-------
                                           EPA 747-R-95-002a
                                                    May  1995
A FIELD TEST OF LEAD-BASED  PAINT  TESTING TECHNOLOGIES:

                 SUMMARY REPORT
      U.S. EPA Region II Library
        790 Broadway
               Technical Programs Branch
             Chemical Management Division
       Office  of  Pollution Prevention and Toxics
Office of Prevention, Pesticides, and Toxic  Substances
         U.S.  Environmental Protection Agency
                  401 M Street, S.W.
                Washington, D.C. 20460
                     Hecycled/Recyclable
                     Printed with Soy/Canola Ink on paper that
                     contains at least 50% recycled fiber

-------
The material in this document has been subject to Agency technical
and policy review and approved  for  publication  as  an EPA report.
Mention of trade names,  products, or services does not convey, and
should not  be interpreted as  conveying,  official  EPA approval,
endorsement, or recommendation.

             This report is copied on recycled paper.

-------
                   CONTRIBUTING ORGANIZATIONS

     The study  described in this  report  was funded by  the U.S.
Environmental Protection Agency and the U.S. Department of Housing
and  Urban  Development.    The  study  was  managed by  the  U.S.
Environmental  Protection  Agency.    The  study  was   conducted
collaboratively  by  two  organizations  under  contract  to  the
Environmental Protection  Agency,  Midwest Research Institute and
QuanTech.  Each organization's responsibilities are listed below.


                   Midwest Research Institute

     Midwest  Research   Institute   (MRI)    was   responsible  for
initiating the pilot study on schedule,  for overall production of
the Quality Assurance Project Plan for both  the pilot and the full
study, for providing  input to the design of the study, for planning
and supervising the field work, for collecting paint samples, for
the laboratory  analysis of paint  chip  samples,  and for writing
sections of the technical report.


                             QuanTech

     QuanTech (formerly David C. Cox & Associates)  was responsible
for the  design of  the  study  and  contributions  to the  Quality
Assurance  Project Plan  for  the  pilot  and full studies,  for
participation in  field  work,  for  data  management and statistical
analysis, and for overall production of the technical and summary
reports.


              U.S. Environmental  Protection Agency

     The U.S. Environmental Protection Agency (EPA) co-funded the
study and  was  responsible for managing the  study, for reviewing
study documents, and  for arranging for the peer review of the final
report.   The  EPA  Project Leader was John Schwemberger.   The EPA
Work Assignment Managers were John Scalera and John Schwemberger.
The EPA Project  Officers were Jill  Hacker, Samuel Brown, and Janet
Remmers.    Cindy  Stroup was  the  Branch  Chief  of  the  Technical
Programs Branch and initiated this study.


            U.S. Department of  Housing and Development

     The  Department  of  Housing  and  Urban  Development   (HUD)
co-funded  the  study and  identified sources  of  housing  for the
study.  Bill Wisner was the key HUD staff member.

-------
                        ACKNOWLEDGEMENTS

     The study could not have been done without the assistance and
cooperation of  the Housing Authority  of Louisville,  the Denver
Housing Authority,  and the Philadelphia Housing Authority.

     Special thanks are due to Mike Godfrey and George Adams of the
Housing Authority of Louisville, Mark Ward  and  Ben Roybal of the
Denver Housing Authority,  and John Peduto, Cynthia Jones, and Bill
Zollicoffer of the Philadelphia Housing Authority.

-------
                        TABLE OF CONTENTS
EXECUTIVE SUMMARY 	  xiii
     BACKGROUND	xiii
     TECHNOLOGIES EVALUATED 	  xiii
     STUDY OBJECTIVES	xiv
     FIELD TESTING	xiv
     STUDY RESULTS	xvi
          Laboratory Analysis Results 	   xvi
          Chemical Test Kit Results	   xvi
          XRF Results	xvii
     OVERALL RECOMMENDATIONS FOR TESTING  	 xviii
          XRF Instrument Conclusions  	   xviii
          Chemical Test Kit Conclusions	   xix

1  DESCRIPTION OF THE STUDY 	   1
     1.1  BACKGROUND	   1
     1.2  STUDY OBJECTIVES  	   2
     1.3  APPROACH	   3
     1.4  TECHNOLOGIES  	   4
     1.5  FIELD TESTING 	   6
     1.6  PEER REVIEW	   8

2    STUDY CONCLUSIONS,  TESTING RECOMMENDATIONS,  AND SUMMARY
     OF STUDY RESULTS	11
     2.1  CONCLUSIONS AND RECOMMENDATIONS FOR TESTING ....  11
          2.1.1     XRF Instrument Conclusions  	  11
          2.1.2     Chemical Test Kit Conclusions	11
     2.2  RESULTS FOR STUDY OBJECTIVES  	  12
          2.2.1     Precision    and   Accuracy    of    XRF
                    Instruments	12
          2.2.2     Substrate Interference  	  12
          2.2.3     Large XRF Errors	13
          2.2.4     Field  Quality  Assurance   and  Quality
                    Control Methods 	  13
          2.2.5     Operating Characteristic Curves for Test
                    Kits	13
          2.2.6     Variability of Lead Levels in Paint ...  14

3  DETAILED STUDY RESULTS 	  15
     3 .1  LEAD LEVELS IN THE STUDY SAMPLES	15
     3.2  XRF INSTRUMENTS	16
     3.3  CHEMICAL TEST KITS	28
     3.4  PAINT CHIP SAMPLING AND ANALYSIS	33
                               VII

-------
                         LIST OF TABLES
Table 1.


Table 2.



Table 3.
Table 4.
Table 5.
Table 6.
Table 7.
Table 8.
Table 9.
Cross-Tabulation  of  Paint  Sample  Lead  Levels
mg/cm2  Lead and Percent Lead by Weight   .  .  .  .
in
Estimated Standard Deviation at 0.0 mg/cm2 and 1.0
mg/cm2  Lead  for One Nominal 15-Second Paint Reading
for K-Shell XRF Instruments,  by Substrate .  .  .  .  ,
     15
     17
Estimated Standard Deviation at 0.0 mg/cm2 and 1.0
mg/cm2  Lead for  One  Nominal 15-Second  Reading on
Control  Blocks  for  K-Shell  XRF   Instruments,  by
Substrate 	

Bias at  0.0  mg/cm2  and  1.0  mg/cm2  Lead  for One
Nominal   15-Second   Reading  for  K-Shell   XRF
Instruments, by Substrate 	
Bias  at  0.0  mg/cm2  and  1.0  mg/cm2  Lead  for One
Nominal  15-Second Reading  on Control  Blocks for
K-Shell XRF Instruments, by Substrate  	
Bias  at  0.0  mg/cm2  and  1.0  mg/cm2 Lead  for One
Nominal   15-Second   Reading   for  L-Shell   XRF
Instruments, by Substrate 	
                                                               18
                                                               19
                                                               20
                                                               21
False  Positive,  False  Negative  and Inconclusive
Percentages for K-Shell  XRF Instruments,  Based on
One Nominal 15-Second Reading With an INCONCLUSIVE
RANGE OF 0.4 - 1.6 mg/cm2 (1.0 mg/cm2 Threshold)

False  Positive,  False  Negative  and Inconclusive
Percentages for K-Shell  XRF Instruments,  Based on
One Nominal 15-Second Reading With an INCONCLUSIVE
RANGE OF 0.7 - 1.3 mg/cm2 (1.0 mg/cm2 Threshold)

False Positive and  False Negative Percentages for
K-Shell  XRF  Instruments,  Based on One  Nominal
15-Second Reading With  NO INCONCLUSIVE RANGE  (1.0
mg/cm2 Threshold)  	  ,
                                                               24
                                                               25
                                                               26
Table 10. False  Positive,  False  Negative  and Inconclusive
          Percentages for L-Shell  XRF Instruments,  Based on
          One Nominal 15-Second Reading with an INCONCLUSIVE
          RANGE OF 0.4 - 1.6 mg/cm2 (1.0 mg/cm2 Threshold)

Table 11. Overall False Positive and False Negative Rates for
          Test Kits Compared to  Laboratory Analytical Results
          Using the 1.0 mg/cm2 Threshold  	
                                                     26
                                                     29
                                IX

-------
Table 12.  Overall False Positive and False Negative Rates for
          Test Kits Compared to  Laboratory Analytical Results
          Using the 0.5% Threshold	
Table 13. Probability of  a  Positive Test Kit  Result  at 1.0
          mg/cm2  Lead	
Table 14. Probability of a Positive  Test  Kit Result at 0.5%
          Lead	

Table 15. Lead  Level  in  mg/cm2  at  Which  There  is  a  50%
          Probability of a Positive Test Kit Result  .  . .  .  ,

Table 16. Lead Level in Percent Lead  by Weight at Which There
          is  a  50%  Probability  of a   Positive  Test  Kit
          Result  	
29
31
32
33
33

-------
                         LIST OF FIGURES
Figure 1.  Full study template.
                               XI

-------
                     EXECUTIVE SUMMARY
     BACKGROUND

     This study was undertaken by the U.S.  Environmental
Protection Agency (EPA)  and the U.S.  Department of Housing and
Urban Development (HUD)  to collect information needed for the
development of federal guidance on testing paint for lead.  Prior
to this study, lead testing information was inadequate as little
formal evaluation had been done of the various field testing
methodologies.

     The impetus for this study came  from the passage of Title X
(Section 1017 of the Residential Lead-Based Paint Hazard
Reduction Act of 1992),  which mandated that the federal
government establish guidelines for lead-based paint hazard
evaluation and reduction. This study  was designed to produce the
type of detailed information EPA and  HUD needed in order to
respond to that mandate,  and focused  on two field technologies
that are used for testing for lead in paint:  portable X-ray
fluorescence  (XRF)  instruments and chemical test kits.  A pilot
study was conducted during March and  April 1993 in Louisville,
Kentucky.  The full study was conducted from July through October
1993 in Denver, Colorado and Philadelphia,  Pennsylvania.

     This is a summary report of the  study.  For readers that are
interested in more technical detail on the study, there is also a
comprehensive technical report available:  A Field Test of Lead-
Based Paint Testing Technologies: Technical Report (EPA 747-R-95-
002b).    Both reports are available from the National Lead
Information Center Clearinghouse (1-800-424-LEAD).
TECHNOLOGIES EVALUATED

     This study evaluated XRF instruments and chemical test kits.
XRF instruments measure lead in paint by directing high energy X-
rays and gamma rays into the paint,  causing the lead atoms in the
paint to emit X-rays which are detected by the instrument and
converted to a measurement of the amount of lead in the paint.
Chemical test kits detect the presence of lead in paint by a
chemical reaction that occurs when chemicals in the kit are
exposed to lead.   This reaction causes a color change to occur if
lead is present in the paint.
                              Xlll

-------
     Laboratory spectroscopic analysis of paint samples was
conducted to determine  the actual levels of lead in the paint.
The laboratory results were used as a benchmark for comparison to
the XRF and test kit results.
STUDY OBJECTIVES

     The overall study goal was to collect information about
field measurement methodologies sufficient to allow EPA and HUD
to establish guidance and protocols for lead hazard
identification and evaluation.   In order to achieve that goal,
the study had to be designed and conducted with sufficient rigor
and appropriate quality assurance.

