EPA Contract EP-W-09-024
Work Assignment 1-06
December 2010
Environmental Technology
Verification Program
Environmental and Sustainable
Technology Evaluations
Report
SILVER LAKE RESEARCH
LEADAVERT™ TEST KIT
QUALITATIVE SPOT TEST KIT FOR LEAD IN PAINT
Prepared by
Battelle
Baireiie
The HusiiiesiS of Innovation
for
v/EPA
U.S. Environmental Protection Agency
ET1/ET1/ET1/
-------
EPA Contract EP-W-09-024
Work Assignment 1-06
December 2010
Environmental Technology
Verification Program
Environmental and Sustainable
Technology Evaluations
Report
SILVER LAKE RESEARCH
LEADAVERT™ TEST KIT
QUALITATIVE SPOT TEST KIT FOR LEAD IN PAINT
by
Stephanie Buehler, Dale Rhoda, and Bruce Buxton Battelle
Julius Enriquez and Evelyn Hartzell, U.S. EPA
Battelle
Columbus, Ohio 43201
-------
Notice
Funding for this verification test was provided under Contract No. EP-W-09-024, Work
Assignments 4-16, 0-06, and 1-06, Office of Pollution Prevention, and Toxics, US EPA. The U.S.
Environmental Protection Agency, through its Office of Research and Development, managed
the research described herein. It has been subjected to the Agency's peer and administrative
review and has been approved for publication. Any opinions expressed in this report are those of
the author(s) and do not necessarily reflect the views of the Agency, therefore, no official
endorsement should be inferred. Any mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
11
-------
Foreword
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting the
Nation's land, air, and water resources. Under a mandate of national environmental laws, the
Agency strives to formulate and implement actions leading to a compatible balance between
human activities and the ability of natural systems to support and nurture life. To meet this
mandate, EPA's research program is providing data and technical support for solving
environmental problems today and building a science knowledge base necessary to manage our
ecological resources wisely, understand how pollutants affect our health, and prevent or reduce
environmental risks in the future.
The National Risk Management Research Laboratory (NRMRL) is the Agency's center for
investigation of technological and management approaches for preventing and reducing risks
from pollution that threaten human health and the environment. The focus of the Laboratory's
research program is on methods and their cost-effectiveness for prevention and control of
pollution to air, land, water, and subsurface resources; protection of water quality in public water
systems; remediation of contaminated sites, sediments and groundwater; prevention and control
of indoor air pollution; and restoration of ecosystems. NRMRL collaborates with both public
and private sector partners to foster technologies that reduce the cost of compliance and to
anticipate emerging problems. NRMRL's research provides solutions to environmental
problems by: developing and promoting technologies that protect and improve the environment;
advancing scientific and engineering information to support regulatory and policy decisions; and
providing the technical support and information transfer to ensure implementation of
environmental regulations and strategies at the national, state, and community levels.
This publication has been produced as part of the Laboratory's strategic long-term research plan.
It is published and made available by EPA's Office of Research and Development to assist the
user community and to link researchers with their clients.
Sally Gutierrez, Director
National Risk Management Research Laboratory
in
-------
Acknowledgments
The authors wish to acknowledge the support of all those who helped plan and conduct the
verification test, analyze the data, and prepare this report. We also would like to thank Al
Liabastre, ret. U.S. Army Center for Health Promotion and Preventive Medicine; David Jacobs,
National Center for Healthy Housing; Kenn White, American Industrial Hygiene Association;
Larry Franklin, Centers for Disease Control and Prevention, Coordinating Center for
Environmental Health and Injury Prevention, National Center for Environmental Health, Lead
Poisoning Prevention Branch; and Moira Lataille and Michael Crane, U.S. EPA for their careful
review of the test/quality assurance plan and this verification report. Quality assurance oversight
was provided by Michelle Henderson, U.S. EPA, and Zachary Willenberg and Rosanna Buhl,
Battelle.
IV
-------
Contents
Notice ii
Foreword iii
Acknowledgments iv
List of Abbreviations and Acronyms viii
Chapter 1 Background 1
Chapter 2 Technology Description 2
Chapters Test Design and Procedures 3
3.1 Introduction 3
3.2 Test Facility 4
3.3 Test Procedures 4
3.3.1 Test Sample Collection, Storage, and Shipment 7
3.3.2 Test Sample Analysis Procedure 7
Chapter 4 Quality Assurance/Quality Control 9
4.1 Quality Control Samples 9
4.1.1 ICP-AES Blank Sample Results 9
4.1.2 ICP-AES Matrix Spike Samples and Calibration Verification Standards 9
4.1.3 Test Kit Quality Controls and Blank PEMs 10
4.2 Audits 11
4.2.1 Performance Evaluation Audits 11
4.2.2 Technical Systems Audit 11
4.2.3 Audit of Data Quality 11
Chapters Statistical Methods 13
5.1 False Positive and False Negative Rates 13
5.2 Precision 14
5.3 Sensitivity 14
5.4 Modeled Probability of Test Kit Response 14
5.4.2 Accounting for Measurement Error - SIMEX Background and Intuition 15
5.4.3 SIMEX Input and Analysis 16
5.4.4 Goodness of Fit 17
5.5 Matrix Effects 18
5.6 Operational Factors 18
Chapter 6 Test Results 19
6.1 False Positive and False Negative Rates 19
6.2 Precision 23
6.3 Sensitivity 27
6.4 Modeled Probability of Test Kit Response 28
6.5 Matrix Effect 40
-------
6.6 Operational Factors 41
Chapter 7 Performance Summary 42
Chapters References 45
Appendix A Performance Evaluation Materials Summary Information A-1
Al Preparation of Performance Evaluation Materials A-3
A2 Comparison of Expected vs. Actual Lead Concentrations of Performance Evaluation
Materials A-19
A3 Q A/QC Results for the ICP-AES Analysis of Performance Evaluation Materials A-22
Appendix B Vendor Comments B-l
Figures
Figure 2-1. LeadAVERT™ Test Kit test strip (showing result line) 2
Figure 6-1. Probability curves that represent test kit results that are both perfect (red line) and
within RRP rule criteria (black solid line) 33
Figure 6-2. Probability curves with shaded region to denote performance results that meet RRP
rule false positive and negative criteria. Test kits with curves that fall within the white region
and avoid the shaded region meet the RRP rule 34
Figure 6-3. LeadAVERT™ Test Kit predicted probability of positive test result (solid lines)
with 90% prediction interval (dotted lines) for the technical operator and yellow lead paint on
various substrates 35
Figure 6-4. LeadAVERT™ Test Kit predicted probability of positive test result (solid lines)
with 90% prediction interval (dotted lines) for the non-technical operator and yellow lead paint
on various substrates 36
Figure 6-5. LeadAVERT™ Test Kit predicted probability of positive test result (solid lines)
with 90% prediction interval (dotted lines) for the technical operator and white lead paint on
various substrates 37
Figure 6-6. LeadAVERT™ Test Kit predicted probability of positive test result (solid lines)
with 90% prediction interval (dotted lines) for the non-technical operator and white lead paint on
various substrates 38
Tables
Table 3-1. PEMs Testing Scheme for Each Test Kit 5
Table 6-1. The number of panels in each false positive and false negative analysis category.... 20
vi
-------
Table 6-2. LeadAVERT™ Test Kit false positive results for panels with confirmed lead levels <
0.8 mg/cm2 and false negative results for panels with confirmed lead levels > 1.2 mg/cm2 21
Table 6-3. LeadAVERT™ Test Kit false positive results for panels with confirmed lead levels <
1 mg/cm2 and false negative results for panels with confirmed lead levels > 1 mg/cm2 22
Table 6-4. Actual lead levels and their replicate set labels 23
Table 6-5. The number of panels at each target level and the number in each replicate set bin . 24
Table 6-6. LeadAVERT™ Test Kit consistency results by operator type, lead type, substrate, and
lead level 26
Table 6-7. LeadAVERT™ Test Kit precision results by lead type and operator type 27
Table 6-8. LeadAVERT™ Test Kit sensitivity results - lowest lead level for which the kit gave
consistent positive results (mg/cm2) 27
Table 6-9. LeadAVERT™ Test Kit univariate associations between probability of positive
response and explanatory variables 28
Table 6-10. LeadAVERT™ Test Kit multivariable Stata SIMEX logistic regression parameter
estimates 29
Table 6-11. LeadAVERT™ Test Kit modeled probability of positive test results and upper 95%
prediction bound when lead level = 0.8 mg/cm2 30
Table 6-12. LeadAVERT™ Test Kit modeled probability of positive test results, lower 95%
prediction bound, and corresponding conservative estimate of the false negative rate when lead
level = 1.2 mg/cm2 31
Table 6-13. LeadAVERT™ Test Kit false positive and negative threshold values (95%
confidence) based on the modeled probability of test results 40
vn
-------
List of Abbreviations and Acronyms
AMS
ASTM
CCV
COC
CRM
EPA
ESTE
ETV
ICP-AES
LCS
mg/cm2
MSDS
NLLAP
PE
PEM
ppb
PT
QA
QC
QCS
QMP
RRP
SOP
ISA
Advanced Monitoring Systems
American Society for Testing and Materials
continuing calibration verification
chain of custody
certified reference material
U.S. Environmental Protection Agency
Environmental and Sustainable Technology Evaluations
Environmental Technology Verification
inductively coupled plasma-atomic emission spectrometry
laboratory control spike
milligrams per centimeter squared
material safety data sheets
National Lead Laboratory Accreditation Program
performance evaluation
performance evaluation material
parts per billion
performance test
quality assurance
quality control
quality control sample
quality management plan
Renovation, Repair, and Painting
standard operating procedure
technical systems audit
Vlll
-------
Chapter 1
Background
The U.S. Environmental Protection Agency (EPA) supports the Environmental Technology
Verification (ETV) Program to facilitate the deployment of innovative environmental
technologies through performance verification and dissemination of information. The goal of the
ETV Program is to further environmental protection by accelerating the acceptance and use of
improved and cost-effective technologies. ETV seeks to achieve this goal by providing high-
quality, peer-reviewed data on technology performance to those involved in the design,
distribution, financing, permitting, purchase, and use of environmental technologies.
ETV works in partnership with recognized testing organizations; with stakeholder groups
consisting of buyers, vendor organizations, and permitters; and with the full participation of
individual technology developers. The program evaluates the performance of innovative
technologies by developing test plans that are responsive to the needs of stakeholders,
conducting field or laboratory tests (as appropriate), collecting and analyzing data, and preparing
peer-reviewed reports. All evaluations are conducted according to rigorous quality assurance
(QA) protocols to ensure that data of known and adequate quality are generated and that the
results are defensible.
This verification test was conducted under the U.S. EPA ETV program. Testing was performed
by Battelle, which served as the verification organization under the Environmental and
Sustainable Technology Evaluations (ESTE) arm of ETV. Battelle evaluated the performance of
qualitative spot test kits for lead in paint.
This verification test was developed with the support of a stakeholder technical panel. A
voluntary stakeholder technical panel consisting of individuals from the American Industrial
Hygiene Association (Kenn White), U.S. Department of Housing and Urban Development
(Warren Friedman), National Institute for Occupational Safety and Health (Kevin Ashley), U.S.
Army Center for Health Promotion and Preventative Medicine (Al Liabastre), National Center
for Healthy Housing (David Jacobs), National Association of Homebuilders (Matt Watkins), the
U.S. Consumer Product Safety Commission (Joanna Matheson), the Center for Disease Control
and Prevention (Larry Franklin), and U.S. EPA (Paul Carroll and Moira Lataille) was formed for
this verification test. Participants on this panel were reviewed and approved by EPA. This panel
gave input during the entire ETV process, including providing guidance and input on the
development of the performance evaluation materials used in this test, on the development of the
test design and test/QA plan, and comments on this report.
-------
Chapter 2
Technology Description
This report provides results for the verification testing of LeadAVERT™ Test Kit for lead-based
paint by Silver Lake Research. The following is a description of the LeadAVERT™ Test Kit,
based on information provided by the vendor. The information provided below was not verified
in this test.
The LeadAVERT™ Test Kit is an antibody-based test for the detection of lead in paint samples.
The test uses specific monoclonal antibodies that recognize and bind to lead atoms extracted
from paint with a weakly acidic, low-toxicity extraction solution. The antibodies are
incorporated in a competitive immunoassay in a lateral-flow test strip format, so that the entire
immunoassay is contained in a small test strip activated by the flow of sample. Results are read
visually as blue lines appearing in the result window of the test strip. The LeadAVERT™ Test
Kit is calibrated to give a "positive" or "negative", relative to the 1 milligrams per centimeter
squared (mg/cm2) standard for lead based paint.
The LeadAVERT™ Test Kit is provided as a 20-test package, and includes all components
necessary to run the test: a reusable stencil to trace a specific area of paint, disposable test vials,
an extraction solution in a dropper bottle, and a container of 20 single-use test strips. Paint is
removed from the traced area into the test vial, and the extraction
solution is added to the vial. After 3 minutes, the test strip is placed
into the vial, and allowed to run for 10 minutes. Results are read by
I comparing the intensity of two blue lines appearing in the "result
window" of the test strip (see Figure 2-1). The entire procedure is
completed in less than 15 minutes and requires no power sources or
instrumentation.
The LeadAVERT™ Test Kit package is a container of approximately
5" X 8" X 1.5". Recommended storage is at room temperature (50-
Figure 2-1. 86°F / 10-30°C). The 20-test LeadAVERT™ Test Kit has a suggested
LeadAVERT™ Test Kit retail price of $39.95 (as of May 2010).
test strip (showing result
line)
-------
Chapter 3
Test Design and Procedures
3.1 Introduction
This verification test was conducted according to procedures specified in the Test/QA Plan for
Verification of Qualitative Spot Test Kits for Lead in Paint1 Lead-based paints were commonly
used in houses in both interior and exterior applications prior to 1978, when the US government
banned the use of lead-based paint in residential applications. The term lead-based paint means
paint or other surface coatings that contain lead at contents that equal or exceed a level of 1.0
mg/cm2 or 0.5 percent by weight.2 This paint still exists in many of these houses across the
country. The accurate and efficient identification of lead-based paint in housing is important to
the Federal government as well as private individuals living in residences containing such paints.
Renovation, repair, and painting (RRP) activities may disturb painted surfaces and produce a
lead exposure hazard. Such disturbances can be especially harmful to children and pregnant
women as lead exposure can cause neurological and developmental problems in both children
and fetuses. In fact, because of the large amount of pre-1978 housing stock, a report by the
President's Task Force on Environmental Health Risks and Safety Risks to Children found that
approximately 24 million US dwellings were at risk for lead-based paint hazards.3
There are lead-based paint test kits available to help home owners and contractors identify lead-
based paint hazards before any RRP activities take place so that proper health and safety
measures can be taken. However, many of these test kits have been found to have high rates of
false positives (i.e., test kit indicates that lead in excess of 1.0 mg/cm2 is present, while in fact
the true lead level is below 1.0 mg/cm2).4 This verification test was conducted in response to the
call of the Renovation, Repair, and Painting rule2 for an EPA evaluation and recognition program
for test kits that are candidates to meet the goal of a demonstrated probability (with 95%
confidence) of a false negative response less than or equal to 5% of the time for paint containing
lead at or above the regulated level, 1.0 mg/cm2 and a demonstrated probability (with 95%
confidence) of a false positive response less than or equal to 10% of the time for paint containing
lead below the regulated level, 1.0 mg/cm2. This test incorporated ASTM International's El828,
Standard Practice for Evaluating the Performance Characteristics of Qualitative Chemical Spot
Test Kits for Lead in Paint guidelines into the test design.
The objective of this verification test was to evaluate the performance of the test kits for the
detection of lead in paint. This evaluation assessed the capabilities of the lead paint spot test kits
against laboratory-prepared performance evaluation material (PEM) samples and compared the
lead paint test kit results with those of a standard technique, inductively coupled plasma-atomic
-------
emission spectrometry (ICP-AES). Additionally, this verification test relied on verification
testing staff observations to assess other performance characteristics of the lead paint test kits.
Only qualitative results (e.g., detect/non-detect of lead at specified levels) were considered for
each technology.
The LeadAVERT™ Test Kit was verified by evaluating the following parameters:
• False positive and false negative rates
• Precision
• Sensitivity
• Modeled probability of test kit response
• Matrix effects
• Operational factors.
Verification testing of the test kit was conducted from January to June 2010. This timeframe
included testing of the test kit and also completion of all ICP-AES and QC analyses. False
positive and negative rates were determined by comparing test kit responses to actual lead
concentrations of the PEM as determined through ICP-AES. Precision was determined by
reproducibility of responses for replicate samples. Sensitivity was determined as the lowest
detectable level of the test kit. The modeled probability and matrix effects were determined
using logistic regression models.
Operational factors such as ease of use, operator bias, average cost, average time for kit
operation, helpfulness of manuals, and sustainability metrics such as volume and type of waste
generated from the use of each test kit, toxicity of the chemicals used, and energy consumption
were determined based on documented observations of the testing staff and the Battelle
Verification Test Coordinator. Operational factors were described qualitatively, not
quantitatively; therefore, no statistical approaches were applied to the operational factors.
3.2 Test Facility
Laboratory analyses of the LeadAVERT™ Test Kit were conducted in Battelle laboratories in
Columbus, Ohio. No field testing was conducted during this technology verification.
3.3 Test Procedures
Qualitative spot test kits for lead in paint were evaluated against a range of lead concentrations in
paint on various substrates through the use of PEMs. PEMs were 3 inch by 3 inch square panels
of wood (pine and poplar), metal, drywall, or plaster that were prepared by Battelle.6 Pine and
poplar were chosen for the wood panels as they are representative of woods most commonly
found in homes. Table 3-1 shows the PEMs prepared for each test kit. Poplar and pine PEMs
were distributed in random mixtures (e.g., two poplar and one pine or one poplar and two pine)
for each set of three wood PEMs listed in Table 3-1. Each PEM was coated with either white
lead (lead carbonate) or yellow lead (lead chromate) paint. The paint contained lead targeted at
-------
Table 3-1. PEMs Testing Scheme for Each Test Kit
Lead Type
Control Blank
White Lead
(Lead Carbonate)
Yellow Lead
(Lead Chromate)
Lead Level
(mg/cm2)
0
0.3
0.6
1.0
1.4
2.0
6.0
0.3
0.6
1.0
1.4
2.0
6.0
Substrate
Wood
Metal
Dry wall
Plaster
Wood
Metal
Drywall
Plaster
Wood
Metal
Drywall
Plaster
Wood
Metal
Drywall
Plaster
Wood
Metal
Drywall
Plaster
Wood
Metal
Drywall
Plaster
Wood
Metal
Drywall
Plaster
Wood
Metal
Drywall
Plaster
Wood
Metal
Drywall
Plaster
Wood
Metal
Drywall
Plaster
Wood
Metal
Drywall
Plaster
Wood
Metal
Drywall
Plaster
Wood
Metal
Drywall
Plaster
Painted PEMs Subtotal
PEMs Analyzed Per Test Kit by Topcoat Color
White
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
156
Red-Orange
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
156
Grey-Black
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
156
Unpainted PEMs Subtotal (2 per each substrate)
Total
Total
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
468
8
476
Actual number of PEMs used to evaluate performance at specific lead levels varied based on actual concentrations observed during analysis.
-------
0.3, 0.6, 1.0, 1.4, 2.0, and 6.0 mg/cm2. These lead concentrations were chosen with input from
the stakeholder technical panel based on criteria provided in EPA's lead RRP rule as well as to
represent potential lead levels in homes. Paint containing no lead (0.0 mg/cm2) was also applied
to each substrate and tested.
Two different layers of paint were applied over the leaded paint. One was a primer designed for
adhesion to linseed oil-based paint and the second coat was a typical interior modern latex paint
tinted to one of three colors: white, red-orange, or grey-black. These colors were chosen by
EPA, with input from the stakeholder technical panel, based on the potential of certain colors to
interfere or not with lead paint test kit operations. The topcoat paint manufacturers'
recommended application thickness was used. Two coats at the recommended thickness were
applied. Details on the PEM production process can be found in Appendix A.
