1445 ROSS AVENUE
DW.LAS, TEXAS 7520?
                EPA 560/5-90-002
                  December 1989
     TRANSMISSION ELECTRON MICROSCOPY ASBESTOS LABORATORIES
                        QUALITY ASSURANCE GUIDELINES
                                  Prepared by:
      Chesson Consulting
      Suite 601
      1717 Massachusetts Ave., N.W.
      Washington, D.C. 20036
Chatfield Technical
Consulting, Ltd.
2071 Dickson Road
Mississauga, Ontario
Canada L5B 1Y8
                                      and
                                  BATTELLE
                                Arlington Office
                              2101 Wilson Boulevard
                             Arlington, Virginia 22201
                             Contract No. 68-02-4294
                           Exposure Evaluation Division
                            Office of Toxic Substances
                      Office of Pesticides and Toxic Substances
                       U.S. Environmental Protection Agency
                             Washington, D.C. 20460

-------
                         AUTHORS AND CONTRIBUTORS
      This document was prepared under contract to the U.S. Environmental Protection Agency.
The primary authors were Jean Chesson of Chesson Consulting and Eric Chatfield of Chatfield
Technical Consulting, Ltd. Jerry Rench and Ted Berner of Battelle were
responsible for contract administration and coordinated document production.  Betsy Dutrow
(Task Manager), Brad Schultz (Co-Task Manager), Jay Glatz (Co-Task Manager), Edie Sterrett
(Project Officer), Mike Beard, Joe Breen and Cindy Stroup participated in early discussions and
subsequent reviews and revisions.
                             ACKNOWLEDGEMENTS
      We acknowledge the support and critical evaluation provided  by Eric Steel, Shirley
Turner, and John Donaldson of the National Institute of Standards and Technology.  Their
comments and those of the peer review panel improved the document.  Gordon Cash assisted
in the cover design.

-------
                                      PREFACE
On October 22, 1986, the Asbestos Hazard Emergency Response Act of 1986 (AHERA) was
signed into law. AHERA required the U.S. Environmental Protection Agency (EPA) to develop
regulations for the control of asbestos-containing materials in schools and the National Institute
of Standards and Technology (NIST) to formulate  a laboratory accreditation program for
transmission electron microscopy (TEM) laboratories involved in the analysis of air samples for
asbestos.

EPA's regulation for schools included new analytical protocols for TEM analysis of air samples,
the Interim Transmission Electron Microscopy Analytical Methodologies (40 CFR 763, Appendix
A to Subpart E).  NIST, in its development of a National Voluntary Laboratory Accreditation
Program (NVLAP) for TEM laboratories, determined the test method for accreditation would
be EPA's methodology.

NIST accredits laboratories in the NVLAP based on a multistep process. First, laboratories are
evaluated by an expert technical assessor in an on-site assessment.  The assessor reviews all
aspects of the laboratory's operation.  In addition, the assessor examines the quality manual and
quality assurance system instituted by the laboratory.  The quality assurance system and its
associated procedures are critical components of routine laboratory operation.  Second, NIST
issues proficiency testing samples to the laboratory. The laboratory analyzes the samples and
returns the results to NIST. Finally, based on the results of the site visit and the proficiency
testing, NIST determines whether the laboratory meets all requirements for accreditation.

These EPA guidelines have been produced to assist TEM laboratories in developing and refining
their QA programs.  The guidelines themselves do not constitute a QA manual.  Instead, they
provide background information on the  role of QA and suggestions  on its implementation.
Following these guidelines does not guarantee accreditation by NIST.  Quality assurance is an
ongoing process specific to each laboratory. Therefore, laboratories should continually endeavor
to extend and improve their programs beyond the suggestions in these guidelines.
                                           11

-------
                           TABLE OF CONTENTS
1.0    INTRODUCTION

2.0    HOW TO USE THESE GUIDELINES

3.0    BACKGROUND
      3.1 WHAT IS QUALITY ASSURANCE?
      3.2 QUALITY ASSURANCE FOR TEM ANALYSIS
           3.2.1 Bias
           3.2.2 Precision
      3.3 OBJECTIVES OF A QUALITY SYSTEM

4.0  ESTABLISHING AND MAINTAINING A
      QUALITY SYSTEM
      4.1 QUALITY MANUAL CONCEPTS
      4.2 LABORATORY METHODOLOGY
      4.3 PROCEDURES FOR  RECEIVING AIR
           SAMPLES
           4.3.1 Documentation
           4.3.2 Acceptance/Rejection
                   Criteria
      4.4 SAMPLE CUSTODY  CONSIDERATIONS
      4.5 MEASUREMENT OF GRID OPENING
           DIMENSIONS
      4.6 CONTROL OF CONTAMINATION
           4.6.1 General Considerations
           4.6.2 Laboratory Blank
                   Measurements
           4.6.3 Filter Lot Blanks
           4.6.4 Field Blanks
           4.6.5 Reagent, Equipment and
                   Material Blanks
           4.6.6 What to Do When
                   Contamination is Detected
      4.7 QUALITY ASSURANCE REANALYSIS
           OF SAMPLES
           4.7.1 Types of Reanalysis
           4.7.2 Determination of Operator
                   Precision
           4.7.3 Determination of Laboratory
                   Precision
           4.7.4 Determination of Intra-
                   Laboratory Method Precision
           4.7.5  Determination of Inter-
                   Laboratory Method Precision
Pages Revision Date
2
1
7
1
3
2
2
1
38
2
1
1
2
1
1
2
4
2
2
1
2
1
1
6
3
1
2
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
                                    111

-------
                        TABLE OF CONTENTS
                              (Continued)
                                                   Pages Revision Date
      4.7.6  Analysis of Proficiency
              Testing Samples
      4.7.7  Interlaboratory Sample
              Analyses
      4.7.8  How to Correct Counting
              Discrepancies

4.8  INTERPRETATION OF RECOUNT DATA
4.9  USE OF STANDARD SAMPLES
      4.9.1  Calibration of Plasma
              Asher Performance
      4.9.2  Calibration of TEM
              Magnification
      4.9.3  Cab'bration of the
              Diffraction Camera Constant
      4.9.4  Calibration of Electron
              Beam Dose
      4.9.5  Calibration of Operators
              Using SRM 1876a
      4.9.6  Quality of TEM Specimen
              Preparation
      4.9.7  Quality of Electron
              Diffraction Patterns
      4.9.8  Calibration of Electron
              Beam Spot Size and EDXA
              System
      4.9.9  Preparation and Use of
              In-House Reference Standards
      4.9.10 Use of Samples and Proficiency
              Testing Samples as Operator
              Training Materials
4.10  DATA RECORDING
      4.10.1 Sample Identification
      4.10.2 Filter Type and Specimen
               Preparation
      4.10.3 Grid and Grid Opening
               Selection
      4.10.4 Fiber Identification
      4.10.5 Measurement of the
               Dimensions of Asbestos
               Structures
2
2
1
9
7
2
1
1
1
2
2
2
0
0
0
0
0
0
0
0
0
0
0
0
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
1    0  12/27/89

1    0  12/27/89
2
3
1
2
2
1
0
0
0
0
0
0
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
12/27/89
     0  12/27/89
                                   IV

-------
                            TABLE OF CONTENTS
                                 (Continued)
            4.10.6 Labeling of Micrographs
                    and EDXA Spectra
            4.10.7 Reporting of Grid
                    Opening Dimensions
            4.10.8 Structure of Sample
                    Data File
            4.10.9 Checking of Analytical
                    Data
      4.11 ELECTRON DIFFRACTION PATTERN
            MEASUREMENTS
      4.12 MEASUREMENTS ON EDXA SPECTRA
      4.13 CALCULATIONS AND REPORTING OF
            RESULTS
            4.13.1 Bias
            4.13.2 Precision
            4.13.3 Other Calculations
                                                     Pages  Revision Date
     0  12/27/89

     0  12/27/89
     0  12/27/89
2
2
5
1
2
2
3
0  12/27/89

0  12/27/89
0  12/27/89
0  12/27/89
0  12/27/89
0  12/27/89
0  12/27/89
0  12/27/89
5.0    REFERENCES
     0  12/27/89

-------
                                  LIST OF FIGURES
Figure 3-1.   Examples of different types of repeated analyses that  may be used to assess
             precision.   (A)  Measurement of interlaboratory variability  by  having  two
             laboratories  analyze  the  same sample. (B) Measurement of intralaboratory
             variability  by making two  separate  preparations of the  same  sample.   (C)
             Measurement of between analyst variability by having two analysts count the same
             set of grid openings. Note that the subsampling step contributes to the variability
             in (A) and (B), but not in (C).  Subsampling places an upper limit on the level of
             precision that can be  achieved.

Figure 4-l(a). Percentiles for the calculated coefficient of variation as a function of the mean
             number of asbestos  structures present for different values of the number  of
             repeated analyses (n  = 2)

Figure 4-l(b). Percentiles for the calculated coefficient of variation as a function of the
             mean number of asbestos structures present for different values of the number of
             repeated analyses (n  = 3)

Figure 4-l(c). Percentiles for the calculated coefficient of variation as a function of the mean
             number of asbestos  structures present for different values of the number  of
             repeated analyses (n  = 4)

Figure 4-l(d). Percentiles for the calculated coefficient of variation as a function of the
             mean number  of asbestos structures present for different values of the number
             of repeated analyses (n = 5)

Figure 4-l(e). Percentiles for the calculated coefficient of variation as a function of the mean
             number of asbestos  structures present for different values of the number  of
             repeated analyses (n  = 6)
                                           VI

-------
                         LIST OF APPENDICES
APPENDIX A: SOURCES OF INFORMATION ON
             QUALITY ASSURANCE

APPENDIX B: EXAMPLES OF QUALITY ASSURANCE
             PROBLEMS AND CORRECTIVE
            ACTIONS
Pages Revision Date


 1   0  12/27/89



 4   0  12/27/89
                                Vll

-------
                                                                QA Guidelines for TEM
                                                                           Revision: 0
                                                               Date: December 27, 1989
                                                                   Section 1  Page 1 of 2
1.0    INTRODUCTION
             These EPA guidelines have been produced to assist laboratories in developing
and refining quality assurance programs for analysis of airborne asbestos by transmission electron
microscopy (TEM). While much of the information will be useful for TEM analysis of airborne
asbestos in general, the guidelines are directed specifically toward the analytical protocol required
by the U.S. Environmental Protection Agency (EPA) to determine completion of response actions
in schools (Appendix A to Subpart E, 40 CFR 763).
             Regulatory language tends  to be terse and, by necessity, lacks  discussion and
explanation.  These guidelines explain the purpose of a quality assurance program and provide
examples of how a laboratory might establish a program to achieve and maintain specific data
quality objectives.   The  information should assist laboratories  in the preparation of quality
manuals as required by the National Voluntary Laboratory Accreditation Program (NVLAP) for
TEM administered by  the  National  Institute of Standards and Technology  (NIST).  This
document is not concerned with the administrative aspects of quality manuals; these aspects are
clearly and adequately described in the NVLAP Airborne Asbestos Handbook (NIST 1989).
             The suggested procedures  are intended to be compatible with existing regulations,
but may not cover all requirements and are not necessarily the only way in which compliance may
be achieved.  Also, regulations are subject to revision over time. Readers  should regard  the
guidelines as supplementary information to be used in association with regulatory or other
requirements, not as a substitute for the requirements themselves.  The guidelines will be updated
as required to reflect new information and changing requirements.
             The remainder of this document is divided into two parts. Part  1 provides general
background information. Quality assurance is described from the perspective of the laboratory
and the laboratory's client. The statistical concepts of bias and precision are introduced and the
objectives of a  quality system stated.  A "How to Use these Guidelines" section is directed
particularly to the new laboratory establishing a QA program for the first time. Part 2 of the

-------
                                                                 QA Guidelines for TEM
                                                                            Revision: 0
                                                                Date:  December 27, 1989
                                                                    Section 1 Page 2 of 2

guidelines describes specific procedures designed to achieve quality assurance objectives.   The

organization of the section follows the progress of a sample through the analytical procedure,

beginning with general laboratory methodology  and receipt of samples, through control of

contamination, re-counting of certain samples,  and  use of standards,  to data recording,

calculation and reporting of results.

-------
                                                                QA Guidelines for TEM
                                                                            Revision: 0
                                                               Date:  December 27, 1989
                                                                  Section 2 Page 1 of 1

2.0  HOW TO USE THESE GUIDELINES

             As stated  earlier, these guidelines  alone  are not sufficient  to develop and

implement a QA program. For inexperienced laboratories the following steps are recommended:


       •     Read Section 3.0 (Background) of these guidelines and consult other  references
             to gain a general understanding of quality  assurance.  Information sources are
             listed in Appendix A.

       •     Read the remainder of the document to get  an  overview of items that need to be
             considered in a TEM laboratory.

       •     Collect all  the information relevant to the  types of analyses and mandatory or
             voluntary accreditation programs with which you expect to be  involved (e.g.,
             AHERA,  NVLAP,   State  and  local  regulations,   and  other accrediting
             organizations).

       •     Plan the sections of your quality manual to cover all the items you have identified
             in the first  three steps. Arrange the sections of the manual so that it will be easy
             to use in the lab. For example, you might organize it according to the analytical
             process (receipt of samples, preparation, analysis,  reporting)  or according to
             personnel responsibilities (laboratory manager, technicians, analysts).

