages for a given threshold differ by less than one percentage point between the two
DRAFT -- DO NOT CITE OR QUOTE 120 August 28, 2000
-------
1
2
3
4
o
o
1
o
m
O
B
O
I
Table 3-23a. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentrations. According to Study and Housing Aue
Category and Within Specific Subsets of the Sampled Housing Within a Study
Study
Subset of Units or
Type of Soil/Cover
Yard-Wide Average Soil-Lead Concentration (ug/g)
N
Geometric
Mean
Geometric
Std. Dev.
Minimum
25th
Percentile
Median
75th
Percentile
Maximum
Percentage of Homes with Yard-Wide Average
Soli-Lead Concentration 2
400 ug/g
1200 ug/g
2000 ug/g
5000 ug/g
Houses Built Prior to 1940
HUD National
Survey
(unweighted)
HUD National
Survey (403
RA)2
Rochester
Lead-in-Dust
All
City Only
Northeast
Midwest
South
West
Northeast • City Only
Midwest -City Only
South -Cily Only
West - City Only
All
City Only
Northeast
Midwest
South
West
Northeast -City Only
Midwest -City Only
South -City Only
West -City Only
Total Soil
Fine Soil Fraction
SB
29
11
15
19
13
3
6
13
7
77
45
26
19
19
13
16
9
13
7
75
75
2983
3756
4538
4769
1663
2862
5451
7942
2214
4497
4627
5094
4907
9407
1736
2955
5254
14051
2238
4344
10186
7493
38
40
19
39
43
40
24
37
49
28
31
3.0
16
27
37
38
16
21
40
28
33
2.6
174
174
1367
498
174
259
2558
804
174
842
174
174
1367
498
174
259
2558
804
174
842
917
1029
1094
1102
2896
109.4
474
1127
2558
3715
596
1377
137.7
2588
2896
1623
474
1127
2789
6419
596
1377
5086
4385
3465
5339
4436
6791
1252
3935
4436
14278
2581
5856
3935
6134
5541
8347
1252
3935
6421
8407
2581
5856
9119
6860
8053
11592
6279
14970
6134
7112
14271
14970
7175
11592
8407
8407
8407
14630
6134
7112
8407
14630
7175
11592
19718
12050
4619
4619
1427
2752
4619
2019
1427
2752
4619
1244
4619
4619
1427
2752
4619
2019
1427
2752
4619
1244
55617
10721
43.1%
552%
545%
53.3%
263%
46.2%
66.7%
667%
385%
714%
592%
67.4%
647%
773%
277%
455%
720%
924%
398%
669%
813%
827%
172%
241%
91%
400%
53%
154%
333%
667%
7.7%
143%
196%
207%
08%
541%
29%
180%
1 5%
778%
42%
166%
360%
253%
69%
69%
00%
13.3%
53%
77%
00%
167%
77%
00%
94%
86%
00%
252%
29%
90%
00%
340%
42%
00%
240%
133%
0.0%
00%
0.0%
00%
00%
0.0%
00%
00%
00%
00%
00%
0.0%
00%
00%
00%
00%
00%
00%
00%
00%
107%
4.0%
ro
I?
li
(O
-------
o
3)
Table 3-23a. (cont.)
Study
Subset of Units or
Type ol Soil/Cover
Yard-Wide Average Soil-Lead Concentration (ug/g)
N
Geometric
Mean
Geometric
Std. Dev.
Minimum
25th
Percentlle
Median
75th
Percentlle
Maximum
Percentage ol Homes with Yard-Wide Average
Soil-Lead Concentration Z
400 ug/g
1200 ug/g
2000 ug/g
5000 ug/g
Houses Built Prior to 1940 (cont.)
HUD National
Survey (HUD
NS weights)1
HUD
Grantees
Evaluation
All
City Only
Northeast
Midwest
South
West
Northeast -City Only
Midwest - City Only
South -City Only
West - City Only
All Grantees
Alameda
California
Cleveland
Minnesota
Rhode Island
Wisconsin
Milwaukee
Vermont
58
29
11
15
19
13
3
6
13
7
181
39
7
64
18
11
28
6
a
4339
4787
4333
9554
1736
2955
4449
15596
2238
4344
7572
6509
3251
16291
442.8
8351
3034
10859
3959
34
36
15
28
37
38
14
21
40
28
42
27
22
22
26
36
11 9
15
36
174
174
1367
498
174
259
2558
804
174
842
00
395
580
4310
495
65.0
00
7660
385
1094
1102
2896
1094
474
1127
2558
3715
596
1377
4310
3180
3250
9220
2650
2815
2603
8040
1765
3465
5339
4436
6791
1252
3935
443.6
14278
258.1
5856
8350
5820
4050
14300
471.8
12055
5560
1005.0
503.8
8053
11592
6279
14970
6134
7112
14271
14970
717.5
11592
14550
13480
5400
28575
790.0
14960
9085
11885
1033.9
4619
4619
1427
2752
4619
2019
1427
2752
4619
1244
14180
12648
560
14180
4492
5648
3852
2288
2078
561%
675%
625%
751%
277%
45.5%
908%
91.1%
398%
669%
790%
692%
571%
1000%
61 1%
727%
679%
1000%
500%
242%
29.9%
1.4%
'593%
29%
180%
46%
91 1%
42%
166%
35.4%
33.3%
00%
594%
56%
545%
107%
16.7%
25.0%
11 6%
12.5%
00%
27.7%
29%
90%
0.0%
398%
4.2%
0.0%
18.8%
7.7%
0.0%
37.5%
56%
18.2%
71%
16.7%
125%
0.0%
0.0%
00%
0.0%
0.0%
0.0%
0.0%
00%
0.0%
3.9%
2.6%
0.0%
7.8%
0.0%
91%
00%
00%
00%
Houses Built From 1940-1959
HUD National
Survey
(unweighted)
All
City Only
Northeast
Midwest
South
West
Northeast • City Only
Midwest - City Only
South -City Only
West - City Only
77
37
10
19
33
15
4
3
20
10
831
810
969
852
841
707
96.4
1043
915
550
29
32
42
24
33
22
18
31
41
21
54
54
337
93
54
249
639
363
54
249
443
435
524
524
435
346
706
363
462
260
758
773
620
905
810
60.4
773
905
945
497
1416
1296
77.3
1453
135.1
1455
1514
3459
1395
1085
7025
7025
4318
346
7025
214
225
346
7025
214
39%
27%
100%
00%
61%
00%
00%
00%
50%
00%
2.6%
27%
100%
00%
3.0%
0.0%
00%
00%
50%
00%
2.6%
27%
100%
0.0%
30%
0.0%
00%
00%
5.0%
0.0%
1.3%
2.7%
00%
00%
30%
00%
00%
0.0%
5.0%
00%
§
11
m
O
0)
ro
ro
a
ro
-------
o
3J
O
o
O
m
O
3D
O
Q
Table 3-23a. (cont.)
Study
Subset of Units or
Type of Soil/Cover
Yard-Wide Average Soil-Lead Concentration (ug/g)
N
Geometric
Mean
Geometric
Std. Dev.
Minimum
25th
Percentile
Median
75th
Percentile
Maximum
Percentage of Homes with Yard-Wide Average
Soil-Lead Concentration 2
400 ug/g
1200 ug/g
2000 ug/g
5000 ug/g
Houses Built From 1940-1959 (cont.)
HUD National
Survey (403
RA)»
HUD National
Survey (HUD
NS weights)1
HUD
Grantees
Evaluation
Rochester
Lead-in-Dust '
All
City Only
Northeast
Midwest
West
Northeast • City Only
Midwest - City Only
South • City Only
West • City Only
All
City Only
Northeast
Midwest
South
West
Northeast - City Only
Midwest -City Only
South -City Only
West - City Only
All Grantees
Alameda
Rhode Island
Wisconsin
Vermont
Total Soil
B7
46
17
21
33
16
10
5
20
11
77
37
10
19
33
15
4
3
20
10
11
2
5
3
1
5
5
926
101 5
1364
926
831
721
2558
1406
918
563
839
808
1028
857
831
699
1388
104.3
91.8
531
4920
10590
4092
4048
4790
1668
1863
32
34
44
24
33
22
23
24
41
21
31
32
45
24
33
22
18
2.5
41
20
30
21
50
12
26
26
54
54
337
93
54
249
639
363
54
249
54
54
337
93
54
249
639
36.3
54
24.9
355
6320
355
3160
4790
467
51 1
476
490
538
589
435
394
773
905
462
260
443
435
524
524
435
34.6
706
363
462
260
3280
6320
3280
3160
4790
113.9
1040
814
1037
773
1233
810
709
2694
2167
945
518
758
773
620
905
810
604
773
905
945
497
4790
12032
5095
4500
4790
1806
1985
1707
2182
3133
1820
1351
1717
3133
3459
1395
1296
141 6
1296
773
1453
1351
1455
1514
3459
1395
1085
6400
17745
6400
4665
4790
2128
4585
7025
7025
4318
372
7025
220
1412
372
7025
220
7025
7025
4318
346
7025
214
225
346
7025
214
3024
1774
3024
466
479
632
465
52%
47%
143%
00%
61%
00%
126%
00%
57%
00%
44%
28%
123%
00%
61%
00%
00%
00%
57%
00%
727%
1000%
600%
667%
1000%
200%
400%
43%
47%
143%
00%
33%
0.0%
126%
00%
57%
0.0%
34%
28%
123%
00%
33%
00%
00%
00%
57%
00%
182%
500%
200%
00%
00%
00%
00%
23%
93%
33%
00%
00%
00%
57%
0.0%
34%
28%
123%
00%
33%
00%
00%
0.0%
57%
00%
9 1%
00%
200%
00%
00%
00%
00%
23%
33%
00%
00%
5.7%
00%
1 2%
00%
00%
57%
00%
00%
0.0%
00%
00%
00%
00%
ro
w
ro
GO
-------
Table 3-23a. (cent.)
Study
Subset of Units or
Type of Soil/Cover
Yard-Wide Average Soil-Lead Concentration (ug/g)
N
Geometric
Mean
Geometric
Std. Dev.
Minimum
25th
Percentile
Median
75th
Percentile
Maximum
Percentage of Homes with Yard-Wide Average
Soli-Lead Concentration I
400 ug/g
1200 ug/g
2000 ug/g
5000 ug/g
Houses Built From 1960-1979 (1977 for HUD Grantees Evaluation)
HUD National
Survey
(unwe\ghted)
HUD National
Survey (403
RA)!
HUD National
Survey (HUD
NS weights)'
All
City Only
Northeast
Midwest
South
West
Northeast -City Only
Midwest - City Only
South -City Only
West -City Only
All
City Only
Northeast
Midwest
South
West
Northeast - City Only
Midwest -City Only
South - City Only
West -City Only
All
City Only
Northeast
Midwest
South
West
Northeast - City Only
Midwest -City Only
South -City Only
West - City Only
115
43
10
29
59
17
3
6
25
9
120
46
10
29
64
17
3
6
28
9
115
43
10
29
59
17
3
6
25
9
339
392
61.2
265
364
284
987
196
48.5
254
328
362
607
271
365
238
1150
201
488
235
324
35.9
607
271
35.7
238
1150
201
493
235
26
27
22
24
25
37
22
1 4
27
26
26
24
22
23
23
30
19
1 4
22
21
26
25
22
23
23
30
19
14
23
21
46
52
148
46
52
48
425
137
52
54
46
52
148
46
52
48
425
137
52
54
46
52
148
46
52
48
425
137
52
54
200
211
41 1
171
226
142
425
138
264
188
204
213
41 1
171
230
142
425
138
268
188
200
21 1
41 1
171
226
142
425
138
264
188
301
333
622
23.4
321
237
1152
206
39.4
237
31 5
348
622
23.4
351
237
1152
206
427
237
301
333
622
234
321
237
1152
206
39.4
237
583
79.3
1152
392
664
395
1962
21 1
816
318
625
685
1152
392
649
395
1962
21.1
804
318
583
793
1152
392
664
395
1962
21 1
816
318
996
996
196
355
996
604
196
33
996
186
996
996
196
355
996
604
196
33
996
186
996
996
196
355
996
604
196
33
996
186
1 7%
23%
00%
0.0%
17%
5.9%
00%
00%
40%
00%
1 2%
0.9%
00%
00%
08%
37%
00%
0.0%
20%
00%
12%
09%
0.0%
0.0%
0.8%
3.7%
00%
0.0%
22%
00%
00%
0.0%
0.0%
0.0%
0.0%
00%
00%
0.0%
0.0%
00%
00%
00%
00%
00%
00%
0.0%
00%
00%
00%
00%
00%
00%
0.0%
00%
0.0%
0.0%
0.0%
00%
00%
0.0%
0.0%
0.0%
0.0%
00%
00%
00%
00%
00%
00%
00%
00%
00%
0.0%
0.0%
0.0%
00%
00%
0.0%
0.0%
00%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
00%
00%
00%
0.0%
0.0%
0.0%
00%
00%
0.0%
00%
00%
0.0%
0.0%
0.0%
00%
0.0%
00%
00%
00%
0.0%
0.0%
0.0%
0.0%
00%
0.0%
0.0%
00%
00%
00%
00%
00%
8
1
o
m
O
3)
O
m
»
ro
-------
o
0)
Table 3-23a. (cont.)
Study
Subset of Units or
Type of Soil/Cover
Yard-Wide Average Soil-Lead Concentration (ug/g)
N
Geometric
Mean
Geometric
Std. Dev.
Minimum
25th
Percentile
Median
75th
Percentile
Maximum
Percentage of Homes with Yard-Wide Average
Soil-Lead Concentration 2
400 ug/g
1200 ug/g
2000 ug/g
5000 ug/g
Houses Built From 1960-1979 (1977 for HUD Grantees Evaluation) (cont.)
Rochester
Lead-in-Dust *
Total Soil
Fine Soil Fraction
1
1
1062
1245
1062
1245
1062
1245
1062
1245
1062
1245
106
124
00%
00%
00%
00%
00%
00%
00%
00%
Houses Built After 1979 (1977 for HUD Grantees Evaluation)
HUD National
Survey (403
RA)'
HUD
Grantees
Evaluation
Rochester
Lead-in-Dust 4
All
City Only
Midwest
South
West
Midwest - City Only
South - City Only
West - City Only
Minnesota
Total Soil
Fine Soil Fraction
28
9
4
18
6
1
5
3
1
1
1
224
248
115
297
150
204
385
128
4055
5219
5455
23
23
17
21
23
10
20
20
54
54
67
56
54
204
210
54
4055
521 9
5455
136
204
70
210
62
204
21 3
54
4055
5219
5455
212
213
124
250
136
204
245
130
4055
5219
5455
450
297
190
583
297
204
793
297
4055
521 9
5455
97
97
20
97
62
20
97
30
406
522
546
00%
00%
00%
00%
00%
00%
00%
00%
1000%
1000%
1000%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
o
Fl
o
3
o
I
ro
Ol
1 Mass-weighted arithmetic average soil-lead concentration as reported in the HUD National Survey, summarized without weighting by sample weights
2 Mass-weighted arithmetic average soil-lead concentration as calculated in Chapter 3 of the 403 risk analysis, summarized by weighting each average to reflect the 1997 U S housing stock
3 Mass-weighted arithmetic average soil-lead concentration as reported in the HUD National Survey, summarized by weighting with the National Survey sample weights
4 Some houses in this housing age category may belong to an earlier age category, as some houses may have actually been built earlier than the year specified within the study's database
-------
1
2
3
4
o
3)
§
O
m
O
3J
O
I
m
Table 3-23b. Descriptive Statistics of Average Soil-Lead Concentrations in Specific Yard Areas and/or for Certain Subsets of
the Sampled Housing Within a Study, Presented bv Housing Age Category
Study
Location
Average Soli-Lead Concentration (ug/g)
N
Geometric
Mean
Geometric
Std. Dev.
Minimum
25th
Percentlle
Median
75th
Percentlle
Maximum
Percentage of Homes with Average Soil-Lead
Concentration 2
400M9/9
1200 M9/9
2000 M9/9
5000 M9/9
Houses Built Prior to 1940
HUD National
Survey
(unweighted)
HUD National
Survey (403
RA)*
Dnpline/
Entryway
Dnpline/
Entryway
All
City Only
Northeast
Midwest
South
West
Northeast -City Only
Midwest -City Only
South -City Only
West -City Only
All
City Only
Northeast
Midwest
South
West
Northeast - City Only
Midwest -City Only
South - City Only
West - City Only
64
33
15
17
19
13
6
7
13
7
77
45
26
19
19
13
16
9
13
7
3472
4262
5569
5159
2075
2544
5093
7020
2994
4280
5290
6227
6644
9247
2159
241 4
7436
13871
301.3
3750
40
40
20
40
50
41
24
38
57
31
34
33
16
36
43
38
16
30
46
29
197
23.1
2118
347
231
197
2118
937
231
665
197
231
2118
347
231
197
2118
937
231
665
1279
1829
3779
2598
494
881
2419
2616
590
1238
1840
2616
4530
2598
494
881
4530
4248
590
1238
4182
504.3
5237
4248
1829
4236
4585
7020
2963
6528
4664
7008
6224
7020
1829
4236
9137
11145
2963
6528
9876
11065
8877
11655
8429
9323
7008
19328
11065
10382
11265
11265
11265
11655
8429
9323
11265
11265
11065
10382
8960
8960
2334
5336
8960
1149
2334
5336
8960
1149
8960
2334
5336
8960
1149
2334
5336
8960
1149
516%
606%
73.3%
529%
316%
538%
667%
71 4%
46.2%
714%
78.7%
936%
647%
338%
502%
941%
924%
487%
669%
152%
13.3%
235%
105%
0.0%
167%
286%
15.4%
00%
127%
24%
397%
91%
00%
1.5%
389%
131%
00%
91%
6.7%
17.6%
5.3%
00%
16.7%
143%
7.7%
0.0%
92%
0.8%
37.9%
29%
00%
15%
340%
42%
0.0%
6.1%
00%
5.9%
5.3%
00%
00%
143%
00%
86%
00%
126%
29%
00%
00%
34.0%
4.2%
00%
ro
O)
I?
(Q
I
ro
CO
ro
-------
Table 3-23b. (cont.)
Study
Location
Average Soil-Lead Concentration (ug/g)
N
Geometric
Mean
Geometric
Std. Dev.
Minimum
25th
Percentlle
Median
75th
Percentlle
Maximum
Percentage of Homes with Average Soil-Lead
Concentration t
400 ug/g
1200 ug/g
2000 ug/g
5000 ug/g
Houses Built Prior to 1940 (cont.)
HUD National
Survey (HUD
NS weights)3
Baltimore
R&M
California
HUD
Grantees
Evaluation
New Haven,
Connecticut
Rochester
Lead-in-Dust
Drlplme/
Entryway
All
City Only
Northeast
Midwest
South
West
Northeast -City Only
Midwest -City Only
South - City Only
West -City Only
Dnpline
Oakland, LA, Sacramento
Oakland
Dnpline
All Grantees
Alameda
California
Cleveland
Minnesota
Rhode Island
Wisconsin
Milwaukee
Vermont
Near (near the house)
Far (near the street)
Dnpline
Dnpline (fine soil only)
64
33
15
17
19
13
6
7
13
7
26
377
174
266
57
7
64
21
13
44
6
54
112
112
158
158
4818
5458
5620
9178
2159
2414
572.4
14188
301 3
3750
4445
NA
NA
10259
7333
321 9
24917
4557
1251 5
5049
12247
13558
12525
8165
13298
9378
37
38
15
36
43
38
16
31
46
29
51
NA
NA
39
28
19
23
28
30
76
22
32
NA
NA
35
32
197
231
2118
347
231
197
2118
937
231
665
289
NA
NA
01
300
940
5400
450
1120
0.1
3270
280
NA
NA
291
123
1279
182.9
3779
259.8
494
881
2419
261.6
590
1238
71 5
NA
NA
5340
3700
2500
13700
2750
7440
3535
10700
6450
NA
NA
6688
6400
4182
5043
5237
4248
1829
4236
4585
7020
2963
6528
6869
NA
NA
10775
652.0
3500
21500
4000
15110
7965
11665
14250
NA
NA
13220
10765
987.6
11065
6877
11655
842.9
9323
7008
19328
11065
10382
17675
NA
NA
21500
13170
5200
46380
7700
24010
12050
19100
31800
NA
NA
27550
18160
8960
8960
2334
5336
8960
1149
2334
5336
8960
1149
3539
NA
NA
50600
21131
780
16380
8120
10209
5733
3727
50600
NA
NA
110834
21049
62.9%
717%
904%
633%
338%
50.2%
868%
91.6%
487%
669%
607%
660%'
900%'
808%
719%
286%
1000%
524%
846%
750%
833%
889%
NA
NA
880%
854%
17.5%
168%
36%
412%
9.1%
00%
33%
431%
131%
00%
429%
NA
NA
462%
28.1%
00%
797%
95%
615%
295%
500%
556%
NA
NA
551%
443%
147%
12 2%
12%
393%
29%
00%
33%
377%
42%
00%
107%
NA
NA
293%
140%
00%
59.4%
48%
308%
9.1%
167%
407%
NA
NA
367%
21.5%
52%
114%
00%
131%
2.9%
00%
00%
377%
42%
00%
00%
NA
NA
86%
35%
00%
219%
48%
77%
45%
00%
56%
NA
NA
133%
44%
8
O
O
0)
O
O
ro
14
H
I
I
ro
-------
Table 3-23b. (cont.)
Study
Location
Average Soil-Lead Concentration (ug/g)
N
Geometric
Mean
Geometric
Std. Dev.
Minimum
25th
Percentlle
Median
75th
Percentlle
Maximum
Percentage of Homes with Average Soil-Lead
Concentration 2
400 ug/g
1200 ug/g
2000 ug/g
5000 ug/g
Houses Built From 1940-1959
HUD National
Survey
(unweighted)
HUD National
Survey (403
RA)»
HUD National
Survey (HUD
NS weights) '
Dnpline/
Entryway
Driplme/
Entryway
Dnplme/
Entryway
All
City Only
Northeast
Midwest
South
West
Northeast -City Only
Midwest • City Only
South - City Only
West -City Only
All
City Only
Northeast
Midwest
South
West
Northeast -City Only
Midwest -City Only
South -City Only
West -City Only
All
City Only
Northeast
Midwest
South
West
Northeast -City Only
Midwest -City Only
South - City Only
West -Cily Only
82
42
13
20
33
16
7
4
20
11
87
46
17
21
33
16
10
5
20
11
82
42
13
20
33
16
7
4
20
11
1077
1168
1704
1218
979
772
2568
2185
1140
589
1086
1193
152.7
1253
968
757
2944
2240
1140
552
1030
1057
1349
1187
968
757
2552
1934
1140
552
32
41
46
26
34
25
51
33
44
25
31
38
36
25
33
25
28
27
44
24
31
39
38
25
33
25
35
29
44
24
8.0
80
359
116
80
95
735
513
80
95
80
80
359
11.6
80
95
735
513
80
95
80
80
359
116
80
95
735
513
80
95
558
558
735
61 9
455
431
837
913
539
318
558
57.4
735
619
455
43.1
837
131 4
539
318
558
558
735
619
455
431
837
913
539
318
896
877
837
1167
89.6
710
880
311 1
942
574
900
942
880
131 4
896
710
3096
3732
942
574
896
877
837
1167
896
710
880
311 1
942
574
1794
1875
2460
244.9
1453
178.7
25705
590.0
1753
177.9
2189
2460
3732
2490
1453
178.7
3732
4907
1753
1779
1794
187.5
2460
2449
145.3
1787
25705
5900
1753
1779
13596
13596
2571
689
13596
284
2571
689
13596
188
13596
13596
2571
689
13596
284
2571
689
13596
188
13596
13596
2571
689
13596
284
2571
689
13596
188
85%
143%
231%
10.0%
61%
0.0%
286%
50.0%
100%
00%
65%
96%
14.3%
70%
49%
00%
126%
33.0%
86%
00%
69%
106%
172%
74%
49%
00%
20.2%
425%
86%
00%
49%
71%
231%
00%
3.0%
00%
286%
0.0%
50%
0.0%
43%
4.7%
143%
00%
33%
00%
126%
00%
57%
00%
45%
5.2%
172%
00%
33%
00%
202%
00%
5.7%
00%
3.7%
71%
15.4%
0.0%
3.0%
0.0%
28.6%
0.0%
5.0%
0.0%
22%
4.7%
5.0%
0.0%
3.3%
0.0%
12.6%
0.0%
5.7%
0.0%
2.3%
5.2%
6.0%
00%
33%
00%
20.2%
0.0%
57%
00%
12%
2.4%
0.0%
0.0%
30%
0.0%
0.0%
00%
5.0%
0.0%
1 1%
23%
0.0%
0.0%
33%
0.0%
00%
00%
57%
00%
11%
2.6%
00%
0.0%
3.3%
0.0%
00%
00%
57%
00%
O
o
I
o
m
O
3J
m
GO
to
09
ro
-------
Table 3-23b. (cont.)
Study
California
HUD
Grantees
Evaluation
New Haven,
Connecticut
Rochester
Lead-in-Dust '
Location
Average Soil-Lead Concentration (ug/g)
N
Oakland. LA, Sacramento
Oakland
Dnpline
All Grantees
Alameda
Rhode Island
Wisconsin
Vermont
Near (near the house)
Far (near the street)
Dnpline
Dnpline (fine soil only)
163
17
17
4
6
5
2
115
115
13
13
Geometric
Mean
NA
NA
4780
4843
5091
5165
317.5
5349
5002
2826
2765
Geometric
Std. Dev.
NA
NA
32
22
48
27
66
NA
NA
35
34
Minimum
25th
Percentile
Median
Houses Built From 1940-1959 (cont.)
NA
NA
660
1740
660
1390
84.0
NA
NA
275
297
NA
NA
1740
2730
1400
400.0
840
NA
NA
1705
1460
NA
NA
5300
6455
5360
5160
642.0
NA
NA
2590
2720
75th
Percentile
Maximum
NA
NA
9250
9220
12170
5930
12000
NA
NA
8430
8510
NA
NA
5389
925
5389
2160
1200
NA
NA
1790
1788
Percentage of Homes with Average Soil-Lead
Concentration 2
400 ug/g
196%'
706%'
647%
500%
667%
80.0%
50.0%
NA
NA
38.5%
308%
1200 ug/g
NA
NA
235%
00%
333%
200%
500%
NA
NA
7.7%
77%
2000 ug/g
NA
NA
11 8%
00%
167%
200%
00%
NA
NA
00%
0.0%
5000 ug/g
NA
NA
59%
00%
167%
00%
00%
NA
NA
00%
0.0%
Houses Built From 1960-1979 (1977 for HUD Grantees and New Haven)
HUD National
Survey
(unweighted)
Dnpline/
Entryway
All
City Only
Northeast
Midwest
South
West
Northeast - City Only
Midwest - City Only
South -City Only
West - City Only
117
45
10
29
61
17
3
6
27
9
391
448
668
298
417
362
101 1
213
561
287
28
29
21
25
28
40
25
1.7
28
29
52
52
206
58
52
75
353
101
52
79
213
217
353
194
270
159
353
158
282
203
336
373
731
284
359
266
141.2
204
495
266
700
79.3
1186
399
723
365
2072
355
917
343
1713
1713
207
685
1713
910
207
40
1713
337
26%
22%
00%
3.4%
16%
59%
00%
0.0%
37%
00%
09%
22%
00%
00%
16%
00%
00%
00%
37%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
0.0%
00%
00%
00%
00%
o
o
o
o
DO
O
I
m
1
2
10
ro
(O
(Q
a
-------
Table 3-23b. (cont.)
Study
Location
Average Soil-Lead Concentration (ug/g)
N
Geometric
Mean
Geometric
Std. Dev.
Minimum
25th
Percentlle
Median
75th
Percentile
Maximum
Percentage of Homes with Average Soil-Lead
Concentration 2
400 ug/g
1200 ug/g
2000 ug/g
5000 ug/g
Houses Built From 1960-1979 (1977 for HUD Grantees and New Haven) (eont.)
HUD National
Survey (403
RA)2
HUD National
Survey (HUD
NS weights)3
California
New Haven.
Connecticut
Rochester
Lead-in-Dust '
Dnplme/
Entryway
Dnplme/
Entryway
All
City Only
Northeast
Midwest
South
West
Northeast -City Only
Midwest -City Only
South • City Only
West -City Only
All
City Only
Northeast
Midwest
South
West
Northeast -City Only
Midwest -City Only
South - City Only
West -City Only
Oakland, LA, Sacramento
Oakland
Near (near the house)
Far (near the street)
Dnplme
Dnplme (fine soil only)
120
46
10
29
64
17
3
6
28
9
117
45
10
29
61
17
3
6
27
9
93
12
33
33
4
4
380
409
667
306
419
296
1186
221
573
260
375
406
667
306
408
296
1186
221
568
260
NA
NA
2865
3822
663
663
27
27
21
24
26
33
21
16
25
24
27
27
21
24
26
33
21
16
25
24
NA
NA
NA
NA
21
18
52
52
206
58
52
75
353
101
52
79
52
52
206
58
52
75
353
101
52
79
NA
NA
NA
NA
275
290
21 4
21 7
353
194
272
159
353
158
303
203
213
217
353
194
270
159
353
158
282
203
NA
NA
NA
NA
379
495
342
37.8
731
284
368
26.6
1412
204
508
266
336
373
731
284
359
266
1412
204
495
266
NA
NA
NA
NA
696
780
758
839
1186
399
834
365
2072
355
879
343
700
793
1186
399
723
365
2072
355
917
343
NA
NA
NA
NA
1256
985
1713
1713
207
685
1713
910
207
40
1713
337
1713
1713
207
685
1713
910
207
40
1713
337
NA
NA
NA
NA
160
111
20%
09%
00%
37%
08%
37%
00%
00%
20%
00%
20%
09%
00%
37%
08%
3.7%
00%
00%
20%
00%
16 1%4
583%'
NA
NA
00%
00%
03%
0.9%
0.0%
00%
08%
00%
00%
00%
2.0%
0.0%
0.3%
09%
00%
00%
0.8%
00%
00%
0.0%
20%
00%
NA
NA
NA
NA
00%
00%
0.0%
00%
0.0%
00%
00%
0.0%
00%
00%
0.0%
00%
00%
00%
0.0%
0.0%
00%
00%
00%
0.0%
00%
00%
NA
NA
NA
NA
00%
00%
00%
0.0%
0.0%
0.0%
00%
00%
00%
00%
0.0%
00%
00%
0.0%
00%
0.0%
00%
00%
00%
0.0%
00%
00%
NA
NA
NA
NA
00%
00%
O
o
o
m
O
33
O
O
8
.8
li
I
I
-------
o
33
o
o
1
o
o
3J
O
I
m
Table 3-23b. (cont.)
Study
Location
Average Soil-Lead Concentration (ug/g)
N
Geometric
Mean
Geometric
Std. Dev.
Minimum
25th
Percentile
Median
75th
Percentile
Maximum
Percentage of Homes with Average Soil-Lead
Concentration *
400 ug/g
1200 ug/g
2000 ug/g
5000 ug/g
Houses Built After 1979 (1977 for HUD Grantees and New Haven)
HUD National
Survey (403
RA)1
HUD
Grantees
Evaluation
Rochester
Lead-in-Dust •
Dnpline/
Entryway
Dnpline
All
City Only
Midwest
South
West
Midwest -City Only
South • City Only
West - City Only
Minnesota
Dnpline
Dnpline (fine soil only)
28
9
4
18
6
1
5
3
1
10
10
274
324
154
345
202
355
527
140
3300
1476
1353
25
25
IB
25
24
10
22
18
37
31
56
79
75
56
79
355
213
79
3300
178
260
119
213
90
199
95
355
276
79
3300
660
720
283
315
154
328
179
355
373
110
3300
1258
1255
523
373
279
700
315
355
1283
31 5
3300
7050
1690
144
144
35
144
105
35
144
31
330
874
876
00%
00%
00%
0.0%
00%
00%
00%
00%
00%
300%
200%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
0.0%
00%
0.0%
00%
0.0%
0.0%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
00%
10 CO NA = Not Available
I Mass-weighted arithmetic average soil-lead concentration as reported in ihe HUD National Survey, summarized without weighting by sample weights
2 Mass-weighted arithmetic average soil-lead concentration as cah.ul.iied in Chapter 3 ol the 40? risk analysis, summarized by weighting each average to reflect Ihe 1997 U S housing slock
3 Mass-weighted arithmetic average sal-lead concentration as reported in Ihe HUD National Survey, summarized by weighting with the National Survey sample weights
4 Percent of samples that exceed 500 ug/g
S Some houses in this housing age category may belong to an earlier age category, as some houses may have actually been built earlier than Ihe year specified within the study's database
(Q
a
»
ro
o
8
-------
o
ID
o
O
1
o
H
m
O
D
I
m
SECTION 403 RISK ANALYSIS
eg
c>
Homes Above 400 ppm :
1 1 .8% (8.9 M to 14.7 M )
Homes Above 1200 ppm :
3.4% (2 M to 4.9 M )
(Total Number of Horn es: 99.3 M)
0.1
10 100
Soil-Lead Concentration (ppm)
Homes Above 2000 ppm
1.7% (0.9 M to 2.7 M )
Homes Above 5000 ppm
0.4% (0.2 M to 0.8 M )
1000
10000
ro
INTERIM
NSLAH
CO
o
Homes Above 400 ppm :
11.1% (8.2 M to 11.5 M)
Homes Above 1200 ppm :
3.2% (2.1 M to 3.7 M)
(Total Number of Homes: 89.2 M)
P J
o
0.1
10 100
Soil-Lead Concentration (ppm)
Homes Above 2000 ppm
1.6% (1 M to 2 M)
Homes Above 5000 ppm
0.4% (0.2 M to 0.5 M)
1000
10000
Figure 3-18. Estimated Distribution of Yardwide Average Soil-Lead Concentration in the Nation's Housing Stock, and
£ Corresponding Estimates of the Percentage of Homes Exceeding Specified Thresholds (with 95% Confidence
I Intervals on the Corresponding Number of Homes, in Millions), Based on Data from the HUD National Survey
oo (top plot) and the Interim NSLAH (bottom plot)
ro
§ Note: The estimated exceedance percentages are calculated based on the fitted distribution (solid curve).
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
surveys. Each estimated exceedance percentage is accompanied by an approximate 95%
confidence interval on the number of homes in the U.S. housing stock that exceeds the threshold
(given in millions).
In Figure 3-18, the distribution based on the HUD National Survey data used in the §403
risk analysis was determined by censoring data values below 24.5 ppm (i.e., the bottom 30
percent of the data, taking into account the sample weights). The distribution based on the
interim NSLAH data was determined by censoring data values below 2.01 ppm, which
corresponds to the bottom 5% of the observed weighted distribution, including negative values.
Among these four thresholds, the estimated percentage of residences that exceed the
threshold vary widely. For a threshold of 2000 ppm, the estimated percentage is 1.6% to 1.7%
for the two surveys, while the percentage increases to from 11.1% to 11.8% for the two surveys
when the threshold is lowered to 400 ppm.
For both surveys, the estimated exceedance percentages specified within Figure 3-18 for
yardwide average soil-lead concentration, based on the fitted lognormal distribution, are also
included within Table 3-24 (columns 2 and 4) for the same four thresholds. Also included in
Table 3-24 (columns 3 and 5) are estimated exceedance percentages that were determined solely
by the proportion of total sampling weights in the survey that corresponded to surveyed units
whose household average floor dust-lead loadings exceeded the given threshold (i.e., information
from the bar charts within Figure 3-18). The two types of estimates are very similar for the
interim NSLAH data except at the highest threshold, while for the HUD National Survey data,
differences between the estimates increase as the threshold decreases. It should be noted that the
lognormal-based estimates for the exceedance percentages (which were also portrayed in Figure
3-18) should be used when making inferences on the nation's housing stock.
Table 3-24. Estimated Percentages of 1997 U.S. Housing Exceeding Specified
Thresholds of Yardwide Average Soil-Lead Concentration
Soil-Lead
Cone.
Threshold
(ppm)
400
1200
2000
5000
§403 Risk Analysis - Based on Data from the
HUD National Survey (n = 284)
Based on the Fitted
Lognormal Distribution
(i.e.. the curve in
Figure 3-18)
1 1 .8%
3.4%
1 .7%
0.4%
Based on the
Weighted Observed
Data (i.e., the bar
chart in Figure 3-18)
13.2%
4.7%
2.5%
0.2%
Data from the Interim NSLAH (n = 706)
Based on the Fitted
Lognormal
Distribution (i.e., the
curve in Figure 3-18)
11.1%
3.2%
1 .6%
0.4%
Based on the
Weighted Observed
Data (i.e., the bar
chart in Figure 3-18)
11.2%
2.9%
1.7%
0.1%
Note: Data are imputed for those surveyed units with missing data prior to calculating the above statistics (34 observations
in the HUD National Survey had either driplme or remote soil-lead concentration imputed prior to calculating a yardwide
average; 42 observations in the interim NSLAH had an imputed yardwide average). The estimates based on the weighted
observed data are simple weighted percentiles that do not originate from a fitted distribution.
DRAFT - DO NOT CITE OR QUOTE
133
August 28, 2000
-------
1 It was also desired to calculate exceedance percentages for only urban residences within
2 the U.S. housing stock, as urban soil has the potential for being more likely to be contaminated
3 by lead than non-urban soil (in the absence of a particular lead source). Thus, the procedure used
4 to fit the distributions in Figure 3-18 was also applied to the HUD National Survey data for only
5 the 146 surveyed homes labeled as being located in urban areas. (The interim NSLAH data were
6 not included in this exercise because homes were not characterized by urbanicity.)
7 Figure 3-19 plots the distribution and documents the exceedance percentages for urban
8 residential soil-lead concentrations as estimated using the HUD National Survey data. Based on
9 the fitted lognormal distribution, this figure indicates that approximately 2.8 percent of the
10 roughly 40 million homes in urban areas are estimated to exceed a yardwide average soil-lead
11 concentration of 2000 ppm". This corresponds to approximately 1.1 million homes. However,
12 because the sampling weights in the HUD National Survey (and revised in the §403 risk analysis)
13 were not necessarily determined to ensure that the weights assigned to the homes in urban areas
14 would be representative of the entire urban housing stock, caution must be taken in making
15 inferences on the national urban housing stock based on these estimates.
16 3.2.2.4 Interpreting the Observed Differences with Other Studies. Contrasting the
17 measured soil-lead concentrations from one study to another is complicated by differences in
18 study designs, sampling locations, and sampling and laboratory protocols and practices used by
19 these studies. As areal patterns in the lead concentration of residential soil have long been
20 recognized, different locations within the same yard can have widely different soil-lead
21 concentrations. For example, levels along the foundation of the residence are typically highest,
22 reflecting the presence of deteriorated lead-based paint formerly on the residence or deposited
23 leaded gasoline emissions washed off the roof. Also, distinct sampling protocols may impact the
24 amount of lead measured in a collected sample. The Rochester and Milwaukee studies, for
25 example, partitioned a collected soil sample into fine- and coarse-sieved fractions. Finally,
26 various laboratory practices and procedures can leach more or less lead from the digested soil
27 sample. Some studies seek to mimic "bioavailable" lead by using an acidic digestion meant to
28 mimic human stomach acids.
29 Unfortunately, insufficient data were available from the various studies in Table 3-17 to
30 consider fully any distinctions in soil-lead concentration that would be prompted exclusively by a
31 study's collection and measurement practices. Undoubtedly, soil collection and measurement
32 practices partially explain the observed differences across the studies, but their effects cannot be
33 quantified at this stage. The data summaries in Section 3.2.2.2 attempted to express soil-lead
34 concentrations in the Rochester study as reflecting the total sample (as is done in many studies)
35 rather than only the fine-sieved portion of the sample by adjusting the data based on relationships
36 observed in the Milwaukee study among fine-, coarse- and total-sieved soil fraction data.
1' The sum of the sampling weights (adjusted in the §403 risk analysis to represent the 1997 housing stock) for the
146 urban homes in the HUD National Survey is roughly 40 million. The fitted lognormal distribution in Figure 3-19 treats the
bottom 20 percent of the HUD National Survey (based on the sample weights) as censored data at 21.3 ug/g.
DRAFT -- DO NOT CITE OR QUOTE 134 August 28, 2000
-------
o
0)
CD
0)
jo
CD
CC
CO
CD
C\J
CD
o
CD
Hornes above 400 ppm
16.3%{4.7Mtoa6M)
Ho
Total Nuiter of Homes
40.3 M
Homes above 2000 ppm
28%(d5Mto20l\/l
Homes abcve 5000 ppm
I I I I Mill 1 I I I I INI 1 I I I Mill 1 I I I I MM
10 100 1000 10000
Soil-Lead Concentration (ppm)
Figure 3-19. Estimated Distribution of Yardwide Average Soil-Lead Concentration Among
Urban Housing in the HUD National Survey, and Corresponding Estimates of
the Percentage of Urban Homes That Exceed Specified Thresholds (with
95% Confidence Intervals on the Corresponding Number of Urban Homes in
the Nation, in Millions)
Note: Because the HUD National Survey was not necessarily conducted in a manner such that the sample
weights for urban housing are representative of urban housing in the entire country, caution should be made
when attempting to use this information to infer about urban housing for the entire nation.
DRAFT -- DO NOT CITE OR QUOTE
135
August 28, 2000
-------
1 It is possible to discuss study-specific caveats about how the housing selection procedure
2 and sample collection and analysis procedures differ between the studies and, therefore, can
3 contribute to the differences observed in the plots and tables in Section 3.2.2.2. For the
4 Baltimore R&M, the Rochester, HUD Grantee, and HUD National Survey studies, this
5 information was summarized in Tables 3-3a through 3-3f of the §403 risk analysis report. For
6 the interim NSLAH, this information was summarized in Section 3.1 of this report. Some of the
7 study differences mentioned in Section 3.2.1.4 as possibly contributing to differences in dust-lead
8 loading data would also be contributors to differences in the reported soil-lead concentration
9 data. Other differences among the studies in Table 3-17 include the following:
10 • The neighborhoods surveyed within the Baltimore R&M study, 3-City study,
11 Cincinnati Longitudinal study, California Lead study, and HUD Grantees
12 evaluation had a high prevalence of homes with lead-based paint hazards, along
13 with a history of children with elevated blood-lead concentrations and/or
14 considered at high-risk for lead poisoning.
15
16 • For the HUD Grantees evaluation, 28% of the homes were single-family
17 buildings, 32% were single-family detached, and 12% were single-family attached
18 (rowhouses). All homes in the R&M intervention group within the Baltimore
19 R&M study were urban rowhouses (single-family attached). Eighty percent of the
20 homes in the HUD National Survey were single-family dwellings. In the 3-Cities
21 study, 100% of the Boston homes were single-family detached residences, most of
22 the Baltimore homes were single-family attached dwellings, and the majority of
23 Cincinnati homes were multi-story, multi-family structures.
24 • The dates of environmental sampling were 11/89-3/90 for the HUD National
25 Survey, 12/93-1/99 for the HUD Grantees evaluation, 8/93-11/93 for the
26 Rochester study, 3/93-11/94 for the Baltimore R&M study, 2/89-2/90 for the
27 Baltimore 3-City study, 7/89-12/89 for the Boston 3-City study, and 1/89-8/89 for
28 the Cincinnati 3-City study. Therefore, the HUD National Survey performed
29 sampling roughly three years before the three major studies in this report, but near
30 in time to others (such as the 3-Cities study).
31 • The New Orleans, Baltimore Garden, Minneapolis Clean-Up and Minnesota
32 studies have sometimes been identified as using distinct laboratory practices,
33 producing higher soil-lead concentrations than might be otherwise measured. The
34 published literature regarding these studies, however, cites nothing unusual.
35 Because the HUD Grantees evaluation emphasizes local control of the individual
36 programs, each grantee is responsible for designing and implementing lead-hazard reduction
37 approaches applicable to its specific needs and objectives. These responsibilities include the
38 recruitment methods, enrollment criteria, and intervention strategies. However, to enable
39 comparison of results from the various approaches, grantees participating in the evaluation
40 follow the same sampling protocols and use standard data collection forms developed
DRAFT - DO NOT CITE OR QUOTE 136 August 28, 2000
-------
1 specifically for this evaluation. Table 3-4 of the §403 risk analysis report documented the
2 differences between grantees in their enrollment/recruitment criteria. As a result, the summaries
3 in Section 3.2.2.2 were also presented by grantee.
4 3.2.2.5 Conclusions of the Soil-Lead Data Comparisons. The following can be
5 concluded from review of the boxplots and tables within Section 3.2.2 of this report, especially in
6 regard to how the reported soil-lead concentration data for various studies compare with data
7 from the HUD National Survey (as portrayed in the §403 risk analysis):
8 • Geometric mean (yard-wide) average soil-lead concentration was quite lower for
9 the HUD National Survey relative to the yardwide estimates for most of the other
10 studies cited in Table 3-17. However, the interim NSLAH (Section 3.2.2.1), as
11 well as such studies as the Cincinnati 3-City and Baltimore Garden studies, did
12 report geometric mean soil-lead concentrations that were comparable to that for
13 the HUD National Survey. Otherwise, the distributions of soil-lead
14 concentrations were rather consistent across the studies and available grantees.
15 • Among the housing age categories, the greatest difference in observed soil-lead
16 concentration between the HUD National Survey (as portrayed in the §403 risk
17 analysis) and the interim NSLAH was for housing built prior to 1940, where
18 nearly a 50% decline in the estimated median was seen from the HUD National
19 Survey to the interim NSLAH. The two sets of results were comparable among
20 the other housing age categories.
21
22 • The low geometric mean soil-lead concentration in the HUD National Survey
23 compared to other studies within Table 3-17 was most dramatic for homes built
24 from 1940 to 1959. For homes built prior to 1940, the geometric mean reported in
25 the §403 risk analysis (463 ug/g) was within 150 ug/g of that for three grantees
26 within the HUD Grantees evaluation: California, Minnesota, and Vermont.
27 However, for homes built from 1940 to 1959, the geometric mean soil-lead
28 concentration across all units in the HUD Grantees evaluation (492 ug/g) was
29 over four times higher than that reported in the §403 risk analysis (92.6 ug/g).
30 Insufficient numbers of housing units built after 1959 in the other studies prevent
31 reliable comparisons of soil-lead concentrations with these studies.
32 • Overall, the importance of housing age is evident in the summaries within the four
33 housing age categories. Older housing is more likely to contain higher average
34 soil-lead concentrations compared to newer housing. However, within an age
35 category, the summaries were reasonably consistent across studies.
36
37 • As expected, dripline/entryway soil-lead concentrations consistently exceeded
38 yard-wide average levels for all studies with sampling plans permitting such
39 comparisons. That soil-lead concentrations exhibit an area! pattern is well-known
40 and documented throughout the scientific literature, and suggests caution when
DRAFT - DO NOT CITE OR QUOTE 137 August 28, 2000
-------
1 comparing the HUD National Survey yard-wide average results to those of other
2 studies.
3 In general, the soil-lead concentrations observed in the HUD National Survey seem lower
4 than many studies, but not necessarily beyond reason. Several of these other studies were
5 conducted in urban neighborhoods already recognized to either have elevated environmental-lead
6 levels or high incidence rates of elevated blood-lead concentrations among resident children. As
7 such, higher soil-lead concentrations among these residences may be entirely consistent.
8 Furthermore, soil-lead levels in the HUD National Survey were found to be comparable with
9 those reported in the interim NSLAH, which reflects the entire nation's housing stock, and in
10 other studies such as the Cincinnati 3-City and the Baltimore Garden study. Even studies
11 conducted within the same urban area can differ considerably in the reported soil-lead
12 concentrations; for example, the Baltimore 3-City study had levels about five times higher than
13 the Baltimore Garden study.
14 3.3 EVALUATION OF SOIL PICA IN CHILDREN
15 This section investigates what has been published in the literature concerning the
16 potential effects that pica for soil may have on children's exposure to lead, over and above the
17 exposure associated with pica for paint that was considered when estimating risks in the §403
18 risk analysis. While the analysis did not consider the independent impact of soil pica over and
19 above paint pica, it considered the impact of soil pica as part of the relation between soil-lead
20 concentration and blood-lead concentration. While this section does not change the approach
21 taken in the original §403 risk analysis, it documents information obtained on the component of
22 soil-lead exposure that may be attributable to soil pica.
23 This section summarizes information on pica behavior for soil and paint for the three
24 studies constituting the Urban Soil Lead Abatement Demonstration Project (USLADP) (USEPA,
25 1996a), the Rochester Lead-in-Dust study (USHUD, 1995a; Lanphear et ah, 1996a), and the
26 Baltimore Repair and Maintenance (R&M) study (USEPA, 1996e). The percentage of children
27 who ingest soil, the frequency of soil ingestion episodes, and the amount of soil ingested by
28 children with pica are estimated.
29 3.3.1 What Is Soil Pica?
30 Definitions. The literature provides varying definitions of pica. Pica is generally
31 accepted to be the consumption of non-food items and there are at least nine different types of
32 pica, including soil pica (Lacey, 1990). Some authors also consider mouthing of non-food items
33 a pica behavior. Usually, pica is seen as normal behavior in young children, but abnormal in
34 older children and adults. Exceptions occur, however, for some individuals, such as children and
35 pregnant women in certain ethnic groups, the socially disadvantaged, groups of low income and
36 socioeconomic status, developmentally delayed individuals, and the mentally retarded.
DRAFT--DO NOT CITE OR QUOTE 138 August 28,2000
-------
1 The American Psychiatric Association (DSM-ffl-R) has clinically defined pica as the
2 ingestion of non-nutritive or inedible substances and requires repeated ingestion of a
3 non-nutritive substance for at least one month before pica is considered a diagnosis. However, in
4 practical research, authors tend to use less rigorous definitions of pica. For example,
5 Shellshear et al., (1975) defined pica simply as an unusual appetite for non-food items.
6 Some authors consider pica a common occurrence in young children while others view
7 pica behavior as abnormal. Sedman (1989) included in his definition of pica the ingestion of
8 foreign substances by children that occurs during the course of normal development. This is
9 consistent with Karam et al., (1990) who stated that pica includes the ingestion of some non-food
10 items and that pica is a relatively common occurrence in small children. Barltrop et al., (1974)
11 defined soil pica as the habitual insertion of soiled fingers or toys into a child's mouth, in
12 addition to the direct consumption of soil. In contrast, Lyngbye et al., (1990) loosely defined
13 pica as a mouthing habit more pronounced than in other children at the same age. Calabrese
14 et al., (1991) defined soil pica as the ingestion of soil in amounts far exceeding those observed in
15 the average child.
16 Pica for soil is considered by most authors to be the purposeful ingestion of soil. This
17 definition is used throughout this report. Estimates of intentional soil ingestion, such as would
18 occur in an actual "pica" episode, range from 500 to 13,000 mg soil/day, according to the studies
19 cited in Table 3-25. To put this in perspective, quantitative estimates of inadvertent soil
20 ingestion by normal children range from 9 to 246 mg/day (see Table 3-25), which are consistent
21 with the estimates used in the §403 Risk Analysis.
22 Methods Used to Measure Soil Inqestion. Average daily soil ingestion can be
23 quantified using a mass-balance approach, in which concentrations of tracer elements in fecal
24 matter are measured and used to estimate the amount of soil ingested. The tracer elements
25 typically used in soil ingestion studies include barium (Ba), manganese (Mn), silicon (Si),
26 aluminum (Al), titanium (Ti), vanadium (V), yttrium (Y), and zirconium (Zr). However, in an
27 adult validation study investigating the recovery of different tracer elements, Calabrese et al.,
28 (1989) concluded that the most reliable elements for this type of study are Al, Si, and especially
29 Y. In addition, the authors indicated that when using these tracers, 500 mg/day could reliably be
30 detected, and 100 mg/day could also be reliably detected but with a higher degree of variability.
31 These levels are greater than most estimates of average daily soil ingestion in children. Tracer
32 elements are generally selected due to their high concentration in soil relative to food products,
33 and their low level of absorption in the gastrointestinal tract. Thus, the quantities of these tracer
34 elements present in the feces, corrected for "background" or intake levels, can be attributed to the
35 ingestion of soil (assuming there is no other non-food ingestion occurring, e.g., paint). Using the
36 concentration of a tracer element in the bulk soil, the total quantity of soil ingested can be
37 calculated. Concentrations of the tracer elements in the bulk soil are determined from soil
38 samples around the child's home and play area. Samples are typically taken from the upper
39 layers of soil (as this is where children are assumed to play), and finer size fractions may be
40 separated out (as this size fraction is preferentially ingested) (Calabrese et al., 1989; Sheppard,
DRAFT - DO NOT CITE OR QUOTE 139 August 28, 2000
-------
DRAFT
Table 3-25. Results of Literature Review on Children's Exposure to Lead Through Soil Pica
DO NOT CITE OR QUOTE 1 40 August 28, 200C
ojm •* «nvo t^oo o\o — 10 g soil on
1-2 days/year
• 1 6% of children are
expected to ingest
> 1 g soil on
35-40 days/year
High degree of daily
variation
• 3 of 6 pica children
exhibited behavior on
only 1 of 4 days, others
did more often
• pica child ingested
between 0.5-3.0 g/day
on 4 of 7 days
-------
Table 3-25. (cont.)
Author(s) /
Publication Data
Davis et al.,
(1990)
- Methods
104 children, aged 2-7 years; ingestion
measured by fecal analysis using
chemical tracers and mass-balance
methodology; study over 7 consecutive
days.
Pica Prevalence
(% exhibiting soil pica)
0% pica reported
Soil Intake Amount
(soil ingestion rate mg/day)
• median soil ingestion estimates
ranged from 25.3-81.3 mg/day
(n = 101 ), depending on the tracer
• mean soil ingestion estimates
ranged from 38.9 ± 1 2.2 to
245.5 ±119. 7 mg/day (n = 101),
depending on the tracer
Pica Frequency
(# days on which soil
was ingested)
Information Not Given
Studies Based on Interviews and Questionnaires
Stanek et al.,
1998
Abadin et al.,
(1997)
Bates et al.,
(1995)
Sedman and
Mahmood,
(1994)
Greene et al.,
(1992)
NCHS1 19841
Shellshear et al.,
(1975)
553 children assessed at well-child
pediatric visits; presence of pica
behavior assessed via parent interview*
Discussion of ATSDR method to
estimate blood-lead levels in children,
where soil ingestion is one predictor
143 children, aged 12-23 months; soil
ingestion assessed via questionnaire '
Estimated average daily and lifetime
soil ingestion in young children using
results of two previous chemical tracer
studies (Davis et al., 1990 and
Calabrese et al, 1 989); age adjusted for
a 2 year-old child
270 socioeconomically disadvantaged
children; pica assessed via caretaker
interview at ages 2 years, 3 years, and
4 years 10 months *
2,195 children, ages 1-5 years,
presence of pica behavior assessed by
household interview '
170 children, ages 1-5 years; pica
assessed via parent interview '
Reported by age; pica prevalence of 38%
at age 1 year declines to 21% at age 2
years and < 10 % for ages 3-6 years;
overall estimate of 1 8% for children aged
1 to 6 years
Not Applicable
• 62% (89 of 143) ate soil
• 38% (54 of 143) never ate soil
Not Applicable
• 1 9% at 2 years of age
• 1 3% at 3 years of age
• 6% at 4 years 10 months of age
• 6 month-3 year age group: 1 1 .0%
• 4-5 year age group: 3.2%
• annual family income <$10K: 11.9%
• annual family income >_ S10K: 6.0%
10.6% (18 children of 170)
Information Not Given
• use assumed soil ingestion rate of
200 mg/day
Information Not Given
• mean estimate (2 year-old):
1 95 mg/day (std. err. - 53 mg/day)
• estimated average daily soil
ingestion over a lifetime: 70 mg/day
(accounting for changes in soil
ingestion with age)
Information Not Given
Information Not Given
Information Not Given
• 38% at least monthly,
24% at least weekly,
1 1 % daily at age 1 year
• 21% at least monthly,
7% weekly, 0% daily at
age 2 years
Not Applicable
Information Not Given
Information Not Given
Information Not Given
Information Not Given
Information Not Given
O
o
1
o
o
3J
O
I
m
(Q
ro
co
-------
§
Table 3-25. (com.)
o
z
o
m
O
3)
O
I
m
Author(s) /
Publication Date
Methods
Pica Prevalence
(% exhibiting soil pica)
, Soil Intake Amount
(soil ingestion rate mg/day)
Pica Frequency
(# days on which soil
was ingested)
1
2
3
4
5
6
7
8
9
10
11
12
13
Barltrop et al.,
(1974)
119 children in two towns, ages
2-3 years; 48 children in high soil lead
area; presence of pica behavior
assessed via parent interview"
• 51 of 119 (43%) conformed to pica
definition; 11 of 119 (9%) known to
swallow soil
• 33% (16 of 48) of children had pica for
soil (in high soil lead area)
Information Not Given
Information Not Given
Cohen et al.,
(1973)
230 rural and 272 urban children, mean
age of 4 years; history pica assessed
via parent questionnaire*
• 50% of the children had a history of
pica at some time for dirt, cigarettes, or
other non-food items
• 10% reported ingestion of paint or
plaster
Information Not Given
Information Not Given
• soil pica defined as ingestion of soil in quantities far exceeding those observed in the average child
b subject population consisted of Jamaican children of normal intelligence in long-term institutional settings, with the exception of one child (i.e., the child exhibiting extreme
pica in the older group) who was mentally retarded
" pica defined as ingestion of > 1 g soil/day
* high soil ingestion defined as >0 5 g soil/day on a regular or nearly daily basis
' soil pica not quantitatively defined in this study; pica only indicates the occurrence of ingestion
' soil pica defined as "an unusual appetite' for soil
1 definition of pica for soil included children who habitually put fingers, etc , in mouth while playing in their gardens, as well as children who actually put soil directly into
their mouths
-------
1 1998). Most quantitative estimates of the amount of soil ingested that were reviewed in this
2 report were obtained using the mass-balance/tracer element approach.
3 Incidence rates (i.e., prevalence) of pica for soil in young children may be estimated from
4 parental questionnaires. This approach can yield biased results, however, as it relies on the
5 observation and accurate reporting of pica by the adult. In addition, the response depends on the
6 wording of the question. Various surveys have asked whether the child eats dirt (NCHS, 1984;
7 Bates et al., 1995; USEPA, 1996a - Boston and Baltimore portions), whether the child puts dirt
8 or sand in mouth while playing outside (USEPA, 1996a - Cincinnati portion; USEPA, 1996e), or
9 whether the child puts fingers or toys in mouth while playing outside (Barltrop et al., 1974).
10 Clearly, these questions would elicit different responses from the same caregiver. In addition,
11 response choices may be simply yes or no, may specify a timeframe (e.g., in the past month), or
12 may be open-ended. These choices, too, would result in differing responses. Thus, care must be
13 taken in comparing soil pica prevalence rates originating from parental questionnaire data.
14 While mass-balance studies provide soil ingestion rates to support prevalence data, these
15 studies are also subject to error and have disadvantages. For example, Calabrese et al., (1989)
16 acknowledge that analyzing chemical tracers without the use of a mass-balance approach
17 (i.e., not correcting for intake) can result in soil ingestion estimates that are increased by factors
18 of 2 to 6. In addition, the particular tracer used, the duration of the study, and the frequency of
19 sampling may also influence reported results (Calabrese et al., 1989; Wong, 1988; Calabrese
20 et al., 1997). For example, the short duration of most mass-balance studies makes it difficult to
21 determine a "normal" rate of soil ingestion for a child. Approaches using chemical tracers also
22 have disadvantages in that they are more expensive and generally have small sample sizes.
23 3.3.2 How Does the §403 Risk Analysis Account for Soil Pica?
24 Within the exposure assessment (Chapter 3) portion of the §403 risk analysis report
25 (USEPA, 1998a), soil was considered an indirect source of lead exposure, although summary
26 information on soil pica frequency from two lead exposure studies was presented in Table 3-3b
27 of the §403 risk analysis report.
28 An indicator of soil pica was considered as a candidate predictor variable in the
29 development of the empirical model for the §403 risk analysis. The soil pica variable was based
30 on the parental questionnaire administered in the Rochester Lead-in-Dust study. This variable
31 measured the child's tendency to put dirt or sand in the mouth using a scale of 0 (never) to
32 4 (always). The soil pica variable was borderline significant in single media models, which
33 assessed the relationship between blood-lead concentration and each predictor variable under
34 consideration. These single media models were the first step in developing the empirical model.
35 Variable selection for the multimedia exposure model was based on several properties, including
36 the strength of the relationship with blood-lead concentration as estimated using the bivariate
37 statistical models, predictive power of each variable when included into a model with competing
38 sources of lead exposure, and interpretability of parameter estimates. The soil pica variable was
DRAFT - DO NOT CITE OR QUOTE 143 August 28, 2000
-------
1 dropped during this phase of the empirical model development. Additional information on the
2 development of the empirical model can be found in Appendix G of the §403 risk analysis report.
3 Age-dependant soil and dust ingestion rates for the Integrated Exposure, Uptake, and
4 Biokinetic (EEUBK) model were taken from the IEUBK guidance manual (USEPA, 1994) and
5 represent central tendencies within the range of values seen in different studies. Combined soil
6 and dust ingestion amounts ranged from 85 to 135 mg/day, as shown in Table 4-1 of the
7 §403 risk analysis report, of which 45 percent was assumed to be from soil. Thus, soil ingestion
8 was assumed to be between 38 and 61 mg/day for children aged 0 to 7 years, with the maximal
9 ingestion estimated for children aged 1 to 3 years. These ingestion rates are consistent with
10 estimates of inadvertent soil ingestion presented in this report, but are not representative of pica
11 episodes. While IEUBK model predicted blood-lead levels were adjusted in the §403 risk
12 analysis to allow consideration of paint pica in homes with damaged lead-based paint, as
13 described in Section 4.1 and Appendix Dl of the §403 risk analysis report, no such adjustment
14 was made for the effect of soil pica.
15 It should be noted that while neither model used in the §403 risk analysis had explicitly
16 accounted for soil pica as a separate factor independent of paint pica, the impact of soil pica was
17 included in the analysis as part of the relation between soil-lead concentration and blood-lead
18 concentration which the analysis characterized.
19 3.3.3 Prevalence of Soil Pica Behavior
20 Estimates reported in the scientific literature of the percentage of children who ingest soil
21 are summarized in this section. From the literature, it was not possible to estimate the percentage
22 of children who exhibit pica for soil but not paint. The §403 risk analysis did account for the
23 effect of paint pica on blood-lead concentration estimates. For children who ingest both paint
24 chips and soil, it is reasonable to assume that the effect of soil pica is insignificant compared to
25 that of paint pica. Thus, in estimating the percentage of children who ingest soil, it is important
26 to exclude those who also ingest paint chips. It was possible to estimate the percentage of
27 children who exhibit soil pica, but not paint pica, using information from parental questionnaires
28 administered in the USLADP study (USEPA, 1996a), Baltimore R&M study (USEPA, 1996e),
29 and Rochester Lead-in-Dust study (USHUD, 1995a; Lanphear et al., 1996a). This information is
30 also summarized in this section.
31 3.3.3.1 Literature Review. Most sources in the literature reported prevalence rates for
32 general pica behavior (mouthing or eating non-food items) or for soil pica (eating dirt). One
33 source (Stanek et al., 1998) reported pica rates for a variety of specific non-food items (soil, paint
34 chips, paper, toys, etc), but did not cross-tabulate [e.g., 18 percent of children ages 1 - 6 years, as
35 assessed by parent interview, were reported to ingest/mouth dirt at least monthly and 3 percent to
36 ingest/mouth paint chips, but information was not provided on how many children eat both soil
37 and paint chips (Stanek et al., 1998)]. An overview of selected studies estimating pica
38 prevalence is shown in Table 3-25 above. It is important to note that in the cited studies, various
39 definitions of "pica" were used in reporting the prevalence of pica behavior. For example,
DRAFT -- DO NOT CITE OR QUOTE 144 August 28, 2000
-------
1 Greene et al., (1992) defined soil pica only as the occurrence of soil ingestion and reported the
2 percentage of children who ingest soil based on caretaker interview. In comparison, Calabrese et
3 al., (1997) defined soil pica as consumption exceeding 0.5 grams per day and reported the
4 prevalence of pica behavior as assessed quantitatively by mass-balance methods. It should be
5 noted that only primary research findings are reported in Table 3-25, with the exception of
6 Calabrese and Stanek (1993), which summarized and critiqued the dissertation of Wong (1988),
7 which was not easily obtained. Several review articles were also obtained. These were excluded
8 from Table 3-25, as insufficient details of the source studies were provided. General findings of
9 the review articles are cited in the text.
10 Table 3-25 shows that the estimated percentage of children ingesting soil ranged from
11 1.6 to 62 percent and varied with definition/criteria for soil pica used, age group of children, and
12 socioeconomic status. In general, 12 of 16 observations in the table report a prevalence of soil
13 ingestion in children of 13 percent or lower (where, at a minimum, limiting criteria are defined as
14 ingesting soil at least once). "Normal" mouthing behavior, however, is typically exhibited more
15 commonly, particularly in the younger age groups. For example, Barltrop et al., (1974) reported
16 that 43 percent of children exhibited pica defined to include mouthing behavior, but that
17 9 percent were estimated to swallow soil. Stanek et al., (1998) assessed non-food ingestion and
18 mouthing behaviors in 533 children, ages 1 to 6, by parental interview. Results of the survey
19 indicated that 38 percent of 1 year old children and 21 percent of 2-year old children
20 ingest/mouth soil at least monthly. In contrast, at ages 3 to 6 years, less than 10 percent of
21 children were observed to ingest/mouth soil at least monthly. At age 1 year, 11 percent of
22 children were observed to ingest/mouth soil daily compared to one percent or less among
23 children aged 2 to 6 years.
24 Of the studies that used mass-balance methodology, prevalence of soil pica ranged from
25 1.6 to 20.8 percent. For studies that employed parent or caretaker interview methodology, soil
26 pica prevalence ranged from 3.2 to 19 percent, although one study reported a rate of 62 percent,
27 which appears to be more consistent with studies that monitored general mouthing behavior. For
28 both methodologies, the prevalence of soil ingestion tended to be higher in the younger age
29 groups and for children in families with lower socioeconomic status. Although, as mentioned in
30 Section 3.3.1 above, techniques utilizing parental interview are generally considered less reliable
31 than quantitative methodologies, the issue of consistently defining "pica" when reporting study
32 results is an issue not only in studies using questionnaire methodology, but also in the
33 mass-balance/chemical tracer studies. Differences in values reported, both between and within
34 the various assessment techniques, may largely be due to differences in how pica behavior is
35 being defined in the study. Thus, many studies estimating soil ingestion prevalence may not
36 consistently monitor the actual overall risk associated with soil pica behavior in children.
37 3.3.3.2 Prevalence of Soil Pica Separate from Paint Pica. Although the literature
38 review provided several estimates of the prevalence of soil pica behavior, none of the cited
39 sources provided information about concurrent paint pica behavior. For the purpose of this
40 report, the prevalence of soil pica behavior in the absence of paint pica is of interest, as the §403
41 risk analysis did account for paint pica behavior. For children who ingest both paint chips and
DRAFT -- DO NOT CITE OR QUOTE 145 August 28. 2000
-------
1 soil, it is reasonable to assume that the effect of soil pica is insignificant compared to that of
2 paint pica. Unless there is industrial contamination, or the home is in an area with heavy traffic,
3 where residual leaded gasoline emissions are present, lead in residential soil is usually derived
4 primarily from lead-based paint. Thus, soil pica can be considered an indirect pathway of
5 exposure to leaded paint, whereas paint pica is a direct exposure pathway.
6 Information on pica behavior for paint, soil, and other objects was collected in the three
7 USLADP studies (USEPA, 1996a), the Rochester Lead-in-Dust study (USHUD, 1995a;
8 Lanphear et al., 1996a), and the Baltimore R&M study (USEPA, 1996e), through parental
9 reporting of observed behaviors. Therefore, it is possible to use these data to estimate the
10 prevalence of soil pica separately from paint pica. This information is summarized in Table
11 3-26. As can be seen in this table, rates of soil pica only range from 9.1 to 40.9 percent, while
12 rates of both soil and paint pica range from 1.4 to 7.4 percent.
13 Some of the disparity in the rates reported in Table 3-26 can be explained by the survey
14 questions and other factors associated with the study. For example, in the Rochester Lead-in-
15 Dust, Baltimore R&M, and Cincinnati USLADP studies, parents were asked how frequently the
16 child put dirt or sand in his or her mouth. In contrast, parents in the Boston and Baltimore
17 portions of the USLADP were asked how frequently the child ate dirt or sand. The paint pica
18 questions were more consistent across studies, querying how frequently the child put paint chips
19 , in his or her mouth. In the Cincinnati USLADP study, the time-period of observation for both
20 soil and paint was limited to the previous month, whereas the other studies used open-ended time
21 periods. Response rates in Rochester were consistent with literature estimates of soil pica that
22 included mouthing behavior, while the Baltimore R&M and Cincinnati USLADP studies
23 provided substantially lower estimates. Because most homes in the Baltimore R&M study had
24 small or no yards, the low estimates of soil mouthing behavior are not unexpected. The lower
25 response in Cincinnati is probably due to the limited period of observation.
26 Since inadvertent soil ingestion due to mouthing behavior was included in the IEUBK
27 model analysis for the §403 risk analysis, the prevalence of soil ingestion, rather than mouthing
28 behavior, is of interest in the context of this report. Thus, the Boston and Baltimore portions of
29 the USLADP study provide the best estimates of soil pica behavior in the absence of paint pica.
30 These estimates are 14.4 and 16.3 percent in Boston and Baltimore, respectively. These
31 estimates are greater than those derived from the mass-balance studies, but consistent with other
32 studies that rely on parental reporting methods. The prevalence of pica for both paint and soil
33 was low in Boston (1.4 %), but somewhat higher in Baltimore (6.0 %). Adding these rates to the
34 reported rates for soil pica alone does not substantially increase the estimates, however, which
35 remain in the range of other studies that rely on parental reporting.
DRAFT - DO NOT CITE OR QUOTE 146 August 28, 2000
-------
1
2
3
Table 3-26. Estimated Rates of Paint and Soil Pica Behavior Reported in the
USLADP Studies, the Rochester Lead-in-Dust Study, and the
Baltimore R&M Study
Study
Boston
USLADP
(146 children)
Baltimore
USLADP
(400 children)
Cincinnati
USLADP
(220 children)
Baltimore R&M
Pre-mtervention
(165 children)
Rochester
Lead-in-Dust
(203 children)
Type of Pica
Behavior
Soil only
Paint only
Soil and Paint
neither
Soil only
Paint only
Soil and Paint
neither
Soil only
Paint only
Soil and Paint
neither
Soil only
Paint only
Soil and Paint
neither
Soil only
Paint only
Soil and Paint
neither
Study Children Exhibiting Such Pica Behavior
Percent (#) of
Study Children1
14.4% (21)
9.6% (14)
1.4% (2)
74.7% (109)
16.3% (65)
10.5% (42)
6.0% (24)
67.3% (269)
23.2% (51)
2.7% (6)
2.3% (5)
71.8% (158)
9.1% (15)
7.3% (12)
7.3% (12)
76.4% (126)
40.9% (83)
2.5% (5)
7.4% (15)
49.3% (100)
Average Age
(months)
NA
NA
NA
NA
35.7
33.1
26.5
41.7
35.9
19.7
28.0
28.1
27.5
26.7
24.2
31.1
21.0
23.0
20.0
21.4
Geometric Mean.
Blood-Lead Cone. (/sg/dL)
12.5
12.7
13.4
11.7
12.0
11.7
15.2
10.3
12.8
15.3
14.7
8.9
10.5
15.3
20.7
9.4
6.1
11.5
8.5
6.2
5
6
1
8
9
10
11
12
13
14
15
16
17
18
19
20
21
1 A response of 'Unknown* was treated as missing and was not included in the calculation of these percentages.
NA = Not applicable
22 3.3.4 Estimating the Frequency of Ingestion and Amount
23 of Soil Ingested by Children Who Exhibit Soil Pica
24 As discussed in Section 3.3.3.1 above, studies reporting soil ingestion prevalence have
25 the potential to misrepresent the extent of soil pica behavior in children due to differences in
26 methodology and criteria for defining pica. Therefore, estimates of ingestion quantity and
27 frequency may also be employed to assess the severity of soil pica behavior.
28 Mass-balance studies provide data on the frequency of ingestion and amount of soil
29 ingested by children who exhibit soil pica. These studies estimate the typical amounts of soil
30 inadvertently ingested by normal children as ranging from 9 to 246 mg/day. The estimated
DRAFT » DO NOT CITE OR QUOTE
147
August 28, 2000
-------
1 quantities ingested in actual "pica" episodes are between 500 and 13,000 mg/day (Table 3-25).
2 The literature generally reports pica behavior to be episodic in nature, varying both amongst
3 different children and within individual children. In addition, the occurrence (including both
4 frequency and quantity) of soil ingestion was observed to be influenced by the age of the child
5 (Stanek et al., 1998; NCHS, 1984), as well as by variety of factors that may alter the child's
6 access to soil, including seasonal variation and/or climate/vegetation differences (Simon, 1998;
7 Wong, 1988, as cited in Calabrese and Stanek, 1993), socioeconomic status (NCHS, 1984;
8 Bhatia, 1988), and parental supervision (Wong, 1988, as cited in Calabrese and Stanek, 1993;
9 Bhatia, 1988). However, Davis et al., (1990) found that although there was considerable
10 variability in soil ingestion estimates among children, there was no consistent demographic or
11 behavioral factor that was predictive of soil ingestion.
12 Calabrese et al., (1989) estimated median soil ingestion rates, including those involved in
13 non-pica behavior, between 9 and 40 mg/day (n = 64). Calabrese et al., (1997) also observed
14 median soil ingestion rates under 40 mg/day in 12 children (selected from the population
15 described in Stanek et al., 1998) identified by their parents as likely to ingest soil at a high rate.
16 These levels of soil ingestion typically would not be considered pica behavior. Each of these
17 studies, however, did report observations of a child exhibiting extreme soil pica behavior, with
18 one child ingesting from 5 to 8 grams of soil per day (Calabrese et al., 1989) and another child
19 ingesting between 0.5 and 3.0 grams of soil per day on 4 of 7 days (Calabrese et al., 1997).
20 Calabrese et al., (1991) found that the soil pica behavior for the former child occurred only on
21 two days during the two weeks of observation with an ingestion rate ranging from 10-13 grams
22 of soil per day, suggesting that the issue of variability in soil pica behavior may be very
23 important, meriting further research. Implications of these patterns were demonstrated by
24 Calabrese et al., (1993), who observed that on the two days when the child displayed soil pica
25 behavior, she also displayed striking increases in fecal lead excretory values. In contrast, the pica
26 child reported in Calabrese et al., (1997) consistently ingested large quantities of soil
27 (0.5-3 g soil/day on 4 of 7 days).
28 Calabrese and Stanek (1993), in a review of a dissertation by Wong (1988), presented the
29 results of a 4-month mass-balance/chemical tracer study of 52 Jamaican children of generally
30 normal intelligence in an institutional setting. The children were partitioned into a younger
31 (0.3-7.5 years) group and an older (1.8-14 years) group. One of the children in the older group
32 exhibited mental retardation. This was the only child in the older group (of 28 children) that
33 exhibited soil pica exceeding one gram of soil per day. This child had an average ingestion rate
34 of 41 g soil/day over 4 months (observations on 1 day per month). In the younger group,
35 10.5 percent of total observations (n = 84) included soil pica, and five of the 24 children
36 exhibited soil pica on at least one occasion. The Wong study showed that soil pica occurred
37 more frequently in younger children, and there was a fairly high degree of daily variation in soil
38 ingestion among the children exhibiting soil pica. For example, 3 of 6 children displayed pica on
39 only 1 of 4 days. Furthermore, even for the children who consumed soil more consistently with
40 regards to frequency, the rates were still variable (e.g., 1.0-10.3 g/day). Calabrese and Stanek
41 (1993) suggest that although this study confirms that soil pica, strictly defined as ingestion
DRAFT -- DO NOT CITE OR QUOTE 148 August 28, 2000
-------
1 greater than 1.0 g/day, is likely to be rare in older children, the Wong (1988) study is important in
2 that it challenges the idea that pica is a rare event in younger children.
3 Using daily soil ingestion data from their 1989 study, Stanek and Calabrese (1995)
4 developed annual soil ingestion distribution estimates as follows. First, the mean and variance of
5 daily soil ingestion were estimated for each of the 64 children in the 1989 study, based on 4 to
6 8 daily estimates for each child. Then 365 daily soil ingestion amounts for each child were
7 calculated as percentiles of a log-normal distribution with the estimated mean and variance, in
8 increments of 1/365. Based on these distributions, Stanek and Calabrese conclude that 33
9 percent of children are expected to ingest more than 10 grams of soil on 1-2 days per year and
10 that 16 percent of children are expected to ingest more than 1 gram of soil on 35-40 days per
11 year. These ingestion levels are consistent with amounts estimated for soil pica episodes. The
12 median and 95th percentile for average daily soil ingestion resulting from this method were
13 75 mg/day and 1,751 mg/day, respectively. While the median estimate is similar to previous
14 estimates, the estimated 95th percentile is substantially greater than most other estimates.