The LeadAVERT™ Test Kit for lead paint was operated by a technical and non-technical
operator. The technical operator was a Battelle staff member with laboratory experience. The
technical operator was trained by a representative of the vendor company in the operation of its
test kit. The same technical operator operated this test kit throughout testing. Because this lead
paint test kit is anticipated to be used by certified remodelers, renovators, and painters, it was
also evaluated by a non-technical operator. The non-technical operator was a certified renovator
with little to no experience with lead analysis. The non-technical operator was provided the
instruction manual provided by the vendor with the test kit for training. The non-technical
operator viewed the materials himself to understand how to operate the test kit. The non-
technical operator was also permitted to ask questions or clarifications of the vendor on the
operation of the test kit. This scenario approximated the training renovators are expected to
receive under the RRP rule.
Tests were performed in duplicate on each PEM by each operator, technical and non-technical
(i.e., two samples were taken from each PEM by each operator). Duplicates were tested in
succession by each operator on a given PEM. PEMs were analyzed blindly by each operator in
that the PEMs used for analysis were marked with a non-identifying number. Test kit operators
were not made aware of the paint type, lead level, or substrate of the PEM being tested. PEMs
were tested in random order (i.e., PEMs were placed in plastic bins and the operators arbitrarily
selected a PEM for analysis). To determine whether the substrate material affected the
performance of the test kits, two unpainted PEMs of each substrate were tested using each test
kit, in the same manner as all other PEMs (i.e., per the test kit instructions). Three PEMs at each
lead level, substrate, and topcoat color were prepared for use in this test. In total, 468 painted
PEMs were prepared for use in the verification test of each test kit.
Paint chip samples from each PEM were analyzed by a National Lead Laboratory Accreditation
Program (NLLAP) recognized laboratory, Schneider Laboratories, Inc., using ICP-AES to
confirm the lead level of each PEM used for testing. The paint chip samples for reference
analyses were collected by Battelle according to a Battelle SOP7, which was based on ASTM
E1729.8 The reference analyses confirmed the lead level of each PEM. Lead levels determined
through the reference analysis were used for reporting and statistical analyses.
The procedures for collecting, storing, and shipping test samples are provided below.
-------
3.3.1 Test Sample Collection, Storage, and Shipment
Chips of lead paint were taken from each PEM and sent for ICP-AES analysis at a NLLAP-
recognized laboratory, Schneider Laboratories, Inc. A glass screw-top vial was labeled with the
PEM identification number located on the back of the panel. The number was also recorded on
the Chain of Custody (COC) form. Sampling was performed per the Battelle SOP for collection
of dried paint samples for lead determination. All safety precautions and personal protective
equipment were used. A one inch square, metal template was placed adjacent to the tested area.
A utility knife was used to trace around the template. Tweezers and a utility knife were used to
scrape and remove the paint within the one inch area, using caution to minimize introduction of
the substrate into the paint sample. The topcoat and remaining paint were transferred to a
glassine weighing paper with the assistance of a paintbrush. The sample was then transferred
from the glassine paper into a glass vial using the paintbrush. All instruments and templates were
wiped with tissue paper and the bench top was cleaned and gloves were changed between each
sample to minimize contamination. The paint brush was carefully flicked and tapped over a trash
can to remove any residual lead dust. All wipes and gloves were disposed of as lead waste. The
vials were then collected into a zip-top bag and taped up securely for shipping. The bags and
COC were then shipped together using overnight delivery to Schneider Laboratories, Inc.
Paint chip samples were stored at room temperature as received by Schneider Laboratories, Inc.
and then analyzed by ICP-AES. Analytical results were reported to Battelle within 2-3 days.
Sample digests were stored separately by Schneider Laboratories, Inc. at room temperature.
PEMs were stored individually in zip-top bags. The back of each PEM was labeled with an
identifying number. The outside of the zip-top bag was labeled with the same number. Each
PEM was wrapped in a Kimwipe and each zip-top bag was sealed when not in use. The zip-top
bags containing the PEMs were housed in large plastic bins in the laboratory during testing.
3.3.2 Test Sample Analysis Procedure
Materials needed that were not part of the LeadAVERT™ Test Kit included a pencil, adhesive
tape, a utility knife, and a timer. Everything else was incorporated in the kit including test strips,
extraction solution, test vials, and a sample stencil.
Using the stencil provided with the LeadAVERT™ Test Kit, a circle was drawn on the PEM
sample. Less than a quarter inch under the circle, the test vial was attached to the PEM with the
adhesive tape. With the utility knife (an X-acto® knife), the circle was cut out on the outside of
the pencil markings. The circle was placed or "scooted" into the vials, making sure all powders
were removed from the substrate; if removing the circle of paint into the vial was difficult, the
inside of the circle was quarterly cut. The taped vial was removed from the surface of the PEM;
the underlying paint was removed on some panels when removing the tape, prohibiting the
taking of a duplicate sample in that area of the PEM. Eleven drops of the extraction solution
were added to the vial and allowed to sit for three minutes. During the start of those three
minutes, the vial was stirred briefly by gently tapping it to the table or swirling the vial for a few
seconds. After the three minutes passed, a fresh testing strip was inserted into the sample for ten
seconds with the arrow on the test strip pointing into the solution. The strip was then removed
from the solution and placed across the top of the vial for ten minutes, with the bottom end of the
-------
strip not touching anything. Results were read by noting the presence of Line 1 and Line 2 on
the test strip. A positive result was indicated by either two equally dark lines or the presence of
Line 2 as darker than Line 1. A negative result was indicated either by Line 1 only or by the
appearance of Line 1 darker than Line 2. The test strips had a built-in quality control measure.
The test was considered to not run properly and thus invalid if no lines appeared on the test strip
after the 10 minute wait period or if both lines were very light. Invalid samples were re-analyzed
by collecting another paint sample from the same PEM. Samples were collected and analyzed
from each PEM in duplicate, and recorded with the results and any observations.
-------
Chapter 4
Quality Assurance/Quality Control
QA/QC procedures were performed according to the quality management plan (QMP) for the
Battelle ETV Advanced Monitoring Systems (AMS) Center9, except where differences were
noted for ESTE per the EPA ETV Program QMP10, and the test/QA plan for this verification
test.1 Test procedures were as stated in the test/QA plan; however a deviation to the test/QA plan
was made during the ICP-AES analyses. For some sample runs, continuous calibration
verification (CCV) samples were run once every 20 instead of 10 samples. This deviation is
described below. This change was assessed to have no impact on the quality of the results as
described below. QA/QC procedures and results are described below. Additional information
on QA/QC outcomes for the PEMs are provided in Appendix A.
4.1 Quality Control Samples
Steps were taken to maintain the quality of data collected during this verification test. This
included analyzing specific quality control samples for the reference method (ICP-AES) and the
test kit.
4.1.1 ICP-AES Blank Sample Results
Various blank samples were analyzed for the ICP-AES analyses. Method blank samples were
analyzed in each set of 10-20 paint samples to ensure that no sources of contamination were
present. An initial calibration blank was analyzed at the beginning of each run and used for
initial calibration and zeroing the instrument. A continuing calibration blank was analyzed after
each CCV to verify blank response and freedom from carryover. No blank samples failed during
the analyses.
4.1.2 ICP-AES Matrix Spike Samples and Calibration Verification Standards
Initial calibration standards were run at the beginning of each set of analyses. The acceptance
criterion for the calibration coefficient of the calibration standards was > 0.998. If this criterion
was not met, the analysis was stopped and recalibration was performed before samples were
analyzed. A 500 parts per billion (ppb) CCV standard was analyzed at the beginning of each run
(following the initial calibration), at the end of each run, and every 10-20 samples. CCV
recoveries ranged from 96% to 108%. Per the test/QA plan, CCV sample frequency was once
every 10 samples. For most of the sample sets CCVs were performed with this frequency.
-------
However, for later sample sets CCVs were run once every 20 samples. CCV samples were used
to verify instrument performance. CCV samples were run every 10 samples as a preventative
measure so that large amounts of samples do not need to be re-run if a CCV sample fails. In the
course of this study, one CCV sample failed. All samples from the last passing CCV of that
sample set were re-analyzed.
A matrix spike sample and laboratory control sample (LCS), as well as duplicates of these
samples, were also analyzed. Duplicate samples were run once every 10-20 samples. Acceptable
recoveries for matrix spike samples were between 80-120%. Acceptable recoveries for LCS
samples were between 80-120%. Duplicate samples had acceptance criteria of ±25% relative
percent difference (RPD).
All matrix spike samples were performed as post-digestion spikes as there was insufficient
sample volume to perform a pre-digestion spike. Matrix spike recoveries ranged from 86% to
207%. Six matrix spike samples failed, with recoveries above the specified acceptance criteria.
In these instances, the lead concentration in the sample was well above the spike level. Matrix
spike results indicated that matrix interferences were not observed. Duplicate samples were
within the specified RPD.
LCS samples were analyzed once every 10-20 samples. LCS recoveries ranged from 17% to
225%. Schneider Laboratories, Inc. noted that LCS failures on one sample set were attributed to
improper spiking technique. Training on spiking procedures was immediately implemented by
Schneider Laboratories for all analysts spiking samples. All LCS failures occurred prior to a
revision to the Schneider Laboratories, Inc. SOP11 for analyzing paint samples written
specifically for this verification test. In the original version of the SOP, LCS samples were
prepared by spiking a known amount of lead onto a certified reference material (CRM). This
practice was changed because there were over-recovery issues. This was because the spike was
not >3x the background lead concentration because of the high lead concentrations in the actual
CRM samples. In the revised SOP, the LCS was prepared by spiking a piece of lead-free latex
paint. There were no LCS failures after that. In addition, a QC check sample containing only
the CRM, which had a known concentration of lead weighed out to a particular amount, was
analyzed with each sample set throughout the verification test. These QC check samples all
passed acceptance criteria.
4.1.3 Test Kit Quality Controls and Blank PEMs
As indicated in Section 3.3.2, quality control measures were built into the test strip itself. Some
test results were considered invalid according to this quality control measure. If an invalid result
was obtained, another sample was taken from the same PEM and those valid results were
reported. Painted PEMs containing no lead as well as each of the PEM substrates containing no
paint were also run as part of the verification test. All samples of PEM substrates containing no
paint returned negative results from the test kit (i.e., no lead was present). All but three samples
of painted PEMs containing no lead returned negative results. There was no reason to suspect
contamination, so these results were interpreted to be false positive results.
10
-------
4.2 Audits
Three types of audits were performed during the verification test: a performance evaluation (PE)
audit of the reference method measurements made in this verification test, a technical systems
audit (ISA) of the verification test performance, and a data quality audit. Audit procedures are
described below.
4.2.1 Performance Evaluation Audits
A PE audit was conducted to assess the quality of the reference method measurements made in
this verification test. The reference method PE audit was performed by supplying an
independent, NIST-traceable lead paint standard (Reference Material 8680, panel CBS), to the
reference laboratory. The PE audit samples were analyzed in the same manner as all other
samples and the analytical results for the PE audit samples were compared with the nominal
concentration. The target criterion for this PE audit was in agreement with the analytical result
within 20% of the nominal concentration. The specified acceptable concentration range for the
NIST standard panel was 1.13 - 1.75 mg/cm2 (1.44 ±0.31 mg/cm2). The PE samples taken from
this standard panel were 1.38, 1.38, 1.19, and 1.31 mg/cm2. The PE audit result met the target
criterion. This audit was performed once at the start of the test.
4.2.2 Technical Systems A udit
The Battelle Quality Manager performed one TSA during this verification test to ensure that the
verification test was being performed according to the Battelle AMS Center and ETV Program
QMPs, the test/QA plan, any published reference methods, and standard operating procedures. In
the TSA, the Battelle Quality Manager reviewed the reference methods used, compared actual
test procedures with those specified or referenced in the test/QA plan, and reviewed data
acquisition and handling procedures. Also in the TSA, the Battelle Quality Manager observed
testing, observed reference method sample preparation and analysis, inspected documentation,
and reviewed technology-specific record books. He also checked standard certifications and
technology data acquisition procedures and conferred with the technical staff. A TSA report was
prepared. There were no findings. The records concerning the TSA are permanently stored with
the Battelle Quality Manager.
The EPA ETV Quality Manager also performed a TSA of both the reference laboratory and the
testing conducted at Battelle Columbus, OH facilities. No findings were reported in the TSA of
the reference laboratory, Schneider Laboratories, Inc. In the TSA of the lead paint test kit
evaluations at Battelle's Columbus, OH facilities, the EPA ETV Quality Manger cited two
findings. These findings were related to ease of use observations and were immediately and
adequately addressed and did not affect the quality of the test.
4.2.3 Audit of Data Quality
Records generated in the verification test received a one-over-one review (i.e., review by a
Battelle technical staff who did not generate the records) before these records were used to
calculate, evaluate, or report verification results. A Battelle technical staff member involved in
the verification test reviewed the data. Datasheets generated by the operators during testing were
11
-------
reviewed for completeness and errors. The person performing the review added his/her initials
and the date to a hard copy of the record being reviewed. At least 10% of the data acquired
during the verification test, including the ICP-AES results, were audited by Battelle. At least
25% of the ICP-AES data acquired during the verification test were audited by EPA. Battelle's
Quality Manager traced the data from the initial acquisition, through reduction and statistical
analysis, to final reporting to ensure the integrity of the reported results. All calculations
performed on the data undergoing the audit were checked. Minor transcription errors were
identified and corrected before the results were used for the calculations described in Chapter 5.
Battelle's and EPA's Quality Managers also reviewed the PEM ICP-AES results thoroughly to
ensure that all data quality indicators as stated in the test/QA plan were followed and that
reported results matched the data generated on the instrument. Findings were cited by the EPA
Quality Manager. Appropriate corrective actions were taken. Significant QA/QC concerns
identified during EPA's audit are discussed in Section 4.1.
12
-------
Chapter 5
Statistical Methods
The statistical methods used to evaluate the performance factors listed in Section 3.1 are
presented in this chapter. The LeadAVERT™ Test Kit was evaluated for qualitative results (i.e.,
positive/negative responses to samples). All data analyses were based on these qualitative
results. QC samples and unpainted PEM substrates were not included in any of these analyses.
Results are provided in Chapter 6.
5.1 False Positive and False Negative Rates
A false positive response was defined as a positive result when regulated lead-based paint was
not present. The test/QA plan1 defined false positive rates as being based on target lead levels at
and below 0.6 mg/cm2 with confirmed values not to exceed 0.8 mg/cm2. Because confirmed lead
levels of particular PEMs did not sometimes match target concentrations for those PEMs, false
positive rates were assessed on panels with confirmed lead levels at 0.8 mg/cm2 and lower.
Consistent with the EPA's April 22, 2008 RRP rule2, panels with an ICP-AES confirmed lead
level between 0.8 and 1.0 mg/cm2 were not used in the false positive analysis.
A false negative response was defined as a negative response when regulated lead-based paint
was present. The test/QA plan defined false negative rates as being based on target lead levels at
and above 1.4 mg/cm2 with confirmed values not to exceed 1.2 mg/cm2. Because confirmed
lead levels of particular PEMs did not sometimes match target concentrations for those PEMs,
false negative rates were assessed on panels with confirmed lead levels at 1.2 mg/cm2 and
higher. Consistent with the EPA's April 22, 2008 RRP rule, panels with an ICP-AES confirmed
lead level between 1.0 and 1.2 mg/cm2 were not used in the false negative analysis.
Based on stakeholder technical panel input, the EPA lead paint action level of 1.0 mg/cm2 lead
was included for analysis as part of the verification test. Though evaluations of test kit
performance based on this level is not in the EPA RRP rule, false positive and negative rates, in
addition to those stated above, were also calculated for each test kit based on 1.0 mg/cm2 lead.
Thus, false positive rates were assessed on PEMs with confirmed lead levels at 1.0 mg/cm2 and
lower and false negative rates were assessed on PEMs with confirmed lead levels at 1.0 mg/cm2
and higher. For panels that measure 1.0 mg/cm2, positive results were considered "correct" and
negative results were considered false negative. If the confirmed lead concentration of the PEM
was greater than 1.0 mg/cm2 (e.g., 1.1 mg/cm2), then negative results were considered false
13
-------
negatives. If the confirmed lead concentration of the PEM was less than 1.0 mg/cm2 (e.g., 0.9
mg/cm2), then positive results were considered false positives.
False positive and negative rates were calculated as shown in Equations 1 and 2, respectively:
„ , _ # of positive results ,..,
False Positive Rate = J—*- 7 (1)
total # of PEMs with lead level below 0.8 (or 1.0) mg/cm
„ , „ # o/ negative results
False Negative Rate = •
total # of PEMs with lead level above 1.2 (or 1.0) mg/cm2
5.2 Precision
Precision was measured by the reproducibility of responses for replicate samples within a group
of PEMs. Precision results were reported as the percentage of consistent responses from all
replicate sets for those paint types (see Equation 3). Responses were considered inconsistent if
25% or more of the replicates differed from the response of the other samples in the same group
of PEMs.
_,..„, . , x # of consistent responses of replicate sets n__ ,„.
Precision (% consistent results) = xWO (3)
total number of replicate sets
5.3 Sensitivity
The sensitivity or lowest detectable lead level for each test kit was identified based on the
detection results across all PEM lead levels. The lowest PEM lead level with consistent (>75%)
positive or "detect" responses was considered the lowest detectable level. The identified lowest
detectable lead level was reported and discussed.
5.4 Modeled Probability of Test Kit Response
Logistic regression models were used to determine the probabilities of positive or negative
responses of the test kit at the 95% confidence level, as a function of lead concentration and
other covariates, such as substrate type, lead paint type, operator type, and topcoat color. An
evaluation of the bivariate relationship between the response variable and each candidate
explanatory variable was performed by fitting single covariate logistic models to assess the
predictive ability of each of the PEM parameters. Using the results from these bivariate
analyses, a parsimonious multivariate model was developed including a set of explanatory
variables which were most predictive of the probability of the test kit response variable. The
potential logistic regression model took the form below:
14
-------
where Yt is the outcome of the test kit, Xt is a vector of explanatory variables associated with Yt
and ft represent a vector of unknown parameters which was estimated with the model. Test
results that indicated that lead was present were represented with Y=l; negative results were
represented with Y=0. Candidate independent variables associated with the response variable
were lead level (continuous), operator type (categorical), lead type (categorical), substrate type
(categorical), and topcoat color (categorical). Interactions between categorical predictor
variables were also assessed. Categorical covariates were modeled using indicator variables.
SAS's PROC LOGISTIC was used to evaluate the association between each explanatory variable and
the probability of a positive test kit result. Then multivariable models were fit using a backward
selection process whereby all explanatory variables were included in the initial model. In a multi-
step backwards elimination process, the variable with the weakest association (highest Type HI p-
value) was eliminated from the model until all of the variables that remained had Type III p-values
less than 0.05. The list of variables that remained formed the basis for evaluating interactions.
Measured lead level was retained as an explanatory variable in all multivariable models. Two-way
interactions were tested between all pairs of categorical explanatory variables that had p-values
below 0.05. Interactions were retained in the multivariable models if their p-values were smaller
than 0.05.
5.4.2 Accounting for Measurement Error — SIMEX Background and Intuition
Categorical covariates in this experiment were measured without error, but the lead level
measurements were subject to some measurement error due both to variability inherent in the
measurement (ICP-AES) process and possibly due to spatial heterogeneity in lead concentrations
in paint on the PEMs themselves. The experimental design did not include multiple ICP-AES
analyses per PEM so there is no direct estimate of the variability in measurements for these data.