       •     Prepare each section using material from these  guidelines and any other relevant
             sources.  Where appropriate, define data  quality objectives and  specify how
             compliance with them is to be measured on a section by section basis.  Start
             describing  what  you  do now.  If present procedures are  obviously inadequate,
             decide how to modify them.

       •     Organize the quality manual so that it can be revised easily. Add new procedures
             or modify  existing ones as  the need arises.  Ensure  that the manual is read,
             understood, and followed by everyone in  the laboratory.

-------
                                                                QA Guidelines for TEM
                                                                            Revision:  0
                                                               Date: December 27, 1989
                                                                   Section 3  Page 1 of 7
3.0  BACKGROUND
3.1 WHAT IS QUALITY ASSURANCE?
             Quality assurance (QA) is a system of activities intended to assure the producer
(the laboratory) or the user of reported data (the client) that the data are of known  quality.
Since almost  every activity taking place in a laboratory potentially affects the quality  of data
produced, QA  covers most aspects of laboratory practice including facilities, personnel, and
procedures. Quality control (QC) is the subset of QA activities that involves measuring quality
on an ongoing basis and taking corrective action when necessary.  For example, applying regular
proficiency tests to monitor performance falls under the definition of QC, whereas requiring that
personnel have  certain education qualifications, although part of QA, is not part of QC. These
guidelines tend to emphasize QC, not because  other aspects of QA are less important, but
because QC procedures often require greater explanation.
             QA is important to the laboratory because it establishes the quality of its product.
Superior quality attracts and keeps clients. In addition, a realistic statement of quality safeguards
the laboratory from misinterpretation by clients who might otherwise interpret individual results
as being more precise than they really are.  QA measurements have important legal implications
when the results of a particular  analysis or set of analyses are questioned.  A knowledgeable
client will ask for details of the QA program when selecting a laboratory. Measures of quality
are needed by  the client in order to design a monitoring program (the  quality of individual
measurements affects  the number of analyses required to achieve a particular objective) and to
interpret the  results.  Decisions made on the  basis of laboratory measurements may involve
considerable cost and have significant health impacts.  Therefore, it is critical for the client to be
adequately informed regarding data quality.

3.2  QUALITY ASSURANCE FOR TEM ANALYSIS OF AIRBORNE ASBESTOS
             The general principles of QA and QC apply to TEM analysis of airborne asbestos.
EPA has defined Good Laboratory Practice (GLP) Standards (40 CFR 792). The standards are
not specific for TEM laboratories because they address topics to be considered in any laboratory
QA program. According to EPA regulations, laboratories must address these requirements when
establishing a QA program. The GLP Standards may be obtained by contacting EPA at US EPA

-------
                                                                 QA Guidelines for TEM
                                                                             Revision:  0
                                                                Date: December 27, 1989
                                                                   Section 3 Page 2 of 7
(TS-799), 401 M Street, S.W., Washington, D.C. 20460 or by phoning (202) 554-1404.  Other
sources of general information on QA are listed in Appendix A.
             There are aspects of TEM analysis that differ from other types of chemical analyses
and require special consideration when formulating the details of a QA program.  Unlike many
analyses that involve a calibration curve approach,  TEM  is a  counting protocol  in which
individual particles  are identified and counted.  The data are essentially discrete, rather than
continuous. Furthermore, there is often a need to measure low asbestos concentrations for which
approximations using continuous statistical methods (for example, the normal distribution) are
least satisfactory.  The approaches suggested in  this document have been developed taking into
account these special features of TEM.  Since the AHERA TEM protocol is relatively new, there
is  a lack of data and experience on which to  base specific recommendations.   Approaches,
particularly those involving quantitative measures of quality, are expected to evolve as laboratories
experiment and find approaches that work best in their situations.  This document will be revised
periodically to reflect new developments. These  guidelines present one way of monitoring errors
in the laboratory. As NIST conducts proficiency testing rounds, additional information will be
gathered.  A reasonable system for monitoring and reporting errors will be developed using the
NIST data.
             TEM asbestos laboratories are responsible for  estimating the  concentration of
asbestos structures  per unit  surface area of a filter.  To do so, the laboratory makes two
measurements:  the  area of filter examined and the number of asbestos structures present in that
area.  The number  of asbestos structures  present is divided by the  area  of filter examined to
obtain an estimate of the number of asbestos structures per unit surface area of filter. Of course,
many laboratories  will  make additional  calculations such as  estimating airborne asbestos
concentration using data provided  by the client. The  quality of these additional  data depends
on factors beyond the control of the laboratory.  The factors include the accuracy with which the
air volume was measured, filter handling during sampling and transport, and filter contamination
from various sources before the filter reached the  laboratory.  Realization that the laboratory
analysis provides only an estimate of the number of asbestos structures per surface area of filter
simplifies the approach to laboratory QA and reduces confusion over certain analytical terms.
Blanks (filters through which no air has been drawn), regular air samples (filters through which

-------
                                                                  QA Guidelines for TEM
                                                                              Revision:  0
                                                                Date:  December 27, 1989
                                                                    Section 3 Page 3 of 7
air has been drawn), standard material, and filters of unknown origin are all treated in the  same
manner.   This argument is  not meant to suggest that  laboratories should lose sight of the
objective of their clients, which in most, but not all, cases is related to asbestos in the sampled
air. Rather, the intention is  to clarify the distinction between factors affecting quality that can
be controlled by the laboratory and those which cannot, so that efforts to improve the quality of
data can be properly focused.
              Having reduced TEM analysis to the generation of two quantities—the area of filter
examined and the number of asbestos structures present in that area~QA is directed to ensuring
that these quantities are sufficiently accurate  for their intended purpose.  Accuracy is usually
defined in terms  of bias and precision as discussed in the following sections.

3.2.1  Bias
              Bias is a systematic error that is either inherent in a method  or caused by  some
artifact or idiosyncrasy of the measurement system. Bias may be positive or negative, and several
kinds can exist concurrently. Examples of positive bias include

       •      use of contaminated reagents  that artificially inflate the number of asbestos
              structures counted; and

       •      inclusion of nonasbestos structures in the asbestos structure count because an
              analyst is insufficiently  skilled  in identifying electron diffraction patterns and
              consequently overestimates asbestos concentration.

Examples of negative bias include

       •      etching too much of the filter through improper calibration of the plasma etcher
              resulting in a loss of asbestos structures; and

       •      incorrectly rejecting some asbestos structures as too small because of inaccurate
              magnification  calibration and thereby underestimating asbestos concentration.

-------
                                                                 QA Guidelines for TEM
                                                                             Revision: 0
                                                                Date:  December 27, 1989
                                                                    Section 3  Page 4 of 7
              Although many sources of bias can be identified and steps taken to reduce their
contribution, the net bias associated with a final measurement may be difficult to assess because
there is sometimes no way to determine the "correct" answer. Assessment of net bias is often
by consensus.  For example, if five laboratories consistently obtain comparable results on
interlaboratory tests and a sixth laboratory consistently obtains lower results, the conclusion is
that  the sixth laboratory has a consistent negative bias, even though it is possible that the sixth
laboratory is least biased and the other five laboratories have a consistent positive bias. Standard
materials (See Section 4.9.10) can be used to measure bias for a restricted set of cases. Unlike
other types of chemical analyses, asbestos samples cannot be spiked easily in order to assess bias.
One source of bias that can be measured and reported in a relatively straightforward manner is
the positive bias due to contamination from sources other than  the sample received from the
client. Contamination may occur during filter manufacture, through the use of reagents that
contain asbestos, or through  handling in  the laboratory.   Contamination is measured with
laboratory blanks as discussed in Section 4.6.2.

3.2.2  Precision
              Precision is the  degree  of mutual agreement betweeen repeated independent
measurements of the same quantity.  It is concerned with the scatter of the results. Unlike bias,
precision can be readily measured.  Repeated  measurements on the same sample following the
complete preparation and analysis procedure each time provide an overall measure of precision.
Repeated measurements that are close  together indicate high precision.  The coefficient of
variation (also called the relative standard deviation) is one measure of precision that has proved
convenient for asbestos analysis. The coefficient of variation is defined as the standard deviation
divided by  the mean. A small coefficient of variation indicates high precision.
              Sources of variability,  and hence contributors to lowered precision, include
       •      variability introduced during the sample receiving  and preparation steps;

       •       subsampling  of  the  filter  (selection of  a  small portion  for microscopic
              examination); and

-------
                                                                 QA Guidelines for TEM
                                                                             Revision:  0
                                                                Date:  December 27, 1989
                                                                   Section 3 Page 5 of 7

       •     variability  introduced during  microscopic examination due  to  variability in
             equipment and analyst performance.

Each of these major sources can be broken down into smaller components,  such as variability
introduced during plasma etching, or day-to-day variability in the performance of a single analyst.
             The subsampling step in which a portion of filter  is selected for microscopic
examination has special significance because it places an upper bound on precision.  No matter
how meticulously procedures are followed, or how skilled the analyst, repeated measurements will
give different results because there are different numbers of asbestos structures on different
sections of the  filter.  Since the  laboratory has no  control over  the spatial distribution of
structures on the filter, variability introduced  through  subsampling  cannot be controlled by the
laboratory QA program.
             Laboratory QA includes procedures for minimizing variability and hence increasing
precision. Almost any action that defines and refines the procedures to be followed (educational
qualifications, written standard operating procedures, recording requirements) has the potential
to improve precision. Laboratory QC, as a subset of the QA program, is involved in the active
on-going assessment of precision and the taking of corrective action when precision falls below
a predetermined level.  Different types of repeated measurements are used to measure different
components of precision.   Figure  3-1 provides three  examples of  the many possibilities.  See
Section 4.7 for a more detailed discussion.  Note that whenever repeated analyses involve a
different subsampling of the filter surface,  the spatial distribution of structures will contribute
variability and provide an upper limit to the precision that can be obtained.

-------
  «a-
  o  ~,
F—*

B
•sis a
|*££
•I  e^
«  8
a  D
o
   o
     o
   ..  o n
                                               •^ •!-• J— S

                                              °^ s n«
                                              05 ^ ex c _5
                                              v >r> w 3 •"
                    (M
                       :r
                       ' £
                       '.:ct'
                                      CM
                                      CO
                                      111
                                      cc
                                      CO
                                              s
                                                 •§ 1
                                                 -i a
                                              CO

                                              W O .-
                                                  50
                                                 .   .

                                              mill
                                              S S 'C . 'C X!
                                              ^l S « ^ * iJ
                                               .S > .o > «
                                              SJ3

-------
                                                                 QA Guidelines for TEM
                                                                             Revision:  0
                                                                Date: December 27, 1989
                                                                    Section 3  Page 7 of 7
3.3  OBJECTIVES OF A QUALITY SYSTEM
              The quality system, the system of procedures that represents the QA program,
has three main objectives:

       1.      To achieve acceptable data quality;
       2.      To identify and correct problems that are having an undesirable effect on data
              quality; and
       3.      To characterize the quality of the data produced.
       The  fourth and final section of this document  discusses specific procedures that  are
designed to achieve these objectives.  To the extent  permitted by applicable regulations and
requirements, these procedures will vary from laboratory to laboratory depending on individual
circumstances. First, each laboratory must define its own acceptable data quality objectives. A
data quality objective is a statement of the quality of data needed for a particular purpose. For
example, a laboratory might specify that asbestos structure levels measured on blanks should not
exceed a certain  concentration, or an  analyst  should correctly identify  at least a certain
percentage of electron diffraction patterns presented  during monthly QA checks. Data quality
objectives may change with time as the laboratory gains experience, acquires  new equipment, or
makes other modifications.  Second, the emphasis on particular problem areas may shift as some
problems are eliminated (e.g., a supplier of contamination-free filters is located) and others are
introduced (e.g., an inexperienced analyst joins the staff).  Third, as stated earlier, although the
general principles have been widely used by laboratories for many types of analyses, the specific
quantitative methods suggested in this document for characterizing TEM data quality are largely
untested and are expected to be modified as experience is gained. The procedures described in
Section 4.0 should be adapted and incorporated into a Quality Manual written specifically for an
individual laboratory (See Section 4.1).

-------
                                                               QA Guidelines for TEM
                                                                           Revision:  0
                                                              Date:  December 27, 1989
                                                                 Section 4  Page 1 of 38
4.0  ESTABLISHING AND MAINTAINING A QUALITY SYSTEM
4.1  QUALITY MANUAL CONCEPTS
             There  is often some  confusion between the  terms "quality manual", "quality
assurance manual," and "quality assurance records." The term "quality manual" has been used
by NVLAP to denote a manual which contains not only a complete quality assurance manual, but
also information on various other topics such as staff qualifications and training programs. A
quality manual should be  an up-to-date  document which completely describes  all of the
procedures used in the TEM laboratory. In the quality manual, a new employee should be able
to find all of the procedures used by the laboratory from log-in of the sample to the issuing of
the final  report.  It should also contain examples of all of the forms used by the laboratory in
processing of a sample through the system.
             The quality assurance  section of the quality manual provides a description of the
various types of QA activities and samples, how frequently these activities will be performed,
along with details of the methods by which QA data are summarized, and a description of when
and how corrective actions are to be taken.
             Quality assurance  records are  the data  output from calibrations and from the
analyses  of  quality  assurance samples.  These data are more conveniently maintained  and
summarized in a  separate file, and  computer records  are often used.  The  quality assurance
records should contain details of the conclusions drawn from the data, and any corrective actions
which have been taken.  If the corrective actions involve a change of analytical procedure, the
new procedure should then appear as a revision of the  quality manual.
             The quality manual should be a document which progressively evolves along with
changes in the size and nature of the analytical laboratory. A quality manual which is appropriate
in content for a  one-person laboratory  will not be satisfactory for a large laboratory with
numerous TEM operators and several instruments. In setting up a manual, it must be  recognized
that frequent revisions and additions will be necessary,  and the system of page numbering and
format should be  chosen to accommodate these changes easily. The system of numbering used
in this document (Abbreviated  title, Revision  #,  Date, Section #, Page # of #) is useful.
Changed or  deleted sections should be archived, not destroyed.