     To ensure adequacy of the resulting data,  six specific study
objectives were developed:  three primary and three secondary.
The results are presented in this report in two ways:  overall
conclusions and testing recommendations are made in light of the
overall study goal,  and results are provided in terms  of the
specific study objectives.

     The three primary study objectives were:  (1)  to
characterize the performance (precision and accuracy)  of portable
XRF instruments under field conditions;   (2)  to evaluate the
effect on XRF performance of interference from the material  (the
substrate) underlying the paint; and   (3) to characterize the
relationship between test kit results and the actual lead level
in the paint (operating characteristic curves).

     The three secondary study objectives were:   (4) to
understand XRF behavior in the field through the investigation of
XRF measurements that were very different than their
corresponding lab result;   (5)  to evaluate field quality
assurance and control methods;  and   (6) to investigate the
variability of lead levels in the paint within the study sampling
locations.
FIELD TESTING

     Three primary concerns of the field testing portion of the
study were consistency, real world comparability, and quality
control.   Due to the differences among the three measurement
methods:  XRF, test kits, and laboratory analysis, field testing
approaches necessarily varied somewhat.  In order to ensure
consistency, testing was standardized as much as possible.  A

                               xiv

-------
template was designed for test locations throughout the study
housing units, and the different measurement methods were
systematically assigned to consistent test locations within the
template.  This approach ensured results could be compared across
different test locations and measurement methods.

     At each test location, chemical test kits were tested first.
The individuals who did the field testing of the test kits were
selected to represent typical homeowners who might purchase test
kits for their personal use.  That is,  they did not have any
specific scientific background nor prior training.  To further
replicate "real world" use, the test kits were rotated among the
testers during the study.  One of the test kits was an exception
to this.  It was a kit which is only used by state-certified
inspectors.   For that kit, a state-certified inspector was
brought in and that particular kit was not included in the kit
rotation.  After each tester completed a test location, the used
area of the template was covered to prevent subsequent testers
from observing the results obtained by prior testers.

     Once test kit testing was finished, paint samples were
taken.   Paint was removed from a specified location on the
template and sent to a laboratory for spectroscopic analysis.
A modified NIOSH method 7082 was followed with all appropriate
quality control samples including laboratory and field
duplicates.

     XRF testing was the final step in the field portion of the
study.   It was conducted by trained and licensed XRF instrument
operators employed by independent testing companies.  XRF testing
was carried out on the portions of the templates designated for
this purpose.   A number of quality control procedures were
employed, including the use of National Institute of Standards
and Technology (NIST) Standard Reference Material (SRM) paint
films.   The NIST SRM paint film is a thin layer of paint with a
known level of lead enclosed between two layers of plastic.  A
portion of the template was scraped bare of paint, revealing the
material underneath the paint,  the substrate, which was either
brick,  concrete,  drywall, metal, plaster or wood.  The NIST SRM
paint film was placed on the bare substrate and a reading was
taken in order to determine if the substrate interfered with the
XRF reading.  In addition, blocks of known substrate materials,
called control blocks, were utilized in the field.  The NIST SRM
paint film was placed on the appropriate block and XRF readings
taken in order to determine if control block substrates could be
surrogates for the substrates underlying the painted areas
tested.

                                xv

-------
STUDY RESULTS;

Laboratory Analysis Results

     There were two key results of the laboratory analyses.
First, laboratory analysis results exhibited a wide range of lead
levels with a distribution similar to that reported in the 1990
HUD National Survey of Lead-Based Paint in Housing.  Second, lead
levels appear to vary significantly across the same painted
surface.

     Two federal thresholds have been established to define lead-
based paint on painted architectural components.   If paint is
found to contain lead equal to or greater than these thresholds,
it is characterized as lead-based paint.  The federal threshold
in milligrams lead per unit area is 1.0 mg/cm2.   The federal
threshold in percent lead by weight is 0.5%.  Approximately 20%
of the samples analyzed in this study were equal  to or greater
than the federal threshold of 1.0 mg/cm2,  while 29% were  equal  to
or greater than the federal threshold of 0.5% lead.  A rough
numerical equivalence between results reported as mass of lead
per unit are.a (mg/cm2)  and as percent  lead by weight (%)  was
found in the study data.  That is, 1.0 mg/cm2 lead was found to
be roughly equivalent to 1% lead by weight.

     The variability of a set of test results is  the extent to
which the results in the set differ from one another.  The
standard deviation is a statistical measure of the extent that
actual test results tend to spread about an average value.  The
typical relative standard deviation for laboratory analytical
measurements in the study samples was 13%.  Variability between
field duplicate samples, taken nine inches apart  at a subset of
test locations,  was much larger, between 30% - 60%, indicating
significant variability in lead levels across the same painted
surface.   The statistical analysis of the data took variability
in lead levels into account.
Chemical Test Kit Results

     The primary result of the test kit evaluation is that they
varied widely in their performance in classifying paint against
either the 1.0 mg/cm2 or 0.5% threshold.   No single kit  achieved
a low rate of both false positive and false negative results and
their performance varied across substrates.
                               xvi

-------
     A false negative result occurs when the kit fails to detect
the presence of lead in paint equal to or greater than the
federal threshold, but in fact, the paint is shown by laboratory
analysis to contain lead equal to or greater than the threshold.
Similarly, a false positive result occurs when the kit detects
lead equal to or greater than the federal threshold, but
laboratory analysis shows that the paint does not contain lead
equal to or greater than the threshold.

     No kit in the study achieved low rates of both false
positive and false negative results.  Two out of six kits were
prone to false negative results.  Negative test results obtained
with these two kits do not necessarily indicate the absence of
lead.  The other four kits had a tendency to produce false
positive results,  even at levels of lead well below the federal
thresholds.

     Further, the performance of the test kits varied with
different types of substrates.  Most kits usually produced a
positive result on at least one substrate, even for very low lead
levels.  This suggests positive interferences with the chemicals
in the kits.  On the other hand, some test kits demonstrated
negative interferences on some substrates, as indicated by not
always giving a positive result for high levels of lead.
XRF Results

     The primary result of the XRF testing is that K-shell
instruments were often effective in classifying paint samples
against the federal threshold of 1.0 mg/cm2,  when using  an
inconclusive classification range, laboratory confirmation, and
substrate correction,  as needed.  Generally,  L-shell instruments
had extremely high false negative rates, making them ineffective
in classifying paint against the 1.0 mg/cm2 threshold.

     In this study, measurement bias, or bias, is the tendency of
a set of test results to be either greater or less than the
laboratory measurements of the lead content of the paint.  If
test results tend to be greater than the laboratory results,  they
are said to exhibit positive bias.  If the test results tend to
be less than the laboratory results, they exhibit negative bias.
Results of tests using XRF instruments showed both positive and
negative bias.  Biases of the K-shell XRF instruments were
strongly dependent on the underlying substrate.  One K-shell
instrument exhibited much less bias than the other XRF
instruments.  L-shell instruments generally had large negative

                              xvii

-------
biases at the 1.0 mg/cm2  threshold that were usually independent
of the substrate.

     Substrate correction,  using NIST SRM readings on  either the
scraped substrates or the control blocks,  did not reduce bias for
L-shell instruments.  For K-shell instruments,  results were
mixed.  Control block correction reduced bias for two  instruments
on some substrates.  Correction using NIST SRM readings on the
scraped substrate was effective for two instruments on most
substrates, and for another instrument on some substrates.

     The variability of the results from each XRF instrument was
estimated by calculating a standard deviation.   The results of
most K-shell instruments exhibited high variability at the
federal threshold of 1.0 mg/cm2.  The  variability in the  results
from the L-shell instruments was significantly lower than that of
K-shell instruments.

     Despite their generally high variability and bias, K-shell
instruments were often effective in classifying the paint samples
in this study against the federal threshold of 1.0 mg/cm2 when
using an inconclusive classification range of 0.4 to 1.6 mg/cm2
with mandatory laboratory confirmation.   Without using an
inconclusive range and laboratory confirmation, only two of the
K-shell instruments had both false positive and false  negative
rates below 10%.

     Generally, L-shell instruments had extremely high false
negative rates.  One L-shell instrument had moderate to high
false negative rates, depending on the width of the inconclusive
range, but still gave low readings on some samples with high
levels of lead.
OVERALL RECOMMENDATIONS FOR  TESTING

XRF Instrument Conclusions

     The primary XRF conclusion is that testing by K-shell XRF
instruments, with laboratory confirmation of inconclusive XRF
results, and with substrate correction in cases where this is
effective in reducing bias, is a viable way to test for lead-
based paint.  This approach can produce satisfactory results for
classifying the paint on architectural components using the
federal threshold of 1.0 mg/cm2.
                              xviii

-------
     Further, the variability found in paint samples located
approximately nine inches apart supports the conclusion that the
most effective method of XRF testing of a single architectural
component, such as a window sill, wall, or door, is to obtain
readings at different points on the component, and compute their
average.  This would replace the current practice which is to
average a number of XRF readings taken at a single point.

Chemical Tast Kit Conclusions

     The conclusion of this study is that test kits should not be
used for lead paint testing.  Test kits cannot determine the
extent of lead-based paint in a home and the need for protecting
the occupants, especially when repairs or renovations are carried
out.  Homeowners and renters cannot be confident that test kits
will discriminate accurately between lead-based paint and non-
lead based paint.  They should not make decisions on repairs,
renovations or abatements based on test kit results.
                               xix

-------
               1   DESCRIPTION  OF  THE  STUDY

1.1 BACKGROUND

     Lead-based paint  (LBP) in older housing, especially lead-
based paint in poor condition, is recognized as a major cause,
both direct and indirect, of elevated blood lead levels in
children between 1 and 6 years old.  Exposure to lead in paint
can come from the paint chips themselves,  from dust caused by
abrasion of paint on friction surfaces, or from chalking of
exterior paint.  The Lead-Based Paint Poisoning Prevention Act of
1971, as amended by the Housing and Community Development Act of
1987, established 1.0 mg/cm2 as  the federal  threshold  requiring
abatement of lead-based paint in public and Indian housing
developments nationwide.  To implement this legislation, Congress
required the U.S. Department of Housing and Urban Development
(HUD) to complete testing for lead-based paint in all public and
Indian housing by December, 1994.  In response to this
requirement, HUD, with substantial input from the Environmental
Protection Agency  (EPA), published interim guidelines for testing
and abatement of LBP in public and Indian housing in April, 1990.
At the time the HUD Guidelines were published, the research
conducted to evaluate the performance of X-ray fluorescence (XRF)
instruments and chemical test kits in detecting LBP at or above
the federal threshold was limited.  The recommended approach was
to perform XRF testing, with laboratory confirmation of
inconclusive results.  The Guidelines recommended that test kits
should not be used as a primary testing method.  Federal guidance
documents available from the National Lead Information Center
Clearinghouse also did not recommend the use of test kits by
homeowners or renters.