15 Assumptions and limitations of this approach include:
16 1. The assumption of a log-normal distribution for daily soil ingestion. Insufficient
17 data were available to determine whether this assumption is reasonable.
18 2. The estimation of the mean and variance for each child based on very small
19 sample sizes. The annual estimates were strongly affected by the tails of the
20 distribution, which are imprecise due to large variability in the estimates of the
21 mean and variance.
22 3. The extrapolation of daily soil ingestion estimates from a 2-week period in
23 autumn to the remainder of the year, without regard to possible seasonal effects.
24 In addition, the children studied were a nonrandom sample residing in or near an
25 academic community in western Massachusetts. Thus, the soil ingestion behavior
26 of these children may not be representative of those living in other climates,
27 geographic regions, or in inner-city or rural areas.
28 4. The presence of trace elements in fecal matter was assumed to be entirely due to
29 soil consumption, after correcting for food consumption, with no contribution
30 from indoor dust.
31 Many of these assumptions and limitations serve to introduce positive bias to the daily soil
32 ingestion estimate, while the effect of others is unclear. Nonetheless, this analysis is at present,
33 the only available source of both frequency of soil pica episodes and amount ingested during soil
34 pica episodes.
DRAFT -- DO NOT CITE OR QUOTE 149 August 28, 2000
-------
1 3.3.5 Conclusions on Soil Pica
2 The following conclusions can be made from the findings presented in this section:
3 • The prevalence of soil pica, exclusive of paint pica, is most likely between 10 and
4 20 percent in young children. For the purpose of this report, the Boston and
5 Baltimore portions of the USLADP provide the best estimates of soil pica
6 behavior in the absence of paint pica (14.4 and 16.3 percent, respectively).
7 • Soil pica behavior is episodic in nature. The frequency of soil pica episodes
8 depends on many factors, including climate, access to bare soil, socioeconomic
9 standing, age of child, and parental supervision. In one study of 12 children
10 identified by their parents to be predisposed to pica for soil, only one child
11 displayed soil pica during the two week observation period (Calabrese et al.,
12 1997). Only one study estimated annual rates for pica episodes (Stanek and
13 Calabrese, 1995). This study suggested that 33 percent of children would ingest
14 more than 10 grams of soil on 1-2 days per year, and that 16 percent of children
15 are expected to ingest more than 1 gram of soil on 35-40 days per year.
16 • Estimates of the amount of soil ingested during pica episodes vary widely among
17 the mass balance studies, from 500 to 13,000 mg/day. The average daily ingestion
18 over a year, however, may be much lower. Assuming the frequencies estimated
19 by Stanek and Calabrese (1995), children who ingest 15 grams of soil on 1-2 days
20 per year and 50 mg/day on remaining days would have an average daily soil intake
21 of 132 mg/day over the course of a year. Children who ingest 1.5 grams of soil on
22 40 days per year and 50 mg/day on remaining days would have an average daily
23 soil intake of 209 mg/day. A question, however, is whether the amount of lead in
24 soil ingested on the small number of days where pica episodes occurred would be
25 sufficient to elevate the blood-lead concentration to unsafe levels.
26 3.4 CHARACTERIZING THE POPULATION OF CHILDREN
27 IN THE NATION'S HOUSING STOCK
28 For the §403 risk analysis, it was necessary to estimate numbers of children of specific
29 age groups who reside within the 1997 national housing stock in order to characterize the extent
30 to which various environmental-lead levels provide exposures to children and to characterize the
31 benefits associated with performing interventions under §403 rules. These estimates were based
32 on numbers of housing units determined by sampling weights within the HUD National Survey
33 (conducted in 1989-1990), revised to represent the 1997 national occupied housing stock and on
34 average numbers of children per housing unit determined from the 1993 American Housing
35 Survey (AHS). The estimates used in the §403 risk analysis were presented in Section 3.3.2 and
36 Appendix C of the §403 risk analysis report. This section provides alternative estimates using
DRAFT -- DO NOT CITE OR QUOTE 150 August 28, 2000
-------
1 more recent data (i.e., interim data from the NSLAH and data from the 1997 American Housing
2 Survey).
3 The method to calculating the alternative estimates involved determining the numbers of
4 children in a given age group for each of the 706 housing units surveyed within the NSLAH
5 whose interim data were made available to this effort. Methods used to obtain these estimated
6 numbers of children were similar to those presented in Section 1.2 of Appendix C1 of the §403
7 risk analysis report. For a given age group of children, the estimated number of children
8 associated with a given NSLAH-surveyed unit was determined by the following formula:
9
# children = (1997 weight) *(Average # residents per unit) *(# children per person)
(1)
10 The "1997 weight" factor in equation (1) was the interim sampling weight from the NSLAH for
11 the unit. The factor "average # residents per unit" in equation (1) was calculated for the housing
12 group based on information obtained from the 1997 AHS. The 1997 AHS database provided
13 information on up to 18 residents within each housing unit in the AHS. Once units surveyed in
14 the 1997 AHS were placed within the four year-built categories (pre-1940, 1940-1959, 1960-
15 1979, post-1979), the average number of people residing in a unit (regardless of their ages) was
16 calculated for each group. This average ranged from 2.5 to 2.7 across the four year-built
17 categories. Therefore, a common average of 2.6 residents per unit was used for the entire
18 national housing stock. The third factor in equation (1), "# children per person," represented the
19 average number of resident children (of the given age group) in a housing unit. This factor was
20 calculated from information contained in the 1997 AHS, instead of from forecasted birth rate and
21 population estimates from the Bureau of the Census as was done in the §403 risk analysis.
22 Table 3-27 contains estimates of average number of children per unit in the 1997 national
23 housing stock, according to age group. These number are the product of the final two factors in
24 equation (1). Therefore, these number are multiplied by the sampling weights for each housing
25 unit in the interim NSLAH to obtain a revised number of children per housing unit. For children
26 aged 12-35 months, the estimated average of 0.073 children per unit is about 9% lower than the
27 estimate of 0.080 used in the §403 risk analysis.
28
29
30
31
32
Table 3-27. Alternative Estimates of the Average Number of Children Per Unit in the
1997 National Housing Stock, by Age of Child
Age Group
1 2-35 months
12-71 months
Estimated Average Number of
Children Per Unit
2.6*0.0281 = 0.073
2.6*0.0722 = 0.188
DRAFT - DO NOT CITE OR QUOTE
151
August 28, 2000
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
By summing the estimates across surveyed units in the interim NSLAH, the updated
number of children aged 12-35 months and 12-71 months residing within the 1997 national
housing stock is obtained by year-built category and for the nation. Table 3-28 provides these
alternative estimates on the number of children residing in the 1997 housing stock according to
age of housing unit and age of child. The overall estimate of approximately 6.51 million children
aged 1-2 years is approximately 18% lower than the estimate of 7.96 million made in the §403
risk analysis. The lower estimates are due to the lower per-unit estimate from Table 3-27 and on
the lower sample weight total in the interim NSLAH data compared to the HUD National Survey
(Table 3-2). They are also likely to be underestimates of the numbers of children of the given
age category, based upon population projections previously published by the U.S. Bureau of the
Census (e.g., Day, 1993).
Table 3-28. Alternative Estimates of the Average Number of Children in the 1997
National Housing Stock, by Age of Child and Year-Built Category. Based on
Data Obtained Since the §403 Risk Analysis
Years in Which Housing
Units Were Built
Prior to 1 940
1940-1959
1960-1977
After 1977
Unknown1
All Housing2
Age of Child Within These Housing Units
1-2 Years
1.053,000
1 ,234.000
1,877.000
1,759.000
591,000
6,513,000
1-5 Years
2,705,000
3,170,000
4,822,000
4,519,000
1,518,000
16,735,000
1 There are 66 units in the interim NSLAH which have missing age of house data.
2 Values in this row may differ from sum of previous rows due to rounding.
25 3.5 SUMMARIES OF DUST-LEAD LEVELS ON SURFACES OTHER
26 THAN UNCARPETED FLOORS AND WINDOW SILLS
27 The exposure assessment in Chapter 3 of the §403 risk analysis report concluded that
28 even at low to moderate lead levels, lead-contaminated dust can affect children's blood-lead
29 concentration. The assessment focused on dust-lead found on floor and window sill surfaces, for
30 which §403 regulatory standards were proposed. However, dust found on other surfaces, such as
31 exterior dust and dust in air ducts, carpeted floors, window troughs (also known as window
32 wells), and upholstery, may also potentially present a lead exposure hazard. Many issues
33 concerning potential exposure to dust-lead on these other surfaces were raised throughout the
34 §403 Dialogue Process as well as in comments received on preliminary drafts of the §403 risk
35 analysis and on the proposed rule. For example, there was extensive discussion during the
36 Dialogue Process concerning whether standards were necessary for window troughs (i.e.,
37 window wells) as long as there are standards for window sills and the window troughs are
DRAFT - DO NOT CITE OR QUOTE
152
August 28, 2000
-------
1 thoroughly cleaned. Concern was also expressed about sampling on carpeted and upholstered
2 surfaces.
3 The purpose of this section is to supplement information in the original exposure
4 assessment by assessing the potential exposure to dust-lead found on surfaces other than floors
5 and window sills. In particular this assessment seeks to answer the following questions:
6 1. What information is available to assess residential lead exposure resulting from dust
7 on surfaces other than floors and window sills?
8 2. What does this information say about the distribution of environmental lead levels
9 for dust on these other types of surfaces?
10 3. Is there evidence of a relationship between these exposures and children's blood-
11 lead concentrations?
12 4. Is the information sufficient to set a regulatory standard and is a standard necessary?
13 The exposure assessment is based on a review of the literature to identify studies with
14 potentially useful data for assessing lead hazards due to dust-lead on these other surfaces. It
IS should be noted that this section deals only with several specific surfaces other than uncarpeted
16 floors and window sills. These surfaces include exterior dust, air ducts, window troughs, and
17 upholstery. Carpeted floors are not considered in this section because lead exposure from
18 carpeted floors is being addressed in a separate report.
19 The literature review for this assessment drew upon previous literature reviews conducted
20 for the §403 risk analysis and reviews conducted for other EPA published reports (e.g., USEPA,
21 1997d). Most of the studies that were found that addressed the various surfaces are included
22 here, regardless of whether there is any information specifically relating dust-lead levels and
23 blood-lead levels. For those studies where blood samples were collected from resident children,
24 those results are also presented. Table 3-29 provides a summary of the studies that were
25 examined. The table indicates the surfaces from which dust samples were collected and in the
26 case of exterior dust samples, where those samples were collected.
27 Table 3-29 indicates that the review of the literature found fourteen studies that have
28 examined exterior dust as a source of lead exposure and seven studies that similarly assessed
29 window troughs. Only four studies were located with significant information on dust-lead levels
30 in air ducts, and four with similar information on upholstery.
31 3.5.1 Distribution of Dust-Lead on Surfaces Other than Floors
32 and Window Sills
33 Tables 3-30 through 3-33 present summary information from the studies related to
34 exterior dust samples, air duct samples, window trough samples, and upholstery samples,
DRAFT--DO NOT CITE OR QUOTE 153 August 28,2000
-------
1 Table 3-29. Studies for Which Dust Samples Have Been Collected from Exterior Areas,
2 Air Ducts, Window Troughs, and Upholstery for Lead Analysis
= medium was sampled in the given study
Study
Lead-Based Paint Abatement and Repair & Maintenance
(R&M) Study (USEPA. 1997a,e)
University of Rochester Lead-in-Dust Study (USHUD.
1995a)
Urban Soil Lead Abatement Demonstration Project (USEPA.
1996a)
The National Survey of Lead-Based Paint in Housing
(USEPA, 1995)
The HUD Lead-Based Paint Abatement Demonstration
(USHUD, 1991)
The Comprehensive Abatement Performance (CAP) Study
(USEPA, 1996d)
Birmingham Urban Lead Uptake Study (Davies et al., 1990)
Mexico City Study (Romieu et al., 1995)
Butte-Silver Bow Environmental Health Lead Study (Butte-
Silver Bow Dept. of Health et al., 1991)
Midvale Community Lead Study (Bornschem et al., 1990)
Philadelphia Neighborhood Lead Study (USDHHS, 1991)
The Arnhem, Netherlands Lead Study (Brunekreef et al.,
1981)
Belgium Lead Smelter Study (Roels et al., 1980)
Mount Pleasant Household Lead Study (Francek et al..
1994)
Baltimore Experimental Paint Abatement Study (Farfel et al.,
1991)
Traditional vs. Modified Practices of Lead Abatement in
Baltimore (Farfel et al., 1990)
Wales Environmental Lead Study (Gallacher et al., 1984)
New Orleans Day Care Center Lead Study (viverette et al.,
1996)
Omaha Study of Childhood Lead (Angle et al., 1995)
Renovation & Remodeling Study (USEPA, 1 997b,c)
HVFS Pilot Study (Roberts et al., 1993)
Throop, PA, Superfund Cleanup (Steuteville, 1990)
Exterior Dust
(Sampling Location)
/ (Entryway)
/ (Play area. Porch.
Entryway7)
/" (Entryway, Mat)
/ (Entryway)
/ (Playground,
Doormat, Pavement,
Roadside)
/ (Street)
/ (Entrance)
S (Entrance)
/ (Street)
/ (Street)
/
(School playground)
/ (Entrance')
/ (Pavement)
/ (Play area)
Air
Ducts
/
/
/
/
Window
Troughs
/
/
^
/
/
/
/
Uphol-
stery
/
/
/
Blood
/
/
/
/
/
/
/
/
/
/
/
S
/
/
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
DRAFT -- DO NOT CITE OR QUOTE
154
August 28, 2000
-------
Table 3-30. Summary of Data from Studies Where Exterior Dust Samples Were Collected for Lead Analysis
Study
Lead-Based Paint
Abatement and Repair
& Maintenance (R&M)
Study (1993-present)
University of
Rochester Lead-
in-Oust Study
(1993)
The
Comprehensive
Abatement
Performance
(CAP) Study
(1997)
BirminQham
Urban Lead
Uptake Study
(1984-1985)
EnvironmentaH.ead Measurements
Group of Homes/
Sampling Location
Entryway
R&M Unit Camp
R&MII,lmt.Camp
R&M III. Init. Camp.
Pre Abt . Init. Camp.
Mod Urb , Init Camp
Porch
Entryway
External
Combined
Entryway
Doormat
Pavement
Roadside
Sampling
Method
BRM
BRM
DVM
Wipe
BRM
DVM
Wipe
BRM
DVM
Cyclone
vacuum
Vacuum
Vacuum
Vacuum
N
25
23
26
15
15
145
150
97
42
97
97
Concentration Statistics
(units)
GM [95% Cl] (ug/g)
2219 [1218-4043)
4265 [2588-7029]
6936 [3549-13555]
2073 [1232-3488]
137 P5 -250]
GM (±2SD](ug/g)
1132 (42-30150]
557 [52-6017]
468 [19-11243]
329 [18-5967]
GM [Range] (ug/g)
237 [9-16,355]
GM [5", 95"%]
AM [Range] (ug/g)
615 [120-4300]
1436 [79-15000]
360 [127-1340]
506 [62-5100]
527 [195-1170]
805 [80-2100]
Loading Statistics (units)
GM [95% Cl] tug/ft2)
242 [109-539]
187 [102-340]
342 [184-637]
227 [76-676]
335 [188-597]
GM [±2SD](pg/ft2)
548 [7-43370]
17 [1-446]
57 [4-871]
88 [0-15881]
3 [0-124]
18 [2-215]
335 [7-17271]
18 [1-576]
GM [Range] (up/ft*)
384 [40-14021]
Blood-Lead Measurements
Group of Homes/Children
R&M Unit Camp.
R&M II. Init. Camp.
R&M III, Init Camp.
Pre Abt., Init. Camp.
Mod Urb , Init. Camp.
All Children
Levels < 10 ug/dL
Levels 2 10 ug/dL
N
33
32
33
23
19
205
157
48
Statistics (units)
GM [95%CI](ug/dL)
99 [7.9-123]
138 [112-16.9]
142 [113-16.1]
12.8 [102-161]
4.8 [38-6.1]
GM (SD](ug/dL)
7.7 [5.1]
5.5 [22]
15.1 [5.0]
GM p". 95"%] (ug/g)
11 7 (624)
O
o
2 1
•i O
I a
i s
D
Q
10
2
3
4
I
17
18
19
20
8
-------
Table 3-30. (cont.)
Study
Urban Soil Lead
Abatement
Demonstration
Project (1988-1992)
Environmental-lead Measurements
Group of Homes/
Sampling Location
_. .
Cincinnati
Mat:
Soil. Dust AbtRd 1
Soil, Dust Abt Rd.2
Soil,DustAbt.Rd.3
Soil. Dust Abt. Rd 4
Soil, Dust Abt. Rd. 5
Dust. (Soil) Abt. Rd.1
Dust, (Soil) Abt. Rd 2
Dust. (Soil) Abt Rd 3
Dust. (Soil) Abt Rd 4
Dust. (Soil) Abt. Rd 5
No Treatment Rd 1
No Treatment Rd.2
No Treatment Rd. 3
No Treatment Rd. 4
No Treatment Rd. 5
Cincinnati
Entryway
Soil, Dust Abt. Rd 1
Soil, Dust Abt. Rd 2
Soil. Oust Abt Rd 3
Soil, Dust Abt. Rd 4
Soil, Dust Abt Rd 5
Soil, Dust Abt. Rd. 6
Soil, Dust Abt. Rd. 7
Dust. (Soil) Abt. Rd 1
Dust, (Soil) Abt Rd2
Dust. (Soil) Abt Rd 3
Dust, (Soil) Abt. Rd. 4
Dust, (Soil) Abt RdS
Dust, (Soil) Abt Rd.6
Dust. (Soil) Abt Rd 7
No Treatment Rd. 1
No Treatment Rd 2
No Treatment Rd. 3
No Treatment Rd 4
No Treatment RdS
No Treatment Rd.6
No Treatment Rd. 7
Sampling
Method
Personal Air
Monitoring
Vacuum
Pump
Personal Air
Monitoring
Vacuum
Pump
N
Concentration Statistics
(units)
Med (ug/g)
109
738
549
767
659
132
939
702
722
889
100
373
349
405
332
334
606
433
491
211
382
488
425
492
468
632
102
598
615
290
367
317
286
84
317
284
Loading Statistics (units)
Blood-Lead Measurements
Group of Homes/Children
Soil. Dust Abt. Rd.1
Soil. Dust Abt. Rd.2
Soil, Dust Abt. Rd. 3
Soil, Dust AbtRd. 4
Soil. Dust Abt RdS
Soil, Dust Abt Rd.6
Soil, Dust Abt. Rd. 7
Dust. (Soil) Abt Rd 1
Dust. (Soil) Abt Rd2
Dust. (Soil) Abt. Rd. 3
Dust. (Soil) Abt. Rd 4
Dust, (Soil) Abt. Rd. 5
Dust, (Soil) Abt. Rd.6
Dust, (Soil) Abt. Rd. 7
No Treatment Rd 1
No Treatment Rd.2
No Treatment Rd 3
No Treatment Rd 4
No Treatment Rd. 5
No Treatment Rd 6
No Treatment Rd 7
N
Statistics (units)
GM (ug/dL)
8.8
.
6.9
8.8
.
82
8.7
108
.
9.3
8.6
.
76
8.9
8.3
•
5.7
6.8
7.2
7.8
•j H
4 S
3)
m
in
o>
ro
oo
ro
-------
o
•3}
Table 3-30. (cont.)
Study
Mexico City Study
(1992-1994)
Butte-Silver Bow
Envirorunonkil
Health Lead Study
(1990)
Midvale Community
Lead Study (1989)
Philadelphia
Neighborhood
Lead Study (1989)
Environmental-Lead Measurements
Group of Homes/
Sampling Location
Street
Entrance
All Locations
Location A
Location B
Location C
Location D
Location E
Location F
Location G
Enbyway
Street:
1 Block from Facility
2 Blocks from Facility
3 Blocks from Facility
4 Blocks from Facility
Sampling
Method
Broom
DVM
Vacuum
Spatula
N
200
210
141
10
7
9
21
11
11
112
Concentration Statistics
(units)
AM |iQR](ug/g)
206 [895-270]
GM (GSD](ug/g)
541 [296]
921 [259]
302 [172]
439 [356]
218 [243]
188 [226]
273 [2.56]
924 [145]
GM [[Range] (ug/g)
466 [[79-2984]
AM (ug/g)
1087
1078
907
882
Loading Statistics (units)
Blood-Lead Measurements
Group of Homes/Children
<1 8 months old
18-35 months old
35-49 months old
50 months old
Total
All Locations
Location A
Location B
Location C
Location D
Location E
Location F
Location G
LPLRichm 0-5 yrs.
Comparison 0-5 yrs
U. PI Richm 0-5 yrs.
ManayunkO-Syrs.
LPLRichm. 6-15 yrs
Comparison 6-1 5 yrs
U. PI. Richm. 6-15 yrs
Manayunk 6-15 yrs.
L PI Richm. 2 16 yrs
Comparison i16 yrs
U.Pt Richm. 2 16 yrs.
Manayunk 2 16 yrs.
L Pt Richm Total
Comparison Total
U.Pt. Richm Total
Manayunk Total
N
52
55
44
49
200
183
15
12
11
27
17
17
13
122
96
55
41
41
41
29
12
197
239
142
97
360
376
226
150
Statistics (units)
AM [SD](ug/dL)
7.38 [4.81]
10.13 [592]
11.07 [5.83]
1140 [5.76]
9.91 [578]
GM (GSD](ug/dL)
369 (1.84)
2.27 [167]
4.59 [1.89]
456 [1.79]
2.72 [1.50]
3.02 [152]
302 [1.52]
3.81 [167]
GM [Range] (ug/dL)
52 [05-145]
AM [SD](ug/dL)
97 (4.8)
9.5 (35)
9 1 (3.7)
100 (3.3)
78 (3.3)
74 (29)
6.7 (2.9)
90 (2.4)
64 (3.3)
73 (4.3)
69 (5.0)
7.7 (30)
7.7 (41)
7.8 (41)
7.4 (46)
84 (3.2)
O
o
o
o
3)
O
§
m
-------
Table 3-30. (cont.)
Study
The Amhem
Netherlands
Study (1978)
Mount Pleasant
Household Study
(1991)
Wales Environmental
Lead Study (unknown)
New Orleans Day Care
Center Lead Study
(unknown)
Environmental-Lead Measurements
Group of Homes/
Sampling Location
Street
Entrance?
Pavement
Areal
Area 2
Play area
Private inner city
Private outer aty
Public Inner aty
Public outer city
Sampling
Method
Vacuum
Wipe
Unknown
Unknown
N
42
42
30
5
4
5
5
Concentration Statistics
(units)
GM [Range] (mg/kg)
AM
690 [77-2667]
659
GM [SD]
AM [Range] (ug/g)
Med.
12 [105]
53 [0-594]
7
AM [95%CI](umol)
2 [08-35]
1 [04-4.2]
Loading Statistics (units)
Med [Range] (ug/fr3)
412 [44-690]
3 [22-80]
11 [8-33]
11 [9.5-184]
Blood-Lead Measurements
Group of Homes/Children
Areal, Children
Areal, Mother
Area2, Children
Area2, Mother
N
42
42
30
30
Statistics (units)
GM (ug/dL)
16.1
AM ISDKugWL)
16.56 [8.49]
890 [2.48]
14.70 [3.73]
8.90 [2.90]
o
o
if
1 —I
3 m
O
3)
4 o
n
7
8
tn
9 co
10
11
12
13
14
15
16
17
18
19
20
21
N = Sample size
GM = Geometric mean
AM = Arithmetic mean
Med. = Median
SD = Standard deviation
GSD = Geometric standard deviation
IQR = Interquartile range (75th percentile - 25th percentile)
BRM = Baltimore R&M vacuum method
DVM = Dust vacuum method
Rd = Sampling round
ro
-------
Table 3-31. Summary of Data from Studies Where Air Duct Dust Samples Were Collected for Lead Analysis
Study
Lead-Based Paint
Abatement and
Repair &
Maintenance
(R&M) Study
(1993-present)
The
Comprehensive
Abatement
Performance
(CAP) Study
(1993)
R&R Study
(1993-1995)
Omaha Study of
Childhood Lead
(1995)
Environmental-Lead Measurements
Group of Homes/
Sampling Location
R&M Unit. Camp
R&M II. Imt. Camp.
R&M III. Init. Camp.
Pre.Abt., Imt. Camp
Mod. Urb., Init. Camp
Denver Pilot Study
Denver Full Study
Baltimore & Denver
Omaha City
Sampling
Method
BRM
Blue Nozzle
CAPS
Cyclone
Wipe
Unknown
N
1
12
15
1
0
10
109
21
21
Concentration Statistics
(units)
GM [95%CI](ug/g)
1445 [617-3,388]
1491 [945-2,354]
GM [Range] (ug/g)
749(363-1.699]
427 [59-5,640]
Loading Statistics (units)
GM [95%CI](|ig/ft2)
51405 [33,671-78.480]
30046 [18.399-49,066]
GM [Range) (ug/ft2)
308 [27-3,910]
120(2-40.900]
GM [Range] Gig/ft7)
2.900 [205-30,900]
AM (ug/g)
383
Blood-Lead Measurements
Group of
Homes/Children
R&M I, Imt. Camp
R&M II. Init. Camp
R&M III. Imt. Camp
Pre Abt., Init. Camp.
Mod Urb.. Imt Camp
N
33
32
33
23
19
Statistics
(units)
GM [SD](ug/dL)
9.9 [79-12.3]
138 [112-16.9]
142 [113-16.1]
128 [10.2-161]
48 [38-6.1]
8
* o
1 H
3 m
4 O
5 =
6 o
,0
11
12
13
14
ic tn
1 j co
16
17
18
19
20
21
22
23
24
25
26
27
N= Sample size
GM = Geometric mean
AM = Arithmetic mean
Med. = Median
SD = Standard deviation
BRM = Baltimore R&M vacuum method
DVM = Dust vacuum method
Rd = Sampling round
>
-------
o
?
o
o
Q
i H
3 m
4 O
5 *
6 o
i c
8
Table 3-32. Summary of Data from Studies Where Window Trough Dust Samples Were Collected for Lead Analysis
Study
Lead-Based Paint
Abatement and
Repairs
Maintenance
(R&M) Study (1993-
present)
University of
Rochester Lead-In-
Dust Study (1993)
The National Survey
of Lead-Based
Paint m Housing
(1989-1990)'
The HUD Lead-
Based Paint
Abatement
Demonstration
(1990-1993)
The Comprehensive
Abatement
Performance (CAP)
Study (1993)
Environmental-Lead Measurements
Group of Homes/
Sampling Location
R&M I. Init Camp
R&M II, mil. Camp
R&M III. Init. Camp.
Pre Abt , Init. Camp
Mod Urb., Init. Camp
Rochester. NY
Homes Built Prior to 1940
Homes Built 1940-1959
Homes Build 1960-1979
All Surveyed Homes
Homes Built Prior to 1940
Homes Built 1940-1959
Homes Build 1960-1979
AD Surveyed Homes
Albany
Cambridge
Omaha
Sampling
Method
BRM
BRM
DVM
Wipe
Blue Nozzle
Wipe-
Equivalents
Wet Wipe
Cyclone
Vacuum
N
43
45
54
31
30
77
87
120
284
77
87
120
284
98
119
161
98
Concentration Statistics
(units)
GM (95%CI](ug/g)
22144 (15.091-32.495]
20462 [15.106-27.717]
21600 [12,751-36.590)
2251 [1, 247-4.062]
338 [239-479]
GM[i2SD](ug/g)
6114 [65-579.533]
1709 [17-171.081]
GM (Range]
(M9/9)
8389 [1898-74980]
1972 [5.2-41429]
1016 [192-17725]
1965 [52-74980]
Loading Statistics
(units)
GM [95% Cl| dig/ft2)
7051 [4,896-10,156]
9900 [7,245-13,529]
13916 [10,104-19,167]
802 [501-1,284]
1021 [515-2,024]
GM (±2SD](ugrtr?)
11874 [26-5,365.819]
370 [3-45.177]
2759 [29-264.752]
GM [Range]
(MO/lf)
929 [16-23798]
140 [03-5312]
110 [004-3244]
177 [0.04-23798]
2017 [87-33391]
389 [19-9446]
296 [0.4-6169]
460 [04-33391]
> 800 ugfff [#][%]
[2] [20%]
[5] [42%]
[0][0%]
GM [Range] fog/ft2)
25156(191-244581]
Blood-Lead Measurements
Group of
Homes/Children
R&M 1. Init Camp.
R&M II, Init Camp
R&M III, Init Camp.
Pre Abt., Init. Camp.
Mod. Urb, Init Camp.
All Children
Levels < 10 ug/dL
Levels * 10 ug/dL
N
33
32
33
23
19
205
157
48
Statistics
(units)
GM (SD](ug/dL)
99 [79-12.3]
13.8 [11.2-16.9]
14.2 [11.3-16.1]
12.8 (10.2-16.1)
4.8 [3.8-6.1]
AM [SD](ug/dL)
77 (5.1]
5.5 [22]
15 1 [5.0]
o>
o
21 >
22 |
23 5
24
-------
o
3J
Table 3-32. (cont.)
Study
Baltimore
Experimental Paint
Abatement Study
(1986-1987)
Traditional vs.
Modified Practices
of Lead Abatement
in Baltimore (1984-
1985)
Envlronmer
Group ol Homes/
Sampling Location
NewWndw..Pre-Abt.
NewWndw.Pst-Abt
NewWndw..Pst-Trt
NewWndw..Pst C-u
NewWndw.lm Pst
NewWndw.Sm Pst
NewWndw.,6-9m Pst
On-sit C S . Pre-Abt
On-sit C.S.,Pst-Abt.
On-sitC S.Pst-Trt
On-sit. C. S , Pst C -u
On-sit.C.S.,1m.Pst
On-sit. C.S., 3m. Pst
On-sit C. S.,6-9m Pst.
Traditional
Modified
Sampling
Method
Wipe
Wet Wipe
N
11
13
13
14
10
13
14
17
16
12
17
15
15
10
Concentration Statistics
(units)
Loading Statistics
(units)
GM (iig/ft2)
15663
2238
1212
466
186
466
466
44566
29462
5687
1585
5314
7179
2984
GM big/ft2)
15570
18367
Blood-Lead Measurements
Group of Statistics
Homes/Children N (units)
o
m
O
31
i
2
3
4
i
10
11
12
1!
\l
1 Area-weighted household averages are summarized in this table for this study, using sampling weights modified in the §403 risk analysis to represent the 1997 housing stock
Wipe-equivalents represent converting the Blue-Nozzle dust-lead loadings to wipe-equivalent loadings as documented in Chapter 4 of the §403 risk analysis report. Data were
Imputed for surveyed homes without data (see Chapter 3 of the §403 nsk analysis report).
N = Sample size
GM = Geometric mean
AM = Arithmetic mean
Med. = Median
SD = Standard deviation
BRM = Baltimore R&M vacuum method
DVM = Dust vacuum method
Rd = Sampling round
8
-------
Table 3-33. Summary of Data from Studies Where Upholstery Dust Samples Were Collected for Lead Analysis
Study
Lead-Based Paint Abatement
and Repair & Maintenance
(R&M) Study (1998-present)
Mexico City Study
(1992-1994)
HVFS Pilot Study (1993)
Throop, PA Superiund
Cleanup (1989-1990)
Environmental-Lead Measurements
Group of Homes/
Sampling Location
R&M Unit Camp
R&MII,lnit.Camp.
R&M III. Imt. Camp.
Pre Abt, mil. Camp.
Mod. Utb.lnit Camp
Seattle
Throop, PA Pre-Clean
Throop, PA Post-Clean
Sampling
Method
HVFS
Vacuum (15 21
mm air pump)
Vacuum (1 5 21
mm air pump)
N
23
7
0
14
16
5
5
5
Concentration Statistics
(units)
GM [95%CI](Mg/g)
699 [493-992]
700 [180-2,722]
503 [353-718]
142 [101-200]
AM [Range] ug/g
229 [130-380]
Loading Statistics
(units)
GM [95%CI](ug/fl2)
95 [49-186]
92 [39-218]
101 [55-186]
61 [35-138]
AM [SD] (ug/fP)
70 011 [019]
AM (Range) (jg/ft2
27.8 [27-94.9]
AM [Range] ug/fP
285 [199-343]
231 [137-355]
Blood-Lead Measurements
Group of
Homes/Children
R&M l.lnit Camp.
R&MII,lnit.Camp.
R&M III, Imt. Camp.
Pre. Abt, (nil Camp.
Mod. Urb., Imt. Camp.
<18 months old
18-35 months old
35-49 months old
50 months old
Total
N
33
32
33
23
19
52
55
44
49
200
Statistics (units)
GM (95%CI)(pg/dL)
9.9 [7.9-123)
13.8 [11.2-169]
14.2 [11.3-16.1]
12.8 (10.2-16.1)
4.8 [3.8-6.1]
AM (SD](ug/dL)
7.38 [4.81]
10.13 [5.92]
11.07 [5.83J
11.40 [576]
991 [578]
o
o
2 1
1 H
3 m
4 O
5 *
m
6
11
12
13
4
i
N = Sample size
GM = Geometric mean
AM = Arithmetic mean
Med. = Median
SD = Standard deviation
BRM = Baltimore R&M vacuum method
DVM = Dust vacuum method
8
-------
1 respectively. Included in the tables are sampling location, dust collection methods used, number
2 of dust samples taken, and distribution statistics available from the literature on dust-lead
3 concentrations and loadings. Also presented in the tables are simple summaries of blood-lead
4 concentrations for those studies which also sampled blood.
5 Exterior Dust
6 Table 3-30 summarizes the data from the fourteen studies in which exterior dust samples
7 were collected and analyzed for lead content. In addition, there are also results for several
8 studies in which dust samples were collected inside the home in the entryway. These can
9 occasionally be good representations of exterior dust samples. The summary consists of
10 descriptive statistics for the dust-lead measures and also for blood-lead measures, when
11 available. Summary statistics are presented separately for different study groups and different
12 sampling periods, where appropriate. As indicated in Table 3-29, exterior dust samples were not
13 collected at a consistent location across all studies. Sampled locations included the house
14 entrance (including doormats and porches), street, and the child's play area. Likewise, samples
IS were collected by a variety of methods including surface scraping, vacuum sampling, and wipe
16 sampling. The sampling location and method have a significant impact on the lead loading and
17 concentration estimates.
18 Overall, even with the variability in sampling location and method, Table 3-30 indicates
19 the potential for significant amounts of lead in exterior dust, with concentrations often exceeding
20 400 ug/g and loadings often exceeding 100 ug/ft2.
21
22 Air Ducts
23 As indicated in Table 3-31, only four studies were identified that contained information
24 on lead levels in dust within air ducts in residential housing. While only limited information was
25 encountered, a consensus across studies was that air ducts can contain high amounts of dust and
26 lead. This was due partially to the general lack of cleaning of air ducts over time and the ability
27 of lead particles to enter air ducts from outside of the unit via ventilation filters.
28 In units with a potential for containing lead hazards, dust-lead loadings in air ducts
29 typically exceeded 100 fig/ft2, with individual samples often exceeding 1,000 ug/ft2. Lead levels
30 can vary considerably among dust samples within the same unit and in different units. Older
31 ductwork and HVAC systems, as well as vacant units in which no cleaning is performed and
32 HVAC systems may not be used, tend to have high dust loadings, and therefore, higher dust-lead
33 loadings when a lead source is present. Several methods were used across studies to collect dust
34 in air ducts. As air ducts often have metal surfaces, issues concerning static electricity must be
35 considered when sampling dust from air ducts.
36 The EPA Comprehensive Abatement Performance (CAP) study involving occupied
37 homes assumed to be free of lead-based paint for at least two years, provided the greatest amount
38 of information on lead in dust within air ducts; levels were relatively low in this study compared
DRAFT--DO NOT CITE OR QUOTE 163 August 28,2000
-------
1 to the others. Nevertheless, in a typical housing unit in the CAP study, average dust-lead
2 loadings from air ducts exceeded all other sampled surfaces except for window troughs and
3 entryways. In general, air duct dust-lead levels in the Baltimore R&M study and the Renovation
4 and Remodeling study were considerably higher than in the CAP study, as these studies included
5 older, vacant units in need of repair and maintenance.
6 The relative sparsity of published information indicates that many open questions exist on
7 the nature of lead contamination of dust within residential air ducts and whether the lead in this
8 dust is available for exposure to residents (especially children). Nevertheless, evidence exists
9 that air ducts can contain some of the highest levels of lead in dust within a housing unit.
10 Window Troughs
11 Table 3-32 presents summary information on seven studies that sampled dust-lead levels
12 in window troughs (also known as window wells). The HUD Grantees evaluation is a significant
13 data source on window trough dust-lead levels in high risk housing that is not included in Table
14 3-32, as these data are still being collected and reported.
15 In general, partially because of the published data summarized in Table 3-32, and
16 partially because a standard for window troughs has been historically used in risk assessments
17 and to determine clearance (following EPA's Interim Guidance for §403 standards and the HUD
18 Guidelines), window troughs are widely recognized as a major reservoir of dust-lead in
19 residences. As shown in Table 3-32, levels often exceed 800 ug/ft2 and it is not uncommon to
20 see levels above 10,000 ug/ft2 in high risk housing. However, unlike the other surfaces discussed
21 in this report, national estimates of the distribution of dust-lead in window troughs are available
22 from the HUD National Survey. The estimated national geometric mean dust-lead loading in
23 window troughs from the HUD National Survey (as modified in the §403 risk analysis to reflect
24 the 1997 housing stock and wipe techniques) was 460 ug/ft2 (Table 3-32), with 30% of homes
25 estimated to have average window trough dust-lead loadings at or above 800 ug/ft2.
26 Upholstery
27 Table 3-33 summarizes the data from the four studies which collected dust-lead samples
28 from upholstery. In general, dust-lead loadings for these surfaces averaged below 100 ug/ft2.
29 As the sample sizes in all of these studies were small, and sampling techniques, sampling
30 locations, and study goals varied considerably from study to study, more information would be
31 necessary to fully characterize potential lead hazards associated with upholstery.
DRAFT -- DO NOT CITE OR QUOTE 164 August 28, 2000
-------
1 3.5.2 Evidence of a Relationship Between Children's Blood-Lead
2 Concentrations and Dust-Lead on Surfaces Other Than
3 Floors and Window Sills
4 The information available to assess children's exposure to dust-lead on surfaces covered
5 in this section is discussed in detail for each surface type below. It should be noted, however,
6 that in general it is difficult to establish the causal link between these surfaces and children's
7 blood-lead concentrations. This is true for many reasons. Often other important sources of lead
8 exposure are not well characterized in the studies that provide data on these special surfaces.
9 Correlations are often estimated based on small sample sizes and without adjusting for other
10 exposure variables such as lead in floor-dust and soil. Moreover, there is often correlation
11 between lead levels on these surfaces and lead levels on floors, window sills, and in soil. For all
12 of these reasons, it must be noted that even significant correlation coefficients should not be
13 interpreted as the degree to which dust-lead on these surfaces causes a change in blood-lead
14 concentration. In almost all cases, in order to characterize the pathway of lead from these
IS surfaces to children's blood, additional data collection or analyses are needed.
16 Exterior Dust
17 Table 3-30 (in the previous subsection) contains a summary of the dust-lead data and
18 blood-lead data separately for studies which collected exterior dust. However, it contains no
19 results providing information about the relationship between exterior dust-lead levels and blood-
20 lead levels. The reports describing these analyses were examined to assess the relationship
21 between exterior dust-lead levels and blood-lead levels. In the Repair and Maintenance Study
22 (USEPA, 1997a,e), exterior dust samples were collected at five separate times. The (Pearson)
23 correlation coefficients between blood-lead concentrations and entry way dust-lead concentrations
24 ranged from 0.23 to 0.49, and the correlation coefficients between blood-lead concentrations and
25 entryway dust-lead loadings ranged from 0.10 to 0.46. In most cases, those correlation
26 coefficients were statistically significant at the oc=0.01 level. In the Rochester Study (USHUD,
27 1995a), the University of Cincinnati Dust Vacuum Method (DVM) and Baltimore Repair and
28 Maintenance vacuum method (BRM) were used to collect external dust samples. The correlation
29 coefficients for the BRM and DVM techniques were 0.21 and 0.27 for the correlation between
30 blood-lead concentrations and exterior dust-lead concentrations and 0.34 and 0.18 for the
31 correlation between blood-lead concentrations and exterior dust-lead loadings, respectively. The
32 correlations were all statistically significant at the oc=0.05 level, with the correlation for the DVM
33 measured loading significant at the oc=0.01 level. On the other hand, the Mexico City Study
34 (Romieu et. al., 1995), Arnhem Study (Brunekreef et. ah, 1981) and the Midvale Study
35 (Bomschein et. al., 1990) reported the correlations between external dust measurements and
36 blood-lead levels to be statistically insignificant.
37 Multivariate regression and structural equation modeling was used in some of the studies
38 to examine how multiple sources of environmental lead exposure and other factors affect blood-
39 lead levels. Regression analyses were carried out in most of the studies where both external dust-
40 lead and blood-lead measurements were collected, but the external dust-lead measurements were
DRAFT -- DO NOT CITE OR QUOTE 165 August 28, 2000
-------
1 not included as an explanatory variable in any of the reported regression models. Reasons for
2 excluding external dust-lead were not clearly stated. Speculatively, such reasons may include a
3 lack of interest in the relationship, poor data quality in the external dust-lead measurements,
4 colinearity of external with internal dust-lead measurements and omission of the variable through
5 step-wise regression. Structural equation modeling was carried out in the (Three Cities) Urban
6 Soil Lead Abatement Demonstration Project (USEPA, 1996a), Butte-Silver Bow Study (Butte-
7 Silver Bow DoH et. al., 1991) and Midvale Study (Bornschein et. al., 1990). In the Three Cities
8 Study, exterior dust was lead considered as a component of the lead exposure pathway in the
9 general structural equation model, but the component was excluded in the actual implementation
10 of the model. In the Butte-Silver Bow Study, external dust-lead was also excluded from the
11 structural equation model, but external dust-lead was observed to be correlated (Pearson
12 correlation r=0.64) with soil lead, which was included as a component of the lead exposure
13 pathway in the model. The Midvale Study was the only one to include external dust-lead in the
14 actual implementation of the structural equation model, but the dust-lead to blood-lead
15 relationship was reported as being statistically insignificant.
16 In summary, there is much difficulty in distinguishing between direct and indirect
17 exposure in cases where external dust-lead levels is closely related to levels in other sources of
18 environmental lead. Correlation and univariate regressions with external dust-lead and blood-
19 lead fail to account for the possibility that external dust-lead by itself may only play a small part
20 in aggregate lead exposure when other sources of lead and exposure pathways are considered.
21 Multivariate regressions using lead measurements from multiple sources do not solve this
22 problem due to problems with colinearity. The preferred approach would be to use structural
23 equation models, which allow multiple source and exposure pathways to be modeled in a
24 reasonable way, but this approach requires more effort in terms of implementation and
25 interpretation of the model, and is not well-reported in the literature. Therefore, quantitative
26 estimates of the effect of external dust-lead on children's blood-lead concentrations have not
27 been well established in the literature.
28 Air Ducts
29 Most of the encountered articles provided only preliminary information on lead exposures
30 associated with air ducts. It is unclear to what extent dust-lead in air ducts is accessible to
31 children. Children would not typically be expected to encounter the dust lodged in air ducts
32 directly. One case study found that dust-lead levels in living areas outside of contaminated air
33 ducts can be orders of magnitude lower than what is found in the air ducts. However, if dust in
34 air ducts is disturbed, it is more likely to be introduced to the air and to nearby surfaces with
35 which children can come into direct contact. In particular, HVAC ductwork removal can yield
36 extensive contamination of surfaces in the general area of the ductwork.
37 Only one study (the Baltimore R&M Study) estimated (in a quantitative manner) the
38 association between blood-lead concentrations in children and dust-lead levels found in air ducts.
39 This relationship was expressed as a simple correlation coefficient. Unlike correlations between
40 blood-lead concentrations and dust-lead levels on other surfaces, the correlation coefficient
DRAFT - DO NOT CITE OR QUOTE 166 August 28, 2000
-------
1 involving dust-lead levels from air ducts was not significant at the 0.05 level. However, this
2 analysis was based on a small sample size and did not adjust for the effects of other exposure
3 variables such as lead in floor-dust and soil. Moreover, as evidence of a significant correlation
4 was observed between air duct dust-lead levels and lead levels on other surfaces, such as floors,
5 even significant correlation coefficients should not be interpreted as the degree to which air duct
6 dust causes a change in blood-lead concentration. In order to characterize the pathway of lead
7 from air ducts to children's blood, additional data collection and analyses are needed.
8 Window Troughs
9 Of the seven studies listed in Table 3-32 above that collected information on dust-lead in
10 window troughs (also known as window wells), three also collected blood-lead data from
11 resident children. Correlation coefficients between blood-lead levels and window-trough dust-
12 lead concentrations in the R&M study ranged from 0.20 to 0.39, and correlation coefficients
13 between blood-lead levels and window-trough dust-lead loadings ranged from 0.06 to 0.44. The
14 correlations between dust-lead concentrations and blood-lead concentrations were statistically
15 significant in 4 of the 5 sampling campaigns, and the correlation between dust-lead loading and
16 blood-lead concentration was statistically significant only in the pre-maintenance sampling. In
17 the Rochester Study, correlation coefficients between blood-lead concentrations and dust-lead
18 loadings were 0.35 for the BRM samples, 0.31 for the DVM samples, and 0.29 for wipe samples,
19 while correlation coefficients between blood-lead concentrations and dust-lead concentrations
20 were 0.23 for both BRM and DVM samples.
21 Multiple analyses (to be documented in a companion §403 report) have been conducted to
22 assess the relative contribution of window troughs to blood-lead concentrations after accounting
23 for the effect of floor-dust, window sill-dust, and soil. Others (USEPA, 1996b; USEPA, 1996c)
24 have examined whether the predictive ability of a model improves when adding window trough
25 lead levels to a model which already accounts for dust-lead on floors and window sills. Results
26 of these analyses on the Rochester Lead-in-Dust Study and Baltimore R&M study data indicated
27 that the estimated effect of window trough dust-lead on blood-lead was either not statistically
28 significant or only marginally significant after adjusting for the effects of floor lead, sill lead, and
29 temporal variation. A pathways analysis (USEPA, 1999) using structural equations modeling
30 concluded that window troughs were a significant pathway for lead exposure, both as a direct
31 pathway of lead to children's blood-lead concentration (Rochester Lead-in-Dust Study) and as an
32 indirect pathway through window sills and floors to blood-lead concentrations (Baltimore R&M
33 Study).
34 In summary, the association between blood-lead concentrations and window trough dust-
35 lead has been well established in the literature. The more difficult question of the degree to
36 which window troughs contribute directly or indirectly to children's lead exposure is not well
37 established.
DRAFT - DO NOT CITE OR QUOTE 167 August 28,2000
-------
1 Upholstery
2 Table 3-33 in the previous subsection included results for two studies which measured
3 both children's blood-lead concentrations and dust-lead on upholstery. In the Baltimore R&M
4 study, correlation coefficients were calculated between blood-lead concentrations in children and
5 both the loading and concentration of lead in upholstery dust. These correlation coefficients
6 ranged from 0.19 to 0.61 for dust-lead concentrations and from 0.06 to 0.47 for dust-lead
7 loading. These correlations were statistically significant in the pre-intervention sampling. As
8 with most of the other surfaces discussed in this section, upholstery dust-lead levels were not
9 included in any analyses to determine which lead sources were most significantly related to
10 blood-lead levels. In the Mexico City Study, the correlation between upholstery dust-lead levels
11 was not statistically significant, resulting in the absence of upholstery dust-lead levels from
12 models linking blood-lead levels and environmental lead levels.
13 The results of the two studies assessing the importance of upholstery dust as a source for
14 lead exposure in children differ. In one case, the relationship between blood-lead and upholstery-
15 dust-lead is significant, while in the other it is not. Moreover, as upholstery dust-lead is often
16 correlated with other lead exposure variables, such as floor dust-lead and soil-lead, as cautioned
17 earlier, the positive correlation coefficient should not be interpreted as the degree to which
18 upholstery dust causes a change in blood-lead concentration. In order to characterize the
19 pathway of lead from upholstery to children's blood (and perhaps hands), additional data
20 collection and analyses are needed.
21 3.5.3 Implications of the Available Information For Regulatory Standards
22 Two primary questions related to the need and feasibility of regulatory standards for dust-
23 lead on surfaces other than floors and window sills are:
24 1. Is there sufficient information available on which to base a standard?
25 2. Is the standard necessary to either identify a lead hazard at a residence or to
26 characterize the risk to determine appropriate corrective actions?
27 The answers to these questions are discussed for each surface type below.
28 Exterior Dust
29 In general, there is a fair amount of data on exterior dust, including studies where exterior
30 dust has been measured along with other lead exposure variables and blood-lead concentrations.
31 The amount of data implies that analyses could be conducted to provide a quantitative basis for
32 an exterior dust standard. However, implementation and interpretation of such analyses for
33 exterior dust will face many difficulties. For example, in many of the studies it is difficult to
34 distinguish between exterior dust and soil samples because of aggregation of the samples or of
35 the measurements. Some external sampling for lead was carried out using surface scrapings
DRAFT--DO NOT CITE OR QUOTE 168 August 28,2000
-------
1 which measures lead levels from a mix of both soil and dust-lead and some analyses averaged the
2 external soil and dust measurements and recorded the value as a single external lead
3 measurement. (Hence only studies for which a clear distinction between external soil and dust-
4 lead levels is possible were included in this summary.) It is also difficult to determine what
5 locations for exterior dust should be included. Should the focus be on enclosed spaces or also
6 include unenclosed areas such as sidewalks, stoops, and unenclosed porches? One primary
7 reason to focus only on enclosed areas is because exposures to unenclosed areas are not under the
8 direct control of property owner. Exposure and cleaning scenarios for enclosed versus
9 unenclosed areas are likely to be very different as well. In conclusion, decisions on the specific
10 focus of a standard for exterior dust would impact the feasibility of establishing a good
11 quantitative basis on which to set the standard.
12 The question of whether a standard for exterior dust is necessary is also a difficult one,
13 for which the literature does not have a clear answer. While it is reasonable to assume that
14 measurements of lead in interior dust and exterior soil might capture a lead hazard if one exists,
15 there is not a strong body of information on which to base this conclusion. A separate standard
16 may not be necessary if risk assessors are aware of the potential hazard from exterior dust, and
17 include testing or corrective actions in cases where it is suspected to be an important pathway of
18 exposure (for example, in the case where a child spends a considerable amount of time on a
19 paved surface, such as a driveway or patio).
20 Air Ducts
21 There is insufficient data upon which to develop a hazard standard for lead in air duct
22 dust, or upon which to draw conclusions about the necessity of a standard to either identify a
23 hazard or determine corrective actions.
24 Window Troughs
25 The fact that regulatory standards have been proposed for dust-lead on floors and window
26 sills based on data sets (most notably the Rochester Lead-in-Dust Study and the HUD National
27 Survey) that also include window troughs implies that sufficient data exists on which to base a
28 standard for window troughs.
29 However, while there is sufficient information on which to base a standard, analyses
30 conducted to assess the necessity of a window trough standard given the existence of a floor and
31 window sill standard suggest that a window trough standard may not be necessary to identify a
32 residence with a lead-based paint hazard. These analyses include the sensitivity/specificity
33 analyses included in a companion §403 report as well as the analyses that examine the effect of
34 adding window sills to a statistical model that already includes floors and window sills (USEPA,
35 1996b; USEPA, 1996c). Given the correlation between window trough and window sill lead
36 levels, it is likely that if more sampling is to be done beyond a minimal risk assessment, more
37 benefit will be obtained from sampling more windows at the sill rather than sampling fewer
38 windows but at both the sill and trough. Moreover, cleaning of window troughs is
DRAFT -- DO NOT CITE OR QUOTE 169 August 28, 2000
-------
1 recommended for all homes that require a dust intervention, and clearance standards have been
2 proposed to guide assessment of the effectiveness of the cleaning. For these reasons, it does not
3 appear that an additional standard for window troughs is necessary either to identify a home with
4 a hazard or to guide corrective actions.
5 Upholstery
6 There is insufficient data upon which to develop a hazard standard for lead in upholstery
7 dust, or upon which to draw conclusions about the necessity of a standard to either identify a
8 hazard or determine corrective actions.
9 3.6 DISTRIBUTION OF CHILDHOOD BLOOD-LEAD
10 This section updates the information presented in Section 3.4 of the §403 risk analysis
11 report on the distribution of childhood blood-lead concentration in the United States, with a focus
12 on the 1-2 year (12-35 month) age range as the population of interest. In addition to a national
13 characterization based on data from Phase 2 of the Third National Health and Nutrition
14 Examination Survey (NHANES HI), Section 3.4 of the §403 risk analysis report summarized data
IS from other studies (e.g., the Baltimore R&M study, the Rochester Lead-in-Dust study, and the
16 HUD Grantees evaluation) to provide supporting information on the prevalence of elevated
17 blood-lead concentrations in children living in urban locations and in older housing or housing
18 likely to contain lead-based paint. Blood-lead data from these other studies were also considered
19 because the NHANES HI did not collect environmental-lead data, despite having the most
20 nationally representative data on blood-lead levels.
21 Section 3.6.1 below is an update of Section 3.4.4 of the §403 risk analysis report. It
22 contains revised data summaries of pre-intervention blood-lead concentrations in children
23 monitored within the HUD Grantees evaluation and revised regression model fits to predict
24 blood-lead concentration as a function of dust-lead loading for each individual grantee, as well as
25 for the Rochester Lead-in-Dust study (i.e., the study that provided the data used to developed the
26 empirical model developed within the §403 risk analysis. These revisions were possible as
27 additional pre-intervention data from the HUD Grantees evaluation (through 1/99) have been
28 made available to the risk analysis since the report was released.
29 Section 3.6.2 provides information from the Cincinnati Prospective Lead study (Clark et
30 al., 1985) and summarized by the Centers for Disease Control and Prevention (CDC) on the
31 relationship between children's blood-lead concentration and housing age/condition, and how
32 this relationship may change with the age of the child (CDC, 1991). CDC used this information
33 in their recommendations for blood-lead screenings of young children.
DRAFT - DO NOT CITE OR QUOTE 170 August 28. 2000
-------
1 3.6.1 Evaluation of the HUD Lead-Based Paint Hazard Control
2 Grant Program ("HUD Grantees'")
3 Blood-lead concentrations of children residing in households participating in the
4 evaluation phase of the HUD Grantees evaluation (Section 3.2.2.3 of the §403 risk analysis
5 report) were measured, along with environmental-lead levels in various media. The population
6 of children targeted for participation in the program differed among the fourteen grantee
7 recipients, due to the different enrollment criteria among the grantees (see Table 3-4 of the §403
8 risk analysis report). These criteria included targeting high-risk neighborhoods, enrolling only
9 homes with a lead-poisoned child, and considering unsolicited applications. Pre-intervention
10 data collected through January 1999 are presented in this section; these data provide some of the
11 most recent information on the relationship between children's blood-lead concentration and
12 environmental-lead levels.
13 Across all grantees, pre-intervention blood-lead concentration data through 1/99 were
14 available for 526 children aged 1-2 years and for 764 children aged 3-5 years. For these children,
15 Table 3-34 summarizes measured blood-lead concentration for each combination of blood
16 collection type (venipuncture, fingerstick) and age of child (1-2 years, 3-5 years, and 1-5 years).
17 Table 3-34 also summarizes measured blood-lead concentration for children aged 1-2 years for
18 each combination of blood collection type and grantee. Note that fingerstick methods were
19 predominant for Wisconsin, Milwaukee, and Vermont, while Rhode Island used both methods
20 for similar numbers of children. The remaining nine grantees (excluding New Jersey) used the
21 venipuncture either exclusively or predominantly.
22 According to Table 3-34, the geometric mean blood-lead concentration via the
23 venipuncture collection method was 9.3 ug/dL for children aged 1-2 years and 8.0 ug/dL for
24 children aged 3-5 years. In contrast, the geometric means based on data from Phase 2 of
25 NHANES m were 3.1 ug/dL for children aged 1-2 years and 2.5 ug/dL for children aged 3-5
26 years (Table 3-36 of the §403 risk analysis report). The larger values in the HUD Grantees
27 evaluation reflect the HUD Grantees program's procedure of selecting high-risk children for
28 monitoring. The differing enrollment criteria across grantees also contributed to considerable
29 differences in the geometric mean blood-lead concentration among the grantees.
30 Under venipuncture, the geometric means of children aged 1-2 years for individual
31 grantees reporting more than one blood-lead result ranged from 4.2 ug/dL (California, which
32 only targeted older units) to 15.9 ug/dL (Cleveland, which targeted units with lead-poisoned
33 children).
34 The geometric mean blood-lead concentration via the fingerstick collection method was
35 9.6 ug/dL for children aged 1-2 years and 8.7 ug/dL for children aged 3-5 years. When data were
36 available for more than one child under fingerstick collection methods, the geometric means for
37 children aged 1-2 years ranged from 5.9 ug/dL (Wisconsin) to 13.5 ug/dL (Milwaukee).
DRAFT -- DO NOT CITE OR QUOTE 171 August 28, 2000
-------
1 Table 3-34. Summary of Children's Pre-lntervention Blood-Lead Concentration in the
2 HUD Grantees Evaluation According to Blood Collection Method, Child Age
3 Category, and Grantee (ages 1 -2 years only)
Age Category
1-2 Years
3-5 Years
1-5 Years
Grantee
Alameda County
Baltimore
Boston
California
Cleveland
Massachusetts
Minnesota
New Jersey
Rhode Island
Wisconsin
Milwaukee
Chicago
New York City
Vermont
Age Range
1-2 Years
3-5 Years
1-5 Years
Grantee
Cleveland
Massachusetts
Minnesota
Rhode Island
Wisconsin
Milwaukee
Vermont
Number
of
Children
Blood-Lead Concentration (pg/dL)
ArithniBtic
Mean
Geometric
Mean
Geometric
Standard
Deviation
Minimum
Blood Collection Method =
361
534
895
12.5
10.7
11.4
9.3
8.0
8.5
2.3
2.2
2.2
0.7
0.0
0.0
25th
Percentile
Median
75th
Percentile
Maximum
= Venipuncture
5.4
4.4
5.0
10.0
8.9
9.0
17.0
15.0
16.0
53.0
48.0
53.0
Blood Collection Method = Venipuncture (Children Aged 1-2 Years only)
27
24
20
21
64
44
75
1
13
9
4
28
23
8
165
230
395
6.5
9.3
12.7
5.3
19.3
11.0
14.5
3.0
10.4
10.2
21.0
13.9
5.2
13.5
12.2
11.2
11.6
4.7
7.6
10.4
4.2
15.9
9.0
10.7
3.0
9.0
8.7
17.5
11.7
4.7
12.4
2.2
1.9
2.0
2.0
1.9
1.9
2.4
-
1.8
1.8
2.1
2.0
1.6
1.6
1.4
2.0
3.0
1.4
4.0
3.0
0.7
3.0
3.0
4.0
6.0
1.0
2.0
6.0
Blood Collection Method
9.6
8.7
9.1
2.0
2.0
2.0
2.0
2.0
2.0
3.0
6.0
6.0
3.2
11.5
6.0
6.0
3.0
7.0
6.0
12.0
9.5
4.0
8.5
4.7
7.0
14.5
3.8
17.0
9.0
11.0
3.0
9.0
8.0
21.5
12.0
5.0
14.5
6.6
10.0
19.0
6.0
28.0
15.5
22.0
3.0
14.0
12.0
30.0
19.0
7.0
17.0
24.8
26.0
27.0
16.9
53.0
40.0
43.0
3.0
21.0
24.0
35.0
35.0
12.0
22.0
= Fingerstick
6.0
5.0
5.0
9.0
8.5
9.0
15.0
15.0
15.0
62.0
49.0
62.0
Blood Collection Method = Fingerstick (Children Aged 1-2 Years only)
1
4
1
9
42
84
24
13.0
7.8
33.0
8.8
6.3
16.7
7.9
13.0
7.3
33.0
8.2
5.9
13.5
7.0
-
1.5
--
1.5
1.4
2.0
1.6
13.0
4.0
33.0
5.0
3.5
2.0
3.5
13.0
6.0
33.0
7.0
4.0
9.0
5.0
13.0
8.5
33.0
7.0
6.0
14.5
6.5
13.0
9.5
33.0
11.0
8.0
20.5
11.0
13.0
10.0
33.0
15.0
14.0
62.0
16.0
Note: All pre-intervention blood-lead concentration data available and collected through 1/99 are included in the above
summaries.
DRAFT ~ DO NOT CITE OR QUOTE
172
August 28, 2000
-------
1 The percentages of children with elevated blood-lead concentrations (i.e., concentrations
2 at or above 10, 15, 20 or 25 ug/dL) at pre-intervention are summarized in Table 3-35. According
3 to this table, 51 percent of children aged 1-2 years sampled via venipuncture methods had blood-
4 lead concentrations at or above 10 ug/dL, compared to the estimates of 5.88% for Phase 2 of
5 NHANES m, 53.8% for the Baltimore R&M study (pre-intervention), and 23.4% for the
6 Rochester Lead-in-Dust study (Tables 3-37, 3-41, and 3-42, respectively, of the §403 risk
7 analysis report). For individual grantees having more than one child with a measured blood-lead
8 concentration, the percentage of children aged 1-2 years with blood-lead concentrations
9 (venipuncture) at or above 10 ug/dL varied from 4% (New York City, which targeted housing
10 and neighborhoods rather than lead-poisoned children) to 80% (Cleveland). The range of
11 percentages under the fmgerstick method were similar to that under the venipuncture method, but
12 less data were available to estimate them.
13 Figures 3-20 and 3-21 illustrate the nature of the linear relationship observed in the HUD
14 Grantees evaluation between a child's (log-transformed) blood-lead concentration and the
15 household's (log-transformed) area-weighted arithmetic average wipe dust-lead loading for floors
16 and window sills, respectively. The figures portray fitted linear regression models for each
17 grantee, as well as for the Rochester Lead-in-Dust study and, in Figure 3-21, the Baltimore R&M
18 study (for comparison purposes). The regression model used only the log-transformed average
19 dust-lead loading as a predictor variable; the impact of other potentially important predictor
20 variables on blood-lead concentration was not considered in the model fittings. The regression
21 lines span the ranges of the observed area-weighted average dust-lead loadings, except data for
22 five HUD Grantees households (three from Cleveland and one each from Baltimore and Rhode
23 Island) were omitted from Figure 3-21 as their average window sill dust-lead loadings were
24 extremely low (less than 0.05 ug/ft2) compared to the other households and were considered too
25 influential to the model fittings.
26 When fitting the regression models in Figures 3-20 and 3-21 to the HUD Grantees data, it
27 was desired to have each household having blood-lead and dust-lead data be represented by only
28 a single data point. This was possible only if blood-lead data were considered for a single child
29 in that household. In situations where data for multiple children were available for a single
30 household, only data for the youngest child older than 12 months of age were considered. This
31 approach resulted in a single blood-lead result for each household with blood-lead data. In
32 addition, only data for children meeting the following criteria were included in the regression
33 modeling:
34 • Children who lived in the sampled housing unit for at least three months and
35 before dust and soil samples were collected;
36 • Children whose blood samples were taken within four months of dust and soil
37 sample collection;
38 • Children not having medical treatment for lead poisoning.
DRAFT -- DO NOT CITE OR QUOTE 173 August 28, 2000
-------
1
2
3
Table 3-35. Percentage of Children with Elevated Blood-Lead Concentration (at Pre-
Intervention) in the HUD Grantees Evaluation According to Blood Collection
Method, Child Age Category, and Grantee (ages 1-2 years only)
Age Range
1-2 Years
3-5 Years
1-5 Years
Grantee
Alameda County
Baltimore
Boston
California
Cleveland
Massachusetts
Minnesota
New Jersey
Rhode Island
Wisconsin
Milwaukee
Chicago
New York City
Vermont
Age Range
1-2 Years
3-5 Years
1-5 Years
Grantee
Cleveland
Massachusetts
Minnesota
Rhode Island
Wisconsin
Milwaukee
Vermont
Number of
Children
Percentage of Children with Elevated Blood-Lead Concentration (%)
* 10/ig/dL
2 15/ig/dL
* 20 //g/dL
& 25/ig/dL
Blood Collection Method = Venipuncture
361
534
895
51
43
46
35
28
31
18
12
14
12
7
9
Blood Collection Method = Venipuncture (Children Aged 1-2 Years only)
27
24
20
21
64
44
75
1
13
9
4
28
23
8
22
33
55
14
80
45
61
0
38
44
75
75
4
63
11
21
50
5
59
30
44
0
23
22
75
39
0
50
4
8
15
0
38
9
31
0
8
11
50
14
0
13
0
4
5
0
30
7
21
0
0
0
50
7
0
0
Blood Collection Method = Fingerstick
165
230
395
46
44
45
28
26
27
14
14
14
10
7
9
Blood Collection Method = Fingerstick (Children Aged 1-2 Years only)
1
4
1
9
42
84
24
100
25
100
33
10
70
29
0
0
100
11
0
50
13
0
0
100
0
0
26
0
0
0
100
0
0
19
0
35
36
Note: All pre-intervention blood-lead concentration data available and collected through 1/99 are included in the above
summaries.
DRAFT - DO NOT CITE OR QUOTE
174
August 28, 2000
-------
100
10
CD
0.1 1.0 10.0 100.0 1000.0 10000.0
Carpeted and Uncarpeted Floor Wipe Area-Weighted Arithmetic Average Dust-Lead Loading ( ^g/ft2)
AJameda County D B B Baltimore
Cleveland F—F—F Massachusetts
Wisconsin K K K Milwaukee
Vermont A A A Rochester
Boston
Minnesota
Chicago
ODD
California
Rhode Island
New 'York City
Figure 3-20. Fitted Regression Models Predicting Children's Blood-Lead Concentration as
a Function of Area-Weighted Arithmetic Average Floor Dust-Lead Loading
(Wipe Collection Method), for the Various Grantees in the HUD Grantees
Evaluation and for the Rochester Lead-in-Dust Study
(Note: Venipuncture blood-lead data were exclusively used in each fitting except for Wisconsin, Milwaukee, and Vermont,
where fingerprick blood-lead data were exclusively used.)
DRAFT -- DO NOT CITE OR QUOTE
175
August 28, 2000
-------
100
I
10H
1
0.1
1.0 10.0 100.0 1000.0 10000.0
Window Sill Area—Weighted Arithmetic Average Dust—Lead Loading (
100000.0
A A A Alameda County
E E E Cleveland
J— J — j Wisconsin
N N N Vermont
&' o o Baltimore
F F F Massachusetts
K K K Milwaukee
000 Baltimore R&M
G G G Boston
G G G Minnesota
ILL Chicago
A A A Rochester
D D D California
-1 — 1 — h Rhode Island
M M M New ID* City
Figure 3-21. Fitted Regression Models Predicting Children's Blood-Lead Concentration as
a Function of Area-Weighted Arithmetic Average Window Sill Dust-Lead
Loading (Wipe Collection Method), for the Various Grantees in the HUD
Grantees Evaluation and for the Rochester Lead-in-Dust Study
(Note: Venipuncture blood-lead data were exclusively used in each fitting except for Wisconsin, Milwaukee, and Vermont,
where fingerprick blood-lead data were exclusively used.)
1 The regression models in Figures 3-20 and 3-21 were fitted to blood-lead concentration
2 data only under the venipuncture method for all but the three grantees (Wisconsin, Milwaukee
3 and Vermont) for which fingerstick sample results were predominant. For these three grantees,
4 only fingerstick blood-lead concentration data were used in the regressions.
5 Note that the slopes of the fitted regression lines in Figures 3-20 and 3-21 are generally
6 similar in sign and magnitude (given expected ranges of variability) across the grantees and the
7 two other studies. This suggests that the relationships between blood-lead concentration and
8 household average dust-lead loading were relatively consistent across grantees. In particular,
DRAFT -- DO NOT CITE OR QUOTE
176
August 28, 2000
-------
1 these relationships were similar to that observed for data from the Rochester study (i.e., the data
2 used to develop the empirical model presented in Chapter 4 of the §403 risk analysis). This
3 conclusion is important in that the data from the HUD Grantees evaluation reflect a much larger
4 geographical area than the Rochester study and represent several types of exposure conditions.
5 3.6.2 Evidence of the Impact of Housing Age/Condition
6 on Blood-Lead Concentration
7 The role that housing age plays in the increased likelihood of a resident child having an
8 elevated blood-lead concentration has been well-documented and is accepted by many experts in
9 residential lead exposure. Older housing is more likely to contain lead-based paint in a
10 deteriorated condition, which contributes to lead in other environmental media within the
11 residence, especially those media that is most likely to come into direct contact with children. In
12 particular, the importance that the level of deterioration plays in the accessibility of lead-based
13 paint hazards implies that housing condition is an additional key factor in predicting blood-lead
14 concentration.
15 Table 3-39 of the §403 risk analysis report summarized data from Phase 2 of NHANES
16 m to illustrate how geometric mean blood-lead concentration and the percentage of elevated
17 blood-lead concentrations (i.e., percentage exceeding a given threshold) for children are related
18 to housing age category. For example, the percentage of children aged 1-5 years with blood-lead
19 concentration of at least 10 ug/dL increases from 1.6% for children living in post-1973 housing
20 to 8.6% for children living in pre-1946 housing, with a corresponding geometric mean increase
21 from 2.0 to 3.8 ug/dL. The Centers for Disease Control and Prevention (CDC) cited these same
22 results in their 1997 document, Screening Young Children for Lead Poisoning, to support their
23 conclusion that older housing (i.e., housing built prior to 1950) contained the greatest risk for
24 lead-based paint hazards.
25 Figure 6-1 of the CDC's 1991 document, Preventing Lead Poisoning in Young Children
26 - A Statement by the Centers for Disease Control, presents results from the Cincinnati
27 Prospective Lead Study (Clark et al., 1985) to illustrate how the combination of housing age and
28 condition is related to children's blood-lead concentration and how this relationship changes with
29 the age of the child. This figure is duplicated in Figure 3-22. This figure shows that children's
30 blood-lead concentration tends to peak at 18-24 months, with the most rapid increase occurring
31 between 6-12 months. The highest blood-lead levels are associated with housing built prior to
32 World War n, as well as older housing (predominantly 19th century) that once contained
33 considerable lead-based paint but which later underwent rehabilitation. Within these groups of
34 housing, children living in units in a deteriorated or dilapidated condition had consistently higher
35 geometric mean-blood-lead concentrations through their first three years, with this geometric
36 mean exceeding 20 ug/dL from about 12 to 24 months of age. CDC used the information
37 presented in Figure 3-22 to prepare a recommended screening schedule for testing children's
38 blood-lead levels.
DRAFT -- DO NOT CITE OR QUOTE 177 August 28, 2000
-------
0369
12 15 U 21 24 27 30 33 36 39 42
Child's Age (in Months)
DtL - Deteriorated
hr- WWII
$*t « S«trtf»ctor)r
Prt-WWll
Pub. - NMk
Figure 3-22. Geometric Mean Blood-Lead Concentration Versus Child Age, As Reported
Within the Cincinnati Prospective Lead Study and Presented According to
Housing Age and Condition
(Note: Duplicated from Figure 6-1 of CDC, 1991. Blood-lead concentrations for the same cohort of children
were measured over time. Numbers in parentheses indicate numbers of children with blood-lead information at
18 months of age.)
DRAFT -- DO NOT CITE OR QUOTE
178
August 28, 2000
-------
1 4.0 DOSE-RESPONSE ASSESSMENT
2 The objective of dose-response assessment was to characterize the relationship between
3 environmental-lead exposure and the resulting adverse health effects in young children. The
4 foundation of this characterization was the relationship between environmental-lead levels and
5 blood-lead concentration. EPA's Integrated Exposure, Uptake, and Biokinetic (IEUBK) model
6 and an empirical model developed for this risk assessment (Sections 4.1 and 4.2 of the §403 risk
7 analysis report) were employed to make this characterization.
8 Section 4.1 of this chapter documents an additional tool, obtained since the §403 risk
9 analysis report was published, for predicting blood-lead concentration as a function of
10 environmental-lead levels. This tool is a regression model developed from epidemiological data
11 collected from 12 studies and was suggested for use in the §403 risk analysis by some
12 commenters on the §403 proposed rule. As the U.S. Department of Housing and Urban
13 Development (HUD) sponsored the development of this model, it is referred to in this report as
14 the "HUD Model." The goal of this model was to "estimate the contribution of lead-
15 contaminated house dust and soil to children's blood-lead levels" (Lanphear et al., 1998). This
16 goal is consistent with the objectives of the §403 risk assessment, and so the model merits
17 consideration for this analysis. Section 4.1 includes documentation on the HUD model, key steps
18 that were taken in its development, and issues that are necessary to consider when interpreting
19 the results of the model fits.
20 Section 4.2 of this chapter contains a revision of the Rochester Multimedia model,
21 introduced in Section 4.2.3 of the §403 risk analysis report, to allow the model to predict results
22 that are more comparable to the results of the performance characteristics analysis presented in
23 the preamble to the §403 proposed rule. The Rochester Multimedia model predicted a geometric
24 mean blood-lead concentration as a function of average dust-lead loadings in floors and window
25 sills, dripline soil-lead concentration, and a variable which indicates the presence of deteriorated
26 lead-based paint and a child with paint pica tendencies. In contrast, the performance
27 characteristics analysis in the preamble estimated risks associated with dust-lead loadings on
28 uncarpeted floors, dust-lead loadings on window sills, yardwide average soil-lead concentration,
29 and the percentage of painted components with deteriorated lead-based paint. Because the
30 definitions of the data inputs were not always consistent between these two statistical
31 approaches, their findings were not comparable. Thus, the revised model presented in Section
32 4.2 uses the same types of data inputs as those used for the performance characteristics analysis.
33 Section 4.2 also documents multimedia models that omit one or more of the dust, soil, and paint
34 input variables to obtain predicted blood-lead concentration in instances where data for one or
35 more of these media were not available.
36 Section 4.3 provides additional information regarding key assumptions made in the risk
37 characterization process: the "scaling" algorithm used to determine a post-intervention blood-
38 lead concentration distribution that is comparable to the baseline distribution, and the issue of
DRAFT - DO NOT CITE OR QUOTE 179 August 28. 2000
-------
1 adjusting for measurement error when deriving the empirical model used in the §403 risk
2 analysis.
3 4.1 HUD MODEL
4 The HUD model (Lanphear et al., 1998) was developed by a team of researchers and
5 sponsored by HUD's Office of Lead Hazard Control. This modeling effort used data from 12
6 epidemiologic studies (hence its frequent reference as a "pooled analysis" model) to make
7 statistical inferences on the contribution of lead-contaminated house dust and residential soil to
8 children's blood-lead concentration.
9 The HUD model predicts a geometric mean blood-lead concentration for children aged 6-
10 36 months (i.e., the age range of data considered from the 12 studies) as a function of exposure to
11 specific lead levels in dust, soil, paint, and water, and as a function of other important
12 demographic variables. Therefore, the model is used to estimate individual risks, or the risks
13 associated with a class of children determined by specified environmental-lead levels to which
14 they are exposed. EPA addressed minimizing individual risks when establishing "levels of
15 concern" for lead in dust and soil within the §403 proposed rule. However, EPA was obliged to
16 consider population-based risks within a cost-benefit analysis to establish lead hazard standards
17 within the proposed rule.