-------
                                                               QA Guidelines for TEM
                                                                           Revision:  0
                                                              Date: December 27, 1989
                                                                 Section 4 Page 2 of 38
             The responsibility for writing and incorporating changes into the quality manual
must  be delegated to  a particular individual.   Periodic meetings  should  be  held  with
administrative and laboratory staff to determine what changes are necessary in the quality manual.
An organizational change of senior analytical staff, change in facilities, or the introduction of
changes in either the analytical methodology or the sample recording procedures are examples
of situations where revisions in the quality manual are warranted.

4.2 LABORATORY METHODOLOGY
             All of the analytical procedures used by the  laboratory for measurement of
asbestos structure concentrations in air by TEM should be fully documented in the  quality
manual; however, merely inserting a copy of the published methodology in the quality manual
is  not considered sufficient.  Each laboratory implements methods  in different ways because
different equipment and facilities are used.  Also,  analytical methodologies are  not  static;
experienced laboratories continue to incorporate significant improvements, which are  not part of
the published methodologies,  into their  procedures.  Where the analytical methodology  is
complex, as it is for TEM asbestos analysis, and the methodology is available as a published
document, the basic method may be referenced and bound separately, or it may be included in
full in the quality manual.  In particular, the method described in Appendix A to Subpart E, 40
CFR 763, October 30,1987, must be practiced by the laboratory, and a copy of this method must
be available in the laboratory.  However, the quality manual must contain detailed descriptions
of the analytical procedures exactly as practiced by the particular laboratory, in order to ensure
that all analysts in the laboratory have  the required  information enabling them to  work to a
uniform standard.

4.3 PROCEDURES FOR RECEIVING AIR SAMPLES
4.3.1  Documentation
             The method of handling documentation of air  samples upon receipt  by the
laboratory must be fully described in the quality manual. For example, procedures  and forms
should exist for recording of the method of shipping, the date and time of receipt,  how many
samples were in the shipment, and also for recording  of the  sample descriptions and data, such

-------
                                                               QA Guidelines for TEM
                                                                          Revision:  0
                                                             Date:  December 27, 1989
                                                                Section 4  Page 3 of 38
as air volumes, provided by the client. The action to be taken in the event of incomplete sample
documentation should be described in the quality manual.

4.3.2 Acceptance/Rejection Criteria
             The quality manual should contain a clear description of the criteria used by the
laboratory for  rejection of samples. There may be cause for rejection of a sample at various
points in the receiving and analysis process. For example, if air samples are received with bulk
samples in the same container, then in accordance with  the AHERA method, they must  be
rejected. As the sample is processed, it may be discovered that the filter had been damaged in
some way, or there may be evidence of air leakage during sampling. Any such defect which could
possibly compromise the final result must be noted, and the action taken must be recorded.

4.4  SAMPLE CUSTODY CONSIDERATIONS
             Documentation of the chain of custody for  samples is very  important. Chain of
custody involves recording the passage of samples from person to person and place to place and
identifying where and under whose responsibility samples are at any particular time. Any gap
in knowledge concerning the custody of samples  can be used to  discredit the  results of the
analyses. The procedures for recording chain of custody on all samples must be described, and
a copy of the chain of custody form used by the laboratory should be included in the manual.
Chain of custody records and the samples should be stored securely, and their locations should
be specified in the quality manual.  The policy of the laboratory concerning breaks in the chain
of custody and the actions required in this event must also be specified.

4.5  MEASUREMENT OF GRID OPENING DIMENSIONS
             The  area  of filter  examined is one of the two measurements for which the
laboratory is responsible. The number of asbestos structures present in that area will be divided
by the area to give the estimated concentration of  asbestos structures per surface area of filter.
The method used by the laboratory to measure the grid opening dimensions must be specified,
and the accuracy of the measurement method must be established.  One option is to measure a
sufficient number of grid openings from a batch of grids and to derive a mean dimension which

-------
                                                               QA Guidelines for TEM
                                                                           Revision:  0
                                                              Date:  December 27, 1989
                                                                 Section 4  Page 4 of 38
is then used for all samples.  Another option is to measure grid openings on each of the actual
TEM grids examined for a sample, and calculate an area for each sample.
             When the first  method  is  used,  the  area  estimate  affects the estimated
concentration in two ways.  If the  area estimate is greater than the true  average  area, all
concentrations estimated using that area estimate will suffer a negative bias; if the area estimate
is less than the true average area, all concentrations estimated using that area estimate will suffer
a positive bias.  The area estimate also  contributes variability to the estimated concentration
because, even if the estimate is an unbiased estimate of the true average area, it will  not be
exactly equal to the area examined in any particular analysis since grid opening size varies from
grid opening to grid opening. In general, the effect of the area estimate on the accuracy of the
estimated concentration is expected to be small relative to inaccuracies surrounding the estimate
of the number of asbestos structures.  The following quality control procedure is suggested to
ensure that this is the case.
             Select 20 grids at random from each lot of grids (approximately 1,000 grids).
Measure the area of at least 20 grid openings on each grid.  Calculate the mean of the 400
measurements and use  the mean as the estimated area of one grid opening for all calculations
using grids from that lot.
             When the second method is used, at least one grid opening must be measured
on each grid. Additional measurements may be needed if there is significant variability among
grid openings within a single grid.
             Depending on lot size and the  amount of variability within and between grids,
the first method  may  be   more practical for routine laboratory operations.   Individual
measurements on each  grid may be called for in special cases where specific documentation is
needed.  Maximum accuracy is achieved when the area of each grid opening examined in the
analysis is measured.

4.6  CONTROL OF CONTAMINATION
4.6.1  General Considerations
             Chrysotile asbestos is ubiquitous in the general environment. Therefore, materials
used in preparation of TEM specimen grids  from air sample filters can be contaminated by

-------
                                                                QA Guidelines for TEM
                                                                            Revision: 0
                                                               Date:  December 27, 1989
                                                                 Section 4 Page 5 of 38
asbestos fibers. Accordingly, the methods used by the laboratory to ensure that results are not
affected by contamination must be addressed in detail.  The methods used by each  laboratory
will vary, depending on what has been found satisfactory.  Some laboratories specify that all TEM
specimen preparation will be performed in a positive pressure clean room. Others have found
that reliable data can be produced by less extreme measures. The precise procedures used by
the particular laboratory  should be fully described.  The actions to be taken in the case of a
contamination incident must be specified.  If a source of contamination is identified, procedures
should be modified to minimize the possibility of a recurrence.  The modifications  should be
documented by a revision to the quality manual.

4.6.2  Laboratory Blank Measurements
              The preparation of unused filters along with actual sample filters provides a means
of detecting whether the result of any TEM analysis has been compromised by contamination
from  sources in the  laboratory during preparation.  The unused  filter material employed for
laboratory blanks should  be of the same type (mixed cellulose ester (MCE) or polycarbonate
(PC)) as that  employed  for the samples,  and it should be selected from  a stock of  filters
maintained by the laboratory for the purposes of its own laboratory blank program.  These filters
should have a  low  and well-characterized rate of asbestos contamination.  The AHERA TEM
protocol specifies one laboratory blank be prepared for each group of samples prepared together,
or a minimum  of one blank for every 10 samples prepared. At least one of the laboratory blanks
will be analyzed for every 25 samples.
             The  AHERA protocol specifies the minimum number of blanks. The  laboratory
may exceed these minima. It is recommended that at least one laboratory blank be included with
every  group of samples processed as a group through the plasma asher and the Jaffe washer in
order to provide blank control over every sample. No sample should be passed  through  the
preparation sequence unless it is accompanied by a laboratory blank. If high concentrations are
observed on a group of samples, and no blank was processed  simultaneously along with  the
group, there is no evidence available to prove that the results are not a consequence of  laboratory
contamination.   A blank filter incorporated with some earlier group of samples is of little
diagnostic value, and re-preparation of the filters would be necessary because  the validity of the

-------
                                                                QA Guidelines for TEM
                                                                            Revision: 0
                                                               Date:  December 27, 1989
                                                                  Section 4  Page 6 of 38
high results cannot be demonstrated.
             The policy of the  laboratory concerning the point at  which blank filters  are
introduced, the number of blank filters to be prepared, and the number of these blank specimens
to be  examined  in  the TEM should be described.   The AHERA protocol specifies 53
structures/mm2 as the acceptable upper limit for a single background measurement, for both PC
and MCE filters, and the long-term average of these background measurements may not exceed
18 structures/mm2. The laboratory may apply more stringent standards, and should attempt to
achieve the lowest possible laboratory blank values. The quality manual should specify the upper
limits of acceptability for blank measurements, along with the actions to be taken if these limits
are exceeded.

4.6.3  Filter Lot Blanks
              Before air samples are collected, it is important to establish that the particular
batch of filters to be employed for air sampling has an acceptably low background contamination.
Filters are selected at random from the batch for analysis.  These measurements are known as
"filter lot blanks." For blank measurements to be truly representative of the filter materials used
to collect the  actual samples, the laboratory should analyze unused filters from the same lot
number.  The AHERA protocol requires that the maximum permissible levels of contamination
on filter  lot  blanks are 53 structures/mm2 with a  long-term average  not  exceeding 18
structures/mm2.
              The relevance of filter lot  blanks to the operation of a  particular laboratory
depends on the range of services offered by the laboratory. If the laboratory supplies filter
cassettes to the client, or if the laboratory personnel collect field samples, the responsibility and
methods for ensuring that the filters are of sufficiently low  background must be clearly defined
in the quality manual.

4.6.4 Field Blanks
              The AHERA TEM protocol specifies that one closed and two open field blanks
shall be submitted along with each group of air sample filters. Closed field blanks are cassettes
which have been transported to the sampling site and sent to the laboratory without having been

-------
                                                                QA Guidelines for TEM
                                                                            Revision: 0
                                                               Date:  December 27, 1989
                                                                  Section 4  Page 7 of 38
opened; open field blanks are cassettes which have been transported to the sampling site, opened
for a short time without any air having passed through the filter and sent to the laboratory.  The
sealed blank should not be regarded by the laboratory as equivalent  to  a laboratory blank,
because it has been out of the control of the laboratory.  The laboratory should always add
laboratory blanks from its own stock of well-characterized filter medium to each preparation
series, because if the sealed blank exhibits any contamination, a laboratory blank provides the
only way by which a laboratory can confidently demonstrate the integrity of the results.

4.6.5  Reagent, Equipment and Material Blanks
             The extent to which reagents, equipment and materials used in the TEM specimen
preparation require investigation for the possible presence of asbestos contamination depends on
the results of laboratory blanks.  If, with  the  existing reagents, equipment and materials,
laboratory blanks consistently meet the required specifications, there is no need to  investigate
further.  Each of the items used in  the  specimen preparation can  potentially be  sources of
asbestos contamination. The procedures used by the laboratory to establish that these materials
do not contribute fiber contamination to the  measurement should be described in  the quality
manual.   If contamination is  detected on a laboratory blank,  then it will be  necessary to
investigate every item used in the analysis, including the reagents, tools, glassware, plasma asher
and vacuum evaporator. It is recommended that records be kept of any changes in the analytical
procedure, particularly such events as opening of a new bottle of solvent or the introduction of
new glassware.

4.6.6  What to Do When Contamination is Detected
             A contamination incident is usually detected by finding asbestos structures on a
blank, or on any other sample which was not anticipated  to have  asbestos structures.  The
procedures which the laboratory follows in this event should be described.  When a laboratory
blank is found to be contaminated, all of the samples which were prepared following the previous
satisfactory blank measurement must be assumed to have been compromised. An example of
these  procedures is described in Appendix B.

-------
                                                                QA Guidelines for TEM
                                                                           Revision: 0
                                                              Date:  December 27, 1989
                                                                 Section 4  Page 8 of 38

4.7  QUALITY ASSURANCE REANALYSIS OF SAMPLES

4.7.1 Types of Reanalysis
             The precision of TEM asbestos analysis is determined by a statistical evaluation

of repeat analyses of samples. The terms "duplicate analysis" and "replicate

analysis" have often been used, but the meanings of these terms are ambiguous, and they are not

sufficiently descriptive  to accommodate the range of  reanalysis possibilities.  The following

combinations are possible for repeat analyses:


       (a)    Reanalysis of the same grid openings of the specimen grids by the same operator;

       (b)    Reanalysis of the same grid openings of  the specimen grids by another operator
             who is a staff member of the same laboratory;

       (c)    Reanalysis of the same grid openings of  the specimen grids by another operator
             who is a staff member of another laboratory operating under a different  quality
             assurance program;

       (d)    Reanalysis of the same specimen grids by the same operator, without restricting
             analysis to the same  grid openings;

       (e)    Reanalysis of the same specimen grids by another operator who is a staff member
             of the same laboratory, without restricting analysis to the same grid openings;

       (f)    Reanalysis of the same specimen grids by another operator who is a staff member
             of another laboratory operating under  a  different  quality  assurance  program,
             without restricting the analysis to the same grid openings;

       (g)    Preparation  of new specimen grids from the original  filter by the original
             laboratory, followed  by analysis of the grids according to protocols (d), (e) or (f);
             and

       (h)    Preparation of new specimen grids from the original filter by a different laboratory
             operating under a different quality assurance program, followed by analysis of the
             grids according to protocols (d),  (e) or (f).