     The Residential Lead-Based Paint Hazard Reduction Act of
1992 ("Title X") mandated the evaluation and reduction of
lead-based paint hazards in the nation's existing housing.  Title
X also established 0.5% lead as an alternative to the 1.0 mg/cm2
threshold.   Section 1017 of Title X required HUD to develop
guidelines for federally-supported lead-based paint hazard
evaluation and reduction activities.  HUD is complying with this
requirement by preparing a major revision and expansion of the
1990 Guidelines.  To support the testing and inspection portion
of the revised Guidelines, EPA and HUD funded this field study of
technologies used to detect and measure lead in paint.  It is the
first comprehensive evaluation of XRF instruments and test kits
under field conditions.

-------
     This report is a summary of the study procedures and
provides results,  conclusions and recommendations for testing for
lead-based paint.   Study conclusions and testing recommendations,
and a summary of study results are in chapter 2.   Chapter 3
contains detailed study results.  Further information on all
aspects of the study can be found in the detailed report entitled
A Field Test of Lead-Based Paint Testing Technologies:  Technical
Report (EPA 747-R-95-002b).
1.2  STUDY OBJECTIVES

     The overall study goal was to collect information about
field measurement methodologies sufficient to allow EPA and HUD
to establish guidance and protocols for lead hazard
identification and evaluation.  In order to achieve that goal,
the study had to be designed and conducted with sufficient rigor
and appropriate quality assurance.

     To ensure adequacy of the resulting data, six specific study
objectives were developed:  three primary and three secondary.
The results are presented in this report in two ways: overall
conclusions and testing recommendations are made in light of the
overall study goal, and results are provided in terms of the
specific study objectives.

     The three primary study objectives were:  (1)  to
characterize the performance  (precision and accuracy) of portable
XRF instruments under field conditions;   (2) to evaluate the
effect on XRF performance of interference from the material  (the
substrate) underlying the paint; and   (3) to characterize the
relationship between test kit results and the actual lead level
in the paint (operating characteristic curves).

     The three secondary study objectives were:   (4) to
understand XRF behavior in the field through the investigation of
XRF measurements that were very different than their
corresponding lab result;  (5) to evaluate field quality
assurance and control methods; and   (6) to investigate the
variability of lead levels in the paint within the study sampling
locations.

     This study differs from previous studies conducted to
measure lead in paint because the study included a larger number
of samples and more diverse testing locations, and was designed
so that test results obtained at different locations could be
compared.  Paint from a total of 1,290 locations in 22 housing

-------
units in three cities was tested.  The tested locations were free
from identifiable biases and represent a variety of paint types,
substrates, architectural designs, and lead levels in paint.  The
study was designed to evaluate field testing technologies used to
identify lead-based paint that were commercially available or
were working prototypes as of June, 1993.  These technologies
included six types of XRF instruments and six chemical test kits.
Spectroscopic laboratory analysis was used to verify results
obtained by the XRF instruments and chemical test kits.
1.3  APPROACH

     The study began in March 1993 in Louisville, Kentucky, with
a pilot conducted at a vacant public housing development built in
1937.  Testing was conducted at 100 locations in 4 units in 2
buildings.  The pilot had several objectives.  First, it was
important to determine the feasibility of collecting large
numbers of paint samples in the field while ensuring the quality
of the samples,  and to develop and test a system for labelling
and tracking the samples.   Removal of paint with a heat gun and
paint scraper proved to be a successful technique.  A barcode
system that labelled and tracked samples was developed and
tested.  A working system for selecting and marking test
locations was developed.  The field practicality of the test kits
for large testing programs was evaluated.  Procedures for
monitoring XRF testing and recording of data were developed.
Field testing sequences to minimize the potential for variability
in XRF results caused by frequent substrate changes were used.
Time estimates for all aspects of sample collection and testing
were made.  The schedule and logistics for the full study were
based on these time estimates.  A database structure was
developed for storing and retrieving study data.

     The full study was conducted in two cities, Denver in July
and August 1993  and Philadelphia in September and October 1993.
Denver and Philadelphia were specifically chosen because housing
was available that met study criteria and because the public
housing authorities in those cities were willing to work closely
with EPA and its contractors.  The study tested units from both
multifamily housing, where units tend to be quite similar to each
other,  and from single-family homes.  A total of 10 scattered-
site single-family homes were tested in Denver; eight were built
between 1943 and 1952,  while two were older, dating from 1890 and
1905.  In Philadelphia, eight units in two buildings in a single
multifamily development built in 1942 were tested.  Including
those in the pilot study,  a total of 1,290 individual test

-------
locations on 6 substrate types in the 22 housing units were
tested.   There were 100 test locations in Louisville,  750 in
Denver and 440 in Philadelphia.   The breakdown of testing
locations by substrate was:  93 brick,  226 concrete,  124 drywall,
217 metal, 242 plaster, and 388 wood substrates.
1.4  TECHNOLOGIES

     Chemical test kits detect the presence of lead in paint by a
chemical reaction that occurs when chemicals in the kit are
exposed to lead.  This reaction causes a color change to occur if
lead is present in the paint.  The test kits in the study
represented the range of kits available at the time the study was
conducted.  Test kits from five different manufacturers were
examined in this study:  three rhodizonate based kits, two sodium
sulfide based kits, and one proprietary kit.  Both of the most
common types of chemical test kits, rhodizonate based kits and
sodium sulfide based kits,  were used in the pilot study.  The
rhodizonate kits included were LeadCheck (also called LeadCheck
II)  and the sanding and coring versions of Lead Alert; the sodium
sulfide kits were Lead Detective and the Massachusetts state-
approved kit.  The pilot study also included the Lead Zone kit,
which utilizes proprietary chemistry.   It was expected that the
results of the pilot study would be similar for kits based on
similar chemistry, that is, rhodizonate or sodium sulfide, so
that fewer kits would need to be included in the full study.
However, the test results were not similar for kits utilizing
similar chemistry, so the same six kits were included in the full
study.

     Portable XRF instruments direct high energy X-rays and gamma
rays into paint.  These high-energy rays strike lead atoms,
causing electrons to be ejected from their electron orbits, or
shells.  In a process called fluorescence,  other electrons refill
the voids left by the ejected electrons, producing X-rays.  These
X-rays have specific frequencies based on differences in energy
between the electron shells which contained the emitted electrons
and the electron shells which received the electrons.  The amount
of X-ray energy emitted at several specific frequencies, in this
case called K-shell or L-shell X-ray energy, is measured by
detectors on XRF instruments and used to calculate the amount of
lead in paint.

     XRF instruments are classified by the type of X-ray energy
that they detect, K-shell X-rays, L-shell X-rays, or both.
K-shell X-rays are more highly penetrating than L-shell X-rays

-------
since L-shell X-rays have lower energy.  For this reason, K-shell
X-rays are more useful for detecting lead in deeper layers of
paint.  Two of the XRF instruments in this study detected K-shell
X-rays, two XRF instruments detected L-shell X-rays, and two
instruments detected both K-shell and L-shell X-rays.

     Efforts were made to include a representative example of
every XRF instrument available at the time of the study.  Six
types of XRF instruments were in the study.  The MAP-3, the
Microlead I, and the XK-3 were included because they were the
most commonly used instruments for LBP testing when the study
began.  The X-MET 880 was included because it performed
successfully in the pilot study.  After completion of the pilot
study, all other known manufacturers of XRF instruments or
working prototypes were invited to participate in a day of
ruggedness testing to determine whether the instruments were
portable and could function reliably throughout a full day of
field testing.  As a result,  two additional instruments, the Lead
Analyzer and a prototype of the XL, were included in the full
study.  Since the conclusion of the field portion of the study,
new XRF instruments and modified versions of some tested
instruments have become commercially available.

     The third type of technology in the study was laboratory
analysis which was used to verify results obtained by the two
field technologies:  chemical test kits and XRF instruments.  For
this study, the laboratory instrument used was an atomic emission
spectrophotometer.   The laboratory procedure involved dissolving
paint samples in acid, then filtering and diluting them.  A
portion of the dissolved sample was placed in the
spectrophotometer and heated to extremely high temperatures by a
device inside the spectrophotometer called a high temperature
atomizer.  At very high temperatures, most of the sample is
broken down into individual atoms.   Individual atoms absorb and
re-emit energy produced by the atomizer.  Atoms of different
chemical elements re-emit energy at different energy levels.  A
detector in the spectrophotometer sorts and measures the energy
re-emitted by the atoms of different chemical elements.  In this
way, the amount of energy re-emitted by lead atoms is measured
and then used to calculate the amount of lead in the sample.  The
particular type of spectrophotometer used in this study was an
inductively coupled plasma atomic emission spectrophotometer
(ICP).  The analytical laboratory results were continually
evaluated by using reference materials to assure the accuracy of
the laboratory analysis of field samples.

-------
     Chemical test kit results were reported as either negative
or positive indicating the absence of lead or presence of lead,
respectively.  XRF instruments and laboratory analysis results
were reported as quantitative measures of lead.  XRF instruments
report their results as mass of lead per unit area (mg/cm2) .
Laboratory analysis results were reported both as mass of lead
per unit area (mg/cm2)  and percent  lead by weight  (%) .
1.5  FIELD TESTING

     Templates were designed for marking test locations in the
study housing units so that results could be compared for
different test technologies and locations.  The most commonly
used template, shown in Figure 1, was a rectangle 14 inches long
and 4 inches wide.  For certain locations such as door frames, a
thin version of the template, 2 inches by 14 inches, was needed.
On the left of the most commonly used template was a square 4x4
inches; in the center, a second 4x4 inch square was divided
into four 2x2 inch subsquares; the remaining 6x4 inch
rectangle on the right of the template was divided into six
vertical strips each 1x4 inches.  One of the 2x2 inch
subsquares was randomly selected as the location for paint
sampling for laboratory analysis.  At 10% of locations in the
full study, a duplicate paint sample was taken adjacent to the
right end of the template for use in assessing variability in the
paint lead levels.  Following paint sampling, the remainder of
the center 4x4 inch square was scraped to remove all remaining
paint.   It was then used for taking XRF measurements on bare
substrates both with and without the standard reference material
paint films  (SRM 2579) developed by the National Institute of
Standards and Technology  (NIST).  The NIST SRM paint film is a
thin layer of paint with a known level of lead enclosed between
two layers of plastic.  The 4x4 inch square on the left of the
template was used for XRF measurements on paint. The six 1x4
inch strips were randomly assigned as testing locations for the
six chemical test kits.  Each of the testing locations in the
study was selected and marked by the field statisticians using
the template and an indelible ink marker.  Each test location was
numbered for identification and sample tracking.