18 4.1.1 Form of the HUD Model
19 The HUD model takes the following form:
20 (1) Ln(PbB) = 1.496 + 0. !83-Ln(DustLead) + 0.0/395 Ln(WaterLead) + 0.02116-Ln(ExtLead) +
21 0.005787-Ln(ExtLead)-ExiType + 0.4802-Ln(ExtLead)-ExtLoc - 0.1336-ExtType + 0.5858-ExtLoc
22 - 0.02199-Ln(MaxXRF) + 0.0381 l-Ln(MaxXRF)-PamtCond - 0.0808-PaintCond + 0.02126-Age -
23 0.001399-Age2 + 0.00007854-Age3 - 0.3932-Boston - 0.01167-Butte + 0.2027-Bcreek +
24 0.2392-Cpgm + 0.5383-Csoil + 0.05717-Leadville + 0.1761-Magna • 0.04209-RochLong +
25 0.07257-RochLlD - 0.3712-Sandy + 0.1777-Midvale + 0.123-Race + 0.3175-SES1 + 0.2138-SES2
26 + 0.1799-SES3 + 0.169J-SES4 - 0.03233-MouthOfien - 0.2454-MouthRare - 0.1397-MouthSome
27 + 0.002649-Ln(DustUad)-Age - 0.0003381-Ln(DustLead)-Age2 - 0.0000128 !-Ln(DustLead)-Age3
28 + 0.2212-Ln(ExtLead)-MouthOften + 0.07892-Ln(ExtLead)-MouthRare +
29 0.1663-Ln(ExtLead)-MouthSome + 0.5305-Ln(WaterLead)-SESl - 0.0136-Ln(WaterLead)-SES2 +
30 0.1033-Ln(WaterLead)-SES3 - 0.09098-Ln(WaterLead)-SES4 + 0.01192-Age-Race -
31 0.01023-Age-SESl + 0.003849-Age-SES2 + 0.00008468-Age-SES3 - 0.01679-Age-SES4 + error
32 where
33 Ln(PbB) = log-transformed blood-lead concentration (ug/dL)
34 Ln(DustLead) = log-transformed interior (wipe) floor dust-lead loading (ug/ft2), minus
35 the mean of the log-transformed data used to develop the model (2.605 fig/ft2)
36 Ln(WaterLead) = log-transformed water-lead concentration (ppb), minus the mean of the
37 log-transformed data used to develop the model (0.785 ppb)
38 Ln(ExtLead) = log-transformed exterior-lead concentration (ppm), minus the mean of the
39 log-transformed data used to develop the model (6.232 ppm), where the exterior
DRAFT -- DO NOT CITE OR QUOTE 180 August 28, 2000
-------
1 sample is either soil collected at the perimeter of the foundation, soil from the
2 child's play area, or exterior dust
3 ExtType = indicator of the type of exterior sample (1 = dust, 0 = soil)
4 ExtLoc = indicator of whether exterior sample is represented by soil at the perimeter of
5 the house's foundation (1 = exterior sample is not from perimeter soil, 0 =
6 exterior sample is from perimeter soil)
7 Ln(MaxXRF) = log-transformed maximum lead-paint measurement on interior surfaces
8 (mg/cm2, as measured by XRF), minus the mean of the log-transformed data used
9 to develop the model (0.921 mg/cm2)
10 PaintCond = indicator of paint condition (1 = damaged, 0 = undamaged)
11 Age = age of child (months) minus the mean age of children whose data were used to
12 develop the model (16.3 months)
13 Age2 = Age2 - (85.5 + 4.82-Age) (quadratic orthogonal polynomial)
14 Age3 = Age3 - (-490.71 + 10.32-Age2 + 122.3-Age) (cubic orthogonal polynomial)
15 Boston, Butte, Bcreek, Cpgm, Csoil, Leadville, Magna, RochLong, RochUD, Sandy, and
16 Midvale are indicators that the data come from the particular study being represented (1 =
17 data comes from the particular study, 0 = otherwise)
18 Race = Race indicator (0 = white, 1 = other)
19 SES1 = indicator of whether the pseudo-Hollingshead measure of socioeconomic status'is
20 equal to 1 (1 = yes, 0 = no)
21 SES2 = indicator of whether the pseudo-Hollingshead measure of socioeconomic status is
22 equal to 2(1 = yes, 0 = no)
23 SES3 = indicator of whether the pseudo-Hollingshead measure of socioeconomic status is
24 equal to 3 (1 = yes, 0 = no)
25 SES4 = indicator of whether the pseudo-Hollingshead measure of socioeconomic status is
26 equal to 4(1 = yes, 0 = no)
27 MouthOften = indicator of whether mouthing behavior occurs often in the child (1 =
28 often, 0 = otherwise)
29 MouthRare = indicator of whether mouthing behavior occurs rarely in the child (1 =
30 rarely, 0 = otherwise)
31 MouthSome = indicator of whether mouthing behavior occurs sometimes in the child (1 =
32 sometimes, 0 = otherwise)
33 error - random error between the observed log-transformed blood-lead concentration and
34 what is predicted by the model.
35 4.1.2 Development of the HUD Model
36 This section presents several issues on how the HUD model was developed that have a
37 direct impact on the predicted blood-lead concentration and how this prediction should be
38 interpreted. These issues include how studies were selected, how study effects were represented
39 in the model, and how data were handled or adjusted prior to or during the model development
40 exercise.
DRAFT -- DO NOT CITE OR QUOTE 181 August 28, 2000
-------
1 Study Selection and Potential Selection Bias
2 The HUD model was developed from environmental-lead, blood-lead, and demographic
3 data from 12 studies performed over a 15-year time frame (1982-1997). These studies
4 investigated the relationship between environmental-lead levels and children's blood-lead levels
5 in various locations and subpopulations. Five of the studies (representing 62% of the data used
6 to fit the model) were conducted in urban environments:
7 • Boston Longitudinal Study (Rabinowitz et al., 1985)
8 • Cincinnati Longitudinal Study (Bomschein et al., 1985b)
9 • Cincinnati Soil Study (Clark et al., 1991)
10 • Rochester Longitudinal Study (Lanphear et al., unpublished)
11 • Rochester Lead-in-Dust Study (Lanphear et al., 1996a,b)
12 The remaining seven studies (representing 38% of the data used to fit the model) were conducted
13 in milling, mining, or smelter environments:
14 • Bingham Creek, Utah (1993)
15 • Butte, Montana (1990)
16 • Leadville, Colorado (1991)
17 • Magna, Utah (1994)
18 • Midvale, Utah (1989)
19 • Palmerton, Pennsylvania (1994)
20 • Sandy, Utah (1994)
21 According to Lanphear et al. (1998), these 12 studies were selected based on the following
22 criteria:
23 • The studies had well-defined sampling protocols for blood and environmental
24 media (particularly dust, soil, and paint).
25 • The studies took measures of dust-lead levels, soil-lead levels, paint-lead content
26 (via XRF), and paint condition.
27 • The original data were available and could be reanalyzed.
28 • Dust samples were collected via wipe techniques or by the Dust Vacuum Method
29 (DVM)
30 • Dust samples were taken within three months of collecting blood samples from
31 the resident child(ren) (to address seasonal variation in blood-lead concentration).
32 • Children were not selected on the basis of having a high blood-lead concentration.
33 In addition, only cross-sectional data were considered (i.e., data were not considered that could
34 reflect changes in environmental-lead levels over time).
DRAFT - DO NOT CITE OR QUOTE 182 August 28, 2000
-------
1 As a result of the inclusion criteria, data for at least seven other studies considered by
2 HUD were excluded from the model development process. These studies and the primary
3 reasons for their exclusion (when specified within Lanphear et al., 1998) were
4 • UK Study (Davies et al., 1990) - dust collection method not wipe or DVM
5 • Boston and Baltimore segments of the EPA Urban Soil Lead Abatement
6 Demonstration Study (USEPA, 1996a; Weitzman et al., 1993) - dust collection
7 method not wipe or DVM
8 • Australian National Survey (Donovan et al., 1996) - lack of XRF paint-lead levels
9 • Baltimore R&M Study (USEPA, 1997a; USEPA, 1997e) - method for selecting
10 children did not meet the criteria, and data were not available to the analysis
11 • Telluride, CO (1987) - reason for exclusion not given
12 • Trail, BC (1992) - reason for exclusion not given.
13 Lanphear et al. (1998) indicates that the 12 studies were not chosen to represent the entire
14 nation or even communities like those in which the studies were conducted. In fact, these studies
15 were conducted in communities with a recognized environmental-lead hazard, and any abatement
16 efforts within each study targeted those hazards. Furthermore, the study effects included in the
17 model were treated as fixed effects (i.e., they are the only studies of interest in the model-
18 building process) rather than random effects (i.e., they are assumed to be a random sample of a
19 larger population of studies). Thus, if the model is used to estimate risks to a broader population
20 of children than simply those within the 12 studies, additional information is needed to determine
21 the extent to which the pooled data used to develop the model are representative of the U.S.
22 housing stock.
23 Fixed vs. Random Study Effects and Interaction with the Study Effects
24 As stated in the previous paragraph, the study effect in the HUD model is a series of fixed
25 effects. If the study effect was assumed to be random instead of fixed (i.e., the studies can be
26 considered a random selection of all such residential-lead exposure studies), then study-to-study
27 variation would become a contributor to total variation in the prediction. Based on work with
28 - previous models that incorporate a random study effect, the study-to-study component is typically
29 a major portion of total variability in the prediction. Thus, the variability associated with
30 predictions by the HUD model is likely underestimated.
31 Additional underestimation in variability may result from the absence of interaction terms
32 in the model between study effects and other environmental exposure factors. This can
33 underestimate variability associated with inferences involving the environmental exposure
34 factors, including the principal inferences which involve dust-lead and exterior-lead levels.
35 Adjusting for Measurement Error in Environmental-Lead Predictor Variables
36 The HUD model parameter estimates were determined from Simulation Extrapolation
37 (SIMEX) methods, which attempted to quantify the theoretical relationship between blood-lead
DRAFT -- DO NOT CITE OR QUOTE 183 August 28. 2000
-------
1 and "error-free" measures of environmental-lead. As a result, the parameter estimates were
2 adjusted to reflect measurement error present among the environmental-lead variables. However,
3 the goal is to predict children's blood-lead concentrations as a function of wipe dust-lead
4 loadings and soil-lead concentrations as they would be measured in a risk assessment, not their
5 true, "error-free" (but unobservable) values. Carroll et al. (1995) states that for prediction
6 problems, adjusting for the effects of measurement error in predictor variables is rarely
7 necessary. Adjusting for measurement error in these predictor variables tends to increase the
8 values of the slope parameters associated with these variables, which in turn can inflate predicted
9 blood-lead concentrations. Thus any predictions from the HUD model fits should be properly
10 labeled that values of the predictor variables are assumed to be "error-free" rather than measures
11 of environmental-lead levels taken from activities such as a risk assessment (as the predictor
12 variables in the models used in the §403 risk analysis were assumed to represent).
13 Making Survey Variable Definitions Consistent Across Studies
14 Because different survey designs in different studies can result in different definitions for
IS a common survey measure (e.g., SES, mouthing behavior, paint condition), which in turn can
16 introduce considerable complication when interpreting model predictions, certain survey
17 measures were redefined to make them more consistent across studies. Each redefinition
18 transforms the original data values to values generated from a domain that is consistent across the
19 studies. However, such a transformation does not remove all study-to-study differences in these
20 values. In particular, it does not consider factors that impact how the specific study
21 measurements were obtained and which differ from study to study, such as the use of different
22 survey instruments and different approaches to administering the instruments.
23 Converting DVM Dust-Lead Loadings to Wipe-Equivalents
24 While it was desired to have floor dust-lead loading assuming wipe dust collection as a
25 predictor variable in the HUD model, some of the 12 studies used DVM methods to collect dust
26 samples. Rather than exclude data from these studies from consideration in the model
27 development effort, a procedure was derived to convert DVM dust-lead loadings reported in
28 these studies to wipe-equivalent loadings. This procedure used data from the Butte study to
29 develop the following conversion equations:
30 • Log, d(Wipe) = 0.7727 + 0.9821 -Log, (JDVM) for carpeted floors,
31 « Logl(JWipe) = 0.1762 + 0.4839-Log,(j(DVM) for hard floors.
32 where "Wipe" and "DVM" indicate wipe dust-lead loadings and DVM dust-lead loadings,
33 respectively (Westat, 1998). Note that these same two equations were used to convert DVM
34 dust-lead loadings in each study, regardless of whether the relationship between DVM and wipe
35 dust-lead loadings differed among the studies.
36 The method for deriving these conversion equations included a procedure to adjust for
37 measurement error in the DVM dust-lead loading measurements. However, the purpose of the
DRAFT -- DO NOT CITE OR QUOTE 1 84 August 28, 2000
-------
1 conversion was to predict a measured wipe dust-lead loading based on a measured DVM dust-
2 lead loading, not the (unobservable) "true" DVM dust-lead loading. Thus, some bias may have
3 been introduced in this conversion process.
4 Interpreting the "Exterior-Lead" Predictor
5 Certain households whose data were used in the HUD model development effort did not
6 have soil-lead data for various reasons (e.g., no bare soil). In these instances, exterior dust-lead
7 concentration was generally measured instead. As a result, among the predictor variables in the
8 HUD model was an indicator variable that identified whether or not an exterior-lead
9 measurement was from dust or soil. This indicator allowed both the model intercept and the
10 slope factor associated with exterior-lead concentration to change according to its value.
11 However, no other consideration was made for differences in sampling and analysis methods
12 between soil and exterior dust and the impact such differences can have on the reported lead
13 levels. In addition, no indication was given that differences in bioavailability between soil-lead
14 and exterior dust-lead were considered in developing the model.
15 When a household had soil-lead data available, data from foundation perimeter (i.e.,
16 dripline) soil were used when available; otherwise, play-area soil-lead data were used. Like the
17 soil vs. exterior-dust issue in the previous paragraph, the HUD model includes an indicator
18 variable that identified whether or not soil-lead levels represent dripline soil. This indicator
19 allowed both the model intercept and the slope factor associated with exterior-lead concentration
20 to change according to its value. However, certain study-to-study differences in collecting soil
21 samples or obtaining a soil-lead measurement were not considered in model development, such
22 as the depth of soil sampling, soil surface type (e.g., covered vs. bare), chemical methods for the
23 digestion and analysis of soil samples, and soil compositing.
24 Handling Missing Water-Lead Measurements
25 While the HUD model included water-lead concentration as a predictor variable, it was
26 necessary to impute values for this measurement during model development when data for a
27 given household were not available. Water-lead data were unavailable for all households in two
28 studies and up to 12% of study households in the other ten studies. In such instances, imputed
29 measurements were randomly generated from a lognormal distribution with geometric mean
30 equal to that observed from data for other study households (if data for other households were
31 available) or to the community-wide average (if data for other households were not available).
32 Handling Data Reported Below a Detection Limit
33 When data values reflected measurements at or below some detection limit, the HUD
34 model development effort replaced them with probability-based values between zero and the
35 detection limit, where probabilities were determined from a lognormal distribution associated
36 with data above the detection limit. According to Table 2.15 of Westat (1998), the incidence of
37 not-detected results was high with XRF paint-lead level, and to a lesser extent, with water-lead
DRAFT - DO NOT CITE OR QUOTE 185 August 28, 2000
-------
1 concentration. In half of the studies, the percentage of not-detected paint-lead results ranged
2 from 20% to 86% (with the detection limit being reported at either 0.1 or 0.7 mg/cm2). The
3 percentage of not-detected water-lead concentrations (i.e., results somewhere below 5 ppb)
4 exceeded 80% in two studies. In contrast, none of the lead levels in blood, dust, or soil samples
5 were reported below a detection limit in 10 of the 12 studies, and lead levels in no more than 9%
6 of these samples were reported below a detection limit in the other two studies.
7 Selecting from Multiple Observations Within a Household
8 When blood-lead concentration data were available for multiple children within a
9 household (as occurred in nine of the studies), only data for one child selected at random from
10 the household were considered in the HUD model development. If blood-lead data existed at
11 multiple time points for the same child (such as at 6, 18, and 24 months of age in the Boston
12 Longitudinal study), those time points whose data met the initial inclusion criteria were identified
13 (e.g., lead interventions did not occur between the time points), and data for the time point
14 having dust-lead measurements taken more closely in time were selected for the model
15 development. Similarly, when environmental-lead measurements were repeatedly taken over
16 time for a given household, data for the time point closest to a blood-lead measurement were
17 used.
18 Handling Seasonalitv Effects
19 The effect of seasonality on blood-lead concentrations was given some consideration in
20 the modeling effort (e.g., blood and dust samples must have been collected within three months
21 of each other for their data to be included). However, there is no effect of seasonality included in
22 the final model. It is unclear whether seasonality was determined not to be a significant effect
23 among the pooled data, or whether a seasonality term was intentionally left out of the model.
24 4.1.3 Interpreting Results of Fitting the HUD Model
25 The previous section discussed issues concerning the pooled study data and development
26 of the HUD model that should be understood when using the model to estimate risks, as is done
27 in Section 5.1.1 and Appendix F. This section addresses the interpretation of results from fitting
28 the HUD model and, in particular, caveats associated with certain interpretations.
29 Individual risks vs. population-based risks
30 As mentioned earlier, the HUD model estimates individual risks associated with lead
31 exposure to children aged 6-36 months. While the §403 risk analysis included individual risks
32 analyses, which EPA used in efforts to establish levels of concern for lead in environmental
33 media, EPA was required to employ cost-benefit analysis to select §403 hazard standards. The
34 cost-benefit analyses used population-based risks (i.e., risks posed by childhood lead exposure to
35 the nation as a whole) to estimate the benefit and cost associated with performing interventions
36 and other activities in response to §403 rules.
DRAFT -- DO NOT CITE OR QUOTE 186 August 28, 2000
-------
1 Individual risks and population-based risks are generally not comparable. This must be
2 understood when attempting to compare the individual risks estimated by fitting the HUD model
3 at specified environmental-lead levels with the population-based risk estimates found in the §403
4 risk analysis.
5 Interpreting Model Parameter Estimates
6 The prediction parameters in the HUD model are not independent. For example, it is
7 known that soil-lead and dust-lead measures are correlated. Therefore, it is not appropriate to
8 interpret the parameter estimates in the HUD model (or in the models developed for the §403 risk
9 analysis) in isolation. Using the parameter estimates to characterize a cause-and-effect
10 relationship that is attributable to a single parameter alone, such as measuring the extent of an
11 increase in blood-lead concentration associated with a given increase in dust-lead loading, is very
12 problematic.
13 One example of how correlation among the predictor variables can influence the model
14 parameter estimates is seen with maximum XRF paint-lead measurement. One would expect a
15 positive correlation between maximum XRF paint-lead measurement and blood-lead
16 concentration, and as a result, a positive slope parameter. However, the estimated slope
17 parameter is negative (-0.022), although not significantly different from zero. The negative
18 estimate is likely due to confounding between paint-lead measurements and other predictor
19 variables. The likelihood of confounding increases with the number of parameters in the model.
20 Problems with interpreting model parameter estimates in isolation emphasizes the need to
21 consider total exposure (i.e., prediction based on considering the joint effect of all model
22 parameters) rather than exposure associated with a single environmental medium. In the §403
23 situation, protectiveness needs to be judged by recognizing that hazard standards exist for dust,
24 soil, and paint, and that resulting actions from these multiple standards will determine the level
25 of protection, not just the actions associated with a single standard. For example, the level of
26 protection associated with a dust-lead loading standard of 5 |Jg/ft2, without consideration of other
27 standards, may equal that associated with a joint set of standards that involve a higher dust-lead
28 loading standard.
29 Interpreting Results at Low Environmental-Lead Exposures
30 The HUD model and the models developed for the EPA risk analysis are "log-log"
31 models. That is, they predict log-transformed blood-lead concentration as a linear function of
32 log-transformed environmental-lead levels. As the log transformation "stretches out" the lower
33 portion of the scale and contracts the upper portion of the scale, very low environmental-lead
34 levels and blood-lead concentrations have undue influence on inferences made from the models.
35 For example, the effect of increasing dust-lead loading from 1 to 10 ug/ft2 is equal to the effect of
36 increasing dust-lead loading from 10 to 100 ug/ft2. Therefore, inferences at such low levels can
37 be overestimated and misleading. Thus, any inferences at very low dust-lead loadings, such as 1
38 or 5 ug/ft2, should be made with caution.
DRAFT - DO NOT CITE OR QUOTE 187 August 28, 2000
-------
1 4.1.4 Conclusions
2 The following conclusions can be made on the HUD model and the comparison of risk
3 estimates originating from this model versus those originating from models used in the §403 risk
4 analysis:
5 • As the HUD model parameters associated with environmental-lead measurements
6 in specific media have been adjusted for measurement error, the input parameters
7 to this model are assumed "true" lead levels in these media. This can provide
8 biased results when the model is used to predict blood-lead concentration
9 associated with lead levels measured in current risk assessments. The Rochester
10 multimedia model and the empirical model did not have such an adjustment
11 incorporated.
12 ' • While the HUD model contains study effects, they are considered fixed effects
13 and therefore allow the model to make predictions for only the group of children
14 represented by the 12 studies. Furthermore, the study effects impact only the
15 intercept of the model, and any study-to-study differences that may be present in
16 other model terms (such as in environmental-lead measurements) are not
17 represented.
18 • The HUD model handles "exterior-lead measurements" (e.g., soil) differently than
19 the §403 models; the impact of such difference has not been determined.
20 4.2 ALTERNATIVE MULTIMEDIA MODELS FOR PREDICTING
21 A GEOMETRIC MEAN BLOOD-LEAD CONCENTRATION
22 BASED ON ENVIRONMENTAL-LEAD LEVELS
23 As discussed in the introduction to this chapter, the Rochester multimedia model,
24 presented in Section 4.2.3 of the §403 risk analysis report, was developed using data from the
25 Rochester Lead-in-Dust study to explain children's blood-lead concentration as a function of
26 dust-lead loadings from floors (carpeted and uncarpeted) and window sills, dripline soil-lead
27 concentration, and an indicator variable on the presence of deteriorated lead-based paint and a
28 child with paint pica tendencies. This model was used in the risk characterization (Section 5.3 of
29 the §403 risk analysis report) to determine the probability that a child exposed to specific levels
30 of lead in paint, dust, and soil would have a blood-lead concentration at or above 10 ug/dL. EPA
31 used these estimates of individual risk, as well as the findings of performance characteristics
32 analyses detailed in Section 6.1 of this report, in proposing levels of concern for lead in dust
33 (page 30318 in the §403 proposed rule).
34 The §403 proposed rule considered uncarpeted floors and a yard-wide average soil-lead
35 concentration when proposing dust and soil standards and levels of concern. The performance
36 characteristics analysis cited in the proposed rule considered these types of dust-lead and soil-
DRAFT -- DO NOT CITE OR QUOTE 188 August 28, 2000
-------
1 lead measures. However, in the Rochester multimedia model, the floor dust-lead loading
2 measure did not limit the type of floor surface to uncarpeted floors, the soil-lead measure
3 represented only dripline soil, and the paint/pica indicator variable was different from the paint
4 measure used in the performance characteristics analysis (the percentage of tested components in
5 the home with deteriorated lead-based paint). For these reasons, it was difficult to compare
6 estimates of individual risks based on this model to results obtained from the performance
7 characteristics analyses. Thus, it was desired to fit an alternative multimedia model (cited as
8 "Model A" in this section) that replaced the floor dust-lead and soil-lead predictor variables used
9 in the Rochester multimedia model with uncarpeted floor dust-lead loading and yard-wide
10 average soil-lead concentration, respectively, and replaced the paint/pica indicator variable with a
11 measure of the percentage of tested components containing lead-based paint.
12 While a household risk assessment for lead-based paint hazards is expected, at a
13 minimum, to characterize lead levels in floor dust and to identify the extent of deteriorated lead-
14 based paint, it is possible that some risk assessments may not measure lead levels in soil or
15 window sill dust. Therefore, to investigate how individual risks would be characterized in these
16 types of risk assessments, two alternative multimedia models were fitted that were reduced
17 versions of Model A. One model excluded soil-lead concentration as a predictor variable
18 ("Model B"), and the other model excluded both soil-lead concentration and window sill dust-
19 lead loading as predictor variables ("Model C").
20 The three alternative multimedia models were fitted using data from the Rochester Lead-
21 in-Dust study, using the same approach used to fit the Rochester multimedia model in the §403
22 risk analysis. The models were log-linear in nature, where the dust-lead and soil-lead measures
23 were log-transformed, and the models predicted a log-transformed blood-lead concentration. For
24 example, Model A took the following form:
25 log(PbB) = Po + P,*log(PbF) + p2*log(PbW) + P3*log(PbS) + p4*PbP
26 where PbB represents blood-lead concentration (ug/dL), PbF represents household average dust-
27 lead loading for uncarpeted floors (ug/ft2), PbW represents household average dust-lead loading
28 for window sills (ug/ft2), PbS represents yard-wide average soil-lead concentration (ug/g), and
29 PbP represents the larger (between the interior and exterior of the housing unit) of the
30 percentages of tested components containing deteriorated lead-based paint. As with the
31 Rochester multimedia model, ordinary least squares regression methods were used to fit the
32 models to the Rochester data.
33 Table 4-1 presents the estimates of the model parameters for each of the three alternative
34 multimedia models. Note that the model fits were based on different numbers of housing units,
35 as eliminating certain predictor variables from the above model resulted in more housing units
36 that had all necessary data available for fitting the model. An investigation of model diagnostics
37 showed that the extent of collinearity among the predictor variables in these models was low.
38 Generally, the slope estimates associated with the paint variable were very low, and except for
DRAFT - DO NOT CITE OR QUOTE 189 August 28, 2000
-------
1
2
3
4
5
Table 4-1. Parameter Estimates and Associated Standard Errors for the Three
Alternative Multimedia Models Fitted to Rochester Study Data to Predict
Log-Transformed Blood-Lead Concentration1
Parameter
Po
P,
P2
Pa
P4
R2
O&ror
n
Predictor Variable
Intercept
log (PbF): Area-Weighted Arithmetic Mean
(Wipe) Dust-Lead Loading from Uncaroeted
Floors
log (PbW): Area-Weighted Arithmetic Mean
(Wipe) Dust-Lead Loading from Window
Sills
log (PbS): Yardwide Average Soil-Lead
Concentration2 (fine soil fraction)
PbP: The larger of the following two
percentages: % of interior tested surfaces
that contain deteriorated LBP, and % of
exterior tested surfaces that contain
deteriorated LBP3
Coefficient of Determination
Error
# data points included in model fitting
Parameter Estimate (Standard Error)
Model A
0.331
(0.263)
0.114
(0.049)
0.082
(0.037)
0.115
(0.040)
0.001
(0.002)
20.03%
0.561
177
Model B
0.899
(0.183)
0.130
(0.048)
0.101
(0.035)
~
0.002
(0.002)
15.62%
0.572
188
Model C
1.337
(0.122)
0.140
(0.043)
-
-
0.004
(0.002)
11.38%
0.580
196
10
11
12
13
14
15
16
17
18
19
20
21
•--• indicates that the variable is not included in the model as a predictor. The models are log-linear in nature.
1 One housing unit (cid=01689) had an uncarpeted floor dust-lead loading measurement of 18130.0//g/ft2 (only one
uncarpeted floor wipe sample was collected in this unit). This data value was omitted when fitting the above models as it
was highly influential and led to a noticeable reduction in the estimate of P,.
2 Yardwide soil-lead concentration at a given housing unit was calculated as the unweighted arithmetic average of driplme
and play area soil-lead concentrations. If one or the other is missing (but both are not missing), yardwide concentration was
set equal to the non-missing value. If both are missing, yardwide concentration is missing.
3 If one or the other of these two percentages is missing (but both are not missing), the value of this variable is set equal to
the non-missing value. If both are missing, the value is missing.
DRAFT - DO NOT CITE OR QUOTE
190
August 28, 2000
-------
1 Model C, were not significantly different from zero at the 0.05 level. See footnotes to this table
2 for additional details on the model fits.
3 Other alternative multimedia models were considered when initially developing the
4 Rochester multimedia model. These models, and information used to evaluate these models in
5 selecting the final version used in the §403 risk analysis, were presented in Appendix G of the
6 §403 risk analysis report.
7 4.3 SUPPLEMENTAL INFORMATION ON MODEL-BASED
8 APPROACHES IN THE $403 RISK ANALYSIS
9 Some comments on the §403 proposed rule addressed issues concerning approaches taken
10 in the §403 risk analysis in which statistical models were used to predict a post-intervention
11 distribution of blood-lead concentration, and therefore, a means of assessing how health risks
12 associated with lead-based paint hazards would change as a result of implementing the §403 rule.
13 The types of models used in this analysis and the approach to characterize post-intervention
14 health risks were presented in Chapters 4 and 6 of the §403 risk analysis report. In addition,
15 technical details on these models and approaches were provided in appendices to the report.
16 However, certain comments on the §403 proposed rule indicated that this information may not
17 have been provided in sufficient detail or would have benefited from additional clarity. In
18 particular, two issues in question involved how the §403 risk analysis estimated a post-
19 intervention blood-lead concentration distribution that was comparable to the baseline
20 distribution that was characterized by data from Phase 2 of NHANES ffl, and how the empirical
21 model (Section 4.2 of the §403 risk analysis report) account for measurement error issues
22 associated with the predictor variables. The following subsections provide additional
23 information on these two issues, specifically geared toward addressing the specific areas raised
24 by selected public comments.
25 4.3.1 The "Scaling" Algorithm Used to Determine a Post-
26 Intervention Blood-Lead Concentration Distribution
27 In the §403 risk analysis, EPA used data from Phase 2 of NHANES ffl (collected from
28 1991-1994) as the basis for the baseline ("pre-§403 rule") characterization of children's blood-
29 lead concentration in the U.S. housing stock. As discussed in Section 3.5 of the §403 risk
30 analysis report, EPA took this approach because these data were considered the best available
31 data (as well as the most recent data) on blood-lead measures, the data consisted of actual blood-
32 lead measurements from a nationally-representative survey, and it was preferred (and considered
33 more defensible) to use such data to characterize the baseline distribution rather than data
34 generated from statistical prediction models. However, because the "post-§403 rule" time period
35 has not yet occurred, and it was desired to compare the blood-lead distribution in this time period
36 to the baseline blood-lead distribution, it was necessary to use statistical prediction models to
37 generate how the baseline distribution would change following interventions performed as a
38 result of the §403 rule.
DRAFT - DO NOT CITE OR QUOTE 191 August 28. 2000
-------
1 When estimating the blood-lead concentration and health effect endpoints used in the
2 §403 risk analysis, it was assumed that the national distribution of blood-lead concentration was
3 lognormally distributed. Initial investigations into the weighted NHANES EQ blood-lead
4 concentration data used in the §403 risk analysis (Figure 5-3 of the §403 risk analysis report)
5 suggested that this was a satisfactory assumption. The lognormal distribution is characterized by
6 the geometric mean and geometric standard deviation (GSD) of the data (or, equivalently, the
7 exponentiated mean and exponentiated standard deviation of the log-transformed data). Once the
8 geometric mean and GSD were calculated from the weighted NHANES HI data, the additional
9 assumption of lognormality was used to obtain baseline estimates of the health effect and blood-
10 lead concentration endpoints considered in the §403 risk analysis (e.g., probability that a child's
11 blood-lead level was at or above 10 ug/dL). These estimates were presented in Table 5-1 of the
12 §403 risk analysis report.
13 To estimate how the blood-lead distribution changed between the "pre-§403 rule" and
14 "post-§403 rule" environments (based on a given set of candidate §403 standards and on
15 assumed changes in environmental-lead levels resulting from implementing the §403 rule),
16 model-based estimates of the blood-lead distribution were made for both environments. For
17 reasons explained in the §403 risk analysis report, EPA chose to characterize both the pre-§403
18 and post-§403 blood-lead distributions twice, with the first characterization using the IEUBK
19 model and the second characterization using the empirical model. Each characterization
20 involved fitting the given model to environmental-lead data separately for each home in the HUD
21 National Survey and weighting each prediction appropriately to represent a given proportion of
22 the nation's children. Therefore, although the HUD National Survey homes do not represent a
23 random sample of homes in the national housing stock, and therefore, the set of predicted blood-
24 lead concentrations do not themselves represent a random sample of blood-lead levels in the
25 nation's children, the fact that each prediction is weighted appropriately to represent a given
26 proportion of the nation's children allows the total set of predictions to be a good estimate of the
27 national blood-lead distribution, in either a pre-§403 or post-§403 environment.
28 Appendix E2 of the §403 risk analysis report discusses how model-predicted blood-lead
29 concentrations generated for the HUD National Survey homes are used to estimate the geometric
30 mean and GSD associated with the national blood-lead distribution. Recall that for a given HUD
31 National Survey home, the model-predicted blood-lead level represents a geometric mean of
32 children whose exposure is characterized by the environmental-lead levels in that home.
33 Therefore, the estimated GSD of the national blood-lead distribution is characterized not only by
34 the variability among the predicted blood-lead levels, but also by the assumed variability in
35 blood-lead levels among individual children exposed to the same environmental-lead levels.
36 Under a given model (i.e., either the IEUBK or empirical model), the "scaling" algorithm
37 (Appendix Fl of the risk analysis report) involves calculating the proportional change in the
38 geometric mean and GSD of the model-predicted blood-lead distribution from the "pre-§403
39 rule" to the "post-§403 rule" environment. Then, the same proportional change in both statistics
40 was applied to the geometric mean and GSD of the baseline distribution determined from the
41 NHANES m data:
DRAFT -- DO NOT CITE OR QUOTE 192 August 28, 2000
-------
2 GSDpo.,,^3 = GSDbaseline * (GSDmode|.based ^,^03 /
3 The resulting geometric mean and GSD, along with an assumption that lognormality still holds,
4 characterized a blood-lead concentration that represented the post-§403 environment and that
5 was considered comparable to the baseline distribution.
6 This type of algorithm was necessary due to the difficulties associated with comparing a
7 model-based post-§403 blood-lead distribution directly with a baseline distribution that was
8 characterized from observed (NHANES TH) data. While the empirical model was calibrated so
9 that its estimate for the baseline national geometric mean blood-lead concentration for children
10 aged 1-2 years (as obtained within the approach taken in the §403 risk analysis) equaled the
11 NHANES HI Phase 2 estimate (although the predicted GSD was not similarly calibrated), such a
12 calibration was not possible for the EEUBK model. As a result, using the HUD National Survey
13 data as input, these models could not both predict the same national estimate of the geometric
14 mean blood-lead concentration as Phase 2 of NHANES ffl. In addition, the empirical model
IS could not be developed based on data from a national survey that measured both blood-lead and
16 environmental-lead levels, which would have facilitated direct comparisons of the predicted
17 blood-lead distribution between pre-§403 and post-§403 environments. Therefore, the "post-
18 §403 rule" blood-lead distributions predicted by the two models had some inconsistency with the
19 baseline distribution estimated from the NHANES HI data, making direct comparisons
20 problematic. For example, if a model underestimates the geometric mean, the benefits associated
21 with the §403 rule could be overestimated, while if a model overestimates the geometric mean,
22 this could result in estimates of negative benefits. Therefore, the "scaling" algorithm used in the
23 risk analysis used the models to predict the change in the geometric mean and GSD that occurs
24 from a pre- to post-§403 rule environment, then applied this same change to the baseline
25 distribution.
26 Note that the scaling algorithm does not require that the two model-based blood-lead
27 distributions (pre-§403 and post-§403) be independent of each other. In fact, the two
28 distributions are dependent, because the post-§403 environmental-lead levels, used to predict
29 post-§403 blood-lead levels, are dependent on the pre-§403 levels. Only the geometric mean and
30 GSD of these two distributions are necessary to characterize, and they are used simply to
31 estimate the proportional change in the geometric mean and GSD of the national blood-lead
32 distribution between pre-§403 and post-§403 conditions.
33 While the geometric mean and GSD are scaled separately, one change is not necessarily
34 independent of the other. For example, if the pre-§403 geometric mean and GSD both have high
35 values, they are both likely to be reduced at a greater rate than at lower values. The approach
36 was kept as simple as possible while retaining scientific defensibility, in order that it be easily
37 applied during risk characterization and in the economic analysis.
38 In the peer review of the §403 risk analysis report, EPA specifically asked the peer
39 reviewers to comment on whether the scaling procedure was scientifically defensible in general,
DRAFT - DO NOT CITE OR QUOTE 193 August 28,2000
-------
1 and in particular, whether it was relevant in the situation where the environmental-lead data
2 (from the HUD National Survey) and the blood-lead data (from NHANES HI) were collected at
3 different periods in time. None of the peer reviewers specifically criticized the scaling algorithm.
4 Furthermore, the Science Advisory Board reviewing the §403 risk analysis stated that, in general,
5 the approach was scientifically defensible as presented, and specifically, the multi-step approach
6 was warranted due to the need to use various datasets of differing sources and representing
7 different time periods to make the characterization.
8 See Section 6.4.4 below for an alternative approach to applying this scaling algorithm
9 where the probability of a child's blood-lead concentration exceeding 10 ug/dL is scaled rather
10 than the GSD.
11 4.3.2 Adjusting the Empirical Model Parameter
12 Estimates to Reflect Measurement Error
13 The approach to the measurement error adjustment, discussed in Section 4.2.4 of the §403
14 risk analysis report, attempted to correct for the fact that while the empirical model was
15 developed using data from the Rochester Lead-in-Dust study, it was used to predict a (pre-403)
16 geometric mean blood-lead concentration assuming that the data input to the model originated
17 from the HUD National Survey. Because the Rochester study and the HUD National Survey
18 used different sampling schemes involving different collection devices and instruments, as well
19 as different analytical methods, and because the ranges of observed environmental-lead levels
20 differed between the two studies, it was necessary to adjust the model parameter estimates to
21 reflect these differences prior to allowing the model to accept HUD National Survey
22 environmental-lead data as input to the prediction.
23 Note that the measurement error adjustment made to the empirical model was not to
24 address the more-standard "errors in variables" issue (Carroll et al., 1995) which attempts to take
25 into account that a value input to the model represents a measurement subject to error, rather than
26 a "true" value. As discussed in Section 4.2.4 and Section G4.2 (Appendix G) of the §403 risk
27 analysis report, the empirical model was not intended to be used in the risk analysis as a dose-
28 response model, which would have required the predictor variables to reflect actual exposures.
29 Instead, the model assumed that its input environmental-lead information reflected measurements
30 that would have been made as a result of a risk assessment within a home. Therefore, adjusting
31 the model for the fact that its inputs reflect measured rather than actual lead levels was
32 considered inappropriate for this analysis. This decision in the type of application that is
33 represented by the §403 risk analysis has been concurred upon in the published literature (e.g.,
34 Carroll and Galindo, 1999).
35 When using the empirical model to predict a post-403 geometric mean blood-lead
36 concentration, some of the HUD National Survey dust-lead and soil-lead data (i.e., those data for
37 homes that exceed the candidate 403 standards) were modified to reflect the impact of
38 performing interventions in response to the 403 rule on these measured data values (see Table
39 6-2 of the §403 risk analysis report), then the model is fitted to the modified data. These
DRAFT - DO NOT CITE OR QUOTE 194 August 28, 2000
-------
1 modified data are still considered to be measured lead levels, rather than actual (or "true") lead
2 levels. However, the modified data values must first be transformed to represent measurements
3 that would have made under the methods used in the HUD National Survey. For example, the
4 assumed post-intervention floor wipe dust-lead loading of 40 ug/ft2 must be converted to a Blue
5 Nozzle vacuum-equivalent loading prior to using it as input to the empirical model.
DRAFT - DO NOT CITE OR QUOTE 195 August 28, 2000
-------
This page intentionally blank.
DRAFT -- DO NOT CITE OR QUOTE 196 August 28, 2000
-------
1 5.0 RISK CHARACTERIZATION
2 Chapter 5 of the §403 risk analysis report documented the final portion of the risk
3 assessment phase of the §403 risk analysis, in which the methods introduced in the earlier
4 chapters of the §403 risk analysis report were applied to characterize risks associated with current
5 (i.e., baseline) lead exposures for children aged 1-2 years. The baseline distribution of blood-
6 lead concentration in this population was characterized using data from Phase 2 of the Third
7 National Health and Nutritional Examination Survey (NHANES HI), conducted from 1991 to
8 1994. Alternative pre-§403 risk estimates were also calculated as a function of environmental-
9 lead levels by using data from the HUD National Survey as input to the BEUBK and empirical
10 models. Both individual risk estimates (i.e., risks associated with specific environmental-lead
11 levels) and population-based risk estimates (i.e., average risks over the entire nation) were
12 presented. As mentioned in Section 2.0, the specific blood-lead concentration and health effect
13 endpoints used to measure the risks of lead exposure to children aged 1-2 years were
14 • Incidence of blood-lead concentration greater than or equal to 10 ug/dL
15 • Incidence of blood-lead concentration greater than or equal to 20 ug/dL
16 • Incidence of IQ score less than 70 resulting from lead exposure
17 • Incidence of IQ score decrement greater than or equal to 1 resulting from lead
18 exposure
19 • Incidence of IQ score decrement greater than or equal to 2 resulting from lead
20 exposure
21 • Incidence of IQ score decrement greater than or equal to 3 resulting from lead
22 exposure
23 • Average IQ decrement in a child, resulting from lead exposure.
24 The risk characterization included a sensitivity and uncertainty analysis where possible
25 alternatives to various approaches taken and assumptions made in the risk characterization were
26 identified and incorporated into the analysis, and the resulting impact on the risk estimates was
27 evaluated. This analysis resulted in a measure of the uncertainty associated with the risk
28 estimates due to methodological assumptions, thereby producing a range of estimates within
29 which the true risk may reasonably be expected to fall. Section 5.1 of this chapter contains the
30 following additional sensitivity and uncertainty analyses that were performed and documented
31 since the §403 risk analysis report was published:
32 • Calculate individual risks associated with specified lead levels in floor-dust and
33 soil, as predicted by the HUD model introduced in Section 4.1 and the alternative
34 multimedia models presented in Section 4.2.
35 • Calculate estimates assuming a 50% decline in the estimated geometric mean
36 blood-lead concentration of children aged 1-2 years from the estimate generated
37 from Phase 2 of NHANES ffl (in addition to the estimates associated with 10%,
DRAFT - DO NOT CITE OR QUOTE 197 August 28, 2000
-------
1 20%, and 30% declines that were presented in Section 5.4.3 of the §403 risk
2 analysis report).
3 • Calculate model-based estimates of the pre-§403 blood-lead distribution under
4 revised environmental-lead levels (from the HUD National Survey) input to the
5 models, with the revisions representing the potential change in these levels that
6 may have occurred since the survey was performed.
7 • Calculate baseline estimates of the IQ-related health effect endpoints assuming
8 that specified non-zero thresholds exist in the relationship between blood-lead
9 concentration and IQ.
10 5.1 RISK CHARACTERIZATION SENSITIVITY AND UNCERTAINTY ANALYSIS
11 The following subsections present the results of additional sensitivity and uncertainty
12 analyses performed to gauge the level of uncertainty in baseline risk estimates associated with
13 methodological assumptions. These results should be considered with those presented in the
14 sensitivity and uncertainty analyses in Section 5.4 of the §403 risk analysis report to characterize
15 overall uncertainty associated with the methods and assumptions taken in the risk assessment.
16 5.1.1 Estimates of Individual Risks from Applying the HUD Model
17 In Section 5.3 of the §403 risk analysis report, the concept of individual risks was
18 introduced, and estimates of individual risks associated with lead exposure in children were
19 generated by fitting the IEUBK and Rochester multimedia models to specified environmental-
20 lead levels. Briefly, within the context of the §403 risk analysis, individual risks refer to the risks
21 associated with a young child's exposure to specified levels of environmental-lead. Once
22 environmental-lead levels were specified for each medium, the model-predicted blood-lead
23 concentration at these levels, along with the assumption that blood-lead concentrations have a
24 lognormal distribution with a specified variability, were used to estimate the percentage of
25 children exposed to the specified set of environmental-lead levels that would have elevated
26 blood-lead concentrations (i.e., at or above 10 |Jg/dL). Then, those sets of environmental-lead
27 levels associated with estimated elevated blood-lead percentages of 1%, 5%, and 10% were
28 identified and presented in Tables 5-5 through 5-7 and Figures 5-7 and 5-8 of Section 5.3 of the
29 §403 risk analysis report. The BEUBK model was used to identify soil-lead concentrations
30 associated with these elevated blood-lead percentages (at specified dust-lead loadings), while the
31 Rochester multimedia model was used to identify (wipe) dust-lead loadings associated with these
32 elevated blood-lead percentages (at specified soil-lead concentrations). These results contributed
33 to the information which EPA used in proposing dust and soil levels of concern in the §403
34 proposed rule. See Section 5.3 of the §403 risk analysis report for additional details.
35 As discussed in Section 4.1 of this report, the HUD model was published shortly after the
36 §403 risk analysis report was finalized, and some commenters to the §403 proposed rule
37 suggested that it be considered within the §403 risk analysis. The HUD model can be used in the
DRAFT--DO NOT CITE OR QUOTE 198 August 28.2000
-------
1 same manner as the §403 risk analysis models to predict individual risks associated with
2 exposure to a specified set of environmental-lead levels. Therefore, this section presents
3 estimates of individual risks based on fitting the HUD model to specified environmental-lead
4 levels and compares these risk estimates with those based on the IEUBK model (for soil) or the
5 Rochester multimedia model (for dust). Supporting summaries and discussion for comparing
6 HUD model results with those of the multimedia models developed for the §403 risk analysis are
7 presented in Appendix F.
8 Soil-Lead Concentrations
9 When fitting the HUD model to evaluate individual risks associated with yard-wide
10 average soil-lead concentration, the model's soil-lead parameters were set to the following
11 values:
12 • ExtType = 0 (indicating that soil was sampled rather than exterior dust)
13 • ExtLoc = 0.5 (indicating that the total soil sampled was a rough average of drip-
14 line and non-drip-line soil)
15 Figure 5-1 plots estimates of the percentage of children's blood-lead concentrations at or
16 above 10 ug/dL as a function of soil-lead concentration, as predicted by the IEUBK model and
17 the HUD model. The left-most panel in Figure 5-1 corresponds to the results of IEUBK model
18 fits at specified dust-lead concentrations (100,200, and 500 ug/g) and is identical to Figure 5-7
19 in the §403 risk analysis report. The middle and right-most panels of Figure 5-1 correspond to
20 fits of the HUD model at specified dust-lead loadings (5, 10,20,25,40,50, 100 and 200 ug/ft2)
21 and differ according to the assumed geometric standard deviation (GSD) associated with the
22 blood-lead concentration distribution (GSD=1.6 for the middle panel; GSD=1.72 for the right-
23 most panel). A GSD of 1.72 was estimated within the HUD model publication (Lanphear et al.,
24 1998).
25 The IEUBK and HUD model fits portrayed in Figure 5-1 are not directly comparable as
26 the IEUBK model controls for dust-lead concentration while the HUD model controls for dust-
27 lead loading. However, the plots do suggest that the predicted patterns of change in blood-lead
28 concentration with soil-lead concentration differ considerably for the two models.
29 Table 5-la, identical to Table 5-5 of the §403 risk analysis report, presents the soil-lead
30 concentrations for which the IEUBK model-predicted percentages of children having blood-lead
31 concentrations of at least 10 ug/dL equal 1%, 5%, and 10% assuming dust-lead concentrations of
32 100, 200, and 500 ug/g. Table 5-lb presents the soil-lead concentrations for which the HUD
33 model-predicted percentages of children having blood-lead concentrations of at least 10 ug/dL
34 equal 1%, 5%, and 10% assuming the dust-lead loadings considered in the HUD model panels of
35 Figure 5-1. Table 5-lb (HUD model) indicates that, for all dust-lead loadings considered, the
36 soil-lead concentrations estimated to maintain risks at 1% and 5% are less than 400 ug/g; even a
37 soil-lead concentration of less than 6 ug/g would not achieve these levels of protection if children
DRAFT--DO NOT CITE OR QUOTE 199 August 28,2000
-------
O
O
O
o
H
m
O
•u
0
O
m
o
o
500 750 1000 1250 1500
Soil-Lead Concentration (>ig/g):
Dust-Lead Concentration (/ig/g): — 100
-- 200
— 500
IEUBK Model
1750 2000
500 750 1000 1250
Soil-Lead Concentration
1500 1750 2000
250
Floor Du!t-L«od Loading (/jg/lt 2): 5 -• 10 - - 20
- 25 -• 40 - 50
- 100 •- 200
HUD Model (CSO = 1.60)
500 750 1000 1250 1500
Soil-Leod Concentration {/ig/g)
1750 2000
Floor Dust-Lead loading L>g/ll2):
5
10 --- 20
- 25 - - 40 - - 50
- 100 • - 200
HUD Model (GSD = 1.72)
>
CO
Figure 5-1. Percentage of Children's Blood-Lead Concentrations, as Predicted by the IEUBK and HUD Models. That Will
Exceed or Equal 10 pg/dL as a Function of Yard-Wide Average Soil-Lead Concentration and at Fixed Levels of
Dust-Lead Concentrations or Loadings
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Table 5-1 a. Yard-Wide Average Soil-Lead Concentrations at Which the Percentage of
Children Aged 1-2 Years With Blood-Lead Concentration At or Above 10
Aig/dL is Estimated by the IEUBK Model at 1, 5, or 10%, for Three Assumed
Dust-Lead Concentrations (Table 5-5 in §403 risk analysis report).
Floor Dust-Lead
Concentration (/ig/g)
100
200
500
Soil-Lead Concentration (/ig/g)
1%
155
35
Not achievable
5%
365
245
Not achievable
10%
515
395
25
Table 5-1b. Yard-Wide Average Soil-Lead Concentrations at Which the Percentage of
Children Aged 1-2 Years With Blood-Lead Concentration At or Above 10
A/g/dL is Estimated by the HUD Model at 1, 5, or 10%, for Eight Assumed
Dust-Lead Loadings and Two Assumed Geometric Standard Deviations
Floor Dust-Lead
Loading (/ig/ft2)
Soil-Lead Concentration (/ig/g)
1%
5% | 10%
GSD = 1.60
5
10
20
. 25
40
50
100
200
74.4
49.4
32.9
28.8
21.8
19.1
12.7
8.5
GSD
5
10
20
25
40
50
100
200
45.9
30.5
20.3
17.8
13.5
11.8
7.9
5.2
186.2
123.7
82.3
72.1
54.7
47.9
31.9
21.2
303.7
201.8
134.2
117.6
89.2
78.2
52.0
34.5
= 1.72
132.4
88.0
58.5
51.3
38.9
34.1
22.7
15.1
232.8
154.8
102.9
90.2
68.4
60.0
39.9
26.5
DRAFT - DO NOT CITE OR QUOTE
201
August 28, 2000
-------
1 were exposed to dust-lead loadings of 40 ug/ft2 or more. Based on Table 5-1 b, if dust-lead
2 loadings are equal to 5 \ig/ft2, a soil-lead concentration of approximately 300 ug/g (100 ug/g)
3 maintains the percentage of blood-lead concentrations greater than or equal to 10 ug/dL at 5% for
4 a GSD of 1.60(1.72).
5 Floor Dust-Lead Loadings
6 Figures 5-2 and 5-3 plot the estimated percentages of children having blood-lead
7 concentrations at or above 10 ug/dL as a function of floor dust-lead loadings as predicted by the
8 HUD model and the Rochester multimedia model assuming GSDs of 1.60 and 1.72, respectively,
9 on the blood-lead distribution. Soil-lead concentrations are assumed to be fixed at 100,400,
10 1200, 2000 and 5000 ug/g and, for the Rochester multimedia model, window sill dust-lead
11 loadings are assumed to be fixed at 200 and 500 ug/ft2. In each figure, the left-most panel
12 contains estimates based on fitting the HUD model. The Rochester multimedia model panels for
13 GSD equal to 1.60 and soil-lead concentrations of 100 and 400 ug/g were presented in Figure 5-8
14 in the §403 risk analysis report.
15 Tables 5-2 and 5-3 present the floor dust-lead loadings that are predicted by the HUD
16 model and Rochester multimedia model, respectively, to maintain the percentage of children
17 having blood-lead concentrations above or equal to 10 ug/dL at 1%, 5%, and 10% for specified
18 levels of soil-lead concentration, window sill dust-lead loading and GSD. Approximate 95%
19 upper confidence bounds, which account for the variability of parameter estimates from the
20 Rochester multimedia model, are also provided in Table 5-3. See Appendix C2, Section 5.0 of
21 the §403 risk analysis report for the methodology used to compute these confidence bounds. The
22 rows of Table 5-3 corresponding to a GSD of 1.60 and soil-lead concentrations of 100 and 400
23 ug/g are identical to Table 5-6 in the §403 risk analysis report (after correcting an error in the
24 computation of the confidence bounds).
25 Some key findings noted when comparing the individual risk estimates presented in
26 Appendix F between the HUD model and Rochester multimedia model include the following:
27 • At very low floor dust-lead loadings (i.e., 1-5 ug/ft2), the HUD model and the
28 Rochester multimedia model yield similar predictions for the geometric mean
29 blood-lead concentration, which also results in similar predictions for the health-
30 effect endpoints that are calculated directly from this geometric mean (e.g.,
31 percentage of children with blood-lead concentration at or above a specified
32 threshold; average IQ decrement resulting from lead exposure). However, due to
33 the forms of these models and concerns involving the accuracy of very low dust-
34 lead measurements, any conclusions made at such low dust-lead loadings must be
35 made with caution.
36 • The predicted geometric mean blood-lead concentration under the HUD model
37 ranges from 20% to nearly 60% higher than the prediction under the Rochester
38 multimedia model as floor dust-lead loadings increase from 15 to 100 |Jg/ft2 and
DRAFT -- DO NOT CITE OR QUOTE 202 August 28,2000
-------
D
DO
D
O
Z
O
o
O
ID
O
O
rn
ro
o
100 200 300
floor Dust-Leod Loodin
Soil-Leod Concentration (jjg/g): - 100
-- 1200
— 5000
HUD Model
500
400
2000
100 200 300 WO
Floor Dusl-Leod Looding (^g/ft?)
l-lwd Concentration (jiq/g): - 100 • • 400
" 1200 — 2000
— 5000
Rochester Multimedia Uodel
Wirdo« Sill Dust-Leod Looding ; 200 jig/It'
500
N 70
01
o 60
2 50
a.
o 40
1 »
0
| 20
a.
ID
0
100 200 300 400
Floor Dusl-Leod Looding (/Jg/ft!)
Soil-LMd Concwlrolion (jij/j): - 100 " 400
-- 1200 — 2000
- 5000
Rochester Multimedia Model
Window Sill Dusl-Leod Loading = 500 ^g/fl!
500
>
a Figure 5-2.
ro
oo
IVJ
o
o
o
Percentage of Children's Blood-Lead Concentrations, As Predicted By the HUD Model and the
Rochester Multimedia Model, That Will Exceed or Equal 10 ug/dL as a Function of Floor Dust-Lead
Loading for Five Soil-Lead Concentrations and Two Window Sill Dust-Lead Loadings (Geometric
Standard Devi at ion=1,60)
-------
o
X
o
o
z
o
o
H
m
O
3D
D
O
m
ro
100 200 300 tOO
floor Dusl-Liod boding ta/tl?)
ConcenttatkxiOVj):
- 100 - 400
-- 1200 -- 2000
— 5000
500
100 200 300 «00
FlooiDusHwd Leading to/ll!)
- 100 '400
-- 1200 — 2000
— 5000
500
100
200
300
400
HUDUoM
«r.do. $11 DusHMdUodinj =
Sol-liodC»K.nlrotiai(«/g): - 100 •• 400
-- 1200 — 2000
5000
RochKlir kuilimtdio liooel
Windoi Sill Ousi-lMd Loading - 500 jig/ll'
500
>
« Figure 5-3.
CD
Percentage of Children's Blood-Lead Concentrations, As Predicted By the HUD
Model and the Rochester Multimedia Model, That Will Exceed or Equal 10 ug/dL as a
Function of Floor Dust-Lead Loading for Five Soil-Lead Concentrations and Two
Window Sill Dust-Lead Loadings (Geometric Standard Deviation^ .72)
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Table 5-2. Floor Dust-Lead Loadings at Which the Percentage of Children Aged 1-2
Years With Blood-Lead Concentration At or Above 10 //g/dL is Estimated by
the HUD Model at 1, 5, or 10%, for Five Assumed Soil-Lead Concentrations
and Two Assumed Geometric Standard Deviations
Soil-Lead
Concentration (/ig/g)
Floor Dust-Lead Loading (//g/ft2) .
1%
5%
10%
GSD = 1.60
100
400
1200
2000
5000
1.88
0.90
0.50
0.39
0.24
8.93
4.29
2.40
1.83
1.12
20.49
9.83
5.49
4.19
2.58
GSD = 1.72
100
400
1200
2000
5000
0.83
0.40
0.22
0.17
0.10
5.01
2.40
1.34
1.02
0.63
13.05
6.26
3.50
2.67
1.64
DRAFT - DO NOT CITE OR QUOTE
205
August 28, 2000
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Table 5-3. Floor Dust-Lead Loadings at Which the Percentage of Children Aged 1-2
Years With Blood-Lead Concentration At or Above 10/ig/dL is Estimated by
the Rochester Multimedia Model at 1, 5, or 10%, for Five Assumed Soil-
Lead Concentrations, Two Assumed Window Sill Dust-Lead Loadings, and
Two Assumed Geometric Standard Deviations (expanded version of Table 5-
6 in §403 risk analysis report).
Soil-Lead
Concentration
0*9/9)
Window Sill
Dust-Lead
Loading (fig/ft2)
Roor Dust-Lead Loading (pg/ft*) - " "'
1%
Estimate
95% Upper
Confidence
Bound1
5%
EstimstB
95% Upper
ConfidBnco
Bound1
10%
Estimate
95% Upper
Confidence
Bound1
GSD = 1.60
luu
400
1200
200O
5000
200
500
200
500
200
500
200
500
200
500
0.05
0.02
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.37
0.14
0.04
0.02
0.01
0.00
0.00
0.00
0.00
0.00
6.70
2.00
0.61
0.18
0.09
0.03
0.04
0.01
0.01
0.00
22.00
8.93
2.81
1.12
0.59
0.22
028
0.10
0.07
0.02
89.08
26.62
8.13
2.43
1.22
0.36
0.50
0.15
0.10
0.03
327.73
92.01
20.32
9.24
4.86
1.99
2.47
0.96
0.69
0.25
GSD = 1.72
100
400
1200
2000
5000
200
500
200
500
200
500
200
500
200
500
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.04
0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.11
0.33
0.10
0.03
0.02
0.00
0.01
0.00
0.00
0.00
4.90
1.95
0.64
0.24
0.12
0.04
0.06
0.02
0.01
0.00
21.87
6.54
2.00
0.60
0.30
0.09
0.12
0.04
0.03
0.01
67.96
24.54
7.04
3.00
1.59
0.61
0.77
0.29
0.20
0.07
1 The 95% upper confidence bounds here differ from those in Table 5-6 of the S403 risk analysis report. The values here
are corrected for a mistake in the original computations.
DRAFT -- DO NOT CITE OR QUOTE
206
August 28. 2000
-------
1 as soil-lead concentrations decrease from 2000 ppm to 10 ppm (assuming, for the
2 Rochester multimedia model, that window sill dust-lead loadings are at their
3 estimated national median level; Tables F-1 and F-2 of Appendix F). Note that
4 for a fixed value of the geometric standard deviation (GSD) for the blood-lead
5 distribution, the average IQ decrement in the population that is associated with
6 lead exposure is a multiple of the geometric mean (as calculated in the §403 risk
7 analysis). Therefore, similar differences in predictions between the two models
g would occur for average IP decrement.
9 • If the geometric standard deviation (GSD) associated with the blood-lead
10 distribution is fixed, then as floor dust-lead loadings increase beyond 10 jig/ft2,
11 the predicted percentaee of children with blood-lead levels at or above 10 ug/dL
12 increases at a much faster rate under the HUD model (at a constant soil-lead
13 level). For example, if window sill dust-lead loading is at its estimated national
14 median and soil-lead concentration is below 2000 ppm, the predicted percentage
15 under the HUD model is at a minimum twice as large as the prediction under the
16 Rochester multimedia model. This difference in predictions gets even greater as
17 the assumed soil-lead concentration gets lower. For example, at a GSD of 1.6, a
18 floor dust-lead loading of 100 ug/ft2, and a soil-lead concentration of 10 ppm, the
19 prediction is over 7 times higher for the HUD model compared to the Rochester
20 multimedia model (13.1% versus 1.76%; Tables F-3 and F-4 of Appendix F). -
21 5.1.2 Estimates of Individual Risks from Applying the
22 Alternative Rochester Multimedia Model
23 The approach to estimating individual risks that was discussed and applied (using the
24 EUBK model and Rochester multimedia model) in Section 5.3 of the §403 risk analysis report
25 and was applied using the HUD model in Section 5.1.1 above was also applied using the
26 alternative Rochester multimedia model ("Model A") that was documented in Section 4.2 of this
27 report. Recall that this model uses 1) average dust-lead loadings on uncarpeted floors, 2) average
28 dust-lead loadings on window sills, 3) yardwide average dust-lead concentration, and 4) the
29 percentage of painted components containing deteriorated lead-based paint as predictor variables.
30 The blood-lead concentration predicted by this model is an estimate of the geometric mean of the
31 distribution of blood-lead concentration, which is assumed to be lognormal with geometric
32 standard deviation 1.6. The resulting distribution is then used to estimate the percentage of
33 children with blood-lead concentration at or above 10 ug/dL. This subsection documents the
34 results of estimating individual risks based on the alternative Rochester multimedia model.
35 In this analysis, it is assumed that the given residential environment for which individual
36 risks will be estimated has no deteriorated lead-based paint (as was done in Section 5.3 of the
37 §403 risk analysis report). Then, levels of two of the remaining three predictor variables are
38 fixed, and the level of the third variable is determined so that either 1 %, 5%, or 10% of children
DRAFT - DO NOT CITE OR QUOTE 207 August 28,2000
-------
1 exposed to the combined lead levels given by the three predictor variables would be predicted to
2 have a blood-lead concentration at or above 10 ug/dL.
3 Table 5-4a contains the estimated floor dust-lead loading at which the predicted
4 percentage of children with blood-lead concentration at or above 10 ug/dL is either 1%, 5%, or
5 10%, given that window sill dust-lead loading is at either 200 or 500 Mg/ft2, and soil-lead
6 concentration is at either 100 or 400 Mg/g. Table S-4b contains the estimated window sill dust-
7 lead loading at which the predicted percentage of children with blood-lead concentration at or
8 above 10 ug/dL is either 1%, 5%, or 10%, given that floor dust-lead loading is at either 25 or 100
9 Mg/ft2. and soil-lead concentration is at either 100 or 400 ug/g. These two tables correspond to
10 Tables 5-6 and 5-7, respectively, in Section 5.3 of the §403 risk analysis report, where the
11 estimates were based on the original Rochester multimedia model. Finally, Table 5-4c contains
12 the estimated yard-wide average soil-lead concentration at which the predicted percentage of
13 children with blood-lead concentration at or above 10 ug/dL is either 1%, 5%, or 10%, given that
14 floor dust-lead loading is at either 25 or 100 ug/ft2, and window sill dust-lead loading is at either
15 200 or 500 ug/ft2. (A corresponding table for the soil-lead concentration does not exist in
16 Section 5.3 of the §403 risk analysis report as the soil-lead concentration predictor variable in the
17 original Rochester multimedia model assumed only dripline soil rather than a yard-wide average,
18 and the IEUBK model did not accept dust-lead loadings as input.)
19 Effect on risk analysis: If the target percentage of children with elevated blood-lead
20 concentration is 5%, the estimates in Table 5-4a (fourth column) are only slightly higher than the
21 corresponding estimates in Table 5-6 of the §403 risk analysis report, suggesting that at the fixed
22 levels of soil-lead concentration and window sill dust-lead loading, the corresponding floor dust-
23 lead loading is nearly the same for the two methods determined by the two forms of the
24 multimedia model. However, the estimated floor dust-lead loadings are considerably smaller
25 (and below 50 ug/ft2) than in Table 5-6 of the §403 risk analysis report under 10% risk and
26 considerably larger (but still all below 1 ug/ft2) under a 1% risk.
27 For window sills (Table 5-4b), estimated dust-lead loadings are reduced under the
28 alternative Rochester multimedia model compared to the estimates in Table 5-7 of the §403 risk
29 analysis report at the 5% and 10% risk levels. At the 5% risk level, the estimated window sill
30 dust-lead loadings were at 40 ug/ft2 or lower at the specified soil-lead and uncarpeted floor dust-
31 lead levels (Table 5-4b, fourth column).
32 Yardwide average soil-lead concentration needed to be below 150 ug/g at each of the
33 specified levels of floor and window sill dust-lead loadings to achieve risks of 10% or lower, and
34 at 32 ug/g or lower to achieve risks of 5% or lower (Table 5-4c). The estimated soil-lead
35 concentration increases as the two specified dust-lead loadings decrease, suggesting that the soil-
36 lead standard would need to be less stringent as the dust-lead loading standards became more
37 stringent.
DRAFT - DO NOT CITE OR QUOTE 208 August 28,2000
-------
1
2
3
4
5
6
7
8
Table 5-4a. Uncarpeted Floor Dust-Lead Loadings at Which the Percentage of Children
Aged 1-2 Years With a Blood-Lead Concentration At or Above 10 //g/dL is
Estimated by the Alternative Rochester Multimedia Model (A) at 1 %. 5%, or
10%, Under Fixed Levels of Yard wide Average Soil-Lead Concentration and
Window Sill Dust-Lead Loading
Yardwide Average
Soil-Lead Cone.
0/g/g)
100
400
Window Sill Dust-
Lead Loading
U/g/ft2)
200
500
200
500
Uncarpeted Floor Dust-Lead Loading (//g/ft2) at Which the
Estimated % of Children With Blood-Lead Concentration
At or Above 10 //g/dL is ...
1%
0.47
0.24
0.12
0.06
5%
7.8
4.0
1.9
1.0
10%
35
18
8.6
4.5
10
11
12
13
Note: The percentages of 1%, 5%, and 10% were determined assuming that the blood-lead distribution is
lognormal with geometric mean as predicted by the alternative Rochester multimedia model (A) and a
geometric standard deviation of 1.6.
14
15
16
17
18
19
20
21
22
23
24
25
26
Table 5-4b. Window Sill Dust-Lead Loadings at Which the Percentage of Children Aged
1-2 Years With a Blood-Lead Concentration At or Above 10 //g/dL is
Estimated by the Alternative Rochester Multimedia Model (A) at 1%, 5%, or
10%, Under Fixed Levels of Yardwide Average Soil-Lead Concentration and
Uncarpeted Floor Dust-Lead Loading
Yardwide Average
Soil-Lead Cone.
(//g/g)
100
400
Uncarpeted Floor
Dust-Lead
Loading (//g/ft2)
25
100
25
100
Window Sill Dust-Lead Loading (/ig/ft2) at Which the
Estimated % of Children With Blood-Lead Concentration
At or Above 10 //g/dL is ...
1%
0.80
0.12
0.11
0.02
5%
40
5.8
5.7
0.80
10%
320
46
46
6.6
Note: The percentages of 1%, 5%, and 10% were determined assuming that the blood-lead distribution is
lognormal with geometric mean as predicted by the alternative Rochester multimedia model (A) and a
geometric standard deviation of 1.6.
DRAFT ~ DO NOT CITE OR QUOTE
209
August 28, 2000
-------
1
2
3
4
5
6
7
8
Table 5-4c. Yardwide Average Soil-Lead Concentration at Which the Percentage of
Children Aged 1-2 Years With a Blood-Lead Concentration At or Above 10
//g/dL is Estimated by the Alternative Rochester Multimedia Model (A) at
1%, 5%, or 10%, Under Fixed Levels of Dust-Lead Loadings for Uncarpeted
Floors and Window Sills
Uncarpeted Floor
Dust-Lead
Loading (jig/ft2)
25
100
Window Sill Dust-
Lead Loading
(/ig/ft2)
200
500
200
500
Yardwide Average Soil-Lead Concentration (/ig/g) at
Which the Estimated % of Children With Blood-Lead
Concentration At or Above 10 //g/dL is ...
1%
2.0
0.49
1.0
0.26
5%
32
8.0
16
4.2
10%
140
35
73
18
10
11
12
13
Note: The percentages of 1%, 5%, and 10% were determined assuming that the blood-lead distribution is
lognormal with geometric mean as predicted by the alternative Rochester multimedia model (A) and a
geometric standard deviation of 1.6.
14 5.1.3 Considering Potential Declines in Blood-Lead Concentration
15 from NHANES III Phase 2 Measures
16 The results of this subsection are an extension of the analysis in Section 5.4.3 of the §403
17 risk analysis report. In that subsection, the geometric mean blood-lead concentration of 3.14
18 ug/dL for children aged 1-2 years, estimated from data collected in Phase 2 of NHANES EH
19 (1991-1994), was assumed to be either 10%, 20%, or 30% lower, and the resulting impact on the
20 baseline risk estimates was investigated. This analysis was performed due to the likelihood of
21 continued decline in blood-lead concentrations in the U.S. population that has occurred in recent
22 years. It was desired to augment this analysis by considering an additional assumption on the
23 percentage decline since Phase 2 of NHANES HI: 50%.
24 Table 5-5 presents the baseline estimates of the blood-lead concentration and health effect
25 endpoints for children aged 1-2 years, where each blood-lead concentration measurement in
26 Phase 2 of NHANES m was reduced by the same amount: 10%, 20%, 30%, or 50%. Thus, the
27 analysis assumed a constant percentage decline for the entire blood-lead concentration
28 distribution as characterized by Phase 2 of NHANES IQ. This table is an extension of the results
29 presented in Table 5-11 of the §403 risk analysis report and includes the baseline estimates
30 reported in the risk analysis for comparison purposes (i.e., where no reduction is assumed).
31 Note that within NHANES ffl, the estimated geometric mean blood-lead concentration
32 for children aged 1-2 years declined from 4.05 ug/dL in Phase 1 to 3.14 ug/dL in Phase 2,
33 representing a 22.5% decline. This is within the range of declines being considered in the
DRAFT - DO NOT CITE OR QUOTE
210
August 28, 2000
-------
1
2
3
4
Table 5-5. Sensitivity Analysis for the Estimated Baseline Number and Percentage of
Children Aged 1-2 Years Having Specific Health Effect and Blood-Lead
Concentration Endpoints, Assuming Various Percentage Declines in Blood-
Lead Concentration Since Phase 2 of NHANES III
Health Effect and Blood-Lead
Concentration Endpoints
PbB2 20/sg/dL
PbB2 10//g/dL
IQ score less than 70
IQ score decrement 2 1
IQ score decrement 2 2
IQ score decrement 2 3
Average IQ score decrement
Geometric Mean (/sg/dL)
Numbers (%) of Children Aged 1-2 Years
Risk Analysis
Estimate (Table 5-1
of the §403 risk
analysis report)
46,800
(0.588%)
458,000
(5.75%)
9,130
(0.115%)
3,060,000
(38.5%)
863,000
(10.8%)
294,000
(3.70%)
1.06
3.14
Percentage Decline in Blood-Lead Concentration
Since NHANES III Phase 2
10%
30,900
(0.388%)
340,000
(4.27%)
8,610
(0.108%)
2,640,000
(33.2%)
669,000
(8.40%)
213,000
(2.68%)
0.951
2.82
20%
18,900
(0.238%)
239,000
(3.00%)
8,160
(0.102%)
2,190,000
(27.6%)
493,000
(6.19%)
146,000
(1.83%)
0.845
2.51
30%
10,600
(0.133%)
156,000
(1.96%)
7.760
(0.098%)
1,740,000
(21.8%)
340,000
(4.27%)
91.900
(1.15%)
0.740
2.20
50%
2,130
(0.0268%)
46,800
(0.588%)
7,140
(0.0897%)
863,000
(10.8%)
117,000
(1.47%)
25,200
(0.317%)
0.528
1.57
5
6
7
8
9
10
11
12
13
14
15 sensitivity analysis within Table 5-5. However, due to the NHANES HI survey design and how
16 this survey was performed, caution must be taken when interpreting observed differences in
17 results between the two phases of this survey.
18 Effect on risk analysis: According to Table 5-5, if it were assumed that a 50% across-
19 the-board decline in blood-lead concentration (resulting in a national geometric mean blood-lead
20 concentration of 1.57 ug/dL), this would reduce the estimated number of children whose blood-
21 lead concentration was at or above 20 ug/dL from 46,800 to 2,130, a decline of 95%, while the
22 estimated number at or above 10 ug/dL would be reduced by nearly 90%, to 46,800 children.
23 The 50% decline resulted in percentage declines of 72%, 86%, and 91 % for numbers of children
24 with IQ score decrements of 1,2, or 3, respectively, as a result of lead exposure. The estimated
25 average IQ decrement in the population due to lead exposure is cut in half under this assumption
26 (from 1.06 to 0.53 points), matching the assumed 50% decline in blood-lead concentration
27 because the IQ/blood-lead concentration relationship is assumed to be linear across the entire
28 range of blood-lead concentration. The effects of the lower assumed percentage declines (10%-
29 30%) were discussed in Section 5.4.3 of the §403 risk analysis report.
DRAFT » DO NOT CITE OR QUOTE
211
August 28, 2000
-------
1 The results in Table 5-5 are based on the assumption that the blood-lead concentrations
2 for each child in the population have been reduced by the same percentage since Phase 2 of
3 NHANES ffl. In reality, different subgroups have achieved different rates of change over this
4 time. However, considering different percentage declines for different subgroups would be very
5 difficult, and the resulting estimates of the health effect and blood-lead concentration endpoints
6 would likely differ only slightly from that observed in Table 5-5.
7 5.1.4 Considering How Baseline Environmental-Lead Levels May Have
8 Changed Since the HUD National Survey
9 Although interim data from the NSLAH (Section 3.1) have recently been made available
10 to this risk analysis and have been summarized throughout this report, the fact that the public
11 could not have reviewed these summaries during the public comment period limits the extent to
12 which these data could be considered in the rulemaking. Therefore, for purposes of the
13 rulemaking, data from the HUD National Survey continue to be the only nationally-
14 representative data source on baseline environmental-lead levels in the nation's housing stock.
15 Nevertheless, it was desired to estimate how changes in these environmental-lead levels that have
16 occurred since the HUD National Survey was conducted would affect the baseline (i.e., pre-
17 §403) risk characterization. Therefore, a sensitivity analysis was performed where the HUD
18 National Survey data was adjusted to reflect possible change in the distribution over time. The
19 adjusted data would yield a surrogate distribution of baseline environmental-lead levels. Several
20 alternative adjustments would be considered, and risk estimates based on each set of adjusted
21 data would be calculated.
22 To help in determining appropriate adjustments to the HUD National Survey data, the
23 summaries presented in Section 3.2 of this document compared the distribution of dust-lead and
24 soil-lead data reported in the HUD National Survey with distributions from other studies
25 performed more recently, but typically in specific locations that may not necessarily be
26 nationally-representative. These summaries showed that the distributions were quite consistent
27 across studies, suggesting that the distributions based from the HUD National Survey data, even
28 after converting from Blue Nozzle dust-lead loadings to wipe-equivalent loadings, are likely
29 adequate for characterizing environmental-lead levels even up to ten years after the survey. In
30 fact, the HUD National Survey data distributions were often centered at lower lead levels than in
31 the other studies. (A primary exception was household average dust-lead loadings, where data
32 from the interim NSLAH were considerably lower than in other studies, including the HUD
33 National Survey. It is currently uncertain of the degree to which the observed distribution from
34 the interim NSLAH reflects actual declines in dust-lead levels since the HUD National Survey.)
35 Furthermore, the nature of the distribution appears to be affected by housing age, with higher
36 lead levels associated with older housing.
37 For this sensitivity analysis, the following five alternatives for adjusting the HUD
38 National Survey data were made, in an effort to reflect more current environmental-lead levels in
39 households:
DRAFT - DO NOT CITE OR QUOTE 212 August 28, 2000
-------
1 • Average dust-lead loading and dust-lead concentration reduced by 20%, (yard-
2 wide) average soil-lead concentration reduced by 20%
3 • Average dust-lead loading/concentration reduced by 50%, average soil-lead
4 concentration reduced by 50%
5 • Average dust-lead loading/concentration reduced by 50%, average soil-lead
6 concentration reduced by 0%
7 • Average dust-lead loading/concentration reduced by 0%, average soil-lead
8 concentration reduced by 50%
9 • Average dust-lead loading/concentration increased by 25%, average soil-lead
10 concentration increased by 25%.
11 The dust-lead loading assumptions are assumed to be for both floors and window sills and are
12 made to the reported Blue Nozzle loadings (i.e., those estimates used as input to the empirical
13 model). The same changes are assumed for Blue Nozzle floor dust-lead concentration, which is
14 used as input to the DEUBK model.
15 Each of the above five sets of alternatives implies that the same percentage change would
16 be applied to data from each housing unit in the HUD National Survey. Thus, the resulting
17 national distribution of baseline environmental-lead levels would be a simple shift in the current
18 distribution used in the §403 risk analysis, with no change in the variability associated with the
19 distribution. Insufficient data exist to determine how a distribution's variability may have
20 changed, so it is assumed to remain unchanged. Within each set, different percentage changes
21 are considered for dust-lead loading and soil-lead concentration to allow for added flexibility in
22 how lead levels may have changed in different media. Four of the five sets represent declines
23 over time, which are expected due to the increased prevalence of homes with no lead-based paint
24 in the housing stock and the reduced likelihood of residual contamination associated with leaded
25 gasoline emissions. Nevertheless, one set representing an increase is considered, due to the
26 potential for the new survey to include housing with generally higher levels than those houses
27 included in the original survey and the continued potential for deteriorated lead-based paint and
28 other lead sources to contaminate dust and soil.
29 Note that in all five sets of adjustments, the assumed within-house geometric standard
30 deviation (GSD) remains equal to 1.6. Alternative values of this GSD assumption were
31 considered in the sensitivity analysis presented in Section 5.4.6 of the §403 risk analysis report.
32 Table 5-6 presents the pre-§403 model-based estimates for the health effect and blood-
33 lead concentration endpoints, under each of the above five sets of data adjustments, as calculated
34 based on blood-lead distribution generated from IEUBK and empirical model fits. For
35 comparison purposes, the table includes the estimates assuming no adjustments (i.e., as the data
36 were used in the §403 risk analysis) as reported in Table 5-2 of the §403 risk analysis report.