             In the case of reanalysis protocols (a), (b) and (c), another and more sophisticated

level of control is possible.  This is called "verified counting" (Steel and  Small 1985; Ogden, et

al. 1986).  Verified counting is the  only method known for determining  many of the causes of

analysis discrepancies  between  TEM operators.   The  use of  verified counting  allows the

-------
                                                                  QA Guidelines for TEM
                                                                              Revision:  0
                                                                Date:  December 27,1989
                                                                   Section 4  Page 9 of 38
classification of discrepancies into categories based on identification, complex counting rules,
stage traverse methods, fiber sizing, etc. Verified counting is performed by having the same grid
opening analyzed by more than one operator.
              The only requirement for verified counting is simply that each structure in a grid
opening be uniquely identifiable, so that cross comparison can be made between operators and
verification can be performed on structures for which there is disagreement or those structures
found by only one operator.  One way of implementing verified counting is using an identical
orientation of the same grid opening, using a similar scanning pattern so that asbestos structures
are encountered by  each operator in approximately the same  order.  Each asbestos structure
detected is then identified and measured, after which a sketch is made of the appearance of the
structure and its orientation.  The asbestos structure counts obtained and the sketches of each
structure made by the different operators are then compared and evaluated to determine which
asbestos structures were correctly classified by each operator. It is then possible to specify each
operator's performance in terms of true positive counts, false positive counts and false negative
counts.  Since it is often impractical to define all nonasbestos structures, the false positive and
false negative error rates  are defined in terms of the total number of asbestos structures.  For
example, suppose the area examined during verified counting contains 20 asbestos structures and
an  analyst correctly identifies 18 of the  asbestos structures as asbestos and classifies the 2
remaining asbestos structures as nonasbestos.  The true positive  rate for that analyst is 90%
(18/20 x 100), and the false negative rate is 10%  (2/20 x 100). If the analyst also  incorrectly
classifies 3 nonasbestos structures as asbestos, the  false positive rate is 15%  (3/20 x 100).
              The data  on precision generated  by repeat  analyses  according  to  the above
protocols have  different meanings. Laboratories  do not need to  apply all of these reanalysis
types on a routine basis, but it  is important to understand the various possibilities and the
meaning of data derived from each.
             It is necessary for a laboratory to generate information concerning the precision
attained by individual TEM operators and the precision attained by the laboratory overall. It is
also necessary to establish that the results are comparable with any consensus standards which
exist, and with  those generated by other laboratories.  Absolute standards are not available to
assist in these tasks, and the air filters to be analyzed do not generally exhibit good  uniformity

-------
                                                                 QA Guidelines for TEM
                                                                             Revision:  0
                                                                Date:  December 27,1989
                                                                 Section 4  Page 10 of 38
of particle deposition.
             The quality manual must show that the laboratory staff understand the types of
reanalyses which are possible, and it must contain a description of how the laboratory uses the
various kinds of reanalyses to characterize the precision and accuracy of the operators and the
laboratory.  Complete characterization of the precision and accuracy of the overall analytical
method is not a responsibility for the individual laboratory, although some of the data generated
from the reanalyses will provide estimates.

4.7.2  Determination of Operator Precision
             The number of asbestos structures recorded on each grid opening  of a TEM
specimen grid is usually not constant. Accordingly, if the performance of an operator is checked
by allowing the operator to select different sets of grid openings for each successive recount of
the same sample, it is  not certain  whether any variability in the results  is a consequence of
operator variability or nonuniformity of particulate deposition on the specimen grid. Therefore
operator precision can be measured only by reanalysis of the same grid openings.  On  a routine
basis, these measurements should be made by simple  recounts of the  same grid openings.  If
significant discrepancies are detected by these routine QA procedures, the  full verified counting
procedure, incorporating a sketch of each asbestos structure and using defined grid orientations,
should be used to investigate the origin of the discrepancies.  It should be recognized that the
use of the full verified counting procedure could alert the operator prior to the first examination
of the grid that a test is being conducted. The operator is likely to take more care and  therefore
the precision value  derived from verified counting will usually be more optimistic than that
normally achieved by the operator  under routine conditions.  Laboratories can counteract this
effect by recording all information needed for verified counts during their routine counts.  Then
verified counts can subsequently be performed on any  previously counted  squares.

4.7.3  Determination of Laboratory Precision
              Recounts made by different operators on the same grid openings of specimens
provide a measurement of the precision of fiber counting in the laboratory. As in  the case of
determination of individual operator precision, the most realistic data representative of normal

-------
                                                                 QA Guidelines for TEM
                                                                              Revision:  0
                                                                Date:  December 27,1989
                                                                  Section 4 Page 11 of 38
operation are obtained  using  routine  re-counts of identical  grid openings.   If the analyst
conducting the first examination is alerted to the fact that the count is a test, the verified counting
procedure will yield an optimistic value for the laboratory precision. Laboratories can counteract
this effect in several ways. One way would be to use the full verified counting procedure for all
samples so that the operator is not aware of which grid openings are later to be verified by a
second examination.  This approach, however, is somewhat  time consuming.  An alternative
approach is to select a proportion of the grid openings from routine sample examination and to
examine  these by the verified counting method  in  order  to generate  QA data.  In the
interpretation of verified  counts, it should be recognized that chrysotile can suffer beam damage
during the initial TEM  examination, and this will compromise the  ability of operators in
successive examinations to obtain diffraction patterns.

4.7.4  Determination of Intralaboratory Method Precision
              As discussed in Section 3.2.2, the distribution of asbestos structures collected on
a filter is not within the control of the laboratory, and different areas of the filter will generally
yield different results. This variability is an additional component related to the overall method
of analysis.  Repreparation of  several sets of specimen grids from the original air filters and
counting of  asbestos structures  on each  of these sets by  a different operator provides a
measurement of the overall method precision as practiced by the particular laboratory.

4.7.5  Determination of Interlaboratory Method Precision
              Interlaboratory measurements on different areas of the collection  filter will
generally exhibit the most variability.  This type of interlaboratory measurement, in which each
laboratory prepares its own grids from the original filters, provides information on the precision
of the overall method since  it replicates the entire protocol.  All  other precision estimates
described here encompass a portion of the total.

4.7.6  Analysis of Proficiency  Testing Samples
              Proficiency testing  samples  will be issued to TEM asbestos laboratories as part
of NVLAP. All operators in the laboratory will be required to count asbestos structures on these

-------
                                                                  QA Guidelines for TEM
                                                                              Revision:  0
                                                                Date:  December 27, 1989
                                                                  Section 4 Page 12 of 38
samples in order for the laboratory to maintain its accredited status.  This type of sample will
provide an external reference.  It is not possible to issue such samples frequently enough to serve
as the only interlaboratory proficiency testing standards.  Therefore, laboratories  should, in
addition, have arrangements for interlaboratory sample analyses to be conducted regularly with
other organizations.
              As described in section 4.7.2 (Determination of Operator Precision), a stringent
internal proficiency testing program must also be incorporated  into the QA program  of a
laboratory.  A continuing program of this type has been shown to improve the performance of
operators.  Proficiency testing samples can be generated internally by the laboratory, using either
previously-analyzed samples, or samples generated artificially. Three types of proficiency testing
sample should be used:  one type to test specimen preparation; one to  test routine identification
and quantification; and a third type to challenge the identification skills of the operators.

4.7.7  Interlaboratory Sample Analyses
              If a laboratory is operating in isolation under an internal QA program, the results
from  the laboratory can be strongly biased.  This is a consequence  of the  fact that asbestos
analysis has a subjective component, and bias will only be detected when the laboratory results
are compared with those of another laboratory for the same samples. A positive or negative bias
can occur as a result  of improper fiber identification criteria, or as a  consequence of improper
application of fiber counting criteria.  A negative bias can arise if the specimen preparations are
too thick, if the fiber identification criteria are excessively rigid, if fiber crystallinity is destroyed
due to electron beam  damage  during examination, or if  fibers are  lost  during the specimen
preparation. Thick preparations lead to  low image contrast and failure  to obtain discernible
electron  diffraction  patterns,  leading to  reporting of lower results.  Excessively  rigid  fiber
identification criteria may  not  take  account  of the instrumental  limitations resulting in the
rejection of many fibers when they are actually asbestos.  Over-irradiation of the specimen in the
TEM may lead to an inability to identify asbestos fibers in any succeeding analysis. Loss of fibers
during preparation can  occur as a result of some systematic feature of the specimen preparation
procedure used. Any of these situations can exist indefinitely within one laboratory, because each
operator uses the same criteria as specified originally by the laboratory supervisor.  Moreover,

-------
                                                                 QA Guidelines for TEM
                                                                             Revision:  0
                                                                Date:  December 27, 1989
                                                                 Section 4 Page 13 of 38
there may be good agreement between results generated by all of the operators, despite the fact
that the results are actually strongly biased (precision but not accuracy).
              Each laboratory should establish a relationship with at least one external laboratory
for the purposes of conducting interlaboratory analyses. These analyses must be conducted blind,
so the operator in the second laboratory is not aware of the result already produced by the first
laboratory. The samples should be selected to have suitable fiber loadings so that the results will
be diagnostic for any major discrepancies. If major discrepancies are detected, the reasons must
be investigated, and verified counting  is recommended  as a  useful tool for this purpose.
              The number of samples  to be analyzed by  an external laboratory and the frequency
with which this should be done are dependent on the number of analysts in the laboratory and
the volume of samples analyzed.  The number of analyses required can be minimized  by  an
agreement between the laboratories to share all of the interlaboratory data. A sample exchange
resulting in four  interlaboratory results made four  times per year is an acceptable minimum
number. Filter sectors should be exchanged, so that the entire analytical  sequence is  tested and
TEM specimen grids should also be exchanged for counting of identical grid openings by the two
laboratories.   If  discrepancies  are found, verified  counting procedures should be instituted
between the laboratories in order to resolve  them.
              The choice  of  external  laboratory  is  important, because  any discrepancies
generated (possibly as a result of selection of an inexperienced partner) will form part of the
laboratory's QA records. The external laboratory must, however, be totally independent; sample
exchanges with a laboratory from the same corporation, the  same organization, or one which is
associated in any formal manner with the first laboratory, cannot be considered as interlaboratory
exchanges.

4.7.8 How to  Correct Counting Discrepancies
              When discrepancies arise between two operators or two laboratories, they must  be
resolved before either the operators or laboratories can have total confidence in their results.
An example of the resolution of such  a discrepancy is discussed in
Appendix B.

-------
                                                                 QA Guidelines for TEM
                                                                            Revision: 0
                                                               Date:  December 27, 1989
                                                                 Section 4 Page 14 of 38
4.8  INTERPRETATION OF RECOUNT DATA
             Repeated analyses of the same sample are intended to identify problems (failure
to follow operating procedures in a consistent manner, inability of  an analyst to distinguish
asbestos  structures from nonasbestos,  different interpretations  of  counting rules)  and  to
characterize performance over time both for the laboratory as a whole and for individual analysts.
Repeated analyses are expected to give different results.  To achieve  the  first objective it is
necessary to decide when the differences are sufficiently large to indicate  a problem. Ideally, this
is based  on accumulated performance data.  A repeated  analysis that falls outside the usual
performance standards calls for investigation and corrective action.  Performance data for a new
laboratory, or a new protocol such as the AHERA TEM protocol, are not available when a QA
system is originally developed.   The  following discussion provides some  suggestions for
accumulating and interpreting the necessary data.
             There is an important distinction between repeated analyses that include different
subsamplings of the filter (for example, (A) and (B) in Figure 3-1) and those that do not (for
example, (C) in Figure 3-1).  Different subsamplings of the filter introduce a component of
variability that cannot be controlled by the laboratory. When a repeated analysis is done using
the same grid openings that were examined in the original analysis, no subsampling is involved
and differences between results  cannot be attributed to  the  spatial distribution of asbestos
structures on the filter.  Verified asbestos analysis which does not involve a subsampling step is
described in detail in  Steel and Small (1985). and will not be discussed further in this section.
When a subsampling step is involved, variability, as measured by the coefficient of variation, will
depend on the number of asbestos structures counted. This fact complicates QC, especially at
low concentrations (Ogden, et al.  1986).   Since the  AHERA TEM protocol is a clearance
protocol, low concentrations are expected and cannot be ignored.  The following procedure is
designed for low concentrations.
             For each type of repeated analysis involving  a subsampling step, accumulate data
for plotting the coefficient of variation against the mean. The coefficient of variation is given by:

                                       CV = (s/m)

-------
                                                                  QA Guidelines for TEM
                                                                              Revision:  0
                                                                Date: December 27, 1989
                                                                  Section 4 Page 15 of 38
where m = Ex/n is the mean of the results obtained from n analyses of a single sample and s
= (E(x-m)2/(n-l))^ is the standard deviation. If n varies from sample to sample, create a separate
graph for each n, or distinguish the different values of n by using different colors or symbols on
the graph.
              Figure 4-1 indicates how the graphs of the coefficient of variation would look if
the only source of variability were that introduced by a Poisson distribution describing the spatial
distribution of structures on the surface of the filter1.  The Poisson distribution represents the
best case in which asbestos structures are distributed randomly across the surface of the filter.
In  practice, the distribution of asbestos structures across the surface of the filter  may be
somewhat patchy and other distributions  such  as the negative binomial  may be more realistic
(Javitz and Fowler 1981). If asbestos structures  are distributed according to a negative binomial,
the coefficient of  variation will be greater than  the coefficient of variation predicted by (.ie
Poisson distribution.
              Ideally, if a Poisson distribution were appropriate, 50 percent of the coefficients
of variation calculated by a laboratory should lie below the 50th percentile  line,2 75 percent
should lie  below  the 75th percentile  line,  and so on.  In practice, coefficients of variation
calculated  by a  laboratory from actual samples will be  higher  because, (1) the true spatial
distribution may be more variable than that represented by the Poisson distribution, and (2) there
are many other sources of variability in addition to the effect of subsampling. A laboratory might
find, for example, that only 35 percent of its values fall below the 50th percentile, only 60 percent
of its values fall below the 75th percentile, etc.  The discrepancy between the ideal performance
and actual  performance  is  a result  of additional variability  beyond that expected from
subsampling a Poisson distribution.  There is insufficient experience at present to indicate the
         graphs were generated by computer simulation. A random sample of size n was
generated from a Poisson distribution with a specified mean and the empirical coefficient of
variation calculated. The process was repeated 500 times for each mean value.
    2The theoretical percentiles were estimated from the 500 values of the coefficient of
variation generated in the computer simulation. For example, the  75th percentile was
determined by finding a value that was greater than 75 percent of the observations and
smaller than 25 percent of the observations.