     The first step in the full study was to test the six
chemical test kits.  Testers for five of the six test kits were
individuals without any special scientific background or prior
training.  They were selected to represent typical homeowners who
might purchase kits for their personal use.  The testers were
trained by field supervisors to ensure that study protocols were

-------
         PAINTED
     XRF TEST SURFACE
    BARE
XRF TEST SURFACE
TEST KIT TEST SURFACES








PRIMARY
PAINT
SAMPLE


















FIELD
LJUrLIUAl t
PAINT
SAMPLE

                             t
                      A   C
                      Test Kits
Figure 1. Full study template.

followed.  The training did not provide the  testers with
knowledge about test kit operation beyond the  information
contained in the manufacturer's instructions.  These  five  kits
were rotated among the testers during the study.  The sixth kit,
tested by a state-certified inspector, was not part of the kit
rotation.  After each tester had completed the testing at  a
location, the strip of the test location where the color change
could be observed was taped over to prevent  subsequent testers
from knowing the result of the test.

     After test kit testing was completed, paint chip samples
were taken and sent to the laboratory for ICP  spectroscopic
analysis.  Paint samples were homogenized by grinding to a
powder, and, if necessary, subsampled prior  to analysis.
Subsampling was necessary because the total  mass of many samples
was too large for a single laboratory analysis.

     The third and final step in the field study was  XRF testing.
It was conducted by trained and licensed XRF instrument  operators
employed by independent testing companies.   Within each  unit,
test locations from each substrate type were tested as a group.
For example, all locations on metal substrates were tested,  then
all locations on wood substrates were tested,  etc.  This was done
to minimize the potential for XRF variability  caused  by  repeated
substrate changes.  However, the order of substrates  tested
within a unit was varied.  Quality control checks were also
performed on six control blocks, each composed of a different
substrate,  combined with the NIST SRM paint  films.  To ensure
that the testing protocol was followed exactly, and to ensure
accurate recording of data, during testing each XRF instrument

-------
operator was observed by a full-time monitor who recorded the
results and reported to a field supervisor.
1.6  PEER REVIEW

     The technical report on this study was reviewed
independently by members of a peer review panel.   Comments which
are important for interpreting the study results  or which had an
important impact on the report are discussed below.

     A comment from a number of reviewers related to the
representativeness of the study paint samples and the fact that
the sample was not selected randomly from the national housing
stock.  Although the sample was not randomly selected, the sample
did include different substrate materials, housing components,
paint thicknesses, and lead levels.  The housing  in the study
included both single-family homes and multifamily housing.  The
distribution of lead levels in the study is similar to the
distribution in the HUD National Survey of pre-1980 housing.

     A comment from the reviewers related to the  training
received by the individuals who, as representative homeowners or
renters, applied the test kits.  There were concerns that it
would have been more appropriate to have no training to better
simulate what a homeowner or renter would encounter.  However,
the training did not give the individuals in the  study any more
information beyond what could have been obtained  from a careful
reading of the kit instructions.  The kits were rotated among the
testers to reduce the chance of an individual becoming an expert
with a single kit.  Nevertheless, it is probably  fair to say that
the training, the availability of on-site supervisors, and the
large number of tests performed by the individual testers
provided conditions that exceeded what would be typical for a
homeowner or renter who purchased a test kit.

     A comment was made concerning the impact of  spatial
variation and laboratory measurement error on the false positive
and false negative rates calculated from the study data.  A
simulation study was conducted to address this comment and the
results included in the final technical report.  The simulation
study demonstrated that the false positive and false negative
rates were robust, and therefore accurately portrayed performance
of the technologies in the study.  Another reviewer comment on
the statistical analysis of paint samples with lead below
detection levels led to an improvement in the approach for
estimating model parameters.

                                8

-------
     A number of reviewers commented on the length of the
technical report.  In response to those comments, a summary
report was developed from the technical report to make the
information in the technical report accessible to a wider
audience.

     EPA has established a public record for the peer review
under administrative record 142.  The record is available in the
TSCA Nonconfidential Information Center, which is open from noon
to 4 PM Monday through Friday, except legal holidays.  The TSCA
Nonconfidential Information Center is located in Room NE-B607,
Northeast Mall, 401 M Street SW, Washington, D.C.

-------
10

-------
 2   STUDY CONCLUSIONS,  TESTING  RECOMMENDATIONS,
              AND  SUMMARY OF STUDY RESULTS

     This section provides  conclusions and recommendations for
testing as well as  a summary of  results  from the study.  The
section is divided  into two subsections.  Section 2.1 contains
conclusions and recommendations  for testing for lead-based paint
and section 2.2 contains a  summary of results organized by study
objectives.  The conclusions, recommendations, and results are
based on the samples and data collected  in this study, and are
specific to the laboratory  analysis method, chemical test kits,
and XRF instruments used.
2.1   CONCLUSIONS AND RECOMMENDATIONS  FOR TESTING

2.1.1     XRF Instrument Conclusions

     The primary XRF  conclusion  is  that testing using K-shell XRF
instruments,  with laboratory confirmation of inconclusive XRF
results, and with substrate  correction in cases where this is
effective in reducing bias,  is a viable way to test for lead-
based paint.   This approach  can  be  expected to produce
satisfactory results  for classifying the paint on architectural
components as either  above or below the federal threshold of 1.0
mg/cm2.

     Currently,  a common practice is to average a number of
readings taken at a single point on an architectural component.
The study demonstrated that  the  most effective method of XRF
testing is to obtain  readings at different points on the
component and compute their  average.  This recommendation is
supported by the variability found  in paint samples located
approximately nine inches apart, and evidence that a single XRF
reading at one point  provided almost as much information as an
average of three XRF  readings at the same point.


2.1.2     Chemical Test Kit  Conclusions

     The conclusion of this  study is that test kits should not be
used for lead paint testing.  Test  kits cannot determine the
extent of lead-based  paint in a  home and the need for protecting
the occupants,  especially when repairs or renovations are carried
out.   Homeowners and  renters cannot be confident that test kits
will discriminate accurately between lead-based paint and non-

                               11

-------
lead based paint.   They should not make decisions on repairs,
renovations or abatements based on test kit results.
2.2   RESULTS FOR  STUDY  OBJECTIVES

2.2.1     Precision and Accuracy of XRF Instruments

     The first primary objective of this study was to
characterize the precision and accuracy of XRF instruments on
common substrates under field conditions.   The results of the
study showed that most K-shell instruments exhibited relatively
high variability and a high degree of bias at lead levels close
to the federal threshold of 1.0 mg/cm2.  Nevertheless,  K-shell
XRF instruments reliably classified the paint samples in this
study vis-a-vis the federal threshold of 1.0 mg/cm2,  provided a
suitable inconclusive range and substrate correction (where
appropriate) were used.

     Test results using L-shell instruments generally exhibited
large negative biases which increased with the lead level in the
paint.  Bias for L-shell instruments was usually substantial at
1.0 mg/cm2  lead.   L-shell instruments were less  variable  than K-
shell instruments.  As a consequence of the large negative
biases,  L-shell instruments exhibited a high rate of false
negative results when classifying paint using the 1.0 mg/cm2
threshold.   When an inconclusive range was added, L-shell
instruments, with one exception, still had high rates of false
negatives.   The one exception exhibited reductions in the rate of
false negatives as the inconclusive range was lengthened.
2.2.2     Substrate Interference

     The second primary objective of the study was to evaluate
the effect on the performance of XRF instruments of interference
or bias attributable to the underlying substrate and, hence, to
evaluate the utility of different approaches for adjusting XRF
readings for this bias.  The results of the study showed that
biases of most K-shell instruments were strongly substrate
dependent.  Test results using L-shell instruments generally
exhibited large negative biases at the 1.0 mg/cm2  threshold that
were usually independent of the substrate.

     Substrate correction obtained using readings on NIST SRM
paint films placed on test location areas scraped bare of paint
reduced bias for two of the K-shell instruments, and for a third

                                12

-------
on metal and wood substrates.  The already low bias of the fourth
K-shell instrument's results was unchanged.  Substrate correction
using NIST SRM paint films over control blocks was effective in
reducing bias for one K-shell instrument, and somewhat effective
for a second on plaster, concrete and metal.  No method of
substrate correction reduced the bias of L-shell readings.
2.2.3     Large XRF Errors

     A secondary objective of the study was to investigate large
errors in the XRF measurements, i.e., measurements that were very
different than their corresponding lab results.  The results of
the study showed that the incidence of large XRF errors was very
low (0.6%).  Moreover, many of the large errors occurred for
several instruments at the same test location.  This suggests a
common cause other than mere erratic behavior on the part on any
single XRF instrument.
2.2.4     Field Quality Assurance and Quality Control Methods

     Another secondary objective of the study was to evaluate
field quality assurance and quality control methods.  The study
results showed that NIST SRM readings on control blocks were
unable to predict XRF instrument performance on painted
components in most cases.  In particular,  the study results
showed that erratic behavior in XRF readings taken on control
blocks was not necessarily predictive of similarly erratic
behavior on actual paint samples.  Finally, with the exception of
two K-shell instruments used on some substrates, substrate
correction using readings on NIST SRM paint films placed on
control blocks of substrate materials brought to the site was not
effective in reducing biases of readings attributable to
substrate interference.
2.2.5     Operating Characteristic Curves for Test Kits

     The third primary objective of the study was to estimate the
operating characteristic curve for each test kit under field
conditions.  The results of the study showed that the probability
of a positive classification when the sample's lead level was
equal to the federal thresholds varied depending on the kit and
substrate and that high levels of lead would not always be
detected by some test kits.  Furthermore, there were numerous
cases of positive test results at lead levels well below the

                                13

-------
federal thresholds.   None of the test kits used in this study
demonstrated low rates of both false positive and false negative
results when compared to laboratory analytical results using the
federal thresholds,  1.0 mg/cm2  and  0.5%.
2.2.6     Variability of Lead Levels in Paint

     The third secondary objective of the study was to
investigate the variability of lead levels in paint using
laboratory measurements of field duplicate samples.  The study
results showed that the typical relative standard deviation for
laboratory analytical measurements in the study samples was 13%.
Variability between field duplicate samples was much larger,
between 30% - 60% at one standard deviation,  indicating
significant variability in lead levels between paint samples
approximately 9 inches apart.  This variability in lead levels
within single architectural components, called spatial
variability, was the primary cause of variability in the paint
samples.
                                14

-------
                3   DETAILED  STUDY  RESULTS
     This section contains details of the study results
3.1  LEAD LEVELS IN  THE STUDY SAMPLES
1.   Of the 1,290 paint samples collected and analyzed in the
     laboratory in this study, approximately 20% contained lead
     at a level equal to or greater than 1.0 ing/cm2,  one of the
     federal thresholds for defining LBP on painted surfaces.
     Approximately 29% of the samples contained lead equal to or
     greater than 0.5% by weight, the other federal threshold for
     LBP on painted surfaces.

     Lead levels in the samples were reported by the laboratory
     as mass per unit area  (mg/cm2 lead)  and percent  lead by
     weight (%).   Table 1 presents a cross-tabulation of lead
     levels expressed in mg/cm2 and percent  lead by weight.   The
     arithmetic mean lead level in the study samples was 1.17
     mg/cm2  (1.12%).   The  median lead level  of  the study samples
     was 0.20 mg/cm2  (0.20%).   The 25th and  75th percentiles were
     0.03 mg/cm2   (0.05%)  and  0.62 mg/cm2  (0.72%).  The minimum
     and maximum values were 0.0001 mg/cm2 (0.0004%)  and 37.29
     mg/cm2  (34.56%).