DRAFT -- DO NOT CITE OR QUOTE 213 August 28,2000
-------
1
2
3
4
5
6
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
Table 5-6. Sensitivity Analysis on How Changes in Household Average Baseline Dust-
Lead Loadings/Concentrations and Soil-Lead Concentration Impact Pre-§403
Estimates of Health Effect and Blood-Lead Concentration Endpoints for
Children Aged 1 -2 Years
Assumed Percentage Change in Average Dust-Lead Loadings and Concentrations
(Both Floor and Window Sill) and in Yard-wide Average Soil-Lead Concentration
Dust:
Soil:
No change
No change
20%
decrease
20%
decrease
50%
decrease
50%
decrease
50%
decrease
No change
No change
50%
decrease
25%
increase
25%
increase
Predicted Health Effect And Blood-Lead Concentration Endpoints (Based on Empirical Model)
PbB 220 (%)
PbB2lO(%)
IQ < 70 (%)
IQ decrement 2 1 (%)
IQ decrement 2 2 (%)
IQ decrement 2 3 (%)
Avg. IQ decrement
0.0278
1.54
0.0997
34.5
4.53
0.718
0.932
0.0212
1.28
0.0983
31.8
3.87
0.584
0.896
0.0117
0.849
0.0957
26.5
2.74
0.373
0.825
0.0187
1.17
0.0977
30.6
3.61
0.532
0.880
0.0176
1.13
0.0974
30.1
3.49
0.509
0.873
0.0364
1.85
0.101
37.2
5.27
0.877
0.969
Predicted Health Effect And Blood-Lead Concentration Endpoints (Based on IEUBK Model)
PbB 220 (%)
PbB2lO(%)
IQ < 70 (%)
IQ decrement 2! (%)
IQ decrement 22 (%)
IQ decrement 28 (%)
Avg. IQ decrement
2.24
12.4
0.146
50.4
19.9
8.95
1.40
1.39
9.33
0.131
45.1
15.8
6.46
1.24
0.427
4.60
0.110
34.6
8.97
2.90
0.978
0.957
7.28
0.121
40.3
12.8
4.92
1.12
1.44
9.70
0.132
46.4
16.4
6.72
1.26
3.06
15.3
0.160
55.4
23.8
11.3
1.56
Effect on risk analysis: The greatest total decline in baseline environmental-lead levels
being considered is the set containing 50% declines in both dust- and soil-lead levels (i.e., the
fourth column of Table 5-6). Under the empirical model, Table 5-6 indicates that the most
sensitive endpoints to the 50% decline in both dust-lead and soil-lead are the incidence of IQ
decrement of at least 3 and the incidence of blood-lead concentration of at least 10 ug/dL, where
declines of 48% and 45%, respectively, were observed in these estimates relative to no decline in
environmental-lead levels. The empirical model-based estimates appear to be more sensitive to
changes in soil-lead concentration than in dust-lead concentration, as lower estimates were
observed when soil-lead concentrations declined by 50% (and no change was made to dust-lead
loadings) than when dust-lead loadings declined by 50% (and no change was made to soil-lead
concentrations). This is explained by the empirical model's larger slope estimate for soil-lead
DRAFT - DO NOT CITE OR QUOTE
214
August 28, 2000
-------
1 concentration than for dust-lead loading in either floors or window sills (Table 4-3 of the §403
2 risk analysis report).
3 When the IEUBK model is used to estimate the distribution of blood-lead concentration,
4 corresponding declines of 68% and 63% were observed for incidence of IQ score decrement at or
5 above 3 and for incidence of blood-lead concentration at or above 10 ug/dL, respectively, when
6 50% declines are assumed for both dust-lead and soil-lead concentration (Table 5-6). However,
7 in this same scenario, the greatest decline (81%) among the endpoints is observed with the
8 incidence of blood-lead concentration at or above 20 ug/dL. This is a considerable decline
9 compared to the 24% decline for this endpoint observed under the empirical model. Contrary to
10 the type of finding observed under the empirical model, the IEUBK model-based estimates
11 appear to be more sensitive to changes in dust-lead concentration than in soil-lead concentration,
12 as lower estimates were observed when dust-lead concentrations declined by 50% (and no change
13 was made to soil-lead concentrations) than when soil-lead concentrations declined by 50% (and
14 no change was made to dust-lead concentrations).
15 Under both models, the last column in Table 5-6 shows that only modest increases in the
16 risk estimates were observed under the one adjustment assumption involving increases in
17 environmental-lead levels (i.e., 25% increases in both dust-lead and soil-lead levels).
18 5.1.5 Impact on the Estimated Incidence of IQ Point Decrement
19 Assuming Certain Thresholds on the IQ/Blood-Lead Relationship
20 As discussed in Chapter 4 of the §403 risk analysis report, results of the meta-analysis
21 documented in Schwartz (1994) indicate that an average IQ point loss of 0.257 is predicted for
22 every 1.0 ug/dL increase in blood-lead concentration, with no evidence of a threshold in this
23 relationship (i.e., non-zero blood-lead concentration below which the predicted IQ point loss is
24 zero). These results were used in the §403 risk analysis to characterize the IQ/blood-lead
25 relationship. Section 2.3 of this report provides additional justification for making these
26 assumptions.
27 As discussed in Section 2.3, some researchers have suggested that a non-zero threshold
28 exists in the IQ/blood-lead relationship. While no consensus on a single threshold has been
29 adopted among those making this conclusion, and such conclusions are occasionally made by
30 visual inspection of data rather than on statistical criteria, this sensitivity analysis considers the
31 impact of assuming non-zero blood-lead concentration thresholds on the baseline and model-
32 based pre-§403 risk estimates. A non-zero threshold will result in reduced estimates for health
33 effects measured by IQ decrement, as children with blood-lead concentrations below the
34 threshold will have an estimated IQ decrement of zero due to lead exposure.
35 Estimates of the IQ decrement parameters under the following thresholds are presented in
36 this subsection: 1,2,3,5,8, and 10 ug/dL. In addition, the estimates under an assumed
37 "threshold" of 0 ug/dL (i.e., those measured in the §403 risk analysis and meaning that any
38 blood-lead level, regardless of how small, would have an adverse effect on a child's IQ score) are
DRAFT - DO NOT CITE OR QUOTE 215 August 28. 2000
-------
1 presented for comparison purposes. The candidate threshold of 8 ug/dL has been suggested by
2 Rabinowitz et al. (1992), as discussed in Section 2.3. The candidate threshold of 10 ug/dL was
3 selected as it represents the action level reported by the Centers for Disease Control and
4 Prevention (Section 2.5.1 of the §403 risk analysis report). It also is representative of the higher-
5 level thresholds reported by some early studies; thresholds any higher than 10 ug/dL would result
6 in extremely low (and likely very underestimated) risk estimates and have been discounted by
7 more recent studies. Levels of 1, 2, 3, and 5 ug/dL represents possible candidates of a very low,
8 but positive, threshold. Such thresholds would not be detectable by many studies in the literature
9 as they tend to fall below the range of observed data or the detection limit of the blood-lead
10 measurement procedure.
11 The assumption of a positive threshold requires a minor modification to the method used
12 to predict IQ score decrement based on blood-lead concentration. An average IQ point loss of
13 0.257 continues to be predicted for every 1.0 ug/dL increase in blood-lead concentration, but
14 only above the assumed blood-lead concentration threshold value. Thus, if T represents the
15 threshold, then the predicted IQ score decrement at a blood-lead concentration of C would equal
16 0.257*(C-T) if C is greater than T, or zero if C is less than or equal to T. While the
17 methodology used to obtain risk estimates remains the same as that documented in Appendices
18 El and E2 of the §403 risk analysis report, slight differences were required for calculating the
19 average and standard deviation of IQ decrement, as this measure was no longer assumed to be
20 lognormally distributed. See Appendix B for how these statistics are calculated assuming a non-
21 zero threshold.
22 Table 5-7 presents the estimated percentages of children with IQ score decrements greater
23 than or equal to 1,2, or 3, and the average and standard deviation IQ point decrement under
24 assumptions of an IQ score decline of 0.257 points for every 1.0 Mg/dL increase in blood-lead
25 concentration above the specified threshold. These estimates are presented assuming the baseline
26 blood-lead distribution (top section of the table), the pre-§403 distribution as generated by
27 IEUBK model fits (middle section of the table), and the pre-§403 distribution as generated by the
28 empirical model fits (bottom section of the table) for children aged 1-2 years.
29 Effect on risk analysis: The magnitude of the assumed blood-lead concentration
30 threshold has a considerable impact on the percentage of children affected by decrements in IQ
31 score. As seen in Table 5-7, while the §403 risk analysis estimated an average IQ decrement of
32 1.06 points occurs due to lead exposure across the population of children aged 1-2 years, this
33 average declines by approximately 44% under a assumed threshold of 2 ug/dL (0.588 points) and
34 by 90% under a threshold of 8 ug/dL (0.103 points). An estimated 38.5% of children aged 1-2
35 years were expected to experience an IQ score decrement of at least 1 if a threshold was not
36 assumed. This percentage is decreased by approximately 50% under a threshold of 2 ug/dL
37 (19.6%) and by 90% under a threshold of 8 ug/dL (3.5%). The percentage decline is decreased
38 in magnitude as the lower limit of IQ score decrement increases to 3, but it remains at least a
39 39% decline for a threshold of 2 ug/dL and 83% for a threshold of 8 ug/dL.
DRAFT -- DO NOT CITE OR QUOTE 216 August 28, 2000
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
Table 5-7. Sensitivity Analysis on the Assumed Blood-Lead Concentration Threshold on
IQ Decrement and Its Impact on the Pre-§403 Estimates of IQ Decrement
Endpoints for Children Aged 1-2 Years
Assumed
Threshold
(pg/dU
% of Children Aged 1-2 Years with a Specified IQ
Decrement Due to Lead Exposure1
IQ Decrement *
1
IQ Decrement *
2
Baseline Estimates (Section 5.1
0
1
2
3
5
8
10
38.5
27.3
19.6
14.2
7.83
3.50
2.15
10.8
8.08
6.10
4.66
2.80
1.40
0.915
IQ Decrement *
3
Average IQ
Decrement
iff points)2
Standard
Deviation of IQ
Decrement2
.1 of §403 risk analysis report)
3.70
2.88
2.26
1.80
1.16
0.627
0.429
1.06
0.804
0.588
0.428
0.233
0.103
0.0638
0.895
0.891
0.860
0.802
0.666
0.494
0.408
Pre-§403 Estimates Based on IEUBK Model-Generated PbB Distribution
(Section 5.1.2 of §403 risk analysis report)
0
1
2
3
5
8
10
50.4
39.3
30.8
24.4
15.7
8.58
5.96
19.9
16.0
13.0
10.6
7.27
4.31
3.13
8.95
7.42
6.19
5.20
3.73
2.35
1.76
1.40
1.15
0.921
0.738
0.483
0.273
0.194
1.35
1.35
1.33
1.28
1.15
0.964
0.854
Pre-§403 Estimates Based on Empirical Model-Generated PbB Distribution
(Section 5.1.2 of §403 risk analysis report)
0
1
2
3
5
8
10
34.5
20.4
12.0
7.14
2.61
0.652
0.278
4.53
2.76
1.71
1.07
0.443
0.130
0.0613
0.718
0.464
0.330
0.202
0.0926
0.0312
0.0158
0.932
0.675
0.442
0.271
0.0972
0.0224
0.00912
0.538
0.537
0.514
0.453
0.309
0.162
0.108
1 A 0.257 IQ decrement is assumed for each 1.0;/g/dL increase in PbB above the assumed threshold (see Section 4.4.1 of
the §403 risk analysis report). Thus, the following hold:
• PIIQ 2 11 = PIPbB 2 (threshold + 3.9//g/dU
• PIIQ * 2] = PIPbB 2 (threshold + 7.8;ig/dL)l
• PIIQ 2 3] = PIPbB 2 (threshold + 11.7 j/g/dL)]
1 Average and standard deviation of IQ decrement are calculated assuming no IQ decrement occurs below the assumed
threshold, and a 0.257 IQ decrement is assumed for each 1.0 pg/dL increase in PbB above the threshold.
DRAFT - DO NOT CITE OR QUOTE
217
August 28, 2000
-------
1 Similar patterns of decline were seen for the pre-§403 estimates generated under the
2 EUBK and empirical model-based blood-lead distributions, with the empirical model predicting
3 greater reductions for the larger thresholds. These model-based estimates were used in the
4 procedure to characterize changes from baseline that occur in a post-§403 environment, which is
5 addressed in Chapter 6.
DRAFT -- DO NOT CITE OR QUOTE 218 August 28, 2000
-------
1 6.0 ANALYSIS OF EXAMPLE OPTIONS FOR THE §403 STANDARDS
2 Chapter 6 of the §403 risk analysis presented the methodology used to characterize
3 reductions to childhood health effect and blood-lead concentration endpoints expected to result
4 after interventions are conducted in response to the proposed §403 rule and applied this
5 methodology to a broad range of example options for standards. Assumptions were made on
6 post-intervention environmental-lead levels, which were applied to those HUD National Survey
7 housing units where a particular intervention was triggered as a result of having environmental-
8 lead levels that exceeded an example standard. Then, the IEUBK and empirical models were
9 used to generate the post-§403 blood-lead concentration distribution given post-§403
10 environmental-lead levels. These results, combined with similar model-based estimates in the
11 pre-§403 environment presented in Chapter 5, were used to obtain a final post-§403 blood-lead
12 distribution which was comparable to the baseline distribution generated by data from Phase 2 of
13 NHANES HI. This procedure was detailed in Chapter 6 and Appendix Fl of the §403 risk
14 analysis report. This was the distribution upon which the health effects and blood-lead
15 concentration endpoints were estimated in the post-§403 environment.
16 The risk management procedure in Chapter 6 of the §403 risk analysis report considered
17 example standards for the following risk assessment measures:
18 • Average floor dust-lead loading
19 • Average window sill dust-lead loading
20 • Average soil-lead concentration
21 • Amount of deteriorated lead-based paint requiring paint maintenance
22 • Amount of deteriorated lead-based paint requiring paint abatement
23 Note that the lead-based paint standards considered in the risk management procedure differed
24 somewhat from the standards proposed in the §403 rule (see Chapter 1 of this report), as the rule
25 considered only a single tier rather than a two-tiered standard.
26 Section 6.1 presents additional detail and results on the performance characteristics
27 analyses, a non-modeling data analysis procedure used by EPA to help establish levels of concern
28 within the §403 rule. Performance characteristics analyses cited in the §403 proposed rule are
29 detailed, and additional performance characteristics analyses performed after the proposed rule to
30 address public comments and to finalize the rule are presented.
31 Section 6.2 investigates the incidence of children with elevated blood-lead concentrations
32 in homes where no candidate standard is met or exceeded (i.e., children who would be "missed"
33 by a specified set of candidate standards).
34 Since the §403 risk analysis report was published, public comment resulted in an
35 additional investigation into the assumptions made in the risk management on average dust-lead
36 loading following an intervention involving dust cleaning (40 ug/ft2 on floors, 100 ug/ft2 on
DRAFT - DO NOT CITE OR QUOTE 219 August 28, 2000
-------
1 window sills). The results of this investigation are presented in Section 6.3. Based on this
2 investigation, the impact of alternative assumptions on post-intervention dust-lead loadings on
3 characterizing the reduction in risk as a result of implementing §403 rules was evaluated through
4 a sensitivity analysis presented in Section 6.4. Also included in Section 6.4 are sensitivity
5 analyses applied to baseline (pre-§403) data within Section 5.1 of this report to evaluate the
6 impact of potential changes to the HUD National Survey data and assumptions on non-zero
7 thresholds for the IQ/blood-lead relationship, where the analyses are implemented on data
8 representing the post-§403 environment.
9 6.1 PERFORMANCE CHARACTERISTICS ANALYSES
10 The procedures defined and discussed in the §403 risk analysis report used statistical
11 modeling techniques to characterize risks of lead exposure to children in the nation's housing
12 stock and how these risks may be reduced as a result of interventions performed to reduce lead-
13 based paint hazards in the housing stock under the §403 rule. While using the findings of this
14 risk analysis to evaluate options for the standards specified in the §403 rule, EPA also wished to
15 base its evaluation partially on a non-modeling approach using data from field studies that
16 measured lead levels in both children's blood and in the same environmental media targeted by
17 the §403 rule. In particular, given the data reported in these studies, EPA was interested in
18 observing how often a specified set of candidate standards would "trigger" interventions in
19 housing units within these studies and the extent to which these units contained a child with an
20 elevated blood-lead concentration (^ 10 ug/dL). Such an investigation provided useful
21 information on the performance of a specified set of candidate standards without some of the
22 complexities associated with making conclusions from statistical modeling analyses.
23 EPA employed performance characteristics analysis, sometimes referred to as
24 sensitivity/specificity analysis, as a non-modeling approach to evaluating candidate §403
25 standards. The underlying statistical principle of this approach involves conditional probabilities
26 and has been documented in references such as Fleiss (1981). This chapter presents the findings
27 of performance characteristics analyses applied to data from the Rochester Lead-in-Dust study.
28 Applying data from this study was highly appropriate under the objective to evaluate candidate
29 lead standards in the §403 rulemaking. The form of the study data used in this analysis is
30 discussed in detail within Section 6.1.1. The methods used to perform this performance
31 characteristics analysis are presented in Section 6.1.2. Section 6.1.3 presents the results of
32 performance characteristics analysis presented in the preamble and which were used in the §403
33 rulemaking. Finally, Section 6.1.4 presents additional performance characteristics analyses
34 performed after the §403 proposed rule was published, where these analyses considered other
35 sets of standards (including the standards specified in the §403 proposed rule) and other means of
36 handling data on amount of deteriorated paint within a household.
37 6.1.1 Data Used in The Performance Characteristics Analysis
38 The performance characteristics analysis was applied to data from the recently-conducted
39 Rochester Lead-in-Dust study. A summary of objectives and design information for this study is
DRAFT -- DO NOT CITE OR QUOTE 220 August 28.2000
-------
1 found in Section 3.2.2.2 of the §403 risk analysis report. The Rochester study data were selected
2 for this analysis for the following reasons:
3 • The study reported information for all media for which §403 standards were
4 proposed (e.g., dust-lead on floors and window sills, soil-lead, condition of lead-
5 based paint).
6 • The study measured blood-lead concentration in 205 children aged 12-31 months
7 who resided in the selected homes.
8 • The dust sampling methods used in this study included the wipe technique, from
9 which dust-lead loadings were measured.
10 • For some homes, soil was sampled from multiple locations (i.e., dripline and play
11 areas), allowing for yardwide average soil-lead concentration to be estimated.
12 • While homes and children were targeted for selection in this study, the selection
13 process was more random and more representative of a general population than is
14 the case with other lead exposure studies.
15 The primary concern with using data from the Rochester study in this analysis is the degree to
16 which the study may be considered representative of the nation as a whole. The study selected a
17 targeted sample which was limited to a single geographic area. The sample consisted of children
18 who had moderate exposure to lead in their home environment and did not necessarily include
19 children with very high or very low exposure to lead. In particular,
20 • 22.9% of the children in this study (47 children total) had blood-lead
21 concentrations at or above 10 ug/dL, compared to the national estimate of 5.9%
22 for children aged 1-2 years according to Phase 2 of the Third National Health and
23 Nutrition Examination Survey (NHANES HI) (CDC, 1997).
24 • The geometric mean blood-lead concentration for the study children was 6.38
25 ug/dL with a geometric standard deviation (GSD) of 1.85. This compares with a
26 geometric mean of 3.1 ug/dL and GSD of 2.09 estimated for U.S. children aged 1-
27 2 years according to Phase 2 of NHANES m (CDC, 1997).
28 • At least 84% of the housing units included in this study were built prior to 1940,
29 compared to the estimated 20% of the entire U.S. housing stock made within the
30 §403 risk analysis (Table 3-5 of USEPA, 1998). There is a well-documented
31 relationship between age of housing and presence of lead-based paint hazards.
32 • While geometric mean floor dust-lead loadings were comparable between the
33 Rochester study and the HUD National Survey (after converting the data to wipe-
34 equivalent loadings), whose results are considered nationally-representative,
DRAFT - DO NOT CITE OR QUOTE 221 August 28.2000
-------
1 geometric mean estimates of window sill dust-lead loading and soil-lead
2 concentration were higher for the Rochester study relative to HUD National
3 Survey estimates (Section 3.2).
4 Despite these limitations, the Rochester study is considered one of the best resources of data for
5 characterizing the relationship between children's blood-lead concentration and residential
6 environmental-lead levels, and therefore, for evaluating national standards for lead in the nation's
7 housing stock.
8 While data were available for 205 units in the Rochester study, somewhat fewer of these
9 units had values for all required data endpoints for this analysis. In particular, 177 units had data
10 reported on the amount of deteriorated lead-based paint, plus lead measurements for floor dust
11 (wipe), window sill dust (wipe), and soil (dripline and/or play area). Of these units, 77 had soil-
12 lead data for both dripline and play areas, thereby allowing an average concentration across these
13 two areas to be calculated.
14 For the analysis presented in the §403 proposed rule, the following five data endpoints
15 were calculated for each Rochester study housing unit:
16 • Area-weighted average uncarpeted floor wipe dust-lead loading (i.e., the measured
17 loading for each sample was weighted by the area of the sample when averaged)
18 • Area-weighted average window sill wipe dust-lead loading
19 • Average of dripline and play area soil-lead concentrations
20 • The percentage of interior painted components tested in the study that contained
21 lead-based paint (measurements at or above 1.0 mg/cm2) and some level of
22 deterioration (paint condition listed as fair or poor)
23 • The percentage of exterior painted components tested in the study that contained
24 lead-based paint and some level of deterioration.
25 Note that these endpoints are comparable to the standards included in the §403 proposed rule,
26 with the exception of the latter two paint-lead measurements. While the proposed §403 standard
27 for the paint component is expressed as a square footage of deteriorated lead-based paint for
28 components with large surface areas (2 ft2 for interior surfaces, 10 ft2 for exterior surfaces) or as
29 the percentage of total painted surface area that is deteriorated for components with small surface
30 areas (10%), no indication on the amount of deteriorated lead-based paint on a given component
31 (either in square feet or as a percentage of the total surface area) was recorded in the Rochester
32 study. Instead, each paint-lead measurement was associated with an indicator of the paint's
33 condition (good, fair, poor). Therefore, for this analysis, the amount of deteriorated lead-based
34 paint in a housing unit was taken to be the percentage of tested components in the housing unit
35 that contained lead-based paint along with some level of deterioration (i.e., condition of paint
DRAFT -- DO NOT CITE OR QUOTE 222 August 28,2000
-------
1 either fair or poor). This result was assumed to be a good estimate of the total amount of lead-
2 based paint in the unit that was deteriorated.
3 6.1.2 Analysis Approach
4 The performance characteristics analysis classified each housing unit in the Rochester
5 study according to two different criteria:
6 1. Whether or not the unit exceeded any of the candidate standards for the various
7 media being controlled.
8 2. Whether or not the unit contained a child with elevated blood-lead concentration
9 UlOug/dL).
10 The first criterion represented whether a housing unit was "triggered" for any intervention by
11 exceeding at least one candidate standard, while the second represented whether the unit
12 contained a child requiring attention as a result of having an elevated blood-lead concentration.
13 The first criterion was determined by noting whether the value for at least one of the five
14 endpoints mentioned at the end of the previous section exceeded the standard associated with the
15 type of measurement represented by that endpoint.
16 For a given set of candidate standards, the set of housing units in the Rochester study was
17 identified that had data for all of the above five endpoints. These units were classified according
18 to whether or not they achieved the above two criteria. These results are summarized in the
19 manner illustrated within the 2x2 frequency table in Table 6-1. From this information, the four
20 performance characteristics defined in Table 6-1 were then calculated: sensitivity, specificity,
21 positive predictive value (PPV), and negative predictive value (NPV). These characteristics
22 provide the necessary information for evaluating the sets of standards on their ability to target the
23 proper set of units for intervention.
24 In this analysis, a "false positive" corresponds to triggering a housing unit for intervention
25 when it does not contain a child with an elevated blood-lead concentration, while a "false
26 negative" corresponds to not triggering a housing unit containing a child with an elevated blood-
27 lead concentration. Note that the proportion of false positives is equal to one minus the
28 specificity, while the proportion of false negatives is equal to one minus the sensitivity.
29 While information from all four performance characteristics are important for evaluating
30 the performance of a given set of standards, typically one or two characteristics are given more
31 weight than the others in the performance evaluation process. For example, in the preamble,
32 EPA evaluated candidate standards for dust-lead loading on uncarpeted floors and window sills
33 according to whether the performance characteristics analysis yielded a value of NPV from 95 to
34 99 percent under the given set of standards. This implied that no more than 5% of children living
35 in housing units with environmental-lead levels below the standards would have elevated blood-
36 lead concentrations (i.e., at or above 10 jig/dL). More recent Agency inquiries have focused on
DRAFT - DO NOT CITE OR QUOTE 223 August 28,2000
-------
1
2
3
Table 6-1. Definitions of Performance Characteristics Used to Evaluate How Various
Combinations of Environmental-Lead Standards Classify Housing Units in the
Rochester Lead-in-Dust Study
Blood-Lead Concentration
At or Above 10 //g/dL?
In the above table, the lette
or above 10 //g/dL who live
specified standards. Letter;
equals a + b + c+d. From th
Performance
Characteristic
Sensitivity
(or True Positive Rate, or
1 - False Negative Rate)
Specificity
(or True Negative Rate, or
1 - False Positive Rate)
Positive Predictive Value
(PPV)
Negative Predictive Value
(NPV)
Yes
No
Any of the Standards Exceeded?
No
a
c
Yes
b
d
r 'b' represents the number of children which have a blood-lead concentration at
in a residence with environmental-lead levels that exceed at least one of the
i 'a', 'c', and 'd' represent similar counts. The total number of housing units
ese counts, the following performance characteristics are calculated:
Definition
Probability of a housing unit exceeding at least one
standard given that there is a resident child with an
elevated blood concentration U 1 0 //g/dL)
Probability of a housing unit not exceeding at least
one standard given that a resident child has a low
blood-lead concentration « 10 //g/dL).
Probability of a resident child having an elevated
blood-lead concentration (z 10 //g/dL) given that the
housing unit exceeds at least one standard.
Probability of a resident child having a low blood-
lead concentration « 1 0 //g/dL) given that the
housing unit does not exceed at least one standard.
Calculation
b/(a + b)
c/(c+d)
b/(b+d)
c/(a + c)
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22 the ability of candidate standards to "trigger" housing units containing elevated blood-lead
23 children, which corresponds to maximizing the sensitivity.
24 Figure 6-1 provides an example (based on hypothetical data) of an ideal situation for
25 selecting a single standard (e.g., dripline soil-lead concentration). In this example, a dripline
26 soil-lead concentration standard of 400 ug/g would result in all four performance characteristics
27 achieving their maximum value of 1 (or 100%). Thus, all homes triggered for intervention (i.e.,
28 exceeding the standard) would contain a child with an elevated blood-lead concentration, and all
29 homes containing a child with an elevated blood-lead concentration would be triggered for
30 intervention. This situation is very unlikely to occur typically. Therefore, in a less than ideal
31 situation (i.e., with typical data), one may wish to maximize each characteristic or some subset of
32 the most important characteristic(s). If all four characteristics are equally important, one
33 approach is to maximize the unweighted sum of the four characteristics. In the ideal situation
34 represented by Figure 6-1, this sum would equal 4 (or 400%). With actual data, however, this
35 sum will be less than 4. Figure 6-2 illustrates a situation (again, based on hypothetical data)
36 where both the NPV and sensitivity equal 100%, but the PPV and specificity are less than 100%.
37 This situation would be acceptable if only the NPV and sensitivity needed to be maximized.
DRAFT - DO NOT CITE OR QUOTE
224
August 28, 2000
-------
50
40
I
c
.2 30
a
u
*J
u
o
o
o
T3
O
20
I
•a
o
_o
S
10
400 800 1.200 1,600
Dripline Soil-Lead Concentration
2.000
3,400
Figure 6-1. Example of an Ideal Situation for Establishing Potential Dripline Soil-
Lead Standards
DRAFT -- DO NOT CITE OR QUOTE
225
August 28, 2000
-------
50
40
.2 30
a
c
ID
O
O
O
"§ 20
I
•O
o
o
CQ
10
400 800 1,200 1.600
Dripline Soil-Lead Concentration Cug/g)
2,000
2.400
Figure 6-2. Example of a Situation Where the Negative Predictive Value and
Sensitivity Equal 100%, but the Positive Predictive Value and Specificity
are Less than 100%
1 The performance characteristics analysis was repeated for different sets of standards. For
2 each analysis, the information within Table 6-1 was calculated, and those sets of standards that
3 maximized the desired performance criteria were identified.
4 The different analyses presented in the subsequent sections of this chapter were
5 performed on different subsets of housing units in the Rochester study. The results are purely
6 descriptive in that they represent combinations of candidate standards that meet the specified
7 performance criteria when considering the housing units in the Rochester study and are not based
8 on any underlying probability model. Different results are possible if this analysis were to be
9 applied to data from different studies. In addition, only point estimates of the performance
10 characteristics are presented. The uncertainty in these estimates is primarily dependent on
11 sample size, and to a lesser degree on measurement error.
DRAFT -- DO NOT CITE OR QUOTE
226
August 28, 2000
-------
1 6.1.3 Results Cited in the §403 Proposed Rule
2 The analysis presented in this section were cited in section B. 1 .d of Part IV of the
3 preamble. This section of the preamble contained a brief presentation of the information
4 presented in Section 6.1.2 above, then cited findings of analyses documented in a memorandum
5 dated 9/3/97 from Battelle (Ronald Menton and Warren Strauss) to EPA (Todd Holderman).
6 EPA requested that Battelle perform this analysis in an action item of a meeting between Battelle
7 and EPA on August 27, 1997. A copy of the cited memorandum is found in Appendix G.
8 The analyses presented in Appendix G were performed on data for the 77 housing units in
9 the Rochester study that had all necessary data for the analysis, including soil-lead concentrations
10 for both dripline and play areas. As the §403 proposed rule was to contain a yardwide average
11 soil-lead standard, it was desired to consider only those housing units that had soil-lead data for
12 both locations. The considerable reduction in the number of Rochester study housing units
13 whose data were considered in this analysis (from 205 to 77 units) was due primarily to the fact
14 that play-area soil-lead concentration was measured for less than half of the study units.
15 The combinations of candidate standards considered in this analysis were those requested
16 by EPA at the time, when EPA was actively considering candidate standards in the rulemaking.
17 These combinations included all 8x4x9x3=864 combinations of the following:
18 • uncarpeted floor dust-lead loading: 50,75,100,125, 150,175,200,400 ug/ft2
19 • window sill dust-lead loading: 100, 300,500, 800 pg/ft2
20 • average soil-lead concentration: 200, 300,400, 500,600,700, 900, 1000, 1500
21 Mg/g
22 • maximum of percent of interior/exterior painted surfaces with deteriorated lead-
23 based paint: 5, 10,20%
24 Note that the type of endpoint that represented the paint-lead measurement in this analysis (i.e.,
25 the last bullet) differed from the type of paint-lead standard that EPA ultimately proposed in the
26 §403 proposed rule.
27 The purpose of this analysis was to identify those sets of candidate standards (from the
28 864 combinations above) which, when applying the performance characteristics analysis under
29 those sets of standards, resulted in values of negative predictive value (as defined in Table 6-1
30 above) that met one of the following three criteria:
31 • NPV * 99%
32 • 95% z NPV < 99%
33 • 90% s NPV < 95%.
34 The findings of this analysis are documented in Tables 1 and 2 of Appendix G.
35 Twenty-one of the 77 housing units whose data were included in the analysis did not
36 exceed any of the candidate standards in at least one of the 864 combinations of candidate
DRAFT-DO NOT CITE OR QUOTE 227 August 28.2000
-------
1 standards. These housing units, with the values of the endpoints used to compare to the
2 candidate standards and children's blood-lead concentration, are listed in Table 6-2. This means
3 that the denominator of NPV (i.e., the number of housing units that do not exceed at least one of
4 the candidate standards being considered) never exceeded 21 across the 864 combinations. For
5 some combinations, the denominator was as small as 2. Furthermore, all but two of the 21 units
6 in Table 6-2 contained children with blood-lead concentrations below 10 ug/dL. As a result, the
7 value of NPV was no lower than 84.6% across all 864 combinations of candidate standards. At
8 least one of the above three criteria for NPV was met for 808 (93.5%) of the combinations. Of
9 these 808 combinations, NPV equaled 100% for 690 of the combinations, equaled 95% for seven
10 combinations, and was at least 90% but below 95% for the remaining 111 combinations.
11 All of the remaining 56 housing units in the analysis that are not represented in Table 6-2
12 exceeded either the soil-lead standard or one of the two paint standards (i.e., interior and/or
13 exterior) in each of the 864 combinations of candidate standards. That is, each of these houses
14 had at least one of the following:
15 • average soil-lead concentration of at least 1500 ug/g
16 • at least 20% of painted surfaces with deteriorated lead-based paint in the interior
17 and/or exterior.
18 Therefore, the 56 housing units not represented in Table 6-2 were triggered in each of the 864
19 combinations of candidate standards, without regard to the floor or window sill standards.
20 The results presented in this section led to the following conclusions stated in Part IV of
21 the preamble:
22 "For uncarpeted floors, dust-lead loadings ranged from 50 jug/ft2 to 400 \ig/f?
23 depending on the dust-lead loading on interior window sills and the soil-lead
24 concentration. For interior window sills, dust-lead loadings ranged from 100
25 Vg/ft2 to 800 ng/jf depending on the dust-lead loading on uncarpeted floors and
26 the soil-lead concentration. These ranges are significantly higher than the ranges
27 yielded by the multimedia approach."
28 "Soil-lead concentrations ranged from 200 ppm to 1,500 ppm depending on dust-
29 lead loadings on uncarpeted floors and interior window sills and the exceedance
30 probability."
31 The ranges cited in the preamble were precisely the lower and upper ranges of the candidate
32 standards considered in this analysis. These findings reflect the very high values of the NPV
33 across the combinations of standards considered in this analysis.
DRAFT -- DO NOT CITE OR QUOTE 228 August 28,2000
-------
1
2
Table 6-2. Set of 21 Housing Units in the Rochester Study in Which No Standard Was
Exceeded in at Least One of the 864 Combinations of Candidate Standards
Housing ID
00034
00132
00302
00637
00874
00974
01047
01062
01195
01228
01930
01971
01991
02290
02411
02837
03174
03360
03527
05343
05498
Statistics Compared to the Candidate Standards1
Floor Dust-
Lead Loading
fog/ft2)
63.60
17.30
2.55
59.00
12.90
14.90
20.83
12.40
12.25
3.37
19.35
5.10
15.50
2.65
10.48
4.29
18.60
12.43
6.08
10.30
19.15
Window Sill
Dust-Lead
Loading
(pg/ft2)
349.9
90.6
70.7
74.9
293.7
45.6
372.3
87.1
32.2
16.2
118.8
41.9
398.9
74.1
178.5
2.8
235.6
702.0
148.8
75.7
66.0
Average Soil-
Lead Cone.
(/ig/g)
438.5
268.0
124.5
950.0
102.9
51.1
574.3
447.5
830.5
419.0
773.4
506.0
104.0
465.0
828.5
458.5
625.5
912.0
539.5
552.0
1150.5
% of Interior
Components
with
Deteriorated
LBP
17
0
0
0
0
0
18
0
0
0
0
11
0
11
10
0
0
0
14
13
0
% of Exterior
Components
with
Deteriorated
LBP
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Blood-Lead
Cone. (//g/dL)
7.1
6.0
4.8
13.3
2.1
8.9
3.9
7.4
6.9
4.6
4.6
6.1
7.5
4.9
9.0
8.9
5.8
11.3
4.5
5.6
5.8
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
See Section 6.1.1 for the definitions of these statistics.
DRAFT -- DO NOT CITE OR QUOTE
229
August 28. 2000
-------
1 6.1.4 Results of Analysis on Specified Sets of Standards
2 The analyses presented in Section 6.1.3 were performed prior to release of the §403
3 proposed rule and contributed to the information presented in the preamble. Since the proposed
4 rule was released, EPA has requested additional performance characteristics analyses be
5 performed on various combinations of candidate standards, to address various issues raised
6 within the public comments to the proposed rule and in support of preparing the final §403 rule.
7 This section presents the results of these additional performance characteristics analyses.
8 As discussed in Section 6.1.3, one of the limitations of the analyses presented in the
9 preamble was the relatively small number of housing units (77) in the Rochester study whose
10 data were used in the analyses. This small number was primarily due to the lack of available
11 soil-lead concentrations from play areas and the desire to have soil-lead data for both dripline and
12 play areas in order to calculate a yardwide average. Thus, the additional analyses presented in
13 this section re-defined how the soil-lead measure was calculated (with different approaches taken
14 to this re-definition), thereby increasing the number of units whose data could be included in the
15 analysis.
16 6.1.4.1 Analyses Performed on 41 Combinations of Candidate Standards, in Three
17 Iterations. The candidate standards that were considered in this analysis were the following:
18 • uncaroeted floor dust-lead loading: 5, 10,20,25,40,50,100, 200 ug/ft2
19 • window sill dust-lead loading: 250 Mg/ft2
20 • vardwide average soil-lead concentration: 400, 1200,2000, 5000 ug/g
21 • amount of deteriorated lead-based paint: 2% of interior painted surfaces or 10% of
22 exterior painted surfaces.
23 Thus, different candidate standards for floor dust-lead loading and soil-lead concentration were
24 considered, while only a single candidate standard was considered for window sills (i.e., that
25 specified in the §403 proposed rule) and deteriorated lead-based paint. This analysis considered
26 a total of 41 combinations of candidate standards, corresponding to the 8x4=32 combinations of
27 the above candidates, as well as the additional 9 combinations:
28 • only the paint standards (1 additional combination)
29 • only the paint and soil-lead concentration standards (4 additional combinations)
30 • only the paint, soil-lead concentration, and window sill dust-lead loading
31 standards (4 additional combinations).
32 For each combination, the four performance characteristics were calculated and presented, as
33 well as the number of housing units that exceed at least one of the specified standards.
34 Note that the above candidate paint standard (percentage of paint that is deteriorated lead-
35 based paint) is not expressed in the manner that the proposed paint standard in the §403 proposed
36 rule was expressed (amount of deteriorated lead-based paint, in square feet). As discussed in
DRAFT - DO NOT CITE OR QUOTE 230 August 28,2000
-------
1 Section 6.1.1 above, the Rochester study measured only lead content in paint plus an indicator of
2 paint condition, and therefore, did not measure the surface area containing deteriorated lead-
3 based paint. For the Rochester study data, the above paint standard triggered all units with
4 deteriorated lead-based paint present, as the lowest observed non-zero percentage of deteriorated
5 lead-based paint among the study units was 8% for interior surfaces and 14% for exterior
6 surfaces.
7 Three iterations of this analysis was performed, with each iteration involving data for a
8 different number of housing units:
9 Iteration #1: Instead of requiring soil-lead concentrations be reported for both dripline
10 and play areas, as was done within the analysis cited in Section 6.1.3 above, average soil-
11 lead concentration was set equal to the reported concentration at one of these areas if no
12 concentration is reported for the other area. This approach permitted data for 177 housing
13 units to be used in the analysis.
14 Iteration #2: After taking the approach in iteration #1, any units that did not have soil-
15 lead concentration reported due to having no bare soil available from which to sample
16 were assigned a soil-lead concentration of 0 ppm. This approach was taken as Title IV of
17 TSCA restricts the §403 soil-lead hazard standard to bare soil and further assuming that
18 any covered soil at these units would not pose a soil-lead hazard. This approach
19 permitted data for 184 housing units to be used in the analysis.
20 Iteration #3: After taking the approach in iterations #1 and #2, the 21 remaining units
21 having missing data for at least one endpoint had an imputed value assigned to the
22 endpoint(s) equal to the average value across units within the same year-built category
23 (pre-1940,1940-1959, 1960-1979, post-1979) and having the same indicator of whether
24 or not lead-based paint is present in the unit. This method followed the same approach
25 taken in the §403 risk analysis (Section 3.3.1.1 of the §403 risk analysis report) to impute
26 data for housing units in the HUD National Survey. This approach permitted data for 205
27 housing units to be used in the analysis.
28 The results of each iteration are now presented.
29 Iteration #1: Data for 177 Housing Units.
30 Table 6-3 presents the results of the performance characteristics analyses performed on
31 data for 177 housing units (#1 above) under the 41 combinations of standards listed above. Note
32 from this table that the fixed paint standards (which were equivalent to finding any deteriorated
33 lead-based paint in the unit) triggered an intervention for nearly three-fourths of the 177 units.
34 These paint standards considered jointly with a soil-lead concentration standard of 400 ug/g
35 resulted in 100% sensitivity and negative predictive value regardless of the dust standards.
36 Sensitivity and negative predictive values of 100% were also met at a soil-lead concentration
37 standard of 1200 jig/g if the floor dust-lead loading standard was at 10 ug/ft2 and the window sill
DRAFT-DO NOT CITE OR QUOTE 231 August 28,2000
-------
Table 6-3. Results of Performance Characteristics Analysis Performed on Data for 177 Units in the Rochester Lead-in-Dust
Study for Specified Sets of Standards1
LBP = lead-based paint; EBL = elevated blood-lead level (:> 10//g/dL)
o
o
33
8
5
6
7
8
9
10
11 ro
12 g
13
14
15
16
17
18
19
20
21
22 |
23 |
24 »
25 |
Set of Standards
%of
Interior
Paint that
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
%of
Exterior
Paint that
is
Damaged
LBP
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Soil-
Lead
Cone.
(ppm)
—
400
400
400
400
400
400
400
400
400
400
1200
1200
1200
1200
Window
Sill Dust-
Lead
Loading
0/g/ffl
--
--
250
250
250
250
250
250
250
250
250
--
250
250
250
Floor Dust-
Lead
Loading
Uig/ft2)
--
--
--
200
100
50
40
25
20
10
5
-
--
200
100
# {%) of the
177 Housing
Units That
Are At or
Above At
Least One
Standard
132(74.6%)
154(87.0%)
156(88.1%)
156(88.1%)
156(88.1%)
156(88.1%)
156(88.1%)
156(88.1%)
159(89.8%)
168(94.9%)
173(97.7%)
137(77.4%)
141 (79.7%)
141 (79.7%)
141 (79.7%)
SENSITIVITY
# (%) of the 43
Housing Units
with EBL
Children That
Are At or
Above At Least
One Standard
36 (83.7%)
43(100%)
43 (100%)
43 (100%)
43 (100%)
43 (100%)
43 (100%)
43 (100%)
43 (100%)
43 (100%)
43 (100%)
39 (90.7%)
40 (93.0%)
40 (93.0%)
40 (93.0%)
SPECIFICITY
# (%) of the
134 Housing
Units with No
EBL Children
That Are At or
Above No
Standards
38 (28.4%)
23(17.2%)
21 (15.7%)
21 (15.7%)
21 (15.7%)
21 (15.7%)
21 (15.7%)
21 (15.7%)
18(13.4%)
9 (6.7%)
4 (3.0%)
36 (26.9%)
33 (24.6%)
33 (24.6%)
33 (24.6%)
POSITIVE
PREDICTIVE
VALUE
n <%) of
Housing Units
That Are At or
Above At Least
One Standard
That Have EBL
Children2
36/132(27.3%)
43/154(27.9%)
43/156(27.6%)
43/156(27.6%)
43/156(27.6%)
43/156(27.6%)
43/156(27.6%)
43/156(27.6%)
43/159(27.0%)
43/168(25.6%)
43/173(24.9%)
39/137(28.5%)
40/141 (28.4%)
40/141 (28.4%)
40/141 (28.4%)
NEGATIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above No
Standards That
Do Not Have
EBL Children3
38/45 (84.4%)
23/23 (100%)
21/21 (100%)
21/21 (100%)
21/21 (100%)
21/21 (100%)
21/21 (100%)
21/21 (100%)
18/18(100%)
9/9 (100%)
4/4 (100%)
36/40 (90.0%)
33/36(91.7%)
33/36(91.7%)
33/36(91.7%)
Sum of Four
Performance
Character-
istics (%)
223.8
245.1
243.2
243.2
243.2
243.2
243.2
243.2
240.5
233.3
227.8
236.0
237.7
237.7
237.7
-------
o
DO
Table 6-3. (cont.)
Set of Standards
%of
Interior
Paint that
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
%of
Exterior
Paint that
is
Damaged
LBP
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Soil-
Lead
Cone.
(ppm)
1200
1200
1200
1200
1200
1200
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
Window
Sill Dust-
Lead
Loading
fc/g/ft2)
250
250
250
250
250
250
--
250
250
250
250
250
250
250
250
250
Floor Dust-
Lead
Loading
(/ig/ft2)
50
40
25
20
10
5
-
-
200
100
50
40
25
20
10
5
# (%) of the
177 Housing
Units That
Are At or
Above At
Least One
Standard
142 (80.2%)
143 (80.8%)
143 (80.8%)
147(83.1%)
165(93.2%)
171 (96.6%)
135(76.3%)
140(79.1%)
140(79.1%)
140(79.1%)
141 (79.7%)
142 (80.2%)
142 (80.2%)
147(83.1%)
165(93.2%)
171 (96.6%)
SENSITIVITY
# (%) of the 43
Housing Units
with EBL
Children That
Are At or
Above At Least
One Standard
41 (95.3%)
41 (95.3%)
41 (95.3%)
42 (97.7%)
43 (100%)
43 (100%)
38 (88.4%)
39 (90.7%)
39 (90.7%)
39 (90.7%)
40 (93.0%)
40 (93.0%)
40 (93.0%)
42 (97.7%)
43 (100%)
43 (100%)
SPECIFICITY
# (%) of the
134 Housing
Units with No
EBL Children
That Are At or
Above No
Standards
33 (24.6%)
32 (23.9%)
32 (23.9%)
29(21.6%)
12(9.0%)
6 (4.5%)
37 (27.6%)
33 (24.6%)
33 (24.6%)
33 (24.6%)
33 (24.6%)
32 (23.9%)
32 (23.9%)
29(21.6%)
12(9.0%)
6 (4.5%)
POSITIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above At Least
One Standard
That Have EBL
Children2
41/142 (28.9%)
41/143 (28.7%)
41/143 (28.7%)
42/147 (28.6%)
43/165(26.1%)
43/171 (25.1%)
38/135(28.1%)
39/140(27.9%)
39/140(27.9%)
39/140(27.9%)
40/141 (28.4%)
40/142 (28.2%)
40/142 (28.2%)
42/147 (28.6%)
43/165(26.1%)
43/171 (25.1%)
NEGATIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above No
Standards That
Do Not Have
EBL Children3
33/35 (94.3%)
32/34(94.1%)
32/34(94.1%)
29/30 (96.7%)
12/12(100%)
6/6(100%)
37/42(88.1%)
33/37 (89.2%)
33/37 (89.2%)
33/37 (89.2%)
33/36(91.7%)
32/35(91.4%)
32/35(91.4%)
29/30 (96.7%)
12/12(100%)
6/6(100%)
Sum of Four
Performance
Character-
istics (%)
243.1
242.0
242.0
244.6
235.0
229.6
232.2
232.4
232.4
232.4
237.7
236.5
236.5
244.6
235.0
229.6
§
1
O
m
O
m
w
I
-------
Table 6-3. (cont.)
Set of Standards
%of
Interior
Paint that
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
%of
Exterior
Paint that
is
Damaged
LBP
10
10
10
10
10
10
10
10
10
10
Soil-
Lead
Cone.
(ppm)
5000
5000
5000
5000
5000
5000
5000
5000
5000
5000
Window
Sill Dust-
Lead
Loading
l/ig/ft2)
-
250
250
250
250
250
250
250
250
250
Floor Dust-
Lead
Loading
(Mg/ft2)
--
--
200
100
50
40
25
20
10
5
# (%) of the
177 Housing
Units That
Are At or
Above At
Least One
Standard
133(75.1%)
138(78.0%)
138(78.0%)
138(78.0%)
139(78.5%)
140(79.1%)
140(79.1%)
145 (81.9%)
164(92.7%)
171 (96.6%)
SENSITIVITY
# (%) of the 43
Housing Units
with EBL
Children That
Are At or
Above At Least
One Standard
37 (86.0%)
38 (88.4%)
38 (88.4%)
38 (88.4%)
39 (90.7%)
39 (90.7%)
39 (90.7%)
41 (95.3%)
43 (100%)
43 (100%)
SPECIFICITY
# (%) of the
134 Housing
Units with No
EBL Children
That Are At or
Above No
Standards
38 (28.4%)
34 (25.4%)
34 (25.4%)
34 (25.4%)
34 (25.4%)
33 (24.6%)
33 (24.6%)
30 (22.4%)
13(9.7%)
6 (4.5%)
POSITIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above At Least
One Standard
That Have EBL
Children2
37/133(27.8%)
38/138(27.5%)
38/138(27.5%)
38/138(27.5%)
39/139(28.1%)
39/140(27.9%)
39/140 (27.9%)
41/145 (28.3%)
43/164(26.2%)
43/171 (25.1%)
NEGATIVE
PREDICTIVE
VALUE
# {%) of
Housing Units
That Are At or
Above No
Standards That
Do Not Have
EBL Children3
38/44 (86.4%)
34/39 (87.2%)
34/39 (87.2%)
34/39 (87.2%)
34/38 (89.5%)
33/37 (89.2%)
33/37 (89.2%)
30/32 (93.8%)
13/13(100%)
6/6(100%)
Sum of Four
Performance
Character-
istics 1%)
228.6
228.5
228.5
228.5
233.6
232.4
232.4
239.8
235.9
229.6
§
1
o
m
O
31
O
§
m
1
2
3
4
6
7
8
9
10
11
12
13
14
15
16
1 Calculations are based on data from 177 of 205 units in the Rochester Lead-in-Oust study that had available data for average (wipe) floor dust-lead loading, average (wipe)
window sill dust-lead loading, average soil-lead concentration (across driphne and play areas, with only one of the two areas represented if no data existed for the other
area), percentage of interior lead-based paint that is deteriorated, and percentage of exterior lead-based paint that is deteriorated. Of these 177 units, 43 have children with
elevated blood-lead concentrations U 10//g/dL).
1 Cell entries are as follows: (number of homes at or above at least one standard that have EBL children)/(total number of homes at or above at least one standard), followed
by the corresponding percentage (in parentheses).
17 {= 3 Cell entries are as follows: (number of homes not at or above at least one standard that do not have EBL childrenl/ltotal number of homes not at or above any standard),
18 JJ followed by the corresponding percentage (in parentheses).
-------
1 dust-lead loading was at 250 ug/ft2, although the specificity declined considerably at these
2 standards. When the floor dust-lead loading standard was raised to 20 ug/ft2 in this situation,
3 both the sensitivity and negative predictive value remained above 95%. However, at soil-lead
4 standards of 1200 jig/g or higher, the 95% criterion for both sensitivity and negative predictive
5 value were no longer achieved once the floor dust-lead loading standard exceeded 20 ug/ft2.
6 Among the 32 combinations of standards included in Table 6-3, the sum of the four
7 performance characteristics (i.e., the last column of the table) was maximized at 244.6% at a
8 floor dust-lead loading standard of 20 ug/ft2 and a soil-lead standard of either 1200 or 2000 ug/g.
9 (The paint and window sill standards were fixed in each combination.)
10 The proposed §403 standards, assuming the different approach taken in this analysis to
11 interpreting the paint standards, resulted in a 93% sensitivity (40 of 43 units containing an
12 elevated blood-lead child are triggered) and nearly a 92% negative predictive value (see
13 shaded/bold row within Table 6-3). The sum of the four performance characteristics was
14 237.7%. Nearly 80% of the 177 units exceeded at least one of the proposed §403 standards.
15 Iteration #2: Data for 184 Housing Units.
16 Table 6-4 presents the same types of results as in Table 6-3, but it reflects analyses that
17 included data for seven additional housing units where soil-lead concentration was assumed to be
18 0 ug/g due to having no bare soil present for sampling (i.e., a total of 184 housing units). Only
19 one of these seven additional units contained a child with an elevated blood-lead concentration.
20 Slight reductions in the values of the performance characteristics were seen from Table
21 6-3 to Table 6-4 with the addition of these seven units. The one additional unit containing a
22 child with elevated blood-lead concentration did not exceed any of the paint, soil, or window sill
23 standards in the table and exceeded only floor dust-lead loading standards below 50 ug/ft2.
24 However, as in Table 2-3, sensitivity and negative predictive values of 100% (and the
25 considerable declines in specificity) continued to occur at a soil-lead concentration standard of
26 1200 ug/g if the floor dust-lead loading standard was at 10 ug/ft2 and the window sill dust-lead
27 loading was at 250 ug/ft2.
28 Despite the general declines in the values of the four performance characteristics from
29 Table 6-3, the largest observed value of the sum of these characteristics among the 32
30 combinations of standards (245.0) was slightly larger than in Table 6-3. This value was observed
31 for the same two combinations of standards for which the maximum occurred in Table 6-3: a
32 floor dust-lead loading standard of 20 ug/ft2 and a soil-lead standard of either 1200 or 2000 ug/g.
33 The proposed §403 standards, assuming the different approach taken in this analysis to
34 interpreting the paint standards, resulted in nearly a 91 % sensitivity and nearly a 90% negative
35 predictive value, which were slight declines from Table 6-3 (see shaded/bold row within Table
36 6-4). The sum of the four performance characteristics was 233.2%.
DRAFT--DO NOT CITE OR QUOTE 235 August 28,2000
-------
1 TJ Table 6-4. Results of Performance Characteristics Analysis Performed on Data for 184 Units in the Rochester Lead-in-Dust
2 • Study for Specified Sets of Standards1
o
3 o LBP = lead-based paint; EBL = elevated blood-lead level U 10/vg/dL)
o
m
O
33
Set of Standards
%of
Interior
Paint that
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
%of
Exterior
Paint that
is
Damaged
LBP
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Soil-
Lead
Cone.
(ppm)
--
400
400
400
400
400
400
400
400
400
400
1200
1200
1200
1200
Window
Sill Dust-
Lead
Loading
(/ig/ft2)
--
--
250
250
250
250
250
250
250
250
250
--
250
250
250
Floor Dust-
Lead
Loading
fcig/ft2)
--
-
-
200
100
50
40
25
20
10
5
--
--
200
100
# (%) of the
184 Housing
Units That
That Are At
or Above At
Least One
Standard
136(73.9%)
158(85.9%)
160(87.0%)
160(87.0%)
160(87.0%)
160(87.0%)
161 (87.5%)
161 (87.5%)
164(89.1%)
173(94.0%)
179(97.3%)
141 (76.6%)
145 (78.8%)
145 (78.8%)
145 (78.8%)
SENSITIVITY
» (%) of the 44
Housing Units
with EBL
Children That
That Are At or
Above At Least
One Standard
36(81.8%)
43 (97.7%)
43 (97.7%)
43 (97.7%)
43 (97.7%)
43 (97.7%)
44 (100%)
44 (100%)
44 (100%)
44 (100%)
44 (100%)
39 (88.6%)
40 (90.9%)
40 (90.9%)
40 (90.9%)
SPECIFICITY
# (%} of the
140 Housing
Units with No
EBL Children
That Are At or
Above No
Standards
40 (28.6%)
25(17.9%)
23(16.4%)
23(16.4%)
23(16.4%)
23(16.4%)
23(16.4%)
23(16.4%)
20 (14.3%)
11 (7.9%)
5 (3.6%)
38(27.1%)
35 (25.0%)
35 (25.0%)
35 (25.0%)
POSITIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above At Least
One Standard
That Have EBL
Children2
36/136(26.5%)
43/158(27.2%)
43/160(26.9%)
43/160(26.9%)
43/160(26.9%)
43/160(26.9%)
44/161 (27.3%)
44/161 (27.3%)
44/164(26.8%)
44/173(25.4%)
44/179(24.6%)
39/141 (27.7%)
40/145 (27.6%)
40/145 (27.6%)
40/145 (27.6%)
NEGATIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above No
Standards That
Do Not Have
EBL Children3
40/48 (83.3%)
25/26 (96.2%)
23/24 (95.8%)
23/24 (95.8%)
23/24 (95.8%)
23/24 (95.8%)
23/23 (100%)
23/23 (100%)
20/20 (100%)
11/11 (100%)
5/5 (100%)
38/43 (88.4%)
35/39 (89.7%)
35/39 (89.7%)
35/39 (89.7%)
Sum of Four
Performance
Character-
istics (%)
220.2
239.0
236.9
236.9
236.9
236.9
243.8
243.8
241.1
233.3
228.2
231.8
233.2
233.2
233.2
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
ro
o
o
-------
Table 6-4. (cont.)
Set of Standards
%of
Interior
Paint that
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
%of
Exterior
Paint that
is
Damaged
LBP
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Soil-
Lead
Cone.
(ppm)
1200
1200
1200
1200
1200
1200
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
Window
Sill Dust-
Lead
Loading
(//g/ftz)
250
250
250
250
250
250
—
250
250
250
250
250
250
250
250
250
Floor Dust-
Lead
Loading
fcig/ft2)
50
40
25
20
10
5
—
--
200
100
50
40
25
20
10
5
# (%) of the
184 Housing
Units That
That Are At
or Above At
Least One
Standard
146 (79.3%)
148 (80.4%)
148 (80.4%)
152(82.6%)
170(92.4%)
177(96.2%)
139(75.5%)
144 (78.3%)
144 (78.3%)
144 (78.3%)
145 (78.8%)
147 (79.9%)
147 (79.9%)
152(82.6%)
170(92.4%)
177(96.2%)
SENSITIVITY
# (%) of the 44
Housing Units
with EBL
Children That
That Are At or
Above At Least
One Standard
41 (93.2%)
42 (95.5%)
42 (95.5%)
43 (97.7%)
44 (100%)
44(100%)
38 (86.4%)
39 (88.6%)
39 (88.6%)
39 (88.6%)
40 (90.9%)
41 (93.2%)
41 (93.2%)
43 (97.7%)
44 (100%)
44(100%)
SPECIFICITY
» (%) of the
140 Housing
Units with No
EBL Children
That Are At or
Above No
Standards
35 (25.0%)
34 (24.3%)
34 (24.3%)
31 (22.1%)
14(10.0%)
7 (5.0%)
39 (27.9%)
35 (25.0%)
35 (25.0%)
35 (25.0%)
35 (25.0%)
34 (24.3%)
34 (24.3%)
31 (22.1%)
14(10.0%)
7 (5.0%)
POSITIVE
PREDICTIVE
VALUE
# (%) of
Housing Unjts
That Are At or
Above At Least
One Standard
That Have EBL
Children2
41/146(28.1%)
42/148 (28.4%)
42/148 (28.4%)
43/152(28.3%)
44/170(25.9%)
44/177(24.9%)
38/139(27.3%)
39/144(27.1%)
39/144(27.1%)
39/144(27.1%)
40/145 (27.6%)
41/147 (27.9%)
41/147 (27.9%)
43/152(28.3%)
44/170(25.9%)
44/177 (24.9%)
NEGATIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above No
Standards That
Do Not Have
EBL Children3
35/38(92.1%)
34/36 (94.4%)
34/36 (94.4%)
31/32 (96.9%)
14/14(100%)
7/7 (100%)
39/45 (86.7%)
35/40 (87.5%)
35/40 (87.5%)
35/40 (87.5%)
35/39 (89.7%)
34/37 (91.9%)
34/37(91.9%)
31/32(96.9%)
14/14(100%)
7/7 (100%)
Sum of Four
Performance
Character-
istics (%)
238.4
242.6
242.6
245.0
235.9
229.9
228.2
228.2
228.2
233.2
237.3
237.3
245.0
235.9
229.9
8
1
o
m1
O
DO
O
I
ro
co
ro
o
o
o
-------
Table 6-4. (cont.)
Set of Standards
%of
Interior
Paint that
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
%of
Exterior
Paint that
is
Damaged
LBP
10
10
10
10
10
10
10
10
10
10
Soil-
Lead
Cone.
(ppm)
5000
5000
5000
5000
5000
5000
5000
5000
5000
5000
Window
Sill Dust-
Lead
Loading
U/g/ft2)
--
250
250
250
250
250
250
250
250
250
Floor Dust-
Lead
Loading
foig/ft2)
~
--
200
100
50
40
25
20
10
5
# (%) of the
184 Housing
Units That
That Are At
or Above At
Least One
Standard
137(74.5%)
142(77.2%)
142 (77.2%)
142 (77.2%)
143 (77.7%)
145 (78.8%)
145 (78.8%)
150(81.5%)
169(91.8%)
177(96.2%)
SENSITIVITY
» (%] of the 44
Housing Units
with EBL
Children That
That Are At or
Above At Least
One Standard
37(84.1%)
38 (86.4%)
38 (86.4%)
38 (86.4%)
39 (88.6%)
40 (90.9%)
40 (90.9%)
42 (95.5%)
44 (100%)
44 (100%)
SPECIFICITY
» (%} of the
140 Housing
Units with No
EBL Children
That Are At or
Above No
Standards
40 (28.6%)
36 (25.7%)
36 (25.7%)
36 (25.7%)
36 (25.7%)
35 (25.0%)
35 (25.0%)
32 (22.9%)
15(10.7%)
7 (5.0%)
POSITIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above At Least
One Standard
That Have EBL
Children2
37/137(27.0%)
38/142 (26.8%)
38/142 (26.8%)
38/142 (26.8%)
39/143 (27.3%)
40/145 (27.6%)
40/145 (27.6%)
42/150(28.0%)
44/169(26.0%)
44/177(24.9%)
NEGATIVE
PREDICTIVE
VALUE
tt (%) of
Housing Units
That Are At or
Above No
Standards That
Do Not Have
EBL Children3
40/47(85.1%)
36/42 (85.7%)
36/42 (85.7%)
36/42 (85.7%)
36/41 (87.8%)
35/39 (89.7%)
35/39 (89.7%)
32/34(94.1%)
15/15(100%)
7/7 (100%)
Sum of Four
Performance
Character-
istics (%)
224.8
224.6
224.6
224.6
229.4
233.2
233.2
240.4
236.7
229.9
o
o
I
o
o
DO
i
m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
ro
w
oo
1 Calculations are based on data from 184 of 205 units in the Rochester Lead-in-Dust study that had available data for average (wipe) floor dust-lead loading, average (wipe)
window sill dust-lead loading, average soil-lead concentration (across driplme and play areas, with only one of the two areas represented if no data existed for the other
area), percentage of interior lead-based paint that is deteriorated, and percentage of exterior lead-based paint that is deteriorated. Homes having no reported soil-lead
concentration but with no bare soil reported are assumed to have a soil-lead concentration of 0 ppm for these calculations. Of these 184 units, 44 have children with
elevated blood-lead concentrations (2 10pg/dL).
1 Cell entries are as follows: (number of homes at or above at least one standard that have EBL children)/(total number of homes at or above at least one standard), followed
17 £ by the corresponding percentage (in parentheses).
18 ~ ' Cell entries are as follows: (number of homes not at or above at least one standard that do not have EBL children)/(total number of homes not at or above any standard),
19 _°° followed by the corresponding percentage (in parentheses).
§
-------
1 Iteration #3: Data for 205 Housing Units.
2 The third set of performance characteristics analyses was performed on data for all 205
3 housing units in the Rochester study. The previous analyses involved data for fewer housing
4 units as some units did not have recorded data for the key endpoints used in the analyses to
5 compare to the various candidate standards. Therefore, this analysis replaced incidences of
6 missing data with data values that were imputed from information available from other study
7 units. It was assumed that these imputed values were accurate estimates of what would have
8 been reported for these units. This estimate for a housing unit could vary considerably from what
9 would have been reported, however, based on actual conditions and behaviors in the household.
10 As all 205 housing units had reported values for child's blood-lead concentration and for
11 the percentage of tested interior components containing deteriorated lead-based paint, no
12 imputation was necessary for these two endpoints. The other four endpoints had at least one
13 housing unit with missing data. For each of these four endpoints, Table 6-5 contains the number
14 of housing units with missing data according to year-built category and whether or not the unit
15 contains lead-based paint, along with the imputed data value assigned to these units, which
16 equaled the average value across all units in that same category that had non-missing data. The
17 imputed data values depended on the year-built category and lead-based paint indicator as these
18 two variables are typically important predictors of these values. This same approach was used in
19 the §403 risk analysis to impute environmental-lead data values for HUD National Survey units
20 having missing data (see Section 3.3.1.1 of USEPA, 1998).
21 The data imputation process documented in Table 6-5 resulted in assigning imputed data
22 to 21 units: 19 built prior to 1940, one built from 1940-1959, and one built after 1979. A total of
23 eight average uncarpeted floor dust-lead loadings, nine average window sill dust-lead loadings,
24 six average soil-lead concentrations, and one percentage of deteriorated lead-based paint on
25 exterior surfaces were imputed.
26 Table 6-6 presents estimates of the four performance characteristics for the 41
27 combinations of standards, using reported and imputed data for 205 housing units in the
28 Rochester study. These estimates are very similar to those in Table 6-4 that were calculated from
29 data for 184 housing units. The same conclusions can be drawn from these results as were made
30 from the results in Tables 6-3 and 6-4. This implies that at the given combinations of candidate
31 standards considered in these analyses, the methods used in this section to estimate performance
32 characteristics were relatively robust across the different sets of data used in the analyses (i.e.,
33 177, 184, or 205 units).
34 As sensitivity and negative predictive value are the two performance characteristics of
35 most interest to Agency reviewers, the results for these two characteristics from Tables 6-3,6-4,
36 and 6-6 are summarized in Table 6-7. This summary emphasizes the relative stability of the
37 estimates across the different approaches used to make the calculations.
DRAFT - DO NOT CITE OR QUOTE 239 August 28,2000
-------
1
2
3
4
Table 6-5. Numbers of Housing Units with Missing Data for Four Endpoints and the
Imputed Data Values Assigned to These Units in This Analysis
Year-Built
Category
Pre-1940
1940-1959
1960-1979
Post-1979
Lead-
Based
Paint
Present?
Yes
No
Yes
No
-
Yes
No
Area-Weighted
Average Unearpeted
Floor Dust-Lead
Loading
# Units
with
Missing
Data
6
1
0
0
0
1
0
Imputed
Value
Uig/ft2)1
160.2
(157)
13.3
(8)
-
-
-
91.3
(3)
-
Area-Weighted
Average Window
Sill Dust-Lead
Loading
# Units
with
Missing
Data
5
3
1
0
0
0
0
Imputed
Value
(fig/ft2)1
633.2
(158)
95.2
(6)
569.0
(12)
-
--
—
-
Average Soil-Lead
Concentration
» Units
with
Missing
Data
5
1
0
0
0
0
0
Imputed
Value
U/g/g)1
1258
(158)
631.7
(8)
~
-
-
—
-
% of Exterior
Components
Containing
Deteriorated Lead-
Based Paint
# Units
with
Missing
Data
1
0
0
0
0
0
0
Imputed
Value
(%)'
25.2%
(162)
—
—
-
-
—
-
7
8
9
10
1 Number in parentheses equals the number of values (i.e., housing units) entering into calculation of the imputed value,
which is the average of these values.
DRAFT » DO NOT CITE OR QUOTE
240
August 28, 2000
-------
3 1
Table 6-6. Results of Performance Characteristics Analysis Performed on Data for 205 Units in the Rochester Lead-in-Dust
Study for Specified Sets of Standards1
LBP = lead-based paint; EBL = elevated blood-lead level U 10/yg/dL)
Set of Standards
%of
Interior
Paint that
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
%of
Exterior
Paint that
is
Damaged
LBP
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Soil-
Lead
Cone.
(ppm)
..
400
400
400
400
400
400
400
400
400
400
1200
1200
1200
1200
Window
Sill Dust-
Lead
Loading
0/g/ft2)
--
..
250
250
250
250
250
250
250
250
250
—
250
250
250
Floor Dust-
Lead
Loading
Oig/ft2)
--
«
~
200
100
50
40
25
20
10
5
--
--
200
100
# (%) of the
205 Housing
Units That
Are At or
Above At
Least One
Standard
151 (73.7%)
177(86.3%)
179(87.3%)
179(87.3%)
179(87.3%)
180(87.8%)
181 (88.3%)
181 (88.3%)
184(89.8%)
193(94.1%)
199(97.1%)
159(77.6%)
163(79.5%)
163(79.5%)
163 (79.5%)
SENSITIVITY
# (%) of the 48
Housing Units
with EBL
Children That
Are At or
Above At Least
One Standard
39(81.3%)
47 (97.9%)
47 (97.9%)
47 (97.9%)
47 (97.9%)
47 (97.9%)
48(100%)
48(100%)
48(100%)
48 (100%)
48 (100%)
43 (89.6%)
44(91.7%)
44(91.7%)
44(91.7%)
SPECIFICITY
# (%) of the
157 Housing
Units with No
EBL Children
That Are At or
Above No
Standards
45 (28.7%)
27(17.2%)
25(15.9%)
25(15.9%)
25(15.9%)
24(15.3%)
24(15.3%)
24(15.3%)
21 (13.4%)
12(7.6%)
6 (3.8%)
41 (26.1%)
38 (24.2%)
38 (24.2%)
38 (24.2%)
POSITIVE
PREDICTIVE
VALUE
» (%) of
Housing Units
That Are At or
Above At Least
One Standard
That Have EBL
Children2
39/151 (25.8%)
47/177(26.6%)
47/179(26.3%)
47/179(26.3%)
47/179(26.3%)
47/180(26.1%)
48/181 (26.5%)
48/181 (26.5%)
48/184(26.1%)
48/193(24.9%)
48/199(24.1%)
43/159(27.0%)
44/163(27.0%)
44/163(27.0%)
44/163(27.0%)
NEGATIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above No
Standards That
Do Not Have
EBL Children3
45/54 (83.3%)
27/28 (96.4%)
25/26 (96.2%)
25/26 (96.2%)
25/26 (96.2%)
24/25 (96.0%)
24/24(100%)
24/24 (100%)
21/21 (100%)
12/12(100%)
6/6 (100%)
41/46(89.1%)
38/42 (90.5%)
38/42 (90.5%)
38/42 (90.5%)
Sum of Four
Performance
Character-
istics (%)
219.1
236.3
236.3
236.3
235.3
241.8
241.8
239.5
232.5
227.9
233.3
233.3
233.3
O
33
4 g
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22 >
-------
Table 6-6. (cont.)
Set of Standards
%of
Interior
Paint that
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
%of
Exterior
Paint that
is
Damaged
LBP
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Soil-
Lead
Cone.
(ppm)
1200
1200
1200
1200
1200
1200
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
Window
Sill Dust-
Lead
Loading
(pg/ft2)
250
250
250
250
250
250
--
250
250
250
250
250
250
250
250
250
Floor Dust-
Lead
Loading
(pg/ft2)
50
40
25
20
10
5
--
--
200
100
50
40
25
20
10
5
# (%) of the
205 Housing
Units That
Are At or
Above At
Least One
Standard
165(80.5%)
167 (81.5%)
167(81.5%)
171 (83.4%)
190(92.7%)
197(96.1%)
155 (75.6%)
162(79.0%)
162(79.0%)
162(79.0%)
164 (80.0%)
166(81.0%)
166(81.0%)
171 (83.4%)
190(92.7%)
197 (96.1%)
SENSITIVITY
# (%) of the 48
Housing Units
with EBL
Children That
Are At or
Above At Least
One Standard
45 (93.8%)
46 (95.8%)
46 (95.8%)
47 (97.9%)
48 (100%)
48 (100%)
42 (87.5%)
43 (89.6%)
43 (89.6%)
43 (89.6%)
44(91.7%)
45 (93.8%)
45 (93.8%)
47 (97.9%)
48 (100%)
48 (100%)
SPECIFICITY
# (%) of the
157 Housing
Units with No
EBL Children
That Are At or
Above No
Standards
37 (23.6%)
36 (22.9%)
36 (22.9%)
33(21.0%)
15(9.6%)
8(5.1%)
44 (28.0%)
38 (24.2%)
38 (24.2%)
38 (24.2%)
37 (23.6%)
36 (22.9%)
36 (22.9%)
33(21.0%)
15 (9.6%)
8(5.1%)
POSITIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above At Least
One Standard
That Have EBL
Children2
45/165(27.3%)
46/167(27.5%)
46/167(27.5%)
47/171 (27.5%)
48/190(25.3%)
48/197(24.4%)
42/155(27.1%)
43/162(26.5%)
43/162(26.5%)
43/162(26.5%)
44/164 (26.8%)
45/166(27.1%)
45/166(27.1%)
47/171 (27.5%)
48/190(25.3%)
48/197(24.4%)
NEGATIVE
PREDICTIVE
VALUE
n i%) of
Housing Units
That Are At or
Above No
Standards That
Do Not Have
EBL Children3
37/40 (92.5%)
36/38 (94.7%)
36/38 (94.7%)
33/34(97.1%)
15/15(100%)
8/8 (100%)
44/50 (88.0%)
38/43 (88.4%)
38/43 (88.4%)
38/43 (88.4%)
37/41 (90.2%)
36/39 (92.3%)
36/39 (92.3%)
33/34(97.1%)
15/15(100%)
8/8 (100%)
Sum of Four
Performance
Character-
istics (%)
237.1
241.0
241.0
243.5
234.8
229.5
230.6
228.7
228.7
228.7
232.3
236.1
236.1
243.5
234.8
229.5
o
o
o
o
33
o
m
B
ro
(O
a
8
ro
-------
o
3)
Table 6-6. (cont.)
Set of Standards
%of
Interior
Paint that
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
%of
Exterior
Paint that
is
Damaged
LBP
10
10
10
10
10
10
10
10
10
10
Soil-
Lead
Cone.
(ppm)
5000
5000
5000
5000
5000
5000
5000
5000
5000
5000
Window
Sill Dust-
Lead
Loading
(//g/ft2)
—
250
250
250
250
250
250
250
250
250
Floor Dust-
Lead
Loading
0/g/ft2)
~
--
200
100
50
40
25
20
10
5
# {%) of the
205 Housing
Units That
Are At or
Above At
Least One
Standard
152(74.1%)
159(77.6%)
159(77.6%)
159(77.6%)
161 (78.5%)
163(79.5%)
163(79.5%)
168(82.0%)
189(92.2%)
197 (96.1%)
SENSITIVITY
# (%) of the 48
Housing Units
with EBL
Children That
Are At or
Above At Least
One Standard
40 (83.3%)
41 (85.4%)
41 (85.4%)
41 (85.4%)
42 (87.5%)
43 (89.6%)
43 (89.6%)
45 (93.8%)
48 (100%)
48 (100%)
SPECIFICITY
» (%) of the
157 Housing
Units with No
EBL Children
That Are At or
Above No
Standards
45 (28.7%)
39 (24.8%)
39 (24.8%)
39 (24.8%)
38 (24.2%)
37 (23.6%)
37 (23.6%)
34(21.7%)
16(10.2%)
8(5.1%)
POSITIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above At Least
One Standard
That Have EBL
Children2
40/152(26.3%)
41/159(25.8%)
41/159(25.8%)
41/159(25.8%)
42/161 (26.1%)
43/163(26.4%)
43/163(26.4%)
45/168(26.8%)
48/189 (25.4%)
48/197 (24.4%)
NEGATIVE
PREDICTIVE
VALUE
# (%) of
Housing Units
That Are At or
Above No
Standards That
Do Not Have
EBL Children3
45/53 (84.9%)
39/46 (84.8%)
39/46 (84.8%)
39/46 (84.8%)
38/44 (86.4%)
37/42(88.1%)
37/42(88.1%)
34/37 (91.9%)
16/16(100%)
8/8 (100%)
Sum of Four
Performance
Character-
istics (%)
223.2
220.8
220.8
220.8
224.2
227.6
227.6
234.1
235.6
229.5
o
m
o
3)
O
I
m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1 This analysis used the same data values used in Table 6-4, except missing values for the given endpomts were replaced by imputed numbers given in Table 6-5. Homes
having no reported soil-lead concentration but with no bare soil reported are assumed to have a soil-lead concentration of 0 ppm for these calculations. Of these 205 units,
48 have children with elevated blood-lead concentrations U 10/ig/dL)
2 Cell entries are as follows: (number of homes at or above at least one standard that have EBL children)/(total number of homes at or above at least one standard), followed
by the corresponding percentage (in parentheses).
16 ,§ 3 Cell entries are as follows: (number of homes not at or above at least one standard that do not have EBL children)/(total number of homes not at or above any standard),
17 g followed by the corresponding percentage (in parentheses).
3
-------
1 Table 6-7. Estimates of Sensitivity and Negative Predictive Value Presented in Tables
2 6-3, 6-4, and 6-6
• Sot of Stondords
%of
Interior
Paint that
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
%of
Exterior
Paint that
is
Damaged
LBP
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Soil-Lead
Cone.