To account for the uncertainty associated with that error, the final multivariable model for each
test kit was subjected to a simulation and extrapolation (SIMEX) analysis.12"16
A detailed description of SIMEX is beyond the scope of this report, but in short, it is a robust
method of accounting for measurement error. The method requires either replicate measures of
the quantity that is measured with error, or a characterization of the variability in the
measurements. It then estimates what the regression model coefficients would be in the absence
of measurement error. The technique estimates standard errors for the regression model
coefficients using the bootstrap technique. SIMEX analyses were carried out in Stata version
11.1 using the programs described in Hardin et al (2003c).15
The premise of the analysis is that one of the independent variables, namely lead concentration,
has been measured with error. In the logistic regression models considered here, lead
concentration is the only continuous independent variable; all of the other covariates are
categorical. Thus, lead concentration may be considered the 'x' in a simple linear regression.
The observed variability in 'x' is comprised of two components, actual variation in lead
15
-------
concentration and measurement error. If we were able to remove the measurement error then we
would observe less variability in that independent variable.
There are two important points of intuition that will inform expectations about what is seen in the
SIMEX results. First, the data along the x-axis of a scatterplot would "tighten up" if
measurement error were removed. "Tightening up" the independent variable in a regression
analysis will result in a steeper slope or a regression coefficient with a larger magnitude. This is a
fundamental consequence of any technique that adjusts for measurement error in the independent
variable in a regression analysis. In the lead paint analysis, steeper logistic regression curves
will result from the SIMEX analysis than would result from a non-SIMEX analysis where lead
levels were considered to be fixed and known.
Second, when the statistical analyses acknowledge and account for the measurement error, then
the regression output prediction intervals may be wider than those for a non-SEVIEX analysis
where 'x' is considered to be fixed and known. For any given predicted value of the outcome
variable, the prediction interval will most likely be wider, or at least not narrower. But for a
fixed value of 'x', (such as 0.8 or 1.2 mg/cm2) whether the SIMEX prediction intervals are wider
or narrower than the non-SIMEX intervals depends on how much the slopes of the SIMEX and
non-SIMEX regression line differ. For typical logistic regression models, prediction intervals
are very narrow at the extreme low and high asymptotic ends of the x-axis, and only appreciably
wide in the region where the probability of the outcome is not near zero and not near one. So if
the SIMEX analysis has only a moderate impact on the slope then wider prediction intervals
might be observed at 0.8 and 1.2 mg/cm2. But if the slope changes dramatically, then 0.8 or 1.2
mg/cm2 might now be in the part of the prediction curve that is near zero or one and the SIMEX
prediction interval might be dramatically more narrow than a non-SIMEX interval.
Thus, the prediction curves for every SIMEX analysis are expected to be steeper than, or at least
not less steep than, a non-SIMEX analysis. However, the assessment of test kit performance is
based on the upper and lower bounds of prediction intervals at 0.8 and 1.2 .mg/cm2, respectively.
5.4.3 SIMEX Input and Analysis
During pre-production of the PEMs, replicate paint chip samples were analyzed from selected
metal PEMs that served as reference panels (see Appendix A). Three metal panels were
prepared for the pre-production homogeneity testing. Four paint chip samples, one from each
quadrant of the PEM, were taken and analyzed via ICP-AES for their lead levels. Data are
available on the coefficients of variation for these metal PEMs for both white and yellow lead.
These data are shown below in Table 5-1. Though these data did not come from actual PEMs
used during the lead paint test kit verification test, this information was used as a surrogate
measure of homogeneity variability on the PEMs.
For each PEM in the study, nine random pseudo-replicates were generated from a normal
distribution with a mean equal to the confirmed lead concentration for that panel, and a standard
deviation computed from the metal reference PEM data in Table 5-1 and indexed by the panel's
lead type and target lead level. The nine measurements were used as inputs to the Stata SIMEX
algorithm as if they were true replicate measurements.
16
-------
Table 5-1. Results from Final Homogeneity Testing for each Set of ETV PEMs
Target Lead Mean Levels CoV*
Lead Type Level ICP (mg/cm2) ICP
White Lead
Yellow Lead
0.3
0.6
1.0
1.4
2.0
6.0
0.3
0.6
1.0
1.4
2.0
6.0
0.30
0.65
0.99
1.56
1.85
5.97
0.30
0.62
1.07
1.42
1.92
6.88
13.3
7.1
3.9
7.2
5.6
14.2
9.6
4.1
11.0
4.1
10.1
5.2
* Coefficient of Variation (Standard Deviation/Mean x 100)
There are two user-specified parameters for the Stata SIMEX algorithm: 1) the number of
replicate measurements for the covariate measured with error, and 2) the number of bootstrap
samples used to estimate standard errors on regression parameters. In testing not detailed here,
the sensitivity of the SIMEX algorithm to different settings of these parameters was investigated.
It was determined that the qualitative results were not sensitive to the values used in the analysis.
The values used were nine pseudo-replicates per PEM and 199 bootstrap samples, respectively.
The predicted regression curves and associated prediction intervals were generated in the interval
0.0 to 6.0 mg/cm2 using Stata. The relevant prediction bounds (the upper bound at 0.8 mg/cm2
and lower bound at 1.2 mg/cm2) were assessed and the predicted false positive and false negative
rates based on these prediction bounds were determined.
5.4.4 Goodness of Fit
To assess whether the logistic regression models fit the data well, standardized Pearson residuals
were computed for every observation and those with an absolute value greater than two were
flagged and plotted versus lead level. Standardized Pearson residuals greater than two are
associated with observations that are not well fit by the model. In the logistic regression context
observations that are not well fit might be those with high lead levels where the kit results were
negative or very low lead levels where the kit results were positive. In the absence of categorical
variables the standardized Pearson residuals should be normally distributed, so we would expect
approximately 5% of the observations to have residuals with absolute value greater than two. In
this case there are categorical covariates so the residuals are not strictly expected to be
distributed normally but the proportion of observations with large residuals is still informative.
That proportion is reported in Section 6.4.
17
-------
5.5 Matrix Effects
The covariate-adjusted logistic regression model described in Section 5.4 was used to assess the
significance of PEM parameters and the interactions among them on the performance of the test
kits. PEM parameters were included in the model as explanatory variables associated with the Y;
response variable.
Comparison of the observed values of the response variable to predicted values obtained from
models with and without the predictor variable in question was the guiding principle in the
logistic regression model. The likelihood function is defined as
where ft(Yi') is the conditional probability of Y; =1 and [1 - ^r(^)] is the conditional probability
of Y; =0 given the vector of explanatory variables (X). For purposes of assessing the
significance of a group of p predictor variables (where p can be 1 or more), we computed the
likelihood ratio test statistic, G, as follows:
G = -2 loge [likelihood without the p variables / likelihood with the p variables] (6)
Under the null hypothesis, this test statistic followed a chi-square distribution with p degrees of
freedom. If the test statistic was greater than the 95th percentile of the chi-square distribution,
then the group of variables, taken together, were statistically significant.
5.6 Operational Factors
There were no statistical calculations applicable to operational factors. Operational factors were
determined qualitatively based on assessments from the Operator (both technical and non-
technical) and the Battelle Verification Test Coordinator. Operational factors such as ease of
use, operator bias, average cost, average time for kit operation, and helpfulness of manuals, were
determined. Sustainability metrics such as volume and type of waste generated from the use of
each test kit, toxicity of the chemicals used, and energy consumption are discussed. This
discussion is based on how much waste was generated and what the waste was composed of,
information from the vendor on how the waste should be properly handled, a summary of the
pertinent MSDS information, when available, and noting whether the test kit used batteries, a
power supply, or no energy source was needed. Information on how many tests each kit could
perform as well as the shelf life of the test kit and chemicals used as part of the test kit was also
reported.
18
-------
Chapter 6
Test Results
The results for the LeadAVERT™ Test Kit are presented below for each of the performance
parameters.
In this report each PEM is associated with three definitions of lead levels:
• Target lead level - the expected concentration of each PEM as outlined in Table 3-1.
These target lead levels were 0, 0.3, 0.6, 1.0, 1.4, 2.0, or 6.0 mg/cm2.
• Confirmed lead level - the concentration as measured by the reference laboratory using
ICP-AES analysis.
• Closest target lead level - the target level that is closest to the confirmed level. If a panel
has a target lead level of 1.4 mg/cm2 and a confirmed lead level of 1.9 mg/cm2 then the
closest target level is 2.0 mg/cm2.
Under ideal circumstances the confirmed lead level would equal the target lead level, but this
was sometimes not the case. Analyses where lead level was a categorical variable (i.e.,
consistency, precision, and sensitivity analyses) characterized the panels by their closest target
lead level. Analyses where lead level was a continuous variable (i.e., the false positive/negative
and logistic regression analyses) characterized the panels by their confirmed lead level. Each
analysis described clearly which level was used to characterize the lead level.
6.1 False Positive and False Negative Rates
Observed false positive and negative rates were calculated based on confirmed lead levels as
measured though ICP-AES analysis. For example, if the PEM was confirmed to have a lead
level of 1.4 mg/cm2, and the test kit returned a negative result, this would be considered a false
negative. Table 3-1 details the target lead levels for the PEMs and the number of PEMs that
were anticipated at each lead level. Because of variations in PEM production, the confirmed
lead level of a particular PEM did not always match the target lead level. Table 6-1 compares
the number of PEMs at the confirmed and target lead levels used for the false positive and
negative analyses. The data are divided into three categories: those panels eligible for false
positive analysis (lead levels up to and including 0.8 mg/cm2), those excluded from false positive
and false negative analyses (lead levels between 0.8 and 1.2 mg/cm2) and those eligible for false
negative analysis (lead levels 1.2 mg/cm2 and above). If the confirmed lead levels had been
equal to the target lead levels, all of the numbers would lie along the shaded diagonal. Because
the confirmed levels sometimes differed significantly from the target levels, (i.e., the target lead
level was at 0.6 mg/cm2 but confirmed near 1.4 mg/cm2) some panels appear in the off-diagonal
table entries and were therefore included in portions of the analysis other than those for which
they had been targeted.
19
-------
Table 6-1. The number of panels in each false positive and false negative analysis category
Confirmed Lead Levels
Target
Lead
Levels
Eligible for False Positive Analysis
Excluded from Analysis
Eligible for False Negative Analysis
Total
Eligible
for False
Positive
Analysis
146
4
0
150
Excluded from
Analysis
26
42
9
77
Eligible
for False
Negative
Analysis Total
8
26
207
180
72
216
241 468
Tables 6-2 and 6-3 list the observed false positive and false negative rates for the LeadAVERT™
Test Kit under two sets of conditions:
• Table 6-2 shows the observed false positive results for panels with confirmed lead levels
< 0.8 mg/cm2 and observed false negative results for panels with confirmed lead levels >
1.2 mg/cm2, per the RRP ruling2.
• Table 6-3 shows observed false positive results for panels with confirmed lead levels < 1
mg/cm2 and observed false negative results for panels with confirmed lead levels > 1
mg/cm2.
Results for both the technical and non-technical operator are presented. Results are presented as
overall rates (i.e., false positive and negative results across all applicable PEMs combined) and
also false positive and negative rates based on lead paint type (i.e., white or yellow lead),
substrate (i.e., drywall, metal, plaster, or wood), and topcoat paint color (i.e., grey red or white).
The overall observed false positive rate for the technical operator, based on confirmed lead levels
of <0.8 mg/cm2 (see Table 6-2) was 22%. Observed false positive rates based on PEM
characteristics ranged from 2% for yellow lead PEMs to 55% for white lead PEMs. Higher
observed false positive rates were found for metal and wood PEMs as opposed to drywall and
plaster, which had similar, low observed false positive rates (9%). Topcoat color did not appear
to impact the observed false positive rate.
The overall observed false positive rate for the non-technical operator was 12%. Similar to the
technical operator, evaluation of white lead PEMs produced the highest observed false positive
rate. Unlike with the technical operator, metal and plaster PEMs resulted in higher observed
false positive rates (20% and 16%, respectively) than drywall and wood (5% for both).
Observed false negative rates for both the technical and non-technical operator indicated that the
LeadAVERT™ Test Kit did not perform well on yellow lead paint. Substrate and topcoat color
did not appear to make a difference, as these observed false negative rates were similar to the
overall rates found for each operator. Observed false negative rates for the technical operator
were 37% overall and 71% on yellow lead PEMs. Observed false negative rates for the non-
technical operator were 56% overall and 96% on yellow lead PEMs.
20
-------
Table 6-2. LeadAVERT™ Test Kit false positive results for panels with confirmed lead
levels < 0.8 mg/cm2 and false negative results for panels with confirmed lead levels > 1.2
mg/cm2
Overall
None
White
Yellow
Drywall
Metal
Plaster
Wood
Grey
Red
White
Lead AVER
False Positives'
Technical Non-technical
Operator Operator
66 / 300 = 22%
6 / 72 = 8%
58/106 = 55%
2 / 122 = 2%
6 / 66 = 9%
35 / 90 = 39%
6 / 64 = 9%
19 / 80 = 24%
20 / 102 = 20%
25 / 102 = 25%
21 / 96 = 22%
35 / 300 = 12%
0 / 72 = 0%
33 / 106 = 31%
2 / 122 = 2%
3 / 66 = 5%
18 / 90 = 20%
10 / 64 = 16%
4 / 80 = 5%
14 / 102 = 14%
7 / 102 = 7%
14 / 96 = 15%
T Test Kit™
False Negatives"
Technical Non-technical
Operator Operator
177 / 480 = 37% 266 / 476 = 56%
NA NA
4 / 238 = 2% 34 / 234 = 15%
173 / 242 = 71% 232 / 242 = 96%
52 / 128 = 41% 70 / 128 = 55%
35 / 90 = 39% 52 / 86 = 60%
43 / 146 = 29% 72 / 146 = 49%
47 / 116 = 41% 72 / 116 = 62%
58/156 = 37% 86/152 = 57%
56/162 = 35% 86/162 = 53%
63 / 162 = 39% 94 / 162 = 58%
'False positives on PEMs with confirmed lead levels < 0.8 mg/cm2
"False negatives on PEMs with confirmed lead levels > 1.2 mg/ cm2
NA: If the paint did not contain lead then a false negative is not possible, those entries are 'NA'
(not applicable).
The observed false negative rates for both the technical and non-technical operator using 1.0
mg/cm2 as the deciding concentration (see Table 6-3) were almost identical to those found using
RRP rule concentration limits (see Table 6-2). The observed false positive rates, however, were
higher overall for both operators. As with Table 6-2, plaster appeared to have less of an impact
on observed false positive results for the technical operator at the 1.0 mg/cm2 deciding
concentration. However, it appears that metal might have been a factor in the observed false
positive rates for the technical operator. Observed false positive rates for the substrates and
topcoat colors were similar to the overall observed rate for the non-technical operator when
panels were divided based on 1.0 mg/cm2. As in Table 6-2, white lead panels showed the highest
observed false positive results for the technical and non-technical operators.
21
-------
Table 6-3. LeadAVERT™ Test Kit false positive results for panels with confirmed lead
levels < 1 mg/cm2 and false negative results for panels with confirmed lead levels > 1
mg/cm2
Overall
None
White
Yellow
Drywall
Metal
Plaster
Wood
Grey
Red
White
LeadAVER
False Positives'
Technical Non-technical
Operator Operator
122/400 = 31%
6/72 = 8%
114 / 164 = 70%
2 / 164 = 1%
28 / 92 = 30%
53 / 128 = 41%
12/76 = 16%
29 / 104 = 28%
42 / 138 = 30%
41/132 = 31%
39 / 130 = 30%
77 / 402 = 19%
0 / 72 = 0%
75 / 166 = 45%
2 / 164 = 1%
17/92 = 18%
30 / 128 = 23%
14 / 76 = 18%
16 / 106 = 15%
30 / 138 = 22%
21 / 132 = 16%
26 / 132 = 20%
T Test Kit™
False Negatives"
Technical Non-technical
Operator Operator
199 / 534 = 37%
NA
4 / 266 = 2%
195 / 268 = 73%
58 / 144 = 40%
43 / 104 = 41%
49/158 = 31%
49 / 128 = 38%
64 / 172 = 37%
66 / 180 = 37%
69 / 182 = 38%
304 / 530 = 57%
NA
48 / 262 = 18%
256 / 268 = 96%
82 / 144 = 57%
60 / 100 = 60%
80/158 = 51%
82 / 128 = 64%
94 / 168 = 56%
102 / 180 = 57%
108 / 182 = 59%
'False positives on PEMs with confirmed lead levels < 1.0 mg/cm2
"False negatives on PEMs with confirmed lead levels > 1.0 mg/ cm2
NA: If the paint did not contain lead then a false negative is not possible, those entries are 'NA'
(not applicable).
Note that the observed false positive and negative rates presented in this section provide a
general representation of the ability of the LeadAVERT™ Test Kit to correctly identify regulated
lead paint when it is present or absent. The results presented in Table 6-2 provide rates based on
the cut-off concentration (0.8 or 1.2 mg/cm2) as well as all levels evaluated below or above those
concentrations. To evaluate test kit performance based on the RRP rule, lead paint test kits
should have a demonstrated probability (with 95% confidence) of a negative response at or
above the regulated lead level <5% of the time. Test kits should also have a demonstrated
probability (with 95% confidence) of a positive response below the regulated lead level <10% of
the time. Because the RRP rule also indicated that test kit performance would not be based on
lead levels between 0.8 and 1.2 mg/cm2, the false positive and negative probabilities assessed in
this report were then based around the excluded concentrations (of 0.8 and 1.2 mg/cm2). False
positive and negative rates associated with these criteria are discussed in Section 6.4.
22
-------
6.2 Precision
To compute precision, it is first necessary to compute the number of replicate sets with consistent
responses. Replicate sets are defined in the test/QA plan1 to be groups of panels with similar
lead levels. The target lead levels in this experiment were 0, 0.3, 0.6, 1.0, 1.4, 2, and 6 mg/cm2
but the lead levels that were achieved, as confirmed by ICP-AES, sometimes varied from those
target levels. To assemble replicate sets that represented the target lead levels, the panels were
assigned to the replicate set that was nearest their confirmed lead level. In other words, if a
particular panel was targeted for 0.3 mg/cm2 but was measured to have 0.9 mg/cm2 then it was
assigned to the replicate set nearest 0.9 mg/cm2, which is the set labeled 1.0 mg/cm2. Table 6-4
shows the thresholds that defined the replicate set bins as well as the range of measured levels
that fell in each bin.
Table 6-4. Actual lead levels and their replicate set labels
Replicate Set Bin
Label (mg/cm2)
(Closest Target
Lead Level)
0
0.3
0
1
1
6
0
4
2.0
6
0
Bin Thresholds
(mg/cm2)
Targeted to have zero lead
0
-------
Table 6-5. The number of panels at each target level and the number in each replicate set
bin
Replicate Set Bin
(Target level that is closest to the panel's actual measured lead level)
Target Lead Level
(mg/cm2)
0.3 0.6
1.4
0
0.3
0.6
1.0
1.4
2.0
6.0
36
59 12
2 37
4
-
.
-
Total 36 61 53
-
1
25
42
8
1
-
77
-
-
7
22
50
3
-
82
-
-
1
4
14
66
-
85
-
-
-
-
-
2
72
74
Total
36
72
72
72
72
72
72
468
Table 6-6 lists consistency results for the LeadAVERT™ Test Kit by operator type, lead type,
substrate, and lead level. Each table entry lists the number of test results with those
characteristics (N) as well as the proportion of the results that were positive for lead (Pos). Table
entries where the proportion is below 25% or above 75% are 'consistent', meaning that more
than three-quarters of the results were the same (negative or positive). Table entries where the
proportion of positive results ranges from 25% to 75% are considered to be 'inconsistent'.