Table 1.    Cross-Tabulation of Paint Sample Lead Levels in mg/cm2 Lead and
          Percent Lead by Weight.
Percent
Lead
by Weight
< 0.5
0.5 - 1.0
a 1.0
Totals
mg/cm2 Lead
< 0.5
874
36
16
926
0.5 - 1.0
42
44
25
111
2 1.0
2
14
237
253
Totals
918
94
278
1,290
2.
For the paint samples, lead levels expressed in mg/cm2 and
lead levels expressed in percent lead by weight were  roughly
equivalent, as shown by the distribution in Table 1.  A
level of 1.0 mg/cm2 was roughly equivalent to 1.0% by weight
and a level of 0.5% by weight was roughly equivalent  to  0.5
mg/cm2.
                                15

-------
     The overall average ratio between the  two types  of
     measurement units for the 1,290  primary paint  samples
     analyzed in the laboratory was  1.00.   In 80% of  the samples,
     the ratio was between 0.25 and  2.34.   A regression  plot of
     results expressed in percent  lead by weight  (%)  versus mass
     of lead per unit area (mg/cm2) using a  logarithmic  scale
     showed good agreement between the two  types of measurement
     units (R2  =  0.91),  with the following relationship:

               PERCENT LEAD =  0.96 x (AREA  LEAD)0'85, where

               PERCENT LEAD =  percent lead  by weight  (%) and

               AREA LEAD = mass of lead per unit area (mg/cm2) .

     This relationship suggests that  0.5%  lead is roughly
     equivalent to 0.5 mg/cm2  lead, while 1.0  mg/cm2 lead is
     roughly equivalent to 1.0% lead.  This demonstrates that the
     threshold of 1.0 mg/cm2 lead  is  typically less stringent
     than 0.5% lead.


3.2  XRF  INSTRUMENTS

1.    Most K-shell instruments  exhibited relatively  high
     variability, even for paint with low levels of lead.  The
     amount of variability was sometimes related to the  level of
     lead in the sample.

     Table 2 shows estimated standard deviations for  each
     substrate for results using the K-shell XRF instruments at
     lead levels of 0.0 mg/cm2 and 1.0  mg/cm2, for  a single
     15-second (nominal) reading taken on the painted surface of
     each test location.  For  XRF  instrument results  that showed
     significant variation between instruments and/or cities in
     the study, a range of values  for the standard  deviation is
     also presented.  In these cases, the single value in the
     table represents the single instrument, or a group  of
     similar instruments, with the largest  number of  readings
     taken.  These estimated standard deviations take into
     account several sources of variability in addition  to
     instrumental variation.  These include  site-specific factors
     such as the substrate composition and the age  and thickness
     of the paint.  The MAP-3, Microlead I,  and XK-3  results
     exhibited similar high levels of variability.   The  Lead
     Analyzer's results were significantly less variable than the
     other three.  Generally,  the  instruments' results showed

                                16

-------
Table 2.
Estimated Standard Deviation at 0.0 mg/cm2 and 1.0 mg/cm2 Lead for
One Nominal 15-Second Paint Reading for K-Shell XRF Instruments,
by Substrate.
SUBSTRATE
Brick
Concrete
Drywall
Metal
Plaster
Wood
LEAD
ANALYZER
K-SHELL
0.0
mg/cm2
0.17
0.11
0.08
0.18
0.14
0.08
1.0
mg/cm2
0.23
0.37
0.35
0.41
0.24
0.43
MAP -3
K-SHELL
0.0
mg/cm3
0.93
0.90
0.38
0.37
0.81
0.49
1.0
mg/cm2
0.93
1.00
0.38
0.55
0.87
0.67
MICROLEAD I
0.0
mg/cm2
0.59
0.61
(0.48-1.24)
0.34
(0.34-0.53)
0.62
(0.37-0.81)
0.55
(0.37-1.01)
0.62
(0.50-1.06)
1.0
mg/cm2
0.55
0.72
(0.48-1.31)
0.34
(0.34-0.53)
0.68
(0.55-0.81)
0.64
(0.46-1.01)
0.92
(0.55-1.06)
XK-3
0.0
mg/cm'
0.60
0.64
(0.51-0.85)
0.36
(0.21-0.36)
0.52
(0.34-0.70)
0.55
(0.40-0.55)
0.49
(0.25-0.51)
1.0
mg/cm2
0.60
0.64
(0.51-0.85)
0.56
(0.55-0.56)
1.06
(0.49-1.63)
0.63
(0.40-0.81)
0.69
(0.44-1.15)
Ranges presented for XRFs demonstrating significant variability between different instruments.
     higher variability at 1.0 mg/cm2 lead than at  0.0 mg/cm2.
     The difference in variability at the  two levels was greatest
     for the Lead Analyzer's results and least for the MAP-3's
     results.  Variability of control block  quality control test
     results was significantly lower than  results for field test
     locations.  Table 3 is the companion  to Table 2 for control
     block test results.  The standard  deviation  at 0.0  mg/cm2
     was estimated using XRF test results  on the  bare control
     blocks.  The standard deviation at 1.0  mg/cm2  was estimated
     using XRF test results from control blocks covered with the
     NIST SRM 2579 paint film that has  a lead level of 1.02
     mg/cm2.   As  in Table  2,  the Lead Analyzer's  results were
     less variable than the results of  the other  three
     instruments.  For tests on control blocks, the Lead
     Analyzer's results were more variable at 1.0 mg/cm2 than  at
     0.0 mg/cm2.   However,  the  other three instruments' results
     showed similar variability on the  control blocks at the two
     levels, 0.0 and 1.0 mg/cm2.
                                17

-------
Table 3.    Estimated Standard Deviation at 0.0 mg/cm2 and 1.0 mg/cmj Lead for
          One Nominal 15-Second Reading on Control Blocks for K-Shell XRF
          Instruments, by Substrate.
SUBSTRATE
Brick
Concrete
Drywall
Metal
Plaster
Wood
LEAD
ANALYZER
K- SHELL
0.0
mg/cm1
0.11
0.11
0.07
0.15
0.09
0.03
1.0
mg/cm1
0.24
0.24
0.19
0.22
0.20
0.18
MAP -3
K- SHELL
0.0
mg/cm'
0.72
0.64
0.28
0.21
0.69
0.24
1.0
mg/cm1
0.61
0.67
0.34
0.25
0.57
0.24
NICROLEAD I
0.0
mg/cm1
0.48
(0.28-0.76)
0.38
(0.31-0.51)
0.29
(0.21-0.42)
0.27
(0.22-2.39)
0.49
(0.33-0.66)
0.26
(0.24-2.08)
1.0
mg/cm1
0.40
(0.26-0.61)
0.50
(0.41-0.68)
0.29
(0.25-0.44)
0.36
(0.21-2.14)
0.47
(0.30-0.69)
0.33
(0.23-2.25)
XK-3
0.0
mg/cm1
0.33
0.41
0.32
0.38
0.50
0.39
Ranges presented for XRFs demonstrating significant variability between different
1.0
mg/cm'
0.41
0.50
0.45
0.47
0.70
0.43
instruments .
     2.   Biases of most  K-shell instruments were strongly
          substrate dependent.

     Bias of an XRF instrument  is defined as the average
     difference between XRF readings and the true lead level  in
     the paint.  Table  4  shows  biases of the K-shell XRF
     instruments on the field samples.  The results of the Lead
     Analyzer  exhibited low bias on all substrates.  The MAP-3's
     results showed negative bias on brick, concrete, and
     plaster;  positive  bias on  metal; and low bias on wood and
     drywall with  the exception of wood at 1.0 mg/cm2.   The
     Microlead I's results were mostly positively biased, but
     with large differences between individual instruments.   The
     XK-3's results showed large positive biases except on wood
     and drywall,  and also exhibited substantial variation
     between individual instruments.  Table 5 shows biases for
     the K-shell instruments'  results, estimated using control
     block readings.  For the Lead Analyzer, control block biases
                                18

-------
Table 4.    Bias at 0.0 mg/cm2 and 1.0 mg/cm2 Lead for One Nominal 15-Second
          Reading for K-Shell XRF Instruments, by Substrate.
SUBSTRATE
Brick
Concrete
Drywall
Metal
Plaster
Wood
LEAD
ANALYZER
K- SHELL
0.0
mg/cm1
0.08
0.02
-0.02
0.06
0.03
0.01
1.0
mg/cm1
-0.21
-0.01
0.18
0.02
-0.11
0.28
MAP -3
K- SHELL
0.0
mg/cm1
-0.60
-0.66
0.01
0.33
-0.68
-0.05
1.0
mg/cm1
-0.80
-0.45
-0.12
0.42
-0.55
0.36
MICROLEAD I
0.0
mg/cm1
0.10
0.28
(-0.03-0.89)
0.02
(0.00-0.66)
0.35
(-0.42-1.08)
0.01
(-0.09-0.22)
0.00
(0.00-0.60)
1.0
mg/cm1
-0.33
0.38
(0.01-1.23)
0.22
(0.16-1.79)
0.45
(-0.17-1.36)
0.06
(-0.32-0.18)
0.43
(0.18-0.90)
XK-3
0.0
mg/cm1
0.86
1.08
(0.66-1.84)
-0.33
(-0.33-0.25)
0.45
(0.26-1.48)
0.54
(0.38-1.68)
-0.07
(-0.07-0.93)
1.0
mg/cm1
0.88
1.75
(0.23-2.57)
-0.09
(-0.09-0.18)
0.86
(0.81-1.69)
0.57
(0.18-1.63)
0.35
(0.31-1.23)
Ranges presented for XRFs demonstrating significant variability between different instruments.
     were very small.   For  the  MAP-3,  the control block result
     biases were generally  of the same sign, positive or negative,
     as the field sample  result biases,  but the magnitudes were
     very different.   For the Microlead I,  sporadic agreement
     existed between control block and field sample result biases.
     For example, the  control block results showed negative  bias
     on metal, while the  field  sample  results showed a positive
     bias on the same  substrate.   For  the XK-3, the control  block
     result biases usually  tracked the field sample result biases.
                                 19

-------
Table 5.
Bias at 0.0 mg/cm2 and 1.0 mg/cm2 Lead for One Nominal 15-Second
Reading on Control Blocks for K-Shell XRF Instruments,  by
Substrate.
SUBSTRATE
Brick
Concrete
Drywall
Metal
Plaster
Wood
LEAD ANALYZER
K-SHELL
0.0
mg/cm1
0.05
-0.01
-0.01
-0.01
-0.03
-0.00
1.0
mg/cm1
0.08
0.06
0.06
0.11
0.05
0.04
MAP -3
K-SHELL
0.0
mg/cm2
-1.18
-1.20
-0.10
0.23
-1.38
-0.27
1.0
mg/cm2
-0.05
-0.18
0.04
0.18
-0.64
-0.14
MZCROLEAD I
0.0
mg/cm2
0.47
(-0.10-0.51)
0.57
(0.15-1.43)
0.03
(-0.62-0.14)
-0.34
(-0.82-2.25)
0.45
(0.06-1.13)
0.15
(-0.22-1.57)
1.0
mg/cm2
0.45
(-0.31-0.54)
0.70
(0.25-1.59)
0.12
(-0.56-0.18)
-0.35
(-0.84-2.00)
0.40
(0.09-1.02)
0.18
(-0.05-1.47)
XK-3
0.0
mg/cm2
0.97
0.89
0.17
1.10
0.83
0.25
1.0
mg/cm2
1.10
1.00
0.48
1.34
0.83
0.49
Ranges presented for XRFs demonstrating significant variability between different instruments.
3.    With the exception  of  the  XL prototype,  test results using
     L-shell instruments exhibited large negative biases at the
     1.0 mg/cm2 threshold.  However, test results using L-shell
     instruments were  less  variable than results obtained using
     K-shell instruments.