(ppm)
-
400
400
400
400
400
400
400
400
400
400
1200
1200
1200
1200
1200
1200
1200
1200
1200
1200
2000
2000
2000
2000
2000
2000
Window
Sill Dust-
Lead
Loading
0/g/ft1)
-
-
250
250
250
250
250
250
250
250
250
-
250
250
250
250
250
250
250
250
250
--
250
250
250
250
250
Floor
Dust-Lead
Loading
big/ft2)
-
-
-
200
100
50
40
25
20
10
5
-
-
200
100
50
40
25
20
10
5
-
-
200
100
50
40
SENSITIVITY
(% of Housing Units with EBL
Children That Are At or Above
At Least One Standard)
Data for
177 units
(Table
6-3)
83.7%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
90.7%
93.0%
93.0%
93.0%
95.3%
95.3%
95.3%
97.7%
100%
100%
88.4%
90.7%
90.7%
90.7%
93.0%
93.0%
Data for
184 units
(Table
6-4)
81.8%
97.7%
97.7%
97.7%
97.7%
97.7%
100%
100%
100%
100%
100%
88.6%
90.9%
90.9%
90.9%
93.2%
95.5%
95.5%
97.7%
100%
100%
86.4%
88.6%
88.6%
88.6%
90.9%
93.2%
Data for
205 units
(Table
6-6)
81.3%
97.9%
97.9%
97.9%
97.9%
97.9%
100%
100%
100%
100%
100%
89.6%
91.7%
91.7%
91.7%
93.8%
95.8%
95.8%
97.9%
100%
100%
87.5%
89.6%
89.6%
89.6%
91.7%
93.8%
NEGATIVE PREDICTIVE VALUE
(% of Housing Units At or
Above No Standards That Do
Not Have EBL Children)
Data for
177 units
(Table
6-3)
84.4%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
90.0%
91.7%
91.7%
91.7%
94.3%
94.1%
94.1%
96.7%
100%
100%
88.1%
89.2%
89.2%
89.2%
91.7%
91.4%
Data for
184 units
(Table
6-4)
83.3%
96.2%
95.8%
95.8%
95.8%
95.8%
100%
100%
100%
100%
100%
88.4%
89.7%
89.7%
89.7%
92.1%
94.4%
94.4%
96.9%
100%
100%
86.7%
87.5%
87.5%
87.5%
89.7%
91.9%
Data for
205 units
(Table
6-6)
83.3%
96.4%
96.2%
96.2%
96.2%
96.0%
100%
100%
100%
100%
100%
89.1%
90.5%
90.5%
90.5%
92.5%
94.7%
94.7%
97.1%
100%
100%
88.0%
88.4%
88.4%
88.4%
90.2%
92.3%
DRAFT - DO NOT CITE OR QUOTE
244
August 28, 2000
-------
Table 6-7. (cont.)
Set of Standards
%of
Interior
Paint that
is
Damaged
LBP
2
2
2
2
2
2
2
2
2
2
2
2
2
2
%of
Exterior
Paint that
is
Damaged
LBP
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Soil-Lead
Cone.
(ppm)
2000
2000
2000
2000
5000
5000
5000
5000
5000
5000
5000
5000
5000
5000
Window
Sill Dust-
Lead
Loading
(fig/ft2)
250
250
250
250
-
250
250
250
250
250
250
250
250
250
Floor
Dust-Lead
Loading
(lig/ft2)
25
20
10
5
-
«
200
100
50
40
25
20
10
5
SENSITIVITY
(% of Housing Units with EBL
Children That Are At or Above
At Least One Standard)
Data for
177 units
(Table
6-3)
93.0%
97.7%
100%
100%
86.0%
88.4%
88.4%
88.4%
90.7%
90.7%
90.7%
95.3%
100%
100%
Data for
184 units
(Table
6-4)
93.2%
97.7%
100%
100%
84.1%
86.4%
86.4%
86.4%
88.6%
90.9%
90.9%
95.5%
100%
100%
Data for
205 units
(Table
6-6)
93.8%
97.9%
100%
100%
83.3%
85.4%
85.4%
85.4%
87.5%
89.6%
89.6%
93.8%
100%
100%
NEGATIVE PREDICTIVE VALUE
(% of Housing Units At or
Above No Standards That Do
Not Have EBL Children)
Data for
177 units
(Table
6-3)
91.4%
96.7%
100%
100%
86.4%
87.2%
87.2%
87.2%
89.5%
89.2%
89.2%
93.8%
100%
100%
Data for
184 units
(Table
6-4)
91.9%
96.9%
100%
100%
85.1%
85.7%
85.7%
85.7%
87.8%
89.7%
89.7%
94.1%
100%
100%
Data for
205 units
(Table
6-6)
92.3%
97.1%
100%
100%
84.9%
84.8%
84.8%
84.8%
86.4%
88.1%
88.1%
91.9%
100%
100%
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15 6.1.4.2 Considering only Soil and Dust Standards. The analysis in the previous
16 subsection emphasized the difficulty in evaluating candidate paint standards using the Rochester
17 data, not only due to the fact that the Rochester study did not measure total area corresponding to
18 deteriorated lead-based paint, but also that most of the housing units with deteriorated lead-based
19 paint exceeded the candidate standards that were considered in that analysis. In the analysis
20 presented in this subsection, a paint standard was not considered. Instead, the performance
21 characteristics analysis considered only candidate standards for soil-lead, floor dust-lead, and
22 window sill dust-lead, and then investigated the percentage of painted surfaces that contained
23 deteriorated lead-based paint for those houses that did not exceed any of these three candidate
24 standards, in an effort to characterize the extent to which these houses would possibly exceed a
25 paint standard. The candidate standards for dust and soil in this analysis were the same as in the
26 previous subsection:
27 • uncaroeted floor dust-lead loading: 5, 10,20,25,40,50,100,200 ug/ft2
DRAFT - DO NOT CITE OR QUOTE
245
August 28. 2000
-------
1 • window sill dust-lead loading: 250 fig/ft2
2 • vardwide average soil-lead concentration: 400, 1200,2000,5000 jig/g
3 The following 57 combinations of candidate standards were considered in this analysis:
4 • 8x 1x4=32 combinations of the candidate floor-dust, sill-dust, and soil standards
5 • 4x 1 =4 combinations of only the candidate soil and sill-dust standards
6 • 1x8=8 combinations of only the candidate floor-dust and sill-dust standards
7 • 4 candidate soil standards without the others
8 • 1 sill-lead standard without the others
9 • 8 candidate floor-lead standards without the others.
10 The analysis was applied to data for housing units in the Rochester study having data that could
11 be compared to each of the standards included in the given combinaton. Average soil-lead
12 concentration for housing units equaled the average of the dripline and play area soil-lead
13 measures. Units having either dripline soil-lead data or play area soil-lead data, but not both, had
14 an average soil-lead concentration equal to the reported concentration at the area represented by
15 the available data. An average soil-lead concentration of 0 ppm was assigned to housing units
16 having no soil-lead data and no bare soil from which to sample.
17 Table 6-8 contains the results of the performance characteristics analysis, with each row
18 of the table corresponding to one of the 57 combinations of candidate standards being
19 considered. The following are examples of how to interpret the findings within Table 6-8:
20 • Consider combinations of all three standards where the candidate soil-lead
21 standard is 400 ppm and window sill-dust standard is 250 ug/ft2. At an
22 uncarpeted floor-dust standard of 50 ug/ft2, only one of the 44 homes containing
23 children with elevated blood-lead concentration did not exceed any of these three
24 standards and did not contain any deteriorated lead-based paint. (Two other
25 homes with an elevated blood-lead child also do not exceed these dust or soil
26 standards, but they do contain some deteriorated lead-based paint.) Therefore,
27 under these standards, this particular unit would not be triggered for intervention,
28 regardless of the paint standard, despite the unit containing a child with an
29 elevated blood-lead concentration. However, if the uncarpeted floor-dust standard
30 was lowered to 40 ug/ft2, the house would exceed this lower floor standard.
31 • Consider the combination involving only a floor dust-lead standard of 20 Mg/ft2
32 and a window sill dust-lead standard of 250 fig/ft2. A total of 106 of the 188
33 homes met or exceeded at least one of these two standards, including 36 of the 45
34 homes with elevated blood-lead children. Of the 82 homes that did not meet or
35 exceed either dust standard, 9 contained an elevated blood-lead child, of which 2
36 had no deteriorated lead-based paint in either the interior or exterior. This means
37 that if only dust and paint standards were considered, these two homes would not
38 be triggered for any intervention, despite containing elevated blood-lead children.
DRAFT - DO NOT CITE OR QUOTE 246 August 28,2000
-------
Table 6-8. Results of Performance Characteristics Analysis Performed on Data for Housing Units in the Rochester Lead-in-
Dust Study, for Specified Sets of Candidate Standards for Lead in Dust and Soil Only
o
31
o
5 i
o
6 |
8 m
9
10
11
12
13
14
15
16
17
18
LBP = lead-based paint U 1.0 mg/cm2); EBL = elevated blood-lead level U 10 A/g/dL)
"Deteriorated lead-based paint" on a tested surface implies >5% of the lead-based paint is peeling, cracking, worn, chalking,
flaking, blistering, or otherwise separating from the substrate.
Sat of Candidate
Standards for Lead
in ...'
Soil
(ppm)
400
1200
2000
5000
--
--
--
--
Window
Sill Dust
Ovg/ft1)
-
--
--
--
250
-•
--
--
Floor
Dust
Ovg/ft2)
--
--
•-
-
--
200
100
50
# Units
Above
At Least
One
Standard
/Total #
Units2
142/198
55/198
26/198
6/198
73/195
5/196
9/196
19/196
Performance Characteristics
Sensitivity
# (%) of
Units with
EBL
Children
That Are
At or
Above At
Least One
Standard3
40/47
(85.1%)
22/47
(46.8%)
10/47
(21.3%)
3/47
(6.4%)
25/45
(55.6%)
3/47
(6.4%)
5/47
(10.6%)
9/47
(19.1%)
Soecificitv
# (%) of
Units with
No EBL
Children
That Are
At or
Above No
Standard*
49/151
(32.5%)
118/151
(78 1%)
135/151
(89.4%)
148/151
(98.0%)
102/150
(68.0%)
147/149
(98 7%)
145/149
(97.3%)
139/149
(93.3%)
PPV
# (%) Of
Units At
or Above
At Least
One
Standard
That Have
EBL
Children*
40/142
(28.2%)
22/55
(40.0%)
10/26
(38.5%)
3/6
(50.0%)
25/73
(34.2%)
3/5
(60 0%)
5/9
(55.6%)
9/19
(47.4%)
NPV
# (%) Of
Units At
or Above
No
Standard
That Do
Not Have
EBL
Children*
49/56
(87.5%)
118/143
(82.5%)
135/172
(78.5%)
148/192
(77.1%)
102/122
(83.6%)
147/191
(77.0%)
145/187
(77.5%)
139/177
(78.5%)
Sum of
the 4
Perfor-
mance
Charac-
teristics
(%)
233.2
247.5
227.6
231.5
241.4
242.0
241.0
238.3
# Units
with EBL
Children
That Are
At or
Above
No
Standard
and Have
No Deter-
iorated
LBP
1
5
6
8
7
8
8
7
# Units with EBL Children That
Are At or Above No Standard.
Where the % of Tested Interior
Paint Surfaces Having
Deteriorated LBP equals7 ...
0%
1
8
10
12
10
12
12
11
10-
30%
3
9
13
15
4
16
16
15
31-
50%
1
4
8
10
4
9
9
7
>50%
2
4
6
7
2
7
5
5
# Units with EBL Children That
Are At or Above No Standard,
Where the % of Tested
Exterior Paint Surfaces Having
Deteriorated LBP equals7 ...
0%
3
10
14
19
12
20
20
19
20-
50%
2
8
11
13
3
12
11
10
51-
75%
0
4
7
7
2
7
7
5
>75%
2
3
5
5
3
5
4
4
-------
o
3)
Table 6-8. (cont.)
Set of Candidate
Standards for Lead
in ...'
Soil
(ppm)
--
-
-•
-
--
400
1200
2000
5000
--
--
Window
Sill Dust
(pg/ft2)
--
--
--
--
--
250
250
250
250
250
250
Floor
Dust
(pg/ft2)
40
25
20
10
5
--
--
--
--
200
100
# Units
At or
Above
At Least
One
Standard
Hotel #
Units2
31/196
58/196
84/196
150/196
179/196
147/190
93/190
81/190
72/190
70/188
71/188
Performance Characteristics
Sensitivity
ff (%) of
Units with
EBL
Children
That Are
At or
Above At
Least One
Standard3
16/47
(34.0%)
26/47
(55.3%)
31/47
(66.0%)
44/47
(93.6%)
45/47
(95.7%)
41/44
(93.2%)
33/44
(75.0%)
27/44
(61.4%)
25/44
(56.8%)
25/45
(55.6%)
25/45
(55.6%)
Specificity
8 (%) of
Units with
No EBL
Children
That Are
At or
Above No
Standard*
134/149
(89.9%)
117/149
(78.5%)
96/149
(64.4%)
43/149
(28.9%)
15/149
(10.1%)
40/146
127.4%)
86/146
(58.9%)
92/146
(63.0%)
99/146
(67.8%)
98/143
(68.5%)
97/143
(67.8%)
PPV
tt (%) of
Units At
or Above
At Least
One
Standard
That Have
EBL
Children5
16/31
(51.6%)
26/58
(44.8%)
31/84
(36.9%)
44/150
(29.3%)
45/179
(25.1%)
41/147
(27.9%)
33/93
(35.5%)
27/81
(33.3%)
25/72
(34.7%)
25/70
(35.7%)
25/71
(35.2%)
NPV
# (%) of
Units At
or Above
No
Standard
That Do
Not Have
EBL
Children*
134/165
(81.2%)
117/138
(84.8%)
96/112
(85.7%)
43/46
(93.5%)
15/17
(88.2%)
40/43
(93.0%)
86/97
(88.7%)
92/109
(84.4%)
99/118
(83.9%)
98/118
(83.1%)
97/117
(82.9%)
Sum of
the 4
Perfor-
mance
Charac-
teristics
(%)
256.8
263.5
253.0
245.3
219.2
241.5
258.0
242.1
243.2
242.9
241.5
# Units
with EBL
Children
That Are
At or
Above
No
Standard
and Have
No Deter-
iorated
LBP
6
5
3
0
0
1
4
5
6
7
7
# Units with EBL Children That
Are At or Above No Standard,
Where the % of Tested Interior
Paint Surfaces Having
Deteriorated LBP equals7 ...
0%
9
8
6
0
0
1
6
8
9
10
10
10-
30%
13
7
6
3
2
1
1
4
4
4
4
31-
50%
7
4
3
0
0
0
2
3
4
4
4
>50%
2
2
1
0
0
1
2
2
2
2
2
ft Units with EBL Children That
Are At or Above No Standard.
Where the % of Tested
Exterior Paint Surfaces Having
Deteriorated LBP equals7 ...
0%
15
12
7
2
2
2
7
10
11
12
12
20-
60%
8
6
6
1
0
0
1
2
3
3
3
51-
75%
4
2
2
0
0
0
2
2
2
2
2
>75%
4
1
1
0
0
1
1
3
3
3
3
o
o
1
o
m
O
3J
O
m
00
to
-------
o
33
Table 6-8. (cont.)
Set of Candidate
Standards for Lead
in ...'
Soil
(ppm)
--
--
-
--
--
--
400
400
400
400
400
Window
Sill Dust
(pg/ft1)
250
250
250
250
250
250
250
250
250
250
250
Floor
Dust
(pg/fta>
50
40
25
20
10
5
200
100
50
40
25
# Units
At or
Above
At Least
One
Standard
/Total #
Units2
75/188
80/188
93/188
106/188
150/188
175/188
144/184
144/184
144/184
145/184
146/184
Performance Characteristics
Sensitivity
# (%) of
Units with
EBL
Children
That Are
At or
Above At
Least One
Standard3
27/45
(60.0%)
30/45
(66.7%)
34/45
(75.6%)
36/45
(80.0%)
44/45
(97.8%)
44/45
(97.8%)
41/44
(93.2%)
41/44
(93.2%)
41/44
(93 2%)
42/44
(95.5%)
42/44
(95.5%)
Soecificitv
# (%) of
Units with
No EBL
Children
That Are
At or
Above No
Standard*
95/143
(66.4%)
93/143
(65.0%)
84/143
(58.7%)
73/143
(51.0%)
37/143
(25.9%)
12/143
(8.4%)
37/140
(26.4%)
37/140
(26.4%)
37/140
(26.4%)
37/140
(26.4%)
36/140
(25.7%)
PPV
# (%) of
Units At
or Above
At Least
One
Standard
That Have
EBL
Children5
27/75
(36 0%)
30/80
(37.5%)
34/93
(36.6%)
36/106
(34.0%)
44/150
(29.3%)
44/175
(25.1%)
41/144
(28.5%)
41/144
(28.5%)
41/144
(28.5%)
42/145
(29.0%)
42/146
(28.8%)
NPV
# (%) of
Units At
or Above
No
Standard
That Do
Not Have
EBL
Children1
95/113
(84.1%)
93/108
(86.1%)
84/95
(88.4%)
73/82
(89.0%)
37/38
(97.4%)
12/13
(92.3%)
37/40
(92.5%)
37/40
(92.5%)
37/40
(92.5%)
37/39
(94.9%)
36/38
(94.7%)
Sum of
the 4
Perfor-
mance
Charac-
teristics
(%)
246.5
255.3
2593
254.0
250.4
223.6
240.6
240.6
240.6
245.7
244.7
# Units
with EBL
Children
That Are
At or
Above
No
Standard
and Have
Mo Deter-
iorated
LBP
6
5
4
2
0
0
1
1
1
0
0
# Units with EBL Children That
Are At or Above No Standard,
Where the % of Tested Interior
Paint Surfaces Having
Deteriorated LBP equals7 ...
0%
9
7
6
4
0
0
1
1
1
0
0
10-
30%
4
4
2
2
1
1
1
1
1
1
1
31-
50%
3
3
2
2
0
0
0
0
0
0
0
>50%
2
1
1
1
0
0
1
1
1
1
1
# Units with EBL Children That
Are At or Above No Standard.
Where the % of Tested
Exterior Paint Surfaces Having
Deteriorated LBP equals' ...
0%
11
9
7
5
1
1
2
2
2
1
1
20-
60%
3
2
2
2
0
0
0
0
0
0
0
61-
75%
1
1
1
1
0
0
0
0
0
0
0
>75%
3
3
1
1
0
0
1
1
1
1
1
O
o
Q
m
o
3)
O
I
m
S
to
ro
-------
Table 6-8. (cont.)
Set of Candidate
Standards for Lead
in ...'
Soil
(ppm)
400
400
400
1200
1200
1200
1200
1200
1200
1200
1200
Window
Sill Dust
(pg/ft2)
250
250
250
250
250
250
250
250
250
250
250
Floor
Dust
(pg/ft*)
20
10
5
200
100
50
40
25
20
10
5
# Units
At or
Above
At Least
One
Standard
/Total #
Units2
153/184
169/184
177/184
91/184
91/184
95/184
100/184
107/184
118/184
155/184
173/184
Performance Characteristics
Sensitivity
# (%) of
Units with
EBL
Children
That Are
At or
Above At
Least One
Standard3
42/44
(95.5%)
43/44
(97.7%)
43/44
(97.7%)
33/44
(75.0%)
33/44
(75.0%)
35/44
(79.5%)
38/44
(86.4%)
38/44
(86.4%)
39/44
(88.6%)
43/44
(97.7%)
43/44
(97.7%)
Specificity
# (%) of
Units with
No EBL
Children
That Are
At or
Above No
Standard*
29/140
(20.7%)
14/140
(100%)
6/140
(4.3%)
82/140
(58.6%)
82/140
(58.6%)
80/140
(57.1%)
78/140
(55.7%)
71/140
(50.7%)
61/140
(43.6%)
28/140
(20.0%)
10/140
(7.1%)
PPV
» (%) of
Units At
or Above
At Least
One
Standard
That Have
EBL
Children5
42/153
(27.5%)
43/169
(25.4%)
43/177
(24.3%)
33/91
(36 3%)
33/91
(36.3%)
35/95
(36.8%)
38/100
(38.0%)
38/107
(35.5%)
39/118
(33.1%)
43/155
(27.7%)
43/173
(24.9%)
NPV
ff (%) of
Units At
or Above
No
Standard
That Do
Not Have
EBL
Children8
29/31
(93.5%)
14/15
(93.3%)
6/7
(85.7%)
82/93
(88 2%)
82/93
(88 2%)
80/89
(89.9%)
78/84
(92.9%)
71/77
(92.2%)
61/66
(92.4%)
28/29
(96.6%)
10/11
(90.9%)
Sum of
the 4
Perfor-
mance
Charac-
teristics
(%)
2372
2265
212.0
258.0
258.0
263.4
272.9
264.8
257.7
242.0
220.6
# Units
with EBL
Children
That Are
At or
Above
No
Standard
and Have
No Deter-
iorated
LBP
0
0
0
4
4
3
2
2
1
0
0
# Units with EBL Children That
Are At or Above No Standard.
Where the % of Tested Interior
Paint Surfaces Having
Deteriorated LBP equals7 ...
0%
0
0
0
6
6
5
3
3
2
0
0
10-
30%
1
1
1
1
1
1
1
1
1
1
1
31-
50%
0
0
0
2
2
1
1
1
1
0
0
>50%
1
0
0
2
2
2
1
1
1
0
0
# Units with EBL Children That
Are At or Above No Standard.
Where the % of Tested
Exterior Paint Surfaces Having
Deteriorated LBP equals7 ...
0%
1
1
1
7
7
6
4
4
3
1
1
20-
60%
0
0
0
1
1
1
0
0
0
0
0
61-
75%
0
0
0
2
2
1
1
1
1
0
0
>75%
1
0
0
1
1
1
1
1
1
0
0
o
o
I
o
o
3)
i
m
ro
§
-------
Table 6-8. (cont.)
Set of Candidate
Standards for Lead
in ...'
Soil
(ppm)
2000
2000
2000
2000
2000
2000
2000
2000
5000
5000
5000
Window
Sill Dust
Uig/fta)
250
250
250
250
250
250
250
250
250
250
250
Floor
Dust
U/g/ft*>
200
100
50
40
25
20
10
5
200
100
50
tt Units
At or
Above
At Least
One
Standard
/Total #
Units1
79/184
79/184
83/184
88/184
99/184
112/184
152/184
172/184
70/184
71/184
75/184
Performance Characteristics
Sensitivity
#. (%) of
Units with
EBL
Children
That Are
At or
Above At
Least One
Standard3
27/44
(61.4%)
27/44
(61.4%)
29/44
(65.9%)
32/44
(72.7%)
35/44
(79.5%)
37/44
(84.1%)
43/44
(97.7%)
43/44
(97.7%)
25/44
(56.8%)
25/44
(56.8%)
27/44
(61.4%)
Specificity
# (%) of
Units with
No EBL
Children
That Are
At or
Above No
Standard*
88/140
(62.9%)
88/140
(62.9%)
86/140
(61.4%)
84/140
(60.0%)
76/140
(54.3%)
65/140
(46.4%)
31/140
(22.1%)
11/140
(7.9%)
95/140
(67.9%)
94/140
(67.1%)
92/140
(65.7%)
PPV
# (%) of
Units At
or Above
At Least
One
Standard
That Have
EBL
Children5
27/79
(34.2%)
27/79
(34.2%)
29/83
(34.9%)
32/88
(36 4%)
35/99
(35.4%)
37/112
(33.0%)
43/152
(28.3%)
43/172
(25.0%)
25/70
(35.7%)
25/71
(35.2%)
27/75
(36.0%)
NPV
# (%) of
Units At
or Above
No
Standard
That Do
Not Have
EBL
Children"
88/105
(83.8%)
88/105
(83.8%)
86/101
(85.1%)
84/96
(87.5%)
76/85
(89.4%)
65/72
(90.3%)
31/32
(96.9%)
11/12
(91.7%)
95/114
(83.3%)
94/113
(83.2%)
92/109
(84.4%)
Sum of
the 4
Perfor-
mance
Charac-
teristics
(%)
242.2
242.2
247.4
256.6
258.6
253.8
245.0
222.3
243.7
242.4
247.5
# Units
with EBL
Children
That Are
At or
Above
No
standard
and Have
No Deter-
iorated
LBP
5
5
4
3
3
1
0
0
6
6
5
# Units with EBL Children That
Are At or Above No Standard,
Where the % of Tested Interior
Paint Surfaces Having
Deteriorated LBP equals7 ...
0%
8
8
7
5
5
3
0
0
9
9
8
10-
30%
4
4
4
4
2
2
1
1
4
4
4
31-
50%
3
3
2
2
1
1
0
0
4
4
3
>50%
2
2
2
1
1
1
0
0
2
2
2
# Units with EBL Children That
Are At or Above No Standard,
Where the % of Tested
Exterior Paint Surfaces Having
Deteriorated LBP equals7 ...
0%
10
10
9
7
6
4
1
1
11
11
10
20-
60%
2
2
2
1
1
1
0
0
3
3
3
51-
76%
2
2
1
1
1
1
0
0
2
2
1
>76%
3
3
3
3
1
1
0
0
3
3
3
§
O
m
O
3)
ro
01
(Q
I
11 8
-------
Table 6-8. (cont.)
Set of Candidate
Standards for Lead
in ...'
Soil
(ppm)
5000
5000
5000
5000
5000
Window
Sill Dust
(/ig/ft2)
250
250
250
250
250
Floor
Dust
(pg/ft2)
40
25
20
10
5
ft Units
At or
Above
At Least
One
Standard
/Total #
Units2
80/184
92/184
105/184
148/184
172/184
Performance Characteristics
Sensitivity
« (%) of
Units with
EBL
Children
That Are
At or
Above At
Least One
Standard3
30/44
(68.2%)
33/44
(75.0%)
35/44
(79.5%)
43/44
(97.7%)
43/44
(97.7%)
Specificity
# (%) of
Units with
No EBL
Children
That Are
At or
Above No
Standard4
90/140
(64.3%)
81/140
(57.9%)
70/140
(50.0%)
35/140
(25.0%)
11/140
(7.9%)
PPV
# (%) of
Units At
or Above
At Least
One
Standard
That Have
EBL
Children"
30/80
(37.5%)
33/92
(35.9%)
35/105
(33.3%)
43/148
(29.1%)
43/172
(25.0%)
NPV
# (%) of
Units At
or Above
No
Standard
That Do
Not Have
EBL
Children"
90/104
(86.5%)
81/92
(88.0%)
70/79
(88.6%)
35/36
(97.2%)
11/12
(91.7%)
Sum of
the 4
Perfor-
mance
Charac-
teristics
(%)
256.5
2568
251.5
249.0
222.3
# Units
with EBL
Children
That Are
At or
Above
No
Standard
and Have
No Deter-
iorated
LBP
4
4
2
0
0
ff Units with EBL Children That
Are At or Above No Standard,
Where the % of Tested Interior
Paint Surfaces Having
Deteriorated LBP equals7 ...
0%
6
6
4
0
0
10-
30%
4
2
2
1
1
31-
60%
3
2
2
0
0
>50%
1
1
1
0
0
n Units with EBL Children That
Are At or Above No Standard.
Where the % of Tested
Exterior Paint Surfaces Having
Deteriorated LBP equals7 ...
0%
8
7
5
1
1
20-
60%
2
2
2
0
0
51-
75%
1
1
1
0
0
>75%
3
1
1
0
0
1
2
3
4
5
ro
1 The data compared to these standards are average (wipe) floor dust-lead loading, average (wipe) window sill dust-lead loading, and average soil-lead concentration (across dnpline end play
areas, with only one of the two areas represented if no data existed for the other area). Units having no reported soil-lead concentration but with no bare soil reported were assumed to have a
soil-lead concentration of 0 ppm.
1 Total number of units having available data that could be compared to ell specified candidate standards, as well as data on the percentage of tested interior lead-based paint that is deteriorated
end the percentage of tested exterior lead-based paint that is deteriorated
1 Cell entries arelnumber of homes at or above at least one standard that have EBL children)/ number of homes containing EBL children), followed by the corresponding percentage (in
parentheses)
4 Cell entries are (number of homes not at or above at least one standard that do not have EBL childrenl/ltotel number of homes not containing EBL children), followed by the corresponding
percentage (in parentheses)
' Cell entries are (number of homes at or above at least one standard that have EBL children)/(total number of homes at or above at least one standard), followed by the corresponding percentage
(m parentheses)
* Cell entries are (number of homes not at or above at least one standard that do not have EBL children)/(total number of homes not at or above any standard), followed by the corresponding
percentage (in parentheses).
' No housing units had between 0 and 10% deteriorated lead-based point on interior tested surfaces or between or between 0 and 20% deteriorated lead-based pemt on exterior tested surfeces.
-------
1 6.1.4.3 Analysis Involving Only Dust-Lead Standards and a Standard on the
2 Amount of Deteriorated Paint. In some cases, a risk assessment may involve only dust
3 sampling (of floors and window sills) and a visual inspection of painted surfaces for
4 deterioration. That is, no testing of painted surface for lead within the paint would be done, and
5 no soil sampling would be done. In this setting, it was of interest to investigate the extent to
6 which candidate dust-lead loading standards, with standards on the maximum percentage of
7 surfaces with deteriorated paint, performed in the absence of soil standards, within a performance
8 characteristics analysis. The combinations of standards considered in this analysis were the
9 following:
10 • uncarpeted floor dust-lead loading: 5, 10,20,25,40, 50,100 ug/ft2
11 • window sill dust-lead loading: 125,250 ug/ft2
12 • maximum amount of deteriorated paint on a tested surface: >5%, >15%.
13 The candidate paint standards were defined to coincide with the type of paint condition
14 measurement made in the Rochester study. The following 63 combinations of these candidate
15 standards were considered in this analysis:
16 • 7x2x2=28 combinations of the candidate floor-dust, sill-dust, and paint standards
17 • 7x2=14 combinations of only the candidate floor-dust and sill-dust standards
18 • 7x2= 14 combinations of only the candidate floor-dust and paint standards
19 • 7 candidate floor-lead standards without the others.
20 Table 6-9 contains the results of the performance characteristics analysis, with each row
21 of the table corresponding to one of the 63 combinations of candidate standards being
22 considered. The following are examples of what can be concluded from Table 6-9:
23 • While, on their own, the higher candidate floor dust-lead standards trigger few
24 units containing elevated blood-lead children, the number of these homes that are
25 triggered with the addition of a deteriorated paint standard increases dramatically
26 (e.g., from 10.6% to 70.2% at a floor dust-lead standard of 100 ug/ft2, if the 15%
27 paint standard is added).
28 • The performance characteristics do not appear to increase substantially with an
29 increase in the sill standard from 125 to 250 ug/ft2.
30 If the risk assessment does, in fact, do paint testing for lead, then the above standard for
31 paint can be re-defined to represent the maximum amount of deteriorated lead-based paint on a
32 tested surface. Table 6-10 contains the results of the performance characteristics analysis where
33 the paint standard is modified in this manner.
DRAFT - DO NOT CITE OR QUOTE 253 August 28.2000
-------
1 Table 6-9. Results off Performance Characteristics Analysis Performed on Data for
2 Housing Units in the Rochester Lead-in-Dust Study, for Specified Sets of
3 Candidate Standards for Dust-Lead Loadings and Observed Amount of
4 Damaged Paint on a Tested Surface
5 EBL = elevated blood-lead level U10//g/dL)
Set of Candidate Standards
Uncarpeted
Floor Dust-
Lead
Loading
1/ig/ft1)
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
Window
Sill Dust-
Lead
Loading
(/ig/ft2)
-
-
--
-
-
--
--
250
250
250
250
250
250
250
125
125
125
125
125
125
125
--
-
--
--
--
--
Max.
Arm. Of
Damaged
Paint on a
Tested
Surface
(%)'
-
-
-
-
--
-
-
-
-
-
-
--
-
-
--
-
-
--
-
-
--
>15%
>15%
>15%
>15%
>15%
>15%
# Units
At Or
Above
At Least
One
Standard
/ Total #
Units2
9/197
19/197
31/197
58/197
84/197
150/197
180/197
71/189
75/189
80/189
93/189
106/189
150/189
176/189
116/189
118/189
122/189
128/189
134/189
159/189
180/189
101/197
105/197
108/197
116/197
127/197
170/197
Performance Characteristics
Sensitivity
# (%) of Units
with EBL
Children That
Are At or Above
At Least One
Standard3
5/47 (10.6%)
9/47(19.1%)
16/47 (34.0%)
26/47 (55.3%)
31/47(66.0%)
44/47 (93.6%)
45/47 (95.7%)
25/45 (55.6%)
27/45 (60.0%)
30/45 (66.7%)
34/45 (75.6%)
36/45 (80.0%)
44/45 (97.8%)
44/45 (97.8%)
35/45 (77.8%)
36/45 (80.0%)
38/45 (84.4%)
39/45 (86.7%)
40/45 (88.9%)
45/45 (100%)
45/45 (100%)
33/47 (70.2%)
34/47 (72.3%)
35/47 (74.5%)
35/47 (74.5%)
38/47 (80.9%)
46/47 (97.9%)
Specificity
# (%) of Units
with No EBL
Children That Are
At or Above No
Standard4
146/150(97.3%)
140/150(93.3%)
135/150(90.0%)
118/150(78.7%)
97/150(64.7%)
44/150(29.3%)
15/150(100%)
98/144(68.1%)
96/144(667%)
94/144 (65.3%)
85/144(59.0%)
74/144(51.4%)
38/144(26.4%)
12/144(8.3%)
63/144 (43.8%)
62/144(43.1%)
60/144(41.7%)
55/144(38.2%)
50/144(34.7%)
30/144(20.8%)
9/144(6.3%)
82/150(54.7%)
79/150(52.7%)
77/150(51.3%)
69/150(46.0%)
61/150(40.7%)
26/150(17.3%)
PPV
# (%) of Units
At or Above At
Least One
Standard That
Have EBL
Children*
5/9 (55.6%)
9/19(47.4%)
16/31 (51.6%)
26/58 (44.8%)
31/84 (36.9%)
44/150(29.3%)
45/180(25.0%)
25/71 (35.2%)
27/75 (36.0%)
30/80 (37.5%)
34/93 (36.6%)
36/106 (34.0%)
44/150(29.3%)
44/176(25.0%)
35/116(30.2%)
36/118(30.5%)
38/122(31.1%)
39/128(30.5%)
40/134(29.9%)
45/159(28.3%)
45/180(25.0%)
33/101 (32.7%)
34/105 (32.4%)
35/108 (32.4%)
35/116(30.2%)
38/127(29.9%)
46/170(27.1%)
NPV
# (%) of Units At
or Above No
Standard That Do
Not Have EBL
Children8
146/188(77.7%)
140/178(78.7%)
135/166(81.3%)
118/139(84.9%)
97/113(85.8%)
44/47 (93.6%)
15/17(88.2%)
98/118(83.1%)
96/114(84.2%)
94/109 (86.2%)
85/96 (88.5%)
74/83 (89.2%)
38/39 (97.4%)
12/13(92.3%)
63/73 (86.3%)
62/71 (87.3%)
60/67 (89.6%)
55/61 (90.2%)
50/55 (90.9%)
30/30(100%)
9/9 (100%)
82/96 (85.4%)
79/92 (85.9%)
77/89 (86.5%)
69/81 (85.2%)
61/70(87.1%)
26/27 (96.3%)
DRAFT - DO NOT CITE OR QUOTE
254
August 28, 2000
-------
Table 6-9. (cont.)
Set of Candidate Standards
Uncarpeted
Floor Dust-
Lead
Loading
(//g/ft2)
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
Window
Sill Dust-
Lead
Loading
U/g/ft2)
-
-
-
--
-
-
-
-
250
250
250
250
250
250
250
125
125
125
125
125
125
125
250
250
250
250
250
250
250
Max.
Amt. Of
Damaged
Paint on a
Tested
Surface
(%)'
>15%
>5%
>5%
>5%
>5%
>5%
>5%
>5%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>5%
>5%
>5%
>5%
>5%
>5%
>5%
» Units
At Or
Above
At Least
One
Standard
/ Total #
Units2
188/197
164/197
165/197
167/197
167/197
168/197
185/197
192/197
118/189
120/189
123/189
127/189
134/189
167/189
182/189
142/189
144/189
147/189
149/189
152/189
171/189
182/189
162/189
163/189
165/189
165/189
166/189
179/189
185/189
Performance Characteristics
Sensitivity
#(%> of Units
with EBL
Children That
Are At or Above
At Least One
Standard3
47/47 (100%)
43/47(91.5%)
44/47 (93.6%)
45/47 (95.7%)
45/47 (95.7%)
45/47 (95.7%)
47/47 (100%)
47/47 (100%)
36/45 (80.0%)
37/45 (82.2%)
38/45 (84.4%)
38/45 (84.4%)
40/45 (88.9%)
45/45 (100%)
45/45 (100%)
39/45 (86.7%)
40/45 (88.9%)
41/45(91.1%)
41/45(91.1%)
42/45 (93.3%)
45/45 (100%)
45/45 (100%)
41/45(91.1%)
42/45 (93.3%)
43/45 (95.6%)
43/45 (95.6%)
43/45 (95.6%)
45/45 (100%)
45/45 (100%)
Specificity
# (%) of Units
with No EBL
Children That Are
At or Above No
Standard4
9/150(6.0%)
29/150(19.3%)
29/150(19.3%)
28/150(18.7%)
28/150(18.7%)
27/150(18.0%)
12/150(8.0%)
5/150(3.3%)
62/144(43.1%)
61/144(42.4%)
59/144(41.0%)
55/144(38.2%)
50/144(34.7%)
22/144(15.3%)
7/144(4.9%)
41/144(28.5%)
40/144(27.8%)
38/144(26.4%)
36/144(25.0%)
34/144(23.6%)
18/144(12.5%)
7/144 (4 9%)
23/144(16.0%)
23/144(16.0%)
22/144(15.3%)
22/144(15.3%)
21/144(14.6%)
10/144 (6.9%)
4/144(2.8%)
PPV
# (%) of Units
At or Above At
Least One
Standard That
Have EBL
Children5
47/188 (25.0%)
43/164(26.2%)
44/165(26.7%)
45/167(26.9%)
45/167(26.9%)
45/168 (26.8%)
47/185(25.4%)
47/192(24.5%)
36/118(30.5%)
37/120(30.8%)
38/123(30.9%)
38/127(29.9%)
40/134(29.9%)
45/167(26.9%)
45/182 (24.7%)
39/142 (27.5%)
40/144(27.8%)
41/147(27.9%)
41/149(27.5%)
42/152(27.6%)
45/171 (26.3%)
45/182(24.7%)
41/162(25.3%)
42/163(25.8%)
43/165(26.1%)
43/165(26.1%)
43/166(25.9%)
45/179(25.1%)
45/185(24.3%)
NPV
# (%) of Units At
or Above No
Standard That Do
Not Have EBL
Children*
9/9 (100%)
29/33 (87.9%)
29/32 (90.6%)
28/30 (93.3%)
28/30 (93.3%)
27/29(93.1%)
12/12(100%)
5/5 (100%)
62/71 (87.3%)
61/69 (88.4%)
59/66 (89.4%)
55/62 (88.7%)
50/55 (90.9%)
22/22 (100%)
7/7 (100%)
41/47 (87.2%)
40/45 (88.9%)
38/42 (90.5%)
36/40 (90.0%)
34/37(91.9%)
18/18(100%)
7/7 (100%)
23/27 (85.2%)
23/26 (88.5%)
22/24(91.7%)
22/24(91.7%)
21/23(91.3%)
10/10 (100%)
4/4 (100%)
DRAFT - DO NOT CITE OR QUOTE
255
August 28, 2000
-------
Table 6-9. (cont.)
Set of Candidate Standards
Uncarpeted
Floor Dust-
Lead
Loading
(pg/ft2)
100
50
40
25
20
10
5
Window
Sill Dust-
Lead
Loading
Uig/ff)
125
125
125
125
125
125
125
Max.
Amt. Of
Damaged
Paint on a
Tested
Surface
(%)'
>5%
>5%
>5%
>5%
>5%
>5%
>5%
# Units
At Or
Above
At Least
One
Standard
/ Total #
Units2
166/189
167/189
169/189
169/189
170/189
179/189
185/189
Performance Characteristics
Sensitivity
» (%) of Units
with EBL
Children That
Are At or Above
At Least One
Standard3
42/45 (93.3%)
43/45 (95.6%)
44/45 (97.8%)
44/45 (97.8%)
44/45 (97.8%)
45/45(100%)
45/45 (100%)
Specificity
# (%) of Units
with No EBL
Children That Are
At or Above No
Standard*
20/144(13.9%)
20/144(13.9%)
19/144(13.2%)
19/144 (13.2%)
18/144(12.5%)
10/144 (6.9%)
4/144(2.8%)
PPV
# (%) of Units
At or Above At
Least One
Standard That
Have EBL
Children6
42/166(25.3%)
43/167(25.7%)
44/169(26.0%)
44/169(26.0%)
44/170(25.9%)
45/179(25.1%)
45/185(24.3%)
NPV
# (%) of Units At
or Above No
Standard That Do
Not Have EBL
Children*
20/23 (87.0%)
20/22 (90.9%)
19/20(95.0%)
19/20(95.0%)
18/19(94.7%)
10/10(100%)
4/4 (100%)
1
2
3
4
5
6
7
4
1 In the Rochester study, each measurement of lead in paint had the amount of damaged paint specified as "<5%" (good condition), "5-
15%' (fair condition), or * > 15%' (poor condition) of the tested surface, with no indication of total damaged surface area.
1 Total number of units having available data that could be compared to all specified candidate standards.
9 Cell entries are(number of homes at or above at least one standard that have EBL children)/ number of homes containing EBL children),
followed by the corresponding percentage (in parentheses)
* Cell entries are (number of homes not at or above at least one standard that do not have EBL childrenl/ltotal number of homes not
containing EBL children), followed by the corresponding percentage (in parentheses).
5 Cell entries are (number of homes at or above at least one standard that have EBL childrenl/ltotal number of homes at or above at least
one standard), followed by the corresponding percentage (in parentheses).
8 Cell entries are (number of homes not at or above at least one standard that do not have EBL childrenl/ltotal number of homes not at or
above any standard), followed by the corresponding percentage (in parentheses).
DRAFT ~ DO NOT CITE OR QUOTE
256
August 28, 2000
-------
1 Table 6-10. Results of Performance Characteristics Analysis Performed on Data for
2 Housing Units in the Rochester Lead-in-Dust Study, for Specified Sets of
3 Candidate Standards for Dust-Lead Loadings and Observed Amount of
4 Damaged Lead-Based Paint on a Tested Surface
5 EBL = elevated blood-lead level U10//g/dL); LBP = Lead-Based Paint
Set of Candidate Standards
Floor Dust-
Lead
Loading
(pg/ft2)
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
Window
Sill Dust-
Lead
Loading
Uig/ft2)
-
-
..
-
—
—
-
250
250
250
250
250
250
250
125
125
125
125
125
125
125
-
..
—
—
..
--
Max.
Amt. Of
Damaged
LBP on a
'Tested
Surface
(ft)1
-
—
-
-
-
—
-
--
-
-
-
--
--
-
--
-
—
-
-
--
-
>15%
>15%
>15%
>15%
>15%
>15%
# Units
At Or
Above
At Least
One
Standard
/ Total #
Units2
9/197
19/197
31/197
58/197
84/197
150/197
180/197
71/189
75/189
80/189
93/189
106/189
150/189
176/189
116/189
118/189
122/189
128/189
134/189
159/189
180/189
84/197
88/197
94/197
104/197
115/197
162/197
Performance Characteristics
Sensitivity
# (%) of Units
with EBL
Children That
Are At or Above
At Least One
Standard3
5/47 (10.6%)
9/47(19.1%)
16/47(34.0%)
26/47 (55.3%)
31/47 (66.0%)
44/47 (93.6%)
45/47 (95.7%)
25/45 (55.6%)
27/45 (60.0%)
30/45 (66.7%)
34/45 (75.6%)
36/45 (80.0%)
44/45 (97.8%)
44/45 (97.8%)
35/45 (77.8%)
36/45 (80.0%)
38/45 (84.4%)
39/45 (86.7%)
40/45 (88.9%)
45/45 (100%)
45/45 (100%)
27/47 (57.4%)
28/47 (59.6%)
31/47 (66.0%)
33/47 (70.2%)
36/47 (76.6%)
44/47 (93.6%)
Specificity
# (%) of Units
with No EBL
Children That Are
At or Above No
Standard4
146/150(97.3%)
140/150(93.3%)
135/150(90.0%)
118/150(78.7%)
97/150(64.7%)
44/150(29.3%)
15/150(10.0%)
98/144(68.1%)
96/144(66.7%)
94/144 (65.3%)
85/144(59.0%)
74/144(51.4%)
38/144(26.4%)
12/144(8.3%)
63/144(43.8%)
62/144(43.1%)
60/144(41.7%)
55/144(38.2%)
50/144(34.7%)
30/144(20.8%)
9/144 (6.3%)
93/150(62.0%)
90/150(60.0%)
87/150(58.0%)
79/150(52.7%)
71/150(47.3%)
32/150(21.3%)
PPV
# (%) of Units
At or Above At
Least One
Standard That
Have EBL
Children1
5/9 (55.6%)
9/19(47.4%)
16/31 (51.6%)
26/58 (44.8%)
31/84(36.9%)
44/150(29.3%)
45/180(25.0%)
25/71 (35.2%)
27/75 (36 0%)
30/80 (37.5%)
34/93 (36.6%)
36/106(34.0%)
44/150(29.3%)
44/176(25.0%)
35/116(30.2%)
36/118(30.5%)
38/122(31.1%)
39/128(30.5%)
40/134(29.9%)
45/159(28.3%)
45/180(25.0%)
27/84(32.1%)
28/88(31.8%)
31/94(33.0%)
33/104(31.7%)
36/115(31.3%)
44/162(27.2%)
NPV
# (%) of Units At
or Above No
Standard That Do
Not Have EBL
Children*
146/188 (77.7%)
140/178(78.7%)
135/166(81.3%)
118/139(84.9%)
97/113(85.8%)
44/47 (93.6%)
15/17(88.2%)
98/118(83.1%)
96/1 14 (84.2%)
94/109 (86.2%)
85/96 (88.5%)
74/83 (89.2%)
38/39 (97.4%)
12/13(92.3%)
63/73 (86.3%)
62/71 (87.3%)
60/67 (89.6%)
55/61 (90.2%)
50/55 (90.9%)
30/30 (100%)
9/9 (100%)
93/113(82.3%)
90/109 (82.6%)
87/103(84.5%)
79/93 (84.9%)
71/82 (86.6%)
32/35(91.4%)
DRAFT -- DO NOT CITE OR QUOTE
257
August 28, 2000
-------
Table 6-10. (cont.)
Set of Candidate Standards
Uncarpeted
Floor Dust-
Lead
Loading
(f/g/ft1)
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
100
50
40
25
20
10
5
Window
Sill Dust-
Lead
Loading
(pg/ft2)
-
-
-
--
--
-
-
-
250
250
250
250
250
250
250
125
125
125
125
125
125
125
250
250
250
250
250
250
250
Max.
Amt. Of
Damaged
LBP on a
Tested
Surface
(%)'
>15%
>5%
>5%
>5%
>5%
>5%
>5%
>5%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>15%
>5%
>5%
>5%
>5%
>5%
>5%
>5%
# Units
At Or
Above
At Least
One
Standard
/Total*
Units1
183/197
146/197
147/197
149/197
150/197
155/197
181/197
189/197
107/189
109/189
114/189
119/189
126/189
160/189
178/189
135/189
137/189
141/189
144/189
147/189
167/189
181/189
147/189
148/189
150/189
151/189
156/189
175/189
182/189
Performance Characteristics
Sensitivity
# (%) of Units
with EBL
Children That
Are At or Above
At Least One
Standard3
45/47 (95.7%)
39/47 (83.0%)
40/47(85.1%)
41/47 (87.2%)
42/47 (89.4%)
44/47 (93.6%)
47/47 (100%)
47/47 (100%)
32/45(71.1%)
33/45 (73.3%)
36/45 (80.0%)
37/45 (82.2%)
39/45 (86.7%)
44/45 (97.8%)
44/45 (97.8%)
37/45 (82.2%)
38/45 (84.4%)
40/45 (88.9%)
41/45(91.1%)
42/45 (93.3%)
45/45 (100%)
45/45 (100%)
38/45 (84.4%)
39/45 (86.7%)
40/45 (88.9%)
41/45(91.1%)
43/45 (95.6%)
45/45 (100%)
45/45 (100%)
Specificity
# (%) of Units
with No EBL
Children That Are
At or Above No
Standard*
12/150(8.0%)
43/150(28.7%)
43/150(28.7%)
42/150(28.0%)
42/150(28.0%)
39/150(26.0%)
16/150(10.7%)
8/150(5.3%)
69/144(47.9%)
68/144 (47.2%)
66/144 (45.8%)
62/144(43.1%)
57/144 (39 6%)
28/144(19.4%)
10/144(6.9%)
46/144 (31.9%)
45/144(31.3%)
43/144(29.9%)
41/144(28.5%)
39/144(27.1%)
22/144(15.3%)
8/144 (5.6%)
35/144 (24.3%)
35/144(24.3%)
34/144(23.6%)
34/144(23.6%)
31/144(21.5%)
14/144(9.7%)
7/144(4.9%)
PPV
# (%) of Units
At or Above At
Least One
Standard That
Have EBL
Children1
45/183(24.6%)
39/146 (26.7%)
40/147 (27.2%)
41/149 (27.5%)
42/150(28.0%)
44/155(28.4%)
47/181 (26.0%)
47/189(24.9%)
32/107 (29.9%)
33/109 (30.3%)
36/114(31.6%)
37/119(31.1%)
39/126(31.0%)
44/160(27.5%)
44/178(24.7%)
37/135(27.4%)
38/137(27.7%)
40/141 (28.4%)
41/144 (28.5%)
42/147 (28.6%)
45/167 (26.9%)
45/181 (24.9%)
38/147 (25.9%)
39/148 (26.4%)
40/150(26.7%)
41/151 (27.2%)
43/156(27.6%)
45/175(25.7%)
45/182(24.7%)
NPV
# (%) of Units At
or Above No
Standard That Do
Not Have EBL
Children*
12/14(85.7%)
43/51 (84.3%)
43/50 (86.0%)
42/48 (87.5%)
42/47 (89.4%)
39/42 (92.9%)
16/16(100%)
8/8 (100%)
69/82(84.1%)
68/80 (85.0%)
66/75 (88.0%)
62/70 (88.6%)
57/63 (90.5%)
28/29 (96 6%)
10/11 (90.9%)
46/54 (85.2%)
45/52 (86.5%)
43/48 (89.6%)
41/45(91.1%)
39/42 (92.9%)
22/22 (100%)
8/8 (100%)
35/42 (83.3%)
35/41 (85.4%)
34/39 (87.2%)
34/38 (89.5%)
31/33 (93.9%)
14/14(100%)
7/7 (100%)
DRAFT - DO NOT CITE OR QUOTE
258
August 28, 2000
-------
Table 6-10. (cont.)
Set of Candidate Standards
Uncarpeted
Floor Dust-
Lead
Loading
Oig/ft2)
100
50
40
25
20
10
5
Window
Sill Dust-
Lead
Loading
(pg/ft*)
125
125
125
125
125
125
125
Max.
Amt. Of
Damaged
LBP on a
Tested
Surface
<%)'
>5%
>5%
>5%
>5%
>5%
>5%
>5%
# Units
At Or
Above
At Least
One
Standard
/ Total #
Units2
156/189
157/189
159/189
160/189
163/189
176/189
183/189
Performance Characteristics
Sensitivity
# (%) of Units
with EBL
Children That
Are At or Above
At Least One
Standard3
40/45 (88.9%)
41/45(91.1%)
42/45 (93.3%)
43/45 (95.6%)
44/45 (97.8%)
45/45 (100%)
45/45 (100%)
Specificity
# (%) of Units
with No EBL
Children That Are
At or Above No
Standard*
28/144(19.4%)
28/144(19.4%)
27/144(18.8%)
27/144(18.8%)
25/144(17.4%)
13/144(9.0%)
6/144(4.2%)
ppy
# (%) of Units
At or Above At
Least One
Standard That
Have EBL
Children*
40/156(25.6%)
41/157(26.1%)
42/159(26.4%)
43/160(26.9%)
44/163(27.0%)
45/176(25.6%)
45/183(24.6%)
NPV
» (%) of Units At
or Above No
Standard That Do
Not Have EBL
Children*
28/33 (84.8%)
28/32 (87.5%)
27/30 (90.0%)
27/29(93.1%)
25/26 (96.2%)
13/13(100%)
6/6 (100%)
1
2
3
4
5
6
7
' In the Rochester study, each measurement of lead in paint had the amount of damaged pemt specified as *<5%* (good condition), "5-
15%" (fair condition), or '> 15%' (poor condition) of the tested surface, with no indication of total damaged surface area.
1 Total number of units having available data that could be compared to all specified cendidate standards.
3 Cell entries arelnumber of homes at or above at least one standard that have EBL children)/ number of homes containing EBL children),
followed by the corresponding percentage (in parentheses).
4 Cell entries ere (number of homes not at or above at least one standard that do not have EBL childrenl/ltotal number of homes not
containing EBL children), followed by the corresponding percentage (in parentheses).
5 Cell entries are (number of homes at or above at least one standard that have EBL childrenl/ltotal number of homes at or above at least
one standard), followed by the corresponding percentage (in parentheses)
• Cell entries are (number of homes not at or above at least one standard that do not have EBL childrenl/ltotal number of homes not at or
above any standard), followed by the corresponding percentage (in parentheses)
DRAFT •- DO NOT CITE OR QUOTE
259
August 28, 2000
-------
1 6.2 INVESTIGATING INCIDENCE OF ELEVATED BLOOD-LEAD
2 CONCENTRATION IN HOUSING UNITS MEETING ALL
3 EXAMPLE OPTIONS FOR STANDARDS
4 An alternative to the performance characteristics analysis approach (Section 6.1) to
5 evaluating a set of candidate standards is to use statistical modeling techniques to predict a
6 distribution of blood-lead concentration as a function of environmental-lead levels found in
7 homes which do not exceed any of the candidate standards, then estimate the percentage of
8 children residing in these homes that are expected to have elevated blood-lead levels (i.e., at or
9 above 10 ug/dL). It is desired to select a set of candidate standards so that the likelihood of
10 children with elevated blood-lead concentration residing in homes that do not exceed any of the
11 candidate standards would be very low. This section presents a modeling approach to estimate
12 this likelihood, using the alternative Rochester multimedia model presented in Section 4.2 of this
13 report ("Model A" in Table 4-1), and applies this approach to data from the Rochester study.
14 Recall from Section 4.2 that the reason for developing the alternative Rochester
15 multimedia model was to have the risk estimates from model-based analyses be more comparable
16 to the results of the performance characteristics analysis presented in the §403 proposed rule
17 (Section 6.1.3) and the results of the follow-up performance characteristics analyses (Section
18 6.1.4). In particular, both the performance characteristics analysis and the model-based approach
19 involving the alternative Rochester multimedia model use the following types of data as input
20 when characterizing risk:
21 • household average (wipe) dust-lead loading from uncarpeted floors
22 • household average (wipe) dust-lead loading from window sills
23 • yard-wide average soil-lead concentration
24 • the larger of the following two percentages: % of interior tested surfaces that
25 contain deteriorated lead-based paint (LBP), and % of exterior tested surfaces that
26 contain deteriorated LBP
27 In the model-based analysis approach presented below, the candidate standards were used to
28 identify a subset of homes in the Rochester study that were below all of the candidate standards,
29 calculate the average (across homes) of the above three measures of lead levels in dust and soil,
30 and fit the multimedia model to these average lead levels in order to predict a distribution of
31 blood-lead concentrations for children residing in these homes. For simplicity, this analysis
32 assumes that the homes do not contain deteriorated lead-based paint. Because the slope estimate
33 for the paint variable in the alternative Rochester multimedia model is nearly zero (Table 4-1 of
34 Section 4.2), making the assumption that no deteriorated lead-based paint exists in these homes
35 should have a very minor impact on the resulting risk estimates.
36 6.2.1 The Model-Based Approach
37 This model-based approach had the following four steps:
DRAFT - DO NOT CITE OR QUOTE 260 August 28,2000
-------
1 1. For a given set of candidate standards for floor dust-lead loading, window sill
2 dust-lead loading, and soil-lead concentration, identify those homes in the
3 Rochester study that exceed none of the candidate standards in this set.
4 2. For each of the following three household measures, calculate the average across
5 the homes identified in step #1: the household average floor dust-lead loadings,
6 household average window sill dust-lead loading and for yard-wide average soil-
7 lead concentration. These three averages are assumed to represent lead levels in
8 housing represented by the Rochester study homes in step #1 (i.e., homes not
9 exceeding any of the candidate dust and soil standards).
10 3. Use the three averages calculated in step #2 as input to the alternative Rochester
11 multimedia model from Section 4.2 (assuming no deteriorated lead-based paint
12 exists in the units).
13 4. Assume that log-transformed blood-lead concentration for children residing in the
14 homes identified in step #1 is normally distributed with mean equal to the
15 predicted log-transformed blood-lead concentration that is output from the model
16 fitting in step #3, and standard deviation equal to ln( 1.6). (Recall that this
17 assumption on variability was made throughout the §403 risk analysis.) Using
18 normal distribution theory, determine the percentage of children represented by
19 this blood-lead distribution that have log-transformed blood-lead concentration or
20 above log(lO), or equivalently, that have blood-lead concentration at or above 10
21 ug/dL.
22 6.2.2 Example of Applying the Model-Based Approach
23 As an example of how the approach in Section 6.2.1 is applied to data from the Rochester
24 study, consider the following combination of candidate dust-lead and soil-lead standards:
25 • (uncarpeted) floor dust-lead loading = 40 ug/ft2
26 • window sill dust-lead loading = 250 ug/ft2
27 • yard-wide soil-lead concentration = 400 ug/g.
28 According to the performance characteristics analyses documented in Table 6-8 of Section 6.1
29 (i.e., the row of Table 6-8 corresponding to the above three candidate standards), 39 of the 184
30 Rochester study homes having dust-lead and soil-lead data do not exceed any of these three
31 candidate standards. Across these 39 homes, the following averages are calculated from the
32 Rochester study data:
33 • household average (uncarpeted) floor dust-lead loading: 12.7 ug/ft2
34 • household average window sill dust-lead loading: 87.0 ug/ft2
35 • yard-wide average soil-lead concentration: 125.3 ug/g
DRAFT-DO NOT CITE OR QUOTE 261 August 28,2000
-------
1 When fitting the alternative Rochester multimedia model to these three averages (assuming no
2 deteriorated lead-based paint), the model predicts a geometric mean blood-lead concentration of
3 4.68 ug/dL. If the standard deviation of log-transformed data is assumed to be 1.6 and normal
4 distribution theory is applied as described above, the estimated percentage of children with PbB
5 at or above 10 ug/dL is 5.3%.
6 For this same group of 39 homes, two (5.1%) contain children with elevated blood-lead
7 concentration (Table 6-8). Therefore, the estimated risk generated by the performance
8 characteristics analysis is 5.1 %. This agrees closely with the model-based analysis results in this
9 example.
10 Because this analysis focuses on a subset of homes whose lead levels fall below all
11 specified candidate standards, the number of homes in this subset can be small in instances where
12 at least one of the candidate standards is set at the low end of the distribution of lead levels (i.e.,
13 most homes have data that fall above the candidate standard). Therefore, as the set of candidate
14 standards becomes more stringent, and as the size of the sample from which the environmental-
15 lead data originate becomes smaller as a result, the variability associated with the estimated risk
16 increases. Furthermore, as the set of candidate standards becomes less stringent (i.e., as the
17 standards increase), the group of homes not exceeding any of the candidate standards is more
18 likely to remain the same, and as a result, the estimated risk eventually achieves a plateau. This
19 occurs in the above example, as increasing the candidate floor dust-lead loading standard above
20 40 ug/ft2 does little, if any, to increase the estimated risk beyond 5.3% under this approach and
21 under the given set of data, assuming the candidate standards for the other media (window sill
22 dust, soil) remain fixed.
23 The Rochester study data were used in this analysis as the alternative multimedia model
24 was fitted based on the Rochester data. If data from other studies were used instead, it would be
25 necessary to verify that the model parameter estimates adequately reflect the underlying
26 variability in these data in the same manner that they reflect variability in the Rochester study
27 data.
28 While the approach presented in this section is relatively easy to implement, it could be
29 modified even further in an attempt to achieve more accurate risk estimates. Such a modification
30 could reduce the level of simplicity associated with applying the approach. For example, rather
31 than calculate average environmental-lead levels across all homes and fit the model once to these
32 averages, a simulation approach could be applied in an attempt to more accurately represent the
33 entire distribution of environmental-lead levels in these homes and the resulting blood-lead
34 distribution associated with exposure across the entire distribution of environmental-lead levels.
DRAFT - DO NOT CITE OR QUOTE 262 August 28,2000
-------
1 6.3 REVIEW OF PUBLISHED INFORMATION ON POST-INTERVENTION
2 DUST-LEAD LOADINGS
3 This section summarizes published information on lead loadings (amount of lead per unit
4 surface area) in dust samples collected by wipe techniques, as reported by earlier lead
5 intervention studies. This information is used to evaluate assumptions made on post-intervention
6 dust-lead loadings (40 ug/ft2 for floors, 100 ug/ft2 for window sills) within the §403 risk analysis.
7 Details to supplement the summaries in this section are presented in Appendix H.
8 The following seven studies have been identified in which some type of paint or dust
9 intervention was performed, dust samples were collected using wipes or some other technique
10 (e.g., BRM vacuum) whose results could be converted to wipe-equivalent dust-lead loadings, and
11 post-intervention dust-lead loadings on floors and/or window sills were reported (references for
12 these studies are included in Appendix H):
13 • Baltimore Experimental Paint Abatement Studies
14 • Baltimore Follow-up Paint Abatement Study
15 • Baltimore Repair & Maintenance (R&M) Study
16 • Boston Interim Dust Intervention Study
17 • HUD Grantees Evaluation (data available through September 1997)
18 • Denver Comprehensive Abatement Performance (CAP) Study
19 • Jersey City Children's Lead Exposure and Reduction (CLEAR) Study
20 These studies employed a variety of intervention strategies, including single or repeated dust
21 cleanings and interim control or complete abatement of lead-based paint. Dust-lead loadings
22 were measured at varying intervals following intervention. Post-intervention dust-lead loadings
23 were summarized for 19 groups of housing units across these seven studies. These study groups
24 are defined in Appendix H.
25 For both floors and window sills, geometric mean and median dust-lead loadings were
26 observed below the post-intervention assumptions established in the §403 risk analysis in a
27 majority of the study groups. However, this does not preclude results for individual housing
28 units from being above the assumed levels. Furthermore, the extent to which results for these
29 studies represent the nation's housing stock has not been determined. Results are now presented
30 separately for floors and window sills (with more detailed presentations found in Appendix H).
31 6.3.1 Post-Intervention Floor Dust-Lead Loadings
32 Summaries of post-intervention floor (wipe) dust-lead loadings are presented in Table
33 6-11 according to housing group within each study. According to Table 6-11, all but two of the
34 19 study groups reported geometric mean or median floor dust-lead loadings at or below 41
35 ug/ft2 from 6 months to 6 years post-intervention. The other two study groups were from the
36 Baltimore Experimental Paint Abatement Study, where pre-intervention geometric mean dust-
37 lead loadings were much greater (556 ug/ft2 and 1261 ug/ft2) than any other study group (at most
DRAFT - DO NOT CITE OR QUOTE 263 August 28,2000
-------
1
2
Table 6-11. Summaries of Pre- and Post-Intervention Floor Wipe Dust-Lead Loadings for
Housing Groups Within Seven Studies
Study
Baltimore
Experimental Paint
Abatement
Studies2
Baltimore Follow-
up Paint
Abatement Study2
Baltimore R&M
Study3
Boston Interim
Dust Intervention
Study2
HUD Grantees4
Denver CAP Study5
Jersey City CLEAR
Study
Study
Group
Study 1
Study 2
1 2-Month Follow-up
19-Month Follow-up
Previously-Abated Units
Units Slated for R&M
Intervention
Automatic Intervention
Randomized Intervention
All Grantees
Baltimore
Boston
Massachusetts
Milwaukee
Minnesota
Rhode Island
Vermont
Wisconsin
Abated Units
Intervention Group
Pre-lntervention Floor
Dust-Lead Loadings1
(Wl/ft2)
1261
556
NA
NA
45.6
58.6
33.2
37.3
19
41
24
24
14
18
26
28
9
NA
22
Post-Intervention
Floor Dust-Lead Loadings1
Time Following
Intervention
(Months)
6-9
1.5-3.5 Years
10-14
14-24
4 - 6 Years
24
6
6
12
12
12
12
12
12
12
12
12
2 Years
12
Summary Value
fog/ft2)
99
69
20
36
33.0
35.0
23.9
31.4
14
41
18
9
10
18
6
21
5
21.0
15
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
1 Values are geometric means except for the HUD Grantees studies, where values are medians. "NA" indicates not
available.
1 Results are adjusted to reflect total dust-lead loadings by exponentiating the "bioavailable" dust-lead loadings as reported in
the study to the 1.1416 power.
3 Results for the Baltimore R&M Study are converted from BRM dust-lead loadings to wipe-equivalent loadings.
* Data collected through September, 1997
9 Results for the Denver CAP study are converted from CAP cyclone dust-lead loadings to wipe-equivalent loadings.
27
DRAFT - DO NOT CITE OR QUOTE
264
August 28, 2000
-------
1 58.6 ug/ft2). Eleven study groups reported geometric mean or median floor dust-lead loadings at
2 or below 21 ug/ft2 at follow-up periods ranging from 12 months to 2 years. Of these 11 groups,
3 four of the HUD Grantees study groups reported median floor dust-lead loadings at or below 10
4 ug/ft2 at 12 months post-intervention. Median pre-intervention floor dust-lead loadings in these
5 four groups ranged from 9 to 26 fig/ft2.
6 In the HUD Grantees evaluation, seven of the eight largest grantees have median floor
7 dust-lead loadings at or below 21 ug/ft2 at 12 months post-intervention, compared to a median of
8 14 ug/ft2 across all grantees. Although pre-intervention floor dust-lead loadings were lower in
9 the HUD Grantees evaluation compared to other studies, these preliminary results suggest that
10 floor dust-lead loadings can be maintained at levels below 40 ug/ft2 for at least 12 months post-
11 intervention.
12 Results from the Denver CAP study, the Baltimore Follow-up Paint Abatement study, the
13 Baltimore R&M study, the Boston Interim Dust Intervention study, and the Jersey City CLEAR
14 study suggest that geometric mean floor dust-lead loadings of below 40 ug/ft2 can be observed
15 even beyond 12 months post-intervention and up to six years post-intervention, under the same
16 conditions experienced by the housing units in these studies.
17 6.3.2 Post-Intervention Window Sill Dust-Lead Loadings
18 Summaries of post-intervention window sill wipe dust-lead loadings are presented in
19 Table 6-12 according to housing group. Post-intervention geometric means or medians range
20 from 24 ug/ft2 to 958 ug/ft2, which are considerably higher than the summaries for floors.
21 Eleven study groups had geometric mean or median post-intervention window sill dust-lead
22 loadings below 100 ug/ft2,6 groups were at or below 51 ug/ft2, and 3 groups were at or below 41
23 ug/ft2.
24 All but one of the HUD Grantees study groups (the Milwaukee grantee) had median
25 window sill dust-lead loadings below 100 ug/ft2 at 12 months post-intervention. As the
26 intervention strategy for homes in the HUD Grantees evaluation frequently included partial or
27 complete window replacement, these results may not be representative of the outcomes of
28 interventions prompted by the §403 rule.
29 Geometric mean window sill dust-lead loadings were below 100 ug/ft2 for up to two years
30 post-intervention in the Baltimore Follow-up Paint Abatement study, Denver CAP study, and
31 Jersey City CLEAR study. However, in the Baltimore R&M study, Baltimore Experimental
32 Paint Abatement studies, and Boston Interim Dust Intervention study, geometric mean dust-lead
33 loadings remain above 100 ug/ft2 over time. In addition, the 19-month follow-up study group
34 within the Baltimore Follow-up Paint Abatement study and study group #2 of the Baltimore
35 Experimental Paint Abatement studies suggest that geometric mean dust-lead loadings can dip
36 below 100 ug/ft2 immediately after intervention, but then exceed this level after one year or so.
DRAFT--DO NOT CITE OR QUOTE 265 August 28.2000
-------
1
2
Table 6-12. Summaries of Pre- and Post-Intervention Window Sill Wipe Dust-Lead
Loadings for Housing Groups Within Seven Studies
Study
Baltimore
Experimental Paint
Abatement
Studies2
Baltimore Follow-
up Paint
Abatement Study2
Baltimore R&M
Study3
Boston Interim
Dust Intervention
Study2
HUD Grantees4
Denver CAP
Study5
Jersey City CLEAR
Study
Study
Group
Study 1
Study 2
1 2-Month Follow-up
1 9-Month Follow-up
Previously-Abated Units
Units Slated for R&M
Intervention
Automatic Intervention
Randomized Intervention
All Grantees
Baltimore
Boston
Massachusetts
Milwaukee
Minnesota
Rhode Island
Vermont
Wisconsin
Abated Units
Intervention Group
Pre-lntervention Sill
Dust-Lead Loadings1
0/g/ft2)
15215
2784
NA
NA
163.5
778.4
787
205
258
1191
174
328
264
266
314
147
150
NA
75
Post-Intervention
Sill Dust-Lead Loadings1
Time Following
Intervention
6-9
1 .5 - 3.5 Years
10-14
14-24
4 - 6 Years
24
6
6
12
12
12
12
12
12
12
12
12
2 Years
12
Summary Value
(//g/ft2)
958
199
41
147
97.6
204.9
210
110
90
68
49
50
217
77
85
40
51
66.4
24
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
1 Values are geometric means except for the HUD Grantees studies, where values are medians. "NA" indicates not
available.
2 Results are adjusted to reflect total dust-lead loadings by exponentiating the "bioavailable" dust-lead loadings as reported in
the study to the 1.1416 power.
3 Results for the Baltimore R&M Study are converted from BRM dust-lead loadings to wipe-equivalent loadings.
4 Data collected through September, 1997
5 Results for the Denver CAP study are converted from CAP cyclone dust-lead loadings to wipe-equivalent loadings.
DRAFT - DO NOT CITE OR QUOTE
266
August 28, 2000
-------
1 6.4 SENSITIVITY AND UNCERTAINTY ANALYSES FOR RISK
2 MANAGEMENT ANALYSES
3 The following subsections present the results of additional sensitivity and uncertainty
4 analyses performed to gauge the level of uncertainty in the post-§403 risk estimates (and the
5 associated decline from baseline estimates) associated with methodological assumptions. These
6 results should be considered with those presented in the sensitivity and uncertainty analyses in
7 Section 6.4 of the §403 risk analysis report to characterize overall uncertainty associated with the
8 methods and assumptions taken in the risk management.
9 6.4.1 Considering How Baseline Environmental-Lead Levels May
10 Have Changed Since the HUD National Survey
11 Section 5.1.4 of this report addressed the sensitivity of the pre-§403 model-based blood-
12 lead distribution and the resulting health effects and blood-lead concentration endpoint estimates
13 under the IEUBK and empirical models under different assumptions on how the national
14 distribution of baseline environmental-lead levels as estimated using HUD National Survey data
15 may have changed since the time of the survey (1989-1990). The same five sets of adjustments
16 (i.e., percentage changes) made to the average baseline dust-lead loadings, dust-lead
17 concentrations, and soil-lead concentrations for each housing unit in the HUD National Survey
18 were considered in this sensitivity analysis to observe the impact on post-§403 risk estimates
19 under the following set of example options for standards:
20 • Average floor dust-lead loading = 100 ug/ft2
21 • Average window sill dust-lead loading = 500 ug/ft2
22 • Average soil-lead concentration = 2,000 ug/g
23 • Amount of deteriorated lead-based paint requiring paint maintenance = 5 ft2
24 • Amount of deteriorated lead-based paint requiring paint abatement = 20 ft2
25 This set of options was the primary set considered in the sensitivity analyses within Section 6.4
26 of the §403 risk analysis report.
27 Table 6-13 presents the post-§403 estimates for the health effect and blood-lead
28 concentration endpoints under both the IEUBK and empirical models, for each of the five sets of
29 adjustments to the post-§403 environmental-lead levels in housing units within the HUD
30 National Survey and under the above assumption on example standards. Also included in this
31 table are the percentage of homes exceeding the various example standards, which will be lower
32 than in the §403 risk analysis when declines in the appropriate environmental-lead levels are
33 considered and higher when increases are considered. The table also lists the baseline risk
34 estimates for comparison purposes.
35 Effect on risk analysis: Under the five sets of assumptions involving lower assumed
36 baseline environmental-lead levels, the percentage of houses that exceed at least one of the
37 example standards declined by at most about three percentage points (from 21.8% to 18.7%;
DRAFT -- DO NOT CITE OR QUOTE 267 August 28.2000
-------
1
2
3
4
5
6
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
Table 6-13. Sensitivity Analysis on How Changes in Household Average Baseline Dust-
Lead Loadings/Concentrations and Soil-Lead Concentration Impact Post-
§403 Estimates of Health Effect and Blood-Lead Concentration Endpoints for
Children Aged 1-2 Years Under a Specified Set of Example Standards1
Assumed Percentage Change in Average Dust-Lead Loadings and Concentrations
(Both Floor and Window Sill) and in Yard-wide Average Soil-Lead Concentration
Dust:
Soil:
No
change
No
change
20%
decrease
20%
decrease
50%
decrease
50%
decrease
50%
decrease
No
change
No
change
50%
decrease
25%
increase
25%
increase
Baseline
Estimate
(from
Table 5-1
of the
§403 risk
analysis
report)
Percentage of Homes Exceeding Example Standards/Triggers
Floor Dust
Window Sill Dust
Soil
Interior Paint Maintenance
Exterior Paint
Maintenance
Interior Paint Abatement
Exterior Paint Abatement
Any Standard/Trigger
4.04
12.5
2.49
2.92
3.49
2.43
5.77
21.8
2.34
10.8
1.52
2.92
3.49
2.43
5.77
20.6
0.694
9.10
0.746
2.92
3.49
2.43
5.77
18.7
0.694
9.10
2.49
2.92
3.49
2.43
5.77
18.9
4.04
12.5
0.746
2.92
3.49
2.43
5.77
21.6
5.68
14.3
3.27
2.92
3.49
2.43
5.77
24.1
Predicted Health Effect And Blood-Lead Concentration Endpoints (Based on Empirical Model)
PbB 220 {%)
PbB2lO(%)
IQ < 70 (%)
IQ decrement *1 (%)
IQ decrement a 2 (%)
IQ decrement 23 (%)
Avg. IQ decrement
0.406
4.70
0.110
36.3
9.30
2.93
1.00
0.429
4.85
0.111
36.7
9.53
3.04
1.01
0.469
5.10
0.112
37.3
9.90
3.21
1.03
0.445
4.95
0.111
36.9
9.69
3.11
1.02
0.427
4.84
0.111
36.7
9.51
3.03
1.01
0.378
4.52
0.110
35.9
9.02
2.80
0.995
0.588
5.75
0.115
38.5
10.8
3.70
1.06
Predicted Health Effect And Blood-Lead Concentration Endpoints (Based on IEUBK Model)
PbB 220 (%)
PbB 2 10(%)
IQ < 70 (%)
IQ decrement 2! (%)
IQ decrement 22 (%)
IQ decrement 28 (%)
Avg. IQ decrement
0.0539
1.66
0.0984
28.3
4.31
0.858
0.848
0.117
2.48
0.102
31.0
5.77
1.37
0.894
0.166
2.98
0.104
32.7
6.65
1.71
0.924
0.121
2.55
0.102
31.6
5.94
1.42
0.904
0.0681
1.86
0.0992
28.8
4.67
0.983
0.857
0.0542
1.64
0.0982
27.7
4.22
0.847
0.839
0.588
5.75
0.115
38.5
10.8
3.70
1.06
1 Example dust and soil standards were set at: 100/sg/ft2 for floor dust-lead loading, 500pg/ft2 for window sill
dust-lead loading, and 2,000 fjglg for soil-lead concentration. Paint maintenance is performed if more than 5
ft2, but less than 20 ft2 of deteriorated lead-based paint exists. Paint abatement is performed if more than 20
ft2 of deteriorated lead-based paint exists.
DRAFT - DO NOT CITE OR QUOTE
268
August 28, 2000
-------
1 Table 6-13), or about three million homes. The assumption that baseline environmental-lead
2 levels are 25% higher than assumed in the §403 risk analysis results in an increase in the
3 percentage of homes exceeding at least one standard from 21.8% to 24.1 %, an increase of about
4 2.3 million homes (Table 6-13).
5 As would be expected, Table 6-13 shows that all assumptions on baseline environmental-
6 lead levels result in post-§403 estimates of the predicted health effect and blood-lead
7 concentration endpoints that are lower than baseline (the last column of the table). However, as
8 the assumed baseline environmental-lead levels become lower in magnitude, the predicted post-
9 §403 risks actually increase, converging to the baseline estimates. For example, as seen in Table
10 6-3, baseline lead levels that are 20% below what was assumed in the §403 risk analysis resulted
11 in an estimated percentage of children with blood-lead concentrations at or above 10 |jg/dL of
12 4.85%, compared to the §403 risk analysis estimate of 4.70%. When baseline lead levels are
13 50% below the §403 risk analysis estimates, the estimate of this percentage increases to 5.10%.
14 Such a finding appears counter-intuitive when first reviewing the table. However, the alternative
15 assumptions being considered in this sensitivity analysis are to baseline (i.e., pre-§403)
16 environmental-lead levels. As assumptions on these baseline levels move lower, fewer homes
17 are triggered by the §403 standards, and the post-§403 distribution of environmental-lead levels
18 becomes less removed from the baseline distribution. As a result, post-§403 estimates of
19 predicted health effects and blood-lead concentration are not as different from pre-§403
20 estimates. In contrast, as assumed baseline environmental-lead levels increase, more homes are
21 triggered by the §403 standards and, therefore, have their environmental-lead levels drop as a
22 result of interventions, and lower post-§403 risk estimates relative to baseline are observed.