Inconsistent entries are shaded in the tables. Overall consistency results across all substrates for
white and yellow lead panels for each operator type are also provided in the last row of Table 6-
6. Results across both operators and lead paint types are provided in the last column of the table.
Overall inconsistencies for the non-technical operator were found at the 0.6 and 1.0 mg/cm2 lead
levels for white lead PEMs. There were no inconsistencies for the yellow lead PEMs. This does
not, however, indicate that correct results were provided for all yellow lead PEMs. In fact, there
were no inconsistencies found for the yellow lead PEMs for the non-technical operator because
very few positive results were obtained from those panels, regardless of the PEM's
concentration. This may be indicative of an issue with the LeadAVERT™ Test Kit in operating
properly in the presence of lead chromate (yellow lead). Overall inconsistencies for all PEMs for
the non-technical operator were found at >1.0 mg/cm2 lead. This was also true for most
substrates.
The technical operator had only one inconsistency across all white lead PEMs, and that was at
the 0.3 mg/cm2 lead level with wood as the substrate. Inconsistencies were also found for some
substrates coated with yellow lead. Overall inconsistencies for the technical operator were found
for PEMs with lead concentration between (and including) 0.6 mg/cm2 and 2.0 mg/cm2.
24
-------
In the analysis of overall results, the LeadAVERT™ Test Kit was shown to be inconsistent.
Across both operators and all substrates and lead paint type, the test kit was inconsistent across
all lead levels except 0.0 and 0.3 mg/cm2. These total, overall inconsistencies were influenced
by the lack of positive responses for most yellow lead PEMs.
The consistency results provided in Table 6-6 were used to calculate precision. Precision was
estimated for panels with no lead, white lead, and yellow lead and broken out by type of operator
and then aggregated across both types of operators. For any column in Table 6-6, the precision
is simply the proportion of consistent (unshaded) table entries in the rows for the four different
substrates. The 'All' rows are not counted in the precision calculation because those table entries
are summaries of the entries for the four substrates. Thus, precision was calculated as:
# of unshaded table entries in the
n . . fn/ ... 7. N drywall,metal,plaster,and wood sections
Preciswn(% consistent results) = — • — • n\
total entries in those sections \l)
Table 6-7 lists the results of the precision calculations for the LeadAVERT™ Test Kit. Higher
proportions of consistent results indicate more consistency and higher precision.
25
-------
Table 6-6. LeadAVERT™ Test Kit consistency results by operator type, lead type,
substrate, and lead level
Lead Type
DRYWALL
0
0.3
0.6
1.0
1.4
2.0
6.0
METAL
0
0.3
0.6
1.0
1.4
2.0
6.0
PLASTER
0
0.3
0.6
1.0
1.4
2.0
6.0
WOOD
0
0.3
0.6
1.0
1.4
2.0
6.0
ALL
0
0.3
0.6
1.0
1.4
2.0
6.0
Non-technical Operator
None
N Pos
18 0%
18 0%
18 0%
18 0%
72 0%
White
N Pos
18 6%
32 56%
18 78%
22 73%
18 100%
20 55%
16 44%
24 75%
14 57%
16 88%
12 100%
18 22%
6 100%
10 80%
32 100%
26 100%
18 89%
18 11%
10 20%
22 55%
16 50%
22 73%
20 100%
74 24%
32 47%
88 64%
80 78%
86 84%
68 97%
Yellow
N Pos
14 14%
16 0%
10 0%
30 7%
20 20%
20 20%
20 0%
16 0%
28 0%
10 0%
16 0%
18 0%
4 0%
18 0%
14 0%
24 0%
28 0%
18 0%
10 0%
24 0%
16 13%
20 0%
20 0%
18 0%
48 4%
74 0%
68 3%
84 2%
84 5%
74 5%
Total
N Pos
18 0%
32 10%
16 0%
42 28%
48 42%
42 46%
38 60%
18 0%
40 28%
32 22%
52 38%
24 29%
32 44%
30 50%
18 0%
22 11%
24 50%
24 40%
56 50%
54 50%
36 44%
18 0%
28 6%
34 10%
38 34%
36 25%
42 36%
38 50%
72 0%
122 15%
106 21%
156 34%
164 39%
170 45%
142 51%
Technical Operator
None
N Pos
18 11%
18 11%
18 0%
18 11%
72 8%
White
N Pos
18 22%
32 100%
18 100%
22 100%
18 100%
20 90%
16 88%
24 100%
14 100%
16 100%
16 100%
18 0%
6 100%
10 100%
32 94%
26 100%
18 100%
18 44%
10 80%
20 90%
16 100%
22 91%
20 100%
74 41%
32 88%
86 98%
80 98%
86 98%
72 100%
Yellow
N Pos
14 0%
16 0%
10 0%
30 27%
20 10%
20 40%
20 5%
16 0%
28 0%
10 20%
16 0%
18 39%
4 0%
18 0%
14 14%
24 46%
28 7%
18 89%
10 10%
24 0%
16 13%
20 15%
20 0%
18 56%
48 4%
74 0%
68 6%
84 29%
84 5%
74 55%
Total
N Pos
18 11%
32 11%
16 0%
42 50%
48 63%
42 55%
38 70%
18 11%
40 48%
32 44%
52 50%
24 60%
32 50%
34 69%
18 0%
22 0%
24 50%
24 57%
56 70%
54 54%
36 94%
18 11%
28 27%
34 40%
36 51%
36 58%
42 45%
38 78%
72 8%
122 25%
106 37%
154 51%
164 64%
170 51%
146 78%
ALL
Total
N Pos
36 6%
64 11%
32 0%
84 39%
96 53%
84 51%
76 65%
36 6%
80 38%
64 33%
104 44%
48 44%
64 47%
64 60%
36 0%
44 6%
48 50%
48 49%
112 60%
108 52%
72 69%
36 6%
56 16%
68 25%
74 42%
72 41%
84 41%
76 64%
144 4%
244 20%
212 29%
310 43%
328 51%
340 48%
288 65%
N = number of test results in each bin of the table
Pos = Proportion of those N test results that were 'Positive' for the presence of lead.
Lead levels in the left-most column represent the target level closest to the measured level of lead in the panel.
Shaded cells represent 'inconsistent' results, i.e., % positive is between 25% and 75%
26
-------
Table 6-7. LeadAVERT™ Test Kit precision results by lead type and operator type
No Lead White Lead Yellow Lead1
Non-technical
Technical
All
4/4 = 100%
4/4 = 100%
8/8 = 100%
14/23 = 61%
22/23 = 96%
36/46 = 78%
24/24 = 100%
19/24 = 79%
43/48 = 90%
1 Results were consistently negative across all lead levels for this test kit on yellow lead paint panels, even those
samples containing detectable levels of lead.
The LeadAVERT™ Test Kit was precise on PEMs that contained no lead and yellow lead. Note
that high levels of precision do not necessarily correspond to high likelihood of obtaining correct
test results. The precision for yellow lead panels is indicative of the lack of positive responses
across most of these types of PEMs, including those with detectable levels of lead. The technical
operator results were precise for the white lead PEMs, but those for the non-technical operator
were not.
6.3 Sensitivity
Sensitivity was calculated using the bottom six rows in Table 6-6. These rows aggregate results
across all four substrates. For the white lead and yellow lead columns in these tables, the
sensitivity is the lowest lead level > 1 mg/cm2 that is consistently detected with positive results
(unshaded and > 75%). Ideally the kit would give consistently negative results for lead levels <
1 mg/cm2 and consistently positive results for levels > 1 mg/cm2 so the optimal sensitivity results
would be 1 across every row of Table 6-8. Note that the LeadAVERT™ Test Kit is qualitative
in nature and the goal of the test kit is to indicate to the end user if lead paint is present or not at
the level of 1.0 mg/cm2.
Table 6-8. LeadAVERT™ Test Kit sensitivity results - lowest lead level for which the kit
gave consistent positive results (mg/cm2)
Lead Type
Sensitivity
Non-technical Operator
White Yellow Total
1.4 NA NA
Technical Operator
White Yellow Total
0.6 NA 6.0
All
Total
NA
NA = the kit did not give consistent positive results at any lead concentration level
Across all lead paint types and operators, the LeadAVERT™ Test Kit did not generate consistent
positive results at any lead level. When sensitivity is evaluated by operator type, consistently
positive results were found at 1.4 mg/cm2 on white lead PEMs, but no consistently positive
results were obtained at any lead level for yellow lead panels as well as overall for the non-
technical operator. Consistently positive responses were found at the 0.6 mg/cm2 lead level for
the technical operator on white lead PEMs. The 0.6 mg/cm2 lead level is below the lead level
that should produce a positive response, so sensitivity as measured through positive responses is
indicating false positive associations in this case. Otherwise, the LeadAVERT™ Test Kit as
operated by the technical operator did not generate any consistently positive responses for yellow
27
-------
lead PEMs, and the overall sensitivity for the LeadAVERT™ Test Kit as used by the technical
operator was 6.0 mg/cm2, above the desired 1.0 mg/cm2 lead level.
6.4 Modeled Probability of Test Kit Response
Table 6-9 lists the explanatory variables which had significant (p<0.05) univariate associations
with the probability of obtaining a positive test kit result. All potential explanatory variables
except for topcoat color showed a statistically significant univariate association with the
probability of a positive response. Lead level, lead type, operator type, and substrate type were
retained in the multivariable model after backward selection. Table 6-10 lists the parameter
estimates for the multivariable logistic regression models from the Stata SIMEX program. There
were no statistically significant interactions between the categorical covariates.
Table 6-11 lists the modeled probability of a positive test result for the LeadAVERT™ Test Kit
when the lead level is 0.8 mg/cm2 (PREDICTION) along with the upper bound of a 95%
prediction interval (UPPER). That upper bound can be considered to be a worst-case estimate of
the false positive probability when the true lead level is 0.8 mg/cm2 (FALSE POS RATE).
Ideally the numbers in the UPPER/FALSE POS RATE column would be < 10%. Note that the
FALSE POS RATE in Table 6-11 is higher than those in Tables 6-2 and 6-3. In the earlier tables
the rates considered panels at a variety of comparatively low lead levels so some cases should
have been easier for the kit to obtain the correct answer. In Table 6-11, the false positive rate is
evaluated only at 0.8 mg/cm2 so the rate does not benefit from the comparatively lower lead
concentrations. Evaluating at only this level also ensures that a test kit can adequately perform at
concentrations of lead paint closest to the current regulatory level.
Based on the upper prediction bound estimates shown in Table 6-11, the LeadAVERT™ Test Kit
met the false positive criteria for yellow lead. Thus a false positive rate of less than 10% would
be expected to be achieved by both the technical and non-technical operator on all substrates
with yellow lead. Table 6-11 indicates that the lowest upper prediction bounds (1.7% to 2.7%)
would be expected when the non-technical operator was using the LeadAVERT™ Test Kit to
evaluate yellow lead paint on any substrate.
Table 6-9. LeadAVERT™ Test Kit univariate associations between probability of positive
response and explanatory variables
Explanatory Variable Significant Univariate Association? Included in Multivariable Model?
Lead Level Yes (p-value < 0.0001) Yes
Lead Type Yes (p-value < 0.0001) Yes
Operator Type Yes (p-value < 0.0001) Yes
Substrate Type Yes (p-value = 0.0473) Yes
Topcoat Color No (p-value = 0.9385) No
28
-------
Table 6-10. LeadAVERT™ Test Kit multivariable Stata SIMEX logistic regression
parameter estimates
Simulation extrapolation
Residual df = 2003
Variance Function: V(u) = u(l-u)
Link Function : g(u) = log(u/(l-u))
No. of obs =
Bootstraps reps =
Wald F(6,2003)
Prob > F
[Bernoulli]
[Logit]
2010
199
79.77
0.0000
I Bootstrap
result |
Operator type:
technical |
Substrate :
drywall |
metal |
plaster |
Lead type :
white |
lead level |
constant |
Coef. Std. Err.
(non- technical
1.237943
is the
1355222
(wood is the reference
.3039538
.4827346
.4604897
(yellow is the
4.036347
.3670463
-4.768359
1622059
1822251
1746209
t
reference
9.13
level)
1.87
2.65
2.64
P>| 1 1
level)
0.000
0.061
0.008
0.008
[95% Conf.
.9721642
-.0141561
.125364
.1180321
Interval]
1.503723
.6220638
.8401051
.8029473
reference level)
2211071
0272954
2702506
18.26
13.45
-17.64
0.000
0.000
0.000
3. 602723
.3135161
-5.298361
4.469971
.4205766
-4.238357
Table 6-12 lists the modeled probability of a positive test result for the LeadAVERT™ Test Kit
when the lead level is 1.2 mg/cm2 (PREDICTION) along with the lower bound of a 95%
prediction interval (LOWER). The difference between the lower bound and 100% could be
considered to be a worst-case estimate of the false negative probability when the true lead level
is 1.2 mg/cm2 (FALSE NEG RATE). Ideally, for purposes of the RRP rule, the numbers in the
FALSE NEG RATE column would be < 5%. The model results shown in Table 6-12 indicate
that in no scenario could a false negative rate of <5% be obtained using the LeadAVERT™ Test
Kit.
Note that the FALSE NEG RATE in Table 6-12 is higher than those in Tables 6-2 and 6-3. In
the earlier tables, the false negative rates considered panels at a variety of comparatively high
lead levels so some cases should have been easier for the kit to obtain the correct answer. In
Table 6-12, the false negative rate is evaluated only at 1.2 mg/cm2 so the rate does not benefit
from the comparatively higher lead concentrations. Evaluating at only this level also ensures
that a test kit can adequately perform at concentrations of lead paint closest to the current
regulatory level.
29
-------
Table 6-11. LeadAVERT™ Test Kit modeled probability of positive test results and upper
95% prediction bound when lead level = 0.8 mg/cm2
OPERATOR LEAD TYPE SUBSTRATE
NON-TECHNICAL
TECHNICAL
DRYWALL
METAL
WHITE
PLASTER
WOOD
DRYWALL
METAL
YELLOW
PLASTER
WOOD
DRYWALL
METAL
WHITE
PLASTER
WOOD
DRYWALL
METAL
YELLOW
PLASTER
WOOD
LEAD LEVEL
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.8
PREDICTION
46.6%
51.1%
50.6%
39.2%
1.5%
1.8%
1.8%
1.1%
75.1%
78.3%
77.9%
69.0%
5.1%
6.0%
5.9%
3.8%
UPPER (FALSE POS RATE)
52.4%
56.8%
56.6%
44.6%
2.4%
2.6%
2.7%
1.7%
79.3%
82.4%
82.0%
73.8%
7.7%
8.5%
8.6%
5.7%
30
-------
Table 6-12. LeadAVERT™ Test Kit modeled probability of positive test results, lower
95% prediction bound, and corresponding conservative estimate of the false negative rate
when lead level = 1.2 mg/cm2
OPERATOR LEAD TYPE SUBSTRATE
NON-TECHNICAL
TECHNICAL
DRYWALL
METAL
WHITE
PLASTER
WOOD
DRYWALL
METAL
YELLOW
PLASTER
WOOD
DRYWALL
METAL
WHITE
PLASTER
WOOD
DRYWALL
METAL
YELLOW
PLASTER
WOOD
LEAD LEVEL
1.2
1.2
1.2
1.2
1.2
1.2
1.2
1.2
1.2
1.2
1.2
1.2
1.2
1.2
1.2
1.2
PREDICTION
50.3%
54.8%
54.2%
42.8%
1.8%
2.1%
2.0%
1.3%
77.7%
80.7%
80.3%
72.0%
5.8%
6.9%
6.7%
4.4%
LOWER
44.7%
49.0%
48.1%
37.5%
1.1%
1.5%
1.4%
0.9%
73.4%
76.3%
76.0%
67.1%
3.8%
4.9%
4.6%
2.9%
FALSE NEG RATE
55.3%
51.0%
51.9%
62.5%
98.9%
98.5%
98.6%
99.1%
26.6%
23.7%
24.0%
32.9%
96.2%
95.1%
95.4%
97.1%
As another means of reporting the results for the LeadAVERT™ Test Kit, modeled probability
curves were also plotted based on the results of the regression analysis. To better understand the
information being provided in these probability curves, a brief explanation is presented here.
Figure 6-1 shows that for the perfect or ideal test kit, the probability of a positive test result
would be a step function. The probability of a positive result would be zero below 1.0 mg/cm2
and 100% at or above 1.0 mg/cm2. Under the RRP rule, a test kit must yield a demonstrated
probability (with 95% confidence) of no more than 10% false positives at lead concentrations
below 0.8 mg/cm2 and a demonstrated probability (with 95% confidence) of no more than 5%
false negatives at concentrations above 1.2 mg/cm2. Figure 6-1 also shows a performance curve
for a hypothetical test kit that achieves those rates. The upper bound of the 90% prediction
interval is at 10% at 0.8 mg/cm2 and the lower bound of the prediction interval is at 95% at 1.2
mg/cm2.
One way to think of the test kit performance guidelines is in terms of regions of the probability
plots. Figure 6-2 demonstrates this concept. For the kit to be within limits set up by the RRP
rule, the probability curve must trace a path through the white region in the figure and must not
stray into the shaded regions. If the curve crosses the shaded region at the left side of the graph
then there are lead levels < 0.8 mg/cm2 where the false positive rate is > 10%. If the curve
crosses the shaded region at the right side of the graph then there are lead levels > 1.2 mg/cm2
where the false negative rate is > 5%. Either type of intersection between the curve and the
shaded region indicates that the kit does not meet the performance levels stipulated in the RRP
rule.
31
-------
Note that results for the region between 0.8 and 1.2 mg/cm2 were not discussed in this report.
This is consistent with the RRP rule stipulation that lead concentrations between 0.8 and 1.2
mg/cm2 were not to be considered for the evaluation of the performance of lead paint test kits4.
Figures 6-3 through 6-6 show the predicted probability of obtaining a positive test result using
the LeadAVERT™ Test Kit over the full range of explanatory variables along with the bounds of
a 90% prediction interval. Note that the upper and lower bounds of the 90% prediction interval
may also be considered to be upper and lower 95% prediction bounds for one-sided inference.
In every instance the upper end of the probability curves (above 1.2 mg/cm2) pass through the
shaded regions of the plot. This indicates with 95% confidence that there are lead levels above
1.2 mg/cm2 where the predicted false negative rate is greater than 5%, which would be outside of
the false negative criteria established in the RRP rule. For the yellow lead plots the low ends of
the curves (below 0.8 mg/cm2) avoid the shaded regions indicating that the false positive
performance of the test kit meets the RRP rule for that combination of variables.
32
-------
tO
8.
0>
Ł
.Q
CO
o
a!
100%-
80%-
60%-
40%-
20%-
95%
Perfect Kit
Acceptable Kit: Upper 90% Bound (1)
Acceptable Kit
Acceptable Kit: Lower 90% Bound (2)
1 Upper bound is above and left of the "Acceptable Kit" curve
2 Lower bound is below and right of the "Acceptable Kit" curve
10%
0
0.8 { 1.2
Lead Level (mg/cm^Z)
Figure 6-1. Probability curves that represent test kit results that are both perfect (red line) and within RRP rule criteria (black
solid line).
33
-------
tO
8.
0>
Ł
.Q
CO
o
a!
100% H
80%-
60%-
40% H
20%-
95%
10%
0
0.8 i 1.2
Lead Level (ing/cm^)
Figure 6-2. Probability curves with shaded region to denote performance results that meet RRP rule false positive and
negative criteria. Test kits with curves that fall within the white region and avoid the shaded region meet the RRP rule.
34
-------
Operatorjype = TECHNICAL, Leadjype = YELLOW
100%-
^ 75%-
3
8 50%-
et
.ti
S 75%-
(Q
A
2 50%-
&
25%-
0%-
MM
=»
M.
»
...
*
DRYWALL
PLASTER
„
•&«
•Ml
METAL
WOOD
1
0
0.81.2
6 0 °-81'2
Lead Level (mg/cm^Z)
95%
10%
95%
10%
6
Figure 6-3. LeadAVERT™ Test Kit predicted probability of positive test result (solid lines) with 90% prediction interval
(dotted lines) for the technical operator and yellow lead paint on various substrates.