     Table 6 shows estimated  biases of field sample results using
     L-shell instruments at 0.0 mg/cm2  and  1.0  mg/cm2.  The
     instruments' results show  little  bias  at 0.0 mg/cm2.
     However, large negative  biases,  typically between -0.7 and
     -0.9 mg/cm2,  at 1.0 mg/cm2  lead,  are shown for all L-shell
     instruments' results except those obtained using the XL.

     Standard deviations were usually 0.2 mg/cm2  or less for
     field sample test results  at both 0.0  and 1.0 mg/cm2 lead,
     although the MAP-3's L-shell results showed slightly higher
     variability than  this  on metal.   Variability of control
     block results was significantly lower for all L-shell
     instruments compared to  K-shell  instruments' results.
                                20

-------
Table 6.    Bias at 0.0 mc,/cm2 and 1.0 mg/cm2 Lead for One Nominal 15-Second
          Reading for L-Shell XRF Instruments, by Substrate.
SUBSTRATE
Brick
Concrete
Drywall
Metal
Plaster
Wood
LEAD ANALYZER
L-SHELL
0.0
mg/cm2
0.01
0.01
-0.01
0.01
0.002
-0.02
1.0
mg/cm2
-0.77
-0.84
-0.70
-0.79
-0.80
-0.74
MAP -3
L-SHELL
0.0
mg/cm2
0.01
-0.14
-0.12
0.04
-0.12
-0.08
1.0
mg/cma
-0.88
-0.94
-0.62
-0.69
-0.96
-0.65
XL
0.0
mg/cm2
0.11
0.07
0.08
0.07
0.08
0.06
1.0
mg/cm2
-0.40
-0.15
-0.63
-0.10
-0.26
-0.30
X-MET 880
0.0
mg/cm2
0.03
0.05
0.04
0.11
0.05
0.04
1.0
mg/cm2
-0.74
-0.89
-0.74
-0.77
-0.88
-0.70
     The XL results showed smaller biases at  1.0  mg/cm2 than
     results of the other L-shell instruments,  but still  showed
     large negative biases at higher lead levels.

     Biases of the XL's results at 1.0 mg/cm2 lead range  from
     -0.10 to -0.63 mg/cm2.   There was some variation in  bias
     between different XL machines on metal and wood at  1.0
     mg/cm2.   The  instrument's results showed large negative
     biases at higher lead levels.  For example,  it  read  1.0
     mg/cm2  or less on 26% of the samples with lead levels  of
     10.0 mg/cm2 or greater.   The XL instruments used in  this
     study were prototype models.

     Substrate correction obtained using readings for NIST SRM
     paint films placed on test location areas  scraped bare of
     paint reduced bias for results using the Microlead I and the
     XK-3, and for the MAP-3 K-shell instrument results on metal
     and wood substrates.  The already low bias of the Lead
     Analyzer's K-shell results was unchanged.
                                21

-------
     Two methods of substrate  correction using NIST SRM paint
     films placed on the bare  substrate  were  analyzed.   In the
     first method,  called "full"  correction,  readings  were taken
     at each individual  test location after the NIST SRM paint
     film was placed on  the bare  area of the  substrate.   These
     readings provided an offset  value used to correct the paint
     sample readings taken at  that  location.   The  second method,
     called "average" correction, used the  average of  all
     readings taken after the  NIST  SRM paint  film  was  placed on
     the bare area at test locations of  the same substrate in the
     entire dwelling unit. These average readings provided an
     offset value used to correct paint  sample readings taken on
     the same substrate  in a dwelling unit.  Full  correction is
     not a practical method, while  average  correction
     approximates the method recommended in the 1990 HUD
     Guidelines.  The two methods were found  to give
     approximately the same results.

6.    With the exception  of the XK-3 and  the MAP-3  on some
     substrates, substrate correction using readings for NIST SRM
     paint films placed  on control  blocks of  substrate materials
     brought to the site was not  effective  in reducing biases of
     K-shell readings attributable  to substrates.

     A third method of correcting for bias  attributable to
     substrates, called  "control  block"  correction, used the
     average of readings taken on control blocks after the SRM
     paint film was placed on  the control block.  These average
     readings provided an offset  value used to correct paint
     sample readings taken on  the same substrate.   Control block
     correction was not  a generally effective technique to detect
     location-dependent  substrate characteristics  which cause the
     results to show bias. An exception was  the XK-3  instrument.
     This instrument's results typically exhibited positive bias
     which was reduced significantly by  control block  correction.
     For the MAP-3, control block correction  was somewhat
     effective in reducing bias for plaster,  concrete, and metal.
     For the Microlead I, control block  correction actually
     increased bias for  metal  and plaster.

7.    No method of substrate correction reduced the bias of
     L-shell readings.

     Neither the use of  control blocks nor  readings taken after
     placing NIST SRM paint films on scraped  substrates was
     effective in reducing the biases in L-shell readings.  This
     is because L-shell  result bias is caused by difficulty in

                               22

-------
     detecting lead in deeper layers of paint,  which was not
     simulated by usage of the NIST SRM paint films.

8.    Despite the generally high variability and bias of their
     results,  K-shell XRF instruments reliably classified the
     paint samples in this study using the federal threshold of
     1.0 mg/cm2, with laboratory  confirmation of XRF readings
     between 0.4 and 1.6 mg/cm3 and  correction  of  biases
     attributable to substrates as needed.

     Classify a paint sample as positive if the first 15-second
     (nominal)  K-shell XRF reading (substrate corrected as
     appropriate)  taken on paint is 1.6 mg/cm2  or  greater,  as
     negative if the reading is 0.4 mg/cm2 or less;  otherwise  the
     paint sample is classified as inconclusive.   Inconclusive
     readings are to be resolved by laboratory analysis.   Using
     the ICP spectroscopic analysis of the paint sample to
     determine whether the lead level was actually greater than
     or equal to 1.0 mg/cm2,  the  overall  false  positive,  false
     negative and inconclusive rates for the K-shell XRF
     instruments are shown in Table 7.  With the exception of the
     XK-3 false positive rate,  all error rates  were below 10%.
     The false positive rate for the XK-3 was dramatically
     reduced by either method of substrate correction.   For each
     substrate type, most error rates were still below 10%.  The
     exceptions were MAP-3 false negative rates on concrete and
     plaster,  the Microlead I false positive rate on wood, and
     the XK-3 false negative rate on metal.  It is important to
     remember that these classification results apply strictly
     only to the set of samples and instruments in this study.
     Classification results for a different set of samples or
     instruments could be different.
                               23

-------
Table 7.    False Positive, False Negative and inconclusive Percentages for
          K-Shell XRF Instruments,  Based on One Nominal 15-Second Reading
          With an INCONCLUSIVE RANGE OF 0.4 - 1.6 mg/cm2 (1.0 mg/cm2
          Threshold).
INSTRUMENT
Lead Analyzer K-shell
MAP- 3 K-shell
Microlead I
XK-3
XK-3 (Average Corrected)
XK-3 (Control Block Corrected)
FALSE POSITIVE
PERCENTAGE
0.5%
2.3%
7.5%
22%
2.3%
3.5%
FALSE NEGATIVE
PERCENTAGE
1.4%
3.7%
1.1%
1.1%
4.2%
4.0%
INCONCLUSIVE
PERCENTAGE
18%
23%
30%
35%
25%
25%
9.   When the  laboratory confirmation range was narrowed to 0.7
     to 1.3 mg/cm2, thereby substantially reducing the
     inconclusive  percentages,  the K-shell instruments  continued
     to reliably classify paint samples in this study.

     Table 8 shows similar data to Table 7 with the  narrower
     inconclusive  range.   Results of the Microlead I and the XK-3
     both needed substrate correction to achieve  satisfactory
     false positive rates.  For each substrate type,  error rates
     were generally below 10%.   The exceptions were  MAP-3 false
     negative  rates on concrete and plaster, the  Microlead I
     false negative rate on concrete, XK-3 false  negative rates
     on metal  and  plaster, and the XK-3 false positive  rate on
     concrete.  Inconclusive  percentages are reduced by at least
     50% for all XRF instruments compared to the  inconclusive
     percentages when classifying paint samples using the 0.4 -
     1.6 mg/cm2 inconclusive  range.
                                 24

-------
Table 8.    False Positive,  False Negative and Inconclusive Percentages for
          K-Shell XRF Instruments, Based on One Nominal 15-Second Reading
          With an INCONCLUSIVE RANGE OF 0.7 - 1.3 mg/cma (1.0 mg/cm2
          Threshold).
INSTRUMENT
Lead Analyzer K-shell
MAP-3 K-shell
Microlead I
Microlead I (Average Corrected)
XK-3
XK-3 (Average Corrected)
XK-3 (Control Block Corrected)
FALSE POSITIVE
PERCENTAGE
1.2%
4.1%
12%
4.9%
30%
5.5%
6.5%
FALSE NEGATIVE
PERCENTAGE
2.7%
4.6%
2.1%
5.3%
1.7%
6.6%
6.8%
INCONCLUSIVE
PERCENTAGE
6.0%
11%
15%
12%
17%
12%
12%
10.  Without a laboratory confimation range,  the K-shell
     instruments' performance differed when classifying paint
     samples in this study using  the  federal threshold of 1.0
     mg/cm2.

     Based on readings obtained using the K-shell instruments,
     paint samples were classified as positive if the XRF reading
     was 1.0 mg/cm2 or higher and negative  otherwise.   There was
     no inconclusive range.  False positive and false negative
     rates for the K-shell instruments'  results are shown in
     Table 9.  As expected, these  rates are higher than when
     inconclusive ranges were used, but still no greater than 11%
     overall when substrate correction methods are employed as
     needed.  False positive and  false negative rates for
     readings on particular substrates were substantially higher
     than the overall rates as exemplified by the following
     ranges.  For all of the K-shell  instruments, the lowest
     false positive or false negative rate on a particular
     substrate was less than 2.0%.  However,  on the high end, the
     Lead Analyzer's false negative rate on concrete was 11%, the
     MAP-3's false negative rate on concrete was 24%,  the
     Microlead I's false positive  rate on wood was 26%, and the
     XK-3's false positive rate on concrete was 66%.
                                25

-------
Table 9.    False Positive and False Negative Percentages for K-Shell XRF
           Instruments, Based on One Nominal 15-Second Reading With NO
           INCONCLUSIVE RANGE  (1.0 mg/cm2 Threshold).
INSTRUMENT
Lead Analyzer K-shell
MAP- 3 K-shell
Microlead I
Microlead I (Average Corrected)
XK-3
XK-3 (Average Corrected)
XK-3 (Control Block Corrected)
FALSE POSITIVE
PERCENTAGE
3.1%
8.0%
20%
9%
40%
11%
11%
FALSE NEGATIVE
PERCENTAGE
5.9%
8.3%
3.8%
9%
3.6%
10%
11%
11.  With the exception of the XL, L-shell instruments  performed
     poorly when classifying paint using the 1.0 mg/cm2
     threshold, because of a high rate  of false negative results.