23 As seen in Table 6-13, the effect that different assumptions on baseline environmental-
24 lead levels have on the risk estimates is considerably greater under the IEUBK model than the
25 empirical model. The percentage of children with blood-lead concentrations at or above 20
26 ug/dL more than triples under the IEUBK model approach when 50% declines in both dust-lead
27 and soil-lead levels were assumed (from 0.054% to 0.166%), compared to a 16% increase under
28 the empirical model (from 0.406% to 0.469%). Smaller percentage differences are observed for
29 the other endpoints for both models.
30 6.4.2 Impact on the Estimated Incidence of IQ Point Decrement
31 Assuming Certain Thresholds on the IQ/Blood-Lead Relationship
32 The sensitivity of baseline and pre-§403 model-based estimates of IQ decrements greater
33 than 1, 2, or 3, and of the average and standard deviation of the distribution of IQ point
34 decrements was addressed in Section 5.1.5 of this report for various assumptions of a non-zero
35 threshold of blood-lead concentration on the IQ/blood-lead relationship. The following
36 thresholds were considered: 1,2, 3,5, 8 and 10 ug/dL. In this section, post-§403 estimates of
37 these health effect endpoints are estimated (under the same set of options presented in Section
38 6.4.1, using both the IEUBK and empirical models) under these same alternative blood-lead
39 concentration thresholds. These estimates are presented in Table 6-14.
DRAFT - DO NOT CITE OR QUOTE 269 August 28,2000
-------
1 Table 6-14. Sensitivity Analysis on the Assumed Blood-Lead Concentration Threshold on
2 IQ Decrement and Its Impact on the Post-§403 Estimates of IQ Decrement
3 Endpoints for Children Aged 1-2 Years, Under a Specified Set of Example
4 Standards1
Assumed
Threshold
U/g/dL)
% of Children Aged 1-2 Years with a Specified IQ
Decrement Due to Lead Exposure2
IQ Decrement *
1
IQ Decrement *
2
IQ Decrement z
3
Average IQ
Decrement
(# points)3
Standard
Deviation of IQ
Decrement3
Baseline Estimates (Section 5.1.1 of §403 risk analysis report)
0
1
2
3
5
8
10
38.5
27.3
19.6
14.2
7.83
3.50
2.15
10.8
8.08
6.10
4.66
2.80
1.40
0.915
3.70
2.88
2.26
1.80
1.16
0.627
0.429
1.06
0.804
0.588
0.428
0.233
0.103
0.0638
0.895
0.891
0.860
0.802
0.666
0.494
0.408
Post- §40 3 Estimates Based on IEUBK Model-Generated PbB Distribution
0
1
2
3
5
8
10
28.3
17.1
10.4
6.48
2.65
0.790
0.380
4.31
2.78
1.82
1.21
0.566
0.199
0.105
0.858
0.589
0.410
0.289
0.149
0.0593
0.0335
0.848
0.594
0.379
0.234
0.0907
0.0250
0.0116
0.567
0.564
0.529
0.462
0.325
0.188
0.134
Post- 5403 Estimates Based on Empirical Model-Generated PbB Distribution
0
1
2
3
5
8
10
36.3
25.1
17.6
12.5
6.56
2.76
1.64
9.30
6.79
5.02
3.75
2.18
1.03
0.653
2.93
2.24
1.73
1.35
0.838
0.434
0.289
1.00
0.752
0.537
0.380
0.197
0.0812
0.0480
0.817
0.814
0.781
0.721
0.584
0.417
0.337
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
1 Example dust and soil standards were set at: 100//g/ft2 for floor dust-lead loading, 500//g/ft2 for window sill
dust-lead loading, and 2,000 //g/g for soil-lead concentration. Paint maintenance is performed if more than 5
ft2, but less than 20 ft2 of deteriorated lead-based paint exists. Paint abatement is performed if more than 20
ft2 of deteriorated lead-based paint exists.
2 A 0.257 IQ decrement is assumed for each 1.0 j/g/dL increase in PbB above the assumed threshold (see Section 4.4.1 of
the §403 risk analysis report). Thus, the following hold:
• PIIQ i 1| = PIPbB 2 (threshold + 3.9jug/dL)|
• PIIQ z 21 = PIPbB * (threshold + 7.8 j/g/dUl
• PIIQ z3] = PIPbB 2 (threshold + 11.7//g/dUl
3 Average and standard deviation of IQ decrement are calculated assuming no IQ decrement occurs below the assumed
threshold, and a 0.257 IQ decrement is assumed for each 1.0//g/dL increase in PbB above the threshold.
DRAFT - DO NOT CITE OR QUOTE
270
August 28, 2000
-------
1 Effect on risk analysis: As was also seen in Table 5-7 of this report, Table 6-14 shows
2 that the post-§403 risk estimates decrease as the assumed blood-lead concentration threshold
3 increases (i.e., smaller percentages of children experience IQ score decrements under larger
4 threshold assumptions). The IEUBK model is more sensitive than the empirical model to the
5 threshold level. For example, the probability of a child experiencing an IQ decrement of at least
6 1 point decreases by 63% under the IEUBK model (from 28.3% to 10.4%) when the threshold
7 increases from 0 to 2 ug/dL, compared to only a 52% decrease under the empirical model (from
8 36.3% to 17.6%). As the assumed threshold increases, the likelihood of experiencing an IQ
9 decrement of at least 1 point as a result of lead exposure decreases to very low values under both
10 models, and the average IQ score decrement in the population declines to small fractions of
11 points.
12 6.4.3 Considering Alternative Assumptions on Post-Intervention
13 Dust-Lead Loadings
14 In the risk management portion (Chapter 6) of the §403 risk analysis report, it was
15 necessary to make assumptions on predicted post-intervention lead levels when characterizing
16 the blood-lead concentration and health effect endpoints in a post-§403 environment. These
17 assumptions were documented in Table 6-2 of the §403 risk analysis report. Among these
18 assumptions were that dust cleaning activities impacted interior dust-lead loadings in the
19 following way:
20 • Post-intervention household average floor (wipe) dust-lead loadings equaled the
21 minimum of 40 ug/ft2 and the pre-intervention value.
22 • Post-intervention household average window sill (wipe) dust-lead loadings
23 equaled the minimum of 100 ug/ft2 and the pre-intervention value.
24 A dust cleaning was assumed to be included among the interventions performed when either the
25 floor-dust, window sill-dust, soil, or interior paint abatement standards were exceeded within a
26 home. These two assumptions on post-intervention dust-lead loadings were made within the
27 §403 risk analysis based on data reported in EPA's Comprehensive Abatement Performance
28 study and in the Baltimore Experimental Paint Abatement study (see Section 6.1.2 of the §403
29 risk analysis report and Section H2.0 of Appendix H of this report).
30 Tables 6-11 and 6-12 within Section 6.3 of this report presented additional information
31 on household average (wipe) dust-lead loading at pre- and post-intervention for floors and
32 window sills, respectively, from several recent lead intervention studies. This information, some
33 of which was received after the §403 risk analysis report was completed, suggests that it may be
34 common in some instances to observe household average post-intervention dust-lead loadings
35 below the assumptions made above, even from 12 months to six years post-intervention. These
36 findings prompted a sensitivity analysis to investigate how setting assumptions on post-
37 intervention household average dust-lead loadings to below the 40 ug/ft2 and 100 ug/ft2
38 specifications would impact the outcome of the risk management analyses.
DRAFT - DO NOT CITE OR QUOTE 271 August 28,2000
-------
1 In this sensitivity analysis, two alternative assumptions on household average post-
2 intervention floor dust-lead loadings were made: 10 Mg/ft2 and 25 Mg/ft2- As the geometric mean
3 (12-month) post-intervention floor dust-lead loading in the HUD Grantees evaluation was 14
4 Mg/ft2 (Table 6-8) and was even lower for certain grantees, an alternative of 10 ug/ft2 was
5 selected. The alternative of 25 jig/ft2 for floors was selected as it fell halfway between the
6 assumptions of 10 and 40 fig/ft2 and was within the range of expected variability in the
7 summaries for several of the studies in Section 6.3.1.
8 Similarly, two alternative assumptions on household average post-intervention window
9 sill dust-lead loadings were made: 50 Mg/ft2 and 75 ug/ft2. Evidence from Table 6-12 indicates
10 that average window sill dust-lead loadings following intervention could approach SO ug/ft2 in
11 some instances, especially when floor dust-lead loadings are low. The alternative of 75 ug/ft2
12 was selected as it fell halfway between the assumptions of 50 and 100 Mg/ft2, and it was similar
13 to the average levels observed by grantees within the HUD Grantees evaluation (although the
14 HUD Grantees evaluation included window replacement, which was not among the assumed
15 interventions in the §403 risk analysis).
16 In the sensitivity analysis, if a given household's pre-intervention average floor dust-lead
17 loading fell below the given post-intervention assumption, its post-intervention household
18 average floor dust-lead loading was assumed to be equal to its pre-intervention average (as was
19 done in Chapter 6 of the §403 risk analysis report). Second, this sensitivity analysis considers
20 predictions made only by the empirical model, as the IEUBK model does not accept dust-lead
21 loading as input. Finally, the assumptions made in determining post-intervention soil-lead
22 concentrations (150 ug/g following soil removal) and amount of deteriorated lead-based paint
23 (none is present following paint intervention) remained the same as specified in Table 6-2 of the
24 §403 risk analysis report.
25 Table 6-15 presents the estimated post-§403 health effect and blood-lead concentration
26 endpoints associated with the set of example options for standards specified in Section 6.4.1
27 above, for the alternative assumptions on post-intervention floor and window sill dust-lead
28 loadings specified above. Note that each alternative assumption is evaluated on its own (i.e., it is
29 the only change from the §403 risk analysis assumptions). In addition, considering the high
30 correlation in dust-lead loadings between floors and window sills, the two lower alternatives (10
31 Mg/ft2 for floors and 50 Mg/ft2 for window sills) and the two higher alternatives (25 Mg/ft2 for
32 floors and 75 Mg/ft2 for window sills) are evaluated together. For comparison purposes, post-
33 intervention estimates under the §403 risk analysis (i.e., assuming 40 Mg/ft2 for floors and 100
34 Mg/ft2 f°r window sills) and the estimates generated under baseline (pre-§403) conditions (both
35 presented in Table 6-7 of the §403 risk analysis report) are also included in Table 6-15.
36 Effect on risk analysis. Relative to the results reported in the §403 risk analysis report
37 (column 2 of Table 6-15), the greatest deviation occurs with the most substantial change in the
38 assumptions, i.e., the assumptions of 10 Mg/ft2 for floors and 50 Mg/ft2 for window sills (column 7
39 of Table 6-15). Under this particular set of alternative assumptions, the percentage of the
40 nation's children aged 1-2 years that are anticipated to have blood-lead concentration at or above
DRAFT -- DO NOT CITE OR QUOTE 272 August 28,2000
-------
4
Q
nl
§
o
Table 6-15. Sensitivity Analysis on How Changing the Assumption on the Post-Intervention Household Average (Wipe)
Dust-Lead Loadings on Floors and Window Sills Impact Post-§403 Estimates (Based on the Empirical Model) of
the Health Effect and Blood-Lead Concentration Endpoints for Children Aged 1-2 Years Under a Specified Set
of Example Standards1
Health Effect And
Blood-Lead
Concentration
Endpoint
PbB 220 (%)
PbB*10(%)
IQ < 70 (%)
IQ decrement 2! (%)
IQ decrement *2 (%)
IQ decrement *3 (%)
Avg. IQ decrement
Predicted Estimates of the Endpoint (Based on the Empirical Model)
Assumed Post-Intervention Household Average Dust-Lead Loading for Floors and Window Sills2
Floors =
40/ig/ft2
Sills =
100//g/ft2
0.406
4.70
0.110
36.3
9.30
2.93
1.00
Floors =
10/ig/ft2
Sills =
100/ig/ft2
0.389
4.59
0.110
36.1
9.13
2.85
0.999
Floors =
25/ig/ft2
Sills =
100//g/ft2
0.401
4.67
0.110
36.3
9.25
2.90
1.00
Floors =
40//g/ft2
Sills =
50/ig/ft2
0.396
4.64
0.110
36.2
9.20
2.88
1.00
Floors =
40jig/ft2
Sills =
75j/g/ft2
0.402
4.68
0.110
36.3
9.26
2.91
1.00
Floors =
10//g/ft2
Sills =
50/ig/ft2
0.380
4.53
0.110
35.9
9.03
2.81
0.995
Floors =
25//g/ft2
Sills =
75//g/ft2
0.397
4.64
0.110
36.2
9.21
2.89
1.00
Baseline
Estimate
(From Table
5-1 of the
§403 Risk
Analysis
Report)
0.588
5.75
0.115
38.5
10.8
3.70
1.06
5
6
7
8
11
12
13
14
15
16
17
18
19
20
21
22
23
to
oo
8
§
1 Example dust and soil standards were set at 100//g/ft2 for floor dust-lead loading, 500 jug/ft2 for window sill dust-lead loading, and 2,000 //g/g for soil-
lead concentration. Paint maintenance is performed if more than 5 ft2, but less than 20 ft2 of deteriorated lead-based paint exists. Paint abatement is
performed if more than 20 ft2 of deteriorated lead-based paint exists. This analysis follows the same approach conducted in Section 6.3.4 of the §403
risk analysis report. Assumptions on post-intervention soil-lead concentrations and amounts of deteriorated lead-based paint are unchanged from those
specified in Table 6-2 of the §403 risk analysis report.
2 Within a housing unit, the assumed post-intervention average floor dust-lead loading is the minimum of its pre-intervention average and the value for
floors specified in the column heading. Similarly, the unit's assumed post-intervention average window sill dust-lead loading is the minimum of its pre-
intervention average and the value for sills specified in the column heading.
-------
1 10 ug/dL following interventions conducted in response to the §403 rule (given the example
2 standards specified in the footnote to this table) is reduced from 4.70% to 4.53% (a 3.7% decline,
3 equivalent to approximately 13,700 children12). The corresponding reduction in the percentage
4 of children with blood-lead concentration at or above 20 ug/dL is from 0.406% to 0.380% (a
5 6.3% decline, equivalent to approximately 2,000 children).
6 Under the assumptions of 25 Mg/ft2 for floors and 75 ug/ft2 for window sills (column 8 of
7 Table 6-15), the percentage of the nation's children aged 1-2 years that are anticipated to have
8 blood-lead concentration at or above 10 ug/dL is reduced from 4.70% to 4.64% (a 1.2% decline,
9 equivalent to approximately 4,800 children). The corresponding reduction in the percentage of
10 children with blood-lead concentration at or above 20 ug/dL is from 0.406% to 0.397% (a 2.3%
11 decline, equivalent to approximately 750 children).
12 Generally, even lower percentage declines occur for the IQ endpoints compared to the
13 blood-lead concentration endpoints. The exception occurs with the percentage of children with
14 IQ decline of at least 3 points, where a 4.2% decline from the §403 risk analysis assumptions was
15 observed under assumptions of 10 ug/ft2 for floors and 50 ug/ft2 for window sills.
16 This sensitivity analysis indicates that while more housing units may achieve reductions
17 in average dust-lead levels on floors and window sills following a dust cleaning if the assumed
18 post-intervention floor dust-lead loadings are lowered from those made in the §403 risk analysis,
19 the corresponding reduction in the estimated blood-lead concentration and health effect endpoints
20 appears to be modest, especially when compared to the reduction observed from pre- to post-
21 §403 conditions.
22 6.4.4 Characterizing the Post-Intervention Blood-Lead Distribution
23 Based on Relative Change from Baseline in the Geometric
24 Mean and the Probability of a Child's Blood-Lead
25 Concentration Exceeding 10 /sg/dL
26 As discussed in Section 4.3.1 above and in Appendix Fl of the §403 risk analysis report,
27 a "scaling algorithm" was used in the §403 risk analysis to characterize the distribution of blood-
28 lead concentration in the nation's children following interventions that would be performed as a
29 result of implementing the §403 rule (where the algorithm was applied under a specified set of
30 example options for the standards, using a specified blood-lead prediction model, and under
31 assumptions made on the changes in environmental-lead levels that result from the
32 interventions). This distribution is labeled the "post-§403" distribution. This approach
33 calculated the geometric mean (CM) and geometric standard deviation (GSD) of the post-§403
34 blood-lead distribution in the following manner:
12 Assuming that 7.96 million children aged 1-2 years reside in the U.S. housing stock (Table 3-35 of the
§403 nsk analysis report).
DRAFT -- DO NOT CITE OR QUOTE 274 August 28.2000
-------
= GMbasehne*(GMlnodel.basedpos|jW3/GMinodel.based pre-403) (0
*
(GSDmodeM>as(:d ^,^,3 / GSDn,^,.,^,, pre^,3) (2)
3 where the subscripts indicate the blood-lead distribution which either the GM or the GSD
4 represents. See Section 4.3.1 for additional information on this approach.
5 One comment received on the §403 risk analysis was that because the blood-lead
6 concentration endpoints utilized in the risk analysis were exceedance probabilities (i.e., the
7 likelihood of a child's blood-lead concentration exceeding a specified value), it was more
8 important to accurately characterize the right tail of the post-§403 distribution compared to the
9 remainder of the distribution, especially at blood-lead levels beyond 10 ug/dL. Therefore, a
10 variant of the scaling approach was considered that involved scaling the probability of a child's
1 1 blood-lead concentration exceeding 10 ug/dL rather than the GSD. If P10 was used to represent
12 this probability, then the alternative scaling algorithm would involve scaling the geometric mean
13 as in (1) above, but replacing (2) above with the following calculation:
14 P10poSIJ(03 =P10base|ine * (P10mode|.basedposl^03/P10mode|.basedprejM)3) (3)
15 The resulting value is the estimate of the probability of a child's blood-lead concentration
16 exceeding 10 ug/dL in a post- §403 environment. It is calculated by multiplying the probability
1 7 as calculated in the baseline distribution by the relative change, in the probability from the pre-
18 §403 to post-§403 environment as estimated from model-based blood-lead distributions. Then,
19 in order to calculate the other blood-lead concentration and health effect endpoints, the GSD of
20 the post-§403 distribution would be calculated by assuming that this distribution is lognormal.
21 Therefore,
22 GSD^,^ = exp{(log(10) - log(GMpolM(D)y*-|(l - PIO^,^)} (4)
23 where $'' denotes the inverse of the standard normal distribution function.
24 Table 6- 1 6 presents the estimated blood-lead concentration and health effect endpoints
25 that result when applying this alternative scaling algorithm, under both the IEUBK and empirical
26 models. The example options for standards that are assumed in this analysis are the same as
27 those considered in Section 6.4.1 above and are specified in a footnote to Table 6-16. For
28 comparison purposes, this table also contains the estimates under the original version of the
29 scaling approach that was utilized in the §403 risk analysis.
30 Effect on risk analysis. As indicated in Table 6-16, when the probability of exceeding
31 10 ug/dL is scaled instead of the GSD, the estimated probability is approximately 64% higher
32 under the IEUBK model (1.66% to 2.72%), but nearly 20% lower under the empirical model
33 (4.70% to 3.78%). Note that under the alternative approach, estimates based on the IEUBK and
34 empirical models are more similar to each other than under the original scaling algorithm. In the
35 alternative approach, the estimated post-§403 GSD is the same under both models: 1.96. Note
DRAFT - DO NOT CITE OR QUOTE 275 August 28, 2000
-------
1
2
3
4
5
Table 6-16. Estimated Post-§403 Health and Blood-Lead Concentration Endpoints Under
the Original and Alternative Scaling Algorithms for Characterizing the Post-
§403 Blood-Lead Distribution
Original scaling algorithm: Geometric mean and GSD are scaled.
Alternative scaling algorithm: Geometric mean and the probability of PbB exceeding 10/yg/dL are scaled.
Health Effect and Blood-Lead
Concentration Endpoints
% of Children with PbB 2 20 A/g/dL
% of Children with PbB 2 10/yg/dL
%of Children with IQ < 70 due to
lead exposure
% of Children with IQ decrement 2 1
due to lead exposure
% of Children with IQ decrement 2 1
due to lead exposure
% of Children with IQ decrement 2 1
due to lead exposure
Avg. IQ decrement due to lead
exposure
Geometric Mean PbB (GSD)
Post-§403 Estimates Under the
Risk Management Analysis
(Original Scaling Algorithm)
IEUBK Model
0.0539
1.66
0.0984
28.3
4.31
0.858
0.848
2.74(1.84)
Empirical
Model
0.406
4.70
0.110
36.3
9.30
2.93
1.00
3.03 (2.04)
Post-§403 Estimates Under the
Alternative Scaling Algorithm
IEUBK Model
0.156
2.72
0.102
30.1
6.05
1.56
0.884
2.74(1.96)
Empirical
Model
0.249
3.78
0.107
35.5
8.03
2.24
0.977
3.03 (1.96)
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Note: Example dust and soil standards were set at: 100 //g/ft* for floor dust-lead loading. 500//g/ft* for window sill dust-
lead loading, and 2,000 //g/g for soil-lead concentration Paint maintenance is performed if more than 5 ft1, but less than
20 ft2, of deteriorated lead-based paint exists. Paint abatement is performed if more than 20 ft1 of deteriorated lead-based
paint exists. GSD = geometric standard deviation. PbB = blood-lead concentration
25 that there was no change in the manner in which the geometric mean blood-lead concentrations
26 were determined, and therefore, no change is noted between the two approaches.
27 The above results indicate that the alternative scaling approach has a more significant
28 impact on the IEUBK model-based estimates compared to the empirical model-based estimates.
29 The impact of the approach on the empirical model-based estimates is a reduction in the risk
30 estimates due to a 4% reduction in the estimated GSD, while the impact on IEUBK model-based
31 estimates is an increase in the risk estimates due to a 6.5% increase in the estimated GSD.
32 However, because the two approaches did not differ in how the post-§403 geometric mean blood-
33 lead level was calculated, the empirical model estimates remain higher than the IEUBK model
34 estimates.
DRAFT - DO NOT CITE OR QUOTE
276
August 28, 2000
-------
1 7.0 REFERENCES
2 Abadin, H.G., Wheeler, J.S., Jones, D.E., De Rosa, C.T., and MSPH (1997) "A Framework to
3 Guide Public Health Assessment Decisions at Lead Sites." Journal of Clean Technology
4 Environmental Toxicology and Occupational Medicine. 6(3):25-237.
5 Altmann, L., Weinsberg, F., Sveinsson, K., Lilienthal, H., Wiegand, H., and Winneke, G. (1993)
6 "Impairment of Long-Term Potentiation and Learning Following Chronic Lead Exposure."
7 Toxicol. Lett. 66:105-112.
8 Angell, N.F., and Weiss, B. (1982) "Operant Behavior of Rats Exposed to Lead Before or After
9 Weaning." Toxicology and Applied Pharmacology. 63:62-71.
10 Angle, C.R., Manton, W.L, and Stanek, K.L. (1995) "Stable Isotope Identification of Lead
11 Sources in Preschool Children: the Omaha Study." Journal of Toxicology -- Clinical
12 Toxicology. 33(6):657-662.
13 Angle, C. R., and Mclntire, M. S. (1979) "Environmental Lead and Children: The Omaha
14 Study." Journal of Toxicology and Environmental Health. 5:855-870.
15 Baghurst, P.A., McMichael, A.J., Wigg, N.R., Vimpani, G., Robertson, E.F., Roberts, R.J., and
16 Tong, S. (1992) "Environmental Exposure to Lead and Children's Intelligence at the Age of
17 Seven Years: The Port Pirie Cohort Study." New England Journal of Medicine. 327:1279-
18 1284.
19 Banks, E.G., Ferretti, L.E., and Shucard, D.W. (1997) "Effects of Low Level Lead Exposure on
20 Cognitive Function in Children: a Review of Behavioral, Neuropsychological and
21 Biological Evidence." Neurotoxicology. 18(l):237-282.
22 Barltrop, D., and Meek, F. (1979) Effect of particle size on lead absorption from the gut.
23 Archives of Environmental Health. 34: 280-285.
24 Barltrop, D., Strehlow, C.D., Thorton, I., and Webb, J.S. (1974) "Significance of High Soil Lead
25 Concentrations for Childhood Lead Burdens." Environmental Health Perspectives. 7:75-
26 82.
27 Bates, M., Malcolm, M., Wyatt, R., Galloway, Y., Garrett, N., Speir, T., and Read, D. (1995)
28 "Lead in Children from Older Housing Areas in the Wellington Region." New Zealand
29 Medical Journal. 108(1009):400-404.
30 Bellinger, D., Stiles, K.M., and Needleman, H. (1992) "Low-Level Lead Exposure, Intelligence,
31 and Academic Achievement: A Long-term Follow-up Study." Pediatrics. 90(6): 855-861.
DRAFT-DO NOT CITE OR QUOTE 277 August 28.2000
-------
1 Bellinger, D., Sloman, J., Leviton, A., Rabinowitz, M., Needleman H.L., and Waternaux, C.
2 (1991) "Low-Level Lead Exposure and Children's Cognitive Function in the Preschool
3 Years." Pediatrics. 87(2):219-227.
4 Bellinger, D.C., Leviton, A., Waternaux, C., Needleman, H., and Rabinowitz, M. (1987)
5 "Longitudinal Analyses of Prenatal and Postnatal Lead Exposure and Early Cognitive
6 Development." New England Journal of Medicine. 316:1037-1043.
7 Bellinger, D.C., Leviton, A., Needleman, H.L., Waternaux, C., and Rabinowitz, M. (1986a)
8 "Low-Level Lead Exposure and Infant Development in the First Year." Neurobehavioral
9 Toxicology and Teratology. 8:151-161.
10 Bellinger, D.C., Leviton, A., Rabinowitz, M., Needleman, H., and Waternaux, C. (1986b)
11 "Correlates of Low-Level Lead Exposure in Urban Children at 2 Years of Age."
12 Pediatrics. 77:826-833.
13 Bellinger, D.C., Leviton, A., Watemaux, C., et al. (1985a) "A Longitudinal Study of the
14 Developmental Toxicity of Low-Level Lead Exposure in the Prenatal and Early Postnatal
15 Periods." In: Lekkas TD, ed. International Conference on Heavy Metals in the
16 Environment, Athens, Greece, September, Vol 1. Edinburgh, United Kingdom: CEP
17 Consultants, Ltd, 32-34.
18 Bellinger, D.C., Leviton, A., Watemaux, C., and Allred, E. (1985b) "Methodological Issues in
19 Modeling the Relationship Between Low-Level Lead Exposure and Infant Development:
20 Examples from the Boston Lead Study." Environmental Research. 38:119-129.
21 Bellinger, D.C. and Needleman, H.L. (1983) "Lead and the Relationship Between Maternal and
22 Child Intelligence." Journal of Pediatrics. 102:523-527.
23 Bhatia, M.S., Rai, S., Singbhal, P.K., Nigam, V.R., Bohra, N., and Malik, S.C. (1988) "Pica:
24 Prevalence and Etiology." Indian Pediatr. 25:1165-1170.
25 Bornschein, R., Clark, S., Pan, W., and Succop, P. (1990) "Midvale Community Lead Study."
26 University of Cincinnati.
27 Bornschein, R. L., Succop, P., Krafft, K., Clark, S., Peace, B., and Hammond, P. B. (1986)
28 "Exterior Surface Dust Lead, Interior House Dust Lead and Childhood Lead Exposure in an
29 Urban Environment." Conference in Trace Metals in Environmental Health, Columbia,
30 MO.
31 Bornschein, R.L., Succop, P., Dietrich, K.N., Clark, C.S., Que Hee, S., and Hammond, P.B.
32 (1985a) "The Influence of Social and Environmental Factors on Dust Lead, Hand Lead, and
33 Blood Lead Levels in Young Children." Environmental Research. 38:108-118.
DRAFT -- DO NOT CITE OR QUOTE 278 August 28.2000
-------
1 Bornschein, R. L., Hammond, P. B., Dietrich, K. N., Succop, P., Krafft, K., Clark, S., Berger, O.,
2 Pearson, D., and Que Hee, S. (1985b) 'The Cincinnati Prospective Study of Low-Level
3 Lead Exposure and Its Effects on Child Development: Protocol and Status Report."
4 Environmental Research, 38:4-18.
5 Brunekreef, B., Veenstra, S.J., Biersteker, K., and Boleij, J.S.M. (1981) 'The Arnhem Lead
6 Study: Lead Uptake by 1- to 3- Year-Old Children Living in the Vicinity of a Secondary
7 Lead Smelter in Arnhem, Netherlands." Environmental Research. 25:441-448.
8 Burger, J., Carruth-Hinchey, C., Ondroff, J., and McMahon M. (1998) "Effects of Lead on
9 Behavior, Growth, and Survival of Hatchling Slider Turtles." Journal of Toxicology and
10 Environmental Health. 55:495-502.
11 Burns, J.M., Baghurst, P.A., Sawyer, M.G., et al. (1999) "Lifetime Low-Level Exposure to
12 Environmental Lead and Children's Emotional and Behavioral Development: The Port
13 Pirie Cohort Study." American Journal of Epidemiology. 149(8):740-749.
14 Bushnell, P.J. and Bowman, R.E. (1979) "Persistence of Impaired Reversal Learning in Young
15 Monkeys Exposed to Low Levels of Dietary Lead." J. Toxicology and Environmental
16 Health. 5: 1015-1023.
17 Butte-Silver Bow Department of Health, and Department of Environmental Health, University of
18 Cincinnati. (1991) "The Butte-Silver Bow Environmental Health Lead Study." Draft Final
19 Report.
20 Calabrese, E., Barnes, R., Stanek, E.J. m, Pastides, H., Gilbert, C.E., Veneman, P., Wang, X.,
21 Lasztity, A., and Kostecki, P.T. (1989) "How Much Soil Do Young Children Ingest: An
22 Epidemiological Study." Regul Toxicol Pharmacol. 10:123-137.
23 Calabrese, E.J., and Stanek, E.J. (1993) "Soil Pica: Not a Rare Event. " Journal of
24 Environmental Sci. Health, A28(2):373-384.
25 Calabrese, E., Stanek, E.J., and Barnes, R. (1997) "Soil Ingestion Rates in Children Identified
26 by Parental Observation as Likely High Soil Ingesters." /. Soil Contam ,6(3):271-279.
27 Calabrese, E., Stanek, E.J., and Gilbert, C.E. (1991) "Evidence of Soil Pica Behavior and
28 Quantification of Soil Ingested." Hum Exp Toxicol. 10:245-249.
29 Calabrese, E., Stanek, E.J., and Gilbert, C.E. (1993) "Lead Exposures in a Soil Pica Child."
30 J. Environ. Sci. Health. A28 (2):353-362.
31 Carroll, R.J., and Galindo, C.D. (1999) "Measurement Error, Biases, and the Validation of
32 Complex Models." Environmental Health Perspectives. 106:1535-1539.
DRAFT -- DO NOT CITE OR QUOTE 279 August 28.2000
-------
1 Carroll R.J., Ruppert, D., and Stefanski. L.A. (1995) Measurement Error in Nonlinear Models.
2 London: Chapman & Hall.
3 CDC (1997) "Update: Blood-Lead Levels - United States, 1991-1994." Morbidity and Mortality
4 Weekly Report. Centers for Disease Control and Prevention, Public Health Service, U.S.
5 Department of Health and Human Services. 21 February 1997. 46(7): 141-146.
6 Chen, H., Ma, T., Hume, A. S., and Ho, I. K. (1998) "Developmental Lead Exposure Alters the
7 Distribution of Protein Kinase C Activity in the Rat Hippocampus." Biomedical and
8 Environmental Sciences. 11:61-69.
9 Clark, S., Bornschein, R.L., Succop, P., Roda, S., and Peace, B. (1991) "Urban Lead Exposures
10 of Children in Cincinnati, Ohio." Chem. Spec. Bioavail. 3:163-171.
11 Cohen, C.J., Bowers, G.N., and Lepow, M.L. (1973) "Epidemiology of Lead Poisoning: A
12 Comparison Between Urban and Rural Children." JAMA. 226:1430-1433.
13 Cory-Slechta, D.A. (1997) "Relationships Between Pb-Induced Changes in Neurotransmitter
14 System Function and Behavioral Toxicity." Neurotoxicology. 18(3): 673-688.
15 Cory-Slechta, D.A., Weiss, B., and Cox, C. (1985) "Performance and Exposure Indices of Rate
16 Exposed to Low Concentrations of Lead." Toxicology and Applied Pharmacology.
17 78:291-299.
18 Cory-Slechta, D.A. and Thompson, T. (1979) "Behavioral Toxicity of Chronic Post-weaning
19 Lead Exposure in the Rat." Toxicology and Applied Pharmacology. 47:151 -159.
20 Cotter-Howells, J. (1994) "Lead Phosphate Formation in Soils." Environmental Chemistry and
21 Health. 16(2): 86.
22 Cutler, M.G. (1977) "Effects of Exposure to Lead on Social Behavior in the Laboratory Mouse."
23 Psychopharmacology. 52: 279-282.
24 Davies, D.J.A, Thornton, I., Watt, J.M., Culbard, E.B., Harvey, P.O., Delves, H.T., Sherlock,
25 J.C., Smart, G.A., Thomas, J.F.A., and Quinn, M.J. (1990) "Lead Intake and Blood Lead in
26 Two-Year-Old U.K. Urban Children." The Science of the Total Environment. 90:13-29.
27 Davis, A., Ruby, M.V. and Bergstrom, P.D. (1995) "Factors Controlling Lead Bioavailability in
28 the Butte Mining District, Montana, USA." Environmental Geochemistry and Health.
29 16(3-4): 147-157.
30 Davis, J.M., Otto, D.A., Weil, D.E., and Grant, L.D. (1990) "The Comparative Developmental
31 Neurotoxicity of Lead in Humans and Animals." Neurotoxicology and Teratology. 12:
32 215-229.
DRAFT - DO NOT CITE OR QUOTE 280 August 28,2000
-------
1 Davis, S., Waller, P., Buschom, R., Ballou, J., and White, P. (1990) "Quantitative Estimates of
2 Soil Ingestion in Normal Children Between the Ages of 2 and 7 Years: Population-Based
3 Estimate Using Aluminum, Silicon, and Titanium as Soil Tracer Elements." Archives of
4 Environmental Health. 45:112-122.
5 Day, J.C. (1993) Population Projections of the United States, by Age, Sex, Race, and Hispanic
6 Origin: 1993 to 2050. U.S. Bureau of the Census, Current Population Reports P25-1104.
7 Dersimonian, R. and Laird, N. (1986) "Meta-Analysis in Clinical Trials." Controlled Clinical
8 Trials. 7:177-188.
9 Diemel, J.A., Brunekreef, B., Boleij, J.S., Biersteker, K. and Veenstra, S.J. (1981) 'The Arnhem
10 Lead Study n. Indoor Pollution, and Indoor/Outdoor Relationships." Environmental
11 Research. 25:449-456.
12 Dieter, M.P, Matthews, H.B., Jeffcoat, R.A. and Moseman, R.F. (1993) "Comparison of Lead
13 Bioavailability in F344 Rats Fed Lead Acetate, Lead Oxide, Lead Sulfide or Lead Ore
14 Concentrate from Skagway, Alaska." Journal of Toxicology and Environmental Health.
15 39(1): 79-93.
16 Dietrich, K.N., Berger, O.G., Succop, P.A., Hammond, P.B., and Bornschein, R.L.. (1993) 'The
17 Developmental Consequences of Low To Moderate Prenatal and Postnatal Lead Exposure:
18 Intellectual Attainment in the Cincinnati Lead Study Cohort Following School Entry."
19 Neurotoxicology and Teratology. 15:37-44.
20 Donovan, J., Anderson, P., Daley, C., Lea, T., and Luhse, P. (1996) "Lead in Australian
21 Children. Report on the National Survey of Lead in Children." Australian Institute of
22 Health and Welfare, Canberra, Commonwealth of Australia.
23 Duggan, M.J. and Inskip, M.J. (1985) "Childhood Exposure to Lead in Surface Dust and Soil: A
24 Community Health Problem." Public Health Reviews. 13:1-54.
25 Elhelu, M. A., Caldwell, D., and Hirpassa, W. (1995) "Lead in Inner-City Soil and Its Possible
26 Contribution to Children's Blood Lead." Archives of Environmental Health. 50(2): 165-
27 169.
28 Ernhart, C.B., Morrow-Tlucak, M., Worf, A.W., Super, D., and Drotar, D. (1989) "Low Level
29 Lead Exposure in the Prenatal and Early Preschool Periods; Intelligence Prior to School
30 Entry." Neurotoxicology and Teratology. 11:161 -70.
31 Farfel, M.R. and Chisolm, J.J., Jr. (1991) "An Evaluation of Experimental Practices for
32 Abatement of Residential Lead-Based Paint: Report on a Pilot Project." Environmental
33 Research. 55:199-212.
DRAFT-DO NOT CITE OR QUOTE 281 August 28,2000
-------
1 Farfel, M.R. and Chisolm, J.J., Jr. (1990) "Health and Environmental Outcomes of Traditional
2 and Modified Practices for Abatement of Residential Lead-Based Paint." American
3 Journal of Public Health. 80(10): 1240-1245.
4 Fleiss, J.L. (1981) Statistical Methods for Rates and Proportions. Second Edition. New York:
5 John Wiley & Sons.
6 Francek, M.A., Makimaa, B., Pan, V., and Hanko, J.H. (1994) "Small Town Lead Levels: A Case
7 Study from the Homes of Pre-schoolers in Mt. Pleasant, Michigan." Environmental
8 Research. 84(2): 159-166.
9 Freeman, G.B., Johnson, J.D., Killinger, J.M., Liao, S.C., Feder, P.I., Davis, A.O., Ruby, M.V.,
10 Channey, R.L., Lovre, S.C., and Bergstrom, P.D. (1992) "Relative Bioavailability of Lead
11 from Mining Waste Soil in Rats." Fund. Appl. Toxicol. 19: 399-398.
12 Fulton, M., Raab, G., Thomson, G., Laxen, D., Hunter, R., and Hepburn, W. (1987) "Influence of
13 Blood Lead on the Ability and Attainment of Children in Edinburgh." Lancet. 1:1221 -
14 1226.
15 Galke, W. A., Hammer, D.I., and Lawrence, S. W. (1975) "Environmental Determinants of Lead
16 Burdens in Children." In: International Conference on Heavy Metals in the Environment:
17 Symposium Proceedings, T. C. Hutchinson, S. Epstein, A. I. Page, J. VanLoon and T.
18 Davey (eds.), Institute for Environmental Studies, Toronto, ON, Canada, 3:53-74.
19
20 Gallacher, J.E., Elwood, P.C., Phillips, K.M., Davies, B.E., and Jones, D.T. (1984) "Relation
21 Between Pica and Blood Lead in Areas of Differing Lead Exposure." Archives of Diseases
22 in Childhood. 59(1): 40-44.
23 Geist, C.R., Balko, S.W., Morgan, M.E., and Angiak, R. (1985) "Behavioral Effects Following
24 Rehabilitation from Postnatal Exposure to Lead Acetate." Percept. Mot. Skills. 60: 527-
25 536.
26 Gilbert, S.G., and Rice, D.C. (1987) "Low-level Lifetime Lead Exposure Produces Behavioral
27 Toxicity (Spatial Discrimination Reversal) in Adult Monkeys." Toxicology and Applied
28 Pharmacology. 91:484-490.
29 Goldstein, G. (1992) "Developmental Neurobiology of Lead Toxicity." In Human Lead
30 Exposure (Needleman, H., ed.). 125-135.
31 Goldstein, G. (1990) "Lead Poisoning and Brain Cell Function." Environ. Health Perspect.
32 89:91-94.
DRAFT -- DO NOT CITE OR QUOTE 282 August 28,2000
-------
1 Greene, T., Emhart, C.B., and Boyd, T.A. (1992) "Contributions of Risk Factors to Elevated
2 Blood and Dentine Lead Levels in Preschool Children." The Science of the Total
3 Environment. 115:239-260.
4 Harrison, G. (1987) "A Survey of the Lead Distribution in the Soil of Corpus Christi, Texas."
5 Texas Journal of Science. 39(1): 15-22.
6 Harvey, P.G., Hamlin, M.W., Kumar, R., Morgan, G., Spurgeon, A., and Delves, H.T. (1988)
7 "Relationship Between Blood Lead, Behaviour, Psychometric and Neuropsychological Test
8 Performance in Young Children." British Journal of Developmental Psychology. 29:43-52.
9 Hastings, L., Cooper, G.P., Bomschein, R.L., and Michaelson, LA. (1977) "Behavioral Effects
10 of Low Lead Neonatal Lead Exposure." Pharmac. Biochem. Behavior. 7: 37-42.
11 Hatzakis, A., Kolevi, A., Katsouyanni, K., et al. (1987) "Psychometric Intelligence and
12 Attentional Performance Deficits in Lead Exposed Children." International Conference on
13 Heavy Metals in the Environment. Edinburgh, Scotland; CEP Consultants. 16:57-67.
14 Hatzakis, A., Kokkeni, A., Maranelias, C, Katsouyanni, K., Salaminios, F., Kalandidi, A., et al.
15 (1989) "Psychometric Intelligence Deficits in Lead-exposed Children." In: Lead Exposure
16 and Child Development (Smith MA., Grant, I.D., Sora Al, eds.). London: Kluwer,
17 1989:211-23.
18 Hawk, B.A., Schroeder, S.R., Robinson, G., et al. (1986) "Relation of Lead and Social Factors to
19 IQ of Low-SES Children: A Partial Replication." American Journal of Mental Deficiency.
20 91:178-183.
21 Karam, H.S., and Beck, B.D. (1990) "Current Issues in Determining Acceptable Lead
22 Concentrations in Soils." Comments on Toxicology. 3(6):509-529.
23 Kaufman, A.S. (1996) "IQ, Lead Level, and Inferences from Research Studies: Comments
24 Addressing the Underlying Science Forming the Basis of HUD's Regulatory Impact
25 Analysis." Submitted to the Environmental Health Committee of the Science Advisory
26 Board, U.S. Environmental Protection Agency, in their technical review of the proposed
27 TSCA Section 403 regulations (Identification of Dangerous Levels of Lead). August 1996.
28 Kenny, D.A. (1979) Correlation and Causality. New York: John Wiley & Sons, Inc.
29 Krueger, J. A. and Duguay, K. M. (1989) "Comparative Analysis of Lead in Maine Urban Soils."
30 Bulletin of Environmental Contamination and Toxicology. 42:574-581.
31 Kuhlmann, A.C., McGlothan, J.L., and Guilarte, T.R. (1997) "Developmental Lead Exposure
32 Causes Spatial Learning Deficits in Adult Rats." Neuroscience Newsletters. 233:101-104.
DRAFT -- DO NOT CITE OR QUOTE 283 August 28,2000
-------
1 Lacey, E.P. (1990) "Broadening the Perspective of Pica: Literature Review." Public Health
2 Reports. 105(l):29-35.
3 Lanphear, B.P., Matte, T.D., Rogers, J., Clickner, R.P., Dietz, B., Bomschein, R.L., Succop, P.,
4 Mahaffey, K.R., Dixon, S., Galke, W., Rabinowitz, M., Farfel, M., Rohde, C., Schwartz, J.,
5 Ashley, P., and Jacobs, D. (1998) 'The Contribution of Lead-Contaminated House Dust
6 and Residential Soil to Children's Blood Lead Levels: A Pooled Analysis of 12
7 Epidemiologic Studies." Environmental Research (Section A). 79:51-68.
8 Lanphear, B.P., Weitzman, M., and Eberly, S. (1996a) "Racial Differences in Urban Children's
9 Environmental Exposures to Lead." American Journal of Public Health. 86( 10): 1460-1463.
10 Lanphear, B.P., Weitzman, M., Winter, N.L., Eberly, S., Yakir, B., Tanner, M., Emond, M., and
11 Matte, T.D. (1996b) "Lead-Contaminated House Dust and Urban Children's Blood Lead
12 Levels." American Journal of Public Health. 86(10):1416-1421.
13 Lansdown, R., Yule, W., Urbanowicz, M.A., et al. (1986) 'The Relationship Between Blood
14 Lead Concentrations, Intelligence, Attainment and Behavior in a School Population: The
15 Second London Study." International Archives of Occupational Environmental Health.
16 57:225-235.
17 Laperche, V., Logan, T.J., Gaddam, P., and Traina, S.J. (1997) "Effect of Apatite Amendments
18 on Plant Uptake of Lead from Contaminated Soil." Environmental Science and
19 Technology. 31(10): 2745-2753.
20 Lilienthal, H., Winneke, G., and Ewert, T. (1990) "Effects of Lead on Neurophysiological and
21 Performance Measures: Animal and Human Data." Environmental Health Perspectives.
22 89:21-25.
23 Lyngbye, T., Hansen, O.N., and Grandjean, P. (1990) "Predictors of'Tooth Lead Level with
24 Special Reference to Traffic: A Study of Lead-Exposure in Children." Int. Arch. Occup.
25 Environ. Health. 62(6):417-422.
26 Maddaloni, M., Manton, W., Blum, C., Lolacono, N., and Graziano, J. (Date unknown)
27 Bioavailability of Soil-Borne Lead in Adults, by Stable Isotope Dilution. Columbia
28 University, New York, NY; University of Texas at Dallas, Richardson, TX.
29 Marais, M.L., and Wecker, W.E. (1998) "Correcting for Omitted Variables and Measurement
30 Error Bias in Regression with an Application to the Effect of Lead on IQ." Journal of the
31 American Statistical Association. 93:494-505.
32 McKinney, J. (1993) "Metals Bioavailability and Disposition Kinetics Research Needs
33 Workshop, July 18-19, 1990." Toxicological and Environmental Chemistry. 38: 1-71.
DRAFT -- DO NOT CITE OR QUOTE 284 August 28,2000
-------
1 Mello.CF., Kraemer, C.K., Filippin.A., Morsch, V.M., Rodrigues, A.L., Martins, A.F., and
2 Rubin, M.A. (1998) "Effect of Lead Acetate on Neurobehavioral Development of Rats."
3 Brazilian Journal of Medical and Biological Research. 31 (7): 943-950.
4 Mielke, H. W. (1995) "Lead in New Orleans Soils: New Images of an Urban Environment."
5 Environmental Geochemistry and Health. 16(3/4): 123-128.
6 Mielke, H. W. (1993) "Lead Dust Contaminated U.S.A. Communities: Comparison of Louisiana
7 and Minnesota." Applied Geochemistry. Supp. Issue No. 2,257-261.
8 Mielke, H. W., Adams, J. E., Huff, B., Pepersack, J., Reagan, P. L., Stoppel, D., and Mielke, P.
9 W. (1992) "Dust Control as a Means of Reducing Inner-City Childhood Pb Exposure." In:
10 Trace Substances in Environmental Health - XXV. Supplement to Volume 14,
11 Environmental Geochemistry and Health (B. D. Beck, ed.) 121 -128.
12 Mielke, H. W., Adams, J. L., Reagan, P. L., and Mielke, P. W. (1989) "Soil-Dust Lead and
13 Childhood Lead Exposure as a Function of City Size and Community Traffic Flow: The
14 Case for Lead Abatement in Minnesota." In: Lead in Soil: Issues and Guidelines,
15 Supplement to Volume 9 of Environmental Geochemistry and Health (B. E. Davis and B. G.
16 Wixson, eds.). 253-271.
17 Mielke, H. W., Anderson, J.C., Berry, K. J., Mielke, P. W., Chancy, R. L., and Leech, M. (1983)
18 "Lead Concentrations in Inner-City Soils as a Factor in the Child Lead Problem." American
19 Journal of Public Health. 73(12): 1366-1369.
20 Nagymajtenyi, L., Desi, L, Schulz, H., and Papp, A. (1990) "Consequences of Lead Exposure of
21 Rats During Pregnancy, Lactation, and Postweaning. A Combined Behavioural and
22 Neurotoxicological Study." International Journal of Environmental Health Research.
23 8(2): 121-135.
24 NCHS (1984) "Blood Lead Levels for Persons Ages 6 Months - 74 Years: United States,
25 1976-1980." Vital. & Health Statistics. National Center for Health Statistics. 11:233.
26 Needleman, H.L. (1998) "IQ, Lead Level, and Inferences from Research Studies." Comments to
27 the Environmental Health Committee of the Science Advisory Board, U.S. Environmental
28 Protection Agency, in their technical review of the proposed TSCA Section 403 regulations
29 (Identification of Dangerous Levels of Lead). 2 September 1998.
30 Needleman, H.L., and Gatsonis, C.A. (1990) "Low-Level Lead Exposure and the IQ of
31 Children." Journal of the American Medical Association. 263(5):673-678.
32 Petit, T.L., Alfano, D.P., and LeBoutillier, J.C. (1983) "Early Lead Exposure and the
33 Hippocampus: a Review and Recent Advances." Neurotoxicology. 4(l):79-94.
DRAFT - DO NOT CITE OR QUOTE 285 August 28,2000
-------
1 Pendleton, S. (personal communication to Brad Schultz, U.S. EPA).
2 Pocock, S.J., Smith, M., and Baghurst, P. (1994) "Environmental Lead and Children's
3 Intelligence: A Systematic Review of the Epidemiological Evidence." BMJ. 309:1189-
4 1197.
5 Que Hee, S., Peace, B., Clark, S., Boyle, J. R., Bornschein, R. L., and Hammond, P. B. (1985)
6 "Evolution of Efficient Methods to Sample Lead Sources, Such as House Dust and Hand
7 Dust, in the Homes of Children." Environmental Research. 38:77-95.
8 Rabinowitz, M.B., Wang, J.D., and Soong, W.T. (1992) "Apparent Threshold of Lead's Effect
9 on Child Intelligence." Bulletin of Environmental Contamination and Toxicology. 48:688-
10 695.
11 Rabinowitz, M., Leviton, A., Needleman, H., Bellinger, D., and Waternaux, C. (1985)
12 "Environmental Correlates of Infant Blood Lead Levels in Boston." Environmental
13 Research. 38:96-107.
14 Rice, D. C. (1996) "Behavioral Effects of Lead: Commonalities Between Experimental and
15 Epidemiologic Data." Environmental Health Perspectives. 104 (Suppl. 2): 337-351.
16 Rice, D.C. (1988) "Chronic Low-Lead Exposure Does Not Affect Simple Reaction Time."
17 Neurotoxicology. 9:105-108.
18 Rice, D.C. (1985) "Chronic Low-Lead Exposure from Birth Produces Deficits in Discrimination
19 Reversal in Monkeys." Toxicology and Applied Pharmacology. 77:201-210.
20 Rice, D.C. (1984) "Behavior Deficit (Delayed Matching to Sample) in Monkeys Exposed from
21 Birth to Low Levels of Lead." Toxicology and Applied Pharmacology. 75:337-345.
22
23 Rice, D.C. and Gilbert, S.G. (1990a) "Sensitive Periods for Lead-Induced Behavioral
24 Impairment (Nonspatial Discrimination Reversal) in Monkeys." Toxicology and Applied
25 Pharmacology. 102:101-109.
26 Rice, D. C. and Gilbert, S.G. (1990b) "Lack of Sensitive Period for Lead-Induced Behavioral
27 Impairment on a Spatial Delayed Alternation Task in Monkeys." Toxicology and Applied
28 Pharmacology. 103:364-373.
29 Rieuwerts, J.S., and Farago, M.E. (1995) "Lead Contamination in Smelting and Mining
30 Environments and Variations in Chemical Forms and Bioavailability." Chemical
31 Speciation and Bioavailability. 7(4): 113-123.
DRAFT-DO NOT CITE OR QUOTE 286 August 28,2000
-------
1 Roberts, J.W., Han, W., and Ruby, M.G. (1993) "Evaluation of Dust Samplers for Bare Floors
2 and Upholstery." Draft Report to Atmospheric Research and Exposure Assessment
3 Laboratory, Office of Research and Development, U.S. Environmental Protection Agency.
4 October 1993.
5 Roels, H.A., Buchet, J.P., Lauwerys, R.R., Bruaux, P., Claeys-Thoreau, F., LaFontaine, A., and
6 Verduyn, G. (1980) "Exposure to Lead by the Oral and the Pulmonary Routes of Children
7 Living in the Vicinity of a Primary Lead Smelter." Environmental Research. 22:81-94.
8 Romieu, I., Carreon, T., Lopez, L., Palazuelos, E., Rios, C, Manuel, Y., and Hernandez-Avila,
9 M. (1995) "Environmental Urban Lead Exposure and Blood Lead Levels in Children of
10 Mexico City." Environmental Health Perspectives. 103(11): 1036-1040.
11 Schmitt, M.D.C., Trippler, D.J., Wachtler, J.N. and Lund, G.V. (1988) "Soil Lead
12 Concentrations in Residential Minnesota as Measured by ICP-AES." Water, Air, and Soil
13 Pollution. 39:157-168.
14 Schroeder, S.R., Hawk, B., Otto, D., Mushak, P., and Hicks, R.E. (1985) "Separating the Effects
15 of Lead and Social Factors on IQ." Environmental Research. 91:178-183.
16 Schwartz, J. (1994) "Low-Level Lead Exposure and Children's IQ: A Meta-Analysis and Search
17 for a Threshold." Environmental Research. 65:42-55.
18 Schwartz, J. (1993) "Beyond LOEL's, p Values, and Vote Counting: Methods for Looking at the
19 Shapes and Strengths of Association." Neurotoxicology and Teratology. 14(2-3):237-246.
20 Sedman, R.M. (1989) "The Development of Applied Action Levels for Soils Contact: A
21 Scenario for the Exposure of Humans to Soils in a Residential Setting." Environmental
22 Health Perspectives. 79:291-313.
23 Sedman, R.M., and Mahmood, R.J. (1994) "Soil Ingestion by Children and Adults Reconsidered
24 Using the Results of Recent Tracer Studies." Journal of Air & Waste Management
25 Association. 44:141-144.
26 Shellshear, I.D., Jordon, L.D., Hogan, D.J., and Shannon, F.T. (1975) "Environmental Lead
27 Exposure in Christchurch Children: Soil Lead as a Potential Hazard." New Zealand Med.
28 J. 81:382-386.
29 Sheppard, S.C. (1998) "Geophagy: Who Eats Soils and Where Do Contaminants Go?"
30 Environmental Geology (Berlin). 33(2-3): 109-114.
31 Silbergeld, E.K. (1991) "Aneurological Perspective on Lead Toxicity." In Human Lead
32 Exposure (Needleman, H.L. ed.). Boca Raton, Florida: CRC Press, 89-103.
DRAFT-DO NOT CITE OR QUOTE 287 August 28,2000
-------
1 Silva, P.A., Hughes, P., Williams, S., and Faed, J.M. (1988) "Blood Lead, Intelligence, Reading
2 Attainment, and Behaviour in Eleven Year Old Children in Dunedin, New Zealand."
3 Journal of Child Psychology and Psychiatry. 29:43-52.
4 Simon, S.L. (1998) "Soil Ingestion by Humans: A Review of History, Data, and Etiology with
5 Application to Risk Assessment of Radioactively Contaminated Soils." Health Physics,
6 74(6):647-672.
7 Stanek, E.J. m, and Calabrese, E.J. (1995) Daily Estimates of Soil Ingestion in Children.
8 Environmental Health Perspectives. 103: 276-285.
9 Stanek E.J., Calabrese, E.J., Mundt, K., Pekow, P., and Yeatts, K.B. (1998) "Prevalence of Soil
10 Mouthing/Ingestion Among Healthy Children Aged 1 to 6." Journal of Soil
11 Contamination. 7(2):227-242.
12 Stark, A. D., Quah, R. F., Meigs, J. W., and DeLouise, E. R. (1982) 'The Relationship of
13 Environmental Lead to Blood-Lead Levels in Children." Environmental Research. 27:372-
14 383.
15 Steuteville, W.D. (1990) "Interior Remediation of Residences: Lead Abatement (A Case
16 History of a Superfund Project and Public Involvement)." In: Procedures of the National
17 Conference of Hazardous Wastes and Hazardous Materials. Silver Springs, MD:
18 Hazardous Materials Research Control Institute.
19 Sutton, P. Athanasoulis, M., Flessel P., Guirguis, G., Haan, M., Schlag, R., and Goldman, L.
20 (1995) "Lead Levels in the Household Environment of Children in Three High-Risk
21 Communities in California." Environmental Research. 68:45-57.
22 Teichmean, J., Coltrin, D., Prouty, K., and Bir, W. A. (1993) "A Survey of Lead Contamination
23 in Soil Along Interstate 880, Alameda County, California." Journal of the American
24 Industrial Hygiene Association. 54(9):557-559.
25 Teng, S. T. (1990) "Roadside Dusts and Soils Contamination in Cincinnati, Ohio."
26 Environmental Management. 14(1): 107-114.
27 Tong, S. (1998) "Lead Exposure and Cognitive Development: Persistence and a Dynamic
28 Pattern." J. Pediatr. Child Health. 34:114-118.
29 Tong, S., Baghurst, P., McMichael, A., Sawyer, M., and Mudge, J. (1996) "Lifetime Exposure to
30 Environmental Lead and Children's Intelligence at 11-13 Years: the Port Pirie Cohort
31 Study." British Medical Journal. 312:1569-1575.
DRAFT - DO NOT CITE OR QUOTE 288 August 28,2000
-------
1 U.S. Department of Health and Human Services (1999) Toxicological Profile for Lead. Update,
2 7999. Atlanta: Agency for Toxic Substances and Disease Registry, Division of
3 Toxicology/Toxicology Information Branch, U.S. Department of Health and Human
4 Services.
5 U.S. Department of Health and Human Services (1992) Impact of Lead-Contaminated Soil on
6 Public Health. Atlanta: Agency for Toxic Substances and Disease Registry, Division of
7 Toxicology/Toxicology Information Branch, U.S. Department of Health and Human
8 Services. May 1, 1992.
9 U.S. Department of Health and Human Services (1991) "Philadelphia Neighborhood Lead
10 Study." Final Report. Atlanta: Agency for Toxic Substances and Disease Registry,
11 Division of Toxicology/Toxicology Information Branch, U.S. Department of Health and
12 Human Services.
13
14 U.S. Department of Housing and Urban Development (1998) "Evaluation of the HUD Lead-
15 Based Paint Hazard Control Grant Program: Fifth Interim Report, Progress as of September
16 1, 1997." Prepared by the National Center for Lead Safe Housing and the University of
17 Cincinnati Department of Environmental Health for the U.S. Department of Housing and
18 Urban Development. March 1998.
19 U.S. Department of Housing and Urban Development (1995a) 'The Relation of Lead-
20 Contaminated House Dust and Blood Lead Levels Among Urban Children." Volumes I
21 and n. Final Report to the U.S. Department of Housing and Urban Development Grant
22 from the Departments of Pediatrics, Biostatistics, and Environmental Medicine, The
23 Rochester School of Medicine, Rochester, New York, and The National Center for Lead-
24 Safe Housing, Columbia Maryland.
25 U.S. Department of Housing and Urban Development (1995b) "Guidelines for the Evaluation
26 and Control of Lead-Based Paint Hazards in Housing." Office of Lead-Based Paint
27 Abatement and Poisoning Prevention.
28 U.S. Department of Housing and Urban Development (1991) 'The HUD Lead-Based Paint
29 Abatement Demonstration (FHA)." Final Report, Office of Policy Development and
30 Research, U.S. Department of Housing and Urban Development.
31 U.S. Environmental Protection Agency (1999) "Analysis of Pathways of Residential Lead
32 Exposure in Children." Office of Prevention Pesticides, and Toxic Substances. EPA 747-
33 R-98-007.
34 U.S. Environmental Protection Agency (1998a) "Risk Analysis to Support Standards for Lead in
35 Paint, Dust, and Soil." Office of Pollution Prevention and Toxics, U.S. Environmental
36 Protection Agency. EPA 747-R-97-006.
DRAFT--DO NOT CITE OR QUOTE 289 August 28,2000
-------
1 U.S. Environmental Protection Agency (1998b) "An SAB Report: Technical Review of the
2 Proposed TSCA Section 403 Regulation (Identification of Dangerous Levels of Lead)."
3 Environmental Health Committee, Science Advisory Board, U.S. Environmental Protection
4 Agency. EPA-SAB-EHC-99-004, November 1998.
5 U.S. Environmental Protection Agency (1997a) "Lead-Based Paint Abatement and Repair and
6 Maintenance Study in Baltimore." Office of Pollution Prevention and Toxics, U.S.
7 Environmental Protection Agency. EPA 747-R-95-012. May 1997.
8 U.S. Environmental Protection Agency (1997b) "Lead Exposure Associated with Renovation
9 and Remodeling Activities: Environmental Field Sampling Study. Volume I: Draft
10 Technical Report." Office of Pollution Prevention and Toxics, U.S. Environmental
11 Protection Agency. EPA 747-R-97-007.
12 U.S. Environmental Protection Agency (1997c) "Lead Exposure Associated with Renovation
13 and Remodeling Activities: Environmental Field Sampling Study. Volume I: Draft
14 Technical Report." Office of Pollution Prevention and Toxics, U.S. Environmental
15 Protection Agency. EPA 747-R-96-007.
16 U.S. Environmental Protection Agency (1997d) "Summary and Assessment of Published
17 Information on Determining Lead Exposures and Mitigating Lead Hazards Associated with
18 Dust and Soil in Residential Carpets, Furniture, and Forced Air Ducts." Office of
19 Prevention Pesticides, and Toxic Substances, U.S. Environmental Protection Agency. EPA
20 747-S-97-001.
21 U.S. Environmental Protection Agency (1997e) "Lead-Based Paint Abatement and Repair and
22 Maintenance Study in Baltimore: Finding Based on Two Years of Follow-Up." Office of
23 Prevention Pesticides, and Toxic Substances, U.S. Environmental Protection Agency. EPA
24 747-R-97-005.
25 U.S. Environmental Protection Agency (1996a) "Urban Soil Lead Abatement Demonstration
26 Project, Volume I: Integrated Report." Office of Research and Development, U.S.
27 Environmental Protection Agency. EPA 600-P-93-001aF, April 1996.
28 U.S. Environmental Protection Agency (1996b) "Statistical Evaluation of the Relationship
29 Between Blood-Lead and Dust-Lead Based on Data from the Rochester Lead-in-Dust
30 Study." Office of Prevention Pesticides, and Toxic Substances. Final Report.
31 U.S. Environmental Protection Agency (1996c) "Statistical Evaluation of the Relationship
32 Between Blood-Lead and Dust-Lead Based on Pre-Intervention Data from the R&M
33 Study." Office of Prevention Pesticides, and Toxic Substances. Final Report.
DRAFT -- DO NOT CITE OR QUOTE 290 August 28,2000
-------
1 U.S. Environmental Protection Agency (1996d) "Comprehensive Abatement Performance
2 Study. Volume II: Detailed Statistical Results." Office of Prevention, Pesticides, and Toxic
3 Substances, U.S. Environmental Protection Agency. EPA 230-R-94-013b.
4 U. S. Environmental Protection Agency (1996e) "Lead-Based Paint Abatement and Repair and
5 Maintenance Study in Baltimore: Pre-Intervention Finding." EPA 747-R-95-012, August
6 1996.
7 U.S. Environmental Protection Agency (1995) "Report on the National Survey of Lead-Based
8 Paint in Housing. Base Report." Office of Pollution Prevention and Toxics, U.S.
9 Environmental Protection Agency. EPA747-R95-003.
10 U.S. Environmental Protection Agency (1994) "Guidance Manual for the Integrated Exposure
11 Uptake Biokinetic Model for Lead in Children." Office of Emergency and Remedial
12 Response, U.S. Environmental Protection Agency. EPA/540/R-93/081, PB93-963510.
13 U.S. Environmental Protection Agency (1989) "Supplement to 1986 Air Quality Criteria for
14 Lead - Volume I Addendum." (p. A1-67). Research Triangle Park NC: Environmental
15 Criteria and Assessment Office, U.S. Environmental Protection Agency. EPA/600/8-
16 89/049A.
17 U.S. Environmental Protection Agency (1986) "Air Quality Criteria for Lead, Volume IV of IV."
18 Research Triangle Park NC: Environmental Criteria and Assessment Office, U.S.
19 Environmental Protection Agency. EPA-600/8-83/028dF. June 1986.
20 Viverette, L., Mielke, H.W., Brisco, M., Dixon, A., Schaefer, J., and Pierre, K. (1996)
21 "Environmental Health in Minority and Other Underserved Populations: Benign Methods
22 for Identifying Lead Hazards at Day Care Centres of New Orleans." Environmental
23 Geochemistry and Health. 18(1): 41-45.
24 Walsh, T.J., and Tilson, H.A. (1984) "Neurobehavioral Toxicology of the Organoleads."
25 Neurotoxicology. 5:67-86.
26 Wang, L., Xu S.E., Zhang, G.D, and Wang, W.Y. (1989) "Study of Lead Absorption and its
27 Effect on Children's Development." Biomedical and Environmental Sciences. 2:325-330.
28 Weitzman, M., Aschengrau, A., Bellinger, D., Jones, R., Hamlin, J.S., and Beiser, A. (1993)
29 "Lead-Contaminated Soil Abatement and Urban Children's Blood Lead Levels." Journal of
30 the American Medical Association. 269(13):1647-1654.
31 Westat (1998) "A Report on the Analysis of the Blood Lead-Environmental Lead Relationship."
32 Report for the U.S. Department of Housing and Urban Development. 7 August 1998.
DRAFT -- DO NOT CITE OR QUOTE 291 August 28.2000
-------
1 Winneke, G., Lilienthal, H., and Kramer, U. (1996) 'The Neurobehavioural Toxicology and
2 Teratology of Lead." Arch Toxicol. Suppl. 18:57-70.
3 Winneke, G., Brockhaus, A., Ewers, V., Kramer, V., and Neuf, M. (1990) "Results from the
4 European Multicenter Study on Lead Neurotoxicity in Childen: Implications for Risk
5 Assessment." Neurotoxicol Teratol 12:553-559.
6 Winneke, G., Beginn, U., Ewert, T., et al. (1985) "Comparing the Effects of Perinatal and Later
7 Childhood Lead Exposure on Neurophysiological Outcome." Environmental Research.
8 38:155-167.
9 Wong, M.S. (1988) "The Role of Environmental and Host Behavioral Factors in Determining
10 Exposure to Infection With Ascaris Lumbricoides And Trichuris Trichluta." [Ph.D. thesis],
11 University of the West Indies, Mona, Jamaica.
12 Yule, W., Lansdown, R., Millar, I., and Urbanowiez, M. (1981) "The Relationship Between
13 Blood Lead Concentration, Intelligence, and Attainment in a School Population: a Pilot
14 Study." Dev Med Child Neuro. 23:567-576.
DRAFT - DO NOT CITE OR QUOTE 292 August 28,2000
-------
APPENDIX A
GLOSSARY FOR SECTION 2.1
-------
Sources: The Concise Columbia Encyclopedia. 1995. Columbia University Press; Solomon et al.
1993. Biology, Third Edition. Harcourt Brace Publishing
astrocyte - a star-shaped cell, especially a neuroglial cell of nervous tissue.
axon - the long, tubular extension of the neuron that conducts nerve impulses away from the cell
body.
blood-brain barrier - system of capillaries that regulates the movement of chemical substances,
ions, and fluids in and out of the brain.
central nervous system - the portion of the vertebrate nervous system consisting of the brain
and spinal cord.
cerebellum - the trilobed structure of the brain, lying posterior to the pons and medulla
oblongata and inferior to the occipital lobes of the cerebral hemispheres, that is responsible for
the regulation and coordination of complex voluntary muscular movement as well as the
maintenance of posture and balance.
cerebral cortex - the extensive outer layer of gray matter of the cerebral hemispheres, largely
responsible for higher brain functions, including sensation, voluntary muscle movement, thought,
reasoning, and memory.
cerebrum - the large, rounded structure of the brain occupying most of the cranial cavity,
divided into two cerebral hemispheres that are joined at the bottom by the corpus callosum. It
controls and integrates motor, sensory, and higher mental functions, such as thought, reason,
emotion, and memory.
cognitive development - various mental tasks and processes (e.g. receiving, processing, storing,
and retrieving information) that mediate between stimulus and response and determine problem-
solving ability.
demyelination - to destroy or remove the myelin sheath of (a nerve fiber), as through disease.
dendrite - a branched protoplasmic extension of a nerve cell that conducts impulses from
adjacent cells inward toward the cell body.
EEG (electroencephalogram) - a graphic record of the electrical activity of the brain as recorded
by an electroencephalograph. Also called encephalogram.
ECoG (electrocorticogram) - a graphic record of the electrical activity of the brain; used to
calculate parameters of activity, such as wave amplitude and frequency.
encephalitis - inflammation of the brain.
DRAFT-DO NOT CITE OR QUOTE A-1 August 28,2000
-------
encephalopatby - any of various diseases of the brain.
enzyme - any of numerous proteins or conjugated proteins produced by living organisms and
functioning as biochemical catalysts.
gavage - introducing material directly into the stomach using a tube.
genotoxic - causing chromosomal/genetic aberrations.
glial cells (neuroglia) - the delicate network of branched cells and fibers that supports the tissue
(neurons) of the central nervous system.
gray matter - brownish-gray nerve tissue, especially of the brain and spinal cord, composed of
nerve cell bodies and their dendrites and some supportive tissue.
heme (hematin) - ferrous component of hemoglobin, as well as a functional group in other
hemoproteins involved in various functions throughout the body.
hematological - science encompassing the medical study of the blood and blood-producing
organs.
hepatic - of, relating to, or resembling the liver.
hippocampus - a ridge in the floor of each lateral ventricle of the brain that consists mainly of
gray matter and has a central role in memory processes.
histopathology - the study of the microscopic anatomical changes in diseased tissue.
hormone - a chemical messenger, usually a peptide or steroid, produced by one tissue and
conveyed by the bloodstream to another to effect physiological activity, such as growth or
metabolism.
limbic system - a group of interconnected deep brain structures, common to all mammals, and
involved in olfaction, emotion, motivation, behavior, and various autonomic functions.
microtubules - any of the proteinaceous cylindrical hollow structures that are distributed
throughout the cytoplasm of eukaryotic cells, providing structural support and assisting in
cellular locomotion and transport.
DRAFT - DO NOT CITE OR QUOTE A-2 August 28.2000
-------
mitochondrion (plural mitochondria) - a spherical or elongated organelle in the cytoplasm of
nearly all eukaryotic cells, containing genetic material and many enzymes important for cell
metabolism, including those responsible for the conversion of food to usable energy.
morphology - the form and structure of an organism or one of its parts; without consideration of
function.
mutagenic - inducing or increasing the frequency of mutation in an organism.
myelin sheath - the insulating envelope of myelin that surrounds the core of a nerve fiber or
axon and facilitates the transmission of nerve impulses. In the peripheral nervous system, the
sheath is formed from the cell membrane of the Schwann cell and, in the central nervous system,
from oligodendrocytes. Also called medullary sheath.
necrosis - death of cells or tissues through injury or disease, especially in a localized area of the
body.
nerve - many neurons bound together by connective tissue.
neuroglia - see glial cells.
neuron - cell specialized for the conduction of electrochemical nerve impulses that constitute the
brain, spinal column, and nerves, consisting of a nucleated cell body with one or more dendrites
and a single axon. Also called nerve cell.
neurotransmitter - a chemical substance that transmits information (nerve impulses) across the
junction (synapse) that separates one nerve cell (neuron) from another nerve cell or a muscle.
There are more than 300 known neurotransmitters, including dopamine and glutamine.
parasympathetic nervous system - the part of the autonomic nervous system originating in the
brain stem and the lower part of the spinal cord that, in general, inhibits or opposes the
physiological effects of the sympathetic nervous system, as in tending to stimulate digestive
secretions, slow the heart, constrict the pupils, and dilate blood vessels.
peripheral nervous system - the part of the vertebrate nervous system constituting the nerves
outside the central nervous system and including the cranial nerves, the spinal nerves, and the
sympathetic and parasympathetic nervous systems.
perseveration - uncontrolled, incessantly repetitive behavior, occurring even when it directly
results in rewards being withheld.
renal - of, relating to, or in the region of the kidneys.
somatosensory - of or relating to the perception of sensory stimuli from the skin and internal
organs.
DRAFT - DO NOT CITE OR QUOTE A-3 August 28.2000
-------
sympathetic nervous system - the part of the autonomic nervous system originating in the
thoracic and lumbar regions of the spinal cord that in general inhibits or opposes the
physiological effects of the parasympathetic nervous system, as in tending to reduce digestive
secretions, speeding up the heart, and contracting blood vessels.
synapse - the junction across which a nerve impulse passes from an axon terminal to a neuron, a
muscle cell, or a gland cell.
teratogenic - of, relating to, or causing malformations of an embryo or a fetus.
tubulin - a globular protein that is the basic structural constituent of microtubules.
DRAFT-DO NOT CITE OR QUOTE A-4 August 28.2000
-------
APPENDIX B
CALCULATING AVERAGE IQ DECREMENT
ASSUMING A NON-ZERO THRESHOLD ON THE
IQ/BLOOD-LEAD CONCENTRATION RELATIONSHIP
-------
This appendix is an update to Appendix El of the §403 risk analysis report, which
provided details on how the health effect and blood-lead concentration endpoints are calculated
given that blood-lead concentration is lognormally distributed with a geometric mean and
geometric standard deviation specified by GM and GSD, respectively. In estimating average IQ
decrement due to lead exposure and the percentages of children whose IQ decrement as a result
of lead exposure was at or above 1,2, or 3 points, the §403 risk analysis (as detailed in Appendix
El) assumed an average IQ decrement of 0.257 points for every 1.0 ng/dL increase in blood-lead
concentration, and that no blood-lead threshold existed in this relationship (i.e., no non-zero
blood-lead concentration existed below which the predicted IQ decrement was zero). To
evaluate how the assumption of no threshold affects the estimates of these IQ decrement
parameters, the sensitivity analyses presented within Chapters 5 and 6 of this document includes
analyses that estimate these parameters under specified assumptions on a non-zero threshold
(Sections S.I.4 and 6.2.2). This appendix shows how these estimates were calculated in these
sensitivity analyses (i.e., given a non-zero threshold). (Note that the assumption of a threshold
does not affect how the probability of having a blood-lead concentration at or above a specified
value or the probability of observing an IQ less than 70 due to lead exposure are calculated.)
PriO decrement s x] for x=l. 2.3
Let Y denote the IQ decrement associated with a blood-lead concentration specified by
PbB. Assume that the non-zero blood-lead threshold in the blood-lead/IQ relationship is denoted
byT. Then
Y = 0.257*(PbB - T) when PbB * T
= 0 when PbB < T.
Thus, for any positive value x, the probability of observing an IQ decrement (Y) at or above x is
determined by the following:
P[Y * x] = P[0.257*(PbB-T) ;> x] = P[PbB * (x/0.257 + T)] = Pfln(PbB) * ln(x/0.257 + T)]
where ln(.) denotes the natural logarithm transformation. Then, since PbB is assumed to have a
lognormal distribution,
DRAFT-DO NOT CITE OR QUOTE B-1 August 28,2000
-------
PpQ decrement > x] = 1 -
where 0
F(T) when y = 0
Then, the average IQ decrement, denoted by E[Y], is given by
00 CO <*>
E[Y]= fy-f(y/0.257+T)-(l/0.257)dy=[0.257}x-f(x)dx]-[0.257-T}f(x)dx]
6 T T
This equates to the following:
Avg. IQ decrement = E[Y] =
- ln(GM) - In(GSD)2
In(GSD)
-0.257.[l-*(^GSD)
Note that when T=0, average IQ decrement = 0.257*GM*exp(ln(GSD)2/2), which is equation (4)
DRAFT - DO NOT CITE OR QUOTE B-2 August 28.2000
-------
specified within Appendix El of the §403 risk analysis report.
The standard deviation of the distribution of IQ decrement (Y) equals
S. D. (IQ decrement) = VE(Y2 ) - [EQQ]2
The value of E[Y] is given above, and the value of E(Y2) can be found to equal
E[Y2]= 0.2572 • (exp(2(ln(GM) + In(GSD)2))• [l- $( "(AGgp)—- 21n(GSD))J
- 2T • exp(ln(GM) + In(GSD)2 / 2) [ 1 - 0 ( lM^gm In(GSD)) J
+ T2-[l-
-------
APPENDIX C
METHOD TO IMPUTING HOUSEHOLD AVERAGE
ENVIRONMENTAL-LEAD LEVELS FOR HOUSING UNITS IN THE
NATIONAL SURVEY OF LEAD AND ALLERGENS IN HOUSING (NSLAH)
-------
Method to Imputing Household Average Environmental-lead Levels for
Housing Units in the National Survey of Lead and Allergens in Housing (NSLAH)
Occasionally, some of the 706 housing units included in the interim NSLAH database
had no data available to calculate one or more of the following five environmental-lead
parameters:
• area-weighted household average floor dust-lead loading
• area-weighted household average window sill dust-lead loading
• household average soil-lead concentration at dripline/entryway
• household average soil-lead concentration at mid-yard
• yard-wide average soil-lead concentration (taken to be the average of the previous
two measures, or only one of these two measures if no data exist for the other).
In order to apply the risk analysis to the NSLAH data (specifically, the modeling analysis), it was
necessary to estimate these parameters in situations where their values could not be calculated for
a given housing unit due to a lack of available data (i.e., no floor dust-lead loading data, no
window sill dust-lead loading data, or no soil-lead concentration data). Otherwise, those housing
units having missing data, and the portion of the national housing stock represented by their
sampling weights, could not be represented in the risk analysis. The method of assigning
estimated data values to housing units having missing data is called imputation.