35
-------
Operatorjype = NON-TECHNICAL, Leadjype = YELLOW
100%-
^ 75%-
3
8 50%-
et
.! 25%-
^^
<5r u / o
M-
0 100%-
1 75%-
(Q
2 50%-
25%-
0%-
• • • I
• •PI
DRYWALL
PLASTER
B^pVf
METAL
WOOD
0
0.81.2
6 0 °-81'2
Lead Level (mg/cm^Z)
95%
10%
95%
10%
6
Figure 6-4. LeadAVERT™ Test Kit predicted probability of positive test result (solid lines) with 90% prediction interval
(dotted lines) for the non-technical operator and yellow lead paint on various substrates.
36
-------
Operatorjype = TECHNICAL, Leadjype = WHITE
100%-
^ 75%-^
a ***
8 50%-^
et
I
~
S 75%-^
flj **
-°
2 50% -J
&
25%-^
0%-|
DRYWALL
a^BJ
PLASTER
— .
,**
METAL
•• JMl
,^**
WOOD
1
*ii^
i^™
0
0.81.2
6 0 °-81'2
Lead Level (mg/cm^Z)
95%
10%
95%
10%
6
Figure 6-5. LeadAVERT™ Test Kit predicted probability of positive test result (solid lines) with 90% prediction interval
(dotted lines) for the technical operator and white lead paint on various substrates.
37
-------
Operatorjype = NON-TECHNICAL, Leadjype = WHITE
3
0}
0)
55
o
A
M-
o
.ti
S
RJ
0
o.
1
100% -f
75%-^
50% -J
25%-
_
u /o
100%-
75% -J
50% -J
25%-^
0%-|
DRYWALL
.^^^
PLASTER
^ i^-
METAL
.^
WOOD
^^
»*^j«*^
0
0.81.2
6 0 °-81'2
Lead Level (mg/cm^Z)
95%
10%
95%
10%
6
Figure 6-6. LeadAVERT™ Test Kit predicted probability of positive test result (solid lines) with 90% prediction interval
(dotted lines) for the non-technical operator and white lead paint on various substrates.
38
-------
Based on the modeled probabilities shown in Figures 6-3 through 6-6, threshold values for false
positive and negative rates were established for the LeadAVERT™ Test Kit. For the false
positive rate, this threshold value is the lead level, with 95% confidence, below which the
LeadAVERT™ Test Kit would yield fewer than 10% false positive results. For the false
negative rate, this threshold value is the lead level, with 95% confidence, above which the
LeadAVERT™ Test Kit would yield fewer than 5% false negative results. These threshold
values are then the lead levels where the LeadAVERT™ Test Kit is predicted to meet the false
positive and negative criteria set forth in the RRP rule.
Table 6-13 presents the false positive and negative threshold values for the LeadAVERT™ Test
Kit. Threshold lead levels are provided for each substrate, lead type, and operator combination
shown in Tables 6-11 and 6-12.
Table 6-13 indicates that overall, across all factors, no false positive or negative threshold could
be established. The modeled probabilities for white lead did not yield thresholds for either false
positives or false negatives. The modeled probabilities for yellow lead did not yield any false
negative thresholds, though false positive thresholds were established for some combinations of
variables. The table lists false positive thresholds for yellow lead, all of which exceed 1.2
mg/cm2.
Standardized Pearson residuals were calculated to assess goodness of fit of the logistic regression
models. For the LeadAVERT™ Test Kit model, 91.5% of the residuals had absolute values
smaller than two. Note that Figures 6-3 through 6-6 did not show typical logistic regression
curve sigmoid shapes with asymptotes at 0% and 100% probability of positive result. For yellow
lead, the predicted probabilities were always low (never higher than 50%) and for white lead the
predicted probabilities were always fairly high (never lower than 25%). The modeled
probabilities did not approximate those of the ideal or acceptable test kit curves illustrated in
figures 6-1 and 6-2.
39
-------
Table 6-13. LeadAVERT™ Test Kit false positive and negative threshold values (95%
confidence) based on the modeled probability of test results
OPERATOR TYPE LEAD TYPE SUBSTRATE
NON-TECHNICAL
TECHNICAL
DRYWALL
METAL
WWte PLASTER
WOOD
DRYWALL
METAL
Ye"OW PLASTER
WOOD
DRYWALL
METAL
WWte PLASTER
WOOD
DRYWALL
METAL
Ye"OW PLASTER
WOOD
FALSE POSITIVE
THRESHOLD
(mg/cm2)
NA
NA
NA
NA
5.18
4.85
4.81
6
NA
NA
NA
NA
1.63
1.32
1.28
2.58
FALSE NEGATIVE
THRESHOLD (mg/cm2)
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
OVERALL
NA
NA
NA in the FALSE POSITIVE THRESHOLD column means that the false positive rate was > 10% for all lead
levels.
NA in the FALSE NEGATIVE THRESHOLD column means that the false negative rate was > 5% for all lead
levels.
6.5 Matrix Effect
The matrix effects for the LeadAVERT™ Test Kit were evaluated with results in Table 6-10.
The variables that were retained in the multivariable logistic regression model each add
significant explanatory power to their respective models. Those variables are significantly
associated with the probability of obtaining a positive test result from the kits tested in this study.
For the LeadAVERT™ Test Kit, Table 6-10 indicates that after controlling for the significant
covariates, the likelihood of a positive test result is positively and significantly associated with:
higher lead levels, testing by a technical operator, metal and plaster substrates, and white lead.
Note that there was one technical operator and one non-technical operator for the LeadAVERT™
Test Kit. Although the variable Operator Type 'technical vs. non-technical' is statistically
significant for the LeadAVERT™ Test Kit, it is not clear from these data whether the differences
are due to technical training or due to some other factor associated with those individuals. TSAs
40
-------
were performed to ensure that each operator was using the test kit properly and according to the
manufacturer's instructions.
6.6 Operational Factors
Both the technical and non-technical operator found the LeadAVERT™ Test Kit instructions to
be clear, informative, and easy to follow. The non-technical operator received no training from
the vendor and relied solely on the test kit instructions for his understanding of the operation of
the test kit. Both the technical and non-technical operator stated that after reading the instruction
manual and a couple of trial runs, they were prepared to use the test kit. The solutions used for
different steps were easily identifiable within the kit and the storage conditions of the reagents
were readily marked. All reagents came prepared and ready to use.
The LeadAVERT™ Test Kit came in packs of 20 strips. Each kit included a stencil, 20 square
test vials, one bottle of extraction solution, and one container of 20 test strips. The user was
expected to supply a pencil (for tracing the stencil), tape (for adhering the vial to the sampling
area), a utility knife (for cutting to paint from the sampling area), and a timer or stopwatch (for
tracking the time periods noted in the instructions). As per the vendor's instructions, the kit
needed to be stored at room temperature. Even though the kits had no expiration dates, the
instructions indicated to discard anything that might be expired. As indicated in the package, the
test strips had a 90 day shelf life after opening the container holding them.
The test kit instructions indicated that appropriate safety precautions should be taken when
working with lead-based paint, such as wearing the necessary protective equipment. The
extraction solution was weak nitric acid and the kit instructions note that protective equipment
should be used when handling the acid. Both the technical and non-technical operators followed
general laboratory safety procedures and wore a lab coat, protective eyewear, and gloves at all
times. No material safety data sheets (MSDS) were provided with the test kit.
Only one test strip, one vial, and 11 drops of extraction solution are produced as waste for a
single test. If a positive result is obtained with the test kit, the used test vial and extraction
solution as well as the used test strip would be considered lead waste. As such, the
LeadAVERT™ Test Kit instructions indicate that local regulations pertaining to disposal of lead
waste should be followed. Otherwise, according to the vendor non-positive tests (indicating that
no lead at the regulated level is present) would be considered non-lead waste and could be
disposed of with normal waste procedures, with any unused extraction solution flushed down the
sink or toilet with plenty of water. (Note: Because regulations for the disposal of wastes
generated from the use of lead test kits may vary from state to state, EPA recommends that test
kit users contact their state government agency for proper waste disposal requirements.)
Interpretation of the test strips for the LeadAVERT™ Test Kit was sometimes difficult. It was
difficult to discern if the two lines were equal or if one line was darker than the other in some
samples. In all samples, it was up to the discretion of the user to determine the darkness of lines
on the test strip.
The LeadAVERT™ Test Kit was quick and easy to operate. Operation of the test kit took
approximately 15 minutes for one sample for both the technical and non-technical operator. No
power supply was needed for the operation of the test kit. As of the writing of this report, the
cost for each 20-sample LeadAVERT™ Test Kit package is $39.95.
41
-------
Chapter 7
Performance Summary
The overall observed false positive rate for the Lead A VERT™ Test Kit on PEMs with
confirmed lead levels of < 0.8 mg/cm2 was 22% for the technical operator and 12% for the non-
technical operator. Evaluation of white lead PEMs produced the highest observed false positive
rate for both operators. Observed false negative rates for both the technical and non-technical
operator indicated that the LeadAVERT™ Test Kit did not perform well on yellow lead paint.
Substrate and topcoat color did not appear to have an impact, as these observed false negative
rates were similar to the overall rates found for each operator. Observed false negative rates for
the technical operator were 37% overall, 71% on yellow lead PEMs, and 2% on white lead
PEMs. Observed false negative rates for the non-technical operator were 56% overall, 96% on
yellow lead PEMs, and 15% on white lead PEMs. It is possible that the chromate in the yellow
lead paint interfered with the proper operation of the LeadAVERT™ Test Kit test strip.
Overall observed false negative rates on PEMs with confirmed lead levels <1.0 mg/cm2 for both
the technical and non-technical operator were almost identical to those found on PEMs with
confirmed lead levels of <0.8 mg/cm2. The observed false positive rates, however, were higher
overall for both operators than those found on PEMs with confirmed lead levels >1.2 mg/cm2:
31% for the technical operator and 19% for the non-technical operator.
Across both operators and all substrates and lead paint type, the LeadAVERT™ Test Kit
responses were inconsistent across all lead levels except 0.0 and 0.3 mg/cm2. These total, overall
inconsistencies were influenced by the lack of positive responses for most yellow lead PEMs.
Results from the LeadAVERT™ Test Kit indicated 100% precision on PEMs that contained no
lead and 90% precision on PEMs that contained yellow lead. The precision for yellow lead
panels is indicative of the lack of positive responses across most of these types of PEMs,
including those with detectable levels of lead. The technical operator provided results with 96%
precision for the white lead PEMs, but those for the non-technical operator were 61%.
Across all lead paint types and operators, the LeadAVERT™ Test Kit did not generate consistent
positive results at any lead level. When sensitivity was evaluated by operator type, the lowest
lead level for which consistently positive results were found was 1.4 mg/cm2 on white lead
PEMs, but consistently positive results were not obtained at any lead level for yellow lead
panels. Consistently positive responses were found at the 0.6 mg/cm2 lead level for the technical
operator on white lead PEMs. The 0.6 mg/cm2 lead level was below the lead level that should
produce a positive response, so sensitivity as measured through positive responses was indicating
false positive associations in this case. Otherwise, the LeadAVERT™ Test Kit as operated by
the technical operator did not generate any consistently positive responses for yellow lead PEMs,
42
-------
and the overall sensitivity for the LeadAVERT™ Test Kit as used by the technical operator was
6.0 mg/cm2, well above the desired 1.0 mg/cm2 lead level.
Under the RRP rule4, a test kit must yield a demonstrated probability (with 95% confidence) of
no more than 10% false positives at lead concentrations below 0.8 mg/cm2 and a demonstrated
probability (with 95% confidence) of no more than 5% false negatives at concentrations above
1.2 mg/cm2 to meet the rule criteria. Based on the upper bound estimates of the modeled
probability of the LeadAVERT™ Test Kit, the technology met the false positive criteria for
yellow lead. Thus, a false positive rate of less than 10% would only be expected to be achieved
by both the technical and non-technical operator on all substrates with yellow lead. The lowest
false positive rates (from 1.7% to 2.7%) would be expected when the non-technical operator was
using the LeadAVERT™ Test Kit to evaluate yellow lead paint on wood, drywall, plaster, and
metal. The model results indicate that a false negative rate <5% could not be obtained using the
LeadAVERT™ Test Kit.
Based on the modeled probabilities, no overall false positive threshold value (i.e., the lead level,
with 95% confidence, below which the test kit would yield fewer than 10% false positive results)
could be established. Similarly, across all factors of significance, no false negative threshold
value (the lead level, with 95% confidence, above which the test kit would yield fewer than 5%
false negative results) could be established for the LeadAVERT™ Test Kit. The modeled
probabilities for yellow lead did not yield any false negative thresholds that meet the RRP rule,
though thresholds were established for some combinations of variables. All of these thresholds
0 0
exceed 1.2 mg/cm ; ideally those thresholds would lie below 0.8 mg/cm .
After controlling for the significant covariates, the likelihood of a positive test result is positively
and significantly associated with: higher lead levels, testing by a technical operator, metal and
plaster substrates, and white lead. It is not significantly and positively associated with testing by
a non-technical operator, wood and drywall substrates, and yellow lead.
Both the technical and non-technical operator found the LeadAVERT™ Test Kit instructions to
be clear, informative, and easy to follow. The solutions used for different steps were easily
identifiable within the kit and the storage conditions of the reagents were readily marked. All
reagents came prepared and ready to use.
The LeadAVERT™ Test Kit came in packs of 20 strips. Each kit included a stencil, 20 square
test vials, one bottle of extraction solution, and one container of 20 test strips. The user was
expected to supply a pencil (for tracing the stencil), tape (for adhering the vial to the sampling
area), a utility knife (for cutting to paint from the sampling area), and a timer or stopwatch (for
tracking the time periods noted in the instructions).
Only one test strip, one vial, and 11 drops off extraction solution are produced as waste for a
single test. If a positive result is obtained with the test kit, the used test vial and extraction
solution as well as the used test strip would be considered lead waste. As such, the
LeadAVERT™ Test Kit instructions indicate that local regulations pertaining to disposal of lead
waste should be followed. Otherwise, according to the vendor non-positive tests (indicating that
no lead is present) would be considered non-lead waste and could be disposed of with normal
waste procedures, with any unused extraction solution flushed down the sink or toilet with plenty
of water. (Note: Because regulations for the disposal of wastes generated from the use of lead
43
-------
test kits may vary from state to state, EPA recommends that test kit users contact their state
government agency for proper waste disposal requirements.)
Interpretation of the test strips for the LeadAVERT™ Test Kit was sometimes difficult. It was
difficult to discern if the two lines were equal or if one line was darker than the other in some
samples. In all samples, it was up to the discretion of the user to determine the darkness of lines
on the test strip.
The LeadAVERT™ Test Kit was quick and easy to operate. Operation of the test kit took
approximately 15 minutes for one sample for both the technical and non-technical operator. No
power supply was needed for the operation of the test kit. The cost for each 20-sample
LeadAVERT™ Test Kit package at the time of the test was $39.95.
44
-------
Chapter 8
References
1. Test/QA Plan for Verification of Qualitative Spot Test Kits for Lead in Paint, Battelle,
Columbus, Ohio, March 29, 2010.
2. "Lead Renovation, Repair, and Painting Program Final Rule", Federal Register, 73:78
(April 22, 2008), p.21692.
3. President's Task Force on Environmental Health Risks and Safety Risks to Children.
Eliminating Childhood Lead Poisoning: A Federal Strategy Targeting Lead Paint Hazards.
2000. Washington, DC.
4. NISTIR 6398, "Spot Test Kit for Detecting Lead in Household Paint: A Laboratory
Evaluation," NIST, May 2000.
5. ASTM E1828, "Standard Practice for Evaluating the Performance Characteristics of
Qualitative Chemical Spot Test Kits for Lead in Paint," ASTM International.
6. Revised Plan For Development And Production Of Performance Evaluation Materials For
Testing Of Test Kits For Lead In Paint Under The Environmental Technology Verification
Program, Battelle, Columbus, Ohio, September 2008.
7. Standard Operating Procedure for Collection of Dried Paint Samples for Lead Determination,
Battelle, Columbus, OH.
8. ASTM El 729, "Standard Practice for Field Collection of Dried Paint Samples for
Subsequent Lead Determination," ASTM International.
9. Quality Management Plan (QMP) for the ETV Advanced Monitoring Systems Center, U.S.
EPA Environmental Technology Verification Program, prepared by Battelle, Columbus, OH,
Version 7.0, 2008.
10. "Environmental Technology Verification Program Quality Management Plan", December
2002, EPA/600/R-03/021.
11. Standard Operating Procedure for Analysis of Lead in Paint Samples for Battelle, Schneider
Laboratories, Inc., Doc #111-044-10-011, January 20, 2010, revised February 24, 2010 and
April 25, 2010.
45
-------
12. Cook, J. and L. A. Stefanski. 1994. A simulation extrapolation method for parametric
measurement error models. Journal of the American Statistical Association 89: 1314-1328.
13. Hardin, J. W., H. Schmiediche, and R. J. Carroll. 2003a. The simulation extrapolation
method for fitting generalized linear models with additive measurement error. Stata Journal
3(4): 373-385.
14. Hardin, J. W. and R. J. Carroll. 2003b. Measurement error, GLMs, and notational
conventions. Stata Journal 3(4): 328-340.
15. Hardin, J. W., H. Schmiediche, and R. J. Carroll. 2003c. The regression-calibration method
for fitting generalized linear models with additive measurement error. Stata Journal 3(4):
360-371.
16. Stefanski, L. A. and J. Cook. 1995. Simulation extrapolation: The measurement error
jackknife. Journal of the American Statistical Association 90(432): 1247-1256.
46
-------
Appendix A
Performance Evaluation Materials
Summary Information
-------
List of Abbreviations and Acronyms
AMS
ASTM
CCV
CoV
CRM
DI
EPA
ESTE
ETV
FT
HVLP
ICP-AES
ICS
ICV
LCS
mg/cm2
mg/kg
mL
MSDS
NLLAP
PEM
ppb
QA
QC
QMP
RH
RPD
SOP
Advanced Monitoring Systems
American Society for Testing and Materials
continuing calibration verification
coeffi ci ent of vari ati on
certified reference material
deionized
U.S. Environmental Protection Agency
Environmental and Sustainable Technology Evaluations
Environmental Technology Verification
film thickness
high volume/low pressure
inductively coupled plasma-atomic emission spectrometry
interference check sample
initial calibration verification
laboratory control spike
micrograms per liter
microliters
milligrams per centimeter squared
milligrams per kilogram
milliliter
material safety data sheets
National Lead Laboratory Accreditation Program
performance evaluation material
parts per billion
quality assurance
quality control
quality management plan
relative humidity
relative percent difference
standard operating procedure
A-2
-------
Section Al
Preparation of
Performance Evaluation Materials
A-3
-------
Executive Summary
Battelle prepared a batch of performance evaluation materials (PEMs) for use in an
Environmental Technology Verification (ETV) program evaluation of the performance of lead
paint test kits. These PEMs encompass two lead types (white lead [lead carbonate] and yellow
lead [lead chromate]), four separate substrates (metal, wood, drywall, and plaster), and six lead
levels within each lead type (0.3, 0.6, 1.0, 1.4, 2.0, and 6.0 mg/cm2). The goal of the production
was to produce panels at a specified lead level with minimal variability across and within panels.
The study design called for a verification and homogeneity study involving inductively coupled
plasma (ICP) testing of the painted metal panels to determine applied lead levels. Initial
application procedures included spray application for paints at 2.0 and 6.0 mg/cm2, but testing
indicated that spray application yielded high variability in lead levels. As a result, the Battelle
team, in consultation with U.S. Environmental Protection Agency (EPA), decided to apply all
lead paint layers via drawdown bar, which enables more precision in the thickness of the paint
layer applied. Later in the development process, continued high variability measurements led to
the team's decision to include silica in the formulations of each lead paint to thicken the paint
and allow for a more even coating.