     Table 10 shows  false positive,  false negative and
     inconclusive percentages for tests using L-shell instruments
     and an inconclusive range of 0.4 to 1.6 mg/cm2.   With  the
     exception of the  XL, the false  negative rates for  the
     L-shell instruments' results were  very high, due to the
     large negative  biases shown in  the results using these
     instruments.  False positive rates were very low for all
     L-shell instruments' results.

Table 10.   False Positive, False Negative  and Inconclusive Percentages for
           L-Shell XRF  Instruments, Based  on One Nominal 15-Second Reading
           with an INCONCLUSIVE RANGE OF 0.4 - 1.6 mg/cm2 (1.0 mg/cm2
           Threshold).
INSTRUMENT
Lead Analyzer L-shell
MAP-3 L-shell
XL
X-MET 880
FALSE POSITIVE
PERCENTAGE
0.0%
0.0%
0.1%
0.0%
FALSE NEGATIVE
PERCENTAGE
66%
37%
11%
66%
INCONCLUSIVE
PERCENTAGE
6%
12%
15%
7%
                                  26

-------
12.   Although the XL prototype had a lower rate of false negative
     results than the other L-shell instruments,  it still
     exhibited false negative results at very high lead levels.

     As shown in Table 10,  the XL had a false negative rate of
     approximately 11% and a false positive rate of 0.1% using an
     inconclusive range of 0.4 to 1.6 mg/cm2.   However,  of  the 38
     instances where the ICP measurement exceeded 10 mg/cm2,  2 of
     the XL readings were below 0.4 mg/cm2  and  one  was  equal  to
     0.4 mg/cm2.   In all  3  cases,  a paint  sample  with  an ICP
     result above 10 mg/cm2  was classified  as negative  for  lead-
     based paint.  With a narrower inconclusive range of 0.7 to
     1.3 mg/cm2,  the XL had  an overall  false  negative  rate  of
     24.1% and a 0.2% false positive rate.   Classifying the XL
     results without an inconclusive range yielded a 41.8% false
     negative rate and a 0.5% false positive rate.

13.   Generally, a single XRF reading at one point of an
     architectural component provided almost as much accuracy a*
     an average of three XRF readings at the same point.

     When paint samples were classified as positive for XRF
     results 1.6 mg/cm2 or greater,  negative  for  XRF results  0.4
     mg/cm2  or less,  or inconclusive, otherwise,  and the results
     were compared to the lead level obtained from the ICP
     spectroscopic analysis of the paint sample,  there was very
     little difference in the false positive and false negative
     rates for the average of three 15-second readings versus a
     single 15-second reading.  The small improvement in
     classification accuracy did not justify the additional time
     and expense of taking three readings at the same point.
     This remained true when substrate  correction and different
     inconclusive ranges were employed.

     A similar conclusion was reached when the precision of the
     average of three 15-second readings,  as measured by its
     standard deviation, was compared to that of a single
     reading.  If the three readings were statistically
     independent, one would expect the  standard deviation of the
     average to be 58% of the standard deviation of a single
     reading.  However, it was found that the standard deviation
     of the average was much greater than this.  For L-shell
     instruments, the standard deviation of the average was
     typically at least 95% of the standard deviation of a single
     reading.  For K-shell instruments, the standard deviation of
     the average was typically between 76% and 93% of the
     standard deviation of a single reading.

                               27

-------
     There are two reasons why taking the average of  three
     readings did not produce the expected gains  in precision.
     First,  with the exception of the MAP-3 K-shell instrument's
     readings, successive readings at the same point  were
     positively correlated.   Thus,  the reduction  in variability
     from averaging repeat readings was less than would be
     achieved if successive readings had been statistically
     independent.   The second reason why the average  produced a
     smaller reduction in variability than expected is that
     repeated readings reduced only the component of  variability
     due solely to the performance of the instrument.   The study
     data clearly demonstrated that there were additional sources
     of variability that were generally at least  as large as the
     component due to the performance of the XRF  instrument.
     Taking repeated readings does not reduce the variability due
     to these other sources.   The additional variability was due
     to location-specific factors,  such as paint  and  substrate
     composition.
3.3  CHEMICAL TEST KITS

1.    None of the test kits used in this study demonstrated low
     rates of both false positive and false negative results when
     compared to laboratory analytical results using the federal
     thresholds,  1.0 mg/cm2  and 0.5%.

     Table 11 shows overall false positive and false negative
     rates for the test kits compared to laboratory analytical
     results using the 1.0 mg/cm2  threshold.   Table  12  shows  the
     corresponding rates for the 0.5% threshold.   Rates for the
     Lead Alert kits exclude results of tests on painted plaster
     substrates since the manufacturer does not recommend use of
     these kits on plaster.   For the 1.0 mg/cm2 threshold,  State
     Sodium Sulfide and LeadCheck had low false negative rates,
     but high false positive rates.   Lead Alert:   Sanding had a
     low false positive rate,  but a high false negative rate.
     The other three kits tested,  Lead Zone, Lead Detective,  and
     Lead Alert:   Coring, had moderate to high rates of both
     false positive and false negative results.   For the 0.5%
     threshold, State Sodium Sulfide had a low false negative
     rate and Lead Alert:  Sanding had a low false positive rate.
     False negative rates for LeadCheck and false positive rates
     for Lead Alert:  Coring were slightly above 10%.  Lead Zone
     and Lead Detective had high rates of both false positive and
     false negative results.  As was pointed out for XRFs, it is
     important to remember that these classification results

                                28

-------
Table 11.   Overall False Positive and False Negative Rates for Test Kits
          Compared to Laboratory Analytical Results Using the 1.0 mg/cm2
          Threshold.
TEST KIT
LeadCheck
Lead Alert : Coring
Lead Alert : Sanding
Lead Detective
Lead Zone
State Sodium Sulfide
FALSE POSITIVE
PERCENTAGE
46%
15%
9%
36%
28%
65%
FALSE NEGATIVE
PERCENTAGE
6%
24%
53%
23%
14%
1%
Table 12.   Overall False Positive and False Negative Rates for Test Kits
          Compared to Laboratory Analytical Results Using the 0.5%
          Threshold.
TEST KIT
LeadCheck .
Lead Alert : Coring
Lead Alert : Sanding
Lead Detective
Lead Zone
State Sodium Sulfide
FALSE POSITIVE
PERCENTAGE
42%
11%
10%
32%
25%
62%
FALSE NEGATIVE
PERCENTAGE
11%
36%
67%
27%
25%
6%
     apply strictly only to the set  of  samples and kits  in  this
     study.   Classification results  for a different set  of
     samples  or  kits could be different.

2.    The substrate underlying the paint sometimes affected  false
     positive and false negative rates  for test kits.

     LeadCheck:   For both federal thresholds,  the false  positive
     rate on  drywall was considerably lower than on the  other
     five substrates.   False negative rates in mg/cm2 on concrete
     and plaster were higher than on the  other substrates.   For
     percent  by  weight,  false negative  rates were higher on
     concrete, drywall,  metal, and plaster than on brick and
     wood.  Some of these differences in  false negative  rates may
     be caused by sulfates found in  plaster dust, gypsum and
                                 29

-------
stucco.  The kit includes a confirmation procedure to guard
against false negative results caused by sulfates.

Lead Alert:  Coring:  The manufacturer states that this kit
is prone to negative interferences from gypsum and plaster
dust.  High false negative rates were observed on plaster
and drywall for percent lead by weight measurements and on
plaster for mg/cm2  measurements.   However,  the  sample size
for drywall was very small.  False negative rates on brick
were much lower than on the other substrates for both types
of measurements.  For mg/cm2 measurements,  false  positive
rates were lowest on plaster and drywall substrates, and
highest on brick.  For percent lead by weight measurements,
false positive rates were lowest on drywall, plaster, and
wood substrates, and highest on brick.

Lead Alert:  Sanding:  This kit had a very similar pattern
to Lead Alert: Coring with high false negative rates on
plaster and drywall, and the highest false positive rate on
brick.

Lead Detective:  The manufacturer does not recommend use on
metal, but does recommend application on wood,  drywall, and
plaster.  False positive rates were consistent for both
types of measurements on all substrates except brick, which
had a higher false positive rate.  False negative rates were
lowest on wood and highest on brick and concrete substrates.
(Results were observed showing that drywall had the highest
false negative rate for percent lead by weight units, but
the sample size was very small.)  Thus, this kit did not
perform much better on wood, plaster, and drywall than on
metal so that the manufacturer's recommendations were not
borne out by the study data.

Lead Zone:  The manufacturer's instructions only mention
testing on wood and metal.  False positive rates were the
same on all substrates for both types of measurements.
False negative rates were lower on brick, wood, and
concrete, and higher on the other substrates.  The false
negative rate on metal was the highest of all  substrates
using percent lead by weight measurements.  The
manufacturer's instructions do not include mention of using
this kit on substrates where it performed similarly  to its
performance on wood, but do mention its use on metal, where
its false negative rate was substantially larger than its
false negative rate on wood.
                           30

-------
     State Sodium Sulfide:  The instructions contain a caution
     not to test directly on metal.  For metal substrates, a
     paint chip can be removed and tested separate from the
     substrate.  This kit had very high false positive rates for
     both types of measurements on all substrates except drywall.
     False negative rates were low on all substrates for mg/cm2
     measurements.  For percent lead by weight measurements, this
     kit had higher false negative rates on metal, plaster, and
     drywall than on the other substrates.

3.    The probability of a positive classification when the
     sample's lead level was equal to the federal thresholds
     varied depending on the kit and substrate.  High levels of
     lead would not always be detected using test kits alone.

     Table 13 shows the probability of a positive result using a
     test kit on paint with a lead level equal to the 1.0 mg/cm2
     federal threshold, as estimated from the statistical model
     developed in this study.  Table 14 is the companion table
     for the other federal threshold of 0.5% by weight.
     Considerable variation among results for each kit and each
     substrate is seen in the tables.

     High levels of lead were not always detected with complete
     certainty using test kits.  The statistical model estimated
     the limiting probability of a positive test kit result at
     high levels of lead using the laboratory ICP spectroscopic
     results reported in mg/cm2 units.   In a number of  cases,  the
     limiting probability was much lower than the desired value
     of 100%.  This occurred for four of the six kits:   Lead
     Alert:  Coring on metal; Lead Alert:  Sanding on concrete,
     metal, and wood; Lead Detective on concrete, metal, and
     plaster; and Lead Zone on plaster.