-------
  o o\ oo
    oo co


  ls-°
  •S3 (N —•
   » >->
UC4

*s
3

o
       4>
       oo
O
8^
& §

il
Q
•=
"^
o
J£
0)
Q.


i
                         8
-=
^
0)
a
(D
Q.
0>
a
(D
Q.
-=

"c
CD

S
(1)
Q.
                                                              §
                                                              ts

                                                              •<—»
                                                              wa
CO
<4-4
O


1


G
                                                                                1
                        CM
                      c
                     .Q  CO

                     15
                     T3  O
                     .8
                     £5
                                                                    O
                                                                    CM
                                                       CO
                                                       s

                                                       1
                                                       CO

                                                       B

                                                       CD


                                                       E
                                                                         CO
                                                                         CD
                                                                                •
                                                 a 8
                                                  - -^
                                                  ea ca
                                                  CA g
                                                  C5 a
                                                  e "V
                                                  o B
                                                 •J3 ca

                                                 •   8.
                                                                                *J 1)
                                                                                C X)
                                                               31
                                                               CO >
                                                               O ^


                                                              «§
                                                                                T^
                                                                                CS

                                                                                8
               cv
                                      iq
                                      d


-------
   0)
   Q.
0)
a
(D
Q.
        O)
-=
"•5

0)

a
0)
a

4-1
in
        co
     c w

     II
     .Q (0

     I*
     T3 O

     O «
     ° -i
     .52 £
     O 3
     Q. Z
                                    10
                                        (O
                                    o  °

                                    "  5


                                        z

                                        (0
                                        0)

                                    u>  5
                                                       1
                                                       s
CO
<*H
O


I
                                                       eo
                                                       u
                                                       U
                                                       eo
                                                 s
                                               •  21

in
         in
               in
               o
3 3
a "3
sr

^1

si
.2 -6
c« VH
££

c S
8 g
5 2
PH CX
                                                       PL.

-------
        .2
        '^
        o
     O
     O
              i
CD
o
.2
"c
§
|

S
in
                                                          I
                                                          .0
                                                          6
                                                          03
                                               O
                                               CVJ
                                                   w
                                               o  ?

                                                          I A
                                                 ca  ca
                                                 3  §
                                                 - -O
                                                 •§8
                                                 .3  cO
                                                 •S  8-
                                                 §£
                                                 •30
                                                 3.JS
in
cvi
CM
                 in
                 o
                                                          51

                                 <+H *O
                                  c^ >-<
                                  (D O
                                 W ^

                                  11
                                  •i 0)
                                  w £5
                                                           o

-------
"5! O o oo
<   op co
   §^.0

  •S3 cs S
s
                                                                                          CA


                                                                                          WH
|o:
          )
        CO
8
a
§
   o
   S
to C/D
                    1
                    0)
                    a
                    O
                    a


                    f
                    O)
                           .2



                           §
                             Oi


                             S
                             Q.

                                          (D
                                          a


                                         I
.0)


1
0)
a
                                                                                           1/3
 co

<4-H

 O




I

 E
                           in
                                                                             m
                                                                             T-   w
                                                                                  I

                                                                             o   ?
                                                                                  |
                                                                                  CO
                                                                                  03
                                                                                           ca
                                                                                           0)
                                                                                          a

                                                                                          TO  {Q

                                                                                          x  c
                                                                                          co  ™

                                                                                          S  o
                                                                                          o  w
                                                                                          •^3  CO
                                                                                          CO  O
                                                                                          •F^  ^^


                                                                                          IB
                                                                                          4->  D

                                                                                          a) "a

                                                                                         :s  I
                                                                                         sa  «
                                                                                             0)
                in
                           OJ
                                                         in

                                                         C)
                                                                                          PH
                                  UOI;BUBA jo

-------
^ O O OO
^   oo
rrl .. Os

p 8*1
   o ^-
l^^c^
.S   S &<
1   8"
a   Q|
s   y
   I
   Q)
   Q.
   O)
                        0)
                        8.
                        in
                        o>

0)

a
O
Q.


§
O)
8
CD
a
0)
Q.


S
in
                                                                 3
                                                                 ts


                                                                 to
                                                                               03
                                          1
                                           C

                                           C8
                                                         co

                                                         1
                                                         I
                                                     o
                                                     i-   d)
                                                         .a
                                                          co
                                                          CD

                                                                 ca
                                                                               c
                                                                               O S
                                                                               'S cu

                                                                               •S 21
                                                                               II
                                                                 3 3
                                                                4i T3
                                                                 ca >
                                                                 o ._

                                                                 » 5
in
cvi
                        CVJ
     in
                  in
                  o
                                     » S
                                     O en
                                     t-i Oi
                UOjlBUBA p
                                                                2?

-------
                                                                 QA Guidelines for TEM
                                                                             Revision: 0
                                                                Date: December 27,1989
                                                                  Section 4  Page 21 of 38
typical magnitude of the additional variability. As data are accumulated, achievable guidelines
will emerge, and laboratories should strive to improve beyond them.
              Each laboratory should develop its own quality objective or standard based on its
performance record. For example, after collecting data for several months, a laboratory could
derive its own 95th percentile line which would lie somewhat above  the ideal 95th percentile
shown in the figures.  Any repeated  analysis falling above the laboratory's 95th percentile line
would receive special scrutiny. If over a period of time, or within a particular batch of samples,
more than 5 percent of repeated analyses fell above the laboratory's 95th percentile line, a
systematic effort to pinpoint and  correct the decrease in performance  would be undertaken.
Note that one reason for an apparent decrease in performance could be due to an unusually
clumped  distribution of  asbestos structures on  the filter  surface, a factor  over which  the
laboratory has no control. This should not be used as an explanation for every repeated analysis
that falls  outside the designated range, but  it should be kept in mind when investigating a
problem.
              Figure 4-1  shows that the coefficient of variation depends on the number of
repeated analyses (n) used in its calculation and it is particularly sensitive to (n) when both (n)
and the average structure count are small. Keeping the number of repeated analyses as constant
as possible and preferably greater than 3 simplifies interpretation of the results. While this may
be possible in some cases (e.g., a within laboratory program involving four analysts), it may be
impractical in others (e.g., an interlaboratory program with a variable number of participants).
The behavior of the coefficient of variation is more erratic at low structure counts.  This difficulty
can be reduced by increasing the  number of grid openings  counted on QC samples with low
structure concentrations.  The same number of grid openings must be counted for each repeated
analysis, otherwise the coefficient of variation will not be estimated correctly.
              The relative contributions of different parts of the process to total variability can
be assessed by comparing the coefficient of variation graphs for different types of
repeated analyses.  For example,  the percentile lines for  intralaboratory samples of the type
illustrated in Figure  4-1 (A) should  lie above the corresponding percentile lines for repeated
analyses of a single grid in which different grid openings are counted each time. If the difference
between the percentile lines is large,  the sample preparation step is introducing a large amount

-------
                                                                 QA Guidelines for TEM
                                                                             Revision:  0
                                                               Date:  December 27, 1989
                                                                 Section 4 Page 22 of 38
of variability into the process and may be a future target of attempts to reduce total variability.
(Note that the subsampling step is involved in both types of repeated analysis, therefore any
consistent differences between the percentile lines in this example cannot be attributed to the
spatial distribution of structures on the filter.) The coefficient of variation graphs also provide
a means of indicating to the client the amount of uncertainty associated with an individual data
value.  This is discussed in Section 4.13.
             Although the coefficient of variation is a familiar and useful way of summarizing
information on variability it does not capture all the  relevant information that can emerge from
repeated analyses. For example, the coefficient of variation is not sensitive to an analyst in an
intralaboratory program who consistently  obtains slightly lower counts than other analysts.
Laboratories are encouraged to develop additional  and alternative procedures to display and
interpret the  results of repeated analyses.
4.9 USE OF STANDARD SAMPLES
4.9.1  Calibration of Plasma Asher Performance
              In the preparation of TEM specimen grids from MCE filters, the filter is collapsed
and a plasma etching treatment is used to remove a thin layer of the MCE polymer from the
surface.  This is done to remove polymer which may have engulfed asbestos fibers during the
filter  collapsing procedure.   However, since the performance of plasma  ashers cannot  be
accurately  specified, it  is necessary for each laboratory to  determine satisfactory operating
conditions for its particular unit.  The operating conditions can be specified approximately in
terms of a performance  standard. An uncollapsed 0.45 /xm pore size MCE filter is placed in the
plasma  asher,  and conditions are  established  such that the  filter completely  oxidizes in
approximately 15 minutes.  The laboratory then  should  develop calibration data to show that,
under these operating conditions, the time period selected  for the  etching treatment produces
satisfactory recovery of fibers and an acceptable level of background  structure in the final replica.
In other words, the laboratory needs to carry out tests to determine that the etching time and
operating conditions produce the proper etching results.  It has been found for one manufacture
of MCE filters that an etching period of approximately 8 minutes on the collapsed filter under

-------
                                                                QA Guidelines for TEM
                                                                            Revision:  0
                                                               Date:  December 27, 1989
                                                                Section 4 Page 23 of 38
these conditions provides satisfactory samples.  An alternate approach, specified in the AHERA
TEM method, is to establish conditions in the plasma asher such that 10% of the filter thickness
is etched away.
             This performance  standard,  however, is very approximate.  The thickness and
precise mixture of polymers used  in filters varies with the manufacturer, and etching rates of the
different makes of filter may vary. Laboratories must perform their own experiments to establish
that their procedure for plasma  etching of the collapsed filters  does  not generate excessive
background detail on the replica,  or cause asbestos structures to fail to transfer to the final TEM
specimen grids. These data should form part of the QA records,  and the procedures used for
obtaining the data should be fully documented in the quality manual. The performance of any
particular plasma asher varies with aging of the electronic components;  hence a need exists for
the calibration to be repeated on a regular basis.  The
policy of the laboratory for performing  this  calibration should be documented in the quality
manual.

4.9.2 Calibration of TEM Magnification
             The magnification  of the TEM can change.  Particularly after any maintenance of
the TEM,  and  also  at regular intervals, the magnification  of the  TEM at  which  fiber
measurements are made  must  be  determined,  both on the  fluorescent screen  and on  a
photograph, to ensure the measurement of fiber  dimensions continues to be accurate.  If the
required measurements of grid opening dimensions are performed in the TEM, it will also be
necessary to calibrate the lower magnification used for this measurement. The measurements
must form part of the QA records, and the procedures used for making the measurements must
be specified in  the  quality manual.  Carbon  replicas of ruled  gratings are  normally used to
calibrate the magnification of the TEM. A replica with 2160 lines/mm is suitable for calibration
of the high magnification used for fiber counting.  For calibration of the low magnification used
for grid opening dimension measurement, a replica of approximately 600 lines/mm is suitable.
These can be obtained from electron microscopy accessory suppliers.  The calibration standards
used must be specified in the quality manual.

-------
                                                                  QA Guidelines for TEM
                                                                              Revision:  0
                                                                Date:  December 27, 1989
                                                                  Section 4 Page 24 of 38
4.9.3  Calibration of the Diffraction Camera Constant
              The interpretation of electron diffraction patterns requires that the camera constant
for the diffraction  pattern be  known precisely.  Calibration of the  camera constant  can be
achieved using any known crystalline material.  A thin polycrystalline film of gold is a suitable
choice. The procedures for determination of a camera constant for the fluorescent screen, and
for a photograph, must be clearly described in the quality manual.  These procedures must take
into account specimen eccentricity, objective lens excitation, and the effects of lens distortions
on the camera constant.  The procedures for correction of measurements for these distortions
must also be described in the quality manual.  References that may be consulted for electron
diffraction calibration include Lee (1978), Andrews et al. (1967), and Hirsch et al. (1965).