Verification and homogeneity testing was conducted for all 12 lead paints as well as the one no-
lead control paint. Verification testing determined the formulation and drawdown bar best suited
to yield a particular lead level. Homogeneity test results were assessed for proximity to target
lead levels, lead level range, and variability within and between panels. All paints passed
verification and homogeneity testing.
After completing the verification and homogeneity testing, base paint layers were applied for all
12 sets of lead paints (two lead types by six lead levels) and the no-lead paint. Paint chips were
sampled and analyzed from the metal reference panels within each set of PEMs. The metal
reference panel measurements met target specifications for all sets of PEMs. All nine sets of 468
panels each were appropriately labeled and packaged. All reference PEM concentrations and
homogeneity results were reviewed and approved by EPA prior to full-scale production of a set.
Study Design
The initial study design specified production of the ETV PEMs using six lead levels (0.3, 0.6,
1.0, 1.4, 2.0, and 6.0 mg/cm2), two lead types (white and yellow lead), four substrates (wood,
metal, drywall, and plaster), and three topcoat colors (white, red-orange, and grey-black), as
specified in Table A-l. For the wood substrates, both poplar and pine wood panels were
produced, segregated, and uniquely labeled to be consistent with the design in Table A-l.
The final design specified production of 624 panels for each of seven test kits for a total of 4,368
panels. Late in the development process, the planned evaluation design changed so that only 468
panels were required to test each of nine test kits for a total of 4,212 PEMs needed for the ETV
test.
A-4
-------
Table A-l: PEMs Produced for ETV Evaluation
Lead Type
Control Blank
White Lead
(Lead
Carbonate)
Yellow Lead
(Lead
Chromate)
Lead Level
(mg/cm2)
0
0.3
0.6
1.0
1.4
2.0
6.0
0.3,0.6,1.0,
1.4,2.0, 6.0
Substrate
Wood-Poplar
Wood-Pine
Metal
Drywall
Plaster
Wood-Poplar
Wood-Pine
Metal
Drywall
Plaster
Wood-Poplar
Wood-Pine
Metal
Drywall
Plaster
Wood-Poplar
Wood-Pine
Metal
Drywall
Plaster
Wood-Poplar
Wood-Pine
Metal
Drywall
Plaster
Wood-Poplar
Wood-Pine
Metal
Drywall
Plaster
Wood-Poplar
Wood-Pine
Metal
Drywall
Plaster
Wood-Poplar
Wood-Pine
Metal
Drywall
Plaster
Subtotal - Per Test Kit
# Samples Produced Per Test Kit by To
White
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
Red-Orange
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
Grey-Black
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2
2
4
4
4
2 panels per cell for Wood substrates, 4
panels per cell for other substrates (same
design as White Lead panels)
208
208
208
pcoat Color
Total
6
6
12
12
12
6
6
12
12
12
6
6
12
12
12
6
6
12
12
12
6
6
12
12
12
6
6
12
12
12
6
6
12
12
12
36
36
72
72
72
624
7 Test Kits
42
42
84
84
84
42
42
84
84
84
42
42
84
84
84
42
42
84
84
84
42
42
84
84
84
42
42
84
84
84
42
42
84
84
84
252
252
504
504
504
4,368
The original design plan called for a target lead concentration of 0.3 mg/cm for a set of PEMs.
During the writing of the ETV test/quality assurance plan, preliminary ICP results indicated that
the target level for this set of PEMs might be closer to 0.4 mg/cm2. The preliminary results were
used in the ETV test plan.
A-5
-------
Substrate Preparation
The ETV PEMs included four different substrate types - metal, wood, drywall, and plaster;
although two types of wood (pine and poplar) were utilized. The following bulleted lists
describe the steps taken to prepare each of the types of substrates.
Metal
• Iron Phosphate Steel panels 0.032" x 3" x 3" were placed in an isopropyl alcohol bath and
carefully wiped and dried before being placed in plastic bags prior to coating.
• The solvent wipe step was performed to ensure that residue oils/fingerprints from the
manufacturing processes were removed.
Wood
• Wood (pine and poplar) was purchased in 4" widths, planed, and cut into 3" x 3" panels.
• PEMs were placed into constant temperature and humidity conditioning rooms prior to
coating application to ensure uniform water content through each panel prior to coating.
[Note that plaster and drywall panels are less sensitive to water absorption prior to
coating.]
Drywall
• 4" x 8" x 3/8" gypsum drywall sheets were cut into 3" x 3" panels.
Plaster
• Two j oint compound materials were evaluated for ease of application and smoothness to
ensure the best surface for coating. USG Joint Compound provided the smoothest
surface and was used to coat panels at about 1/32" thickness.
• A 3 " x 4' strip of 3/8" thick gypsum drywall was placed into jig, then plaster joint
compound was smoothed over top surface to a precise 1/32" thickness. Plastered drywall
strips were then cut down into 3" x 3" panels.
Sealer Application to Drywall and Plaster PEMs
• Stacks of drywall and plaster PEMs were sealed on cut edges with no lead latex
primer/sealer to eliminate dusting.
All panels were then placed in constant temperature and humidity conditioning rooms prior to
coating application.
Spray Application Facilities and Equipment
Battelle's laboratory includes a walk-in spray booth capable of this type of production as well as
air handling equipment and monitors to ensure the safety of Battelle staff. Although the 0.3 and
6.0 mg/cm2 white lead and no-lead paints were applied via spray application, all other
application of lead paint layers was performed using drawdown bars in a laboratory setting. All
topcoats were applied by spray application in the spray booth. Details on the equipment used in
these processes are listed below.
A-6
-------
Spray Booth
• 10' x 10' x 7.5' double door spray booth
• Compressed air supply for spray equipment
• Spray equipment consists of a high volume/low pressure (HVLP) gravity fed DeVilbiss
spray gun
• Plastic sheeting covering walls and floor to minimize clean-up time
Conditioning Rooms
• Constant temperature (75°Farenheit)) and humidity (50% relative humidity [RH])) rooms
for substrate conditioning (the variability in temperature and RH is not tracked in those
rooms)
• Substrates were conditioned both before and after coating application. Wood substrates
were conditioned a minimum of two weeks prior to coating. All substrates were
conditioned a minimum of 48 hours after coating and before bagging and wrapping.
• Plastic covering was placed on the floor to minimize clean-up time after transporting
drying racks from the coating application lab into the conditioning rooms.
Environmental Health and Safety
Battelle developed a health and safety plan related to producing lead-based paint and PEMs
coated with these paints. The plan was approved internally by appropriate environmental safety
and health personnel. Environmental monitoring during paint mixing and spraying activities
determined that lead exposure levels for workers were below Occupational Safety and Health
Administration standards. Some of the components of the safety plan included:
• All staff and any visitors were required to have documented hazard communication
training on lead.
• Baseline and post-work blood-lead levels were obtained for those Battelle staff that
conducted the paint mixing and spray painting.
• Respirators were used during leaded paint production
• Spray application operations staff were required to have a physical, appropriate training,
and to pass a respirator fit test.
• The interior of the spray booth was covered with plastic or other material that could be
easily removed and was then disposed of as hazardous waste.
• The area in front of the booth was set up as a change-out area where personal protective
equipment, such as coveralls, etc., could be removed without spreading lead outside of
the area.
• Warning signs restricting access were posted at the paint booth door.
Preparation of Linseed Oil Based Leaded Paints
To formulate historically accurate lead-based paints to apply to PEMs, Battelle consulted
Bennett's The Chemical Formulary - A Collection of Valuable, Timely, Practical Commercial
Formulae and Recipes for Making Thousands of Products in Many Fields of Industry, Volume
VI.l The Chemical Formulary had been printed with revisions every year until at least 1998.
Sample formulations from this reference are listed below in Table A-2. Since the paints
produced for the ETV verification of lead test kits were being applied to metal, drywall, plaster
A-7
-------
and wood, Battelle used a combination of formulations from Chapter Thirteen - Paint, Varnish,
Lacquer and Other Coatings to ensure adhesion to all substrates. Battelle reviewed the various
relevant historical formulations and developed formulations to apply to the PEMs that would
work best for application to the four substrates being used, i.e. would provide the best adhesion
to the variety of substrates required while achieving desired target lead levels.
Table A-2: Paint Formulations from The Chemical Formulary
Floor Painting and Finishing (p.
281) (for raw wood)
Soft Paste White Lead, 100 Ib.
Raw Linseed Oil, 3 gal.
Turpentine, 2 gal.
Liquid Drier, 1 pt.
Plaster, Primer (p. 332)
White Lead, Semi-Paste, 100 Ib.
Interior Varnish, 4 gal.
Linseed Oil, Kettle Bodied, 2 gal.
Turpentine, % gal.
Exterior House Paint Pigments
White (p. 328)
35% Leaded Oxide, 45 Ib.
White Lead, 181b.
Titanium Dioxide, 15 Ib.
Inert, 22 Ib. (Battelle used Zinc
Oxide)
In preparing the lead-based paints for the PEMs, Battelle used a combination of raw and boiled
linseed oil to ensure realistic drying time and good adhesion to the variety of substrates. A
variety of other formulas in the reference also mix these two resins.
A similar formulation was also found in Charles Uebele' s Paint Making and Color Grinding: A
Practical Treatise for Paint Manufacturers and Factory Managers2. The excerpt below explains
the difference in formulation requirements based on the substrate to which the paint will be
applied.
"CHAPTER XXV - DIPPING PAINTS.
Dipping Paints for Wood or Metal require to be made specially for either surface, as that
intended for wood will not always serve the purpose for metal. The paint for wood
requires to contain a pigment that acts as a filler, while tin or smooth sheet iron or steel
does not necessarily need it, in fact, it is best without it for certain metallic surfaces. The
function of a dipping paint is, first of all, to economize in labor, to cover uniformly
any article immersed in it, and to dip freely without leaving fringes of paint at the edges
and dry equally all over the surface thus coated.
Metal Preservative Red may be made by grinding a base of 40 pounds bright red oxide of
95 per cent, purity, 8 pounds red lead, 2 pounds zinc chr ornate, 25 pounds floated silex or
silica in 25 pounds raw linseed oil thinning same with 5 gallons raw linseed oil, 1 gallon
hard gum japan and Y% gallon turps. This will produce 12 gallons of paint weighing a
trifle over II pounds per gallon. By substituting a long stock of hard gum varnish for part
of the 5 gallons raw oil a hard drying product will be the result. "
In support of achieving consistent application of the lead-based paints in terms of film thickness
and lead level, Battelle investigated additions of various elements to mitigate settling and
improve application. Silicon dioxide was selected for this purpose because it was present in pre-
1978 leaded paints, is used for thickening and anti-settling properties in modern paint
formulations, and achieved the most consistent results. Battelle established the historical
A-8
-------
precedent for including silica in paints in a technical report submitted to EPA on February 19,
20093.
The primer and topcoats applied to the PEMs on top of the lead-based paints (or base paint for
the no-lead panels) all contain some form of Diatomaceous Silica, as well. The primer and three
topcoats applied are listed below.
• Sherwin Williams brand PreRite Bonding Primer
• Sherwin Williams Classic 99 Interior Satin Latex color Extra White
• Sherwin Williams Classic 99 Interior Satin Latex color 7047 Software (Grey)
• Sherwin Williams Accents Interior Satin Latex color 6867 Fireworks (red-orange)
Section 2 of the primer Material Safety Data Sheet (MSDS) specifies that the primer contains 9%
quartz. Quartz is referred to as "Crystalline Silica" in Section 11 of the MSDS. The MSDSs for
the three topcoats specify Cristobalite (CAS 14464-46-1) as an ingredient, which is a synonym
for silicon dioxide and also referred to as Crystalline Silica in Section 11. All panels have some
level of silica in the topcoat layers.
The paint formulations used for this effort were based on historical records. Primary ingredients
included zinc oxide, raw and boiled linseed oil, turpentine, Japan drier, either lead carbonate or
lead chromate, and titanium dioxide (used to balance the levels of lead). Nine different paint
formulations were produced as dictated by the two lead pigments (lead carbonate and lead
chromate) and the six different lead levels in addition to the zero lead level control. The
formulations were designed to consistently achieve the lead levels required when applied at
typical wet film builds.
The paint formulations are shown in Tables A-3 and A-4 below. Since the molecular
compositions of the two lead pigments are different, the formulations have accounted for these
differences by adjusting the load levels. However, the formulations for the 0% lead chromate
and carbonate were the same because no lead pigment was used in either.
A-9
-------
Table A-3. White Lead (Lead Carbonate) Paint Formulations
0% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7.0
7.0
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
OLead
14.79
0.00
6.16
1.48
0.15
2.16
0.05
24.8
0 Lead % by wt.
59.67%
0.00%
24.86%
5.97%
0.60%
8.70%
0.20%
100%
Gram wt
1491.75
0.00
621.56
149.18
14.92
217.55
5.04
2500
Sample reduced to 60% solids, 0% of TS-100 silica added then sprayed to thickness.
0.3% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
0.3 Lead
14.85
1.49
4.95
1.49
0.15
2.17
0.05
25.1
0.3 Lead % by wt.
59.08%
5.91%
19.69%
5.91%
0.59%
8.62%
0.20%
100%
Gram wt
1477.08
147.71
492.36
147.71
14.77
215.41
4.97
2500
Sample reduced to 60% solids, 0% of TS-100 silica added then sprayed to 3 milswet.
0.6% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
0.6 Lead
14.77
1.77
4.92
1.48
0.15
2.15
0.03
25.3
0.6 Lead % by wt.
58.45%
7.00%
19.47%
5.86%
0.59%
8.51%
0.12%
100%
Gram wt
1461.22
175.11
486.74
146.42
14.84
212.70
2.97
2500
Sample reduced to 70% solids, 0.7% of TS-100 silica added then drawdown with # 24 bar.
1.0% Lead Carbonate Paint Formulation
Materials
ZnO
PbC03
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
Thisformulation wi
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
1.0 Lead
14.40
3.00
4.80
1.44
0.14
2.10
0.05
25.9
1.0 Lead % by wt
55.53%
11.57%
18.51%
5.55%
0.56%
8.10%
0.19%
100%
Gram wt
832.88
173.52
277.63
83.29
8.33
121.46
2.89
1500
I be used to produce 0.6% and 1.4% lead levels at different coatina thickness.
1.4% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
1.4 Lead
13.22
4.21
4.81
1.44
0.14
2.1
0.03
26.0
1.4 Lead % by wt
50.94%
16.22%
18.54%
5.55%
0.54%
8.09%
0.12%
100%
Gram wt
764.16
243.35
278.03
83.24
8.09
121.39
1.73
1500
Sample reduced to 70 % solids, 1.5% of TS-100 silica added then drawdown with # 54 bar.
2.0% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
2.0 Lead
12.88
6.08
4.12
1.41
0.14
2.06
0.05
26.7
2.0 Lead % by wt
48.16%
22.73%
15.41%
5.28%
0.53%
7.70%
0.19%
100%
Gram wt
722.42
340.98
231.17
79.20
7.92
115.50
2.80
1500
Sample reduced to 65% solids, 1.5% of TS-100 silica added then drawdown with #40 bar.
6.0% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
6.0 Lead
4.70
18.10
1.57
1.43
0.14
2.09
0.05
28.1
6.0 Lead % by wt
16.73%
64.49%
5.58%
5.09%
0.51%
7.43%
0.18%
100%
Gram wt
250.89
967.34
83.63
76.40
7.64
111.42
2.67
1500
Sample reduced to 70% solids, 1% of TS-100 silica added then sprayed to thickness.
A-10
-------
0% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7.0
7.0
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
OLead
14.79
0.00
6.16
1.48
0.15
2.16
0.05
24.8
0 Lead % by wt.
59.67%
0.00%
24.86%
5.97%
0.60%
8.70%
0.20%
100%
Gram wt
1491.75
0.00
621.56
149.18
14.92
217.55
5.04
2500
Sample reduced to 60% solids, 0% of TS-100 silica added then sprayed to thickness.
0.3% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
0.3 Lead
14.85
1.49
4.95
1.49
0.15
2.17
0.05
25.1
0.3 Lead % by wt.
59.08%
5.91%
19.69%
5.91%
0.59%
8.62%
0.20%
100%
Gram wt
1477.08
147.71
492.36
147.71
14.77
215.41
4.97
2500
Sample reduced to 60% solids, 0% of TS-100 silica added then sprayed to 3 mils wet.
0.6% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
0.6 Lead
14.77
1.77
4.92
1.48
0.15
2.15
0.03
25.3
0.6 Lead % by wt.
58.45%
7.00%
19.47%
5.86%
0.59%
8.51%
0.12%
100%
Gram wt
1461.22
175.11
486.74
146.42
14.84
212.70
2.97
2500
Sample reduced to 70% solids, 0.7% of TS-100 silica added then drawdown with # 24 bar.
1 .0% Lead Carbonate Paint Formulation
Materials
ZnO
PbC03
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
Thisformulation wi
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
1.0 Lead
14.40
3.00
4.80
1.44
0.14
2.10
0.05
25.9
1.0 Lead % by wt
55.53%
11.57%
18.51%
5.55%
0.56%
8.10%
0.19%
100%
Gram wt
832.88
173.52
277.63
83.29
8.33
121.46
2.89
1500
be used to produce 0.6% and 1.4% lead levels at different coating thickness
1 .4% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
1.4 Lead
13.22
4.21
4.81
1.44
0.14
2.1
0.03
26.0
1.4 Lead % by wt
50.94%
16.22%
18.54%
5.55%
0.54%
8.09%
0.12%
100%
Gram wt
764.16
243.35
278.03
83.24
8.09
121.39
1.73
1500
Sample reduced to 70 % solids, 1.5% of TS-100 silica added then drawdown with # 54 bar.
2.0% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
2.0 Lead
12.88
6.08
4.12
1.41
0.14
2.06
0.05
26.7
2.0 Lead % by wt
48.16%
22.73%
15.41%
5.28%
0.53%
7.70%
0.19%
100%
Gram wt
722.42
340.98
231.17
79.20
7.92
115.50
2.80
1500
Sample reduced to 65% solids, 1.5% of TS-100 silica added then drawdown with #40 bar.
6.0% Lead Carbonate Paint Formulation
Materials
ZnO
PbCO3
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-XD13
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
6.0 Lead
4.70
18.10
1.57
1.43
0.14
2.09
0.05
28.1
6.0 Lead % by wt
16.73%
64.49%
5.58%
5.09%
0.51%
7.43%
0.18%
100%
Gram wt
250.89
967.34
83.63
76.40
7.64
111.42
2.67
1500
Sample reduced to 70% solids, 1% of TS-100 silica added then sprayed to thickness.
A-ll
-------
Tal
)le A-4. Yellow Lead (Lead Chromate) Paint Formulations
0.3% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Lot*
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
0.3 Lead
14.97
1.10
4.99
1.50
0.15
2.18
0.05
Total
24.9
0.3 Lead0/, bywt.
60.03%
4.40%
20.01%
6.00%
0.60%
8.75%
0.20%
100%
Gram wt
1500.74
110.05
500.25
150.07
15.01
218.86
5.01
2500
Sample reduced to 70 % solids, 0.7% of TS-100 silica added then drawdown with #34 bar.
0.6% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TIO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
0.6 Lead
14.65
2.15
4.88
1.47
0.15
2.14
0.03
25.5
0.6Lead% bywt.
57.52%
8.44%
19.16%
5.77%
0.59%
8.40%
0.12%
100%
Gram wt
1437.97
211.03
478.99
144.29
14.72
210.05
2.94
2500
Sample reduced to 70 % solids, 1.5% of TS-100 silica added then drawdown with #24 bar.
1.0% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TIO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot#
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
1.0 Lead
14.40
3.00
4.80
1.44
0.14
2.10
0.05
25.9
1.0 Lead % bywt
55.53%
11.57%
18.51%
5.55%
0.56%
8.10%
0.19%
100%
Gram wt
832.88
173.52
277.63
83.29
8.33
121.46
2.89
1500
This formulation will be used to produce 0.6% and 1.4% lead levels at different coating thickness.