Table 13.   Probability of a Positive Test Kit Result at 1.0 mg/cm2 Lead.
TEST KIT
LeadCheck
Lead Alert : Coring
Lead Alert : Sanding
Lead Detective
Lead Zone
State Sodium Sulfide
Brick
0.95
0.93
N/A
0.81
0.82
0.99
Concrete
0.69
0.27
0.50
0.58
0.27
0.95
Drywall
0.49
N/A
N/A
0.34
0.64
0.68
Metal
0.93
0.66
0.39
0.74
0.59
0.94
Plaster
0.69
N/A
N/A
0.51
0.55
0.95
Wood
0.91
0.57
0.02
0.78
0.80
0.95
                                31

-------
Table 14.   Probability of a Positive Test Kit Result at 0.5% Lead.
TEST KIT
LeadCheck
Lead Alert : Coring
Lead Alert : Sanding
Lead Detective
Lead Zone
State Sodium Sulfide
Brick
0.95
0.73
N/A
0.80
0.81
0.998
Concrete
0.68
0.23
0.13
0.55
0.51
0.93
Drywall
0.48
N/A
N/A
0.31
0.55
0.59
Metal
0.62
0.26
0.05
0.43
0.19
0.83
Plaster
0.68
N/A
N/A
0.46
0.53
0.91
Wood
0.83
0.28
0.03
0.58
0.62
0.87
4.    The lead level at which there was a 50% chance of the
     occurrence of a positive test kit result varied depending on
     the kit and substrate.  In many cases, positive results
     occurred even when paint with very low lead levels was
     tested.

     Table 15 shows the lead level in mg/cm2 at  which each kit
     had an estimated 50% probability of a positive result, by
     substrate.  Table 16 is the companion table in percent lead
     by weight measurements.  There was significant variation in
     50% probability levels for different kits used on the same
     substrate.  There was also significant variation in the 50%
     probability levels for the same kit used on different
     substrates.  One exception, the State Sodium Sulfide kit,
     reached a 50% probability of a positive result at low lead
     levels on all substrates for both types of measurements.

     The statistical model used to analyze the test kit data also
     provided estimates of the limiting probability of a positive
     result as the lead level in the paint sample approached zero
     using the laboratory ICP spectroscopic results reported in
     mg/cm2 units.   It  is  desirable that  this limiting
     probability be zero; otherwise, the kit will produce some
     positive results even for paint samples with very low lead
     levels.  However,  every kit exhibited a non-zero limiting
     probability of a positive result on at least one substrate.
     This occurred on metal substrates for all six kits.  With
     the sodium sulfide kits, Lead Detective and State Sodium
     Sulfide, most substrates had a non-zero limiting probability
     of a positive result.  For the other 4 test kits, limiting
     probabilities of a positive result equaled or exceeded 20%
     for LeadCheck on metal and plaster,  Lead Alert:  Coring on
     brick, and Lead Zone on concrete.  For LeadCheck, Lead
                                32

-------
Table 15.
Lead Level in mg/cm2 at Which There is a 50% Probability of a
Positive Test Kit Result.
TEST KIT
LeadCheck
Lead Alert : Coring
Lead Alert : Sanding
Lead Detective
Lead Zone
State Sodium Sulfide
Brick
0.02
0.33
N/A
0.05
0.08
0.01
Concrete
0.19
1.84
N/A
0.60
1.38
0.01
Drywall
1.14
N/A
N/A
N/A
0.31
0.08
Metal
0.34
0.65
N/A
0.55
0.82
0.08
Plaster
0.13
N/A
N/A
0.98
0.71
0.02
Wood
0.03
0.77
1.24
0.20
0.15
0.04
Table 16.   Lead Level  in Percent Lead by Weight at Which There is a 50%
          Probability of a Positive Test Kit Result.
TEST KIT
LeadCheck
Lead Alert : Coring
Lead Alert : Sanding
Lead Detective
Lead Zone
State Sodium Sulfide
Brick
0.02
0.13
N/A
0.01
0.07
0.01
Concrete
0.16
1.14
0.88
0.33
0.49
0.01
Drywall
0.56
N/A
N/A
N/A
0.35
0.13
Metal
0.32
1.09
N/A
0.63
1.03
0.08
Plaster
0.14
N/A
N/A
0.58
0.44
0.02
Wood
0.07
0.97
1.68
0.36
0.26
0.09
     Detective and State Sodium  Sulfide,  limiting probabilities
     for the wood substrate were positive.

3.4  PAINT CHIP SAMPLING AND ANALYSIS

1.    Lead levels in paint showed significant variation within
     individual architectural components  such as doors, walls,
     and baseboards.

     Duplicate paint samples were taken approximately 9 inches
     apart on the same component at  10% of  the test locations in
     the full study in Denver and Philadelphia.   Duplicate paint
     samples taken from the same component  were called duplicate
     pairs.  The estimated median ratio of  the larger to the
     smaller ICP spectroscopic result,  measured in mg/cm2,  for
     duplicate pairs was 1.6 in  Denver  and  1.3 in Philadelphia.
     The corresponding median ratios for  percent lead by weight
     units were 1.5 and 1.2.  The estimated 95th percentile for
     the ratio in mg/cm2 was 3.7 in Denver and 2.1  in
                                33

-------
     Philadelphia.   The  corresponding 95th percentile  ratios  for
     percent lead by weight  units  were 3.1 and 1.9.  There  was
     slightly greater variability  in lead levels  within
     architectural components  when measured in mg/cm2 than  in
     percent lead by weight.   The  extent  to which greater
     variability would be observed between samples taken farther
     apart than 9 inches is  not  addressed by the  study data.

     Variability in duplicate  samples could result in  different
     classification of paint depending on which member of the
     pair was compared to the  federal threshold.   If the lead
     level of a paint sample was equal to or greater than the
     federal threshold,  it was classified as positive  for lead-
     based paint.  Likewise, if  the sample was less than the
     federal threshold,  then it  was classified as negative.  Of
     128 total duplicate pairs in  the study,  10 (8%) had
     different classifications,  one sample positive and the other
     negative for lead,  compared to the 1.0 mg/cm2  threshold,
     while 8 (6%) had different  classifications compared to the
     0.5% threshold.

     Spatial variation in lead levels within single architectural
     components complicated  the  statistical analysis of XRF and
     test kit performance data in  the study.   Complex  statistical
     models were needed  to account for the impact of spatial
     variation on estimates  of XRF measurement bias and standard
     deviation.  Spatial variation had a smaller  impact on the
     test kit data analysis.

2.    Variation between members of  laboratory duplicate subsample
     pairs was much smaller  than variation between members of
     duplicate samples obtained in the field.

     Laboratory analytical measurement error for  ICP
     spectroscopic analysis  of 2 x 2 inch paint chip samples,
     including homogenization, subsampling and instrumental
     error, can be quantified  using the ratio of  the larger to
     the smaller ICP measurement for a pair of subsamples of the
     same sample.  The estimated median for this  error ratio was
     1.13 for samples taken  from smooth substrates with no
     unusual difficulty in paint removal.  The estimated 95th
     percentile for the  error  ratio was 1.4.   These ratios apply
     to laboratory results reported in both in mg/cm2  and percent
     lead by weight units.
                                34

-------
Laboratory measurement error was approximately constant
across metal, wood, plaster, and drywall substrates, across
cities, and across samples within a substrate or within a
city.  For samples taken on rough substrates such as brick
or concrete,  total laboratory analytical measurement error
was higher:  the estimated median ratio was 1.2 and the
estimated 95th percentile ratio was 1.8.

Only two laboratory duplicate pairs out of a total of 171
(1%)  had different classifications, one of the pair positive
and one negative, with respect to the 1.0 mg/cm2  threshold.
For the 0.5% threshold, three subsample pairs out of 171
(2%)  had different classifications.
                           35

-------
50272-101
  REPORT DOCUMENTATION
              PAGE
1  REPORT NO.
        EPA 747-R-95-002a
3. Recipient's Accession No.
 4. Title and Subtitle
         A FIELD TEST OF LEAD-BASED PAINT TESTING TECHNOLOGIES:
         SUMMARY REPORT
                                                                                       5 Report Date
                                                                                               May 1995
                                                     6.
 7. Author(s)
                                                     8. Performing Organization Rept. No
         Cox, D.C.;  Dewalt, F.G.; Haugen, M.M.; Koyak, R.A.; Schmehl, R.L.
 9. Performing Organization Name and Address

         QuanTech, Inc.
         1911 North Fort Myer Drive, Suite 1000
         Rosslyn, Virginia  22209
                                                     10. ProjectH'ask/Work Unit No.
                      Midwest Research Institute
                 &    425 Volker Boulevard
                      Kansas City, Missouri  64110
11. Contract (C) or Grant (G) No.

        68-DO-0137
 12. Sponsoring Organization Name and Address

         U.S. Environmental Protection Agency
         Office of Pollution, Pesticides and Toxic Substances
         Washington, DC  20460
                                                     13. Type of Report & Period Covered

                                                             Summary Report
                                                     14.
 15. Supplementary Notes
         In addition to the authors listed above, the following key staff members were major contributors to the
         study:  Paul Constant, Donna Nichols, Jack Balsinger, Nancy Friederich, and John Jones of Midwest
         Research Institute; and Connie Reese of QuanTech.
 16. Abstract (Limit:  200 words)

         A large field study was conducted to compare three methods commonly used to test for lead in paint:
         portable X-ray fluorescence (XRF) instruments, lead paint test kits, and laboratory analysis of paint
         chip samples.  Laboratory analysis is considered to be the most accurate of the three methods and was
         the benchmark for comparisons. The study concludes that use of K-shell XRFs, with laboratory
         confirmation of readings designated as inconclusive and with correction of substrate biases where
         appropriate, is an acceptable way to classify painted architectural components versus the federal
         threshold of 1.0 mg/cm1.  The study concludes that test kits should not be used to test for lead in  paint.
         No test kit in the study achieved low rates of both false positive and false negative results.  Some  kits
         yielded a positive result at low levels of lead. Other kits were prone to a negative result when lead in
         paint was above the federal thresholds of 1.0 mg/cm2 and 0.5% by weight.
 17. Document Analysis a. Descriptors

         Lead-based paint, lead-based paint testing, comparability study, field evaluation, recommendations for testing
         for lead in paint


         b. Identifiers/Open-Ended Terms

         X-ray fluorescence instrument, XRF instrument, portable XRF, lead paint test kit, chemical test kit, test kit,
         inductively coupled plasma-atomic emission spectrometry, ICP-AES, ICP


         c COSATI Field/Group
 18 Availability Statement
                                    19 Security Class (This Report)
                                            Unclassified
                                                                      20 Security Class (This Page)
                                                                              Unclassified
            21 No of Pages
                   53
                                                                                                   22. Price
(See ANSI-Z39.18)
                                                              OPTIONAL FORM 272 (4-77)
                                                                     (Formerly NTIS-35)
                                                                 Department of Commerce

-------