4.9.4  Calibration of Electron Beam Dose
              Chrysotile asbestos is easily damaged by excessive irradiation in an electron beam.
During such an irradiation, the morphological appearance of a single fibril of chrysotile changes.
Even if an electron diffraction pattern was visible initially, it rapidly disappears.  This change is
irreversible, and since the electron diffraction pattern is the primary identification criterion for
chrysotile, degradation of fibers in this way may cause low values to be reported.  It  is most
important, therefore, that the electron beam dose administered to a fiber be calibrated in some
way, so that excessive irradiation can be avoided. It is not required that this calibration be an
actual measurement; but the operating conditions must be specified such that excessive irradiation
does not  occur.  A demonstration that the electron diffraction  pattern obtained from a single
fibril of chrysotile can still be seen after an irradiation period of 15 seconds or more constitutes
adequate calibration.  The operating conditions normally used by the  laboratory to achieve this
condition and the  requirement to confirm these conditions  must be specified in the quality
manual.

4.9.5  Calibration of Operators Using SRM 1876a
              NIST Standard Reference Material (SRM)  1876a is a portion of a filter which
has been  examined by a  number of microscopists to determine the fiber  density.  This
material can be used to  prepare  specimen grids  which have a well-characterized fiber density.

-------
                                                                 QA Guidelines for TEM
                                                                              Revision:  0
                                                                Date:  December 27,1989
                                                                  Section 4  Page 25 of 38

These grids should be  used for training of new operators, and also to provide an external

reference for the accuracy of fiber counting. SRM 1876a is a very simple material, and the

fibrous  structures consist mainly of single fibers.   This sample, therefore, is limited  in  its

usefulness for testing of operators' ability to apply the fiber identification and counting criteria

correctly. SRM  1876a  should not be relied on as the only reference sample. It has limited

complexity and is available only as a polycarbonate filter.


4.9.6 Quality of TEM Specimen Preparation

             The quality of the TEM specimen  grids prepared by a laboratory seriously affects

the results obtained. Although some aspects of the quality of the TEM specimens are necessarily

subjective, others  can be put on a more quantitative basis. The aspects of TEM specimen quality

which must be addressed in the quality manual include the following:
       (a)     The TEM specimen must not be too thick. If the carbon replica is too thick or
              if substrates are used and the combined replicas and substrates are  too thick, the
              contrast of small fibers will be reduced, and operators will more readily overlook
              them during scanning. The  substrates  used and the carbon replica may also be
              too thick to permit satisfactory electron diffraction patterns to be obtained.  On
              typical TEM specimens prepared by the laboratory, more than 90% of the single
              fibrils of  reference chrysotile  should  exhibit  satisfactory electron diffraction
              patterns.  Data which demonstrate that  this can be routinely achieved by the
              laboratory must form  part of the QA records.

       (b)     Undissolved filter polymer should not remain on the TEM specimens after solvent
              extraction.  This has the same consequences as those discussed in (a), and the
              specific  procedures used by  the laboratory to solve this problem when it occurs
              should be described.

       (c)     In MCE preparations, there should be no evidence of residual filter structure due
              to incomplete collapsing of the spongy structure of the membrane filter.  Residual
              filter structure  results in a replica containing  confusing  detail which acts as
              camouflage for fibers.  Operators readily overlook  short fibers when they are
              presented against a  highly detailed and contrast background  structure.  Criteria
              for satisfactory preparations  must be specified.

-------
                                                                 QA Guidelines for TEM
                                                                             Revision:  0
                                                                Date:  December 27, 1989
                                                                 Section 4 Page 26 of 38

       (d)    In MCE filter preparations, the collapsed filter should  not be either  over- or
             under-etched during the  plasma etching step.  The degree of etching can be
             estimated by tilting the specimen in the TEM to an  angle of 45°, and observing
             the  structure generated  by the etching treatment.  Criteria  for  satisfactory
             preparations must be specified.

       (e)    For PC filter  preparations,  there  should not be evidence  that the filter  was
             over-heated during the carbon evaporation.  This leads to  a "connected pore"
             appearance in the final TEM specimen, which may  conceal fibers.  Criteria for
             satisfactory preparations must be specified.

       (f)    There should be no evidence that significant numbers of particles have been lost
             from the TEM specimens.  Holes  in the carbon replica are clues that this has
             happened, and if these form a significant proportion of the features in a TEM
             specimen,  the  preparations are  unsatisfactory.  The policy of the laboratory in
             assessing suitability in this regard must be described in the quality manual.

       (g)    Sample filters which are  over-loaded by particulate will generally  yield TEM
             specimens on which a large proportion of the grid openings will exhibit broken
             carbon replicas. Since breakage normally occurs in regions of the specimen where
             particles are closely spaced, the intact areas tend to represent regions of the
             original filter which were more lightly loaded.  Counting of such a specimen may
             yield results which are biased downwards. The criteria used by the laboratory for
             rejection of specimens should be described.

       (h)    Where a large proportion of the  area of each grid opening is obscured by
             particulate, asbestos structures may be  obscured. A count  on such a specimen
             may yield results which are biased downwards. As for (g), the criteria used by the
             laboratory for rejection of such specimens should  be included in the quality
             manual.

       (i)    The quality manual should address the  procedures followed by a laboratory if a
             replica shows any other major defects, such as regions of folding, or unexplained
             debris on blank preparations.
4.9.7  Quality of Electron Diffraction Patterns

              Chrysotile is  identified on the  basis of an  electron diffraction pattern.  It is

therefore important that the procedures used by the laboratory yield patterns on which  the
relevant properties can be seen or measured. For example, the spots should be sharply defined,
and a camera constant which is sufficiently large so that spot positions on recorded diffraction

patterns can be accurately measured should be used. The area of the specimen from which the

-------
                                                                QA Guidelines for TEM
                                                                            Revision:  0
                                                               Date: December 27, 1989
                                                                 Section 4 Page 27 of 38
diffraction pattern is obtained should be sufficiently small so that interferences from adjacent
particles  do not  compromise the interpretation.   The  precise procedures  used to  obtain
satisfactory patterns should be documented in the quality manual.

4.9.8  Calibration  of Electron Beam Spot Size and EDXA System
             The AHERA TEM protocol specifies that the TEM must be capable of producing
an electron beam with a diameter less than 250 nm. This ensures that adequate spatial resolution
is available during acquisition of EDXA spectra from small fibers.  The quality manual  must
contain a description of how the beam diameter can be measured (Williams, 1986;  1984).
             Identification  of asbestos fibers, particularly amphibole asbestos fibers, requires
the use of an energy dispersive x-ray analysis (EDXA) system.  It is important that the energy
(horizontal) scale  of the x-ray spectrum be calibrated to ensure that the x-ray peaks appear at the
correct energy values. This calibration must be performed frequently and is easily accomplished
using a standard specimen referred to below.  Two variables may require adjustment  on the x-ray
system: the amplifier gain and the zero point. In order to perform  this calibration, a standard
specimen is required which yields a low  and a high energy peak.  An evaporated film of
aluminum on a carbon  film, supported on a copper TEM grid is  a suitable  specimen.  This
specimen gives  rise to a  low  energy aluminum peak and a high energy copper peak.
             If misidentification of amphibole fibers is to be avoided, the intensity scale of the
EDXA system must be calibrated carefully, using known standard materials.  Calibrations are
required  for the intensities of the K (alpha) lines for sodium, magnesium, aluminum, silicon,
potassium,  calcium, titanium, manganese and iron.  NIST's  SRM  2063  provides  a means of
calibrating  the  EDXA unit  for magnesium,  silicon, calcium  and iron.  Calibrations are also
required  for sodium, aluminum, potassium,  titanium and manganese in order to  confidently
discriminate asbestos fibers from nonasbestos mineral species.
             It is important to recognize that changes in detector efficiency can occur, and
periodic  recalibration is required.  The  intervals for recalibration  should be  specified in the
quality manual.

-------
                                                                 QA Guidelines for TEM
                                                                             Revision: 0
                                                                Date: December 27, 1989
                                                                  Section 4  Page 28 of 38
4.9.9  Preparation and Use of In-House Reference Standards
              The TEM operator should have available a series of mineral reference standards
consisting of TEM specimen grids with different kinds of mineral fiber on them. These mineral
fibers should include the common asbestos varieties and also other nonasbestos minerals which
have been incorrectly identified as asbestos (e.g., halloysite, palygorskite, vermiculite, talc). Small
samples of minerals can be obtained from some museums and also from the U.S. Geological
Survey. The laboratory should have a written procedure for use of such reference standards in
its training program, and for regular checking of any fiber  identification procedures which are
used.
              Suitable  reference specimen grids may be  prepared using  well-characterized
minerals.  It is often not sufficient to accept identification of mineral standards from geological
collections, since many of the classifications have been done using less specific techniques  than
are required for reference standards. Before incorporating a new mineral reference standard into
the set used by the laboratory, sufficient data or analysis should be documented to establish the
suitability of the material as a reference standard.  This documentation should form part of the
quality manual.

4.9.10  Use of Samples and Proficiency Testing Samples as Operator Training Materials
              Maintenance of  proficiency  in established  operators must  be a  continuing
operation, and proof of proficiency in new operators must be demonstrated before they are
allowed to report results independently.  Reference samples and proficiency testing samples
play a vital role in both activities. The quality manual should contain details  on how the
laboratory maintains  and demonstrates proficiency in its overall analytical program,  from the
initial preparation of the filter to reporting of the result.  Although  the NVLAP will  regularly
issue proficiency testing samples,   each laboratory should  also have an internal  proficiency
development and maintenance program based on samples from other sources.
              Proficiency must  be maintained and demonstrated  in  three separate areas:
specimen  preparation, fiber identification, and asbestos structure counting.  These three
areas require different types of proficiency testing samples.  For specimen preparation, the key
factors to  test for are listed in Section 4.9.6.  Proficiency testing  filters for this purpose  can be

-------
                                                                QA Guidelines for TEM
                                                                            Revision:  0
                                                               Date: December 27, 1989
                                                                 Section 4 Page 29 of 38
prepared by filtration of liquid suspensions of any well-characterized paniculate or asbestos
dispersion. TEM specimen grids which exhibit any of the defects discussed  in Section 4.9.6
should be cause to suspend preparation of samples and to review the preparation techniques in
detail. Any modifications in the preparation procedures which are made should be documented
and  added to the  quality manual as revisions.   Proficiency  in fiber identification can  be
demonstrated only by periodic submission of specimens containing a series of reference materials.
Proficiency in asbestos structure counting can be maintained by periodic analysis of specimens,
some of which contain only simple structures such as fibers and bundles,  and others which
contain a high proportion of complex structures.  It is possible to accumulate a library of such
training samples  from those analyzed routinely.

4.10  DATA RECORDING
             The system used for recording of data must be fully described in the quality
manual.   It should incorporate elements  concerning sample identification,  filter  type and
specimen preparation,  grid  selection,  fiber  identification, asbestos structure  dimension
measurement,  micrograph labeling, spectrum labeling,  TEM magnification, grid opening
dimensions, and any file structure and numbering system which is used for computer data entry.

4.10.1  Sample Identification
             The data recording system must provide a clear tracking for each sample through
the analytical system.  If the laboratory assigns sample numbers, the cross-referencing system
should be fully documented in the quality manual.  At various points during TEM analyses, there
are opportunities for loss of sample identification and interchange of samples. The system used
should record, for example, which  samples were prepared as a single group, and which blank
samples would be representative of the preparation conditions under which the group of samples
were prepared.

4.10.2  Filter Type and Specimen Preparation
             The specimen preparation procedure used  depends on whether the sample was
collected on a PC or an MCE filter; therefore, the filter type must form part of the data records.

-------
                                                                QA Guidelines for TEM
                                                                            Revision:  0
                                                               Date:  December 27, 1989
                                                                Section 4 Page 30 of 38
If special procedures were employed during specimen preparation, details of the variances from
the standard procedure  should be recorded.  For example,  if the filter was found  to  be
overloaded, and the TEM grids were prepared by an indirect transfer procedure,  all details of
the procedure must form part of the sample preparation records. The relevant details for this
example would be the area of filter ashed, the volume of liquid in which the ash from the filter
was dispersed, the volume of this dispersion which was filtered for each of the secondary filters
prepared, the area of each secondary filter, and the type of secondary filter.
              Minor variances from the AHERA protocol, such as redesign of the Jaffe washer,
or  use of a different solvent for dissolution, may still allow the  analysis to  conform to the
requirements of AHERA. The use of such variances, however, must be supported by sufficient
analytical work and documentation to demonstrate equivalence with the standard procedures.
Major variances, such as the indirect protocol used as the example above, do not conform to the
requirements of AHERA. All variances normally used  in the laboratory must be fully described
in the quality manual, and when they are used they must form part of the sample preparation
records.