Sample reduced to 70 % solids, 1% of TS-100 silica added then drawdown with #48 bar.
1.4% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TIO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
1.4 Lead
13.21
5.09
4.2
1.44
0.14
2.1
0.03
26.2
1.4 Lead % bywt
50.40%
19.42%
16.02%
5.49%
0.53%
8.01%
0.11%
100%
Gram wt
1260.02
485.50
400.61
137.35
13.35
200.31
2.86
2500
Sample reduced to 70 % solids, 1% of Aerosil 200 silica added then drawdown with #60 bar.
2.0% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TIO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Lot#
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
2.0 Lead
10.90
7.17
3.81
1.49
0.15
2.18
0.05
Total
25.8
2.0 Lead % bywt
42.32%
27.83%
14.81%
5.80%
0.58%
8.46%
0.19%
100%
Gram wt
1058.12
695.72
370.34
145.00
14.50
211.46
4.85
2500
Sample reduced to 70% solids, 1.5% of TS-100 silica added then drawdown with #42 bar.
6.0% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TIO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Lot#
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
6.0 Lead
1.65
21.43
0.55
1.51
0.15
2.20
0.05
Total
27.5
6.0 Lead % bywt
5.99%
77.84%
2.00%
5.47%
0.55%
7.98%
0.18%
100%
Gram wt
149.69
1946.00
49.90
136.75
13.68
199.43
4.54
2500
Sample reduced to 70% solids, 2% of TS-100 silica added then drawdown with #54 bar.
A-12
-------
0.3% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TiO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Lot*
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
0.3 Lead
14.97
1.10
4.99
1.50
0.15
2.18
0.05
Total
24.9
0.3 Lead % bywt.
60.03%
4.40%
20.01%
6.00%
0.60%
8.75%
0.20%
100%
Gram wt
1500.74
110.05
500.25
150.07
15.01
218.86
5.01
2500
Sample reduced to 70 % solids, 0.7% of TS-100 silica added then drawdown with #34 bar.
0.6% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TIO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
0.6 Lead
14.65
2.15
4.88
1.47
0.15
2.14
0.03
25.5
0.6 Lead % bywt.
57.52%
8.44%
19.16%
5.77%
0.59%
8.40%
0.12%
100%
Gram wt
1437.97
211.03
478.99
144.29
14.72
210.05
2.94
2500
Sample reduced to 70 % solids, 1.5% of TS-100 silica added then drawdown with *24 bar.
1.0% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TIO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
1.0 Lead
14.40
3.00
4.80
1.44
0.14
2.10
0.05
25.9
1.0 Lead % bywt
55.53%
11.57%
18.51%
5.55%
0.56%
8.10%
0.19%
100%
Gram wt
832.88
173.52
277.63
83.29
8.33
121.46
2.89
1500
This formulation will be used to produce 0.6% and 1.4% lead levels at different coating thickness.
Sample reduced to 70 % solids, 1% of TS-100 silica added then drawdown with #48 bar.
1.4% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TIO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Total
Lot*
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
1.4 Lead
13.21
5.09
4.2
1.44
0.14
2.1
0.03
26.2
1.4 Lead % bywt
50.40%
19.42%
16.02%
5.49%
0.53%
8.01%
0.11%
100%
Gram wt
1260.02
485.50
400.61
137.35
13.35
200.31
2.86
2500
Sample reduced to 70 % solids, 1% of Aerosil 200 silica added then drawdown with #60 bar.
2.0% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TIO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Lot*
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
2.0 Lead
10.90
7.17
3.81
1.49
0.15
2.18
0.05
Total
25.8
2.0 Lead % bywt
42.32%
27.83%
14.81%
5.80%
0.58%
8.46%
0.19%
100%
Gram wt
1058.12
695.72
370.34
145.00
14.50
211.46
4.85
2500
Sample reduced to 70% solids, 1.5% of TS-100 silica added then drawdown with *42 bar.
6.0% Lead Chromate Paint Formulation
Materials
ZnO
PbCrO4
TIO2
Linseed Oil
Boiled Linseed Oil
Turpentine
Japan Drier
GW
47.3
51
37
7.8
7.7
7
7
Lot*
ZC-X013
1401047-267
931407T.12
83734
83404
83304
PJD40
Supplier
The Carry Co.
American Elements
DuPont
Recochem Inc.
Recochem Inc.
Recochem Inc.
Barr
6.0 Lead
1.65
21.43
0.55
1.51
0.15
2.20
0.05
Total
27.5
6.0 Lead % bywt
5.99%
77.84%
2.00%
5.47%
0.55%
7.98%
0.18%
100%
Gram wt
149.69
1946.00
49.90
136.75
13.68
199.43
4.54
2500
Sample reduced to 70% solids, 2% of TS-100 silica added then drawdown with #54 bar.
A-13
-------
Paint Formulation Procedures
The paint samples were produced using standard painting production procedures in the Battelle
laboratories, including pre-mixing, media grinding of pigment and binder resin, and paint
letdown with resin and solvents. This procedure has been used for paint production both in the
laboratory and in commercial paint manufacturing for over 50 years. The equipment utilized in
this procedure includes the following:
• Variac that controls the speed of the dispersator
• High speed dispersator using a 5" diameter blade on the end of the mixing shaft
• Ice bath and ice
• Balance
• Paint cans
• Medium paint filters
• Red Devil paint shaker
Following are the detailed steps in the paint formulation procedure:
1. Add enough turpentine to cover mixing blade.
2. Start mixer at low speed.
3. Add zinc oxide slowly for 3-5 minutes, increasing mixing speed as needed to maintain
appropriate grind viscosity as visually evaluated by an operator skilled in the art.
4. Add turpentine as needed to keep the batch rolling.
5. Mix additional 10 minutes after addition of zinc oxide.
6. Add lead pigment slowly for 2-4 minutes, increasing mixing speed as needed.
7. Add turpentine as needed to keep the batch rolling.
8. Mix additional 10 minutes after addition of lead pigments.
9. Add titanium dioxide slowly for 3-5 minutes, increasing mixing speed as needed.
10. Add turpentine as needed to keep the batch rolling.
11. Mix for 60-90 minutes, or until batch viscosity decreases, determined by rolling action of
the batch.
12. Check Hegman, if < 5 continue to mix, and check Hegman every 10 minutes.4
13. When Hegman reaches ^ or > 5, start the let down, which includes adding all remaining
liquid raw materials after the pigment and extenders have been dispersed adequately.
14. Add boiled and raw linseed oil slowly and decrease mixing speed.
15. Add turpentine to wash out linseed oil container.
16. Mix additional 10 minutes after addition of linseed oils.
17. Add Japan drier drop wise to batch.
18. Mix additional 10 minutes after addition of Japan drier.
19. Tare quart cans.
20. Filter batch with medium paint filters into tared quart cans.
21. Note net weight and log book number of batch on quart cans.
22. Yields about ll/2 quarts of lead paint per batch.
23. Allow paint to set overnight.
24. Shake paint with Red Devil paint shaker for about 10 minutes take samples for % solids
check.
25. Check paint solids with moisture balance and record average of three test results on
formulation sheet.
A-14
-------
26. Store paint in aluminum cans in laboratory hood until future use.
Verification and Homogeneity Studies
Various batches of paint were prepared for the initial verification tests - one targeting each lead
level. Each paint was applied via drawdown or hand spraying to 3.5" x 5" metal panels attached
to a wooden rack. For each paint type and concentration batch, panels were coated to determine
proper film thickness, formulations, and drawdown bars to use, if applicable, to achieve each
desired lead level. Subsequently, homogeneity panels were coated to investigate ability to
achieve target lead levels and variability within and across panels. Verification and homogeneity
studies were performed on metal panels only due to ease and accuracy of sample extraction, i.e.,
it was easiest to obtain a 1 inch square sample from the metal surface which led to the most
accurate measurements of lead content in the sampled area, which was critical for verification
purposes.
After drying, paint chip samples were obtained from the metal panels following ASTM E1729.5
Laboratory analysis for lead by inductively coupled plasma-atomic emission spectrometry (ICP-
AES) was planned and conducted at an independent National Lead Laboratory Accreditation
Program (NLLAP)-accredited laboratory, Schneider Laboratories, Inc. ICP-AES testing was
conducted on three panels for each lead level with four samples obtained from each panel,
referred to as Homogeneity Panels since the primary purpose of the samples was to assess
consistency of lead levels across and within panels. The paint chips were digested using EPA
Method 3050B6 and the ICP-AES analysis was conducted following EPA Method 6010C7 as
well as the Schneider Laboratories, Inc. ICP SOP.8 The laboratory electronically reported lead
level measurements along with quality control (QC) sample results. Laboratory spike and
duplicate results as well as calibration verification sample results were supplied and reviewed for
each batch of samples analyzed. Acceptable recoveries for spike samples ranged from 80% to
120%. Acceptable recoveries for calibration verification samples were 90-110%. Acceptable
duplicate samples had a relative percent difference of 25% or less. Percent recoveries for
calibration verification samples ranged from 93-110%. Recoveries for QC spike samples ranged
from 92-115%. All duplicate samples had less than 25% relative percent difference. There were
no QC failures or problems.
Film thickness measurements were obtained by Battelle for each paint sample taken. Results of
the final batches of homogeneity samples for each set of PEMs are included in Table A-5.
Results were evaluated to determine correspondence to target lead levels and level of variability
as measured by the coefficient of variation (CoV), the standard deviation divided by the mean.
The production plan, agreed to in advance, specified a minimum acceptability of a CoV of less
than 15 percent. Following analysis, the results were forwarded to EPA with recommendations
regarding ability to proceed with production or the need for additional homogeneity testing. The
results shown in Table A-5 met the acceptability requirements and were thus deemed acceptable
for proceeding with the production of sets of PEMs at each lead level.
A-15
-------
Table A-5. Results from Final Homogeneity Testing on Metal Substrates for Each Set of
ETV PEMs
Lead
Type
White
Lead
Lead
Chromate
Target Lead
Level
0.3
0.6
1.0
1.4
2.0
6.0
0.3
0.6
1.0
1.4
2.0
6.0
Mean Levels
ICP (mg/cm2)
0.30
0.65
0.99
1.56
1.85
5.97
0.30
0.62
1.07
1.42
1.92
6.88
FT (mils)
0.79
0.95
1.26
1.72
1.48
1.94
1.16
0.98
1.50
1.89
1.38
1.81
CoV*
ICP
13.3
7.1
3.9
7.2
5.6
14.2
9.6
4.1
11.0
4.1
10.1
5.2
FT*
6.1
5.7
3.4
3.5
7.0
8.3
4.0
9.1
7.4
6.8
2.4
3.3
* Coefficient of Variation (Standard Deviation/Mean x 100)
** Film thickness
Production Application of Lead Paint Coatings
Based on the results from the Verification and Homogeneity Study summarized in Table A-5,
production proceeded using the paint formulation and application method (spray or a particular
size drawdown bar) that achieved the target lead levels. During production application,
reference panels were coated along with the production panels at a rate of 18 for each set of 468
panels. For sets of PEMs that were sprayed, reference panels were placed at previously-
assigned, randomly selected locations on the racks containing all the PEMs awaiting paint
application. For sets of PEMs that had the lead paint applied via drawdown bar, production
panels were drawdown in sets of two to three panels each for the wood, metal and drywall
substrates and one at a time for the plaster substrates. At the discretion of the operator, a
reference panel was prepared approximately every 10 sets or 25 panels.
Metal panels were used as the reference panels since metal panels yield the most accurate
measurements of film thickness and lead levels. The reference PEMs were tested for film
thickness during application and for lead level by ICP analysis after the paint had dried. This test
procedure was used to check that the application process resulted in appropriate lead levels.
Despite the use of the metal substrate only for the reference panels, the lead levels and paint
thickness on these reference panels served as representative of the coatings applied to all wood,
drywall, plaster, and metal substrate panels.
Table A-6 presents the average lead levels, CoV, minimum, and maximum of each set of 18
reference panel measurements. Most sets are very close to target lead levels, such as the 2.03
mg/cm2 average for the 2.0 mg/cm2 target yellow lead set, the 0.32 mg/cm2 for the 0.3 mg/cm2
target yellow lead set, and the 0.64 mg/cm2 average for the 0.6 mg/cm2 target white lead set.
There also were a few sets that were a bit off target, but were sufficient to meet the verification
A-16
-------
needs. Despite the high average lead level of 9.2 mg/cm2, the 6.0 mg/cm2 white lead PEMs were
accepted by EPA because they still met the needs of the verification for a set of PEMs at a high
lead level. In the 0.6 mg/cm2 yellow lead batch, the measured lead levels of 17 of the 18
reference panels ranged from 0.51 to 0.66 mg/cm2, yielding a mean of 0.55 mg/cm2, and a CoV
of 7.5%. Because only one reference panel of 18 yielded a high lead level, the set of panels was
accepted.
Table A-6. Reference Panel Results from Final Production for Each Set of ETV PEMs
Lead Type
No Lead
White
Lead
(Lead
Carbonate)
Yellow
Lead
(Lead
Chromate)
Target Lead
Level
0.0
0.3
0.6
1.0
1.4
2.0
6.0
0.3
0.6
1.0
1.4
2.0
6.0
Lead Levels
Mean (mg/cm2)
0.00
0.40
0.64
1.00
1.48
2.29
9.18
0.32
0.57
1.00
1.39
2.03
5.15
CoV
8.2
17.8
13.5
5.1
8.0
5.6
31.2
13.1
16.6
7.1
12.0
9.4
9.6
Range
Min
0.002
0.234
0.425
0.918
1.322
2.018
5.65
0.252
0.511
0.879
1.194
1.483
3.929
Max
0.003
0.505
0.761
1.095
1.748
2.525
18.4
0.428
0.920*
1.148
1.601
2.314
6.247
* Next highest measurement was 0.659
Topcoating
The linseed oil based paints were applied to the PEMs and stored in the constant temperature and
humidity rooms during a four to seven day drying time. The panels were then all topcoated with
Sherwin Williams brand Prep Rite bonding Primer to ensure good adhesion between the linseed
oil based paint and the latex emulsion topcoat paints. The final latex paint topcoat was then
applied to the PEMs. The topcoat paints are described in more detail below:
• Primer - Sherwin Williams Prep Rite bonding primer, diluted with deionized (DI) water
at a ratio of 3:1 parts by volume. Spray application was done with a 50 percent overlap on
the PEMs in both horizontal and vertical directions with a total wet film build of
approximately 4-5 mils (a measure of dry film thickness). The PEMs then were allowed
1-2 hours to air dry before top coats were applied.
• Top coat number 1 is Sherwin Williams Classic 99 interior satin latex; color Extra White,
diluted with DI water at a ratio of 3:1 parts by volume. Spray application was done with a
50 percent overlap on the PEMs in both horizontal and vertical directions, with a total
wet film build of approximately 4-6 mils. Then the PEMs were allowed to air dry for
three days. The PEMs were then bagged for further testing.
• Top coat number 2 is Sherwin Williams Classic 99 interior satin latex; color 7047
(software gray), diluted with DI water at a ratio of 3:1 parts by volume. Spray application
was done with a 50 percent overlap on the PEMs in both horizontal and vertical
A-17
-------
directions for a total wet film build of approximately 4-6 mils. Then the PEMs were
allowed to air dry for three days. The PEMs were then bagged for further testing.
• Top coat number 3 is Sherwin Williams Color Accents interior satin latex; color 6867
(Fireworks orange red), diluted with DI water at a ratio of 3:1.5 parts by volume. Spray
application was done with a 50 percent overlap on the samples in both horizontal and
vertical directions for a total wet film build of approximately 4-6 mils. The PEMs were
then allowed to air dry for three days. The PEM samples were then bagged for further
testing.
PEM Labeling, Packing and Storage
The PEMs were stored in the constant temperature and humidity conditioning rooms prior to
being packed up for transfer to the evaluation location. Each PEM was labeled on the back with
an individual identification number, wrapped in a single laboratory towel to protect the front
surface, and placed inside an individual zip seal bag also labeled with the identification number.
References
1. Bennett, H. The Chemical Formulary - A Collection of Valuable, Timely, Practical
Commercial Formulae and Recipes for Making Thousands of Products in Many Fields of
Industry, Volume VI. 1943. Chemical Publishing Co., Inc. Copyright 1943.
2. Uebele, Charles. Paint Making and Color Grinding: A Practical Treatise for Paint
Manufacturers and Factory Managersa The Trade Papers Publishing Co., Ltd., 1913.
3. Battelle. Addition of Silica to Lead-based Paints Used for Production of PEMs in
Support of ETV Evaluation of Lead Test Kits: References. Technical report submitted to
EPA on February 19, 2009.
4. ASTM D1210-05(2010), "Standard Test Method for Fineness of Dispersion of Pigment-
Vehicle Systems by Hegman-Type Gage," ASTM International.
5. ASTM El 729, "Standard Practice for Field Collection of Dried Paint Samples for
Subsequent Lead Determination," ASTM International.
6. United States Environmental Protection Agency, "Method 3050B: Acid Digestion of
Sediments, Sludges, and Soils", SW846 Online, Revision 2. December 1996.
7. United States Environmental Protection Agency, "Method 6010C: Inductively Coupled
Plasma-Atomic Emission Spectrometry", SW846 Online, Revision 3. February 2007.
8. Schneider Laboratories, Inc. ICP SOP Document# III-017-08-002
A-18
-------
Section A2
Comparison of Expected vs. Actual Lead Concentrations
of Performance Evaluation Materials
A-19
-------
The following tables present a comparison of the expected vs. confirmed lead concentration for
each PEM used during the testing of the lead test kits. Expected concentrations are based on lead
levels defined for sets of PEMs during the PEM production process. That is, PEMs were being
made at expected lead concentrations of 0, 0.3, 0.6, 1.0, 1.4, 2.0, or 6.0 mg/cm2. These are the
expected lead levels as defined in the test/quality assurance (QA) plan. Confirmed
concentrations are based on ICP-AES results from individual paint chip samples taken from each
PEM during testing (see Section 3.3. lin the test/QA plan).
Table A-7 presents the results by substrate and across all PEMs. Table A-8 presents the results
by lead type. The average and standard deviation for the confirmed lead levels, as well as the
CoV, are presented for each expected concentration level.
Table A-7. Confirmed lead level statistics for PEMs compared to expected lead level
concentrations by substrate type.
Substrate
Drywall
Metal
Plaster
Wood
All
N
Confirmed Lead
Confirmed Lead
CoV (%)
N
Confirmed Lead
Confirmed Lead
CoV (%)
N
Confirmed Lead
Confirmed Lead
CoV (%)
N
Confirmed Lead
Confirmed Lead
CoV (%)
N
Confirmed Lead
Confirmed Lead
CoV (%)
Level:
Level:
Level:
Level:
Level:
Level:
Level:
Level:
Level:
Level:
Average
StdDev
Average
StdDev
Average
StdDev
Average
StdDev
Average
StdDev
Expected PEM
0
144
0.00
0.00
143.16
144
0.00
0.01
368.04
140
0.00
0.01
258.82
144
0.00
0.02
470.54
572
0.00
0.01
451.36
0.3
288
0.34
0.07
20.59
288
0.31
0.07
22.91
290
0.44
0.18
40.17
288
0.32
0.16
48.20
1154
0.35
0.14
38.91
0.6
282
0.83
0.22
26.01
288
0.56
0.14
24.39
292
1.25
0.53
42.40
288
0.72
0.22
30.83
1150
0.84
0.41
48.34
Lead Level (mg/cm
1
290
1.15
0.21
17.95
288
0.85
0.10
11.31
296
1.65
0.84
50.92
275
1.07
0.31
28.80
1149
1.18
0.55
46.76
1.4
296
1.48
0.29
19.79
288
1.26
0.19
14.91
288
1.79
0.65
36.22
282
1.45
0.29
20.29
1154
1.50
0.44
29.38
2)
2
288
2.52
0.33
13.22
288
1.91
0.28
14.49
288
2.91
0.85
29.24
284
2.39
0.71
29.56
1148
2.43
0.69
28.49
6
292
9.04
2.32
25.69
286
8.18
1.86
22.76
284
10.11
3.01
29.81
288
8.71
1.59
18.29
1150
9.01
2.36
26.24
CoV = Coefficient of Variation (Standard Deviation/Mean x 100)
Table A-7 indicates that overall confirmed lead levels were similar to expected concentrations.