The imputation method applied to the interim NSLAH data was the same method used in
the §403 risk analysis to impute environmental-lead levels for HUD National Survey units. This
method was documented in Section 3.3.1.1 and Appendix C of the §403 risk analysis report.
This method involved the following:
1. Each NSLAH housing unit was placed into one of 15 categories defined by the
combination of five housing age categories (pre-1940,1940-1959,1960-1977,
post-1977, unknown) and three categories determined by whether or not lead-
based paint (LBP, defined as paint with an x-ray fluorescence measurement of at
least 1.0 mg/cm2) was observed in the unit (yes, no, unknown).
2. Within the eight categories in which both the housing age group and the presence
of LBP were known, the weighted averages of the first four environmental-lead
parameters above were calculated across the housing units having nonmissing
data (where the weights corresponded to the interim NSLAH sampling weights).
Then, within a given category, if a housing unit had missing data for one of these
four parameters, the weighted average for that parameter was assigned to the unit.
3. For the category in which both the housing age group and the presence of LBP
were unknown, housing units having missing data for a given parameter among
the first four parameters above were assigned the weighted average for that
parameter calculated across all units in the interim NSLAH database having
nonmissing data for that parameter.
DRAFT - DO NOT CITE OR QUOTE C-1 August 28, 2000
-------
4. For the four categories in which the housing age group was specified but the
presence of LBP was unknown, housing units having missing data for a given
parameter among the first four parameters above were assigned the weighted
average for that parameter calculated across units within the same housing age
group (without regard to the presence of LBP) that had nonmissing data for that
parameter.
5. For the two categories in which the presence of LBP was known but the housing
age group was not specified, housing units having missing data for a given
parameter among the first four parameters above were assigned the weighted
average for that parameter calculated across units having the same indicator of
LBP (without regard to housing age group) that had nonmissing data for that
parameter.
6. If a housing unit had a missing value for yard-wide average soil-lead
concentration (i.e., no soil-lead concentration data for any soil samples), the
parameter's imputed value assigned to this unit was the arithmetic average of the
unit's imputed values for average dripline/entryway soil-lead concentration and
average mid-yard soil-lead concentration. (Note that if soil-lead data existed for
one location but not for the other, the unit's yard-wide average equaled the
average for only the location having soil-lead data.)
Table C-l presents the weighted averages that were assigned to units having missing data as part
of this imputation scheme, according to category. Note that only those weighted averages that
were assigned to at least one housing unit with missing data are displayed in this table. The
numbers in parentheses correspond to the numbers of housing units in the category to which the
given weighted average was assigned. Only 11 of the 15 housing unit categories are included in
Table C-l, as no imputations were necessary in the other four categories.
As indicated in Table C-l, the above imputation procedure was applied twice to the
NSLAH data: once when making no adjustments to not-detected values, and once after replacing
not-detected values with one-half of the detection limit. Both of these scenarios were considered
in the data summaries and risk analysis. In both cases, the imputed values were the same in a
majority of situations, and those differences which did occur between the two cases were minor.
DRAFT - DO NOT CITE OR QUOTE C-2 August 28,2000
-------
Table C-1. Imputed Environmental-Lead Measurements, by Housing Age Category and Presence of Lead-Based Paint (LBP)1,
and Numbers of Units in the Interim NSLAH to Which Imputed Measurements Were Assigned
Household Average
Environmental-Lead
Measurement
Imputed Measurement2
(Number of Interim NSLAH units In which Imputed measurements were assigned)
Pre-1940 Units
LBP
Present
LBP Not
Present
1940-1959 Units
LBP
Present
LBP Not
Present
1960-1977 Units
LBP
Present
LBP Not
Present
Post-1977 Units
LBP
Present
LBP Not
Present
LBP
Presence
Unknown
Units with Housing
Age Unspecified
LBP
Present
LBP Not
Present
No Adjustment Made to Not-Detected Values
Floor Dust-Lead Loading
1/ig/ft2)
Window Sill Dust-Lead
Loading (//g/ft2)
Yard-Wide Average Soil-Lead
Concentration3 (//g/g)
Soil-Lead Concentration at
Dripline/Entryway (//g/g)
Soil-Lead Concentration at
Mid-Yard (//g/g)
35.30
ID
449.06
(3)
710.77
(7)
1094.6
(8)
326.95
(8)
--
15.45
(1)
176.71
(3)
223.48
(5)
129.93
(3)
494
(2)
144.42
(4)
276.07
(4)
399.75
(6)
152.39
(7)
-
94.66
(4)
242.58
(3)
344.61
(3)
140.55
(31
-
~
161.91
(3)
245.35
(3)
78.47
(4)
1.24
(3)
28.95
(12)
52.33
(5)
64.45
(8)
40.20
(8)
-
28.81
(1)
--
-
-
1.18
(1)
13.99
(12)
24.85
(7)
27.15
(8)
22.56
(11)
1.20
(1)
15.62
(1)
27.78
(1)
31.88
(1)
23.68
(1)
21.20
(1)
285.64
(1)
392.05
(5)
591.39
(5)
192.71
(5)
-
32.73
(2)
63.84
(4)
80.84
14)
46.84
(5)
Not-Detected Values Replaced by LOD/2 (I.e.. one-half of the detection limit)
Floor Dust-Lead Loading
(//g/ft2)
Window Sill Dust-Lead
Loading |//g/ft2)
Yard-Wide Average Soil-Lead
Concentration (//g/g)
Soil-Lead Concentration at
Dripline/Entryway (//g/g)
Soil-Lead Concentration at
Mid- Yard (//g/g)
35.47
- (D
449.10
(3)
710.82
(7)
1094.6
(8)
327.01
(8)
--
15.81
(1)
176.62
(3)
223.48
15)
129.75
(3)
5.19
(2)
144.76
(4)
276.10
(4)
399.76
(6)
152.45
(7)
~
94.88
(4)
242.76
(3)
344.66
(3)
140.86
(3)
-
~
162.07
(3)
245.47
(3)
78.67
(4)
1.72
(3)
29.28
(12)
52.86
(5)
64.85
(8)
40.87
(8)
-•
28.90
(1)
—
--
-
1.71
(1)
14.43
(12)
25.73
(7)
27.86
(8)
23.60
(11)
1.71
(1)
16.02
(1)
28.57
(1)
32.52
(1)
24.63
(1)
21.45
ID
285.81
(1)
392.15
15)
591.46
(5)
192.84
(5)
--
33.09
(2)
64.43
(4)
81.30
(4)
47.56
(5)
s
ISJ
co
ro
' Units with lead-based paint have a maximum observed XRF reading of at least 1 0 mg/cm1 on interior or exterior painted surfaces.
1 See text for details on method of determining imputed measurements.
1 Imputed only when unit has no soil-lead data for either dripline/entryway or mid-yard.
-------
DRAFT - DO NOT CITE OR QUOTE C-4 August 28,2000
-------
APPENDIX D1
SUMMARIES OF INTERIM DUST-LEAD LOADING DATA
FROM THE NATIONAL SURVEY OF LEAD AND ALLERGENS IN HOUSING,
(NSLAH), WHERE IMPUTED DATA ARE EXCLUDED
-------
Summaries of Interim Dust-Lead Loading Data
from the National Survey of Lead and Allergens in Housing (NSLAH),
Where Imputed Data Are Excluded
This appendix presents descriptive statistics of average household dust-lead loadings for
floors and window sills from the §403 risk analysis and from the interim NSLAH dust-lead
loading data where imputed data values calculated based on the methods presented in Appendix
C are omitted. These summaries complement the summary tables and boxplots presented in
Tables 3-4 through 3-1 Ib and Figures 3-1 through 3-6 in the main body of this report, which
included imputed household averages for housing units having no dust-lead loading data.
The statistics on the interim NSLAH data are provided in this appendix under five
different approaches to handling sample results that fall below the instrument's detection limit.
As noted in Table 3-1, the interim NSLAH database reported dust-lead amounts as they were
measured by the analytical instruments, regardless of whether these amounts were below the
instrument's detection limit. While using these actual reported lead amounts rather than a
censored result based on the detection limit can lead to more accurate portrayals of the actual
lead amounts in the samples, some of these reported amounts are zero or below. This can cause
problems in the risk analysis, as the empirical model takes natural logarithms of the household
averages, and logarithms can only be taken on positive values. Therefore, the descriptive
statistics of the interim NSLAH data are presented in this appendix under five approaches to
handling not-detected values associated with individual sample analyses:
No adjustment (i.e., using data as reported in the database)
Replacing the value with zero
Replacing the value with the detection limit (LOD) divided by two
Replacing the value with the detection limit divided by the square root of two
Replacing the value with the detection limit
Replacement with zero introduces the greatest amount of negative bias (i.e., underestimation),
while replacement with the detection limit introduces the greatest amount of positive bias. The
detection limit divided by the square root of two is an efficient estimator of the true amount when
the data are lognormally distributed, while the detection limit divided by two is recommended
when the distribution is highly skewed. Results are presented under these different approaches to
illustrate the impact that any one approach has on the characterized distribution.
The following tables appearing in this appendix are associated with the specified tables in
Chapter 3 of the report:
• Tables Dl-1 andDl-2: national estimates complementing Tables 3-4 and 3-5
DRAFT - DO NOT CITE OR QUOTE Dl-1 August 28.2000
-------
• Tables D1-3 and Dl-4: estimates by housing age category, complementing
Tables 3-6 and 3-7
• Tables D1-5 and Dl-6: estimates by Census region, complementing Tables 3-8
and 3-9
• Tables Dl-7a through Dl-8b: estimates by combinations of Census region and
housing age category, complementing Tables 3-10a through 3-1 Ib.
The following boxplots appearing in this appendix are associated with the specified boxplots in
Chapter 3 of the report:
• Figures Dl-1 and Dl-2: national estimates complementing Figures 3-1 and 3-2
• Figures Dl-3 and Dl-4: estimates by housing age category, complementing
Figures 3-3 and 3-4
• Figures Dl-5 and Dl-6: estimates by Census region, complementing Figures 3-5
and 3-6.
While Tables Dl-1 through Dl-4 and Figures Dl-1 through Dl-2 contain interim NSLAH data
summaries under all five approaches to handling not-detected values, the remaining tables and
figures in this appendix present interim NSLAH data summaries only for the two approaches (no
adjustment; replace by one-half of the level of detection) most likely to be used in the
supplemental risk analysis and considered in the interim NSLAH data summaries presented in
Chapter 3.
DRAFT - DO NOT CITE OR QUOTE D1-2 August 28.2000
-------
Table D1-1. Descriptive Statistics of Area-Weighted Average Floor Wipe Dust-Lead
Loadings for Households, As Reported in the §403 Risk Analysis Versus the
Interim NSLAH Data (imputed data omitted for the NSLAH)
Study
How Not-
D0t8ctod
and
Negative
Data were
Handled
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced
byO
Replaced
by LOD/2
Replaced
by LODA/2
Replaced
by LOD
Area-Weighted Average Floor Dust-Lead Loading (//g/ftY
#
Surveyed
Units with
Positive
Averages
284
624
417
697
697
697
Arith-
metic
Mean
16.5
10.4
10.1
10.8
11.1
11.4
Geo-
metric
Mean2
6.27
1.21
1.95
1.80
2.21
2.73
Geo-
metric
CtH
Dev.2
3.49
4.56
3.89
2.76
2.50
2.29
Minimum
0.508
-1.23
0.00
0.750
1.06
1.50
25"1
Percen-
tile
2.65
0.300
0.00
0.950
1.25
1.60
Median
5.32
1.03
0.500
1.31
1.68
2.10
75"1
Percen-
tile
12.2
2.30
2.00
2.46
2.84
3.20
Maximum
375
5940
5940
5950
5950
5950
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with floor dust-lead data are
used to calculate the remaining statistics).
DRAFT - DO NOT CITE OR QUOTE
D1-3
August 28.2000
-------
Table D1-2. Descriptive Statistics of Area-Weighted Average Window Sill Wipe Dust-Lead
Loadings for Households, As Reported in the §403 Risk Analysis Versus the
Interim NSLAH Data (imputed data omitted for the NSLAH)
Study
How Not-
Detected
and
Negative
Data were
Handled
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced
byO
Replaced
by LOD/2
Replaced
by LODA/2
Replaced
by LOD
Area-Weighted Average Window Sill Dust-Lead Loading U/g/ft2)1
#
Surveyed
Units with
Positive
Averages
284
649
563
665
665
665
Arith-
metic
Mean
550
140
139
140
141
141
Geo-
metric
Mean2
23.0
13.6
20.2
14.9
16.2
17.6
Geo-
metric
Std.
Dev.2
15.8
8.05
6.72
6.71
6.22
5.77
Minimum
0.0118
-9.43
0.00
0.445
0.629
0.889
25m
Percen-
tile
4.35
2.71
1.94
3.09
3.75
4.39
Median
19.5
11.0
10.8
11.1
11.6
12.1
75*
Percen-
tile
198
50.3
50.1
50.1
50.3
50.3
Maximum
43700
11100
11100
11100
11100
11100
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with window sill dust-lead
data are used to calculate the remaining statistics).
DRAFT - DO NOT CITE OR QUOTE
D1-4
August 28.2000
-------
D
O
Z
O
o
o
3
D
o
3
D
—»
i
01
0
HUONS
(403)
NSLAH
(0)
NSLAH
(LOO/2)
NSLAH
(LOO/^Z)
NSLAH
(LOO)
CO
c
ro
CO
Figure D1-1. Boxplots of Area-Weighted Average Floor Wipe Dust-Lead Loadings (pg/ft2) As Observed in the §403 Risk
Analysis (Using HUD National Survey Data) and in the NSLAH (under 5 approaches to handling not-detected
values) (imputed data omitted for the NSLAH)
g (Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using the methods documented in the
§ §403 risk analysis report. See text for definitions of labels along the horizontal axis.)
-------
o
o
o
^ 100000 -
m
0
33
o
Q IOOOO -
0
-1 c-
m
^
1000 -
•0
g 100 -
£
2 i
° i '°~
1 -
O.I -
O.OI -
•
: : t t s
• • • • • •
• «
i :
i
» t
<
4
t 1
1
1
» 1
•
•
•
•
1
1 I
HUDNS
(40J)
NSLAH
(O)
NSLAH
(LOO/2)
NSLAH
(LOD/%/2)
NSLAH
(LOO)
Figure D1-2. Boxplots of Area-Weighted Average Window Sill Wipe Dust-Lead Loadings (ug/ft2) As Observed in the §403
Risk Analysis (Using HUD National Survey Data) and in the NSLAH (under 5 approaches to handling not-
detected values) (imputed data omitted for the NSLAH)
(Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using the methods documented in the
§403 risk analysis report. See text for definitions of labels along the horizontal axis.)
-------
Table D1-3. Descriptive Statistics of Area-Weighted Average Floor Wine Dust-Lead
Loadings for Households, Presented bv Housing Aoe Category. As Reported
in the §403 Risk Analysis Versus the Interim NSLAH Data (imputed data
omitted for the NSLAH)
Study
How Not-
Detected
and
Negative
Data were
Handled
Area-Weighted Average Roor Dust-Lead Loading (pg/ft2)1
# Units
with
Positive
Averages
Arith-
metic
Mean
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
251"
Percen-
tile
Units Built Prior to 1940
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LOD/^2
Replaced by
LOD
77
110
97
113
113
113
47.9
36.9
36.6
37.0
37.2
37.5
22.6
3.66
4.12
3.92
4.36
4.89
3.63
4.49
4.64
3.94
3.62
3.34
0.991
-0.600
0.00
0.750
1.06
1.50
8.84
1.30
0.750
1.45
1.68
2.00
Units Built from 1940 - 1959
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LODA/2
Replaced by
LOD
87
132
96
143
143
143
18.1
4.10
3.75
4.37
4.63
4.99
8.74
1.88
2.38
2.29
2.70
3.22
3.34
3.58
3.33
2.64
2.37
2.15
0.508
-0.720
0.00
0.750
1.06
1.50
4.07
0.719
0.00
1.05
1.37
1.77
Median
17.7
2.42
2.20
2.71
3.05
3.40
7.81
1.77
1.40
1.98
2.22
2.52
75th
Percen-
tile
79.7
9.25
9.25
9.25
9.27
9.38
22.4
3.66
3.40
3.55
3.92
4.83
Maximum
375
5940
5940
5950
5950
5950
171
71.0
71.0
71.0
71.0
71.0
Units Built from 1960-1977 (1960 • 1979 for the §403 risk analysis)
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LOD/v/2
Replaced by
LOD
120
173
107
198
198
198
6.74
1.51
1.20
1.96
2.28
2.73
4.14
0.905
1.32
1.45
1.83
2.32
2.45
3.52
2.69
1.94
1.76
1.63
0.657
-0.733
0.00
0.750
1.06
1.50
2.25
0.206
0.00
0.900
1.24
1.60
3.62
0.880
0.400
1.20
1.53
1.98
7.59
1.70
1.38
1.94
2.19
2.76
106
28.5
28.6
28.8
28.8
28.9
DRAFT - DO NOT CITE OR QUOTE
D1-7
August 28,2000
-------
Table D1-3. (cont.)
Study
How Not-
Dataeted
and
Negative
Data were
Handled
Area-Weighted Average Floor Dust-Lead Loading (f/g/ftT
# Units
with
Positive
Arith-
metic
Mean
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
2501
Percen-
tile
Median
75"1
Percen-
tile
Maximum
Units Built After 1977 (after 1979 for the §403 risk analysis)
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
Replaced by
0
Replaced by
LOD/2
Replaced by
LODA/2
Replaced by
LOD
28
149
72
178
178
178
4.16
1.20
0.949
1.71
2.03
2.47
3.14
0.542
0.959
1.14
1.49
1.96
2.06
3.35
2.53
1.72
1.59
1.50
1.06
-1.05
0.00
0.750
1.06
1.50
1.76
0.146
0.00
0.750
1.06
1.50
2.84
0.400
0.00
1.00
1.34
1.70
5.66
1.07
0.500
1.35
1.72
2.25
12.9
265
265
265
265
265
NSLAH Units with Unspecified Year-Built Indicator
Interim
NSLAH
No
Replaced by
0
Replaced by
LOD/2
Replaced by
LODA/2
Replaced by
LOD
60
45
65
65
65
31.9
31.7
32.3
32.6
32.9
1.30
2.17
2.11
2.53
3.08
6.49
5.44
3.82
3.51
3.24
-1.23
0.00
0.750
1.06
1.50
0.300
0.00
1.00
1.38
1.70
1.24
0.660
1.40
1.84
2.22
2.50
2.20
2.53
2.75
3.10
1040
1040
1040
1040
1040
1 All statistics are calculated by weighting each household by its sampling weight.
1 Only household averages greater than zero are used to calculate this value (data for all units with floor dust-lead data are
used to calculate the remaining statistics).
DRAFT - DO NOT CITE OR QUOTE
D1-8
August 28, 2000
-------
Table D1-4. Descriptive Statistics of Area-Weighted Average Window Sill Wipe Dust-Lead
Loadings for Households, Presented bv Housing Aae Category. As Reported
in the §403 Risk Analysis Versus the Interim NSLAH Data (imputed data
omitted for the NSLAH)
Study
How Not-
Detected
and
Negative
Data were
Handled
Area-Weighted Average Window Sill Dust-Lead Loading (//g/ft2)1
# Units
with
Positive
Averages
Arith-
metic
Mean
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
25*
Percen-
tile
Median
75*
Percen-
tile
Maximum
Units Built Prior to 1940
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LOD/,/2
Replaced by
LOD
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LOD/,/2
Replaced by
LOD
77
109
107
110
110
110
2060
400
400
400
400
400
168
72.9
76.3
72.2
73.3
74.7
16.7
6.62
6.35
6.47
6.30
6.12
0.0155
-0.152
0.00
1.03
1.46
2.06
35.6
21.1
21.1
21.1
21.1
21.1
198
78.2
78.2
78.2
78.2
78.2
1220
284
284
284
284
284
43700
11100
11100
11100
11100
11100
Units Built from 1940 - 1959
87
136
122
137
137
137
285
130
129
130
130
131
22.0
22.7
30.3
24.2
25.7
27.5
10.7
6.91
5.90
604
5.64
5.27
Units Built from 1960-1977 (1960 - 1979
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LOD/v/2
Replaced by
LOD
120
183
163
189
189
189
184
37.3
36.3
37.6
38.1
38.8
16.2
9.78
12.1
10.4
11.2
12.3
14.6
4.89
4.47
4.31
4.05
3.82
0.0118
-1.73
0.00
0.923
1.31
1.66
6.47
6.35
5.53
6.10
6.48
7.56
19.1
21.0
19.5
21.5
21.7
21.9
107
69.1
68.4
69.6
70.1
70.9
16100
3630
3630
3630
3630
3630
for the 5403 risk analysis)
0.0164
-2.32
0.00
1.02
1.36
1.47
2.05
2.82
2.07
3.06
3.60
4.20
16.6
8.03
6.95
7.86
8.29
8.83
217
25.4
21.5
26.4
26.5
27.5
5790
1390
1390
1390
1390
1390
DRAFT - DO NOT CITE OR QUOTE
D1-9
August 28,2000
-------
Table D1-4. (cont.)
Study
How Not-
Detected
and
Negative
Data were
Handled
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LODA/2
Replaced by
LOD
Area-Weighted Average Window Sill Dust-Lead Loading (f/g/ft2)1
# Units
with
Positive
Averages
Units
28
160
115
166
166
166
Arith-
metic
Mean
Built After
83.0
15.6
14.8
16.0
16.5
17.3
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
25"1
Percen-
tile
Median
7501
Percen-
tile
Maximum
1977 (after 1979 for the §403 risk analysis)
8.17
3.26
5.40
4.25
4.95
5.83
9.94
5.32
4.38
3.80
3.50
3.25
0.0164
-9.43
0.00
0.445
0.629
0.889
2.58
0.916
0.00
1.69
2.07
2.61
8.11
2.80
1.71
3.33
4.01
4.80
57.8
8.17
7.29
8.50
9.48
10.0
1590
426
409
427
434
445
NSLAH Units with Unspecified Year-Built Indicator
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LODA/2
Replaced by
LOD
61
56
63
63
63
379
379
379
379
380
38.5
54.2
38.9
40.4
42.1
7.55
5.45
6.91
6.53
6.19
-0.629
0.00
0.720
1.02
1.44
14.3
14.3
17.7
18.8
18.8
36.4
36.4
36.4
36.4
36.4
116
116
116
116
116
9030
9030
9030
9030
9030
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with window sill dust-lead
data are used to calculate the remaining statistics).
DRAFT - DO NOT CITE OR QUOTE
D1-10
August 28, 2000
-------
o
o
I
o
m
O
D
C
o
m
^
S
I
I •
I 1
HUONS
4O3
NSLAH
LOO/2
HUONS
4O3
NSLAH
LOO/2
Prior to 1 940
1940 - 1999
HUONS NSLAH NSLAH
403 LOO/2
I960 - 1977(79)
HUONS NSLAH NSLAH
4O3 LOO/2
Aft«r 1977(79)
>LAH NSLAH
LOO/2
Un.p.cld.d
ID
C
-
Figure D1-3. Boxplots of Area-Weighted Average Floor Wipe Dust-Lead Loadings (ug/ft2), by Housing Age Category, As
Observed in the §403 Risk Analysis (Using HUD National Survey Data) and in the NSLAH (under 2 approaches
to handling not-detected values) (imputed data omitted for the NSLAH)
(Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using the methods documented in the
§403 risk analysis report. See text for definitions of labels along the horizontal axis.)
-------
o
o
H
O
m
O
m
1000 -
1
g too -
J_
85 10 -
£
•o
1 -
O.I -
0.01 -
1
r
i
r
• • «
I i
L •
]
«
]
4
[ -
-
. :
. .
1
i
[ '
1 •
1 1
I
1 : i
[
i i t
i
|
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
HUONS NSLAH NSLAH HUONS NSLAH NSLAH HUONS NSLAH NSLAH HUONS NSLAH NSLAH NSLAH NSLAH
403 LOO/2 403 LOO/2 403 LOO/2 4O3 LOO/2 LOO/2
Prtor to 1940 194O— 1939
1*80-1977(79) Aft«r 1977(79) Unknown
,, Figure D1-4. Boxplots of Area-Weighted Average Window Sill Wipe Dust-Lead Loadings (ug/ft2), by Housing Age Category,
As Observed in the §403 Risk Analysis (Using HUD National Survey Data) and in the NSLAH (under 2
approaches to handling not-detected values) (imputed data omitted for the NSLAH)
10
I
(Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using the methods documented in the
§403 risk analysis report. See text for definitions of labels along the horizontal axis.)
-------
Table D1-5. Descriptive Statistics of Area-Weighted Average Floor Wine Dust-Lead
Loadings for Households, Presented bv Census Region. As Reported in the
§403 Risk Analysis Versus the Interim NSLAH Data (imputed data omitted
for the NSLAH)
Study
How Not-
and
Negative
Data were
Handled
Area-Weighted Average Floor Dust-Lead Loading (pg/ft2)1
#
Surveyed
Units with
Positive
Arith-
metic
Mean
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
25*
Percen-
tile
Median
75"
Percen-
tile
Maximum
Northeast
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
Replaced by
LOD/2
53
103
109
35.6
10.0
10.3
14.9
2.28
2.90
3.95
4.42
3.15
0.632
-0.620
0.750
4.79
0.800
1.20
11.0
1.90
2.13
76.3
6.00
6.00
375
617
617
Midwest
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
Replaced by
LOD/2
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
LOD/2
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
Replaced by
LOD/2
73
135
149
134
230
260
52
156
179
14.7
14.6
14.9
13.3
2.58
3.00
9.81
19.0
19.5
6.32
1.31
2.00
5.01
0.962
1.53
4.97
0.927
1.44
3.26
5.74
3.34
South
3.28
3.92
2.22
West
2.75
3.68
2.31
0.508
-0.733
0.750
0.735
-1.05
0.750
1.06
-1.23
0.750
2.83
0.283
0.760
2.00
0.253
0.970
2.65
0.250
0.780
6.32
1.16
1.29
3.89
0.900
1.20
4.01
0.760
1.20
11.0
2.48
3.15
10.0
1.76
1.89
8.43
1.62
1.88
173
1040
1040
236
265
265
197
5940
5950
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with floor dust-lead data are
used to calculate the remaining statistics).
DRAFT - DO NOT CITE OR QUOTE
D1-13
August 28,2000
-------
Table D1-6. Descriptive Statistics of Area-Weighted Average Window Sill Wipe Dust-Lead
Loadings for Households, Presented bv Census Region. As Reported in the
§403 Risk Analysis Versus the Interim NSLAH Data (imputed data omitted
for the NSLAH)
Study
How Not-
Detected
and
Negative
Data were
Handled
Area-Weighted Average Window Sill Dust-Lead Loading (i/g/ft2)1
#
Surveyed
Units with
Positive
Averages
Arith-
metic
Mean
Geo-
metric
Mean2
Geo-
metric
Std.
Dew.2
Minimum
25m
Percen-
tile
Median
75"
Percen-
tile
Maximum
Northeast
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
LOD/2
53
106
108
1440
170
170
92.2
21.0
22.1
16.1
7.93
6.99
0.0155
-1.89
0.578
15.3
5.94
5.94
173
14.6
14.8
335
89.5
90.0
14600
5530
5530
Midwest
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
LOD/2
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
LOD/2
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
LOD/2
73
143
148
134
231
237
52
169
172
564
216
216
432
121
121
62.2
55.3
55.3
48.5
19.9
20.5
19.6
12.4
14.2
4.45
6.96
7.93
13.2
7.13
6.37
South
12.4
8.68
6.77
West
12.7
6.80
5.68
0.0706
-2.32
1.12
0.118
-9.43
0.646
0.0118
-0.115
0.445
7.76
4.00
4.67
4.60
2.33
2.88
1.68
1.74
2.18
83.0
16.0
15.7
15.0
10.2
10.3
5.40
6.08
6.26
309
54.9
56.1
127
53.8
53.8
28.0
25.6
25.5
43700
9630
9630
28400
11100
-11100
1400
3630
3630
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with window sill dust-lead
data are used to calculate the remaining statistics).
DRAFT - DO NOT CITE OR QUOTE
01-14
August 28.2000
-------
8
I
o
m
o
71
O
§
m
100000 -
o
tri
J
T3
g
8
£Z
I i
HUONS
403
NSLAH
LOO/2
HUONS
403
NSLAH
LOO/2
HUONS
403
NSLAH
LOO/2
HUONS
403
NSLAH
LOO/2
ID
c
to
00
Figure D1-5. Boxplots of Area-Weighted Average Floor Wipe Dust-Lead Loadings (ug/ft2), by Census Region, Observed in
the §403 Risk Analysis (Using HUD National Survey Data) and in the NSLAH (under 2 approaches to handling
not-detected values) (imputed data omitted for the NSLAH)
(Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using the methods documented in the
§403 risk analysis report. See text for definitions of labels along the horizontal axis.)
-------
8
m
O
S;
5
•— 1000 -
I
1
g 100 -
0 1
^ «" 10-
0) J
1
0.1
O.O1
: ' :
• i : • . ...
II « ! i '
]
•
i
•
•
.
.L
: :
i )
! i
1
i
I ;
i 1
i
i
•
i 1
i 1 —
HUDNS NSLAH NSLAH
403 LOO/2
North.o.t
HUONS NSLAH NSLAH
4O3 LOO/2
MMwnt
NSLAH
South
NSLAH
LOO/2
HUONS NSLAH NSLAH
403 LOO/2
Wml
Figure D1-6. Boxplots of Area-Weighted Average Window Sill Wipe Dust-Lead Loadings (ug/ft2), by Census Region, As
£ Observed in the §403 Risk Analysis (Using HUD National Survey Data) and in the NSLAH (under 2 approaches
i. to handling not-detected values) (imputed data omitted for the NSLAH)
po
g (Note: Dust-lead loadings from the HUD National Survey have been converted to wipe-equivalents in the §403 risk analysis using the methods documented in the
S §403 risk analysis report. See text for definitions of labels along the horizontal axis.)
-------
Table D1-7a. Descriptive Statistics of Area-Weighted Average Floor Wipe Dust-Lead
Loadings for Households, Presented bv Housing Aae and Census Region. As
Reported in the §403 Risk Analysis Versus the Interim NSLAH Data Where
No Adjustments Were Made to Not-Detected Results (imputed data omitted
for the NSLAH)
Census
Region
Northeast
Midwest
South
West
Study
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Housing Age
Category
Prior to 1940
1940-1959
1960-1977
(1960-79 for §403)
After 1977
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
( 1979 for §403)
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Area-Weighted Average Floor Dust-Lead Loading (jig/ft2)
#
Surveyed
Units
26
41
17
21
10
19
15
19
32
21
35
29
32
4
25
19
26
33
42
64
69
18
70
13
11
16
34
17
53
6
39
Arithmetic
Mean
63.5
23.7
13.2
3.75
7.00
3.34
1.12
31.3
7.78
15.8
5.48
6.33
1.52
3.32
0.913
50.7
11.0
25.4
3.66
8.06
1.16
4.19
1.04
34.9
264
14.6
2.73
4.50
1.16
4.60
1.75
Geometric
Mean
36.5
5.02
8.84
2.37
4.73
1.72
0.714
14.7
2.42
6.69
2.05
4.58
0.737
2.77
0.545
20.8
3.66
10.3
1.63
4.13
0.814
3.16
0.543
16.2
3.84
9.04
1.59
3.53
0.937
3.36
0.454
Geometric
Std. Dev.
3.39
4.31
2.54
3.36
2.23
3.76
2.78
3.01
4.26
3.95
4.16
2.35
4.77
1.83
3.86
4.01
3.93
3.91
3.40
2.74
3.09
2.05
3.13
3.51
6.17
2.46
2.91
2.03
2.46
2.21
3.67
Median
76.3
4.20
7.81
2.38
4.76
1.46
0.867
8.94
1.97
5.79
1.59
4.44
1.12
2.80
0.320
19.0
2.74
10.0
1.77
3.39
0.880
2.84
0.480
17.2
2.30
7.47
1.24
3.35
0.880
3.00
0.270
DRAFT - DO NOT CITE OR QUOTE
D1-17
August 28, 2000
-------
Table D1-7b. Descriptive Statistics of Area-Weighted Average Floor Wipe Dust-Lead
Loadings for Households, Presented bv Housing Age and Census Region. As
Reported in the §403 Risk Analysis Versus the Interim NSLAH Data Where
Not-Detected Results Were Replaced bv LOD/2 (imputed data omitted for the
NSLAH)
Census
Region
Northeast
Midwest
South
West
Study
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Housing Age
Category
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1940
1940- 1959
1960-1977
( 1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1940
1940-1959
1960-1977
( 1960-79 for §403)
After 1977
(1979 for §403)
Area-Weighted Average Floor Dust-Lead Loading (//g/ft2)
#
Surveyed
Units
26
41
17
23
10
21
16
19
35
21
36
29
37
4
30
19
26
33
48
64
79
18
82
13
11
16
36
17
61
6
50
Arithmetic
Mean
63.5
23.8
13.2
4.03
7.00
3.58
1.68
31.3
8.09
15.8
5.80
6.33
2.00
3.32
1.31
50.7
11.1
25.4
3.94
8.06
1.67
4.19
1.54
34.9
264
14.6
2.94
4.50
1.62
4.60
2.34
Geometric
Mean
36.5
5.47
8.84
2.86
4.73
2.16
1.43
14.7
2.70
6.69
2.57
4.58
1.50
2.77
1.09
20.8
3.87
10.3
1.99
4.13
1.30
3.16
1.13
16.2
4.03
9.04
1.88
3.53
1.39
3.36
1.07
Geometric
Std. Dev.
3.39
3.91
2.54
2.23
2.23
2.60
1.72
3.01
3.23
3.95
3.20
2.35
2.03
1.83
1.67
4.01
3.76
3.91
2.35
2.74
1.74
2.05
1.57
3.51
5.91
2.46
2.32
2.03
1.66
2.21
1.95
Median
76.3
4.35
7.81
2.40
4.76
1.68
1.29
8.94
2.19
5.79
1.53
4.44
1.20
2.80
0.938
19.0
2.70
10.0
1.54
3.39
1.16
2.84
1.06
17.2
2.19
7.47
1.38
3.35
1.26
3.00
0.900
DRAFT - DO NOT CITE OR QUOTE
D1-18
August 28. 2000
-------
Table D1-8a. Descriptive Statistics of Area-Weighted Average Window Sill Wipe Dust-Lead
Loadings for Households, Presented bv Housing Aoe and Census Region. As
Reported in the §403 Risk Analysis Versus the Interim NSLAH Data Where
No Adjustments Were Made to Not-Detected Results (imputed data omitted
for the NSLAH)
Census
Region
Northeast
Midwest
South
West
Study
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Housing Age
Category
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
( 1979 for §403)
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Area-Weighted Average Window Sill Dust-Lead Loading (/sg/ftz)
#
Surveyed
Units
26
39
17
23
10
20
16
19
35
21
34
29
33
4
30
19
25
33
43
64
73
18
68
13
10
16
36
17
57
6
46
Arithmetic
Mean
2700
395
98.5
62.7
499
13.9
18.3
1660
355
98.2
103
223
27.9
62.5
21.0
2450
606
657
164
149
59.1
112
18.4
125
49.5
107
188
58.7
25.7
9.66
5.21
Geometric
Mean
265
95.9
32.6
20.1
38.9
7.88
3.28
435
64.3
17.7
18.9
20.9
9.94
27.5
6.57
64.0
105
38.9
27.1
24.0
12.9
9.09
3.37
11.5
14.2
7.35
26.3
3.83
7.00
2.65
1.79
Geometric
Std. Dev.
6.37
5.55
4.31
20.8
2.67
5.69
5.79
6.13
11.6
6.38
11.6
4.75
6.78
3.64
5.95
9.93
9.13
12.6
5.98
8.60
6.20
5.44
13.2
7.34
11.5
4.25
11.6
3.92
Median
91.7
50.7
18.5
217
6.49
2.06
60.1
17.4
16.0
48.3
9.54
83.0
5.86
115
26.2
27.3
32.0
10.3
7.58
3.62
17.1
6.96
33.4
4.35
4.74
5.94
1.39
DRAFT - DO NOT CITE OR QUOTE
01-19
August 28.2000
-------
Table D1-8b. Descriptive Statistics of Area-Weighted Average Window Sill Wine Dust-Lead
Loadings for Households, Presented bv Housing Age and Census Region. As
Reported in the §403 Risk Analysis Versus the Interim NSLAH Data Where
Not-Detected Results Were Replaced bv LOD/2 (imputed data omitted for the
NSLAH)
Census
Region
Northeast
Midwest
South
West
Study
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Housing Age
Category
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1 977
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1 977
(1979 for §403)
Prior to 1940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
( 1979 for §403)
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1 977
(1979 for §403)
Area-Weighted Average Window Sill Dust-Lead Loading b/g/ft2)
tt
Surveyed
Units
26
40
17
23
10
21
16
19
35
21
35
29
37
4
30
19
25
33
43
64
74
18
72
13
10
16
36
17
57
6
48
Arithmetic
Mean
2700
395
98.5
62.7
499
14.7
18.6
1660
355
98.2
104
223
28.4
62.5
21.4
2450
606
657
165
149
59.4
112
19.0
125
49.8
107
188
58.7
25.5
9.66
5.32
Geometric
Mean
265
86.8
32.6
19.6
38.9
8.39
4.80
435
67.3
17.7
19.9
20.9
10.3
27.5
7.01
64.0
105
38.9
31.8
24.0
13.9
9.09
4.63
11.5
15.9
7.35
27.9
3.83
7.39
2.65
2.35
Geometric
Std. Oev.
15.8
6.95
5.55
4.49
20.8
2.55
3.80
5.79
5.61
11.6
5.51
11.6
3.81
6.78
3.54
23.1
5.94
9.93
7.16
12.6
5.32
8.60
3.93
14.7
4.41
13.2
6.61
11.5
3.92
11.6
3.01
Median
176
91.7
50.7
18.9
217
7.37
3.73
542
60.1
17.4
15.7
48.3
9.54
83.0
6.20
24.4
115
26.2
27.3
32.0
12.6
7.58
3.62
7.05
17.2
6.96
33.3
4.35
6.26
5.94
1.68
DRAFT - DO NOT CITE OR QUOTE
D1-20
August 28.2000
-------
APPENDIX D2
SUMMARIES OF INTERIM YARD-WIDE AVERAGE SOIL-LEAD
CONCENTRATION DATA FROM THE NATIONAL SURVEY OF
LEAD AND ALLERGENS IN HOUSING (NSLAH),
WHERE IMPUTED DATA ARE EXCLUDED
-------
Summaries of Interim Yard-Wide Average Soil-Lead Concentration Data
from the National Survey of Lead and Allergens in Housing (NSLAH),
Where Imputed Data Are Excluded
This appendix presents descriptive statistics of yard-wide average soil-lead concentration
from the §403 risk analysis and from the interim NSLAH dust-lead loading data where imputed
data values calculated based on the methods presented in Appendix C are omitted. These
summaries complement the summary tables and boxplots presented in Tables 3-18 through 3-2Ib
and Figures 3-12 through 3-14 in the main body of this report, which included imputed
household averages for housing units having no soil-lead concentration data from anywhere in
the yard.
As in Appendix Dl, the statistics on the interim NSLAH data are provided in this
appendix under the following five different approaches to handling sample results that fall below
the instrument's detection limit.
No adjustment (i.e., using data as reported in the database)
Replacing the value with zero
Replacing the value with the detection limit (LOD) divided by two
Replacing the value with the detection limit divided by the square root of two
Replacing the value with the detection limit
(See Appendix Dl for details.) Results are presented under these different approaches to
illustrate the impact that any one approach has on the characterized distribution.
The following tables appearing in this appendix are associated with the specified tables in
Chapter 3 of the report:
• Table D2-1: national estimates complementing Table 3-18
• Table D2-2: estimates by housing age category, complementing Table 3-19
• Table D2-3: estimates by Census region, complementing Table 3-20
• Tables D2-4a and D2-4b: estimates by combinations of Census region and
housing age category, complementing Tables 3-2la and 3-2Ib.
The following boxplots appearing in this appendix are associated with the specified boxplots in
Chapter 3 of the report:
• Figure D2-1: national estimates complementing Figure 3-12
• Figure D2-2: estimates by housing age category, complementing Figure 3-13
• Figure D2-3: estimates by Census region, complementing Figure 3-14.
While Tables D2-1 and D2-2 and Figure D2-1 contain interim NSLAH data summaries under all
five approaches to handling not-detected values, the remaining tables and figures in this appendix
present interim NSLAH data summaries only for the two approaches (no adjustment; replace by
DRAFT - DO NOT CITE OR QUOTE D2-1 August 28.2000
-------
one-half of the level of detection) most likely to be used in the supplemental risk analysis and
considered in the interim NSLAH data summaries presented in Chapter 3.
Table D2-1. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentrations for
Households, As Reported in the §403 Risk Analysis Versus the Interim
NSLAH Data (imputed data omitted for the NSLAH)
Study
How Not-
Detected
and
Negative
Data were
Handled
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced
byO
Replaced
by LOD/2
Replaced
by LODA/2
Replaced
by LOD
Yard-Wide Average Soil-Lead Concentration (pg/g)1
#
Surveyed
Units with
Positive
Averages
284
647
608
664
664
664
Arith-
metic
Mean
235
198
197
198
199
199
Geo-
metric
Mean2
61.9
50.5
58.2
50.1
52.7
55.8
Geo-
metric
C*H
Dev.2
4.46
5.13
4.72
4.74
4.45
4.17
Minimum
4.63
0.00
0.00
4.62
6.53
9.23
2501
Percen-
tile
21.3
16.1
14.3
15.6
16.4
17.0
Median
49.2
40.6
39.2
40.6
40.6
40.6
75th
Percen-
tile
142
145
145
145
145
145
Maximum
7030
9270
9270
9270
9270
9270
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with soil-lead data are used to
calculate the remaining statistics).
Note: The yard-wide average for a household is the average of the following two statistics: 1) the average of the mid-yard
sample results, and 2) the average of results for the driplme and entryway samples.
DRAFT - DO NOT CITE OR QUOTE
D2-2
August 28, 2000
-------
.1
1
a
w
0.01
HUDNS
NSLAH
(0)
NSLAH
(LOO/2)
NSLAH
(LOO/V2)
NSLAH
(LOO)
ho
O
O
O
Figure D2-1. Boxplots of Yard-Wide Average Soil-Lead Concentrations (ug/g) As Observed in the §403 Risk Analysis
(Using HUD National Survey Data) and in the NSLAH (under 5 approaches to handling not-detected values)
(imputed data omitted for the NSLAH)
-------
Table D2-2. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentration for
Households, Presented bv Housing Aae Category. As Reported in the §403
Risk Analysis Versus the Interim NSLAH Data (imputed data omitted for the
NSLAH)
Study
How Not-
Detected
and
Negative
Data were
Handled
Yard-Wide Average Soil-Lead Concentration 0/g/g)1
# Units
with
Positive
Averages
Arith-
metic
Mean
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
25"1
Percen-
tile
Median
7501
Percen-
tile
Maximum
Units Built Prior to 1940
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LODA/2
Replaced by
LOD
77
104
104
104
104
104
761
651
651
651
651
651
463
284
283
284
284
285
3.09
3.66
3.71
3.67
3.66
3.65
17.4
12.8
8.33
10.8
11.9
13.3
259
132
132
132
132
132
569
279
277
279
280
281
1030
571
571
571
571
571
4620
9270
9270
9270
9270
9270
Units Built from 1940 - 1959
5403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LOD/>/2
Replaced by
LOD
87
138
137
138
138
138
287
264
264
264
264
264
92.6
107
109
108
109
109
3.15
3.49
3.36
3.39
3.35
3.31
5.40
1.65
0.00
4.62
6.53
9.23
44.3
43.1
43.1
43.1
43.1
43.1
77.3
91.9
91.9
91.9
91.9
91.9
162
223
223
223
223
223
7030
4340
4340
4340
4340
4340
Units Built from 1960-1977 (1960 - 1979 for the 5403 risk analysis)
5403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LODA/2
Replaced by
LOD
120
190
182
193
193
193
55.0
76.7
76.0
77.2
77.7
78.4
32.8
31.1
33.9
32.6
34.2
36.2
2.56
3.69
3.45
3.27
3.08
2.91
4.63
0.00
0.00
4.83
6.83
9.66
19.7
13.7
12.1
14.7
15.3
16.3
29.7
27.7
27.2
28.3
28.4
28.6
61.6
59.3
59.3
59.3
59.3
59.3
996
1120
1120
1120
1120
1120
DRAFT - DO NOT CITE OR QUOTE
D2-4
August 28, 2000
-------
Table D2-2. (cent.)
Study
How Not-
Detected
and
Negative
Data were
Handled
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LOD/,/2
Replaced by
LOD
Yard-Wide Average Soil-Lead Concentration d/g/g)1
# Units
with
Positive
Averages
Units
28
160
131
172
172
172
Arith-
metic
Mean
Built After
31.3
27.6
26.1
28.3
29.3
30.6
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.a
Minimum
2501
Percen-
tile
Median
75-
Percen-
tile
Maximum
1977 (after 1979 for the §403 risk analysis)
22.4
15.2
18.6
15.7
17.7
20.2
2.31
3.29
2.98
2.71
2.43
2.18
5.35
0.00
0.00
4.65
6.57
9.30
13.6
5.67
1.89
6.24
7.87
10.3
21.2
14.3
12.0
14.5
15.2
16.0
45.0
32.9
32.9
32.9
32.9
32.9
97.4
474
472
475
476
477
NSLAH Units with Unspecified Year-Built Indicator
Interim
NSLAH
No
adjustment
Replaced by
0
Replaced by
LOD/2
Replaced by
LODA/2
Replaced by
LOD
55
54
57
57
57
169
168
169
169
170
66.6
70.3
62.7
64.8
67.2
4.26
3.99
4.21
4.02
3.84
0.00
0.00
4.74
6.70
9.47
19.4
17.9
19.4
19.4
19.4
49.6
49.6
49.6
49.6
49.6
158
158
158
158
158
2290
2290
2290
2290
2290
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with soil-lead data are used to
calculate the remaining statistics).
Note: The yard-wide average for a household is the average of the following two statistics: 1) the average of the mid-yard
sample results, and 2) the average of results for the driplme and entryway samples.
DRAFT - DO NOT CITE OR QUOTE
D2-5
August 28.2000
-------
0
O 100000 -
o
O
— 1
rn 10000 -
O
D
c
o
m 1000-
T
Q.
a.
.0 1OO -
|
"c
o
u
8
V 10-
o> S
46
l/l
i -
• •
• * •
4
<
1
:
1
i
• • •
•
: :
• • •
•
i $
• • •
j
|
|
i
•
I
i
•
I
CO
c
W-
O
O
o
I
I
I
I
r
HUONS NSLAH NSLAH
403 LOO/2
Priol to 194O
HUONS NSLAH NSLAH
403 LOO/2
1940 - 1939
NSLAH
1
HUONS NSLAH NSLAH
»O3 LOO/2
1960 - 1977(79)
HUONS
403
NSLAH
NSLAH
LOO/2
Alter 1977(79)
I
ISLAM NSLAH
LOO/2
Unip.cld.d
y Figure D2-2. Boxplots of Yard-Wide Average Soil-Lead Concentration (ug/g), by Housing Age Category, As Observed in
the §403 Risk Analysis (Using HUD National Survey Data) and in the NSLAH (under 2 approaches to handling
not-detected values) (imputed data omitted for the NSLAH)
-------
Table D2-3. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentration for
Households, Presented bv Census Region. As Reported in the §403 Risk
Analysis Versus the Interim NSLAH Data (imputed data omitted for the
NSLAH)
Study
How Not-
Detected
and
Negative
Data were
Handled
Yard-Wide Average Soil-Lead Concentration (pg/g)1
#
Surveyed
Units with
Positive
Averages
Arith-
metic
Mean
Geo-
metric
Mean2
Geo-
metric
Std.
Dev.2
Minimum
25th
Porcon-
tile
Median
75*
Percen-
tile
Maximum
Northeast
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
LOD/2
53
95
95
437
435
435
206
160
161
3.58
4.29
4.20
14.8
3.92
6.24
60.1
56.1
56.1
279
176
176
569
396
396
4320
3460
3460
Midwest
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
LOD/2
§403 Risk Analysis
(HUO Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
LOD/2
§403 Risk Analysis
(HUD Natl. Survey)
Interim
NSLAH
No
adjustment
Replaced by
LOD/2
73
143
144
134
250
257
52
159
168
404
221
221
125
161
161
112
61.7
62.5
81.4
63.6
63.8
44.5
36.4
35.5
34.4
28.0
29.3
6.33
5.05
4.77
South
2.94
4.60
4.36
West
3.92
4.35
3.48
4.63
0.00
4.90
5.22
0.00
4.65
4.79
0.00
4.62
19.7
20.8
20.6
22.6
11.5
12.6
14.2
10.4
11.2
51.6
59.5
59.5
40.8
27.2
27.2
27.2
29.4
29.4
264
206
206
79.3
78.6
78.6
61.6
70.0
70.0
2750
7070
7070
7030
9270
9270
2020
776
776
1 All statistics are calculated by weighting each household by its sampling weight.
2 Only household averages greater than zero are used to calculate this value (data for all units with soil-lead data are used to
calculate the remaining statistics).
Note: The yard-wide average for a household is the average of the following two statistics:. 1) the average of the mid-yard
sample results, and 2) the average of results for the dripline and entryway samples.
DRAFT - DO NOT CITE OR QUOTE
D2-7
August 28. 2000
-------
o
Tl
O
o
I
o
H
m
O
D
§
m
i
•
i
•
i
o
ho
60
HUGHS
403
CD
C
ho
o
o
o
NSLAH
LOO/2
HUONS NSLAH NSLAH
403 LOO/2
Uldwnl
HUONS
403
NSLAH
South
NSLAH
LOO/2
HUONS NSLAH NSLAH
403 LOO/2
W««(
Figure D2-3. Boxplots of Yard-Wide Average Soil-Lead Concentration (ug/g), by Census Region, As Observed in the §403
Risk Analysis (Using HUD National Survey Data) and in the NSLAH (under 2 approaches to handling not-
detected values) (imputed data omitted for the NSLAH)
-------
Table D2-4a. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentrations for
Households, Presented bv Housing Aae and Census Region. As Reported in
the §403 Risk Analysis Versus the Interim NSLAH Data Where No
Adjustments Were Made to Not-Detected Results (imputed data omitted for
the NSLAH)
Census
Region
Northeast
Midwest
South
West
Study
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Housing Age
Category
Prior to 1 940
1940-1959
1960-1977
(1960-79 for §403)
After 1977
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
( 1979 for §403)
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1 940
1940-1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Yard-Wide Average Soil-Lead Concentration1 b/g/g)
# Surveyed
Units
26
35
17
20
10
19
15
19
35
21
35
29
35
4
28
19
24
33
47
64
78
18
79
13
10
16
36
17
58
6
38
Arithmetic
Mean
542
903
573
292
79.1
138
62.6
1310
505
127
233
42.7
95.5
13.0
34.3
417
694
327
366
54.6
68.9
38.5
22.2
594
153
96.8
136
56.2
44.6
21.7
16.1
Geometric
Mean1
491
471
136
193
60.7
66.3
42.9
941
225
92.6
102
27.1
37.8
11.5
12.8
174
270
83.1
95.2
36.5
26.8
29.7
15.6
295
119
72.1
81.6
23.8
23.4
15.0
9.01
Geometric
Std. Dev.3
1.57
3.49
4.40
2.31
2.15
3.07
2.76
2.68
3.39
2.41
3.18
2.32
3.42
1.66
3.97
3.68
3.84
3.27
4.43
2.30
3.61
2.11
2.47
3.76
2.27
2.19
3.08
3.02
3.77
2.34
3.73
Median
444
461
60.1
194
69.7
50.9
43.1
1390
273
123
75.7
23.4
32.0
12.4
9.36
159
186
81.0
64.5
34.7
26.1
25.0
15.0
394
158
60.4
89.5
20.0
26.3
13.6
5.88
1 All statistics are calculated by weighting each household by its sampling weight.
3 Only household averages greater than zero are used to calculate this value (data for all units with soil-lead data are used to
calculate the remaining statistics).
Note: The yard-wide average for a household is the average of the following two statistics: 1) the average of the mid-yard
sample results, and 2) the average of results for the driplme and entryway samples.
DRAFT - DO NOT CITE OR QUOTE
D2-9
August 28,2000
-------
Table D2-4b. Descriptive Statistics of Yard-Wide Average Soil-Lead Concentrations for
Households, Presented bv Housing Age and Census Region. As Reported in
the §403 Risk Analysis Versus the Interim NSLAH Data Where Not-Detected
Results Were Replaced bv LOD/2 (imputed data omitted for the NSLAH)
Census
Region
Northeast
Midwest
South
West
Study
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
§403 Risk Anal.
Interim NSLAH
Housing Age
Category
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Prior to 1 940
1940- 1959
1960-1977
(1960-79 for §403)
After 1977
( 1979 for §403)
Prior to 1 940
1 940 - 1 959
1960-1977
(1960-79 for §403)
After 1977
(1979 for §403)
Yard-Wide Average Soil-Lead Concentration1 U/g/g)
# Surveyed
Units
26
35
17
20
10
19
15
19
35
21
35
29
35
4
29
19
24
33
47
64
80
18
82
13
10
16
36
17
59
6
46
Arithmetic
Mean
542
903
573
292
79.1
138
62.8
1310
505
127
233
42.7
95.8
13.0
34.9
417
694
327
366
54.6
69.5
38.5
22.7
594
154
96.8
136
56.2
45.2
21.7
17.9
Geometric
Mean
491
469
136
193
60.7
66.1
45.1
941
225
92.6
103
27.1
38.5
11.5
13.8
174
270
83.1
96.3
36.5
277
29.7
15.3
295
120
72.1
84.5
23.8
26.4
15.0
10.8
Geometric
Std. Dev.
1.57
3.53
4.40
2.31
2.15
3.08
2.45
2.68
3.38
2.41
3.15
2.32
3.34
1.66
3.09
3.68
3.84
3.27
4.37
2.30
3.25
2.11
2.30
3.76
2.25
2.19
2.76
3.02
2.85
2.34
2.44
Median
444
461
60.1
194
69.7
50.9
43.1
1390
273
123
75.7
23.4
32.0
12.4
9.67
159
186
81.0
64.5
34.7
26.1
25.0
14.7
394
158
60.4
89.5
20.0
26.3
13.6
7.68
' All statistics are calculated by weighting each household by its sampling weight.
Note: The yard-wide average for a household is the average of the following two statistics: 1) the average of the mid-yard
sample results, and 2) the average of results for the dnpline and entryway samples.
DRAFT - DO NOT CITE OR QUOTE
D2-10
August 28, 2000
-------
APPENDIX E
METHOD TO ESTIMATING TOTAL SOIL-LEAD CONCENTRATION
FROM ANALYTICAL RESULTS FOR FINE AND COARSE SOIL FRACTIONS
-------
Method to Estimating Total Soil-Lead Concentration
from Analytical Results for the Fine and Coarse Soil Fractions
In an effort to reflect bioavailable lead in soil, the Rochester Lead-in-Dust study
partitioned their collected soil samples into fine- and coarse-sieved fractions. The soil-lead
concentration of the complete sample (i.e., total soil) was not measured. The absence of such a
measure limits the ability to compare the soil results from the Rochester study with those of other
studies. The recent Milwaukee study, however, also fractioned their soil samples but made
provisions to simultaneously measure total soil-lead. This appendix describes an effort to use the
results of the Milwaukee study to estimate the soil-lead concentration of total soil for samples
collected in the Rochester study.
The Milwaukee study data available for this analysis represented 66 paired samples
collected at the child's play area and the residence's drip line. The same sieve-fraction used in
the Rochester was employed in Milwaukee. For each collected sample, the lead concentration of
fine-sieved, coarse-sieved and total soil was measured. The mass of each soil fraction was not
reported.
Figures E-l and E-2 compare the Milwaukee and Rochester study data. In particular,
these figures plot the coarse versus the fine soil-lead concentrations for the play area and drip
line measurements, respectively. Distinct plotting symbols delineate samples from the two
studies. These plots show that the data range and scatter about the trend line are considerably
greater in the Rochester study than in the Milwaukee study.
A likelihood ratio test was used to assess whether linear models for the two studies were
statistically different. Results for play area samples in the two studies (Figure E-l) do evidence
statistically (p<.01) distinct linear relationships between fine- and coarse-sieved soil-lead
concentrations. Results for drip line samples in the two studies (Figure E-2) were not
statistically distinct at the 0.05 level. These analyses suggest there are some differences in the
fine- versus coarse-sieved soil-lead concentration relationships measured in these studies. These
differences should be acknowledged when considering the merits of the Rochester total soil
estimation procedure outlined below.
To estimate the soil-lead concentration of total soil, it is useful to consider how total soil-
lead concentration may be calculated from fine- and coarse-sieve soil-lead concentrations and
masses. Specifically, let x/yfand xfyc represent the micrograms of lead (x) per gram of soil (y)
for fine- and coarse-sieved fractions, respectively, of a soil sample. The sample's total soil-lead
concentration, then, can be written as follows:
DRAFT - DO NOT CITE OR QUOTE E-1 August 28. 2000
-------
-1
0
o
1
0
— 1
m
0
XJ
O
C
g
—\
m
m
N)
C
CQ
C
M
co
o
0
o
100000 -
•
n ^ i*mf^i
! ^b ^^^BVV
— ^f m^^f^
O ^rh^5H^r
n en ra^B o
^sfcl^r""'
•C 31^^ n n
rO [-ffl r—i CH
UJB
1 — 1 ' — 'i — i
dfn Q n
i-ppP Tn
d r~i^
n
n
n
-T . 1 1 1 . • 1 1 1 . 1 1 1 . . 1 r — |
10 100 1000 10000 100OOO
Fine-Sieved Soil-Lead Concentration at Play Area
• • • MILWAUKEE ODD ROCHESTER
Figure E-1 . Coarse- versus Fine-Sieved Soil Lead Concentration Measured at Child's Play Area during Rochester and
Milwaukee Studies
-------
-H
i
0
0
z
0
H
O
1
m
O
;o
D
c
o
H
m
m
to
CD
c
en.
N>
CO
ro
o
o
0
1000000 :
a>
e
-H
a 100000 -
'^j
a
4-»
id
a
o
-H
4->
2 10000 -
c '
o
c
o
o
•o
(O
0)
71 1000 -
•-1
o
•O
0)
0)
co
i, 100 -
<0
m
o
u
10-
n
• n
n n
n
n n ^••^D
m^ta £]•
-*3-,rF^fin^^^— ' '— '
1 — 1[ rlBMT^^^ T •
i — I — f — iLcf-]' pQB^^Jn
CD L-J — iT~h ~jBh^3SPIU
'"^ SS^^TO^K^
Q Ji^^O3PBPFI ^P
n cf^ fi3 oSra^? a
nfi, 1 DiH LjT^— I—1
^l^LD
n
1 [ l,r.-tj"^Q
n ad^iT^ D
§P^D
1 — 1
n
r"°p i— i
n
ua
a
10 10O 1000 10000 1OOOOO 1000000
Fine-Sieved Soil-Lead Concentration at Drip Line
• • • MILWAUKEE a a a ROCHESTER
Figure E-2. Coarse- versus Fine-Sieved Soil Lead Concentration Measured at the Drip Line during Rochester and Milwaukee
Studies
-------
Thus, a sample's total soil-lead concentration can be written as a function of the sample's fine-
sieved soil mass fraction and the sample's fine- and coarse-sieved soil-lead concentrations. Since
the sieved soil mass fractions were not reported in the Milwaukee study, some assumptions
regarding these fractions were required. For the sake of simplicity, the fine-sieved soil mass
fraction was assumed constant. The total soil-lead concentration, then, is a weighted combination
of the fine- and coarse-sieved soil-lead concentrations,
Such a simple model is critical since the fine- and coarse-sieved soil-lead concentrations were the
only soil results reported in the Milwaukee study (i.e., no mass fraction data are available).
The model equation specified above was fit to both the play area and drip line data in the
Milwaukee study using the NLIN procedure in the SAS® System. This module was used because
it permitted the necessary link between the coefficients on fine- and coarse-sieved soil-lead
concentration. The estimated value for /7was approximately 0.25 when fitting the
aforementioned relationship to the play area samples alone, the drip line samples alone, and to
both sets of samples together. That is, the Milwaukee data suggested the following:
Total soil-lead concentration = 0.25-(Fine) + 0.75 -(Coarse).
Figure E-3 presents the results of fitting the above model to the Milwaukee data. The
plot is of the predicted total soil-lead concentration versus the observed total soil-lead
concentrations. Distinct plotting symbols represent the different sampling locations (drip line or
play area). As expected, the fit is more than reasonable for both locations.
DRAFT - DO NOT CITE OR QUOTE E-4 August 28,2000
-------
o
o
o
3
m
m
Ul
10OOO
1000
100
1O -L,
10
100
1000 10000
Observed Total Soil—Lead Concentration
1OOOOO
• • • Drip Line
A Flay Axes
t.
M Figure E-3. Predicted versus Observed Total Soil-Lead Concentration by Sampling Location (Milwaukee Study)
K)
O
O
o
-------
APPENDIX F
COMPARISON AND CONTRAST OF RISK ESTIMATES FROM
THE HUD MODEL AND THE ROCHESTER MULTIMEDIA MODEL
DEVELOPED IN THE §403 RISK ANALYSIS
-------
Comparison and Contrast of Risk Estimates from the HUD Model
and the Multimedia Models Developed in the §403 Risk Analysis
To determine how blood-lead concentration as predicted by the HUD model differs from
that predicted by the Rochester multimedia model, the HUD model results presented in Tables 4
and 5 of Lanphear et al., 1998, were compared to results under the Rochester multimedia model
given the same sets of input values considered in these two tables. HUD model results presented
in this appendix were taken from these two tables. However, when interpreting how these results
compare across the two models in this exercise, one should recall that the HUD model assumes
that input environmental-lead levels are "true" levels. This is the result of measurement error
adjustments made to this model, which were not made to the Rochester multimedia model.
Thus, estimates under the Rochester multimedia model assume that environmental-lead levels
input to the model are measurements that result from a risk assessment.
Tables 4 and 5 of Lanphear et al., 1998, reflected HUD model fits for all combinations of
the following:
• Floor (wipe) dust-lead loadings of 1, 5, 10, 15, 20, 25, 40, 50, 55, 70, and 100
• Soil-lead concentrations of 10, 72, 100, 400, 500, 1000, 1500, 2000, and 4000
ppm.
These same input values were also considered in this exercise. This list includes the proposed
§403 hazard standard for soil (2000 ppm) and national median levels (according to Lanphear et
al., 1998) for floor dust-lead loading (5 ug/ft2) and soil-lead concentration (72 ppm). In addition,
for the Rochester multimedia model, a floor dust-lead loading of 50 ug/ft2 (i.e., the proposed
§403 hazard standard for floor-dust) and a soil-lead concentration of 400 ppm (i.e., the proposed
§403 soil-lead level of concern) were added to the list of input values.
As the Rochester multimedia model requires window sill (wipe) dust-lead loading as
input, a value of 27.5 ug/ft2 was used. This value represents the national median dust-lead
loading for window sills, as estimated within the §403 risk analysis using HUD National Survey
data, with sampling weights updated to reflect the 1997 housing stock (the §403 risk analysis
report) and Blue Nozzle vacuum dust-lead loadings converted to wipe-equivalents using
conversion equations found in USEPA, 1997.
According to Lanphear et al., 1998, all HUD model fits assumed that maximum interior
paint-lead concentration was set at 1.6 mg/cm2 and water-lead concentration at 1 ppb; these
values represented national median levels. The age of child was specified as 16 months (the
mean age across all of the pooled data on which the model was developed), and values of
categorical variables were taken to be the average across the population represented by the
pooled data. The HUD model fits assumed no exposure to damaged paint, and exterior-lead
exposures were estimated from dripline soil samples.
DRAFT - DO NOT CITE OR QUOTE f-1 August 28,2000
-------
F.I COMPARING THE ESTIMATED GEOMETRIC MEAN
BLOOD-LEAD CONCENTRATIONS
Tables F-l and F-2 present geometric mean blood-lead concentrations (ug/dL) under each
combination of the floor dust-lead loading and soil-lead concentration values mentioned above,
as predicted by the HUD model and the Rochester Multimedia model, respectively.
Table F-1. Geometric Mean Blood-Lead Concentrations (fjgldl), as Predicted by the HUD
Model for Specified Values of Environmental-Lead Levels1
Interior
Floor
Dust-Lead
Loading
(//g/ft2)2
1
53
10
15
20
25
40
55
70
100
Soil-Lead Concentration at the Foundation Perimeter (ppm)
10
2.3
3.2
3.7
4.0
4.2
4.4
4.9
5.2
5.5
5.9
723
2.8
4.0
4.6
5.0
5.3
5.5
6.1
6.5
6.8
7.3
100
2.9
4.1
4.7
5.1
5.4
5.7
6.3
6.7
7.0
7.6
500
3.5
4.9
5.6
6.1
6.5
6.8
7.5
8.0
8.4
9.0
1000
3.8
5.3
6.1
6.6
7.0
7.3
8.1
8.6
9.1
9.7
1500
4.0
5.5
6.3
6.9
7.3
7.7
8.4
9.0
9.5
10.2
2000
4.1
5.7
6.5
7.4
7.6
7.9
8.7
9.3
9.8
10.5
4000
4.4
6.1
7.1
7.7
8.1
8.5
9.4
10.0
10.5
11.3
1 Taken from Table 4 of Lanphear et al., 1998. Table entries represent blood-lead concentrations for a 16-
month old child (i.e., the mean age in HDD's pooled analysis). Water-lead concentration is assumed to be 1.0
ppb, an estimate of the national median as determined in Lanphear et al., 1998, from the pooled data and other
sources. Maximum XRF paint-lead measurement is assumed to be 1.6 mg/cm2, which is the median level
based on data from the HUD National Survey. No exposure to damaged paint was assumed. The effects for
other categorical model predictors (i.e., study, race, SES, mouthing behavior) were set to the arithmetic mean
effect across the population represented by the study data.
2 Assumes wipe dust collection techniques.
3 Estimated median level based on data from the HUD National Survey, as determined in Lanphear et al., 1998.
The median wipe dust-lead loading was determined by converting Blue Nozzle vacuum loadings from the HUD
National Survey to wipe-equivalent loadings using a conversion equation published in Farfel et al., 1994.
DRAFT - DO NOT CITE OR QUOTE
F-2
August 28. 2000
-------
Table F-2. Geometric Mean Blood-Lead Concentrations (//g/dL), as Predicted by the
Rochester Multimedia Model for Specified Values of Environmental-Lead
Levels1
Interior
Floor
Dust-Lead
Loading
U/g/ft2)2
1
53
10
15
20
25
40
50
55
70
100
Soil-Lead Concentration at the Drip Line (ppm)
10
2.74
3.05
3.19
3.28
3.34
3.39
3.50
3.55
3.57
3.63
3 72
72s
3.43
3.82
4.00
4.11
4.18
4.25
4.38
4.45
4.47
4.55
4.65
100
3.56
3.96
4.15
4.26
4.34
4.41
4.55
4.61
4.64
4.72
4.83
400
4.18
4.64
4.86
4.99
5.09
5.16
5.33
5.40
5.44
5.53
5.66
500
4.28
4.76
4.99
5.12
5.22
5.30
5.46
5.54
5.58
5.67
5.80
1000
4.63
5.15
5.40
5.54
5.65
5.73
5.91
6.00
6.04
6.13
6.28
1500
4.85
5.40
5.65
5.80
5.92
6.00
6.19
6.28
6.32
6.43
6.58
2000
5.02
5.58
5.84
6.00
6.11
6.20
6.40
6.49
6.53
6.64
6.80
4000
5.43
6.04
6.32
6.49
6.61
6.71
6.92
7.03
7.07
7.19
7.36
1 Window sill (wipe) dust-lead loading is assumed to be 27.5 fjglft2. the median area-weighted household
average determined from HUD National Survey data (after converting Blue Nozzle dust-lead loadings to wipe-
equivalent loadings and after updating the sample weights to reflect the 1997 housing stock, using methods
developed for the §403 risk analysis). The reported geometric means in this table equal (0.91 *A + 0.09*B),
where A is the predicted geometric mean assuming PbP=0 (i.e., no deteriorated lead-based paint or paint pica
tendencies in the child - see Section 3.2), and B is the predicted geometric mean assuming PbP= 1.5.
2 Assumes wipe dust collection techniques.
3 Estimated median level based on data from the HUD National Survey, as determined in Lanphear et al., 1998.
The median wipe dust-lead loading was determined by converting Blue Nozzle vacuum loadings from the HUD
National Survey to wipe-equivalent loadings using a conversion equation published in Farfel et al., 1994.
At median environmental-lead levels, the HUD model and Rochester Multimedia model
estimates are very similar. The HUD model estimate of 4.0 ug/dL is only 4.7% above the
Rochester Multimedia model estimate of 3.82 ng/dL. At the proposed §403 standards for floor-
dust and soil (50 ug/ft2 and 2000 ppm, respectively), the HUD model predicts a geometric mean
blood-lead concentration of approximately 9.1 ug/dL, which is 40% above the Rochester
Multimedia model estimate (6.49 ug/dL).
To more easily observe how model estimates change as dust-lead and soil-lead levels
vary, Figures F-la and F-lb portray the information in Tables F-l and F-2 graphically. For each
model, the two figures demonstrate how predicted geometric mean blood-lead concentration
DRAFT - DO NOT CITE OR QUOTE
F-3
August 28, 2000
-------
r
S °
30 40 50 M TO
InUrior Floor Duit-U.4 U>«dlnj On/ft)
Figure F-1a. Predicted Geometric Mean Blood-Lead
Concentration vs. Floor Dust-Lead Loading
(//g/ft2). Assuming Soil-Lead Concentration
72 ppm
(see footnotes to Tables F-1 and F-2)
I *•
1500 2000 2500
Soil Lead COQC. (ppm)
Figure F-1b. Predicted Geometric Mean Blood-Lead
Concentration vs. Soil-Lead Concentration
(ppm). Assuming Floor Dust-Lead Loading = 5
//g/ft2
(see footnotes to Tables F-1 and F-2)
DRAFT - DO NOT CITE OR QUOTE
F-4
August 28, 2000
-------
increases as either floor dust-lead loading (Figure F-la) or soil-lead concentration (Figure F-lb)
increases. While results for the empirical model (Section 4.2.5 of the §403 risk analysis report)
are included in these figures, they should not be considered in the interpretation of results across
models. In both figures, environmental-lead levels in media other than that specified on the
horizontal axis are set at estimated national median levels, as indicated in the footnotes of Tables
F-landF-2.
Figure F-la shows that HUD model estimates become considerably higher than those for
the Rochester multimedia model when floor dust-lead loadings increase. As floor dust-lead
loading increases from 1 to 100 ug/ft2 and other environmental media are at their estimated
national median levels (e.g., soil-lead concentration = 72 ppm), predicted blood-lead
concentrations under the HUD model increase three-fold. In contrast, estimates under the
Rochester multimedia model increase by 35%. In the settings represented within Figure 3- la, the
HUD model estimates are similar to or lower than those for the Rochester multimedia model
only at very low floor dust-lead loadings (i.e., less than 10 ug/ft2). However, inferences at such
low loadings must be done with extreme caution.
Figure 3-lb shows a different relationship than that seen in Figure 3- la. In this plot, soil-
lead concentration increases from 10 to 4000 ppm, but floor dust-lead loading is fixed at 5 jig/ft2.
In this setting, estimates between the HUD model and the Rochester multimedia model are
nearly the same across the range of soil-lead concentrations. However, inferences at such a low
floor dust-lead loading must be made with caution in these models.
The extent of difference in the predicted geometric mean blood-lead concentration
between the HUD and Rochester multimedia model estimates gets larger as the assumed dust-
lead loading increases and as soil-lead concentration decreases. Among the different
combinations of dust-lead and soil-lead levels utilized in the model fits, the HUD model estimate
differs greatly at the largest dust-lead loading (100 ug/ft2) and the lowest soil-lead concentration
(10 ppm), where this estimate (5.9 ug/dL) is a 59% increase over the Rochester multimedia
model estimate (3.72 ug/dL).
F.2 Comparisons of the Estimated Percentage of Children With Blood-Lead
Concentrations At or Above 1 0 //g/dL
When an estimated geometric mean (GM) from the previous sub-section is combined
with an assumed geometric standard deviation (GSD) on the distribution of blood-lead
concentration, and if this distribution is assumed to be lognormal, then the probability of
observing blood-lead concentrations at or above 10 ug/dL (the lowest blood-lead concentration
considered elevated by the Centers for Disease Control and Prevention) is calculated as
io] - i -
ln(GSD)
where Q(z) is the probability of observing a value less than z under the standard normal
distribution. This sub-section presents estimates of this probability (expressed in percentage
DRAFT -DO NOT CITE OR QUOTE F-5 August 28. 2000
-------
terms) under the estimated geometric means in Tables F-l and F-2 and under three different
assumptions on the geometric standard deviation (GSD):
• GSD=1.6, used to represent within-house variability in the §403 risk analysis
• GSD=1.72, assumed in Lanphear et al., 1998
• GSD=1.75, calculated from data in the Rochester Lead-in-Dust study
Tables F-3 and F-4 present the estimated percentages under the HUD model and the Rochester
Multimedia model, respectively.
When GSD=1.72 and at estimated median environmental-lead levels, Tables F-3 and F-4
indicate that the estimated percentages are similar between the HUD model (4.56%) and the
Rochester multimedia model (3.79%). While the similarity was expected given the similar
geometric means observed in the previous sub-section, the HUD model estimate is
approximately 20% higher than the Rochester multimedia model estimate, which is a higher rate
of increase than the 4% increase observed in the estimated geometric mean. Furthermore, these
estimates can change considerably with the GSD. For example, under GSD=1.6, the estimates
are 45-55% lower (2.56% under the HUD model, 2.03% under the Rochester multimedia model)
than their respective values under GSD=1.72.
Figures F-2a and F-2b portray how the estimated percentages of blood-lead
concentrations at or above 10 ug/dL increase as dust-lead and soil-lead levels, respectively, are
increased. These estimates coincide with the geometric mean estimates plotted in Figures F-l a
and F-lb and are calculated under the same underlying assumptions (i.e., national median levels
are assumed for media not specified on the horizontal axis). Each figure contains three plots, one
for each assumed GSD value.
Figure 3-2a shows that at an assumed soil-lead concentration of 72 ppm, the HUD model
estimates become markedly increased as floor dust-lead loading increases to 100 fig/ft2. At 100
ug/ft2, the HUD model estimates from 25% to 29% of children have blood-lead concentrations at
or above 10 ug/dL (under GSD values from 1.6 to 1.75), while these estimates range from 5% to
9% under the Rochester multimedia model.
In contrast, Figure 3-2b shows that at an assumed floor dust-lead loading of 5 ug/ft2, the
HUD model and Rochester multimedia model provides nearly identical estimates of the
probability at or above 10 ug/dL, across the entire range of soil-lead concentration (10-4000
ppm). This is due to the similar geometric mean estimates observed in Figure 3-lb. At this floor
dust-lead loading and at GSD=1.72, the estimated probabilities range from approximately 1.5%
to 18% under both models as the soil-lead concentration increases.