4.10.3  Grid and Grid Opening Selection
              There is often reason  to believe  that  the particulate on  a collection filter is
restricted to localized areas of the filter.  Accordingly, it is important to record the point in a
filter count at which another grid is examined.  It is also necessary as part  of the NVLAP to be
able to relocate, at a later time, all of the individual grid openings on which the reported results
are based.  The data recording system must therefore accommodate some  means of identifying
each particular grid and grid opening for which data are reported.  The simplest way to achieve
this is to use finder TEM grids, in which each grid opening can be uniquely identified by a letter
or number  incorporated into the design of the grid. Alternatively, if the grid design incorporates
a center marker which is asymmetric, grid openings can be relocated  using a grid map.  For
verified counting activities, the data recording system must permit the orientation of the grid as
it is placed in the TEM specimen holder to be recorded.
              The precise openings to be examined should be selected randomly, although care
should be taken to ensure that those selected appear representative of the overall specimen. This

-------
                                                                 QA Guidelines for TEM
                                                                             Revision:  0
                                                                Date:  December 27, 1989
                                                                 Section 4 Page 31 of 38
can be judged by an examination of the grid at low magnification.  If the sample appears to be
nonhomogeneous, a representative grouping of grid openings should be selected for examination.
It  is advantageous to avoid selection  of grid openings near the edge of the  grid,  since the
specimen holder profile may obstruct the EDXA detector's view of particles on these  openings.

4.10.4  Fiber Identification
              In a TEM examination, not all fibers can be unequivocally identified as asbestos
or nonasbestos.  QA interpretation is greatly assisted if the basis for each fiber identification is
recorded.  For example, during an initial examination of a specimen containing chrysotile, the
first operator may damage chrysotile fibers to the extent that the second operator, no matter how
experienced, cannot obtain a high proportion of electron
diffraction patterns.  In other types of samples, the specimen preparation may be too thick to
allow electron diffraction patterns from chrysotile to be discerned, or the crystal structure of the
chrysotile  may have been damaged by acid treatment  or heating before air sampling took place.
None of these situations reflect on the competence of the microscopist, but the failure to observe
a high proportion of fibers which exhibit satisfactory electron diffraction patterns often provides
a clue to other, and perhaps more serious QA problems associated with the analyses.
              The quality manual must have  a complete description of the techniques used by
the laboratory for identification of asbestos fibers on  the basis of their morphology, electron
diffraction patterns and EDXA spectra. This description should include or refer to data showing
that the techniques used by the laboratory reliably discriminate asbestos fibers from nonasbestos
fibers.  If, in  the system of fiber identification used by  the laboratory, ambiguities of fiber
identification exist (e.g.,  the inability to discriminate chrysotile from antigorite on the basis of the
EDXA spectrum alone), these ambiguities should be  described in the quality manual.

4.10.5  Measurement of the Dimensions of Asbestos Structures
              TEM analyses performed by the AHERA method do not require  the dimensions
of asbestos structures to be reported, other than a discrimination of fibers with lengths between
0.5 fim and 5  ^m, and  those longer than 5 /«m.  Other analytical methods  may require the
dimensions of each asbestos structure to be reported.  The data recording system must allow for

-------
                                                                QA Guidelines for TEM
                                                                           Revision: 0
                                                              Date:  December 27, 1989
                                                                Section 4  Page 32 of 38
these dimensions to be recorded for each fibrous structure detected.

4.10.6 Labeling of Micrographs and EDXA Spectra
             For the purposes of the AHERA TEM method, it is necessary to photographically
record some electron diffraction patterns, and also to record EDXA spectra. The quality manual
should contain a description of the system used to identify micrographs and spectra, as they relate
to specific fibrous structures in the fiber counting data.

4.10.7 Reporting of Grid Opening Dimensions
             The data record for each sample must contain the value of the  grid opening
dimensions to be used in the calculation of structure concentration.

4.10.8 Structure of Sample Data File
             The majority of laboratories use computers to manipulate the large amounts of
data generated by TEM analysis of air samples for asbestos.  The structure of the data file used
must be fully described in the quality manual.

4.10.9 Checking of Analytical Data
             Proof-reading of data is recommended to guard against major errors which may
be introduced by simple keyboard entry errors.  The methods in use by the laboratory to carry
out a second check on the accuracy of raw data should be described in the quality manual.

4.11 ELECTRON DIFFRACTION PATTERN MEASUREMENTS
             Measurements on electron diffraction patterns must be made very accurately
if a confident identification is  to be achieved.  The method by which these measurements are
made must be  included  in  the quality manual.   For those patterns which are recorded
photographically, the method by which the camera constant, d-spacing and angles between the
diffraction maxima are measured must be described, along with estimates of error  for each of
these measurements. For example, in the case of chrysotile, it is necessary to observe the spacing
of the layer lines and the separation between  the (002) reflections,  in addition  to qualitative

-------
                                                                QA Guidelines for TEM
                                                                            Revision:  0
                                                               Date:  December 27, 1989
                                                                Section 4 Page 33 of 38
aspects of the pattern such as the streaking of the (110) and (130) reflections.  Although
observation of a layered ED pattern with a layer line spacing of
0.53 nm does not constitute conclusive identification for amphibole, this observation can often
be made on amphibole asbestos fibers and contributes to the identification process
when supported by morphological and EDXA data.
             It is not economically feasible to  record every pattern  photograpically  and to
measure  each pattern from the photograph.  The manual must therefore contain a description
of the procedures used for visual identification of ED patterns. Visual observation of ED can
be achieved with least beam damage if small camera lengths are used and observations of the
pattern are made through the binoculars. The tilted screen introduces an elliptical distortion into
the image. However, with experience, an operator can recognize the essential features of  the ED
patterns semi-quantitatively even when the image is distorted in this way. There must be some
arrangement to allow estimation of significant measurements directly on the fluorescent screen
of the TEM.  For chrysotile, it is necessary to measure the separation of the 0.73 nm (002)
maxima and the separation of the 0.53 nm layer lines. With some care, these measurements can
be made with sufficient accuracy directly on the screen. For those patterns which are recorded
photographically, the method by which the camera constant, d-spacings  and angles between the
diffraction maxima are measured must  be described, along with estimates of error for each of
these measurements.

4.12 MEASUREMENTS  ON EDXA SPECTRA
             Identification  of chrysotile  can usually be achieved by electron  diffraction,
combined with observations of morphology.  However, if the morphology is not distinct, which
may be the case in some complex structures, the EDXA spectrum may be the only identification
aid available. The area of the magnesium peak relative  to the silicon peak allows discrimination
between talc and chrysotile. For the amphiboles, the relative areas of the relevant peaks  are the
only means of discriminating between the amphiboles,  and it is therefore important that these
peak  areas be measured accurately.  The method by  which statistically-valid peak areas are
obtained must be described in the quality manual. Where a fiber is identified on the basis of the
EDXA spectrum, the peak areas obtained must be quoted in  the data records, or a spectrum

-------
                                                                QA Guidelines for TEM
                                                                            Revision: 0
                                                               Date:  December 27, 1989
                                                                 Section 4  Page 34 of 38
must be stored such that the peak areas can be derived.
4.13  CALCULATIONS AND REPORTING OF RESULTS
             As discussed in Section  3.0,  the laboratory is responsible for estimating  the
concentration of asbestos structures on the surface of the filter. This estimate is calculated by
dividing the number of asbestos structures counted by an estimate of the area of filter examined.
The  number of significant digits reported  should be consistent  with  the precision of  the
measurement. Structure concentrations  per square millimeter based on a count of less than 50
asbestos structures should be reported to no  more than two significant digits, although an extra
digit should  be  carried to avoid rounding errors when  calculations are being made.  If no
asbestos  structures are  counted  in  the  area of  filter examined, the reported  estimated
concentration is zero structures per area  of filter. Although this estimate may be imprecise, zero
is the proper estimate of concentration. Substituting some other number for zero biases the
estimated concentration and distorts interpretation of sets of measurements.  To be useful to the
client, the estimated concentration should be accompanied by indicators of data  quality. These
include an estimate of the positive bias introduced by  contamination,  and an indication of
precision as discussed below.

4.13.1  Bias
              Laboratory contamination may add to the number of asbestos structures present
on the filter.  Section 4.6 described how  to minimize contamination.  Laboratory blanks provide
an upper bound for the magnitude of laboratory contamination. They cannot provide an actual
estimate of laboratory contamination because it is not usually possible to distinguish laboratory
contamination  from contamination  introduced  during  filter  manufacture.   Note  that  the
laboratory cannot account for initial contamination levels on filters that the client obtains from
a source other than the laboratory, nor can the  laboratory control contamination or particulate
loss  introduced during handling and transport from the field.  The  laboratory can only report
there are an estimated x structures per square millimeter on the filter and up to y structures per
square millimeter could have been introduced during the laboratory analysis.

-------
                                                                 QA Guidelines for TEM
                                                                             Revision: 0
                                                                Date:  December 27, 1989
                                                                  Section 4  Page 35 of 38
             The value of y is tied closely to the idea of detection limit.  The detection limit
from the laboratory's perspective is the smallest measurement that is unlikely (probability less
than a specified value) to be due entirely to contamination from the laboratory. This number
should be derived from laboratory blank  data. A typical approach is to use the 99th percentile,
i.e., the concentration  per surface area of filter below which 99 percent of measurements on
laboratory blanks fall.  (EPA has determined a detection  limit of 70 structures per  square
millimeter (s/mm2) for the AHERA TEM protocol for the purposes of applying the first step of
the AHERA clearance test.   This  level  is  based  on experience  with laboratory blank
measurements  and  is  thought  to reflect  contamination  during  manufacture.   Individual
laboratories may achieve detection limits  that are higher or lower than this value.) The method
of deriving the detection limit should be  stated clearly in the Quality Manual and in the report
to the client.

4.13.2  Precision
             Section 4.8 described measuring precision in terms of the coefficient of variation
as a function of the mean. The average coefficient of variation for repeated analyses of samples
of a particular concentration (where the repeated analysis consists of all preparation and analysis
steps) may be reported along with each  reported concentration.  Alternatively, the laboratory
might choose to provide the client with a graph showing the coefficient of variation as a function
of the mean.
             Confidence intervals are often used to indicate precision. The method used to
construct a 95 percent confidence interval, for example,  should result in an interval that contains
the true asbestos  structure concentration  95 percent of the time. A narrow confidence interval
indicates high precision. A wide confidence interval indicates low precision.
             EPA is developing a method for constructing confidence intervals for TEM results
using laboratory-specific coefficients  of  variation.   This research effort  is needed because
commonly used confidence intervals based on the normal distribution are inappropriate for TEM
data especially when only a small number  of structures are  counted.  Until a satisfactory
alternative is developed, it is suggested that confidence intervals, if reported, be based  on  the
Poisson distribution.  The report  should clearly state that the Poisson distribution is an optimistic

-------
                                                                 QA Guidelines for TEM
                                                                             Revision:  0
                                                                Date:  December 27, 1989
                                                                  Section 4  Page 36 of 38
assumption that ignores most sources of variability (including laboratory performance) and that
the true confidence interval is likely to be larger (possibly much larger).  For many clients, the
laboratory's graph of  the  coefficient of variation  as  a function  of  the mean  will be more
informative than a confidence interval for each concentration.
              As stated above, samples on which  no asbestos  structures  have  been counted
should be reported as 0 s/mm2.  This does not necessarily imply,  of course, that there are no
asbestos structures on the filter, any more than an estimate of 105 s/mm2, implies that there are
exactly 105 s/mm2 present.  The information provided on precision should make this clear to the
client. Although the coefficient of variation is not  defined at zero, the graph of coefficient of
variation plotted against the mean will indicate the tendency for the coefficient  of variation to
increase as the mean decreases.  If confidence  intervals are used, the confidence interval should
be provided for observations of zero just as it  is provided for any other estimate. All measures
of precision should be clearly  described in the Quality Manual and in the report to the client.
              The  coefficient of  variation  or  a confidence  interval  is appropriate for
characterizing individual estimates of concentration,  and is particularly useful for clients planning
air monitoring programs for purposes other than AHERA clearance. However, clients who are
applying the AHERA TEM clearance test to their results do not need to use this  information as
part of the clearance test calculation. The AHERA TEM clearance test is based on an aggregate
of at least five individual estimates and the test already takes into account the variability of the
individual data values.

4.13.3  Other Calculations
              Although the laboratory's control extends only to estimating the concentration of
asbestos structures on the surface of the filter,  most clients are interested in estimating airborne
asbestos concentrations. Consequently, many laboratories routinely make additional calculations
using data supplied by the client.   The estimated airborne asbestos  concentration (ignoring
contribution from contamination) is given by:

                                   C = (n/a) x (A/V)

-------
                                                                 QA Guidelines for TEM
                                                                             Revision:  0
                                                                Date:  December 27,1989
                                                                 Section 4 Page 37 of 38
where:
              C is the estimated  airborne  asbestos concentration in structures  per cubic
              centimeter,
              n is the number of asbestos structures counted,
              a is the estimated area of filter surface examined in square millimeters,
              A is the collection area of the filter in square millimeters, and
              V is the volume of air sampled in cubic centimeters.
Estimates of n and a are provided by the laboratory. Estimates of A and V are provided by the
client.
              The analytical sensitivity (S), the airborne asbestos concentration corresponding
to the counting of a single asbestos structure, is given by:

                                   S  = (I/a) x (A/V).