However, there are substrate types for which, comparatively, the confirmed lead levels were
higher than the expected levels. Average confirmed levels for drywall and plaster PEMs were
higher than expected levels, especially when compared to average confirmed lead levels from
metal and wood. The PEMs used in the verification test were produced mainly using a
drawdown technique (for all panels except no lead, 0.3 mg/cm2 and 0.6 mg/cm2 white lead).
This involved applying the paint to the PEM and pulling it down with a specially designed bar.
Being porous substrates, it is possible that the plaster and drywall panels absorbed some of the
A-20
-------
paint, causing more paint to be applied to the PEM to accommodate the thickness required on the
PEM. This would then lead to higher lead concentrations on these substrates. The most
significant potential impact of this effect can be seen on the plaster PEMs. This potential effect
is based on observations during the production of the PEMs but has not been studied or
confirmed.
Table A-8. Confirmed lead level statistics for PEMs compared to expected lead level
concentrations by lead paint type.
Lead Type
None
White
Yellow
All
N
Confirmed Lead
Confirmed Lead
CoV (%)
N
Confirmed Lead
Confirmed Lead
CoV (%)
N
Confirmed Lead
Confirmed Lead
CoV (%)
N
Confirmed Lead
Confirmed Lead
CoV (%)
Level:
Level:
Level:
Level:
Level:
Level:
Level:
Level:
Average
StdDev
Average
StdDev
Average
StdDev
Average
StdDev
Expected PEM
0
572
0.00
0.01
451.36
572
0.00
0.01
451.36
0.3
576
0.30
0.08
25.56
578
0.40
0.16
40.64
1154
0.35
0.14
38.91
0.6
574
0.88
0.41
46.53
576
0.80
0.40
49.95
1150
0.84
0.41
48.34
Lead Level
1
573
1.24
0.72
58.18
576
1.13
0.30
26.32
1149
1.18
0.55
46.76
(mg/cm
1.4
578
1.53
0.52
34.08
576
1.46
0.33
22.87
1154
1.50
0.44
29.38
2)
2
576
2.36
0.58
24.60
572
2.51
0.78
31.25
1148
2.43
0.69
28.49
6
572
8.37
2.05
24.47
578
9.64
2.48
25.75
1150
9.01
2.36
26.24
CoV = Coefficient of Variation (Standard Deviation/Mean x 100)
The results in Table A-8 show that there was no significant difference in confirmed lead levels
between white and yellow lead PEMs. The CoVs values were all <50% at all levels except 0.0
mg/cm2. The larger CoV at this level is reflective of small changes around the zero lead level
and most likely represent ICP-AES measurement variability near the detection limit, since no
lead was used in preparing these PEMs. It should be noted, as discussed in Section Al of this
appendix, that the PEMs prepared at the expected lead level of 6.0 mg/cm2 were known to be on
average higher than 6.0 mg/cm2 and that it was purposefully decided to accept the variation
present at this expected lead level.
Though there were some differences between the confirmed and expected lead levels, it should
be noted that when evaluated for proper responses, test kit results were compared to confirmed
lead levels. That is, test kit results were always compared to the actual PEM lead levels, not the
expected.
A-21
-------
Section A3
QA/QC Results for the ICP-AES Analysis of
Performance Evaluation Materials
A-22
-------
Summary of Lead Level Confirmation ICP-AES Analysis of PEMs
All paint chip samples from the PEMs used in this verification test were analyzed using ICP-
AES by Schneider Laboratories, Inc.
Sample preparation procedures followed the SOP generated by Schneider Laboratories, Inc. for
this study (Schneider Laboratories, Inc., SOP Battelle Paint Samples, Doc # 111-044-10-001).
Information on how QC samples were spiked and final concentrations is provided in the SOP.
Three versions of this SOP (the original and two revisions) were used dated 1/20/10, 2/24/10,
and 4/25/10. Approximately 27% of the PEMs were analyzed prior to the 2/24/10 revision to the
SOP. In the 2/24/10 version, revisions were made to clarify that post-digestion matrix spikes and
duplicates were being evaluated. Additionally, the laboratory control spike (LCS) procedures
changed such that a separate LCS and a QC check sample were now being performed.
Originally, in the 1/20/10 version, the LCS was prepared by spiking the QC check sample, which
was a certified reference material (CRM) (as stated in Section 6.11.2 of the 1/20/10 SOP)
containing a known quantity of lead. This practice was changed because there were recovery
issues. The spike concentration of 1000 micrograms per liter (ng/L) was not >3x the background
lead concentration because of the high lead concentrations in the actual CRM samples (4630
milligrams per kilogram [mg/kg]). Thus, as of the 2/24/10 SOP, one LCS, one QC check
sample, and one QC check sample duplicate were being evaluated for every 20 samples. The
LCS (Blank Paint QC) sample in the 2/24/10 SOP was defined as a piece of non-lead containing
paint that was spiked with lead solution to a resulting concentration of 1000 [ig/L. The QC check
sample in the 2/24/10 SOP contained 10 mg of the CRM, a known lead-containing material. The
QC check (CRM) was purchased to contain 4630 ± 266 mg/kg lead. To prepare the sample, 10
mg of the CRM was weighed out and diluted to 10 mL, resulting in a final concentration of 4.630
mg/L.
The 4/25/10 revision of the SOP clarified the acceptance criteria for the LCS samples, as it did
not appear to be clearly defined in previous versions.
Because of the high lead concentration in the PEM samples, dilutions were made to the samples
prior to initial analysis. The dilutions were prepared by spiking 10 microliters (|jL) of the
original digested sample into 9.990 milliliters (mL) of reagent water for a 1:1000 dilution. The
samples were thoroughly mixed by inverting, and then analyzed for lead content. If the result
was below the reporting limit, the sample was reanalyzed either non-diluted or at a lower dilution
level. If samples were rerun at a different dilution level, this was noted in the QC summary
report for that particular sample set.
The MDL for lead was 2.91 |ig/L.
The reporting limit was 40 |ig/L. Therefore all blank results should be <40 |ig/L.
A-23
-------
Summary of Quality Control Measures for PEMs ICP-AES Analysis
QC procedures were performed in accordance with the quality management plan (QMP) for the
Battelle ETV Advanced Monitoring Systems (AMS) Center, except where differences were
noted for Environmental and Sustainable Technology Evaluations (ESTE) per the EPA ETV
Program QMP, and the test/QA plan for this verification test. Test procedures were conducted as
stated in the test/QA plan; however a deviation to the test/QA plan was made during the ICP-
AES analyses. For some sample runs, continuous calibration verification (CCV) samples were
run once every 20 instead of 10 samples. This deviation is further described below. This change
was assessed to have no impact on the quality of the results as described below. QC results for
the analysis of paint chip samples from the PEMs are described below.
ICP-AES Blank Sample Results
Various blank samples were analyzed for the ICP-AES analyses. Method blank samples were
analyzed in each set of 10-20 paint samples to ensure that no sources of contamination were
present. An initial calibration blank was analyzed at the beginning of each run and used for
initial calibration and zeroing the instrument. A continuing calibration blank was analyzed after
each CCV to verify blank response and freedom from carryover. No blank samples failed QC
during the analyses.
Calibration Verification Standards
Initial calibration standards were run at the beginning of each set of analyses. The acceptance
criterion for the calibration coefficient of the calibration standards was > 0.998. If this criterion
was not met, the analysis was stopped and recalibration was performed before samples were
analyzed. A 500 parts per billion (ppb) CCV standard was analyzed at the beginning of each run
(following the initial calibration), at the end of each run, and every 10-20 samples. CCV
recoveries ranged from 96% to 108%. Per the test/QA plan, CCV sample frequency was once
every 10 samples. For most of the sample sets, CCVs were performed with this frequency.
However, for later sample sets, CCVs were run once every 20 samples. CCV samples are used
to verify instrument performance and are evaluated usually at a specified frequency as a
preventative measure so that large amounts of samples do not need to be re-run if a CCV sample
fails. In the course of this study, only one CCV sample failed, and it was when the CCV was
being run once every 10 samples. All samples from the last passing CCV of that sample set were
re-analyzed.
QC samples also included an initial calibration verification (ICV) standard and interference
check sample (ICS). Both samples were 500 ppb. ICV samples were analyzed once at the
beginning of each sample run and were required to have percent recoveries between 90-110% to
be acceptable. ICS samples were analyzed at the beginning and end of every run and every 10-
20 samples. ICS samples had to have percent recoveries between 80-120% to be acceptable. All
reported ICV and ICS samples met the acceptance criteria. Recoveries for ICV samples ranged
from 96% to 108%. Recoveries for ICS samples ranged from 93% to 112%.
A-24
-------
Matrix Spike Samples/Duplicates
Matrix spike samples, as well as duplicates of these samples, were analyzed once every 10-20
samples. Acceptable recoveries for matrix spike samples were between 80-120%. Duplicate
samples had acceptance criteria of ±25% relative percent difference (RPD).
All matrix spike samples were performed as post-digestion spikes as there was insufficient
sample volume to perform a pre-digestion spike. Matrix spike recoveries ranged from 86% to
207%. Six matrix spike samples failed with recoveries above the specified acceptance criteria.
In these instances, the lead concentration in the sample was well above the spike level. Matrix
spike results indicate that matrix interferences were not observed. Duplicate samples were
within the specified RPD.
LCS Samples
LCS samples were analyzed once every 10-20 samples. Acceptable recoveries for LCS samples
were between 80-120%. LCS recoveries ranged from 17% to 225%. Schneider Laboratories,
Inc. noted that LCS failures on one sample set were attributed to improper spiking technique.
Training on spiking procedures was immediately implemented by Schneider Laboratories, Inc.
for all analysts spiking samples. All LCS failures occurred prior to a revision to the Schneider
Laboratories, Inc. SOP for analyzing paint samples for this verification test. In the original
version of the SOP, LCS samples were prepared by spiking a known amount of lead onto a
CRM. This practice was changed on 2/24/10 because there were recovery issues. The spike was
not >3x the background lead concentration because of the high lead concentrations in the actual
CRM samples. In the revised SOP, the LCS was prepared by spiking a piece of lead-free latex
paint. There were no LCS failures after that In addition, a QC check sample containing only
the CRM, which has a known concentration of lead weighed out to a particular amount, was
analyzed with each sample set throughout the verification test. These QC samples all passed
acceptance criteria.
A-25
-------
Appendix B
Vendor Comments
-------
Silver Lake Research submitted the following comments on the draft report. These
comments have not been reviewed by Battelle or U.S. EPA for accuracy, and do not
necessarily reflect the opinions or views of U.S. EPA. Any questions regarding the
comments in this section should be addressed to the vendor. EPA's response to the
comments below is on file.
Silver Lake Research Corporation disputes the data analysis and the conclusions of the ETV
ESTE Verification Report of the LeadAVERT™ Test Kit for Lead-Based Paint. The data
analysis and the conclusions of the report are contrary to both the intent and the guidelines of the
Lead Renovation, Repair, and Painting Rule, and are also contrary to the provisions of the
Test/QA Plan for this verification. In fact, the verification and data analysis were performed is
such a way that no test kit could possibly achieve the Rule's performance specifications in this
verification.
Specifically, the following deficiencies are evident in the data analysis and conclusions of the
Verification Report:
1. The data analysis does not adequately consider the inherent variability in lead content in
different spots of the PEM panels. The results of the LeadAVERT Test Kits were evaluated as
true or false (negative or positive) as compared to a single ICP-AES measurement performed on
another spot of the same PEM panel. Clearly, if there is a significant difference between the lead
levels at the spot used for the LeadAVERT measurement and the spot used for the ICP-AES
measurement, the results of the two measurements would be expected to differ - especially if the
PEM panel was prepared to be near the 1.0 mg/cm2 level.
The only measurement of spot-to-spot variability in the PEM lead levels was done as described
in Appendix A (p. A-15). This homogeneity analysis was performed on samples that were
produced only for that purpose and were substantially different than the actual PEMs used in the
verification - 1) they were only on metal substrate (this was by far the most consistent substrate
of the five used in the verification PEMs), 2) they were produced with a drawdown technique
that was not used on all of the actual verification PEMs, and 3) the production of actual PEMs
was a completely separate process, and no quality checks or measurements were done to ensure
that homogeneity of the actual PEMs was the same as that of the preliminary samples.
Furthermore, the homogeneity analysis was performed by testing 1 in2 areas of paint, averaging
out the spot-to-spot variability over a relatively large area - a much greater area than that used by
the LeadAVERT test (.024 in2). This homogeneity testing is in no way indicative of the true
spot-to-spot variability in the actual PEMs used in the verification.
Actual verification PEMs are very likely to be non-homogeneous to a much greater extent than
the metal-substrate "homogeneity study" samples. No homogeneity testing was performed on
actual verification PEMs, but panel-to-panel variation was reported (Table A-7). This data shows
extremely high coefficients of variation (at 1.0 mg/cm2 , a minimum of 11.31% for metal
substrate and >50% for plaster substrate!). It is simply not possible that panel-to-panel variability
would be so high that a standard deviation would be half the mean, yet inside each panel the
distribution of lead was completely uniform. Yet the results of a LeadAVERT test were
compared to a single measurement of another spot on the PEM as if the entire panel was
B-2
-------
uniformly the same lead level as that single measurement. As a result, the LeadAVERT results
appear inaccurate, when in fact it is likely that the ICP-AES measurement is correct at its
location on the PEM and the result of the LeadAVERT is also correct at its location.
2. It must be noted that the ICP-AES method has its own measurement error inherent in the
technique, measured by recovery rates of matrix spiked samples and calibrations. Appendix A
states that acceptable variation in matrix spiked samples for ICP-AES was ±25%, and the
acceptable recovery range for LCS spiked samples was 80-120%. This is another source of
significant variation that is not directly addressed by the data analysis.
3. These major deficiencies in the data analysis were noted by SLRC in its comments on the
Draft Verification Report, and the revised version of the Report attempted to use a simulation
and extrapolation program to address this problem. However, since there is only one ICP-AES
measurement for each PEM panel, it is not possible to provide a reasonable model of the true
spot-to-spot variation on that panel. SIMEX may be used in some situations to address
measurement error (that is, the error inherent in the measurement procedure) - but not actual
variation in lead concentration, which is independent of measurement. In the SIMEX analysis
performed in Section 5.4 of the Verification Report, the few pre-production samples produced in
"homogeneity testing" serve as the surrogate data point for modeling the variability across all
PEMs. This is invalid modeling, since the "homogeneity" samples were on metal only (the least
inconsistent substrate for actual PEMs in Table A-7), were produced outside of the actual
PEM production process, and the assessed variability was an average of a sample area more than
40X larger than the LeadAVERT sample area.
4. The Lead RRP Rule specifically defines 0.8 and 1.2 mg/cm2 as the single-point levels at which
false positive and false negative rates must be assessed, respectively. Clearly, to be able to
determine that false positive rates are below 10% and false negative rates below 5%, with >95%
confidence level, the determination of the true lead level at the point of measurement must be
precise and accurate with a much lower error than the RRP Rule criteria. This is simply not the
case in this verification. Homogeneity of actual PEMs was not directly measured, and it can be
presumed that PEMs are highly variable because of the observed panel-to-panel variability.
Allowed recoveries for the ICP-AES measurement itself are outside of the error rates of the Rule.
Attempted SIMEX to correct these deficiencies is invalid, as it is based on data that is not
representative of the actual PEMs. This data analysis cannot possibly approve any test kit,
since the error that is assigned to the test kit is actually inherent in the variability of the
PEMs and their measurement - and that variability is higher than the allowed error rate
for test kits.
5. Statistical modeling programs were used to predict the performance of LeadAVERT with
respect to the Lead RRP Rule criteria. These models are not appropriately predicting the actual
performance of the LeadAVERT test kit. As an example, Figure 6-5 of the Verification Report
graphically shows the predicted performance of the LeadAVERT test Kit on white lead paint
samples when performed by a technical operator. On all types of substrates, the predicted
"probability of positive result" appears almost independent of lead levels, far different from the
"acceptable" test kit performance shown in Figure 6-2. The predicted rate of LeadAVERT
positive results with zero lead is >60% for all substrate types (Figure 6.5). The actual rate of
B-3
-------
positive results with zero lead is 4% (8% for the technical operator, 0% for the non-technical
operator). The difference between predicted and actual rate is huge, and it is indicative of an
inappropriate use of a model that does not fit the performance.
The determination of LeadAVERT's performance with respect to the Lead RRP Rule criteria is
based on modeling, not directly on actual results at the criteria points. The use of an
inappropriate model results in inappropriate prediction, and dooms the test in this evaluation.
It is noted that the Test/QA Plan specified what type of regression models were to be used in the
data analysis, based on previous studies of other test kits, all of which were based on color-
producing chemical reactions. It should also be noted that LeadAVERT is an immunoassay, not a
chemical test kit, and its results (the relationship of signal to lead concentration) do not follow
the same type of linear kinetics that is assumed in the regressions used in the data analysis. The
performance of LeadAVERT across a range of variables had not been tested at the time the
Test/QA Plan was finalized. Other deviations from the Test/QA Plan were made during the
verification and data analysis (e.g., "binning" PEMs of different target lead levels into a single
analysis group based on a single ICP-AES measurement - never considered nor approved in the
Test/QA Plan). The use of an inappropriate statistical model to determine passing or failing with
respect to Lead RRP Rule criteria runs counter to the intentions of the Rule.
6. The LeadAVERT Test Kit was not designed to detect "yellow" lead, lead chromate.
This pigment is not commonly found on indoor residential painted surfaces, with the possible
exception of metal substrates and bright yellow to red paint colors. Users can easily identify
sample types that could contain yellow lead and avoid using LeadAVERT on those samples.
Using LeadAVERT on yellow lead samples can produce false negative results. SLRC requested
that test results on yellow lead samples be reported separately from results on white lead
samples, since reporting "total" results would be an unfair evaluation of test kit performance.
SLRC believes that ETV's insistence on reporting only "total" results and, separately, "yellow
lead" results in the Verification Statement - while not reporting "white lead" results separately -
is an additionally unfair statement that portrays the LeadAVERT Test in the most negative light.
As most potential users will not read the full report, only the summary statement, this type of
selective analysis and data reporting is not warranted.
The intent of the test kit-related sections of Lead RRP Rule was to provide renovators with an
accurate and inexpensive on-site method of determining whether a location has lead-based paint.
The verification Test/QA Plan had the intent of providing an independent and fair evaluation of
test kits with respect to the criteria of the Lead RRP Rule. It is to be expected that unforeseen
variables will come up in the course of a long and laborious verification study - and these
difficulties must be addressed properly in the data analysis in order to obtain valid conclusions to
the verification process. In this verification, the unforeseen variable was the unexpected
variability of the PEMs. However, nothing in the data analysis adequately addressed this
unforeseen variable - the data analysis was conducted as if there is no spot-to-spot variability in
PEMs and no error rate to the ICP-AES measurement. SIMEX modeling based on assumptions
from a different process is not a valid patch for the problem. Furthermore, key conclusions with
regard to the performance criteria of the Lead RRP Rule were based on inappropriate predictive
B-4
-------
models that were wildly inaccurate in predicting the actual data obtained in the verification itself
- far beyond what one might expect in any valid regression analysis. The conclusions of the data
analysis are therefore invalid.
Silver Lake Research Corporation
October 29, 2010
B-5
------- |