DRAFT-DO NOT CITE OR QUOTE F-6 August 28,2000
-------
Table F-3. Percentage of Children with Blood-Lead Concentration At or Above 10 //g/dL.
as Predicted by the HUD Model for Specified Values of Environmental-Lead
Levels and Under Different Estimates for GSD1
Interior
Floor
Dust-Lead
Loading
U/g/ft2)2
1
5
10
15
20
25
40
55
70
100
Soil-Lead Concentration at the Yard Perimeter (ppm)
10
0.09
0.77
1.72
2.56
3.25
4.03
6.45
8.21
10.2
13.1
721
0.34
2.56
4.92
7.01
8.84
10.2
14.6
18.0
20.6
25.2
100
0.42
2.89
5.41
7.60
9.49
11.6
16.3
19.7
22.4
28.0
500
GSD = 1.6
1.28
6.45
10.9
14.6
18.0
20.6
27.0
31.7
35.5
41.1
1000
1.98
8.84
14.6
18.8
22.4
25.2
32.7
37.4
42.0
47.4
1500
2.56
10.2
16.3
21.5
25.2
28.9
35.5
41.1
45.7
51.7
2000
2.89
11.6
18.0
26.1
28.0
30.8
38.4
43.9
48.3
54.1
GSD = 1.72
1
5
10
15
20
25
40
55
70
100
0.34
1.78
3.34
4.56
5.48
6.50
9.42
11.4
13.5
16.5
0.95
4.56
7.61
10.1
12.1
13.5
18.1
21.4
23.9
28.1
1.12
5.01
8.19
10.7
12.8
15.0
19.7
23.0
25.5
30.6
2.64
9.42
14.3
18.1
21.4
23.9
29.8
34.0
37.4
42.3
3.72
12.1
18.1
22.2
25.5
28.1
34.9
39.0
43.1
47.8
4.56
13.5
19.7
24.7
28.1
31.5
37.4
42.3
46.2
51.5
5.01
15.0
21.4
28.9
30.6
33.2
39.9
44.7
48.5
53.6
4000
4.03
14.6
23.3
28.9
32.7
36.5
44.8
50.0
54.1
60.3
6.50
18.1
26.4
31.5
34.9
38.2
45.5
50.0
53.6
58.9
GSD = 1.75
1
5
10
15
20
25
40
55
70
100
0.43
2.09
3.78
5.08
6.06
7.12
10.1
12.1
14.3
17.3
1.15
5.08
8.26
10.8
12.8
14.3
18.9
22.1
24.5
28.7
1.35
5.56
8.86
11.4
13.5
15.8
20.5
23.7
26.2
31 2
3.03
10.1
15.0
18.9
22.1
24.5
30.4
34.5
37.8
42.5
4.19
12.8
18.9
22.9
26.2
28.7
35.3
39.4
43.3
47.8
5.08
14.3
20.5
25.4
28.7
32.0
37.8
42.5
46.3
51.4
5.56
15.8
22.1
29.5
31.2
33.7
40.2
44.8
48.6
53.5
7.12
18.9
27.0
32.0
35.3
38.6
45.6
50.0
53.5
58.6
1 Footnotes are indicated within Table F-1.
DRAFT - DO NOT CITE OR QUOTE
F-7
August 28, 2000
-------
Table F-4. Percentage of Children with Blood-Lead Concentration At or Above 10 //g/dL,
as Predicted by the Rochester Multimedia Model for Specified Values of
Environmental-Lead Levels and Under Different Estimates for GSD1
Interior Floor
Dust-Lead
Loading
(//0/ftV
1
5
10
15
20
25
40
50
55
70
100
Soil-Lead Concentration at the Drip Line (ppm)
10
723
100
400
500
1000
1500
2000
4000
GSD =1.6
0.30
0.57
0.76
0.88
0.98
1.07
1.27
1.38
1.42
1.55
1.76
1.15
2.03
2.55
2.91
3.19
3.42
3.95
4.23
4.35
4.67
5.18
1.41
2.45
3.06
3.48
3.80
4.07
4.68
5.00
5.13
5.50
6.08
3.16
5.13
6.24
6.97
7.53
7.98
9.01
9.53
9.75
10.35
11.28
3.56
5.72
6.93
7.72
8.32
8.81
9.92
10.47
10.72
11.36
12.35
5.09
7.92
9.46
10.46
11.21
11.82
13.17
13.86
14.15
14.93
16.12
6.20
9.48
11.23
12.35
13.20
13.88
15.39
16.15
16.48
17.33
18.64
7.10
10.71
12.62
13.83
14.75
15.48
17.10
17.91
18.26
19.18
20.57
9.68
14.14
16.44
17.89
18.96
19.82
21.71
22.64
23.05
24.09
25.68
GSD = 1.72
1
5
10
15
20
25
40
50
55
70
100
0.85
1.43
1.76
1.99
2.16
2.31
2.64
2.81
2.88
3.08
3.40
2.44
3.79
4.54
5.03
5.41
5.71
6.40
6.75
6.90
7.30
7.92
2.86
4.40
5.24
5.79
6.21
6.55
7.31
7.69
7.86
8.30
8.99
5.36
7.86
9.17
10.01
10.64
11.14
12.27
12.83
13.07
13.71
14.68
5.89
8.57
9.97
10.86
11.52
12.06
13.25
13.84
14.10
14.76
15.79
7.81
11.08
12.76
13.82
14.61
15.24
16.63
17.31
17.61
18.38
19.56
9.13
12.78
14.63
15.79
16.65
17.33
18.84
19.58
19.90
20.73
22.00
10.16
14.09
16.06
17.29
18.20
18.93
20.52
21.30
21.63
22.50
23.83
13.00
17.60
19.87
21.27
22.30
23.12
24.90
25.77
26.14
27.11
28.56
GSD = 1.75
1
5
10
15
20
25
40
50
55
70
100
1.04
1.69
2.06
2.31
2.51
2.66
3.02
3.21
3.29
3.51
3.84
2.81
4.27
5.06
5.58
5.98
6.29
7.01
7.37
7.53
7.94
8.58
3.26
4.91
5.80
6.38
6.81
7.17
7.95
8.35
8.52
8.98
9.68
5.93
8.52
9.87
10.72
11.36
11.88
13.01
13.58
13.82
14.46
15.44
6.48
9.25
10.68
11.58
12.26
12.80
14.00
14.59
14.85
15.52
16.55
8.47
11.81
13.51
14.58
15.37
16.00
17.38
18.07
18.36
19.13
20.30
9.83
13.53
15.39
16.55
17.41
18.09
19.59
20.32
20.64
21.46
22.71
10.88
14.84
16.82
18.05
18.95
19.67
21.25
22.02
22.35
23.21
24.51
13.75
18.35
20.60
21.99
23.01
23.82
25.57
26.42
26.79
27.74
29.16
Footnotes are indicated within Table F-2.
DRAFT - DO NOT CITE OR QUOTE
F-8
August 28.2000
-------
Model:
GSD=1.6
1O BO 3O 4O SO BO TO
Interior Floor Duat~L**d Lo»dln« (M*/ft )
9O 10O
1O I'D
30 40 SO BO TO
Interior floor Du«l-L*»d Lo«din« (M
-------
GSD=1.6
Model:
ooo 1000 l»oo eooo esoo aooo 3000 «ooo
Soil L-»d Cone (ppra)
M
ja zox -
GSD=1.72
BOO 1000 1500 ZOOO Z6OO 3OOO 30OO
Soil L«»d Cone (ppm)
GSD=1.75
1SOO BOOO
Soil L«*d Cono (ppm)
BOO 3OOO 3BOO 4OOO
HUD Model
Rochester Multimedia Model
Empirical Model
Figure F-2b. Predicted Percentage of Children with Blood-Lead Concentration At or Above
10 /yg/dL vs. Soil-Lead Concentration (ppm). Assuming Floor Dust-Lead
Loading = 5 //g/ft2 (see footnotes to Tables F-1 and F-2)
DRAFT - DO NOT CITE OR QUOTE F-10 August 28,2000
-------
Across Tables F-3 and F-4, the largest deviation in the estimated percentage of children
with blood-lead concentration at or above 10 ug/dL between the HUD model and the Rochester
multimedia model exists at the lowest soil-lead concentration (10 ppm) and the highest floor
dust-lead concentration (100 ug/ft2). Here, the HUD model estimate (16.5%) is nearly five times
that under the Rochester multimedia model (3.4%) when GSD=1.72.
Table F-5 presents the predicted geometric mean blood-lead concentration and percentage
of children with blood-lead concentration at or above 10 ug/dL, at the proposed §403 hazard
standards for floors and soil (50 ug/ft2 and 2000 ppm, respectively). For the Rochester
multimedia model, the window sill dust-lead loading is assumed to be 27.5 ug/ft2 (the estimated
national median). At these levels, the GSD assumption has less of an impact on the predicted
percentages than was seen at national median levels. However, the HUD model predicts
considerably higher percentages than the other.
Table F-5. Predicted Geometric Mean Blood-Lead Concentration and Percentage of
Children with Blood-Lead Concentration At or Above 10 //g/dL, at the
Proposed §403 Hazard Standards for Floors and Soil (50 //g/ft2 and 2000
ppm. Respectively) and at a Window Sill Dust-Lead Loading of 27.5 //g/ft2
(An Estimated Median Level for the Nation)
Model
HUD Model*
Rochester Multimedia
Model
Predicted Geometric
Mean Blood-Lead
Concentration
(/ig/dL)
9.1
6.49
Predicted Percentage of Children With
Blood-Lead Concentrations At or Above
10 //g/dL
GSD =1.6
42%
17.9%
GSD =1.72
43%
21.3%
GSD = 1.75
44%
22.0%
Values are interpolated from results presented in Lanphear et al., 1998. This model does not
use window sill dust-lead loading at an input value.
DRAFT - DO NOT CITE OR QUOTE
F-11
August 28, 2000
-------
APPENDIX G
PERFORMANCE CHARACTERISTICS ANALYSIS
CITED IN THE §403 PROPOSED RULE
-------
ilBaitelle
Putting Technology To Work
Date Septembers, 1997
TO Todd Holderman
From Ronald Menton and Warren Strauss
subject Requested Analyses for WA 3-28 EPA Contract No.
68-D5-0008
Attached are two tables describing the results of analyses performed to identify example options
for combined multi-media standards which achieve negative predictive values of 99,95 and 90
percent for detecting a childhood blood-lead concentration of 10 ug/dL. The negative predictive
value is defined in this analysis as the probability of a resident child in the Rochester Lead-in-
Dust study having a blood-lead concentration below 10 ug/dL, given that lead-levels in
residential environmental media are below the combined standard. The example standards
provided in this memorandum are based on an empirical sensitivity/specificity analysis
performed on a subset of 77 homes/children from the Rochester Lead-in-Dust Study. These 77
homes included measurements of children's blood-lead concentration, soil-lead concentration,
uncarpeted floor and window sill dust-lead loading and the percentage of interior and exterior
painted surfaces with deteriorated lead-based paint. For each home, soil-lead concentrations
measured for the drip-line and play-area sampling locations were averaged to produce a yard-
wide average soil-lead concentration. The sensitivity/specificity analyses focussed on all
possible combinations of the following potential standards for environmental lead:
Environmental Media
Uncarpeted Floor Dust-Lead Loading
Window Sill Dust-Lead Loading
Average Soil-Lead Concentration
Maximum of Percent of Intenor/Extenor
Potential Standards Considered in Analysis
50, 75, 100, 125, 150, 175, 200 and 400 ug/ft2
800, 500, 300 and 100 ug/ft2
200, 300, 400, 500, 600, 700, 900, 1000, 1500 ug/g
5,10,20%
Table 1 provides the maximum lead-levels identified in each of the above four environmental
media, which when combined, achieve a negative predictive value (NPV](of 99,95 and 90
percent or above. Note that combined standards that achieve a NPV of 99% also achieve NPV s
of 95% and 90%, and that combined standards that achieve a NPV of 95% also achieve a NPV of
90%.
Table 2 provides a summary of all the potential combinations of standards in the above four
environmental media that achieved negative predictive values of 99,95 and 90 percent or above.
In Table 2 the negative predictive value achieved corresponds to any combination of potential
standards in a row. For example, all combinations of standards of 50 - 400 ug/fT for dust on
uncarpeted floors, 100 - 800 ug/ft2 for dust on window sills, 200 - 900 ug/g for average soil and
5 - 20 percent of painted surfaces having deteriorated lead-based paint resulted in negative
DRAFT - DO NOT CITE OR QUOTE
G-1
August 28.2000
-------
Septembers, 1997
predictive values of 99 percent or above.
Please note that the results provided in Tables 1 and 2 are based on an analysis of data from 77
homes, and that since there were relatively few homes that had environmental lead-levels below
the combination of standards under consideration, the denominator for the negative predictive
value estimates are small in most cases (i.e. less than 25).
DRAFT-DO NOT CITE OR QUOTE G-2 August 28.2000
-------
Septembers, 1997
Table 1 Example Options For the Maximum Combined Multi-Media Standard which
Achieves a NPV of 99,95 and 90% for Detecting a Blood-Lead Concentration
of 10 ug/dL, Based on Data from the Rochester Lead-in-Dust Study.
NPV
Achieved
99%
95%
90%
1 =^^
Uncarpeted Floor
Dust-Lead Standard
fog/ft2)
400
50
400
400
Average Soil-Lead
Concentration
G»g/g)
900
1500
1500
1500
Window Sill Dust-
Lead Standard
(ug/ft1)
800
500
500
800
Maximum of
Percent of
Interior/Exterior
Components with
Deteriorated LBP
20
20
20
20
Table 2. Example Options For All Combinations of Multi-Media Standards which
Achieve a NPV of 99,95 and 90% for Detecting a Blood-Lead Concentration
of 10 ug/dL, Based on Data from the Rochester Lead-in-Dust Study.
NPV
Achieved
99%
95%
90%
Uncarpeted Floor
Dust-Lead Standard
(US/ft2)
400, 200, 175, 150,
125, 100, 75, 50
50
400, 200, 175, 150,
125, 100, 75
400, 200, 175, 150,
125, 100, 75
Average Soil-Lead
Concentration
(Mg/g)
900, 700, 600, 500,
400, 300, 200
1500, 1000
1500
1500
1000
1500, 1000
Window Sill Dust-
Lead Standard
(ug/ft1)
800, 500, 300, 100
500, 300, 100
500
800, 300, 100
500, 300
500, 300
100
800
Maximum of
Percent of
Interior/Exterior
Components with
Deteriorated LBP
20, 10, 5
20, 10, 5
20
20
10,5
20, 10, 5
20
20, 10, 5
The options for combined multi-media standards in these tables are based on a sensitivity/specificity analysis of
empirical data from 77 homes in the Rochester Lead-in-Dust Study which included measurements of children's
blood-lead concentration, drip-line and play-area soil-lead concentration, uncarpeted floor and window sill dust-lead
loading, and the percentage of interior and exterior painted surfaces with deteriorated lead-based paint.
DRAFT - DO NOT CITE OR QUOTE
G-3
August 28, 2000
-------
APPENDIX H
REVIEW OF PUBLISHED INFORMATION ON POST-INTERVENTION
WIPE DUST-LEAD LOADINGS ON FLOORS AND WINDOW SILLS
-------
H1.0 INTRODUCTION
One goal of the §403 risk analysis was to determine how the likelihood of children with
blood-lead concentrations exceeding certain thresholds (10 and 20 ug/dL) declines as a result of
reducing environmental-lead levels when interventions are performed in response to §403 rules.
An empirical model was used in both a pre- and post-intervention setting to predict geometric
mean blood-lead concentration as a function of environmental-lead levels, including average
dust-lead loadings for floors and window sills. It was assumed that pre-intervention average
dust-lead loadings on floors and window sills were reduced when performing the following
interventions:
• Dust cleaning (as triggered by exceeding either the floor or window sill dust-lead
standards)
• Interior paint abatement
• Soil removal
For each of these interventions, the assumed post-intervention wipe dust-lead loadings are as
follows:
• Floors: 40 ug/ft2 or the pre-intervention value, whichever is smaller
• Window sills: 100 ug/ft2 or the pre-intervention value, whichever is smaller.
Note that both assumptions are below their respective §403 standards. Post-intervention dust-
lead loadings are assumed to hold for four years following a dust cleaning, 20 years following
interior paint abatement, and permanently following soil removal.
Since the §403 risk analysis was performed, additional information has been identified
which could be used to refine the assumptions on post-intervention wipe dust-lead loadings.
This appendix examines some of that information and summarizes existing data from
intervention studies to characterize pre- and post-intervention wipe dust-lead loadings.
H2.0 REVIEW OF AVAILABLE INFORMATION
According to Section 6.1.1 of the §403 risk analysis report, the post-intervention dust-
lead loadings of 40 ug/ft2 for floors and 100 ug/ft2 for window sills were selected based on data
from EPA's Comprehensive Abatement Performance (CAP) study and the Baltimore
Experimental Paint Abatement study. Justification was as follows:
• Geometric mean vacuum dust-lead loadings from abated units in the CAP study were
29 ug/ft2 for floors (187 samples) and 92 ug/ft2 for window sills (78 samples), where
the samples were collected approximately two years after paint intervention
performed within the HUD Lead-Based Paint Abatement Demonstration.
DRAFT-DO NOT CITE OR QUOTE H-1 August 28,2000
-------
• Geometric mean wipe dust-lead loadings in the Baltimore Experimental Paint
Abatement study were 41 ug/ft2 for floors and 103 ug/ft2 for window sills, in 13
housing units approximately 18-42 months after complete paint intervention.
Intervention studies that contain information on pre- and post-intervention dust-lead
loadings (assuming either wipe dust collection methods or a method in which the reported
loadings can be converted to wipe-equivalent loadings) and that can be used to evaluate the §403
assumptions on post-intervention dust-lead loadings are identified in Table H-l. These studies
were included in USEPA, 1995a, and USEPA, 1998, which contain summary information on
studies available in the scientific literature whose findings could be used to make conclusions on
the effectiveness of lead hazard intervention (defined as "any non-medical activity that seeks to
prevent a child from being exposed to the lead in his or her surrounding environment"). A
summary of key information on study design and conclusions for the studies in Table H-l is
found in Appendix H2.
When comparing dust-lead loading results across the studies in Table H-l, the following
issues should be considered:
Converting vacuum dust-lead loadings to wipe-equivalent loadings
Two of the studies in Table H-l used dust collection methods other than the wipe
method. The Baltimore R&M study used the BRM vacuum method, while the CAP study used a
cyclone vacuum specifically developed for the study. While post-intervention wipe dust-lead
loadings are of interest here, these two studies are included in Table H-l as previous efforts allow
the vacuum dust-lead loadings to be converted to wipe-equivalent loadings. These conversions
were made prior to displaying results from these two studies in this appendix.
The Baltimore R&M study collected composite dust samples using the BRM vacuum
method. The conversion of BRM dust-lead loadings to wipe-equivalent loadings for the
Baltimore R&M study was developed within the §403 risk analysis effort (USEPA, 1997a) and
takes the following form:
Floors: Wipe = (px8.34xBRM0371) + ((l-p)*3.01 xBRM0227)
Window sills: Wipe= 14.8xBRM0453
where Wipe is the average wipe dust-lead loading, BRM is the average BRM dust-lead loading,
and p is the proportion of a composite floor-dust sample obtained from uncarpeted floors. These
conversion equations were determined based on side-by-side BRM/wipe dust-lead loading data
from four studies.
Dust-lead loadings for samples collected by the CAP study's cyclone vacuum were
converted to wipe-equivalent loadings based on the conclusion made within the CAP study that
vacuum dust-lead loadings were, on average, 1.38 times larger than wipe dust-lead loadings
DRAFT-DO NOT CITE OR QUOTE H-2 August 28.2000
-------
Table H-1. Studies Containing Information on Pre-lntervention and Post-Intervention
Dust-Lead Loadings on Floors and Window Sills, Where Wipe Collection
Methods or a Method Whose Loadings Can Be Converted to Wipe-
Equivalents Were Used
Study
Baltimore (MD) Dust
Control Study
Baltimore (MD)
ixperimental Paint
Abatement Studies
Baltimore (MD) Follow-
up Paint Abatement
Baltimore (MD) Repair
and Maintenance (R&M)
Study
Baltimore (MD)
Traditional/Modified
Paint Abatement Study
Boston (MA) Interim
Dust Intervention Study
East St. Louis (ID
Educational Intervention
Evaluation of the HUD
Lead-Based Paint
Hazard Control Grant
Program (HUD
Grantees) (data
collected through
August, 1997)
HUD Abatement
Demonstration
Program/EPA
Comprehensive
Abatement Performance
(CAP) Study
Study
Duration
1981
1986-87
(Study #1)
12/91 -01/92
(Study #2)
01/91 -06/92
1993-95
1984-85
05/93 - 04/95
1994-95
1994-
present
1989-90
(HUD Demo)
03/92 - 04/92
(CAP Study)
Type of Interventions
Considered
aint interventions
ome units received periodic
ust control
aint interventions using
xperimental procedures, with
xtensive cleanup
Paint interventions with
extensive clean-up
Various types of R&M paint
nterventions (including
cleanup, prevention of
recontamination, and
education)
"Traditional" and "modified"
paint abatements, with some
cleanup.
Intervention groups received
paint and/or dust intervention
low-tech). Comparison group
received an outreach visit.
Educational materials provided,
potential lead hazards
identified.
Wide range of interventions to
reduce/eliminate lead-based
paint hazards.
Encapsulation/enclosure
Various paint removal methods
Type of
Wipe
Digestion
Method
Cold HCI
Cold HCI
Cold HCI
BRM
vacuum
method was
used
Cold HCI
Cold HCI
Cold HCI
(likely)
Heated
HNO3/H202
Heated
HNO3/H202
(CAP Study
cyclone was
used in the
CAP Study
Reference(s)
Charney et al., 1983
Farfel and Chisolm,
991
USEPA, 1987
Farfel et al.. 1994
MDE, 1995
USEPA, 1996c
USEPA, 1997b
USEPA, 1997c
Farfel and Chisolm,
1990
Aschengrau et al.,
1998
Mackey et al., 1996
Copley, 1995
NCLSH and UC,
1997
NCLSH and UC,
1998
HUD, 1991
USEPA, 1996a
USEPA, 1996b
DRAFT - DO NOT CITE OR QUOTE
H-3
August 28.2000
-------
Table H-1. (cont.)
Study
Jersey City (NJ)
Children's Lead
Exposure and Reduction
(CLEAR) Dust
Intervention Study
Paris Paint Abatement
Study
Rochester (NY)
Educational Intervention
Study
Study
Duration
1992-94
01/90-02/92
08/93 - 06/94
Type of Interventions
Considered
Biweekly dust control
assistance and educational
sessions
Paint interventions with dust
cleaning
Intervention group received
direction on performing periodic
dust control. Control group
received educational materials
only.
Type of
Wipe
Digestion
Method
Heated
HN03/H202
Unspecified
Heated
HN03/H202
Reference(s)
Adgate et al., 1995
Rhoads et al., 1 996
Lioy et al., 1 997
Nedellec et al., 1995
Lanphear et al.,
1995
Lanphear et al.,
1996
(page 147 of USEPA, 1996b), regardless of lead level or sampling component. This conclusion
was made by fitting a log-linear regression model, using an errors-in-variables approach, on lead
loading data for 33 pairs of side-by-side vacuum/wipe dust samples collected within the CAP
study. The model predicted vacuum dust-lead loading as a function of wipe dust-lead loading.
Therefore, the conversion of vacuum dust-lead loading data from the CAP study (for both floors
and window sills) involved dividing each vacuum dust-lead loading by 1.38 to obtain a wipe-
equivalent loading. The estimated geometric mean wipe dust-lead loading equals the geometric
mean vacuum dust-lead loading, divided by 1.38.
Handling differences in wipe digestion methods
The studies in Table H-1 are identified according to the type of wipe digestion method
used in the analytical process. Generally, one of two categories of digestion methods was used
by each study. The "heated HNO3/H2O2" method, which is the method recommended in EPA's
National Lead Laboratory Accreditation Program (NLLAP), allows total lead amounts in the
sample to be determined. The "cold HC1" method, documented in Vostal et al., 1974, and used
at the Kennedy Krieger Institute in Baltimore, MD, generally allows only "bioavailable" lead
amounts to be measured in the sample. Therefore, in order to make wipe dust-lead loadings
comparable across all studies in Table H-1, it is necessary to adjust the "bioavailable" lead
loadings that are reported in the studies that used the "cold HC1" digestion method to reflect total
lead amounts. Appendix A of USEPA, 1997a, provided a means by which this adjustment can
be made:
T = B1.416
where T is the total dust-lead loading, and B is the "bioavailable" dust-lead loading. This
adjustment was developed by fitting a log-linear regression model (with no intercept term) on
DRAFT - DO NOT CITE OR QUOTE
H-4
August 28, 2000
-------
Table H-1. (cont.)
uncarpeted floor dust-lead loading data that were collected in a pilot study (NCLSH, 1993). This
pilot study investigated how dust-lead loadings changed across five different sampling and
analysis methods.
In this appendix, summary statistics for studies labeled in Table H-1 as utilizing the "cold
HC1" wipe digestion method were calculated on dust-lead loadings that were adjusted by the
method in the previous paragraph. This implies taking geometric means calculated on the study
data to the 1.1416 power.
Considering different intervention methods across studies
As seen in the second column of Table H-1, the studies utilized different intervention
approaches. The HUD Grantees evaluation program is the most widely-encompassing of the
studies, containing dust-lead loading data at up to 12 months post-intervention for floors and
window sills in over 500 housing units as measured by 14 Grantees across the country.
Therefore, the impact of intervention activities on dust-lead loading will likely vary considerably
across these studies. Furthermore, caution should be used in considering the results of certain
studies, such as the educational intervention studies, when the aim is to evaluate the effect of
performing highly-intensive dust and paint abatements on dust-lead loading.
H3.0 RESULTS
For eight studies in Table H-1 that measured and documented post-intervention dust-lead
loadings and which considered paint and/or dust interventions (i.e., not just educational
interventions), Tables H-2 and H-3 provide summaries of the measured dust-lead loadings from
these studies, both prior to intervention (if available) and at specified time points following the
interventions, for floors and window sills, respectively. Summaries are presented according to
study group within each study. These tables contain geometric mean dust-lead loadings for all
studies but the HUD Grantees evaluation, whose references provided only median dust-lead
loadings. Note that not all studies in these tables provided information on pre-intervention dust-
lead loadings. Also, as discussed in the previous chapter, the measured dust-lead loadings in the
Baltimore R&M study and the CAP study have been converted from vacuum to wipe-equivalent
loadings, and dust-lead loadings in studies using the "cold HC1" wipe digestion method have
been adjusted to reflect total lead loadings, prior to preparing the summaries in Tables H-2 and
H-3.
More detailed dust-lead loading summaries are provided in the tables in Appendix H3.
These tables include the information in Tables H-2 and H-3, along with sample sizes associated
with the summaries, 95% confidence intervals for selected estimates, and reported differences in
dust-lead loadings from pre-intervention which were measured in the Paris Paint Abatement
study and the two educational intervention studies in Rochester and East St. Louis.
DRAFT-DO NOT CITE OR QUOTE H-5 August 28.2000
-------
Table H-2. Summaries of Pre- and Post-Intervention Floor Dust-Lead Loadings
from Studies Evaluating Paint and/or Dust Interventions
Study
Baltimore
Experimental Paint
Abatement Studies2
Baltimore Follow-up
Paint Abatement
Study2
Baltimore R&M
Study3
Baltimore Traditional/
Modified Paint
Abatement Study2
Boston Interim Dust
Intervention Study2
Study
Group
Study 1
(6 homes)
Study 2
(13 homes)
6-Month Follow-up
1 2-Month Follow-up
1 9-Month Follow-up
All Occupied Units
Previously-Abated Units
Units Slated for R&M
Intervention
Modern Urban Units
Traditional
Modified
Automatic Intervention
Randomized Intervention
Pro-Intervention Floor
Dust-Lead Loadings1
big/ft2)
1261
556
40.9
45.6
58.6
10.0
549
642
33.2
37.3
Post-Intervention
Floor Dust-Lead Loadings1
Time Following
Intervention
Immediately
6-9 Months
Immediately
1.5-3.5 Years
Immediately
5-7 Months
Immediately
10- 14 Months
Immediately
14-24 Months
Immediately
2 Months
6 Months
1 2 Months
1 8 Months
24 Months
48 Months
6 Months
1 2 Months
1 8 Months
24 Months
Immediately
2 Months
6 Months
1 2 Months
1 8 Months
24 Months
6 Months
1 2 Months
1 8 Months
24 Months
48 Months
Immediately
6 Months
Immediately
6 Months
6 Months
6 Months
Summary Value b/g/ft*)
259
99
20
69
47
22
41
20
24
36
52.5
40.2
26.5
27.1
24.8
24.1
8.4
41.1
39.8
37.3
33.0
52.5
40.2
36.3
39.9
33.3
35.0
8.1
7.3
7.8
7.1
8.4
4033
714
1626
714
23.9
31.4
DRAFT - DO NOT CITE OR QUOTE
H-6
August 28, 2000
-------
Table H-2. (cont.)
HUD Grantees
CAP Study4
Study
Group
All Grantees
Baltimore
Boston
Massachusetts
Milwaukee
Minnesota
Rhode Island
Vermont
Wisconsin
Abated Units
Dust-Lead Loadings1
(//g/ft2)
19
41
24
24
14
18
26
28
9
22
Post-Intervention
Floor Dust-Lead Loadings1
Time Following
Intervention
Immediately
6 Months
12 Months
Immediately
6 Months
1 2 Months
Immediately
6 Months
12 Months
Immediately
6 Months
1 2 Months
Immediately
6 Months
12 Months
Immediately
6 Months
1 2 Months
Immediately
6 Months
1 2 Months
Immediately
6 Months
1 2 Months
Immediately
6 Months
1 2 Months
2 Years
1 2 Months
Summary Value U/g/ft2!
17
14
14
18
42
41
54
16
18
20
11
9
15
10
10
18
18
18
7
6
6
17
21
21
8
6
5
21.0
15
1 Values are geometric means except for the HUD Grantees studies, where values are medians.
2 Results are adjusted to reflect total dust-lead loadings by exponentiating the "bioavailable" dust-lead loadings as reported in
the study to the 1.1416 power.
3 Results for the Baltimore R&M Study are converted from BRM dust-lead loadings to wipe-equivalent loadings.
4 Results for the CAP study are converted from CAPS cyclone dust-lead loadings to wipe-equivalent loadings.
DRAFT - DO NOT CITE OR QUOTE
H-7
August 28.2000
-------
Table H-3. Summaries of Pre- and Post-Intervention Window Sill Dust-Lead Loadings
from Studies Evaluating Paint and/or Dust Interventions
Study
Baltimore
Experimental Paint
Abatement Studies2
Baltimore Follow-up
Paint Abatement
Study2
Baltimore R&M
Study3
Baltimore Traditional/
MnHiliaH Paint
Abatement Study2
Boston Interim Oust
Intervention Study2
Study
Group
Study 1
(6 homes)
Study 2
(13 homes)
6-Month Follow-up
1 2-Month Follow-up
1 9-Month Follow-up
All Occupied Units
Previously-Abated Units
Units Slated for R&M
Intervention
Modern Urban Units
Traditional
Modified
Automatic Intervention
Randomized Intervention
Pre-lntervention Sill Dust-
Lead Loadings1
Oig/ft2!
15215
2784
356.2
163.5
778.4
45.6
e<)/\Q
787
205
Post-Intervention
Sill Dust-Lead Loadings1
Time Following
Intervention
Immediately
6-9 Months
Immediately
1.5-3.5 Years
Immediately
5-7 Months
Immediately
10- 14 Months
Immediately
14-24 Months
Immediately
2 Months
6 Months
1 2 Months
1 8 Months
24 Months
48 Months
6 Months
1 2 Months
1 8 Months
24 Months
Immediately
2 Months
6 Months
12 Months
1 8 Months
24 Months
6 Months
1 2 Months
1 8 Months
24 Months
48 Months
Immediately
6 Months
Immediately
6 Months
6 Months
6 Months
Summary Value
(f/g/ft2)
737
958
19
199
50
71
50
41
50
147
185.4
241.4
138.2
136.2
135.1
117.5
37.1
107.4
116.0
89.1
97.6
185.4
241.4
247.0
237.6
246.8
204.9
41.7
40.0
40.5
34.8
37.1
11460
4360
1496
4662
210
110
DRAFT - DO NOT CITE OR QUOTE
H-8
August 28. 2000
-------
Table H-3. (cont.)
HUD Grantees
CAP Study4
Study
Group
All Grantees
Baltimore
Boston
Massachusetts
Milwaukee
Minnesota
Rhode Island
Vermont
Wisconsin
Abated Units
Lead Loadings'
1/ig/ft1)
258
1191
174
328
264
266
314
147
150
75
Post-Intervention
Sill Dust-Lead Loadings1
Time Following
Intervention
Immediately
6 Months
1 2 Months
Immediately
6 Months
1 2 Months
Immediately
6 Months
1 2 Months
Immediately
6 Months
1 2 Months
Immediately
6 Months
12 Months
Immediately
6 Months
1 2 Months
Immediately
6 Months
1 2 Months
Immediately
6 Months
1 2 Months
Immediately
6 Months
12 Months
2 Years
1 2 Months
Summary Value
U/g/ft2)
52
97
90
49
87
68
53
48
49
32
77
50
84
231
217
66
86
77
18
87
85
21
60
40
22
37
51
66.4
24
1 Values are geometric means except for the HUD Grantees studies, where values are medians.
2 Results are adjusted to reflect total dust-lead loadings by exponentiating the "bioavailable" dust-lead loadings as reported in
the study to the 1.1416 power.
3 Results for the Baltimore R&M Study are converted from BRM dust-lead loadings to wipe-equivalent loadings.
4 Results for the CAP study are converted from CAPS cyclone dust-lead loadings to wipe-equivalent loadings.
DRAFT - DO NOT CITE OR QUOTE
H-9
August 28, 2000
-------
Floor dust-lead loadings
Table H-2 contains post-intervention floor dust-lead loading summaries for 24 study
groups, including two control groups from the Baltimore R&M study and a total of nine groups
from the HUD Grantees evaluation.
Eighteen study groups in Table H-2 contain information on dust-lead loading
measurements immediately after intervention. Of these 18 groups, 10 had geometric mean or
median dust-lead loadings ranging from 7-24 jig/ft2 immediately after intervention. Eight of
these 10 groups were from the HUD Grantees evaluation, whose pre-intervention median dust-
lead loadings were no higher than 41 ug/ft2. Eight of the 18 groups had geometric mean or
median dust-lead loadings above 40 ug/ft2 immediately after intervention.
Among the nine study groups in the HUD Grantees evaluation, seven groups had median
dust-lead loadings that remained constant or steadily declined to below 20 ug/ft2 for up to 12
months post-intervention. The other two study groups had median loadings increase to
approximately pre-intervention levels over this 12-month period. In addition, the CAP study, the
Baltimore Follow-up Paint Abatement study, the Baltimore R&M study, and Boston Interim
Dust Intervention study, and the CLEARS suggest that geometric mean dust-lead loadings of
below 40 ug/ft2 can be observed for up to two years post-intervention. Only in study #1 of the
Baltimore Experimental Paint Abatement studies and the Baltimore Traditional/Modified Paint
Abatement study did geometric mean dust-lead loadings exceed 40 fig/ft2 at approximately six
months post-intervention; however, pre-intervention levels were higher than in the other studies.
Window sill dust-lead loadings
The same 24 study groups represented in Table H-2 also are included in Table H-3, where
post-intervention window sill dust-lead loading summaries are presented. Results in Table H-3
indicate that post-intervention window sill dust-lead loadings are generally higher (up to double
the value) than those for floors. The post-intervention geometric means (or medians) range from
18 ug/ft2 to over 11,000 ug/ft2.
As in Table H-2, 18 study groups in Table H-3 contain information on dust-lead loading
measurements immediately after intervention. In the nine study groups of the HUD Grantees
evaluation, the three groups of the Baltimore Follow-up Paint Abatement study, and study #2 of
the Baltimore Experimental Paint Abatement studies, geometric mean or median dust-lead
loadings immediately after intervention were below 100 ug/ft2 (range: 18-84 ug/ft2). In
particular, study #2 of the Baltimore Experimental Paint Abatement studies saw a substantial
decline in the geometric mean from pre-intervention (2,784 ug/ft2) to immediately post-
intervention (19 ng/ft2). The remaining five study groups (study #1 of the Baltimore
Experimental Paint Abatement studies, and study groups from the Baltimore R&M study and the
Baltimore Traditional/Modified Paint Abatement study) had geometric mean dust-lead loadings
exceeding 180 ug/ft2 immediately post-intervention, but these groups had geometric mean pre-
intervention dust-lead loadings above 300 ug/ft2.
DRAFT-DO NOT CITE OR QUOTE H-10 August 28.2000
-------
Except for the Milwaukee grantee, the study groups within the HUD Grantees evaluation
had median window sill dust-lead loadings below 100 ug/ft2 for up to 12 months post-
intervention. Only two grantees (Boston and Wisconsin) did not have a decline in median
window sill dust-lead loadings over the 12-month period.
In addition to the HUD Grantees evaluation, geometric mean window sill dust-lead
loadings remain below 100 ug/ft2 for up to 12 months post-intervention in the Baltimore Follow-
up Paint Abatement study, the CAP study, and the CLEARS (Table H-3). However, in studies
such as the Baltimore R&M study, the Baltimore Traditional/Modified Paint Abatement study,
the Baltimore Experimental Paint Abatement studies, and the Boston Interim Dust Intervention
study, geometric mean dust-lead loadings remain above 100 ug/ft2 over time. In addition, the 19-
month follow-up study group within the Baltimore Follow-up Paint Abatement study and the
Baltimore Experimental Paint Abatement studies suggest that geometric mean dust-lead loadings
can dip below 100 ug/ft2 immediately after intervention, but then increase substantially after one
year or so.
The summaries in Tables H-2 and H-3 are calculated across housing units in specified
study groups. With the lack of results for individual housing units and the absence of variability
estimates associated with these summaries, these summaries do not necessarily indicate what
may be occurring in specific units (such as those housing units that see little, if any, change from
pre- to post-intervention). Additional information on results within housing units should also be
considered if such information is available.
H4.0 REFERENCES TO APPENDIX H
Adgate, JL, Weisel, C, Wang, Y, Rhoads, GG, and Lioy, PJ. (1995) "Lead in House Dust:
Relationships between Exposure Metrics." Environmental Research. 70:134-147.
Aschengrau, A, Hardy, S, Mackey, P, and Pultinas, D. (1998) "The Impact of Low Technology
Lead Hazard Reduction Activities among Children with Mildly Elevated Blood Lead
Levels." Environmental Research. 79:41-50.
Chamey, E, Kessler, B, Farfel, M, and Jackson, D. (1983) "Childhood Lead Poisoning: a
Controlled Trial of the Effect of Dust-Control Measures on Blood Lead Levels." New
England Journal of Medicine. 309:1089-1093.
Copley, C. (1995) "East St. Louis Lead Dust Reduction in Homes of At-Risk Children: Pilot
Project." Report on Grant NE995974-01 to Region V of the U.S. Environmental
Protection Agency. 23 October 1995.
Farfel, MR, Chisolm, JJ, and Rohde, CA. (1994) "The Longer-Term Effectiveness of Residential
Lead Paint-Abatement." Environmental Research. 66:199-212.
Farfel, MR, and Chisolm, JJ. (1991) "An Evaluation of Experimental Practices for Abatement of
Residential Lead-Based Paint: Report on a Pilot Project." Environmental Research.
55:199-212.
DRAFT - DO NOT CITE OR QUOTE H-11 August 28.2000
-------
Farfel, MR, and Chisolm, JJ. (1990) "Health and Environmental Outcomes of Traditional and
Modified Practices for Abatement of Residential Lead-Based Paint." American Journal
of Public Health. 80(10): 1240-1245.
HUD (1991) "The HUD Lead-Based Paint Abatement Demonstration (FHA)." Office of Policy
Development and Research, U.S. Department of Housing and Urban Development.
August 1991.
Lanphear, BP, Winter, NL, and Weitzman, M. (1996) "A Randomized Trial of the Effect of
Dust Control on Children's Blood Lead Levels." Pediatrics. 98(1):35.
Lanphear, BP, Winter, NL, Apetz, LJ, Eberly, S, and Weitzman, M. (1995) "A Randomized Trial
of the Effect of Dust Control on Children's Blood Lead Levels." Report on Grant
NYLPR002-94 Submitted to the U.S. Department of Housing and Urban Development
and the National Center for Lead-Safe Housing.
Lioy, PJ, Yiin, L, Adgate, J.Weisel, CP, and Rhoads, GG. (1997) 'The Effectiveness of a Home
Cleaning Intervention Strategy in Reducing Potential Dust and Lead Exposures."
(Submitted to Journal of Exposure Analysis and Environmental Epidemiology).
Mackey, P, Aschengrau, A, Balasko, C, Pultinas, D, and Hardy, S. (1996) "Blood Lead Levels
Following Environmental Intervention Study," Final Report on Grant H64/CCH108235-
03 to the U.S. Department of Health and Human Services, Public Health Service, Centers
for Disease Control and Prevention, Environmental Hazards & Health Effects, Childhood
Lead Poisoning Prevention.
MDE (1995) Final Report on Grant H64/CCH 30 7067-03 by the Maryland Department of the
Environment to U.S. Department of Health and Human Services, Public Health Service,
Centers for Disease Control and Prevention, Environmental Hazards & Health Effects,
Childhood Lead Poisoning Prevention. March 1995.
NCLSH and UC (1998) "Evaluation of the HUD Lead-Based Paint Hazard Control Grant
Program," Fifth Interim Report. Prepared by the National Center for Lead-Safe Housing
and The University of Cincinnati Department of Environment Health for the U.S.
Department of Housing and Urban Development. March 1998.
NCLSH and UC (1997) "Evaluation of the HUD Lead-Based Paint Hazard Control Grant
Program," Fourth Interim Report. Prepared by the National Center for Lead-Safe
Housing and The University of Cincinnati Department of Environment Health for the
U.S. Department of Housing and Urban Development. March 1997.
NCLSH (1993) "A Comparison of Five Sampling Methods for Settled Lead Dust: A Pilot
Study." Preliminary draft report prepared by the National Center for Lead-Safe Housing.
15 June 1993.
DRAFT - DO NOT CITE OR QUOTE H-12 August 28. 2000
-------
Nedellec, V, Fontaine, A, Luciolli, E, and Bourdillon, F. (1995) "Evaluation of Abatement
Interventions in 59 Homes of Lead Poisoned Children," Rev. Epidem. Et Sante Publ.
43:485-493.
Rhoads, GG, Ettinger, AS, Goldman, KD, Weisel, CP, Buckley, TJ, and Lioy, PJ. (1996) "The
Effect of a Dust Lead Control Program Combined with Health Education on Blood Lead
in Toddler: A Randomized Study," (Abstract) Submitted for Presentation at the Annual
Conference of the American Public Health Association, November 1996.
USEPA (1998) "Review of Studies Addressing Lead Abatement Effectiveness: Updated
Edition." Office of Pollution Prevention and Toxics, U.S. Environmental Protection
Agency. EPA 747-B-98-001, December 1998.
USEPA (1997a) "Conversion Equations for Use in Section 403 Rulemaking." Office of
Pollution Prevention and Toxics, U.S. Environmental Protection Agency. EPA 747-R-
96-012, December 1997.
USEPA (1997b) "Lead-Based Paint Abatement and Repair and Maintenance Study in Baltimore:
Findings Based on the First Year of Follow-up." Office of Pollution Prevention and
Toxics, U.S. Environmental Protection Agency. EPA 747-R-97-001, August 1997.
USEPA (1997c) "Lead-Based Paint Abatement and Repair and Maintenance Study in Baltimore:
Findings Based on the Two Years of Follow-up." Office of Pollution Prevention and
Toxics, U.S. Environmental Protection Agency. EPA 747-R-97-005, December 1997.
USEPA (1996a) "Comprehensive Abatement Performance Study. Volume I: Summary
Report." Office of Pollution Prevention and Toxics, U.S. Environmental Protection
Agency. EPA 230-R-94-013a, April 1996.
USEPA (1996b) "Comprehensive Abatement Performance Study. Volume II: Detailed
Statistical Results." Office of Pollution Prevention and Toxics, U.S. Environmental
Protection Agency. EPA230-R-94-013b, April 1996.
USEPA (1996c) "Lead-Based Paint Abatement and Repair and Maintenance Study in Baltimore:
Pre-Intervention Findings." Office of Pollution Prevention and Toxics, U.S.
Environmental Protection Agency. EPA 747-R-95-012, August 1996.
USEPA (1995a) "Review of Studies Addressing Lead Abatement Effectiveness." Office of
Pollution Prevention and Toxics, U.S. Environmental Protection Agency. EPA 747-R-
95-006, July 1995.
USEPA (1987) "Baltimore Integrated Environmental Management Project - Phase II Report:
Reducing the Hazards from Abatement of Lead Paint." Final Report to the U.S.
Environmental Protection Agency.
DRAFT - DO NOT CITE OR QUOTE H-13 August 28, 2000
-------
Vostal, JJ, Taves, E, Sayre, JW, and Chamey, E. (1974) "Lead Analysis of House Dust: A
Method for the Detection of Another Source of Lead Exposure in Inner City Children."
Environmental Health Perspectives. May 1974, 91-97.
DRAFT-DO NOT CITE OR QUOTE H-14 August 28.2000
-------
APPENDIX H2
INFORMATION ON THE INTERVENTION STUDIES INCLUDED IN TABLE H-1
DRAFT - DO NOT CITE OR QUOTE H-15 August 28. 2000
-------
Baltimore (MD) Dust Control Study
• Conducted in 1981 to assess whether lead-based paint abatement followed by periodic
dust control would be more effective in reducing blood-lead concentration than
performing only lead-based paint abatement.
• The study targeted housing units containing lead-based paint and children aged 15-72
months of age with at least two confirmed blood-lead concentration measurements
between 30-49 ug/dL.
• Two groups of housing units (a control group of 35 homes and an experimental group of
14 homes) underwent lead-based paint abatement which entailed removing all peeling
lead-containing interior and exterior paint from the residence. In addition, all child
accessible surfaces (below 1.2 m) which may be chewed on were covered or rendered
lead-free. No extensive clean-up procedures were required following the abatements.
• The experimental group received periodic dust-control (twice-monthly visits by a dust-
control team) involving wet-mopping all rooms in the residence where dust-lead loadings
in an initial survey exceeded 100 ug/ft2.
In the experimental group, dust samples were collected from all areas within the
residence where the child spent time. The samples were collected with alcohol-treated
wipes within a 1 ft2 area of floor or from the entire window sill. The samples were
collected at recruitment and both before and after each dust-control measure was
performed.
Baltimore (MD) Experimental Paint Abatement Studies
Studies to demonstrate and evaluate experimental lead-based paint abatement practices
developed in response to the inadequacies uncovered in the Baltimore (MD)
Traditional/Modified Paint Abatement Study.
• The experimental practices called for floor-to-ceiling abatement of all interior and
exterior surfaces where lead content of the paint exceeded 0.7 mg/cm2 by XRF or 0.5%
by weight by wet chemical analysis. Several methods were tested, including
encapsulation, off-site and on-site stripping, and replacement. The abatements took place
either in unoccupied dwellings or the occupants were relocated during the abatement
process. Lead-contaminated dust was contained and minimized during the abatement,
and extensive clean-up activities included HEPA vacuuming and off-site waste disposal.
In addition, extensive worker training and protection were provided.
One study involving 6 housing units (poorly-maintained, had multiple lead-based paint
hazards, built in the 1920s) received abatements from 10/86-1/87 as part of a pilot study
examining the experimental procedures. Four units were vacant, and two contained lead-
poisoned children. This study evaluated short-term abatement efficacy (up to 9 months).
DRAFT - DO NOT CITE OR QUOTE H-16 August 28. 2000
-------
Dust samples from the 6 housing units were collected immediately before abatement,
during abatement, after the final clean-up, and at 1, 3, and 6-9 months following
abatement.
Another study which evaluated longer-term abatement efficacy (1.5-3.5 years) involved
13 occupied housing units which received experimental abatements from 1988-1991 by
local pilot projects.
Dust samples from the 13 housing units were collected from 12/91 - 01/92 at the same
locations, where possible, that had been sampled pre- and immediately post-abatement.
Alcohol-treated wet wipes were used to collect dust samples.
Baltimore (MD! Follow-up Paint Abatement Study
• Paint interventions (encapsulation, off-site and on-site stripping, and replacement) were
performed (from floor to ceiling) on all interior and exterior surfaces where lead content
of paint exceeded 0.7 mg/cm2 by XRF or 0.5% by weight by wet chemical analysis.
Abatements took place in unoccupied dwellings or after occupants were relocated.
• Lead-contaminated dust was contained and minimized during the abatement.
• Extensive clean-up activities (including HEPA vacuuming and off-site waste disposal)
followed the abatement to ensure clearance. Clearance levels for floors, window sills,
and window wells were set at 200 ug/ft2, 500 ug/ft2, and 800 ng/ft2, respectively.
• Wipe dust-lead loading samples were taken upon clearance and at approximately 6,12,
and 19 months post-intervention from floors, window sills, and window wells in rooms
where the child spent time.
• By 19 months post-intervention, only 5% of the homes were above clearance for floors,
while 42% and 47% of the homes were above clearance levels for window sills and
window wells, respectively.
Baltimore (MD) Repair & Maintenance (R&JVD Study
• Study begun in 1993 to measure the short-term (2 to 6 months) and long-term (12 to 24
months) changes in dust-lead loadings and concentrations and in children's blood lead
concentrations associated with conducting R&M interventions, and to make comparisons
with houses that had undergone previous comprehensive abatement, as well as a group of
modem urban houses.
DRAFT-DO NOT CITE OR QUOTE H-17 August 28,2000
-------
Three types of dwellings were recruited in this study: 16 dwellings that were previously
abated (in 1988-1992), 75 dwellings slated to receive R&M interventions, and 16 modem
urban dwellings (assumed to be free of lead-based paint).
The 75 R&M dwellings were older (mostly pre-1940), low-income dwellings which were
divided into three equal groups according to the intervention performed in this study; the
R&M-I group had low-level interventions (wet scraping, limited repainting, wet cleaning
with TSP, HEPA vacuuming, placing an entryway mat, exterior surface stabilization,
cleaning supplies and education to residents), the R&M-n group had intermediate-level
interventions (R&M-I interventions plus treatments to floors, windows, and doors to
reduce abrasion), and the R&M-EII group had high-level interventions (R&M-II
interventions plus trim replacement and encapsulation). The remaining dwellings acted
as control dwellings.
The BRM vacuum method was used to collect dust samples in this study (a modified
HVS3 cyclone collector). Floor and window sill dust samples were composites across
multiple rooms. The environmental sampling design was as follows:
Campaign
Initial
Immediate Post-R&M
2 Months Post-R&M
6 Months Post-R&M
1 2 Months Post-R&M
1 8 Months Post-R&M
24 Months Post-R&M
Type of Data '
Blood
RM2
/•
/
/
/
/
/
/
Control 3
/
/
V
/
/
Oust
RM
/
/
/
V
/
/
/
Control
V
/
/
/
V
Soil
RM
/
/
/
/
Control
V
/
V
Water
RM
/•
/*
/
/
Control
/
/
V
1. A '/' indicates that the data were collected for all R&M groups or all control groups. Symbol'/" indicates that
data collected only for R&M I and II groups, and '/" ' only for R&M II and III.
2. RM denotes the component including three R&M groups: R&M I, R&M II and R&M III.
3. Control denotes the component including two control groups: Previously Abated and Modern Urban.
DRAFT - DO NOT CITE OR QUOTE
H-18
August 28, 2000
-------
Baltimore (MD) Traditional/Modified Paint Abatement Study
• Conducted from 1984-1985 to evaluate the health and environmental impact of
"traditional" and "modified" Baltimore practices for abating lead-based paint.
• The study contained housing units with multiple interior surfaces coated with lead-based
paint and containing at least one child with a blood-lead concentration exceeding 30
ug/dL.
• "Traditional" abatements (conducted in 53 housing units) addressed deteriorated paint on
surfaces up to four feet from the floor, and all hazardous paint on accessible surfaces
which may be chewed on. Paint with a lead content greater than 0.7 mg/cm2 by XRF or
0.5% by weight by wet chemical analysis was denoted hazardous. Open-flame burning
and sanding techniques were commonly used, the abated surfaces were not repainted, and
clean-up typically entailed, at most, dry sweeping.
"Modified" abatements (conducted in 18 housing units) included the use of heat guns for
paint removal and the repainting of abated surfaces. Furnishings were protected during
abatement. In addition, clean-up efforts were conducted that involved wet-mopping with
a high phosphate detergent, vacuuming with a standard shop vacuum, and off-site
disposal of debris. In addition, worker training, protection, and supervision were
provided.
• Neither traditional nor modified abatements considered window wells.
• Dust samples were obtained using a alcohol-treated wipe within a defined area template
(1
Increased dust-lead loadings were measured immediately following traditional
abatements (usually within two days) on or in close proximity to abated surfaces. Dust-
lead levels measured after modified abatements were also higher than pre-abatement
levels, but not to the extent seen for traditional practices. At six months post-abatement,
PbD levels were comparable to, or greater than, their respective pre-abatement loadings
in both study groups.
Despite the implementation of improved practices, modified abatements, like traditional
abatements, did not result in any long-term reductions of levels of lead in house dust. In
addition, the activities further elevated blood-lead concentrations.
Boston (MA) Interim Dust Intervention Study
Children under 4 years of age with modestly-elevated blood-lead concentration (11 -24
ug/dL) and living in homes containing lead-based paint on at least two window sills or
wells were targeted for participation. Lead hazard reduction activities were not
previously conducted in these homes.
DRAFT-DO NOT CITE OR QUOTE H-19 August 28,2000
-------
Units with severe household lead hazards (i.e., paint chips on floors, large amounts of
loose dust or paint chips in window wells, or holes larger than one inch wide in walls
containing lead-based paint) were placed into an "automatic intervention" group (n=22).
Remaining units were randomly assigned to a "randomized intervention" group (n=22) or
a "randomized comparison" group (n=19).
Units in the two intervention groups received a one-time paint and/or dust intervention.
The intervention was considered "low-technology" and consisted of HEP A vacuuming all
window well, window sill, and floor surfaces; washing window well and window sill
surfaces with a tri-sodium phosphate (TSP) and water solution; repairing holes in walls;
and re-painting window well and window sill surfaces to seal chipping or peeling paint.
These units also received outreach and educational information including a demonstration
of effective housekeeping techniques and monthly reminders with instructions to wash
hard surface floors, window sills and wells with a TSP and water solution at least twice a
week.
The "randomized comparison" group received only the outreach visit, in which the home
was visually assessed for lead hazards and the family was educated about the causes and
prevention of lead poisoning. They were also provided with cleaning instructions and a
free sample of TSP cleaning solution.
16 study units had permanent lead-based paint hazard remediation performed outside of
the study protocol during the 6-month follow-up period. It is uncertain whether data for
these units were treated differently in the study as a result.
Dust samples were collected from floors, window sills, and window wells at baseline and
6 months post-intervention in all units, and at one month post-intervention for the two
intervention groups. However, results were not reported for the one-month post-
intervention campaign.
Dust, soil, and water samples were analyzed using atomic absorption spectrophotometry
(AAS). The detection limit for dust-lead loading results was 30 ug/ft2.
At 6 months post-intervention, geometric mean floor dust-lead loadings had decreased
slightly for both intervention groups and increased in the comparison group. Geometric
mean window sill dust-lead loadings decreased in all three groups, and geometric mean
window well dust-lead loadings decreased for both intervention groups, but remained the
same for the comparison group. None of the changes in dust-lead loadings was
statistically significant.
DRAFT - DO NOT CITE OR QUOTE H-20 August 28.2000
-------
East St. Louis (ID Educational Intervention Study
• Children were identified for the study through a screening program for children receiving
public assistance.
• During an initial visit to each home, an XRF paint survey was conducted, and areas with
high lead loadings were pointed out to the residents. Lead educators, hired from within
the community, provided instruction on cleaning and hygiene. Written materials and a
videotape on reducing lead exposure were also provided.
• At the initial visit and at three months following this visit (for homes where the residents
reported that they had cleaned at least once using the recommended procedures), at least
three dust-wipe samples were collected from smooth surfaces in the home. In addition,
two vacuum (HVS-3) samples were collected from carpets in the entry and bedroom, or
other play area.
• Participating families were contacted regularly throughout the course of the study to
reinforce the importance of regular, thorough cleaning.
Evaluation of the HUD Lead-Based Paint Hazard Control Grant Program (HUD Grantees)
• A formal evaluation of this ongoing study is being conducted to determine the
effectiveness of various abatement methods used by State and local governments (who
are HUD grantees) to reduce lead-based paint hazards in housing.
Data collection began in 1994 and is expected to continue through 1999.
• Enrollment criteria varied among the different grantees and included targeting high-risk
neighborhoods, homes with a lead-poisoned child, and unsolicited applications.
Grantees were given the flexibility to select the type and intensity of the lead treatments
for any particular unit. The intensity of an intervention is reported by location (interior,
exterior, or site) and consists of a number representing the type of intervention performed
in that location. The interventions range from taking no action, to a simple cleaning, to
window replacement or full lead-based paint abatement. Some interim controls on soil
(e.g., cover), as well as soil removal, were also performed.
• The grantees followed the same sampling protocols when collecting environmental
samples (including dust using wipe techniques) and used standard forms developed
specifically for the evaluation.
• Dust samples are collected from occupied housing units at four times during the study: at
pre-intervention, immediately after intervention, and at 6 and 12 months following
intervention. Nine of the 14 grantees participating in this evaluation are also collecting
data at 24 and 36 months following intervention (these data have not yet been collected).
DRAFT-DO NOT CITE OR QUOTE H-21 August 28.2000
-------
HUD Abatement Demonstration Program/
EPA Comprehensive Abatement Performance (CAP) Study
The FHA portion of the HUD Abatement Demonstration Program ("HUD Demo") was
conducted to estimated the comparative costs of alternative methods of lead-based paint
abatement, to assess the efficacy of these methods, and to confirm the adequacy of
worker protection safeguards during abatement.
• In the HUD Demo, lead-based paint abatements were performed in 172 HUD-owned,
single-family properties located in seven cities across the country.
• Wipe dust samples were collected immediately following intervention and cleaning in the
HUD Demo to evaluate whether lead levels were below 200 ug/ft2 for floors and 500
u.g/ft2 for window sills. Repeated iterations of cleaning and dust sampling were
performed if additional cleaning was deemed necessary.
• The CAP study was a follow-up to the HUD Demo performed in Denver, CO. The
objectives of the CAP study were to assess the long-term efficacy of two primary
abatement methods (encapsulation/enclosure and removal methods), to characterize lead
levels in dust and soil in unabated homes and homes abated by different methods, to
investigate the relationship between household dust-lead and lead from other sources (i.e.,
soil and air ducts), and to compare dust-lead loading results from cyclone vacuum
sampling and wipe sampling protocols.
• The CAP study collected approximately 30 dust and soil samples at each of 52 occupied
houses in Denver. Of these houses, 39 had lead-based paint abatements performed
approximately two years earlier as part of the HUD Demo. The remaining 17 houses
were considered within the HUD Demo, but were found to be free of lead-based paint and
therefore had no abatements performed.
• The CAP study used a cyclone vacuum for collecting dust samples, where this vacuum
was designed especially for this study. Dust samples were collected from the floor
perimeter, window sills, window wells, entryway floors, and air ducts in either two or
three rooms. Some wipe dust samples were also collected to make comparisons between
wipe and vacuum dust-lead loadings.
• For window sills within 10 houses, pre-abatement dust-lead loadings and loadings
measured during the CAP study both averaged between 175-200 jig/ft2 (i.e., there was no
evidence of significant differences between pre- and post-intervention dust-lead
loadings). However, no adjustment was made between the wipe and vacuum methods
used in pre- and post-intervention, respectively. A similar comparison between pre- and
post-intervention dust-lead loadings for floors was not possible due to a lack of sufficient
pre-intervention data.
DRAFT - DO NOT CITE OR QUOTE H-22 August 28,2000
-------
Abatements were found to be effective in that no significant difference in dust-lead
loadings were observed between abated and unabated units in the CAP study (with the
exception of dust from air ducts).
Jersey City (NJ) Children's Lead Exposure and Reduction (CLEAR) Dust Intervention
Study
Children under 3 years of age and at risk for elevated blood-lead concentration were
targeted for participation.
• Lead hazard intervention consisted of biweekly assistance with home dust control (which
included wet mopping of floors, damp-sponging of walls and horizontal surfaces, and
HEP A vacuuming) and a series of educational sessions about lead. The cleaning teams
provided the education during the course of their visits and mainly focused on teaching
the caretakers how to clean the home.
• Dust-wipe samples were collected from uncarpeted floors in the kitchen and the floor of
one other room frequented by the enrolled child.
• This analysis indicated that a thorough cleaning program reduced the geometric mean of
the dust and lead loading and found that 68%, 75%, and 81% of the Lead Group (Study)
homes had a reduction in lead loading on the kitchen floors, bedroom floors, and window
sills, respectively.
Paris Paint Abatement Study
Children less than 6 years of age, identified as severely lead-poisoned, and living in
homes with lead-based paint were targeted for participation.
A one-time paint intervention was performed, consisting of chemical stripping with
caustic products, encapsulation (consisting of covering the toxic paint with coating
material which prevents the dispersion of chips and particles into the home), replacement
of antiquated elements and paint coatings of lead-based paints, and a final dust cleaning.
Chemical stripping, using Peel Away™, was used on 52% of the items abated, a
combination of stripping and encapsulation was used on 36% of the items abated, and a
combination of encapsulation and replacement was used on 12% of the abated items.
Families were relocated during abatement.
• Dust samples were collected in 29 homes at baseline, during the intervention, and at 1 to
2 months, 3 to 6 months, and 7 to 12 months post-intervention. Dust sampling was done
by wiping the floor 1 meter from the wall, over an area of 30x30 cm2, with a paper towel
impregnated with alcohol.
DRAFT-DO NOT CITE OR QUOTE H-23 August 28.2000
-------
For 11 homes having an initial dust-lead loading greater than 92.9 ng/ft2, median
decreases were 144 ug/ft2 at 1 to 2 months follow up and 157 fig/ft* at 3 to 6 months post-
intervention.
By 6 to 28 months post-intervention, the maximum dust-lead loadings were less than 92.9
ug/ft2 for 40 out of 45 households.
Rochester fiVY) Educational Intervention Study
Included 104 of the 205 children in the Rochester Lead-in-Dust study, aged 12-31 months
at enrollment, with low to moderate blood-lead concentration. Households were
randomly assigned to an intervention or control group.
• Aim of the study was to determine the effectiveness of simple dust control by household
members as a means of reducing children's blood-lead concentration.
A trained interviewer visited families assigned to the intervention group. The interviewer
stressed the importance of dust control as a means of reducing lead exposure and
provided the household with cleaning supplies (paper towels, spray bottles and Ledisolv,
a detergent developed specifically for lead contaminated house dust). Families were
instructed to clean the entire house once every three months, interior window sills,
window wells and floors near windows once every month, and carpets once a week with a
vacuum cleaner, if available.
• For families assigned to the control group, only a brochure was provided containing
information about lead poisoning and its prevention.
Dust samples (using a K-mart brand of baby wipes) were collected at the time of the
home visit (baseline) and at seven months following the visit. Locations of dust samples
included entryway floors and the kitchen, as well as from the floors, interior window sills
and window wells of the child's principal play area.
DRAFT - DO NOT CITE OR QUOTE H-24 August 28,2000
-------
APPENDIX H3
DETAILED SUMMARY TABLES
DRAFT-DO NOT CITE OR QUOTE H-25 August 28.2000
-------
Table H3-1. Summary of Floor Dust-Lead Loadings, Under Wipe Dust Sampling Techniques, at Pre- and Post-Intervention
Name of Study
Baltimore
Experimental
Paint
Abatement
Studies2
Baltimore
Follow-up Paint
Abatement
Study2
Group of
Housing
Units Within
the Study
Study 1
(6 homes)
Study 2
(13 homes)
6-Month
Follow-up
1 2-Month
Follow-up
1 9-Month
Follow-up
Pre-lntervention Floor
Dust-Lead Loadings
N
70
42
Type of
Statistic1
GM
(95% Cl)
GM
(95% Cl)
Value of
Statistic
(/ig/ft2)
1261
(908,
1761)
556
(289.
1074)
Post-Intervention Floor Dust-Lead
Loadings
Time
Following
Intervention
Immediately
6-9 Months
Immediately
1.5-3.5
Years
Immediately
Following
Clearance
5-7
Months
Immediately
Following
Clearance
10-14
Months
Immediately
Following
Clearance
14-24
Months
N
70
63
47
71
29
27
22
Type of
Statistic1
GM
(95% Cl)
GM
(95% Cl)
GM
(95% Cl)
GM
(95% Cl)
GM
(95% Cl)
GM
(95% Cl)
GM
(95% Cl)
GM
(95% Cl)
GM
(95% Cl)
GM
(95% Cl)
Value of
Statistic
(//g/ft2)
259
(196.
366)
99
(79, 136)
20
(9.8, 40)
69
(40, 125)
29
(20.41)
22
(15.31)
41
(25. 63)
20
(15.29)
24
(14. 38)
36
(20, 63)
Difference from Pre-lntervention
Time Value of
Following Type of Statistic
Intervention N Statistic1 (//g/ft2)
o
m
O
30
O
I
m
I
«
M
-------
Table H3-1. (cont.)
Name of Study
Baltimore R&M
Study3
Housing
Units Within
the Study
All Occupied
Units
Previously-
Abated Units
Units Slated
for R&M
Intervention
Pre-lntervention Floor
Dust-Lead Loadings
N
90
16
58
Type of
Statistic1
GM
GM
GM
Value of
Statistic
(//g/ft2)
40.9
45.6
58.6
Post-Intervention Floor Dust-Lead
Loadings
Time
Following
Intervention
Immediately
2 Months
6 Months
1 2 Months
1 8 Months
24 Months
48 Months
6 Months
1 2 Months
1 8 Months
24 Months
Immediately
2 Months
6 Months
1 2 Months
1 8 Months
24 Months
N
37
37
66
66
64
62
7
14
14
13
13
37
37
37
37
37
35
Type of
Statistic1
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
Value of
Statistic
(//g/ft2)
52.5
40.2
26.5
27.1
24.8
24.1
8.4
41.1
39.8
37.3
33.0
52.5
40.2
36.3
39.9
33.3
35.0
Difference from Pre-lntervention
Time Value of
Following Type of Statistic
Intervention N Statistic1 (//g/ft2)
8
o
nl
O
70
O
m
£
(Q
o
o
-------
Table H3-1. (cont.)
Name of Study
Baltimore R&M
Study3
Baltimore
Traditional/
Modified Paint
Abatement
Study2
Boston Interim
Dust
Intervention
Study2
East St. Louis
Educational
Intervention
Study2
Group of
Housing
Units Within
the Study
Modern
Urban Units
Traditional
Modified
Automatic
Intervention
Randomized
Intervention
Intervention
Group
Pre-lntervention Floor
Dust-Lead Loadings
N
16
280
82
10
9
30
Type of
Statistic1
GM
GM
(95% Cl)
GM
(95% Cl)
GM
GM
AM
GM
Value of
Statistic
(/ig/ft2)
10.0
549
(482.
645)
642
(433,
908)
33
37
1095
1080
Post-Intervention Floor Dust-Lead
Loadings
Time
Following
Intervention
6 Months
1 2 Months
1 8 Months
24 Months
48 Months
Immediately
6 Months
Immediately
6 Months
6 Months
6 Months
3 Months
N
15
15
14
14
7
271
234
50
57
10
9
30
Type of
Statistic1
GM
GM
GM
GM
GM
GM
(95% Cl)
GM
(95% Cl)
GM
(95% Cl)
GM
(95% Cl)
GM
GM
AM
GM
Value of
Statistic
(//g/ft2)
8.1
7.3
7.8
7.1
8.4
4033
(3269,
4936)
714
(594,
834)
1626
(1082,
2418)
714
(526,
983)
24
31
486
103
Difference from Pre-lntervention
Time Value of
Following Type of Statistic
Intervention N Statistic1 (//g/ft2)
-56%
1 MnnlliL in PerC6nt
Chan9e -87%
o
O
1
O
o
3J
O
§
m
to
oo
a.
o
o
-------
Table H3-1. (cont.)
Name of Study
HUD Grantees
Group of
Housing
Units Within
the Study
All Grantees
Baltimore
Boston
Mass.
Milwaukee
Pre-lntervention Floor
Dust-Lead Loadings
N
557
32
28
42
170
Type of
Statistic1
Median
Median
Median
Median
Median
Value of
Statistic
(//g/ft2)
19
41
24
24
14
Post-Intervention Floor Dust-Lead
Loadings
Time
Following
Intervention
Immediately
Post
6 Months
1 2 Months
Immediately
Post
6 Months
1 2 Months
Immediately
Post
6 Months
1 2 Months
Immediately
Post
6 Months
12 Months
Immediately
Post
6 Months
1 2 Months
N
557
32
28
42
170
Type of
Statistic1
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Value of
Statistic
-------
Table H3-1. (cont.)
Name of Study
HUD Grantees
CAP study4
Jersey City
(NJ) CLEARS
Group of
Housing
Units Within
the Study
Minnesota
Rhode Island
Vermont
Wisconsin
Unabated
homes
Abated
homes
Intervention
Group
Pre-lntervention Floor
Dust-Lead Loadings
N
105
31
43
48
Type of
Statistic1
Median
Median
Median
Median
Value of
Statistic
(/ig/ft2)
18
26
28
9
42
GM
22
Post-Intervention Floor Dust-Lead
Loadings
Time
Following
Intervention
Immediately
Post
6 Months
12 Months
Immediately
Post
6 Months
1 2 Months
Immediately
Post
6 Months
12 Months
Immediately
Post
6 Months
1 2 Months
2 years
2 years
12 Months
N
10b
31
43
48
51
187
40
Type of
Statistic1
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
GM
(25th %ile)
(75th %ile)
GM
25th %ile
75th %ile
GM
Value of
Statistic
(//g/ft2)
18
18
18
7
6
6
17
21
21
8
6
5
15
(4.1)
(47)
21
(4.9)
(76}
15
Difference from Pre-lntervention
Time Value of
Following Type of Statistic
Intervention N Statistic1 (/ig/ft2)
I
1
o
o
7)
D
(Q
I
«
o
o
-------
Table H3-1. (cont.)
o
o
o
m
O
30
O
Q
Name of Study
Paris Paint
Abatement
Study
Rochester
Educational
Intervention
Study
Group of
Housing
Units Within
the Study
Intervention
Group
Intervention
Group -
Uncarpeted
Floors
Intervention
Group -
Carpeted
Floors
Pre-lntervention Floor
Dust-Lead Loadings
N
24
Type of
Statistic1
Median
Value of
Statistic
-------
o
o
I
o
m
O
O
I
Table H3-2. Summary of Window Sill Dust-Lead Loadings, Under Wipe Dust Sampling Techniques, at Pre- and Post-
Intervention
Name of Study
Baltimore
Experimental
Paint
Abatement
Studies2
Baltimore
Follow-up Paint
Abatement
Study2
firnun of
Housing
Units Within
the Study
Study 1
(6 homes)
Study 2
(13 homes)
6-Month
Follow-up
1 2-Month
Follow-up
Pre-lntervention Sill Dust-
Lead Results
N
34
53
Type of
Statistic1
GM
(95% CD
GM
(95% CD
Value of
Statistic
(/ig/ft2)
15215
(9389,
24618)
2784
(1322,
5891)
Post-Intervention Sill Dust-Lead Results
Time
Following
Intervention
Immediately
Post
6-9 Months
Immediately
Post
1.5-3.5
Years
Immediately
Following
Clearance
5-7
Months
Immediately
Following
Clearance
10-14
Months
N
35
31
54
59
27
27
26
26
Type of
Statistic1
GM
(95% CD
GM
(95% CD
GM
(95% CD
GM
(95% CD
GM
(95% CD
GM
(95% CD
GM
(95% CD
GM
(95% CD
Value of
Statistic
(/ig/ft2)
737
(411,
1364)
958
(526.
1681)
19
(9.8, 35)
199
(119,
331)
50
(32, 81)
71
(43, 119)
50
(31,81)
41
(49, 132)
Difference from Pre-lntervention
Time Value of
Following Type of Statistic
Intervention N Statistic1 (pg/ft2)
I
IN)
IO
CO
8
8
-------
Table H3-2. (cont.)
Name of Study
Baltimore
Follow-up Paint
Abatement
Study2
Baltimore R&M
Study3
Group of
Housing
Units Within
the Study
1 9-Month
Follow-up
All Occupied
Units
Previously-
Abated Units
Units Slated
for R&M
Intervention
Pre-lntervention Sill Dust-
Lead Results
N
90
16
58
Type of
Statistic1
GM
GM
GM
Value of
Statistic
(//g/ft2)
356.2
163.5
778.4
Post-Intervention Sill Dust-Lead Results
Time
Following
Intervention
Immediately
Following
Clearance
14-24
Months
Immediately
2 Months
6 Months
1 2 Months
1 8 Months
24 Months
48 Months
6 Months
1 2 Months
1 8 Months
24 Months
Immediately
2 Months
6 Months
1 2 Months
18 Months
24 Months
N
19
19
37
37
66
66
64
62
7
14
14
13
13
37
37
37
37
37
35
Type of
Statistic1
GM
(95% CD
GM
(95% CD
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
GM
Value of
Statistic
(//g/ft2)
50
(19. 52)
147
(66, 324)
185.4
241.4
138.2
136.2
135.1
117.5
37.1
107.4
116.0
89.1
97.6
185.4
241.4
247.0
237.6
246.8
204.9
Difference from Pre-lntervention
Time Value of
Following Type of Statistic
Intervention N Statistic1 l/ig/ft2)
o
o
1
o
m
o
30
O
I
*
to
I
O
O
-------
Table H3-2. (cont.)
Name of Study
Baltimore R&M
Study3
Baltimore
Traditional/
Modified Paint
Abatement
Study2
Boston Interim
Dust
Intervention
Study2
Group of
Housing
Units Within
the Study
Modern
Urban Units
Traditional
Modified
Automatic
Intervention
Randomized
Intervention
Pro-Intervention Sill Dust-
Lead Results
N
16
249
45
10
9
Type of
Statistic1
GM
GM
(95% CD
GM
(95% CD
GM
GM
Value of
Statistic
(/ig/ft2)
45.6
3708
(2953,
4600)
5209
(3765,
7246)
787
205
Post-Intervention Sill Dust-Lead Results
Time
Following
Intervention
6 Months
1 2 Months
1 8 Months
24 Months
48 Months
Immediately
Post
6 Months
Immediately
Post
6 Months
6 Months
6 Months
N
15
15
14
14
7
246
199
64
66
10
9
Type of
Statistic1
GM
GM
GM
GM
GM
GM
(95% CD
GM
(95% CD
GM
(95% Cl)
GM
(95% Cl)
GM
GM
Value of
Statistic
Uig/ft2)
41.7
40.0
40.5
34.8
37.1
11460
(8929,
14654)
4360
(3356,
5674)
1496
(1058,
2114)
4662
(3126,
6961)
210
110
Difference from Pre-lntervention
Time Value of
Following Type of Statistic
Intervention N Statistic1 (//g/ft2)
o
o
1
o
m
O
o
I
o
o
-------
Table H3-2. (cont.)
Name of Study
HUD Grantees
Group of
Housing
Units Within
the Study
All Grantees
Baltimore
Boston
Mass.
Milwaukee
Pre-lntervention Sill Dust-
Lead Results
N
547
32
29
43
166
Type of
Statistic1
Median
Median
Median
Median
Median
Value of
Statistic
(/ig/ft2)
258
1191
174
328
264
Post-Intervention Sill Dust-Lead Results
Time
Following
Intervention
Immediately
Post
6 Months
1 2 Months
Immediately
Post
6 Months
1 2 Months
Immediately
Post
6 Months
1 2 Months
Immediately
Post
6 Months
1 2 Months
Immediately
Post
6 Months
1 2 Months
N
547
32
29
43
166
Type of
Statistic1
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Value of
Statistic
l/ig/ft2)
52
97
90
49
87
68
53
48
49
32
77
50
84
231
217
Difference from Pre-lntervention
Time Value of
Following Type of Statistic
Intervention N Statistic1 U/g/ft2)
Immediately Bn,»
Post Median "ou%
6Mon,hs «' ££ -62*
12 Months -65%
o
o
1
o
o
73
O
I
m
to
CO
-------
Table H3-2. (cont.)
Name of Study
HUD Grantees
CAP study4
Jersey City
(NJ) CLEARS
Group of
Housing
Units Within
the Study
Minnesota
Rhode Island
Vermont
Wisconsin
Unabated
homes
Abated
homes
Intervention
Group
Pre-lntervention Sill Dust-
Lead Results
N
108
31
32
45
Type of
Statistic1
Median
Median
Median
Median
Value of
Statistic
(//g/ft2)
266
314
147
150
39
GM
75
Post-Intervention Sill Dust-Lead Results
Time
Following
Intervention
Immediately
Post
6 Months
1 2 Months
Immediately
Post
6 Months
1 2 Months
Immediately
Post
6 Months
1 2 Months
Immediately
Post
6 Months
1 2 Months
2 years
2 years
1 2 Months
N
108
31
32
4b
38
78
36
Type of
Statistic1
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
Median
GM
(25th %ile)
(75th %ile)
GM
25th %ile
75lh %ile
GM
Value of
Statistic
(//g/ft2)
66
86
77
18
87
85
21
60
40
22
37
51
34
(7.1)
(163)
66
(11)
(339)
24
Difference from Pre-lntervention
Time Value of
Following Type of Statistic
Intervention N Statistic1 U/g/ft2)
o
o
1
o
m
O
30
O
I
m
M
00
8
8
-------
Table H3-2. (cont.)
-<
1
3
o
m
u
7i
O
3
m
Name of Study
Rochester
Educational
Intervention
Study
firm in of
Housing
Units Within
the Study
Intervention
Group
Pre-lntervention
Sill Dust-
Lead Results
Type of
N Statistic1
Value of
Statistic
(//g/ft2)
Post-Intervention Sill Dust-Lead Results
Time
Following
Intervention N £
Value of
Type of Statistic
Statistic1 (//g/ft2)
•
Difference from Pre-lntervention
Time
Following
Intervention
7 Months
N
80
Type of
Statistic1
Median
Absolute
Change
(IQ
Range)
Value of
Statistic
(//g/ft2)
-58
(-154.-10)
' GM = geometric mean. AM = arithmetic mean Cl = Confidence Interval.
2 Results (for geometric means and medians ONLY) are adjusted to reflect total dust-lead loadings by exponentiating the "bioavailable" dust-lead loadings as reported in the study to
the 1.1416 power.
3 Results for the Baltimore R&M Study are converted from BRM dust-lead loadings to wipe-equivalent loadings.
4 Results for the CAP study are converted from CAPS cyclone dust-lead loadings to wipe-equivalent loadings.
CO
I
o
o
-------
|