The analytical sensitivity  can be regarded as a conversion factor between structure  count and
estimated airborne asbestos concentration.   It indicates the coarseness of the measurement
method.  Note that the analytical sensitivity is determined completely by the values of a, A, and
V which are selected by the laboratory (a) or supplied by the client (A and V). The analytical
sensitivity does not depend on the concentration of asbestos structures. The AHERA TEM
protocol requires an analytical sensitivity of 0.005 s/cm3 or smaller.
              The detection limit and confidence intervals expressed in terms of s/mm2 may
be converted to s/cm3 by multiplying by  (A/V). The detection limit is harder to interpret on this
scale  because  contamination is not associated with a volume of air.  If the detection limit is
expressed as  0.01 s/cm3  , for  example,  one  could  say "The estimated airborne  asbestos
concentration is 0.06 s/cm3 , however, up to 0.01 s/cm3 of this could be due to contamination."
Coefficients of variation are unaffected by the change of scale. (The AHERA clearance protocol
does  not address  the question of correcting for contamination because the same  correction
applied to two sets of samples will not  affect a  comparison between the sets.  Correction for

-------
                                                                QA Guidelines for TEM
                                                                            Revision:  0
                                                               Date:  December 27, 1989
                                                                Section 4 Page 38 of 38
contamination only becomes an issue when the absolute airborne asbestos concentration is of
concern.)
             Strictly speaking, the laboratory analyzes individual samples without knowledge of
their origin or  application and reports the results  individually.  Some laboratories  provide
additional  consulting services such  as applying  the  AHERA clearance test to  determine
completion of an asbestos abatement project.  These activities are beyond the scope of this
document.

-------
                                                               QA Guidelines for TEM
                                                                           Revision:  0
                                                              Date: December 27, 1989
                                                                  Section 5  Page 1 of 1
5.0    REFERENCES
Andrews KW, Dyson DJ, Keown SR. (1967):  Interpretation of Electron Diffraction Patterns.
Hilger & Watts Ltd., London.

Hirsch PB, Howie A, Nicholson RB, Pashley DW, Whelan MJ. (1965): Electron Microscopy
of Thin  Crystals. Butterworths, London.

Javitz HS, Fowlwer DP.  (1981): Statistical analysis of microscopic counting data. In Electron
Microscopy and X-Ray Applications, Russell PA, ed. Ann Arbor Science.

Lee RJ. (1978): Basic Concepts of Electron Diffraction and Asbestos Identification Using
SAD. SEM/1978/I (Ed. O. Johari) AMF O'Hare,  Chicago, Illinois, 677-694.

National Institute of Standards and Technology. (1989):  National Voluntary Laboratory
Accreditation Program, Airborne Asbestos Handbook:  Operational and Technical
Requirements of the Laboratory Accreditation Program for Transmission Electron
Microscopy (Draft, March 1989).

Ogden TL, Shenton-Taylor T, Sherrie JW, Crawford NP, Moorcroft S, Duggan MJ, Jackson
PA, Treble RD. (1986):  Within-laboratory quality control of asbestos counting. Annals of
Occupational Hygiene, 30,  411-425.

Small JA, Steel EB, Sheridan PJ. (1985): Analytical Standards for the Analysis of Chrysotile
Asbestos in Ambient Environments.  Analytical Chemistry, 57, 204-208.

Steel EB, Small JA. (1985):  Accuracy of Transmission Electron Microscopy for the Analysis
of Asbestos in Ambient Environments.  Analytical Chemistry, 57, 209-213.

Williams, DB. (1986):  Standardized Definitions of X-Ray Analysis Performance Criteria in
the AEM. Microbeam Analysis - 1986, (AD Romig Jr. and WF Chambers, Eds),  San
Francisco Press, Inc., Box 6800, San Francisco, CA 94101-6800, 443-448.

Williams, DB. (1984):  Practical analytical electron microscopy in materials science.  Philips
Electronics Instruments,  Inc., Mahwah, New Jersey, 34-35.

-------
                                                         QA Guidelines for TEM
                                                                    Revision:  0
                                                       Date:  December 27, 1989
                                                        Appendix A  Page 1 of 1


APPENDIX A: SOURCES OF INFORMATION ON QUALITY ASSURANCE

Garfield, Frederick  M.,  Quality  Assurance Principles for  Analytical Laboratories,
Association of Official Analytical Chemists, Arlington, VA, 1984.

Taylor, John Keenan, Quality Assurance of Chemical Measurements, Lewis Publishers, Inc.,
Chelsea, MI, 1987.

United States Environmental Protection Agency, OTS Guidance  Document for the
Preparation  of Quality Assurance Project Plans (Final, October 1, 1987).

Youden, W.J. and E.H. Steiner, Statistical Manual of the Association of Official Analytical
Chemists, Association of Official Analytical Chemists, Arlington, VA, 1975.

-------
                                                              QA Guidelines for TEM
                                                                          Revision: 0
                                                             Date: December 27, 1989
                                                              Appendix B Page  1 of 4

APPENDIX B:      EXAMPLES  OF  QUALITY ASSURANCE  PROBLEMS  AND
                   CORRECTIVE ACTIONS
             Asbestos analysis by TEM is very complex, and it is therefore not practical to write
a complete fault-tracing guide.  The following are examples of situations which may arise, and
possible solutions are discussed.  The relevance of some of the quality assurance practices will
be clarified by these examples.
EXAMPLE Bl:     OBSERVATION OF ASBESTOS STRUCTURES IN A LABORATORY
                   BLANK
             During a series of routine AHERA analyses of MCE filters, a laboratory blank is
found to be contaminated by asbestos structures containing chrysotile.  Two of the five samples
identified as originating from inside the work-site have approximately the same level of chrysotile
structures as the laboratory blank, and the other three samples have no asbestos structures. The
average value of the five samples exceeds 70 s/mm2.
             Before any investigation is made into  the possible source of contamination, the
data should be reviewed to establish that the asbestos structures have been properly identified
by an experienced operator.  If satisfactory blank analyses had been consistently obtained before
preparation of this particular group of samples, any change in the procedures or reagents which
may have occurred immediately prior to preparation of this group of samples should be examined
critically.
             If no changes in the procedure had been made prior to the preparation of these
samples, the reliability of the laboratory blank should be investigated.  Was the filter from the
same lot number as the other five samples?  How well-characterized is the average blank value
for filters from this lot number?  If no data are available, it  cannot be determined whether the
contamination was already present on the unused filters or whether the contamination could have
arisen  during the specimen preparation.  This observation emphasizes the importance  of
well-characterized filter material for use as laboratory blanks. The laboratory should maintain
a stock of well-characterized filter material, and should always incorporate a laboratory blank.
             The next question concerning these results is whether the samples were at any
point processed individually through the collapsing procedure, the plasma etching,  or the carbon
evaporation. If they were individually prepared on separate slides, it is possible that three of the

-------
                                                                QA Guidelines for TEM
                                                                            Revision:  0
                                                               Date: December 27,1989
                                                                Appendix B Page 2 of 4
microscope slides were contaminated before use, or that contamination was transferred to two
of the samples and also to the blank sample by one of these procedures. For this reason, it is
good practice to prepare groups of samples on the same slide, along with a laboratory blank on
the same slide.  By preparing samples in this way, all of the samples experience exactly the  same
conditions, and the laboratory blank is a control for the whole group of samples. Assuming that
all samples were prepared on the same  slide, and all processed simultaneously through the
plasma etching and carbon evaporation  steps, it can  be  stated that the contamination  most
probably arose from another source.
             The order in which the sample filters were handled should be inspected.  Was
the order such that the laboratory blank was prepared after handling the two positive samples?
If this was the case, is it possible that contamination could have been transferred from the two
samples to the laboratory blank by tools such as scalpels, tweezers or punches?  This would be
unlikely if the two high samples were the  first two handled, followed by  three low samples and
then by a contaminated laboratory  blank.  If contamination from the  handling  tools  can  be
rejected  on this basis, the filter dissolution should be investigated.   Was the  Jaffe  washer
thoroughly washed and filled with fresh solvent prior to its use for this set of samples? In order
to eliminate this question entirely, it is good practice to use a freshly washed Jaffe washer and
new solvent for each batch of grids. Assuming that this was done, the precise method for cutting
of collapsed and coated  filter portions should be examined.  Were the tweezers and scalpel
cleaned between handling of each of the samples? What was the order in which the collapsed
filter portions were cut and placed in the  Jaffe washer?
             If no answer is found, there may be sporadic contamination from the reagents or
other materials  used.  All reagents and  other  materials which have contacted the filters or
specimens should be checked for contamination. If no source of contamination specific to this
particular group of samples is identified,  it would be necessary to review all of the data from
samples which have been prepared since the last satisfactory laboratory blank was prepared.  If
a source of contamination is identified,  then procedures must be modified to eliminate this
source, and a series of laboratory blanks prepared to demonstrate that satisfactory performance
can be  obtained.   Preparation  of  actual sample filters should  not be recommenced  until
satisfactory laboratory blanks can be consistently obtained.

-------
                                                                QA Guidelines for TEM
                                                                            Revision:  0
                                                               Date: December 27, 1989
                                                                Appendix B Page 3 of 4
             When incidents such as that described above occur, it is important to document
all of the measurements and actions in the quality assurance records. It is sometimes found that
no  source  of contamination can be confidently identified, and after  demonstrating that
satisfactory blanks  can be  obtained,  analysis of actual samples is recommenced.   If  a
contamination incident reoccurs, the quality assurance records may provide valuable clues as to
the source,  particularly if a source was not identified on a previous occasion.

EXAMPLE B2:      THE ASBESTOS  STRUCTURE  COUNTS BY TWO  OPERATORS
                    DISAGREE
             As part of the routine quality assurance recounting, it may be found that two
operators in the laboratory obtain significantly different asbestos structure counts for a particular
specimen.  If the two operators were instructed to  count asbestos structures on the same set of
grid openings, and the discrepancy is a significant one, there may be several reasons:

       (a)    The grids or grid openings examined by the two operators were not actually the
             same, due to mislabeling or simple selection of the incorrect openings;
       (b)    The operators have not used identical fiber identification criteria, resulting in some
             structures being classified  incorrectly;
       (c)    The operators have not counted structures in  the same way.  One operator may
             have  perceived structures  differently, resulting in a different numerical structure
             count;
       (d)    If the asbestos structures were chrysotile, the first operator may have destroyed the
             crystallinity of the fibers so that they could not be identified by electron diffraction
             by the second operator; and
       (e)    The two operators overlooked different proportions of asbestos structures during
             the scanning of the  specimens.

             A major disagreement of this type  is very  serious, because the structures on a
particular grid opening are the same for both operators, and ideally the structure counts obtained
by the two operators should be identical. Before any action is taken, it is important to establish
that a disagreement really exists, and that the problem is not simply one of incorrect selection

-------
                                                                 QA Guidelines for TEM
                                                                             Revision:  0
                                                                Date: December 27, 1989
                                                                 Appendix B Page 4 of 4
or labeling of the grid openings to be examined.  This can be established by a  review of the
specimens, conducted simultaneously by both operators. If an incorrect grid opening selection
has been made by one of the operators, the sample should be reexamined so that counts are
made on identical sets of grid openings.  In addition, the reasons  for the incorrect selection
should be investigated and appropriate corrective actions taken to minimize the possibility of a
recurrence.
              If it is certain that the grid openings examined by the two operators were identical,
the data should be examined carefully to determine whether there are any consistent differences.
If the operator reporting  the higher  result has  reported more short fibers for  which the
identifications are indisputable, it may indicate that the other operator is not scanning correctly
and requires additional training. It could also be that one operator is applying the fiber counting
criteria differently.  Whatever the reason may be, two operators presented with the same image
on the TEM, and with access to the same identification procedures,  are reporting different
results. Under these conditions, the analysis must be considered out of control and no further
analysis can be conducted by these two operators until the problem has been resolved using the
verified counting technique.   The  problem is considered  to  have  been  resolved when both
operators have demonstrated competence within the criteria (greater than or equal to 80% true
positive counts, less than 10% false positive  and less than 20% false negative counts) before
actual sample counting is allowed to proceed.

-------
REPORT DOCUMENTATION \ »• REPORT NO. I 2.
PAGE i 560/S-90-002 	 . J_ . 	 ... . .
I. Title and Subtitle
ualitv Ascarranrf Ruirtolin^e! for Trancarrj ssion KT^< *l mn
icroscopy Asbestos Laboratories
'. AuthoKs)
prnw/T.r Chsrtfi0!^, E • (Tr3S'!Son J • Rench J
i. Performinc Organization Name and Address
•attelle
rlington Office
102 Wilson Boulevard
rlington, VA 22201
2. Sponsoring Organization Name and Address
3 Environmental Protection Agency
:f ice of Toxic Substances/Exposure Evaluation Division
31 M Street, SW, Washington, DC 20460
S. Supplementary Notes
3. Recipient's Accc.s\ton No. i
i
5. Report Date
J3ecember_1989
6.
8. Performing Organization Rept. No.
10. Project/Task/Work Unit No.
11. Contract(C) or Grant(G) No.
cc> 68-02-4294
(G)
13. Type of Report & Period Covered
Peer-reviewed report
14.

:. Abstract (Limit: 200 words)
Suidance  is provided for refining quality assurance programs for transmission electron
licrosoopy asbestos  laboratories.   Emphasis is provided for  interpreting quality
assurance data and controlling  laboratory contamination problems.
. Document Analysis  a. Descriptors
AHERA, quality assurance, asbestos, transmission electron microscopy,  TEM, NIST,
National Voluntary Laboratory Accredition Program  (NVLAP)
 b. Identifiers/Open-Ended Terms
 c. COSATI Field/Group
Availability Statement
19. Security Class (This Report)
TTnc'Irissified
20. Security Class (This Pace)
Unclassified
21. No. of Paces
62
22. Price .
 ANSI-Z39.18)
                                       See Instructions on Reverse
OPTIONAL FORM 272 (4-7T)
(Formerly NTIS-35)
Department of Commerce

-------